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
Breast Cancer ResBreast Cancer Research1465-54111465-542XBioMed Central London bcr11981598746110.1186/bcr1198Research ArticleEffect of reproductive factors on stage, grade and hormone receptor status in early-onset breast cancer Largent Joan A [email protected] Argyrios [email protected] Hoda [email protected] Epidemiology Division, Department of Medicine, University of California, Irvine, Irvine, California, USA2005 16 5 2005 7 4 R541 R554 22 1 2005 25 2 2005 5 4 2005 21 4 2005 Copyright © 2005 Largent 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
Women younger than 35 years who are diagnosed with breast cancer tend to have more advanced stage tumors and poorer prognoses than do older women. Pregnancy is associated with elevated exposure to estrogen, which may influence the progression of breast cancer in young women. The objective of the present study was to examine the relationship between reproductive events and tumor stage, grade, estrogen receptor and progesterone receptor status, and survival in women diagnosed with early-onset breast cancer.
Methods
In a population-based, case–case study of 254 women diagnosed with invasive breast cancer at age under 35 years, odds ratios (ORs) and 95% confidence intervals (CIs) were estimated using unconditional logistic regression with tumor characteristics as dependent variables and adjusting for age and education. Survival analyses also examined the relationship between reproductive events and overall survival.
Results
Compared with nulliparous women, women with three or more childbirths were more likely to be diagnosed with nonlocalized tumors (OR = 3.1, 95% CI = 1.3–7.7), and early age (<20 years) at first full-term pregnancy was also associated with a diagnosis of breast cancer that was nonlocalized (OR = 3.0, 95% CI = 1.2–7.4) and of higher grade (OR = 3.2, 95% CI 1.0–9.9). The hazard ratio for death among women with two or more full-term pregnancies, as compared with those with one full-term pregnancy or none, was 2.1 (95% CI = 1.0–4.5), adjusting for stage. Among parous women, those who lactated were at decreased risk for both estrogen receptor and progesterone receptor negative tumors (OR = 0.2, 95% CI = 0.1–0.5, and OR = 0.4, 95% CI = 0.2–0.8, respectively).
Conclusion
The results of the present study suggest that pregnancy and lactation may influence tumor presentation and survival in women with early-onset breast cancer.
==== Body
Introduction
Invasive breast cancer among women younger than 35 years is relatively rare in comparison with breast cancer in older women; only 2% of diagnoses occur in the younger age group [1]. However, it is the leading type of cancer and cause of cancer-related death in women aged 15–34 years [2]. Prognosis following diagnosis for younger women is worse; they have shorter recurrence-free and overall survival than do older premenopausal [3] and post-menopausal [4] women.
Malignant breast tumors in women younger than 35 years tend to have features that are characteristic of more aggressive or advanced tumors compared with tumors in older women. The tumors in younger women tend to be larger, have more nodal involvement, are more poorly differentiated, and are more often estrogen receptor (ER) negative than are tumors in older women [3,5]. These tumor variables have been associated with poorer overall survival. Although there is evidence linking tumor factors with recurrence-free and overall survival, little is known about host characteristics that may contribute to differences in tumor variables observed at diagnosis, particularly in young women, who as a group are not subject to the same screening practices (namely mammography) as their older counterparts. An understanding of host variables that influence stage at presentation and other tumor characteristics may help to elucidate the mechanisms that are involved in tumor progression in early-onset breast cancer.
Pregnancy and childbirth are generally believed to reduce a woman's long-term risk for breast cancer. However, some studies have documented a transient increase in risk in the years immediately following childbirth [6-8], and a decrease in survival for women diagnosed in the years following childbirth [9,10].
Early detection is a key factor in determining whether a woman will survive a breast cancer diagnosis. Advances in screening methods have lead to a significant decrease in the percentage of diagnoses of late-stage disease in the past decade [11]. However, women younger than 40 years are generally not included in recommendations regarding who should undergo screening mammograms in the USA.
The observation that women with breast cancer under age 35 years are more likely to be diagnosed with advanced stage disease [3,5] and to have poorer survival than older women, combined with the fact that these women are not included in mammogram screening recommendations, underscores the importance of uncovering the processes of disease progression in these women so that effective, targeted strategies for early detection and management of this disease may be developed. The present study examines the relationships between reproductive and hormonal factors, stage, grade, hormone receptor status, and survival among women diagnosed with breast cancer at age under 35 years.
Materials and methods
Study design
A population-based, case–case design was used to examine the relationships between host reproductive and other variables, and tumor characteristics at presentation in breast cancer patients diagnosed at an early age.
Population under study
The study included 298 patients with early-onset (diagnosed at age <35 years) breast cancer. Eligible participants were all women aged between 18 and 34 years residing in one of 38 of the 58 California counties (n = 169), in Connecticut (n = 60), or in Massachusetts (n = 69) at the time of diagnosis. The participation rate was 65% for patients from California, 61% for patients from Connecticut, and 55% for patients from Massachusetts.
The women were diagnosed with invasive breast cancer between 1 January 1995 and 31 December 1996. The patients had to be living at the time of contact. Women who met these criteria were identified through the state or regional cancer registries of California, Connecticut, and Massachusetts. Twenty-eight women in California, one woman in Massachusetts, and no women in Connecticut died before study contact. Limited data on nonparticipants from California allowed for a comparison of demographic variables between participants and nonparticipants. Although there were no significant age differences between participants (mean 31.3 years) and nonparticipants (mean 30.9 years), participants (73%) were more likely to be non-Hispanic white than were nonparticipants (59%).
Protocol
Before contact with any potential participants, approval was obtained from the local institutional review boards at participating centers. Consent to participate was first obtained from the treating physician. Potential participants were then mailed introductory letters and response forms. Trained interviewers telephoned participants to conduct a structured telephone interview and to schedule an appointment to collect a blood sample. Participants completed a written informed consent and a release of medical information form. The mean time between diagnosis and interview was 11.5 months (range 0–27 months).
Data collection and measurement tools
The interviews obtained information on host characteristics including age, demographic information, reproductive variables (age at menarche, pregnancy and childbirth history, history of spontaneous or induced abortions, lactation history), self-reported family history of cancer, and oral contraceptive and other hormone medication use.
Data on tumor characteristics were available for 254 (85.2%) participants from the cancer registry databases and pathology reports, and these women comprised the study group. Information on tumor characteristics was collected at the time of diagnosis, including tumor site, histology, laterality, behavior, grade, stage, tumor size, and ER and progesterone receptor (PgR) status.
Cancer registry and SEER linkage
The California Cancer Registry abstracts for all of the California patients were available and were the primary source of tumor characteristic and follow-up data for these patients.
The 60 Connecticut patients were linked with the National Cancer Institute's Surveillance, Epidemiology, and End Results (SEER) Program's 1973–1999 public use data files [12] to obtain tumor characteristics and vital status updates. Because personal identifiers were used only by the centers that conducted the telephone interviews and were not available in the analysis data set for these patients, linkage was performed using the following variables, if these tumor characteristics were also available from pathology reports: year of birth, age at diagnosis, month and year of diagnosis, marital status, race, and tumor characteristics (site, histology, laterality, behavior, grade, stage, ER/PgR status, and size).
There were 60 patients who participated in the Early Onset Breast Cancer Study and who were Connecticut residents at the time of diagnosis. All of the 60 participants were linked to the SEER records using the variables listed above.
Pathology reports
Tumor data for the Massachusetts patients were ascertained using pathology reports. Pathology reports were available for 25 of the patients from Massachusetts and were reviewed by a certified tumor registrar and coded for tumor site, histology, laterality, behavior, grade, size, and SEER summary stage. Where available, pathology reports were also used to supplement tumor data for the patients from California and Connecticut.
Stage determination
The SEER summary stage classifies tumors into one of the following categories [13]: in situ, localized, regional by direct extension, regional by nodes, regional by nodes and direct extension, distant, or unstaged based on the SEER guidelines. Another staging system reported for the patients from California was the TNM staging system, which follows the American Joint Commission on Cancer (AJCC) criteria and consists of assigning appropriate letters or numbers to the following three fields [14]: T (primary tumor), N (nodal involvement), and M (distant metastasis). This classification is recorded for both clinical and pathological staging.
These sources of tumor staging information resulted in a total sample of 247 with stage classification as either localized (invasive carcinoma confined to the breast) or nonlocalized (invasive carcinoma spread beyond the breast, by direct extension and/or to regional lymph nodes, and/or by direct extension beyond adjacent organs specified as regional, metastasis to distant lymph nodes, or development of discontinuous secondary or metastatic tumors).
Grade
Tumors were classified according to histologic grade or degree of differentiation, as coded in the registry databases or pathology reports. This measures the degree to which the tumor cells have the differentiated or specialized characteristics of the tissue or organ in which they are found. Tumor grade information was available for 213 participants.
Estrogen receptor/progesterone receptor status
Tumors classified as borderline for hormone receptor status were considered positive (n = 3 for ER status; n = 3 for PgR status). A total of 182 and 176 records contained information on ER and PgR status, respectively.
Reproductive/hormonal variables
Reproductive variables and hormonal exposures were assessed using the personal interview data available for all 298 study participants.
Family history status
Family history of cancer was assessed during the personal interview. A positive family history was considered to be at least one first-degree female relative (mother or sister) diagnosed with breast or ovarian cancer. Adopted participants were excluded from the family history analyses (n = 2), leaving 296 with complete family history information.
Statistical analysis
Tumor presentation
Analyses were completed to assess whether host characteristics (including reproductive/hormonal and family history variables) were associated with tumor characteristics indicative of advanced or aggressive disease (including tumor stage, grade, and ER/PgR status), adjusting for age at diagnosis and education (higher of self or spouse). Odds ratios (ORs) were calculated using unconditional logistic regression with tumor characteristics as the dependent variables (localized versus nonlocalized disease; grade 1–2 tumors versus grade 3–4 tumors; ER positive versus ER negative; PgR positive versus PgR negative). Trend tests for select variables were performed by regressing categorical variables as continuous. All analyses were performed using SAS version 8.2 (SAS Institute Inc., Cary, NC, USA).
Survival analysis
Data on follow up were obtained for the California cases through 31 December 2001 from the California Cancer Registry and for the Connecticut cases through 31 December 1999 from the SEER 1999 public use data file [12]. Follow-up data were not available for the patients from Massachusetts. Survival time was truncated at 60 months for the 40 California cases with follow-up information beyond 60 months. A woman's survival time was censored if the woman was known to be alive at the date of last follow up. Kaplan–Meier product limit curves were generated to examine the associations between demographic and reproductive characteristics and overall survival following breast cancer diagnosis. The Wilcoxon test was used to test the equality of the overall survivor functions across groups. The effects of reproductive factors on survival were also quantified with relative risk estimates (and 95% confidence intervals [CIs]) calculated using Cox proportional hazards regression analysis, adjusting for breast cancer stage.
Results
Demographic and tumor characteristics of the 254 women studied are summarized in Table 1. Seventy-six per cent of tumors were classified as infiltrating duct carcinoma. Stage was localized in 51.6%, nonlocalized in 45.7%, and unknown in 2.8% of cases. The majority of tumors were self-detected (80.7%). Age at diagnosis ranged from 20 to 34 years, with 75.6% of cases occurring between ages 30 and 34 years. The majority of patients (72.8%) were non-Hispanic Caucasians.
Reproductive and hormonal factors and tumor characteristics
Age- and education-adjusted analyses of reproductive events occurring before diagnosis and tumor characteristics are summarized in Tables 2, 3, 4, 5. Reproductive variables include age at menarche, age at first full-term pregnancy, total number of live births, total number of miscarriages and induced abortions, breastfeeding history, months between last full-term pregnancy and diagnosis, and oral contraceptive use. Tumor characteristics include stage at diagnosis, grade, ER, and PgR status.
Analyses of breast cancer stage included 247 women for whom tumor stage data were available (Table 2). Women with a first full-term pregnancy at age under 20 years were three times as likely to be diagnosed with nonlocalized disease than were women with no full-term pregnancies (OR = 3.0, 95% CI 1.2–7.4). This OR decreased as age at first full-term pregnancy increased (P for trend < 0.01). However, when this analysis was restricted to parous women there were no significant associations between age at first full-term pregnancy and stage (for age 20–24 years: OR = 0.7, 95% CI = 0.3–1.7; for age 25–29 years: OR = 0.5, 95% CI = 0.2–1.4; for age 30–34 years: OR = 0.3, 95% CI = 0.1–1.1; versus age <20 years). Women who had three or more live births were about three times more likely to be diagnosed with nonlocalized disease than were women who did not have any live births. When this analysis was restricted to only gravid women, those with three or more pregnancies were more likely to be diagnosed with more advanced stage breast cancer than were women who reported one to two pregnancies (OR = 2.3, 95% CI = 1.2–4.3). Number of births and age at first full-term pregnancy were highly correlated, and when terms for both were included in a logistic regression model neither was found to remain statistically significant (data not shown). Age at menarche, total number of miscarriages, total number of induced abortions, time since last full-term pregnancy, breastfeeding history, and oral contraceptive use were not found to be significantly associated with stage.
Analyses of tumor grade included the 213 women for whom data on tumor grade were available. Early age at first full-term pregnancy was the only reproductive variable examined that was found to be significantly associated with higher tumor grade. Women who had a first full-term pregnancy at age under 20 years were significantly more likely to have tumors of grade 3–4 than were nulliparous women. The other reproductive and family history variables examined were not found to be significantly associated with tumor grade (Table 3).
The analyses of ER and PgR status included the 182 and 176 women, respectively, for whom data on ER and PgR status were available (Tables 4 and 5). Breastfeeding history was the only reproductive variable found to be significantly associated with both ER and PgR status. Among parous women, those who reported ever breastfeeding were significantly more likely to have ER-positive tumors and PgR-positive tumors. Other reproductive variables examined were not found to be significantly associated with either ER or PgR status.
Family history and tumor characteristics
A family history of breast or ovarian cancer in a mother or sister was examined in relation to tumor characteristics. Having a mother or sister with a diagnosis of breast or ovarian cancer was not significantly associated with tumor stage, grade, ER or PgR status.
Survival analysis
Follow-up data on vital status and date of last follow up were available for 221 cases. Median 5-year survival time was 45 months (range 2–59 months) for the cases from Connecticut and 54 months (range 3–60 months) for the cases from California, and 52 months overall. Six of the 60 (10.0%) cases from Connecticut and 24 of the 161 cases from California (14.9%) died during the follow-up period.
Reproductive variables examined in relation to survival were stratified by stage and included number of full-term pregnancies (0–1, ≥ 2), age at first full-term pregnancy (<25 years, 25–34 years, no full-term pregnancies), time since last pregnancy (gravid women: 0–23 months, ≥ 24 months), and lactation history (parous women: never, ever). When stratified by stage, women with either localized or nonlocalized disease who had two or more full-term pregnancies did not have significantly decreased survival as compared with women with no or one full-term pregnancies (P = 0.19 and P = 0.15, respectively; Fig. 1). However, in an unstratified analysis women with two or more full-term pregnancies had significantly decreased survival as compared with women with no or one full-term pregnancies (P < 0.05). Age at first full-term pregnancy and lactation history were not found to be significantly associated with overall survival (P = 0.08 and P = 0.45, respectively). However, gravid women with nonlocalized tumors who had their last pregnancy within 2 years of diagnosis had borderline decreased survival as compared with gravid women with nonlocalized tumors who had their last pregnancy more than 2 years before diagnosis (P = 0.06), but this association was not significant among women with localized tumors (P = 0.24; Fig. 2).
Cox proportional hazards regression was used to test the association of reproductive variables (total full-term pregnancies, age at first full-term pregnancy, and time since last pregnancy) with survival, adjusting for stage. In these separate analyses, increased parity ≥ 2 full-term pregnancies as compared with 0–1 full-term pregnancies) was a significant predictor of decreased survival (hazard ratio [HR] = 2.1, 95% CI = 1.0–4.5). Age at first full-term pregnancy and time since last pregnancy were not found to be significantly associated with risk for decreased survival (HR = 0.8, 95% CI = 0.3–1.9, and HR = 0.5, 95% CI = 0.2–1.1, respectively).
Discussion
Steroid hormones play an important role in the etiology of breast cancer. Evidence has consistently linked endogenous hormone exposure to breast cancer risk [15]. Estrogen has been implicated to induce tumor promotion and/or progression by enhancing cell proliferation [16]. In the present study it was hypothesized that events associated with increases in estrogen levels in young women may influence the progression of breast tumors, resulting in a more rapidly developing disease that presents at a later stage and ultimately has a poorer prognosis. Breast cancer patients younger than 35 years were selected for the study because tumors in this group tend to have the most aggressive profiles.
Factors associated with tumor characteristics
Parity
Assessment of reproductive histories of early-onset breast cancer patients indicated that increased parity was associated with diagnosis of nonlocalized disease. However, this association did not extend to pregnancies ending in either spontaneous or induced abortion. This finding is consistent with a recent study of breast cancer patients aged under 45 years [17] that reported an association between increased parity and overall mortality.
Recent pregnancy
Having a full-term pregnancy within the 2 years preceding diagnosis was not found to be associated with later stage disease, higher grade, or ER or PgR status. These findings differ from a recent study of breast cancer patients aged under 45 years [17], which found a recent full-term pregnancy (<2 years before diagnosis) to be associated with disease that had spread to the regional lymph nodes. It should be noted that the present study included a younger population, which may represent a group more likely to have a delay in diagnosis as compared with older premenopausal women.
Age at first full-term pregnancy
Another finding of the present study was that early age at first full-term pregnancy was also associated with later stage disease compared with nulliparity. As the age at first full-term pregnancy increased, the risk for having a diagnosis of nonlocalized disease decreased, although risk estimates for any age at first full-term pregnancy were elevated as compared with nulliparity. However, when this analysis was restricted to include only parous women, no significant association between age at first full-term pregnancy and stage was apparent.
There is some previous evidence [18,19] that an early age at first full-term pregnancy may be associated with poorer prognosis. It should also be noted that an early age at first full-term pregnancy has been demonstrated to be protective in terms of lifetime risk for breast cancer [20], and this protective effect may be explained by the process of differentiation of breast tissue induced by a full-term pregnancy, resulting in lower susceptibility to carcinogenic influences. However, the present study addresses the issue of tumor progression rather than the risk for developing breast cancer, because only breast cancer cases were included in these analyses. Therefore, events that are believed to be protective in terms of breast cancer risk, such as early age at first full-term pregnancy and increased parity, may have an adverse effect on breast cancer progression for women who are diagnosed with the disease. Perhaps the timing of reproductive events in relation to the initiation and promotion of breast cancers is critical in determining whether the effect may be harmful for women who develop breast cancer during their principal reproductive years. Several findings of the present study support such a theory.
Lactation
Lactation is believed to be associated with decreased circulating levels of estrogen and progesterone, and was hypothesized to be associated with earlier stage tumors, with less aggressive profiles. Lactation among parous women was not found to be associated with tumor stage. However, parous women who reported that they had ever breastfed a child were more likely to have ER-positive and PgR-positive tumors than were parous women who had never breastfed a child. Lactation, therefore, was associated with tumor markers indicative of better prognosis. Another case–control study of women aged 20–74 years [21] found that women who had ever breastfed were at decreased risk for developing ER-positive/PgR-negative (OR = 0.5, 95% CI = 0.3–0.8) and ER-negative/PgR-positive tumors (OR = 0.4, 05% CI = 0.2–0.8). In a case-only analysis, lactation was not associated with ER-positive/PgR-positive or ER-negative/PgR-negative status.
Perhaps the difference in results between the present study and the previous one [21] is in part attributable to the different subpopulations of breast cancer patients. Both ER and PgR status were demonstrated to vary with age, with older women more likely to be diagnosed with tumors positive for these markers [21]. It is possible that, again, the timing of events may be important in determining the effect of lactation on hormone receptor status of breast cancers. In other words, the stage of progression of the breast tumor at the time of lactation may influence whether there is an effect on expression of hormone receptors, and this may differ in general by age of the patient.
Exogenous hormones
Exogenous hormone exposure, in the form of oral contraceptives, was also examined in relation to tumor characteristics. This source of exogenous hormones was not associated with tumor stage, grade, or ER or PgR status. Oral contraceptive use was further examined by months of use, but no significant associations between months of use and any of the tumor characteristics emerged. Therefore, the present study does not support an association between exogenous hormone use and tumor presentation among early-onset breast cancer patients.
Survival analysis
Survival analyses stratified by stage suggested that a recent pregnancy (within 2 years of diagnosis) was associated with decreased survival as compared with women who had not had a recent pregnancy among women with nonlocalized tumors. Thirty-eight per cent of the women with nonlocalized tumors who had a recent pregnancy died within 5 years of diagnosis as compared with 15% of the women who had not had a recent pregnancy. This finding corresponds with previous research reporting that a recent pregnancy has an adverse effect on survival following breast cancer diagnosis [9,10,17,22,23] and supports the idea that a recent pregnancy may be responsible for more rapid progression of disease, resulting in poorer prognosis. The association was not observed among women with localized tumors, perhaps because the survival rate in these women is generally high.
Strengths and limitations
The hypotheses and findings presented here have biologic plausibility because of the established relationship between estrogen levels and both breast tissue proliferation and breast cancer risk. Estrogen and other hormone exposures are known to be affected substantially by pregnancy and other reproductive events. Therefore, it is biologically plausible that reproductive events may influence the progression of breast cancer among women in their childbearing years via a hormonal pathway.
Including only living cases at the time of the study might have introduced a survival bias among participants. This might have reduced the proportion of women in the study with advanced stage disease. Participants were interviewed within 2 years after diagnosis and, given that breast cancer is generally not a rapidly fatal disease, this may not have introduced much bias. In fact, fewer than 6% of potential participants died before study contact overall.
Another limitation is that there was not much information available regarding screening practices. Although we do not expect women in this age group to be receiving screening mammography, it was not possible to adjust for the effect of screening behavior (breast self examinations or clinical breast examinations) in the analyses. Therefore, if some of the independent variables such as parity and age at first full-term pregnancy are associated with screening practices, then the effects of these independent variables may have been misconstrued. However, it is likely that reproductive events such as pregnancy and childbirth would be associated with more frequent screening procedures because these events place women in the health care system. If so, then the independent effects of these variables on tumor characteristics may have been underestimated.
It is possible that socioeconomic status (SES) confounded the association between reproductive variables and tumor characteristics. Study participants were not well characterized with respect to SES. Education as a surrogate measure of SES was included as a covariate in the analysis, although it was not found to be significantly associated with any of the tumor characteristics under study (stage, grade, or ER or PgR status). Future studies should take care in characterizing participants with respect to SES.
Adjuvant therapy information was not available for all participants and was not included in the survival analyses. It should also be noted that a fairly small sample size limited the extent of multivariate analyses that could be performed. Furthermore, follow-up information was not available for the patients diagnosed in Massachusetts, which further limited the sample size available for survival analyses. However, because breast cancer diagnosis at age under 35 years is a rare event, large sample sizes for population-based studies of this subgroup are often not available, and this study made use of a resource of such patients.
Conclusion
Reproductive events such as pregnancy play an important role in the etiology of breast cancer. Whether the effect of pregnancy on breast cancer initiation or progression is protective or harmful may depend on the timing of the event in relation to the carcinogenesis process. Women who have had full-term pregnancies, particularly beginning at an early age, have consistently been observed to be at decreased lifetime risk for developing breast cancer as compared with nulliparous women or those who have a first full-term pregnancy at an older age. This may be due in part to the differentiation of breast tissue that occurs during a full-term pregnancy.
The findings of the present study suggest that for women diagnosed with breast cancer before age 35 years, increased parity and an early age at first full-term pregnancy are associated with more advanced stage at presentation. Because all of the women in this study were breast cancer patients, it is possible that the pregnancies had an adverse effect on progression of a breast cancer that was already developing. However, this study suggests that the pregnancy need not be particularly recent to have the adverse effect. Therefore, if a woman has a full-term pregnancy and has not already had a prior initiation event, then the pregnancy would be expected to leave the breast tissue more differentiated and less susceptible to carcinogenic influences in the long term. However, the findings of the present study suggest that a full-term pregnancy may have an adverse effect on a breast cancer that is developing. Studies documenting a transient increase in breast cancer risk following pregnancy also support this theory [6-8].
An understanding of the role of reproductive and hormonal exposures in the etiology and progression of breast cancer in young women highlights some of the age-related issues in breast cancer research by emphasizing that well established risk factors in older women may not have similar effects in the disease process in younger women, whose diagnoses occur nearer in time to reproductive events such as pregnancy.
Abbreviations
CI = confidence interval; ER = estrogen receptor; HR = hazard ratio; OR = odds ratio; PgR = progesterone receptor; SEER = Surveillance, Epidemiology, and End Results; SES = socioeconomic status.
Competing interests
The author(s) declare that they have no competing interests.
Authors' contributions
JAL performed the statistical analysis under the supervision of AZ and wrote the manuscript under the supervision of HA-C. All authors read and approved the final manuscript.
Acknowledgements
Funding for the study resource was provided by Department of Defense grant number DAMD17-94-J-4450 and by National Cancer Institute grant number R01 CA058860. We gratefully acknowledge the support and assistance of Dr Frederick Li and his staff at the Dana Farber Institute, as well as the staff of the participating regional and state cancer registries. In addition, we should like to extend special thanks to Trini Scott, CTR, for assistance with coding of pathology reports, and to the women who participated in the study.
Figures and Tables
Figure 1 Survival by number of full-term pregnancies and stage. (a) Localized tumors, 0–1 full-term pregnancies; (b) localized tumors, ≥ 2 full-term pregnancies; (c) nonlocalized tumors, 0–1 full-term pregnancies; and (d) non-localized tumors, ≥2 full-term pregnancies (P = 0.19 for localized tumors; P = 0.15 for nonlocalized tumors).
Figure 2 Survival by time since last pregnancy and stage. (a) Localized tumors, time since last pregnancy <24 months; (b) localized tumors, time since last pregnancy ≥ 24 months; (c) nonlocalized tumors, time since last pregnancy <24 months; and (d) nonlocalized tumors, time since last pregnancy ≥ 24 months (P = 0.24 for localized tumors; P = 0.06 for nonlocalized tumors).
Table 1 Demographic and tumor characteristics of early onset breast cancer patients
Characteristic Frequency (n) %
Age at diagnosis (years)
20–24 7 2.7
25–29 55 21.6
30–34 192 75.6
Race/ethnicity
Non-Hispanic white 185 72.8
African-American 17 6.7
Hispanic 33 13.0
Asian/Pacific islander 5 2.0
Native American 1 0.4
Other 13 5.1
Education
Less than college degree 143 56.3
College graduate 111 43.7
Histology
Infiltrating duct carcinoma 193 76.0
Comedocarcinoma 15 5.9
Medullary carcinoma 9 3.5
Lobular carcinoma 8 3.2
Infiltrating duct and lobular carcinoma 8 3.2
Other 21 8.3
Stage
Localized 131 51.6
Nonlocalized 116 45.7
Unknown 7 2.8
Grade/differentiation
Grade I/well differentiated 9 3.0
Grade II/moderately differentiated 72 24.2
Grade III/poorly differentiated 124 41.6
Grade IV/undifferentiated 8 2.7
Grade not determined/not stated/NA 85 28.5
Laterality
Right 132 52.0
Left 122 48.0
Estrogen receptor status
Positive 92 36.2
Negative 87 34.2
Borderline 3 1.2
Unknown 72 28.3
Progesterone receptor status
Positive 88 34.6
Negative 85 33.5
Borderline 3 1.2
Unknown 78 30.7
Method of breast cancer detection
Self examination 205 80.7
Physician examination 19 7.5
Mammography 17 6.7
Other 12 4.7
Missing 1 0.4
Table 2 Association of stage of breast cancer diagnosis and reproductive and family history variables
Reproductive and family history characteristics Localized (n = 131) Nonlocalized (n = 116) ORa 95% CI P
Age at menarche (years)b
<12 26 (19.8) 29 (25.4) 1.0 (ref)
≥ 12 105 (80.2) 85 (74.6) 0.8 0.4–1.4 0.35
Age at first full-term pregnancy (years)
No full-term pregnancies 48 (36.6) 29 (25.0) 1.0 (ref)
<20 12 (9.2) 19 (16.4) 3.0 1.2–7.4 0.02
20–24 31 (23.7) 35 (30.2) 2.1 1.0–4.3 0.04
25–29 25 (19.1) 24 (20.7) 1.7 0.8–3.6 0.15
30–34 15 (11.4) 9 (7.8) 1.1 0.4–2.9 0.86
Trend <0.01
Total live births
0 48 (36.6) 29 (25.0) 1.0 (ref)
1–2 71 (54.2) 67 (57.8) 1.7 0.9–3.1 0.07
3 or more 12 (9.2) 20 (17.2) 3.1 1.3–7.7 0.01
Trend 0.01
Total miscarriages (gravid women)
0 69 (74.2) 67 (72.0) 1.0 (ref)
1 or more 24 (25.8) 26 (28.0) 1.2 0.6–2.3 0.59
Total induced abortions (gravid women)
0 65 (69.9) 59 (63.4) 1.0 (ref)
1 or more 28 (30.1) 34 (36.6) 1.2 0.7–2.3 0.51
Breastfeeding (parous women)
Never 25 (30.1) 26 (29.9) 1.0 (ref)
Ever 58 (69.9) 61 (70.1) 1.1 0.5–2.2 0.80
Months between last full-term pregnancy and diagnosis
No full-term pregnancy 48 (36.6) 29 (25.0) 1.0 (ref)
0–23 23 (17.6) 26 (22.4) 1.9 0.9–3.9 0.09
24 or more 60 (45.8) 61 (52.6) 1.9 1.0–3.5 0.05
Trend 0.06
Oral contraceptive use
Never 19 (14.5) 21 (18.1) 1.0 (ref)
Ever 112 (85.5) 95 (81.9) 0.8 0.4–1.5 0.43
Total months of oral contraceptive usec
None 19 (14.6) 21 (18.9) 1.0 (ref)
1–35 35 (26.9) 29 (26.1) 0.6 0.3–1.3 0.22
36–83 33 (25.4) 29 (26.1) 0.7 0.3–1.4 0.30
84 or more 43 (33.1) 32 (28.8) 0.6 0.3–1.2 0.17
Trend 0.46
Family history of breast or ovarian cancerd
No family history 109 (83.8) 95 (82.6) 1.0 (ref)
Mother or sister 21 (16.2) 20 (17.4) 1.2 0.6–2.3 0.67
aAnalyses adjusted for age at diagnosis and education (higher of self or spouse). bTwo records missing age at menarche. cSix records missing total months of oral contraceptive use. dTwo records were excluded because proband reported being adopted. CI, confidence interval; OR, odds ratio.
Table 3 Association of breast cancer grade by reproductive and family history variables
Reproductive and hormonal Grade 1–2 Grade 3–4 OR (for grade 95% CI P
Age at menarche (years)
<12 15 (18.8) 33 (25.4) 1.0 (ref)
≥ 12 65 (81.2) 97 (74.6) 0.7 0.4–1.5 0.38
Age at first full-term pregnancy (years)
No full-term pregnancies 27 (33.3) 35 (26.5) 1.0 (ref)
<20 5 (6.2) 23 (17.4) 3.2 1.0–9.9 0.05
20–24 25 (30.9) 34 (25.8) 1.0 0.4–2.1 0.95
25–29 17 (21.0) 26 (19.7) 1.4 0.6–3.1 0.43
30–34 7 (8.6) 14 (10.6) 2.5 0.8–7.5 0.10
Trend 0.27
Total live births
0 27 (33.3) 35 (26.5) 1.0 (ref)
1–2 45 (55.6) 78 (58.3) 1.5 0.8–2.8 0.25
3 or more 9 (11.1) 19 (15.2) 1.7 0.6–4.6 0.30
Trend 0.22
Total miscarriages (gravid women)
0 50 (82.0) 73 (70.2) 1.0 (ref)
1 or more 11 (18.0) 31 (29.8) 2.0 0.9–4.5 0.08
Total induced abortions (gravid women)
0 40 (65.6) 68 (65.4) 1.0 (ref)
1 or more 21 (34.4) 36 (34.6) 1.0 0.5–1.9 0.92
Breastfeeding (parous women)
Never 15 (27.8) 34 (35.0) 1.0 (ref)
Ever 39 (72.2) 63 (65.0) 0.8 0.4–1.6 0.51
Months between last full-term pregnancy and diagnosis
No full-term pregnancy 27 (33.3) 35 (26.5) 1.0 (ref)
0–23 19 (23.4) 27 (20.4) 1.2 0.5–2.6 0.71
≥ 24 35 (43.2) 70 (53.0) 1.7 0.9–3.5 0.12
Trend 0.12
Oral contraceptive use
Never 10 (12.4) 21 (15.9) 1.0 (ref)
Ever 71 (87.6) 111 (84.1) 0.7 0.3–1.6 0.42
Total months of oral contraceptive useb
None 10 (13.2) 21 (16.0) 1.0 (ref)
1–35 24 (31.6) 32 (24.4) 0.8 0.4–2.0 0.68
36–83 17 (22.4) 38 (29.0) 1.5 0.6–3.7 0.35
84 or more 25 (32.9) 40 (30.5) 1.2 0.5–2.8 0.64
Trend 0.78
Family history of breast or ovarian cancerc
No family history 65 (80.2) 113 (86.9) 1.0 (ref)
Mother or sister 16 (19.8) 17 (13.1) 0.7 0.3–1.4 0.29
aAnalyses adjusted for age at diagnosis and education (higher of self or spouse). bFive records missing total months of oral contraceptive use. cTwo records were excluded because proband reported being adopted. CI, confidence interval; OR, odds ratio.
Table 4 Association of estrogen receptor status and reproductive and family history variables
Reproductive and family history characteristics ER-positive or borderline tumors (n = 95) ER-negative tumors (n = 87) ORa 95% CI P
Age at menarche (years)b
<12 16 (17.0) 19 (22.4) 1.0 (ref)
≥ 12 78 (83.0) 66 (77.6) 0.7 0.3–1.5 0.42
Age at first full-term pregnancy (years)
No full-term pregnancies 27 (28.4) 27 (31.0) 1.0 (ref)
<20 10 (10.5) 16 (18.4) 1.5 0.6–4.0 0.43
20–24 26 (27.4) 19 (21.8) 0.7 0.3–1.6 0.39
25–29 23 (24.2) 16 (18.4) 0.7 0.3–1.7 0.45
30–34 9 (9.5) 9 (10.3) 1.2 0.4–3.6 0.80
Trend 1.0
Total live births
0 27 (28.4) 27 (31.0) 1.0 (ref)
1–2 57 (60.0) 49 (56.3) 0.9 0.4–1.7 0.68
3 or more 11 (11.6) 11 (12.6) 1.0 0.3–2.8 0.97
Trend 0.86
Total miscarriages (gravid women)
0 61 (80.3) 42 (65.6) 1.0 (ref)
1 or more 15 (19.7) 22 (34.4) 2.1 1.0–4.6 0.05
Total induced abortions (gravid women)
0 48 (63.2) 44 (68.8) 1.0 (ref)
1 or more 28 (26.8) 20 (31.2) 0.7 0.4–1.5 0.42
Breastfeeding (parous women)
Never 12 (17.6) 29 (48.3) 1.0 (ref)
Ever 56 (82.4) 31 (51.7) 0.2 0.1–0.5 <0.001
Months between last full-term pregnancy and diagnosis
No full-term pregnancy 27 (28.4) 27 (31.0) 1.0 (ref)
0–23 19 (20.0) 16 (18.4) 0.9 0.4–2.0 0.71
≥ 24 49 (51.6) 44 (50.6) 0.9 0.4–1.8 0.77
Trend 0.78
Oral contraceptive use
Never 19 (20.0) 12 (13.8) 1.0 (ref)
Ever 76 (80.0) 75 (86.2) 1.6 0.7–3.4 0.28
Total months of oral contraceptive usec
None 19 (20.2) 12 (14.6) 1.0 (ref)
1–35 24 (25.5) 23 (28.0) 1.1 0.5–2.6 0.81
36–83 16 (17.0) 20 (24.4) 1.5 0.6–3.7 0.42
84 or more 35 (37.2) 27 (32.9) 0.9 0.4–2.1 0.85
Trend 0.78
Family history of breast or ovarian cancerd
No family history 75 (79.8) 71 (82.6) 1.0 (ref)
Mother or sister 19 (20.2) 15 (17.4) 0.9 0.4–1.9 0.72
aAnalyses adjusted for age at diagnosis and education (higher of self or spouse).
bThree records missing age at menarche. cSix records missing total months of oral contraceptive use. dTwo records were excluded because proband reported being adopted. CI, confidence interval; ER, estrogen receptor; OR, odds ratio.
Table 5 Association of progesterone receptor status and reproductive and family history variables
Reproductive and family history characteristics PgR-positive or borderline tumors (n = 91) PgR-negative tumors (n = 85) ORa 95% CI P
Age at menarche (years)b
<12 14 (15.6) 21 (25.3) 1.0 (ref)
≥ 12 76 (84.4) 62 (74.7) 0.6 0.3–1.2 0.14
Age at first full-term pregnancy
No full-term pregnancies 27 (29.7) 25 (29.4) 1.0 (ref)
<20 10 (11.0) 15 (17.6) 1.6 0.6–4.4 0.37
20–24 25 (27.5) 20 (23.5) 0.8 0.4–2.0 0.71
25–29 19 (20.9) 18 (21.2) 1.1 0.5–2.6 0.82
30–34 10 (11.0) 7 (8.2) 0.9 0.3–2.9 0.88
Trend 0.65
Total live births
0 27 (29.7) 25 (29.4) 1.0 (ref)
1–2 53 (58.2) 50 (58.8) 1.1 0.5–2.1 0.87
3 or more 11 (12.1) 10 (11.8) 1.0 0.3–2.9 1.00
Trend 0.95
Total miscarriages (gravid women)
0 54 (76.1) 45 (69.2) 1.0 (ref)
1 or more 17 (23.9) 20 (30.8) 1.4 0.7–3.0 0.38
Total induced abortions (gravid women)
0 46 (64.8) 44 (67.7) 1.0 (ref)
1 or more 25 (35.2) 21 (32.3) 0.8 0.4–1.7 0.65
Breastfeeding (parous women)
Never 14 (21.9) 25 (41.7) 1.0 (ref)
Ever 50 (78.1) 35 (58.3) 0.4 0.2–0.8 0.02
Months between last full-term pregnancy and diagnosis
No full-term pregnancy 27 (29.7) 25 (29.4) 1.0 (ref)
0–23 17 (18.7) 16 (18.8) 1.0 0.4–2.5 0.95
≥ 24 47 (51.6) 44 (51.8) 1.1 0.5–2.2 0.87
Trend 0.87
Oral contraceptive use
Never 19 (20.9) 10 (11.8) 1.0 (ref)
Ever 72 (79.1) 75 (88.2) 2.0 0.8–4.6 0.11
Total months of oral contraceptive usec
None 19 (21.3) 10 (12.2) 1.0 (ref)
1–35 23 (25.8) 23 (28.0) 1.6 0.6–3.9 0.33
36–83 14 (15.7) 20 (24.4) 2.3 0.9–6.1 0.10
84 or more 33 (37.1) 29 (35.4) 1.5 0.6–3.5 0.38
Trend 0.34
Family history of breast or ovarian cancerd
No family history 71 (78.9) 70 (83.3) 1.0 (ref)
Mother or sister 19 (21.1) 14 (16.7) 0.8 0.4–1.8 0.58
aAnalyses adjusted for age at diagnosis and education (higher of self or spouse).
bThree records missing age at menarche. cFive records missing total months of oral contraceptive use. dTwo records were excluded because proband reported being adopted. CI, confidence interval; PgR, progesterone receptor; OR, odds ratio.
==== Refs
Ries LAG Eisner MP Kosary CL Hankey BF Miller BA Clegg L Mariotto A Feuer EJ Edwards BK editors SEER Cancer Statistics Review, 1975–2001 2004 Bethesda, MD: National Cancer Institute Last accessed 29 April 2005
Perkins CI Cohen R Morris CR Allen M Kwong SL Schlag R Wright WE Cancer in California: 1988–1995 1998 Sacramento, CA: California Department of Health Services, Cancer Surveillance Section
Winchester DP Osteen RT Menck HR The National Cancer Data Base report on breast carcinoma characteristics and outcome in relation to age Cancer 1996 78 1838 1843 8859200 10.1002/(SICI)1097-0142(19961015)78:8<1838::AID-CNCR27>3.0.CO;2-Y
Chung M Chang HR Bland KI Wanebo HJ Younger women with breast carcinoma have a poorer prognosis than older women Cancer 1996 77 97 103 8630946 10.1002/(SICI)1097-0142(19960101)77:1<97::AID-CNCR16>3.0.CO;2-3
Gajdos C Tartter PI Bleiweiss IJ Bodian C Brower ST Stage 0 to stage III breast cancer in young women J Am Coll Surg 2000 190 523 529 10801018 10.1016/S1072-7515(00)00257-X
Lambe M Hsieh C Trichopoulos D Ekbom A Pavia M Adami H-O Transient increase in the risk of breast cancer after giving birth N Engl J Med 1994 331 5 9 8202106 10.1056/NEJM199407073310102
Tavani A Gallus S La Vecchia C Negri E Montella M Dal Maso L Franceschi S Risk factors for breast cancer in women under 40 years Eur J Cancer 1999 35 1361 1367 10658528 10.1016/S0959-8049(99)00139-2
Wohlfahrt J Andersen PK Mouridsen HT Melbye M Risk of late-stage breast cancer after a childbirth Am J Epidemiol 2001 153 1079 1084 11390326 10.1093/aje/153.11.1079
Guinee VF Olsson H Moller T Hess KR Taylor SH Fahey T Gladikov JV van den Blink JW Bonichon F Dische S Effect of pregnancy on prognosis for young women with breast cancer Lancet 1994 343 1587 1589 7911917 10.1016/S0140-6736(94)93054-6
Phillips KA Milne RL Friedlander ML Jenkins MA McCredie MR Giles GG Hopper JL Prognosis of premenopausal breast cancer and childbirth prior to diagnosis J Clin Oncol 2004 22 699 705 14966094 10.1200/JCO.2004.07.062
American Cancer Society, California Division, and Public Health Institute, California Cancer Registry California Cancer Facts and Figures, 1999 1999 Oakland, CA: American Cancer Society, California Division
SEER Surveillance, Epidemiology, and End Results (SEER) Program Public-Use Data (1973–1999) 2002 Bethesda, MD: National Cancer Institute, DCCPS, Surveillance Research Program, Cancer Statistics Branch based on the November 2001 submission
SEER Summary Staging Guide for the Cancer Surveillance, Epidemiology and End Results Reporting (SEER) Program 1977 Bethesda, MD: US Department of Health and Human Services, Public Health Services, National Institutes of Health reprinted July 1986
American Joint Committee on Cancer Manual for Staging of Cancer 1997 5 Philadelphia, PA: Lippincott-Raven Publishers
Bernstein L Ross RK Henderson BE Prospects for the primary prevention of breast cancer Am J Epidemiol 1992 135 142 152 1311141
Hiraku Y Yamashita N Nishiguchi M Kawanishi S Catechol estrogens induce oxidative DNA damage and estradiol enhances cell proliferation Int J Cancer 2001 92 333 337 11291067 10.1002/ijc.1193
Daling JR Malone KE Doody DR Anderson BO Porter PL The relation of reproductive factors to mortality from breast cancer Cancer Epidemiol Biomarkers Prev 2002 11 235 241 11895871
Schouten LJ Hupperets PSGJ Jager JJ Volovics L Wils JA Verbeek ALM Blijham GH Prognostic significance of etiological risk factors in early breast cancer Breast Cancer Res Treat 1997 43 217 223 9150901 10.1023/A:1005790531122
Kroman N Wohlfahrt J Andersen KW Mouridsen HT Westergaard T Melbye M Parity, age at first childbirth and the prognosis of primary breast cancer Br J Cancer 1998 78 1529 1533 9836489
Chie WC Hsieh CC Newcomb PA Longnecker MP Mittendorf R Greenberg ER Clapp RW Burke KP Titus-Ernstoff L Tentham-Dietz AM Age at any full-term pregnancy and breast cancer risk Am J Epidemiol 2000 151 715 722 10752799
Huang W Newman B Millikan RC Schell MJ Hulka BS Moorman PG Hormone-related factors and risk of breast cancer in relation to estrogen receptor and progesterone receptor status Am J Epidemiol 2000 151 703 714 10752798
Kroman N Wholfahrt J West Andersen K Mouridsen HT Westergaard T Melbye M Time since childbirth and prognosis in primary breast cancer: population based study BMJ 1997 315 851 855 9353505
Olson SH Zauber AG Tang J Harlap S Relation of time since last birth and parity to survival of young women with breast cancer Epidemiology 1998 9 669 671 9799180 10.1097/00001648-199811000-00015
| 15987461 | PMC1175072 | CC BY | 2021-01-04 16:04:33 | no | Breast Cancer Res. 2005 May 16; 7(4):R541-R554 | utf-8 | Breast Cancer Res | 2,005 | 10.1186/bcr1198 | oa_comm |
==== Front
BMC BioinformaticsBMC Bioinformatics1471-2105BioMed Central London 1471-2105-6-1381593875010.1186/1471-2105-6-138Softwarelibcov: A C++ bioinformatic library to manipulate protein structures, sequence alignments and phylogeny Butt Davin [email protected] Andrew J [email protected] Christian [email protected] Faculty of Computer Science, Dalhousie University, 6050 University Ave. Halifax, NS, B3H 1W5, Canada2 Dept. of Biochemistry and Molecular Biology, Dalhousie University, Tupper Medical Building, Halifax, NS, B3H 1X5, Canada3 Canadian Institute for Advanced Research (CIAR)2005 6 6 2005 6 138 138 5 11 2004 6 6 2005 Copyright © 2005 Butt 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
An increasing number of bioinformatics methods are considering the phylogenetic relationships between biological sequences. Implementing new methodologies using the maximum likelihood phylogenetic framework can be a time consuming task.
Results
The bioinformatics library libcov is a collection of C++ classes that provides a high and low-level interface to maximum likelihood phylogenetics, sequence analysis and a data structure for structural biological methods. libcov can be used to compute likelihoods, search tree topologies, estimate site rates, cluster sequences, manipulate tree structures and compare phylogenies for a broad selection of applications.
Conclusion
Using this library, it is possible to rapidly prototype applications that use the sophistication of phylogenetic likelihoods without getting involved in a major software engineering project. libcov is thus a potentially valuable building block to develop in-house methodologies in the field of protein phylogenetics.
==== Body
Background
With the development of genomics, research in biology and systems biology is becoming increasingly data-driven. The feedback between available data and hypotheses has accelerated the pace at which innovative ideas are generated. Life scientists are in a position to design novel methodologies but do not necessarily have the in-house skills to produce software implementations. Simple methods, made of complex building blocks such as maximum likelihood calculations, require major software development projects before they can be prototyped. The use of libraries can help to rapidly prototype software implementations.
We present libcov, an object-oriented library to perform phylogenetic inference and the manipulation of protein sequences and structures. The library is written in C++, is compliant with the GNU standards and packaged as a dynamic library that can be installed on most Unix distributions (including MacOS X).
There are other bioinformatic libraries available, many of which overlap with libcov in their functionalities. The PAL library, for example, [1] is a Java implementation which offers a versatile object set for nucleotide and protein phylogeny. More generally, interested readers can visit the Open Bioinformatics Foundation [2] that links to a series of libraries written in various popular scripting languages such as Perl and Python. Further, there are other libraries available in C++ such as the Bioinformatics Template library BTL [3], and the compBioTool++[4], both of which focus on sequence manipulation.
The scope of libcov is to offer a series of high-level functions that can be invoked in one line of code, and which does not force an implementation to adopt specialized custom types. As for any open source project, it is possible to use or extend the low-level Application Programming Interface (API) to add functionalities or entirely new modules.
Implementation
Libcov offers a high-level programming interface, using an Object-Oriented (OO) approach with classes to represent distinct identities. For example, the class covTree represents a phylogenetic tree, covAlignment is used to store alignments, and class PDBentity handles 3D protein structures from the PDB format. The PDBentity class is a hierarchical structure of peptide chains, residues and atoms. Other classes handle elements such as geometric transformations and substitution matrices.
Libcov is designed as a protein phylogeny library. The data structures and methods that its public interface offers can be integrated within application prototypes with a minimal impact on software design. Most of the return types are Standard Template Library (STL) containers, which can be seamlessly integrated into ongoing software projects. Specialized classes can be derived by consulting the online API documentation. Examples of integration of libcov within C++ source codes are presented in Figure 1.
A summary of the functions offered by libcov is presented in TABLE 1. A more complete list of methods is available at the project's website.
Currently, we have implemented three major applications using libcov. covTREE is our protein sequence simulator that has the ability to simulate complex patterns of protein evolution and phylogenetic artifacts[5]. It uses the Monte Carlo-based simulation functions that libcov provides. covSEARCH is a tree searching program using the maximum likelihood and tree re-arrangement algorithms in libcov. covARES maps sequence and phylogenetic information on to protein models [6]. These applications are also available to the research community under a GNU GPL license.
Conclusion
The libcov library is actively under development, and we will be frequently releasing updated versions. As libcov is the engine powering the phylogenetic application covSEARCH, future work will involve new algorithms of tree searching, confidence interval determination and the integration of structure-based models of substitution.
External contributions are welcomed as the functionality of the library will evolve to match the research interests of the developers of phylogenetics applications.
Availability and Requirements
Project's name: libcov
Project's website:
Operating System: GNU C++ library. Tested on Linux, MacOSX and other Unix-based operating systems.
License: GPL
Non-academic licensing: None.
Authors' contributions
C. Blouin – Scientific Functionalities, High-level design, Redaction of manuscript.
D. Butt – Software design, implementation and testing.
A.J. Roger – Scientific functionalities, redaction of manuscript.
Acknowledgements
This work was supported by Genome Atlantic grant on Prokaryotic genome diversity and evolution, and by the NSERC Discovery grant 298397-04 (CB). The author would like to thank J. Murdoch for her contribution to the treeSPACE module.
Figures and Tables
Figure 1 The likelihood of a consensus tree. In this example, a file containing trees in NEWICK format is parsed and a consensus tree is resolved using the greedy majority-rule consensus algorithm [9, 21]. Finally, the likelihood of the resulting tree is calculated. Bolded lines are libcov API calls.
Table 1 High Level functionalities
Category Method Reference
I/O Tree (NEWICK)
Sequences (FASTA, PHYLIP)
Protein Structure (PDB)
Tree manipulation Random/exhaustive:
Subtree Pruning Regrafting (SPR)
Tree Bissection Reconnection (TBR)
Nearest Neighbor interchange (NNI)
Branch Swapping
Stepwise addition [7]
Phylogeny Neighbor Joining [8]
Greedy Majority-rule consensus [9]
Maximum Likelihood Rates across site modeling Estimation of shape parameter α [10]
ML confidence intervals KH [11]
SH [12]
RELL [13]
Expected Likelihood Weights [14]
ML performance P-matrix caching Chebyshev Polynomial approximation [15]
Substitution matrices JTT [16]
PAM [17]
WAG [18]
Simulation Protein Sequence Simulation (Rates across sites, rate shifts, site specific frequencies, multiple datasets, likelihood computation) [5, 19]
Random Number generation [20] Acknowl. Z. Yang for implementation in PAML
Structural Biology Manipulation / mapping Neighboring site anisotropy (NSA) Geometric transformations Distance/Contact Matrices [6]
==== Refs
Drummond A Strimmer K PAL: an object-oriented programming library for molecular evolution and phylogenetics Bioinformatics 2001 17 662 663 11448888 10.1093/bioinformatics/17.7.662
O|B|F Open Bioinformatics Foundation
Williams M The Bioinformatics Template Library (BTL)
Durbin KJ CompBioTools++
Blouin C Butt DJ Roger AJ The impact of taxon sampling on the estimation of rates of evolution at sites. Mol Biol Evol 2005 22 784 791 15590908 10.1093/molbev/msi065
Blouin C Boucher Y Roger AJ Inferring functional constraints and divergence in protein families using 3D mapping of phylogenetic information Nucleic Acids Res 2003 31 790 797 12527789 10.1093/nar/gkg151
Felsenstein J Inferring Phylogenies 2004 1 Sunderland, MA, Sinauer Associates, Inc. 664
Saitou N Nei M The neighbor-joining method: a new method for reconstructing phylogenetic trees Mol Biol Evol 1987 4 406 425 3447015
Felsenstein J PHYLIP (Phylogeny Inference Package) version 3.6 2002 Seattle, Wa., Distributed by the author, Dept. of Genetics, U. of Washington
Yang Z Among-site rate variation and its impact on phylogenetic analyses. Trends Ecol Evol 1996 11 367 372 10.1016/0169-5347(96)10041-0
Kishino H Hasegawa M Evaluation of the maximum likelihood estimate of the evolutionary tree topologies from DNA sequence data, and the branching order in hominoidea J Mol Evol 1989 29 170 179 2509717
Shimodaira H Hasegawa M Multiple Comparisons of Log-Likelihoods with Applications to Phylogenetic Inference Mol Biol Evol 1999 16 1114 1116
Kishino H Miyata T Hasegawa M Maximum Likelihood inference of protein phylogeny and the origin of chloroplasts J Mol Evol 1990 30 151 160
Strimmer K Rambaut A Inferring confidence sets of possibly misspecified gene trees Proc R Soc Lond B Biol Sci 2002 269 137 142 10.1098/rspb.2001.1862
Pupko T Graur D Fast computation of maximum likelihood trees by numerical approximation of amino acid replacement probabilities Computational Statistics & Data Analysis 2002 40 285 291 10.1016/S0167-9473(02)00008-7
Jones DT Taylor WR Thornton JM The rapid generation of mutation data matrices from protein sequences Comput Appl Biosci 1992 8 275 282 1633570
Dayhoff MO Schwartz RM Orcutt BC Dayhoff MO A model of evolutionary change in proteins Atlas of protein sequence and structure 1978 5 Silver Spring, MA, National Biomedical Research Foundation 345 352
Whelan S Goldman N A general empirical model of protein evolution derived from multiple protein families using a maximum-likelihood approach Mol Biol Evol 2001 18 691 699 11319253
Grassly NC Adachi J Rambaut A PSeq-Gen: an application for the Monte Carlo simulation of protein sequence evolution along phylogenetic trees Comput Appl Biosci 1997 13 559 560 9367131
Wichmann BA Hill ID An efficient and portable pseudo-random number generator. Appl Stat 1982 31 188 190
Bryant D Janowitz M, Lapointe FJ, McMorris FR, Mirkin B and Roberts FS A Classifcation of Consensus Methods for Phylogenetics BioConsensus 2003 , DIMACS. AMS. 164 184
| 15938750 | PMC1175080 | CC BY | 2021-01-04 16:02:50 | no | BMC Bioinformatics. 2005 Jun 6; 6:138 | utf-8 | BMC Bioinformatics | 2,005 | 10.1186/1471-2105-6-138 | oa_comm |
==== Front
BMC BioinformaticsBMC Bioinformatics1471-2105BioMed Central London 1471-2105-6-1561596976910.1186/1471-2105-6-156SoftwaretransAlign: using amino acids to facilitate the multiple alignment of protein-coding DNA sequences Bininda-Emonds Olaf RP [email protected] Lehrstuhl für Tierzucht, Technical University of Munich, Hochfeldweg 1, 85354 Freising-Weihenstephan, Germany2005 22 6 2005 6 156 156 14 4 2005 22 6 2005 Copyright © 2005 Bininda-Emonds; licensee BioMed Central Ltd.This is an Open Access article distributed under the 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
Alignments of homologous DNA sequences are crucial for comparative genomics and phylogenetic analysis. However, multiple alignment represents a computationally difficult problem. For protein-coding DNA sequences, it is more advantageous in terms of both speed and accuracy to align the amino-acid sequences specified by the DNA sequences rather than the DNA sequences themselves. Many implementations making use of this concept of "translated alignments" are incomplete in the sense that they require the user to manually translate the DNA sequences and to perform the amino-acid alignment. As such, they are not well suited to large-scale automated alignments of large and/or numerous DNA data sets.
Results
transAlign is an open-source Perl script that aligns protein-coding DNA sequences via their amino-acid translations to take advantage of the superior multiple-alignment capabilities and speed of an amino-acid alignment. It operates by translating each DNA sequence into its corresponding amino-acid sequence, passing the entire matrix to ClustalW for alignment, and then back-translating the resulting amino-acid alignment to derive the aligned DNA sequences. In the translation step, transAlign determines the optimal orientation and reading frame for each DNA sequence according to the desired genetic code. It also checks for apparent frame shifts in the DNA sequences and can handle frame-shifted sequences in one of three ways (delete, align as amino acids regardless, or profile align as DNA). As a set of comparative benchmarks derived from six protein-coding genes for mammals shows, the strategy implemented in transAlign always improves the speed and usually the apparent accuracy of the alignment of protein-coding DNA sequences.
Conclusion
transAlign represents one of few full and cross-platform implementations of the concept of translated alignments. Both the advantages accruing from performing a translated alignment and the suite of user-definable options available in the program mean that transAlign is ideally suited for large-scale automated alignments of very large and/or very numerous protein-coding DNA data sets. However, the good performance offered by the program also translates to the alignment of any set of protein-coding sequences. transAlign, including the source code, is freely available at http://www.tierzucht.tum.de/Bininda-Emonds/ (under "Programs").
==== Body
Background
Alignments of homologous DNA sequences are crucial for comparative genomics and phylogenetic analysis [1]. The most accurate multiple alignment tool arguably remains the human eye. However, the increasing amount of sequence data and the increasing scope of projects using these data mean that an automated alignment procedure is often necessary at some point to achieve the final alignment.
For protein-coding DNA sequences, alignments obtained from the amino-acid residues specified by the DNA sequences will often be superior to those obtained directly from the DNA for several reasons (see also [2]). First and foremost, aligning the amino-acid residues preserves the codon structure of the coding sequence, thereby avoiding the introduction of any frame shifts through the alignment process. Second, because amino acids are more conserved evolutionarily than DNA, and possibly because the amino-acid alphabet is larger than the DNA one and therefore less likely to become saturated with convergent substitutions over longer timeframes, it is often easier to align amino-acid sequences between more distantly related organisms. Third, unlike for nucleotide data, the transition matrices that exist for amino-acid data (e.g., BLOSUM [3], GONNET [4], or PAM [5]) are empirically derived and thus perhaps more "biologically realistic". The many different possible models of nucleotide evolution (see [6]) and the fact that different genes evolve according to different models makes the likelihood of obtaining an equivalent, global nucleotide transition matrix small. Finally, because the translated amino-acid sequence is one-third as long as the original DNA sequence, the alignment procedure will be faster. Based purely on the differences in sequence length, the speedup would be on the order of a factor of nine, given that the Smith-Waterman [7] algorithm for the pairwise alignment of sequences that underlies many multiple-alignment programs runs in O(n2), where n = length of the sequence (i.e., is proportional to the product of the lengths of the sequences). However, other considerations, including the speed of the different scoring routines that could be implemented for DNA versus amino-acid data or the memory usage and general implementation of the system, will also be important determinants of the final relative speed increase.
One limitation to aligning amino-acid residues is that the redundancy of the genetic code, whereby up to six sets of nucleotide triplets can specify the same amino acid, means that it is not possible to back-translate an amino-acid sequence without recourse to the corresponding DNA sequence. Numerous programs exist to back-translate aligned amino-acid sequences – for example, the standalone version of RevTrans [2] and mrtrans [8] – but most require both the aligned sequences and the corresponding, unaligned DNA sequences as input. As such, the investigator must determine the proper translation frame for each sequence and perform the amino-acid alignment beforehand, which does not lend itself to the automated alignment of large numbers of DNA sequences.
The server version of RevTrans [9] goes a step further by optionally taking DNA sequences as input, virtually translating them into their respective amino-acid sequences, aligning the latter using DIALIGN2 [10], and then back-translating to achieve the DNA alignment. Altogether, this strategy makes use of the superior and faster alignments produced by amino-acid data, while retaining the greater information content of the DNA sequences for future analyses. Similar functionality is also built directly into DIALIGN2. However, the RevTrans server is limited to only 75 DNA sequences and does not perform any preprocessing of them. As such, is not well-suited to the automated alignment of large numbers of sequences. Both RevTrans and DIALIGN2 also make use of only the BLOSUM transition matrix. LAGAN and Multi-LAGAN [11] also offer the possibility of "translated alignments" (via the translated anchoring option), but both programs are geared more toward the alignment of long, genomic sequences.
Building on these solutions, transAlign (for translated alignments) provides the same basic functionality as the RevTrans server, but with no constraints on the number of input sequences (beyond the memory of the user's computer) and a wider selection of amino-acid transition matrices. More importantly, transAlign also offers a suite of user-defined options (described below) for manipulating either the raw sequence data or the aligned sequences. The most of important of these options relate to DNA sequences that do not translate into "clean" amino-acid sequences and thus could impact negatively on the amino-acid alignment. Together with it being a standalone program, these features make transAlign suitable for both individual data sets and as part of a pipeline for the automated alignment of large numbers of sequences downloaded directly from any of the sequence databases.
Implementation
transAlign can automatically read DNA sequences in any of four formats: fasta, nexus [12], classic [13] or "extended" [14] PHYLIP, and Se-Al [15]. It can also write the final alignment in any of these same formats. (Conversion to or from additional formats can be accomplished through other programs such as readSeq [16] or sreformat, part of the HMMER package [17].) Some basic filtering of the DNA sequences is also implemented, including the stripping of gaps (either all gaps or only those flanking a sequence) and deleting sequences with more than a user-defined percentage of ambiguous nucleotides (i.e., Ns).
After initial processing of the DNA sequences, transAlign will determine the optimal translation for each sequence according to any of the genetic codes listed by the NCBI [18]. It is also possible for Se-Al formatted data to have different genetic codes specified for each sequence. As far as possible, transAlign translates codons containing ambiguous nucleotides (but not explicit gaps). The optimal translation is held to be that yielding the fewest stop codons excluding the terminal codon. By default, only the three reading frames for the input orientation are examined; however, it is possible to examine the complemented, reversed, and reverse-complemented orientations as well. For equally optimal orientations, transAlign favours the one perturbing the original DNA sequence the least: in order, 1) the orientation as input, followed by the second and third reading frames in that orientation, and then the respective reading frames in each of the 2) complementary, 3) reverse, and 4) reverse-complementary orientations.
transAlign then passes the translated sequences to ClustalW [19,20] for alignment (according to any of the BLOSUM, GONNET or PAM protein weighting matrices) and back-translates the resulting aligned residue sequences into aligned DNA sequences. ClustalW was chosen because it is perhaps the best known and most widely used multiple-alignment program. It also offers the largest choice of amino-acid transition matrices (BLOSUM, GONNET, and PAM) and the ability to do profile alignments (see below). However, slight modifications to the transAlign code would allow the use any suitable multiple-alignment program that accepts protein sequence data as input (e.g., DIALIGN2 with its Clustal-like output in particular). Regardless of the alignment program used, it is expected that increases in both speed and accuracy compared to aligning the sequences as DNA would still occur given the many advantages for aligning protein-coding DNA sequences as amino acids (see above).
An option is also provided to automatically delete any poorly aligning sequences as determined by the initial pairwise alignments performed by ClustalW. This feature is intended largely to remove problematic sequences from alignment pipelines, where it is difficult to (manually) improve the global alignment afterwards. For each sequence, the mean of its pairwise alignment scores is compared to that between all the remaining sequences according to a one-tailed two-sample t-test corrected for multiple comparisons. As such, the procedure is most effective at identifying isolated problematic sequences, which might derive from the inclusion of a potential paralog or simply a misidentified sequence. Families of such sequences (e.g., if the data set contains numerous copies of each of the paralogs from a gene family) are less likely to be detected.
Because ClustalW ignores ambiguous amino acids and stop codons (neither being present in the amino-acid transition matrices), transAlign translates them initially as gaps to permit back-translation. This procedure is unproblematic unless the ambiguous residue or stop codon is adjacent to a gap inferred by the alignment procedure, where it could be placed at either the start or end of the gap. For ambiguous residues arising from incomplete codons, transAlign determines the more optimal of the two placements based on the concordance of the missing nucleotide(s) with the gap. However, all such instances should still be examined and, if necessary, corrected for on an individual basis during the manual inspection that follows any automated alignment procedure.
Obviously, the use of transAlign is restricted to coding DNA sequences only and should not be used for non-coding DNA, whether for genes such as 18S rDNA (= MTRNR2; [21]); flanking UTR, regulatory, or intronic regions of genes; or microsatellite sequences. The procedure is also adversely affected by frame shifts (e.g., from sequencing errors). Therefore, transAlign will minimally issue a warning for each sequence that contains more than a user-specified threshold of stop codons (excluding the terminal codon) in the optimal orientation. This threshold can either be an absolute number of stop codons (default) or a percentage of stop codons in the remaining sequence after the first stop codon is encountered. Although this procedure is generally robust, it is less likely to detect frame shifts that occur near either end of a given sequence because of the reduced probability of an erroneous stop codon arising in the few remaining resides.
Three global solutions for any frame-shifted sequences are implemented in transAlign: 1) deletion, 2) alignment using the translated sequences regardless (with the associated errors), or 3) subsequent profile alignment as DNA to the aligned set of non-frame-shifted sequences (default). The latter option is the slowest of the three, but allows all sequences to be aligned as robustly as possible. Moreover, even a partial profile alignment will always be faster than aligning all sequences as DNA (Figure 1), regardless of the actual speedup inherent to aligning the shorter amino-acid sequences. However, performance will drop off quickly as the proportion of frame-shifted sequences in the data set increases. For instance, assuming a speedup of 9x for aligning amino acids compared to DNA (which, as mentioned, is the value expected based only on length considerations), the overall time saving will only be about 2x if frame-shifted sequences comprise 25% of all sequences (see Figure 1). Finally, to facilitate the manual inspection of the dataset, transAlign will also attempt to infer putative locations for frame-shifting indels based on a comparison of gaps between the amino-acid aligned and DNA profile-aligned sequences.
As mentioned above, transAlign will output the aligned DNA sequences in any or all of fasta, nexus, (classic or extended) PHYLIP, or Se-Al formats. By default, the sequences are output in alphabetical order according to their name. However, it is also possible to output them to match their order in the original input file or as they were output from the ClustalW alignment. The latter option is particularly useful at identifying "families" of similar sequences or those sequences that were profile-aligned to facilitate any manual correction of the global alignment.
transAlign is written in Perl and is open source. It will run on any operating system with a Perl interpreter and is command-line driven. However, it also features a user-interactive mode where the user is prompted to set all the relevant variables. It requires that a remotely-callable version of ClustalW is present either in the global path or in a user-specified one. Again, however, slight modifications to the code would allow the use any suitable multiple-alignment program.
Results and discussion
To test the potential performance advantages offered by a translated alignment of protein-coding DNA sequences, six mammalian coding genes were each aligned either directly using ClustalW (default parameters) or via their amino-acid translations using transAlign (genetic code specified, otherwise default parameters). All alignments used ClustalW v1.83 on an 800-MHz dual-processor Macintosh G4 running OS 10.3.5. The qualities of the respective alignments were judged relative to a manual alignment of the same data set, each of which was completed for other purposes prior to transAlign being written. As such, the manual alignments represent reasonable, independent reference points. Quality was quantified by calculating the opposite of the Hamming distance (i.e., matching nucleotides score +1; mismatches score +0) between the same sequence in the test alignment and the manually produced one. These values were then averaged for each data set to essentially reveal how many nucleotides, on average, were correctly aligned.
The benchmark data (Table 1) show that transAlign indeed delivers alignments of often superior quality compared to a DNA alignment of the same data set, but always with a significant savings in time. In particular, the speedup was usually 7x or greater, and approximately the theoretical 9x for the three cases where a profile alignment was not performed. The only exception was for RBP3, where the many sequences that were identified as having possible frame shifts (61 of the 484 in the data set) necessitated an extensive DNA profile alignment. Even so, the overall speedup for this data set remained greater than 3x, in line with theoretical expectations based on the proportion of frame-shifted sequences (see Figure 1). In all cases, accuracy was either comparable to or significantly exceeded that of a DNA alignment. For MTCYB, the largest data set examined, the improvement in the alignment score was substantial (~2x), with the translated alignment requiring only 1.6 days as compared to over two weeks for the DNA sequence data.
It should be kept in mind that these benchmarks serve largely to point out the performance advantages inherent to performing a translated alignment. Other multiple-alignment programs that are faster than ClustalW do exist. But, the same advantages would also apply to these programs, such that alignments for the benchmark data sets could be obtained in even less time.
Conclusion
The principle underlying transAlign – that of aligning protein-coding DNA via its amino-acid translation – is not novel, having been suggested at least since the initial release of mrtrans (circa 1993). However, together with LAGAN, Multi-LAGAN, DIALIGN2, and the RevTrans server, transAlign represents one of the few complete implementations of the principle, with most of the remaining methods requiring the user to manually translate the DNA sequences and perform the amino-acid alignment. However, transAlign, in addition to being cross-platform, also includes a diverse suite of user-definable options relating to the processing of the DNA sequence data, its alignment as amino-acid data, and subsequent back-translation into aligned DNA data. In particular, transAlign uniquely offers different options to process sequences that do not translate into clean amino-acid sequences and, as such, may disrupt the alignment procedure. All these options mean that transAlign is well suited for the large-scale automated alignment of very large and/or very numerous data sets. As the benchmark studies show, the use of translated alignments provides alignments of at least comparable and often improved quality compared to a DNA alignment and always with a significant savings in time.
Availability and requirements
Project name: transAlign
Project home page: http://www.tierzucht.tum.de/Bininda-Emonds/ (under "Programs")
Operating system: Unix-based systems including OS X and Linux; DOS
Programming language: Perl; no additional modules required
Other requirements: ClustalW or, with suitable modifications to the source code, most other multiple-alignment programs
License: None; open-source
Any restrictions to use by non-academics: None
Acknowledgements
I thank Antonis Rokas and especially Bernhard Haubold and Alexis Stamatakis for initial discussion and encouragement. Several anonymous reviewers also provided helpful comments that improved the MS. Robin Beck generated the six data sets used in the benchmark test and helped with the manual alignment of several of them. These data sets are freely available on the download page for transAlign. This work was funded as part of the NGFN-funded project "Bioinformatics for the Functional Analysis of Mammalian Genomes" (BFAM).
Figures and Tables
Figure 1 Theoretical gain in speed from performing a translated alignment. The figure reveals there is always a performance advantage in aligning any given proportion of the protein-coding DNA sequences in a data set via their amino-acid translations with the remaining DNA sequences subsequently profile-aligned to them. The curve as shown is based on the assumption that the translated alignment is 9x faster, on average, than the respective DNA alignment; other values produce nearly identical curves of different scales.
Table 1 Benchmark data for the comparative performance of a translated alignment. Six mammalian protein-coding genes were aligned either as DNA (using ClustalW; default parameters) or via their translations as amino acids (using transAlign; genetic code specified, otherwise default parameters). All analyses used ClustalW v1.83 on an 800-MHz dual-processor Macintosh G4 running OS 10.3.5. The alignment score is taken relative to the corresponding sequence from a manually aligned data set and is the opposite of the Hamming distance (i.e., matching bases score +1, mismatches score +0). The alignment score was calculated for each individual sequence and then averaged over all sequences in each data set. Gene symbols follow the HUGO Gene Nomenclature Committee (HGNC; [21]).
Amino-acid alignment
DNA alignment Time (sec)
Data set No. of sequences Unaligned sequence length Alignment time (sec) Average alignment score Amino-acid alignment DNA profile alignment transAlign processing Total Average alignment score
BDNF 100 256-768 475 579.28 52 14 0 66 774.61
MTCYB 2484 388-1200 1216963 437.54 127309 13823 34 141166 860.75
RAG1 128 543-3141 2804 2346.46 307 n/a 3 310 2345.13
RAG2 196 326-1584 6492 1583.85 733 n/a 3 736 1583.95
RBP3 484 627-1292 45122 598.26 4004 10636 9 14649 579.71
VWF 182 711-1310 8384 862.06 921 n/a 4 925 1002.16
==== Refs
Haubold B Wiehe T Comparative genomics: methods and applications Naturwissenschaften 2004 91 405 421 15278216
Wernersson R Pedersen AG RevTrans: Multiple alignment of coding DNA from aligned amino acid sequences Nucleic Acids Res 2003 31 3537 3539 12824361 10.1093/nar/gkg609
Henikoff S Henikoff JG Amino acid substitution matrices from protein blocks Proc Natl Acad Sci USA 1992 89 10915 10919 1438297
Gonnet GH Cohen MA Benner SA Exhaustive matching of the entire protein sequence database Science 1992 256 1443 1445 1604319
Dayhoff MO Schwartz RM Orcutt BC Dayhoff MO A model of evolutionary change in proteins Atlas of Protein Sequence Structure 1978 5 Washington, D.C.: National Biomedical Research Foundation 345 352
Posada D Crandall KA Selecting the best-fit model of nucleotide substitution Syst Biol 2001 50 580 601 12116655 10.1080/106351501750435121
Smith TF Waterman MS Identification of common molecular subsequences J Mol Biol 1981 147 195 197 7265238 10.1016/0022-2836(81)90087-5
MRTRANS – CDNA alignment based on protein alignment
RevTrans Server
Morgenstern B DIALIGN 2: improvement of the segment-to-segment approach to multiple sequence alignment Bioinformatics 1999 15 211 218 10222408 10.1093/bioinformatics/15.3.211
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
Maddison DR Swofford DL Maddison WP NEXUS: an extensible file format for systematic information Syst Biol 1997 46 590 621 11975335
Felsenstein J PHYLIP (Phylogeny Inference Package), version 3.6 2004 Seattle: Department of Genome Sciences, University of Washington
Guindon S Gascuel O A simple, fast, and accurate algorithm to estimate large phylogenies by maximum likelihood Syst Biol 2003 52 696 704 14530136 10.1080/10635150390235520
Se-Al Homepage
Readseq Homepage
HMMER: sequence analysis using profile hidden Markov models
NCBI Taxonomy Homepage
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
Chenna R Sugawara H Koike T Lopez R Gibson TJ Higgins DG Thompson JD Multiple sequence alignment with the Clustal series of programs Nucleic Acids Res 2003 31 3497 3500 12824352 10.1093/nar/gkg500
Wain HM Lush M Ducluzeau F Povey S Genew: the human gene nomenclature database Nucleic Acids Res 2002 30 169 171 11752283 10.1093/nar/30.1.169
| 15969769 | PMC1175081 | CC BY | 2021-01-04 16:02:50 | no | BMC Bioinformatics. 2005 Jun 22; 6:156 | utf-8 | BMC Bioinformatics | 2,005 | 10.1186/1471-2105-6-156 | oa_comm |
==== Front
BMC BiotechnolBMC Biotechnology1472-6750BioMed Central London 1472-6750-5-171593509310.1186/1472-6750-5-17Research ArticleCpG-island fragments from the HNRPA2B1/CBX3 genomic locus reduce silencing and enhance transgene expression from the hCMV promoter/enhancer in mammalian cells Williams Steven [email protected] Tracey [email protected] Tony [email protected] Mark [email protected] David [email protected] Michael [email protected] Alistair [email protected] Andrew [email protected] Robert [email protected] ML Laboratories PLC-Research Division, MED IC4, Keele University Science and Business Park, Keele, Staffordshire, ST5 5SP, UK2 Nuclear Biology Group, Division of Medical and Molecular Genetics, GKT School of Medicine, Kings College London, Guy's Campus, London Bridge, London, SE1 9RT, UK2005 3 6 2005 5 17 17 22 12 2004 3 6 2005 Copyright © 2005 Williams 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 hCMV promoter is very commonly used for high level expression of transgenes in mammalian cells, but its utility is hindered by transcriptional silencing. Large genomic fragments incorporating the CpG island region of the HNRPA2B1 locus are resistant to transcriptional silencing.
Results
In this report we describe studies on the use of a novel series of vectors combining the HNRPA2B1 CpG island with the hCMV promoter for expression of transgenes in CHO-K1 cells. We show that the CpG island gives at least twenty-fold increases in the levels of EGFP and EPO observed in pools of transfectants, and that transgene expression levels remain high in such pools for more than 100 generations. These novel vectors also allow facile isolation of clonal CHO-K1 cell lines showing stable, high-level transgene expression.
Conclusion
Vectors incorporating the hnRPA2B1 CpG island give major benefits in transgene expression from the hCMV promoter, including substantial improvements in the level and stability of expression. The utility of these vectors for the improved production of recombinant proteins in CHO cells has been demonstrated.
==== Body
Background
Despite recent progress in elucidating the molecular basis of gene expression the practicality of achieving reliable, stable, high-level transgene expression in mammalian cells remains a major challenge. This inefficiency of transgene expression is largely attributable to transcriptional silencing, which typically involves methylation at CpG DNA sequences, histone deacetylation and chromatin condensation in the vicinity of the integration site [1,2]. When exogenous genes are introduced into cultured cells, many of the integration events lead to rapid transgene silencing, whilst the remainder give widely varying expression levels [3,4]. For this reason the use of cultured mammalian cells to produce recombinant proteins (except at very small scale) usually requires labour-intensive isolation of rare, clonal cell lines that stably express the transgene at high levels. Similarly, chromatin structure-related silencing and extreme variability in expression hinder the use of transgenic animals for purposes such as defining novel gene function and manufacture of therapeutic proteins [5,6]. Furthermore, achieving clinical utility with gene therapies (especially for chronic diseases), is greatly hindered by inadequate levels and duration of therapeutic transgene expression, resulting partly from transcriptional silencing [7,8]. The identification of elements capable of maintaining a transcriptionally competent ("open") chromatin domain resistant to silencing, irrespective of tissue type or integration site, is therefore an important objective in the development of technology giving more efficient transgene expression in mammalian cells, for many important applications.
We have reported previously that large DNA fragments containing CpG islands from the human TBP-PSMB1 and HNRPA2B1-CBX3 loci (designated TBP and RNP below) are resistant to heterochromatin-mediated silencing of these genes [9]. These regions are structurally similar, containing dual divergently transcribed promoters embedded within an extended methylation-free CpG island. The divergent promoters drive expression of ubiquitously expressed housekeeping genes, encoding the TATA binding protein and proteasome component-B1 in the case of the TBP locus, and the heterogeneous ribonucleoprotein A2/B1 and chromobox homolog 3 in the case of the RNP locus. Our previous studies showed that with these large CpG island-containing fragments, active transcription from the endogenous promoters was maintained even upon integration into centromeric heterochromatin, that single copy integration at different sites gave very similar levels of gene expression, and that position effect variegation was greatly reduced [9].
The endogenous promoters of the housekeeping genes at the TBP and RNP loci are relatively weak. In this paper we report studies with a series of novel expression vectors in which the CpG island from the RNP locus was combined with the immediate early promoter/enhancer of the human cytomegalovirus (hCMV), the promoter most commonly used to achieve strong, non-selective expression in mammalian cells for many applications. We show that the RNP CpG island confers major benefits in transgene expression from this promoter, including substantial improvements in the level of expression and proportion of transfected cells that express at detectable levels. We show that with these novel vectors expression remains high on prolonged subculturing. We also demonstrate the utility of these vectors for the improved production of recombinant proteins in Chinese Hamster Ovary (CHO) cells.
Results and Discussion
RNP CpG island fragments enhance transgene expression and confer resistance to silencing when linked to the hCMV promoter
Our previous report concerned the ability of large DNA fragments containing the RNP CpG island to inhibit transcriptional silencing of genes expressed from the promoters of the endogenous housekeeping genes [9]. With a view to developing vectors allowing facile generation of constructs for improved, high-level transgene expression, we first evaluated smaller fragments containing this CpG island. As shown in Figure 1A the HNRPA2B1 and CBX3 genes are divergently transcribed from two separate promoters embedded within a methylation-free CpG island approximately 3 kb in length [9]. We evaluated three CpG island-containing fragments of length 8.0 kb, 4.0 kb and 1.5 kb (Figure 1B). All three fragments contained the dual divergent promoters. The 8.0 kb and 4.0 kb fragments both contained the entire methylation-free CpG island, while the 1.5 kb fragment contained approximately half of this region. These fragments were incorporated into vectors for expression of EGFP from the hCMV promoter as indicated in Figure 1B.
Following stable transfection of these constructs into CHO-K1 cells, EGFP expression was quantified by FACScan analysis of transfectant pools after 33 and 200 generations of culture (Figure 2). After 33 generations pools derived with hCMV-EGFP showed a characteristically wide range of expression levels, manifest as a broad plateau rather than a discreet peak of EGFP expressing cells. Interestingly, 48% of the cells showed no detectable expression and only 6% of the cells showed expression exceeding 1000 fluorescence units at this stage of culture. By contrast, after 33 generations of culture for the 8.0 and 4.0 kb RNP-EGFP containing populations only 3% and 5% of the cells respectively showed no detectable expression, with 88% and 84% of the cells respectively showing median expression exceeding 1000 fluorescence units. The majority of cells in these populations showed very high expression levels, as evidenced by the steep shoulder at the end of the detectable range. Similarly, after 200 generations of culture these stably transfected cell populations contained 75% and 73% of cells with median fluorescence exceeding 1000 units. At the same stage of culture only 2% of cells in the pool transfected with hCMV-EGFP showed fluorescence exceeding 1000 units.
Figure 3 (A and B) shows the results (in histogram format) of similar experiments, conducted over 199 or 107 generations of culture, with vectors containing the 8.0, 4.0 and 1.5 kb CpG islands. The 1.5 kb RNP CpG island fragment conferred improvements in level of expression from the hCMV promoter and percentage of cells showing detectable EGFP expression that were very similar to those observed for the larger fragments. FACScan analysis also showed the characteristic steep shoulder associated with the larger RNP fragments (data not shown). Overall, the 8.0, 4.0 and 1.5 kb RNP fragments reproducibly gave 20- to 40-fold increases in the median expression levels of EGFP compared to that observed for the hCMV promoter alone, and these increased levels were maintained through at least 107 generations of continuous culture.
The effects of the CpG island fragments on EGFP expression level and stability were further studied using clonal CHO-K1 cell lines rather than transfected pools. Figure 4A shows the results of quantifying EGFP expression for clones isolated following stable transfection with the hCMV-EGFP and 8.0 kb RNP-EGFP vectors. Of more than a hundred clonal lines containing the hCMV-EGFP construct the majority (60 of 112) showed no detectable EGFP expression, with only a single clone displaying a median expression level that exceeded 5000 fluorescence units. In contrast, for clones generated with the 8.0 kb RNP-EGFP construct, the great majority (74 of 86) did show detectable EGFP expression and a large proportion (63 of 86) had a median expression level exceeding 5000 units. 41 of 86 clones with the 8.0 kb RNP-EGFP construct showed expression exceeding that of the best hCMV-EGFP clone. Of the 12 clones generated with this construct that showed no detectable EGFP expression, all those examined by Southern blot analysis proved to have deletions extending (at least) into the RNP sequences (data not shown). These deletions presumably occurred prior to or during the integration process. Similar analysis of non-expressing clones generated with the hCMV-EGFP construct revealed many with no detectable deletion. These results suggest that the 8.0 kb RNP fragment confers a substantial increase in the proportion of integration events that lead to detectable gene expression.
Ideally, for large-scale manufacture of protein therapeutics clonal cell lines producing the proteins are expanded without drug selection from a master cell bank. Where proteins are manufactured for use in humans the regulatory authorities require evidence that the manufacturing process gives production that is stable (i.e that falls within defined specification limits) over 25 to 30 generations of culture. To investigate the suitability of vectors containing the RNP CpG island fragments for this purpose, four clonal lines generated with the 8.0 kb RNP-EGFP construct and showing median expression that exceeded 5000 fluorescence units after 118 generations of culture in the presence of G418, were cultured with and without drug selection for a further 213 generations. For all four clones no reduction was observed in the percentage of cells showing expression exceeding 1000 fluorescence units after 213 generations of culture with drug selection. Also for all four clones no reduction was observed in the percentage of such cells after 38 generations without drug selection. Slight instability of expression was observed for one clone, #54, on very prolonged culture without drug selection. As shown in Figure 4B for this clone the percentage of cells showing expression exceeding 1000 units declined from 95% to 79% between 38 and 213 generations, while for the other three clones, exemplified by #67, no such decline was observed.
RNP CpG island vectors give improved yields of recombinant proteins in CHO cells
The utility of the RNP CpG island vectors for recombinant protein production was further examined using the erythropoietin (EPO) gene. Constructs were generated with the EPO cDNA sequence under the control of either the hCMV promoter alone or in combination with the 8.0 kb RNP fragment (Figure 1B). Following stable transfection into CHO-K1 cells EPO production was measured for pools of transfectants over 174 generations of culture (Figure 5A). The 8.0 kb RNP-EPO construct gave a substantially higher yield of EPO at all time points than that observed with hCMV-EPO. On average the RNP CpG island vector gave a twenty-fold improved yield of EPO, and the improved production was maintained through at least 100 generations of subculture.
Clonal CHO-K1 lines were derived from transfectants generated with the hCMV-EPO and 8.0 kb RNP-EPO expression constructs. Twenty clones were chosen at random and assessed for EPO production (Figure 5B). For clones derived with hCMV-EPO, 6/20 produced no detectable EPO and only 1/20 (C4) showed a yield (0.85 μg/ml) exceeding 0.5 μg/ml. In marked contrast, 20/20 clones harbouring 8.0 kb RNP-EPO showed detectable EPO production, 18/20 produced more than 0.5 μg/ml, 16/20 produced more than 1 μg/ml, and one clone (R16) produced almost 14 μg/ml. Thus inclusion of the 8.0 kb RNP fragment in the expression construct resulted in the facile isolation of clones with improved EPO production, the average yield for the twenty clones being twenty-one-fold higher than with hCMV-EPO, and the best producing clone with 8.0 kb RNP-EPO giving sixteen-fold higher productivity than C4, the highest producing hCMV-EPO line.
Increased expression with RNP CpG island vectors is not due to increased transgene copy number
The possibility that CpG island vectors give improved transgene expression through integration at increased copy number was evaluated by conducting copy number analysis on genomic DNA for clonal lines expressing EGFP following transfection with hCMV-EGFP (13 lines) or 8.0 kb RNP-EGFP (9 lines). All these lines proved to have copy numbers of two or three. The highest expressing clones with both constructs had a copy number of three, but so did some of the lowest expressing clones for both constructs. Overall, statistical analysis of variance (by the ANOVA single factor method) indicated no significant correlation between level of expression and copy number for either construct.
Conclusion
The hCMV promoter has been by far the most commonly used promoter for high level transgene expression in mammalian cells, for a wide range of applications that includes the production of protein therapeutics in cultured cells, transgenic animals and gene therapies. Its utility in expressing transgenes for all these applications, however, has been hindered by its susceptibility to silencing, largely through effects involving adverse chromatin structure. For example, the isolation of clonal cell lines or transgenic animals showing stable, high level expression of transgenes is usually a slow and labour-intensive procedure because most integration events lead to silencing.
The results we report here show that incorporating a RNP CpG island fragment immediately upstream of the hCMV promoter gives major benefits in expression from the latter. These benefits include a substantial increase in the median level of expression observed in pools of transfectants, together with a substantial improvement in the proportion of cells in the pool that express. The CpG island fragments reproducibly gave 20 to 40-fold increases in the level of expression observed. With both EGFP and EPO these dramatically increased expression levels were maintained through at least 100 generations of subculture. The CpG island fragments also enabled the rapid and facile isolation of clonal cell lines showing stable, high level expression of EGFP and EPO.
Our previous work [9] showed that a 16 kb DNA fragment containing the RNP CpG island conferred resistance to heterochromatin-mediated silencing of expression from the endogenous RNP promoter and reduced position effect variegation, giving very consistent expression levels in tissue culture cells. In the studies we report here, constructs with the 8.0 kb RNP CpG island fragment preceding the hCMV promoter gave detectable levels of transgene expression for all integration events in which the construct remained intact, together with substantial increases in the level of expression from the hCMV promoter. Analysis of variance (by the single factor ANOVA method) showed these increases to be statistically significant (p = 8.13E-27). These constructs, however, gave much greater variability in levels of expression than those observed in our previous work with a larger CpG island fragment and the endogenous RNP promoter. It is not clear whether the greater variability in expression level for the CpG island constructs in the studies we report here than our previous ones is due to the use of a different promoter or different cell line, or occurs because sequences giving more complete isolation from the effects of chromatin structure or transcriptional activity adjacent to the integration site are present on the 16 kb fragment but not on the 8 kb fragment.
The present results demonstrate two very important applications of this novel transgene expression technology. One is the rapid production of recombinant proteins, in the quantities required for basic research, drug discovery and preclinical studies, using stable pools of transfectants. The CpG island technology circumvents the need for slow, labour intensive screening of clonal lines for this purpose. The other is faster and easier identification of clonal cell lines that show stable, high level production as candidates for large scale manufacture of protein therapeutics. With the CpG island technology screening of only 20 clonal CHO cell lines was sufficient to identify lines showing very high yields of EPO, the latter being a blockbuster protein drug that is widely used to treat anaemia associated with renal failure and chemotherapy. It is noteworthy that the hCMV and CHO cells used in these experiments are the promoter and cells most commonly used for commercial manufacture of protein therapeutics.
Through their capacity to confer an increase in the proportion of integration events that are productive, together with improvements in the level of transgene expression, RNP CpG island vectors have many other potential applications, including use in mammalian cell-based in vitro screens for drug discovery, transgenic animals for basic research and drug discovery, and gene therapy. For some of these applications, notably those involving transgene delivery with integrating viral vectors, gene expression elements of small size are required. The data we present here suggest that the 1.5 kb RNP fragment confers major benefits in expression level from the hCMV promoter. Efforts to further characterise and minimise these elements are ongoing, but the 1.5 kb fragment should be small enough for incorporation into most viral vectors.
The mechanism by which the RNP CpG island reduces silencing and improves transgene expression is under investigation in our laboratories. The data we report here show that the mechanism does not involve increasing the copy number of the integrated transgene. Our results to date, reported previously [9] and in this paper, are consistent with these elements being able to establish and maintain a more open chromatin domain irrespective of the local chromosome environment. We propose that the promoter-containing CpG-islands of housekeeping genes possess a chromatin remodelling function and that this is designated a "Ubiquitously-acting Chromatin Opening Element", or "UCOE".
Methods
Expression vectors
hCMV-EGFP was vector pEGFP-N1 (Clontech, Cambridge, UK), which contains the hCMV promoter/enhancer on a 589 bp fragment. RNP CpG island-containing vectors were constructed by inserting genomic fragments from the RNP locus into the blunted Ase I site of hCMV-EGFP. These fragments were blunted versions of the 8.0 kb Hind III fragment, 4.0 kb BamH I-Hind III fragment and 1.5 kb Esp3I fragments to give 8.0 kb RNP-EGFP, 4.0 kb RNP-EGFP and 1.5 kb RNP-EGFP respectively (Figure 1B).
The erythropoietin (EPO) cDNA was isolated by PCR amplification from a Quick Clone Foetal Liver cDNA library (Clontech, Cambridge, UK). The resulting 705 bp product was subcloned using the TA-cloning vector pCR3.1 (Invitrogen, Paisley, UK) generating pCR-EPO. hCMV-EPO was constructed by subcloning an NheI-NotI fragment from pCR-EPO into the respective sites within the vector pEGFP-NI. 8.0 kb RNP-EPO was constructed by subcloning the blunt-ended 8.0 kb Hind III fragment into the blunted Ase I site of hCMV-EPO (Figure 1B).
Cell lines and transfections
CHO-K1 cells were grown in HAMS F12 (Invitrogen, Paisley, UK) plus 4500 mg/l L-ananyl-L-glutamine, 10 μg/ml each of penicillin and streptomycin, and 10% (v/v) heat inactivated foetal calf serum (FCS; Invitrogen, Paisley, UK). Transfection was carried out by electroporation using approximately 107 cells from 80% confluent cultures and a BioRad Gene Pulser II ™ set to deliver a single pulse of 975 μF at 250 V. Transfections used 1 μg of linearised hCMV plasmid and equivalent molar quantities for expression vectors of different size. Stably transfected cells were selected and maintained in growth medium containing 400 μg/ml geneticin sulphate (G418; Sigma, Poole, UK). Clonal cell lines were derived from stable transfected pools by standard limiting dilution techniques.
Quantification of transgene expression
Analysis of cells transfected with EGFP reporter constructs was with a Becton-Dickinson FACScan using the parental CHO-K1 cell line as a background, autofluorescence control. For EPO quantitation cells were seeded at 106 cells/well in 6-well plates and incubated in FCS-containing medium over 48 hours. Conditioned medium samples were assayed for EPO using a Quantikine IVD ELISA kit (R&D Systems Europe, Abingdon, UK)
Transgene copy number analysis
Copy number was determined by quantitative PCR (QPCR) using oligonucleotide primers and Beacon probes designed using Beacon Designer v2.0 software (Premier Biosoft International, CA, USA) specific to hCMV or EGFP sequences and to the endogenous housekeeping gene β-actin as a copy number control. Multiplex QPCR reactions containing optimised oligonucleotide and probe concentrations and ~500 ng of template DNA were performed on the MX4000™ multiplex PCR quantitative PCR system according to the manufacturers protocol (Stragene, La Jolla, CA), and subsequent data manipulations performed using the MX4000™ analysis software (Stratagene, La Jolla, CA). Copy number was assigned by comparison of the hCMV and/or EGFP: β-actin endpoint fluorescence ratio with that of a verified single copy integrant.
Authors' contributions
SW carried out vector construction, EGFP pool transfection and analysis and experimental design.
TM and TM participated in vector construction, CHO-KI transfection, EGFP pool and clonal analysis.
MG carried out EPO cloning, CHO-KI pool and clonal analysis.
DS carried out copy number analysis and manuscript preparation.
MA was the original investigator and participated in the conception of the study.
AI participated in coordination of the project and manuscript preparation.
AM and RC were involved in the conception, experimental design and manuscript preparation.
Acknowledgements
This work was partially funded under the UK Department of Trade and Industry LINK programme with support from the Medical Research Council and Biotechnology and Biological Sciences Research Council.
Figures and Tables
Figure 1 The human HNRPA2B1/CBX3 (RNP) locus and constructs used in EGFP and EPO expression studies. A. Organisation of the RNP genomic locus. The figure indicates the locations of the divergent promoters (arrows), the exons (black boxes) and the 8.0 kb Hind III fragment that includes the 3 kb unmethylated CpG island. B. EGFP and EPO expression constructs. All constructs carried the Neor gene preceded by the SV40 promoter for selection of stable transfectants of CHO-K1 cells.
Figure 2 FACScan analysis of EGFP expressing CHO-K1 cells. Following stable transfection with hCMV-EGFP, 8.0 kb RNP-EGFP, or 4.0 kb RNP-EGFP, pools of >100 transfectants were cultured continuously with drug selection for 200 generations, and FACScan analysis conducted after 33 and 200 generations. The results are plotted as green fluorescence on a logarithmic scale (FL1-H) versus number of events on a linear scale (counts). Figures for gate M1 are % cells with undetectable fluorescence; those for gate M2 are % cells showing greater than 1000 fluorescence units.
Figure 3 RNP CpG island fragments enhance the level and stability of EGFP expression in stably transfected CHO-K1 cells. A. Histograms representing the median fluorescence and percentage cells with detectable EGFP expression for CHO-K1 cells stably transfected with hCMV-EGFP, 8.0 kb RNP-EGFP or 4.0 kb RNP-EGFP. Following transfection pools of >100 transfectants were cultured continuously with drug selection for 107 generations and FACScan analysis conducted every few days. B. Histograms representing the median fluorescence and percentage cells with detectable EGFP expression for CHO-K1 cells stably transfected with hCMV-EGFP, 4.0 kb RNP-EGFP or 1.5 kb RNP-EGFP. Following transfection pools of >100 transfectants were cultured continuously with drug selection for 107 generations and FACScan analysis conducted every few days
Figure 4 The 8.0 kb RNP CpG island fragment gives a high proportion of clonal CHO-K1 cell lines showing high-level, stable expression of EGFP. Following stable transfection of CHO-K1 cells clonal lines were derived by limiting dilution. Median fluorescence levels were measured for each clone by FACScan analysis after 22 generations of culturing with drug selection. A. Histograms representing the median fluorescence levels for clones generated with hCMV-EGFP (n = 112) and 8.0 kb RNP-EGFP (n = 86). B. FACscan analysis of two clonal lines described in A. Clonal cell lines were subcultured continuously with or without drug selection for a further 213 generations, with analysis after 38 and 213 generations. Figures for gate M1 are % cells showing greater than 1000 fluorescence units.
Figure 5 The 8.0 kb RNP fragment increases the yield of EPO in stably transfected CHO-K1 cells. A EPO production from pools of CHO-K1 cells stably transfected with hCMV-EPO or 8.0 kb RNP-EPO. Following selection on G418 pools of transfectants were cultured continuously with drug selection for 174 generations. Assays for EPO production were conducted every few days. B EPO production from clonal CHO-K1 cell lines derived with hCMV-EPO or 8.0 kb RNP-EPO. Clonal lines were derived by limiting dilution from pools of transfectants selected on G418. Colonies were expanded for 35 generations with drug selection prior to assay for EPO production.
==== Refs
Razin A CpG methylation, chromatin structure and gene silencing – a three-way connection EMBO 1998 17 4905 4908 10.1093/emboj/17.17.4905
Bird AP Wolffe AP Methylation-induced repression – belts, braces, and chromatin Cell 1999 99 451 454 10589672 10.1016/S0092-8674(00)81532-9
Fusenegger M Bailey JE Hauser H Mueller P Genetic optimization of recombinant glycoprotein production by mammalian cells Trends Biotech 1999 17 35 42 10.1016/S0167-7799(98)01248-7
Doerfler W Schubbert R Heller H Kammer C Hilger-Eversheim K Knoblauch M Remus R Integration of foreign DNA and its consequences in mammalian systems Trends Biotech 1997 15 297 301 10.1016/S0167-7799(97)01061-5
Furth PA Hennighausen L Baker C Beatty B Woychick R The variability in activity of the universally expressed human cytomegalovirus immediate early gene 1 promoter/enhancer in transgenic mice Nucleic Acids Res 1991 19 6205 8 1956779
Houdebine LM Transgenic animal bioreactors Transgenic Res 2000 9 301 304 11131008 10.1023/A:1008934912555
Kuriyama S Sakamoto T Kikukawa M Nakatani T Toyokawa Y Tsujinoue H Ikenaka K Fukui H Tsujii T Expression of a retrovirally transduced gene under control of an internal housekeeping gene promoter does not persist due to methylation and is restored partially by 5-azacytidine treatment Gene Ther 1998 5 1299 1305 9930334 10.1038/sj.gt.3300738
Mountain A Gene therapy – the first decade Trends Biotech 2000 18 119 128 10.1016/S0167-7799(99)01416-X
Antoniou M Harland L Mustoe T Williams S Holdstock J Yague E Mulcahy T Griffiths M Edwards S Ioannou PA Mountain A Crombie R Transgenes encompassing dual-promoter CpG islands from the human TBP and HNRPA2B1 loci are resistant to heterochromatin-mediated silencing Genomics 2003 82 269 279 12906852 10.1016/S0888-7543(03)00107-1
| 15935093 | PMC1175082 | CC BY | 2021-01-04 16:02:57 | no | BMC Biotechnol. 2005 Jun 3; 5:17 | utf-8 | BMC Biotechnol | 2,005 | 10.1186/1472-6750-5-17 | oa_comm |
==== Front
BMC CancerBMC Cancer1471-2407BioMed Central London 1471-2407-5-681598969310.1186/1471-2407-5-68Research ArticleOverexpression of cathepsin f, matrix metalloproteinases 11 and 12 in cervical cancer Vazquez-Ortiz Guelaguetza [email protected] Patricia [email protected] Karla [email protected] Alfonso [email protected] Lucia [email protected] Patricia [email protected] José A [email protected] Mauricio [email protected] Oncogenomics Laboratory, Oncology Research Unit, Oncology Hospital, National Medical Center SXXI-IMSS, Mexico2 Division of Basic Research, INCAN, SS, Mexico3 Laboratory of Theoretical Biology, Research Department, La Salle University, Mexico2005 30 6 2005 5 68 68 16 3 2005 30 6 2005 Copyright © 2005 Vazquez-Ortiz 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
Cervical carcinoma (CC) is one of the most common cancers among women worldwide and the first cause of death among the Mexican female population. CC progression shows a continuum of neoplastic transitions until invasion. Matrix metalloproteinases (MMPs) and cathepsins play a central role on the enhancement of tumor-induced angiogenesis, cell migration, proliferation, apoptosis and connective tissue degradation. MMPs -2 and -9 expression has been widely studied in cervical cancer. Nevertheless, no other metalloproteinases or cathepsins have been yet related with the progression and/or invasion of this type of cancer.
Methods
Three HPV18 CC cell lines, two HPV16 CC cell lines and three HPV16 tumor CC tissues were compared with three morphologically normal, HPV negative, cervical specimens by cDNA arrays. Overexpression of selected genes was confirmed by end point semiquantitative reverse transcription-PCR with densitometry. In situ hybridization and protein expression of selected genes was further studied by means of two tissue microarrays, one consisting of 10 HSIL and 15 CC and the other one of 15 normal cervical and 10 LSIL tissues.
Results
TIMP1, Integrins alpha 1 and 4, cadherin 2 and 11, Cathepsins F, B L2, MMP 9, 10 11 and 12 were upregulated and Cathepsin S, L, H and C, Cadherins 3 and 4, TIMP3, MMP 13, Elastase 2 and Integrin beta 8 were found to be downregulated by cDNA arrays. Endpoint RT-PCR with densitometry gave consistent results with the cDNA array findings for all three genes selected for study (CTSF, MMP11 and MMP12). In situ hybridization of all three genes confirmed overexpression in all the HSIL and CC. Two of the selected proteins were detected in LSIL, HSIL and CC by immunohistochemistry.
Conclusion
Novel undetected CC promoting genes have been identified. Increased transcription of these genes may result in overexpression of proteins, such as CTSF, MMP11 and MMP12 which could contribute to the pathogenesis of CC.
==== Body
Background
Cervical carcinoma (CC) is one of the most common cancers among women worldwide and the first cause of death among the Mexican female population[1]. High-risk human papillomavirus (HPV) infection is considered the most important risk factor associated with the development of this tumor, and it is present in the 99.7% of invasive cervical tumors worldwide[2]. CC progression shows a continuum of neoplastic transitions from low grade squamous intraepithelial lesion (LSIL), High grade squamous intraepithelial lesion (HSIL), to invasive cervical cancer. Tumor invasion and metastasis are key steps in the progression of malignant tumors, which involves both extracellular matrix (ECM) and basement membrane degradation[3]. Although several classes of proteolytic enzymes are involved in these processes[4], Matrix metalloproteinases (MMPs) play a central role on the enhancement of tumor-induced angiogenesis, cell migration, proliferation, apoptosis and connective tissue degradation [5,6]. Different MMP family members have been identified, including collagenases (MMP-1, -8 and -13), gelatinases (MMP-2 and -9), stromelysins (MMP-3, -10, -11), membrane associated (MMP-14, -15, -16, -17, -23, -24, -25) and other types of MMPs, such as the Metalloeslastase (MMP-12)[7]. MMPs -2 and -9 have been widely studied. Increased mRNA and protein levels of both MMP-2 and MMP-9 have been detected in breast, colon, pancreatic and cervical cancers[8].
Human macrophage metalloelastase (MMP-12) has been identified in alveolar macrophages of cigarette smokers as an elastolytic MMP[9]. It can degrade elastin, and other substrates, such as type IV collagen, fibronectin, laminin, gelatin, vitronectin, entactin, heparin, and chondroitin sulphates[10]. MMP-12 is overexpressed by macrophages in atherosclerotic lesions[11] and in intestinal ulcerations[12]. It has been demonstrated that besides macrophages, transformed epithelial cells can express MMP-12 in skin and vulvar cancers; MMP12 expression levels correlate with epithelial dedifferentiation and histological aggressiveness [13,14].
Stromelysin 3 ST3 (MMP-11) is another protease that can modulate cancer progression by remodelling extracellular matrix. It cleaves α1-antitripsin and IGF-BP1[15]. Normal ST3 expression is present during embriogenesis and wound healing, and its expression in stressed epithelial cells is detected in the vicinity of fibroblasts[16]. Malignant epithelial cells depend on local stromal cells including fibroblastic, endothelial and inflammatory cells to develop primary and secondary tumors. ST3 expression is observed in the area which surrounds malignant epithelial tumour cells and sometimes in tumor cells of esophageal, oral, papillary thyroid, colorectal, skin and ovarian carcinomas [17-22]. Hence, ST3 gene expression appears to be associated with tumor progression[23].
Cathepsins are cysteine proteases from the papain family that are responsible for protein breakdown in lysosomes. The human Cathepsin family is composed by Cathepsins B, C, F, H, K, L, O, S V, W and X[24]. The recently described Cathepsin F has been demonstrated by Northern blot analysis, to be ubiquitously expressed in several tissues with a higher expression in skeletal muscle and testis. Cathepsin F transcripts were also found in several cancer cell lines, suggesting that this enzime could be involved in degradative processes during tumor progression[25].
In the present study we examined mRNA expression levels of 8,000 genes including MMPs and cathepsins by means of cDNA microarrays in CC samples and cell lines. Selected genes were confirmed to be up-regulated by endpoint RT-PCR and densitometry. Moreover, RNA and protein expression of the selected genes were further examined by in situ hybridization and IHC applied on Tissue Microarrays (TMAs)[26]. Several genes that were either up- or down-regulated in CC were detected, raising the possibility that these genes could play an active or a passive role in the malignant progression of CC.
Methods
Cell lines and tissue samples for cDNA arrays and RT-PCR
Hela (HPV18), SiHa (HPV16) and CaSki (HPV16) cell lines were grown in the recommended media. The cell lines designated CaLO (derived from Invasive squamous carcinoma Stage IIB, HPV18) and INBL (derived from Invasive Squamous carcinoma IV A, HPV18), were established at the FES Zaragoza-UNAM, Mexico [27]. Fresh biopsies of dysplastic lesions were taken during colposcopy from 20 patients at the National Center for Dysplasias, in Mexico City. Twenty two normal cervical tissues from pre-menopausal women who died from non-gynecological related causes were obtained within the first two hours post-mortem at the Pathology Department of the Hospital General de Mexico, S. S. All the described procedures were evaluated and approved by the local committee of ethics of the Mexican Institute of Social Security; we also obtained written informed consent from the patients. All tissue samples were divided in three sections, the central part was snapped frozen in liquid nitrogen and stored at -70°C until nucleic acid extraction, and the extremes from the biopsies were fixed overnight in 70% ethanol and paraffin embedded at the Department of Pathology, Oncology Hospital, National Medical Center SXXI, Mexico. Serial sections from the extremes of the biopsies stained by Haematoxilin/Eosin were inspected for representativity of the tissue. Carcinoma samples were considered as representative when at least 70% of all cells in the tissue section consisted of cancer cells. Normal tissue samples were considered as such when at least 80% of the sample consisted of microscopically ectocervical normal tissue and HPV infection absence. All the invasive samples were diagnosed as squamous lesions.
Tissue samples for ISH and IHC
Paraffin blocks of 35 patients: 10 HSIL, 15 CC, and 10 LSIL were randomly collected from the files of the National Center of Dysplasias in Mexico City. Fifteen normal cervical tissue samples (obtained post-mortem) were fixed in buffered formaldehyde and embedded in paraffin for preparation of the TMA blocks. The TMAs were constructed as previously described[28]. One of the TMAs contained 10 HSIL tissues and CC samples; the other was made with the 15 normal cervical tissues and 10 LSIL samples. The TMAs did not include the investigated samples by the cDNA arrays or by RT-PCR.
Nucleic acid isolation for probe preparation and RT-PCR
Both RNA and DNA were Isolated from all tissue samples which met the required criteria: squamous cell carcinoma, HPV type 16 or 18, at least 70% cancer cells in the biopsy, and from the cervical cancer cell lines using the Trizol Reagent (Gibco BRL, Life Technologies, Grand Island NY USA) according to manufacturer's conditions. Each DNA sample was subjected to HPV typing by polymerase chain reaction (PCR) using the MY09/MY11 consensus primers and direct sequencing. Potentially contaminating DNA was removed by RNAse-free DNAse I treatment (Ambion Inc. Austin Texas USA). The resulting RNA concentration was measured spectrophotometrically and the quality of both nucleic acids was confirmed in agarose gels.
cDNA probe preparation
cDNAs were synthesized according to Ambion Advantage System User Manual. These samples were: cell lines HeLa, SiHa, CaSki, CALO and INBL, three HPV16+ tumor samples (T07, T31, T64) and three HPV- normal samples (N03, N11, N22). Unfortunately no HPV18+ tumor samples were available for array hybridization due to insufficient RNA yield for probe preparation. The cDNA probes were obtained by RT-PCR from 500 ng RNA in the presence of (α-33P) dATP (3000 Ci/mmol, NEN). Briefly, RNAs were denatured at 70°C for 10 min and cDNAs were synthetized at 42°C by oligo-dT priming in a final volume of 30 μl. The labelled cDNAs were purified by spin column chromatography. The final incorporation of the radiolabel was approximately 1 × 108 cpm per reaction.
Hybridization and array analysis
ULTRArray Advantage System array blots (Ambion Inc. Austin Texas USA) containing 8400 genes were prehybridized at 68°C for at least 90 min before probe addition in hybridization buffer. Then, 1.5 × 107 cpm of each labeled cDNA was added to the buffer. Hybridization was performed at 60°C overnight in a rolling bottle. The arrays were washed twice with 2X SSC and 0.5 % SDS at 60°C for 30 min; followed by two stringent washes with 0.5X SSC, 0.5% SDS at the same temperature and for the same length of time. Finally, damp arrays were sealed in plastic wrap and exposed to imaging plates (BASMP 2040S; Fuji, Nakamura, Japan) for 24 hours, which were then scanned with a STORM 860 phosphorimager/fluorimager (Amersham Biosciences, Buckinghamshire UK) to obtain 16-bit images. ArrayVision software (Image Research Inc) was used for the analysis. The artifacts were eliminated and the intensity of each spot was analyzed after background substraction. Then, data was normalized and analyzed by using limma bioconductors and marray package [29,30] for all genes presented in the array and the means of the normal array (three normal samples) were obtained. The frequency of positive cases was obtained by comparing the mean normal array with each individual tumor or cell line array. The threshold frequency was set at 40% and the intensity ratio (tumor;normal) threshold values of 4 up-regulation and -1 for down-regulation present in all samples were used in an attempt to detect significant expression changes.
RT-PCR analysis
Estimation of the reliability of the cDNA results for the three selected genes was performed by endpoint RT-PCR with densitometry. These samples were previously described; four normal samples, 5 cell lines, 3 HPV16 and 2 HPV18 tumors; displaying a clear change in their expression in respect to the normal, independently of the HPV type present. Two-hundred ng of total RNA from all samples were reverse transcribed using the RT-PCR Access System (Promega, Madison, WI). Primers for the genes of interest were: Cathepsin F (CTSF) 396 bp, TM = 67°C, (5'-GTGCTGATCAGAGTGGCTGCTGC-3' and 5'-AGTTTCCYGGACATGGAGAGGGAC-3'); Matrix Metalloproteinase 12 (MMP12), 370 bp TM = 55°C (5'-TCACGAGATTGGCCATTCCTT-3' and 5'-TCTGGCTTCAATTTCATAAGC-3')[31]; and Matrix Metalloproteinase 11 (MMP11/ STMY3), 399 bp, TM = 66°C, (5'-CCATGGCAGTTGGTGCAGGAGCAG-3' and 5'-TGCAGTCATCTGGGCTGAGACTCA-3'); and β-actin: (5'-TGAAGTCTGACGTGGACATC-3' and GTTCGTTCCTCATACTGCTCA-3') 243 bp, TM = 55°C primers were designed using Integrated DNA technologies Biotools software . The RT-PCR conditions were: for the first strand synthesis of cDNA 48°C for 45 min and 94°C for 2 min to denature template; and for second strand synthesis and DNA amplification: 94°C for 30 sec, (specific TM°C for each set of primers) for 1 min, and 68°C for 2 min for a total of 24 cycles, followed by a single step at 68°C for 7 min. The products were visualized on 1.5% agarose gels stained with ethidium bromide, and signals were quantified by densitometry using MetaView analysing system (version 4.5 Universal Imaging Corp., USA). MMP11, MMP12 and CTS expression was standardized to β-actin expression assessed from the same cDNA in separate PCR reactions and run in parallel on separate gels. The standardized mean of each triplicate PCR was then expressed relative to the levels in β-actin cDNA. Statistical comparisons among groups were analyzed by a Kruskall Wallis-test. A P value of <0.05 was considered as statistically significant.
In situ hybridization (ISH)
Briefly, five-micron tissue sections were obtained from TMA paraffin blocks, deparaffinized and rehydrated in a graded ethanol series (100, 90, 70, and 30%), and transferred to PBS solution (137 mM NaCl, 2.7 mM KCl, 4.3 mM Na2HPO4, 1.4 mM KH2PO4) for 10 min. Tissues were treated with DNase solution (1 μg/ml) for 10 min at 37°C and washed three times in PBS. Endogenous peroxidase inactivation was carried out by incubating the samples in hydrogen peroxide in methanol for 3% hydrogen peroxide (H2O2) for 40 min. Sense and antisense probes for CTSF, MMP11 and 12 genes were generated by single-strand PCR using specific cDNA obtained from SiHa cells RNA as template and labeled with biotin-16-dUTP (Roche). Each tissue was covered with 50 μl hybridization cocktail and a coverslip. The hybridization cocktail consisted of 50% formamide, 10% dextran sulfate, 2X SSC (20X SSC: 3 M NaCl, 300 mM Na3C6H5O7), PBS, 2% SDS, 100 μg/ml sonicated salmon sperm and 50 ng of dUTP-biotin-labeled probe (sense or antisense). Both the tissue RNA and probe were denatured at 65°C for 10 min. After hybridization, the coverslips were soaked off in Tris-buffered saline with Tween buffer 1X (TBST 10X: 500 nmol/L TrisHCL, pH 7.6, 3 mol/L NaCl, 1% Tween 20), and the tissues were incubated at 55°C for 20 min in stringent wash solution. The horseradish peroxidase labelled antibody (SA-HRP) (GenPoint System from DAKO) was immediately applied on the tissues and incubated for 15 min in a humidifier chamber. The sections were washed in TBST 1X. Biotinyl tyramide was applied to the tissue sections for signal amplification for 15 min at room temperature, and washed in TBST. A second SA-HRP step was carried out and the color reaction was developed with 0.06% diaminobenzidine (DAB) in 3% H2O2. Finally, the slides were washed, haematoxylin/counterstained, dehydrated in graded ethanol and mounted. Negative controls for ISH were carried out with the sense probes or with a treatment with RNAse solution (100 mg/ml for 30 min at 37°C) prior ISH. Cells were scored as positive for MMP11, MMP12 and CTSF when they showed cytoplasmic expression under the light microscope. Only the neoplastic region of each tissue section was evaluated. To asses the stained cytoplasm, the slides were viewed at 40× magnification. The positivity of cells in each tissue section was estimated by the mean signal intensity, where: (0) positive staining present in at least half of the studied tissue, (1) positive staining of half of the tumor area, and (2) when all tissue had positive staining. The slides were evaluated blindly by three independent observers.
Immunohistochemical assays (IHC)
Briefly, the tissue sections were deparaffinized and rehydrated in a graded ethanol series. Endogenous peroxidase inactivation was carried out as described in ISH, and the slides were preincubated with DAKO Protein Block Serum free medium Cat X0909 (DAKO, Carpinteria, CA) for 30 min at 37°C to prevent non specific immunoreaction. Excess medium was decanted ant tissues were incubated with the primary antibody as follows: MMP 11 (Biomeda, Cat V10221, Lot 10441, 1:100), MMP12 (R&D Systems, Cat MAB917, Lot AGEO22051. 1:100) at 37°C for 30 min, after which, DAKO Envision System Peroxidase (DAKO) was applied and the slides were counterstained with hematoxylin. As for the In situ hybridization, only cells in the neoplastic region of each tumor were evaluated and only when staining for the two proteins were present in the cytoplasm. A rough scoring was done to quantitate the intensity of the staining by three authors (GVO, MSV and PPS). Levels of MMP11 and MMP12 expression in tissue sections were scored under the light microscope. Scores were obtained by estimating mean signal intensity (scale 0 to 2). Scoring was achieved as for ISH. After the examination of the slides by the three independent observers, a global agreement regarding the results was reached.
Results
Changes in cDNA expression in CC
Initial analysis of the three normal cervical tissues by limma and marray package showed that the expression profiles were similar. Thus, the data of the histologically normal tissue samples were pooled to generate a mean normal tissue array. Data from the cell lines and tissue arrays were compared with the mean normal tissue arrays. The frequency of increased or decreased gene expression changes was determined by comparing the average normal tissue array with each of the cell lines and tissue arrays. The cut-off values were 4 fold over the normal for up-regulated, and -1 fold for down-regulated or suppressed genes. Fourteenl genes were overexpressed (TIMP1, Integrins alpha 1 and 4, cadherin 2 and 11, Cathepsins F, B L2, MMP 9, 10 11 and 12), and 10 genes were down-regulated (Cathepsin S, L, H and C, Cadherins 3 and 4, TIMP3, MMP 13, Elastase 2 and Integrin beta 8, Table 1).
Verification of cDNA expression array results by RT-PCR
The gene expression profile findings in cDNA arrays were confirmed by endpoint RT-PCR with densitometry for selected genes (CTSF, MMP11, MMP12, and β-actin; Fig. 1). In general, these analyses showed significant results consistent with those obtained by cDNA arrays. All 8 samples investigated by means of cDNA arrays showed increased CTSF expression, and in line with this, the tumor band intensities in agarose gels, as compared with the control band intensity, were increased from an average of 1.2 to 7 fold in all tumors for CTSF in the RT-PCR analyses. All samples studied by cDNA microarray analysis had increased MMP11 expression, and had a mean of 6 fold increased expression by RT-PCR. The greatest difference between the two methods was found in MMP12 expression. All samples studied showed increased MMP12 expression by cDNA array analysis and RT-PCR. All comparisons were statistically significant.
ISH was performed on HSIL and CC cases. A strong focal signal (+++) was observed in all CC for the three genes with 90% and 100% of positive. The number of positive cells present in CC (+++) was higher than in HSIL (++). CTSF was expressed in 90%, MMP11 in 90%, and MMP12 in 90% of HSIL studied samples in the tissue array. ALL HSIL showed numerous positive cells for the three transcripts studied. (Fig. 2C, G and 2K; Table 2). In most HSIL MMP11, MMP12 and CTS staining was seen exclusively in the epithelium. The number of positive cells in HSIL was higher than in LSIL (++, +). All three genes were stained in 90% of LSIL (+) samples, while the signal was weaker or undetectable in histological normal cervical tissues (Table 2, Fig. 2). H&E staining at higher magnification in an adjacent tissue slice revealed that the expression of CTS, MMP11 and MMP12 was usually confined to the cytoplasm of a subset of epithelial tumor cells with morphological features of squamous differentiation. Almost all samples showed heterogeneous staining along the tumor, with a few cases showing a stronger reaction in basal layer cells. No staining was observed in the adjacent stroma.
By immunostaining, the MMP11 and MMP12 proteins were present only in the cytoplasm. There was a strong correlation between ISH and IHC positive reactions. All CC samples had a large number of positive cells (+++, ++) for both proteins, the number of positive cells was lower in HSIL (+++, ++). As seen with ISH, LSIL staining was lower than in HSIL (+) and none of the normal tissues stained for any of the studied proteins (Table 2, Figure 3). Immnunostaining was heterogeneous in CC and in HSIL, whereas in LSIL the staining was preferentially seen on the basal epithelial cells. No staining was observed in the adjacent stroma.
Discussion
To identify novel genes that could be associated with the development CC, we screened a cDNA-based gene expression array of more than 8000 genes. From those genes found to be significantly upregulated by gene expression analysis, we initially focused on 24 because of their possible involvement in invasiveness and metastasis, 14 of these genes were found to be expressed only in tumor tissue. The present report constitutes the detailed analysis of three such genes: MMP11 MMP12 and cathepsin F. Overexpression of CTSF, MMP11 and MMP12 in CC was confirmed by means of end point RT-PCR. All transcripts were detected in CC, suggesting that these genes are essentially active in the invasive processes of cervical cancer. It is important to note that by ISH these genes were found to be progressively expressed through different stages of cervical carcinogenesis, except for two cases of HSIL, one probable reason of this could be: a possible regression of the disease.
Many normal biological processes, including reproduction, fetal development and wound healing, are critically dependent on controlled degradation of extracellular matrix (ECM). However, excessive degradation of matrix components also occurs in pathologic tissue destruction such as cancer [32]. MMP11 expression in the immediate vicinity or within cancer cells has been associated with some human carcinomas and it is a consistently active partner of invading cancer cells. Its function differs throughout cancer progression, where in can be a tumor enhancer or a repressor in a number of processes leading to local or distal invasion and promoting cancer survival in the stromal environment [33]. We found expression of MMP11 in cervical precancerous lesions but not in histological normal cervix tissues. MMP11 expression was significantly higher in invasive carcinomas. Positive immunoreaction was detected in the cytoplasm of cervix epithelium tumor cells, differing from other tumor tissues, where MMP11 expression is restricted to stromal cells that surround the neoplastic area[34].
MMP11 expression has been reported in pre-invasive bronchial lesions and in carcinomas. MMP11 expression in bronchial lesions starts in dysplasia and carcinoma in situ, increasing in invasive lesions, as it appears to occur in CC. Expression of MMP11 in intraepithelial squamous lesions suggest that, as in squamous lung carcinomas, it could be related to progression of phenotypic alterations acquired early during the malignant transformation pathway of cervical epithelium and it is maintained after invasion[35]. It has been previously reported that when MMP11 is increased in tumorigenesis, this is not due to increased neoangiogenesis or cancer cell proliferation, but from a decrease of cancer cell death by apoptosis and necrosis. These data suggest that MMP11 could play different roles trough distinct pathways at different stages of cervical tumor development and progression. This needs to be further studied in CC.
The biological role of MMP12 in tumor progression is not clearly understood but it is thought to be involved in the degradation of components of the basement membrane[36]. Moreover, previous reports have shown that transformed epithelial cells in skin and vulvar cancers express MMP12 and such expression correlates with epithelial dedifferentiation and invasive aggressiveness. In oral verrucous and squamous cell cancer, the absence of MMP12 from epithelial cells has been reported to be a marker of good prognosis in non-invasive oral carcinoma[37].
MMP12 expression in cervical dysplasias and carcinomas has not been previously reported. Expression of MMP12 was detected in most LSIL, HSIL, and all CC samples, but it was absent in normal tissues, probably meaning that this protein is closely involved in early stages of transformation and invasion. By immunohistochemical assays, MMP11 and 12 proteins were detected in LSIL and HSIL with similar patterns as in ISH. Staining was mainly detected in basal cells, probably indicating their degradative and invasive capability.
Although only little data are available on CTSF, analysis of its expression in humans has revealed that it is present in most normal tissues, suggesting a general role for this enzyme in lysosomal function in all cell types. However, CTSF expression levels in normal tissues exhibit a wide variability depending on the tissue type. Several human cancer cell lines have increased expression of CTSF compared to its normal counterpart. This suggests that this enzyme could play an important role in carcinogenesis. Our results are in agreement with a previous study which revealed high levels of CTSF in HeLa cells[38], but it has not been described in cervical cancer. Due to a lack of availability of a specific CTSF antibody, we did not perform IHC assays for this protein. However, by means of ISH, CTSF expression pattern resembles that of MMP11 and 12.
Finally, two more genes found to be overexpressed in this study were MMP9 and TIMP1 (tissue inhibitor of metalloproteinase 1, Table 1). Interestingly, MMP9 and MMP2 up-regulation in cervical cancer was previously reported [33-43] and the combined overexpression of MMP9, MMP11 and TIMP1 has been associated with the invasive features of some cancers[44]. These proteins are clearly associated with CC and their overxpression in this work validates the approach employed.
Conclusion
In conclusion, our results show that CTSF, MMP11 and 12 are expressed in dysplastic and tumoral epithelial cells suggesting all of these proteins could be used as "potential" progression markers for cervical cancer. The understanding of the role, if any, of these proteins in the pathogenesis of CC deserves further studies.
List of abbreviations
HPV: Human Papilloma Virus, ISH: In Situ Hybridization, IHC: Immunohistochemistry, CTSF: Cathepsin F.
Competing interests
The author(s) declare that they have no competing interests.
Authors' contributions
GVO performed all molecular assays, designed the study, data collection and analysis, and drafted the manuscript.
PPS carried out the IHC and in the tissue block collection, participated in drafting the manuscript.
KV and PM were involved in fresh tissue collection and preparation of tissue blocks.
AD, LT, were involved in carrying out the tissue analysis and data acquisition.
JAG was involved in statistical analysis and writing and revising the manuscript critically.
MS conceived and designed the study, coordinated and managed the study, performed data analysis.
All authors read and approved the final manuscript.
Pre-publication history
The pre-publication history for this paper can be accessed here:
Acknowledgements
We are grateful to Dr. José Moreno, Dr. Diego Arenas (CMN SXXI-IMSS) and C. Barba MD for their helpful support. This work was partially supported by Grants No. 34686-M and 7114 (Fondos Sectoriales, CONACyT-México). During this work GVO, PP, KV and PM were recipients of a CONACyT and IMSS fellowship.
Figures and Tables
Figure 1 Relative mRNA expression levels of CTSF, MMP11 and MMP12 in cervical tumors detected by RT-PCR. One hundred of total RNA was employed for RT-PCR experiments. The used samples were; line a) positive control, line b) negative control, line 1) normal HPV- tissue #N07, 2) normal HPV- tissue #N11, 3) normal HPV- tissue #N20, 4) normal HPV- tissue #N23, 5) HeLa, 6) CaLO, 7) INBL, 8) CaSki, 9) SiHa, 10) tumoral sample #T04, 11) tumoral sample #T24, 12) tumoral sample #T43, 13) tumoral sample #T45, 14) Tumoral sample T63. A 100 bp ladder was used as molecular weight marker. Values in the graph are presented as ratio of densitometric readings of samples to corresponding beta actin samples.
Figure 2 In situ hybridization of cervical tissue microarrays with CTSF, MMP11 and MMP12 probes. A) Breast cancer (as positive control of CTSF), B) normal cervical epithelium negative for CTSF, C) HSIL with positive immunostaining, D) CC tissue positive for CTSF hybridization, E) Breast cancer (as positive control for MMP11), F) normal cervical epithelium, negative for MMP11 hybridization, G) as in C, HSIL positive for MMP11 hybridization, H) positive hybridization for MMP11 in CC, I) Breast cancer as positive control for MMP12, J) MMP12 negative hybridization in normal epithelium, K) as in C and G HSIL positive for MMP12, and L) Positive hybridization for MMP12 in CC. Original magnifications were 40 X for A, C, E, I and L; and 10X were used for B, D, F, G, H, J, and K.
Figure 3 Immunohistochemical detection of MMP11 and MMP12 by cervical tissue microarrays. A) Breast cancer sample (as positive control for MMP11 staining), B) negative control; a CC sample without primary antibody, C) one normal cervical tissue showing negative staining for MMP11, D) LSIL which presents positive immunostaining for MMP11, E) HSIL positive staining for MMP11 F) a CC showing positive immunoreaction for MMP11, G) Positive control for MMP12, H) as in B, negative control for MMP12, I) normal cervical tissue shows negative immunostaining for MMP12, J) LSIL showing positive immunoreaction for MMP12, K) HSIL showing immunodetection for MMP12 and L) a CC showing positive immunoreaction for MMP12. Original magnifications were 40 X for A, C, D, E, F, G, J, and K; and 10X were used for B, H, I, and L.
Table 1 Summary of altered gene expression in cervical cancer by cDNA array. Ratio (T/N), intensity ratio of the corresponding signals between tumor and normal cervical tissue.
Gene Name Unigene Locus Hela/n CaLO/n INBL/n SiHa/n CaSki/n T07/n T31/n T64/n
tissue inhibitor of metalloproteinase 1 Hs.446641 Xp11.3 7 4 4 3 4 4 4 4
integrin, alpha 9 Hs.222 3p21.3 7 6 3 4 3 4 4 3
cadherin 2, type 1, N-cadherin Hs.334131 18q11.2 6 4 4 5 4 5 4 6
cathepsin L2 Hs.87417 9q22.2 9 5 4 4 4 4 4 4
matrix metalloproteinase 12 Hs.1695 11q22.3 5 9 8 9 6 5 4 5
integrin, alpha 4 Hs.145140 2q31 8 3 5 4 3 4 5 4
serine B member 9 Hs.104879 6p25.2 5 4 4 5 4 3 4 4
serine, proteinase inhibitor, clade B member 3 Hs.227948 18q21.3 6 4 4 5 4 4 4 3
cathepsin B Hs.297939 8p22 6 4 3 5 5 4 4 4
cathepsin F Hs.11590 11q13 8 8 6 7 6 5 5 5
matrix metalloproteinase 10 Hs.2258 11q22 7 5 5 3 4 4 4 5
matrix metalloproteinase 11 (stromelysin 3) Hs.143751 22q11.23 6 8 6 6 6 5 5 5
matrix metalloproteinase 9 Hs.151738 20q13.12 5 6 5 4 4 5 4 3
cadherin 11, type 2 Hs.443435 16q22.1 6 3 3 3 4 5 5 4
cathepsin S Hs.181301 1q21 0 1 1 0 1 1 0 1
cathepsin L Hs.418123 9q21.33 0 0 1 1 1 0 0 1
cathepsin H Hs.114931 15q25.1 0 1 0 1 0 0 1 0
cathepsin C Hs.128065 11q14.2 1 0 1 0 1 1 1 0
matrix metalloproteinase 13 Hs.2936 11q22 0 0 0 1 0 1 0 0
tissue inhibitor of metalloproteinase 3 Hs.245188 22q12.3 0 1 0 1 1 1 1 0
cadherin 3, type 1, P-cadherin (placental) Hs.191842 16q22.1 0 1 1 0 0 1 0 1
cadherin 4, type 1, R-cadherin (retinal) Hs.376792 20q13.3 1 0 1 0 1 1 0 1
integrin, beta 8 Hs.355722 7p21.1 0 0 0 1 1 0 0 1
elastase 2, neutrophil Hs.99863 19p13.3 0 1 0 0 1 0 0 0
Table 2 Summary of the In situ hybridizations and immunohistochemical stainings results of cervical tissue microarrays.
% Positivity in ISH
CTSF MMP11 MMP12
Total of cases - (+) (++) (+++) - (+) (++) (+++) - (+) (++) (+++)
Normal tissues 15 14 1 0 0 13 2 0 0 13 2 0 0
LSIL 10 1 9 0 0 1 9 0 0 1 9 0 0
HSIL 10 1 1 8 0 1 2 7 0 1 2 7 0
CC tissues 15 0 0 0 15 0 0 0 15 0 0 0 15
% Positivity in IHC
MMP11 MMP12
Total of cases - (+) (++) (+++) - (+) (++) (+++)
Normal tissues 15 15 0 0 0 15 0 0 0
LSIL 10 0 10 0 0 0 9 1 0
HSIL 10 0 2 8 0 0 0 8 0
CC tissues 15 0 0 5 10 0 0 1 14
==== Refs
Mexican Ministry of Health Registro Histopatológico de Neoplasias Malignas, Compendio de mortalidad y morbilidad Secretaría de Salud, México 1998
Walboomers J Jacobs M Manos M Bosch X Kummer A Shah K Snijders P Peto J Meijer C Muñoz N Human Papilloma Virus is a necessary cause of invasive cervical cancer worldwide J Pathol 1999 189 12 19 10451482 10.1002/(SICI)1096-9896(199909)189:1<12::AID-PATH431>3.0.CO;2-F
Liotta LA Kohn EC The microenvironment of the tumor-host interface Nature (Lond) 2001 411 375 379 10.1038/35077241
Egeblad M Werb Z New functions for the matrix metalloproteinases in cancer progression Nature Rev 2002 2 163 176 10.1038/nrc745
McCawley LJ Matrisian LM Matrix metalloproteinases: multifunctional contributors to tumor progression: a review Mol Med Today 2000 6 149 156 10740253 10.1016/S1357-4310(00)01686-5
Nagase H Woessner JF Jr Matrix metalloproteinases: a review J Biol Chem 1999 274 491 21 10.1074/jbc.274.31.21491
McCawley LJ Matrisian LM Matrix metalloproteinases:they're not just for matrix anymore! Curr Opin Cell Biol 2001 13 534 540 11544020 10.1016/S0955-0674(00)00248-9
Van Trappen PO Ryan A Carroll M Lecoeur C Goff L Gyselman VG Young BD Lowe DG Pepper MS Shepherd JH Jacobs IJ A model for co-expression pattern analysis of genes implicated in angiogenesis and tumour cell invasion in cervical cancer Br J Cancer 2002 87 537 544 12189553 10.1038/sj.bjc.6600471
Shapiro SD Kobayashi DK Ley TJ Cloning and characterization of a unique elastolytic metalloproteinase produced by human alveolar macrophages J Biol Chem 1993 268 824 23
Gronski TJ JrMartin RL Kobayashi DK Walsh BC Holman MC Huber M Van Wart HE Shapiro SD Hydrolysis of a broad spectrum of extracellular matrix proteins by human macrophage metalloelastase J Biol Chem 1997 272 12189 12194 9115292 10.1074/jbc.272.18.12189
Curci JA Liao S Huffman MD Shapiro SD Thompson RW Expression and localization of macrophage elastase (matrix metalloproteinase-12) in abdominal aortic aneurysms J Clin Invest 1999 102 1900 1910
Vaalamo M Karjalainen-Lindsberg ML Puolakkainen P Kere J Saarialho-Kere U Distinct expression profiles of stromelysin-2 (MMP-10), collagenase-3 (MMP-13), macrophage metalloelastase (MMP-12), and tissue inhibitor of metalloproteinases-3 (TIMP-3) in intestinal ulcerations Am J Pathol 1998 152 1005 1014 9546361
Kerkela E Ala-Aho R Jeskanen L Rechardt O Grenman R Shapiro SD Kahari VM Saarialho-Kere U Expression of human macrophage metalloelastase (MMP-12) by tumor cells in skin cancer J Invest Dermatol 2000 114 1113 1119 10844553 10.1046/j.1523-1747.2000.00993.x
Kerkela E Ala-aho R Klemi P Grenman S Shapiro SD Kahari VM Saarialho-Kere U Metalloelastase (MMP-12) expression by tumour cells in squamous cell carcinoma of the vulva correlates with invasiveness, while that by macrophages predicts better outcome J Pathol 2002 198 258 269 12237887 10.1002/path.1198
Noel A Boulay A Kebers F Kannan R Hajitou A Calberg-Bacq CM Basset P Rio MC Foidart JM Demonstration in vivo that stromelysin-3 functions through its proteolytic activity Oncogene 2000 19 1605 1612 10734321 10.1038/sj.onc.1203465
Boulay A Masson R Chenard MP El Fahime M Cassard L Bellocq JP Sautes-Fridman C Basset P Rio MC High cancer cell death in syngeneic tumors developed in host mice deficient for the stromelysin-3 matrix metalloproteinase Cancer Res 2001 61 2189 2193 11280785
Yamashita K Tanaka Y Mimori K Inoue H Mori M Differential expression of MMP and uPA systems and prognostic relevance of their expression in esophageal squamous cell carcinoma Int J Cancer 2004 110 201 207 15069682 10.1002/ijc.20067
Soni S Mathur M Shukla NK Deo SV Ralhan R Stromelysin-3 expression is an early event in human oral tumorigenesis Int J Cancer 2003 107 309 316 12949813 10.1002/ijc.11366
Wasenius VM Hemmer S Kettunen E Knuutila S Franssila K Joensuu H Hepatocyte growth factor receptor, matrix metalloproteinase-11, tissue inhibitor of metalloproteinase-1, and fibronectin are up-regulated in papillary thyroid carcinoma: a cDNA and tissue microarray study Clin Cancer Res 2003 9 68 75 12538453
Wlodarczyk J Stolte M Mueller J E-cadherin, beta-catenin and stromelysin-3 expression in de novo carcinoma of the colorectum Pol J Pathol 2001 52 119 124 11769398
Thewes M Worret WI Engst R Ring J Stromelysin-3 (ST-3): immunohistochemical characterization of the matrix metalloproteinase (MMP)-11 in benign and malignant skin tumours and other skin disorders Clin Exp Dermatol 1999 24 122 126 10233668 10.1046/j.1365-2230.1999.00431.x
Mueller J Brebeck B Schmalfeldt B Kuhn W Graeff H Hofler H tromelysin-3 expression in invasive ovarian carcinomas and tumours of low malignant potential Virchows Arch 2000 437 618 624 11193473 10.1007/s004280000261
Basset P Bellocq JP Lefebre O Noel A Chenard MP Wolf C Anglard P Rio MC Stromelysin-3: a paradigm for stroma-derived factors implicated in carcinoma progression Crit Rev Oncol Hematol 1997 26 43 53 9246540
Wex T Levy B Wex H Bromme D Human cathepsins W and F form a new subgroup of cathepsins that is evolutionary separated from the cathepsin B- and L-like cysteine proteases Adv Exp Med Biol 2000 477 271 280 10849754
Santamaria I Velasco G Pendas AM Paz A Lopez-Otin C Molecular cloning and structural and functional characterization of human cathepsin F, a new cysteine proteinase of the papain family with a long propeptide domain J Biol Chem 1999 274 13800 13809 10318784 10.1074/jbc.274.20.13800
Cross SS Hamdy FC Deloulme JC Rehman I Expression of S100 proteins in normal human tissues and common cancers using tissue microarrays: S100A6, S100A8, S100A9 and S100A11 are all overexpressed in common cancers Histopathology 2005 46 256 269 15720411 10.1111/j.1365-2559.2005.02097.x
Caceres-Cortes JR Alvarado-Moreno JA Waga K Rangel-Corona R Monroy-Garcia A Rocha-Zavaleta L Urdiales-Ramos J Weiss-Steider B Haman A Hugo P Brousseau R Hoang T Implication of tyrosine kinase receptor and steel factor in cell density-dependent growth in cervical cancers and leukemias Cancer Res 2001 61 6281 6289 11507083
Hidalgo A Piña P Guerrero G Salcedo M A simple method for the construction of small format tissue array J Clin Pathol 2003 56 144 146 12560397 10.1136/jcp.56.2.144
Smyth GK Gentleman R, Carey V, Dudoit S, Irizarry R, Huber W Limma: linear models for microarray data Bioinformatics and Computational Biology Solutions using R and Bioconductor, Chapter 23 Springer, New York
Smyth GK Linear models and empirical Bayes methods for assessing differential expression in microarray experiments Statistical Applications in Genetics and Molecular Biology Smyth, G K 2004 3 33
Hoon N Hong K Hui S Kang S Young K Ho S MMP expression profiling in recurred stage IB lung cancer Oncogene 2004 23 845 851 14647437 10.1038/sj.onc.1207505
Boulay A Masson R Chenard MP El Fahime M Cassard L Bellocq JP Sautes-Fridman C Basset P Rio MC High cancer cell death in syngeneic tumors developed in host mice deficient for the stromelysin-3 matrix metalloproteinase Cancer Res 2001 61 2189 2193 11280785
Basset P Bellocq J Wolf C Stoll I Hutin P Limacher J Podhajcer O Chernard M Rio M Chambon P A novel metalloproteinase gene specifically expressed in stromal cells of breast carcinomas Nature 1990 348 699 703 1701851 10.1038/348699a0
Rouyer N Wolf C Chenard MP Rio M Chabon P Belloq J Basset P Stromelysin-3 gene expression in human cancer: an overview Invasion Metastasis 1994 95 41269 41275
Basset P Bellocq J Lefebvre O Noel A Chenard M Wolf C Anglard P Rio m Stromelysin 3: a paradigm for stroma derived factors implicated in carcinoma progression Crit Rev Oncol Hematol 1997 26 43 53 9246540
Shipley M Wesselchmidt R Kobashi K Ley T Shapiro S Metalloelastase is required for macrophage mediated proteolysis and matrix invasion in mice Proc Natl Acd Sci USA 1996 93 3942 3946 10.1073/pnas.93.9.3942
Impola U Uitto VJ Hietanen J Hakkinen K Zhang L Larjava H Isaka K Saarialho-Kere U V Differential expression of matrilysin-1 (MMP-7), 92 kD gelatinase (MMP-9), and metalloelastase (MMP-12) in oral verrucous and squamous cell cancer J Pathol 2004 202 14 22 14694517 10.1002/path.1479
Chapman H Riese RJ Guo-Ping S Emerging roles for cysteine proteases in human biology Ann Rev Physiol 1997 59 63 88 9074757 10.1146/annurev.physiol.59.1.63
Van Trappen P Ryan A Carroll M Lecoeur C Goff L Gyselman V Young BD Lowe DG A model of co-expression pattern analysis of genes implicated in angiogenesis and tumor cell invasion in cervical cancer Brit J Can 2002 87 537 544 10.1038/sj.bjc.6600471
Arguello-Ramirez J Perez-Cardenas E Delgado-Chavez R Solorza-Luna G Villa-Trevino S Arenas-Huertero F Matrix metalloproteinases-2, -3, and -9 secreted by explants of benign and malignant lesions of the uterine cervix Int J Gynecol Cancer 2004 14 333 40 15086734 10.1111/j.1048-891x.2004.014218.x
Kato Y Yamashita T Ishikawa M Relationship between expression of matrix metalloproteinase-2 and matrix metalloproteinase-9 and invasion ability of cervical cancer cells Oncol Rep 2002 9 565 9 11956628
Davidson B Goldberg I Kopolovic J Lerner-Geva L Gotlieb WH Weis B Ben-Baruch G Reich R Expression of matrix metalloproteinase-9 in squamous cell carcinoma of the uterine cervix-clinicopathologic study using immunohistochemistry and mRNA in situ hybridization Gynecol Oncol 1999 72 380 6 10053110 10.1006/gyno.1998.5285
Davidson B Goldberg I Liokumovich P Kopolovic J Gotlieb WH Lerner-Geva L Reder I Ben-Baruch G Reich R Expression of metalloproteinases and their inhibitors in adenocarcinoma of the uterine cervix Int J Gynecol Pathol 1998 17 295 301 9785129
Veli-Matti Wasenius Hemmer S Kettunen E Sakari Knuutila Fanssila K Heikki J Hepatocyte Growth Factor Receptor, Matrix Metalloproteinase-11, Tissue Inhibitor of Metalloproteinase-1, and Fibronectin Are Up-Regulated in Papillary Thyroid Carcinoma: A cDNA and Tissue Microarray Study Clin Can Res 2003 9 68 75
| 15989693 | PMC1175083 | CC BY | 2021-01-04 16:03:03 | no | BMC Cancer. 2005 Jun 30; 5:68 | utf-8 | BMC Cancer | 2,005 | 10.1186/1471-2407-5-68 | oa_comm |
==== Front
BMC Evol BiolBMC Evolutionary Biology1471-2148BioMed Central London 1471-2148-5-341592705710.1186/1471-2148-5-34Research ArticleAutomatic selection of representative proteins for bacterial phylogeny Bern Marshall [email protected] David [email protected] Palo Alto Research Center, 3333 Coyote Hill Road, Palo Alto, CA 94304, USA2005 31 5 2005 5 34 34 2 11 2004 31 5 2005 Copyright © 2005 Bern and Goldberg; licensee BioMed Central Ltd.2005Bern and Goldberg; licensee BioMed Central Ltd.This is an Open Access article distributed under the 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 there are now about 200 complete bacterial genomes in GenBank, deep bacterial phylogeny remains a difficult problem, due to confounding horizontal gene transfers and other phylogenetic "noise". Previous methods have relied primarily upon biological intuition or manual curation for choosing genomic sequences unlikely to be horizontally transferred, and have given inconsistent phylogenies with poor bootstrap confidence.
Results
We describe an algorithm that automatically picks "representative" protein families from entire genomes for use as phylogenetic characters. A representative protein family is one that, taken alone, gives an organismal distance matrix in good agreement with a distance matrix computed from all sufficiently conserved proteins. We then use maximum-likelihood methods to compute phylogenetic trees from a concatenation of representative sequences. We validate the use of representative proteins on a number of small phylogenetic questions with accepted answers. We then use our methodology to compute a robust and well-resolved phylogenetic tree for a diverse set of sequenced bacteria. The tree agrees closely with a recently published tree computed using manually curated proteins, and supports two proposed high-level clades: one containing Actinobacteria, Deinococcus, and Cyanobacteria ("Terrabacteria"), and another containing Planctomycetes and Chlamydiales.
Conclusion
Representative proteins provide an effective solution to the problem of selecting phylogenetic characters.
==== Body
Background
In molecular phylogeny, a great deal of attention has gone to computational methods for building phylogenetic trees [1], but much less to methods for selecting phylogenetic characters. Most sequence-based studies of prokaryotic and universal phylogeny have used either small-subunit rRNA genes [2-4] or highly conserved proteins such as ribosomal proteins, elongation factors, chaperones, and tRNA synthetases [4-6], arguing that these core sequences are unlikely to be horizontally transferred and hence should reflect vertical descent. Another sequence-based method relies upon the presence/absence of hand-picked "signature sequences" (conserved insertions/deletions) [7,8] to infer descent. This method does not specifically handle horizontal transfers, but can sometimes resolve a short internal branch that cannot be unambiguously resolved by a continuous evolutionary model. Early phylogenetic studies [2,3,8], limited by availability, were forced to use manually selected characters, but with the recent proliferation of full bacterial genomes, this restriction no longer applies, as all genes and genome data have now become potential characters.
Whole-genome phylogeny has the potential to discern vertical descent even in the case of widespread horizontal transfer. Many recent attempts at whole-genome phylogeny have used automatically computed characters other than sequence, such as gene order [6], dinucleotide frequencies [9], presence/absence of orthologous pairs [10], and presence/absence of gene families [11,12]. These non-sequence studies have given bacterial phylogenies with substantial areas of disagreement, and indeed it has been found that for deep prokaryotic phylogeny, sequence generally carries a stronger signal than dinucleotide frequencies [9], gene order [13], or number of common orthologs [6]. Hence there arises a need for a phylogenetic methodology that combines the power of sequence-based approaches with the objectivity and completeness of whole-genome approaches.
Researchers have responded to this need with a number of approaches that seek a predominant set of "concordant" [14] genes compatible with the same phylogeny. The approaches vary in whether they use gene trees or distance matrices to evaluate the genes, and also in their levels of automation and completeness. The use of gene trees is more common than the use of distance matrices. Brochier et al. [4] start with 57 translational apparatus proteins, ubiquitous over the set of organisms under study. They automatically screen these genes to obtain 44 concordant genes by principal component analysis of vectors of likelihoods for 375 test tree topologies; they associate the first principal component with gene length and the second with "incongruence". Along similar lines, Zhaxybayeva and Gogarten [15,16] evaluate genes by probability mapping of small trees such as four-taxon "quartets". Battistuzzi et al. [17] manually curated 60 ubiquitous COGs [18] down to 32 by rejecting all those that gave unstable gene trees, gene trees with either archaebacteria or eubacteria non-monophyletic, or "deep nesting" of a species from one phylum within another phylum. Daubin et al. [19] cluster the topologies of 310 computed gene trees in order to find a concordant set of 121, and then combine the concordant trees using the "supertree" approach [20-22], which can accommodate the missing data resulting from non-ubiquitous genes.
The use of organismal distance matrices to evaluate genes is less common, but as we argue below, it has a number of advantages over the use of gene trees. Clarke et al. [14] and Gophna et al. [23] compute distance matrices from reciprocal-best BLAST scores on a large number of genes. They measure the concordance of genes by correlation with the median distance matrix. They then use the consensus distance matrix of the concordant genes [14] or a weighted combination of the gene distance matrices [23] to compute the tree directly, rather than employing a more principled, model-based method such as maximum likelihood [1,24,25]. Novichkov et al. [26] improve upon the distance matrices of Clarke et al. by using a linear measure of evolutionary distance rather than BLAST scores and by assigning rate parameters to correct for the differing mutability of genes. They do not, however, go on to compute phylogenies based on concordant genes.
In this paper, we introduce and validate a fully automatic, whole-genome methodology based on "representative sequences". We start by assembling all of the highly conserved proteins (families of orthologs) within the set of genomes. We then use these genes to compute a consensus distance matrix by an algorithm similar to that of Novichkov et al. [26]. We use the consensus distance matrix to select representative sequences, but not to build the phylogenetic tree directly. Representative sequences are contiguous subsequences – typically 300 residues – from ubiquitous, conserved proteins, such that each orthologous family of representative sequences taken alone gives a distance matrix in close agreement with the consensus matrix. The phylogenetic tree is then computed using maximum-likelihood methods on a multiple alignment of a concatenation of representative sequences. We validate the methodology on a set of small phylogenetic problems with accepted answers, before going on to compute automatically a phylogeny for all of Bacteria. Our overall bacterial phylogeny shows striking agreement with the tree produced by Battistuzzi et al. [17] using genes that were manually curated for agreement with accepted clades.
We chose to use distance matrices rather than gene trees for the evaluation of gene concordance, because trees are brittle: a small change in sequence can dramatically change tree probabilities or the topology of the likeliest tree. (Even in continuous tree-space [27], small changes in input can force large changes in trees if the sequence data is not consistent with any one tree.) Moreover, distance matrices can more easily incorporate missing data: each sufficiently conserved gene, ubiquitous or not, can contribute to the pairwise distances between the organisms containing that gene. Combining trees on different sets of organisms is not as straightforward; indeed supertree methods typically combine matrix encodings [20-22]. Our method is conservative in its use of non-ubiquitous genes, including only those well-conserved proteins appearing in at least three-fourths of the taxa under study, but even at this level it can typically use more than three times as many genes as are completely ubiquitous over the taxa. Finally, anomalous pairwise distances directly locate likely horizontal gene transfers and other discontinuous evolutionary events (such as large insertions or deletions, loss of a domain, hidden paralogy, or rapid evolution due to change of function), and indeed our method rediscovers a number of previously proposed horizontal transfers.
Results
Our computational experiments produced: (1) an evaluation of the methodology of representative proteins, (2) findings concerning which proteins make the best phylogenetic characters, and (3) a fairly complete and well-resolved phylogeny for sequenced Bacteria.
Methodology
We evaluated our methodology on a test set of 10 deep phylogeny problems with known answers, comparing a maximum-likelihood tree-building method using representative sequences with the same method using an identical amount of randomly chosen, highly conserved, ubiquitous protein sequence. The same data-mining program compiled the families of orthologs in each case, but the randomly chosen proteins were picked without regard to their "representativeness", and hence are equivalent to small subsamples from all highly conserved ortholog families. We used small subsamples, because concatenations of 20 or more highly conserved genes almost invariably give correct trees for our relatively easy test set problems. As seen in Table 1 (also see Table 3), representative proteins generally outscore randomly chosen proteins in number of correct single-gene trees, number of accepted clades found over all gene trees, and number of correct trees on concatenations of genes. The consensus gene tree (that is, consensus over all single-gene trees) made with representative proteins succeeded (included all accepted clades) on all problems except 4 and 9, and a consensus (over bootstraps) concatenated tree succeeded on all problems except 4. Randomly chosen proteins succeeded less often. The consensus gene tree made with randomly chosen proteins failed on problems 1, 2, 7, and 9, and the consensus concatenated tree failed on problems 1, 2, 7, 9, and 10. The one problem on which randomly chosen ubiquitous proteins outscored representative proteins was problem 4, for which representative proteins often divided the organisms as ((Buchnera, Rickettsia), Mycoplasma) (Staphylococcus, (Mycobacterium, Bifidobacterium)), not finding the relatedness of Mycoplasma and Staphylococcus. As intended, representative proteins accurately represent the consensus distance matrix, which includes the contributions of many non-ubiquitous genes and has large distances to the endocellular organisms Buchnera and Mycoplasma, whereas the randomly chosen ubiquitous genes underrepresent the genomic distances to these two organisms and thus do not imply a tree with such long branches. Altogether the results on the 10 test problems suggest that representative proteins reconstruct short internal branches more efficiently than do randomly chosen ubiquitous proteins, but at the cost of greater susceptibility to long-branch attraction.
Table 1 Validation of our methodology on 10 deep phytogeny problems. Organism abbreviations are shown in Table 3, and the accepted clades are shown with parentheses. The column labeled "# Clades" gives the number of accepted clades to be found. The column labeled "# Genes" gives the number of genes used. The Trees column gives the number of gene trees that find all the accepted clades; results for representative proteins are on the left, and results for randomly picked ubiquitous proteins are on the right. For each gene, the most conserved 300-residue sequence was used, and randomly picked proteins were matched to the representative proteins in overall conservation level. Consensus gives the number of accepted clades found over all gene trees; an asterisk indicates that the consensus tree (computed using CONSENSE from the PHYLIP package [52]) finds all the accepted clades. Concatenation gives the number of clades found in 100 bootstraps from a concatenated alignment of all genes; an asterisk here indicates the success of the consensus over bootstrap trees. In problem 6 for example, there are 5 accepted clades, 8 single-gene trees, and 100 bootstrap trees, so a perfect "Consensus" score would be 40, and a perfect "Concatenation" score would be 500.
Organisms # Clades # Genes Trees Consensus Concatenation
1. (Borr, Trep) (Chlor, Bac) (Campy, Bruc) 3 8 8* 2 24* 12 299 112
2. (Neiss, Rals) (Xyl, Haem) (Rick, Meso) 3 8 5* 3 21* 19 247 207
3. (Clost, Lacto) (Mycob, Bifid) (Campy, Rick) 3 8 6* 4* 18* 18* 294 283
4. (Buch, Rick) (Mycob, Bifid) (Staph, Mycop) 3 8 2 1* 13 15* 235 297
5. (Urea, Mycop) (Strep, Lacto) (Staph, List) 3 8 8* 5* 24* 21* 300 300
6. (Syn, Pro) (Rick, Buch) (Chlor, Bac) (Staph, Strep) (Borr, Trep) 5 8 7* 2* 37* 26* 481 472
7. ((Rick, Bruc) ((Vib, Esch, Haem), Neiss) (Heli, Campy)) (Syn, Pro) (Clost, Staph) (Borr, Trep) 8 17 3* 3 129* 108 762 741
8. ((Caul, Meso), Esch) (Chlor, Bac) (Pro, Nos) 4 8 7* 3* 30* 27* 400 398
9. ((Geo, Desulf), (Wol, Campy), (Caul, Rick)) (Borr, Lep) (Chlor, Bac) 6 8 1 2 31* 32 554 512
10. (Chlor, Bac) (Mycop, Strep, Clost) (Mycob, Bifid) 3 8 1* 2* 15* 13 255 245
Proteins as phylogenetic characters
We find that protein representativeness cannot be determined a priori, as many favored proteins such as elongation factors and ribosomal subunits turn out to be poor representatives for certain sets of genomes. For example, EF-G is a relatively weak representative for Proteobacteria; the pairwise distance matrix computed from EF-G alone has a correlation coefficient of .64 with the 200-protein consensus distance matrix, whereas the best proteins such as GroEL have correlation coefficients greater than .90 as shown in Table 2. Yet EF-G is an acceptable protein for rooting the bacterial tree or for computing the initial, overall tree for Figure 1 using a diverse set of 30 bacteria; for these tasks its correlation coefficients of .77 and .70 are not much lower than those of the best proteins. Brochier et al. [4] also found EF-G suspect; it fell outside of their main cluster, and with further analysis they found probable horizontal gene transfer (HGT) between β- and γ-Proteobacteria. Ribosomal proteins S1, S14, and L4 are poor representatives for the diverse set of 30 bacteria, with anomalously small distances between Deinococcus and α-Proteobacteria, Actinobacteria and Proteobacteria (also noted by [28]), and Actinobacteria and Firmicutes, respectively. Yet S1 turns out to be a good representative for Proteobacteria alone; in this case S1 has a correlation coefficient of .88, whereas for the diverse set the correlation coefficient is only .29. Generally tRNA synthetases are less representative than elongation factors and ribosomal subunits.
Table 2 Representative proteins used to compute Figure 2. Class is the COG functional code [18]. Rank is rank in a list of most conserved proteins (families of orthologs), from 0 to 199, for the set of genomes under study; thus FtsH (rank 2) is more conserved than DNA polymerase I (rank 59). Coeff, S. Dev, and Max are respectively the correlation coefficient, the standard deviation, and the maximum elementwise difference between the scaled distance matrix given by this protein and the consensus distance matrix. Distances for this set of organisms were approximately 0–150. Each sequence was limited to the most conserved 300-long amino acid sequence for the protein.
Gene Class Name COG Rank Coeff S.Dev Max
GidA D glucose-inhibited division protein 0445 23 .92 4.27 11.90
- R GTP-binding protein 0012 43 .94 3.84 12.43
RuvB L Holliday junction DNA helicase 2255 19 .89 4.32 12.51
Pnp J polynucleotide phosphorylase 1185 25 .86 4.94 13.11
PyrG F CTP synthetase 0504 26 .91 4.19 13.95
LepA N GTP-binding elongation factor 0481 15 .92 4.88 14.00
DnaX L DNA polymerase III subunits gamma and tau 2812 86 .90 4.37 14.59
Mfd LK transcription-repair coupling factor 1197 31 .88 4.82 14.94
UvrB L DNA excision nuclease subunit B 0556 12 .93 4.44 16.29
InfB J translation initiation factor IF-2 0532 32 .90 4.59 17.46
Exo L DNA polymerase I 0258 59 .89 4.85 17.60
PolC L DNA polymerase III, alpha chain 0587 61 .77 6.43 17.81
RecA L RecA protein 0468 4 .85 6.22 19.18
GyrA L DNA gyrase subunit A 0188 10 .88 5.60 19.74
HflB 0 cell division protein FtsH 0465 2 .86 5.29 19.89
ClpX O ATP-dependent Clp protease, ClpX 1219 13 .89 5.12 20.10
ThrS J threonyl-tRNA synthetase 0441 33 .77 6.94 20.19
Rho K transcription termination factor rho 1158 3 .87 5.83 20.20
GroL O GroEL, chaperone Hsp60 0459 8 .92 5.90 20.35
ClpB 0 ClpB protein 0542 7 .75 6.23 21.01
- R putative GTP-binding protein 1160 165 .94 7.93 21.07
DnaK 0 dnaK, chaperone Hsp70 0443 5 .81 6.48 21.27
RpSA J 30S ribosomal subunit protein S1 0539 38 .88 8.14 22.48
RpoA K DNA-directed RNA polymerase alpha chain 0202 102 .91 10.93 32.41
TrxB 0 thioredoxin reductase 0492 66 .87 8.67 32.54
UvrC L excinuclease ABC subunit C 0322 133 .93 6.67 32.68
NusA K transcription pausing 0195 106 .83 11.11 39.39
QRI7 O o-sialoglycoprotein endopeptidase 0533 123 .89 7.53 43.68
YidC N 60 kD inner membrane protein 0706 179 .85 13.69 53.24
SecY N subunit of translocase 0201 82 .86 8.00 58.78
Figure 1 A rooted phylogenetic tree of Bacteria computed from representative proteins. As explained in the text, this tree was computed first altogether, then with an outgroup of Aeropyrum and Methanopyrus to place the root, and then again in overlapping halves using different proteins, with the split at the doubled edge near Chlorobium. The numbers indicate bootstrap support for clades out of 100 trials; omitted numbers are all 100. The bootstrap support for the root is 44; the second choice is shown dashed. The weakest bootstrap support is for the Spirochaetes and Chlamydiales clade; again the second choice is shown dashed. The Chloroflexus genome is available only in a draft; we give it a tentative placement, without bootstrap support or edge lengths, based on about 1200 columns.
Methionyl-tRNA synthetase seems to have been horizontally transferred between Cyanobacteria and Firmicutes, isoleucy1-tRNA synthetase between Actinobacteria and Rickettsiales, and at least one domain of alanyl-tRNA synthetase between Aquifex and Bacteroidetes/Chlorobi. Researchers have been somewhat divided about tRNA synthetases. Brown et al. [5] used them as phylogenetic characters, but Brochier et al. [4] used them as a gene sample enriched in HGT. By individually screening protein families for representativeness for a specific set of organisms, we can use the representative tRNA synthetases and avoid the anomalous ones; for example we use threonyl-tRNA synthetase (correlation coefficient .77) for the left half of Figure 1.
Despite the examples just given, representative proteins do tend to come disproportionately from core functional categories such as transcription and translation, consistent with the hypothesis [29,30] that HGT is less common for informational proteins than for metabolic proteins. Table 2 gives the list of representative proteins used to compute Figure 2. Several poorly characterized GTPases (Ffh, Obg, LepA, COG0012, COG1160), rarely used in phylogeny, repeatedly turned out to be representative [see Additional file 1]. Table 2 also lists three different measures of representativeness – agreement with the consensus distance matrix – for the representative proteins. One of these measures, the maximum elementwise difference between the single-gene and the consensus distance matrices, is generally quite large; every protein gave some pairwise distance that differed by at least 10% from the corresponding consensus distance. This observation means that very few proteins land near the center of all pairwise histograms (Figure 6) for a diverse set of bacteria; most proteins are well away from the mode in at least one such histogram. All measures of representativeness improve quite dramatically, however, with decreasing taxon diversity. The computation of Figure 3 for β- and γ-proteobacteria used only proteins with correlation coefficients at least .88, whereas the overall bacterial phylogeny necessarily used proteins with correlation coefficients as low as .62.
Figure 2 An unrooted phylogenetic tree of Proteobacteria and related bacteria. This tree shows the left half of Figure 1, including a number of additional genomes. The numbers associated with edges give bootstrap support as before; the best supported alternative choices are shown dashed. This tree slightly favors breaking the Spirochaetes/Chlamydiales clade of Figure 1.
Figure 3 An unrooted phylogenetic tree of β- and γ-Proteobacteria. The edge labeled 67 is the only edge without bootstrap support of 100; the one alternative topology covers the other 33 bootstrap trials. This tree switches the branching order of Vibrio and Haemophilus from that shown in Figures 1 and 2. This tree should be more reliable, due to better taxon sampling and protein representativeness. For this less diverse set of taxa, many proteins had correlation coefficients greater than .90 with the consensus distance matrix. Notice that the two species of Vibrio are much more diverged than Escherichia and Salmonella.
Figure 6 Histograms of evolutionary distances. Plotted are the evolutionary distances, between E. coli and three other bacteria, Streptococcus pneumoniae, Neisseria meningitidis, and Haemophilus influenzae. Each distance D(i, j, k), described in the Methods section, is given by a pairwise alignment of amino acid sequences of a given length (typically 300 residues), the most conserved subsequences for a family of orthologous proteins. We can interpret distances as times, with greater time towards the left. All three histograms are roughly bell-shaped but with rather high variances, which suggests that reliable phylogenetic inference requires either a great many sequences or representative sequences that sit near the center in all pairwise histograms. The peaks at 100+ indicate missing orthologs. There are several apparent horizontal transfers (right-side outliers) in S.pneumoniae and N.meningitidis. Even discounting the peaks at 100+, the left-side outliers (rapid evolution, large insertions or deletions, missing domains, hidden paralogs, and horizontal transfers from more distant organisms) outnumber right-side outliers; this pattern holds true even for very distant pairs such as S.pneumoniae and E.coli.
Bacterial phylogeny
Figures 1, 2, 3, 4, 5 give our phylogenies for Bacteria. These trees provide further validation of the methodology in the sense that they correctly identify all accepted clades, and are fairly robust under bootstrapping and under varying the choices of representative proteins and species. These trees were computed by maximum-likelihood methods on alignments of 5000–8000 columns, concatenations of (typically 300-residue) subsequences of the 20–40 most representative proteins for the set of genomes under study. The same number – or even a greater number – of columns of randomly chosen ubiquitous proteins gives worse results, not always finding the monophyly of Proteobacteria and of γ-Proteobacteria, two well-accepted clades that are nontrivial to resolve [23]. As explained in the Methods section, Figure 1 was computed all at once with a subsample of organisms, and again in two overlapping pieces (which gave compatible trees), each piece again using 5000–8000 columns. A separate run with the same subsample of organisms along with two archaea (Aeropyrum and Methanopyrus) was used to root the tree, and yet another separate run was used to place Chloroflexus, a partially sequenced genome. Figures 1, 2, 3, 4, 5 are in almost complete agreement with each other, as were the several different runs used to compute Figure 1, so our results seem to be robust under different taxon samples. One exception is Bacteroidetes/Chlorobi; if this phylum is omitted as in [6] deep branches rearrange, for example, Actinobacteria/Deinococcus/Cyanobacteria forms a high-level clade with Firmicutes. Figure 1 suggests that Chlorobium is the organism that has diverged least from the difficult-to-resolve central area of the tree.
Figure 4 An unrooted phylogenetic tree of Firmicutes, Actinobacteria, and Cyanobacteria. Chlorobium and Bacteroides were included here and in Figure 2 in order to orient the trees relative to each other.
Figure 5 An unrooted phylogenetic tree of Firmicutes and Actinobacteria. This tree includes the reduced genomes of Mycoplasma and Ureaplasma. We found that if these genomes were included in larger phylogenies, such as those in Figures 1 and 4, we obtained unreliable results with poor bootstrap support.
The trees given in Figures 1, 2, 3, 4, 5 show substantial areas of agreement with those of other researchers [4-7,17]. Especially striking is the agreement with the recently published phylogeny by Battistuzzi et al. [17], computed from about 8000 columns from 32 manually selected genes. Figure 1 disagrees with the tree of Battistuzzi et al. only at three internal branches with poor bootstrap support in both studies: the root (they use the dashed edge), the edge marked 81 (they join Actinobacteria/Deinococcus/Cyanobacteria with Firmicutes), and the edge marked 85 between the branches to Haemophilus and Vibrio (they switch the order as in our Figures 2 and 3).
Our study supports a basal position for Aquifex as in [4,6,17], rather than placing it near Bacteroidetes/Chlorobi [7,8] or Proteobacteria [31]. We place Bacteroidetes/Chlorobi as a bridge between Gram-positive bacteria and Proteobacteria as in [17] and in the rRNA tree of Brochier et al. [4]. We form a clade containing Actinobacteria, Deinococcus, and Cyanobacteria ("Terrabacteria" [17]) as in [4-6,17], but (just barely) reject a higher-level clade [6,17] combining Actinobacteria/Deinococcus/Cyanobacteria with Firmicutes. We place Planctomycetes (Pirellula) with Chlamydiales as in [32]. We find that Spirochaetes, Chlamydiales (plus Pirellula), and ε-Proteobacteria form a close trio, with exact branching order hard to resolve, due to closely spaced branching events and/or numerous horizontal transfers. The bootstrap values at the branches in this trio stand out as weak in our generally well-resolved tree. We also consider the root to be relatively weak, not only because of its bootstrap value but also because it appears next to two thermophilic eubacteria showing the most HGT with Archaea. Our program rejects obvious cases of HGT as non-representative, but it cannot screen out subtle ancient transfers or genome-wide biases in amino acid composition. We reject grouping δ- and ε-Proteobacteria into a clade as in GenBank's taxonomy, because such a grouping had zero support out of 100 trials. We place Buchnera/Wigglesworthia farther than Haemophilus, Vibrio, and Shewanella from E. coli, in agreement with [17], but in disagreement with a recent phylogeny of γ-Proteobacteria [33] and two more focused studies [34,35]. Our placement of Buchnera/Wigglesworthia, however, may be an artifact related to long-branch attraction as in problem 4 of Table 1, so the true position of these reduced genomes may indeed be with enterobacteria.
Our tree supports the history of photosynthetic organisms determined by signature sequences [8]; moreover it places Chloroflexus, Chlorobium, and Cyanobacteria in close proximity. Even so, our tree implies either HGT or a common ancestor with both RC-1 and RC-2 types of reaction centers [8,36,37]. One explanation of the paraphyly of RC-1 and RC-2 would be horizontal transfer from Chloroflexus/Cyanobacteria to Proteobacteria, and indeed our program flags protochlorophyllide reductase, both ChlB and ChlN subunits, as anomalously close between Cyanobacteria and photosynthetic proteobacteria. Our program does not, however, find an unusually large amount of HGT among photosynthetic bacteria.
Discussion
Any attempt to screen proteins for anomalous evolution inevitably leads to questions concerning horizontal gene transfer. The rate of HGT is notoriously difficult to estimate [38,39]. One study [40] indirectly estimates that about 30% of genes have been subject to HGT within Bacteria, whereas a more direct analysis finds evidence for HGT in about 50% of ubiquitous genes over all prokaryotes [17]. A recent study [26] finds that about 30% of single-ortholog COGs deviated significantly from "clock-like evolution" over smaller taxa, such as γ-Proteobacteria, with about half of these deviations due to HGT and half due to other anomalies. In what is perhaps the most detailed study to date, Lerat et al. [33] found only two strong examples of horizontal transfers among 205 genes within γ-Proteobacteria.
We did not derive estimates of the rate of HGT from our experiments; however, our finding that representativeness improves dramatically with decreasing taxon diversity suggests that any study relying on a set of bacterial genomes from a single phylum, as in the recent work on γ-Proteobacteria [26,33], is likely to underestimate the rate of HGT. Our finding (Figure 6) that unusually large pairwise evolutionary distances are much more common than unusually small distances confirms [26] that many proteins deviate from clock-like evolution due to discontinuous evolutionary events besides HGT.
In a recent study, Raymond et al. [41] concluded that HGT has obscured the history of photosynthetic organisms on the basis of widespread disagreement among 188 gene trees relating five such organisms: Synechocystis, Chloroflexus, Chlorobium, Rhodobacter, and Heliobacterium (close to Clostridium). We think, however, that the phylogeny of these five diverse and sparsely sampled [42] bacteria is simply too hard a problem for single gene sequences to resolve. We found that 100 bootstraps from single genes also gave widespread disagreement among resulting trees. An important point is that the number of alignment sites (columns) required to build a robust tree depends not on the number of organisms but rather on the length of the shortest internal tree edge. Daubin et al. [43] recently made this same point, and further argued that HGT has been overused as an explanation for the difficulty of bacterial phylogeny. Horizontal transfer, however, at the rate of one core gene replaced – "xenologous displacement" [26] – in each genome per 100 million years would be sufficient to pose an obstacle to deep bacterial phylogeny, yet remain undetected by the technique of Daubin et al., which looked for transfers into exactly one of a closely related pair (probably less than 100 million years since divergence) such as E. coli and Salmonella typhimurium.
Conclusion
Ultimately prokaryotic phylogeny faces many obstacles beyond horizontal transfer. Even in eukaryotes, where HGT is extremely rare, Rokas et al. [44] found that concatenations of a large number of genes gave more consistent trees and better bootstrap support than single genes and concatenations of fewer than 20 genes. Representative proteins are clock-like proteins, ubiquitous over the set of genomes under study, selected with reference to a distance matrix computed by robust statistical methods from hundreds of well-conserved, but not necessarily ubiquitous, genes. Concatenations of representative proteins give results consistent with much larger concatenations of randomly selected proteins for phylogenetic problems with known answers, so there is reason to believe that representative proteins can also provide the statistical leverage necessary to shed light on deep unresolved branches that remain controversial even with the use of all ubiquitous genes. The method of representative proteins, however, like all phylogenetic methods, has strengths and weaknesses. By incorporating information from many faster-evolving, non-ubiquitous genes, the method may be more susceptible to long-branch attraction and convergent evolution than reliance on ubiquitous genes.
In our view, genes have rich and varied evolutionary histories, contingent upon selective pressures and random events at several different levels. Hence genes are likely to show continuous variation in quality as phylogenetic characters, rather than falling into two universal categories, bad and good, "subject to horizontal transfer" and "conserved core" [4,19]. There may well be genes whose entire evolutionary histories are smooth and regular, but there are surely not enough of them to resolve all phylogenetic questions, and we will be forced to use other, locally reliable genes to resolve certain parts of the tree of life. Hence we believe that the tools we developed for mining and ranking potential characters fill an important niche in the evolving methodology of phylogenetic inference.
Methods
We used the publicly available genomes from GenBank [45] and the Joint Genome Institute [46]. We relied on the annotations to identify protein-coding genes, but not to identify orthologous sequences. All the software developed for this project is available from the authors upon request.
Orthologous sequences
We used our own data-mining program (manuscript in submission), similar to an all-against-all BLAST search [47], to find orthologous sequences and rank them by degree of conservation. Given a set of genomes and a sequence length, say 300 amino acid residues, this program returns a ranked list of the 200 most conserved proteins at that length, along with pointers into the genomes for the locations of the sequences. If there are not 200 well-conserved orthologous families within the set of genomes – for example if the set includes a mix of eubacteria and archaea – then the program returns only as many families as are deemed to be obvious homology (approximately 30% pairwise identity). The choice of 200 is somewhat arbitrary, but seemed to be close to the maximum for genome sets containing more than one bacterial phylum, without incurring much misalignment or mutational saturation. The program is not designed to find remote homology, and existing tools such as PSI-BLAST [48] are in fact better for this task. We ran the program for several different sequence lengths from 60 to 300, and included each representative protein only once, at the maximum length for which it was representative.
We could have used existing tools such as BLAST or existing databases such as COGs for ortholog assembly, but our own program offered several advantages. First, it ranks ortholog families by conservation level, measured by the quartile log odds similarity over all pairs, that is, the similarity score greater than 1/4 and smaller than 3/4 of the pairwise similarities. (Thus if a gene is missing from more than 1/4 of the genomes under study, it has a conservation level near zero.) The program limits attention to sequences of fixed length, such as 300 residues, in order to compare conservation levels fairly. Then by examining ortholog families in order of decreasing conservation level, the program screens out families that are not sufficiently conserved. Second, it is much faster than BLAST, so that we could perform ortholog assembly separately for each set of genomes, thereby finding proteins conserved within the taxa under study, but not conserved more generally. Third, because it finds ortholog families using all pairwise alignments of candidate orthologs, rather than by a reciprocal or circular BLAST search, it also minimizes the problem of hidden paralogy [49]. We rarely observed paralogs (as identified by GenBank annotations) when the sequence length was greater than 100 residues. Moreover, selection of representative proteins should filter out remaining hidden paralogy that could mislead the tree-building program.
Representative sequences
The next step condenses the large amount of information in the conserved sequences to a matrix {D(i, j)} of pairwise evolutionary distances. The step also retains a matrix {D(i, j, k)} for each protein k, that is, the pairwise evolutionary distances given by the k-th set of orthologous sequences. The amount of agreement between {D(i, j, k)} and {D(i, j)} determines whether protein k is representative. We start by computing, for each pair of organisms and each protein, a pairwise alignment and a log odds score (Smith-Waterman algorithm using the BLOSUM50 substitution matrix with default gap costs). This gives a three-dimensional array of log odds similarity scores S(i, j, k), where i and j index the organisms and k indexes the proteins. There are many alternative choices of similarity scores, simpler scores such as percent identity and more complicated scores involving multiple BLOSUM matrices; the alternatives we tried gave nearly identical choices of representative proteins. We assume that each protein has its own "clock" as in Yang's proportional model [50], so that in any given time period, the log odds of a given amount of evolution for protein k is a fixed multiple of the log odds of the same amount of evolution for protein k'. We compute the multiplier Mk for protein k with an iterative procedure. We first detect whether an organism i is missing protein k; for such an i and k, score S(i, j, k) is very low for all j. We discard these S(i, j, k) scores altogether. Then starting with Mk = 1 for all k, we alternate the following steps. Convergence (to three decimal places) resulted after only three iterations.
1. For each i and j, set S(i, j) ← trimmed mean MkS(i, j, k). For each i and j, we drop the top and bottom 20% of the MkS(i, j, k) values and compute the mean of the middle 60%. We chose this trimmed mean after estimating the size of distribution tails (anomalous distances) in histograms such as those shown in Figure 6.
2. For each k, set Mk ← median S(i, j) / S(i, j, k). Then normalize Mk ← Mk·c, where c = m / ∑kMk and m is the number of proteins, so that the Mk's average 1.0.
Step 1 forms a consensus over proteins and step 2 forms a consensus over organism pairs. We used robust statistics (trimmed means and medians) instead of ordinary means due to the many outliers. Multipliers typically varied from about 0.5 for the most conserved protein to about 1.6 for the 200-th most conserved protein. Novichkov et al. [26] independently developed a similar procedure; however, they used the median rather than trimmed mean in Step 1 and simply ran the two steps once each, without iteration. Because the consensus similarities S(i, j) improve with estimates of the multipliers Mk, iteration gives better results, as we discuss below.
We convert the scaled similarity scores Mk·S(i, j, k) to evolutionary distances D(i, j, k) = C - log (Mk·S(i, j, k)), where C is a constant chosen so that all the D(i, j, k) are positive. In the case that organism i or j was identified as missing protein k, we set D(i, j, k) = ∞. Figure 6 shows histograms of D(i, j, k) values for three different pairs of organisms; the distances for missing proteins appear at 100+.
We set the consensus distance D(i, j) between organisms i and j to be the trimmed mean (again using the middle 60%) of the finite (not from missing proteins) D(i, j, k) values. We tested how well distance matrices conformed to trees by running the Fitch-Margoliash algorithm [1,51] (program FITCH in the PHYLIP package [52]). For example, the Fitch-Margoliash tree [see Figure S1 in Additional file 1] gives 0.199 relative squared errror, that is, 0.199 = ∑i, j (T(i, j) - D(i, j))2/D(i, j)2, where T(i, j) is the pairwise distance given by the tree, and gives 1.51% average percent standard deviation (APSD) [51,52], a measure of the typical error of T(i, j) relative to D(i, j). For comparison, we tested the median procedure of Novichkov et al. by computing consensus similarities with SN(i, j) = median S(i, j, k) and consensus distances with DN(i, j) = C - log SN(i, j). On 8 out of the 9 distance matrices used in our study, {D(i, j)} conformed to tree distances (computed with FITCH) more closely than did {DN(i, j)}, with APSDs ranging from 0.63 to 3.67 for {D(i, j)} and from 0.92 to 4.26 for {DN(i, j)}. For the organisms in Figure S1 the FITCH tree produced from {DN(i, j)} broke several accepted clades and achieved a relatively poor APSD of 4.24%. The one case on which {DN(i, j)} outperformed {D(i, j)} was the taxon sample that included the partially sequenced genome Chloroflexus.
The distances {D(i, j, k)} given by a protein k can be regarded as a vector with N = n(n - 1)/2 entries, where n is the number of organisms. We computed three different measures of how well {D(i, j, k)} represents the consensus matrix {D(i, j)}. The three measures are: (1) correlation coefficient of {D(i, j, k)} with {D(i, j)}; (2) standard deviation, that is, ; and (3) max distance, that is, maxi, j |D(i, j, k) - D(i, j)|. We chose the top-ranking proteins by max distance, enough proteins to give approximately 10,000 columns; we marked the worst 20% of proteins by each of measures (1) and (2), and dropped all marked proteins, thereby obtaining 6000–8000 columns. Thus a protein had to be good on all three measures to be considered representative; in particular, a protein had to be ubiquitous over the organisms under study. For this reason, we initially left out organisms with very reduced genomes, such as Mycoplasma and Buchnera; these organisms are best added to the tree (Figures 3 and 5) later using representative proteins chosen for a smaller range of taxa.
Multiple alignment and tree building
We used CLUSTAL W [53] to compute a multiple alignment for each set of orthologs. We removed columns of possibly incorrect alignment using a method recommended by Sidow (personal communication). We first removed all columns containing a gap character -, and then removed columns to the left and right of a gap column until reaching a column of chemical agreement. A column of chemical agreement is one in which all the amino acids in the column are from a single group, where the groups are acidic residues {D, E}, aromatic residues {F, W, Y}, basic residues {H, K, R}, cysteine {C}, nonpolar residues {A, G, I, L, P, V}, and polar residues {M, N, Q, S, T}. We also removed blocks of more than eight consecutive columns in which no column was one of chemical agreement; even if such a highly variable block is a correct alignment (one-for-one amino acid substitution over evolutionary time), it may have too many superimposed mutations to be helpful for phylogeny. Finally we concatenated all the cleaned alignment files, obtaining a multiple alignment of about 5000–7000 columns.
We used SEMPHY Version 0.9 to compute phylogenetic trees [54]. SEMPHY is a relatively fast EM (expectation maximization) program for ML (maximum likelihood) phylogeny, which assumes a Markov model of evolution. We used the JTT model [55], which is SEMPHY's default. The newer SEMPHY Version 1.0 models mutation rate variation among columns with a discrete gamma distribution, but we found only insignificant differences in the resulting ML trees, so we preferred the simpler, homogeneous-rate model. Cleaned alignments of representative proteins tend to be more homogeneous than alignments of random proteins. To estimate the confidence of clades, we used a bootstrapping procedure or – more precisely – a jackknife procedure. To create a random subsample of the entire alignment file we included each block of 80 consecutive columns in the subsample with probability 0.5; this gives a slightly harsher test than the standard bootstrap. We tested one tree (the overall tree for Figure 1) both ways; the jackknife numbers were all 0–10% smaller than the corresponding bootstrap numbers. The rationale for randomly sampling blocks rather than individual columns was to effectively vary the set of representative proteins as in [44]. For each ML tree computation, we created 100 random subsamples and ran SEMPHY on each subsample. Typical running times were 15–20 minutes per subsample, or about 30 hours for 100 subsamples. We also tried PhyML [56], another fast program for ML phylogeny. With a single substition rate category, PhyML Version 2.4.4 is faster than SEMPHY (7 minutes versus 17 minutes for a computation with 29 organisms and 5424 columns). On the other hand, PhyML may be more prone to converge to a suboptimal local optimum, as we sometimes obtained a greater likelihood within PhyML by inputting SEMPHY's solution as PhyML's starting tree.
To produce Figure 1, we made five different runs of tree computations with different sets of genomes. We chose representative proteins separately for each run, because representativeness depends upon the set of genomes under study. An initial run with 30 diverse bacteria [the 30 bacteria shown in Figure S1 of Additional file 1], with all phyla except Planctomycetes represented, sketched out an unrooted phylogeny. The root was placed by a run with the same 30 eubacteria along with two archaea, Aeropyrum and Methanopyrus. We then split the tree in two pieces ("left" and "right") by cutting it at the deep interior edge with best bootstrap support, the doubled edge shown in the figure. Computing a large phylogenetic tree in pieces allows the use of more proteins of greater representativeness, but such a split must be done very cautiously because each organism potentially affects all the unknown sequences at interior nodes. We added more organisms, such as Xanthomonas and Vibrio, assigning them to the left or right half according to the accepted taxonomy. We then added a right-half organism (Chlorobium) to the left half, and a left-half organism (Escherichia) to the right half, and made two more runs of tree computations, one for the left and one for the right. These runs corroborated the split by their placements of Chlorobium and Escherichia. In Figure 1 all edge lengths and bootstrap-support numbers refer to these last two runs, except for the numbers on the doubled edge and the root. (Edge lengths on the left and right sides of the tree are thus not strictly comparable as they refer to different proteins.) Finally, we added Chloroflexus by rerunning the overall computation, that is, 31 organisms in all. The draft genome, however, is missing many typically representative proteins such as GroEL and elongation factor TU, so we could harvest only about 1200 columns of representative proteins, and we indicate its position by a dashed edge without bootstrap support. (The 84 in Figure 1 is the support for the Actinobacteria/Deinococcus/Cyanobacteria clade.) In order to produce Figure 2, we reran the organisms of the left half along with some other fully sequenced Proteobacteria, Spirochaetes, and Chlamydiales. For Figure 3, we reran our procedures on γ-and β-proteobacteria, now including the reduced genomes of Buchnera and Wigglesworthia. For Figure 4, we reran most of the organisms of the right half along with some other genomes, and finally for Figure 5 we ran Actinobacteria and Firmicutes (along with Fusobacterium and Deinococcus), now including the reduced genomes of Mycoplasma and Ureaplasma. We attempted to include these reduced genomes at the higher level of Figure 4, but obtained inconsistent results depending upon the proteins selected.
Authors' contributions
DG wrote the program that assembles the orthologs and ranks them by degree of conservation. MB wrote the program that selects representative proteins, and carried out all the computational experiments.
Table 3 Bacterial genomes used in this paper. All phyla in GenBank as of December 2004 are represented. Bold letters give abbreviations used in Table 1.
Specific Genome Taxonomy (Phylum; Class)
Corynebacterium glutamicum Actinobacteria;Actinobacteria
Mycobacterium tuberculosis H37Rv Actinobacteria;Actinobacteria
Bifidobacterium longum Actinobacteria;Actinobacteria
Streptomyces avermitilis Actinobacteria;Actinobacteria
Aquifex aeolicus Aquificae;Aquificae
Bacteroides thetaiotaomicron VPI-5482 Bacteroidetes/Chlorobi;Bacteroidetes
Chlorobium tepidum TLS Bacteroidetes/Chlorobi;Chlorobi
Porphyromonas gingivalis W83 Bacteroidetes/Chlorobi;Chlorobi
Chlamydia trachomatis Chlamydiae/Verrucomicrobia;Chlamydiae
Chlamydophila pneumoniae AR39 Chlamydiae/Verrucomicrobia;Chlamydiae
Chloroflexus aurantiacus Chloroflexi;Chloroflexi
Gloeobacter violaceus Cyanobacteria;Chroococcales
Synechococcus sp WH8102 Cyanobacteria;Chroococcales
Synechocystis PCC6803 Cyanobacteria;Chroococcales
Nostoc sp Cyanobacteria;Nostocales
Prochlorococcus marinus MIT9313 Cyanobacteria;Prochlorophytes
Deinococcus radiodurans Deinococcus-Thermus;Deinococci
Bacillus subtilis Firmicutes;Bacilli
Oceanobacillus iheyensis Firmicutes;Bacilli
Listeria monocytogenes Firmicutes;Bacilli
Staphylococcus aureus subsp. aureus N315 Firmicutes;Bacilli
Lactococcus lactis Firmicutes;Bacilli
Streptococcus pneumoniae R6 Firmicutes;Bacilli
Clostridium tetani E88 Firmicutes;Clostridia
Thermoanaerobacter tengcongensis Firmicutes;Clostridia
Mycoplasma Firmicutes;Mollicutes
Ureaplasma Firmicutes;Mollicutes
Fusobacterium nucleatum Fusobacteria;Fusobacteria
Pirellula sp Planctomycetes;Planctomycetacia
Caulobacter crescentus Proteobacteria;Alphaproteobacteria
Rhodopseudomonas palustris Proteobacteria;Alphaproteobacteria
Brucella melitensis Proteobacteria;Alphaproteobacteria
Bradyrhizobium japonicum Proteobacteria;Alphaproteobacteria
Mesorhizobium loti Proteobacteria;Alphaproteobacteria
Rickettsia conorii Proteobacteria;Alphaproteobacteria
Ralstonia solanacearum Proteobacteria;Betaproteobacteria
Neisseria meningitidis Z2491 Proteobacteria;Betaproteobacteria
Chromobacterium violaceum Proteobacteria;Betaproteobacteria
Bordetella pertussis Proteobacteria;Betaproteobacteria
Nitrosomonas europaea Proteobacteria;Betaproteobacteria
Coxiella burnetii Proteobacteria;Gammaproteobacteria
Escherichia coli K12 Proteobacteria;Gammaproteobacteria
Haemophilus influenzae Proteobacteria;Gammaproteobacteria
Pseudomonas aeruginosa Proteobacteria;Gammaproteobacteria
Shigella flexneri 2a Proteobacteria;Gammaproteobacteria
Shewanella oneidensis Proteobacteria;Gammaproteobacteria
Vibrio cholerae Proteobacteria;Gammaproteobacteria
Xanthomonas campestris Proteobacteria;Gammaproteobacteria
Xylella fastidiosa Proteobacteria;Gammaproteobacteria
Desulfovibrio desulfuricans Proteobacteria;delta/epsilon subdivisions
Geobacter sulfurreducens Proteobacteria;delta/epsilon subdivisions
Campylobacter jejuni Proteobacteria;delta/epsilon subdivisions
Helicobacter pylori J99 Proteobacteria;delta/epsilon subdivisions
Wolinella succinogenes Proteobacteria;delta/epsilon subdivisions
Borrelia burgdorferi Spirochaetes;Spirochaetes
Leptospira interrogans Spirochaetes;Spirochaetes
Treponema pallidum Spirochaetes;Spirochaetes
Thermotoga maritima Thermotogae;Thermotogae
Supplementary Material
Additional File 1
(1) a phylogenetic tree computed directly from a distance matrix for 30 diverse bacteria, (2) the representative proteins used in all the phylogenetic tree computations, and (3) some probable horizontal transfers detected by the study.
Click here for file
==== Refs
Swofford DL Olsen GJ Waddell PJ Hillis DM Hillis DM, Moritz C, Mable BK Phylogenetic inference Molecular Systematics 1996 Sinauer Associates
Ludwig W Strunk O Klugbauer S Klugbauer N Weizenegger M Neumaier J Bachleitner M Schleifer KH Bacterial phylogeny based on comparative sequence analysis Electrophoresis 1998 19 554 568 9588802 10.1002/elps.1150190416
Olsen GJ Woese CR Overbeek R The winds of (evolutionary) change: breathing new life into microbiology J Bacteriology 1994 176 1 6
Brochier C Bapteste E Moreira D Philippe H Eubacterial phylogeny based on translational apparatus proteins TRENDS in Genetics 2002 18 1 5 11750686 10.1016/S0168-9525(01)02522-7
Brown JR Douady CJ Italia MJ Marshall WE Stanhope MJ Universal trees based on large combined protein sequence data sets Nature Genetics 2001 28 281 285 11431701 10.1038/90129
Wolf YI Rogozin IB Grishin NV Tatusov RL Koonin EV Genome trees constructed using five different approaches suggest new major bacterial clades BMC Evol Biol 2001 1 8 11734060 10.1186/1471-2148-1-8
Gupta RS Protein phylogenies and signature sequences: a reappraisal of evolutionary relationships among Archaeabacteria, Eubacteria, and Eukaryotes Microbiol and Mol Biol Rev 1998 62 1435 1491 9841678
Gupta RS Mukhtar T Singh B Evolutionary relationships among photosynthetic prokaryotes (Heliobacterium chlorum, Chloroflexus aurantiacus, cyanobacteria, Chlorobium tepidum, and proteobacteria): implications regarding the origin of photosynthesis Mol Microbiol 1999 32 893 906 10361294 10.1046/j.1365-2958.1999.01417.x
Campbell A Mrázek J Karlin S Genome signature comparisons among prokaryote, plasmid, and mitochondrial DNA Proc Natl Acad Sci USA 1999 96 9184 9189 10430917 10.1073/pnas.96.16.9184
Snel B Bork P Huynen MA Genome phylogeny based on gene content Nature Genetics 1999 21 108 110 9916801 10.1038/5052
Fitz-Gibbon ST House CH Whole genome-based phylogenetic analysis of free-living microorganisms Nucleic Acids Research 1999 27 4218 4222 10518613 10.1093/nar/27.21.4218
Tekaia F Lazcano A Dujon B The genomic tree as revealed from whole proteome comparisons Genome Research 1999 9 550 557 10400922
Mushegian AR Koonin EV Gene order is not conserved in bacterial evolution Trends in Genetics 1996 12 289 290 8783936 10.1016/0168-9525(96)20006-X
Clarke GDP Beiko RG Ragan MA Charlebois RL Inferring genome trees by using a filter to eliminate phylogenetically discordant sequences and a distance matrix based on mean normalized BLASTP scores J Bacteriology 2002 184 2072 2080 10.1128/JB.184.8.2072-2080.2002
Zhaxybayeva O Gogarten JP Bootstrap, Bayesian probability, and maximum likelihood mapping: exploring new tools for comparative genome analyses BMC Genomics 2002 3 4 11918828 10.1186/1471-2164-3-4
Zhaxybayeva O Gogarten JP An improved probability mapping approach to assess genome mosaicism BMC Genomics 2003 4 37 12974984 10.1186/1471-2164-4-37
Battistuzzi FU Feijao A Hedges AB A genomic timescale of prokaryote evolution: insights into the origin of methanogenesis, phototrophy, and the colonization of land BMC Evol Biol 2004 4 44 15535883 10.1186/1471-2148-4-44
Tatusov RL Natale DA Garkavtsev IV Tatusova TA Shankavaram UT Rao BS Kiryutin B Galperin MY Fedorova ND Koonin EV The COG database: new developments in phylogenetic classification of proteins from complete genomes Nucleic Acids Research 2001 29 22 28 11125040 10.1093/nar/29.1.22
Daubin V Gouy M Perrière G A phylogenomic approach to bacterial phylogeny: evidence of a core of genes sharing a common history Genome Research 2002 12 1080 1090 12097345 10.1101/gr.187002
Baum BR Combining trees as a way of combining data sets for phylogenetic inference, and the desirability of combining gene trees Taxon 1992 41 3 10
Ragan MA Phylogenetic inference based on matrix representation of trees Mol Phyl Evol 1992 1 53 58 10.1016/1055-7903(92)90035-F
Daubin V Gouy M Perrière Bacterial molecular phylogeny using supertree approach Genome Informatics 2001 12 155 164 11791234
Gophna U Doolittle WF Charlebois RL Weighted genome trees: refinements and applications J Bacteriology 2005 187 1305 1316 10.1128/JB.187.4.1305-1316.2005
Felsenstein J Evolutionary trees from DNA sequences: a maximum likelihood approach J Molecular Evolution 1981 17 368 376 10.1007/BF01734359
Felsenstein J Phylogenies from molecular sequences: inference and reliability Annual Reviews in Genetics 1988 22 521 565 10.1146/annurev.ge.22.120188.002513
Novichkov PS Omelchenko MV Gelfand MS Mironov AA Wolf YI Koonin EV Genome-wide molecular clock and horizontal gene transfer in bacterial evolution J of Bacteriology 2004 186 6575 6585 10.1128/JB.186.19.6575-6585.2004
Billera LJ Holmes SP Vogtmann K Geometry of the space of phylogenetic trees Advances in Appl Math 2001 27 733 767 10.1006/aama.2001.0759
Brochier C Philippe H Moreira D The evolutionary history of ribosomal protein RpS14: horizontal gene transfer at the heart of the ribosome TRENDS in Genetics 2000 16 529 533 11102698 10.1016/S0168-9525(00)02142-9
Garcia-Vallve S Romeu A Palau J Horizontal gene transfer in bacterial and archaeal complete genomes Genome Research 2000 10 1719 1725 11076857 10.1101/gr.130000
Jain R Rivera MC Lake JA Horizontal gene transfer among genomes: the complexity hypothesis Proc Natl Acad Sci USA 1999 96 3801 3806 10097118 10.1073/pnas.96.7.3801
Griffiths E Gupta RS Signature sequences in diverse proteins provide evidence for the late divergence of the order Aquificales Int Microbiology 2004 7 41 52
Teeling H Lombardot T Bauer M Ludwig W Glöckner FO Evaluation of the phylogenetic position of the planctomycete 'Rhodopirellula baltica' SH 1 by means of concatenated ribosomal protein sequences, DNA-directed RNA polymerase subunit sequences and whole genome trees Int J Syst Evol Microbiol 2004 54 791 801 15143026 10.1099/ijs.0.02913-0
Lerat E Daubin V Moran NA From gene trees to organismal phylogeny in prokaryotes: the case of the γ-proteobacteria PLoS Biology 2003 1 1 9 10.1371/journal.pbio.0000019
Itoh T Martin W Nei M Acceleration of genomic evolution caused by enhanced mutation rate in endocellular symbionts Proc Natl Acad Sci USA 2002 99 12944 12948 12235368 10.1073/pnas.192449699
Canback B Tamas IASG A phylogenomic study of endosymbiotic bacteria Mol Biol Evol 2004 21 1110 1122 15014155 10.1093/molbev/msh122
Baymann F Brugna M Mühlenhoff U Nitschke W Daddy, where did (PS)I come from? Biochimica et Biophysica Acta 2001 1507 291 310 11687221
Xiong J Fischer WM Inouse K Nakahara M Bauer CE Molecular evidence for the early evolution of photosynthesis Science 2000 289 1724 1730 10976061 10.1126/science.289.5485.1724
Doolittle WF Phylogenetic classification and the universal tree Science 1999 284 2124 2128 10381871 10.1126/science.284.5423.2124
Daubin V Ochman H Quartet mapping and the extent of lateral transfer in bacterial genomes Mol Biol Evol 2004 21 86 89 12949130 10.1093/molbev/msg234
Kunin V Ouzonis CA The balance of driving forces during genome evolution in prokaryotes Genome Research 2003 13 1589 1594 12840037 10.1101/gr.1092603
Raymond J Zhaxybayeva O Gogarten JP Gerdes SY Blankenship RE Whole-genome analysis of photosynthetic prokaryotes Science 2002 298 1616 1620 12446909 10.1126/science.1075558
Hillis DM Pollock DD McGuire JA Zwickl DJ Is sparse taxon sampling a problem for phylogenetic inference? Syst Biol 2003 52 124 126 12554446
Daubin V Moran NA Ochman H Phylogenetics and the cohesion of bacterial genomes Science 2003 301 829 832 12907801 10.1126/science.1086568
Rokas A Williams BL King N Carroll SB Genome-scale approaches to resolving incongruence in molecular phylogenies Nature 2003 425 798 804 14574403 10.1038/nature02053
Benson DA Karsch-Mizrachi I Lipman DJ Ostell J Rapp BA Wheeler DL GenBank Nucleic Acids Research 2002 30 17 20 11752243 10.1093/nar/30.1.17
DOE Joint Genome Institute
Altschul SF Gish W Miller W Myers EW Lipman DJ Basic local alignment search tool J Mol Biol 1990 215 403 410 2231712 10.1006/jmbi.1990.9999
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 Research 1997 25 3389 3402 9254694 10.1093/nar/25.17.3389
Koski LB Golding GB The closest BLAST hit is often not the nearest neighbor J Mol Evol 2001 52 540 542 11443357
Yang Z Maximum-likelihood models for combined analyses of multiple sequence data J Mol Evol 1996 42 587 596 8662011
Fitch WM Margoliash E Construction of phylogenetic trees Science 1967 228 279 284 5334057
Felsenstein J PHYLIP (Phylogeny Inference Package) version 3.5c Department of Genetics, U of Washington
Thompson JD Higgins DG Gibson TJ CLUSTAL W: improving the sensitivity of progressive multiple sequence alignment through sequence weighting, position-specific gap penalties, and weight matrix choice Nucleic Acids Research 1994 22 4673 4680 7984417
Friedman N Ninio M Pe'er I Pupko T A structural EM algorithm for phylogenetic inference J Computational Biology 2002 9 331 353 10.1089/10665270252935494
Jones D Taylor W Thornton J The rapid generation of mutation data matrices from protein sequences Comput Appl Biosci 1992 8 275 282 1633570
Guindon S Gascuel O A simple, fast, and accurate algorithm to estimate large phylogenies by maximum likelihood Syst Biol 2003 52 696 704 14530136 10.1080/10635150390235520
| 15927057 | PMC1175084 | CC BY | 2021-01-04 16:37:17 | no | BMC Evol Biol. 2005 May 31; 5:34 | utf-8 | BMC Evol Biol | 2,005 | 10.1186/1471-2148-5-34 | oa_comm |
==== Front
BMC GenetBMC Genetics1471-2156BioMed Central London 1471-2156-6-391598515810.1186/1471-2156-6-39Research ArticleThe polymorphic nature of the human dopamine D4 receptor gene: A comparative analysis of known variants and a novel 27 bp deletion in the promoter region Szantai E [email protected] R [email protected] M [email protected] A [email protected] Z [email protected] Department of Medical Chemistry, Molecular Biology and Pathobiochemistry, Semmelweis University, Budapest, Hungary2 Department of Genetics, Eotvos Lorand University, Budapest, Hungary3 Marie Curie Chair of the EC, Horvath Lab. of Bioseparation Science, Institute of Analytical Chemistry and Radiochemistry, Leopold-Franzens University Innsbruck, Austria2005 28 6 2005 6 39 39 14 2 2005 28 6 2005 Copyright © 2005 Szantai et al; licensee BioMed Central Ltd.2005Szantai 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 human dopamine D4 receptor (DRD4) is a candidate gene of great interest in molecular studies of human personality and psychiatric disorders. This gene is unique in having an exceptionally high amount of polymorphic sites both in the coding and in the promoter region.
Results
We report the identification of a new 27 bp deletion starting 524 bp upstream of the initiation codon (27 bp del) of the dopamine D4 receptor (DRD4) gene, in the close vicinity of the -521C>T SNP. The presence of the 27 bp deletion leads to the misgenotyping of the -616C>G SNP by the Sau96 I RFLP method, thus the genotype determination of the mutation is of additional importance. The frequency of this novel sequence variation is considerably low (allele frequency is = 0.16%), as no homozygotes, and only 3 heterozygote carriers were found in a healthy, unrelated Caucasian sample (N = 955).
Conclusion
Remarkably, the deleted region contains consensus sequences of binding sites for several known transcription factors, suggesting that the different alleles may affect the transcriptional regulation of the gene. A comparison of methods and results for the allelic variations of the DRD4 gene in various ethnic groups is also discussed, which has a high impact in psychiatric genetic studies.
==== Body
Background
The dopaminergic system has received a significant amount of attention due to the important role it plays in the central nervous system in motor control, cognition, reward, emotion and endocrine regulation [1]. Recently, the D4 dopamine receptor (DRD4) gene [GenBank:U95122, GenBank:L12397] has been the target of psychogenetic studies mainly because it possesses a high number of polymorphisms [2], presumably as a consequence of its subtelomeric chromosomal localization (11p15.5) [3,4]. Polymorphisms in the DRD4 gene have received particular attention in the past decade because of their possible role in mental disorders [5-8], substance abuse [9-11] and the normal variations of human personality [12-16].
The gene and its regulatory region contain several length variations and numerous single nucleotide polymorphisms (SNPs). Much interest has been focused on these sequence variants in the investigation of complex neurobehavioral disorders using association study methods.
However, results of the psychogenetic association studies are often controversial. There are several reasons for the lack of conclusive results, such as the dangers of population stratification and the small effect size of individual genes complicated by oligogenic interactions [17]. Failure to confirm associations determined by various laboratories is partially explained by differences in demographic and ethnic structure of the subsequent studies, as a consequence of the diversity in allele frequencies of several polymorphisms among different populations. Moreover, differences in phenotyping and genotyping methodology may also obscure the weak effect of a gene variant on the investigated trait.
The human dopamine D4 receptor gene, located proximal to the Harvey-RAS oncogene locus and distal to the tyrosine hydroxylase locus [18], consists of 4 exons. The third exon contains a polymorphic 48 bp repeat region, the repeat number varies between 2 and 11. This part of the gene encodes the putative third cytoplasmic loop, and thus the polymorphism changes the length of the receptor protein [19]. Given that the variable number of 48 bp repeat (VNTR) is in a region that couples to G proteins and mediates postsynaptic effects [20], association studies have generated considerable interest. The DRD4 VNTR has been widely studied since a striking association was described between the seven-repeat allele and the human personality trait of increased novelty seeking [12,13], although the replications of the initial findings have been controversial [9,21]. So far evidence suggests that there is an association between the DRD4 7 repeat allele and attention deficit hyperactivity disorder (ADHD) also, although the effect size is small [6].
Further polymorphisms of the coding region of the DRD4 gene were also identified, [5,22-24], moreover additional mutations and polymorphisms have been described just recently in the introns as well [25-27].
Analysis of the 5' upstream region of DRD4 gene revealed that the promoter region is probably located in the region between -591 and -123 relative to the initiation codon, with a negative modulator between the -770 and -679 positions [28]. This region of the DRD4 gene is also astonishingly abundant in polymorphisms (see Fig. 4), including numerous SNPs [29,30], and a 120 bp tandem duplication [31].
Figure 4 Location of the 27 bp deletion and other polymorphic sites in the promoter region of the human DRD4 gene. The deleted region starts 524 bp upstream of the initiation codon. Figure is drawn to scale and each position is shown relative to the first nucleotide of the initiation codon indicated as +1. A cell type-specific promoter region between -591 and -123 contains the promoter of the gene, and the negative modulator is between the -770 and -679 positions (grey boxes). Open boxes indicate length variants at the DRD4 promoter region. Arrowheads specify the positions of SNPs.
The -521C>T SNP is located on a CpG island at one end of a cell-type specific regulatory element [28]. The -521C allele was shown to be 40% more active than the T allele in an in vitro transient expression system [32]. Moreover, a significantly higher incidence of the CC genotype was found among schizophrenics [32]. The SNP has also been associated with novelty seeking behavior in separate studies of Japanese, Caucasian and Afro-American samples [16,33,34], but negative results were also described [35-37]. Meta-analyses of existing studies have been conducted to provide statistical measures of the small association between -521 C>T SNP genotypes and novelty seeking [38].
A regularly investigated polymorphism in the promoter sequence of the DRD4 gene is the -616C>G SNP that may result in the gain of an AP-2 transcription factor binding site [39] and thereby affects the expression characteristics of the receptor. Associations of the -616C>G SNP with ADHD [39,40], schizophrenia [41] and personality dimensions [33] were studied but did not yield consequent results. Barr et al.[39] added the -616C>G SNP to the investigations of the 5' upstream region by studying the haplotype transmissions of three polymorphisms in the promoter region and the 48 bp VNTR in exon III as genetic risk factors for ADHD. An association between ADHD and the -616C>G SNP was reinforced recently [40].
An additional length polymorphism of the 5' untranslated region of the DRD4 gene has been reported as common in the population. The polymorphism consists of a 120 bp tandem duplication 1.2 kb upstream from the initiation codon [31] and gives rise to the Pst I RFLP previously reported [42]. Association studies between the 120 bp duplication and the occurrence of ADHD hypothesized the potential significance of this region in the regulation of transcription [43]. Experimental evidence suggests enhanced binding capacity of Sp1 transcription element to the duplicated form [44]. The duplication is abundant in the human race, although allele frequencies do vary among different ethnic groups (see Discussion).
Identification of new, perhaps functional length variants is especially intriguing because the up-regulation of D4 receptors may be more significant than any changes in the protein structure [45,46]. In previous publications the density of DRD4 was found to be six fold elevated in the brains of schizophrenic patients [47], and the DRD4 mRNA was elevated in the frontal cortex of schizophrenics in post mortem studies in comparison with controls [47], pointing to possible allelic variants that influence the transcription levels of the DRD4 gene. The -521C>T SNP has been shown to affect transcriptional activity in an in vitro transient expression system [32]. In the present study we characterized a new deletion mutation in the promoter region of the DRD4 gene, which is a promising target of future association studies. This sequence variation has been discovered during the large-scale application of our recently published Sau96 I RFLP genotyping method of the -616C>G SNP [30]. A novel technique was developed for the screening of this mutation, and applied on a Caucasian (Hungarian) population of 959 individuals.
Results
Recently, we described a novel genotyping method for the analysis of the -616C>G SNP in the DRD4 gene [GenBank:L12397] promoter region [GenBank:U95122], consisting of the Sau96 I RFLP and an allele specific amplification procedure. The parallel application of the two independent methods (referred to as double genotyping) highly increases the reliability of the study, especially when investigating outstandingly polymorphic regions such as the DRD4 gene [30]. During the large-scale genotyping of the -616C>G SNP by this newly described double-genotyping method, we noticed some abnormalities in the size of the produced fragments. Fig. (1) unmistakably shows the two different size fragments obtained from the Sau96 I RFLP where the 207 bp long fragment represents the -616 C allele and the 172 bp long product refers to the cleaved -616 G allele. Surprisingly, in lane 5, an unexpected short fragment was produced (indicated by an asterisk) in case of a sample having a -616 GG genotype determined by allele-specific PCR [30] (not shown). The size of the indefinite fragment was approximately 30 bp less than that of the expected 172 bp long PCR product of the -616 G allele. The appearance of the unusual PCR fragment has two feasible explanations: (1) There might be a new SNP in the amplified fragments of the sample resulting in an additional polymorphic Sau96 I restriction site, (2) or a length variation in one of the alleles of the individual might also be responsible for the production of the short fragment. Electrophoretic analysis of the undigested PCR-products yielded two distinct bands demonstrating that the new sequence variant was a deletion.
Figure 1 Sau96 I RFLP genotyping method of the -616C>G SNP reveals a new variation in the DRD4 gene. Representative genotyping results are depicted. Conditions of PCR amplification and electrophoretic separation were carried out as previously described (see Figure 2. in [30]). The asterisk indicates the short PCR fragment containing the deletion.
When investigating the sample containing the deletion by allele-specific amplification no abnormalities were found (not shown). This can be readily explained focusing on the setup of this method, which applies a sense C-specific and an antisense G-specific primer for the simultaneous amplification of two PCR-products with different length corresponding to the two alleles respectively. Consequently, if the new sequence variant is localized upstream from the -616C>G SNP on the same chromosome as the -616 C allele or downstream from the SNP together with the G allele, then the deleted region is not amplified thus the mutation does not show up on the electropherogram. To confirm our notion we carried out the allele-specific PCR of this sample applying primers with opposite orientation (i.e. a sense G-and an antisense C-specific primer). In this arrangement we detected a shortened -616 G specific PCR product, suggesting that the deletion is localized downstream from the -616C>G SNP composing a haplotype with -616 G variant in this sample. Moreover, it is important to note that the deletion does not cause any uncertainty if the genotyping of the -616C>G SNP is carried out by allele specific amplification, as the size difference between the two allele-specific products is fairly big (252 bp) excluding the possibility of misgenotyping.
Sequencing of the shortened fragment of the Sau96 I RFLP (referred to as the product of the "Del" allele) also confirmed the presence of the deletion in this sample. Fig. (2) shows a sequence containing the deletion aligned to that obtained from the GeneBank : "Human dopamine D4 receptor gene, 5' flanking region" (accession number: U95122) and "Homo sapiens Dopamine D4 receptor (DRD4) gene" (accession number: L12397). It is unambiguous that the deletion affects 27 base-pairs, however its position cannot be marked explicitly since it is located between two GGAG sequences and it is hard to agree on which GGAG is enclosed by the deleted section. Therefore the mutation can be placed anywhere between the -524th and the -554th positions relative to the initiation codon. According to the nomenclature of the Human Genome Variation Society we suggest the -524 position to be the assigned start point of the deletion, since this is the closest point to the coding sequence of the DRD4 gene.
Figure 2 Alignment of the deleted allele string with the published DNA sequence of the upstream region [GenBank:U95122, GenBank:L12397.] Positions of the -616C>G, -615A>G, -603T>G, -602G>del and -521C>T SNPs are specified. The bold letters indicate the ambiguous localization of the deletion, however according to the nomenclature of the Human Genome Variation Society the suggested start point is the -524 position.
PCR primers flanking the region of the mutation were designed (see "Methods") to determine the genotype of the 27 bp deletion in our population of unrelated Caucasians. Fig. (3) shows the electrophoretic separation of the obtained PCR amplicons in the analysis of 5 samples. A longer 391 bp (designated "Non-del") and a shorter 364 bp (referred to as "Del") fragment was formed in lane 3 representing a heterozygote for the 27 bp deletion ("Del"/"Non-del"). The rest of the individuals (lanes 1, 2, 4 and 5) do not contain the deletion: only the longer fragment was formed showing a homozygote "Non-del"/"Non-del" genotype. In order to determine the allele frequency of this new sequence variation, we screened for the Del allele in a large (N = 955) healthy, unrelated Hungarian (Caucasian) population. Seeing that our results show a very low allele frequency of 0.16% for the Del allele of the DRD4 promoter region, this new variant can be considered as a mutation rather than a polymorphism. The Mendelian inheritance of the novel 27 bp deletion was also investigated by a three-generation family analysis, where one heterozygote was found in the first and one in the second generation, respectively. We did not find any homozygotes in our sample population, or among the available relatives (N = 4) of the heterozygote individuals.
Figure 3 Genotyping the 27 bp deletion mutation in healthy Hungarian individuals. Amplification conditions: reaction mixtures contained 200 μM dATP, dCTP, dTTP, and 100 μM dGTP and dITP; 1 μM of each primer (see Methods), 1 ng DNA template, 0.25 U DNA polymerase, 1x reaction buffer, and 1x Q solution in a total volume of 10 μL. Thermocycling conditions: 95°C for 15 min, followed by 35 cycles of 94°C for 1 min; 65°C for 30 sec; 72°C for 1 min, finished by 72°C for 10 min. Electrophoresis: fine resolution agarose gel in 40 mM Tris, 10 mM EDTA pH 8.0 buffer solution at room temperature, 6.6 V/cm electric field, 90 min, followed by 1 μg/mL ethidium bromide staining. The faint diffuse bands in lane 3 migrating slower than the representative fragments presumably contain heteroduplex molecules.
To assign the functional role of the 27 bp deletion we carried out an "in silico" transcription factor binding analysis. The Transcription Element Search System (TESS; ) was used to specify the transcription factors that can play a role in the gene expression regulation by binding to this region. Even when applying stringent search parameters (6 bases as "Minimum string length" and allowing no mismatch in the sequence) the following human transcription factors were suggested to bind the region of investigation: NF-E2, AP-2, Sp1. Although the physiological importance of this result needs to be further studied, the presence of the Sp1 site is especially notable, as a zinc finger type transcription factor (dopamine receptor regulating factor, DRRF) was described to effectively bind to the same sequences as Sp1 in the dopamine receptor promoters [48].
Discussion
Here we described a novel 27 bp deletion in the DRD4 promoter region, which has several important aspects. Primarily, it is a source of misgenotyping of the -616C>G SNP when using the standard genotyping method of Sau96 I RFLP. As this SNP is often studied in psychiatric genetics finding the appropriate genotyping methodology is of great importance [39,40]. The problem arises from the hardly detectable difference between the 180 bp amplicon of the 27 bp Del and the -616 C allele-combination (i.e. haplotype) vs. the 172 bp product of the -616 G variant after Sau96 I digestion. Thus, the existence of such a deletion is a source of error, since some of the -616 C alleles might be misgenotyped as -616 G in the presence of the deletion.
On the other hand, using the previously reported [30] allele-specific amplification, it is possible to distinguish between the -616C and G alleles with no doubt, even in the presence of the deletion. We can explain the results assuming that the four samples heterozygous for the deletion contained the deleted allele on the same chromosome as the -616G allele. This assumption was later confirmed by allele specific sequencing. As the -616C-specific primer was the sense and the -616G-specific primer was the antisense primer in our system, the obtained products did not contain the region of the deletion. When applying a -616 G-specific sense primer in another setup of allele specific amplification, the G-specific fragment clearly showed the presence of the 27 bp deletion (data not shown), but the length variation did not influence the accuracy of the -616C>G SNP genotyping after all.
Although the occurrence of the 27 bp deletion is rather low (0.16%) in our healthy Caucasian population, the allele frequency might be different in other ethnic groups and various clinical samples. Therefore, further studies are necessary to gather information on the allele and genotype frequencies of this novel length variant in a range of other populations, and these results should be considered in the psychogenetic association studies of the -616C>G polymorphism as well.
It is of additional importance to perform association studies in different populations since each may represent distinct environments that could possibly interact with genetic variants. The methodological conditions and demographic setting is crucial in evaluating the strength of a relationship and in replicating an association found by different investigations. Here we address the differences in ethnic structure and methodological approach used in the successive studies of commonly investigated DRD4 gene variations and review the diversity of their allele frequencies in different populations around the globe.
The variable number of a 48 bp repeat (48 bp VNTR) in exon III makes the DRD4 gene one of the most frequently studied genes in psychiatric genetics. The "long form" (7R) of this length polymorphism was found to be associated with the personality trait of "novelty seeking" [12,13], children's attention deficit hyperactivity syndrome [8] and drug abuse [10]. In the Hungarian population (as well as in other Caucasian samples) the 4 repeat allele occurs most frequently (65.0%), followed by the 7 repeat (19.5%) and the 2 repeat (8,9%) variations. The rare alleles include the 3 repeat (3.8%), 8 repeat (1.2%), 5 repeat (1.1%) and the 6 repeat (0.4%) variants, whereas no 9 and 11 repeat allele was found (see Table 1). We have identified one 10 repeat allele (0.1%) in our large (2N = 1196) Caucasian sample. This form has only been reported previously in the African population with 1% incidence [49]. In numerous studies, genotype groups were defined only by absence vs presence of the DRD4 exon III 7 repeat, these results are not included in the table.
Table 1 Allele frequency data for the 48 bp VNTR polymorphism of DRD4
Origin of populations Allele frequency (%)
2N 2 3 4 5 6 7 8 10
Caucasian Hungarian (Ronai et al., 2000) 1196 8.9 3.8 65.0 1.1 0.4 19.5 1.2 0.1
European-mixed (Chang et al., 1996) 176 12.0 6.0 57.0 2.0 1.0 21.0 1.0 0.0
Caucasians in New Zealand (Mill et al., 2002) 1760 8.8 4.6 65.0 0.9 0.6 19.4 0.6 0.0
German (Strobel et al., 2003) 230 7.3 4.3 66.5 0.9 0.4 19.1 1.3 0.0
German (Franke et al., 2000) 394 7.1 3.3 70.1 3.3 0.5 15.2 0.5 0.0
Italian (Mochi et al., 2003) 212 10.4 3.3 68.9 2.8 0.9 13.2 0.5 0.0
Italian (DeLuca et al., 2003) 190 11.0 5.3 67.9 0.5 0.5 13.7 1.0 0.0
Mean 8.9 4.3 65.6 1.3 0.5 18.5 0.8 0.0
Standard deviation 1.1 0.6 2.5 0.8 0.2 2.1 0.3 0.0
Asian Chinese (Li et al., 2000) 608 18.8 1.2 76.6 1.3 2.1 0.0 0.0 0.0
Japanese (Ishiguro et al., 2000) 680 12.0 1.0 81.0 4.0 1.0 1.0 0.0 0.0
Mean 15.2 1.1 78.9 2.7 1.5 0.5 0.0 0.0
Standard deviation 3.4 0.1 2.2 1.3 0.5 0.5 0.0 0.0
African Falashan (Chang et al., 1996) 128 3.0 0.0 83.0 0.0 2.0 11.0 0.0 1.0
American Mayans (Chang et al., 1996) 100 1.0 0.0 57.0 0.0 3.0 39.0 0.0 0.0
The prevalence of the most common alleles in the Hungarian population studied matches the frequencies measured by Mill et al. (New Zeland [50]) and Strobel et al. (Germany [51]). Minor differences of the 7 repeat frequencies incidence among Caucasians might originate from population stratification. If, however, allele frequencies of similar populations are quite different this problem might also originate from preferential amplification of shorter PCR products [52] as a methodological artifact in this highly GC rich region, resulting in an underestimation of the 7 repeat allele frequencies in the above populations.
Noticeably, there is considerable variation in the distribution of the alleles as we compare Caucasian and Asian populations. The occurrence of the 7 repeat allele is extremely low (0.5%) in Chinese and Japanese samples, and relatively high in North American Mayans (39%), whereas Caucasians have an intermediate (18.5%) mean prevalence. There are two theories describing the origin of the 7 repeat allele [53,54], however independently on the reason for the varying frequency data of this variation, a high danger of artifacts caused by population admixture should definitely be taken into account when genetic association studies are being performed [55].
Beside the length polymorphism of the DRD4 gene coding sequence, the 5' upstream region of this gene has numerous sequence variations also. Among the promoter polymorphisms, the -521C>T SNP has been widely studied since it was shown to be associated with the personality trait of novelty seeking [16,34]. Frequency data obtained for the -521C>T SNP seem to be considerably constant among the populations (see Tab. 2), as frequency differences between the ethnic groups are smaller than the variations observed within the same study group. Therefore, the alterations in the frequency between the C vs. T alleles can be readily explained by the differences in methodology and sample size. One of the possible sources of technical difficulties might originate from using the PCR-RFLP protocol designed by Okuyama et al. [32], where the 3' end of the reverse primer anneals to the polymorphic -603rd position [41], -603T>del), which might result in unequal amplification of the homologous chromosomes in heterozygotes. Therefore the position of the primers were changed in other protocols, including the previously described PCR-RFLP system, where an internal control restriction site was added to the system [16]. In order to enhance the validity and reliability, two parallel methods were used for genotyping the -521C>T SNP in our laboratory, a PCR-RFLP and a fast allele-specific protocol [56]. Using a large sample (N = 598), we found a somewhat lower (46.5%) allele frequency for the -521 C allele compared to the -521 T allele and the genotype frequencies corresponded to the Hardy-Weinberg equilibrium (see Table 2., line 1).
Table 2 Genotypes and allele frequencies of the -521 CT SNP in different populations
Origin of population Allele frequency (%) Genotype frequency (%)
2N C T N CC CT TT
Caucasian Hungarian (Szantai et al., 2005) 1196 46.5 53.5 598 21.4 50.0 28.6
European-mixed (Barr et al., 2001) 308 45.8 54.2 154
German (Strobel et al., 2003) 230 40.0 60.0 115 20.9 38.3 40.8
Swedish (Jönsson et al., 2001) 776 42.0 58.0 388 15.0 54.0 31.0
mean 44.4 55.6 19.1 50.2 30.7
Standard deviation 2.4 2.4 3.0 4.5 3.6
Asian Japanese (Okuyama et al., 1999) 538 41.0 59.0 269 14.0 53.0 33.0
Japanese (Ishiguro et al., 2000) 538 41.0 60.0 269 14.0 53.0 33.0
Japanese (Mitsuyasu et al., 1999) 294 41.0 59.0 147 17.9 46.3 35.8
Japanese (Okuyama et al., 2000) 172 53.5 46.5 86 32.5 41.9 25.6
Chinese (Xing et al., 2003) 412 39.1 60.9 206 12.1 53.9 34.0
Chinese (Li et al., 2000) 422 40.3 59.7 211 27.5 25.6 46.9
Korean (Lee et al., 2003) 202 46.0 54.0 101
mean 41.8 58.4 17.9 46.7 35.5
Standard deviation 3.5 3.6 6.7 10.4 5.8
American Afro-American (Bookman et al., 2002) 142 55.0 45.0 71 32.0 42.0 26.0
Although the literature of the -616C>G SNP is limited, the available genotype frequencies do vary among different ethnic groups significantly as shown in Tab. (3). The frequency of the two alleles were quite similar (-616C: 51.5%, -616G: 48.5%) in the Hungarian population studied, whereas in Asian and Afro-American populations a higher occurrence of the -616G allele (69.7% and 71.8% respectively) was observed. It is difficult to account for the significant difference between the mixed-European and the Hungarian samples. According to our recent results [30], the Ava II RFLP genotyping protocol [29] overestimated the -616C allele in the presence of the -615G allele of a novel SNP in this position. Sau96 I, in place of the Ava II restriction enzyme, was shown to be the appropriate choice of restriction endonuclease [30], as the newly described -615A>G SNP had no influence on the digestion reaction by Sau96 I. This is another example of how obtained allele frequencies depend on differences in methodology, providing a source of unreplicated association studies.
Table 3 Allele and genotype frequencies of the -616 CG SNP in 5 populations
Origin of population Allele frequency (%) Genotype frequency (%)
2N C G N CC CG GG
Caucasian Hungarian (Szantai et al., 2005) 1196 51.5 48.5 598 27.8 47.5 24.7
European-mixed (Barr et al., 2001) 308 72.7 27.3 154
mean 55.8 44.2 31.1* 49.3* 19.5*
Standard deviation 8.6 8.6
Asian Japanese (Mitsuyasu et al., 1999) 160 28.4 71.6 80 14.9 26.9 58.2
Chinese (Xing et al., 2003) 412 31.1 68.9 206 6.3 49.5 44.2
mean 30.3 69.7 8.7 43.2 48.1
Standard deviation 1.2 1.2 3.9 10.1 6.3
American Afro-American (Bookman et al., 2002) 142 28.2 71.8 71 13.0 31.0 56.0
* Expected genotype frequencies were calculated from measured allele frequencies assuming Hardy-Weinberg equilibrium in the populations.
Recently, significant amount of interest has been focused on a polymorphic tandem repeat element located 1.2 kb upstream of the initiation codon in the DRD4 promoter. The 120 bp of duplicated sequence was linked to ADHD and might be involved in the regulation of transcription [43,57]. Allele frequency variation is high among populations, in Europeans the duplicated form is more common than in other populations (see Tab. 4). Since the region of this length variation is not GC rich and it is further away from the polymorphic hotspot of the DRD4 promoter, no significant genotyping obstacles have been found.
Table 4 Frequency distribution for the 120 bp duplicated alleles of the DRD4 promoter
Origin of population Allele frequency (%) Genotype frequency (%)
2N 1 2 N 1/1 1/2 2/2
Caucasian Hungarian (Szantai et al., 2005) 1196 17.1 82.9 598 3.0 28.3 68.7
European-mixed (Seaman et al., 1999) 174 19.5 80.5
European-mixed (Barr et al., 2001) 308 19.9 80.1
mean 17.9 82.1 3.2* 29.4* 67.4*
Standard deviation 1.5 1.5
Asian Chinese (Xing et al., 2003) 412 37.4 62.6 206 13.6 47.6 38.8
Chinese (Seaman et al., 1999) 122 36.1 63.9
mean 37.1 62.9 13.8* 46.6* 39.6*
Standard deviation 0.5 0.5
African African (Seaman et al., 1999) 136 59.6 40.4 35.5* 48.2* 16.3*
American Mayan (Seaman et al., 1999) 106 48.1 51.9 23.1* 49.9* 26.9*
* Expected genotype frequencies were calculated from measured allele frequencies assuming Hardy-Weinberg equilibrium in the populations.
Fig. (4) summarizes the polymorphisms highlighted by the literature, and the number of the variants keeps increasing. Some of the recently described polymorphisms such as the -1106C>T, -906C>T [27], -615A>G [30] SNPs and the novel 27 bp deletion, described here have not been characterized yet more thoroughly. The 27 bp deletion can be rapidly genotyped with the assay described in this paper and could be a component of further association studies that analyze the 5' region of the DRD4 gene. Additionally, the dbSNP database of the NCBI contains numerous further SNPs in the non-coding region of the gene (-872A>G, -844C>G, -764A>C, -754C>G, -713C>G, -599C>G, -528C>T, -364A>G) that haven't even been published yet. Theoretical approaches have shown that highly variable polymorphic markers are more useful in association analysis than less polymorphic ones, since there is a higher probability to identify allele frequency differences between cases and controls [58]. Similarly, analysis of haplotypes, involving several polymorphic sites, might provide a greater power for association analysis. Haplotypes of the 120 bp duplication [31] and some of the SNPs [23,29] published earlier have already been the targets of numerous genetic association studies. Recently, direct haplotype detection methods for the commonly investigated -521C>T and -616C>G SNPs as well as the 120 bp duplication promoter polymorphisms were developed in our laboratory [59,60]. Linkage disequilibrium analysis of polymorphisms marks the different haplotype blocks on the chromosome segment giving way to future association studies [61].
The polymorphic variations in the regulatory region may directly influence the regulation of transcription of the DRD4 gene. It was demonstrated that some of these polymorphisms have functional effects resulting in different transcriptional activity [32,57]. Furthermore, the 120 bp duplication and the C to G change at the -616th position cause the gain of additional binding sites of known transcription elements [31,39,62]. Accordingly, we examined the deleted sequence for any potential transcriptional binding sites and found the consensus sequences of several factors, including Sp1, AP-2alphaB and NF-E2. It has been reported in a human retinoblastoma cell line that the -521C>T SNP has significant influence on the transcriptional efficiency of the DRD4 gene suggesting the relevance of a single SNP in dopaminergic neurotransmission [32]. As the 27 bp deletion lies in the very same region, it can be assumed that this variation might also have a considerable impact on the transcriptional activity of the gene.
Methods
Participants
DNA was extracted from epithelial cells of 959 healthy subjects (396 male and 563 female) in a Caucasian sample of Hungarian origin. Signed informed consent was obtained from all the participants. The research protocol was approved by the Research Ethics Committee.
Non-invasive DNA sampling
Buccal cells were collected by cotton swabs from the inner surface of the mouth [63]. DNA was isolated by phenol extraction and alcohol precipitation as described earlier [64].
Genotyping protocol for the 27 bp del
The Qiagen® HotStarTaq™ DNA polymerase kit was used for polymerase chain reaction (PCR). Reaction mixtures contained 200 μM dATP, dCTP, dTTP, and 100 μM dGTP and dITP; 1 μM of forward primer (5'-GGA ATG GAG GAG GGA GCG GG-3'), and 1 μM of reverse primer (5'-GAC GCC AGC GCC ATC CTA CC-3'), approximately 1 ng DNA template, 0.25 U DNA polymerase, 1x reaction buffer, and 1x Q solution in a total volume of 10 μL. Thermocycling was initiated at 95°C for 15 minutes to activate the hot start enzyme and to denature genomic DNA, which was followed by 35 cycles of 1 min denaturation at 94°C, 30 sec annealing at 65°C and 1 min extension at 72°C. A final 10 min extension step at 72°C was followed by cooling the samples to 8°C. PCR products were analyzed by conventional submarine horizontal agarose gel-electrophoresis (MidiGel of Biocenter, Szeged, Hungary). A composite gel was applied containing 1.5% low EEO agarose Type I (Sigma Chemicals) and 2% Methaphor agarose "fine resolution" (Cambrex Corporation). Separation was performed at room temperature in 40 mM Tris, 10 mM EDTA.Na2, 1% acetic acid (pH 8.0) buffer and 6.6 V/cm (100 V field strength, 26 mA) for 90 min. After separation, dsDNA fragments were stained in 1 μg/mL ethidium bromide solution at room temperature for 15 minutes. No destaining was necessary. A BioRad Gel-Doc 1000 gel-documentation system (Hercules) was used for visualization of the DNA fragments.
DNA sequencing
Genomic DNA of individuals shown to carry the deleted allele were sequenced by amplification of the region in question using the above described PCR conditions. Polymerase chain reaction was carried out by the forward primer: 5'-GGA ATG GAG GAG GGA GCG GG-3', and reverse primer: 5'-CGC TCC ACC GTG AGC CCA GTA T-3'. The shorter PCR product (deleted allele) was purified from the gel using the Qiagen QIAquick Spin DNA-Extraction Kit. DNA sequencing was performed by an ABI 370 Sequencer, using the reverse primer (see above) of the amplification reaction.
Database searching
The 27 bp of sequence, which is deleted in the mutant allele, was searched for known transcription factor binding sites using the Transcription Element Search Software (TESS) web-based search tool to search the TransFac database; [65].
Authors' contributions
ES carried out the genotyping experiments of the 959 individuals, took part in the elaboration of the novel genotyping procedure and helped to draft the manuscript. RS analyzed the results of the initial experiments suggesting the presence of the novel mutation and drafted the manuscript. MS conceived of the study and participated in its design and coordination. AG carried out the database search and coordinated the experimental work. ZR designed the primers for the investigation of the 27 bp del and optimized the genotyping protocol, analyzed the sequencing results and helped to finalize the manuscript. All authors read and approved the final manuscript.
Acknowledgements
We are grateful to O. Kiraly, M. Szoke and Z. Csapo for valuable discussions and help with sequencing the deleted allele, and to G. Kolmann for her technical assistance. This work was supported by Hungarian grants NKFP 0008/2002, GVOP AKF 311 2004 05 0324_3.0 and OTKA F42730. E. Szantai is indebted to L. Orosz, the head of her Ph.D. Program (Classical and Molecular Genetics) at Eötvös Loránd University, Budapest.
==== Refs
Jaber M Robinson SW Missale C Caron MG Dopamine receptors and brain function Neuropharmacology 1996 35 1503 1519 9025098 10.1016/S0028-3908(96)00100-1
Oak JN Oldenhof J Van Tol HH The dopamine D(4) receptor: one decade of research Eur J Pharmacol 2000 405 303 327 11033337 10.1016/S0014-2999(00)00562-8
Gelernter J Kennedy JL Van Tol HH Civelli O Kidd KK The D4 dopamine receptor (DRD4) maps to distal 11p close to HRAS Genomics 1992 13 208 210 1349574 10.1016/0888-7543(92)90222-E
Petronis A Van Tol HH Lichter JB Livak KJ Kennedy JL The D4 dopamine receptor gene maps on 11p proximal to HRAS Genomics 1993 18 161 163 8276407 10.1006/geno.1993.1445
Catalano M Nobile M Novelli E Nothen MM Smeraldi E Distribution of a novel mutation in the first exon of the human dopamine D4 receptor gene in psychotic patients Biol Psychiatry 1993 34 459 464 8268330 10.1016/0006-3223(93)90236-7
Faraone SV Doyle AE Mick E Biederman J Meta-analysis of the association between the 7-repeat allele of the dopamine D(4) receptor gene and attention deficit hyperactivity disorder Am J Psychiatry 2001 158 1052 1057 11431226 10.1176/appi.ajp.158.7.1052
Grice DE Leckman JF Pauls DL Kurlan R Kidd KK Pakstis AJ Chang FM Buxbaum JD Cohen DJ Gelernter J Linkage disequilibrium between an allele at the dopamine D4 receptor locus and Tourette syndrome, by the transmission-disequilibrium test Am J Hum Genet 1996 59 644 652 8751866
LaHoste GJ Swanson JM Wigal SB Glabe C Wigal T King N Kennedy JL Dopamine D4 receptor gene polymorphism is associated with attention deficit hyperactivity disorder Mol Psychiatry 1996 1 121 124 9118321
Comings DE Gonzalez N Wu S Gade R Muhleman D Saucier G Johnson P Verde R Rosenthal RJ Lesieur HR Rugle LJ Miller WB MacMurray JP Studies of the 48 bp repeat polymorphism of the DRD4 gene in impulsive, compulsive, addictive behaviors: Tourette syndrome, ADHD, pathological gambling, and substance abuse Am J Med Genet 1999 88 358 368 10402503 10.1002/(SICI)1096-8628(19990820)88:4<358::AID-AJMG13>3.0.CO;2-G
Kotler M Cohen H Segman R Gritsenko I Nemanov L Lerer B Kramer I Zer-Zion M Kletz I Ebstein RP Excess dopamine D4 receptor (D4DR) exon III seven repeat allele in opioid-dependent subjects Mol Psychiatry 1997 2 251 254 9152990 10.1038/sj.mp.4000248
Muramatsu T Higuchi S Murayama M Matsushita S Hayashida M Association between alcoholism and the dopamine D4 receptor gene J Med Genet 1996 33 113 115 8929946
Benjamin J Li L Patterson C Greenberg BD Murphy DL Hamer DH Population and familial association between the D4 dopamine receptor gene and measures of Novelty Seeking Nat Genet 1996 12 81 84 8528258 10.1038/ng0196-81
Ebstein RP Novick O Umansky R Priel B Osher Y Blaine D Bennett ER Nemanov L Katz M Belmaker RH Dopamine D4 receptor (D4DR) exon III polymorphism associated with the human personality trait of Novelty Seeking Nat Genet 1996 12 78 80 8528256 10.1038/ng0196-78
Noble EP Ozkaragoz TZ Ritchie TL Zhang X Belin TR Sparkes RS D2 and D4 dopamine receptor polymorphisms and personality Am J Med Genet 1998 81 257 267 9603615 10.1002/(SICI)1096-8628(19980508)81:3<257::AID-AJMG10>3.0.CO;2-E
Rogers G Joyce P Mulder R Sellman D Miller A Allington M Olds R Wells E Kennedy M Association of a duplicated repeat polymorphism in the 5'-untranslated region of the DRD4 gene with novelty seeking Am J Med Genet 2004 126B 95 98 10.1002/ajmg.b.20133
Ronai Z Szekely A Nemoda Z Lakatos K Gervai J Staub M Sasvari-Szekely M Association between Novelty Seeking and the -521 C/T polymorphism in the promoter region of the DRD4 gene Mol Psychiatry 2001 6 35 38 11244482 10.1038/sj.mp.4000832
Ebstein RP Benjamin J Belmaker RH Personality and polymorphisms of genes involved in aminergic neurotransmission Eur J Pharmacol 2000 410 205 214 11134670 10.1016/S0014-2999(00)00852-9
Kennedy JL Sidenberg DG Van Tol HH Kidd KK A HincII RFLP in the human D4 dopamine receptor locus (DRD4) Nucleic Acids Res 1991 19 5801 1682888
Van Tol HH Wu CM Guan HC Ohara K Bunzow JR Civelli O Kennedy J Seeman P Niznik HB Jovanovic V Multiple dopamine D4 receptor variants in the human population Nature 1992 358 149 152 1319557 10.1038/358149a0
Asghari V Sanyal S Buchwaldt S Paterson A Jovanovic V Van Tol HH Modulation of intracellular cyclic AMP levels by different human dopamine D4 receptor variants J Neurochem 1995 65 1157 1165 7643093
Kluger AN Siegfried Z Ebstein RP A meta-analysis of the association between DRD4 polymorphism and novelty seeking Mol Psychiatry 2002 7 712 717 12192615 10.1038/sj.mp.4001082
Cichon S Nothen MM Catalano M Di Bella D Maier W Lichtermann D Minges J Albus M Borrmann M Franzek E . Identification of two novel polymorphisms and a rare deletion variant in the human dopamine D4 receptor gene Psychiatr Genet 1995 5 97 103 8746407
Nothen MM Cichon S Hemmer S Hebebrand J Remschmidt H Lehmkuhl G Poustka F Schmidt M Catalano M Fimmers R . Human dopamine D4 receptor gene: frequent occurrence of a null allele and observation of homozygosity Hum Mol Genet 1994 3 2207 2212 7881421
Seeman P Ulpian C Chouinard G Van Tol HH Dwosh H Lieberman JA Siminovitch K Liu IS Waye J Voruganti P . Dopamine D4 receptor variant, D4GLYCINE194, in Africans, but not in Caucasians: no association with schizophrenia Am J Med Genet 1994 54 384 390 7726213
Barr CL Kennedy JL Lichter JB Van Tol HH Wetterberg L Livak KJ Kidd KK Alleles at the dopamine D4 receptor locus do not contribute to the genetic susceptibility to schizophrenia in a large Swedish kindred Am J Med Genet 1993 48 218 222 8135305 10.1002/ajmg.1320480409
Shimada MK Inoue-Murayama M Ueda Y Maejima M Murayama Y Takenaka O Hayasaka I Ito S Polymorphism in the second intron of dopamine receptor D4 gene in humans and apes Biochem Biophys Res Commun 2004 316 1186 1190 15044110 10.1016/j.bbrc.2004.03.006
Wang E Ding YC Flodman P Kidd JR Kidd KK Grady DL Ryder OA Spence MA Swanson JM Moyzis RK The Genetic Architecture of Selection at the Human Dopamine Receptor D4 (DRD4) Gene Locus Am J Hum Genet 2004 74
Kamakura S Iwaki A Matsumoto M Fukumaki Y Cloning and characterization of the 5'-flanking region of the human dopamine D4 receptor gene Biochem Biophys Res Commun 1997 235 321 326 9199190 10.1006/bbrc.1997.6770
Mitsuyasu H Ozawa H Takeda Y Fukumaki Y Novel polymorphisms in the upstream region of the human dopamine D4 receptor (DRD4) gene J Hum Genet 1999 44 416 418 10570917 10.1007/s100380050191
Ronai Z Szantai E Szmola R Nemoda Z Szekely A Gervai J Guttman A Sasvari-Szekely M A novel A/G SNP in the -615th position of the dopamine D4 receptor promoter region as a source of misgenotyping of the -616 C/G SNP Am J Med Genet 2004 126B 74 78 10.1002/ajmg.b.20112
Seaman MI Fisher JB Chang F Kidd KK Tandem duplication polymorphism upstream of the dopamine D4 receptor gene (DRD4) Am J Med Genet 1999 88 705 709 10581493 10.1002/(SICI)1096-8628(19991215)88:6<705::AID-AJMG22>3.0.CO;2-F
Okuyama Y Ishiguro H Toru M Arinami T A genetic polymorphism in the promoter region of DRD4 associated with expression and schizophrenia Biochem Biophys Res Commun 1999 258 292 295 10329380 10.1006/bbrc.1999.0630
Bookman EB Taylor RE Adams-Campbell L Kittles RA DRD4 promoter SNPs and gender effects on Extraversion in African Americans Mol Psychiatry 2002 7 786 789 12192624 10.1038/sj.mp.4001075
Okuyama Y Ishiguro H Nankai M Shibuya H Watanabe A Arinami T Identification of a polymorphism in the promoter region of DRD4 associated with the human novelty seeking personality trait Mol Psychiatry 2000 5 64 69 10673770 10.1038/sj.mp.4000563
Ekelund J Suhonen J Jarvelin MR Peltonen L Lichtermann D No association of the -521 C/T polymorphism in the promoter of DRD4 with novelty seeking Mol Psychiatry 2001 6 618 619 11673788 10.1038/sj.mp.4000943
Jonsson EG Ivo R Gustavsson JP Geijer T Forslund K Mattila-Evenden M Rylander G Cichon S Propping P Bergman H sberg M Nothen MM No association between dopamine D4 receptor gene variants and novelty seeking Mol Psychiatry 2002 7 18 20 11803441 10.1038/sj.mp.4001950
Strobel A Lesch KP Hohenberger K Jatzke S Gutzeit HO Anacker K Brocke B No association between dopamine D4 receptor gene exon III and -521C/T polymorphism and novelty seeking Mol Psychiatry 2002 7 537 538 12140774 10.1038/sj.mp.4001027
Schinka JA Letsch EA Crawford FC DRD4 and novelty seeking: results of meta-analyses Am J Med Genet 2002 114 643 648 12210280 10.1002/ajmg.10649
Barr CL Feng Y Wigg KG Schachar R Tannock R Roberts W Malone M Kennedy JL 5'-untranslated region of the dopamine D4 receptor gene and attention-deficit hyperactivity disorder Am J Med Genet 2001 105 84 90 11425008 10.1002/1096-8628(20010108)105:1<84::AID-AJMG1068>3.0.CO;2-Q
Lowe N Kirley A Mullins C Fitzgerald M Gill M Hawi Z Multiple marker analysis at the promoter region of the DRD4 gene and ADHD: Evidence of linkage and association with the SNP -616 Am J Med Genet 2004 131B 33 37 10.1002/ajmg.b.30071
Mitsuyasu H Hirata N Sakai Y Shibata H Takeda Y Ninomiya H Kawasaki H Tashiro N Fukumaki Y Association analysis of polymorphisms in the upstream region of the human dopamine D4 receptor gene (DRD4) with schizophrenia and personality traits J Hum Genet 2001 46 26 31 11289715 10.1007/s100380170120
Paterson AD Ying DJ Petronis A Schoots O Lieberman JA Van Tol HH Kennedy JL A PstI restriction fragment length polymorphism in the 5' untranslated region of DRD4 is not associated with schizophrenia Psychiatr Genet 1996 6 191 193 9149324
McCracken JT Smalley SL McGough JJ Crawford L Del'Homme M Cantor RM Liu A Nelson SF Evidence for linkage of a tandem duplication polymorphism upstream of the dopamine D4 receptor gene (DRD4) with attention deficit hyperactivity disorder (ADHD) Mol Psychiatry 2000 5 531 536 11032387 10.1038/sj.mp.4000770
Ronai Z Guttman A Keszler G Sasvari-Szekely M Capillary electrophoresis study on DNA-protein complex formation in the polymorphic 5' upstream region of the dopamine D4 receptor (DRD4) gene Curr Med Chem 2004 11 1023 1029 15078164 10.2174/0929867043455503
Seeman P Guan HC Van Tol HH Dopamine D4 receptors elevated in schizophrenia Nature 1993 365 441 445 8413587 10.1038/365441a0
Seeman P Guan HC Nobrega J Jiwa D Markstein R Balk JH Picetti R Borrelli E Van Tol HH Dopamine D2-like sites in schizophrenia, but not in Alzheimer's, Huntington's, or control brains, for [3H]benzquinoline Synapse 1997 25 137 146 9021894 10.1002/(SICI)1098-2396(199702)25:2<137::AID-SYN4>3.0.CO;2-D
Stefanis NC Bresnick JN Kerwin RW Schofield WN McAllister G Elevation of D4 dopamine receptor mRNA in postmortem schizophrenic brain Brain Res Mol Brain Res 1998 53 112 119 9473618 10.1016/S0169-328X(97)00285-4
Hwang CK D'Souza UM Eisch AJ Yajima S Lammers CH Yang Y Lee SH Kim YM Nestler EJ Mouradian MM Dopamine receptor regulating factor, DRRF: a zinc finger transcription factor Proc Natl Acad Sci U S A 2001 98 7558 7563 11390978 10.1073/pnas.121635798
Chang FM Kidd JR Livak KJ Pakstis AJ Kidd KK The world-wide distribution of allele frequencies at the human dopamine D4 receptor locus Hum Genet 1996 98 91 101 8682515 10.1007/s004390050166
Mill JS Caspi A McClay J Sugden K Purcell S Asherson P Craig I McGuffin P Braithwaite A Poulton R Moffitt TE The dopamine D4 receptor and the hyperactivity phenotype: a developmental-epidemiological study Mol Psychiatry 2002 7 383 391 11986982 10.1038/sj.mp.4000984
Strobel A Lesch KP Jatzke S Paetzold F Brocke B Further evidence for a modulation of Novelty Seeking by DRD4 exon III, 5-HTTLPR, and COMT val/met variants Mol Psychiatry 2003 8 371 372 12740593 10.1038/sj.mp.4001253
Ronai Z Guttman A Nemoda Z Staub M Kalasz H Sasvari-Szekely M Rapid and sensitive genotyping of dopamine D4 receptor tandem repeats by automated ultrathin-layer gel electrophoresis Electrophoresis 2000 21 2058 2061 10879966 10.1002/1522-2683(20000601)21:10<2058::AID-ELPS2058>3.0.CO;2-1
Ding YC Chi HC Grady DL Morishima A Kidd JR Kidd KK Flodman P Spence MA Schuck S Swanson JM Zhang YP Moyzis RK Evidence of positive selection acting at the human dopamine receptor D4 gene locus Proc Natl Acad Sci U S A 2002 99 309 314 11756666 10.1073/pnas.012464099
Chen CS Burton M Greenberger E Dmitrieva J Population migration and the variation of dopamine D4 receptor (DRD4) allele frequencies around the globe Evolution and Human Behavior 1999 20 309 324 10.1016/S1090-5138(99)00015-X
Hamer D Sirota L Beware the chopsticks gene Mol Psychiatry 2000 5 11 13 10673763 10.1038/sj.mp.4000662
Ronai Z Barta C Guttman A Lakatos K Gervai J Staub M Sasvari-Szekely M Genotyping the -521C/T functional polymorphism in the promoter region of dopamine D4 receptor (DRD4) gene Electrophoresis 2001 22 1102 1105 11358133 10.1002/1522-2683()22:6<1102::AID-ELPS1102>3.0.CO;2-3
D'Souza UM Russ C Tahir E Mill J McGuffin P Asherson PJ Craig IW Functional effects of a tandem duplication polymorphism in the 5'flanking region of the DRD4 gene Biol Psychiatry 2004 56 691 697 15522254 10.1016/j.biopsych.2004.08.008
Sham PC Zhao JH Waldman I Curtis D Should ambiguous trios for the TDT be discarded? Ann Hum Genet 2000 64 575 576 11281220 10.1046/j.1469-1809.2000.6460575.x
Ronai Z Guttman A Nemoda Z Gervai J Sasvari-Szekely M Direct haplotype detection of adjacent polymorphic sites in the regulatory region of the dopamine D4 receptor (DRD4) gene Electrophoresis 2002 23 1512 1516 12116162 10.1002/1522-2683(200205)23:10<1512::AID-ELPS1512>3.0.CO;2-B
Szantai E Kiraly O Nemoda Z Kereszturi E Csapo Z Sasvari-Szekely M Gervai J Ronai Z Linkage analysis and molecular haplotyping of the dopamine D4 receptor gene promoter region Psychiatr Genet 2005 in press
Lai E Bowman C Bansal A Hughes A Mosteller M Roses AD Medical applications of haplotype-based SNP maps: learning to walk before we run Nat Genet 2002 32 353 12410232 10.1038/ng1102-353
Williams T Tjian R Analysis of the DNA-binding and activation properties of the human transcription factor AP-2 Genes Dev 1991 5 670 682 2010091
Meulenbelt I Droog S Trommelen GJ Boomsma DI Slagboom PE High-yield noninvasive human genomic DNA isolation method for genetic studies in geographically dispersed families and populations Am J Hum Genet 1995 57 1252 1254 7485180
Sambrook J Fritsch EF Maniatis T Molecular cloning: A Laboratory Manual, 2nd Ed., 2005 New York, Cold Spring Harbor Laboratory, Cold Spring Harbor
Schug J Overton GC Modeling transcription factor binding sites with Gibbs Sampling and Minimum Description Length encoding Proc Int Conf Intell Syst Mol Biol 1997 5 268 271 9322048
| 15985158 | PMC1175085 | CC BY | 2021-01-04 16:38:18 | no | BMC Genet. 2005 Jun 28; 6:39 | utf-8 | BMC Genet | 2,005 | 10.1186/1471-2156-6-39 | oa_comm |
==== Front
BMC Health Serv ResBMC Health Services Research1472-6963BioMed Central London 1472-6963-5-421593509910.1186/1472-6963-5-42Research ArticleVolume-based referral for cardiovascular procedures in the United States: a cross-sectional regression analysis Epstein Andrew J [email protected] Saif S [email protected] Harlan M [email protected] Kevin GM [email protected] Division of Health Policy and Administration, Department of Epidemiology and Public Health, Yale University School of Medicine, New Haven, Connecticut, USA2 Section of Cardiovascular Medicine, Department of Internal Medicine, Yale University School of Medicine, New Haven, Connecticut, USA3 Yale-New Haven Hospital Center for Outcomes Research and Evaluation, New Haven, Connecticut, USA4 Center for Health Equity Research and Promotion, Philadelphia Veterans' Affairs Hospital, Philadelphia, Pennsylvania, USA5 Department of Health Care Systems, Wharton School of Business, and Section of General Internal Medicine, Department of Medicine, School of Medicine, both at University of Pennsylvania, Philadelphia, Pennsylvania, USA2005 3 6 2005 5 42 42 24 1 2005 3 6 2005 Copyright © 2005 Epstein et al; licensee BioMed Central Ltd.2005Epstein et al; licensee BioMed Central Ltd.This is an Open Access article distributed under the terms of the Creative Commons Attribution License (), which permits unrestricted use, distribution, and reproduction in any medium, provided the original work is properly cited.
Background
We sought to estimate the numbers of patients affected and deaths avoided by adopting the Leapfrog Group's recommended hospital procedure volume minimums for coronary artery bypass graft (CABG) surgery and percutaneous coronary intervention (PCI). In addition to hospital risk-adjusted mortality standards, the Leapfrog Group recommends annual hospital procedure minimums of 450 for CABG and 400 for PCI to reduce procedure-associated mortality.
Methods
We conducted a retrospective analysis of a national hospital discharge database to evaluate in-hospital mortality among patients who underwent PCI (n = 2,500,796) or CABG (n = 1,496,937) between 1998 and 2001. We calculated the number of patients treated at low volume hospitals and simulated the number of deaths potentially averted by moving all patients to high volume hospitals under best-case conditions (i.e., assuming the full volume-associated reduction in mortality and the capacity to move all patients to high volume hospitals with no related harms).
Results
Multivariate adjusted odds of in-hospital mortality were higher for patients treated in low volume hospitals compared with high volume hospitals for CABG (OR 1.16, 95% CI 1.10–1.24) and PCI (OR 1.12, 95% CI 1.05–1.20). A policy of hospital volume minimums would have required moving 143,687 patients for CABG and 87,661 patients for PCI from low volume to high volume hospitals annually and prevented an estimated 619 CABG deaths and 109 PCI deaths. Thus, preventing a single death would have required moving 232 CABG patients or 805 PCI patients from low volume to high volume hospitals.
Conclusion
Recommended hospital CABG and PCI volume minimums would prevent 728 deaths annually in the United States, fewer than previously estimated. It is unclear whether a policy requiring the movement of large numbers of patients to avoid relatively few deaths is feasible or effective.
==== Body
Background
Patients treated at hospitals with higher volumes of cardiovascular procedures, including coronary artery bypass graft (CABG) surgery and percutaneous coronary intervention (PCI), are reported to have better outcomes than patients treated at hospitals with lower volumes [1-3]. These data have led to a growing interest in using volume to characterize hospital quality of care for cardiovascular procedures by purchaser and consumer organizations [4,5]. As part of its "Evidence-Based Hospital Referral" guidelines, the Leapfrog Group, a coalition of large health care purchasers that insure 34 million Americans, recommended until 2003 that its members contract for selected procedures, including CABG and PCI, only with hospitals that met minimum volume thresholds [6]. Hospital risk-adjusted mortality criteria were added to the Leapfrog guidelines in 2003 [7].
Although proponents contend that implementing volume-based thresholds will reduce procedure-associated mortality, there are limited data on the number of patients affected by the adoption of volume minimums and the magnitude of any potential benefits. One study of patients hospitalized in California in 1997 suggested that 338 deaths might be prevented in that state each year by adopting volume minimums for CABG and PCI [8], and another study estimated that applying hospital volume minimums for CABG and PCI nationwide would save 1,871 lives annually [9]. However, both studies relied on estimates of hospital CABG [1] and PCI [3] volume-mortality associations using New York state data from the late 1980s and early 1990s. Recent research by Birkmeyer and Dimick [7] uses data from 2000 to estimate the volume-mortality association and the national impact of the Leapfrog volume standards, and finds that 148,508 CABG cases would have to be moved to avert 594 deaths and 91,153 PCI cases would have to be moved to avert 547 deaths.
Because many purchasers have begun selective referral to providers that meet these Leapfrog criteria, it is important to understand the potential benefits and costs of the Leapfrog quality improvement recommendations. To provide an alternative, generalizable forecast of the potential consequences of adopting a volume-based referral policy, we conducted an evaluation of a hypothetical nationwide implementation of the Leapfrog Group's volume-based standards, the only criteria in effect until 2003, for CABG and PCI using nationally representative data from 1998–2001. We specifically sought to estimate the number of patients at low volume hospitals who would need to be moved to high volume hospitals and the number of deaths potentially averted by the adoption of hospital volume minimums.
Methods
National Inpatient Sample
Our analysis was based on the Nationwide Inpatient Sample (NIS), a hospital discharge database from the Agency for Healthcare Research and Quality's Health Care Utilization Project [10]. As the largest publicly available all-payer inpatient database in the United States, the NIS contains administrative records for all hospitalizations in a randomly selected national sample of non-governmental, acute care hospitals. The 2001 NIS, the most recent version of the NIS available at the time of our study, contains information on more than 7.4 million discharges from nearly 1,000 hospitals in 33 states, corresponding to nearly 20% of all admissions to US non-federal hospitals [11]. The NIS contains de-identified, hospitalization-level data, including information on primary and secondary diagnoses, demographic characteristics, procedure use, length of stay, payer, total charges, and admission and discharge status. Our study pooled data from the 1998, 1999, 2000, and 2001 NIS releases.
Study sample
We created separate, procedure-specific cohorts for hospitalizations in which a patient had any procedure code indicating a CABG (International Classification of Diseases, 9th Edition Clinical Modification [ICD-9-CM] codes 36.10–36.2) or PCI (ICD-9-CM codes 36.00–36.06 and 36.09). Of the nearly 29 million records in the 1998, 1999, 2000, and 2001 NIS, we identified 306,942 hospitalizations with a CABG and 517,178 hospitalizations with a PCI. We excluded patients under the age of 18 and neonatal or obstetric admissions in order to restrict our evaluation to a typical adult population. Hospitalizations with missing data for sex, age, or mortality were also excluded. To limit administrative data coding errors, we excluded patients treated at hospitals with fewer than 10 CABGs in any year from the CABG cohort and patients treated at hospitals with fewer than 5 PCIs in any year from the PCI cohort. Finally, following the Leapfrog Group's policy recommendation [12], we restricted our analysis to admissions at hospitals located in US Census Bureau-defined Metropolitan Statistical Areas, which are clusters of counties comprising large population centers. The two final unweighted procedure cohorts consisted of 296,135 hospitalizations for CABG drawn from 746 hospital-years of data and 496,252 hospitalizations for PCI drawn from 851 hospital-years of data. With the appropriate NIS sampling weights, these data represented 1,496,937 hospitalizations for CABG from 3,365 hospital-years and 2,500,796 hospitalizations for PCI from 4,141 hospital-years.
Hospital volume groups
To assess the association of hospital CABG and PCI volume and patient mortality, hospitals were divided into separate groups based on their annual volume. Hospitals were categorized as low volume if their annual volume was below the procedure-specific volume minimum recommended by the Leapfrog Group (450 cases for CABG, 400 cases for PCI) [7]. All other hospitals were considered to be high volume.
Statistical analysis
Patient characteristics, including demographics, admission type, comorbidities, and payer, were compared between patients treated in low volume hospitals and high volume hospitals within each procedure cohort using global chi-square analyses for categorical variables and simple t-tests for continuous variables.
The principal study outcome was in-hospital mortality. We compared crude rates of mortality between patients treated in high volume hospitals and low volume hospitals in each procedure cohort using global chi-square analysis. Unadjusted and multivariable logistic regressions that accounted for the NIS survey design were conducted within each procedure cohort to assess the association between treatment at a low volume hospital and patient mortality. Patient characteristics incorporated in the multivariable models, which were derived from previous administrative data-based evaluations of CABG and PCI and clinical judgment, included: sex, race (white, black, other), age (<65 years, 65–74 years, ≥75 years), year, admission source, urgency of admission (emergent, urgent, elective, unknown/missing), coronary artery disease (principal diagnosis of MI [ICD-9-CM code 410], secondary diagnosis of MI, any non-MI coronary disease diagnosis [ICD-9-CM codes 411–414], none), diabetes (ICD-9-CM code 250), chronic obstructive pulmonary disease (ICD-9-CM codes 490–496), hypertension (ICD-9-CM codes 401–405), renal dysfunction (ICD-9-CM codes 580–586), congestive heart failure (ICD-9-CM codes 428, 402.01, 402.11, 402.91, 404.01, 404.11, 404.91), and peripheral vascular disease (ICD-9-CM codes 440, 443).
In addition to factors common to both procedure models, we added selected covariates to specific procedure volume analyses. The multivariable model for CABG mortality accounted for concomitant valve repair and other open heart surgery procedures (ICD-9-CM procedure code 35), use of an internal mammary graft (ICD-9-CM procedure codes 36.15, 36.16), and a same admission PCI (ICD-9-CM procedure codes 36.00–36.06, 36.09). The analysis of hospital PCI volume controlled for multiple vessel PCI (ICD-9-CM procedure code 36.05).
Estimating the effect of establishing hospital volume minimums
To assess the impact of hospital volume minimum policy for CABG and PCI, we calculated the average annual number of patients treated at low volume hospitals in each procedure cohort. Observations were weighted using NIS sampling weights to obtain nationally generalizable estimates.
We then estimated the number of deaths that could be prevented by the universal adoption of hospital volume minimums using mortality estimates obtained from the procedure-specific multivariable logistic regression models. Volume at low volume hospitals was modeled using a logarithmic transformation based on previous studies suggesting hospital volume-mortality associations exhibit a log-linear relationship [13-15]. Volume at high volume hospitals was modeled with a single dummy variable to reflect the average volume-associated mortality effect in high volume hospitals. An initial risk-adjusted probability of mortality was calculated for each hospitalization using the current distribution of patients across hospital volume groups. To derive a "best case" estimate of the impact of a hospital volume minimum policy, we assumed that all patients at low volume hospitals could be transported to high volume hospitals. A second risk-adjusted probability of mortality was then calculated for hospitalizations treated in low volume hospitals assuming that they had been treated in a typical high volume hospital by setting the hospital volume effect for low volume hospital patients to the average hospital volume effect for patients treated in high volume hospitals. This process removed any measured volume-associated difference in mortality between patients treated in low volume and high volume hospitals. Differences in weighted risk-adjusted probabilities between the two scenarios were summed across all low volume hospital patients to determine the estimated number of deaths averted by the adoption of hospital volume minimums for CABG and PCI.
Statistical analyses were conducted using SAS 8.2 (SAS Institute, Cary, NC) and Stata 8.2 (Stata Corporation, College Station, TX). Analysis of the NIS database was approved by the University of Pennsylvania Institutional Review Board.
Results
Patient characteristics
The proportion of patients treated at low volume hospitals was 14.0% for PCI and 38.4% for CABG. Mean patient age was 64.2 for PCI and 66.0 for CABG, and was generally comparable for patients treated in low volume and high volume hospitals. A greater proportion of patients treated at low volume CABG and PCI hospitals were non-white, while a lower proportion represented elective admissions or patients received in transfer as compared with patients treated at high volume hospitals for both procedures. The proportion of CABG patients receiving internal mammary artery grafts was slightly greater in high volume hospitals while the proportion of PCI patients with a myocardial infarction was slightly higher in low volume hospitals. Other patient characteristics, including sex distribution and prevalence of comorbid conditions, were generally comparable between patients at low volume and high volume hospitals (Table 1).
Table 1 Patient characteristics across hospital volume groups
Characteristics Hospital CABG Volume Hospital PCI Volume
Overall Low (<450) High (≥450) P Overall Low (<400) High (≥400) P
% of patients (weighted) 100.0 38.4 61.6 - 100.0 14.0 86.0 -
% of hospital year groups* 100.0 69.8 30.2 - 100.0 45.1 54.9 -
Mean age, (SD) years 66.0 (0.07) 65.9 (0.08) 66.1 (0.11) 0.12 64.2 (0.08) 63.7 (0.12) 64.2 (0.09) <0.001
Age 0.11 <0.001
Less than 65 41.1 41.7 40.7 48.7 50.7 48.3
65–74 years 34.5 34.2 34.7 28.8 27.6 29.0
75 years of age and older 24.4 24.1 24.6 22.5 21.7 22.6
Male 69.5 69.4 69.5 0.90 65.5 64.4 65.7 <0.001
Race 0.014 0.016
White 64.2 59.2 67.3 63.3 56.7 64.4
Black 3.8 3.8 3.9 4.5 5.1 4.4
Other 7.5 9.5 6.2 7.3 10.1 6.9
Race not reported/missing 24.5 27.6 22.6 24.9 28.1 24.4
Primary payer <0.001 <0.001
Medicare 53.8 51.9 55.0 49.0 45.7 49.5
Medicaid 4.0 4.7 3.5 4.0 5.3 3.8
Private 37.3 37.4 37.3 41.2 40.6 41.4
Other/missing 4.9 6.0 4.2 5.8 8.4 5.4
Diabetes 29.5 29.8 29.3 0.14 24.9 25.4 24.8 0.12
Hypertension 58.6 58.0 58.9 0.098 54.1 53.0 54.3 0.022
COPD 17.3 17.5 17.2 0.37 11.2 11.9 11.0 0.002
Congestive heart failure 17.6 17.5 17.7 0.64 9.9 10.7 9.8 <0.001
Peripheral vascular disease 7.6 7.5 7.6 0.55 5.6 5.4 5.6 0.43
Renal disease 6.0 5.8 6.1 0.11 2.6 2.9 2.5 <0.001
Coronary disease <0.001 <0.001
MI as primary diagnosis 20.3 20.1 20.4 31.0 37.6 29.9
MI as secondary diagnosis 5.7 6.3 5.3 4.5 5.0 4.4
Other coronary artery disease 65.6 65.5 65.7 60.2 53.1 61.4
No coronary disease 8.4 8.1 8.6 4.3 4.3 4.3
Admission type <0.001 <0.001
Emergency 23.7 23.8 23.6 31.9 36.9 31.1
Urgent 25.2 22.1 27.1 26.1 23.6 26.5
Elective 41.5 38.6 43.4 32.3 21.5 34.1
Other/missing 9.6 15.6 5.9 9.6 18.1 8.2
Arrived by inter-hospital transfer <0.001 <0.001
Yes 18.0 11.5 22.0 18.2 9.0 19.7
No 78.8 85.4 74.7 78.4 88.4 76.8
Unknown 3.2 3.2 3.3 3.4 2.7 3.6
Year 0.59 0.17
1998 25.4 26.0 25.1 22.0 25.6 21.4
1999 23.7 21.5 25.1 22.2 24.2 21.9
2000 25.5 25.2 25.7 26.0 25.6 26.1
2001 25.3 27.4 24.0 29.8 24.6 30.6
Procedure-specific variables
Same admission PCI 2.8 3.3 2.4 <0.001 - - - -
Concomitant valve procedure 10.6 9.5 11.3 <0.001 - - - -
Internal mammary artery graft 16.8 13.4 18.8 <0.001 - - - -
Multivessel PCI - - - - 14.5 14.8 12.6 <0.001
Unless noted otherwise, findings are expressed as percentages
Percentages may not sum to 100 due to rounding
* Hospital year groups refer to the number of hospitals that contributed data in each year of the NIS. A hospital participating in the NIS over the 3 year period would be considered to have contributed 3 hospital year groups to the analysis.
Hospital procedure volume and mortality
Crude in-hospital mortality was 3.64% for patients undergoing CABG and 1.50% for patients undergoing PCI. In-hospital mortality rates were higher for patients treated in low volume hospitals compared with high volume hospitals for CABG (3.85% vs. 3.51%, P = 0.002) and PCI (1.96% vs. 1.43%, P < 0.001). Patients at low volume hospitals remained at increased risk of in-hospital mortality after multivariable adjustment for CABG (odds ratio [OR] 1.16, 95% confidence interval [CI] 1.10–1.24) and PCI (OR 1.12, 95% CI 1.05–1.20) compared with patients at high volume hospitals (Table 2).
Table 2 Patient mortality by hospital volume groups
Hospital CABG Volume
Overall Low (<450) High (≥450) P
Crude rates 3.6 3.9 3.5 0.002
Unadjusted odds ratio (95% CI) - 1.10 (1.04–1.16) 1.00 [referent] 0.001
Adjusted odds ratio (95% CI) - 1.16 (1.10–1.24) 1.00 [referent] <0.001
Hospital PCI Volume
Overall Low (<400) High (≥400) P
Crude rates 1.5 2.0 1.43 <0.001
Unadjusted odds ratio (95% CI) - 1.37 (1.28–1.49) 1.00 [referent] <0.001
Adjusted odds ratio (95% CI) - 1.12 (1.05–1.20) 1.00 [referent] 0.001
Impact of hospital procedure volume minimums
Implementation of a hospital procedure volume minimum policy for cardiovascular procedures would require the transfer of an estimated 231,348 total patients each year, 143,687 patients for CABG, and 87,661 for PCI. A best-case estimate suggests the transfer of patients from low volume to high volume hospitals annually could have prevented 728 in-hospital deaths. The majority of annual deaths prevented by the transfer of patients to high volume hospitals were for patients undergoing CABG (619 deaths, 4.5% of all CABG deaths), with 109 deaths (1.2% of all PCI deaths) avoided among patients undergoing PCI. Adoption of a hospital volume minimum policy would thus require the transfer of 232 patients from low volume to high volume CABG hospitals to avert a single death, and 805 patients from low volume to high volume PCI hospitals to avert a single death (Table 3).
Table 3 Volume and mortality estimates
CABG PCI
Current
Total procedures, n(%) 374,234 625,199
Performed at LVH, n(%) 143,687 87,661
Performed at HVH, n(%) 230,547 537,538
Total in-hospital deaths, n(%) 13,633 9,405
Performed at LVH, n(%) 5,535 1,719
Performed at HVH, n(%) 8,098 7,686
Adopting volume minimum policy
Total procedures, n(%) 374,234 625,199
Performed at LVH, n(%) 0 0
Performed at HVH, n(%) 374,234 625,199
Total in-hospital deaths, n(%) 13,014 9,296
Performed at LVH, n(%) 0 0
Performed at HVH, n(%) 13,014 9,296
Impact of adopting volume minimum policy
Procedures moved from LVH to HVH, n 143,687 87,661
Deaths averted, n 619 109
Number of procedures moved from LVH to HVH to avoid 1 death 232 805
Percent reduction in deaths 4.54% 1.16%
Discussion
The nationwide implementation of a hospital volume minimum policy for cardiovascular procedures based on volume thresholds promoted by the Leapfrog Group [7] would have required the annual redistribution of more than 231,000 patients who underwent CABG or PCI at low volume hospitals between 1998 and 2001. At best, this redistribution would have resulted in approximately 728 fewer deaths annually, concentrated primarily among patients who underwent CABG (619 deaths). This suggests that previously reported mortality benefits associated with volume minimums for PCI may be overstated (547 deaths averted vs. 109) [7]. Differences between our estimates and previous ones are driven more by methodology than data. Whereas previous studies assumed that all patients moved from low- to high-volume centers would receive the same average mortality benefit, our methodology calculated the expected benefit for each patient at a low-volume center based on the volume of the center and the patient's comorbidities.
The potential to avert up to 728 deaths each year through the treatment of CABG and PCI patients at only high volume hospitals may be interpreted as sufficient evidence for the adoption of a procedure volume minimum policy. However, this benefit must also be considered in the context of the required transfer of over 231,000 patients each year from low volume to high volume hospitals. Because the average absolute incremental increase in mortality associated with treatment at a low volume hospital compared with a high volume hospital is generally small (<0.5%), particularly for patients undergoing PCI, a large number of patients would need to be treated at high volume hospitals (805 for PCI) in order to avert a single death. This number needed to treat is larger than that of most current cardiovascular drugs and therapies, suggesting only a modest benefit for any individual patient [16]. Moreover, the number of deaths attributable to treatment at low versus high volume hospitals represents only a small proportion of overall procedure mortality (4.5% of CABG deaths, 1.2% of PCI deaths). Recent studies also suggest substantial heterogeneity in CABG and PCI outcomes among hospitals, including low volume hospitals with better than predicted outcomes and high volume hospitals with worse than predicted outcomes [17,18]. As such, hospital volume may be both a modest and unreliable measure of any individual hospital's performance [19].
Policies regulating hospital procedure volume minimums may also have potential adverse consequences. Concentrating services among fewer providers may adversely affect access to procedures in many areas of the country [20]. Differences in patient characteristics between low and high volume hospitals in our analysis suggest the suspension of services at low volume hospitals may disproportionately affect minorities and patients with Medicaid insurance, groups with historically limited access to cardiovascular care. A reduction in the number of CABG and PCI providers may also result in higher prices for health care purchasers as provider competition is reduced. Adoption of volume thresholds may unwittingly motivate providers to treat patients with borderline indications in order to meet volume minimums. Patients may not support regionalization of procedures associated with hospital volume minimums if they prefer receiving care at local, low volume hospitals [21]. Each of these and other factors requires consideration prior to adopting any hospital volume minimum policy.
Limitations
Although our analysis provides estimates of the mortality reductions that may be achieved through hospital volume minimums, we necessarily make several assumptions. First, we assumed that high volume hospitals achieved superior outcomes because of their higher volumes [22]. If high volume hospitals have better outcomes because of patient selection or factors other than hospital volume itself, the mortality benefit and number of lives saved reported by transferring patients from low volume hospitals may be overstated. Second, our analysis assumes that all patients can be shifted from low volume hospitals to high volume hospitals and accrue a hospital volume-associated mortality benefit. No study to date has tested this assertion, and its validity remains unknown [23]. Third, because it is impossible to know the exact volume of the high volume hospital to which a low volume hospital patient would be transferred, we assumed that all low volume hospital patients would receive the average benefit of being treated at a high volume hospital. Finally, our analysis assumes that redistributing patients across hospitals has no adverse consequences.
Our study has a number of limitations. First, the NIS database is based on administrative data, and may be susceptible to hospital-based variations in coding practices. Previous studies, however, have demonstrated that administrative databases contain sufficient information to evaluate hospital differences in procedure quality [24], and the NIS is a comprehensive, nationally representative, all-payer database that includes information on cardiovascular procedure use. Second, we evaluated in-hospital mortality and were unable to assess other outcomes, including procedural complications or post-discharge events. However, the Leapfrog Group's volume recommendations are predicated on a mortality reduction, not improvement on other outcomes [6,7]. Third, the NIS does not collect data concerning physician volume, and thus we could not assess the effect of operator volume. However, previous studies suggest hospital volume is associated with outcomes even after accounting for operator volume [13,25,26]. Learning by doing and/or (dis)economies of scale, which are not captured in this analysis, may also influence estimates of the impact of a hospital volume minimum policy. Finally, the NIS does not contain unique patient identifiers, and the possible inclusion of multiple patient admissions in our cohort may violate statistical assumptions of independence.
Conclusion
Implementation of a hospital volume minimum policy for CABG and PCI based on the Leapfrog Group's Evidence Based Hospital Referral guidelines in effect until 2003 would have required the annual redistribution of over 231,000 patients from low volume hospitals to high volume hospitals between 1998 and 2001. This policy would have resulted in, at best, an estimated 728 fewer deaths annually, primarily among patients undergoing CABG. These estimates rely on the unproven assumption that simply directing all patients to high volume hospitals would eliminate the full differential in mortality between low and high volume hospitals. Further, given the uncertain feasibility and potential adverse consequences, requiring the movement of 232 patients for CABG or 805 patients for PCI between hospitals to avert a single death may not be an effective policy. These issues deserve further study before hospital procedure volume minimum policies for CABG and PCI are adopted more widely by purchasers.
Competing interests
The authors declare that they have no competing interests.
Authors' contributions
AJE and SSR were the principal authors of the manuscript. AJE was the principal data analyst. AJE, SSR, HMK and KGMV participated in study conception, interpretation, and drafting of the manuscript. AJE and KGMV were involved with data acquisition. All authors read and approved the final manuscript.
Pre-publication history
The pre-publication history for this paper can be accessed here:
Acknowledgements
At the time this research was conducted, AJE was supported by a National Research Service Award (T32-HS00009) from the Agency for Healthcare Research and Quality at the University of Pennsylvania. SSR was supported by a Medical Scientist Training Grant (GM07205) from the National Institutes of Health National Institute of General Medical Sciences. KGMV was supported by a Veterans' Affairs Health Services Research and Development Career Development Award and a Doris Duke Foundation Clinical Scientist Development Grant.
==== Refs
Hannan EL Kilburn HJ Bernard H O'Donnell JF Lukacik G Shields EP Coronary artery bypass surgery: the relationship between inhospital mortality rate and surgical volume after controlling for clinical risk factors Med Care 1991 29 1094 1107 1943270
Hannan EL Popp AJ Tranmer B Fuestel P Waldman J Shah D Relationship between provider volume and mortality for carotid endarterectomies in New York state Stroke 1998 29 2292 2297 9804636
Hannan EL Racz M Ryan TJ McCallister BD Johnson LW Arani DT Guerci AD Sosa J Topol EJ Coronary angioplasty volume-outcome relationships for hospitals and cardiologists JAMA 1997 277 892 898 9062327 10.1001/jama.277.11.892
Anonymous Hospital ratings
Center for Medical Consumers Center for Medical Consumers
The Leapfrog Group Evidence-based hospital referral
Birkmeyer JD Dimick JB Potential benefits of the new Leapfrog standards: effect of process and outcomes measures Surgery 2004 135 569 575 15179361 10.1016/j.surg.2004.03.004
Dudley RA Johansen KL Brand R Rennie DJ Milstein A Selective referral to high-volume hospitals: estimating potentially avoidable deaths JAMA 2000 283 1159 1166 10703778 10.1001/jama.283.9.1159
Birkmeyer JD Finlayson EVA Birkmeyer CM Volume standards for high-risk surgical procedures: potential benefits of the Leapfrog initiative Surgery 2001 130 415 422 11562662 10.1067/msy.2001.117139
Steiner C Elixhauser A Schnaier J The Healthcare Cost and Utilization Project: an overview Eff Clin Pract 2002 5 143 151 12088294
Anonymous National Inpatient Sample
Birkmeyer JD Birkmeyer CM Wennberg DE Young M Leapfrog patient safety standards: the potential benefits of universal adoption 2004 Leapfrog Group
Birkmeyer JD Stukel TA Siewers AE Goodney PP Wennberg DE Lucas FL Surgeon volume and operative mortality in the United States N Engl J Med 2003 349 2117 2127 14645640 10.1056/NEJMsa035205
Jollis JG Peterson ED DeLong ER Mark DB Collins SR Muhlbaier LH Pryor DB The relation between the volume of coronary angioplasty procedures at hospitals treating Medicare beneficiaries and short-term mortality N Engl J Med 1994 331 1625 1629 7969344 10.1056/NEJM199412153312406
Jollis JG Peterson ED Nelson CL Stafford JA DeLong ER Muhlbaier LH Mark DB Relationship between physician and hospital coronary angioplasty volume and outcome in elderly patients Circulation 1997 95 2485 2491 9184578
Anonymous Number Needed to Treat (NNT)
Rathore SS Epstein AJ Volpp KGM Krumholz HM Hospital coronary artery bypass graft surgery volume and patient mortality, 1998-2000 Ann Surg 2004 239 110 117 14685108 10.1097/01.sla.0000103066.22732.b8
Epstein AJ Rathore SS Volpp KGM Krumholz HM Hospital percuatenous coronary intervention volume and patient mortality, 1998 to 2000: does the evidence support current procedure volume minimums? J Am Coll Cardiol 2004 43 1755 1762 15145095 10.1016/j.jacc.2003.09.070
Peterson ED Coombs LP DeLong ER Haan CK Ferguson TB Procedural volume as a marker of quality for CABG surgery JAMA 2004 291 195 201 14722145 10.1001/jama.291.2.195
Petersen LA Normand SL Leape LL McNeil BJ Regionalization and the underuse of angiography in the Veterans Affairs health care system as compared with a fee-for-service system N Engl J Med 2003 348 2209 2217 12773649 10.1056/NEJMsa021725
Finlayson SRG Birkmeyer CM Tosteson ANA Nease RFJ Patient preferences for location of care: implications for regionalization Med Care 1999 37 204 209 10024124 10.1097/00005650-199902000-00010
Shahian DM Normand SLT The volume-outcome relationship: from Luft to Leapfrog Ann Thorac Surg 2003 75 1048 1058 12645752 10.1016/S0003-4975(02)04308-4
Epstein AJ Rathore SS Coronary artery bypass surgery, hospital volume, and risk Circulation 2003 108 e6 e7 12848167 10.1161/01.CIR.0000079058.43607.8B
Jones RH Hannan EL Hammermeister KE DeLong ER O'Connor GT Luepker RV Parsonnet V Pryor DB for the Working Group Panel on the Cooperative CABG Database Project Identification of preoperative variables needed for risk-adjustment of short-term mortality after coronary artery bypass graft surgery J Am Coll Cardiol 1996 28 1478 1487 8917261 10.1016/S0735-1097(96)00359-2
Hannan EL Wu C Ryan TJ Bennett E Culliford AT Gold JP Hartman A Isom OW Jones RH McNeil B Rose EA Subramanian VA Do hospitals and surgeons with higher coronary artery bypass graft surgery volumes still have lower risk-adjusted mortality rates? Circulation 2003 108 795 801 12885743 10.1161/01.CIR.0000084551.52010.3B
McGrath PD Wennberg DE Dickens JDJ Siewers AE Lucas FL Malenka DJ Kellett MAJ Ryan TJJ Relation between operator and hospital volume and outcomes following percutaneous coronary interventions in the era of the coronary stent JAMA 2000 284 3139 3144 11135777 10.1001/jama.284.24.3139
| 15935099 | PMC1175086 | CC BY | 2021-01-04 16:31:53 | no | BMC Health Serv Res. 2005 Jun 3; 5:42 | utf-8 | BMC Health Serv Res | 2,005 | 10.1186/1472-6963-5-42 | oa_comm |
==== Front
BMC Infect DisBMC Infectious Diseases1471-2334BioMed Central London 1471-2334-5-441593874410.1186/1471-2334-5-44Research ArticleCongenital rubella syndrome in Iran Sadighi Jila [email protected] Hasan [email protected] Kazem [email protected] Institute for Health Sciences Research, Tehran, Iran2 Tehran University of Medical Sciences, Tehran, Iran2005 6 6 2005 5 44 44 15 9 2004 6 6 2005 Copyright © 2005 Sadighi et al; licensee BioMed Central Ltd.2005Sadighi 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
Congenital rubella syndrome (CRS) can be prevented with appropriate vaccination programs. The prevalence rates of rubella and CRS in Iran are unknown; therefore, the risk of exposure in pregnant women is not clear. The prevalence of CRS in the pre-vaccine period can be estimated by evaluating the proportion of children in the population with sensorineural hearing loss attributable to rubella.
Methods
This was a case-control study to estimate prevalence of CRS in Tehran (Iran) by evaluating the proportion of children with sensorineural hearing loss attributable to rubella. The study used rubella antibody titer as an indicator, and compared the prevalence of rubella antibody between children with and without sensorineural hearing loss. Using these findings, the proportion of cases of sensorineural hearing loss attributable to rubella was estimated.
Results
A total of 225 children aged 1 to 4 years were entered into the study (113 cases and 112 controls). There was a significant difference between cases and controls with regard to rubella antibody seropositivity (19.5% vs. 8.9%, respectively, odds ratio = 2.47, 95% CI = 1.04–5.97). The proportion of sensorineural hearing loss cases attributable to rubella was found to be 12%, corresponding to a CRS prevalence of 0.2/1000.
Conclusion
The prevalence of CRS was approximately 0.2/1000 before rubella vaccination in Iran, Moreover; the results suggest that implementation of appropriate rubella vaccination programs could potentially prevent about 12% of cases of sensorineural hearing loss in Iranian children. This data could potentially be used as baseline data, which in conjunction with an appropriate method, to establish a surveillance system for rubella vaccination in Iran. An appropriate surveillance system is needed, because the introduction of a rubella vaccine without epidemiological data and an adequate monitoring program could result in the shifting of rubella cases to higher ages, and increasing the incidence of CRS.
==== Body
Background
Rubella is a common, normally mild disease that mainly affects children aged 2–12 years. Rubella in pregnancy may cause abortion, stillbirth and congenital anomalies, or congenital rubella syndrome (CRS). Prior to the introduction of rubella vaccine in 1969, the disease was distributed evenly throughout the world. In temperate regions, the incidence was usually highest in late winter and early spring. Minor epidemics occurred every 6–9 years, with major epidemics occurring at intervals ranging from 10 to 30 years [1,2].
The rubella pandemic in the 1960's clearly demonstrated the extraordinary teratogenic potential of the rubella virus. In spite of the fact that 80% of pregnant women were immune to rubella in the United States, it is estimated that more than 12,500,000 cases of rubella occurred. Congenital rubella occurred in an estimated 30,000 pregnancies, with 10,000 resulting in fetal death or therapeutic abortion, and 20,000 resulting in infants born with CRS [3]. The estimated cost to the US economy was approximately $2 billion [4].
The incidence of congenital rubella varies in different populations and depends on the number of susceptible pregnant women, the circulation of rubella virus, and rubella vaccination coverage. According to the World Health Organization (WHO), at least 236,000 CRS cases occur in every non-epidemic year in developing countries, and this increase by up to 10 fold during epidemic years The CRS cases are rarely reported in these countries, and the extent of the problem remains unknown. However, the indiscriminate introduction of rubella vaccine without epidemiological data and an adequate monitoring program should be avoided because the occurrence of rubella cases can shift to higher ages and increase the incidence of CRS [5].
Rubella is often not notified, as many cases are not seen by a doctor or even recognized by the patient; consequently, rubella outbreaks can occur without clinical recognition. Nevertheless, studies in Central and South America, Africa, India and the Middle and Far East suggest that rubella is widespread and endemic in most developing countries [6,7].
The percentage of infection in the fetuses of mothers infected by rubella during the first trimester of pregnancy is greater than 80%. As a result, the target group for the vaccination is all women of childbearing age. Therefore, the fundamental reason for using the vaccine containing the rubella antigen is to prevent congenital rubella syndrome [8].
In October 2004, CDC convened an independent panel of internationally recognized authorities on public health, infectious disease, and immunization to assess progress toward elimination of rubella and congenital rubella syndrome in the United States, a national health objective for 2010. Since rubella vaccine licensure in 1969, substantial decline in rubella and CRS have occurred, and absence of endemic transmission in the United States is supported by recent data: fewer than 25 reported rubella cases each year since 2001, at least 95% vaccination coverage among school aged children, estimated 91% population immunity, adequate surveillance to detect rubella outbreaks and a pattern of virus genotypes consistent with virus originating in other parts of the world [9].
A rubella vaccination program in the United Kingdom (UK) was initiated in 1970. Reported cases of CRS declined from about 50 a year 1971–75 to just over 20 a year 1986–90, and rubella associated terminations from an average of 750 to 50 a year [10].
The European Region (EUR) of the World Health Organization (WHO) comprises 52 member countries, with an estimated population of 876 million. In 1998, the Regional Committee for EUR resolved to reduce the incidence of congenital rubella syndrome (CRS) in all countries to <1 per 100,000 live births by 2010. Large rubella outbreaks continue to occur in countries that only recently introduced rubella vaccination (e.g., Russian Federation and Romania). In many countries, CRS surveillance is not fully implemented, resulting in underestimates of CRS disease burden, both at country and regional levels [11]. The introduction of rubella vaccine in the infant immunization schedule should only be considered when coverage>80% can be assured on a long-term basis [12].
Between December 2003 and January 2004, the Ministry of Health and Medical Education of the Islamic Republic of Iran implemented a nationwide campaign to vaccinate about 32,000,000 people aged 5 to 25 years with a combined measles and rubella (MR) vaccine. Before this campaign, rubella vaccination was not included in the childhood vaccination schedule. After the campaign, the Ministry changed the childhood vaccination schedule so as to include 2 doses of the measles, mump and rubella (MMR) vaccine, one given at 15 months and the other at 4 to 6 years of age. Before the MR campaign in Iran, the epidemiology of rubella and congenital rubella was not clear. There is still no adequate surveillance system in place for congenital rubella after vaccination. An adequate surveillance system should monitor the following parameters: incidence of CRS, incidence of rubella, rubella immunity in women of childbearing age, rubella vaccination coverage and rubella outbreaks.
Retrospective studies on children with abnormalities similar to the complications of CRS in developing countries can estimate the rate of congenital rubella risk in these children. Cases of rubella-related deafness in children have been identified by comparing the prevalence of rubella antibody in children with and without sensorineural deafness. Therefore, a reduction in the number of deaf children has been used as an indication of reduced maternal rubella infection after the introduction of rubella vaccination programs [6]. Sensorineural hearing loss is one of the most common abnormalities (50%) associated with CRS [13]. Hearing loss present at birth is often not detected until a later age. Hearing loss can also occur as a delayed manifestation of CRS; and is the most common complication (80%) with late onset [14]. The hearing loss is bilateral, sensorineural type in all grades of severity [12].
Moreover, up to 50% of infections during pregnancy are sub-clinical, and many go unrecognized. Thus, the estimated incidence of rubella-related deafness (like the other CRS defects) is likely less than the true incidence [5]. Hearing impairment can result from fetal rubella not only during the first trimester, but also in the second and third trimesters of pregnancy [13]. Rubella serology is useful in epidemiological studies to examine the role of rubella as a cause of sensorineural hearing loss, because the number of acquired infections can be estimated from data gathered from controls and cases of rubella-related deafness in populations of children. The aim of this study was to use a similar method to estimate the prevalence of CRS by indicating the proportion of children in the population with sensorineural hearing loss attributable to rubella.
Methods
A case-control study of 225 children aged 1–4 years was conducted from November 1995 to May 1996 in Tehran, Iran. The study compared the prevalence of rubella antibody between children with and without sensorineural hearing loss, and tested the hypothesis that congenital rubella is associated with an increased risk of sensorineural hearing loss.
The cases were 113 medically confirmed deaf children admitted to deaf educational centers. The controls were 112 children with normal hearing selected from the surgery ward of Amirkabir hospital in Tehran. One limitation of this study was the selection of the control group, because many parents with healthy children were reluctant to allow blood sampling of their children. Therefore, the control group was taken from among patients at a hospital surgery ward, excluding subjects who may have had an infectious disease.
It is significant that children under 1 and over 4 years of age were excluded from the study. In infants (mainly under 6 months), high levels of rubella antibody passively transferred from mother to fetus could result in an overestimate of seropositivity, while in children over 4 years old, rubella antibody is usually acquired from a postnatal infection. Detection of antibody in children 1 to 4 years of age was used to make a retrospective diagnosis of congenital rubella infection. It is also significant that rubella infection is most frequent among children 5 to 14 years of age [15].
During the time period covered by this study, rubella vaccination was not included in the national routine childhood vaccination; however, some physicians administered MMR vaccine instead of measles vaccine to their patients. Therefore, we excluded all children who had a history of MMR vaccination. History of MMR vaccination was obtained from the vaccination cards of children.
For all children, data were collected on maternal (prenatal and delivery) and neonatal histories. In addition, blood samples were taken from children and the titer of rubella antibody was determined. The serological technique employed was the hemagglutination inhibition (HI) test. Children with rubella HI antibody titers of 1:8 or greater were regarded as seropositive, and those with titers of less that 1:8 as seronegative. The Virology Department of the Public Health School (Tehran University of Medical Sciences) carried out the HI tests. Seropositive children in the case and control groups were compared using the chi-squared test and odds ratio (OR). The attributable risk (AR) was then estimated. The formula for calculating the AR using the odds ratio for disease in the exposed population is:
% Attributable Risk = P (OR-1)/ 1+ P (OR-1)
where P is the exposure prevalence in the controls as long as the disease is rare and the control group is reasonably representative of all non-cases in the population (proportion of rubella antibody in the control group), and OR is the odds ratio for disease in the exposed population.
Finally, we developed the following formula to estimate the prevalence of CRS:
"Prevalence of CRS = B * C * (1/D) * 1000"
This formula is derived from the following formula:
"Prevalence of CRS = [A * B * C * (1/D) * 1000] / A"
where A is the number of children during the year of study, B is the prevalence of hearing loss, C is the AR, and D is the frequency of sensorineural hearing loss following CRS.
Results
There were 113 children in the case group and 112 children in the control group. The case and control groups consisted of 53% and 47% male subjects, respectively. This difference was not statistically significant (P = 0.2). No significant difference was found with regard to age distribution of cases as compared to controls (P = 0.12). The mean (SD) age was higher in cases than controls, but this difference was not significant: 2.9 years (SD = 1.1) compared with 2.6 years (SD = 1.3) (P = 0.4).
The percentage of children with rubella antibody in the case and control groups is shown in Table 1. The percentage of children with rubella antibody was significantly higher in cases than controls: 19.5% compared with 8.9%, respectively (OR = 2.47, 95% CI 1.04–5.97, P = 0.02). Therefore, congenital rubella was associated with a significantly increased (greater than two times) sensorineural hearing loss in children. The AR was estimated to be about 12%, which indicates that about 12% of the children born with hearing loss were damaged as a result of congenital rubella, and thus that it may be possible to reduce the incidence of sensorineural hearing loss by up to 12% by preventing rubella infection in pregnant women.
Table 1 Rubella antibody status in hearing impaired (case) and normal control children aged 1–4 years, Tehran, Iran
Rubella antibody status Case n (%) Control n (%) Odds Ratio (95% CI)
Seropositive 22 (19.5) 10 (8.9)
Seronegative 91 (80.5) 102 (91.1) 2.47 (1.04–5.94)
Total 113 (100) 112 (100)
According to the results of this study, the prevalence of CRS in Tehran was estimated to be about 0.2/1000 children. This prevalence was estimated using the following data:
A = number of children aged 1 – 4 years in Tehran (1995): 775000 (using national data)
B = prevalence of hearing loss in children: 1/1000 (using national data)
A*B = children with hearing loss: 775 (expected)
C = attributable risk (derived from the present study): 12%
A*B*C = children with rubella-related deafness: 93 children (expected)
D = frequency of sensorineural hearing loss associated with CRS: 50% (derived from previous studies)
A*B*C*1/D = the number of children with CRS: 186
Therefore:
Prevalence of CRS in Tehran = 186 / 775000 = 0.2/1000
Considering the number of children aged 1 – 4 years in Iran and the prevalence of CRS estimated here (0.2/1000 children), the present results suggest that tens of CRS cases could be prevented each year by appropriate vaccination programs.
Maternal history in sensorineural hearing-impaired children (case group) according to rubella antibody status is shown in Table 2. Nine mothers (41%) of 22 deaf seropositive children reported a history of rubella infection, rash, or rubella exposure during pregnancy, whereas none of 91 mothers of seronegative deaf children reported these events (p = 0.000).
Table 2 Rubella antibody status in sensori-neural hearing-impaired children aged 1–4 years according to maternal history of rubella infection, Tehran, Iran
Maternal history of rubella Seropositive n (%) Seronegative n (%) P-value
Positive 9 (41) 0 (0)
Negative 13 (59) 91 (100) 0.000
Total 22 (100) 91 (100)
Discussion
The present findings indicate that the prevalence of CRS in Iran is approximately 0.2/1000 (Before rubella vaccination in Iran). World Health Organization reported that in the absence of widespread rubella vaccination, the incidence of CRS varies between 0.1–0.2 per 1,000 live births, with higher rates (1–4 per 1,000 live births) during epidemics [2].
During rubella outbreaks, rates of CRS per 1000 live births were at least 1.7 in Jamaica, 0.7 in Oman, 2.2 in Panama, 1.5 in Singapore, 0.9 in Sri Lanka, and 0.6 in Trinidad and Tobago. These rates are similar to those reported from industrialized countries during the pre-vaccine era [16].
The prevalence of bilateral sensorineural hearing loss is 0.5 to 1 newborns per 1000 live births. In addition, the onset of hearing loss can occur at any time throughout childhood. Thus, it is estimated that the prevalence of bilateral hearing loss increases to 1.5–2/1000 children under the age of 6 years [15]. In Iran, the prevalence of hearing loss is 1/1000 children [17]. Considering the prevalence of sensorineural hearing loss, the estimated AR in this study (12%), it is estimated that congenital rubella was the cause of deafness in approximately 93 children aged 1–4 in Tehran in the year considered in the present study. If we assume that the epidemiology of rubella in Tehran, the capital city of Iran, is similar to that in other areas of Iran, we obtain the estimate that in the year of the present study there were approximately 620 children (1–4 years) in Iran with deafness that could have been prevented by rubella vaccination.
Some investigations in Iran showed that rubella immunity in women of childbearing age from 1968 through 1995 (the time of this study) fluctuated between 70% and 95% [18-22]. The rate of rubella immunity in this population in 1995 was estimated as 80% [23]. Therefore, the year considered in the present study was a non-epidemic year, and the number of children born with deafness due to rubella would be expected to increase in epidemic years.
Nine mothers (41%) of 22 deaf, seropositive children in the present study reported a history of rubella, rash, or rubella exposure during pregnancy. Other studies have also reported between 40% and 75% of deaf seropositive children had such a maternal history [24].
In this study, the degree of hearing loss in children who attended deaf educational centers was often higher than 50 dB (severe to profound hearing loss), and their hearing loss was bilateral. Thus, children with low severity of deafness were not included in this study, and the relation between severity of deafness and congenital rubella could not be estimated.
About 20% of the children in deaf educational centers were not included because of a history of MMR (Measles, Mump and Rubella) vaccination, which may lead to an underestimation of the OR. Meanwhile, experience in other countries suggests that if MMR vaccination coverage is less than 60%–70%, it may actually increase the age of infection, and therefore the incidence of CRS [1].
After comparing vaccination cards with parents' reports about vaccination history, it was found that five deaf children among the cases actually had a history of MMR vaccination, and they were consequently excluded from the study. Contrary to expectations, rubella antibody titer was negative in two of these children. Therefore, the efficacy of MMR vaccine should be investigated in future studies.
According to research done in Iran (during the year of this study), the rate of rubella immunity has reached about 80% [23]; however, this rate of immunity is similar to that in other countries during the rubella pandemic of the 1960's, which claimed thousands of victims [3,4]. Epidemiological evidence has shown that while rubella virus continues to circulate among children, there is still a risk of infection in pregnant women, even though only 3% of them are non-immune, and there is little prospect of eliminating CRS [1,25].
The world has now cumulated 35 years of lessons on use of rubella vaccine, and some striking examples of how rubella vaccination strategies should and should not be applied [10,11,26-31]. Most importantly, studies in developed countries have generated the following recommended vaccination program: routine MMR vaccination at 12–15 months of age followed by a second dose of MMR vaccine at 4–6 years (both sexes) [10,32,33]. This study clearly showed the necessity for suitable rubella vaccination program in Iran. However, inadequately implemented childhood vaccination runs the risk of altering rubella transmission dynamics and can lead to increase insusceptibility in women of childbearing age with the potential of increased numbers of cases of CRS. Consequently, it is essential that childhood vaccination programs achieve and maintain high levels of coverage [12].
Before the MR campaign in Iran, the epidemiology of rubella and congenital rubella was not clear. In spite of this unclear epidemiology, rubella vaccination was launched in Iran (after the time period covered by the present study). Now, due to limited disease surveillance and reporting systems, data on the incidence of rubella and CRS in Iran are scant. This study provides data on the prevalence of CRS in Iran; this data could potentially be used as baseline data, which in conjunction with an appropriate method, to establish a surveillance system for rubella vaccination in Iran.
Conclusion
Rubella is a common communicable viral disease of childhood, and rubella in pregnancy may cause CRS. Sensorineural hearing loss is the most common abnormality associated with CRS, and it is the most common complication with late onset. The present findings indicate that the prevalence of CRS was approximately 0.2/1000 before rubella vaccination in Iran, Moreover; the results suggest that implementation of appropriate rubella vaccination programs could potentially prevent about 12% of cases of sensorineural hearing loss in Iranian children. This data could potentially be used as baseline data, which in conjunction with an appropriate method, to establish a surveillance system for rubella vaccination in Iran. An appropriate surveillance system is needed because the introduction of rubella vaccine without epidemiological data and an adequate monitoring program may result in the shifting of rubella cases to higher ages, and increased incidence of CRS.
List of abbreviations
CRS: Congenital Rubella Syndrome
AR: Attributable Risk
OR: Odds Ratio
HI: Heamagglutination Inhibition
SD: Standard Deviation
Competing interests
The author(s) declare that they have no competing interests.
Authors' contributions
JS: Study idea, literature review, methodology development, data analysis and interpretation, and writing the final article.
HE: Supervisor of the research project.
KM: Counseling the methodology and supervised the statistical analysis.
Pre-publication history
The pre-publication history for this paper can be accessed here:
Acknowledgements
This study was supported by grant of the Tehran University of Medical Sciences. The authors acknowledge helpful discussions with Dr. Ali Montazeri from Iranian Institute for Health Sciences Research.
==== Refs
Galazka A Rubella in Europe Epidemiol Infect 1991 107 43 54 1879489
World Health Organization Rubella vaccines. WHO Position Paper Wkly Epidemiol Rep 2000 75 161 69
Herrmann KL Rubella in the United States: toward a strategy for disease control and elimination Epidemiol Infect 1991 107 55 61 1879490
Preblud SR Serdula MK Frank JA JrBrandling-Bennett AD Hinman AR Rubella vaccination in the United States: a ten years review Epidemiol Rev 1980 2 171 94 7000534
World Health Organization (EPIGAG) Rubella and congenital rubella syndrome in developing countries 14th Meeting 1991
Miller CI Rubella in the developing world Epidemiol Infect 1991 107 63 68 1879491
Mingle JAA Frequency of rubella antibodies in the population of some African countries Rev Infect Dis 1985 7 S68 71 4001737
Pan American Health organization. Introduction of new vaccines in the national vaccination. MMR vaccine
Centers for Disease Control and Prevention (CDC) Achievements in public health: Elimination of rubella and congenital rubella syndrome – United States, 1969–2004 MMWR 54 279 82 2005 Mar 25
Tookey P Rubella in England, Scotland and Wales Euro Surveill 9 21 22 2004 Apr 1 15075481
Centers for Disease Control and Prevention (CDC) Progress toward elimination of measles and prevention of congenital rubella infection – European region, 1990–2004 MMWR 54 175 8 2005 Feb 25
Report of a meeting on preventing congenital rubella syndrome: immunization strategies, surveillance needs. Geneva, 12–14 January 2000
Northern JL Downs MP Hearing in Children 1991 248 4 Baltimore: Williams & Wilkins 371
Nelson WE Behrman RE Kliegman RM Arvin AM Nelson Textbook of Pediatrics 1992 14 Philadelphia: Saunders 520 522
Nelson WE Behrman RE Kliegman RM Arvin AM Nelson Textbook of Pediatrics 1996 15 Philadelphia: Saunders 571 572
Cutts FT Robertson SE Diaz-Ortega JL Samuel R Control of rubella and congenital rubella syndrome (CRS) in developing countries, Part 1: Burden of disease from CRS Bull World Health Organ 1997 75 55 68 9141751
Ministry of Health and Medical Education Health and Disease in Iran National report 1990 1999
Naficy K Saidi S Serological survey on viral antibodies in Iran Tropical and geographical medicine 1970 22 183 8 4317129
Saidi S Epidemiological survey of rubella immunity in Iran Bulletin of the World Health Organization 1972 46 563 5 4538202
Kabiri M Moattari A The rubella immunosurveillance of Iranian females: an indication of the emergence of rubella outbreak in Shiraz, Iran Iranian journal of medical sciences 1993 18 134 7
Modarres S Modarres S Oskoii NN The immunity of children and adult females to rubella virus infection in Tehran Iranian journal of medical sciences 1996 21 69 73
Pakzad P Ghafourian M Rubella survey among pregnant women and congenitally infected infants in Khouzestan province Medical journal of Ahwaz 1996 19 University of Medical Sciences 56 66
Rahimi F Sarafnegad A Salarbehzadi SH Survey of rubella immunity in Tehran Thesis 1994 Azad University
Gumpel SM Hayes K Dudgeon JA Congenital perceptive deafness: role of intrauterine rubella Br Med J 2 300 4 1971 May 8 5575233
Miller E Rubella in the United Kingdom Epidemiol-Infect 1991 107 31 42 1879488
Levy-Bruhl D Six C Parent I Rubella control in France Euro Surveill 9 13 14 2004 Apr 1 15192259
Katow S Surveillance of congenital rubella syndrome in Japan, 1978–2002: effect of revision of the immunization law Vaccine 22 4084 91 2004 Sep 28 15364460
Davidkin I Peltola H Leinikki P Epidemiology of rubella in Finland Euro Surveill 9 11 12 2004 Apr 1
Ciofi Degli Atti M Filia A Revello M Buffolano W Salmaso S Rubella control in Italy Euro Surveill 9 17 18 2004 Apr 1
Rafila A Marin M Pistol A Nicolaiciuc D Lupulescu E Uzicanin A Reef S A large rubella outbreak, Romania – 2003 Euro Surveill 9 7 8 2004 Apr 1 15192257
Panagiotopoulos T Georgakopoulou T Epidemiology of rubella and congenital rubella syndrome in Greece, 1994–2003 Euro Surveill 9 15 16 2004 Apr 1
American Academy of Pediatrics Committee on Infectious Diseases Recommended childhood and adolescent immunization schedule: United States, 2005 Pediatrics 2005 115 182 15630000
Centers for Disease Control and Prevention (CDC) Recommended childhood and adolescent immunization schedule -United States, 2005 MMWR 53 Q1 3 2005 Jan 7
| 15938744 | PMC1175087 | CC BY | 2021-01-04 16:28:14 | no | BMC Infect Dis. 2005 Jun 6; 5:44 | utf-8 | BMC Infect Dis | 2,005 | 10.1186/1471-2334-5-44 | oa_comm |
==== Front
BMC MicrobiolBMC Microbiology1471-2180BioMed Central London 1471-2180-5-251590449410.1186/1471-2180-5-25Research ArticleRole of core promoter sequences in the mechanism of swarmer cell-specific silencing of gyrB transcription in Caulobacter crescentus England Jennifer C [email protected] James W [email protected] Department of Chemistry and Biochemistry and Molecular Biology Institute, University of California, Los Angeles Los Angeles, CA, 90095-1569, USA2005 17 5 2005 5 25 25 1 3 2005 17 5 2005 Copyright © 2005 England and Gober; licensee BioMed Central Ltd.This is an Open Access article distributed under the 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
Each Caulobacter crescentus cell division yields two distinct cell types: a flagellated swarmer cell and a non-motile stalked cell. The swarmer cell is further distinguished from the stalked cell by an inability to reinitiate DNA replication, by the physical properties of its nucleoid, and its discrete program of gene expression. Specifically, with regard to the latter feature, many of the genes involved in DNA replication are not transcribed in swarmer cells.
Results
We show that for one of these genes involved in DNA replication, gyrB, its pattern of temporal expression depends upon an 80 base pair promoter region with strong resemblance to the Caulobacter crescentus σ73 consensus promoter sequence; regulation does not appear to be affected by the general strength of the promoter activity, as mutations that increased its conformity with the consensus did not affect its cell-cycle expression pattern. Transcription from the gyrB promoter in vitro required only the presence of the σ73 RNA polymerase (from E. coli) and the requisite nucleoside triphosphates, although a distinct binding activity, present in crude whole-cell extracts, formed a complex gyrB promoter DNA. We also assayed the effect on gyrB expression in strains containing mutations in either smc or dps, two genes encoding proteins that condense DNA. However we found there was no change in the temporal pattern of gyrB transcription in strains containing deletions in either of these genes.
Conclusion
These experiments demonstrate that gyrB transcription does not require any auxiliary factors, suggesting that temporal regulation is not dependent upon an activator protein. Swarmer-specific silencing may not be attributable to the observed physical difference in the swarmer cell nucleoid, since mutations in either smc or dps, two genes encoding proteins that condense DNA, did not alter the temporal pattern of gyrB transcription in strains containing deletions in either of these genes. Rather a repressor that specifically recognizes sequences in the gyrB promoter region that are also probably essential for transcription, is likely to be responsible for controlling cell cycle expression.
==== Body
Background
The bacterium Caulobacter crescentus divides asymmetrically once during each cell cycle, yielding a sessile stalked cell and a motile swarmer cell [1,2]. These two cell types differ not only in morphology but also in their ability to replicate their genomes and to divide. The stalked cell immediately reinitiates DNA replication and cell division. The swarmer cell, in contrast, is incompetent for these two processes until it sheds its single, polar flagellum and differentiates into a stalked cell. The stalked cell then passes through several predivisional stages, during which time the genome is replicated and segregated, the newly synthesized DNA is methylated, and a new flagellum is synthesized and constructed at the pole opposite the stalk [1-4]]. The asymmetry that distinguishes the two daughter cells is founded in the predivisional cell before cell separation occurs. Asymmetric cell division depends upon several general mechanisms, including intracellular protein and/or mRNA localization, regulated phosphorylation and proteolysis, and cell-cycle dependent transcription and translation.
The inability of the Caulobacter swarmer cell to replicate DNA is due in part to the activity of the response regulator CtrA, which binds to sites in the origin of replication (Cori) [5,6]. CtrA is synthesized and phosphorylated when DNA replication is approximately half-way complete; the phosphorylated protein persists in the swarmer cell after cell division [7,8]. Phosphorylated CtrA represses transcription of DNA adjacent to the Cori, that is necessary for replication initiation, and also prevents binding of the DnaA initiator protein [6,9,10]. When the swarmer cell differentiates into a stalked cell, CtrA is proteolyzed and DNA replication commences [7]. In addition to its role in preventing DNA replication, CtrA is also a transcriptional activator of many genes that affect several different aspects of Caulobacter development. CtrA regulates transcription of its own gene, activates early genes in flagellar biogenesis, and represses ftsZ, the earliest known gene in cell division [5,8,11]. Interestingly, while microarray analysis has implicated CtrA in the control of approximately 25% of the cell-cycle regulated genes and it appears to directly regulate 55 operons, it does not regulate the transcription of several genes involved in DNA replication [12,13]. Thus, control of DNA replication initiation also depends on CtrA-independent transcription regulation.
Transcription of many DNA replication genes begins just before DNA replication initiates at the swarmer-to-stalked cell transition. These swarmer-cell-silenced genes include dnaA, dnaN (encoding the β subunit of DNA polymerase III), dnaX (γ and τ subunits of DNA polymerase III), dnaK (replication-initiation chaperone), and gyrB (B subunit of DNA gyrase) [14-18]]. The predicted promoter sequences for most of these genes align well with the published consensus sequence for σ73 [19]. As the promoters of these DNA replication-associated genes appear to contain neither a CtrA binding site or any obvious elements that would suggest transcription activation in stalked cells, the mechanism responsible for their cell-cycle dependent regulation remains unknown.
In this study we analyze the role of the gyrB promoter in swarmer cell-specific silencing of gene expression. We map the site at which transcription of gyrB starts and create deletions to further delineate the promoter sequences required for temporal regulation of the gene to an 80 bp region. Using site-directed mutagenesis we further show that transcriptional silencing in swarmer cells does not appear to depend upon the overall strength of the promoter activity. In vitro experiments demonstrate that gyrB transcription does not require any auxiliary factors, suggesting that temporal regulation is not dependent upon an activator protein. However, gel-mobility shift assays demonstrate that an activity, present in C. crescentus crude whole-cell extracts and unrecognized by anti-RNAP antibody, is able to bind to the gyrB promoter. One possibility is that swarmer-specific silencing is linked to the observed difference in the physical characteristics between the swarmer cell nucleoids and stalked cell nucleoids [20-23]. Thus, a protein that affects the overall folding and structure of the nucleoid might also be regulating transcription. In this regard, we also assayed the effect on gyrB expression in strains containing mutations in either smc or dps, two genes encoding proteins that condense DNA. However we found there was no change in the temporal pattern of gyrB transcription in strains containing deletions in either of these genes.
Results
Mapping and deletion analysis of the gyrB promoter
Previous studies of the gyrB gene found that transcription is silenced in swarmer cells and begins when the swarmer cell differentiates into a stalked cell [17]. These experiments also demonstrated that gyrB is transcribed from its own promoter and is not part of an operon, this despite the fact that the predicted start codon overlaps with the upstream recF stop codon [17]. Because the original gyrB-promoterless lacZ reporter fusion used in these studies contained a relatively large region of DNA upstream of the predicted start of translation, we wished to map the gyrB promoter to the minimal region capable of sustaining wild-type temporal regulation of the gene. Primer extension revealed two adjacent transcription start sites, separated from the translation start site by a relatively long leader sequence of approximately 170 base pairs (Fig. 1A). Note that in a separate study a slightly different start site, four base pairs further upstream, was also identified [14]. While we did not detect this start site in either the primer extension assay or in an in vitro transcription/primer extension experiment (see below), the alternative transcription start site was taken into consideration in subsequent mutagenesis experiments.
Based on the transcription start sites found by primer extension, alignment and comparison with several other known σ73 genes allowed identification of possible -10 and -35 promoter elements of gyrB (Fig. 1B) [14-16,19,24-26]. The consensus is derived from σ73 transcribed genes that are expressed at relatively constant levels throughout the cell cycle; expression of gyrB, along with other DNA-replication genes, varies during the cell cycle. In analyzing the gyrB promoter and surrounding DNA, we considered two broad possibilities for gyrB transcription regulation. gyrB expression might be controlled by the activity of a swarmer cell-specific factor that inhibits transcription or by a stalked cell-specific factor that activates transcription. An (A+T)-rich region, beginning approximately 60 bp upstream of the transcription start site, at first appeared to be a good candidate for a possible regulatory protein binding site. This region consists of two stretches of repeated thymidine residues and one stretch of repeated adenine residues, approximately phased with respect to each other on the DNA strand (data not shown). Such an arrangement could create a bend in the DNA, a secondary structure that is preferred by some DNA-binding proteins and that can affect transcription independent of any additional protein constituents [27,28]. Another part of the gyrB gene that might possibly contain a cis-acting regulatory site is the extremely long (170 bp), untranslated leader sequence between the transcription and translation start sites. Alternatively, gyrB transcription might not depend on any remote DNA elements or sites for sequence-dependent DNA binding proteins. As has been reported, the putative gyrB promoter appears more similar to a subset of σ73 genes that mirror its temporal regulation than to those genes that are constitutively expressed throughout the cell cycle (Fig. 1B) [14,17,24,26]. This would seem to suggest that temporal regulation is dictated by the promoter itself rather than by a more distantly located DNA element.
In order to begin dissection of the gyrB promoter, we used PCR to create deletions both upstream and downstream of the core σ73 promoter. We measured the transcriptional activity from these shortened promoters, each fused to a promoter-less lacZ gene, by β-galactosidase activity. Deletion of the upstream sequences, including the (A+T)-rich region, resulted in a slight, but not statistically significant, decrease in gyrB transcription (pJEZP2 and pJEZP3) (Fig. 2). A similar decrease was seen when approximately 100 bp of the untranslated leader sequence was deleted (pJEZP2s). However, wild-type transcription levels were restored by combined removal of the leader and upstream sequences (pJEZP3s, pJEZP4) and/or by removal of an additional 49 bp of the leader sequence (pJEZP5). These results indicate that the overall strength of gyrB promoter activity depends upon at most an 80 bp promoter region, covering -53 to +27 bp relative to the start of transcription. The temporal regulation of the deleted promoters was then investigated by immunoprecipitation of β-galactosidase produced from each construct during growth in synchronized cultures. The expression pattern of the deleted promoter fusions during the cell cycle was similar to that observed for the full-length fusion (Fig. 3). In all cases, transcription was barely detectable in swarmer cells (0 division units) followed by a marked increase in transcription at the swarmer-to-stalked cell transition (0.2 division units) and then a subsequent decrease in late predivisional cells (0.8 division units) (Fig. 3). Thus, the temporal control of the gyrB gene is also dictated by the core, 80 bp promoter.
Mutagenesis of the gyrB promoter
One attractive idea to explain the temporal regulation of gyrB transcription is that a deviation in promoter sequences from that of the σ73 consensus results in a expression levels that are somewhat weaker than that of constitutively expressed promoters. This property might make gyrB transcription more susceptible to a general silencing mechanism present in swarmer cells, whereas stronger σ73promoters would be expressed at normal levels. We investigated the importance of the basic architectural elements of the gyrB promoter, relative to the consensus sequence derived from constitutive σ73 promoters, by site-directed mutagenesis (Fig. 4). The mutations were designed to bring the gyrB promoter into better agreement with the C. crescentus σ73 consensus [19,26]. The P6m2 mutation changed the bases CT at -16 to -15 to GC, conserved at those positions in the consensus; likewise, P6m3 created a T at the -37 position. Because an alternate start site and promoter were published while these experiments were in progress, we also designed mutations based on these findings, so designated by the letter "L" in the name of the mutation [14]. The mutations P6mL2 and P6mL3 were aimed at increasing the consensus agreement of the -35 region published previously [19,26]; P6mL4 targeted a proposed 13-mer motif located between the -10 and -35 region that was also previously identified [24].
Each mutated promoter was fused to a promoterless lacZ gene and expression levels were quantitatively measured by β-galactosidase assays (Fig. 4A). The mutations P6m2 and P6m3 resulted in transcription levels that were twice that of the wild-type level, while the P6mL2, -3, and -4 were transcribed at normal or (P6mL3) reduced amounts. The increased promoter activity of P6m2 and P6m3 and the decreased expression of P6mL3 support the proposed promoter sequence reported here (Fig. 1) as the P6m2 and P6m3 mutations increase the agreement with the σ73 consensus and while the P6mL3 mutation decreases the conformity. One possibility that could account for the increased promoter activity of the P6m2 and P6m3 mutant promoters was inappropriate expression in swarmer cells. Therefore, we wished to examine the temporal transcription patterns of these promoters in order to test this possibility (Fig. 4B). Surprisingly, transcription from the mutant promoters, though stronger, was still regulated in the wild-type manner during the cell cycle (Fig. 4B).
Transcription of gyrB does not require an activator
Because simply creating a stronger σ73 promoter failed to have any effect on the temporal pattern of gyrB transcription, we wished to rule out the possibility that gyrB requires a transcriptional activator for expression within stalked cells. We employed an in vitro transcription assay to address this issue, using a commercially obtained preparation of E. coli (σ70) RNA polymerase (RNAP). In this experiment, transcription from a promoter on a super-coiled plasmid was assayed by primer extension. Transcription products were detected using both preparations of RNAP, indicating the lack of a requirement for an auxiliary transcriptional activation factor (Fig. 5). Of the several start sites observed, one site was present when either RNAP species was used in the in vitro assay (Fig. 5) as well as when RNA was obtained directly from cultured cells (see Fig. 1A). The presence of one common start site is further evidence for the correct identification of the gyrB promoter reported here. From this experimental result, we can conclude that the sole protein requirement for in vitro transcription of the gyrB gene is the σ70 (σ73)RNAP holoenzyme.
An activity, distinct from RNA polymerase, binds the gyrB promoter in vitro
Because transcription from the gyrB promoter did not appear to require an activator protein, it seemed likely to us that temporal regulation of the gene during the cell cycle was not attributable to stalked cell-specific transcriptional activation. An alternative possibility is that the gene is specifically repressed in swarmer cells and in the swarmer pole of the predivisional cell, possibly by a protein that binds to the promoter. We therefore decided to use gel-mobility shift assays to look for any binding activities present in C. crescentus crude cell extracts that interacted with the gyrB promoter. A very weak shift was observed when fresh crude extract was incubated with the probe (Fig. 6A, lanes 2–3, white broken arrow); this shift is apparently due to the binding of RNAP, evidenced by the further retardation of mobility when anti-RNAP antibody was added to the incubation mixture (Fig. 6A, lanes 5–7). An additional complex was seen with increasing volumes of crude extract (Fig. 6A, lanes 3–4, gray arrow). This binding activity appears to be distinct from RNAP binding as the shifted band is unaffected by the addition of anti-RNAP antibody (Fig. 6A, lanes 5–7).
Interestingly, we observed different results when we compared samples containing fresh extracts (and extracts that had been frozen only once) with those containing extracts that had been previously thawed and then refrozen. The non-RNAP binding activity was labile under conditions in which the extract had been subjected to one round of freeze-thaw (Fig. 6B, lanes 2 and 3 vs. lanes 4 and 5). Furthermore, when the extract that had been refrozen and thawed was used in the binding assay (Fig. 6B, lanes 2,3), the disappearance of the specific, labile protein band was accompanied by the appearance a new specific super-shifted band upon the addition of anti-RNAP antibody.
The role of DNA condensing proteins in the regulation of gyrB transcription
Previous studies have shown that, in addition to being unable to initiate DNA replication, C. crescentus swarmer cells also differ from stalked cells in their chromosomal architecture [21-23]. These experiments revealed that isolated swarmer cell nucleoids sediment faster in a sucrose gradient than those of stalked cells. One hypothesis is that the swarmer cell nucleoid possesses a different complement of DNA 'condensing' proteins such as histone-like proteins. The gel mobility shift activity present in crude cell extracts may reflect the binding of a protein that alters the overall architecture of the swarmer cell nucleoid. Two candidates for such an activity may be either Dps or SMC. In E. coli, the Dps protein is induced both in stationary phase and during starvation; it is able to physically protect the nucleoid from oxidative and nuclease damage, as well as affect transcription under these conditions [29-32]. SMC (Structural Maintenance of Chromosomes), is a member of the eukaryotic SMC family [reviewed in [33-36]], large, multidomain proteins that associate with each other in an antiparallel fashion. This association occurs through long, coiled-coil domains that flank a central hinge region, resulting in protein dimers that are shaped like old-fashioned hairpins. Dimers of SMC associate with non-SMC factors to form specific complexes in eukaryotic cells, with functions in DNA organization and movement, ranging from condensation to dosage compensation to repair [33,37]. Experiments in Bacillus subtilis have demonstrated that chromosomal DNA was visibly decondensed in smc mutants [38].
The Caulobacter dps gene is approximately 50 to 55% identical and 60 to 65% similar to dps genes in other organisms, including Burkholderia pseudomallei, Bordetella pertussis and Pseudomonas syringae [data not shown]. Furthermore, a search of the NCBI CDD (conserved domain database) revealed significant alignment with a COG (cluster of orthologous genes) for Dps/ferritin-like DNA-binding proteins. The gene appears to be the first in a two-gene operon; the expected product of the second gene is a conserved hypothetical protein of unknown function. A lacZ transcription fusion of a DNA fragment spanning 600 bp upstream of the predicted dps translation start site was expressed at a low level (350–400 units of β-galactosidase activity) in late log-phase cultures [data not shown]. Although it has been shown to be involved in DNA protection and starvation and stationary-phase survival in other organisms, it seemed possible that Dps might be involved in developmental events specific to the swarmer cell: the results of genome-wide DNA microarray experiments showed that dps mRNA is maximally present in swarmer cells [13,30]. As one way to investigate a possible role for Dps in Caulobacter development, we engineered a strain in which 90% of the dps coding sequence was replaced with a gene conferring resistance to gentamicin. The strain was indistinguishable from wild-type cells in logarithmic growth and morphology, and we detected no overt aberration in swarmer-cell motility or nucleoid appearance (data not shown). Additionally, we constructed a strain with a null mutation in smc by disrupting the chromosomal copy of the s m c gene by insertion of a gene conferring spectinomycin resistance [see Methods]. The deletion was confirmed by Southern blot and by immunoblot with antibodies against SMC [data not shown].
Finally, we examined the cell-cycle transcription pattern of gyrB in the smc and dps mutant strains. The temporal expression pattern of a gyrB fused to a promoter-less lacZ gene was assayed by immunoprecipitation of pulse-labeled β-galactosidase in synchronized cultures of NA1000 (wild-type), JG3003 and JG3402 (Fig. 7). In both mutant strains, the lacZ gene was transcribed in a manner comparable to the transcription observed when the same promoter fusion was expressed in NA1000. Thus, SMC and Dps do not appear to participate in swarmer-specific transcription silencing in Caulobacter crescentus.
Discussion
In Caulobacter crescentus, many of the genes involved in DNA replication, including dnaN, dnaX, dnaA, dnaC, dnaK and gyrB, are maximally expressed at the time that this process occurs [14-18,24,25]. Thus, transcription of these genes does not occur in swarmer cells, or in the swarmer pole of predivisional cells, and begins only when the swarmer cells differentiate into stalked cells or when DNA replication commences anew in the stalked cells at the completion of cell division. The means whereby this regulatory pattern is achieved remains unknown. In this study, we used gyrB as a representative gene with which to investigate the mechanism responsible for swarmer-specific transcription silencing in C. crescentus. As revealed by primer extension and in vitro transcription/primer extension assays, the transcription start site of the gene is separated from the translation start site by a fairly long leader sequence. There is apparent variation in the precise beginning of the transcript, as observed here and as reported elsewhere [14]. The lack of a precise transcription start site may be due to other factors that influence gyrB regulation. For example, transcription of the gyrB gene in Caulobacter is induced by relaxation of the DNA [17], and this induction is distinct from the increased expression that occurs during the swarmer-to-stalked cell transition. It is also possible that the transcription start site varies depending upon undefined regulatory cues such as growth rate and/or growth phase of the culture.
Having delineated the gyrB promoter to a minimal, 80 base pair region, we compared this promoter to other σ73 promoters and to the published consensus sequence, taking care to explore different alignments based on the several possible start sites. The alignment shown [see Fig. 1] is that best fitting the major start site obtained by both primer extension and in vitro transcription/primer extension (for an alternative, see [14]). Examination of the -10 and -35 regions of the promoter, based upon our alignment, revealed that the gyrB promoter greatly differs from the C. crescentus σ73 consensus sequence. This suggested an interesting possibility for a swarmer-specific silencing mechanism. We reasoned that greater deviation from the consensus sequence might make these promoters slightly inferior binding sites for the σ73 RNAP, and thus slightly weaker. Such promoters might be more susceptible to a situation in which RNAP binding is partially inhibited or blocked, a possible scenario occurring in swarmer cells. Site-directed mutagenesis allowed us to test this possibility by creating versions of the gyrB promoter that more closely matched the consensus sequence and therefore should, in theory at least, have been transcribed at higher levels. However, while in two cases transcription from the mutated promoters was greater than from the wild-type promoter, temporal regulation of transcription was unaffected. Thus, it seems that the strength of the promoter does not intrinsically determine cell cycle regulation of gyrB.
The experiments presented here show that the gyrB promoter can be transcribed by E. coli σ70 RNA polymerase holoenzyme in the absence of additional protein factors, suggesting the unlikelihood of temporal regulation depending solely upon an activator. However, it is possible that a stalked-cell specific transcriptional activator is required to overcome a swarmer-cell specific repression complex or DNA topological conformation. When a probe containing the gyrB promoter was used in gel-mobility shift assays, two binding activities, both of which are components of crude Caulobacter extracts, were apparent. One of these activities was loosely identified as RNA polymerase based on the ability of anti-RNAP antibody to recognize and alter the migration of this band through the gel. The other binding activity was not recognized by the antibody. This second species was vulnerable to freeze-thaw treatment, as it was only present in extract that had not been re-frozen. One might imagine a scenario in which a swarmer-specific DNA protein(s) binds to the gyrB promoter and interferes with the ability of the polymerase to access the DNA.
Previous experiments with another swarmer cell-repressed promoter, the dnaX promoter, revealed the RRF motif, that by sequence comparison is located between the -10 and -35 elements in all of the swarmer cell-silenced DNA replication gene promoters [24]. Within the RRF of the dnaX promoter, a C residue at position -21 appears to be involved in the temporal regulation of the gene [24]. A substitution of T at this position results in a more than 3-fold increase in transcription and, more importantly, in a loss of swarmer-specific silencing of dnaX. Additionally it was shown that the RRF motif is bound by an unknown species found in crude extracts [24]. The same shift was seen when DNA fragments encompassing the -10 and -35 regions of several different DNA-replication genes, including gyrB and dnaX, were used as probes in gel-mobility shift assays. It is possible that the binding activity reported here for the gyrB promoter is the same as that observed previously with the RRF motif. If this observed binding activity were responsible for the swarmer cell-specific silencing of gyrB transcription, the one might expect that swarmer cell extracts would be enriched in binding activity. However, we were unable to detect this activity in extracts derived from either isolated swarmer or stalked cells, probably owing to limitations in obtaining a sufficient quantity of material. The relevance of these DNA binding activities and a experimental definition of their binding sequences in similarly regulated promoters will need to be explored.
The condensed nature of nucleoids isolated from swarmer cells, relative to stalked cells, when subjected to centrifugation through a sucrose gradient, suggests that the two cell types differ in the proteins that affect the organization and folding of their chromosomes [20]. Growth-phase dependent variation in the relative composition of the DNA-binding protein cadre in E. coli has been documented [39]. A similar developmentally-regulated phenomenon might be responsible for silencing gyrB and other genes involved in DNA replication in Caulobacter swarmer cells. Proteins such as HU or H-NS bind DNA nonspecifically and have been shown to affect transcription regulation [40]. While neither protein binds to a specific sequence on the DNA, each has a preference for certain secondary DNA structures, including gaps and junctions (HU) and (A+T)-rich bends (H-NS). Of particular interest, H-NS represses the virF gene by binding to two sites, one overlapping the RNAP binding site in the promoter and another 200 bp upstream [41]. However, repression only occurs at temperatures below 32°C because below this temperature the DNA is more intrinsically bent, which allows interaction between H-NS bound at both sites. HU, the most abundant DNA-binding protein in bacteria, also regulates transcription. Similar to integration host factor, HU is thought to exert its influence by creating a bend in the DNA and thus affecting the activity of other, more specific transcription regulators [reviewed in [40,42,43]]. Based on searches of the published genome, we have concluded that C. crescentus does not possess a gene that might encode H-NS, nor does it appear to have a homolog of the closely related gene, stpA [42]. C. crescentus, does however, possess genes encoding proteins that are closely homologous to the two subunits of E. coli HU (HUα and HUβ). We suspect that HU is probably not involved in the regulation of the DNA replication genes since we have recently found that the abundance of both HU subunits does not change during the course of the cell cycle [unpublished observation].
In order to determine whether or not two other well-known DNA condensation proteins might affect swarmer-specific transcription, we constructed strains in which smc or dps were disrupted. There is strong evidence that the large bacterial SMC protein, very different from the small DNA-binding proteins discussed above, is involved in chromosome condensation and partitioning, similar to its eukaryotic homologs [33,36]. A deletion of the gene encoding SMC in Bacillus subtilis results in a temperature-sensitive strain with visibly decondensed chromosomes and defects in chromosome segregation [44,45]. Similar results have also been reported in Caulobacter crescentus [46]. Importantly, however, temporal transcription of gyrB was unchanged in the Δsmc mutant, indicating that SMC is not necessary for swarmer-specific transcriptional silencing.
The final target of our studies was the Dps protein. in E. coli, Dps protects DNA during periods of starvation or in stationary phase [31,32]. When bound to DNA, Dps forms a crystalline structure that is proposed to render the DNA physically resistant to oxidative damage as well as degradation by some nucleases [29]. Dps is also important for induced DNA protection during non-stationary phase growth, as dps mutants are more sensitive to hydrogen peroxide [47]. The nature of the Caulobacter swarmer cell (DNA replication- and cell division-incompetent) makes it conceivable that the intracellular environment may be likened to that of a cell in stationary phase or starved for nutrients. Thus, perhaps Dps plays a different role in Caulobacter. Supporting this hypothesis, DNA microarray experiments have revealed that the dps mRNA level is greatest in swarmer cells and then decreases dramatically at the swarmer-to-stalked cell transition [13]. A deletion of dps, however, did not have a visible effect on the Caulobacter cell cycle. Also, similar to the Δsmc strains, the temporal expression pattern of the gyrB promoter was unaffected in the Δdps strain.
Conclusion
In summary, these results show that a core 80 bp gyrB promoter sequence is sufficient to confer a cell cycle-regulated pattern of transcription. In vitro assays indicated that gyrB transcription does not require any factors in addition to RNAP. Thus, the transcription of gyrB is likely to be regulated by a repressor protein in swarmer cells, and not an activator protein. This repressor may be related to the DNA binding activity present in crude cell extracts that bound specifically to the gyrB promoter region. Deletions in the genes encoding the global DNA binding proteins, SMC and Dps had no effect on viability, growth rate or the temporal transcription pattern of gyrB. Thus, the results of this study do not provide definitive support for the idea that DNA-binding proteins that alter the global architecture of chromosomal DNA are involved in transcriptional repression in the C. crescentus swarmer cell. However, it is possible that the condensation of the swarmer-cell nucleoid is the result of the combined activity of two or more proteins and this may be important in influencing gyrB transcription.
Methods
Bacterial strains, growth conditions, and plasmid construction
Table 1 shows the strains and plasmids employed in this study. The gyrB fragments used in construction of promoter fusions were created by polymerase chain reaction (PCR), unless otherwise indicated, with the indicated restriction sites introduced on oligonucleotides otherwise complementary to the wild-type DNA sequence. The exception to this is the plasmid pJEZ1.1, in which the gyrB promoter fragment is derived from an upstream, native PstI site (~-460 bp from the transcription start site), and the plasmid pJEKS-gyrB500, in which the 500 bp gyrB fragment is derived from the same native PstI site and a native StuI site at position +76 bp from the transcription start site. Generally, DNA fragments were first inserted into the plasmid pBluescript-KS+ before subsequent subcloning. The template used for the PCR reactions was purified C. crescentus (NA1000) chromosomal DNA; the base plasmid (placZ/290) [48] containing the fusions is a low-copy-number, plasmid, introduced into Caulobacter from a host E. coli strain (S17-1) by conjugation [49]. All E. coli cultures were grown at 37°C with aeration in LB broth containing (if applicable) the appropriate antibiotic [50]. C. crescentus strains were grown with aeration at 31°C in either peptone yeast extract (PYE) [51] or M2 minimal medium containing glucose [52].
The dps deletion strain was generated by the sacB selection method [53]. In order to accomplish this, the dps gene was first isolated by PCR as two separate DNA fragments that were subcloned into pBluescript KS+, reconstituting the gene with a newly-created internal MluI site. The coding region was then disrupted by introducing a gentamicin resistance cassette from pKnock GM [54] into the unique MluI site. This entire fragment, dps2Δg, was then subcloned into the sacB selection vector, pNPTS139 [M.R.K. Alley, unpublished]. The plasmid, pNPTS-dps2Δg, harbored in E. coli strain S17-1, was introduced into C. crescentus strain NA1000 by conjugation. Transconjugants were selected from PYE agar plates containing kanamycin and naladixic acid (20 μg/ml). A 2 ml culture inoculated with several transformants was grown for approximately 8 hours in PYE without selection. The culture was then diluted 1/10 and plated (0.2 ml/plate) on PYE agar containing 2.5% sucrose. One to three hundred sucrose resistant colonies were than spotted onto two different growth media: PYE sucrose plates containing gentamicin and PYE plates containing kanamycin. Sucrose-resistant, kanamycin-sensitive, gentamicin-resistant colonies where isolated for further study. The deletion of dps in strain pJG3402 was confirmed by PCR. The smc deletion strain was also isolated by this basic method except that the smc gene was disrupted with a spectinomycin resistance cassette. The smc deletion was confirmed by Southern blot (data not shown) and by immunoblot with anti-SMC antibodies (not shown). Following allelic exchange, the disrupted dps and smc genes were each transduced into a new wild-type C. crescentus strain [53].
Cell cycle experiments
The transcription of the lacZ reporter fusions was assayed in cultures synchronized essentially as described previously [21]. Isolated swarmer cells were suspended in M2 glucose medium and allowed to continue growth. At various time points, 5 ml portions of growing culture were removed and the proteins were pulse-labeled for 5 minutes with 35S-Trans-label (ICN). Labeled β-galactosidase was immunoprecipitated as described previously [17].
Site-directed mutagenesis and molecular biology procedures
The initial template used for mutagenesis was the 180 bp gyrBP4 promoter fragment. The fragment was cloned into M13-BM20 and site-directed mutagenesis was performed on single-stranded DNA isolated from E. coli TG-1, as previously described [55]. The mutant gyrB promoters were confirmed by sequencing performed according to the dideoxy chain-termination method [56]. The β-galactosidase was assayed as previously described [57,58], and the reported values represented the mean activity from at least three independently grown cultures assayed in triplicate. The standard deviation in all cases was less than 5%. All other routine molecular biology manipulations were performed as previously described [50].
Primer extension and in vitro transcription/primer extension
Primer extension, both singly and following in vitro transcription, was performed as previously described [50]. The oligonucleotides (PE1: 5'-GCGTCGCCACGCGAACGC-3'; P E 2 : 5 ' – T G A G T T C G T C A G C C A G A G C – 3 ' ; B S 2 : 5 ' -GAGCAATATTACAGGATTCG-3') used in these experiments and for sequencing (data not shown) were complimentary to sequences near the 5' end of the predicted mRNA sequence and radioactively end-labeled with [γ-32P] ATP and T4 polynucleotide kinase [50]. In each primer extension assay, the same primer was also used for a concurrent dideoxy sequencing reaction [56]. For primer extension, total RNA was isolated from strain NA1000 (O.D. at 600 nm of 1.0) grown in M2 glucose minimal medium. Pelleted cells were resuspended in 20 mM sodium acetate, 1 mM EDTA (pH 5.4). 10% SDS was then added to a final concentration of 4%, followed by addition of an equal volume of phenol equilibrated with 20 mM sodium acetate, 1 mM EDTA (pH 5.4). Extraction, consisting of vortexing and incubation at 65°C for 10 minutes, was performed twice. The RNA was then ethanol precipitated from the aqueous phase, rinsed twice with 70% ethanol, and dried. Approximately 50 μg of the isolated RNA was used for primer extension.
In vitro transcription was performed as described [50]. The template was a cesium chloride gradient-purified preparation of pBluescript KS+ containing either the 900 bp gyrBP3 promoter fragment (see Fig. 3) or an approximately 500 bp PstI/StuI fragment (gyrB500) containing the promoter and regions further upstream (Table 1). It was transcribed by either E. coli RNA Polymerase σ70 holoenzyme (Epicenter Technologies, Madison, WI). The resultant mRNA was then reverse transcribed from the PE2 primer, as detailed above.
Gel mobility shift assays
The crude extracts used in these experiments were prepared from mid-log phase, 0.5-liter cultures of wild-type Caulobacter crescentus (NA1000). The cells were collected by centrifugation and then rinsed three to four times with HEMGK (20 mM HEPES, pH 7.6; 0.1 mM EDTA, pH 8.0; 12.5 mM MgCl2; 5% glycerol; 100 mM KCl) before being re-suspended in 4 ml of the same. PMSF in 95% ethanol was then added to 0.1% and the cells were lysed by sonication. The lysate was cleared by centrifugation at 34,500 × g in a Sorvall SS-34 rotor; the cleared lysate, containing approximately 1 to 3 mg/ml protein, was immediately frozen (-80°C) in 0.4 ml aliquots. Unless otherwise indicated, each aliquot was thawed for use only once. The DNA probe was prepared by digesting pKSgyrB500 with HindIII (a site derived from the polylinker) and PstI, purifying the fragment from a 1% agarose gel and then labeling with [α-32P] dGTP by filling in the HindIII site with DNA polymerase I, Klenow fragment (Promega). The gel shift assays were performed essentially as described, with some minor modifications [60]. Briefly, each reaction contained 10–20 × 103 cpm of the labeled DNA probe and was incubated in Incubation Buffer (20 mM Tris-Cl, pH 7.5; 80 mM NaCl; 1 mM EDTA, pH 8.0; 0.1 mg/ml salmon-sperm DNA, sonicated) for 10 minutes at room temperature, after which loading dye was added and the samples run on a 6% non-denaturing polyacrylamide gel at 4°C. Where antibody to RNA polymerase [59] was included in the reaction, 1 to 3 μl were added and all samples were incubated an additional 5 minutes at room temperature.
Authors' contributions
J.C.E. carried out the experimental portion of this study. J.C.E. and J.W.G. conceived of the study and wrote the manuscript.
Acknowledgements
J.C.E. was supported by USPHS predoctoral fellowship GM-07185. Work in our laboratory is supported by Public Health Service Grant GM48417 from the National Institutes of Health to J.W.G.
Figures and Tables
Figure 1 The gyrB transcription start site and promoter. (A) The transcriptional start site of gyrB was mapped by primer extension analysis, using an oligonucleotide primer complementary to a sequence just upstream of the translational start site. Transcription initiates at two adjacent sites, indicated by stars, approximately 170 bp upstream of the predicted start of translation. (B) The gyrB promoter was compared with other known σ73 promoters, both cell-cycle regulated and constitutive; predominantly the former are shown [14-16,19,24-26]. The start sites and predicted -10 and -35 promoter elements are indicated above the sequence alignment; the published C. crescentus consensus sequence for genes transcribed by the common σ73 RNAP holoenzyme is shown below, where S stands for G or C, W indicates A or T, and N means any nucleotide [19,26]. The lines above the dnaX sequence indicate the previuosly identified RRF element [24].
Figure 2 Deletion analysis of the gyrB promoter region. PCR-produced gyrB deletions were fused to a promoter-less lacZ gene in the placZ/290 vector. Transcriptional activity, measured by β-galactosidase production, is shown as a percent of wild-type (pJEZ1.1) activity (3,968 units). For each construct, the percentage is based on a mean value calculated from β-galactosidase activities of three or more different mid-log phase cultures. The base pair positions relative to the start of transcription are marked below each fusion. White boxes indicate the promoter-less lacZ gene and gray boxes represent the gyrB coding sequence.
Figure 3 Temporal expression patterns of the gyrB promoter fusions during the C. crescentus cell cycle. Transcription of the gyrB-lacZ fusions contained on the indicated plasmids was examined during synchronized growth by immunoprecipitation of pulse-labeled β-galactosidase, as described in Material and Methods. The lanes, each revealing the β-galactosidase synthesized during one five-minute interval, are aligned beneath a diagram showing the corresponding stage in the cell cycle, as determined by light microscopy. 'Cell division units' refer to fractions of one complete cell cycle, the duration of which is approximately 140 minutes under the experimental conditions employed here. Samples were labeled every 30 minutes or every 20 minutes (pJEZP5).
Figure 4 Transcription of gyrB promoter mutations designed to conform to the C. crescentus σ73 promoter consensus. (A) The plasmid pJEZP4 contains the template promoter construct used for mutagenesis (see Fig. 3). Shown below is the DNA sequence of the promoter, with the -10 and -35 consensus regions in reverse highlight above. The mutations, shown as boxed sequences with the altered bases in bold-face, were based upon sequence alignment and were aimed at increasing the promoter's conformity with the accepted σ73 consensus sequence. In a separate study, an alternative start site and subsequent alignment located the gyrB promoter further upstream; the mutations designed to take these data into account include an "L" in their names [14]. The transcriptional activity from the mutant promoters, as determined by β-galactosidase measurements of the corresponding lacZ fusions, is presented in the box in the lower right-hand corner as a percentage of wild-type (pJEZ1.1). (B) Temporal expression patterns of the gyrB promoter mutants. Transcription from the gyrB promoter mutants was analyzed during the C. crescentus cell cycle. The results for the mutants exhibiting the strongest quantitative transcription levels (pJEZP6m2 and pJEZP6m3) are shown beneath a schematic of the cell cycle that shows the corresponding developmental stage of the synchronized culture. Samples were labeled every 20 minutes.
Figure 5 In vitro transcription/primer extension of gyrB. In vitro transcription, followed by primer extension, was performed on the gyrB promoter template pJEKS-P3. A σ70 RNAP holoenzyme preparation, obtained from E. coli σ70 was used for transcription assays. Shown at the top are the DNA species, produced from the extended messages, next to a sequencing ladder generated from the same primer. Strong start sites are indicated by larger arrows, weaker start sites by smaller arrows. Below, is the complimentary sequence of the gyrB promoter, with the -10 and -35 promoter elements indicated. The major transcription start sites found by this assay are shown by upward-facing arrows, while the adjacent start sites previously found by primer extension (see Fig. 1) are shown in bold; the largest upward-facing arrow indicates a start site appearing in both assays and with both RNA polymerase species.
Figure 6 Gel-mobility shift assays reveal a distinct, labile activity that binds to the gyrB promoter. (A) A 500 base pair [32P]-end-labeled DNA fragment containing the gyrB promoter was used as a binding template for incubation with increasing amounts of fresh C. crescentus crude cell extract (approximately 1 to 3 mg/ml), without (lanes 1–4) or with (lanes 5–7) increasing amounts of anti-RNAP antibody. (B) The same binding template was incubated with 3 μl of fresh crude extract (lanes 4–5) or extract that had been previously re-frozen (lanes 2–3), with and without anti-RNAP antibody (3 μl). All binding mixtures were incubated at room temperature for 10 minutes and immediately loaded onto native 6% polyacrylamide gels. Black solid arrows indicate probe alone; gray solid arrows indicate a potential freeze-thaw-labile, shifted species; white, broken arrows indicate a binding activity that is super-shifted by anti-RNAP antibody.
Figure 7 The gyrB promoter is expressed in the wild-type temporal pattern in smc and dps null strains. The plasmid pJEZP5, containing the gyrB promoter fused to a promoter-less copy of lacZ, was moved into NA1000 (wt, top), JG3003 (middle) and JG3402 (bottom) by conjugation. Transcription from the gyrB promoter was monitored during synchronized growth of the two strains by immunoprecipitation of β-galactosidase pulse-labeled at intervals during the cell cycle. For each strain, the protein labeled during each five-minute interval is shown beneath a diagram of the corresponding stage in the cell cycle, as determined by light microscopy.
Table 1 Strains and Plasmids
Strain Reference
E. coli
S17-1 Rp4-2, Tc::Mu, Km::Tn7 [49]
TG-1 F_, lacIq, proA+B+ lacZ_M15 [61]
C. crescentus
NA1000 syn-1000 (synchronizable strain) [22]
JG3003 syn-1000, smc7 (spec resistance) [this study]
JG3402 syn-1000, dps2 (gent resistance) [this study]
Plasmids Reference
pBluescript-KS+ cloning vector [Stratagene]
placZ/290 pRK290 derivative containing a promoterless lacZ gene [48]
pNPTS139 sacB counter selection vectors; (kanr) [M.R.K. Alley]
pJEKS- gyrB500 500 bp, PstI/StuI gyrB fragment in pBluescript-KS+ [this study]
pJEKS-P3 904 bp PstI/HindIII gyrB promoter fragment [this study]
pJGZ1.65 1.65 kb PstI/EcoRI fragment in placZ/290, gyrB-lacZ transcriptional reporter fusion [17]
pJEZ1.1 1.3 kb PstI/HindIII fragment in placZ/290 [this study]
pJEZP2 942 bp PstI/HindIII gyrB-lacZ transcriptional reporter fusion [this study]
pJEZP3 904 bp PstI/HindIII gyrB-lacZ transcriptional reporter fusion [this study]
pJEZP2s 167 bp PstI/(native)StuI gyrB-lacZ transcriptional reporter fusion [this study]
pJEZP3s 132 bp PstI/(native)StuI gyrB-lacZ transcriptional reporter fusion [this study]
pJEZP4 180 bp PstI/HindIII gyrB-lacZ transcriptional reporter fusion [this study]
pJEZP5 80 bp PstI/HindIII gyrB-lacZ transcriptional reporter fusion [this study]
==== Refs
Ausmees N Jacobs-Wagner C Spatial and temporal control of differentiation and cell cycle progression in Caulobacter crescentus Annu Rev Microbiol 2003 57 225 247 14527278 10.1146/annurev.micro.57.030502.091006
Ryan KR Judd EM Shapiro L The CtrA response regulator essential for Caulobacter crescentus cell-cycle progression requires a bipartite degradation signal for temporally controlled proteolysis J Mol Biol 2002 324 443 455 12445780 10.1016/S0022-2836(02)01042-2
England JC Gober JW Cell cycle control of cell morphogenesis in Caulobacter Curr Opin Microbiol 2001 4 674 680 11731319 10.1016/S1369-5274(01)00268-5
Gober JW England JC Brun YV, Shimkets LJ Regulation of flagellum biosynthesis and motility in Caulobacter, Prokaryotic Development 2000 American Society for Microbiology, Washington, D. C 319 339
Quon KC Marczynski GT Shapiro L Cell cycle control by an essential bacterial two-component signal transduction protein Cell 1996 84 83 93 8548829 10.1016/S0092-8674(00)80995-2
Quon KC Yang B Domian IJ Shapiro L Marczynski GT Negative control of bacterial DNA replication by a cell cycle regulatory protein that binds at the chromosome origin Proc Natl Acad Sci USA 1998 95 120 125 9419339 10.1073/pnas.95.1.120
Domian IJ Quon KC Shapiro L Cell type-specific phosphorylation and proteolysis of a transcriptional regulator controls the G1-to-S transition in a bacterial cell cycle Cell 1997 90 415 424 9267022 10.1016/S0092-8674(00)80502-4
Domian IJ Reisenauer A Shapiro L Feedback control of a master bacterial cell-cycle regulator Proc Natl Acad Sci USA 1999 96 6648 6653 10359766 10.1073/pnas.96.12.6648
Marczynski GT Shapiro L Cell-cycle control of a cloned chromosomal origin of replication from Caulobacter crescentus J Mol Biol 1992 226 959 977 1518064 10.1016/0022-2836(92)91045-Q
Marczynski GT Lentine K Shapiro L A developmentally regulated chromosomal origin of replication uses essential transcription elements Genes Dev 1995 9 1543 1557 7601356
Kelly AJ Sackett MJ Din N Quardokus E Brun YV Cell cycle-dependent transcriptional and proteolytic regulation of FtsZ in Caulobacter Genes Dev 1998 12 880 893 9512521
Laub MT Chen SL Shapiro L McAdams HH Genes directly controlled by CtrA, a master regulator of the Caulobacter cell cycle Proc Natl Acad Sci USA 2002 99 4632 4637 11930012 10.1073/pnas.062065699
Laub MT McAdams HH Feldblyum T Fraser CM Shapiro L Global analysis of the genetic network controlling a bacterial cell cycle Science 2000 290 2144 2148 11118148 10.1126/science.290.5499.2144
Roberts RC Shapiro L Transcription of genes encoding DNA replication proteins is coincident with cell cycle control of DNA replication in Caulobacter crescentus J Bacteriol 1997 179 2319 2330 9079919
Winzeler E Shapiro L A novel promoter motif for Caulobacter cell cycle-controlled DNA replication genes J Mol Biol 1996 264 412 425 8969294 10.1006/jmbi.1996.0650
Zweiger G Shapiro L Expression of Caulobacter dnaA as a function of the cell cycle J Bacteriol 1994 176 401 408 8288535
Rizzo MF Shapiro L Gober JW Asymmetric expression of the Gyrase B gene from the replication-competent chromosome in the Caulobacter crescentus predivisional cell J Bacteriol 1993 175 6970 6981 8226640
Gomes S Gober JW Shapiro L Expression of the Caulobacter heat shock gene dnaK is developmentally controlled during growth at normal temperatures J Bacteriol 1990 172 3051 3059 2345134
Malakooti J Wang SP Ely B A consensus promoter sequence for Caulobacter crescentus genes involved in biosynthetic and housekeeping functions J Bacteriol 1995 177 4372 4376 7543475
Gober JW Shapiro L Temporal and spatial regulation of developmentally expressed genes in Caulobacter BioEssays 1991 13 277 283 10.1002/bies.950130604
Evinger M Agabian N Caulobacter crescentus nucleoid: analysis of sedimentation behavior and protein composition during the cell cycle Proc Natl Acad Sci USA 1979 76 175 178 284329
Evinger M Agabian N Envelope-associated nucleoid from Caulobacter crescentus stalked and swarmer cell J Bacteriol 1977 132 294 301 334726
Swoboda KK Dow CS Vitkovic L Nucleoids of Caulobacter crescentus CB15 J Gen Microbiol 1982 128 279 289
Keiler KC Shapiro L Conserved promoter motif is required for cell cycle timing of dnaX transcription in Caulobacter J Bacteriol 2001 183 4860 4865 11466289 10.1128/JB.183.16.4860-4865.2001
Avedissian M Lessing D Gober JW Shapiro L Gomes SL Regulation of the Caulobacter crescentus dnaKJ operon J Bacteriol 1995 177 3479 3484 7768857
Malakooti J Ely B Principal sigma subunit of the Caulobacter crescentus RNA Polymerase J Bacteriol 1995 177 6854 6860 7592478
Perez-Martin J de Lorenzo V Clues and consequences of DNA bending in transcription Annu Rev Microbiol 1997 51 593 628 9343361 10.1146/annurev.micro.51.1.593
Aiyar SE Gourse RL Ross W Upstream A-tracts increase bacterial promoter activity through interactions with the RNA polymerase α subunit Proc Natl Acad Sci USA 1998 95 14652 14657 9843944 10.1073/pnas.95.25.14652
Wolf SG Frenkiel D Arad T Finkel SE Kolter R Minsky A DNA protection by stress-induced biocrystallization Nature 1999 400 83 85 10403254 10.1038/21918
Antelmann H Engelmann S Schmid R Sorokin A Lapidus A Hecker M Expression of a stress- and starvation-induced dps/pexB-homologous gene is controlled by the alternative sigma factor σB in Bacillus subtilis J Bacteriol 1997 179 7251 7256 9393687
Martinez A Kolter R Protection of DNA during oxidative stress by the nonspecific DNA-binding protein Dps J Bacteriol 1997 179 5188 5194 9260963
Almirón M Link AJ Furlong D Kolter R A novel DNA-binding protein with regulatory and protective roles in starved Escherichia coli Genes Dev 1992 6 2646 2654 1340475
Hirano T The ABCs of SMC proteins: two-armed ATPases for chromosome condensation, cohesion, and repair Genes Dev 2002 16 399 414 11850403 10.1101/gad.955102
Graumann PL SMC proteins in bacteria: condensation motors for chromosome segregation? Biochimie 2001 83 53 59 11254975 10.1016/S0300-9084(00)01218-9
Cobbe N Heck MMS Review: SMCs in the world of chromosome biology – from prokayotes to higher eukaryotes J Struct Biol 2000 129 123 143 10806064 10.1006/jsbi.2000.4255
Hirano T SMC-mediated chromosome mechanics: a conserved scheme from bacteria to vertebrates? Genes Dev 1999 13 11 19 9887095
Strunnikov AV Jessberger R Structural maintenance of chromosomes (SMC) proteins: conserved molecular properties for multiple biological functions Eur J Biochem 1999 263 6 13 10429180 10.1046/j.1432-1327.1999.00509.x
Graumann PL Losick R Strunnikov AV Subcellular localization of Bacillus subtilis SMC, a protein involved in chromosome condensation and segregation J Bacteriol 1998 180 5749 5755 9791128
Azam TA Iwata A Nishimura A Ueda S Ishihama A Growth phase-dependent variation in protein composition of the Escherichia coli nucleoid J Bacteriol 1999 181 6361 6370 10515926
McLeod SM Johnson RC Control of transcription by nucleoid proteins Curr Opin Microbiol 2001 4 152 159 11282470 10.1016/S1369-5274(00)00181-8
Falconi M Colonna B Prosseda G Micheli G Gualerz CO Thermoregulation of Shigella and Escherichia coli EIEC pathogenicity: a temperature-dependent structural transition of DNA modulates accessibility of virF promoter to transcriptional repressor H-NS EMBO J 1998 17 7033 7043 9843508 10.1093/emboj/17.23.7033
Dorman CJ Deighan P Regulation of gene expression by histone-like proteins in bacteria Curr Opin Genet Dev 2003 13 179 184 12672495 10.1016/S0959-437X(03)00025-X
Oberto J Drlica K Rouvière-Yaniv J Histones, HMG, HU, IHF: Même combat Biochimie 1994 76 901 908 7748933 10.1016/0300-9084(94)90014-0
Britton RA Lin DC-H Grossman AD Characterization of a prokaryotic SMC protein involved in chromosome partitioning Genes Dev 1998 12 1254 1259 9573042
Moriya S Tsujikawa E Hassan AK Asai M Kodama KT Ogasawara N A Bacillus subtilis gene-encoding protein homologous to eukaryotic SMC motor protein is necessary for chromosome partition Mol Microbiol 1998 29 178 187 10.1046/j.1365-2958.1998.00919.x
Jensen RB Shapiro L The Caulobacter crescentus smc gene is required for cell cycle progression and chromosome segregation Proc Natl Acad Sci USA 1999 96 10661 10666 10485882 10.1073/pnas.96.19.10661
Altuvia S Almiron M Huisman G Kolter R Stortz G The dps promoter is activated by OxyR during growth and by IHF and σS in stationary phase Mol Microbiol 1994 13 265 272 7984106
Gober JW Shapiro L A developmentally regulated Caulobacter flagellar promoter is activated by 3' enhancer and IHF binding elements Mol Biol Cell 1992 3 913 926 1392079
Simons R Priefer U Puhler A A broad host range mobilization system for in vivo genetic engineering: transposon mutagenesis in gram negative bacteria Bio/Technology 1983 1 784 790 10.1038/nbt1183-784
Sambrook J Fritsch EF Maniatis T Molecular cloning: a laboratory manual 1989 2 Cold Spring Harbor Laboratory Press, Cold Spring Harbor, New York
Poindexter JS Biological properties and classification of the Caulobacter group Bacteriol Rev 1964 28 231 295 14220656
Johnson RC Ely B Isolation of Spontaneously Derived Mutants of Caulobacter crescentus Genetics 1977 86 25 32 407126
Ely B Johnson RC Generalized transduction in Caulobacter crescentus Genetics 1977 87 391 399
Schweizer HP Allelic exchange in Pseudomonas aeruginosa using novel ColE1-type vectors and a family of cassettes containing a portable oriT and the counter-selectable Bacillus subtilis sacB marker Mol Microbiol 1992 6 1195 1204 1588818
Alexeyev MF The pKNOCK series of broad-host-range mobilizable suicide vectors for gene knockout and targeted DNA insertion into the chromosome of gram-negative bacteria Biotechniques 1999 26 824 828 10337469
Kunkel TA Robert JD Zakour RA Rapid and efficient site-specific mutagenesis without phenotypic selection Methods Enzymol 1987 154 367 382 3323813
Sanger F Nicklen S Coulson AR DNA sequencing with chain-terminating inhibitors Proc Natl Acad Sci USA 1977 74 5463 5467 271968
Mangan EK Bartamian M Gober JW A mutation that uncouples flagellum assembly from transcription alters the temporal pattern of flagellar gene expression in Caulobacter crescentus J Bacteriol 1995 177 3176 3184 7768816
Miller JH Experiments in molecular genetics 1972 Cold Spring Harbor Laboratory, Cold Spring Harbor, NY
Amemiya K Bellofatto V Shapiro L Feingold J Transcription initiation in vitro and in vivo at a highly conserved promoter within a 16 S ribosomal RNA gene J Mol Biol 1986 187 1 14 2420995 10.1016/0022-2836(86)90401-8
Ausubel FM Brent R Kingston RE Moore DD Seidman JG Smith JA Struhl K Current Protocols in Molecular Biology 1996 John Wiley and Sons, Inc., New York, NY
Carter P Beduouelle H Winter G Improved oligonucleotide site-directed mutagenesis using M13 vectors Nucleic Acids Res 1985 13 4431 4443 2989795
| 15904494 | PMC1175088 | CC BY | 2021-01-04 16:03:40 | no | BMC Microbiol. 2005 May 17; 5:25 | utf-8 | BMC Microbiol | 2,005 | 10.1186/1471-2180-5-25 | oa_comm |
==== Front
BMC MicrobiolBMC Microbiology1471-2180BioMed Central London 1471-2180-5-311591346210.1186/1471-2180-5-31Research ArticleGenetic islands of Streptococcus agalactiae strains NEM316 and 2603VR and their presence in other Group B Streptococcal strains Herbert Mark A [email protected] Catriona JE [email protected] David [email protected] Emmelien [email protected] Nicola [email protected] Lori AS [email protected] Nigel J [email protected] University Departments of Paediatrics, John Radcliffe Hospital, Headington, Oxford, OX3 9DU, UK2 Department of Microbiology, John Radcliffe Hospital, Headington, Oxford, OX3 9DU, UK3 Bacterial Pathogenesis and Functional Genomics Group, The Sir William Dunn School of Pathology, University of Oxford, South Parks Rd, Oxford, OX1 3RE, UK2005 24 5 2005 5 31 31 10 12 2004 24 5 2005 Copyright © 2005 Herbert 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
Streptococcus agalactiae (Group B Streptococcus; GBS) is a major contributor to obstetric and neonatal bacterial sepsis. Serotype III strains cause the majority of late-onset sepsis and meningitis in babies, and thus appear to have an enhanced invasive capacity compared with the other serotypes that cause disease predominantly in immunocompromised pregnant women. We compared the serotype III and V whole genome sequences, strains NEM316 and 2603VR respectively, in an attempt to identify genetic attributes of strain NEM316 that might explain the propensity of strain NEM316 to cause late-onset disease in babies. Fourteen putative pathogenicity islands were described in the strain NEM316 whole genome sequence. Using PCR- and targeted microarray- strategies, the presence of these islands were assessed in a diverse strain collection including 18 colonizing isolates from healthy pregnant women, and 13 and 8 invasive isolates from infants with early- and late-onset sepsis, respectively.
Results
Side-by-side comparison of the strain NEM316 and strain 2603VR genomes revealed that they are extremely similar, with the only major difference being the capsulation loci and mobile genetic elements. PCR and Comparative Genome Hybridization (CGH) were used to define the presence of each island in 39 GBS isolates. Only islands I, VI, XII, and possibly X, met criteria of a true pathogenicity island, but no significant correlation was found between the presence of any of the fourteen islands and whether the strains were invasive or colonizing. Possible associations were seen between the presence of island VI and late-onset sepsis, and island X and early-onset sepsis, which warrant further investigation.
Conclusion
The NEM316 and 2603VR strains are remarkable in that their whole genome sequences are so similar, suggesting that the capsulation loci or other genetic differences, such as pathogenicity islands, are the main determinants of the propensity of serotype III strains to cause late-onset disease. This study supports the notion that GBS strain NEM316 has four putative pathogenicity islands, but none is absolutely necessary for disease causation, whether early- or late-onset sepsis. Mobile genetic elements are a common feature of GBS isolates, with each strain having its own peculiar burden of transposons, phages, integrases and integrated plasmids. The majority of these are unlikely to influence the disease capacity of an isolate. Serotype associated disease phenotypes may thus be solely related to differences in the capsulation loci.
==== Body
Background
Streptococcus agalactiae (Group B Streptococcus, GBS) is a Gram positive, facultative anaerobic bacterium that is the most common cause of neonatal and obstetric sepsis, and is an increasingly important cause of septicaemia in elderly and immunocompromised patients [1]. Serotype III GBS causes approximately 37% of early-onset and 67% of late-onset neonatal GBS sepsis (compared with 13% and 5%, respectively, caused by serotype V), and is the predominant serotype causing late-onset meningitis [1,2]. Serotype V prevails in invasive infection in non-pregnant adults (causing 29% of all such infections) [3]. The genetic determinants of the propensity of serotype III GBS to cause late-onset sepsis and meningitis have not been fully elucidated, but the availability of whole genome sequences of a serotype III isolate (strain NEM316) and a serotype V isolate (strain 2306VR) brings this prospect closer [4,5]. One possibility is that the serotype III GBS has pathogenicity islands (PAIs) that are not present in the other serotypes, and which confer an enhanced invasive potential. Glaser et al. [4] described fourteen regions of strain NEM316 that they considered to be putative PAIs. These islands are composed of 11 to 77 genes and contain most of the mobile elements in the NEM316 genome [4]. Six of the islands are adjacent to tRNA genes, a feature of pathogenicity islands [6], and many known or putative virulence genes of GBS are contained within these regions. For instance, alp2 [7] is in 'island IV', the cyl operon [8] is in 'island VI', and lmb and scpB are in 'island XII' [9]. PAIs are defined by the following criteria: (1) they carry one or more virulence genes, (2) they are present in the genome of pathogenic bacterium but absent in non-pathogenic representatives of the same species, (3) they are frequently located adjacent to tRNA genes, (4) they are associated with mobile genetic elements and are often flanked by direct repeats (DR), (5) they are unstable and either the whole of the PAI or part of it may be deleted, and (6) often represent mosaic like structures rather than homogenous segments of horizontally acquired DNA [10].
We used the C. elegans database genome sequence graphical interface (AceDB) [11,12] to compare the strain NEM316 and the strain 2603VR genome sequences to identify serotype III and V genomic differences, and to further define the putative PAIs in the NEM316 serotype III strain. We then conducted PCR amplification and targeted microarray-based comparative genome hybridization (CGH) studies aimed at delineating the nature of the putative PAIs.
Results
NCBI and AceDB analysis of the sequenced serotype III and V genomes
Side-by-side comparison of the serotype III and V genome sequences, strains NEM316 and 2603VR respectively, identified numerous annotation differences between open reading frames, most generated by true or sequencing error frame shifts and differences in the annotation of initiation codons. The similarity of the two genomes is otherwise remarkable (see figure 1). The other major differences between the two genomes are the capsulation loci and the presence of multiple mobile elements including integrated plasmids, prophages, transposons, and one to two gene integrases/transposases. Much of this acquired DNA appears to be unique to each sequenced strain (represented by triangles in figure 1), in the type of mobile element but not necessarily the genomic location.
Which islands appear to be real PAIs?
PAIs contain virulence and mobilization genes and are flanked by direct repeat (DR) sequences that are recognised by mobilization proteins [10]. Potential PAIs must be distinguished from non-mobile regions of the chromosome that contain virulence genes adjacent to tRNA genes, and which have merely attracted mobile elements. Such mobile elements may themselves be genomic or metabolic islands but by definition they are not PAIs, unless they mobilize virulence genes and are associated with pathogenic strains [10].
Our annotation of the putative PAIs is given in table 1. The putative PAIs are present in both strain NEM316 and strain 2603VR, with the exception of islands III, VII, VIII and X, which are only present in strain NEM316. Islands III, VII, and VIII were described as identical copies of a chromosomally integrated plasmid, designated pNEM316-1. Two further islands are present in strain 2603VR that are not present in strain NEM316: sag0915-0937 (a copy of Tn916) and sag1835-1886 (a prophage). None of these mobile elements contain known virulence genes, and they may therefore not be true PAIs.
Inserted into the ends of islands II, IV, V, XI, XIII and XIV, and the middle of island VI, are mobile elements that contain phage or transposon genes, but no known virulence genes (see table 1). The mobile elements in strain NEM316 are different from those in strain 2603VR at each of these sites of insertion. The putative PAIs do not otherwise contain mobilization genes or flanking DR sequences. Island IX contains a two-component regulator, but has no mobilization genes. These putative PAIs may therefore merely represent non-mobile regions of the genome into which phages and transposons have inserted.
Islands I, VI and XII contain virulence genes (rgg[4,13,14], the cyl locus, and lmb/scpB[9], respectively) flanked by mobilization genes that are present in both sequenced strains. Island X contains mobilization genes and is presumed to be mobile because it is only found in strain NEM316 and not strain 2603VR. It also contains genes encoding surface proteins that have an LPXTG signal sequence; these could potentially have a virulence role. Four regions of the GBS genome (islands I, VI, X and XII) may therefore be real PAIs.
Other genomic differences
Aside from annotation discrepancies, mobile elements and the capsulation loci, there are few other differences between the two sequenced genomes. There is a possible lone example of a Minimal Mobile Element (MME) [15]. Two genes present between purK and purB (genes involved in purine metabolism) in strain NEM316 (gbs0045-0046) compared with a single different gene in the same location in strain 2603VR (sag0046). The putative MME was PCR amplified in each of the 39 strains in our collection. Only two insert types were amplified, of 2,036 bp and 1,636 bp. Representatives of these were sequenced and found to have the exact sequence of either gbs0045-0046 or sag0046, respectively. All strains had either one or other insert between purK and purB. Other inserts between purK and purB are identifiable in the genome sequences of other pathogenic streptococci (figure 2), hence fulfilling the criteria for an MME.
Another disparity between the two GBS whole genome sequences is the gene gbs0048 (a Cro/CI transcriptional regulator) in strain NEM316, which has a different proximal half compared to its homologue sag0048 in strain 2603VR.
The presence of putative pathogenicity islands as defined by CGH
The presence of putative pathogenicity islands as defined by CGH. Results of PCR (figure 3) and CGH (figure 4) analyses. The genes and GBS strains shaded grey in table 2 are those included in CGH experiments, table 3. Sag0001 encodes dnaA, a positive control for PCR. The NEM316 strain is a positive control. The pNEM316-1 plasmid is located three times in the NEM316 genome, and in figure 1 is represented as 'islands III, VII and VIII'. The strains are divided into three groups: colonizing strains from healthy pregnant women, and strains causing early- and late-onset sepsis in babies; and are sub-divided into those strains for which we have PCR results, and those for which we have PCR and CGH data.
Some putative PAIs are almost always present in the strain collection
Islands II, IV, V, IX and XII-XIV are almost always present in every strain from our strain collection. An occasional gene could not be amplified in one or more strains. For instance, sag1246, located in the distal half of island XII, could not be amplified in strains J99, B9, MK3, M1, and J87 (see figure 3). However, sag1233, located in the proximal half of island XII, could be amplified in all strains. The whole genome comparison of strains NEM316 and 2603VR revealed inter-strain sequence divergence at the distal end of island XII, whereas the proximal end of island XII, containing lmb/scpB (sag1234-1235), is highly conserved between these two strains (see table 1). Amplification of sag1233 therefore best reflects the presence of the putative PAI. Sag1233 may be particularly hard to PCR amplify because sequence divergence affects primer annealing. Similar sequence divergence between strains may also explain our inability to amplify occasional genes in islands V and XIII.
These islands are present in all strains tested, whether isolated from disease or colonizing sites and therefore do not meet the PAI definition criteria (2) and (5), above: that they should be present in pathogenic but absent from non-pathogenic strains, and they should be unstable and delete with distinct frequencies. Colonizing is not necessarily synonymous with non-pathogenic, a fact that confounds interpretation of a genetic comparison of invasive and colonizing strains.
Some putative PAIs are almost always absent in the strain collection
Copies of pNEM316-1, represented by islands III, VII and VIII, are only found in strain NEM316, and are consistently absent from the other strains in our collection. Five genes were amplified from island X. They were only all consistently amplifiable from strain NEM316. Two or three genes from island X, however, were amplified in 4 strains other than NEM316, reflecting either the part presence of the island in these strains or marked sequence divergence. These islands, that are absent from most disease causing strains, are unlikely to be PAIs. However, the central part of island X is present in 5 strains known to have caused early-onset sepsis, and absent from all colonizing and late-onset sepsis strains.
Some putative PAIs are variably present in the strain collection
Two genes amplified from each of islands I and VI revealed a variable presence of these islands in the strains of our collection (see figure 3). Island I is at least part-present in 14 of 18 colonizing strains (78%), 8 of 13 early-onset sepsis strains (61%), and 6 of 8 late-onset sepsis strains (75%). Although island I meets the PAI criteria of being variably present in the species, there is no relationship between the whole or part presence of the island and whether the strain was colonizing or disease causing. The two genes amplified from island I were sag0224 and sag0234. Sag0234 is close to the only recognisable virulence gene in island I, rgg (sag0239; homologue of a virulence regulator in S. pyogenes), and thus amplification of this gene reflects the presence of the most important part of the island. Sag0234 homologues are present in 13 of 18 colonizing strains (72%), in 7 of 13 early-onset sepsis strains (53%), and in 6 of 8 late-onset sepsis strains (75%). Thus, in this relatively small collection, there is no relationship between the presence of the distal half of island I and whether the strain was a colonizing or disease causing isolate.
Island VI is at least part present in all colonizing strains, 12 of 13 early-onset sepsis strains, and all late-onset sepsis strains. There is therefore no relationship between the island and disease. The proximal marker gene sag0645 is closer to the cyl locus (encoding the β-hemolysin, a major contributor to virulence in GBS) than the distal marker gene, sag0685, and therefore possibly better reflects the presence of a PAI that contains Tn5252 transposon genes and the cyl locus. Sag0645 is present in 14 of 18 colonizing strains (78%), in 9 of 13 early-onset sepsis strains (69%), and in 7 of 8 late-onset sepsis strains (87.5%). Although these differences are not statistically significant, there is a trend towards the presence of this putative PAI in late-onset sepsis strains. A larger study is required to bear out this finding.
Comparative genomic hybridization analysis
Comparative genomic hybridisation (CGH) analysis was performed on 22 of the 39 strains assessed by PCR. These 22 strains were randomly selected and included 15 of the 18 colonizing strains, 3 of the 13 isolates that caused early-onset sepsis, and 4 of the 8 strains that caused late-onset sepsis.
For probes to the island genes, the results of CGH (figure 4) are near identical to those of PCR (figure 3), with only a few exceptions. Notable is the hybridization of strains Z50 and K1 DNA to the gbs0367 gene probe, suggesting that this gene, and therefore possibly the whole or part of pNEM316-1, is present in these strains. However, the presence of pNEM316-1 was not detected by PCR in these or any other strain except the control NEM316 strain. Thus, perhaps the gene sequence of pNEM316-1 is divergent in strains Z50 and K1 so that the primers for PCR were unable to anneal, or that CGH detected a similar gene to gbs0367. We propose that similar reasons account for the other few discrepancies that exist between the PCR and CGH results. In general, however, the CGH and PCR results are highly consistent.
Although not the main focus of this study, the presence of the other genes for which probes were included on the sub-microarray was also assessed by CGH. Eighty five percent of all the 384 probes included on the sub-microarray gave strong hybridization signals for all strains tested, indicating that at least for the region of the gene chosen for the probe design there is very little variability between the strains. However, hybridization to 15% of the probes was variable in at least three of the 22 strains tested. In most instances there was no probe hybridization, but occasionally the hybridization signal was reduced, suggesting sequence variation within the probe region. The genes with presumed sequence divergence encoded six sortases, ten proteins with an LPXTG signal sequence, two clp proteases, one ABC transporter and five PTS proteins, thirteen putative or known regulators, and sixteen other proteins (see table 2). Of these, several are genes with possible virulence enhancing roles (highlighted bold in table 2), including three virulence regulators rgf (sag1956-7) [16], a putative rofA-like protein (RALP, sag1463) [17] and rogB (sag1409) [18], two genes in the cyl operon (sag0662 and sag0664) [19], cfb (sag2043) encoding the CAMP factor [20], and pavA (sag1190; adherence and virulence protein A) [21], and are therefore worthy of further disease association studies. Of note, putative homologues of the major virulence regulators of Streptococcus pyogenes [22], such as mga (sag0277), rofA/nra (sag1356, sag1359, sag1409, and sag1463), and rgg/ropB (sag1490, sag2158), and all the other identifiable regulators included on the array (reviewed by Herbert et al [23]) are non-variable in their hybridization pattern, across the strain collection.
Discussion
By combining the results of genome comparison and PCR/CGH analysis we can make the following arguments about the likelihood that each of the putative PAIs is a true PAI:
Island I may be a true PAI. It contains the virulence regulator rgg, which is flanked by mobilization genes, and the whole island is variably present in strains of our collection. It does not appear to be found preferentially in GBS isolates that are known to have caused disease, but the number of isolates tested in this study may be too small to tease out small contributions of a PAI to invasiveness. A confounding factor is that the colonizing isolates in our collection may have the capacity to cause disease. Thus, our colonizing and disease isolates do not simply reflect non-pathogenic and pathogenic strains, respectively. This study is not powered to identify small contributions of a putative PAI to the propensity of serotype III to cause late-onset sepsis. Only a very large study is likely to do this.
Island II is unlikely to be a true PAI. In strains NEM316 and 2603VR there are two different mobile elements inserted at the same relative genomic location into the proximal end of the island, neither of which appears to harbour virulence genes. This suggests that the proximal end of island II is a hot spot for the insertion of mobile elements. Furthermore, the distal half of island II does not appear to have mobilization machinery and is present in all the strains within our collection. Islands III, VII and VIII are near-identical copies of a chromosomally integrated plasmid, pNEM316, which contains no known virulence determinants and which is only present in strain NEM316, and is not present within other strains within our collection. Thus, this plasmid is unlikely to be a PAI.
Islands IV and V are unlikely to be PAIs for the same reason as island II. Island VI may be a true PAI as it contains the cyl locus adjacent to Tn5252 (present in both strains NEM316 and 2603VR), has a mosaic-like structure, and is variably present in our strain collection. We cannot show a relationship between the presence of island VI and strains causing disease, but this may be due to limitations of the power of this study. Island IX does not contain mobilization genes and is present in all strains within our collection, making it unlikely that it is a PAI. Island X is mobile, but does not contain obvious virulence determinants. The whole of island X is only found in strain NEM316, and parts of it within four other strains causing early-onset sepsis. There may thus be an association between the middle of island X, gbs1125-1135, and the capacity of an isolate to cause chorioamnionitis. Early-onset sepsis in a newborn baby reflects invasive disease in a pregnant mother, whereas the fetus is merely a vulnerable secondary host. The potential association between island X and early-onset sepsis needs a larger study for clarification. Island XI is mostly composed of a small mobile element present in strain NEM316, but not strain 2603VR, and does not contain known virulence genes.
Island XII contains mobilization and virulence genes, has a mosaic like structure, the distal end of it is variably present in strains of our collection. It could therefore be a PAI. Our study does not have the power to identify an association between the presence of the island and disease. Islands XIII and XIV are unlikely to be PAIs for the same reason as island II.
Conclusion
The majority of late-onset meningitis, and to a lesser extent late-onset sepsis, is caused by serotype III strains. There is likely to be a bacterial genetic basis for this invasive propensity. A comparison of the whole genome sequences of a serotype III isolate, NEM316, and a serotype V isolate, 2603VR, is remarkable in the degree of similarity of the two strains, but there are some dissimilarities. These include open reading frame annotation discrepancies, genes that show sequence divergence between strains, an MME, mobile DNA, and the capsulation loci. This study contributes to our understanding of pathogenesis by further delineating the nature of mobile elements in GBS. Individual GBS isolates probably carry their own unique aliquot of horizontally acquired genetic material. Only four (islands I, VI, X and XII) of 14 putative PAIs are likely to be real PAIs, but there is no absolute association of any of these four PAIs with strains causing disease. The strongest possible disease association is with island X and early-onset sepsis.
Methods
Strains and culture conditions
GBS isolates were cultured overnight in Todd-Hewitt broth (Oxoid). DNA was extracted from 39 isolates of GBS (table 3): 18 colonizing strains; 13 strains derived from babies with early-onset sepsis (early-onset sepsis); and 8 strains from babies with late-onset sepsis. The control strain was NEM316 (CIP82.45, Collection de l'Institut Pasteur).
Genome comparisons
GBS serotype III strain NEM316 and serotype V strain 2603VR genome sequences were compared through NCBI [24,25] and using AceDB [11,12], hosted by the University of Oxford Bioinformatics Centre [26]. Additional information on domains and homologies were obtained through NCBI BLAST searches [27] and the NCBI Conserved Domain Search [28].
Molecular Methods
DNA was extracted from a 3 ml culture of each strain using spin column technology (DNAeasy; Qiagen), following the manufacturer's recommendations with the exception that lysozyme was replaced by mutanolysin (50 units per extraction) and the cell pellet was pre-incubated with this enzyme for 60 minutes at 37°C.
Double strand sequencing was conducted by the Department of Biochemistry Core Sequencing Facility, University of Oxford, using the same primers employed for the PCR using gel extracted (Qiagen) templates. Sequencing reactions used Big Dye version 3 (Applied Biosciences) and were analyzed on an ABI377 sequencer. Sequences were assembled, evaluated, and interpreted using Chromas v2.3 (Technelysium Pty Ltd) and ClustalW [29].
PCR analysis
A standard PCR condition, Taq DNA polymerase (Roche) with 1.5 mM Mg2+, gene-specific primers (table 4) and an annealing temperature of 56°C, was established for amplifying one to five genes from each island in the NEM316 control strain, and the same PCR conditions were used to attempt amplification in the other 38 strains in our collection. The presence of a correct size amplicon was used as a surrogate marker of the presence of the whole island. When an amplicon was not obtained from a strain, the PCR was repeated with lower stringency conditions, by increasing to 2.5 mM Mg2+ and decreasing the primer annealing temperature to 52°C. For all 39 strains, the gene dnaA (gbs0001, sag0001) was successfully PCR amplified, indicating that there were no significant PCR inhibitors in our DNA preparations. Consistent results were achieved with the PCR independently performed twice (by DM and EA). We did not attempt to amplify a gene from 'island XI' as our genome alignment and annotation clearly identified that the major part of this island was a small prophage found in NEM316 but not 2603VR. For MME amplification, primers were designed to the 3'-end of purK and the 5'-end of purB.
Comparative genome hybridization
Fifteen gene-specific probes from within the islands were incorporated into a 384-probe GBS sub-microarray being developed to study regulatory networks in GBS (unpublished). The other probes were designed from 369 genes representing all the identifiable regulators (including homologues of Streptococcus pyogenes regulators such as rofA, rggB, mga), all the known GBS virulence factors, stress adaptation molecules, and proteins with LPXTG sorting signals, and many transporters (focussing on ABC and PTS systems). Probe regions were chosen using AceDB [11,12], so that where the gene was present in both strains, a region of greater than 300 bp region was chosen that was near identical in each of the sequenced genomes that was devoid of repetitive elements. Primers were designed using Primer3 [30], with the product size set at an optimum of 300 bp (range 150–450 bp), the primer size at 19 bp (range 17–21), the primer Tm set at 58°C (range 54–63), the primer GC% at 40 (range 30–80), and the GC clamp option set to 1. The primers were synthesised commercially (Operon). See Additional file 1 for sequence information. Amplicons were generated using DNA extracted from the sequenced serotype III strain NEM316 [4]. The printed probes were amplified from a 1:50 dilution of these products by second-round PCR using the first-round primers, once a single band of the correct size had been obtained from the first reaction, a similar single band was confirmed from the second round PCR. PCR products were checked using 96-well E-gels (Invitrogen). Probes were spotted onto Genetix amine microarray slides in Genetix amine spotting solution for amine slides using a Qarray Mini microarray printer (Genetix) using 150 micron tipped solid tungsten pins (Genetix). FluoroLink™ Cy3-dCTP and FluoroLink™ Cy5-dCTP (Amersham Pharmacia Biotech) were incorporated into 10 μg of chromosomal DNA using random hexamer primers (Invitrogen) and DNA polymerase I, Klenow fragment (Bioline, UK). Labelled DNA:DNA probe microarray hybridizations were conducted in 4x SSC, 0.2875% SDS at 65°C overnight.
Of the 384 gene-specific probes included on the array, seven probes were directed at serotype-specific capsular polysaccharide synthesis genes and were thus hybridization controls. Another five probes were directed at genes in pNEM316-1 and 'island X' that only infrequently hybridized to the DNA from the strains in the collection.
The probes were chosen in gene regions that were of low complexity, contained no repeats, and were identical according to our alignment of the NEM316 and 2603VR genomes using AceDB.
Identification of strain differences within the non-island genes was not the initial purpose of this study. However, such variation, in the context of the relative paucity of differences found in islands genes, indicates that allelic variants of the non-island genes may explain differences in strain behaviour. A larger scale project directed at these genes, using a microarray based upon a greater number of genome sequences than were available for this project, is needed to specifically investigate this type of divergence.
Probes to genes from each of islands I-XIV were included on the array, with the exception of 'islands IX and XI'. PCR analysis demonstrated that the 'island IX' region is consistently present in all strains in our collection, and our analysis did not support the notion that it contains mobile DNA. The major part of 'Island XI' is a small prophage, and we therefore expected it not to be relevant to the virulence of the organism. One probe was included to each 'island', except for two probes to 'islands I and V', and five probes to 'island X'.
Authors' contributions
MAH did the genome comparisons, designed the primers, probes and the sub-microarray, participated and supervised the PCR and CGH experiments, and drafted the manuscript. CJBB participated in the CGH experiments and analysed the CGH data. DMcC participated in PCR and CGHs experiments, and amplified and purified the probes for the microarray. EA helped with genome comparisons and identification of array probes, and participated in PCR experiments. NJ provided the strains (prior to publication of MLST), primers and protocols. LASS established the microarray protocols, provided scientific advice, undertook gene annotations and helped with drafting the manuscript. NJS facilitated the use of AceDB, set up the microarray infrastructure, and advised on MMEs and other aspects of the work, and helped with drafting the manuscript.
Supplementary Material
Additional File 1
submicroarray primers. File containing the sequences for the primers used for probe PCR.
Click here for file
Acknowledgements
CJBB is supported by a Wellcome Trust Entry Level Fellowship and LASS by a Wellcome Trust Project grant awarded to NJS. This project was also part-funded by OHSRC research grant 778.
Figures and Tables
Figure 1 A representation of the serotype III (NEM316; gbs001-2136) and serotype V (2603VR; sag001-2175) genomes (diagrammatic and not to scale). The genome sequences are mostly identical (represented by a horizontal line), triangles above the line represent gene regions unique to NEM316, and the triangles below are those present only in 2603VR. Boxed regions are putative PAIs (marked I to XIV). Grey bars with the PAIs represent genes amplified as surrogate markers for the presence of the whole island. Similar information can now be visualised through GenePlot, the NCBI pairwise comparison of protein homologs .
Figure 2 An example of an MME in GBS. Different intergenic regions are depicted between purK (pale blue block) and purB (lavender block) in various streptococcal species. Homologs of gbs0045 are indicated with an asterix. Hypothetical proteins are designated 'hypo'.
Figure 3 The presence of putative pathogenicity islands as defined by PCR. Results of PCR (figure 3) and CGH (figure 4) analyses. The genes and GBS strains shaded grey in table 2 are those included in CGH experiments, table 3. Sag0001 encodes dnaA, a positive control for PCR. The NEM316 strain is a positive control. The pNEM316-1 plasmid is located three times in the NEM316 genome, and in figure 1 is represented as 'islands III, VII and VIII'. The strains are divided into three groups: colonizing strains from healthy pregnant women, and strains causing early- and late-onset sepsis in babies; and are sub-divided into those strains for which we have PCR results, and those for which we have PCR and CGH data.
Figure 4 The presence of putative pathogenicity islands as defined by CGH. Results of PCR (figure 3) and CGH (figure 4) analyses. The genes and GBS strains shaded grey in table 2 are those included in CGH experiments, table 3. Sag0001 encodes dnaA, a positive control for PCR. The NEM316 strain is a positive control. The pNEM316-1 plasmid is located three times in the NEM316 genome, and in figure 1 is represented as 'islands III, VII and VIII'. The strains are divided into three groups: colonizing strains from healthy pregnant women, and strains causing early- and late-onset sepsis in babies; and are sub-divided into those strains for which we have PCR results, and those for which we have PCR and CGH data.
Table 1 The structure of 14 putative PAIs.
Putative PAI Position No. of genes Characteristics
I gbs0211-0235 25 Island I is adjacent to tRNA-Ala, begins with a phage integrase family site-specific recombinase (gbs0211) and a Cro/CI transcriptional regulator (gbs0212), and harbors other mobilization genes. The region contains rgg (gbs0230), a homologue of a virulence regulator in S. pyogenes.
II gbs0236-0254 19 Inserted into the proximal end of island II, adjacent to tRNA-Leu, are 9 genes in strain NEM316 (gbs0236-0244) not present in strain 2603VR, and 7 genes in strain 2603VR (sag0245-0251) not present in strain NEM316. Gbs0236-0244 consists of genes encoding a phage integrase and other phage proteins. Sag0245-0251 consists of genes encoding hypothetical proteins and a Cro/CI family regulator. Neither of these two regions harbor known virulence genes. In the remainder of island II (gbs0245-0254 or sag0252-0264), there are no other mobilization genes or known virulence genes.
III, VII, VIII gbs0361-0410
gbs0692-0740
gbs0969-1016 50
49
48 Islands III, VII, VIII are near-identical copies of a chromosomally integrated plasmid, designated pNEM316-1.
IV gbs0458-0482 (gbs0458-0486*) 25 (29)* Gbs0458-0470 (or sag0423-0433) contains several transcriptional regulators, including araC family members, and the virulence factor alp, but does not harbor any identifiable mobilization genes. Inserted into the distal end of island IV, adjacent to tRNA-Thr, are 16 genes in strain NEM316 (gbs0471-0486) not present in strain 2603VR, and 6 genes in strain 2603VR (sag0434-0439) not present in strain NEM316. Gbs0471-0486 contains an integrase (gbs0482) and a Cro/CI transcriptional regulator (gbs0475). Sag0434-0439 contains an IS256 family transposase (sag0434), a phage family site-specific recombinase (sag0438). Neither of these elements contain known virulence genes.
V gbs0588-0598 11 Inserted into the proximal end of island V, adjacent to tRNA-Arg, is a single gene in strain NEM316 (gbs0588; an integrase) that is not present in strain 2603VR, and 65 genes in strain 2603VR (sag0545-0609) that are not present in strain NEM316. Sag0545-0609 contains numerous prophage lambda genes. The remainder of island V (gbs0589-0598 or sag0610-0617) harbors genes encoding a cell membrane protein complex and a two-component regulator, vncSR, flanked by two transposase genes (for instance, sag0611 a degenerate transposase and sag0618 a truncated transposase). There are no genes known to be involved in virulence in island V.
VI gbs0616-0678 63 Island VI contains the cyl locus (gbs0644-0655; sag0662-0673), encoding a β-hemolysin that has been shown unequivocally to be involved in virulence. The region preceding the cyl locus (gbs0616-0639) in strain NEM316 contains Tn5252 transposon genes, and is identical in strain 2603VR (sag0636-0657). Downstream of the cyl locus, in strain NEM316, there are neither mobilization genes nor other known virulence genes. In the middle of the island, three genes in strain NEM316 (gbs0656-0658; encoding a permease and hypothetical proteins) are not present in strain 2603VR, and 10 genes in strain 2603VR (sag0674-0683; protease, endopeptidase and permease genes) are not present in strain NEM316. The distal half of island VI contains genes encoding core metabolic enzymes, and does not contain mobile elements or virulence determinants.
IX gbs1049-1076 28 Island IX contains genes with homology to those encoding a two-component regulatory system, a carbon starvation protein, and secreted proteins, but it does not contain any mobilization genes.
X gbs1118-1153 (gbs1118-1152*) 36 (35)* Island X appears to be mobile in that it is present in strain NEM316 but not in strain 2603VR, and it contains transferase, relaxase and some genes homologous with those in Tn5252. It also contains 3 LPXTG genes and a DNA methyltransferase. There are no known virulence genes.
XI gbs1214-1224 11 Island XI is composed of three genes that are present in both strains NEM316 and 2603VR, and these are involved in murein hydrolase export. Eight genes in island XI are present in strain NEM316, but not in strain 2603VR. One of these is an integrase, and the element is adjacent to a tRNA gene. None of the 8 genes appears to have a role in virulence.
XII gbs1296-1373 78 Island XII is a good candidate for a pathogenicity island. The virulence genes lmb (gbs1307), and scpB (gbs1308), encoding laminin binding protein and C5a peptidase, respectively, are at the proximal end of island XII, and are part of a large compound transposon. Upstream of lmb/scpB, gbs1296-1306, are five transposon (ISSdy1) or phage related genes, and downstream of lmb/scpB, gbs1309-1313 and gbs1338-1340, are other transposon (Tn5252) genes. In the distal half of island XII, 24 genes are present in strain NEM316 (gbs1314-1337; encoding phage and plasmid replication genes and the lac operon) that do not occur in strain 2603VR. In the same relative location in the genome, 20 genes (sag1253-1272; encoding heavy metal transporters) are present in strain 2603VR that do not occur in strain NEM316.
XIII gbs1965-2011 47 Inserted into the proximal end of island XIII, adjacent to tRNA-Lys, are 20 genes in strain NEM316 (gbs1965-1984; function mostly unknown, but not obvious virulence genes) that is not present in strain 2603VR, and 47 genes in strain 2603VR (sag1979-2025; containing several phage genes) that are not present in strain NEM316. The downstream half of the island (gbs1987-2011) is identical in strains NEM316 and 2603VR and contains genes encoding the CAMP factor, two proteases, core metabolic enzymes, two transporters, and a two-component regulator. There are no mobilization genes in this half of the island.
XIV gbs2071-2092 (gbs2064*-2092) 22 (29)* Inserted into the proximal end of island XIV are 16 genes in strain NEM316 (gbs2064-2079; containing numerous phage genes) that are not present in strain 2603VR, and 10 genes in strain 2603VR (sag2111-2120; containing phage genes) that are not present in strain NEM316. The remainder of the island contains genes encoding 2 two-component regulators, 2 membrane proteins and enzymes involved in metabolism, but no obvious virulence or mobilization genes.
* Re-annotation of putative pathogenicity islands based upon the location of mobile DNA present in strain NEM316 but absent from strain 2603VR.
Table 2 Genes variably present in GBS strains, as defined by CGH analysis
Functional category Gene
Sortases sag0633, sag0647-8, sag0650, sag1406-7
LPXTG proteins sag0433, sag0645-6, sag0649, sag1333, sag1404, sag1407-8, sag1462 and sag2063
clp proteases sag1294 and sag1585
Transporters sag1517, sag1998-90, sag1902 and sag1934
Regulators sag0048, sag0124, sag0169, sag 0637, sag0644, sag1128, sag1332, sag1359, sag1409 (rogB), sag1463 (encoding a RALP), sag1791, and sag1956-7 (rgf)
Encoding other proteins sag0031, sag 0624, sag0662 (cyl operon), sag0664 (cyl operon), sag0825, sag1190 (pavA), sag1283, sag1417, sag1442, sag1472, sag1510, sag1558, sag1603, sag1675, sag1772, and sag2043 (cfb)
Genes highlighted in bold have a known or likely role in virulence.
Table 3 Strains employed for PCR and CGH analysis.
Serotype Colonizing Early-onset sepsis Late-onset sepsis
Ia Z18A, Z81A J99, J67 -
Ib Z69A, Z72A, Z73, Z87A, Z111 J96 -
II Z77A - J87
III Z73, Z34A, Z50, Z101A, Z117 B9, H11, J81, J88, J100, R1, WC3, NEM316 M1, MK2, J76, J95, B11, J90, K1
V Z84A, Z12A, Z87A, Z95 B3 -
NT Z41 MK3 -
The strains are a subset of those used for multilocus sequence typing (MLST) of GBS. Strains indicated in bold were assessed by PCR and CGH. Non-highlighted strains were only assessed by PCR. NT = nontypeable; NEM316 = CIP82.45 (Collection de l'Institut Pasteur).
Table 4 Primers for PCR
Gene assignment in strain NEM316 Gene assignment in strain 2603VR Putative island Primer Pairs
gbs0001 sag0001 - 5'-gtagctgatagtcctggc-3' and 5'-agtccccaactaaagcgc-3'
gbs0045-46 sag0046 MME 5'-aaatgggacacgtacgg-3' and 5'-attgccgccatctcaggg-3'
gbs0217
gbs0227 sag0224
sag0234 I 5'-caagcctttaatgctcgc-3' and 5'-aactgaaattccaatcgcc-3'
5'-tcatcgcgaaaatatggag-3' and 5'-cggtcttttagaaactgtgtcc-3'
gbs0247 sag0254 II 5'-gacttatttcaagtttatgg-3' and 5'-acccttatatacgacagc-3'
gbs0367
gbs0388
gbs0393 - pNEM316-1 (islands III, VII and VIII) 5'-atcgatttaggattcatgcc-3' and 5'-caacattcgcaaaataagcc-3'
5'-cctagatggcgtagaggcag-3' and 5'-ttgctcacagaccataagcg-3'
5'-tcacccctgagacgtttacc-3' and 5'-gatcgtaaccacggtttgct-3'
gbs0467 sag0430 IV 5'-attgatagatcttacttgcg-3' and 5'-tgatgcaatagctattggc-3'
gbs0589
gbs0598 sag0610
sag0617 V 5'-cagggtgttcaaggctacc-3' and 5'-caagcttacgcacccaag-3'
5'-ctttcctaaaacatatttgg-3' and 5'-atatggtaaaaacttaaggc-3'
gbs0628
gbs0660 sag0645
sag0685 VI 5'-tagctcagtttgcgactgg-3' and 5'-ccaacttttgcatctgctg-3'
5'-aattcttgattgatgagcg-3' and 5'-tcagctttaatcaattccc-3'
- sag0915 Tn916 5'-aagaccaaaagtggcgaac-3' and 5'-gcctttggattcattcctg-3'
gbs1050
gbs1073 sag1015
sag1038 IX 5'-agcagttacttgatttgcc-3' and 5'-tcctgaattagctagtcgc-3'
5'-tctgcttgagataactccc-3' and 5'-caatagcagttatcaaaggg-3'
gbs1120
gbs1125
gbs1135
gbs1143
gbs1145 - X 5'-cctagatggcgtagaggcag-3' and 5'-ttgctcacagaccataagc-3'
5'-tcgacgtgttttacggttg-3' and 5'-accgaagagatgatgacgac-3'
5'-gggccacactagaaactgc-3' and 5'-aaatccttcatcgctcctg-3'
5'-tcacccctgagacgttacc-3' and 5'-gatcgtaaccacggtttgct-3'
5'-tctctcggcgttattgtcc-3' and 5'-acaaaagcacaagcgactg-3'
gbs1306
gbs1313 sag1233
sag1246 XII 5'-ctttactggcttcacttgg-3' and 5'-gttgatacaggcattgagc-3'
5'-gattactctaccagtgagg-3' and 5'-agaatagtctgcttcaccc-3'
gbs1987
gbs2008 sag2029
sag2053 XIII 5'-ctgacaattgctttgtttcg-3' and 5'-ggctaacccaaatgtaccg-3'
5'-gctcctctgattaatgccc-3' and 5'-caagctcttgttcggttgc-3'
gbs2082 sag2123 XIV 5'-tttctgggaaaaatcagtgg-3' and 5'-ttttcccgaacaaatgatg-3'
==== Refs
Schuchat A Group B streptococcus. Lancet 1999 353 51 56 10023965 10.1016/S0140-6736(98)07128-1
Weisner AM Johnson AP Lamagni TL Arnold E Warner M Heath PT Efstratiou A Characterization of group B streptococci recovered from infants with invasive disease in England and Wales Clin Infect Dis 2004 38 1203 1208 15127328 10.1086/382881
Schuchat A Epidemiology of group B streptococcal disease in the United States: shifting paradigms Clin Microbiol Rev 1998 11 497 513 9665980
Glaser P Rusniok C Buchrieser C Chevalier F Frangeul L Msadek T Zouine M Couve E Lalioui L Poyart C Trieu-Cuot P Kunst F Genome sequence of Streptococcus agalactiae, a pathogen causing invasive neonatal disease Mol Microbiol 2002 45 1499 1513 12354221 10.1046/j.1365-2958.2002.03126.x
Tettelin H Masignani V Cieslewicz MJ Eisen JA Peterson S Wessels MR Paulsen IT Nelson KE Margarit I Read TD Madoff LC Wolf AM Beanan MJ Brinkac LM Daugherty SC DeBoy RT Durkin AS Kolonay JF Madupu R Lewis MR Radune D Fedorova NB Scanlan D Khouri H Mulligan S Carty HA Cline RT Van Aken SE Gill J Scarselli M Mora M Iacobini ET Brettoni C Galli G Mariani M Vegni F Maione D Rinaudo D Rappuoli R Telford JL Kasper DL Grandi G Fraser CM Complete genome sequence and comparative genomic analysis of an emerging human pathogen, serotype V Streptococcus agalactiae Proc Natl Acad Sci U S A 2002 99 12391 12396 12200547 10.1073/pnas.182380799
Hacker J Blum-Oehler G Muhldorfer I Tschape H Pathogenicity islands of virulent bacteria: structure, function and impact on microbial evolution Mol Microbiol 1997 23 1089 1097 9106201 10.1046/j.1365-2958.1997.3101672.x
Kong F Gowan S Martin D James G Gilbert GL Molecular profiles of group B streptococcal surface protein antigen genes: relationship to molecular serotypes J Clin Microbiol 2002 40 620 626 11825981 10.1128/JCM.40.2.620-626.2002
Doran KS Chang JC Benoit VM Eckmann L Nizet V Group B streptococcal beta-hemolysin/cytolysin promotes invasion of human lung epithelial cells and the release of interleukin-8 J Infect Dis 2002 185 196 203 11807693 10.1086/338475
Franken C Haase G Brandt C Weber-Heynemann J Martin S Lammler C Podbielski A Lutticken R Spellerberg B Horizontal gene transfer and host specificity of beta-haemolytic streptococci: the role of a putative composite transposon containing scpB and lmb Mol Microbiol 2001 41 925 935 11532154 10.1046/j.1365-2958.2001.02563.x
Schmidt H Hensel M Pathogenicity islands in bacterial pathogenesis Clin Microbiol Rev 2004 17 14 56 14726454 10.1128/CMR.17.1.14-56.2004
Durbin R Thierry-Mieg JT A C. elegans DataBase 1991
Saunders NJ Peden JF Hood DW Moxon ER Simple sequence repeats in the Helicobacter pylori genome Mol Microbiol 1998 27 1091 1098 9570395 10.1046/j.1365-2958.1998.00768.x
Chaussee MS Sylva GL Sturdevant DE Smoot LM Graham MR Watson RO Musser JM Rgg influences the expression of multiple regulatory loci to coregulate virulence factor expression in Streptococcus pyogenes Infect Immun 2002 70 762 770 11796609 10.1128/IAI.70.2.762-770.2002
Chaussee MS Somerville GA Reitzer L Musser JM Rgg coordinates virulence factor synthesis and metabolism in Streptococcus pyogenes J Bacteriol 2003 185 6016 6024 14526012 10.1128/JB.185.20.6016-6024.2003
Saunders NJ Snyder LA The minimal mobile element Microbiology 2002 148 3756 3760 12480877
Spellerberg B Rozdzinski E Martin S Weber-Heynemann J Lutticken R rgf encodes a novel two-component signal transduction system of Streptococcus agalactiae Infect Immun 2002 70 2434 2440 11953380 10.1128/IAI.70.5.2434-2440.2002
Beckert S Kreikemeyer B Podbielski A Group A streptococcal rofA gene is involved in the control of several virulence genes and eukaryotic cell attachment and internalization Infect Immun 2001 69 534 537 11119547 10.1128/IAI.69.1.534-537.2001
Gutekunst H Eikmanns BJ Reinscheid DJ Analysis of RogB-controlled virulence mechanisms and gene repression in Streptococcus agalactiae Infect Immun 2003 71 5056 5064 12933848 10.1128/IAI.71.9.5056-5064.2003
Pritzlaff CA Chang JC Kuo SP Tamura GS Rubens CE Nizet V Genetic basis for the beta-haemolytic/cytolytic activity of group B Streptococcus Mol Microbiol 2001 39 236 247 11136446 10.1046/j.1365-2958.2001.02211.x
Hassan AA Abdulmawjood A Yildirim AO Fink K Lammler C Schlenstedt R Identification of streptococci isolated from various sources by determination of cfb gene and other CAMP-factor genes Can J Microbiol 2000 46 946 951 11068682 10.1139/cjm-46-10-946
Holmes AR McNab R Millsap KW Rohde M Hammerschmidt S Mawdsley JL Jenkinson HF The pavA gene of Streptococcus pneumoniae encodes a fibronectin-binding protein that is essential for virulence Mol Microbiol 2001 41 1395 1408 11580843 10.1046/j.1365-2958.2001.02610.x
Kreikemeyer B McIver KS Podbielski A Virulence factor regulation and regulatory networks in Streptococcus pyogenes and their impact on pathogen-host interactions Trends Microbiol 2003 11 224 232 12781526
Herbert MA Beveridge CJE Saunders NJ Bacterial virulence factors in neonatal sepsis: group B streptococcus Current Opinion in Infectious Diseases 2004 17 225 229 15166825 10.1097/00001432-200406000-00009
Streptococcus agalactiae complete genome NEM316 2002
Streptococcus agalactiae complete genome 2603V/R 2002
Computational Biology Research Group
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
Marchler-Bauer A Anderson JB DeWeese-Scott C Fedorova ND Geer LY He S Hurwitz DI Jackson JD Jacobs AR Lanczycki CJ Liebert CA Liu C Madej T Marchler GH Mazumder R Nikolskaya AN Panchenko AR Rao BS Shoemaker BA Simonyan V Song JS Thiessen PA Vasudevan S Wang Y Yamashita RA Yin JJ Bryant SH CDD: a curated Entrez database of conserved domain alignments Nucleic Acids Res 2003 31 383 387 12520028 10.1093/nar/gkg087
Rozen S Skaletsky H Primer3 on the WWW for general users and for biologist programmers Methods Mol Biol 2000 132 365 386 10547847
| 15913462 | PMC1175089 | CC BY | 2021-01-04 16:03:39 | no | BMC Microbiol. 2005 May 24; 5:31 | utf-8 | BMC Microbiol | 2,005 | 10.1186/1471-2180-5-31 | oa_comm |
==== Front
BMC NephrolBMC Nephrology1471-2369BioMed Central London 1471-2369-6-61591891510.1186/1471-2369-6-6Research ArticleEffects of diabetes and hypertension on macrophage infiltration and matrix expansion in the rat kidney Hartner Andrea [email protected] Roland [email protected] Michael [email protected] Nada [email protected] Karl F [email protected] Children and Youth Hospital, University of Erlangen-Nuremberg, Loschgestrasse 15, D-91054 Erlangen, Germany2 Nephrology and Hypertension, University of Erlangen-Nuremberg, Loschgestrasse 8, D-91054 Erlangen, Germany3 Medicine II, Augsburg City Hospital, Stenglinstrasse 2, D-86156 Augsburg, Germany2005 27 5 2005 6 6 6 18 10 2004 27 5 2005 Copyright © 2005 Hartner 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 experimental models of diabetes mellitus, aggravation of renal injury by concomitant hypertension has been described. Inflammatory mechanisms contribute to renal damage in both diseases. We investigated whether hypertension and diabetes mellitus act synergistically to induce macrophage infiltration and matrix expansion in the kidney.
Methods
Insulin-dependent diabetes mellitus was induced by streptozotocin injections to hypertensive mRen2-transgenic rats (TGR) and normotensive Sprague-Dawley control rats. Quantitative immunohistochemical examination of kidney tissue sections was used to measure macrophage infiltration and matrix expansion. The expression of MCP-1, Osteopontin, RANTES, ICAM-1 and VCAM-1 was evaluated by real-time RT-PCR. The localization of MCP-1 was studied by immunohistochemistry.
Results
Macrophage infiltration was present in the kidney of normotensive diabetic rats. Hypertensive rats exhibited a more marked infiltration of macrophages, regardless of whether diabetes was present or not. Gene expression of ICAM-1, VCAM-1 and RANTES was unaltered whereas Osteopontin and MCP-1 were induced by hypertension. Immunoreactive MCP-1 was slightly increased in diabetic rat kidney podocytes, and more markedly increased in hypertensive animals. Glomerular matrix accumulation was induced by diabetes and hypertension to a similar degree, and was highest in hypertensive, diabetic animals.
Conclusion
Diabetes mellitus caused a mild, and angiotensin-dependent hypertension a more marked infiltration of macrophages in the kidney. Combination of both diseases led to additive effects on matrix expansion but not on inflammation. Hypertension appears to be a much stronger stimulus for inflammation of the kidney than STZ diabetes, at least in mRen2-transgenic rats.
==== Body
Background
Diabetic nephropathy is the most common cause of end-stage renal failure in developed countries and its incidence continues to rise [1]. In most patients with diabetic nephropathy, hypertension is present and contributes significantly to the progression of renal failure in diabetes [1]. Studies in diabetic rats as well as in human volunteers with hyperglycemia indicated that activation of the intrarenal renin-angiotensin system (RAS) plays a key role in the development of the hemodynamic abnormalities in early diabetic nephropathy [2]. Angiotensin II-induced hypertension leads to macrophage infiltration in the kidney, and chemokines have been proposed as mediators of macrophage infiltration [3,4]. For example, the chemokine monocyte chemoattractant protein-1 (MCP-1) can be induced in vascular smooth muscle cells by angiotensin II [5]. The finding that MCP-1 is likewise induced in renal mesangial cells by high glucose concentrations [6] is in keeping with the hypothesis that chronic inflammatory mechanisms may also contribute to the pathogenesis of diabetic nephropathy [7]. Thus, induction of chemokines in the diabetic kidney seems to enhance macrophage infiltration [8-12].
Kelly et al. described that induction of osteopontin is related to macrophage infiltration in streptozotocin diabetic, mRen-2 transgenic hypertensive rats [13]. These authors had previously reported that this animal model resembles aspects of human diabetic nephropathy [14]. We used this animal model to examine the potential contribution of other chemokines (MCP-1, RANTES) and adhesion molecules to macrophage infiltration. Further, we sought to delineate the relative contribution of angiotensin-dependent hypertension from that of diabetes to inflammation by including normotensive diabetic rats of the same genetic background.
Methods
Rat models of hypertension and diabetes mellitus
Rats were housed in a room maintained at 22 ± 2°C, exposed to a 12 hour dark/light cycle. The animals were allowed unlimited access to chow (#1320, Altromin, Lage, Germany) and tap water. All procedures performed on animals were done in accordance with guidelines of the American Physiological Society and were approved by the local government authorities (Regierung von Mittelfranken, AZ # 621-2531.31-19/96). Eighteen male rats heterozygous for the mouse ren-2 transgene (TGR) with angiotensin II dependent hypertension [15] and 18 age-matched Sprague-Dawley-Hannover (SD) controls (Möllegaard, Eijby, Denmark) at an average body weight of 250 g were used for induction of diabetes by intraperitoneal injection of streptozotocin (STZ) (Sigma, Deisenhofen, Germany) (70 mg per kg of body weight) dissolved in 0.1 M sodium citrate buffer (pH 4.5). Two days later, blood was obtained from the tail vein and diabetes was confirmed by measurement of blood glucose using a reflectance meter (Glucometer Elite II, Bayer, Leverkusen, Germany). Only rats with a consistent blood glucose > 250 mg/dl were included (13 TGR and 12 SD). Diabetic and control rats were followed for 5 weeks. Blood glucose and systolic blood pressure (as measured by tail-cuff plethysmography under light ether anesthesia) were monitored weekly (at 8 a.m.). After five weeks, the rats were kept in metabolic cages for determination of urinary albumin excretion (enzyme immunoassay kit, CellTrend, Luckenwalde, Germany) for 24 hours. Subsequently, the rats were equipped with a femoral catheter and arterial blood pressure was measured via transducers (Grass Instruments, Quincy, USA) connected to a polygraph (Hellige, Freiburg, Germany) four hours after termination of anesthesia. Rats were sacrificed and kidneys were weighed and decapsulated. Half of each kidney was immediately snap-frozen in liquid nitrogen for later protein and RNA extraction. The other half was fixed in methyl-Carnoy solution (60% methanol, 30% chloroform and 10% glacial acetic acid) for histology and immunohistochemistry.
Real-time RT-PCR detection of mRNA
Renal cortical tissue extraction and real-time RT-PCR were carried out as described [16]. Briefly, first-strand cDNA was synthesized with TaqMan reverse transcription reagents (Applied Biosystems, Darmstadt, Germany) using random hexamers as primers. Final RNA concentration in the reaction mixture was adjusted to 0,5 ng/μL. Reactions without Multiscribe reverse transcriptase were used as negative controls for genomic DNA contamination. PCR was performed with an ABI PRISM 7000 Sequence Detector System and TaqMan or SYBR Green Universal PCR Master Mix (Applied Biosystems) according to the manufacturers instructions. All samples were run in triplicates. The relative amount of the specific mRNA was normalized with respect to 18S rRNA. Primer design was accomplished with PrimerExpress software (Applied Biosystems). Primer sequences used are as follows. RANTES forward primer: GTCGTCTTTGTCACTCGAAGGA, RANTES reverse primer: GATGTATTCTTGAACCCACTTCTTCTC and RANTES probe: CCGCCAAGTGTGTG CCAACCC. ICAM-1 forward primer: GGGCCCCCTACCTTAGGAA, ICAM-1 reverse primer: GGGACAGTGTCCCAGCTTTC. VCAM-1 forward primer: TGTGGAAGTGT GCCCGAAAT, VCAM-1 reverse primer: TGCCTTGCGGATGGTGTAC. Primers and probes for 18S, MCP-1 and osteopontin were previously described [16-18].
Western blot detection of MCP-1
Protein was extracted from kidneys of 5 rats of each group using Tri-reagent (MRC Inc.). Protein concentration was determined using a protein assay kit (Pierce, Rockford, IL, USA). Western blot analysis was performed as described before [3] with a polyclonal rabbit anti-rat MCP-1 antiserum which was kindly provided by Dr. T. Yoshimura, Frederick, MD and used at a dilution of 1:250.
Immunohistochemistry
After overnight fixation in methyl-Carnoy solution, tissues were dehydrated by bathing in increasing concentrations of methanol, followed by 100% iso-propanol. After embedding in paraffin, 3 μm sections were cut with a Leitz SM 2000 R microtome (Leica Instruments, Nussloch, Germany). After deparaffinization, endogenous peroxidase activity was blocked with 3% H2O2 in methanol for 20 min at room temperature. Immunohistochemical detection of ED-1 was carried out as described [3]. The mouse monoclonal antibody against the macrophage marker ED-1 was purchased from Serotec (Biozol, Eching, Germany) and used at a dilution of 1:250. Renal cortical collagen I was detected by a rabbit polyclonal antibody (Biogenesis, Poole, England, UK) at a dilution of 1:1000. A goat polyclonal antibody to collagen IV (Southern Biotechnology Associates, Birmingham, AL, USA) was used at a dilution of 1:500. Each slide was counterstained with hematoxylin.
Analysis of data
Intraglomerular ED-1 positive cells were counted in all glomeruli of a given kidney section (100–300 glomeruli, no selection) and expressed as cells per glomerular section. Interstitial ED-1 positive cells were counted in 30 medium-power (magnification 250 ×) cortical views per section and expressed as cells per square mm. Counting was begun in a random cortical field and in consecutive non-overlapping cortical fields to the right of the previous view without selection; if necessary, counting was continued at the opposite (left) edge of the section. MCP-1 staining was evaluated in >100 glomeruli and 20 cortical interstitial low-power (100 ×) fields by a blinded observer. Glomeruli were classified as showing no staining (score 0), staining of up to one third of the glomerular tuft area (score 1), staining affecting one to two thirds (score 2) or more than two thirds of the glomerular tuft (score 3). Interstitial fields were classified as showing no staining (score 0), one area of peritubular interstitial staining not spanning the circumference of a tubular cross-section (score 1); two such areas or one area of staining spanning the circumference of a tubular cross-section (score 2), peritubular staining involving less than 4 (score 3) or more than 4 (score 4) tubular cross-sections. Expansion of interstitial collagen I was measured by Metaview software (Visitron Systems, Puchheim, Germany) in 10 non-overlapping medium-power cortical views per section excluding glomeruli and was expressed as percent of stained area per cross section. Glomerular collagen IV staining was measured by Metaview in every third glomerulus per cross section, and the stained area was expressed as percentage of the total area of the glomerular tuft.
Two-way analysis of variance, followed by the post-hoc Bonferroni test with adjustment for multiple comparison, was used to compare groups. A p value < 0.05 was considered significant. The procedures were carried out using the SPSS version 11.5 software (SPSS Inc., Chicago, IL, USA). Values are displayed as means ± standard error of the mean.
Results
Injection of STZ induced diabetes mellitus equally well in SD and in TGR. In STZ-treated animals, blood glucose levels were not different between TGR and SD throughout the development of the disease (see figure 1A). In untreated TGR, blood glucose did not differ from untreated SD rats (figure 1A). Systolic and mean arterial blood pressure were significantly higher in TGR compared to SD (figure 1B). Development of hypertension was not affected by STZ treatment; there were no differences between STZ-treated and untreated control animals with regard to systolic blood pressure (data not shown) and mean intraarterial blood pressure measurements (figure 1B). Albumin excretion was increased in diabetic animals and even more in hypertensive rats; combination of both diseases did not further elevate albuminuria (figure 1C).
STZ diabetes led to reduced body weight gain and kidney hypertrophy. Kidney weight/body weight ratio was significantly increased in SD-STZ, TGR and TGR-STZ versus SD controls as well as in TGR-STZ versus TGR (table 1). TGR hypertensive animals had significantly higher heart weight/body weight ratios than normotensive SD rats, which was not affected by STZ diabetes (table 1). Diabetes led to polyuria (table 1), which was further increased in hypertensive rats. Urine production was slightly higher in TGR than in SD, but significantly elevated in the diabetic groups SD-STZ and TGR-STZ (table 1).
Both, diabetes and hypertension caused macrophage infiltration of the kidney (figure 2). The effect of hypertension, however, was more pronounced. TGR-STZ animals had the highest levels of macrophages in glomeruli but there was no statistically significant difference compared to TGR (figure 2). Minor macrophage infiltration was observed in the interstitial space of SD-STZ rats with diabetes alone while marked macrophage influx was noted in glomeruli of normoglycemic hypertensive TGR (figure 2). In the combined model of diabetes and hypertension in TGR-STZ, the effect of both diseases on macrophage infiltration of the interstitial space appeared to be additive although the numerical difference between TGR and TGR-STZ failed to reach statistical significance (figure 2). Of note, in TGR-STZ rats, interstitial macrophages were often localized in periglomerular areas (not shown).
Expression of mediators regulating macrophage infiltration (MCP-1, osteopontin, RANTES) and adhesion molecules involved in macrophage infiltration (ICAM-1, VCAM-1) was investigated to further elucidate inflammatory pathways in hypertensive and diabetic renal disease: In TGR rats, renal MCP-1 and osteopontin mRNA steady state levels were increased compared to SD controls, as determined by real-time RT-PCR (figure 3A and 3B). STZ-induced diabetes mellitus of five weeks did not significantly increase MCP-1 and osteopontin mRNA in SD rats and did not augment the upregulation of MCP-1 mRNA in TGR (figure 3). In contrast, the chemokine RANTES or the adhesion molecules ICAM-1 and VCAM-1 were not transcriptionally regulated in the kidney in response to hypertension or diabetes (table 2). The expression level of MCP-1 correlated with macrophage counts in the interstitial space (r2 = 0.47, p = 0.002) but not in glomeruli (r2 = 0.002, p = 0.857).
The specificity of the antibody to MCP-1 was confirmed by Western blot analysis of renal protein (figure 4) yielding a characteristic double band. By immunohistochemistry (figure 5), staining of the smooth muscle layer of small arteries and afferent arterioles was present in controls as well as in diseased animals, albeit more markedly in hypertensive rats (figure 5C). However, almost no glomerular MCP-1 staining was observed in control animals (figure 5A). Focal and segmental positive staining for MCP-1 was detected in glomeruli of normotensive SD-STZ rats with diabetes mellitus (figure 5B). Widespread glomerular staining for MCP-1 (figure 5C and 5D) was seen in TGR and TGR-STZ. By high power light microscopy and double-staining for ED-1, it was noted that intrinsic glomerular cells, rather than infiltrating cells, stained positively for MCP-1. The pattern of immunostaining suggested a predominantly podocyte and occasionally endothelial localization of MCP-1 (figure 5). In the cortical interstitium, MCP-1-staining was localized to peritubular spindle-shaped interstitial cells, possibly fibroblasts, in close proximity to ED-1-positive macrophages (figure 5E). Quantification of glomerular and interstitial MCP-1 staining demonstrated a mild increase in SD-STZ rats and a massive increase in TGR and TGR-STZ (figure 6A). Interstitial staining was little affected by STZ diabetes but markedly induced by TGR hypertension (figure 6B).
A moderate matrix expansion was detected in SD-STZ rats in the renal cortex (figure 7A) and in glomeruli (figure 7B) as compared to SD. In TGR, cortical and glomerular matrix expansion was more prominent than in STZ (figure 7A+B), which was further aggravated in TGR-STZ with regard to glomerular matrix (figure 7B) but not to cortical matrix expansion (figure 7A).
Discussion
The results of the present study confirm that in the rat, both hypertension and diabetes mellitus induce macrophage infiltration in the kidney which may contribute to the development of glomerular and interstitial injury. Moreover, our results confirm and extend previous reports of increased MCP-1 expression in the kidney in diabetes [8-12] and in hypertension [3]. In our study, the effect of hypertension on MCP-1 and osteopontin expression as well as on macrophage infiltration was much more prominent than the effects of diabetes mellitus. Streptozotocin diabetes induced some predominantly glomerular MCP-1 expression and macrophage infiltration whereas hypertensive rats exhibited marked interstitial and glomerular inflammation. This finding cannot be explained by the longer duration of hypertension, compared with diabetes. In preliminary experiments (data not shown), we assured that macrophage infiltration in the kidney is highest from 2 to 6 weeks after STZ injection and decreases thereafter, as described previously by others [7,12]. A different time course of macrophage influx, with a prominent late infiltration, has been found in other rodent models of diabetes, for example in db/db mice with type 2 diabetes [11].
Our data did not reveal evidence for a synergistic effect of TGR hypertension and STZ diabetes on kidney inflammation. Macrophage infiltration tended to slightly higher values in hypertensive, hyperglycemic rats, compared to hypertensive, normoglycemic animals but this trend did not reach statistical significance. These results contrast sharply with those reported by Kelly et al. [13] who described very little inflammation in normoglycemic TGR but a very marked macrophage infiltration in diabetic TGR. We cannot fully explain this discrepancy but several factors may contribute to it. Kelly et al. did not include normotensive controls, neither normoglycemic nor hyperglycemic [13]. Therefore, these authors may have missed the effect of TGR hypertension alone on macrophage infiltration. The shorter duration of diabetes in our study is unlikely to explain the different findings, as discussed above, but the different age and gender of the rats may play a role. We induced STZ diabetes in adult male TGR whereas Kelly et al. employed adolescent (6 week old) female TGR [13]. In other respects, our data confirmed the notion that TGR hypertension aggravates the renal sequelae of STZ diabetes, as described by Kelly et al. [14]. Glomerular sclerosis, as assessed by quantification of collagen IV staining, was significantly higher in hypertensive, diabetic TGR, compared to all other groups.
Real-time RT-PCR was used to screen for large differences of the expression of several proinflammatory molecules, and we focused on the factors which exhibited a clear induction in diabetic, hypertensive TGR. The apparently "negative" results on ICAM-1, VCAM-1 and RANTES should not be misinterpreted to exclude a role of these factors. A more subtle investigation might have shown localized induction of these factors which has in fact been described in hypertensive and/or diabetic kidney disease [4,12,19,20]. The most prominent induction was detected for osteopontin and MCP-1. We focused on MCP-1 because the expression of osteopontin in the diabetic TGR had already been investigated by Kelly et al. [13]. Real-time RT-PCR did not detect an induction of MCP-1 in total cortical RNA of kidney of normotensive, diabetic rats. However, a more subtle investigation by means of immunohistochemistry clearly showed that MCP-1 was increased in glomeruli of normotensive, diabetic rats, in agreement with previous reports [8-12]. MCP-1 was more markedly induced in hypertensive rats but the effects of hypertension and diabetes were not additive. Interestingly, glomerular MCP-1 was localized in podocytes, similar to the localization of osteopontin in glomeruli of diabetic TGR described by Kelly et al. [13]. These findings underscore the potential role of podocytes for glomerular inflammation.
Conclusion
In our rat model, the effects of hypertension and/or angiotensin II on the expression of the chemokine MCP-1, and on macrophage infiltration, in the kidney were much more pronounced than the effects of diabetes. Furthermore, the effects of both diseases combined on kidney inflammation were not synergistic, and often not even additive. We speculate that the mechanical or hormonal factor – i.e., angiotensin-dependent hypertension – plays a much greater role for the induction of an inflammatory reaction in the kidney than does the metabolic factor, hyperglycemia. In patients with diabetic nephropathy, the decisive contribution of hypertension to renal damage has often been noted [1,21]. Investigators have combined STZ diabetes with genetic models of hypertension, and acceleration of diabetic nephropathy was reported regardless of whether spontaneously hypertensive rats [22], Dahl salt-sensitive rats [23] or mRen2-transgenic hypertensive TGR [14] were used. Our data do not support the notion that inflammation explains this acceleration of diabetic kidney damage because chemokine expression and macrophage infiltration were largely determined by the presence of hypertension, regardless of whether the animals were diabetic or normoglycemic. Other processes, for example accumulation of extracellular matrix, could contribute to the accelerated kidney damage caused by diabetes mellitus in combination with hypertension.
Competing interests
The author(s) declare that they have no competing interests.
Authors' contributions
A.H. and K.F.H. drafted this manuscript. K.F.H. and R.V. planned and designed the study. N.C. and M.W. performed and evaluated the animal experiments. A.H., N.C. and M.W. performed and evaluated the histological, molecular biology and immunohistochemical studies. N.C. and M.W. revised the manuscript for intellectual content. All authors have approved the final version of the manuscript.
Pre-publication history
The pre-publication history for this paper can be accessed here:
Acknowledgements
We gratefully acknowledge the expert technical assistance of Rainer Wachtveitl, Miroslava Kupraszewicz-Hutzler, Elisabeth Buder and Rita Zitzmann. This study was part B12 of the "Interdisziplinaeres Zentrum fuer Klinische Forschung" at the Hospital of the University of Erlangen-Nuernberg, funded by the "Bundesministerium fuer Bildung und Forschung" (01 KS 0002). In addition, the study was supported by a grant-in-aid (KFO 106, TP2) and a Heisenberg scholarship (Hi 510/7-1, to Karl F. Hilgers) from the "Deutsche Forschungsgemeinschaft". Part of the data were presented in abstract form at the 1998 meeting of the American Society of Nephrology in Philadelphia, PA, U.S.A.
Figures and Tables
Figure 1 Blood glucose (panel A), arterial blood pressure (panel B) and albuminuria (panel C) of diabetic and hypertensive rats. Blood glucose was measured 24 hours before termination of the experiment. Mean arterial blood pressure was determined in awake rats via indwelling catheters inserted into the femoral artery. For determination of albuminuria, urine was collected 24 hours before sacrifice. SD, normotensive normoglycemic control rats; TGR, transgenic hypertensive rats; STZ, streptozotocin treatment. * indicates p < 0.05 versus normotensive normoglycemic SD control rats.
Figure 2 Macrophage infiltration of glomeruli (panel A) and interstitial space (panel B) of the kidney from rats with diabetes mellitus and/or hypertension. Glomerular macrophages are expressed as ED-1-positive cells per glomerular cross-section, interstitial macrophages as ED-1-positive cells per square mm. SD, normotensive normoglycemic control rats; TGR, transgenic hypertensive rats; STZ, streptozotocin treatment. Data are mean ± SEM of n = 5 rats. * indicates p < 0.05 versus normotensive normoglycemic SD control rats.
Figure 3 Real-time RT-PCR analysis of MCP-1 (A) and osteopontin (B) mRNA expression in the renal cortex. Data are expressed as fold control compared to SD control rats. Data are mean ± SEM of n = 5 rats. * indicates p < 0.05 versus normotensive normoglycemic SD control rats. SD, normotensive normoglycemic control rats; TGR, transgenic hypertensive rats; STZ, streptozotocin treatment.
Figure 4 Western blot analysis for MCP-1 protein in cortical protein preparations of two rats of each experimental group, yielding the characteristic MCP-1 double band (14 and 16 kDa). SD, normotensive normoglycemic control rats; TGR, transgenic hypertensive rats; STZ, streptozotocin treatment.
Figure 5 Immunohistochemistry for MCP-1 and the macrophage marker ED-1. Panel A-D: examples of photomicrographs of MCP-1 staining (brown) in glomeruli, hematoxylin counterstain (blue nuclei) Panel A (glomerulus of a normotensive normoglycemic rat) represents score 0, panel B (from a normotensive diabetic rat) score 1, panel C (from a hypertensive normoglycemic rat kidney) score 2, and panel D (from a hypertensive diabetic rat) score 3. Panel E, High power magnification of MCP-1 immunohistochemistry (brown) in a hypertensive diabetic rat. MCP-1 staining localized to spindle-shaped cells, probably fibroblasts, surrounding tubules in a double immunohistochemistry with the macrophage marker ED-1 (blue cytoplasm, arrows), methyl-green counterstain. Macrophages were often localized in close proximity to the MCP-1 positive cells surrounding tubules (asterisk). Scale bars (50 μm) are indicated. Note the identical scale for panels A-D but different scale for panel E.
Figure 6 Semiquantitative evaluation of MCP-1 immunohistochemistry. For glomerular staining (A) more than 100 glomeruli per kidney section were classified 0 to 3 (see methods and figure 5 for details). The percentage of glomeruli assigned to a given score value is shown. For interstitial MCP-1 scores (B), 20 low-power cortical fields were classified 0 to 4 (see methods for details). The percentage of interstitial fields assigned to a given score value is shown. SD, normotensive normoglycemic control rats; TGR, transgenic hypertensive rats; STZ, streptozotocin treatment. Data are mean ± SEM of n = 5 rats. * indicates p < 0.05 versus normotensive normoglycemic Sprague-Dawley (SD) control rats.
Figure 7 Matrix expansion in the renal cortex and the glomerulus. A: measurement of cortical collagen I staining and B: measurement of glomerular collagen IV staining. Data are mean ± SEM, * indicates p < 0.05 versus normotensive normoglycemic SD controls. # indicates p < 0.05 versus SD-STZ. § indicates p < 0.05 versus normoglycemic TGR.
Table 1 Body weight, organ weights and urine production
Group SD SD+STZ TGR TGR+STZ
N 5 7 7 6
Body weight (g) 429.6 ± 9.0 308.0 ± 9.5 * 388.3 ± 14.6 * 316.3 ± 36.4 * §
Heart weight (g) 1.53 ± 0.06 1.24 ± 0.05 * 1.94 ± 0.06 * 1.47 ± 0.18
Heart weight / body weight ratio(mg/g) 3.55 ± 0.11 4.03 ± 0.09 * 5.03 ± 0.19 * 4.63 ± 0.16 *
Kidney weight (g) 1.30 ± 0.03 1.53 ± 0.02 * 1.47 ± 0.06 * 1.67 ± 0.13 *
Kidney weight / body weight ratio(mg/g) 3.04 ± 0.05 4.98 ± 0.15 * 3.82 ± 0.18 * 5.43 ± 0.33 * §
Urine production (ml/24 h) 22.0 ± 2.7 95.7 ± 10.1 * 30.0 ± 1.7 * 111.5 ± 13.6 * §
* p < 0.05 versus SD
§p < 0.05 versus TGR
Table 2 mRNA expression of the chemokine RANTES and the adhesion molecules ICAM-1 and VCAM-1.
SD SD-STZ TGR TGR-STZ
RANTES 1.0 ± 0.26 0.57 ± 0.03 0.47 ± 0.07 0.6 ± 0.12
VCAM-1 1.0 ± 0.14 0.98 ± 0.33 1.34 ± 0.51 0.65 ± 0.16
ICAM-1 1.0 ± 0.25 1.31 ± 0.49 0.95 ± 0.32 0.68 ± 0.23
Data are fold induction relative to SD as assessed by real-time RT-PCR from renal cortex and are displayed as mean ± standard error of the mean. SD Sprague-Dawley Hannover rats. STZ streptozotocin-treated. TGR Ren-2 transgenic rats. There were no significant differences.
==== Refs
Ritz E Orth SR Nephropathy in patients with type 2 diabetes mellitus N Engl J Med 1999 341 1127 1133 10511612 10.1056/NEJM199910073411506
Hollenberg NK Price DA Fisher ND Lansang MC Perkins B Gordon MS Williams GH Laffel LM Glomerular hemodynamics and the renin-angiotensin system in patients with type 1 diabetes mellitus Kidney Int 2003 63 172 178 12472780 10.1046/j.1523-1755.2003.00701.x
Hilgers KF Hartner A Porst M Mai M Wittmann M Hugo C Ganten D Geiger H Veelken R Mann JF Monocyte chemoattractant protein-1 and macrophage infiltration in hypertensive kidney injury Kidney Int 2000 58 2408 2419 11115074 10.1046/j.1523-1755.2000.00424.x
Wolf G Ziyadeh FN Thaiss F Tomaszewski J Caron RJ Wenzel U Zahner G Helmchen U Stahl RA Angiotensin II stimulates expression of the chemokine RANTES in rat glomerular endothelial cells. Role of the angiotensin type 2 receptor J Clin Invest 1997 100 1047 1058 9276721
Capers Q Alexander RW Lou P De Leon H Wilcox JN Ishizaka N Howard AB Taylor WR Monocyte chemoattractant protein-1 expression in aortic tissues of hypertensive rats Hypertension 1997 30 1397 1402 9403559
Ihm CG Park JK Hong SP Lee TW Cho BS Kim MJ Cha DR Ha H A high glucose concentration stimulates the expression of monocyte chemotactic peptide 1 in human mesangial cells Nephron 1998 79 33 37 9609459 10.1159/000044988
Young BA Johnson RJ Alpers CE Eng E Gordon K Floege J Couser WG Seidel K Cellular events in the evolution of experimental diabetic nephropathy Kidney Int 1995 47 935 944 7752595
Wada T Furuichi K Sakai N Iwata Y Yoshimoto K Shimizu M Takeda SI Takasawa K Yoshimura M Kida H Kobayashi KI Mukaida N Naito T Matsushima K Yokoyama H Up-regulation of monocyte chemoattractant protein-1 in tubulointerstitial lesions of human diabetic nephropathy Kidney Int 2000 58 1492 1499 11012884 10.1046/j.1523-1755.2000.00311.x
Kato S Luyckx VA Ots M Lee KW Ziai F Troy JL Brenner BM MacKenzie HS Renin-angiotensin blockade lowers MCP-1 expression in diabetic rats Kidney Int 1999 56 1037 1048 10469372 10.1046/j.1523-1755.1999.00643.x
Banba N Nakamura T Matsumura M Kuroda H Hattori Y Kasai K Possible relationship of monocyte chemoattractant protein-1 with diabetic nephropathy Kidney Int 2000 58 684 690 10916091 10.1046/j.1523-1755.2000.00214.x
Chow F Ozols E Nikolic-Paterson DJ Atkins RC Tesch GH Macrophages in mouse type 2 diabetic nephropathy: Correlation with diabetic state and progressive renal injury Kidney Int 2004 65 116 128 14675042 10.1111/j.1523-1755.2004.00367.x
Sassy-Prigent C Heudes D Mandet C Belair MF Michel O Perdereau B Bariety J Bruneval P Early glomerular macrophage recruitment in streptozotocin-induced diabetic rats Diabetes 2000 49 466 475 10868970
Kelly DJ Wilkinson-Berka JL Ricardo SD Cox AJ Gilbert RE Progression of tubulointerstitial injury by osteopontin-induced macrophage recruitment in advanced diabetic nephropathy of transgenic (mRen-2)27 rats Nephrol Dial Transplant 2002 17 985 991 12032186 10.1093/ndt/17.6.985
Kelly DJ Wilkinson-Berka JL Allen TJ Cooper ME Skinner SL A new model of diabetic nephropathy with progressive renal impairment in the transgenic (mRen-2)27 rat (TGR) Kidney Int 1998 54 343 352 9690200 10.1046/j.1523-1755.1998.00019.x
Mullins JJ Peters J Ganten D Fulminant hypertension in transgenic rats harbouring the mouse Ren-2 gene Nature 1990 344 541 544 2181319 10.1038/344541a0
Veelken R Hilgers KF Porst M Krause H Hartner A Schmieder RE Effects of sympathetic nerves and angiotensin II on renal sodium and water handling in rats with common bile duct ligature Am J Physiol Renal Physiol 2005 288 F1267 F1275 15701819 10.1152/ajprenal.00069.2003
Behr TM Wang X Aiyar N Coatney RW Li X Koster P Angermann CE Ohlstein E Feuerstein GZ Winaver J Monocyte chemoattractant protein-1 is upregulated in rats with volume-overload congestive heart failure Circulation 2000 102 1315 1322 10982549
Uno Y Horii A Umemoto M Hasegawa T Doi K Uno A Takemura T Kubo T Effects of hypergravity on morphology and osteopontin expression in the rat otolith organs J Vestib Res 2000 10 283 289 11455109
Mervaala EM Müller DN Park JK Schmidt F Lohn M Breu V Dragun D Ganten D Haller H Luft FC Monocyte infiltration and adhesion molecules in a rat model of high human renin hypertension Hypertension 1999 33 389 395 9931135
Mai M Hilgers KF Geiger H Experimental studies on the role of intercellular adhesion molecule-1 and lymphocyte function-associated antigen-1 in hypertensive nephrosclerosis Hypertension 1996 28 973 979 8952585
Mogensen CE Combined high blood pressure and glucose in type 2 diabetes: double jeopardy. British trial shows clear effects of treatment, especially blood pressure reduction BMJ 1998 317 693 694 9732334
Cooper ME Allen TJ O'Brien RC Macmillan PA Clarke B Jerums G Doyle AE Effects of genetic hypertension on diabetic nephropathy in the rat – functional and structural characteristics J Hypertens 1988 6 1009 1016 3221096
Korner A Jaremko G Eklof AC Aperia A Rapid development of glomerulosclerosis in diabetic Dahl salt-sensitive rats Diabetologia 1997 40 367 373 9112012 10.1007/s001250050689
| 15918915 | PMC1175090 | CC BY | 2021-01-04 16:03:44 | no | BMC Nephrol. 2005 May 27; 6:6 | utf-8 | BMC Nephrol | 2,005 | 10.1186/1471-2369-6-6 | oa_comm |
==== Front
BMC PediatrBMC Pediatrics1471-2431BioMed Central London 1471-2431-5-181598517010.1186/1471-2431-5-18Research ArticleRetinopathy of prematurity and risk factors: a prospective cohort study Karna Padmani [email protected] Jyotsna [email protected] Linda [email protected] Wilfried [email protected] Division of Neonatology, Pediatrics and Human Development, Michigan State University, East Lansing, USA2 Department of Epidemiology, College of Human Medicine, Michigan State University, East Lansing, USA2005 28 6 2005 5 18 18 30 11 2004 28 6 2005 Copyright © 2005 Karna et al; licensee BioMed Central Ltd.2005Karna 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
Increased survival of extremely low birth infants due to advances in antenatal and neonatal care has resulted in a population of infants at high risk of developing retinopathy of prematurity (ROP). Therapeutic interventions include the use of antenatal and postnatal steroids however, their effects on the severity of ROP is in dispute. In addition, it has not been investigated whether severe ROP is due to therapeutic interventions or due to the severity of illness. The aim of the present study was to assess the association between the incidence of severe retinopathy of prematurity (greater than stage 2 – International classification of ROP) and mechanical ventilation, oxygen therapy, gestational age, antenatal and postnatal steroids in extremely low birth weight infants.
Methods
Neonates admitted to the neonatal intensive care unit in Lansing, Michigan, during 1993–2000 were followed to determine factors influencing the development of severe retinopathy of prematurity. Ophthalmologic examinations were started at 6 weeks and followed until resolution. We used logistic regression to estimate the relative risk (odds ratio) associated with risk factors of ROP.
Results
Of the neonates with ≤ 1500 g birth weight, admitted to the neonatal intensive care unit, 85% (616/725) survived. Severe retinopathy of prematurity was detected in 7.8% of 576 neonates who had eye examinations. Neonates of lower gestational age (≤ 25 weeks and 26–28 weeks) had an increased odds ratio of 8.49 and 3.19 for the development of severe retinopathy of prematurity, respectively, compared to those 29 weeks and older. Late postnatal steroid treatment starting after 3 weeks of life showed 2.9-fold increased odds ratio, in particular administration for two weeks and more (OR: 4.09, 95% CI: 1.52–11.03). With increasing antenatal steroids courses the risk of severe retinopathy of prematurity decreased, however, it was not significant. Lower gestational age, dependence on ventilation, and use of postnatal steroids were intertwined. Simultaneous presence of these factors seems to indicate severe disease status.
Conclusion
Prolonged and late postnatal steroids treatment in very low birth weight infants may pose an increased risk for the development of severe retinopathy of prematurity; however, use of postnatal steroids may also be a marker for severity of illness. Further studies need to focus on biologic markers in the pathogenesis of retinopathy of prematurity and to better understand the influence of therapies.
==== Body
Background
Retinopathy of prematurity (ROP) is the main cause of visual impairment in premature infants [1]. The increased survival of extremely low birth weight (ELBW) infants in recent years, due to advances in neonatal care, has produced a population of infants at very high risk of developing ROP [2]. It has been believed for many years that oxygen therapy increases the risk of ROP in preterm infants [3]. However, ROP can occur even with careful control of oxygen therapy [4]. Several factors increase the risk of ROP, especially those associated with short gestation and low birth weight [5]. Other identified risk factors include sepsis, intraventricular hemorrhage, exposure to light [6], and blood transfusions [7], and mechanical ventilation [8].
One of the recent changes in the care of low birth weight (LBW) infants is administration of antenatal steroids (ANS) to reduce the risk of respiratory distress syndrome (RDS) and neonatal death in preterm neonates [9]. The 1994 National Institute of Health consensus development conference recommended the administration of ANS for pregnancies 24 to 34 weeks threatened with premature delivery [10]. Despite strong support for giving a single course of ANS, the need is far less clear for continuing this regimen in high-risk women who have not delivered one week after their initial course. A survey of members of the Society of Perinatal Obstetricians in 1995 reported that 96% of respondents were willing to give more than one course of ANS and 58% would give more than six courses of ANS [11]. In addition, postnatal steroids (PNS) have been used increasingly for the prevention and treatment of Chronic Lung Disease (CLD) in LBW infants [12]. However, reports concerning the association between postnatal dexamethasone (PNS) use and the incidence of ROP are contradictory [13-16]. Recent Cochrane reviews concluded no significant effect of early therapy with PNS on severe ROP, but late PNS use revealed increased risk of severe ROP (RR = 1.52, 95%CI: 1.09–2.12) [17-19]. However, the conclusion for late PNS use is based only on 241 children from six studies.
Not only are there controversies around the effect of antenatal and postnatal steroids on the incidence and severity of ROP in ELBW infants, effects of multiple antenatal steroid courses on the incidence and severity of ROP has not yet been assessed. The primary focus of this study was to examine the relationship between severe ROP and ANS courses, late use of postnatal steroids and its duration, mechanical ventilation, oxygen therapy, and gestational age.
Methods
Subjects and clinical data
All very low birth weight (VLBW, ≤1500 g) neonates, hospitalized in the Neonatal Intensive Care Unit (NICU) at Sparrow Hospital from January 1993 to July 2000 and who survived until discharge, were eligible for the study. Infants with lethal congenital anomalies were excluded. Data for all patients admitted to the NICU were entered into a computerized neonatal database (medical data systems, Pennsylvania). We retrieved and analyzed data for these patients. Antenatal steroids (Betamethasone) 12 mg, 2 doses at 24-hour intervals were administered prenatally to women with threatened preterm labor at 24 – 34 weeks gestation. Gestational age was determined by either the last menstrual period or ultrasound and confirmed by neonatal examination. Surfactant was given to infants who met clinical and radiologic criteria for respiratory distress syndrome (RDS) as a rescue treatment within 2–6 hours of life. Patent ductus arteriosus was diagnosed by echocardiography and treated per protocol. Intraventricular hemorrhage was diagnosed by serial cranial ultrasound studies. Chronic Lung Disease (CLD) was defined as a requirement for supplemental oxygen at 36 weeks post-conceptional age and chest x-ray changes consistent with CLD. Usually 3–4 weeks after delivery, postnatal dexamethasone treatment was initiated at the discretion of the individual neonatologist for the treatment of CLD, for high oxygen need, or inability to wean infant from respiratory support. A single pediatric ophthalmologist did all eye examinations starting at 6 weeks after delivery. Follow up examinations were done until the resolution of ROP or retinal maturation and used the international classification of ROP to classify the severity [20]. None of the babies received additional supplementation of Vitamin E.
Statistical analyses
In order to determine whether the incidence of ROP was increased in specific exposure levels, we grouped continuous variables: gestational age: ≤ 25 weeks, 26–28 weeks, and > 28 weeks; exposure to ANS: no, 0.5, 1, 2, ≥ 3 courses; duration of ventilation: no, 1, 2, ≥ 3 weeks, exposure to PNS: no, 1, 2, ≥ 3 weeks. To assess the relative risk of PNS in multivariate models, we decided to dichotomize post-natal steroids use into any exposure versus no exposure.
We assessed risk factors for severe ROP (>stage 2). Since we had twins in the sample and data of twins are correlated, the use of logistic regression would not be justified. Hence we used Generalized Estimation Equations (SAS, PROC GENMOD) that allows adjusting for autocorrelation between observations [21]. We estimated odds ratios (OR) and their 95% confidence intervals (95% CI) for all potential risk factors. We then reduced the model to the most parsimonious set of predictors, by eliminating all potential predictors that showed no important association (OR not different from 1) and that did not change the OR of the other predictors by more than 10% when eliminated from a full model. All above mentioned procedures were carried out using SAS software version 8.02 [22].
The study protocol was approved by the human subject review committee of Michigan State University.
Results
During the observation period from 1993 to 2000, 725 very low birth weight (VLBW) infants were admitted to the neonatal intensive care unit (NICU). A total of 616 infants (85%) survived until hospital discharge. Eye examination was performed on 576 infants; 40 infants (6.5%) were lost to follow up. From 1993 to 2000, the incidence of severe retinopathy of prematurity (ROP) changed only slightly in our neonatal intensive care unit (Figure 1). The use of antenatal steroids showed a major increase during 1993 – 1994 and a minor change thereafter. The use of PNS increased during the last four years of the study. Hence on this aggregative level, there seems to be no association between severe ROP with antenatal or postnatal use of steroids.
Figure 1 Incidence of severe retinopathy of prematurity (> stage 2 – International classification of ROP) and treatment with prenatal and postnatal steroids in the neonatal intensive care unit, Lansing, Michigan.
However, on an individual level the incidence of severe ROP was associated with postnatal administration of steroids for greater than 1-week duration (Table 1). Beyond that, there was an obvious increase with gestational age and a reduced incidence in African American neonates. The explanatory model for severe ROP, mutually controlling for all predictors, is provided in Table 2. Compared to Caucasian children, the odds ratio was significantly lower in African American children. Gender, respiratory distress syndrome, oxygen treatment at 36 weeks post-conceptional age, and prenatal use of steroids were not significantly associated with severe ROP. A more parsimonious model included only three significant factors: gestational age, duration of ventilation, and postnatal steroids. The incidence of severe ROP clearly decreased with increasing gestational age (Table 2). Late postnatal treatment with steroids showed a 2.9-fold increased odds ratio (95% CI: 1.17–7.28) for the development of severe ROP. Duration of ventilation was not significant but needed to be controlled for as it changed the effect of postnatal steroids and gestational age.
Table 1 Incidence of Retinopathy of Prematurity (ROP) by patient characteristics.
n No ROP (%) ≤ stage 2 ROP (%) > stage 2 ROP (%)
Total cohort 576 63.4 28.8 7.8
Gender boys 308 64.6 27.0 8.4
girls 268 61.9 31.0 7.1
Gestational age ≤ 25 weeks 61 4.9 59.0 36.1
26–28 weeks 198 42.9 47.0 10.1
> 28 weeks 317 87.4 11.7 1.0
Race White 433 62.8 27.7 9.5
Black 109 68.8 18.7 2.8
Other 34 52.9 44.1 2.9
Use of antenatal steroids no 153 60.1 31.4 8.5
0.5 course 83 59.0 28.9 12.1
1 course 236 66.1 27.5 6.4
2 courses 51 62.8 25.5 11.8
≥ 3 courses 53 67.9 30.2 1.9
Respiratory distress syndrome no 293 78.5 18.1 3.4
yes 283 47.7 39.9 12.4
Ventilator no 200 87 12.5 0.5
1 week 174 82.8 15.5 1.7
>1 – 2 weeks 48 54.2 39.6 6.3
>2 weeks 154 13.6 61.7 24.7
Use of postnatal steroids no 422 78.7 19.2 2.1
1 week 47 37.2 61.7 2.1
> 1 – 2 weeks 38 15.8 52.6 31.6
> 2 weeks 62 6.5 21.7 48.9
Oxygen for 36 weeks no 474 70.9 23.3 4.9
yes 102 28.4 50.0 21.6
Table 2 Odds ratio for risk factors of Retinopathy of Prematurity (ROP).
Full Model Reduced Model
n = 570 n = 570
Odds-Ratio 95% CI Odds-Ratio 95% CI
Gestational age >28 weeks >28 weeks 1 - 1
26–28 weeks 4.12 1.05–16.11 3.19 0.81–12.55
≤ 25 weeks 11.27 2.61–48.66 8.49 2.0–35.94
Race Caucasian 1 -
Other 0.22 0.03–2.0
African American 0.18 0.05–0.68
Gender Boys 1 -
Girls 0.71 0.33–1.52
Antenatal steroids No 1 -
0.5 course 0.78 0.27–2.28
1 course 0.50 0.17–1.26
2 courses 0.52 0.14–1.91
≥ 3 courses 0.17 0.02–1.49
Respiratory distress syndrome No 1 -
Yes 0.96 0.39–2.37
Ventilation No 1 - 1
1 week 2.15 0.2–22.58 2.3 0.23–23.35
1 – 2 weeks 3.45 0.30–39.91 4.57 0.40–51.63
> 2 weeks 8.27 0.86–79.74 9.02 0.96–85.0
Postnatal steroids No 1 - 1
Yes 2.69 1.05–6.89 2.91 1.17–7.28
Oxygen for 36 weeks No 1 -
Yes 1.29 0.59–2.87
At the onset of the analysis, we dichotomized postnatal steroid application (none vs. any, Table 2). However, descriptive data in Table 1 suggested a threshold between one and two weeks of postnatal steroids used. Thus, we additionally analyzed a three-category variable. The analyses revealed that, compared to no postnatal steroid administration, there was no increased relative risk (odds ratio) for postnatal steroids used for seven days and less, but an increased risk if used for two weeks and more (OR: 4.09, 95% CI: 1.52–11.03). Additionally, we had information on the dose of postnatal steroids for 488 of 576 children. A steroid administration of two weeks or more corresponded to a dose of ≥ 7 mg in 73.7% of the neonates.
The three factors, gestational age, duration of ventilation, and late administration of postnatal steroids, were intertwined (Figure 2). Among infants who received postnatal steroids, 65–76% were dependent on ventilation for more than two weeks, but only 4.4% (13/296) of the infants older than 28 weeks with ≤ 2 weeks of ventilation received postnatal steroids. We further attempted to test the significance of combined effects of the three predictors on the incidence of severe ROP. However, high collinearity between them averted further statistical assessments. Nevertheless, when comparing the incidence of severe ROP in different combinations of these three factors, it was notably higher in infants treated with postnatal steroids (Figure 3), in particular infants with a low gestational age and infants dependent on ventilation for more than two weeks. The majority of cases of severe ROP were in infants with lower gestational age (<28 weeks). In this group without the use of postnatal steroids, the incidence of severe ROP increased from 3% (3/100) to 15.2% (5/33); and in the group with postnatal steroids from 20% (4/20) to 28.8% (30/104) with varying duration of ventilation.
Figure 2 Association between gestational age, duration of ventilation and postnatal steroids administration.
Figure 3 Association between gestational age, duration of ventilation, and severe ROP among VLBW infants stratified for administration of postnatal steroids.
Discussion
Our study was comprised of 576 neonates with ≤ 1500 g birth weights, of which 7.8% developed severe ROP (greater than stage 2 – International classification of ROP). Gestational age, duration of ventilation, and late postnatal steroids administration were significantly associated with severity of ROP. Gestational age and duration on ventilation were associated with development of ROP [8,23]. Consistent with published data, incidence of severe ROP was higher in Caucasians than African American infants [24,25].
Antenatal steroid use has been recommended for pregnancies 24 to 34 weeks gestation with threatened premature delivery to decrease the risk of RDS and neonatal death in premature neonates. ANS (dexamethasone) has been reported to be associated with decreased incidence of severe ROP [26]. However, Smith et al. have reported that single or multiple courses of ANS were not protective for the development of severe ROP [27,28]. In our study neither single nor multiple courses of ANS imparted a significant difference for the incidence of severe ROP.
In retrospective studies, postnatal steroid administration has been shown to either have a protective effect on the incidence of severity of ROP (case-control study with 58 newborns) [13], no effect (retrospective review with 147 newborns) [15], or on the contrary, increases the severity of ROP and need for cryotherapy (case control study with 52 newborns) [14].
Meta-analyses of prospective, randomized, placebo-controlled trials found no effect on the incidence of severe ROP when postnatal steroids were used early (within 96 hours after birth, eight studies, 1,453 infants) or within 7–14 days of life (five studies, 247 infants) for treatment of CLD [18,19]. However, a significant increase in severe ROP was reported when PNS was used after 3 weeks of life to treat CLD among 241 VLBW infants in six randomized placebo controlled trials (late administration) [17]. This is in accordance with our findings. The Cochrane meta-analysis reported an increased risk of 1.52 (95%CI 1.09–2.12); we found an increased odds ratio of 2.91 (Table 2).
Regarding duration, the six studies in the Cochrane review on late administration showed a wide variation in the duration of the treatment (from 6 to 42 days) [17]. Duration and dose have not yet been addressed in a meta-analysis. Two observational studies have reported an effect of duration and dose. Termote et al. reported that in VLBW infants only prolonged use of postnatal hydrocortisone was associated with an increased risk for severe ROP [29]. Haroon Parupia et al. suggested that a higher cumulative dose of PNS was a risk factor for severe ROP, controlling for other risk factors [30]. Consistent with these reports, our findings have demonstrated that ≥ 2 weeks duration, which corresponds to a total of ≥ 7 mg steroid dose, was associated with an increased incidence of severe ROP (OR: 4.09, 95% CI: 1.52–11.03). This may be indicative that only a longer duration of delayed PNS use is a risk factor for severe ROP.
Limitations of our study were the long duration of the period from 1993–2000 during which significant advances have been made in the treatment protocol of VLBW infants and could have influenced our results; however, we did not see any difference in the incidence of severe ROP during the 1993 to 1996 and 1997 to 2000 periods (Figure 1). We believe that it was an advantage that this observational study focused on a single tertiary care center with little variability in practice. In particular, the ophthalmologic examinations were consistent since all neonates were seen by the same ophthalmologist for the entire duration of the study. Another limitation is, however, that we do not have information on the follow-up of these infants after 6 months or after resolution of the ROP. Also, at the beginning of our study, reports on the effects of PNS on ROP were contradictory [13-16]. This study was initiated to explore predictive factors for severe ROP, but not to test a priory hypotheses.
In our study we have included babies with birth weight <= 1500 g based on the joint statements of American Academy of Pediatrics and the American Academy of Ophthalmology guidelines for screening premature infants for retinopathy of prematurity [31,32]. Our entry criteria differs from current practice, which is to screen neonates with birth weights <= 1250 g or <= 28 weeks of gestation based on the CRYO-ROP and LIGHT-ROP studies [33]. Our study also showed a substantial lower incidence of severe ROP in infants >28 weeks of gestational age (Table 1)
In our explanatory model for ROP, after excluding non-significant effects and non-confounding variables, the incidence of severe ROP was associated only with three factors (Table 2): gestational age, duration of ventilation, and PNS treatment. These associations were intertwined (Figures 2 and 3). Infants with lower gestational age were more likely to have been ventilated for a longer duration, received late PNS treatment for a longer period and developed severe ROP. We do not know whether late PNS treatment is the risk factor or a marker of underlying illness severity. A recent study reported short gestation, prolonged ventilation, frequent apnea and surfactant use as risk factors for developing severe ROP, however, these factors again may be the marker for the severity of illness [8].
Regarding the pathogenesis of ROP, abnormal vascular growth has been attributed to low insulin-like growth factor-1 (IGF-1) levels [34]. Genetic studies have shown that IGF-1 is critical in vessel development [35], and IGF-1 levels are correlated with birth weight and gestational age [36-39]. The premature infants who develop ROP have lower serum levels of IGF-1 than age-matched infants without disease [40]. Low levels of IGF-1 have been suggested to predict infants who will develop ROP and further suggest that early restoration of IGF-1 in preterm infants to normal levels could prevent this disease [34]. In order to achieve a better understanding of the pathogenesis of ROP, future studies should investigate the role of biologic markers. These studies can also determine the impact of risk factors such as steroids (antenatal and postnatal), gestational age, birthweight, oxygen administration, and duration of ventilation on biological markers and specific mechanisms in the pathogenesis of ROP.
Conclusion
Prolonged use or higher cumulative dose of postnatal hydrocortisone administered to treat chronic lung disease after 3 weeks of life was associated with an increased relative risk for retinopathy of prematurity in very low birth weight neonates.
Abbreviations
ROP, retinopathy of prematurity; ANS, antenatal steroids; PNS, postnatal steroids; NICU, neonatal intensive care unit; OR, odds ratio; 95% CI, 95% confidence intervals; ELBW, extremely low birth weight; RDS, respiratory distress syndrome; CLD, chronic lung disease; IGF-1, insulin-like growth factor-1.
Competing interests
The author(s) declare that they have not competing interests.
Authors' contributions
PK developed and designed the study and helped write the report. JM assisted in analyzing the data, literature review, and writing the discussion. LA conducted the ophthalmologic examinations in the infants and contributed to the interpretation. WK conducted the statistical analysis and worked on the report.
Pre-publication history
The pre-publication history for this paper can be accessed here:
==== Refs
Purohit DM Ellison RC Zierler S Miettinen OS Nadas AS Risk factors for retrolental fibroplasia: experience with 3,025 premature infants. National Collaborative Study on Patent Ductus Arteriosus in Premature Infants Pediatrics 1985 76 339 44 2863804
Gibson DL Sheps SB Uh SH Schechter MT McCormick AQ Retinopathy of prematurity-induced blindness: birth weight-specific survival and the new epidemic Pediatrics 1990 86 405 12 2388790
Kinsey VE Arnold HJ Kalina RE Stern L Stahlman M Odell G Driscoll JM JrJH Elliott J Payne A Patz PaO2 levels and retrolental fibroplasia: a report of the cooperative study Pediatrics 1977 60 655 68 578921
Flynn JT Bancalari E Snyder ES Goldberg RN Feuer W Cassady J Schiffman J Feldman HI Bachynski B Buckley E A cohort study of transcutaneous oxygen tension and the incidence and severity of retinopathy of prematurity N Engl J Med 1992 326 1050 4 1549150
Avery GB Glass P Retinopathy of prematurity: progress report Pediatr Ann 1988 17 520, 530, 532 3
Glass P Avery GB Subramanian KN Keys MP Sostek AM Friendly DS Effect of bright light in the hospital nursery on the incidence of retinopathy of prematurity N Engl J Med 1985 313 401 4 3839567
Brooks SE Marcus DM Gillis D Pirie E Johnson MH Bhatia J The effect of blood transfusion protocol on retinopathy of prematurity: A prospective, randomized study Pediatrics 1999 104 514 8 10469778 10.1542/peds.104.3.514
Kim TI Sohn J Pi SY Yoon YH Postnatal risk factors of retinopathy of prematurity Paediatr Perinat Epidemiol 2004 18 130 4 14996252 10.1111/j.1365-3016.2003.00545.x
Ryan CA Finer NN Antenatal corticosteroid therapy to prevent respiratory distress syndrome J Pediatr 1995 126 317 9 7844686
Effect of corticosteroids for fetal maturation on perinatal outcomes NIH Consensus Statement 1 24 Feb. 28-March 2, 1994
Planer RA Ballard PL Coburn CE Boardman CR Cnaan A Morgan MA Parer J Antenatal corticosteroid (ANCS) use in preterm labor in the USA Pediatric Research 1996 39 110A
Yeh TF Lin YJ Hsieh WS Lin HC Lin CH Chen JY Kao HA Chien CH Early postnatal dexamethasone therapy for the prevention of chronic lung disease in preterm infants with respiratory distress syndrome: a multicenter clinical trial Pediatrics 1997 100 E3 9310536 10.1542/peds.100.4.e3
Sobel DB Philip AG Prolonged dexamethasone therapy reduces the incidence of cryotherapy for retinopathy of prematurity in infants of less than 1 kilogram birth weight with bronchopulmonary dysplasia Pediatrics 1992 90 529 33 1408504
Batton DG Roberts C Trese M Maisels MJ Severe retinopathy of prematurity and steroid exposure Pediatrics 1992 90 534 6 1408505
Wright K Wright SP Lack of association of glucocorticoid therapy and retinopathy of prematurity Arch Pediatr Adolesc Med 1994 148 848 52 8044263
Ramanathan R Siassi B deLemos RA Severe retinopathy of prematurity in extremely low birth weight infants after short-term dexamethasone therapy J Perinatol 1995 15 178 82 7666264
Halliday HL Ehrenkranz RA Delayed (>3 weeks) postnatal corticosteroids for chronic lung disease in preterm infants Cochrane Database Syst Rev 2001 CD001145 11405975
Halliday HL Ehrenkranz RA Early postnatal (<96 hours) corticosteroids for preventing chronic lung disease in preterm infants Cochrane Database Syst Rev 2001 CD001146 11279706
Halliday HL Ehrenkranz RA Moderately early (7–14 days) postnatal corticosteroids for preventing chronic lung disease in preterm infants Cochrane Database Syst Rev 2001 CD001144 11279705
An international classification of retinopathy of prematurity. The Committee for the Classification of Retinopathy of Prematurity Arch Ophthalmol 1984 102 1130 4 6547831
Stokes CS Koch GG Gategorical data analysis using the SAS system SAS Institute 1995
SAS/STAT software SAS Institute Inc 2002 Cary, N.C
Italian multicentre study on retinopathy of prematurity. The Italian ROP Study Group Eur J Pediatr 1997 156 939 43 9453377 10.1007/s004310050747
Saunders RA Donahue ML Christmann LM Pakalnis AV Tung B Hardy RJ Phelps DL Racial variation in retinopathy of prematurity. The Cryotherapy for Retinopathy of Prematurity Cooperative Group Arch Ophthalmol 1997 115 604 8 9152127
Tadesse M Dhanireddy R Mittal M Higgins RD Race, Candida sepsis, and retinopathy of prematurity Biol Neonate 2002 81 86 90 11844875 10.1159/000047189
Higgins RD Mendelsohn AL DeFeo MJ Ucsel R Hendricks-Munoz KD Antenatal dexamethasone and decreased severity of retinopathy of prematurity Arch Ophthalmol 1998 116 601 5 9596495
Smith LM Qureshi N Chao CR Effects of single and multiple courses of antenatal glucocorticoids in preterm newborns less than 30 weeks' gestation J Matern Fetal Med 2000 9 131 5 10902829 10.1002/(SICI)1520-6661(200003/04)9:2<131::AID-MFM9>3.0.CO;2-M
Banks BA Cnaan A Morgan MA Parer JT Merrill JD Ballard PL Ballard RA Multiple courses of antenatal corticosteroids and outcome of premature neonates. North American Thyrotropin-Releasing Hormone Study Group Am J Obstet Gynecol 1999 181 709 17 10486488
Termote J Schalij-Delfos NE Donders AR Cats BP Do postnatal glucocorticoids and retinopathy of prematurity relate? Am J Perinatol 2000 17 291 8 11144310 10.1055/s-2000-13437
Haroon Parupia MF Dhanireddy R Association of postnatal dexamethasone use and fungal sepsis in the development of severe retinopathy of prematurity and progression to laser therapy in extremely low-birth-weight infants J Perinatol 2001 21 242 7 11533841 10.1038/sj.jp.7200531
American Academy of Pediatrics. Section on Ophthalmology Screening examination of premature infants for retinopathy of prematurity. A joint statement of the American Academy of Pediatrics, the American Association for Pediatric Ophthalmology and Strabismus, and the American Academy of Ophthalmology Pediatrics 1997 100 273 9254381 10.1542/peds.100.2.273
American Academy of Pediatrics. Section on Ophthalmology Screening examination of premature infants for retinopathy of prematurity Pediatrics 2001 108 809 811 11533356 10.1542/peds.108.3.809
Reynolds JD Dobson V Quinn GE Fielder AR Palmer EA Saunders RA Hardy RJ Phelps DL Baker JD Trese MT Schaffer D Tung B CRYO-ROP and LIGHT-ROP Cooperative Study Groups Evidence-based screening criteria for retinopathy of prematurity: natural history data from the CRYO-ROP and LIGHT-ROP studies Arch Ophthalmol 2002 120 1470 1476 12427059
Hellstrom A Perruzzi C Ju M Engstrom E Hard AL Liu JL Albertsson-Wikland K Carlsson B Niklasson A Sjodell L Low IGF-I suppresses VEGF-survival signaling in retinal endothelial cells: direct correlation with clinical retinopathy of prematurity Proc Natl Acad Sci U S A 2001 98 5804 8 11331770 10.1073/pnas.101113998
Hellstrom A Carlsson B Niklasson A Segnestam K Boguszewski M de Lacerda L Savage M Svensson E Smith L Weinberger D IGF-I is critical for normal vascularization of the human retina J Clin Endocrinol Metab 2002 87 3413 6 12107259 10.1210/jc.87.7.3413
Bennett A Wilson DM Liu F Nagashima R Rosenfeld RG Hintz RL Levels of insulin-like growth factors I and II in human cord blood J Clin Endocrinol Metab 1983 57 609 12 6348065
Giudice LC de Zegher F Gargosky SE Dsupin BA de las Fuentes L Crystal RA Hintz RL Rosenfeld RG Insulin-like growth factors and their binding proteins in the term and preterm human fetus and neonate with normal and extremes of intrauterine growth J Clin Endocrinol Metab 1995 80 1548 55 7538146 10.1210/jc.80.5.1548
Lineham JD Smith RM Dahlenburg GW King RA Haslam RR Stuart MC Faull L Circulating insulin-like growth factor I levels in newborn premature and full-term infants followed longitudinally Early Hum Dev 1986 13 37 46 3956421 10.1016/0378-3782(86)90096-4
Smith WJ Underwood LE Keyes L Clemmons DR Use of insulin-like growth factor I (IGF-I) and IGF-binding protein measurements to monitor feeding of premature infants J Clin Endocrinol Metab 1997 82 3982 8 9398700 10.1210/jc.82.12.3982
Smith LE Pathogenesis of retinopathy of prematurity Acta Paediatr Suppl 2002 91 26 8 12200894 10.1080/08035250260095771
| 15985170 | PMC1175091 | CC BY | 2021-01-04 16:31:05 | no | BMC Pediatr. 2005 Jun 28; 5:18 | utf-8 | BMC Pediatr | 2,005 | 10.1186/1471-2431-5-18 | oa_comm |
==== Front
BMC Public HealthBMC Public Health1471-2458BioMed Central London 1471-2458-5-551592462510.1186/1471-2458-5-55Research ArticleIncrease of mild disability in Japanese elders: A seven year follow-up cohort study Okochi Jiro [email protected] Department of Health Services Coordination, Graduate School of Medical Sciences, Kyushu University, Maidashi 3-1-1, Higashi-ku, Fukuoka, 812-8582 Japan2005 30 5 2005 5 55 55 13 2 2005 30 5 2005 Copyright © 2005 Okochi; licensee BioMed Central Ltd.2005Okochi; licensee BioMed Central Ltd.This is an Open Access article distributed under the 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
Japan has the highest life expectancy in the world. In a 2002 census government report, 18.5% of Japanese were 65 years old and over and 7.9% were over 75 years old. In this ageing population, the increase in the number of dependent older persons, especially those with mild levels of disability, has had a significant impact on the insurance budget. This study examines the increase of mild disability and its related factors.
Methods
All community-dwelling residents aged 65 and over and without functional decline (n = 1560), of Omishima town, Japan, were assessed in 1996 using a simple illustrative measure, "the Typology of the Aged with Illustrations" to establish a baseline level of function and were followed annually until 2002. The prevalence and incidence of low to severe disability, and their association with chronic conditions present at the commencement of the study, was analyzed. A polychotomous logistic regression model was constructed to estimate the association of each chronic condition with two levels of disability.
Results
An increase in mild functional decline was more prevalent than severe functional decline. The accumulation of mild disability was more prominent in women. The major chronic conditions associated with mild disability were chronic arthritis and diabetes in women, and cerebrovascular accident and malignancy in men.
Conclusion
This study showed a tendency for mild disability prevalence to increase in Japanese elders, and some risk factors were identified. As mild disability increasingly prevalent, these findings will help determine priorities for its prevention in Japanese elders.
==== Body
Background
Japan has the highest life expectancy in the world. In a 2002 Japanese census report, 18.5% of Japanese were 65 years old and over and 7.9% were over 75 years old. A long-term care insurance (LTCI) law was introduced in 2000 to cover both home-based and institutional care services for the large elderly population [1]. Since then, the rapid increase in the number of beneficiaries has enlarged the budgetary balance of calls and its premium rates. To access LTCI-provided services, elderly persons must comply with an eligibility test. This test is based on the physical and mental status, and it divides care needs into six categories or levels, based on the estimated amount of care resource utilization[2]. According to Ministry of Health, Labor and Welfare (MHLW) figures, beneficiaries of the at-home care service and the institutional service increased by 99% and 38%, respectively, between April 2000 and April 2003. During the same period, the number of elderly persons insured by LTCI increased by only 11%. As a result of a recent report of the MHLW, which confirmed an increase in the need for mild level care (grade 1 – support needed) from 46% to 53%, the prevention of mild disability became a focus of attention. A recent government commission on elderly care in Japan also reported that the increase in the number of elderly persons, especially those with mild disability, is endangering the insurance scheme, and the government is in the process of redesigning the scheme to refocus services for the elderly with mild disability away from direct care to preventive services. Thus, prevention or delay of the onset of functional limitation is an important objective in the health care system.
The theory of compression of morbidity suggests that life-style changes and suitable treatment for chronic illnesses can postpone the development of chronic conditions and their unwanted sequelae [3]. In the United States, later levels of disabilities and death rates are predictable from specific chronic conditions [4-7]. Very few such studies have been conducted for the Japanese population [8,9] and even fewer provide information on mild disability, which is the most common and increasing source of dependency[8].
This study aims to describe incidence and prevalence of functional decline, and to determine whether the incidence of disability at mild and severe levels is associated with age, gender and chronic conditions.
Methods
This research combines two distinct methodologies: a longitudinal cohort study, from 1996 to 2002, of functional decline based on the entire population of the elderly in a single town, and a retrospective questionnaire study of chronic conditions in the same population.
Longitudinal cohort study
The base population of the study was the population of Omishima town, Ehime prefecture. According to the 1995 census, the total population of the town was 4782, and elderly persons (aged over 65) numbered 1935 (40% of the population). The local municipality provided the researchers with a list of all elderly persons of 65 years and older taken from the residential register, and 1843 elderly persons living at home (95% of this total) were identified in August 1996. Persons who did not give written informed consent (n = 5) were excluded from the study, and this left an initial cohort of 1838.
Measurement of disability
The majority of studies estimating the incidence of elderly functional decline are based on interview or questionnaire, as opposed to the probably more reliable approach of observation [10]. The present study uses an observational instrument which, because of its simplicity and ease of use, should permit more frequent observational studies of elderly functional decline. This method is the Typology of the Aged with Illustrations (TAI).
The TAI is an instrument for the measurement of elderly function, and is composed of four scales representing mobility, eating, toileting, and mental status (Figures 1,2,3,4) [11,12].
Figure 1 TAI mobility
Figure 2 TAI mental status
Figure 3 TAI eating
Figure 4 TAI toileting
Each item in the TAI has six hierarchical states (5 to 0), representing levels of disability in each domain. Five represents no disability and 0 represents extreme disability. Each state is defined by a threshold and illustrated as shown in the example of the mobility scale (Figure 1). The levels on the mobility scale are as follows: level 5, ability of the elder to climb stairs without aid or assistive devices; level 4, can not climb stairs without aid but can walk on flat surface without aid or assistive devices; level 3, cannot walk on a flat surface without aid, but can move around using assistive devices and perform transfer independently while seated; level 2, cannot either move around or transfer while seated using assistive device or aid from the others, but can sit up and maintain seated position ; level 1, cannot either sit up or maintain seated position but can roll over on the bed without aid; and level 0, cannot roll over on a bed while lying without aid.
Its reproducibility, construct validity and concurrent validity have been established in a previous study [11]. Average weighted kappa of the four scales was 0.65 and there were no significant differences between experienced and non-experienced TAI users. It has high concurrent validity with the Functional Independence Measure (FIM).
Using TAI, 1560 elderly persons were identified as not having any functional decline, and were used to assess the incidence of disability each year for successive seven years from 1996 to 2002, in order to follow its association with age, gender, living status and presence of chronic conditions. Eighteen non-professional district welfare commissioners recorded information pertinent to elderly function in the four above-mentioned domains using the TAI. Following intensive training in its use, they assessed the function of the participants in 1996, and again each August for six consecutive years. The evaluators were asked to observe and classify the present status in the measurement month using TAI mobility, eating and toileting scales. In case of TAI mental status scale, the evaluators were asked to observe and also to interview the relative functional status. In cases of cognitive impairment, the assessment was based on interviews with family members as proxies, together with observation of the elderly person.
Retrospective questionnaire on chronic conditions
In addition to the yearly observation of elderly functions, a questionnaire covering seventeen chronic medical conditions (including the onset) was completed by the participants remaining in February 2003. Those elders who were hospitalized, institutionalized or died by 2002 were excluded from the questionnaire study since they were unable to complete the questionnaire survey. The same district welfare commissioner, who originally carried out the observation with TAI, distributed and collected the questionnaires. They also assisted respondents who had difficulty in completing the questionnaire.
The seventeen chronic medical conditions were decided based on a Ministerial statistical report on long-term care insurance law [13], and modified for the purpose of this study. They were: chronic arthritis, osteoporosis, bone fracture, chronic pain, cerebrovascular accident (CVA), heart disease, high blood pressure, diabetes, hyperlipidemia, chronic lung disease, intestinal disease, renal disease, eye disease, malignancy, depression, Alzheimer's disease and Parkinson's disease.
The questionnaire provided descriptions and definitions of the chronic conditions to facilitate understanding and eliminate recall bias as far as possible. It also included the onset years of chronic conditions. Data of social status, social activities, and health-related behavior was also recorded, but were not used in the current analysis.
Analysis
The initial sample of 1838 was used to describe the correlation of initial disability and future severity. Of these, samples not providing a disability index due to death, emigration or institutionalization were not included in the analysis of disability index.
The data of 1560 elderly persons without initial functional decline were used to assess the incidence of disability in the population studied. Using the TAI scale, the author devised a disability index [14,15], as follows. Each of the four scales of TAI has a six-level structure (Figure 1,2,3,4). Level 5 of elderly function in each scale represents no disability and was scored as 0. At level 4, the elder has one functional problem, for example in TAI mobility, in climbing stairs, and is assigned a score of 1; at level 3, the score is 2, and so on. The results of all four scales are summed to form a single index, theoretically ranging from 0 to 20, and then divided by 20 to give each individual's score for each year of the survey.
Elders with a disability score of 0 in any year were defined as no disability. Those with scores of 0.05 to 0.10 (maximum of two disabilities) were defined as suffering from mild disabilities. Those elders with an index score equal to or greater than 0.15 (more than three disabilities) were defined as suffering from severe disabilities.
The disability-free sample (n = 1560) was used to describe the prevalence and incidence of disability. For the analysis of point prevalence, the result of each year's measurement was applied. For the analysis of incidence of new mild disability from disability-free samples, the person-year method was used. Incidence of severe disability included progression of mild disability to severe disability.
The association of disability index with gender, age and chronic medical condition, over the seven years, was analyzed using applicable data from the 1560 samples.
A polychotomous logistic regression model was constructed to test the effect of each covariate on the development of functional decline at the two levels of severity using eligible data [16]. The covariates were age at base line, living status and the seventeen chronic conditions. Only chronic conditions diagnosed before 1996 were included in the analysis so as to avoid the inclusion of acute episodes of diseases, such as CVA and bone fractures.
The associations of the chronic conditions with each of the two outcome variables were tested independently, using the chi-square or Fisher's exact test, by stratification and non-stratification of gender and endpoint functional status. Only those conditions that achieved a significance level of P < 0.05 were incorporated in the logistic regression.
Finally, as the population studied showed gender differences in functional decline, separate models for men and women were constructed. All p values were two-tailed. The analyses were conducted using SPSS 11.5.1J, Windows.
Results
The cohort of 1838 elderly persons aged 65 and over was 40% male at the beginning of the study in 1996. Age range was 65 to 99 years, and average age was 73.6 years (SD 6.5) for males and 74.8 years (SD 7.1) for females.
When first observed in 1996, 1560 (85 %) of these elders had no functional decline, 180 (10 %) showed a mild level of disability, and 98(5%) showed severe disability. Additional file 1 shows the change of the status from 1996 to 2002. Higher transition to severe disability was more prominent in mild disability group (14%) compared with no disability group (4%). There was a difference of transition from no disability to mild disability between genders (male 10% versus female 23%). The transition from no disability to dead was higher in male (26% versus 13%).
The average age of sample without disability of men (n = 654) and women was 73.0(SD6.0) and 73.6(SD6.3), respectively.
By 2002, 289 of the original 1560 participants without initial disability had died, 53 were lost to follow up or emigrated, and 50 were hospitalized or institutionalized. All of them were excluded from the analysis of risk factors. Of the 1168 participants remaining in the study in 2002, 1107(96%) participants provided the initial living status data and were measured for all consecutive 7 years, and 1067 (93%) responded to the questionnaire in 2003.
Figure 5 summarizes the subjects' progress through the study.
Figure 5 Population of elderly people living at home: Flow of subjects through the study
The disability index in 1996 was significantly higher in those who died before the last measurement in 2002 using the 1273 surviving cases and 433 deaths (Mann-Whitney's U Test, P < 0.001). These 1200 cases who were measured for consecutive 7 years were analyzed to show the result of the rank correlation between the initial disability index and the disability index of subsequent years, after excluding elderly who died (n = 433), emigrated (n = 66), or were unable to participate further due to hospitalization or institutionalization (n = 66) (Table 1).
Table 1 The Mean and median of disability index and rank correlation between disability index of 1996
year 1996 1997 1998 1999 2000 2001 2002
Male (n = 453) Mean 0.006 0.006 0.010 0.012 0.015 0.018 0.027
SD 0.035 0.034 0.053 0.058 0.060 0.067 0.080
Median 0.003 0.003 0.003 0.004 0.007 0.008 0.012
Correlation* 0.629 0.475 0.433 0.363 0.324 0.327
P <0.001 <0.001 <0.001 <0.001 <0.001 <0.001
Female (n = 747) Mean 0.007 0.009 0.009 0.011 0.016 0.022 0.036
SD 0.031 0.032 0.035 0.043 0.055 0.061 0.082
Median 0.005 0.006 0.006 0.007 0.010 0.014 0.020
Correlation* 0.615 0.521 0.460 0.363 0.324 0.256
P <0.001 <0.001 <0.001 <0.001 <0.001 <0.001
*Spearman's rank correlation between disability index of 1996
Table 2 shows the distribution of disabilities, according to TAI grade, of subjects at the two levels of the index of disability, in1996. For example, of those with mild disability (n = 180), 141 (78%) could not climb stairs by themselves, but could walk on a flat surface without aid or assistive devices, and 13 (7%) could move about only on a flat surface with aid. Forty-three (24%) had mild memory problems and 15 (8%) had mild difficulty using toilet. Only two had a problem with eating. In those with mild disability, 145 (81%) showed disability on only one scale, while 35 (19%) showed disability on two scales. 84 percent of the elders had disability only in the mobility scale and this suggested that the mild disability group is composed mostly of the elders with mobility problem, without other functional problem.
Table 2 Disability level and result measured with the typology of the aged with illustrations (TAI) in 1996
mild disability (n = 180)
TAI level mobility mental eating toileting
5 14% 76% 99% 92%
4 78% 24% 1% 8%
3 7% 1% 1% 0%
2 0% 0% 0% 0%
1 0% 0% 0% 0%
0 0% 0% 0% 0%
severe disability (n = 98)
TAI level mobility mental eating toileting
5 4% 44% 61% 16%
4 21% 27% 28% 34%
3 32% 1% 7% 20%
2 21% 21% 2% 16%
1 15% 3% 1% 12%
0 6% 4% 1% 1%
Subjects with severe disability (n = 98) had a variety of functional impairment. In this group, only two cases (2%) showed disability on only one scale, both of which involved mental status dysfunction, and 4 cases (4%) had no problem with mobility.
Only one subject had a TAI mental level of 3 and only four a TAI mental level of 1, all of whom exhibited problem behaviors, as shown in the Figure 2.
Prevalence and incidence of disability
Figure 6 shows the change of the median of the disability index for consecutive seven years (n = 1107). In this analysis, only the samples that were measured for consecutive 7 years were included, and therefore, the elderly persons who were hospitalized, institutionalized, dead or emigrated were excluded. Older age at base line had an effect on the rate of the disability development. In base line age group older than 75, the increase of the disability index was more prominent in woman than that of man after year 2000. Most of the curves, except for the man aged 75 and over, showed an exponential increase pattern.
Figure 6 Pattern of disability index median, in gender and age groups (n = 1107)
The associations of gender and age with scores on the disability index were tested separately on yearly data using the eligible samples from the same population (n = 1107). Gender difference was not obvious from 1997 to 2001, but in 2002 women showed a higher mean disability index score than men (male 0.22, female 0.30, T test, P < 0.05). Age at enrolment, in 1996, correlated positively with disability index scores for every year of measurement (Spearman's rank correlation, P < 0.01).
The point prevalence of disability at the two severity levels, and of institutionalization and death, is shown in Fig. 7. Mild disability was more increased in women than in men, rising to 22% in women versus 10% in men by 2002. By contrast, loss from the study population by death was more common in men, reaching 26% in men versus 13% in women by 2002. The proportion of severe disability in 2002 was 3.4% and 4.2% for men and women, respectively, and the proportion of elders who were institutionalized in 2002 was 3.4% and 4.2%, respectively.
Figure 7 Prevalence of disability, institutionalization and death in men versus women without initial disability (n = 1560)
As shown in Table 3, age-group in 1996 also had an effect on the development of the disability. In men, higher age group showed higher proportion of death in 2002, while institutionalization was higher in younger age group. In women, both the proportion of death and institutionalization were higher in older age group.
Table 3 Age and gender difference of the disability, institutionalization and death without initial disability (n = 1560)
gender age group no disability mild disability severe disability institution* emigrated† death
Male 65 to 74 n 271 38 14 15 11 78
(n = 427) % 63 9 3 4 3 18
75 and over n 82 29 8 4 10 94
(n = 227) % 36 13 4 2 4 41
Total n 353 67 22 19 21 172
(n = 654) % 54 10 3 3 3 26
Female 65 to 74 n 375 94 12 15 11 39
(n = 546) % 69 17 2 3 2 7
75 and over n 108 111 26 16 21 78
(n = 360) % 30 31 7 4 6 22
Total n 483 205 38 31 32 117
(n = 906) % 53 23 4 3 4 13
* including hospitalization
† including loss from the sample with unknown reasons
Figure 8 shows the yearly incidence of new cases at the two levels of disability by gender. A high incidence of mild disability compared to severe disability was particular to females. The incidence of death was higher in men in all 6 consecutive years (data not shown).
Figure 8 Incidence of disability by severity and gender
Factors associated with disability
Additional file 2 shows baseline chronic conditions in 1996, cross-tabulated with the outcome level of disability by gender. The average age of males (n = 405) completing this part of the study was 71.6 years (SD 4.9), and of females, 72.6 years (SD 5.8). At least one chronic conditions was reported by 671 (61.4%) of this group.
The association of number of chronic conditions with scores on the disability index was tested separately on yearly data using this sample (n = 1067). The number of chronic condition was correlated with disability index score in 1998 (P < 0.05), 2000, 2001 and 2002 (P < 0.01, respectively).
The selection of elderly persons without functional decline at the commencement of the study excluded the participants suffered from Alzheimer disease or other dementia. Depression (n = 11), Parkinson's disease (n = 7) and Alzheimer disease (n = 0), which were too low in prevalence in 1996 to permit statistically meaningful analysis, are not included in this table. Chronic arthritis, osteoporosis, bone fracture, cerebrovascular accident (CVA), diabetes, chronic lung disease, eye disease and malignancy all showed significant associations with level of disability (chi-square test or Fisher's exact test, stratifying and non-stratifying the outcome severity level).
The chronic conditions with statistically significant associations and age at initial measurement were used to construct the polychotomous logistic regression model shown in Table 4. The conditions found to be related to mild disability in males were CVA and malignancy, and that to severe disability was CVA. The chronic conditions related to mild disability in women were chronic arthritis and diabetes, and those related to severe disability were chronic arthritis and CVA. The results for severe disability must be interpreted cautiously, because of the limited number in the end-point sample; the confidence interval for relative risk is larger than that for mild disability.
Table 4 Associations of Chronic Conditions and Age with Functional Decline in Participants Without Initial Functional Limitation (n = 1067)
outcome level
mild severe
Gender Covariate R.R 95% C.I. P R.R 95% C.I. P
Male
age* 2.5 (1.4–4.5) P < 0.01 5.3 (2.1–13.2) P < 0.01
chronic arthritis 1.9 (0.6–6.2) 1.6 (0.3–8.6)
osteoporosis† 4.4 (0.5–36.2) n.a
bone fracture 2.6 (0.8–8.1) 1.7 (0.2–14.8)
CVA§ 5.6 (1.7–19.1) P < 0.01 20.3 (5.2–78.6) P < 0.01
diabetes 1.2 (0.4–3.1) 2.5 (0.6–10.1)
chronic lung disease 2.2 (0.9–5.6) 2.9 (0.7–12.0)
eye disease 1.0 (0.4–2.4) 1.0 (0.3–3.6)
malignancy† 5.4 (1.6–18.3) P < 0.01 n.a
Female
age* 4.9 (3.4–7.1) P < 0.01 9.0 (4.4–18.2) P < 0.01
chronic arthritis 2.8 (1.5–5.2) P < 0.01 5.4 (1.9–15.8) P < 0.01
osteoporosis 1.4 (0.6–3.2) 2.6 (0.7–9.9)
bone fracture 1.3 (0.6–3.1) 1.1 (0.3–4.9)
CVA§ 3.4 (0.6–19.8) 22.3 (2.5–198.5) P < 0.01
diabetes 2.6 (1.2–5.9) P < 0.05 1.5 (0.2–13.2)
chronic lung disease 0.3 (0.1–1.1) 1.5 (0.3–8.0)
eye disease 1.1 (0.6–1.8) 0.8 (0.3–2.3)
malignancy† 0.8 (0.3–2.3) n.a
*effect of ten-year increase, † insufficient numbers in category, §cerebrovascular accident
Because of the relatively low prevalence of chronic conditions, the sum of chronic conditions suffered was used to determine the effect of multiple conditions. The relative risks, of adding 1 chronic condition for severe and for mild disability were 1.2 (95%CI 1.0–1.4) and 1.2 (1.1–1.3), respectively, controlling for age and gender.
Discussion
The aim of the current study was to describe incidence and prevalence of disability and to identify the effect of a age, gender, living condition and chronic conditions as risk factors of functional decline in Japanese elders, and to associate them with different degrees of disability. The identification of risk factors that correlate with the development of mild disability, and which serve as are suitable targets for prevention, is of particular importance in today's society, where increasing prevalence of mild disability and of costly dependency of the aged is clearly apparent [17,18]. The present study is also of interest for its use of a base population in which the proportion of elders, aged 65 and over was 40%. To the best of the author's knowledge, this is the most aged society studied epidemiologically to date.
Prevalence and incidence of disability
This study initially used the disability index to show the occurrence of disability in the population. This index had an exponential distribution, i.e. the most of the elders have no disability as shown in Table 1 and Figure 6, since the sample represents a normal population, as is in a previous study[14].
The speed of disability development was different among age-groups and genders, suggesting that there is different underlying process for developing disability among these groups (Figure 6). To examine this difference, the author divided the disability into two categories; mild and severe disability. And the change was prominent in the mild disability group (Figure 7).
Although the mortality rate of this cohort was within the range of that of other studies, the prevalence of overall disability was higher than in some other studies in Japan [19-21]. One study, for example, reported a lower rate of mobility disability compared to the present study [21]. This may be because of fine categorization in TAI definitions, as it classifies the elders who have problem to climb stairs into mild disability. In a previous study, stair climbing was categorized to be the difficult task, compared to other ADL and mobility items[22]. The measurement instrument in this study employs it as a tool to detect mild disability. Repeated measurement will likely show a higher chance of identifying more disability[23], and the very aged population might also have been responsible for this difference in prevalence.
The present study found an increase of mild disability in the cohort, especially in women. These findings appear to differ from those of previous studies which found that men show a faster decline than women in the Japanese population[21]. However, this result is in accordance with that of women having a longer survival time, and therefore the disability accumulates in women [24]. In woman, the transition from no disability to mild disability was higher in both age groups than men (Table 3). Higher disability index after 2000 in woman aged 75 and over also supports the accumulation of disability in woman. In men, higher age group showed higher proportion of death in 2002, but it did not apply to the cases of institutionalization. This suggested the non-exponential pattern of increase of disability index median in men (Figure 6) was attributable to the death, but not to institutionalization.
The gender difference of the proportion of elders with severe disability was not as prominent as with mild disability. These results suggest different factors are associated with the development of disability in two genders, especially in the development of mild functional limitation.
Factors associated with disability
Earlier studies in Japan have identified a variety of chronic conditions as related to the development of task specific ADL or IADL disability [19,25,26]. The association of chronic diseases with both physical and cognitive function has been investigated [8,19]. However, to date, no studies using a cohort design and a Japanese sample have reported the association of number and type of chronic conditions with severity levels of disability to the best of the authors' knowledge.
It is reasonable to hypothesize that different kinds of chronic conditions will have different functional sequelae, and there is some empirical evidence both in the US and in Japan that different risk factors are associated with reduced performance on different levels of disability.
Previous non-Japanese studies have estimated the risk associated with chronic conditions for the development of different levels of functional or ADL disability[4,6,7]. In the present study, the principle associates of both levels of disability for men were CVA and malignancy, while in females they were chronic arthritis, CVA and diabetes, as shown in Table 4. These findings are similar, but not identical, to those found in a previous Japanese study [25].
Some chronic conditions might relate to earlier death of the participants. The weak association between the number of chronic conditions and the disability index in earlier years, namely 1997 and 1999, might be due to exclusion of deceased and institutionalized cases.
Of the chronic condition studied, CVA is the most frequently cited as to have association with functional decline [8,25,27], but it has been shown that, because of the short survival time after stroke, the number of dependent elderly persons does not necessarily increase as a result [21]. This study also showed the association of the severe disability and CVA. And the incidence and prevalence of disability did not increase as much as the mild disability.
In contrast, chronic arthritis is consistently found to be a risk factor for both genders, and shows no association with mortality [4]. As might be expected, studies have indicated that the prevention of disabilities consequent on non-fatal conditions, such as chronic arthritis, is the most cost-effective preventative strategy [17,18]. The present study confirmed the significance of chronic arthritis, in women only, both for its high prevalence (9%) and its high relative risk for the development of both mild and severe disability.
This study also found the association between diabetes and mild disability in woman. In a Japanese population, Kishimoto et al. reported that, a history of diabetes is associated with poor performance on more ADL tasks in women than in men [26]. Diabetes has been shown to be associated with slower walking speed, inferior lower extremity function, and decreased balance[28], all of which meet characteristics of mild disability in the present study.
Many previous studies have suggested bone fracture and osteoporosis are risk factors for functional disability [29,30]. In the present study, however, while bone fracture and osteoporosis, in women only, appeared to be associated using Fisher's exact test, it failed to show a relationship in the logistic regression model. Ross et al. have suggested that the risk of falls among Japanese women is lower than for Caucasian women [31]. The low prevalence of these conditions in non-disabled persons may have contributed to this result. In addition, it is possible that the six-year analysis interval used in the current logistic regression analysis was too long for the detection of effects of bone fracture[23].
The prevalence of the chronic condition that achieved statistical significance with chi-square test was highest in eye disease in women, but it did not show association in the logistic regression model. Next to it was the chronic arthritis, osteoporosis and bone fracture, followed by the diabetes. In men, chronic lung disease is the highest followed by the CVA then chronic arthritis. This result suggested different approach in prophylaxis is required to prevent accumulation of disability in the population.
Study limitation
The present study has a few limitations. The history of physician-diagnosed chronic medical conditions and self-reports of the same were obtained retrospectively. A previous study had found that self-report of chronic conditions in the elderly was accurate [32], but inaccurate recall of the time of onset of chronic conditions was present, especially for arthritis [33]. Current ignorance of the prevalence of chronic conditions among well-functioning Japanese elders also limits the interpretation of the prevalence of chronic conditions among this sample. The exclusion from the analysis of participants who died or were institutionalized or emigrated in the course of the study, some of whom may have exhibited a chronic condition at baseline, may also have affected the results since those who were included for the analysis of risk factors were younger and thus were presumably healthier. The absence of information regarding to the levels of severity of the chronic conditions reported, and the relatively low prevalence of each chronic condition, meant that the associations measured were less specific than could be desired. Chronic conditions such as osteoporosis and Parkinson's disease that did not achieve statistical significance in this study may in fact contribute to the development of disability with a larger sample. Some conditions could be related to the development of disabilities in shorter or longer period of observation.
This study did not incorporate those elders who were institutionalized or dead at the endpoint for the analysis of the risk factors. This is because only 24 percent of the institutionalized cases provided responses to the questionnaire study, and none did so in the deceased cases, compared to 90 percent of the surviving cases. Inclusion of these endpoints could have improved association with the risk factors.
In addition, caution should be exercised with regard to extrapolation of the results to other populations due to the use of a single base population. However, the present study does have the advantage of using a whole population rather than a sample. By using geographically defined area, this study had little loss of the data throughout the 7 consecutive years.
Other methodological approach of analysis, such as the use of Structural Equation Model (SEM) could have been more appropriate with this data. However the stability of the model when applied for this analysis was poor, mainly because of the distribution of the endpoint variables used in this study.
Despite the limitations, this study is significant in that it provides information on the incidence and prevalence in Japan of two levels of disability – mild and severe – and gives indication of priorities in the selection of chronic conditions for prophylaxis, especially as regards to the elderly with mild disability over a lengthy period. In the context of long-term care insurance in Japan, and plans to direct services for mildly impaired elderly persons towards rehabilitation, this study can be employed to develop suitable objectives in the prevention of unwanted sequelae of chronic conditions. This study also suggested that man and woman require different prophylaxis, because different factors were associated with the development of disabilities in two genders.
Population-based studies using TAI in another Japanese town, at two- and six-month observation intervals have been initiated by the author and collaborators, in order better to understand the functional loss process and its risk factors.
Conclusion
This study showed a tendency for mild disability prevalence to increase in Japanese elders, especially in women. This study also identified some risk factors in the development of mild disability; chronic arthritis and diabetes for women and the CVA for men. In Japan, the budgetary balance of the newly instituted long-term care insurance system is endangered by increase in the mildly impaired elderly, and these findings should help determine priorities for prevention.
Competing interests
The author has not received 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.
The authors does not 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. I am not applying for any patents relating to the content of the manuscript. The author of this article has not received reimbursements, fees, funding, or salary from an organization that holds or has applied for patents relating to the content of the manuscript. I do not have any other financial competing interests.
Authors' contributions
Jiro Okochi carried out the study design, data collection, statistical analysis and preparation of the manuscript.
Pre-publication history
The pre-publication history for this paper can be accessed here:
Supplementary Material
Additional File 1
The status change of the participants between 1996 and 2002 by gender (n = 1838)
Click here for file
Additional File 2
Baseline characteristics, prevalence of chronic conditions and endpoint functional status of remaining participants in 2002 (n = 1067)
Click here for file
Acknowledgements
The author gratefully acknowledges the support received from Omishima Health Service Center. Special thanks go to Dr. Tai Takahashi (Professor of Hospital Management at the International University of Health and Welfare, Tochigi, Japan), Dr. Shinya Matsuda (Professor of Public Health at the University of Occupational and Environmental Health, Japan), Dr. Kohei Akazawa (Professor of Medical Statistics at Niigata University), and Dr. Nobufumi Yasuda (Associate Professor of Public Health at Kochi University).
==== Refs
Matsuda S The health and social system for the aged in Japan Aging Clin Exp Res 2002 14 265 270 12462371
Tsutsui T Muramatsu N Care-needs certification in the long-term care insurance system of Japan J Am Geriatr Soc 2005 53 522 527 15743300 10.1111/j.1532-5415.2005.53175.x
Katz S Branch LG Branson MH Papsidero JA Beck JC Greer DS Active life expectancy N Engl J Med 1983 309 1218 1224 6633571
Boult C Kane RL Louis TA Boult L McCaffrey D Chronic conditions that lead to functional limitation in the elderly J Gerontol 1994 49 M28 36 8282978
Fried LP Guralnik JM Disability in older adults: evidence regarding significance, etiology, and risk J Am Geriatr Soc 1997 45 92 100 8994496
Furner SE Rudberg MA Cassel CK Medical conditions differentially affect the development of IADL disability: implications for medical care and research Gerontologist 1995 35 444 450 7557514
Guralnik JM Kaplan GA Predictors of healthy aging: prospective evidence from the Alameda County study Am J Public Health 1989 79 703 708 2729467
Ho HK Matsubayashi K Wada T Kimura M Kita T Saijoh K Factors associated with ADL dependence: A comparative study of residential care home and community-dwelling elderly in Japan. Geriatrics & Gerontology International 2002 2 80 86 10.1046/j.1444-1586.2002.00026.x
Shinkai S Kumagai S Fujiwara Y Amano H Yoshida Y Watanabe S Ishizaki T Suzuki T Shibata H Predictors for the onset of functional decline among initially non-disabled older people living in a community during a 6-year follow-up. Geriatrics & Gerontology International 2003 3 S31 S39 10.1111/j.1444-0594.2003.00094.x
Stuck AE Walthert JM Nikolaus T Bula CJ Hohmann C Beck JC Risk factors for functional status decline in community-living elderly people: a systematic literature review Soc Sci Med 1999 48 445 469 10075171 10.1016/S0277-9536(98)00370-0
Okochi J Takahashi T Takamuku K Matsuda S Takagi Y Reliability of a geriatric assessment instrument with illustrations Geriatr Gerontol Int 2005 5 37 47 10.1111/j.1447-0594.2005.00268.x
Takahashi T Okochi J Takamuku K Matsuda S The introduction of typology of the aged with illustrations. Casemix Quarterly 2001 3 3 14
Ministry of Health Welfare and Labor Survey on the long-term care insurance providers
Mitnitski AB Graham JE Mogilner AJ Rockwood K The rate of decline in function in Alzheimer's disease and other dementias J Gerontol A Biol Sci Med Sci 1999 54 M65 9 10051857
Mitnitski AB Mogilner AJ Graham JE Rockwood K Techniques for knowledge discovery in existing biomedical databases: estimation of individual aging effects in cognition in relation to dementia J Clin Epidemiol 2003 56 116 123 12654405 10.1016/S0895-4356(02)00581-4
Hosmer DW Lemeshow S Applied logistic regression Wiley Series in Probablity and Statistics 2000 New York , JohnWiley and Sons,Inc 161 172
Bonneux L Barendregt JJ Nusselder WJ der Maas PJ Preventing fatal diseases increases healthcare costs: cause elimination life table approach Bmj 1998 316 26 29 9451262
Nusselder WJ van der Velden K van Sonsbeek JL Lenior ME van den Bos GA The elimination of selected chronic diseases in a population: the compression and expansion of morbidity Am J Public Health 1996 86 187 194 8633734
Fujiwara Y Shinkai S Kumagai S Amano H Yoshida Y Yoshida H Kim H Suzuki T Ishizaki T Haga H Watanabe S Shibata H Longitudinal changes in higher-level functional capacity of an older population living in a Japanese urban community. Arch Gerontol Geriatr 2003 36 141 153 12849088 10.1016/S0167-4943(02)00081-X
Ishizaki T Watanabe S Suzuki T Shibata H Haga H Predictors for functional decline among nondisabled older Japanese living in a community during a 3-year follow-up J Am Geriatr Soc 2000 48 1424 1429 11083318
Sauvaget C Yamada M Fujiwara S Sasaki H Mimori Y Dementia as a Predictor of Functional Disability: A Four-Year Follow-Up Study. Gerontology 2002 48 226 233 12053112 10.1159/000058355
Granger CV Hamilton BB Linacre JM Heinemann AW Wright BD Performance profiles of the functional independence measure AM J PHYS MED REHABIL 1993 72 84 89 8476548
Gill TM Hardy SE Williams CS Underestimation of disability in community-living older persons J Am Geriatr Soc 2002 50 1492 1497 12383145 10.1046/j.1532-5415.2002.50403.x
Ferrucci L Guralnik JM Simonsick E Salive ME Corti C Langlois J Progressive versus catastrophic disability: a longitudinal view of the disablement process J Gerontol A Biol Sci Med Sci 1996 51 M123 30 8630705
Kamiyama T Muratani H Kimura Y Fukiyama K Abe K Fujii J Kuwajima I Ishii M Shiomi T Kawano Y Mikami H Ibayashi S Omae T Factors related to impairment of activities of daily living Intern Med 1999 38 698 704 10480299
Kishimoto M Ojima T Nakamura Y Yanagawa H Fujita Y Kasagi F Kodama K Ueda K Suzuki S Kagamimori S Relationship between the level of activities of daily living and chronic medical conditions among the elderly J Epidemiol 1998 8 272 277 9884476
Ikebe T Ozawa H Lida M Shimamoto T Handa K Komachi Y Long-term prognosis after stroke: a community-based study in Japan J Epidemiol 2001 11 8 15 11253911
Volpato S Blaum C Resnick H Ferrucci L Fried LP Guralnik JM Comorbidities and impairments explaining the association between diabetes and lower extremity disability: The Women's Health and Aging Study Diabetes Care 2002 25 678 683 11919124
Greendale GA Barrett-Connor E Ingles S Haile R Late physical and functional effects of osteoporotic fracture in women: the Rancho Bernardo Study J Am Geriatr Soc 1995 43 955 961 7657934
Lorrain J Paiement G Chevrier N Lalumiere G Laflamme GH Caron P Fillion A Population demographics and socioeconomic impact of osteoporotic fractures in Canada Menopause 2003 10 228 234 12792295 10.1097/00042192-200310030-00010
Ross PD Norimatsu H Davis JW Yano K Wasnich RD Fujiwara S Hosoda Y Melton LJ A comparison of hip fracture incidence among native Japanese, Japanese Americans, and American Caucasians Am J Epidemiol 1991 133 801 809 2021147
Bush TL Miller SR Golden AL Hale WE Self-report and medical record report agreement of selected medical conditions in the elderly Am J Public Health 1989 79 1554 1556 2817172
Kehoe R Wu SY Leske MC Chylack LTJ Comparing self-reported and physician-reported medical history Am J Epidemiol 1994 139 813 818 8178794
| 15924625 | PMC1175092 | CC BY | 2021-01-04 16:28:56 | no | BMC Public Health. 2005 May 30; 5:55 | utf-8 | BMC Public Health | 2,005 | 10.1186/1471-2458-5-55 | oa_comm |
==== Front
BMC Public HealthBMC Public Health1471-2458BioMed Central London 1471-2458-5-571593509110.1186/1471-2458-5-57Research ArticleGender and age differences among current smokers in a general population survey John Ulrich [email protected] Monika [email protected] Christian [email protected] Anja [email protected] Institute of Epidemiology and Social Medicine, University of Greifswald, Walther-Rathenau-Str. 48, D-17487 Greifswald, Germany2005 3 6 2005 5 57 57 9 2 2005 3 6 2005 Copyright © 2005 John et al; licensee BioMed Central Ltd.2005John 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
Evidence suggests a higher proportion of current smokers among female than among male ever smokers at the age above 50. However, little is known about the proportion of current smokers among ever smokers in old age groups with consideration of women in comparison to men from general population samples. The goal was to analyze the proportions of current smokers among female and among male ever smokers including those older than 80.
Methods
Cross-sectional survey study with a national probability household sample in Germany. Data of 179,472 participants aged 10 or older were used based on face-to-face in-home interviews or questionnaires. The proportions of current smokers among ever smokers were analyzed dependent on age, age of onset of smoking and cigarettes per day including effect modification by gender.
Results
Proportions of current smokers tended to be larger among female than among male ever smokers aged 40 or above. Women compared to men showed adjusted odds ratios of 1.7 to 6.9 at ages 40 to 90 or older in contrast to men. No such interaction existed for age of onset of smoking or cigarettes per day.
Conclusion
Special emphasis should be given to current smokers among the female general population at the age of 40 or above in public health intervention.
==== Body
Background
Quitting smoking is a major public health goal throughout the life span. Benefits of smoking cessation exist even at the age of over 60 years [1,2]. Although there is large and consistent evidence that the proportion of current smokers declines by age including those above 65 [3-7], there are substantial subpopulations that maintain this health detrimental behavior [6]. Even among the oldest olds, individuals maintain smoking [8]. Little is known about associations with gender in this process. Whereas in 1994 the male population in the USA had higher proportions of current smokers than the female population, women and men aged 65 or older did not differ in their proportions of current smokers (female current smokers: 11.1 %, 95 % confidence interval, CI, 9.8 – 12.4; male current smokers: 13.2 %, CI 11.3 – 15.1 %) [3]. One main outcome of public health intervention is the quit rate, expressed as the proportion of former among the ever smokers in a specified period of time. An alternative measure is the proportion of current smokers among ever smokers, especially when exploring factors that may be barriers for quitting.
The proportion of current smokers among male ever smokers aged 65 or older in the USA was 18.5 % (CI 15.0 – 21.0 %) in 1994, which was lower than the equivalent proportion among female ever smokers aged 65 or more (29.3 %; CI 26.1 – 32.5 %) [3]. Female and male smokers did not differ according to cessation rates. However, female smokers compared to male smokers had an odds ratio of 1.9 for relapse in a study about the prediction of smoking cessation and relapse among individuals aged over 50 years [9]. In an Italian general population sample of individuals aged 65 to 84 the proportion of current smokers among female ever smokers was 51.9 % at age 65 – 69 and 36.1 % at age 80 – 84, and among male ever smokers 34.1 % at age 65 – 69 and 20.7 % at age 80 – 84 [6]. In Great Britain, a trend towards lower decrease in the proportion of current smokers among female ever smokers than among male ever smokers was observed based on household survey data using birth cohorts from 1897 to 1951 (age range 25 – 83) [4]. With increasing age, the proportion of current smokers tended to be higher among female than among male ever smokers [4]. Data from Finland, although limited to the age of 64 or younger, revealed a lower rate of individuals who stopped smoking among the female population at age 60 – 64 in comparison to the male population aged 60 – 64 [10].
One reason for the higher proportion of current smokers among older female compared to older male ever smokers might be that the women may be lighter smokers than men and thus are less prone to smoking-attributable disease. Women at a certain age might have smoked less over the lifetime and might have started smoking later in life than men at that age. Female smokers might feel less burden from consequences of smoking compared to men. This assumption is supported by evidence showing a lower tobacco-attributable mortality and fewer years of potential life lost from smoking in women than in men [11]. Accordingly, current smokers among women might fear less threat from disease and death that is attributable to smoking.
Altogether, little is known about the proportion of individuals who maintain smoking in old age among ever smokers with consideration of women compared to men from general population samples, particularly from countries with little activity in smoking prevention. Studies done so far did not include substantial numbers of female and male smokers aged above 65 years, did not focus on proportions of current smokers among ever smokers or did not consider effect modification by gender in a multivariate data analysis. The goal of the present paper was to analyze the proportions of current smokers among female and among male ever smokers across an age range of 10 or older in a general population sample.
Methods
Sample
We used the "Mikrozensus", a cluster household sample representative for residents in Germany across the whole age range [12]. Data were collected in April 1999. At that time Germany was a country with very few activities in smoking prevention. Accordingly, there was an underdeveloped anti-smoking climate in the nation [13] and a lower intention to quit smoking compared to countries with more activities in smoking prevention [14]. Every member of each selected household was included into the study as a participant. Participation was mandatory by law for a core part of the assessment, mainly including questions about demographic data, housing and employment. Of the eligible individuals, 97.4 % participated in the core part of the study (N = 724,139). An additional, voluntary part included questions about tobacco smoking in a 45 % random subsample. Available for scientific use were the data of a 70 % random subsample of the 724,139 participants from the German Federal Statistical Office (N = 506,897) [12]. Among this subsample, 232,397 (45.8 %) had participated in the voluntary part of the Mikrozensus. We excluded children aged less than 10 years (n = 22,156). Among the remaining individuals, there were 30,769 with missing data for smoking status. They were excluded from the data analysis as well. Subjects with and without information about the smoking status did not differ in gender, age, school education and having been ill during the last four weeks prior to the interview, when effect sizes are considered. Among those without information about the smoking status, 52.7 % were women, and among those with information about smoking status 52.1 % were women (Likelihood chi2 3.0; not significant; Cohen's w .004 [15]). Among women without information about smoking the mean age was 45.5 (Standard deviation, Std, 21.9), and among women with information about smoking status it was 46.7 (Std 20.7; t-test; p < .001; d 0.06 [15]). Among women without information about the smoking status, there were 17.4 % aged 70 or older, and among women with information about the smoking status this figure was 16.6 (Likelihood chi2 632.3; p < .001; Cohen's w .08 [15]). Among men without information about smoking the mean age was 41.0 (Std 18.9), and among men with information about the smoking status it was 43.6 (Std 19.1; t-test, p < .001; d 0.13 [15]). Among men without information according to smoking status, there were 8.1 % at the age of 70 or older, among men with information about smoking status 9.9 % (Likelihood chi2 420.0; p < .001; Cohen's w .06, [15]). For school education Cohen's w was .06 for women and .05 for men, for having had a disease during the last four weeks prior to the interview Cohen's w was .02 for women and .03 for men indicating no effect. The final sample consisted of 179,472 residents, among them 78,959 ever smokers (44.0 %). The data of these female and male ever smokers were used in our analysis.
The final sample is representative for the general population of Germany aged 10 or older with respect to gender and age [16]: In the 10 year age groups, the maximum deviance of the proportion of women in the final sample from the proportion of women in the same 10 year age groups of the general population was 2.4 percentage points (mean deviance: 1.0 percentage points). The maximum deviance in the distribution of gender over the 10 year age groups was 1.3 percentage points (mean deviance: 0.6 percentage points) among women, among men the maximum deviance was 1.7 percentage points (mean deviance: 0.7 percentage points).
Assessments
Every person of the household able to understand and to answer the interview questions and present in the household when the interviewer showed up responded in the face-to-face interview. The persons gave information about those household members that were not present. If nobody could be contacted personally the interviewer left a questionnaire to be filled in for all household members. The interview and the questionnaire included the same questions. According to smoking, the individuals were asked: "Are you currently a smoker?" (Yes, regularly/Yes, occasionally/No), if not: "Did you smoke in the past?" (Yes, regularly/Yes, occasionally/No). If the respondent answered "Yes" to one of the two questions s/he was asked: "How old have you been when you started smoking?" (age in years). "What do/did you smoke predominantly?" (cigarettes, cigars, pipe tobacco). "How many cigarettes do you/did you smoke per day?" (less than 5, 5 – 20, 21 – 40, more than 40). The data included the age of the respondent in years, and those aged 95 or older as one group. The individuals were asked "Have you been ill during the last 4 weeks (including today) or have you been injured by accident?" (Yes, ill/Yes, injured by accident/No), if yes: "How long did your illness or your injury last? (7 categories that were recoded to: 1 year or less, more than 1 year). School education was assessed by the respondent's school graduation (not including university degrees; degrees 1–5 for lowest to highest education). Income per individual was assessed by 18 categories.
Statistical analysis
For the bivariate data analysis, means with standard deviations, proportions and Likelihood chi2 tests were used. For proportions, the effect size estimate Cohen's w was included, and values .10 or higher were interpreted as indicating an effect [15]. For means, the effect size measure d was used with values 0.20 to 0.49 interpreted as indicating a small and values 0.50 to 0.79 indicating a medium effect [15]. For the multivariate data analysis, logistic regression analysis was performed. SPSS 13.0 was used for all analyses.
Results
Proportion of current smokers
According to the bivariate data analysis, among the 179,472 residents there were 78,959 ever smokers (44.0 %). Among the 93,588 women, there were 33.4 %, among the 85,884 men 55.6 % ever smokers, more among men than among women (Chi2 9031.6; df 1; p < .001; w .23). The female ever smokers included 64.0 %, the male ever smokers 58.7 % current smokers (Chi2 224.3; df 1; p < .001; w .05). For age groups below 40, the data revealed lower proportions of current smokers among the female ever smokers (Table 1). For age groups 40 or older, there were higher proportions of current smokers among female than among male ever smokers with effects in the age groups between 60 and 89 years. No effect existed among the age groups 10 – 19 and 30 – 59.
Table 1 Percent current smokers among female and male ever smokers*
Women Men Chi2 P w
n ever
smokers % current smokers
within ever smokers n ever
smokers % current smokers
within ever smokers
Total 31228 64.0 47731 58.7 224.3 .001 .05
Age
10 – 19 1396 92.9 1803 95.5 9.4 .01 .06
20 – 29 4770 78.2 5784 85.6 99.2 .001 .10
30 – 39 8084 69.3 9760 74.2 51.4 .001 .05
40 – 49 6988 64.8 9463 64.1 0.8 ns
50 – 59 4567 56.6 8149 50.7 40.9 .001 .06
60 – 69 2866 48.7 7334 37.2 112.2 .001 .10
70 – 79 2049 34.4 4340 22.3 103.0 .001 .13
80 – 89 468 29.7 1004 19.5 18.2 .001 .11
90 or older 40 30.0 94 20.2 1.5 ns
Age of onset
of smoking
10 – 15 5543 72.2 9793 64.8 90.5 .001 .08
16 5219 66.0 9258 62.2 20.8 .001 .04
17 – 18 7621 62.4 12796 55.6 90.5 .001 .07
19 – 20 4618 60.6 6506 52.8 67.6 .001 .08
21 – 25 3023 57.0 3806 54.7 3.6 ns
26 – 70 2717 59.2 1911 54.3 11.2 .001 .05
Cigarettes per day
less than 5 6894 52.0 5431 52.6 0.4 ns
5 – 20 20083 68.4 29398 61.4 256.0 .001 .07
21 – 40 2625 72.7 7567 62.6 89.0 .001 .09
41 or more 272 50.0 1200 34.8 21.2 .001 .12
Disease
(last 4 weeks)
not diseased 27042 65.4 40755 60.5 163.4 .001 .05
diseased for
1 year or less 2074 58.9 3163 55.6 5.7 .05 .03
diseased for more
than 1 year 1465 47.4 2748 37.6 37.3 .001 .09
* by age, age of onset of smoking, cigarettes per day and disease status.
Chi2 Likelihood chi2 test result.
P Significance level: .001 ≤ .001, .01 ≤ .01, .05 ≤ .05, ns not significant.
w Effect size measure Cohen's w [15].
The mean age at onset of smoking among female ever smokers aged 40 to 59 was higher than among male ever smokers aged 40 to 59 with a small and among those aged 60 or older with a medium effect (Table 2). At the age of 50 or older women disclosed a higher proportion of subjects who smoked 5 – 20 cpd than men with an effect.
Table 2 Mean age at onset of smoking and cigarettes per day by age*
Women Men T Chi2 P Effect size
n ever smokers mean
% current
smokers
within ever
smokers n ever smokers mean
% current
smokers
within ever
smokers mean difference (standard error)
Total
Age
10 – 19
mean age at onset (SD) cigarettes per day 1303 15.2 (1.6) 1663 15.4 (1.6) 0.2 (0.1) -2.7 .01 0.10
4 or less % 430 89.1 496 91.1 1.1 ns
5 – 20 % 854 95.4 1174 97.8 8.7 .01 .07
21 or more % 55 87.3 71 97.2 4.7 .05 .19
20 – 29
mean age at onset (SD) cigarettes per day 4419 16.7 (2.6) 5367 16.8 (2.5) 0.1 (0.1) -1.7 ns 0.04
4 or less % 1104 69.3 796 75.9 10.1 .01 .07
5 – 20 % 3142 81.9 4029 87.7 46.6 .001 .08
21 or more % 356 83.1 741 88.5 5.9 .05 .07
30 – 39
mean age at onset (SD) cigarettes per day 7493 17.3 (3.4) 8976 17.0 (3.2) 0.2 (0.1) 4.1 .001 0.06
4 or less % 1639 58.3 996 62.9 5.3 .05 .04
5 – 20 % 5365 72.2 6420 75.5 16.3 .001 .04
21 or more % 782 79.8 1831 79.2 0.1 ns
40 – 49
mean age at onset (SD) cigarettes per day 6464 18.4 (4.5) 8769 17.5 (3.7) 1.0 (0.1) 13.9 .001 0.23
4 or less % 1248 51.3 836 48.8 1.2 ns
5 – 20 % 4646 68.0 5674 65.7 6.0 .05 .02
21 or more % 802 72.6 2235 67.8 6.3 .05 .04
50 – 59
mean age at onset (SD) cigarettes per day 4156 21.1 (6.6) 7454 18.5 (4.8) 2.7 (0.1) 23.0 .001 0.48
4 or less % 911 42.2 729 42.5 0.2 ns
5 – 20 % 2867 61.8 4736 52.1 69.5 .001 .10
21 or more % 543 61.9 1829 50.8 20.7 .001 .09
60 – 69
mean age at onset (SD) cigarettes per day 2618 24.3 (9.0) 6825 19.3 (6.1) 5.0 (0.2) 26.2 .001 0.68
4 or less % 702 36.5 803 35.0 0.4 ns
5 – 20 % 1795 55.0 4310 38.6 138.5 .001 .15
21 or more % 225 48.0 1382 31.6 22.2 .001 .12
70 – 79
mean age at onset (SD) cigarettes per day 1848 25.4 (10.2) 4060 20.0 (6.6) 5.5 (0.3) 21.1 .001 0.66
4 or less % 662 23.6 601 23.3 0.0 ns
5 – 20 % 1174 40.5 2515 22.6 123.2 .001 .18
21 or more % 107 33.6 564 15.6 17.1 .001 .17
80 or older
mean age at onset (SD) cigarettes per day 440 26.5 (11.5) 956 21.4 (7.5) 5.1 (0.6) 8.5 .001 0.55
4 or less % 198 24.2 174 20.1 0.9 ns
5 – 20 % 240 32.9 540 18.0 20.4 .001 .16
21 or more % 27 51.9 114 9.6 22.0 .001 .44
* Means for current smokers; % current smokers within ever smokers.
T T score for t-test of means.
Chi2 Likelihood chi2 test result for proportions.
P Significance level: .001 ≤ .001, .01 ≤ .01, .05 ≤ .05, ns not significant.
Effect size Cohen's w for proportions and d for differences between means [15].
SD Standard deviation.
The proportion of current smokers among the ever smokers decreased by age (women: Chi2 2463.6; p < .001; w .28; men: Chi2 9103.7; p < .001; w .42). A lower age of onset of smoking was associated with a lower proportion of current smokers among ever smokers (women: Chi2 297.2; p < .001; w .10; men: Chi2 379.5; p < .001; w .09). Smokers who had been ill during the last four weeks prior to the interview showed a lower proportion of current smokers than those who said that they had not been ill. Those with a disease of 1 year or longer showed a lower proportion of current smokers than those who had been ill for a shorter time (women: Chi2 212.4; p < .001; w .08; men: Chi2 561.1; p < .001; w .11).
Prediction of current smoking
According to the multivariate data analysis, there was an interaction effect of gender and age for the odds of being a current smoker (Table 3). Women who were 80 to 89 years old had an odds ratio of 4.3 (CI 2.4 – 7.6) to be current smokers compared to men at this age. No interaction was found for the age of onset of smoking, for cigarettes per day or disease with gender.
Table 3 Prediction of current smoking; logistic regression analysis
Current smoker
Odds ratio Confidence interval*
Age
10 – 19 ref.
20 – 29 0.3 0.25 – 0.40
30 – 39 0.2 0.140 – 0.225
40 – 49 0.1 0.093 – 0.150
50 – 59 0.07 0.055 – 0.088
60 – 69 0.04 0.031 – 0.050
70 – 79 0.02 0.016 – 0.026
80 – 89 0.015 0.011 – 0.020
90 or older 0.013 0.006 – 0.026
Gender
Male ref.
Female 1.2 0.96 – 1.4
Age by gender
10 – 19, female ref.
20 – 29, female 1.3 0.8 – 2.1
30 – 39, female 1.5 0.9 – 2.4
40 – 49, female 1.7 1.1 – 2.8
50 – 59, female 2.2 1.4 – 3.5
60 – 69, female 2.8 1.7 – 4.5
70 – 79, female 3.0 1.9 – 5.0
80 – 89, female 4.3 2.4 – 7.6
90 or older, female 6.9 1.7 – 28.3
Age of onset of smoking
10 – 15 ref.
16 1.0 0.92 – 1.04
17 – 18 1.0 0.98 – 1.1
19 – 20 1.2 1.16 – 1.3
21 – 25 1.4 1.3 – 1.5
26 – 70 2.0 1.8 – 2.2
Age of onset of smoking by gender
10 – 15, female ref.
16, female 1.0 0.9 – 1.1
17 – 18, female 1.1 0.96 – 1.2
19 – 20, female 1.2 1.04 – 1.4
21 – 25, female 0.9 0.8 – 1.1
26 – 70, female 1.1 0.9 – 1.3
Cigarettes per day
less than 5 0.54 0.52 – 0.57
5 – 20 ref.
21 – 40 1.2 1.1 – 1.3
41 or more 0.4 0.38 – 0.5
Cigarettes per day by gender
less than 5, female 0.7 0.6 – 0.8
5 – 20, female ref.
21 – 40, female 1.1 0.97 – 1.2
41 or more, female 1.1 0.8 – 1.5
Disease (last 4 weeks)
not diseased ref.
diseased, 1 year or less 0.8 0.78 – 0.9
diseased, more than 1 year 0.8 0.7 – 0.9
Disease by gender
not diseased, female ref.
diseased, 1 year or less, female 1.0 0.9 – 1.1
diseased, more than 1 year, female 1.0 0.8 – 1.1
N = 58,984 ever smokers.
Logistic regression analysis, odds ratios adjusted for school education (years) and household income per household member. All variables in the table were entered into the model in addition to school education and household income per household member.
* 95 %.
ref. Reference category.
Discussion
The study reveals two main findings: First, there is a strong gender-age interaction indicating that surviving female ever smokers have higher odds of maintaining smoking at the age of 40 or above than surviving male ever smokers. This difference tends to increase until the age of 90 or above. Our data confirm former results from different countries that suggest higher proportions of current smokers among female than among male ever smokers [3,4,6]. Second, the higher proportion of current smokers among female than among male ever smokers cannot be explained by either a later age of onset of smoking or less cigarettes per day among female smokers according to the multivariate data analysis.
One reason for the female-male difference in the proportion of current smokers among ever smokers may be a selective death rate attributable to smoking or smoking and alcohol risk drinking: Evidence about tobacco-attributable mortality suggests that more male than female current smokers die from tobacco-attributable disease and from tobacco- and alcohol-attributable disease at the age before 65 [11]. However, we do not have data from longitudinal studies including data collection at death about smoking status. Furthermore, evidence shows higher relative risks of obstructive pulmonary disease and vascular disease associated with smoking for women than for men [17], and data revealed that there is myocardial infarction after a lower lifetime dosage of tobacco smoking among female than among male smokers [18]. Altogether, evidence according to selective death rates due to smoking is ambiguous.
A second reason for the female-male difference in the proportion of current smokers among ever smokers may be an age-period-cohort effect showing increasing rates of smoking women since the 1950ies. No data are available about smoking status by age and gender from comparable surveys carried out before 1989. However, lung cancer death rates were considerably higher among men than among women. Among men aged 70 or older, 11,000 lung cancer death cases attributable to smoking occurred compared to 2,600 among women aged 70 or older according to an overview from the year 2000 [19]. It may be assumed that rates of current smokers increased in Germany as in other Western countries. In addition, it may have contributed to the male-female difference of current smokers that the German Nazi regime especially forced health behavior including being abstinent from smoking [20,21]. Therefore, particularly many of the women who had been teenagers or young adults during the Nazi regime may have remained never smokers. The increase of smoking might have been accompanied by less openness towards quitting among female compared to male smokers. Accordingly, female current smokers may have less fear of health disturbance from smoking. If in the general population, particularly among female current smokers, the belief exists that tobacco-attributable disease mainly affects male smokers, then this might lead to lack of concern about the susceptibility to health hazards from smoking among women. Furthermore, there could be biological factors or nicotine dependence that in female more than in male smokers might act as a barrier against quitting. Other potential confounders may include socioeconomic status, education, living in an urban environment, and tobacco advertising that especially addresses sexual attractiveness of young women. However, the odds ratios found in our analysis were adjusted for school education and income per household member.
Women may have been more likely than men to misclassify themselves as never smokers what might have artificially increased the proportion of current smokers among ever smokers [cf. [22]]. This effect might be stronger the longer the time between the smoking and the interview and the fewer cpd the individual has smoked which is more likely for women than for men. Furthermore, recall bias towards misclassification as a never smoker becomes more likely with older age, given the likelihood of age-dependent limitations of memory. On the other hand, the long-term memory is rather little impaired by age. Also, smoking as a young women must have been somewhat outstanding until the 1950ies what could have been easy to recall.
Our findings confirm former results that show a decrease of the proportion of current smoking by age [3-7]. Our data show that this is true even until the age of 80 or older. No potential influence from the perception of having been ill during the last four weeks on the maintenance of smoking could be found, even not from disease that had lasted for more than a year. There may be insufficient understanding of the relationships between single tobacco-attributable disease and its risk among smokers, or there might be psychological coping strategies active for not admitting smoking-related reasons for the disease.
Different kinds of bias may have been introduced by old people. First, it seems more likely to meet old people in their household than middle adult age people that are working. Second, a considerable part of the elderly lives in institutions that were included in the Mikrozensus. Third, there may be more recall bias due to age-related memory deficits. Fourth, women have a higher life expectancy than men. However, these assumptions of a bias are rather unlikely in light of the fact that our final sample was representative for the general population aged 10 or older in Germany with respect to gender and age.
A strength of the data is that they include considerable numbers of respondents in old age due to data gathering at home. On the other hand, there are several limitations to this study. First, it is cross-sectional only, and the data do not allow any conclusions about causal relationships. Second, the interview questions about smoking did not refer to clear time frames. Third, we could not determine the lifetime number of cigarettes smoked or pack-years among former smokers since the interview did not include questions about the date of stopping smoking and there were only 4 categories for the number of cigarettes smoked per day. Fourth, data were based on self-statements only, no validation of smoking was used. However, evidence shows that the proportion of smokers who deny or minimize smoking in survey studies may be negligible and does not significantly change the results [23]. Fifth, sample selection bias may exist due to a large number of missing data for the smoking status. Sixth, the sample is representative for only one country with very little activity in the prevention of tobacco-attributable death and disease at the time of the data collection.
Conclusion
Special emphasis should be given to current smokers among the female general population at the age of 40 or above. Female smokers might be confronted with more barriers against quitting than male smokers. Prevention activity that is proactively targeted at the female population and tailored to the needs of women at the age of 40 or above might help to stimulate smoking cessation. Proactive population-based intervention focusing on the health hazards of smoking for these women should be used in public health intervention.
Competing interests
The author(s) declare that they have no competing interests.
Authors' contributions
UJ carried out the data analysis and wrote major parts of the first draft of the paper. MH obtained the data, introduced the idea of the data analysis, analyzed parts of the data and contributed parts of the text to the paper. CM and AS assisted with the writing of the manuscript and in the interpretation of the results.
Pre-publication history
The pre-publication history for this paper can be accessed here:
Acknowledgements
This work was financially supported by grants from the Social Ministry of the State of Mecklenburg-West Pomerania in Germany (grant no. IX 311a 406.68.43.05) and from the German Ministry of Education and Research (grant no. 01EB0120). The data were provided by the German Federal Statistical Office.
==== Refs
Taylor DHJ Hasselblad V Henley SJ Thun MJ Sloan FA Benefits of smoking cessation for longevity Am J Public Health 2002 92 990 996 12036794
Burns DM Cigarette smoking among the elderly: disease consequences and the benefits of cessation Am J Health Promot 2000 14 357 361 11067570
Husten CG Shelton DM Chrismon JH Lin YC Mowery P Powell FA Cigarette smoking and smoking cessation among older adults: United States, 1965-94 Tob Control 1997 6 175 180 9396100
Kemm JR A birth cohort analysis of smoking by adults in Great Britain 1974-1998 J Public Health Med 2001 23 306 311 11873893 10.1093/pubmed/23.4.306
Ruchlin HS An analysis of smoking patterns among older adults Med Care 1999 37 615 619 10386573 10.1097/00005650-199906000-00010
Farchi G Brescianini S Maggi S Mariotti S Scafato E Di Carlo A Baldereschi M Inzitari D Risk factors and health determinants in older Italians Aging Clin Exp Res 2004 16 3 12 15132285
Sulander T Helakorpi S Rahkonen O Nissinen A Uutela A Smoking and alcohol consumption among the elderly: trends and associations, 1985-2001 Prev Med 2004 39 413 418 15226054 10.1016/j.ypmed.2004.02.049
Tafaro L Cicconetti P Tedeschi G Baratta A Ursino R Ettorre E Marigliano V Smoking and longevity: an incompatible binomial? Arch Gerontol Geriatr Suppl 2004 425 430 15207443 10.1016/j.archger.2004.04.054
Falba T Jofre-Bonet M Busch S Duchovny N Sindelar J Reduction of quantity smoked predicts future cessation among older smokers Addiction 2004 99 93 102 14678067 10.1111/j.1360-0443.2004.00574.x
Laaksonen M Uutela A Vartiainen E Jousilahti P Helakorpi S Puska P Development of smoking by birth cohort in the adult population in eastern Finland 1972-97 Tob Control 1999 8 161 168 10478400
John U Hanke M Tobacco- and alcohol-attributable mortality and years of potential life lost in Germany European Journal of Public Health 2003 13 275 277 14533733 10.1093/eurpub/13.3.275
Leim K Christians H Mikrozensus 1999. Dokumentation und Datenaufbereitung [Microcensus 1999. Documentation and data preparation]
Fagerstrom K Boyle P Kunze M Zatonski W The anti-smoking climate in EU countries and Poland Lung Cancer 2001 32 1 5 11282422 10.1016/S0169-5002(00)00203-8
John U Meyer C Rumpf HJ Hapke U Relation among stage of change, demographic characteristics, smoking history, and nicotine dependence in an adult German population Preventive Medicine 2003 37 368 374 14507495 10.1016/S0091-7435(03)00149-X
Cohen J Statistical power analysis 1988 Hillsdale, Lawrence Erlbaum
Federal Statistics Office Statistisches Jahrbuch 2001 für die Bundesrepublik Deutschland [Statistical Yearbook 2001 for the Federal Republic of Germany] 2001 Stuttgart, Metzler-Poeschel
Prescott E Osler M Andersen PK Hein HO Borch-Johnsen K Lange P Schnohr P Vestbo J Mortality in women and men in relation to smoking International Journal of Epidemiology 1998 27 27 232 9563690 10.1093/ije/27.1.27
Prescott E Hippe M Schnohr P Hein HO Vestbo J Smoking and risk of myocardial infarction in women and men: longitudinal population study British Medical Journal 1998 316 1043 11047 9552903
Corrao MA Guindon GE Sharma N Shokoohi DF Tobacco control country profiles 2000 Atlanta, GA, American Cancer Society
Davey Smith G Lifestyle, health, and health promotion in Nazi Germany Bmj 2004 329 1424 1425 15604167 10.1136/bmj.329.7480.1424
Heuer C Becker N Smoking prevalence and lung cancer mortality in Germany J Epidemiol Biostat 1999 4 45 52 10613716
van de Mheen PJ Gunning-Schepers LJ Reported prevalences of former smokers in survey data: the importance of differential mortality and misclassification Am J Epidemiol 1994 140 52 57 8017403
Vartiainen E Seppala T Lillsunde P Puska P Validation of self reported smoking by serum cotinine measurement in a community-based study Journal of Epidemiology and Community Health 2002 56 167 170 11854334 10.1136/jech.56.3.167
| 15935091 | PMC1175093 | CC BY | 2021-01-04 16:28:55 | no | BMC Public Health. 2005 Jun 3; 5:57 | utf-8 | BMC Public Health | 2,005 | 10.1186/1471-2458-5-57 | oa_comm |
==== Front
BMC Pulm MedBMC Pulmonary Medicine1471-2466BioMed Central London 1471-2466-5-81597812910.1186/1471-2466-5-8Research ArticleTh1/Th2 cytokine pattern in bronchoalveolar lavage fluid and induced sputum in pulmonary sarcoidosis Tsiligianni Ioanna [email protected] Katerina M [email protected] Despina [email protected] Nikolaos [email protected] George [email protected] Nikolaos M [email protected] Demosthenes [email protected] Department of Pneumonology, University of Crete, Heraklion, Greece2 Department of Hematology, University of Thessaly, Larissa, Greece3 Department of Pneumonology, Democritus University of Thrace, Alexandroupolis, Greece2005 24 6 2005 5 8 8 19 5 2005 24 6 2005 Copyright © 2005 Tsiligianni et al; licensee BioMed Central Ltd.2005Tsiligianni 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
Sarcoidosis is thought to be a T-helper type 1 cytokine (Th2 cytokine) mediated disorder. Induced sputum (IS) has been proposed as a useful non-invasive method, mainly for the assessment of the airway diseases. The aim of this study was to explore induced sputum (IS) CD4+Th1 T-lymphocyte subpopulation and to compare them with those of bronchoalveolar lavage fluid (BALF) in patients with sarcoidosis.
Methods
We studied prospectively 21 patients (12 female, 9 male) of median age 46 yr (range, 25–65) with sarcoidosis and 10 normal subjects (5 female, 5 male) of median age 39 yr (range, 26–60). IS was performed with hypertonic saline solution using an ultrasonic nebulizer. BALF was performed within 10 days of IS. After stimulation of sputum lymphocytes with phorbol-myristate-acetate, we used double immunocytochemical methods to identify CD4+ IFN-γ positive and IL-4 positive cells (Th1 and Th2, respectively).
Results
Sarcoidosis patients had an increased number of CD4+ -IFN-γ producing cells in IS (p = 0.003) and BALF (p = 0.01) in comparison with normal subjects. No significant differences were detected between CD4+ -IL-4 cells in BALF (p = 0.053, NS) and IS (p = 0.46, NS) between sarcoidosis patients and healthy controls. The ratio of Th1 to Th2 cells in BALF and IS was statistically different in sarcoidosis when compared with normal subjects (p = 0.007 in BALF and IS). A significant correlation was found between CD4+ IFN-γ positive cells in IS and those in BALF in sarcoidosis patients (r = 0.685, p = 0.0006).
Conclusion
These data suggests that a Th1-like cytokine pattern can be observed in CD4+ T-lymphocytes in IS in patients with pulmonary sarcoidosis. Further studies are needed to explore the value of IS vs BALF in the follow-up of these patients.
==== Body
Background
Sarcoidosis is a chronic, systemic disease of unknown origin, characterized by the formation of noncaseating granulomas in affected organs, most commonly the lung. According to the ATS/ERS/WASOG [1] statement on sarcoidosis, the diagnosis is based on clinical and radiological findings and on histologic evidence. The technical availability of the fiberoptic bronchoscope has facilitated the development of the bronchoalveolar lavage procedure, which is useful for the collection of cells and secretions from the lower respiratory tract. Although this technique causes low risk to the patient, it is relatively invasive and inconvenient for the patients. Sputum induction has been proposed as a non-invasive method, useful to the investigation of asthma [3], and also of other interstitial lung disease (ILD) [4,5], more specifically sarcoidosis [4,6-9], pneumoconiosis due to dust exposure [10], Crohn's disease [11] and non-granulomatous ILD [4]. The exact role of induced sputum in the diagnosis and evaluation of disease activity in ILD and especially in sarcoidosis has not been clearly defined. Furthermore, no specific studies have been undertaken to evaluate the safety and functional effects of induced sputum in patients with ILD [5]. Today's attitude and opinion of researchers in this area is that induced sputum is a promising technique in assessing ILD, but its diagnostic role has not yet been clarified, and should be used as a complimentary tool to BAL for research and clinical monitoring [5,12].
The pathogenesis of pulmonary sarcoidosis has been related to an increased production of Th1-like cytokines. However cytokine expression in sarcoidosis has not been systematically studied yet. It is well known that CD4+ T cells can be divided in two subgroups on the basis of their cytokines profile. Th1 cells produce type 1 cytokines {IL-2, tumor necrosis factor-(TNF)-α, interferon gamma-(IFN-γ)}, while Th2 cells produce type 2 cytokines (IL-4, IL-5, IL-6, IL-10, IL13).
As far as we know there are only a few studies regarding the Th1/Th2 balance in sarcoidosis and all of them either in bronchoalveolar lavage fluid (BALF) or in blood [13-15]. Therefore, the aim of this study was to explore induced sputum (IS) CD4+ Th1 T-lymphocyte subpopulation and to compare them with those in the BALF in patients with sarcoidosis.
Methods
Subjects
The demographic and clinical characteristics of the patients and controls are shown in Table 1. Twenty-one consecutive sarcoidosis patients, 9 male and 12 female, of median (range) age 46 (25–65) years who were investigated in the sarcoidosis clinic of our Hospital were enrolled in the study. Three of them were smokers and 18 non-smokers. The ATS/ERS/WASOG statement [1] on sarcoidosis was followed for the diagnosis, based on history, clinical symptoms, standard chest radiography, computed tomography, lung Ga67 scintigraphy and laboratory tests (serum angiotensin-converting enzyme). All of them had transbronchial or open lung biopsy with histopathological evidence of noncaseating epithelioid cell granulomas.
Table 1 Demographic and spirometric characteristics of sarcoidosis patients and control subjects. Values are expressed as mean ± SD, and age as median (range).
Characteristics Sarcoidosis Normal subjects
Number 21 10
Sex: Male/Female 12/9 5/5
Age, yr 46 (25–65) 39 (26–60).
Smokers/non smokers 18/3 0/10
FVC, % pred 93.1 ± 6.4 103 ± 13.7
FEV1, % pred 92.3 ± 6.6 101.1 ± 19
FEV1 / FVC, % pred 95.3 ± 5.9 100.3 ± 8.6
K CO % pred 82.3 ± 8.1 96 ± 6.3
Radiographic Stage (n) I: 8, II: 7,III: 6
Abbreviations: KCO = Carbon monoxide transfer coefficient, FVC % pred : Forced vital capacity % predicted, FEV1 % pred : Forced expiratory volume within the first second % predicted
According to chest radiography classification of sarcoidosis, 8 had type I disease (lymphadenopathy alone, 7 type II disease (lymphadenopathy and parenchymal opacities), and 6 type III disease (only parenchymal opacities).
The control group included 10 healthy nonsmoker volunteers (5 female, 5 male), median (range) age 39 (26 – 60) years who were able to produce adequate sputum samples following sputum induction. Patients and controls with acute respiratory infection during the 6 weeks prior to the study were excluded. The Ethics committee of our hospital approved the protocol and all patients and controls gave their consent.
Spirometry
Spirometry was performed with a computerized system (MasterLab, Jaeger2.12, Germany). The measurement was performed using standard protocols according to ATS guidelines [16].
Sputum induction
Sputum was induced by the inhalation of hypertonic saline aerosol solution, generated by an ultrasonic nebulizer (Ultraneb 2000, DeVilbiss, Somerset, PA, USA), as previously described [6,17,18].
Sputum processing
Sputum was processed within 1 hour after termination of the induction. The method of sputum examination described earlier [17,18] was used with some modifications [6].
Bronchoalveolar lavage
Fiberoptic bronchoscopy with BAL was performed within two weeks from the IS, as part of routine clinical management, according to recommended guidelines and previous reports [6,20].
BALF processing
The recovered BAL fluid was filtered through sterile gauze (Thompson, Ontario, Canada) and centrifuged at 400 g for 15 minutes at 4°C. Total cell counts were determined using an improved Neubauer counting chamber, and expressed as the total number of cells per mL of aspirated fluid. The pellet was washed three times with cold PBS-Dulbecco's and the cells were adjusted to a final concentration of 106cells/mL with RPMI1640 plus 2%FCS. The slide preparation was performed as previously reported [6].
Immunocytochemical analysis
The immunocytochemical analysis and T-cell determination were performed in sputum cytospins. Briefly, sputum lymphocytes were stimulated by incubating suspension B in 24-well plates at a concentration of 2 × 103 cells/μL for 5 h, under 5% CO2, at 37°C, in RPMI-1640 at 10% FCS in the presence of phorbol 12-myristate 13 acetate (PMA) 25 ng/mL, ionomycin 1 μmoL and Brefeldin A 10 μg/Ml (Sigma-AldrichCorp. St.Louis, MO, USA). Cytospins were made using cytocentrifugation of 50 μL of the stimulated suspension and were stored at -80°C for immunocytochemical analysis later. Approximately 175,000 cells were cytospined on each slide (3500 cells/μl × 50 μl = 175,000 cells), among which there was a sufficient number of lymphocytes to stain.
After defrosting the slides, they were fixed in acetone for 10 min and rehydrated in Tris-Buffered Saline (TBS). The double immunocytochemical method was performed in two steps as previously described [21]. Briefly, the specimens were first incubated for 30 min with bovine serum to block the unspecific binding and then were exposed to the first primary antibody at a dilution of 1:50 for 30 min at room temperature. After washing inTBS three times they were exposed to the first secondary antibody, rabbit anti mouse immunoglobulin fluorescein isothiocyanate (FITC)-conjugated for 15 min (ImmunotechMarseille, France) at a dilution of 1:50. After washing in TBS three times they were exposed to the secondary primary antibody at a dilution of 1:50 for 30 min at room temperature. After washing in TBS three times they were exposed to the second secondary antibody, rabbit anti-mouse immunoglobulin phycoerythrin-conjugated (IgG-PE) for 15 min (Immunotech Marseille, France) at a dilution of 1:50. After washing in TBS three times the slides were mounted with 30% glycerol in TBS. Two investigators examined the slides under ultraviolet microscope and their results were averaged. Three replicate measurements were performed by each observer in 10 slides. Both intra- and inter- observer coefficient of variation were <15%. The ratio of CD4+-IFN-γ positive cells to CD4+-IL4 positive cells was calculated. For the estimation of each ratio, 500 T-cells were counted with more than one cytospin stained if necessary, because this number was sufficient to obtain a mean value per subject that remained constant after further increasing the number of cells counted. Results were also expressed as percentage of lymphocytes by dividing the number of lymphocytes stained per slide by total the number of lymphocytes per slide. Negative controls were obtained by the use of mouse-anti-mouse immunoglobulin to estimate the non specific binding. Positive controls were obtained by the use of T-lymphoma cells cytospined on slides. Macrophages were excluded from counting by morphology.
Measurement of CD4+-IFN-γ and CD4+-IL4 producing T-cells
The primary anti CD4 mouse anti-human monoclonal antibody with secondary rabbit anti-mouse IgG- FITC antibody and the primary anti-IFN-γ mouse antihuman monoclonal antibody (Caltag Burlingame, CA, USA) with secondary rabbit antimouse IgG-phycoerythrin-conjugated (IgG-PE) antibody were used. At least 500 CD4+ cells were counted to estimate the number of CD4+-IFN-γ cells. Also the same method was applied for staining CD4+-IL-4 producing cells with anti-IL4 antibody (Caltag Burlingame, CA, USA).
Statistics
All analyses were performed using the statistical software StatsDirect for Windows version2.4.1 (Camcode; Cambridge, UK). Results are expressed as mean ± SD, or median (range), unless otherwise indicated. The Shapiro-Wilk W test for normality was applied to assess normality. Differences between sarcoidosis patients and controls were tested usingt-test for normally and the Mann-Whitney U test for non normally distributed data. Differences between the BALF and sputumwithingroupswere tested using the paired student's t-test for normally and theWilcoxon's signed rank test for non-normally distributed variables. Correlation between cell numbers and T lymphocyte subsets were analyzed using the Pearson's correlation coefficient. A p value of <0.05 was considered as statistically significant
Results
Patients with sarcoidosis were older than normal control subjects and had lower FVC, and FEV1 values but the differences were not statistically significant. Induced sputum and BALF were tolerated well by all subjects, without any adverse events (Table 1).
In BALF the percentage of IFN-γ producing CD4+ Tcells was significantly higher in patients with sarcoidosis than in normal control subjects (mean ± SD, 53.1 ± 17.9 versus 38.1 ± + 7.7, p = 0.019) after the stimulation with PMA/ionomycin. In induced sputum the percentage of IFN-γ producing CD4+ T cells was significantly higher in patients with sarcoidosis than in normal control subjects (mean ± SD, 60.0 ± 23.3 versus 35.60 ± 12.46, p = 0.003) after the stimulation. No significant differences were detected between sarcoidosis patients and control subjects regarding the proportion of IL-4 producing CD4+ T cells in BALF (p = 0.46) as well as in induced sputum (p = 0.055) (Table 2). The ratio Th1/Th2 was found significantly higher in BALF (p = 0.007) and also in induced sputum compared with healthy controls.
Table 2 Mean ± SD of Th1, Th2 lymphocyte subsets in induced sputum (IS) and bronchoalveolar lavage fluid (BALF) in sarcoidosis patients and control subjects.
Induced sputum BALF
Th1 Th2 Th1 Th2
Control 35.6 ± 12.4 0.2 ± 0.15 38.1 ± 7.7 0.3 ± 0.14
Sarcoidosis patients 60.0 ± 23.3 0.2 ± 0.12 53.1 ± 17.9 0.2 ± 0.11
P 0.003 0.46 0.019 0.55
* Paired t-test or Wilcoxon's signed rank test, unadjusted p-values.
No statistically significant difference was found between induced sputum and BALF within groups (paired t test). A significant correlation was found between BALF and IS in the percentage of IFN-γ producing CD4+ T cells counts (r= 0. 685, p = 0.006) (Figure 1).
Figure 1 Correlation of CD4+ IFN-γ cells between bronchoalveolar lavage fluid (BALF) and induced sputum (IS).
Discussion
To the best of our knowledge this is the first study in the literature to investigate the balance Th1/Th2 in sarcoidosis with the method of induced sputum. The main finding of this study is that a Th1-like cytokine pattern can be observed in CD4+ BALF as well as induced sputum T lymphocytes. We also detected a positive correlation between CD4+ IFN-γ producing T cells in induced sputum and those in BALF in sarcoidosis patients (p = 0.0006, r = 0.685).
We have previously shown that inflammation in sarcoidosis could be effectively and non-invasively determined by the analysis of cell differential counts and T-lymphocyte subsets in induced sputum [6]. We demonstrated that induced sputum can discriminate between patients with sarcoidosis and normal subsets on the basis of the profile of airway inflammatory cells, including lymphocytes [6]. Based on our findings that there is a positive correlation between BALF and induced sputum lymphocyte counts (r = 0.616) and CD4+/CD8+ ratio (r = 0.7) we aimed to evaluate the Th1/Th2 balance in sarcoidosis in both BALF and induced sputum. As it has already been reported, the mean CD4+/CD8+ ratio was not significantly different between BALF and induced sputum in 20 sarcoidosis patients (p > 0.253) and in normal controls (p > 0.3). In our study population a CD4+/CD8+ ratio ≥ 2.5 was found on 17 patients in induced sputum, while a CD4+/CD8+ ratio > 4 was found in 12 (60%) patients [6].
The balance between Th1 and Th2 helper cells and their associated cytokine patterns can in an immune response influence the phenotype and progression of several clinical diseases. An imbalance in the expression of Th1 and Th2 type cytokines by alveolar cells is thought to play an important role in the immunopathogenesis of sarcoidosis [22-26]. Previous authors have shown a spontaneous release from alveolar cells of the Th1 type cytokines IFN-γ and IL-2 but not Th2 type cytokines [24,26-29]. Additionally, there are no data on cytokine profiles in the late stages of fibrotic sarcoidosis to assess the contribution of Th1 or Th2 cytokines to the fibrotic process. Experimental models confirm that Th2-mediated granulomatous responses are more fibrotic than Th1-mediated inflammation, so in the absence of human data, there is uncertainty as to the relevant immune processes in fibrotic pulmonary sarcoidosis. Conceivably, patients with fibrotic sarcoidosis are those who switch to a more fibrotic Th2 response later in the course of the disease, or have a persistent dominant Th2 response from the iniatial stages of disease [29,30]. Furthermore, data support the immunopathogenetic concept of compartmentalization and the predominance of CD4+ T cells producing mainly Th1 type cytokines in acute pulmonary sarcoidosis, which becomes less prominent during the course of the disease [15]. However, in a study using T cell clones from BALF, Baumer et al. found equal levels of Th1 and Th2 cytokine gene expression in patients with sarcoidosis [31]. Thus, the purpose of the present study was to further investigate the phenotype of T lymphocytes in BALF and in induced sputum in patients with pulmonary sarcoidosis with focus on expression of markers indicating Th1 and Th2 cellular function. Our findings suggest a shift toward a type-1 phenotype in CD4+ T-cell populations in accordance with previous studies in BALF and in peripheral blood [13-15,32,33]. Intracellular cytokine staining could be useful to study cytokine expression in sarcoidosis [6,14]. These data are in agreement with those investigating the cytokine profile of T cell subsets with other methods like flow cytometry [13,32,33]. Inui et al. [13] investigated the alteration of the Th1/Th2 and Tc1/ Tc2 balance in sarcoidosis, using cytokine flow cytometry. The authors showed for the first time a significantly higher percentage of IFN-γ producing CD4+ T cells together with a markedly increased ratio of IFN-γ/IL-4-producing CD4+ T cells in BALF, after PMA/ ionomycin stimulation in patients with sarcoidosis compared with normal subjects. On the other hand, there were no differences in the percentages of IFN-γ or IL-4- producing CD8+ T cells in either the peripheral blood or BALF between patients and controls. In addition, a recent report in eighteen patients with untreated sarcoidoisis, examined the expression of Th1 associated chemokine and cytokine receptors CXCR3, CCR5, and interleukin IL-12R, IL-18R, respectively, as well as of the Th2 associated chemokines receptors CCR4 and CXCR4 on CD4+ and CD8+ T cells [33]. That report showed the lung accumulation of the above Th1 associated receptors, in agreement with recent data of our group [34,35].
Conclusion
This report is the first to evaluate the critical balance of Th1 to Th2 cells with the method of Induced sputum. In conclusion, these data demonstrate that there are increased numbers of CD4+ IFN-γ producing T cells in induced sputum from patients with pulmonary sarcoidosis, in accordance with the shift towards the Th1 response known to exist in BALF and peripheral blood. The positive correlation between the CD4+ IFN-γ producing T cells in BALF and in induced sputum, suggest that sarcoidosis patients could be followed up with this noninvasive method which has to be investigated in further studies.
Abbreviations
BALF: Bronchoalveolar Lavage Fluid, FEV1 % pred: Forced expiratory volume within the first second % predicted, FVC % pred: Forced vital capacity % predicted, ILD: Interstitial Lung Disease, KCO= Carbon monoxide transfer coefficient, TNF: Tumor Necrosis Factor, IFN-γ: Interferon gamma, IS: Induced Sputum, TBS: Tris-Buffered Saline.
Competing interests
This study supported by a grant from the Society for Respiratory Research at the University of Thrace
Authors' contributions
IT, KMA and DB were involved with the study conception. IT, NT, GC and KMA carried out the data acquisition and interpretation. DK performed the immunocytochemical analysis. NT performed the statistical analysis. KMA and DB prepared the manuscript. DB and NMS were involved in revising the article 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 are grateful to Argyris Tzouvelekis (M.D) for formatting the article.
==== Refs
Hunninghake GW Costabel U Ando M Baughman R Cordier JF du Bois R Eklund A Kitaichi M Lynch J Rizzato G Rose C Selroos O Semenzato G Sharma OP ATS/ERS/WASOG statement on sarcoidosis. American Thoracic Society/European Respiratory Society/World Association of Sarcoidosis and Granulomatous Disorders Sarcoidosis Vasc Diffuse Lung Dis 1999 16 149 173 10560120
Costabel U Guzman J Bronchoalveolar lavage in interstitial lung disease Curr Opin Pulm Med 2001 7 1117 11622 10.1097/00063198-200109000-00002
Fahy JV Wong H Liu J Boushey HA Comparison of samples collected by sputum induction and bronchoscopy from asthmatic and healthy subjects Am J Respir Crit Care Med 1995 152 53 58 7599862
Fireman E Topilsky I Greif J Lerman y Schwarz Y Man A Topilsky M Induced sputum compared to bronchoalveolar lavage for evaluating patients with sarcoidosis and non-granulomatous interstitial lung disease Respir Med 1999 93 827 834 10603633 10.1016/S0954-6111(99)90269-X
Olivieri D D'Ippolito R Chetta A Induced sputum: diagnostic value in interstitial lung disease Curr Opin Pulm Med 2000 6 411 414 10958231 10.1097/00063198-200009000-00004
Tsiligianni I Tzanakis N Kyriakou D Chrysofakis G Siafakas N Bouros D Comparison of sputum induction with bronchoalveolar lavage cell differential counts in patients with sarcoidosis Sarcoidosis Vasc Diffuse Lung Dis 2002 19 205 210 12405490
D'Ippolito R Chetta A Olivieri D Role of induced sputum in interstitial lung disease Eur Respir J 2000 16 573 574 11028676 10.1034/j.1399-3003.2000.016003573.x
D'Ippolito R Foresi A Chetta A Casalini A Castagnaro A Leone C Olivieri D Induced sputum in patients with newly diagnosed sarcoidosis: comparison with bronchial wash and BAL Chest 1999 115 1611 1615 10378557 10.1378/chest.115.6.1611
Vassilakis DA Sourvinos G Pantelidis P Spandidos DA Siafakas NM Bouros D Extended genetic alterations in a patient with pulmonary sarcoidosis, a benign disease Sarcoidosis Vasc Diffuse Lung Dis 2001 18 307 310 11587105
Fireman E Greif J Schwarz Y Man A Ganor E Ribak Y Lernan Y Assessment of hazardous dust exposure by BAL and induced sputum Chest 1999 115 1720 1728 10378572 10.1378/chest.115.6.1720
Fireman Z Osipov A Kivity S Kopelman Y Sternberg A Lazarov E Fireman E The use of induced sputum in the assessment of pulmonary involvement in Crohn's disease Am J Gastrenterol 2000 95 730 734 10.1111/j.1572-0241.2000.01843.x
Fireman E Induced sputum: opening a new window to the lung Sarcoidosis Vasc Diffuse Lung Dis 2001 18 263 271 11587097
Inui N Chida K Suda T Nakamura H Th1/Th2 and Tc1/Tc2 profiles in peripheral blood and bronchoalveolar lavage fluid cells in pulmonary sarcoidosis J Allergy Clin Immunol 2001 107 337 344 11174202 10.1067/mai.2001.112273
Prasse A Georges CG Biller H Hamm H Mathys H Luttman W Virchow JC Th1 cytokine pattern in sarcoidosis is expressed by bronchoalveolar CD4 and CD8 T cells Clin Exp Immunol 2001 122 241 248 11091281 10.1046/j.1365-2249.2000.01365.x
Mollers M Aries SP Dromann D Mascher B Braun J Dalhoff K Intracellular cytokine repertoire in differentT cell subsets from patients with sarcoidosis Thorax 2001 56 187 193 10.1136/thorax.56.6.487
American Thoracic Society Standardization of spirometry, 1994 update Am J Respir Crit Care Med 1995 152 1107 1136 7663792
Popov T Pizzichini M Pizzichini E Kolendowicz R Punthakee Z Dolovich J Hargreave F Some technical factors influencing the induction of sputum for cell analysis Eur Respir J 1995 8 559 565 7664854
Popov T Gottschalk R Kolendowicz R Dolovich J Powers P Hargreave F The evaluation of a cell dispersion method of sputum examination Clin Exp Allergy 1994 24 778 783 7982128
Kips JC Fahy JV Hargraeve FE Ind PW Veen JC Methods for sputum induction and analysis of induced sputum: a method for assessing airway inflammation in asthma Eur Respir J 1998 26 9S 12S
Report of the European Society of Pneumology Task Group Technical recommendations and guidelines for bronchoalveolar lavage (BAL) Eur Respir J 1989 2 561 585 2663535
Kyriakou D Alexandrakis MG Kyriakou ES Liapi D Kourelis TV Mavromanolakis M Vlachonikolis I Eliakis P Aberrant expression of the major sialoglycoprotein (CD43) on the monocytes of patients with myelodysplastic syndromes Ann Hematol 2000 79 198 205 10834507 10.1007/s002770050579
Muller-Quernheim J Sarcoidosis: immunopathogenetic concepts and their clinical application Eur Respir J 1998 12 716 738 9762805 10.1183/09031936.98.12030716
Agostini C Costabel U Semenzato G Sarcoidosis news: immunologic frontiers for new immunosuppressive strategies Clin Immunol Immunopathol 1998 88 199 204 9714698 10.1006/clin.1998.4544
Walker C Bauer W Braun RK Menz G Braun P Schwarz F Hansel TT Villiger B Activated T cells and cytokines in bronchoalveolar lavages from patients with various lung diseases associated with eosinophilia Am J Respir Crit Care Med 1994 150 1038 1048 7921434
Drent M Grutters JC Mulder PG van Velzen-Blad H Wouters EF van den Bosch JM Is the different T helper cell activity in sarcoidosis and extrinsic allergic alveolitis also reflected by the cellular Bronchoalveolar lavage fluid profile? Sarcoidosis Vasc Diffuse Lung Dis 1997 14 31 38 9186987
Minshall EM Tsicopoulos A Yasruel Z Wallaert B Akoum H Vorng H Tonnel AB Hamid Q Cytokine mRNA gene expression in active and nonactive sarcoidosis Eur Respir J 1997 10 2034 2039 9311498 10.1183/09031936.97.10092034
Hoshino T Itoh K Gouhara R Yamada A Tanaka Y Ichikawa Y Azuma M Mochizuki M Oizumi K Spontaneous production of various cytokines except IL-4 from CD4+ T cells in the affected organs of sarcoidosis patients Clin Exp Immunol 1995 102 399 405 7586698
Moller DR Forman JD Liu MC Noble PW Greenlee BM Vyas P Holden DA Forrester JM Lazarus A Wysocka M Trinchieri G Karp C Enhanced expression of IL-12 associated with Th1 cytokine profiles in active pulmonary sarcoidosis J Immunol 1996 156 4952 4960 8648147
Moller DR Cells and cytokines involved in the pathogenesis of sarcoidosis Sarcoidosis Vasc Diffuse Lung Dis 1999 16 24 31 10207939
Moller DR Pulmonary fibrosis of sarcoidosis. New approaches, old ideas Am J Respir Cell Mol Biol 2003 29 S37 S41 14503552
Baumer I Zissel G Schlaak M Muller-Quernheim J Th1/Th2 cell distribution in pulmonary sarcoidosis Am J Respir Cell Mol Biol 1997 16 171 177 9032124
Wahlstrom J Katchar K Wigzell H Olerup O Eklund Grunewald J Analysis of intracellular cytokines in CD4+ and CD8+ lung and blood T cells in sarcoidosis Am J Respir Crit Care Med 2001 163 115 121 11208635
Katchar K Eklund A Grunewald J Expression of Th1 markers by lung accumulated T cells in pulmonary sarcoidosis Journal of InternalMedicine 2003 254 564 571 10.1111/j.1365-2796.2003.01230.x
Antoniou KM Alexandrakis M Sfiridaki K Tzanakis N Symvoulakis EK Tsiligianni I Bouros D Siafakas NM Comparison of sputum induction with Bronchoalveolar Lavage Fluid Cytokine IL-12 and IL-18 levels in patients with idiopathic pulmonary fibrosis (IPF/UIP) Am J Respir Crit Med 2004 169 sA298
Antoniou KM Tsiligianni I Alexandrakis Sfiridaki K Tzortzaki EG Tzanakis N Bouros D Siafakas NM Th1 cytokine profile pattern in peripheral blood, induced sputum and Bronchoalveolar lavage fluid in pulmonary sarcoidosis Am J Respir Crit Med 2004 169 sA549
| 15978129 | PMC1175094 | CC BY | 2021-01-04 16:30:12 | no | BMC Pulm Med. 2005 Jun 24; 5:8 | utf-8 | BMC Pulm Med | 2,005 | 10.1186/1471-2466-5-8 | oa_comm |
==== Front
BMC Womens HealthBMC Women's Health1472-6874BioMed Central London 1472-6874-5-81591889910.1186/1472-6874-5-8Research ArticleAberrations of TACC1 and TACC3 are associated with ovarian cancer Lauffart Brenda [email protected] Mary M [email protected] Roger [email protected] David [email protected] Richard A [email protected] Jennifer D [email protected] Ivan H [email protected] Department of Cancer Genetics, Roswell Park Cancer Institute, Elm and Carlton Streets, Buffalo, New York, 14263, USA2 Department of Pharmacology and Therapeutics, Roswell Park Cancer Institute, Elm and Carlton Streets, Buffalo, New York, 14263, USA3 Gilda Radner Familial Ovarian Cancer Registry, Roswell Park Cancer Institute, Elm and Carlton Streets, Buffalo, New York, 14263, USA2005 26 5 2005 5 8 8 25 2 2005 26 5 2005 Copyright © 2005 Lauffart et al; licensee BioMed Central Ltd.2005Lauffart 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
Dysregulation of the human Transforming Acidic Coiled Coil (TACC) genes is thought to be important in the development and progression of multiple myeloma, breast and gastric cancer. Recent, large-scale genomic analysis and Serial Analysis of Gene Expression data suggest that TACC1 and TACC3 may also be involved in the etiology of ovarian tumors from both familial and sporadic cases. Therefore, the aim of this study was to determine the occurrence of alterations of these TACCs in ovarian cancer.
Methods
Detection and scoring of TACC1 and TACC3 expression was performed by immunohistochemical analysis of the T-BO-1 tissue/tumor microarray slide from the Cooperative Human Tissue Network, Tissue Array Research Program (TARP) of the National Cancer Institute, National Institutes of Health, Bethesda, MD, USA. Tumors were categorized as either positive (greater than 10% of cells staining) or negative. Statistical analysis was performed using Fisher's exact test and p < 0.05 (single comparisons), and p < 0.02 (multiple comparisons) were considered to be significant. Transgenomics WAVE high performance liquid chromatography (dHPLC) was used to pre-screen the TACC3 gene in constitutional DNA from ovarian cancer patients and their unaffected relatives from 76 families from the Gilda Radner Familial Ovarian Cancer Registry. All variant patterns were then sequenced.
Results
This study demonstrated absence of at least one or both TACC proteins in 78.5% (51/65) of ovarian tumors tested, with TACC3 loss observed in 67.7% of tumors. The distribution pattern of expression of the two TACC proteins was different, with TACC3 loss being more common in serous papillary carcinoma compared with clear cell carcinomas, while TACC1 staining was less frequent in endometroid than in serous papillary tumor cores. In addition, we identified two constitutional mutations in the TACC3 gene in patients with ovarian cancer from the Gilda Radner Familial Ovarian Cancer Registry. These patients had previously tested negative for mutations in known ovarian cancer predisposing genes.
Conclusion
When combined, our data suggest that aberrations of TACC genes, and TACC3 in particular, underlie a significant proportion of ovarian cancers. Thus, TACC3 could be a hitherto unknown endogenous factor that contributes to ovarian tumorigenesis.
==== Body
Background
It is apparent that for a normal cell to develop into a highly delocalized metastatic cancer, multiple genetic events are required to overcome the normal mechanisms that control the growth and development of healthy tissue. About 10% of ovarian cancer patients inherit a familial predisposition, and of those cases, only 35–50% can be attributed to the inheritance of defects in the BRCA1 and BRCA2 tumor suppressor genes [1,2]. In addition, BRCA1 and BRCA2 mutations are not directly involved in the initiation events leading to the development of sporadic tumors, indicating that additional, as yet unidentified genes must play a significant role in the etiology of both familial and sporadic ovarian cancer. In ovarian cancer, comparative genomic hybridization (CGH), multicolor spectral karyotyping (SKY), and loss of heterozygosity (LOH) studies have identified several regions of the genome that may contain novel genes involved in the development and progression of ovarian cancer [3-5]. These techniques have indicated that deletions or rearrangements of 4p16 and 8p11, the loci for TACC3 and TACC1 respectively, commonly occur in 40% of ovarian cancer cell lines and primary tumors from both familial and sporadic cases [3-5]. SAGE (Serial Analysis of Gene expression) analysis further suggests that TACC3 and TACC1 are downregulated in ovarian tumors and ovarian cancer cell lines [6]. Thus, based upon both the location of TACC1 and TACC3 in regions consistently associated with ovarian cancer [3,5] and SAGE expression data [6], we have set out to determine the occurrence of alterations of these TACCs in ovarian cancer.
Methods
Serial Analysis of Gene Expression (SAGE)
The results of SAGE analysis of libraries generated by the method of Velculescu et al [7] were downloaded from the SAGEMAP section of the Gene Expression Omnibus website at the National Center for Biotechnology Information [6], and critically assessed for reliability to specifically predict expression of TACC3 and TACC1 in ovarian tissue and tumors. SAGE profiles used for TACC3 and TACC1 were [8] and [9], respectively.
Tissue and tumor microarrays
T-BO-1 and IMH-343 tissue and tumor microarray slides were obtained from the Cooperative Human Tissue Network, Tissue Array Research Program (TARP) of the National Cancer Institute, National Institutes of Health (Bethesda, MD, USA) and Imgenex Corporation (San Diego, CA, USA), respectively. Each of these microarrays contain normal tissues that are known to express TACC1 and TACC3 [10-12]. Clear cell carcinomas on the tissue/tumor slides, which cannot be graded using the World Health Organisation or FIGO systems [13], were classified as grade 3, as recommended by the NCCN practice guidelines [14].
Immunohistochemical staining
Immunohistochemical procedures were first optimized using formalin-fixed, paraffin-embedded MCF7 and HT-29 cell lines prepared in our laboratory. Conditions were then further optimized using mixed normal/tumor human tissue microarrays (Imgenex Corporation).
Tumor microarrays were deparaffinized in three changes of xylene and rehydrated using graded alcohols at room temperature. Endogenous peroxidase was quenched with aqueous 3% H2O2 for 20 min. and washed with PBS-Tween20 (PBS/T). Antigen retrieval was performed with citrate buffer pH 6.0 by heating twice in a microwave for 10 min., and then cooling to room temperature for 15 min. Following a PBS/T wash, the slides were placed in a humidity chamber and 0.03% w/v casein in PBS/T was applied to the tissue sections for 30 min. to block non-specific binding. This was then replaced with the primary anti-TACC3 antibody (#sc5885, Santa Cruz Biotech., Santa Cruz, CA, USA) at 1 μg/ml and left overnight at 4°C. As a negative control, a duplicate slide was incubated with antibody that had been precompeted with 10-fold excess of TACC3 peptide competitor [Santa Cruz #sc5885P] for 2 hr. at room temperature prior to use. The slides were then washed with PBS/T, followed by the biotinylated secondary antibody [Santa Cruz #sc2042 Donkey anti-goat] for 30 min. A PBS/T wash was followed by incubation with streptavidin-peroxidase reagent [#50-242, Zymed, Carlsbad, CA, USA] for 30 min. PBS/T was used as a wash and then chromogen DAB [#K3466, DAKO, Carpinteria, CA, USA] was applied for 5 min. (color reaction product – brown). The slides were then counterstained with Hematoxylin (blue in figures), dehydrated, cleared and finally coverslipped.
Antigen retrieval for TACC1 was essentially as described above. Slides were then incubated overnight at 4°C with primary anti-TACC1 antibody (#07-229, Upstate, Lake Placid, NY, USA) at 1 μg/ml. Concentration matched rabbit IgG (1 μg/ml) was used on a duplicate slide, instead of the primary TACC1 antibody, as a negative control for immunostaining. The slides were then placed on a DAKO autostainer with the following program; 1) one PBS/T wash; 2) incubation with a biotinylated secondary antibody [Vector Elite kit] for 30 min.; 3) one PBS/T wash; and then 4) incubation with ABC reagent [Vector Elite kit] for 30 min. PBS/T was used as a final wash and chromogen-DAB [DAKO] was applied for 5 min. (color reaction product – brown). The slides were then counterstained with Hematoxylin and dehydrated, cleared and coverslipped.
Image analysis
Tumor microarrays were examined by light microscopy and cores were categorized into either positive (+) or negative (-) staining for TACC3 and TACC1, with (+) staining representing those cores where greater than 10% of the cells had detectable staining, otherwise they were scored (-). No staining was observed in the negative controls, which were serial tumor microarray slides that were incubated with either peptide block or IgG depending upon the test antibody. Positive controls were normal tissues on each tissue and tumor array, and normal ovarian surface epithelium. Images were captured using a Nikon eclipse E600 microscope with Spot insight digital camera and Spot Advanced software version 4.0.1 (Diagnostic Instruments, MI, USA).
Statistical analysis
Fisher's exact test was performed using GraphPad Prism Version 3.03, (GraphPad Software, San Diego, California, USA, ). All tests were two-sided and p < 0.05 and p < 0.02 were considered to be significant for single and multiple comparisons respectively.
Mutation analysis using Transgenomic WAVE dHPLC
The TACC3 gene was screened for mutations using the Transgenomic WAVE dHPLC system [15]. The sequence of each fragment to be amplified by the polymerase chain reaction (PCR) was first analyzed using the WAVE melting profile software program to determine gradient and column temperature conditions for each PCR product. Individual exons and intronic regions encompassing the 5' and 3' splice sites were then amplified from a standard commercial normal human placental DNA sample (conditions and primer sequences available on request). These PCR products were then used to establish the actual melting profiles for each exon. Test samples were mixed with the standard normal control PCR product at a 1:1 ratio and denatured at 95°C for 5 min., followed by slow cooling to room temperature at the rate of 0.1°C every four seconds to enhance heteroduplex formation. WAVE dHPLC analysis was then performed and patterns compared with matched normal sibling controls (non-symptomatic) and the unmixed normal control pattern. For all variant patterns detected, aliquots from the original PCR reaction were then directly sequenced by the Roswell Park Cancer Institute Biopolymer core facility. Variant sequences were cross-referenced against GenBANK and Single Nucleotide Polymorphism databases [16], to test for occurrence in a wider population sample. New gene alterations that were found only in cDNAs from normal tissues in the databases, or found in normal and affected individuals in our test subjects were labelled as polymorphisms.
All investigations were performed after approval by the Roswell Park Cancer Institute institutional review board.
Results
Expression analysis of TACC1 and TACC3 in ovarian tumors
TACC1 and TACC3 are expressed in the epithelia of a number of different tissues, including the mammary gland and the ovary [17-19] (Fig. 1), suggesting that these proteins may be required for the normal maintenance of the epithelial component of organs within the body. Published SAGE (Serial Analysis of Gene Expression) analysis suggests that TACC1 expression is absent in 3 out of 7 bulk ovarian tumors and cell lines (Figure 3). Furthermore, TACC3 is downregulated in all 7 of the bulk ovarian tumors and tumor cell lines, compared with the normal human ovarian surface epithelium (HOSE) (Figure 3) [6]. Notably, similar to the two transciption factors GATA4 and GATA6 [20], TACC3 downregulation occurs in IOSE29, an immortalized ovarian surface epithelial cell line, suggesting that loss of TACC3 expression may be an early event during the immortalization process. To determine whether the pattern of gene expression noted by SAGE analysis is also evident in patient tumors, we assessed ovarian tumor microarrays from the National Cancer Institute multi-tumor tissue microarray (TARP) facility for expression of both proteins. 65 of the 75 OSE tumors on the tissue/tumor microarray could be evaluated for both TACC1 and TACC3 expression after the immunostaining procedure. Representative images demonstrating positive or negative tumor staining compared to normal human ovarian surface epithelium are shown in Fig. 1. A total of 21 (32.3%) and 41 (63.1%) showed positive staining for TACC3 and TACC1 respectively (Table 1). Thus, 44 (67.7%) and 24 (36.9%) could be categorized as absent or minimal expression for TACC3 and TACC1 respectively, similar to the results observed in the SAGE analysis. In addition, the allocation of TACC1 and TACC3 status to each core showed that 78.5% of the tumors exhibited negative immunoreactivity for one or both proteins. Within the ovarian tumor set tested (n = 65), there was a significant difference in the expression of TACC1 compared to TACC3 (Fisher's exact test, p = 0.0008) (Fig. 2). Expression of neither protein was associated with tumor grade. In all cases where TACC3 expression was observed, the protein was excluded from the nucleus (Fig. 1), contrary to the predominantly nuclear localization observed in the normal ovarian surface epithelium in vivo [17-19] (Fig 1). These results indicate that tumors derived from the ovarian surface epithelium exhibit low/negative expression or aberrant localization of TACC3, which may biologically contribute to development of these tumors.
Figure 1 Immunohistochemical analysis of tumor micoarrays. Representative normal human ovarian surface epithelium and tumor cores stained for TACC1 (panels A-E), and TACC3 (panels F-J) proteins to show positive and negative staining. TACC protein expression is detected as a brown signal against the blue Hematoxylin counterstain. In all cases where TACC3 expression was observed, the protein was excluded from the nucleus of the ovarian tumor cells, unlike the observable nuclear and cytoplasmic expression of TACC1. A, normal ovarian surface epithelium with nuclear/cytoplasmic TACC1 staining; B, serous papillary TACC1 +ve; C, serous papillary TACC1 -ve; D, endometroid TACC1 +ve; E, endometroid TACC1 -ve; F, normal ovarian surface epithelium with nuclear TACC3 staining; G, serous papillary TACC3 +ve; H, serous papillary TACC3 -ve; I, clear cell TACC3 +ve; J, clear cell TACC3 -ve. A-H: Main panel original magnification ×40; insets show whole tumor core at original magnification ×10.
Figure 2 Summary chart of the expression of TACC1 and TACC3 relative to ovarian tumor type. The difference between expression of TACC3 and TACC1 in all types of tumors, and the serous papillary subtype in particular, was significant (Fisher's exact test, p = 0.0008, and p < 0.0001 respectively).
Figure 3 Serial Analysis of Gene Expression for TACC1 and TACC3 in ovarian cancer.
Table 1 Distribution of TACC1 and TACC3 expression in ovarian tumors
Tumor type TACC3a TACC1a TACC3 & TACC1b
Tumor gradesc Tumor gradesc Tumor gradesc
I II III all I II III all all
serous papillary 1d (7)e 14.3%f 6 (18) 33.3% 2 (17) 11.8% 9 (42) 21.4% 6 (7) 85.7% 13 (18) 72.2% 11 (17) 64.7% 30 (42) 71.4% 6 (42) 14.3%
mucinous 1 (3) 33.3% 0 (1) 0% 1 (1) 100% 2 (5) 40% 1 (3) 33.3% 1 (1) 100% 1 (1) 100% 3 (5) 60% 2 (5) 40%
endometroid 3 (5) 60% 0 (1) 0% 0 (1) 0% 3 (7) 42.8% 2 (5) 40% 0 (1) 0% 0 (1) 0% 2 (7) 28.6% 1 (7) 14.3%
clear cellg 7(11) 63.6% 7 (11) 63.6% 6 (11) 54.5% 6 (11) 54.5% 5 (11) 45.5%
all types 5 (15) 33.3% 6 (20) 30.0% 10 (30) 33.3% 21(65) 32.3% 9 (15) 60.0% 14 (20) 70.0% 18 (30) 60.0% 41 (65) 63.1% 14 (65) 21.5%
aExpression irrespective of the status of the other TACC protein;
bTumor cores expressing both proteins;
cSixty five cores could be graded (1,2,3) and scored for TACC1 and TACC3 expression. Tumors with less than 10% of cells stained were classed as negative (-), greater than 10%, as positive (+). Percentage expression was calculated within each histological subtype.
dNumber of positive cores in a set;
eTotal number of cores in a set;
fPercentage of positive cores in a set;
gClear cell carcinomas cannot be graded using the World Health Organisation or FIGO systems [13], but are classified as grade 3, as recommended by the NCCN practice guidelines [14].
Next, we examined the association of expression of each TACC protein with specific types of tumor to test whether the distribution of each TACC varied between subtypes (Summarized in Table 1). TACC3 expression was observed in 21.4% (9 of 42) serous papillary, 40% (2 of 5) mucinous, 42.8% (3 of 7) endometroid, and 63.6% (7 of 11) clear cell ovarian tumor cores. A significant difference in TACC3 expression between tumor types was noted, with TACC3 being expressed less frequently in serous papillary tumors than clear cell tumors (Fisher's exact test, p = 0.0113). In contrast, TACC1 was detected in 71.4% (30 of 42) serous papillary, 60% (3 of 5) mucinous, 28.6% (2 of 7) endometroid and 54.5% (6 of 11) clear cell ovarian tumor cores. TACC1 expression was less frequent in endometroid than in serous papillary tumors (Fisher's exact test p = 0.0406). Noticeably, TACC3 compared with TACC1 staining in serous papillary tumors was significantly different (Fisher's exact test, p < 0.0001) (Fig. 2). There was no significant difference between their expression in clear cell, endometroid or mucinous tumor types. However, as it is more difficult to obtain large numbers of these less common types of ovarian cancer, it should be stressed that further analysis will be required to investigate whether relationships between tumor type and grade will hold in a larger sample set.
Mutation analysis of the TACC3 gene in familial ovarian cancer
The prevalence of structural aberrations in 4p16 noted during ovarian tumorigenesis indicates that this region could contain one or more genes whose normal role is to control the growth and maintenance of the ovarian surface epithelium [3-5]. Thus, loss of one copy of 4p16 in the developing ovarian tumor may result in haploinsufficiency, or unmask a pathogenic mutation in one or more of the genes in this region. With the evidence from the expression analysis presented above, it would appear that a significant number of ovarian cancers lack TACC3 expression. As the TACC3 gene is located within 4p16 [11], we next investigated whether constitutional mutations occur within the TACC3 gene in familial ovarian cancer. Thus, mutation analysis was performed on the TACC3 coding sequence, and the corresponding intron/exon boundaries of this 16 exon gene. Specifically, we used the Transgenomics WAVE dHPLC technology as a pre-screening approach to sample constitutional DNA from ovarian cancer patients and their unaffected relatives from 76 families from the Gilda Radner Familial Ovarian Cancer Registry. These families had previously tested negative for mutations in known ovarian cancer predisposition genes, including BRCA1 and BRCA2, and thus the underlying genetic predisposition in these families is currently unknown [2].
We identified novel sequence variants in a number of the samples, but most of these were due to polymorphisms in the intronic sequence included in the amplified products. However, novel sequence substitutions were also identified in the TACC3 coding sequence (Table 2), many of which resulted in no change at the protein level. Four nucleotide substitutions, not found in the GenBANK or SNP databases, which did alter the protein sequence were identified in both normal and affected sibs and thus also categorized as novel polymorphisms. In most cases, the alterations resulted from C to T base changes in CpG dinucleotides. Intriguingly, we identified a relatively rare insertion/deletion polymorphism in two 12mer amino acid repeats encoded by exon four [11]. The identification of this polymorphism could explain the previously reported variable number of copies of a distinct 24 amino acid repeat found in mouse TACC3 cDNAs from various sources [11,18,21]; although in only one case was the corresponding exon (exon 4) sequenced [18].
Table 2 TACC3 sequence changes detected in members of the Gilda Radner registry
Exon Codon Change Nucleotide changea Effectb Change in amino acid/protein property Remarks
3 TCA>TTA c.278C>T Ser93Leu Hydrophilic to hydrophobic In proband (ovarian cancer), and mother (uterine cancer); not in two unaffected sisters.
4 GAG> AAG c.427G>A Glu143Lys Acidic to basic Polymorphism
4 AGC>AGT c.531C>T Ser177Ser Silent Polymorphism
4 c.673_708del
AAAGCGGAGACTCCGCACG
GAGCCGAGGAAGAATGC Lys225_Cys236del Removal of one 12 amino acid repeat Polymorphism
4 GGC>GGG c.1086C>G Gly362Gly Silent Polymorphism
4 Ccg>Ctg c.1250C>T Pro417Leu Potential tertiary structure change Polymorphism
5 GCG>GTG c.1406C>T Ala469Val Polymorphism
6 GGG>GAG c.1541G>A Gly514Glu Small hydrophobic to acidic Only in affected sibling, not in unaffected sister or daughter.
9 TTC>TTT c.1809T>C Phe603Leu Aromatic to aliphatic Polymorphism
11 CAC>CAT c.1998C>T His666His Silent Polymorphism
16 c.2621T>A 3' untranslated region. Polymorphism
a: Based upon RefSeq TACC3 nucleotide sequence NM_006342
b: Based upon RefSeq TACC3 Protein sequence NP_006333
Significantly, in two unrelated patients from different families, we identified sequence alterations that were not present in unaffected sib(s) in the respective families or unrelated samples. In one case, a heterozygous C to T mutation in exon 6 was detected in the affected sister, but not her unaffected sister or sibling (Table 2). This mutation results in the dramatic change of amino acid 514 from glycine to negatively charged glutamic acid, and was not identified in any other normal or affected individuals from the registry, so far tested. Cross-referencing this sequence to GenBANK revealed only five occurrences of this mutation in cDNAs from the Expressed Sequence Tag database, all of which were derived from tumor tissues (including brain tumors and leukemias). The underlying c.1541G>A nucleotide substitution was detected in the SNP database from a survey of 71 individuals (subdivided into three panels based upon ethnicity), with the minor A allele frequency ranging from 0.312 in the Caucasian North American panel to 0.109 in the African American panel, frequencies that are higher than our detection of this allele in the Gilda Radner registry i.e. in one affected individual in the 76 families tested. In the second family, a heterozygous constitutional mutation was detected in exon 3 that resulted in a serine-leucine change at amino acid 93 in a patient diagnosed with clear cell ovarian carcinoma. This sequence change was not identified in other individuals from the registry, or in the GenBANK and SNP databases. The mother of this patient was previously diagnosed with uterine cancer (although the exact subtype of tumor was not recorded), and was also heterozygous for the same mutation. Two sisters, who are currently disease-free, do not carry the mutation. These data suggest that TACC3 may be a new familial predisposition or modifier locus for gynecological cancer.
Discussion
TACC1 and TACC3 are expressed in the epithelial components of several tissues in the body, including the mammary gland and the ovary. Based upon inferences from current SAGE data and accumulating genomic analysis, the specific goal of this report was to determine the potential significance of these proteins in the development of ovarian cancer. This has revealed that 78.5% of ovarian tumors lack appreciable expression of TACC1 and/or TACC3. In particular, we have determined that 67.7% (44 of 65) of ovarian tumors lose expression of TACC3. Subdivision of the tumors suggested a difference in the distribution pattern of expression of the two TACC proteins, with TACC3 loss being more common in serous papillary carcinoma compared with clear cell carcinomas, while TACC1 staining was less frequent in endometroid than in serous papillary tumor cores. However, due to the relatively small numbers of tumors in each histological category, particularly the rarer endometroid and mucinous subtypes, firm conclusions about the exact distribution pattern will require analysis of a much larger sample set. Furthermore, we detected constitutional mutations in the TACC3 gene in ovarian cancer patients from the Gilda Radner Familial Ovarian Cancer registry, in the absence of mutations in known predisposition genes.
During growth and progression, tumors can undergo a substantial amount of genomic rearrangement, including translocations, deletions and amplifications. Although many of these changes are random, due to the inherent genomic instability that can occur during tumor cell division, consistent abnormalities in the same or related tumor types can suggest that one or more genes in a particular region may be involved in the pathogenesis of the disease. Several different cancers, and in particular, breast, ovarian, endometrial and cervical cancers exhibit loss of 4p16 [5,22-27], the site of the TACC3 gene [11,10]. Thus, if TACC3 were a significant player in the pathogenesis of cancer, we would expect that a proportion of these tumors could similarly lose expression of TACC3. Indeed, in a survey of 500 resected breast tumors, TACC3 protein was significantly reduced in approximately 50% of the tumors [28]. In contrast, Affymetrix microarray analysis has revealed that levels of TACC3 mRNA increase during the transition of breast cancer from preinvasive ductal carcinoma in situ to invasive ductal carcinoma [29], suggesting that TACC3 may impart a proliferative advantage to a subset of breast cancers. A more recent study has indicated that TACC2 and TACC3 are members of a set of 21 proteins that are strong prognostic indicators of clinical outcome in breast cancer [30]. In addition, TACC3 is located within 200 kb of a translocation breakpoint cluster region associated with multiple myeloma, which results in TACC3 upregulation. This suggests that an increase in TACC3 may contribute to the pathogenesis of this B-Cell disorder [11]. A similar dichotomy in the expression pattern of TACC1 has been observed, in that TACC1 can be upregulated or lost in cancer [28,31,32]. This may in part be explained by as yet unidentified tissue specific functions of these proteins or the existence of cancer-associated alternative splice products, as evidenced for TACC1 in gastric cancer [32]. However, currently there is no firm evidence for alternative splicing of TACC3 in normal or cancerous tissues.
Each human TACC gene maps to a region of the genome that is consistently associated with tumorigenesis and progression. Rearrangements of the short arm of chromosome 8 are noted in several different cancers [3,33-35]. Typically, these rearrangements span a significant portion of 8p21-8p11, encompassing the TACC1 locus. Similarly, loss of chromosome 10, in particular the region of chromosome 10q25-26 flanked by the DNA markers D10S221 to D10S216, which encompasses the TACC2 locus, is a frequent occurrence in cancers of different origins (Reviewed in [36]). While this could suggest that the TACC genes may be mutated in a proportion of these cancers, to our knowledge, no report has directly assessed this possibility. With the prevalence of aberrations in the expression of TACC3 in the ovarian tumor arrays, we screened the coding sequence of the TACC3 gene in 76 ovarian cancer families that do not have mutations in known ovarian predisposition loci, such as BRCA1 and BRCA2 [1,2]. This led to the identification of several new coding polymorphisms, which were found in normal and affected individuals in our test subjects. With the relatively late age of onset of ovarian cancer, it remains possible that some of the siblings that have been classified as normal may actually be asymptomatic carriers. Thus, some of these novel polymorphisms may actually represent low penetrance modifiers of ovarian cancer risk, in a similar manner to BARD1 polymorphisms that were subsequently shown to be associated with increased breast cancer risk in the general population [37-39].
In addition, we identified two germline constitutional missense mutations that were specific to the affected patient, and were not found in currently unaffected siblings or unaffected members of the Gilda Radner registry. In addition, the Ser93Leu substitution was not found in normal individuals or cDNAs from the SNP or GenBANK databases. Significantly, the mother of the patient that carried the Ser93Leu substitution was also heterozygous for the same mutation and had previously been treated for a uterine cancer of undetermined type, while two siblings without this mutation remain disease free to date. In the Gilda Radner registry, the risk for uterine cancer is approximately five fold higher than the general population, and this risk increases with the number of first degree relatives diagnosed with ovarian cancer [40]. A similar finding was observed in a separate genealogy-based study [41]. Indeed, a link between ovarian and uterine cancer may not be surprising given the common embryonic origin of the epithelium of the female reproductive tract i.e. the coelomic epithelium [42]. In an independent study, in one BRCA1 linked family, the daughter of a patient diagnosed with ovarian cancer not only had bilateral breast cancer, but also uterine leiomyomata [43,44]. In addition, a germ line missense mutation in the BRCA1 associated protein BARD1 can give rise to independent tumors in the breast, ovary (clear cell carcinoma) and endometrium (also a clear cell carcinoma) [37]. This connection is particularly intriguing considering the recent observation that the C. elegans TAC protein interacts directly with the C. elegans homologue of BARD1 [45]. Unfortunately, we were unable to obtain additional tumor samples from our patients or material from an additional affected family member for further study. The Gly514Glu mutation, noted in the second family, is also present in five cDNAs derived from cancer cell lines or tumors in the expressed sequence tag (EST) database. Interestingly, the TACC3 sequence cloned by McKeveney et al [21] also contains the Gly514Glu substitution. Once again, this sequence was derived from a leukemia cell line, not from a normal tissue counterpart, suggesting that mutations in TACC3 may also be a feature of other hematological malignancies, in addition to the previously documented aberrant expression observed in multiple myeloma [46]. Intriguingly, two previous studies have reported an increased risk of ovarian cancer in the daughters of mothers with multiple myeloma [47,48], raising the possibility that TACC3 is a candidate gene contributing to the familial association of these two diseases. Further, larger population studies will be required to confirm whether both these changes represent rare polymorphisms, low penetrance modifiers that contribute to ovarian cancer development and/or are direct pathological mutations.
Potential functional significance of alterations of TACC3 in the ovarian surface epithelium
The TACC genes were first identified as potential oncogenes and tumor suppressors in breast cancer [10,36,49]. However, it is apparent from their proposed functions in cell division, transcriptional and posttranscriptional control, that they have the potential to interact with pathways that are commonly mutated in several different forms of cancer. TACC3 has an evolutionarily conserved interaction with the microtubule associated proteins and mitotic regulators, chTOG [50], and Aurora A kinase, and can be phosphorylated by the latter (Reviewed in [51]). In addition, a genetic interaction between TACC3 and p53 has been suggested, based upon work carried out in mice [52]. Intriguingly, we have recently found that TACC3 associates with a BRCA1-containing complex (Lauffart et al unpublished), and Boulton et al [45] have shown that C. elegans TAC directly interacts with the C. elegans homologue of BARD1. These functional interactions suggest that loss of the TACC proteins may promote tumor progression by increasing aberrations in centrosomal duplication and function, or DNA damage responses [10,53]. In addition, TACC3 has been shown to interact with nuclear localized transcription factors [18,51,54] and histone acetyltransferases [55]. Overexpression of TACC3 in hematopoietic cells results in aberrant localization of FOG1 [54], a regulator of the GATA transcription factors. Intriguingly, similar to TACC3, two members of this latter family, GATA 4 and GATA 6, are either lost or mislocalized in ovarian tumors [20]. Given the potential ability of TACC3 to interact with proteins known to be involved in ovarian tumorigenesis, our data therefore raise the possibility that TACC3 mutation, mislocalization or loss may serve as alternative mechanisms of functional inactivation of GATA4/6 and BRCA1 leading to the dedifferentiation and malignant development of the ovarian surface epithelium.
Conclusion
This report documents the first recorded analysis of the TACC genes during the development of ovarian cancer. We have now shown that 78.5% of ovarian tumors lack appreciable expression of TACC1 and/or TACC3 proteins, confirming the inferences made from published SAGE analysis. Together with the novel finding that constitutional mutations in the TACC3 gene may be associated with a subset of familial ovarian/gynecological malignancies, this study, therefore, suggests that the TACCs, and TACC3 in particular, are intimately involved in the mechanisms leading to the development of ovarian cancer.
Competing interests
The author(s) declare that they have no competing interests.
Authors' contributions
BL analyzed the tumor microarrays, and performed all statistical analysis and data presentation. MMV and JDB performed the immunohistochemical procedures. WAVE dHPLC and mutation analysis of families from the Gilda Radner Familial Ovarian Registry was carried out by RE, DC, and RAD. IHS conceived and designed the project, analyzed tumor microarray data and drafted the complete manuscript.
Pre-publication history
The pre-publication history for this paper can be accessed here:
Acknowledgements
This work was supported in part by the Elsa U. Pardee Foundation, developmental funds support from the Roswell Park Cancer Institute, and Core grant P30-CA16056-27 from the National Cancer Institute. We thank Dr James L. Kepner (Department of Biostatistics) for his helpful advice.
==== Refs
Gayther SA Russell P Harrington P Antoniou AC Easton DF Ponder BA The contribution of germline BRCA1 and BRCA2 mutations to familial ovarian cancer: no evidence for other ovarian cancer-susceptibility genes Am J Hum Genet 1999 65 1021 1029 10486320 10.1086/302583
Werness BA Ramus SJ Dicioccio RA Whittemore AS Garlinghouse-Jones K Oakley-Girvan I Tsukada Y Harrington P Gayther SA Ponder BA Piver MS Histopathology, FIGO stage, and BRCA mutation status of ovarian cancers from the Gilda Radner Familial Ovarian Cancer Registry Int J Gynecol Pathol 2004 23 29 34 14668547 10.1097/01.pgp.0000101083.35393.cd
Rao PH Harris CP Yan LX Li XN Mok SC Lau CC Multicolor spectral karyotyping of serous ovarian adenocarcinoma Genes Chromosomes Cancer 2002 33 123 132 11793438 10.1002/gcc.1221
Suzuki S Moore DH Ginzinger DG Godfrey TE Barclay J Powell B Pinkel D Zaloudek C Lu K Mills G Berchuck A Gray JW An approach to analysis of large-scale correlations between genome changes and clinical endpoints in ovarian cancer Cancer Res 2000 60 5382 5385 11034075
Ramus SJ Pharoah PD Harrington P Pye C Werness B Bobrow L Ayhan A Wells D Fishman A Gore M Dicioccio RA Piver MS Whittemore AS Ponder BA Gayther SA BRCA1/2 Mutation Status Influences Somatic Genetic Progression in Inherited and Sporadic Epithelial Ovarian Cancer Cases Cancer Res 2003 63 417 423 12543797
Gene Expression Omnibus repository. http://www ncbi nlm nih gov/SAGE/
Velculescu VE Zhang L Vogelstein B Kinzler KW Serial analysis of gene expression Science 1995 270 484 487 7570003
SAGE Tag to Gene Mapping Retrieve by SAGE tag: ACTCAATAAA http://www ncbi nlm nih gov/SAGE/index cgi?cmd=tagsearch&org=Hs&tag=ACTCAATAAA&anchor=NLAIII
SAGE Tag to Gene Mapping Retrieve by SAGE tag: TTTCATTGCC http://www ncbi nlm nih gov/sage/index cgi?&anchor=NLAIII&cmd=tagsearch&tag=TTTCATTGCC&org=Hs
Still IH Hamilton M Vince P Wolfman A Cowell JK Cloning of TACC1, an embryonically expressed, potentially transforming coiled coil containing gene, from the 8p11 breast cancer amplicon Oncogene 1999 18 4032 4038 10435627 10.1038/sj.onc.1202801
Still IH Vince P Cowell JK The third member of the transforming acidic coiled coil-containing gene family, TACC3, maps in 4p16, close to translocation breakpoints in multiple myeloma, and is upregulated in various cancer cell lines Genomics 1999 58 165 170 10366448 10.1006/geno.1999.5829
National Cancer Institute, Center for Cancer ResearchTechnology Initiatives, Tissue Array Research Program (TARP) http://ccr cancer gov/tech_initiatives/tarp/default asp
Shimizu Y Kamoi S Amada S Akiyama F Silverberg SG Toward the development of a universal grading system for ovarian epithelial carcinoma: testing of a proposed system in a series of 461 patients with uniform treatment and follow-up Cancer 1998 82 893 901 9486579 10.1002/(SICI)1097-0142(19980301)82:5<893::AID-CNCR14>3.0.CO;2-W
Morgan RJ Alvares RR Armstrong DK Copeland L Fiorica J Fishman DD Fowler J Gershenson D Greer BE Johnston C Kessinger A Lele S Locker GY Matulonis V Ozols RF Sabbatini P Teng N NCCN Practice guidelines for ovarian cancer. National Comprehensive Cancer Network, Complete library of NCCN Oncology Practice Guidelines 2000
Kuklin A Munson K Gjerde D Haefele R Taylor P Detection of single-nucleotide polymorphisms with the WAVE DNA fragment analysis system Genet Test 1997 1 201 206 10464646
dbSNP Homepage http://www ncbi nlm nih gov/SNP/index html
Aitola M Sadek CM Gustafsson JA Pelto-Huikko M Aint/Tacc3 is highly expressed in proliferating mouse tissues during development, spermatogenesis, and oogenesis J Histochem Cytochem 2003 51 455 469 12642624
Sadek CM Jalaguier S Feeney EP Aitola M Damdimopoulos AE Pelto-Huikko M Gustafsson J Isolation and characterization of AINT: a novel ARNT interacting protein expressed during murine embryonic development Mech Dev 2000 97 13 26 11025203 10.1016/S0925-4773(00)00415-9
Sadek CM Pelto-Huikko M Tujague M Steffensen KR Wennerholm M Gustafsson JA TACC3 expression is tightly regulated during early differentiation Gene Expr Patterns 2003 3 203 211 12711550 10.1016/S1567-133X(02)00066-2
Capo-chichi CD Roland IH Vanderveer L Bao R Yamagata T Hirai H Cohen C Hamilton TC Godwin AK Xu XX Anomalous expression of epithelial differentiation-determining GATA factors in ovarian tumorigenesis Cancer Res 2003 63 4967 4977 12941822
McKeveney PJ Hodges VM Mullan RN Maxwell P Simpson D Thompson A Winter PC Lappin TR Maxwell AP Characterization and localization of expression of an erythropoietin- induced gene, ERIC-1/TACC3, identified in erythroid precursor cells Br J Haematol 2001 112 1016 1024 11298601 10.1046/j.1365-2141.2001.02644.x
Climent J Martinez-Climent JA Blesa D Garcia-Barchino MJ Saez R Sanchez-Izquierdo D Azagra P Lluch A Garcia-Conde J Genomic loss of 18p predicts an adverse clinical outcome in patients with high-risk breast cancer Clin Cancer Res 2002 8 3863 3869 12473601
Forozan F Mahlamaki EH Monni O Chen Y Veldman R Jiang Y Gooden GC Ethier SP Kallioniemi A Kallioniemi OP Comparative genomic hybridization analysis of 38 breast cancer cell lines: a basis for interpreting complementary DNA microarray data Cancer Res 2000 60 4519 4525 10969801
Wang ZC Lin M Wei LJ Li C Miron A Lodeiro G Harris L Ramaswamy S Tanenbaum DM Meyerson M Iglehart JD Richardson A Loss of Heterozygosity and Its Correlation with Expression Profiles in Subclasses of Invasive Breast Cancers Cancer Res 2004 64 64 71 14729609
Sherwood JB Shivapurkar N Lin WM Ashfaq R Miller DS Gazdar AF Muller CY Chromosome 4 deletions are frequent in invasive cervical cancer and differ between histologic variants Gynecol Oncol 2000 79 90 96 11006038 10.1006/gyno.2000.5922
Nagase S Sato S Tezuka F Wada Y Yajima A Horii A Deletion mapping on chromosome 10q25-q26 in human endometrial cancer Br J Cancer 1996 74 1979 1983 8980400
Sato T Saito H Morita R Koi S Lee JH Nakamura Y Allelotype of human ovarian cancer Cancer Res 1991 51 5118 5122 1655245
Conte N Charafe-Jauffret E Delaval B Adelaide J Ginestier C Geneix J Isnardon D Jacquemier J Birnbaum D Carcinogenesis and translational controls: TACC1 is down-regulated in human cancers and associates with mRNA regulators Oncogene 2002 21 5619 5630 12165861 10.1038/sj.onc.1205658
Ma XJ Salunga R Tuggle JT Gaudet J Enright E McQuary P Payette T Pistone M Stecker K Zhang BM Zhou YX Varnholt H Smith B Gadd M Chatfield E Kessler J Baer TM Erlander MG Sgroi DC Gene expression profiles of human breast cancer progression Proc Natl Acad Sci U S A 2003 100 5974 5979 12714683 10.1073/pnas.0931261100
Jacquemier J Ginestier C Rougemont J Bardou VJ Charafe-Jauffret E Geneix J Adelaide J Koki A Houvenaeghel G Hassoun J Maraninchi D Viens P Birnbaum D Bertucci F Protein expression profiling identifies subclasses of breast cancer and predicts prognosis Cancer Res 2005 65 767 779 15705873
Rhodes DR Barrette TR Rubin MA Ghosh D Chinnaiyan AM Meta-analysis of microarrays: interstudy validation of gene expression profiles reveals pathway dysregulation in prostate cancer Cancer Res 2002 62 4427 4433 12154050
Line A Slucka Z Stengrevics A Li G Rees RC Altered splicing pattern of TACC1 mRNA in gastric cancer Cancer Genet Cytogenet 2002 139 78 83 12547166 10.1016/S0165-4608(02)00607-6
Adelaide J Chaffanet M Imbert A Allione F Geneix J Popovici C van Alewijk D Trapman J Zeillinger R Borresen-Dale AL Lidereau R Birnbaum D Pebusque MJ Chromosome region 8p11-p21: refined mapping and molecular alterations in breast cancer Genes Chromosomes Cancer 1998 22 186 199 9624530 10.1002/(SICI)1098-2264(199807)22:3<186::AID-GCC4>3.0.CO;2-S
Pan Y Lui WO Nupponen N Larsson C Isola J Visakorpi T Bergerheim US Kytola S 5q11, 8p11, and 10q22 are recurrent chromosomal breakpoints in prostate cancer cell lines Genes Chromosomes Cancer 2001 30 187 195 11135436 10.1002/1098-2264(2000)9999:9999<::AID-GCC1075>3.3.CO;2-8
Theillet C Adelaide J Louason G Bonnet-Dorion F Jacquemier J Adnane J Longy M Katsaros D Sismondi P Gaudray P FGFRI and PLAT genes and DNA amplification at 8p12 in breast and ovarian cancers Genes Chromosomes Cancer 1993 7 219 226 7692948
Lauffart B Gangisetty O Still IH Molecular cloning, genomic structure and interactions of the putative breast tumor suppressor TACC2 Genomics 2003 81 192 201 12620397 10.1016/S0888-7543(02)00039-3
Thai TH Du F Tsan JT Jin Y Phung A Spillman MA Massa HF Muller CY Ashfaq R Mathis JM Miller DS Trask BJ Baer R Bowcock AM Mutations in the BRCA1-associated RING domain (BARD1) gene in primary breast, ovarian and uterine cancers Hum Mol Genet 1998 7 195 202 9425226 10.1093/hmg/7.2.195
Karppinen SM Heikkinen K Rapakko K Winqvist R Mutation screening of the BARD1 gene: evidence for involvement of the Cys557Ser allele in hereditary susceptibility to breast cancer J Med Genet 2004 41 e114 15342711 10.1136/jmg.2004.020669
Ishitobi M Miyoshi Y Hasegawa S Egawa C Tamaki Y Monden M Noguchi S Mutational analysis of BARD1 in familial breast cancer patients in Japan Cancer Lett 2003 200 1 7 14550946 10.1016/S0304-3835(03)00387-2
Jishi MF Itnyre JH Oakley-Girvan IA Piver MS Whittemore AS Risks of cancer among members of families in the Gilda Radner Familial Ovarian Cancer Registry Cancer 1995 76 1416 1421 8620417
Kerber RA Slattery ML The impact of family history on ovarian cancer risk. The Utah Population Database Arch Intern Med 1995 155 905 912 7726698 10.1001/archinte.155.9.905
Auersperg N Wong AS Choi KC Kang SK Leung PC Ovarian surface epithelium: biology, endocrinology, and pathology Endocr Rev 2001 22 255 288 11294827 10.1210/er.22.2.255
Paley PJ Swisher EM Garcia RL Agoff SN Greer BE Peters KL Goff BA Occult cancer of the fallopian tube in BRCA-1 germline mutation carriers at prophylactic oophorectomy: a case for recommending hysterectomy at surgical prophylaxis Gynecol Oncol 2001 80 176 180 11161856 10.1006/gyno.2000.6071
Agoff SN Mendelin JE Grieco VS Garcia RL Unexpected gynecologic neoplasms in patients with proven or suspected BRCA-1 or -2 mutations: implications for gross examination, cytology, and clinical follow-up Am J Surg Pathol 2002 26 171 178 11812938 10.1097/00000478-200202000-00003
Boulton SJ Martin JS Polanowska J Hill DE Gartner A Vidal M BRCA1/BARD1 Orthologs Required for DNA Repair in Caenorhabditis elegans Curr Biol 2004 14 33 39 14711411 10.1016/j.cub.2003.11.029
Stewart JP Thompson A Santra M Barlogie B Lappin TR Shaughnessy JJ Correlation of TACC3, FGFR3, MMSET and p21 expression with the t(4;14)(p16.3;q32) in multiple myeloma Br J Haematol 2004 126 72 76 15198734 10.1111/j.1365-2141.2004.04996.x
Hemminki K Granstrom C Familial clustering of ovarian and endometrial cancers Eur J Cancer 2004 40 90 95 14687794 10.1016/S0959-8049(03)00627-0
Stratton JF Thompson D Bobrow L Dalal N Gore M Bishop DT Scott I Evans G Daly P Easton DF Ponder BA The genetic epidemiology of early-onset epithelial ovarian cancer: a population-based study Am J Hum Genet 1999 65 1725 1732 10577927 10.1086/302671
Chen HM Schmeichel KL Mian IS Lelievre S Petersen OW Bissell MJ AZU-1: A Candidate Breast Tumor Suppressor and Biomarker for Tumor Progression Mol Biol Cell 2000 11 1357 1367 10749935
Gard DL Becker BE Josh RS MAPping the eukaryotic tree of life: structure, function, and evolution of the MAP215/Dis1 family of microtubule-associated proteins Int Rev Cytol 2004 239 179 272 15464854
Still IH Vettaikkorumakankauv AK DiMatteo A Liang P Structure-function evolution of the transforming acidic coiled coil genes revealed by analysis of phylogenetically diverse organisms BMC Evol Biol 2004 4 16 15207008 10.1186/1471-2148-4-16
Piekorz RP Hoffmeyer A Duntsch CD McKay C Nakajima H Sexl V Snyder L Rehg J Ihle JN The centrosomal protein TACC3 is essential for hematopoietic stem cell function and genetically interfaces with p53-regulated apoptosis EMBO J 2002 21 653 664 11847113 10.1093/emboj/21.4.653
Gergely F Karlsson C Still I Cowell J Kilmartin J Raff JW The TACC domain identifies a family of centrosomal proteins that can interact with microtubules Proc Natl Acad Sci U S A 2000 97 14352 14357 11121038 10.1073/pnas.97.26.14352
Garriga-Canut M Orkin SH Transforming acidic coiled-coil protein 3 (TACC3) controls friend of GATA-1 (FOG-1) subcellular localization and regulates the association between GATA-1 and FOG-1 during hematopoiesis J Biol Chem 2004 279 23597 23605 15037632 10.1074/jbc.M313987200
Gangisetty O Lauffart B Sondarva GV Chelsea DM Still IH The transforming acidic coiled coil proteins interact with nuclear histone acetyltransferases Oncogene 2004 23 2559 2563 14767476 10.1038/sj.onc.1207424
| 15918899 | PMC1175095 | CC BY | 2021-01-04 16:30:36 | no | BMC Womens Health. 2005 May 26; 5:8 | utf-8 | BMC Womens Health | 2,005 | 10.1186/1472-6874-5-8 | oa_comm |
==== Front
Curr Control Trials Cardiovasc MedCurrent Controlled Trials in Cardiovascular Medicine1468-67081468-6694BioMed Central 1468-6708-6-91591068010.1186/1468-6708-6-9ResearchEffect of rosuvastatin on outcomes in chronic haemodialysis patients – design and rationale of the AURORA study Fellström Bengt [email protected] Faiez [email protected] Roland [email protected] Hallvard [email protected] Alan [email protected] Helen [email protected] Wim [email protected] AURORA Study Group 1 Department of Medical Science, Renal Unit, University Hospital, Uppsala, Sweden2 Clinical Investigation Center INSERM (CIC), Hôpital Jeanne d'Arc, Toul, France3 Med Klinik IV, Univ.-Klinik Erlangen-Nürnberg, Germany4 Department of Nephrology, Rikshospitalet, Oslo, Norway5 Department of Medicine and Therapeutics, Western Infirmary Hospital, Glasgow, United Kingdom6 AstraZeneca, Macclesfield, United Kingdom7 AstraZeneca, Macclesfield, United Kingdom2005 23 5 2005 6 1 9 9 4 4 2005 23 5 2005 Copyright © 2005 Fellström et al; licensee BioMed Central Ltd.2005Fellström 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 end-stage renal disease (ESRD) are at high risk of cardiovascular events. Multiple risk factors for atherosclerosis are present in ESRD and may contribute to the increased risk of cardiovascular mortality in this population. In contrast to patients with normal renal function, the benefits of modifying lipid levels on cardiovascular outcomes in patients with ESRD on haemodialysis have yet to be confirmed in large prospective randomised trials. A study to evaluate the Use of Rosuvastatin in subjects On Regular haemodialysis: an Assessment of survival and cardiovascular events (AURORA) will be the first large-scale international trial to assess the effects of statin therapy on cardiovascular morbidity and mortality in ESRD patients on chronic haemodialysis.
Methods
More than 2,750 ESRD patients who have been receiving chronic haemodialysis treatment for at least 3 months have been randomised (1:1), irrespective of baseline lipid levels, to treatment with rosuvastatin 10 mg or placebo. The primary study endpoint is the time to a major cardiovascular event (first occurrence of cardiovascular death, non-fatal myocardial infarction or non-fatal stroke). Secondary endpoints include all-cause mortality, major cardiovascular event-free survival time, time to cardiovascular death, time to non-cardiovascular death, cardiovascular interventions, tolerability of treatment and health economic costs per life-year saved. Study medication will be given until 620 subjects have experienced a major cardiovascular event.
Conclusion
Our hypothesis is that results from AURORA will establish the clinical efficacy and tolerability of rosuvastatin in patients with ESRD receiving chronic haemodialysis and guide the optimal management of this expanding population.
atherosclerosiscardiovascular diseaseend-stage renal diseasehaemodialysislipidsstatin
==== Body
Background
Patients with end-stage renal disease (ESRD) undergoing haemodialysis have substantially higher cardiovascular disease mortality rates than the general population [1-5]. Accelerated atherosclerosis has been observed in haemodialysis patients [6] and may contribute to this increased cardiovascular event rate. Treatment of ESRD and its cardiovascular consequences places a large burden upon healthcare providers, and the cost and prevalence are expected to increase greatly over the next decade [5]. Therefore, controlling risk factors for atherosclerosis may play an important role in preventing cardiovascular events in these individuals.
Hydroxy-methylglutaryl-coenzyme A (HMG-CoA) reductase inhibitors (statins) have been demonstrated to reduce coronary heart disease morbidity and mortality in several landmark trials [7-14]. The reductions in cardiovascular events also occur in patients with average to lower than average baseline low-density lipoprotein cholesterol (LDL-C) levels [12,14] and the benefits of statin therapy can be independent of lipid lowering [15]. Most haemodialysis patients do not have elevated cholesterol (total and LDL-C) [16,17], and the highest mortality risk is often in patients with very low cholesterol levels [18]. Other chronic diseases and malnutrition may act to lower blood cholesterol and independently increase the risk of death, thereby contributing to the negative association between very low cholesterol and mortality observed in these patients [19,20]. Indeed, a recent prospective study suggests that levels of total cholesterol (TC) may be associated with mortality in haemodialysis patients without evidence of inflammation and malnutrition [20]. Haemodialysis patients have other lipid abnormalities such as lower levels of high-density lipoprotein cholesterol (HDL-C) and elevated intermediate-density lipoprotein (IDL) and triglycerides [16,17].
Although the efficacy of statins is well established in conditions associated with increased cardiovascular risk, dialysis patients have generally been excluded from statin outcome trials because of their related co-morbidities and as a result of pharmacokinetic and safety issues. It is not appropriate to extrapolate evidence from patients with normal renal function to patients with ESRD on haemodialysis, and there is a need to investigate the benefit-risk profile of statins specifically in this population. Data from an observational study indicate that statin treatment may improve survival in patients with ESRD [21]. Mortality was 32% lower in patients with ESRD who received statin treatment compared with patients not receiving statins, a finding that is consistent with other large outcome statin trials [7,8,11]. However, observational studies do not establish a causal relationship and large-scale randomised studies are required [22]. The Die Deutsche Diabetes Dialyse Studie (4D study) investigated the effect of atorvastatin 20 mg or placebo on cardiovascular mortality, non-fatal myocardial infarction and stroke in 1,255 patients with type 2 diabetes who were on haemodialysis treatment for no more than 2 years [23]. Initial results presented at the annual meeting of the American Society of Nephrology indicate that the primary endpoint was reduced by 8% after 4 years' treatment with atorvastatin, but this was not significantly different from placebo [24]. Further large-scale long-term prospective, randomised studies are required to investigate the efficacy and safety of statins for reducing cardiovascular morbidity and mortality, specifically in ESRD patients on haemodialysis [22].
Rosuvastatin is the most efficacious of the available statins for lowering LDL-C levels, causing reductions of 47% at the initial starting dose of 10 mg [25]. In addition, rosuvastatin has benefits across the lipid profile, including increases in HDL-C, reductions in small dense LDL and triglyceride-rich lipoprotein particles [25,26], and has a safety profile consistent with other available statins [27]. These properties make it an ideal agent for a study investigating the benefits of statin treatment for the prevention of cardiovascular events in a population of patients at high risk of cardiovascular events. AURORA (A study to evaluate the Use of Rosuvastatin in subjects On Regular haemodialysis: an Assessment of survival and cardiovascular events) is the first large-scale international trial to assess the effects of a statin on cardiovascular morbidity and mortality in ESRD patients on chronic haemodialysis irrespective of baseline lipid levels.
Methods
Study design
AURORA is a double-blind, randomised, multicentre, phase IIIb, parallel-group study comparing the effects of rosuvastatin (10 mg once daily) with placebo on survival and cardiovascular events in ESRD patients on chronic haemodialysis.
More than 2,750 subjects from approximately 300 centres in Europe, Canada, Australia, Brazil, Mexico and South Korea have been randomised, irrespective of baseline lipid levels. During the recruitment period, patients were screened over 2 weeks for eligibility according to the major inclusion and exclusion criteria (Table 1). Demographic data were collected on age, gender, race, medical history, physical examination (including height, weight, blood pressure and heart rate), smoking status, and method and duration of dialysis (Table 2). Patients were then randomised (1:1) into the two blinded treatment arms of the study, rosuvastatin 10 mg/day or placebo, and visits to assess safety and efficacy are scheduled to occur at 3 months, 6 months and every 6 months thereafter until the end of the study (Figure 1; Table 2).
Table 1 Major eligibility criteria
Inclusion
Men and women aged 50–80 years
End-stage renal failure and chronic haemodialysis for at least 3 months
Provision of written informed consent
Exclusion
Underlying haematological, neoplastic, GI, metabolic (other than diabetes) or infectious condition expected to reduce survival to less than 1 year
Patients likely to require a kidney transplant within 1 year
Statin therapy within the previous 6 months
History of serious reactions to statins
Unexplained CK >3 times ULN
Active liver disease (ALT >3 times ULN)
Uncontrolled hypothyroidism
A disallowed medication, such as another lipid-modifying agent or cyclosporin
ALT: Alanine transaminase, CK: Creatine kinase, GI: Gastrointestinal, ULN: Upper limit of normal
Figure 1 Study design
Objectives and endpoints
The primary objective of the AURORA study is to compare the effects of rosuvastatin with placebo on cardiovascular events. The primary endpoint is the time to a major cardiovascular event (cardiovascular death, fatal myocardial infarction or non-fatal stroke). An independent and blinded Clinical Endpoint Committee (see Appendix) will review all deaths, strokes and myocardial infarctions using a predefined set of event definitions to ensure consistency of event diagnosis across all subjects throughout the study.
Secondary endpoints include all-cause mortality, cardiovascular event-free survival, cardiovascular death, non-cardiovascular death, procedures as a result of stenosis or thrombosis of the vascular access for chronic haemodialysis (arteriovenous fistulas and grafts only), and coronary or peripheral revascularisations. Furthermore, the tolerability of rosuvastatin in ESRD patients undergoing regular chronic haemodialysis will be compared with placebo. The health economic impact of rosuvastatin treatment on the utilisation of resources and costs associated with the occurrence of a major cardiovascular event will also be assessed. Costs related to hospitalisations will be combined with survival analyses to enable the calculation of cost per life-year saved.
Tertiary objectives include assessing the efficacy of treatment at 3 and 12 months post-randomisation on high-sensitivity C-reactive protein (hsCRP) and various fasting lipid parameters: TC, LDL-C, HDL-C, non-HDL-C, TC/HDL-C, LDL-C/HDL-C, triglycerides, apolipoprotein B (Apo B), apolipoprotein AI (Apo AI), Apo B/Apo AI ratio, and oxidised LDL. Additionally, changes in TC, LDL-C, HDL-C, non-HDL-C, TC/HDL-C, LDL-C/HDL-C and triglycerides will be assessed after 24 months, 36 months (then yearly as required) and at the final visit. A central laboratory service (Quintiles Laboratories, Livingston, UK) certified for standardisation of lipid analyses by the Standardization Program of the Center for Disease Control and Prevention and the National Heart, Lung and Blood Institute will perform all laboratory testing of lipids and hsCRP. Additional laboratory work will be performed in subgroup studies analysing the predictive power of various other markers of cardiovascular risk.
Genetic research may also be performed on certain subjects at some study sites on an optional basis. DNA samples will be obtained for future research into the effects of genetic polymorphisms on the response to rosuvastatin and placebo. Susceptibility to, and prognosis of, cardiovascular disease and lipid disorders will also be studied.
Study population
Eligibility and treatment withdrawal
Irrespective of baseline lipid levels, patients aged 50–80 years with ESRD who have been receiving chronic haemodialysis for at least 3 months were considered for inclusion in the study unless they fulfilled any of the exclusion criteria (key criteria detailed in Table 1). In addition, patients who undergo kidney transplant surgery will be withdrawn from study medication, but will continue to follow scheduled study assessments. Treatment will also be discontinued if patients have creatine kinase >10 times upper limit of normal (ULN) accompanied by unexplained muscle pain or if myopathy is suspected or if the patient becomes pregnant. Other reasons that may lead to treatment discontinuation include: an adverse event or endpoint that in the opinion of the investigator warrants treatment withdrawal; alanine transaminase increase >3 times ULN on two occasions; other safety reasons as judged by the investigator; protocol non-compliance.
Ethics
The study protocol and patient consent form have been approved by an Independent Ethics Committee or Institutional Review Board, as appropriate. The study is being performed in accordance with the ethical principles that have their origin in the Declaration of Helsinki and are consistent with International Conference on Harmonisation/Good Clinical Practice. The investigator from each centre will ensure that subjects are given full and adequate oral and written information about the nature, purpose, risks and benefits of the study and that written consent is obtained. The conduct and progress of the study is reviewed by the Steering Committee and Executive Steering Committee (see Appendix). In addition, an independent Data and Safety Monitoring Board (DSMB) (see Appendix) is responsible for monitoring all safety aspects of the study. The DSMB will review and evaluate all serious adverse events and endpoints at least every 6 months using unblinded data and will then report to the Executive Steering Committee in a blinded manner. The DSMB will also be responsible for making recommendations to the Executive Steering Committee regarding modifying or stopping the study early.
Data analysis
The efficacy endpoints will be analysed using the intent-to-treat population, including all randomised subjects. For the time-to-event endpoints, Kaplan-Meier curves will be used to show the proportion of subjects without an endpoint. An unadjusted Cox Proportional Hazards model with treatment as the covariate will be used to analyse the endpoint. An additional Cox Proportional Hazards regression model [28] will also be constructed adjusting for known and potential risk factors as a secondary analysis and to test for interactions between risk factors and treatment. The number of coronary or peripheral revascularisations will also be summarised, but not formally analysed. For the tertiary endpoints, the percentage change from baseline for each parameter will be assessed for each time point by an analysis of covariance, with the baseline value as a covariate. An exploratory analysis will also be conducted to examine whether lipid values at 3, 12, 24, 36 months (yearly if required) and at the final visit are predictive of future cardiovascular events.
The sample size was determined such that the study will be adequately powered for the analysis of the primary endpoint, time from randomisation to a major cardiovascular event. Based on previous studies, it is anticipated that the cardiovascular event rate in the placebo group will be 11% per year. The sample size has been calculated in order to be able to detect a 25% reduction in cardiovascular events rate per year at a two-sided significance level of 4.86%, with 90% power. The study will continue until 620 subjects have experienced a major cardiovascular event. This is expected to be approximately 3.9 years after initiation of the study, which began in January 2003. An interim efficacy analysis will be conducted by the DSMB when 310 subjects have experienced a major cardiovascular event. The purpose of the interim analysis is to investigate, at this time, the possibility of concluding a statistically significant difference between the treatments for time from randomisation to major cardiovascular event.
Discussion
Patients with ESRD have an increased risk of cardiovascular morbidity and mortality [2,4], and this contributes to the large burden this disease places upon healthcare providers [5]. A number of studies have indicated that patients on dialysis with very low TC levels have a higher rate of mortality [18,20]. Both small dense LDL and triglyceride-rich lipoproteins have been implicated in the development of cardiovascular disease [29-32] and are common in patients with ESRD and on haemodialysis [16,33-37]. Triglyceride-rich lipoproteins, in particular IDL, are prevalent and predictive of atherosclerosis in haemodialysis patients [17]. Thus, modifying lipid levels with a statin should have beneficial effects on cardiovascular outcomes in these patients. Large outcome studies also indicate that the positive effects of statins occur irrespective of baseline lipid levels [12,14]. While the efficacy of statins for altering the lipid profile has been studied previously in ESRD patients [38-42], their long-term benefits on cardiovascular outcomes in this population have not yet been examined and dialysis patients have generally been excluded from statin outcome trials.
Several studies examining outcome after statin treatment have recently been completed or are in progress in patients with related conditions. In a study in renal transplant patients, fluvastatin reduced the risk of cardiac death or non-fatal myocardial infarction by 35% compared with subjects receiving placebo (risk ratio 0.65 [95% confidence intervals 0.48–0.88]; p = 0.005) [43]. Initial findings of the 4D study did not demonstrate a significant difference between atorvastatin 20 mg and placebo treatment; however, LDL-C levels were reduced during the study period [24]. Another study is ongoing; the Study of Heart And Renal Protection (SHARP) aims to assess the effects of simvastatin and the cholesterol-absorption inhibitor ezetimibe among patients with chronic kidney disease [18]. AURORA is the first international, prospective, randomised, double-blind, placebo-controlled study to assess whether statin therapy alone can reduce the incidence of coronary events in ESRD patients on chronic haemodialysis.
Most haemodialysis patients do not have elevated LDL-C and, in accordance with findings from the Heart Protection Study, the AURORA trial will randomise patients irrespective of their baseline lipid levels [12]. Patients on haemodialysis have lower levels of HDL-C and higher triglycerides compared with control subjects [16]. Accordingly, a statin with efficacy across the lipid profile would be appropriate to assess the benefits of treatment in ESRD patients. Rosuvastatin is the most efficacious statin reported for lowering LDL-C [44], and has benefits across the lipid profile. In a study of hypercholesterolaemic patients, rosuvastatin has been shown to increase HDL-C by 7.7–9.6% compared with 4.4–5.7% for atorvastatin and 5.2–6.0% for simvastatin across the 10–40 mg dose range [44]. In addition, rosuvastatin has been shown to substantially lower triglyceride-rich lipoproteins in both hypercholesterolaemic and hypertriglyceridaemic patients [45]. The effects of haemodialysis upon excretion of statins should also be considered. On the basis of pharmacokinetic, pharmacodynamic and safety data obtained from a pilot study, rosuvastatin 10 mg can be safely administered to patients with ESRD on chronic haemodialysis (data on file, AstraZeneca).
In addition to their ability to modify lipid levels, statins produce other effects that may benefit ESRD patients. Oxidative stress has been implicated in the pathogenesis of cardiovascular disease, and it is enhanced in patients with renal insufficiency [46,47]. Recently, statins have been reported to have anti-oxidant effects [48], which may be of benefit to these patients. Statins may also have additional non-lipid or pleiotropic effects, for example, improving endothelial function [15]. Endothelial dysfunction has been implicated in the development of cardiovascular disease and has been shown to be closely associated with the degree of renal insufficiency [49]. Pre-clinical experiments have shown that rosuvastatin can improve endothelial function [50], although this has yet to be verified in a clinical situation. Finally, left ventricular hypertrophy and heart failure can also develop in ESRD patients on haemodialysis [51]. Statin therapy has been reported to reduce left ventricular hypertrophy [52] and retrospective analyses indicate that these agents can also help to prevent heart failure [53]. These properties could provide an additional beneficial effect in ESRD patients.
The AURORA study should provide valuable information on the utility of statin treatment for the reduction of cardiovascular events in patients with ESRD. As an atherogenic lipid profile and endothelial dysfunction are often present in individuals with ESRD and contribute to the development of atherosclerosis, it is anticipated that rosuvastatin will improve cardiovascular outcomes in this patient population. Furthermore, it is hoped that AURORA will produce data on the cost effectiveness (cost due to hospitalisation and cost per year of life saved) and long-term safety of rosuvastatin treatment in ESRD patients receiving chronic haemodialysis. This study will also generate a large database of information from a well-documented cohort, which may be valuable in the epidemiological evaluation of cardiovascular risk in patients with ESRD.
Competing interests
This study is sponsored by AstraZeneca. The authors also have affiliations with other pharmaceutical companies including Pfizer (AJ, BF, HH, RS, FZ), Bristol-Myers Squibb (AJ, HH, RS), Novartis (AJ, BF, HH, RS, FZ), Fujisawa (AJ, BF), Roche (AJ, BF, HH, RS), Merck (AJ, BF, HH, RS), Schering-Plough (BF, HH), Wyeth (AJ, BF), GlaxoSmithKline (AJ, BF, HH), Servier (AJ), Actelion (AJ), Pharmalink (BF). These relate to personal or institutional-affiliated receipt of income in the areas: Research grants, honoraria and Consultant fees presently or during the last five years. The authors WW and HR are employed by AstraZeneca.
Authors' contributions
Author BF is the Principal Investigator of the study and contributed to the concept and design of the study
Authors FZ, RS, HH, AJ and WW contributed to the design of the study
Author HR contributed to the planned statistical analyses of the study and sample size determination.
All authors read and approved the final manuscript.
Appendix
Executive Steering Committee
Professor B Fellström (Principal Investigator; Uppsala, Sweden), Professor F Zannad (Toul, France), Professor R Schmieder (Erlangen, Germany), Dr H Holdaas (Olso, Norway), Dr A Jardine (Glasgow, UK).
Steering Committee
Executive Steering Committee members plus Dr K Bannister (Adelaide, Australia), Dr J Beutler (Utrecht, The Netherlands), Professor D Chae (Kyungki-Do, South Korea), Professor SM Cobbe (Glasgow, UK), Dr B Espinoza Vazquez (Colonia Toriello Guerra, Mexico), Professor C Gronhagen-Riska (Helsinki, Finland), Dr J Lima (Sao Paulo, Brazil), Professor R Lins (Antwerpen, Belgium), Dr A McMahon (Edmonton, Canada), Professor G Mayer (Innsbruck, Austria), Professor H Parving (Gentofte, Denmark), Professor G Remuzzi (Bergamo, Italy), Dr O Samuelsson (Goteborg, Sweden), Professor S Sonkodi (Szeged, Hungary), Dr G Suleymanlar (Antalya, Turkey), Professor V Tesar (Prague, Czech Republic), Dr D Tsakiris (Veria, Greece), Professor V Todorov (Pleven, Bulgaria), Professor A Wiecek (Katowice, Poland), Professor R Wûthrich (Gallen, Switzerland).
Data and Safety Monitoring Board
Professor H Dargie (Chair; Glasgow, UK), Professor E Ritz (Heidelberg, Germany), Professor H Wedel (Goteborg, Sweden), Professor AH Zwinderman (Amsterdam, The Netherlands).
Clinical Endpoint Committee
Professor SM Cobbe (Chair; Glasgow, UK), Dr A Brady (Glasgow, UK), Dr C Deighan (Glasgow, UK), Dr A Gaw (Glasgow, UK), Professor P Macfarlane (Glasgow, UK), Professor D Stott (Glasgow, UK).
Supplementary Material
Additional File 1
Table 2. study plan (includes table footnotes)
Click here for file
Acknowledgements
We gratefully acknowledge the individuals and committees listed in the Appendix (below) for their invaluable contributions to the planning and implementation of this study.
==== Refs
Brown JH Hunt LP Vites NP Short CD Gokal R Mallick NP Comparative mortality from cardiovascular disease in patients with chronic renal failure Nephrol Dial Transplant 1994 9 1136 1142 7800214
Foley RN Parfrey PS Sarnak MJ Clinical epidemiology of cardiovascular disease in chronic renal disease Am J Kidney Dis 1998 32 S112 S119 9820470
Levey AS Beto JA Coronado BE Eknoyan G Foley RN Kasiske BL Klag MJ Mailloux LU Manske CL Meyer KB Parfrey PS Pfeffer MA Wenger NK Wilson PW Wright JT Jr Controlling the epidemic of cardiovascular disease in chronic renal disease. National Kidney Foundation Task Force on Cardiovascular Disease Am J Kidney Dis 1998 32 853 906 9820460
Parfrey PS Cardiac disease in dialysis patients: diagnosis, burden of disease, prognosis, risk factors and management Nephrol Dial Transplant 2000 15 Suppl 5 58 68 11073277 10.1093/ndt/15.suppl_5.58
US Renal Data System National Institutes of Health, National Institute of Diabetes and Digestive and Kidney Diseases 2000 Annual Data Report 2000 Bethesda, MD
Lindner A Charra B Sherrard DJ Scribner BH Accelerated atherosclerosis in prolonged maintenance hemodialysis N Engl J Med 1974 290 697 701 4813742
Scandinavian Simvastatin Survival Study Group Randomised trial of cholesterol lowering in 4444 patients with coronary heart disease: the Scandinavian Simvastatin Survival Study (4S) Lancet 1994 344 1383 1389 7968073
Shepherd J Cobbe SM Ford I Isles CG Lorimer AR MacFarlane PW McKillop JH Packard CJ for the West of Scotland Coronary Prevention Study Group Prevention of coronary heart disease with pravastatin in men with hypercholesterolemia N Engl J Med 1995 333 1301 1307 7566020 10.1056/NEJM199511163332001
Sacks FM Pfeffer MA Moye LA Rouleau JL Rutherford JD Cole TG Brown L Warnica JW Arnold JMO Wun C-C Davis BR Braunwald E The effect of pravastatin on coronary events after myocardial infarction in patients with average cholesterol levels N Eng J Med 1996 335 1001 1009 10.1056/NEJM199610033351401
Downs JR Clearfield M Weis S Whitney E Shapiro DR Beere PA Langendorfer A Stein EA Kruyer W Gotto AM Jr Primary prevention of acute coronary events with lovastatin in men and women with average cholesterol levels. Results of AFCAPS/TexCAPS JAMA 1998 279 1615 1622 9613910 10.1001/jama.279.20.1615
The Long-term Intervention with Pravastatin in Ischemic Disease (LIPID) Study Group Prevention of cardiovascular events and death with pravastatin in patients with coronary heart disease and a broad range of initial cholesterol levels N Eng J Med 1998 339 1349 1357 10.1056/NEJM199811053391902
Heart Protection Study Collaborative Group MRC/BHF Heart Protection Study of cholesterol lowering with simvastatin in 20,536 high-risk individuals: a randomised placebo-controlled trial Lancet 2002 360 7 22 12114036 10.1016/S0140-6736(02)09327-3
Shepherd J Blauw GJ Murphy MB Bollen EL Buckley BM Cobbe SM Ford I Gaw A Hyland M Jukema JW Kamper AM MacFlarlane PW Meinders AE Norrie J Packard CJ Perry IJ Stott DJ Sweeney BJ Twomey C Westendorp RG Pravastatin in elderly individuals at risk of vascular disease (PROSPER): a randomised controlled trial Lancet 2002 360 1623 1630 12457784 10.1016/S0140-6736(02)11600-X
Sever PS Dahlöf B Poulter NR Wedel H Beevers G Caulfield M Collins R Kjeldsen SE Kristinsson A McInnes GT Mehlsen J Nieminen M Obrien E Östergren J for the ASCOT investigators Prevention of coronary and stroke events with atorvastatin in hypertensive patients who have average or lower-than-average cholesterol concentrations, in the Anglo-Scandinavian Cardiac Outcomes Trial-Lipid Lowering Arm (ASCOT-LLA): a multicentre randomised controlled trial Lancet 2003 361 1149 1158 12686036 10.1016/S0140-6736(03)12948-0
Laufs U Beyond lipid-lowering: effects of statins on endothelial nitric oxide Eur J Clin Pharmacol 2003 58 719 731 12634978
Shoji T Nishizawa Y Kawagishi T Tanka M Kawasaki K Tabata T Inoue T Morii H Atherogenic lipoprotein changes in the absence of hyperlipidemia in patients with chronic renal failure treated by hemodialysis Atherosclerosis 1997 131 229 236 9199276 10.1016/S0021-9150(97)00054-3
Shoji T Nishizawa Y Kawagishi T Kawasaki K Taniwaki H Tabata T Inoue T Morii H Intermediate-density lipoprotein as an independent risk factor for aortic atherosclerosis in hemodialysis patients J Am Soc Nephrol 1998 9 1277 1284 9644639
Baigent C Landry M Study of Heart and Renal Protection (SHARP) Kidney Int 2003 63 Suppl 84 207 210 10.1046/j.1523-1755.63.s84.4.x
Baigent C Burbury K Wheeler D Premature cardiovascular disease in chronic renal failure Lancet 2000 356 147 152 10963260 10.1016/S0140-6736(00)02456-9
Liu Y Coresh J Eustace JA Longenecker JC Jaar B Fink NE Tracy RP Powe NR Klag MJ Association between cholesterol level and mortality in dialysis patients: role of inflammation and malnutrition JAMA 2004 291 451 459 14747502 10.1001/jama.291.4.451
Seliger SL Weiss NS Gillen DL Kestenbaum B Ball A Sherrard DJ Stehman-Breen CO HMG-CoA reductase inhibitors are associated with reduced mortality in ESRD patients Kidney Int 2002 61 297 304 11786112 10.1046/j.1523-1755.2002.00109.x
Kidney Disease Outcomes Quality Initiative Managing dyslipidemias in chronic kidney disease Am J Kidney Dis 2003 41 Suppl 3 S1 S91 12751048
Wanner C Krane V Marz W Olschewski M Asmus HG Kramer W Kuhn KW Kutemeyer H Mann JF Ruf G Ritz E Deutsche Diabetes-Dialyse-Studie (4D) Study Group Randomized controlled trial on the efficacy and safety of atorvastatin in patients with type 2 diabetes on hemodialysis (4D study): demographic and baseline characteristics Kidney Blood Press Res 2004 27 259 266 15316128 10.1159/000080241
Wanner C Results of a randomized controlled trial with atorvastatin in dialyzed diabetic patients (4D trial) Presented at the annual meeting of the American Society of Nephrology St Louis, Missouri, USA 29 October 2004
Blasetto JW Stein EA Brown WV Chitra R Raza A Efficacy of rosuvastatin compared with other statins at selected starting doses in hypercholesterolemic patients and in special population groups Am J Cardiol 2003 91 3C 10C 12646336 10.1016/S0002-9149(03)00003-1
Chapman MJ Caslake M Packard C McTaggart F New dimension of statin action on apoB atherogenicity Clin Cardiol 2003 26 Suppl I I7 I10 12539816
Brewer HB Jr Benefit-risk assessment of rosuvastatin 10 to 40 milligrams Am J Cardiol 2003 92 Suppl 2 23 29 10.1016/S0002-9149(03)00779-3
Cox D Regression models and life tablets (with discussion) J Royal Stat Soc 1972 74 187 220
Austin MA Breslow JL Hennekens CH Buring JE Willett WC Krauss RM Low-density lipoprotein subclass patterns and risk of myocardial infarction JAMA 1988 260 1917 1921 3418853 10.1001/jama.260.13.1917
Campos H Genest JJ JrBlijlevens E McNamara JR Jenner JL Ordovas JM Wilson PW Schaefer EJ Low density lipoprotein particle size and coronary artery disease Arterioscler Thromb 1992 12 187 195 1543692
Lamarche B Tchernof A Moorjani S Cantin B Dagenais GR Lupien PJ Després JP Small, dense low-density lipoprotein particles as a predictor of the risk of ischemic heart disease in men. Prospective results from the Québec Cardiovascular Study Circulation 1997 95 69 75 8994419
Sacks FM Alaupovic P Moye LA Cole TG Sussex B Stamfer MJ Pfeffer MA Braunwald E VLDL, apolipoproteins B, CIII, and E, and risk of recurrent coronary events in the Cholesterol Recurrent Events (CARE) trial Circulation 2000 102 1886 1892 11034934
O'Neal D Lee P Murphy B Best J Low-density lipoprotein particle size distribution in end-stage renal disease treated with hemodialysis or peritoneal dialysis Am J Kidney Dis 1996 27 84 91 8546142
Königer M Quaschning T Wanner C Schollmeyer P Krämer-Guth A Abnormalities in lipoprotein metabolism in hemodialysis patients Kidney Int 1999 56 Suppl 71 S248 S250 10.1046/j.1523-1755.1999.07166.x
Deighan CJ Caslake MJ McConnell M Boulton-Jones JM Packard CJ Atherogenic lipoprotein phenotype in end-stage renal failure: origin and extent of small dense low-density lipoprotein formation Am J Kidney Dis 2000 35 852 862 10793019
Wanner C Quaschning T Dyslipidemia and renal disease: pathogenesis and clinical consequences Curr Opin Nephrol Hypertens 2001 10 195 201 11224694
Fytili CI Progia EG Panagoustsos SA Thodis ED Passadakis PS Sombolos KI Vargemezis VA Lipoprotein abnormalities in hemodialysis and continuous ambulatory peritoneal dialysis patients Ren Fail 2002 24 623 630 12380908 10.1081/JDI-120013966
Wanner C Horl WH Luley CH Wieland H Effects of HMG-CoA reductase inhibitors in hypercholesterolemic patients on hemodialysis Kidney Int 1991 39 754 760 2051734
Nishizawa Y Shoji T Emoto M Kawasaki K Konishi T Tabata T Inoue T Morii H Reduction of intermediate density lipoprotein by pravastatin in hemo- and peritoneal dialysis patients Clin Nephrol 1995 43 268 277 7606882
Nishizawa Y Shoji T Tabata T Inoue T Morii H Effects of lipid-lowering drugs on intermediate-density lipoprotein in uremic patients Kidney Int 1999 56 Suppl 71 S134 S136 10.1046/j.1523-1755.1999.07133.x
Nishikawa O Mune M Miyano M Nishide T Maeda A Kimura K Takahashi T Kishino M Tone Y Otani H Ogawa A Maeda T Yukawa S Effect of simvastatin on the lipid profile of hemodialysis patients Kidney Int 1999 56 Suppl 71 S219 S221 10.1046/j.1523-1755.1999.07157.x
Saltissi D Morgan C Rogby RJ Westhuyzen J Safety and efficacy of simvastatin in hypercholesterolemic patients undergoing chronic renal dialysis Am J Kidney Dis 2002 39 283 290 11840368
Holdaas H Fellström B Jardine AG Holme I Nyberg G Fauchald P Gronhagen-Riska C Madsen S Neumayer HH Cole E Maes B Ambuhl P Olsson AG Hartmann A Solbu DO Pedersen TR Assessment of LEscol in Renal Transplantation (ALERT) Study Investigators Effect of fluvastatin on cardiac outcomes in renal transplant recipients: a multicentre, randomised, placebo-controlled trial Lancet 2003 361 2024 2031 12814712 10.1016/S0140-6736(03)13638-0
Jones PH Davidson MH Stein EA Bays HE McKenney JM Miller E Cain VA Blasetto JW Comparison of the efficacy and safety of rosuvastatin versus atorvasatatin, simvastatin, and pravastatin across doses (STELLAR Trial) Am J Cardiol 2003 92 152 160 12860216 10.1016/S0002-9149(03)00530-7
Olsson AG McTaggart F Raza A Rosuvastatin: a highly effective new HMG-CoA reductase inhibitor Cardiovasc Drug Rev 2002 20 303 328 12481202
Locatelli F Canaud B Eckardt K-U Stenvinkel P Wanner C Zoccali C Oxidative stress in end-stage renal disease: an emerging threat to patient outcome Nephrol Dial Transplant 2003 18 1272 1280 12808161 10.1093/ndt/gfg074
Annuk M Zilmer M Lind L Linde T Fellstrom B Oxidative stress and endothelial function in chronic renal failure J Am Soc Nephrol 2001 12 2747 2752 11729244
Rosenson RS Statins in atherosclerosis: lipid-lowering agents with antioxidant capabilities Atherosclerosis 2004 173 1 12 15177118 10.1016/S0021-9150(03)00239-9
Annuk M Lind L Linde T Fellström B Impaired endothelium-dependent vasodilation in renal failure in humans Nephrol Dial Transplant 2001 16 302 306 11158404 10.1093/ndt/16.2.302
Stalker TJ Lefer AM Scalia R A new HMG-Co A reductase inhibitor, rosuvastatin, exerts anti-inflammatory effects on the microvascular endothelium: the role of mevalonic acid Br J Pharmacol 2001 133 406 412 11375257 10.1038/sj.bjp.0704070
Foley RN Clinical epidemiology of cardiac disease in dialysis patients: ventricular hypertrophy, ischemic heart disease, and cardiac failure Semin Dial 2003 16 111 117 12641874 10.1046/j.1525-139X.2003.160271.x
Su SF Hsiao CL Chu CW Lee BC Lee TM Effects of pravastatin on left ventricular mass in patients with hyperlipidaemia and essential hypertension Am J Cardiol 2000 86 514 518 11009268 10.1016/S0002-9149(00)01004-3
Kjekshus J Pederson TR Olsson AG Faergeman O Pyörälä L The effects of simvastatin on the incidence of heart failure in patients with coronary heart disease J Card Fail 1997 3 249 254 9547437 10.1016/S1071-9164(97)90022-1
| 15910680 | PMC1175096 | CC BY | 2021-01-04 16:47:32 | no | Curr Control Trials Cardiovasc Med. 2005 May 23; 6(1):9 | utf-8 | Curr Control Trials Cardiovasc Med | 2,005 | 10.1186/1468-6708-6-9 | oa_comm |
==== Front
Epidemiol Perspect InnovEpidemiologic perspectives & innovations : EP+I1742-5573BioMed Central London 1742-5573-2-31591671010.1186/1742-5573-2-3Analytic PerspectiveHistorical Perspective: S. Leonard Syme's influence on the development of social epidemiology and where we go from there Yen Irene H [email protected] University of California, San Francisco, USA2005 25 5 2005 2 3 3 18 4 2005 25 5 2005 Copyright © 2005 Yen; licensee BioMed Central Ltd.2005Yen; licensee BioMed Central Ltd.This is an Open Access article distributed under the 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 article accompanies Len Syme's "Historical Perspective: The social determinants of disease – some roots of the movement." It describes some of Len's role in the development of social epidemiology through mentoring investigators and influencing training programs. It also discusses some challenges for the field and ways to move forward.
==== Body
Introduction
In "Historical Perspective: The social determinants of disease – some roots of the movement," Professor S. Leonard (Len) Syme describes his professional activities as they related to the emergence of research on the social determinants of health within the field of epidemiology [1]. He writes in some detail about his work at the U.S. National Institutes of Health (NIH) starting shortly after completing his doctorate in medical sociology through to about 1980, by which time he had been a professor at the School of Public Health of the University of California (UC), Berkeley for over a decade.
In this historical perspective, I want to highlight Len's contributions to the field of social epidemiology, with a particular focus on mentoring and training programs. I focus on training because of Len's enormous contribution as a teacher and mentor. Because there are so many of us who have benefited from Len's mentoring, I feel it was important to include other voices and have solicited reflections from colleagues, advisees, and other epidemiologists who are familiar with Len's scholarship. Social epidemiology is maturing as a discipline and there is more work to be done. I thought this essay would be a good place to reflect on some challenges that we face and what we might do to overcome them.
Analysis
Research on social determinants – cultivating investigators
I frequently hear people marvel on what an impressive list of names marks Len's career as a professor, both in the quality of the work they are pursuing and the sheer quantity of people. Nancy Adler, professor of medical psychology at UC San Francisco and director of the MacArthur Research Network on Socioeconomic Status and Health, travels all around the U.S. attending meetings on the social determinants of health and health disparities. She told me, "Everywhere I go, when people go around the table to introduce themselves and say how they got there, there are always several people in the room who say, 'I'm here because Len Syme was my mentor.' Len is like the universal blood donor." (Nancy Adler, personal communication, March 8, 2005.)
During the period of the primary focus of Len's essay (1968–1980), he advised Lisa Berkman (chair of the Department of Society, Human Development, and Health, Harvard School of Public Health), Robert Hiatt (Director of Population Sciences, UCSF Comprehensive Cancer Center), George Kaplan (Director, Center for Social Epidemiology and Population Health, University of Michigan), and Michael Marmot (Professor of Epidemiology, University College London and Chair of the World Health Organization Commission on the Social Determinants of Health) among others. Since 1980, he was the major thesis advisor to Jack Guralnik, Mary Haan, Nancy Krieger, John Lynch, Teresa Seeman, Marilyn Winkleby, and many others who are doing important work, though perhaps not in social epidemiology nor necessarily in an academic setting.
I met Len in 1988 when I entered the master's degree program in epidemiology and biostatistics at UC Berkeley School of Public Health. I was a member of the third class of students admitted for a two-year master's degree program. Prior to 1986, Berkeley had only offered a one-year master's degree curriculum in epidemiology for people who had prior graduate training, mostly medical doctors. For the two-year program, we were required to attend a core seminar. The core seminar gave us an overview of epidemiology as a discipline (including historical readings), key concepts, a chance to learn about individual faculty research emphasis, and a place for discussion. The faculty in charge rotated; our year, it was Len's turn. I completed the master's degree and took a job with the Centers for Disease Control and Prevention.
I returned to Berkeley in 1993 to pursue a doctorate. Len had already been a strong influence on my thinking as a budding epidemiologist having been the lead faculty for the core seminar during the master's degree program. Len introduced key epidemiological concepts to me, such as a basic definition of epidemiology, "epidemiology is the study of the distribution of diseases in populations." As he had been doing since he started teaching epidemiology, he had all of us read Durkheim and vividly emphasized Durkheim's contribution. I can still hear Len's voice highlighting Durkheim's work, showing the most personal of all acts (suicide) varied by region over a period of several decades. Len would stress the significance of the finding by asking provocative questions. How could the rates be so stable over time since clearly the people involved were different? Always an accomplished story teller, Len would unfold the mystery explaining how Durkheim had found that these regions differed by dominant religion, the Catholic regions had different rates of suicide from the Protestant. Ultimately Durkheim proposed the concept of anomie. Alongside Durkheim's work, Len talked about the futility of behavioral interventions, pointing out that if we only target behaviors (e.g. smoking, exercise, diet) without targeting root causes, we will have to repeat our work in cohort after cohort of populations. These sorts of stories led me to want to focus on the root factors that underlie the behaviors and eventually I chose to focus on neighborhood environment. Len was my major dissertation advisor.
Len's students appreciate his personalized attention, challenging intellectual arguments, and ongoing support regardless of professional choices. Before taking his current position at UCSF, Robert Hiatt was Deputy Director, Division of Cancer Control & Population Sciences, National Cancer Institute. He reflects, "In an unplanned way Len has been a critical influence on me throughout my career. First, by introducing me to psychosocial perspectives in epidemiology while in training, then by directing me into community intervention research, and finally through indirect, but very real, ways by influencing my work at the National Cancer Institute where I generated research programs in population health." (Robert Hiatt, personal communication, 2005). Here is another reflection from Nancy Krieger, Department of Society, Human Development, and Health, Harvard School of Public Health:
Len Syme was and continues to be a critical mentor for me, in the very best sense of the term. Len consistently has encouraged critical thought and fresh examination of long-standing problems of social inequalities in health. He likewise has always emphasized the relevance of social analysis for understanding public health problems, recognizing that investigating population health requires population level theories. The challenge thus becomes integrating population level social and biological thinking, and Len never once suggested to me this was a waste of time! In fact, when he was my doctoral advisor, he supported strongly my early work on thinking through the intertwined social and biological determinants of breast cancer across the lifecourse, just as he also supported strongly my work on using theories of social class to inform measurement of social inequalities in health and theories of race relation to inform research on how racism harms health. To Len, epidemiology was first and foremost a science of understanding the population distributions of disease, not simply a set of methods to apply to quantitative data that happened to be about health – and it is to his credit that this orientation not only is strong but growing among the many he has taught and influenced. (Nancy Krieger, personal communication, May 4, 2005)
From Lisa Berkman, Chair of the Department of Society, Human Development, and Health, Harvard School of Public Health:
At the time I was thinking about my dissertation, there were only a few studies out there about how social conditions might influence health. The data from Japan was of course very influential. At the time I was also studying social networks in the sociology department with Claude Fisher. Len's perspective was invaluable. He has always encouraged his students to follow their noses and be critical thinkers. His mentoring was invaluable, he questioned and questioned until I finally had a coherent story to tell. (Lisa Berkman, personal communication, May 9, 2005)
Research on social determinants – Explicit training programs
My classmates in the master's degree program at Berkeley noticed that none of the professors in the epidemiology portion of our program had doctoral degrees in epidemiology, but rather were trained in other disciplines, primarily medicine. Our cohort's disciplinary pasts were even more varied including people who had majored in history, creative writing, environmental studies, political science, and anthropology. I believe that the influx into epidemiology of people from outside of medicine has contributed to the widespread interest in non-clinical non-behavioral factors such as discrimination, social capital, and income inequality. This increased interest is leading to formal training programs in the social determinants of health or population health.
Beginning in 1991, the Harvard School of Public Health established the Health and Social Behavior Department. (In 2003, this Department was re-organized together with the Maternal and Child Health department into the department of Society, Human Development, and Health.) At the time, it was one of the first of its kind, with a stated mission "to identify the social and behavioral determinants of health and to develop and evaluate interventions and policies leading to the improvement of the public's health and quality of life"[2].
More social determinants or population health masters and doctoral training programs are emerging. UC Berkeley began a Health and Social Behavior Masters program in 1999. The University of South Florida approved a socio-health sciences MPH in 2004. Also in 2004, University College London established a MSc program in Health and Society: Social Epidemiology.
Multi-discipline approaches to research have been in vogue of late. Len mentions the Robert Wood Johnson Health and Society Scholars (HSS) postdoctoral training program. HSS had its first cohort of scholars in 2003. Pamela Russo, senior program officer in the Health Group of the Robert Wood Johnson (RWJ) Foundation, credits Len with laying the groundwork for the creation of such a program:
I would say that Len'sthinking, mentoring, and approach were seminal in the evolution of transforming the older narrow medical and biological paradigm for epidemiology to a much broader range of determinants of health captured in the population health approach. ... He personally influenced or mentored many of the current H&SS faculty and site directors – including Lisa Berkman, George Kaplan and others. This program would not have developed if not for Len and his mentees having built the foundations of the field. He broadened the horizon of what influences health status. He also has been a powerful champion of research and interventions to reduce health disparities between populations throughout his career. (Pamela Russo, personal communication, March 11, 2005)
Readers may also be familiar with the Kellogg Scholars in Health Disparities which began in 2001 and includes a focus on social determinants. The director of this program, Barbara Krimgold, similarly recognizes Len's strong influence on the structure, emphasis of multiple disciplines, and inclusion of policy perspectives in the program (Barbara Krimgold, personal communication, March 16, 2005). Len continues mentoring fellows in the RWJ HSS and Kellogg programs who are at the UC Berkeley/UC San Francisco joint training sites.
Research on social determinants – institutional structures
As mentioned above, in his teaching Len consistently emphasizes the population health and prevention themes to public health and the concept that epidemiology was supposed to be the core scientific method serving public health. He stresses that even though epidemiology rhetoric was about investigating ways to prevent disease, more often the field was constrained by the medical model of dividing the body into organ systems. Looking at NIH during the period that Len's essay focuses on (1968–1980), one can see that the institutes were predominantly concerned with parts of the body or diseases. For example, the National Heart, Lung, and Blood Institute, National Institute of Diabetes & Digestive & Kidney Diseases, National Cancer Institute, and the National Institute of Allergy and Infectious Diseases are four of the six institutes established in the 1940s. The Office of Behavioral and Social Sciences Research (OBSSR) was established in 1995. The first director of the OBSSR was Dr. Norman Anderson. Dr. Anderson describes some of Len's more recent influence on the NIH research agenda:
Len's research, and that which followed it, was the basis of the first ever NIH-wide conference on social determinants of health and illness, entitled, "Toward Higher Levels of Analysis: Promise and Progress on Social Aspects of Illness", which was sponsored by OBSSR. The conference led to a research agenda for the funding of social science research at NIH. So the foundation that Len created was critical to the infusion of a social science perspective at NIH. (Norman Anderson, personal communication, March 15, 2005)
The conference that Dr. Anderson refers to was held in 2000, the same year that the National Center for Minority Health and Health Disparities was established.
U.S. funding opportunities for research on the social determinants of health are heavily influenced by the NIH agenda. In 1999, the National Institute for Environment Health Sciences (NIEHS) spear-headed research on health disparities, first by hosting a series of workshops around the country which academic researchers, government agency staff, and community organization staff attended to discuss priorities. One result was the November 1999 release of the RFA, "Health disparities: linking biological and behavioral mechanisms with social and physical environments." According to Dr. Frederick Tyson, one of the organizers of the workshops and authors of the RFA, Len's "work on how socioeconomic factors influence health outcomes is a major foundation to the RFA." (Tyson, personal communication, March 10, 2005).
Len has also been influential outside the US. In the mid 1980s, Dr. Fraser Mustard (former Vice President of the Faculty of Health Sciences of McMaster University and first President of the Canadian Institute of Advanced Research (CIAR)) was investigating the underlying causes of heart disease in order to prevent it. He was referred to Len because of the MRFIT Study. Apparently, Len argued that not enough was known about the causes to prevent heart disease. Their discussions led Dr. Mustard to invite Len to serve on the advisory committee of the CIAR Population Health Program, which started in 1987. (John Frank, personal communication, March 24, 2005). In part, as an outgrowth of the work of the Population Health Program, in 2001, The Canadian Institutes of Health Research, the Canadian equivalent of the U.S. National Institutes of Health, established the Institute of Population and Public Health (IPPH). Len is a member of the IPPH advisory board and according to the scientific director of the IPPH, during these four years, "he has materially advanced the Institute's development by his unwavering commitment to excellence in research, and his tenacious support for bold investments that can 'make a difference' in building the field of population and public health research in Canada." (John Frank, personal communication, March 21, 2005).
Research on social determinants – general influence on the field
Aside from directly influencing the training and funding programs mentioned above, social epidemiology was strongly influenced by Len's research. Bruce Dohrenwend, Professor of Epidemiology at Columbia Mailman School of Public Health, reacts to Len's essay:
He was so modest in describing his own achievements and so generous with his praise of others that his own major contributions might be missed by a novice reader. One of these, especially, should be highlighted. This is his leadership role in investigating a series of imaginative hypotheses developed to explain the stunning finding that Japanese men who had migrated to California had rates of coronary heart disease that were five times higher than in their counterparts in Japan. This work is a major achievement, and Len was at the center of it. (Bruce P. Dohrenwend, personal communication, March 28, 2005.)
Not surprisingly, Len's work has reached far. Töres Theorell, director of the National Institute for Psychosocial Factors and Health and Professor of Psychosocial Medicine at the Karolinska Institute in Stockholm, Sweden wrote:
I am of course not the only one here who has been influenced by Len Syme. In particular, social support was an extremely important subject for several researchers here, for instance Per Olof Ostergen in Malmo and Kristina Orth Gomer in Stockholm. So if you look at the reference list of these people, you will find Len Syme's name. Len Syme wrote about our book Healthy Work (Karasek and Theorell 1990) in several places and has been driving the idea that 'control in life' would be an important research dimension. ... I am a great admirer of Len, whose constructive ideas and engaged helpfulness has always been a star. (Töres Theorell, personal communication, March 26, 2005.)
Conclusion
At the end of Len's essay, he reflects that while a tremendous amount of progress has been made, the advances have been met with opposition and suspicion. As someone who is trying to participate and make contributions in the field of social epidemiology, I can attest to the resistance and skepticism. In fact, the idea of "social epidemiology" as a subfield was recently debated in a set of commentaries in the International Journal of Epidemiology [3-5].
There are two steps that we can take to build the credibility and contributions of social epidemiology. One of them Len brings up in his essay, the pressing need to bring social theory into the picture. Social theory should be a necessary piece of the conceptualization of research questions by more people who are engaged in research on social factors or population health and more fully incorporated into our training programs.
The need for deeper understanding and application of social theory has already been taken up by some people in the field[4-9]. Others have made strong arguments for the importance of incorporating theory into public health practice more broadly[10,11]. At the risk of oversimplifying, here is a brief overview of the crux of these arguments: Investigating the social determinants of health has turned our attention to the role of income or wealth, race/ethnicity, and education. Instead of including these variables in our multivariate models as covariates, they could be the central variables. Income/wealth, race/ethnicity, and education are the product of social processes and their meanings cannot be measured simply by individual self-report. Social theories provide containers or frameworks with which to understand the social processes and hypothesize about how and why they are relevant to health. A basic analytic theme in social theory is the structure-agency duality[12]. Structures are social institutions such as the family, political institutions, and economic relations. Agency refers to an individual's capacity to act deliberately or to exercise power[13]. Social theories explain how the tension between structures and agency play out. Incorporating and applying theories can inform and strengthen our research questions.
The second step that is tied into the incorporation of social theory is the need for social epidemiology to have a practice-based anchor. In other words, we need grounding in everyday activities. In this way, the field would gain a sorely needed practice perspective and the accompanying legitimacy and credibility. Recent discussions which press for more participatory research are one means to bring in this element [5]. By including the voices of the people who are most directly affected by the processes under study and by holding close the end goal of designing programs, the research will gain more solid grounding. Another way to think about the practice-research link for social epidemiologists is to look to colleagues in other subfields within epidemiology. For example, clinical epidemiologists work closely with physicians and nurses. Their research findings can directly contribute to clinical practice. Clearly both sides are enriched by the contributions of the other. Social epidemiology as a discipline must establish formal ties to practice-based fields, such as city and regional planning, public policy, and education, in order to inform and be informed.
Social epidemiology owes Len Syme a lot. We can repay him by asking tough questions, holding ourselves to high standards, and serving as mentors.
Abbreviations
CIAR – Canadian Institute of Advanced Research
HSS – Health and Society Scholars
MPH – Masters in Public Health
MRFIT – Multiple Risk Factor Intervention Trial
MSc – Masters in Sciences
NIH – National Institutes of Health
OBSSR – Office of Behavioral and Social Sciences Research
RFA – Request for Applications
RWJ – Robert Wood Johnson
UC – University of California
UCSF – University of California, San Francisco
==== Refs
Syme SL Historical perspective: The social determinants of disease – some roots of the movement Epidemiologic Perspectives & Innovations 2005 2 2 15840175 10.1186/1742-5573-2-2
Accessed March 16, 2005
Zielhuis GA Kiemeney LALM Social epidemiology? No way Int J Epidemiol 2001 30 43 44 11171849 10.1093/ije/30.1.43
Krieger N Commentary: Society, biology and the logic of social epidemiology Int J Epidemiol 2001 30 44 46 11171850 10.1093/ije/30.1.44
Macdonald KI Commentary: Social epidemiology. A way? Int J Epidemiol 2001 30 46 47 11171851 10.1093/ije/30.1.46
Krieger N Theories for social epidemiology in the 21st century: an eco-social perspective Int J Epidemiol 2001 30 668 77 11511581 10.1093/ije/30.4.668
Frohlich KL Corin E Potvin A theoretical proposal for the relationship between context and disease Sociology of Health & Illness 2001 23 776 797 10.1111/1467-9566.00275
Muntañer C Invited commentary: Social mechanisms, race, and social epidemiology Amer J Epidemiol 1999 150 121 6 10412956
Cooper RS Kaufman JS Is there an absence of theory in social epidemiology? The authors respond to Muntañer Amer J Epidemiol 1999 150 127 8
Frohlich KL Mykhalovsky E Miller F Daniel M Advancing the population health agenda: Encouraging the integration of social theory into population health research and practice Can J Public Health 2004 95 392 395 15490933
Potvin L Gendron S Bilodeau A Chabot P Integrating social theory into public health practice Am J Pub Health 2005 95 591 595 15798114 10.2105/AJPH.2004.048017
Szeter S Woolcock M Rejoinder: Crafting rigorous and relevant social theory for public health policy Int J Epidemiol 2004 33 700 704 10.1093/ije/dyh263
Giddens A Central Problems in Social Theory 1979 London: Macmillian Press 53 76
Leung MW Yen IH Minkler M Community based participatory research: a promising approach for increasing epidemiology's relevance in the 21st century Int J Epidemiol 2004 33 499 506 15155709 10.1093/ije/dyh010
| 15916710 | PMC1175097 | CC BY | 2021-01-04 16:36:38 | no | Epidemiol Perspect Innov. 2005 May 25; 2:3 | utf-8 | Epidemiol Perspect Innov | 2,005 | 10.1186/1742-5573-2-3 | oa_comm |
==== Front
Health Qual Life OutcomesHealth and Quality of Life Outcomes1477-7525BioMed Central London 1477-7525-3-371592705410.1186/1477-7525-3-37ResearchOrder effects: a randomised study of three major cancer-specific quality of life instruments Cheung Yin-Bun [email protected] Celestine [email protected] Cynthia [email protected] Julian [email protected] Joseph [email protected] MRC Tropical Epidemiology Group, London School of Hygiene and Tropical Medicine, London WC1E 7HT, UK2 Clinical Trials and Epidemiological Sciences, National Cancer Centre, 11 Hospital Drive, 169610, Singapore3 Department of Palliative Medicine, National Cancer Centre, 11 Hospital Drive, 169610, Singapore 4 Department of Rheumatology and Immunology, Singapore General Hospital, Outram Road, 169608, Singapore2005 31 5 2005 3 37 37 25 4 2005 31 5 2005 Copyright © 2005 Cheung et al; licensee BioMed Central Ltd.2005Cheung 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 methodological studies and outcomes research, questionnaires often comprise several health-related quality of life (HRQoL) measures. Previous psychological studies have suggested that changing the sequential order of measurement scales within a questionnaire could alter the pattern of responses. Yet, information on the presence or absence of order effects on the assessment of HRQoL in cancer patients is limited.
Methods
An incomplete block design was used in this study of 1277 cancer patients. Each patient filled out a questionnaire package that contained two of the three major cancer-specific HRQoL instruments, namely the Functional Assessment of Cancer Therapy – General, the European Organization for the Research and Treatment of Cancer Core Quality of Life Questionnaire and the Functional Living Index – Cancer. Within a questionnaire package the sequential order of the instruments contained were randomised. Measurement properties of the instruments, including the number of missing values, mean HRQoL scores, known-groups validity and internal consistency were compared between samples of different presentation orders.
Results
No effect of presentation order on the four properties aforementioned was found.
Conclusion
Presentation order is unlikely to alter the responses to these HRQoL instruments administered in cancer patients when any two of them are used together.
health-related quality of lifecancerorder effect
==== Body
Background
The order of questions in an interview may affect the responses to each question [1-3]. Conventional wisdom suggests that surveys should begin with simple, descriptive and non-sensitive questions [2,3]. The items used in composite measurement scales may also be subjected to a context effect [4]. Yet, there has been limited information in the area of quality of life research to confirm if the presentation ordering of composite measurement scales within a questionnaire would alter the results.
Jensen et al. [5] and Mook [6] discussed various reasons why order effects could appear. For instance, respondents may experience fatigue or lose concentration towards the end of a questionnaire and as a result, the probability of misinterpretation and omission of items may increase. According to this view, the strength of order effects is related to the length of the questionnaire. Moreover, respondents may produce different patterns of responses as the previous questionnaires desensitise or familiarize them with a topic.
The development of new health-related quality of life (HRQoL) instruments frequently employs multiple instruments in order to determine convergent and divergent validity. It is uncertain whether the validity or other measurement properties of an instrument could be affected by the presentation order. Furthermore, the possibility of an order effect points to the need for caution during the comparison of information across studies in which the HRQoL measurement scales are not presented in the same order, even if the questions are identical. It is therefore important to determine or prevent order effects in such situations.
Using randomised and counterbalanced designs, Jensen et al. [5] and Lucas [7] demonstrated that the presentation order of some psychological instruments had an impact on the scores. In a postal survey that included four health and HRQoL measurement scales, the researchers used two versions of a questionnaire [8]. One began with generic measures followed by specific measures; another presented the constituent groups of items in chronological order according to the time period the items referred to. They found that the questionnaires using chronological order were returned more promptly although the presentation order did not appear to affect the answers. One study investigated this issue in the assessment of the quality of life of cancer patients [9]. The participants self-administered a questionnaire in which the European Organisation for Research and Treatment of Cancer Core Quality of Life Questionnaire (EORTC QLQ-C30) preceded the Functional Assessment of Cancer Therapy – General (FACT-G). The two instruments contained similar questions on four aspects of HRQoL, namely, pain, nausea, meeting family needs, and general satisfaction. The investigators found that the four questions in the two instruments indicated similar level of HRQoL even though the patients had been exposed to the EORTC QLQ-C30 questions before they answered the similar FACT-G questions.
In a recent study, the FACT-G and Quick-FLIC (an abbreviated version of the Functional Living Index – Cancer, FLIC [10]) were used [11]. Alternating sequencing of these two HRQoL instruments were carried out to form two different questionnaire packages. The study showed that there was no major effect of presentation order on the mean scores, amount of missing values, and known-groups validity and internal consistency of the instruments. The inadequacies of the study were that it used a relatively short questionnaire as the mean time to complete was only 15.0 minutes; it involved only two HRQoL instruments; and the sample size was relatively small.
The present study aimed to verify the previous findings about the lack of order effects in the assessment of cancer patients' quality of life. It used a larger sample size and longer questionnaires that involved three major HRQoL instruments commonly used in oncology.
Methods
Design
This study used an incomplete block design [12], in which participants were randomised to receive one of the following six questionnaire packages (in this order of presentation): (1) EORTC QLQ-C30 and FACT-G, (2) FACT-G and EORTC QLQ-C30, (3) EORTC QLQ-C30 and FLIC, (4) FLIC and EORTC QLQ-C30, (5) FACT-G and FLIC, and (6) FLIC and FACT-G. We chose against using a complete block design of having each patient complete all three questionnaires because past experiences suggested that some patients might be unable or unwilling to spend so much time and concentration on it. In the current study, the mean time taken to complete the interview was 20.4 minutes, but the 90th percentile was 39 minutes. Due to logistic considerations, the randomisation used days rather than individuals as units and assigned the six packages in blocks of six days [13]. In the examination of order effects on FACT-G, for instance, the FACT-G data from packages (2) and (5), where FACT-G was administered first, were compared against those from packages (1) and (6), where FACT-G was administered last. For brevity, we used the phrases order A and order B to mean an HRQoL instrument was administered first and last, respectively.
Patient recruitment
Patients were recruited from the National Cancer Centre, Singapore, which serves about 70% of the cancer patients seen by the public sector of the country, from September 2003 to May 2004. The study was approved by the Ethics Committee of the Centre. Patients were approached while they were in the waiting areas of the specialist outpatient clinics, ambulatory treatment unit and the therapeutic radiology department of the Centre. The inclusion criteria were: literate in English or Chinese, aged 18 years or older, and agreeable to give written informed consent. The patients were heterogeneous in clinical profiles, such as having different types of tumour, and were fitting for the study of the three instruments that were designed for application to all cancer patients.
Singapore is a multi-ethnic society with the Chinese forming about 70% of the total population. The Chinese participants had the option of answering either an English or a Chinese questionnaire according to their lingual preference, whereas other participants answered an English questionnaire. Participants were requested to self-administer the questionnaire packages (where possible). Upon request by the patients, interviews would be administered by one of the two research coordinators of the project.
Instruments
The FACT-G version 4 and EORTC QLQ-C30 version 3 were used. The FLIC had been modified in two aspects for use in Singapore [14,15]. Firstly, the word "cancer" was removed from the questions because some patients, particularly the older patients, might be unaware of their diagnosis and sometimes their families might not want them told the diagnosis. In this regard, it is of note that the FACT-G and EORTC QLQ-C30 do not mention the word cancer. Secondly, the visual analogue scale was difficult to some patients, especially the older and less educated. It was replaced by a seven-point Likert format scale. Similar modifications of the FLIC have also been reported in other countries [16,17].
The questionnaire packages each began with a page of demographic and health questions on information such as Eastern Cooperative Oncology Group (ECOG) performance status [18] and whether the patients were on chemotherapy and/or radiotherapy. Treatment status was classified as whether the patient was on chemotherapy and/or radiotherapy or not (yes or no).
Statistical considerations
All HRQoL items were recoded such that a higher score reflects a better quality of life. Missing values in the FACT-G, FLIC and EORTC QLQ-C30 were imputed by the half-rule [19]. ANOVA and Chi-square tests were used to compare continuous and categorical variables, respectively, between patients who answered the six questionnaire packages. Fisher's exact test was used to compare the number of missing values in each instrument between orders of presentation. Negative binomial regression was used to estimate the difference in mean number of missing values between presentation orders and the confidence interval (CI) [20]; linear regression was used for HRQoL scores. Cronbach's alpha was calculated for each HRQoL instrument in each order of presentation. There is no established analytic procedure for the estimation of CI for the difference in Cronbach's alpha. We employed the bootstrapping method, with 1000 replications [21].
In line with commonly accepted practice for equivalence studies, 90% confidence intervals (CI) were estimated [22,23]. Equivalence was declared if the 90% CI fell totally within an equivalence zone. For the comparison of mean number of missing values, the equivalence zone was pre-defined as ± 1 item. For the comparison of Cronbach's alpha, the zone was ± 0.1.
There is no consensus to the definition of equivalent HRQoL scores. Using various clinical criteria, Cella et al. [24] suggested that the minimal clinically significant difference on the FACT-G scale is 4 points. Based on the assessment of subjective significance, Osoba et al. [25] suggested that "a little" change on the EORTC QLQ-C30 global quality of life scales was approximately 5 to 10 points, on a scale of 0 to 100. Interestingly, both studies approximately agreed with Cohen's [26] suggestion that an effect size between 0.2 to less than 0.5 standard deviation (SD) is small. In the present data set, 4 points of the FACT-G score and 5 points of the EORTC global functioning score are equivalent to 0.25 and 0.23 of their SD's. Therefore we defined an equivalence margin as ± 0.25 SD, rounded to the nearest integer. It corresponded to 4, 6 and 5 points for the FACT-G, FLIC and EORTC QLQ-C30, respectively. Furthermore, we defined a "small difference" margin as ± 0.5 SD. This took into account Osoba et al. [25] about a little change (10 points). The small difference margins for the FACT-G, FLIC and EORTC QLQ-C30 were 8, 12 and 10 points, respectively.
The main analyses did not adjust for covariates. Supplementary analyses adjusted for covariates shown in table 1 using the multiple regression analysis approach.
Table 1 Respondent characteristics by questionnaire package (N = 1277)
Variable Questionnaire Package
EORTC QLQ-C30 + FACT-G (N = 200) FACT-G + EORTC QLQ-C30(N = 240) EORTC QLQ-C30 + FLIC (N = 233) FLIC + EORTC QLQ-C30 (N = 188) FACT-G + FLIC (N = 215) FLIC + FACT-G (N = 201) p-value(a)
Age Mean (SD) 51.0 (12.0) 51.6 (12.5) 50.6 (9.9) 51.0 (10.3) 51.4 (10.5) 51.4 (11.6) 0.933
Gender Male 43.5% 40.8% 36.1% 41.0% 43.7% 38.3% 0.547
Female 56.5% 59.2% 64.0% 59.0% 56.3% 61.7%
Race Chinese 88.0% 91.3% 90.6% 91.0% 90.7% 89.1% 0.446
Malay 6.5% 4.6% 3.4% 2.7% 3.7% 4.5%
Indian 4.0% 3.8% 3.0% 4.3% 2.3% 5.5%
Others 1.5% 0.4% 3.0% 2.1% 3.3% 1.0%
Education Primary or below 22.1% 20.4% 21.0% 19.2% 23.8% 21.9% 0.212
Secondary 44.7% 43.8% 49.4% 57.9% 45.3% 47.8%
Post-secondary 33.2% 35.8% 29.6% 22.9% 30.8% 30.4%
ECOG 0–1 75.5% 72.5% 75.1% 71.8% 69.8% 72.0% 0.772
2–4 24.5% 27.5% 24.9% 28.2% 30.2% 28.0%
Treatment Inactive 59.0% 69.6% 59.2% 61.7% 60.9% 62.7% 0.186
Active 41.0% 30.4% 40.8% 38.3% 39.1% 37.3%
Tumor Breast 30.5% 33.8% 40.8% 33.0% 33.0% 34.8% 0.719
Lung 7.5% 8.3% 5.6% 6.4% 9.8% 10.0%
Colo-rectal 14.5% 10.8% 12.0% 13.8% 13.5% 11.9%
Gynaecological 5.0% 5.8% 3.4% 9.0% 4.7% 6.5%
Nasopharyngeal 13.5% 13.3% 12.5% 13.3% 14.4% 16.4%
Head & Neck 8.0% 5.8% 5.2% 5.9% 5.1% 7.00%
Others 21.0% 22.1% 20.6% 18.6% 19.5% 13.4%
Self-administered Yes 76.5% 79.6% 76.8% 75.5% 75.4% 73.1% 0.738
No 23.5% 20.4% 23.2% 24.5% 24.7% 26.9%
Language English 57.5% 56.7% 49.8% 51.1% 60.0% 57.2% 0.219
Chinese 42.5% 43.3% 50.2% 48.9% 40.0% 42.8%
(a) Difference between six questionnaire packages tested by ANOVA for age and Chi-square for categorical variables.
A sample size of 270 per instrument per order of presentation would give a power of 80% and a 5% probability of the type I error for confirming equivalence (± 0.25 SD) between different orders of presentation [22]. The sample size here was about 50% larger because the primary purpose of the study (to compare the variability of the different HRQoL instruments [27]) required it.
Results
A total of 1317 patients consented to participate. Some patients' family members insisted on completing the questionnaire on their behalf. These proxy interviews were excluded. After this exclusion the number of subjects was 1277.
Table 1 provides a descriptive summary of the background characteristics of the patients by questionnaire package. The six groups of patients were similar in clinical and demographic characteristics (each p > 0.10). They were also similar in terms of mode of administration of the questionnaires and the language used (each p > 0.10).
Table 2 shows the number of missing values in the FACT-G, FLIC and EORTC QLQ-C30 by presentation order. The Fisher's exact test showed no significant differences in the distribution of the number of missing values between the two presentation orders A and B for the three instruments (each p > 0.10). The mean number of missing FACT-G item values was 0.03 higher (90% CI = -0.11 to 0.18) among patients who answered the FACT-G first than those who answered the FACT-G second. The corresponding figures for the FLIC and EORTC QLQ-C30 were 0.01 (-0.11 to 0.13) and 0.11 (0.04 to 0.18), respectively. All three confidence intervals totally fell within the pre-defined equivalence zone of ± 1 item. Further analysis using multiple regression analysis to adjust for the covariates shown in table 1 gave similar results. The mean difference (90% CI) between presentation orders in FACT-G, FLIC and EORTC QLQ-C30 missing items were, respectively, 0.04 (-0.10 to 0.18), -0.09 (-0.27 to 0.09) and 0.16 (0.04 to 0.28).
Table 2 Number of missing values in FACT-G, FLIC and EORTC QLQ-C30 by presentation order(a)(b)
FACT-G FLIC EORTC QLQ-C30
Number of missing values Order A (N = 455) Order B (N = 401) Order A (N = 389) Order B (N = 448) Order A (N = 433) Order B (N = 428)
0 207 195 303 345 368 380
1 157 139 47 52 46 40
2 48 36 16 27 12 6
3 20 9 15 16 2 0
4 9 8 4 3 3 2
5 5 4 0 4 1 0
6 4 3 1 1 0 0
7 2 2 1 0 0 0
8 3 2 1 0 0 0
≥9 0 3 1 0 1 0
Mean 0.96 0.93 0.44 0.43 0.25 0.14
p-value (Fisher's exact test) 0.825 0.457 0.308
(a) Order A and order B mean, respectively, the HRQoL instrument was placed first and second in the questionnaire.
(b) The FACT-G comprises 27 items, the FLIC comprises 22, and the EORTC QLQ-C30 comprises 30 items. Therefore the results are not comparable across questionnaires.
Table 3 shows the means and standard deviations of the FACT-G, FLIC and EORTC QLQ-C30 total / global functioning scores according to order of presentation. The means and standard deviations were similar between the two orders across all three instruments. The mean FACT-G score was 2.44 points higher in the interviews where FACT-G was administered first. The 90% CI was 0.66 to 4.24, slightly exceeding the pre-defined equivalence zone of ± 4 points but not exceeding the "small difference" zone. The means of FLIC scores were almost identical in the two presentation orders and the confidence interval totally fell within the pre-defined equivalence zone of ± 6 points (difference = -0.62; 90% CI = -3.20 to 1.97). The mean EORTC QLQ-C30 score was 3.43 points lower in interviews where the EORTC QLQ-C30 was administered first; the confidence interval (-5.87 to -0.99) slightly exceeded the pre-defined equivalence zone of 5 points but not the "small difference" zone. The results after adjustment for the covariates in table 1 were similar. The mean difference (90% CI) between presentation orders in FACT-G, FLIC and EORTC QLQ-C30 scores were, respectively, 2.78 (1.28 to 4.28), 0.28 (-1.77 to 2.33) and -4.16 (-6.27 to -2.05).
Table 3 Comparison of FACT-G, FLIC and EORTC QLQ-C30 total/global functioning scores by presentation order(a)
FACT-G FLIC EORTC QLQ-C30
Order A (N = 445) Order B (N = 390) Order A (N = 378) Order B (N = 436) Order A (N = 433) Order B (N = 428)
Mean 85.92 83.48 123.56 124.18 63.68 67.11
Difference in means (95% CI) 2.44 (0.66 to 4.24) -0.62 (-3.20 to 1.97) -3.43 (-5.87 to -0.99)
SD 15.44 15.94 21.84 22.79 21.69 21.83
(a)Order A and order B mean, respectively, the HRQoL instrument was placed first and second in the questionnaire.
Table 4 presents the mean values of the FACT-G, FLIC and EORTC QLQ-C30 total / global functioning scores by performance status and presentation order. All three instruments indicated a statistically significantly poorer quality of life in patients who had a poorer performance status (ECOG score 2 to 4) regardless of presentation order (each p < 0.05). In the case where FACT-G was administered first, the FACT-G score was 9.39 points higher in patients with better performance status. In the case where it was administered last, the FACT-G score was 6.68 points higher in such patients. The difference between the two estimates of between-group difference was 9.39 – 6.68 = 2.71 (90% CI = -1.83 to 7.24). Similarly, the differences (90% CI's) in between-group difference for FLIC and EORTC QLQ-C30 were 1.00 (-5.88 to 7.88) and -3.60 (-9.70 to 2.50), respectively. All three estimates of difference in between-group difference were within the equivalence zone of ± 0.25 SD. Although the three confidence intervals slightly exceeded the equivalence zone of ± 0.25 SD, they fell within the "small difference" zone of ± 0.5 SD. Again, adjustment for covariates did not make any practical difference. The difference between the two estimates of between-group difference for FACT-G, FLIC and EORTC were, respectively, 3.26 (-1.18 to 7.71), -0.79 (-7.14 to 5.56) and -3.58 (-9.47 to 2.32).
Table 4 Comparison of the mean FACT-G, FLIC and EORTC QLQ-C30 total/global functioning scores by ECOG score and presentation order(a)
Order ECOG score FACT-G FLIC EORTC QLQ-C30
A 0–1 83.93 119.37 58.73
2–4 74.54 110.11 52.18
Difference in means 9.39 (p < 0.001) 9.26 (p = 0.002) 6.55 (p = 0.014)
B 0–1 80.64 118.39 64.21
2–4 73.96 110.13 54.06
Difference in means 6.69 (p = 0.001) 8.26 (p = 0.004) 10.15 (p < 0.001)
Difference in A – difference in B 2.70 1.00 -3.60
90% CI (-1.83 to 7.24) (-5.88 to 7.88) (-9.70 to 2.50)
(a) Order A and order B mean, respectively, the HRQoL instrument was placed first and second in the questionnaire.
Table 5 shows the Cronbach's alpha values of the three instruments by presentation order. The values were very similar across presentation orders and all three confidence intervals totally fell within the pre-defined equivalence zone of ± 0.1.
Table 5 Internal inconsistency of Fact-G, FLIC and EORTC QLQ-C30 by presentation order(a)
Order FACT-G FLIC EORTC QLQ-C30
A 0.909 0.925 0.896
B 0.911 0.934 0.866
Difference in alpha -0.002 -0.009 0.030
(90% CI) (-0.022 to 0.099) (-0.024 to 0.006) (-0.024 to 0.082)
(a) Order A and order B mean, respectively, the HRQoL instrument was placed first and second in the questionnaire
Discussion
Experimental evidence on the issue of order effects in the assessment of cancer patients' quality of life is scarce. There is substantial evidence that the measurement of psychological health and psychiatric morbidity are affected by the order of presentation of instruments [5-7]. However, a recent experimental study of 190 cancer patients suggested that the FACT-G and Quick-FLIC were free from such effects [11]. The researchers suggested that questions on HRQoL are less stigmatising and less threatening than questions about psychological problems. They also suggested that the more complicated skip patterns of some psychological / psychiatric measures may give rise to order effects related to the "punishment hypothesis" and "learning hypothesis" and that such patterns are rarely seen in cancer quality of life questionnaires [5,11]. In the present study of a substantially larger sample, we examined three HRQoL instruments commonly used in cancer research. Due to logistic considerations, we chose to use days rather than patients as the units of randomisation. We can think of no reason why bias should arise from this allocation scheme. Comparison of various background characteristics attested to the comparability of the patients randomised to different questionnaire packages. Secondary analyses adjusted for covariates gave similar results. The strength of order effects (if any) may depend on the length of questionnaire. The mean time to completion of the questionnaire packages was 20.4 minutes in the present study, about 5 minutes longer than that of the previous study, and the 90th percentile was 39.0 minutes.
Our findings lend additional support to the previous finding that the order of presentation has little influence over the assessment of quality of life in cancer patients, evidenced by the following results. First, equivalence in the number of missing values and internal consistency of all three instruments across presentation orders was confirmed. Second, the mean values of the FLIC administered in different orders were also equivalent. The FACT-G and EORTC QLQ-C30 administered in different orders also showed similar mean values, although the confidence intervals of the difference slightly excluded the equivalence margin. Since the confidence intervals totally fell within the "small difference" zone of ± 0.5 SD, it can be concluded that at most the order of presentation has a small effect on mean FACT-G and EORTC QLQ-C30 global scores. Third, regardless of presentation order, all three instruments revealed a statistically significant difference in quality of life between patients with better versus poorer performance status. Again, the confidence intervals totally fell within the pre-defined "small difference" zones although not the equivalence zones. Known-groups validity did not seem to be affected. The use of multiple instruments in an interview is a common practice. The findings here should be good news for quality of life researchers as they suggest that previous studies were probably not unduly influenced by different ordering of instruments and that more complicated designs to prevent an order effect are not necessary.
The study of equivalence is often controversial. There is no clear-cut ground for the definition of equivalence. Still, studies reviewed above seem to converge to the conclusion that a difference smaller than 0.25 SD is irrelevant and 0.5 SD is small. Hence the ways we defined the equivalence and small difference zones. Secondly, the use of 90% CI is mainly a matter of common practice in equivalence studies rather than a matter of theoretical justification. In a discussion about the use of confidence intervals in equivalence trials, Senn [23] maintained that "all standards of significance and confidence are in any case arbitrary... little can be done to remove the arbitrary element". Since both 90% and 95% are arbitrary, it is our preference to adopt the common practice of using 90% CI. The response rate to this study was about 60%. Though this is not a high response rate, the findings are relevant because patients who refused to participate in the assessment of quality of life were not the concern of the present study. Whether there is an order effect in questionnaire presentation does not have any relevance to patients who do not participate, and vice versa. One limitation of the present study was that, although the point estimates seemed to suggest the lack of an order effect in mean scores and know-groups validity, some of the confidence intervals concerned slightly stretched across the equivalence margins. As such, a more definite conclusion awaits further studies. Moreover, the issue should be assessed again if the interviews concerned are considerably lengthier than the present one.
Conclusion
There is no evidence of any major impact of the order of presentation on the assessment of cancer patients' quality of life when two of the three questionnaires – FLIC, FACT-G and EORTC QLQ-C30 – are used together.
Authors' contributions
YBC conceived of the study, participated in the experimental design, developed the statistical framework, carried out part of the statistical analysis, and drafted part of the manuscript. CL conducted a large part of the statistical analysis and drafted part of the manuscript. CG participated in the research design, and the interpretation and discussion of findings. JT participated in the research design, the development of the statistical framework for the equivalence analysis, the interpretation of findings, and writing of the manuscript. JW participated in the research design, and the interpretation and discussion of findings. All authors read and approved the final manuscript.
==== Refs
Serdula MK Mokdad AH Pamuk ER Williamson DF Byers T Effects of question order on estimates of the prevalence of attempted weight loss Am J Epidemiol 1995 42 64 67 7785675
Bowling A Research Methods in Health 1997 Buckingham: Open University Press 241 270
Schuman H Presser S Questions and Answers in Attitude Surveys 1981 NY: Academic Press
Streiner DL Norman GR Health Measurement Scales: A Practical Guide to their Development and Use 1989 Oxford: Oxford University Press 144
Jensen PS Watanabe HK Richters JE Who's up first? Testing for order effects in structured interviews using a counterbalanced experimental design J Abnorm Child Psychol 1999 27 439 436 10821625 10.1023/A:1021927909027
Mook DG Psychological Research: Strategy and Tactics 1982 NY: Harper & Row
Lucas CP The order effect: reflections on the validity of multiple test presentations Psychol Med 1992 22 197 202 1574556
Dunn KM Jordan K Croft PR Does questionnaires structure influence response in postal surveys J Clin Epidemiol 2003 56 10 16 12589865 10.1016/S0895-4356(02)00567-X
Kemmler G Holzner B Kopp M Dunser M Margreiter R Greil R Sperner-Unterweger B Comparison of two quality-of-life instruments for cancer patients: the Functional Assessment of Cancer Therapy-General and the European Organization for Research and Treatment of Cancer Quality of Life Questionnaire-C30 J Clin Oncol 1999 17 2932 2940 10561373
Cheung YB Goh C Wong LC Ng GY Lim WT Leong SS Tan EH Khoo KS Quick-FLIC: Validation of a short questionnaire for assessing quality of life of cancer patients Br J Cancer 2004 90 1747 1752 15150626
Cheung YB Wong LC Tay MH Toh CK Koo WH Epstein R Goh C Order effects in the assessment of quality of life of cancer patients Qual Life Res 2004 13 1217 1223 15473500 10.1023/B:QURE.0000037499.80080.07
Senn S Cross-over Trials in Clinical Research 1993 Chichester and New York: Wiley
Pocock S Clinical Trials: A Practical Approach 1983 Chichester and New York: Wiley
Goh CR Lee KS Tan TC Wang TL Tan CH Wong J Ang PT Chan ME Clinch J Olweny CL Schipper H Measuring quality of life in different cultures: translation of the Functional Living Index for Cancer (FLIC) into Chinese and Malay in Singapore Ann Acad Med Singapore 1996 25 323 334 8876896
Cheung YB Ng GY Wong LC Koo WH Tan EH Tay MH Lim D Poon D Goh C Tan SB Measuring quality of life in Chinese cancer patients: a new version of the Functional Living Index – Cancer (Chinese) Ann Acad Med Singapore 2003 32 376 380 12854381
Conner-Spady B Cumming C Nabholtz JM Jacobs P Stewart D Responsiveness of the EuroQol in breast cancer patients undergoing high dose chemotherapy Qual Life Res 2001 10 479 486 11789549 10.1023/A:1013018218360
Takeda F Uki J Recent progress in cancer pain management and palliative care in Japan Ann Acad Med Singapore 1994 23 296 299 7521622
Blagden SP Charman SC Sharples LD Magee LR Gilligan D Performance status score: do patients and their oncologists agree Br J Cancer 2003 89 1022 1027 12966419 10.1038/sj.bjc.6601231
Cella D FACIT Manual: Manual of the Functional Assessment of Chronic Illness Therapy (FACIT) Measurement System 1997 Evanston, IL: Northwestern University
Hardin J Hilber J Generalized Linear Models and Extensions 2001 College Station, TX: Stata Corporation
Efron B Tibshirani R An Introduction to the Bootstrap 1993 New York: Chapman & Hall
Machin D Campbell M Fayers P Pinol A Sample Size Tables for Clinical Studies 1997 2 Oxford: Blackwell
Senn S Statistical Issues in Drug Development 1997 Chichester: Wiley 320
Cella D Eton DT Lai JS Peterman AH Merkel DE Combining anchor and distribution-based methods to derive minimal clinically important differences on the Functional Assessment of Cancer Therapy (FACT) anemia and fatigue scales J Pain Symptom Manage 2002 24 547 561 12551804 10.1016/S0885-3924(02)00529-8
Osoba D Rodriges G Myles J Zee B Pater J Interpreting the significance of changes in health-related quality-of-life scores J Clin Oncol 1998 16 139 144 9440735
Cohen J Statistical Power Analysis for the Behavioural Sciences 1988 2 Hilllsdale, NJ: L.Erlbaum Associates
Cheung YB Goh C Thumboo J Khoo KS Wee J Variability and sample size requirements of quality of life measures: A randomized study of three major questionnaires J Clin Oncol
| 15927054 | PMC1175098 | CC BY | 2021-01-04 16:38:13 | no | Health Qual Life Outcomes. 2005 May 31; 3:37 | utf-8 | Health Qual Life Outcomes | 2,005 | 10.1186/1477-7525-3-37 | oa_comm |
==== Front
J Neuroengineering RehabilJournal of NeuroEngineering and Rehabilitation1743-0003BioMed Central London 1743-0003-2-101592707510.1186/1743-0003-2-10ResearchFinger extensor variability in TMS parameters among chronic stroke patients Butler Andrew J [email protected] Shannon [email protected] Steven L [email protected] Paul [email protected] Departments of Rehabilitation Medicine, Emory University School of Medicine, Emory University, Atlanta, USA 30322, GA 2 Medicine, Emory University School of Medicine, Emory University, Atlanta, USA 30322, GA3 Cell Biology, Emory University School of Medicine, Emory University, Atlanta, USA 30322, GA4 Department of Psychology, Emory College, Emory University, Atlanta, USA 30322, GA5 Department of Biostatistics, Rollins School of Public Health, Emory University, Atlanta, USA 30322, GA2005 31 5 2005 2 10 10 7 11 2004 31 5 2005 Copyright © 2005 Butler et al; licensee BioMed Central Ltd.2005Butler 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 determined the reliability of topographic motor cortical maps and MEP characteristics in the extensor digitorum communis (EDC) evoked by single-pulse TMS among patients with chronic stroke.
Methods
Each of ten patients was studied on three occasions. Measures included location of the EDC hotspot and center of gravity (COG), threshold of activation and average amplitude of the hotspot, number of active sites, map volume, and recruitment curve (RC) slope.
Results
Consistent intrahemispheric measurements were obtained for the three TMS mapping sessions for all measured variables. No statistically significant difference was observed between hemispheres for the number of active sites, COG distance or the RC slope. The magnitude and range of COG movement between sessions were similar to those reported previously with this muscle in able-bodied individuals. The average COG movement over three sessions in both hemispheres was 0.90 cm. The average COG movement in the affected hemisphere was 1.13 (± 0.08) cm, and 0.68 (± 0.04) cm) for the less affected hemisphere. However, significant interhemispheric variability was seen for the average MEP amplitude, normalized map volume, and resting motor threshold.
Conclusion
The physiologic variability in some TMS measurements of EDC suggest that interpretation of TMS mapping data derived from hemiparetic patients in the chronic stage following stroke should be undertaken cautiously. Irrespective of the muscle, potential causes of variability should be resolved to accurately assess the impact of pharmacological or physical interventions on cortical organization as measured by TMS among patients with stroke.
motor mappingreliabilitycenter of gravityupper limbplasticityrehabilitationcortex
==== Body
Background
Single pulse Transcranial Magnetic Stimulation (TMS) is a safe and noninvasive technique for mapping cortical motor representation [1-4]. Recently, TMS has been used to explore mechanisms underlying both spontaneous and therapy-induced post-stroke motor recovery. In this context, most interventional studies have not considered intra-subject variability of TMS maps prior to the provision of a therapy, thus implying that cortical changes are attributable to the intervention. However, our laboratory recently demonstrated significant variability within able-bodied, right hand dominant participants across sessions and between hemispheres, for distance between the lowest resting motor threshold locations for a muscle (hotspot), center of gravity distance, and normalized map volume TMS parameters when mapping the extensor digitorum communis (EDC) muscle [5]. Adjusting for time and examining mean changes for hemispheres across sessions revealed that there was a 9-fold greater movement over sessions in the left hemisphere among these variables. Previous studies have shown reproducible motor maps of abductor pollicis brevis (APB) and abductor digiti minimi (ADM) [6] in both healthy subjects [6] and chronic stroke patients [7] using conventional electrode placement. In addition, Wasserman et al. (2002) found no systematic changes in resting and active motor evoked potential (MEP) thresholds among 19 women across three sessions.
However, few studies have examined the inherent variability in TMS motor maps in chronic stroke subjects not receiving an intervention. This preliminary study represents one of the first efforts to evaluate intra-subject variability in TMS motor maps of chronic stroke patients during three separate mapping sessions. As in a previous report on able-bodied participants [5], we chose to map EDC because this muscle is often affected by a stroke and its volitional activation is important in overcoming the profound flexion posture at the hand and wrist that characterizes many patients. Furthermore, the EDC is near the skin surface, making it a convenient and more precise site for electromyography recording due to its close proximity to other finger and wrist extensors which limits effects of cross talk, undesired overflow effects and, if present, volume-conducted pick up by muscles with comparable function.
Therefore, the present study is unique because of the specificity of recording using closely spaced electrodes and the repetitive sessions permitting examination of variability in TMS-related measures for the EDC muscle in patients greater than two years post stroke. The inherent variability seen in TMS measures following physical or pharmacological interventions would need to be less than that seen under non-interventional conditions to be assured that changes induced by these interventions are associated with cortical reorganization.
Methods
Design
This study used repeated measures, non-random sampling design. Motor maps for the EDC were created for each hemisphere during all three sessions for every subject. Sessions were separated by approximately seven days.
Chronic stroke patients
Ten right-handed patients who suffered a stroke greater than 2 years prior to testing were recruited using consecutive sampling of all chronic stroke patients who had the ability to extend ≥ 20° at the wrist and 10° at the fingers [8]. Specific upper extremity motor deficits were similar to those seen in patients enrolled in a multisite randomized trial to investigate the effect of constraint-induced movement therapy in improving upper extremity function among adults recovering from a cerebrovascular stroke [9]. The medical condition of each patient was stable. Each volunteer was living independently within the community and ambulated independently. For this preliminary study, patients with a wide range of cortical lesions and chronicity were studied. Basic information about age, gender, hand dominance, time since stroke and lesion site is found in Table 1. Four of ten patients had strokes that primarily affected their non-dominant upper extremity. Data from nine able-bodied volunteers collected in a previously reported TMS variability study were used as a comparison group [5].
Table 1 Clinical data for patient volunteers.
Participant Age Gender Hand Dom. Months since Stroke Site of Lesion
1 58 Male R 32 Left Lacunar Infarct CVA
2 55 Male R 34 Left thalamic ICH and right subcortical lacunae
3 78 Male R 35 Right Internal capsule lacunar CVA
4 56 Female R 56 Right cerebral hemisphere
5 46 Female R 54 Right putamen hemorrhage
6 70 Female R 98 Right cerebral hemisphere
7 60 Female R 147 Left cerebral hemisphere
8 56 Male R 85 Left cerebral hemisphere
9 56 Male R 33 Left lacunar infarct corona radiate
10 67 Female R 25 Left cerebellum
Participants were excluded if they had: a history of epilepsy, psychiatric disorders, fracture in the upper extremity within the past two years, diaphoresis, severe spasticity, tendonitis in the upper extremity within the last three months, migraine headaches within the last six months, Attention Deficit Disorder, or Attention Deficit Hyperactivity Disorder. In addition, participants could not be receiving stimulant or relaxant medications, (including anti-spasticity medication or pharmacological injections) demonstrate current exacerbation of osteoarthritis in the upper extremity or of rheumatic disorders, or be participating in sports that require excessive wrist extension for more than once per week over the previous three months. Volunteers read and signed an informed consent form previously approved by the local University Institutional Review Board.
Measurements/Instrumentation
Details about the experimental design and data collection methods have been presented previously [5]. Briefly, the following variables were measured at each session: hotspot and active site locations, hotspot excitability threshold, average MEP amplitude for hotspot and active sites, and recruitment curve slope. The hotspot was defined as the grid location where the motor threshold was the lowest while evoking the largest response [10]. Given the comparatively closer inter-electrode recording distances, active sites were designated as the grid locations where a response of ≥ 25 μV in 5 out of 10 trials at 110 percent of resting motor threshold was obtained. Each site with five consecutive responses less than 25 μV was considered non-active. Mapping was complete when locations adjacent to the active sites were identified as non-active. Recruitment curves were generated to evaluate the relationship between MEP amplitudes at the hotspot and progressively increasing stimulus intensities until the curve flattened. The slope of the recruitment curve is thought to be a function of the physical distribution of stimulus excitation from the coil and yields a measure of distribution of the excitability in the cortex [11].
The average MEP amplitude for the hotspot, center of gravity (COG), normalized map volume, and slope of the recruitment curve, were calculated following data collection. COG was defined as the map location representing the amplitude-weighted center of the area of excitability [12]. Normalized map volume was defined as the area of the map multiplied by the normalized MEP amplitudes.
Normalization of mean amplitudes (nMEP) was completed for all coordinates for each participant by dividing the mean amplitudes by the maximum mean amplitude. The normalized map volume (nMV) was calculated by adding all of the nMEP amplitudes and multiplying by the area [13]. The X and Y coordinates for each active site were multiplied by the normalized MEP amplitude (X*nMEP and Y*nMEP), and the sum of all the values was calculated respectively. The center of gravity (COG) X coordinate was calculated by and COG Y coordinate was calculated by [12].
The recruitment curve (RC) was generated by examining MEP amplitudes at the hotspot over progressively increasing intensities, thus providing information about cortical excitability. This was done by placing the coil at the hotspot and recording 5 stimuli in 10% increments beginning at an intensity of 10 % below threshold. Data collection for the RC was terminated when a plateau of the sigmoidal curve was observed. When calculating the RC slope, the first two data points collected were omitted because they were at sub-threshold levels, and the end point of the recruitment curve was determined to be either at 80% stimulator output, where a supra-threshold motor response was observed, or once a plateau in the recruitment curve was noted. The slope of the recruitment curve was generated from the resultant data points using linear regression.
The MEPs were recorded using two 7 mm × 4 mm silver-silver chloride surface electrodes (Medtronic, Inc., Minneapolis, MN) separated by approximately 1.5 centimeters. The peak-to-peak amplitude of the unrectified MEP was measured automatically using custom established routines created in LabView 6.0 (National Instruments, Austin, TX) in each of the 10 trials in each block, and their average was calculated for each stimulus site to give the mean peak-to-peak amplitude.
Reliability
The reliability of data acquisition was assessed by two investigators. One investigator performed the stimulation, while the other monitored the recordings for all sessions. Each investigator performed the same duties throughout the study to decrease the chance of experimenter variability [5]. Potential participants were screened using an inclusion/exclusion criteria questionnaire. To ensure consistent electrode placement for all sessions, the EDC muscle belly was isolated by palpation and then marked at the first session. A clear acetate sheet was applied to each forearm. Marks were then placed on the acetate sheet for electrode placement and relevant anatomical landmarks to assure consistent placement during subsequent sessions. To maintain consistent cap placement across sessions, detailed distance recordings were made from the nasion, inion, and bilateral pre-tragus to the vertex.
Procedure
Patient preparation
After isolating each EDC with the wrist in flexion to determine optimal placement of the electrodes, the skin surface over the EDC on the forearms was shaved and abraded with alcohol until erythemic responses appeared. Recording electrodes were placed on the skin over the EDC muscle bellies, and a reference electrode was applied ipsilaterally and proximally to the recording electrodes to reduce EMG noise levels. Skin impedance between active electrodes and between each active electrode and the reference were kept below 2 kilo-ohms (kΩ), and below 20 kΩ respectively.
Each participant was seated in a relaxed position with pillows placed under the forearms and hands. A firm-fitting cap upon which 1 cm2 grids had been imprinted was placed on the participant's head and secured appropriately to serve as a reference for reproducible coil placement and orientation.
Data collection
EMG data were measured bilaterally through surface electrode pairs, but responses to cortical stimulation were only recorded from the electrodes contralateral to the hemisphere being stimulated. Surface EMG signals were amplified and filtered with an Isolated Bioelectric Amplifier (James Long, Caroga Lake, NY), with bandpass filter settings of 30 and 1000 Hz, and digitally sampled at 1 KHz. 100 ms of prestimulation activity and 200 ms of post-stimulation activity were recorded. Trials in which active contraction contaminated the MEP were omitted, and the trial was repeated. To facilitate subject alertness throughout data collection, the investigator monitoring recordings engaged in neutral conversation with each volunteer between blocks of presentations of stimuli.
Stimulation of each hemisphere at the motor cortex using a 9 cm diameter figure-8 coil MAGSTIM 200 (Magstim Company Ltd., Whitland, Dyfed, UK) was performed in a systematic fashion at 0.2 Hz. The coil was oriented with the handle facing backward so the induced current in the brain was in the posterior-anterior direction during the rising phase of the monophasic pulse. Approximately 300–400 stimuli were delivered in sequential order during the mapping procedure.
Potential hotspot sites were identified using a stimulus intensity that evoked MEPs ≥ 25 μV, in five out of ten trials. Once these cortical sites were identified, the intensity was reduced until the hotspot and the hotspot's excitability threshold for the EDC were determined. Thereafter, the stimulus intensity was increased by ten percent and cortical sites beginning at the hotspot were stimulated to identify the active sites. Mapping was complete when all surrounding inactive sites were identified.
Data Analysis
The assumption of sphericity was ensured using the Greenhouse-Geisser correction. A two-way repeated measures analysis of variance (ANOVA) was used to explore the difference between sessions, hemispheres, lesion location and the interaction within participants for the following variables: resting motor threshold, map area, mean peak-to-peak MEP amplitude for the hotspot, normalized map volume, slope of recruitment curve, COG centroid and COG distance A and B. For all tests the alpha level was set at α = 0.05. The Euclidean equation was applied to determine the distance the hotspot and COG locations traveled from sessions: one to two (distance A) and two to three (distance B).
To allow for comparison between sessions in a single hemisphere, a centroid point, Xc, Yc, was calculated from the three x-and y-co-ordinates for the COG and hotspot positions. The x and y co-ordinates represent the medial-lateral and anterior-posterior distance (cm) from an arbitrary origin (0,0).
Results
The scalp overlying the motor cortex was stimulated at 110% of motor threshold, while recording from EDC. A representative MEP amplitude of 60 μV beginning approximately 20 ms after the stimulus artifact is depicted in Figure 1.
Figure 1 A representative MEP amplitude of 70 μV beginning approximately 20 ms after the stimulus artifact.
Patient data for the affected and less affected hemispheres are provided in Table 2 (see Additional file 1). The resting motor threshold (RMT) in the affected hemisphere had a minimum value of 43% (case #2, session 3) and maximum value of 100% (case #1, session 3). The RMT values in the less affected hemisphere ranged from 31% (case #4, session 3) to 63% (case #8, session 1).
Map volume in the affected hemisphere ranged from 3.13 cm2 (case #2, session 3) to 13.26 cm2 (case #1, session 1), while in the less affected hemisphere values ranged from 0.038 cm2 (case #6, session 3) to 0.385 cm2 (case #3, session 1) respectively. The minimum MEP amplitude in the affected hemisphere was observed in case #1, session 2 (0.0122 μV) while the maximum 0.1828 μV was observed in case #2, session 1.
The number of active sites in the affected hemisphere ranged from 0 (case #1, session 2) to 19 (case #10, session 2). While the range in the less affected hemisphere was from 3 active sites (case #7, session 3) to 11 (case #1,2,5). Collectively these data would appear to illustrate a substantial degree of variability in all values among these 10 patients with stroke.
However, analysis of variance showed no between session variability for any of the measured parameters (Table 3). There were no statistically significant interhemispheric (between hemispheres) difference in the number of active sites (F1,7 = 0.28; p= 0.6157), and RC slope (F1,7 = 3.34 ; p = 0.1106). In contrast, greater interhemispheric variability was observed for: average MEP amplitude (F1,6 = 85.01; p < 0.0001), normalized map volume (F1,7 = 5.98; p = 0.044), and resting motor threshold (F1,8 = 12.79; p = 0.0072) (Table 3). As shown in Figure 2, resting motor threshold was larger for the affected 63.1% (2.1) than the less affected 44.7% (2.1) hemisphere. Normalized map volume was also larger for the affected 8.7 cm (0.5) compared to the less affected 6.3 cm (0.5) hemisphere. Larger MEP amplitudes were recorded in the less affected hemisphere compared to the more affected hemisphere [(0.15 μV (± 0.01) and 0.05 μV (± 0.01)].
Table 3 Analysis of Variance for Dependent Variables
Hemisphere Session Interaction
F value P value F value P value F value P value
Motor Threshold 12.79 0.0072* 0.47 0.6336 1.25 0.3139
Average MEP Amplitude 85.01 0.0001* 1.50 0.2628 2.78 0.1016
# Active Sites 0.28 0.6157 0.52 0.6061 0.29 0.7532
Normalized Map Volume 5.98 0.0444* 0.02 0.9759 1.35 0.2914
COG distance 1.22 0.2833 0.53 0.4781 0.06 0.8165
Recruitment Curve Slope 3.34 0.1106 0.67 0.5264 1.17 0.3380
* Indicates statistically significant value; MEP = Motor Evoked Potential
Figure 2 Inter-hemispheric variability collapsed across the three mapping sessions for the parameters: average MEP amplitude, normalized map volume, and resting motor threshold. P-values are depicted in the lower right corner of each plot.
When considering lesion location as a factor (i.e. cortical vs. subcortical), ANOVA revealed no significant differences in any of the dependent variables measured across hemisphere, session or their interaction.
The ANOVA comparing COG distances A versus B were not significant between hemispheres or sessions (Table 3). The magnitude and range of COG movement between sessions were similar (Figure 3, Table 4) to those reported in a previous mapping study of this muscle with able-bodied individuals [5]. The average COG movement over three sessions in both hemispheres was 0.90 cm. The average COG movement in the affected hemisphere was 1.13 (± 0.08) cm, and for the less affected hemisphere 0.68 (± 0.04) cm among our stroke participants.
Figure 3 2-D representation of the overall COG movement (cm) across three sessions for each participant and both hemispheres. First session is demarcated by a larger symbol. The COG was calculated using mean MEP amplitudes shown for active sites only. Larger numbers on the x-coordinate and y-coordinates represent lateral and anterior scalp stimulus locations, respectively. Note that locations are unadjusted for the repeated measures on hemisphere and session. Each grid location represents one centimeter. The hatched circle represents the COG centroid location for a single subject in one hemisphere. All centroids are displayed in Figure 4.
Table 4 Average and range of COG movement across session 1 (S1), session 2 (S2) and session 3 (S3) and between hemispheres. SD = standard deviation.
Hemisphere Mean (cm) SD Range (cm)
Affected
S1→S2 1.04 0.55 0.21→ 1.75
S2→S3 1.14 0.54 0.68A1.87
S3→S1 1.20 0.70 0.45A1.70
Ave 1.13 0.08 0.21→1.87
Less affected
S1→S2 0.90 0.41 0.49→ 1.70
S2→S3 0.62 0.39 0.16A0.90
S3→S1 0.53 0.34 0.02A0.91
Ave 0.68 0.04 0.02→1.70
To allow for comparison between sessions in a single hemisphere, a centroid point was calculated. No significant difference was observed between the affected and less affected hemispheres across three sessions for COG centroid (Figure 4). No significant interhemispheric (between hemisphere) or intrahemispheric (between session) variability was observed for the COG centroids (p = 0.6611).
Figure 4 Centroid location of COG for the affected and less affected hemisphere for individual patients along with 9 able-bodied adults. The left hemisphere corresponds to the dominant arm in able-bodied participants and the affected and less affected hemispheres are of mixed hand dominance for the patients No significant variability exists when comparing left or right hemisphere of right handed able-bodied individuals with affected (p = 0.996) and less affected (p = 0.68) hemispheres of patients. Symbols with asterisks (*) represent centroids for left hemisphere (triangle*) and right hemisphere (square*) of able-bodied individuals. Each grid location represents one centimeter.
There were no significant differences in movement of COG centroid between the left or right hemisphere of healthy right handed individuals [5] and the affected (p = 0.996) or less affected (p = 0.68) hemisphere of right handed patients with stroke. All of our able-bodied volunteers were right hand dominant, and all of our patients were right hand dominant. Therefore, both groups could be compared. Figure 4 indicates that the COG centroid location for the affected and less affected hemisphere for individual patients along with 9 able-bodied adults show considerable overlap.
Discussion
This study demonstrated consistent between session measures for all the recorded variables. Consistent between hemisphere measures were obtained for the number of active sites, COG distance and recruitment curve slope, when recording EDC maps using single pulse TMS among patients greater than 2 years after stroke. In contrast, between hemispheres variability was observed in three measures: the average MEP amplitude, normalized map volume and resting motor threshold.
These findings support previous studies which report reproducible motor maps of the abductor pollicis brevis [6,7,10] and abductor digiti minimi [6] in both healthy subjects [6,10] and chronic stroke patients [7].
Interhemispheric variability collapsed across the three mapping sessions
Our data are in accord with previous reports on patients with stroke showing that resting motor threshold is significantly higher and MEP amplitudes are smaller in the affected hemisphere compared to the less affected hemisphere and that the relationship is reproducible between sessions [7,14,15].
The larger normalized map volume of EDC in the damaged hemisphere may be due to the dynamic alteration in the pattern of brain activity in response to change in afferent signals, efferent signals and/or adjustment to injury (i.e. neuroplasticity). In the current study, six of ten patients reported strokes that primarily affected their dominant upper extremity. Although behavioral data were not collected prior to TMS mapping, all patients reported living within their communities and using their more impaired upper extremities for many activities of daily living. None of the volunteers were receiving formal training (i.e. constraint induced therapy) at the time of testing, however, they would have met the inclusion criteria to participate in a randomized clinical trial of constraint induced therapy that required initiation of wrist and finger extension [9]. Their repetitive efforts at using the more impaired arm may have contributed to modifying functional reorganization of remaining cortical tissue in the corresponding hemisphere. This use may have consequently led to a comparably larger map size.
Motor or sensory activity in one arm can affect the other arm. There is the potential for input from the ipsilateral (ie. less impaired hand) side to the damaged side of the brain. Frequent use of the less impaired limb may have led to a map volume increase on the ipsilateral (affected hemisphere). There is now evidence that such modulatory effects can occur with practice [16] and has the potential to occur with mild or strong voluntary contractions [17]. Further data collection is necessary to completely explore this theory.
The much greater COG movement across sessions in the damaged hemispheres of stroke patients than in undamaged hemispheres of both stroke patients and comparison group is likely related to greater map volume in the damaged hemispheres. The calculation of COG x- and y-coordinates is dependent upon MEP amplitude (nMEP), and normalized map volume (nMV). The normalized map volume is directly proportional to the number of active sites. Large intersession variation in either of these values will affect the COG value and subsequent calculation of displacement between sessions. Although the variability in MEP amplitude was comparable between hemispheres, closer inspection of the data indicated up to a 58% greater variation in the number of active sites between sessions on the affected hemisphere (mean = 8.13 ± 03.94) compared to the unaffected hemisphere (mean = 8.3 ± 02.30). The increased variability in the number of active sites in the affected hemisphere is a contributing factor to the greater COG movement between sessions observed in the affected hemisphere.
Overall COG movement across three sessions for each participant and both hemispheres
The center of gravity remained consistent over the three sessions, with the majority of movement occurring in the anterior or posterior directions, along the Y-axis (Figure 3), an observation consistent with the TMS-induced field generated from the figure of eight coil orientation [18]. The average COG movement in the less affected hemisphere, 0.68 (± 0.04) cm is equivalent to the average COG movement 0.68 (± 0.02) cm measured from EDC in nine able bodied adults [5]. The average COG movement in the affected hemisphere reported here is about 60% greater when compared to the less affected hemisphere (Table 4). These changes in COG shift between session and across hemispheres are considerably larger than measures reported by Liepert et al. in a previous study of stroke patients' undergoing an intervention [7]. Their measurement for COG displacement in the abductor pollicis brevis (APB) was 0.234 ± 0.21 cm in the media-lateral axis in the affected hemisphere and 0.153 ± 0.18 cm in the less affected hemisphere and 0.71 ± 0.47 and 0.50 ± 0.426 cm in the anterior-posterior axis for the affected and less affected hemispheres, respectively.
The difference in magnitude may be a function of how COG displacement is determined between sessions. Our calculation of the Euclidean distance is fundamentally different than that described by Liepert et al. [19,20]. Liepert's description of the shift in COG between sessions using displacement is useful because it provides an indication of both distance and directional change along one axis. However, concern should be given to the use of a mean displacement, expressed as the difference between two consecutive x- or y-coordinates without considering the absolute value of the calculation. Failure to consider the overall positive and negative directionality of displacement may have led to artificially lower COG shifts in value than seen in the current study (i.e. if first value is negative and second is of equal value positive). In contrast the resulting Euclidean distance between two points is an absolute value. Calculating the Euclidean distance between two points in a plane using the Pythagorean Theorem allows for the creation of a 2-dimensional displacement vector which can better describe the overall change in location between sessions independent of direction.
The calculation of COG is dependent on MEP characteristics which differ for distal and more proximal muscles. For instance, the MEP thresholds in proximal muscles (i.e. deltoid, biceps brachii) are higher and the responses vary more in amplitude from trial to trial than in distal muscles such as abductor pollicis brevis and flexor carpi radialis [21]. Furthermore, the form and structure of MEPs in proximal muscles is often more complex than in distal muscles. Although not statistically different, larger variation in EDC amplitude from trial to trial could be linked to the observed increases in COG movement between sessions because a single large amplitude MEP can have a significant weighting on the overall mean of 10 samples which is then used for subsequent statistical calculation.
Additionally, McDonnell et al. (2004) have noted sufficiently large variability in MEPs recorded under standard conditions so that no significant differences in their magnitude over time can be revealed by conventional statistical analysis (ANOVA). They suggested that if a change in MEP size is expected as a result of an intervention, the change in magnitude must be large or many trials must be included in the analysis, before significant differences can be demonstrated [22]. Therefore a reproducibly large change in MEP amplitude is necessary for significant movement in COG over sessions. However substantial variability in MEPs over trials may also increase COG movement.
Centroid of COG in both hemispheres among individual patients and able-bodied adults
The calculation of a centroid permits visualization of a geometric locus for COG among cerebral hemispheres of our stroke and able-bodied participants. One would predict slight variations in cortical representation of the EDC between hemispheres. However, there are no predicable shifts in COG from session1 to session 3. Our data provide evidence that there is relative consistency in chronic stroke patients not receiving an intervention.
MEP characteristics displaying stability between sessions
In this study we observed large fluctuations in MEP amplitude, even under carefully controlled conditions. A previous study [23] found that regardless of the variation in the MEP amplitude, TMS map positions and areas are remarkably stable, with variations on the order of 1 mm for map position and less than 5% for map area. Likewise, our standard deviation in COG values was very small, with a mean value of 1.1 mm in latitude and 1.3 mm in longitude across subjects. In addition, the standard deviation of mean map area was only 1.1 cm2 (3.0%) across subjects. This high stability in COG has been observed in serial studies of patients with unilateral motor problems, in which the less affected side was stable to within 2–3 mm over periods from weeks to years [24].
In our patients with chronic stroke, the average COG movement across sessions in the less affected hemisphere was comparable to those shown previously in able bodied individuals [5], while COG movement across sessions in the affected hemisphere varied more on average than seen among those individuals (Figure 3 and Table 4). This increased variability may have resulted from our patients sustaining an uncontrolled 'relaxed' state compared to a controlled low-level voluntary contraction at 10% of maximum root-mean-square EMG activity [23]. Nonetheless, the COG variability found in our study was insignificant, producing a reliable measure for each patient's hemispheres for all three mapping sessions.
MEP characteristics showing high variability
The high trial-to-trial variability of MEP amplitude may be attributed to a number of factors. First, a range of cellular excitability levels in both spinal and upper motoneurons, which, under some circumstances, may bring these cells very close to firing threshold without actually discharging. In the case of the upper motoneuron, inherent excitability levels could allow some neurons to reach their discharge threshold but without the appropriate numbers or sufficient spinal synaptic excitability to temporally or spatially enhance EDC motoneuron discharge. Thus, mapping in the relaxed state is complicated by the variations that may occur in corticospinal excitability but could not be measured by monitoring EMG activity.
A second factor that can contribute to the intrinsic differences in MEP amplitude is variability in the desynchronization of the efferent volley [25]. Spontaneous physiological oscillations in motoneuron excitability at both the cortical and spinal levels are uncontrollable and unobservable factors potentially causing significant fluctuations in response size [10]. Changes in the state of the participant's alertness [26], levels of muscle tonicity, or anticipation of movement-specific factors, such as mental imagery [27,28], may also contribute to intra-trial MEP variability.
A third factor contributing to MEP variability may be stimulus intensity. This study recorded MEPs using stimulus intensities of 110% of motor threshold. When using higher stimulus intensities, as reported in some studies [29] there are more motoneurons activated and, therefore, fewer are available to spontaneously discharge in concert and contribute to the MEP amplitude, thereby affecting variability.
Small alterations in the position of the coil also can produce a source of within-subject variability [30]. Although the experimenter can make every effort to hold the coil in a uniform manner on a given scalp location, the identical spot is probably not stimulated at each session. The figure-8 coil can be rotated slightly, yet be the source of immense change in the area of the cortex being stimulated.
Although stimulation with a figure-of-eight coil is often described as focal; 'focusing' the electromagnetic field is in fact not practical. The maximum field is generated at the point under the intersection of the two wings of a figure of 8 coil; however, a divergent field is created surrounding this point. As a result, the spatial distribution of induced current flow can still be quite large, and the possibility of exciting cells under the wings and even cells located some distance from their intersection exists [1].
These factors can produce substantial variations in results obtained from TMS mapping studies. Although some of these factors are near impossible to control in a mapping session, others, such as coil placement and focal stimulation or the level of subject attentiveness, should be addressed as a precursor to TMS mapping studies designed to exploit variables using this modality to interpret results from specific interventions.
Electrode Placements
We chose the use of closely spaced surface electrodes to measure motor evoked potentials from the EDC because this placement limits the evoked MEPs to the underlying muscle and its associated movement. Previous TMS studies used widely spaced electrode arrays. For example, the placement of electrodes over the abductor pollicis brevis records a wide range of movements caused by the flexor pollicis brevis, adductor pollicis, opponens pollicis or interossei [31]. When comparing evoked responses using typical montage and closer spaced electrode arrays, we demonstrated larger map volumes with the montage configuration compared to close electrode placement. The summation of MEPs that represent multiple muscles seen by the more widely spaced electrode configuration results in greater MEP amplitudes, map volumes, number of active sites, and steeper recruitment curve slopes; however, there is greater difficulty identifying which muscles contribute to the response with each cortical stimulation and the representative movements they subsume. This consideration is important in TMS studies that relate changes in map attributes to function. For example, in mapping the APB [19], by knowing that the traditional placements also monitor volume conducted responses from muscles with a flexion function, would increased maps be teleologically relevant to an intervention designed to enhance movement in patients with stroke?
Functional Ramifications
Another unique aspect of our study was the focus on activation of the EDC and extension of the fingers. This movement is important in retraining function among patients with stroke, because hand extensor muscles are typically weak or inactive while muscles with a flexion function are disinhibited. Placement of wide spaced EDC surface electrodes previously [32] may actually record motions that are counterproductive to the benefits inherent in the very therapy being instituted.
Conclusion
This study is one of the very few to examine variability in TMS responses among a small group of patients with chronic stroke. Even with the use of chronic stroke patients and closely spaced electrodes, similarities to previous studies using able-bodied subjects were found. However, not surprisingly, findings that were significantly different from these prior studies were also observed.
Closely spaced electrode placement is important for properly isolating movements in limb muscles. Therefore, TMS studies using this placement array need to be undertaken to determine if resultant maps replicate those generated from previous studies employing more traditional, wider spaced electrode configurations. The potential causes of variability identified in this study, the precision of electrode recordings, or entirely new analysis methods should be considered in an effort to accurately assess pharmacological or physical interventions and their impact on cortical organization.
Authors' contributions
AB, SK and SW conceived of the study, participated in its design, participated in the data collection and drafted the manuscript. PW participated in the design of the study and performed the statistical analysis. All authors read and approved the final manuscript.
Abbreviations
COG = center of gravity; TMS = transcranial magnetic stimulation; EDC= extensor digitorum communis; MEP = motor evoked potential; EMG= electromyography
Supplementary Material
Additional File 1
Table 2 represents patient data for affected and less affected hemispheres.
Click here for file
Acknowledgements
Support received from the Emory University School of Medicine and the Emory Department of Rehabilitation Medicine, and NIH Grants HSD 37606 and 40984. Special thanks to Sarah Blanton, DPT, NCS, Jean Ko, and Amir Ahmadian, Emory Department of Rehabilitation Medicine, for the recruitment of participants and data analysis. The authors are grateful to Dr. Warren G. Darling for his thoughtful comments on the manuscript.
==== Refs
Barker AT Paulus W, Hallet M, Rossini P, Rothwell J The history and basic principles of magnetic nerve stimulation Transcranial magnetic stimulation 1999 51 New York: Elsevier Science 3 18
Beric A Devinsky O, Beric A, Dogali M Transcranial electrical and magnetic stimulation Electrical and magnetic stimulation of the brain and spinal cord Advances in Neurology 1993 63 New York: Raven Press 29 42
Wassermann EM Risk and safety of repetitive transcranial magnetic stimulation: report and suggested guidelines from the International Workshop on the Safety of Repetitive Transcranial Magnetic Stimulation, June 5–7, 1996 Electroencephalogr & Clin Neurophysiol 1998 108 1 16 9474057 10.1016/S0168-5597(97)00096-8
Pascual-Leone A Tormos JM Keenan J Tarazona F Canete C Catala MD Study and modulation of human cortical excitability with transcranial magnetic stimulation J Clin Neurophysiol 1998 15 333 343 9736467 10.1097/00004691-199807000-00005
Wolf SL Butler AJ Campana GI Parris TA Struys DM Weinstein SR Weiss P Intra-subject reliability of parameters contributing to maps generated by transcranial magnetic stimulation in able-bodied adults Clin Neurophysiol 2004 115 1740 1747 15261852 10.1016/j.clinph.2004.02.027
Mortifee P Stewart H Schulzer M Eisen A Reliability of transcranial magnetic stimulation for mapping the human motor cortex Electroencephalogr Clin Neurophysiol 1994 93 131 137 7512919 10.1016/0168-5597(94)90076-0
Liepert J Graef S Uhde I Leidner O Weiller C Training-induced changes of motor cortex representations in stroke patients Acta Neurol Scand 2000 101 321 326 10987321 10.1034/j.1600-0404.2000.90337a.x
Wolf SL Binder-MacLeod SA Electromyographic biofeedback applications to the hemiplegic patient. Changes in upper extremity neuromuscular and functional status Phys Ther 1983 63 1393 1403 6611660
Winstein CJ Miller JP Blanton S Taub E Uswatte G Morris D Nichols D Wolf S Methods for a multisite randomized trial to investigate the effect of constraint-induced movement therapy in improving upper extremity function among adults recovering from a cerebrovascular stroke Neurorehabil Neural Repair 2003 17 137 152 14503435 10.1177/0888439003255511
Wassermann EM Variation in the response to transcranial magnetic brain stimulation in the general population Clin Neurophysiol 2002 113 1165 1171 12088713 10.1016/S1388-2457(02)00144-X
Siebner HR Rothwell J Transcranial magnetic stimulation: new insights into representational cortical plasticity Exp Brain Res 2003 148 1 16 12478392 10.1007/s00221-002-1234-2
Wassermann EM McShane LM Hallett M Cohen LG Noninvasive mapping of muscle representations in human motor cortex Electroencephalogr Clin Neurophysiol 1992 85 1 8 1371738
Thickbroom GW Mastaglia FL Pascual-Leone A, Davey NJ, Rothwell J, Wassermann EM, Puri BK Mapping studies Handbook of Transcranial Magnetic Stimulation 2002 London: Arnold: New York: Oxford University Press (distributor) 127 140
Pennisi G Alagona G Rapisarda G Nicoletti F Costanzo E Ferri R Malaguarnera M Bella R Transcranial magnetic stimulation after pure motor stroke Clin Neurophysiol 2002 113 1536 1543 12350429 10.1016/S1388-2457(02)00255-9
Cicinelli P Traversa R Rossini PM Post-stroke reorganization of brain motor output to the hand: a 2–4 month follow-up with focal magnetic transcranial stimulation Electroencephalogr Clin Neurophysiol 1997 105 438 450 9448645 10.1016/S0924-980X(97)00052-0
Strens LH Fogelson N Shanahan P Rothwell JC Brown P The ipsilateral human motor cortex can functionally compensate for acute contralateral motor cortex dysfunction Curr Biol 2003 13 1201 1205 12867030 10.1016/S0960-9822(03)00453-6
Hortobagyi T Taylor JL Petersen NT Russell G Gandevia SC Changes in segmental and motor cortical output with contralateral muscle contractions and altered sensory inputs in humans J Neurophysiol 2003 90 2451 2459 14534271
Ruohonen J Ilmoniemi RJ Pascual-Leone A, Davey NJ, Rothwell J, Wassermann EM, Puri BK Physical principles for transcranial magnetic stimulation Handbook of Transcranial Magnetic Stimulation 2002 London: Arnold: New York: Oxford University Press (distributor) 17 29
Liepert J Bauder H Wolfgang HR Miltner WH Taub E Weiller C Treatment-induced cortical reorganization after stroke in humans Stroke 2000 31 1210 1216 10835434
Liepert J Miltner WH Bauder H Sommer M Dettmers C Taub E Weiller C Motor cortex plasticity during constraint-induced movement therapy in stroke patients Neurosci Lett 1998 250 5 8 9696052 10.1016/S0304-3940(98)00386-3
Brasil-Neto JP McShane LM Fuhr P Hallett M Cohen LG Topographic mapping of the human motor cortex with magnetic stimulation: factors affecting accuracy and reproducibility Electroencephalogr Clin Neurophysiol 1992 85 9 16 1371748 10.1016/0168-5597(92)90095-S
McDonnell MN Ridding MC Miles TS Do alternate methods of analysing motor evoked potentials give comparable results? J Neurosci Methods 2004 136 63 67 15126046 10.1016/j.jneumeth.2003.12.020
Thickbroom GW Byrnes ML Mastaglia FL A model of the effect of MEP amplitude variation on the accuracy of TMS mapping Clin Neurophysiol 1999 110 941 943 10400209 10.1016/S1388-2457(98)00080-7
Byrnes ML Thickbroom GW Wilson SA Sacco P Shipman JM Stell R Mastaglia FL The corticomotor representation of upper limb muscles in writer's cramp and changes following botulinum toxin injection Brain 1998 121 977 988 9619198 10.1093/brain/121.5.977
Magistris MR Rosler KM Truffert A Myers JP Transcranial stimulation excites virtually all motor neurons supplying the target muscle. A demonstration and a method improving the study of motor evoked potentials. [comment] Brain 1998 121 437 450 9549520 10.1093/brain/121.3.437
Kiers L Cros D Chiappa KH Fang J Variability of motor potentials evoked by transcranial magnetic stimulation Electroencephalogr Clin Neurophysiol 1993 89 415 423 7507428 10.1016/0168-5597(93)90115-6
Abbruzzese G Trompetto C Schieppati M The excitability of the human motor cortex increases during execution and mental imagination of sequential but not repetitive finger movements Exp Brain Res 1996 111 465 472 8911941 10.1007/BF00228736
Ikai T Findley TW Izumi S Hanayama K Kim H Daum MC Andrews JF Diamond BJ Reciprocal inhibition in the forearm during voluntary contraction and thinking about movement Electromyogr Clin Neurophysiol 1996 36 295 304 8877322
Herwig U Kolbel K Wunderlich AP Thielscher A von Tiesenhausen C Spitzer M Schonfeldt-Lecuona C Spatial congruence of neuronavigated transcranial magnetic stimulation and functional neuroimaging Clin Neurophysiol 2002 113 462 468 11955990 10.1016/S1388-2457(02)00026-3
Ellaway PH Davey NJ Maskill DW Rawlinson SR Lewis HS Anissimova NP Variability in the amplitude of skeletal muscle responses to magnetic stimulation of the motor cortex in man Electroencephalogr Clin Neurophysiol 1998 109 104 113 9741800 10.1016/S0924-980X(98)00007-1
Butler AJ Wolf SL Transcranial magnetic stimulation to assess cortical plasticity: a critical perspective for stroke rehabilitation J Rehabil Med 2003 20 26 12817653 10.1080/16501960310010106
Wittenberg GF Chen R Ishii K Bushara KO Taub E Gerber LH Hallett M Cohen LG Constraint-induced therapy in stroke: magnetic-stimulation motor maps and cerebral activation Neurorehabil Neural Repair 2003 17 48 57 12645445 10.1177/0888439002250456
| 15927075 | PMC1175099 | CC BY | 2021-01-04 16:37:41 | no | J Neuroengineering Rehabil. 2005 May 31; 2:10 | utf-8 | J Neuroeng Rehabil | 2,005 | 10.1186/1743-0003-2-10 | oa_comm |
==== Front
Proteome SciProteome Science1477-5956BioMed Central London 1477-5956-3-41592704610.1186/1477-5956-3-4ResearchProteome analysis of a recombinant Bacillus megaterium strain during heterologous production of a glucosyltransferase Wang Wei [email protected] Rajan [email protected]ürch Tobias [email protected] Manfred [email protected] Marco [email protected] Dieter [email protected] Wolf-Dieter [email protected] TU-BCE, German Research Centre for Biotechnology, Mascheroder Weg 1, D-38124 Braunschweig, Germany2 Department of Structural Biology, German Research Centre for Biotechnology, Mascheroder Weg 1, D-38124 Braunschweig, Germany3 Institute of Microbiology, Technical University Braunschweig, Spielmannstrasse 7, D-38106 Braunschweig, Germany2005 31 5 2005 3 4 4 10 2 2005 31 5 2005 Copyright © 2005 Wang et al; licensee BioMed Central Ltd.2005Wang et al; licensee BioMed Central Ltd.This is an Open Access article distributed under the terms of the Creative Commons Attribution License (), which permits unrestricted use, distribution, and reproduction in any medium, provided the original work is properly cited.
A recombinant B. megaterium strain was used for the heterologous production of a glucosyltransferase (dextransucrase). To better understand the physiological and metabolic responses of the host cell to cultivation and induction conditions, proteomic analysis was carried out by combined use of two-dimensional gel electrophoresis and mass spectrometry (2-DE/MS) for protein separation and identification.
2-DE method was optimized for the separation of intracellular proteins. Since the genome of B. megaterium is not yet available, peptide sequencing using peptide fragment information obtained from nanoelectrospray ionization quadrupole-time-of-flight tandem mass spectrometry (ESI-QqTOF MS/MS) was applied for protein identification. 167 protein spots were identified as 149 individual proteins, including most enzymes involved in the central carbon metabolic pathways and many enzymes related to amino acid synthesis and protein synthesis. Based on the results a 2-DE reference map and a corresponding protein database were constructed for further proteomic approaches on B. megaterium.
For the first time it became possible to perform comparative proteomic analysis on B. megaterium in a batch culture grown on glucose with xylose induction for dextrasucrase production. No significant differences were observed in the expression changes of enzymes of the glycolysis and TCA cycle, indicating that dextransucrase production, which amounted to only 2 % of the entire protein production, did not impose notable metabolic or energetic burdens on the central carbon metabolic pathway of the cells. However, a short-term up-regulation of aspartate aminotransferase, an enzyme closely related to dextransucrase production, in the induced culture demonstrated the feasibility to use 2-DE method for monitoring dextransucrase production. It was also observed that under the cultivation conditions used in this study B. megaterium tended to channel acetyl-CoA into pathways of polyhydroxybutyrate production. No expression increases were found with cytosolic chaperones such as GroEL and DnaK during dextransucrase production and secretion, whereas a strong up-regulation of the oligopeptide-binding protein OppA was observed in correlation with an increased secretion of dextransucrase into the culture medium.
==== Body
Background
The Gram-positive bacterium B. megaterium has been proven as a promising host for the production of diverse heterologous proteins and vitamins due to its intrinsic favourable properties such as low protease activity and high secretion capability [1]. Using recombinant B. megaterium strains for the heterologous production of a glucosyltransferase, namely dextransucrase from Leuconostoc mesenteroides NRRL B-512F, has been under investigation and improved production and secretion of dextransucrase was achieved compared with the recombinant production of dextransucrase in E. coli [2]. Dextransucrase can be used to catalyze polymerization reactions leading to the production of dextran. Dextran is widely used as a blood plasma substitute or as a basic chromatographic support material.
To optimize the cell cultivation and the recombinant protein production processes, it is important to understand the physiological and metabolic responses of the host cell to the cultivation and induction conditions. To this end we carried out proteomic analysis with a recombinant B. megaterium strain. Unlike Bacillus subtilis, which is the best characterized Gram-positive bacterium with its genome already completely sequenced in 1997 [3] and comprehensive proteomic analysis has been since accomplished [4-6], the genome of B. megaterium has not yet been sequenced and to our knowledge no information on proteomic analysis of B. megaterium has been published. In this work for the first time proteomic analysis of a recombinant B. megaterium strain based on two-dimensional gel electrophoresis in combination with mass spectrometric techniques (2-DE/MS) for protein separation and characterization was carried out. Comparative proteomic analysis was performed to study cellular protein expression changes related to defined cultivation and induction conditions for the production of recombinant dextransucrase by the recombinant B. megaterium strain.
Results and discussion
1. Proteome mapping of the strain B. megaterium MS941dsrS by 2-DE/MS
Methods used for the proteomic mapping of B. megaterium are the characterization of protein expression changes by 2-DE and the identification of proteins of interest by MS. This is aimed at establishing a functional metabolic network of B. megaterium, especially those involved in the central carbon metabolism, amino acid biosynthesis and protein biosynthesis, as well as the identification of metabolic pathways and cellular processes closely related to the production and secretion of the recombinant protein. Figure 1 shows a typical image of 2-DE separation of intracellular proteins of B. megaterium in the pH range of 4–7. When 250 μg of a protein sample were applied, about 580 – 800 protein spots can be detected on the different gels after coomasie staining, The spots matching rates between individual gels were between 58% to 75%.
Figure 1 2-D IEF/SDS-PAGE gel electrophoretic separation of an intracellular protein extract of B. megaterium MS941dsrS at the pH range of 4–7. Cell sample was taken from a batch culture grown on glucose during the exponential growth phase and 6 h after xylose-induction. Proteins identified by ESI-QqTOF MS/MS are marked with their abbreviation names.
1.1 Peptide mass fingerprinting with MALDI-TOF MS
At present, the most straightforward and real high-throughput mass spectrometric method of protein identification is peptide mass fingerprinting. This approach of stringent matching of measured peptide masses with computer-generated masses relies, however, to a large degree on the availability of a completely sequenced genome of the investigated organism. Alternatively, if a protein of interest is not present in a database, peptide sequences deduced from tandem mass spectra can be employed for protein identification via sequence similarity searching. In this way a putative function can also be assigned to an unknown protein. Because of the rapid growth of completed genome sequences, it is becoming increasing possible to identify proteins of an unsequenced organism via sequence similarity searching, especially when sequences of organisms within the same kingdom or related species already exist in any of the public-accessible protein sequence databases [7-10].
In the case of B. megaterium its genome is not yet available. Up to now 76 protein sequences of B. megaterium can be found in the annotated protein database Swiss-Prot. However, only four of them are enzymes directly involved in the central carbon metabolic pathways, namely glyceraldehyde-3-phosphate dehydrogenase (GAP), 2,3-biphosphoglycerate-independent phosphoglycerate mutase (PGM), phosphoglycerate kinase (PGK) and triphosphate isomerase (TPI) for glycolysis. Another 215 protein sequences exist in the protein database TrEMBL, a computer-annotated supplement of Swiss-Prot . However, most of these proteins are hypothetical proteins and almost no enzymes of the central carbon metabolism or amino acid and protein biosynthesis are present. On the other hand, genome sequencing of several microorganisms from the genus Bacillus, i.e. B. subtilis [3], B. halodurans [11], Oceanobacillus iheyensis [12], B. anthracis [13], B. cereus [14] and recently B. licheniformis [15] have been finished. These Bacillus species show highly conserved orthologous genes, including those for central carbon metabolism and for amino acid and protein biosynthesis. In addition, as a model system of gram-positive bacteria the genome sequence of B. subtilis has been well annotated. Therefore, it is conceivable that the sequence information of these Bacilli can help the identification of unknown proteins of B. megaterium through sequence similarity searching.
At first, about 200 relatively highly expressed protein spots excised from 2-D gels were subjected to in-gel tryptic digestion and MALDI-TOF-MS analysis, followed by cross-species peptide mass fingerprinting against the protein sequence databases NCBInr and SWISS-PROT/TrEMBL. Regardless whether or not a constraint on species of origin was imposed, only 30 protein spots could be identified with significant scoring as 10 B.megaterium own proteins and 8 proteins showing high similarities to homologous proteins of other Bacillus species. Among the B.megaterium proteins identified are the four enzymes GAP, PGM, PGK and TPI present in the Swiss-Prot database.
1.2 Peptide sequencing with ESI-QqTOF MS/MS
The result indicates that the homology between B. megaterium and other Bacilli species with completed genome sequences are still not high enough for an unambiguous identification of most of the unknown proteins of B. megaterium only through peptide mass fingerprinting. Consequently, ESI-QqTOF MS/MS analysis was carried out to acquire additional peptide sequence information for protein identification via sequence similarity searching for homologous proteins.
Among the available sequence similarity searching programs using peptide sequences produced by MS/MS analysis, MS BLAST has been developed to overcome specific limitations imposed by mass spectrometric data, such as the limited completeness and confidence of predicted sequences. It is targeted at matching of closely related short peptides typically obtained from, for example, ESI-QqTOF MS/MS analysis [7-10,16]. However, the success of MS BLAST identification still depends on the number and quality of sequenced peptides. In our study, two to four tandem mass spectra of peptide precursors were normally available for a protein of interest from ESI-QqTOF MS/MS measurements. To obtain peptide sequences of better quality and to improve the possibility of protein identification, performing manual sequencing was still necessary for the interpretation of most of the MS/MS spectra. Peptide sequences obtained were assembled into query strings and subjected to MS BLAST searching. The results are presented as high-scoring pairs (HSPs) which are defined as regions of high local sequence similarity between individual peptides in the query and a protein sequence from the database entry. Homologous proteins shown in a hit list were sorted by their total scores, which are the sum of the scores of high-scoring pairs for each protein, and categorized according to their statistic significance into three groups: positive hit, borderline hit and negative hit. In this work we set the criteria for a positive identification as follows: the candidate protein is generally the top hit protein and the score of its top-ranked HSP should be higher than the statistic threshold. In addition, the candidate protein is normally a homologous protein from an organism of the genus Bacillus. Figure 2A shows, as an example, a MS spectrum of peptide precursors of an unknown protein digested with trypsin. Three double charged precursors of m/z 702.81, 784.31 and 892.43 (arrow indicated) were further fragmented to obtain their MS/MS spectra. Figure 2B is the MS/MS spectrum of the precursor m/z 892.43. Three amino acid sequences derived by manual sequencing of the three MS/MS spectra were used for similarity searching of homologous proteins. The NAD-dependent malic enzyme 3 of B. subtilis (MalS) was found as the top homologue protein candidate with a high score of 94 for the best aligned HSP and a total score of 242 which are significant enough for functional assignment of the unknown protein of B. megaterium as a malic enzyme.
About 300 relatively highly expressed protein spots were excised from 2-D gels and subjected to the ESI-QqTOF MS/MS analysis. 167 spot were identified as 149 individual proteins, because some proteins appeared as isoforms on the 2-D gels, namely as several spots having similar molecular weights (Mw) but different isoelectric points (pI). According to the categorization used in the KEGG PATHWAY database most identified proteins can be classified into the following functional categories based on their functions or, at least, putative functions assigned: 53 proteins of the carbohydrate metabolism, mainly enzymes for the central carbon metabolism, including nearly all enzymes involved in the glycolysis and tricarboxylic acid cycle (TCA cycle), 4 enzymes of the pentose phosphate pathways, as well as 12 enzymes related to pyruvate metabolism; 31 proteins related to amino acid biosynthesis and metabolism; 14 proteins associated with protein biosynthesis; 15 proteins for nucleotide metabolism and genetic information (DNA, RNA) processing; as well as proteins involved in energy metabolism (3), cellular processes (5), membrane transport (9), stress responses (14) and other pathways like metabolism of complex carbohydrates (1), biosynthesis of secondary metabolites (2) and metabolism of cofactors and vitamins (3). Only 8 proteins can not be assigned any functions or putative functions. The identified proteins are summarized in Table 1 (see additional file 1).
Figure 2 Sequencing of tryptic peptides by ESI-QqTOF MS/MS for protein identification via sequence-similarity searching. (A) MS spectrum of the peptide precursors of an unknown protein digested with trypsin, three double charged precursors of m/z 702.81, 784.31 and 892.43 were further fragmented to obtain their MS/MS sequence spectra; (B) MS/MS spectrum of the precursor m/z 892.43, shown at the bottom of the spectrum is the sequencing result of this peptide precursor.
This work represents the first effect of large scale 2-DE separation and quantification of intracellular proteins of B. megaterium combined with identification of many of these proteins by tandem mass spectrometric analysis. This enabled us to create a 2-DE reference map as highlighted in Figure 1 and a corresponding protein database of B. megaterium. Based on these information a metabolic network of the central carbon metabolism as well as a part of the amino acid biosynthesis and metabolism are constructed (Figure 3).
Figure 3 Metabolic Network of the central carbon metabolism and partial amino acid biosynthesis and metabolism of B. megaterium MS941dsrS constructed based on the identified proteins (italic labeled are proteins not yet identified).
2. Batch culture of the B. megaterim strain MS941dsrS for recombinant dextransucrase production
Our previous study on batch cultures of B. megaterium MS941dsrS revealed that an early xylose induction at OD578 of 0.3 and cultivation at pH 5.2 gave the best production and secretion of dextransucrase (data not published). Consequently, for proteomic analysis two batch cultures with B. megaterium MS941dsrS, one with xylose induction and the other one without xylose induction as control, were carried out under the same conditions. Conducting the control culture was aimed at distinguishing between protein expression changes resulting from recombinant dextransucrase production and those aroused from cell growth. As shown in Figure 4(A), the two batch cultures demonstrate very similar time course of cell growth and glucose consumption. Xylose added for induction was not consumed by the cells because of carbon catabolic repression by glucose.
Figure 4 Batch cultures of B. megaterium MS941dsrS at pH 5.2. (A) Time course of cell growth on glucose. Induction was carried out with 5 g/L xylose at OD578 = 0.3 for the production of the recombinant dextransucrase (DsrS). Samples were taken for proteomic analysis shortly before induction (I0), then 1 h (I1), 6 h (I6) and 10 h (I10) after induction. Also shown is the time course of a control batch culture of B. megaterium MS941dsrS without xylose induction. Samples were also taken for proteomic analysis according to the sampling periods used for the induced culture. (B) Production, distribution and secretion of DsrS during the xylose induced batch culture. DsrS was detected as three cell-associated forms: precursor DsrS with signal peptide still attached, mature DsrS without signal peptide and degraded DsrS in three fractions: secretion, cell-associated soluble fraction and cell-associated insoluble fraction, respectively.
To determine the formation and secretion of dextransucrase in B. megaterium cell samples were taken from the induced culture shortly before xylose induction, then 1 h, 2 h, 3 h, 6 h and 9 h after the induction. Three protein fractions, namely secretion fraction, cell-associated soluble fraction and cell-associated insoluble fraction, were prepared according to the method described in the Materials and Methods section and separated on SDS-PAGE activity staining gels to obtain dextransucrase in the different fractions. The results are shown in Figure 4(B). Three forms of dextransucrase of different molecular weights can be distinguished. They represent the precursor dextransucrase with signal peptide (200 kDa) that is still localized in the cytoplasm, the mature dextransucrase without signal peptide (188 KDa) that is already secreted outside the cell membrane and the degraded dextransucrase (165 kDa), respectively. A maximal production of dextransucrase reached already 1 h after the xylose induction which is shown as two thickest bands in the cell-associated soluble and insoluble fractions on the activity staining gels, respectively. In addition, the produced dextransucrase was readily translocated over the cell membrane, as is evidenced by the fact that no precursor dextransucrase is visible on the activity staining gels. However, dextransucrase aggregated in the space between cell membrane and cell wall, since no bands could be detected in the secretion fraction. The further passage of dextransucrase through the cell wall was hampered as it was also observed by Malten et al. [2]. Afterwards both cell-associated soluble and insoluble dextransucrase decreased steadily with time. Meanwhile the mature dextransucrase diffused continually through the cell wall to the surrounding growth medium. The diffusion barrier of the cell wall explains the time delay between maximal dextransucrase production and its maximal secretion into the medium which was reached 9 hours after induction.
3. Comparative proteomic analysis of metabolism
3.1 Regulations of enzymes involved in the glycolysis and TCA cycle
As depicted in Figure 5 expressions of most enzymes of the glycolysis showed no obvious changes during the sampling period of cultivation. Slightly reduced expression levels can be observed 1 h after xylose induction, except for pyruvate kinase (PYK). The non-induced culture demonstrated a similar time course of expression. As also shown in Figure 5, except for fumarate hydratase (FumA) enzymes involved in the TCA cycle were up-regulated with most of them showing continuous increases up to 6 h after xylose induction. However, the batch culture without xylose induction showed again similar time courses of expression changes of the most TCA cycle enzymes. This indicates that the up-regulations were more likely a result of cell growth rather than a consequence attributed to the demand for the recombinant protein dextransucrase production. We found that during the first hour with maximal dextransucrase production only 6.75 mg dextransucrase / g CDW were built, compared to a biomass increase of 750 mg/L within this time span. Taken into account that the biomass of B. megaterium contains about 40% proteins [17], dextransucrase amounted to only 2% of entire proteins synthesized during this time period. Therefore, we believe that in this case the production of dextransucrase did not impose any noticeable metabolic and energetic burdens on the cells central carbon metabolic pathways.
Figure 5 2-D electrophoresis determined expression changes of some enzymes involved in glycolysis (PFK, FBA, TPI, GAP, PGK, PGM, ENO and PYK) and TCA cycle (CitZ, CitB, CitC, OdhB, SucC, SucD, SdhA, FumA and MDH). The x-axis gives the time after xylose-induction and the y-axis gives the average value of normalized protein spot volume on 2-D gel. Abbreviations of protein names are referred to Table 1. Samples were taken from (A): a batch culture of B. megaterium MS941dsrS shortly before xylose-induction, then 1 h, 6 h and 10 h after induction for recombinant DsrS production (solid curves), and (B): a control batch culture of B. megaterium MS941dsrS without xylose-induction with the same time course of sampling as (A) (dashed curves).
The expression of the enzyme pyruvate dehydrogenase was an exception. As shown in Figure 6, all subunits of the pyruvate dehydrogenase complex (PDH), the component E1 pyruvate dehydrogenase with its two subunits (PdhA and PdhB), the component E2 dihydrolipoamide acetyltransferase (PdhC) and the component E3 dihydrolipoamide dehydrogenase (PdhD) showed continuously decreased expression levels, especially in the early cell growth phase. As a key enzyme at the interface between glycolysis and TCA cycle, PDH is inhibited by ATP, NADH and its product acetyl-CoA. These findings propose that the reduced expression level of PDH was caused by an accumulation of acetyl-CoA in cells, indicating an inherent discrepancy between glycolysis and TCA cycle of this B. megaterium strain. Consequently acetyl-CoA generated could not be readily oxidized through the TCA cycle, even though many enzymes of TCA cycle were also up-regulated during this early growth phase.
Figure 6 2-D electrophoresis determined expression changes of some enzymes involved in pyruvate metabolism (PdhA, PdhB, PdhC, PdhD and PTA), polyhydroxybutyrate synthesis (MmgA, PhaR and PhaP) and amino acid biosynthesis (AspB, GlnA, AroA, SerA, IlvB, IlvC, IlvD and IlvE), as well as the malic enzyme MalS. Additional descriptions for this figure are the same as given in Figure 5.
3.2 Regulations of enzymes involved in dissipation of acetyl-CoA
Contrary to the decreased expression of PDH, a notable up-regulation of acetyl-CoA acetyltransferase (MmgA), which converts acetyl-CoA to acetoacetyl-CoA, was observed in the early cell growth phase in both batch cultures (Figure 6), regardless whether or not there was a xylose induction. This can be considered as an additional evidence of the accumulation of acetyl-CoA in B. megaterium cells and how the cells cope with the oversupply of acetyl-CoA. It is well known that B. megaterium has a remarkable ability of producing polyhydroxyalkanoates (PHAs), a group of carbon and energy storage polymers that accumulate as inclusion bodies in many bacteria in response to environmental conditions [18-23]. Polyhydroxybutyrate (PHB) is the most common type of the family PHAs. B. megaterium cells taken from both induced and non-induced batch cultures were stained with Sudan black according to the method described by Burdon [24], and similar round PHB granule were visualized inside cells. Acetyl-CoA is used as substrate for the synthesis of PHB by a sequence of three reactions catalyzed by acetyl-CoA acetyltransferase (MmgA), 3-hydroxybutyryl-CoA dehydrogenase (MmgB) and poly(3-hydroxybutyrate) synthase (PHB Synthase). MmgA is the key regulatory enzyme of PHB synthesis from glucose [23]. Besides MmgA we have also identified two other proteins involved in PHB synthesis, namely PhaR and PhaP. PhaR functions directly or indirectly with another PHA inclusion body-associated protein PhaC to produce an active PHA synthase. It is required for PHA accumulation [20]. PhaP is characterized according to McCool [21] as a storage protein. It increased during the late lag phase and early to mid-exponential phase, decreased in mid- to late-exponential phase, and increased during stationary phase growth. PhaP belongs to the highly expressed intracellular proteins of B. megaterium as shown on the 2-D gels (Figure 1). In this work both PhaR and PhaP showed changes of expression levels similar to that of MmgA, namely a strong increase during the early cell growth phase both in induced and non-induced batch cultures (Figure 6). The expressions were then leveled off in the later growth phase in the batch culture without xylose induction, whereas in the induced batch culture the expressions increased further up to 6 h after xylose induction, then the expressions of PhaR and PhaP reduced from 6 h to 10 h after induction. Being a storage protein PhaP can be degraded as a source of amino acids for the cells [21]. Therefore, the decreased expression level of PhaP may imply its degradation for overcoming possible nutrient limitations and sustaining the cell growth in the stationary phase. The reason for the difference in expressions of MmgA, PhaR and PhaP between induced and non-induced cultures is not clear.
It is interesting that the phosphate acetyltransferase (PTA) which channels acetyl-CoA through acetyl phosphate into acetate under aerobic conditions was also strongly down-regulated in accordance with the decreased expression of PDH (Figure 6). This was verified by the very low concentration of acetate found in the culture medium. It seems that B. megaterium is in favor of using acetyl-CoA for PHB synthesis than switching on its overflow mechanism for the dissipation of the accumulated acetyl-CoA. This is receivable, since acetate accumulation is toxic for the cell, on the other hand accumulation of PHB is advantageous for the cell to store large quantities of carbon and energy without significantly affecting the osmotic pressure of the cell. In addition, PHB also serves as a sink of reducing power and could be regarded as a redox regulator within the cell. It has been even suggested that PHB may play a role in the regulation of intracellular calcium concentration and in calcium signalling in the plasma membrane of some Gram-positive and Gram-negative bacteria [23]. Indeed, the ability of B. megaterium to accumulate PHB is so dominant that the PHB content in the cells could reach up to 32% of the cell dry weight [19]. For the purpose of recombinant overproduction of dextransucrase it would be of interest to find out whether knock-out of PHB-producing genes could lead to better redirection of energy and carbon sources into the production of the heterologus protein in B. megaterium.
3.3 Regulations of some enzymes involved in amino acid biosynthsis
As shown in Figure 6 four enzymes implicated in the synthesis of branched-chain amino acids valine, leucine and isoleucine using pyruvate as precursor were remarkably up-regulated. They are acetolactate synthase (IlvB), ketol-acid reductoisomerase (IlvC), dihydroxy-acid dehydratase (IlvD) and branched-chain amino acid aminotransferases (IlvE). Since the increases of the expression levels are comparable between the induced and the non-induced cultures, they should be resulted mainly from the demand on amino acids for cell growth. It is worth to mention that expression changes of these enzymes showed very similar time courses to the expression changes of MmgA, PhaR and PhaP as also shown in Figure 6. It has been reported [18] that branched-chain amino acids such as valine and isoleucine can induce the synthesis of phosphotransbutyrylase (YqiS), an enzyme which enhances the accumulation of PHB in B. megaterium. B. megaterium grown in culture media supplemented with valine and isoleucine showed drastically increased accumulation of PHB. Therefore, the enhanced expression levels of these enzymes appear to be related to the PHB synthesis and accumulation as discussed above.
Aspartate aminotransferase (AspB) is the enzyme that catalyzes the conversion of oxaloacetate to aspartate. Aspatate can be further converted to other amino acids of the oxaloacetate family. As shown in Figure 6, the expression of AspB remained at a quite constant level in the non-induced culture, whereas it showed about 50% increase 1 h after xylose induction in the induced culture. This was in agreement with the highest production of the recombinant dextransucrase during this time period as discussed above and can be due to the fact that aspartate, as well as threonine and asparagine of the oxaloacetate family belong to the most needed amino acids for dextransucrase synthesis. Amino acids of oxaloacetate family make together up to 37 % of the amino acid composition of dextransucrase. This demonstrates a high demand on the precursor oxaloacetate. Thus, the proteomic analysis can indicate the metabolic changes as shown for the short-term synthesis of dextransucrase.
The enhanced conversion of oxaloacetate to aspartate might be evidenced by the different expressions of the NAD-dependent malic enzyme MalS in the induced and non-induced cultures (Figure 6). While its expression maintained nearly constant in the non-induced culture, the expression of MalS decreased by 40% 1 h after xylose induction in the induced culture, indicating a reduced channeling of malate to pyruvate. During this time period the expression of malate dehadrogenase (MDH) increased by 160% in the induced culture compared with 36% increase in the non-induced cuture (Figure 5) which might also imply an enhanced conversion of malate to oxaloacetate for amino acid synthesis.
As was reported for B. subtilis [25,26] and B. clausii [27], malic enzymes form a "futile cycle" with pyruvate carboxylase (PycA) and malate dehydrogenase (MDH) (Figure 3). PycA fulfills an important role of catalyzing the anaplerotic reaction to directly replenish oxaloacetate from the pyruvate pool. Although PycA has not yet been identified from 2-DE/MS analysis in this work, our preliminary enzyme activity measurements using the method described by Schröder et al [28] has confirmed the existence of this enzyme in B. megaterium (data not published). It is conceivable that a similar "futile cycle" also exists in the metabolism of B. megaterium. Advantages of having this kind of "futile cycle" is still not clearly understood. It has been assumed that "futile cycle" is important for keeping the metabolic flexibility between the interface of glycolysis and TCA cycle. It plays a key role in anapleosis and metabolic regulations, especially for replenishing the TCA cycle under the stress of enhanced demand on precursors such as oxaloacetate and 2-oxoglutarate for biosynthesis. What an impact this kind of "futile cycle" may have on the production of the recombinant dextransucrase in B. megaterium will be further studied, not merely by proteomic analysis but in combination with metabolic flux analysis.
3.4 Regulations of enzymes related to dextransucrase translocation
It is often reported that microorganisms normally respond to the strong overproductions of recombinant proteins by increased expression levels of chaperones such as GroEL, DnaK and the trigger factor TIG [29-36]. However, from our proteomic analysis as shown in Figure 7 we found expression levels of these chaperones remained quite constant both in xylose-induced culture and in non-induced culture. This supports the observation that dextransucrase produced was readily translocated through the cytoplasmic membrane and therefore, required no over expression of these cytosolic chaperones. On the other hand, we observed an strong up-regulation of the oligopeptide-binding protein OppA in the induced culture compared with that in non-induced culture as also shown in Figure 7. Two isoforms of OppA were identified in this work. Expressions of both isoforms were steadily up-regulated after xylose induction. For a better visualization of the changes of the OppA spots zoomed sections on the 2-D gels are also shown in Figure 7. OppA is normally known as a component of the oligopeptide permease, a binding-protein dependent transport system. It plays an essential role in the uptake of peptides as nutrients and is also required for sporulation and competence for B. subtilis. The increased expression of OppA might be a kind of response of cells to the aggregation and eventual degradation of dextransucrase in the space between cell membrane and cell wall. In addition, competent cells are normally characterized by porous membranes that might facilitate further secretion of dextransucrase into the medium. This could be an explanation of the maximal secretion of dextransucrase happened 9 h after xylose induction as shown in Figure 4.
Figure 7 2-D electrophoresis determined expression changes of the chaperones GroEL, DnaK, TIG as well as the oligopeptide-binding protein OppA (A). Zoomed visualization of the expression alterations of the two OppA spots on the 2-D gels from samples taken from the xylose induced culture. Additional descriptions for this figure are the same as given in Figure 5.
In addition, a chaperone-like property has been also suggested for the periplasmically located OppA of E. coli in addition to its function in transport by Richarme and Caldas [37]. They found that OppA interacts with unfolded and denaturated proteins to form stable complexes. It might bind a notable amount of unfolded proteins until permissive renaturation conditions are restored. Further the authors proposed that other binding proteins, including the substrate-binding lipoproteins of Gram-positive bacteria, should possess similar properties. Therefore, it would be of interest to further study the possible impact of OppA for the secretion of dextransucrase, so as to find out whether OppA is just an indicator of dextransucrase translocation and aggregation between the cytoplasmic membrane and the cell wall or it may act as an extracellular chaperone that can be exploited to help the passage of dextransucrase through the cell wall.
Conclusion
In this work a 2DE method and a protein reference map for the proteomic analysis of B. megaterium were established. Despite the missing of genome sequence 149 individual proteins were identified through public protein database supported homologue protein searching using peptide fragmentation information acquired from ESI-QqTOF MS/MS analysis. Out of them 35 proteins could be assigned to enzyme functions of the central carbon metabolism (glycolysis, pentose phosphate pathway, TCA cycle and pyruvate metabolism) and 31 to amino acid synthesis and metabolism, leading to the construction of a partial metabolic network which is useful for metabolic pathway analysis.
During batch growth of B. megaterium on glucose expressions of glycolytic enzymes remained approximately constant, while most enzymes of the TCA cycle were up-regulated. The components of the PDH complex enzyme as well as phosphate acetyltransferase (PTA) were remarkably down-regulated, whereas some enzymes related to PHB synthesis were strongly up-regulated, indicating a metabolic discrepancy between glycolysis and TCA cycle of this B. megaterium strain and the channeling of acetyl-CoA into the biosynthesis of PHB as a carbon and energy storage source
Except for a few cases the protein expression profiles of the non-induced and induced B. megaterium batch cultures, the latter producing additionally the heterologous protein dextransucrase, did not differ significantly. This indicates that protein expression of B. megaterium concerning the central carbon metabolism was predominantly governed by growth and little affected by the xylose induced generation of the heterologous gene product. Indeed, the mass of dextransucrase was estimated as only 2% of the entire protein produced. Only some enzymes of amino acid synthesis exhibited discrepancies between induced and non-induced cultures. Specifically, the aspartate aminotransferase (AspB) was up-regulated in the induced culture. This enzyme channels oxaloacetate into a large family of amino acids which is strong required (37%) in dextransucrase synthesis. Expression levels of cytosolic chaperones needed for posttranslational processing hardly changed, whereas the oligopeptide-binding protein (OppA) exhibited increased expression in the induced culture, suggesting that this protein may be involved in the translocation of the heterologous dextransucrase.
Materials and methods
Batch cultivation of B. megaterium for recombinant dextransucrase production
The B. megaterium strain MS941 transformed with the plasmid pMM1520dsrS which carries the gene of a dextransucrase from Leuconostoc mesenteroides [38] was chosen as the candidate strain for the investigation on heterologous dextransucrase production. The medium used was developed for an optimal growth of the B. megaterium strain investigated in this study [39]. It contains per liter 33 g glucose monohydate, 5.0 g (NH4)2SO4, 2.2 g KH2PO4, 0.3 g MgSO4·7H2O, 0.5 g yeast extract, 2.0 mL trace element solution and 0.1 mL Sigma antifoam 204 per liter. The trace element solution contained 40 g MnCl2·4H2O, 53 g CaCl2·2H2O, 2.5 g FeSO4·7H2O, 2.5 g (NH4)6Mo7O24·4H2O and 2.5 g CoCl2· 6H2O.
At first, shaking flasks containing 50 mL of the medium were inoculated with a glycerol stock of the B. megaterium strain MS941 carrying the plasmid pMM1520dsrS (B. megaterium MS941dsrS) and cultivated overnight at 37°C and 250 rpm. Upon reaching an OD578 of approximately 3, 10 mL culture were added to 990 mL batch medium in a bioreactor (Biostat B2) and batch culture was carried out at 37°C, pH 5.2, 500 rpm and aeration of 1 L air/min. Analysis of biomass, glucose and metabolites were carried out as described in detail by Hollmann and Deckwer (2004). Biomass as cell dry weight (CDW) was calculated from measured OD value according to a linear relationship between OD and CDW that was determined in our previous works (data not published). Production of recombinant dextransucrase was induced by adding xylose at a concentration of 0.5% w/v after the OD578 reached about 0.3. For comparative 2-DE analysis, another batch culture without xylose induction was carried out under the same conditions.
For 2-DE analysis cell samples were taken from the xylose induced culture immediately before the induction (I0), then 1 h (I1), 6 h (I6) and 10 h (I10) after the induction. Similarly, after the OD578 reached about 0.3 the first cell sample (NI0) was taken from the non-induced culture. Further samples were then taken 1 h (NI1), 6 h (NI6) and 10 h (NI10) after the first sampling. Cells were immediately chilled in ice water after sampling, then centrifuged at 6500 rpm (Sorvall RT 6000B, DuPont) for 30 min at 4°C. Cell pellets were washed twice with phosphate-buffered saline (PBS) solution and stored at -80°C until use.
For the detection of the production and secretion of dextransucrase in B. megaterium, samples taken from the induced culture were centrifuged at 4°C and 5000 rpm for 10 min to obtain cell-free supernatants and cell pellets. The supernatants were centrifuged again at 4°C and 13000 rpm for 10 min to obtain protein sediments which contained the secreted dextransucrase as the secretion fractions. The cell pellets were resuspended with a lysis buffer (100 mM Na3PO4, 5 mg/mL lysozyme, pH 6,5 with H3PO4) and incubated at 37°C for 30 min and centrifuged at at 4°C and 13000 rpm for 10 min to separate the supernatant as the cell-associated soluble fractions from the sediments, which were then resuspended in 8M urea and centrifuged again to collect the supernatant as the cell-associated insoluble fractions. All three fractions were subjected to the determination of the activities of dextransucrase according to an activity staining method described by Malten et al [2]. Briefly, proteins of the three fraction were separated with normal SDS-PAGE gels. DsrS which was still bound to the gels was renaturated and incubated with sucrose to catalyze the formation of dextran. The amount of formed dextran was set into relation to the catalytic activity of DsrS for the quantification of the activity.
Extraction and separation of intracellular protein of B. megaterium by 2-D IEF/SDS-PAGE
The 2-D IEF/SDS-PAGE gel electrophoresis (2-DE) method established in our laboratory for the proteomic analysis of different microorganisms [40] has been optimized for the separation of intracellular proteins of B. megaterium. To obtain raw protein extracts cell pellets were resuspended with a lysis buffer containing 7 M urea, 2 M thiourea, 4% (w/v) CHAPS, 1% (w/v) dithiothreitol (DTT), 0.8% (w/v) Pharmalyte™ pH 3–10, and 5 mM Pefabloc, and disrupted by ultrasonication in a ice bath for 5 × 60 s and 2 × 30 s with a 30 s interval between each ultrasonic cycle for better cooling effect. Insoluble materials were separated by centrifugation at 13,000 g for 30 min at 4°C. For an improved 2-DE separation of B. megaterium proteins, raw protein extracts should be further treated to diminish other interfering cell-intrinsic components. Different purification methods such as membrane dialysis using the Mini Dialysis Kits, 1 kDa cut-off (Amersham Biosciences), or precipitation with various organic solvents have been tested. A phenol precipitation with subsequent acetone extraction resulted in the best 2-DE separation performance. Briefly, raw protein extracts were extracted with a TE-buffer (10 mM EDTA, pH 7.4) saturated phenol by vigorous shaking and incubation. Proteins form a white interphase between the phenolic and the aqueous phases. After discarding the aqueous phase and washing the protein interphase twice with Milli-Q water, proteins were precipitated with cold acetone (-20°C), washed additionally two times with cold acetone, air-dried and stored at -80°C until use.
Aliquots of protein pellets were diluted with an adequate volume of rehydration buffer (7 M urea, 2 M thiourea, 4% (w/v) CHAPS, 1% (w/v) DTT, 0.5% IPG buffer pH 4–7 and trace amount of bromphenol blue). The total protein concentration in the supernatant was determined by the Bradford method using the RotiQuant reagent (Bio-Rad) according to the manufacturer's instruction. Duplicate gels were analyzed for each sample and all the samples from the same batch culture were run under the exactly same conditions. The first-dimensional gel electrophoresis of isoelectric focusing (IEF) was run with the IPGphor Isoelectric Focusing System (Amersham Biosciences) at a temperature of 20°C. 250 μg of each protein sample were loaded onto Immobiline DryStrip gels (IPG strips) of pH 4–7 by in-gel rehydration. IEF was performed with the following parameters: 30 V × 6 h, 60 V × 6 h, 200 V × 1 h, 500 V × 1 h, 1000 V × 1 h, gradient from 1000 V to 8000 V within 30 min, then 8000 V × 10 h. The second-dimensional gel electrophoresis of SDS-PAGE was carried out using the vertical slab separation unit Ettan Dalt II System and pre-cast polyacrylamide gels Ettan Dalt II Gel 12.5% (Amersham Biosciences). Prior to SDS-PAGE the IPG strips were equilibrated in the SDS equilibration buffer as recommended in the user's manual provided by the manufacturer for the pre-cast gels. SDS-PAGE separation was performed at 25°C in constant power mode as follows: 2 W/gel for 1 h and then 20 W/gel until the bromophenol blue dye front reached the bottom of the gel. Subsequently, gels were stained using Brilliant Blue G-Colloidal Concentrate (Sigma, Saint Louis, Missouri, USA) by fixing in a glacial acetic acid : methanol : water (7:40:53) solution for 1 h; staining overnight with Brilliant Blue G-Colloidal according to the manufacturer's instruction and rinsing several times with Milli-Q water. The gels were then scanned with a UMAX PowerLook III scanner at 300 dpi resolution to acquire the gel images. Computer analysis of the gels for protein spot detection, matching and quantification were performed with the Phoretix 2D Advanced Software Version 2003.02 (Phoretix, Newcastle upon Tyne, UK). Exactly same parameters such as sensitivity and operator size were used for spot detection. The same background subtraction method "lowest on boundary" was applied for all gels. User seeds were added to help spot matching between gels. To overcome the problem of electrophoretic variations between gels, minor manual adjustments during spot detection and matching were performed when it was obviously necessary. Average values of protein spot volume intensities and their corresponding standard deviations were calculated to characterize expression changes of proteins. Protein spot intensity was defined as the normalized spot volume which is the ratio of the single spot volume to the total spots volumes on a 2-D gel. Normally only proteins showing more than 2-fold increase or decrease in expressions were considered to be up- or down-regulated.
Protein digestion and mass spectrometric analysis
Protein spots cut out from 2-DE gels were subjected to in-gel digestion with trypsin according to a method described previously [41] with some modification, namely, after rinse with water and dehydration with acetonitrile gel pieces were directly tryptic digested without further treatment. Peptides obtained were then extracted and purified with reversed-phased C18 ZipTips (Millipore, Bedford, USA).
Matrix-assisted laser desorption ionization time-of-flight mass spectrometry (MALDI-TOF MS) with a Bruker Ultraflex time-of-flight mass spectrometer (Bruker Daltonics GmbH, Germany) and nanoelectrospray ionization quadrupole-time-of-flight tandem mass spectrometry (ESI-QqTOF MS/MS) with a Q-TOF 2 mass spectrometer (Micromass, Manchester, England) equipped with a nanospray ion source were carried out as described by Wang et al [41].
Protein identification by homologue protein searching
Peptide masses obtained from MALDI-TOF MS analysis were used for cross-species homologue protein searching in the public protein databases NCBInr and SWISS-PROT/TrEMBL by peptide mass fingerprinting. The search program Mascot (Matrix Science Ltd., UK, ) was used and search parameters were given as follows: trypsin was the digestion enzyme used, one missed cleavage sites was allowed, cysteine was modified by iodoacetamide and methionine was assumed to be partially oxidized. All peptide mass values are monoisotopic and the mass tolerance was set at 100 ppm.
MS/MS spectra of selected peptide precursors from ESI-QqTOF MS/MS analysis were enhanced using the Max Ent 3 software (Micromass), followed by automatic or manual sequencing using the PepSeq program of the software package Masslynx™ Version 3.5 (Micromass). Peptide sequences obtained were merged and submitted for similarity searching of homologous proteins using the protein database search program MS BLAST [7,16] with the scoring matrix PAM30MS and against the comprehensive non-redundant protein sequence database nrdb95. No constraints on protein molecular weights (Mw), isoelectric point (pI) or species of origin were imposed.
Competing interests
The author(s) declare that they have no competing interests.
Authors' contributions
W. Wang designed and carried out 2-DE experiments, did protein identification with MS data and contributed to the preparation of this manuscript. R. Hollmann and T. Fuerch carried out the batch culture experiments and the dextransucrase activity tests. M. Nimtz participated in ESI-QqTOF MS/MS and MALDI-TOF MS analysis as well as in discussions for the preparation of this manuscript. M. Malten and D. Jahn supplied the recombinant B. megaterium strain and were well involved in discussions for the preparation of this manuscript. W.-D. Deckwer initiated and coordinated this study, and contributed to the preparation of this manuscript. All authors have read and approved the final manuscript.
Supplementary Material
Additional File 1
Table 1. Overview of intracellular proteins of Bacillus megaterium MS941dsrS separated by two dimensional gel electrophoresis and identified after in-gel tryptic digestion by ESI-QqTOF MS/MS analysis and homologue protein searching using MS BLAST against the non-redundant protein database nrdb95.
Click here for file
Acknowledgements
This work was financially supported by the Sonderforschungsbereich 578 der Deutschen Forschungsgemeischaft (DFG). The authors thank P. Westphal for her excellent technical assistance in 2-DE experiments. The authors are grateful to the GBF MS analytic group, especially A. Abrahamik, U. Felgenträger and J. Majewski for their help in the ESI-QqTOF MS/MS and MALDI-TOF MS analysis.
==== Refs
Vary PS Prime time for Bacillus megaterium Microbiology 1994 140 1001 13 8025666
Malten M Hollmann R Deckwer W-D Jahn D Production and secretion of recombinant Leuconostoc mesenteroides dextransucrase DsrS in Bacillus megaterium Biotechnol Bioeng 2005 89 206 218 15593264 10.1002/bit.20341
Kunst F Ogasawara N Moszer I Albertini AM Alloni G Azevedo V Bertero MG Bessieres P Bolotin A Borchert S Borriss R Boursier L Brans A Braun M Brignell SC Bron S Brouillet S Bruschi CV Caldwell B Capuano V Carter NM Choi SK Codani JJ Connerton IF Danchin A The complete genome sequence of the Gram-positive bacterium Bacillus subtilis Nature 1997 390 249 56 9384377 10.1038/36786
Hecker M Volker U Towards a comprehensive understanding of Bacillus subtilis cell physiology by physiological proteomics Proteomics 2004 4 3727 50 15540212 10.1002/pmic.200401017
Hecker M A proteomic view of cell physiology of Bacillus subtilis - bringing the genome sequence to life Adv Biochem Eng Biotechnol 2003 83 57 92 12934926
Buttner K Bernhardt J Scharf C Schmid R Mader U Eymann C Antelmann H Volker A Volker U Hecker M A comprehensive two-dimensional map of cytosolic proteins of Bacillus subtilis Electrophoresis 2001 22 2908 35 11565787 10.1002/1522-2683(200108)22:14<2908::AID-ELPS2908>3.0.CO;2-M
Shevchenko A Sunyaev S Liska A Bork P Shevchenko A Nanoelectrospray tandem mass spectrometry and sequence similarity searching for identification of proteins from organisms with unknown genomes Methods Mol Biol 2003 211 221 34 12489434
Shevchenko A Chernushevic I Shevchenko A Wilm M Mann M "De novo" sequencing of peptides recovered from in-gel digested proteins by nanoelectrospray tandem mass spectrometry Mol Biotechnol 2002 20 107 18 11876295 10.1385/MB:20:1:107
Liska AJ Shevchenko A Expanding the organismal scope of proteomics: cross-species protein identification by mass spectrometry and its implications Proteomics 2003 3 19 28 12548630 10.1002/pmic.200390004
Habermann B Oegema J Sunyaev S Shevchenko A The power and the limitations of cross-species protein identification by mass spectrometry-driven sequence similarity searches Mol Cell Proteomics 2004 3 238 249 14695901 10.1074/mcp.M300073-MCP200
Takami H Nakasone K Takaki Y Maeno G Sasaki R Masui N Fuji F Hirama C Nakamura Y Ogasawara N Kuhara S Horikoshi K Complete genome sequence of the alkaliphilic bacterium Bacillus halodurans and genomic sequence comparison with Bacillus subtilis Nucleic Acids Res 2000 28 4317 31 11058132 10.1093/nar/28.21.4317
Takami H Takaki Y Uchiyama I Genome sequence of Oceanobacillus iheyensis isolated from the Iheya Ridge and its unexpected adaptive capabilities to extreme environments Nucleic Acids Res 2002 30 3927 3935 12235376 10.1093/nar/gkf526
Read TD Peterson SN Tourasse N Baillie LW Paulsen IT Nelson KE Tettelin H Fouts DE Eisen JA Gill SR Holtzapple EK Okstad OA Helgason E Rilstone J Wu M Kolonay JF Beanan MJ Dodson RJ Brinkac LM Gwinn M DeBoy RT Madpu R Daugherty SC Durkin AS Haft DH Nelson WC Peterson JD Pop M Khouri HM Radune D Benton JL Mahamoud Y Jiang L Hance IR Weidman JF Berry KJ Plaut RD Wolf AM Watkins KL Nierman WC Hazen A Cline R Redmond C Thwaite JE White O Salzberg SL Thomason B Friedlander AM Koehler TM Hanna PC Kolsto AB Fraser CM The genome sequence of Bacillus anthracis Ames and comparison to closely related bacteria Nature 2003 423 81 6 12721629 10.1038/nature01586
Ivanova N Sorokin A Anderson I Galleron N Candelon B Kapatral V Bhattacharyya A Reznik G Mikhailova N Lapidus A Chu L Mazur M Goltsman E Larsen N D'Souza M Walunas T Grechkin Y Pusch G Haselkorn R Fonstein M Ehrlich SD Overbeek R Kyrpides N Genome sequence of Bacillus cereus and comparative analysis with Bacillus anthracis Nature 2003 423 87 91 12721630 10.1038/nature01582
Rey MW Ramaiya P Nelson BA Brody-Karpin SD Zaretsky EJ Tang M Lopez de Leon A Xiang H Gusti V Clausen IG Olsen PB Rasmussen MD Andersen JT Jorgensen PL Larsen TS Sorokin A Bolotin A Lapidus A Galleron N Ehrlich SD Berka RM Complete genome sequence of the industrial bacterium Bacillus licheniformis and comparisons with closely related Bacillus species Genome Biol 2004 5 R77 15461803 10.1186/gb-2004-5-10-r77
Shevchenko A Sunyaev S Loboda A Shevchenko A Bork P Ens W Standing KG Charting the proteomes of organisms with unsequenced genomes by MALDI-quadrupole time-of-flight mass spectrometry and BLAST homology searching Anal Chem 2001 73 1917 26 11354471 10.1021/ac0013709
Zalabak V Babicka J Chaloupka J Bacillus megaterium as a possible source of protein Mikrobiologiia 1975 44 720 6 809643
Vazquez GJ Pettinari MJ Méndez BS Evidence of an association between poly(3-hydroybutyrate) accumulation and phosphotransbutyrylase expression in Bacillus megaterium Int Microbiol 2003 6 127 129 12827525 10.1007/s10123-003-0120-5
Hori K Kaneko M Tanji Y Xing XH Unno H Construction of self-disruptive Bacillus megaterium in response to substrate exhaustion for polyhydroxybutyrate production Appl Microbiol Biotechnol 2002 59 211 216 12111148 10.1007/s00253-002-0986-8
McCool GJ Cannon MC PhaC and PhaR are required for polyhydroxyalkanoic acid synthase activity in Bacillus megaterium J Bacteriol 2001 183 4235 4243 11418564 10.1128/JB.183.14.4235-4243.2001
McCool GJ Cannon MC Polyhydroxyalkanoate inclusion body-associated proteins and coding region in Bacillus megaterium J Bacteriol 1999 181 585 592 9882674
McCool GJ Fernandez T Cannon MC Polyhydroxyalkanoate inclusion bodies growth and proliferation in Bacillus megaterium FEMS Microbiol Lett 1996 138 41 48 10.1016/0378-1097(96)00079-1
Anderson AJ Dawes EA Occurrence, metabolism, metabolic role, and industrial uses of bacterial polyhydroxyalkanoates Microbiol Rev 1990 54 450 72 2087222
Burdon KL Fatty material in bacteria and fungi revealed by staining dried, fixed slide preparations J Bacteriol 1946 52 665 678 16561232
Fry B Zhu T Koepel RR Phalakornkule C Ataai MM Characterization of growth and acid formation in a Bacillus subtilis pyruvate kinase mutant Appl Environ Microbiol 2000 66 4045 4049 10966427 10.1128/AEM.66.9.4045-4049.2000
Dauner M Storni T Sauer U Bacillus subtilis metabolism and energetics in carbon-limited and excess carbon chemostat culture J Bacteriol 2001 183 7308 7317 11717290 10.1128/JB.183.24.7308-7317.2001
Christiansen T Christensen B Nielsen J Metabolic network analysis of Bacillus clausii on minimal and semirich medium using 13C-labeled glucose Metabolic Eng 2002 4 159 169 10.1006/mben.2001.0219
Schröder C Selig M Schönheit P Glucose fermentation of acetate, CO2 and H2 in the anaerobic hyperthermophilic eubacterium Thermotoga maritima: involvement of the Embden-Meyerhof pathway Arch Microbiol 1994 161 460 470
Dürrschmid K Marzban G Dürrschmid E Striedner G Clementschitsch F Cserjan-Puschmann M Bayer K Monitoring of protein profiles for the optimization of recombinant fermentation process using public domain databases Electrophoresis 2003 24 303 310 12652602 10.1002/elps.200390027
Hoffmann F Weber J Rinas U Metabolic adaptation of Escherichia coli during temperature-induced recombinant protein production: 1. Readjustment of metabolic enzyme synthesis Biotechnol Bioeng 2002 80 313 319 12226864 10.1002/bit.10379
Champion KM Nishihara JC Joly JC Arnott D Similarity of the Escherichia coli proteome upon completion of different biopharmaceutical fermentation processes Proteomics 2001 1 1133 1148 11990508 10.1002/1615-9861(200109)1:9<1133::AID-PROT1133>3.0.CO;2-S
Han MJ Yoon SH Lee SY Proteome analysis of metabolically engineered Escherichia coli producing poly(3-hydroxybutyrate) J Bacteriol 2001 183 301 308 11114930 10.1128/JB.183.1.301-308.2001
Kabir MM Shimizu K Proteome analysis of a temperature-inducible recombinant Escherichia coli for poly-β-hydroxybutyrate production J Biosci Bioeng 2001 92 277 284 16233096 10.1263/jbb.92.277
Juergen B Hanschke R Sarvas M Hecker M Schweder T Proteome and transcriptome based analysis of Bacillus subtilis cells overproducing an insoluble heterologous protein Appl Microbiol Biotechnol 2001 55 326 332 11341315 10.1007/s002530000531
Juergen B Lin HY Riemschneider S Scharf C Neubauer P Schmid R Hecker M Schweder T Monitoring of genes that respond to overproduction of an insoluble recombinant protein in Escherichia coli glucose-limited fermentations Biotechnol Bioeng 2000 70 217 224 10972933 10.1002/1097-0290(20001020)70:2<217::AID-BIT11>3.0.CO;2-W
Rinas U Synthesis rates of cellular proteins involved in translation and protein folding are strongly altered in response to overproduction of basic fibroblast growth factor by recombinant Escherichia coli Biotechnol Prog 1996 12 196 200 8857188 10.1021/bp9600039
Richarme G Caldas TD Chaperone properties of the bacterial periplasmic substrate-binding proteins J Biol Chem 1997 272 15607 15612 9188448 10.1074/jbc.272.25.15607
Malten M Jahn D Produktion und Sekretion von Glukosyltransferasen 2002 German Patent Application No. 102 25 380.3
Hollmann R Deckwer W-D Pyruvate formation and suppression in recombinant Bacillus megaterium cultivation J Biotechnol 2004 111 89 96 15196773 10.1016/j.jbiotec.2004.03.006
Wang W Sun J Hartlep M Deckwer W-D Zeng A-P Combined use of proteomic analysis and enzyme activity assays for metabolic pathway analysis of glycerol fermentation by Klebsiella pneumoniae Biotechnol Bioeng 2003 83 525 536 12827694 10.1002/bit.10701
Wang W Sun J Nimtz M Zeng A-P Deckwer W-D Protein identification from two-dimensional gel electrophoresis analysis of Klebsiella pneumoniae by combined use of mass spectrometry data and raw genome sequences Proteome Sci 2003 1 6 14653859 10.1186/1477-5956-1-6
| 15927046 | PMC1175100 | CC BY | 2021-01-04 16:37:15 | no | Proteome Sci. 2005 May 31; 3:4 | utf-8 | Proteome Sci | 2,005 | 10.1186/1477-5956-3-4 | oa_comm |
==== Front
BMC AnesthesiolBMC Anesthesiology1471-2253BioMed Central London 1471-2253-5-81594638410.1186/1471-2253-5-8Research ArticleInhibition of sarcoplasmic Ca2+-ATPase increases caffeine- and halothane-induced contractures in muscle bundles of malignant hyperthermia susceptible and healthy individuals Schuster Frank [email protected]üller Rainer [email protected] Edmund [email protected] Norbert [email protected] Martin [email protected] Department of Anaesthesiology, University of Wuerzburg Oberduerrbacher Strasse 6, 97080 Wuerzburg, Germany2 Department of Anaesthesiology and Intensive Care, Hospital of Stralsund, 18410 Stralsund, Germany2005 9 6 2005 5 8 8 24 11 2004 9 6 2005 Copyright © 2005 Schuster et al; licensee BioMed Central Ltd.2005Schuster 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
Malignant hyperthermia (MH) is triggered by halogenated anaesthetics and depolarising muscle relaxants, leading to an uncontrolled hypermetabolic state of skeletal muscle. An uncontrolled sarcoplasmic Ca2+ release is mediated via the ryanodine receptor. A compensatory mechanism of increased sarcoplasmic Ca2+-ATPase activity was described in pigs and in transfected cell lines. We hypothesized that inhibition of Ca2+ reuptake via the sarcoplasmic Ca2+-ATPase (SERCA) enhances halothane- and caffeine-induced muscle contractures in MH susceptible more than in non-susceptible skeletal muscle.
Methods
With informed consent, surplus muscle bundles of 7 MHS (susceptible), 7 MHE (equivocal) and 16 MHN (non-susceptible) classified patients were mounted to an isometric force transducer, electrically stimulated, preloaded and equilibrated. Following 15 min incubation with cyclopiazonic acid (CPA) 25 μM, the European MH standard in-vitro-contracture test protocol with caffeine (0.5; 1; 1.5; 2; 3; 4 mM) and halothane (0.11; 0.22; 0.44; 0.66 mM) was performed. Data as median and quartiles; Friedman- and Wilcoxon-test for differences with and without CPA; p < 0.05.
Results
Initial length, weight, maximum twitch height, predrug resting tension and predrug twitch height of muscle bundles did not differ between groups. CPA increased halothane- and caffeine-induced contractures significantly. This increase was more pronounced in MHS and MHE than in MHN muscle bundles.
Conclusion
Inhibition of the SERCA activity by CPA enhances halothane- and caffeine-induced contractures especially in MHS and MHE skeletal muscle and may help for the diagnostic assignment of MH susceptibility. The status of SERCA activity may play a significant but so far unknown role in the genesis of malignant hyperthermia.
==== Body
Background
In skeletal muscle, the action potential passes along the surface membrane of the muscle fibre into the transverse tubular system. Depolarisation of the voltage sensitive dihydropyridine receptor leads to an opening of the ryanodine receptor in the nearby sarcoplasmic reticulum (SR). Sarcoplasmic calcium (Ca2+) release via the ryanodine receptor raises cytosolic Ca2+ and activates muscle contraction. Energy-dependent Ca2+ reuptake into the SR is caused by the SR Ca2+-ATPase (SERCA) and enables skeletal muscle relaxation [1]. In individuals susceptible to the autosomal dominant skeletal muscle disorder malignant hyperthermia (MH), electro-mechanical coupling is disturbed. Due to MH-associated mutations in the ryanodine receptor, triggering agents such as halogenated anaesthetics cause an excessive Ca2+ release from the SR resulting in intracellular hypermetabolism, increased mitochondrial energy-turnover and metabolic failure with a deficiency of adenosine-triphosphate [2]. This may also lead to energetic exhaustion of the SERCA, the main transporter for Ca2+ ions across the sarcoplasmic membrane. Cyctosolic Ca2+ concentration is determined by sarcoplasmic Ca2+ release and it's reuptake via the SERCA [3]. The mycotoxin cyclopiazonic acid (CPA) is a selective inhibitor of SR Ca2+ reuptake [4] that has been used previously to study SERCA in different tissues [5,6].
We hypothesized that in skeletal muscle, preincubation with CPA enhances halothane- and caffeine-induced contractures in MH susceptible (MHS) more than in non-susceptible (MHN) skeletal muscle.
Methods
Muscle bundles of 30 patients undergoing a diagnostic in-vitro contracture test (IVCT) were investigated to detect MH susceptibility. With informed consent, surplus muscle bundles were studied by the same IVCT protocol following SERCA inhibition by CPA.
Muscle biopsy
A muscle biopsy of the vastus lateralis muscle was performed following a femoral nerve block. Muscle bundles were immediately placed in carboxygenated (95% oxygen, 5% carbon dioxide) Krebs-Ringer's solution (NaCl 118.1 mM; KCl 3.4 mM; CaCl2 2.5 mM; MgSO4 0.8 mM; KH2PO4 1.2 mM; NaHCO3 25.0 mM; Glucose 11.1 mM) and transported to the laboratory.
Standard IVCT
In brief, after length and wet weight of each muscle bundle was measured, single muscle strips were mounted vertically in the experimental bath perfused with carboxygenated Krebs-Ringer's solution at 37°C, fixed to an isometric force transducer (Lectromed Type 4150, UK) and stimulated electrically with a supramaximal square wave stimulus at 1 ms duration and a frequency of 0.2 Hz (Hugo-Sachs-Elektronik, Type 215/I, Germany). Resting tension and twitch height of the muscle strips were recorded continuously by a digital recording system (MusCo, RS BioMed, Germany). After equilibration, caffeine (Sigma Chemicals, Germany) respectively halothane (Abott, Germany) were given at increasing concentrations of 0.5; 1; 1.5; 2; 3; 4; and 32 mM respectively 0.11; 0.22; 0.44 and 0.66 mM at 3 min intervals. A contracture < 2 mN at caffeine 2 mM and halothane 0.44 mM was classified MHN. A stronger contracture following only one of both drugs lead to the diagnosis MH equivocal (MHE). If both drugs developed a significant contracture the patient was assigned as MHS. Investigations were performed within 5 hours after muscle biopsy [7].
CPA-IVCT
A modified contracture test was carried out studying the drug CPA (M = 336.38 g mol-1) that was prepared in a stock solution at 2.5 mM dissolved in dimethylsulphoxide 0,5% (DMSO) (all Sigma Chemicals, Germany). Following equilibration as described above, muscle bundles were incubated with CPA 25 μM for 15 min. The contracture test was then carried out as described above.
Statistics
Data are shown as median and quartiles. IVCT results of skeletel muscle contractures with CPA were statistically evaluated in comparison to the results without CPA by using the Friedman- and Wilcoxon-test for differences with and without CPA. p < 0.05 was considered significant.
Results
Thirty patients, 9 female and 21 male, with a mean age of 28 (15 – 32) years and a mean weight of 74 (62 – 87) kg were studied. 7 patients were classified as MHS, 7 as MHEh (susceptible only for halothane) and 16 as MHN according to the criteria of the European Malignant Hyperthermia group diagnostic protocol. In every patient, an additional IVCT with CPA was performed. Muscle bundles used for the IVCT and CPA-IVCT did not differ regarding to length, weight, maximum twitch height, predrug resting tension and predrug twitch height (Table 1).
Table 1 Biometric data of muscle bundles used for the In-vitro Contracture-Test without (IVCT) and with preincubation with cyclopiazonic acid (CPA-IVCT); median and quartiles.
IVCT CPA-IVCT
Length (mm) 18 (16 – 20) 18 (15 – 20)
Weight (mg) 220 (190 – 233) 205 (180 – 240)
Maximum twitch height (mN) 24 (22 – 27) 25 (23 – 27)
Predrug resting tension (mN) 11 (9 – 14) 11 (8 – 14)
Predrug twitch height (mN) 57 (37 – 75) 42 (16 – 82)
In the caffeine contracture test, prior incubation with CPA resulted in significant higher contractures compared to the diagnostic IVCT in the MHS and MHEh group (Table 2). At the diagnostic threshold dose of caffeine 2 mM, MHS muscles developed significantly higher contractures with 32 (25 – 38) mN following preincubation with CPA vs. 8 (4 – 12) mN without CPA. In the MHEh group CPA preincubation lead to significantly higher contractures with 12 (11 – 27) mN vs. 1 (0 – 1) mN without CPA, while the contractures of MHN muscle bundles did not differ with or without CPA.
Table 2 Caffeine-induced contractures with and without preincubation by cyclopiazonic acid 25 μM (CPA); median and quartile; * p < 0.05 for differences with CPA and without CPA.
Caffeine [mM] 0.5 1 1.5 2 3 4 32
MHS [mN] 1 (0 – 1) 1 (1 – 1) 3 (1 – 6) 8 (4 – 12) 20 (15 – 31) 21 (14 – 35) 171 (136 – 137)
MHSCPA [mN] 3 (2 – 8)* 14 (11 – 31)* 27 (17 – 47)* 32 (25 – 38)* 31 (27 – 37) 34 (29 – 38) 131 (108 – 191)
MHEh [mN] 1 (1 – 2) 1 (1 – 1) 0 (0 – 0) 1 (0 – 1) 2 (1 – 3) 4 (3 – 7) 127 (108 – 190)
MHEhCPA [mN] 0 (0 – 1) 3 (0 – 5) 7 (3 – 19)* 12 (11 – 27)* 31 (29 – 37)* 46 (31 – 47)* 199 (156 – 227)
MHN [mN] 1 (0 – 1) 1 (0 – 1) 1 (1 – 1) 1 (0 – 1) 1 (0 – 1) 1 (0 – 2) 158 (108 – 176)
MHNCPA [mN] 2 (1 – 2) 2 (1 – 2) 1 (1 – 5) 1 (1 – 6) 4 (1 – 21)* 17 (4 – 27)* 167 (153 – 180)
At halothane 0.44 mM, CPA preincubation increased contractures of MHS and MHEh muscle bundles significantly to 59 (33 – 73) mN respectively 45 (24 – 55) mN compared to standard IVCT conditions with 20 (16 – 26) mN respectively 4 (2 – 4) mN. In addition, in the MHN group at halothane 0.44 mM contractures were significantly increased by CPA preincubation to 16 (4 – 34) mN vs. 1 (1 – 1) mN without CPA (Table 3).
Table 3 Halothane-induced contractures with and without cyclopiazonic acid 25 μM (CPA) pre-incubation; median and quartile; * p < 0.05 for differences between IVCT and CPA-IVCT.
Halothane [mM] 0.11 0.22 0.44 0.66
MHS [mN] 5 (2 – 6) 14 (13 – 20) 20 (16 – 26) 19 (11 – 24)
MHSCPA [mN] 40 (26 – 58)* 52 (29 – 76)* 59 (33 – 73)* 48 (30 – 57)*
MHEh [mN] 0 (0 – 0) (0 – 1) 4 (2 – 4) 3 (2 – 4)
MHEhCPA [mN] 7 (1 – 11)* 25 (13 – 39)* 45 (24 – 55)* 45 (25 – 48)*
MHN [mN] 1 (0 – 2) 1 (0 – 1) 1 (1 – 1) 1 (0 – 1)
MHNCPA [mN] 1 (0 – 3) 2 (0 – 25) 16 (4 – 34)* 20 (9–31)*
Discussion
In MH uncontrolled SR Ca2+ release, caused by MH associated mutations mainly in the ryanodine receptor gene, is widely accepted as the underlying pathophysiological mechanism of hypermetabolism [8]. However, the detection of a mutation in the alpha 1-subunit of the voltage sensitive dihydropyridine receptor in a French MH family suggests a more complex pathogenesis of MH [9]. According to the unique mechanism of intracellular Ca2+ cycling that induces contraction and relaxation in vertebrate skeletal muscle, sarcoplasmic Ca2+ release and sarcoplasmic Ca2+ reuptake determine the mainstays of Ca2+ regulation. Undoubtly, an altered SR Ca2+ release plays a crucial role in the development of MH. However, it is completely unclear why many MHS individuals may suffer from MH only after several uneventful exposures to trigger agents during anaesthesia. Several modulating factors have been postulated to modulate cytosolic Ca2+ concentrations, e.g. magnesium [10], sympathetic activity [11], temperature [12], volatile anesthetics [13] or channel's redox state [14]. While SR Ca2+ release was extensively studied in MH [15], the impact of an altered SR Ca2+ reuptake on the pathogenesis of MH by intrinsic or extrinsic factors is poorly understood. Theoretically, a reduced activity of the skeletal muscular SERCA type 1 may result in an elevated cytosolic Ca2+ level due to a persistent slow Ca2+ efflux out of the SR that is otherwise balanced by reuptake [16]. A critical threshold of cytosolic Ca2+ may then be exceeded and may lead to contracture development in vitro and to the MH syndrome in susceptible patients. Interestingly, CPA alone did not induce skeletal muscle contractures at 25 μM [17]. We assume that in our study SERCA was inhibited almost completely, since CPA 10 μM reduced the SERCA activity approximately by 70% in frog skinned fibres [16] and nearly by 100% in rat skinned fibres [18].
In the presented study, CPA preincubation lead to a high variability of halothane- respectively caffeine-induced contractures especially in the MHS and MHEh group, despite SERCA distribution does not differ between MHS and MHN muscle [19]. Interestingly, the response of MHEh muscle bundles to caffeine was enhanced by CPA preincubation. However, at this stage, our results do not suggest CPA as an alternative approach to improve differentiation of MHE from MHN respectively MHS individuals.
Ca2+ uptake capacity and SERCA activity was found to be significantly increased in MHS pigs [20] and in HEK-293 cells transfected by MH mutants [21] but was described to be lower in MHS muscle compared to normal human skeletal muscle [22]. Since a leaky ryanodine receptor in MHS individuals may lead to increased cytosolic calcium, it looks feasible that SR-Ca-ATPase may be upregulated by a compensatory mechanism.
Another option is that CPA itself modulates directly the effect of the trigger agent. This is less likely since halothane and caffeine do have different binding sites at the sarcoplasmic membrane [23].
The role of a reduced SERCA activity in the pathogenesis of Brody's disease, a skeletal muscular myopathy, is well known and characterized by painless muscle cramping and exercise-induced muscle stiffness linked to a mutation in the gene encoding SERCA [24,25]. The left-shift of the dose-response curve for halothane- and caffeine-induced contractures following inhibition of the sarcoplasmic Ca2+ reuptake by CPA points out the essential part of SERCAs in the regulation of cytoplasmic Ca2+. We believe this may be an explanation why some MH susceptible patients develop a MH crisis while others never or only after several trigger exposures suffer from MH despite a proven in vitro susceptibility. In this context, an altered activity of SERCA due to intrinsic or extrinsic factors may play a crucial role in the evolution of MH.
Conclusion
The present study demonstrates that CPA preincubation enhances halothane- and caffeine-induced muscle contractures in the IVCT of MHS, MHEh more than in MHN patients.
Modulation of SERCA may play a significant role in the development of malignant hyperthermia. Patients with a high activity may compensate an increased Ca2+ release or leakage from the SR while patients with a low activity of the SERCA do not. Further investigations with focus on extrinsic and intrinsic factors that modulate SERCA activity may be helpful to understand why MH patients may have had several anaesthesias including trigger agents without a significant reaction while developing a fulminate MH crisis at another occasion.
Abbreviations
Ca2+ Calcium
CPA Cyclopiazonic acid
IVCT In-Vitro Contracture Test
MH Malignant hyperthermia
MHEh Malignant hyperthermia equivocal; susceptible only for halothane
MHN Malignant hyperthermia non-susceptible
MHS Malignant hyperthermia susceptible
SERCA Sarcoplasmic calcium adenosine triphosphatase
SR Sarcoplasmic reticulum
Authors' contributions
FS collected and analysed the data and drafted the manuscript. RM collected data and performed the statistical analysis. EH conceived the study. NR participated in the design of the study. MA designed the study protocol, accompanied the data acquisition and helped writing the manuscript. All authors read and approved the final manuscript.
Pre-publication history
The pre-publication history for this paper can be accessed here:
Acknowledgements
The study was performed at the Department of Anesthesiology at the University of Wuerzburg, Germany.
==== Refs
Berchtold MW Brinkmeier H Müntener M Calcium ion in skeletal muscle: Its crucial role for muscle function, plasticity, and disease Physiol Rev 2000 80 1215 1265 10893434
Gronert GA Antognini JF Pessah IN Miller RD Malignant Hyperthermia Anesthesia 2000 5 Philadelphia: Churchill Livingstone 1033 1052
Gommans IMP Vlak MHM De Haan A Van Engelen BGM Calcium regulation and muscle disease J Muscle Res Cell Motil 2002 23 59 63 12363286 10.1023/A:1019984714528
Seidler NW Joan I Vegh M Martonosi A Cyclopiazonic acid is a specific inhibitor of the Ca2+ ATPase of the sarcoplasmic reticulum J Biol Chem 1989 264 17816 17823 2530215
Duke AM Steele DS Effects of cyclopiazonic acid on Ca2+ regulation by the sarcoplasmic reticulum in saponin-permeabilized skeletal muscle fibres Eur J Physiol 1998 436 104 111 10.1007/s004240050610
Enzmann NR Balog EM Gallant EM Malignant Hyperthermia: Effects of sarcoplasmic reticulum Ca2+ pump inhibition Muscle Nerve 1998 21 361 366 9486865 10.1002/(SICI)1097-4598(199803)21:3<361::AID-MUS10>3.0.CO;2-2
The European Malignant Hyperpyrexia Group A protocol for the investigation of malignant hyperpyrexia (MH) susceptibility Br J Anaesth 1984 56 1267 1269 6487446
Urwyler A Deufel T McCarthy T West S for the European Malignant Hyperthermia Group Guidelines for the molecular genetic detection of susceptibility to malignant hyperthermia Br J Anaesth 2001 86 283 287 11573677 10.1093/bja/86.2.283
Monnier N. Procaccio V. Stieglitz P. Lunardi J. Malignant-hyperthermia susceptibility is associated with a mutation of the alpha 1-subunit of the human dihydropyridine-sensitive L-type voltage-dependent calcium-channel receptor in skeletal muscle Am J Hum Genet 1997 60 1316 1325 9199552
Meissner G Henderson JS Rapid calcium release from cardiac sarcoplasmic reticulum vesicles is dependent on Ca2+ and is modulated by Mg2+, adenine nucleotide, and calmodulin J Biol Chem 1987 262 3065 2434495
Gronert GA Milde JH Theye RA Role of sympathetic activity in porcine malignant hyperthermia Anesthesiology 1997 47 411 415 911050
Nelson TE Porcine malignant hyperthermia: critical temperatures for in vivo and in vitro responses Anesthesiology 1990 73 411 5
Kunst G Graf BM Schreiner R Martin E Fink RH Differential effects of sevoflurane, isoflurane and halothane on Ca2+ release from the sarcoplasmic reticulum of skeletal muscle Anesthesiology 1999 91 179 86 10422943 10.1097/00000542-199907000-00026
Xia R Stangler T Abramson JJ Skeletal muscle ryanodine receptor is a redox sensor with a well defined redox potential that is sensitive to channel modulators J Biol Chem 2000 275 36556 61 10952995 10.1074/jbc.M007613200
MacLennan DH Phillips MS Malignant hyperthermia Science 1992 256 789 194 1589759
Du GG Ashley CC Lea TJ Effects of thapsigargin and cyclopiazonic acid on the saroplasmatic reticulum Ca2+ pump of skinned fibres from skeletal muscle Pflugers Arch 1994 429 169 175 7892102 10.1007/BF00374309
Anetseder M Sixt S Hartung E Cyclopiazonic acid increases halothane induced contractures Minerva Anestesiologica 1994 60 59 64
Kurebayashi N Ogawa Y Discrimination of Ca2+ ATPase activity of the sarcoplasmic reticulum from actomyosin-type ATPase activity of myofibrils in skinned mammalian skeletal muscle fibres: distinct effects of cyclopiazonic acid on the two ATPase activities J Muscle Res Cell Motil 1991 12 355 365 1834695 10.1007/BF01738590
Everts ME Ørding H Hansen O Nielsen PA Ca(2+)-ATPase and Na(+)-K(+)-ATPase content in skeletal muscle from malignant hyperthermia patients Muscle Nerve 1992 15 162 167 1312675 10.1002/mus.880150206
O'Brien PJ Porcine malignant hyperthermia susceptibility: increased calcium-sequestering activity of skeletal muscle sarcoplasmic reticulum Can J Vet Res 1986 50 329 37 3742368
Tong J McCarthy TV MacLennan DH Measurement of resting cytosolic Ca2+ concentrations and Ca2+ store size in HEK-293 cells transfected with malignant hyperthermia or central core disease mutant Ca2+ release channels J Biol Chem 1999 274 693 702 9873004 10.1074/jbc.274.2.693
Conrescu M Lopez JR Medina P Alamo L Deficient function of the sarcoplasmic reticulum in patients susceptible to malignant hyperthermia Muscle Nerve 1987 10 238 241 2951595 10.1002/mus.880100307
Zucchi R Ronca-Testoni S The sarcoplasmic reticulum Ca2+ channel/ryanodine receptor: Modulation by endogenous effectors, drugs and disease states Pharmacol Rev 1997 49 1 51 9085308
Brody IA Muscle contracture induced by exercise. A syndrome attributable to decreased relaxing factor N Engl J Med 1969 281 187 192 4239835
Odermatt A Taschner PE Khanna VK Busch HF Karpati G Jablecki CK Breuning MH MacLennan DH Mutations in the gene-encoding SERCA1, the fast-twitch skeletal muscle sarcoplasmic reticulum Ca2+ ATPase are associated with Brody disease Nat Genet 1996 14 191 194 8841193 10.1038/ng1096-191
| 15946384 | PMC1175794 | CC BY | 2021-01-04 16:28:05 | no | BMC Anesthesiol. 2005 Jun 9; 5:8 | utf-8 | BMC Anesthesiol | 2,005 | 10.1186/1471-2253-5-8 | oa_comm |
==== Front
PLoS BiolPLoS BiolpbioplosbiolPLoS Biology1544-91731545-7885Public Library of Science San Francisco, USA 1600850210.1371/journal.pbio.0030250Research ArticleEvolutionMolecular Biology/Structural BiologyBiochemistryEubacteriaA Conserved Mechanism for Sulfonucleotide Reduction Mechanism of Sulfonucleotide ReductionCarroll Kate S
1
Gao Hong
1
2
Chen Huiyi
3
Stout C. David
4
Leary Julie A
2
Bertozzi Carolyn R [email protected]
1
2
5
1Department of Chemistry, University of California, Berkeley, California, United States of America,2Departments of Chemistry and Molecular Cell Biology, Genome Center, University of California, Davis, California, United States of America,3Department of Molecular and Cell Biology, University of California, Berkeley, California, United States of America,4Department of Molecular Biology, The Scripps Research Institute, La Jolla, California, United States of America,5Howard Hughes Medical Institute, University of California, Berkeley, California, United States of AmericaMatthews Rowena G. Academic EditorUniversity of MichiganUnited States of America8 2005 19 7 2005 19 7 2005 3 8 e2508 3 2005 12 5 2005 Copyright: © 2005 Carroll 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.
Reducing the Mysteries of Sulfur Metabolism
Sulfonucleotide reductases are a diverse family of enzymes that catalyze the first committed step of reductive sulfur assimilation. In this reaction, activated sulfate in the context of adenosine-5′-phosphosulfate (APS) or 3′-phosphoadenosine 5′-phosphosulfate (PAPS) is converted to sulfite with reducing equivalents from thioredoxin. The sulfite generated in this reaction is utilized in bacteria and plants for the eventual production of essential biomolecules such as cysteine and coenzyme A. Humans do not possess a homologous metabolic pathway, and thus, these enzymes represent attractive targets for therapeutic intervention. Here we studied the mechanism of sulfonucleotide reduction by APS reductase from the human pathogen Mycobacterium tuberculosis, using a combination of mass spectrometry and biochemical approaches. The results support the hypothesis of a two-step mechanism in which the sulfonucleotide first undergoes rapid nucleophilic attack to form an enzyme-thiosulfonate (E-Cys-S-SO3
−) intermediate. Sulfite is then released in a thioredoxin-dependent manner. Other sulfonucleotide reductases from structurally divergent subclasses appear to use the same mechanism, suggesting that this family of enzymes has evolved from a common ancestor.
A diverse family of enzymes that catalyze the first step in sulfur assimilation share the same mechanism.
==== Body
Introduction
Carbon, nitrogen, and sulfur undergo energy cycles in our environment that are essential to support life. While carbon and nitrogen are available primarily as gases that require fixation, sulfur occurs abundantly as inorganic sulfate. Despite this apparent advantage, a substantial energetic hurdle must still be overcome to convert the sulfur in its most oxidized form into sulfide, the oxidation state of sulfur required for the synthesis of essential biomolecules [1]. This transformation is accomplished by a group of enzymes that, together, make up the sulfate assimilation pathway [2].
We have studied the Mycobacterium tuberculosis sulfur assimilation pathway (Figure 1) as a possible venue for novel therapeutic targets [3]. In M. tuberculosis, ATP sulfurylase (Figure 1, reaction a) activates inorganic sulfate via adenylation. The resulting compound, adenosine 5′-phosphosulfate (APS), contains a unique high-energy anhydride bond. This vital intermediate resides at a metabolic hub within M. tuberculosis. In the nonreductive branch of sulfate metabolism, APS is phosphorylated a second time by APS kinase (Figure 1, reaction b) to form 3′-phosphoadenosine 5′-phosphosulfate (PAPS), the universal sulfate donor for sulfotransferases [4]. Alternatively, for the purposes of thiol or sulfide-containing metabolite production mentioned above, APS is shuttled through the reductive branch of this pathway (Figure 1, reactions c1, d, and e).
Figure 1 Routes of Sulfate Assimilation
Inorganic sulfate is adenylated by ATP sulfurylase (reaction a) to form APS. Higher plants and the majority of sulfate reducing bacteria use APS as their source of sulfite (reactions c1 → d → e). In some organisms, APS kinase (reaction b) phosphorylates APS at the 3′-hydroxyl group to form PAPS for use as a sulfate donor for sulfotransferases or as a source of sulfite. The lower pathway of sulfate reduction (reactions c2 → d → e) is utilized by γ-proteobacteria such as E. coli and some fungi. Depending on the organism, APS or PAPS is reduced to sulfite by APS reductase (reaction c1) and PAPS reductase (reaction c2), respectively. Sulfite is reduced to sulfide by sulfite reductase (reaction d) and incorporated into cysteine by O-acetylserine-(thiol) lyase (reaction e). Important metabolites such as methionine and coenzyme A are, in turn, synthesized from cysteine.
In mycobacteria, APS reductase (Figure 1, reaction c1) catalyzes the first committed step in the biosynthesis of reduced sulfur compounds (Figure 2). In this reaction, APS is reduced to sulfite and adenosine 5′-phosphate (AMP) [5]. The input of electrons, or reduction potential, necessary for this reaction is provided by thioredoxin, a 12-kDa ubiquitous protein cofactor [6–8]. Thioredoxin contains a -Cys-XX-Cys- motif, and together these two cysteine residues form a redox-active disulfide bond [9]. While other interesting functions for thioredoxin have been reported, its most commonly known role is participation in thiol/disulfide exchange reactions, in which it donates electrons to proteins with concomitant oxidation of its own thiols. Notably, APS reductase has been identified in a screen for essential genes in Mycobacterium bovis bacillus Calmette-Guérin [10]. Furthermore, in a murine model of tuberculosis infection, APS reductase is critical for virulence (R. Senaratne, personal communication). Since humans do not possess an analogous enzyme, from a therapeutic standpoint, APS reductase represents a unique target for antibiotic therapy.
Figure 2 Reaction Catalyzed by Sulfonucleotide Reductases
The requirement for reduced sulfur in the biosynthesis of cysteine, methionine, and other primary metabolites is widespread among many organisms, such as algae, plants, fungi, and a diverse array of bacteria [11–13]. However, variations within this metabolic pathway exist, and interestingly, not all organisms reduce APS. In a subtle change of substrate specificity, some organisms reduce PAPS (Figure 1, reaction c2) as the source of sulfite (Figure 2) [14,15]. Known collectively as sulfonucleotide reductases, this enzyme family can be subdivided into three groups based on their substrate specificity, cofactors, and domain organization (Figure 3A) [11,12,16].
Figure 3 Domain Organization and Cysteine Conservation within the Sulfonucleotide Reductase Family
(A) APS reductases from higher plants (group a) possess a reductase domain and a unique C-terminal domain with high homology to thioredoxin. Bacterial APS reductases (group b) lack this specialized domain, but share the cysteine motif -CC-X∼80-CXXC- present in the reductase domain. PAPS reductases (group c) lack this cysteine motif as well as the thioredoxin domain. All sulfonucleotide reductases have a cysteine at the end of the C terminus in the reductase domain. This residue is essential for catalysis.
(B) A dendrogram illustrating the sequence homology between enzymes within the sulfonucleotide reductase family. Each of the three subclasses of sulfonucleotide reductases is clearly delineated: APS reductases from higher plants with their unique C-terminal thioredoxin domain (A. thaliana and L. minor), bacterial APS reductases (M. tuberculosis, P. aeruginosa, and R. meliloti) and PAPS-reducing organisms (E. coli and S. cerevisae).
Burkholderia cepacia, Shigella flexneri,
B. subtilis, and Bacillus anthracis were included for comparison. The sequence alignment was performed using ClustalW and the tree was constructed with the Drawtree program [53].
The first category of reductases consists of higher plants (Figure 3A, group a), such as Arabidopsis thaliana and Lemina minor [13,17,18]. These enzymes reduce APS. The main characteristic that distinguishes this group from other sulfonucleotide reductases is the presence of a 16 kDa domain fused to the C-terminal end of the reductase, with an extremely high degree of homology to thioredoxin—in particular, the two redox-active cysteines. Notably, these enzymes also possess an iron-sulfur cluster consisting of four iron and four inorganic sulfur atoms (4Fe-4S) [18]. This cluster is ligated by the cysteines found in the sequence motif -CC-X∼ 80-CXXC-, although at present it is not clear whether three or all four cysteines within this motif participate in iron-sulfur cluster ligation [12,18]. On the other end of this continuum are PAPS reductases (Figure 3A, group c) found in γ-proteobacteria and in fungi, such as Escherichia coli and Saccharomyces cerevisiae, respectively [14,15]. In addition to being distinct from the first class of reductases based on substrate specificity, these enzymes lack the iron-sulfur cluster cysteine motif and are thus devoid of this cofactor.
More recently, a new subclass of sulfonucleotide reductases was discovered, designated as bacterial APS reductases (Figure 3A, group b) [11]. Representative examples from this group are found in organisms such as Mycobacterium tuberculosis, Pseudomonas aeruginosa, and Rhizobium meliloti. Interestingly, another member of this subclass, from Bacillus subtilis, can reduce PAPS in addition to APS [19]. These bacterial enzymes possess a higher degree of homology to plant APS reductases than to the bacterial PAPS reductases, particularly in their absolute conservation of the iron-sulfur cluster cysteine motif and 4Fe-4S cluster [12]. A dendrogram that illustrates the amino acid sequence relationship between the sulfonucleotide reductases described above is depicted in Figure 3B. Their segregation at the primary sequence level mirrors their substrate and cofactor distinctions.
Two fundamental questions remain unanswered for this family of catalysts, each subgroup with its distinct cofactors, substrate specificity and domain organization. Namely, at a molecular level, how do sulfonucleotide reductases catalyze this unique reaction and, on a broader level, do sulfonucleotide reductases utilize similar or distinct catalytic mechanisms? APS and PAPS reductases share a high degree of sequence homology and possess an absolutely conserved cysteine residue at the C terminus of the reductase domain (Figure 3A). This cysteine residue does not play a role in iron-sulfur cluster ligation, and mutation of this cysteine to serine abolishes detectable catalytic activity [12,14,18]. From these data, there is a general consensus on the essential catalytic role of this cysteine residue.
In early studies using S. cerevisiae and E. coli PAPS reductases, steady-state kinetics yielded parallel lines in Lineweaver-Burk double reciprocal plots [14,15]. Such results can be indicative of a “ping-pong” reaction mechanism. The term ping-pong is used describe a mechanism in which the enzyme first reacts with one substrate to give a covalently modified enzyme and releases one product [20,21]. The modified enzyme then reacts with a second substrate in a subsequent step. In the case of sulfonucleotide reductases, one substrate can be thought of as the sulfonucleotide and the other, thioredoxin. For simplicity, throughout the text we use the term “first step” to refer collectively to the steps that lead up to the covalently modified enzyme and the term “second step” to refer collectively to the steps subsequent to this modification.
Several mechanisms have been proposed for the reduction of both APS and PAPS. The fundamental difference between these mechanisms lies in the specific role of thioredoxin during the catalytic cycle. Initially, a model was hypothesized in which thioredoxin would act in the first step of the reaction catalyzed by PAPS reductase to reduce an intermolecular disulfide bond proposed to form between the two C-terminal cysteine residues of the reductase homodimer (Figure 4A) [14]. One of the liberated thiolates would subsequently execute a nucleophilic attack on the sulfur atom in PAPS, generating an enzyme-thiosulfonate intermediate, E-Cys-S-SO3
−. In the second step, nucleophilic attack by the second thiolate would generate a sulfite product with concomitant reoxidation of the intermolecular disulfide bond.
Figure 4 Previously Proposed Models of Sulfonucleotide Reduction
(A) Prior to covalent intermediate formation, thioredoxin (Trx) reduces an intermolecular disulfide bond [14,15].
(B) Prior to covalent intermediate formation, thioredoxin (Trx) reduces an intramolecular disulfide bond [22].
In a more recent study, it was proposed that APS reductase from P. aeruginosa, and likely all bacterial APS reductases, employ a slight variant of the PAPS reductase catalytic strategy (Figure 4B) [22]. In this work, Kim and colleagues suggested that thioredoxin reduces an intramolecular disulfide bond proposed to form between the C-terminal thiol in the reductase domain and the second cysteine in the -CC- sequence motif. The remainder of the reaction would proceed as previously put forward for PAPS reductase; the partner thiolate in the reduced disulfide bond would attack the thiosulfonate intermediate to regenerate the intramolecular disulfide and yield sulfite product.
An alternative mechanism has been suggested for the reduction of APS in higher plants such as A. thaliana and related organisms that possess the C-terminal thioredoxin domain [23]. In this model, APS binds to the reductase in the first step and forms an enzyme-thiosulfonate intermediate via the essential C-terminal cysteine. In a subsequent step, the authors propose that the thioredoxin domain would be responsible for the release of bound sulfite. In a series of experiments using only the N-terminal APS reductase domain of the enzyme, Weber and colleagues convincingly established the presence of a covalent sulfite adduct at the C-terminal cysteine in the reductase domain [23]. Demonstration of a specific requirement for thioredoxin in product release remained more elusive, as the N-terminally truncated enzyme was only 3-fold more active in the presence of recombinant thioredoxin m, as compared to dithiothreitol (DTT) alone. A second thioredoxin, thioredoxin f, provided no catalytic advantage over DTT. Furthermore, as this small group of reductases from higher plants possessed different domain organization from the bulk of the enzyme family, it was unclear if the results from this work could be more generally applied.
Herein, we describe experiments that draw upon a combination of approaches—biophysical analysis, site-directed mutagenesis, kinetics, cysteine labeling, and direct analysis of enzyme intermediates by Fourier transform ion-cyclotron resonance (FT-ICR) mass spectrometry—to address the question of mechanism in sulfonucleotide reductases. Our results demonstrate that M. tuberculosis APS reductase catalyzes the reduction of APS in a two-step process: In the first step, APS undergoes rapid nucleophilic attack to form a unique enzyme-thiosulfonate intermediate. In the second step, thioredoxin mediates sulfite release.
We also investigated the mechanisms of other bacterial APS reductases, including that from P. aeruginosa. In contrast to previous suggestions, all bacterial APS reductases we studied utilize the same general mechanism as we found for M. tuberculosis APS reductase. Additionally, we have reexamined the mechanism proposed for PAPS reduction by E. coli PAPS reductase. Like the APS reductases, E. coli PAPS reductase forms an enzyme-thiosulfonate intermediate and employs the same overall catalytic strategy as all other sulfonucleotide reductases investigated in this study. Taken together, these data suggest that the family of sulfonucleotide reductases share the same general catalytic strategy and provide functional evidence for evolution from a common ancestor.
Results
The first mechanism proposed for sulfonucleotide reduction was generated from studies with S. cerevisiae and E. coli PAPS reductases [14,15]. Both enzymes were isolated as homodimers and possessed a single cysteine at their C termini. It was proposed that reduction of a dynamic intermolecular disulfide bond formed between the two C-terminal cysteine residues initiated the catalytic cycle (Figure 4A). One of the liberated thiolates would attack the substrate sulfur atom to form a thiosulfonate intermediate. Subsequent attack by the second cysteine upon this intermediate would regenerate the enzyme disulfide and produce sulfite. There are four clear predictions from this model: (1) Sulfonucleotide reductases should be homodimers; (2) In the first step of the reaction, thioredoxin should reduce an intermolecular disulfide bond between the catalytic cysteines; (3) A thiosulfonate enzyme intermediate should be formed; and (4) The second thiolate should be essential for regeneration of the proposed disulfide bond and the production of sulfite. As M. tuberculosis APS reductase is a compelling target for drug development, we wanted to test the various features of this mechanism in order to better inform our efforts at rational drug design.
M. tuberculosis APS Reductase Is a Monomer
We set out to confirm that M. tuberculosis APS reductase was a dimer, but to our surprise, size exclusion chromatography of M. tuberculosis APS reductase indicated that this enzyme was a monomer (Figure S1A). Under the same conditions, E. coli PAPS reductase eluted exclusively as the homodimer, as reported in previous work (Figure S1A) [14]. To rule out the possibility of anomalous migration on the gel filtration column, we used analytical ultracentrifugation to determine the molecular weight of M. tuberculosis APS reductase. Figure 5 depicts representative data from these experiments from a sample containing 10 μM enzyme. Analysis of these data yielded an estimated molecular weight for M. tuberculosis APS reductase of 28,513 Daltons (Da). This value is in excellent agreement with the expected molecular weight of the holoenzyme, 28,705.88 Da. As an independent confirmation of these results, an electrospray ionization (ESI) mass spectrum was acquired for APS reductase. The enzyme was sprayed in ammonium acetate buffer, a mild condition that enables the analysis of the folded protein with retention of noncovalently associated cofactors such as an iron-sulfur cluster (referred to as “native mass analysis” throughout the text) [24,25]. The mass spectrum of APS reductase showed a charge distribution of 9+, 10+, and 11+ with a mass of 28,706.0 ± 0.06 Da, corresponding to the monomeric protein (Figure S1B).
Figure 5 M. tuberculosis APS Reductase Is a Monomer
Equilibrium sedimentation experiments of M. tuberculosis APS reductase were performed using a Beckman Optima XL-I centrifuge at 4 °C. The concentration of APS reductase shown in this figure was 10 μM. However, as described in Materials and Methods, multiple concentrations of enzyme were analyzed, and in all cases we observed that the enzyme was a monomer. Fitting the data to a model of a single ideal species yielded a molecular mass of approximately 28.5 kDa.
To ensure that the histidine (His) tag was not disrupting oligomerization, untagged M. tuberculosis APS reductase was expressed and purified. This enzyme migrated with the same retention volume on the size exclusion column as His-tagged APS reductase (Figure S2A). To test for oligomerization in the presence of substrate, AMP, or thioredoxin, M. tuberculosis APS reductase was analyzed by successive runs on a gel filtration column whose buffer contained a saturating amount of the ligand or protein tested. There was no evidence of dimerization under any condition (unpublished data). APS reductase activity for the untagged and His-tagged protein was also assayed across the protein elution profile of the size exclusion column. In both cases, only a single peak of activity was observed; the elution volume of this activity corresponded to the monomeric molecular weight (Figure S2). The specific activity of His-tagged and untagged M. tuberculosis APS reductase using E. coli thioredoxin were 5.1 and 4.6 μmol min−1 mg protein−1, respectively. These values are within range of those previously reported for APS reductases from other organisms such as P. aeruginosa (5.8 μmol min−1 mg protein−1), R. meliloti (7 nmol min−1 mg protein−1), B. subtilis (0.28 μmol min−1 mg protein−1), and A. thaliana (30 μmol min−1 mg protein−1) [11,16,19,23]. Taken together, these experiments demonstrate that the active form of M. tuberculosis APS reductase is monomeric. Furthermore, this finding was not confined to a single species as other bacterial APS reductases such as those from Mycobacterium smegmatis and R. meliloti were also isolated as monomers (unpublished data). These data, in addition to the continuing controversy regarding differing mechanisms proposed for this family of enzymes, prompted us to investigate the mechanism of M. tuberculosis and other sulfonucleotide reductases in more detail.
M. tuberculosis APS Reductase Does Not Contain an Intramolecular Disulfide Bond
One possible explanation for these data was that monomeric sulfonucleotide reductases might possess an intramolecular disulfide bond instead of an intermolecular disulfide bond. Indeed, Kim and colleagues have recently made this proposal to describe the catalytic behavior of P. aeruginosa and bacterial APS reductases in general [22]. In this slight variation of the original model, thioredoxin would reduce an intramolecular disulfide bond in the first step with all subsequent steps remaining the same (Figure 4B). To investigate whether an intramolecular disulfide bond existed in M. tuberculosis APS reductase, cysteine-labeling experiments were carried out using 4-vinylpyridine (VP), a small and highly specific reagent used to covalently modify free thiols. After cysteine-labeling and buffer exchange to remove unreacted reagent, the number of covalently modified cysteines was determined by denatured mass analysis (treatment of enzyme with 80% acetonitrile and 1% formic acid). First, we confirmed that all six cysteines could be labeled in the unfolded enzyme. When the enzyme was denatured with guanidinium hydrochloride in the presence of the reducing agent tris-(2-carboxyethyl)phosphine (TCEP), reaction with a 2-fold excess of VP over the total concentration of cysteine residues resulted in five labeled cysteines. Reaction with a 10-fold excess of reagent successfully labeled all six cysteines (unpublished data). Next, we probed for cysteine modifications in the folded protein. Incubation of the native enzyme with a 10-fold excess of VP under reducing conditions yielded a molecular weight shift that corresponded to the addition of only two vinyl pyridine labels (Table 1). No additional labels were observed upon increasing the amount of VP to a 100-fold excess over total cysteine concentration in the presence of either TCEP or DTT. Moreover, no change in labeling was observed when reductant was omitted from labeling reactions (unpublished data). These results suggested that only two reactive thiols were accessible within the folded M. tuberculosis APS reductase protein.
Table 1 Mass Measurements of Wild-Type and Mutant M. tuberculosis APS Reductase Labeled with VP
From sequence gazing, we speculated that the two most likely candidates responsible for the modifications were cysteine 59 (Cys59), a nonconserved cysteine present in the N-terminal portion of the enzyme, and cysteine 249 (Cys249), the most C-terminal cysteine and putative catalytic nucleophile. To test whether these were indeed the cysteines accessible to VP labeling, the same labeling experiments were carried out with two cysteine-to-serine mutant enzymes. Since VP does not label serine residues, mutation of Cys59 or Cys249 to serine would result in the loss of a single label if they represented the labeled species in the wild-type enzyme. As predicted, each mutant resulted in the loss of precisely one VP label (Table 1). Furthermore, no labels were obtained with the double mutant Cys59Ser-Cys249Ser (Table 1). These experiments confirmed the identity of the VP-modified residues in the wild-type enzyme as Cys59 and Cys249.
The thiol labeling experiments demonstrated that M. tuberculosis APS reductase contains two accessible cysteine side chains that could, in theory, form an intramolecular disulfide bond. However, when the molecular weight of native M. tuberculosis APS reductase holoenzyme was evaluated by native mass spectrometry, we found no evidence for the presence of a disulfide bond, as the measured mass, in the absence or presence of reductant, was 28,706.0 ± 0.06 Da. This mass is in excellent agreement with the theoretical value of 28,705.88 Da, the predicted molecular weight for this enzyme without an intramolecular disulfide bond. Furthermore, when we assayed the mutant enzymes for catalytic activity, we found that Cys59 was entirely dispensable for catalysis (Figure 6), which strongly suggests that no catalytically essential disulfide bond with this residue was formed. In contrast, the Cys249Ser mutation resulted in an inactive enzyme, as anticipated from previous work demonstrating the essential nature of this residue (Figure 6). Taken together, these experiments demonstrate that intramolecular disulfide bond formation does not play a role in substrate reduction by M. tuberculosis APS reductase.
Figure 6 Cysteine 59 Is Not Required for M. tuberculosis APS Reductase Activity
The activity of wild-type (closed circles), Cys59Ser (open circles) and Cys249Ser (closed squares) APS reductase was determined as described in Materials and Methods. The concentration of sulfonucleotide reductase in this experiment was 10 nM and the concentration of APS was 20 μM.
A Stable Thiosulfonate Intermediate Is Formed in the Absence of Thioredoxin
Prior work with sulfonucleotide reductases from other organisms, as well as the data described above, had shown that the C-terminal cysteine residue (Cys249) in the reductase domain of M. tuberculosis APS reductase was essential for activity [14,23]. However, we have also shown that this cysteine is not involved in the formation of a disulfide bond that would require reduction by thioredoxin prior to substrate attack. We therefore considered the possibility that this cysteine would act directly as a nucleophile in attacking APS, without prior reduction by thioredoxin, to yield an enzyme-thiosulfonate intermediate. While predicted by the original model proposed for PAPS reduction, the existence of such an intermediate has been under debate for some time in the sulfonucleotide reductase literature [15,22,23,26]. Incubation of APS with a truncated version of A. thaliana APS reductase under physiological conditions yielded a sulfite bound to the C-terminal cysteine within the reductase domain [23]. Despite these intriguing data, no definitive evidence for such an intermediate has been presented for any bacterial APS or PAPS reductase.
With M. tuberculosis APS reductase, we hypothesized that the intermediate could be accumulated simply by omitting thioredoxin from the reaction. In order to test this model, we used mass spectrometry to probe for modifications to the enzyme during the reaction. In these experiments, APS was incubated with enzyme in ammonium acetate buffer; the mixture was sprayed directly for native mass spectrometry analysis. As reported above, the mass spectrum of APS reductase in the absence of substrate is characterized by a major series of ions with a molecular weight corresponding to the holoenzyme (Figure 7A, solid circles). Coincubation of APS reductase and APS resulted in the formation of a new series of ions with a molecular weight approximately 80 Da higher than the holoenzyme (Figure 7A, asterisks). Moreover, the signal of these ions intensified in response to increasing APS concentration. High-resolution FT-ICR mass analysis of the new species confirmed a mass shift of +80.04 ± 0.07 Da relative to the holoenzyme, consistent with the molecular weight expected for a covalently bound sulfite. Finally, proteolytic digest of the intermediate followed by mass analysis of the resulting peptides confirmed the exact mass difference of +80 Da relative to the unmodified peptide (Figure 7B). Thus, the simplest interpretation of these data is that incubation of substrate with enzyme produces a covalently bound sulfite modified intermediate.
Figure 7 M. tuberculosis APS Forms a Thiosulfonate-Enzyme Intermediate in the Absence Thioredoxin
(A) ESI mass spectra in low resolution showing the titration of APS reductase with APS in 50 mM NH4OAc (pH 7.5). In tracings labeled a to e, the concentration of APS was 0, 1, 2, 4, and 10 μM, respectively. The concentration of APS reductase is 15 μM in all cases. Peaks indicate ions corresponding to the 10+ charge states of the enzyme (E, solid circles), the covalent intermediate (E-SO3
−, asterisks), and the noncovalent complex between the intermediate and AMP (E-SO3
− • AMP, triangles). The dashed lines highlight the ions of APS reductase that lack a mature iron-sulfur cofactor. The concentrations of APS reductase reported in this figure were measured prior to mass analysis. A small fraction of protein loss occurs during the mass analysis, and thus the reported concentrations of protein should be treated as an upper limit.
(B) ESI mass spectra of trypsin-digested APS reductase incubated with a 10-M excess of APS. The C-terminal peptide (T27, residues 247–262) containing Cys249 showed a +80 Da shift (right tracing) corresponding to the molecular weight of a sulfite compared to the unmodified peptide (left tracing). The measured m/z values of the peptide with and without bound sulfite are indicated in the spectra. Shown in parenthesis are the theoretical values.
To support the mass spectrometry findings, biochemical assays were also performed. Incubation of [35S]-labeled APS together with enzyme, and subsequent analysis of the reaction by nonreducing SDS-PAGE, produced a single radioactive band—indicating the transfer of the 35S label from the substrate to the enzyme—at the expected molecular weight for APS reductase (Figure S3, lane 1). This radiolabel could be eliminated in the unfolded enzyme by heat in combination with β-mercaptoethanol (unpublished data). Likewise, preincubation of enzyme with saturating amounts of nonradioactive substrate prior to addition of [35S]APS, blocked radiolabeling (Figure S3, lane 2). Taken together, these data demonstrate the formation of a stable thiosulfonate enzyme-intermediate in M. tuberculosis APS reductase.
In the mass spectrometry analysis of APS reductase incubated with APS, we also detected a third series of ions with a calculated mass corresponding to the holoenzyme plus 427 Da (Figure 7A, triangles). The additional 427 Da could represent the intact holoenzyme bound to unreacted substrate, APS. Alternatively, this species could represent the enzyme-thiosulfonate intermediate bound to the second product, AMP. To distinguish between these two possibilities, we carried out solution kinetic experiments using conditions identical to those used for mass spectrometry to measure the rate at which substrate was converted to intermediate. By the first measurable time point (15 s), all substrate had been consumed, suggesting that the +427 Da ion was not due to binding of unreacted substrate (unpublished data). Therefore, the simplest interpretation of these data is that the third ion represents APS reductase bound covalently to sulfite and noncovalently to the product, AMP. Finally, in this experiment we also observed two minor series of ions whose molecular weights were less than the holoenzyme (Figure 7A, dashed lines). The mass of these ions was 28,352.4 Da and 26,529.3 Da. It has been well established that iron-sulfur clusters are vulnerable to oxidation and dissociation from their protein scaffolds [18,19,25,27]. For M. tuberculosis APS reductase, the theoretical mass for the apoenzyme (enzyme without iron-sulfur cluster) is 28,352.2 Da and for the 2Fe-2S intermediate it is 28,529.1 Da. These values are within 0.2 Da of the measured masses of the ions noted above. Therefore, a likely possibility is that these ions represent different stages in cluster decomposition of APS reductase. In contrast to the holoenzyme, these lower molecular weight ions did not exhibit any mass shift upon the addition of APS.
The Thiosulfonate Intermediate Is Bound to the C-Terminal Cysteine
To test whether the C-terminal cysteine, Cys249, was essential for the formation of the enzyme-sulfite intermediate, we carried out experiments analogous to those above with the Cys249Ser mutant. If Cys249 is the nucleophile for attack on APS, the Cys249Ser mutant should not form the enzyme-sulfite intermediate. In these experiments, APS was incubated with the Cys249Ser mutant APS reductase. When this reaction was analyzed by native mass spectrometry, the mutant protein exhibited a mass shift of +427 Da, consistent with the molecular weight of APS (Figure 8A). However, despite the ability of the mutant to bind APS, this variant did not form the thiosulfonate enzyme-intermediate. In separate experiments, after incubation with substrate, the Cys249Ser mutant enzyme was treated with 80% acetonitrile and 1% formic acid to unfold the protein and protonate the side chains for denatured mass spectrometry analysis. This treatment resulted in release of the noncovalently bound APS from the mutant enzyme, and only the intact apoprotein was observed (Figure 8A, inset). In contrast, native mass analysis of the reaction between APS and wild-type APS reductase yielded the expected thiosulfonate-enzyme intermediate. Subsequent analysis of this intermediate under denaturing conditions demonstrated that the sulfite modification persisted, as expected for a covalently bound intermediate (Figure 8B; Figure 8B, inset).
Figure 8 M. tuberculosis APS Reductase Cys249Ser Mutant Enzyme Binds APS, but Does Not Form a Covalent Intermediate
ESI mass spectra of 20 μM Cys249Ser APS reductase with 20 μM APS (A) and wild-type enzyme (B) with 5 μM APS in 50 mM NH4OAc (pH 7.5). Insets: The corresponding deconvoluted spectra acquired in 80% acetonitrile showing the presence and absence of a covalent modification to the mutant and wild-type enzymes, respectively.
As an independent confirmation of these results and a more direct demonstration that the sulfite was attached specifically to Cys249, cysteine-labeling experiments were carried out with wild-type APS reductase and the two mutants, Cys59Ser and Cys249Ser, using VP as a probe of cysteine reactivity after incubation with APS. If the sulfite intermediate were bound to Cys249, this residue would be unreactive toward VP labeling. Therefore, in the presence of substrate, both wild-type and Cys59Ser enzymes should lose a single VP label with a concomitant gain of a sulfite, while Cys249Ser should only lose a single VP label. When the products of these reactions were analyzed by denatured mass spectrometry, each of these expectations was confirmed (Table 1). Finally, when Cys249Ser was incubated with [35S]-labeled APS and analyzed by nonreducing SDS-PAGE, no radiolabeled protein resulted (Figure S3, lane 4), in contrast to Cys59Ser (Figure S3, lane 7) and wild-type enzyme (Figure S3, lane 1). These data demonstrate that the C-terminal cysteine residue (Cys249) is essential for intermediate formation, and confirm that the sulfite was bound to this catalytic cysteine.
Thioredoxin Is Not Required for Intermediate Formation but for Product Formation
The previous experiments suggested that incubation of substrate together with enzyme is sufficient to form the thiosulfonate enzyme-intermediate. The simplest interpretation of these data is that formation of the covalent intermediate occurs prior to a step that requires thioredoxin. Two predictions from this model are that (1) one equivalent of covalent intermediate per equivalent of active enzyme is formed in the absence of thioredoxin, and (2) the addition of thioredoxin will lead to product formation and turnover of the enzyme.
To test the first prediction, we quantified formation of thiosulfonate enzyme-intermediate as a function of APS reductase concentration. The data from this experiment are shown in Figure 9; for ease of analysis we have plotted the ratio of enzyme to substrate concentration (E:S) as the x-coordinate. The fraction of intermediate formed increased linearly with enzyme concentration. At an equimolar concentration of enzyme and substrate, 80% of the substrate would be converted into the covalent intermediate. Depletion of substrate at an exact one-to-one ratio of enzyme to substrate would require that our enzyme preparation contain 100% active molecules. Thus, taking into account that oxidation of the iron-sulfur cluster results in inactive enzyme (see Figure 7) [18,19], these experiments show that the fraction of covalent intermediate formed relative to enzyme concentration is robust and relevant to the mechanism.
Figure 9 Covalent Enzyme-Intermediate Formation Is Stoichiometric with APS Reductase
The fraction of APS converted to thiosulfonate enzyme-intermediate was monitored as a function of APS reductase concentration as described in the Materials and Methods. The concentration of APS in each reaction was 0.5 μM. Enzyme concentration was varied between 0 and 1.25 μM. The dashed line is a theoretical fit of the fraction intermediate formed and its dependence upon enzyme concentration to a linear equation (R
2 ≥ 0.98).
The second prediction of our model is that thioredoxin will reduce the enzyme-thiosulfonate intermediate to form product. This hypothesis was first tested using native mass analysis. In the absence of substrate, the spectrum of APS reductase was characterized by three series of ions; the major series of ions corresponded to the holoenzyme (Figure 10). Upon supplementation with substrate, the enzyme-thiosulfonate intermediate was formed. Addition of reduced, but not oxidized, thioredoxin resulted in the release of the intermediate. Control experiments demonstrated that release of sulfite was not due to residual DTT in the reduced thioredoxin sample (unpublished data).
Figure 10 Thioredoxin Reduces the Thiosulfonate Enzyme-Intermediate
ESI mass spectra, in low resolution, of 10 μM APS reductase (A); 2.5 μM APS and 10 μM APS reductase (B); 2.5 μM APS, 3.7 μM oxidized thioredoxin, and 10 μM APS reductase (C); and 2.5 μM APS, 3.7 μM thioredoxin reduced with 33 μM DTT, and 10 μM APS reductase (D). The ions corresponding to the 10+ charge state of the enzyme-associated species are labeled.
Solution kinetics and gel-labeling experiments were also performed to verify these mass spectrometry data. Addition of thioredoxin to a reaction mixture containing APS reductase and APS enabled multiple turnover of substrate (Figure 11). Sulfite was not released by the addition of high concentrations of DTT in place of thioredoxin as the reductant. Additionally, a panel of small-molecule reductants with varying reduction potentials was screened for their ability to release sulfite from the intermediate. In each case, APS reduction was below our detection limit over the time scale of the experiment (≤ 10−5 min−1, unpublished data). Finally, gel-labeling experiments demonstrated that incubation of [35S]-SO3
2− radiolabeled reductase together with thioredoxin resulted in release of the radiolabel, as expected (Figure S3, lanes 3 and 9). These experiments demonstrated that efficient turnover of APS specifically requires thioredoxin.
Figure 11 Sulfite Release Is Thioredoxin-Dependent
Comparison of substrate turnover for M. tuberculosis APS reductase supplemented with 10 μM thioredoxin (closed circles), with 10 mM DTT (open circles), or in the absence of all reductants (closed squares) as described in the Materials and Methods. The concentration of APS reductase in this experiment was 5 nM, and of APS was 20 μM.
Sulfonucleotide Reduction Revisited
Given the results for M. tuberculosis APS reductase and their consistency with the mode of action proposed for A. thaliana APS reductase, we wondered whether all sulfonucleotide reductases utilized the same overall catalytic strategy. To further test this hypothesis, we chose to investigate E. coli PAPS reductase and P. aeruginosa APS reductase, which have been proposed to act via the mechanisms presented in Figure 4A and 4B, respectively.
With these two enzymes, we performed mass spectrometry analysis and biochemical characterization similar to that described for M. tuberculosis APS reductase. First, we probed for thiosulfonate formation. Like M. tuberculosis APS reductase, both E. coli PAPS reductase and P. aeruginosa APS reductase formed the enzyme-thiosulfonate intermediate without prior reduction by thioredoxin (Table 2). As expected, mutation of the catalytic cysteine in each enzyme to serine prevented intermediate formation (Table 2). Also consistent with M. tuberculosis APS reductase, the covalently bound sulfite was released upon addition of reduced thioredoxin and not by DTT (Table 2). These conclusions were further corroborated as described above for M. tuberculosis APS reductase using data obtained from gel-labeling and biochemical assays (Figures S4 and S5). Finally, two additional bacterial APS reductases from R. meliloti and M. smegmatis were analyzed in similar biochemical experiments (unpublished data). These studies reiterated the findings observed with all other sulfonucleotide reductases analyzed in this study—the reaction proceeded via the covalent sulfite-bound intermediate; sulfite production was dependent upon reduction of this intermediate by thioredoxin.
Table 2 Mass Measurements of Wild-Type and Mutant E. coli and P. aeruginosa Reductases Reacted with Substrate, in the Absence or Presence of Thioredoxin
To complete our analysis, we carried out cysteine-labeling with E. coli PAPS reductase and P. aeruginosa APS reductase using VP. Since E. coli PAPS reductase contains only one cysteine residue in its amino acid sequence, incorporation of one VP label was clearly predicted. A single label was obtained and, as described for M. tuberculosis APS reductase, this labeling could be blocked by preincubation of substrate and enzyme prior to the addition of VP (Table 3). The primary amino acid sequence of P. aeruginosa APS reductase contains five cysteine residues, four from the -CC-X∼ 80-CXXC- sequence motif and one from the C-terminal cysteine nucleophile. When P. aeruginosa APS reductase was incubated with VP and the reaction products were analyzed by mass spectrometry, as in the labeling studies with E. coli and M. tuberculosis reductases, only a single VP label was observed (Table 3). To ensure that this pattern of labeling was not an artifact specific to VP modification, we repeated the experiments with iodoacetamide, another cysteine-modifying reagent. As in labeling studies with VP, a single label was obtained with iodoacetamide under reducing or oxidizing conditions (unpublished data). Site-directed mutagenesis as well as preincubation of the enzymes with substrate prior to VP addition positively identified the labeled residue as the C-terminal cysteine (Tables 2 and 3). These experiments provide no evidence to support the existence of a catalytic disulfide bond in either E. coli PAPS reductase or P. aeruginosa APS reductase. Moreover, based on the predictions made by the models presented in Figure 4A and 4B, reoxidation of such a disulfide bond would preclude the isolation of a stable enzyme-thiosulfonate intermediate and directly contradict the observed requirement for thioredoxin in sulfite release. These results demonstrate in E. coli PAPS and P. aeruginosa APS reductases that (1) incubation of sulfonucleotide reductase together with substrate yields a covalent thiosulfonate-enzyme intermediate; (2) intermediate formation is not dependent upon prior reduction of the sulfonucleotide reductase by thioredoxin; and (3) thioredoxin is essential for reduction of the thiosulfonate intermediate.
Table 3 Mass Measurements of Wild-Type and Mutant E. coli PAPS and P. aeruginosa APS Reductases Labeled with VP
Discussion
Over the last decade, a large body of work has contributed substantially to our understanding of the unique family of sulfonucleotide reductase enzymes. The authors of these studies have proposed models for sulfonucleotide reduction that have predictive power. We have anchored our studies by systematically testing some of the central predictions from the leading models previously proposed for sulfonucleotide reduction. In this study, we have used new experimental approaches to investigate this reaction from a different perspective—namely, through the use of high-resolution FT-ICR mass spectrometry in conjunction with cysteine-labeling, solution-based kinetic analysis, and other biochemical approaches. Our results strongly suggest that a general catalytic strategy is shared by all sulfonucleotide reductases and provide the first functional evidence for evolution from a common ancestor.
A Universal Mechanism for Sulfonucleotide Reduction
The data presented in this paper demonstrate that each of the bacterial sulfonucleotide reductases operates via a two-step mechanism, as originally proposed by Weber and coworkers to describe APS reduction in higher plants [23]. In Figure 12, we take the original proposal by Weber and coworkers a step further and provide a detailed picture of the molecular events that might take place during sulfonucleotide reduction. During the first step, an absolutely conserved cysteine residue carries out nucleophilic attack on the sulfonucleotide sulfate group to yield an enzyme-thiosulfonate intermediate. The stable formation of this intermediate has been demonstrated for each enzyme investigated here. In all cases, addition of substrate to enzyme with subsequent analysis of the reaction mixture by mass spectrometry yielded holoenzyme with a molecular weight shift that corresponds the covalent addition of a sulfite group (see Figures 7 and 8; Table 2). Additionally, gel-labeling experiments have demonstrated the requisite transfer of radiolabel from [35S]-substrate to enzyme (Figures S3 and S4). We have also shown through site-directed mutagenesis and cysteine-labeling studies that the conserved nucleophile is not only essential for intermediate formation, but is also the site of sulfite attachment (see Figures 8, S3, and S4; Tables 2 and 3). Pre-steady state kinetic analysis shows that sulfonucleotide reductases are blocked after covalent intermediate formation in the absence of thioredoxin and cannot perform multiple rounds of substrate turnover (see Figures 9 and S5). Finally, native mass analysis of M. tuberculosis APS reductase demonstrates that there are no disulfide bonds in the holoenzyme or in the thiosulfonate enzyme-intermediate; cysteine-labeling studies further support this conclusion for each of the sulfonucleotide reductases analyzed in this work (see Tables 1 and 3). Taken together, these experiments demonstrate that thioredoxin is not required for sulfonucleotide binding and enzyme-thiosulfonate intermediate formation.
Figure 12 Mechanism of Sulfonucleotide Reduction
In the second half of the reaction, we demonstrated that thioredoxin is required for product formation. Mass spectrometry and gel-labeling experiments clearly show the resolution of thiosulfonate enzyme-intermediate with concomitant regeneration of free sulfonucleotide reductase when thioredoxin is included in the reaction mixture (see Figures 10, 11, and S5; Table 2). In addition, these studies demonstrate that reduction of the thiosulfonate bond requires thioredoxin; DTT by itself does not efficiently support the reduction of the intermediate (see Figures 10, 11, and S5; Table 2). In the final steps of the reaction, we propose that the more reactive N-terminal thioredoxin cysteine would carry out nucleophilic attack upon the sulfonucleotide reductase catalytic cysteine (Figure 12) [9]. This reaction would result in the formation of a mixed disulfide between thioredoxin and the sulfonucleotide reductase and, concomitantly, release sulfite. Subsequent thiol/disulfide exchange with the second thioredoxin thiolate would yield oxidized thioredoxin and regenerate reduced sulfonucleotide reductase. The proposed covalent protein-protein intermediate is reminiscent of those predicted during the thioredoxin-mediated reduction of protein disulfides [9]. However, the subsequent chemical step is very rapid and we find no evidence for accumulation of this species. This result is not unexpected, because of the intramolecular nature of this reaction; we are currently using different experimental approaches to test this aspect of the proposed model in greater experimental detail.
A commonly employed assay to measure sulfonucleotide reductase includes the addition of 20–40 mM nonradioactive sulfite during the reaction or at its conclusion [13,28]. In their work with A. thaliana APS reductase, Weber et al. [23] demonstrated that the sulfite bound to the C-terminal cysteine is released upon exposure to millimolar concentrations of sulfite, a natural reductant. We have reiterated these findings in our own studies (unpublished data). Thus, the inclusion of sulfite in the assay of sulfonucleotide reductase activity results in thioredoxin-independent reduction of a thiosulfonate enzyme-intermediate to produce sulfite. It is likely that this phenomenon has hampered prior discovery of the covalent intermediate in bacterial sulfonucleotide reductases, and can account for the modest activity observed in the absence of thioredoxin reported by previous studies [11,12,14].
A Novel Enzyme-Thiosulfonate Intermediate
The thiosulfonate species (E-Cys-S-SO3
−) formed during sulfonucleotide reduction is novel among the diverse array of characterized covalent enzyme intermediates. The most closely related covalent enzyme-intermediate is the persulfide (E-Cys-S-S−) found in the family of NifS-like enzymes that includes NifS from Azotobacter vinelandii and E. coli IscS [29,30]. These catalysts have cysteine desulfurase activity that is utilized to mobilize sulfur from L-cysteine [29–31]. The sulfur is bound to a conserved active site cysteine residue from which is it subsequently transferred to form a persulfide group on a variety of other targets, such as the scaffold proteins NifU and IscU, for use in the assembly of iron-sulfur clusters [32]. In addition to cluster formation, NifS-like enzymes are also required for other biological processes such as the biosynthesis of thiamin, biotin, and molybdenum cofactors [32,33]. Another catalyst that has been demonstrated to form a persulfide intermediate is the sulfur transferase rhodanese. Like NifS-like proteins, this enzyme cycles between a sulfur-free and a stable persulfide-containing form [34]. In vitro, it has been demonstrated that rhodanese can act as a sulfur insertase and regenerate iron-sulfur clusters in proteins [35]. However, the in vivo role of rhodanese has not been well established. Notably, the structure of sulfur-free rhodanese and its persulfide-containing form have been solved for a number of organisms [34,36]. A comparison of the two enzyme species indicates a significant conformational rearrangement in the vicinity of the catalytic cysteine upon persulfide formation. By analogy, a similar conformational change could occur upon thiosulfonate formation in sulfonucleotide reductases that would facilitate the recognition and subsequent reduction of this intermediate by thioredoxin.
Another related protein modification is the oxidation of cysteine to sulfenic acid (R-S-OH) [37]. This oxidation is reversible and can be reduced by thioredoxin and glutathione. Cysteines can also be hyperoxidized to sulfinic (R-SO2H) and sulfonic acids (R-SO3H) [37,38]. This process is thought to occur by reaction of susceptible cysteine residues with oxidizing metabolic byproducts, such as hydrogen peroxide. A family of enzymes known as the peroxiredoxins is reversibly inactivated by hyperoxidation of a cysteine residue to the corresponding sulfinic acid in vivo [39,40]. A protein termed sulfiredoxin was later identified and can reduce this sulfinic acid modification back to cysteine [41]. Finally, in humans sulfite is thought to be detoxified by conversion to S-sulfocysteine (Cys-S-SO3
−) [42]. From these few examples, it is apparent that the thiol is subject to unique biological chemistry and is often the site of a diverse array of modifications that include the covalent intermediate we observed in APS reductase.
The Role of the Iron-Sulfur Center in APS Reductase
All sulfonucleotide reductases analyzed in this study utilized the same overall two-step strategy to catalyze sulfonucleotide reduction. Nevertheless, there are clear differences between APS and PAPS reductases. Specifically, these two classes of enzymes differ in the presence of the iron-sulfur center; APS reductases contain a 4Fe-4S cluster that is notably absent from PAPS reductases. The logical question that emerges from this difference is: What role does the iron-sulfur cluster play in APS reductase?
In this study, we observed ions of M. tuberculosis APS reductase that have a measured mass in excellent agreement with the theoretical mass expected for the apoenzyme and the 2Fe-2S intermediate. In support of this assignment, it has recently been demonstrated that exposure to oxygen generates the apo and 2Fe-2S cluster forms of B. subtilis sulfonucleotide reductase [19]. In contrast to M. tuberculosis APS reductase holoenzyme, these ions do not undergo any mass shift upon addition of substrate. Thus, forms of APS reductase that appear to lack a mature cofactor did not form a stable association with substrate (in contrast to that observed with the Cys249Ser mutant [Figure 8]) or catalyze the formation of the enzyme-thiosulfonate intermediate. These observations are also consistent with previous studies that have correlated the loss of iron with loss of APS reductase activity [18,19]. One explanation for these data is that the loss of cofactor could result in gross active site structural perturbation with concomitant disruption of substrate binding interactions. However, an alternative possibility is that a direct interaction between the iron-sulfur cluster and APS is required for binding. An intriguing corollary of this model could be that PAPS reductases compensate for the lack of this cofactor through additional binding energy gained from the 3′-phosphate moiety of PAPS. In support of this second proposal, all APS reductases reported on thus far exhibit a decrease in activity over time that correlates with decomposition of the iron-sulfur cluster [12,17–19,22]. In one such study, the addition of AMP to L. minor APS reductase storage buffer prevented this loss of enzyme activity [17]. A plausible explanation for the additional stability is that the terminal phosphate of AMP interacts with the cluster, preventing oxidation and stabilizing enzyme activity. By analogy, the sulfate moiety of APS could also interact with the cluster.
Inspection of the motif that serves as a marker for the iron-sulfur cluster in APS reductases reveals the presence of an unusual cysteine pair—Cys140 and Cys141 (using M. tuberculosis APS reductase numbering). This Cys-Cys tandem motif is atypical within the family of iron-sulfur cluster-containing proteins and has therefore spurred debate as to the number of cysteines that ligate the 4Fe-4S cluster. Mössbauer spectroscopy has been used to probe the cluster in L. minor and P. aeruginosa APS reductases [12,18]. In these experiments, only three of the four iron sites exhibited similar Mössbauer parameters. Therefore, the authors concluded that the APS reductase 4Fe-4S cluster is ligated by three cysteines. While there is no doubt that the iron-sulfur cluster in APS reductase exhibits unusual spectroscopic properties, alternative interpretations of these data with respect to cluster coordination are possible. The cysteine-labeling studies carried out in this study serve as an independent test of the three-cysteine theory. If only three cysteines participate in iron-sulfur coordination, the simplest expectation would be that two covalent modifications could occur—on Cys141 and Cys249. However, we found that M. tuberculosis Cys59Ser and wild-type P. aeruginosa APS reductases incorporated only one cysteine modification, specifically at Cys249. These cysteine-labeling data would be consistent with a role for Cys141 in iron-sulfur cluster ligation.
Kim et al. [22] have recently reported on the reaction of P. aeruginosa APS reductase with dithio-1,4-nitrobenzoic acid (DTNB). In this experiment, reduction of DTNB gave rise to a spectroscopic signal that was proposed to represent the presence of two free thiols in P. aeruginosa APS reductase. As noted above, the data in the present study indicate that P. aeruginosa APS reductase has only one free thiol—Cys249. The experiments reported in this work have probed for free thiols using the alkylating reagents VP and iodoacetamide. After the labeling reaction, covalent modifications were quantified by mass spectroscopy. In addition, the identity of labeled cysteines was unambiguously established using site-directed mutagenesis. Unfortunately, quantitation of free thiols by measuring the release of 5-thio-2-nitrobenzoate (TNB−) from DTNB by monitoring the absorbance at 412 nm allows for determination only of total thiol content in a sample. In the absence of additional experiments to confirm that a protein cysteine residue has been labeled, and that this label can be eliminated via site-directed mutagenesis, it is not possible to establish the basis of the TNB− signal. For example, iron-sulfur cluster oxidation occurs normally under aerobic conditions and produces free sulfide. Since every mole of APS reductase protein is normally associated with 4 moles of sulfide, decomposition of the cluster can result in significant concentrations of sulfide in the sample [25]. Free sulfide is very reactive, and if present during a DTNB assay, will generate TNB− signal that is not attributable to free protein cysteine thiol content [43]. Furthermore, the study by Kim et al. [22] purified P. aeruginosa APS reductase only via affinity chromatography. In the current work, enzymes were purified in an additional step using size-exclusion chromatography. It is also possible, then, that differences in protein preparation could contribute to the observed discrepancy. Interestingly, our analysis indicated that P. aeruginosa APS reductase was a homotetramer (Figure S1A).
Additional data obtained from site-directed mutagenesis of Cys140 and Cys141 also support a role for these cysteines in cluster coordination [18,22]. Mutation of Cys140 to serine in P. aeruginosa APS reductase results in an enzyme with little detectable iron incorporation. The same mutation to Cys141 results in a less dramatic phenotype, with a 20% decrease in iron and a 35% decrease in sulfide content assayed soon after purification. However, the long-term stability of the iron-sulfur cluster is substantially decreased in the context of the Cys141 mutant as compared to the wild-type enzyme. One explanation for the enhanced short-term stability of the iron-sulfur cluster in P. aeruginosa Cys141Ser mutant could involve the homotetrameric nature of this enzyme. Such a higher order assembly could provide stabilizing protein-protein contacts for the iron-sulfur cluster. This hypothesis is supported by the finding that cluster loss in the P. aeruginosa Cys141Ser mutant was more rapid at lower protein concentrations [22].
Figure 13 illustrates a model that can account both for the unusual properties of the cluster and the role we propose for it in substrate binding. While the Cys-Cys motif is uncommon, a recent computational study has suggested that it is theoretically possible for each of these cysteines to ligate different iron atoms within the same cluster [44]. In addition, the importance of each of the four cysteines is strongly suggested by the cysteine-labeling and site-directed mutagenesis experiments detailed above. Therefore, we propose that in the absence of substrate, each cysteine would participate in cluster coordination. However, the interaction with Cys141 would be unique because of the geometrical constraints imposed by the neighboring cysteine and its own interaction with the cluster. When APS binds the reductase, the oxygens on the sulfate moiety could displace Cys141. In our cysteine-labeling experiments, the addition of substrate blocks labeling by VP; no additional labels are obtained. Thus, for the displacement model to hold true, reaction of VP with the liberated Cys141 would have to be prohibited, for example by the acquisition of a new binding interaction. Alternatively, Cys141 could remain bound, and the additional binding of APS would augment the coordination sphere of the metal. In either scenario, once the substrate is bound, we propose that the unique iron in the cluster acts as a Lewis acid by activating the substrate for subsequent nucleophilic attack and by stabilizing the developing negative charge. A similar role for iron-sulfur cluster participation in substrate binding and activation has been demonstrated for aconitase [45]. This enzyme contains an iron-sulfur cluster with a unique iron site bound by three cysteine ligands and a water molecule. During the conversion of citrate to isocitrate, the terminal oxygen atoms of the substrate interact with this unique iron. However, in our model for APS reductase we are proposing a “substrate-activated” aconitase-like arrangement in which the fourth iron interacts with Cys141; APS binding to the cluster would result in the displacement or shift of this cysteine. An important requirement for iron-sulfur cluster involvement in substrate binding and/or catalysis is that the cluster resides near the active site. Unfortunately, no structural data on assimilatory APS reductases are currently available. However, two studies have reported structural modeling via sequence homology using the crystal structure of E. coli PAPS reductase for P. aeruginosa and B. subtilis APS reductase [19,22]. These models position the cysteines implicated in cluster ligation within the vicinity of the enzyme active site. Thus, the location of the cluster would be consistent with our proposed catalytic function. Nevertheless, additional biochemical investigation together with high-resolution structural information on APS reductase will be required to probe the model presented in Figure 13 in further molecular detail.
Figure 13 Proposed Role of the Iron-Sulfur Cluster in APS Reduction
Summary
The evolutionary relatedness within the family of sulfonucleotide reductases has been strongly suggested by sequence homology. Nevertheless, issues of differing substrate specificity, the presence or absence of an iron-sulfur cofactor and the continued conflict in the literature with respect to mechanism had obscured this vision. Hence, the general mechanism of sulfonucleotide reduction established in this work revalidates the evolutionarily relatedness of these enzymes and will help inform future studies to address the experimental challenges outlined above, such as structural characterization, identification of the determinants that govern substrate specificity, and the catalytic involvement of the iron-sulfur cluster. Finally, given that discrete snapshots in the evolution of sulfonucleotide reduction have been biologically retained—APS reductases containing an iron-sulfur cofactor (M. tuberculosis), the bispecific APS and PAPS reductase containing an iron-sulfur cofactor (B. subtilis), and PAPS reductases without an iron-sulfur cluster cofactor (E. coli)—we are presented with an exciting opportunity to probe more general questions regarding the evolution of catalytic diversity in the work to come.
Materials and Methods
Materials
Nonradioactive APS was purchased from Biolog Life Sciences Institute, (Bremen, Germany). [35S]SO4
2− (specific activity1,491 Ci/mmol) was obtained from MP Biochemicals (Irvine, California, United States). Molecular biology grade DTT was from Invitrogen (Carlsbad, California, United States). Nonradioactive PAPS, TCEP, and E. coli thioredoxin protein were all purchased from EMD Biosciences (San Diego, California, United States). VP, bis-tris propane, methionine, and iodoacetamide were all purchased from Sigma-Aldrich (St. Louis, Missouri, United States). Depending upon availability, PEI-cellulose thin-layer chromatography (TLC) plates (20 cm × 20 cm) were purchased from J. T. Baker (Phillipsburg, New Jersey, United States) or EMD Biosciences. Pfu and DpnI polymerase were from Stratgene (La Jolla, California, United States). Restriction enzymes and T4 DNA ligase were from New England Biolabs (Beverly, Massachusetts, United States). All other chemicals were purchased from J. T. Baker and were of the highest purity available (≥ 95%). DNA oligonucleotides were purchased from Qiagen (Valencia, California, United States).
Preparation of the sulfonucleotide reductase expression vectors
Table S1 lists the oligonucleotides used in this study. The gene encoding the M. tuberculosis APS reductase was amplified from H37Rv M. tuberculosis genomic DNA and cloned in a protein expression vector as described [3]. The gene encoding M. smegmatis APS reductase was amplified from M. smegmatis genomic DNA. The gene encoding P. aeruginosa APS reductase was amplified from P. aeruginosa genomic DNA
ATCC 47085D (
ATCC, Manassas, Virginia, United States). The gene encoding E. coli PAPS reductase was amplified from E. coli genomic DNA
ATCC 700926 (
ATCC). PCR reactions contained 0.25 μM of each primer, 10–100 ng of genomic DNA template, and 2.5 units of Pfu DNA polymerase in reaction buffer supplied by the manufacturer. PCR was performed in a PTC-200 thermocycler (Bio-Rad Laboratories, South San Francisco, California, United States) using the following program: 2 min at 95 °C; 30 cycles of 30 s at 95 °C, 1 min at 55 °C, and 1 min at 68 °C; and a 5-min extension at 68 °C. In some cases, the annealing temperature had to be adjusted in order to achieve the desired product. PCR reactions were screened for the expected DNA fragment via agarose gel electrophoresis. Subsequently, each PCR fragment was ligated into a Zero Blunt Topo cloning vector (Invitrogen) and subsequently digested by NdeI and XhoI (pET24b) or NdeI and BamHI (pET14b). Alternatively, the PCR fragment was directly digested without additional cloning steps. The purified gene fragment was then ligated into a NdeI- and XhoI-digested, CIP-treated pET24b or a NdeI- and BamHI-digested CIP-treated pET14b vector using T4 DNA ligase. The ligation reaction was transformed into chemically competent XL-1 Blue cells (Stratagene). After growth on LB plates containing 50 mg/ml kanamycin (pET24b) or 100 mg/ml ampicillin (pET14b), colonies were selected and cultures were grown up overnight. Plasmid DNA minipreps were screened for the presence of the gene by PCR. DNA sequencing using forward T7 and reverse T7 primers was performed to confirm the identity of the desired gene product.
Site-directed mutagenesis
Site-specific mutations were made using Quik Change PCR mutagenesis kit (Stratagene) using the appropriate plasmid template according to the manufacturer's specifications. The DNA oligonucleotides used in these reactions are listed in Table S1. Successful incorporation of the desired mutation was confirmed by sequencing and once verified, the plasmid was transformed into BL21(DE3) (Novagen, San Diego, California, United States) for protein expression as described above.
Expression and purification of sulfonucleotide reductases
Proteins were expressed by transforming a reductase-containing plasmid into BL21(DE3) cells (Novagen) and grown on LB-agarose containing 50 mg/ml kanamycin. An isolated colony was grown in 5 ml of LB broth containing 50 mg/ml kanamycin. The culture was grown at 37 °C overnight. This culture was used to inoculate 1 l of LB broth containing 50 mg/ml kanamycin. The culture was grown with shaking (250 rpm) at 37 °C to an OD of 0.6, and IPTG was added to a final concentration of 0.4 mM. The shaker flasks were then shifted to18 °C and grown 12–16 h. Subsequently, 1 l of cells were collected by centrifugation and resuspended in 30 mls lysis buffer (20 mM sodium phosphate [pH 7.4], 0.5 M NaCl, 10 mM imidazole, and 1 mM methionine) together with an EDTA-free protease inhibitor tablet (Roche, Indianapolis, Indiana, United States) before disruption by sonication on ice. After sonication, DNase and RNase (Sigma) were added to the lysate at 10and 5 μg/ml, respectively, and stirred for 10 min on ice. The cell lysate was cleared by centrifugation and the supernatant was applied to a 1 or 5 ml HiTrap Chelating column (Amersham, Piscataway, New Jersey, United States). The column was washed with ten column volumes in 20 mM phosphate (pH 7.4), 0.5 M NaCl, and 50 mM imidazole, and was eluted with 20 mM phosphate (pH 7.4), 0.5 M NaCl, and 250 mM imidazole. Fractions containing the desired protein were pooled and concentrated using Amicon 10,000 Da molecular weight cutoff centrifugal filters (Millipore, Billerica, Massachusetts, United States) prior to injection onto a 10/30 Superdex 200 or a 16/60 Superdex 200 prep grade gel filtration column. The standard gel filtration buffer was 50 mM Tris-HCl (pH 8.0), 10% glycerol, and 5 mM DTT with ionic strength adjusted to 150 mM with NaCl. The inclusion of DTT in the gel filtration buffer for sulfonucleotide reductases purification has been well established [14–16]. DTT was included in the gel filtration buffer to slow down oxidation and decomposition of the iron-sulfur cluster. No change in retention volume during size exclusion chromatography was observed with any of the proteins purified for this study when DTT was omitted from the buffer. Fractions containing sulfonucleotide reductase were pooled, aliquoted into single-use portions, snap-frozen in liquid nitrogen, and stored at −80 °C. Protein concentrations were determined precisely by quantitative amino acid analysis (AAA Service Laboratory, Boring, Oregon, United States). Iron content, typically ≥ 3.5 mol iron/mol protein, of each preparation was determined as previously described [46,47].
Untagged M. tuberculosis APS reductase was expressed as reported above. Cells from 1 l of bacterial culture were resuspended in 30 ml of 20 mM bis-tris propane (pH 7.0) and lysed by sonication. The resulting lysate was clarified by centrifugation, and the supernatant was applied to a 5-ml Fast Flow Q anion exchange column (Amersham). The column was washed with ten column volumes of lysis buffer, and proteins were eluted over ten column volumes with a salt gradient of 0–1 M NaCl. Column fractions were assayed for APS reductase activity and for absorbance at 390 nm to detect the iron-sulfur cluster. APS reductase eluted between 250 and 350 mM NaCl. APS reductase-containing fractions were pooled and ammonium sulfate added to a final concentration of 30%. APS reductase activity was retained in the supernatant. Proteins that precipitated at 30% ammonium sulfate were removed by centrifugation. Ammonium sulfate was added to a final concentration of 70% to precipitate APS reductase. The protein pellet was resuspended in gel filtration buffer and was further purified by size exclusion chromatography, as described above.
Synthesis of [35S]APS and [35S]PAPS
35S-labeled APS and PAPS were prepared by incubating [35S]Na2SO4, ATP, ATP sulfurylase(Sigma), inorganic pyrophosphatase (Sigma), and recombinant APS kinase together as previously described [48,49]. Analysis of the reaction by PEI-cellulose TLC plates developed in 1 M LiCl indicated complete conversion of [35S]SO4
2− to [35S]PAPS. The reaction was then terminated by heating for 2 min in a boiling water bath. The precipitate formed was removed by centrifugation. [35S]PAPS was further purified by treatment with activated charcoal as previously described [50]. For the synthesis of [35S]APS, nuclease P1 was incubated with the [35S]PAPS reaction supernatant for 30 min at room temperature. Complete conversion of [35S]PAPS to [35S]APS was verified by TLC and [35S]APS further purified as above. Single-use aliquots of [35S]APS and [35S]PAPS, always containing less than 1.5% contaminating [35S]SO4
2− , were stored at −80 °C. Minimizing the amount of [35S]SO4
2− present in the substrate served only to facilitate product quantification, as sulfate had no observed effect on reaction rates at concentrations measured into the millimolar regime. Furthermore, while APS is susceptible to nonenzymatic hydrolysis of APS to sulfate and AMP, this reaction was undetectable over the time scale of our assays as measured in reactions that contained only APS or APS incubated with the catalytically inactivated M. tuberculosis APS reductase Cys249Ser mutant.
General kinetic methods
Reactions were performed at 30 °C in assay buffer (50 mM sodium phosphate [pH 7.0], adjusted to 100 mM ionic strength with NaCl) unless noted otherwise. Nonradioactive PAPS or APS were doped with a trace amount of [35S]APS or [35S]PAPS. Different buffers were used at the same pH value to test for buffer-specific effects on reductase activity. No significant buffer effects on enzymatic activity were observed. Reactions were performed at pH 7.0, the pH value of the maximal rate based on the observed pH dependence of M. tuberculosis APS reductase (unpublished data). Multiple turnover reactions contained 10 μM thioredoxin and 5 mM DTT and were typically initiated by the addition of APS or PAPS. As shown by the data presented in Figures 10 and 11 and Table 2, DTT was unable to reduce the thiosulfonate-enzyme intermediate. DTT was included in multiple turnover reactions to regenerate the thioredoxin that becomes oxidized as a consequence of APS reduction (Figure 2). While multiple turnover reactions are typically followed using initial rates, with the exception of very slow or undetectable reactions, we followed our reactions to completion (five half-lives or more). For the purposes of these assays, inhibition by AMP product did not have a significant effect upon the reaction rates measured. The reaction progress curve was plotted as a function of time, and the fractional extent of reaction and fit by a single exponential (Kaleidagraph, Synergy Software, Reading, Pennsylvania, United States). Good first-order fits to the data with end points of 90% were obtained (R
2 ≥ 0.98). The kinetic data presented in each figure were measured in at least three independent experiments to ensure reproducibility of results.
TLC-based assay for sulfonucleotide reduction
Traditionally, sulfonucleotide reductase activity has been quantified by quenching the [35S]SO3
2− producing reaction with 2 M sulfuric acid, together with a large excess of cold sulfite, to yield [35S]-labeled sulfoxide gas [13,28]. The reaction vessel is rapidly placed into a vial containing an organic base, such as octyl amine, to trap the gas. The vial is closed and incubated overnight to maximize gas absorption. Subsequently, the amount of trapped radiolabeled sulfide is quantified by scintillation counting.
While the original assay continues to be in use, a TLC-based method has more recently been developed that avoids the production of radioactive gas and allows analysis of results on a faster time scale [16]. Perhaps most importantly, the fraction of reaction can be directly determined by quantifying both substrate and product spots as visualized by TLC, which helps to minimize quantification of reaction artifacts and improves the accuracy of measured rates [51]. In this method, a 2-min heat-kill step is used to terminate the sulfonucleotide reductase reaction and is followed by analysis on PEI-cellulose TLC plates developed in 1 M LiCl. To facilitate kinetic analysis of sulfonucleotide reductase activity, we have modified this assay slightly based on methods developed by Peluso et al. to measure GTPase activity [52]. Our modification has been designed to take advantage of the fact that sulfonucleotide reductase activity is pH-dependent; no detectable activity was observed below pH 5 for any of the sulfonucleotide reductases used in this study. At specified times, an aliquot was removed from the reaction mixture and quenched with 0.75 M potassium phosphate (pH 3.3). The products and unreacted substrate were then separated by TLC on PEI-cellulose plates (prepared by prewashing for 5 min in 10% NaCl, followed by three 5-min washes in water and then drying) and developed in 1 M LiCl and 0.3 M sodium phosphate (pH 3.8). Control reactions showed that the quench rapidly extinguished all reductase activity. Under aerobic conditions, free sulfite product naturally oxidizes to sulfate. The rate at which this oxidation occurs is enhanced both by low pH and by the presence of metal ions. Thus, termination of the reaction with the low pH quench and subsequent development in the running buffer described above converts the free [35S]SO3
2− product generated in the enzymatic reaction quantitatively into [35S]SO4
2− . The product was then quantified as [35S]SO4
2− . Addition of 1 mM MgSO4 to the quench buffer improved the resolution of substrate and product on the PEI-cellulose plates. TLC plates were analyzed using Phosphorimager analysis (Amersham) with Image Quant quantitation software. In side-by-side experiments, reaction rates measured using the low-pH quench were within experimental error (± 15%) of the previously published heat-kill assay. Representative data for this modified TLC assay are presented in Figure S6.
Measurement of intermediate formation by TLC
For reactions in which multiple turnover was not desired, thioredoxin was omitted from the reaction mixture. Sulfonucleotide reductase enzyme was used at concentrations in the high nanomolar to low micromolar range to detect a single turnover. Under conditions that had quantitatively produced the thiosulfonate intermediate, E-SO3
−, as observed by mass spectrometry and gel labeling, a new signal on the TLC plate was observed as a discrete spot at that remained the origin of the TLC plate (see Figure S5). In these experiments, intermediate formation was followed in two ways. Since radiolabel would be transferred from substrate to enzyme upon intermediate formation, in the first method we quantified the decrease in APS over time. In a second method, we quantified the appearance of the new spot at the TLC origin, which was assumed to be E-SO3
−. For each enzyme concentration, the reaction endpoint was determined at the point where no further substrate was consumed or intermediate was formed. The endpoint of each reaction (quantified either as fraction APS consumed or as fraction intermediate formed) was plotted as the ratio of enzyme to substrate concentration. As expected, the amount of substrate consumed was identical to the amount of new spot generated, and either treatment of the data yielded the same results.
Analytical ultracentrifugation
Sedimentation equilibrium experiments were conducted using a Beckman XL-I analytical ultracentrifuge (Beckman Instruments, Fullerton, California, United States). Purified M. tuberculosis APS reductase protein in gel filtration buffer was analyzed at concentrations ranging from 5 to 30 μM at 4 °C. All concentrations analyzed gave very similar molecular weights. In independent experiments, samples were centrifuged at speeds ranging from 20,000 to 30,000 rpm, and the absorbance versus radial distribution was determined between 380 and 400 nm, the wavelength region in which the iron-sulfur cluster cofactor bound to the enzyme absorbs. Absorbance data were obtained at the wavelength indicated and at radial increments of 0.003 cm, each data point being an average of three measurements. Data were acquired at intervals over a period of 18 h, with scans at 3-h intervals to ensure that the system had attained equilibrium. Equilibrium distributions were then acquired at 0.001-cm radial increments, each data point being the average of five measurements. The sedimentation equilibrium data were analyzed using a vbar of 0.729 calculated from the amino acid composition, a calculated solvent density of 1.026, and best fit to a model of a single ideal species using a data analysis program supplied by Beckman.
4-vinylpyridine and iodoacetamide labeling
N-methylmorpholine at final concentration of 10 mM was added to approximately 1 mg/ml protein concentration in gel filtration buffer. Under reducing conditions, either 2 mM TCEP or 5 mM DTT was used. Results obtained were independent of reducing agent used. VP was then added at 2- to 100-fold excess per mole of cysteine residues. The cysteine-labeling reaction proceeded in the dark at room temperature for 60 min and was terminated by snap freezing in liquid nitrogen. For the reactions with substrate, sulfonucleotide was added at a 2- to 5-fold excess concentration per protein monomer prior to the addition of VP. Reactions with iodoacetamide were performed as above, except that iodoacetamide was used at 25 mM final concentration. Labeled samples were buffer-exchanged into NH4OAc as described below to remove small molecule reagents, and then diluted with 80:20 acetonitrile:water containing 1% formic acid for mass analysis.
Buffer exchange and sample preparation for mass spectrometry
Aliquots (100 μl) of purified sulfonucleotide reductase or thioredoxin were buffer exchanged into 50 mM NH4OAc (pH 7.5), using Amicon 10,000 Da molecular weight cutoff centrifugal filters with the temperature of the centrifuge set at 4 °C. The buffer was exchanged three times, and the final protein concentration was determined using Bradford assay. Stock solutions of APS, PAPS, and DTT were prepared in 50 mM NH4OAc as well. To generate reduced thioredoxin, a ten-fold molar excess of DTT was incubated with the protein for 15 min on ice. Mass analysis was performed to confirm the expected +2 Da shift, relative to oxidized thioredoxin. For enzyme-substrate incubation experiments with M. tuberculosis APS reductase, appropriate volumes of each component were mixed in the NH4OAc buffer and the mixtures were chilled on ice for at least 15 min before being introduced into the mass spectrometer. For the trypsin digest of APS reductase, 15 μM APS reductase was incubated in 50 mM NH4OAc (pH 7.5) with and without 10 molar excess APS at room temperature for 1 h with 0.4% (w/w) sequence grade trypsin. Some mixtures were further diluted with 80:20 acetonitrile:water containing 1% formic acid to detect possible covalent modifications (denatured mass analysis). For the experiments designed to observe intermediate formation and release with E. coli PAPS reductase and P. aeruginosa APS reductase, incubations were carried out for 15 min at the following concentrations: 10 μM enzyme, 100 μM APS or PAPS, 5 μM thioredoxin, and 50 μM DTT. After incubation, samples were dissolved in 80:20 acetonitrile:water containing 1% formic acid for mass analysis.
Mass spectrometry
Mass spectrometry data for M. tuberculosis APS reductase were acquired on a Bruker FT-ICR MS (Bruker, Billerica, Massachusetts, United States) equipped with an actively shielded 7 tesla superconducting magnet. Solutions were infused at a rate of 2 μl/min into an Apollo electrospray source (Bruker) operated in the positive mode. To analyze the enzyme and its reaction mixture in the native state, the syringe pump and spray chamber were wrapped with ice bags to prevent the protein sample from precipitating out of solution. The N2 nebulizing and drying gas pressure was maintained at 50 and 25 psi, respectively. The bias on the glass capillary was kept at 4,600 V, and 140 °C drying gas was used to assist in the desolvation process. Further desolvation was achieved by collisions of the ions with neutral buffer gas at the nozzle-skimmer region using a ∼180-V capillary exit voltage. A throttle valve was installed at the nozzle-skimmer region to adjust the pressure to approximately 1 × 10−6 mbar. Ions were externally accumulated in a radio frequency-only hexapole for 1 s before transfer into the ICR cell. Excessive kinetic energy was removed by colliding the ions with Argon pulsed into the cell to a pressure of approximately 1 × 10−8 mbar. Usually for one transient, two loops of Argon pulse were applied with a series of pump downs applied to lower the pressure in the cell to approximately 1 × 10−10 mbar before ion detection. All ions were collected using gated trapping and detected using chirp excitation and broad band data acquisition using an average of 16–64 time domain transients containing 32 K or 1 M data points. The original time domain free induction decay spectra were zero filled, Gaussian-multiplied and Fourier transformed. All the data were acquired and processed using Bruker Xmass version 6.0.0 software. The parameters of the ESI source, ion optics and cell were tuned for the best signal-to-noise ratio and were kept the same for systematic experiments.
Mass spectrometry data for E. coli PAPS reductase and P. aeruginosa APS reductase were acquired on a Q-TOF micro mass spectrometer (Waters, Milford, Massachusetts, United States) with electrospray source operated in positive mode. The spray probe used was silicon capillary, which was drawn down to a fine taper above a flame and cut manually to give the required diameter and flow. Samples were infused at a flow rate of 3 μl/min. The ESI source voltages were as follows: 3,200 V capillary, 35 V sample cone, and 1.5 V extraction cone. Source temperature was kept at 120 °C and analyzer pressure read-back was 4.5 × 10−5 mbar. A denatured myoglobin solution was used as the calibrant solution. The acquisition range was 500–2,500 m/z with an acquisition step of 1.2 s. All spectra were processed using MassLynx software (Waters).
Supporting Information
Figure S1 Gel Filtration Profile of Wild-Type and Mutant M. tuberculosis APS Reductase, E. coli PAPS Reductase, and P. aeruginosa APS Reductase
(A) Superdex 200 protein profile of M. tuberculosis APS reductase (filled circles), E. coli PAPS reductase (filled squares) and P. aeruginosa APS reductase (filled diamonds) followed by absorbance at 280 nm. Inset shows the calibration of the Superdex 200 column with known protein standards (blue dextran 2,000,000 Da, thyroglobulin 670,000 Da, bovine IgG 158,000 Da, human IgG 150,000 Da, bovine serum albumin 67,000 Da, ovalbumin 44,000 Da, chymotrypsin 25,000 Da, myoglobin 17,000 Da, and ribonuclease A 13,700 Da). Based on elution volume, the molecular weight (calculated from the standard curve of known protein standards) of M. tuberculosis APS reductase is 34,033 Da, E. coli PAPS reductase is 69,085 Da, and P. aeruginosa APS reductase is 128,078 Da. For each sulfonucleotide reductase the gel filtration profile of the catalytic cysteine to serine mutant has also been plotted (M. tuberculosis, open circles; E. coli, open squares; and P. aeruginosa, open diamonds).
(B) ESI-mass spectrometry spectrum of 10 μM APS reductase in 50 mM NH4OAc. Three charge states are observed, 9+, 10+ and 11+. The calculated mass is 28,706.00 Da using the deconvolution function on the Bruker Xmass software.
(2.3 MB EPS).
Click here for additional data file.
Figure S2 Gel Filtration and Activity Profile of Untagged and His-Tagged M. tuberculosis APS Reductase
(A) Superdex 200 profile of untagged M. tuberculosis APS reductase, purified in prior steps by anion exchange and ammonium sulfate fractionation, was followed by analysis of absorbance at 390 nm to detect the iron-sulfur cluster (solid line). The predicted molecular weight of the major peak is 34,334 Da, close to the expected value of 27,638 Da. Assay of untagged M. tuberculosis APS reductase activity across the gel filtration column (dashed line) as described in Materials and Methods.
(B) Activity of His-tagged M. tuberculosis APS reductase plotted in conjunction with absorption at 390 nm.
(708 KB EPS).
Click here for additional data file.
Figure S3 Gel Labeling of Wild-Type, Cys59Ser, and Cys249Ser M. tuberculosis APS Reductase
In this experiment, 5 μM wild type (lanes 1–3), Cys249Ser (lanes 4–6), and Cys59Ser (lanes 7–9) M. tuberculosis APS reductase was incubated at room temperature in 50 mM bis-tris propane (pH 7.0), 100 mM NaCl with [35S]APS only (lanes 1, 4, and 7); with 200 μM APS for 5 min prior to the addition of [35S]APS (lanes 2, 5, and 8); or with [35S]APS for 5 min followed by the addition of 10 μM thioredoxin (lanes 3, 6, and 9). SDS-PAGE load dye (without reductant) was added to terminate the reaction. The samples were heated at 60 °C for 3 min and size-fractionated by 12% nonreducing SDS-PAGE. The incorporation of radioactivity was analyzed with a Phosphorimager.
(2.1 MB TIF).
Click here for additional data file.
Figure S4 Gel Labeling of Wild-Type and Mutant E. coli and P. aeruginosa Sulfonucleotide Reductases
(A) Gel labeling of wild type (lanes 1–3) and Cys239Ser (lanes 4–6) E. coli PAPS reductase was carried out as described in the legend to Figure S3, except that PAPS was used as the substrate.
(B) Gel labeling of wild type (lanes 1–3) and Cys256Ser (lanes 4–6) P. aeruginosa reductase was carried out as described in the legend to Figure S3.
(1.7 MB TIF).
Click here for additional data file.
Figure S5 Intermediate Formation and Sulfite Release by M. tuberculosis, E. coli, and P. aeruginosa Sulfonucleotide Reductases
(A) Here, 5 μM M. tuberculosis APS reductase (lanes 3–6) or E. coli PAPS reductase (lanes 9–12) was incubated at room temperature in 50 mM bis-tris propane (pH 7.0), 100 mM NaCl with [35S]APS or [35S]PAPS in the presence (lanes 5, 6, 11, and 12) or absence (lanes 3, 4, 9, and 10) of 10 μM thioredoxin. Lanes 1, 2, 7, and 8 are control reactions in which sulfonucleotide reductase and thioredoxin have been omitted. Each odd lane is a sample of the reaction prior to addition of sulfonucleotide reductase. Each even lane is a sample from the reaction that has been terminated 10 min after the addition of sulfonucleotide reductase. No intermediate release was observed using 10 mM DTT in place of thioredoxin (unpublished data).
(B) Experiment exactly as above except with P. aeruginosa APS reductase.
(3.7 MB EPS).
Click here for additional data file.
Figure S6 Multiple Turnover TLC Assay for Sulfonucleotide Reduction
(A) TLC analysis to follow the extent of the sulfonucleotide reductase reaction, as described in Materials and Methods. A complete reaction progress curve for the reaction of trace [35S]APS, 20 μM APS, 10 μM thioredoxin, 5 mM DTT, and 20 nM M. tuberculosis APS reductase. The relative amounts of 35S in the product and substrate spots were quantified with a Phosphorimager.
(B) The time course of the reaction shown in (A) follows a single-exponential function.
(2 MB TIF).
Click here for additional data file.
Table S1 Oligonucleotide Primers Used in This Study
(33 KB DOC).
Click here for additional data file.
Accession Numbers
Available GenBank accession numbers for the sulfonucleotide reductase genes discussed in this paper are as follows: M. tuberculosis (CAB03733.1), R.
meliloti (AAD55759.1), P. aeruginosa (AAG05145.1), B. cepacia (AF170343), L. minor (CAB65911.1), A. thaliana (APR2) (AAC26977.1), E. coli (BAB37040.1), S. flexneri (AAP18092.1), S. cerevisiae (AAA34774.1), B. subtilis (CAB13431.1), and B.
anthracis (AAT30539.1).
We thank S. Long for the expression plasmid for R. meliloti APS reductase and K. Karbstein for helpful discussion and comments on the manuscript. This work was supported by National Institutes of Health Grant AI51622 (to CRB). KSC is a Cancer Research Fund Fellow of the Damon Runyon-Walter Winchell Foundation.
Competing Interests. The authors have declared that no competing interests exist.
Author contributions. KSC conceived and designed the experiments. KSC, HG, and HC preformed the experiments. KSC, HG, CDS, JAL, and CRB analyzed the data. KSC wrote the paper.
Citation: Carroll KS, Gao H, Chen H, Stout CD, Leary JA, et al. (2005) A Conserved mechanism for sulfonucleotide reduction. PLoS Biol 3(8): e250.
Abbreviations
4Fe-4Sfour iron-four sulfur cluster
AMPadenosine 5′-phosphate
APSadenosine 5′-phosphosulfate
Cyscysteine
DaDalton
DTNBdithio-1,4-nitrobenzoic acid
DTTdithiothreitol
Eenzyme
ESIelectrospray ionization
FT-ICRFourier transform ion-cyclotron resonance
Hishistidine
PAPS3′-phosphoadenosine 5′-phosphosulfate
Ssubstrate
Serserine
TCEPtris-(2-carboxyethyl)phosphine
TLCthin layer chromatography
TNB−5-thio-2-nitrobenzoate
VP4-vinylpyridine
==== Refs
References
Schwenn JD Photosynthetic sulphate reduction Z Naturforsch 1994 49c 531 539
Kredich NM Escherichia coli and Salmonella typhimurium Cellular and molecular biology 1996 Washington (DC) ASM Press 514 527
Williams SJ Senaratne RH Mougous JD Riley LW Bertozzi CR 5′-adenosinephosphosulfate lies at a metabolic branch point in mycobacteria J Biol Chem 2002 277 32606 32615 12072441
Chapman E Best MD Hanson SR Wong C Sulfotransferases: Structure, mechanism, biological activity, inhibition and synthetic utility Angew Chem Int Ed 2004 43 3526 3548
Lampreia J Pereira AS Moura JJG Adenylylsulfate reductases from sulfate-reducing bacteria Methods Enzymol 1994 243 241 260
Gonzalez Porque P Baldesten A Reichard P The involvement of the thioredoxin system in the reduction of methionine sulfoxide and sulfate J Biol Chem 1970 245 2371 2374 4392601
Lillig CH Prior A Schwenn JD Aslund F Ritz D New thioredoxins and glutaredoxins as electron donors of 3′-phosphoadenylylsulfate reductase J Biol Chem 1999 274 7695 7698 10075658
Tsang ML Schiff JA Assimilatory sulfate reduction in an Escherichia coli mutant lacking thioredoxin activity J Bacteriol 1978 134 131 138 25880
Holmgren A Thioredoxin and glutaredoxin systems J Biol Chem 1989 264 13963 13966 2668278
Sassetti CM Boyd DH Rubin EJ Comprehensive identification of conditionally essential genes in mycobacteria Proc Natl Acad Sci U S A 2001 98 12712 12717 11606763
Bick JA Dennis JJ Zylstra GJ Nowack J Leustek T Identification of a new class of 5′-adenylylsulfate (APS) reductases from sulfate-assimilating bacteria J Bacteriol 2000 182 135 142 10613872
Kopriva S Buchert T Fritz G Suter M Benda R The presence of an iron-sulfur cluster in adenosine 5′-phosphosulfate reductase separates organisms utilizing adenosine 5′-phosphosulfate and phosphoadenosine 5′-phosphosulfate for sulfate assimilation J Biol Chem 2002 277 21786 21791 11940598
Setya A Murillo M Leustek T Sulfate reduction in higher plants: Molecular evidence for a novel 5′-adenylylsulfate reductase Proc Natl Acad Sci U S A 1996 93 13383 13388 8917600
Berendt U Haverkamp T Prior A Schwenn JD Reaction mechanism of thioredoxin: 3′-phospho-adenylylsulfate reductase investigated by site-directed mutagenesis Eur J Biochem 1995 233 347 356 7588765
Schwenn JD Krone FA Husmann K Yeast PAPS reductase: Properties and requirements of the purified enzyme Arch Microbiol 1988 150 313 319 3060034
Abola AP Willits MG Wang RC Long SR Reduction of adenosine-5′-phosphosulfate instead of 3′-phosphoadenosine-5′-phosphosulfate in cysteine biosynthesis by Rhizobium meliloti and other members of the family Rhizobiaceae J Bacteriol 1999 181 5280 5287 10464198
Suter M von Ballmoos P Kopriva S den Camp RO Schaller J Adenosine 5′-phosphosulfate sulfotransferase and adenosine 5′-phosphosulfate reductase are identical enzymes J Biol Chem 2000 275 930 936 10625629
Kopriva S Buchert T Fritz G Suter M Weber M Plant adenosine 5′-phosphosulfate reductase is a novel iron-sulfur protein J Biol Chem 2001 276 42881 42886 11553635
Berndt C Lillig CH Wollenberg M Bill E Mansilla MC Characterization and reconstitution of a 4Fe-4S adenylyl sulfate/phosphoadenylyl sulfate reductase from Bacillus subtilis
J Biol Chem 2004 279 7850 7855 14627706
Fersht A Structure and mechanism in protein science: A guide to enzyme catalysis and protein folding 1999 New York W. H. Freeman 650
Segel IH Enzyme kinetics: Behavior and analysis of rapid equilibrium and steady-state enzyme systems 1975 Hoboken (New Jersey) John Wiley and Sons 957
Kim SK Rahman A Bick JA Conover RC Johnson MK Properties of the cysteine residues and iron-sulfur cluster of the assimilatory 5′-adenylyl sulfate reductase from Pseudomonas aeruginosa
Biochemistry 2004 43 13478 13486 15491155
Weber M Suter M Brunold C Kopriva S Sulfate assimilation in higher plants characterization of a stable intermediate in the adenosine 5′-phosphosulfate reductase reaction Eur J Biochem 2000 267 3647 3653 10848982
Hernandez H Hewitson KS Roach P Shaw NM Baldwin JE Observation of the iron-sulfur cluster in Escherichia coli biotin synthase by nanoflow electrospray mass spectrometry Anal Chem 2001 73 4154 4161 11569804
Johnson KA Verhagen MF Brereton PS Adams MW Amster IJ Probing the stoichiometry and oxidation states of metal centers in iron-sulfur proteins using electrospray FTICR mass spectrometry Anal Chem 2000 72 1410 1418 10763234
Tsang ML Schiff JA Sulfate-reducing pathway in Escherichia coli involving bound intermediates J Bacteriol 1976 125 923 933 3497
Khoroshilova N Popescu C Munck E Beinert H Kiley PJ Iron-sulfur cluster disassembly in the FNR protein of Escherichia coli by O2 : [4Fe-4S] to [2Fe-2S] conversion with loss of biological activity Proc Natl Acad Sci U S A 1997 94 6087 6092 9177174
Brunold C Suter M Sulphur metabolism B. Adenosine 5′-phosphosulphate sulphotransferase Methods Plant Biochem 1990 3 339 342
Schwartz CJ Djaman O Imlay JA Kiley PJ The cysteine desulfurase, IscS, has a major role in in vivo Fe-S cluster formation in Escherichia coli
Proc Natl Acad Sci U S A 2000 97 9009 9014 10908675
Zheng L White RH Cash VL Jack RF Dean DR Cysteine desulfurase activity indicates a role for NIFS in metallocluster biosynthesis Proc Natl Acad Sci U S A 1993 90 2754 2758 8464885
Zheng L White RH Cash VL Dean DR Mechanism for the desulfurization of L-cysteine catalyzed by the nifS gene product Biochemistry 1994 33 4714 4720 8161529
Frazzon J Dean DR Formation of iron-sulfur clusters in bacteria: An emerging field in bioinorganic chemistry Curr Opin Chem Biol 2003 7 166 173 12714048
Frazzon J Dean DR Biosynthesis of the nitrogenase iron-molybdenum-cofactor from Azotobacter vinelandii
Met Ions Biol Syst 2002 39 163 186 11913125
Bordo D Deriu D Colnaghi R Carpen A Pagani S The crystal structure of a sulfurtransferase from Azotobacter vinelandii highlights the evolutionary relationship between the rhodanese and phosphatase enzyme families J Mol Biol 2000 298 691 704 10788330
Pagani S Eldridge M Eady RR Nitrogenase of Klebsiella pneumoniae . Rhodanese-catalysed restoration of activity of the inactive 2Fe species of the Fe protein Biochem J 1987 244 485 488 3311031
Cicero DO Melino S Orsale M Brancato G Amadei A Structural rearrangements of the two domains of Azotobacter vinelandii rhodanese upon sulfane sulfur release: Essential molecular dynamics, 15N NMR relaxation and deuterium exchange on the uniformly labeled protein Int J Biol Macromol 2003 33 193 201 14607364
Jacob C Holme AL Fry FH The sulfinic acid switch in proteins Org Biomol Chem 2004 2 1953 1956 15254616
Woo HA Jeong W Chang TS Park KJ Park SJ Reduction of cysteine sulfinic acid by sulfiredoxin is specific to 2-cys peroxiredoxins J Biol Chem 2005 280 3125 3128 15590625
Woo HA Chae HZ Hwang SC Yang KS Kang SW Reversing the inactivation of peroxiredoxins caused by cysteine sulfinic acid formation Science 2003 300 653 656 12714748
Wood ZA Poole LB Karplus PA Peroxiredoxin evolution and the regulation of hydrogen peroxide signaling Science 2003 300 650 653 12714747
Biteau B Labarre J Toledano MB ATP-dependent reduction of cysteine-sulphinic acid by S. cerevisiae sulphiredoxin Nature 2003 425 980 984 14586471
Kagedal B Kallberg M Sorbo B A possible involvement of glutathione in the detoxication of sulfite Biochem Biophys Res Commun 1986 136 1036 1041 2872886
Hausinger RP Howard JB Thiol reactivity of the nitrogenase Fe-protein from Azotobacter vinelandii
J Biol Chem 1983 258 13486 13492 6580291
Gurrath M Friedrich T Adjacent cysteines are capable of ligating the same tetranuclear iron-sulfur cluster Proteins 2004 56 556 563 15229887
Beinert H Kennedy MC Stout CD Aconitase as iron-sulfur protein, enzyme, and iron-regulatory protein Chem Rev 1996 96 2335 2374 11848830
Beinert H Micro methods for the quantitative determination of iron and copper in biological material Methods Enzymol 1978 54 435 445 732579
Kennedy MC Kent TA Emptage M Merkle H Beinert H Evidence for the formation of a linear [3Fe-4S] cluster in partially unfolded aconitase J Biol Chem 1984 259 14463 14471 6094558
Leyh TS Taylor JC Markham GD The sulfate activation locus of Escherichia coli K12: Cloning, genetic, and enzymatic characterization J Biol Chem 1988 263 2409 2416 2828368
Schwedock J Long SR ATP sulphurylase activity of the nodP and nodQ gene products of Rhizobium meliloti
Nature 1990 348 644 647 2250719
Segel IH Renosto F Seubert PA Sulfate-activating enzymes Methods Enzymol 1987 143 334 349 2821345
O'Brien PJ Herschlag D Alkaline phosphatase revisited: Hydrolysis of alkyl phosphates Biochemistry 2002 41 3207 3225 11863460
Peluso P Shan SO Nock S Herschlag D Walter P Role of SRP RNA in the GTPase cycles of Ffh and FtsY Biochemistry 2001 40 15224 15233 11735405
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
| 16008502 | PMC1175818 | CC BY | 2021-01-05 08:21:48 | no | PLoS Biol. 2005 Aug 19; 3(8):e250 | utf-8 | PLoS Biol | 2,005 | 10.1371/journal.pbio.0030250 | oa_comm |
==== Front
PLoS BiolPLoS BiolpbioplosbiolPLoS Biology1544-91731545-7885Public Library of Science San Francisco, USA 1600850310.1371/journal.pbio.0030262Research ArticleEcologyEvolutionInfectious DiseasesMicrobiologyZoologyParasitologyInsectsPrevalence-Dependent Costs of Parasite Virulence Prevalence-Dependent VirulenceBedhomme Stephanie
1
Agnew Philip
2
Vital Yuri
2
Sidobre Christine
2
Michalakis Yannis [email protected]
2
1Department of Biology, Queen's University, Kingston, Ontario, Canada,2Génétique et Evolution des Maladies Infectieuses, Montpellier, FranceEbert Dieter Academic EditorUniversity of BaselSwitzerland8 2005 19 7 2005 19 7 2005 3 8 e26225 1 2005 26 5 2005 Copyright: © 2005 Bedhomme 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.
Prevalence of Infection in a Population Can Shape Parasite Virulence
Costs of parasitism are commonly measured by comparing the performance of infected groups of individuals to that of uninfected control groups. This measure potentially underestimates the cost of parasitism because it ignores indirect costs, which may result from the modification of the competitiveness of the hosts by the parasite. In this context, we used the host-parasite system consisting of the yellow fever mosquito Aedes aegypti and the microsporidian parasite Vavraia culicis to address this question: Do infected individuals exert a more or less intense intraspecific competition than uninfected individuals? Our experimental results show that, indeed, infected hosts incur a direct cost of parasitism: It takes them longer to become adults than uninfected individuals. They also incur an indirect cost, however, which is actually larger than the direct cost: When grown in competition with uninfected individuals they develop even slower. The consequence of this modification of competitiveness is that, in our system, the cost of parasitism is underestimated by the traditional measure. Moreover, because the indirect cost depends on the frequency of interactions between infected and uninfected individuals, our results suggest that the real cost of parasitism, i.e., virulence, is negatively correlated with the prevalence of the parasite. This link between prevalence and virulence may have dynamical consequences, such as reducing the invasion threshold of the parasite, and evolutionary consequences, such as creating a selection pressure maintaining the host's constitutive resistance to the parasite.
The outcome of competition between mosquito larvae infected with a parasite depends on the infection status of the individual concerned and on that of its competitor, with implications for measuring virulence and for host-parasite dynamics.
==== Body
Introduction
The presence of parasitism creates heterogeneity in host populations: Parasitised hosts may have different behaviour [1], different food requirements [2], different feeding rates [3], or different sensitivity to stress, such as pollution [4]. This heterogeneity can modify intraspecific competition among individuals within host populations. There are three types of competitive interaction to consider: Among uninfected hosts, among infected hosts, or between uninfected and infected hosts. If these interactions differed in their effects on host life-history traits, the presence of parasitism would induce not only a direct cost of infection but also indirect costs through the modification of intraspecific competitiveness. If infected hosts differ in competitive strength from uninfected hosts, the outcome of competitive interactions will depend on which individuals are involved, and thus indirect costs will depend on the parasite's prevalence.
These potential indirect effects of parasitism are especially important within the context of measuring the costs of parasitism. Indeed, what is typically measured is the direct cost only; normally, this is done by comparing the performance of individuals coming from either infected or uninfected populations, and the indirect costs resulting from the modification of the competitive ability of infected individuals are not taken into account.
We used the system involving the yellow fever mosquito Aedes aegypti and the microsporidian parasite Vavraia culicis to investigate whether infected hosts exert more or less intense intraspecific competition than uninfected hosts. That is to say, we investigated whether the parasitism modifies the host's intraspecific competitiveness. We addressed this issue in our experiment by measuring both direct and indirect costs, by comparing intraspecific competition between and among parasitised and unparasitised individuals. More specifically, two individuals, which were infected or not, competed for resources in standard growing vials. This resulted in four treatments, according to the infection status of competing larvae, which were: (i) infected versus infected (+/+), (ii) infected versus control (+/−), (iii) control versus infected (−/+), and (iv) control versus control (−/−), where the first term refers to the focal individual and the second to its competitor. To take into account possible interactions with the amount of resources available, we considered two food regimes, 100% and 50% of a standard regime detailed in the Materials and Methods section. We measured the direct and indirect effects of parasitism through the modifications it induced on the probability to emerge, i.e., the probability to reach the adult stage, and on developmental time of the focal individual as a function of the infection status of the focal individual and its competitor.
Results
The probability of both individuals emerging when receiving a 100% diet was high and similar for the “−/−” and “+/+” treatments (89.6% and 86.3%, respectively). This probability remained high when receiving a 50% regime for the “−/−” treatment (94.4%) but was lower for the “+/+” treatment (67.8%) and caused a significant food-by-treatment interaction (chi-square = 8.823, degrees of freedom [d.f.] = 1, p = 0.003). The frequency of pre-emergence mortality observed in the “+/−” treatments however was not different from that which would be expected on the combined mortality of infected and uninfected individuals estimated from the “+/+” and “−/−” treatments (100% regime, chi-square = 1.458, d.f. = 1, p = 0.227; 50% regime, chi-square = 0.635, d.f. = 1, p = 0.235). Furthermore, of the 19 cases where only one individual emerged from a “+/−” treatment receiving the 50% regime, 11 were uninfected individuals and eight were infected individuals, showing that pre-emergence mortality in “+/−” treatments was not strongly biased towards the infected individual. Thus, the 50% regime decreased the overall probability of an infected individual emerging, but there was no evidence that being in competition with an uninfected individual amplified this effect.
We found significant effects for the effect of food on developmental time (Table 1): Individuals with a 100% regime developed faster. A significant effect of the “infection status” was found for developmental time (Figure 1): Infected individuals had a longer developmental time. The effect of infection is also illustrated by considering the status of the first individual to emerge in “+/−” vials. Infected individuals emerged first only in 22.5% cases (Fisher's exact test p < 0.001). Controlling for gender composition showed that the infection effect is even stronger: In vials with two males, the infected individual emerged first in 19.6% of the cases (Fisher's exact test p < 0.001), while in vials with two females, the infected individual emerged first in 10.1% of the cases (Fisher's exact test p < 0.001). Moreover, as Figure 1 shows, individuals having an infected competitor (i.e., “−/+” and “+/+” individuals) had a shorter developmental time than individuals having an uninfected competitor (i.e., “+/−” and “−/−” individuals). The subsequent analyses allowed us to verify this influence of the nature of the competitor: “+/+” individuals had a shorter developmental time (p = 0.019) than “+/−” individuals, and “−/+” individuals had a shorter developmental time (p < 0.001) than “−/−” individuals. Sex status had a significant effect corresponding to known differences between male and female mosquitoes: Males had a shorter developmental time than females. However, the interaction between sex status and infection status was not significant.
Figure 1 Developmental Time in Days and 95% Confidence Intervals for the Four Categories of Infection Status
In the infection status category denominations (e.g., +/+, +/−, −/+, and −/−), the first symbol indicates the infection status of the focal individual, and the second indicates that of its competitor. The filled black arrow represents the commonly measured direct costs of parasitism, and the open grey arrows represent the indirect costs through the modification of intraspecific competitiveness by parasitism (see text).
Table 1 ANOVAS for the Effects of Food Regime, Sex Status, and Infection Status on Developmental Time
Discussion
Our data indicate that infection causes an increase in developmental time of the host, which, when combined with lower larval food availability, decreases the probability of infected individuals surviving to emerge as adults. Moreover, we show that parasitism modifies intraspecific competitiveness for the host. Indeed, individuals grown with an infected individual had a shorter developmental time. This suggests that the intraspecific competition exerted by infected individuals is less intense than for uninfected individuals.
An increase in developmental time is costly for hosts not only because it lengthens generation time [5], but also because the rate of pre-adult mortality due to V. culicis increases with time [6].
For male mosquitoes, an increase in developmental time is likely to entail reduced reproductive success. This is because the development of larval populations often begins synchronously following the immersion of eggs by water and because adult females become refractory to copulation after being mated once. Thus, in male-male competition for females, slower larval development will be associated with reduced access to mating.
The less intense competition exerted by infected host individuals could be explained by a reduced feeding rate due to a less intense competitiveness for food. They could also show less physical activity and thus generate less stress by reduced contact with their competitors.
Our results could be interpreted as a simple variation in the expression of the costs of parasitism depending on environmental conditions, as has been shown for many other host-parasite systems [6–9]. However, in the present experiment, the variation in the expression of virulence is due to the presence of the parasite itself in other individuals of the host population and, therefore, implies feedbacks in dynamic and evolutionary processes, as discussed below.
Less intense competitiveness of infected individuals has already been reported in host-parasitoid interactions: Sisterson and Averill [10] showed that larvae of Acrobasis vaccinii (Lepidoptera), which are parasitised by the parasitoid Phanerotoma franklini (Hymenoptera), defended their resources better against parasitised rather than against unparasitised hosts. Other experimental studies have shown that parasitoidism may considerably modify the effects of intraspecific competition among hosts [10–15], and theoretical studies have suggested that it can be an important factor in their population dynamics [16–18]. However, the consequences of a modification in the intensity of intraspecific competition by parasitism on parasite virulence and its evolutionary consequences were not explored.
Consequences for Measures of Virulence
The intensity of intraspecific competition was modified by parasitism in our system. This implies that the costs of parasitism are not completely captured by comparing individuals from a completely infected population to individuals of a completely uninfected population. If we concentrate on our results for developmental time, this kind of classical measurement of virulence is symbolised by the filled black arrow in Figure 1. This comparison between infected and uninfected populations underestimates the full cost of parasitism because it ignores the cost induced by the modifications of intraspecific competitiveness, represented by the open grey arrows in Figure 1. A more realistic evaluation of the full cost of infection should incorporate all types of costs, the relative weight of each type depending on the frequency of each type of competitive interaction in the population. In populations with a low prevalence of parasitism, infected individuals interact mainly with uninfected competitors. This type of competitive interaction is the most costly for infected hosts. Thus, the virulence expressed in populations with a low prevalence of infection may be high. In contrast, where the prevalence of infection is high, infected individuals mainly interact with other infected individuals, and thus will not pay the indirect cost of parasitism. Virulence may thus be negatively correlated with parasite prevalence. However, we used here a minimalist representation of prevalence, and at least two other phenomena could influence the shape of the relationship between prevalence and virulence. The first is that infected individuals could aggregate to reduce the fitness cost of infection by reducing their contact with uninfected individuals. The second is that high-prevalence conditions favor high parasite burden and multiple infections. High parasite burden increases the negative effects of parasites. Multiple infection, on the other hand, may either increase [19] or decrease [20–22] parasite virulence, depending on how various factors interact to determine parasitic virulence. The slope of the relationship between prevalence and virulence could thus be modified by integrating the other correlates of high prevalence.
The same type of frequency-dependent costs have also been found in the evaluation of inbreeding depression for several plant species [23–25]: The depression is generally greater when the competitors of an inbred individual are mostly outbred, as opposed to when they are mostly inbred. The existence of frequency-dependent costs for these two very different biological phenomena suggests that indirect costs due to the modification of intraspecific competitiveness could be common. Taking the composition of the population into account when evaluating a cost could thus be worthwhile for other phenomena.
Evolutionary Implications
Modifications of the competitive ability of hosts by parasites have been shown to affect the population dynamics of hosts and parasites [17,18]. For example, in our system, in low-prevalence conditions, the strength of intraspecific competition against infected hosts is intense and leads to a large increase in developmental time. This enhances the parasite's potential transmission success and may help it to become established in naïve host populations by reducing the invasion threshold.
However, the evolutionary consequences of the modification of host competitiveness by parasites have not been considered, and in particular, models of virulence evolution [19,26–30] have not included it.
Prevalence-dependent virulence may also influence the evolution of constitutive host resistance to parasites. Indeed, in models for the evolution of resistance to parasites [31], low prevalence represents conditions in which the cost of resistance greatly reduces the fitness of resistant individuals, because a majority of them pay the cost without receiving the benefit of carrying a resistance allele. Low prevalence is thus a condition in which resistance to a parasite is counterselected. A negative relation between prevalence and virulence, as suggested by our results, would increase the relative fitness advantage of resistant individuals in populations with low prevalences of infection and could partially counterbalance selection against resistant individuals. This verbal model needs to be substantiated by a theoretical approach that includes a more realistic view of infection characteristics such as environmental and temporal variation in prevalence, parasite burden, or multiple infections.
In conclusion, our experiment reveals that, in our system, the outcome of competition among host larvae not only depends on the infection status of the individual concerned but also on that of its competitors. If this result is true in other host-parasite systems, it has several possible consequences: From an experimental point of view, the measurement of the parasite's virulence can be affected by the prevalence under which the measure is made. At the population scale, the virulence of the parasite varies with its prevalence, and this may affect the dynamics of the host-parasite system. From an evolutionary point of view, this phenomenon may influence the evolution of parasite virulence and host resistance.
Materials and Methods
Biological system
The microsporidium V. culicis naturally infects several genera of mosquitoes [32]. Host larvae ingest the parasites' spores along with their food and become horizontally infected when these spores germinate and infect host gut cells. Within host cells, V. culicis undergoes a series of developmental stages before starting to produce its spores from 8 to 10 days postinfection. Physical damage to host tissues occurs when spore-laden cells rupture and disseminate their contents. As the parasites' spores do not resist desiccation [33], its transmission success is more likely to be assured by spores released from the body of dead and decaying larvae or pupae rather than from infected individuals that emerge and leave the aquatic environment as adult mosquitoes (p. 455, [34]). Thus, larval developmental time and survival to adulthood are important life-history traits influencing the parasite's transmission success.
Ae. aegypti is a subtropical mosquito whose larvae grow in natural or artificial containers [35]. These sites show temporal variation in size and food availability, so larvae experience variable conditions of intraspecific competition. Larvae feed by filtering water, while pupae do not feed.
In previous studies, we found that intraspecific competition among uninfected Ae. aegypti larvae can strongly influence their developmental time and survival to adulthood [36,37]. Intraspecific competition not only involves competition for food [38], but includes all the detrimental environmental changes induced by the presence of conspecifics. These modifications can be stress-generated by physical contacts [39] or chemical pollution (e.g., by nitrogenous waste [40]), as previously demonstrated for Ae. aegypti larvae [41]. Concerning the effects of parasitism, we have shown that V. culicis prolongs developmental time and reduces the chances of reaching adulthood for Ae. aegypti larvae reared in the absence of intraspecific competition [42]. Furthermore, the expression of the parasite's virulence varied along a gradient of environmental resource availability, reflecting the strength of interspecific competition between host and parasite for host resources [42]. These results suggest that intraspecific competition is likely to interact and influence the expression of the parasite's virulence.
Experimental design
Our strain of Ae. aegypti is derived from a large number of eggs collected in Tingua, Brazil and provided by Ricardo Lourenço de Oliveira of the Instituto Oswaldo Cruz (Rio de Janeiro, Brazil). It had been reared in standardised and outbred conditions (3,000 reproductive adults in each generation) in our laboratory for three generations at the time of the experiment. The spores of V. culicis were derived from a stock isolated from Ae. albopictus in Florida and provided by Dr. J. J. Becnel (United States Department of Agriculture, Gainesville, Florida, United States).
Recently hatched Ae. aegypti were split into 30 groups of 60 larvae each, and each group was put in a petri dish (diameter 55 mm) containing 10 ml of softened water. We added 6 × 104
V. culicis spores per larva in 0.05 ml of softened water to 15 of the dishes. These latter individuals are subsequently called “exposed individuals.” To the remaining 15 petri dishes, we added 0.05 ml of softened water; we refer to these as “control individuals.” To each dish, we added 3.6 mg of Tetramin (powdered fish food). Spores and larvae were kept in contact for 24 h. Contact between larvae and spores was restricted to 24 h in order to synchronise the age structure of infections.
After this infection period, larvae were rinsed and transferred to individual Drosophila vials (diameter 25 mm × 95 mm), two per vial, containing 5 ml of softened water.
We had three categories of vials, representing the three cases of intraspecific competitive interactions in a partially infected host population: (i) Vials containing two control larvae, (ii) vials containing two exposed larvae, and (iii) vials containing a control larva and an exposed larva. This minimalistic approach of manipulating density has already been shown to capture the general effects of intraspecific competition, while avoiding pitfalls of more traditional approaches [37]. During their development, larvae were provided daily with their food dissolved in 1 ml of softened water. Prior to feeding, 1 ml of water was removed to maintain a constant volume, because the depth of the environment in which larvae grow has been shown to influence mosquito life history traits [43].
Two food regimes were adopted: 50% and 100% of a standard regime. The standard regime consisted of 0.08 mg on day 1, 0.16 mg on day 2, 0.32 mg on day 3, and 0.64 mg of Tetramin per vial from day 4 onwards. In the 50% food regime, these quantities were divided by two. Food was provided to vials until both individuals in each vial had either died or pupated, and the amount of food was not adjusted to the number of larvae remaining in the vial. This procedure was followed to allow the maximum number of individuals to reach adulthood and their gender to be determined, thus maximizing statistical power in the analysis of gender effects. Vials were arranged in racks of four-by-ten vials where each rack was assigned to a food treatment, half receiving the 50% regime and half the 100% food regime. The three infection treatments were distributed randomly among racks. The number of vials in each category was calculated, allowing for a 95% infection rate and sex-ratio variation, so as to obtain a minimum of 30 replicates in each infection/food/sex combination. There was a total of 775 vials. The experiment was divided into seven blocks in order to reduce the effects of uncontrolled environmental variations. The experiment was conducted in a room maintained at 25 °C and a photoperiod of 12 h of light to 12 h of dark.
Vials were examined every 12 h, and age at pupation was recorded. Pupae were transferred to individual vials containing 5 ml of softened water and the vials were covered with a fine nylon gauze. At emergence, the adult sex was noted. For all individuals (larvae, pupae, and adults) having an age at death of more than 8 d and coming from “+/+” and “+/−” treatments and 89 individuals coming from the “−/−” treatment, spore load was evaluated by homogenising the body of the individual in 0.2 ml of water and counting the number of spores with a Neubauer cell counter and phase-contrast light microscope. We considered as infected all individuals for which we observed more than one spore, each spore observed under the microscope corresponding to 2,000 spores within an infected mosquito. Subsequent analyses were restricted to vials whose a posteriori infection status matched the a priori status. Dead mosquitoes were stored at −20 °C.
Statistical analyses
The probability of emerging as an adult in the “+/+” and “−/−” treatments was estimated by comparing the frequency of vials in which both individuals emerged as adults as opposed to vials in which at least one individual died before this stage. From these data we estimated the probability of an infected or uninfected individual emerging. These estimates were used to generate predictions for the frequency of emergence in the “+/−” treatments and against which the observed frequencies were tested. Only vials we could confirm as not having their a priori infection status were excluded from these analyses, those which could not be confirmed because of mortality before spore production began were assumed to have their a priori infection status.
Only replicates in which both individuals survived to pupation were used for the subsequent analyses. This insured that larval mortality did not influence traits of the surviving individuals. Moreover, as mentioned above, we included in the analyses only individuals from the vials in which the observed infection status matched the expected status. The size of the dataset was thus reduced to 539 vials.
The analysis presented here was specifically designed to answer the question, Do infected hosts exert more or less intense intraspecific competition than uninfected hosts? In a previous study [36], it was shown that the sex of a competitor could influence the intensity of competition. The sex of the competitor was thus included in our analyses. One of the two mosquitoes in each tube was randomly selected as the focal individual to be analysed, with the other designated as its competitor. We combined our knowledge of the infection status and sex of both individuals into two factors, infection status and sex status. There were four categories in each factor. For infection status, these categories were (i) infected versus infected (+/+), (ii) infected versus control (+/−), (iii) control versus infected (−/+), and (iv) control versus control (−/−), where the first term refers to the focal individual and the second to its competitor. Correspondingly, the categories for sex status were (i) female versus female, (ii) female versus male, (iii) male versus female, and (iv) male versus male, with first and second terms referring to the focal and competitor individuals, respectively. As it is only data from the focal individual being analysed, the degrees of freedom in our analysis are based on the number of replicate vials and not on the number of individuals in the data set.
The factors of infection status, sex status, and food regime were analysed by fully factorial ANOVA. However, the results of such an analysis may have depended on the particular combination of mosquitoes randomly selected as the focal individual. Consequently, we resampled the data by repeating the randomisation process and analysis 200 times. We had previously verified that the mean of the estimated F-values was stable with fewer than 200 resamplings. The results presented correspond to the mean of estimated values obtained by the resampling procedure. The p-values presented are from the upper 95% quantile of the distribution of p-values obtained and are, thus, conservative estimates. Based on the results of this first analysis, we decided to test for the effect of the nature of the competitor. To do so, we first performed the same type of analysis (with resampling and ANOVA) on a data file containing only “+/+” and “+/−” individuals to test the effect of the nature of the competitor on infected individuals. We then performed the same type of analysis on a data file containing only “−/+” and “−/−” individuals to test for the effect of the nature of the competitor on control individuals. These analyses had sex status, infection status, and food regime as factors, and are equivalent to a contrasts analysis.
Statistical analyses were performed with JMP, version 3.2.2 (SAS Institute, http://www.sas.com/) and Splus 2000 (MathSoft, http://www.mathsoft.com/).
Life history traits
We analysed the effects of competition and parasitism on the probability of emergence and developmental time. More specifically, we did not examine their effects on adult body size or starved adult longevity because of the food distribution pattern we followed in this experiment. Indeed, as mentioned above, a given amount of food was distributed to each vial daily until both individuals of the vial had either pupated or died, implying that after the first individual had pupated the second received the assigned amount of food for two individuals. Therefore, slow-developing individuals received more food than fast-developing individuals, which renders body size and longevity data uninterpretable. The food provisioning bias can only weaken the effects we reveal on survival and developmental time, which renders our analyses on these traits conservative in this respect.
This research was funded by a CNRS ATIP grant to YM. Many thanks to M. Choisy for help in statistical analyses and to four anonymous reviewers for their useful comments on the manuscript.
Competing Interests. The authors have declared that no competing interests exist.
Author contributions. SB, PA, and YM conceived and designed the experiments. SB, PA, YV, and CS performed the experiments. SB, PA, and YM analyzed the data. SB, PA, and YM wrote the paper.
Citation: Bedhomme S, Agnew P, Vital Y, Sidobre C, Michalakis Y (2005) Prevalence-dependent costs of parasite virulence. PLoS Biol 3(8): e262.
Abbreviation
d.f.degree(s) of freedom
==== Refs
References
Thomas F Renaud F de Meeus T Manipulation of host behaviour by parasites: Ecosystem engineering in the intertidal zone Proc R Soc Lond B Biol Sci 1998 265 1091 1096
Thompson SN Redak RA Wang L-W Altered dietary nutrient intake maintains metabolic homeostasis in parasitized larvae of the insect Manduca sexta L J Exp Biol 2001 204 4065 4080 11809781
Rivero A Ferguson HM The energetic budget of Anopheles stephensi infected with Plasmodium chabaudi : Is energy depletion a mechanism for virulence? Proc R Soc Lond B Biol Sci 2003 270 1365 1371
Brown AF Pascoe D Parasitism and host sensitivity to cadmium: An acanthocephalan infection of the freshwater amphipod Gammarus pulex
J Appl Ecol 1989 26 473 487
Stearns SC The evolution of life histories 1992 Oxford Oxford University Press 249
Agnew P Berticat C Bedhomme S Sidobre C Michalakis Y Parasitism increases and decreases the costs of insecticide resistance in mosquitoes Evolution 2004 58 579 586 15119441
Blanford S Thomas MD Pugh JK Pell JK Temperature checks the Red Queen? Resistance and virulence in fluctuating environment Ecol Lett 2003 6 2 5
Brown MJF Loosli R Scmid-Hempel P Condition-dependent expression of virulence in a trypanosome infecting bumblebees Oïkos 2000 91 421 427
Ferguson HM Read AF Genetic and environmental determinants of malaria parasite virulence in mosquitoes Proc R Soc Lond B Biol Sci 2002 269 1217 1224
Sisterson MS Averill AL Interactions between parasitized and unparasitized conspecifics: parasitoids modulate competitive dynamics Oecologia 2003 135 362 371 12721825
Prevost G Etude expérimentale des interactions entre parasitisme et compétition larvaire chez Drosophila melanogaster meigen [dissertation] 1985 Lyon Université Claude Bernard Lyon I 119 p. Available from the Université Claude Bernard Lyon I library, Lyon, France
Wajnberg E Prevost G Boulétreau M Genetic and epigenetic variation in Drosophila larvae suitability to a hymenopterous endoparasitoid Entomophaga 1985 30 187 191
Reed DJ Begon M Thompson DJ Differential cannibalism and population dynamics in a host-parasitoid system Oecologia 1996 105 189 193
Washburn JO Mercer DR Anderson JR Regulatory role of parasites: Impact on host population shifts with resource availability Science 1991 253 185188
Bernstein C Heizmann A Desouhant E Intraspecific competition between healthy and parasitised hosts in a host-parasitoid system: Consequences for life-history traits Ecol Entomol 2002 27 415 423
Bernstein C Density-dependence and the stability of host-parasitoid systems Oïkos 1986 47 176 180
Hochberg ME Population dynamic consequences of the interplay between parasitism and intraspecific competition for host-parasite systems Oïkos 1991 61 297 306
Spataro T Bernstein C Combined effects of intraspecific competition and parasitoid attacks on the dynamics of a host population: A stage-structured model Oïkos 2004 105
Frank S Models of parasite virulence Q Rev Biol 1996 71 37 78 8919665
Chao L Hanley KA Burch CL Dahlberg C Turner PE Kin selection and parasite evolution: Higher and lower virulence with hard and soft selection Q Rev Biol 2000 75 261 275 11008699
Brown SP Hochberg ME Grenfell BT Does multiple infection select for raised virulence? Trends Microbiol 2002 10 401 405 12217504
Schjorring S Koella JC Sub-lethal effects of pathogens can lead to the evolution of lower virulence in multiple infections Proc R Soc Lond B Biol Sci 2003 270 189 193
Cheptou P-O Schoen D Frequency-dependent inbreeding depression in Amsinckia
Am Nat 2003 162 744 753 14737712
Cheptou P-O Lepart J Escarré J Inbreeding depression under intraspecific competition in a highly outcrossing population of Crepis sancta (Asteracae): Evidence for frequency-dependent variation Am J Bot 2001 88 1424 1429 21669674
Schmitt J Ehrhardt DW Enhancement of inbreeding depression by dominance and supression in Impatiens capensis
Evolution 1990 44 269 278
Day T Burns JG A consideration of patterns of virulence arising from host-parasite coevolution Evolution 2003 57 671 676 12703956
Day T Proulx SR A general theory for the evolutionary dynamics of virulence Am Nat 2004 163 E40 E63 15122509
Ebert D Weisser WW Optimal killing for obligate killers: The evolution of life-histories and virulence of semelparous parasites Proc R Soc Lond B Biol Sci 1997 264 985 991
Gandon S Jansen VAA van Baalen M Host life history and the evolution of parasite virulence Evolution 2001 55 1056 1062 11430642
van Baalen M Coevolution of recovery ability and virulence Proc R Soc Lond B Biol Sci 1998 265 317 325
Gillespie JH Natural selection for resistance to epidemics Ecology 1975 56 493 495
Weiser J Data sheet on the biological control agent Vavraia (Pleistophora) culicis (Weiser 1946) 1980 Geneva World Health Organisation 1 5
Kelly JF Anthony DW Dillard CR A laboratory evaluation of the microsporidian Vavraia culicis as an agent for mosquito control J Invert Pathol 1981 37 117 122
Becnel JJ Andreadis TG Wittner M Microsporidia in insects The microsporidia and microsporidiosis 1999 Washington (DC) ASM Press 447 501
Southwood TR Murdie G Yasuno M Tonn RJ Reader PM Studies on the life budget of Aedes aegypti in Wat Samphaya, Thailand Bull World Health Organ 1972 46 211 226 4537483
Bedhomme S Agnew P Sidobre C Michalakis Y Sex-specific reaction norms to intraspecific larval competition in the mosquito Aedes aegypti
J Evol Biol 2003 16 721 730 14632235
Agnew P Hide M Sidobre C Michalakis Y A minimalist approach to the effects of density-dependent competition on insect life-history traits Ecol Entomol 2002 27 396 402
Prout T McChesney F Competition among immatures affects their adult fertility: Population dynamics Am Nat 1985 126 521 558
Renshaw M Service MW Birley MH Density-dependent regulation of Aedes cantans (Diptera: Culicidae) in natural and artificial populations Ecol Entomol 1993 18 223 233
Borash DJ Gibbs AG Mueller LD A genetic polymorphism maintained by natural selection in a temporally varying environment Am Nat 1998 151 148 156 18811414
Bedhomme S Agnew P Sidobre C Michalakis Y Pollution by conspecifics as a component of intraspecific competition among Aedes aegypti larvae Ecol Entomol 2005 30 1 7
Bedhomme S Agnew P Sidobre C Michalakis Y Virulence reaction norms across a food gradient Proc R Soc Lond B Biol Sci 2004 271 739 744
Wynn G Paradise CJ Effects of microcosm scaling and food resources on growth and survival of larval Culex pipiens
BMC Ecol 2001 1 3 11527507
| 16008503 | PMC1175819 | CC BY | 2021-01-05 08:28:15 | no | PLoS Biol. 2005 Aug 19; 3(8):e262 | utf-8 | PLoS Biol | 2,005 | 10.1371/journal.pbio.0030262 | oa_comm |
==== Front
PLoS BiolPLoS BiolpbioplosbiolPLoS Biology1544-91731545-7885Public Library of Science San Francisco, USA 1601872010.1371/journal.pbio.0030265Research ArticleNeuroscienceHomo (Human)Primary Visual Cortex Activity along the Apparent-Motion Trace Reflects Illusory Perception V1 Activity Reflects Illusory PerceptionMuckli Lars [email protected]
1
2
Kohler Axel
1
2
Kriegeskorte Nikolaus
1
¤Singer Wolf
1
2
1Department of Neurophysiology, Max Planck Institute for Brain Research, Frankfurt am Main, Germany,2Brain Imaging Center Frankfurt, Frankfurt am Main, Germany,Zatorre Robert Academic EditorMcGill UniversityCanada8 2005 19 7 2005 19 7 2005 3 8 e26529 9 2004 31 5 2005 Copyright: © 2005 Muckli 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.
Now You Don't See It, Now You Do: Filling In Creates the Illusion of Motion
The illusion of apparent motion can be induced when visual stimuli are successively presented at different locations. It has been shown in previous studies that motion-sensitive regions in extrastriate cortex are relevant for the processing of apparent motion, but it is unclear whether primary visual cortex (V1) is also involved in the representation of the illusory motion path. We investigated, in human subjects, apparent-motion-related activity in patches of V1 representing locations along the path of illusory stimulus motion using functional magnetic resonance imaging. Here we show that apparent motion caused a blood-oxygenation-level-dependent response along the V1 representations of the apparent-motion path, including regions that were not directly activated by the apparent-motion-inducing stimuli. This response was unaltered when participants had to perform an attention-demanding task that diverted their attention away from the stimulus. With a bistable motion quartet, we confirmed that the activity was related to the conscious perception of movement. Our data suggest that V1 is part of the network that represents the illusory path of apparent motion. The activation in V1 can be explained either by lateral interactions within V1 or by feedback mechanisms from higher visual areas, especially the motion-sensitive human MT/V5 complex.
Using fMRI in humans, the authors reveal a clear role for V1 cortex in forming an illusory perception of motion when stationary stimuli are successively flashed in different locations.
==== Body
Introduction
Apparent motion can be perceived when two spatially segregated visual stimuli are presented in succession [1]. The illusion persists even when the stimuli are widely separated, a phenomenon called “long-range apparent motion” (here we use the term “apparent motion” to refer to long-range apparent motion) [2]. In order to respond to apparent motion, neurons have to integrate information over a large part of visual space, spanning at least the distance between the two inducing stimuli. Neurons in the middle temporal (MT) area of the macaque have pronounced directional selectivity and receptive-field sizes of up to a 25° visual angle [3–6], which makes them ideally suited for the integration of apparent-motion-inducing stimuli. Several studies have shown that the middle temporal area in the macaque and other primates and its human homolog, the human MT/V5 complex (hMT/V5+), respond to stimulus conditions that induce apparent motion [7–9].
In contrast, receptive-field sizes in early visual areas and, in particular, the primary visual cortex (visual cortex area 1 [V1]) are too small to account for long-range interactions between stimuli [10–13]. The fact that one actually observes spatially resolved movement between the inducing stimuli in apparent motion suggests that there could be a filling-in process in early visual areas that is driven by feedback from extrastriate regions with larger receptive-field sizes. In particular, back-projections from hMT/V5+ have been shown to be relevant for perception of motion and apparent motion [14–17]. A psychophysical study of visual interference along the illusory path of apparent motion has indeed suggested feedback processes acting at the level of V1 as a possible mechanism [18]. Liu et al. [19] have used functional magnetic resonance imaging (fMRI) to test whether the perceptual “filling-in” has an early or late cortical locus. Using a visual display comprising two concentric rings, Liu and colleagues found a late cortical locus (hMT/V5+ and lateral occipital complex) but no activity in the path of apparent motion in early visual areas [19].
We used a different visual display to test the same hypotheses, and we report here about V1 activity along the illusory motion path. In order to investigate the topographic pattern of apparent-motion-related activity in V1, we used fMRI in humans to map the retinotopic representation of apparent motion. We specifically looked for activity in subregions that were not directly activated by the apparent-motion-inducing stimuli. Our results show that there is activity in V1 sites representing the illusory motion path that cannot be explained by the local characteristics of the stimulus alone. The influence of attention was controlled with a demanding center task, in which subjects had to detect numbers in a rapidly changing stream of letters and numbers. Using a bistable motion-quartet stimulus [20,21], we confirmed that the activity in V1 changes as a function of the subjectively perceived motion trace. Considering the separation between the apparent-motion stimuli, we argue that this activity is most likely caused by feedback projections from extrastriate regions with larger and direction-selective receptive fields.
Results
Experiment 1
We determined the extent and borders of early visual areas V1–V3 (Figure S1) in a retinotopic-mapping experiment [8]. In a second session, subjects were presented with five different stimulus conditions (Figure 1). The apparent-motion stimulus (Figure 1A) consisted of two white squares blinking in alternation, presented on a dark screen to the right side of a fixation cross. For mapping of the apparent-motion path, a real-motion stimulus was generated that consisted of a white square moving along the perceived trajectory of the apparent-motion stimulus (Figure 1B). The remaining three conditions consisted of contrast-inverting checkerboards presented at three different locations to map the cortical representation of the apparent-motion-inducing stimuli (upper and lower site, Figure 1D and 1F) and one location on the path of apparent motion (middle, Figure 1E).
Figure 1 Stimuli Used in Experiments 1 to 3
(A) The apparent-motion stimulus consisted of two white squares (size, 1.8°) blinking in alternation (2.3 Hz). The squares were presented at an eccentricity of 9.5° visual angle (distance between squares, 16.5°).
(B) The real-motion stimulus consisted of a white square moving in harmonious oscillation (average speed, 66.5°s−1) on the perceived path of apparent motion. The end points of the movement were adjacent to the position of the apparent-motion squares (maximum eccentricity, 8.0°).
(C) The flicker control stimulus from experiment 2 was composed of two white squares (1.8°) that were blinking simultaneously at 4.6 Hz.
(D–F) For the mapping conditions, high-contrast inverting checkerboards were presented in the upper (D), middle (E), and lower (F) right visual field. Sizes were adjusted according to the cortical magnification factor in V1 to produce activated regions of a similar spatial extent (3.6° for the upper and lower stimuli presented at 9.5° eccentricity, and 1.8° for the middle stimulus presented at 4.7° eccentricity).
(G) For the five conditions of experiment 1 (A, B, D–F), subjects saw a white cross at the middle of the screen, which they had to fixate (G, upper part). In experiment 2, a randomly generated stream of letters and digits was presented at 2 Hz in the middle of the screen (G, lower part). Subjects had to fixate the character stream and either passively view the stream or perform a digit-detection task.
(H) The motion quartet (experiment 3) was composed of four white squares presented at eccentricities similar to the apparent-motion stimulus. Two squares from diagonally opposite corners were presented at the same time. The motion quartet can be seen in vertical (left part) or horizontal (right part) motion without any changes in the physical characteristics of the stimulus. Subjects had to fixate a white cross in the middle of the screen and had to report the perceived direction of movement (vertical or horizontal).
To demonstrate specific effects of apparent motion, data were analyzed in experiment 1 using two different approaches. First, we described the spatial distribution of apparent-motion-driven activity in V1 using simple activation maps (apparent motion > baseline) Second, in an explorative manner, we searched for regions of interest (ROIs) within these activity maps that were not activated in any way by the inducing stimuli. It is clearly expected that at the beginning and at the end of the motion streak, locations are also activated by the upper or lower blinking square. Therefore, activity patterns from the apparent-motion map (see above) could be influenced by the effects of the inducing stimuli. The second strategy provides proof of principle that there is activity within the streak that is not activated in any way by the inducing squares.
Following the first strategy (green map in Figures 2 and S1), we found that the activity evoked during apparent motion spanned the region of V1 representing the illusory motion path between the two apparent-motion-inducing stimuli (t > 2, p < 0.05 for each single-subject analysis). This streak of activity was always located on the real-motion path and, in four out of five subjects, was also activated by the middle mapping stimulus (Figure S1). Moreover, the activation maps for real motion and for apparent motion covered additional parts of V1 peripheral to the direct connection between the end points (Figure 2B and 2E). Following the second strategy, we found in four subjects ROIs within V1 between the mapped end points of the motion path that were activated in the apparent-motion condition but did not respond at all when the upper or lower mapping stimulus was presented alone (Figures 2 and S1).
Figure 2 Activation Pattern in V1 for Experiment 1
Medio-posterior view on the inflated left occipital cortex of subject HP (A and B) and subject AK (C). All five subjects are shown in Figure S1. Gray-scale coloring of cortex indicates the extent of retinotopic visual areas (light gray: V1 and V3/VP; dark gray: V2, V4, and V3A) and the gyral pattern (concave surfaces are indicated in darker gray in the respective areas).
(A) The cortical representations of the mapping stimuli are marked in color. The maps were obtained by calculating a balanced contrast between the respective mapping condition of interest and the two other mapping conditions (e.g., middle versus upper and lower). The thresholds of t-maps were individually adjusted to obtain patches of comparable size in V1: upper (yellow), t(600) > 9.3; middle (orange), t(600) > 11.9; and lower (red), t(600) > 15.1 (all p << 0.001).
(B) Cortical activation maps for the apparent-motion (green) and real-motion (blue) condition compared to baseline (apparent motion in V1: t[600] > 3.6, p < 0.0004; real motion in V1: t[600] > 18.6, p << 0.001). Cortical representations of the mapping stimuli are indicated by colored outlines taken from (A). The solid white line indicates a patch of significant activation (t[600] > 3, p < 0.003) for the following conjunction of contrasts, which represents the ideal activation pattern: (middle > upper) and (middle > lower) and (apparent motion > upper) and (apparent motion > lower). The dashed white line marks a peripheral region in V1 with significant activation in response to apparent motion and real motion.
(C) Same as (B) for subject AK. The white region with the optimal response pattern overlaps largely with the representation of the middle stimulus (in orange; see Figure S1 for details).
(D) Event-related BOLD signal change for subject HP plotted over time from the patch outlined by the solid white line in (B) (solid black line indicates stimulus onset, dotted line stimulus offset).
(E) Event-related BOLD signal change of the apparent-motion-activated (green) region from the eccentric parts of V1 outlined by the dashed white line in (B) (solid black line indicates stimulus onset, dotted line stimulus offset). This eccentric region responds to real motion and apparent motion but not to the middle stimulus.
(F) Same as (D) for subject AK.
Error bars correspond to standard errors of the mean.
Experiment 2
In the second experiment, we wanted to replicate our findings and to control for possible alternative explanations for the apparent-motion-related activity. We added a task that required the subjects to divert their attention to the center of the visual field. We presented a stream of rapidly changing digits and letters instead of the fixation cross and instructed the subjects to respond to the appearance of the digits (see Figure 1G). Separate fMRI scans were acquired while the subjects either performed the attention task or passively viewed the character stream. At the same time, six different conditions were presented in the periphery: five conditions from experiment 1 (real motion, apparent motion, upper, middle, and lower) and an additional control condition, in which two squares were blinking simultaneously at the upper and lower positions (flicker, see Figure 1C).
As in experiment 1, we first described the spatial distribution of apparent-motion-driven activity in V1 using contrast maps (apparent motion > flicker) and searched for ROIs within these activity maps that were not activated in any way by the inducing stimuli. In addition to the two analysis strategies applied in experiment 1, we used a further two approaches that were more objective. As a third strategy, we selected ROIs in individual subjects from the representation of the apparent-motion path by use of the middle mapping condition (conjunction map: middle > upper and middle > lower). In the fourth approach, ROIs were selected by contrasting the real-motion condition with the outer mapping conditions. These ROI-based approaches are comparable to the one followed in the study by Liu et al. [19].
Using strategies 1 and 2, we replicated the previous findings (n = 5, three of whom had already participated in experiment 1). Again, we found apparent-motion-related activity between the end points (green map in Figures 3 and S2) in patches that were overlapping with regions activated by real motion (blue map) and the middle mapping stimulus. Moreover, the activation maps for real motion and for apparent motion covered additional parts of V1 peripheral to the direct connection between the end points. The ROI time courses show that activated regions in the middle of the motion streak responded exclusively to middle mapping, real motion, and apparent motion but not to upper or lower mapping stimulation. The activation of these not directly stimulated areas remained significant in seven out of ten single comparisons when we compared the apparent-motion-related activity to the activity caused by the flicker control condition (Figure S2). In this condition, the end points were stimulated in the same way as during apparent-motion induction, except that the squares were presented simultaneously and did not produce apparent motion (contrast apparent motion versus flicker; t[1445] > 2, p < 0.05 in the significant single-subject analyses).
Figure 3 Cortical Activation in V1 for Experiment 2
(A) Left occipital cortex of subject HP (all five subjects are shown in Figure S2) with superimposed contrast maps indicating the cortical representation of the stimulus positions: upper (yellow), t(1445) > 15; middle (orange), t(1445) > 12; and lower (red), t(1445) > 9.3 (all p << 0.001). For comparison, the patches from experiment 1 (see Figure 1) are marked with dotted lines.
(B) Cortical activation maps for apparent motion (green) and real motion (blue). In this case, apparent motion is contrasted with the flicker control condition (apparent motion > flicker in V1: t[1445] > 3.6, p < 0.0004); real motion is compared to fixation baseline (real motion > baseline in V1: t[1445] > 18.6, p << 0.001). The white line indicates an example of a significant (t > 2.2, p < 0.05) contrast conjunction: (middle > upper) and (middle > lower) and (apparent motion > flicker).
(C) BOLD activity profile of the region indicated by the white line in (B). Bars indicate average activity during the respective conditions expressed in beta weights from a GLM analysis.
Following strategies 3 and 4, we computed event-related averages of V1 activity for different ROIs (Figure 4). The four ROIs were defined separately in individual subjects as parts of V1 that showed a significant response to one of the three mapping conditions (upper, lower, and middle) or to the real-motion condition. The peak response relative to baseline was then compared for the different conditions and t-tests were computed across subjects. The ROIs for the upper and lower condition showed the expected response profile. Activation was strong for the respective mapping stimuli as well as for flicker and apparent motion, which were presented at the same location as the mapping conditions. The regions also showed a substantial response to real motion, probably because of larger receptive-field sizes in the periphery and because the real-motion square overlapped with the outer mapping stimuli. There was no significant difference in the responses to the apparent-motion and flicker conditions (p > 0.10), suggesting that the two conditions were equivalent in terms of the local activity produced by the inducing stimuli. For the two ROIs on the path of apparent motion, middle mapping, and real motion, the response profile looked different: For both regions, real motion produced the strongest activation, followed by the middle condition. In addition, there was a reliable response to apparent motion that was higher than the activity induced by the flicker squares. This effect decreased when the subjects' attention was focused on the letters–digits discrimination task (Figures 4 and S2) but was still significant (p < 0.05). Notably, the decrease was not due to a reduction of apparent-motion-related activity but to an increased activity during the flicker control condition.
Figure 4 Response Profile for Different ROIs in V1 for Experiment 2
The four panels correspond to four ROIs that were defined in V1 by the respective conditions. To determine the ROIs for the three mapping conditions—upper, middle, and lower—the relevant condition was contrasted with the remaining two mapping conditions in a conjunction analysis (e.g., upper > lower and upper > middle). For the real-motion ROI, the real-motion condition was contrasted with the upper and lower mapping conditions (conjunction analysis). The bars designate percentage signal change of the peak response relative to baseline for the different conditions (see Materials and Methods). The hatched bars are activations from runs with passive viewing only, and the solid bars are from runs with center task. For both the real-motion and middle ROI, the response to the apparent-motion stimulus is significantly larger than the response to the flicker squares. This is not the case for the upper and lower ROIs. Error bars correspond to standard errors of the mean.
It has been demonstrated in previous fMRI studies that spatial attention strongly modulates activation in visual cortical areas down to V1 [22–27]. To validate the effectiveness of our attention-demanding center task, we compared the activation in hMT/V5+ for the different conditions in runs with center task to the activation found with passive viewing (Figure 5). In addition, we looked at the attentional modulation in V1 at the upper and lower locations (see Figure 4). In hMT/V5+, the general activation level was highest for real motion and apparent motion followed by flicker and the mapping stimuli (see Figure 5). For both task conditions, passive-viewing and center task, hMT/V5+ was activated significantly more strongly for apparent motion than for flicker (t[7247] > 2.7, p < 0.01), in correspondence with the apparent-motion sensitivity of hMT/V5+ [8,9].
Figure 5 Attentional Modulation of Activity in hMT/V5+ in Experiment 2
Responses to the different conditions in experiment 2 for the region hMT/V5+ are shown. The bars designate percentage signal change of the peak response relative to baseline (see Materials and Methods). The hatched bars are activations from runs with passive viewing only, and the solid bars are from runs with center task. The static conditions—upper, middle, lower, and flicker—induced only low activation in hMT/V5+. In contrast, the response to apparent motion and real motion was higher, corresponding with the motion sensitivity of hMT/V5+. The attentional modulation of the responses was very strong. Under all conditions, the activity was smaller during runs with center task than during runs with passive viewing. Error bars correspond to standard errors of the mean.
We computed an attentional-modulation index (AMI) [26] for the three main conditions of interest: real motion, apparent motion, and flicker. The AMI equals the difference between the responses of the passive-viewing and center task runs normalized by the response of the passive-viewing runs ([passive viewing − center task]/passive viewing). A positive AMI value indicates a reduction in activation for the center task runs. In hMT/V5+, the AMI was 0.57 for flicker, 0.48 for apparent motion, and 0.41 for real motion. The very high AMI values show that the subjects' attention was efficiently diverted from the peripherally presented stimuli. In contrast to the results in hMT/V5+, the pattern of attentional modulation was very different in the V1 ROIs (see Figure 4). Only real motion showed a positive AMI, with 0.20 for the lower ROI and 0.34 for the upper ROI. The other relevant conditions showed an enhanced response with the center task; their AMI values ranged from −0.55 to −0.08.
Experiment 3
In a third experiment, we presented a bistable apparent-motion stimulus that consisted of four blinking squares (motion quartet; see Figure 1H). The motion quartet induces spontaneous switches between the perceptions of vertical or horizontal apparent motion without any changes in the physical characteristics of the stimulus [20,21]. This allows us to identify activity that is closely related to the conscious perception of apparent motion [9]. We presented a motion quartet with the two squares in the right hemi-field at approximately the same locations as the apparent-motion stimuli in experiments 1 and 2. Subjects (n = 6; five subjects from experiment 2) continuously reported their current percept using the left and right response buttons for vertical and horizontal movement, respectively.
For experiment 3, we calculated contrast maps that indicate higher activation in response to vertical as compared to horizontal apparent-motion perception. We found patches of activity that showed a selective increase of activity following perceptual switches from horizontal apparent motion to vertical apparent motion in all six subjects (green regions in Figures 6 and S3). These patches were located within V1 between the cortical representations of the inducing stimuli (red regions in Figures 6 and S3) consistent with the cortical representation of the vertical motion streak.
Figure 6 Results from the Bistable Motion Quartet (Experiment 3)
(A and B) Medio-posterior view on the inflated left occipital cortex of subject HP (A) and AK (B) (all six subjects are shown in Figure S3). Gray shading indicates the extension of V1 (light gray for V1) and the cortex curvature (dark gray, concave; light gray, convex). Activation maps show the cortical representation of the stimulated locations in red (motion quartet > baseline). Contrast maps in green indicate regions that are more active for vertical apparent motion than for horizontal apparent motion (ROI-based GLM at indicated locations; HP: t[301] > 2.5, p < 0.02; AK: t[708] > 2, p < 0.05). The dotted line is a spline-interpolated curve connecting the stimulated locations and the region that is more active during the perception of vertical apparent motion. (This line does not necessarily indicate the path of apparent motion from experiments 1 and 2 since stimulus parameters had to be adjusted.)
(C and D) The solid white lines mark regions from which event-related averages were calculated. Event-related averages are shown for subject HP (C) and for subject AK (D). The time courses are aligned to the time point at which the subject indicates a switch in perception (t = 0, black line) and are shown for the time period from 4 s before to 24 s after the perceptual switch. The first perceptual period following each stimulation onset is omitted from the analysis (see Materials and Methods). Error bars correspond to standard errors of the mean. For HP, the vertical percept lasted on average for 11.9 s (standard deviation, 10.3 s), the horizontal for 13.4 s (standard deviation, 9.2 s). For AK, the vertical lasted for 14.6 s (standard deviation, 7.1 s), and the horizontal for 16.6 s (standard deviation, 6.7 s).
Discussion
We have shown in three experiments that there is V1 activity that correlates with the perception of long-range apparent motion. We observed activity patterns that corresponded to the expected motion streak. In ROI analyses, the apparent-motion activity was found to be significantly higher than in the flicker control condition and was also unmodulated by attentional set. In an event-related design with the bistable motion quartet, the activity could be closely linked to the conscious perception of apparent motion in response to perceptual switches.
Horizontal Connections
Our results indicate that there are subregions in V1 that are active during illusory apparent-motion perception even though the inducing stimuli are presented outside the receptive fields of the neurons in these regions. This spreading extra-receptive-field response could be mediated by horizontal connections in V1 [13,28]. However, three arguments render it unlikely that horizontal interactions within V1 are a major source of activity along the illusory motion path. First, the flicker stimuli did not induce a comparable extra-receptive-field response although they should have caused the same local activation as the apparent-motion-inducing stimuli. Second, the horizontal connections in V1 span only short distances that are well below the distance between the representations of the apparent-motion-inducing stimuli [13]. Third, in a motion quartet there is perceptual competition between vertical and horizontal apparent motion, leading to mutually exclusive percepts. There is evidence that this competition is resolved in hMT/V5+ [29].
Attention and Feedback
A more likely candidate for the apparent-motion-related activity is top-down influence from higher areas [30]. Attention increases the blood-oxygenation-level-dependent (BOLD) signal in early visual areas including V1 [23–25,27,31,32], even in the absence of visual stimulation [26]. However, data from our control experiment show that the apparent-motion-related activation persisted when we diverted the subjects' attention away from the stimuli. Thus, apparent-motion-related activation is not solely due to spatial attention. This suggests an attention-independent but motion-specific filling-in process associated with the illusory motion percept.
The best candidate area for a possible top-down influence on V1 related to processing of apparent motion is hMT/V5+. Several studies have emphasized the important influence that feedback from higher areas has on functions of V1 [30,33–36], and feedback from hMT/V5+ to V1 has been assigned a role in the perception of real motion and apparent motion [14–18]. Neurons in hMT/V5+ have receptive fields large enough to span the distance between the apparent-motion-inducing stimuli, and they respond to apparent motion and real motion in similar ways [8,9,37]. The back-projections from higher to lower cortical areas fan out and can span at least the size of their receptive fields [13,38,39].
We observed illusion-related activity and real-motion-related activity not only on the direct motion path but also in the periphery of the motion streak. Coactivation of more peripheral sections might be a result of feedback activity that is spreading out to larger sections in the periphery. It is particularly those cells that have sufficiently large receptive fields (16.5°) to cover both apparent-motion-inducing stimuli (illusion-related activity and real-motion-related activity) that are expected to be found in more peripheral sections of higher visual areas and are therefore expected to back-project, especially to peripheral parts in early visual areas.
Previous and Related Findings
Why was the earlier attempt by Liu et al. [19] unsuccessful in finding apparent-motion-related activity in V1? The experimental strategy that they employed was comparable to ours in most aspects, but there was a significant difference in the stimulation material that they used: large rings inducing radial inward–outward apparent motion. The stimulation of large sections of the visual cortex might have induced a complex pattern of excitation and inhibition in directly adjacent parts of V1. Moreover, we showed that in most subjects apparent-motion-related activity is displaced to the periphery. So in the case of inward–outward apparent motion, much of the apparent-motion-related activity might be displaced towards the outer ring.
A number of recent studies have demonstrated a close relationship between activation in V1 and conscious perception of visual stimuli. In these studies, different imaging and electrophysiological methods have been used to investigate the functional properties of V1 during binocular rivalry [40–43], perception of moving phosphenes [15,17], figure–ground segregation [34,44–46], the “line-motion” illusion [28], and color filling-in [47]. In contrast to previous results [48], it was found that V1 activity can be tightly linked to visual awareness. This has been supported in other experiments showing that complex features, such as motion-defined edges [49] and second-order motion [50,51], are represented as early as in V1. The new view on early visual processing has been applied to other sensory systems. Chen et al. [52] demonstrated that correlates of a tactile illusion could already be found in primary somatosensory cortex. In our study, we extended the previous findings to the domain of apparent motion.
Current findings from simultaneous fMRI and electrophysiological recordings in monkeys have suggested that the BOLD signal might be especially sensitive to changes in local field potentials [53]. A more recent study established a tight link between BOLD signal and neuronal synchronization [54]. This could explain why the spiking activity of neurons located along the illusory motion trace has been found unchanged in earlier electrophysiological studies of V1 in macaques [4]. Logothetis and colleagues showed that the BOLD signal is correlated with an increase in neuronal activity but further suggested that it primarily reflects the input of a cortical area and its local intra-cortical processing, including excitatory and inhibitory inter-neurons [55]. Our data are consistent with the interpretation that at the level of hMT/V5+, motion features are extracted from neurons with sufficiently large receptive fields to cover both apparent-motion-inducing stimuli. Feedback from hMT/V5+ could cause synaptic processes at the level of V1 that produce a BOLD response without causing major increases in spiking activity. Whether and how interactions between hMT/V5+ and V1 contribute to the perception of apparent motion remains to be investigated in future studies.
Materials and Methods
Subjects
Eight subjects (six male) participated in the fMRI experiments; their mean age was 29.2 y (range, 22–34 y). All subjects had normal or corrected-to-normal (one subject) vision. The participants received information on fMRI and a questionnaire to check for potential health risks and contraindications. Volunteers gave their informed consent after being introduced to the procedure in accordance with the Helsinki declaration (www.wma.net/e/ethicsunit/helsinki.htm). All subjects participated in retinotopic mapping. Five subjects participated in experiment 1, five subjects in experiment 2, and six subjects in experiment 3. Of the subjects who participated in experiment 1, three also participated in experiments 2 and 3. Of the subjects who participated in experiment 2, five also participated in experiment 3.
Stimuli
Stimuli were generated with custom-made software based on the Microsoft DirectX library (StimulDX, Brain Innovation, Maastricht, The Netherlands). The stimuli were back-projected onto a frosted screen with a liquid-crystal-display projector (VPL PX 20, Sony, Tokyo, Japan) and a custom-made lens. Subjects viewed the screen through a mirror. Mirror and projection screen were fixed onto the head coil. Seven different types of stimuli were used in the experiments (see Figure 1; stimulus in Figure 1C was used only in experiment 2 and stimulus in Figure 1H only in experiment 3). The apparent-motion stimulus (see Figure 1A) consisted of two white blinking squares (size, 1.8° visual angle) presented in alternation on a dark screen to the right side of the fixation cross. The squares were presented at an eccentricity of 9.5° visual angle (distance between squares, 16.5°). Stimulus duration was 150 ms with an inter-stimulus interval of 67 ms, corresponding to a presentation frequency of 2.3 Hz. This specific frequency was chosen based on previous results showing that the perception of apparent motion is strongest in an envelope from 2 to 3 Hz [56,57]. A real-motion stimulus (see Figure 1B) was generated by presenting a white square in harmonious oscillation (location on the path equals the sine of time) on the perceived path of the apparent-motion stimulus (size, 1.8° visual angle).
The end points of the oscillation were directly adjacent to the position of the apparent-motion-inducing squares. The maximum eccentricity of the squares was 8.0° and the distance between the end points of the motion path 13.3° (average speed, 66.5°s−1). Two static stimuli were used to map the locations where the apparent-motion squares were presented (upper and lower checkerboard in Figure 1D and 1F) and one static stimulus to map the center region of the apparent-motion path (middle checkerboard in Figure 1E). The middle stimulus was located halfway between the two squares of the apparent-motion stimulus at an eccentricity of 5.0° (size, 1.8°). The static stimuli were checkerboards consisting of a 4 × 4 matrix of alternating black and white squares. The checkerboards inverted their contrast every 100 ms, corresponding to a presentation frequency of 5 Hz. The size of the squares was 3.6° (upper and lower) and 1.8° (middle).
In experiment 2, we used an additional control stimulus (see Figure 1C) that was identical to the apparent-motion stimulus except that the two squares were presented in synchrony, resulting in a presentation frequency of 4.6 Hz. To control for attention effects, we also introduced a demanding center task [58]. Subjects saw a stream of alphanumeric characters instead of the fixation cross and had to press a button whenever they detected a numeric character (see Figure 1G). The presentation frequency of the characters was 2 Hz and targets appeared with a probability of 0.125.
In experiment 3, we presented four blinking squares in a rectangular configuration in which the diagonally opposing dots blinked simultaneously (see Figure 1H). We presented two dots at approximately the same positions in the right visual field as in experiments 1 and 2. Two further dots were presented in the left visual field and were therefore not processed by ipsilateral V1 [59]. We adjusted for each subject the vertical distance of the squares in order to make the stimulus bistable.
Procedure
Each of the conditions in experiment 1 (five conditions) and experiment 2 (six conditions) was presented for approximately 12.5 s (six volumes) in a block design, separated from the next block by a fixation period of the same length. A complete presentation cycle consisted of the following sequence of conditions (with intervening fixations): middle, apparent motion, upper, real motion, and lower (60 volumes) in experiment 1; and middle, apparent motion, upper, real motion, lower, and flicker (72 volumes) in experiment 2. In each following cycle, the presentation order of the conditions was reversed with respect to the preceding cycle, providing a control for serial-order effects. A complete scan comprised five cycles of experimental conditions plus an additional eight volumes of fixation at the beginning of the scan (308 volumes for experiment 1 and 368 volumes for experiment 2). In experiment 1, two complete scans were acquired. Subjects were asked to fixate and be attentive during the scans. In experiment 2, four scans were carried out with each subject. In two scans of experiment 2, the subjects had to perform the attention task, while in the other two scans they only had to fixate the alphanumeric stream of the attention task. In the two runs of experiment 3, we presented the bistable motion quartet five times for 125 s (60 volumes each). The five presentation periods were separated by 21 s (ten volumes) of fixation. During the motion-quartet trials, subjects had to indicate their current percept (vertical versus horizontal motion) by continuously pressing one of two buttons with their right index and middle fingers. They were instructed to apply a strict criterion for changes of perception.
Imaging
fMRI scanning was performed on a 1.5 Tesla Siemens Magnetom Vision scanner (Siemens, Erlangen, Germany) at the University Clinic in Frankfurt am Main. A gradient-recalled echo-planar-imaging sequence was used with the following parameters: 16 or 18 slices, oriented approximately in parallel to the anterior commissure–posterior commissure plane; recording time for one volume (TR), 2,081 ms; TE, 60 ms; FA, 90°; FOV, 210 mm; in-plane resolution, 3.44 × 3.44 mm; slice thickness, 4 mm; and gap thickness, 0.4 mm. In addition, a T1-weighted anatomical scan was acquired for all subjects using a Siemens fast low-angle-shot (FLASH) sequence (isotropic voxel size, 1 mm3).
Data analysis
Data were analyzed using the BrainVoyager 4.9 software package (Brain Innovation). The first four volumes of each scan were discarded to preclude T1 saturation effects. Preprocessing of the functional data included the following steps: (i) three-dimensional motion correction using the Levenberg-Marquardt algorithm, (ii) linear-trend removal and temporal high-pass filtering at 0.01 Hz, and (iii) slice-scan-time correction with sinc interpolation.
The statistical analysis was performed with multiple linear regression. For every voxel, the time course was regressed on a set of dummy-coded predictors representing the experimental conditions. To account for the shape and delay of the hemodynamic response [60], the predictor time courses (box-car functions) were convolved with a gamma-variate function.
With the data from experiment 2, five ROIs were defined in individual subjects for a more detailed analysis of activation patterns in V1 and hMT/V5+. The three mapping conditions—upper, middle, and lower—and the real-motion condition were used to determine four ROIs in V1. We used conjunction analyses to effectively restrict the ROIs to parts of V1 with a specific functional profile. For example, to find regions responding specifically to the lower mapping, the activation produced by the lower stimulus had to be significantly higher than the response to the middle as well as the upper stimulus. For the real-motion ROI in V1, the real-motion condition was contrasted with the upper and lower condition. The threshold was set to a Bonferroni-corrected p-value of 0.1. The hMT/V5+ ROI was defined by a balanced contrast between real motion and the three mapping conditions (upper, middle, and lower). For four subjects, p-values were set to less than 0.2 (Bonferroni-corrected). In one subject only, a more liberal significance level of p < 0.002 (uncorrected) had to be used.
To ensure that the increases in event-related activity were not influenced by the preceding conditions, we inspected the event-related time-course averages (see Figures 2 and S1), which are aligned to an average baseline of 0% signal change for the period of 4 s before stimulation onset. The activation patterns for the ROIs were visualized by plotting the percentage signal change of the peak response (corresponding to the average of the three time points around 8 s, 10 s, and 12 s) for the different conditions. The measurements with and without center task were analyzed separately. Statistical significance was assessed across subjects to validate the reliability of the effects in our group of participants.
For experiment 3, we mapped the cortical representation of our inducing stimuli by a general linear model (GLM) analysis in which we looked for higher activity during motion-quartet stimulation as compared to baseline (mapping of all stimulus locations). In addition, we identified regions with higher activation for perceived vertical motion than for perceived horizontal motion and baseline (conjunction analysis). From these regions, we extracted event-related time courses to visualize the activation changes due to the perceptual switches form horizontal to vertical motion. To avoid unspecific stimulus-onset effects, the first perceptual phase of each stimulation block was excluded from the analysis.
Retinotopic mapping
Phase-encoded retinotopic mapping was assessed in each subject and included mapping of eccentricity and polar angle [8,61–63]. In the eccentricity-mapping experiment, black and white checkerboard patterns were presented in a ring-shaped configuration and were flickered at a rate of 4 Hz. The ring started with a radius of 1° and slowly expanded to a radius of 12° within 96 s. In the polar-angle mapping experiment, the checkerboard pattern consisted of a ray-shaped disk segment subtending 22.5° of polar angle. The ray started at the right horizontal meridian and slowly rotated clockwise for a full cycle of 360° within 96 s. Each mapping experiment consisted of seven repetitions of a full expansion or ten repetitions of rotation, respectively, with each cycle lasting for 64 s.
The analysis of the retinotopic-mapping experiment was conducted by the use of a cross-correlation analysis. We used the predicted hemodynamic signal time course for the first 1/8 of a stimulation cycle (corresponding to a 45° visual angle in the polar mapping experiment) and shifted this reference function successively in time (time steps correspond to the recording time for one volume) [63]. Sites activated at particular eccentricities and polar angles were identified through selecting the lag value that resulted in the highest cross-correlation value for a particular voxel. The obtained lag values at particular voxels were encoded in pseudo-color on corresponding surface patches (triangles) of the reconstructed cortical sheet. Based on the polar-angle mapping experiment, the boundaries of retinotopic cortical areas V1, V2, V3, VP, V3A, and V4v were estimated manually on the inflated cortical surface and colored in shades of light and dark gray.
Cortical-surface reconstruction
The recorded high-resolution T1-weighted three-dimensional recordings were used for surface reconstruction of both cortical hemispheres of each subject [64]. The white/gray-matter border was segmented with a region-growing method preceded by inhomogeneity correction of signal intensity across space. The borders of the two resulting segmented subvolumes were tessellated to produce a surface reconstruction of the left hemisphere. The resulting surface was used as the reference mesh for projecting functional data on inflated representations. A morphed surface always possesses a link to the folded reference mesh so that functional data can be shown at the correct location on folded as well as inflated representations. This link was also used to keep geometric distortions to a minimum during inflation and flattening through inclusion of a morphing force that keeps the distances between vertices and the area of each triangle of the morphed surface as close as possible to the respective values of the folded reference mesh.
Supporting Information
Figure S1 Occipital BOLD Activation of Five Subjects in Experiment 1
Color coding for subjects HP, MN, AK, DL, and SW for (B) to (D) is identical to the coding of Figure 2. In addition, maps from a retinotopic-mapping experiment are provided in (A). Gray-scale coloring of cortex indicates the extent of retinotopic visual areas (light gray: V1 and V3/VP; dark gray: V2, V4, and V3A) and the gyral pattern. Borders between visual regions are marked with black lines (A–C).
(A) Retinotopic mapping of early visual areas. Phase-encoded activation maps show regions responding to stimuli in the lower-right quadrant (green to blue) and upper-right quadrant (red to yellow) of the visual field. Red color indicates the representation of the upper vertical meridian, blue and yellow the horizontal meridian, and green the lower vertical meridian. The overlaid outlines (red, orange, and yellow) are from (B) and indicate the cortical representation of the mapping stimuli used in experiment 1.
(B) The cortical representations of the mapping stimuli are marked in color. The maps were obtained by calculating a balanced contrast between the respective mapping condition of interest and the two other mapping conditions (e.g., middle versus upper and lower). The thresholds of t-maps were individually adjusted to obtain patches of comparable size in V1: upper (yellow), t(600) > 3.8, p < 0.0002 (HP > 9.3; MN > 4.0; AK > 4.1; DL > 5.8; SW > 3.8); middle (orange), t(600) > 3.2, p < 0.002 (HP > 11.9; MN > 5.8; AK > 3.2; DL > 5.6; SW > 6.2); and lower (red), t(600) > 4.3, p < 0.00003 (HP > 15.1; MN > 8.1; AK > 4.6; DL > 5.6; SW > 4.3).
(C) Cortical activation maps for the apparent-motion (green) and real-motion (blue) conditions compared to baseline—apparent motion in V1: t(600) > 2.2, p < 0.03 (HP > 3.6; MN > 4.5; AK > 2.2; DL > 8.6; SW > 5.9); real motion in V1: t(600) > 5.5, p << 0.001 (HP > 18.6; MN > 9.6; AK > 5.5; DL > 8.0; SW > 8.0). Cortical representations of the mapping stimuli are indicated by colored outlines taken from (B). The solid white line indicates a patch of significant activation (t[600] > 2.7, p < 0.008 [HP > 3.0; MN > 4.1; AK > 2.9; DL > 2.7; SW > 3.4]) for the following conjunction of contrasts, which represents the ideal activation pattern: (middle > upper) and (middle > lower) and (apparent motion > upper) and (apparent motion > lower).
(D) Event-related BOLD signal change plotted over time (from 6 s before to 24 s after stimulus onset) from the respective patch outlined by the solid white line in panel (C). Error bars correspond to standard errors of the mean.
(2.5 MB ZIP).
Click here for additional data file.
Figure S2 Occipital BOLD Activation of Five Subjects in Experiment 2
Color coding for subjects HP, MN, AK, LM, and MW is identical to the coding for HP described in Figure 3.
(A) Left occipital hemispheres with superimposed contrast maps indicating the cortical representation of the stimulus positions: upper (yellow), t(1445) > 8.3, p << 0.001 (HP > 9.3; MN > 12.4; AK > 8.9; LM > 11.4; MW > 8.3), middle (orange), t(1445) > 4, p << 0.001 (HP > 11.9; MN > 10.4; AK > 4.0; LM > 11.4; MW > 5.9), and lower (red), t(1445) > 4.2, p << 0.001 (HP > 15.1; MN > 11.0; AK > 11.0; LM > 6.3; MW > 4.2). For comparison, the patches from experiment 1 are marked with dotted lines.
(B) Cortical activation maps for apparent motion (green) and real motion (blue). In this case, apparent motion is contrasted with the flicker control condition (apparent motion > flicker in V1: t[1445] > 2.1, p < 0.05 [HP > 3.6; MN > 2.4; AK > 3.4; LM > 3.0; MW > 2.6]); real motion is compared to fixation baseline (real motion > baseline in V1: t[1445] > 10.1, p << 0.001 [HP > 18.6; MN > 20.3; AK > 10.1; LM > 12.6; MW > 11.1]). The white line in (B) indicates the regions from which examples of BOLD activity were taken and presented in (C). Bars indicate average activity during the respective conditions expressed in beta weights from a GLM analysis. Differences between apparent motion and flicker were analyzed separately for runs in which subjects performed the center task (solid bar) and runs in which subjects viewed passively (hatched bar). Although the overall apparent motion-versus-flicker contrasts were significant in each example (p < 0.05), not all of the more specific subcomparisons reached the significance level (p < 0.05; significant contrasts are marked with asterisks; all p-values are corrected for serial correlation).
(1.4 MB ZIP).
Click here for additional data file.
Figure S3 Group Results from Bistable Motion Quartet (Experiment 3)
Results of experiment 3 are shown on medio-posterior views of the inflated left occipital cortex of all six subjects ([A and B] LM, HP, and AK; [C and D] VV, MW, and MN). Color coding for subjects LM, VV, MW, and MN is identical to the coding for HP and AK described in Figure 6.
(A and C) Activation maps show the cortical representation of the stimulated locations in red (motion quartet > baseline). Contrast maps in green indicate regions that are more active for vertical apparent motion than for horizontal apparent motion (ROI-based GLM at locations indicated by the white line—LM: t[708] > 3.3, p < 0.001; HP: t[301] > 2.5, p < 0.02; AK: t[708] > 2.0, p < 0.05; VV: t[708] > 2.6, p < 0.01; MW: t[353] > 3.0, p < 0.01; MN: t[708] > 2.3, p < 0.02). The dotted line is a spline-interpolated curve connecting the stimulated locations and the region that is more active during the perception of vertical apparent motion. (This line does not necessarily indicate the path of apparent motion from experiments 1 and 2 since stimulus parameters had to be adjusted.) The solid white lines mark regions from which event-related averages were calculated.
(B and D) Event-related averages are aligned to the time point at which the subjects indicated a switch in perception (t = 0, black line). The first perceptual period following each stimulation onset is omitted from the analysis (see Materials and Methods). Error bars correspond to standard errors of the mean.
(856 KB ZIP).
Click here for additional data file.
This work was supported by the Deutsche Forschungsgemeinschaft, the Bundesministerium für Bildung und Forschung (grant DLR/BMBF 01 GO 0203), and the Max Planck Society. We thank Ralf A. W. Galuske, Rainer Goebel, and Jean-Michel Hupé for helpful discussions.
Competing interests. The authors have declared that no competing interests exist.
Author contributions. LM, AK, NK, and WS conceived and designed the experiments. LM and AK performed the experiments. LM and AK analyzed the data. LM, AK, NK, and WS wrote the paper.
¤ Current address: National Institutes of Health, Bethesda, Maryland, United States of America
Citation: Muckli L, Kohler A, Kriegeskorte N, Singer W (2005) Primary visual cortex activity along the apparent-motion trace reflects illusory perception. Plos Biol 3(8): e265.
Abbreviations
AMIattentional-modulation index
BOLDblood-oxygenation-level-dependent
fMRIfunctional magnetic resonance imaging
GLMgeneral linear model
hMT/V5+human MT/V5 complex
ROIregion of interest
V[number]visual cortex area [number]
==== Refs
References
Wertheimer M Experimentelle Studien über das Sehen von Bewegung Z Psychol 1912 61 161 265
Braddick OJ Low-level and high-level processes in apparent motion Philos Trans R Soc Lond B Biol Sci 1980 290 137 151 6106234
Dubner R Zeki SM Response properties and receptive fields of cells in an anatomically defined region of the superior temporal sulcus in the monkey Brain Res 1971 35 528 532 5002708
Mikami A Newsome WT Wurtz RH Motion selectivity in macaque visual cortex. II. Spatiotemporal range of directional interactions in MT and V1 J Neurophysiol 1986 55 1328 1339 3734858
Newsome WT Mikami A Wurtz RH Motion selectivity in macaque visual cortex. III. Psychophysics and physiology of apparent motion J Neurophysiol 1986 55 1340 1351 3734859
Albright TD Cortical processing of visual motion Rev Oculomot Res 1993 5 177 201 8420549
Mikami A Direction selective neurons respond to short-range and long-range apparent motion stimuli in macaque visual area MT Int J Neurosci 1991 61 101 112 1809728
Goebel R Khorram-Sefat D Muckli L Hacker H Singer W The constructive nature of vision: Direct evidence from functional magnetic resonance imaging studies of apparent motion and motion imagery Eur J Neurosci 1998 10 1563 1573 9751129
Muckli L Kriegeskorte N Lanfermann H Zanella FE Singer W Apparent motion: Event-related functional magnetic resonance imaging of perceptual switches and states J Neurosci 2002 22 RC219 11978860
Gattass R Gross CG Sandell JH Visual topography of V2 in the macaque J Comp Neurol 1981 201 519 539 7287933
Smith AT Singh KD Williams AL Greenlee MW Estimating receptive field size from fMRI data in human striate and extrastriate visual cortex Cereb Cortex 2001 11 1182 1190 11709489
Angelucci A Levitt JB Walton EJS Hupé JM Bullier J Circuits for local and global signal integration in primary visual cortex J Neurosci 2002 22 8633 8646 12351737
Stettler DD Das A Bennett J Gilbert CD Lateral connectivity and contextual interactions in macaque primary visual cortex Neuron 2002 36 739 750 12441061
Seghier M Dojat M Delon-Martin C Rubin C Warnking J Moving illusory contours activate primary visual cortex: An fMRI study Cereb Cortex 2000 10 663 670 10906313
Pascual-Leone A Walsh V Fast backprojections from the motion to the primary visual area necessary for visual awareness Science 2001 292 510 512 11313497
Antal A Kincses TZ Nitsche MA Paulus W Modulation of moving phosphene thresholds by transcranial direct current stimulation of V1 in human Neuropsychologia 2003 41 1802 1807 14527543
Silvanto J Cowey A Lavie N Walsh V Striate cortex (V1) activity gates awareness of motion Nat Neurosci 2005 8 143 144 15643428
Yantis S Nakama T Visual interactions in the path of apparent motion Nat Neurosci 1998 1 508 512 10196549
Liu T Slotnick SD Yantis S Human MT+ mediates perceptual filling-in during apparent motion NeuroImage 2004 21 1772 1780 15050597
Neuhaus W Experimentelle Untersuchung der Scheinbewegung Arch Gesamte Psychol 1930 75 315 458
Ramachandran VS Anstis SM Perceptual organization in multistable apparent motion Perception 1985 14 135 143 4069943
Kastner S De Weerd P Desimone R Ungerleider LG Mechanisms of directed attention in the human extrastriate cortex as revealed by functional MRI Science 1998 282 108 111 9756472
Tootell RBH Hadjikhani N Hall EK Marrett S Vanduffel W The retinotopy of visual spatial attention Neuron 1998 21 1409 1422 9883733
Brefczynski JA DeYoe EA A physiological correlate of the “spotlight” of visual attention Nat Neurosci 1999 2 370 374 10204545
Gandhi SP Heeger DJ Boynton GM Spatial attention affects brain activity in human primary visual cortex Proc Natl Acad Sci U S A 1999 96 3314 3319 10077681
Kastner S Pinsk MA De Weerd P Desimone R Ungerleider LG Increased activity in human visual cortex during directed attention in the absence of visual stimulation Neuron 1999 22 751 761 10230795
Somers DC Dale AM Seiffert AE Tootell RBH Functional MRI reveals spatially specific attentional modulation in human primary visual cortex Proc Natl Acad Sci U S A 1999 96 1663 1668 9990081
Jancke D Chavane F Naaman S Grinvald A Imaging cortical correlates of illusion in early visual cortex Nature 2004 428 423 426 15042090
Sterzer P Eger E Kleinschmidt A Responses of extrastriate cortex to switching perception of ambiguous visual motion stimuli Neuroreport 2003 14 2337 2341 14663187
Bullier J Hupé JM James AC Girard P The role of feedback connections in shaping the responses of visual cortical neurons Prog Brain Res 2001 134 193 204 11702544
Watanabe T Harner AM Miyauchi S Sasaki Y Nielsen M Task-dependent influences of attention on the activation of human primary visual cortex Proc Natl Acad Sci U S A 1998 95 11489 11492 9736764
Slotnick SD Schwarzbach J Yantis S Attentional inhibition of visual processing in human striate and extrastriate cortex NeuroImage 2003 19 1602 1611 12948715
Hupé JM James AC Payne BR Lomber SG Girard P Cortical feedback improves discrimination between figure and background by V1, V2 and V3 neurons Nature 1998 394 784 787 9723617
Supèr H Spekreijse H Lamme VAF Two distinct modes of sensory processing observed in monkey primary visual cortex (V1) Nat Neurosci 2001 4 304 310 11224548
Galuske RA Schmidt KE Goebel R Lomber SG Payne BR The role of feedback in shaping neural representations in cat visual cortex Proc Natl Acad Sci U S A 2002 99 17083 17088 12477930
Lee SH Blake R Heeger DJ Traveling waves of activity in primary visual cortex during binocular rivalry Nat Neurosci 2005 8 22 23 15580269
Mikami A Newsome WT Wurtz RH Motion selectivity in macaque visual cortex. I. Mechanisms of direction and speed selectivity in extrastriate area MT J Neurophysiol 1986 55 1308 1327 3016210
Salin PA Girard P Kennedy H Bullier J Visuotopic organization of corticocortical connections in the visual system of the cat J Comp Neurol 1992 320 415 434 1629397
Salin PA Bullier J Corticocortical connections in the visual system: Structure and function Physiol Rev 1995 75 107 154 7831395
Polonsky A Blake R Braun J Heeger DJ Neuronal activity in human primary visual cortex correlates with perception during binocular rivalry Nat Neurosci 2000 3 1153 1159 11036274
Tong F Engel SA Interocular rivalry revealed in the human cortical blind-spot representation Nature 2001 411 195 199 11346796
Lee SH Blake R V1 activity is reduced during binocular rivalry J Vis 2002 2 618 626 12678633
Tong F Primary visual cortex and visual awareness Nat Rev Neurosci 2003 4 219 229 12612634
Skiera G Petersen D Skalej M Fahle M Correlates of figure-ground segregation in fMRI Vision Res 2000 40 2047 2056 10828471
Lamme VAF Why visual attention and awareness are different Trends Cogn Sci 2003 7 12 18 12517353
Supèr H van der Togt C Spekreijse H Lamme VAF Internal state of monkey primary visual cortex (V1) predicts figure-ground perception J Neurosci 2003 23 3407 3414 12716948
Sasaki Y Watanabe T The primary visual cortex fills in color Proc Natl Acad Sci U S A 2004 101 18251 18256 15596726
Crick F Koch C Are we aware of neural activity in primary visual cortex? Nature 1995 375 121 123 7753166
Reppas JB Niyogi S Dale AM Sereno MI Tootell RBH Representation of motion boundaries in retinotopic human visual cortical areas Nature 1997 388 175 179 9217157
Nishida S Sasaki Y Murakami I Watanabe T Tootell RB Neuroimaging of direction-selective mechanisms for second-order motion J Neurophysiol 2003 90 3242 3254 12917391
Seiffert AE Somers DC Dale AM Tootell RBH Functional MRI studies of human visual motion perception: Texture, luminance, attention and after-effects Cereb Cortex 2003 13 340 349 12631563
Chen LM Friedman RM Roe AW Optical imaging of a tactile illusion in area 3b of the primary somatosensory cortex Science 2003 302 881 885 14500850
Logothetis NK Pauls J Augath M Trinath T Oeltermann A Neurophysiological investigation of the basis of the fMRI signal Nature 2001 412 150 157 11449264
Niessing J Ebisch B Schmidt KE Niessing M Singer W Hemodynamic signals correlate tightly with synchronized gamma oscillations Science 2005 In press
Logothetis NK The underpinnings of the BOLD functional magnetic resonance imaging signal J Neurosci 2003 23 3963 3971 12764080
Finlay D von Grünau MW Some experiments on the breakdown effect in apparent motion Percept Psychophys 1987 42 526 534 3696947
Selmes CM Fulham WR Finlay DC Chorlton MC Manning ML Time-till-breakdown and scalp electrical potential maps of long-range apparent motion Percept Psychophys 1997 59 489 499 9158324
Chaudhuri A Modulation of the motion aftereffect by selective attention Nature 1990 344 60 62 2304555
Tootell RBH Mendola JD Hadjikhani NK Liu AK Dale AM The representation of the ipsilateral visual field in human cerebral cortex Proc Natl Acad Sci U S A 1998 95 818 824 9448246
Boynton GM Engel SA Glover GH Heeger DJ Linear systems analysis of functional magnetic resonance imaging in human V1 J Neurosci 1996 16 4207 4221 8753882
Engel SA Rumelhart DE Wandell BA Lee AT Glover GH fMRI of human visual cortex Nature 1994 369 525 8031403
Sereno MI Dale AM Reppas JB Kwong KK Belliveau JW Borders of multiple visual areas in humans revealed by functional magnetic resonance imaging Science 1995 268 889 893 7754376
Linden DE Kallenbach U Heinecke A Singer W Goebel R The myth of upright vision. A psychophysical and functional imaging study of adaptation to inverting spectacles Perception 1999 28 469 481 10664787
Kriegeskorte N Goebel R An efficient algorithm for topologically correct segmentation of the cortical sheet in anatomical mr volumes NeuroImage 2001 14 329 346 11467907
| 16018720 | PMC1175820 | CC BY | 2021-01-05 08:21:26 | no | PLoS Biol. 2005 Aug 19; 3(8):e265 | utf-8 | PLoS Biol | 2,005 | 10.1371/journal.pbio.0030265 | oa_comm |
==== Front
PLoS BiolPLoS BiolpbioplosbiolPLoS Biology1544-91731545-7885Public Library of Science San Francisco, USA 1600450910.1371/journal.pbio.0030266Research ArticleEcologyEvolutionPaleontologyMammalsThe Impact of the Species–Area Relationship on Estimates of Paleodiversity Paleospecies-Area CurvesBarnosky Anthony D [email protected]
1
2
Carrasco Marc A
1
Davis Edward B
1
1Department of Integrative Biology and Museum of Paleontology, University of California, Berkeley, California, United States of America,2Museum of Vertebrate Zoology, University of California, Berkeley, California, United States of AmericaLevin Simon Academic EditorPrinceton UniversityUnited States of America8 2005 19 7 2005 19 7 2005 3 8 e2664 3 2005 1 6 2005 Copyright: © 2005 Barnosky 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.
Space Matters: Estimating Species Diversity in the Fossil Record
Estimates of paleodiversity patterns through time have relied on datasets that lump taxonomic occurrences from geographic areas of varying size per interval of time. In essence, such estimates assume that the species–area effect, whereby more species are recorded from larger geographic areas, is negligible for fossil data. We tested this assumption by using the newly developed Miocene Mammal Mapping Project database of western North American fossil mammals and its associated analysis tools to empirically determine the geographic area that contributed to species diversity counts in successive temporal bins. The results indicate that a species–area effect markedly influences counts of fossil species, just as variable spatial sampling influences diversity counts on the modern landscape. Removing this bias suggests some traditionally recognized peaks in paleodiversity are just artifacts of the species–area effect while others stand out as meriting further attention. This discovery means that there is great potential for refining existing time-series estimates of paleodiversity, and for using species–area relationships to more reliably understand the magnitude and timing of such biotically important events as extinction, lineage diversification, and long-term trends in ecological structure.
The importance of accounting for geographic area in estimating biodiversity from the fossil record is revealed through an analysis of the Miocene Mammal Mapping Project.
==== Body
Introduction
Paleontology-based estimates of how species diversity fluctuates through time [1–6] typically do not take into account the species–area curve, which is one of the best understood relationships in ecology and predicts how many species can be expected as the geographic area of sampling increases [7–11]. The relationship generally is expressed by S = cAz, where S is the number of species, A is the geographic area that was sampled, and c and z are empirically derived constants that express the slope of the power function and its intercept. Generally, z-values are lowest for accumulations of species within continental biogeographic provinces (z of approximately 0.15), intermediate for island archipelagos (z of approximately 0.25–0.45), and highest for data accumulated across continental biogeographic provinces (z of approximately 0.9). These relationships hold for nearly all taxonomic groups, including plants, aquatic invertebrates, insects, birds, reptiles, amphibians, and mammals [8–17], and have led to a universal recognition that the species–area relationship has high predictive power for estimating biodiversity on the modern landscape and modeling such long-term processes as extinction [8,10,11,13,18–22, but see 23,24].
Despite that recognition, it has never been possible to directly determine species–area relationships for paleontological data, which form the basis for understanding biotic processes such as how diversity fluctuates over long time periods, timing and causes of lineage diversification, and the magnitude of mass extinctions [1–6,25–29]. The importance of such assessment has been recognized [30] and therefore attempted in indirect ways, primarily by using various proxies of rock volume or outcrop area to estimate the potential for sampling taxa from various geologic time intervals [6,31,32]. Such studies have concentrated on marine invertebrate taxa and have demonstrated correlations between rock volume, outcrop area, collection effort, and paleodiversity. However, correcting for a species–area bias differs from these indirect proxies for standardization because a species–area correction must take into account the actual geographic area encompassed by the sampling sites instead of the area potentially available for sampling.
The reason that the paleospecies–area relationship has not been addressed directly stems from technological constraints and the way paleontological data were traditionally compiled, i.e., with more attention to placing them in a time interval and analyzing differences through time than to placing them in geographic context and analyzing changes across space within a given time interval. Now both of these constraints have been removed with the advent of (i) electronic paleontological databases, which include both geographic coordinates for fossil sites and lists of taxa per site [33], and (ii) Geographic Information Systems and imaging programs that allow direct measurement of the geographic area encompassed by a set of fossil species.
Here we utilize a paleontological database, Miocene Mammal Mapping Project (MIOMAP [34]), to directly derive aspects of the paleospecies–area relationship for fossil terrestrial mammals (Figure 1). The MIOMAP database is a compilation of all western North American mammal occurrences and associated geologic information between 5 and 30 million years old (details of which are fully documented online [34]). It includes approximately 3,100 georeferenced localities and 14,000 occurrences of taxa. The temporal bins are those shown in Table 1. We use this dataset because (i) we built it and are intimately familiar with its potentials and limitations; (ii) the novel mapping interfaces allow direct measurement of the area that encompasses the fossil data; (iii) the fossil localities can be well-constrained, both geographically (Figure 1) and temporally, and are well documented with metadata [34]; (iv) fossil mammals have been used for some of the most comprehensive examinations of paleodiversity fluctuations and their causes [2,4,30,35–37]; and (v) potential paleospecies–area biases have been recognized as problematic for mammals but have not been exhaustively explored [30]. Although we focus on fossil mammals, our results have implications for paleodiversity curves for any taxon.
Figure 1 Distribution of Fossil Localities in the MIOMAP Database and Oligo-Miocene Biogeographic Provinces
Gray dots show localities for all time intervals; black dots show localities that are Early Barstovian age. Black lines and numbers demarcate and label the following biogeographic provinces: 1, Columbia Plateau; 2, Northern Rockies; 3, Northern Great Plains; 4, Northern Great Basin; 5, Colorado Plateau; 6, Coastal California; 7, Mojave Desert; 8, Southern Great Basin; 9, Southern Great Plains.
Table 1 Temporal Bins into Which Species Occurrences Were Sorted
We tested for species–area effects in the fossil data in two ways. First, we derived the paleospecies–area curve for a single temporal bin for which subcontinental-scale spatial coverage was best—the Early Barstovian. This curve was produced from a nested set of faunal lists (i.e., a type I curve of reference [10]), which started with one of the biogeographic provinces (Figure 1) and successively added species from the other ones, such that species accumulated as the sampling radius encompassed more and more of the western United States until all localities and provinces were included. Second, we examined unnested data, whereby species richness within each temporal bin for each of the nine different biogeographic regions was plotted against the geographic sampling area (type IV curve of reference [10]). Today, the continental biogeographic provinces we used are regarded as biogeographically distinct from one another [38–40], and it is likely that provincial differences also existed through the Oligocene and Miocene, although characteristics of each province were different from those of today [4,41].
Our metric for diversity is paleospecies richness, i.e., the maximum number of species identified in a sample. Because paleospecies richness is strongly dependent on many biases that influence sample size, we first standardized the data using rarefaction [42–46] based on occurrences (basically whether a taxon was present or absent at a given locality; see the Materials and Methods section for the alternative data parameters tested).
With the rarefaction results, we plotted the species richness values determined at 75 occurrences (a value that was high enough to see divergence in the rarefaction curves, but low enough to include many time intervals) against geographic area (the smallest polygon that enclosed all the localities contributing to a given species list). Analysis of the rarefaction curves and plots allowed us to estimate the impact of the species–area relationship on paleodiversity, and to distinguish peaks in diversity that are artifacts of the species–area effect from those that may indicate times of real diversity fluctuation.
Results/Discussion
Species–Area Effect with Nested Data
The species–area curve derived from nested data (Figure 2) within a single well-sampled time interval, the Early Barstovian (see Figure 1 and [34] for localities included), clearly demonstrates that species–area effects might underlie much of the diversity fluctuation apparent in paleontological datasets (Figure 2). We found that, just as for the modern species–area curves, the paleospecies–area curves for our data are described by a power function with a very high R2 value and a highly significant slope. It is not surprising that uncorrected species counts (Figure 2A) show a high R2 value (0.95) and significant slope (p < 0.0002), given the numerous biases in fossil data that are related to sample size. However, even after rarefying the data, a very strong species–area relationship (R2 = 0.85, p <0.0034) still existed (Figure 2B). This demonstrates that occurrence rarefaction by itself is insufficient to correct for size of geographic area, and that many of the apparent diversity fluctuations in paleontological datasets may be the result of the species–area relationship. This effect should be especially pronounced with the unequal geographic sampling that characterizes most paleontologic data.
Figure 2 Paleospecies–Area Curve for the Early Barstovian Temporal Interval
(A) Uncorrected counts of total species found in each successively larger geographic area.
(B) Species richness values standardized by using the rarefaction value for 75 occurrences. Dotted lines are 95% confidence intervals around the best-fit line. Equation abbreviations: S, species; A, area.
Species–Area Effect with Unnested Data
Plotting unnested data from each temporal bin for each of the nine biogeographic provinces (see Figure 1) further underscores how geographic coverage contributes to paleospecies diversity fluctuations (Figure 3). The paleospecies–area effect is highly significant (p < 0.001) in both uncorrected (Figure 3A) and occurrence-standardized assessments, but becomes a better fit with the standardized data (Figure 3B). This suggests that as other sampling biases are stripped away, the underlying control on paleospecies richness per time bin is geographic sampling area.
Figure 3 Number of Species Plotted against the Total Geographic Area of the Fossil Localities
(A) Uncorrected counts of total species found in each temporal bin within each biogeographic province.
(B) Species richness values standardized by using the rarefaction value for 75 occurrences. Each data point represents one biogeographic province in one of the temporal bins listed in Table 1. Dotted lines are 95% confidence intervals around the best-fit line.
Paleodiversity Curves
Usually paleodiversity patterns are depicted as time series of sample-standardized data. In assessing apparent diversity change in such time series in different geographic regions (regions 1–3 and 6 in Figure 1), we found that the paleospecies–area effect by itself explains most of the apparent diversity fluctuations (Figure 4). The strength of the paleospecies–area effect becomes evident by comparing geographic sampling area through time with the “traditional” time-series depictions (Figure 4). In general, the peaks in the traditional time-series curves are species lists accumulated over large geographic areas, and the declines are species lists accumulated within smaller areas. This is clearly demonstrated by comparing paleodiversity fluctuations with geographic sampling area fluctuations (black lines versus gray lines in Figure 4). By determining the departures from this general concordance between geographic area and paleodiversity, it is possible to recognize which diversity fluctuations most likely have a cause rooted in biological process or sampling biases that are not predominantly caused by the species–area effect, as illustrated by the following examples.
Figure 4 Comparison of Species Diversity and Geographic Sampling Area through Time
Black lines and black boxes are time-series plots of occurrence-rarefaction species richness at 75 occurrences for each of the four biogeographic provinces. The numbers by the data points identify the time bins as indicated in Table 1; bin 1 is oldest and bin 14 is youngest. Gray lines and open boxes show the geographic area sampled for each time bin.
(A) Northern Great Plains (region 3 in Figure 1).
(B) Northern Rockies (region 2).
(C) Columbia Plateau (region 1).
(D) Coastal California (region 6).
In the Northern Great Plains (Figure 4A), the only fluctuations that are not predominantly explained by sampling differently sized geographic areas are in time bins 1 and 9 (lower diversity than expected) and possibly bin 13 (higher diversity than expected). The other fluctuations that would be interpreted from looking at the traditional time series alone (black line in Figure 4A), e.g., the dip and rise from time bins 4 through 8, are simply artifacts of the species–area effect. Likewise, in the Northern Rockies (Figure 4B), variable geographic sampling area explains most of the apparent time series fluctuations except low diversity in time bin 8; and in the California Coast (Figure 4D), only time bin 10 stands out as a true decline with respect to geographic area.
In the Columbia Plateau (Figure 4C), the time-series diversity fluctuations can generally be explained by the species–area effect, except for time bin 8, which exhibits a particularly pronounced deviation from that expected from geographic sampling area. In fact, that time features an unusual sampling situation, in which two geographically close faunas (Red Basin and Quartz Basin, Oregon) accumulated in two very different depositional environments [47], thus accounting for the rise in diversity even though geographic area decreased. In this case, the discrepancy between the diversity curve and the geographic area curve highlights an unusual sampling situation that must be taken into account when interpreting diversity fluctuations.
In addition, intraprovincial regression plots for these four regions were inspected to determine if they were consistent with the overall pattern seen in Figure 3. For the Northern Great Plains, Northern Rockies, and California Coast, the paleospecies–area regressions result in the expected positive slopes, but lack statistical significance likely because of the small number of points. For the Columbia Plateau, time bin 8 is such an outlier that it actually makes the slope slightly negative (though not statistically significant); however, with post hoc removal of time bin 8, the slope becomes significantly positive (R2 = 0.8524, p = 0.009), despite being based on only six remaining points. Like the time-series curves, such inspections are valuable in identifying time bins that may be recording unusual sampling biases, which are recognizable as extreme outliers.
Of additional relevance is the fact that the time series indicate that temporal coverage varies considerably from region to region. None of the regions have data for all of the temporal bins, and only the Northern Great Plains has a reasonably continuous time series. This suggests that paleospecies–area effects that we have demonstrated region by region may be magnified, rather than averaged out, by combining all geographic regions into a single dataset for time-series analysis. It will be enlightening to apply paleospecies–area corrections to the many paleodiversity curves that have been generated with other datasets as a prelude to interpreting the cause of apparent changes in diversity.
Interpretation of Z-Values
Z-values are informative for species–area curves because they differ in datasets sampled from within provinces versus those that accumulate species from multiple provinces. Figure 2 exemplifies an interprovincial paleo-z-value. In general, the z-value for this Early Barstovian paleospecies–area curve is lower than for modern interprovincial species–area curves, as would be expected from the less complete sampling of the fossil biota as compared to the modern biota. The lower slope in the occurrence-standardized curve (Figure 2B) as compared to the uncorrected curve (Figure 2A) and to modern species–area curves also results from rarefaction of the localities [48]. The z-values for regressions illustrated in Figure 3 are for unnested data and thus are not directly comparable to z-values for type I species–area curves. However, it is worth noting that their z-values also are low compared to either inter- or intraprovincial species–area curves [8], as would be expected from the rarefaction effect.
Conclusions
Correcting the species–area bias in paleontological data is different from the indirect proxies for sample standardization that have been applied in the past, such as standardizing for rock volume or outcrop area, because it actually takes into account dispersion of sampling sites, rather than area potentially available for sampling. Our demonstration that the paleospecies–area relationship explains many of the spatiotemporal differences that have been traditionally seen in species richness of Miocene mammals has widespread implications.
First, it is likely that the influence of the paleospecies–area relationship is similarly strong in most fossil datasets, meaning that the shapes of global and regional paleodiversity curves may change substantially once the area effect is accounted for. This indicates that it may now be possible to refine our understanding of widely recognized events in the history of life by correcting for the paleospecies–area effect. For example, taking into account paleospecies–area effects will be important in clarifying the controversy about whether marine diversity substantially increased in the Cenozoic—the so-called pull of the recent [5,49]. Paleospecies–area corrections can be expected to adjust magnitudes and perhaps timing of extinction events documented in the fossil record, which in turn are of particular value in comparisons of past with current extinction rates [9,25,50,51]. Interpretations of ecological dynamics through time [6,30,35,49,52], the cornerstone of inferring macroecological processes from pattern [11,53], also are subject to paleospecies–area considerations.
Second, it has been prohibitively difficult to correct for area effects on paleospecies richness, but with interactive mapping and image analysis software applied to paleontological databases as we have done here, such corrections are now feasible. Finally, the recognition that paleospecies–area curves exhibit some of the same properties (high R2 values, significant slopes) as modern species–area curves portends an important way to merge macroecology with paleontology.
Materials and Methods
Temporal bins
We followed published literature in assigning fossil occurrences to subdivisions of the North American Land Mammal Ages as specified in Tedford et al. [41]. Although the temporal bins are not equal in duration, we determined that this has little influence on diversity counts per bin because (i) there is no correlation between bin length and number of localities (R2 = 0.04); (ii) the localities do not span the entire time represented by each bin [34]; and (iii) recent work [54] has indicated that bins of the sort used here, based on maximum taxon associations, are best suited to comparisons of diversity through time, as they produce a series of biologically meaningful groupings that do not change much within each bin.
Species counts
We calculated both maximum and minimum number of species, the former including all specimens that were only identified to genus or higher taxon as belonging to unique species, the latter including all such taxa as belonging to a species represented by more diagnostic material. We present the minimum counts here, but using maximum counts does not substantively alter our results.
Geographic area calculation
Geographic areas were calculated by using the MIOMAP mapping interface to zoom in on the set of localities at appropriate scales, importing the resulting image into the image analysis program ImageJ (http://rsb.info.nih.gov/ij/, and then using the ImageJ software to trace the minimum convex polygon that would enclose all the localities of interest and to calculate the area enclosed by the polygon. For our purposes this was the most appropriate measure of geographic area (rather than summing radii around localities or smaller polygons that minimally enclosed spatially discrete clusters of localities, but not counting the area that separates the clusters) because the intent was to see how species accumulate as total geographic area is added.
Rarefaction methods
Because paleospecies richness is strongly dependent on biases that influence sample size, we standardized the species counts using rarefaction [42–46]. Rarefaction was accomplished with S. Holland's analytic rarefaction software (http://www.uga.edu/strata/software/, based on the analytical solutions to rarefaction presented by Raup [42] and originally derived by Hurlbert [44] and Heck et al. [45]. A review of the development of this methodology can be found in Tipper [43]. Detailed study of the taphonomic and collection biases of all included fossil sites can identify whether rarefaction using number of identified specimens (NISP—how many specimens represented a given taxon) or occurrences (whether a taxon was present or absent at a given locality) is more appropriate; however, that is usually prohibitively labor-intensive. Because either method might be appropriate in a given instance, we analyzed our data in both ways to understand the range of results that might be produced. We found that many of the NISP counts are based only on “high-graded” specimen counts—i.e., publication of only the best specimens, rather than all the specimens that were collected from a given locality [55]. Time intervals that have many of these high-graded NISP counts, combined with many localities that have very poor NISP information, exhibit a rarefaction curve that rises artificially steeply. Rarifying by occurrences removes the effect of high-graded localities and missing data. In addition, this method produces results equivalent to another commonly used method of sample standardization—occurrences weighted resampling [46]. Given the benefits of employing occurrence rarefaction, we report only those results here. However, it is worth noting that using the NISP rarefactions would not alter our conclusions. We determined species richness at 50, 75, and 100 occurrences. We report only the richness values at 75 occurrences, because that occurrence value provided a large number of data points while at the same time eliminating points that were based on more suspect, spotty data. Interpreting results on the bases of 50 or 100 occurrences would result in substantially similar conclusions.
We thank B. Kraatz for helping to develop analytical software and A. Shabel, S. Hopkins, J. Tseng, J. McGuire, and B. Kraatz for helpful comments. Kelley Smith Feranec developed the graphics for the MIOMAP site. We appreciate reviews by J. Valentine, M. Fortelius, and an anonymous reviewer. National Science Foundation grants EAR-9909353 and EAR-0310221 provided funding. This is University of California Museum of Paleontology Contribution No. 1885.
Competing interests. The authors have declared that no competing interests exist.
Author contributions. ADB, MAC, and EBD conceived and designed the experiments. ADB, MAC, and EBD performed the experiments. ADB, MAC, and EBD analyzed the data. ADB wrote the paper.
Citation: Barnosky AD, Carrasco MA, Davis EB (2005) The impact of the species–area relationship on estimates of paleodiversity. PLoS Biol 3(8): e266.
Abbreviations
MIOMAPMiocene Mammal Mapping Project
NISPnumber of identified specimens
==== Refs
References
Raup DM Taxonomic diversity during the Phanerozoic Science 1972 177 1065 1071 17840596
Alroy J New methods for quantifying macroevolutionary patterns and processes Paleobiology 2000 26 707 733
Sepkoski JJ Bambach RK Raup DM Valentine JW Phanerozoic marine diversity and the fossil record Nature 1981 293 435 437
Barnosky AD Carrasco MA Effects of Oligo-Miocene global climate changes on mammalian species richness in the northwestern quarter of the USA Evol Ecol Res 2002 4 811 841
Jablonski D Roy K Valentine JW Price RM Anderson PS The impact of the pull of the recent on the history of marine diversity Science 2003 300 1133 1135 12750517
Sepkoski JJ A kinetic model of Phanerozoic taxonomic diversity. III. Post-Paleozoic families and mass extinctions Paleobiology 1984 10 246 267
Godron M Comparison of a species area curve with its model Oecolog Plantar 1971 6 189 196
Rosenzweig ML Species diversity in space and time 1995 New York Cambridge University Press 436
Rosenzweig ML Loss of speciation rate will impoverish future diversity Proc Natl Acad Sci U S A 2001 98 5404 5410 11344286
Scheiner SM Six types of species-area curves Glob Ecol Biogeogr 2003 12 441 447
Brown JH Macroecology 1995 Chicago University of Chicago Press 269
Anderson MJ Effects of patch size on colonisation in estuaries: Revisiting the species-area relationship Oecologia 1999 118 87 98 20135164
Crawley MJ Harral JE Scale dependence in plant biodiversity Science 2001 291 864 868 11157164
Kattan GH Franco P Bird diversity along elevational gradients in the Andes of Colombia: Area and mass effects Glob Ecol Biogeogr 2004 13 451 458
Ricklefs RE Lovette IJ The roles of island area per se and habitat diversity in the species-area relationships of four Lesser Antillean faunal groups J Anim Ecol 1999 68 1142 1160
Schoener TW Spiller DA Losos JB Natural restoration of the species-area relation for a lizard after a hurricane Science 2001 294 1525 1528 11711674
Schoereder JH Galbiati C Ribas CR Sobrinho TG Sperber CF Should we use proportional sampling for species-area studies? J Biogeogr 2004 31 1219 1226
Barnosky AD Hadly EA Maurer BA Christie MI Temperate terrestrial vertebrate faunas in North and South America: Interplay of ecology, evolution, and geography with biodiversity Conserv Biol 2001 15 658 674
Harte J Kinzig A Green J Self-similarity in the distribution and abundance of species Science 1999 284 334 336 10195901
Ostling A Harte J Green JL Kinzig AP A community-level fractal property produces power-law species-area relationships Oikos 2003 103 218 224
Thomas CD Cameron A Green RE Bakkenes M Beaumont LJ Extinction risk from climate change Nature 2004 427 145 148 14712274
Weiher E The combined effects of scale and productivity on species richness J Ecol 1999 87 1005 1011
Arita HT Rodriguez P Geographic range, turnover rate and the scaling of species diversity Ecography 2002 25 541 550
Seabloom EW Dobson AP Stoms DM Extinction rates under nonrandom patterns of habitat loss Proc Natl Acad Sci U S A 2002 99 11229 11234 12177416
Raup DM Sepkoski JJ Periodic extinction of families and genera Science 1986 231 833 836 11542060
Valentine JW Phanerozoic taxonomic diversity: A test of alternate models Science 1973 180 1078 1079 17806588
Bambach RK Species richness in marine benthic habitats through the Phanerozoic Paleobiology 1977 3 152 167
Foote M Origination and extinction components of taxonomic diversity: Paleozoic and post-Paleozoic dynamics Paleobiology 2000 26 578 605
Raup DM Species diversity in the Phanerozoic: A tabulation Paleobiology 1976 2 279 288
Alroy J McKinney ML Drake JA Equilibrial diversity dynamics in North American mammals Biodiversity dynamics: Turnover of populations, taxa, and communities 1998 New York Columbia University Press 232 287
Crampton JS Beu AG Cooper RA Jones CM Marshall B Estimating the rock volume bias in paleobiodiversity studies Science 2003 301 358 360 12805555
Raup DM Species diversity in the Phanerozoic: An interpretation Paleobiology 1976 2 289 297
Barnosky AD Paleontology Database Network [database] 2005 Available: http://www.ucmp.berkeley.edu/pdn/index.html . Accessed 7 June 2005
Carrasco MA Kraatz BP Davis EB Barnosky AD Miocene Mammal Mapping Project [database] 2005 Available: http://www.ucmp.berkeley.edu/miomap/ . Accessed 7 June 2005
Alroy J Constant extinction, constrained diversification, and uncoordinated stasis in North American mammals: New perspectives on faunal stability in the fossil record Palaeogeogr Palaeoclimatol Palaeoecol 1996 127 285 311
Alroy J Koch PL Zachos JC Erwin DH Wing SL Global climate change and North American mammalian evolution Deep time: Paleobiology's perspective 2000 Lawrence (Kansas) Allen Press 259 288
Barnosky AD Hadly EA Bell CJ Mammalian response to global warming on varied temporal scales J Mamm 2003 84 354 368
Hagmeier EM A numerical analysis of the distributional patterns of North American mammals. II. Re-evaluation of the provinces Syst Zool 1966 15 279 299
Hagmeier EM Stults CD A numerical analysis of the distributional patterns of North American mammals Syst Zool 1964 13 125 155
Lugo AE Brown SL Dodson R Smith TS Shugart HH The Holdridge life zones of the conterminous United States in relation to ecosystem mapping J Biogeogr 1999 26 1025 1038
Tedford RH Albright LB Barnosky AD Ferrusquia-Villafranca I Hunt RM Woodburne MO Mammalian biochronology of the Arikareean through Hemphillian interval (late Oligocene through early Pliocene epochs) Late Cretaceous and Cenozoic mammals of North America 2004 New York Columbia University Press 169 231
Raup DM Taxonomic diversity estimation using rarefaction Paleobiology 1975 1 333 342
Tipper JC Rarefaction and rarefiction—The use and abuse of a method in paleoecology Paleobiology 1979 5 423 434
Hurlbert S The nonconcept of species diversity: A critique and alternative parameters Ecology 1971 52 577 586
Heck KLJ Van Belle G Simberloff D Explicit calculation of the rarefaction diversity measurement and the determination of sufficient sample size Ecology 1975 56 1459 1461
Bush AM Markey MJ Marshall CR Removing bias from diversity curves: The effects of spatially organized biodiversity on sampling standardization Paleobiology 2004 30 666 686
Shotwell JA Miocene mammals of southeast Oregon Bull Univ Oreg Mus Nat Hist 1968 14 1 67
Olszewski TD A unified mathematical framework for the measurement of richness and evenness within and among multiple communities Oikos 2004 104 377 387
Alroy J Marshall CR Bambach RK Bezusko K Foote M Effects of sampling standardization on estimates of Phanerozoic marine diversification Proc Natl Acad Sci U S A 2001 98 6261 6266 11353852
Regan HM Lupia R Drinnan AN Burgman MA The currency and tempo of extinction Am Nat 2001 157 1 10 18707231
Nichols JD Pollock KH Estimating taxonomic diversity, extinction rates, and speciation rates from fossil data using capture-recapture models Paleobiology 1983 9 150 163
Kirchner JW Evolutionary speed limits inferred from the fossil record Nature 2002 415 65 68 11780116
Isaac NJB Mallet J Mace GM Taxonomic inflation: Its influence on macroecology and conservation Trends Ecol Evol 2004 19 464 469 16701308
Escarguel G Bucher H Counting taxonomic richness from discrete biochronozones of unknown duration: A simulation Palaeogeogr Palaeoclimatol Palaeoecol 2004 202 181 208
Davis EB Pyenson ND Assessing mammalian paleofaunal diversity: Discrepancies between published and museum collection data for the Miocene of northwestern Nevada, USA GSA Abst Prog 2003 35 498
| 16004509 | PMC1175821 | CC BY | 2021-01-05 08:21:26 | no | PLoS Biol. 2005 Aug 19; 3(8):e266 | utf-8 | PLoS Biol | 2,005 | 10.1371/journal.pbio.0030266 | oa_comm |
==== Front
PLoS BiolPLoS BiolpbioplosbiolPLoS Biology1544-91731545-7885Public Library of Science San Francisco, USA 1600850410.1371/journal.pbio.0030273Research ArticleBiophysicsBiotechnologyEcologyEvolutionGenetics/Genomics/Gene TherapyMicrobiologyEubacteriaIn VitroNew Insights into Metabolic Properties of Marine Bacteria Encoding Proteorhodopsins New Insights into Marine ProteorhodopsinsSabehi Gazalah
1
Loy Alexander
2
Jung Kwang-Hwan
3
¤Partha Ranga
3
Spudich John L
3
Isaacson Tal
4
Hirschberg Joseph
4
Wagner Michael [email protected]
2
Béjà Oded [email protected]
1
1Department of Biology, Technion-Israel Institute of Technology, Haifa, Israel,2Department of Microbial Ecology, University of Vienna, Austria,3Center for Membrane Biology, Department of Biochemistry and Molecular Biology, The University of Texas Medical School, Houston, Texas, United States of America,4Department of Genetics, The Hebrew University of Jerusalem, IsraelEisen Jonathan Academic EditorInstitute for Genomic ResearchUnited States of America8 2005 19 7 2005 19 7 2005 3 8 e27320 4 2005 6 6 2005 Copyright: © 2005 Sabehi 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.
Light-Sensing Protein Illuminates Sun-Loving Ocean Bacteria
Proteorhodopsin phototrophy was recently discovered in oceanic surface waters. In an effort to characterize uncultured proteorhodopsin-exploiting bacteria, large-insert bacterial artificial chromosome (BAC) libraries from the Mediterranean Sea and Red Sea were analyzed. Fifty-five BACs carried diverse proteorhodopsin genes, and we confirmed the function of five. We calculate that proteorhodopsin-exploiting bacteria account for 13% of microorganisms in the photic zone. We further show that some proteorhodopsin-containing bacteria possess a retinal biosynthetic pathway and a reverse sulfite reductase operon, employed by prokaryotes oxidizing sulfur compounds. Thus, these novel phototrophs are an unexpectedly large and metabolically diverse component of the marine microbial surface water.
Metagenomic analyses of the Mediterranean Sea and Red Sea estimate that proteorhodopsin genes are exploited by a surprisingly high percentage of bacteria. Experimental studies reveal metabolically diverse roles for these phototrophs.
==== Body
Introduction
Proteorhodopsin (PR) proteins are bacterial retinal-binding membrane pigments that function as light-driven proton pumps in the marine ecosystem [1,2]. A gene encoding such a pigment was originally discovered on a large genome fragment [1] derived from an uncultured marine gammaproteobacterium of the SAR86 group [3,4]. Subsequently, many diverse PRs have been detected in marine plankton, via PCR-based gene surveys [5,6], environmental bacterial artificial chromosome (BAC) and fosmid libraries screening [7,8], or environmental shotgun libraries [9]. Recently, through comparative analyses of SAR86 rRNA-bearing genomic fragments, it was shown that diverse SAR86 members contain PR pigments belonging to different groups [7]. Furthermore, in another environmental genomics study, it was proposed that a Pacific PR is encoded by a planktonic alphaproteobacterium [8]. Although retrieval and comparative analyses of large genome fragments carrying PR genes is the most promising approach to phylogenetically assign and better understand uncultured PR-carrying organisms, the data accumulated to this day come from only five different PR genes contained within large insert BAC or fosmid clones: the original Pacific 31A08 clone [1], Antarctic ANT32C12 fosmid clone [8], Pacific Alphaproteobacteria-related clone HOT2C01 [8], Pacific clone HOT4E07, and eBAC20E09 clone from the Red Sea [7].
Results/Discussion
To better understand the extent of naturally occurring PR variability and physiological traits associated with PR-carrying organisms, we surveyed large insert BAC libraries (with inserts up to 170 Kb) from the photic zone of the Mediterranean Sea and Red Sea using Southern hybridization and newly designed general degenerated PR primers. The primers were designed based on alignments of PR sequences from the North Atlantic Ocean, the Mediterranean and Red Seas [5,6], the Pacific Ocean [7,8], and from the Sargasso Sea environmental shotgun project [9]. These primers amplified diverse PR sequences (red in Figure 1), which were not restricted to the three PR families we previously amplified using non-degenerative primers (orange in Figure 1). The diversity of PRs observed in the BAC library was comparable to recent findings from randomly sequenced small-insert shotgun libraries from the Sargasso Sea [9]. Fifty-five different BAC clones were found to contain PRs in the Mediterranean library, representing 0.52% of the total clones. Assuming (i) that an average marine bacterium had a genome size of 2.0 Mb, (ii) that the cloned DNA was recovered from exclusively prokaryotes, and (iii) that each PR-carrying microorganism carried only one PR gene copy on its genome, this PR abundance suggests that 13% of the bacteria in the photic zone of the Mediterranean Sea possess a PR gene (10,560 BAC clones × 80-Kb average insert size = 844.8 Mb; 844.8 Mb / 2.0 Mb = 422.2 genomes represented in the library; 55 PR genes in 422.2 genomes represent 13%). This estimation does not consider possible biases of environmental BAC libraries prepared in Escherichia coli against foreign DNA [10]. Interestingly, 50% of these PR-containing BAC clones fall into two distinct groups (red circles in Figure 1), which might represent the most abundant PR-containing bacteria in Mediterranean surface waters.
Figure 1 Phylogenetic Tree of PR Proteins from the Mediterranean and Red Seas along with PR Homologs in GenBank
The tree was divided into what we propose are distinct subfamilies of sequences, based on bootstrap values significance. The tree was constructed as follows: (i) All homologs of PR proteins were identified in GenBank including predicted proteins from the Sargasso Sea assemblies using BLASTp [36] searches with representatives of previously identified PR-like protein families as query sequences. (ii) All sequences greater than 300 nucleotides in length were aligned to each other using CLUSTALx [37], and a neighbor-joining phylogenetic tree was inferred using the neighbor programs of PAUP* [38]. Bootstrap resampling (1,000 pseudoreplications) of neighbor-joining and maximum parsimony trees were performed in all analyses to provide confidence estimation for the inferred topologies. Bootstraps values greater than 50% are indicated above the branches (neighbor-joining/maximum parsimony). The scale bar represents the number of substitutions per site. The sequences are colored according to the type of sample in which they were found: blue, cultured species; orange, sequences from uncultured organisms obtained using PCR-based methods; and red, BAC-derived sequences from uncultured species in the Mediterranean Sea and Red Sea (this study) or from previously reported Pacific, Antarctic, and Red Sea [1,7,8] BAC/fosmids. Black squares mark sequenced BACs in this study; red squares label BACs sequenced in previous reports. α, Alphaproteobacteria; γ, Gammaproteobacteria. Red circles mark the two abundant PR groups discussed in the manuscript.
BAC clones representing each PR family (black squares in Figure 1) were partially or completely sequenced and annotated (11 clones in total). Two and seven out of these 11 BAC clones are suggested to be coming from prokaryotes related to the Gamma- and Alphaproteobacteria, respectively, based on top BLAST hits criteria (see Tables S1–S6) and previously published information [8]. Based on homology searches, we were able to assign BAC clone MED49C08 from one of the gammaproteobacterial groups to the SAR86 clade; thus, 14 other BAC clones with almost identical PR genes (Figure 1) were also considered as members of this group (assuming no lateral gene transfer in the case of PR). Three of the retrieved BAC clones (MED86H08, MEDPR45, and MED42A11) are predicted to be from the SAR11 group because they carry PR genes with high sequence homology to a PR recently identified by proteome analysis of a cultured alphaproteobacterium (SAR11) [11], and data from other genes on the BACs support alphaproteobacterial affiliations. The high abundance of genome fragments from SAR86 and Alphaproteobacteria found here is consistent with previous reports, which determined members of the SAR86 clade to account for up to 8% of the active bacteria in the photic zone of a coastal North Sea sample [3] while SAR11 members were found to represent as much as 50% of the total marine surface water microbial community [12]. Based on 16S rRNA surveys, both the SAR86 and SAR11 clades harbor very diverse populations [13]. This “microdiversity” is also reflected on the PR level (Figure 1). All PR representatives (Alphaproteobacteria MED18B02, MED46A06, MED66A3; Gammaproteobacteria MED49C08; and unassigned group MED13K09 and MED82F10) checked using the E. coli heterologous expression system showed light-driven proton pumping activity as well as fast photocycles typical of retinylidene transporters [14] (Figure 2). The photochemical reaction cycles observed are among the most rapid seen for proton-pumping rhodopsins. Of interest is that the pigments exhibiting blue absorption spectra (MED18B02, MED49C08, and MED13K09) have fast photocycles indicative of efficient proton pumps operating in a high solar radiation environment as found in surface water (12-m depth) from which the BAC library was prepared. In contrast, the only previously characterized blue absorbing PR, HOT75 [15], has an order-of-magnitude slower photocycle. This was previously attributed to its retrieval from 75-m depth, where solar flux intensities are greatly reduced [15]. Taken together, these data imply that the widespread marine SAR86 and SAR11 groups, as well as other bacterial groups, are using light-driven PR-based phototrophy as a way to harvest additional energy in oligotrophic marine environments.
Figure 2 Laser-Flash-Induced Absorbance Changes in Suspensions of E. coli Membranes Containing PR Proteins
(A–E) PR proteins are from BACs MED46A06, MED66A03, MED18B02, MED49C08, and MED13K09, respectively.
A 532-nm pulse (6-ns duration, 40 mJ) was delivered at time 0, and absorption changes were monitored at wavelengths near the absorption maximum of the main absorption band in the visible range of the unphotolyzed pigment (520 nm for A and B, 480 nm for C–E) and the final photointermediate (the O intermediate) which is the longest-lived species in each of the photochemical reaction cycles (620 nm for A and B, 580 nm for C–E). 150–2,000 transients were collected at one to two flashes/sec and averaged for each trace as previously described [39]. The bar in each panel indicates the scale of the absorption change (× 10−3). Panel E exhibits greater noise because of the lower amplitudes of absorption changes due to lower expression level of the pigment.
Insets: E. coli membranes containing PR apoproteins in 50 mM Tris-HCl (pH 9.0) were reconstituted with an ethanolic solution of 2 μM all-trans retinal. The retinal-reconstitution of PR pigments were recorded in a Cary 4000 spectrophotometer at room temperature. The spectra were taken 40 min after retinal addition, which produced between 0.035 to 0.078 absorption units at the absorption maxima indicated.
Several interesting operons providing new insights into the metabolisms of PR-encoding microorganisms were linked to PR genes or found on PR-containing BACs. On clone MED13K09, an entire dsr operon containing the genes for both subunits of a reverse siroheme sulfite reductase (dsrAB), typically used by chemotrophic or anaerobic phototrophic bacteria for exploiting reduced sulfur compounds as electron donor [16,17], was found. The reverse sulfite reductase encoded on this BAC clone forms a highly supported monophyletic cluster with nine reverse sulfite reductases for which genes (or gene fragments) were retrieved from the Sargasso Sea shotgun library [9] and with the respective enzyme of the anaerobic phototrophic purple sulfur bacterium Allochromatium vinosum [18] (Figure 3A), a member of the Gammaproteobacteria. This grouping is further supported by a highly conserved gene order of other dsr genes on the genome fragments (Figure 3B). Furthermore, some but not all phylogenetic analyses of three ribosomal proteins encoded on the genome fragment from BAC clone MED13K09 also suggest that the organism is a deep-branching gammaproteobacterium (Figure S1).
Figure 3 Dsr Operons in PR-Carrying BAC and Sargasso Sea Scaffolds
(A) Phylogenetic tree showing the affiliation of DsrAB from MED13K09. Alignment regions of insertions and deletions were omitted in DsrAB amino acid sequence analyses. Polytomic nodes connect branches for which a relative order could not be determined unambiguously by using distance-matrix (FITCH with the Dayhoff PAM matrix, global rearrangements, and randomized input order of species), maximum-parsimony, and maximum-likelihood (with JTT-f as the amino acid replacement model) methods. Maximum-parsimony bootstrap values (%) are indicated at each node (1,000 re-samplings). The bar represents 10% sequence divergence as estimated from distance-matrix analysis. α, Alphaproteobacteria; β, Betaproteobacteria; γ, Gammaproteobacteria. In total, nine Sargasso Sea shotgun clones contained complete (IBEA_CTG_1982486, AACY01045584; IBEA_CTG_2027414, AACY01063972) or partial (IBEA_CTG_UAAO864TF, AACY01493489; IBEA_CTG_SSBMN57TR, AACY01327066; IBEA_CTG_SKBEW15TR, AACY01199346; IBEA_CTG_2002781, AACY01059482; IBEA_CTG_1960714, AACY01122073; IBEA_CTG_2018072, AACY01005285; IBEA_CTG_UAAYT68TR, AACY01523913) dsrAB sequences that formed a monophyletic cluster with MED13K09 and A. vinosum. Whole-genome shotgun sequence data for Thiobacillus denitrificans, Magnetospirillum magnetotacticum, and Magnetococcus sp. MC-1 were produced by the US Department of Energy Joint Genome Institute (http://www.jgi.doe.gov/). The yet-uncompleted genome sequence of T. denitrificans contains a frame shift in dsrB. Dissimilatory (bi)sulfite reductase sequences of sulfate-/sulfite reducers were taken from Wagner et al.[40], Klein et al.[41], and Zverlov et al. [42]].
(B) Organization of the dsr operons on MED13K09, Sargasso Sea shotgun clones IBEA_CTG_2027414 and IBEA_CTG_1982486, and in A. vinosum, Chlorobium tepidum TLS, and the sulfate-reducer Archaeoglobus fulgidus. Asterisk indicates an authentic frame shift in the second copy of dsrB in the genome of C. tepidum.
Since we could demonstrate that BAC MED13K09 is not a chimera (Figure S2), the close relationship of the reverse sulfite reductase from the PR-carrying MED13K09 clone with the enzyme of the gammaproteobacterium A. vinosum might suggest the existence of a novel anoxygenic phototroph exploiting light for energy generation not only by its bacteriochlorophyll-containing photosystem but also by PR. Alternatively, these genes might originate from a novel chemotrophic oxidizer of reduced sulfur compounds. In this context, it is interesting to note that some anoxygenic phototrophs [19] closely related to A. vinosum as well as thiobacilli [20], which both possess dsrAB genes [17] (Figure 3), are capable of gaining energy from aerobic oxidation of dimethyl sulfide to sulfate. In contrast to reduced inorganic sulfur compounds, dimethyl sulfide is present in the analyzed oxygenated marine surface waters [21], and PR- and DsrAB-exploiting marine bacteria might thus be involved in degradation of this compound, which plays the key role in the transport of sulfur from oceanic to terrestrial systems [22] and as a precursor for cloud condensation nuclei [23]. Together with the recent finding that SAR11 bacteria consume significant amounts of dimethylsulfoniopropionate [24], an osmoprotectant produced by marine algae and plant halophytes that is degraded by marine bacteria to DMS [25], our results suggest that bacteria exploiting PR phototrophy might be of importance for sulfur cycling in the marine photic zone.
Another interesting genomic feature linked to PR genes was a carotenoid biosynthesis gene cluster found on clones MED66A03, MED13K09, RED17H08, and MED82F10 (Figure 4 and Tables S1–S4). The arrangement of the respective genes was similar, containing the gene order crtIBY in all BACs. These genes are predicted to encode for phytoene desaturase, phytoene synthase, and lycopene cyclase, respectively, which catalyze the formation of β-carotene from geranylgeranyl pyrophosphate through phytoene and lycopene intermediates [26]. In addition, the first gene in the carotenoid biosynthesis pathway coding for geranylgeranyl diphosphate synthase (crtE) was found in the same operon in MED66A03, RED17H08. MED13K09 carries the crtE gene outside the operon approximately 25 kilobases downstream. This suggests that bacteria carrying these operons can synthesize β-carotene. Interestingly, the first reported bacterial gene coding for a homolog of the bacteriorhodopsin-related-protein-like homolog protein (Blh) from the archaeon Halobacterium sp. NRC-1 was found in the operons of MED66A03, RED17H08, and MED13K09, leading to the operonal arrangement of crtEIBY, blh on MED66A03, RED17H08 and crtIBY, blh on MED13K09. Bacteriorhodopsin-related protein was recently implicated in retinal biosynthesis [27] and was suggested to be the protein converting β-carotene to retinal, similar to the activity of 15,15′-β-carotene dioxygenase from Drosophila melanogaster [28]. Although highly speculative, as the identity between the archaeal Blh and the bacterial proteins is only 20%, this may imply that bacteria possessing PR apoproteins also carry the ability to synthesize the retinal chromophore and to potentially form functional PR holoproteins. Indeed, expression of the Blh homolog in β-carotene-producing E. coli cells resulted in the loss of the yellow color of these cells (Figure 4). When checked via HPLC, a clear all-trans retinal signal was seen only in cells expressing the Blh gene. Moreover, co-expression of the bacterial Blh homolog on a β-carotene-producing and PR-expressing E. coli background produced red-colored cells, indicating that the β-carotene is cleaved by the Blh homolog to retinal, which enters the membrane to form an active PR. The β-carotene cleaving enzyme Blh is the first one of its kind found in bacteria. The recently reported retinal biosynthetic enzyme from Synechocystis PCC 6803 [29] cleaves apo-carotenoids only (i.e., single-ringed carotenes), while the bacterial Blh cleaves β-carotene. In addition, a predicted gene encoding for isopentenyl diphosphate isomerase was found in the carotenoid biosynthetic operons containing the blh gene. This protein was shown to enhance isoprenoid biosynthesis when expressed in E. coli cells [30].
Figure 4 Retinal Biosynthesis Pathways in PR-Carrying BACs
(A) Schematic comparison of different carotenoid biosynthesis gene clusters linked to PR genes. ORF marked in gray represent predicted carotenoid biosynthesis genes while PR is marked in black.
(B) HPLC separation of the retinoids formed in the β-carotene producing E. coli and expressing the Blh protein. Left panel, extract from non-induced cells; right panel, after 60 min of induction (L-arabinose). Insights, absorption spectra of peaks 1 (β-carotene) and 2 (all-trans retinal).
(C) Color shift due to the cleavage of β-carotene to retinal in E. coli cells. Color shift from orange (β-carotene; non induced) to almost white (retinal; L-arabinose induced cells) in β-carotene producing and accumulating E. coli cells caused by expression of the blh gene and, the same β-carotene producing cells co-expressing the blh and a PR gene; color shift to red (L-arabinose and IPTG induced cells).
By taking advantage of large insert environmental BAC libraries and heterologous expression assays, we were able to show that PR-carrying bacteria are an important component of the microbial communities in the photic zone of the Mediterranean Sea and Red Sea, and that several phylogenetically diverse PR genes encode functional light-driven proton pumps. Furthermore, we revealed previously unrecognized links between PR genes and different and partly unexpected metabolic traits and thus gained novel insights into the biology of some uncultured PR-carrying bacteria. Some of these PR-carrying bacteria are apparently energy scavengers, ideally adapted to oligotrophic marine surface waters by exploiting not only light but possibly also some reduced organic sulfur compounds for energy generation.
Materials and Methods
BAC library construction
BAC libraries were constructed from plankton samples collected in the Red Sea or from 12-m water collected on a transect from Haifa to Cyprus (33°25′N, 33°56'E to 32°54'N, 34°44'E). BAC library construction was carried out as previously described [10] with minor modifications (for more details, see http://www.tigr.org/tdb/MBMO/MBMO.shtml). The approximately 800 l of pre-filtered waters (Whatman GF/A filter) (Middlesex, United Kingdom) were collected and filtered by tangential flow filtration with a Millipore (Billerica, Massachusetts, United States) Pellicon-2 unit equipped with a C screen Biomax (Martinsried, Germany) 30 polysulfone membrane cartridge (30,000 MW cutoff and 0.5 m2 cassette). Bacterioplankton cells were collected by centrifugation of the retentate (4°C, 38,900 × g, 1 h). The bacterioplankton pellet was embedded in agarose plugs, and DNA was extracted and cloned into the pBACindigo356 vector. The library consists of 10,560 clones with an average insert size of 80 Kb with a coverage of approximately 850 Mb.
BAC clones chosen for sequencing (black squares in Figure 1) were first screened by low coverage and then followed by a deeper coverage (BAC RED22E04: [estimated size 40 Kb] [72 reads, genomic coverage by trimmed read bases of quality ≥ 20 is 1.06×], 13 contigs assembled that sum to 21 Kb; BAC MED42A11: [estimated size 75 Kb] [458 reads, genomic coverage by trimmed read bases of quality ≥ 20 is 5.10×], eight contigs assembled that sum to 14.5 Kb; MED86H08: [estimated size 35Kb] [45 reads, genomic coverage by trimmed read bases of quality ≥ 20 is 1.04×], six contigs assembled summing to 16 Kb; BAC MED18B02: [estimated size 35Kb], 4 contigs assembled that sum to 32.5 Kb, [364 reads, genomic coverage by trimmed read bases of quality ≥ 20 is 6.12×]; BAC MED46A06: [estimated size 70 Kb], 1 contigs assembled that sum to 69.2 Kb [645 reads, genomic coverage by trimmed read bases of quality ≥ 20 is 7.65×]). BAC clones listed in Tables S1–S6 were completely sequenced (BACs MED13K09 [938 reads, genomic coverage by trimmed read bases of quality ≥ 20 is 6.78×], MED66A03 [621 reads, genomic coverage by trimmed read bases of quality ≥ 20 is 7.48×], MED49C08 [342 reads, genomic coverage by trimmed read bases of quality ≥ 20 is 4.14×], MED35C06 [256 reads, genomic coverage by trimmed read bases of quality ≥ 20 is 4.10×] and RED17H08 [1,800 reads, genomic coverage by trimmed read bases of quality ≥ 20 is 9.4×]) or sequenced to approximately 8× coverage (BAC MED82F10 [1,009 reads, genomic coverage by trimmed read bases of quality ≥ 20 is 8.81×; 11 contigs were assembled that sum to 82 Kb, BAC MED82F10 insert size is approximately 85 Kb]). All BAC insert size estimations were based on NotI restriction digest and CHEF pulse field gel electrophoresis separation, 13 h at 12 °C in 0.5 × TBE buffer at 6 V cm−1 with 5–15-s pulses. Sequencing of subclone shotgun libraries from the different BACs were performed by Macrogen (Seoul, Korea). Sequencing reads were trimmed of vector using SEQUENCHER 4.1.2 software (Gene Codes, Ann Arbor, Michigan, United States) and further trimmed manually after chromatograms were inspected for ambiguities. Trimmed and edited sequences were assembled using SEQUENCHER 4.1.2. Additional ambiguities were corrected by examining chromatograms manually, and a consensus for base calling was determined. Gap filling was performed by primer walking or PCR amplification using gap end sequences as primers and sequencing of the PCR product. Direct BAC end-sequencing reads were used to confirm the assembly of the different BACs.
PR detection in BAC libraries
PR presence in the BAC libraries was detected via Southern hybridizations or by multiplex PCR [31]. Southern hybridizations were performed essentially according to [8] using a mix of different PR genes (eBAC31A08, PalE6, medA15r8ex4, and NA13r8_9) as probes. PR detection via PCR was achieved by using degenerated primers 5′-MGNTAYATHGAYTGGYT-3′, 5′-WWNMGNTAYGTNGAYTGG-3′, and 5′- GGRTADATN
GCCCANCC-3′ targeting the conserved RYIDWL, [L/Y/F/Y]RYVDW and GWAIYP regions, respectively
PR expression and activity
PR representatives (MED13K09, MED18B02, MED46A06, MED49C08, MED66A3, and MED82F10) from different PR groups were expressed in E. coli cells using the pBAD expression system [1], and their light-driven proton-pumping activity was measured as previously described [32].
β-carotene dioxygenase activity
XLI-Blue E. coli cells transformed with pBCAR [33] with the crtE, crtB, crtI, and crtY genes for β-carotene biosynthesis from Erwinia herbicola, pGB-Ipi carrying ipi (IPP isomerase to DMAPP) from Haematococcus
pluvialis [34] and plasmid pBAD-Blh carrying the blh gene under the arabinose promoter were grown overnight at 37 °C in the dark to early stationary phase. Bacteria were harvested from samples of 10 ml of culture for carotenoids and retinoids analysis at time 0, 1, 2, and 6 h after addition of 0.1% (w/v) L-arabinose. Uninduced cells were harvested at 6 h of growth. Cells were resuspended in 200 μl of 6M formaldehyde and incubated for 2 min at 37 °C. Two ml of dichloromethane were added, and carotenoids and retinoids were extracted twice with 4 ml of hexane. The solvent was dried under a stream of nitrogen and the carotenoids dissolved in 75 μl hexan:ethanol 99.5:0.5 to be injected to the HPLC. Carotenoids and retinoids were separated by HPLC using a Waters (Milford, Massachusetts, United States) system and a Spherisorb ODS2 C18 (5 μm, 4.6 × 250 mm) reversed-phase column. Samples of 25 μl were injected to a Waters 600 pump. A gradient of acetonitrile:water (9:1) containing 0.1% (w/v) ammonium acetate (A) and ethylacetate (B), at a constant flow rate of 1.6 ml/min was used as follows: 100% A during the first 15 min; 100% to 80% A during 8 min; 80% to 65% A during 4 min, followed by 65% to 45% A during 14 min and a final segment at 100% B. Light absorption peaks were detected in the range of 200–600 nm using a Waters 996 photodiode array detector. Carotenoids and retinoids were identified by their absorption spectra and characteristic retention time. Pure all-trans-retinal and pure β-carotene were used as standards.
Alternatively, for the co-expression with the PR, Blh homolog from MED66A03 was amplified using the BLH66A3fwd 5′-
ACCATGGGTGGCTTGATGTTAATTGATTGGTG-3′ and BLH66A3rev 5′-
ATTTTTGATTTTAATTCTGGAAGAGTGTGGTC-3′ primers, cloned into the pBAD-TOPO (Invitrogen, Carlsbad, California, United States) expression vector and transformed into β-carotene accumulating E. coli cells carrying plasmid p
ACCAR16ΔcrtX with the crtE, crtB, crtI, and crtY genes for β-carotene biosynthesis from E. herbicola [26]. For the co-expression with the PR gene, the blh gene was cut out using NcoI and PmeI restriction enzymes and cloned into a pBAD-TOPO-derived plasmid carrying the 31A08 PR gene under the control of the lacUV-promoter [K.-H. Jung, V. D. Trivedi, E. N. Spudich, J. L. Spudich, unpublished data]. β-carotene-accumulating E. coli cells were grown with L-arabinose and IPTG to induce the blh and PR genes, respectively.
Supporting Information
Figure S1 Phylogenetic Analysis of Ribosomal Proteins L21, L27, and S20 from BAC Clone MED13K09
Ribosomal protein L31, which is also present on BAC clone MED13K09, was excluded from the analysis because lateral gene transfer of this protein has been reported [35]. The dataset consisted of 93 reference organisms, which represent all bacterial genera for which whole genome sequences have been reported. A concatenated dataset and a 30% amino acid sequence conservation filter (234 alignment positions) was used for phylogeny inference. Polytomic nodes connect branches for which a relative order could not be determined unambiguously by using distance-matrix, maximum-parsimony, and maximum-likelihood methods. In contrast to the consensus tree, trees inferred by distance matrix (DM) and maximum-likelihood (ML) methods do support a clustering of MED13K09 proteins with the Gamma-/Betaproteobacteria (see insets).
(690 KB JPG).
Click here for additional data file.
Figure S2 BAC Clone MED13K09 Is Not a Chimera
Schematic illustration showing that BAC clone MED13K09 is not a chimera and that the dsr genes identified are linked to the PR gene on the genome of the respective unknown marine bacterium. In addition to BAC clone MED13K09, a partially overlapping BAC clone (MED47G02) was detected by BAC end sequencing. This clone does also carry dsrA (>100% identity on DNA level) as demonstrated by PCR amplification and sequencing. Specific primer sets were designed and used to amplify overlapping 4-kilobase PCR fragments (shown in red) (using DNA isolated directly from the environment as a template), which demonstrate that the sequence region of MED13K09 identical to the 3′ end of MED47G02 is actually connected to the PR gene. DsrAB, crtE, PR, and BAC MED47G02 end positions relative to BAC MED13K09 are marked. In addition, a shotgun sequence scaffold from Sargasso Sea carrying dsr genes and a PR has been deposited by Venter et al.[9], providing independent evidence for co-occurrence of these genes on bacterial genomes.
(213 KB JPG).
Click here for additional data file.
Table S1 List of Genes on BAC Clone MED13K09
This clone contains four genes encoding ribosomal proteins (S20, L27, L21, L31). Based on these proteins, a phylogenetic analysis was performed (see Figure S1). Of the 100 ORFs annotated, 54%, 12%, and 34% were provisionally assigned based on the top BLAST hit to the Gammaproteobacteria, Alphaproteobacteria, and other prokaryotes, respectively.
(124 KB DOC).
Click here for additional data file.
Table S2 List of Genes on BAC Clone MED66A03
Of the 40 ORFs annotated, 15%, 50%, and 35% were provisionally assigned based on the top BLAST hit to the Gammaproteobacteria, Alphaproteobacteria, and other prokaryotes, respectively.
(60 KB DOC).
Click here for additional data file.
Table S3 List of Genes on BAC Clone RED17H08
Of the 38 ORFs annotated, 16%, 42%, and 42% were provisionally assigned based on the top BLAST hit to the Gammaproteobacteria,
Alphaproteobacteria, and other prokaryotes, respectively.
(62 KB DOC).
Click here for additional data file.
Table S4 List of Genes on BAC Clone MED82F10
Of the 18 ORFs annotated, 28%, 22%, and 50% were provisionally assigned based on the top BLAST hit to the Gammaproteobacteria,
Alphaproteobacteria, and other prokaryotes, respectively.
(41 KB DOC).
Click here for additional data file.
Table S5 List of Genes on BAC Clone MED49C08
aORF has highest homology to a protein from the SAR86-related environmental BAC clone EBAC31A08 [1]. Of the 67 ORFs annotated, 60%, 25%, and 15% were provisionally assigned based on the top BLAST hit to the Gammaproteobacteria, Alphaproteobacteria, and other prokaryotes, respectively.
(103 KB DOC).
Click here for additional data file.
Table S6 List of Genes on BAC Clone MED35C06
aORF has highest homology to a protein from the SAR86-related environmental BAC clone EBAC31A08 [1]. Of the 39 ORFs annotated, 77%, 13%, and 10% were provisionally assigned based on the top BLAST hit to the Gammaproteobacteria, Alphaproteobacteria, and other prokaryotes, respectively.
(58 KB DOC).
Click here for additional data file.
Accession Numbers
The GenBank (http://www.ncbi.nlm.nih.gov/Genbank) accession numbers for the genes and gene products discussed in this paper are: A. vinosum (U84760); Archaeoglobus fulgidus (NC_000917); BAC sequences (BAC MED13K09–DQ068067, BAC MED18B02 [DQ088853–DQ088855], BAC MED35C06–DQ077553, BAC MED42A11 [DQ088869–DQ088876], BAC MED46A06–DQ088847, BAC MED49C08–DQ077554, BAC MED66A03–DQ065755, BAC MED82F10–DQ073796, BAC MED86H08 [DQ088848–DQ088852], BAC RED17H08–DQ068068, BAC RED22E04 [DQ088856–DQ088868]); Chlorobium tepidum TLS (NC_002932); dsr (IBEA_CTG_1960714, AACY01122073; IBEA_CTG_1982486, AACY01045584; IBEA_CTG_1982486; IBEA_CTG_2002781, AACY01059482; IBEA_CTG_2018072, AACY01005285; IBEA_CTG_2027414, AACY01063972; IBEA_CTG_SKBEW15TR, AACY01199346; IBEA_CTG_SSBMN57TR, AACY01327066; IBEA_CTG_UAAO864TF, AACY01493489; IBEA_CTG_UAAYT68TR, AACY01523913); Magnetococcus sp. MC-1 (NZ_AAAN02000009); Magnetospirillum magnetotacticum (NZ_AAAP01003703); PR (AY599897–AY599909, CH024977, DQ062244–DQ062293); and Thiobacillus denitrificans (NZ_AAFH01000002).
We thank G. D. Eytan for helping in constructing the Mediterranean BAC library, M. Taylor for help with ribosomal protein phylogeny, E. F. DeLong for consulting and moral support, M. Simon and C. Dahl for valuable discussion, and N. Misawa for kindly providing plasmid p
ACCAR16DcrtX. This research was supported in part by the Human Frontiers Science Program P38/2002 (OB, JLS), Israel Science Foundation grant 434/02 (OB), Welch Foundation (JLS), the BIOLOG2-program of the bmb+f (MW), and a Marie Curie Intra-European fellowship (VENTSULFURMICDIV) to AL.
Competing Interests. The authors have declared that no competing interests exist.
Author contributions. GS, AL, JLS, JH, and OB conceived and designed the experiments. GS, AL, RP, TI, and JH performed the experiments. GS, AL, RP, JLS, TI, JH, MW, and OB analyzed the data. GS, KHJ, JLS, and JH contributed reagents/materials/analysis tools. AL, JLS, JH, MW, and OB wrote the paper.
¤ Current address: Department of Life Science, Interdisciplinary Program of Integrated Biotechnology, Sogang University, Seoul, Korea
Citation: Sabehi G, Loy A, Jung KH, Partha R, Spudich JL, et al. (2005) New insights into metabolic properties of marine bacteria encoding proteorhodopsins. PLoS Biol 3(8): e273.
Abbreviations
BACbacterial artificial chromosome
Blhbacteriorhodopsin-related-protein-like homolog protein
PRproteorhodopsin
==== Refs
References
Béjà O Aravind L Koonin EV Suzuki MT Hadd A Bacterial rhodopsin: Evidence for a new type of phototrophy in the sea Science 2000 289 1902 1906 10988064
Béjà O Spudich EN Spudich JL Leclerc M DeLong EF Proteorhodopsin phototrophy in the ocean Nature 2001 411 786 789 11459054
Pernthaler A Pernthaler J Schattenhofer M Amann R Identification of DNA-synthesizing bacterial cells in coastal North Sea plankton Appl Environ Microbiol 2002 68 5728 5736 12406771
Pernthaler A Pernthaler J Amann R Fluorescence in situ hybridization and catalyzed reporter deposition for the identification of marine bacteria Appl Environ Microbiol 2002 68 3094 3101 12039771
Man D Wang W Sabehi G Aravind L Post AF Diversification and spectral tuning in marine proteorhodopsins EMBO J 2003 22 1725 1731 12682005
Sabehi G Massana R Bielawski JP Rosenberg M DeLong EF Novel proteorhodopsin variants from the Mediterranean and Red Seas Environ Microbiol 2003 5 842 849 14510837
Sabehi G Béjà O Suzuki MT Preston CM DeLong EF Different SAR86 subgroups harbour divergent proteorhodopsins Environ Microbiol 2004 6 903 910 15305915
de la Torre JR Christianson L Béjà O Suzuki MT Karl D Proteorhodopsin genes are widely distributed among divergent bacterial taxa Proc Natl Acad Sci U S A 2003 100 12830 12835 14566056
Venter JC Remington K Heidelberg J Halpern AL Rusch D Environmental genome shotgun sequencing of the Sargasso Sea Science 2004 304 66 74 15001713
Béjà O Suzuki MT Koonin EV Aravind L Hadd A Construction and analysis of bacterial artificial chromosome libraries from a marine microbial assemblage Environ Microbiol 2000 2 516 529 11233160
Stapels MD Cho JC Giovannoni SJ Barofsky DF Proteomic analysis of novel marine bacteria using MALDI and ESI mass spectrometry J Biomol Tech 2004 15 191 198 15331585
Morris RM Rappé MS Connon SA Vergin KL Siebold WA SAR11 clade dominates ocean surface bacterioplankton communities Nature 2002 420 806 810 12490947
Fuhrman JA Community structure and function in prokaryotic marine plankton Antonie van Leeuwenhoek 2002 81 521 527 12448747
Spudich JL Yang CS Jung KH Spudich EN Retinylidene proteins: Structures and functions from archaea to humans Annu Rev Cell Dev Biol 2000 16 365 392 11031241
Wang WW Sineshchekov OA Spudich EN Spudich JL Spectroscopic and photochemical characterization of a deep ocean proteorhodopsin J Biol Chem 2003 278 33985 33991 12821661
Hipp WM Pott AS Thum-Schmitz N Faath I Dahl C Towards the phylogeny of APS reductases and sirohaem sulfite reductases in sulfate-reducing and sulfur-oxidizing prokaryotes Microbiology 1997 143 2891 2902 9308173
Dahl C Rakhely G Pott-Sperling AS Fodor B Takacs M Genes involved in hydrogen and sulfur metabolism in phototrophic sulfur bacteria FEMS Microbiol Lett 1999 180 317 324 10556728
Pott AS Dahl C Sirohaem sulfite reductase and other proteins encoded by genes at the dsr locus of Chromatium vinosum are involved in the oxidation of intracellular sulfur Microbiology 1998 144 1881 9695921
Jonkers HM Jansen M Van der Maarel MJEC Van Gemerden H Aerobic turnover of dimethyl sulfide by the anoxygenic phototrophic bacterium Thiocapsa roseopersicina
Arch Microbiol 1999 172 150 156 10460885
Visscher PT Taylor BF A new mechanism for the aerobic catabolism of dimethyl sulfide Appl Env Microbiol 1993 59 3784 3789 8285684
Liss P Take the shuttle—from marine algae to atmospheric chemistry Science 1999 285 1217 1218
Bates TS Lamb BK Guenther A Digon J Stoiber RE Sulfur emissions to the atmosphere from natural sources J Atmos Chem 1992 14 315 337
Charlson RJ Lovelock JE Andreae MO Warren SG Oceanic phytoplankton, atmospheric sulphur, cloud albedo and climate Nature 1987 326 655 661
Malmstrom RR Kiene RP Cottrell MT Kirchman DL Contribution of SAR11 bacteria to dissolved dimethylsulfoniopropionate and amino acid uptake in the North Atlantic Ocean Appl Env Microbiol 2004 70 4129 4135 15240292
Yoch DC Dimethylsulfoniopropionate: Its sources, role in the marine food web, and biological degradation to dimethylsulfide Appl Env Microbiol 2002 68 5804 5815 12450799
Misawa N Satomi Y Kondo K Yokoyama A Kajiwara S Structure and functional analysis of a marine bacterial carotenoid biosynthesis gene cluster and astaxanthin biosynthetic pathway proposed at the gene level J Bacteriol 1995 177 6575 6584 7592436
Peck RF Echavarri-Erasun C Johnson EA Ng WV Kennedy SP brp and blh are required for synthesis of the retinal cofactor of bacteriorhodopsin in halobacterium salinarum
J Biol Chem 2001 276 5739 5744 11092896
von Lintig J Vogt K Filling the gap in vitamin A research. Molecular identification of an enzyme cleaving beta-carotene to retinal J Biol Chem 2000 275 11915 11920 10766819
Ruch S Beyer P Ernst H Al Babili S Retinal biosynthesis in Eubacteria: In vitro characterization of a novel carotenoid oxygenase from Synechocystis sp. PCC 6803 Mol Microbiol 2005 55 1015 1024 15686550
Kajiwara S Fraser PD Kondo K Misawa N Expression of an exogenous isopentenyl diphosphate isomerase gene enhances isoprenoid biosynthesis in Escherichia coli
Biochem J 1997 324 421 426 9182699
Stein JL Marsh TL Wu KY Shizuya H DeLong EF Characterization of uncultivated prokaryotes: Isolation and analysis of a 40-kilobase-pair genome fragment from a planktonic marine archaeon J Bacteriol 1996 178 591 599 8550487
Man-Aharonovich D Sabehi G Sineshchekov OA Spudich EN Spudich JL Characterization of RS29, a blue-green proteorhodopsin variant from the Red Sea Photochem Photobiol Sci 2004 3 459 462 15122363
Isaacson T Ohad I Beyer P Hirschberg J Analysis in vitro of the enzyme CRTISO establishes a poly-cis carotenoid biosynthesis pathway in plants Plant Physiol 2004 136 4246 4255 15557094
Lotan T Hirschberg J Cloning and expression in Escherichia coli of the gene encoding b-C-4-oxygenase, that converts β-carotene to the ketocarotenoid canthaxanthin in Haematococcus pluvialis
FEBS Lett 1995 364 125 128 7750556
Makarova KS Ponomarev VA Koonin EV Two C or not two C: Recurrent disruption of Zn-ribbons, gene duplication, lineage-specific gene loss, and horizontal gene transfer in evolution of bacterial ribosomal proteins Genome Biol 2001 2 research/0033.0033
Altschul SF Madden TL Schaffer AA Zhang J Zhang Z Gapped BLAST and PSI-BLAST: A new generation of protein database search programs Nucleic Acids Res 1997 25 3389 3402 9254694
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
Swofford DL PAUP*. Phylogenetic Analysis Using Parsimony (*and other methods). 4.0b10 ed 2002 Sunderland (Massachusetts) Sinauer Associates
Sasaki J Spudich JL The transducer protein HtrII modulates the lifetimes of sensory rhodopsin II photointermediates Biophys J 1998 75 2435 2440 9788938
Wagner M Roger AJ Flax JL Brusseau GA Stahl DA Phylogeny of dissimilatory sulfite reductases supports an early origin of sulfate respiration J Bacteriol 1998 180 2975 2982 9603890
Klein M Friedrich M Roger AJ Hugenholtz P Fishbain S Multiple lateral transfers of dissimilatory sulfite reductase genes between major lineages of sulfate-reducing prokaryotes J Bacteriol 2001 183 6028 6035 11567003
Zverlov V Klein M Lucker S Friedrich MW Kellermann J Lateral gene transfer of dissimilatory (bi)sulfite reductase revisited J Bacteriol 2005 187 2203 2208 15743970
| 16008504 | PMC1175822 | CC BY | 2021-01-05 08:21:26 | no | PLoS Biol. 2005 Aug 19; 3(8):e273 | utf-8 | PLoS Biol | 2,005 | 10.1371/journal.pbio.0030273 | oa_comm |
==== Front
PLoS BiolPLoS BiolpbioplosbiolPLoS Biology1544-91731545-7885Public Library of Science San Francisco, USA 10.1371/journal.pbio.0030284SynopsisEvolutionMolecular Biology/Structural BiologyBiochemistryEubacteriaReducing the Mysteries of Sulfur Metabolism Synopsis8 2005 19 7 2005 19 7 2005 3 8 e284Copyright: © 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 Conserved Mechanism for Sulfonucleotide Reduction
==== Body
Sulfur is one of life's essential atoms. It is found in two amino acids—methionine and cysteine—and in several vitamins. Outside of living organisms, sulfur is mostly found in its fully oxidized form, the sulfate ion (SO4
2−). To incorporate it into biomolecules, enzymes must reduce it, stripping it of oxygens. As is the case with so many other metabolic pathways, microorganisms possess a richer and more varied sulfur metabolism than humans do. In particular, bacteria, including the human pathogen Mycobacterium tuberculosis, use enzymes called sulfonucleotide reductases to begin the reduction process, creating reduced sulfite (SO3
2−). In this issue of PLoS Biology, Carolyn Bertozzi and colleagues elucidate the novel molecular mechanism at the heart of this reaction, and show that it is conserved among a wide variety of organisms, from bacteria to plants.
The process begins when sulfate links up with an ATP molecule to form a species, abbreviated APS. The authors used a variety of biochemical techniques to discover what happens next. When APS was incubated with one kind of sulfonucleotide reductase, called APS reductase, they found that the weight of the enzyme increased by 80 Daltons, exactly the weight of a covalently bound sulfite, suggesting that the sulfur of the APS had been reduced and linked to the enzyme. By mutating the enzyme, they showed that the sulfite links to a critical cysteine amino acid, which itself contains a sulfur (sulfur–sulfur covalent bonds are common in biochemical molecules). To regenerate the enzyme and liberate the sulfite, yet another sulfur-containing compound, thioredoxin, joins the action, the result of which is that sulfite is released and the enzyme is restored to its original form, ready to react again.
Researchers deduced the mechanism of the highly conserved sulfonucleotide reductase enzyme in Mycobacterium tuberculosis pictured here (Image: US Centers for Disease Control and Prevention)
Analysis of the same pathway in two other types of bacteria confirmed that this same mechanism occurs in each. Combined with previous observations of a similar mechanism in the model plant Arabidopsis, these results demonstrate a remarkable bit of functional conservation over many millions of years of evolution. As well as elucidating an important piece of sulfur biochemistry, the deeper understanding of this critical biochemical process provided by this study may have practical implications. Since humans do not possess this pathway, drugs that target it might make an effective antibiotic with a low risk for side effects.
| 0 | PMC1175823 | CC BY | 2021-01-05 08:21:24 | no | PLoS Biol. 2005 Aug 19; 3(8):e284 | utf-8 | PLoS Biol | 2,005 | 10.1371/journal.pbio.0030284 | oa_comm |
==== Front
PLoS BiolPLoS BiolpbioplosbiolPLoS Biology1544-91731545-7885Public Library of Science San Francisco, USA 10.1371/journal.pbio.0030286SynopsisEcologyEvolutionPaleontologySpace Matters: Estimating Species Diversity in the Fossil Record Synopsis8 2005 19 7 2005 19 7 2005 3 8 e286Copyright: © 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.
The Impact of the Species-Area Relationship on Estimates of Paleodiversity
==== Body
To get a sense of the challenges in measuring biodiversity, consider this: estimates for the number of living species on earth range from 3.5 million to over 30 million. Only 1.9 million species have been classified and described. (Of these, 15,589 currently face extinction.) Now imagine trying to calculate patterns of biodiversity through the paleontological record. One tool ecologists rely on to identify patterns of biological diversity is a long-established rule of thumb called the species–area effect: the tendency for species number, or richness, to increase in a regular way with area.
Paleontologists typically have been unable to apply the species–area rule in estimating paleodiversity and instead use species counts for a given time interval to calculate historical biodiversity, with the assumption that other sampling considerations override the species–area effect. In a new study, Anthony Barnosky, Marc Carrasco, and Edward Davis test this assumption and discover that the golden rule of ecology holds for the rock record as well. Just as geographic sampling influences diversity counts in the modern landscape, the species–area effect influences counts in the fossil record.
To get a true picture of geographic conditions, Barnosky et al. used mapping and imaging systems that generate direct measures of the geography for a given set of fossil species. To get a sense of diversity across time and space, the authors used a recently completed archival database (which they also built) that integrates the geographic data with fossil datasets, called the Miocene Mammal Mapping Project (MIOMAP). MIOMAP includes all western North American mammals from 5–30 million years ago—3,100 localities and 14,000 occurrences of species in all.
The authors first tested the fossil data for species–area effects with species from a time period with robust geographic data (the Early Barstovian, about 14.8–15.9 million years ago). They plotted species richness against geographic area, using what's known as nested sets of fauna (species that occur in a smaller area represent a subset of species in a larger area) by starting with one biogeographic region and successively adding species from others. They also used unnested sets of fauna to plot species richness within a given time period for nine different geographic regions against the geographic sampling area. After correcting for possible biases in sample size that might influence the number of species, Barnosky et al. found a strong species–area effect in both analyses.
These results, they argue, suggest that many fluctuations in diversity seen in fossil analyses actually arise from the species–area effect. Given the lack of uniform geographic sampling in paleontological data, the impact of this effect may be significant—and likely applies to other taxa as well. Once the effect is factored in, one might expect significant adjustments in accepted patterns of global and regional paleodiversity. And because an important metric for understanding current extinctions relies on descriptions of past extinction events, controlling for a paleodiversity–area effect may provide a better frame of reference for understanding the current biodiversity crisis. Estimates of paleodiversity also have important evolutionary implications for understanding how and when new species emerge. Thanks to the innovative text-mining tools and approach presented here, future studies can more easily correct for area effects and explore these issues. And given the parallels between species–area relationships in paleontology and ecology, collaborations across disciplines may offer valuable insights into ecological dynamics through time.
A new, comprehensive database compiles mammalian fossils including this upper jaw of the Sthenicitis campestris, a weasel from about 12 million years ago (Photo: Alan B. Shabel)
| 0 | PMC1175824 | CC BY | 2021-01-05 08:21:24 | no | PLoS Biol. 2005 Aug 19; 3(8):e286 | utf-8 | PLoS Biol | 2,005 | 10.1371/journal.pbio.0030286 | oa_comm |
==== Front
PLoS BiolPLoS BiolpbioplosbiolPLoS Biology1544-91731545-7885Public Library of Science San Francisco, USA 10.1371/journal.pbio.0030286SynopsisEcologyEvolutionPaleontologySpace Matters: Estimating Species Diversity in the Fossil Record Synopsis8 2005 19 7 2005 19 7 2005 3 8 e286Copyright: © 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.
The Impact of the Species-Area Relationship on Estimates of Paleodiversity
==== Body
To get a sense of the challenges in measuring biodiversity, consider this: estimates for the number of living species on earth range from 3.5 million to over 30 million. Only 1.9 million species have been classified and described. (Of these, 15,589 currently face extinction.) Now imagine trying to calculate patterns of biodiversity through the paleontological record. One tool ecologists rely on to identify patterns of biological diversity is a long-established rule of thumb called the species–area effect: the tendency for species number, or richness, to increase in a regular way with area.
Paleontologists typically have been unable to apply the species–area rule in estimating paleodiversity and instead use species counts for a given time interval to calculate historical biodiversity, with the assumption that other sampling considerations override the species–area effect. In a new study, Anthony Barnosky, Marc Carrasco, and Edward Davis test this assumption and discover that the golden rule of ecology holds for the rock record as well. Just as geographic sampling influences diversity counts in the modern landscape, the species–area effect influences counts in the fossil record.
To get a true picture of geographic conditions, Barnosky et al. used mapping and imaging systems that generate direct measures of the geography for a given set of fossil species. To get a sense of diversity across time and space, the authors used a recently completed archival database (which they also built) that integrates the geographic data with fossil datasets, called the Miocene Mammal Mapping Project (MIOMAP). MIOMAP includes all western North American mammals from 5–30 million years ago—3,100 localities and 14,000 occurrences of species in all.
The authors first tested the fossil data for species–area effects with species from a time period with robust geographic data (the Early Barstovian, about 14.8–15.9 million years ago). They plotted species richness against geographic area, using what's known as nested sets of fauna (species that occur in a smaller area represent a subset of species in a larger area) by starting with one biogeographic region and successively adding species from others. They also used unnested sets of fauna to plot species richness within a given time period for nine different geographic regions against the geographic sampling area. After correcting for possible biases in sample size that might influence the number of species, Barnosky et al. found a strong species–area effect in both analyses.
These results, they argue, suggest that many fluctuations in diversity seen in fossil analyses actually arise from the species–area effect. Given the lack of uniform geographic sampling in paleontological data, the impact of this effect may be significant—and likely applies to other taxa as well. Once the effect is factored in, one might expect significant adjustments in accepted patterns of global and regional paleodiversity. And because an important metric for understanding current extinctions relies on descriptions of past extinction events, controlling for a paleodiversity–area effect may provide a better frame of reference for understanding the current biodiversity crisis. Estimates of paleodiversity also have important evolutionary implications for understanding how and when new species emerge. Thanks to the innovative text-mining tools and approach presented here, future studies can more easily correct for area effects and explore these issues. And given the parallels between species–area relationships in paleontology and ecology, collaborations across disciplines may offer valuable insights into ecological dynamics through time.
A new, comprehensive database compiles mammalian fossils including this upper jaw of the Sthenicitis campestris, a weasel from about 12 million years ago (Photo: Alan B. Shabel)
| 0 | PMC1175825 | CC BY | 2021-01-05 08:21:25 | no | PLoS Biol. 2005 Aug 19; 3(8):e287 | latin-1 | PLoS Biol | 2,005 | 10.1371/journal.pbio.0030287 | oa_comm |
==== Front
PLoS BiolPLoS BiolpbioplosbiolPLoS Biology1544-91731545-7885Public Library of Science San Francisco, USA 10.1371/journal.pbio.0030290SynopsisNeuroscienceHomo (Human)Now You Don't See It, Now You Do: Filling In Creates the Illusion of Motion Synopsis8 2005 19 7 2005 19 7 2005 3 8 e290Copyright: © 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.
Primary Visual Cortex Activity Along the Apparent-Motion Trace Reflects Illusory Perception
==== Body
The parade of lights flashing on a theater marquee provides an important lesson in how the brain creates the illusion of motion. While we know each bulb remains stationary, the lighting and dimming of each in succession makes it appear that light is moving across the marquee. Even when successive bulbs are separated by a large space, our brains fill in the missing data to create the illusion that the motion has occurred smoothly from one point to the next. But where in the brain does the illusion occur? In this issue of PLoS Biology, Lars Muckli and colleagues demonstrate that the first cortical area in the visual processing stream, the primary visual cortex, participates in sustaining the illusion, probably under the influence of higher visual centers in the brain that likely create it.
In this study, human subjects observed a simple white square on a computer screen, flashed first in one spot, and then again several centimeters away. A functional magnetic resonance image of their brain activity was recorded, and the most active areas were determined. In the primary visual cortex, or V1, activity was found in one area that corresponded to the location in the visual field of the first flash, and in another area corresponding to the location of the second flash. Remarkably, there was also activity in between these two locations, corresponding to a region of the visual field that, while not itself illuminated, was on the path between the two flashes. This same area was active when the subjects viewed the white square moving smoothly between the two spots, and was absent when the square was flashed at only the initial or only the final spot, or when squares in both spots flickered simultaneously. The subjects' brains, it seems, filled in the missing information when a pattern of activity suggesting motion was detected. Further evidence that such “fill-in” activity in V1 corresponds to the conscious perception of motion came from an additional experiment, in which the flashing pattern of a quartet of white squares suggested alternating, and mutually exclusive, horizontal or vertical motion. When the observers reported the quartet as moving vertically, there was more activity in the V1 region corresponding to the vertical apparent motion.
The source of this filling in is not likely to be in V1 itself, the authors argue, because the middle area is too far from either of the ends to be stimulated by them directly. Instead, they propose the most likely source is activity in a visual area much higher in the processing chain, called hMT/V5+. Projections from here are known to influence V1 activity, and cover large enough areas to encompass the entire region of V1 activation in these experiments.
The brain regions highlighted above allow us to see a series of flashing images as smooth motion. This activity occurs in the primary visual cortex
These results provide further demonstration that the brain masterfully creates a continuous, yet sometimes imagined, whole from individual, and often incomplete, parts. Most surprisingly, even the brain areas originally thought to be “literalist” in their representation of the environment are actually accomplices in the construction of illusory views. While such synthetic activity occasionally leaves us prey to optical illusions, it's a small price to pay for what we get in return: a seamless understanding of often fractured perspectives.
| 0 | PMC1175826 | CC BY | 2021-01-05 08:21:26 | no | PLoS Biol. 2005 Aug 19; 3(8):e290 | utf-8 | PLoS Biol | 2,005 | 10.1371/journal.pbio.0030290 | oa_comm |
==== Front
PLoS BiolPLoS BiolpbioplosbiolPLoS Biology1544-91731545-7885Public Library of Science San Francisco, USA 10.1371/journal.pbio.0030295SynopsisEcologyEvolutionInfectious DiseasesMicrobiologyZoologyParasitologyInsectsPrevalence of Infection in a Population Can Shape Parasite Virulence Synopsis8 2005 19 7 2005 19 7 2005 3 8 e295Copyright: © 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.
Prevalence-Dependent Costs of Parasite Virulence
==== Body
If necessity is the mother of invention, the coevolutionary arms race is the mother of adaptation. For parasites and hosts engaged in an ongoing battle to gain advantage, those adaptations take many forms. A host continually tinkers with resistance mechanisms while the parasite adjusts its means of infection. But if infection causes too much damage to its host, the parasite could destroy its chances of transmission and, thus, survival. Thus, from the parasite's point of view, natural selection must create an optimal balance between the costs of parasitic infection (or virulence)—host death—and the benefits—eluding host defenses and establishing an infection.
Stephanie Bedhomme, Yannis Michalakis, and their colleagues study the coevolution of parasite virulence and host life history traits. In a new study, the researchers modify the standard approach to studying the costs of infection (comparing infected and uninfected groups) by introducing another variable: intraspecific competition between hosts. Specifically, they ask, does infection affect competitive interactions between individuals? As expected, the authors find that infected individuals pay a cost compared to their healthy counterparts. But surprisingly, both infected and uninfected individuals do better when their competitor is infected: parasite costs depend on the infection status of the competitors.
To study the interplay between parasitism and intraspecific competition, Bedhomme et al. worked with the yellow fever mosquito Aedes aegypti and its natural enemy, the single-celled parasite Vavraia culicis. In the wild, mosquitoes get V. culicis infections when larvae ingest spores of the parasite, which enter larval gut cells, where they develop and reproduce. Because the spores need water to survive, it's thought that successful transmission depends on larvae, which live in water, and on two host life history traits: larval developmental time and probability of emergence of the host.
By accounting for competition between infected and uninfected mosquitoes, researchers discovered a link between parasitic virulence and the prevalence of infection in a population
To study this process in the lab, Bedhomme et al. divided recently hatched mosquito larvae into groups of 60 larvae, and exposed half of the groups to the parasite. Larvae were then placed two by two into vials. Vials contained either two uninfected larvae, two infected larvae, or one infected and one uninfected individual.
Larvae were treated with high- and low-food diets. Competitive performance was measured by the probability of reaching adulthood and the time to develop. On the high-food diet, infected and uninfected larvae mostly grew normally and reached adulthood. On a low-food diet, however, infected larvae grew slowly and were less likely to reach adulthood—probably because the parasite had more time to kill them. Being paired with an infected versus uninfected partner did not influence this outcome.
As for time to develop, infected pairs took longer to develop than uninfected pairs, as expected. But with infected and uninfected pairs, infected larvae took longer to develop than their healthy partners, meaning they should be more likely to succumb to the parasite. Competing against a healthy partner increased virulence by increasing development time. Interestingly, infected mosquitoes also fared better when paired with an infected competitor.
These results suggest that a high incidence, or prevalence, of parasitic infection in the population means that healthy larvae face less competition and do better than they would if they had to compete with healthy individuals. Infected individuals will also do better if there's a high prevalence of infection because they are more likely to compete against equally poor competitors. Thus, by ignoring the effects of competition, standard models underestimate the full costs of virulence—and, more important, miss a significant link between a parasite's prevalence in a population and its virulence.
Because the virulence of the parasite varies with its distribution in the population, this phenomenon may affect the population dynamics of parasite and host: infected hosts can't compete as well and take longer to develop, increasing the parasite's chances of transmission. The evolutionary implications of this relationship may be a selective pressure for host resistance to parasites. In traditional models, in low-virulence conditions, resistance is selected against because the cost of resistance reduces the fitness of the uninfected individual. But when low prevalence is associated with high virulence, the benefits of resistance would increase the individual's fitness. For an infected mosquito, at least, you stand a better chance of getting your wings and leaving the natal lagoon if more of your larval neighbors are infected too.
| 0 | PMC1175827 | CC BY | 2021-01-05 08:21:25 | no | PLoS Biol. 2005 Aug 19; 3(8):e295 | utf-8 | PLoS Biol | 2,005 | 10.1371/journal.pbio.0030295 | oa_comm |
==== Front
PLoS BiolPLoS BiolpbioplosbiolPLoS Biology1544-91731545-7885Public Library of Science San Francisco, USA 1752326310.1371/journal.pbio.0030296EditorialScience PolicyMeasures of Impact EditorialParthasarathy Hemai 8 2005 19 7 2005 19 7 2005 3 8 e296Copyright: © 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.
PLoS Biology receives a preliminary impact factor.
==== Body
im·pact (i˘m'pa˘kt′) n.
1. The striking of one body against another; collision.
2. The force or impetus transmitted by a collision.
3. The effect or impression of one thing on another.
4. The power of making a strong, immediate impression.
—Dictionary.com
Open-access publishing was headed on a collision course with traditional models of scientific publishing since well before the Public Library of Science launched its first journal. The force of that collision has seen dramatic shifts in the publishing landscape that include increased support from funding agencies for open-access publishing models and institutional archiving, greater availability of free-access articles and options from subscription-based publishers, and the launch of new open-access journals.
PLoS Biology was launched in October 2003, less than two years ago, as an open-access home to the very best in biological research. By any measure, the impact of this launch was noticeable. The online publication of our first issue was accompanied by strong and favorable media attention to our articles. The New York Times alone covered articles from nine of our first twelve issues. Content from these issues was downloaded, redistributed, and reanalyzed. In 2004, PLoS Biology articles were downloaded more than 1 million times. Because the reuse of open-access content is allowed and encouraged, the only restriction (aside from proper citation of the authors) is the creativity of the user. And with the launch of the journal and the attendant excitement about the content, manuscript submissions and presubmission enquiries rose dramatically.
But why did anyone submit great work to a journal that didn't even exist yet, from a publisher with no established reputation? The answer is that it was on the strength of promises made by our in-house editors and academic editorial board to uphold high standards and rigorous peer review, to launch an open-access alternative to the best journals, and to drive a transformation in scholarly publishing. On that promise, more than 250 authors published the 30 research articles that composed our first three issues. And it is on the basis of those first three issues that Thompson ISI has calculated a 2004 preliminary impact factor for PLoS Biology of 13.9.
Artist's conception of the Deep Impact spacecraft observing the birth of the new crater on Tempel 1
(Image: NASA/JPL/UMD; art: Pat Rawlings)
Since even before PLoS Biology was launched (and plenty of times since then), we've received queries from prospective authors asking about our impact factor. However, because of the way impact factors are calculated, it is not possible to have an impact factor until a specific time has lapsed. Thompson ISI calculates the impact factors that it announces this year by adding up all the citations in 2004 to articles that appeared in a journal in 2002 and 2003, and then dividing the total number of citations by the number of articles published by that journal in 2002 and 2003. For a long-standing journal, therefore, this number reflects the average number of citations over the course of a year to articles published in the two previous years. For PLoS Biology, this number therefore refers to citations during 2004 to articles published in only the three months of the journal's lifetime prior to 2004, which is why the initial impact factor can only be considered preliminary.
Of PLoS Biology's article types, Thompson ISI has chosen to define Research Articles, Primers, and Unsolved Mysteries as potentially citeable articles, and, hence, has divided the total number of citations accordingly. As we did not intend the latter two categories to contain articles that would garner citations from the publications monitored by ISI, it does not surprise us that these articles were in fact only cited in scholarly journals 2.4 times on average. Journal editors know that there are various ways to deliberately improve an impact factor, for example, by publishing topical review articles and by weighting content towards more highly cited fields. This begs the question of whether the editors of PLoS Biology should play the impact factor game and discontinue some of our educational material in favor of higher citations. On the contrary, our goal is to eventually expand and further develop these components of PLoS Biology.
Although our magazine content is an actively evolving section of our young journal, we consider it to be a part of the overall mission of the Public Library of Science to make scientific publishing accessible to more than just the research community. The eventual impact we hope to have on education and policy far outweighs the narrow scope of impact as defined by the impact factor. Our sister journal, PLoS Medicine, has outstripped her older sibling in the variety of content designed to educate and spark debate, rather than garner citations. A recently published Policy Forum article, “Nanotechnology and the Developing World” (DOI: 10.1371/journal.pmed.0020097), which identified and ranked the ten applications of nanotechnology most likely to benefit developing countries, was featured in the popular media in nine languages in 22 countries, including reports by BBC and Reuters. While it remains to be seen how an impact factor for PLoS Medicine will be calculated, what is more exciting to us is to think about ways to measure impact more broadly.
PLoS Biology was launched to give those who believe in the goals of open-access publishing a home for their very best biological research papers, and to show once and for all that open-access publishing is compatible with maintaining standards for the best science. Scientists need to publish in journals that are highly regarded by their peers, and the impact factor is one measure of that judgment. But there are so many more measures of impact. Publishing as a Primer an engaging and personal account by Frans de Waal of the relationship between primatology and sociology (DOI: 10.1371/journal.pbio.0020101) may not have helped our citation numbers, but it will have impacted nonetheless the readers directed to the related paper by an educational supplement in the New York Times. In schools and colleges, educators are free to use our content to inspire the next generation to a greater scientific literacy. And in our technological society, scientific literacy is more important than ever.
Comparisons are natural, but the top-tier journals that we aim to challenge were established long before the impact factor was even a twinkle in the eye of ISI's founder, Eugene Garfield. We hope that this number will give those who have wished to support unrestricted dissemination of scientific information, but who have held back for lack of a quantitative measure of the impact of publishing in PLoS Biology, one more incentive to submit their best work to this journal. Now is the time to impact the future of scientific publishing for the better.
Hemai Parthasarathy is Managing Editor for PLoS Biology. E-mail: [email protected]
| 17523263 | PMC1175828 | CC BY | 2021-01-05 08:21:24 | no | PLoS Biol. 2005 Aug 19; 3(8):e296 | utf-8 | PLoS Biol | 2,005 | 10.1371/journal.pbio.0030296 | oa_comm |
==== Front
BMC BiochemBMC Biochemistry1471-2091BioMed Central London 1471-2091-6-101592706110.1186/1471-2091-6-10Research ArticleCharacterization of the aggregates formed during recombinant protein expression in bacteria Schrödel Andrea [email protected] Marco Ario [email protected] EMBL, Protein Expression Core Facility, Meyerhofstr. 1, D-69117, Heidelberg – Germany2005 31 5 2005 6 10 10 10 2 2005 31 5 2005 Copyright © 2005 Schrödel and de Marco; licensee BioMed Central Ltd.2005Schrödel and de Marco; licensee BioMed Central Ltd.This is an Open Access article distributed under the 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 first aim of the work was to analyze in detail the complexity of the aggregates formed upon overexpression of recombinant proteins in E. coli. A sucrose step gradient succeeded in separating aggregate subclasses of a GFP-GST fusion protein with specific biochemical and biophysical features, providing a novel approach for studying recombinant protein aggregates.
Results
The total lysate separated into 4 different fractions whereas only the one with the lowest density was detected when the supernatant recovered after ultracentrifugation was loaded onto the sucrose gradient. The three further aggregate sub-classes were otherwise indistinctly precipitated in the pellet. The distribution of the recombinant protein among the four subclasses was strongly dependent on the DnaK availability, with larger aggregates formed in Dnak- mutants. The aggregation state of the GFP-GST recovered from each of the four fractions was further characterized by examining three independent biochemical parameters. All of them showed an increased complexity of the recombinant protein aggregates starting from the top of the sucrose gradient (lower mass aggregates) to the bottom (larger mass aggregates). These results were also confirmed by electron microscopy analysis of the macro-structure formed by the different aggregates. Large fibrils were rapidly assembled when the recombinant protein was incubated in the presence of cellular extracts, but the GFP-GST fusion purified soon after lysis failed to undergo amyloidation, indicating that other cell components probably participate in the active formation of large aggregates. Finally, we showed that aggregates of lower complexity are more efficiently disaggregated by a combination of molecular chaperones.
Conclusion
An additional analytical tool is now available to investigate the aggregation process and separate subclasses by their mass. It was possible to demonstrate the complexity of the aggregation pattern of a recombinant protein expressed in bacteria and to characterize biochemically the different aggregate subclasses. Furthermore, we have obtained evidence that the cellular environment plays a role in the development of the aggregates and the problem of the artifact generation of aggregates has been discussed using in vitro models. Finally, the possibility of separating aggregate fractions with different complexities offers new options for biotechnological strategies aimed at improving the yield of folded and active recombinant proteins.
==== Body
Background
The concept of protein aggregation suggests a non-physiological process resulting in the formation of large structures, often chaotic, and in which the proteins have lost their original function/activity. Nevertheless, the collapse of the native conformation can also produce very regular structures, as in the case of amyloid fibrils [1]. Such a process can originate from sensitive protein intermediates during folding as well as from partially denatured proteins that lost their native conformation as a consequence of stress conditions.
Cells possess a sophisticated quality control system to prevent the accumulation of protein aggregates. Molecular chaperones are engaged to promote the correct (re)-folding of misfolded molecules that otherwise undergo protease degradation. Misfolded proteins escaping the quality control may form aggregates that can be trapped in precipitates (aggresome in eukaryotic cells, inclusion bodies in bacteria) to limit their interference with the cell physiology [2]. Inclusion bodies also have a storage function and parts of the trapped proteins are in a dynamic equilibrium with their soluble fraction [3]. Under pathological conditions aggregates develop into structures that hinder the cell functions, as in the case of neuron degenerative diseases.
In bacteria the stress-dependent development of aggregates has been exploited to study the function of the chaperone network. Aggregation has been reversed in vivo and the identification of the chaperone combinations necessary for the re-folding of the proteins from aggregates was performed using in vitro conditions [4-7]. Nevertheless, the biophysical features of the aggregates have never been investigated. Heat shock is the most studied stress factor but recombinant protein expression can also dramatically modify the cell balance. In fact, the exploitation of highly efficient polymerases increases the rate of protein synthesis so that as much as 50% of the totally accumulated protein can be represented by the recombinant one and the cell folding machinery can become limiting. The optimization of some growth parameters, like the use of low growth temperatures and non-saturating amounts of expression inducer as well as the over-expression of chaperones by means of short heat shock, ethanol stress or recombinant co-expression [8,9], has often improved the yields of recombinant soluble proteins. Nevertheless, in most of the cases part or all of the recombinant protein expressed in bacteria is recovered as precipitates in the inclusion bodies.
Both amorphous and organized inclusion bodies have been isolated [10]. Their composition varies from almost homogeneous to cases in which 50% of the material is represented by contaminants [11,12]. The structural heterogeneity of the inclusion bodies has recently been shown [13,14] and it could be a consequence of the variable aggregation pattern to which a single protein can undergo under different conditions [15]. Proteins trapped in the inclusion bodies can be re-solubilised in vivo by impairing the de novo protein synthesis because the block of new protein production makes available larger amounts of chaperones and foldases for refolding precipitated proteins [3]. The temporal separation between recombinant expression of chaperones and target proteins has also been successfully used to improve the yield of soluble recombinant proteins [8]. These results suggest a model for which soluble proteins are in a dynamic equilibrium with aggregates. In conclusion, modifications of the cell conditions can modulate the aggregation rate and the protein aggregation process can be reversed by conditions favorable for the folding machinery.
This dynamic view for which proteins can pass from soluble to insoluble and back to soluble state suggests the presence of different degrees of aggregation complexity. Soluble aggregates of recombinant proteins have been described [16,17] and in a recent paper we have shown that the GFP-GST fusion protein expressed in bacteria forms aggregates with an estimated mass ranging from a few hundred kDa to more than 1000 kDa [18]. The separation of the aggregates using a blue native gel electrophoresis followed by SDS-PAGE indicated an almost continuous distribution with few regions of concentrated accumulation. This kind of analysis allows for precise identification of aggregate patterns and comparison among different samples but is not suitable for the further characterization of the aggregates. Therefore, we present here an alternative protocol to separate sub-classes of aggregates using a sucrose step gradient and the results concerning the biophysical organization and biochemical specificities of such aggregates.
Results and Discussion
Separation of protein aggregate sub-classes by sucrose step gradient
Preliminary experiments showed that the recombinant GFP-GST produced in bacteria grown at temperature higher than 30°C was mainly recovered in the pellet after ultracentrifugation of the lysates. Nevertheless, decreasing growth temperatures enabled the proportionally inversed recovery of the fusion protein in the supernatant. At 20°C roughly half of the total GFP-GST was in the supernatant (data not shown).
Density gradients have been widely used to separate biological material according to mass. We loaded cell fractions from bacteria induced to express the GFP-GST fusion recombinant protein on a sucrose step gradient to recover sub-classes of aggregates. The fluorescence of GFP-GST simplified the identification of the sucrose concentrations which enabled the separation of the aggregates only at the interface between two different sucrose cushions. Finally, four fractions of GFP-GST were separated when loading a total lysate recovered from bacteria grown at 20°C onto a 0%, 30%, 50%, 70%, 80% sucrose step gradient (Fig. 1A, tube number 2). SDS analysis confirmed that the recombinant GFP-GST was the major protein in all the fractions, however, the co-migrated bacterial proteins were specific for a particular fraction (data not shown). We have already shown that aggregates of GFP-GST can trap other proteins [18] and that chaperones can strongly bind to aggregated recombinant proteins [19]. Dot blot analysis performed using antibodies against the major chaperones showed that DnaK and ClpB were concentrated mostly in the upper gradient fractions -in which the low-density material accumulated- while GroEL and IbpB co-migrated with the larger GFP-GST aggregates (Fig. 1B). These data are in agreement with previous reports that indicated a preferential binding of the different chaperones to aggregates with different degree of complexity [6,7].
Figure 1 Separation of recombinant GFP-GST fractions by a sucrose step gradient. A) Distribution of the recombinant protein using cell fractions recovered from different bacterial strains and from bacteria grown at different temperatures. Tube number 1 was loaded with the supernatant separated after lysate ultracentrifugation while total lysates were used for the other experiments. B) Dot-blot for the fractions separated by sucrose step gradient. Each fraction was tested with specific antibodies for the chaperones DnaK, ClpB, IbpB and GroEL.
The recombinant protein from the four fractions was purified by metal affinity chromatography and both fluorescence and SDS-PAGE analysis indicated that the entire recombinant protein was bound and specifically eluted (data not shown). Protein amount determined by Bradford indicated that, on average, 39% of the total GFP-GST accumulated in the fraction 1, 14%, 22% and 25% in the other three, respectively, from the top to the bottom.
After ultracentrifugation of the lysate, the supernatant was loaded onto the sucrose gradient and the GFP-GST migrated exclusively to the interface between 0% and 30% sucrose (Fig. 1A, tube number 1). We knew from the preliminary experiments that bacteria grown at 30°C produced only insoluble GFP-GST. The fusion protein present in the total lysate from such bacteria was distributed almost exclusively in the fractions 3 and 4 and the fluorescence was almost undetectable (Fig. 1A, tube number 3).
The role of chaperones in limiting the protein aggregation has been widely demonstrated and DnaK has a key role in the chaperone network [4-7]. The sucrose step gradient demonstrated what kind of aggregate pattern modifications occur when the DnaK concentrations vary. No GFP-GST was recovered anymore in the upper fraction when DnaK- mutant bacteria were grown at 20°C and non-fluorescent aggregates largely accumulated in the lower fractions and even on the bottom of the tube (Fig. 1A, tube number 4). In contrast, both soluble GFP-GST and stronger fluorescence were detected after separation of a lysate from bacteria over-expressing DnaK grown at 30°C (Fig. 1A, tube number 5), suggesting that DnaK can improve the GFP-GST stability.
This first set of experiments showed the complexity of the aggregation pattern. In fact, the previously non-characterized insoluble fraction recovered in the pellet was distributed in three classes according to mass and it was possible to separate soluble and insoluble recombinant protein by means of a sucrose gradient. Noteworthy is also the fact that fluorescence can be found in all the four fractions (Fig. 1A), indicating that even in the insoluble aggregates of a larger mass at least part of the trapped recombinant protein conserved a native-like structure. This is in agreement with the report that part of the protein present in the inclusion bodies conserves its secondary structure [20]. Aggregate sub-classes with different complexity and protease resistance have previously been identified in inclusion bodies and also in that case a protein fraction was still active [13,14,21]. In this study, the structural hetereogenity of the proteins trapped in the aggregates is confirmed by our data.
Biophysical characterization of the GFP-GST fractions separated by the sucrose gradient
The separation of the recombinant GFP-GST on the sucrose gradient is an indication of a mass difference among the aggregates and we wished to confirm these data by size exclusion chromatography (SEC). First, the GFP-GST proteins affinity purified from the four sucrose gradient fractions were dialysed and analysed in the fluorimeter according to the method proposed by Nominé et al. [22], namely the absorbance at 280 and 340 nm was measured and the ratio calculated. This value (aggregation index) indicates the relative aggregation, is quickly determined, and allows the comparison of different fractions of the same protein. Low values indicate a lower aggregation state and our data show that there is a gradient of increasing aggregation from the top fraction to the bottom fractions (Table 1).
Table 1 Biophysical characterization of the different aggregate fractions separated by sucrose gradient. The 4 fractions were analysed for their aggregation index, their elution profile using size exclusion chromatography (SEC) and calculating the ratio between aggregated and monodispersed protein, and their binding to the dye ThioflavinT, indicative of amyloid formation. The results refer to one experiment representative of three repetitions.
Aggregation index Abs 280/340 nm SEC index monodispersed/ aggregated protein ThioflavinT Abs 482 nm
Fraction 1 0.38 1.8 4.8
Fraction 2 2.83 0.5 8.8
Fraction 3 3.95 0.4 9.6
Fraction 4 5.96 0.25 13.4
The 4 GFP-GST fractions were also subjected to SEC and the ratio between the areas of the peaks corresponding to the monodispersed and the aggregated protein was calculated (SEC index). Such an index confirmed an increasing state of aggregation from sucrose fraction 1 to 4 (Table 1). Surprisingly, the SEC experiments showed that both aggregated and functional forms of the fusion protein were present in both the three fractions corresponding to the insoluble GFP-GST and the (soluble) fraction 1. Soluble aggregates have been described before and are probably common when fusion proteins are expressed [16,17]. It was not possible to separate monodispersed GFP-GST from soluble aggregates by means of sucrose gradients of decreasing concentrations (data not shown).
We finally tried to characterize the aggregates according to their specific structure. ThioflavinT (ThT) is a dye that preferentially binds to amyloid-like fibrils [23]. We measured an increasing binding when aggregates of higher complexity were used (Table 1). In contrast, there was not significant binding of any aggregate to 8-anilino-1-naphtalenesulfonic acid (ANSA) that has been used as a marker of the amorphous aggregates [24]. This suggests that the aggregates formed by GFP-GST probably have a regular structure involving β-sheets rather than being a chaotic complex held together by hydrophobic interactions. Instead, a micellar organization has been proposed for the soluble aggregates [17,22].
Aggregate identification by electron microscopy
In the case of the GFP-GST fractions we showed that the degree of amyloidation detected by ThT-binding progressively increased from fraction 1 to fraction 4 (Table 1). The capacity to form fibrils is sequence specific [25] and it seems a generic feature of polypeptide chains [26]. The development into fibrils is characterized by a log phase during which the aggregation seeds are formed followed by a period of rapid growth [27]. Once formed, the fibrils act as aggregation seeds, speeding up the process. Therefore, it could be expected that larger aggregate networks have the possibility to develop faster into structures of higher complexity. In order to test this hypothesis, the GFP-GST from the four sucrose gradient fractions was recovered immediately after centrifugation and mounted for electron microscopy analysis.
Some aggregation seeds (20–40 nm in diameter) were visible even when the GFP-GST from the upper fraction was used (Figure 2A, fraction 1). Sort of chains composed by globular elementary structures and measuring several hundreds of nm were observed when GFP-GST from the fraction 2 was exploited (Figure 2A) while protofilaments and higher ordered fibrils [28] longer than 1 μm (Figure 2A) were visible when samples from fractions 3 and 4 were used. Therefore, it was possible to demonstrate the relation between the biochemical indexes used to characterize the aggregation of GFP-GST and the macro-aggregation complexity visible by electron microscopy.
Figure 2 Electron microscopy characterization of GFP-GST macro-aggregates. A) Samples recovered from the 4 aggregation fractions were mounted soon after the sucrose gradient separation and observed by electron microscopy. B) The samples for the electron microscopy grids were from the fractions 1 and 4 recovered after sucrose gradient separation but incubated 24 hours with the co-migrated cell fraction before being mounted.
Fibrils are the end product of GFP-GST aggregation but the different classes of aggregates separated by sucrose gradient can be considered as dynamic intermediates that can either develop to larger structures or be reversed into lower-complexity aggregates [29]. Both the initial complexity and the incubation time of polypeptides prone to aggregation are crucial for the building of the aggregates. We wished to demonstrate the importance of these factors in a control experiment. GFP-GST was separated into fractions by sucrose gradient and the fractions 1 and 4 were mounted for electron microscopy only after 24 hours of incubation in the presence of the co-migrated cell components. Both samples raised similar large fibrils (Figure 2B), indicating that the incubation period was sufficient for both, independent of their initial aggregation state, to reach the rapid growth phase that leads to the fibril formation.
This experiment underlines once more the importance of the parameter time in studies dealing with aggregation and questions the meaning of some in vitro experiments. In fact, the fibril maturation outside the bacterial cell could have peculiar features. For instance, the lack of space-constrain or limitations in the disaggregation processes could enable the formation of fibrils the length of which are difficultly compatible with the size of E. coli cells (Figure 2B). The experiments described in the two last paragraphs will show the impact of cell components in promoting aggregation and disaggregation.
Finally, the presence of aggregation seeds smaller than 40 nm in diameter shows that it is not possible to discriminate between soluble and aggregated fractions by the use of simplified methods in high-throughput protocols as, for instance, the exploitation of a 0.65 μm pore size filter [30].
Is the aggregation of GFP-GST actively supported?
In the previous experiments we showed that even the moderately aggregated GFP-GST recovered from the upper fraction of the sucrose gradient could form fibrils if the sample was incubated with the cell fraction for at least 1 day before it was prepared for the electron microscopy analysis. In a recent paper it was claimed that bacterial chaperones play an active role in the formation of the aggregates [31]. The possible participation of cell components in catalyzing the GFP-GST fibril formation was investigated in a control experiment. The process of aggregate maturation of the soluble recombinant protein in the presence of other cell components was limited to 1 hour performing the affinity purification of the GFP-GST immediately after lysis to avoid a seeding process during the 15 hour centrifugation of the cell components upon the sucrose gradient. The sample was incubated at room temperature for 4 weeks and the modifications of the secondary structure were monitored by CD while corresponding samples were mounted for electron microscopy. No significant modification was observed in the first two weeks and a slight increase of the β-sheet content was measured only after 4 weeks (Figure 3). The use of different protein concentrations and the addition of sucrose to the proteins did not modify the pattern and no detectable aggregate was observed at the electron microscopy using the corresponding samples (data not shown).
Figure 3 Circular dichroism spectra of GFP-GST. Protein purified by metal affinity from the supernatant obtained after lysate ultracentrifugation was directly analysed (day 0) or incubated at room temperature before the collection of further spectra the days 10, 15, and 27.
Therefore, these results strongly suggest that the co-presence of other molecules is necessary to trigger the process of regular aggregation of the recombinant protein, probably by facilitating the formation of aggregation seeds. Chaperones can play a role in the aggresome formation [32] and GroEL has been claimed to be actively involved in bacterial inclusion body formation [31]. Our data can only confirm that GroEL co-migrates with the aggregates of larger mass (Fig. 1B). Finally, we are looking for an analytical method to determine if the process of cell lysis is crucial for the development of the aggregates.
Aggregate complexity and re-folding
Both in vivo and in vitro experiments illustrated the co-operative action of chaperone networks in disaggregating misfolded proteins [4-7] but the features of the real aggregates that are the target of the chaperones in the cells have never been investigated. We used the aggregates from fractions 3 and 4 to test if they could be a substrate for chaperone-dependent refolding and if the different structure complexity had a role on the refolding kinetic.
An equimolar combination of DnaK, DnaJ, GrpE, and ClpB [6] quickly disaggregated the large precipitates (Figure 4). Specifically, the complexity of the aggregates from fraction 3 was reduced in a faster and more efficient way. In fact, the aggregation index dropped by half in only 4 min while it took 10 min in the case of the aggregates from fraction 4. Furthermore, there was a higher residual aggregation: the aggregation indexes measured were 1.2 and 0.7 for the aggregates from fractions 4 and 3, respectively. In comparison, the GFP-GST from fraction 1 scored 0.38 (Table 1). The addition of equimolar amounts of BSA to the aggregates in absence of chaperones had no disaggregation effect.
Figure 4 Chaperone-dependent in vitro disaggregation. Purified GFP-GST aggregates recovered from the fractions 3 and 4 of the sucrose gradient were incubated in the presence of an equimolar mixture of DnaK, DnaJ, GrpE, and ClpB in the presence of a system constantly providing ATP. The aggregation index was repeatedly measured during a 45 min incubation.
The preferential disaggregation of subclasses of aggregates with lower complexity observed in vitro is reminiscent of previous works indicating that specific subclasses of the proteins trapped in the inclusion bodies are preferentially refolded under physiological conditions [3,13] and that the reversibility is increasingly difficult and dependent on the size of the aggregates [29]. The limit of this experiment is that it is difficult to scale up and the small amount of the protein used was insufficient for undertaking further biophysical analysis. The aggregation index gives only relative values and, therefore, we can state that the degree of aggregation decreased but cannot conclude that the disaggregated protein was also correctly folded. Nevertheless, the results suggest that it would be of biotechnological interest to separate the aggregate subclasses and use the lower complexity aggregates in refolding protocols.
Conclusion
There is increasing evidence that aggregates are heterogeneous in size and complexity [2,12-16,26]. The aggresomes are actively built in eukaryotic cells and the physiological meaning of the process would be the packing of disorganized aggregates that could interfere with the normal cell functions by non-specifically binding to other cell components [33,34]. The possibility to recover functional proteins from the insoluble aggregates [3] would indicate that at least in bacteria they can function as a reserve in dynamic equilibrium with soluble fractions.
The expression of recombinant proteins is a stress factor because they compete for energy and substrates with native expression and can interfere with the normal metabolism by forming aggregates, both in prokaryotic and eukaryotic cells [2,34]. The possibility to store the excess of misfolded recombinant protein could be a way to get rid of dangerous aggregating material when misfolded proteins escaped the quality control of chaperones and proteases [2]. The cellular mechanisms that favor the generation of amyloids (Figure 2) might also be useful in preventing amorphous aggregates in non-specifically trapping native proteins [18]. The aggregate organization would consider an aggregate mash that grow from small entities towards larger insoluble structures [34] composed by a core of protease-resistant fibrils [13,14], homologous proteins at different levels of misfolding and some heterologous and non-specifically trapped proteins [18] (Figure 5B).
Figure 5 Schematic representation of the aggregation. A) Dynamic of the aggregation. GFP-GST aggregates progressively form both soluble and insoluble aggregates. Chaperone activity can reverse the process of aggregation in a way that is inversely proportional to the degree of complexity reached by the aggregates and could also play a role in the aggregate maturation towards more structured complexes. B) Aggregate model. The aggregation of GFP-GST probably starts with misfolded single proteins that collapse into pre-fibrillar structures. These catalyze the aggregation of new molecules to form larger amyloid fibrils. In the initial phases, the co-presence of molecules with different degree of misfolding and amyloidation seems apparent. Pre-fibrils could form the core of the aggregation seeds to which partially misfolded GFP-GST molecules bind. Some of these still conserve a native-like structure compatible with fluorescence functionality. The aggregation nets can trap other proteins in a probably non-specific manner.
In this paper we present data supporting the idea of a progressive maturation of recombinant GFP-GST aggregates into amyloid fibrils. Furthermore, it seems that the process is facilitated by some other cell components since the fibril maturation was extremely slower when the recombinant protein was separated from the other cell components soon after the lysis (Fig. 3). For instance, GroEL has been reported having an active role in inclusion body formation [31] and specifically co-migrate with the larger aggregates could (Fig. 1B). Conversely, the combination of DnaK, DnaJ, GrpE and ClpB could disaggregate large insoluble structures (Figures 4 and 5A).
It seems that the aggregation process of recombinant proteins is extremely more complicated than normally accepted and our separation protocol turned out to be a useful tool for characterizing the aggregates. Furthermore, such an aggregation process shares many features with the maturation of pathological amyloids in eukaryotic cells and, therefore, the bacterial system -experimentally easy to modify- would be considered as a model to integrate the results obtained using in vitro systems and to study the impact of chemical and biophysical parameters on the aggregation development. We simplified the work by using a fluorescent construct but any protein for which antibodies are available could be used for following the aggregation development.
Methods
Cell culture and protein preparation
A fusion construct His-GST-GFP cloned in a Gateway destination vector (Invitrogen, kindly provided by D. Waugh) was transformed and expressed in the following bacterial strains: BL 21 (DE3), BL 21 (DE3) RIL codon plus, GK2 (dnak-), BL 21 (DL3) co-expressing the chaperone combinations GroELS and GroELS/DnaK/DnaJ/GrpE/ClpB, respectively (kindly provided by B. Bukau). Bacteria were grown at 37°C until the OD600 reached 0.4, then the cultures were adapted to different temperatures (20°C, 25°C, 30°C, 37°C), induced at an OD600 of 0.6 with 0.1 mM IPTG and grown for further 20 h. The bacteria were pelleted by centrifugation (6000 g × 15 min), washed in 10 mL of PBS and finally stored at -20°C.
The pellet was resuspended in 10 mL of lysis-buffer (50 mM potassium phosphate buffer, pH 7.8, 0.5 M NaCl, 5 mM MgCl2, 1 mg/mL lysozyme, 10 μg/mL DNase), sonicated in a water bath (Branson 200) for 5 min and the lysate was incubated for 30 min on a shaker at room temperature. The supernatant was recovered after ultracentrifugation (35 min at 150000 × g).
Fractions from sucrose gradients were recovered using a bent Pasteur pipette and affinity purified using a HiTrap chelating affinity column (Amersham Biosciences) pre-equilibrated with 20 mM Tris HCl, pH 7.8, 500 mM NaCl, 15 mM imidazole. The His-tagged recombinant protein was eluted in 20 mM Tris, pH 7.8, 125 mM NaCl, and 250 mM imidazole. Protein quantification was based on the absorbance at 280 nm.
Sucrose gradients and gel filtration
Total cell lysates or supernatants from ultracentrifugation of total cell lysates (1 mL) were loaded onto 14 × 95 mm Ultra-Clear centrifuge tubes (Beckman) prepared with a step gradient formed by four layers of 20 mM TrisHCl buffer, pH 8, containing 80%, 70%, 50%, 30%, and 0% sucrose, respectively. The tubes were centrifuged 15 hours at 180,000 × g at 4°C using a SW40Ti rotor and a L-70 Beckman ultracentrifuge. The protein fractions were recovered from the interfaces between two sucrose layers, affinity purified as described above and used for further analysis. The samples for gel filtration were concentrated and the buffer replaced with 50 mM TrisHCl, pH8.0, 150 mM NaCl using a Vivapore concentrator (Vivascience) and then separated by gel filtration using a Superose 12 HR 10/30 column (Amersham).
Bioanalytical assays
The aggregation rate of the proteins was analysed according to Nominé et al. [22] using an AB2 Luminescence Spectrometer (Aminco Bowman Series 2) equipped with SLM 4 software. The excitation was induced at 280 nm and the emission scan was recovered between 260 and 400 nm.
Amyloid aggregates were estimated according to their binding to the specific dye thioflavin-T (ThT), as described by LeVine [23], and protein surface hydrophobicity was determined using the fluorescent probe 8-anilino-1-naphtalenesulfonic acid (ANSA) [24].
Circular dichroism (CD) spectra were recorded between 250 and 190 nm using suprasil precision cells (Hellma) and a Jasco J-710 instrument.
Western and dot blotting
Western blots were performed as previously described [18] using anti-GST primary antibodies. For dot blotting the proteins were transferred onto a PVDF membrane using a Bio-Rad Criterion blotter. The primary rabbit antibodies were a gift from Dr. Bukau and were purified from sera using Protein G Plus/Protein A Agarose (Oncogene) to minimize the background. Peroxidase-conjugated secondary antibodies for chemioluminescent detection were purchased from Dianova and the detection performed using the SuperSignal® West Femto Maximum Sensitivity Substrate (Pierce), following the supplier's instructions. Blots were used repeatedly by effectively removing the antigen-antibody interaction using the Western Blot Recycling it (Alpha Diagnostic Int.).
Sample preparation for electron microscopy
Protein samples were purified by affinity chromatography and equal amounts fixed by using the "single-droplet" parafilm protocol. 5 μL of each protein sample were pipetted on a grid (Agar Scientific) and incubated 1 min at room temperature. Excess fluid was removed using filter paper, the unbound protein was washed and the grids were placed on a 50 μL drop of 1% uranyl acetate with the section side downwards. Finally, the grids were dried, placed in the grid-chamber and stored in desiccators before the samples were observed with a CM120 BioTwin electron microscope (Philips).
In vitro re-folding assay
The conditions for the chaperone-dependent disaggregation of GST-GFP in vitro were chosen according to Mogk et al. [35] and the process was monitored using the fluorimetric assay described above [22]. 1 μM of aggregated protein was resuspended in 50 mM Tris HCl, pH 7.5, 20 mM MgCl2, 150 mM KCl, 2 mM DTT, in the presence of 1 μM ClpB, 1 μM DnaK, 0.2 μM DnaJ, 0.1 μM GrpE, 3 mM phosphoenolpyruvate, and 20 ng/mL of pyruvate kinase. The reaction was started by the addition of 2 mM NaATP.
Authors' contributions
Andrea Schrödel performed all the experiments at least once. Ario de Marco conceived the study, repeated some of the experiments and wrote the manuscript.
Acknowledgements
The authors wish to thank Dr. M. Lopez de la Paz for the assistance with the electron microscopy, Dr. B. Bukau for having provided the chaperone vectors and antisera, Dr. A. Mogk for his refolding protocol, and Dr. D Waugh for his GFP-GST construct.
==== Refs
Harper JD Lansbury PT Models of amyloyd seeding in Alzheimer's disease and Scrapie Annu Rev Biochem 1997 66 385 407 9242912 10.1146/annurev.biochem.66.1.385
Kopito RR Aggresomes, inclusion bodies and protein aggregation Trends Cell Biol 2000 10 524 530 11121744 10.1016/S0962-8924(00)01852-3
Carrió MM Villaverde A Protein aggregation as bacterial inclusion bodies is reversible FEBS Lett 2001 489 29 33 11231008 10.1016/S0014-5793(01)02073-7
Veinger L Diamant S Buchner J Goloubinoff P The small heat-shock protein IbpB from Escherichia coli stabilizes stress-denatured proteins for subsequent refolding by a multichaperone network J Biol Chem 1998 273 11032 11037 9556585 10.1074/jbc.273.18.11032
Mogk A Tomoyasu T Goloubinoff P Rüdiger S Röder D Langen H Bukau B Identification of thermolabile Escherichia coli proteins: prevention and reversion of aggregation by DnaK and ClpB EMBO J 1999 18 6934 6949 10601016 10.1093/emboj/18.24.6934
Goloubinoff P Mogk A Ben Zvi AP Tomoyasu T Bukau B Sequential mechanism of solubilization and refolding of stable protein aggregates by a bichaperone network Proc Natl Acad Sci USA 1999 96 13732 13737 10570141 10.1073/pnas.96.24.13732
Mogk A Deuerling E Vorderwulbecke S Vierling E Bukau B Small heat shock proteins, ClpB and the DnaK system form a functional triade in reversing protein aggregation Mol Microbiol 2003 50 585 595 14617181 10.1046/j.1365-2958.2003.03710.x
de Marco A De Marco V Bacteria co-transformed with recombinant proteins and chaperones cloned in independent plasmids are suitable for expression tuning J Biotechnol 2004 109 45 52 15063613 10.1016/j.jbiotec.2003.10.025
Steczko J Donoho GA Dixon JE Sugimoto T Axelrod B Effect of ethanol and low-temperature culture on expression of soybean lipoxygenase L-1 in Escherichia coli Prot Expr Purif 1991 2 221 227 10.1016/1046-5928(91)90075-T
Bowden GA Peredes AM Georgiou G Structure and morphology of protein inclusion bodies in Escherichia coli Biotechnol (NY) 1991 9 725 730 10.1038/nbt0891-725
Valax P Georgiou G Molecular characterization of beta-lactamase inclusion bodies produced in Escherichia coli. 1. Composition Biotechnol Prog 1993 9 539 547 7764166 10.1021/bp00023a014
Speed MA Wang DIC King J Specific aggregation of partially folded polypeptide chains: the molecular basis of inclusion body composition Nat Biotechnol 1996 14 1283 1287 9631094 10.1038/nbt1096-1283
Carrió MM Cubarsi R Villaverde A Fine architecture of bacterial inclusion bodies FEBS Lett 2000 471 7 11 10760503 10.1016/S0014-5793(00)01357-0
Carrió MM Gonzalez-Montalban N Vera A Villaverde A Ventura S Amyloid properties of bacterial inclusion bodies J Mol Biol 2005
Ben Zvi AP Goloubinoff P Proteinaceous infectious behavior in non-pathogenic proteins is controlled by molecular chaperones J Biol Chem 2002 277 49422 49427 12377766 10.1074/jbc.M209163200
Sachdev D Chirgwin JM Properties of soluble fusions between mammalian aspartic proteases and bacterial maltose-binding protein Biochem J 1999 338 77 81 9931301 10.1042/0264-6021:3380077
Nominé Y Ristriani T Laurent C Lefevre J-F Weiss E Travé G Formation of soluble inclusion bodies by HPV E6 oncoprotein fused to maltose-binding protein Prot Expr Purif 2001 23 22 32 10.1006/prep.2001.1451
Stegemann J Ventzki R Schrödel A de Marco A Comparative analysis of protein aggregates by blue native electrophoresis and subsequent SDS-PAGE in a three-dimensional geometry gel Proteomics 2005 5 2002 9 15841497 10.1002/pmic.200401091
Carrió MM Corchero JL Villaverde A Dynamics of in vivo protein aggregation: building inclusion bodies in recombinant bacteria FEMS Microbiol Lett 1998 169 9 15 9851031 10.1016/S0378-1097(98)00444-3
Oberg K Chrunyk BA Wetzel R Fink AL Native-like secondary structure in interleukin-1β inclusion bodies by attenuated total reflectance FT-IR Biochemistry 1994 33 2628 2634 8117725 10.1021/bi00175a035
Tokatlidis K Dhurjati P Millet J Beguin P Albert JP High activity of inclusion bodies formed in Escherichia coli overproducing Clostridium thermocellum endoglucanase D FEBS Lett 1991 282 205 208 2026260 10.1016/0014-5793(91)80478-L
Nominé Y Ristriani T Laurent C Lefevre J-F Weiss E Travé G A strategy for optimizing the monodispersity of fusion proteins: application to purification of recombinant HPV E6 oncoprotein Prot Engineer 2001 14 297 305 10.1093/protein/14.4.297
LeVine H Quantification of beta-sheet amyloid fibril structures with ThioflavinT Methods Enzymol 1999 309 274 284 10507030
Busby TF Atha DH Ingham KC Thermal denaturation of antithrombin III. Stabilization by heparin and lyotropic anions J Biol Chem 1981 256 12140 12147 7298649
Linding R Schymkowitz J Rousseau F Diella F Serrano L A comparative study of the relationship between protein structure and beta-aggregation in globular and intrinsically disordered proteins J Mol Biol 2004 342 345 353 15313629 10.1016/j.jmb.2004.06.088
Dobson CM Protein folding and misfolding Nature 2003 426 884 890 14685248 10.1038/nature02261
Caughey B Lansbury PTJr Protofibrils, pores, fibrils, and neurodegeneration: separating the responsible protein aggregates from the innocent bystanders Annu Rev Neurosci 2003 26 267 298 12704221 10.1146/annurev.neuro.26.010302.081142
Holm Nielsen E Nybo M Svehag S-E Electron microscopy of prefibrillar structures and amyloid fibrils Methods Enzymol 1999 309 491 496 10507043
Calamai M Canale C Relini A Stefani M Chiti F Dobson CM Reversal of protein aggregation provides evidence for multiple aggregated states J Mol Biol 2005 346 603 616 15670608 10.1016/j.jmb.2004.11.067
Dyson MR Shadbolt SP Vincent KJ Perera RL McCafferty J Production of soluble mammalian proteins in Escherichia coli: identification of protein features that correlate with successful expression BMC Biotechnology 2004 4 32 15598350 10.1186/1472-6750-4-32
Carrió MM Villaverde A Role of molecular chaperones in inclusion body formation FEBS Lett 2003 537 215 221 12606060 10.1016/S0014-5793(03)00126-1
Garcia-Mata R Bebok Z Sorscher EJ Sztul ES Characterization and dynamics of aggresome formation by a cytosolic GFP chimera J Cell Biol 1999 146 1239 1254 10491388 10.1083/jcb.146.6.1239
Johnston JA Wand CL Kopito RR Aggresomes: a cellular response to misfolded proteins J Cell Biol 1998 143 1883 1898 9864362 10.1083/jcb.143.7.1883
Stefani M Dobson CM Protein aggregation and aggregate toxicity: new insights into ptotein folding, misfolding diseases and biological evolution J Mol Med 2003 81 678 699 12942175 10.1007/s00109-003-0464-5
Mogk A Schlieker C Friedrich KL Schönfeld H-J Vierling E Bukau B Refolding of substrates bound to small Hsps relies on a disaggregation reaction mediated most efficiently by ClpB/DnaK J Biol Chem 2003 278 31033 31042 12788951 10.1074/jbc.M303587200
| 15927061 | PMC1175841 | CC BY | 2021-01-04 16:36:49 | no | BMC Biochem. 2005 May 31; 6:10 | utf-8 | BMC Biochem | 2,005 | 10.1186/1471-2091-6-10 | oa_comm |
==== Front
BMC BioinformaticsBMC Bioinformatics1471-2105BioMed Central London 1471-2105-6-1251591069010.1186/1471-2105-6-125Methodology ArticleDetection of nuclei in 4D Nomarski DIC microscope images of early Caenorhabditis elegans embryos using local image entropy and object tracking Hamahashi Shugo [email protected] Shuichi [email protected] Hiroaki [email protected] Kitano Symbiotic Systems Project, ERATO, Japan Science and Technology Corporation, M31 6A, 6-31-15 Jingumae, Shibuya, Tokyo 150-0001, Japan2 Graduate School of Science and Technology, Keio University, 3-14-1 Hiyoshi, Kohoku, Yokohama 223-8522, Japan3 Institute for Bioinformatics Research and Development (BIRD), Japan Science and Technology Agency, 5-3 Yonbancho, Chiyoda, Tokyo 102-0081, Japan4 Sony Computer Science Laboratories, Inc., 3-14-13 Higashi-Gotanda, Shinagawa, Tokyo 141-0022, Japan2005 24 5 2005 6 125 125 16 12 2004 24 5 2005 Copyright © 2005 Hamahashi et al; licensee BioMed Central Ltd.This is an Open Access article distributed under the terms of the Creative Commons Attribution License (), which permits unrestricted use, distribution, and reproduction in any medium, provided the original work is properly cited.
Background
The ability to detect nuclei in embryos is essential for studying the development of multicellular organisms. A system of automated nuclear detection has already been tested on a set of four-dimensional (4D) Nomarski differential interference contrast (DIC) microscope images of Caenorhabditis elegans embryos. However, the system needed laborious hand-tuning of its parameters every time a new image set was used. It could not detect nuclei in the process of cell division, and could detect nuclei only from the two- to eight-cell stages.
Results
We developed a system that automates the detection of nuclei in a set of 4D DIC microscope images of C. elegans embryos. Local image entropy is used to produce regions of the images that have the image texture of the nucleus. From these regions, those that actually detect nuclei are manually selected at the first and last time points of the image set, and an object-tracking algorithm then selects regions that detect nuclei in between the first and last time points. The use of local image entropy makes the system applicable to multiple image sets without the need to change its parameter values. The use of an object-tracking algorithm enables the system to detect nuclei in the process of cell division. The system detected nuclei with high sensitivity and specificity from the one- to 24-cell stages.
Conclusion
A combination of local image entropy and an object-tracking algorithm enabled highly objective and productive detection of nuclei in a set of 4D DIC microscope images of C. elegans embryos. The system will facilitate genomic and computational analyses of C. elegans embryos.
==== Body
Background
The position of the nucleus is a crucial piece of information in any study of the development of multicellular organisms. A fertilized egg – a single cell – develops into a multicellular organism through many spatially and temporally dynamic cellular activities, including cell division, cell migration, cell differentiation, cell fusion, and cell death. Often, these dynamic cellular activities are described in terms of the positions of the nuclei, and the roles and mechanisms of those cellular activities are studied using these descriptions of cellular activities because the nucleus is generally positioned at the center of a cell and is the most noticeable organelle in a cell [1]. The position of the nucleus is usually identified from images captured through a microscope. Therefore, detection of the nucleus in microscope images is essential for studying the development of multicellular organisms.
The nucleus is usually detected manually on these microscope images. However, manual detection reduces the objectivity and productivity of identification of nuclear position. The objectivity and productivity of such measurements are becoming critical in modern biology, where the importance of bioinformatics, computational biology, and genomics is increasing. High objectivity of measurements is strongly expected in bioinformatics and computational biology. In the large-scale data analyses typical of bioinformatics, the quality of the analysis depends largely on that of the data analyzed [2]. In the simulation analyses typical of computational biology, the decision-making step is a comparison between the simulation and in vivo measurement [3]. High productivity of measurements is strongly expected in genomics. Organisms have thousands of genes [4,5], and systematic study of the functions of all of these genes – a typical strategy in genomics – needs thousands of measurements [6].
The soil nematode Caenorhabditis elegans is the simplest multicellular organism that has been most extensively studied in biology [7,8]. Because of the simplicity of this organism, results from its study constitute a foundation for our understanding of higher multicellular organisms. In C. elegans, the position of the nucleus is usually identified from images obtained through a Nomarski differential interference contrast light microscope, hereafter called a DIC microscope [9]. Three-dimensional (3D) positions of the nuclei are identified from a set of images recorded in multiple focal planes, and time-dependent changes in these positions are followed in a set of images recorded in multiple focal planes and at multiple time points. The 4D DIC microscope is an automated system that records DIC microscope images in multiple focal planes and at multiple time-points [10,11]. To help follow time-dependent changes in the 3D positions of nuclei in a set of images recorded by the 4D DIC microscope system (hereafter called a set of 4D DIC microscope images), two computer-assisted systems have been developed, namely SIMI BioCell [12] and 3D-DIASemb [13]. SIMI BioCell is a graphical user interface that displays a set of 4D DIC microscope images, helps to identify the positions of nuclei, and records these identified positions. 3D-DIASemb is similar to SIMI BioCell but can also record and display the perimeter of the nucleus and cell. Although both of these systems help greatly to follow time-dependent changes in the 3D positions of nuclei, the nuclei are still detected manually and nuclear detection is therefore still a laborious task. As a result, the objectivity and productivity of identification of nuclear positions are still low.
Automation of nuclear detection increases the objectivity and productivity of identification of nuclear positions. Yasuda et al. [14] attempted to automate nuclear detection by using several edge detection operators [15,16]. Their automated system detected nuclei from the two- to eight-cell stages in a specific set of 4D DIC microscope images. However, their system required laborious hand-tuning of parameters every time a new set of 4D DIC microscope images was applied, because the edge detection operators were very sensitive to differences in image quality (e.g., brightness, contrast) among sets of images; the differences could be controlled but not eliminated. In addition, their system could not detect nuclei that were in the process of cell division, because detection of nuclei relied on the nucleus being round (and therefore not in the process of division). Unless the positions of the dividing nuclei are known, it is difficult to follow the cell division pattern of embryos. Therefore, the system of Yasuda et al. [14] requires marked improvement before it can be used in research.
We developed a system that automates the detection of nuclei in C. elegans embryos. Our system uses local image entropy [17] and an object-tracking algorithm [18-20] to automate the detection of nuclei in sets of 4D DIC microscope images. Because local image entropy is not sensitive to differences in image quality among sets of images, our system can be applied to different sets without the need to change the system parameters. Because the object-tracking algorithm is independent of the process of cell division, our system detects nuclei both in and not in the process of cell division. Here, we show that our system can effectively detect nuclei in a C. elegans embryo from fertilization to the onset of gastrulation, i.e., from the one- to 24-cell stages.
Results
Appearance of nuclei in images obtained by the 4D DIC microscope system
The appearance of the nuclei of C. elegans embryos in 4D DIC microscope images (Figure 1A, B) varies among different focus levels and different developmental stages. The nucleus appears as a smooth, round region in the center of the cell, the cytoplasm of which appears as a rough region at all developmental stages. The boundary of the nucleus is apparent when the focus level is close to the level of the center of the nucleus (0 μm, 0 s in Figure 1B). As the focus level becomes higher or lower, the nucleus becomes smaller, reflecting the 3D shape of the nucleus, and the boundary of the nucleus becomes blurred (-3.5 μm and +3.5 μm in Figure 1B). The nucleus becomes invisible when the focus level goes beyond the level of the upper or lower end of the nucleus (-7.0 μm and +7.0 μm in Figure 1B). As the embryo develops, the number of cells in the embryo increases through repeated cell divisions, each of which produces two daughter cells from a single mother cell. When cell division begins, the nucleus begins to elongate and the boundary of the nucleus becomes blurred (160 s in Figure 1B). As cell division progresses, the nucleus continues to elongate (320 s in Figure 1B). The elongated nucleus is fragmented into several pieces (480 s in Figure 1B), which then form daughter nuclei in two daughter cells (640 s in Figure 1B). The size of the nuclei gradually decreases as the embryo develops and the number of nuclei increases (8 μm in diameter at the one-cell stage and 5 μm at the 24-cell stage). Although the appearance of the nuclei in the images varies among different focal planes and different developmental stages, a smooth image texture is a common feature of the appearance of nuclei. Our image-processing algorithm uses this feature to detect nuclei in the images (see next section).
Detection of nuclei using regions of low local image entropy
To detect nuclei in the 4D DIC microscope images, we used a common feature of nuclei in the images, that is, their smooth image texture (see previous section, Figure 1B). To quantify the smoothness of image texture in various regions of an image, we used local image entropy [17], which computes the image entropy [21] of a small area surrounding a point of interest in an image. Image entropy represents the smoothness of image texture; its value becomes high when the texture is rough and low when the texture is smooth. Because smooth image texture is a common feature of the appearance of nuclei in 4D DIC microscope images, we expected local image entropy to be lower in the nuclei than in the cytoplasm. An important feature of image entropy is low sensitivity to differences in image quality, particularly in terms of the brightness of the image. Therefore, we expected that local image entropy would quantify the smoothness of image texture in multiple images in a manner that was not sensitive to differences in quality among images.
We defined an image conversion using local image entropy as follows. Let [xij] be the matrix representing a digitized input image. Then the result of image conversion using local image entropy in an X × Y pixel window is an image [yij], where the value of yij equals the entropy of the input image lying in the X × Y pixel window Wij whose top left is pixel xij. The image entropy is , where N is the number of gray levels and P(k) is the probability of occurrence of gray level k in window Wij. Because of the presence of the window, the number of columns and rows of [yij] is smaller than those of [xij] by X - 1 and Y - 1, respectively.
To determine whether local image entropy could effectively distinguish nuclei from cytoplasm in 4D DIC microscope images, we converted the images using various window sizes (from 2 × 2 to 50 × 50 pixels, results for 4 × 4, 10 × 10 and 50 × 50 pixels are shown in Figure 2). As expected, local image entropy was lower (darker) in the nuclei than in the cytoplasm (e.g., 10 × 10 window size in Figure 2). When we used a large (50 × 50) window, the difference in local image entropy between nuclei and cytoplasm became smaller. When we used a small (4 × 4) window, high-entropy spots (bright spots) appeared throughout the images. These results indicate that local image entropy effectively distinguishes nuclei from cytoplasm in 4D DIC microscope images. For our images, 10 × 10 pixels (1 μm × 1 μm) appeared likely to be the optimal size of the window. We investigated 25 widely-used texture measures selected from all four texture analysis methods categorized by Tuceryan and Jain [22] and confirmed that local image entropy provides the best performance among those texture measures to distinguish between nuclei and cytoplasm (see Additional file 1).
To detect nuclei using this difference in local image entropy between nuclei and cytoplasm, we applied thresholding [23] to the images resulting from the image conversion and produced low-entropy regions (Figure 2E–M). The low-entropy regions were produced as follows: neighboring pixels whose local image entropy was lower than the threshold were grouped, and the resulting group was defined as a low-entropy region. As expected, many of these low-entropy regions corresponded to nuclei in the original images, whereas the size and number of the regions depended on the threshold value. The shapes of the low-entropy regions approximated those of corresponding nuclei when the threshold value was set to 175 (Figure 2F, I, L). As the threshold value decreased, the regions became smaller and more fragmented (Figure 2G, J, M). As the threshold value increased, the regions became larger and more aggregated (Figure 2E, H, K). These results indicate that low-entropy regions can be used to detect nuclei in 4D DIC microscope images. For our images, 175 was likely to be the optimal threshold value. In addition to the low-entropy regions that corresponded to nuclei, many low-entropy regions were produced that did not correspond to nuclei. These low-entropy regions corresponded to regions that have similar (smooth) image textures to that of the nucleus, such as the boundaries between cells and the spaces between the embryo and the eggshell.
Nuclear detection using low-entropy regions
We evaluated the performance of nuclear detection in a set of 4D DIC microscope images by using low-entropy regions. For the evaluation, we produced low-entropy regions from five sets of images of C. elegans embryos using a 10 × 10 pixel window and a threshold value of 175 (Figure 3). Each set of images consisted of 10,080 images (56 focal planes × 180 time points = 10,080 images). We then calculated the sensitivity and specificity as measures of performance.
Sensitivity was defined as the ratio of the sum of the number of nuclei detected at each time point to the sum of the number of nuclei existing at each time point. A nucleus was considered to be "detected" at a specific time point when it was detected by at least one low-entropy region at any focal plane at this specific time point. This definition of sensitivity is reasonable because of the difficulty in specifying the number of low-entropy regions that are expected to detect a given nucleus. The following three factors underlie this difficulty. First, a single nucleus is usually detected by several low-entropy regions in different focal planes at a single time point. Second, a single nucleus is sometimes detected by several low-entropy regions in the same focal plane at a single time point. Third, it is difficult to determine which focal plane is the top end and which is the bottom end of the focal planes at which a given nucleus is expected to be detected in low-entropy regions, because the appearance of the nucleus becomes gradually blurred as the focal plane becomes farther from the center of the nucleus (Figure 1B).
Specificity was defined as the ratio of the number of low-entropy regions detecting nuclei to the number of low-entropy regions produced. Because local image entropy is not sensitive to differences in image quality, particularly in terms of the brightness of the image, we expected that the performance of nuclear detection by examination of low-entropy regions would differ little among sets of 4D DIC microscope images.
We obtained perfect (= 1.0) sensitivity for all sets of images from the one- to the 24-cell stages (Table 1). All nuclei were detected at any time point independently of whether or not they were in the process of cell division. To confirm that this perfect sensitivity was not solely a feature of the five sets of images examined, we produced low-entropy regions from 44 sets of images of C. elegans embryos using 10 × 10 pixel windows and threshold values of 175, and then calculated the sensitivity. We obtained perfect sensitivity for all 44 sets of images of embryos from the one- to the 24-cell stages (data not shown). Sensitivity became imperfect in the later stages of embryogenesis, i.e., around the 44-cell stage or later (data not shown). In contrast, very low (< 0.10) specificity was obtained for all sets of images (Table 1). In summary, low-entropy regions could be used to detect nuclei in a set of 4D DIC microscope images of C. elegans embryos from the one- to 24-cell stages with very high sensitivity and very low specificity. The performance of nuclear detection by low-entropy regions differed little among sets of images.
Selection of low-entropy regions using object-tracking algorithm in the forward direction of time
The very high sensitivity and very low specificity of nuclear detection by using low-entropy regions motivated us to develop a process that selected low-entropy regions that actually detected nuclei. To develop this process, we used spatial and temporal information on the nucleus. In terms of spatial information, we expected the nucleus to be detected by several low-entropy regions, each of which would overlap with another region in an adjacent focal plane at the same time point, because the radius of the nucleus (> 2.5 μm) was much larger than the distance between two adjacent focal planes (0.5 μm). Therefore, a low-entropy region would be more likely to detect a nucleus than others when it overlapped with a region that detected the nucleus in an adjacent focal plane at the same time point. In terms of temporal information, we expected the nucleus to be detected by several low-entropy regions, each of which would overlap with another region in the same focal plane at an adjacent time point, because the nucleus rarely moves more than a distance equal to its diameter (> 5 μm) within the time equal to the interval between two adjacent time points (40 s). Therefore, a low-entropy region would be more likely to detect a nucleus than others when the region overlapped with a region that detected the nucleus in the same focal plane at an adjacent time point.
To select low-entropy regions by using this spatial and temporal information, we used an object-tracking algorithm [18-20] (Figure 4). The tracking algorithm was composed of the following two recursive processes. First, a low-entropy region in focal plane f at time point t is selected if the region overlaps with a region that has been selected in either focal plane f - 1 or f + 1 at time point t. Second, a low-entropy region at focal plane f at time point t is selected if the region overlaps with a region that has been selected in focal plane f at time point t - 1. Manual selection of a low-entropy region at time point 0 triggers these processes. We call this algorithm forward tracking because it tracks nuclei in the forward direction of time.
To examine whether forward tracking effectively selects low-entropy regions that can actually detect nuclei, we applied this algorithm to the low-entropy regions produced from five sets of 4D DIC microscope images of C. elegans embryos from the one- to 24-cell stages (Table 1). As expected, we obtained perfect sensitivity for nuclear detection by the selected low-entropy regions. All nuclei were detected at any time point, independently of whether or not they were in the process of cell division. Specificity was about 6.7 times better than before selection, although it was still far from perfect. These results indicate that forward tracking effectively selects low-entropy regions that can actually detect nuclei.
Further selection of low-entropy regions using object-tracking algorithm in the backward direction of time
To further select low-entropy regions, we used another tracking algorithm. This algorithm, called backward tracking, used the same recursive processes as forward tracking, with the exception of the direction of tracking, i.e., it tracked nuclei in the backward direction of time (Figure 4). We expected that this backward tracking would be effective for selecting low-entropy regions after forward tracking, because forward tracking usually creates many dead-end branches (Figure 4), which consist of low-entropy regions that do not detect nuclei. Backward tracking selected low-entropy regions that were not included in these dead-end branches (Figure 4).
Backward tracking was composed of the following two recursive processes. First, a low-entropy region in focal plane f at time point t is selected if the region overlaps with a region that has been selected in either focal plane f - 1 or f + 1 at time point t. Second, a low-entropy region in focal plane f at time point t is selected if the region overlaps with a region that has been selected in focal plane f at time point t + 1. Manual selection of low-entropy regions at the last time point triggers the processes.
To examine whether backward tracking is effective for selection of low-entropy regions after forward tracking, we applied backward tracking to the five sets of low-entropy regions selected by forward tracking (Table 1). Again, we obtained perfect sensitivity for nuclear detection by low-entropy regions selected by backward tracking. All nuclei were detected at any time point independently of whether or not they were in the process of cell division. Sensitivity was markedly better than before backward tracking, although it was still far from perfect. These results indicate that backward tracking is effective for selection of low-entropy regions after forward tracking.
Excellent selection of low-entropy regions using object-tracking algorithm, depending on the extent of overlap between two regions
The very high sensitivity but far lower perfect specificity (0.56 in average) of low-entropy regions selected by the combination of forward and backward trackings motivated us to develop a process that would more effectively select low-entropy regions that could detect nuclei. To develop this process, we used more detailed spatial and temporal information on the nucleus. In terms of more detailed spatial information, we expected the nucleus to be detected by several low-entropy regions, each of which overlapped to a large extent with one of the others in an adjacent focal plane at the same time point, because the 3D shape of the nucleus is usually simple. Therefore, a low-entropy region would become more likely to detect a nucleus when the region overlapped to a large extent with a region that detected the nucleus in an adjacent focal plane at the same time point. In terms of more detailed temporal information, we expected that a nucleus would be detected by several low-entropy regions, each of which overlapped to a certain extent with another in the same focal plane at two adjacent time points, because the nucleus usually moves much less than a distance equal to its diameter within the time equal to the interval between two adjacent time points. Therefore, a low-entropy region would become more likely to detect a nucleus when the region overlapped with a region that detected the nucleus in the same focal plane at two adjacent time points, and when both regions overlapped by a large extent.
To select low-entropy regions using this more detailed spatial and temporal information, we introduced a minimum overlap ratio to the forward and backward trackings. The minimum overlap ratio between two low-entropy regions was defined as the smallest ratio of the number of pixels shared by these two regions to the number of pixels making up each region. Thus, when the minimum overlap ratio between two overlapping regions increases, the two regions overlap to a greater extent, i.e., the two regions are more likely to detect the same nucleus. In the forward and backward trackings, we used this minimum overlap ratio to select pairs of low-entropy regions that overlapped to an extent greater than a prefixed value – i.e., pairs of low-entropy regions that were more likely to detect the same nucleus than a prefixed likelihood.
Forward tracking with a minimum overlap ratio was composed of the following two recursive processes. First, a low-entropy region in focal plane f at time point t is selected if the region overlaps with a region that has been selected either at focal plane f - 1 or f + 1 at time point t by a minimum overlap ratio more than the threshold Tf. Second, a low-entropy region in focal plane f at time point t is selected if the region overlaps with a region that has been selected in focal plane f at time point t - 1 by a minimum overlap ratio more than the threshold Tt. Manual selection of a low-entropy region at time point 0 triggers the processes in the same way as with the original forward tracking.
Backward tracking with a minimum overlap ratio is composed of the same recursive processes as forward tracking with a minimum overlap ratio, except that the direction of tracking is reversed – i.e., it tracks nuclei in the backward direction of time in the same way as with the original backward tracking. Manual selection of low-entropy regions at the last time point triggers the processes in the same way as with the original backward tracking. We expected that, as Tf and Tt increased, the selected low-entropy regions would become more likely to detect nuclei.
To examine whether the combination of forward and backward trackings with minimum overlap ratio (hereafter called advanced forward and backward trackings) would more effectively select low entropy regions than a combination of the original forward and backward trackings, we applied this combination of advanced forward and backward trackings to the low-entropy regions produced from five sets of 4D DIC microscope images of C. elegans embryos from the one- to 24-cell stages. Various sets of Tf and Tt were examined (Tables 2, 3). As expected, as Tf and Tt increased, the specificity of detection by the selected low-entropy region increased, whereas the sensitivity of detection by the region decreased. We found many sets of Tf and Tt that provided very high specificity (= 1.0), and several of them also provided perfect sensitivity (for example, Tf = 70% and Tt = 4% in Table 3). In this set of Tf and Tt, the selected low-entropy regions nearly perfectly detected all nuclei at any time point, independently of whether or not the nuclei were in the process of cell division. These results indicate that the combination of advanced forward and backward trackings more effectively selected low-entropy regions than did the combination of original forward and backward trackings. When an optimal set of Tf and Tt was applied, the combination of advanced forward and backward trackings nearly perfectly selected low-entropy regions that could detect nuclei.
Discussion
We developed a system that automates the detection of nuclei in a set of 4D DIC microscope images of C. elegans embryos. One major advantage of this system is the use of local image entropy to quantify the appearance of the nucleus in the images. Our previous system used edge detection operators to quantify the appearance of the nucleus [14]. Because these operators were sensitive to differences in image quality (e.g., brightness, contrast) among sets of images, the previous system required laborious hand-tuning of system parameters each time a new image set was used (see Additional file 1). Local image entropy is not sensitive to differences in image quality among sets of images because it represents the smoothness of the image texture (see Additional file 1). Therefore, our system can be applied to different image sets without the need to change the system parameters. We applied five sets of 4D DIC microscope images to our system, and the system detected the nuclei in these sets with similar sensitivity and specificity when we used the same parameter values (Table 1). This reduced sensitivity to differences in image quality makes our system applicable to research. We can apply this system to sets of 4D DIC microscope images of mutant C. elegans embryos (see Additional file 2) and embryos in which specific genes are silenced by RNA interference (see Additional file 3).
Another major advantage of our system is the use of object-tracking algorithms to examine all regions with the features of the image texture of the nucleus (i.e., low local image entropy) in a set of 4D DIC microscope images and to select regions that can actually detect nuclei. A DIC image of a C. elegans embryo contains many regions that have similar (smooth) image textures to that of the nucleus but that do not actually correspond to the nucleus, such as the boundaries between cells and the spaces between the embryo and the eggshell (Figure 3). Thus, in addition to image texture, other features of the nucleus are needed to completely distinguish the nucleus. Our previous system used the (round) shape of the nucleus that was not in the process of cell division in addition to the feature of image texture, as quantified by edge detection operators [14]. This previous system could not detect nuclei in the process of cell division. The object-tracking algorithm in our new system uses spatial and temporal information on the nucleus, and this information is independent of the process of cell division. Thus, our system detects all nuclei – whether or not the cell is dividing – at every time point from one- to 24-cell stages. This continuous detection of nuclei is a great help in following the cell division pattern of the embryo.
Our system effectively detected nuclei over a markedly longer developmental period than did the previous system, i.e., from the one- to 24-cell (Tables 2, 3) stages compared with only the two- to eight-cell stages [14]. This extension of the period of effective nuclear detection primarily results from the very high sensitivity of nuclear detection by low-entropy regions before forward and backward trackings (Table 1). The sensitivity and specificity of nuclear detection by these "original" low-entropy regions depend on the parameters used to produce the regions (i.e., window size and entropy threshold): the higher the sensitivity, the lower the specificity. Our system uses a set of values for these parameters that provides very high sensitivity and very low specificity of nuclear detection by the original low-entropy regions (Table 1), because subsequent forward and backward trackings effectively distinguish those regions that actually detect nuclei from those that do not.
The previous system used a two-step strategy similar to ours: i.e., regions that had the image texture of the nucleus were produced using edge detection operators, and from these "likely nuclear" regions, those that actually detected nuclei were selected using the shape of the nucleus. The sensitivity and specificity of nuclear detection by these "original" likely nuclear regions depended on the parameters used to produce the regions. However, the shape-dependent selection of likely nuclear regions was far less effective than the selection of low-entropy regions by forward and backward trackings. Thus, the previous system used a set of parameter values that provided markedly lower sensitivity and markedly higher specificity of nuclear detection by the original likely nuclear regions than by the original low-entropy regions. In the current study, we found very high sensitivity of nuclear detection by the original low-entropy regions up to the 44-cell stage (data not shown). Thus, improvement in the selection of low-entropy regions will further extend the period of effective nuclear detection. We are developing an improved system that uses both a tracking algorithm and the known shape and size of nuclei in non-dividing cells to select low-entropy regions.
Fluorescent labeling of nuclei is a method that has recently been developed for identifying the positions of the nuclei in living C. elegans embryos [24,25]. With this method, the genetic information of an embryo is artificially modified so that the embryo expresses nuclear protein fused with fluorescent protein, such as histone H2B fused with green fluorescent protein (GFP) [25]; the embryo is illuminated by excitatory light (e.g., blue or UV light for GFP), and the expressed fusion protein produces light of a specific color (e.g., green for GFP). Because the nuclei are labeled with a specific color, detection of the nuclei is much easier than that using the DIC microscope. However, the development of the embryo expressing the fusion protein may differ from that of the intact embryo because of the presence of GFP or the modification of genetic information [26-28]. Fluorescent labeling can be used to visualize nuclei for a markedly shorter period than with the DIC microscope because of photobleaching: i.e., the intensity of fluorescence of the fusion protein decreases because of exposure of the protein to the excitatory light [29], although the amount of photobleaching can be reduced by the use of multiphoton fluorescence imaging [30]. In contrast, the DIC microscope can be used to visualize the nuclei of an intact embryo throughout the development of C. elegans. Therefore, to describe the precise position of nuclei in living C. elegans embryos, identification of the position of the nucleus using the DIC microscope seems more suitable than that using fluorescent labeling of nuclei.
A major drawback of our system is the need for manual selection of low-entropy regions at the first and last time points. These manual operations may reduce the objectivity and productivity of our system, because selection is determined by the operator. However, slight differences in manual selection at the first and the last time points does not influence the automated selection of low-entropy regions in between these points, because the automated selections select all regions that overlap with other selected regions in the adjacent focal plane at the same time point or in the same focal plane at the adjacent time point. Thus, usually our system objectively detects nuclei in between the first and last time points. Manual selection of low-entropy regions at the first and the last time points could still reduce the productivity of our system, because these manual selections usually take about 10 min. However, our system still markedly increases the productivity of identification of the positions of the nuclei in C. elegans embryos, because manual selection of low-entropy regions for all time points from the one- to 24-cell stages (56 focal planes × ~120 time points = ~6720 images) takes more than 50 h. Our system needs about 135 min for computation (120 min for the production of low-entropy regions and 15 min for the forward and backward trackings) and 10 min for manual operations to detect all the nuclei in a set of 4D DIC microscope images of a C. elegans embryo recorded from the one- to 24-cell stages. These times for computation and manual operations are acceptable in research. The selection of low-entropy regions at the first and last time points will be automated, most likely by using known properties of nuclei, such as the known shapes and sizes of nuclei in non-dividing cells.
The low-entropy regions before selection by the forward and backward trackings failed to detect nuclei at around the 44-cell stage or later. Because the window size (10 × 10 pixels) and the threshold value (175) used in this experiment appear likely to be optimal for our system, the result indicates that the limit of the nuclear detection system presented here is around the 44-cell stage. We believe that this limit comes from the reduction in size of the cells during embryogenesis. As the size of the cells decreases during embryogenesis, the distance between the nucleus and cell membrane decreases. Usually at around the 44-cell stage, some nuclei are positioned so close to the cortex of the embryo that a 10 × 10 pixel window cannot produce a high-entropy (> 175) boundary between the nucleus and the image background; the texture of the image background is smooth (Figure 2), and thus the local image entropy in the image background is as low as that in the nucleus. In this situation, the low-entropy regions corresponding to the cortically positioned nucleus merge with the low-entropy regions corresponding to the image background. Because our nuclear detection system removes the low-entropy regions corresponding to the image background, the low-entropy regions produced by our system fail to detect the cortically positioned nucleus. To overcome this limitation, modulation of the window size and/or the threshold value depending on the embryonic stage and/or position of the nucleus within the embryo (central or cortical) might be effective. We observed that low-entropy regions produced using a smaller (< 10 × 10 pixel) window size and/or smaller (< 175) threshold value successfully discriminated between such cortically positioned nuclei and the image background in the later stages of embryogenesis.
Our system is applicable to research programs that require high objectivity and/or productivity of identification of the positions of the nuclei in C. elegans embryos. Because the sensitivity and specificity of nuclear detection by our system depend on the thresholds for minimum overlap ratios (Tf and Tt), the values of these thresholds should be specified when the system is applied to a specific study. We often use Tf = 70% and Tt = 4%, because sensitivity is often more important than specificity in our research. We applied this system to our automated cell division pattern measurement system for C. elegans embryos; the measurement system was used in our large-scale cell division pattern analysis of gene-knockout C. elegans embryos [31]. The cell division pattern analysis will provide new opportunities for bioinformatics in studies of the development of multicellular organisms [32]. In addition, this system has been used to measure the positions of the male pronucleus (the sperm-derived nucleus) in a very early C. elegans embryo; the measurements were compared with computer simulations to determine the mechanism that specifies the positions of the male pronucleus during the very early period of C. elegans development [33]. To calculate the precise 3D shape and/or position of a nucleus from the low-entropy regions produced by this system, we need to consider the DIC shear angle, because the angle makes a substantial artifact in DIC images [34]. Because of its high objectivity and productivity of measurement, our system will contribute greatly to studies of the development of multicellular organisms.
Conclusion
We have presented a system that automates the detection of nuclei in a set of 4D DIC microscope images of C. elegans embryos. The system can be applied to multiple image sets without the need to change parameter values. It can be used to detect nuclei that are in the process of cell division and can detect nuclei with very high sensitivity and specificity from fertilization to the onset of gastrulation, i.e., from the one- to 24-cell stages, enabling highly objective and productive identification of the positions of nuclei in C. elegans embryos. The system is applicable to comparisons between in vivo measurement and computer simulation and to systematic cell division pattern analysis of knockout embryos.
Methods
Preparation of 4D DIC microscope images of C. elegans embryos
The Bristol N2 C. elegans was cultured under standard conditions [35]. An embryo immediately after fertilization (before meeting of the female and male pronuclei) was dissected from a hermaphrodite and mounted on a 2% agar pad on a glass slide, covered with a coverslip, and sealed with petroleum jelly. Nomarski DIC images were obtained using a Leica DMRE microscope equipped with an HCX PL APO 100×/1.40 NA objective, whose illumination intensity and objective-side Wollaston prism were adjusted to obtain images of the same quality. Digital images of 600 × 600 pixels with 256 gray levels (0.1 μm per pixel) were recorded with an ORCA CCD Camera (Hamamatsu Photonics), and the recording system was controlled by IP Lab 3.5 software (Scanalytics). Digital images of the developing embryo were recorded at 22°C in 56 focal planes, with a distance of 0.5 μm between two focal planes, and a set of 56 focal plane images was recorded every 40 s for 2 h.
Calculation of sensitivity and specificity of nuclear detection
Low-entropy regions that actually detected nuclei were manually selected in five sets of 4D DIC microscope images of an embryo. The resulting five sets of low-entropy regions were used as references to calculate the sensitivity and specificity of nuclear detection. Sensitivity was calculated using low-entropy regions of all time points from time point 0 to that corresponding to the end of the 24-cell stage. Specificity was calculated using low-entropy regions at 11 time points, obtained by sampling every 10 time points from the beginning of the two-cell stage to the end of the 24-cell stage. The number of time points from the beginning of the two-cell stage to the end of the 24-cell stage was 106 on average.
Hardware and software environment
Because we needed to process many images, low-entropy regions were produced from sets of 4D DIC microscope images using a Beowulf-class PC cluster [36] consisting of 48 nodes, each of which used a 2-GHz Intel Pentium 4 processor, 1 GB of SDRAM memory, and a 100 Base-TX Ethernet card. Parallel Virtual Machine (PVM) software [37] was used for communications between the nodes. In our implementation, a single image in an image set was processed by a single CPU in the cluster. The forward and backward trackings for selection of low-entropy regions were processed on a single processor PC (2.2 GHz Intel Pentium 4 processor and 1 GB of RDRAM memory). The programs were written in C.
Availability
Our software implementations are readily available on the web at .
Authors' contributions
SH participated in conception, designed the system, tested the system, and drafted the manuscript. SO conceived of the study, contributed to the design, tested the system, and drafted the manuscript. HK participated in system design. All authors read and approved the final manuscript.
Supplementary Material
Additional File 1
Detailed discussions about various measures of image texture.
Click here for file
Additional File 2
Figure 5 Low-entropy regions in a par-1 embryo. Input images of a par-1 embryo are shown in the left column and low-entropy regions (black) are shown in the right column. Bar is 10 μm.
Click here for file
Additional File 3
Figure 6 Low-entropy regions in a tba-2 (RNAi) embryo. Input images are shown in the left column and low-entropy regions (black) are shown in the right column. Bar is 10 μm.
Click here for file
Acknowledgements
We are grateful to K. Kemphues for providing the par-1 (KK288) mutant. We thank J. Ahringer, Cambridge University Technical Services Limited and UK MRC Human Genome Mapping Project Resource Center, for providing C. elegans chromosome 1 RNAi library. We thank M. Urai for help with software development and Y. Kitamura for support in experiments. We also thank H. Amano, K. Oka, A. Kimura, K. Kyoda, M. Morohashi, and all members of the Onami lab for their discussions and advice. This work was supported in part by Special Coordination Funds for Promoting Science and Technology (to SO and HK) and by a Grant-in-Aid for Scientific Research on Priority Areas (to SO), from the Ministry of Education, Culture, Sports, Science and Technology of Japan.
Figures and Tables
Figure 1 Overview of 4D DIC microscope images of C. elegans embryo. (A) Schematic of 4D DIC microscope images. Digital images of a developing embryo were recorded in multiple focal planes and a set of multifocal images was recorded with a fixed time interval, α. (B) Example of 4D DIC microscope images of a C. elegans embryo. Each column shows multifocal images recorded at a specific time point, with 3.5 μm between two focal planes. Each row shows time-lapse images recorded in a specific focal plane with 160 s between two time points. Bar is 10 μm.
Figure 2 Effect of window size and threshold value on production of low-entropy regions. (A) Input image. Low-entropy regions were produced from an image of a four-cell-stage embryo using various window sizes and threshold values. (B–D) Effect of window size on image conversion using local image entropy. The input image was applied to the image conversion using window sizes of 4 × 4 (B), 10 × 10 (C), and 50 × 50 (D) pixels. Darker colors represent lower local image entropies. (E–M) Effect of window size and threshold value on low-entropy regions. Low-entropy regions (black) were produced using threshold values of 200 (E, H, K), 175 (F, I, L), or 150 (G, J, M) from the images resulting from the image conversion, using window sizes of 4 × 4 (E–G), 10 × 10 (H–J) or 50 × 50 (K–M) pixels. A window of 10 × 10 pixels corresponds to that of 1 μm × 1 μm.
Figure 3 Low-entropy regions of different focal planes and different time points. (A – E and K–O) Input images. (F–J and P–T) Low-entropy regions (black) produced from the input images. Low-entropy regions were produced from multifocal images of an embryo at a specific time point in the four-cell stage with 4.5 μm between two focal planes (A–E) and from time-lapse images of an embryo at a specific focal plane with 30 min between two time points (K–L). The low-entropy regions produced are displayed to the right of each input image. The window size was 10 × 10 pixels and the threshold value was 175.
Figure 4 Overview of forward and backward trackings. Low-entropy regions selected by forward tracking are shown in the left column and those selected by backward tracking after the selection by forward tracking are shown in the right column. These low-entropy regions (white) are overlaid on their corresponding input images. Solid arrows represent the tracking of low-entropy regions. The broken arrow represents a dead-end branch of tracking.
Table 1 Performance of nuclear detection by low-entropy regions and those selected by forward and backward trackings
set 1 set 2 set 3 set 4 set 5
Low-entropy regions 0.06 (1.0) 0.08 (1.0) 0.06 (1.0) 0.08 (1.0) 0.08 (1.0)
Forward Tracking 0.38 (1.0) 0.48 (1.0) 0.55 (1.0) 0.44 (1.0) 0.53 (1.0)
Forward and backward trackings 0.44 (1.0) 0.55 (1.0) 0.67 (1.0) 0.50 (1.0) 0.63 (1.0)
Data are specificity and (sensitivity) for five sets of 4D DIC microscope images of a C. elegans embryo from the one- to 24-cell stages (1 set = 56 focal planes × ~120 time points = ~6720 images).
Table 2 Specificity of nuclear detection by low-entropy regions selected by advanced forward and backward trackings
Tf Tt
1 pixel† 4% 8% 12% 16% 20%
Forward tracking 1 pixel‡ 0.48 (0.06) 0.54 (0.11) 0.56 (0.11) 0.56 (0.10) 0.58 (0.10) 0.59 (0.10)
30% 0.57 (0.08) 0.69 (0.11) 0.74 (0.11) 0.78 (0.10) 0.84 (0.09) 0.87 (0.07)
50% 0.66 (0.11) 0.79 (0.11) 0.84 (0.08) 0.87 (0.07) 0.90 (0.06) 0.94 (0.03)
70% 0.74 (0.12) 0.86 (0.08) 0.90 (0.07) 0.93 (0.05) 0.94 (0.05) 0.97 (0.03)
90% 0.83 (0.10) 0.93 (0.04) 0.95 (0.03) 0.98* (0.01*) 0.99* (0.01*) 1.00* (0.00*)
Forward and backward trackings 1 pixel‡ 0.56 (0.08) 0.62 (0.13) 0.64 (0.13) 0.65 (0.12) 0.67 (0.13) 0.69 (0.13)
30% 0.70 (0.10) 0.83 (0.12) 0.87 (0.09) 0.90 (0.07) 0.93 (0.05) 0.95 (0.03)
50% 0.82 (0.12) 0.91 (0.07) 0.93 (0.05) 0.95 (0.05) 0.97 (0.04) 0.98 (0.02)
70% 0.87 (0.11) 0.96 (0.05) 0.97 (0.04) 0.99 (0.02) 0.99 (0.01) 1.00 (0.00)
90% 0.93 (0.08) 0.99 (0.01) 1.00 (0.00) 1.00* (0.00*) 1.00* (0.00*) 1.00* (0.00*)
Data are mean and (SD) for five sets of 4D DIC microscope images of a C. elegans embryo from the one- to 24-cell stages.
*Data are for four sets of DIC microscope images, because no low-entropy regions were selected in one image set.
† Low-entropy regions were selected when they shared at least one pixel with regions already selected in the same focal plane at an adjacent time point, i.e., the same condition as used for the original forward and backward trackings.
‡ Low-entropy regions were selected when they shared at least one pixel with regions already selected in an adjacent focal plane at the same time point, i.e., the same condition as used for the original forward and backward trackings.
Table 3 Number of image sets for which advanced trackings detected nuclei with perfect sensitivity
Tt
1 pixel 4% 8% 12% 16% 20%
Tf 1 pixel 5 5 5 4 4 4
30% 5 5 4 4 4 3
50% 5 5 4 4 2 1
70% 5 5 4 4 2 1
90% 1 1 1 1 0 0
Sensitivities of nuclear detection by low-entropy regions selected by advanced forward and backward trackings were calculated for five sets of 4D DIC microscope images of a C. elegans embryo from the one- to 24-cell stages. Number of image sets for which sensitivity of nuclear detection were 1.0 is shown.
==== Refs
Sulston JE Schierenberg E White JG Thomson JN The embryonic cell lineage of the nematode Caenorhabditis elegans Dev Biol 1983 100 64 119 6684600 10.1016/0012-1606(83)90201-4
Bhan A Galas DJ Dewey TG A duplication growth model of gene expression networks Bioinformatics 2002 18 1486 1493 12424120 10.1093/bioinformatics/18.11.1486
Sprague BL Pearson CG Maddox PS Bloom KS Salmon ED Odde DJ Mechanisms of microtubule-based kinetochore positioning in the yeast metaphase spindle Biophys J 2003 84 3529 3546 12770865
Blattner FR Plunkett G Bloch CA 3rdPerna NT Burland V Riley M Collado-Vides J Glasner JD Rode CK Mayhew GF Gregor J Davis NW Kirkpatrick HA Goeden MA Rose DJ Mau B Shao Y The complete genome sequence of Escherichia coli K-12 Science 1997 277 1453 1462 9278503 10.1126/science.277.5331.1453
The C.elegans Sequencing Consortium Genome sequence of the nematode C.elegans : a platform for investigating biology Science 1998 282 2012 2018 9851916 10.1126/science.282.5396.2012
Hughes TR Marton MJ Jones AR Roberts CJ Stoughton R Armour CD Bennett HA Coffey E Dai H He YD Kidd MJ King AM Meyer MR Slade D Lum PY Stepaniants SB Shoemaker DD Gachotte D Chakraburtty K Simon J Bard M Friend SH Functional discovery via a compendium of expression profiles Cell 2000 102 109 126 10929718 10.1016/S0092-8674(00)00015-5
Riddle DL Blumenthal T Meyer BJ Priess JR (Eds) C ELEGANS II 1997 New York: Cold Spring Harbor Laboratory Press
Wood W (Ed) The Nematode Caenorhabditis elegans 1988 New York: Cold Spring Harbor Laboratory Press
Nomarski G Weill AR Application à la métallographie des méthodes interférentielles à deux ondes polarisées Rev Metall 1955 2 121 128
Hird SN White JG Cortical and cytoplasmic flow polarity in early embryonic cells of Caenorhabditis elegans J Cell Biol 1993 121 1343 1355 8509454 10.1083/jcb.121.6.1343
Thomas C DeVries P Hardin J White J Four-dimensional imaging: computer visualization of 3D movements in living specimens Science 1996 273 603 607 8662545
Schnabel R Hutter H Moerman D Schnabel H Assessing normal embryogenesis in Caenorhabditis elegans using a 4D microscope: variability of development and regional specification Dev Biol 1997 184 234 265 9133433 10.1006/dbio.1997.8509
Heid PJ Voss E Soll DR 3D-DIASemb: a computer-assisted system for reconstructing and motion analyzing in 4D every cell and nucleus in a developing embryo Dev Biol 2002 245 329 347 11977985 10.1006/dbio.2002.0631
Yasuda T Bannai H Onami S Miyano S Kitano H Towards automatic construction of cell-lineage of C.elegans from Nomarski DIC microscope images Genome Inform Ser Workshop Genome Inform 1999 10 144 154 11072351
Kirsch R Computer determination of the constituent structure of biological images Comput Biomed Res 1971 4 315 328 5562571 10.1016/0010-4809(71)90034-6
Prewitt J Lipkin B, Rosenfeld A Object enhancement and extraction Picture Processing and Psychopictorics 1970 New York: Academic Press 75 149
Handmann U Kalinke T Tzomakas C Werner M Seelen WV An image processing system for driver assistance Image Vision Comput 2000 18 367 376 10.1016/S0262-8856(99)00032-3
Awasthi V Doolittle KW Parulkar G McNally JG Cell tracking using a distributed algorithm for 3-D image segmentation Bioimaging 1994 1 98 112 10.1002/1361-6374(199406)2:2<98::AID-BIO4>3.3.CO;2-X
Geerts H De Brabander M Nuydens R Geuens S Moeremans M De Mey J Hollenbeck P Nanovid tracking: a new automatic method for the study of mobility in living cells based on colloidal gold and video microscopy Biophys J 1987 52 775 782 3427186
Lee GM Ishihara A Jacobson KA Direct observation of brownian motion of lipids in a membrane Proc Natl Acad Sci U S A 1991 88 6274 6278 1712486
Pratt WK Digital Image Processing 1991 2 New York: John Wiley & Sons
Tuceryan M Jain AK Chen CH, Pau LF, Wang PSP Texture Analysis The Handbook of Pattern Recognition and Computer Vision 1998 2 New Jersey: World Scientific Publishing Co 207 248
Otsu N A thresholding selection method from gray-level histogram IEEE Trans Sys Man Cybern 1979 9 62 66
Kelly WG Xu S Montgomery MK Fire A Distinct requirements for somatic and germline expression of a generally expressed Caenorhabditis elegans gene Genetics 1997 146 227 238 9136012
Praitis V Casey E Collar D Austin J Creation of low-copy integrated transgenic lines in Caenorhabditis elegans Genetics 2001 157 1217 1226 11238406
Fire A Integrative transformation of Caenorhabditis elegans EMBO J 1986 5 2673 2680
Liu HS Jan MS Chou CK Chen PH Ke NJ Is green fluorescent protein toxic to the living cells? Biochem Biophys Res Commun 1999 260 712 717 10403831 10.1006/bbrc.1999.0954
Zhang J Campbell RE Ting AY Tsien RY Creating new fluorescent probes for cell biology Nat Rev Mol Cell Biol 2002 3 906 918 12461557 10.1038/nrm976
Strome S Powers J Dunn M Reese K Malone CJ White J Seydoux G Saxton W Spindle dynamics and the role of gamma-tubulin in early Caenorhabditis elegans embryos Mol Biol Cell 2001 12 1751 1764 11408582
Denk W Strickler JH Webb WW Two-photon laser scanning fluorescence microscopy Science 1990 248 73 76 2321027
Onami S Hamahashi S Nagasaki M Miyano S Kitano H Kitano H Automatic acquisition of cell lineage through 4D microscopy and analysis of early C. elegans embryogenesis Foundations of Systems Biology 2001 Cambridge, Massachusetts: The MIT Press 39 55
Braun V Azevedo RB Gumbel M Agapow PM Leroi AM Meinzer HP ALES: cell lineage analysis and mapping of developmental event Bioinformatics 2003 19 851 858 12724295 10.1093/bioinformatics/btg087
Kimura A Onami S Computer simulations and image processing reveal length-dependent pulling force as the primary mechanism for C. elegans male pronuclear migration Dev Cell 2005 8 765 775 15866166 10.1016/j.devcel.2005.03.007
Preza C Snyder DL Theoretical development and experimental evaluation of imaging models for differential-interference-contrast microscopy, J Opt Soc Am A 1999 16 2185 2199
Brenner S The genetics of Caenorhabditis elegans Genetics 1974 77 71 94 4366476
Sterling T Savarese D Becker DJ Dorband JE Ranawake UA Packer CV BEOWULF: a parallel workstation for scientific computation Proceedings of the 1995 International Conference on Parallel Processing: August 14–18, 1995 1995 1 Boca Raton: CRC Press 11 14
Sunderam V PVM: a framework for parallel distributed computing Concurrency Pract Exp 1990 2 315 339
| 15910690 | PMC1175842 | CC BY | 2021-01-04 16:02:49 | no | BMC Bioinformatics. 2005 May 24; 6:125 | utf-8 | BMC Bioinformatics | 2,005 | 10.1186/1471-2105-6-125 | oa_comm |
==== Front
BMC BioinformaticsBMC Bioinformatics1471-2105BioMed Central London 1471-2105-6-1351592978910.1186/1471-2105-6-135SoftwareYANA – a software tool for analyzing flux modes, gene-expression and enzyme activities Schwarz Roland [email protected] Patrick [email protected] Kamp Axel [email protected] Bernd [email protected] Heiner [email protected] Stefan [email protected] Thomas [email protected] Dept of Bioinformatics, Biocenter, University of Würzburg; Germany2 Dept of Theoretical Chemistry, Organikum, University of Würzburg, Germany3 Dept of Bioinformatics, University of Jena, Germany4 Center for Biochemistry (BZH), University of Heidelberg, Germany5 Structural and Computational Biology, EMBL, Heidelberg, Germany2005 1 6 2005 6 135 135 3 2 2005 1 6 2005 Copyright © 2005 Schwarz et al; licensee BioMed Central Ltd.This is an Open Access article distributed under the terms of the Creative Commons Attribution License (), which permits unrestricted use, distribution, and reproduction in any medium, provided the original work is properly cited.
Background
A number of algorithms for steady state analysis of metabolic networks have been developed over the years. Of these, Elementary Mode Analysis (EMA) has proven especially useful. Despite its low user-friendliness, METATOOL as a reliable high-performance implementation of the algorithm has been the instrument of choice up to now. As reported here, the analysis of metabolic networks has been improved by an editor and analyzer of metabolic flux modes. Analysis routines for expression levels and the most central, well connected metabolites and their metabolic connections are of particular interest.
Results
YANA features a platform-independent, dedicated toolbox for metabolic networks with a graphical user interface to calculate (integrating METATOOL), edit (including support for the SBML format), visualize, centralize, and compare elementary flux modes. Further, YANA calculates expected flux distributions for a given Elementary Mode (EM) activity pattern and vice versa. Moreover, a dissection algorithm, a centralization algorithm, and an average diameter routine can be used to simplify and analyze complex networks. Proteomics or gene expression data give a rough indication of some individual enzyme activities, whereas the complete flux distribution in the network is often not known. As such data are noisy, YANA features a fast evolutionary algorithm (EA) for the prediction of EM activities with minimum error, including alerts for inconsistent experimental data. We offer the possibility to include further known constraints (e.g. growth constraints) in the EA calculation process. The redox metabolism around glutathione reductase serves as an illustration example. All software and documentation are available for download at .
Conclusion
A graphical toolbox and an editor for METATOOL as well as a series of additional routines for metabolic network analyses constitute a new user-friendly software for such efforts.
==== Body
Background
Elementary mode analysis (EMA) analyzes complex metabolic networks
Metabolic networks include many enzymes. These operate together in a complex way as metabolites of one reaction may be processed (consumed or provided) by a number of different enzymes. Whereas in biochemistry textbooks such networks are often described as linear pathways or simple, separate subnetworks, real metabolic webs show an astonishing complexity regarding the number of possible routes a metabolite can take through the network.
EMA is an algorithm that systematically enumerates all possibilities how enzymes can operate together without violating the steady state condition of the system (see below). Using EMA, complex networks can be analyzed in terms of contained pathways, robustness, central enzymes, medical targets, optimum yield and effector compounds, such as signaling phospholipids, with interesting applications in medicine and biotechnology [1].
EMA – algorithm and related approaches
To perform a holistic network analysis, the stoichiometric and thermodynamic feasibility of all possible pathways has to be tested. We therefore assume the system to be in a steady-state, in which intermediate or internal metabolites are balanced [2]. Their concentrations do not change in the timescale of study as the amount of production of these metabolites equals their consumption.
To find all pathways through a given network we look for all vectors v of enzyme coefficients, the so called flux vectors or flux distributions, which satisfy the steady-state condition of
N*v = 0 (1)
for all internal metabolites (stoichiometric feasibility). Here, N is the m × r stoichiometric matrix of the system with m being the number of metabolites in the system and r being the number of reactions (in eq. (4), upper case R is used). To solve such systems under consideration of additional irreversibility constraints imposed by the reactions in the system (thermodynamic feasibility), the mathematical theory of convex analysis [3] is used to project the equation above and the irreversibility constraints into what is called a pointed convex polyhedral cone. This approach is used by several algorithms to determine the possible pathways through the system, out of which recent analyses have focused on two concepts [4]: Extreme Pathways [5] and Elementary Mode Analysis (EMA) [2].
Both algorithms return the edges of the calculated cone, the convex basis, as pathways. In addition, EMA returns all possible non-decomposable pathways through the network, the so called Elementary Modes (EMs) or (Elementary) Flux Modes. Both methods yield a complete description of the metabolic network in which every concrete "state" of the system can be described as a non-negative linear combination of the set of pathways or EMs returned.
Elementary Mode Analysis has been successfully applied to numerous biochemical systems [6-8] and its capability to determine maximum conversion yields [9] and minimal cut sets [10] of biochemical systems makes it an important tool to predict the effect of enzyme inactivations, pharmacological effects, growth calculations and biotechnological applications [11]. We previously developed the software METATOOL [12] as an implementation of the Elementary Mode Analysis and enhanced EMA further by developing new techniques to cope with complex networks. These include the dissection of the network at metabolites with especially high connectivity [13] and an approach to reduce the complexity of the network by systematic variation of the internal and external status of the involved metabolites, thus reducing the number of EMs returned [14].
Development and advantages of YANA
METATOOL represents an efficient implementation of the EMA algorithm and has been integrated as an analysis option in large software packages such as GEPASI [15]. However, as a command line driven program, it lacks the comfort and usability of a graphical user interface (GUI) as well as the ability to perform further analyses out of the box. Alternatively, phpMetatool [16] provides some predefined biochemical networks but offers no further analysis options or processing of the METATOOL results. The program FluxAnalyzer [17] provides a graphical interface and some processing of elementary modes, for example, computation of minimal cut sets [10]. For dissection of complex networks, other tools such as SEPARATOR [13] have to be installed and import and export data using the METATOOL text file format. This gives rise to several common data-exchange and formatting problems.
YANA offers now an integrated modeling environment with standardized data exchange capabilities. It is a platform for integrating future analysis modules and includes strategies to address one of the most important issues in current metabolic modeling, the combinatorial explosion of EMs in complex networks. Further, it allows the convenient modification editing of metabolic networks with a comfortable user interface and the possibility of performing EMA analyses using the established METATOOL algorithm. It additionally incorporates strategies to reduce network complexity by using threshold operations on the metabolites and brings a variety of visualization options for concrete flux distributions of a network. It calculates for a user-defined EM activity pattern the resulting flux distribution, and is further able to predict a valid EM activity pattern from a given flux distribution even when only few or inaccurate enzyme activity data are available from experiments.
Implementation
Elementary Mode Analysis
To perform pathway analysis on the network under study, YANA acts as a front-end to METATOOL and computes the Elementary Modes of a network if the following information is provided:
• Metabolites used in the system, including information whether they are treated as internal or external
• Enzymes / reactions involved in the network, including their substrates and products and irreversibility constraints
Parsing the output from METATOOL, YANA shows a tabular overview of the calculated EMs and is able to display detailed information for each of the EMs obtained, including partaking reactions, their reaction equations, as well as the overall net reactions of the Elementary Modes.
Additionally the average diameter (path length) of the EMs is displayed, an information of importance for the dissection of larger networks.
Reducing network complexity
To prevent combinatorial explosion of the number of EMs in well connected networks, YANA offers and implements two strategies to reduce network complexity. Both change systematically the internal / external status of the metabolites using their connectivity values as the basic criterion.
In the first strategy [13], the network is divided by automatically setting metabolites with a connectivity value above a certain user-defined threshold as "external". This results in a split of the metabolic network into sub-networks, which are convenient to analyze. Individual sub-networks can be obtained using the program SEPERATOR [13], and the new routine from YANA directly gives the resulting complete but simplified network.
Alternatively, YANA offers the option to set all metabolites with a connectivity value below the threshold as external. In this way, only connections between the core nodes of a metabolic system are included, neglecting those on the outskirts. The resulting pathway set still holds the most important EMs, shortened and focused on the central hub metabolites [18].
To get an estimate on the average size of the metabolic network before and after dissection the average diameter (path length) for the modes can be used.
Translating EM activities into flux distributions
As described in the background section, convex analysis returns the spanning vectors of the cone that describes the solution to the steady-state equation system and thus every actual flux distribution (vector v in equation 1) is a linear combination of the obtained EMs.
By assigning an activity value in percent to each EM, except for a scalar factor, every flux distribution possible for the system can be reached. These valid flux distributions, or flux vectors v, hold an integer value for each enzyme in the system. Those values, which are responsible for sustaining the steady state in the system (v satisfies equation 1 for all internal metabolites) represent the relative flux through the respective enzyme and thus must be a combination of both the reaction velocity of the enzyme (real enzyme kinetics) and the amount of protein available.
YANA offers the possibility to compute flux distributions both in absolute (ignoring reaction directions) and relative values. This is done by iterating over all calculated modes, and summing the absolute (eq. 2) or relative (eq. 3) flux coefficients of each enzyme multiplied with the activity of the EM.
The resulting flux distribution is visualized and presented to the user either in form of graphical bar and pie charts (Figure 2) or in tabular form (Table 4).
Calculating flux distributions from a given theoretical set of EM activities is important e.g. to estimate the relevance of an enzyme in a metabolic network [19], but it would be a desirable goal if one could somehow measure flux distributions in living cells and map them onto the EM activities [20]. It would then be possible to take an experimental snapshot of the system and from it derive actual pathway activities.
Translating (partially) known flux distributions into EM activities
To obtain flux distributions of a living cell, one could either measure metabolite fluxes directly or estimate fluxes from protein quantification and enzyme turnover rates. Protein amounts will, in practice, be measured either by proteomics or gene expression. For the latter, an estimate from a comprehensive RNA and protein expression analysis in yeast indicates that for each mRNA copy on average there are 4000 molecules of synthesized protein found [21], with individual variation depending on mRNA stability, translatory regulation and promotor activities. To measure all these different factors involved in expression levels and, further, the enzyme activity itself is a non-trivial undertaking. There is a complex interplay between mRNA expression level, protein expression level, enzyme activity level and resulting metabolite fluxes to get optimal responses to different environmental conditions. The user should take into account that any of these expression levels are only crude estimates for the other levels and their effects. However, for most practical purposes it is sufficient if the user knows roughly the order of activity differences between the modeled enzymes, and which important regulatory signals influencing expression levels have to be considered (e.g. an unstable protein or unstable mRNA for a given enzyme should not be neglected). Taking these variables into account, flux distributions can be estimated.
To find the resulting EM activities from the estimated or observed flux distributions, there are, in general, many solutions possible. One could first choose a certain preferred flux mode, adjust its activity, and try to fit the observed flux distribution as accurately as possible. Next, select the next mode and so on. As the modes can be given by different preference schemes, it is absolutely possible that several schemes will fit the distribution equally well.
To find a rational and compact criterion for mode selection, here, we have chosen to first select the modes which are the shortest. It has already been shown in an earlier work [22] that these are the modes which contribute most to gene expression, at least in the central metabolism of E. coli (these are actually preferred to be kept by the well connected metabolite choosing routine above). In addition, metabolic webs have been shown to grow selectively around central "hub" metabolites to favor short metabolic paths [18,23].
For calculating EM activities from observed or estimated flux distributions, there are analytical treatments possible based on criteria other than pathway length [20]. However, all experimental measurements have errors. In particular, this applies to gene expression data where detection problems, background and standardization are routine challenges. Similarly, proteome data are selective, and protein levels measured are influenced by factors such as gel resolution, multi-spot detection and similar technicalities. As protein quantifications can only be measured with certain error margins and asinformations about enzyme turnover rates are not always accurate, we do not demand an exact solution to a flux distribution found experimentally (see e.g. Ref [20] for recent advances in this area). We focus on minimizing the difference to the target flux distribution instead. The error function uses the sum of squared differences between calculated enzyme activities (EC) and target enzyme activities (ET) which is to be minimized and, with R being the number of reactions in the system:
To achieve a fast and robust convergent solution for this error criterion, an evolutionary algorithm (EA) was successfully applied here. The algorithm starts to calculate flux distributions, even if only one enzyme activity or very few are known. A Pareto-optimal solution [24] for such limited experimental data is also found by the evolutionary strategy implemented in YANA.
The algorithm uses a randomly initialized population of 100 individuals with a per feature mutation probability M. This depends on the number of features F taken into account and the number of iterations I already run through, thus introducing a cool-down factor to the mutation probability scaling logarithmically with the number of time steps.
Furthermore, recombination between individuals is achieved by uniform crossover, randomly selecting one of the individuals as a parent for the feature in question. Selection pressure is induced by calculating a rank-based fitness from the square deviation of each individual to the target distribution thus giving each individual a probability R to take part in the recombination process that depends upon its rank r in the population and the population size P.
The evolutionary algorithm routine allows the inclusion of further fitness parameters and helps to fit enzyme activities including these additional constraints. For example, these could be (i) correlations between enzyme expression levels (or just ratios) according to gene-chip experimental results or (ii) constraints based on biochemical data and knowledge on enzyme activities; (iii) metabolite constraints, for instance production of certain amino acids has to be above a certain threshold (given by growth demands or again experimental data), (iv) genetic constraints (certain enzyme genes are known to have modified expression or enzymatic activities), (v) necessary activity or certain levels for specific enzyme pools (e.g. all enzymes connected to redox protection). The fitness function is implemented in such a way, that further positive or negative functions can easily be added with desired weights to the total fitness by the user. Also in that case, the evolutionary algorithm searches for the best possible solution describing the enzyme fluxes with minimum error according to observed enzyme expression data, while including these additional constraints.
In each refinement step, the flux mode fitting routine implemented in YANA selects the shortest modes. If two modes are equal, it picks the better connected inside the network, counting all metabolites according to the reaction they participate in.
For an overview of EA performance, see Results section c).
Implementation details
For the development of the YANA software package Borland JBuilder 2005 was used in combination with EJ-technologies real-time profiling suite JProfiler. The profiling of the software and the evolutionary algorithm in particular was carried out on a dual Intel Xeon 3.06 GHz CPU with Hyper Threading and 8 GB of RAM. For development and testing, a standard PC with a single 1.7 GHz Pentium 4 processor was used.
The YANA program flow includes the initial editing of a metabolic network in terms of enzymes and their respective substrates and products or, alternatively, loading networks from files in the METATOOL or SBML format.
Subsequently, elementary mode analysis is performed by YANA using the provided precompiled METATOOL binaries. It then parses the resulting output file and presents the arising elementary modes to the user, giving the possibility to retrieve detailed information about a specific elementary mode including partaking enzymes, net reactions, and the stoichiometry involved.
Once the analysis is performed, flux distributions can be visualized in several presentation formats simulating either a virtual spot intensity on a gel, or displaying comparative enzyme copy numbers in a virtual cell. From there on, the user can enter a target flux distribution and YANA tries to adjust the elementary mode activities to approximate the entered flux distribution using the EA.
In any part of the program, diagrams and tables can be printed and exported to graphic files in the portable network graphics (PNG) format or into text files using comma separated values (CSV) for easy import in e.g. Microsoft Excel.
To gain the advantage of platform independence, YANA was implemented using the Java SDK 1.5 and we provide, with our download precompiled METATOOL, binaries for both Windows and UNIX systems. The graphical user interface is based on the SWING Java framework, making strict use of the model-view-controller (MVC) paradigm.
To further add to the usability of the program, support for the Systems Biology Markup Language [25] (SBML Level 2 [26]) was integrated, an XML-based file format which enjoys increasing popularity in current bioinformatics and chemical applications. The software is also able to import and export data in the traditional METATOOL file format.
Results
a1) YANA program package
Required Inputs
(i) YANA analyses metabolic networks using EMA:
The required input for YANA (and the integrated METATOOL software) to perform an EMA is the set of enzymes and metabolites in the network under study. Their specific reactions and reversibility can be obtained from textbooks and databases. Metabolites must be defined as internal or external according to available biochemical knowledge. Kinetic data, RNA or protein expression data are not required for this part of the analysis.
(ii) For the calculation of flux distributions, the user has to choose as input the activities of the different EMs. If there is no information on this available, YANA assumes all EMs to be equally active. For accurate predictions of enzyme activities experimental data on flux ratios is helpful.
(iii) To analyze how the predicted elementary modes from step 1 fit expression data, some experimental data on protein or gene expression are required. For most accurate predictions in this step, kinetic data on enzyme activities, on translation speed, protein stability and transcription are required. However, YANA needs, as minimum information for predictions, only the estimated activity levels of some of the enzymes involved. Nevertheless, it calculates an optimal solution, mapping the information on enzyme activity available to a predicted EM activity pattern.
Program usage and outputs
Output
(i) All pathways in the network are calculated, a list of EMs is given, both as enzyme cascades as well as the overall reactions of the elementary modes with educts and products.
(ii) Calculation of specific flux distributions and visualization in form of pie and bar charts and in tabular form.
(iii) A specific EM activity pattern best fitting the user given flux distribution with preferably high activities on short EMs.
Usage
The YANA main screen is divided into two parts. On the left hand side, the user is able to enter the metabolites involved in the network with information about whether they are considered internal or external. On the right hand side, the enzymes are defined using the metabolites entered earlier. User actions are usually invoked using the toolbar at the top of the YANA main window. By clicking the analysis button, elementary mode analysis is performed, showing the results in a new frame in a tabular form. YANA gives the user detailed information about the elementary flux modes, their irreversibility constraints, partaking enzymes and so on. The user has the possibility to adjust the activity of each elementary mode on a percent scale using the slider incorporated into the table. By clicking the diagram button, YANA calculates the enzyme activity pattern using the activity vector entered before and displays the results either in a diagram or table format. The flux calculation button brings up a new screen on which the user can enter the target flux distribution for use with the evolutionary algorithm. All table data in YANA can be exported into text files using comma separated values; for convenience the target flux vector can be imported the same way. For further and more detailed information see the readme file that comes with the software package and which is also available via online help.
a2) Visualization and analysis of METATOOL output by YANA
To demonstrate the YANA package we use a running example (Figure 1) of the human redox metabolism (75 metabolites (46 internal, 29 external) and 58 enzymes), around the central enzyme glutathione reductase [see Additional file 1]; see also Ref [14]; extended from Ref [27]).
Under stationary conditions, this system yields a total of 134 EMs. From these, 46 include glutathione reductase, 117 involve energy consumption (ATP), whereas 128 involve redox reactions. The complete listing of these modes is given in supplementary material [see Additional file 1].
b1) YANA dissects and analyzes a metabolic web according to well connected metabolites
To illustrate the complexity reduction power of the YANA strategies we performed a series of EMAs with rising threshold values using both threshold types.
Dissection (Table 1)
Using a threshold of 7 (metabolites participating in more than seven reactions are considered external), only membrane phosphorylation is placed in a separate sub-network. A threshold of 5 splits the system into seven sub-networks with instructive specific biochemical functions and flux modes: (i) a well connected sub-network includes salvage pathway, pentose phosphate cycle, NOS, SOD and redox protection by uric acid and GSH; other sub-networks are (ii) GSH synthesis, (iii) GSH degradation and GSH protection of protein groups, (iv) membrane phosphorylation (as with threshold 7), (vi) lower glycolysis (trioses), (vii) adenylate kinase. Threshold 3 splits these sub-networks further into a total of 18 components, e.g. the well connected sub-network (i) is now put into its single pathways as named above.
Hub metabolites (Table 2)
The complex system of 134 elementary modes is first reduced to a 87-mode system (GR modes and pathways which are more central than other redox enzyme paths remain, if the threshold is set to 5 reactions). Only 24 modes remain if the threshold is set to the best connected metabolite, the currency metabolite ATP. The very short diameter obtained shows that this analysis zooms in on hub metabolites [18] and well connected next-neighbor reactions, showing the quickly equilibrated central parts of the system which one could consider more (high threshold) or less (low threshold) as external and well buffered central pools, the most pronounced being the reactions with the central currency metabolite ATP.
b2) YANA translates EM activities into specific flux distributions
Table 4 shows the calculated flux distribution for the system under study if all EMs are considered equally active (100% activity).
In the example, GR as a central enzyme of the network has an activity of 399. Besides this, the most active enzymes are: GAPDH (598), PGM (598), LDH (598), PGK (560), PK (598) and EN (598), as a parts of glycolysis, and the enzymes G6PD (576); PGLase (576) and GL6PDH (576), as components of the oxidative part of the pentose phosphate pathway. For the obtained flux distribution, we notice a tight connection between glycolysis and the glutathione reductase metabolism. The main pathways of glycolysis and PPP supply energy and reduction equivalents for strong redox protection provided by the glutathione reductase network. In contrast, several other enzymes are downregulated, in particular, those which use uric acid as an antioxidant as well as catalase.
The program also quickly calculates and visualizes flux distributions for any other chosen EM activities as given in Table 4. Thus, one notices that selective activation of EMs related to the pentose phosphate pathway leads to similar results as above. Setting only HGPRT-containing modes at a maximum activity (and all others to 0%) gives a more selective response with several enzymes completely deactivated. Finally, when all modes containing glutathione reductase are active, the graph shows the central position of GR in the network by a peak, and underlines even more the importance of critical energy providing pathways for redox protection (Figure 2).
c) Out of (partially) known flux distributions, YANA predicts and identifies EM activities with minimal error
Using our illustration example, we give
a) the results for the situation where only the enzyme fluxes for glycolytic enzymes are all set to 100 (equal activity, for convenience assumed to represent international enzyme units [micromol/minute]) and all others are known to be at zero.
b) The same as before, but all other fluxes are unknown or simply have not been measured (the enzyme activity is then set to -1 in order to indicate lack of knowledge).
Situation a) reveals a flux distribution in which, after upregulation of glycolytic enzymes, the three enzymes forming the oxidative part of PPP are also highly active. In addition, glutathione reductase (GR), NO synthase (NOS) and TrxRI (thioredoxin reductase) are upregulated as well, showing that a major part of the metabolite flux uses the path from glycolysis via oxidative PPP to redox protection enzymes. Not connected to glycolysis at all, and thus set to zero activity, are again the use of uric acid as an antioxidant and catalase.
Situation b) – a scenario where the measured data are similar but more incomplete – gives similar results, underlining that glycolysis or its side-products are important for many reactions in this network. For this case, uric acid as anti-oxidant and catalase are predicted not to be used. Detailed results for both situations are given in supplementary material. [see Additional file 1].
For comparison, experimental data on the activity of glutathione reductase and the connected redox network have been reported by Krauth-Siegel et al. (1996) [28] and others. The concentration of glutathione reductase is approx. 0.2 μM in human red blood cells and in the cytosol of various eukaryotic cells [28,29]. In erythrocytes, this corresponds to a maximal enzyme activity of 2 U/ml at 25°C. Assuming that the concentration of the substrate glutathione disulfide is 1 to 10 μM under physiological conditions, the turnover of substrate can be estimated to be 30 μM/min to 270 μM/min (30 mU/ml to 270 mU/ml).
Transcriptome analyses have been reported for antioxidant proteins of the malaria parasite Plasmodium falciparum in its various developmental stages [30]. The other side of the coin, the proteomics of oxidatively modified proteins has been reviewed by Ghezzi and Bonetto (2003) [31].
The still sparse and incomplete data support the scenarios discussed here, in particular regarding the high activity of glutathione reductase modes as well as the importance of energy providing reactions. However, a detailed kinetic and experimental metabolic flux analysis of the whole system has not yet been achieved.
The convergence criterion for the EA was to achieve a sum-of-squares error of less then 5% of the best evolved flux distribution to the target flux distribution. Regarding measurement or experimental errors and constraints, the user is alerted in case measurements are incompatible with the calculated theoretical flux distribution but also about which data are responsible for maximizing the difference between observed and calculated flux distribution.
In Table 3 EA convergence is tested using randomly generated flux distributions as test datasets, working on our example system with 134, 48 and 24 modes.
Using the example network above, with a threshold of 8, more than 50% convergence could be reached after 100 iterations (22 seconds).
Discussion
After its conceptual description [2], Elementary Mode Analysis has continuously been improved including new algorithms [12,19,32], visualization (php-Metatool [33]) and dissection algorithms [13,14]. Computation of elementary modes and visualization of these is also feasible by the program FluxAnalyzer [17]. Furthermore, alternative approaches also allow enumerating feasible routes in complex metabolic networks, for instance extreme pathway analysis [4] and hierarchical decomposition [34]. All these further implementations and algorithmic developments have specific advantages, but also limitations.
The current software package allows user-friendly post-processing of the METATOOL output. In particular visualization of the modes, editing metabolites and reactions, and graphical comparisons of enzymes and their involvement in reactions of the metabolic network are available for the user. YANA is a stand-alone visualization tool with its focus on user intervention, the quick comparison of results and thorough data exchange capabilities. In contrast, there are a number of more complex and integrated packages available such as GEPASI [15,35] which have less visualization options and offer other calculation possibilities.
For addressing the major problem of combinatorial explosion of the number of EMs in complex networks, YANA implements a decomposition method proposed earlier [13]. In this method, all highly connected metabolites are set to external status. Moreover, a new simplification strategy is offered to reduce complex metabolic networks. Earlier studies on metabolite databases show that the well connected "hub" metabolites dominate the overall architecture of a metabolic web and represent its core [18]. Here we offer the option to consider only those reactions where well connected metabolites are involved – the threshold can be chosen by the user. In fact, the results here show that such a procedure reduces a metabolic web considerably. This is particularly useful to dissect and put apart those larger parts of the metabolic web which are not well connected, so that they do not add to the central part of this metabolic map.
Metabolic fluxes are difficult to measure. YANA offers a specific approach to correlate metabolic fluxes with EM activities. Alternative algorithms for such an effort have been proposed [20,22]. The YANA routine offers several advantages. Firstly, most experimental data on protein or gene expression are always prone to errors and noisy. To account for this, in YANA no exact EM activity solution for the corresponding flux distribution is sought. Instead, the experimental input is critically examined in regard to whether it is realistic and can be satisfied by any combination of EMs. Next, the error between the observed values of enzyme fluxes and the theoretical calculated flux distribution is minimized. Accordingly, YANA also accepts rather incomplete measurements, for instance, when only two enzyme flux values are known. Furthermore, the evolutionary strategy allows incorporating any further user-desired multiple constraints into the fitness function.
The calculated EM activity pattern should additionally satisfy metabolite restrictions, as well as growth or genetic considerations on the enzyme or metabolite profile. Further constraints, which might be considered, are, for example, expression constraints dependent on promoter structure, RNA stability or protein stability. In spite of this flexibility, the evolutionary strategy converges swiftly to a solution. The great advantage of this is that we have both robust optimization and already take into account that there is noise, and that no perfect solution is possible. If desired, more criteria could be added with ease to the EA.
Conclusion
YANA adds a compact, user-friendly software package to the analysis of metabolic webs, offering several new implementations for typical challenges in such analyses including modeling of expression data. The results illustrate the application for a central redox network around glutathione reductase. Further developments will consider additional regulatory constraints profiting from the evolutionary strategy applied as well as a graphical editor for the metabolic networks including dedicated algorithms for the automatic layout of the graphs.
Availability and Requirements
All software and documentation are available for download at .
The package requires at least Java Runtime Environment (JRE) Version 1.5.0 and the following libraries, which are included in the download bundle and can be found in the /lib subdirectory:
• GenJava-CSV (© 2003, Henri Yandell)
• Jakarta Common Collections 3.1 (© 2004, The Apache Software Foundation)
• JFreeChart 0.9.21 (© 2004, Object Refinery Limited and Contributors)
• JigCell Modelbuilder (© 2004, Virginia Polytechnic Institute and State University)
• JMat 5.0 (© 2004, Yann Richet)
• Mosfet Liquid L&F (© 2004, Miroslav Lazarevic)
• Noia KDE 1.00 (© Carles Carbonell Bernado)
All libraries are licensed under either GNU General Public License (GPL) [36], Lesser GNU General Public License (LGPL) [37], BSD OpenSource License [38], DARPA BioComp OpenSource License, or other proprietary open source licenses that allow the use, redistribution, and modification of the application or parts of it. The copyright stays with the corresponding authors.
A 1.4 GHz CPU and 256 MB RAM are recommended for running the YANA software package. Installation requires at least 30 MB of hard disk space. YANA is supposed to run on any 32-bit Windows or Linux platform.
List of abbreviations
• EMA – Elementary Mode Analysis
• EM – Elementary Mode, also known as Elementary Flux Mode or Flux Mode
• EA – Evolutionary Algorithm
Authors' contributions
All authors read and approved the ms and made critical comments, adding to the final version presented here. In addition they contributed
RS: Architecture and implementation, graphical design, design of user interface.
PM: Tested and wrote an early implementation of the software and simplification routine.
AVK: Tested YANA, Metatool expertise, compatibility with Metatool.
BE: Provided theoretical insights and chemistry knowledge.
RHS: Provided experimental insights and discussion points.
SS: Biophysical knowledge, expertise in flux balance analysis, metabolic modelling and interpretation of obtained modes as well as for algorithm strategy.
TD: Concept; plan for the software and strategy, lead and guided the study.
Supplementary Material
Additional File 1
Metabolic network around GR reductase and flux distribution examples (Microsoft Excel 2003): The file contains the complete metabolic network used for elementary mode analysis including the metabolites, reactions / enzymes and elementary modes. Additionally, two flux distributions for upregulated glycolysis are given in the file, as discussed in the main section of the article.
Click here for file
Acknowledgements
We thank K. Langner for stylistic corrections and proof-reading (native speaker) and DFG for support (SFB 544/B2; BO-1099/5-2; Da 208/7-1).
Figures and Tables
Figure 1 Screenshot of the GR (glutathione reductase) system in YANA. The YANA main screen showing the GR redox network involving 75 metabolites (left side view) and 58 enzymes (right side view), resulting in 134 flux modes (not shown here).
Figure 2 Screenshot of the simulated enzyme activities diagram. Diagram of simulated spot intensities on a gel, after activation of GR containing elementary modes. Obviously glutathione reductase is indeed most active whereas other enzymes not involved in the core GR part of the system are downregulated.
Table 1 Simplification of the GR system by dissection at highly connected metabolites (cutting)
Threshold No. of elem. modes GR involved ATP involved Redox reactions Diameter
>11 134 46 (34%) 117 (87%) 128 (95%) 22.35
7 215 68 (31%) 131 (60%) 199 (92%) 22.26
5 35 4 (11%) 18 (52%) 16 (47%) 6.17
3 10 0 (0%) 5 (50%) 2 (20%) 3.0
Table 2 Simplification of the GR system by concentration on highly connected metabolites (centralization)
Threshold No. of elem. modes GR involved ATP involved Redox reactions Diameter
0 134 46 (34%) 117 (87%) 128 (95%) 22.35
5 87 22 (25%) 45 (52%) 32 (37%) 2.75
10 24 0 (0%) 24 (100%) 0 (0%) 2.38
Table 3 EA performance for three levels of complexity
No. of elementary modes Average time to convergence
134 1147.3 sec
48 81.7 sec
24 13.2 sec
Table 4 Individual enzyme activities summed over all elementary modes Calculation of individual enzyme activities according to a given flux distribution: The 134 modes obtained from the input system [see Additional file 1] are all assumed to be active with standard (1 flux unit) activity. Alternatively, fractions of full activity of individual modes (given in percentages) can be set by the user and included in the calculation. For the standard flux vector, the total enzyme activities are calculated by YANA as follows (arbitrary units, only relative fluxes are calculated):
Name Act. (a.u.) Name Act. (a.u.) Name Act. (a.u.) Name Act. (a.u.)
ALD 203.0 ApK 111.0 DPGM 38.0 EN 598.0
GAPDH 598.0 GpoI 209.0 Gr 399.0 HYPXLeak 74.0
LACex 598.0 PGI 203.0 PGK 560.0 PGLase 576.0
PGM 598.0 Pmr 201.0 PNPase 111.0 PRM 111.0
PRPPsyn 111.0 R5PI 192.0 TA 192.0 TKI 192.0
TKII 192.0 TPI 203.0 TrxRI 589.0 Xu5PE 384.0
ADA 37.0 AdPRT 74.0 AK 38.0 AMPase 75.0
AMPDA 37.0 Cat 6.0 Cca 37.0 CgdI 37.0
CgdII 75.0 CytI 38.0 DPGase 38.0 G6PD 576.0
Gcl 112.0 GL6PDH 576.0 GLCim 395.0 Gls 112.0
GtfI 37.0 GtfII 37.0 GtfIII 38.0 Gtr 37.0
Har 7.0 HGPRT 37.0 HK 395.0 IMPase 74.0
LDH 598.0 MemPhos 38.0 Nos 196.0 Opr 38.0
Pdo 99.0 PFK 203.0 PK 598.0 Sod 196.0
Tdi 196.0 Xen 196.0
==== Refs
Ziebuhr W Xiao K Coulibaly B Schwarz R Dandekar T Pharmacogenomic strategies against resistance development in microbial infections Pharmacogenomics 2004 5 361 379 15165173 10.1517/14622416.5.4.361
Schuster S Hilgetag C On elementary flux modes in biochemical systems at steady state Journal of Biological Systems 1994 2 165 182 10.1142/S0218339094000131
Rockafellar RT Convex analysis 1970 Princeton, Princeton University Press
Papin JA Stelling J Price ND Klamt S Schuster S Palsson BO Comparison of network-based pathway analysis methods Trends Biotechnol 2004 22 400 405 15283984 10.1016/j.tibtech.2004.06.010
Schilling CH Letscher D Palsson BO Theory for the systemic definition of metabolic pathways and their use in interpreting metabolic function from a pathway-oriented perspective J Theor Biol 2000 203 229 248 10716907 10.1006/jtbi.2000.1073
Poolman MG Fell DA Raines CA Elementary modes analysis of photosynthate metabolism in the chloroplast stroma Eur J Biochem 2003 270 430 439 12542693 10.1046/j.1432-1033.2003.03390.x
Carlson R Srienc F Fundamental Escherichia coli biochemical pathways for biomass and energy production: creation of overall flux states Biotechnol Bioeng 2004 86 149 162 15052634 10.1002/bit.20044
Carlson R Srienc F Fundamental Escherichia coli biochemical pathways for biomass and energy production: identification of reactions Biotechnol Bioeng 2004 85 1 19 14705007 10.1002/bit.10812
Schuster S Fell DA Dandekar T A general definition of metabolic pathways useful for systematic organization and analysis of complex metabolic networks Nat Biotechnol 2000 18 326 332 10700151 10.1038/73786
Klamt S Gilles ED Minimal cut sets in biochemical reaction networks Bioinformatics 2004 20 226 234 14734314 10.1093/bioinformatics/btg395
Dandekar T Sauerborn R Comparative genome analysis and pathway reconstruction Pharmacogenomics 2002 3 245 256 11972445 10.1517/14622416.3.2.245
Pfeiffer T Sanchez-Valdenebro I Nuno JC Montero F Schuster S METATOOL: for studying metabolic networks Bioinformatics 1999 15 251 257 10222413 10.1093/bioinformatics/15.3.251
Schuster S Pfeiffer T Moldenhauer F Koch I Dandekar T Exploring the pathway structure of metabolism: decomposition into subnetworks and application to Mycoplasma pneumoniae Bioinformatics 2002 18 351 361 11847093 10.1093/bioinformatics/18.2.351
Dandekar T Moldenhauer F Bulik S Bertram H Schuster S A method for classifying metabolites in topological pathway analyses based on minimization of pathway number Biosystems 2003 70 255 270 12941488 10.1016/S0303-2647(03)00067-4
Mendes P Biochemistry by numbers: simulation of biochemical pathways with Gepasi 3 Trends Biochem Sci 1997 22 361 363 9301339 10.1016/S0968-0004(97)01103-1
phpMetatool
Klamt S Stelling J Ginkel M Gilles ED FluxAnalyzer: exploring structure, pathways, and flux distributions in metabolic networks on interactive flux maps Bioinformatics 2003 19 261 269 12538248 10.1093/bioinformatics/19.2.261
Schmidt S Sunyaev S Bork P Dandekar T Metabolites: a helping hand for pathway evolution? Trends Biochem Sci 2003 28 336 341 12826406 10.1016/S0968-0004(03)00114-2
Gagneur J Klamt S Computation of elementary modes: a unifying framework and the new binary approach BMC Bioinformatics 2004 5 175 15527509 10.1186/1471-2105-5-175
Poolman MG Venakatesh KV Pidcock MK Fell DA A method for the determination of flux in elementary modes, and its application to Lactobacillus rhamnosus Biotechnol Bioeng 2004
Ghaemmaghami S Huh WK Bower K Howson RW Belle A Dephoure N O'Shea EK Weissman JS Global analysis of protein expression in yeast Nature 2003 425 737 741 14562106 10.1038/nature02046
Stelling J Klamt S Bettenbrock K Schuster S Gilles ED Metabolic network structure determines key aspects of functionality and regulation Nature 2002 420 190 193 12432396 10.1038/nature01166
Wagner A Fell DA The small world inside large metabolic networks Proc R Soc Lond B Biol Sci 2001 268 1803 1810 10.1098/rspb.2001.1711
Yen G Lu H Hierarchical genetic algorithm for near optimal feedforward neural network design Int J Neural Syst 2002 12 31 43 11852443
Hucka M Finney A Sauro HM Bolouri H Doyle JC Kitano H Arkin AP Bornstein BJ Bray D Cornish-Bowden A Cuellar AA Dronov S Gilles ED Ginkel M Gor V Goryanin II Hedley WJ Hodgman TC Hofmeyr JH Hunter PJ Juty NS Kasberger JL Kremling A Kummer U Le Novere N Loew LM Lucio D Mendes P Minch E Mjolsness ED Nakayama Y Nelson MR Nielsen PF Sakurada T Schaff JC Shapiro BE Shimizu TS Spence HD Stelling J Takahashi K Tomita M Wagner J Wang J The systems biology markup language (SBML): a medium for representation and exchange of biochemical network models Bioinformatics 2003 19 524 531 12611808 10.1093/bioinformatics/btg015
Finney A Hucka M Systems biology markup language: Level 2 and beyond Biochem Soc Trans 2003 31 1472 1473 14641091
Schuster S Fell DA Pfeiffer T Dandekar T P. B Elementary mode analysis illustrated with human red cell metabolism BioThermoKinetics in the Post Genomic Era (C Larsso, I-L Pahlman, L Gustafsson, eds) 1998 Chalmers, Göteborg 332 339
Krauth-Siegel RL Muller JG Lottspeich F Schirmer RH Glutathione reductase and glutamate dehydrogenase of Plasmodium falciparum, the causative agent of tropical malaria Eur J Biochem 1996 235 345 350 8631352 10.1111/j.1432-1033.1996.00345.x
Schirmer RH Krauth-Siegel RL Schulz GE Glutathione reductase Coenzymes and Cofactors 1989 IIIA 553 596
Bozdech Z Ginsburg H Antioxidant defense in Plasmodium falciparum--data mining of the transcriptome Malar J 2004 3 23 15245577 10.1186/1475-2875-3-23
Ghezzi P Bonetto V Redox proteomics: identification of oxidatively modified proteins Proteomics 2003 3 1145 1153 12872215 10.1002/pmic.200300435
Wagner C Nullspace approach to determine elementary modes of chemical reaction systems J Phys Chem 2004 B 108 2425 2431
Hofestädt R Lautenbach K Lange M Modellierung und Simulation Metabolischer Netzwerke DFG-Workshop Preprint 2000 10
Gagneur J Jackson DB Casari G Hierarchical analysis of dependency in metabolic networks Bioinformatics 2003 19 1027 1034 12761067 10.1093/bioinformatics/btg115
Martins AM Mendes P Cordeiro C Freire AP In situ kinetic analysis of glyoxalase I and glyoxalase II in Saccharomyces cerevisiae Eur J Biochem 2001 268 3930 3936 11453985 10.1046/j.1432-1327.2001.02304.x
GNU General Public License
Lesser GNU General Public License
BSD OpenSource License
| 15929789 | PMC1175843 | CC BY | 2021-01-04 16:02:50 | no | BMC Bioinformatics. 2005 Jun 1; 6:135 | utf-8 | BMC Bioinformatics | 2,005 | 10.1186/1471-2105-6-135 | oa_comm |
==== Front
BMC Cardiovasc DisordBMC Cardiovascular Disorders1471-2261BioMed Central London 1471-2261-5-171598242110.1186/1471-2261-5-17Study ProtocolA cluster randomized trial to assess the impact of opinion leader endorsed evidence summaries on improving quality of prescribing for patients with chronic cardiovascular disease: rationale and design [ISRCTN26365328] Majumdar Sumit R [email protected] Finlay A [email protected] Ross T [email protected] Department of Medicine, University of Alberta, Edmonton AB and the Institute of Health Economics, Edmonton AB, Canada2005 27 6 2005 5 17 17 6 6 2005 27 6 2005 Copyright © 2005 Majumdar et al; licensee BioMed Central Ltd.2005Majumdar et al; licensee BioMed Central Ltd.This is an Open Access article distributed under the terms of the Creative Commons Attribution License (), which permits unrestricted use, distribution, and reproduction in any medium, provided the original work is properly cited.
Background
Although much has been written about the influence of local opinion leaders on clinical practice, there have been few controlled studies of their effect, and almost none have attempted to change prescribing in the community for chronic conditions such as heart failure (HF) or ischemic heart disease (IHD). These two conditions are common and there is very good evidence about how to best prevent morbidity and mortality – and good evidence that quality of care is, in general, suboptimal. Practice audits have demonstrated that about one-half of eligible HF patients are prescribed ACE inhibitors (with fewer still reaching appropriate target doses) and less than one-third of patients with established IHD are prescribed statins (with many fewer reaching recommended cholesterol targets). It is apparent that interventions to improve quality of prescribing are urgently needed. We hypothesized that an intervention that consisted of patient-specific one-page evidence summaries, generated and then endorsed by local opinion leaders, would be able to change prescribing practices of community-based primary care physicians.
Methods (study design)
A pragmatic single-centre cluster randomized controlled trial comparing an opinion leader-based intervention to usual care for patients with HF or IHD. Randomization will be clustered at the level of the primary care physician; as the design effect is anticipated to be negligible, the unit of analysis will be the patient. Patients with HF or IHD (not receiving ACE inhibitors or statins, respectively) will be recruited from community pharmacies and allocated to intervention or usual care based on the randomization status of their primary care physician. The primary outcome is improvement in prescription of proven efficacious therapies for HF (ACE inhibitors) or IHD (statins) within 6 months of the intervention.
Conclusion
If the methods used in this intervention are found to improve prescribing practices, similar interventions could be designed for other chronic conditions dealt with in the outpatient setting.
==== Body
Background
Opinion leaders – an untapped resource for quality improvement
Local opinion leaders are able to influence the practice of other physicians because they are well-known, respected, and trusted to evaluate medical innovations within the local context [1-4]. Because they influence patterns of practice in the community, and may accelerate the uptake of knowledge, their participation in any program of quality improvement is essential. Certainly, surveys of physicians [5,6] have consistently confirmed the importance of colleagues and local consultants on individual patterns of practice. Yet, the use of local opinion leaders to influence physician practice has only been tested in six randomized controlled trials [1,2]. While two of these trials [3,4] demonstrated an important impact upon practice, both assessed labour-intensive, expensive, hospital-based educational interventions spearheaded by a small number of opinion leaders (four in one study, 16 in the other) for conditions treated in a hospital setting (delivery by cesarean section, treatment of acute myocardial infarction). Although the use of opinion leaders to influence the outpatient management of common conditions holds great promise, this is yet a hypothesis to be rigorously tested [7].
Underuse of proven medications for cardiac diseases is common
In fact, most patients with chronic conditions (such as heart failure [HF] or atherosclerotic ischemic heart disease [IHD]) are treated as outpatients. Previous practice audits have documented significant "care gaps" between the available evidence and actual practice (e.g., only half of eligible patients with HF receive an ACE inhibitor [7,8], despite overwhelming evidence of benefit [7,8], and those who are prescribed an ACE inhibitor are usually given doses below those tested in clinical trials [10]; less than a third of patients with established IHD receive lipid-lowering therapy [7,10,11], despite multiple trials [7,12,14] showing benefit, and of those prescribed lipid-lowering therapies, only 15% achieve recommended cholesterol goals [11]). As these two conditions represent an important burden of illness for the community [15], closure of these care gaps should be a public health priority. Moreover, we believe methods need to be developed that can reliably and efficiently improve the quality of care for these high-risk populations [7].
Community-based interventions must be simple and practical
When testing the effect of opinion leader influence in the outpatient setting, we must keep in mind the potential generalizability and applicability of any proposed intervention. Thus, the focus must be on a practical means of incorporating opinion leaders into practice, because we know that opinion leaders may exert considerable influence on community practice. In addition, previous studies have established that the methods of information transfer most favoured by physicians are one-page summaries of guidelines or evidence [5,17]. Moreover, other work has suggested that when the transfer of clinically relevant information is directly linked to a specific patient encounter with a specific recommendation (e.g., real-time reminders, whether manual or computerized) there is a much greater likelihood of affecting change [2,7,18]. The ideal way to create such a scenario would be to have a "real-time" reminder generated in the physicians' office and linked to the actual clinical encounter. This is not currently practical or feasible within the vast majority of community practices in Canada – most of these practices are still paper-chart based, and even of those practices that have electronic medical records and some capacity for computerized decision support, most are not linked to other databases such as vital statistics, hospital discharge databases, or pharmacy dispensing records.
Potential role of community pharmacies and pharmacists
Prior research, and our own ongoing work in other clinical arenas, leads us to believe that the community pharmacy may be a reasonable location to generate the requisite linkages needed to undertake an intervention to improve quality of outpatient prescribing as well as a convenient site for unobtrusively recruiting patients [19]. This is because patients with chronic conditions can be identified by "marker medications" at the time of presentation to their pharmacy (more than 90% of HF patients are prescribed loop diuretics [8], and almost 100% of patients with known IHD are given prescriptions for short-acting nitrates [20]), and because pharmacists already interact on a regular basis with both prescribing physicians and their patients. Therefore, the community pharmacy may serve as a potential site for patient recruitment and education or intervention. For example, the SCRIP study used 54 community pharmacies in Western Canada, and demonstrated improvements in lipid management in a 675 patient randomized trial [19]. This pharmacy network has evolved and expanded to bring together community pharmacists, physicians and other healthcare professionals into broad-based community-level studies through The Epidemiology Coordinating and Research (EPICORE) Centre and its subsidiary, the Centre for Community Pharmacy Research and Interdisciplinary Strategies (COMPRIS) at The University of Alberta.
Evidence-based interventions to improve quality of prescribing
With these basic considerations in mind, we designed an evidence-based multi-faceted intervention to improve the quality of prescribing in the community. We propose having locally-nominated opinion leaders generate and endorse one-page evidence summaries for two common and chronic cardiovascular conditions. These evidence summaries, linked with specific patient-level medication profiles (generated at the community pharmacy), will be distributed to practicing physicians and attached to their patients' chart. Our hypothesis is that this will act as both a source of credible and convincing information and a specific reminder for action at the next patient encounter. Our study is designed to test this hypothesis, by assessing the impact of this intervention on the quality of prescribing for patients with heart failure or ischemic heart disease.
Methods
Purpose
The purpose of this study is to determine whether "evidence summaries," generated and endorsed by local opinion leaders, can improve the prescribing of proven efficacious therapies by primary care physicians for their patients with chronic cardiovascular diseases.
Objectives
This study has three main objectives:
1. To determine whether evidence summaries generated and endorsed by local opinion leaders can increase the prescription of efficacious therapies for patients with HF or IHD.
2. To compare the effectiveness of the proposed intervention with two different conditions, to begin to understand the potential for generalizability across different diseases and conditions.
3. To evaluate whether distribution of local opinion leader endorsed evidence summaries, through community pharmacies, is a viable quality improvement tool within the constraints of current healthcare delivery systems.
Study design
A single centre pragmatic cluster randomized controlled trial in which primary care physicians and their patients will be randomly allocated to intervention (relevant single page evidence summary endorsed by local opinion leaders and a copy of the patient's current medication profile faxed back to the primary care physician) or control (only the patient's current medication profile faxed back to the primary care physician). To prevent the potential for "contamination" within an individual physicians' practice (e.g., allocation of an intervention and a control patient to the same physician) a modified form of cluster allocation will be undertaken. Specifically, if a physician is randomized to the intervention arm for IHD, her other IHD patients that present to a participating pharmacy will also be subject to the IHD intervention, to a maximum of five patients. In addition, she will also act as her own control, because she will be assigned to the control arm for her HF patients, again to a maximum of five patients. For a schematic of the overall study design, please see Additional File: 1.
Identifying opinion leaders and generating evidence summaries
All primary care physicians within Capital Health (the greater Edmonton metropolitan area, Alberta, Canada) were mailed a one-page survey (based on a previously validated opinion leader nomination instrument [3,4,21]) asking them to nominate local colleagues who best match standard descriptions of "educationally influential" opinion leaders (see Additional File: 2). The overall response rate was 30% (225 of 788 surveys returned), typical of the response rates for physician surveys. Five opinion leaders were overwhelmingly identified by their peers (1 physician for HF only, 1 physician for IHD only, and 3 physicians nominated for both HF and IHD; 3 opinion leaders were cardiologists and 2 were general internists) and were approached to participate in the development of the one-page evidence summaries for HF and for IHD. All 5 nominated opinion leaders participated in the development of the intervention materials. The survey instrument also informed primary care physicians about the study (in general terms) and provided them the option to "opt-out" of having themselves or their patients recruited or enrolled in the study – only 19 physician respondents declined to participate. Physician survey non-respondents (i.e., those who did not explicitly and actively opt-out of the study) will be considered eligible for the study, although any physician will be able to withdraw themselves and their patients at any time.
The data for the evidence summaries will be initially generated by experts in evidence-based medicine and cardiovascular care using standardized techniques for research synthesis [22]; the opinion leaders and study investigators will then ensure that treatment recommendations are consistent with available clinical practice guidelines, and then carefully incorporate this material into the one-page evidence summaries. The final content and format of these summaries will be arrived at by standard consensus methods, and these evidence summaries will form the core of our intervention.
Description of the intervention
The intervention consists of a disease-specific and patient-specific one-page evidence summary. It will be a patient-specific letter addressed to the patients' primary care physician, along with a description of the potential risks of undertreatment and current evidence-based treatment recommendations. The letter will be signed and endorsed by all 5 of the study opinion leaders. Examples of the IHD and HF letters are provided in Additional Files: 3 and 4.
Accompanying the letter will be the most recent pharmacy record of medications dispensed to the study patient. It is intended that the evidence summary and the pharmacy medication profile will become part of the patients' medical record and act as a reminder or prompt at the next patient visit. These materials will be faxed to the primary care physician from the patients' community pharmacy.
Controls ("usual care")
Physicians of control patients will only receive the pharmacy medication dispensing record. This will be done for two reasons. First, to disentangle the effects of having a complete medication record (in and of itself, a departure from "usual care") versus the impact of the evidence summaries. Second, to control for the possibility of any attention-related or study-related "Hawthorne effects" – all patients and all physicians will receive the same number of contacts and follow-ups, and all that will differ between the experimental arms is the content of the intervention. Again, control materials will be faxed from the patients' community pharmacy. It should be noted here that "control," as we have defined it, is actually more rigorous than the current standard of care; it might also be argued simply providing a medication profile would be sufficient in an of itself to change practice, although a recent systematic review concluded that this would not be the case [2].
Study setting
Capital Health (Edmonton, Alberta) is one of the largest integrated health service delivery organizations in Canada. It provides comprehensive health services for almost one million people and has an annual budget of almost two billion dollars Canadian. Primary care is delivered by approximately 800 fee-for-service physicians. Community dwelling patients with cardiovascular disease (either HF or IHD) will be recruited from a convenience sample of 40 different community pharmacies, some of which have previously participated in pharmacist-based research studies. It should be noted that, at the time of study design, none of these pharmacies were electronically linked to either physicians' practice records or any form of patients' medical charts.
Subjects
Involvement of primary care physicians has already been described, and we potentially will be able to draw subjects from the population-base of all fee-for-service physicians in the entire region. Patients (and their prescribing physicians) will be identified by community pharmacists from lists obtained from their computerized dispensing records. Patients with a prescription for a "marker" medication (a loop diuretic for the proxy diagnosis of HF; a short acting nitrate for the proxy diagnosis of IHD) who are not currently taking the respective study medications of interest will be notified by their own pharmacist about the study and informed that (with their permission) a member of the research team will contact them with more information about the study. A research assistant (pharmacist) will then initiate telephone contact once potential eligibility has been determined, and obtain verbal consent from eligible patients to access their pharmacy records twice (study entry and 6-month followup) for the purposes of the study. It should be noted that these patient's pharmacists, providing routine services in the community, already have access to their pharmacy records. Specific criteria for inclusion and exclusion include the following:
Inclusion criteria
Patients with HF or IHD who are not currently taking the study medications of interest (ACE inhibitors/angiotensin receptor blockers for HF or statins for IHD), and whose primary care physician of record is part of the study. For patients who happen to be eligible for both HF and for IHD, only one condition will be selected at random.
Exclusion criteria
Patients who decline enrollment, or who are unable or unwilling to give informed consent, or who have previously taken the study medications according to dispensing records, or who have a documented allergy or intolerance to study medications according to pharmacist records, or who are in long-term care facilities or institutions, or who do not confirm on the basis of self-report that they have a diagnosis of either HF or IHD, or whose primary care physician has already contributed 5 patients to the study.
Allocation to experimental arms
Simple randomization will occur at the level of the primary care physician, before patient recruitment and enrollment begins, using a computer-generated sequence with allocation concealment. Each participating physician will be randomly allocated to intervention or control arm for HF; physicians allocated to intervention for HF will be automatically assigned to the control arm for IHD, and vice versa. This design prevents "contamination" within an individual physicians' practice (e.g., having one intervention HF patient and one control HF patient), while simultaneously increasing study efficiency (i.e., physicians can contribute patients to both the HF and IHD arms of the study, thus potentially reducing the overall number of physicians required). Furthermore, no one physician will be permitted to contribute more than 5 patients to the study, to minimize any issues related to the potential for physician-level clustering of study outcomes [23,24].
That said, this might be considered a form of "cluster" randomization [2,23,24]. However, on a population-wide basis we anticipate that most regional physicians will contribute no patients, and the majority of physicians will contribute no more than 1 study patient. This has been the case in previous studies we have recently undertaken using similar recruitment and allocation methods [19,25]. Furthermore, by our study design, no physician will be able to contribute more than 5 patients to the study as a whole. Therefore, the "design effect" should be negligible, particularly since previous studies have demonstrated that important cluster effects do not come into play until the physician-to-patient ratio exceeds 1-to-5, and we expect our ratio to be close to 1-to-1; thus, all sample size and analytic considerations will be based on the patient as the unit of analysis and the unit of causal inference [23-25].
Outcome measures
The primary outcome measure will be the "improvement" of prescribing for efficacious therapies in patients with a chronic cardiovascular disease within 6 months of the intervention. By study design, none of the study patients will be taking the medications of interest. For HF, starting any ACE inhibitor or angiotensin receptor blocker will be considered a positive outcome. For IHD, starting any statin will be considered a positive outcome. For the primary outcome all positive study-related medication changes will be pooled for an overall estimate of effect, compared with usual care controls.
The main secondary outcomes will be condition-specific "improvement" in prescribing after 6 months. In addition, we will attempt to assess optimization of dosage for each of the medications prescribed (i.e., ACE inhibitors or angiotensin receptor blockers and statins). We will also evaluate patient adherence (using prescription refill rates based on dispensing records) and examine the potential influence of age and sex on outcomes.
Study procedures and data collection
The primary source of data for the study will be the patient-level medication profiles generated at each community pharmacy. All data will be collected by a research assistant traveling to each pharmacy. Data will be collected using a standardized abstraction instrument. Only data without unique personal identifiers, but with a unique study ID# (and therefore anonymized), will be entered into a secure database housed at the EPICORE Centre, University of Alberta. Investigators will be blinded to physician- and patient-level unique identifiers and investigators and study patients will be masked to allocation status. Primary care physicians themselves cannot be blinded to allocation status. Follow-up data collection is scheduled at 6-months, again based on patient-level medication profiles generated by participating pharmacists, and follow-up data will be collected without knowledge of allocation status in an independent and blinded fashion. All statistical analyses will be conducted by an independent statistician also masked to allocation status.
Covariates that will be collected include age, sex, and self-reported medical diagnoses. Using medication profiles, we will also be able to generate a previously validated measure of comorbidity, the Chronic Disease Score. This measure has been previously demonstrated to provide valid and reliable risk adjustment for comorbidity or case-mix, and it is able to predict long-term morbidity, hospitalization, mortality, and health care utilization [26].
The time period chosen for outcome assessment (6-months) was selected as the vast majority of HF patients (over 80% [10]) are seen at least every 6 months and a poll of general internal medicine specialists and cardiologists at the University of Alberta confirmed that most IHD patients are also seen within this same time period. Data collation, quality assurance, and statistical analyses will be carried out at the EPICORE Centre.
Statistical analyses
Although the physician will be the unit of allocation, the patient is the unit of analysis and causal inference. This is justified by the anticipated small design effect, and the fact we expect that the outcomes for individual patients to be clinically and statistically independent of each other – because each intervention is itself both patient- and condition-specific. The main analysis will be a comparison of the proportion of all patients who successfully achieve the primary outcome ("improved" prescribing of efficacious therapies, as described above) at 6-months. This will be tested using the chi square statistic. There will be no planned interim analyses, and we will consider a p-value <0.05 to be statistically significant. Secondary analyses will compare the proportions of patients who successfully achieve the primary outcome in each of the two cardiovascular conditions (HF and IHD), to determine whether observed effects are consistent across different disease states. Two prespecified subgroups (male vs. female and age less than 70 years vs. 70 years of age and older) will also be considered. Secondary outcomes will be studied in an analogous fashion, and will be considered only exploratory in nature, and as such, comparisons will be made without corrections for multiple testing.
In order to investigate what factors are associated with changes in the primary outcome (our dependent binary variable), and to control for the possibility of potential imbalances in patient-level characteristics induced by our randomization scheme, multivariable logistic regression analyses will be used to examine those variables that are deemed to be clinically important (i.e., age, sex, diabetes status) or that differ statistically at a p-value <0.10 between experimental arms. In addition, to examine the remote possibility of "cluster-associated" study design effects, two secondary sensitivity analyses will be considered: first, the main analysis will be repeated using the physician as the unit of analysis [2,23,24]; and second, the aforementioned logistic regression models will be re-analyzed using generalized estimating equations to control for the potential lack of statistical independence among patients treated by the same study physician [2,23,24].
Sample size considerations
Using surveys of physicians and policy-makers, we determined that the "minimal" clinically important difference for this particular intervention to be considered both useful (and perhaps) fundable on an ongoing basis, was a 20% improvement over and above usual care. By design, use of any of the study medications at baseline will be zero. After 6 months, we assume no more than 10% of control patients will have started a study medication. Acknowledging that a 20% absolute increase in the primary outcome (starting a study medication, either an ACE inhibitor or angiotensin receptor blocker [HF] or a statin [IHD]) would be a clinically important effect size, setting the α error rate at 0.05 (2-sided), and the β error at 0.20 (power 80%), a total sample size of 140 will be required. Allowing for losses to follow-up, the ability to examine each of the conditions separately, and the possibility of a very small design effect associated with patient clustering, the total sample size has been adjusted upwards to 160 patients.
Ethical considerations
The protocol and procedures as described have received institutional approval from the Health Research Ethics Board of the University of Alberta. In addition, any individual physician or patient will have the opportunity to withdraw from the study at any time, and none of the investigators or analysts will have access to any physician-identifiable or patient-identifiable data. The funding for the study is from two peer-reviewed grants. The funding sources (Alberta Heritage Foundation for Medical Research and the Institute of Health Economics, Edmonton, Alberta) had no role in the design, conduct, analysis, interpretation, or reporting of the study and will not have access to the data. Finally, it should be noted that the 5 opinion leaders received no compensation (financial or otherwise) whatsoever for their participation in the study or endorsement of the evidence summaries.
Discussion
Herein, we have reported the background, rationale, and study protocol for a pragmatic cluster randomized controlled trial of an intervention that will test the hypothesis that opinion leader-generated and endorsed one page evidence summaries that are patient-specific will be able to improve the quality of cardiovascular medication prescribing. The opinion leader literature is sparse at best, and yet the potential role and educational influence of these important physician champions is vastly underappreciated. To our knowledge, this will be the only rigorous study of the "opinion leader hypothesis" undertaken outside the acute care or hospital setting. In addition, we will be able to gather valuable data about the potential role for community pharmacies and pharmacists in improving the quality of care for patients with chronic diseases that (by definition) require careful and longterm polypharmacy.
Perhaps a more ideal study could be undertaken in a more hospitable setting, such as a large managed care organization or a large academic medical centre – one that has a captive population and where all health providers have access to a universal electronic medical record that is directly linked to pharmacy dispensing databases and where there is potential for computerized decision support with the capacity to generate real time reminders and prompts. Indeed, this has been tested in a randomized trial for outpatients with chronic HF or chronic IHD (an almost identical population in most respects to the one we are enrolling), and it has been found wanting [27]. It could be that information technology alone will be unable to improve the quality of care if the treatment recommendations are not first endorsed by local opinion leaders. This possibility, as well as the fact that most community-based practices are years to decades away from having clinically useful and integrated electronic health information systems, will mean that the results of our study will likely have broader implications than for just the two conditions we chose to study.
Conclusion
If we are able to demonstrate that opinion leaders can change clinical practice in the community, then larger multi-centre studies of different conditions (e.g., controller medications in asthma, antiresorptive therapies in osteoporotic patients with fractures, antihypertensive agents in people with diabetes) will need to be undertaken.
Abbreviations
IHD, ischemic heart disease; HF, heart failure; ACE inhibitor, angiotensin-converting enzyme inhibitor
Competing interests
The author(s) declare that they have no competing interests.
Authors' contributions
All of the authors contributed equally to all aspects of the manuscript and fulfill all standard criteria for authorship. Sumit Majumdar also drafted the first version of the manuscript and will act as guarantor.
Pre-publication history
The pre-publication history for this paper can be accessed here:
Supplementary Material
Additional File 1
Overview of Study Design
Click here for file
Additional File 2
Opinion Leader Survey Instrument
Click here for file
Additional File 3
Example of the Evidence Summary for Heart Failure
Click here for file
Additional File 4
Example of the Evidence Summary for IHD
Click here for file
Acknowledgements
Supported by grants from the Alberta Heritage Foundation for Medical Research (AHFMR) and the Institute of Health Economics. Sumit Majumdar and Finlay McAlister receive salary support from AHFMR and the Canadian Institutes of Health Research. Finlay McAlister and Ross Tsuyuki are supported by the Merck Frosst/Aventis Chair in Patient Health Management at the University of Alberta.
==== Refs
Thomson MA Oxman AD Haynes RB Davis DA Freemantle N Harvey EL Local opinion leaders to improve health professional practice and health care outcomes (Cochrane Review) The Cochrane Library 1998 Oxford: Update Software
Majumdar SR Lipton HL Soumerai SB Strom B Evaluating and improving physician prescribing Pharmacoepidemiology 2005 4 Toronto: John Wiley and Sons 419 438
Soumerai SB McLaughlin TJ Gurwitz JH Gaudagnoli E Hauptman PJ Borbas C Morris N McLaughlin B Gao X Willison DJ Asinger R Gobel F Effect of local medical opinion leaders on quality of care for acute myocardial infarction. A randomized controlled trial JAMA 1998 279 1358 63 9582043 10.1001/jama.279.17.1358
Lomas J Enkin M Anderson GM Hannah WJ Vayda E Singer J Opinion leaders vs audit and feedback to implement practice guidelines. Delivery after cesarean section JAMA 1991 265 2202 7 2013952 10.1001/jama.265.17.2202
Hayward RSA Guyatt GH Moore KA McKibbon KA Carter AO Canadian physicians' attitudes about and preferences regarding clinical practice guidelines CMAJ 1997 156 1715 23 9220923
McAlister FA Graham I Karr G Laupacis A Evidence-based medicine and the practising physician: a survey of Canadian general internists J Gen Intern Med 1999 14 236 42 10203636 10.1046/j.1525-1497.1999.00323.x
Majumdar SR McAlister FA Furberg CD From knowledge to practice in chronic cardiovascular disease: a long and winding road J Am Coll Cardiol 2004 43 1738 42 15145092 10.1016/j.jacc.2003.12.043
The Clinical Quality Improvement Network Investigators Mortality risk and patterns of practice in 4,606 acute care patients with congestive heart failure. The relative importance of age, sex, and medical therapy Arch Intern Med 1996 156 1669 73 8694665 10.1001/archinte.156.15.1669
Garg R Yusuf S for the Collaborative Group on ACE Inhibitor Trials Overview of randomized trials of angiotensin-converting enzyme inhibitors on mortality and morbidity in patients with heart failure JAMA 1995 273 1450 6 7654275 10.1001/jama.273.18.1450
McAlister FA Teo KK Taher M Montague TJ Humen D Cheung L Kiaii M Yim R Armstrong PW Insights into the contemporary epidemiology and outpatient management of congestive heart failure Am Heart J 1999 138 87 94 10385769
The Clinical Quality Improvement Network Investigators Low incidence of assessment and modification of risk factors in acute care patients at high risk for cardiovascular events, particularly among females and the elderly Am J Cardiol 1995 76 570 2 7677079
Majumdar SR Gurwitz JH Soumerai SB Undertreatment of hyperlipidemia in the secondary prevention of coronary artery disease J Gen Intern Med 1999 14 492 499 10.1046/j.1525-1497.1999.02229.x
Scandinavian Simvastatin Survival Study Group Randomised trial of cholesterol lowering in 4444 patients with coronary heart disease: the Scandinavian Simvastatin Survival Study (4S) Lancet 1994 344 1383 9 7968073
Sacks FM Pfeffer MA Moye LA for the CARE Investigators The effect of pravastatin on coronary events after myocardial infarction in patients with average cholesterol levels N Engl J Med 1996 335 1001 9 8801446 10.1056/NEJM199610033351401
The Long-term Intervention with Pravastatin in Ischemic Disease (LIPID) Study Group Prevention of cardiovascular events and death with pravastatin in patients with coronary heart disease and a broad range of initial cholesterol levels N Engl J Med 1998 339 1349 1357 9841303 10.1056/NEJM199811053391902
Chan B Coyte P Heick C Economic impact of cardiovascular disease in Canada Can J Cardiol 1996 12 1000 06 9191493
Hayward RSA Wilson MC Tunis SR Guyatt GH Moore KA Bass EB Practice guidelines: what are internists looking for? J Gen Intern Med 1996 11 176 8 8667096
Oxman AD Thomson MA Davis DA Haynes RB No magic bullets: a systematic review of 102 trials of interventions to improve professional practice CMAJ 1995 153 1423 31 7585368
Tsuyuki RT Johnson JA Teo KK Simpson SH Ackman ML for the Study of Cardiovascular Risk Intervention by Pharmacists (SCRIP) Investigators A randomized trial of the effect of community pharmacist intervention on cholesterol risk management Arch Intern Med 2002 162 1149 1155 12020186 10.1001/archinte.162.10.1149
Cannon PJ Connell PA Stockley IH Garner ST Hampton JR Prevalence of angina as assessed by a survey of prescriptions for nitrates Lancet 1988 1 979 981 2896837 10.1016/S0140-6736(88)91790-4
Hiss RG MacDonald R David WR Identification of physician educational influentials in small community hospitals Res Med Educ 1978 17 283 8
NHS Centre for Reviews and Dissemination Undertaking systematic reviews of research on effectiveness CRD Guidelines for those carrying out or commissioning reviews 1996 CRD Report 4, University of York
Donner A Birkett N Buck C Randomization by cluster: sample size requirements and analysis Am J Epdiemiol 1981 114 906 14
Divine GW Brown JT Frazier LM The unit of analysis error in studies about physicians' patient care behavior J Gen Intern Med 1992 7 623 629 1453246
Majumdar SR Rowe BH Folk D Johnson JA Holroyd BH Morrish DW Maksymowych WP Steiner IP Harley CH Wirzba B Hanley DA Blitz S Russell AS A controlled trial to increase detection and treatment of osteoporosis in older patients with a wrist fracture Ann Intern Med 2004 141 366 373 15353428
Clark DO Von Korff M Saunders K Baluch WM Simon GE A chronic disease score with empirically derived weights Med Care 1995 33 783 795 7637401
Tierney WM Overhage JM Murray MD Harris LE Zhou XH Eckert GJ Smith FE Nienaber N McDonald CJ Wolinsky FD Effects of computerized guidelines for managing heart disease in primary care: a randomized controlled trial J Gen Intern Med 2003 18 967 976 14687254 10.1111/j.1525-1497.2003.30635.x
| 15982421 | PMC1175844 | CC BY | 2021-01-04 16:30:07 | no | BMC Cardiovasc Disord. 2005 Jun 27; 5:17 | utf-8 | BMC Cardiovasc Disord | 2,005 | 10.1186/1471-2261-5-17 | oa_comm |
==== Front
BMC Dev BiolBMC Developmental Biology1471-213XBioMed Central London 1471-213X-5-91591891010.1186/1471-213X-5-9Research Article Creatine synthesis and transport during rat embryogenesis: Spatiotemporal expression of AGAT, GAMT and CT1 Braissant Olivier [email protected] Hugues [email protected] Anne-Marie [email protected] Oliver [email protected] Theo [email protected] Claude [email protected] Clinical Chemistry Laboratory, University Hospital, CH-1011 Lausanne, Switzerland2 Institute of Cell Biology, Swiss Federal Institute of Technology, CH-8093 Zürich, Switzerland3 Institute of Molecular Biology, University of Zurich, CH-8057 Zürich, Switzerland2005 26 5 2005 5 9 9 16 3 2005 26 5 2005 Copyright © 2005 Braissant et al; licensee BioMed Central Ltd.2005Braissant 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
Creatine (Cr) is synthesized by a two-step mechanism involving arginine:glycine amidinotransferase (AGAT) and guanidinoacetate methyltransferase (GAMT), and is taken up by cells through a specific Cr transporter, CT1. Recently, genetic defects of this pathway have been described, that lead to Cr deficiency, neurological symptoms in early infancy and severe neurodevelopmental delay. To investigate the involvement of Cr synthesis and uptake pathways during embryonic development, we determined the spatiotemporal expression of AGAT, GAMT and CT1 during the rat embryogenesis, at the mRNA and protein level.
Results
We show that AGAT and GAMT are expressed in hepatic primordium as soon as 12.5 days, then progressively acquire their adult pattern of expression, with high levels of AGAT in kidney and pancreas, and high levels of GAMT in liver and pancreas. AGAT and CT1 are prominent in CNS, skeletal muscles and intestine, where they appear earlier than GAMT. High levels of CT1 are found in epithelia.
Conclusion
Our results suggest that de novo synthesis of Cr by AGAT and GAMT, as well as cellular Cr uptake by CT1, are essential during embryonic development. This work provides new clues on how creatine can be provided to developing tissues, and suggests that Cr deficiencies might induce irreversible damages already in utero, particularly on the nervous system.
==== Body
Background
Central nervous system (CNS) is the main organ affected in patients suffering from creatine (Cr) deficiency syndromes due either to AGAT, GAMT or CT1 deficiency [1-3]. As recently described, these patients present neurological symptoms in early infancy and show severe neurodevelopmental delay [4-6]. All three deficiencies are characterized by an absence, or a severe decrease, of Cr in CNS [7,8].
The Cr / phosphocreatine (P-Cr) / creatine kinase (CK) system is essential for the buffering and transport of high energy phosphates [9]. Cr is taken up by food, or synthesized endogenously by a two-step mechanism involving L-arginine:glycine amidinotransferase (AGAT) and S-adenosyl-L-methionine:N-guanidinoacetate methyltransferase (GAMT). Cr is taken up by cells through CT1, a specific Cr transporter belonging to the Na+-dependent neurotransmitter transporter family. In adult mammals, AGAT is predominantly expressed in kidney and pancreas, and GAMT is mainly localized in liver and pancreas. In addition, both enzymes are also expressed in various other tissues, albeit at lower levels. The highest expression of CT1 is found in kidney, heart and skeletal muscle (see [10] and references therein). Cr synthesis has been observed in CNS [11]. AGAT and GAMT mRNAs have been revealed in neurons, astrocytes and oligodendrocytes [12,13]. By contrast, CT1 has been found in neurons, oligodendrocytes and microcapillary endothelial cells, but is not detectable in astrocytes [13-18]. Cr plays an essential role in CNS, where it is involved in Na+-K+-ATPase activity, neurotransmitter release, maintenance of membrane potentials, Ca++ homeostasis or restoration of ion gradients (for a review, see [10]). We have further shown recently that Cr might be involved in axonal growth [19]. Cr poorly crosses the blood brain barrier of rodents [20,21]; high doses of Cr given over a long period of treatment only partially replenish brain Cr of AGAT and GAMT deficient patients [7,8]. It has thus been suggested that the postnatal and adult CNS might depend, at least for a part of its needs, on its own Cr synthesis [13]. This is however in contradiction with the fact that CT1 deficient patients, who should express AGAT and GAMT correctly in their CNS, are nevertheless depleted in intracerebral Cr stores [22].
Little information is available on AGAT, GAMT and CT1 in embryonic development. AGAT (mRNA) and GAMT (protein) were found in whole extracts of the developing mouse embryo [23,24]. CT1 mRNA has been shown in the E14 rat embryo, in the entire neuraxis as well as in non-neural tissue [15]. The materno-fetal transport of Cr has been demonstrated [25,26]. As Cr deficiencies lead to severe developmental delay, our aim was to investigate at what time and in which tissues the system for Cr synthesis and transport is expressed during embryonic development. We determined therefore the tissue distribution of AGAT, GAMT and CT1 gene expression in rat embryos aged of 12.5, 15.5 and 18.5 days, at mRNA and protein level using in situ hybridization and immunohistochemistry respectively.
Results
The developmental expression of AGAT, GAMT and CT1 genes was analyzed in E12.5, E15.5 and E18.5 rat embryos at the mRNA and protein levels. For each embryonic stage, patterns of AGAT, GAMT and CT1 expression were validated by i) the specificity of anti-AGAT, anti-GAMT and anti-CT1 antibodies (Figure 1), ii) the specificity of in situ hybridization probes (Figure 3 and [13]), and iii) the remarkable coherence between in situ hybridization (mRNA, blue) and immunohistochemistry (protein, red) stainings (see Figures 3 and 4 for E12.5, Figures 5 and 6 for E15.5, Figures 7 and 8 for E18.5). Brain structures enlarged in Figures 3 to 8 (neocortical epithelium, choroids plexus) are illustrated in Figure 2 at lower magnification.
Figure 1 Specificity of the anti-AGAT, anti-GAMT and anti-CT1 antibodies. Western blot analysis of cell extract from rat kidney, by anti-AGAT, anti-GAMT and anti-CT1 antibodies. 10 μg of proteins were loaded in each lane. MM is the molecular mass marker.
Figure 2 Neocortex of the rat embryo at E12.5, E15.5 and E18.5. Hematoxylin staining. A: E12.5. B: E15.5. C: E18.5. Neocortical and plexus choroid structures depicted by rectangles in A, B and C are enlarged in Figures 3 A,E,I; 4 B,E,H; 5 A,B,F,G,K,L; 6 A,E,I; 7 A,B,M,N; and 8 A,B. CP: choroid plexus; LV: lateral ventricle; N: neocortex; P: pallidum. Bar : 200 μm.
Figure 3 Expression of the AGAT, GAMT and CT1 mRNAs in the E12.5 rat embryo. In situ hybridization (mRNA, blue signal) experiments performed with antisense probes against AGAT (A-C), GAMT (E-G) and CT1 (I-K) mRNAs, and with the sense counterpart probes for AGAT (D), GAMT (H) and CT1 (L). A,E,I and D,H,L: dorsal telencephalic neuroepithelium; B,F,J: somites; C,G,K: liver. Bar : 100 μm.
Figure 4 Expression of the AGAT, GAMT and CT1 proteins in the E12.5 rat embryo. Immunohistochemistry (protein, red signal) experiments performed with anti-AGAT (A-C), anti-GAMT (D-F) and anti-CT1 (G-I) antibodies. A,D,G : coronal section through hindbrain; B,E,H: dorsal telencephalic neuroepithelium; C,F,I: liver. Bar : A,D,G: 200 μm; B,C,E,F,H,I: 100 μm.
Figure 5 Expression of the AGAT, GAMT and CT1 mRNAs in the E15.5 rat embryo. In situ hybridization (mRNA, blue signal) experiments performed with antisense probes against AGAT (A-E), GAMT (F-J) and CT1 (K-O) mRNAs. A,F,K: neocortical neuroepithelium; B,G,L: choroid plexus; C,H,M: gonadal primordium; D,I,N: kidney; E,J,O: liver. Bar : 100 μm.
Figure 6 Expression of the AGAT, GAMT and CT1 proteins in the E15.5 rat embryo. Immunohistochemistry (protein, red signal) experiments performed with anti-AGAT (A-D), anti-GAMT (E-H) and anti-CT1 (I-L) antibodies. A,E,I: neocortical neuroepithelium; B,F,J: gonadal primordium; C,G,L: kidney; D,H,L: liver. Bar : 100 μm.
Figure 7 Expression of AGAT and GAMT in the E18.5 rat embryo. In situ hybridization (mRNA, blue signal) and immunohistochemistry (protein, red signal) experiments against AGAT (A-L) and GAMT (M-X). A,B,M,N: neocortical neuroepithelium; C,D,O,P: skeletal muscle; E,F,Q,R: kidney; G,H,S,T: liver; I,J,U,V: pancreas; K,L,W,X: intestine. Bar : 100 μm.
Figure 8 Expression of CT1 in the E18.5 rat embryo. In situ hybridization (mRNA, blue signal) and immunohistochemistry (protein, red signal) experiments against CT1. A,B: neocortical neuroepithelium; C,D: skin; E,F: kidney; G,H: intestine. Bar : 100 μm.
Figure 9 Expression of AGAT mRNA in the main tissues of the rat embryo. i : isolated cells ; - : absent ; ± : barely detectable ; + : moderate expression, ++ : strong expression, +++ : very strong expression. More differentiated structures of older stages are connected with the “{“ sign to the younger structure from which they originate.
Figure 10 Expression of GAMT mRNA in the main tissues of the rat embryo. i : isolated cells ; - : absent ; ± : barely detectable ; + : moderate expression, ++ : strong expression, +++ : very strong expression. More differentiated structures of older stages are connected with the “{“ sign to the younger structure from which they originate.
Figure 11 Expression of CT1 mRNA in the main tissues of the rat embryo. i : isolated cells ; - : absent ; ± : barely detectable ; + : moderate expression, ++ : strong expression, +++ : very strong expression. More differentiated structures of older stages are connected with the “{“ sign to the younger structure from which they originate.
E12.5 rat embryos
AGAT was detected in all parts of the E12.5 central nervous system (Figure 9; Figures 3A; 4A and 4B). Highest levels were detected in the middle part of somites (Figure 3B) as well as in the hepatic primordium (Figures 3C; 4C). AGAT was also detected in the wall of dorsal aorta, but was not detectable in epidermis and heart (Figure 9).
GAMT at E12.5 was detectable in the hepatic primordium only (Figures 3G; 4F), with all other tissues being negative (Figure 10, Figures 3E–F; 4D and 4E).
CT1 was expressed by all parts of the E12.5 CNS (Figure 11; Figures 3I; 4G and 4E). In contrast to AGAT which was restricted to the central part of somites, CT1 was found expressed throughout the whole somites (Figure 3J). CT1 was also present in the dorsal aorta and the hepatic primordium (Figures 3K; 4I). At E12.5, CT1 was not detectable in epidermis and heart (Figure 11).
In situ hybridization control sections hybridized with the respective AGAT, GAMT and CT1 sense probes were negative (Figure 3D, 3H and 3L).
E15.5 rat embryos
AGAT was expressed in most regions of the E15.5 CNS, with increased levels detected in isolated cells throughout the developing brain (Figure 9; Figures 5A; 6A). AGAT was not detected in choroid plexus (Figure 5B), nor in the dorsal root ganglia (Figure 9). High levels of AGAT were observed in skeletal muscles and primordia of gonads (Figures 5C; 6B), and AGAT was also detected in caudal somites (as shown earlier at E12.5, Fig. 3B), liver (Figures 5E; 6D) and pancreas, with, as in CNS, higher levels in isolated cells (Figure 9). At E15.5, AGAT could not be detected in kidney (Figures 5D; 6C), nor in all the remaining tissues observed (Figure 9).
GAMT could not be detected in most of the E15.5 CNS (Figures 5F; 6E), with the exception of striatum and pons (Figure 10). High levels of GAMT were found in liver (Figures 5J; 6H), and it was also expressed in pancreas. All the remaining tissues observed were negative for GAMT at E15.5 (Figure 10).
CT1 was strongly expressed in most of the E15.5 CNS (Figure 11; Figures 5K; 6I). CT1 was detected in choroid plexus (Figure 5L). CT1 was observed in skeletal muscles and caudal somites, as well as in kidney (Figures 5N; 6K), lung, stomach and intestine epithelial cells (Figure 11). CT1 could not be detected in gonads (Figures 5M; 6J), liver (Figures 5O; 6L) and pancreas, nor in all the remaining tissues observed (Figure 11).
E18.5 rat embryos
AGAT mRNA was expressed in most regions of the E18.5 CNS (Figure 9). Interestingly, high levels of AGAT mRNA were detected in endothelial cells of the developing cerebral capillaries (Figure 7A), but was absent from choroid plexus (Figure 9). A high expression of AGAT, both at the mRNA and protein levels, was also detected in skeletal muscle (Figure 7C,D), kidney (Figure 7E,F) and pancreas (Figure 7I,J), while a lower level of AGAT was found in liver (Figure 7G,H) and intestine epithelial cells (Figure 7K,L). At E18.5, AGAT could not be detected in epidermis and dermis, in olfactory epithelium, trachea and lung, nor in stomach and heart (Figure 9).
GAMT was expressed in neocortex (Figure 7M,N), hippocampus, striatum, pallidum and spinal cord, but could not be detected in other structures of the E18.5 brain (Figure 10). GAMT was absent from endothelial cells of cerebral capillaries, as well as from choroid plexus. In peripheral tissues, the highest expression of GAMT, both at the mRNA and protein levels, was found in skeletal muscles (Figure 7O,P) and liver (Figure 7S,T), while it was also present in pancreas (Figure 7U,V) and intestinal epithelial cells (Figure 7W,X). GAMT could not be detected in the E18.5 kidney (Figure 7Q,R), nor in all the remaining tissues observed (Figure 10).
CT1 was highly expressed in most regions of the E18.5 CNS, both at the mRNA and protein levels (Figure 11, Figure 8A,B). CT1 was absent from endothelial cells of cerebral capillaries, but highly expressed in choroid plexus (Figure 11). Most peripheral tissues expressed CT1 (Figure 11), as illustrated for epidermis and dermis (Figure 8C,D), kidney (Figure 8E,F) and intestine epithelial cells (Figure 8G,H). However, no signal for CT1 could be detected in liver and pancreas (Figure 11).
Discussion
Creatine in embryonic development
The Cr / P-Cr / CK system plays an essential role in energy homeostasis during the vertebrate embryonic development, with prominent activities in tissues such as developing CNS and muscles [9]. CK genes are expressed very early in many structures of the vertebrate embryo [27,28], whereas Cr concentration between 2 and 4 mmol/kg wet weight are found in the whole rat fetus, depending on the developmental stage [29]. In CNS, Cr concentration between 5 and 8 mmol/kg wet weight have been found in rat and human fetus, depending on the gestational stage [29,30]. Parts of the developmental needs for Cr can be fulfilled by the active transport of Cr from the mother to the embryo [25,26,31]. GAMT knock-out mice, which develop biochemical alterations comparable to those found in human GAMT deficient patients, present an increase in perinatal mortality [24]. It is not known however whether alterations in Cr pathways, as found in AGAT, GAMT or CT1 deficiencies, impairs the development of the embryo.
AGAT and GAMT for de novo synthesis of Cr in the embryo
As both substrates arginine (for AGAT) and S-adenosylmethionine (for GAMT) are available for most tissues (including liver, brain and muscle) during the mammalian embryonic development [32-34], our study suggests that the rat embryo might be able of its own Cr synthesis as soon as E12.5, in the hepatic primordium which expresses both AGAT and GAMT. From E12.5 to E18.5, both enzymes then progressively acquire their expression pattern found in aduldhood, with AGAT mainly expressed in kidney and pancreas, and GAMT preferentially localized in liver and pancreas (see [10] and references therein). As in adulthood however, many other different embryonic tissues retain low levels of AGAT and/or GAMT; we found that skeletal muscles and intestinal epithelial cells at E18.5 are equipped to synthesize their own Cr by expressing both AGAT and GAMT. A few embryonic structures express only AGAT (muscles before E18.5, regions of CNS, blood vessels) suggesting that they have to release the intermediate GAA, that has to be transported to GAMT-expressing cells for Cr synthesis to occur, as it is generally thought between kidney (AGAT) and liver (GAMT) in adult mammals. A few embryonic structures (i.e. somites at E12.5 and E15.5, skeletal muscles at E15.5 and E18.5, gonadal primordium at E15.5) showed a striking and very high expression of AGAT. This might be explained by the strong and positive regulation of AGAT, at the transcriptional level, by thyroid, growth and sex hormones [26,35,36], which do control the embryonic development of somites, skeletal muscles and gonads [37-40].
While this study shows that AGAT is expressed as soon as E12.5 in the whole CNS parenchyme and increases towards E18.5, it demonstrates also that GAMT expression in the developing CNS is delayed, and remains at low levels as compared to AGAT. Adult rat CNS appears to be able of synthesizing its own Cr by expressing AGAT and GAMT in most cell types of the brain [13]; in contrast, our work suggests that independance of CNS for its own Cr synthesis is probably limited to the end of embryogenesis and restricted to discrete regions of CNS. Thus, the embryonic brain might depend mainly on extra-CNS supply of Cr, be it of maternal origin or synthesized in other tissues of the embryo (see below).
CT1 for the supply of Cr to embryonic tissues
From E12.5 to E18.5, most of the embryonic tissues progressively express CT1, allowing cellular Cr uptake. Notable exceptions are liver (apart from a low and transient expression at E12.5) and pancreas, which express GAMT, and endothelial cells of brain capillaries (see below). These results are in accordance with earlier work on the E14 rat embryo [15]. Highest levels of CT1 are found as soon as E12.5, in tissues requiring high amounts of Cr, like somites, skeletal muscles and CNS. In gonads, CT1 and GAMT appeared at E18.5 only, while AGAT was highly expressed at E15.5 and decreased at E18.5. This is in accordance with their respective expression in the adult reproductive tract [12,41]. CT2, a second potential Cr transporter showing 95% nucleotide identity with the CT1 coding sequence, has been found expressed in testis solely, at the RNA level [41]. However, CT2 most probably represents a pseudogene, as it cannot be fully translated into a functional protein [42], suggesting that the needs in Cr of the reproductive tract are fulfilled by CT1, as well as AGAT and GAMT.
The delayed and low level of GAMT expression during CNS embryogenesis suggests that the embryonic brain might depend on extra-CNS supply of Cr. This is supported by our results on CT1, which is expressed at high levels in the whole embryonic CNS, with a peak at E15.5. Interestingly, CT1 is expressed in the choroid plexus of the rat embryo (E15.5 and E18.5), but not of the adult rat brain [13]. In contrast, CT1 is absent from endothelial cells of brain capillaries in the embryo, while these cells express it in the adult rat CNS [13,20]. As choroid plexus differentiates earlier than brain capillaries and participates to early trophic supply for CNS [43,44], one might speculate that before angiogenesis occurs in CNS parenchyme, extra-CNS Cr is supplied from blood to cerebrospinal fluid through the choroid plexus. Cr would then be available for the whole embryonic brain through cerebrospinal fluid circulation [45] and the observed high levels of CT1 in the developing neuroepithelium, particularly in ependymal epithelium along ventricles (see below CT1 in surface epithelia exposed to amniotic fluid). As brain develops and enlarges, and CNS angiogenesis progresses, the ratio of exchange surfaces in choroid plexus and CNS microcapillaries shifts to predominance of brain microcapillaries [43-45]. Thus, at the end of embryonic development and then postnatally, Cr supply to the brain may occur preferentially at microcapillaries, with CT1 being up-regulated in capillary endothelial cells and down-regulated in choroid plexus, as in adulthood [13,20]. It should be emphasized however that supply of Cr from blood to postnatal or adult brain is very likely of less quantitative importance than intra-cerebral Cr synthesis, as astrocytes around blood brain barrier do not express CT1 [13].
At E15.5 and E18.5, CT1 was found highly expressed in all epithelia of the rat embryo being in contact with amniotic fluid, i.e. epidermis, olfactory epithelium, trachea, lung, stomach, and intestine. Cr uptake has been confirmed recently in E16 rat embryo intestine [46]. Amniotic fluid contains significant amounts of Cr (50 to 100 μM in human depending on gestational age [47]; 320 μM in rat [46]). It is produced by amniotic cells (at the surface of chorion) and the foetus, and is continuously renewed by oral intake of the foetus and excretion (urine). Amniotic fluid may thus represent an easy way to supply Cr to many structures of the embryo, through epithelial expression of CT1 in embryonic parts where vasculature is not yet fully developed.
It is known that Cr is also supplied by active transport from the mother to the embryo, accumulating in the chorioallantoic placenta and yolk sac at concentrations higher than found in maternal or fetal blood, then diffusing down its concentration gradient into the fetal circulation [25,26,31]. Thus previous studies, as well as our data, suggest that materno-fetal transport of Cr and de novo synthesis of Cr in the embryo are both necessary for a normal development to occur.
In utero consequences of AGAT, GAMT or CT1 deficiencies
Patients suffering of AGAT, GAMT or CT1 deficiency present neurological symptoms in early infancy and show severe neurodevelopmental delay [7,22]. AGAT and GAMT deficiencies can be treated with oral Cr, which slowly replenishes brain levels of Cr [1,3-5,48-50]. Treatment of CT1 deficient patients with oral Cr does not replenish their CNS Cr level [51]. Despite developmental improvement and recovery of AGAT and GAMT deficient patients treated with Cr, sequelae to brain development and mental retardation remain [7,8]. For GAMT deficiency, this may be due to the toxicity of the GAA accumulating in CNS. Most patients with AGAT, GAMT or CT1 deficiencies are diagnosed during infancy, and significant damage to their brain occur postnatally. However, AGAT, GAMT and CT1 expression patterns during the rat embryogenesis suggests that some of the irreversible damages observed in Cr deficient patients, lacking either AGAT, GAMT or CT1, may already occur in utero.
Conclusion
We have shown that AGAT, GAMT and CT1 are expressed by various tissues throughout the development of the rat embryo. This study suggests that de novo synthesis of Cr and Cr uptake are important for embryonic development. This work provides new clues on how creatine can be provided to developing tissues, and suggests moreover that irreversible damage observed in Cr deficient patients, lacking either AGAT, GAMT or CT1, may already occur in utero.
Methods
Preparation of E12.5, E15.5 and E18.5 rat embryos
All animal procedures were in compliance with the directives of the Swiss Academy of Medical Science. Sprague-Dawley pregnant female rats (Charles River Laboratories, France) were fed a standard chow formula. Embryonic stages of rat embryos were determined from the appearance of vaginal plug in pregnant females. 2 pregnant females were used for each embryonic stage (E12.5: 12.5 days of gestation, 7 embryos; E15.5: 15.5 days of gestation, 5 embryos; E18.5: 18.5 days of gestation, 5 embryos). For each embryonic stage, females were sacrificed by decapitation, and embryos were removed from uterus, rinsed in diethylpyrocarbonate (DEPC)-treated PBS and fixed for 15 h at room temperature in 4% paraformaldehyde in DEPC-treated PBS. Subsequently, embryos were cryoprotected at 4°C in 12% and 18% sucrose in DEPC-treated PBS for 18 h and 24 h respectively, then embedded in tissue-freezing medium (Jung, Nussloch, Germany) and frozen in liquid nitrogen cooled isopentane. Embryos were stored at -80°C until used for cryosections.
In situ hybridization and immunohistochemistry
Partial cDNAs of the rat sequences AGAT (nt 182-1314, Gene Bank accession number U07971), GAMT (nt 131-734, Gene Bank J03588) and CT1 (nt 901-2544, Gene Bank NM_01738) were used to synthesize antisense and sense digoxigenin-labeled AGAT, GAMT and CT1 riboprobes as described previously [13]. 12 μm thick cryosections (Leica CM 1800) were prepared for each embryonic stage, which were analyzed by a sensitive technique of non-radioactive in situ hybridization [52]. Briefly, cryosections were postfixed 10 min in 4% paraformaldehyde in DEPC-treated PBS, washed 2 × 15 min in PBS containing 0.1% fresh DEPC and equilibrated 15 min in 5 × SSC. Sections were hybridized (58°C for 40 h in 5 × SSC, 50% formamide and 40 μg/ml salmon sperm DNA) with the digoxigenin-labeled antisense and sense riboprobes (400 ng/ml) for rat AGAT, GAMT and CT1. Sections were then washed (30 min in 2 × SSC at room temperature, 1 h in 2 × SSC at 65°C, 1 h in 0.1 × SSC at 65°C) and stained with alkaline phosphatase-coupled anti-digoxygenin antibody (Roche, Basel, Switzerland) using nitroblue tetrazolium and 5-bromo-4-chloro-3-indolyl-phosphate. The specificity of hybridization was ascertained by the use of sense probes for AGAT, GAMT and CT1 genes having the same length, GC content and digoxigenin incorporation as their antisense counterparts. In each in situ hybridization experiment, a section hybridized with an antisense probe was always followed by an adjacent section hybridized with the corresponding sense control probe. After staining, sections were dehydrated and mounted (Eukit, O. Kindler Co, Freiburg, Germany).
AGAT, GAMT and CT1 proteins were detected with rabbit polyclonal antibodies, that were made through injection of the following antigenic peptides: AGAT N-terminal amino acids (aa) 62-77 and C-terminal aa 410-423(SwissProt, accession number P50442); GAMT aa 27-227(SwissProt, accession number P10868); and CT1 N-terminal aa 15-29 (SwissProt, accession number P28570). Specific immunoglobulins against AGAT, GAMT and CT1 were obtained by peptide affinity chromatography. These antibodies recognize specific bands for AGAT (46 kDa), GAMT (26 kDa) and CT1 (61 kDa) respectively, by western blotting experiments (Figure 1). AGAT, GAMT and CT1 proteins were analyzed by immunohistochemistry on 8 μm thick cryosections using the 3 polyclonal antibodies described above. Briefly, cryosections were postfixed for 1 h in 4%PFA in PBS. Endogenous peroxidase activities were bleached using 1.5% H2O2 in PBS for 15 min. After blocking 1 h in 1% bovine serum albumin in PBS, anti-AGAT, anti-GAMT and anti-CT1 antibodies were incubated for 1 h at room temperature in the same buffer. Their detection by peroxidase staining was performed using the Histostain Plus Kit (Zymed Laboratories Inc) with aminoethyl carbazole and H2O2.
The expression patterns of AGAT, GAMT and CT1 genes observed during the embryonic development were considered specific and validated by (i) the excellent correlation, in the multiple embryonic tissues analysed, between the signals observed at the mRNA level by in situ hybridization and at the protein level by immunohistochemistry, (ii) the negative signals observed with the in situ hybridization sense probes (see Figure 3 and [13]), and (iii) by the absence of any labelling in immunohistochemical controls without primary antibodies or in presence of pre-immune serum (data not shown).
Sections were observed and photographed on an Olympus BX50 microscope equipped with a DP-10 digital camera (Olympus Opticals, Japan). Structures were identified according to [53-55], after staining of sections by hematoxylin (see Figure 2). In Figures 9,10 and 11 semi-quantitative levels of AGAT, GAMT and CT1 mRNA were determined based on ISH experiments, as described [56]. Levels of transcripts observed by optical microscopy are indicated by - and + signs which however do not represent a strict linear mesure of mRNA.
List of abbreviations used
AGAT: arginine:glycine amidinotransferase; CK: creatine kinase; CNS: central nervous system; Cr: creatine; CT1: creatine transporter 1; DEPC: diethylpyrocarbonate; GAMT: guanidinoacetate methyltransferase; P-Cr: phosphocreatine
Authors' contributions
OB conceived of the study, wrote the manuscript and performed immunohistochemistry experiments. HH designed and characterized the anti-GAMT and anti-CT1 antibodies, performed the western blotting experiments and participated in the writing of the manuscript. AMV carried out the in situ hybridization and immunohistochemistry experiments. OS designed and characterized the anti-AGAT antibody, and participated in the writing of the manuscript. TW and CB participated in the writing of the manuscript. All authors read and approved the final manuscript.
Acknowledgements
This work was funded by the Swiss National Science Foundation (grants n° 3100-063892 and 3100A0-100778 to OB, HH and CB; n° 3100A0-102075 to TW) and the Swiss Society for Research on Muscle Disease (to TW). Marc Loup and Barbara Eilers are gratefully acknowledged for their excellent technical work.
==== Refs
Stöckler S Holzbach U Hanefeld F Marquardt I Helms G Requart M Hänicke W Frahm J Creatine deficiency in the brain: a new, treatable inborn error of metabolism Pediatr Res 1994 36 409 413 7808840
Salomons GS van Dooren SJ Verhoeven NM Cecil KM Ball WS DeGrauw TJ Jakobs C X-linked creatine-transporter gene (SLC6A8) defect: a new creatine-deficiency syndrome Am J Hum Genet 2001 68 1497 1500 11326334 10.1086/320595
Item CB Stöckler-Ipsiroglu S Stromberger C Muhl A Alessandri MG Bianchi MC Tosetti M Fornai F Cioni G Arginine:glycine amidinotransferase deficiency: the third inborn error of creatine metabolism in humans Am J Hum Genet 2001 69 1127 1133 11555793 10.1086/323765
Schulze A Hess T Wevers R Mayatepek E Bachert P Marescau B Knopp MV De Deyn PP Bremer HJ Rating D Creatine deficiency syndrome caused by guanidinoacetate methyltransferase deficiency: diagnostic tools for a new inborn error of metabolism J Pediatr 1997 131 626 631 9386672
Battini R Leuzzi V Carducci C Tosetti M Bianchi MC Item CB Stöckler-Ipsiroglu S Cioni G Creatine depletion in a new case with AGAT deficiency: clinical and genetic study in a large pedigree Mol Genet Metab 2002 77 326 331 12468279 10.1016/S1096-7192(02)00175-0
DeGrauw TJ Salomons GS Cecil KM Chuck G Newmeyer A Schapiro MB Jakobs C Congenital creatine transporter deficiency Neuropediatrics 2002 33 232 238 12536364 10.1055/s-2002-36743
Stromberger C Bodamer OA Stöckler-Ipsiroglu S Clinical characteristics and diagnostic clues in inborn errors of creatine metabolism J Inherit Metab Dis 2003 26 299 308 12889668 10.1023/A:1024453704800
Sykut-Cegielska J Gradowska W Mercimek-Mahmutoglu S Stöckler-Ipsiroglu S Biochemical and clinical characteristics of creatine deficiency syndromes Acta Biochim Pol 2004 51 875 882 15625559
Wallimann T Wyss M Brdiczka D Nicolay K Eppenberger HM Intracellular compartmentation, structure and function of creatine kinase isoenzymes in tissues with high and fluctuating energy demands: the 'phosphocreatine circuit' for cellular energy homeostasis Biochem J 1992 281 21 40 1731757
Wyss M Kaddurah-Daouk R Creatine and creatinine metabolism Physiol Rev 2000 80 1107 1213 10893433
Van Pilsum JF Stephens GC Taylor D Distribution of creatine, guanidinoacetate and enzymes for their biosynthesis in the animal kingdom. Implications for phylogeny Biochem J 1972 126 325 345
Lee H Kim JH Chae YJ Ogawa H Lee MH Gerton GL Creatine synthesis and transport systems in the male rat reproductive tract Biol Reprod 1998 58 1437 1444 9623603
Braissant O Henry H Loup M Eilers B Bachmann C Endogenous synthesis and transport of creatine in the rat brain: an in situ hybridization study Mol Brain Res 2001 86 193 201 11165387 10.1016/S0169-328X(00)00269-2
Guimbal C Kilimann MW A Na+-dependent creatine transporter in rabbit brain, muscle, heart, and kidney. cDNA cloning and functional expression J Biol Chem 1993 268 8418 8421 8473283
Schloss P Mayser W Betz H The putative rat choline transporter CHOT1 transports creatine and is highly expressed in neural and muscle-rich tissues Biochem Biophys Res Commun 1994 198 637 645 8297374 10.1006/bbrc.1994.1093
Happe HK Murrin LC In situ hybridization analysis of CHOT1, a creatine transporter, in the rat central nervous system J Comp Neurol 1995 351 94 103 7896942 10.1002/cne.903510109
Hiel H Happe HK Warr WB Morley BJ Regional distribution of a creatine transporter in rat auditory brainstem: an in situ hybridization study Hear Res 1996 98 29 37 8880179 10.1016/0378-5955(96)00046-9
Saltarelli MD Bauman AL Moore KR Bradley CC Blakely RD Expression of the rat brain creatine transporter in situ and in transfected HeLa cells Dev Neurosci 1996 18 524 534 8940628
Braissant O Henry H Villard AM Zurich MG Loup M Eilers B Parlascino G Matter E Boulat O Honegger P Bachmann C Ammonium-induced impairment of axonal growth is prevented through glial creatine J Neurosci 2002 22 9810 9820 12427837
Ohtsuki S Tachikawa M Takanaga H Shimizu H Watanabe M Hosoya K Terasaki T The blood-brain barrier creatine transporter is a major pathway for supplying creatine to the brain J Cereb Blood Flow Metab 2002 22 1327 1335 12439290 10.1097/00004647-200211000-00006
Perasso L Cupello A Lunardi GL Principato C Gandolfo C Balestrino M Kinetics of creatine in blood and brain after intraperitoneal injection in the rat Brain Res 2003 974 37 42 12742622 10.1016/S0006-8993(03)02547-2
Salomons GS van Dooren SJ Verhoeven NM Marsden D Schwartz C Cecil KM DeGrauw TJ Jakobs C X-linked creatine transporter defect: an overview J Inherit Metab Dis 2003 26 309 318 12889669 10.1023/A:1024405821638
Sandell LL Guan XJ Ingram R Tilghman SM Gatm, a creatine synthesis enzyme, is imprinted in mouse placenta Proc Natl Acad Sci USA 2003 100 4622 4627 12671064 10.1073/pnas.0230424100
Schmidt A Marescau B Boehm EA Renema WK Peco R Das A Steinfeld R Chan S Wallis J Davidoff M Ullrich K Waldschutz R Heerschap A De Deyn PP Neubauer S Isbrandt D Severely altered guanidino compound levels, disturbed body weight homeostasis and impaired fertility in a mouse model of guanidinoacetate N-methyltransferase (GAMT) deficiency Hum Mol Genet 2004 13 905 921 15028668 10.1093/hmg/ddh112
Davis BM Miller RK Brent RL Koszalka TR Materno-fetal transport of creatine in the rat Biol Neonate 1978 33 43 54 656521
Walker JB Creatine: biosynthesis, regulation, and function Adv Enzymol 1979 50 177 242 386719
Lyons GE Muhlebach S Moser A Masood R Paterson BM Buckingham ME Perriard JC Developmental regulation of creatine kinase gene expression by myogenic factors in embryonic mouse and chick skeletal muscle Development 1991 113 1017 1029 1668275
Dickmeis T Rastegar S Aanstad P Clark M Fischer N Plessy C Rosa F Korzh V Strahle U Expression of brain subtype creatine kinase in the zebrafish embryo Mech Dev 2001 109 409 412 11731259 10.1016/S0925-4773(01)00536-6
Miller TJ Hanson RD Yancey PH Developmental changes in organic osmolytes in prenatal and postnatal rat tissues Comp Biochem Physiol A Mol Integr Physiol 2000 125 45 56 10779730 10.1016/S1095-6433(99)00160-9
Kreis R Hofmann L Kuhlmann B Boesch C Bossi E Huppi PS Brain metabolite composition during early human brain development as measured by quantitative in vivo 1H magnetic resonance spectroscopy Magn Reson Med 2002 48 949 958 12465103 10.1002/mrm.10304
Koszalka TR Jensh RP Brent RL Placental transport of creatine in the rat Proc Soc Exp Biol Med 1975 148 864 869 1129310
Wu G Meininger CJ Knabe DA Bazer FW Rhoads JM Arginine nutrition in development, health and disease Curr Opin Clin Nutr Metab Care 2000 3 59 66 10642085 10.1097/00075197-200001000-00010
Nishimura K Nakatsu F Kashiwagi K Ohno H Saito T Igarashi K Essential role of S-adenosylmethionine decarboxylase in mouse embryonic development Genes Cells 2002 7 41 47 11856372 10.1046/j.1356-9597.2001.00494.x
Zhu X Mar MH Song J Zeisel SH Deletion of the Pemt gene increases progenitor cell mitosis, DNA and protein methylation and decreases calretinin expression in embryonic day 17 mouse hippocampus Brain Res Dev Brain Res 2004 149 121 129 15063092 10.1016/j.devbrainres.2004.01.004
McGuire DM Tormanen CD Segal IS Van Pilsum JF The effect of growth hormone and thyroxine on the amount of L-arginine:glycine amidinotransferase in kidneys of hypophysectomized rats. Purification and some properties of rat kidney transamidinase J Biol Chem 1980 255 1152 1159 6766137
Guthmiller P Van Pilsum JF Boen JR McGuire DM Cloning and sequencing of rat kidney L-arginine:glycine amidinotransferase. Studies on the mechanism of regulation by growth hormone and creatine J Biol Chem 1994 269 17556 17560 8021264
Wang JJ Immunocytochemical demonstration of the binding of growth-related polypeptide hormones on chick embryonic tissues Histochemistry 1989 93 133 141 2613553 10.1007/BF00315966
Muscat GE Downes M Dowhan DH Regulation of vertebrate muscle differentiation by thyroid hormone: the role of the myoD gene family Bioessays 1995 17 211 218 7748175 10.1002/bies.950170307
Harvey S Johnson CD Sanders EJ Extra-pituitary growth hormone in peripheral tissues of early chick embryos J Endocrinol 2000 166 489 502 10974643 10.1677/joe.0.1660489
Luna M Huerta L Berumen L Martinez-Coria H Harvey S Aramburo C Growth hormone in the male reproductive tract of the chicken: heterogeneity and changes during ontogeny and maturation Gen Comp Endocrinol 2004 137 37 49 15094334 10.1016/j.ygcen.2004.02.005
Iyer GS Krahe R Goodwin LA Doggett NA Siciliano MJ Funanage VL Proujansky R Identification of a testis-expressed creatine transporter gene at 16p11.2 and confirmation of the X-linked locus to Xq28 Genomics 1996 34 143 146 8661037 10.1006/geno.1996.0254
Eichler EE Lu F Shen Y Antonacci R Jurecic V Doggett NA Moyzis RK Baldini A Gibbs RA Nelson DL Duplication of a gene-rich cluster between 16p11.1 and Xq28: a novel pericentromeric-directed mechanism for paralogous genome evolution Hum Mol Genet 1996 5 899 912 8817324 10.1093/hmg/5.7.899
Dziegielewska KM Ek J Habgood MD Saunders NR Development of the choroid plexus Microsc Res Tech 2001 52 5 20 11135444 10.1002/1097-0029(20010101)52:1<5::AID-JEMT3>3.0.CO;2-J
Engelhardt B Development of the blood-brain barrier Cell Tissue Res 2003 314 119 129 12955493 10.1007/s00441-003-0751-z
Segal MB The choroid plexuses and the barriers between the blood and the cerebrospinal fluid Cell Mol Neurobiol 2000 20 183 196 10696509 10.1023/A:1007045605751
Peral MJ Galvez M Soria ML Ilundain AA Developmental decrease in rat small intestinal creatine uptake Mech Ageing Dev 2005 126 523 530 15722111 10.1016/j.mad.2004.11.005
Groenen PM Engelke UF Wevers RA Hendriks JC Eskes TK Merkus HM Steegers-Theunissen RP High-resolution 1H NMR spectroscopy of amniotic fluids from spina bifida fetuses and controls Eur J Obstet Gynecol Reprod Biol 2004 112 16 23 14687733 10.1016/S0301-2115(03)00279-3
Stöckler S Isbrandt D Hanefeld F Schmidt B Von Figura K Guanidinoacetate methyltransferase deficiency: the first inborn error of creatine metabolism in man Am J Hum Genet 1996 58 914 922 8651275
Ganesan V Johnson A Connelly A Eckhardt S Surtees RA Guanidinoacetate methyltransferase deficiency: new clinical features Pediatr Neurol 1997 17 155 157 9367297 10.1016/S0887-8994(97)00083-0
Schulze A Mayatepek E Bachert P Marescau B De Deyn PP Rating D Therapeutic trial of arginine restriction in creatine deficiency syndrome Eur J Pediatr 1998 157 606 607 9686828 10.1007/s004310050890
Cecil KM Salomons GS Ball WS Wong B Chuck G Verhoeven NM Jakobs C DeGrauw TJ Irreversible brain creatine deficiency with elevated serum and urine creatine: a creatine transporter defect? Ann Neurol 2001 49 401 404 11261517 10.1002/ana.79
Braissant O Measurement of nitric oxide-related enzymes in brain by in situ hybridization Meth Mol Biol 2004 279 113 224
Hebel R Stromberg MW Anatomy and embryology of the laboratory rat 1986 Wörthsee: BioMed Verlag
Kaufman M The atlas of mouse development 1992 London: Academic Press
Altman J Bayer S Atlas of prenatal rat brain development 1995 Boca Raton: CRC Press Inc
Braissant O Gotoh T Loup M Mori M Bachmann C Differential expression of the cationic amino acid transporter 2(B) in the adult rat brain Mol Brain Res 2001 91 189 195 11457509 10.1016/S0169-328X(01)00113-9
| 15918910 | PMC1175845 | CC BY | 2021-01-04 16:40:17 | no | BMC Dev Biol. 2005 May 26; 5:9 | utf-8 | BMC Dev Biol | 2,005 | 10.1186/1471-213X-5-9 | oa_comm |
==== Front
BMC Evol BiolBMC Evolutionary Biology1471-2148BioMed Central London 1471-2148-5-351592707110.1186/1471-2148-5-35Research ArticleEvolutionary history of Wolbachia infections in the fire ant Solenopsis invicta Ahrens Michael E [email protected] Dewayne [email protected] Department of Entomology, 643 Russell Labs, 1630 Linden Drive, University of Wisconsin, Madison, WI 53706 USA2005 31 5 2005 5 35 35 5 1 2005 31 5 2005 Copyright © 2005 Ahrens and Shoemaker; licensee BioMed Central Ltd.2005Ahrens and Shoemaker; licensee BioMed Central Ltd.This is an Open Access article distributed under the 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
Wolbachia are endosymbiotic bacteria that commonly infect numerous arthropods. Despite their broad taxonomic distribution, the transmission patterns of these bacteria within and among host species are not well understood. We sequenced a portion of the wsp gene from the Wolbachia genome infecting 138 individuals from eleven geographically distributed native populations of the fire ant Solenopsis invicta. We then compared these wsp sequence data to patterns of mitochondrial DNA (mtDNA) variation of both infected and uninfected host individuals to infer the transmission patterns of Wolbachia in S. invicta.
Results
Three different Wolbachia (wsp) variants occur within S. invicta, all of which are identical to previously described strains in fire ants. A comparison of the distribution of Wolbachia variants within S. invicta to a phylogeny of mtDNA haplotypes suggests S. invicta has acquired Wolbachia infections on at least three independent occasions. One common Wolbachia variant in S. invicta (wSinvictaB) is associated with two divergent mtDNA haplotype clades. Further, within each of these clades, Wolbachia-infected and uninfected individuals possess virtually identical subsets of mtDNA haplotypes, including both putative derived and ancestral mtDNA haplotypes. The same pattern also holds for wSinvictaA, where at least one and as many as three invasions into S. invicta have occurred. These data suggest that the initial invasions of Wolbachia into host ant populations may be relatively ancient and have been followed by multiple secondary losses of Wolbachia in different infected lineages over time. Finally, our data also provide additional insights into the factors responsible for previously reported variation in Wolbachia prevalence among S. invicta populations.
Conclusion
The history of Wolbachia infections in S. invicta is rather complex and involves multiple invasions or horizontal transmission events of Wolbachia into this species. Although these Wolbachia infections apparently have been present for relatively long time periods, these data clearly indicate that Wolbachia infections frequently have been secondarily lost within different lineages. Importantly, the uncoupled transmission of the Wolbachia and mtDNA genomes suggests that the presumed effects of Wolbachia on mtDNA evolution within S. invicta are less severe than originally predicted. Thus, the common concern that use of mtDNA markers for studying the evolutionary history of insects is confounded by maternally inherited endosymbionts such as Wolbachia may be somewhat unwarranted in the case of S. invicta.
==== Body
Background
Innumerable insects and other terrestrial arthropods are infected with maternally transmitted endosymbionts. While many endosymbionts spread by increasing the fitness of their hosts, others spread by manipulating host reproduction in ways that specifically enhance transmission of infected cytoplasm, even if this results in reduced transmission of nuclear genes [1]. In these latter cases, such symbionts act as parasites. Parasitic endosymbionts are extremely prevalent in nature, and include many bacteria in the genus Wolbachia [2-4]. These endosymbiotic bacteria infect a wide variety of arthropods and filarial nematodes [2-4]. Although Wolbachia infecting filarial nematodes generally are considered mutualists, most Wolbachia strains infecting insects act as parasites. Recent surveys suggest that Wolbachia infect a substantial proportion of insect species, with estimates ranging from 17% [5-7] to 76% [8]. Extrapolation of these estimates suggests that millions of insect species are currently infected with Wolbachia, making these bacteria among the most widespread parasites on earth.
Wolbachia transmission within host species mainly occurs maternally through the egg cytoplasm, and as such, these microbes have evolved several mechanisms to enhance their own transmission that either increase their host's investment in daughters or decrease the reproductive success of uninfected females. These mechanisms include cytoplasmic incompatibility (CI), thelytokous parthenogenesis, feminization of genetic males, and male-killing [for recent reviews see [1,9,10]]. In addition to their vertical (maternal) transmission from mother to offspring, several independent lines of evidence clearly show Wolbachia are also horizontally transmitted both within and among different host species [3,11-18]. However, despite knowledge that Wolbachia can be transmitted horizontally, a general understanding of the frequency and mode of horizontal transmission within natural host populations is poorly documented.
One approach often employed to infer the transmission patterns and evolutionary history of Wolbachia infections within a given host species is to compare patterns of Wolbachia and host mtDNA genetic variation [19-40]. If the two genomes are strictly co-transmitted vertically from mother to offspring as predicted, then there should be strong linkage between a host's mtDNA genome and the associated Wolbachia genome. Depending on the age of infection, such linkage should be observable in patterns of molecular variation of the two genomes such that a given Wolbachia strain is associated with a particular mtDNA haplotype or clade of haplotypes [24,30,38,41-47]. On the other hand, this tight association is lost if horizontal transmission of Wolbachia occurs, in which case one would not necessarily expect concordant patterns of variation between the two genomes. As an example of using this approach, extensive studies of Drosophila simulans have revealed that this species is infected with at least four genetically distinct strains of Wolbachia, presumably representing four independent invasions across three distinct clades of mitochondrial haplotypes [26,30-37,44].
Several studies have been conducted examining the distribution and prevalence of Wolbachia infections among native South American populations of the fire ant Solenopsis invicta, as well as the effects of Wolbachia on mtDNA variation in this species. The general findings of these previous studies were: 1) the prevalence of Wolbachia infections varies significantly among different native geographic populations of S. invicta, 2) two divergent mtDNA haplotype lineages and two Wolbachia variants occur within S. invicta, and 3) a strong association between each Wolbachia variant and host mtDNA lineage exists, albeit these latter two conclusions were based on a relatively small number of samples from only two populations [38,46,48]. Interestingly, despite the apparent strong association between genomes, as well as evidence for a high fidelity of maternal transmission of Wolbachia within colonies of S. invicta in the field, Shoemaker et al. [46] found no consistent correlation between the presence of Wolbachia and either levels or patterns of mtDNA diversity. That is, levels of mtDNA variation in Wolbachia-infected and uninfected populations were similar and patterns of mtDNA variation within Wolbachia-infected populations did not differ consistently from neutral expectations, despite the prediction that strong positive selection acting on Wolbachia influences the evolutionary dynamics of other cytoplasmic genomes [46]. There are three potential non-mutually exclusive explanations for these puzzling results: 1) Wolbachia infections in S. invicta are sufficiently ancient so that levels of mtDNA variation have re-equilibrated to their levels prior to invasion of Wolbachia, 2) Wolbachia infections are horizontally transmitted within S. invicta such that the two genomes are not strictly co-transmitted as previously suggested, or 3) the evolutionary history of Wolbachia infections within S. invicta involves multiple independent invasions of one or more Wolbachia variants.
The major goal of the present study was to infer the transmission patterns and evolutionary history of Wolbachia infections within S. invicta. To accomplish our objective, we generated sequence data from two portions of the Wolbachia genome present in numerous infected individuals of S. invicta collected throughout the species' native range and subsequently compared these data to patterns of mtDNA variation to determine the extent of Wolbachia strain variation as well as the predominant mode of Wolbachia transmission in this species. In addition, we also use these data to address the issue of whether or not the significant variation in Wolbachia prevalence among fire ant populations is simply due to the presence of different Wolbachia variants in these populations. As we show below, our results based on these extensive sequence data lead to new insights regarding the history of Wolbachia infections in S. invicta, and in so doing, partly explain the paradoxical findings of previous studies on these ants.
Results and discussion
Diversity of Wolbachia strains in S. invicta
Our Wolbachia (wsp) sequence data, which includes partial wsp sequences from 138 Wolbachia-infected individuals, revealed only three unique variants within S. invicta. All three variants are identical to previously reported Wolbachia (wsp) variants from fire ants and fall into one of the two divergent major Wolbachia subgroups comprising Wolbachia strains specific to New World ants (InvA and InvB) [49,50]. Two of the variants were identical to Wolbachia variants previously reported to infect S. invicta (wSinvictaA and wSinvictaB; InvA and InvB subgroups, respectively), whereas the third "new" variant is identical to a variant previously reported to infect the closely related fire ant species S. richteri (wSrichteriA; InvA subgroup) [38].
Additionally, the Wolbachia 16S sequence data from a subset of infected individuals did not reveal any new Wolbachia variants within S. invicta. The 16S sequences from all individuals infected with the variants wSinvictaA and wSrichteriA (based upon wsp sequences) were identical to each other as were the 16S sequences from individuals infected with wSinvictaB. However, the 16S sequences from individuals infected with the variants wSinvictaA and wSrichteriA differed by a single nucleotide substitution from those in individuals infected with wSinvictaB. All 16S sequences belong to the A group of Wolbachia (as opposed to the A and B groups for wsp sequences). This discrepancy between the two genes most likely results from an historical recombination event within the Wolbachia genome, which perhaps is not unexpected given previous studies showing recombination of Wolbachia genomes commonly occurs [51,52].
Transmission patterns of Wolbachia in S. invicta
Both Wolbachia (wsp) and mtDNA sequence data were available for 133 of 138 infected individuals (Table 1 and Figure 1): MtDNA sequence data were lacking for the remaining five Wolbachia-infected individuals, which are excluded from the comparative analyses below. The distribution of all mtDNA haplotypes within each of the eleven populations is shown in Table 2 and the particular haplotypes that are associated with Wolbachia infections is shown in Table 2 and Figure 1. A comparison of wsp and mtDNA sequence variation suggests a complex evolutionary history of Wolbachia infections in S. invicta, involving multiple independent invasions of Wolbachia into S. invicta followed by frequent secondary loss of infections in different maternal lineages. Indeed, these data suggest that at least six independent invasions involving three different Wolbachia variants have occurred into S. invicta (scenario 1 of Figure 1). The variant wSinvictaB apparently invaded S. invicta on two separate occasions. One of these invasions is most likely a rather recent event, as it is associated with only three individuals, all of which harbour an identical mtDNA haplotype (haplotype #51; incidentally, all three individuals also are infected with the wSrichteriA variant). The other invasion of wSinvictaB into S. invicta is presumably more ancient as evidenced by the strong association of this variant with a highly divergent mtDNA clade comprising closely related mtDNA haplotypes (i.e., clade I in Figure 1). There appears to have been a single invasion of the wSrichteriA variant into S. invicta, as its presence is limited to a single clade of mtDNA haplotypes (clade IV), all of which come from individuals collected from two populations in southern Brazil: Arroio dos Ratos and Rincão dos Cabrais (Figure 2; see Ahrens et al. [53]). Finally, the association of Wolbachia variant wSinvictaA with three highly divergent clades of mtDNA haplotypes (clades II, III, and V) is consistent with three separate, rather ancient invasions of this Wolbachia strain into S. invicta.
Table 1 Prevalence of Wolbachia variants in eleven sampled populations of S. invicta. N represents the number of individuals of S. invicta surveyed for Wolbachia. The total number of infected individuals is represented by ninf whereas nwsp and n16S represent the number of individuals for which the wsp and 16S genes, respectively, were sequenced. The data in column "wsp Strains" indicate the Wolbachia variants present in each population (based on wsp sequences) as well as the number of individuals infected with each variant (in parentheses). The number of individuals from each population where mtDNA sequence data were available is also indicated.
City Country N ninf nwsp wsp Strains n16S mtDNA
Corrientes Argentina 79 53 53 wSinvictaA (22), wSinvictaB (31) 6 54
Formosa Argentina 68 3 3 wSinvictaA (2) 1 38
Roldán Argentina 14 13 13 wSinvictaA (1), wSinvictaB (12) 5 14
Rosario Argentina 30 25 25 wSinvictaA(4), wSinvictaB (21) 2 29
Arroio dos Ratos Brazil 34 20 20 wSrichteriA (20) 6 33
Rincão dos Cabrais Brazil 35 5 5 wSrichteriA (1), wSrichteriA+wSinvictaB (4) 1 10
Campo Grande Brazil 43 7 7 wSinvictaA (7) 2 29
Ceu Azul Brazil 80 11 11 wSinvictaA (11) 3 66
Pontes E Lacerda Brazil 30 0 0 - 0 28
Pedra Preta Brazil 63 1 1 wSinvictaB (1) 1 48
São Gabriel do Oeste Brazil 79 0 0 - 0 51
Totals: 555 138 138 27 400
Figure 1 Bayesian phylogenetic tree (A) and minimum spanning network (B) of mtDNA haplotypes from S. invicta. Both the Bayesian phylogenetic tree and minimum spanning network of mtDNA haplotypes from S. invicta reprinted from Ahrens et al. [53]. Haplotypes associated with the three Wolbachia variants in S. invicta are indicated by coloured bars/circles. For each mtDNA haplotype, the coloured areas of bars/circles are proportional to the number of Wolbachia-infected individuals, also indicated by the values in parentheses. The five haplotype clades in the Bayesian tree harbouring Wolbachia infected individuals are linked to their corresponding haplotype clusters by Roman numerals I-V. Purported invasion/horizontal transmission events of Wolbachia into S. invicta under scenarios 1 and 2 are indicated by the grey and black coloured bars, respectively, on the Bayesian tree. Also indicated is the evolutionary transition of variant wSinvictaA to variant wSrichteriA (black box to blue box). See text for more details.
Table 2 Distribution of different mtDNA haplotypes within the eleven sampled populations of S. invicta. h represents the number of different mtDNA haplotypes occurring in each population (see Table 1 for total number of mtDNA sequences generated from individuals of each population). Haplotypes occurring in more than one population are underlined, and haplotypes found in Wolbachia-infected individuals are in bold italics.
City Country h Haplotypes Occurring in Populations
Corrientes Argentina 20 1, 2, 3, 4, 5, 6, 7, 8, 9, 10, 11, 12, 13, 14, 15, 16, 17, 35, 54, 55
Formosa Argentina 18 18, 19, 20, 21, 22, 23, 24, 25, 26, 27, 28, 29, 30, 31, 32, 33, 34, 36
Roldán Argentina 5 1, 7, 75, 76, 77
Rosario Argentina 6 1, 5, 7, 78, 79, 80
Arroio dos Ratos Brazil 7 69, 70, 71, 72, 73, 74, 81
Rincão dos Cabrais Brazil 5 51, 52, 53, 67, 68
Campo Grande Brazil 5 46, 47, 48, 49, 50
Ceu Azul Brazil 11 37, 38, 39, 40, 41, 42, 43, 56, 57, 58, 59
Pontes E Lacerda Brazil 6 45, 62, 63, 64, 65, 66
Pedra Preta Brazil 3 44, 60, 61
São Gabriel do Oeste Brazil 2 48, 49
Figure 2 Distribution and prevalence of Wolbachia variants in eleven sampled populations of S. invicta. Each pie diagram shows the proportions of Wolbachia-infected (separately for each variant) and uninfected individuals in each geographic population (sample sizes in parentheses). The native range of S. invicta as currently understood is indicated by green shading and is based on Buren et al. [63], Trager [57], and Pitts [64].
An alternative scenario, however, is that there have been only three independent invasions of Wolbachia into S. invicta: Two of these invasions involve wSinvictaB (as described above) whereas the third invasion involves Wolbachia variant wSinvictaA (scenario 2 of Figure 1). Under this scenario such a single invasion of wSinvictaA into S. invicta presumably would have to be quite ancient, since it requires that the infection would have had to be present in the common ancestor of clades II-V (see Figure 1). Assuming a divergence rate of 2% per million years [54], our estimate of the net average nucleotide divergence among all mtDNA haplotypes comprising clades II-V (2.4%; Ahrens et al. [53]) would suggest this invasion of wSinvictaA (or most recent Wolbachia sweep) occurred roughly 1.2 mya. An additional caveat of this scenario is that wSrichteriA is not a novel, independently acquired Wolbachia infection but instead represents a derived variant of wSinvictaA (see Figure 1).
Although these rather restrictive conditions might lead one to conclude that a single invasion of wSinvictaA into S. invicta seems unlikely, this is not necessarily the case. First, whilst it is tempting to interpret the high levels of divergence among mtDNA haplotypes as indicating an ancient invasion of Wolbachia, one must be cautious when using estimates of mtDNA sequence divergence for inferring evolutionary rates simply because such high divergence may be the result a Wolbachia-driven increase in mtDNA substitution rates [39]. If true, then we may have substantially overestimated the time of invasion of wSinvictaA into S. invicta. A necessary requirement, however, is that individuals comprising the separate clades correspond to different lineages or populations that themselves are connected by little or no migration, since significant gene exchange would erase the signature of high divergence among clades. Otherwise, the most plausible explanation for the association of wSinvictaA with these divergent clades is that multiple independent invasions have occurred into S. invicta (i.e., scenario 1 above). For most populations currently infected with wSinvictaA, the condition of substantially reduced gene flow among populations holds: Ahrens et al. [53] found that genetic divergence among populations is very high and that mtDNA genetic variation is correlated with geography such that 76 out of 81 mtDNA haplotypes identified in S. invicta were exclusive to single populations. Thus, the likelihood of a single invasion seems much more reasonable when we consider not only the possible effects of Wolbachia within populations but also how Wolbachia infections can accelerate divergence among populations (divergence among mtDNA lineages), especially those connected by very limited gene flow. Finally, we should also point out that the additional above requirement that wSrichteriA is a derived variant of wSinvictaA also is quite reasonable given that these two strains differ by only a single nucleotide substitution at the highly evolving wsp gene.
Regardless of the presumed number of invasions of Wolbachia into S. invicta (three, six, or perhaps more), it is clear that the secondary loss of Wolbachia infections from host lineages following invasion is very common. Such frequent loss of infections is most obvious when one considers the fact that uninfected individuals harbour both derived mtDNA haplotypes and ancestral haplotypes inferred to be associated with the original infection (Figure 1). Previously, Shoemaker et al. [48] estimated that the fidelity of maternal transmission of Wolbachia in S. invicta in nature generally is very high (>99%), but nonetheless is not perfect, ranging from 90–100% within different matrilines.
Thus, our data partly resolve the paradox of a lack of a consistent correlation between the presence of Wolbachia and either levels or patterns of mtDNA diversity in S. invicta. Clearly, the previous assertion of strictly vertical transmission of Wolbachia in fire ants breaks down upon finer-scale analysis. Multiple independent invasions of Wolbachia into S. invicta have occurred, and in every case these have been followed by frequent secondary loss of infections. Thus, although we predicted a strong association between the mtDNA and Wolbachia genomes since both are co-transmitted from mother to offspring, the strong association of the two genomes in fire ants clearly has broken down over time due to frequent horizontal transmission and secondary loss of Wolbachia strains [26,36,44-46,55].
Wolbachia distribution and prevalence in S. invicta
Variation in the distribution and prevalence of Wolbachia in natural populations of S. invicta may be due to: 1) presence of different Wolbachia variants within and among populations 2) genetic differences among host individuals from different populations or 3) genetic drift [48]. To attempt to address this issue, we examined the Wolbachia strain identities and their corresponding frequencies within each of the eleven sampled populations of S. invicta. If variation in Wolbachia prevalence is due simply to differences in the particular Wolbachia variants or combinations of variants within and among these populations, then we might expect that despite differences in overall Wolbachia prevalence among populations the prevalence of any particular Wolbachia variant is similar in each of the host populations where it occurs. Thus, a simple explanation for the observed variation in prevalence may be that the array of Wolbachia variants differs among host populations. On the other hand, if this variation results from genetic differences among host individuals from different populations, then one might expect that the prevalence of specific Wolbachia variants varies among different host populations, and possibly that the variants are associated with quite different mtDNA haplotypes in each population. Although the effects of Wolbachia on S. invicta are currently unknown, we would expect infection prevalence to vary stochastically if Wolbachia do not have any measurable fitness or sex ratio effects on their fire ant hosts.
The distribution and prevalence of the three Wolbachia variants within the eleven sampled populations of S. invicta is shown in Figure 2. The wSinvictaA variant occurs at similar prevalence (11.4–27.8% of individuals) in four of the five populations where it is found, possibly indicating this low prevalence represents the stable equilibrium frequency of this variant. The similar prevalence of wSinvictaA in different populations that are both genetically differentiated and separated by great geographical and ecological differences [53] suggests that the dynamics and prevalence of this variant are most likely not strongly affected by its host or environment. The wSinvictaB variant is largely confined to individuals collected from the southwestern populations of Corrientes and Roldán/Rosario. This variant occurs at relatively high prevalence in these populations (39.2–75.0%). Finally, the wSrichteriA variant has a very restricted distribution and is found only in individuals from the Arroio dos Ratos and Rincão dos Cabrais populations in the southernmost portion of Brazil. Our survey data revealed that 36% of all colonies surveyed from these populations harbour this Wolbachia variant.
Together, these data indicate that the variation in Wolbachia prevalence among populations can be explained largely by differences in the array of Wolbachia variants within host populations. Even so, we cannot discount completely a role for host effects in determining Wolbachia prevalence given the very high levels of genetic differentiation among populations [46,53,56]. Additionally, while our data do not imply an obvious role for environmental conditions affecting Wolbachia dynamics, it is interesting to note the apparent positive correlation between Wolbachia prevalence and latitude. An analogous pattern previously has been reported for a Wolbachia variant infecting the beetle Chelymorpha alternans. In this host species, Wolbachia prevalence apparently is lower in areas experiencing longer dry seasons and higher average daily temperatures [24]. Thus, although unlikely, it remains possible that the overall Wolbachia infection dynamics in S. invicta are influenced by differences in environmental conditions as well, with higher Wolbachia prevalence occurring in the more southerly temperate populations.
Finally, our results combined with mtDNA data from earlier studies argue against the previous hypothesis that variation in Wolbachia prevalence is simply due to the recent invasion and ongoing spread of Wolbachia in S. invicta. First, a substantial number of polymorphic sites were found in the mtDNA sequences comprising each of five clades (I-V), indicating the Wolbachia infections are sufficiently ancient enough that numerous mtDNA mutations have accumulated since the most recent invasion(s) of Wolbachia. Assuming a divergence rate of 2% per million years [54], estimates of the average sequence divergence among mtDNA haplotypes within clades I-V (0.1–1.2%; Ahrens et al. [53]) would suggest the most recent invasion of Wolbachia (or most recent Wolbachia sweep) within any of these clades roughly occurred at least 50,000 years BP (Although Wolbachia endosymbionts may accelerate divergence between lineages or populations, recurrent Wolbachia sweeps have the opposite effect on differentiation within populations and result in substantially reduced mtDNA variation within populations [30,39]). The finding that the composition and diversity of mtDNA haplotypes found in infected and uninfected individuals within populations are virtually identical clearly suggests that uninfected individuals are derived from infected lineages via incomplete maternal transmission of Wolbachia and lends further support to the hypothesis that Wolbachia infections in S. invicta are evolutionarily old. An alternative possibility, which we consider less likely, is that there has been rampant horizontal transmission of the same Wolbachia variants within and among S. invicta populations.
Conclusion
The evolutionary history of Wolbachia in S. invicta is far more complex than previously recognized: at least three and possibly as many as six horizontal transmission events involving three different variants have occurred into S. invicta. Further, in every case these independent acquisitions of Wolbachia have been followed by multiple independent losses of Wolbachia infections over time. Indeed, we should note that if loss of Wolbachia infection occurs as commonly as our data suggest, then we likely have underestimated the number of invasions or horizontal transmissions of Wolbachia. These extensive sequence data also suggest that the significant variation in Wolbachia prevalence among fire ant populations most likely is due simply to the presence of different variants limited to specific regions of S. invicta's range, but roles for both host effects and the environment in accounting for the observed patterns cannot be excluded. Our results also partly explain the previous puzzling findings of no clear effects of Wolbachia infection on patterns of mtDNA variation and substitution in fire ants [46]. Wolbachia transmission over evolutionary time appears to be uncoupled from that of the mtDNA genome such that the predicted effect of Wolbachia in reducing host mtDNA variation is not clearly evident as originally predicted. Thus, our previous concern that recurrent Wolbachia sweeps within fire ant populations may confound the use of mtDNA markers for studying the evolutionary history of fire ants (i.e. phylogeographic studies, identification of source populations), as the invasion of new strains would erase all pre-existing variation, seems somewhat unwarranted.
On the other hand, the high levels of divergence among mtDNA haplotype clades (~3.2% [53]) are analogous to patterns reported for the two Wolbachia-infected insect species, Drosophila recens and D. simulans, and may be the footprint of another predicted effect of Wolbachia infections, namely, an accelerated mtDNA substitution rate as a result of recurrent Wolbachia sweeps (see Shoemaker et al. [39] for full discussion). For example, Shoemaker et al. [39] observed an mtDNA-specific accelerated rate of evolution in D. recens, a species in which virtually all individuals are infected by a single Wolbachia strain, relative to the closely related uninfected species D. subquinaria. In D. simulans, previous studies have revealed that despite very little sequence variation within each of the three defined mtDNA haplotype clades, substantial differentiation exists among these clades [26,30-37,44]. Although no formal comparative analyses have been conducted in either S. invicta or D. simulans to test the above hypothesis, one possible explanation for the high level of divergence among these well-defined mtDNA haplotype clades in both species is that it results from Wolbachia-driven acceleration in the mtDNA substitution rate [39]. Together, these three studies lend support to the hypothesis that maternally-inherited endosymbiont infections may increase the rate of substitution in mtDNA [39]. Clearly, additional comparative studies in other insects are needed to test the generality of this hypothesis, especially since such effects have important consequences for the assumptions of neutrality and use of mtDNA as a molecular clock in insects.
Methods
Collection and identification of ants
Individuals of S. invicta were collected from native populations in Argentina and Brazil in 1992 and 1998 (Table 1). Multiple workers and winged virgin queens were collected from each of 555 colonies representing eleven geographic populations distributed over much of the known native range of S. invicta (see Figure 1 of Ahrens et al. [53] for locations). All collected individuals were identified as S. invicta by J. P. Pitts using species-informative morphological characters [57,58].
Sequencing of Wolbachia strains
DNA was extracted from a single individual from each of the 555 colonies using the Puregene DNA isolation kit (Gentra Systems) [38,59]. We previously screened all 555 DNA extracts for the presence of Wolbachia by means of PCR using the primers wsp81F and wsp691R [48,59,60]. These wsp primers amplify a portion of a highly-variable gene encoding the Wolbachia outer surface protein [59,60]. Our previous survey of S. invicta revealed that 138 of the 555 individuals (colonies) were Wolbachia-infected (see Table 1). For the present study, we sequenced a portion of the wsp gene from all 138 infected individuals using the above primers. Wolbachia DNA was PCR-amplified in 30-μL volumes, with the PCR reaction components and thermal cycling conditions identical to those described in Shoemaker et al. [38]. Wsp PCR amplicons were purified for sequencing using Ampure magnetic beads (Agencourt Bioscience Corp.) and subsequently used directly in standard fluorescent cycle-sequencing PCR reactions (ABI Prism Big Dye terminator chemistry, Applied Biosystems). Sequencing reactions were purified using CleanSEQ magnetic beads (Agencourt Bioscience Corp.) and run on an ABI 3700 sequencer at the UW Biotechnology Center DNA Sequencing Laboratory.
Initial sequencing results of the wsp gene revealed the presence of more than one Wolbachia strain in three individuals of S. invicta (i.e., multiple peaks or frameshifts in electropherogram profiles were observed). For these three individuals, Wolbachia DNA was PCR-amplified as described above, except the final extension at 72°C was increased to 30 minutes. PCR amplicons were cloned directly into a vector following manufacturer's suggestions (Topo TA cloning kit, Invitrogen corp.) and resulting colonies screened for the presence of the desired wsp PCR insert using the wsp primers. For each individual, PCR-amplified products from ten colonies (which presumably had the wsp insert) were purified and sequenced as described above.
We also PCR-amplified and sequenced a 945 base portion of the Wolbachia 16S gene using primers specific to this region [3] from a subset of the infected individuals within each population in an attempt to further characterize and identify unique Wolbachia strains (27 sequences total). PCR reaction components and thermal cycling conditions were identical to those described in O'Neill et al. [3]. Purification and sequencing of 16S amplicons, as well as cloning and sequencing of individuals possessing more than one Wolbachia strain, were carried out as described for the wsp gene above.
Comparing Wolbachia (wsp) and mtDNA variation
Both the phylogeny and minimum spanning network of 81 unique mtDNA haplotypes representing 400 individuals (colonies) from the eleven populations used in the present study were generated previously by Ahrens et al. [53] using MrBayes 3.0 [61] and ARLEQUIN ver. 2.000 [62], respectively. Both methods of analysis identified six well-supported clades (clusters) of closely related mtDNA haplotypes, with each clade separated from the others by at least 18 mutational steps. With few exceptions, each clade is comprised of mtDNA haplotypes present in individuals from only one or two geographically proximal populations of S. invicta [for a more detailed description, see [53]]. For the present study, we used our wsp gene sequence data to determine the infection status, infection frequency, and strain identity for individuals of each mtDNA haplotype within the pre-existing networks.
Authors' contributions
MEA carried out the majority of the molecular work and performed phylogenetic data analyses. DDS designed and coordinated the study, collected all of the ants used for the study, carried out a portion of the molecular work, and performed most of the data analyses. Both authors contributed to writing the manuscript and approved the final manuscript.
Acknowledgements
We wish to thank Laurent Keller, Mark Mescher, and Ken Ross for their invaluable assistance in the collecting of ants used for the present study. We also thank Ken Ross and three anonymous reviewers for their comments on an earlier version of the manuscript. This study was supported by grants from the College of Agriculture and Life Sciences at the University of Wisconsin, the United States Department of Agriculture NRICGP, and the U.S. National Science Foundation to DDS.
==== Refs
Stouthamer R Breeuwer JAJ Hurst GDD Wolbachia pipientis: Microbial manipulator of arthropod reproduction Annual Review of Microbiology 1999 53 71 102 10547686 10.1146/annurev.micro.53.1.71
Werren JH O'Neill SL O'Neill SL, Hoffmann AA, Werren JH The evolution of heritable symbionts Influential Passengers: Inherited Microorganisms and Arthropod Reproduction 1997 New York , Oxford University Press 1 41
O'Neill SL Giordano R Colbert AME Karr TL Robertson HM 16S rRNA phylogenetic analysis of the bacterial endosymbionts associated with cytoplasmic incompatibility in insects Proceedings of the National Academy of Sciences, USA 1992 89 2699 2702
Bandi C Anderson T Genchi C Blaxter M Phylogeny of Wolbachia in filarial nematodes Proceedings of The Royal Society of London: Series B, Biological Sciences 1998 265 2407 2413 10.1098/rspb.1998.0591
Werren JH Guo L Windsor DW Distribution of Wolbachia in Neotropical arthropods Proceedings of the Royal Society of London: Series B, Biological Sciences 1995 262 197 204
Werren JH Windsor DM Wolbachia infection frequencies in insects: Evidence of a global equilibrium? Proceedings of the Royal Society of London: Series B, Biological Sciences 2000 267 1277 1285 10.1098/rspb.2000.1139
West SA Cook J Werren JH Godfrey HCJ Wolbachia in two insect host-parasitoid communities mol ecol 1998 7 1457 1465 9819901 10.1046/j.1365-294x.1998.00467.x
Jeyaprakash A Hoy MA Long PCR improves Wolbachia DNA amplification: wsp sequences found in 76% of sixty-three arthropods Insect Molecular Biology 2000 9 393 405 10971717 10.1046/j.1365-2583.2000.00203.x
Werren JH Biology of Wolbachia Annual Review of Entomology 1997 42 537 609 10.1146/annurev.ento.42.1.587
Stevens L Giordano R Fialho RF Male-killing, nematode infections, bacteriophage infection, and virulence of cytoplasmic bacteria in the genus Wolbachia Annu Rev Ecol Syst 2001 32 519 545 10.1146/annurev.ecolsys.32.081501.114132
Huigens ME deAlmeida RP Boons PAH Luck RF Stouthamer R Natural interspecific and intraspecific horizontal transfer of parthenogensis-inducing Wolbachia in Trichogramma wasps Proceedings of the Royal Society of London: Series B, Biologial Sciences 2004 271 509 515 10.1098/rspb.2003.2640
Vavre F Fleury F Lepetit D Fouillet P Bouletreau Phylogenetic evidence for horizontal transmission of Wolbachia in host-parasitoid associations Molecular Biology and Evolution 1999 16 1711 1723 10605113
Heath BD Butcher RDJ Whitfield WGF Hubbard SF Horizontal transfer of Wolbachia between phylogenetically distant insect species by a naturally occurring mechanism Current Biology 1999 9 313 316 10209097 10.1016/S0960-9822(99)80139-0
van Meer MMM Witteveldt J Stouthamer R Phylogeny of the arthropod endosymbiont Wolbachia based on the wsp gene Insect Molecular Biology 1999 8 399 408 10469257 10.1046/j.1365-2583.1999.83129.x
Huigens ME Luck RF Klaassen RHG Maas MFPM Timmermans MJTN Stouthamer R Infectious parthenogenesis Nature 2000 405 178 179 10821272 10.1038/35012066
Werren JH Zhang W Guo LR Evolution and phylogeny of Wolbachia: Reproductive parasites of arthropods Proceedings of the Royal Society of London: Series B, Biological Sciences 1995 261 55 71
Noda H Miyoshi T Zhang Q Watanabe K Deng K Hoshizaki S Wolbachia infection shared among planthoppers (Homoptera: Delphacidae) and their endoparasite (Strepsiptera: Elenchidae): a probable case of interspecies transmission Mol Ecol 2001 10 2101 2106 11555254 10.1046/j.0962-1083.2001.01334.x
Schilthuizen M Stouthamer R Horizontal transmission of parthenogenesis-inducing microbes in Trichogramma wasps Proceedings of the Royal society of London: Series B, Biological Sciences 1997 264 361 366 10.1098/rspb.1997.0052
Cordaux R Michel-Salzat A Frelon-Raimond M Rigaud T Bouchon D Evidence for a new feminizing Wolbachia strain in the isopod Armadillidium vulgare: Evolutionary implications Heredity 2004 93 78 84 15138452 10.1038/sj.hdy.6800482
Rigaud T Bouchon D Souty-Grosset C Raimond R Mitochondrial DNA polymorphism, sex ratio distorters, and population genetics in the isopod Armadillidium vulgare Genetics 1999 152 1669 1677 10430591
Michel-Salzat A Cordaux R Bouchon D Wolbachia diversity in the Porcellionides pruinosus complex of species (Crustacea : Oniscidea): evidence for host-dependent patterns of infection Heredity 2001 87 428 434 11737290 10.1046/j.1365-2540.2001.00920.x
Guillemaud T Pasteur N Rousset F Contrasting levels of variability between cytoplasmic genomes and incompatibility types in the mosquito Culex pipiens Proceedings of The Royal Society of London: Series B, Biological Sciences 1997 264 245 251 10.1098/rspb.1997.0035
Rousset F Solignac M Evolution of single and double Wolbachia symbioses during speciation in the Drosophila simulans complex Proceedings of the National Academy of Sciences, USA 1995 92 6389 6393
Keller GP Windsor DM Saucedo JM Werren JH Reproductive effects and geographical distributions of two Wolbachia strains infecting the Neotropical beetle, Chelymorpha alternans Boh. (Chrysomelidae, Cassidinae) Molecular Ecology 2004 13 2405 2420 15245413 10.1111/j.1365-294X.2004.02213.x
Behura SK Sahu SC Mohan M Nair S Wolbachia in the Asian rice gall midge, Orseolia oryzae (Wood- Mason): correlation between host mitotypes and infection status Insect Mol Biol 2001 10 163 171 11422512 10.1046/j.1365-2583.2001.00251.x
Turelli M Hoffmann AA McKechnie SW Dynamics of cytoplasmic incompatibility and mtDNA variation in natural Drosophila simulans populations Genetics 1992 132 713 723 1468627
Marcade I Souty-Grosset C Bouchon D Rigaud T Raimond R Mitochondrial DNA variability and Wolbachia infection in two sibling woodlice species. Heredity 1999 83 71 78 10447705 10.1038/sj.hdy.6885380
Rokas A Atkinson RJ Brown GS West SA Stone GN Understanding patterns of genetic diversity in the oak gallwasp Biorhiza pallida: demographic history or a Wolbachia selective sweep? Heredity 2001 87 294 304 11737276 10.1046/j.1365-2540.2001.00872.x
Jiggins FM Male-killing Wolbachia and mitochondrial DNA: Selective sweeps, hybrid introgression and parasite population dynamics Genetics 2003 164 5 12 12750316
Turelli M Hoffmann AA Rapid spread of an inherited incompatibility factor in California Drosophila Nature 1991 353 440 442 1896086 10.1038/353440a0
Ballard JWO Comparative genomics of mitochondrial DNA in Drosophila simulans Journal of Molecular Evolution 2000 51 64 75 10903373
Ballard JWO Sequential evolution of a symbiont inferred from the host: Wolbachia and Drosophila simulans Molecular Biology and Evolution 2004 21 428 442 14660690 10.1093/molbev/msh028
Ballard JWO Chernoff B James AC Divergence of mitochondrial DNA is not corroborated by nuclear DNA, morphology, or behavior in Drosophila simulans Evolution 2002 56 527 545 11989683
Ballard JWO Hatzidakis OJ Karr TL Kreitman M Reduced variation in Drosophila simulans mitochondrial DNA Genetics 1996 144 1519 1528 8978041
Dean MD Ballard KJ Glass A Ballard JWO Influence of two Wolbachia strains on population structure of East African Drosophila simulans Genetics 2003 165 1959 1969 14704179
James AC Dean MD McMahon ME Ballard JWO Dynamics of double and single Wolbachia infections in Drosophila simulans from New Caledonia Heredity 2002 88 182 189 11920119 10.1038/sj.hdy.6800025
James AC Ballard JWO Expression of cytoplasmic incompatibility in Drosophila simulans and its impact on infection frequencies and distribution of Wolbachia pipientis Evolution 2000 54 1661 1672 11108593
Shoemaker DD Ross KG Keller L Vargo EL Werren JH Wolbachia infections in native and introduced populations of fire ants (Solenopsis spp.) Insect Molecular Biology 2000 9 661 673 11122476 10.1046/j.1365-2583.2000.00233.x
Shoemaker DD Dyer KA Ahrens M McAbee K Jaenike J Decreased diversity but increased substitution rate in host mtDNA as a consequence of Wolbachia endosymbiont infection Genetics 2004 168 2049 2058 15611174 10.1534/genetics.104.030890
Charlat S Nirgianaki A Bourtzis K Mercot H Evolution of Wolbachia-induced cytoplasmic incompatibility in Drosophila simulans and D. Sechellia Evolution 2002 56 1735 1742 12389718
Caspari E Watson GS On the evolutionary importance of cytoplasmic sterility in mosquitoes Evolution 1959 13 568 570
Fine PEM On the dynamics of symbiote-dependent cytoplasmic incompatibility in Culicine mosquitoes Journal of Invertebrate Pathology 1978 30 10 18 415090 10.1016/0022-2011(78)90102-7
Prout T Some evolutionary possibilities for a microbe that causes incompatibility in its host Evolution 1994 48 909 911
Turelli M Hoffmann AA Cytoplasmic incompatibility in Drosophila simulans: Dynamics and parameter estimates from natural populations Genetics 1995 140 1319 1338 7498773
Turelli M Evolution of incompatibility-inducing microbes and their hosts Evolution 1994 48 1500 1513
Shoemaker DD Keller G Ross KG Effects of Wolbachia on mtDNA variation in two fire ant species Mol Ecol 2003 12 1757 1771 12803629 10.1046/j.1365-294X.2003.01864.x
Charlat S Bonnavion P Mercot H Wolbachia segregation dynamics and levels of cytoplasmic incompatibility in Drosophila sechellia Heredity 2003 90 157 161 12634822 10.1038/sj.hdy.6800211
Shoemaker DD Ahrens M Sheill L Mescher M Keller L Ross KG Distribution and prevalence of Wolbachia infections in native populations of the fire ant Solenopsis invicta (Hymenoptera : Formicidae) Environmental Entomology 2003 32 1329 1336
Tsutsui ND Kauppinen SN Oyafuso AF Grosberg RK The distribution and evolutionary history of Wolbachia infection in native and introduced populations of the invasive argentine ant (Linepithema humile) Mol Ecol 2003 12 3057 3068 14629385 10.1046/j.1365-294X.2003.01979.x
Van Borm S Wenseleers T Billen J Boomsma JJ Cloning and sequencing of wsp encoding gene fragments reveals a diversity of co-infecting Wolbachia strains in Acromyrmex leafcutter ants. Molecular Phylogenetics and Evolution 2003 26 102 109 12470942 10.1016/S1055-7903(02)00298-1
Jiggins FM The rate of recombination in Wolbachia bacteria Molecular Biology and Evolution 2002 19 1640 1643 12200493
Werren JH Bartos J Recombination in Wolbachia Current Biology 2001 11 431 435 11301253 10.1016/S0960-9822(01)00101-4
Ahrens M Ross KG Shoemaker DD Phylogeographic structure of the fire ant Solenopsis invicta in its native South American range: Roles of natural barriers and habitat connectivity Evolution 2005 in press
Brower AVZ Rapid morphological radiation and convergence among races of the butterfly Heliconius errato inferred from patterns of mitochondrial DNA evolution Proceedings of the National Academy of Sciences, USA 1994 91 6491 6495
Shoemaker DD Katju V Jaenike J Wolbachia and the evolution of reproductive isolation between Drosophila recens and Drosophila subquinaria Evolution 1999 53 1157 1164
Ross KG Shoemaker DD Species delimitation in native South American fire ants Molecular Ecology 2005 in press
Trager JC A revision of the fire ants, Solenopsis geminata group (Hymenoptera: Formicidae: Myrmicinae) Journal of the New York Entomological Society 1990 99 141 198
Pitts JP McHugh JV Ross KG Revision of the fire ants of the Solenopsis saevissima species-group (Hymenoptera: Formicidae) Zootaxa 2005 in press
Zhou W Rousset F O'Neill S Phylogeny and PCR-based classification of Wolbachia strains using wsp gene sequences Proceedings of The Royal Society of London: Series B, Biological Sciences 1998 265 509 515 10.1098/rspb.1998.0324
Braig HR Zhou W Dobson S O'Neill SL Cloning and characterization of a gene encoding the major surface protein of the bacterial endosymbiont Wolbachia Journal of Bacteriology 1998 180 2373 2378 9573188
Huelsenbeck JP Ronquist F Nielsen R Bollback JP Evolution - Bayesian inference of phylogeny and its impact on evolutionary biology Science 2001 294 2310 2314 11743192 10.1126/science.1065889
Schneider S Roessli D Excoffier L Arlequin ver. 2.000: A software for population genetic data analysis 2000 University of Geneva, Switzerland , Genetics Biometry Laboratory
Buren WF Allen GE Whitcomb WH Lennartz FE Williams RN Zoogeography of the imported fire ants Journal of the New York Entomological Society 1974 82 113 124
Pitts JP A cladistic analysis of the Solenopsis saevissima species-group (Hymenoptera: Formicidae). Entomology 2002 Athens, GA , University of Georgia
| 15927071 | PMC1175846 | CC BY | 2021-01-04 16:37:17 | no | BMC Evol Biol. 2005 May 31; 5:35 | utf-8 | BMC Evol Biol | 2,005 | 10.1186/1471-2148-5-35 | oa_comm |
==== Front
BMC GastroenterolBMC Gastroenterology1471-230XBioMed Central London 1471-230X-5-171592979610.1186/1471-230X-5-17Research ArticleHepatocellular carcinoma and the penetrance of HFE C282Y mutations: a cross sectional study Willis Gavin [email protected] Vicky [email protected] Ian W [email protected] Ray [email protected] Jennie Z [email protected] Barbara A [email protected] Department of Molecular Genetics, Norfolk and Norwich University Hospital, Norwich, NR47UY, UK2 Department of Histopathology, Norfolk and Norwich University Hospital, Norwich, NR47UY, UK3 Department of Gastroenterology, Norfolk and Norwich University Hospital, Norwich, NR47UY, UK4 Department of Haematology, Norfolk and Norwich University Hospital, Norwich, NR47UY, UK5 School of Medicine, Health Policy and Practice, University of East Anglia, Norwich, NR47PT, UK2005 1 6 2005 5 17 17 7 2 2005 1 6 2005 Copyright © 2005 Willis 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 most patients with hereditary haemochromatosis have HFE C282Y mutations, the lifetime risk to HFE C282Y homozygotes of developing fatal diseases such as hepatocellular carcinoma is uncertain. We have carried out a cross-sectional study to determine the proportion of diagnosed hepatocellular carcinoma patients who are homozygous for the HFE C282Y mutation; and to estimate the penetrance of this genotype with respect to hepatocellular carcinoma in East Anglia.
Methods
Tissue biopsies were analysed from 144 cases of hepatocellular carcinoma for HFE C282Y mutations; the data produced were compared with the frequency of HFE mutations in a large sample of the local population. Data were also retrieved from the East Anglian Cancer Intelligence Unit to determine the annual incidence of hepatocellular carcinoma; and from appropriate life tables.
Results
Eight out of 144 of the cases were homozygous for the HFE C282Y mutation, all 8 cases were male. 6 of these 8 cases had a previous diagnosis of hereditary haemochromatosis. Male HFE C282Y homozygotes were more likely to be diagnosed with hepatocellular carcinoma (odds ratio [OR] = 14, 95% confidence interval [CI] = 5–37). For this population, we estimate that the penetrance of the HFE C282Y homozygous genotype, with respect to hepatocellular carcinoma, was between 1.31 % and 2.1% for males and was zero for females.
Conclusion
In this population, we found that only a very small proportion of homozygotes for the HFE C282Y mutation developed hepatocellular carcinoma. However, individuals with this genotype have a significantly increased risk of this rare disease relative to those who do not carry the mutations.
==== Body
Background
Hereditary haemochromatosis is an autosomal recessive genetic condition in which excess iron is absorbed by the intestine. Individuals with the clinical manifestations of the disease (which include liver cirrhosis, hepatocellular carcinoma, diabetes mellitus, cardiomyopathy and arthropathy) will have accumulated iron over many years of adult life resulting in progressive tissue damage. Liver disease is the commonest cause of death of patients with hereditary haemochromatosis [1,2]. A recent cohort study of patients diagnosed with haemochromatosis in Sweden found that at ten years follow-up, the absolute risk of liver cancer was 6% among men and 1.5% among women [3]. This patient cohort is likely to be at higher risk of liver cancer than those HFE C282Y homozygotes who do not display the signs and symptoms of haemochromatosis [3].
Haemochromatosis is unusual among genetic diseases because it can easily be treated. Individuals diagnosed and treated by regular venesection before symptoms of cirrhosis occur have a normal life expectancy [4]. The discovery of mutations in the HFE gene that are present in most haemochromatosis patients has provided a useful test in families affected by the disease [5]. Two HFE genotypes are commonly associated with haemochromatosis: homozygosity for the C282Y (845A) mutation and compound heterozygosity with the C282Y and H63D (187G) mutation [5-7].
The high frequency of HFE mutations in the normal population and the availability of a treatment for haemochromatosis led to suggestions that population screening for these mutations would be justified on the basis of both health and cost [8,9]. The value of screening depends on the penetrance of the HFE genotypes. Penetrance in this case can be defined as the frequency with which individuals of particular genotypes develop either iron overload or clinical manifestations.
We have previously studied the penetrance of the HFE mutations with respect to haemochromatosis disease manifestations by comparing the predicted birth rate of HFE C282Y homozygotes in our study population of 500,000 with the incidence of HFE C282Y-homozygous patients diagnosed with haemochromatosis, cirrhosis [10], liver cancer [11], arthritis [12] or diabetes [13]. We showed that, in this large population, few HFE C282Y homozygotes (1.4%) were diagnosed with haemochromatosis and of the remainder few were diagnosed with liver disease (2.7% – 8%) or diabetes (0 – 1.3%).
Our findings suggested an unexpectedly low figure for penetrance which made our estimates for the potential benefit of genetic screening marginal.
Because of the need for more data to inform decisions about the value of population screening we have now collected a much larger cohort of hepatocellular carcinoma (HCC) cases (n = 144) for HFE genotyping. The samples were drawn from histology archives, collected over a thirty year period, from the two largest hospitals in East Anglia. We have used genotyping data from this cross-section of 144 cases and HCC incidence data from the cancer registry to deduce the incidence of HFE C282Y homozygotes diagnosed with HCC annually. We have compared this with the proportion of the general population who are HFE C282Y homozygotes and who reach the at-risk age for HCC annually; this corresponds to the penetrance of HFE C282Y homozygosity with respect to developing HCC.
Methods
Patient samples; hepatocellular carcinoma cases
The Norfolk and Norwich University Hospital has a catchment area of 500,000 people. 41 suitable HCC cases were identified from histopathology records from 1974 to 2001 regardless of any previous diagnosis of haemochromatosis. 29 cases were from men and 12 were from women. 28 of these 41 cases from Norwich were included in our previous studies [10,11].
Addenbrooke's Hospital, Cambridge, is a tertiary referral centre for hepatic transplantation and referral. Therefore, cases of liver cancer derive from the hospital's catchment area of approximately 350,000 individuals and from other hospitals in East Anglia, and less commonly from the rest of the U.K. and Italy. For this study, cases were rejected if the patient's name was Italian to exclude tertiary referral cases of southern European origin (because of differences in the prevalence of HFE mutations). 103 HCC cases from Addenbrooke's Hospital were identified by searching files from 1969 to 2000 for cases of HCC regardless of any previous diagnosis of haemochromatosis and included both biopsy cases and explanted livers. 73 cases were from men and 30 from women.
None of the 103 cases from Addenbrooke's Hospital had been analysed in the cohort of cases from the Norfolk and Norwich University Hospital.
Slides from all 144 cases from Norfolk and Cambridge were reviewed and the presence of HCC confirmed.
Analysis of HFE genotype
Formalin-fixed, paraffin-wax embedded specimens were retrieved from the histology archives. DNA was extracted from 10 mm2 of 10 μM tissue sections by sequential octane and acetone extractions; followed by digestion overnight at 55°C in 50 μl of 500 μg/ml proteinase K, 10 mM Tris, 1 mM EDTA and 0.5% Tween 20 and subsequent incubation for 10 minutes at 96°C.
Analysis of the HFE codon 282 genotypes was carried out on the PE Biosystems 7700 (Applied Biosystems), using standard allelic discrimination assay software and Taqman Universal PCR Mastermix (Applied Biosystems). In most cases, 1 μl of template DNA (approximately 50 ng) was used per assay, the primers and probes are described in table 1. The annealing temperature in a standard PE Biosystems 7700 protocol for the assay was 65°C. Four each of CY heterozygote and CC and YY homozygote and no DNA controls were included on each 96 well plate. The primers used amplify a 106 base pair fragment of the HFE gene and therefore the assay is suitable even for relatively degraded DNA.
Population data
The control HFE genotype data (see table 2) discussed in this paper are pooled from two recent publications looking at HFE mutations in the Norfolk and Cambridgeshire populations [12,14].
The number of C282Y homozygotes reaching the at-risk age for HCC annually was calculated using data in the 1985–1988 life tables [15].
The annual incidence of HCC in East Anglia between 1970 and 2001 was obtained from the East Anglian Intelligence Unit (cancer registry) [16].
Ethical approval
The analysis of the previously archived tumour and DNA samples was carried out with local research ethics committee (LREC) approval from Norwich (NDEC97/090) and Cambridge (LREC 00/153). All genetic analysis was carried out on anonymous samples.
Statistical analysis
95% confidence intervals (CI) are exact binomial or, for ratios, the normal approximation; p values are by Fisher's exact test.
Results
Analysis of the HFE genotypes
8/144 (5.6 %, 95% CI = 2.4–10.7%) of the cases of HCC were homozygous for the HFE C282Y mutation. 102 of the 144 cases were males and all HFE C282Y homozygous cases were male. Thus 8/102 male HCC cases were HFE C282Y homozygotes while 9/1508 of the control population had this genotype (see table 2). Male HFE C282Y homozygotes were therefore more likely to be diagnosed with HCC (OR = 14, 95% CI = 5–37).
The 8 C282Y homozygous cases are described in table 3. Five of the HFE C282Y homozygous cases were from Cambridge; all these had been diagnosed previously with haemochromatosis. The three other HFE C282Y homozygous cases were from Norwich. One of these cases had been diagnosed with haemochromatosis. The remaining two cases had not been previously diagnosed with haemochromatosis. Because Addenbrooke's Hospital is a tertiary referral centre the details of the Norwich HFE C282Y homozygotes were checked against the Cambridge samples to ensure that there was no duplication.
17/144 (11.8%) cases were heterozygous for the HFE C282Y mutation. This frequency is essentially the same as that in the normal population (see table 2).
Population data
C282Y homozygotes reaching the at-risk age for HCC annually
54 was chosen as the at-risk age for HCC because it is the mean age for presentation with HCC in our cohort. The proportion of the male population reaching the age of 54 annually (1260 per 100,000) was calculated from the birth rate (1390 per 100,000) and the proportion surviving to the age of 54 (lx = 0.912) in the 1985–1988 life tables [15]. The proportion of the population who are C282Y homozygotes reaching the age of 54 annually was calculated from the above figure multiplied by the observed proportion of C282Y homozygotes in the normal population: 7.5 per 100,000 [12,14].
The penetrance of the C282Y homozygous genotype with respect to HCC
The mean annual incidence of HCC in East Anglia between 1971 and 2001 was 1.26 per 100,000 males [16]. Our sample allows us to estimate that 7.8% (8 /102 males in the study) of these were HFE C282Y homozygotes (0.099 per 100,000). This figure contrasts with the proportion of the population who are C282Y homozygotes reaching the at-risk age for HCC annually: 7.5 per 100,000 (see above).
We thus estimate that only 1.31% (95% CI = 0.52–3.32%) of males homozygous for the HFE C282Y genotype are diagnosed and recorded with HCC in this population. This corresponds to an estimate of penetrance if most HCC cases are recorded by the cancer registry.
An alternative method to estimate the normal population frequency of C282Y homozygosity is to calculate the square of the C282Y allele frequency (0.0622 = 0.0038). This figure has a lower standard error, being derived from the larger number of heterozygotes, but does not take account of population effects such as non-random mating or mixing. Repeating the above calculation using this approach gives an estimate for penetrance of 2.1% (95% CI = 0.89–4.05%).
In this study we have failed to see any penetrance for the C282Y homozygous genotype with respect to HCC for females.
Histology
Slides for cases that were HFE C282Y homozygotes were reanalysed to examine the background liver abnormalities in more detail (see table 3). Two of the five C282Y homozygous samples from Addenbrooke's Hospital were biopsy specimens, and the remaining three were explanted livers. Two cases showed significant fibrosis with prominent linking of many portal tracts by fibrous bands, but did not show established cirrhosis. The remaining 3 cases all showed micronodular cirrhosis. These cases were staged for fibrosis using the method described by Ishak et.al. [17].
The needle biopsies of 3 C282Y homozygotes from Norwich were analysed, one biopsy consisted of tumour only, a second showed a minute area of cirrhotic liver and a third showed minimal fibrosis and was not cirrhotic; however, there was very little background liver tissue in this specimen.
All HFE C282Y homozygous cases with an adequate amount of background liver showed siderosis of grades 1 to 4, with or without a history of venesection treatment (see table 3).
Discussion
The discovery of the HFE gene in 1996, the high prevalence of C282Y mutations, and the morbidity and mortality associated with untreated hereditary haemochromatosis have presented molecular diagnostics with a potentially attractive test for population screening.
Our results show that male HFE C282Y homozygotes were more likely to be diagnosed with HCC (OR = 14, 95% CI = 5–37); the HFE C282Y homozygous genotype could therefore be a significant cause of liver cancer. This is consistent with the results of longitudinal studies of haemochromatosis patient cohorts showing that primary liver cancer is a common cause of death [1-3]. We found 8 HFE C282Y homozygotes in a cohort of 144 HCC cases; of these 6 had been previously diagnosed with hereditary haemochromatosis. The genetic data from the 2 other cases could be interpreted as evidence of undiagnosed hereditary haemochromatosis leading to HCC. However, the clinical implication of this finding is uncertain because these cases were diagnosed with HCC in 1985 and 1990; before which there was less awareness of, and surveillance for, hereditary haemochromatosis.
Our previous studies have shown that most people with HFE mutations can survive to old age and do not suffer from signs of iron overload and haemochromatosis [18,19]. Large population screens also suggest that only a minority of HFE C282Y homozygotes develop clinical signs and symptoms of iron overload [20-22].
We have now studied a large cohort of HCC patients collected over three decades in a well-defined population. We estimate that between 1.31% (95% CI = 0.52–3.32%) and 2.1% (95% CI = 0.89–4.05%) of males homozygous for the C282Y genotype have diagnosed and recorded HCC. We found zero penetrance for the HFE C282Y homozygous genotype with respect to HCC in women. We have previously shown that, in Norfolk, only a small proportion of HFE C282Y homozygotes have been diagnosed with and treated with venesection for haemochromatosis [23]; therefore pre-cirrhotic management of haemochromatosis does not explain the low penetrance described.
These estimates of penetrance in men and women are higher and lower respectively than our previous combined estimate for men and women of 0.4% (95% CI = 0–1%) [11]. A combined figure for men and women based on the data presented here would be slightly higher than our previous estimate. This difference results mainly from a lower normal population frequency for the HFE C282Y allele in the much larger and better age-matched normal control population presented here.
To estimate the penetrance of the HFE mutations we have carried out a cross-sectional study of histologically confirmed cases of HCC and used cancer registry data that is reliant on an accurate clinical diagnosis. The main source of systematic error in estimating penetrance in this study is likely to be unrecorded or mis-classified HCC. If HCC cases have not been reported to the cancer registry or were not accurately classified (e.g. recorded as liver cancer not otherwise specified) then our calculation of the penetrance of these HFE mutations would be an under-estimate.
Any error is likely to be small for two reasons. First, cancer is a notifiable disease and, 18 months after diagnosis in East Anglia, ascertainment for all tumours is nearly 100% complete (Sara Godward, East Anglian Cancer Intelligence Unit; personal communication). Secondly, HCC usually develops as a long-term complication of cirrhosis which will often have been detectable several years beforehand.
The frequency of HFE C282Y homozygosity in HCC patients in this study (5.6%) is similar to those seen in other studies of northern European populations [24,25]. Blanc et al. found that 5.7% of individuals, in a selected group of French HCC cases developing without cirrhosis, were HFE C282Y homozygotes [24]. Cauza et al. found that 3.1% of a cross-section of 162 Austrian HCC cases were HFE C282Y homozygotes [25]. For males, Austria is an area with high rates of HCC (10.5/100,000) [26], low HFE C282Y allele frequencies (5%) [27,28] and similar life expectancy relative to Britain. Using this limited data we calculate that this corresponds to a higher penetrance, at about 10%, of HFE C282Y homozygosity with respect to HCC in Austria.
Studies from Italy [29,30] and Spain [31] have reported no HFE C282Y homozygotes among cohorts of HCC patients. However, these were small studies in populations with low HFE C282Y prevalence. One large study also found no HFE C282Y homozygotes among a cohort of French HCC patients [32]. However, the case ascertainment was different to our own; having excluded any patients suspected to have genetic haemochromatosis thus precluding comparison with our data.
The expression of the life-threatening clinical manifestations of haemochromatosis has been shown to be affected by environmental modifying factors that may also be population specific. Italian haemochromatosis patients who were older than 55 years, had cirrhosis, a history of high alcohol consumption and were positive for antibodies for hepatitis B at diagnosis, had a 150 times higher relative risk of HCC [33]. Excessive alcohol consumption was also shown to accentuate the expression of haemochromatosis in French HFE C282Y homozygotes [34]. One limitation of this study is that we did not collect data on any environmental risk factors that our patients with HCC were exposed to; such as high alcohol use and chronic viral hepatitis. The low mean annual incidence of HCC (1.26 per 100,000 males) in East Anglia over the last three decades could reflect low exposure to environmental risk factors.
The value of population screening for HFE C282Y mutations partly depends on the penetrance with respect to the life-threatening manifestations of haemochromatosis. The zero penetrance described in this study for female HFE C282Y homozygotes developing HCC is at odds with proposals for whole population screening. However, male HFE C282Y homozygotes have a high relative risk of developing HCC. A targeted screening strategy that considers synergistic factors could prove effective for the prevention of this fatal disease, on the grounds of both cost and health.
Conclusion
• In this U.K. population we have shown that male HFE C282Y homozygotes are over-represented in a cross-section of confirmed HCC cases collected over three decades. Most of the HFE C282Y homozygotes had been previously diagnosed with hereditary haemochromatosis; we therefore found little evidence of undiagnosed haemochromatosis-related HCC over a thirty year period.
• We have used our genotyping data to estimate that between 1.31% and 2.1% of males homozygous for the C282Y genotype but no females have diagnosed and recorded HCC which corresponds to a low estimate of penetrance.
• The annual incidence of hepatocellular carcinoma in East Anglia is low relative to other populations which may reflect low exposure to environmental risk factors for HCC. In other populations these environmental risk factors have been shown to synergise with HFE mutations.
Abbreviations
Hepatocellular carcinoma (HCC), odds ratio (OR) confidence interval (CI).
Competing interests
The author(s) declare that they have no competing interests.
Authors' contributions
GW and BAJ conceived of the study, participated in the design of the study*, carried out the genetic analysis and drafted the manuscript. GW also carried out the statistical analysis.
JZW and IWF participated in the design of the study* and helped to draft the document.
VB and RL participated in the design of the study*, sample collection, helped to draft the document and carried out histological analysis.
*The study design included LREC and research governance applications.
All authors read and approved the final manuscript.
Pre-publication history
The pre-publication history for this paper can be accessed here:
Acknowledgements
We are grateful to The Big C Appeal and The BUPA Foundation for the grants for equipment, consumables and salary used in this study. We are also grateful to Katy Smith and Della Heron for assistance with DNA extraction and to Leigh Roberts and Sara Godward from the East Anglian Cancer Intelligence Unit for providing cancer registry data.
Figures and Tables
Table 1 HFE C282Y allelic discrimination assay using taqman probes. The primers and probes used to amplify and detect the portion of the HFE gene around codon 282.
Primers and Probes Concentration
forward primer ggctggataaccttggctgtac 300 nM
reverse primer gtcacataccccagatcacaatga 900 nM
mutant probe fam tgctccacctggtacgtatatctctgct 100 nM
wild type probe vic ctccacctggcacgtatatctctgct 100 nM
Table 2 HFE genotypes for codon 282 of HCC cases and control cohorts. Ages in years and numbers (percentages) of cases with each genotype.
HCC Patients Controls *
Mean Age 54 a 57
CC (normal) 119 (82.6%) 1331 (88.2%)
CY (heterozygotes) 17 (11.8%) 168 (11.1%) p = 0.34
YY (homozygotes) 8 (5.6%) 9 (0.6%) p = 0.00003
Total Number 144 1508
p values are single tailed probabilities of over-representation of mutant genotypes in HCC patients by Fisher's exact test.
* Norfolk data presented in Willis et al. 2002 [12], Cambridge data presented in Halsall et.al. 2003 [14].
a age data available for only 138/144 cases.
Table 3 Features of HFE C282Y homozygotes. Cases 1–3 were from Norwich, cases 1 and 2 were included in our previous study [11].
Histology (and histological staging of fibrosis) Grade of Siderosis (liver iron) Venesectiona Age# Sex
1 HCC (inadequate for staging) NA* No 66 M
2 HCC, cirrhosis (inadequate for staging) Grade 1 No 70 M
3 HCC (inadequate for staging) Grade 3/4 Yes 64 M
4 HCC, fibrosis (stage 4) Grade 2/3 Yes 64 M
5 HCC, cirrhosis (stage 6) Grade 4 No 73 M
6 HCC, fibrosis (stage 4) Grade 1 Yes 64 M
7 HCC, cirrhosis (stage 6) Grade 4 Not known 63 M
8 HCC, cirrhosis (stage 6) Grade 2/3 Yes 62 M
a Information about previous venesection treatment obtained from histology request forms and reports.
# Age at biopsy.
* Not applicable; only tumour was present in the needle biopsy sample.
HCC, hepatocellular carcinoma.
==== Refs
Niederau C Fischer R Purschel A Stremmel W Haussinger D Strohmeyer G Long term survival in patients with hereditary haemochromatosis Gastroenterology 1996 110 1107 1119 8613000
Fracanzani AL Conte D Fraquelli M Taioli E Mattioli M Losco A Fargion S Increased cancer risk in a cohort of 230 patients with hereditary hemochromatosis in comparison to matched control patients with non-iron-related chronic liver disease Hepatology 2001 33 647 651 11230745 10.1053/jhep.2001.22506
Elmberg M Hultcrantz R Ekbom A Brandt L Olsson S Olsson R Lindgren S Loof L Stal P Wallerstedt S Almer S Sandberg-Gertzen H Askling J Cancer risk in patients with hereditary hemochromatosis and in their first-degree relatives Gastroenterolology 2003 125 1733 1741 14724826
Niederau C Fischer R Sonnenberg A Stremmel W Trampisch HJ Strohmeyer G Survival and causes of death in cirrhotic and in non cirrhotic patients with primary haemochromatosis N Engl J Med 1985 313 1256 62 4058506
Feder JN Gnirke A Thomas W Tsuchihashi Z Ruddy DA Basava A Dormishian F Domingo R JrEllis MC Fullan A Hinton LM Jones NL Kimmel BE Kronmal GS Lauer P Lee VK Loeb DB Mapa FA McClelland E Meyer NC Mintier GA Moeller N Moore T Morikang E Prass CE Quintana SM Schatzman RC Brunke KJ Drayna DT Risch NJ Bacon BR Wolff RK A novel MHC class I-like gene is mutated in patients with hereditary hemochromatosis Nature Genet 1996 13 399 408 8696333 10.1038/ng0896-399
Beutler E Gelbart T West C Lee P Adams M Blackstone R Pockros P Kosty M Venditti CP Phatak PD Seese NK Chorney KA Ten Elshof AE Gerhard GS Chorney M Mutation analysis in hereditary hemochromatosis Blood Cells Mol Dis 1996 22 187 194 8931958 10.1006/bcmd.1996.0027
The UK Haemochromatosis Consortium A simple genetic test identifies 90% of UK patients with haemochromatosis Gut 1997 41 841 44 9462220
Bassett ML Leggett BA Halliday JW Webb S Powell LW Analysis of the cost of population screening for haemochromatosis using biochemical and genetic markers Hepatology 1997 27 517 524 10.1016/S0168-8278(97)80357-1
Burt MJ George PM Upton JD Collett JA Frampton CM Chapman TM Walmsley TA Chapman BA The significance of haemochromatosis gene mutations in the general population: implications for screening Gut 1998 43 830 836 9824612
Willis G Wimperis JZ Lonsdale R Fellows IW Watson MA Skipper LM Jennings BA Incidence of liver disease in people with HFE mutations Gut 2000 46 401 404 10673304 10.1136/gut.46.3.401
Willis G Wimperis JZ Lonsdale R Jennings BA Haemochromatosis gene mutation in hepatocellular cancer Lancet 1997 350 365 366 10.1016/S0140-6736(05)63143-1
Willis G Scott DGI Jennings BA Smith K Bukhari M Wimperis JZ HFE mutations in an inflammatory arthritis population Rheumatology 2002 41 176 179 11886966 10.1093/rheumatology/41.2.176
Sampson MJ Williams T Heyburn PJ Greenwood RH Temple RC Wimperis JZ Jennings BA Willis GA HFE Prevalence of HFE Haemochromatosis gene mutations in unselected male patients with Type 2 diabetes J Lab Clin Med 2000 135 170 173 10695662 10.1067/mlc.2000.104464
Halsall DJ McFarlane I Luan J Cox TM Wareham NJ Typical type 2 diabetes mellitus and HFE gene mutations: a population-based case-control study Hum Mol Genet 2003 12 1361 1365 12783844 10.1093/hmg/ddg149
Government Actuary's Department Life tables
East Anglian Cancer Intelligence Unit
Ishak K Baptista A Bianchi L Callea F De Groote J Gudat F Denk H Desmet V Korb G MacSween RNM Phillips MJ Portmann BG Poulsen H Scheuer PJ Schmid M Thaler H Histological grading and staging of chronic hepatitis J Hepatol 1995 22 696 699 7560864 10.1016/0168-8278(95)80226-6
Willis G Wimperis JZ Smith KC Fellows IW Jennings BA HFE (haemochromatosis gene) C282Y homozygotes in an elderly male population Lancet 1999 354 221 222 10421310 10.1016/S0140-6736(99)02195-9
Willis G Wimperis JZ Smith K Fellows IW Jennings BA HFE mutations in the elderly Blood Cells Mol Dis 2003 31 240 246 12972032 10.1016/S1079-9796(03)00158-X
Jackson HA Carter K Darke C Guttridge MG Ravine D Hutton RD Napier JA Worwood M HFE mutations, iron deficiency and overload in 10500 blood donors Br J Haematol 2001 114 474 84 11529872 10.1046/j.1365-2141.2001.02949.x
Beutler E Felitti VJ Koziol JA Ho NJ Gelbart T Penetrance of 845G→A (C282Y) HFE hereditary haemochromatosis mutation in the USA Lancet 2002 359 211 218 11812557 10.1016/S0140-6736(02)07447-0
Andersen RV Tybjaerg-Hansen A Appleyard M Birgens H Nordestgaard BG Hemochromatosis mutations in the general population: iron overload progression rate Blood 2004 103 2914 2919 15070663 10.1182/blood-2003-10-3564
Willis G Jennings BA Goodman E Fellows IW Wimperis JZ A high prevalence of HLA-H 845A mutations in hemochromatosis patients and the normal population in eastern England Blood Cells Mol Dis 23 288 291 9410472 10.1006/bcmd.1997.0145
Blanc JF De Ledinghen V Bernard PH de Verneuil H Winnock M Le Bail B Carles J Saric J Balabaud C Bioulac-Sage P Increased incidence of HFE C282Y mutations in patients with iron overload and hepatocellular carcinoma developed in non-cirrhotic liver J Hepatol 2000 32 805 811 10845668 10.1016/S0168-8278(00)80250-0
Cauza E Peck-Radosavljevic M Ulrich-Pur H Datz C Gschwantler M Schoniger-Hekele M Schoniger-Hekele M Hackl F Polli C Rasoul-Rockenschaub S Muller C Wrba F Gangl A Ferenci P Mutations of the HFE gene in patients with hepatocellular carcinoma Am J Gastroenterol 2003 98 442 447 12591066 10.1111/j.1572-0241.2003.07222.x
Bray F Sankila R Ferlay J Parkin DM Estimates of cancer incidence and mortality in Europe in 1995 Eur J Cancer 2002 38 99 166 11750846 10.1016/S0959-8049(01)00350-1
Kazemi-Shirazi L Datz C Maier-Dobersberger T Kaserer K Hackl F Polli CT Steindl PE Penner E Ferenci P The relation of iron status and hemochromatosis gene mutations in patients with chronic hepatitis C Gastroenterology 1999 116 127 134 9869610
Datz C Haas T Rinner H Sandhofer F Patsch W Paulweber B Heterozygosity for the C282Y mutation in the hemochromatosis gene is associated with increased serum iron, transferring saturation, and haemoglobin in young women: a protective role against iron deficiency? Clin Chem 1998 44 2429 2432 9836708
Racchi O Mangerini R Rapezzi D Gaetani GF Nobile MT Picciotto A Ferraris AM Mutations of the HFE gene and the risk of hepatocellular carcinoma Blood Cells Mol Dis 1999 25 350 353 10660482 10.1006/bcmd.1999.0263
Pirisi M Toniutto P Uzzau A Fabris C Avellini C Scott C Apollonio L Beltrami CA Bresadola F Carriage of HFE mutations and outcome of surgical resection for hepatocellular carcinoma in cirrhotic patients Cancer 2000 89 297 302 10918159 10.1002/1097-0142(20000715)89:2<297::AID-CNCR14>3.0.CO;2-N
Lauret E Rodriguez M Gonzalez S Linares A Lopez-Vazquez A Martinez-Borra J Rodrigo L Lopez-Larrea C HFE gene mutations in alcoholic and virus-related cirrhotic patients with hepatocellular carcinoma Am J Gastroenterol 2002 97 1016 1021 12003382 10.1111/j.1572-0241.2002.05553.x
Boige L Castera L de Roux N Ganne-Carrie N Ducot B Pelletier G Beaugrand M Buffet C Lack of association between HFE gene mutations and hepatocellular carcinoma in patients with cirrhosis Gut 2003 52 1178 1181 12865278 10.1136/gut.52.8.1178
Fargion S Fracanzani AL Piperno A Braga M D'Alba R Ronchi G Fiorelli G Prognostic factors for hepatocellular carcinoma in genetic hemochromatosis Hepatology 1994 20 1426 1431 7982640
Scotet V Merour MC Mercier AY Chanu B Le Faou T Raguenes O Le Gac G Mura C Nousbaum JB Ferec C Hereditary hemochromatosis: effect of excessive alcohol consumption on disease expression in patients homozygous for the C282Y mutation Am J Epidemiol 2003 158 129 134 12851225 10.1093/aje/kwg123
| 15929796 | PMC1175847 | CC BY | 2021-01-04 16:03:26 | no | BMC Gastroenterol. 2005 Jun 1; 5:17 | utf-8 | BMC Gastroenterol | 2,005 | 10.1186/1471-230X-5-17 | oa_comm |
==== Front
BMC Health Serv ResBMC Health Services Research1472-6963BioMed Central London 1472-6963-5-371589288210.1186/1472-6963-5-37CorrespondenceIdentification of ambiguities in the 1994 chronic fatigue syndrome research case definition and recommendations for resolution Stouten Bart [email protected] Violierstraat 27, 5402 LA Uden, The Netherlands2005 13 5 2005 5 37 37 16 8 2004 13 5 2005 Copyright © 2005 Stouten; licensee BioMed Central Ltd.2005Stouten; licensee BioMed Central Ltd.This is an Open Access article distributed under the 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 recent article by Reeves et al. on the identification and resolution of ambiguities in the 1994 chronic fatigue syndrome (CFS) research case definition recommended the Checklist Individual Strength, the Chalder Fatigue Scale, and the Krupp Fatigue Severity Scale for evaluating fatigue in CFS studies. To be able to discriminate between various levels of severe fatigue, extreme scoring on the individual items of these questionnaires must not occur too often.
Methods
We derived an expression that allows us to compute a lower bound for the number of items with the maximum item score for a given study from the reported mean scale score, the number of reported subjects, and the properties of the fatigue rating scale. Several CFS studies that used the recommended fatigue rating scales were selected from literature and analyzed to verify whether abundant extreme scoring had occurred.
Results
Extreme scoring occurred on a large number of the items for all three recommended fatigue rating scales across several studies. The percentage of items with the maximum score exceeded 40% in several cases. The amount of extreme scoring for a certain scale varied from one study to another, which suggests heterogeneity in the selected subjects across studies.
Conclusion
Because all three instruments easily reach the extreme ends of their scales on a large number of the individual items, they do not accurately represent the severe fatigue that is characteristic for CFS. This should lead to serious questions about the validity and suitability of the Checklist Individual Strength, the Chalder Fatigue Scale, and the Krupp Fatigue Severity Scale for evaluating fatigue in CFS research.
==== Body
Text
Since ambiguities in the 1994 chronic fatigue syndrome (CFS) research case definition [1] do indeed contribute to inconsistenties in the identification of cases, I welcome the publication by Reeves et al. [2] and the authors' efforts to resolve these problems. However, I have to express my deepest concerns about the three instruments that the authors have recommend for measuring fatigue in research studies on CFS. Because all three instruments easily reach the extreme ends of their scales on a large number of the individual items, they do not accurately represent the severe fatigue that is required to satisfy any of the published CFS research case definitions [1,3-5]. This low ceiling effect seriously distorts the fatigue measurements, which will inevitably result in bias and potentially misleading results.
To verify that the three recommended instruments do indeed exhibit low ceiling effects, one can study the mean scale scores that are reported in the literature. The recommended instruments were the Checklist Individual Strength (CIS) [6], the Chalder Fatigue Scale [7], and the Krupp Fatigue Severity Scale [8]. Each of these questionnaires consists of a fixed number of questions or statements. The answer to each question or the degree to which the participant agrees with a statement is scored on a certain scale. A question or statement with its corresponding scale is referred to as an item, and the assigned value corresponding to the participant's answer as the item score. A participant's fatigue rating scale score Y is computed by summing his individual item scores.
We can derive a lower bound L for the number of items with a maximum score for a given study by combining the reported mean fatigue rating scale score with the properties of the scale. Let us denote the reported number of subjects by n and the mean scale score of these subjects by . We consider instruments that consist of N items, with m possible scores for each item. Each item score is an element of the set {S1, S2,..., Sm - 1, Sm}, where Si - 1 <Si. Hence, S1 and Sm are respectively the minimum and maximum possible item scores. We count the number of items with a certain score Si, and denote this number by ki. Because we have n individuals who each answered N questions, the ki's add up to nN. Consequently,
The sum of the item scores of all individuals together is equal to n. Moreover, it is also equal to . Since Si - 1 <Si, we find that
Hence, we find that the lower bound L that we were looking for is given by
If L should be negative, which happens when is less than N Sm - 1, then we set L to zero. A lower bound for the percentage of items with the maximum score is . Note that this percentage is independent of the number of subjects in the study.
Lower bounds L for the number of items with the maximum score corresponding to data reported in literature were computed for each of the recommended fatigue rating scales. Because a recent Dutch article [9] recommended the Shortened Fatigue Questionnaire (SFQ) for assessing fatigue in clinical practice, this scale was also included in the analysis. The SFQ is simply a reduced version of the CIS 'fatigue severity' subscale, so the two are closely related.
At least two articles per fatigue rating scale were selected on a rather arbitrary basis. Subjects fulfilled the CDC88-CFS [3], Oxford-CFS [5], CDC94-CFS [1], or CDC94-UCF (unexplained chronic fatigue, i.e. either CFS or idiopathic chronic fatigue) [1] criteria. In particular, the study by Vercoulen et al. [10] was selected because it contains detailed data on the distribution of the scores for each CIS subscale. The study by Alberts et al. [11] was included because it contained normative data for the SFQ. The study by Vermeulen et al. [12] was selected to also include data on the SFQ from another source than the University Medical Centre Nijmegen. The article by Jason et al. [13] was selected because it was specifically concerned with the reliability and validity of a screening instrument for CFS. A recent Cochrane review [14] has investigated the relative effectiveness of exercise therapy and control treatments for CFS. All four studies that were included in that review and that have already been published [15-18] were analyzed here (one study by Moss-Morris et al. that was included in the review was submitted but not yet published). The other studies were selected because they were easily available to the author. Baseline data for Friedberg and Krupp [19] and Deale et al. [20] were read from the graphs presented in the articles. It is remarked that the 'matched ambulant group' in Van der Werf et al. [21] is a subset of the 'total ambulant group' in that study. Furthermore, the 'research participants' in Van der Werf et al. [22] are the same subjects as the 'total ambulant group' in [21].
The lower bounds for the number of items with the maximum score are presented in Table 1. From the lower bounds listed in the last column of the table we see that for several studies the number of items with the maximum score is larger than 40%. It is emphasized that the lower bounds were derived assuming a worse case scenario for the distribution of the item scores, i.e. participants have either the highest or the second highest possible score on each item. Since the worse case distribution is quite unrealistic, in reality the percentages of items with the maximum score are generally (even) higher than the values reported in the table. For example, according to the table it is not possible to conclude that extreme scoring occurred on the 'physical activity' subscale of the CIS in the study by Vercoulen et al. [10]. However, according to additional data listed in that article the 80th percentile of the 'physical activity' subscale is equal to the maximum possible subscale score of 3 × 7 = 21. Thus approximately 20% of the subjects reached the extreme score on all of their items, from which we can infer that extreme scoring occurred on at least 20% of the items.
Table 1 Lower bounds for the number of items with the maximum score for several studies. N is the number of items that constitute the (sub)scale, Sm is the maximum possible individual item score, n is the reported number of subjects, is the reported mean (sub)scale score, and L is the derived lower bound for the number of items with the maximum score. The last column lists a lower bound for the percentage of items with the maximum score based on L. The second highest possible item score Sm - 1 is equal to Sm - 1 for all considered (sub)scales.
Scale N Sm n L
Checklist Individual Strength Oxford-CFS, CDC94-UCF; Vercoulen et al. [10]
-fatigue severity subscale 8 7 758 51.7 2805 46%
-physical activity subscale 3 7 758 16.9 0 0%
-reduced motivation subscale 4 7 758 17.0 0 0%
-concentration subscale 5 7 758 27.5 0 0%
Checklist Individual Strength CDC94-UCF; van der Werf et al. [21]
-homebound group; fatigue severity subscale 8 7 18 53.6 101 70%
-matched ambulant group; fatigue severity subscale 8 7 32 52.8 154 60%
-total ambulant group; fatigue severity subscale 8 7 270 52.1 1107 51%
-homebound group; physical activity subscale 3 7 18 15.8 0 0%
-matched ambulant group; physical activity subscale 3 7 32 17.0 0 0%
-total ambulant group; physical activity subscale 3 7 270 17.6 0 0%
-homebound group; concentration subscale 5 7 15 22.4 0 0%
Shortened Fatigue Questionnaire van der Werf et al. [22]
-survey respondents (Dutch ME-Association members) 4 7 1955 23.9 0 0%
-research participants (CDC94-UCF) 4 7 270 26.1 567 53%
Shortened Fatigue Questionnaire Oxford-CFS, CDC94-UCF; Alberts et al. [11]
-normative data for CFS 4 7 445 26 to 27 890 50%
Shortened Fatigue Questionnaire CDC94-CFS; Vermeulen et al. [12]
-study group 4 7 35 24.8 28 20%
Krupp Fatigue Severity Scale CDC88-CFS; Friedberg et al. [19]
-treatment group 9 7 22 58 88 44%
-no-treatment group 9 7 22 51 0 0%
Krupp Fatigue Severity Scale CDC88-CFS; DeLuca et al. [23]
-subjects with concurrent axis 1 psychiatric disorder 9 7 12 58.5 54 50%
-subjects without concurrent psychiatric disorder 9 7 21 57.2 67 36%
14-item Chalder Fatigue Scale Oxford-CFS; Wearden et al. [15]
-'exercise and fluoxetine group' 14 3 33 35.9 261 56%
-'exercise and placebo group' 14 3 34 33.7 194 41%
-'exercise control and fluoxetine group' 14 3 35 34.4 224 46%
-'exercise control and placebo group' 14 3 34 34.0 204 43%
14-item Chalder Fatigue Scale Oxford-CFS; Fulcher et al. [16]
-exercise group 14 3 33 28.9 30 6%
-fiexibility group 14 3 33 30.5 83 18%
11-item Chalder Fatigue Scale CDC94-CFS; Jason et al. [13]
-physical subscale 7 3 15 18.40 66 63%
-mental subscale 4 3 15 9.13 17 28%
11-item Chalder Fatigue Scale CDC94-CFS; Wallman et al. [17]
-exercise group; physical subscale 7 3 32 11.6 0 0%
-exercise group; mental subscale 4 3 32 6.3 0 0%
-relaxation/flexibility group; physical subscale 7 3 29 11.4 0 0%
-relaxation/flexibility group; mental subscale 4 3 29 5.6 0 0%
11-item bimodal Chalder Fatigue Scale Oxford-CFS and CDC94-CFS; Deale et al. [20]
-cognitive behavior therapy group 11 1 30 10.1 303 92%
-relaxation group 11 1 30 9.3 279 85%
11-item bimodal Chalder Fatigue Scale Oxford-CFS; Powell et al. [18]
-control group 11 1 34 10.6 360 96%
-minimum intervention group 11 1 37 10.4 385 95%
-telephone intervention group 11 1 39 9.9 386 90%
-maximum intervention group 11 1 38 10.2 388 93%
It should be clear that extreme scoring on a large number of items occurred for all scales across several studies. Only the 'concentration' and 'reduced motivation' subscales of the CIS did not show evidence of extreme scoring. That the amount of extreme scoring for a certain scale varies from one study to another suggests heterogeneity in the selected subjects across studies. Since the studies that were analyzed were selected on a rather arbitrary basis and not in a systematic way, the data in Table 1 should not be regarded as a true reflection of the CFS literature as a whole. The main point is that it does prove that abundant extreme scoring occurred for all the recommended fatigue rating scales in at least some of the CFS studies published in literature.
One only needs to glance at the three recommended instruments to understand why extreme scoring occurs so often. The CIS and the Krupp Fatigue Severity Scale consist of statements like "I feel tired" and "I am easily fatigued" that are scored on seven-point scales (from "yes, that is true" to "no, that is not true" for the CIS; from "strongly disagree" to "strongly agree" for the Krupp scale). Thus it does not matter whether a subject feels 'extremely tired,' 'severely tired' or 'just tired,' and is 'easily extremely fatigued,' 'easily severely fatigued' or 'easily fatigued;' he will score on the extreme end of the scale for all these cases. A similar argument applies to the Chalder Fatigue Scale, where the participant has to choose from one of four answers like "less than usual," "no more than usual," "more than usual" and "much more than usual" to questions such as "Do you feel weak?" For the continuous version of the Chalder scale answers are rated from 0 to 3, for the bimodal version the scoring system is {0, 0, 1, 1}. This explains why the binary version performs even worse than the continuous version.
Interestingly, the ceiling effect has been noted before by members of the International CFS Study Group in their individual publications: "The CIS-fatigue score [i.e. the 'fatigue severity' subscale of the CIS] involves an overall rating and in CFS samples easily reaches the extreme end of its scale" [21]; "a ceiling effect in the [Krupp] Fatigue Severity Scale may limit its utility to assess severe fatigue-related disability" [24]. A publication that examined the distribution of the 14 items of the Chalder Fatigue Scale in 136 CFS patients found that "Scores on eight items were normally distributed, but six items ('tiredness,' 'resting more,' 'lacking energy,' 'feeling weak,' 'feeling sleepy or drowsy,' and 'starts things without difficulty but gets weaker as goes on') were highly skewed with the majority of patients reaching the maximum score" [25].
Abundant extreme scoring and the corresponding inability to discriminate between various levels of severe fatigue can lead to misleading results in several ways. For example, van der Werf et al. [21] compared a group of 18 homebound CF(S) patients with a group of 32 matched ambulant CF(S) patients. No significant difference was found when fatigue was measured with the CIS 'fatigue severity' subscale (p = 0.39). But when fatigue was measured with the 'Daily Observed Fatigue' scale that does not exhibit such a strong ceiling effect, it was concluded that the homebound group was significantly more fatigued than the ambulant group (p < 0.01). Another problem occurs when studying the relation between the experienced level of fatigue and another factor such as social support. Then the correlation between the two will certainly be distorted if the fatigue measurement has a low ceiling effect and the other measure has not. The most dangerous situation however arises when a scale with low ceiling is used as a primary outcome measure to evaluate a CFS treatment. Consider five patients with a baseline CIS-fatigue score of 52 (e.g. the mean baseline score in Prins et al. [26] was 52.1). Suppose one patient improves (e.g. CIS-fatigue = 16 at follow-up) and the other four patients become extremely fatigued due to treatment (CIS-fatigue = 56 at follow-up, i.e. the maximum scale score). Then still the overall mean has improved from 52 to 48, even though 80% of the subjects are substantially more fatigued after treatment. In particular, participants who already have the maximum scale score at baseline can never get worse according to the 'recommended' fatigue rating scales. Systematic errors that may result in artificial treatment effects opposite to the true situation should be avoided at all times.
Unfortunately, the reasons for recommending the CIS, the Krupp and the Chalder scales in the main article text are limited to 'they have been used before,' 'normative data have been collected' and 'receiver-operating characteristics have been published.' In the Author's response to reviews (25 July 2003) that is available on the pre-publication site of the article, the authors remark that these are all 'standardized, validated, internationally accepted instruments' without giving any reference to support this statement. Although the recommended fatigue rating scales might indeed be accepted by numerous scientists of various nationalities, the evidence presented here must lead to serious questions about their validity and suitability for CFS research.
Noticeably, the Profile of Fatigue-Related Symptoms (PFRS) that was developed more than a decade ago by Ray et al. [27,28] is a rating scale that does not has the flaw of low ceiling in CFS samples. It consists of the four subscales 'Emotional Distress,' 'Cognitive Difficulty,' 'Fatigue' and 'Somatic Symptoms.' All subscales have high reliability and showed good convergence with comparison measures. Why was the PFRS not included in the authors' advice? To shed some light on the underlying scientific process that has ultimately led to their recommendations, I would like to ask the authors to make the workshop summaries and the focus group reports available.
Strictly speaking, the CIS, the Krupp Fatigue Severity Scale and the Chalder Fatigue Scale are all able to discriminate between CFS subjects and healthy subjects. Thus all three might indeed be used to improve the precision of CFS case ascertainment for research studies. However, if one really wishes to take CFS research forwards instead of three steps backwards, then it would be wise to abandon these low ceiling fatigue rating scales and start focussing on instruments that accurately represent the severe fatigue that is currently defined to be so characteristic for CFS.
Competing interests
The author(s) declares that he has no competing interests.
Authors' contributions
BS wrote the paper and performed the analysis.
Pre-publication history
The pre-publication history for this paper can be accessed here:
Acknowledgements
The author thanks Dr. Ellen Goudsmit, psychologist, for proofreading the original manuscript and providing valuable information on the various fatigue rating scales.
==== Refs
Fukuda K Straus SE Hickie I Sharpe MC Dobbins JG Komaroff A the International Chronic Fatigue Syndrome Study Group The chronic fatigue syndrome: a comprehensive approach to its definition and study Ann Intern Med 1994 121 953 959 7978722
Reeves WC Lloyd A Vernon SD Klimas N Jason LA Bleijenberg G Evengard B White PD Nisenbaum R Unger ER the International Chronic Fatigue Syndrome Study Group Identification of ambiguities in the 1994 chronic fatigue syndrome research case definition and recommendations for resolution BMC Health Serv Res 2003 3 25 14702202 10.1186/1472-6963-3-25
Holmes GP Kaplan JE Gantz NM Komaroff AL Schonberger LB Straus SE Jones JF DuBois RE Cunningham-Rundles C Pahwa S Tosato G Zegans LS Purtilo DT Brown N Schooley RT Brus I Chronic fatigue syndrome: a working case definition Ann Intern Med 1988 108 387 389 2829679
Lloyd AR Hickie I Boughton CR Spencer O Wakefield D Prevalence of chronic fatigue syndrome in an Australian population Med J Aust 1990 153 522 528 2233474
Sharpe MC Archard LC Banatvala JE Borysiewicz LK Clare AW David A Edwards RHT Hawton KEH Lambert HP Lane RJM McDonald EM Mowbray JF Pearson DJ Peto TEA Preedy VR Smith AP Smith DG Taylor DJ Tyrrell DAJ Wessely S White PD A report – chronic fatigue syndrome: guidelines for research J R Soc Med 1991 84 118 121 1999813
Bültmann U de Vries M Beurskens AJHM Bleijenberg G Vercoulen JHMM Kant IJ Measurement of prolonged fatigue in the working population: determination of a cutoff point for the checklist individual strength J Occup Health Psychol 2000 5 411 416 11051524 10.1037//1076-8998.5.4.411
Chalder T Berelowitz G Pawlikowska T Watts L Wessely S Wright D Wallace EP Development of a fatigue scale J Psychosom Res 1993 37 147 153 8463991 10.1016/0022-3999(93)90081-P
Krupp LB LaRocca NG Muir-Nash J Steinberg AD The fatigue severity scale: application to patients with multiple sclerosis and systemic lupus erythematosus Arch Neurol 1989 46 1121 1123 2803071
van Engelen BGM Kalkman JS Schillings ML van der Werf SP Bleijenberg G Zwarts MJ Moeheid bij neuromusculaire aandoeningen Ned Tijdschr Geneeskd 2004 148 1336 1341 15283024
Vercoulen JHMM Alberts M Bleijenberg G De checklist individual strength (CIS) Gedragstherapie 1999 32 131 136
Alberts M Smets EMA Vercoulen JHMM Garssen B Bleijenberg G 'Verkorte vermoeidheidsvragenlijst': een praktisch hulpmiddel bij het scoren van vermoeidheid Ned Tijdschr Geneeskd 1997 141 1526 1530 9543741
Vermeulen RCW Scholte HR Chronic fatigue syndrome and sexual dysfunction J Psychosom Res 2004 56 199 201 15016578 10.1016/S0022-3999(03)00546-4
Jason LA Ropacki MT Santoro NB Richman JA Heatherly W Taylor R Ferrari JR Haney-Davis TM Rademaker A Dupuis J Golding J Plioplys AV Plioplys S A screening instrument for chronic fatigue syndrome: reliability and validity Journal of Chronic Fatigue Syndrome 1997 3 39 59
Edmonds M McGuire H Price J Exercise therapy for chronic fatigue syndrome (Cochrane Review) The Cochrane Library 2004 Chichester, UK: John Wiley & Sons, Ltd
Wearden AJ Morriss RK Mullis R Strickland PL Pearson DJ Appleby L Campbell IT Morris JA Randomised, double-blind, placebo-controlled treatment trial of fluoxetine and graded exercise for chronic fatigue syndrome Br J Psychiatry 1998 172 485 490 9828987
Fulcher KY White PD Randomised controlled trial of graded exercise in patients with the chronic fatigue syndrome BMJ 1997 314 1647 1652 9180065
Wallman KE Morton AR Goodman C Grove R Guilfoyle AM Randomised controlled trial of graded exercise in chronic fatigue syndrome Med J Aust 2004 180 444 448 15115421
Powell P Bentall RP Nye FJ Edwards RHT Randomised controlled trial of patient education to encourage graded exercise in chronic fatigue syndrome BMJ 2001 322 387 390 11179154 10.1136/bmj.322.7283.387
Friedberg F Krupp LB A comparision of cognitive behavioral treatment for chronic fatigue syndrome and primary depression Clin Infect Dis 1994 18 S105 S110 8148435
Deale A Chalder T Marks I Wessely S Cognitive behavior therapy for chronic fatigue syndrome: a randomized controlled trial Am J Psychiatry 1997 154 408 414 9054791
van der Werf S Prins J Klein-Rouweler E Alberts M van der Meer J Bleijenberg G van der Werf SP Homebound chronic fatigue syndrome patients Determinants and consequences of experienced fatigue in chronic fatigue syndrome and neurological conditions PhD thesis 2000 Katholieke Universiteit Nijmegen 31 41
van der Werf S Prins J Jansen T van der Meer J Bleijenberg G van der Werf SP Results of a large survey among members of the Dutch ME-Association Determinants and consequences of experienced fatigue in chronic fatigue syndrome and neurological conditions PhD thesis 2000 Katholieke Universiteit Nijmegen 15 22
DeLuca J Johnson SK Ellis SP Natelson BH Cognitive functioning is impaired in patients with chronic fatigue syndrome devoid of psychiatric disease J Neurol Neurosurg Psychiatry 1997 62 151 155 9048715
Friedberg F Jason LA Selecting a fatigue rating scale The CFS Research Review 2002 35 7 11
Morriss RK Wearden AJ Mullis R Exploring the validity of the Chalder fatigue scale in chronic fatigue syndrome J Psychosom Res 1998 45 411 417 9835234 10.1016/S0022-3999(98)00022-1
Prins JB Bleijenberg G Bazelmans E Elving LD de Boo TM Severens JL van der Wilt GJ Spinhoven P van der Meer JWM Cognitive behaviour therapy for chronic fatigue syndrome: a multicentre randomised controlled trial Lancet 2001 357 841 847 11265953 10.1016/S0140-6736(00)04198-2
Ray C Weir WRC Phillips S Cullen S Development of a measure of symptoms in chronic fatigue syndrome: the profile of fatigue-related symptoms (PFRS) Psychol Health 1992 7 27 43
Ray C Weir WRC Cullen S Phillips S Illness perception and symptom components in chronic fatigue syndrome J Psychosom Res 1992 36 243 256 1564677 10.1016/0022-3999(92)90089-K
| 15892882 | PMC1175848 | CC BY | 2021-01-04 16:31:50 | no | BMC Health Serv Res. 2005 May 13; 5:37 | utf-8 | BMC Health Serv Res | 2,005 | 10.1186/1472-6963-5-37 | oa_comm |
==== Front
BMC NeurolBMC Neurology1471-2377BioMed Central London 1471-2377-5-101594904310.1186/1471-2377-5-10Research ArticleEffect of pre-stroke use of ACE inhibitors on ischemic stroke severity Selim Magdy [email protected] Sean [email protected] Italo [email protected] Louis [email protected] Gottfried [email protected] Department of Neurology, Beth Israel DeaconessMedical Center, Boston, USA2 Interventional Neuroradiology, Jackson Memorial Hospital, Miami, USA2005 10 6 2005 5 10 10 24 1 2005 10 6 2005 Copyright © 2005 Selim et al; licensee BioMed Central Ltd.2005Selim 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 trials suggest that angiotensin-converting enzyme inhibitors (ACEI) are effective in prevention of ischemic stroke, as measured by reduced stroke incidence. We aimed to compare stroke severity between stroke patients who were taking ACEI before their stroke onset and those who were not, to examine the effects of pretreatment with ACEI on ischemic stroke severity.
Methods
We retrospectively studied 126 consecutive patients presenting within 24 hours of ischemic stroke onset, as confirmed by diffusion-weighted magnetic resonance imaging (DWI). We calculated the NIHSS score at presentation, as the primary measure of clinical stroke severity, and categorized stroke severity as mild (NIHSS [less than or equal to] 7), moderate (NIHSS 8–13) or severe (NIHSS [greater than or equal to] 14). We analyzed demographic data, risk-factor profile, blood pressure (BP) and medications on admissions, and determined stroke mechanism according to TOAST criteria. We also measured the volumes of admission diffusion- and perfusion-weighted (DWI /PWI) magnetic resonance imaging lesions, as a secondary measure of ischemic tissue volume. We compared these variables among patients on ACEI and those who were not.
Results
Thirty- three patients (26%) were on ACE-inhibitors. The overall median baseline NIHSS score was 5.5 (range 2–21) among ACEI-treated patients vs. 9 (range 1–36) in non-ACEI patients (p = 0.036). Patients on ACEI prior to their stroke had more mild and less severe strokes, and smaller DWI and PWI lesion volumes compared to non-ACEI treated patients. However, none of these differences were significant. Predictably, a higher percentage of patients on ACEI had a history of heart failure (p = 0.03). Age, time-to-imaging or neurological evaluation, risk-factor profile, concomitant therapy with lipid lowering, other antihypertensives or antithrombotic agents, or admission BP were comparable between the two groups.
Conclusion
Our results suggest that ACE-inhibitors may reduce the clinical severity of stroke, as measured by NIHSS score. Further, larger-scale, prospective studies areneeded to validate our findings, and to elucidate the mechanism(s) of ACEImediated benefits in patients with ischemic stroke.
==== Body
Background
Data from the heart outcomes prevention evaluation study (HOPE) suggest that angiotensin-converting enzyme inhibitors (ACEI) are effective in prevention of ischemic stroke, as measured by reduced stroke incidence in subjects randomized to treatment with ACEI [1]. In this trial, the use of the ACEI, ramipril, resulted in a 32% reduction in ischemic stroke risk despite minimal reduction in blood pressure (BP) [1], leading some to suggest that ACEI may also exert direct neuroprotective effects.
To further elucidate if ACEI have potential neuroprotective effects, we tested whether their use prior to ischemic stroke onset might also reduce the severity of stroke. We examined clinical and admission magnetic resonance imaging (MRI) data from patients with ischemic stroke to determine the effects of prestroke use of ACEI on stroke severity.
Methods
Study design and patient selection
We retrospectively reviewed our prospectively collected stroke database over a 30-month period from 1998 to 2000, and identified consecutive patients who presented with acute ischemic stroke within 24 hours of onset and had DWI/PWI upon presentation. Onset time was defined, as the last time the patient was known to be in his/her usual state of health. The diagnosis of ischemic stroke was confirmed by diffusion-weighted imaging (DWI) showing evidence of acute cerebral infarction, combined with serial neurological examinations performed by stroke-trained neurologists. We included patients who had received thrombolytic, endovascular or experimental neuroprotective treatment. We only excluded patients who had transient ischemic attacks (TIAs), in whom DWI/PWI was negative.
Data collection and assessments
We retrieved the following data for each patient: (1) demographics; (2) risk factors for stroke, i.e. hypertension (HTN), diabetes mellitus (DM), hyperlipidemia, coronary artery disease (CAD), atrial fibrillation (AF), heart failure (CHF), history of TIA and smoking, as reported by the patient andhis/her family; (3) vital signs at presentation (BP and temperature); (4) blood glucose level at admission; (5) medications upon admission, with particular attention to antiplatelets, anticoagulants, lipid-lowering agents, and antihypertensives including ACEI. We did not collect information about the duration of medication(s) use, daily use or compliance. Patients and families were only questioned about patient's use of medication(s), including ACEI, in the week before stroke; (6) the baseline National Institute of Health Stroke Scale (NIHSS) score [2], which was recorded by stroke-trained neurolgistscertified in the application of NIHSS at admission; and (7) time from strokedetection to imaging.
Outcome measures
We used the NIHSS score at presentation as the primary measure of clinical stroke severity, and categorized stroke severity as mild (NIHSS score = 7), moderate (NIHSS score 8–13) or severe (NIHSS score = 14). We measured the total DWI and PWI lesion volumes, as secondary radiological measures of stroke severity, in 110/126 patients. All MRI studies were performed on a Siemens Medical Systems Vision 1.5-T MR whole body scanner with echoplanar imaging capabilities. An experienced researcher blinded to clinical data and patient's identity, performed MRI measurements. The volume of the perfusion abnormality was measured on relative Mean Transit Time (rMTT) maps. The specific MRI sequence parameters, imaging processing and volumetric analysis are described in details in previous publications [3,4]. We classified stroke mechanisms, after completing the diagnostic work-up, according to the Trial of Org 10172 in Acute Treatment (TOAST) criteria [5].
Statistical analysis
We divided patients into 2 groups, those who were taking ACEI before their stroke onset and those who were not. We compared inter-group differences between individual categorical variables by using student's t-test or Wilcoxon rank sum test for continuous variables, and Fisher's exact test forcategorical comparisons, as appropriate. We compared the median baseline NIHSS scores with the Mann-Whitney U test to evaluate the severity of the stroke in both groups. The Cochran-Mantel-Haenszel row mean score rank test, adjusted for various confounding variables (age, sex, risk factors, use of concomitant medications, time-from-stroke-to-evaluation, and stroke mechanism/subtype) was used to control for the differences in NIHSS scores between ACEI and non-ACEI users. A p-value of < 0.05 was considered statistically significant for all analyses.
Results
Patient characteristics (demographic and clinical Features)
A total of 126 patients met all of our inclusion and none of the exclusion criteria, and were included in subsequent analyses. Approximately, 26% (33 patients) were on ACEI before stroke onset. Fourteen were taking lisinopril (20 to 40 mg per day), 13 enalapril (10 to 40 mg per day), 5 captopril (75 to 150 mg per day) and 1 accupril (40 mg per day). None of our patients was taking perindopril or ramipril, or a combination of different ACEI. Table 1 summarizes the demographic and clinical features of patients in both ACEI- and non ACEI-treated groups. There were no significant differences in the mean age or sex distribution between the 2 groups. Slightly higher percentages of ACEI-treated patients had history of HTN, DM, hyperlipidemia, and smoking in comparison to the non-ACEI group. A slightly higher percentage of non-ACEI group reported history of prior TIA. However, none of these differences were statistically significant. There was a trend towards a higher frequency of cardiac disease among ACEI-treated patients. This was mostly driven by a significantly higher percentage of heart failure among ACEI-treated patients.
Table 1 Comparison of demographic and clinical features between ACEI- and non ACEI-treated groups.
Number of patients ACEI group 33 (26%) Non-ACEI group 93 (74%) p-value
Sex (women/men) 13/20 43/50
• % women 39% 46% 0.55
• % men 61% 54% 0.55
Mean age (year) ± SD 73 ± 11 70 ± 13 0.23
Risk factors:
History of hypertension 78% 65% 0.19
History of diabetes 32% 24% 0.36
History of hyperlipidemia 28% 22% 0.48
History of cardiac disease 39% 21% 0.06
• CHF 23% 9% 0.03*
• AF 10% 7% 0.69
• CAD 6% 5% 0.98
History of smoking 16% 11% 0.53
History of prior TIA 10% 16% 0.39
Concomitant medications:
Antiplatelets 37% 41% 0.68
Anticoagulants 12% 10% 0.74
Statins 20% 21% 0.99
Other BP lowering agents 56% 52% 0.84
• Diuretics 25% 18% 0.45
• B-blockers 18% 27% 0.35
• Ca++ blockers 13% 13% 0.99
• ARB 0% 4% 0.57
Time from stroke-to evaluation
• 0–6 h 61% 63% 0.83
• 6–25 h 39% 37% 0.83
Clinical features:
NIHSS score, median 5.5 9 0.036*
SBP (mean ± SD), mmHg 162 ± 27 158 ± 31 0.35
DBP (mean ± SD), mmHg 84 ± 16 81 ± 20 0.38
Temperature (mean ± SD), F° 97 ± 7 98 ± 6 0.34
A roughly equal percentage of patients in both groups presented to our emergency room and were imaged within 6 hours from stroke onset (61% in ACEI-treated patients vs. 63% in non-ACEI group; p = 0.83). The mean time from stroke-symptom onset to evaluation was 10.9+5.2 h in ACEI-treated patients vs. 11.3+6.4 h in non-ACEI group (p = 0.62). There were no significant differences in admission temperature (97+7 F° vs. 98+6 F°; p = 0.34) or glucose levels (137+23 mg/dL vs. 144+29 mg/dL; p = 0.26) between the 2 groups. The mean SBP upon admission was 162+27 mmHg in ACEI-treated patients vs. 158+31 mmHg in non-ACEI group, and the mean DBP was 84+16 vs. 81+20 mmHg. None of these differences were statistically significant.
A roughly equal percentage of patients in each group were using antiplatelets, anticoagulants, statins and other BP lowering agents. Similarly, the frequency of using other classes of antihypertensive agents was not significantly different in either group. Four patients were taking angiotensin receptor blockers (ARBs) at the time of their stroke. None of these 4 patients was on ACEI. They were all included in non ACEI-treated group for purposes of statistical analysis.
Patient outcomes
Approximately, 48% of ACEI-treated patients had baseline NIHSS score = 7 compared with 40% of non-ACEI group (p = 0.42); 28% had NIHSS score between 8 – 13 vs. 20% in non-ACEI group (p = 0.46); and 24% had NIHSS score = 14 vs. 40% in non-ACEI users (p = 0.18).
Figure 1 shows the distribution of stroke mechanisms, according to TOAST criteria, among ACEI- and non ACEI-treated patients. The stroke mechanisms were roughly equivalent in both groups. Although, cardioembolic cause was more frequent among non ACEI-treated patients (34% vs. 24%) and lacunar etiology was more commonly seen among patients who were taking ACEI prior to stroke onset (32% vs. 23%), these differences were not statistically significant.
Figure 1 Comparison of stroke mechanisms between ACEI- and non ACEI-treated patients.
The overall median NIHSS score at admission was significantly lower in ACEI-treated patients (5.5; range 2 – 21) than in non-ACEI patients (9; range 1 – 36; p = 0.036). This difference remained statistically significant after controlling for confounding variables, such as history of hypertension, TIA, DM hyperlipidemia and cardiac disease, including CHF, stroke mechanism, onsetto- evaluation time, and concomitant medications, using the Cochran-Mantel-Haenszel row mean score test using ranks adjusted for these covariates (p = 0.042). Similarly, the median NIHSS score at admission was lower in the ACEI group when the analysis was limited to patients with non-lacunar strokes (8.5 vs. 12; p = 0.03) or to those who presented within 6 hours of stroke symptom onset (6 vs. 8; p = 0.046).
The DWI/PWI lesions volumetric measurements were performed in 87% of the patients (110/126). Sixteen patients were not included in MRI data analyses because their images were of poor quality to allow adequate quantitative measurements. As table 2 indicates, there were no significant differences between both groups with regard to the mean diffusion, perfusion or perfusion-diffusion (mismatch) lesion volumes.
Table 2 Comparison of MRI between ACEI- and non ACEI-treated group
ACEI group Non-ACEI group p-value
DWI lesion volume (mean ± SD), cm3 25.2 ± 23.4 28.7 ± 25.0 0.55
PWI lesion volume (mean ± SD), cm3 72.6 ± 56.6 75.1 ± 68.5 0.86
Mismatch (PWI – DWI) volume 47.6 ± 39.5 46.6 ± 28.2 0.93
Discussion
We found that the baseline NIHSS score was lower in patients who were taking ACEI prior to their stroke compared to those who were not taking ACEI at the time of stroke onset. The NIHSS is accepted widely for measuring acute stroke deficits to assess the degree of severity of neurological deficits from stroke and its reliability has been tested in several clinical trials [6-8].
We found no difference in admission BP between ACEI and non-ACEI users, suggesting that the beneficial effects of ACEI use may not be directly related to their BP-lowering effect. This is concordant with the results from the HOPE trial [1].
Several factors could potentially account for the observed difference in baseline NIHSS between ACEI and non-ACEI groups, such as differences in risk factors, stroke mechanism/subtype, or baseline hemodynamic parameters. The risk factors that influence ACEI use, such as history of HTN, CHF and DM, were dissimilarly distributed between the two groups, and their impact on stroke type and severity cannot be entirely excluded. However, ourfindings are unlikely to be related to differences in baseline risk factor profile between the ACEI- and non-ACEI treated patients since patients who were on ACEI had a higher prevalence of HTN, DM and heart failure, which may have biased our data toward higher stroke severity in ACEI-treated patients, and thus limited our ability to detect larger differences in favor of ACEI use. It is possible that ACEI use reflected a greater degree of medical attention and more aggressive risk factor reduction in these patients, which subsequently lessened stroke severity.
Recent studies have shown that TIAs before stroke can induce tolerance (ischemic preconditioning) to subsequent strokes by raising the threshold of brain tissue vulnerability, which results in smaller infarct volumes, and better recovery [9-11]. We found no significant differences in the frequency of prior history of TIAs, as reported by the patient or his/her family, between the ACEI and non-ACEI treated patients. In fact, a slightly higher percentage of patients in the non-ACEI group reported history of TIAs prior to their presenting stroke.
Some epidemiological studies show that greater stroke severity at onset is associated with a shorter interval between symptom onset and time to emergency department arrival [12], suggesting that the observed difference in baseline NIHSS could be attributed to dissimilar distribution of patients' arrival time to the hospital. However, a roughly identical proportion of patients in both groups presented to our hospital within 6 hours of stroke-symptom onset.
Since the observed beneficial effect of ACEI on stroke severity could potentially be secondary to ACEI effects on stroke mechanism, we examined the impact of ACEI use on stroke mechanism using TOAST criteria. We found a greater preponderance of lacunar strokes among ACEI-treated patients and cardioembolic strokes among non-ACEI patients. However, these differences were not significant and the difference in baseline NIHSSS remained significant even after excluding patients with non-lacunar strokes from analysis. This suggests that ACEI use did not seem to influence stroke mechanism in our cohort of patients, and that our findings were unrelated to the higher frequency of lacunar strokes among ACEI-treated patients. Similarly, the beneficial effect of ACEI in our patients is unlikely to be related to other concomitant treatments. Although, several patients in both groups were on statins, antithrombotics and other antihypertensive agents, we found no significant difference between ACEI- and non ACEI-treated patients receiving any of these classes of drugs. Finally, it is noteworthy that the difference in baseline stroke severity between ACEI and non-ACEI groups remained statistically significant after adjusting for the above confounding variables.
A recent prospective observational study of 507 patients with first-ever ischemic stroke showed that treatment with ACEI at the time of stroke onset is associated with reduced plasma concentration of C-reactive protein and better long-term outcomes [13], suggesting that ACEI may have anti-inflammatory properties and reduce the acute-phase inflammatory response after stroke onset. There are several other potential mechanisms by which ACEI may provide benefit to stroke patients. Experimental data suggest that the rennin angiotensin system modulates the atherosclerotic process, and that angiotensin II exerts pro-inflammatory actions in the vascular wall, which induce the production of reactive oxygen species and hydroxyl radicals, cytokines and adhesion molecules [14-19]. Angiotensin converting enzyme inhibitors could provide neuroprotection via blockade of angiotensin II-mediated endothelial dysfunction, lipid peroxidation and subsequent oxidative stress, and vascular smooth muscle intracellular calcium accumulation and hypertrophy [14-20]. Furthermore, ACEI may help maintain homeostatic balance of fibrinolytic and procoagulant factors [21] and increase cerebral blood flow [22]. Recent studies using transcranial Doppler ultrasonography have shown that perindopril can improve the cerebral vasomotor reactivity in patients with lacunar infarcts beyond any effect on BP [22], and that treatment with quinapril can ameliorate cerebrovascular reactivity caused by methionine-induced hyperhomocysteinemia in healthy volunteers [23].
We found that ACE I use had no effect on MRI measures of ischemic lesion volume. This discrepancy between the beneficial effects of ACEI on clinical, but not radiological, measures of stroke severity is reconcilable since the correlation between infarct volume and NIHSS is only moderate, particularly in non-dominant hemispheric strokes [24-26]. We explored the possibility that the lower NIHSS scores in ACEI-treated patients might be secondary to a higher percentage of non-dominant hemispheric strokes in this group [26]. However, we found no significant difference in the preponderance of non-dominant hemispheric strokes between the 2 groups (data not reported). Infarct location, not only size, is also an important determinant of the severity of clinical deficits and our small sample size may have limited our ability to detect a difference in favor of ACEI.
We acknowledge that our study has inherent limitations imposed by its retrospective nature, non-randomization of treatment allocation and small sample size. The small number of ACEI-treated patients does not allow us to test for possible differences among the various ACEIs or dose regimens. Similarly, we cannot be certain of the duration of treatment or compliance with daily use of ACEI in our patients. We used an arbitrary cut-off for NIHSS scores to categorize stroke severity. It is possible that different cut-off values could lead to different results. Most importantly, our study lacks follow-up data regarding the effect of ACEI use on long-term outcomes since a large percentage of our patients were either enrolled in experimental neuroprotective trials or treated with thrombolysis upon presentation.
Conclusion
Our results show that pre-stroke use of ACEI is associated with milder stroke severity, as assessed by NIHSS score. Our findings need to be prospectively validated in larger-scale randomised studies, and the mechanism(s) of ACEI-mediated benefits in patients with ischemic stroke need to be elucidated.
Competing interests
The author(s) declare that they have no competing interests.
Authors' contributions
MS designed the study, collected and analyzed data, wrote the paper, and carried out critical revision of the manuscript. SS collected and analyzed data, and reviewed the manuscript. IL collected data and reviewed the manuscript. LC assisted with data interpretation and critically reviewed the manuscript. GS collected data, and carried out critical revision of the manuscript.
Pre-publication history
The pre-publication history for this paper can be accessed here:
Acknowledgements
This study was supported, in part, by grants from the Harvard Center for Neurodegeneration (Dr Selim) and the Doris Duke Charitable foundation, the NINDS (1R01NS045049-01A1) and the Dana foundation (Dr Schlaug).
==== Refs
The Heart Outcome Prevention Evaluation Study Investigators Effects of an angiotensin-converting enzyme inhibitor, ramipril, on cardiovascular events in high-risk patients N Engl J Med 2000 342 145 153 10639539 10.1056/NEJM200001203420301
Brott T Adams HP JrOlinger CP Marler JR Barsan WG Biller J Spilker J Holleran R Eberle R Hertzberg V Measurements of acute cerebral infarction: a clinical examination scale Stroke 1989 20 864 870 2749846
Schlaug G Benfield A Baird AE Siewert B Lovblad KO Parker RA Edelman RR Warach S The ischemic penumbra: Operationally defined by diffusion and perfusion MRI Neurology 1999 53 1528 1537 10534263
Selim M Fink JN Kumar S Caplan LR Horkan C Yi C Linfante I Schlaug G Predictors of hemorrhagic transformation after intravenous recombinant tissue plasminogen activator: Prognostic value of the initial apparent diffusion coefficient and diffusion-weighted lesion volume Stroke 2002 33 2047 2052 12154261 10.1161/01.STR.0000023577.65990.4E
Adams HP JrBendixen BH Kappelle LJ Biller J Love BB Gordon DL Marsh EE 3rd Classification of subtype of acute ischemic stroke: Definitions for use in a multicenter clinical trial. TOAST. Trial of Org 10172 in acute stroke treatment Stroke 1993 24 35 41 7678184
Brott T Adams HP JrOlinger CP Marler JR Barsan WG Biller J Spilker J Holleran R Eberle R Hertzberg V Measurements of acute cerebral infarction: a clinical examination scale Stroke 1989 20 864 879 2749846
Spilker J Kongable G Barch C Braimah J Brattina P Daley S Donnarumma R Rapp K Sailor S Using the NIH Stroke Scale to assess stroke patients. The NINDS rt-PA Stroke Study Group J Neurosci Nurs 1997 29 384 92 9479660
Wilterdink JL Bendixen B Adams HP JrWolson RF Clarke WR Hansen MD Effect of prior aspirin use on stroke severity in the trial of Org 10172 in acute ischemic stroke (TOAST) Stroke 2001 32 2836 2840 11739983
Schaller B Ischemic preconditioning as induction of ischemic tolerance after transient ischemic attacks in human brain: its clinical relevance Neurosci Lett 2005 377 206 211 15755527 10.1016/j.neulet.2004.12.004
Arboix A Cabeza N Garcia-Eroles L Massons J Oliveres M Targa C Balcells M Relevance of transient ischemic attack to early neurological recovery after nonlacunar ischemic stroke Cerebrovasc Dis 2004 18 304 311 15331877 10.1159/000080356
Wegener S Gottschalk B Jovanovic V Knab R Fiebach JB Schellinger PD Kucinski T Jungehulsing GJ Brunecker P Muller B Banasik A Amberger N Wernecke KD Siebler M Rother J Villringer A Weih M MRI in Acute Stroke Study Group of the German Competence Network Stroke. Transient ischemic attacks before ischemic stroke: preconditioning the human brain? A multicenter magnetic resonance imaging study Stroke 2004 35 616 621 14963288 10.1161/01.STR.0000115767.17923.6A
Rossnagel K Jungehulsing GJ Nolte CH Muller-Nordhorn J Roll S Wegscheider K Villringer A Willich SN Out-of-hospital delays in patients with acute stroke Ann Emerg Med 2004 44 476 483 15520707
Di Napoli M Papa F Angiotensin-converting enzyme inhibitor use isassociated with reduced plasma concentration of C-reactive protein in patients with first-ever ischemic stroke Stroke 2003 34 2922 2929 14605324 10.1161/01.STR.0000099124.84425.BB
Daugherty A Manning WM Cassis LA Angiotensin II promotes atherosclerotic lesions and aneurysms in apolipoprotein E-deficient mice J Clin Invest 2000 105 1605 1612 10841519
Yoshimoto T Fukai N Sato R Sugiyama T Ozawa N Shichiri M Hirata Y Antioxidant effect of adrenomedullin on angiotensin II-induced reactive oxygen species generation in vascular smooth muscle cells Endocrinology 2004 145 3331 3337 15070851 10.1210/en.2003-1583
Tummala PE Chen XL Sundell CL Laursen JB Hammes CP Alexander RW Harrison DG Medford RM Angiotensin II induces vascular cell adhesion molecule-1 expression in rat vasculature: A potential link between the renin-angiotensin system and atherosclerosis Circulation 1999 100 1223 1229 10484544
Skurk T Van Harmelen V Hauner H Angiotensin II Stimulates the Release of Interleukin-6 and Interleukin-8 From Cultured Human Adipocytes by Activation of NF-{kappa}B Arterioscler Thromb Vasc Biol 2004 24 1199 1203 15130920 10.1161/01.ATV.0000131266.38312.2e
Kranzhofer R Schmidt J Pfeiffer CA Hagl S Libby P Kubler W Angiotensin induces inflammatory activation of human vascular smooth muscle cells Arterioscler Thromb Vasc Biol 1999 19 1623 1629 10397679
Michel JB Renin-angiotensin system and vascular remodelling Med Sci (Paris) 2004 20 409 413 15124112
Francis GS ACE inhibition in cardiovascular disease N Engl J Med 2001 342 201 202 10639547 10.1056/NEJM200001203420309
Makris TK Stavroulakis GA Krespi PG Hatzizacharias AN Triposkiadis FK Tsoukala CG Votteas VV Kyriakidis MK Fibrinolytic/hemostatic variables in arterial hypertension: response to treatment with irbesartan or atenolol Am J Hypertens 2000 13 783 788 10933570 10.1016/S0895-7061(00)00262-4
Walters M Muir S Shah I Lees K Effect of Perindopril on Cerebral Vasomotor Reactivity in Patients With Lacunar Infarction Stroke 2004 35 1899 1902 15166388 10.1161/01.STR.0000131748.12553.ed
Chao CL Lee YT Impairment of cerebrovascular reactivity by methionine-induced hyperhomocysteinemia and amelioration by quinapril treatment Stroke 2000 31 2907 2911 11108747
Mitsias PD Jacobs MA Hammoud R Pasnoor M Santhakumar S Papamitsakis NI Soltanian-Zadeh H Lu M Chopp M Patel SC Multiparametric MRI ISODATA ischemic lesion analysis: correlation with the clinical neurological deficit and single-parameter MRI techniques Stroke 2002 33 2839 2844 12468779 10.1161/01.STR.0000043072.76353.7C
Tong DC Yenari MA Albers GW O'Brien M Marks MP Moseley ME Correlation of perfusion- and diffusion-weighted MRI with NIHSS score in acute (<6.5 hour) ischemic stroke Neurology 1998 50 864 870 9566364
Fink JN Selim MH Kumar S Silver B Linfante I Caplan LR Schlaug G Is the association of National Institutes of Health Stroke Scale scores and acute magnetic resonance imaging stroke volume equal for patients with right- and left-hemisphere ischemic stroke? Stroke 2002 33 954 958 11935043 10.1161/01.STR.0000013069.24300.1D
| 15949043 | PMC1175849 | CC BY | 2021-01-04 16:28:53 | no | BMC Neurol. 2005 Jun 10; 5:10 | utf-8 | BMC Neurol | 2,005 | 10.1186/1471-2377-5-10 | oa_comm |
==== Front
BMC NeurosciBMC Neuroscience1471-2202BioMed Central London 1471-2202-6-391593264110.1186/1471-2202-6-39Research ArticleSynaptic and genomic responses to JNK and AP-1 signaling in Drosophila neurons Etter Paul D [email protected] Radhakrishnan [email protected] Zaneta [email protected] Chirag [email protected] Dirk [email protected] Heinrich [email protected] Mani [email protected] Department of Molecular & Cellular Biology, University of Arizona, Tucson, USA2 Department of Brain and Cognitive Sciences, MIT, Cambridge, USA3 Department of Biomedical Genetics, University of Rochester, Rochester, USA4 ARL Division of Neurobiology, University of Arizona, Tucson, USA2005 2 6 2005 6 39 39 18 3 2005 2 6 2005 Copyright © 2005 Etter 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 transcription factor AP-1 positively controls synaptic plasticity at the Drosophila neuromuscular junction. Although in motor neurons, JNK has been shown to activate AP-1, a positive regulator of growth and strength at the larval NMJ, the consequences of JNK activation are poorly studied. In addition, the downstream transcriptional targets of JNK and AP-1 signaling in the Drosophila nervous system have yet to be identified. Here, we further investigated the role of JNK signaling at this model synapse employing an activated form of JNK-kinase; and using Serial Analysis of Gene Expression and oligonucleotide microarrays, searched for candidate early targets of JNK or AP-1 dependent transcription in neurons.
Results
Temporally-controlled JNK induction in postembryonic motor neurons triggers synaptic growth at the NMJ indicating a role in developmental plasticity rather than synaptogenesis. An unexpected observation that JNK activation also causes a reduction in transmitter release is inconsistent with JNK functioning solely through AP-1 and suggests an additional, yet-unidentified pathway for JNK signaling in motor neurons. SAGE profiling of mRNA expression helps define the neural transcriptome in Drosophila. Though many putative AP-1 and JNK target genes arose from the genomic screens, few were confirmed in subsequent validation experiments. One potentially important neuronal AP-1 target discovered, CG6044, was previously implicated in olfactory associative memory. In addition, 5 mRNAs regulated by RU486, a steroid used to trigger conditional gene expression were identified.
Conclusion
This study demonstrates a novel role for JNK signaling at the larval neuromuscular junction and provides a quantitative profile of gene transcription in Drosophila neurons. While identifying potential JNK/AP-1 targets it reveals the limitations of genome-wide analyses using complex tissues like the whole brain.
==== Body
Background
Conserved neuronal signaling pathways regulate synaptic plasticity, the ability of neurons to modify synaptic connections. "Long-term" forms of neuronal plasticity require new gene expression that results in persistent synaptic change (altered synaptic strength and morphology). Thus, long-term forms of plasticity may be perturbed, in a variety of model systems, by protein synthesis inhibitors, or manipulation of either specific signaling kinases or critical downstream transcription factors [1-4].
A major requirement in long-term plasticity for the ERK/MAPK (extracellular signal-regulated kinase subfamily of mitogen-activated protein kinases) signaling cascade acting through CREB (the cAMP response element binding protein, a basic leucine zipper – bZIP – transcription factor) has been supported in diverse experimental paradigms [1,4,5]. Activation of CREB has been reported to enhance long-term memory in Drosophila and rodents, and long-term facilitation in the sea slug Aplysia [6-11]. Normal ERK signaling is required for hippocampal LTP formation, for BDNF-induced forms of structural plasticity, as well as for several forms of learning and long-term memory [12-15]. For example, ERK activation is necessary for the formation of conditioned taste aversion and spatial learning in rodents, and blockade of ERK signaling affects long-term, but not short-term, fear conditioning [16-18]. In addition, ERK regulates synapse plasticity in flies and LTF and memory in Aplysia [19-21].
Many "immediate-early genes" (IEGs), including members of the AP-1 family of transcription factors (heterodimeric transcription factor complexes consisting of the bZIP proteins Fos and Jun), are induced in response to diverse stimuli in the brain, such as electrical stimulation, stress, psychotropic drugs, novel experience and spatial learning [22,23]. Induction of AP-1 mRNA in neurons requires CREB activation [24]. Although roles have been established for AP-1 components ΔFosB and c-fos in synaptic and behavioral plasticity [25-28], the specific mechanisms and signal transduction pathways that initiate and sustain AP-1 dependent neuronal processes have yet to be elaborated [29]. For instance, the requirement for kinase-mediated modification of immediate-early transcription factors remains poorly studied in the context of neural plasticity, and early-response genes downstream of these critical IEGs, have not yet been identified.
While the majority of plasticity studies have focused on CREB and the ERK signaling cascade [5,30,31], recent studies, especially of other MAP-kinase family members [32], lead to a broader view of the molecules involved in neuronal plasticity and memory formation. The JNK/MAPK (Jun N-Terminal kinase) signaling cascade and AP-1 proteins have recently been shown to play critical roles in long-term plasticity and memory formation in mammals [22,26-28,33]. Similarly, p38/MAPK mediates memory formation in the rat hippocampus [34] and both short- and long-term synaptic depression in Aplysia [32].
As suggested by its importance in the control of processes underlying cocaine addiction [29], new data indicate that AP-1 may widely influence transcriptional events that underlie long-term synaptic plasticity. AP-1, under regulation by JNK, positively regulates both synaptic growth and synaptic strength at the Drosophila neuromuscular junction (NMJ) [35]. At this synapse AP-1 shows a wider range of influence than CREB whose effects at the same synapse are limited to controlling synaptic strength [36]. Thus, while neural induction of Fos and Jun together is sufficient to cause increases in synaptic size and efficacy at the NMJ, CREB activation, apparently dispensable for synaptic growth, is only essential for AP-1 induced changes in synaptic strength [35-37]. In other neural contexts, the exact roles and mechanisms of AP-1 and JNK signaling in long-lasting forms of plasticity are largely unknown [26-28,33,35,38]. While many genes regulated during CREB and ΔFosB (a splice variant of the FosB gene) mediated cocaine reward [39] have been recently identified, this study identified exclusively late-response genes whose expression levels were altered following 5-days to 8-weeks of either ΔFosB and CREB overexpression in the nucleus accumbens, or cocaine injection.
Here, we address two outstanding questions regarding JNK and AP-1 function in synaptic change. First, using temporally controlled induction of a JNK-activating kinase in the fly nervous system, we address synaptic consequences of JNK activation at the Drosophila neuromuscular junction. Second, using DNA microarray or SAGE (serial analysis of gene expression) to profile neuronal transcripts in control and experimental animals, we identify groups of neuronal genes potentially regulated by either: a) JNK, or b) AP-1 in the fly CNS within 6 hours of pathway activation. Some of these changes were confirmed by quantitative real-time RT-PCR (Q-PCR), including CG6044, that was previously identified in a screen as a potential gene required for normal memory formation in Drosophila [40]. We found five genes are responsive to the progesterone-related steroid RU486 commonly used for temporal control of GAL4-restricted transgene expression in Drosophila [35,41]. In addition, the mini-white gene, a common marker used in most Drosophila transgenes, is induced strongly by AP-1 and JNK signaling. These observations and their wider implications are discussed below.
Results
Neuronal JNK activation triggers synaptic growth
To assess the effect of neuronal JNK activation on synaptic change we expressed an activated JNK-kinase – hemipterous (hepact) [42] – in the nervous system and analyzed associated phenotypic consequences. Chronic overexpression of hepact with neuronal GAL4 drivers (C155, C380, D42 and OK6) caused lethality ranging from late embryonic to early 2nd instar larval stages.
In order to overcome this deleterious effect, we drove expression using the inducible GeneSwitch GAL4 (GS-GAL4) system to express hepact acutely in postembryonic neurons [35,41]. Age-selected larvae were exposed to the inducing ligand RU486 between late 1st instar and early 2nd instar larval stages and allowed to develop to climbing 3rd instar larval stage. Neural overexpression of hepact resulted in a 30% increase in bouton number at the larval NMJ compared to the non-RU486 fed sibling controls (Figure 1). Changes in synapse size may not be attributed to the effect of RU486 since wild-type animals fed the steroid did not show a significant change in bouton number (Figure 1C). Thus, postembryonic activation of JNK signaling in the CNS leads to synaptic growth at the larval motor synapse.
JNK activation disrupts transmitter release and alters presynaptic composition
To evaluate the effect of JNK activation on transmitter release and postsynaptic response, we measured both spontaneous and evoked junctional potentials with and without JNK activation in postembryonic CNS. Increased hepact resulted in an unexpected 60% decrease in the amplitude of excitatory junctional potentials (EJP) (Figure 2A, B). Increased neuronal JNK signaling also decreased the amplitude of spontaneous responses by ~50% (Figure 2C). The quantal content of presynaptic transmitter release shows a 30% decrease when hepact is overexpressed in postembryonic neurons (Figure 2D). Therefore, while sustained postembryonic JNK activation in the CNS triggers synaptic growth, the strength of the synapse is reduced. A potential cellular mechanism that underlies this reduction in quantal content was suggested by immunohistochemical analyses of NMJs in JNK-activated animals.
Presynaptic proteins including synaptic vesicle proteins, Synaptotagmin, Csp and antigen(s) recognized by anti-HRP were substantially decreased when JNK was activated in motor neurons (Figure 3A–F). Levels of Syt staining were reduced by 45%, Csp by 30% and anti-HRP by 50% (Figure 3G). In contrast, postsynaptically enriched proteins, Fasciclin II and Dlg, do not show any change in intensity.
Because JNK has been implicated in axonal transport, we asked whether transport defects could possibly explain how JNK alters presynaptic composition [43,44]. Defects in axonal cytoskeletal assembly or anterograde axonal transport cause accumulation of Syt positive puncta on axonal tracts [44]. Such organelle jams were not present in axonal tracts of larvae overexpressing hepact. The nerves were indistinguishable from control animals, indicating that visible axonal transport defects are not present after overexpression of hepact (data not shown). For the purposes of this study, these results simply point to potential effects of JNK activation in the CNS that go beyond its previously defined role as a positive regulator of AP-1 and, thereby, of synaptic growth and synaptic strength [35].
Genome-wide screen to identify JNK targets in neurons
To identify transcriptional targets of JNK signaling in the nervous system we performed a genome-wide analysis of JNK-responsive genes in the Drosophila larval CNS using Serial Analysis of Gene Expression (SAGE). SAGE is an approach that has been extensively used in analyzing expression changes in cancer cells and other disease states as well as to analyze gene expression in the Drosophila embryo and developing eye [45-47]. SAGE is based on generating unique 14 bp tags at a defined position in almost every transcript and, following random sequencing of some 20,000 cDNAs, analyzing the frequency at which each tag (and hence each transcript) occurs in a sample RNA. We used SAGE to a) profile gene expression in the fly central nervous system; and b) identify transcriptional targets of neuronal JNK signaling. To identify early mediators of synaptic change, we analyzed RNA expression 6 hours after JNK activation. We induced a 6-hour burst of neural hepact expression in third-instar larval nervous systems using RU486 feeding to induce transcription mediated by neural GS-GAL4. In experimental ElavGS-GAL4-hepact animals, we confirmed that JNK signaling was significantly activated by performing the following tests. Quantitative PCR demonstrated a 16-fold induction of hep mRNA in larval CNS after hepact overexpression (P < 0.001) (Figure 5). RNA in situ hybridization of RU486 exposed larval CNS showed a marked increase in hep mRNA localization in the entire larval CNS (Figure 4A). Immunostaining with an antibody specific for phsophorylated JNK showed induction of hepact mRNA leads to activation of JNK (Figure 4B). Finally, we observed that downstream gene expression of a JNK target gene puc occurs after hepact induction in the CNS. puc mRNA is induced nearly 3-fold (P < 0.01) by Q-PCR analysis (Figure 5) and in an "enhancer trap" lacZ line we were able to visualize puc promoter activity in the larval CNS (Figure 4C). Thus, our protocol to stimulate neural JNK is sufficient to induce an established downstream target of JNK signaling. Exposure of identically cultured, wild-type animals to RU486 did not lead to induction of hep or puc mRNA or lead to activation of JNK (data not shown). Hence, changes observed between RU486 treated and untreated animals should be largely attributable to JNK signaling in the larval CNS.
We sequenced approximately 20,000 tags from individual libraries prepared from dissected larval nervous systems of either control or hepact expressing animals. About 9900 unique SAGE tags represented in these libraries were associated with specific genes/genomic sequences using either a database containing predicted tags of all genes annotated by the BDGP [48] or BLAST searches to identify other transcription units [45,47]. Approximately 10% of tags with less than 3 matches to the genome mapped to regions with no predicted gene. About 12.5% of all tags did not match the genome probably due to polymorphisms, errors in sequencing or possible gaps in the published Drosophila genome sequence. Comparison of the top 60 expressed genes in the CNS SAGE library to embryonic and photoreceptor SAGE libraries revealed that while 32% of these genes are highly expressed in all 3 tissues such as the cytoskeletal protein betaTub56D, 37% are enriched in the nervous system like the translation elongation factor Ef1alpha100E (see Figure 6). Such comparisons could prove useful for understanding transcriptional regulation and other processes in different tissues (see Discussion).
Potential JNK-target genes were identified by comparing the relative representation of specific tags in control and hepact expressing nervous systems (Figure 4D). A tag was considered up or downregulated when present 3 or more times in a given library and changed at least 3-fold between the two libraries. By these criteria, 346 tags were increased while 271 were decreased following JNK induction. Of these, 25 were "upregulated" and 32 "downregulated" more than 8-fold. Approximately 50% of the induced or repressed tags in the hepact library mapped to genes that fell into different functional classes, ~35% of these tags mapped to genes that have no predicted function and ~10% mapped to parts of the genome without any predicted genes (see Figure 4E, F).
To determine whether predictions of SAGE could be confirmed by more careful single gene analyses, we performed Q-PCR to measure relative levels of expression of selected candidate JNK-target genes in control and JNK-induced nervous systems. We selected 61 candidate genes for such Q-PCR verification based on: 1) an abundance of tags for that gene in the hepact induced library (9 genes); 2) an interesting known function for the gene (12 genes); 3) presence of AP-1 binding sites in the promoter region of the gene (11 genes); and 4) random selection of genes that did not fall into the above criteria (29 genes). From the 61 genes so examined, 15 that showed induction in at least two independent RT-PCR analyses were analyzed further, namely more extensive Q-PCR analyses using RNAs from 5 independent JNK-induction experiments. In the end, only three genes showed consistent JNK-responsiveness (P < 0.05) (Figure 5). white showed consistent and robust increases in mRNA levels, whereas appl and cher showed smaller magnitude inductions. To test if some of the candidate JNK target genes were robustly regulated in subsets of neurons, but diluted out in the Q-PCR analysis, we examined the expression of RNA in the larval CNS using in situ hybridizations with probes for several candidate mRNAs. We did not see a clear increase in expression in any of the putative target genes in hepact expressing larval CNSs other than white (Figure 10; also see microarray screen validation).
The small number of SAGE-predicted JNK target genes confirmed by RNA in situ and Q-PCR analyses was difficult to explain without multiple repetitions for which SAGE, being expensive and time-consuming, is not ideally suited. Also, we speculated that genes expressed at lower levels than those identified by SAGE may be true JNK/AP-1 target genes. To test and further these considerations, we used a different genomic approach – oligonucleotide microarrays – to search for neuronal AP-1 target genes.
Whole-genome microarray screen to identify direct AP-1 target genes in the nervous system
We performed comprehensive analyses of transcript levels in fly heads using Affymetrix Drosophila Genome1 GeneChip arrays representing the entire annotated genome at the time of its release (~13600 unique genes). An outline of the microarray screen design is illustrated in Figure 7. The analysis compared mRNA levels, with or without AP-1 induction using the same conditional GS-GAL4 strategy described for the previous SAGE analysis. After treatment with RU486 for 6 hours to induce fos and jun we consistently achieved, respectively, ~2.5- and 9-fold induction of fos and jun transcripts in fly heads (quantified by Q-PCR in 1–3 day old adult flies, Figure 8A). Untreated control animals showed no significant difference of either gene when levels were compared between age-matched siblings from the same experiment (average untreated change).
Each array "experiment" included sibling flies split into three groups: group A was the experimental (AP-1 induced) population (and groups B and C were independent controls). Thus, each experiment typically allowed transcript levels (normalized hybridization signals) to be compared between experimental and control samples ("A/B" or "A/C" comparisons), and between two identically treated controls (a "B/C" comparison). Candidate AP-1 responsive transcripts would be identified as those with "A/B" and "A/C" ratios significantly different from control "B/C" ratios. This experimental design was useful because hybridization signals for some mRNAs varied significantly more than others and could potentially confound a more straightforward analysis. Through 5–7 repetitions of this basic experiment, we obtained 12 independent experimental versus control ratios, and 5 control-control ratios from which means, variances and SEMs could be determined. Sibling, age-matched controls used in each experiment ensured that genetic background, which can have a large effect on transcriptional variance [49], was not a confounding factor in our analyses.
Based on analyses of 19 hybridizations we established that basic elements of the array technology, probe labeling, hybridization and scanning, were working efficiently and reproducibly (See Methods for a complete description). Microarray hybridization data were passed through three statistical filters to select the most promising AP-1 responsive genes (Methods). Filter 1: We asked that the average ratio of hybridization signal from AP-1 induced versus control mRNA hybridization was significantly (P < 0.01) different from 1.0 by Student's t-test. Filter 2: We asked that the signal ratio be greater than 1.2. Filter 3: Through analysis of variation observed in identical control-control comparisons, we ensured that genes passing filters 1 and 2 did not show wide variability, for instance based on physiological states of the flies.
Using filter 1: of the ~5200 genes considered for analysis (those with relatively strong and specific hybridization signals), 269 showed altered expression after AP-1 induction, with a significance of P < 0.01 (Student's t-test). Strikingly, 167 genes showed significant upregulation while only 102 were downregulated, a skew consistent with AP-1's expected role as a transcriptional activator. 269 candidates, at P < 0.01, is substantially larger than predicted by random chance (52 genes – 0.01 × 5200 genes). However, when a second filter – a requirement that the signal ratio modulus be greater than 1.2 – was applied, the number of candidates dropped to 115. Though small, such signal ratios could correspond to higher mRNA ratios and have been reported as meaningful in previous microarray experiments. Filter 3, to eliminate "variable" genes, trimmed the list of candidate genes that respond consistently to AP-1 overexpression in the fly head to either 4 (P < 0.01) or 16 (P < 0.05, listed in Figure 9) for which "A/B" and "A/C" ratios were significantly greater than control "B/C" values by Student's t-test.
An internal control for the array screen and analysis was provided by Drosophila Jun (Jra), whose mRNA was experimentally induced. We found that jun ranked highest once all three filters were applied and showed robust induction with an average log2 treated expression ratio of 1.26 (P = 5.9e-11) and an average untreated expression change of only -0.07 (Figure 8A). In contrast, fos did not pass these stringent filters although we consistently observed an average 2.3-fold increase in fos transcript levels by quantitative RT-PCR (Figure 8A). This discrepancy may arise from either of two limitations: a) that fos is a low-abundance transcript in the fly head, below the threshold for quantifiable gene expression change detection using Affymetrix GeneChip arrays; or b) the fos probe on this particular array may not perform reliably [50], perhaps hybridizing to other non-specific RNA probes. Many gene probes could have similar problems; indeed, other transcripts may exhibit altered expression levels beyond the scope and sensitivity of this assay.
We searched promoters (sequences 3 kb upstream of the translation start sites) of the top 15 candidate AP-1 responsive genes for conserved AP-1 or CREB binding sites and compared their frequencies of occurrence in this group with frequencies observed in a control group of 15 genes that appeared insensitive to AP-1 induction. This analysis revealed no significant enrichment of CREB or AP-1 binding elements in promoters selected based on the microarray experiments (data not shown).
Microarray screen validation
A major task after initial microarray screening has been completed is confirmation of candidate gene transcript level changes using secondary, independent tests for gene expression. Although the frequency of false-positives is substantially reduced through repetition, a subset of observed expression differences should be validated by other methods.
To confirm positives, a subset of the most robustly changing AP-1-responsive genes, exhibiting significant up- or down-regulation by microarray analysis, were selected as candidates for real-time quantitative RT-PCR validation using gene specific primers (see Figure 9). 12 genes chosen from the group of 15 top candidates mentioned above, in addition to more than 30 genes from outside this stringent set – those with very low "P" values or specific predicted biological functions – were selected for these more careful confirmatory experiments. Increases in transcript levels following AP-1 overexpression, detected by Q-PCR, for fos, jun and 2 confirmed candidate genes (white and CG6044) are shown in Figure 8A. All mRNA levels are normalized to the control gene rp49. Transcript levels for a second control gene, gapdh1, are shown to demonstrate its levels do not change significantly by either AP-1 induced versus control ("A/B", "A/C") or control-control ("B/C") comparisons.
Five uncharacterized genes (CG2016, CG11191, CG15438, CG5853 and CG3348) were confirmed by Q-PCR to be consistently altered in RU486-treated, AP-1 induced samples (Figure 9, Figure 8B – data for CG5853 and CG3348 not shown). In addition, overexpression of fbz with RU486 treatment also caused a similar change in transcript levels of these 5 genes. When treated with the steroid, wild type flies and all other transgenic lines tested showed consistent alterations of these 4 mRNA transcripts in the head, suggesting they are hormone-responsive genes in the fly.
As in the larval CNS, Q-PCR experiments confirmed white gene induction in the adult fly head. white transcript levels are significantly increased in the head following AP-1 overexpression in the brain (Figure 8A, Figure 9, Figure 10). Further Q-PCR experiments demonstrated white transcripts are increased to an even greater extent when hepact is induced in combination with AP-1 or by itself in the adult nervous system (Figure 10A), although its levels are not increased to the degree seen in the larval CNS (Figure 5). white is not induced when fbz is overexpressed, nor in wild type flies treated with RU486. Only primers designed to the 3' portion of the white transcript showed altered levels (Figure 10A), which is consistent with the background strain used in all the experiments(w1118). This strain lacks the 5' portion of the white gene locus [51], yet still contains sequence for and expresses the second through fifth exons (data not shown) that are induced in response to JNK signaling in the fly head. The same transcriptional induction profile is observed in ElavGS-GAL4-hepact flies with the wild-type (w+) copy of the white gene on the X chromosome as well as in a white null (w11E4 [52]) background (data not shown). This suggests white induction occurs via the mini-white cassette present in pUAST transgenes. RNA in situ hybridization experiments confirmed the increase in mini-white transcript levels in the CNS of larvae in which hepact has been induced (Figure 10B). Increases were also observed in larvae overexpressing AP-1 (data not shown), albeit with smaller magnitude changes consistent with our findings from Q-PCR analyses in the adult head (see Figure 10A).
CG6044 induction following AP-1 overexpression was also confirmed in independent Q-PCR experiments (Figure 8A). Consistent increases in transcript levels were observed in all AP-1 overexpressing heads but not in treated fbz or wild type heads (data not shown). The induction observed by quantitative RT-PCR and microarray experiments was not reflected in follow-up in situ experiments; however, this is likely because the small magnitude increases in transcript levels observed by other means (1.2-fold – microarray; 1.4-fold – Q-PCR) are below the detection range for this method.
Discussion
This extensive study makes three contributions: (A) it demonstrates unexpected and novel interactions between JNK and cellular processes that underlie synapse plasticity; (B) by SAGE analyses, it provides a genomic profile of mRNAs expressed in the fly larval nervous system; (C) it presents two large-scale genomic approaches to identify JNK and AP-1 targets in the fly CNS providing useful data pertinent to JNK/AP-1 signaling in neurons as well as to genomic analyses in the Drosophila nervous system.
Effects of JNK activation in postembryonic motorneruons
The immediate-early transcription factor AP-1 positively regulates both synapse size and synapse strength at the Drosophila larval NMJ [35]. While JNK signaling is necessary for the effect of AP-1 on synapse structure and function, it is not clear whether JNK signaling is sufficient for synaptic change. We show, first, that activation of JNK in post-embryonic neurons leads to significant synaptic alterations; second, that these alterations are inconsistent with JNK functioning solely through AP-1. Our finding that activation of JNK signaling leads to an increase in synapse number but decreases synapse strength indicates that JNK activates not only AP-1, a positive regulator of growth and strength, but also a pathway that negatively influences synaptic strength.
The neural transcriptome, and its regulation by JNK and AP-1
The ability of SAGE to evaluate absolute expression levels of gene transcripts enables relatively facile, quantitative, profiling of gene expression in any given tissue (or RNA source). Given the intense interest in Drosophila neurobiology, a previous painstaking sequence analysis of some 1000 cDNAs from a fly brain cDNA library provided useful new information on the neural transcriptome [53]. The analysis presented here, following sequencing of about 20,000 ESTs from two independent brain libraries, substantially extends the previous study. The use of this resource is demonstrated by our simple survey of highly expressed neuronal RNA-binding proteins, potentially involved in important neural-specific, post-transcriptional functions such as translational repression, mRNA transport or RNA editing. 10% of the 60 most highly expressed (non-ribosomal) mRNAs in nervous system encode RNA-binding proteins, 2 of which are enriched in neurons versus embryonic tissue. A significant fraction of these (3/6) have conserved homologs recently found on RNA granules, organelles containing translationally repressed mRNAs which are actively transported to synaptic sites [54]. We have recently begun functional analyses of some of these RNA-binding proteins. Similarly, we anticipate that identification of tissue-specific genes could provide unanticipated launch points for investigation into their cellular functions.
Given the evidence to indicate wide effects of AP-1 and JNK on synaptic properties, we searched for AP-1 and JNK-target genes using both SAGE and microarray approaches to determine effects of JNK and AP-1 signaling on neuronal gene expression. Of the two approaches, microarray analysis, being dependent on parameters such as hybridization and labeling efficiencies that vary among individual transcripts, is not ideal for quantitative analyses as outlined in the previous section. However, it provides information on transcripts with low to moderate levels of expression, is fast, and allows multiple iterations of each experiment at a small cost relative to SAGE.
In order to identify early transcriptional targets, most likely to link JNK and AP-1 activation to synaptic change, we used the steroid-inducible GAL4 system, an increasingly popular strategy to achieve conditional, tissue-specific transgene expression in Drosophila [35,41,55,56]. SAGE-derived transcript profiles of RNA extracted from whole larval CNSs showed several potentially significant targets. However, very few were confirmed by secondary low-throughput, gene-specific analyses. Microarray-derived transcript profiles of adult head mRNA showed similar results. Several statistically significant targets of AP-1 signaling were initially identified; however, few were confirmed by carefully controlled application of the most commonly used transcript-specific analyses (quantitative RT-PCR and in situ RNA hybridizations). While the implications of these results for neurogenomics are briefly discussed in the next section, we first consider the "positive" genes identified by SAGE and microarray screens.
Quantitative RT-PCR validation of the generated SAGE data resulted in the identification of 3 genes, cher, appl and white, which were consistently upregulated following JNK activation in the larval CNS. Though we were unable to evaluate induction of cher and appl by RNA in situ hybridization, white showed robust increases by this method as it did by Q-PCR. A total of seven expression changes identified in the microarray screen were verified by Q-PCR analysis; remarkably, five turned out to be genes responding to RU486 treatment itself rather than to consequent AP-1 induction. These steroid-responsive genes may be of significant biological interest. However, from our point of view they serve primarily to: a) further establish the bonafides of our experimental and analytical protocols; and b) as a useful caution for Drosophilists and others using the steroid-inducible conditional expression system. The remaining two confirmed AP-1 target genes were w hite, also identified in the SAGE screen but shown eventually to be expressed from the P-element associated mini-white locus, and CG6044. Of potential significance, CG6044 has been implicated in olfactory associative memory [40].
AP-1 responsiveness of CG6044 was verified in Q-PCR validation experiments (Figure 8). The gene was previously found in a mutational screen for putative memory genes required for normal olfactory conditioning in Drosophila [40]. In addition, it is one of the few genes from the list of likely AP-1 targets (listed in Figure 9) that has a conserved AP-1 binding site within 500 base pairs of its translation start site. It is therefore a promising candidate warranting further investigation into the role it plays in synaptic plasticity and memory formation.
Lessons and limitations
It appears unlikely, if not inconceivable, that the 4 probable downstream genes enumerated above could mediate the demonstrated effects of AP-1 or JNK induction on motor-synapse properties. Thus, the genomic approaches we have followed, while informative, have likely not led to the identification of JNK/AP-1 targets that link these signals to synaptic change. One possible interpretation, that the experiments were technically flawed, appears to be ruled out, not only because internal controls (Jun, Hep, Puckered and steroid-responsive genes) were identified in the screens, but also because various standards for microarray hybridization data and SAGE library complexity were evaluated and shown to be well within the technically optimal range. Thus, we are left with the second interpretation, that analysis of whole-brain mRNA may not allow targets of signaling pathways to be unambiguously identified. A major issue is likely to be cell-type heterogeneity within the brain. If different subsets of neurons show substantially different genomic responses to JNK/AP-1 (including the absence of a response), then altered expression of the meaningful JNK/AP-1 targets in a subset of cells may be diluted by the large background of mRNA deriving from other neuronal types.
At a conceptual level, Barolo and Posakony have nicely articulated the concept of "activator insufficiency" and the need for cooperative activation of multiple transcription factors for turning on transcriptional pathways governing developmental processes [57]. Considerable evidence argues that neurons are a diverse class of cells with a range of distinct transcriptional ground states. For example, cell-type-specific binding of CREB to known target gene promoters has been shown in various cell types under basal and stimulated conditions [58]. Similarly, the response of different neuronal populations to TGFβ has been shown to be highly context dependent and to derive from variations in expression of specific TGFβ insensitive transcription factors [59]. Thus, genomic analyses when applied to whole nervous systems may have significant intrinsic limitations.
Nevertheless, some conserved downstream genes may still be revealed [60-63]. For instance, the steroid hormone, RU486, used to induce transgene expression in our experiments presumably activates a set of hormone-responsive genes in a large subset of neural cells. However, for incisive mechanistic analyses for which Drosophila is so convenient, we suggest that genome-wide screens described to study signaling responses in the nervous system be applied with specific refinements, such as emerging methodologies to prepare sufficient mRNA from a homogeneous population of cells in which biological function of these signaling pathways have been evaluated [64]. Various GFP transgene lines should make it possible to sort specific cell populations prior to genomic screens to identify transcriptional targets.
The availability of new genetic and molecular tools and refined functional genomic approaches should result in continued understanding of how kinases and transcription factors regulate molecular changes that occur in the Drosophila nervous system, as well as intrinsic flexibility and constraints of these signaling pathways.
Conclusion
This study revealed unexpected relationships between JNK signaling and synaptic plasticity in Drosophila that are inconsistent with a role for JNK acting solely through AP-1 to affect strength of the synapse. It also presents a profile of the transcriptome of the larval nervous system and, while providing potential transcriptional targets of JNK and AP-1 signaling in neurons, points out the pitfalls of genome-wide analyses in complex tissues such as the whole fly nervous system.
Methods
Fly strains and genetics
We used the following strains: wild type (Oregon R; D. Brower); GAL4-responsive UAS-hepact (M. Mlodzik), UAS-fbz, UAS-fos, UAS-jun (M. Bienz), puc-lacZ line – puce69(A. Martinez Arias); neural GAL4 lines – C155, C380, D42 and OK6 were from C. Goodman, V. Budnik, G. Boulianne and B McCabe, respectively; ElavGS-GAL4 line was from T. Osterwalder and H. Keshishian.
Postembryonic and acute induction in neurons
Induction in larvae
All animals were generated by crossing males homozygous for UAS-transgenes (or wild-type males) with virgin females homozygous for the ElavGS-Gal4 driver. All animals were raised at 25°C and parents transferred to a new vial each day for age-selection of larval instar stages. For postembryonic induction, larvae in vials that should contain a majority of late 1st instar-early 2nd instar were transferred into a standard vial containing 0.015 mg/ml RU486 (Sigma) for 48 hrs before climbing third instar larvae were selected for further analysis. Control animals were exposed to food containing only 4% ethanol (same as treated) and analyzed accordingly. For acute induction in 3rd instar larvae, age-selected larvae were transferred to a 1.5 ml sample tube containing 0.5 ml of 3 mg/ml RU486 for 2 min, before they were washed and transferred into a standard vial containing 0.015 mg/ml RU486 (Sigma) for 6 hrs before the CNS was dissected for further analysis.
Induction in adults
As in larvae, all animals were generated by crossing males homozygous for UAS-transgene constructs (or wild-type males) with virgin females homozygous for the ElavGS-GAL4 driver. Progeny reared at 25°C were aged to be 1–3 days old at time of treatment. 16–64 hour old adults were starved for 8 hours in a Tupperware container filled with desiccant to keep the humidity level at ~16% and ensure ingestion of the treatment medium. Flies were then split into separate bottles and fed for 6 hours. Each treatment consisted of ElavGS-GAL4-UAS flies handled identically (aged and starved in the same bottle) except RU486 was added to the sucrose fed to the experimental group. Experimental animals were fed on a kimwipe soaked with RU486 in 2% sucrose at a final concentration of 0.04 mg/ml, taped to the bottom of a large, dry, empty bottle. Sibling control flies of the same genotype were fed sucrose alone.
Immunostaining
Larvae were raised at 25°C after postembryonic induction of UAS transgenes, dissected, stained with anti-Syt antibody and mounted. Bouton number was counted from projections of confocal sections at 60X magnification. Boutons at segment A2 in muscle 6 and 7 were manually counted without knowledge of the genotype (blind counting), using Metamorph imaging software. No significant difference in muscle surface area, measured using a drawing tool in Metamorph was observed in the different genotypes. To quantify levels of synaptic proteins synapses labeled with specific antibodies; anti-syt, anti-csp, anti-HRP, anti-fas II, anti-dlg, were identically imaged for control and induced animals and the average pixel intensity of terminal boutons (3–4) was measured and analyzed. To quantify organelle accumulation on axons after hepact induction, larval segmental nerves were imaged at high resolution using a cooled charge-coupled device camera (Princeton Instruments) and Metamorph imaging software (Universal Imaging). After background subtraction, images were analyzed for organelle jams and compared with control animals.
Electrophysiology
All electrophysiological recordings were made from muscle 6 within A2, with the larval preparation immersed in a low volume of the HL3 saline with 1 mM Ca2+. Electrophysiology was performed as described previously. In all experiments, the CNS was gently removed to prevent endogenous motor firing. Motor nerves were stimulated with glass-tipped suction electrodes. For intracellular recordings, electrodes pulled from borosilicate capillary tubes were backfilled with 3 M KCl, yielding resistances of 6–10 MΩ. To ensure good recordings, preparations with resting potentials more positive than -60 mV were discarded. For recording excitatory junctional potentials (EJPs), an isolated pulse stimulator (A-M systems, Everett, WA) was used to deliver 1 msec pulses at a frequency of 1 Hz to elicit an evoked response. All recordings were acquired with an axoclamp 2B amplifier in conjunction with pClamp 6 software (Axon Instruments, Foster City, CA). The EJP amplitude for each preparation was determined from an average of 15 consecutive evoked responses. For quantifying mini frequencies, the number of mEJPs occurring consecutively within 30 sec was counted for each preparation. The mEJP amplitude for each preparation was determined from an average of 30 consecutive mEJPs. At least 5 animals were analyzed for each genotype. For each animal examined that was exposed to RU486 treated food, we examined control animals and expressed the quantal content of transmitter release as a percentage of control.
Serial Analysis of Gene Expression (SAGE)
SAGE was performed as previously described [45,47]. Briefly, polyA mRNA from 50 CNSs dissected out of drug treated 3rd instar larvae was purified with dynabeads mRNA direct kit (Dynal). Double-stranded cDNA was synthesized on the beads and digested with the anchoring enzyme (NlaIII; NEB). After linker ligation, digestion with the tagging enzyme (BsmFI, NEB), and ligation of the ditags, PCR amplification (29 cycles) was carried out with 20% of the ligation product as template. The 100 bp PCR products were purified and submitted to a secondary PCR (10–12 cycles) with biotinylated primers to generate enough material for the concatemerization. After NlaIII digestion, the released ditags were purified by polyacrylamide gel electrophoresis and subsequently incubated with 100 ul of Dynabeads Streptavidin to eliminate any remaining biotinylated linkers. Concatemerization was carried out for four hours. Concatemers were cloned into the SphI site of pZero1 (Invitrogen), and resulting colonies were screened for inserts by PCR and submitted for sequencing. All sequencing reactions and SAGE tag generation was performed at Agencourt Inc. (Boston).
Analysis of SAGE Data and Annotation of SAGE Tags
Sequenced SAGE concatemers were analyzed using the SAGE2000 program obtained from The Johns Hopkins University (see also [65]). The database linking SAGE tags to data of the Berkeley Drosophila Genome Project was built using datasets downloaded from the BDGP site [66] and extracting the 10 bp sequence downstream of the 3'-most CATG site. These putative tags were linked to the GadFly site of the corresponding gene. Annotation of experimental data was performed using Microsoft Access to link the experimental dataset and the Tag annotation database [45,47].
Microarray analysis
Total RNA was extracted from 200–300 heads of RU486-treated (AP-1 induced) and untreated control ElavGS-GAL4-AP1 flies (w1118;UAS-fos/+;UAS-jun/ElavGS-GAL4) using the RNeasy kit (Qiagen). 5 ug of total RNA was used as a starting template for 19 microarray hybridizations (7 treated and 12 untreated RNA samples). Two sets of control flies were included for analysis in five of the seven experimental AP-1 overexpression treatments used for the microarray hybridizations. Transcript quantification was performed with Affymetrix Drosophila Genome1 GeneChip [67] arrays using biotinylated cRNA targets prepared according to standard Affymetrix protocols by the GATC Affymetrix Core Facility at the University of Arizona [68]. Hybridized arrays were scanned using Affymetrix MicroArraySuite software as described in the manufacturer's protocol. All hybridizations were normalized with a global scaling factor of 500 so that transcript levels could be compared directly. Text files containing raw, normalized values were exported into Excel for further analysis.
Internal control, 3'-5' probe signal ratios (a measure of how well the biochemical reactions went prior to hybridization of the biotinylated probe to the oligonucleotide array) were within the range recommended by the manufacturer for all hybridizations. R2 values for all comparisons of control versus control samples from the same experimental group were high (≥ 0.97).
Between 41% and 49% of all genes were scored present or marginal on each of the arrays by MicroArraySuite. In order to avoid spurious data, only the 5188 genes present or marginal in all 7 AP-1 induced samples were considered for further analysis (~38% of all probes on the array). Ratios between AP-1 induced and the 1 or 2 control samples from a given treatment were calculated in Excel. Fold-differences were converted to log2 values so that increasing and decreasing levels of mRNA could be compared directly. Log2 values (n = 12) were tested against the value of 0, expected if there were no change in expression, using the Student's t-test (unpaired t-test, two-sided P, samples with unequal variance estimates). The P-values accepted for our analysis (P < 0.01) therefore reflect a 99% probability that the null hypothesis (there is no difference in the expression of a given transcript in AP-1 induced samples) should be rejected.
Secondary filters to eliminate false positives and randomly fluctuating transcripts included: 1) the average AP-1 induced versus control ratio (n = 12) for a given gene had to be 1.2 or higher; 2) Log2 values for the expression ratio, comparing AP-1 induced to control signals for a given gene (n = 12), were tested against the values of control versus control ratios (n = 5), again using the Student's t-test – genes passing this statistical filter (P < 0.05) were considered to be changed beyond the dynamic nature of the transcript.
Quantitative real time RT-PCR and in situ hybridization
Larval CNS
To quantify RNA expression, approximately, 25 larval brains were dissected for each sample. PolyA mRNA was isolated using the Dynabeads mRNA direct kit (Dynal) and oligo dT-primed cDNA was synthesized with the Omniscript cDNA synthesis kit (Qiagen). The cDNA was diluted 1:5 for Q-PCR reactions performed on a Cepheid SMARTCycler using QuantiTect SYBR Green PCR kit (Qiagen). Transcript levels were determined using gene-specific primer sets (details available on request). Expression differences are shown as the average change in cycle number at which PCR product (determined by fluorescent signal) is detected as statistically significant above background. This is referred to as the crossing threshold and the more cDNA template present at the start of the reaction, the fewer number of cycles it takes to reach this point. A one-cycle difference represents a two-fold difference in starting template concentration. All transcript levels are normalized to the control gene, ribosomal protein 49 (rp49), as previously described [35].
Adult heads
Independent RNA samples were extracted as for microarray experiments for all Q-PCR comparisons. Equal amounts of total RNA (4 ug) for RU486-treated (induced) and untreated control samples were purified from genomic DNA with the DNA-free DNase kit (Ambion) prior to oligodT-primed cDNA synthesis using the Omniscript cDNA synthesis kit (Qiagen). The cDNA was diluted 1:20 with nuclease-free H2O (Invitrogen) for Q-PCR reactions performed as described above. Each PCR reaction was repeated in triplicate for 3–5 independent RNA preparations from separate RU486 treatments. Sample sets were compared using the Student's t-test as for the array analysis and only results showing a P-value <0.05 were considered statistically significant.
In situ hybridizations were performed using probes prepared with PCR DNA (400–600 bp) from primers specific for gene of interest containing T7 RNA polymerase binding site in the sense orientation and SP6 RNA polymerase-binding site in the antisense orientation. RNA probes were labeled with DIG and visualized using either NBT/BCIP (blue reaction product). Larval CNSs were dissected after drug treatment and the tissue was processed using standard protocols [69].
Authors' contributions
PDE and MR conceived of and designed the AP-1 overexpression experiments. PDE performed the microarray experiments and statistical analysis; PDE and CP performed the microarray confirmation experiments. RN, HJ, MR and DB conceived of and designed the SAGE experiments. RN and HJ performed the SAGE experiments and analysis; RN and ZN performed the hepact overexpression and SAGE validation experiments. PDE, RN and MR drafted the manuscript with input from the other authors.
Acknowledgements
The authors would like to thank Charles Hoeffer and Leona Mukai for their technical assistance and discussion in developing and optimizing the AP-1 induction protocols. We thank Brian Coullahan, Kevin Kiesler and all the GATC Microarray Core Facility members at the University of Arizona – where the microarray probes were synthesized, hybridized and scanned – for their expert technical assistance. We also thank Agencourt Inc. (Boston) for their assistance with the SAGE experiments. This work was supported by grants RO1-DA15495 and KO2-DA17749 from the NIDA and the Science Foundation of Ireland Research Professorship to MR, as well as Institutional, pre-doctoral training grants (T32) 532GM08659 and (T32) AG07434-04 to the University of Arizona.
Figures and Tables
Figure 1 Postembryonic expression of hepact in larval neurons increases synaptic growth. Confocal projections of synaptic arbors show that synapse size is increased after hepact induction in postembryonic neurons (B) compared to control (A). C) A histogram representation of bouton number shows that hepact overexpression leads to a 31% increase in synapse size (P < 0.001), while exposing wild-type larvae to the inducible ligand does not cause a significant change in synapse size.
Figure 2 hepact expression leads to decreased transmitter release. A) EJP traces from larvae in which hepact expression is induced (lower trace) or control (upper trace). Expression of hepact in postembryonic neurons leads to decreases in EJP and miniature (m)EJP amplitude by nearly 50% (P < 0.01 for both) compared to control (B, C). D) Quantal content of presynaptic transmitter release is reduced by 35% after hepact induction in postembryonic neurons (P < 0.04).
Figure 3 Presynaptic protein levels decrease with hepact overexpression. Confocal projections of synapses show that levels of presynaptic protein synaptotagmin (Syt) (B) and an antigen recognized by anti-HRP (D) are reduced after hepact induction compared to controls (A, C), whereas levels of the postsynaptically enriched protein dlg is similar to control (E) after hepact induction (F). G) Quantification of fluorescent intensities show that levels of presynaptic proteins Syt, Csp and anti-HRP, are reduced by 45%, 34% and 50% respectively (P < 0.001 for all) compared to control. Levels of postsynaptically enriched proteins FasII and Dlg go not significantly change after hepact induction in postembryonic neurons.
Figure 4 Acute induction of hepact in larval neurons. A) RNA in situ hybridization using a probe specific for hep shows inducible expression of hep mRNA in third instar larval CNS. B) Western blot analysis of larval CNS protein extracts shows increased levels of activated JNK (P-JNK) after hepact induction. C) A lacZ enhancer trap line of puc shows increased lacZ expression after hepact induction in larval CNS. D) A distribution of up- and down- regulated SAGE tags comparing hepact induced and control libraries, indicates most tags are present in similar numbers in both the induced and control libraries. E) Functional classification of SAGE tags upregulated after hepact induction. Approximately 50% of tags map to genes with no known function and to regions of the genome without an identified gene.
Figure 5 Q-PCR validation of SAGE results. Quantitative comparisons of transcript levels in larval CNS RNA from RU486 treated (hepact induced) versus control samples. Values represent average cycle difference in PCR product between induced and control samples (N = 5). Each cycle change corresponds to a 2-fold difference in mRNA levels (see Methods). After induction of hepact, hep RNA levels increase 16-fold relative to control (P < 0.001); puc RNA, not identified by SAGE, is induced three-fold relative to control (P < 0.01). While SAGE targets appl and cher show induction above control RNA levels, white is induced more than 32-fold (P < 0.001).
Figure 6 Comparison of the top 60 expressed genes in larval CNS with expression profiles from embryo and photoreceptor cells identified by SAGE. The top 60 highly expressed tags from the hepact CNS control library were compared with control libraries from embryonic [45] and photoreceptor tissues (Jasper and Bohmann, unpublished data). The number of tags for each gene is indicated on the left for all three tissues and all libraries examined, normalized, to the same number – 20,000 – of total tags sequenced. Tag rankings are sorted in descending order for the control library for each tissue using hepact CNS control library as reference, after eliminating tags with more than 3 matches to the genome and excluding any ribosomal RNA binding proteins (highly enriched) and selecting only tags that mapped to an identified gene. 19/60 (green) highly expressed genes in the CNS libraries were also in the top 60 of highly expressed genes in the embryo and photoreceptor libraries while 22/60 (white) are found in the top 60 only in the CNS libraries. There were 13/60 (pink) genes found only in the top 60 of CNS and photoreceptor libraries and 6/60 (gray) genes found only in the top 60 of CNS and embryo libraries.
Figure 7 Microarray experimental design, analysis and validation. To induce AP-1 in the nervous system, 1–3 day old adult ElavGS-GAL4-AP1 flies were treated with the synthetic steroid hormone RU486 in 2% sucrose or sucrose alone for six hrs. Biotinylated RNA from heads was created and hybridized to Affymetrix Drosophila Genome1 GeneChip arrays. Gene expression changes between AP-1 induced and control samples were considered significant if they passed a statistical (P < 0.01, Student's t-test) and secondary filters looking at variance in untreated control samples from the same experiment. Validation of candidate gene expression changes was carried out using quantitative real-time RT-PCR and in situ hybridization experiments.
Figure 8 Q-PCR validation of AP-1 induction and microarray results. A) Quantitative comparisons of transcript levels in adult head RNA from RU486 treated (AP-1 induced) versus control (black) and control-control (gray) samples. Values represent average cycle difference in PCR product between samples being compared (N = 5). Positive values indicate an increase in transcript compared to unchanging reference gene, negative values a decrease. fos, jun, white and CG6044 RNA levels are increased in the fly head after AP-1 induction in the nervous system while untreated control levels show no significant difference. B) Comparisons of transcript levels in adult head RNA from RU486 treated versus control samples. CG2016, CG11191 and CG15438 levels are induced in all samples from flies fed the steroid hormone RU486 (N = 3). X-axis in panel B indicates the UAS-transgene(s) induced by ElavGS-GAL4 (AP1: UAS-fos;UAS-jun). *Average difference between samples significant at P < 0.05 (Student's t-test).
Figure 9 Top 15 candidate AP-1 responsive genes identified by microarray analysis. Genes altered following neuronal AP-1 overexpression, passing statistical (Student's t-test, P < 0.01) and secondary filters based on ratio thresholds and variance in untreated control samples from the same experiment. Arrows on left indicate directionality of expression change listed in order of magnitude from largest positive ratio (induction) on top to largest negative ratio (repression) on bottom. jun induction shows the largest expression change by these criteria. Predicted functions from Flybase [70]. Bold type highlights expression changes confirmed by Q-PCR (n.t. = not tested). Blue type highlights RU486-responsive genes. *VGA – volatile general anesthetic.
Figure 10 white transcript levels are induced when positive JNK pathway components are overexpressed in the fly nervous system. A) Quantitative comparisons of white transcript levels in adult head RNA from RU486 treated versus control samples. PCR primers designed to the 3' end of white (black), but not 5' primers (gray), show increased levels in response to AP-1 and hepact induction (N = 3). Wild-type flies exposed to hormone or flies overexpressing fbz do not show induction of white. X-axis indicates the UAS-transgene(s) induced. *Average treated difference significant at P < 0.05 (Student's t-test). B)in situ confirmation of increase in white transcript levels following hepact induction in the larval CNS.
==== Refs
Bailey CH Bartsch D Kandel ER Toward a molecular definition of long-term memory storage Proc Natl Acad Sci U S A 1996 93 13445 13452 8942955 10.1073/pnas.93.24.13445
Yin JC Tully T CREB and the formation of long-term memory Curr Opin Neurobiol 1996 6 264 268 8725970 10.1016/S0959-4388(96)80082-1
Alberini CM Genes to remember J Exp Biol 1999 202 2887 2891 10518471
Kandel ER The molecular biology of memory storage: a dialogue between genes and synapses Science 2001 294 1030 1038 11691980 10.1126/science.1067020
Thomas GM Huganir RL MAPK cascade signalling and synaptic plasticity Nat Rev Neurosci 2004 5 173 183 14976517 10.1038/nrn1346
Bartsch D Ghirardi M Skehel PA Karl KA Herder SP Chen M Bailey CH Kandel ER Aplysia CREB2 represses long-term facilitation: relief of repression converts transient facilitation into long-term functional and structural change Cell 1995 83 979 992 8521521 10.1016/0092-8674(95)90213-9
Yin JC Del Vecchio M Zhou H Tully T CREB as a memory modulator: induced expression of a dCREB2 activator isoform enhances long-term memory in Drosophila Cell 1995 81 107 115 7720066 10.1016/0092-8674(95)90375-5
Josselyn SA Shi C Carlezon WAJ Neve RL Nestler EJ Davis M Long-term memory is facilitated by cAMP response element-binding protein overexpression in the amygdala J Neurosci 2001 21 2404 2412 11264314
Barco A Alarcon JM Kandel ER Expression of constitutively active CREB protein facilitates the late phase of long-term potentiation by enhancing synaptic capture Cell 2002 108 689 703 11893339 10.1016/S0092-8674(02)00657-8
Guan Z Giustetto M Lomvardas S Kim JH Miniaci MC Schwartz JH Thanos D Kandel ER Integration of long-term-memory-related synaptic plasticity involves bidirectional regulation of gene expression and chromatin structure Cell 2002 111 483 493 12437922 10.1016/S0092-8674(02)01074-7
Alarcon JM Malleret G Touzani K Vronskaya S Ishii S Kandel ER Barco A Chromatin acetylation, memory, and LTP are impaired in CBP+/- mice: a model for the cognitive deficit in Rubinstein-Taybi syndrome and its amelioration Neuron 2004 42 947 959 15207239 10.1016/j.neuron.2004.05.021
English JD Sweatt JD A requirement for the mitogen-activated protein kinase cascade in hippocampal long term potentiation J Biol Chem 1997 272 19103 19106 9235897 10.1074/jbc.272.31.19103
Alonso M Medina JH Pozzo-Miller L ERK1/2 activation is necessary for BDNF to increase dendritic spine density in hippocampal CA1 pyramidal neurons Learn Mem 2004 11 172 178 15054132 10.1101/lm.67804
Patterson SL Pittenger C Morozov A Martin KC Scanlin H Drake C Kandel ER Some forms of cAMP-mediated long-lasting potentiation are associated with release of BDNF and nuclear translocation of phospho-MAP kinase Neuron 2001 32 123 140 11604144 10.1016/S0896-6273(01)00443-3
Adams JP Sweatt JD Molecular psychology: roles for the ERK MAP kinase cascade in memory Annu Rev Pharmacol Toxicol 2002 42 135 163 11807168 10.1146/annurev.pharmtox.42.082701.145401
Berman DE Hazvi S Rosenblum K Seger R Dudai Y Specific and differential activation of mitogen-activated protein kinase cascades by unfamiliar taste in the insular cortex of the behaving rat J Neurosci 1998 18 10037 10044 9822758
Blum S Moore AN Adams F Dash PK A mitogen-activated protein kinase cascade in the CA1/CA2 subfield of the dorsal hippocampus is essential for long-term spatial memory J Neurosci 1999 19 3535 3544 10212313
Schafe GE Atkins CM Swank MW Bauer EP Sweatt JD LeDoux JE Activation of ERK/MAP kinase in the amygdala is required for memory consolidation of pavlovian fear conditioning J Neurosci 2000 20 8177 8187 11050141
Martin KC Michael D Rose JC Barad M Casadio A Zhu H Kandel ER MAP kinase translocates into the nucleus of the presynaptic cell and is required for long-term facilitation in Aplysia Neuron 1997 18 899 912 9208858 10.1016/S0896-6273(00)80330-X
Koh YH Ruiz-Canada C Gorczyca M Budnik V The Ras1-mitogen-activated protein kinase signal transduction pathway regulates synaptic plasticity through fasciclin II-mediated cell adhesion J Neurosci 2002 22 2496 2504 11923414
Sharma SK Carew TJ The roles of MAPK cascades in synaptic plasticity and memory in Aplysia: facilitatory effects and inhibitory constraints Learn Mem 2004 11 373 378 15286179 10.1101/lm.81104
Nestler EJ Kelz MB Chen J DeltaFosB: a molecular mediator of long-term neural and behavioral plasticity Brain Res 1999 835 10 17 10448191 10.1016/S0006-8993(98)01191-3
Guzowski JF Setlow B Wagner EK McGaugh JL Experience-dependent gene expression in the rat hippocampus after spatial learning: a comparison of the immediate-early genes Arc, c-fos, and zif268 J Neurosci 2001 21 5089 5098 11438584
Konradi C Cole RL Heckers S Hyman SE Amphetamine regulates gene expression in rat striatum via transcription factor CREB J Neurosci 1994 14 5623 5634 8083758
Kelz MB Chen J Carlezon WAJ Whisler K Gilden L Beckmann AM Steffen C Zhang YJ Marotti L Self DW Tkatch T Baranauskas G Surmeier DJ Neve RL Duman RS Picciotto MR Nestler EJ Expression of the transcription factor deltaFosB in the brain controls sensitivity to cocaine Nature 1999 401 272 276 10499584 10.1038/45790
Guzowski JF Insights into immediate-early gene function in hippocampal memory consolidation using antisense oligonucleotide and fluorescent imaging approaches Hippocampus 2002 12 86 104 11918292 10.1002/hipo.10010
Fleischmann A Hvalby O Jensen V Strekalova T Zacher C Layer LE Kvello A Reschke M Spanagel R Sprengel R Wagner EF Gass P Impaired long-term memory and NR2A-type NMDA receptor-dependent synaptic plasticity in mice lacking c-Fos in the CNS J Neurosci 2003 23 9116 9122 14534245
Gass P Fleischmann A Hvalby O Jensen V Zacher C Strekalova T Kvello A Wagner EF Sprengel R Mice with a fra-1 knock-in into the c-fos locus show impaired spatial but regular contextual learning and normal LTP Brain Res Mol Brain Res 2004 130 16 22 15519672 10.1016/j.molbrainres.2004.07.004
Nestler EJ Common molecular and cellular substrates of addiction and memory Neurobiol Learn Mem 2002 78 637 647 12559841 10.1006/nlme.2002.4084
Sweatt JD The neuronal MAP kinase cascade: a biochemical signal integration system subserving synaptic plasticity and memory J Neurochem 2001 76 1 10 11145972 10.1046/j.1471-4159.2001.00054.x
Thiels E Klann E Extracellular signal-regulated kinase, synaptic plasticity, and memory Rev Neurosci 2001 12 327 345 11783718
Guan Z Kim JH Lomvardas S Holick K Xu S Kandel ER Schwartz JH p38 MAP kinase mediates both short-term and long-term synaptic depression in aplysia J Neurosci 2003 23 7317 7325 12917365
Bevilaqua LR Kerr DS Medina JH Izquierdo I Cammarota M Inhibition of hippocampal Jun N-terminal kinase enhances short-term memory but blocks long-term memory formation and retrieval of an inhibitory avoidance task Eur J Neurosci 2003 17 897 902 12603281 10.1046/j.1460-9568.2003.02524.x
Alonso M Bevilaqua LR Izquierdo I Medina JH Cammarota M Memory formation requires p38MAPK activity in the rat hippocampus Neuroreport 2003 14 1989 1992 14561935 10.1097/00001756-200310270-00022
Sanyal S Sandstrom DJ Hoeffer CA Ramaswami M AP-1 functions upstream of CREB to control synaptic plasticity in Drosophila Nature 2002 416 870 874 11976688 10.1038/416870a
Davis GW Schuster CM Goodman CS Genetic dissection of structural and functional components of synaptic plasticity. III. CREB is necessary for presynaptic functional plasticity Neuron 1996 17 669 679 8893024 10.1016/S0896-6273(00)80199-3
Sanyal S Narayanan R Consoulas C Ramaswami M Evidence for cell autonomous AP1 function in regulation of Drosophila motor-neuron plasticity BMC Neurosci 2003 4 20 12969508 10.1186/1471-2202-4-20
Kelz MB Nestler EJ deltaFosB: a molecular switch underlying long-term neural plasticity Curr Opin Neurol 2000 13 715 720 11148675 10.1097/00019052-200012000-00017
McClung CA Nestler EJ Regulation of gene expression and cocaine reward by CREB and DeltaFosB Nat Neurosci 2003 6 1208 1215 14566342 10.1038/nn1143
Dubnau J Chiang AS Grady L Barditch J Gossweiler S McNeil J Smith P Buldoc F Scott R Certa U Broger C Tully T The staufen/pumilio pathway is involved in Drosophila long-term memory Curr Biol 2003 13 286 296 12593794 10.1016/S0960-9822(03)00064-2
Osterwalder T Yoon KS White BH Keshishian H A conditional tissue-specific transgene expression system using inducible GAL4 Proc Natl Acad Sci U S A 2001 98 12596 12601 11675495 10.1073/pnas.221303298
Weber U Paricio N Mlodzik M Jun mediates Frizzled-induced R3/R4 cell fate distinction and planar polarity determination in the Drosophila eye Development 2000 127 3619 3629 10903185
Chang L Jones Y Ellisman MH Goldstein LS Karin M JNK1 is required for maintenance of neuronal microtubules and controls phosphorylation of microtubule-associated proteins Dev Cell 2003 4 521 533 12689591 10.1016/S1534-5807(03)00094-7
Gunawardena S Her LS Brusch RG Laymon RA Niesman IR Gordesky-Gold B Sintasath L Bonini NM Goldstein LS Disruption of axonal transport by loss of huntingtin or expression of pathogenic polyQ proteins in Drosophila Neuron 2003 40 25 40 14527431 10.1016/S0896-6273(03)00594-4
Jasper H Benes V Schwager C Sauer S Clauder-Munster S Ansorge W Bohmann D The genomic response of the Drosophila embryo to JNK signaling Dev Cell 2001 1 579 586 11703947 10.1016/S1534-5807(01)00045-4
Gorski SM Chittaranjan S Pleasance ED Freeman JD Anderson CL Varhol RJ Coughlin SM Zuyderduyn SD Jones SJ Marra MA A SAGE approach to discovery of genes involved in autophagic cell death Curr Biol 2003 13 358 363 12593804 10.1016/S0960-9822(03)00082-4
Jasper H Benes V Atzberger A Sauer S Ansorge W Bohmann D A genomic switch at the transition from cell proliferation to terminal differentiation in the Drosophila eye Dev Cell 2002 3 511 521 12408803 10.1016/S1534-5807(02)00297-6
Adams MD Celniker SE Holt RA Evans CA Gocayne JD Amanatides PG Scherer SE Li PW Hoskins RA Galle RF George RA Lewis SE Richards S Ashburner M Henderson SN Sutton GG Wortman JR Yandell MD Zhang Q Chen LX Brandon RC Rogers YH Blazej RG Champe M Pfeiffer BD Wan KH Doyle C Baxter EG Helt G Nelson CR Gabor GL Abril JF Agbayani A An HJ Andrews-Pfannkoch C Baldwin D Ballew RM Basu A Baxendale J Bayraktaroglu L Beasley EM Beeson KY Benos PV Berman BP Bhandari D Bolshakov S Borkova D Botchan MR Bouck J Brokstein P Brottier P Burtis KC Busam DA Butler H Cadieu E Center A Chandra I Cherry JM Cawley S Dahlke C Davenport LB Davies P de Pablos B Delcher A Deng Z Mays AD Dew I Dietz SM Dodson K Doup LE Downes M Dugan-Rocha S Dunkov BC Dunn P Durbin KJ Evangelista CC Ferraz C Ferriera S Fleischmann W Fosler C Gabrielian AE Garg NS Gelbart WM Glasser K Glodek A Gong F Gorrell JH Gu Z Guan P Harris M Harris NL Harvey D Heiman TJ Hernandez JR Houck J Hostin D Houston KA Howland TJ Wei MH Ibegwam C Jalali M Kalush F Karpen GH Ke Z Kennison JA Ketchum KA Kimmel BE Kodira CD Kraft C Kravitz S Kulp D Lai Z Lasko P Lei Y Levitsky AA Li J Li Z Liang Y Lin X Liu X Mattei B McIntosh TC McLeod MP McPherson D Merkulov G Milshina NV Mobarry C Morris J Moshrefi A Mount SM Moy M Murphy B Murphy L Muzny DM Nelson DL Nelson DR Nelson KA Nixon K Nusskern DR Pacleb JM Palazzolo M Pittman GS Pan S Pollard J Puri V Reese MG Reinert K Remington K Saunders RD Scheeler F Shen H Shue BC Siden-Kiamos I Simpson M Skupski MP Smith T Spier E Spradling AC Stapleton M Strong R Sun E Svirskas R Tector C Turner R Venter E Wang AH Wang X Wang ZY Wassarman DA Weinstock GM Weissenbach J Williams SM WoodageT Worley KC Wu D Yang S Yao QA Ye J Yeh RF Zaveri JS Zhan M Zhang G Zhao Q Zheng L Zheng XH Zhong FN Zhong W Zhou X Zhu S Zhu X Smith HO Gibbs RA Myers EW Rubin GM Venter JC The genome sequence of Drosophila melanogaster Science 2000 287 2185 2195 10731132 10.1126/science.287.5461.2185
Jin W Riley RM Wolfinger RD White KP Passador-Gurgel G Gibson G The contributions of sex, genotype and age to transcriptional variance in Drosophila melanogaster Nat Genet 2001 29 389 395 11726925 10.1038/ng766
Giles PJ Kipling D Normality of oligonucleotide microarray data and implications for parametric statistical analyses Bioinformatics 2003 19 2254 2262 14630654 10.1093/bioinformatics/btg311
Hazelrigg T Levis R Rubin GM Transformation of white locus DNA in drosophila: dosage compensation, zeste interaction, and position effects Cell 1984 36 469 481 6319027 10.1016/0092-8674(84)90240-X
Zachar Z Bingham PM Regulation of white locus expression: the structure of mutant alleles at the white locus of Drosophila melanogaster Cell 1982 30 529 541 6291773 10.1016/0092-8674(82)90250-1
Posey KL Jones LB Cerda R Bajaj M Huynh T Hardin PE Hardin SH Survey of transcripts in the adult Drosophila brain Genome Biol 2001 2 RESEARCH0008 11276425 10.1186/gb-2001-2-3-research0008
Kanai Y Dohmae N Hirokawa N Kinesin transports RNA: isolation and characterization of an RNA-transporting granule Neuron 2004 43 513 525 15312650 10.1016/j.neuron.2004.07.022
Rogina B Helfand SL Sir2 mediates longevity in the fly through a pathway related to calorie restriction Proc Natl Acad Sci U S A 2004 101 15998 16003 15520384 10.1073/pnas.0404184101
Chang KT Shi YJ Min KT The Drosophila homolog of Down's syndrome critical region 1 gene regulates learning: implications for mental retardation Proc Natl Acad Sci U S A 2003 100 15794 15799 14668437 10.1073/pnas.2536696100
Barolo S Posakony JW Three habits of highly effective signaling pathways: principles of transcriptional control by developmental cell signaling Genes Dev 2002 16 1167 1181 12023297 10.1101/gad.976502
Cha-Molstad H Keller DM Yochum GS Impey S Goodman RH Cell-type-specific binding of the transcription factor CREB to the cAMP-response element Proc Natl Acad Sci U S A 2004 101 13572 13577 15342915 10.1073/pnas.0405587101
Sanyal S Kim SM Ramaswami M Retrograde regulation in the CNS; neuron-specific interpretations of TGF-beta signaling Neuron 2004 41 845 848 15046717 10.1016/S0896-6273(04)00152-7
Claridge-Chang A Wijnen H Naef F Boothroyd C Rajewsky N Young MW Circadian regulation of gene expression systems in the Drosophila head Neuron 2001 32 657 671 11719206 10.1016/S0896-6273(01)00515-3
Etter PD Ramaswami M The ups and downs of daily life: profiling circadian gene expression in Drosophila Bioessays 2002 24 494 498 12111731 10.1002/bies.10109
McDonald MJ Rosbash M Microarray analysis and organization of circadian gene expression in Drosophila Cell 2001 107 567 578 11733057 10.1016/S0092-8674(01)00545-1
Ueda HR Matsumoto A Kawamura M Iino M Tanimura T Hashimoto S Genome-wide transcriptional orchestration of circadian rhythms in Drosophila J Biol Chem 2002 277 14048 14052 11854264 10.1074/jbc.C100765200
Colosimo ME Brown A Mukhopadhyay S Gabel C Lanjuin AE Samuel AD Sengupta P Identification of thermosensory and olfactory neuron-specific genes via expression profiling of single neuron types Curr Biol 2004 14 2245 2251 15620651 10.1016/j.cub.2004.12.030
SAGE.net
BDGP
Affymetrix
GATC
Park JH Schroeder AJ Helfrich-Forster C Jackson FR Ewer J Targeted ablation of CCAP neuropeptide-containing neurons of Drosophila causes specific defects in execution and circadian timing of ecdysis behavior Development 2003 130 2645 2656 12736209 10.1242/dev.00503
Flybase
| 15932641 | PMC1175850 | CC BY | 2021-01-04 16:03:48 | no | BMC Neurosci. 2005 Jun 2; 6:39 | utf-8 | BMC Neurosci | 2,005 | 10.1186/1471-2202-6-39 | oa_comm |
==== Front
BMC Plant BiolBMC Plant Biology1471-2229BioMed Central London 1471-2229-5-81592706510.1186/1471-2229-5-8Research ArticleBiomarker metabolites capturing the metabolite variance present in a rice plant developmental period Tarpley Lee [email protected] Anthony L [email protected] Tesfamichael H [email protected] Lloyd W [email protected] Texas A&M Agricultural Research and Extension Center, 1509 Aggie Dr, Beaumont, Texas, 77713, USA2 Soil and Crop Sciences Department, Texas A&M University, College Station, Texas, USA3 Analytical Research Laboratories, Oklahoma City, Oklahoma, USA4 Samuel Roberts Noble Foundation, Ardmore, Oklahoma, USA2005 31 5 2005 5 8 8 24 12 2004 31 5 2005 Copyright © 2005 Tarpley 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 analyzes metabolomic data from a rice tillering (branching) developmental profile to define a set of biomarker metabolites that reliably captures the metabolite variance of this plant developmental event, and which has potential as a basis for rapid comparative screening of metabolite profiles in relation to change in development, environment, or genotype. Changes in metabolism, and in metabolite profile, occur as a part of, and in response to, developmental events. These changes are influenced by the developmental program, as well as external factors impinging on it. Many samples are needed, however, to characterize quantitative aspects of developmental variation. A biomarker metabolite set could benefit screening of quantitative plant developmental variation by providing some of the advantages of both comprehensive metabolomic studies and focused studies of particular metabolites or pathways.
Results
An appropriate set of biomarker metabolites to represent the plant developmental period including the initiation and early growth of rice tillering (branching) was obtained by: (1) determining principal components of the comprehensive metabolomic profile, then (2) identifying clusters of metabolites representing variation in loading on the first three principal components, and finally (3) selecting individual metabolites from these clusters that were known to be common among diverse organisms. The resultant set of 21 biomarker metabolites was reliable (P = 0.001) in capturing 83% of the metabolite variation in development. Furthermore, a subset of the biomarker metabolites was successful (P = 0.05) in correctly predicting metabolite change in response to environment as determined in another rice metabolomics study.
Conclusion
The ability to define a set of biomarker metabolites that reliably captures the metabolite variance of a plant developmental event was established. The biomarker metabolites are all commonly present in diverse organisms, so studies of their quantitative relationships can provide comparative information concerning metabolite profiles in relation to change in plant development, environment, or genotype.
==== Body
Background
Variation in crop development due to genotype and environment strongly impacts yield. Increases in crop production efficiency are needed on a global basis because of projected expanding human populations coincident with regional decreases in area of arable land [1,2]. "An understanding of crop responses to environment will provide the fundamental basis for developing methods for achieving these increases in efficiency" (Hall,[2]). Plants interact with environment in both chemical and physical ways, but we have very little systematic understanding of how the plant responds chemically during development and in developmental response to environment [3,4]. This lack of knowledge of the broad changes in metabolite patterns during development limits our efficiency to manipulate the cellular or molecular aspects of plant development with intent to influence yield or sustainability of production.
Recent advances in plant metabolomics, that is large-scale phytochemical analysis of plants [5-7], are paving the way for identifying broad changes in metabolite patterns. Metabolomics has typically been used to characterize the comprehensive changes due to specific environmental or genetic perturbations [6]. Gas chromatography-mass spectrometry (GC-MS) methods currently are being used in many of the metabolomics studies and can provide an accurate and reproducible quantitative and qualitative assessment of a large complement of the metabolome [5,8,7]. A potential disadvantage of the GC-MS methods lies in the serial processing of samples. The time required to analyze large sample numbers can be lengthy for some studies.
Many samples are needed to characterize quantitative aspects of developmental variation. In these situations, methods that use parallel processing of samples to allow high-throughput assay would complement the traditional comprehensive, but serial, procedures, such as GC-MS. A potential disadvantage of the parallel processing methods lies in their dependence on predetermination of the metabolites to be assayed, which presents a possible bias in observed metabolite patterns.
The use of biomarker metabolites is common in many biological fields, including clinical chemistry. Foyer et al. [9] have recently proposed the use of certain amino acids or combinations of them as biomarker metabolites of several metabolic processes or states of plants. In the clinical-chemistry approach, the metabolites are typically chosen based on their diagnostic value, whereas in our study an approach was sought that combined the advantages of the diagnostic approach and the comprehensive metabolomic approach. The comprehensiveness is approached when the set of biomarkers captures much of the variance of the metabolome. The diagnostic value is approached through interpretation of the pattern of the biomarkers relative to each other, and the shifts or distortions in this pattern under various conditions. We anticipate that a biomarker metabolite set constructed through data reduction methods will substantially overlap or capture biomarker sets developed through knowledge of plant physiology.
Representatives from clusters of metabolites can probably capture much of the metabolite variance of a metabolomics study because multiple correlations among metabolites are commonly observed in metabolomics studies [10]. If a set of representatives could be identified for which: (1) the elements (metabolites) represented much of the metabolite variance within a study potentially impacting the improvement of crop production efficiency, (2) the elements were relatively independent of each other, and (3) the elements were common and found in any typical plant sample, then the resultant set of biomarker metabolites could be used in comparative screening of metabolite patterns of plant developmental periods, of plant response to specific environmental factors, or of genotypes in set conditions, and could provide a complementary tool of the comprehensive metabolomic technologies and of diagnostic biomarker approaches.
Metabolite composition is expected to vary consistently in response to development and environment. Core primary metabolites are known to provide good metabolite discrimination between genotypes [11]. This is expected because their quantities are typically affected by many genetic changes, i.e. they are involved in highly regulated activities. Also likely would be an effect on their quantities due to developmental change partially triggered by the internal programming influencing development, and by the need for certain metabolic transitions to occur with a change in growth pattern [9]. Many of these core primary or central metabolites show significant change in response to environmental conditions [12,13]. Knowledge of variation in central metabolism is furthermore considered fundamental for the progress of metabolic engineering [14], indicating a broad belief in its consistent and impacting variation. Primary and central metabolites need to change differentially in development and in response to environment because of biological reasons, and have been demonstrated to do so in a variety of organisms (the examples above include higher plants and Escherichia coli). In addition, a fair amount of general knowledge and assay procedures exist for these metabolites. The likelihood of meeting the three requirements itemized in the above paragraph was considered good, if the following simple procedure was used: 1) partition the variance in metabolites of the developmental event into independent components, 2) ensure that the variation within each of the major components was represented, and then 3) identify a minimal set of central (or nearly central) metabolites that satisfied these conditions.
The utility of a biomarker metabolite set for a developmental study depends on the ability of the biomarkers to provide a snapshot of an aspect of plant development. The ability of the biomarkers to faithfully represent the pattern of variation among the tissues sampled at various locations within the plant and at different plant ages provides a type of internal validation. An appropriate representation of the pattern of relationships among the tissue samples of this study is the set of correlations, based on the metabolite data, existing among them. The reliability of the biomarker metabolite set to capture the metabolite variance can thus be based on the ability to discern the same pattern of relationships among the tissues based on the correlations among them relative to the relationships based on correlations utilizing a more comprehensive metabolite dataset.
If a reliable, validated set of biomarker metabolites could be developed, then a final objective was to provide a demonstration of the type of output from comparative screening of metabolite patterns.
The metabolomics dataset used for development of the proposed biomarker metabolite set examined metabolite composition during initial tiller development of rice (Oryza sativa L.). Tillering is a major yield component of rice, as well as of wheat and many of the other small grains, because the number of tillers per land area strongly influences the number of panicles (heads of grain) per land area. Tiller initiation is sensitive to genotype and environmental effects. Environmental factors affecting tiller initiation include most known to affect plant developmental events, such as radiation quantity and spectral quality, adequacy of nutrition, extent of oxidative stress, and presence of growth inhibitors. A longer-term goal of our project is to understand the commonalities and differences in how these factors affect tiller initiation, so that schemes can be developed to minimize their effects and increase the consistency and manipulability of rice tillering and thus rice crop yield and quality.
Results and discussion
First, some interpretation is provided of the principal components of the metabolite space in rice tiller development. This includes some examination of the patterns in metabolite loadings on Principal Component 1. Next, the results from a metabolite clustering based on the ranked principal component loadings are provided. This section includes the resulting selection of the biomarker metabolites, and a check of their relative independence. The next section discusses the physiological relevance of the biomarker metabolites with a focus on their loadings on the top three principal components. After this, some internal validation is provided by analyzing the reliability of the biomarker set to represent the pattern of metabolite variation observed among the tissues. An external validation is then provided by testing the ability of individual biomarkers to predict the changes in concentration of other metabolites in response to an environmental variable. Finally, a type of output from comparative screening of metabolite patterns is demonstrated.
Interpretation of principal components of the metabolite space in rice tiller development
The metabolite profiles from the rice tiller development study were redistributed into independent subsets through the application of principal component analysis of the metabolite space (as opposed to plant developmental series space). Principal components in standardized centered metabolite space were determined, thus the results are based on analysis performed on the magnitude and pattern of the variation in concentrations of the individual metabolites (rather than on their absolute concentrations). The first five principal components, which explained 83% of the total metabolite variance (the first three principal components explained 62% of the total), were evaluated in this study.
The first pass at interpreting the principal components involved plotting the scores against the two developmental variables of days post-emergence and height of the sample's mid-section (Figure 1). If a principal component was strongly influenced by development, then its score would be expected to demonstrate a pattern of change with respect to both developmental variables. Such a pattern was observed in the plots of Principal Component 1, 3, and 5 scores (Figure 1). If a principal component was strongly influenced by environment, independently of development, then the scores would be expected to demonstrate a pattern of change relative to days post-emergence but not height of the tissue section at sampling. An example of a possible environmental variation that could have an influence only on days post-emergence would be change in photosynthetic radiation intensity on the day of harvest due to variation in cloud cover. Such a pattern was observed in the plots of Principal Component 4 (and probably also Principal Component 2, if the strong influence of the most basal tissue location is temporarily disregarded) (Figure 1). The tendency of the main influence on principal component scores' variations to alternate among two main categories (possibly developmental and environmental) of influences is not unusual (for another example, see Tarpley et al. [15], in which there was demonstrated alternation between physiological and environmental influences).
Patterns in metabolite loadings on Principal Component 1
Patterns in metabolite loadings on Principal Component 1 suggested some metabolite variation could be interpreted via well-known metabolism. For example, Ireland [16] discusses the primary routes of nitrogen flow in amino acid synthesis in plants. A number of the involved amino acids were also relatively strong – positive or negative – contributors (loaders) on the Principal Component 1 of our study. About half of the relatively strong (top 8 or 9 positive or negative loaders of the 155 metabolites that were identified with a standards library, and remaining after removal of some members of highly correlated sets of metabolites) were amino acids, including those indicated by Ireland, indicating that patterns of metabolites observed in relation to plant developmental or environmental factors can sometimes be related to well-known metabolism. For Principal Component 1 in this study, several of the relatively strong positive loaders (serine, glycine, alanine, aspartate) are fairly close in metabolic space to glutamate, which is a relatively strong negative loader, and thus somewhat distant from them in our metabolite space. Glutamate and aspartate, for example, are separated by a single metabolic step – the glutamate:oxaloacetate aminotransferase [16].
These same four metabolites opposing glutamate in the Principal Component 1 loading in metabolite space (alanine, aspartate, and the glycine/serine ratio) have been proposed as a biomarker metabolite set of the relative rate of photorespiration in many C3 crop species based on physiological understanding [17,9]. Photorespiratory activity usually increases with advancement in leaf development [18,19], and would be expected to increase during the developmental period in this study. The high positive loadings of the photorespiration markers on Principal Component 1 supports this expectation because Principal Component 1 scores tend to increase with development, thus these high positive loaders are increasing in relative concentration during development also. The metabolic links among these amino acids, or of the biomarker metabolite set capturing their behaviour along with those of other metabolites in more dimensions of the metabolite space, will be of interest in interpreting the metabolite variance present in tiller initiation and early development in rice.
Metabolite clustering based on ranked principal component loadings
Metabolite selection was initiated via K-means clustering into 27 clusters based on ranked loadings on the three top principal components. Clusters representing 20 of these 27 combinations were found. From 17 of these clusters, representative metabolites were selected based foremost on their proximity to the center of the cluster and nextmost on their perceived commonality as a metabolite. For four of the remaining ten desired combinations, a metabolite from a neighboring cluster was considered sufficiently close to be useful when it had loadings on the first two principal components categorically identical to that sought and with "near-miss" location in loadings on Principal Component 3. No representative metabolites were found for the remaining six combinations. The strongest pattern in common among the six unfilled combinations is that only one has a strong negative loading on Principal Component 1. In other words, because of the tendency of Principal Component 1 scores to increase with development (grasses have basal meristematic tissue, so an increase in mid-section height is also a progression in development) (Figure 1), a representative metabolite could almost always be found in this study when a requirement for that metabolite was to possess a large specific proportional decline in its concentration during development relative to other metabolites. The resultant set of metabolites and the combinations they represent are illustrated in Table 1. An additional data file (see Additional file: 1) lists the metabolites that were used in the principal component analysis and subsequent clustering, and have been at least partially or tentatively identified. The actual ranks based on loading values on the top three principal components are provided.
The distribution of the Pearson correlation values of pairwise comparisons among the selected biomarker metabolites was determined. The systematic method used to select the biomarker metabolites would be expected to yield a range of Pearson values. For example, a metabolite picked from a cluster of metabolites with strong negative loadings on Principal Components 1, 2, and 3 (i.e., trehalose) would be expected to have a fairly positive correlation with citrate (negative on Principal Components 1 and 2, little loading on 3), not much correlation with galactose (not much loading on Principal Components 1 and 2, negative on 3), and a strong negative correlation with uracil (positive loader on all three Principal Components). The mean Pearson value was 0.06 ± 0.06 (95% confidence interval) and the dispersion was 0.41 ± 0.04 (95% confidence interval) (for comparison, a normal distribution would have a mean of 0 and a dispersion of 1, but remember that a non-pathological distribution of correlation values cannot have a dispersion approaching 1 because the correlation values are always between -1 and 1), indicating a satisfactory – relatively independent – distribution of biomarker metabolites based on their correlations.
Interpretation of physiological relevance of biomarker metabolites in relation to their principal component loadings
The biomarker metabolites presented here were selected partially based on their range of loadings on the top three principal components, and any interpretation of their physiological relevance starts with an examination of physiologically relevant groups of them contributing in common to a particular principal component.
The Principal Component 1 score tends to increase with development (Fig. 1), thus the relatively high proportion of amino acids among the biomarkers with strong negative loading on Principal Component 1 (gamma-amino butyric acid, glutamate, leucine, lysine and phenylalanine) is consistent with other observations of a decrease in amino acid contents of leaves during development [20].
The biomarkers present in the tricarboxylic acid cycle (malate, succinate and citrate) are all strong negative loaders on Principal Component 2. This component tends to vary more with days post-emergence than with the more strictly developmental variable of height of sampling of the tissue. These results suggest the tricarboxylic acid cycle is subject to coarse control in response to the environmental factor/s influencing Principal Component 2 values.
A strong opposition exists between the positive- and negative-loading biomarkers on Principal Component 3 with respect to the nitrogen content of the metabolites. Nearly all of the positive-loading biomarker metabolites are nitrogen-containing – glutamate, leucine, phenylalanine, pyroglutamate, thymine, uracil and valine; the sole exception is trans-aconitate. Trans-aconititate is the stable form of aconitate and is capable of being formed from cis-aconitate [21]. Nearly all of the negative-loading biomarker metabolites are sugar or organic acid, non-nitrogen compounds – galactose, mannose, trehalose, carbonate, malate, oxalate and shikimate; the sole exception is gamma-amino butyric acid (GABA). Among the higher plants, GABA is the most widely distributed of the amino acids in which the amino group is not in the alpha position [22].
Reliability of the biomarker set to represent the pattern of metabolite variation observed among the tissues
If the resultant set of biomarker metabolites captures much of the metabolite variance in development as obtained through the metabolomic profiling, then we should be able to "flip" the analysis around and detect natural patterns of tissue relationships that relate to development, and possibly environment, based on their biomarker metabolite concentrations.
In order to evaluate the biomarker metabolite set for ability to consistently detect patterns in metabolite distributions among the tissues during development, we compared the set of all correlations among the sampled tissues based on the standardized and centered concentrations of the 21 biomarker metabolites vs. the set of all correlations based on the values of the five principal components that explained most (83%) variance in the original metabolomics data set. Figure 2 is a scatter plot of the pairs of values for these two data sets. These metabolite concentrations are standardized and centered in the figure because our study was mainly interested in the magnitude and pattern of the variation in the metabolites during development. The correlation values plotted in the figure have also been transformed to a Z-scale to bring out the accuracy (slope of a fitted line would be near 1 with an intercept near 0) in the ability of the biomarker metabolite set to mimic the pattern among the tissues, and with reasonable precision (r = 0.82).
The biomarker metabolite set was determined to be reliable (P = 0.001) in capturing the metabolite-based relationships among tissues present in this particular study of the advent of tillering in rice. We chose to not try to explain any more of the variance of the original dataset because our objective was to develop a biomarker metabolite set to detect patterns among the tissues during development, but we could not, using biological reasoning, explain any of the principal components after Principal Component 5 in terms of patterns among tissues during development. Thus, using only the first five principal components served as a way of filtering out noise [23] in the comprehensive dataset. This helps avoid overfitting of the data [24].
External validation of the biomarker metabolite set
The proposed biomarker metabolite set gains value if used in comparative screening, but its use in comparative situations requires some confidence in its transferability to different situations. Partial confidence is provided by the internal validation described in the above section, which indicates the set was well-constructed, and also by some natural groupings of metabolites associated with the detected principal components. Full confidence, however, requires a demonstration that the set is transferable. An external validation was developed by re-analysing the data of Sato et al. (2004) [25], which is a capillary electrophoresis (CE) – mass spectrometer: CE-diode array detector metabolomics study of rice leaves, with an emphasis on the day/night transition (their Table 2). Pairs of metabolites were identified, in which one member of a pair is a biomarker metabolite that they also measured in their study, and the other member of a pair was a metabolite measured by Sato et al. for which a metabolite had a definitive pairing with the biomarker metabolite. The changes in the day/night ratio in metabolite concentration were compared for the members of each pair that could be constructed. For the six pairs that could be used in the validation, 4 predicted a higher day/night ratio with a higher day/night ratio being observed for each of them, 1 predicted a lower day/night ratio with a lower ratio being observed for it, and one predicted no change in day/night ratio with a higher ratio being observed. The results indicate that the biomarker metabolite set is likely (P = 0.05) to be useful in some other situations, for which these primary and central metabolites would be expected to change in a highly regulated manner, because the tested biomarkers worked well as predictors in the different environmental situation of the day/night transition in leaves.
Example output from comparative screening of biomarker metabolite patterns
A possible way of presenting the results of a biomarker metabolite study is provided in Figure 3, in which four of the tissue sections representing a developmental range in the tillering study are profiled with respect to biomarker metabolite variation. Clearly, these biomarkers of metabolite change in the tillering study change in definable ways. Some (such as several of the organic acids) tend to increase in concentration during development, others (leucine, phenylalanine, trehalose, glutamate) decrease, and others exhibit a more complex pattern. The combination of these patterns reliably reflects the changes in the comprehensive metabolomic profile, and provides a fresh view of the biology of this developmental event. The presentation of metabolite variation based on biomarker variation allows viewing in a single chart whereas the variation of the original 332 metabolites (actually those remaining after the removal of some members of highly intercorrelated sets of metabolites) cannot be easily captured in a single chart. In contrast, the presentation of the variation in only one or a few metabolites fails to capture the broader patterns of change in composition that the presentation of the variation in the biomarker metabolites provides. The distortions in the shape (the relationships of the biomarker metabolites to each other based on their patterns of variation) in the biomarker metabolite space can be captured graphically and/or mathematically and evaluated as an approximation of the change in the comprehensive set of metabolites.
Applicability of biomarker metabolite set for comparative screening
Tillering is an important, well-regulated developmental event in cereal crops and many other grasses. The ability to capture the metabolite variance of this developmental profile using a small set of common metabolites as biomarkers suggests the ability to capture a large portion of it in other plant developmental events using the same set of biomarker metabolites. These metabolites are typically involved in heavily regulated metabolic activities, and it comes as no surprise that they can be useful as biomarkers in plant development. Differences in the use of various metabolic pathways need to occur in different developmental or growth response events, but these are differences in degree. Some of these differences can help modulate the different developmental or growth response events [9]. The interrelationships among the biomarker metabolites would be expected to change in consistent ways in response to change in development, genotype and environment. Comparisons of the changes in the interrelationships of the biomarkers (of the stretches and tucks – the distortions – in the biomarker space) under various conditions will provide new information about plant physiological response in development and in response to environment. The biomarker metabolite set cannot be optimal for any set of conditions, but is likely to be fairly robust in capturing physiologically real differences among them, while being responsive to eventual metabolic interpretation.
Conclusion
Variation in crop development due to genotype and environment greatly impacts yield, yet the community's understanding of the quantitative biochemical variation in plant development is small.
This paper has presented an approach to developing a biomarker metabolite set that captures much of the metabolite variation present in a comprehensive metabolomics set of a plant developmental event, tillering, in rice. The resulting biomarker set is intended to provide some of the advantages of a metabolomics approach and of the use of one or a few diagnostic metabolites.
The approach uses simple and commonly available multivariate statistics, namely principal component analysis and K-means clustering, to assist with the biomarker metabolite selection. The resulting set of 21 biomarkers was shown to be reliable in capturing the metabolite variance in the comprehensive metabolomics study and valid for predicting metabolite changes observed in another metabolomics study, while capturing variation that has reasonable potential to be related to well-known metabolism.
The selection of the biomarker metabolites was further constrained to primary or central or common metabolites. Because the selected metabolites are common among diverse organisms and often have specific assay procedures already developed (e.g., Bergmeyer [26], Passoneau and Lowry [27], Gibon et al. [4], and Kiianitsa et al. [28]), the biomarker set is amenable to future assay via high-throughput technologies, and in application to diverse situations. The biomarker metabolite set can serve as a basis for comparative screening of metabolite patterns of plant developmental periods, of plant response to specific environmental factors, or of genotypes in set conditions.
Methods
Culture of plants and initial preservation of tissue samples
Seedlings of rice cv. IR-36 were grown in a black clay soil, typical of the area's rice fields, in flats in the greenhouse under typical temperature and supplemental lighting regimes at the Texas A&M Agricultural Research and Extension Center at Beaumont, Texas, USA. Nitrogen fertilizer was applied as urea at the 2–3 leaf stage. Sampling started one week after emergence and continued at two- to four-day intervals for a total of five dates over a 10-day period. Separate sets of seedlings were used at each sampling date, thus avoiding injury to existing plants. The seedlings developed to about the 3-leaf to 5-leaf stage during this interval. All samples were collected, processed and stored between 1000 and 1400 h CDT (near solar mid-day). The soil was washed off the seedlings within five minutes, and the seedlings placed with roots in tap water until dissection. This is an established procedure for maintaining rice plant integrity for short periods during sampling.
Seedlings were sectioned in 2-mm intervals along the developing culm, starting at the base of the plant and continuing for a total of ten sections. At this developmental stage, 2-mm sections are thick enough to ensure that the large majority of cells are not disrupted from the slicing action of the new ethanol-cleansed razor blades. This minimizes injury response. Each replicate contained tissue from an average of 50 seedlings, i.e. 2-mm sections from the same culm position for each of 50 seedlings. Slow-growing seedlings were avoided. The obtained sections were plunged within a few seconds into liquid nitrogen until all sections were collected. Sections were then stored at -80°C in nitrogen-purged vials until lyophilized for use in the metabolomics procedures. There were three replicates, each with sections from 50 seedlings.
Only sections of positions 1, 3, 5, 7 and 9 from the base of the plant were used for the metabolomics. The sections were collected along two gradients: along the culm and during development.
After lyophilization (Labconco Freezone 6, Kansas City, Missouri, USA), the samples were capped in amber glass vials under nitrogen and then sealed externally around the cap rim with polyethylene homopolymer film (Parafilm M; Pechiney Plastic Packaging, Neenah, Wisconsin, USA) to further minimize gas penetration prior to shipment to Noble Foundation laboratories for metabolomic analysis.
Extraction, derivatization, and instrumental analysis
A total of 6.02 mg (± 0.02 mg) of the lyophilized pulverized tissue was weighed into 3.7-mL (1 dram) vials containing teflon inlays. Metabolite extractions were performed by adding 1.5 mL chloroform containing 5 mg/L phenanthrene internal standard and 1.5 mL bottled water containing 25 mg/L ribitol internal standard, followed by vortexing for 1 min. The samples were incubated at 50°C for 1 h with shaking and placed at -20°C overnight. The samples were again incubated at 50°C for 4 h with shaking. The samples were then sonicated 30 s and centrifuged in a swinging bucket rotor at 2,900 × g for 30 min. Aliquots of 1.2 mL were taken from both the polar and lipophilic layers and transferred to 2.0 mL autosampler vials (Agilent, Palo Alto, California, USA) with teflon/silicon septa. The polar layer was dried in a speed vac (Savant, Albertville, Minnesota, USA) for 4 h, and the lipophilic layer dried under a stream of compressed nitrogen for 2 h. Extracts were stored at -20°C until ready for analysis. Both the polar and lipophilic extracts were analyzed.
Dried polar extracts were prepared by methoximation in 120 μl of 15 g/L methoxyamine hydrochloride in pyridine at 50°C for 4 h followed by a brief sonication (<30 s) to dislodge any pellet. Samples were derivatized by adding 120 μl N-methyl-N-(trimethylsilyl)trifluoroacetamide) +1% trimethylchlorosilane followed by incubation at 50°C for 1 h.
Dried lipophilic extracts were prepared by transmethylation of fatty acid and lipids. Dried extracts were dissolved in 100 μl chloroform and 300 μl methanol containing 1.25 M HCl and then incubated at 50°C for 24 h. Samples were then dried under a stream of nitrogen for 2 h. Dried transmethylated samples were resuspended in 70 μl pyridine and derivatized in 30 μl N-methyl-N-(trimethylsilyl)trifluoroacetamide) +1% trimethylchlorosilane at 50°C for 1 h.
Derivatized metabolite mixtures were analyzed using a Hewlett Packard 6890 gas chromatograph, 5973 mass selective detector, and 6890 series injector. The integrated system was operated under HP Chemstation (Agilent). Polar samples were analyzed by injecting 1 μl with a split injection ratio of 5:1; lipophilic samples of 1 μl were analyzed using a 1:1 split injection ratio. All samples were injected in duplicate. Analyses were performed using a 60-meter DB-5MS capillary separation column (J&W Scientific, Palo Alto, California, USA). Injection temperature was 280°C, interface 280°C. Separations were achieved using the following temperature program: 3 min isothermal heating at 80°C, followed by a 5°C min-1 oven ramp to 315°C, and a final isothermal heating at 315°C for 14 min for polar and 12 min for lipophilic samples. Mass spectra were recorded at 2.48 scans s-1 with a mass scanning range of 50 to 650 m/z. Each run required approximately 70 min including machine equilibration time.
Metabolite identifications were determined using GC-MS spectral database matching against the current National Institute of Standards and Technology library (NIST02) and a Noble Foundation in-house custom database focused on plant metabolites. Standards for construction of the in-house library were prepared by methoximation and derivatization as described above. The Automated Mass Spectral Deconvolution and Identification Software (AMDIS) (National Institute of Standards and Technology (NIST), Gaithersburg, MD) was utilized for library construction and metabolite identification in the raw metabolite profiles.
The GC-MS chromatographic data alignment was performed according to Duran et al. [29]. Selected ions were extracted and aligned from raw metabolomic data files using a custom Perl script. This provides a more comprehensive interrogation of the data and is capable of resolving coeluting chromatographic peaks based on the underlying mass data.
Correlation among metabolites
The correlations between pairs of metabolites were evaluated as Pearson correlation coefficients [30]. One or more metabolites from highly intercorrelated sets of metabolites were omitted from the dataset when necessary to ensure subsequent calculations did not involve singular matrices [31]. Singular matrices do not provide unique solutions with many multivariate statistical methods.
Principal component analysis
Principal components in standardized centered metabolite space were determined. This analysis included metabolites that were not matched against a standards library, but did not include those omitted to avoid singular matrices. The analysis was performed in MathCad 2001 (MathSoft, Inc., Cambridge, Massachusetts, USA) using the matrix manipulations described by Pielou [23]. Although there are methods available that can assist in determining which principal components to retain (e.g., scree analysis), the cut-off was made based on the observations that other principal components neither explained much of the variance nor exhibited any pattern in metabolite loadings that could be easily related to known metabolism, or as a response to developmental or environmental variables [32].
Metabolite selection via K-means clustering
Individual metabolites were selected to provide a spread of variation in loadings on the first three principal components, while being potentially easy to assay. Metabolite selection involved K-means clustering of the ranked loadings on the three top principal components, followed by individual selection of promising metabolites from the clusters. The K-means clustering was performed using Cleaver (Classification of Expression, Array Version 1.0) software available through the site for microarray analysis maintained by the Helix Bioinformatics Group at the Stanford School of Medicine [33]. The analysis sought 27 clusters using the Euclidean distance metric. This set-up maximized the potential to obtain clusters representing all possible combinations of strong positive loading, weak loading, and strong negative loading elements for each of the three principal components. In other words, 3 loading strengths × 3 principal components, taken three at a time = 27 combinations for which a representative metabolite was desired. Representative metabolites were selected based foremost on their proximity to the center of the cluster and then on their perceived commonness as a metabolite. For the remaining clusters, a metabolite from a neighboring cluster was considered sufficiently close to be useful when it had loadings on the first two principal components categorically identical to that sought and with "near-miss" location in loadings on Principal Component 3. If no sufficiently representative metabolite could be found with this approach, then no further effort was made and such clusters remained unrepresented.
Reliability of the biomarker set to represent the pattern of metabolite variation observed among the tissues
In order to evaluate the biomarker metabolite set for ability to consistently detect patterns in metabolite distributions among the tissues during development, the set of all pair-wise Pearson correlations [30] among the sampled tissues based on the standardized and centered concentrations of the biomarker metabolites were obtained. The equivalent set of all pair-wise correlations was obtained based on the unranked values of the five principal components that explained most (83%) variance in the original metabolomics data set. The reliability of the biomarker metabolite set to capture the metabolite variation present among the tissues in development was obtained by comparing the two sets of correlation values using the methods of Fisher [30]. There are 435 pairs of values to compare, but these are not truly independent variables, so the reliability was analyzed with 28 degrees of freedom (30 tissue samples minus 2 degrees of freedom) to account for the lack of independence among pairs of correlation values.
External validation of the biomarker metabolite set
An external validation analysis was developed by re-analysing the data of Sato et al. (2004) [25], which is a capillary electrophoresis (CE)- mass spectrometer: CE-diode array detector metabolomics study of rice leaves, with an emphasis on the day/night transition (their Table 2). Pairs of metabolites were identified, in which one member of a pair was a biomarker metabolite that they also measured in their study, and the other member of a pair was a metabolite measured by Sato et al. for which a metabolite had a definitive pairing with the biomarker metabolite. The changes in the day/night ratio in metabolite concentration were compared for the members of each pair that could be constructed, and the directions of changes in the biomarker metabolites were used to predict the changes in the other metabolites. The average percentage change in concentration of the metabolites used in this validation test was considered fairly small at 51% of the night-time value, thus justifying the use of the measure of the ability of the biomarkers to indicate the direction of change in concentration of the respective predicted metabolites in the day/night transition as a reasonably powerful validation test of the biomarkers' ability to mark broad changes in the other metabolites.
Authors' contributions
Culture of plants, sampling of tissues and initial preservation of tissue samples (THK, LT); Extraction, derivatization, and instrumental analysis through and including mass spectral interpretation and metabolite identification (ALD; LWS); Data analysis related to development of the biomarker metabolite set; planning and writing the manuscript (LT). All authors read and approved the final manuscript.
Supplementary Material
Additional File 1
Ranked principal component loadings of identified metabolites. List of metabolites that were used in the principal component analysis and subsequent clustering, and have been at least partially or tentatively identified. The actual ranks based on loading values on the top three principal components are presented. The selected biomarker listings are bolded. When a biomarker has multiple metabolite entries, then these entries are matched through use of a non-black font color.
Click here for file
Acknowledgements
We thank the Texas Rice Research Foundation for providing funding to LT, including support for THK, during conductance of this study.
Figures and Tables
Figure 1 Principal component scores during a rice plant developmental period bridging first tillering. The scores (categorized by value using a grey-scale as indicated in the legend) of Principal Components 1 to 5 (Panels 1 to 5, respectively) are plotted against the progression in sampling of days post-emergence (horizontal axis) and the height of the sampled tissue section (as height [mm] of mid-section – the vertical axis of each panel). Rice plants have a basal meristem, so an increase in mid-section height is also a progression in development. The principal components are of a standardized, centered metabolite space from an analysis of a comprehensive metabolomics dataset. Principal Components 1, 3 and 5 show a pattern of change relative to both developmental variables (axes), while Principal Component 4 and Principal Component 2 (if the influence of the 1-mm mid-section height samples are ignored) vary mainly with days post-emergence and are probably influenced by environment more than development. Each value is the mean of three replicates, each of which pooled sections representing 50 different plants.
Figure 2 Plot of correlations among samples based on biomarker metabolites vs. based on principal component scores. The biomarker metabolite set does a reasonable job of mimicking the pattern among the sampled tissues based on the top five principal components. The pattern among the tissues with respect to their metabolite composition is discerned by the set of all pairwise Pearson correlation values among the tissues. This "correlation measure" of the pattern among the tissues was applied using two different sets of markers. The set represented by the horizontal axis used the scores of the top five principal components from the analysis of the comprehensive metabolomics dataset. The set represented by the vertical axis used the biomarker metabolite concentrations. These metabolite concentrations are standardized and centered because our study was mainly interested in the magnitude and pattern of the variation in the metabolites during development. The correlation values plotted in the figure have been transformed to a Z-scale to bring out the accuracy (slope of a fitted line would be near 1 with an intercept near 0) in the ability of the biomarker metabolite set to mimic the pattern among the tissues, and with reasonable precision (r = 0.82).
Figure 3 Magnitude and pattern of variation in biomarker metabolite concentrations in samples ranging in development. Four of the tissue samples representing a developmental range are profiled with respect to biomarker metabolite variation. The samples progress in height at mid-section of the sampled tissue and in days post-emergence, thus they represent a cross-section of the larger set of tissue samples. The biomarker metabolites are listed along the horizontal axis, and each dot plot shows the Z-scores for the biomarker metabolite concentration. The metabolites vary a lot in absolute concentration, so the Z-score is used to equalize the overall variation in concentration of the metabolites during the study. Thus the figure shows the pattern and magnitude of the variation among the presented tissues, but also the amount of this variation relative to that of the metabolite concentration for the whole study. For example, oxalic acid and glutamate both have a fairly wide range of Z-scores for these samples, although the pattern of variation is opposite.
Table 1 The selected biomarker metabolites, and the combinations of loadings on principal components that they represent. The set of biomarker metabolites selected to capture much of the variance in metabolite composition of a rice tillering event is listed. The selection procedure constrained the biomarker metabolites to represent variation in loading on the top three principal components in the standardized centered metabolite space of an analysis of a comprehensive metabolomics dataset. The loading of the individual metabolite on the principal components is symbolically represented: high positive loading (POS), weak loading (---), or high negative loading (NEG).
Biomarker metabolite Principal Component 1 loading Principal Component 2 loading Principal Component 3 loading
Trehalose NEG NEG NEG
Citric Acid NEG NEG ---
Glutamic Acid NEG NEG POS
Mannose NEG --- NEG
Phenylalanine NEG --- POS
gamma-Aminobutyric Acid (GABA) NEG POS NEG
Lysine NEG POS ---
Leucine NEG POS POS
Shikimic Acid --- NEG NEG
Succinic Acid --- NEG ---
Pyroglutamic Acid --- NEG POS
Galactose --- --- NEG
Valine --- --- POS
p-Hydroxybenzoic Acid --- POS ---
Thymine --- POS POS
Malic Acid POS NEG NEG
Salicylic Acid POS NEG ---
Oxalic Acid POS --- NEG
trans-Aconitic Acid POS --- POS
Carbonate POS POS NEG
Uracil POS POS POS
==== Refs
Evans LT Crop Evolution, Adaptation and Yield 1993 Cambridge, England , Cambridge University Press 500 8422359
Hall AE Crop Responses to Environment 2001 Boca Raton, FL, USA , CRC Press LLC 232
Lambers H Poorter H Inherent variation in growth rate between higher plants: a search for physiological causes and ecological consequences Advances in Ecological Research 1992 23 187 261
Gibon Y Vigeolas H Tiessen A Geigenberger P Stitt M Sensitive and high throughput metabolite assays for inorganic pyrophosphate, ADPGlc, nucleotide phosphates, and glycolytic intermediates based on a novel enzymic cycling system Plant Journal 2002 30 221 236 12000458 10.1046/j.1365-313X.2001.01278.x
Fiehn O Kopka J Dormann P Altmann T Trethewey RN Willmitzer L Metabolite profiling for plant functional genomics Nature Biotechnology 2000 18 1157 1161 11062433 10.1038/81137
Hall R Beale M Fiehn O Hardy N Sumner LW Bino R Plant metabolomics as the missing link in functional genomics strategies Plant Cell 2002 14 1437 1440 12119365 10.1105/tpc.140720
Sumner LW Mendes P Dixon RA Plant metabolomics: large-scale phytochemistry in the functional genomics era Phytochemistry 2003 62 817 836 12590110 10.1016/S0031-9422(02)00708-2
Fiehn O Metabolomics - the link between genotypes and phenotypes Plant Molecular Biology 2002 48 155 171 11860207 10.1023/A:1013713905833
Foyer CH Parry M Noctor G Markers and signals associated with nitrogen assimilation in higher plants Journal of Experimental Botany 2003 54 585 593 12508069 10.1093/jxb/erg053
Steuer R Kurths J Fiehn O Weckwerth W Observing and interpreting correlations in metabolomic networks Bioinformatics 2003 19 1019 1026 12761066 10.1093/bioinformatics/btg120
Taylor J King RD Altmann T Fiehn O Application of metabolomics to plant genotype discrimination using statistics and machine learning Bioinformatics 2002 18 S241 S248 12386008
Tweeddale H Notley-McRobb L Ferenci T Effect of slow growth on metabolism of Escherichia coli, as revealed by global metabolite pool ("metabolome") analysis Journal of Bacteriology 1998 180 5109 5116 9748443
Bailey NJC Oven M Holmes E Nicholson JK Zenk MH Metabolomic analysis of the consequences of cadmium exposure in Silene cucubalus cell cultures via 1H NMR spectroscopy and chemometrics Phytochemistry 2003 62 851 858 12590112 10.1016/S0031-9422(02)00719-7
Buchholz A Hurlebaus J Wandrey C Takors R Metabolomics: quantification of intracellular metabolite dynamics Biomolecular Engineering 2002 19 5 15 12103361 10.1016/S1389-0344(02)00003-5
Tarpley L Sassenrath GF VanToai T, Major D, McDonald M, Schepers J, Tarpley L Environmental and physiological components of the cotton leaf reflectance spectrum Digital Imaging and Spectral Techniques: Applications to Precision Agriculture and Crop Physiology 2003 Madison, WI, USA , American Society of Agronomy, Inc., Crop Science Society of America, Inc., Soil Science Society of America, Inc. 95 109
Ireland R Dennis DT, Layzell DB, Lefebvre DD, Turpin DH Amino acid and ureide biosynthesis Plant metabolism 1997 2nd Singapore , Addison Wesley Longman Ltd. 478 494
Novitskaya L Trevanion SJ Driscoll S Foyer CH Noctor G How does photorespiration modulate leaf amino acid contents? a dual approach through modelling and metabolite analysis Plant, Cell and Environment 2002 25 821 835 10.1046/j.1365-3040.2002.00866.x
Lennon AM Pratt J Leach G Moore AL Developmental regulation of respiratory activity in pea leaves Plant Physiology 1995 107 925 932 12228412
Thompson P Bowsher CG Tobin AK Heterogeneity of mitochondrial protein biogenesis during primary leaf development in barley Plant Physiology 1998 118 1089 1099 9808754 10.1104/pp.118.3.1089
Jeong ML Jiang H Chen HS Tsai CJ Harding SA Metabolic profiling of the sink-to-source transition in developing leaves of quaking aspen Plant Physiology 2004 136 3364 3375 15448196 10.1104/pp.104.044776
Cai H Strouse J Dumlao D Jung ME Clarke S Distinct reactions catalyzed by bacterial and yeast trans-aconitate methyltransferases Biochemistry 2001 40 2210 2219 11329290 10.1021/bi0022902
Robinson T The Organic Constituents of Higher Plants. Their Chemistry and Interrelationships 1980 4th North Amherst, Massachusetts, U.S.A. , Cordus Press 352
Pielou EC The Interpretation of Ecological Data 1984 New York, NY, USA , John Wiley & Sons, Inc. 263
Seasholtz MB Kowalski B The parsimony principle applied to multivariate calibration Analytica Chimica Acta 1993 277 165 177 10.1016/0003-2670(93)80430-S
Sato S Soga T Nishioka T Tomita M Simultaneous determination of the main metabolites in rice leaves using capillary electrophoresis mass spectrometry and capillary electrophoresis diode array detection Plant Journal 2004 40 151 163 15361149 10.1111/j.1365-313X.2004.02187.x
Bergmeyer HU Bergmeyer J Graßl M Methods of Enzymatic Analysis 1984 3rd Weinheim, Germany , Verlag Chemie
Passoneau JV Lowry OH Enzymatic Analysis: A Practical Guide 1993 Totowa, NJ, USA , Humana Press, Inc. 403
Kiianitsa K Solinger JA Heyer WD NADH-coupled microplate photometric assay for kinetic studies of ATP-hydrolyzing enzymes with low and high specific activities Analytical Biochemistry 2003 321 266 271 14511695 10.1016/S0003-2697(03)00461-5
Duran AL Yang J Wang L Sumner LW Metabolomics spectral formatting, alignment and conversion tools (MSFACTs) Bioinformatics 2003 19 2283 2293 14630657 10.1093/bioinformatics/btg315
Snedecor GW Cochran WG Statistical Methods 1980 7th Ames, IA, USA , Iowa State University Press 507
Carroll JD Green PE Chaturvedi A Mathematical Tools for Applied Multivariate Analysis 1997 revised San Diego, CA, USA , Academic Press 376
Brereton RG Chemometrics: Data Analysis for the Laboratory and Chemical Plant 2003 Chichester, West Sussex, England , John Wiley & Sons, Ltd. 489
Cleaver (Classification of Expression, Array Version 1.0) [http://classify.stanford.edu/documentation.html]
| 15927065 | PMC1175851 | CC BY | 2021-01-04 16:03:52 | no | BMC Plant Biol. 2005 May 31; 5:8 | utf-8 | BMC Plant Biol | 2,005 | 10.1186/1471-2229-5-8 | oa_comm |
==== Front
Biomed Eng OnlineBioMedical Engineering OnLine1475-925XBioMed Central London 1475-925X-4-391596323710.1186/1475-925X-4-39ResearchA 63 element 1.75 dimensional ultrasound phased array for the treatment of benign prostatic hyperplasia Saleh Khaldon Y [email protected] Nadine Barrie [email protected] Department of Bioengineering Graduate Program in Acoustics College of EngineeringThe Pennsylvania State University 206 Hallowell Building University Park, PA 16802, USA2 Graduate Program in Acoustics College of Engineering The Pennsylvania State University 206 Hallowell Building University Park, PA 16802, USA2005 17 6 2005 4 39 39 16 12 2004 17 6 2005 Copyright © 2005 Saleh and Smith; licensee BioMed Central Ltd.2005Saleh and Smith; licensee BioMed Central Ltd.This is an Open Access article distributed under the 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
Prostate cancer and benign prostatic hyperplasia are very common diseases in older American men, thus having a reliable treatment modality for both diseases is of great importance. The currently used treating options, mainly surgical ones, have numerous complications, which include the many side effects that accompany such procedures, besides the invasive nature of such techniques. Focused ultrasound is a relatively new treating modality that is showing promising results in treating prostate cancer and benign prostatic hyperplasia. Thus this technique is gaining more attention in the past decade as a non-invasive method to treat both diseases.
Methods
In this paper, the design, construction and evaluation of a 1.75 dimensional ultrasound phased array to be used for treating prostate cancer and benign prostatic hyperplasia is presented. With this array, the position of the focus can be controlled by changing the electrical power and phase to the individual elements for electronically focusing and steering in a three dimensional volume. The array was designed with a maximum steering angle of ± 13.5° in the transverse direction and a maximum depth of penetration of 11 cm, which allows the treatment of large prostates. The transducer piezoelectric ceramic, matching layers and cable impedance have been designed for maximum power transfer to tissue.
Results
To verify the capability of the transducer for focusing and steering, exposimetry was performed and the results correlated well with the calculated field. Ex vivo experiments using bovine tissue were performed with various lesion sizes and indicated the capability of the transducer to ablate tissue using short sonications.
Conclusion
A 1.75 dimensional array, that overcame the drawbacks associated with one-dimensional arrays, has been designed, built and successfully tested. Design issues, such as cable and ceramic capacitances, were taken into account when designing this array. The final prototype overcame also the problem of generating grating lobes at unwanted locations by tapering the array elements.
ultrasound transducer1.75 dimensional arrayfocusingmatching layer
==== Body
1. Background
Treating prostate diseases such as prostate cancer and benign prostatic hyperplasia (BPH) is of great importance. In the United States, most of the new diagnosed prostate cancer cases appear in men who are over the age of 55 while most of the BPH cases appear after the age of 60. According to the National Cancer Institute, 50 percent of men between the ages of 60 and 70, and 90 percent of men between the ages of 70 and 90, have BPH symptoms. Prostate cancer is a life threatening disease while BPH is a benign growth that needs to be treated since normal urine flow can be blocked as a result of the prostate pushing against the urethra and the bladder (National Cancer Institute 1999).
Existing techniques for treating such diseases include hyperthermia, focus surgery, radiotherapy, chemotherapy and surgery. Currently, surgical techniques are widely used over the other modalities; that is due to the inefficiency and the unpleasant side effects those modalities have. However surgical techniques have numerous complications that appear in about one in four cases, which include impotence, incontinence, and urinary tract infections and often require lengthy hospitalization [1,2].
Due to its noninvasiveness, focus surgery is gaining more attention than the other modalities in the past decade [3]. With focus surgery, ultrasound or microwave devices are used to generate a focused beam at a certain location in the prostate, which kills the cells at that location by raising their temperature to 60°C for about ten seconds. Attention is given more to ultrasound rather than microwave. That is because microwave has either a shallow penetration depth (when high frequencies are used) or a lack of the ability to generate a significant focus (when low frequencies are used) [4].
With focused ultrasound (FUS), tissue is noninvasively ablated by elevating the temperature at the focal point above 60°C using short sonications (10–30 seconds). In this kind of treatment, the target volume can be ablated by focusing the ultrasound beam at a certain position, and then steering the focus to cover the whole enlarged volume. Thus FUS can be used for prostate ablation to remove a non-desirable growth of the prostate [5-7]. Since the tissue volume to be ablated is larger than the geometric focus of the array, the transducer needs to be moved repeatedly to destroy the desired volume and unnecessarily extend the treatment time. Phased arrays overcome this problem by electrically steering the focal point from one location to another by changing the phase and power to the individual elements of the array. Previous effective prostate ultrasound devices include both mechanically and electrically steered designs. Electrically steered include a one-dimensional (1-D) 120 × 1 aperiodic, linear array design (90 × 15 mm2) which reduced grating lobes and could steer the focus in the radial and transverse but not the longitudinal direction [8]. Another experimental design was a 62 × 1, linear array (75 × 15 mm2) with a mechanical translation that could electrically steer the focus in the radial and transverse but not the longitudinal direction [9]. The drawbacks behind these designs are that they can only steer the focus in the radial and transverse directions or require complex mechanisms to move the focus. Improvements over 1-D arrays for the treatment of localized prostatic cancer can be achieved. Many multi-dimensional ultrasound phased arrays have been designed and built for the treatment of prostate diseases; that includes a 1.5-dimensional (1.5-D) phased array [10] (a 1.5-D array consists of three individual linear array that can be driven individually or connected together to form a single linear array), a 1.75-dimensional (1.75-D) phased array [11] (a 1.75-D array consists of many individual linear arrays that are driven separately), and a two-dimensional (2-D) phased array [12]. The advantage with a multi-dimensional phased array is that it has the capability of focusing and steering in a 3-dimensional (3-D) representation of the prostate without the need to physically move the array.
Issues regarding the construction of an array used for FUS of the prostate initially deal with the frequency and size of the ceramic to be diced into an array. The resonant frequency should be greater than 500 kHz [13] while the size of the transducer needs to be large enough to be able to deliver high power but small enough to be an intracavitary device. Before construction, computer simulations can be performed to determine the acoustic field. Pressure wave and temperature simulations indicated that a tapered array design reduced grating lobes significantly compared to equal element size arrays. Based on the computer model, a tapered array that satisfied grating lobes, frequency, and size limitations was designed. Lead zirconate titanate (PZT- 8) was chosen as the ceramic material of the array since it has the capability of handling the high electrical powers used in focused ultrasound. To maximize the acoustical power transmission from the elements and improve the structural integrity of the array face, two matching layers were designed and fabricated. Issues regarding the cabling and electrical matching of the elements were also considered. Exposimetry of the acoustic field from the array was performed to compare experimental and calculated theoretical results. Ex vivo experiments using bovine tissue were also performed to demonstrate the feasibility of the array to ablate tissue. This paper describes the design, construction and evaluation of a 1.75-D ultrasound phased array that is capable of focusing and steering in a 3-D volume to be used in the treatment of BPH.
2. Methods
2.1 Simulations
2.1.1 Acoustic pressure field simulations
MATLAB computer simulation programs were written to determine the number and the size of the phased array elements in addition to determining the pressure and temperature fields from the device. The array was modeled (Figure 1) as a 1.75-D tapered array in order to have focusing and steering capabilities in both x and z directions (x = transverse, y = longitudinal and z = radial). Focusing in the y direction is done in a different way; the array is divided into three identical rows, each one represents a single linear array. If the focus is required at y = 0, the middle row should be used. A focus at y = -0.9 cm requires driving the lower row, while a focus at y = +0.9 cm requires the operation of the upper row. Although the degree of freedom in the y direction is not perfect, the size of the lesion generated by a single sonication compensates for that, since the focus length is about 9 mm in the y direction. With these requirements, this array was capable of focusing and steering with a steering angle of ± 13.5° with maximum focal depth of 11 cm. The phase of each element was determined such that signals from individual elements were coherent at the focal point. Measuring the difference in path length between each element to the focus in comparison to the path from the center of the array to the focus determined the element phase calculation. The phase, φi, (degrees) of element i was given by:
Figure 1 Based on the simulations, a diagram of the 1.75-D 63 element (3 × 21) tapered array with total size of 27 × 53 mm2 with the proportions of the ceramic and matching layer illustrated. The diced face of the ceramic was cut 100% through and each individual element was attached to the electrical cabling using low temperature soldering material.
Where λ is the wavelength (m), di is the distance (m) from the centre of element i to the focal point, dois the distance (m) from the centre of the array to the focus and n is an integer to keep 0 ≤ φi ≥ 360°. Huygen's principle was used to model the pressure field as a summation of simple sources [14] and the total acoustic pressure at any point in the field was calculated using the discrete approximation of the Rayleigh- Sommerfeld equation:
Where p is the total acoustic pressure in Pascals (Pa), P is the total acoustic power emitted by the array in watts (W), ρ is the density of the medium (998 kg·m-3), c is the speed of sound (m·s-1), A is the total surface area of the array (m2), f is the resonant frequency (1.2 MHz), S is the area of the corresponding element (m2) and α is the attenuation in soft tissue (10 Np·m-1·MHz-1).
The acoustic pressure field simulations started with a 1-D model that was used to simulate different tapering techniques to see their effect on the grating lobe values. Equal, linear, Hanning and Hamming tapering techniques were simulated. Improvements to the tapered array design started with a 27 × 53 mm2 solid piezoceramic cut into a 3 × 21 pattern with 63 individual elements with lengths (Li) of 1.68, 1.73, 1.81, 1.91, 2.02, 2.14, 2.26, 2.36, 2.43, 2.48, 2.50, 2.48, 2.43, 2.36, 2.26, 2.14, 2.02, 1.91, 1.81, 1.73, 1.68 mm for elements i = 1 through 21, respectively, and widths (Wi) of 9.0 mm for all elements i = 1 through 3, respectively (Figure 1). The maximum possible steering angle was calculated to be tan-1(1.2/5.0) = 13.5° with maximal focal depth of 11 cm. Off-axis focusing and the grating lobe level are directly related to each other since increasing the steering angle causes a nonlinear increase in the grating lobe level. However, the designed array described in this paper kept a good grating lobe level when aiming the focus at a point that was 5 mm away from the z direction. When focusing at (0.2, 0, 5) and (0.5, 0, 5) cm, the grating lobe level was kept around -12 dB, as can be seen in Figures 2(a) and 2(b), respectively.
Figure 2 Off-axis focusing has a direct impact on the grating lobe level. Increasing the steering angle by changing the focal point position in the x direction increases the grating lobe level. For a focus aimed at (0.2, 0, 5) and (0.5, 0, 5) cm, a fair grating lobe level of about -12dB was observed, as seen in (a) and (b), respectively.
2.1.2 Temperature distribution simulations
From the pressure field of the simulated array, the temperature distribution in the tissue was modeled using the Pennes' bioheat transfer equation (BHTE) [15]:
Where Ct is the specific heat of the tissue (3770 J·kg-1·°C-1), K is the thermal conductivity (0.5 W·m-1·°C-1), T is the temperature at time t at the point x, y, z in °C, Ta is the arterial blood temperature (37°C), w is the perfusion in the tissue in kg·m-3·s-1, Cb is the specific heat of the blood (3770 J·kg-1·°C-1) and q(x, y, z) is the power deposited at the point x, y, z. The power was calculated from the pressure field of the array design while the BHTE was determined using a numerical finite difference method with the boundary conditions set at 37°C. The total intensity at point (x, y, z) was also calculated from the pressure field of the simulated array and is given by [16]:
Where I(x, y, z) is the intensity at point (x, y, z) in W·m-2.
Temperature simulations were used to verify the potential to increase the tissue temperature to about 60°C with short sonications. Both on- and off-axis simulations were performed to see what impact they have on grating lobe values. The effect of off-axis focusing on the temperature distributions becomes more evident at high steering angles. For the case where the steering angle was set to 4.75°, i.e., focus at (5, 0, 60) mm, the temperature distribution was calculated and plotted in Figures 3(a–c) as a distribution at the plane of interest, a cross section along the line a-a and a cross section along the line b-b, respectively. Those three figures show that the simulated temperature at the focal point was about 54°C, while the temperature elsewhere was kept below 41°C.
Figure 3 A temperature map (a) for a focus aimed at (5, 0, 60) mm and cross section temperatures (b) and (c) across the lines a-a and b-b, respectively, as a function of distances x and z, respectively.
2.2 Transducer construction
The 1.75-D array described in this chapter vibrates in the thickness mode, which means that ε33 is the permittivity value of interest. Although Lead zirconate-titanate (PZT-5H) has a higher permittivity, which would lead to a higher capacitance, it cannot handle the large power that is used in FUS. PZT-4 and PZT-8 are good candidates concerning power, with an advantage for PZT-8 over PZT-4. The capacitance of a certain element in the array depends on the thickness (which is constant for all elements), the permittivity (which is a material characteristic) and the surface area of that element. Since the areas of the elements of the array are small, this will result in a small capacitance and thus large element impedance.
PZT-8 can handle the large electrical power needed for tissue ablation, has an extremely high mechanical quality factor and extremely low loss factor. Thus PZT-8 material (TRS Ceramics, State College, PA, USA) was chosen at a frequency of 1.2 MHz and diced, in house, into 3 × 21 elements forming the complete array. The cuts were made by dicing the material 100% through its thickness with a kerf width of 300 μm using a dicing saw (Model 780, K & S-Kulick and Soffa Industries, Willow Grove, PA, USA) in our lab. For maximum acoustical power transfer from the individual elements to the tissue, two matching layers were designed and constructed. The thickness and material selection of the matching layers were designed based on the solution to a four-layer problem (transducer, first matching layer, second matching layer, and tissue), which ensured the required maximum power transfer. The acoustic impedance of the two matching layers (Z1 and Z2) was calculated using a criterion determined by Fraser:
Z1 = (Zpiezo)4/7 (Ztissue)3/7 (5)
Z2 = (Zpiezo)1/7 (Ztissue)6/7 (6)
Each of the two matching layers was designed for a quarter wavelength thickness. Accordingly, the thickness of the first and second matching layers was determined to be 0.396 and 0.429 mm, respectively. The first matching layer, mixed in-housed, was a 2:1, epoxy to silver mixture of Insulcast 501 (Insulcast, Roseland, NJ, USA) and 2–3 micron silver epoxy (Aldrich, Milwaukee, WI, USA), while the second matching layer was a SPURR (Spi Supplies, West Chester, PA, USA) four-part low viscosity material. For this array design (Figure 4), the specially machined, waterproof cylindrical applicator housing (30 mm diameter) was made from magnet compatible Delrin® (Dupont, Wilmington, DE, USA) at the Penn State engineering shop.
Figure 4 Photograph of the constructed, waterproof array with 7.0 m low capacitance cable that connected to the amplifier system.
Transmission line theory applies to the coaxial cables used in ultrasound applications. The simplest approximation for the transmission line is a lumped capacitance, where the total cable capacitance can be measured by multiplying the cable capacitance per meter, Cm, by the length of the cable L in meters. Since the load impedance (element impedance) is high, the cable has to be a low capacitance cable, which effectively means high cable impedance. A cable with a characteristic impedance of 75 Ω was found to be suitable
The 1.75-D array contains 63 elements. Since those elements are of different size, each one of them will have different electrical impedance depending on the surface area of that element which determines the capacitance of that element and thus the electrical impedance. The target is to match each one of these impedances to the common value of 50Ω∠0°. A simple LC (L = inductor, C = capacitor) impedance matching circuit was built for each of the 63 elements.
2.3 Exposimetry
To determine the acoustic field generated by the array, an automated computer controlled positioning system, which could translate a hydrophone (needle one with 0.5 mm diameter, Precision Acoustics, UK) throughout the acoustic field of the array placed in a water tank, was used. The transducer was submerged in water (room temperature, approximately 20°C) in a tank (120 × 50 × 52 cm3) made almost anechoic with sound absorbing rubber. A custom made degasser, built in-house, was used to reduce the dissolved oxygen content of the distilled water to 1–2 ppm to reduce cavitation. The system was controlled using a personal computer connected to a four-motor positioning system (Velmex Inc., Bloomfield, NY, USA) via the RS232 serial port and also connected, via the general purpose interface bus (GPIB), to a digital oscilloscope (Agilent 54622A, Agilent Technologies, Palo Alto, CA, USA) which recorded the voltage amplitudes detected by the hydrophone. Custom written, Quick Basic (Microsoft Corporation, Redmond, WA, USA) programs were used for automated control of the motors and data acquisition from the oscilloscope. Initially, multiple on-axis (i.e. where the focus is along the major z axis, zf) exposimetry experiments were performed. With the focus set to 0, 0, zf mm, zf was varied from 10 mm to 110 mm with a step size of 5 mm. To determine the repeatability or standard deviation of the focusing, 5–10 experiments were performed at each location. For off-axis studies (i.e., where the focus was not on z but aimed toward the x axis, xf), the focus was located at (xf, 0, 60) mm while the steering angle was adjusted to the desired value by choosing appropriate values for xf. The steering angle was varied from -12° to +12° with a step size of 2° in both x and y directions with multiple experiments (5–10) performed at each angle. In both the on-axis and off-axis experiments, the scanning step size was 0.5 mm while the scanning area was 40 × 40 mm2. The hydrophone voltage recordings were used to calculate the normalized intensities based on the pressures that were plotted as the mean and standard deviation of the results (x ± s.d.) and compared against the calculated values.
2.4 Ex vivo experiments
To test the ability of the array to generate lesions in non-perfused bovine tissue, the array was submerged 6 cm in water and aimed perpendicular to the surface of water. Fresh bovine tissue was obtained, placed in the water tank and held 2.5 cm in front of the array. For both on- and off-axis focusing, a single linear array was driven with an average electrical power of 1 Watt per element for six to seven minutes.
3. Results
To compare experimental and theoretical results, more than fifty exposimetry experiments were performed throughout the desired ablation volume to determine the focusing capability of the array. As an example of a typical exposimetry result at the location (x, y, z) = (0, 0, 40) mm, Figure 5(a) shows a comparison plot along the z-axis of the calculated and experimental normalized intensities. Figure 5(b) plots similar theoretical and experimental data but instead along the x-axis for the same focus (0, 0, 40 mm). As can be seen for both plots, the theoretical intensity data correlated well with the experimental results. To evaluate the feasibility of the array to steer the focus, a typical three dimensional normalized intensity result from a focal point directed at 0, 0, 40 mm. The results were plotted as a mesh (Figure 6(a)) and contour (Figure 6(b)) with contour levels at 0, -1, -2, -3, -6 and -9 dB of the normalized intensity with the grating lobe levels at about -9.0 dB or less.
Figure 5 Comparison of calculated and experimental normalized intensities for a focus at 0, 0, 40 mm plotted along the (a) z axis and (b) x axis.
Figure 6 Exposimetry results of the normalized intensity for off-axis focusing with the focal point aimed at 0, 0, 40 mm plotted as a (a) mesh or (b) contour with levels indicated at 0, -1, -2, -3, -6 and -9 dB. These results indicate acceptable grating lobes of less than -9 dB.
Ex vivo experiments were also performed to verify the ability of the array to generate lesions in bovine tissue. In one experiment, the array was turned on for about six minutes. After turning the array off, the bovine piece was cut at the position where the focal point was aimed. A lesion that has the dimensions of 1 cm × 0.3 cm was observed, as Figure 7(a) shows. An unmarked version of Figure 7(a) is shown in Figure 7(b) for better visualization of the lesion. During the experiment, thermocouples were used to monitor the temperature at two locations, the focal point and the grating lobe locations. The temperature recording, Figure 7(c), shows that the focal point temperature reached about 52°C, while the grating lobe temperature was kept below 40°C, as compared to the simulated values (using the BHTE) of 51.7°C and 40.2°C, respectively. In another experiment, the array was turned on for about seven minutes and then turned off. The observed lesion was approximately 1.3 cm × 0.38 cm in size, as shows in the marked picture of the lesion, Figure 8(a). The same picture, but without marking the lesion, is shown in Figure 8(b) for better visualization of the lesion. The temperature recordings for this experiment show that the temperature at the focal point position increased to reach about 49.5°C while the temperature at the grating lobe position was about 39°C at the end of the experiment, as shown in Figure 8(c). Although the sonication time for this experiment was seven minutes while it was six minutes for the experiment shown in Figure 8, the final temperature value at the focal point was less for the seven minute experiment as compared to the six minute one. This might be due to the uncertainty of the location of the thermocouples.
Figure 7 (a) A lesion, with dimensions of about 1 cm × 0.3 cm, generated by a six minute sonication time experiment (b) Temperature recordings at the locations of the focal point and grating lobe.
Figure 8 A marked (a) and unmarked (b) lesion, with dimensions of 1.3 cm × 0.38 cm, generated by a seven minute sonication time experiment, and the temperature recordings (c) at the locations of the focal point and grating lobe.
4. Discussion
Intracavitary ultrasound offers an attractive means of focused ultrasound treatment for BPH with significant advantages over other treatment methods due to the relatively short treatment time, its noninvasive nature and reduced complications. One compelling reason for using an intracavitary device with focused ultrasound is that the prostate is easily accessible via transrectal applicators, which allow for heating of the target volume in the prostate with minimal heating of normal tissue. Using phased arrays to electrically focus the ultrasound beam provides a controlled localized power deposition into tissue and reduces significantly the treatment time since the focus is electronically scanned instead of manually.
In designing this array, several issues were taken into account to address its application for BPH treatment. The dimensions of array were designed for an intracavitary rectal device. With appropriate housing, the array dimensions of 2.7 × 5.3 cm2 are suitable. Another issue concerning this design was the grating lobe level, which was reduced significantly by tapering the array element lengths.
To treat the prostate, the array was aimed toward the intended target volume, and the elements were driven at a calculated amplitude and phase to generate a single focal point with electrical steering.
The array can ablate a volume that lies in its steering volumes. Assuming that the volume to be ablated is 1 × 1 × 2 cm3 and that its center is 3 cm away from the array face, two techniques can be used as a treatment plan to ablate the whole volume; the first one is using a single focal point regime in which the target volume is divided into small volumes. The size of these small volumes is chosen based on the size of the lesion and the sonication time. Assuming that the lesion was a 4 mm long cigar shape with 2 mm diameter for a 10 second sonication, dividing the 1 × 1 × 2 cm3 volume into 5 × 5 × 5 points indicates that 125 sonications are needed to ablate the target based on a single 10 second sonication that is electronically steered between the 125 positions. Starting at the center of the target volume, a single focal point is generated there and then electronically steered 125 times to cover the whole volume. To avoid uncontrolled heat buildup and pre-focal heating, the switching between the focal points is done in a way that any two focal points consecutive in time are far away from each other in distance. By doing that, enough time is given to the pre-focal positions to cool down. A second technique to ablate a large volume is by generating multi-focal points at the same time. Dividing the array into three areas, each responsible for generating a single focal point, will result in reducing the overall treatment time by a factor of three. This technique is time efficient, but the drawback behind it is that the driving electrical power per unit area should be increased.
If the maximum possible steering angle is 13.5° in the transverse direction, as the case for this array design, attempting to focus outside this volume will add a significant amount of energy to the grating lobes which will cause an unwanted heating. This puts a limitation for the array if the target volume extends beyond the 13.5° limit.
When coupled with MR temperature mapping, FUS provides an efficient way to treat BPH and at the same time gives a quick feedback about the temperature distribution inside the prostate [8]. Although ultrasound imaging for the prostate has shown to give good results [5,17], the array described here was designed to accompany an MRI.
Similar to prostate cancer treatment with focused ultrasound, benign fibroadenomas in the breast are currently treated clinically using multiple sonications from a single element transducer, which is mechanically scanned [18], in conjunction with MRI for guidance of thermotherapy of the procedure [19]. Although the treatment has been shown to be effective, the process includes an unnecessary delay due to the mechanical scanning protocol. When closely spaced locations are targeted with focused ultrasound, thermal buildup results from the accumulation of neighboring sonications and the nearfield heating. As a result, a lengthy delay between sonications (cooling time) is required to reduce thermal buildup. Investigators have shown that a cooling time of 50–60 seconds was necessary to reduce the heat from the near neighbor sonications [20] however this can add considerable time to the procedure and initiate inaccuracies to the MR thermometry through patient motion. With phased arrays a focal pattern can be arranged such that there is enough time for the heat to dissipate by sonicating non-neighboring regions within the tumor [21]. A treatment planning routine can be plotted over the entire tumor region such that the volume is ablated through distant and non-adjacent ablations to avoid thermal buildup yet destroy the volume in the least amount of time. This research demonstrates the feasibility of electrically steered arrays that can be used to ablate tissue for the intended treatment of benign prostatic hyperplasia.
5. Conclusion
A 1.75-D ultrasound phased array, that can focus and steer in a 3-D representation of the prostate without the need to physically move the array, had been successfully built and tested for the treatment of prostate cancer and BPH. Previous focused ultrasound array designs were problematic since they required complex methods to move the focus, as for the case of annular arrays, or had linear (1-D) designs that were only capable of focusing along one axis. These drawbacks were the motivation to design a new array that can be used in FUS and at the same time be systematically controlled to reposition the focus throughout a specific volume with an acceptable level of grating lobes. Further improvement over this array design seems to be feasible due to recent developments in building focused transducers using piezocomposite technology [22]. A three-layer PZT-8 material may also be used to increase the capacitance and thus make it easier to electrically match the small elements.
6. Authors' contributions
The first author (Saleh) contributed 60% of the work done for this paper. He performed the computer simulations required for the study, participated in performing the experiments that were performed to verify the study, and participated in the drafting and revising of the article. The other Author (Smith) contributed 40% of work done. She came up with idea, participated in some of the experiments performed and participated in the drafting and revising of the article. So both authors have made substantial contributions, have been involved in writing the article and have given final approval for the final submitted version.
7. Acknowledgements
This work was supported by the Department of Defense Congressionally Directed Medical Prostate Cancer Research Program (DAMD17-0201-0124).
==== Refs
Barrett D Mayo Clinic on prostate health 2000 1 Rochester, Minn: Mayo Clinic; New York
DelaRosette JJ Zlotta AR Alternative Instrumental Treatments in BPH. Future Perspectives European Urology 1999 35 173 176 9933812 10.1159/000019839
Diederich CJ Hynenen K Induction of hyperthermia using an intracavitary multielement ultrasonic applicator, IEEE Trans Biomed Eng 1989 36 432 438 2714822 10.1109/10.18749
Hutchinson E Intracavitary ultrasound phased arrays for thermal therapies PhD thesis 1997 Massachusetts Institute of Technology
Mahoney K Fjield T McDannold N Clement G Hynynen K Comparison of modelled and observed in vivo temperature elevations induced by focused ultrasound: implications for treatment planning Phys Med Biol 2001 46 1785 1798 11474925 10.1088/0031-9155/46/7/304
Sanghvi NT Foster RS Bihrle R Casey R Uchida T Phillips MH Noninvasive surgery of prostate tissue by high intensity focused ultrasound: an updated report Eur J Ultrasound 1999 9 19 29 10099163 10.1016/S0929-8266(99)00010-5
Hurwitz MD Kaplan I Svensson GK Hansen J Hynynen K Feasibility and patient tolerance of a novel transrectal ultrasound hyperthermia system for treatment of prostate cancer Int J Hyperthermia 2001 17 31 37 11212878 10.1080/02656730150201570
Hutchinson EB Hynynen K Intracavitary ultrasound phased arrays for noninvasive prostate surgery IEEE Trans Ultrason Ferroelect Freq Control 1996 43 1032 1042 10.1109/58.542048
Sokka S Hynynen K The feasibility of MRI guided whole prostate ablation with a linear aperoidic intracvitary ultrasound phased array Phys Med Biol 2000 45 3373 3383 11098911 10.1088/0031-9155/45/11/319
Curiel L Chavrier F Souchon R Birer A Chapelon JY 1.5-D high intensity focused ultrasound array for non-invasive prostate cancer surgery IEEE Trans Ultrason Ferroelect Freq Control 2002 49 231 242 10.1109/58.985707
Saleh K Smith NB Design and evaluation of a 3 × 21 element 1.75 dimensional tapered ultrasound phased array for the treatment of prostate disease Materials Research Innovation 2004 8 121 124
Saleh K Smith NB Two Dimensional Ultrasound Phased Array Design for Tissue Ablation for Treatment of Benign Prostatic Hyperplasia International Journal of Hyperthermia 2004 20 7 31 14612311 10.1080/0265673031000150867
Buchanan MT Hynynen K Design and experimental evaluation of an intracavitary ultrasound phased array system for hyperthermia IEEE Trans Biomed Eng 1994 41 1178 1187 7851919 10.1109/10.335866
Zemanek J Beam behavior within the nearfield of a vibrating piston J Acoust Soc Am 1971 49 181 191
Pennes HH Analysis of tissue and arterial blood temperatures in the resting human forearm J Appl Physiol 1948 1 93 122
Nyborg WL Heat generation by ultrasound in a relaxing medium J Acoust Soc Am 1981 70 310 312
Chapelon J Ribault M Birer A Vernier F Souchon R Gelet A Treatment of localised prostate cancer with transrectal high intensity focused ultrasound Eur J Ultrasound 1999 9 31 38 10099164 10.1016/S0929-8266(99)00005-1
Hynynen K Pomeroy O Smith DN Huber PE McDannold NJ Kettenbach J MR imaging-guided focused ultrasound surgery of fibroadenomas in the breast: a feasibility study Radiology 2001 219 176 185 11274554
Quesson B de Zwart JA Moonen CT Magnetic resonance temperature imaging for guidance of thermotherapy J Magn Reson Imaging 2000 4 525 533 11042633 10.1002/1522-2586(200010)12:4<525::AID-JMRI3>3.0.CO;2-V
McDannold NJ Jolesz FA Hynynen KH Determination of the optimal delay between sonications during focused ultrasound surgery in rabbits by using MR imaging to monitor thermal buildup in vivo Radiology 1999 211 419 426 10228523
Daum DR Hynynen K Thermal dose optimization via temporal switching in ultrasound surgery IEEE Trans Ultrason Ferroelect Freq Control 1998 45 208 215 10.1109/58.646926
Fleury G Berriet R Le Baron O Huguenin B New piezocomposite transducers for therapeutic ultrasound Proceedings of the Workshop on MRI-Guided Focused Ultrasound Surgery: 19–21 June 2002; Cambridge, 2002 39
| 15963237 | PMC1175852 | CC BY | 2021-01-04 16:37:36 | no | Biomed Eng Online. 2005 Jun 17; 4:39 | utf-8 | Biomed Eng Online | 2,005 | 10.1186/1475-925X-4-39 | oa_comm |
==== Front
Cardiovasc DiabetolCardiovascular Diabetology1475-2840BioMed Central London 1475-2840-4-91598515710.1186/1475-2840-4-9ReviewThe central role of vascular extracellular matrix and basement membrane remodeling in metabolic syndrome and type 2 diabetes: the matrix preloaded Hayden Melvin R [email protected] James R [email protected] Suresh C [email protected] Department of Family and Community Medicine, University of Missouri School of Medicine Columbia, Missouri PO BOX 1140 Lk. Rd. 5–87 Camdenton, Missouri 65020 USA2 Department of Internal Medicine, University of Missouri School of Medicine Columbia, Missouri Health Sciences Center, MA410, DC043.00 Columbia, Missouri 65212 USA3 Department of Physiology and Biophysics, University of Louisville, School of Medicine 500 South Preston Street University of Louisville Louisville, Kentucky 40292 USA2005 28 6 2005 4 9 9 23 3 2005 28 6 2005 Copyright © 2005 Hayden et al; licensee BioMed Central Ltd.2005Hayden 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 vascular endothelial basement membrane and extra cellular matrix is a compilation of different macromolecules organized by physical entanglements, opposing ionic charges, chemical covalent bonding, and cross-linking into a biomechanically active polymer. These matrices provide a gel-like form and scaffolding structure with regional tensile strength provided by collagens, elasticity by elastins, adhesiveness by structural glycoproteins, compressibility by proteoglycans – hyaluronans, and communicability by a family of integrins, which exchanges information between cells and between cells and the extracellular matrix of vascular tissues.
Each component of the extracellular matrix and specifically the capillary basement membrane possesses unique structural properties and interactions with one another, which determine the separate and combined roles in the multiple diabetic complications or diabetic opathies.
Metabolic syndrome, prediabetes, type 2 diabetes mellitus, and their parallel companion (atheroscleropathy) are associated with multiple metabolic toxicities and chronic injurious stimuli. The adaptable quality of a matrix or form genetically preloaded with the necessary information to communicate and respond to an ever-changing environment, which supports the interstitium, capillary and arterial vessel wall is individually examined.
AtherosclerosisCollagenelastinproteoglycanstructural glycoproteinIntegrinoxidative stressredox stressMMPTIMPremodeling
==== Body
Background
A matrix may be defined as something within or from which something else originates, develops, or takes form. The extracellular matrix (ECM) is a post-natally developed mesenchyme and provides scaffolding and structural support for cells and organs. Additionally, it is capable of exchanging information with cells and thereby modulates a whole host of processes including development, cell migration, attachment, differentiation, and repair. The repairing aspect of the ECM allows it to play a crucial role in wound healing via its chemotactic, haptotactic, opsonic, and ultimate attachment properties.
Metabolic syndrome (MetS) and type 2 diabetes mellitus (T2DM), which are now considered to be of pandemic proportions are conditions associated with multiple metabolic toxicities (table 1) and chronic injurious stimuli (figure 1). When uncontrolled by chronic injurious stimuli, there is chronic activation of these above processes resulting in fibrosis, structural derangement, tissue or organ dysfunction, and ultimate failure as a result of loss of form – structure and function.
Table 1 The multiple metabolic toxicities of the A-flight-u Acronym
Multiple injurious stimuli responsible for the production of ROS.
A Angiotensin II (also induces protein kinase C – β isoform)
Amylin (hyperamylinemia) islet amyloid polypeptide toxicity
AGEs/AFEs (advanced glycosylation/fructosylation endproducts)
Apolipoprotein B
Antioxidant reserve compromised
Absence of antioxidant network
Aging
ADMA (Asymmetrical DiMethyl Arginine)
Adipose toxicity: Obesity toxicity – Lipid Triad (decreased HDL-C, increased triglycerides and small dense LDL-C)
Adipocytokine toxicity: Increased TNF alpha, Il-6, TGF beta, PAI-I and the increased hormones resistin, leptin and decreased adiponectin.
F Free fatty acid toxicity: Obesity toxicity – Lipid Triad
L Lipotoxicity – Hyperlipidemia – Obesity toxicity – Lipid Triad Leptin toxicity
I Insulin toxicity (endogenous hyperinsulinemia-hyperproinsulinemia) IGF-1 – (GH-IGF-I axis) toxicity: This may serve to increase bone metabolism within the media of the AVW
Inflammation toxicity
G Glucotoxicity (compounds peripheral insulin resistance) and induces reductive stress through the sorbitol/polyol pathway
Pseudohypoxia (increased NADH/NAD ratio)
H Hypertension toxicity
Homocysteine toxicity
hs-CRP
T Triglyceride toxicity: Obesity toxicity – Lipid Triad
U Uric Acid – Xanthine Oxidase toxicity: Uric acid is an antioxidant early in physiological range of atherogenesis and a conditional prooxidant late when elevated through xanthine oxidase enzyme and generation of ROS: thus generating the paradoxical (antioxidant → prooxidant):
URATE REDOX SHUTTLE
Endothelial cell dysfunction with eNOS uncoupling, decreased eNO and increased ROS
Figure 1 multiple injurious stimuli to the Endothelium, intima, media, and adventitia. The endothelial cell is exposed to multiple injurious stimuli consisting of: modified LDL-cholesterol, various infection insults (viral and bacterial), angiotensin II, hemodynamic stress, LPa, glucose, homocysteine, uric acid, Ca++, phosphorus, parathyroid hormone, and intimal redox stress or reactive oxygen species. These multiple injurious stimuli (A-FLIGHT-U) cause a chronic injury and a response to injury with resultant remodeling of the arterial vessel wall and in particular the ECM. In the MetS, prediabetes, and overt T2DM, these stimuli act in concert to result in this detrimental remodeling with structural-functional abnormalities and dysfunction. The endothelium and its BM act as the first line of defense and are therefore the first cell and matrix to be affected with resulting dysfunction and structural changes. MetS, prediabetes, and T2DM undergo an accelerated atherosclerosis we term atheroscleropathy. Oxidation, glycation, glycoxidation, or homocysteinylation must modify LDL-cholesterol for LDL-C to become atherogenic. Multiple injurious stimuli acting alone and synergistically to modify LDL-cholesterol with resultant matrix structural defects accelerating atherogenesis and angiogenesis are observed. Each layer of the arterial vessel wall is eventually affected by these injurious stimuli initially from the lumen outward (inside-out) and later in the process to effect the plaque vulnerability from the outside-in (adventitial layer) by an inducible set of custom delivery vessels called the vasa vasorum.
The Component quintology of the ECM
The ECM consists of the following quintet: basement membrane (BM), collagen, elastin, proteoglycans (glycosaminoglycans – GAGs) and hyaluronan, and structural – adhesive glycoproteins.
I. Basement membrane (BM): (intimal and capillary)
The BM is important for the physical support and cellular attachment of cells and maintenance of their structural integrity, thus allowing cells to create and maintain their own special environment and provides a filtering – sieving mechanism due to the strong anionic charges of its matrix.
The importance of the ECM and the thickened capillary BM in diabetes was brought to "prime time" attention of diabetologists and researchers in 1968 with the publication of a paper by Siperstein MD and colleagues [1]. During the decade of the 70s others became interested in this phenomenon of matrix expansion within the basement membrane [2-10]. Specifically, Williamson JR and Kilo C were very strong contributors to this exciting area of science and they contributed strongly to the concept that diabetics have "leaky" blood vessels and that glucose was toxic to the endothelial cell and instigated capillary BM thickening [11-29].
Throughout the decades of the 1970's and 1980's; a common vernacular terminology used to describe diabetes was the following:
"Diabetes is a basement membrane disease."
This terminology is infrequently used today even though it is a widely accepted concept (capillary BM thickening is an ultrastructural hallmark in diabetic patients). This review will focus on the importance of the remodeled thickened CBM and update each component of the ECM. Additionally, an attempt will be made to show how the ECM comes "preloaded" with multiple reparative mechanisms to undergo the morphological structural change of remodeling in response to the metabolic and pathobiomolecular mechanisms associated with MetS, prediabetes, and overt T2DM.
The BM is a specialized extracellular matrix, which provides support – cell regulatory and filtering – sieving functions. Endothelial cells and most other epithelial cells are capable of synthesizing their BM.
MetS, prediabetes and T2DM are characterized by perturbations of the arterial vasculature, especially the endothelium and capillary BM, which are integrally involved with profound cardiovascular and microvascular complications. The endothelium and its BM are the first line of defense against injurious stimuli at the vascular lumen and capillary bed and are responsible for the regulation of vascular tone, circulation, fluidity, coagulation, inflammatory responses, oxidative stress, and remodeling in response to injurious stimuli (figure 1, 2).
Figure 2 The basement membrane exploaded. This image expands – explodes the BM and demonstrates the importance of each of its components that are involved in the expansion and thickening of the BM in MetS, prediabetes, and T2DM. The BM is an integral part of the ECM and plays such an integrating role in the structural – functional changes associated with MetS, prediabetes, and T2DM. An integrin has been placed in this image to verify its important-integrating role in cell-cell, cell-matrix communication. This image demonstrates the existence of a shape and a form to the PAS+, hyaline staining, thickened BM in MetS, prediabetes, and T2DM.
The interactions of the endothelial cell, its endothelial capillary BM, and their associated ECM become major players in the developing complications of MetS, prediabetes, and T2DM and are the central issue of this review. Proposed mechanisms of increased ECM accumulation in the BM (table 2) are rooted in multiple metabolic toxicities and reactive oxygen species (ROS) associated with MetS and T2DM and have a multifactorial pathogenesis.
Table 2 Observations and proposed mechanisms of increased capillary bm thickening with appliations to the myocardial, intima, islet, neuronal unit, Endothelial, renal, retinal and skeletal capillary basement membranes (Increased synthesis and decreased degradation tips the balance to accumulation)
Observations Proposed mechanisms
Glucotoxicity: IGT postprandial, IFG, and overt T2DM Protein Kinase C (PKC) activation.
Altered integrin expression of podocyte and pericyte: (podocyte-pericyte loss)
Increased Synthesis Of: Type Iv Collagen TGF beta, VEGF, and possibly PDGF beta 1. All associated with PKC activation and induction of growth hormones – factors from glucotoxicity and → ROS.
Increased Maintenance of: type IV Collagen
AGE cross-linking of type IV collagen.
AGE – RAGE connection. Increased resistance to protease (MMP) degradation, allowing type IV collagen to accumulate.
Decreased Degradation. Decreased MMP-2 MMP-3
Decreased Degradation. Increased expression of TIMP-2
Increased ROS.
In general, ROS promotes ECM fibrosis under the influence of chronic injurious stimuli and is associated with the chronic inflammatory state. ROS
increases all aspects of type IV accumulation: Glucotoxicity → PKC activation, AGE cross-linking activation, Decreased eNOS and eNO activity resulting in increased MMP activation.*
*Comment: It seems whenever there is robust MMP activation the result is robust newly synthesized collagen, which is more susceptible to ROS oxidation and accumulation. In general, it is more difficult to degrade newly synthesized – oxidized collagen than non-oxidized collagen. Other observations include an association of decreased eNO and increased activation of MMPs and in a like manner when eNO is normal or elevated MMP activity is suppressed. This comment points to the importance of a healthy eNOS and eNO generating endothelium in order for the ECM to maintain homeostasis.
Endothelial nitric oxide synthase (eNOS) – Endothelial nitric oxide (eNO)
Accumulation of BM material in renal tubular cells, the endothelial capillary beds of the renal, retinal, neuronal unit, myocardial and skeletal muscle, and the arterial vasculature itself are at the very core of these disease processes and diabetic complications. Even though these remodeled BMs appear thickened on microscopic examination, they lose their filtering – sieving (permselectivity) function and become dysfunctional due to a leakiness of larger proteins (such as albumin and lipoproteins into the intima and sub-capillary interstitium) and inflammatory cells [1-29]. The remodeling of the endothelial BM may also make the endothelium more prone to erosion and thrombosis in patients with metabolic abnormalities.
II. Collagen is the most abundant protein in humans and provides the framework for all multi-cellular organisms. There are characteristic triplet repeats of amino acids in the collagen molecule consisting of glycine XY, which results in glycine being present in every third amino acid. The collagen molecule is formed by three polypeptide chains, which intertwine to form triple helical rope-like collagen fibrils. These fibrils are cross-linked by hydroxyl groups between alpha chains (a major contributor to their tensile strength) to form the collagen fiber and these fibers, in turn, form collagen bundles. Gaps in the collagen fibril give the cross-banding appearance of types I and II collagen fibers at a characteristic length of 67 nm when viewed by electron microscopy. In type III collagen there is a structurally beaded appearance instead of the characteristic cross-banding appearance observed in types I and II collagen.
The physical and tensile strength of collagens are typified by collagen type I (having the tensile strength of steel), which predominates in bones, tendons, skin, and mature scars, while type II collagen is thinner and predominates in cartilage, vitreous humor and nucleus pulposus. Type III collagen is found in organs requiring more plasticity such as blood vessels, heart, gastrointestinal tract, uterus, and the dermis.
Types I, II, and III collagen are the fibrillar – interstitial collagens and are the most abundant collagen types. They are important in diabetic remodeling fibrosis within the myocardium (cardiomyopathy), the tubulo-interstitium of the kidney (nephropathy-interstitiopahy), the intima (intimopathy) in atheroscleropathy, dermopathy, interstitial changes within the retina (retinopathy), and possibly the neuronal unit of neuropathy. In contrast, collagens IV, V, and VI are non-fibrillar or amorphous and are found in BMs and interstitial tissue.
Historically, BMs have been shown to be highly insoluble and possess a distinct stability against mechanical forces. These findings are correlated with the presence of large amounts of a collagenous protein, which differ from the fiber-forming fibrillar collagens type I through III and thus the term type IV collagen has emerged.
One very unique feature of type IV collagen is the presence of seven to eight cysteine residues, which are involved in intra and intermolecular disulfide bonds, which aid in the stabilization of this polymer. This presence of cysteine in type IV collagen is in contrast to mature fibrillar collagens type I, II, and III, as they lack a cysteine moiety.
Type IV collagen is found exclusively in BM and it does not form individual fibers with electron microscopic cross-banding like the other collagens but instead forms an amorphous polygonal matrix, which is associated with laminin and other matrix macromolecules to form the unique BM matrix (figure 2).
Turnover of type IV collagen is known to be very slow unless there are inciting injurious stimuli that activate the specific BM degrading enzymes: matrix metalloproteinase(s) MMP-2 (constitutive) and MMP-9 (inducible) [31-34].
III. Elastin is known to provide support and elasticity. This elasticity is important for many tissues and organs such as the blood vessels, heart, skin, lung, and uterus. Elastin is a 70-kd glycoprotein and constitutes the central core of elastic fibers. It is similar to collagen, in that it is rich in glycine and proline, but unlike collagen, it contains almost no hydroxylated amino acids. It is cross linked, but unlike most other proteins it does not form definite folds but rather oscillates between different states to form random coils. It is this cross-linked, random-coiled structure of elastin that determines the capacity of the elastic network to stretch and recoil. Fibrillin microfibrils (a unique glycoprotein microfibril) are stiffer reinforcing fibers in compliant tissues and have been recently identified to be associated with elastic fibers [35].
Elastin is not felt to be a primary component of the capillary BM and it is interesting to note that the capillary tuft of the glomerulus was found to be devoid of elastin and present only in the mesangial stalk and afferent and efferent arterioles [36]. This may be one of the reasons that the capillary BM of the glomerular tuft undergoes remodeling expansion and results in a thickening of its BM when exposed to increased volume or increased pressure (intraglomerular hypertension), which occurs in the MetS and early in the natural history of T2DM.
Elastin provides an elastic molecular recoil phenomenon to the ECM and this is why there is a distinct internal and external elastic lamina on either side of the medial vascular smooth muscle layer of the arterial vessel wall (figure 1).
IV. Proteoglycan(s) (PG) – (glycosaminoglycans – GAGs) and Hhyaluronan
PG and hyaluronan are ubiquitous and found within the intima. They are synthesized primarily by the vascular smooth muscle cell (VSMC), other cells of mesenchymal origin, and in BM by the endothelium.
PG consist of a core protein(s) covalently linked to one or more highly sulfated polysaccharide chains termed glycosaminoglycans (GAGs). These molecules are highly diverse with multiple combinations of core proteins and polysaccharide chains. Examples are: heparan sulfate proteoglycans (HSPG), chondroitin sulfate proteoglycans, keratan sulfate proteoglycans, and dermatan sulfate proteoglycans (table 4).
Table 4 A representation of proteoglycans I – IV present in vascular ECM
Family (location) Common name Function
I. Large (interstitial) Versican (CSPG) Compression resilience. Similar to Hyaluronan
II. Small (leucine-rich) Decorin
Biglycan
Lumican Collagen organization.
III. Basement membrane Recently discovered →
In diabetic renal BM the CSPG Bamacan may be substituted for the HSPG Perlecan Perlecan (HSPG)
Bamacan (CSPG) (galactosaminoglycan chains)
Agrin (HSPG) Anionic filtration barrier Binds growth factors
Neural tight junctions (Blood Brain Barrier) and in renal BM
IV. Cell surface (Plasma membrane).
I.- IV.
Present in vascular ECM The vascular SMC is the principal source for these vascular proteoglycans. Syndecan-1
Fibroglycan
N-Syndecan (HSPG)
Ryudocan
Glypican (HSPG) ECM receptors, growth factor receptors. Binds coagulant enzymes, cytokines, and lipases.
V. Cerebral Proteoglycans
Others: Cerebrocan
Neurocan (CSPG)
Phosphacan (CSPG)
Brevican (CSPG)
_______
Aggrecan
Betaglycan Prominent in cartilage
They have multiple roles in regulating matrix structure such as cell growth and differentiation and permeability. They are highly sulfated and possess an anionic or negative charge, which makes them ideal to play the important role of selective filtering in the BM, especially in the renal glomerulus.
In this review we are mainly focused on the heparan sulfate proteoglycans (HSPG) (specifically perelecan of the capillary BM) and their role in BM filtering function due to the anionic charge provided by sulfation of the polysaccharide chains. In diabetes there is known to be decreased levels of perlecan in the glomerular BM and in the BM of endothelial, epithelial, and renal tubular cells, which would allow for the loss of an effective filtering function and these observations play a central role in the development of diabetic micro and macroalbuminuria.
This loss of filtering function is associated with the loss of perlecan and is also associated with the increased permeability of the microvessels throughout the vasculature affecting most of the diabetic complications and vasculopathies [37]. There are undoubtedly multiple causes for this decrease in perlecan and the multiple metabolic A-FLIGHT-U toxicities are related to this decrease in perlecan within the BM (table 1, 4). For example: Elevated levels of LDL-cholesterol and oxidized LDL-cholesterol, as well as, lysolecithin decrease not only perlecan core protein synthesis but also enhance heparan sulfate degradation by stimulating endothelial secretion of heparanase. ApoE and apoE-HDL, in contrast, increase perlecan core protein as well as sulfation of heparan sulfate [38]. Additionally non-esterified fatty acid or free fatty acid elevation has been shown to alter PG synthesis within the intima and contribute to LDL-cholesterol retention as well as allowing for increased permeability through an alteration in PG synthesis [39].
Recently it has been suggested that the PGs and the structural – adhesive glycoproteins and their associated glycosaminoglycans (GAG) side chains form a unit, which has been termed the glycocalyx. This unit may serve as a mechanosensor for both endothelial nitric oxide and prostacyclin responses of the endothelium to shear stress [40].
Syndecans form the largest group of HSPG on the endothelial surface and are set apart by being the only HSPG that penetrates the cytoplasm, allowing for an interaction with the cytosolic cytoskeleton (enabling an "outside in" mechanosensing capability). Glypicans form the second most common HSPG group, and have structural similarities to syndecan, typically differing only in the number of GAG attachment sites while perlecan remains in the basement membrane as discussed earlier [41].
This allows the glycocalyx to sense changes in shear stress from the outside and communicate with the G-protein receptors, including those that form a cytoplasmic bond with endothelial nitric oxide synthase and cytoskeletal elements like actin that can transduce physical forces throughout the cell to affect cellular function [42].
The nomenclature regarding PGs will undoubtedly undergo changes in the near future as an attempt to relate the specific PGs to their genomic origins. An attempt to aid in the classification of various PGs relating to the vascular ECM is presented at this point in time with some names of PGs in other tissues also being represented (table 4). Since there are present an unlimited number of possible interactions and combinations of the various PGs, there will undoubtedly become a new and improved nomenclature in the near future.
Other PGs such as versican, biglycan, and decorin accumulate in developing atherosclerotic and restenotic lesions. They contribute to plaque burden and influence cellular and extracellular events associated with the pathogenesis of vascular lesions, such as migration and proliferation, lipid metabolism and retention, and thrombosis.
Additionally, PGs also interact with other components of the ECM and contribute to their ability to regulate biomechanical properties of vascular lesions and even the ability of plaques to resist rupture.
IV. Hyaluronan (HA) is a huge molecule consisting of disaccharides stretched end-to-end, while lacking a core protein. It binds large amounts of water and forms a viscous hydrated gel, which gives the ECM turgor and allows it to resist compressive forces. Because of this unique ability it is found in abundance in cartilage of joints as it provides resilience and lubrication. It serves as a ligand for core proteins and is often a backbone for large proteoglycan complexes. It facilitates cell migration and inhibits cell-cell adhesion. It is synthesized primarily by the VSMC and is important in the development and progression of atherosclerosis, as well as, the process of post angioplasty restenosis. It communicates primarily through the integrin CD44 and is associated with angiogenesis. Hyaluronan is increased along with VSMC in atherosclerotic plaque erosion and is decreased in the vulnerable thin-cap atheroma associated with plaque rupture.
V. Structural – adhesive glypoproteins
Fibronectin, a large glycoprotein (approximately 450 kDa), is one of the first structural macromolecules to be deposited during embryonic development. It forms a primitive matrix that allows the initial organization to be replaced by the definitive, organ-specific matrix. The embryologic role of tissue fibronectin as the initial undifferentiated matrix is recapitulated in the early phases of injurious wound healing.
Fibronectin is a multifunctional adhesive protein whose primary function is to attach cells to a variety of matrices. Structurally, it consists of two polypeptide chains held together by two disulfide bonds. In addition to providing structural support it is associated with cell surfaces, pericellular matrices, and BMs. It is synthesized primarily by the cells of mesenchymal origin such as fibroblasts, monocytes, and endothelial cells.
Fibronectin binds to collagen, fibrin, and proteoglycans via specific domains and to cells via receptors that recognize the specific amino acid tripeptide RGD sequence (arginine-glycine-aspartic acid). This RGD integrin-binding motif is felt to be important for the haptotaxsis migration of cells within the ECM. A good example is the migration of capillary endothelial cells within the ECM during the process of capillary angiogenesis. Laminin (820 kDa) is the most abundant glycoprotein in BMs. This structural- adhesive glycoprotein binds to cells, heparan sulfate proteoglycans, and type IV collagen. Laminin is a hetero-trimeric polypeptide and appears as a cross-like structure with a single central polypeptide A chain and two flanking polypeptide B chains, which turns outward at right angles. This adhesive glycoprotein is felt to be important in cellular alignment (figure 2) [40-44].
Enactin and nidogen are different names for the same macromolecule. It is a dumbbell shaped structural – adhesive glycoprotein of ~150 kDa consisting of a 1217 amino acid residue. It has binding properties to both laminin and type IV collagen and thus can act as an adhesive bridge and is important in assembling these two major BM proteins [43,44] (figure 2). In addition to its role of assembling type IV collagen and laminin, Lebel SP et al. has been able to demonstrate it has permselective properties through an alteration of the anionic charges, as well as, promoting a morphological thickening of the BM as demonstrated in the enactin null transgenic mouse model [45].
Thus, these initially reparative mechanisms (the remodeling of the ECM), when stimulated by the chronic injurious stimuli associated with MetS, prediabetes, and T2DM result in devastating structural – functional complications.
Cellular integrins and ECM ligand binding
As mentioned previously the ECM comes genetically preloaded with a vast amount of information in addition to its scaffolding and structural supporting capabilities. Therefore, it is imperative that a brief discussion of the mechanisms allowing this communication of information be discussed.
The integrins comprise a molecular family of cell surface receptors and communicators that bind to the components of the ECM including collagen, laminin, and fibronectin. This interaction between the ECM and cellular integrins allow for a bi-directional (outside in and inside out) exchange of information facilitating a cell matrix communication (figure 3). A specific matrix ligand-binding site of the ECM proteins, consisting of a tripeptide sequence known as the RGD, binds to the integrins on the cell surface.
Figure 3 The integrins. This image portrays the integrin family of transmembrane molecules (receptors), which interact with the molecules of the ECM (ligands) and ligands associated with other cellular elements. Integrins are heterodimers, consisting of an alpha unit on the left made up of two disulfide bonded polypeptide chains. The beta unit on the right consists of a single polypeptide chain. Integrins bind to matrix ligand binding sties, which are specific amino acid sequences (usually 3–8) and at the top of this image the RGD (arginine, glycine, and aspartyl amino acids) matrix ligand – binding site is demonstrated. There are three domains: the extracellular domain – the transmembrane domain – the cytosolic domain, which interact with the cytosolic cytoskeletal proteins: Actin and Talin. These special transmembrane molecules allow for the "outside in" and "inside out" communication between cells and the ECM. Glucotoxicity affects integrin function through decreased perlecan within BMs and this may increase the susceptibility of endothelial cell dysfunction and demise (apoptosis), allowing for endothelial cell erosion and loss of endothelial cell stability upon its BM increasing the possibility of plaque erosion – rupture and thrombosis.
A simple analogy to the personal computer can be made, as integrins are the modern day computers of the cell, which span the plasma membrane with an extracellular domain, a transmembrane domain, and a cytosolic domain. These above properties allow this unique polypeptide – heterodimeric family consisting of 1 to 18 alpha chains and 1 to 8 beta chains (currently allowing for 24 different functional integrins to occur in humans) to connect to the outside world (i.e. the world wide web of the ECM). In addition to the individual 24 identified integrins identified there exists another additional mode of communication as they can cluster together just as multiple computers can be clustered to create a "google-like" search engine to expand exponentially the cells communication skills. RO Hynes [47] has presented interesting information that there are altered states of the integrins resting on the plasma membrane: The inactive state, whereintegrins rest flaccidly upon the cell surface and an active state, where the integrin stands erect and at full attention, and thus involve the previously mentioned clustering of integrins expanding their communication skills.
This marvelous communication system of integrins allows the cell to share information and communicate bi-directionally with the ECM. In order to survive as an organism the dictum of "no cell is an island" holds true. The cell must stay "connected" either through cell-cell adhesion (connexins) or to its matrix (through integrin-matrix ligand binding sites of the ECM) or undergo apoptosis – or more specifically anoikis (self suicide) triggered by loss of contact with the ECM [46-54].
The role of matrix metalloproteinases: MMPs and their inhibitors TIMPs
The Interstitial or fibrillar collagen types I-III are the primary collagens in the interstitial ECM. They are maintained and under the control of the family of zinc-dependent, redox sensitive, endopeptidases: matrix metalloproteinases (MMPs) (table 5). There is a delicate physiologic balance between the tearing down, rebuilding, tailoring, and sculpturing (remodeling) of the collagens within the ECM.
Table 5 Extracellular matrix degradation mechanisms: Focus on MMP
Type Examples Location Examples of Substrates
Serine protease Plasmin, urokinase, cathepsin G, TPA Pericellular Extracellular Fibrin, fibronectin, laminin, some proteoglycans
Cysteine protease Cathepsins B, D, H, L, N, and S Generally Cytosol-lysosomal Collagen, elastin. & proteoglycans
MMPs: Matrix Metalloproteinases Interstitial collagenases (MMP-1) Extracellular Collagens I, II, III, VII, and X
Basement Membrane Gelatinase A (MMP-2) 72 kDa Extracellular Collagens IV – BM, V, VII, and X Elastinolytic
Stromelysin-1 (MMP-3) Extracellular Collagens IV, III, V, and IX; laminin, fibronectin, elastin, proteoglycans
PUMP-1 (MMP-7) Extracellular Gelatin, fibronectin, laminin, collagen type IV, procollagenase, and proteoglycan core protein
Of Emerging Importance! [134] Neutrophil Elastase - collagenase
(MMP-8): Activated by CD-40 ligand Extracellular Macrophage
Endothelial Cell of Vasa Vasorum Collagens I, II, and III and proteoglycans Elastinolytic Internal Elastic Lamina. Activates MMP-2 -9
Basement Membrane Gelatinase B (MMP-9) 92 kDa Extracellular Collagens IV – BM V, VII, and plus Elastinolytic
Stromelysin-2 (MMP-10) Extracellular Similar to stromelysin-1
Stromelysin-3 (MMP-11) Extracellular Gelatin, fibronectin, and proteoglycans
Metalloelastase (MMP-12) Extracellular Elastin
Membrane type MMP MT-MMP (MMP-14) Cell surface Collagen IV, gelatin, and progelatinase A
Cardiac integrin MMP Disintegrin Metalloproteinse (DMP) Membrane type integrin matrix degrading MMP Endothelial Cardiac Integrin
MMPs show a wide range of specificity for various substrates, which include: native and partially degraded fibrillar collagens, basement membrane collagens, proteoglycans, elastin, and fibronectin. The ability of certain MMPs, such as MMP-2, MMP-3, MMP-9, and MMP-12, to hydrolyze elastin are of particular importance in terms of their effects on the vasculature not only within the arterial-vascular wall of vulnerable plaques but also within the vulnerable renal-mesangial stalk, which may result in plaque rupture and glomerular collapse, respectively.
MMP-2 or gelatinase A (72 kDa) and MMP-9 or gelatinase B (92 kDa) are the two enzymatic proteinases, which are primarily responsible for tearing down type IV collagen BM. These are synthesized by multiple vascular cell types including the endothelial cell and its supportive cell types: the pericyte and podocyte, VSMCs, renal mesangial cells, the fibroblast and the vascular fibroblast and myofibroblast, and the systemic-circulatory derived monocyte derived macrophage, as well as, the local tissue macrophage. The following two statements are necessary for a better understanding of their complicated roles in the ECM remodeling process.
MMP-2 may be considered to be a constitutive enzyme, while MMP-9 may be considered to be inducible in these various cell types.
The more robust the MMP signal and actions within the ECM, the more robust the repair mechanism of newly formed collagen synthesis.
There is a delicate balance between MMPs and their naturally occurring inhibitors (tissue inhibitors of matrix metalloproteinases or TIMPs). In the physiologic state the organism attempts at all times to achieve homeostasis. As a result there are checks and balances in the MMP – TIMP ratio. Additionally, it is important to understand that MMPs reside not only in the secreted form in the circulatory system, but also reside within the zymogen form within the ECM and remain in an inactive-latent or pro MMP state until they are activated by the tissue or urokinase plasminogen activator (tPA – uPA) driven plasmin. MMP-9 specifically can be activated by MMP-2 and MMP-3, as well as, membrane anchored MT1-MMP at the cell surface, converting proMMP-2 to active MMP-2 [55].
As will discussed later, the elevations of plasminogen activator inhibitor, elevated in MetS, prediabetes, and overt T2DM may have a devastating and detrimental effect on plasmin production and thus activation of latent or pro MMPs. This may play a role in matrix accumulation within the capillary BM, impaired fibrinolysis, impaired wound healing, and the impaired arteriogenesis associated with the vascular paradox.
MMP-9 has been shown to be elevated in T2DM and, in addition, the role of redox stress was shown to play an important role [56]. In addition to the endothelial cell, tissue and circulatory monocyte derived macrophage of chronic inflammation, the mesangial cell of the renal glomerular mesangium, the endothelial supporting podocyte and pericyte, and the cardiac myofibroblast each play an important role in synthesizing the inducible MMP-9. Because of the finding of an elevated MMP-9 in T2DM, both of these supportive cells may play an important role in the maintenance and the over-expression of type IV collagen in the endothelial CBM.
The integrin receptor for hyaluronan is CD-44 and it has been shown that MMP-9, in its active form, is associated with the cell surface via this CD-44 – hyaluronan integrin, which demonstrates just how connected the proteolytic MMP-9 enzyme activity is related to proteoglycans of the matrix and the cell surface membrane anchored MT1-MMP and the integrins. Also, the inactive or pro MMP-9 has a strong binding affinity via its gelatin-binding domain to bind to the alpha 2 (IV) chain of BM type IV collagen [55]. Recently, in our laboratory, we have been able to demonstrate decreased endothelial cell density with increased apoptosis of endothelial cells in the hearts of mice treated with alloxan vs. controls. Additionally, there was a decrease in NO and an increase in peroxynitrite and ROS in these same animals thus, linking the importance of cellular apoptosis, MMP-9 and redox stress. We then compared these findings of alloxan-induced diabetes in MMP-9 knockout mice to alloxan-induced diabetes in the wild type. Alloxan-induced diabetes MMP-9 -/- mice did not have induced apoptosis and did not have a decrease in endothelial cell density when compared to wild type alloxan-induced diabetes [57,58].
These findings may apply to the beta cell within the islet, as all cells require an integrin-matrix ligand binding for survival. The MMP-9 may also decrease the larger size amylin derived islet amyloid fibrils to the more intermediate size toxic amyloid particles and contribute to apoptosis as described by Janson et al. [59]
Death AK et al. have recently been able to demonstrate that MMP-1, MMP-2, and MMP-9 had an increased expression and activity by endothelial and monocyte derived macrophage cells under the influence of an elevated glucose in diabetic relevant concentrations. Additionally, they were able to show a decrease in MMP-3, while there was no significant effect on TIMP-1 expression. This dysregulation of MMP/TIMP system could lead to a net activation and a robust matrix degradation of type IV collagen within the basement membrane leading to a more robust laying down of new and reassembled type IV collagen as well as other BM matrix constituents [60]. This could also add to the vulnerability of vulnerable plaques, as well as, accelerating the underlying atherosclerotic process within the arterial vessel wall.
Tsilibary EC has been able to demonstrate an increase in type IV collagen, a decreased expression of MMP-2 and MMP-3, and an increased expression of TIMP-2 under high glucose conditions [34]. Their group has also been able to elegantly demonstrate a dysregulation of integrin expression, in that, under high glucose conditions the normal pattern of type IV collagen – integrin expression was shifted from alpha(3)beta(1) and alpha(2)beta(1) to a pattern of expression for alpha (v)beta(3) and alpha(5)beta(1). This alteration between the integrin – type IV interaction could certainly be playing a role in the loss of foot processes and the narrowed filtration slits of the supportive glomerular endothelial podocyte. Also, if there was robust MMP-9 production the integrin – matrix ligand binding could also become disrupted resulting in the loss of attachment of the podocyte to the ECs resulting in anoikis (apoptosis as a result of loss of attachment by integrin-matrix ligand binding sites: see previous section on integrins) [34].
The guardian angels of the capillary endothelial cell: the pericyte and podocyte
Capillary endothelial cells are supported and nourished by the pericyte in the systemic vascular bed and by the podocyte (the renal visceral epithelial cell) of the renal glomerular vascular bed. These cells play a similar supportive role for the endothelium and may be considered to be their guardian angels.
Each of these cells is very sensitive to oxidative – redox stress and the toxicity of hyperglycemia, be it intermittent (postprandial as in prediabetes) or sustained in overt T2DM. Once the protective effect of the pericyte and podocyte are lost by dysfunction or loss by apoptosis, the capillary endothelium becomes highly vulnerable to the multiple toxicities (A-FLIGHT-U toxicities) (table 1) and injurious stimuli (figure 1) associated with the MetS, prediabetes, and overt T2DM.
The initial structural findings demonstrated by electron microscopy were the loss of the foot process between the pericyte, podocyte, and capillary endothelial cells and eventually the loss of these two supportive cells, in part, through apoptosis. The vulnerability of these two specific cells are quite reminiscent of the beta cell within the islets of the pancreas in regards to their being unable to properly handle the elevated tension of oxidative – redox stress. These two unique cells support the capillary endothelium in synthesizing and maintaining their shared tri-laminar BM consisting of a lamina rara – lamina densa – lamina rara [61-65].
The role of advanced glycosylation endproducts (age) and ECM remodeling
AGE and the resultant cross-linking of proteins make the AGE-collagen adducts less likely to be degraded by MMPs (thus allowing for accumulation). The BM thickening associated with obesity and MetS are probably reversible, while the BM thickening associated with diabetes, hyperglycemia, ROS, and PKC activation are irreversible due to AGE formation and cross-linking.
Additionally, AGE directly quench endothelial nitric oxide and the oxidative – redox stress generated in their formation contributes to endothelial nitric oxide quenching, as well. This contributes to the endothelial cell dysfunction associated with the MetS, prediabetes, and T2DM. These affects of AGE not only contribute to the thickening of the capillary BM but also contribute to the endothelial dysfunction seen early on and contribute to its progressive deterioration as the underlying glycosylated type IV collagen accumulates within the BM [66-68]
The formation of AGE also will contribute to the accumulation of the interstitial fibrillar collagens responsible for the interstitiopathy associated with diabetic nephropathy and diabetic cardiomyopathy (figure 4).
Figure 4 Formation of age. The formation of AGE, as a result of chronic hyperglycemia, is complex and this figure demonstrates the steps and time frames involved in the formation of AGE and cross-linking of proteins. This complex process is reversible until the NH-R(n) 's are crosslinked.
Remodeling of ECM in metabolic syndrome
When T2DM is clinically diagnosed there may already be diabetic complications, such as: retinopathy (20%), nephropathy (8%), neuropathy (9%), atheroscleropathy – macrovascular disease (50%), and endotheliopathy (endothelial dysfunction: approaching 100 %) [69-72]. These clinical findings have lead clinicians to the hypothesis that either impaired glucose tolerance and impaired fasting glucose have preexisted for some time prior (in the 5–10 year range) to the diagnosis of overt T2DM or that the natural history of T2DM with its origins rooted in the MetS have contributed to the preexisting diabetic complications at the time of clinical diagnosis. Other possible explanations could be that polygenic T2DM is a vascular disease rooted in endothelial genetic defects and occurs as a result of interactions with environmental stressors such as over nutrition, obesity, and under exercise in the MetS with hyperglycemia being a late manifestation [73].
The MetS (figure 5) consists of four major components: I. Hyperinsulinemia, II. Hypertension, III, Dyslipidemia (Lipid Triad of increased triglycerides, increased small dense LDL-cholesterol, and decreased HDL-cholesterol) – Obesity, and IV.
Figure 5 The metabolic syndrome "Reloaded". MetS (Syndrome X) "reloaded" is a unique clustering of clinical syndromes and metabolic derangements. Reaven initially described the MetS in 1988. He initially discussed the four major determinants consisting of: I. Hypertension. II. Hyperinsulinemia. III. Hyperlipidemia (Dyslipidemia of elevated VLDL – triglycerides, decreased HDL-cholesterol, and elevated small dense atherogenic LDL-cholesterol). IV. Hyperglycemia or impaired glucose tolerance, impaired fasting glucose, or even overt T2DM and the central importance of insulin resistance and hyperinsulinemia. The important association of polycystic ovary syndrome (PCOS), hyperuricemia, fibrinogen, hsCRP, microalbuminuria, PAI-1, and more recently reactive oxygen species (ROS), NASH, and the damaging oxidative potential of Hcy and endothelial dysfunction have all contributed to a better understanding of this complicated clustering phenomenon. ROS, and those with a white background: Hyperuricemia, microalbuminuria, hyperhomocysteinemia, highly sensitive CRP, indicate the newer additions giving rise to the new terminology: Metabolic Syndrome Reloaded.
Hyperglycemia. The hyperglycemia – glucotoxicity section [G] of the A-FLIGHT-U toxicities are discussed elsewhere in this article and therefore the focus will be on the remaining four categories of the MetS: Volume, Pressure, Dyslipidemia – Obesity, and Hyperglycemia.
I. Volume
Hyperinsulinemia, hyperproinsulinemia, and hyperamylinemia all three independently and synergistically activate angiotensin II and increase renal blood flow resulting in renal hyperfiltration (Section [A] amylin toxicity and Ang II toxicity and section (I) insulin toxicity of table1). This results in both increased volume and pressure. Increased volume and hyperfiltration results in dilated glomerular capillaries, expansion of Bowman's space, glomerular hypertrophy and expansion and capillary BM thickening.
II. Pressure: hypertension
Hypertension is part and parcel of the MetS and results in vascular remodeling consisting of arteriolosclerosis of the BM especially in the afferent arteriole of the kidney (figure 6) as well as, remodeling of type I-III collagen of the renal tubular interstitium. Additionally there is arterial intimal remodeling in thehypertensive MetS patient [74].
Figure 6 Renal glomerular Remodeling. This image portrays a normal nephron unit on the left transitioning to an abnormal remodeled nephron unit with changes representative of diabetic nephropathy and changes of glomerulosclerosis. Left: Normal renal capillary glomerular and tubulo-interstitial structures. Transitioning to the Center of the image is the mesangial stalk with mesangial cell hyperplasia (yellow) and mesangial expansion with loss of foot processes of the podocyte (also termed visceral epithelium) (blue) and increasing thickness of the glomerular BM (red). Right: Increased capillary glomerular BM thickening (red) with atrophic podocytes and loss of foot processes of the podocyte (blue) to the capillary glomerular endothelial cell. Right: Also depicts tubulo-interstitial fibrosis with expansion of the peritubular (blue) extracellular matrix (fibrosis) with an increased thickening of the tubular BM (red). Just below the efferent (blue) arteriole is depicted hyaline arteriolosclerosis and just above the afferent arteriole (red) is depicted hyperplastic arteriolosclerosis with its characteristic "onion skin" like changes. The thickened BMs, arteriolar changes, and the mesangial expansion all are PAS+, hyaline staining, and contain large amounts of type IV collagen with increased laminin and fibronectin with concurrent decreased amounts of heparin sulfate proteoglycan (perlecan). Continuous with the proximal tubules (green) is the outer parietal epithelial cells (green), which constitutes the outer structure of Bowman's capsule and Bowman's space.
Hypertension is associated with oxidative stress of the arterial intima (associated with the production of ROS), which can activate protein kinase C (PKC) and transforming growth factor beta affecting both the BM of arterioles, as well as, the interstitium associated with the vascular intima and the interstitium of the renal tublular epithelium (figure 6).
III. Dyslipidemia – obesity
Lipid peroxidation results in ROS, which may activate the PKC beta isoform resulting transforming growth factor-beta activation and glomerular BM, matrix expansion, and interstitial renal tubular remodeling.
Obesity is associated with insulin resistance (IR) and compensatory hyperinsulinemia, hyperproinsulinemia, and hyperamylinemia, which are known to activate the renin angiotensin system and Ang II. Likewise, Ang II is known to induce ROS through the membraneous reduced nicotinamide adenine dinucleotide phosphate oxidase enzyme system and transforming growth factor-beta, which results in glomerular and renal tubular interstitial remodeling. We have been able to demonstrate in an (high-fat diet) obesity dog model an increase in arterial pressure, hyperinsulinemia, activation of the renin-angiotensin system, glomerular hyperfiltration, a trend to elevation of transforming growth factor beta and structural changes including: expansion of Bowman's capsule, increased mesangial matrix and thickening of the glomerular and tubular basement membranes and the number of dividing cells in the kidney [75].
The natural progressive history of T2DM with associated MetS and IR – associated compensatory hyperinsulinemia may result in a remodeling of the ECM prior to the diagnosis of overt T2DM. The volume, pressure, dyslipidemia, and obesity can also stimulate these same mechanisms and affect the intima, BM, and interstitial collagen within the myocardium resulting in a periodic acid Shiff positive staining of arterioles, capillary BMs, and the myocardial interstitium, in addition too, the remodeling of collagens type I-III of the interstitial matrix in target organs including the myocardium [76].
IV. Hyperglycemia impaired glucose tolerance (IGT), impaired fasting glucose, and overt T2DM are discussed later in the section entitled: Central role for protein kinase C beta isoform. Hyperglycemia mechanisms.
Remodeling of ECM in diabetic complications
Regarding macrovascular disease, Norhammer A et al., were able to demonstrate that 70% of patients with an acute myocardial infarction have either diabetes or IGT [77]. This elevated association of IGT and or diabetes points to atheroscleropathy and macrovascular disease. Cardiologists have noted a strong correlation of acute coronary syndromes and diabetes or IGT for some time and this study now validates their clinical suspicions. This information provides the clinician with an opportunity to possibly reverse the progressive nature of macrovascular disease in these patients by aggressive treatment through weight loss and exercise or pharmacological intervention to treat the underlying MetS state of IR and possibly the progressive beta cell dysfunction with resultant accelerated atherosclerosis and macrovascular disease. This needs to be accomplished in concert with the primary care clinician.
Regarding microvascular pathology, diabetes is the leading cause of blindness, end-stage renal disease, and a variety of debilitating neuropathies. Diabetic patients are the fastest-growing group of renal dialysis and transplant recipients, and in the USA, their 5-year survival rate is only 21 percent, which is worse than all forms of cancer combined. Over 60% of diabetic patients suffer from neuropathy, which accounts for 50% of all nontraumatic amputations in the USA [78].
A central role for protein kinase C beta isoform (PKC)
Each of the microvascular diabetic complications share a common microvascular metabolic signaling pathway through activation of PKC (figure 7) [79,80]. ROS and PKC play such important roles in each of the microvascular diabetic complications in both T1DM and T2DM. Hyperglycemia has been thought to be the stimulus for activation of PKC and the subsequent complications.
Figure 7 Pkc activation. This figure demonstrates the multiple deleterious actions and mechanisms of PKC beta II isoform on cellular function and ECM remodeling and BM thickening.
Hyperglycemia-induced mechanisms that may induce vascular dysfunction in specific sites of diabetic microvascular damage include the following:
1. Increased polyol pathway flux.
2. Altered cellular redox state with elevations of ROS.
3. Increased formation of diacylglycerol
4. Subsequent activation of specific PKC isoforms
5. Accelerated nonenzymatic formation of AGE and the AGE-RAGE connection. Activation of the receptor for AGE plays an important role.
6. Elevations of ROS.
Each of these mechanisms may contribute to the known pathophysiologic features of diabetic complications by a number of mechanisms, including the upregulation of cytokines and growth factors. Recently Brownlee M et al. has demonstrated that hyperglycemia (glucotoxicity) results in the formation of ROS, which then activate the deleterious PKC mechanism (figure 7) [81].
While the above may help to understand the increase in microvascular disease, it is felt that macrovascular disease may be an earlier occurrence and more deeply rooted in IR than from the later onset of hyperglycemia. While, at the same time a better understanding of the above numbered mechanisms play such an important role of accelerating and destabilizing atherosclerotic vulnerable plaques in the diabetic patient [82].
Conclusion
The adaptability of the ECM and its individual components in response to an ever changing environment including its response to multiple injurious stimuli resulting in an oxidative – redox stress resulting in an excess of ROS, known to be present in MetS and T2DM, allows tissues and organs to survive. However, this adaptability – survival mechanism results in a change in form and structure resulting in fibrosis or scarring, which results in abnormal function or disease. The ECM with its communications skills enhanced through a family of cellular integrins allows for information to be exchanged in order to adapt to its ever-changing environment. This review has focused on MetS, prediabetes, T2DM, and atheroscleropathy in an effort to better understand these mechanisms in the clustering of clinical syndromes (MetS) and the specific disease state of T2DM, which are each tightly associated with the current epidemic of obesity – diabesity and a genetic predisposition of a large number of patients in order to expand our current database of knowledge.
A central theme to the injury process regarding gene activation and transcription of various factors in an attempt to respond to the multiple injurious stimuli can be related to each organ, tissue, and cell, in that, whenever there is injury the cell recapitulates its embryonic genetic memory in an attempt to heal through growth (re-growth), differentiation, development, and repair. As a result of this chronically activated wound healing mechanism, which allows for survival; we as clinicians and researchers in this field of study must constantly review and expand our knowledge in an attempt to alter the wound healing – ECM response in a manner to decrease the morbidity and mortality and the progressive nature of MetS, prediabetes, T2DM, atheroscleropathy, and their associated complications.
List of abbreviations
AGE: advanced glycosylation endproducts
BM(s): basement membrane(s)
CBM: capillary basement membrane
ECM: extracellular matrix
GAG(s): glycosaminoglycan(s)
HSPG: heparan sulfate proteoglycans
IGT: impaired glucose tolerance
IR: insulin resistance
MetS: metabolic syndrome
MMP(s): matrix metalloproteinase(s)
PG: proteoglycan
PKC: protein kinase C
RAGE: receptor for advanced glycosylation endproducts
RGD: arginine-glycine-aspartic acid sequence
ROS: reactive oxygen species
T2DM: type 2 diabetes mellitus
TIMP: tissue inhibitor of metalloproteinase
Competing interests
MRH, SCT, and JRS: None
Authors' contributions
MRH conceived the idea to write this paper, MRH, SCT, and JRS collaborated to write and edit this manuscript equally.
Table 3 Components of the basement membrane: See Figure 2
Component Constituent Chains Molecular Composition Function
Type IV Collagen: alpha 1(IV), alpha 2(IV)
alpha 3(IV)
alpha 4(IV)
alpha 5(IV) Three alpha chains
Structure:
Polygonal shaped Network structure Provides a structural-lattice base for the attachment of other BM macromolecules such as HSPG, laminin, enactin and Fn.
Perlecan:
Heparan sulfate proteoglycan (HSPG):
Proteoglycan (PG) Polypeptide chain, side chains of GAGs Protein Core GAG side chains Highly anionic sulfated.
Structure:
Multiple globular protein core with multiple polypeptide chains. See figure 2. Electrostatic charge important for filtering. Especially in renal glomerulus.
Enactin – Nidogen: [31]
Structural – Adhesive Glycoprotein Single polypeptide chain Structure:
Dumbbell-shaped sulfated glycoprotein Bridges Laminin and Type IV collagen.
Important in assembly of the BM and changes in permselectivity properties.
Fibronectin (Fn):
Structural – Adhesive Glycoprotein Two polypeptide chains connected by two disulfide bridges. Structural glycoprotein One of the most primitive ECM macromolecules: The first to be deposited in the embryo. Parallel to V-shaped joined by two disulfide bonds. Connecting cells with other components of the ECM, which integrates the cell into a functional unit. Very important in wound healing.
Laminin:
The most abundant glycoprotein in BMs. Structural – Adhesive Glycoprotein
CABLIN: NEW
Capillary Basement membrane lamina A, B1, B2
First unique protein of the capillary basement membrane One A and two B chains. Structure: Cruciform shape
Rod like structure found only in the lamina rara of capillaries Cell attachment Assembly of the BM Stabilization of type IV Collagen
Cell-matrix attachment providing stability to the basement membrane
See Figure 2
Table 6 The positve protective effects of Endothelial Nitric Oxide Synthase (Enos) and Endothelial Nitric Oxide (eNO)
The positve protective effects of eNOS – eNO
1. Promotes vasodilatation of vascular smooth muscle.
2. Counteracts smooth muscle cell proliferation.
3. Decreases platelet adhesiveness.
4. Decreases adhesiveness of the endothelial layer to WBCs (monocytes). Thus, the .... "Teflon effect".
5. Anti- inflammatory.
6. Anti- oxidant. It scavenges reactive oxygen species, locally.
Acts as a chain – breaking antioxidant to scavenge ROS.
7. Anti- fibrotic. When NO is normal or elevated MMPs are low and conversely if NO is low MMPs are elevated and active. MMPs are redox sensitive.
8. NO has diverse anti-atherosclerotic actions on the arterial vessel wall: including antioxidant effects by direct scavenging of ROS – RNS acting as chain breaking antioxidants and anti-inflammatory effects
Acknowledgements
The authors wish to acknowledge the large body of work by Drs. Siperstein, Kilo, and Williamson and their colleagues. They have contributed greatly to extend the knowledge of diabetes and especially the important role of the basement membrane during the decades of the 70s and the 80s. They have shown us the way in this field of research.
==== Refs
Siperstein MD Unger RH Madison LL Studies of muscle capillary basement membranes in normal subjects, diabetic, and prediabetic patients J Clin Invest 1968 47 1973 1999 5675423
Siperstein MD Unger RH Madison LL Further electron microscopic studies of diabetic microangiopathy Adv Metab Disord 1970 1 261 5446895
Merimee TJ Siperstein MD Fineberg SE McKusick VA The microangiopathic lesions of diabetes mellitus: an evaluation of possible causative factors Trans Assoc Am Physicians 1970 83 102 112 5505385
Merimee TJ Siperstein MD Hall JD Fineberg SE Capillary basement membrane structure: a comparative study of diabetics and sexual ateliotic dwarfs J Clin Invest 1970 49 2161 2164 5480844
Siperstein MD Capillary basement membranes and diabetic microangiopathy Adv Intern Med 1972 18 325 344 4197108
Feingold KR Browner WS Siperstein MD Prospective studies of muscle capillary basement membrane width in prediabetics J Clin Endocrinol Metab 1989 69 784 789 2778035
Siperstein MD Diabetic microangiopathy, genetics, environment, and treatment Am J Med 1988 85 119 130 3057889 10.1016/0002-9343(88)90404-4
Klein RF Feingold KR Morgan C Stern WH Siperstein MD Relationship of muscle capillary basement membrane thickness and diabetic retinopathy Diabetes Care 1987 10 195 199 3582080
Feingold KR Lee TH Chung MY Siperstein MD Muscle capillary basement membrane width in patients with vacor-induced diabetes mellitus J Clin Invest 1986 78 102 107 3722372
Siperstein MD Diabetic microangiopathy and the control of blood glucose N Engl J Med 1983 309 1577 1599 6656851
Kilo C Vogler N Williamson JR Muscle capillary basement membrane changes related to aging and to diabetes mellitus Diabetes 1972 21 881 905 4558085
Williamson JR Vogler NJ Kilo C Early capillary basement membrane changes in subjects with diabetes mellitus Adv Metab Disord 1973 2 363 371 4720373
Fajans SS Williamson JR Weissman PN Bogler NJ Kilo C Conn JW Basement membrane thickening in latent diabetes Adv Metab Disord 1973 2 393 399 4720376
Koenig RJ Peterson CM Kilo C Cerami A Williamson JR Hemoglobin AIc as an indicator of the degree of glucose intolerance in diabetes Diabetes 1976 25 230 232 1254113
Williamson JR Rowold E Hoffman P Kilo C Influence of fixation and morphometric technics on capillary basement-membrane thickening prevalence data in diabetes Diabetes 1976 25 604 613 819319
Williamson JR Kilo C Current status of capillary basement-membrane disease in diabetes mellitus Diabetes 1977 26 65 73 318626
Ganda OP Soeldner JS Gleason RE Smith TM Kilo C Williamson JR Monozygotic triplets with discordance for diabetes mellitus and diabetic microangiopathy Diabetes 1977 26 469 479 192616
Kilo C Williamson JR Basement membranes in muscle capillaries Adv Exp Med Biol 1979 124 133 139 506830
Olson ND Nuttall FQ Sinha A Kilo C Williamson JR Thin muscle capillary basement membranes in myotonic dystrophy Diabetes 1979 28 686 689 446922
Chang K Uitto J Rowold EA Grant GA Kilo C Williamson JR Increased collagen cross-linkages in experimental diabetes: reversal by beta- aminopropionitrile and D-penicillamine Diabetes 1980 29 778 781 7439537
Tilton RG Hoffmann PL Kilo C Williamson JR Pericyte degeneration and basement membrane thickening in skeletal muscle capillaries of human diabetics Diabetes 1981 30 326 334 7202865
Uitto J Perejda AJ Grant GA Rowold EA Kilo C Williamson JR Glycosylation of human glomerular basement membrane collagen: increased content of hexose in ketoamine linkage and unaltered hydroxylysine-O-glycosides in patients with diabetes Connect Tissue Res 1982 10 287 296 6218960
Williamson JR Kilo C Capillary basement membranes in diabetes Diabetes 1983 32 96 100 6400674
Ganda OP Williamson JR Soeldner JS Gleason RE Kilo C Kaldany A Miller JP Garovoy MR Carpenter CB Muscle capillary basement membranewidth and its relationship to diabetes mellitus in monozygotic twins Diabetes 1983 32 549 556 6685073
Barnett AH Spiliopoulos AJ Pyke DA Stubbs WA Rowold E Hoffmann P Faller A Kilo C Miller JP Williamson JR Muscle capillary basement membrane in identical twins discordant for insulin-dependent diabetes Diabetes 1983 32 557 560 6685074
Tilton RG Faller AM Burkhardt JK Hoffmann PL Kilo C Williamson JR Pericyte degeneration and acellular capillaries are increased in the feet of human diabetic patients Diabetologia 1985 28 895 900 4092858 10.1007/BF00703132
Tilton RG LaRose LS Kilo C Williamson JR Absence of degenerative changes in retinal and uveal capillary pericytes in diabetic rats Invest Ophthalmol Vis Sci 1986 27 716 721 3700020
Rogers DG White NH Santiago JV Miller JP Weldon VV Kilo C Williamson JR Glycemic control and bone age are independently associated with muscle capillary basement membrane width in diabetic children after puberty Diabetes Care 1986 9 453 459 3769715
Williamson JR Tilton RG Chang K Kilo C Basement membraneabnormalities in diabetes mellitus: relationship to clinical microangiopathy Diabetes Metab Rev 1988 4 339 370 3292174
Tsilibary EC Microvascular basement membranes in diabetes mellitus J Pathol 2003 200 537 54 12845621 10.1002/path.1439
Hollertz O Sulphur: the vulnerable factor X in atherosclerosis Med Hypotheses 2002 59 35 38 12160678 10.1016/S0306-9877(02)00193-7
Yurchenco PD Schittny JC Molecular architecture of basement membranes FASEB J 1990 4 1577 1590 2180767
Weber S Dolz R Timpl R Fessler JH Engel J Reductive cleavage and reformation of the interchain and intrachain disulfide bonds in the globular hexameric domain NC1 involved in network assembly of basement membrane collagen (type IV) Eur J Biochem 1988 175 229 233 3402452 10.1111/j.1432-1033.1988.tb14188.x
Timpl R Structure and biological activity of basement membrane proteins Eur J Biochem 1989 180 487 502 2653817 10.1111/j.1432-1033.1989.tb14673.x
Sherratt MJ Baldock C Haston JL Holmes DF Jones CJ Shuttleworth CA Wess TJ Kielty CM Fibrillin microfibrils are stiff reinforcing fibres in compliant tissues J Mol Biol 2003 332 183 193 12946356 10.1016/S0022-2836(03)00829-5
Sterzel RB Hartner A Schlotzer-Schrehardt U Voit S Hausknecht B Doliana R Colombatti A Gibson MA Braghetta P Bressan GM Elastic fiber proteins in the glomerular mesangium in vivo and in cell culture Kidney Int 2000 58 1588 1602 11012893 10.1046/j.1523-1755.2000.00320.x
Conde-Knape K Heparan sulfate proteoglycans in experimental models of diabetes: a role for perlecan in diabetes complications Diabetes Metab Res Rev 2001 17 412 421 11757076 10.1002/dmrr.236
Pillarisetti S Lipoprotein modulation of subendothelial heparan sulfate proteoglycans (perlecan) and atherogenicity Trends Cardiovasc Med 2000 10 60 65 11150731 10.1016/S1050-1738(00)00048-7
Olsson U Bondjers G Camejo G Fatty acids modulate the composition of extracellular matrix in cultured human arterial smooth muscle cells by altering the expression of genes for proteoglycan core proteins Diabetes 1999 48 616 622 10078565
Florian JA Kosky JR Ainslie K Pang Z Dull RO Tarbell JM Heparan sulfate proteoglycan is a mechanosensor on endothelial cells Circ Res 2003 93 e136 e142 14563712 10.1161/01.RES.0000101744.47866.D5
Bernfield M Kokenyesi R Kato M Hinkes MT Spring J Gallo RL Lose EJ Biology of the syndecans: a family of transmembrane heparan sulfate proteoglycans Annu Rev Cell Biol 1992 8 365 393 1335744 10.1146/annurev.cb.08.110192.002053
Rapraeger AC Syndecan-regulated receptor signaling J Cell Biol 2000 149 995 998 10831602 10.1083/jcb.149.5.995
Yurchenco PD Schittny JC Molecular architecture of basement membranes FASEB J 1990 4 1577 1590 2180767
Timpl R Structure and biological activity of basement membrane proteins Eur J Biochem 1989 180 487 502 2653817 10.1111/j.1432-1033.1989.tb14673.x
Lebel SP Chen Y Gingras D Chung AE Bendayan M Morphofunctional studies of the glomerular wall in mice lacking entactin-1 J Histochem Cytochem 2003 51 1467 1478 14566019
Danen EH Sonnenberg A Erratum: Integrins in regulation of tissue development and function J Pathol 2003 201 632 641 J Pathol; 200: 471–480 14648669 10.1002/path.1472
Hynes RO Integrins: bidirectional, allosteric signaling machines Cell 2002 110 673 687 12297042 10.1016/S0092-8674(02)00971-6
van der Flier A Sonnenberg A Function and interactions of integrins Cell Tissue Res 2001 305 285 298 11572082 10.1007/s004410100417
Goldsmith EC Borg TK The dynamic interaction of the extracellular matrix in cardiac remodeling J Card Fail 2002 8 S314 S318 12555138 10.1054/jcaf.2002.129258
Arnaout MA Integrin structure: new twists and turns in dynamic cell adhesion Immunol Rev 2002 186 125 140 12234368 10.1034/j.1600-065X.2002.18612.x
Tanaka Y [Adhesion molecules: relevance to basic and clinical research] J UOEH 2001 23 421 429 11789144
Brown E Hogg N Where the outside meets the inside: integrins as activators and targets of signal transduction cascades Immunol Lett 1996 54 189 193 9052876 10.1016/S0165-2478(96)02671-5
Halvorson MJ Coligan JE Sturmhofel K The vitronectin receptor (alpha V beta 3) as an example for the role of integrins in T lymphocyte stimulation Immunol Res 1996 15 16 29 8739562
Faull RJ Ginsberg MH Inside-out signaling through integrins J Am Soc Nephrol 1996 7 1091 1097 8866399
Fridman R Toth M Chvyrkova I Meroueh SO Mobashery S Cell surface association of matrix metalloproteinase-9 (gelatinase B) Cancer Metastasis Rev 2003 22 153 166 12784994 10.1023/A:1023091214123
Uemura S Matsushita H Li W Glassford AJ Asagami T Lee KH Harrison DG Tsao PS Diabetes mellitus enhances vascular matrix metalloproteinase activity: role of oxidative stress Circ Res 2001 88 1291 1298 11420306
Camp TM Tyagi SC Senior RM Hayden MR Tyagi SC Gelatinase B(MMP-9) an apoptotic factor in diabetic transgenic mice Diabetologia 2003 46 1438 1445 12928773 10.1007/s00125-003-1200-y
Hayden MR Tyagi SC Intimal redox stress: Accelerated atherosclerosis in metabolic syndrome and type 2 diabetes mellitus. Atheroscleropathy Cardiovasc Diabetol 2002 1 3 12392600 10.1186/1475-2840-1-3
Janson J Ashley RH Harrison D McIntyre S Butler PC The mechanism of islet amyloid polypeptide toxicity is membrane disruption by intermediate-sized toxic amyloid particles Diabetes 1999 48 491 498 10078548
Death AK Fisher EJ McGrath KC Yue DK High glucose alters matrix metalloproteinase expression in two key vascular cells: potential impact on atherosclerosis in diabetes Atherosclerosis 2003 168 263 269 12801609 10.1016/S0021-9150(03)00140-0
Nielsen BS Sehested M Kjeldsen L Borregaard N Rygaard J Dano K Expression of matrix metalloprotease-9 in vascular pericytes in human breast cancer Lab Invest 1997 77 345 355 9354769
Arihiro S Ohtani H Hiwatashi N Torii A Sorsa T Nagura H Vascular smooth muscle cells and pericytes express MMP-1, MMP-9, TIMP-1 and type I procollagen in inflammatory bowel disease Histopathology 2001 39 50 59 11454044 10.1046/j.1365-2559.2001.01142.x
von Luttichau I Djafarzadeh R Henger A Cohen CD Mojaat A Jochum M Ries C Nelson PJ Kretzler M Identification of a signal transduction pathway that regulates MMP-9 mRNA expression in glomerular injury Biol Chem 2002 383 1271 1275 12437116 10.1515/BC.2002.142
Asanuma K Shirato I Ishidoh K Kominami E Tomino Y Selective modulation of the secretion of proteinases and their inhibitors by growth factors in cultured differentiated podocytes Kidney Int 2002 62 822 831 12164864 10.1046/j.1523-1755.2002.00539.x
Nakamura T Ushiyama C Suzuki S Hara M Shimada N Ebihara I Koide H Urinary excretion of podocytes in patients with diabetic nephropathy Nephrol Dial Transplant 2000 15 1379 1383 10978394 10.1093/ndt/15.9.1379
Bucala R Tracey KJ Cerami A Advanced glycosylation products quench nitric oxide and mediate defective endothelium-dependent vasodilatation in experimental diabetes J Clin Invest 1991 87 432 438 1991829
Bucala R Tracey KJ Cerami A Advanced glycosylation products quench nitric oxide and mediate defective endothelium-dependent vasodilatation in experimental diabetes J Clin Invest 1991 87 432 438 1991829
Clancy RM Leszczynska-Piziak J Abramson SB Nitric oxide, an endothelial cell relaxation factor, inhibits neutrophil superoxide anion production via a direct action on the NADPH oxidase J Clin Invest 1992 90 1116 1121 1325992
Mudaliar SR Henry RR Management and prevention of diabetic complications Atlas of Clinical Endocrinology Series Editor: Korenman SG Diabetes Editor: Kahn Ronald C 2000 2 Blackwell Science Inc. Developed by Current Medicine, Inc., Philadelphia 81 94
Harris MI Harris MI, Cowie CC, Stern MP Summary Diabetes in America, NIH Publication No 95-1468 1995 Washington D.C: US Government Printing Office 1 14
Fagan TC Deedwania PC The cardiovascular dysmetabolic syndrome Am J Med 1998 105 77S 82S 9707273 10.1016/S0002-9343(98)00216-2
Garber AJ Vascular disease and lipids in diabetes Med Clin North Am 1998 82 931 948 9706127 10.1016/S0025-7125(05)70030-4
Hayden MR Tyagi SC Is type 2 diabetes mellitus a vascular disease (atheroscleropathy) with hyperglycemia a late manifestation? The role of NOS, NO, and redox stress Cardiovasc Diabetol 2003 2 2 12628022 10.1186/1475-2840-2-2
Sowers JR Insulin resistance and hypertension Am J Physiiol Heart Circ Physiol 2004 286 H1597 H1562 10.1152/ajpheart.00026.2004
Henegar JR Bigler SA Henegar LK Tyagi SC Hall JE Functional and structural changes in the kidney in the early stages of obesity J Am Soc Nephrol 2001 12 1211 1217 11373344
Hayden MR Tyagi SC Myocardial redox stress and remodeling in metabolic syndrome, type 2 diabetes mellitus, and congestive heart failure Med Sci Monit 2003 9 SR35 SR52 12883468
Norhammar A Tenerz A Nilsson G Hamsten A Efendic S Ryden L Malmberg K Glucose metabolism in patients with acute myocardial infarction and no previous diagnosis of diabetes mellitus: a prospective study Lancet 2002 359 2140 2144 12090978 10.1016/S0140-6736(02)09089-X
National Diabetes Data Group Diabetes in America 1995 2 Bethesda, Maryland, USA 1 730 NIH Publication no. 95-1468
Inoguchi T Battan R Handler E Sportsman JR Heath W King GL Preferential elevation of protein kinase C isoform beta II and diacylglycerol levels in the aorta and heart of diabetic rats: differential reversibility to glycemic control by islet cell transplantation Proc Natl Acad Sci U S A 1992 89 11059 11063 1438315
Koya D Jirousek MR Lin YW Ishii H Kuboki K King GL Characterization of protein kinase C beta isoform activation on the gene expression of transforming growth factor-beta, extracellular matrix components, and prostanoids in the glomeruli of diabetic rats J Clin Invest 1997 100 115 126 9202063
Nishikawa T Edelstein D Du XL Yamagishi S Matsumura T Kaneda Y Yorek MA Beebe D Oates PJ Hammes HP Giardino I Brownlee M Normalizing mitochondrial superoxide production blocks three pathways of hyperglycaemic damage Nature 2000 404 787 790 10783895 10.1038/35008121
Aronson D Rayfield EJ How hyperglycemia promotes atherosclerosis: molecular mechanisms Cardiovasc Diabetol 2002 1 1 12119059 10.1186/1475-2840-1-1
| 15985157 | PMC1175853 | CC BY | 2021-01-04 16:25:03 | no | Cardiovasc Diabetol. 2005 Jun 28; 4:9 | utf-8 | Cardiovasc Diabetol | 2,005 | 10.1186/1475-2840-4-9 | oa_comm |
==== Front
J Exp Clin Assist ReprodJournal of Experimental & Clinical Assisted Reproduction1743-1050BioMed Central London 1743-1050-2-81592704910.1186/1743-1050-2-8ResearchLow-dose aspirin does not improve ovarian stimulation, endometrial response, or pregnancy rates for in vitro fertilization Hurst Bradley S [email protected] Jennifer T [email protected] Paul B [email protected] Margaret A [email protected] Terry A [email protected] Michelle L [email protected] Department of Obstetrics and Gynecology, Carolinas Medical Center, 1000 Blythe Blvd, Charlotte, NC 28203, USA2005 31 5 2005 2 8 8 21 2 2005 31 5 2005 Copyright © 2005 Hurst et al; licensee BioMed Central Ltd.2005Hurst et al; licensee BioMed Central Ltd.This is an Open Access article distributed under the terms of the Creative Commons Attribution License (), which permits unrestricted use, distribution, and reproduction in any medium, provided the original work is properly cited.
Background
The purpose of this study is to determine if low-dose aspirin improved ovarian stimulation, endometrial response, or IVF pregnancy rates in our program.
Methods
Retrospective analysis of 316 consecutive IVF cycles from 1995 through 2001. Aspirin 80 mg daily was initiated at the start of luteal leuprolide in 72 cycles. The 244 controls received no aspirin during treatment.
Results
The live birth rate in aspirin users was 29%, slightly lower compared to 41% in the no aspirin control group (p = 0.07). Implantation rates were 21% with aspirin and 30% in the control population (p = 0.01). There was no difference in the maximal endometrial thickness between aspirin and non-aspirin groups. The two groups were similar regarding age, gonadotropin ampules, embryos, number of embryos transferred, prior parity, diagnosis, use of intracytoplasmic sperm injection, and stimulation protocol.
Conclusion
Low-dose aspirin was not beneficial to IVF patients in our program. Aspirin does not enhance endometrial thickness, augment the ovarian response, or improve pregnancy rates.
aspirinembryo transferendometriumfertilization in vitroinfertility
==== Body
Background
Numerous measures have been employed in an attempt to increase implantation and pregnancy rates in assisted reproduction. Aspirin has been utilized as one such potential therapy. This drug has been shown to increase uterine blood flow [1], hence clinicians have postulated that aspirin could improve the receptiveness of the endometrium, thereby increasing implantation and birth rates.
Our institution at one time used aspirin routinely during IVF cycles, based on the work of studies which showed that low-dose aspirin increased implantation and pregnancy rates in women undergoing IVF [2,3]. Contrary data from Urman and co-investigators found no improvement in IVF outcomes with low-dose aspirin [4]. Subsequently, the use of aspirin was stopped in our program early in 2000. Since conflicting results have been reported in the literature, we sought to compare pregnancy rates along with other IVF outcome variables retrospectively in the two groups of women (aspirin vs. non-aspirin) at our institution.
Methods
This study was a retrospective analysis of 316 consecutive IVF cycles from 1995 – 2001 at Carolinas Medical Center comparing women who were treated with low-dose aspirin versus those who did not receive aspirin treatment. Aspirin was used in all initial cycles from 1995, and excluded from most, but not all cycles beginning early in 2000 at the discretion of the attending physician. Demographic data including age, parity, cycle number, basal FSH, diagnosis, method of stimulation, and use of intracytoplasmic sperm injection was obtained from our database. For the purpose of this study, we divided method of stimulation into GnRH antagonists, long luteal leuprolide, and micro-dose flare. The infertility diagnoses were categorized into male factor, endometriosis, tubal factor, ovulatory dysfunction, unexplained, and other, which included uterine factors and immunological causes. The pregnancy and delivery rates were stable in our program from 1995 to 2001.
Seventy-two aspirin cycles were reviewed along with 244 non-aspirin cycles. For the aspirin cycles, 80 mg of aspirin daily was initiated at the start of down-regulation with luteal leuprolide. Aspirin was started on the first day of leuprolide in microdose flare stimulations. Patients were instructed to continue aspirin until they received the results of their pregnancy tests. The controls received no aspirin at any point during treatment. The outcome measures from the completed cycles were then reviewed. Of interest were the number of gonadotropin ampules used, endometrial thickness, number of eggs fertilized, number of embryos transferred, implantation rate, pregnancy rate, and live birth rate.
Statistics
The main independent variable was treatment with aspirin (yes/no). Demographic and clinical characteristics for each treatment group were reported and compared with two-tailed t-test, Wilcoxon Rank Sum test, Chi-Square or Fisher's Exact tests, as appropriate. The study outcomes were analyzed in two stages: the first with Chi-Square tests followed by a confirmatory analysis using a regression method generalized estimating equations (GEE). Further analysis of the outcomes assessed their association with aspirin treatment after controlling for other patient and clinical characteristics. The power of the study to determine a difference in pregnancy rates with and without aspirin based on previous studies was approximately 60–72% with an alpha of 0.05 [2,3].
Results and Discussion
There was no significant difference between age, previous pregnancy, infertility diagnosis, prior IVF, basal FSH, and method of stimulation between the aspirin and non-aspirin groups. (Table 1) More women in the non-aspirin group had been pregnant before (15.9% v. 9.7%) compared to the aspirin group, but this did not achieve statistical significance (p = 0.06).
Table 1 Demographic Data
Aspirin No aspirin p
Number patients 72 (23%) 244 (77%)
Age 34 ± 4 34 ± 4 0.7
Previously pregnant 7 (10%) 39 (16%) 0.06
Diagnosis
• Unexplained 4 (6%) 12 (5%)
• Male factor 23 (32%) 51 (21%)
• Endometriosis 9 (13%) 51 (21%)
• Tubal factor 13 (18%) 54 (22%)
• Ovulatory dysfunction 12 (17%) 24 (10%)
• Other 6 (8%) 17 (7%)
• Multiple diagnoses 4 (6%) 54 (22%)
Prior IVF 32% 22% 0.12
Basal FSH (mIU/mL) 7 ± 2 7 ± 8 0.14
Stimulation method
• Antagonist 0 2 (1%)
• Long luteal leuprolide 62 (86%) 195 (80%)
• Flare 8 (11%) 34 (14%)
Low-dose aspirin did not improve any IVF outcomes analyzed in this study, even though more embryos were transferred to women who used aspirin (p = 0.03) (Table 2). In fact, the pregnancy rate in aspirin users was 48%, slightly lower compared to non-users, 57% (p = 0.18). Clinical pregnancy rates were 45% and 54% for users and non-users, respectively. Live birth rates tended to be lower with aspirin, 29% and 41%, respectively (p = 0.07). Implantation rates were significantly lower in patients who received aspirin, 21% and 30%, respectively (p = 0.01). Maximal endometrial thickness was not improved with aspirin compared to non-aspirin controls (p = 0.26). The percentage of ICSI cycles was similar in each group, as was the number of eggs fertilized.
Table 2 Results
Aspirin No Aspirin P
Ampules (75 IU) 42 ± 15 44 ± 17 0.35
Endometrial thickness 12 ± 2 12 ± 2 0.26
ICSI 23 (32%) 67 (28%) 0.46
Oocytes fertilized 9 ± 4 9 ± 6 0.7
Embryos transferred 4 ± 1 3 ± 1 0.03
Pregnancy rate 48% 57% 0.18
Live birth rate 29% 41% 0.07
Implantation rate 21% 30% 0.01
Low-dose aspirin did not benefit IVF patients in our program. Aspirin therapy did not enhance endometrial thickness, augment the ovarian response, or improve pregnancy rates. The demographics were similar between the two groups of patients, with similarities in diagnosis, stimulation protocol, as well as number of ICSI cycles.
Our results conflict with several studies that have shown that aspirin is beneficial for infertility therapy. Rubenstein et al found that aspirin 100 mg starting in the luteal phase of the preceding cycle improved blood flow velocity, ovarian responsiveness, implantation and pregnancy rates in a randomized, controlled trial of 149 patients undergoing IVF compared to 149 placebo-treated controls [2,5]. Weckstein et al also found enhanced uterine blood flow and significantly higher implantation and clinical pregnancy rates with low-dose aspirin in women who had a thin endometrium undergoing embryo transfer from oocyte donation in a randomized controlled study [3].
Interestingly, endometrial thickness was not improved with aspirin. In an prospective, randomized insemination study of women with a thin endometrium undergoing insemination, aspirin improved the percentage of trilaminar endometrium and pregnancy rates from 9 to 18%, but not endometrial thickness or ultrasound flow patterns [6].
Waldenstrom et al randomized 1380 unselected IVF cycles on alternate days to receive aspirin 75 mg or no aspirin starting on the day of embryo transfer and continuing until 18 days after retrieval [7]. In this study, the live birth rate was 27% with aspirin and 23% in the control population, with an odds ratio 1.2 (95% CI 1.0–1.6). A non-controlled study found that IVF outcome was significantly improved when aspirin, heparin, and intravenous immunoglobulin therapy was administered to women with repeat IVF failures and antiphospholipid antibodies, but not to women with negative antiphospholipid antibodies [8]. Other studies have also found a beneficial effect with aspirin/heparin, and aspirin plus prednisolone in IVF patients [9-13]. In vitro studies have shown that aspirin attenuates placental apoptosis, and this could be a possible explanation of how aspirin is beneficial, even in the absence of endometrial or oocyte improvement [14]. Proponents of aspirin consider treatment to be a simple, inexpensive, and harmless means to improve IVF outcomes [7].
However, some studies have shown anticoagulation therapy to be ineffective, and sometimes detrimental, during IVF. A large randomized controlled trial of low-dose aspirin by Urman et al found no difference in implantation or pregnancy rates in patients undergoing ICSI [4]. A higher incidence of ectopic pregnancy was found in the aspirin group. A prospective, randomized, placebo-controlled IVF trial by Stern and colleagues found no benefit with aspirin and heparin for women with prior IVF implantation failure and antiphospholipid or antinuclear antibodies [15]. Another small matched study of women undergoing frozen embryo transfer found an 11% pregnancy rate with aspirin compared to 33% in controls, although the results were not statistically different [16]. Implantation rates were also lower with aspirin therapy, 2.9%, compared to 10.9% in untreated patients in this study. An uncontrolled study of IVF likewise found that outcomes were not improved with aspirin and heparin compared to conventional therapy [17]. Finally, a prospective, randomized, double-blind, placebo-controlled trial of poor responders by Lok et al found no benefit with daily aspirin 80 mg for cancellation rates, total dose of hMG used, number of mature follicles, or number of oocytes retrieved [18]. Furthermore, there was no difference in intraovarian or uterine artery pulsatility index with daily aspirin.
Randomized controlled trials have repeatedly shown that combined aspirin plus heparin improves pregnancy outcomes for women with recurrent pregnancy losses attributed to antiphospholipid antibodies [19,20]. This benefit is also shown in a prospective series [21]. Outcomes are better with aspirin plus heparin therapy than with aspirin alone in most [20,21], but not all studies [22,23]. Aspirin plus corticosteroid therapy, on the other hand, may be harmful. Combined low-dose aspirin plus prednisone increased the risk of preterm birth in two randomized controlled trials [24,25]. With a minimal benefit of aspirin alone for women with recurrent pregnancy losses and antiphospholipid antibodies, it is not surprising that we failed to find a beneficial effect of aspirin therapy in our general IVF population.
In our study, we did not test for uterine blood flow or routinely test for antiphospholipid antibodies. Therefore, we were not able to sub-divide the women in our study into groups that might be more responsive to aspirin. However, an ASRM Practice Committee Report in 1999 concluded that antiphospholipid antibodies do not affect IVF success, and therapy is not justified [26]. Furthermore, we believe that implantation rates, pregnancy rates, and live birth rates are more important indicators of IVF outcome compared to indirect measurements such as endometrial blood flow. In our study, pregnancy, implantation, and live birth rates were higher in the non-aspirin control group.
Another weakness in our study is the six-year period over which our IVF cycles were reviewed. It is possible that subtle differences could bias results in the aspirin and control groups in a retrospective analysis. Additionally, the small study population yields a limited statistical power to detect minor differences in pregnancy outcomes with aspirin. There are actual and sometimes large differences between the two groups of women, which could affect the outcomes. The differencesare not significant, but might be due to the small population studied. There certainly could be minor changes in treatment protocols over that span of time, but our age-related pregnancy and live birth rates remained stable during the years of this study.
Based on the results from our study and the prospective randomized trials by Urman and colleagues [4] and Stern et al [15], aspirin is not beneficial for a general IVF population. Since implantation, pregnancy, and delivery rates are higher for non-aspirin users, our study raises the possibility that aspirin may lower IVF success. A potential fertility reducing effect of aspirin is plausible, since prostaglandins affect ovulation, fertilization, and implantation [27]. Since aspirin inhibits prostaglandin synthesis, implantation could be compromised. Clearly, a larger, prospective randomized study with adequate power would be needed to determine if low-dose aspirin reduced IVF success.
There is some risk associated with aspirin therapy for infertility, although the extent of the risk for a healthy infertility population is unclear. One population based cohort study found that aspirin and nonsteroidal anti-inflammatory agents increased the risk of miscarriage, although a recent meta-analysis showed no increased risk of miscarriage with aspirin [28,29]. Although aspirin does not appear to alter the risk of congenital anomalies, first trimester aspirin consumption may increase the incidence of gastroschisis [30]. Acetylsalicylic acid may reach the uteroplacental circulation and exert antiplatelet effects in the fetus and newborn, although the incidence of neonatal bleeding does not appear to be increased with maternal aspirin [31,32]. However, maternal aspirin may raise the risk of placental abruption and antenatal, intrapartum, and postpartum hemorrhage [32,33]. Additionally, there is at least one reported maternal death due to complications of cerebral hemorrhage in a woman treated with aspirin and heparin after IVF [34]. Although these risks may be small, treatment with aspirin is not justified in the absence of a proven benefit.
Conclusion
Low-dose aspirin did not enhance endometrial thickness, augment the ovarian response, or improve pregnancy rates in our study. There is no apparent benefit in the routine use of aspirin during IVF cycles, and this practice should be abandoned.
Competing interests
The author(s) declare that they have no competing interests.
Authors' contributions
* BSH conceived of the study, participated in the analysis and interpretation of the data, and drafting and revising the manuscript. JTB made substantial contributions to the design and acquisition of data, and drafting the manuscript. PBM, MAP, TAL, and MLM made substantial contributions to the acquisition of data and revising the manuscript. All authors read and approved the final manuscript
Acknowledgements
We thank Howell Sasser, Ph.D., Director of Research Epidemiology, Dickson Institute at Carolinas Medical Center for his assistance with statistical evaluation.
==== Refs
Wada I Hsu CC Williams G Macnamee MC Brinsden PR The benefits of low-dose aspirin therapy in women with impaired uterine perfusion during assisted conception Hum Reprod 1994 9 1954 7 7844233
Rubinstein M Marazzi A Polak de Fried E Low-dose aspirin treatment improves ovarian responsiveness, uterine and ovarian blood flow velocity, implantation, and pregnancy rates in patients undergoing in vitro fertilization: a prospective, randomized, double-blind placebo-controlled assay Fertil Steril 1999 71 825 9 10231040 10.1016/S0015-0282(99)00088-6
Weckstein LN Jacobson A Galen D Hampton K Hammel J Low-dose aspirin for oocyte donation recipients with a thin endometrium: prospective, randomized study Fertil Steril 1997 68 927 30 9389827 10.1016/S0015-0282(97)00330-0
Urman B Mercan R Alatas C Balaban B Isiklar A Nuhoglu A Low-dose aspirin does not increase implantation rates in patients undergoing intracytoplasmic sperm injection: a prospective randomized study J Assisted Reprod Genet 2000 17 586 90 10.1023/A:1026491426423
Polak de Fried E Errata Fertil Steril 1999 72 755 10.1016/S0015-0282(99)00351-9
Hsieh YY Tsai HD Chang CC Lo HY Chen CL Low-dose aspirin for infertile women with thin endometrium receiving intrauterine insemination: a prospective, randomized study J Assist Reprod Genet 2000 17 174 7 10911579 10.1023/A:1009474307376
Waldenstrom U Hellberg D Nilsson S Low-dose aspirin in a short regimen as standard treatment in in vitro fertilization: a randomized, prospective study Fertil Steril 2004 81 1560 4 15193477 10.1016/j.fertnstert.2004.02.082
Sher G Zouves C Feinman M Maassarani G Matzner W Chong P Ching W A rational basis for the use of combined heparin/aspirin and IVIG immunotherapy in the treatment of recurrent IVF failure associated with antiphospholipid antibodies Am J Reprod Immunol 1998 39 391 4 9645271
Hasegawa I Hamanoto Y Suzuki M Murakawa H Kurabayshi T Takakuwa K Tanaka K Prednisolone plus low-dose aspirin improves the implantation rate in women with autoimmune conditions who are undergoing in vitro fertilization Fertil Steril 1998 70 1044 8 9848293 10.1016/S0015-0282(98)00343-4
Sher G Matzner W Feinman M Maassarani G Zouves C Chong P Ching W The selective use of heparin/aspirin therapy, alone or in combination with intravenous immunoglobulin G, in the management of antiphospholipid antibody-positive women undergoing in vitro fertilization Am J Reprod Immunol 1998 40 74 82 9764348
Sher G Maassarani G Zouves C Feinman M Sohn S Matzner W Chong P Ching W The use of combined heparin/aspirin and immunoglobulin G therapy in the treatment of in vitro fertilization patients with antithyroid antibodies Am J Reprod Immunol 1998 39 223 5 9553645
Sher G Feinman M Zouves C Kuttner G Maassarani G Salem R Matzner W Ching W Chong P High fecundity rates following in-vitro fertilization and embryo transfer in antiphospholipid antibody seropositive women treated with heparin and aspirin Hum Reprod 1994 9 2278 83 7714144
Geva E Amit A Lerner-Geva L Yaron Y Daniel Y Schwartz T Azem F Yovel I Lessing JB Prednisone and aspirin improve pregnancy rate in patients with reproductive failure and autoimmune antibodies: a prospective study Am J Reprod Immunol 2000 43 36 40 10698039 10.1111/j.8755-8920.2000.430107.x
Bose P Black S Kadyrov M Weissenborn U Neulen J Regan L Huppertz B Heparin and aspirin attenuate placental apoptosis in vitro: implications for early pregnancy Am J Obstet Gynecol 2005 192 23 30 15671997 10.1016/j.ajog.2004.09.029
Stern C Chamley L Norris H Hale L Baker HW A randomized, double-blind, placebo-controlled trial of heparin and aspirin for women with in vitro fertilization implantation failure and antiphospholipid or antinuclear antibodies Fertil Steril 2003 80 376 83 12909502 10.1016/S0015-0282(03)00610-1
Check JH Dietterich C Lurie D Nazari A Chuong J A matched study to determine whether low-dose aspirin without heparin improves pregnancy rates following frozen embryo transfer and/or affects endometrial sonographic parameters J Assist Reprod Genet 1998 15 579 82 9866064 10.1023/A:1020373009043
Kutteh WH Yetman DL Chantilis SJ Crain J Effect of antiphospholipid antibodies in women undergoing in-vitro fertilization: role of heparin and aspirin Hum Reprod 1997 12 1171 5 9221995 10.1093/humrep/12.6.1171
Lok IH Yip SK Cheung LP Yin Leung PH Haines CJ Adjuvant low-dose aspirin therapy in poor responders undergoing in vitro fertilization: a prospective, randomized, double-blind, placebo-controlled trial Fertil Steril 2004 81 556 61 15037402 10.1016/j.fertnstert.2003.07.033
Triolo G Ferrante A Ciccia F Accardo-Palumbo A Perino A Castelli A Giarratano A Licata G Randomized study of subcutaneous low molecular weight heparin plus aspirin versus intravenous immunoglobulin in the treatment of recurrent fetal loss associated with antiphospholipid antibodies Arthritis Rheum 2003 48 728 31 12632426 10.1002/art.10957
Rai R Cohen H Dave M Regan L Randomised controlled trial of aspirin and aspirin plus heparin in pregnant women with recurrent miscarriage associated with phospholipids antibodies (or antiphospholipid antibodies) BMJ 1997 314 253 7 9022487
Kutteh WH Antiphospholipid antibody-associated recurrent pregnancy loss: treatment with heparin and low-dose aspirin is superior to low-dose aspirin alone Am J Obstet Gynecol 1996 174 1584 9 9065133
Farquharson RG Quenby S Greaves M Antiphospholipid syndrome in pregnancy: a randomized, controlled trial of treatment Obstet Gynecol 2002 100 408 13 12220757 10.1016/S0029-7844(02)02165-8
Pattison NS Chamley LW Birdsall M Zanderigo AM Liddell HS McDougall J Does aspirin have a role in improving pregnancy outcome for women with the antiphospholipid syndrome? A randomized controlled trial Am J Obstet Gynecol 2000 183 1008 12 11035355 10.1067/mob.2000.106754
Laskin CA Bombardier C Hannah ME Mandel FP Ritchie JW Farewell V Farine D Spitzer K Fielding L Soloninka CA Yeung M. Prednisone and aspirin in women with autoantibodies and unexplained recurrent fetal loss NEJM 1997 337 148 53 9219700 10.1056/NEJM199707173370302
Silver RK MacGregor SN Sholl JS Hobart JM Neerhof MG Ragin A Comparative trial of prednisone plus aspirin versus aspirin alone in the treatment of anticardiolipin antibody-positive obstetric patients Am J Obstet Gynecol 1993 169 1411 7 8267038
Practice Committee Report Antiphospholipid antibodies do not affect IVF success American Society for Reproductive Medicine 1999
Rock JA Hurst BS Clinical significance of prostanoid concentration in women with endometriosis Prog Clin Biol Res 1990 23 61 80 2406757
Li DK Liu L Odouli R Exposure to non-steroidal anti-inflammatory drugs during pregnancy and risk of miscarriage: population based cohort study BMJ 2003 327 368 12919986 10.1136/bmj.327.7411.368
Kozer E Nikfar S Costei A Boskovic R Nulman I Koren G Aspirin consumption during the first trimester of pregnancy and congenital anomalies: a meta-analysis Am J Obstet Gynecol 2002 187 1623 30 12501074 10.1067/mob.2002.127376
Kozer E Costei AM Boskovic R Nulman I Nikfar S Koren G Effects of aspirin consumption during pregnancy on pregnancy outcomes: meta-analysis Devel Reprod Toxicology 2003 68 70 84 10.1002/bdrb.10002
Leonhardt A Bernert S Watzer B Schmitz-Zeigler G Seyberth HW Low-dose aspirin in pregnancy: maternal and neonatal aspirin concentrations and neonatal prostanoid formation Pediatrics 2003 111 e77 81 12509599 10.1542/peds.111.1.e77
Sibai BM Caritis SN Thom E Klebanoff M NcNellis D Rocco L Paul RH Romero R Witter F Rosen M Prevention of preeclampsia with low-dose aspirin in healthy, nulliparous pregnant women. The National Institute of Child Health and Human Development Network of Maternal-Fetal Medicine Units NEJM 1993 329 1213 8 8413387 10.1056/NEJM199310213291701
Golding J A randomized trial of low dose aspirin for primiparae in pregnancy. The Jamaica Low Dose Aspirin Study Group Br J Obstet Gynaecol 1998 105 293 9 9532989
Centers for Disease Control and Prevention Pregnancy-related death associated with heparin and aspirin treatment for infertility, 1996 JAMA 1998 279 1860 1 9634247 10.1001/jama.279.23.1860
| 15927049 | PMC1175854 | CC BY | 2021-01-04 16:40:17 | no | J Exp Clin Assist Reprod. 2005 May 31; 2:8 | utf-8 | J Exp Clin Assist Reprod | 2,005 | 10.1186/1743-1050-2-8 | oa_comm |
==== Front
J Neuroengineering RehabilJournal of NeuroEngineering and Rehabilitation1743-0003BioMed Central London 1743-0003-2-121592706410.1186/1743-0003-2-12ResearchMathematical models use varying parameter strategies to represent paralyzed muscle force properties: a sensitivity analysis Frey Law Laura A [email protected] Richard K [email protected] Graduate Program in Physical Therapy and Rehabilitation Science, 1-252 Medical Education Bldg., The University of Iowa, Iowa City, IA, USA2005 31 5 2005 2 12 12 22 12 2004 31 5 2005 Copyright © 2005 Law and Shields; licensee BioMed Central Ltd.2005Law and Shields; licensee BioMed Central Ltd.This is an Open Access article distributed under the 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
Mathematical muscle models may be useful for the determination of appropriate musculoskeletal stresses that will safely maintain the integrity of muscle and bone following spinal cord injury. Several models have been proposed to represent paralyzed muscle, but there have not been any systematic comparisons of modelling approaches to better understand the relationships between model parameters and muscle contractile properties. This sensitivity analysis of simulated muscle forces using three currently available mathematical models provides insight into the differences in modelling strategies as well as any direct parameter associations with simulated muscle force properties.
Methods
Three mathematical muscle models were compared: a traditional linear model with 3 parameters and two contemporary nonlinear models each with 6 parameters. Simulated muscle forces were calculated for two stimulation patterns (constant frequency and initial doublet trains) at three frequencies (5, 10, and 20 Hz). A sensitivity analysis of each model was performed by altering a single parameter through a range of 8 values, while the remaining parameters were kept at baseline values. Specific simulated force characteristics were determined for each stimulation pattern and each parameter increment. Significant parameter influences for each simulated force property were determined using ANOVA and Tukey's follow-up tests (α ≤ 0.05), and compared to previously reported parameter definitions.
Results
Each of the 3 linear model's parameters most clearly influence either simulated force magnitude or speed properties, consistent with previous parameter definitions. The nonlinear models' parameters displayed greater redundancy between force magnitude and speed properties. Further, previous parameter definitions for one of the nonlinear models were consistently supported, while the other was only partially supported by this analysis.
Conclusion
These three mathematical models use substantially different strategies to represent simulated muscle force. The two contemporary nonlinear models' parameters have the least distinct associations with simulated muscle force properties, and the greatest parameter role redundancy compared to the traditional linear model.
==== Body
Background
Chronic complete spinal cord injury (SCI) induces musculoskeletal deterioration that can be life threatening. Initially muscle atrophy occurs [1], followed by muscle fiber and motor unit transformation [2-5], and ultimately lower extremity osteoporosis develops [6-10]. Maintaining paralyzed muscle tissue may prove to be a valuable means for improving the general health and well-being of individuals with SCI. Neuromuscular electrical stimulation (NMES) can be used to restore function or to impart physiologic stresses to the skeletal system in an attempt to minimize muscle atrophy and ultimately osteoporosis [11-18]. However, well-defined NMES initiated muscle forces are needed as high forces can result in bone fracture [19].
Mathematical muscle models may be essential for the determination of the necessary musculoskeletal stresses that will safely maintain the integrity of muscle and bone following SCI. Further, a clear understanding of the relationships between model parameters and muscle contractile properties or their underlying physiological processes would benefit the practical use of models for therapeutic applications. Accordingly, several approaches have been used to mathematically model electrically induced muscle forces [20-24] in able-bodied human and animal muscle.
Although muscle force production is an inherently nonlinear response of the neuromuscular system, reasonable force approximations have been achieved using linear systems [25]. A nonlinear version of a traditional 2nd order system was developed by Bobet and Stein [20], and validated using cat soleus (slow) and plantaris (fast) muscle. A variation of the traditional Hill model, with additional Huxley-type modeling components (similar to the Distribution-Moment Model described by Zahalak and Ma,[26]), has evolved since its introduction [27], successfully representing submaximally activated, able-bodied, human quadriceps muscle [28-32]. While other models are available these three examples represent a diverse range of modeling approaches that allow a wide variety of discrete input patterns using constant parameter coefficients.
We are not aware of any previous comparisons of these types of models to elucidate their differences in modeling strategies. Although model parameter roles are often reported with physiologic interpretations, rarely has evidence been provided to support these physiologic (vs. mathematic) characterizations. The purpose of this study was to systematically compare one traditional linear model and two contemporary nonlinear models, using a sensitivity analysis to examine how each model's parameters influenced select simulated force properties.
The three models used different strategies to represent select force properties (peak force, force time integral, time to peak tension, half relaxation time, catch-like property, and force fusion). Further, previously reported definitions were not consistently supported by the sensitivity analyses for one of the nonlinear models. These results are important for the implementation and interpretation of future studies aimed at modeling chronically paralyzed muscle and are necessary precursors for the optimization of therapeutic stresses in attempts to maintain the integrity of paralyzed extremities and/or restore function after SCI.
Methods
This study consists of simulated sensitivity analyses of three mathematical muscle models currently available in the literature (see below). A common, but unique, feature of each of these models is that they can accommodate inputs consisting of any number of pulses at any combination of interpulse intervals (IPIs). This input flexibility allows each model to predict a wide-range of force responses, including the impulse-response, variable or constant frequency trains, doublets, and/or randomly spaced stimulation pulses that could be useful for electrical stimulation of paralyzed human muscle.
Linear Model
The simplest model in this study is a traditional 2nd order linear model consisting of one differential equation and three constant parameters. Second order linear systems are widely used to represent a variety of dynamic systems [33] and have been used in various formats to represent muscle [25,34,35]. Although a second order linear model can be mathematically represented in several ways, the traditional linear system theory configuration was used for this analysis (1).
The parameters for this modeling strategy have well-documented mathematical definitions. Parameter β is the system gain, ωn is the undamped natural frequency, and ζ is the damping ratio (a measure of output oscillation).
Investigating the sensitivity of this traditional modeling approach for predicting simulated muscle force properties provides a valuable basis for the interpretation and comparison of more complex muscle modeling approaches, where the parameters may not be clearly defined. In addition, this model may be easily modulated with more complex feedback control systems, making clear interpretations of the parameter roles in terms of muscle force properties desirable.
2nd Order Nonlinear Model
A nonlinear variation of a 2nd order linear model was introduced by Bobet and Stein [20]. In addition to two first order differential equations (2 and 4), it includes a saturation nonlinearity (3) which saturates force at higher levels as well as a variable time constant parameter (5), which generally decreases (becomes slower) with increasing force.
q(t) = ∫exp(-aT)u(t - T)dT (2)
x(t) = q(t)n /(q(t)n + kn) (3)
F(t) = Bb ∫exp(-bT)x(t - T)dT (4)
b = b0 (1 - b1F(t) / B)2 (5)
In Equation 2 the input, u(t), is a time series of the stimulation pulse train, with values of zero as the baseline and equal to 1/(delta t) at each pulse. The final output, F(t), is the modeled force over time (4), using (5) to define the variable parameter, b, as force varies over time. Parameter b varies with force based on constant parameters b0 and b1. This model has six constant parameters, B, a, b0, b1, n, and k, acting as the gain, two rate constants, and three "muscle specific constants" [20], respectively. See Table 1 for previously reported parameter definitions. Although in the original model, parameter b1 is constrained to values between o and 1, pilot studies using human paralyzed muscle observed better model fits when this constraint was relaxed to allow for negative values as well [36].
Table 1 Summary of reported parameter definitions for three mathematical muscle models.
Model Parameter Definition
2nd Order Linear β (Ns) output gain [25, 33, 35]
ωn (rad/s) natural undamped frequency [25, 33, 35]
ζ (-) damping coefficient [25, 33, 35]
2nd Order Nonlinear B (N) force gain, "maximum tetanic force" [20]
a (1/s) "muscle specific" rate constant [20]
b0(1/s) rate constant; maximum value of variable rate constant parameter, b, when b1 = zero. [20]
b1 (-) force feedback mechanism for variable rate constant, b; higher values = greater modulation of parameter b [20]
n (-) "muscle specific constant" used in static force saturation equation [20]
k (-) "muscle specific constant" used in static force saturation equation [20]
Hill-Huxley Nonlinear A (N/ms) Force scaling factor [21, 28, 29, 31, 32, 41, 42], and scaling factor for the muscle shortening velocity [29, 31, 41, 42]
τ1(ms) Force decay time constant when CN is absent, i.e. "in absence of strongly bound cross-bridges" [21, 28-32, 41, 42]
τ2(ms) Force decay time constant when CN is present; "extra friction due to bound cross-bridges" [21, 28-32, 41, 42]
τc(ms) Time constant controlling rise and decay of CN [21, 28-31, 41, 42] or the transient shape of CN [32] and time constant controlling the duration of force enhancement due to closely spaced pulses [30]
km(-) "Sensitivity of strongly bound cross-bridges to CN" [29, 31, 32, 41, 42]
R0(-) Magnitude of force enhancement due to closely-spaced pulses [28, 30] and/or from the following stimuli [29, 31, 41, 42]
Hill Huxley Nonlinear Model
The second nonlinear mathematical muscle model has been described by its authors as an extension of the Hill modeling approach [21,27]. However, one equation in the model represents calcium kinetics not typical of Hill-based modeling approaches, and contains model components that resemble the Distribution-Moment Model [26], an extension of the Huxley model. Thus, we will use the term Hill Huxley nonlinear model to represent this modeling approach.
The most current version of this model incorporates two nonlinear differential equations, (6) and (7) [27,29-31].
Equation 6 is reported to represent the calcium kinetics involved in muscle contraction (both the release/reuptake of Ca2+ as well as the binding to troponin, state variable = Cn), where variable parameter, Ri, is defined in (9). Ri decays as a function of each successive interpulse interval (ti-ti-1) rather than as a function of force as for the 2nd order nonlinear model [27,29-31]. Equation 7 predicts force (state variable, F), based on the state variable, Cn, but has no analytical solution, requiring numerical analysis techniques to solve for force. The Hill Huxley model incorporates a total of six constant parameters, A, τ1, τc, τ2, Ro, and km, as the gain, three time constants, a doublet parameter, and a "sensitivity" parameter [29], respectively. Please see Table 1 for previously reported parameter definitions.
Sensitivity Analysis
Simulated force trains were calculated for six different input patterns using Matlab 6.0 (Release 12, The Mathworks, Inc. USA): three constant frequency trains (CT) at 5, 10, and 20 Hz (using 8, 10, and 12 pulses, respectively), and three doublet frequency trains (DT) with base frequencies of 5, 10, and 20 Hz, but with an added pulse (doublet) 6 ms after the first pulse (using 9, 11, and 13 pulses, respectively). Please see figure 1 for a schematic representation of the input patterns.
Figure 1 Schematic representation of simulated force stimulation patterns. Simulated stimulation patterns at three frequencies, 5, 10, and 20 Hz, and two types of patterns, constant train (CT) with constant interpulse intervals, and doublet train (DT) with one additional doublet pulse occurring 6 ms after the first pulse.
These input patterns and frequencies were chosen to approximately correspond to a set of safe and most plausible stimulation patterns for a patient population. The risk of fracture with high frequency stimulation in individuals with SCI is considerable [19,37,38] and must be considered for the ultimate aim of validating this model for paralyzed muscle. Secondarily, to best consider parameter sensitivities at various points along the sigmoidal portion of the force frequency relationship in paralyzed muscle[39], frequencies ranging from 5 to 20 Hz were chosen in concert with 6 ms doublets (167 Hz).
The role of each parameter, in each mathematical muscle model, was determined by altering one parameter at a time, keeping all other parameters set at baseline values. The parameter increment, range, and baseline values were based on both previously reported values (Table 2) and extensive experimental pilot data (means ± 4 SD) from chronically paralyzed human soleus muscle with and without previous electrical stimulation training [36]. Previously reported parameter values varied by species [21,25,27,40] and varied through model evolutions [21,27,30,31]. Using parameter values based on pilot studies helps to provide a consistent basis necessary for between model comparisons. As no other reports of model applications in human SCI muscle were available, a wide range of values were incorporated in this study (~ +/- 4 SD of baseline) to maximize the potential for these results to be meaningful for various human paralyzed muscle applications.
Table 2 Parameter baselines, increments, and ranges used for the sensitivity analysis.
Model Parameter Range Baseline ± Increment Previously Reported Values
Human Animal
2nd Order Linear β (Ns) 15 – 60 30 ± 5 0.05 – 0.5A 0.10 – 0.62B
ωn(rad/s) 7 – 25 13 ± 2 12.6 – 18.8A 12.6 – 50.3B
ζ (-) 0.4 – 1.3 0.7 ± 0.1 0.6 – 1.0A 1.0 – 2.0B
2nd Order Nonlinear B (N) 375 – 1050 600 ± 75 - 9.0 – 46C
a (s-1) 10 – 28 16 ± 2 - 9.4 – 40C
b0 (s-1) 6 – 24 12 ± 2 - 11 – 40C
b1 (-) -0.8 – 0.8 -0.2 ± 0.2 - 0.4 – 0.95C
n (-) 1 – 10 4 ± 1 - 3.2 – 4.0C
k (-) 0.1 – 1.0 0.4 ± 0.1 - 0.78 – 1.0C
Hill-Huxley Nonlinear A (N/ms) 5 – 14 8 ± 1 3 – 5 D - †
τ1(ms) 5 – 95 35 ± 10 42 – 51 D -
τ2(ms) 30 – 165 75 ± 15 NA – 124‡ D -
τc (ms) 5 – 50 20 ± 5 20* D -
km (-) 0.025 – 0.25 0.1 ± 0.025 0.1 – 0.3‡ D -
R0 (-) 1 –10 4 ± 1 1.14* – 2* D -
A Approximate values of submaximally-activated human soleus muscle when positioned ~ neutral ankle dorsiflexion [25].
B Approximate values of maximally activated cat soleus muscle [40].
C Range of reported values for maximally activated cat soleus and plantaris muscle[20].
D Values for submaximally-activated human quadriceps muscle in the non-fatigued state. [29, 31, 42]
† The original Hill Huxley model parameters are too different for direct comparisons [27]
* Parameter values preset at constant values.
‡ Only one representative single subject value available.
NA No reported values available in 2 of the 3 studies.
Simulated force trains were calculated for eight values of each parameter for each of the six input patterns, as well as a single twitch (for doublet analyses, see below), creating a total of 56 force profiles per model parameter. Force was simulated at 1000 Hz.
Simulated Force Properties
For each of the CT force profiles, five specific force characteristics were determined using Matlab (Mathworks, USA): peak force (PF), defined as the maximum force at any time in the force profile; force-time integral (FTI), defined as the area under the force profile; half-relaxation time (1/2 RT), defined as the time required for force to decay from 90% to 50% of the final peak value; late relaxation time (LRT), defined as the time required for force to decay from 40% to 10% of the final peak value; and relative fusion index (RFI), defined as the mean of the last four pulses' minima divided by their succeeding four peaks (a RFI value of 1.0 indicates full fusion with no drop in force between pulses, whereas a value of 0.0 indicates no summation at all – a series of twitches reaching baseline between pulses). The time to peak tension (TPT) property, defined as the time (ms) required to reach 90% of the first peak force from time zero was determined using the 5 Hz CT pattern only. Using the DT and CT patterns at each frequency, the relative doublet PF (DPF) and doublet FTI (DFTI) were calculated. The DPF (and DFTI) were defined as the PF (FTI) of the DT and CT force differential (DT-CT) at each frequency normalized by the PF (FTI) of a single twitch. Values greater than (less than) 1.0 for either doublet property indicate more (less) force output than would be expected from a single twitch.
Statistical Analysis
The change in each of these force characteristics with each parameter increment was calculated (7 increments for 8 parameter values) using Matlab and Excel (Microsoft Office, USA). Analysis of Variance (ANOVA) was used to determine if any parameter had a significant influence on each force property, using α ≤ 0.05. Tukey's follow-up tests were used to determine which parameters had significant influences on each force property and relative to one another, to maintain the family wise error of 0.05 for each model.
Results
Examples of individual parameter increments on two of the six simulated force trains (5 Hz doublet train, DT, and 20 Hz constant train, CT) for the linear model, the 2nd order nonlinear model, and the Hill Huxley nonlinear model are shown in figures 2, 3, and 4, respectively. The results for specific force properties are presented by model as follows.
Figure 2 Linear model simulated force examples. Two simulated force trains are shown: 5 Hz doublet train, DT (left column), and 20 Hz constant train, CT (right column), with variations in each of the three individual parameter, β, ωn and ζ. Only odd numbered parameter increments are included (· -· -1st, - - 3rd, 5th and – 7th) for clarity.
Figure 3 2nd order nonlinear model simulated force examples. Two simulated force trains are shown: 5 Hz doublet train, DT (left column), and 20 Hz constant train, CT (right column), with variations in each of the six individual parameter, B, a, bo, b1, n, and k. Only odd numbered parameter increments are included (· -· -1st, - - 3rd, 5th and – 7th) for clarity.
Figure 4 Hill Huxley nonlinear model simulated force examples. Two simulated force trains are shown: 5 Hz doublet train, DT (left column), and 20 Hz constant train, CT (right column), with variations in each of the six individual parameter, A, τ1, τ2, τc, km, and Ro. Only odd numbered parameter increments are included (· -· -1st, - - 3rd, 5th and – 7th) for clarity.
Linear Model
The select simulated force characteristics for the three linear model parameters are shown in figure 5 using 10 Hz, consistent with the results at 5 and 20 Hz. Peak force (PF) and force time integral (FTI) were most strongly influenced at all three constant frequency trains (CT) (5, 10, and 20 Hz) by the gain parameter, β, with overall mean increases of 65.3 N and 50.0 Ns per 5 Ns increase in β, respectively (p < 0.05, figures 5 and 8), as would be expected based on previous definitions [33]. Changes in the natural frequency and the damping ratio, ωn and ζ respectively, produced relatively small, but significant (p < 0.05) effects on PF, but had no significant effect on FTI. No linear model parameter had any (nonlinear) effect on the doublet response relative to the twitch at any frequency (figures 5 and 8); i.e. additional pulses produced exactly the same amount of additional force a single pulse would produce in isolation, consistent with the definition of a linear system.
Figure 5 Representation of the parameter effects on simulated force characteristics for the linear model. Linear Model parameter effects on select force characteristics for the 10 Hz constant frequency pattern. Panel A: peak force (PF); B: force time integral (FTI); C: relative doublet PF; D: relative doublet FTI; E: time to peak tension (TPT); F: 1/2 relaxation time (HRT); G: late relaxation time (LRT); and H: relative fusion index (RFI, see text for operational definitions). Please see Table 2 for parameter baseline and increment values.
Figure 8 Mean (SD) change in force magnitude characteristics per parameter increment for three muscle models. The linear model (left column), the 2nd Order Nonlinear model (middle column), and the Hill Huxley nonlinear model (right column) are shown at 5, 10, and 20 Hz. Peak force (PF) and force time integral (FTI) for the constant frequency trains (CT) are shown in rows 1 and 2, respectively. Relative doublet (Dblt) to twitch (Tw) PF and FTI, (DT-CT)/Tw, are shown in rows 3 and 4. Significant (p < 0.05) parameter influences (for 5, 10, and 20 Hz inclusive) are indicated by an asterisk (*).
The natural frequency, ωn, was the most influential parameter for three of the four speed properties examined as expected based on its parameter definition (Table 1): time to peak tension (TPT), half relaxation time (1/2 RT), and relative fusion index (RFI), and was a secondary influence on the late relaxation time (LRT); see figures 5 and 9. Two rad/s increments in ωn resulted in overall mean decreases of 9.6 ms, 12.5 ms, 13.1 ms, and 6.0 % for TPT, 1/2 RT, LRT, and RFI, respectively. The damping coefficient, ζ, also had significant (p < 0.05) influences on each force time property, but was a primary influence only for LRT, due to its strong influence on the final decay and oscillation of the system [33]. The gain parameter, β, had no significant effects on any of the force time characteristics, as would be expected. The simulated baseline force fusion (RFI) levels were 39.1, 80.8, and 95.3 % fused at 5, 10, and 20 Hz, respectively, indicating the simulated force baselines roughly represented a range of the force-frequency curve.
Figure 9 Mean (SD) change in select force time characteristics per parameter increment for three muscle models. The linear model (left column), the 2nd Order Nonlinear model (middle column), and the Hill Huxley nonlinear model (right column) are shown. Row 1 shows the time to peak tension (TPT) for the 5 Hz constant train (CT). Rows 2, 3, and 4 show the 1/2 relaxation time (HRT), late relaxation time (LRT), and the relative fusion index (RFI), respectively, for 5, 10, and 20 Hz CTs. See text for operational definitions. Significant (p < 0.05) parameter influences (for 5, 10, and 20 Hz inclusive) are indicated by an asterisk (*).
In summary, the force magnitude and force time properties were clearly divided between parameters in the linear model. Parameter β, the gain parameter, was the primary influence on the PF and FTI, whereas ωn and ζ, the natural frequency and damping ratio, were the primary and secondary influences on the four force speed properties.
2nd Order Nonlinear Model
Figure 6 displays the effects of incremental changes in each of the six 2nd order nonlinear model parameters on eight force characteristics using 10 Hz force trains. Similar results were found for 5 and 20 Hz. Peak force was significantly (p < 0.05) influenced by parameters B, k, and a, previously defined as the gain, a force saturation parameter and a rate constant [20]. The gain produced the greatest mean change in peak force (56.4 N per 75 N change in B, p < 0.05) followed by the saturation and rate constants, -43.2 and -25.2 N per parameter increment of 0.1 (unitless, k), and 2 s-1 (a); see figures 6 and 8. Similar results were observed for the FTI, however the magnitudes of the mean FTI change per parameter increment were not different between k and B or between B and a.
Figure 6 Representation of the parameter effects on simulated force characteristics for the 2nd order nonlinear model. 2nd order nonlinear model parameter effects on select force characteristics for the 10 Hz constant frequency pattern. Panel A: peak force (PF); B: force time integral (FTI); C: relative doublet PF; D: relative doublet FTI; E: time to peak tension (TPT); F: 1/2 relaxation time (HRT); G: late relaxation time (LRT); and H: relative fusion index (RFI, see text for operational definitions). Please see Table 2 for parameter baseline and increment values.
The relative doublet PF and FTI (doublet trains minus constant trains, normalized by the twitch) were only significantly (p < 0.05) affected by one parameter, one of the force saturation parameters, k (figures 6 and 8, with mean changes of 23.6 % and 30.5 % for the relative DPF and DFTI per 0.1 increment in k, respectively). Thus, as k increased, the added force due to a doublet increased. However, the simulated doublet at baseline parameter values consistently produced less force than a single isolated twitch, and decreased with frequency, with added peak force values of 79.7, 76.9, and 17.9% of the twitch at 5, 10, and 20 Hz, respectively (see figure 6 for 10 Hz representation only).
There were not any direct relationships between specific parameters and force time properties for the 2nd order nonlinear model. Different combinations of parameters influenced each of the force time characteristics (TPT, 1/2 RT, LRT, and RFI) (see figures 6 and 9). Consistent with Bobet's parameter definitions (Table 1), parameter b0, a rate constant parameter [20], most consistently influenced speed related properties overall (1/2 RT, LRT, and RFI), whereas B, the model gain, had no effect on any force time properties. The remaining four parameters, a, b1, n, and k, each produced significant (p < 0.05) mean changes in one or more of the force time properties evaluated (figure 9), supporting their somewhat vague previous definitions (Table 1). The force fusion (RFI) was equally influenced by parameters a, b0, b1, and k, with mean increases or decreases in force fusion ranging from 2.5 – 3.9 % per parameter increment (significant at p < 0.05). The simulated baseline force fusion levels (RFIs) encompassed a slightly wider range of the force frequency curve than observed with the linear model: 21.8, 75.3 and 99.5 % at 5, 10, and 20 Hz, respectively.
In summary, while most parameters were clearly differentiated as affecting solely force magnitude (B) or force time properties (b0, b1, and n) for the 2nd order nonlinear model, this was not universally observed. Parameters k and a, a force saturation parameter and a rate constant [20], had strong influences on both force time and force magnitude characteristics, with parameter k having more primary influences (p < 0.05 per Tukey's follow-up test groupings) than any other parameter in this model. Further, more specific parameter definitions than previously provided (see Table 1) do not appear to be warranted based on this sensitivity analysis.
Hill Huxley Nonlinear Model
Figure 7 shows the effects of incremental changes in each of the six Hill Huxley nonlinear model parameters on eight force characteristics using 10 Hz force trains. Similar results were observed at 5 and 20 Hz. Peak force and FTI for the constant trains were significantly affected by five of the six parameters, but the primary influence(s) based on Tukey's groupings were a time constant parameter and the gain [29], parameters τc and A (PF: 71.7 and 55.6 %, respectively) and τc (FTI: 75.8 Ns). Secondary influences on PF, based on Tukey's follow-up tests, included two additional time constants and a "sensitivity" parameter [29], parameters τ1, τ2, and km, respectively (mean PF change 44.6 – 45.6 N). Secondary influences on the FTI included the gain as well: parameters A, τ1, τ2, and km (mean FTI change 30.5 – 42.1 Ns). Ro, the parameter intended to control force enhancement due to doublets [29], had no significant effect on either PF or FTI for the constant stimulation (see figure 8). The strong influences of the three time constants and the "sensitivity" parameter [29] in addition to the primary gain factor on force magnitude properties were not expected based on previous published definitions (Table 1), and suggests that one or more of these time constant parameters may play a larger role in this model than previously described.
Figure 7 Representation of the parameter effects on simulated force characteristics for the Hill Huxley nonlinear model. Hill Huxley nonlinear model parameter effects on select force characteristics for the 10 Hz constant frequency pattern. Panel A: peak force (PF); B: force time integral (FTI); C: relative doublet PF; D: relative doublet FTI; E: time to peak tension (TPT); F: 1/2 relaxation time (HRT); G: late relaxation time (LRT); and H: relative fusion index (RFI, see text for operational definitions). Please see Table 2 for parameter baseline and increment values.
Both the relative doublet PF and FTI (representing aspects of the "catch-like" property of muscle) were equally affected by the "sensitivity" and doublet parameters, km and Ro, (figure 8) although only the Ro parameter was specifically added to the Hill Huxley model to better represent closely spaced pulses [30]. Increments in km and Ro resulted in equivalent mean increases of 8.2 and 6.2 % for the DPF and 11.0 and 7.9 % for the DFTI, respectively (figure 8). The time constant, τc, played a secondary role in relative doublet force with mean decreases of 4.4 and 5.1 % for doublet PF and doublet FTI, respectively (figure 8). Time constant, τ1, had an equal effect on DPF as τc, but had no significant effect on DFTI. As with the 2nd order nonlinear model, the additional peak force resulting from the simulated doublet relative to the twitch at baseline parameter values was less than 100%, and decreased with increasing frequency: 81.7, 69.4, and 26.3 % at 5, 10, and 20 Hz, respectively.
The four speed property measures were significantly influenced (p < 0.05) by three parameters in the Hill Huxley nonlinear model: time constants τc and τ1 and "sensitivity" parameter km (figure 9), however τ2 had no significant effect on any simulated force speed property despite its previous definition (Table 1). Time constant, τc, was consistently the primary influence (based on Tukey's follow-up tests) with mean increases of 9.7 ms, 19.2 ms, 24.8 ms, and 8.8 %, TPT, 1/2 RT, LRT, and RFI, respectively, per 5 ms increment in τc. Secondary influences on the relaxation times (1/2 RT and LRT) included time constant, τ1, and "sensitivity" parameter, km, with mean changes of 6.7 and -5.1 ms (magnitudes not significantly different) for the 1/2 RT and 19.8 and -5.1 ms for the LRT, respectively. This finding was surprising given that in most previous publications, τc has been kept at a constant value of 20 ms [29,30,32], and τ1 has been based on experimental late decay rates [21,29,30,32], which is not well supported by these results. Further, the only significant influence on fusion (RFI) was parameter τc. The simulated baseline fusion levels were 10.3, 78.1, and 98.5 % at 5, 10, and 20 Hz, respectively, providing a similar range of the simulated force frequency curve as the 2nd order nonlinear model.
In summary, five of the six parameters (gain, time constants, and "sensitivity") had nearly equal influences on the force magnitude properties, whereas only parameters τc, τ1, and km (two time constant and the "sensitivity" parameter) had significant influences on force time properties, only partially supporting previously published parameter definitions. Further, parameter τc was a primary influence for all but the doublet force characteristics.
Discussion
A common finding between models in this sensitivity analysis was that the "gain" factors (β, B, and A for the linear, 2nd order nonlinear, and Hill Huxley nonlinear models, respectively) each significantly altered only force magnitude characteristics, but were not the sole influence (or even the primary influence for the Hill Huxley model) on peak force. While the mathematical gain may relate to physiologic measures such as maximal tetanic force [20] or physiologic cross-sectional area, ultimately muscle force production is a result of several factors including muscle speed properties. Further, the two 2nd order system models had the most clearly discernible gain parameters (β and B), whereas the Hill Huxley nonlinear model had equivalent gain effects from A, τ1, τ2, and km – all less than parameter τc. Indeed, the definition of one parameter may be valid (e.g. a force gain parameter) but it is noteworthy that the parameter definition does not necessarily indicate the extent to which other parameters may also alter the physical property most commonly associated with that definition (e.g. peak force versus force gain).
Definitive physiologic force property associations were not always apparent for each model's parameters; however, parameter classifications as primarily force magnitude or force time modulators may be more appropriate. This was most clearly observed in the simple linear model, where β affected only force gain properties and the natural frequency, ωn, and damping ratio, ζ, influenced primarily the force time properties, consistent with traditional linear systems theory definitions for these parameters which have little overlap [33].
In the 2nd order nonlinear model, parameters b0, b1, and a behaved primarily as rate constants and B was a pure gain factor, consistent with previous definitions (Table 1). The rate constants were not clearly differentiated by the specific force time properties commonly considered in the muscle literature (e.g. TPT and 1/2 RT), but each had varying degrees of influence on the specific speed properties. Parameters n and k from the force saturation equation in the 2nd order nonlinear model played minimal and maximal roles in the model, respectively, when considering the eight force properties included in this study. Again, neither of these parameters can be easily defined physiologically, but k in particular provides a valuable contribution to the model, both due to its numerous primary influences (TPT, RFI, FTI, DPF, and DFTI) as well as its sole significant influence on the relative doublet force output.
The Hill Huxley model displayed the most parameter role redundancy with the least clearly defined individual parameter roles of the three modeling approaches. This redundancy may be beneficial for representing actual muscle forces, but it complicates the physiologic parameter interpretations often attributed to Hill-based models. Consistent with previous definitions for the Hill Huxley model parameters (Table 1), parameter A displayed purely gain characteristics, τ1 and τc proved to be important time constants, and Ro did influence the magnitude of additional doublet force. However, τ1 was not the primary nor the sole influence on the late decay time as its definition would suggest; doublet PF and FTI were equally influenced by km and Ro, despite the definition of Ro; and τ2 had no significant effects on any force time properties contrary to expectations for a time constant. Further investigation of parameter τ2, approaching previously reported values (Table 2) in non paralyzed human muscle, further diminished the overall influence of this parameter on the force magnitude and force time properties, suggesting that the discrepancies between these results and previous definitions are not due to differences in the range investigated.
To use any of these models for experimental muscle conditions, mathematical optimization would be used to solve the underdetermined series of equations. Due to the overlapping roles of the nonlinear model parameters (figures 8 and 9), it is possible that mathematical optimization of any one parameter (and more so with multiple parameters) may alter its "physiological" meaning, as changes in one parameter can often be offset by concomitant changes in others. Although Hill- and Huxley-type models are often credited as providing physiologically meaningful parameter values [21], with the intent of using parameter values for insight into the underlying muscle contractile and fatigue mechanisms [21], this sensitivity analysis would suggest parameter values should be interpreted with caution. However, this conclusion may not extend to all nonlinear or Hill-based models, but could be the result of the many parameter substitutions and equation evolutions of this particular Hill-based model and its inclusion of Huxley-type components.
The discrepancies between the simulated parameter roles and previous definitions for the Hill Huxley nonlinear model might suggest that some previously reported parameterization techniques and assumptions may be less than ideal. Parameter τc has been kept constant at 20 ms [21,30,31], potentially neglecting the numerous influences this parameter has on muscle force properties. The experimentally derived late decay rate has been used to estimate values for τ1 as described by Ding et al [21,30]. While this study does not directly assess the validity of this approach, it should be noted that τ1 was not the strongest influence on the late decay time. Most recently, Ro has been defined as having a linear, constant relationship with km: Ro = 1.04 + km [31]. This linear relationship was not apparent in this sensitivity study. Parameters km and Ro had similar effects on the doublet "catch-like" property of muscle, however they displayed disparate changes with increasing frequency (figure 8). Indeed, this simple linear relationship may hold true for isolated muscle conditions, such as the submaximally activated, able-bodied quadriceps muscle tested by Ding and colleagues, but possibly may not hold for human paralyzed muscle. The use of mathematical optimization techniques to determine all model parameters may circumvent the dependence of the model on potentially erroneous or incomplete parameter definitions. This optimization approach has been used for both 2nd order models [20,25,40].
The linear, individual investigation of parameter sensitivities is a potential limitation of this study. Particularly for the nonlinear models, interactions between parameters are likely to exist, which may not be fully exhibited within these results. However, altering multiple parameters at a time, while in theory useful, could produce highly complex results, making study assessments practically infeasible. This systematic sensitivity analysis approach provides valuable information regarding the different parameters' influences on force characteristics and illuminates each model's approach to mathematically representing physiologic phenomena that has not been previously investigated. Clinical scientists in rehabilitation must continue to understand the meaning of various muscle models in an effort to develop effective therapeutic interventions. This sensitivity analysis provides a framework for investigators to compare and choose a model that is most appropriate for the clinical application.
Conclusion
The key findings of this study were 1) the linear model parameters were clearly separated between simulated muscle force gain and speed properties, whereas this delineation was blurred for the two nonlinear models; 2) simulated force magnitude (PF) was generally influenced by multiple parameters for the nonlinear models, not solely by the defined force gain factors; 3) the reported physiologic parameter definitions were not consistently supported by the results for the Hill Huxley nonlinear model; and 4) these three mathematical models utilize substantially different approaches for representing muscle force, as indicated by the differences in parameter roles observed for each model.
This sensitivity analysis provides a strong framework to better understand the roles and sensitivities of each parameter for three mathematical muscle models as well as a means to compare their different modeling strategies. The results of this study will help researchers better understand the similarities and differences among three possible modeling approaches, assist in the interpretation of parameter values with varying muscle conditions (e.g. fatigue or contractile protein adaptations), and may provide valuable information necessary for choosing the most appropriate modeling approach for a particular application. The three models evaluated each use constant parameters to modulate their force outputs; given the same inputs these results conclude that they employ notably different strategies using constant parameters that do not consistently match previously reported definitions (Hill Huxley nonlinear model in particular). Further experimental studies will be needed to assess which model is best suited for use with human paralyzed muscle applications.
Abbreviation List
CT constant frequency trains
DT doublet frequency trains (single doublet at the start of a CT)
DPF doublet peak force normalized by twitch peak force
DFTI doublet force time integral normalized by twitch force time integral
FTI force time integral
1/2 RT half relaxation time
Hz Hertz
LRT late relaxation time
N Newtons
PF peak force
RFI relative fusion index
s seconds
TPT time to peak tension
Competing interests
The author(s) declare that they have no competing interests.
Authors' contributions
LAFL carried out all force simulations and calculations, performed statistical analysis and drafted the manuscript. RKS participated in the design and coordination of the study and critical revisions of the manuscript. Both authors read and approved of the final manuscript.
Acknowledgements
The authors would like to acknowledge funding from NIH RO1 HD39445 (RKS) and the Foundation for Physical Therapy (LAFL).
==== Refs
Castro MJ Apple DFJ Hillegass EA Dudley GA Influence of complete spinal cord injury on skeletal muscle cross-sectional area within the first 6 months of injury European Journal of Applied Physiology & Occupational Physiology 1999 80 373 378 10483809 10.1007/s004210050606
Shields RK Fatigability, relaxation properties, and electromyographic responses of the human paralyzed soleus muscle Journal of Neurophysiology 1995 73 2195 2206 7666132
Talmadge RJ Castro MJ Apple DFJ Dudley GA Phenotypic adaptations in human muscle fibers 6 and 24 wk after spinal cord injury Journal of Applied Physiology 2002 92 147 154 11744654
Scelsi R Marchetti C Poggi P Lotta S Lommi G Muscle fiber type morphology and distribution in paraplegic patients with traumatic cord lesion. Histochemical and ultrastructural aspects of rectus femoris muscle Acta Neuropathologica 1982 57 243 248 7136501 10.1007/BF00692178
Round JM Barr FM Moffat B Jones DA Fibre areas and histochemical fibre types in the quadriceps muscle of paraplegic subjects Journal of the Neurological Sciences 1993 116 207 211 8336167 10.1016/0022-510X(93)90327-U
Demirel G Yilmaz H Paker N Onel S Osteoporosis after spinal cord injury Spinal Cord 1998 36 822 825 9881730 10.1038/sj.sc.3100704
Lee TQ Shapiro TA Bell DM Biomechanical properties of human tibias in long-term spinal cord injury Journal of Rehabilitation Research & Development 1997 34 295 302 9239622
Szollar SM Martin EM Sartoris DJ Parthemore JG Deftos LJ Bone mineral density and indexes of bone metabolism in spinal cord injury American Journal of Physical Medicine & Rehabilitation 1998 77 28 35 9482376 10.1097/00002060-199801000-00005
Biering-Sorensen F Bohr HH Schaadt OP Longitudinal study of bone mineral content in the lumbar spine, the forearm and the lower extremities after spinal cord injury Eur J Clin Invest 1990 20 330 335 2114994
Shields RK Muscular, skeletal, and neural adaptations following spinal cord injury Journal of Orthopaedic & Sports Physical Therapy 2002 32 65 74 11838582
Shields RK Dudley-Javoroski S Deshpande P Long term electrical stimulation training prevents soleus muscle adaptation after spinal cord injury: ; New Orleans, LA. 2003
Frey Law LA Shields RK Femoral loads during passive, active, and active-resistive stance after spinal cord injury: a mathematical model Clinical Biomechanics 2004 19 313 321 15003348 10.1016/j.clinbiomech.2003.12.005
Rochester L Barron MJ Chandler CS Sutton RA Miller S Johnson MA Influence of electrical stimulation of the tibialis anterior muscle in paraplegic subjects. 2. Morphological and histochemical properties Paraplegia 1995 33 514 522 8524604
Mohr T Andersen JL Biering-Sorensen F Galbo H Bangsbo J Wagner A Kjaer M Long-term adaptation to electrically induced cycle training in severe spinal cord injured individuals.[erratum appears in Spinal Cord 1997 Apr;35(4):262] Spinal Cord 1997 35 1 16 9025213 10.1038/sj.sc.3100343
Gerrits HL Hopman MT Sargeant AJ Jones DA De Haan A Effects of training on contractile properties of paralyzed quadriceps muscle Muscle & Nerve 2002 25 559 567 11932974 10.1002/mus.10071
Chilibeck PD Bell G Jeon J Weiss CB Murdoch G MacLean I Ryan E Burnham R Functional electrical stimulation exercise increases GLUT-1 and GLUT-4 in paralyzed skeletal muscle Metabolism: Clinical & Experimental 1999 48 1409 1413 10582549
Shields RK Dudley-Javoroski S Musculoskeletal adaptations after spinal cord injury are prevented with a minimal dose of daily electrical stimulation exercise 2004 abstract
Hartkopp A Murphy RJ Mohr T Kjaer M Biering-Sorensen F Bone fracture during electrical stimulation of the quadriceps in a spinal cord injured subject Arch Phys Med Rehabil 1998 79 1133 1136 9749697 10.1016/S0003-9993(98)90184-8
Bobet J Stein RB A simple model of force generation by skeletal muscle during dynamic isometric contractions IEEE Transactions on Biomedical Engineering 1998 45 1010 1016 9691575 10.1109/10.704869
Ding J Binder-Macleod SA Wexler AS Two-step, predictive, isometric force model tested on data from human and rat muscles J Appl Physiol 1998 85 2176 2189 9843541
Dorgan SJ O'Malley MJ A nonlinear mathematical model of electrically stimulated skeletal muscle IEEE Transactions on Rehabilitation Engineering 1997 5 179 194 9184904 10.1109/86.593289
Durfee WK Palmer KI Estimation of force-activation, force-length, and force-velocity properties in isolated, electrically stimulated muscle IEEE Transactions on Biomedical Engineering 1994 41 205 216 8045573 10.1109/10.284939
Gollee H Murray-Smith DJ Jarvis JC A nonlinear approach to modeling of electrically stimulated skeletal muscle IEEE Transactions on Biomedical Engineering 2001 48 406 415 11322528 10.1109/10.915705
Bawa P Stein RB Frequency response of human soleus muscle Journal of Neurophysiology 1976 39 788 793 966038
Zahalak GI Ma SP Muscle activation and contraction: constitutive relations based directly on cross-bridge kinetics Journal of Biomechanical Engineering 1990 112 52 62 2308304
Wexler AS Ding J Binder-Macleod SA A mathematical model that predicts skeletal muscle force IEEE Transactions on Biomedical Engineering 1997 44 337 348 9125818 10.1109/10.568909
Ding J Wexler AS Binder-Macleod SA A predictive model of fatigue in human skeletal muscles Journal of Applied Physiology 2000 89 1322 1332 11007565
Ding J Wexler AS Binder-Macleod SA A mathematical model that predicts the force-frequency relationship of human skeletal muscle Muscle & Nerve 2002 26 477 485 12362412 10.1002/mus.10198
Ding J Wexler AS Binder-Macleod SA Development of a mathematical model that predicts optimal muscle activation patterns by using brief trains J Appl Physiol 2000 88 917 925 10710386
Ding J Wexler AS Binder-Macleod SA Mathematical models for fatigue minimization during functional electrical stimulation Journal of Electromyography & Kinesiology 2003 13 575 588 14573372 10.1016/S1050-6411(03)00102-0
Perumal R Wexler AS Ding J Binder-Macleod SA Modeling the length dependence of isometric force in human quadriceps muscles Journal of Biomechanics 2002 35 919 930 12052394 10.1016/S0021-9290(02)00049-0
Close CM Frederick DK Modeling and Analysis of Dynamic Systems 1995 2nd Edition New York, John Wiley & Sons 681
Baratta RV Zhou BH Solomonow M Frequency response model of skeletal muscle: effect of perturbation level, and control strategy Medical & Biological Engineering & Computing 1989 27 337 345 2601461
Bobet J Stein RB Oguztoreli MN A linear time-varying model of force generation in skeletal muscle IEEE Transactions on Biomedical Engineering 1993 40 1000 1006 8294124 10.1109/10.247798
Frey Law LA Predicting human paralyzed muscle force properties: an assessment of three mathematical muscle models Physical Rehabilitation Science 2004 Iowa City, The University of Iowa 138
Lazo MG Shirazi P Sam M Giobbie-Hurder A Blacconiere MJ Muppidi M Osteoporosis and risk of fracture in men with spinal cord injury Spinal Cord 2001 39 208 214 11420736 10.1038/sj.sc.3101139
Vestergaard P Krogh K Rejnmark L Mosekilde L Fracture rates and risk factors for fractures in patients with spinal cord injury Spinal Cord 1998 36 790 796 9848488 10.1038/sj.sc.3100648
Shields RK Law LF Reiling B Sass K Wilwert J Effects of electrically induced fatigue on the twitch and tetanus of paralyzed soleus muscle in humans Journal of Applied Physiology 1997 82 1499 1507 9134899
Mannard A Stein RB Determination of the frequency response of isometric soleus muscle in the cat using random nerve stimulation Journal of Physiology 1973 229 275 296 4353409
Ding J Wexler AS Binder-Macleod SA A predictive fatigue model--II: Predicting the effect of resting times on fatigue IEEE Transactions on Neural Systems & Rehabilitation Engineering 2002 10 59 67 12173740 10.1109/TNSRE.2002.1021587
Ding J Wexler AS Binder-Macleod SA A predictive fatigue model--I: Predicting the effect of stimulation frequency and pattern on fatigue.[erratum appears in IEEE Trans Neural Syst Rehabil Eng. 2003 Mar;11(1):86] IEEE Transactions on Neural Systems & Rehabilitation Engineering 2002 10 48 58 12173739 10.1109/TNSRE.2002.1021586
| 15927064 | PMC1175855 | CC BY | 2021-01-04 16:37:41 | no | J Neuroengineering Rehabil. 2005 May 31; 2:12 | utf-8 | J Neuroeng Rehabil | 2,005 | 10.1186/1743-0003-2-12 | oa_comm |
==== Front
Mol PainMolecular Pain1744-8069BioMed Central London 1744-8069-1-191593265210.1186/1744-8069-1-19ResearchPre-injury administration of morphine prevents development of neuropathic hyperalgesia through activation of descending monoaminergic mechanisms in the spinal cord in mice Rashid Md Harunor [email protected] Hiroshi [email protected] Division of Molecular Pharmacology and Neuroscience, Nagasaki University Graduate School of Biomedical Sciences, Nagasaki 852-8521, Japan2 Dept of Integrative Physiology, Kyushu University Graduate School of Medical Sciences, 3-1-1 Maidashi, Higashi-ku, Fukuoka 812-8582, Japan2005 3 6 2005 1 19 19 15 2 2005 3 6 2005 Copyright © 2005 Rashid and Ueda; licensee BioMed Central Ltd.2005Rashid and Ueda; licensee BioMed Central Ltd.This is an Open Access article distributed under the 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 present study examined whether pre-injury administration of morphine can prevent partial sciatic nerve injury-induced neuropathic pain in mice. We observed that pre-injury administration of subcutaneous (s.c.) and intracerebroventricular (i.c.v.) morphine dose-dependently prevented the development of both thermal and mechanical hyperalgesia at 7 days following nerve injury in mice. The pre-injury morphine (s.c.)-induced analgesia was significantly blocked by pretreatment with naloxone injected s.c. or i.c.v., but not i.t., suggesting that systemic morphine produced the pre-emptying effects mainly by acting at the supra-spinal sites. Since it is believed that activation of descending monoaminergic mechanisms in spinal cord largely contributes to the supra-spinal analgesic effects of morphine, we investigated the involvement of serotonergic and noradrenergic mechanisms in spinal cord in the pre-injury morphine-induced analgesic effects. We found that pre-injury s.c. morphine-induced analgesic effect was significantly blocked by i.t. pretreatment with serotonergic antagonist, methysergide and noradrenergic antagonist, phentolamine. In addition, pre-injury i.t. injection of serotonin uptake inhibitor, fluoxetine and α2-adrenergic agonist, clonidine significantly prevented the neuropathic hyperalgesia. We next examined whether pre-injury morphine prevented the expression of neuronal hyperactivity markers such as c-Fos and protein kinase C γ (PKCγ) in the spinal dorsal horn. We found that pre-injury administration of s.c. morphine prevented increased expressions of both c-Fos and PKCγ observed following nerve injury. Similar results were obtained with i.t. fluoxetine and clonidine. Altogether these results suggest that pre-injury administration of morphine might prevent the development of neuropathic pain through activation of descending monoaminergic pain inhibitory pathways.
==== Body
Background
One of the critical factors that initiate and maintain chronic pain is central sensitization where neurons in the spinal dorsal horn become more excitable due to prior repetitive noxious stimuli [1]. Thus, preventing the initial cascade of neural events may eliminate the long-term hypersensitivity. Initiating an analgesic regimen before onset of such noxious stimulus in an attempt to prevent the central sensitization is known as preemptive analgesia [2]. The concept of preemptive analgesia was originally proposed at the beginning of the last century by Crile [3]. Since the revival of the concept again by Woolf in 1983 in experimental animals [4], it has been practiced in the clinic in order to lessen post-operative pain following various surgical operations [2,5-8]. In spite of some controversies regarding the effectiveness of preemptive analgesia in some clinical settings, it may have tremendous economic benefits due to savings from reduced length of hospital stay, fewer post-operative complications, and improved quality of life [9]. Preemptive analgesia strategies mainly include infiltration with local anesthetics, nerve block, epidural block, use of analgesics such as morphine, NSAIDS, cyclooxygenase (COX)-2 inhibitors, inhibition of pain pathways by NMDA antagonists etc. [2,7-9].
Both clinical and preclinical studies suggest that pre-operative administration of morphine and other opioid analgesics can improve post-operative pain management [10-12]. Recent studies also demonstrate that opioids are able to prevent central sensitization in animal models of pain [13]. However, the effectiveness of pre-injury morphine to prevent induction of nerve injury-induced neuropathic pain has been largely unknown. Smith et al., [14] reported that pre-injury administration of systemic morphine was less effective than α2-adrenergic receptor agonist, clonidine in preventing the mechanical hyperalgesia in a rat model of mononeuropathy. On the other hand, Puke and Wiesenfeld-Hallin [15] showed that pre-operative intrathecal administration of morphine, but not clonidine, prevented the autotomy behavior in a rat peripheral axotomy model. Therefore the exact mechanism of preemptive analgesic effect of morphine in nerve injury-induced pain is yet to be clarified. It is well known that μ-opioid receptors (MOP) are largely distributed in different brain areas with some distribution in the spinal dorsal horn and dorsal root ganglion neurons [16]. The analgesic effect of systemic morphine is, however, mainly produced by activation of MOP in the periaqueductal grey (PAG), and brainstem nucleus raphe magnus (NRM) and locus coeruleous (LC), ultimately activating the descending pain inhibitory pathways consisting mainly of the noradrenergic and serotonergic neuronal terminals to the spinal cord [17]. Direct activation of the spinal MOP by intrathecal morphine is also reported to produce potent acute analgesia in experimental animals [18,19]. However, the efficacy of both systemic and spinal morphine is reduced in neuropathic pain [19,20]. Therefore, the concept of pre-operative application of morphine could provide a way out to circumvent the limitations associated with acute administration of morphine against such painful conditions.
In the present study, we utilized a systematic approach to see the exact contribution of supra-spinal and spinal μ-opioid receptors in the pre-injury morphine-induced analgesic effects by administering it through various routes. We also examined the contribution of spinal monoaminergic systems in the pre-injury morphine-induced analgesic effects. In addition, we observed the effects of pre-injury administration of morphine on nerve injury-induced increases in expression of c-Fos and PKCγ, two important markers of neuronal hyperactivity, in the spinal cord.
Results
Pre-, but not post-, injury administration of morphine prevented the development of thermal and mechanical hyperalgesia in nerve-injured mice
Morphine, injected subcutaneously (s.c.) 30 min before partial sciatic nerve injury in mice, dose-dependently prevented the development of both thermal and mechanical hyperalgesia observed at 7 days after nerve injury with a significant effect at doses of 3 and 10 mg/kg s.c. (Fig. 1A,D). However, 10 mg/kg of s.c. morphine, injected 30 min after the nerve injury operation, failed to prevent the development of thermal or mechanical hyperalgesia in mice (Fig. 1A,D; last column). The pre-emptying effects of pre-injury morphine on post-injury pain continued at day 14 after nerve injury (data not shown). When we examined the effect of pre-injury intrathecal (i.t.) and intracerebroventricular (i.c.v.) morphine, only i.c.v. morphine significantly prevented the development of thermal and mechanical hyperalgesia (Fig. 1C,F). Although i.t. morphine produced some pre-emptying effects, it was statistically insignificant (Fig. 1B,E). Post-injury injection of both i.t. and i.c.v. morphine at 30 min after nerve injury did not produce any post-operative analgesia in the nerve-injured mice at 7 day after injury (data not shown). Pre-operative injection of morphine in sham-operated group of mice had no effect on the withdrawal latencies or thresholds observed at 7 days after sham operation (data not shown).
Figure 1 Pre-injury morphine-induced analgesia in partial sciatic nerve injury model mice. A-F: Pre-injury single administration of s.c. (1, 3, 10 mg/kg), and i.c.v. (0.3, 1, 3 nmol), but not i.t. (3, 10, 30 nmol), morphine prevented the development of thermal (A-C) and mechanical hyperalgesia (D-F) in the nerve-injured mice at 7 days following injury compared with vehicle pretreatment (Pre-injury Morph 0). Post-injury injection of 10 mg/kg s.c. morphine (30 min after surgery) failed to prevent the development of thermal or mechanical hyperalgesia. 'Sham' indicates thermal paw withdrawal latencies or mechanical paw withdrawal thresholds in control sham-operated mice. 'Morph' is morphine. Each data represents mean ± SE from 6–7 mice; *p < 0.05 compared with the group receiving vehicle as pretreatment.
Pre-injury subcutaneous (s.c.) morphine-induced analgesia was mediated by MOP in the supra-spinal sites
Since the above results suggest involvement of supra-spinal sites for pre-emptying effects of morphine, we further examined the contribution of μ-opioid receptors (MOP) in the brain and spinal cord in mediating the pre-injury systemic morphine-induced analgesic effect. As shown in Fig. 2A,B, pre-injury s.c. morphine (10 mg/kg)-induced analgesia was significantly blocked by pretreatment with 1 mg/kg of s.c. naloxone, a MOP antagonist. Moreover, pre-injury s.c. morphine-induced preemptive analgesia was blocked by pretreatment with 1 nmol of i.c.v. naloxone, but not with 10 nmol of i.t. naloxone, suggesting the involvement of supra-spinal MOPs in mediating the systemic morphine-induced pre-emptying effects. The s.c., i.t. and i.c.v. doses of naloxone were chosen from previous studies in mice [21].
Figure 2 Involvement of supra-spinal MOPs in the pre-injury morphine-induced analgesic effects. A,B: Blockade of pre-injury morphine (10 mg/kg, s.c.)-induced analgesia by pretreatment with s.c. (1 mg/kg) or i.c.v. (1 nmol), but not i.t. (10 nmol), injection of the μ-opioid receptor antagonist naloxone in thermal paw withdrawal (A) and mechanical paw pressure (B) tests. Thermal paw withdrawal latencies (PWL) or mechanical paw withdrawal thresholds (PWT) were measured at 7 days following nerve injury. 'Morph' is morphine. Each data represents mean ± SE from 6–7 mice; *p < 0.05 compared with the vehicle pretreatment. 'Veh' indicates group of mice receiving only vehicle saline before nerve injury operation.
Pre-injury morphine-induced analgesia is mediated through activation of the descending monoaminergic pathways in the spinal cord
Since the analgesic effect of systemic morphine is believed to be largely mediated through activation of the descending monoaminergic pain inhibitory pathways in the spinal cord [17], we next examined the effects of i.t. injections of antagonists of serotonergic and noradrenergic systems. As shown in Fig. 3A and 3B, i.t. pretreatment with 3 nmol of methysergide, a serotonergic antagonist, significantly blocked the pre-injury morphine-induced pre-emptying effects. Intrathecal (i.t.) pretreatment with 10 nmol of phentolamine, a noradrenergic antagonist also blocked the post-operative analgesia produced by pre-injury administration of 10 mg/kg of s.c. morphine. The antagonists alone had no effects in the nerve-injured mice (data not shown). We performed further experiments to explore whether i.t. injection of serotonergic or noradrenergic agonists could produce similar pre-emptying effects. As shown in Fig. 3C, i.t. pre-injury injection of serotonin uptake inhibitor fluoxetine and α-2 adrenergic agonist clonidine, dose-dependently prevented the thermal and mechanical hyperalgesia in injured mice. The doses of methysergide and phentolamine were chosen from previous reports in mice [21].
Figure 3 Involvement of descending monoaminergic pathways for the pre-injury morphine-induced analgesic effects. A,B: Pre-injury morphine (10 mg/kg. s.c.)-induced analgesia in thermal paw withdrawal (A) and mechanical paw pressure (B) tests were blocked by i.t. pretreatment with 3 nmol of serotonergic antagonist methysergide (Meth) or 10 nmol of adrenergic antagonist phentolamine (Phen). C: Dose-dependent pre-emptying effects induced by pre-injury i.t. injection of adrenergic agonist clonidine and serotonin uptake inhibitor fluoxetine in thermal paw withdrawal test. Thermal paw withdrawal latencies (PWL) or mechanical paw withdrawal thresholds (PWT) were measured 7 days following nerve injury. 'Morph' is morphine. Each data represents mean ± SE from 6–7 mice; *p < 0.05 compared with the group receiving vehicle as pretreatment.
Pre-injury administration of morphine prevented nerve injury-induced expression of c-fos in the spinal cord
The immediate early gene c-fos is an important marker of neuronal activity. Expression of Fos protein is rapidly increased in the spinal cord in response to peripheral noxious stimuli [22]. Induction of neuropathic pain has been correlated with nerve injury-induced short-term as well as long-term c-Fos expression in the spinal dorsal horn [23]. In the rat chronic constriction injury (CCI) model of neuropathic pain, c-Fos expression was increased in the spinal dorsal horn with a peak increase at 3 days after injury, and persisted for 30 days before returning to baseline level [23]. In the present report we examined whether pre-injury administration of morphine could prevent the c-Fos expression in the spinal cord following partial sciatic nerve ligation. As shown in Fig. 4C, partial sciatic nerve ligation increased expression of c-Fos in the ipsilateral side at 7 days after the injury compared with the ipsilateral side of the sham-operated mice (Fig. 4A). Some increase in c-Fos expression was also observed in the contralateral side of nerve-injured mice (Fig. 4D), compared with the control sham-operated mice (Fig. 4B). Pre-injury injection of 10 mg/kg of s.c. morphine 30 min before surgery prevented the injury-induced c-Fos expression in the spinal cord dorsal horn (Fig. 4E,F). When we counted the number of c-Fos-positive cells in the dorsal horn of the spinal cord of these groups of mice, we found that there was a significant increase in number of Fos-positive cells in the ipsilateral side of nerve-injured mice compared with the ipsilateral side of sham-operated mice at 7 days following injury (Fig. 4G). Pre-injury administration of 10 mg/kg of s.c. morphine significantly reduced the number of Fos-positive cells in the ipsilateral dorsal horn of nerve-injured mice (Fig. 4G).
Figure 4 Prevention of nerve injury-induced expression of c-Fos protein in the spinal cord by pre-injury administration of morphine. A,B: Only a few c-Fos-positive neurons were found the spinal dorsal horn of control sham-operated mice. C: At 7 days following peripheral nerve injury, numerous c-Fos-positive cells were observed in most dorsal horn laminas of ipsilateral side of the spinal cord. D: Some c-Fos-positive cells were also observed in the contralateral side. E,F: Pre-injury administration of 10 mg/kg of s.c. morphine prevented the injury-induced expression of c-Fos in the spinal dorsal horn. G: Histogram of the number of c-Fos positive cells in ipsi- and contralateral sides of sham-operated (sham), nerve-injured (NI) and pre-injury morphine treated (Pre-morp) mice from three animals of each treatment group taking three sections from each animal. * p < 0.05 indicates statistically significant difference compared between the ipsilateral sides of sham and nerve-injured mice. # p < 0.05 indicates significant compared between the ipsilateral sides of the nerve-injured and pre-injury morphine treated mice. Scale bar represents 100 μm for all images.
Pre-injury morphine, clonidine and fluoxetine prevented injury-induced increase in PKCγ expression in the spinal dorsal horn
We also examined the effects of pre-injury administration of morphine on expression of PKCγ in the spinal dorsal horn. Increased expression of PKCγ, an important component of central sensitization, is well documented in animal models of peripheral neuropathic pain [24,25]. In the present study we observed an increased expression of PKCγ in ipsilateral side of the spinal dorsal horn of partial sciatic nerve-injured mice at 7 days following injury compared with the ipsilateral side of control sham-operated mice (Fig. 5A–D,K). Pre-injury administration of 10 mg/kg of s.c. morphine almost completely prevented the increase in PKCγ expression observed following nerve injury (Fig 5E,F,K). Pre-injury i.t. administration of clonidine (30 nmol) and fluoxetine (30 nmol) also significantly prevented the injury-induced increase in PKCγ expression in spinal dorsal horn (Fig. 5G–J,K).
Figure 5 Prevention of nerve injury-induced increase in PKCγ expression in the spinal dorsal horn by pre-injury morphine. A,B: PKCγ immunoreactivity in the ipsilateral (A) and contralateral (B) sides of spinal dorsal horn of sham-operated mice. C,D: PKCγ-IR in the ipsilateral (C) and contralateral (D) sides of nerve-injured mice at 7 days after nerve injury. E,F: PKCγ-IR following pre-injury administration of morphine (10 mg/kg, s.c.) in ipsilateral (E) and contralateral (F) sides of nerve-injured mice at day 7. G-J: Pre-injury i.t. injection of clonidine (30 nmol) and fluoxetine (30 nmol) also prevented the injury-induced increase in PKCγ expression in spinal dorsal horn. K: Quantification of PKCγ immunoreactive fluorescence in ipsilateral sides of sham operated (sham), nerve-injured (injured), pre-injury morphine (PEM), pre-injury clonidine (PEC) and pre-injury fluoxetine (PEF) treated mice. Quantification of staining intensity was done using Scion imaging software for Macintosh. The data were represented as the ratio of staining intensity between ipsilateral and contralateral sides of each section in the treatment groups. Data were taken from three animals of each treatment group taking three separate sections from each animal. *, # p < 0.05. The scale bar represents 100 μm for all images.
Discussion
In the present study, we demonstrated that a pre-injury single administration of morphine could prevent development of thermal and mechanical hyperalgesia in the partial sciatic nerve injury model of neuropathic pain. We further demonstrated that pre-injury morphine-induced analgesia might be mediated through activation of the descending monoaminergic pathways in the spinal cord. In the present study, while pre-injury subcutaneous (s.c.) and intracerebroventricular (i.c.v.) morphine produced significant pre-emptying effects, intrathecal (i.t.) morphine only slightly prevented the injury-induced hyperalgesia which was statistically insignificant. These results suggest that the density of μ-opioid receptors (MOP) in the spinal cord might be insufficient to produce the necessary analgesic effect that could prevent the nerve injury-induced initial barrage of neuronal stimulation, ultimately leading to the development of central sensitization. Such results might be in contrary to the observation of strong acute analgesia by intrathecal morphine as reported previously [18,26]. However, we speculate that the distribution of MOP in the spinal cord is indeed sufficient to produce acute analgesia but might be insufficient to produce the necessary analgesia to prevent the nerve injury-induced initial barrage of neuronal stimulation compared with the supra-spinal sites, where MOPs are densely distributed [16]. Moreover, blockade of systemic morphine-induced preemptive analgesia by s.c. and i.c.v., but not i.t. naloxone further indicates the involvement of supra-spinal MOPs in the pre-emptying effects of morphine.
It is well known that the analgesic effect of systemic morphine is largely mediated by activation of MOPs in brainstem nuclei such as nucleus raphe magnus (NRM) and locus coeruleous (LC) that exert a net inhibitory effect on nociceptive transmission through descending monoaminergic pain inhibitory pathways in the spinal cord [17]. The serotoninergic and noradrenergic systems in the spinal cord also mediate the antinociception produced by intracerebroventricular (i.c.v.) injection of morphine [27]. Consistent with these lines of evidence, pre-injury systemic morphine-induced analgesia in our study was significantly blocked by intrathecal (i.t.) pretreatment with both serotonergic and noradrenergic antagonists. Moreover, pre-injury i.t. injection of the serotonin uptake inhibitor fluoxetine and the α-2 adrenergic agonist clonidine produced significant analgesia, further indicating ability of the spinal serotonergic and noradrenergic system to produce sufficient level of pre-emptying effects. It has been reported that both serotonergic and noradrenergic terminals innervate the presynaptic terminals of small nociceptive primary afferents [28], and can inhibit neurotransmitter release [29].
Pre-injury administration of morphine also prevented the injury-induced increases in expression of c-Fos and PKCγ in the spinal cord. Induction of neuropathic pain has been correlated with nerve injury-induced short-term as well as long-term c-Fos expression in the spinal dorsal horn [23]. The blockade of nerve injury-induced c-Fos expression in the spinal cord by pre-injury morphine indicates that preemptive systemic morphine is able to prevent injury-induced neuronal cascade that ultimately might cause neuropathic pain. Activation of PKC in the spinal cord dorsal horn, which triggers sustained activation of N-methyl-D-aspartate (NMDA) receptors, also serves as a marker of central sensitization [30]. Among different isoforms of PKC, the γ isoform was well studied with regard to neuropathic pain. Increased expression of PKCγ is well documented in animal models of peripheral neuropathic pain [24,25,31]. Reduced hyperalgesia was also observed following peripheral nerve injury in mice lacking PKCγ [32]. In the present study, we observed increased expression of PKCγ in the spinal dorsal horn following partial sciatic nerve injury, and pre-injury administration of morphine prevented such increased expression. It has been reported that majority of PKCγ-containing cells in the spinal dorsal horn are mainly excitatory interneurons, and does not contain the μ-opioid receptors (MOP) [33]. This might be also one of the reasons for the ineffectiveness of intrathecally injected morphine to produce significant preemptive analgesia in our studies. Finally, the blockade of injury-induced increase in PKCγ expression by i.t. clonidine and fluoxetine suggest that activation of descending monoaminergic system in spinal cord by systemic morphine might have prevented the development of central sensitization.
Conclusion
In conclusion, results of the present study demonstrate that pre-injury administration of morphine could prevent the development of peripheral nerve injury-induced neuropathic pain through activation of descending pain inhibitory mechanisms. These results may improve management of chronic neuropathic pain by proper use of morphine.
Methods
Animals
Male ddY mice weighing 20–25 g were used throughout the experiments. The mice were housed in a room maintained at 21 ± 2°C, 55 ± 5 % relative humidity and an automatic 12-h light/dark cycle with free access to standard laboratory diet and tap water. The animals were adapted to the testing environment (maintained at 21 ± 2°C, 55 ± 5 % relative humidity and 12-h light/dark cycle) by keeping them in the testing room 24 h before the experiments. Experiments were performed during the light phase of the cycle (10:00 – 17:00). All procedures were approved by Nagasaki University Animal Care Committee and complied with the recommendations of International Association for the Study of Pain [34].
Drugs and injection methods
Following drugs were purchased: morphine hydrochloride (Takeda Pharma. Co. Ltd., Japan), naloxone hydrochloride, methysergide maleate, phentolamine hydrochloride, fluoxetine hydrochloride, and clonidine hydrochloride (all from Sigma Co., St Louis, MO, USA). All drugs were dissolved in physiological saline. Physiological saline was used for control injections. The intrathecal (i.t.) injections were performed free hand between spinal L5 and L6 segments according to the method of Hylden and Wilcox [35]. The intracerebroventricular (i.c.v.) injections were carried out into the left lateral ventricle of mice. Injections were performed using a Hamilton microsyringe fitted with a 26-gauge i.c.v. needle, according to the method of Haley and McCormick [36]. The site of injection was 2 mm caudal and 2 mm lateral to the bregma, and 3 mm in depth from the skull surface. Both i.t. and i.c.v. injections were given in a volume of 5 μl. The mice received the subcutaneous (s.c.) injections in a volume of 0.1 ml/10 g body weight.
Partial ligation of sciatic nerve
Partial ligation of the sciatic nerve of mice was performed under pentobarbital (50 mg/kg i.p.) anesthesia, following the methods of Malmberg and Basbaum [37]. Briefly, the common sciatic nerve of the right hind limb of mice was exposed at high thigh level through a small incision and dorsal 1/3 to 1/2 of the nerve thickness was tightly ligated with a silk suture. The wound was closed with a single muscle suture and antibiotic powder was dusted over the wound area following surgery. Sham operation was performed similarly except without touching the sciatic nerve. Immediately following surgery, the animals were kept in a soft bed cage with some food inside so that the animals could feed themselves without difficulty in standing. The wound healed within 1–2 days and the animals behaved normally. Experiments were carried out at 7 or 14 days post-ligation.
Hargreaves thermal paw withdrawal test
Analgesia was measured from the latency to withdrawal evoked by exposing the right hind paw to a thermal stimulus. Mice were placed under Plexiglas cages on top of a glass sheet. The thermal stimulus (IITC Inc., Woodland Hills, CA, USA) was positioned under the glass sheet to focus the projection bulb exactly on the middle of plantar surface of the animals. A mirror attached to the stimulus permitted visualization of the undersurface of the paw. After one hour of adaptation, paw withdrawal latencies were measured at every 10 min interval until 60 min with vehicle or drug pretreatment. A cut-off thermal latency of 20 s was set in order to prevent tissue damage.
Paw pressure test
Experiments were performed as described previously [38]. Briefly, mice were placed under a Plexiglas chamber on a 6 mm × 6 mm wire mesh grid floor and were allowed to accommodate for a period of one hour. The mechanical stimulus was then delivered onto the middle of the plantar surface of right hind-paw using a Transducer Indicator (Model 1601, IITC Inc., Woodland Hills, USA). The paw withdrawal thresholds were measured at every 10 min interval until 60 min with vehicle or drug pretreatment. In this experiment, a cut-off pressure of 20 g was set to avoid tissue damage.
DAB immunostaining for c-Fos
Mice were deeply anesthetized with i.p. pentobarbital and perfused transcardially with 40 ml of 0.1 M potassium free phosphate buffered saline (K+ free PBS, pH 7.4) followed by 40 ml of 4% paraformaldehyde (PFA) in 0.1 M K+ free PBS. The spinal cord between L4 – L5 segments was removed and post-fixed in 4% PFA for 1 hour. Then, the sample was transferred to 25% sucrose solution (in 0.1 M K+ free PBS) overnight for cryoprotection. Next day, the spinal cord sample was fast-frozen in cryoembedding compound on a mixture of ethanol and dry-ice and stored at -80°C until use. The spinal cord sample was cut as 20 μm thick transverse sections with a cryostat, thaw-mounted on silane-coated glass slide and air dried overnight at room temperature (RT). For c-Fos immunolabeling, spinal cord sections were washed 3 times with K+ free PBS for 5 min each then incubated in excess 100% methanol with 0.1% H2O2. After 3 washings with K+ free PBS, the sections were incubated in excess blocking buffer containing 10% normal goat serum and 2% bovine serum albumin in PBST (2% NaCl, 0.1% Triton-X 100 in K+ free PBS) for 60 min at RT. The sections were washed and reacted with rabbit polyclonal antibody raised against the c-Fos protein (1:1000 in 2% BSA in PBST solution; sc-7202, Santa Cruz Biotechnology, CA, USA) at 4°C overnight. After thorough washings, the sections were incubated with secondary antibody, biotinylated anti-rabbit IgG (1:200 in 2% BSA in PBST solution; Vector, CA, USA) at RT for 60 min, and subsequently with ABC complex (Vector, CA) at RT for 60 min. The antigen-antibody reaction sites were visualized by incubation with a solution containing 0.005% 3,3'-diaminobenzidine tetrahydrochloride (DAB; Dojindo, Japan), 0.002% H2O2, 0.001% nickel ammonium sulfate and 0.002% cobalt chloride in 0.1 M K+ free PBS until the black reaction products appear. The reaction was stopped by washing with ice-cold PBS. After 3–4 washings, the sections were cover-slipped and visualized under a light microscope. The number of c-Fos-positive cells in the ipsi-and contralateral sides of dorsal horn gray mater of the spinal cord was then counted from lamina I-VI and plotted in a bar graph.
Fluorescence immunohistochemistry for PKCγ
The spinal cord sections were prepared as described above. For immunostaining of PKCγ, the spinal cord sections were first pre-blocked with blocking buffer containing 10% normal goat serum and 2% bovine serum albumin in PBST. The sections were then reacted with a rabbit polyclonal antibody raised against the γ isoform of protein kinase C (1:500 in 2% BSA in PBST solution; sc-211, Santa Cruz Biotechnology, CA, USA) at 4°C overnight. The sections were then incubated with a FITC-conjugated anti-rabbit IgG (1:200; Santa Cruz Biotechnology) for 60 min at RT. The sections were washed thoroughly, cover-slipped with Perma Fluor (Thermo Shandon, Pittsburgh, PA, USA) and examined under a fluorescence microscope (Olympus, Tokyo, Japan). Quantification of the intensity of PKCγ-positive fluorescence was then done using Scion imaging software for Macintosh (Scion Corporation, USA).
Statistical analysis
Statistical evaluations of the data were performed by comparison with repeated measures analysis of variance (ANOVA) with suitable post-hoc tests. The criterion of significance was set at p < 0.05. All results are expressed as the mean ± SEM.
List of Abbreviations
s.c., subcutaneous; i.t., intrathecal; i.c.v., intracerebroventricular; MOP, μ-opioid receptor; PBS, phosphate-buffered saline; AUC, area under the curve; ANOVA, analysis of variance; PKCγ, protein kinase C γ isoform.
Competing interests
There are no financial as well as non-financial competitions with any other people or organizations.
Authors' contributions
H. Ueda contributed to the conception and drafting of manuscript and has final approval of this version to be published.
M.H. Rashid contributed to the conception, design, data acquisition and drafting of the manuscript, and has final approval of this version to be published.
Acknowledgements
Parts of this study were supported by Grants-in-Aid from the Ministry of Education, Science, Culture and Sports of Japan, the Human Frontier Science Program (H.U.), and the Japan Society for the Promotion of Science (M.H.R.). We appreciate the valuable advice of Drs. Megumu Yoshimura and Makoto Inoue during preparation of this manuscript.
==== Refs
Woolf CJ Salter MW Neuronal plasticity: increasing the gain in pain Science 2000 288 1765 1769 10846153 10.1126/science.288.5472.1765
Kissin I Preemptive analgesia Anesthesiology 2000 93 1138 43 11020772 10.1097/00000542-200010000-00040
Crile GW The kinetic theory of shock and its prevention through anoci-association Lancet 1913 185 7 16 10.1016/S0140-6736(01)65552-1
Woolf CJ Evidence for a central component of postinjury pain hypersensitivity Nature 1983 308 686 8 6656869 10.1038/306686a0
Filos KS Vagianos CE Pre-emptive analgesia: how important is it in clinical reality? Eur Surg Res 1999 31 122 132 10213850 10.1159/000008630
Farris DA Fiedler MA Preemptive analgesia applied to postoperative pain management AANA J 2001 69 223 8 11759566
Kelly DJ Ahmad M Brull SJ Preemptive analgesia II: recent advances and current trends Can J Anaesth 2001 48 1091 101 11744585
Moiniche S Kehlet H Dahl JB A qualitative and quantitative systematic review of preemptive analgesia for postoperative pain relief: the role of timing of analgesia Anesthesiology 2002 96 725 41 11873051 10.1097/00000542-200203000-00032
Gottschalk A Smith DS New concepts in acute pain therapy: preemptive analgesia Am Fam Physician 2001 63 1979 1984 11388713
Richmond CE Bromley LM Woolf CJ Preoperative morphine pre-empts postoperative pain Lancet 1993 342 73 75 8100911 10.1016/0140-6736(93)91284-S
Abram SE Yaksh TL Morphine, but not inhalation anesthesia, blocks post-injury facilitation. The role of preemptive suppression of afferent transmission Anesthesiology 1993 78 713 721 8385425
Reichert JA Daughters RS Rivard R Simone DA Peripheral and preemptive opioid antinociception in a mouse visceral pain model Pain 2001 89 221 227 11166478 10.1016/S0304-3959(00)00365-1
Sandkuhler J Ruscheweyh R Opioids and central sensitisation: I. Preemptive analgesia Eur J Pain 2005 9 145 8 15737804 10.1016/j.ejpain.2004.05.012
Smith GD Harrison SM Wiseman J Elliott PJ Birch PJ Pre-emptive administration of clonidine prevents development of hyperalgesia to mechanical stimuli in a model of mononeuropathy in the rat Brain Res 1993 632 16 20 8149225 10.1016/0006-8993(93)91132-C
Puke MJ Wiesenfeld-Hallin Z The differential effects of morphine and the alpha 2-adrenoceptor agonists clonidine and dexmedetomidine on the prevention and treatment of experimental neuropathic pain Anesth Analg 1993 77 104 109 8100405
Minami M Satoh M Molecular biology of the opioid receptors: structures, functions and distributions Neurosci Res 1995 23 121 145 8532211 10.1016/0168-0102(95)00933-K
Sawynok J The 1988 Merck Frosst Award. The role of ascending and descending noradrenergic and serotonergic pathways in opioid and non-opioid antinociception as revealed by lesion studies Can J Physiol Pharmacol 1989 67 975 988 2688867
Yaksh TL Analgetic actions of intrathecal opiates in cat and primate Brain Res 1978 153 205 10 98219 10.1016/0006-8993(78)91146-0
Ossipov MH Lopez Y Nichols ML Bian D Porreca F The loss of antinociceptive efficacy of spinal morphine in rats with nerve ligation injury is prevented by reducing spinal afferent drive Neurosci Lett 1995 199 87 90 8584250 10.1016/0304-3940(95)12022-V
Rashid MH Inoue M Toda K Ueda H Loss of peripheral morphine analgesia contributes to the reduced effectiveness of systemic morphine in neuropathic pain J Pharmacol Exp Ther 2001 309 380 387 14718584 10.1124/jpet.103.060582
Kawabata A Kasamatsu K Takagi H L-Tyrosine-induced antinociception in the mouse: involvement of central delta-opioid receptors and bulbo-spinal noradrenergic system Eur J Pharmacol 1993 233 255 260 8385625 10.1016/0014-2999(93)90058-P
Hunt SP Pini A Evan G Induction of c-fos-like protein in spinal cord neurons following sensory stimulation Nature 1987 328 632 634 3112583 10.1038/328632a0
Yamazaki Y Maeda T Someya G Wakisaka S Temporal and spatial distribution of Fos protein in the lumbar spinal dorsal horn neurons in the rat with chronic constriction injury to the sciatic nerve Brain Res 2001 914 106 114 11578603 10.1016/S0006-8993(01)02783-4
Mao J Price DD Phillips LL Lu J Mayer DJ Increases in protein kinase C gamma immunoreactivity in the spinal cord dorsal horn of rats with painful mononeuropathy Neurosci Lett 1995 198 75 78 8592645 10.1016/0304-3940(95)11975-3
Miletic V Bowen KK Miletic G Loose ligation of the rat sciatic nerve is accompanied by changes in the subcellular content of protein kinase C beta II and gamma in the spinal dorsal horn Neurosci Lett 2000 288 199 202 10889342 10.1016/S0304-3940(00)01237-4
Wegert S Ossipov MH Nichols ML Bian D Vanderah TW Malan TP JrPorreca F Differential activities of intrathecal MK-801 or morphine to alter responses to thermal and mechanical stimuli in normal or nerve-injured rats Pain 1997 71 57 64 9200174 10.1016/S0304-3959(97)03337-X
Yaksh TL Direct evidence that spinal serotonin and noradrenaline terminals mediate the spinal antinociceptive effects of morphine in the periaqueductal gray Brain Res 1979 160 180 185 581478 10.1016/0006-8993(79)90616-4
Bourgoin S Pohl M Mauborgne A Benoliel JJ Collin E Hamon M Cesselin F Monoaminergic control of the release of calcitonin gene-related peptide-and substance P-like materials from rat spinal cord slices Neuropharmacology 1993 32 633 640 7689707 10.1016/0028-3908(93)90076-F
Levine JD Fields HL Basbaum AI Peptides and the primary afferent nociceptor J Neurosci 1993 13 2273 2286 8501507
Willis WD Role of neurotransmitters in sensitization of pain responses Ann N Y Acad Sci 2001 933 142 156 12000017
Inoue M Rashid MH Fujita R Contos JJA Chun J Ueda H Initiation of neuropathic pain requires lysophosphatidic acid receptor signaling Nature Med 2004 10 712 718 15195086 10.1038/nm1060
Malmberg AB Chen C Tonegawa S Basbaum AI Preserved acute pain and reduced neuropathic pain in mice lacking PKCgamma Science 1997 278 279 283 9323205 10.1126/science.278.5336.279
Polgar E Fowler JH McGill MM Todd AJ The types of neuron which contain protein kinase C gamma in rat spinal cord Brain Res 1999 833 71 80 10375678 10.1016/S0006-8993(99)01500-0
Zimmermann M Ethical guidelines for investigations of experimental pain in conscious animals Pain 1983 16 109 110 6877845 10.1016/0304-3959(83)90201-4
Hylden JL Wilcox GL Intrathecal morphine in mice: a new technique Eur J Pharmacol 1980 67 313 316 6893963 10.1016/0014-2999(80)90515-4
Haley TJ McCormick WG Pharmacological effects produced by intracerebral injection of drugs in the conscious mouse Br J Pharmacol 1957 12 12 15 13413144
Malmberg AB Basbaum AI Partial sciatic nerve injury in the mouse as a model of neuropathic pain: behavioral and neuroanatomical correlates Pain 1998 76 215 222 9696476 10.1016/S0304-3959(98)00045-1
Rashid MH Ueda H Nonopioid and neuropathy-specific analgesic action of the nootropic drug nefiracetam in mice J Pharmacol Exp Ther 2002 303 226 231 12235255 10.1124/jpet.102.037952
| 15932652 | PMC1175856 | CC BY | 2021-01-04 16:40:08 | no | Mol Pain. 2005 Jun 3; 1:19 | utf-8 | Mol Pain | 2,005 | 10.1186/1744-8069-1-19 | oa_comm |
==== Front
Reprod Biol EndocrinolReproductive biology and endocrinology : RB&E1477-7827BioMed Central London 1477-7827-3-211594147210.1186/1477-7827-3-21ResearchEffect of electro-acupuncture on ovarian expression of α (1)- and β (2)-adrenoceptors, and p75 neurotrophin receptors in rats with steroid-induced polycystic ovaries Manni Luigi [email protected] Thomas [email protected]äng Agneta [email protected] Luigi [email protected] Elisabet [email protected] Cardiovascular Institute and Wallenberg Laboratory, Sahlgrenska Academy, Göteborg University, SE-413 45 Göteborg, Sweden2 Institute of Neurobiology and Molecular Medicine (CNR), Rome, Italy3 Rehabilitation Medicine, Karolinska Hospital, SE-171 77 Stockholm, Sweden4 Department of Obstetrics and Gynaecology, Sahlgrenska University Hospital, Sahlgrenska, SE-413 45 Göteborg, Sweden5 Institute of Occupational Therapy and Physical Therapy, Sahlgrenska Academy, Göteborg University, SE-405 30 Göteborg, Sweden2005 7 6 2005 3 21 21 26 5 2005 7 6 2005 Copyright © 2005 Manni et al; licensee BioMed Central Ltd.2005Manni 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
Estradiol valerate (EV)-induced polycystic ovaries (PCO) in rats is associated with an increase in ovarian sympathetic outflow. Low-frequency (2 Hz) electro-acupuncture (EA) has been shown to modulate sympathetic markers as well as ovarian blood flow as a reflex response via the ovarian sympathetic nerves, in rats with EV-induced PCO.
Methods
In the present study, we further tested the hypothesis that repeated 2 Hz EA treatments modulate ovarian sympathetic outflow in rats with PCO, induced by a single i.m. injection of EV, by investigating the mRNA expression, the amount and distribution of proteins of α1a-, α1b-, α1d-, and β2-adrenoceptors (ARs), as well as the low-affinity neurotrophin receptor (p75NTR).
Results
It was found that EV injection results in significantly higher mRNA expression of ovarian α1b- and α1d-AR in PCO rats compared to control rats. The p75NTR and β2-ARs mRNA expression were unchanged in the PCO ovary. Low-frequency EA resulted in a significantly lower expression of β2-ARs mRNA expression in PCO rats. The p75NTR mRNA was unaffected in both PCO and control rats. PCO ovaries displayed significantly higher amount of protein of α1a-, α1b- and α1d-ARs, and of p75NTR, compared to control rats, that were all counteracted by repeated low-frequency EA treatments, except for α1b-AR.
Conclusion
The present study shows that EA normalizes most of the EV-induced changes in ovarian ARs. Furthermore, EA was able to prevent the EV-induced up regulation of p75NTR, probably by normalizing the sympathetic ovarian response to NGF action. Our data indicate a possible role of EA in the regulation of ovarian responsiveness to sympathetic inputs and depict a possible complementary therapeutic approach to overcoming sympathetic-related anovulation in women with PCOS.
==== Body
Introduction
Polycystic ovary syndrome (PCOS) is a heterogeneous endocrine and metabolic disorder recognized as the primary cause of infertility in women of the reproductive age [1]. The syndrome is associated with ovulatory dysfunction, abdominal obesity, hyperandrogenism, and profound insulin resistance [1].
The precise etiology of the disease is unknown, even though the disturbances detected in PCOS has been attributed to primary defects in the hypothalamus-pituitary-adrenal (HPA) axis, the ovarian microenvironment, the adrenal gland, and the insulin/insulin-like growth factor (IGF)-I metabolic regulatory system [1]. That the sympathetic nervous system may be a primary factor in the development and maintenance of PCOS has been suggested by several investigators [2-5].
The utility of murine models of polycystic ovaries (PCO) has been discussed [6]. Even though it is impossible to reproduce human PCOS in an animal model, such a model may provide important leads. Studies on adult normal cycling rats found that a single intramuscular (i.m.) injection of estradiol valerate (EV) causes acyclicity and formation of PCO [7]. The EV-induced rat PCO model reflects some endocrinological and morphological characteristics of human PCOS, and it is assumed that activity in the ovarian sympathetic nerves is higher than in normal rats [8-10]. This is evidenced by an early increase in ovarian levels of norepinephrine (NE), an enhanced release of NE from ovarian nerve terminals, an increased activity of the catecholamine synthesis-limiting enzyme tyrosine hydroxylase (TH), and down-regulation of β2-adrenoceptors (ARs) in theca-interstitial cells [8-10].
The expression of other types of ARs in the ovary, namely the α1-ARs, has been evaluated by functional studies. α1-ARs are members of the G protein-coupled receptors and play critical roles in the regulation of a variety of physiological processes [11]. Within this classification, there are three subtypes: α1a, α1b, and α1d [11]. The α1a-AR subtype has been reported to be implicated in the maintenance of vascular basal tone, the α1b-AR subtypes to participate in the response to exogenous agonists, and that the α1d-AR subtype is a predominant mediator of arterial vasoconstriction. In vitro studies have demonstrated that α-AR are involved in the regulation of ovarian blood flow [12] and most probably in the ovarian steroidogenesis [13]. In a recent study, we found that the expression of all the α1-AR subtypes in the ovaries of PCO rats significantly differs from that of controls and varies at different time points after EV injection, indicating a possible participation of this ARs in the development of EV-induced PCO [14].
It has been demonstrated that the development of ovarian follicular cysts in steroid-induced PCO in rats is preceded by an increased synthesis of ovarian nerve growth factor (NGF) and low-affinity neurotrophin receptor (p75NTR) mRNA [10]. Thus, blocking the actions of intra-ovarian NGF restores estrus cyclicity as well as structural and functional features of the ovary in EV-induced PCO in rats [10], suggesting that hyper activation of sympathetic input in PCO is related to an overproduction of NGF.
Electro-acupuncture (EA) is a non-pharmacological method known to initiate a number of reactions at the spinal level and centrally in the brain [15,16]. We have recently demonstrated that repeated low-frequency EA treatments induced regular ovulations in more than one-third of the women affected by PCOS and normalized endocrine and neuroendocrine parameters without any negative side-effects [17]. These observation suggest that EA effects are mediated through inhibition of the activity of the sympathetic nervous system since EA is known to modulate various autonomic functions [17]. Moreover, using the steroid-induced PCO model, we found that repeated treatments of low-frequency EA in somatic segment related to the innervation of the ovary, reduced high concentrations of ovarian NGF, corticotrophin-releasing factor (CRF), and endothelin-1 as well as increased low concentrations of hypothalamic β-endorphin [18-21]. Furthermore, low-frequency EA increases ovarian blood flow as a reflex response via the ovarian sympathetic nerves, whereas high frequency decreases ovarian blood flow as a passive response following systemic circulatory changes in both normal and PCO rats [22,23]. These results suggest that repeated treatments of low-frequency, but not high frequency EA, can inhibit high activity in the autonomic nervous system. However, the mechanism implicated in this event is not clearly known.
The present study was undertaken to investigate the effect of repeated treatments of low-frequency EA on ovarian sympathetic innervation in rats with steroid-induced PCO. To address this question, we studied the mRNA expression and protein amount and distribution of the sympathetic markers α1a-, α1b-, α1d-, and β2-AR, and of p75NTR.
Materials and methods
Animals
Thirty-two virgin adult cycling Wistar Kyoto rats (Möllegaard, Denmark) weighing 205-230g were housed four to a cage at a controlled temperature of 22°C with a 12-h light:12-h dark cycle for at least 1 week before and throughout the experimental periods. The rats had free access to pelleted food and tap water. Sixteen rats, those in the two PCO groups described below, were each given a single i.m. injection of 4 mg EV (Riedeldehaen, Germany) in 0.2 ml oil, to induce well-defined PCO [7,18]. Sixteen rats, those in the two Oil groups described below, received a single i.m. injection of 0.2 ml oil (arachidis oleum, Apoteket AB, Umeå, Sweden) only. Thirty to thirty-three days after i.m. injection of EV, i.e. 2 days after the last EA treatment, the rats was killed by decapitation. The injections and the finalizing of the experiment was done independent of cycle day [7,18]. The experiments were carried out according to the principles and procedures outlined in the National Institute of Health (NIH) Guide for the Care and Use of Laboratory Animals and were approved by the local animal ethics committee at Göteborg University, Göteborg, Sweden
Electro-acupuncture treatment
The rats were divided into four experimental groups: i) an Oil group (control, n = 8), ii) an Oil group receiving EA (EA, n = 8), iii) a PCO group (PCO, n = 8), and iv) a PCO group receiving EA (PCO+EA, n = 8).
All groups were anaesthetized for 25 minutes on 12 occasions as described below. The EA and the PCO+EA groups received EA every second day during anesthesia. The EA and the PCO+EA group underwent the first EA treatment 2 days after the EV injection. The points chosen for stimulation were bilateral in the mm. biceps femoris and erector spinae, in somatic segments corresponding to the innervation of the ovaries (Figure 1). The needles (Hegu: Hegu AB, Landsbro, Sweden) were inserted to depths of 0.5-0.8 cm and then attached bilaterally to an electrical stimulator (CEFAR ACUS 4, Cefar, Lund, Sweden). The points were electrically stimulated with a low burst frequency of 2 Hz; each pulse had a duration of 180 μsec, a burst length of 0.1 sec, and a burst frequency of 80 Hz. The intensity (1.5-2 mA) was adjusted until local muscle contractions were observed to reflect the activation of muscle-nerve afferents (A-delta fibers and possibly C fibers). The location and type of stimulation were the same in all rats.
Figure 1 Schematic drawing of the placement and stimulation of the acupuncture needles. Two needles were placed bilaterally in m. erector spinae at the level of Th12 and two were placed in m. quadriceps bilaterally. The needles were then attached to an electrical stimulator for electro-acupuncture (EA) treatment. Reprinted with permission from Biol Reprod (2000) 63:1507-1513.
Anaesthetization
During each treatment, all rats were anaesthetized superficially with an intraperitoneal (i.p.) injection of a mixture of Ketamin (50 mg/kg; PARKE-DAVIS, Warner Lambert Nordic AB, Solna, Sweden) and Rompun (20 mg/kg; Bayer, Bayer AG, Leverkusen, Germany). On day 30 after the i.m. injection of EV, the rats was decapitated, that is, 1-2 days after the last EA treatment.
Tissues
At the completion of the experiment, the ovaries were quickly dissected on dry ice. One ovary was divided in two pieces, weighed, and snap frozen in liquid nitrogen and stored at -80°C until extraction. The second ovary was fixed in buffered 4% formaldehyde for at least 24 hours in preparation for AR, and p75NTR immunohistochemistry.
Real-Time PCR for adrenoceptors
Total RNA from the ovary was extracted using RNeasy Mini kits (Qiagen, Hilden, Germany). First-strand cDNA was synthesized from 1 μg of total RNA with TaqMan reverse transcription reagents (Applied Biosystems., Foster City, CA). Each 100 μl RT-PCR reaction contained 1 μg of total, 1X TaqMan RT buffer, 5 mM MgCl2, 2.5 mM random hexamers, 1 mM dNTP, 0.4 U/ml RNase inhibitor, and 1.25 U/ ml Multiscribe RT (PE Applied Biosystems, Foster City, CA, USA). Reverse transcription was carried out in a PTC-200 PCR system (MJ Research., Boston, MA, USA) at 25°C for 10 min, 48°C for 30 min and 95°C for 5 min.
The polymerase chain reaction (PCR) analysis was performed using the ABI Prism 7700 Sequence Detection System (PE Applied Biosystems, Stockholm, Sweden) and FAM-labeled probe specific for α1a-AR (Rn00567876m1), α1b-AR (ADRA A1B-EX 152027A02), α1d-AR (Rn00577931ml), and β2-AR (Rn00560650s1) (PE Applied Biosystems). Designed primers and a VIC-labeled probe for Glyceraldehyde-3-phosphate dehydrogenase (GAPDH) (NM_031144) were included in the reactions as an internal standard. cDNA was amplified under the following conditions: 1 cycle at 50°C for 2 min and 95°C for 10 min, followed by 40 cycles at 95°C for 15 s and 60°C for 1 min. The amount of mRNA of each gene was calculated using the standard curve method (following the instructions in User Bulletin no. 2, PE Applied Biosystems) and adjusted for the expression of GAPDH.
Reverse Transcriptase-PCR-ELISA for p75NTR
The expression of p75NTR-mRNA was evaluated using the reverse transcriptase (RT)-PCR enzyme-linked immunosorbent assay (ELISA) protocol, exactly as previously described by Tirassa and co-workers [24]. Total RNA was extracted from the ovaries using the method of Chomczynski and Sacchi [25] as modified in the TRIzol Kit (Invitrogen AB, Lidingö, Sweden). Complementary DNA was synthesized from 1 μg of total RNA using 250 ng Oligo (dT)12–18 primer and 200 Units of M-MLV RT (Promega Italia, Milan, Italy) in 20 μl of total volume reaction. p75 and GAPDH genes were co-amplified in a single-tube PCR reaction (35 cycles: 1 min at 95°C; 1 min at 55°C; 2 min at 72°C) using 5'-biotinylated specific primers to generate biotinylated PCR products detectable by digoxygenin-labeled probes in an immuno-enzymatic assay. Primer/probe sequences are as follows: p75NTR biotinylated forward: 5'CGTGTT CTCCTGCCAGGACA3'; p75NTR reverse: 5'GAGATGCCACTGTCGCTGTG3'; p75NTR digoxygenin-labeled probe: 5'ACAGCAGCCAAGATGGAGCAATAGACAGG3'; GAPDH biotinylated forward: 5'CACCACCATGGAGAAGGCC3'; GAPDH reverse: 5'GATGGATGCCTTGGCCAGG3'; GAPDH digoxygenin-labeled probe: 5'ACAATCTTGAGTGAGTTGTCATATTTCTCG3'. The amount of amplified products was measured at an optical density (O.D.) of 450/690 nm (O.D. 450/690) using a Dynatech ELISA Reader 5000. A GAPDH level of O.D. 450/690 was used to normalise the relative differences in sample size, differences in the integrity of the individual RNA, and variations in RT efficiency. For exact methodological details see Tirassa et al. [24].
Immunohistochemistry for adrenoceptors and p75NTR
Commercially available antibodies were used to detect α1a-AR (α1a-AR [C-19]: sc-1477, Santa Cruz, California, USA), α1b-AR (α1b-AR [C-18]: sc-1476, Santa Cruz, California, USA), α1d-AR (α1d-AR [H-142]: sc-10721, Santa Cruz, California, USA), and β2-AR (β2-AR [M-20]: sc-1570, Santa Cruz, California, USA) by immunohistochemistry. The monoclonal antibody anti-p75NTR (clone 192) [26] was produced and purified in our laboratory.
Serial, 15-μm thick sections of each ovary were cut with a cryostat and processed for immunohistochemistry. Briefly, sections were blocked with a 10 minutes incubation in 3% hydrogen peroxide and 10% methanol in PBS containing 0.1% Triton X-100 (PBST), followed by a 30 minutes incubation in 10% normal serum dissolved in PBST. Then, sections were incubated overnight at 4°C with the primary antibody diluted in PBST (rabbit and goat anti-ARs: 5 μg/ml; monoclonal anti p75NTR: 1 μg/ml). Sections were then incubated with biotinylated anti-rabbit IgG (α1d- and β2-AR), anti-goat IgG (α1a- and α1b-AR) or anti-mouse IgG (p75NTR) antibodies (all of them from Vector Lab. Inc., Burlingame, CA, and used according to manufacturer instructions at 1:300 dilution) diluted in PBST. Diaminobenzidine was used to detect the immuno-complex. To assess staining specificity, sections were incubated with non-specific rabbit, goat or mouse IgG (Zymed Lab Inc, San Francisco, Ca) and used as controls. Immunostained sections were evaluated under the Nikon Eclypse E600 microscope equipped with the Nikon DMX 1200 digital camera connected to a PC computer. Sections were coded, and positive cells were counted in 10 sections coming from 5 different ovaries (i.e. 2 sections per ovary) per experimental group. Cell count was carried out using the image processing and analysis program Nikon-Lucia, and measurements were standardized between the experimental groups using the same calibration system and threshold (see below). The number of immunoreactive cells (mean ± SEM) was determined in 20× magnification images over an image area of 40000 μm2. Five non-overlapping areas per section were counted. Since the image analyzer determines the optical density of immunoreactions using a grey scale thresholding operation, measurements were standardized between groups using the following criteria: 1) all measurements were conducted after the same calibration of the image analysis system, 2) thresholding was carried out to the same value for each image, 3) the grey scale was calibrated to a range of 25-150 arbitrary units. Objects with higher or lower grey levels were not considered. A morphological program, which selects only cell bodies – but not small fragments or cells that do not have a complete soma – was also used to quantify immunopositive cells.
Statistical analyses
All statistical evaluations were performed using the Stat View package for Macintosh (Abacus Concepts Inc., Berkeley, CA, USA). mRNA expression and immunopositive cell counts of α1- and β2-ARs, and p75NTR in the ovaries were evaluated using one-way analysis of variance (ANOVA), and the groups were tested using multiple comparisons with the correction of Fisher PSD. All results are reported as a mean ± standard error of the mean (SEM). A p-value less than 0.05 was considered significant.
Results
Ovarian expression and distribution of α1a-AR
The mRNA expression of α1a-AR in the ovary was unaltered in the PCO group compared to the control group. Repeated low-frequency EA treatments did not affect the mRNA expression of α1a-AR in the EA or the PCO+EA groups (Figure 2A).
Figure 2 Ovarian α1a-AR mRNA and protein expression. As shown in panel A, no significant differences were found in α1a-AR mRNA expression between the control, EA, and PCO groups. Values are given as means ± SEMs normalized to GAPDH. The results of α1a-AR immunopositive cell count is shown in panel B. Values are given as means ± SEMs. Immunostaining revealed that ovarian α1a-AR protein is expressed in control group ovaries. EA treatments did not affect the number of immunopositive cells in the ovaries of control rats. PCO ovaries had significantly higher amounts of α1a-AR protein compared with control ovaries. EA treatments in PCO rats decreased α1a-AR protein immunoreactivity when compared with untreated PCO rats. ap < 0.05 vs control group. bp < 0.05 vs PCO group. Representative pictures showing the distribution of α1a-AR positive cells in the ovaries of the experimental groups are showed in Panel: C-D. Immunostained cells (arrows) were localized around blood vessels (C) and in the granulosa cells of an early antral follicle (D) and corpora lutea. F: Follicle; bv: blood vessel; gr: granulose cells; th: thecal layer. Magnification C: ×400; D: ×200.
Significantly higher number of immunopositive cells of α1a-AR was found in the PCO group compared with the control group. EA treatments prevented the increase in α1a-AR protein immunoreactivity, since the number of immunostained cells in the PCO+EA group was not different from the control group. EA treatments did not affect the number of immunopositive cells in the ovaries of control rats. Immunohistochemical analysis on serial sections showed that α1a-AR protein (Figure 2C and 2D) was expressed mainly around blood vessels and granulosa regions.
Ovarian expression and distribution of α1b-AR
The mRNA expression of α1b-AR in the ovary was significantly higher in the PCO group than in the control and EA group. The mRNA expression of α1b-AR was significantly higher in the PCO+EA group compared to the two control groups but not to the PCO group (Figure 3A).
Figure 3 Ovarian α1b-AR mRNA and protein expression. As shown in panel A, no significant differences were found in α1b-AR mRNA between the control and EA groups. A significant increase in ovarian α1b-AR mRNA was found in the PCO group when compared to controls. No differences were found between the PCO+EA group and the PCO group. Values are given as means ± SEMs normalized to GAPDH. ap < 0.05 vs control group. EA treatments did not affect the amount of α1b-AR immunopositive cells in the EA group(panel B). Significantly higher number of immunopositive cells of α1b-AR was found in the ovaries of PCO rats. EA treatments did not affect the amount or distribution of α1b-AR protein in PCO ovaries. Values are given as means ± SEM. ap < 0.05 vs control group. Representative pictures showing the distribution of α1b-AR positive cells in the ovaries of the experimental groups are showed in Panels C-D. Immunostained cells (arrows) were localized around blood vessels (C) and in the granulosa cells of mature follicles (D). F: Follicle; bv: blood vessel; gr: granulose cells; th: thecal layer. Magnification C: ×400; D: ×200.
As illustrated in Figure 3B, significantly higher number of immunopositive cells of α1b-AR was found in the ovaries of PCO and PCO+EA rats when compared to controls. EA treatments did not affect the amount or distribution of α1b-AR protein in the EA and the PCO groups. As shown in Figure 3C–D, α1b-AR protein was located around blood vessels and in the granulose region of mature follicles.
Ovarian expression and distribution of α1d-AR
The mRNA expression of α1d-AR in the ovary was significantly higher in the PCO group than in the control group. The mRNA expression of α1d-AR in the PCO+EA was not different from that in the PCO group (Figure 4A).
Figure 4 Ovarian α1d-AR mRNA and protein expression. As shown in panel A, α1d-AR mRNA expression was significantly higher in the PCO group than in control and EA group. EA significantly decreased the mRNA expression in the PCO+EA group compared with the PCO group. Values are given as means ± SEMs normalized to GAPDH. ap < 0.05 vs control group. The α1d-AR immunopositive cell number (Panel B) was not affect in the EA group when compared to control group. Significantly higher number of immunopositive cells of α1d-AR was found in the PCO group compared with the control group. EA treatment significantly decreased the number of immunopositive cells in the PCO+EA group. Values are given as means ± SEMs. ap < 0.05 vs control group. bp < 0.05 vs PCO group. Representative pictures showing ovarian distribution of α1d-AR expressing cells (pointed by arrows) are showed in Panels C-D. The α1d-AR was found expressed in the granulosa cells of healthy follicles (C) and corpora lutea (D) and around blood vessels (not shown) in all of the experimental groups. F: Follicle; gr: granulose cells; th: thecal layer. Magnification C-D: ×200.
As illustrated in Figure 4B, significantly higher number of immunopositive cells of α1d-AR was found in the PCO group compared with the control group. EA treatment significantly decreased the number of immunopositive cells in the PCO+EA group. The α1d-AR immunopositive cell number was not affected in the EA group when compared to control group. As shown in Figure 4C–D, the α1d-AR was found expressed in the granulosa cells of healthy follicles and corpora lutea, and around blood vessels (not shown) in all of the experimental groups.
Ovarian expression and distribution of β2-AR
The mRNA expression of β2-AR in the ovary of the PCO group was not changed, when compared to the control group. β2-AR mRNA was significantly lower both in the EA group and in the PCO+EA group when compared to the control group (Figure 5A).
Figure 5 Ovarian β2-AR mRNA and protein expression. The expression of β2-AR mRNA in the ovary (panel A) in the PCO group was lower compared to the control group. β2-AR mRNA was unaltered in the PCO+EA group when compared to control. Values are given as means ± SEMs normalized to GAPDH. ap < 0.05 vs control group. β2-AR immunopositive cell number (Panel B) in control ovaries was unchanged by EA treatments. No difference in number of β2-AR immunoreactive cells was found in PCO ovaries, while EA treatments in PCO rats significantly increase the amount of β2-AR immunostained cells when compared to PCO group. Values are given as means ± SEMs. bp < 0.05 vs PCO group. Representative pictures of stained cells (some of them pointed by black arrows) are showed in Panels C- D. The β2-AR was found expressed in degenerating corpora lutea (C) and follicles (D) in all of the experimental groups. Magnification C: ×400; D: ×200.
No difference in number of β2-AR immunoreactive cells was found in PCO ovaries (Figure 5B), while EA treatments in PCO rats (PCO+EA group) significantly increase the amount of β2-AR immunostained cells when compared to PCO group. β2-AR immunopositive cell number in control ovaries was unchanged by EA treatments. The β2-AR was found expressed in degenerating corpora lutea (Figure 5C) and follicles (Figure 5D) in all of the experimental groups.
Ovarian expression and distribution of p75NTR
The p75NTR mRNA expression in the ovary (Figure 6A) was unchanged in the PCO group compared to the control group. Low-frequency EA treatments did not affect ovarian p75NTR mRNA expression in the PCO+EA group compared to the PCO group, and did not differ from the control group.
Figure 6 Ovarian p75NTR mRNA and protein expression. As shown in panel A, ovarian p75NTR mRNA was found unchanged in the PCO group compared to the control and EA group. Low-frequency EA treatments did not significantly affect ovarian p75NTRmRNA expression in the EA, PCO and PCO+EA groups. Values are given as means ± SEMs normalized to GAPDH. As shown in panel B, the number of ovarian p75NTR immunopositive cells was significantly decreased in the EA group when compared to controls. The number of p75NTR-stained cells in the PCO group was significantly higher than in controls, while repeated EA treatments greatly decreased p75NTR protein immunoreactivity in the PCO+EA group. Values are given as means ± SEMs. ap < 0.05 vs control group. bp < 0.05 vs PCO group. Representative pictures of stained cells (pointed by black arrows) are showed in Panels C-D. The p75NTR was found expressed in the thecal layer of healthy follicles (C) and in the stromal region (D) in all of the experimental groups. Magnification C-D: ×200.
As shown in Figure 6B, the number of p75NTR -stained cells in the PCO group was significantly higher than in controls, while repeated EA treatments significantly decreased p75NTR protein immunoreactivity in the PCO+EA group. The number of ovarian p75NTR immunopositive cells was significantly lower in the EA group when compared to controls. As shown in Figure 6C–D, ovarian p75NTR expressing cells were distributed mainly around the follicles in the theca layers, with some immunoreactivity also spread in the ovarian stroma (Figure 6D).
Discussion
The aim of the present study was to investigate whether repeated low-frequency EA treatments modulate the expression of mRNA and the amount and distribution of proteins of α1-, and β2-ARs, and p75NTR in rats with steroid-induced PCO. The results of this study demonstrated that i.m. EV injections result in significantly higher mRNA expression of ovarian α1b- and α1d-AR in PCO rats compared to control rats. EV-induced PCO induced a significantly higher amount of immunostained cells for α1a-, α1b- and α1d proteins, that was prevented by repeated low-frequency EA treatments, except for α1b-AR. The EA treatment also induced an increase of β2 -AR protein in the EV-injected rats.
It has been suggested that high sympathetic drive to the ovary might be important in both EV-induced PCO in rats and PCOS in humans [4,5,9,10,27]. Clinical studies show that women with PCOS temporarily recover normal ovarian function after bilateral wedge resection or ovarian drilling, which partially denervates the ovary [28]. These observations suggest that the ovarian nerves might be involved in the successful outcome of bilateral wedge resection and ovarian drilling. Current pharmacological treatment using clomiphene citrate is the first-line treatment for ovulation induction in women with PCOS [29]. This is effective, but side-effects such as super ovulation are quite common [30]. There is a clear need to identify new therapeutical approaches – including non-pharmacological strategies – to reduce or replace drug intervention.
That EA may reduce hyperactivity in the ovarian peripheral sympathetic nerve fibers is consistent with the theory that EA could modulate sensory, motor, and autonomic outflow at the segmental level [16]. It has also been shown that EA activates higher control systems, resulting in the release of a number of neuropeptides that are important in the modulation of central and segmental autonomic outflow and of the HPO axis [16,31]. We have recently shown that repeated low-frequency EA induces regular ovulations in more than one-third of women with PCOS and normalizes endocrine and neuroendocrine parameters without any negative side-effects [17]. The effects of repeated low-frequency EA were then attributed to an inhibition of hyperactivity in the sympathetic nervous system [16,32]. The present study further indicates that EA is effective in preventing EV-induced dysregulation of ovarian sympathetic markers.
Increased peripheral sympathetic outflow in rats with steroid-induced PCO is evidenced by increased releases of NE, higher concentrations of NE in the ovary, and a reduced number of β2-AR in the ovarian compartment receiving catecholaminergic innervation [8,9]. The role of β2-AR in ovarian physiology and pathophysiology has been related to the regulation of ovarian steroidogenesis [8]. Transection of the superior ovarian nerve in steroid-induced PCO reduces the steroid response, raises β2-AR expression to normal levels, and restores estrus cyclicity and ovulation [8]. Thus the disturbances in steroid production – at least in the present rat PCO model – might be secondary to the elevated adrenergic control over ovarian steroidogenesis mediated by β2-AR. Interestingly, repeated treatments of low-frequency EA induced an increase of β2-AR protein in EV-injected rats, and this is in accordance with the hypothesis that EA down-regulates the activity in the sympathetic nervous system. On the other hand, the mRNA expression of β2-AR was decreased in the EA and in the PCO+EA group compared with the control group. One plausible explanation for the discrepancy between the mRNA and protein levels might be an unbalanced turn over between β2-AR mRNA and protein. Thus the lower level of mRNA, compared to those of the protein, could reflect its utilization for protein synthesis, not balanced by an appropriate mRNA replacement. Such a mechanism could also reflect different regulation levels for our treatments, that could act on protein production separately at both the gene transcription and protein synthesis. Further studies are necessary to clarify this mechanism(s).
The functional significance of the different α1-ARs in the ovary of PCO rats has not been clearly identified. The function of these types of ARs has traditionally been characterized in ovarian physiology as being primarily related to the regulation of ovarian blood flow. Interestingly, α1-agonist stimulation has recently been shown to modulate the progesterone release in cultured granulose cells by potentiation of vasoactive intestinal peptide (VIP) and Pituitary Adenylate Cyclase-Activating Polypeptide (PACAP) [33]. In a recent study, for the first time to our knowledge, we have shown that the expression of all α1-AR subtypes at both the mRNA and the protein level are up-regulated at an early (30 day) and at a late (60 day) stage after EV injection [14]. Thus it can be inferred that α1-ARs participate not only in the physiological regulation of progesterone from the normal rat ovary [33], but also most probably in the up-regulation of progesterone release described in the EV-induced PCO ovary [8]. Furthermore, the dysregulated α1-ARs can be related to high sympathetic activity in the ovaries of PCO rats [8]. Indeed, we found that repeated treatments of low-frequency EA prevents the dysregulation of α1a-AR and α1d-AR protein in rats with steroid-induced PCO that evidence the effectiveness of EA in reducing the sympathetic drive to the ovary. The mRNA expression of α1d-AR was normalized by low-frequency EA, but not that of α1a- and α1b-AR. Again, the discrepancy between the mRNA expression and protein levels might be an unbalanced turn over between α1a- and α1b-ARs mRNA and protein, thus the mRNA that is engaged in the protein translation is not replaced completely, while this is not the case for α1d-AR. These results indicate that a mechanism linked to ovarian presence and function of ARs could be active in this context.
The present results are in line with the results of a recent study demonstrating that low-frequency EA increased blood flow and decreased sympathetic activity in the ovary [22,23]. These observations led to the hypothesis that the effects of low-frequency EA on ovarian blood flow were mediated by α1-ARs [22], and this is in line with another recent study by Uchida et al. [12] which suggested that the regulation of ovarian blood flow via sensory stimulation is mediated by α1-ARs evidenced by blocking α1-ARs.
In the present study, we also demonstrate that repeated treatments of low-frequency EA maintains p75NTR mRNA and protein amount at basal levels in PCO animals. This agrees with results of our previous studies [18,20]. It is known that p75NTR aids the development of specific populations of sympathetic neurons [34], and that this receptor is responsible for the responsiveness of adult sympathetic neurons to target-derived NGF [35]. The evidence that EV injection in adult rats increases the intraovarian synthesis of both NGF and p75NTR [10] suggests a possible functional link between PCO and the NGF/NGF receptor system. Interestingly, these changes were accompanied by selective activation of noradrenergic neurons projecting to the ovary. The activation of the sympathetic nervous system after EV injection has been evidenced by enhanced TH activity in the ovaries of PCO rats [9] and by increase of TH mRNA expression in the catecholaminergic cells of the celiac ganglion selectively projecting to the ovaries [10]. Furthermore, intraovarian blockade of NGF and p75NTR resulted in decreased p75NTR synthesis by ovarian theca cells and restored estrous cyclicity and ovulatory capacity in EV-injected rats [10]. Thus, one possible mechanism underlying the effect of low-frequency EA on sympathetic tone might be decreased p75NTR-mediated sympathetic responsiveness to NGF action. That EA counteracted the EV-induced increase in ovarian expression and amount of p75NTR supports this hypothesis.
Interestingly, CRF a principal neurohormone in the control of the hypothalamus-pituitary-adrenal (HPA) axis, has been shown to be increased in both the median eminence and in the ovary in rats with steroid-induced PCO [19]. In the same study, repeated low-frequency EA restored the increased CRF concentrations indicating that peripheral CRF and the HPA axis plays a crucial role in the regulation of ovarian function in steroid-induced PCO [19]. These data, together with the present one, suggest that EA could act as a modulator of the central control over sympathetic output in rats with steroid-induced PCO. Further studies are necessary to clarify this point.
In conclusion, the present study shows that EA prevented most of the EV-induced changes in ovarian ARs. Furthermore, EA was able to counteract the EV-induced up regulation of p75NTR, probably by normalizing the sympathetic ovarian response to NGF action. Our data indicate the effectiveness of EA in the regulation of ovarian responsiveness to sympathetic inputs and depict a possible complementary therapeutic approach to preventing and/or overcoming sympathetic-related anovulation in women with PCOS.
Authors' contributions
LM participated in the design of the study, carried out part of the animal preparation, performed RT-PCR and immunohistochemical analyses, performed the statistical analysis and drafted the manuscript. TL, AH and LA participated in the design of the study and in writing the manuscript. ES-V participated in the design of the study, carried out part of the animal preparation and drafted the manuscript. All authors read and approved the final manuscript.
Acknowledgements
This study was supported by grants from Wilhelm and Martina Lundgrens's Science Fund, the Hjalmar Svensson Foundation, The Royal Society of Art and Sciences in Göteborg, Magnus Bergwalls stiftelse, the Novo Nordisk Foundation, The Göteborg Medical Society, the Medical Research Council (Project No. 12206, 2004-6399 and -6827), and the Swedish Heart Lung Foundation. The contribution of Luigi Aloe and Luigi Manni is supported by Progetti Strategici FISR/Neurobiotecnologie and by Fondazione CARISBO, Bologna, Italy.
==== Refs
Tsilchorozidou T Overton C Conway GS The pathophysiology of polycystic ovary syndrome Clin Endocrinol (Oxf) 2004 60 1 17 14678281 10.1046/j.1365-2265.2003.01842.x
Lobo RA The role of neurotransmitters and opioids in polycystic ovarian syndrome Endocrinol Metab Clin North Am 1988 17 667 683 2904366
Lobo RA Granger LR Paul WL Goebelsmann U Mishell DRJ Psychological stress and increases in urinary norepinephrine metabolites, platelet serotonin, and adrenal androgens in women with polycystic ovary syndrome Am J Obstet Gynecol 1983 145 496 503 6824043
Semenova II Adrenergic innervation of the ovaries in Stein-Leventhal syndrome Vestn Akad Med Nauk SSSR (Abstract in english) 1969 24 58 62
Heider U Pedal I Spanel-Borowski K Increase in nerve fibers and loss of mast cells in polycystic and postmenopausal ovaries Fertil Steril 2001 75 1141 1147 11384640 10.1016/S0015-0282(01)01805-2
Szukiewicz D Uilenbroek JTJ Polycystic ovary syndrome - searching for an animal model J Med 1998 29 259 275 10503163
Brawer JR Munoz M Farookhi R Development of the polycystic ovarian condition (PCO) in the estradiol valerate-treated rat Biology of Reproduction 1986 35 647 655 3098314
Barria A Leyton V Ojeda SR Lara HE Ovarian steroidal response to gonadotropins and β-adrenergic stimulation is enhanced in polycystic ovary syndrome: role of sympathetic innervation Endocrinology 1993 133 2696 2703 8243293 10.1210/en.133.6.2696
Lara HE Ferruz JL Luza S Bustamante DA Borges Y Ojeda SR Activation of ovarian sympathetic nerves in polycystic ovary syndrome Endocrinology 1993 133 2690 2695 7902268 10.1210/en.133.6.2690
Lara HE Dissen GA Leyton V Paredes A Fuenzalida H Fiedler JL Ojeda SR An increased intraovarian synthesis of nerve growth factor and its low affinity receptor is a principal component of steroid-induced polycystic ovary in the rat Endocrinology 2000 141 1059 1072 10698182 10.1210/en.141.3.1059
Civantos Calzada B Aleixandre de Artinano A α-adrenoceptor subtypes Pharmacol Res 2001 44 195 208 11529686 10.1006/phrs.2001.0857
Uchida S Hotta H Kagitani F Aikawa Y Ovarian blood flow is reflexively regulated by mechanical afferent stimulation of hindlimb in non-pregnant anesthetized rats Autonomic Neuroscience: Basic and Clinical 2003 106 91 97 10.1016/S1566-0702(03)00073-0
Baranowska B Wasilewska-Dziubinska E Radzikowska M Pllonowski A Roguski K Krawczyk E Kawalec M Effects of PACAP and VIP on adrenal progesterone release Ann N Y Acad Sci 1996 805 628 633 8993452
Manni L Holmang A Lundeberg T Aloe L Stener-Victorin E Ovarian expression of α (1)- and β (2)-adrenoceptors and p75 neurotrophin receptors in rats with steroid-induced polycystic ovaries Auton Neurosci 2005 118 79 87 15795180 10.1016/j.autneu.2005.01.004
Andersson S Lundeberg T Acupuncture - from empiricism to science: functional background to acupuncture effects in pain and disease Med Hypotheses 1995 45 271 281 8569551 10.1016/0306-9877(95)90117-5
Sato A Sato Y Uchida S Reflex modulation of visceral functions by acupuncture-like stimulation in anesthetized rats International Congress Series 2002 1238 111 123 10.1016/S0531-5131(02)00418-1
Stener-Victorin E Waldenstrom U Tagnfors U Lundeberg T Lindstedt G Janson PO Effects of electro-acupuncture on anovulation in women with polycystic ovary syndrome Acta Obstet Gynecol Scand 2000 79 180 188 10716298 10.1034/j.1600-0412.2000.079003180.x
Stener-Victorin E Lundeberg T Waldenstrom U Manni L Aloe L Gunnarsson S Janson PO Effects of electro-acupuncture on nerve growth factor and ovarian morphology in rats with experimentally induced polycystic ovaries Biol Reprod 2000 63 1497 1503 11058557
Stener-Victorin E Lundeberg T Waldenstrom U Bileviciute-Ljungar I Janson PO Effects of electro-acupuncture on corticotropin-releasing factor in rats with experimentally-induced polycystic ovaries Neuropeptides 2001 35 227 231 12030806 10.1054/npep.2002.0878
Stener-Victorin E Lundeberg T Cajander S Aloe L Manni L Waldenstrom U Janson PO Steroid-induced polycystic ovaries in rats: effect of electro-acupuncture on concentrations of endothelin-1 and nerve growth factor (NGF), and expression of NGF mRNA in the ovaries, the adrenal glands, and the central nervous system Reprod Biol Endocrinol 2003 1 33 12725645 10.1186/1477-7827-1-33
Stener-Victorin E Lindholm C Immunity and β-endorphin concentrations in hypothalamus and plasma in rats with steroid-induced polycystic ovaries: effect of low-frequency electroacupuncture Biol Reprod 2004 70 329 333 14561641 10.1095/biolreprod.103.022368
Stener-Victorin E Kobayashi R Kurosawa M Ovarian blood flow responses to electro-acupuncture stimulation at different frequencies and intensities in anaesthetized rats Autonomic Neuroscience: Basic and Clinical 2003 108 50 56 10.1016/j.autneu.2003.08.006
Stener-Victorin E Kobayashi R Watanabe O Lundeberg T Kurosawa M Effect of electro-acupuncture stimulation of different frequencies and intensities on ovarian blood flow in anaesthetised rats with steroid-induced polycystic ovaries Reprod Biol Endocrinol 2004 2 16 15046638 10.1186/1477-7827-2-16
Tirassa P Manni L Stenfors C Lundeberg T Aloe L RT-PCR ELISA method for the analysis of neurotrophin mRNA expression in brain and peripheral tissues J Biotechnol 2000 84 259 272 11164267 10.1016/S0168-1656(00)00370-9
Chomczynski P Sacchi N Single-step method of RNA isolation by acid guanidinium thiocyanate- phenol-chloroform extraction Anal Biochem 1987 162 156 159 2440339 10.1016/0003-2697(87)90021-2
Chandler CE Parsons LM Hosang M Shooter EM A monoclonal antibody modulates the interaction of nerve growth factor with PC12 cells J Biol Chem 1984 259 6882 6889 6327698
Lara HE Dorfman M Venegas M Luza SM Luna SL Mayerhofer A Guimaraes MA Rosa E Silva AAM Ramírez VD Changes in sympathetic nerve activity of the mammalian ovary during a normal estrous cycle and in polycystic ovary syndrome: Studies in norepinephrine release Microsc Res Tech 2002 59 495 502 12467025 10.1002/jemt.10229
Nakamura Y Treatment of polycystic ovary syndrome: an overview Horm Res 1990 33 31 2095359
Nestler JE Stovall D Akhter N Iuorno MJ Jakubowicz DJ Strategies for the use of insulin-sensitizing drugs to treat infertility in women with polycystic ovary syndrome Fertil Steril 2002 77 209 215 11821072 10.1016/S0015-0282(01)02963-6
Kousta E White DM Franks S Modern use of clomiphene citrate in induction of ovulation Hum Reprod Update 1997 3 359 365 9459281 10.1093/humupd/3.4.359
Han JS Acupuncture: neuropeptide release produced by electrical stimulation of different frequencies TRENDS in Neurosciences 2003 26 17 22 12495858 10.1016/S0166-2236(02)00006-1
Chen BY Acupuncture normalizes dysfunction of hypothalamic-pituitary-ovarian axis Acupunct Electrother Res 1997 22 97 108 9330669
Wasilewska-Dziubinska E Borowiec M Chmielowska M Wolinska-Witort E Baranowska B Alfa 1 adrenergic potentiation of progesterone accumulation stimulated by vasoactive intestinal peptide (VIP) and pituitary adenylate cyclase-activating polypeptide (PACAP) in cultured rat granulosa cells Neuro Endocrinol Lett 2002 23 141 148 12011800
Lee KF Bachman K Landis S Jaenisch R Dependence on p75 for innervation of some sympathetic targets Science 1994 263 1447 1449 8128229
Cowen T Gavazzi I Plasticity in adult and ageing sympathetic neurons Prog Neurobiol 1998 54 249 288 9481799 10.1016/S0301-0082(97)00071-3
| 15941472 | PMC1175857 | CC BY | 2021-01-04 16:37:13 | no | Reprod Biol Endocrinol. 2005 Jun 7; 3:21 | utf-8 | Reprod Biol Endocrinol | 2,005 | 10.1186/1477-7827-3-21 | oa_comm |
==== Front
RetrovirologyRetrovirology1742-4690BioMed Central London 1742-4690-2-381592707910.1186/1742-4690-2-38ResearchPrevalence of GB virus type C in urban Americans infected with human immunodeficiency virus type 1 Smith Stephen M [email protected] Michael J [email protected] Mahender [email protected] James P [email protected] Lavanya [email protected] Natalia [email protected] Jihad [email protected] Diana [email protected] George [email protected] Saint Michael's Medical Center, Newark New Jersey 07102, USA2 The New Jersey Medical School, Newark New Jersey 07102, USA2005 31 5 2005 2 38 38 3 5 2005 31 5 2005 Copyright © 2005 Smith et al; licensee BioMed Central Ltd.2005Smith 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.
GBV-C virus infection has been linked to improved clinical outcome in HIV-1 co-infected individuals. The epidemiology of GBV-C has, thus far, been limited to the gay male, HIV+ population. Here we describe the prevalence of antibodies against GBV-C envelope glycoprotein E2 and GBV-C viremia in an HIV+ inner city population. This study group is predominantly African-American; 41% of the participants are women. The major risk factor for HIV infection is intravenous drug use. Overall, 56% of the study population had evidence of current or past infection with GBV-C. GBV-C exposure was not associated with hepatitis C virus infection. The group of participants, who had GBV-C viremia and anti-E2 antibodies, had high percentage of patients with an undetectable HIV-1 viral load. These data provide increased insight into the prevalence of GBV-C co-infection in the HIV epidemic in this understudied population.
==== Body
Background
In 1995, several groups independently reported the discovery of two new viruses, which were termed GB virus type C (GBV-C) and hepatitis G virus, respectively (review in [1]). Subsequently, these viruses were found to be two strains of a novel RNA virus belonging to the Flaviviridae family. GBV-C (the designation used in this paper) is distantly related to hepatitis C virus (HCV) with which it shares approximately 30% amino acid homology. While HCV replicates primarily in hepatocytes, GBV-C replicates in both T- (CD4+ and CD8+) and B-lymphocytes. GBV-C is not known to cause disease in humans, but can establish chronic infection in which virus may be present in the blood. After years of infection, infected individuals may spontaneously clear GBV-C [1], although the reasons for this phenomenon are not known. In most cases, clearance of GBV-C is associated with seroconversion to the viral envelope glycoprotein, E2. Paradoxically, viremia may also persist despite the presence of anti-E2 antibodies, and clearance may occur in the absence of seroconversion. GBV-C may be transmitted through several routes, including sexual contact, exposure to contaminated blood and vertical transmission. To date, the epidemiology of GBV-C is incompletely understood.
Of interest, GBV-C infection appears to alter the course of human immunodeficiency virus type 1 (HIV-1) infection. Following an initial report in 1998 [2], several studies have shown that individuals, who are co-infected with GBV-C and HIV-1, have lower levels of HIV-1 viremia and higher CD4+ T cell counts than those infected with HIV-1 alone [3-8]. However, other studies have not supported this association [9-13]. A recent report failed to find evidence that active GBV-C co-infection improved survival 12 to 18 months after HIV-1 seroconversion [6]. Survival rates in persons with persistent GBV-C viremia were, however, significantly better 5 to 6 years after HIV-1 infection.
GBV-C prevalence is known to be significantly higher in HIV-1 seropositive individuals (>75%) [3,5,6,13] compared with healthy blood donors (10–20%) [14]. In most cases, this observation is based on evaluation of patient groups comprised primarily of men, who have sex with men (MSM). The epidemiology of GBV-C among HIV-1 seropositive, inner city residents, whose risk factors, ethnicity and gender are distinct, is not known. In the present study, we evaluated the prevalence of GBV-C infection in a population consisting primarily of HIV-infected, urban African-Americans.
Methods
Study Population
The study population consisted of 353 HIV-1-infected patients who regularly attended a large urban HIV-1 clinic. The patients were recruited over a 3-month period between February and April 2004. The study was approved by the institutional review board of Saint Michael's Medical Center and informed consent was obtained from all participants prior to sample collection. Blood samples were obtained for analysis of GBV-C RNA and anti-E2 antibodies, and for measurement of HIV-1 plasma RNA levels, CD4+ T-cell counts and HCV serology. Treatment was independently determined by the treating physician.
Laboratory Assays
Studies for HIV RNA levels, HIV antibodies, and HCV antibodies were performed by commercial laboratories.
RT-PCR for GBV-C RNA
Total RNA was extracted from 100 μl of serum using an RNAeasy Mini Kit (Qiagen, Valencia, CA). Twenty-five percent of the isolated RNA was used for reverse transcription (RT) and first round PCR. RT-PCR was performed in a single tube using the AccessQuick RT-PCR System (Promega, Madison, WI). Both first- and second-round PCR were carried out using primers that hybridize to 5' non-translated regions of an infectious GBV-C clone (GenBank accession no. AF121950, nt 54 to 389). Primers for the first-round RT-PCR were GBVF1 5'-CCGACGCCTATCTAAGTA GACGC and GBVR1 5'-TCAACTCGCCGGATAAACCTATTGG. Primers for the second-round PCR were GBVF2 5'-GTGACAGGGTTGGTAGG and GBVR2 5'-GACATTGAAGGGCGACGTGG. PCR products were detected on 1.5% agarose gels containing 0.5 μg/ml ethidium bromide. The expected band sizes were 336 and 231 bp for the first- and second-round PCR, respectively. Known GBV-C positive serum (generously provided by Dr. J. Stapleton, University of Iowa, Iowa City, IA) and negative (saline) controls were included in each assay. Samples yielding ambiguous PCR results were re-tested using freshly extracted RNA from the original sera. A reaction was considered positive if either the first- or second-round PCR produced a band of the expected size. The assay was validated using in vitro transcribed GBV-C RNA together with positive and negative control samples.
Detection of Antibodies against GBV-C glycoprotein E2
Antibodies against GBV-C envelope glycoprotein E2 were detected using the μPlate anti-HGenv ELISA test (Roche Diagnostics, Indianapolis, IN) according to the manufacturer's protocol, which is summarized below. A 1:20 dilution of each serum sample was added to an incubation solution containing HGV-E2 antigen-bound, biotinylated anti-E2 antibodies. This solution was then added to a streptavidin-coated microwell plate. After the plate was washed once with the wash-solution, POD-labeled, anti-human Fcγ antibodies were added to the plate, followed by ABTS® substrate solution. Absorbance was read at 405 nm. A sample was considered positive if the A405 ≥ the cut-off value calculated according to the manufacturer's protocol (0.2 times the sum of the positive and negative controls). Samples falling within +/-15% of the cut-off value were repeated using freshly diluted sera.
Statistical Analyses
A chi-square or Fisher's exact test was used to analyze categorical variables. The group means were compared by either the Student's t-test, Mann-Whitney U test or Wilcoxon rank sum test. p values <0.05 were considered to indicate statistical significance and all reported p values were two-sided.
Results
GBV-C prevalence
Of the 353 subjects studied here, 208 (59%) were men and 145 (41%) were women. The mean age within the cohort was 46.4 years, and the mean CD4+ T-cell count was 416 cells/mm3. Plasma from each patient was tested for the presence of GBV-C RNA and anti-glycoprotein E2 antibodies. GBV-C RNA was detected in 23.2% (82/353) of subjects. Among those testing positive for GBV-C RNA, 13 (16%) were also found to be positive for anti-E2 antibodies. Within the study population, 32.3% (114/353) tested positive for anti-E2 antibodies alone, while the remaining subjects (157/353) tested negative for both GBV-C RNA and anti-E2 antibodies. Overall, 56% of the study population had evidence of GBV-C exposure, defined as a positive result for either GBV-C RNA or anti-E2 antibodies.
Among the study subjects, the main HIV-1 risk factors were intravenous drug use (IDU) (53.8%) and heterosexual contact with a person who used intravenous drugs (32.6%) (Table 1). The majority of the enrolled population (71.4%) was African-American. There was no significant difference in GBV-C status between sex, race or HIV-1 risk factor (Table 2).
Table 1 HIV-1 risk factors according to GBV-C status.
GBV-C RNA positive Anti- E2 antibody positive alone Negative for GBV-C RNA and Anti-E2 antibody Total
Hemophilia n (%)* 1 (11%) 2 (22%) 6 (67%) 9 (100%)
Heterosexual 25 (22%) 38 (33%) 52 (45%) 115 (100%)
Heterosexual/IVDU 10 (24%) 10 (24%) 12 (52%) 42 (100%)
IVDU 35 (22%) 58 (37%) 65 (41%) 158 (100%)
MSM 8 (35%) 2 (9%) 13 (66%) 23 (100%)
Transfusion 2 (20%) 3 (30%) 5 (50%) 10 (100%)
Other 1 (17%) 1 (17%) 4 (67%) 6 (100%)
*- percentage by risk factor
Table 2 Demographics of the study population.
GBV-C RNA positive Anti- E2 positive Unexposed to GBV-C All subjects
Total 82 114 157 353
Age* 44.7 ± 7.9 47.0 ± 8.9 46.8 ± 9.5 46.4 ± 9.0
Sex n (%)
Male 52 (63%) 65 (57%) 91 (58%) 208 (59%)
Female 30 (37%) 49 (43%) 66 (42%) 145 (41%)
Race n (%)
Black 65 (79%) 83 (73%) 104 (66%) 252 (71%)
Hispanic 6 (7%) 18 (16%) 31 (20%) 55 (16%)
Caucasian 11 (13%) 13 (11%) 22 (14%) 46 (13%)
* -average ± S.D.
Although GBV-C is thought to be transmitted by similar routes as HCV, HCV antibody status was not strongly associated with GBV-C exposure (Table 3). A total of 163 subjects were positive for HCV antibodies. Among those, 61% (99/163) were also positive for either GBV-C viremia or anti-E2 antibody, while 51% (97/190) of the HCV antibody negative population had evidence of GBV-C exposure.
Table 3 The presence of hepatitis C virus antibody according to GBV-C status.
HCV Antibody Status GBV- C RNA positive n (%) Anti-E2 antibody positive n (%) Negative for GBV-C RNA and Anti-E2 antibody n (%)
HCV (+) 39 (48%) 60 (53%) 64 (41%)
HCV (-) 43 (52%) 54 (47%) 93 (59%)
Total 82 (100%) 114 (100%) 157 (100%)
Implications of GBV-C viremia
Although this study represents a cross-sectional analysis, we evaluated the relationship between GBV-C infection status and HIV-1 viremia or CD4+ T cell count. GBV-C exposure was associated with a larger percentage of patients with CD4+ T cell counts >350 cells/mm3 (p < 0.05, Table 4). In addition, 46.3% of patients with GBV-C viremia had plasma HIV-1 RNA levels <500 copies/ml, compared with 36.9% of patients who tested negative for GBV-C RNA and anti-E2 antibodies. This trend, while suggestive of an effect of GBV-C co-infection on HIV-1, was not statistically significant. However, a separate analysis of the 13 patients who tested positive for both GBV-C RNA and anti-E2 antibodies revealed a significant association between GBV-C infection and low plasma HIV-1 RNA levels (p < 0.05, Table 5). Overall, for the entire cohort, GBV-C status had no effect on the mean plasma HIV-1 RNA levels or the mean CD4+ T-cell count.
Table 4 Patient stratification by CD4+ T-cell count, according GBV-C exposure status.
Exposed to GBV-C n (%) Negative for GBV-C RNA and Anti-E2 antibody n (%)
CD4+ T-cell count
≤ 350 cells/mm3 81 (41%) 82 (52%)
>350 cells/mm3 115 (59%) † 75 (48%)
Total 196 (100%) 157 (100%)
† The percentage (115/196; 59%) of GBV-C exposed subjects with a CD4+ T-cell count >350 cells/mm3 was higher than that (75/157; 48%) of those unexposed to GBV-C (p < 0.05; Chi-square test).
Table 5 Plasma HIV-1 RNA levels according to GBV-C status.
HIV RNA copies/mL GBV-C RNA positive/ anti-E2 antibody positive n (%) GBV-C RNA positive n (%) Anti-E2 antibody positive n (%) Negative for GBV-C RNA and Anti-E2 antibody n (%)
≤ 500 10 (77%) † 28 (41%) 46 (40%) 58 (37%)
>500 3 (23%) 41 (59%) 68 (60%) 99 (63%)
Total 13 (100%) 69 (100%) 114 (100%) 157 (100%)
p < 0.05 when comparing RNA+/E2+ group to other GBV-C status groups.
† Comparison among the four groups were made using the Chi-square test.
Conclusion
This report is the first to describe the prevalence of GBV-C in an inner city population comprised predominantly of HIV-1-infected African-Americans. Previous studies, which focused on HIV-seropositive male homosexuals, found evidence of prior or current GBV-C infection in 74 to 85% of subjects [3,5,6,13]. In contrast, only 56% of the population studied here tested positive for GBV-C RNA and/or anti-E2 antibodies. Analysis of HIV-1 risk factors did not reveal a significant correlation between specific high-risk behavior for HIV-1 and GBV-C exposure.
Our findings comprise data from the largest such cohort studied to date. Within our study population, we found a 61% rate of GBV-C exposure among 163 HCV antibody positive patients; a rate not significantly different from that of HCV antibody negative patients (51%). Taken together with published findings, these data suggest that GBV-C may be transmitted more efficiently by male homosexual contact than by either intravenous drug use or heterosexual contact. Alternatively, infection with GBV-C may have occurred at a higher rate in our population, but either GBV-C viremia or anti-E2 antibody responses may have waned over time in some patients. Pre-exposure to GBV-C may have rendered individuals within the population immune to secondary infection with this flavivirus. Within this cohort, we also studied a large population of HIV-infected women (n = 145) and found a similar rate of GBV-C exposure (55%) to that found in men (57%).
Although the effect of GBV-C on HIV-1 infection remains controversial, most studies have shown higher CD4+ cell counts and lower HIV-1 viral loads in GBV-C viremic patients. Our study suggests a similar effect among a population with distinct characteristics including ethnicity, transmission profiles and gender than those previously reported. In our cohort, 77% of patients who tested positive for both GBV-C RNA and anti-E2 antibodies had plasma HIV-1 RNA levels <500 copies/ml. While these data represent a small number of patients, the percentage with low HIV-1 viral load was statistically higher among these patients when compared to those in the other groups. The co-existence of GBV-C viremia and anti-E2 antibody may be a marker of long-term GBV-C infection, which has recently been shown to correlate with a better outcome in HIV-1-infected individuals [6].
Our study was limited in several aspects, including a lack of longitudinal follow-up. In addition, we were unable to compile adequate data on the duration of HIV-1 infection and the possible impact of antiretroviral therapy in these patients. The latter restriction may be balanced by the fact that nearly all HIV-infected patients in this large cohort received antiretroviral therapy under a relatively consistent standard practiced within a single clinical environment.
Further longitudinal studies will be necessary in HIV-1-infected patients to clarify the potential effects of GBV-C co-infection. Our data support the hypothesis that GBV-C viremic patients with HIV-1 respond better to therapy, which has been suggested by another study [15]. This possibility needs to be tested prospectively. Our data also suggest that the behaviors associated with HIV-1 transmission in the inner city are less associated with GBV-C exposure than in other high-risk settings.
==== Refs
Stapleton JT Williams CF Xiang J GB virus type C: a beneficial infection? J Clin Microbiol 2004 42 3915 3919 15364968 10.1128/JCM.42.9.3915-3919.2004
Toyoda H Fukuda Y Hayakawa T Takamatsu J Saito H Effect of GB virus C/hepatitis G virus coinfection on the course of HIV infection in hemophilia patients in Japan J Acquir Immune Defic Syndr Hum Retrovirol 1998 17 209 213 9495219
Heringlake S Ockenga J Tillmann HL Trautwein C Meissner D Stoll M Hunt J Jou C Solomon N Schmidt RE Manns MP GB virus C/hepatitis G virus infection: a favorable prognostic factor in human immunodeficiency virus-infected patients? J Infect Dis 1998 177 1723 1726 9607857
Lefrere JJ Roudot-Thoraval F Morand-Joubert L Petit JC Lerable J Thauvin M Mariotti M Carriage of GB virus C/hepatitis G virus RNA is associated with a slower immunologic, virologic, and clinical progression of human immunodeficiency virus disease in coinfected persons J Infect Dis 1999 179 783 789 10068572 10.1086/314671
Tillmann HL Heiken H Knapik-Botor A Heringlake S Ockenga J Wilber JC Goergen B Detmer J McMorrow M Stoll M Schmidt RE Manns MP Infection with GB virus C and reduced mortality among HIV-infected patients N Engl J Med 2001 345 715 724 11547740 10.1056/NEJMoa010398
Williams CF Klinzman D Yamashita TE Xiang J Polgreen PM Rinaldo C Liu C Phair J Margolick JB Zdunek D Hess G Stapleton JT Persistent GB virus C infection and survival in HIV-infected men N Engl J Med 2004 350 981 990 14999110 10.1056/NEJMoa030107
Xiang J Wunschmann S Diekema DJ Klinzman D Patrick KD George SL Stapleton JT Effect of coinfection with GB virus C on survival among patients with HIV infection N Engl J Med 2001 345 707 714 11547739 10.1056/NEJMoa003364
Yeo AE Matsumoto A Hisada M Shih JW Alter HJ Goedert JJ Effect of hepatitis G virus infection on progression of HIV infection in patients with hemophilia. Multicenter Hemophilia Cohort Study Ann Intern Med 2000 132 959 963 10858179
Birk M Lindback S Lidman C No influence of GB virus C replication on the prognosis in a cohort of HIV-1-infected patients Aids 2002 16 2482 2485 12461426 10.1097/00002030-200212060-00017
Bjorkman P Flamholc L Naucler A Molnegren V Wallmark E Widell A GB virus C during the natural course of HIV-1 infection: viremia at diagnosis does not predict mortality Aids 2004 18 877 886 15060435 10.1097/00002030-200404090-00005
Quiros-Roldan E Maroto MC Torti C Moretti F Casari S Pan A Carosi G No evidence of benefical effect of GB virus type C infection on the course of HIV infection Aids 2002 16 1430 1431 12131224 10.1097/00002030-200207050-00019
Sabin CA Devereux H Kinson Z Griffioen A Brown D Dusheiko G Lee CA Effect of coinfection with hepatitis G virus on HIV disease progression in hemophilic men J Acquir Immune Defic Syndr Hum Retrovirol 1998 19 546 548 9859971
Van der Bij AK Kloosterboer N Prins M Boeser-Nunnink B Geskus RB Lange JM Coutinho RA Schuitemaker H GB Virus C Coinfection and HIV-1 Disease Progression: The Amsterdam Cohort Study J Infect Dis 2005 191 678 685 15688280 10.1086/427559
Stapleton JT GB virus type C/Hepatitis G virus Semin Liver Dis 2003 23 137 148 12800067 10.1055/s-2003-39943
Rodriguez B Woolley I Lederman MM Zdunek D Hess G Valdez H Effect of GB virus C coinfection on response to antiretroviral treatment in human immunodeficiency virus-infected patients J Infect Dis 2003 187 504 507 12552436 10.1086/368206
| 15927079 | PMC1175858 | CC BY | 2021-01-04 16:36:41 | no | Retrovirology. 2005 May 31; 2:38 | utf-8 | Retrovirology | 2,005 | 10.1186/1742-4690-2-38 | oa_comm |
==== Front
J Transl MedJournal of Translational Medicine1479-5876BioMed Central London 1479-5876-3-231591671510.1186/1479-5876-3-23ResearchAnalysis of memory T lymphocyte activity following stimulation with overlapping HLA-A*2402, A*0101 and Cw*0402 restricted CMV pp65 peptides Ghei Monica [email protected] David F [email protected] Maurizio [email protected] Washington University School of Medicine, St. Louis, MO, USA2 Molecular Immunology Sections, Department of Transfusion Medicine, National Institutes of Health, Bethesda, MD, USA3 Institut für chirurgische Forschung und Spitalmanagement, University of Basel, Switzerland2005 26 5 2005 3 23 23 2 2 2005 26 5 2005 Copyright © 2005 Ghei et al; licensee BioMed Central Ltd.2005Ghei 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 continuous efforts aimed at the identification of new immune epitopes across the MHC system has led to the discovery that more than one peptide may be restricted to the same HLA antigen and function as an immune determinant for that association. The aim of this study was to compare the ability of two overlapping peptides, the nonamer (9-mer) cytomegalovirus (CMV) pp65341–349 (QYDPVAALF) and the decamer (10-mer) CMV pp65341–350 (QYDPVAALFF), and the esadecamer (16-mer) peptide containing both the 9-mer and 10-mer sequences, CMV pp65340–355 (RQYDPVAALFFFDIDL), to stimulate and maintain over time a T cell immune reactivation by HLA-A*2402, A*0101, and Cw*0402 cells from CMV-seropositive subjects. The 9-mer, 10-mer, and 16-mer peptides effectively stimulated CTLs from HLA-A*2402, HLA-A*0101, and HLA-Cw*0402 CMV seropositive donors. This data confirms that both the 9-mer and the 10-mer peptides are promiscuous and are not restricted to a single HLA antigen. CMV pp65341–349 and CMV pp65341–350 have the ability to produce CMV-specific CTLs in subjects with several different HLA types, presenting a practical advantage over other peptides that are restricted only to a single HLA antigen, and thus being optimal for CMV adoptive immune therapy. Moreover, since the 16-mer peptide encompasses both the 9-mer and 10-mer peptides, it may be better than either of these peptides for CMV adoptive immune therapy.
==== Body
Introduction
In healthy subjects, primary infection with Cytomegalovirus (CMV) is usually mild or asymptomatic and is effectively controlled by the cell-mediated immune response [1]. However, in immune compromised individuals, such as those with AIDS or after bone marrow transplantation, CMV reactivation is associated with significant morbidity until the individual's immune system is completely reconstituted [2]. One method of preventing post-transplant CMV infection is adoptive immunotherapy using CMV-specific cytotoxic T cells (CTLs) from the transplant donor [3]. Several HLA class I restricted peptides derived from the immune dominant CMV 65 kd matrix phosphoprotein (pp65) have been shown to produce CMV-specific CTLs. Two overlapping HLA-A*2402 restricted peptides have been described: pp65341–349 and pp65341–350. These peptides are a nonamer (9-mer) and a decamer (10-mer) that overlap except for the last amino acid phenylalanine (F) at the C-terminus [QYDPVAALF(F)]. Despite their similarity, the ability of these peptides to induce a T cell response has been reported to differ [4,5].
Although it has been generally accepted that a unique CMV peptide is bound and presented by each separate HLA class I molecule, recent studies suggest that certain peptides are more promiscuous and may be presented by more than one HLA class I antigen. In this specific case, the 9-mer pp65341–349 has been shown to stimulate CTLs from both HLA-A*2402 and Cw*0402 donors [6], while the 10-mer pp65341–350 has been shown to be reactive with both HLA-A*2402 and A*0101 donors [7].
The current investigation sought to compare the potency of these two peptides and determine the optimum peptide size for effective CMV adoptive immune therapy. Both peptides were tested for their ability to stimulate CMV-specific CTLs in HLA-A*2402, HLA-A*0101, and HLA-Cw*0402 restriction. In addition, a 16-mer pp65340–355 that includes both the 9-mer and the 10-mer peptides was tested for its ability to reactivate memory T cells. This specific 16-mer peptide was selected since it represents the naturally processed peptide that would encompass both the 9-mer and 10-mer peptides. IFN-γ mRNA transcript production was measured by in vitro cell culture assays in which the cells were peptide-induced for 3 hours after a 2-week in vitro sensitization, while IFN-γ protein release was measured using in vitro cell culture supernatant collected at different time points.
The goal of the investigation was to determine whether both the 9-mer and the 10-mer peptides maintain high levels of CTL stimulation over time for all HLA restrictions studied. Moreover, it was important to investigate whether stimulation with the naturally processed 16-mer peptide, followed by re-stimulation by the two smaller peptides embedded within the larger sequence, lead to effective T cell memory immune response.
Materials and methods
Peptide selection, synthesis and nomenclature
Two overlapping peptides derived from the immune dominant CMV 65 kd matrix phosphoprotein (pp65), the nonamer pp65341–349 (QYDPVAALF) and the decamer pp65341–350 (QYDPVAALFF) peptides were used to analyze memory T lymphocyte activity in PBMCs collected from CMV seropositive donors bearing alleles HLA-A*2402, A*0101, or Cw*0402. The esadecamer pp65340–355 (RQYDPVAALFFFDIDL) sequence that encompasses both pp65341–349 (QYDPVAALF) and pp65341–350 (QYDPVAALFF) peptides was selected according to its score, as analyzed by MAPPP (MHC Antigen Peptide Processing Prediction) [8] based on FRAGPREDICT developed by Holzhütter [9] (Figure 1). The original sequence of 20 amino acids in length (CMV pp65340–359), as shown in figure 1, was reduced to 16 amino acids (CMV pp65340–355) in order to allow the complete reconstitution of this sequence. The C-terminus of the peptide was reduced to form a 16-mer sequence and the selection of the 16-mer did not affect the aim of having both 9-mer and 10-mer peptides represented in the sequence. The HLA restriction for this 16 amino acid sequence was not known (Additional File 1).
Figure 1 Scheme of candidate immune regions within the CMV pp65 as analyzed by MAPP. The immune dominant CMV 65 kd matrix phosphoprotein (pp65) is shown above. The highlighted red sequences within the pp65 represent the most immunogeneic polypeptides, as predicted by the MAPPP algorithm (threshold ≤ 0.5). Their positions in the protein are listed below. The two highlighted green peptides, NLVPMVATV and QYDPVAALF(F) are, respectively, the known HLA-A*0201 associated pp65495–503 and the two HLA-A*2402 associated overlapping peptides pp65341–349 and pp65341–350. The table refers to each single polypeptide's degree of cleavage by proteasome activity at either N-terminus or C-terminus. The number of peptides derived from each sequence's proteolitic activity and the corresponding degree of peptide cleavage are also shown. HRC (Highest residue cleavage probability); HFC (Highest fragment cleavage probability).
The three peptides were synthesized by Princeton Biomolecules (Langhorne, PA) with purity from 90% to 100% as analyzed by High Performance Liquid Chromatography (HPLC), dissolved at 100 μg/ml in 50% DMSO and stored at 4°C. To simplify the peptide nomenclature in this paper we refer to peptide CMV pp65341–349 (QYDPVAALF) as the 9-mer peptide, to peptide CMV pp65341–350 (QYDPVAALFF) as the 10-mer peptide, and to sequence CMV pp65340–355 (RQYDPVAALFFFDIDL) as the 16-mer peptide.
Donor collection and cell preparation
Leukocytes were collected from CMV seropositive donors bearing alleles HLA-A*2402, A*0101, or Cw*0402 after obtaining informed consent. The presence of CMV antibodies was analyzed by passive latex agglutination (CMVSCAN kit, Becton Dickinson Microbiology System, Cockeysville, MD). MHC class I genotypes were determined by sequence-specific primer polymerase chain reaction (PCR). Lymphapheresis was performed using a CS3000 Plus blood cell separator (Fenwal Divison, Baxter Health Care, Deerfield, IL), and PBMCs were isolated from the apheresis product by Ficoll (Pharmacia Biotech, Wilkstrom, Sweden) density gradient centrifugation.
PBMCs in vitro sensitization
The in vitro sensitization involved a 2-week cell culture in the presence of peptide and IL-2. Briefly, a 2-week in vitro sensitization was performed using PBMCs from CMV-seropositive donors. PBMCs from each donor were plated at a concentration of 3 × 106 cells/ml in a 24 well/plate with 2 ml RPMI complete medium (10% AB human serum, penicillin, gentamycin, glutamine, and 1% HEPES), and directly stimulated with 3 μg/ml of both test and control peptides (1 μg/ml peptides for each 106 cells). Recombinant human interleukin-2 (rhIL-2, 100 U/ml, PeproTech, Rocky Hill, NJ) was added every other day to the cell culture. At day 15, each batch of cells was washed and directly re-stimulated in fresh medium with either test or control peptide or not re-stimulated. Three hours after re-stimulation cells were harvested to analyze their IFN-γ mRNA transcript production by quantitative real time PCR (qRT-PCR). At three different time points (24, 48, and 72 hours) after re-stimulation supernatants were collected and tested for IFN-γ protein release following the ELISA manufacturer's guidelines.
Quantitative real time PCR (qRT-PCR)
IFN-γ mRNA transcript production by in vitro sensitized PBMCs was evaluated 3 hours after direct peptide re-induction, as previously described [5]. Following 2-week in vitro sensitization (IVS), 2 × 105 PBMCs (final concentration of 1 × 106 cells/ml) were plated in a 96 U-bottom well/plate with 200 μl of RPMI complete medium, incubated overnight, and then directly stimulated with 1 μg/ml of specific peptide. After a 3-hour incubation, total RNA was extracted (RNeasy Mini Kit, Qiagen, Valencia, CA) and 1 μl of synthesized cDNA (Invitrogen, Carlsbad, CA) was used as a template to measure IFN-γ mRNA transcription by qRT-PCR using an ABI Prism 7900 Sequence Detection System (Perkin Elmer, Foster City, CA). Quantitative real time PCR results were reported as the number of IFN-γ gene copies normalized by 105 β-actin gene copies. All PCR assays were performed in triplicate and reported as the average. Stimulation index has been applied based on the negative control values.
ELISA
The release of IFN-γ protein by in vitro sensitized PBMCs after peptide re-stimulation was measured using an enzyme-linked immunosorbent assay (ELISA) kit (Endogen, Woburn, MA). Supernatants of peptide-stimulated PBMCs were collected at 24, 48, and 72 hours. ELISA results were extrapolated from a standard curve generated by linear regression. The assays were performed in duplicate and reported as the average.
Results
IFN-γ mRNA transcript production by 9-mer, 10-mer, and 16-mer sensitized PBMCs from HLA-A*2402, A*0101, and Cw*0402 donors re-stimulated with their cognate determinants
In order to analyze the immediate T cell immune reactivation to each of the three determinants (9-mer, 10-mer and 16-mer) after 2-week in vitro sensitization, qRT-PCR was employed to quantitate levels of IFN-γ mRNA produced by sensitized cells after a 3-hour peptide-induction. Thus, following the 2-week peptide (9-mer, 10-mer, and 16-mer) in vitro sensitization, each batch of sensitized cells was re-stimulated with the cognate peptides. Although both 9-mer and 10-mer peptides were able to maintain high levels of stimulation over this time period for all HLA restrictions tested, the 9-mer peptide induced the highest responses in cells expressing HLA-A*2402 (S.I. 4.07–528) or HLA-Cw*0402 (S.I. 4.15–483) while the 10-mer peptide induced the highest responses in cells expressing HLA-A*2402 (S.I. 3.5–528) or HLA-A*0101 (S.I. 8.25–615). The 16-mer peptide was also able to stimulate T cells from all HLA-A*2402, A*0101 and Cw*0402 donors (S.I. 6.95, 4.96, 5.02) at levels that were approximately equal to the average of those induced by each single 9-mer and 10-mer induction (Figure 2A).
Figure 2 IFN-γ mRNA transcript production by qRT-PCR. The ability of the three CMV peptides; pp65341–349, pp65341–350, and pp65340–355 to reactivate an immune T lymphocyte response after their cognate re-stimulation of self-induced PBMCs (A) or after restimulation with each peptide of 16-mer induced PBMCs (B) was analyzed by qRT-PCR. Levels of IFN-γ mRNA transcript were quantitated after a 3-hour peptide induction. The results were performed in triplicate and reported as the average. Levels of IFN-γ mRNA transcript were normalized by levels of β-actin mRNA transcript and the final values were indicated as the ratio over the negative control.
IFN-γ mRNA transcript production by 16-mer peptide-sensitized PBMCs from HLA-A*2402, A*0101, and Cw*0402 donors re-stimulated with either the 9-mer or the 10-mer determinant
In order to analyze the immediate reactivation of memory T lymphocytes following the 16-mer peptide sensitization, the 2-week in vitro 16-mer peptide sensitized cells were re-stimulated with either 9-mer or 10-mer peptide. Thus, qRT-PCR was employed to quantitate levels of IFN-γ mRNA produced by sensitized cells after a 3-hour peptide-induction. Compared to the previous results, as in figure 2A, that demonstrated the 9-mer peptide's specificity for HLA-A*2402 and HLA-Cw*0402, the re-stimulation of 2-week in vitro 16-mer peptide sensitized cells with the 9-mer peptide confirmed the 9-mer peptide's specificity for HLA-A*2402 and HLA-Cw*0402 plus its ability to enhance the T lymphocytes reactivity in donors bearing HLA-A*0101 (S.I. 3.96–507). Similarly, the re-stimulation of 2-week in vitro 16-mer peptide sensitized cells with the 10-mer peptide confirmed this peptide's specificity for HLA-A*0101 while specifically enhancing the T cell reactivity in HLA-A*2402 donors (S.I. 6.12–236). The HLA-Cw*0402 specificity was also confirmed and maintained for the 10-mer peptides (3.58–467), probably adding a new HLA association to this CMV peptide (Figure 2B).
IFN-γ protein release by 9-mer, 10-mer, and 16-mer peptide-sensitized PBMCs from HLA-A*2402, A*0101, and Cw*0402 donors re-stimulated with their cognate determinants
Following 2-week in vitro sensitization with 9-mer, 10-mer or 16-mer peptides of PBMCs from HLA-A*2402, A*0101, and Cw*0402 donors, supernatants were collected from cell cultures at 24, 48 and 72 hours after the re-stimulation of each sensitized batch of cells with the cognate peptides. Despite the fact that both the 9-mer and 10-mer peptides were able to reactivate cells from each of the three donors at time 0 (Figure 2A), the 9-mer peptide was weaker than the 10-mer peptide in maintaining a consistent immune T response over time in both the HLA-A*2402 and HLA-A*0101 donors. In contrast, the 9-mer peptide was stronger than the 10-mer in stimulating and maintaining a T immune reactivation and response over time in donors bearing HLA-Cw*0402. The stimulation with the 16-mer peptide induced and maintained consistent immune T cell reactivation in all donors tested, as assessed by levels of IFN-γ protein production when compared to the positive control (Figure 3).
Figure 3 IFN-γ protein release by ELISA following re-stimulation of peptide (9-mer, 10-mer, and 16-mer) sensitized PBMCs from all donors with their cognate determinants. Supernatants of peptide (9-mer pp65341–349, 10-mer pp65341–350, and 16-mer pp65340–355) sensitized PBMCs from CMV seropositive donors HLA-A*2402, A*0101, and Cw*0402 were collected at three time point of 24, 48, and 72 hours after cell re-stimulation with each cognate peptide: 9-mer (blue segment), 10-mer (purple segment), and 16-mer (orange segment). The figure shows the peptide induction and the maintenance over time of T lymphocyte reactivation after cognate peptide recall compared to the positive control (green segment) for each individual HLA association: HLA-A24 (top), HLA-A1 (middle), HLA-Cw4 (bottom).
IFN-γ protein release by 16-mer sensitized PBMCs from HLA-A*2402, A*0101, and Cw*0402 donors re-stimulated with either 9-mer or 10-mer determinant
Following 2-week in vitro sensitization of PBMCs from HLA-A*2402, A*0101, and Cw*0402 donors with the 16-mer sequence, supernatants were collected from cell culture at 24, 48 and 72 hours after re-stimulation of 16-mer peptide-sensitized cells with either 9-mer or 10-mer peptides. It seems that in all donors the re-stimulation of 16-mer in vitro sensitized cells with the 9-mer and the 10-mer peptides was able to better induce IFN-γ protein production than 9-mer or 10-mer in vitro sensitized cells that were stimulated with the cognate peptide either by enhancing cytokine protein production at each time point or by better maintaining cytokine protein release over time. Specifically, this was seen in both HLA-A*2402 and HLA-A*0101 donors following 9-mer peptide induction and in the HLA-Cw*0402 donor following 10-mer peptide induction (Figure 4).
Figure 4 IFN-γ protein release ELISA following re-stimulated of 16-mer sensitized PBMCs from all donors with either the 9-mer or 10-mer determinant. Supernatants of 16-mer pp65340–355-sensitized PBMCs from CMV seropositive donors HLA-A*2402, A*0101, and Cw*0402 were collected at the three time points of 24 hours (black), 48 hours (orange), and 72 hours (yellow) after cell re-stimulation with either the 9-mer pp65341–349 or 10-mer pp65341–350 peptide. The figure shows the peptide induction and maintenance over time of the T lymphocyte reactivation from 16-mer pp65340–355-sensitized PBMCs after either 9-mer pp65 341–349 (9/16) or 10-mer pp65 341–350 (10/16) peptide recall compared to the specific 16-mer pp65340–355 re-induction (16) for each individual HLA association: HLA-A24 (top), HLA-A1 (middle), HLA-Cw4 (bottom).
Discussion
The adoptive transfer of immunodominant T lymphocytes to CMV-infected transplanted patients represents one of the treatments of choice to clear the CMV disease and to quickly reconstitute the lost balance between the CMV infection and the immune system [10-12]. The use of specific immune peptides to detect and expand immunocompetent T cells is an important tool that has been already applied in several clinical trials [13,14]. Once CMV epitope mapping had been initiated and led to the identification of epitopes encompassing several HLA class I antigens, great effots were devoted to the identification of HLA cross-reactivity of known immune determinants and the identification of new determinants. In this specific case, the two overlapping peptides pp65341–349 (QYDPVAALF) and pp65341–350 (QYDPVAALFF) have been shown to have a marked cross-reactivity. First, it was shown that pp65341–349 is restricted to HLA-A*2402 and Cw*0402 and pp65341–350 to HLA-A*2402 and A*0101 [6,7]. In this study a new HLA specificity for the 9-mer to HLA-A*0101 and for the 10-mer to HLA-Cw*0402 is well described. Furthermore, this study showed that both peptides were able to effectively stimulate immune T cell responses from HLA-A*2402, A*0101 and Cw*0402 donors, confirming that these peptides are promiscuous and not restricted to a single HLA type. While both the 9-mer and 10-mer peptides are able to maintain high levels of stimulation over time in HLA-A*2402 donors, both are also able to induce and maintain an immune reactivation in donors bearing their other noted HLA associations. In fact, the 9-mer is better able to maintain stimulation in HLA-Cw*0402 donors and the 10-mer is better able to do so in HLA-A*0101 donors.
Interestingly, another new finding emerged: both the 9-mer and 10-mer peptides were able to induce a population of restricted T lymphocytes from a cell population that had been in vitro sensitized with a peptide sequence that encompassed the two peptides. The 16-mer peptide is able to effectively stimulate T cells from HLA-A*2402, A*0101 and Cw*0402 at levels that are well maintained over time. In particular, it seems that the 9-mer and 10-mer peptide reactivation of cells previously sensitized with the 16-mer peptide results in enhanced T cell reactivation in all donors. Although these results need to be confirmed and validated by further in vivo studies, we speculate that the use of this 16-mer region rather than the single 9-mer or 10-mer peptides would be advantageous in clinical modalities such as adoptive transfer of epitope-specific T lymphocytes or epitope-specific vaccinations. These results support the potential use of the 16-mer peptide in CMV adoptive immune therapy [15].
Supplementary Material
Additional File 1
Table 1. to go DOC
Click here for file
Acknowledgements
This research was conducted through the generous support of the NIH Summer Research Fellowship Program for medical/dental students.
==== Refs
Zaja JA Thomas ED, Blume KG, Forman SJ Cytomegalovirus infection Hematopoietic Cell Transplantation 1999 Blackwell Science 560 583
Bronke C Palmer NM Jansen CA Westerlaken GH Polstra AM Reiss P Bakker M Miedema F Tesselaar K Baarle D Dynamics of Cytomegalovirus (CMV)-Specific T Cells in HIV-1-Infected Individuals Progressing to AIDS with CMV End-Organ Disease J Infect Dis 2005 191 873 880 15717261 10.1086/427828
Rauser G Einsele H Sinzger C Wernet D Kuntz G Assenmacher M Campbell JD Topp MS Rapid generation of combined CMV-specific CD4+ and CD8+ T-cell lines for adoptive transfer into recipients of allogeneic stem cell transplants Blood 2004 103 3565 3572 14670917 10.1182/blood-2003-09-3056
Kuzushima K Hayashi N Kimura H Tsurumi T Efficient identification of HLA-A*2402-restricted cytomegalovirus-specific CD8(+) T-cell epitopes by a computer algorithm and an enzyme-linked immunospot assay Blood 2001 98 1872 1881 11535524 10.1182/blood.V98.6.1872
Provenzano M Mocellin S Bettinotti M Preuss J Monsurro V Marincola FM Stroncek D Identification of immune dominant cytomegalovirus epitopes using quantitative real-time polymerase chain reactions to measure interferon-gamma production by peptide-stimulated peripheral blood mononuclear cells J Immunother 2002 25 342 351 12142557 10.1097/00002371-200207000-00006
Kondo E Akatsuka Y Kuzushima K Tsujimura K Asakura S Tajima K Kagami Y Kodera Y Tanimoto M Morishima Y Takahashi T Identification of novel CTL epitopes of CMV-pp65 presented by a variety of HLA alleles Blood 2004 103 630 638 12947002 10.1182/blood-2003-03-0824
Provenzano M Lim JB Mocellin S Monsurro V Bettinotti M Marincola FM Stroncek DF The matrix protein pp65(341–350): a peptide that induces ex vivo stimulation and in vitro expansion of CMV-specific CD8+ T cells in subjects bearing either HLA-A*2402 or A*0101 allele Transfusion 2003 43 1567 157 14617317 10.1046/j.1537-2995.2003.00564.x
Hakenberg J Nussbaum AK Schild H Rammensee HG Kuttler C Holzhutter HG Kloetzel PM Kaufmann SH Mollenkopf HJ MAPPP: MHC class I antigenic peptide processing prediction Appl Bioinformatics 2003 3 155 158 15130801
Holzhütter HG Kloetzel PM A kinetic model of vertebrate 20S proteasome accounting for the generation of major proteolytic fragments from oligomeric peptide substrates Biophysical J 2000 79 1196 1205
Riddell S Greenberg P Principles for adoptive T cell therapy of human viral disease Annu Rev Immunol 1995 13 545 586 7612234 10.1146/annurev.iy.13.040195.002553
Einsele H Roosnek E Rufer N Sinzger C Riegler S Loffler J Grigoleit U Moris A Rammensee HG Kanz L Kleihauer A Frank F Jahn G Hebart H Infusion of cytomegalovirus (CMV)-specific T cells for the treatment of CMV infection not responding to antiviral chemotherapy Blood 2002 99 3916 3922 12010789 10.1182/blood.V99.11.3916
Lim JB Kwon OH Kim HS Kim HO Choi JR Provenzano M Stroncek D Adoptive immunotherapy for cytomegalovirus (CMV) disease in immunocompromised patients Yonsei Med J 2004 45 18 22 15250045
Szmania S Galloway A Bruorton M Musk P Aubert G Arthur A Pyle H Hensel N Ta N Lamb L JrDodi T Madrigal A Barrett J Henslee-Downey J van Rhee F Isolation and expansion of cytomegalovirus-specific cytotoxic T lymphocytes to clinical scale from a single blood draw using dendritic cells and HLA-tetramers Blood 2001 98 505 512 11468143 10.1182/blood.V98.3.505
Watanabe N Kamachi Y Koyama N Hama A Liang J Nakamura Y Yamamoto T Isomura M Kudo K Kuzushima K Kojima S Expansion of human CMV-specific cytotoxic T lymphocytes to a clinical scale a simple culture system using tetrameric HLA-peptide complexes Cytotherapy 2004 6 514 522 15512918 10.1080/14653240410005005
Trivedi D Williams RY O'reilly RJ Koehne G Generation of Cytomegalovirus (CMV)-specific T lymphocytes using protein-spanning pools of pp65-derived pentadecapeptides for adoptive immunotherapy Blood 2005 105 2793 2801 15514011 10.1182/blood-2003-05-1433
| 15916715 | PMC1175859 | CC BY | 2021-01-04 16:39:28 | no | J Transl Med. 2005 May 26; 3:23 | utf-8 | J Transl Med | 2,005 | 10.1186/1479-5876-3-23 | oa_comm |
==== Front
Chiropr OsteopatChiropractic & Osteopathy1746-1340BioMed Central London 1746-1340-13-71596702110.1186/1746-1340-13-7ReviewGolf and upper limb injuries: a summary and review of the literature McHardy Andrew J [email protected] Henry P [email protected] Macquarie Injury Management Group Macquarie University, Sydney 2109 Australia2005 25 5 2005 13 7 7 16 4 2005 25 5 2005 Copyright © 2005 McHardy and Pollard; licensee BioMed Central Ltd.2005McHardy and Pollard; licensee BioMed Central Ltd.This is an Open Access article distributed under the 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
Golf is a popular past time that provides exercise with social interaction. However, as with all sports and activities, injury may occur. Many golf-related injuries occur in the upper limb, yet little research on the potential mechanisms of these injuries has been conducted.
Objective
To review the current literature on golf-related upper limb injuries and report on potential causes of injury as it relates to the golf swing.
Discussion
An overview of the golf swing is described in terms of its potential to cause the frequently noted injuries. Most injuries occur at impact when the golf club hits the ball. This paper concludes that more research into golf-related upper limb injuries is required to develop a thorough understanding of how injuries occur. Types of research include epidemiology studies, kinematic swing analysis and electromyographic studies of the upper limb during golf. By conducting such research, preventative measures maybe developed to reduce golf related injury.
Golfinjuryshoulderelbowwristreview
==== Body
Introduction
Golf is a popular recreational activity that can be played by all ages, genders, and skill levels. Although seemingly uncommon, golf-related injuries do occur, with the three most common injury sites being the lower back, the elbow and the wrist. Together these three sites account for approximately 80% of all injuries sustained by golfers [1-4]. While a number of investigators have conducted research into back-related golfing injuries [5-8] and reviewed how these injuries were sustained [9,10], little research has been identified on how golfing injuries occur in the elbow and wrist [11,12]. The purpose of this paper is to review the function of the upper limb during the golf swing. Also presented is a review of golf-related injuries of the wrist, the elbow and the shoulder as they relate to the golf swing. Finally, there is a discussion on avenues for potential research to understand golf-related upper limb injuries.
Methods
A search of the literature was conducted in the following databases: Medline, Cinahl and Mantis (1966 to present, 1982 to present and 1980 to present respectively). A search of the terms: golf and injury and shoulder or elbow or wrist revealed 45 papers. After setting criteria that required blinded peer-reviewed English language journals only, 42 papers were eventually selected. The literature was collated and sorted according to injury site and relevance. The reference lists of selected papers were examined to determine if any reference papers not found in the original search were relevant. The authors conducted an assessment of methodology and shortcomings of studies, with the findings presented in the discussion section.
The golf swing
The golf swing is a dynamic movement with the potential to cause injury to the golfer. Various injuries occur in different sections of the swing and frequently involve soft tissue injuries [1-4]. An understanding of the mechanics of the golf swing will facilitate appropriate knowledge of the etiology of the injury, thereby improving management. This is particularly true of upper limb golf-related injuries as the arms go through a large range of motion (ROM) during the swing, while providing the link between the fast moving club and the power-generating torso.
The golf swing can be defined as the process of swinging the club to hit the ball. Other than the address position (Figure 1A) it can be divided into seven parts: early backswing (Figure 1B), late backswing (Figure 1C), top of swing (Figure 1D), downswing (Figure 1E), acceleration (Figure 1F), early follow-through (Figure 1G), and late follow-through (Figure 1H).
Figure 1 A-H. Phases of the golf swing. A. Address position, B. Early backswing, C. Late backswing, D. Top of swing, E. Downswing, F. Acceleration, G. Early follow-through, H. Late follow-through.
The golf swing is also often divided into 5 phases: the backswing, the downswing, acceleration, early follow-through and late follow-through [9,13]. In the right-handed golfer, the backswing results in the club being moved away from the direction of intended ball flight and is characterised by a rotation of the shoulder girdle to the right. There is resulting right arm abduction, flexion and external rotation with corresponding left arm adduction, flexion and internal rotation. This takes the golf club in the desired direction. To achieve this movement, the right scapula retracts, while the left scapula protracts and this allows their movement around the trunk in a clockwise movement. The muscles that are predominantly active in this phase and produce these movements are upper and middle trapezius on the right, and the subscapularis and serratus anterior on the left [14-18].
At the top of the backswing, the wrists are in radial deviation, with the right wrist also displaying submaximal extension (Figure 1D).
The downswing phase starts from the top of the backswing and involves the club returning along a similar path to the backswing in preparation to hit the ball, and it involves rapid arm movement. The combined movement of left rotation of the shoulder girdle and scapular rotation, in an anti-clockwise direction around the trunk, is required during the downswing, resulting in increased activity in the left medial scapulae stabilisers/ retractors. To achieve right-sided internal shoulder rotation and flexion, the pectoralis major is very active, while the right upper serratus anterior contracts to assist scapular protraction [14-18].
As seen in Figure 1F, the wrists remain in a similar position to the top of the backswing, a position that is termed 'cocked'.
The acceleration phase of the golf swing is the continuation of the downswing to ball impact. The club head is accelerated to its peak velocity in this phase just prior to contact with the ball, making this the most active phase of the entire golf swing. Bilaterally, the pectoralis muscles are the most active muscles, being the major movers of the shoulder girdle. There is continuation of the right side activity seen during the early downswing, while the left pectoralis appears to maintain an eccentric contraction to control the left arm abduction and external rotation. The muscles involved in scapular movement are also active: the upper serratus on the right to protract the scapula and the levator scapulae on the left side to aid scapular tilting [14-18]. Just prior to impact there is a large increase in wrist flexor muscle activation; what has been termed the 'flexor burst' [11,19,20]. Part of this activity is to return the wrists back (thus club head back) to a position to hit the ball, the 'uncocking' of the wrists.
The early follow-through of the golf swing occurs after ball impact and is the phase where deceleration of trunk rotation occurs. There is a 'rolling' of the forearms at impact that is continued into the early follow-through. This results in left arm supination and right arm pronation followed by left arm external rotation and right arm internal rotation. Bilaterally, the pectoralis major muscles continue to be very active. The active muscles in the shoulder during this phase are the right subscapularis and the left infraspinatus to control the movement seen in the follow-through [14-18].
In the late follow-through, the muscle activity decreases as the golfer nears the end of the swing. The most active muscles in this phase are similar to the early follow-through, but with a lesser degree of activity. The only exception in the upper body is the right serratus anterior, which is more active in this phase as it aids in the protraction of the scapular around the trunk [14-18].
Wrist/Hand injuries
The wrist is one of the most common sites of injury in golfers [3,4]. The wrist accounts for 13–20% of all injuries in amateurs and 20–27% of all injuries in professionals in golf injury epidemiology studies [1-4]. During the golf swing, the wrist is the anchor point between the club and the body. This results in the wrist displaying a large range of motion [19,20]. Wrist injuries commonly occur at the impact point of the golf swing and may result from hitting an object other than the ball. The injury is the result of the sudden change in load applied to the club, and subsequently the golfer, resulting in tissue disruption to the hands and wrist. This commonly occurs in amateurs due to hitting the ball 'fat' (i.e., hitting the ground before the ball). Professionals also sustain wrist injuries but these injuries usually occur in slightly different circumstances. The professional (or amateur) may hit an obscured rock whilst playing from 'the rough' (longer grass that borders the shorter grass of the fairway, the central area that is preferable to hit from). In many major tournaments, particularly "links" courses commonly seen in the United Kingdom, the rough tends to be thick. Whilst attempting to extricate the ball, the long strands of grass tend to wrap themselves around the hosel (that part of the club that joins the shaft to the club head) and shaft of the club. This results in a similar deceleration of the club head during the downswing as hitting the ground, which lends itself to injury. Injury may be either acute where enough force is produced to cause excessive soft tissue elongation in a single swing, or by way of repetitive microtrauma if repeated many times in a short timeframe. Injuries of this nature tend to occur at the hand and wrist but can also occur at the elbow. Muscular strains (particularly the flexor carpi ulnaris [FCU]) and ligamentous strains are common [21,22], but fractures of the hook of hamate may also occur due to this mechanism [23].
Overuse injuries to the wrist are also common and are due mainly to repetitive wrist movement during practice or from alteration to the swing that results in stress to unaccustomed areas. According to a study of the Spain National Insurance Scheme for sportsmen, 10% of golf injuries occur in the wrist. This is contrary to the statistics produced in golf epidemiology studies. A reason for this difference could be differing definitions of what an injury is in each study. The Spanish study found that overuse or sudden changes in swing were the common injury mechanisms, and the FCU was the most common site of injury [21].
Tendonopathy, or more specifically tendonosis has replaced tendonitis as the clinical descriptor of the overuse syndrome [24,25]. The primary reason for this change is due to the majority of overuse tendonopathies displaying collagen degeneration and fibre disorientation. However they do not display the presence of inflammatory cells [24], hence the "itis" is inaccurate. The injury mechanism is either a sudden increase in the volume of practice or alteration of the grip (causing increased loading on an unaccustomed part of the wrist), and then subsequent practice [26]. Onset of the pain is gradual. It tends to have a persistent nature until any aggravating factor(s) are modified or adequate repair (healing) time elapses [24-26].
The FCU of the right wrist in right-handed golfers is vulnerable to injury from microtrauma due to the large forces produced by the swing just prior to impact. This is particularly true when golfers take divots (hit the ground) [26]. As the club hits the ground, a sudden resistance occurs that loads the flexor tendon. If the forces are great enough microtrauma can occur, which combined with repetition through practice may lead to injury. Injury to the FCU results in pain at the proximal border of the trapezium and is increased with wrist flexion.
In the presence of a faulty swing style, the beginner is also susceptible to extensor carpi ulnaris (ECU) injury [26]. Commonly, the beginner 'casts' the club in the early downswing (the early uncocking of the wrist during the downswing and a source of lost power and control), which loads the ECU [26]. Beginners are often overenthusiastic in their practice in an endeavour to improve their game. This may result in repetitive loading, microtrauma and injury to the ECU. A sign of ECU injury includes ulnar wrist pain with tenderness of the dorsal base of the ulnar styloid where the ECU runs through the sixth dorsal compartment. There is often pain on resisted supination and on ulnar deviation in this instance.
An uncommon injury seen in golfers is a fracture to the hook of hamate. Hamate fractures may be acute in nature due to the impingement of the hamate between the hand and the butt end of the club, leading to a fracture in the leading hand (the left hamate in a right-handed golfer) [23]. The literature records acute hamate fractures in golfers as early as 1972 [23]. Stress fractures of the hamate may also occur due to a sudden change in grip positioning or strength with accompanying excessive practice [27]. The ulnar border of the wrist is the site of pain for hamate fractures, with hamate tenderness and positive percussion being an indication for imaging. Care must be taken, however, as x-rays may initially not reveal the fracture [28]. Bone scans or MR imaging will show the fracture.
Other unusual golf-related injuries to the wrist and surrounding structures have also been reported in the literature. These include a case of an amateur golfer with a compression neuropathy of the median nerve in the right palm due to mechanical compression of the median nerve in the right palm by the head of the first metacarpal bone of the left hand [29]. Extensor carpi ulnaris (ECU) tendon dislocation over the ulnar dorsal ridge of the ulnar head aggravated by excessive practice has also been reported [30]. This case was resolved by extensor retinaculum release and partial ulnar head resection after conservative therapy failed. The unusual "hypothenar hammer syndrome" has also been reported in a golfer due to the repetitive hitting of practice balls with a 'faulty' grip causing repeated pressure on the ulnar artery underlying the hypothenar eminence. This practice resulted in thrombus formation in the ulnar artery [31]. While unusual, putting grip alterations have resulted in pain to the volar radial wrist due to a flexor carpi radialis strain. It was reported that this was accentuated by palpation and that a return to the original grip with manual therapy resolved the condition [32].
Elbow injuries
Elbow injuries are common in golfers, especially in amateurs and particularly in females. This is thought to be due to the increased carrying angle seen in the female population [3]. Elbow injuries account for 25–33% of all injuries in amateurs and 7–10% of all injuries in professionals. Ironically, lateral elbow injuries are more common, at a rate of 5:1 when compared to medial elbow injuries (including the so-called Golfer's elbow) [2].
Medial elbow injuries are thought to result from traction-based insults to the elbow, usually to the trailing arm (right elbow in the right-handed golfer). It is the wrist/hand flexors and forearm pronators that are injured at their insertion into the medial epicondyle. These injuries are usually of a traumatic nature and occur at the time of impact. The mechanism is a sudden deceleration of the club head, leading to an increased loading of the medial elbow. This can be due to hitting obscured rocks and tree roots, and in professionals trying to hit repeatedly out of long and thick rough. With amateurs, the hitting of a 'fat' shot is the more likely mechanism. Signs of medial epicondylitis (Golfer's elbow) include pain and tenderness to palpation of the medial epicondyle. Pain is often aggravated by resisted forearm flexion and forearm pronation. There may be trigger point referral along the radial border of the forearm into the dorsum of the hand.
Injury of the lateral aspect of the elbow, the insertion of the wrist/hand extensors into the lateral epicondyle, is more likely to be due to overuse [33]. Gripping the club too tightly during the swing (causing associated extensor eccentric contraction) or changes to the grip with subsequent practice (often fatigue-based) may result in changes in forearm musculature forces and are potentially a source of lateral epicondylitis. Signs of lateral epicondylitis include pain and tenderness to palpation of the lateral epicondyle. Pain is often aggravated by resisted forearm extension and on occasions gripping objects or shaking hands. There may be trigger point referral along the ulnar border of the forearm into the palmar aspect of the hand.
Excessive practice may also result in injury to the lateral elbow. The large increase in flexor activity just prior to impact, the 'flexor burst' [11] accompanied by the rapid wrist movement at the same time places a large stress on the elbow joint and may result in injury due to accumulating microscopic damage [34].
Even though the elbow is a common injury site in golfers, little research has been conducted in this area. Most of the elbow injury mechanisms and management plans are based on racquet sports related injuries. Research focusing on the mechanics of the elbow and related musculature would allow for the accurate aetiology of golf-related elbow injuries to be determined. Understanding how these injuries occur in golfers would ensure the development of appropriate management strategies targeting golf specific injury mechanisms.
Shoulder injuries
Shoulder pain in golfers is a relatively common occurrence compared to other sites of the body, accounting for approximately 8–18% of all golf injuries [1-4].
The shoulder goes through a large ROM during the golf swing including a large degree of left shoulder horizontal adduction and right shoulder external rotation in the backswing. In the follow-through, there is a large degree of left shoulder external rotation and horizontal abduction and right shoulder horizontal adduction [35]. Consequently, excessive practice can produce problems of the shoulder due to overuse.
Injuries to the shoulder in golfers are mainly restricted to the lead shoulder, the left shoulder in right-handed golfers. Studies have found that shoulder pain may be localised to the acromioclavicular (AC) joint, with the potential for either osteoarthritis or distal clavicle osteolysis (which implies horizontal plane compression loading of the joint) [36]. A second study found that posterior instability and subacromial impingement were common findings in golfers with shoulder pain [37]. This pain and instability were reproduced at the top of the backswing (maximal horizontal adduction) [37]. Previously, Bell found that maximal forces about the AC joint occurred in horizontal abduction and adduction. Similar positions are attained by the arm at the top of the back swing (left arm horizontal adduction) and at the end of the follow-through (left arm horizontal abduction), which emphasizes the potential for injury to the AC joint by excessive practice of the golf swing [38].
The practitioner should ascertain the phase of the golf swing that produces the patients shoulder pain; this may facilitate the diagnosis [39]. Posterior shoulder pain in the left shoulder of a right-handed golfer at the top of the backswing should alert the clinician to tightness of the rotator cuff musculature, tightness of the posterior capsule, or posterior capsulitis [39]. Anterior joint line pain at the top of the backswing implies impingement of the humeral head and anterior labrum, while pain localised to the AC joint indicates possible degeneration or impingement of the AC joint [39].
The follow-through phase of the swing may produce posterior shoulder pain due to impingement of either the posterior labrum or the underside of the rotator cuff muscles [39]. Shoulder pain that is generalised and occurs throughout the swing may be due to scapular lag, which alters the mechanics of the shoulder during the swing [39].
A study of golfers who underwent shoulder arthroplasty and were able to return to golf, found that the right shoulder was operated on more frequently (14 out of 26). However, this study made no mention of the cause of the patients shoulder pain. The study also asked a group of surgeons about their opinion of the patient returning to golf after arthroplasty. Out of 44 respondents, 91% encouraged a return to play. This survey showed that shoulder arthroplasty does not necessarily prohibit a return to golf [40].
It is noteworthy that a lack of trunk rotation may require the much smaller shoulder rotators to become excessively active to maintain the momentum of the golf swing. Such a scenario would likely result in the shoulder dysfunction frequently noted in golfers, particularly instability in professionals. It is also worthy to note that those with back problems may potentially induce a shoulder problem in their attempt to reduce the loads on a painful back. Baulbian noted similar observations in his research on a modified golf swing where the back swing is shortened. This research reported that the forces generated in the lower back were reduced by this swing, but the forces generated in the shoulder were greater [41]. This results in the potential for this swing to produce shoulder injury that maybe the result of impingement, instability or rotator cuff tendonopathy. Pain location and shoulder orthopaedic testing helps to differentiate between each clinical entity, though MRI is required to provide a definitive diagnosis.
Discussion
On examining the literature on golf injuries to the upper limb, it is apparent that the majority of papers are case report-based. A case study reports on an individual patient's outcomes and as a result there are inherent limitations such as a lack of control and an inability to generalize findings to the whole population. This type of study, however, provides a platform for the establishment of a testable hypothesis to be made with further research [42]. The studies on golf injury epidemiology allow for a comparison of injury frequency to specific injury sites and also between different groups of golfers (based on gender, skill and age). Most of these studies are retrospective in nature. These types of studies allows for a great deal of data to be gathered, but are susceptible to recall bias. Recall bias occurs when what is thought to have occurred in the past is different to what truly occurred. The use of prospective studies would dramatically reduce recall bias.
How the data are collected influences the accuracy of the data set. Response rates influence how well the results collected can be extrapolated to the population in question. The higher the response rate, the more likely the data are applicable to the whole population in question. Response rates were generally poor ranging from 20.6% to 43%. However, if the sample size is large enough, such data may still be helpful when comparing sites and rates of injury.
An anonymous survey sent in the mail is more likely to be accurate, when compared to a personal interview, particularly with sensitive questions. The majority of the epidemiology studies cited use an anonymous mailed survey that was sent to a group of golfers.
It is apparent that little direct research has been conducted into golf-related upper limb injuries. Much of what if known about injuries relating to the upper limb comes from studies of racquet sports, particularly tennis. While a number of studies have analysed muscle activity in the shoulder musculature during the golf swing, the studies analysed the swing of professional/elite golfers. In many cases, this data may not be applicable to the 'average' golfer (e.g. handicap of 18) due to a difference in the golf swing. To overcome this, research on the swing of the 'average' golfer concentrating on what occurs at the shoulder is needed. This type of study should also look at different swing types: the modern swing, the classic swing and the more recent hybrid swing. Many injuries in golf relate to the wrist and elbow and occur at impact during the golf swing. Research into the forces that occur in the 'perfect' swing and also what occurs in different types of swings/incidents such as miss hits and hitting the ground could provide information on why such injuries occur. Data collected in the research mentioned above may inform injury management (including conditioning / rehabilitation programs) and also potentially prevent upper limb injuries during golf.
Conclusion
The golf swing is a complex body movement involving a large ROM of the upper limb that acts as a link between the golf club and the body. Injuries to the upper limb account for the majority of golf-related injuries recorded. Many injuries occur as the club impacts the ball and are muscle-related. An understanding of how the body moves and the muscle activity achieved during the golf swing helps the health practitioner to understand why these injuries occur. Further study into the different types of golf swing and the different skill levels of golfers is required to fully understand the upper limb function in the golf swing. Such understanding may enable the development of management and prevention programs to reduce the upper limb injuries caused by golf.
Authors' contributions
AJM: Conception and design, search data, paper collection, drafting manuscript, final approval.
HPP: Conception and design, search data, critical review of manuscript, final approval.
==== Refs
Gosheger G Liem D Ludwig K Greshake O Winkelmann W Injuries and overuse syndromes in golf Am J Sp Med 2003 31 438 443
McCarroll Retting AC Shelbourne KD Injuries in the amateur golfer Phys Sports Med 1990 18 122 26
Batt ME A survey of golf injuries in amateur golfers Br J Sports Med 1992 26 63 5 1600459
McCarroll JR Gioe TJ Professional golfers and the price they pay Phys Sports Med 1982 10 64 70
Sugaya H Tschiya A Moriya H Morgan DA Banks SA Farrally MR, Cochran AJ Low-Back Injury in Elite and Professional Golfers An Epidemiologic and Radiographic Study Proceedings of the World Scientific Congress of Golf Science & Golf Ill: 20–24 July 1998; St Andrews 1998 Human Kinetics: Champaign 83 91
Burdorf A Van Der Steenhoven GA Tromp-Klaren EG A one-year prospective study on back pain among novice golfers Am J Sports Med 1996 24 659 64 8883688
Leigh RJ Young DB Farrally MR, Cochran AJ Back pain among junior golfers Proceedings of the World Scientific Congress of Golf Science & Golf Ill: 20–24 July 1998; St Andrews 1998 Human Kinetics: Champaign 92 6
Hosea TM Gatt CJ Galli NA Cochran AJ, E, FN Biomechanical analysis of the golfer's back Proceedings of the World Scientific Congress of Golf Science & Golf I: 20–24 July 1990; St Andrews 1990 Spon: London 43 48
McHardy A Pollard H Lower back pain in golfers: A review J Chiro Med 2005
Hosea TM Back pain in golf Cl Sports Med 1996 15 37 53
Glazebrook MA Curwin S Islam MN Kozey J Stanish WD Medial epicondylitis. An electromyographic analysis and an investigation of intervention strategies Am J Sports Med 1994 22 674 9 7810792
Loftice J Fleisig GS Zheng N Andrews JR Biomechanics of the elbow in sports Clin Sports Med 2004 23 519 30 15474219 10.1016/j.csm.2004.06.003
McHardy A Pollard H Luo K Golf injuries: A review Sports Med 2005
Kao JT Pink M Jobe FW Perry J Electromyographic analysis of the scapular muscles during a golf swing Am J Sports Med 1995 23 19 23 7726345
Pink M Jobe FW Perry J Electromyographic analysis of the shoulder during the golf swing Am J Sports Med 1990 18 137 40 2343980
Jobe FW Perry J Pink M Electromyographic shoulder activity in men and women professional golfers Am J Sports Med 1989 17 782 7 2624291
Moynes DR Perry J Antonelli DJ Jobe FW Electromyography and motion analysis of the upper extremity in sports Phys Ther 1986 66 1905 11 3786421
Jobe FW Moynes DR Antonelli DJ Rotator cuff function during a golf swing Am J Sports Med 1986 14 388 92 3777315
Cahalan TD Cooney WP 3rdTamai K Chao EY Biomechanics of the golf swing as related to club handle design Biomechanics in sport 1987 6 107 111
Cahalan TD Cooney WP 3rdTamai K Chao EY Biomechanics of the golf swing in players with pathologic conditions of the forearm, wrist, and hand Am J Sports Med 1991 19 288 93 1867337
Anonymous Golfers' wrist Br Med J 1977 2 1622 589384
Skolnick AA 'Golfer's wrist' can be a tough break to diagnose JAMA 1998 279 571 572 9486739 10.1001/jama.279.8.571
Torisu T Fracture of the hook of the hamate by a golfswing Clin Orthop 1972 83 91 94 5014836
Maffulli N Wong J Almekinders LC Types and epidemiology of tendinopathy Clin Sports Med 2003 22 675 92 14560540 10.1016/S0278-5919(03)00004-8
Clancy WG JrHagan SV Tendinitis in golf Clin Sports Med 1996 15 27 35 8903707
Murray PM Cooney WP Golf-induced injuries of the wrist Clin Sports Med 1996 15 85 109 8903711
Guha AR Marynissen H Stress fracture of the hook of the hamate Br J Sports Med 2002 36 224 5 12055122 10.1136/bjsm.36.3.224
Walsh JJ IVBishop AT Diagnosis and management of hamate hook fractures Hand Clin 2000 16 397 403 10955213
Hsu WC Chen WH Oware A Chiu HC Unusual entrapment neuropathy in a golf player Neurology 2002 59 646 7 12196674
Oka Y Handa A Recurrent dislocation of the ECU tendon in a golf player: release of the extensor retinaculum and partial resection of the ulno-dorsal ridge of the ulnar head Hand Surg 2001 6 227 30 11901471 10.1142/S0218810401000680
Mueller LP Mueller LA Degreif J Hypothenar hammer syndrome in a golf player: A case report Am J Sports Med 2000 28 741 5 11032235
McHardy A Pollard H Unusual cause of wrist pain in a golfer Br J Sports Med 2004 38 e34 15562149 10.1136/bjsm.2004.011783
Stockard AR Elbow injuries in golf J Am Osteopath Assoc 2001 101 509 516 11575037
Stover CN Wiren G Topaz SR Modern golf swing and stress syndromes Phys Sportsmed 1976 4 42 47
Mitchell K Banks S Morgan D Sugaya H Shoulder motions during the golf swing in male amateur golfers J Orthop Sports Phys Ther 2003 33 196 203 12723676
Mallon WJ Colosimo AJ Acromioclavicular joint injury in competitive golfers J South Orthop Assoc 1995 4 277 82 8925382
Hovis WD Dean MT Mallon WJ Hawkins RJ Posterior instability of the shoulder with secondary impingement in elite golfers Am J Sports Med 2002 30 886 90 12435657
Bell R Acus R Noe D A study of acromioclavicular forces J Sh Elbow Surg 1993 2
Jobe FW Pink MM Shoulder pain in golf Clin Sports Med 1996 15 55 63 8903709
Jensen KL Rockwood CA Shoulder arthroplasty in recreational golfers J Shoulder Elbow Surg 1998 7 362 7 9752645 10.1016/S1058-2746(98)90024-6
Bulbulian R Ball KA Seaman DR The short golf back swing : effects on performance and spinal health implications J Manipulative Physiol Ther 2001 24 569 75 11753330 10.1067/mmt.2001.118982
Portney LG Watkins MP Foundations of clinical research: Applications to practice 2000 Second New Jersey NJ: Prentice Hall Health 268
| 15967021 | PMC1175860 | CC BY | 2021-01-04 16:38:23 | no | Chiropr Osteopat. 2005 May 25; 13:7 | utf-8 | Chiropr Osteopat | 2,005 | 10.1186/1746-1340-13-7 | oa_comm |
==== Front
Crit CareCritical Care1364-85351466-609XBioMed Central London cc34801598738710.1186/cc3480ResearchErythropoietin response in critically ill mechanically ventilated patients: a prospective observational study DeAngelo Alan J [email protected] David G [email protected] Michael W [email protected] Deborah Ebert [email protected] Daniel R [email protected] Physician, Pulmonary and Critical Care Service, Dwight David Eisenhower Army Medical Center, Fort Gordon, Georgia, USA2 Fellow, Pulmonary and Critical Care Service, Brooke Army Medical Center, Fort Sam Houston, Texas, USA3 Physician, Pulmonary and Critical Care Service, David Grant Air Force Medical Center, Travis Air Force Base, California, USA4 Pulmonary and Critical Care Service, Brooke Army Medical Center, Fort Sam Houston, and Assistant Program Director PCCM fellowship, Brooke Army Medical Center, Fort Sam Houston, Texas, USA2005 25 2 2005 9 3 R172 R176 18 11 2004 8 12 2004 19 12 2004 27 1 2004 Copyright © 2005 DeAngelo 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
Anemia is a common problem in critically ill patients. The etiology of anemia of critical illness is often determined to be multifactorial in the clinical setting, but the pathophysiology remains to be elucidated. Erythropoietin (EPO) is an endogenous glycoprotein hormone that serves as the primary stimulus for erythropoiesis. Recent evidence has demonstrated a blunted EPO response as a factor contributing to anemia of critical illness in specific subsets of patients. Critically ill patients requiring mechanical ventilation who exhibit anemia have not been the subject of previous studies. Our goal was to evaluate the erythropoietic response to anemia in the critically ill mechanically ventilated patient.
Methods
A prospective observational study was undertaken in the medical intensive care unit of a tertiary care, military hospital. Twenty patients admitted to the medical intensive care unit requiring mechanical ventilation for at least 72 hours were enrolled as study patients. EPO levels and complete blood count were measured 72 hours after admission and initiation of mechanical ventilation. Admission clinical and demographic data were recorded, and patients were followed for the duration of mechanical ventilation. Twenty patients diagnosed with iron deficiency anemia in the outpatient setting were enrolled as a control population. Control patients had baseline complete blood count and iron panel recorded by primary care physicians. EPO levels were measured at the time of enrollment in conjunction with complete blood count.
Results
The mean EPO level for the control population was 60.9 mU/ml. The mean EPO level in the mechanically ventilated patient group was 28.7 mU/ml, which was significantly less than in the control group (P = 0.035). The mean hemoglobin value was not significantly different between groups (10.6 g/dl in mechanically ventilated patients versus 10.2 g/dl in control patients; P > 0.05).
Conclusion
Mechanically ventilated patients demonstrate a blunted EPO response to anemia. Further study of therapies directed at treating anemia of critical illness and evaluating its potential impact on mechanical ventilation outcomes and mortality is warranted.
==== Body
Introduction
Critically ill patients frequently develop anemia during their intensive care unit (ICU) course. Corwin and coworkers [1] reported that 95% of patients demonstrated abnormal hemoglobin concentration by the third ICU day. Anemia in the ICU patient has been reported to resemble anemia of chronic disease in its metabolic pattern [2]. The etiology of anemia of critical illness is multifactorial; it often results from a combination of primary losses, abnormal coagulation, nutritional deficiencies, depressed bone marrow production, and phlebotomy. Recent evidence has demonstrated a blunted erythropoietin (EPO) response to be a factor contributing to anemia of critical illness in specific subsets of patients, including those with sepsis, multiple trauma, and pediatric critical illness [3-5]. The EPO response in adult patients requiring mechanical ventilation for respiratory failure has not been studied as a primary end-point.
EPO is an endogenous glycoprotein hormone that serves as the primary stimulus for erythropoiesis. The kidney is the primary site of EPO production, but the liver also produces the hormone. EPO acts in the bone marrow, where it promotes terminal differentiation of progenitor cells into erythrocytes [6]. Diminished arterial oxygen content associated with anemia or hypoxia is the major stimulus for EPO production and usually produces an exponential increase [7-9].
Anemia of critical illness and blood management strategies in the ICU continue to be areas of active research. Two recent trials [10,11] demonstrated a reduction in the number of transfusions in critically ill patients treated with recombinant human EPO (rHuEPO). Mortality and adverse clinical events were not statistically different between groups in either study. Hebert and coworkers [12] investigated the effects of a restrictive (threshold 7 g/dl, goal 7–9 g/dl) versus a liberal (threshold 10 g/dl, goal 10–12 g/dl) transfusion strategy in critically ill patients. The authors noted a similar overall 30-day mortality rate between groups but a significantly lower 30-day mortality rate for less acutely ill patients in the restrictive group (Acute Physiology and Chronic Health Evaluation II score <20 and age <55 years). The mortality rate was higher in patients with significant cardiac disease treated with the liberal strategy, but the results did not achieve statistical significance (P = 0.69).
Mechanical ventilation is a common treatment in ICU patients with respiratory failure. A major goal of ICU care is to reduce the number of ventilator days. Numerous clinical factors have an impact on the duration of mechanical ventilation. Improving oxygen delivery to tissues is a recognized goal of ICU care, but its specific impact on outcomes in mechanically ventilated patients is not known. Anemia can lead to a reduction in oxygen delivery. The potential impact of anemia on mechanical ventilation outcomes continues to be evaluated, but there is evidence to suggest a negative impact. Nevins and Epstein [13] found that a low admission hematocrit was significantly associated with death in patients with chronic obstructive pulmonary disease receiving mechanical ventilation. Khamiees and coworkers [14] reported that mechanically ventilated patients with low hemoglobin levels are more likely to be unsuccessfully extubated than are patients with higher hemoglobin levels. Ouellette and colleagues [15] reported that a low hemoglobin level during a period of mechanical ventilation was the most significant risk factor for failure to wean from mechanical ventilation.
We hypothesized that critically ill patients requiring mechanical ventilation have an inadequate EPO response to anemia, which contributes to the development and persistence of anemia of critical illness.
Materials and methods
The study was approved by the Institutional Review Board at Brooke Army Medical Center and was performed in accordance with the ethical standards laid down in the 1964 Declaration of Helsinki. All participants (or surrogates) were counseled and informed consent was obtained before entry into the study.
Study patients
Adult patients (>18 years) admitted to the medical ICU of Brooke Army Medical Center with acute respiratory failure requiring mechanical ventilation for 72 hours and with a hemoglobin level below 13 g/dl were screened for eligibility. Patients with a pre-existing indication for the use of rHuEPO, including anemia associated with end-stage renal disease, cancer, or cancer therapy, and those with HIV infection treated with zidovudine were excluded. Patients with acute or chronic bleeding of any etiology and those who received rHuEPO either before admission or during the ICU course were also excluded. Transfusion thresholds and goals and mechanical ventilation management was at the discretion of the attending physician. Transfusion guidelines outlined by Hebert and coworkers [12] and the American College of Chest Physicians weaning guidelines [16] were provided as a reference, and adherence to these practices was encouraged. In total, 20 study patients were enrolled from January 2003 to December 2003.
Demographic and clinical data including Acute Physiology and Chronic Health Evaluation II scores were recorded at study entry. Admission complete blood count and basic metabolic panel were reviewed. After study enrollment, hemoglobin and EPO levels at day 3 were recorded for statistical analysis, and the arterial oxygen tension (PaO2)/fractional inspired oxygen (FiO2) ratio at day 3 was calculated. Patients were followed for the duration of mechanical ventilation.
Control group
The control group consisted of 20 ambulatory patients with a new diagnosis of iron deficiency anemia (hemoglobin <13 g/dl, ferritin <100 ng/ml, iron <46 μg/dl) screened from a primary care clinic. All patients were free of acute illness, had normal renal function, and had not received rHuEPO during the preceding 30 days. Demographic data and hemoglobin and EPO levels were recorded for statistical analysis.
Erythropoietin assay
Serum EPO levels were measured using a commercial two-site chemiluminescence immunoassay (Nichols Advantage Erythropoietin Assay; Nichols Institute Diagnostics, San Clemente, CA, USA) referenced to the World Health Organization recombinant DNA-derived human EPO 1st International Standard (WHO 87/684). Expected values were determined from data on 119 healthy adults (age range 18–69 years). The results ranged from <5.0 to 25.1 mU/ml. The 95% confidence interval was 5.0–24.6 mU/ml. Reproducibility was determined according to the National Committee for Clinical Laboratory Standards EP5-T2 tentative guidance document [17]. The limit of detection is estimated to be 1.2 mU/ml. The limit of detection was determined from 20 replicate determinations of the zero standard and is defined as the value two standard deviations above the mean of the 20 replicates. The functional sensitivity is estimated at 5.0 mU/ml. The functional sensitivity is based on the lowest concentration of EPO in serum where the interassay precision does not exceed a 20% coefficient of variation.
Statistical analysis
Independent samples t-test was used to evaluate differences in age, hemoglobin, and EPO by group. Paired t-test was used to compare observed versus expected EPO levels by group. A linear regression on EPO as a function of hemoglobin level by group was performed. The results were expressed as mean ± standard deviation. P < 0.05 was considered statistically significant.
Results
Twenty (5 male, 15 female; mean age 70 years, range 49–88 years) critically ill patients requiring mechanical ventilation for acute respiratory failure were enrolled in the study. Table 1 summarizes the study patients' characteristics. Of the 20 study patients, 18 had a PaO2/FiO2 ratio on day 3 of less than 300. Hemoglobin and EPO values were compared with those of 20 (5 male, 15 female; mean age 60 years, range 24–84 years) control patients with iron deficiency anemia.
There was no significant difference in hemoglobin level between the groups (mean hemoglobin 10.6 ± 1.5 g/dl in the study group versus 10.2 ± 1.0 g/dl in the control group; independent samples t-test, P = 0.381). Because there was no difference between groups with respect to hemoglobin, we compared the groups with respect to EPO level. A significantly lower EPO level was recorded in the mechanically ventilated patient group (mean EPO level 28.7 ± 30.4 mU/ml in the study group versus 60.9 ± 58.3 mU/ml in the control group; independent samples t-test, P = 0.035).
A linear regression of EPO as a function of hemoglobin was performed to confirm the difference between expected and observed EPO levels between groups (Fig. 1). There was no significant difference between the observed and expected levels of EPO in the control group (P = 1.000), but there was a statistically significant difference in the study group (P = 0.006).
Discussion
Anemia in the ICU is a common problem, with a multifactorial etiology. We evaluated the relationship of the endogenous EPO response to anemia in the setting of mechanical ventilation and demonstrated a significantly diminished response in this population. Ambulatory iron deficient anemic patients were chosen as control patients in order to match the expected degree of anemia in ICU patients. Additionally, this population demonstrated an elevated EPO response to anemia in a previous study [3]. The EPO response in critical illness has been evaluated in specific subsets of patients but not in mechanically ventilated adult patients in a controlled design.
Rogiers and coworkers [3] compared a mixed population of critically ill patients with iron deficient control patients to determine whether a relationship between EPO response and degree of anemia existed. The study group consisted of 22 septic patients (subgroups with and without renal failure) and 14 nonseptic patients (subgroups with and without renal failure). Patients considered hypoxemic (PaO2 <75 mmHg) were excluded from the analysis. The control group comprised 18 ambulatory iron deficient patients without acute illness. Hematocrit values were similar between study and control patients. A significant inverse correlation between hematocrit and EPO was found in the control patients and in the nonseptic patients without renal failure. The correlation of EPO with hematocrit was lost in the septic patients and in the nonseptic patients with acute renal failure. The authors concluded that the EPO response to anemia is severely blunted in critically ill patients.
Krafte-Jacobs and coworkers [5] demonstrated a blunted EPO response in critically ill pediatric patients with acute anemia and acute hypoxia. Enrolled patients included 21 with acute anemia, 18 with acute hypoxemia (normal hemoglobin), 10 critically ill without anemia or hypoxemia, and 21 outpatients with chronic anemia but no acute illness. Hemoglobin levels were equivalent in the acutely anemic and chronically anemic patients. The EPO levels were similar in the acutely anemic, acutely hypoxemic, and critically ill control patients, but significantly less than the EPO levels in the chronically anemic patients. The authors concluded that the EPO response to known physiologic stimuli is blunted in critically ill children.
Hobisch-Hagen and coworkers [4] found no correlation between EPO and hemoglobin concentrations in 23 adult patients suffering from severe trauma. That observational study did not include a control group for comparison. Trauma patients exhibited anemia (mean hemoglobin 10.0 g/dl) on admission without significant increase during the period of observation. The mean EPO level was 49.8 U/l on day 1 without significant increase throughout the study period (to day 9). The authors concluded that patients with multiple trauma exhibit an inadequate EPO response to low hemoglobin concentrations.
In theory, the treatment of anemia in mechanically ventilated patients with respiratory failure should improve oxygen delivery to the tissues. The interplay of the other principal determinants of oxygen delivery (cardiac output and arterial oxygen saturation) and the overall impact on outcome continues to be evaluated. Hebert and coworkers [18] reported the impact of a liberal (threshold hemoglobin 10.0 g/dl, goal 10–12 g/dl) compared with a restrictive (threshold hemoglobin 7.0 g/dl, goal 7–9 g/dl) transfusion strategy in 713 mechanically ventilated patients, representing a subgroup of a larger study [12]. That study found no difference in the duration of mechanical ventilation between groups.
An adverse impact of anemia on outcome in mechanically ventilated patients has been reported. Khamiees and coworkers [14] conducted a prospective study of predictors of extubation outcome in 91 patients recovering from acute respiratory failure and who successfully completed a spontaneous breathing trial. Patients with hemoglobin values under 10 g/dl were five times as likely to have unsuccessful extubation as those patients with hemoglobin above 10 g/dl. To investigate predictors of outcome, Nevins and Epstein [13] conducted a retrospective cohort study of 166 patients with chronic obstructive pulmonary disease requiring mechanical ventilation for acute respiratory failure of diverse etiologies. Univariate analysis demonstrated lower admission hematocrit to be one of several factors associated with higher in-hospital mortality. Ouellette and colleagues [15] reported that a hemoglobin level under 9 g/dl was the most significant risk factor for unsuccessful extubation in a retrospective review of laboratory parameters and their impact on mechanical ventilation outcomes.
The etiology of anemia of critical illness remains unclear, but a blunted endogenous EPO response appears to play a role. The mechanisms that underlie the blunted endogenous EPO response also remain to be elucidated, although recent studies have demonstrated this response across a spectrum of critically ill patients, suggesting that the presence of critical illness rather than any specific diagnosis is the key factor. Patients with hypoxia – an additional stimulus for endogenous EPO production – were excluded in the aforementioned studies of adult patients. Despite the requirement for mechanical ventilation and the presence of hypoxemia (mean PaO2/FiO2 <300), the critically ill patients in our study also exhibited a blunted EPO response. These results indicate that further investigation into the etiology as well as treatment of anemia of critical ill patients should also include hypoxic patients requiring mechanical ventilation.
Limitations of our data include the small sample size and the observational nature of the study. It was not the objective of the present study to determine the clinical impact of a blunted EPO response on mechanical ventilation outcomes, which therefore cannot be addressed.
Conclusion
In summary, we demonstrated that the EPO response to anemia in the critically ill mechanically ventilated patient is blunted, similar to findings in other previously described subsets of critically ill patients. A negative impact of anemia on outcomes in mechanically ventilated patients has been reported. Further study of therapies directed at treating anemia of critical illness and determining its potential impact on mechanical ventilation outcomes and mortality is warranted.
Key messages
• Anemia in the ICU patient is a common problem with a multifactorial etiology.
• The EPO response to anemia in the critically ill mechanically ventilated patient is blunted.
• Further investigation of therapies directed at anemia of critically ill mechanically ventilated patients are necessary to determine potential morbidity and mortality benefits.
Abbreviations
EPO = erythropoietin; FiO2 = fractional inspired oxygen; ICU = intensive care unit; PaO2 = arterial oxygen tension; rHuEPO = recombinant human erythropoietin.
Competing interests
DRO is a member of the Speaker's Bureau and Consultant, Ortho Biotech, and is on the Speaker's Bureau, Pfizer.
Authors' contributions
AJD modified the original protocol, executed the study, analyzed data, and drafted the manuscript. DGB assisted in executing the study, analyzing the data, and drafting the manuscript. MWQ and DEL participated in the original design and coordination of the study, and in writing the original protocol. DRO, MWQ, and DEL conceived the study. DRO assisted in the original design and drafting of the final manuscript. All authors read and approved the final manuscript.
Acknowledgements
Financial support provided by Ortho-Biotech Products, LP through a Cooperative Research and Development Agreement with the Henry M Jackson Foundation. The opinions or assertions contained herein are the private views of the authors and are not to be construed as reflecting the views of the Departments of the Army, Air Force or 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.
Figures and Tables
Figure 1 Linear regression: erythropoietin as a function of hemoglobin. The line represents the best fit to the values in the control group.
Table 1 Clinical profile of enrolled mechanically ventilated patients
Parameter Value
n 20
Age (years; mean [range]) 70 (49–88)
Male/female 5/15
PaO2/FiO2 ratio (mean [range]) 220 (118–385)
APACHE II score (mean [range]) 19.8 (8–36)
Ventilator days (mean [range]) 12.3 (3–56)
Diagnosis (n)
Pneumonia 8
COPD 5
Pulmonary embolus 1
CHF/cardiac ischemia 4
Acute renal failure 2
APACHE, Acute Physiology and Chronic Health Evaluation; CHF, congestive heart failure; COPD, chronic obstructive pulmonary disease; FiO2, fractional inspired oxygen; PaO2, arterial oxygen tension.
==== Refs
Corwin HL Rodriguez RM Pearl RG Enny C Corwin MJ Gettinger A Erythropoietin response in critically ill patients [abstract 143] Crit Care Med 1997 Suppl 1 A82
Corwin HL Krantz SB Anemia of the critically ill: 'acute' anemia of chronic disease Crit Care Med 2000 28 3098 3099 10966311 10.1097/00003246-200008000-00079
Rogiers P Zhang H Leeman M Nagler J Neels H Melot C Vincent JL Erythropoietin response is blunted in critically ill patients Intensive Care Med 1997 23 159 162 9069000 10.1007/s001340050310
Hobisch-Hagen P Wiedermann F Mayr A Fries D Jelkmann W Fuchs D Hasibeder W Mutz N Klingler A Schobersberger W Blunted erythropoietic reponse to anemia in multiply traumatized patients Crit Care Med 2001 29 743 747 11373460 10.1097/00003246-200104000-00009
Krafte-Jacobs B Levetown ML Bray GL Ruttimann UE Pollack MM Erythropoietin response to critical illness Crit Care Med 1994 22 821 826 8181291
Erslev AJ Erythropoietin N Engl J Med 1991 324 1339 1344 2017231
Jelkmann W Erythropoietin: structure, control of production and function Physiol Rev 1992 72 449 489 1557429
Erslev AJ Caro J Miller O Silver R Plasma erythropoietin in health and disease Ann Clin Lab Sci 1980 10 250 257 7396390
Eckardt KU Boutellier U Kurtz A Schopen M Koller EA Bauer C Rate of erythropoietin formation in humans in response to hypobaric hypoxia J Appl Physiol 1989 66 1785 1788 2732171
Corwin HL Gettinger A Rodriguez RM Pearl RG Gubler KD Enny C Colton T Corwin MJ Efficacy of recombinant human erythropoietin in the critically ill patient: a randomized, double-blind, placebo-controlled trial Crit Care Med 1999 27 2346 2350 10579246 10.1097/00003246-199911000-00004
Corwin HL Gettinger A Pearl RG Fink MP Levy MM Shapiro MJ Corwin MJ Colton T Efficacy of recombinant human erythropoietin in critically ill patients: a randomized controlled trial JAMA 2002 288 2827 2835 12472324 10.1001/jama.288.22.2827
Hebert PC Wells G Blajchman MA Marshall J Martin C Pagliarello G Tweeddale M Schweitzer I Yetisir E A multicenter, randomized, controlled clinical trial of transfusion requirements in critical care N Engl J Med 1999 340 409 417 9971864 10.1056/NEJM199902113400601
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
Khamiees M Raju P DeGirolamo A Amoateng-Adjepong Y Manthous CA Predictors of extubation outcome in patients who have successfully completed a spontaneous breathing trial Chest 2001 120 1262 1270 11591570 10.1378/chest.120.4.1262
Ouellette DR Quinn MW Emmons EE Gallup RA Decreased hemoglobin associated with impaired weaning from mechanical ventilation [abstract] Am J Resp Crit Care Med 2000 160 A560
MacIntyre NR Cook DJ Ely EW Epstein SK Fink JB Heffner JE Hess D Hubmayer RD Scheinhorn DJ Evidence-based guidelines for weaning and discontinuing ventilatory support: a collective task force facilitated by the American College of Chest Physicians; the American Association for Respiratory Care; and the American College of Critical Care Medicine Chest 2001 120 375S 395S 11742959 10.1378/chest.120.6_suppl.375S
National Committee for Clinical Laboratory Standards Evaluation of Precision Performance of Clinical Chemistry Devices: Tentative Guideline: NCCLS document EP5-T2 1992 2 Wayne, PA: National Committee for Clinical Laboratory Standards
Hebert PC Blajchman MA Cook DJ Yetisir E Wells G Marshall J Schweitzer I Do blood transfusions improve outcomes related to mechanical ventilation? Chest 2001 119 1850 1857 11399714 10.1378/chest.119.6.1850
| 15987387 | PMC1175870 | CC BY | 2021-01-04 16:04:52 | no | Crit Care. 2005 Feb 25; 9(3):R172-R176 | utf-8 | Crit Care | 2,005 | 10.1186/cc3480 | oa_comm |
==== Front
Crit CareCritical Care1364-85351466-609XBioMed Central London cc34811598738810.1186/cc3481ResearchOutcome and prognostic factors in critically ill patients with systemic lupus erythematosus: a retrospective study Hsu Chia-Lin [email protected] Kuan-Yu [email protected] Pu-Sheng [email protected] Yeong-Long [email protected] Hou-Tai [email protected] Wen-Yi [email protected] Chia-Li [email protected] Pan-Chyr [email protected] Division of Pulmonary Medicine, Department of Internal Medicine, National Taiwan University Hospital, Taipei, Taiwan2 Assistant Professor, Graduate Institute of Clinical Medicine, College of Medicine, National Taiwan University, Taipei, Taiwan3 Professor, Division of Rheumatology, Department of Internal Medicine, National Taiwan University Hospital, Taipei, Taiwan4 Professor, Division of Pulmonary Medicine, Department of Internal Medicine, National Taiwan University Hospital, Taipei, Taiwan2005 25 2 2005 9 3 R177 R183 25 10 2004 25 11 2004 16 12 2004 1 2 2004 Copyright © 2005 Hsu 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
Systemic lupus erythematosus (SLE) is an archetypal autoimmune disease, involving multiple organ systems with varying course and prognosis. However, there is a paucity of clinical data regarding prognostic factors in SLE patients admitted to the intensive care unit (ICU).
Methods
From January 1992 to December 2000, all patients admitted to the ICU with a diagnosis of SLE were included. Patients were excluded if the diagnosis of SLE was established at or after ICU admission. A multivariate logistic regression model was applied using Acute Physiology and Chronic Health Evaluation II scores and variables that were at least moderately associated (P < 0.2) with survival in the univariate analysis.
Results
A total of 51 patients meeting the criteria were included. The mortality rate was 47%. The most common cause of admission was pneumonia with acute respiratory distress syndrome. Multivariate logistic regression analysis showed that intracranial haemorrhage occurring while the patient was in the ICU (relative risk = 18.68), complicating gastrointestinal bleeding (relative risk = 6.97) and concurrent septic shock (relative risk = 77.06) were associated with greater risk of dying, whereas causes of ICU admission and Acute Physiology and Chronic Health Evaluation II score were not significantly associated with death.
Conclusion
The mortality rate in critically ill SLE patients was high. Gastrointestinal bleeding, intracranial haemorrhage and septic shock were significant prognostic factors in SLE patients admitted to the ICU.
==== Body
Introduction
Systemic lupus erythematosus (SLE) is an archetypal autoimmune disease, involving multiple organ systems and with varying course and prognosis. Even though the survival rate among SLE patients has improved over the past few decades [1-3], there remain a host of factors that are associated with death in SLE patients, including the level of disease activity and demonstrable organ damage at presentation [4,5]. Moreover, coronary artery disease has increasingly been recognized to be an important cause of death in SLE patients [6]. In contrast, infections, which develop in the setting of active SLE under aggressive treatment, are often difficult to identify as a single cause of death [7]. Effective treatment for SLE has led to improved prognosis and extended survival times [8,9]. However, intensive treatment concomitantly results in an increased number of disease- or therapy-associated complications, which also require intensive care. Patients with SLE admitted to the intensive care unit (ICU) mostly present with florid disease manifestations, with a compendium of pathologies precipitating the admissions [10]. However, there is a paucity of clinical data regarding prognostic factors in SLE patients admitted for intensive care.
In the present study we analyzed prognostic factors in a cohort of SLE patients admitted to our ICU over the past 8 years, particularly with respect to causes of ICU admission, severity of illness and clinical course during the patients' ICU stays.
Materials and methods
Patients
All patients with SLE admitted to the medical ICU of the National Taiwan University Hospital from January 1992 to December 2000 were included. Diagnosis of SLE was confirmed if the patient fulfilled at least four of the 1982 American Rheumatism Association revised classification criteria [11]. The exclusion criterion was diagnosis of SLE at or after admission to the ICU. If the patient was admitted to the ICU more than once, only data from the first ICU admission were analyzed.
Data collection
We analyzed the following clinical and laboratory parameters: age, sex, underlying diseases and associated manifestations of SLE, causes of admission, Acute Physiology and Chronic Health Evaluation (APACHE) II score [12], arterial oxygen tension/inspired fractional oxygen ratio, complete blood count, characteristics of lesions on chest radiographs, sites of infection and organisms cultured, treatments administered during the patient's ICU stay, occurrence of complications, duration of ICU study and outcome.
The cause of ICU admission was defined as the major problem necessitating admission to the ICU. This was determined on the basis of clinical data. Cardiogenic pulmonary oedema is due to poor cardiac performance. Noncardiogenic pulmonary oedema is due to fluid overloading of a noncardiogenic cause. APACHE II scores were calculated using clinical data available from the first 24 hours of intensive care. The median APACHE II score was used as a cutpoint to classify the patients into high or low score groups. Renal involvement was defined as urinary excretion of more than 500 mg protein/24 hours, cellular casts not attributable to infection, or abnormal histology on renal biopsy. Abnormal complete blood count was defined as haemolytic anaemia or leucopenia (<4 × 109/l), lymphopenia (<1.5 × 109/l), or thrombocytopenia (<100 × 109/l) in the absence of offending drugs. Neutropenia was defined as an absolute neutrophil count under 1.0 × 109/l. Pneumonia was defined as new and persistent radiographic opacity, positive sputum culture and any three of the following: body temperature above 38°C, white blood cell count above 15 × 109/l, increased airway secretions, or worsening gas exchange [13]. Respiratory failure was defined as arterial oxygen tension below 60 mmHg and/or arterial carbon dioxide tension of 50 mmHg or greater while the patient was breathing room air. Acute respiratory distress syndrome (ARDS) was defined in accordance with to the American–European Consensus Conference on ARDS [14]. Sepsis and septic shock were defined in accordance with the criteria of Bone and coworkers [15].
Gastrointestinal bleeding was defined as the presence of at least one of the following: melena, haematemesis, or blood from nasogastric aspirate over 24 hours. Finally, patient outcome was classed as death while the patient was in the ICU or survival to discharge from the ICU.
Statistical analysis
Values are expressed as median (range) for continuous variables, or as a percentage of the group from which they were derived for categorical variables. Differences in survival among subgroups of variables were analyzed by χ2 test or by Fisher's exact test when necessary. A forward stepwise multivariate logistic regression model was applied (SPSS 10.0 for Windows; SPSS Inc., Chicago, IL, USA), using APACHE II score and variables that were at least moderately associated (P < 0.2) with survival in the univariate analysis. P ≤ 0.05 was considered statistically significant.
Results
Clinical characteristics
From January 1992 to December 2000, a total of 4235 patients were admitted to the ICU. Of these, 51 SLE patients were included in the present study. The clinical features of the 51 SLE patients are summarized in Table 1. Three of the 51 patients had associated autoimmune disease in addition to SLE, including one with polymyositis, one with Graves' disease and one with psoriasis. The most common disease manifestation among the 51 SLE patients before ICU admission was mucocutaneous involvement (44 [86.2%]), followed by renal involvement (37 [72.5%]). The median duration from diagnosis of SLE to ICU admission was 27 months (range 1–288 months). Forty-seven patients (92.2%) were receiving corticosteroid medication before ICU admission, with a mean equivalent dose of 20 mg/day prednisolone.
Causes of admission
A total of 60 ICU admissions were included in the present study, with the annual number of admissions of SLE patients fluctuating. No trend favouring any particular cause of ICU admission was identified during the course of the study. There were seven patients with more than one admission to the ICU, including five patients with two admissions and two with three admissions. The causes of ICU admission are summarized in Table 2. The most common cause of admission to the ICU was pneumonia with ARDS (14 [23%]).
Noninfectious causes
Thirty-three (55.0%) admissions to the ICU were due to noninfectious problems. For patients in the cardiogenic category, heart failure was the major cause of admission, including cardiogenic shock and cardiogenic pulmonary oedema. Nine (15.0%) admissions were for pericardial effusion. Among them, three patients were admitted because of cardiac tamponade. Eleven patients had noninfectious pulmonary problems, and noncardiogenic pulmonary oedema was the most common cause. Among the patients with noncardiogenic pulmonary oedema, all were due to acute deterioration in renal function. For patients in the neurological category, status epilepticus was the most common cause of admission, and most (71.4%) had a previous history of seizures.
Infectious causes
Twenty-seven admissions (45.0%) to the ICU were due to infectious diseases, including pneumonia with ARDS and sepsis of extrapulmonary origin (Table 3). The infectious pathogens identified in SLE patients varied considerably. Eleven had positive blood culture results, including six Gram-negative bacilli, four Gram-positive cocci and one fungaemia. Pseudomonas aeruginosa (n = 3), Salmonella (n = 2; groups B and C) and Escherichia coli (n = 1) accounted for the cases of Gram-negative sepsis, whereas Staphylococcus aureus (n = 2; including one methicillin-resistant S aureus), Staphylococcus epidermidis (n = 1) and Streptococcus pneumoniae (n = 1) were the major pathogens of Gram-positive sepsis. Three patients had confirmed positive pleural effusion culture, including one methicillin-resistant S aureus, one S pneumoniae and one Acinetobacter baumannii. One patient suffered from disseminated tuberculosis with tuberculous bacilli isolated from pleural effusion and ascites. One patient had tuberculous meningitis, with tuberculous bacilli isolated from the cerebrospinal fluid.
Clinical course, treatment and outcome
The clinical courses and outcomes in the 51 patients for their first admissions are summarized in Table 3. In order to assess the possible effect of repeat measurement, the results were analyzed separately by all admissions and first admission only; no significant differences were noted.
Forty-one patients were receiving steroid therapy to control the activity of the disease, including seven receiving pulse therapy (equivalent dose of >625 mg/day prednisolone). Also, 35 patients required mechanical ventilation, with three undergoing tracheotomy because of prolonged intubation. Nineteen patients needed dialysis, including 11 who received continuous venovenous haemofiltration because of unstable haemodynamics.
Fifteen (29.4%) had gastrointestinal bleeding during their ICU stay, which manifested as melena, haematemesis, or blood in the nasogastric aspirate. The rate of steroid use was higher in patients with gastrointestinal bleeding than in those who had no gastrointestinal bleeding (87.5% versus 75%), but the association was not statistically significant (P = 0.253). No evidence of mesenteric vasculitis could be demonstrated in the patients with gastrointestinal bleeding. One of them had colon perforation and underwent surgical intervention, whereas in the others the bleeding was controlled by medication without the need for fluid resuscitation or blood component therapy. Four developed pneumothorax during their ICU stay and were treated by tube thoracotomy for drainage.
Intracranial haemorrhage occurred in six patients (11.7%), including four with brainstem haemorrhage, one with subarachnoid haemorrhage and one with frontal lobe haemorrhage. Three patients were admitted to the ICU because of intracranial haemorrhage; these were not included in the six patients.
Whereas the overall mortality of the non-SLE ICU population was 29.0% from 1992 to 2000, the mortality rate for SLE patients admitted to the ICU was 47.0%.
Prognostic factors
To identify prognostic factors for death in SLE patients admitted to the ICU, univariate and multivariate analyses for these factors were conducted. We performed the analyses using data from the first admission of the patients. Table 4 summarizes the variables with at least moderate influence (P < 0.2) on mortality, as determined by univariate analysis. Patients with abnormal complete blood count on admission (P = 0.005), with intracranial haemorrhage occurring while in the ICU (P = 0.018), with complicating gastrointestinal bleeding in the ICU (P = 0.01), and with concurrent septic shock in the ICU (P < 0.001) were at higher risk of mortality. Patients who had sepsis without pulmonary infection as a cause of admission were at lower risk of mortality (P = 0.04).
Multivariate logistic regression analysis showed that the presence of gastrointestinal bleeding, intracranial haemorrhage and septic shock significantly increased the likelihood of dying, whereas causes of ICU admission and APACHE II scores had no influence (Table 5).
Discussion
We found that the mortality rate was high in SLE patients admitted to the ICU. The most common cause of ICU admission was lung injury/respiratory failure, followed by sepsis/systemic inflammatory response syndrome, cardiogenic causes and neurological disorders. The occurrences of gastrointestinal bleeding, intracranial haemorrhage and septic shock during the ICU stay significantly increased the likelihood of dying.
Recent studies [1-3] have demonstrated a greater reduction in mortality in SLE patients than in the general population over the past few decades. The 10-year survival rate in retrospective series has been 75–85%, with more than 90% of patients surviving longer than 5 years [1-3,16,17]. Nevertheless, outcomes and prognosis in acutely ill SLE patients admitted to the ICU have rarely been investigated. In 1996, Ansell and coworkers [10] reported a retrospective study of SLE patients in the ICUs of two hospitals. They investigated a total of 30 patients and demonstrated high mortality rate in SLE patients in critical care units (47%), similar to the rate in the present study (47%). However, they found that the only pretreatment factor that predicted a poor outcome was the presence of renal involvement due to SLE per se. Survival analysis for patients with and those without renal involvement revealed a difference in long-term survival (maximum follow-up period of 120 months) but not in ICU mortality rate. A multivariate analysis of prognostic factors was not performed in that study because of the small number of patients included. We performed a multivariate analysis in 51 SLE patients admitted to the ICU. Although renal involvement due to SLE was not predictive of patient outcome in the ICU, we identified more than one variable influencing mortality rate in our study.
The average ICU mortality from 1992 to 2000 in our hospital was around 29%, which is lower than the mortality rate in SLE patients admitted to the ICU (47%). The other ICU patients might have different clinical characteristics compared with SLE patients. The data show that the SLE patients requiring ICU admission had poorer outcomes than did other critically ill patients admitted to the ICU.
In one study [4], renal damage, thrombocytopenia, lung involvement, SLE Disease Activity Index greater than or equal to 20 at presentation, and age 50 years or older at diagnosis were all predictive of mortality in univariate and multivariate analyses in SLE patients over a 20-year follow-up period. However, the rate of ICU admission in these patients was not mentioned. In the present study these factors were not associated with ICU and in-hospital mortality in SLE patients. The APACHE II score was of little value in predicting outcome, probably because it could not effectively estimate the influence of underlying systemic diseases and the occurrence of possible complications in the SLE patients admitted to the ICU. Gastrointestinal bleeding, intracranial haemorrhage and septic shock during the ICU stay were associated with a greater risk of death, indicating that clinical course and medical care – not the pretreatment morbidity and acute physiological condition – play key roles in influencing the prognosis of SLE patients in the ICU.
The incidence of gastrointestinal haemorrhage in SLE patients is approximately 5% [18]. Previous studies showed that the incidence of gastrointestinal haemorrhage among the general population of patients admitted to the ICU was 3.5–5% [19,20]. In the present study we found that the incidence of gastrointestinal bleeding among SLE patients was much higher (Table 1) than that in the general cohort of patients admitted to the ICU.
We also found intracranial haemorrhage, including brainstem haemorrhage, subarachnoid haemorrhage and frontal lobe haemorrhage, to be a factor that increases the risk of dying. Acute stroke (infarction or intracranial bleeding) in patients admitted to the ICU with non-neurological problems occurred in 1.25% [21]. Subarachnoid haemorrhage occurred in 10 out of 258 patients with SLE in a previous study [22]. Nevertheless, the actual frequency of and factors contributing to intracranial haemorrhage in SLE patients remain undefined. In the ICU it is often difficult to make a diagnosis of cerebrovascular accident in SLE patients with altered mental status, metabolism-induced focal motor abnormalities, or impaired speech because of mechanical ventilation. On the other hand, many factors may contribute to the pathogenesis of acute stroke, including coagulopathy, hypertension, long-term steroid use and lipid disorders. Early diagnosis and appropriate treatment of intracranial haemorrhage are therefore important aspects of intensive care for SLE patients.
We identified various infectious pathogens in SLE patients. The immunocompromised status associated with the disease itself appears to be primarily responsible for the development of infectious complications [23]. Glucocorticoids and immunosuppressive drugs may increase the risk for infections and the number of types of infections that develop. We found the pathogens in SLE patients in the ICU to vary considerably, and the development of septic shock is a major prognostic factor in these patients. In many patients infections develop in the setting of active lupus undergoing aggressive treatment; alternatively, the manifestations of active lupus can mimic infection clinically. It is sometimes difficult to clarify the site of infection and to initiate antimicrobial therapy promptly. Godeau and coworkers [24] found corticosteroid administration to be related to in-hospital mortality in patients with systemic rheumatic disease who were admitted to the ICU. However, that phenomenon did not present in our study. The differences between studies might be due to several factors. First, our study included a relatively small number of patients. Second, a high percentage of patients received steroid treatment before ICU admission and during the ICU stay (92.2% and 80.4%, respectively); more SLE patients not receiving steroid treatment would be necessary to demonstrate a difference between these two groups. However, Godeau and coworkers [24] found corticosteroid treatment to be related to in-hospital mortality, but other immunosuppressive treatments were not related to outcomes in their study. Further large prospective studies might provide more clinical information about the relationship between immunosuppressive agents and outcomes in this patient population.
There some limitations to the present study. Because of the relatively small number of patients included, the patients studied may not be representative the clinical features of the SLE population. Also, because of the retrospective design, the study lacks information on initial disease activity and laboratory data at the first visit to the hospital, although these clinical features may change after medical treatment but before ICU admission. Initial parameters may have little influence on ICU outcomes, but this could not be tested in the present study.
Conclusion
The mortality rate in critically ill patients with SLE is high. We posit that gastrointestinal bleeding, intracranial haemorrhage and septic shock are significant prognostic factors in SLE patients admitted to the ICU. In contrast, the causes of ICU admission and APACHE II score are not significantly associated with mortality.
Key messages
• The mortality rate in critically ill SLE patients remains high.
• We found that gastrointestinal bleeding, intracranial haemorrhage and septic shock were significant prognostic factors in critically ill patients with SLE.
Abbreviations
APACHE = Acute Physiology and Chronic Health Evaluation; ARDS = acute respiratory distress syndrome; ICU = intensive care unit; SLE = systemic lupus erythematosus.
Competing interests
The author(s) declare that they have no competing interests.
Authors' contributions
C-LH participated in study design and drafted the manuscript. K-YC conceived the study, participated in its design and helped to draft the manuscript. P-SY participated in study design and data collection. Y-LH participated in study design and data collection. H-TC participated in study design and data collection. W-YS performed statistical analysis. C-LY participated in study design. P-CY participated in study design.
Acknowledgments
We thank Dr Fu-Chang Hu at Division of Biostatistics, Graduate Institute of Epidemiology, College of Public Health, National Taiwan University, for his invaluable assistance in data analysis and in the establishment of the multivariate regression model.
Figures and Tables
Table 1 Clinical features of patients with systemic lupus erythematosus admitted to the intensive care unit
Clinical feature Value
Age (years; mean [range]) 29 (12–55)
Female (n [%]) 47 (92.2)
APACHE II score (mean [range]) 19 (9–37)
White blood cell count (×109/l; mean [range]) 8.0 (2.2–136.0)
Platelet count (×109/l; mean [range]) 132.0 (17.0–474.0)
Thrombocytopenia (n [%]) 23 (45.1)
Neutropenia (n [%]) 2 (3.9)
Pulmonary manifestations (n [%])
Consolidation 29 (56.9)
Interstitial 19 (37.3)
Pleural effusion 25 (49.0)
APACHE, Acute Physiology and Chronic Health Evaluation.
Table 2 Causes of admission to the intensive care unit in critically ill patients with systemic lupus erythematosus
Cause of admission Total Noninfectious Infectious
Cardiogenic 11 (18.3)
Cardiogenic shock 4 (6.6)
Ventricular arrhythmia 2 (3.3)
Cardiogenic pulmonary oedema 2 (3.3)
Pericardial effusion with cardiac tamponade 3 (5.0)
Lung injury/respiratory failure 25 (41.6)
Pneumonia with ARDS (including one pulmonary tuberculosis) 14 (23.3)
Noncardiogenic pulmonary oedema 7 (11.6)
Interstitial pneumonitis 1 (1.6)
Pulmonary embolism 1 (1.6)
Haemothorax 1 (1.6)
Upper airway obstruction 1 (1.6)
Sepsis without pulmonary infection 13 (21.7)
Unknown origin of infection 9 (15.0)
Infective endocarditis 1 (1.6)
Peritonitis 1 (1.6)
Cellulitis 1 (1.6)
Meningoencephalitis (tuberculous) 1 (1.6)
Neurological disorder 11 (18.3)
Status epilepticus 7 (11.6)
Intracranial haemorrhage on admission 3 (5.0)
Ischaemic stroke 1 (1.6)
Values are expressed as number (%). ARDS, acute respiratory distress syndrome.
Table 3 Disease courses and outcomes of patients with systemic lupus erythematosus admitted to the intensive care unit
Courses and outcomes Number (%)
Need for mechanical ventilation 35 (68.6)
Steroid use in the ICU 41 (80.4)
Total parenteral nutrition 8 (15.6)
Continuous venovenous haemofiltration 11 (21.6)
Peritoneal dialysis 4 (7.8)
Haemodialysis 16 (31.3)
Operation 6 (11.8)
Gastrointestinal bleeding in the ICU 15 (29.4)
Intracranial haemorrhage in the ICU 6 (11.8)
Pneumothorax in the ICU 4 (7.8)
Septic shock in the ICU 15 (29.4)
Length of ICU stay (days; mean [range]) 7 (1–68)
Death in the ICU 24 (47.0)
Death in the hospital 24 (47.0)
ICU, intensive care unit.
Table 4 Variables that possibly influence the mortality of patients with systemic lupus erythematosus admitted to the intensive care unit: univariate analysis
Variable n Died (n [%]) P
APACHE II score
>19 (median value) 24 11 (45.8) 0.361
≤ 19 27 9 (33.3)
Previous seizure attack before admission
Yes 14 3 (21.4) 0.110
No 37 17 (45.9)
Sepsis without pulmonary infection on admission
Yes 13 2 (15.4) 0.04
No 38 18 (47.4)
Abnormal complete blood count
Yes 41 23 (50.0) 0.005
No 10 0 (0)
Gastrointestinal bleeding in the ICU
Yes 15 11 (68.7) 0.01
No 36 13 (29.5)
Intracranial haemorrhage in the ICU
Yes 6 5 (83.3) 0.018
No 45 15 (33.3)
Concurrent septic shock in the ICU
Yes 15 14 (93.3) <0.001
No 36 6 (16.7)
Included are Acute Physiology and Chronic Health Evaluation (APACHE) II score and variables moderately associated (P < 0.2) with survival. ICU, intensive care unit
Table 5 Variables that significantly influence the mortality of patients with systemic lupus erythematosus admitted to the intensive care unit: multivariate analysis
Variable Death: RR (95% CI) P
Gastrointestinal bleeding in the ICU
Yes 6.97 (0.98–49.68) 0.05
No 1
Intracranial haemorrhage in the ICU
Yes 18.68 (1.13–307.06) 0.04
No 1
Concurrent septic shock
Yes 77.06 (6.85–866.90) <0.001
No 1
CI, confidence interval; ICU, intensive care unit; RR, relative risk.
==== Refs
Uramoto KM Michet CJ Thumboo J Sunku J O'Fallon WM Gabriel SE Trends in the incidence and mortality of systemic lupus erythematosus, 1950–1992. Arthritis Rheum 1999 42 46 50 9920013 10.1002/1529-0131(199901)42:1<46::AID-ANR6>3.0.CO;2-2
Kiss E Regeczy N Szegedi G Systemic lupus erythematosus survival: results from a single center. Clin Exp Rheumatol 1999 17 171 177 10342042
Urowitz MB Gladman DD Abu-Shakra M Farewell VT Mortality studies in systemic lupus erythematosus. Results from a single center. III. Improved survival in SLE. J Rheumatol 1997 24 1061 1065 9195509
Abu-Shakra M Urowitz MB Gladman DD Gough J Mortality studies in systemic lupus erythematosus. Results from a single center. II. Predictor variables for mortality. J Rheumatol 1995 22 1265 1270 7562756
Gladmann DD Prognosis and treatment of systemic lupus erythematosus. Curr Opin Rheumatol 1995 7 402 408 8519613
Sturfelt G Eskilsson J Nived O Truedsson L Valind S Cardiovascular disease in systemic lupus erythematosus. A study of 75 patients from a defined population. Medicine (Baltimore) 1992 71 216 223 1518395
Cohen MG Li EK Mortality in systemic lupus erythematosus: active disease is the most important factor. Aust N Z J Med 1992 22 5 8 1580864
Studenski S Allen NB Caldwell DS Rice JR Polisson RP Survival in systemic lupus erythematosus: A multivariate analysis of demographic factors. Arthritis Rheum 1987 30 1326 1332 3435564
Bresnihan B Outcome and survival in systemic lupus erythematosus. Ann Rheum Dis 1989 48 443 445 2662916
Ansell SM Bedhesi S Ruff B Mahomed AG Richards G Mer M Feldman C Study of critical ill patients with systemic lupus erythematosus. Crit Care Med 1996 24 981 984 8681602 10.1097/00003246-199606000-00018
Tan EM Cohen AS Fries JF Masi AT McShane DJ Rothfield NF Schaller JG Talal N Winchester RJ The 1982 revised criteria for classification of systemic lupus erythematosus. Arthritis Rheum 1982 25 1271 1277 7138600
Knaus WA Draper EA Wagner DP Zimmerman JE APACHE II: a severity of disease classification system. Crit Care Med 1985 13 818 29 3928249
Torres A Aznar R Gatell JM Jimenez P Gonzalez J Ferrer A Celis R Rodriguez-Roisin R Incidence, risk, and prognostic factors of nosocomial pneumonia in mechanically ventilated patients. Am Rev Respir Dis 1990 142 523 528 2202245
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: Definition, mechanism, relevant outcomes, and clinical trial coordination. Am J Respir Crit Care Med 1994 149 818 824 7509706
Bone RC Fisher CJ JrClemmer TP Slotman GJ Metz CA Balk RA Sepsis syndrome: a valid clinical entry. Crit Care Med 1989 17 389 393 2651003
Jacobsen S Petersen J Ullman S Junker P Voss A Rasmussen JM Tarp U Poulsen LH van Overeem Hansen G Skaarup B Mortality and causes of death of 513 Danish patients with systemic lupus erythematosus. Scand J Rheumatol 1999 28 75 80 10229135 10.1080/030097499442522
Urowitz MB Gladman DD Evolving spectrum of mortality and morbidity in SLE. Lupus 1999 8 253 255 10413200 10.1191/096120399678847821
Hoffman BI Katz WA The gastrointestinal manifestations of systemic lupus erythematosus: a review of literature. Semin Arthritis Rheum 1980 9 237 247 6996096 10.1016/0049-0172(80)90016-5
Cook DJ Fuller HD Guyatt GH Marshall JC Leasa D Hall R Winton TL Rutledge F Todd TJ Roy P Risk factors for gastrointestinal bleeding in critical ill patients. N Engl J Med 1994 330 377 381 8284001 10.1056/NEJM199402103300601
Cook DJ Griffith LE Walter SD Guyatt GH Meade MO Heyland DK Kirby A Tryba M Canadian Critical Care Trials Group The attributable mortality and length of intensive care unit stay of clinically important gastrointestinal bleeding in critically ill patients. Crit Care 2001 5 368 375 11737927 10.1186/cc1071
Bleck TP Smith MC Pierre-Louis SJ Jares JJ Murray J Hansen CA Neurologic complications of critical medical illnesses. Crit Care Med 1993 21 98 103 8420739
Mimori A Suzuki T Hashimoto M Nara H Yoshio T Masuyama JI Okazaki H Hirata D Kano S Minota S Subarachnoid hemorrhage and systemic lupus erythematosus. Lupus 2000 9 521 526 11035418
Duffy KN Duffy CM Gladman DD Infection and disease activity in systemic lupus erythematosus: a review of hospitalized patients. J Rheumatol 1991 18 1180 1184 1941820
Godeau B Mortier E Roy PM Chevret S Bouachour G Schlemmer B Carlet J Dhainaut JF Chastang C Short and longterm outcomes for patients with systemic rheumatic diseases admitted to intensive care units: a prognostic study of 181 patients. J Rheumatol 1997 24 1317 1323 9228131
| 15987388 | PMC1175871 | CC BY | 2021-01-04 16:04:52 | no | Crit Care. 2005 Feb 25; 9(3):R177-R183 | utf-8 | Crit Care | 2,005 | 10.1186/cc3481 | oa_comm |
==== Front
Crit CareCritical Care1364-85351466-609XBioMed Central London cc34851598738910.1186/cc3485ResearchThe effect of activated protein C on experimental acute necrotizing pancreatitis Yamenel Levent [email protected] Mehmet Refik [email protected] Bilgin [email protected] Ahmet Turan [email protected] Sezai [email protected] Nuket [email protected] Salih [email protected] Mustafa [email protected] Ilker [email protected] Tahir [email protected] Assistant Professor, Medical Intensive Care Unit, Gülhane School of Medicine, Etlik, Ankara, Turkey2 Associate Professor, Department of Internal Medicine, Gülhane School of Medicine, Etlik, Ankara, Turkey3 Associate Professor, Medical Intensive Care Unit, Gülhane School of Medicine, Etlik, Ankara, Turkey4 Resident, Department of Internal Medicine, Gülhane School of Medicine, Etlik, Ankara, Turkey5 Resident, Department of Surgery, Numune Training Hospital, Sihhiye, Ankara, Turkey6 Resident, Department of Anatomy, Medical Faculty of Hacettepe University, Sihhiye, Ankara, Turkey7 Assistant Professor, Department of Pathology, Gülhane School of Medicine, Etlik, Ankara, Turkey8 Associate Professor, Department of Microbiology, Gülhane School of Medicine, Etlik, Ankara, Turkey9 Assistant Professor, Department of Internal Medicine, Gülhane School of Medicine, Etlik, Ankara, Turkey10 Professor, Department of Internal Medicine, Gülhane School of Medicine, Etlik, Ankara, Turkey2005 4 3 2005 9 3 R184 R190 7 12 2004 12 1 2005 27 1 2005 2 2 2005 Copyright © 2005 Yamenel 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
Acute pancreatitis is a local inflammatory process that leads to a systemic inflammatory response in the majority of cases. Bacterial contamination has been estimated to occur in 30–40% of patients with necrotizing pancreatitis. Development of pancreatic necrosis depends mainly on the degree of inflammation and on the microvascular circulation of the pancreatic tissue. Activated protein C (APC) is known to inhibit coagulation and inflammation, and to promote fibrinolysis in patients with severe sepsis. We investigated the effects of APC on histopathology, bacterial translocation, and systemic inflammation in experimental acute necrotizing pancreatitis.
Materials and method
Forty-five male Sprague-Dawley rats were studied. Rats were randomly allocated to three groups. Acute pancreatitis was induced in group II (positive control; n = 15) and group III (treatment; n = 15) rats by retrograde injection of taurocholate into the common biliopancreatic duct. Group I rats (sham; n = 15) received an injection of normal saline into the common biliopancreatic duct to mimic a pressure effect. Group III rats were treated with intravenous APC 6 hours after induction of pancreatitis. Pancreatic tissue and blood samples were obtained from all animals for histopathological examination and assessment of amylase, tumor necrosis factor-α, and IL-6 levels in serum. Bacterial translocation to pancreas and mesenteric lymph nodes was measured.
Results
Acute pancreatitis developed in all groups apart from group I (sham), as indicated by microscopic parenchymal necrosis, fat necrosis and abundant turbid peritoneal fluid.
Histopathological pancreatitis scores in the APC-treated group were lower than in positive controls (10.31 ± 0.47 versus 14.00 ± 0.52; P < 0.001). Bacterial translocation to mesenteric lymph nodes and to pancreas in the APC-treated group was significantly decreased compared with controls (P < 0.02 and P < 0.007, respectively). Serum amylase, tumor necrosis factor--α, and IL-6 levels were also significantly decreased in comparison with positive controls (P < 0.001, P < 0.04 and P < 0.001, respectively).
Conclusion
APC improved the severity of pancreatic tissue histology, superinfection rates and serum markers of inflammation during the course of acute necrotizing pancreatitis.
See related commentary
==== Body
Introduction
Acute pancreatitis is a local inflammatory process that leads to a systemic inflammatory response in the majority of the cases [1-3]. Severe and life-threatening complications requiring intensive care occur in about 25% of patients with acute necrotizing pancreatitis (ANP) [4]. While the intra-acinar premature activation of digestive enzymes is central to pathophysiological mechanisms of injury, acinar cell apoptosis, increase in oxidative stress, microcirculatory derangements, and release of cytokines contribute to progression of injury and development of extrapancreatic complications [1-5]. Severe acute pancreatitis is usually a result of glandular necrosis [6]. Nuclear factor-κB (NF-κB), a transcription factor that is associated with immediate early gene activation, plays a critical role in the development of necrosis. Although the exact mechanism of NF-κB activation is unknown, once stimulated it leads to production of several inflammatory cytokines, including tumor necrosis factor (TNF)-α [7]. This cytokine is known to increase the severity of pancreatitis by further increasing cytokine production, enhancing pancreatic leukocyte sequestration and accelerating acinar cell apoptosis, ultimately leading to a systemic inflammatory response [8,9]. It has been demonstrated that inhibition of NF-κB activation reduces acinar cell damage and decreases the severity of pancreatitis [10]. Recently, anti-TNF-α treatment in experimental pancreatitis was reported to be of benefit, especially when administered early [11]. However, its effect on established necrotizing pancreatitis is not known.
The protein C pathway serves as a major system for controlling thrombosis, limiting inflammatory responses, and potentially decreasing endothelial cell apoptosis in response to inflammatory cytokines [12]. Recombinant human activated protein C (APC) is known to inhibit coagulation and inflammation, and to promote fibrinolysis in patients with severe sepsis [13]. Binding of APC to the endothelial cell protein C receptor results in a number of actions, including increased activity of APC itself and inhibition of both NF-κB and apoptosis [14].
Edema progresses to necrosis in about 20% of patients with acute pancreatitis [15]. The pancreas is infected in 40–70% of patients with necrotizing pancreatitis, and the mortality rate may be up to 40% when the necrotic tissue becomes superinfected [16]. The most important cause of death in necrotizing pancreatitis is secondary infections, which generally result from translocation of enteric bacteria from the intestine via mainly lymphatic, hematogenous, or transmural routes [17]. On the other hand, prophylactic antibiotic therapy was not found to decrease mortality in controlled clinical trials [18]. Although selective gut decontamination and, to some extent, enteral nutrition were shown to decrease infectious complications [19,20], no specific agent that can strengthen the gut barrier or inhibit translocation of micro-organisms from the gut lumen has yet been identified.
Our aim in the present study was to investigate the effects of recombinant human APC on the progression of experimental ANP. Considering its significant role in inflammatory responses, we hypothesized that APC may alter the degree of local inflammation, development of necrosis and bacterial contamination, and thus the severity of acute pancreatitis.
Materials and methods
The experiment was approved by the Institutional Animal Use and Care Committee of the Gülhane Medical Academy and was performed in accordance with the US National Institutes of Health guidelines for the care and handling of animals.
Animals
Male Sprague–Dawley rats weighing 280–350 g were obtained from the Gülhane School of Medicine Research Center (Ankara, Turkey). Before the experiment the animals were fed standard rat chow, were given free access to water, and were housed in metabolic cages with controlled temperature and 12-hour light–dark cycles for at least 1 week.
Induction of pancreatitis
Anesthesia was induced in rats via inhalation of 250 ml sevoflurane liquid (Abbott, Istanbul, Turkey). Laparotomy was performed through a midline incision. After cannulation of the common biliopancreatic duct with a 28-gauge, 0.5 inch microfine catheter, a microaneurysm clip was placed on the bile duct below the liver and another around the common biliopancreatic duct at its entry into the duodenum to avoid reflux of enteric contents into the duct. Then, 1 ml/kg of 5% sodium taurocholate (Sigma, St. Louis, MO, USA) was slowly infused into the common biliopancreatic duct. The infusion pressure was kept below 30 mmHg, as measured using a mercury manometer. When the infusion was complete, the two microclips were removed and the abdomen was closed in two layers. All procedures were performed using sterile technique.
Study protocol
After the stabilization period, 45 male rats were randomly divided into three groups. Rats in group I (control group; n = 15) underwent laparotomy with manipulation of the pancreas (sham procedure) and received 10 ml/kg saline intravenously (single dose). Groups II and III underwent laparotomy with induction of ANP. Rats in group II (positive control; n = 15) received saline, as in group I but 6 hours after induction of ANP. Rats in group III (treatment group; n = 15) received 100 mg/kg recombinant human APC (Drotrecogin alfa [activated]; Xigris; Lilly, Istanbul, Turkey) intravenously (single dose) 6 hours after induction of ANP. Twenty-four hours after induction of ANP, all surviving animals were killed by intracardiac infection of pentobarbital (200 mg/kg). Blood samples were taken from the heart before the animals were killed in order to measure serum amylase, TNF-α, and IL-6. Animals that died before the end of the study (four in group II and two in group III) were excluded from the analysis.
Histopathologic analysis
A portion of the pancreas from the same anatomical location in each rat, including the main pancreatic duct, was fixed in 10% neutral buffered formalin and embedded in paraffin. One paraffin section stained with hematoxylin and eosin was examined for each pancreas. Two pathologists, who were blinded to the treatment protocol, scored the tissues with respect to edema, acinar necrosis, inflammatory infiltrate, hemorrhage, fat necrosis, and perivascular inflammation in 20 fields. The scores for each histological examination were summed, yielding a maximum score of 24, as defined by Schmidt and coworkers [21].
Amylase measurement
A Hitachi 917 autoanalyzer (Boehringer Mannheim, Mannheim, Germany) was used in the amylase assay.
Tumor necrosis factor-α and interleukin-6 assays
Blood was collected and centrifuged (3000 rpm for 5 min). The serum was stored at -40°C. TNF-α and IL-6 were measured in serum samples using quantitative sandwich enzyme-linked immunosorbent assay kits (R&D Systems Inc., Minneapolis, MN, USA).
Quantitative cultures and bacterial identification
Tissue specimens taken from mesenteric lymph nodes (MLNs) and one portion of the pancreas with macroscopic necrosis were harvested for culture. Each sample was weighed and homogenized. Afterward, the homogenates were diluted serially, quantitatively plated in duplicate on phenylethyl alcohol and MacConkey II agar, and then incubated aerobically at 37°C for 24 hours. Bacterial counts were expressed as colony-forming units/g tissue, and counts of 1000 colony-forming units/g and higher were considered to represent a positive culture. Gram-negative bacteria were identified using the API-20E system (BioMerieux Vitek, Hazelwood, MO, USA). Gram-positive bacteria were identified to the genus level using standard microbiologic methods.
Statistical analysis
Results are expressed as mean ± standard error of the mean. Translocation incidence was evaluated by Fisher's exact test. The significance of differences in total histopathologic scores, serum amylase activities, and cytokine levels were assessed using one-way analysis of variance and Tukey HSD as post hoc tests. Detailed histopathologic scores (e.g. edema and acinar necrosis) were assessed using the Kruskal–Wallis test, and subgroup analyses were conducted using the Mann–Whitney U-test. P < 0.05 was considered statistically significant. All statistical measurements were done using SPSS PC version 9.05 (SPSS Inc., Chicago, IL, USA).
Results
Rats with ANP had extensive parenchyma and fat necrosis, and polymorphonuclear leukocyte infiltration on histologic examination. The total histopathologic score was significantly reduced in group III (10.31 ± 0.47) compared with group II (14.00 ± 0.52; P < 0.001). Although there were marked improvements in pancreatic tissue edema, inflammatory infiltration, fat necrosis, acinar necrosis scores, and perivascular inflammation in APC-treated group III compared with saline-treated group II, there was no significant difference in hemorrhage scores between the two groups (Fig. 1). Histopathologic findings in the groups are summarized in Table 1.
Serum amylase and cytokines assay
Serum amylase, TNF-α, and IL-6 levels in group I (the sham group) were significantly lower than in the other two groups. Significant reductions were found in serum levels of amylase (P < 0.001), TNF-α (P < 0.04), and IL-6 (P < 0.001) in group III (the APC-treated group) compared with group II (the positive control group; Table 2).
Bacterial translocation
Bacteria were cultured from MLNs and pancreatic necrotic tissues in all 11 animals in saline-treated group II. In APC-treated group III, bacterial cultures from MLN samples and pancreatic necrotic tissue samples were positive in seven (54%) and six (46%) of the 13 animals, respectively. MLN and pancreatic tissue infection rates in group III were significantly lower than in group II (P < 0.02 and P < 0.007, respectively). The incidences of bacterial translocation in the three groups are summarized in Fig. 2. Escherichia coli was the most commonly isolated bacteria. Other bacteria isolated from MLNs and pancreatic tissues are listed in Table 3. No organisms were found in either MLNs or pancreatic tissues in rats from group I (sham operated).
Discussion
Acute pancreatitis represents a severe form of inflammation that often leads to severe damage to the gland. Progression from edematous to necrotizing pancreatitis – a process that usually determines the patients' prognosis – is mediated by NF-κB [7]. In the present study, plasma IL-6 and TNF-α levels, together with amylase, were significantly increased after induction of ANP. Stimulation of production of either acute phase proteins and adhesion molecules or several inflammatory cytokines, including TNF-α, IL-1β and IL-6, occurs after NF-κB activation in acute pancreatitis [7]. However, we observed that amylase, and plasma IL-6 and TNF-α levels were all significantly decreased in APC-treated animals.
APC has been shown to inhibit production of TNF-α by decreasing activation of NF-κB [22]. In contrast to many immunomodulatory agents previously tested clinically, recombinant human APC was found to be significantly beneficial in the PROWESS (Recombinant Human Activated Protein C Worldwide Evaluation in Severe Sepsis) study and was approved by the US Food and Drug Administration for use in patients with severe sepsis and septic shock [23,24]. A significant decrease in protein C concentrations was found during the initial phase of experimental ANP [25]. Furthermore, drotrecogin alfa (activated) treatment was recently reported to improve progression of severe sepsis after ANP in two cases [26]. Based on these data and those presented above, APC replacement may interrupt, at least partly, the pathophysiological cascade of inflammation and related events during ANP.
We found significant improvements in pancreatic histology after treatment with recombinant APC. Edema, acinar cell necrosis, fat necrosis, and perivascular inflammation, which occur in almost all inflammatory processes in any organ, resolved in pancreatic tissues from animals treated with APC. However, although we know that the decrease in APC occurs during the initial period of pancreatitis, we only studied its effects in established ANP because we believe that an experimental model should simulate the situation in humans. Indeed, clinically, only a small number of patients with acute pancreatitis present during the early stages of disease. Therefore, the results of the present study are relevant to clinical necrotizing pancreatitis in humans. The concept of administering APC to patients with the disease immediately after the diagnosis is established is rational and should be the focus of research. Nevertheless, more experimental and clinical evidence is needed if we are to evaluate the value of such prophylactic use of APC.
Little is known about effects on the coagulation system in ANP. Because the degree of hemorrhage affects the extent of local and systemic complications in ANP, maintenance of a normal coagulation system in the pancreatic microcirculation in order to prevent thrombosis or bleeding is a desirable objective. Protein C is a critical participant in normal coagulation mechanisms. One interesting finding in the present study was the similarity in hemorrhage scores between groups II and III. In comparison with control animals, APC neither decreased nor increased the incidence of hemorrhagic fields in tissue samples. This not only may refect the anticoagulant effect of APC but also suggests that the coagulation system in pancreas remains intact, even with the organ in a necrotic state.
Evaluation of bacterial translocation after APC treatment was another aim of the study. We found lesser MLN and pancreatic bacterial contamination in APC-administered rats than in control animals. The impact of superinfection of the pancreas is summarized above. The decreased contamination rates may reflect APC-related improvements in gut mucosa. Contamination of necrotic tissues occurs primarily because of translocation of enteric micro-organisms [27]. Our study does not indicate any direct effect of APC on intestinal mucosa, although many factors have been reported to underlie bacterial translocation [27], including intestinal mucosal injury, cecal bacterial overgrowth, decreased gut motility, and compromised host immune functions. Failure of the gut to act as a barrier against bacterial translocation as a result of nitric oxide (NO)-dependent mechanisms [28] has been accepted as one of the most potent origins of sepsis and subsequent organ failure after pancreatitis [29], and inhibition of inducible NO synthase was shown to decrease the incidence of bacterial translocation [30]. Isobe and coworkers [31] found an inhibitory effect of APC on inducible NO synthase induction by decreasing TNF-α production in rats with endotoxin-induced hypotension; a similar action of APC in the impaired intestinal mucosa of the rats with ANP might have been at work in the present study. Future studies addressing the association between APC and NO-dependent damage in the intestine following ANP induction will help to identify a possible second role of APC in the pathogenesis of the disease.
Conclusion
APC improved pancreatic histology and decreased the incidence of bacterial translocation from the intestine in rats with experimental ANP. APC and its reduction appear to play an important role in the pathogenesis of this life-threatening disease. Therefore, the effects of replacement of this mediator in ANP should be a focus of future investigations.
Key messages
• Decrease in APC is important in the pathogenesis of acute pancreatitis related systemic complications.
• Replacement with recombinant APC improved local injury and markers of systemic inflammation, and decreased bacterial translocation from the gut in experimental ANP.
• APC administration may be an alternative treatment option in patients with ANP.
Abbreviations
ANP = acute necrotizing pancreatitis; APC = activated protein C; IL = interleukin; MLN = mesenteric lymph node; NF-κB = nuclear factor-κB; NO = nitric oxide; TNF = tumor necrosis factor.
Competing interests
The author(s) declare that they have no competing interests.
Authors' contributions
LY and MRM conceived the study. BC performed statistical analysis. ATI and SA performed the surgical procedures. NM and SD carried out histological analysis. MO and IT carried out microbiological, amylase and cytokine assays. TU was involved in drafting and revising the manuscript. All authors read and approved the final manuscript.
Figures and Tables
Figure 1 Histology samples from the three groups. (a) Normal pancreatic histology in group I (the sham operated group). (b) Light micrograph showing severe and extensive parenchymal necrosis, with few normal acinar cells in group II (the positive control group). (c) Light micrograph showing mild edema, parenchymal focal necrosis, and inflammation in group III (the activated protein C treated group). All samples were stained with hematoxylin and eosin, and the original magnification for each image is 50×.
Figure 2 Incidences of bacterial translocation to mesenteric lymph nodes (MLNs) and pancreas. Group I, sham operated group; group II, positive control group; group III, activated protein C treated group.
Table 1 Histopathologic scores in the three groups
Feature Group I (sham) Group II (control) Group III (treatment) P (group II versus group III)
Edema 0.67 ± 0.13 3.09 ± 0.16 2.38 ± 0.14 <0.007a
Acinar necrosis 0.07 ± 0.06 1.82 ± 0.12 1.08 ± 0.08 <0.001a
Inflammatory infiltrate 0.73 ± 0.12 2.91 ± 0.16 2.00 ± 0.16 <0.002a
Hemorrhage 0.27 ± 0.12 2.45 ± 0.16 2.08 ± 0.18 NSa
Fat necrosis 0.13 ± 0.09 1.82 ± 0.12 1.31 ± 0.13 <0.02a
Perivascular Inflammation 0.13 ± 0.09 1.91 ± 0.09 1.38 ± 0.14 <0.02a
Histopathologic score 2.00 ± 0.28 14.00 ± 0.52 10.31 ± 0.47 <0.001b
Values are expressed as mean score ± standard error of the mean. aBy Mann–Whitney U test.
bBy Tukey HSD. NS, not significant.
Table 2 Serum amylase, tumor necrosis factor-α, and interleukin-6 levels of groups
Group I (sham) Group II (control) Group III (treatment) P (group II versus group III)
Amylase (pg/ml) 597.2 ± 22.0 1848.3 ± 96.2 1236.1 ± 69.9 <0.001a
TNF-α (pg/ml) 63.4 ± 5.1 114.4 ± 7.9 88.5 ± 7.7 <0.04a
IL-6 (pg/ml) 201.9 ± 17.2 1391.8 ± 106.6 816.2 ± 73.1 <0.001a
Values are expressed as mean ± standard error of the mean. aBy Tukey HSD. IL, interleukin; TNF, tumor necrosis factor.
Table 3 Bacteria isolated from mesenteric lymph node and pancreatic tissue samples
Bacteria Group II (n = 11) Group III (n = 13)
MLNs Pancreas MLNs Pancreas
Escherichia coli 7 8 6 5
Enterococcus sp. 3 2 1 1
Staphylococcus sp. 1 1 0 0
Klebsiella oxytoca 2 1 1 0
Proteus 1 0 0 0
Polimicrobial 2 1 1 0
MLN, mesenteric lymph node.
==== Refs
Neoptolemos JP Raraty M Finch M Sutton R Acute pancreatitis: the substantial human and financial costs Gut 1998 42 886 891 9691932
Norman J The role of cytokines in the pathogenesis of acute pancreatitis Am J Surg 1998 175 76 83 9445247 10.1016/S0002-9610(97)00240-7
Gomez-Cambronero LG Sabater L Pereda J Cassinello N Camps B Vina J Sastre J Role of cytokines and oxidative stress in the pathophysiology of acute pancreatitis: therapeutical implications Curr Drug Targets Inflamm Allergy 2002 1 393 403 14561185 10.2174/1568010023344544
Forsmark CE Toskes PP Acute pancreatitis. Medical management Crit Care Clin 1995 11 295 309 7788533
Whitcomb DC Acute pancreatitis: molecular biology update J Gastrointest Surg 2003 7 940 942 14675701 10.1016/j.gassur.2003.10.001
Baron TH Morgan DE Acute necrotizing pancreatitis N Engl J Med 1999 340 1412 1417 10228193 10.1056/NEJM199905063401807
Pezzilli R Ceciliato R Barakat B Corinaldesi R Immune-manipulation of the inflammatory response in acute pancreatitis. What can be expected? JOP 2004 5 115 121 15138332
Altavilla D Famulari C Passaniti M Galeano M Macri A Seminara P Minutoli L Marini H Calo M Venuti FS Attenuated cerulein-induced pancreatitis in nuclear factor-kappaB-deficient mice Lab Invest 2003 83 1723 1732 14691290 10.1097/01.LAB.0000101734.82054.BE
Gukovskaya AS Gukovsky I Zaninovic V Song M Sandoval D Gukovsky S Pandol SJ Pancreatic acinar cells produce, release, and respond to tumor necrosis factor-alpha. Role in regulating cell death and pancreatitis J Clin Invest 1997 100 1853 1862 9312187
Altavilla D Famulari C Passaniti M Campo GM Macri A Seminara P Marini H Calo M Santamaria LB Bono D Lipid peroxidation inhibition reduces NF-kappaB activation and attenuates cerulein-induced pancreatitis Free Radic Res 2003 37 425 435 12747737 10.1080/1071576031000070093
Oruc N Ozutemiz AO Yukselen V Nart D Celik HA Yuce G Batur Y Infliximab: a new therapeutic agent in acute pancreatitis? Pancreas 2004 28 e1 e8 14707742 10.1097/00006676-200401000-00020
Dhainaut JF Yan SB Claessens YE Protein C/activated protein C pathway: overview of clinical trial results in severe sepsis Crit Care Med 2004 32 Suppl S194 S201 15118517 10.1097/01.CCM.0000128035.64448.45
Lyseng-Williamson KA Perry CM Drotrecogin alfa (activated) Drugs 2002 62 617 630 discussion 631-612 11893230
Haley M Cui X Minneci PC Deans KJ Natanson C Eichacker PQ Activated protein C in sepsis: emerging insights regarding its mechanism of action and clinical effectiveness Curr Opin Infect Dis 2004 17 205 211 15166822 10.1097/00001432-200406000-00006
Banks PA Infected necrosis: morbidity and therapeutic consequences Hepatogastroenterology 1991 38 116 119 1855766
Beger HG Bittner R Block S Buchler M Bacterial contamination of pancreatic necrosis. A prospective clinical study Gastroenterology 1986 91 433 438 3522342
Powell JJ Miles R Siriwardena AK Antibiotic prophylaxis in the initial management of severe acute pancreatitis Br J Surg 1998 85 582 587 9635800 10.1046/j.1365-2168.1998.00767.x
Isenmann R Runzi M Kron M Kahl S Kraus D Jung N Maier L Malfertheiner P Goebell H Beger HG Prophylactic antibiotic treatment in patients with predicted severe acute pancreatitis: a placebo-controlled, double-blind trial Gastroenterology 2004 126 997 1004 15057739 10.1053/j.gastro.2003.12.050
Luiten EJ Hop WC Lange JF Bruining HA Controlled clinical trial of selective decontamination for the treatment of severe acute pancreatitis Ann Surg 1995 222 57 65 7618970
Marik PE Zaloga GP Meta-analysis of parenteral nutrition versus enteral nutrition in patients with acute pancreatitis BMJ 2004 328 1407 15175229 10.1136/bmj.38118.593900.55
Schmidt J Rattner DW Lewandrowski K Compton CC Mandavilli U Knoefel WT Warshaw AL A better model of acute pancreatitis for evaluating therapy Ann Surg 1992 215 44 56 1731649
Joyce DE Grinnell BW Recombinant human activated protein C attenuates the inflammatory response in endothelium and monocytes by modulating nuclear factor-kappaB Crit Care Med 2002 30 Suppl S288 S293 12004250 10.1097/00003246-200205001-00019
Bernard GR Vincent JL Laterre PF LaRosa SP Dhainaut JF Lopez-Rodriguez A Steingrub JS Garber GE Helterbrand JD Ely EW Efficacy and safety of recombinant human activated protein C for severe sepsis N Engl J Med 2001 344 699 709 11236773 10.1056/NEJM200103083441001
Bernard GR Margolis BD Shanies HM Ely EW Wheeler AP Levy H Wong K Wright TJ Extended evaluation of recombinant human activated protein C United States Trial (ENHANCE US): a single-arm, phase 3B, multicenter study of drotrecogin alfa (activated) in severe sepsis Chest 2004 125 2206 2216 15189943 10.1378/chest.125.6.2206
Ottesen LH Bladbjerg EM Osman M Lausten SB Jacobsen NO Gram J Jensen SL Protein C activation during the initial phase of experimental acute pancreatitis in the rabbit Dig Surg 1999 16 486 495 10805548 10.1159/000018774
Machala W Wachowicz N Komorowska A Gaszynski W The use of drotrecogin alfa (activated) in severe sepsis during acute pancreatitis: two case studies Med Sci Monit 2004 10 CS31 CS36 15232511
Dervenis C Smailis D Hatzitheoklitos E Bacterial translocation and its prevention in acute pancreatitis J Hepatobiliary Pancreat Surg 2003 10 415 418 14714160 10.1007/s00534-002-0727-5
Rahman SH Ammori BJ Larvin M McMahon MJ Increased nitric oxide excretion in patients with severe acute pancreatitis: evidence of an endotoxin mediated inflammatory response? Gut 2003 52 270 274 12524412 10.1136/gut.52.2.270
Thomson A Bacterial translocation in acute pancreatitis J Gastroenterol Hepatol 2003 18 1214 12974914 10.1046/j.1440-1746.2003.03145.x
Simsek I Mas MR Yasar M Ozyurt M Saglamkaya U Deveci S Comert B Basustaoglu A Kocabalkan F Refik M Inhibition of inducible nitric oxide synthase reduces bacterial translocation in a rat model of acute pancreatitis Pancreas 2001 23 296 301 11590326 10.1097/00006676-200110000-00011
Isobe H Okajima K Uchiba M Mizutani A Harada N Nagasaki A Okabe K Activated protein C prevents endotoxin-induced hypotension in rats by inhibiting excessive production of nitric oxide Circulation 2001 104 1171 1175 11535575
| 15987389 | PMC1175873 | CC BY | 2021-01-04 16:04:53 | no | Crit Care. 2005 Mar 4; 9(3):R184-R190 | utf-8 | Crit Care | 2,005 | 10.1186/cc3485 | oa_comm |
==== Front
Crit CareCritical Care1364-85351466-609XBioMed Central London cc34881598739010.1186/cc3488ResearchIsolation of Aspergillus spp. from the respiratory tract in critically ill patients: risk factors, clinical presentation and outcome Garnacho-Montero José [email protected] Rosario [email protected] Carlos [email protected]ón Cristóbal [email protected]Álvarez-Lerma Francisco 5Nolla-Salas Juan 6Iruretagoyena José R 7Barcenilla Fernando [email protected] Department of Intensive Care Medicine, Hospital Universitario Virgen del Rocío, Sevilla, Spain2 Department of Intensive Care Medicine, Hospital Universitario Virgen del Rocío, Sevilla, Spain3 Department of Intensive Care Medicine, Hospital Universitario Virgen del Rocío, Sevilla, Spain4 Department of Intensive Care Medicine, Hospital Universitario de Valme, Sevilla, Spain5 Department of Intensive Care Medicine, Hospital Universitari del Mar, Barcelona, Spain6 Department of Intensive Care Medicine, Hospital Universitari del Mar, Barcelona, Spain7 Department of Intensive Care Medicine, Hospital de Cruces, Bilbao, Bikzakia, Spain8 Department of Intensive Care Medicine, Hopsital Universitari Arnau de Vilanova, Lleida, Spain2005 11 3 2005 9 3 R191 R199 30 11 2004 29 12 2004 19 1 2004 2 2 2005 Copyright © 2005 Garnacho-Montero 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
Our aims were to assess risk factors, clinical features, management and outcomes in critically ill patients in whom Aspergillus spp. were isolated from respiratory secretions, using a database from a study designed to assess fungal infections.
Methods
A multicentre prospective study was conducted over a 9-month period in 73 intensive care units (ICUs) and included patients with an ICU stay longer than 7 days. Tracheal aspirate and urine samples, and oropharyngeal and gastric swabs were collected and cultured each week. On admission to the ICU and at the initiation of antifungal therapy, the severity of illness was evaluated using the Acute Physiology and Chronic Health Evaluation II score. Retrospectively, isolation of Aspergillus spp. was considered to reflect colonization if the patient did not fulfil criteria for pneumonia, and infection if the patient met criteria for pulmonary infection and if the clinician in charge considered the isolation to be clinically valuable. Risk factors, antifungal use and duration of therapy were noted.
Results
Out of a total of 1756 patients, Aspergillus spp. were recovered in 36. Treatment with steroids (odds ratio = 4.5) and chronic obstructive pulmonary disease (odds ratio = 2.9) were significantly associated with Aspergillus spp. isolation in multivariate analysis. In 14 patients isolation of Aspergillus spp. was interpreted as colonization, in 20 it was interpreted as invasive aspergillosis, and two cases were not classified. The mortality rates were 50% in the colonization group and 80% in the invasive infection group. Autopsy was performed in five patients with clinically suspected infection and confirmed the diagnosis in all of these cases.
Conclusion
In critically ill patients, treatment should be considered if features of pulmonary infection are present and Aspergillus spp. are isolated from respiratory secretions.
==== Body
Introduction
Aspergillus is a genus of mitosporic fungi, some species of which are known to cause infections in humans, particularly Aspergillus fumigatus (85% of cases) followed by A flavus and A niger [1]. Aspergillus spp. are responsible for a broad spectrum of illnesses, from saprophytic colonization of the bronchial tree to rapidly invasive and disseminated diseases. Invasive aspergillosis remains a major cause of morbidity and mortality in immunosuppressed patients with profound granulocytopenia secondary to haematological malignancies, or solid organ and bone marrow transplantation. Outbreaks of aspergillosis in patients admitted to intensive care units (ICUs) have been reported [2]. Aspergillus spp. can also cause pneumonia in ICU patients without classical predisposing factors, as well as community-acquired pneumonia in otherwise immunocompetent healthy individuals [3,4].
Because the mortality rate with invasive aspergillosis remains high, even in the face of therapy, the work up must be prompt and aggressive. The diagnosis of invasive pulmonary aspergillosis is difficult because definitive diagnosis is based on histological documentation of typical hyphae and a culture positive for an Aspergillus sp. Uncertainty in disease definition is a key contributor to the controversy regarding the optimal method for establishing the diagnosis of invasive infection.
Standard definitions of opportunistic fungal infections in immunocompromised patients with cancer and haematopoietic stem cell transplants were recently proposed [5]. 'Proven' aspergillosis requires histopathological or cytopathological examination showing hyphae with evidence of associated tissue damage, or a positive culture result from a sample obtained using sterile technique along with suggestive clinical or radiological evidence of infection. In addition, 'probable' aspergillosis requires the presence of risk factors in the host, isolation of an Aspergillus sp. and suggestive clinical or radiological findings; 'possible' aspergillosis requires the presence of risk factors in the host and isolation of an Aspergillus sp., or suggestive clinical and radiological findings [5]. Serology is not useful in the diagnosis of aspergillosis, and data regarding the clinical utility of detection of Aspergillus antigenaemia is limited to patients with neutropenia [6].
Treatment is mandatory in severely immunocompromised patients (those with neutropenia or prolonged use of immunosuppressants) with suggestive clinical manifestations or isolation of Aspergillus spp. in respiratory secretions. However, the therapeutic approach is not well defined in critically ill patients without neutropenia or transplantation in whom Aspergillus spp. are cultured in bronchial secretions [7]. Therefore, using data from a large multicentre study designed to assess risk factors and the impact of isolation of fungi in ICU patients, the present study was performed with the following objectives: to determine risk factors for respiratory isolation of Aspergillus spp.; to assess clinical features, treatment and outcomes in patients with Aspergillus spp. recovered from respiratory secretions; and to evaluate the correlation between isolation of Aspergillus spp. in respiratory samples and histopathological findings.
Materials and methods
Study population
A total of 1765 patients older than 18 years of age who were admitted for at last 7 days to 73 medical/surgical ICUs in certain Spanish hospitals between May 1998 and January 1999 were included in the study. The institutional review board of each hospital approved the protocol and waived the need for informed consent.
Design
This was a prospective, cohort, observational, multicentre study. Based on diagnosis at the time of ICU admission, patients were classified as medical, surgical, or trauma. The severity of illness on ICU admission was calculated using the Acute Physiology and Chronic Health Evaluation (APACHE) II scoring system [8]. The definitions of severe sepsis and septic shock used were those of the American College of Chest Physicians/Society of Critical Care Medicine Consensus Conference [9].
In all patients, samples obtained from sputum, tracheal aspirates (intubated patients), urine, pharyngeal exudates and gastric aspirates were cultured for fungi each week. The initial samples were obtained 8 days after admission to the ICU and once a week thereafter. Other samples of peripheral blood or from other infectious foci were obtained at the discretion of the attending physician. Samples were processed by the various reference clinical microbiology laboratories of the participating hospitals using standard procedures, including Sabouraud agar culture and BACTEC method (Becton Dickinson Diagnostic Instrument Systems, Paramus, NJ, USA), for the isolation of fungal species. The A20C system (Byomerieux, Lyon, France) was used for species identification. Candida infection was defined as recovery of Candida spp. from blood samples (in one or more culture bottles), or evidence of endophthalmitis, or a positive culture of tissue specimens or peritoneal fluid culture, or obstruction of the urinary tract by fungal balls.
Risk factors
Various risk factors before ICU admission and during the ICU stay were recorded. These are summarized in Table 1.
Clinical features
Patients with Aspergillus spp. isolated from respiratory samples were retrospectively evaluated. The clinical significance of recovery of Aspergillus spp. was determined individually by the physician in charge, who established whether isolation of Aspergillus spp. represented a case of colonization or infection. Colonization was deemed to be present when the patient did not fulfil criteria for pneumonia; if the patient fulfilled criteria for pneumonia and the clinician in charge considered the isolation of Aspergillus spp. to be clinically valuable, then the patient was considered to be infected. Specific recommendations regarding therapeutic approach when fungi were isolated from culture were not given, and so the decision regarding antifungal treatment was made on an individual basis by the physician in charge. In patients treated with antifungal drugs, adverse events, clinical cure and microbiological eradication (weekly cultures becoming negative) were recorded. For each patient in whom an Aspergillus sp. was detected, clinical data as well as radiographic and computed tomography findings were retrospectively recorded by means of a questionnaire completed by the clinician in charge. Radiographic findings included normal chest radiograph, lobar consolidation, unilateral consolidation, bilateral consolidation and ill-defined nodules [10].
Patients were followed until discharge from the hospital or death during the hospital stay. In patients who died with proven fungal infection or with high suspicion of fungal infection, an autopsy examination was sought.
Statistical analysis
Qualitative variables are expressed as the percentage distribution in each category, and quantitative variables are expressed as mean ± standard deviation in normally distributed variables or median (range) when the distribution was not normal. The Student's t-test or the Mann–Whitney U-test was used for the comparison of categorical and normally distributed and non-normally distributed variables, respectively. Analysis of variance or the Kruskal–Wallis test was used in the comparison of three groups. The χ2 test or the Fisher's exact test was used in the comparison of categorical variables. A comparison of risk factors for the isolation of Aspergillus spp. between groups of patients with Aspergillus spp., patients with Candida spp. infection, and noncolonized, uninfected patients was conducted. For this purpose, a binary logistic regression analysis prior to the bivariate analyses was performed. Variables were included in the model if P ≤ 0.05. Results are expressed as odds ratio (OR), 95% confidence interval (CI). P < 0.05 was considered statistically significant. Statistical analysis was performed using the Statistical Package for the Social Sciences (SPSS) for Windows (version 11.5; SPSS Inc., Chicago, IL, USA).
Results
The study population included 1765 patients (1178 [67%] men; mean [± standard deviation] age 57.8 ± 17.3 years). Underlying diseases were classified as medical in 44% of patients, surgical in 47% and trauma in 9%. A total of 1045 patients were classified as colonized or infected with fungi, and 720 were classified as noncolonized, uninfected patients. Colonization with Candida spp. was diagnosed in 880 (49.8%) patients, Candida spp. infection in 105 (5.9%), and infection with fungi other than Candida spp. in 60 (3.4%). In this group of 60 patients, in whom fungi other than Candida spp. were isolated, Aspergillus spp. were recovered in 38 (63.3%). An Aspergillus sp. was isolated from respiratory secretions in 36 patients (tracheal aspirate 35, sputum 1). A fumigatus was isolated in 35 patients and A niger in one. The length of ICU stay was similar between patients infected with Aspergillus spp. and those infected with Candida spp. (32.1 ± 21.4 days versus 32.8 ± 22.6 days), but it was significantly longer than in noncolonized, uninfected patients (18.4 ± 14.1 days; P < 0.001; Table 2).
Compared with noncolonized, uninfected patients, patients with Aspergillus spp. infection had significantly greater in-hospital mortality (69.4% versus 33%; P < 0.001) and ICU mortality (52.8% versus 24.7%; P < 0.001) rates. Patients with Candida spp. infection also had significantly greater in-hospital mortality (60.9% versus 33%; P < 0.001) and ICU mortality (53.3% versus 24.7%; P < 0.001) rates than did noncolonized, uninfected patients.
Risk factors
The frequency of risk factors for fungal infection before ICU admission were similar in the three groups of patients (Table 2), except for significantly higher rates of chronic obstructive pulmonary disease (COPD), immunosuppression and transplantation in the patients with Aspergillus infection, and a greater prevalence of solid neoplasms in the patients with Candida infection. With regard to risk factors present during the ICU stay, neutropenia and treatment with steroids were significantly more frequent in the Aspergillus group, and total parenteral nutrition was significantly more common in the Candida group (Table 2). Duration of steroid administration was also significantly longer in the Aspergillus group (Table 3). In multivariate analysis, independent factors significantly associated with recovery of Aspergillus spp. compared with noncolonized, uninfected patients were treatment with steroids (OR = 4.5, 95% CI = 1.73–11; P = 0.002) and COPD (OR = 2.9, 95% CI = 1.06–8.08; P = 0.03). When comparisons with patients with Candida infection were performed, immunosuppression (OR = 12.9, 95% CI = 1.34–25; P = 0.001), neutropenia (OR = 9.4, 95% CI = 1.9–19.9; P = 0.02) and COPD (OR = 9.2, 95% CI 1.36–62.5; P = 0.02) emerged as independent factors significantly associated with isolation of Aspergillus spp.
Clinical characteristics
Aspergillus spp. were isolated from respiratory samples in severely ill patients, with a mean APACHE II score on ICU admission of 21.6 ± 6.9 and a mean age of 58.7 ± 16.6 years. Apart for eight patients with Aspergillus infection, the remaining 28 patients had debilitating underlying disorders, with COPD (n = 16), immunosuppression (n = 20) and chronic renal failure (n = 10) being the most common. During their stay in the ICU, 25 patients received steroids and all but one were mechanically ventilated. The mean length of ICU stay before isolation of Aspergillus spp. was 32.1 ± 21.4 days. Previous use of fluconazole was recorded in eight of the 36 patients (22.2%) with isolation of Aspergillus spp., and in 41 of the 105 patients (39%) with invasive candidiasis.
In 14 patients without clinical symptoms of pneumonia, isolation of Aspergillus spp. was interpreted by the clinician in charge as colonization. In two patients Aspergillus spp. were recovered 24 hours before the patient's death, and so the clinical manifestations could not be evaluated. The remaining 20 patients had signs of severe sepsis or septic shock unresponsive to broad-spectrum antibiotics in association with clinical manifestations suggestive of pneumonia. In these cases, isolation of Aspergillus spp. was interpreted to represent infection, and treatment with antifungal agents was started. In seven of these patients, however, bacteria in association with Aspergillus spp. were isolated from the tracheal aspirates, including Pseudomonas aeruginosa (n = 2), Klebsiella pneumoniae (n = 1), Acitenobacter baumannii (n = 1), Stenotrophomonas maltophilia (n = 1), coagulase-negative Staphylococcus spp. and Haemophilus spp. (n = 1). The most frequent radiographic findings were unilateral consolidation and bilateral consolidation.
Treatment and outcome
In the group of 14 patients with Aspergillus colonization, the in-hospital mortality rate was 50% (three patients died in the ICU). Eleven patients were not treated with antifungal drugs, although risk factors were present in seven. Liposomal amphotericin B was prescribed to three patients (one of these patients with predisposing risk factors died in the ICU). The mean cumulative dose of amphotericin B lipid formulation was 3100 mg and the mean duration of treatment was 9 days.
Of the 20 patients with Aspergillus spp. infection 16 died, yielding an in-hospital mortality rate of 80%. All patients were given amphotericin B except one patient, who was treated with intraconazole. Details of treatment are shown in Table 4. The mean APACHE II score at the beginning of antifungal treatment was 22.7 ± 8, as compared with 14.3 ± 2.3 in treated patients colonized with Aspergillus spp. The first choice antifungals were amphotericin B deoxycholate (administered to eight patients), liposomal amphotericin B (eight patients) and amphotericin B lipid complex (three patients). Two patients treated with amphotericin B deoxycholate developed renal failure and treatment was changed to liposomal amphotericin B in one and amphotericin B lipid complex in the other. One patient initially treated with amphotericin B lipid complex was switched to liposomal amphotericin B because of persistence of infection, with positive cultures, after 2 weeks of treatment. After 3 weeks of treatment with liposomal amphotericin B, cultures were negative. Eleven patients died, and in the remaining nine patients treatment was discontinued after clinical cure. Mean duration of treatment in these nine patients was 18 days (range 8–35 days). Clinical resolution of symptoms was achieved with amphotericin B deoxycholate only in one patient and with the lipid formulation in eight (P < 0.05).
Autopsy was performed in five patients with Aspergillus spp. infection. In all cases the examination revealed characteristic hyphae elements within the lung parenchyma with vascular invasion, which is compatible with the diagnosis of invasive aspergillosis. All were COPD patients and had been treated with corticosteroids in the ICU. One patient had a lung cancer. None of these five patients had neutropenia or haematological malignancy.
Discussion
This is the largest study to date in which Aspergillus spp. were isolated from respiratory secretions in a cohort of critically ill patients, including a large number of immunocompetent patients. In this group, isolation of Aspergillus spp. mostly occurred in those with COPD who were treated with steroids during their ICU stay. However, only 13.8% of patients had neutropenia – a classic risk factor for Aspergillus infection.
Various small series and case reports have shown that invasive aspergillosis commonly occurs in critically ill patients admitted to the ICU because of acute exacerbation of COPD and treated with intravenous corticosteroids [11-14]. In those patients steroids were given for a short period (1 week), whereas in our patients treatment was prolonged (3 weeks). In contrast, in a recent study of 250 patients with COPD admitted to the ICU because of acute respiratory failure [15], which did not report on the use of corticosteroids, Aspergillus spp. were not isolated in any respiratory sample. On the other hand, prior treatment with fluconazole was not associated with a higher rate of isolation of Aspergillus spp., as was previously reported in patients with neutropenia [16].
In one-third of cases in the present study recovery of Aspergillus spp. in respiratory secretions, in the absence of signs of pneumonia, was considered to represent colonization. However, three of these patients were given antifungal treatment because of underlying risk factors. An important finding of the study is that systemic antifungal agents were employed in patients with Aspergillus spp. colonization with clinical signs of respiratory infection, despite the fact that associated bacterial pathogens were cultured in almost one-third of cases. Although autopsies were performed in only five patients with Aspergillus infection, histopathological findings confirmed the clinical diagnosis in each case. Our findings are in agreement with those of a recent autopsy study [17] that confirmed the diagnostic value of Aspergillus spp. in respiratory secretions of COPD patients admitted to the ICU and treated with corticosteroids. In contrast, in a study conducted Petri and coworkers [18] in 435 non-neutropenic ICU patients, fungal colonization with Aspergillus spp. was found in 4% of cases, but in none of the patients was a diagnosis of invasive aspergillosis made.
In one study [19], because of the lack of reliable diagnostic tools, up to 60% of patients with invasive aspergillosis diagnosed at autopsy had not received antifungal treatment. Isolation of Aspergillus spp. from respiratory secretions has been regarded as being of limited usefulness in the antemortem diagnosis of invasive aspergillosis. In a study conducted in the 1980 s, Yu and coworkers [20] evaluated 108 patients in whom Aspergillus spp. were isolated from respiratory secretions, but invasive aspergillosis was not demonstrated in non-immunosuppressed patients. In a recent study [21], however, it was shown that malnutrition, diabetes mellitus, pulmonary disorder, or corticosteroid use were underlying risk factors for invasive aspergillosis in patients in whom Aspergillus spp. were isolated from respiratory secretions. On the other hand, invasive aspergillosis does not only occur in immunocompromised patients [3]. In a cohort of 439 non-ICU patients with invasive aspergillosis [22], nine had no apparent underlying conditions before diagnosis. Likewise, acute community-acquired pneumonia due to Aspergillus spp. – a rare infection – has been reported in 12 immunocompetent hosts [4].
It is well known that neutropenia is the main risk factor for aspergillosis because polymorphonuclear neutrophils and macrophages are the first immunological line of defence against Aspergillus spp. [6]. However, T-cell mediated, acquired immunity also plays a role in protecting against fungal infection [23]. Critically ill patients with prolonged stays in the ICU exhibit a complex decrease in immune function, with deactivation of macrophages and altered cellular response [24]. In addition, the immune function of peripheral neutrophils is influenced by acute hyperglycaemia [25]. Furthermore, it has been shown that corticosteroids suppress neutrophil action against Aspergillus hyphae [26]. These mechanisms may explain why Aspergillus infection occurs in ICU patients with a compensatory anti-inflammatory response syndrome or immunoparalysis during multiorgan failure but without any predisposing factors [27,28]; they may also account for the association between corticosteroid use and this invasive fungal infection.
Invasive aspergillosis in ICU patients carries a very high mortality [4,28,29], with an attributable mortality of 18.9% after adjusting for confounding factors [30]. In non-immunocompromised patients, the success of antifungal treatment depends on early diagnosis. However, because delayed diagnosis is the rule, if therapy is not promptly initiated then patients may die from the disease. Amphotericin B deoxycholate was the only therapeutic option in the past and was the antifungal agent used in series with a reported mortality of as high as 100%. In the present study, although there were no differences in in-hospital mortality according to antifungal drug used, clinical cure rates were higher in patients treated with amphotericin B lipid formulations. In two patients amphotericin B deoxycholate was withdrawn because of nephrotoxicity, which increases mortality significantly [31]. Although greater efficacy of amphotericin B lipid formulations compared with amphotericin B deoxycholate in the treatment of invasive aspergillosis has not been demonstrated [7], the use of the lipid formulations appears preferable, especially in critically ill patients, because of better tolerance [32]. New antifungal agents with good activity against Aspergillus spp. have recently become available. Initial treatment of invasive aspergillosis with voriconazole led to better response and improved survival than with the standard approach of initial therapy with amphotericin B [33]. Caspofungine was also effective as salvage therapy in invasive pulmonary aspergillosis, as compared with standard therapy [34].
One of the main limitations of the present study was the retrospective design, in which diagnostic and treatment approaches were not standardized. Also, there were few cases in which the clinical diagnosis of invasive pulmonary aspergillosis was confirmed by histopathological evaluation. Third, mortality rates may be biased by differences in antifungal treatments used at each centre. Nevertheless, the present data add valuable information regarding the significance of isolation of Aspergillus spp. from respiratory samples in critically ill patients.
Conclusion
In summary, COPD and treatment with corticosteroids are major predisposing factors for Aspergillus spp. colonization/infection in critically ill patients. For this reason, in ICU patients with these risk factors, antifungal treatment should be considered in the presence of clinical features of pneumonia and isolation of Aspergillus spp. from respiratory secretions. In contrast, antifungal treatment should not be initiated when Aspergillus spp. are recovered from bronchial aspirates of critically ill patients without predisposing risk factors and in the absence of clinical and radiological signs of pneumonia. In these cases, isolation of Aspergillus spp. should be interpreted as colonization.
Key messages
• COPD and treatment with corticosteroids, and neutropenia are major predisposing factors for respiratory colonization/infection with Aspergillus spp. in critically ill patients.
• In ICU patients with these risk factors, antifungal treatment should be considered in the presence of clinical features of pneumonia and isolation of Aspergillus spp. from respiratory secretions.
• The crude mortality associated with this entity is still very high.
Abbreviations
APACHE = Acute Physiology and Chronic Health Evaluation; CI = confidence interval; COPD = chronic obstructive pulmonary disease; ICU = intensive care unit; OR = odds ratio.
Competing interests
The author(s) declare that they have no competing interests.
Authors' contributions
All of the authors were involved in designing the study and collecting data. JGM and RAV were involved in the statistical analysis. CL obtained funding. JGM drafted the manuscript, which was revised and approved by all of the authors.
Acknowledgements
We thank Gilead Sciences, SL, for financial support in logistic aspects of the study and Marta Pulido, MD, for editing the manuscript and editorial assistance.
This study was supported by an unrestricted grant from Gilead.
This study was carried out with the EPCAN Study Group: J Nolla, F Álvarez-Lerma and M Salvadó (Hospital del Mar, Barcelona); N Carrasco and A Bueno (Hospital de la Princesa, Madrid); F Bobillo and P Ucio (Hospital Clínico, Valladolid); MA León, M Nolla and RA Díaz (Hospital General de Cataluña, Barcelona); JR Iruretagoyena, K Esnaola and I Andetxaga (Hospital de Cruces, Bilbao); A Blanco, F Taboada and R Fernández (Hospital Nuestra Señora de Covadonga, Oviedo); M Nieto, R Diego and F Ortuño (Hospital Clínico San Carlos, Madrid); P Marcos and E Mesalles (Hospital Germans Trias i Pujol, Badalona, Barcelona); A Martínez, M Fernández and F Jaime (Hospital Virgen de la Arrixaca, Murcia); H Sancho and N Izquierdo (Hospital Reina Sofía, Córdoba); M Ulibarrena and F Labayen (Hospital Santiago Apóstol, Vitoria); F Barcenilla, MJ Gil and B Balsera (Hospital Arnau de Villanova, Lleida); R Jordá, M Jurado and J Pérez (Hospital Son Dureta, Palma de Mallorca); E Zavala, A Alcón and N Fabregues (Hospital Clínic i Provincial, Barcelona); MV de la Torre, MA Estecha and A Soler (Hospital Virgen de la Victoria, Málaga); M Bodí and D Castander (Hospital Joan XXIII, Tarragona); A Mendía, J Artaetxebarría and C Reviejo (Hospital Nuestra Señora de Aránzazu, San Sebastián); M Sánchez, A Casamitjana and C Pérez (Hospital Insular, Las Palmas de Gran Canaria); MJ López and E Robles (Hospital General de Segovia, Segovia); Y Insausti and JA Tihistsa (Hospital de Navarra, Pamplona); C García and JM Rubio (Hospital 12 de Octubre, Madrid); R Oltra and O Rodríguez (Hospital Clínico Universitario, Valencia); P Olaechea and R de Celís (Hospital de Galdakao, Bizkaia); JM Soto and J Pomares (Hospital San Cecilio, Granada); J Luna and G Masdeu (Hospital Virgen de la Cinta, Tarragona); R Sierra and A Gordillo (Hospital Puerta del Mar, Cádiz); R Rodríguez and J Fajardo (Hospital Virgen de la Macarena, Sevilla); MA Herranz and JI Gómez (Hospital Río Hortega, Valladolid); RM García and MJ Espina (Hospital de Cabueñes, Gijón); J Garnacho and C Ortiz (Hospital Virgen del Rocío, Sevilla); M Palomar and J Montero J (Hospital Vall d'Hebron, Barcelona); C Cisneros and A Sandiumenje (UCI de Traumatología, Hospital 12 de Octubre, Madrid); M Sánchez and M Álvarez (Hospital Príncipe de Asturias, Madrid); V López and R Julve (Hospital de Sagunto, Valencia); J Solé and M Valerón (Hospital Nuestra Señora del Pino, Las Palmas de Gran Canaria); MA Blasco and S Borrás (Hospital Dr Peset, Valencia); E Maraví and JM Urtasun (Hospital Virgen del Camino, Pamplona); C Sánchez-Díaz (Hospital San Pedro de Alcántara, Cáceres); LM Tamayo (Hospital Río Carrión, Palencia); J Blanco (Complexo Hospitalario Xeral-Calde, Lugo); P Galdós (Hospital General de Móstoles, Madrid); F Barredo (Hospital de Torrecárdenas, Almería); A Rodríguez (Hospital Santa María del Rosell, Cartagena); J Castaño (Hospital Virgen de las Nieves, Granada); A Bonet (Hospital Josep Trueta, Girona); M Cerdá (Hospital de la Creu Roja, L'Hospitalet de Llobregat, Barcelona); A Torres (UVIR, Hospital Clínic i Provincial, Barcelona); F Pérez F (Fundación Jiménez Díaz, Madrid); JM Flores (UCI Traumatología, Hospital Virgen del Rocío, Sevilla); R Diego (Hospital General Universitario, Valencia); C Fernández (Complejo Hospitalario Insalud, León); A Mas (Centre Hospitalari i Cardiologic, Manresa, Barcelona); F Ruiz (Hospital Ciudad de Jaén, Jaén); C León (Hospital Nuestra Señora de Valme, Sevilla); M Casanovas (Hospital de Igualada, Igualada, Barcelona); EA Sanz (Hospital Santa Ana, Motril, Granada); JA Artola (Hospital Naval de San Carlos, Cádiz); MP Luque (UCI de Traumatología, Hospital Clínico Univresitario, Zaragoza); C Palazón (Hospital General Universitario, Murcia); C Sotillo (Hospital Gregorio Marañón, Madrid); A Bisbal (Policlínica Miramar, Palma de Mallorca); MJ Huertos (Hospital de Puerto Real, Cádiz); F Esteban (Hospital Sant Joan de Reus, Reus, Tarragona); P Ugarte (Hospital Marqués de Valdecilla, Santander); R Giral (Hospital General Yagüe, Burgos); V González (Hospital Miguel Servet, Zaragoza); MJ Serralta (Hospital San Juan, Alicante); A Cercas (Hospital de Jerez, Cádiz); A Nebra (Hospital Clínico Universitario, Zaragoza); C Castillo (Hospital Txagorritxu, Vitoria-Gasteiz); A Cercas (Hospital de Jerez, Cádiz); A Nebra (Hospital Clínico Universitario, Zaragoza); C Castillo (Hospital Txagorritxu, Vitoria), A Tejada (UCI Traumatología, Hospital Miguel Servet, Zaragoza); and JI Gómez (REA, Hospital Río Ortega, Valladolid), Spain.
Figures and Tables
Table 1 Risk factors recorded before ICU admission and during ICU stay
Risk factor Comments (where applicable)
Before ICU admission
Surgery before ICU admission Divided into urgent or elective
Diabetes mellitus Only insulin-treated patients
COPD Defined as the presence of a productive cough or expectoration for more than 90 days per year (but on separate days) and for more than 2 consecutive years, provided that a specific disorder responsible for these symptoms was not present
Chronic liver disease With confirmation of the diagnosis by liver biopsy or in patients with signs of portal hypertension, such as oesophageal varices or ascites
Renal failure Defined as need for haemodyalisis or peritoneal dialysis at the time of admission to the hospital
Severe heart failure Defined as New York Heart Association functional class III or IV heart failure
Malignancy Histological evidence required for a diagnosis of solid tumour and definitive diagnosis for the diagnosis of haematological malignancy
HIV infection Defined as HIV-positive status
Neutropenia Total leucocyte count ≤ 500/mm3
Immunosuppression Altered immune status according to APACHE II criteria [8] or in case of a previous diagnosis (congenital or acquired)
Transplant recipients Those patients receiving solid organ or bone marrow transplant
Chemotherapy Use of cytotoxic agents for the treatment of a neoplasm or an autoimmune disease within 30 days before ICU admission
Radiotherapy Radiation therapy within 30 days before ICU admission
During ICU stay
Presence and duration of catheters Urinary bladder, venous, or arterial catheter
Nutrition Enteral or parenteral nutrition
Mechanical ventilation
Dialysis
Use of steroids Patients treated with a daily dose equivalent to 20 mg prednisone
Neutropenia Total leucocyte count ≤ 500/mm3
Drug use Antimicrobial and antifungal agents
APACHE, Acute Physiology and Chronic Health Evaluation; COPD, chronic obstructive pulmonary disease; ICU, intensive care unit.
Table 2 Demographic data and risk factors for fungal infection in critically ill patients admitted to the ICU for more than 7 days
Variable Isolation of Aspergillus spp. Candida spp. infection Noncolonized, uninfected patients P
Patients (n) 36 105 720
Age (years; mean ± SD) 58.7 ± 16.6) 59.5 ± 16.3 56.4 ± 17.4 NS
Men (n [%]) 27 (75) 76 (72.4) 491 (68.2) NS
APACHE II score (mean ± SD) 21.6 ± 6.9 18.5 ± 6.5 18.9 ± 8.1 0.05
ICU stay (days; mean ± SD) 32.1 ± 21.4) 32.8 (22.6) 18.4 (14.1) <0.001
Risk factors before ICU admission (n [%])
Diabetes mellitus 5 (13.9) 18 (17.1) 113 (15.7) NS
COPD 16 (44.4) 16 (15.2) 179 (24.9) 0.002
Solid neoplasm 3 (8.3) 21 (20) 65 (9) 0.002
Hematological neoplasm 1 (2.8) 3 (2.9) 18 (2.5) NS
Transplant recipient 3 (8.3) 0 3 (0.4) <0.001
Immunosuppression 10 (27.8) 8 (7.6) 48 (6.7) <0.001
Chronic renal failure 3 (8.3) 4 (3.8) 26 (3.6) NS
HIV infection 1 (2.8) 2 (1.9) 9 (1.3) NS
Chronic liver disease 2 (5.6) 3 (2.9) 29 (4) NS
Severe heart failure 2 (5.6) 4 (3.8) 41 (5.7) NS
Radiation therapy 0 4 (3.8) 7 (1) NS
Chemotherapy 2 (5.6) 3 (2.9) 13 (1.8) NS
Risk factors during ICU stay
Arterial catheter 31 (86.1) 74 (70.4) 498 (69.1) NS
Venous catheter 36 (100) 104 (99.0) 711 (98.6) NS
Urinary catheter 36 (100) 100 (95.2) 703 (97.5) NS
Mechanical ventilation 35 (97.2) 99 (94.2) 640 (88.8) 0.019
Total parenteral nutrition 20 (55.5) 90 (85.7) 274 (38.0) <0.001
Haemodialysis 11 (30.6) 34 (32.3) 56 (7.8) <0.001
Neutropenia 5 (13.8) 5 (4.7) 20 (2.8) 0.001
Steroids 25 (69.4) 25 (23.8) 139 (19.3) <0.001
Antibiotic treatment 36 (100) 105 (100) 674 (93.5) <0.001
APACHE, Acute Physiology and Chronic Health Evaluation; COPD, chronic obstructive pulmonary disease; ICU, intensive care unit; NS, not significant; SD, standard deviation.
Table 3 Duration (days) of intra-ICU risk factors for fungal infection
Variable Isolation of Aspergillus spp. Candida spp. infection Noncolonized, uninfected patients P
Patients (n) 36 105 720
Arterial catheter 26.0 ± 19.6 25.1 ± 21.4 14.7 ± 11.7 <0.001
Venous catheter 34.0 ± 38.1 32.6 ± 22.6 17.5 ± 12.4 <0.001
Urinary catheter 35.0 ± 37.5 32.7 ± 2.7 17.8 ± 14.1 <0.001
Mechanical ventilation 26.2 ± 17.2 27.0 ± 20.0 14.6 ± 13.9 <0.001
Total parenteral nutrition 20.7 ± 13.8 21.0 ± 19.0 11.4 ± 13.9 <0.001
Haemodialysis 11.7 ± 11.1 15.7 ± 16.4 15.5 ± 13.1 NS
Neutropenia 8.6 ± 8.2 4 ± 6.1 10.8 ± 9.8 NS
Steroids 29.7 ± 45.2 16.7 ± 14.6 14.1 ± 14.3 0.003
Antibiotic treatment 10.8 ± 5.1 11.5 ± 5.5 9.2 ± 4.7 <0.001
Values are expressed as mean ± standard deviation, unless otherwise stated. ICU, intensive care unit; NS, not significant.
Table 4 Characteristics of treatment and outcome in 20 patients with Aspergillus spp. infection
Case APACHE II score at start of treatment Antifungal agent Total doses (mg) Days of treatment Microbiological eradication Clinical cure Outcome
1 18 Liposomal amphotericin B 2450 21 Yes Yes Alive
2 29 Itraconazole - 5 ? No Death in the ICU
3 17 Liposomal amphotericin B 3100 21 ? Yes Alive
4 17 Liposomal amphotericin B 2650 17 Yes Yes Death in the hospital
5 20 Liposomal amphotericin B 600 4 ? No Death in the ICU
6 16 Amphotericin B deoxycholate 450 10 No No Alive
7 ? Amphotericin B deoxycholate 1050 21 Yes Yes Alive
8 23 Amphotericin B deoxycholate 180 5 ? No Death in the ICU
9 20 Amphotericin B deoxycholate 1050 7 No No Death in the ICU
10 19 Amphotericin B lipid complex 2300 9 No No Death in the ICU
11 22 Liposomal amphotericin B 3180 21 ? No Death in the ICU
12 22 Liposomal amphotericin B 3300 15 Yes Yes Death in the ICU
13 11 Amphotericin B deoxycholate/liposomal amphotericin B 1200/1200 12/4 No No Death in the ICU
14 32 Liposomal amphotericin B 4000 16 Yes Yes Death in the ICU
15 45 Amphotericin B lipid complex 2400 8 ? Yes Alive
16 16 Amphotericin B deoxycholate/amphotericin B lipid complex 350/300 6/2 No No Death in the ICU
17 25 Amphotericin B lipid complex/liposomal amphotericin B 3600/6300 14/21 No/Yes No/Yes Death in the ICU
18 ? Amphotericin B deoxycholate 420 7 ? No Death in the ICU
19 34 Amphotericin B deoxycholate 450 7 ? No Death in the ICU
20 23 Liposomal amphotericin B 2200 11 Yes Yes Alive
APACHE, Acute Physiology and Chronic Health Evaluation; ICU, intensive care unit.
==== Refs
Denning DW Mandell G, Douglas J, Bennett D Aspergillus species Principles and Practice of Infectious Disease 2000 Philadelphia: Churchill Livingstone 2675 2685
Humphreys H Johnson EM Warnock DW Willatts SM Winter RJ Speller DC An outbreak of aspergillosis in a general ITU J Hosp Infect 1991 18 167 177 1680899 10.1016/0195-6701(91)90141-T
Chen KY Ko SC Hsueh PR Luh KT Yang PC Pulmonary fungal infection: emphasis on microbiological spectra, patient outcome, and prognostic factors Chest 2001 120 177 184 11451835 10.1378/chest.120.1.177
Clancy CJ Nguyen MH Acute community-acquired pneumonia due to Aspergillus in presumably immunocopetent host. Clues for recognition of a rare but fatal disease Chest 1998 114 629 634 9726758
Ascioglu S Rex JH de Pauw B Bennett JE Bille J Crokaert F Denning DW Donnelly JP Edwards JE Erjavec Z Defining opportunistic invasive fungal infections in immunocompromised patients with cancer and hematopoietic stem cell transplant: an international consensus Clin Infect Dis 2002 34 7 14 11731939 10.1086/323335
Denning DW Invasive aspergillosis Clin Infect Dis 1998 26 781 805 9564455
Stevens DA Kan VL Judson MA Morrison WA Dummer S Denning DW Bennett JE Walsh TJ Patterson TF Pankey GA Practice guidelines for diseases caused by Aspergillus Clin Infect Dis 2000 30 696 709 10770732 10.1086/313756
Knaus WA Draper EA Wagner DP Zimmerman JE APACHE II: a severity of disease classification system Crit Care Med 1985 13 818 829 3928249
Bone RC Balk RA Cerra FB Dellinger RP Fein AM Knaus WA Schein RM Sibbald WJ Definitions for sepsis and organ failure and guidelines for the use of innovative therapies in sepsis. The ACCP/SCCM Consensus Conference Committee. American College of Chest Physicians/Society of Critical Care Medicine Chest 1992 101 1644 1655 1303622
Logan PM Primack SL Miller RR Müller NL Invasive aspergillosis of the airways: radiographic, CT, and pathologic findings Radiology 1994 193 383 388 7972747
Crean JM Niederman MS Fein AM Feinsilver SH Rapidly progressive respiratory failure due to Aspergillus pneumonia: a complication of short-term corticosteroid therapy Crit Care Med 1992 20 148 150 1729033
Rello J Esandi ME Mariscal D Gallego M Domingo C Vallés J Invasive pulmonary aspergillosis in patients with chronic obstructive pulmonary disease: report of eight cases and review Clin Infect Dis 1998 26 1473 1475 9636889
Pittet D Huguenin T Dharan S Sztajzel-Boissard J Ducel G Thorens JB Auckenthaler R Chevrolet JC Unusual cause of lethal pulmonary aspergillosis in patients with chronic obstructive pulmonary disease Am J Resp Crit Care Med 1996 154 541 544 8756836
Bulpa PA Dive AM Garrino MG Delos MA Gonzalez MR Evrard PA Glupczynski Y Installe EJ Chronic obstructive pulmonary disease patients with invasive pulmonary aspergillosis: benefits of intensive care? Intensive Care Med 2001 27 59 67 11280674 10.1007/s001340000768
Afessa B Morales IJ Scanlon PD Peters SG Prognostic factors, clinical course, and hospital outcome of patients with chronic obstructive pulmonary disease admitted to an intensive care unit for acute respiratory failure Crit Care Med 2002 30 1610 1615 12130987 10.1097/00003246-200207000-00035
Meis JF Donnelly JP Hoogkamp-Korstanje JA De Pauw BE Aspergillus fumigatus pneumonia in neutropenic patients during therapy with fluconazole for infection due to Candida spp Clin Infect Dis 1993 16 734 735 8507771
Dimopopulos G Piagnerelli M Berre J Eddafali B Salmon I Vincet JL Disseminated aspergillosis in intensive care unit patients: an autopsy study J Chemother 2003 15 71 75 12678418 10.1159/000070766
Petri MG Konig J Moecke HP Gramm HJ Barkow H Kujath P Dennhart R Schafer H Meyer N Kalmar P Epidemiology of invasive mycosis in ICU patients: a prospective multicenter study in 435 non-neutropenic patients Intensive Care Med 1997 23 317 325 9083235 10.1007/s001340050334
Groll AH Shah PM Mentzel C Schneider M Just-Nuebling G Hubner K Trends in the postmortem epidemiology of invasive fungal infections at a university hospital J Infect 1996 33 23 32 8842991 10.1016/S0163-4453(96)92700-0
Yu VL Muder RR Poorsttar A Significance of isolation of Aspergillus from the respiratory tract in diagnosis of invasive pulmonary aspergillosis. Results from a three-year prospective study Am J Med 1986 81 249 254 3090879 10.1016/0002-9343(86)90259-7
Perfect JR Cox GM Lee JY Kauffman CA de Repentigny L Chapman SW Morrison VA Pappas P Hiemenz JW Stevens DA Mycoses Study Group The impact of culture isolation of Aspergillus species: a hospital-based survey of aspergillosis Clin Infect Dis 2001 33 1824 1833 11692293 10.1086/323900
Patterson TF Kirkpatrick WR White M Hiemenz JW Wingard JR Dupont B Rinaldi MG Stevens DA Graybill JR Invasive aspergillosis. Disease spectrum, treatment practices, and outcome Medicine 2000 79 250 260 10941354 10.1097/00005792-200007000-00006
Latge JP Aspergillus fumigatus and aspergillosis Clin Microbiol Rev 1999 12 310 350 10194462
Lederer JA Rodrick ML Mannick JA The effects of injury on the adaptive immune response Shock 1999 11 153 159 10188766
Kwoun MO Ling PR Lydon E Imrich A Qu Z Palombo J Bistrian BR Immunologic effects of acute hyperglycemia in nondiabetic rats JPEN J Parenter Enteral Nutr 1997 21 91 95 9084011
Roilides E Uhlig K Venzon D Pizzo PA Walsh TJ Prevention of corticoid-induced suppression of human polymorphonuclear leukocyte-induced damage of Aspergillus fumigatus hyphae by granulocyte colony-stimulating factor and gamma interferon Infect Immun 1993 61 4870 4877 7691757
Hartemink KJ Paul MA Spijkstra JJ Girbes AR Polderman KH Immunoparalysis as a cause for invasive aspergillosis? Intensive Care Med 2003 29 2068 2071 12768234 10.1007/s00134-003-1778-z
Meersseman W Vandeecasteele SJ Wilmer A Verbeken E Peetermans WE Wijngaerden EV Invasive aspergillosis in critically ill patients without malignancy Am J Resp Crit Care Med 2004 170 621 625 15229094 10.1164/rccm.200401-093OC
Jannsen JJWM Strack van Schijndel van der Poest Clement EH Ossenkoppele GJ Thijs LG Huijgens PC Outcome of ICU treatment in invasive aspergillosis Intensive Care Med 1996 22 1315 1322 8986479 10.1007/s001340050257
Vandewoude KH Blot SI Benoit D Colardyn F Vogelaers D Invasive aspergillosis in critically ill patients: attributable mortality and excess in length of ICU stay and ventilator dependence J Hosp Infect 2004 56 269 276 15066736 10.1016/j.jhin.2004.01.006
Wingard JR Kubilis P Lee L Yee G White M Walshe L Bowden R Anaissie E Hiemenz J Lister J Clinical significance of nephrotoxicity in patients treated with amphotericin B for suspected or proven aspergillosis Clin Infect Dis 1999 29 1402 1407 10585786 10.1086/313498
Gottfredson M Perfect JR Use of antifungal agents in the intensive care unit Curr Opin Crit Care 1999 5 381 390 10.1097/00075198-199910000-00008
Herbrecht R Denning DW Patterson TF Bennett JE Greene RE Oestmann JW Kern WV Marr KA Ribaud P Lortholary O Voriconazole versus amphotericin B for primary therapy if invasive aspergillosis N Engl J Med 2002 347 408 415 12167683 10.1056/NEJMoa020191
Maertens J Raad I Petrikkos G Boogaerts M Selleslag D Petersen FB for the Caspofungin Salvage Aspergillosis Study Group Efficacy and safety of caspofungin for treatment of invasive aspergillosis in patients refractory to or intolerant of conventional antifungal therapy Clin Infect Dis 2004 39 1563 1571 15578352 10.1086/423381
| 15987390 | PMC1175876 | CC BY | 2021-01-04 16:04:52 | no | Crit Care. 2005 Mar 11; 9(3):R191-R199 | utf-8 | Crit Care | 2,005 | 10.1186/cc3488 | oa_comm |
==== Front
Crit CareCritical Care1364-85351466-609XBioMed Central London cc34951598739110.1186/cc3495ResearchDecreased duration of mechanical ventilation when comparing analgesia-based sedation using remifentanil with standard hypnotic-based sedation for up to 10 days in intensive care unit patients: a randomised trial [ISRCTN47583497] Breen Des [email protected] Andreas [email protected] Manu [email protected] Rex [email protected] Sven [email protected] Inge-Lise [email protected] Pauline [email protected] Andrew JT [email protected] Consultant in Anaesthesia and Intensive Care, Royal Hallamshire Hospital, Sheffield, UK2 Director of Intensive Care Unit, Genimatas General Hospital, Athens, Greece3 Director of Intensive Care Unit, ZiekenhuisNetwerk Antwerpen, Antwerpen, Belgium4 Consultant Anaesthetist, Intensive Care Unit, Dubai Hospital, Dubai, United Arab Emirates5 Deputy Director, Universität Erlangen-Nürnberg, Klinik für Anästhesiologie, Erlangen, Germany6 Senior Registrar, Intensive Care Unit, Righospitalet, Copenhagan, Denmark7 Clinical Scientist, Neurosciences Medicines Development Centre, GlaxoSmithKline, Greenford, Middlesex, UK8 Clinical Development Director, Neurosciences Medicines Development Centre, GlaxoSmithKline, Greenford, Middlesex, UK2005 15 3 2005 9 3 R200 R210 23 12 2004 9 2 2005 Copyright © 2005 Breen 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
This randomised, open-label, multicentre study compared the safety and efficacy of an analgesia-based sedation regime using remifentanil with a conventional hypnotic-based sedation regime in critically ill patients requiring prolonged mechanical ventilation for up to 10 days.
Methods
One hundred and five randomised patients received either a remifentanil-based sedation regime (initial dose 6 to 9 μg kg-1 h-1 (0.1 to 0.15 μg kg-1 min-1) titrated to response before the addition of midazolam for further sedation (n = 57), or a midazolam-based sedation regime with fentanyl or morphine added for analgesia (n = 48). Patients were sedated to an optimal Sedation–Agitation Scale (SAS) score of 3 or 4 and a pain intensity (PI) score of 1 or 2.
Results
The remifentanil-based sedation regime significantly reduced the duration of mechanical ventilation by more than 2 days (53.5 hours, P = 0.033), and significantly reduced the time from the start of the weaning process to extubation by more than 1 day (26.6 hours, P < 0.001). There was a trend towards shortening the stay in the intensive care unit (ICU) by 1 day. The median time of optimal SAS and PI was the same in both groups. There was a significant difference in the median time to offset of pharmacodynamic effects when discontinuing study medication in patients not extubated at 10 days (remifentanil 0.250 hour, comparator 1.167 hours; P < 0.001). Of the patients treated with remifentanil, 26% did not receive any midazolam during the study. In those patients that did receive midazolam, the use of remifentanil considerably reduced the total dose of midazolam required. Between days 3 and 10 the weighted mean infusion rate of remifentanil remained constant with no evidence of accumulation or of a development of tolerance to remifentanil. There was no difference between the groups in SAS or PI score in the 24 hours after stopping the study medication. Remifentanil was well tolerated.
Conclusion
Analgesia-based sedation with remifentanil was well tolerated; it reduces the duration of mechanical ventilation and improves the weaning process compared with standard hypnotic-based sedation regimes in ICU patients requiring long-term ventilation for up to 10 days.
See related commentary
==== Body
Introduction
Most patients who require intensive care need effective analgesia and sedation to control potentially unpleasant symptoms such as pain and anxiety. Analgesics and sedatives are also used to allow patients to tolerate nursing procedures and tracheal intubation and to aid in mechanical ventilation. Most commonly, the combined use of a sedative agent with an opioid is used to achieve this. The sedative agent is titrated towards the degree of sedation and opioids are added as judged necessary for pain control.
The pharmacodynamic effects of traditionally used sedative and opioid drugs are unpredictable and often prolonged in the critically ill patient for various reasons. The pharmacokinetics are altered with different volumes of distribution and elimination half-lives. Multiple drugs are given in large doses for prolonged periods, which can lead to drug interactions and tolerance. There is altered organ function with impaired renal/hepatic function, altered regional blood flow, protein and enzyme system dysfunction and altered receptor response. In addition, the most commonly used sedatives and opioids have a context-sensitive half-time that increases with time. Thus, these drugs will accumulate during prolonged use. All opioids have sedative properties to various degrees at high doses. However, traditionally the opioid part of a sedation regime is kept to a minimum to protect against opioid accumulation and an unpredictable recovery/weaning from mechanical ventilation.
Remifentanil hydrochloride is a potent, selective μ-opioid receptor agonist that has a rapid onset of action (about 1 min) and quickly achieves steady state [1]. Remifentanil is metabolised by non-specific plasma esterases and is therefore independent of organ function [2]. Remifentanil is rapidly metabolised and has a context-sensitive half-time of about 2 to 3 min that is independent of duration of infusion [3]. These features of remifentanil make it an ideal agent for use in critically ill patients. It is easy to titrate and can be given in relatively high doses for prolonged periods without risk of accumulation [4,5] or delayed offset of effects [5]. It allows the opioid to be used as the main drug to provide patient comfort with the sedative agent being kept to a minimum. Remifentanil as part of an analgesic-based sedative regime (analgo-sedation) has been studied in critically ill patients for up to 5 days [4-17].
This present study was conducted to assess the efficacy and safety of a prolonged infusion of remifentanil in critically ill patients for up to 10 days in comparison with a standard sedative regime of midazolam plus a traditional opioid. The efficacy of remifentanil was assessed by the primary endpoint of time from starting the study drug until time of extubation. The safety profile of remifentanil was assessed by monitoring haemodynamic parameters and recording adverse events throughout the study period.
Materials and methods
This study was a randomised, open-label, multicentre, parallel-group study comparing the safety and efficacy of an analgesia-based regime using remifentanil with a conventional hypnotic-based regime in critically ill patients requiring mechanical ventilation for 3 to 10 days. The study was conducted in accordance with good clinical practice and within the guidelines set out in the Declaration of Helsinki. Informed consent/assent was obtained from all patients or their representatives. After approval from local and national ethics committees, 105 patients from 15 centres in 10 countries were recruited. Patients were randomised in a 1:1 ratio to receive either a remifentanil-based regime or a comparator hypnotic-based regime using midazolam with either morphine or fentanyl for analgesia.
Inclusion and exclusion criteria
The target population were those patients requiring long-term mechanical ventilation for medical reasons. Post-surgical patients requiring extended mechanical ventilation as a result of post-surgical complications were also included. Patients were eligible if they were more than 18 years old, had been admitted to the intensive care unit (ICU) within the previous 30 hours, were expected to require mechanical ventilation for longer than 96 hours and required analgesia and sedation. Females were eligible to enter the study if they were of non-childbearing potential or had a negative pregnancy test at screening and agreed not to fall pregnant for 12 days after stopping the study drug.
Patients were excluded from the study if their medical condition prevented assessment of depth of sedation, if it required the frequent down-titration of analgesics/sedatives for assessment, if it was likely to require surgery or tracheostomy during the treatment period, if it required neuromuscular blocking drugs by infusion, if it required epidural blockade, if it required sedatives or anaesthetic agents other than study drugs specified in the treatment period, or if there was a contraindication to the use of remifentanil, morphine, fentanyl or midazolam. Other exclusions were sensitivity to the drugs or class of drugs specified in the study, a history of alcohol or drug abuse, a concurrent or previous entry into this or other investigational drug studies within 30 days, or pregnancy or lactation. Protocol-specified treatment regimes had to be appropriate for the management of the patients. After 30 patients had entered the study, a protocol amendment allowed the inclusion of patients who had been receiving mechanical ventilation for up to 30 hours irrespective of the time in the ICU, allowed the inclusion of patients requiring surgery of less than 6 hours' duration during the treatment period and reduced the required duration of mechanical ventilation from 96 to 72 hours.
Study period
The study was divided into four periods: screening, treatment, post-treatment and follow-up.
Screening period
The screening period was from ICU admission to the start of the study drug and included the time for considering eligibility, obtaining consent/assent, randomisation and assessment of the patient's SAPS II score. Baseline demographics, physiological variables, Sedation–Agitation Scale (SAS) score and pain intensity (PI) score were also assessed during this time. The SAS is a seven-point scoring system, and a SAS score of 3 or 4 was defined as optimal sedation in this study (see Additional file 1, [18]). The PI score is a six-point score where 1 or 2 represents no pain or mild pain (see Additional file 2). Baseline liver function tests and creatinine clearance were also measured.
Treatment period
The treatment period was from the start of the study drug to permanent discontinuation of the study drug, or after 10 days of administration, or death, whichever was the sooner. SAS, PI, heart rate (HR) and mean arterial pressure (MAP) were continuously monitored throughout the treatment period and were recorded at the time of each bolus dose and/or change in infusion rate of any of the study drugs. These parameters were recorded again when optimal sedation and pain control had been established or re-established. These variables were also recorded at least every 8 hours in the event of no change in study drug. Study drugs and amount administered were also recorded during the treatment period. Liver function tests and creatinine clearance were assessed daily. Before the start of administration of the study drugs the existing sedative/analgesic regime was discontinued. Patients were then sedated to an optimal SAS and PI score by a remifentanil-based regime or a hypnotic-based regime.
Remifentanil-based regime
The remifentanil infusion was started at 6 to 9 μg kg-1 h-1 (0.1 to 0.15 μg kg-1 min-1). The remifentanil infusion was titrated in 1.5 μg kg-1 h-1 (0.025 μg kg-1 min-1) increments at 5 to 10 min intervals to achieve an optimum level of sedation/analgesia based on clinical judgement. Bolus doses of remifentanil were not permitted. Once the remifentanil infusion reached a rate of 12 μg kg-1 h-1 (0.2 μg kg-1 min-1), boluses of midazolam (not more than 2 mg) could be used if required after clinical assessment. Remifentanil was not used as the sole agent for sedation at infusion rates greater than 18 μg kg-1 h-1 (0.3 μg kg-1 min-1). Above this rate, midazolam boluses were used. Further increases in the remifentanil rate were allowed for the treatment of pain and in anticipation of short stimulating procedures, up to a maximum rate of 45 μg kg-1 h-1 (0.75 μg kg-1 min-1). The remifentanil dosing regime is depicted in Fig. 1[5,10,11].
Hypnotic-based regime
Midazolam was used by infusion and/or boluses as the sedative agent, and was titrated to an optimum level of sedation based on clinical judgement and in accordance with standard clinical protocols. Either morphine or fentanyl was used as the analgesic agent, titrated to obtain adequate pain control. The initial dose and subsequent adjustments of sedative and analgesic agents were at the investigators' discretion and in accordance with routine clinical practice to obtain an optimum SAS and PI score.
Weaning and extubation
The decision to begin the weaning process was based on clinical judgement and was defined as the time point at which the investigator first adjusted the study drug infusion rate or decided not to give any more boluses of the study drugs so as to encourage spontaneous respiration with the result of extubating the patient. For those patients extubated within 10 days, the study drugs were down-titrated in accordance with clinical judgement until a decision was made to extubate the patient. For patients in the comparator hypnotic-based treatment group this was performed in accordance with routine clinical practice at the investigational site. As a guide, for patients in the remifentanil group who were eligible for extubation, the remifentanil infusion was decreased to 6 μg kg-1 h-1 (0.1 μg kg-1 min-1) either immediately or in increments at the investigator's discretion, and no further midazolam boluses were given. Remifentanil was discontinued after extubation and, if necessary, suitable alternative methods of pain relief were instituted.
For patients not extubated within 10 days, the study drugs were discontinued in both groups and the time to offset of pharmacodynamic effects was recorded. As soon as there was a demonstrable change in haemodynamic variables, SAS score or PI score, alternative sedation and analgesic regimes were instituted as soon as clinically indicated.
Post-treatment period
The post-treatment period was from the end of the treatment period until 24 hours later. MAP, HR, SAS and PI were recorded at 15 min intervals for the first 2 hours, hourly for the next 4 hours then 6-hourly until the end of the post-treatment period.
Follow-up period
The follow-up period was from the end of the post-treatment period until 6 days later.
Study endpoints
Efficacy
The primary endpoint was the time from the start of study drug to extubation. Secondary endpoints were the time from start of study drug until start of weaning, the time from start of weaning until extubation, the time from start of study drug to ICU discharge, descriptive PI and SAS during the treatment and post-treatment periods, total exposure to study drugs and concomitant sedative requirements.
Safety
The safety endpoints were the offset of pharmacodynamic effects of study drugs after permanent discontinuation, haemodynamic effects, clinical adverse events and the requirement for re-intubation. Haemodynamic variables were monitored continuously throughout the study and recorded at the times stated above. Adverse events were recorded from the start of the study drug until the end of the post-treatment period. Serious adverse events were defined as adverse events that resulted in any of the following outcomes: death, life-threatening event, prolongation of hospitalisation, or a disability or incapacity. Important medical events that did not result in death or were not life-threatening were considered serious adverse events when, on the basis of appropriate medical judgement, they jeopardised the patient and required medical or surgical intervention to prevent one of the outcomes listed above. In addition, serious adverse events possibly attributable to study medication were recorded throughout the 6-day follow-up period.
Statistical analysis
The time to event endpoints were analysed with the generalised Wilcoxon test with a two-sided α level of 5% judged to indicate a statistically significant difference between the treatment groups. The data for patients who did not experience the event were censored in accordance with predetermined rules. The results of these analyses were summarised by using 75th centiles, difference between 75th centiles and its 95% confidence interval because too few patients achieved each event to allow estimates based on median times to be determined with any precision. The confidence intervals were calculated with methods described by Collett [19].
The percentage time of optimal analgesia/sedation was calculated and summarised by the median in each treatment group and the median of all possible differences between the groups and the 95% confidence interval around that difference to give the best estimate of median difference. Treatments were compared by using the Wilcoxon rank sum test.
With the exception of the incidence of re-intubation, no formal statistical analyses were performed on the demographic, baseline or safety data. These data were summarised either by means and standard deviations (SD) or by frequency tables as appropriate to the data.
Results
Fifty-seven patients were randomised to receive remifentanil and 48 patients to receive comparator. Of the comparator group, 62% (n = 30) received midazolam with fentanyl, 15% (n = 7) received midazolam with morphine and 23% (n = 11) received midazolam alone.
Patient demographics and baseline characteristics are shown in Table 1. The two treatment groups were well matched in terms of patient characteristics and baseline clinical assessments. The time from ICU entry to the start of study medication was slightly longer in the remifentanil group (remifentanil, mean 23.6 hours, median 14.5; comparator, mean 18.9 hours, median 16.9).
Efficacy
Efficacy endpoints are shown in Table 2. Fewer than 50% of patients were extubated during the 10-day treatment period (45 of 105). The 75th centile has been reported for the efficacy endpoints. This analysis records the time at which 25% of all patients reached the assessment points. There was no difference in the time to the start of the weaning process. There was a statistical and clinically significant difference between the two groups in the study's primary endpoint of time of starting the drug to extubation. A Kaplan–Meier plot analysing the duration of mechanical ventilation is shown in Fig. 2. The time difference was 53.5 hours, being shorter in the remifentanil group (P = 0.033). The time from the start of the weaning process to extubation was also significantly different at 26.6 hours, also in favour of remifentanil (P < 0.001).
The median percentage time of optimal analgesia/sedation was comparable for both groups (remifentanil 96.9%, comparator 97.8%, median difference -0.3, 95% confidence interval -2.7 to 0.2; P = 0.16).
There were 16 patients in each group who survived to 10 days but were not extubated, for whom the time to offset of pharmacodynamic effects on discontinuing the study drugs was assessed on 15 patients receiving remifentanil and 14 patients receiving comparator. This was found to be clinically and significantly faster in the remifentanil group (Fig. 3).
Safety
The incidence of adverse events and the most commonly occurring adverse events (5% or more) is illustrated in Table 3. The profile was similar for the two groups. Liver function tests and creatinine clearance levels were comparable in both groups at baseline and throughout the treatment period. One drug-related serious adverse event was reported. This patient was randomised to receive midazolam and fentanyl. At 1 day after starting study medication the patient developed severe hypotension, which was considered life threatening. There were no reports of muscle rigidity. Comparable MAP and HR values were observed during the post-treatment period.
Exposure to study drugs
The mean total duration, dose and weighted mean infusion rates of study opioids for all the patients treated in this study are illustrated in Table 4. Patients in the remifentanil group received a longer duration of opioid infusion. Of the patients treated with remifentanil, 26% (15 of 57) did not receive any midazolam during the study. The remaining patients received 1 to 100 boluses during the treatment period. Figure 4 illustrates the mean total midazolam dose in patients receiving opioids and no opioids. There was nearly a ninefold difference in mean total midazolam requirements in the fentanyl group compared with the remifentanil group, and a fourfold difference in the morphine group compared with the remifentanil group.
Figure 5 shows the weighted mean infusion rate of opioid by day. Remifentanil infusion requirements rose within the first few days then reached a plateau. Figure 6 represents the mean total daily midazolam requirements in combination with each opioid used. The midazolam requirements with remifentanil were the least and did not change significantly with time.
SAS and PI scores during the post-treatment period
The SAS and PI scores were the same in the remifentanil and comparator groups throughout the post-treatment period. There was no variation over the 24-hour period in either group. The variation in the mean SAS over 24 hours was 3.1 to 3.3 for remifentanil group and 2.7 to 3.0 in the comparator group. The variation in the mean PI over 24 hours was 1.5 to 1.6 for the remifentanil group and 1.5 to 1.7 for the comparator group.
Discussion
The technique of using remifentanil as the primary sedative and analgesic, with the addition of traditional sedatives such as propofol or midazolam only if necessary, has been studied in ICU patients for up to 3 days [5,11,17] and in neurosurgical patients studied for up to 5 days, with good results [12]. Anecdotally, remifentanil has been used in the ICU population for much longer periods [6]. This is the first study to look at the use of remifentanil for prolonged infusions in ICU patients for up to 10 days. The primary aim of the study was to compare the safety of the techniques and the duration for which patients were on mechanical ventilation.
Efficacy
The present study has clearly demonstrated a clinically and statistically significant decrease in duration of mechanical ventilation when using remifentanil-based analgesia and sedation. The difference between remifentanil and the comparator group was more than 2 days (53.5 hours). The time from starting the weaning process to extubation was also significantly different by more than 1 day (26.6 hours). Reducing the duration of mechanical ventilation by a matter of days will potentially help to reduce the complications associated with prolonged intubation and ventilation. These differences cannot be explained by differing eligibilities to start the weaning process, because there was no difference between the times to start the weaning process in either of the two study groups. The SAS and PI scores were the same in both groups from baseline through the treatment period and to the end of the post-treatment period. The difference therefore cannot be explained by different sedation levels between the groups.
There are difficulties of comparing extubation times between sedative regimes in an ICU study. Studies have looked at extubation times after anaesthesia, comparing various drugs [20,21]. The time frame for these studies is much shorter, and the times of decisions to stop study drug(s) and extubate patients are much easier to define. It is possible that a critically ill patient will undergo several increases and decreases in sedative and analgesic agents before the decision is made to wean and extubate. It is possible that patients will have been on no drug for some time. Given these points, the difference in the primary endpoint of the study in a small group of patients is remarkable.
There was a trend towards a shorter ICU stay in the remifentanil group by 1 day although this was not statistically significant. This may be because the numbers of patients who were actually extubated within the study period were small and the study was not sufficiently powered to detect the difference. Whereas discharge from an ICU setting is multifactorial and often depends on factors not related to the medical condition of the patient such as bed shortages on the ward and the time of day. The trend in reduced duration of ICU stay is supported by the work of Matthey and colleagues [22], who showed that remifentanil supplemented with propofol significantly reduced the time on mechanical ventilation and allowed earlier discharge from the ICU than an analgesia-based sedation with fentanyl/midazolam. In another study comparing remifentanil/midazolam with morphine/midazolam in a similar way to this study, there was a significant difference between the extubation times and ICU discharge times between the two groups [17]. However, the duration of mechanical ventilation was relatively short.
The context-sensitive half-time of remifentanil is constant and independent of duration of drug administration when administered for up to 6 hours [3], but does this situation change when remifentanil is given for days rather than hours? One study looked at the pharmacodynamic offset of remifentanil at various time points over 3 days in ICU patients and found that the offset time was constant over this period [5]. In the present study there was no evidence of accumulation of remifentanil over time. Of the patients who were not extubated within 10 days, on discontinuing the remifentanil infusion the mean time to offset of pharmacodynamic effect was only 15 min and was identical to that obtained in the study above [5]. A recent paper has shown that if sedation in ICU patients was stopped on a daily basis, it considerably reduced the duration of mechanical ventilation and ICU stay [23]. However, these patients were heavily sedated. This study and the above study demonstrates that if patients are sedated to an optimum SAS and PI score with remifentanil, then despite the duration of sedation the pharmacodynamic offset time is about 15 min, making a daily 'wake-up call' unnecessary for the reasons of decreasing the amount of total sedation and aiding extubation and ICU discharge.
The duration for which patients had an optimum SAS and PI score was the same in each group and was more than 96% of the duration for which patients were receiving treatment. This is to be expected, because the aim of the study was to titrate the analgesic/sedative regimes to achieve target SAS and PI scores that were optimal. Thus, the only effect on the primary endpoint was the drugs used and not the variation in sedation/analgesic levels.
Safety
The reported incidence of adverse effects in this study was the same in the remifentanil and comparator groups. The incidence was also similar to those in other studies looking at the safety of remifentanil and comparator agents in ICU patients [5,11,12]. The adverse events were also in keeping with events that one would expect in an ICU population. The only serious adverse event that was reported as drug-related occurred in the comparator group. This was of severe life-threatening hypotension. Deaths were the same in each group and at the expected rate for an ICU population. No deaths were reported as a result of the study drug. There were no reports of muscle rigidity or of prolonged μ-opioid effects as a consequence of using remifentanil. There was no statistical difference in the incidence of re-intubation between the two treatment groups.
There was no evidence of adverse SAS or PI scores on discontinuing the remifentanil. The mean SAS and PI scores were the same in each group and stable for the 24-hour post-treatment period. Investigators were able to use the regime of their choice to optimise PI and SAS after extubation, in line with routine clinical practice. There was no evidence that remifentanil sensitised opioid receptors or that the rapid offset of pharmacodynamic effects of remifentanil caused problems with control of pain after extubation.
Exposure to study drugs
In the comparator group the weighted mean infusion rates of fentanyl and morphine were relatively constant throughout the study. The large increase in the morphine requirements on days 8 and 9 represented one patient only.
The weighted mean remifentanil infusion rate to maintain optimum SAS and PI scores was 19.2 μg kg-1 h-1 (0.32 μg kg-1 min-1) (Table 4). This is slightly higher than reported previously, but well within the infusion rates recommended [5,11]. The remifentanil use remained relatively constant throughout the study, and generally the remifentanil was given for longer than the other opioids (Fig. 5, Table 4).
The midazolam requirements were considerably reduced in the remifentanil group and were relatively constant throughout the treatment period (Figs 4 and 6). This is a reflection of the hypnotic-agent-sparing effects of remifentanil and the ease with which the infusion can be titrated to obtain optimum patient comfort [11,17].
In contrast, the mean total daily requirements for midazolam in the comparator group varied considerably throughout the study period to maintain an optimal SAS for 97% of the time. There was an up to 10-fold difference in mean daily midazolam requirements in the fentanyl subgroup (226 mg on day 4; 26 mg on day 10) and a sevenfold difference in the morphine subgroup (189.6 mg on day 5; 28.4 mg on day 10). This accounts for the difference in mean total midazolam requirements between the groups. The requirements for midazolam peaked between days 3 and 5 in the comparator groups, and then tailed off (Fig. 6). The above observations are a reflection of two factors. First, midazolam was used as the primary hypnotic agent in the comparator group and it was therefore adjusted first. Second, although the opioid infusion rates remained relatively constant, as the study progressed, opioid and midazolam accumulation occurred. The accumulation, with large body stores of these drugs, contributed to the sedation in these patients. Therefore the requirement for midazolam administration to maintain a constant SAS and PI score tailed off after a certain period. The context-sensitive half-time of the comparator agents is known to increase with time, and the comparator agent's elimination is organ dependent [1,24].
There was no clinical evidence of the development of tolerance to remifentanil as demonstrated by escalating remifentanil requirements or post-infusion opioid requirements. The weighted mean infusion rate of remifentanil increased slightly until day 3, and was then constant until day 10. The SAS and PI scores at the start and throughout the post-treatment period were the same as in the comparator, and were constant.
There was a group of patients who received only midazolam to provide patient comfort during their stay in the ICU. This occurred at one hospital where this was the preferred practice for the treatment of medical patients without injuries and reflected normal clinical practice at that site.
The present study compares two different sedation techniques. One technique used the analgesic component of the regime as the main variant for sedation (remifentanil). The other used the hypnotic component as the main variant for sedation (midazolam). Although this second group gave rise to three subgroups because either fentanyl, morphine or no analgesic was used, this was unimportant. The main aim was to compare the remifentanil regime with common 'standard practice' for sedation in ICU patients. It would not have been representative of clinical practice to compare two opioids in an analgesia-based regime: the commonly used opioids cannot be used in this way because of the real risk of accumulation. The present study clearly demonstrates that the remifentanil-based regime is superior in terms of reduced time for weaning and, more importantly, reduced duration of mechanical ventilation.
Conclusion
The use of remifentanil in an analgesic-based sedative regime in critically ill patients significantly decreases the duration of mechanical ventilation and of weaning. It is sedative sparing and has a very rapid offset even after a 10-day infusion, with no evidence of accumulation. Remifentanil was well tolerated throughout a 10-day infusion. The adverse event profile was similar in remifentanil-based and hypnotic-based regimes. No adverse events relating to muscle rigidity or prolonged μ-opioid effects were reported. The SAS and PI scores after treatment were comparable. There was no evidence of the development of tolerance to remifentanil and there was no difficulty in maintaining optimal SAS and PI scores after treatment with remifentanil with the use of standard treatment regimes.
Key messages
• The use of remifentanil-based analgesia and sedation significantly reduced the duration of mechanical ventilation.
• Weaning patients from mechanical ventilation can be achieved significantly faster with remifentanil-based analgesia and sedation.
• Remifentanil has a very rapid offset even after a 10-day infusion with no evidence of accumulation.
• There was no evidence of the development of tolerance to remifentanil and there was no increase in opioid requirements after treatment with remifentanil, even after prolonged use.
• A remifentanil-based analgesia and sedation regimen is well tolerated when used for up to 10 days in critically ill patients requiring mechanical ventilation and has a safety profile similar to that observed for hypnotic-based sedation.
Abbreviations
HR = heart rate; ICU = intensive care unit; MAP = mean arterial pressure; PI = pain intensity; SAS = Sedation–Agitation Scale.
Competing interests
DB has no competing interests. AK, MM, RM, SA and I-LJ received payment from GlaxoSmithKline (either personally or to their respective department) depending on the number of patients recruited. PP and AJTK are employees of GlaxoSmithKline.
Authors' contributions
DB made substantial contributions to the conception, design and interpretation of the data collected in this study, and drafted the manuscript. AK, MM, RM SA and I-LJ performed the study and provided critical review of the manuscript. PP coordinated the development and conduct of the study. AJTK contributed to the design of the study and the interpretation of the data and provided critical review of the manuscript. All authors read and approved the manuscript.
Supplementary Material
Additional File 1
A Word file showing the definitions of the scores on the Sedation–Agitation Scale.
Click here for file
Additional File 2
A Word file showing the definitions of pain intensity scores.
Click here for file
Acknowledgements
We acknowledge the contribution of the following to the conduct of the study: in Austria, Professor P Germann (Department of Anaesthesiology and Intensive Medicine, Vienna); in Belgium, Dr M Genard (Hôpital Ambroise Paré, Mons); in Denmark, Dr T Jensen (Rigshospitalet Intensiv terapiafsnit, Blegdamsvej) and Dr B Rasmussen (TV-sektion og Sektion, Aborg); in France, Dr F Lagneau (Beaujon Hospital Service, Clichy) and Professor A Mebazaa (Lariboisière hospital Service, Paris); in Greece, Dr S Stergiopoulos (General Hospital of Nikaia, Athens); in Iran, Dr Giamat (MPO Red Cross, Tehran); in The Netherlands, Dr J Tulleken (AZG Groningen, Groningen) and Dr J Bakker (Isala Klinieken Zwolle, Zwolle); in Portugal, Dr I Miranda (Hospital Sto. António dos Capuchos, Lisbon). Our thanks also go to Steven A Julious, GlaxoSmithKline, for providing statistical support for this study.
Figures and Tables
Figure 1 The dosing algorithm.
Figure 2 Kaplan–Meier survival plot of time to extubation (days).
Figure 3 Median time to offset of effects as measured by the time to therapeutic intervention.
Figure 4 Mean total midazolam dose.
Figure 5 Mean weighted mean opioid infusion over time.
Figure 6 Mean total daily dose of midazolam with concomitant opioids.
Table 1 Patient characteristics and baseline clinical assessments
Characteristic Remifentanil Comparator
Number of patients treated 57 48
Medical (%)/post-surgical (%) 49 (88)/7 (13) 44 (92)/4 (8)
Emergency (%)/elective (%) 27 (84)/5 (16), n = 32 21 (91)/2 (9), n = 23
Age (years) 52.2 ± 18.4 57.3 ± 18.1
Male (%)/female (%) 39 (68)/18 (32) 32 (67)/16 (33)
Height (cm) 171.2 ± 9.7 169.0 ± 7.9
Weight (kg) 78.6 ± 13.41 76.3 ± 15.50
SAPS II on admission 43.0 ± 15.6 43.3 ± 11.2
MAP (mmHg)a 88.8 ± 16.5 88.9 ± 14.8
Heart rate (b.p.m.)a 98.9 ± 20.1 95.9 ± 15.5
SAS scorea 3.3 ± 1.3 3.3 ± 1.4
PI scorea 2.0 ± 1.2 2.1 ± 1.1
aBaseline values. Where errors are given, results are means± SD. MAP, mean arterial pressure; PI, pain intensity; SAS, Sedation–Agitation Scale.
Table 2 Study endpoints
Characteristic Remifentanil (n = 57) Comparator (n = 48) P
Number (%) of patients extubated 29 (51%) 16 (33%)
Time from start of study drugs to weaning (h) 83.0 98.0 0.523
Difference (95% CI) -15.0 (-61.8 to 31.8)
Time from start of study drugs to extubation(h) 94.0 147.5 0.033
Difference (95% CI) -53.5 (-111.4 to 4.4)
Time from weaning time until extubation (h) 0.9 27.5 <0.001
Difference (95% CI) -26.6 (-40.8 to -12.4)
Time from start of study drugs until ICU discharge (h) 187.3 209.8 0.326
Difference (95% CI) -22.5 (-201.5 to 156.5)
Point estimates are 75th centiles. CI, confidence interval.
Table 3 Adverse event profile
Characteristic Number of patients (%)
Remifentanil (n = 57) Comparator (n = 48)
Any adverse event 19 (33) 16 (33)
Any drug-related adverse event 6 (11) 4 (8)
Any serious adverse event 7 (12) 6 (13)
Any drug-related serious adverse event 0 (0) 1 (2)
Premature discontinuation from the study 12 (21) 10 (21)
Deaths 7 (13) 5 (10)
Re-intubated within 10 daysa 7 (25) 2 (12)
Most commonly occurring adverse events (≥ 5%)
Hypotension 3 (5) 4 (8)
Atrial fibrillation 4 (7) 2 (4)
Vomiting 3 (5) 0 (0)
Septic shock 0 (0) 3 (6)
aP value for continuity-corrected χ2. 95% confidence interval -4 to 23. P = 0.193. Time to re-intubation ranged from 2 h to 3 days after stopping remifentanil, and 7 hours to 3 days after stopping comparator agent.
Table 4 Exposure to study drug for all patients treated
Parameter Remifentanil (n = 57) Fentanyl (n = 30) Morphine (n = 7)
Mean duration of infusion (h) 147.2 126.4 120.5
Mean total opioid dose 221,614 μg 20,702 μg 237.5 mg
Mean weighted mean opioid infusion rate(SD) 19.2a (12.2) μg kg-1 h-1 3.0 (3.35) μg kg-1 h-1 0.042 (0.028) mg kg-1 h-1
aEquivalent to 0.32 μg kg-1 min-1.
==== Refs
Egan TD Lemmens HJ Fiset P Hermann DJ Muir KT Stanski DR Shafer SL The pharmacokinetics of the new short-acting opioid remifentanil (GI87084B) in healthy adult male volunteers Anesthesiology 1993 79 881 892 7902032
Westmorland CL Hoke JF Sebel PS Hugg CC JrMuir KT Pharmacokinetics of remifentanil (G187084B) and its major metabolite (GR90291) in patients undergoing elective surgery Anesthesiology 1993 79 893 903 7902033
Kapila A Glass PS Jacobs JR Muir KT Hermann DJ Shiraishi M Howell S Smith RL Measured context-sensitive half-times of remifentanil and alfentanil Anesthesiology 1995 83 968 975 7486182 10.1097/00000542-199511000-00009
Pitsu M Wilmer A Bodenham A Breen D Back V Bonde J Kessler P Fisher G Kirkham A Pharmacokinetics of remifentanil and its major metabolite, remifentanil acid, in ICU patients with renal impairment Br J Anaesth 2004 92 493 503 14766712 10.1093/bja/aeh086
Breen D Wilmer A Bodenham A Bach V Bonde J Kessler P Albrecht S Shaikh S Offset of pharmacodynamic effects and safety of remifentanil in intensive care unit patients with various degrees of renal impairment Crit Care 2004 8 R21 R30 14975051 10.1186/cc2399
Park GR Evans TN Remifentanil in the critically ill: what will its place be? Br J Intensive Care 1996 79 893 903
Main A Remifentanil as an analgesic in the critically ill Anaesthesia 1998 53 823 824 9797530 10.1046/j.1365-2044.1998.0584a.x
Wilhelm W Dorscheid E Schlaich N Niederprum P Deller D The use of remifentanil in critically ill patients. Clinical findings and early experience Anaesthesist 1999 48 625 629 10525595 10.1007/s001010050762
Tipps LB Coplin WM Murry KR Rhoney DH Safety and feasibility of continuous infusion of remifentanil in the neurosurgical intensive care unit Neurosurgery 2000 46 596 601 10719856 10.1097/00006123-200003000-00015
Kirkham A Fisher G Kessler P A dosing algorithm for the use of remifentanil in providing optimal sedation and analgesia in ICU patients Intensive Care Med 2001 S238
Muellejans B Lopez A Cross MH Bonome C Morrison L Kirkham AJT Remifentanil versus fentanyl for analgesia based sedation to provide patient comfort in the intensive care unit: a randomised control trial [ISRCTN43755713] Crit Care 2004 8 R1 R11 14975049 10.1186/cc2398
Karabinis A Mandragos K Stergiopoulos S Komnos A Soukup J Speelberg B Kirkham AJT Safety and efficacy of analgesia-based sedation using remifentanil versus standard hypnotic-based regimens in intensive care unit patients with brain injuries: a randomised, controlled trial [ISRCTN50308308] Crit Care 2004 8 R268 R280 15312228 10.1186/cc2896
Soltēsz S Biedler A Silomon M Schopflin I Molter GP Recovery after remifentanil and sufentanil for analgesia and sedation of mechanically ventilated patients after trauma or major surgery Br J Anaesth 2001 6 763 768 10.1093/bja/86.6.763
Park G Improving sedation and analgesia in the critically ill Minerva Anestesiol 2002 68 505 512 12105406
De Bellis P Gerbi G Pacigalupo P Buscaglia G Massobrio B Montagnani L Servirei L Experience with remifentanil in the intensive care unit Minerva Anestesiol 2002 68 765 773 12496723
Cavaliere F Antonelli M Arcangeli A Conti G Costa R Pennisi MA Proietti R A low-dose remifentanil infusion is well tolerated for sedation in mechanically ventilated, critically ill patients Can J Anaesth 2002 49 1088 1094 12477685
Dahaba AA Grabner T Rehak PH List WF Metzler H Remifentanil versus morphine analgesia and sedation for mechanically ventilated critically ill patients: a randomised double blind study Anesthesiology 2004 101 640 646 15329588 10.1097/00000542-200409000-00012
Riker RR Picard JT Fraser GL Prospective evaluation of the Sedation–Agitation Scale for adult critically ill patients Crit Care Med 1999 27 1325 1329 10446827 10.1097/00003246-199907000-00022
Collett D Modelling survival date in medical research 1994 London: Chapman & Hall
Wilhelm W Schlaich N Harrer J Kleinschmidt S Muller M Larsen R Recovery and neurological examination after remifentanil-desflurane or fentanyl-desflurane anaesthesia for carotid artery surgery Br J Anaesth 2001 86 44 49 11575408 10.1093/bja/86.1.44
Breslin DS Reid JE Mirakhur RK Hayes AH McBrien ME Sevoflurane – nitrous oxide anaesthesia supplemented with remifentanil: effect on recovery and cognitive function Anaesthesia 2001 56 114 119 11167470 10.1046/j.1365-2044.2001.01795.x
Matthey T Schill M Muellejans B Earlier discharge from ICU with remifentanil/propofol versus fentanyl/midazolam Intensive Care Med 2004 30 suppl 1 s49
Kress JP Pohlman AS O'Connor MF Hall JB Daily interruption of sedative infusions in critically ill patients undergoing mechanical ventilation N Engl J Med 2000 342 1471 1477 10816184 10.1056/NEJM200005183422002
Reves JG Miller RD Nonbarbiturate intravenous anesthetics Anesthesia 2000 5 New York: Churchill Livingstone
| 15987391 | PMC1175879 | CC BY | 2021-01-04 16:04:53 | no | Crit Care. 2005 Mar 15; 9(3):R200-R210 | utf-8 | Crit Care | 2,005 | 10.1186/cc3495 | oa_comm |
==== Front
Crit CareCritical Care1364-85351466-609XBioMed Central London cc34971598739210.1186/cc3497ResearchTezosentan-induced attenuation of lung injury in endotoxemic sheep is associated with reduced activation of protein kinase C Kuklin Vladimir [email protected] Mikhail [email protected] Mikhail [email protected] Thomas 3Thomas.V.AndreasenIngebretsen Ole C [email protected] Kirsti [email protected] Lars [email protected] Research Fellow, Department of Anesthesiology, Faculty of Medicine, University of Tromsø, Norway2 Research Fellow, Department of Physiology, Faculty of Medicine, University of Tromsø, Norway3 Departmental engineer, Department of Physiology, Faculty of Medicine, University of Tromsø, Norway4 Professor, Department of Clinical Chemistry, University Hospital of Tromsø, Norway5 Professor, Department of Physiology, Faculty of Medicine, University of Tromsø, Norway6 Professor, Chairman of the Department of Anesthesiology, Faculty of Medicine, University of Tromsø, Norway2005 14 3 2005 9 3 R211 R217 24 11 2004 9 1 2005 27 1 2005 16 2 2005 Copyright © 2005 Kuklin 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
Studies in vitro reveal that endothelin-1 (ET-1) activates the α isoform of protein kinase C (PKC-α) in cultures of endothelial cells, thereby deranging cellular integrity. Sepsis and endotoxemia are associated with increased plasma concentrations of ET-1 that induce acute lung injury (ALI). We recently reported that non-selective ET-1 receptor blockade attenuates ALI in sheep by reducing the endotoxin-induced increase in extravascular lung water index (EVLWI). The aim of this study was to find out whether this attenuation is associated with reduced translocation of PKC-α from the cytosolic to the membrane fraction of lung tissue homogenate.
Methods
Seventeen awake, instrumented sheep were randomly assigned to a sham-operated group (n = 3), a lipopolysaccharide (LPS) group (n = 7) receiving an intravenous infusion of Escherichia coli 15 ng/kg per min for 24 hours, and a tezosentan group (n = 7) subjected to LPS and, from 4 hours, an intravenous injection of tezosentan 3 mg/kg followed by infusion at 1 mg/kg per hour for the reminder of the experiment. Pulmonary micro-occlusion pressure (Pmo), EVLWI, plasma concentrations of ET-1, tumor necrosis factor-a (TNF-a), and interleukin-8 (IL-8) were determined every 4 hours. Western blotting was used to assess PKC-α.
Results
In non-treated sheep a positive correlation was found between the plasma concentration of ET-1 and Pmo in the late phase of endotoxemia (12 to 24 hours). A positive correlation was also noticed between Pmo and EVLWI in the LPS and the LPS plus tezosentan groups, although the latter was significantly reduced in comparison with LPS alone. In both endotoxemic groups, plasma concentrations of ET-1, TNF-α, and IL-8 increased. In the LPS group, the cytosolic fraction of PKC-α decreased by 75% whereas the membrane fraction increased by 40% in comparison with the sham-operated animals. Tezosentan completely prevented the changes in PKC-α in both the cytosolic and the membrane fractions, concomitantly causing a further increase in the plasma concentrations of ET-1, TNF-α, and IL-8.
Conclusion
In endotoxemic sheep, ET-1 receptor blockade alleviates lung injury as assessed by a decrease in EVLWI paralleled by a reduction in Pmo and the prevention of activation of PKC-α.
See related commentary
==== Body
Introduction
Endothelin-1 (ET-1) has been identified as the most potent vasoconstrictor peptide known so far [1,2]. Locally produced ET-1 acts on three types of G-protein-coupled receptor: ETA, ETB1, and ETB2 [3]. The ETA and ETB2 receptors are expressed in vascular smooth muscle cells, whereas ETB1 is localized mainly in the endothelium. Binding of ET-1 to ETA and ETB2 leads to vascular constriction, whereas ETB1 induces relaxation by releasing nitric oxide and prostacyclin [4]. In the lowest concentration range, ET-1 mainly acts on the ETB1 receptor [5].
In sepsis, endotoxin and other microbial products that are released into the bloodstream trigger endothelial cells to the enhanced generation of ET-1 causing local vasoconstriction [6-9]. The effect of ET-1 is most prominent in the pulmonary circulation where the ETA and ETB receptors are widely distributed [10,11]. Previous investigators have noticed that intravenously infused ET-1 results in increased pulmonary artery pressure and lung edema [12,13]. Moreover, in isolated rat lungs in which the vasculature has been paralyzed, ET-1 enhances microvascular permeability, but the mechanisms involved have not yet been settled [14].
Studies in vitro have shown that the binding of ET-1 to its receptor might induce the activation of protein kinase C (PKC) [15,16]. Activation of the α isoform of PKC (PKC-α) might cause disturbances in the shape of the cells as well as of the intercellular junctions. The latter changes might promote acute lung injury (ALI) [17-19]. However, we are unable to determine whether any study in vivo has tested whether PKC-α is activated in endotoxin-induced ALI.
In sheep subjected to continuous infusion of endotoxin, we recently found that the dual ETA and ETB receptor blocker tezosentan precludes ALI as evaluated by improved gas exchange and a partial reversal of the increases in pulmonary vascular pressures and extravascular lung water index (EVLWI) [9]. However, the mechanisms involved in the tezosentan-induced reduction of EVLWI still remain obscure. We speculate whether non-selective blockade of ET-1 receptor by tezosentan alleviates ALI by dampening the activation of PKC-α and modulating inflammatory mediators such as tumor necrosis factor-α (TNF-α) and interleukin-8 (IL-8).
The aim of the present study was twofold: first, to investigate in sheep subjected to endotoxin-induced lung injury whether a relationship exists between the plasma concentration of ET-1 and characteristics of ALI such as the increases in lung microvascular pressure and extravascular lung water content, with or without tezosentan; and second, to assess the effects of tezosentan on the activation of PKC-α in lung tissue in parallel with changes in the plasma concentrations of TNF-α and IL-8.
Methods
The present investigation is based partly on data from a previously published study from our group [9] that was approved by the Norwegian Experimental Animal Board.
In brief, 17 yearling sheep were instrumented with a pulmonary artery thermal dilution catheter introduced via an introducer in the left external jugular vein, a thermo-dye dilution catheter introduced via an introducer in the ipsilateral common carotid artery, and a catheter in the left atrium, as described previously [9].
Experimental protocol
The animals were randomly assigned to a sham-operated group (n = 3), a group (n = 7) receiving an intravenous infusion of Escherichia coli lipopolysaccharide (LPS) 15 ng/kg per min for 24 hours (LPS group), and a group (n = 7) subjected to LPS and, from 4 hours, an intravenous injection of tezosentan 3 mg/kg followed by infusion at 1 mg/kg per hour for the reminder of the 24-hour experiment (LPS plus tezosentan group). During the experiment, sheep had free access to food and water.
EVLWI was assessed by the thermal-dye dilution method (Cold Z-021; Pulsion Medical Systems, Munich, Germany). Pulmonary micro-occlusion pressure (Pmo) was determined every 4 hours, as described previously [20]. In brief, Pmo was determined by advancing the Swan-Ganz catheter into the occlusion position in a distal pulmonary artery with the balloon deflated. The criteria for attainment of the micro-occlusion position included: first, easy retrograde aspiration of blood from the catheter; second, a pH, partial pressure of oxygen (PO2) and carbon dioxide (PCO2) of aspirated blood consistent with occlusion position, that is, partial pressure of oxygen in occlusion position higher than arterial partial pressure of oxygen (PmoO2 > PaO2) and partial pressure of carbon dioxide in occlusion position higher than arterial partial pressure of carbon dioxide (PmoCO2 < PaCO2); third, micro-occlusion pressure greater than proximal occlusion pressure; and fourth, micro-occlusion pressure greater than left atrial pressure, with true zero confirmed by connecting the left atrial catheter and the Swan-Ganz catheter sequentially to the same fixed transducer. Blood for biochemical analysis was sampled at 0, 4, 12, and 24 hours. After the sheep had been killed, lung samples were taken and kept in liquid nitrogen for further analyses.
Western blotting
The activation of PKC-α was assessed by translocation of kinase from cytosolic and/or membrane fractions of lung tissue extracts. In brief, lung tissue samples were homogenized (Polytron homogenizer, blade rotation speed 5,000 r.p.m.) in 1 ml of ice-cold extraction buffer consisting of (in mmol/l): 250 sucrose, 1 EDTA, 1 EGTA, 20 Tris-HCl pH 7.5, 10 2-mercaptoethanol, 20 dithiothreitol and 1 tablet of Complete™ EDTA-free protease inhibitor cocktail per 10 ml. Crude extracts were centrifuged at 200g to remove debris, followed by 100,000g for 60 min at 4°C. The supernatant represented the cytosolic fraction. The pellet was resuspended by sonication in 200 ml of a similar buffer supplemented with 1% Triton X-100 and centrifuged at 25,000g for 15 min at 4°C. The supernatant was collected as the Triton X-100-soluble membrane fraction. For SDS-PAGE, 10% polyacrylamide gels were loaded with 10 mg of protein per lane. After the end of electrophoresis, proteins were electroblotted to nitrocellulose membranes. Membranes were probed overnight with anti-PKC-α primary antibodies (Santa Cruz Biotechnology, Santa Cruz, CA, USA) at 4°C and for 1 hour with sheep anti-rabbit horseradish peroxidase-conjugated secondary antibodies (Zymed, San Francisco, CA, USA) at 22°C. Blots were incubated with ChemiLucent detection kit (Chemicon, Temecula, CA, USA). Immunopositive bands of PKC-α were detected with a Kodak Image Station 1000 (Kodak, Rochester, NY, USA) and densitometry readings were taken for statistical analysis.
Biochemical measurements
The ET-1 plasma levels were measured by chemiluminescent enzyme immunoassay (QuantiGlo QET00; R&D Systems, Minneapolis, MN, USA). Plasma levels of TNF-α and of IL-8 were determined with an Immulite instrument (Diagnostic Products Corporation, Los Angeles, CA, USA).
Statistical analysis
Data were checked for normal distribution by the Kolmogorov–Smirnov test. The relationship between ET-1, Pmo, and EVLWI was evaluated by regression analysis with the Pearson correlation coefficient. Equality of regression lines between the LPS and the LPS plus tezosentan groups was tested by single multiple regression [21]. The detected relative amounts of PKC-α in the groups and tissue fractions were compared by one-way analysis of variance. Plasma concentrations of ET-1, TNF-α, and IL-8 were analyzed by analysis of variance for repeated measurements. If F was statistically significant, Scheffe's test was used for post hoc intergroup analysis. To evaluate differences within groups towards the baseline value (time 0 hours), we used test of contrasts. We regarded P < 0.05 as statistically significant.
Results
In sham-operated sheep, all variables remained unchanged throughout the 24-hour experiments. During the first 12 hours we found no significant correlation between the plasma concentration of ET-1 and Pmo in sheep subjected to LPS (Fig. 1a). In contrast, we found a positive correlation between these variables beyond 12 hours (P < 0.01; Fig. 1b). A positive correlation also existed between Pmo and EVLWI in the LPS and the LPS plus tezosentan groups, as depicted in Fig. 2. However, tezosentan reduced the slope of the regression line compared with LPS alone (P < 0.05; Fig. 2).
As shown in Fig. 3, extracts of lung tissue from sheep exposed to LPS displayed a 75% reduction of the cytosolic fraction of PKC-α in comparison with samples from sham-operated animals (P < 0.05). The membrane fraction of PKC-α simultaneously increased by 40% in the LPS group compared with sham-operated sheep. Administration of tezosentan completely prevented the translocation of PKC-α from the cytosolic to the membrane fractions.
Figure 4 shows that after 4 hours of exposure to LPS, in parallel with the rise in ET-1, the plasma concentrations of TNF-α and IL-8 increased compared with intragroup baseline and sham-operated animals (P < 0.05). Notably, on cessation of the experiment, plasma concentrations of ET-1, TNF-α, and IL-8 were significantly higher in the LPS plus tezosentan group than with LPS alone (P < 0.05).
Discussion
The present study shows that during the late phase of endotoxemia in sheep (12 to 24 hours), the plasma concentration of ET-1 is significantly correlated with the microvascular pressure, whereas no such correlation was found during the early phase. Moreover, we observed a significant and positive correlation throughout the experiment between microvascular pressure and EVLWI in both endotoxemic groups. Interestingly, the regression line had a significantly lower slope in animals receiving tezosentan. To our knowledge, this is the first study demonstrating that blockade of ET-1 receptors precludes endotoxin-induced changes in PKC-α in cytosolic and membrane fractions of lung tissue.
Pulmonary microvascular pressure and permeability are important determinants of lung edema [22]. As reported previously by our group and others, increases in EVLWI in animals exposed to infusion of LPS are associated with enhanced pulmonary microvascular pressure [9,23,24]. However, none of these investigators have focused on the relationship between the plasma concentration of ET-1 and the pulmonary microvascular pressure. There is therefore no general agreement about where in the course of illness, or how, ET-1 exerts its action. One previous report suggests that ET-1 contributes directly to the severity of ALI by increasing the pulmonary microvascular pressure from the first hours of endotoxemia [25]. However, at variance with these results, we found that a fairly strong correlation between the plasma concentration of ET-1 and Pmo exists only in the late phase of endotoxemia. This is also in accordance with investigators who argue that thromboxane A2 is the dominating mediator of vasoconstriction during the first hours of endotoxemia, whereas ET-1 is responsible for vasoconstriction in the late phase [26-29].
The significantly positive correlation between Pmo and EVLWI in endotoxemia, and the decrease in the relationship after treatment with tezosentan, agrees fully with a recent investigation in endotoxemic pigs [30]. However, the beneficial effects of ET-1 blockade cannot be explained solely by attenuation of the endotoxin-induced increase in pulmonary artery pressure. The declining slope of the regression line between Pmo and EVLWI in tezosentan-treated animals indicates that additional factors affecting lung fluid filtration might be active. A few years ago, investigators found that ET-1 increases fluid filtration in isolated blood-perfused rat lungs pretreated with papaverine to deprive the lungs of any vascular tone [14]. Because the ability of the lung microvascular pressure to increase was precluded, the authors interpreted their findings as a result of increased permeability. However, the exact mechanisms involved still remain obscure.
PKC consists of a set of different isoenzymes: classical (α, β, γ), novel (ε, δ, θ, η), and atypical (ξ, λ), of which only the classical isoforms are sensitive to changes in intracellular Ca2+ concentration [31]. Recent studies have shown that ET-1 stimulates the release of Ca2+ from the endoplasmic reticulum and activates PKC-α in the cell membrane [15-19]. After being activated, PKC-α has been shown to mediate the disruption of vascular endothelial cadherin junctions [32]. Moreover, PKC-α activates myosin light chain kinase, which is involved in endothelial cell gap formation and barrier dysfunction [33]. In the lung vasculature, PKC-α-induced disruption might derange the endothelial integrity [19]. We therefore speculate that the increase in vascular permeability and the evolution of ALI might be due to ET-1-induced activation of PKC-α in the cell membrane. We believe that blockade of ET-1 receptors, resulting in a combination of reduced microvascular pressure and decreased activation of PKC-α, is one of the main reasons for the amelioration of ALI in the present study.
It is well established that infusion of LPS stimulates a release of inflammatory mediators such as TNF-α, IL-8, and ET-1 [34-36]. In contrast, ET-1 stimulates monocytes and macrophages to release TNF-α and IL-8 in its own right [37,38]. In the present study, enhanced plasma concentrations of TNF-α, IL-8, and ET-1 were found after 4 hours in both endotoxemic groups. However, at the end of the experiments the plasma concentrations of all three mediators were significantly higher in tezosentan-treated animals than in animals given LPS alone. The increases in ET-1 and TNF-α are consistent with a previous investigation employing the endothelin receptor antagonist bosentan [39], but in contrast to that short-term study, we exposed sheep to 24 hours of endotoxemia. According to another recent study, ETB receptors in the lungs are involved in the clearance of ET-1 from the circulation [40]. Consequently, ET-1 receptor blockade prolongs ET-1 half-life in the plasma and reportedly shifts tissue uptake from the lungs to other organs [41]. In the present study, tezosentan increased the plasma concentration of ET-1 to an extent that might have enhanced the release of TNF-α and IL-8 from the monocytes and macrophages.
The present endotoxin-induced lung injury model in sheep is not ideal for elucidating the effects of ET-1 receptor blockade on permeability because microvascular pressure cannot be deliberately changed. Further studies of ET-1 receptor blockade on permeability are therefore required in a more complex experimental setting on intact animals or in isolated perfused lungs.
Conclusion
In endotoxemic sheep, ET-1 plasma concentration is significantly correlated with Pmo in the late phase. Moreover, Pmo and extravascular lung water content demonstrate a positive correlation from the first hours of endotoxin infusion. Blockade of ET-1 receptors attenuates ALI by reducing the pulmonary microvascular pressure and most probably also by decreasing permeability secondary to reducing the activation of PKC-α. However, further studies are needed to explain the exact mechanisms behind the decrease in extravascular lung water and the prevention of activation of PKC-α after ET-1 receptor blockade.
Key messages
• In endotoxemic sheep, extravascular lung water content correlates positively with pulmonary microvascular pressure.
• Non-selective endothelin-1 receptor blockade attenuates ovine endotoxin-induced lung injury by reducing pulmonary microvascular pressure and probably also by decreasing microvascular permeability secondary to reduced activation of PKCα.
Abbreviations
ALI = acute lung injury; ET-1 = endothelin-1; EVLWI = extravascular lung water index; IL = interleukin; LPS = lipopolysaccharide; PKC = protein kinase C; Pmo = pulmonary capillary micro-occlusion pressure; TNF = tumor necrosis factor.
Competing interests
This study was supported by Helse Nord (Norwegian governmental funds), project number 4001.721.132 and departmental funds of the Departments of Anesthesiology, Physiology and Clinical Chemistry, University of Tromsø, Norway.
Authors' contributions
VK participated in the design of the study, analyzed the data, and drafted the manuscript. MK, MS, TA, OCI and KY contributed to the biochemical analysis and participated in the design of the study. LB administered the study, participated in the design of the study and suggested improvements to the manuscript. All authors read and approved the final manuscript.
Acknowledgements
The authors thank Dr Martine Clozel (Actelion Pharmaceuticals Ltd, Allschwil, Switzerland) for generously giving us tezosentan, and Dr Tormod Brenn PhD and Dr Tom Wilsgaard PhD for review of the statistical methods. A part of this study was presented and awarded a prize at the 17th Annual Congress of the European Society of Intensive Care Medicine, Berlin, Germany, 10 to 13 October, 2004 (Intensive Care Med 2004, 30 (Suppl 1):S32, abstract no 504).
Figures and Tables
Figure 1 Relationship between plasma concentration of endothelin-1 (ET-1) and pulmonary micro-occlusion pressure (Pmo) in sheep. (a) From 0 to 12 hours of LPS infusion (r = -0.51, P = 0.12, n = 10); (b) from 12 to 24 hours of LPS infusion (r = 0.75, P < 0.01, n = 9).
Figure 2 Relationship between extravascular lung water index (EVLWI) and pulmonary micro-occlusion pressure (Pmo) in endotoxemic sheep. LPS alone (r = 0.73, P < 0.01, n = 42). LPS with tezosentan (r = 0.67, P < 0.0001, n = 42).
Figure 3 Protein kinase Cα (PKC-α) in sheep lung tissue homogenates detected by Western blotting. (a) In the cytosolic fraction; (b) in the membrane fraction. Results are means ± SEM. Groups were as follows: sham-operated group (n = 3); lipopolysaccharide group (LPS; n = 4); LPS plus tezosentan group (n = 4). †P < 0.05 between sham-operated and LPS groups; ‡P < 0.05 between LPS and LPS plus tezosentan groups.
Figure 4 Plasma concentration of endothelin-1 (ET-1), tumor necrosis factor-α (TNF-α) and interleukin-8 (IL-8). Results are means ± SEM. Groups were as follows: sham-operated group (n = 3); lipopolysaccharide group (LPS; n = 7); LPS plus tezosentan group (n = 7). ND, not detectable. †P < 0.05 between sham-operated and LPS groups; ‡P < 0.05 between LPS and LPS plus tezosentan groups; *P < 0.05 between sham-operated and LPS plus tezosentan groups; &P < 0.05 from t = 0 hours in the LPS group; §P < 0.05 from t = 0 hours in the tezosentan group.
==== Refs
Yanagisawa M Kurihara H Kimura S Tomobe Y Kobayashi M Mitsui Y Yazaki Y Goto K Masaki T A novel potent vasoconstrictor peptide produced by vascular endothelial cells Nature 1988 332 411 415 2451132 10.1038/332411a0
Mitaka C Hirata Y Nagura T Tsunoda Y Amaha K Circulating endothelin-1 concentrations in acute respiratory failure Chest 1993 104 476 480 8339637
Douglas SA Beck GR JrElliott JD Ohlstein EH Pharmacological evidence for the presence of three distinct functional endothelin receptor subtypes in the rabbit lateral saphenous vein Br J Pharmacol 1995 114 1529 1540 7599920
de Nucci G Thomas R D'Orleans-Juste P Antunes E Walder C Warner TD Vane JR Pressor effects of circulating endothelin are limited by its removal in the pulmonary circulation and by the release of prostacyclin and endothelium-derived relaxing factor Proc Natl Acad Sci USA 1988 85 9797 9800 3059352
Masaki T Possible role of endothelin in endothelial regulation of vascular tone Annu Rev Pharmacol Toxicol 1995 35 235 255 7598493 10.1146/annurev.pa.35.040195.001315
Pernow J Hemsen A Lundberg JM Increased plasma levels of endothelin-like immunoreactivity during endotoxin administration in the pig Acta Physiol Scand 1989 137 317 318 2694766
Nakamura T Kasai K Sekiguchi Y Banba N Takahashi K Emoto T Hattori Y Shimoda S Elevation of plasma endothelin concentrations during endotoxin shock in dogs Eur J Pharmacol 1991 205 277 282 1667911 10.1016/0014-2999(91)90910-I
Mitaka C Hirata Y Makita K Nagura T Tsunoda Y Amaha K Endothelin-1 and atrial natriuretic peptide in septic shock Am Heart J 1993 126 466 468 8338025 10.1016/0002-8703(93)91074-O
Kuklin VN Kirov MY Evgenov OV Sovershaev MA Sjoberg J Kirova SS Bjertnaes LJ Novel endothelin receptor antagonist attenuates endotoxin-induced lung injury in sheep Crit Care Med 2004 32 766 773 15090960 10.1097/01.CCM.0000114575.08269.F6
Henry PJ Rigby PJ Self GJ Preuss JM Goldie RG Relationship between endothelin-1 binding site densities and constrictor activities in human and animal airway smooth muscle Br J Pharmacol 1990 100 786 792 2169940
McKay KO Black JL Diment LM Armour CL Functional and autoradiographic studies of endothelin-1 and endothelin-2 in human bronchi, pulmonary arteries, and airway parasympathetic ganglia J Cardiovasc Pharmacol 1991 17 S206 S209 1725334
Horgan MJ Pinheiro JM Malik AB Mechanism of endothelin-1-induced pulmonary vasoconstriction Circ Res 1991 69 157 164 2054931
Filep JG Sirois MG Rousseau A Fournier A Sirois P Effects of endothelin-1 on vascular permeability in the conscious rat: interactions with platelet-activating factor Br J Pharmacol 1991 104 797 804 1667286
Helset E Kjaeve J Hauge A Endothelin-1-induced increases in microvascular permeability in isolated, perfused rat lungs requires leukocytes and plasma Circ Shock 1993 39 15 20 8481973
Griendling KK Tsuda T Alexander RW Endothelin stimulates diacylglycerol accumulation and activates protein kinase C in cultured vascular smooth muscle cells J Biol Chem 1989 264 8237 8240 2656676
Danthuluri NR Brock TA Endothelin receptor-coupling mechanisms in vascular smooth muscle: a role for protein kinase C J Pharmacol Exp Ther 1990 254 393 399 2166789
Lynch JJ Ferro TJ Blumenstock FA Brockenauer AM Malik AB Increased endothelial albumin permeability mediated by protein kinase C activation J Clin Invest 1990 85 1991 1998 2347922
Siflinger-Birnboim A Goligorsky MS Del Vecchio PJ Malik AB Activation of protein kinase C pathway contributes to hydrogen peroxide-induced increase in endothelial permeability Lab Invest 1992 67 24 30 1378104
Siflinger-Birnboim A Johnson A Protein kinase C modulates pulmonary endothelial permeability: a paradigm for acute lung injury Am J Physiol Lung Cell Mol Physiol 2003 284 L435 L451 12573983
Bjertnaes LJ Koizumi T Newman JH Inhaled nitric oxide reduces lung fluid filtration after endotoxin in awake sheep Am J Respir Crit Care Med 1998 158 1416 1423 9817688
Kleinbaum DG Kupper LL Muller KE Kleinbaum DG Method II: using a single regression equation to compare two straight lines Applied Regression Analysis and Multivariable Methods 1998 3 Pacific Grove, California: Duxbury Press 327 328
Block ER Pulmonary endothelial cell pathology: implications for acute lung injury Am J Med Sci 1992 304 136 144 1503113
Traber DL Adair TH Adams T Jr Hemodynamic consequences of endotoxemia in sheep Circ Shock 1981 8 551 561 7285288
Kutzsche S Lyberg T Bjertnaes LJ Effects of adenosine on extravascular lung water content in endotoxemic pigs Crit Care Med 2001 29 2371 3276 11801842 10.1097/00003246-200112000-00021
Albertini M Ciminaghi B Mazzola S Clement MG Improvement of respiratory function by bosentan during endotoxic shock in the pig Prostaglandins Leukot Essent Fatty Acids 2001 65 103 108 11545627 10.1054/plef.2001.0296
Henry C Ogletree M Brigham K Hammon JW Jr Attenuation of the pulmonary vascular response to endotoxin by a thromboxane synthesis inhibitor (UK-38485) in unanesthetized sheep J Surg Res 1991 50 77 81 1987435
Morel DR Pittet JF Gunning K Hemsen A Lacroix JS Lundberg JM Time course of plasma and pulmonary lymph endothelin-like immunoreactivity during sustained endotoxaemia in chronically instrumented sheep Clin Sci 1991 81 357 365 1655337
Weitzberg E Circulatory responses to endothelin-1 and nitric oxide with special reference to endotoxin shock and nitric oxide inhalation Acta Physiol Scand Suppl 1993 611 1 72 8379340
Schmeck J Heller A Groschler A Recker A Neuhof H Urbaschek R Koch T Impact of endothelin-1 in endotoxin-induced pulmonary vascular reactions Crit Care Med 2000 28 2851 2857 10966261 10.1097/00003246-200008000-00028
Rossi P Wanecek M Konrad D Oldner A Tezosentan counteracts endotoxin-induced pulmonary edema and improves gas exchange Shock 2004 21 543 548 15167683 10.1097/01.shk.0000126147.76311.18
Mellor H Parker PJ The extended protein kinase C superfamily Biochem J 1998 332 281 292 9601053
Sandoval R Malik AB Minshall RD Kouklis P Ellis CA Tiruppathi C Ca2+ signalling and PKCα activate increased endothelial permeability by disassembly of VE-cadherin junctions J Physiol 2001 533 433 445 11389203 10.1111/j.1469-7793.2001.0433a.x
Garcia JG Davis HW Patterson CE Regulation of endothelial cell gap formation and barrier dysfunction: role of myosin light chain phosphorylation J Cell Physiol 1995 163 510 522 7775594 10.1002/jcp.1041630311
Sugiura M Inagami T Kon V Endotoxin stimulates endothelin-release in vivo and in vitro as determined by radioimmunoassay Biochem Biophys Res Commun 1989 161 1220 1227 2662974 10.1016/0006-291X(89)91372-7
Beutler B Cerami A The biology of cachectin/TNF – a primary mediator of the host response Annu Rev Immunol 1989 7 625 655 2540776
Mathison JC Wolfson E Ulevitch RJ Participation of tumor necrosis factor in the mediation of gram negative bacterial lipopolysaccharide-induced injury in rabbits J Clin Invest 1988 81 1925 1237 3384955
Helset E Sildnes T Seljelid R Konopski S Endothelin-1 stimulates human monocytes in vitro to release TNF-1α, IL-1β and IL-6 Mediators Inflamm 1993 2 417 422
Helset E Sildnes T Konopski S Endothelin-1 stimulates monocytes in vitro to release chemotactic activity identified as interleukin-8 and monocyte chemotactic protein-1 Mediators Inflamm 1994 3 155 160
Wanecek M Oldner A Rudehill A Sollevi A Alving K Weitzberg E Cardiopulmonary dysfunction during porcine endotoxin shock is effectively counteracted by the endothelin receptor antagonist bosentan Shock 1997 7 364 370 9165672
Shiba R Yanagisawa M Miyauchi T Ishii Y Kimura S Uchiyama Y Masaki T Goto K Elimination of intravenously injected endothelin-1 from the circulation of the rat J Cardiovasc Pharmacol 1989 13 S98 S101 2473338
Burkhardt M Barton M Shaw SG Receptor- and non-receptor-mediated clearance of big-endothelin and endothelin-1: differential effects of acute and chronic ETA receptor blockade J Hypertens 2000 18 273 279 10726713 10.1097/00004872-200018030-00006
| 15987392 | PMC1175881 | CC BY | 2021-01-04 16:04:53 | no | Crit Care. 2005 Mar 14; 9(3):R211-R217 | utf-8 | Crit Care | 2,005 | 10.1186/cc3497 | oa_comm |
==== Front
Crit CareCritical Care1364-85351466-609XBioMed Central London cc35031598739410.1186/cc3503ResearchContinuously assessed right ventricular end-diastolic volume as a marker of cardiac preload and fluid responsiveness in mechanically ventilated cardiac surgical patients Wiesenack Christoph [email protected] Christoph 2Keyser Andreas 3Laule Sven 4Prasser Christopher 1Keyl Cornelius 51 Consultant, Department of Anesthesiology, University Hospital Regensburg, Regensburg, Germany2 Resident, Department of Anesthesiology, University Hospital Regensburg, Regensburg, Germany3 Staff Surgeon, Department of Cardiothoracic and Vascular Surgery, University Hospital Regensburg, Regensburg, Germany4 Staff Anesthesiologist, Department of Anesthesiology, Heart-Center Bad Krozingen, Bad Krozingen, Germany5 Consultant, Department of Anesthesiology, Heart-Center Bad Krozingen, Bad Krozingen, Germany2005 1 4 2005 9 3 R226 R233 15 10 2004 18 1 2005 1 2 2005 18 2 2005 Copyright © 2005 Wiesenack 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
Assessing cardiac preload and fluid responsiveness accurately is important when attempting to avoid unnecessary volume replacement in the critically ill patient, which is associated with increased morbidity and mortality. The present clinical trial was designed to compare the reliability of continuous right ventricular end-diastolic volume (CEDV) index assessment based on rapid response thermistor technique, cardiac filling pressures (central venous pressure [CVP] and pulmonary capillary wedge pressure [PCWP]), and transesophageal echocardiographically derived evaluation of left ventricular end-diastolic area (LVEDA) index in predicting the hemodynamic response to volume replacement.
Methods
We studied 21 patients undergoing elective coronary artery bypass grafting. After induction of anesthesia, hemodynamic parameters were measured simultaneously before (T1) and 12 min after volume replacement (T2) by infusion of 6% hydroxyethyl starch 200/0.5 (7 ml/kg) at a rate of 1 ml/kg per min.
Results
The volume-induced increase in thermodilution-derived stroke volume index (SVITD) was 10% or greater in 19 patients and under 10% in two. There was a significant correlation between changes in CEDV index and changes in SVITD (r2 = 0.55; P < 0.01), but there were no significant correlations between changes in CVP, PCWP and LVEDA index, and changes in SVITD. The only variable apparently indicating fluid responsiveness was LVEDA index, the baseline value of which was weakly correlated with percentage change in SVITD (r2 = 0.38; P < 0.01).
Conclusion
An increased cardiac preload is more reliably reflected by CEDV index than by CVP, PCWP, or LVEDA index in this setting of preoperative cardiac surgery, but CEDV index did not reflect fluid responsiveness. The response of SVITD following fluid administration was better predicted by LVEDA index than by CEDV index, CVP, or PCWP.
==== Body
Introduction
Accurate evaluation of cardiac performance and preload status, and assessment of fluid responsiveness are important goals in the treatment of critically ill patients. Despite the current controversy surrounding the usefulness of and risks associated with the pulmonary artery catheter (PAC) [1,2], the PAC remains more frequently used for monitoring and is preferred over transesophageal echocardiography (TEE) by cardiovascular anaesthesiologists [3]. However, it has been demonstrated that PAC-derived filling pressures are of little help when making decisions regarding adequate volume therapy. Nevertheless, the majority of intensive care unit (ICU) physicians use filling pressures in their decision making regarding volume replacement to improve hemodynamics. This accentuates the need for reliable indicators of fluid responsiveness so that needless or even deleterious volume replacement associated with increased morbidity and mortality may be avoided in critically ill patients [4]. Several markers of ventricular preload, specifically intrathoracic blood volume [5,6], TEE-derived assessment of left ventricular end-diastolic area (LVEDA) [7,8], and cyclic fluctuation in arterial pressure wave that occurs in mechanically ventilated patients [7,9-14], have been tested as predictors of fluid responsiveness, some with excellent results. However, apart from pulse contour analysis, which has never been found in positive-pressure ventilation to reflect actual stroke volume variation [15,16], none of the techniques for assessing preload can be used continuously or routinely in most patients.
Several studies have emphasized the good correlation between estimates of right ventricular end-diastolic volume (RVEDV) by thermodilution-derived right ventricular ejection fraction (RVEF) and surrogates of stroke volume [17-20]. However, the thermodilution technique for assessing RVEDV is still intermittent, and the value of RVEDV as a marker of fluid responsiveness in critically ill patients is controversial [20-22].
A recently available Swan–Ganz catheter with a rapid response thermistor permits nearly continuous assessment of cardiac output (CO), RVEF and RVEDV, which should be more applicable in the ICU. The measurement variability associated with the intermittent bolus technique is eliminated by this catheter, and continuously assessed RVEDV (CEDV) should be more accurate than RVEDV based on intermittent thermodilution; therefore, CEDV may be a valuable marker of cardiac preload and a predictor of fluid responsiveness.
The purpose of the present study was to compare the accuracy of CEDV derived using a new right-heart ejection fraction catheter and commonly used preload parameters (central venous pressure [CVP], pulmonary capillary wedge pressure [PCWP], and transesophageal echocardiography [TEE]-derived assessment of LVEDA) in predicting the response of stroke volume to volume replacement in mechanically ventilated cardiac surgical patients.
Materials and methods
After obtaining approval from the local ethics committee and written informed consent from all participants, we studied 21 patients (17 male; aged 53–78 years, mean 65.7 years) undergoing elective coronary artery bypass grafting. Patients with valvular heart disease, intracardiac shunts, regional myocardial asynchrony, peripheral vascular disease, preoperative dysrhythmias, and an ejection fraction under 30 % were excluded from the study. Dynamic variables, such as pulse pressure variation, were not measured to assess fluid responsiveness in our investigation.
All patients received an arterial catheter for continuous monitoring of arterial blood pressure (Siemens monitor SC 9000; Siemens AG, Erlangen, Germany). Anesthesia was induced with fentanyl (5 μg/kg) followed by etomidate until loss of consciousness and pancuronium (100 μg/kg), and maintained using 1.5% sevoflurane end-expiratory, supplemented with bolus doses of fentanyl (up to 20 μg/kg) and pancuronium (50 μg/kg) for neuromuscular blockade. Mechanical ventilation (without positive end-expiratory pressure) at a constant tidal volume of 7 ml/kg to an end-tidal partial carbon dioxide tension of 30–35 mmHg was maintained at a inspired fractional oxygen of 0.5 throughout the study.
After induction of anesthesia, a 7.5 Fr right-heart ejection fraction catheter (CCOmboV 774HF75; Edwards Lifesciences, Irvine, CA, USA) was inserted via an 8.5 Fr introducer into the right internal jugular vein and connected to a Vigilance Monitor system (Edwards Lifesciences) for continuous assessment of CO (CCO), CEDV and of RVEF, and for determination of CO using the intermittent thermodilution technique (COTD).
The methodology of CCO measurement, based on the pulsed warm thermodilution technique, was described previously [23] and involves the release of small pulses of heat from a thermal coil mounted on the PAC at the level of the right ventricle. To reflect sudden changes in CO, the Vigilance Monitor provides a STAT mode of operation, which has been shown to permit accurate measurement of CCO [24]. The software algorithm for STAT CCO does not contain a moving average filter but depends on some previous data for artifact suppression. Without user calibration, CCO is computed from the area under the thermodilution curve, and every 30–60 s the displayed CCO is updated.
The new CEDV algorithm uses the slaved electrocardiograph signal and generates a relaxation waveform, which resembles the bolus thermodilution washout decay curve. The waveform is based on the repeating on–off CCO input signal and is generated by accumulating the temperature change for each on and each off segment of the input signal (Fig. 1). Calculation of RVEF is based on estimation of the exponential decay time constant (τ) of this curve and heart rate (HR): RVEF = 1 - exp (-60/ [τ × HR]). CEDV, which is based on CCO, HR and RVEF, is calculated as follows: CEDV = (CCO/HR)/RVEF. It includes the whole range of temperatures of the thermodilution curve (Fig. 1).
COTD measurements were performed by injection of 10 ml iced saline solution via the CVP port and subsequent detection by the thermistor embedded in the PAC. An average of three measurements, all taken within a 10% range randomly distributed over the respiratory cycle, was calculated using the Stewart–Hamilton formula.
The TEE probe (OmniPlane II probe 21369A and SONOS 5500 Phased Array Imaging System; Philips Medical Systems, DA Best, The Netherlands) was positioned to obtain a transgastric midpapillary short-axis view of the left ventricle. This position was maintained over the whole period of data acquisition. Echocardiographic images and electrocardiograms were recorded together, and end-diastole was defined as the greatest left ventricular cross-sectional area immediately after the R-wave peak on the electrocardiogram.
Correspondingly, end-systole was defined as the smallest left ventricular dimension during the last half of the T wave. An independent reviewer, who was blinded to the condition of the trial participants, analyzed TEE images. LVEDA and left ventricular end-systolic area were traced edge to edge, including the papillary muscles. Fractional area change was calculated as (LVEDA – left ventricular end-systolic area)/LVEDA. Three measurements, performed at end-expiration, were analyzed and averaged.
All hemodynamic parameters were measured simultaneously after induction of anesthesia, when CCO had stabilized (T1). A second measurement was performed (T2) 12 min after volume replacement by infusion of 6% hydroxyethyl starch 200/0.5 (7 ml/kg) at a rate of 1 ml/kg per min (mean 579 ml). Measurements were taken in a hemodynamically steady state, in the absence of vasoactive drugs. Patients were classified as responders to volume loading if the increase in thermodilution-derived stroke volume index (SVITD) was 10% or greater, or as nonresponders if the increase in SVITD was under 10%.
Statistical analysis
For statistical analysis, all volume variables were indexed to body surface area. Statistical analysis was performed using the SPSS 12.0 software (SPSS Inc., Chicago, IL, USA). After assessment of normal distribution using the Lilliefors modification to the Kolmogorov–Smirnov test, the Student's t-test was used to compare variables. Because the thermodilution technique still represents the 'gold standard' for assessment of cardiac index (CI), we conducted linear regression analyses between changes in variables that reflect preload (CEDV index, CVP, PCWP, and LVEDA index) and changes in the preload-dependent variable SVITD, and between baseline (T1) values of variables that reflect preload (CEDV index, CVP, PCWP, and LVEDA index) and the change in SVITD (ΔSVITD; expressed as a percentage). P < 0.05 was considered statistically significant.
Results
Demographic data for the patients included in the present study are summarized in Table 1.
Except for HR, all hemodynamic parameters changed significantly after volume replacement (Table 2). The volume-induced increase in SVITD was 10% or greater (range 21.8–93.4%) in 19 patients (responders) and under 10% in two patients (nonresponders).
Linear regression analysis between changes in CEDV index (ΔCEDV index) and ΔSVITD revealed a significant correlation (r2 = 0.55; P < 0.01), but linear regression analysis between changes in CVP, PCWP and LVEDA index, and ΔSVITD did not identify any significant correlations among variables. LVEDA index at baseline and the percentage ΔSVITD were weakly correlated (r2 = 0.38; P < 0.01), but linear regression analysis between the remaining variables reflecting preload (CEDV index, CVP, and PCWP) did not reveal any significant relationships (Fig. 2). Variables reflecting systolic function – RVEF and fractional area change – remained constant, without a significant relationship between them.
Discussion
Over recent years numerous studies have been performed to evaluate the usefulness of thermodilution-derived estimates of RVEDV index in a variety of clinical situations [17-20,25-28]. Several investigators emphasized the good correlation between RVEDV index and CI [17-20], suggesting that a volumetric assessment of cardiac preload may provide a more useful evaluation of ventricular filling than that offered by the assessment of cardiac filling pressures. A previous study found that a RVEDV index greater than 138 ml/m2 was associated with lack of response but that RVEDV index below 90 ml/m2 was associated with a high rate of response to fluid administration [18]. In contrast to these findings, Wagner and Leatherman [22] reported a positive response to volume loading in a number of patients with an RVEDV index above 138 ml/m2 and a lack of response in some patients with an RVEDV index below 90 ml/m2. Furthermore, the response to volume loading was rather unpredictable when RVEDV index ranged between these extremes. Based on those findings, no threshold value may be proposed to discriminate between responders and nonresponders before fluid application [18,21]. Nevertheless, most authors stated that thermodilution-derived estimates of RVEDV index appeared to be better indicators of cardiac preload [19,27,29] and can predict preload recruitable increases in SVI more accurately than can cardiac filling pressures [17,18].
In the present study ΔCEDV index was significantly correlated with ΔSVITD, whereas there was a lack of correlations between changes in the remaining preload-indicating variables and ΔSVITD, suggesting that increased cardiac preload is more reliably reflected by CEDV index than by CVP, PCWP, or LVEDA index. Some investigators questioned the clinical significance of correlation between RVEDV index and continuously assessed CI, but Durham [19] and Nelson [30] and their groups demonstrated that mathematical coupling does not account for the relationship between variables.
Several authors described RVEDV index as a marker of cardiac preload, indicated by the linear correlation between CI and RVEDV index [17-20]. Although a linear correlation between variables seems unlikely because measurements might have been performed at different operating points on the nonlinear curve describing the relationship between end-diastolic volume and stroke volume, the authors stated that RVEDV index could accurately predict preload recruitable increase in CI [17,18]. However, validation of a variable as an indicator of preload requires, in addition to demonstrations that the variable increases with fluid loading and that the increase is related to an increase in stroke volume, the demonstration that this variable does not change with an intervention that alters cardiac contractility (e.g. administration of inotropic agents). In most of studies preload-indicating variables were not tested in the presence of inotropic drugs, and therefore the hypothesis that RVEDV index is an accurate indicator of preload has not yet been proven. Accordingly, assuming that changes in myocardial contractility or afterload did not occur during the study period and measurements were performed in the steep part of the Frank–Starling curve, the significant relationship between CEDV index and CI in our study may merely indicate that an increased cardiac preload is reliably reflected by CEDV index.
It should be noted that the terms 'cardiac preload' and 'fluid responsiveness' are not exchangeable. The increase in SVI depends on ventricular function; a decrease in ventricular contractility decreases the slope of the relationship between end-diastolic volume and stroke volume [31] and moves the Frank–Starling curve to the right. Therefore, patients with a dilated left ventricle could still respond to fluid despite increased measures of static cardiac preload. Consequently, fluid responsiveness, defined as the response of SVI to volume challenge [32], cannot be accurately predicted simply by assessing cardiac preload.
For this reason, the more relevant question concerns the value of RVEDV index as an indicator of fluid responsiveness, but until now only limited and inconsistent information has been available regarding the value of this variable [20,22]. A variable is a predictor of fluid responsiveness if there is a relationship between the baseline value of that variable and changes in SVI after fluid loading. Reuse and coworkers [20] demonstrated a weak correlation (r2 = 0.19; P < 0.01) between the response to fluid challenge and baseline RVEDV index in 41 critically ill patients. Wagner and Leatherman [22] found a comparable, modest correlation among variables (r2 = 0.19; P < 0.05), but they stated that RVEDV index was not a reliable predictor of response to fluid.
In the present study, baseline values of CEDV index were not correlated with changes in SVITD (Fig. 2a). Furthermore, using previously suggested criteria [15], neither a very high (>138 ml/m2) nor a very low (<90 ml/m2) CEDV index proved to be a reliable predictor of hemodynamic response to volume challenge. In accordance with Wagner et al [22], even one patient with markedly elevated CEDVI (159 ml/m2) was able to increase SVI in response to a fluid challenge in this study. This phenomenon may be accounted for by the fact that the left ventricular response to fluid loading may be predicted by the right ventricular volume only in a limited manner. The optimal CEDV index should be determined individually for each patient. Consequently, patients should not be resuscitated to an absolute CEDV index, but rather based upon their individual response of CEDV index and CCI to fluid administration.
A factor that could possibly affect the accuracy of CEDV index is the presence of a low RVEF [22], because CEDV index is calculated as the quotient of SVI and RVEF. The mean RVEF for the patients studied was 30.7 ± 9.1% at baseline, which is markedly lower than in the study conducted by Diebel and coworkers [18] (38 ± 9%). It is possible that CEDV index is a better predictor of response to volume in patients with higher RVEF. Another factor that should be taken into account was mentioned by Michard and coworkers [21,31]. The increase in ventricular end-diastolic volumes as a result of fluid challenge depends on the partitioning of fluid into different cardiovascular compartments organized in series. When ventricular capacitance is increased, volume loading will increase intravascular blood volume but not necessarily cardiac preload [31].
The results of the present investigation suggest that LVEDA index is a better predictor of fluid responsiveness than is CEDV index, and is even better than CVP or PCWP, as indicated by the weak correlation between baseline value of LVEDA index and the resulting increase in SVITD following fluid loading (Fig. 2b). These findings are in accordance with those of other studies [7,13] and emphasize the importance of TEE in detecting acute changes in hemodynamics. However, the short-axis view provides only an area, not a volume, and the assumption that this area correlates with a volume is only valid when there are no regional contraction abnormalities [28,32]. The findings of recent studies demonstrate a limited relationship between hemodynamic and echocardiographic evaluation of left ventricular performance [33] and the minimal value of LVEDA index in discriminating responders from nonresponders [7]. The analysis presented in Fig. 2b shows the considerable influence of two data points corresponding to relative increases in SVI of about 77% and 94%. For the other patients, exhibiting relative increases in stroke volume of 10–40%, LVEDA index could not predict reliably the magnitude of this response. Furthermore, echocardiography requires an experienced investigator, is sometimes impossible to perform, and its availability as a device for continuous assessment of hemodynamics in the ICU is limited.
Limitations
Monitoring of CEDV index can be unreliable in the presence of severe tricuspid valve insufficiency or during conditions of unsteady or rapid changing blood temperature. Furthermore, tachycardia at rates in excess of 150 beats/min will prevent accurate measurement of the patient's R–R interval.
For ethical reasons, assessment of the hemodynamic response of CEDV index was only be performed by a unidirectional preload change. Therefore, this parameter should be evaluated additionally under hemorrhage conditions in an animal experimental setting concerning its relative correctness.
In this setting of preoperative cardiac surgery, characterized by preoperative fasting, diuretic therapy, and the vasodilatory effect of sevoflurane, relative hypovolemia is common and could account for the fact that most of the patients responded to fluid. The small number of patients in the nonresponder group makes any conclusion regarding possible differences in any of the variables between responders and nonresponders difficult.
Conclusion
Despite the limitations mentioned above, the results of the present study demonstrated that an increased cardiac preload is more reliably reflected by CEDV index than by CVP, PCWP or LVEDA index in this setting of preoperative cardiac surgery. However, CEDV index failed to be a variable of fluid responsiveness. The response of SVITD subsequent to fluid administration is better predicted by LVEDA index than by CEDV index.
Key messages
• An increased cardiac preload is more reliably reflected by CEDV index than by CVP, PCWP or LVEDA index.
• But CEDV index did not reflect fluid responsiveness.
• The terms "cardiac preload" and "fluid responsiveness" are not exchangeable.
• Fluid responsiveness is better predicted by LVEDA index than by CEDV index.
Abbreviations
CCO = continuous cardiac output; CEDV = continuous right ventricular end-diastolic volume; CI = cardiac index; CO = cardiac output; CVP = central venous pressure; HR = heart rate; ICU = intensive care unit; LVEDA = left ventricular end-diastolic area; PAC = pulmonary artery catheter; PCWP = pulmonary capillary wedge pressure; RVEF = right ventricular ejection fraction; RVEDV = right ventricular end-diastolic volume; SV = stroke volume; SVITD = thermodilution-derived stroke volume index; TEE = transesophageal echocardiography.
Competing interests
The author(s) declare that they have no competing interests.
Authors' contributions
CW designed the study, processed the data, and wrote the manuscript. CF collected the clinical data. AK collected the clinical data and participated in the study design. SL collected the clinical data. CP designed the study and collected the clinical data. CK performed the statistical analysis and extensively revised the manuscript. All authors read and approved the final manuscript.
Acknowledgements
Departmental funding supported this study financially: Department of Anesthesiology, University Hospital, Regensburg, Germany.
Figures and Tables
Figure 1 CEDV assessment. Shown is a modified algorithm block diagram for continuous right ventricular end-diastolic volume (CEDV) assessment. CCO = continuous cardiac output; CEDV = continuous right ventricular end-diastolic volume; HR = heart rate; PRBS = Pseudo-Random Binary Sequence; REF = right ventricular ejection fraction; τ = exponential decay time constant. Courtesy of Edwards Lifesciences, Unterschleissheim, Germany.
Figure 2 Linear regression analyses. Linear regression analysis between (a) changes in thermodilution-derived stroke volume index (ΔSVITD) and baseline values of continuously assessed right ventricular end-diastolic volume index (CEDVI), and between (b) ΔSVITD and baseline values of transesophageal echocardiographically derived left ventricular end-diastolic area index (LVEDAI).
Table 1 Demographic data and preoperative risk factors
Parameters Value
Demographic data
Age (years) 65.7 ± 6.1
Sex (male/female; % female) 17/4 (19.0%)
BMI (kg/m2) 29.1 ± 2.9
LVEF (%) 52.9 ± 13.7
LVEDP (mmHg) 12.9 ± 5.7
Preoperative risk factors (n [%])
Diabetes 8 (38%)
Hypertension 17 (81%)
Smoking 6 (28%)
Hyperlipidemia 13 (62%)
Myocardial infarction 0 (0%)
Obesity 9 (43%)
History of stroke 2 (9.5%)
PVD 4 (19%)
PAH 0 (0%)
COPD 3 (14%)
Renal disease 4 (19%)
Data are expressed as mean ± standard deviation, or as frequency distributions (n) and simple percentages (%). BMI, body mass index; COPD = chronic obstructive pulmonary disease; LVEDP = left ventricular end-diastolic pressure; LVEF, left ventricular ejection fraction; PAH = pulmonary arterial hypertension; PVD = peripheral vascular disease.
Table 2 Hemodynamic variables at sample points T1 and T2
Variable T1 T2 p-value
HR (beats/min) 62.2 ± 10.7 60.7 ± 8.4 NS
MAP (mmHg) 70.1 ± 10.9 82.2 ± 9.8 <0.01
CVP (mmHg) 10.9 ± 2.2 13.4 ± 2.3 <0.01
PCWP (mmHg) 11.6 ± 2.6 15.2 ± 2.4 <0.01
MPAP (mmHg) 20.5 ± 2.8 23.9 ± 4.1 <0.01
SVRI (dyne· s/cm5·m2) 2878 ± 698 2540 ± 394 <0.01
CITD (l/min·m2) 1.70 ± 0.20 2.22 ± 0.23 <0.01
SVITD (ml/m2) 28.3 ± 6.4 37.2 ± 6.1 <0.01
CCI (l/min·m2) 1.71 ± 0.26 2.20 ± 0.23 <0.01
CEDVI (ml/m2) 95.4 ± 21.0 122.1 ± 24.4 <0.01
RVEF (%) 30.7 ± 9.1 31.7 ± 8.4 NS
SvO2 (%) 73.5 ± 4.3 77.1 ± 3.8 <0.01
LVEDAI (cm2/m2) 11.1 ± 4.1 12.8 ± 4.6 <0.01
FAC (%) 38.2 ± 9.6 37.9 ± 8.4 NS
Times T1 and T2 are before volume replacement and 12 min after volume replacement, respectively. CCI, continuous cardiac index; CEDVI, continuous right ventricular end-diastolic volume index; CI, cardiac index; CVP, central venous pressure; FAC, fractional area change; HR, heart rate; LVEDAI, left ventricular end-diastolic area index; MAP, mean arterial pressure; MPAP, mean pulmonary arterial pressure; PCWP, pulmonary capillary wedge pressure; SVI, stroke volume index; SVRI, systemic vascular resistance index; RVEF, right ventricular ejection fraction; SvO2, mixed venous oxygen saturation.
==== Refs
Richard C Warszawski J Anguel N Deye N Combes A Barnoud D Boulain T Lefort Y Fartoukh M Baud F Early use of the pulmonary artery catheter and outcomes in patients with shock and acute respiratory distress syndrome: a randomized controlled trial JAMA 2003 290 2713 2720 14645314 10.1001/jama.290.20.2713
Sandham JD Hull RD Brant RF Knox L Pineo GF Doig CJ Laporta DP Viner S Passerini L Devitt H A randomized, controlled trial of the use of pulmonary-artery catheters in high-risk surgical patients N Engl J Med 2003 348 5 14 12510037 10.1056/NEJMoa021108
Jacka MJ Cohen MM To T Devitt JH Byrick R The use of and preferences for the transesophageal echocardiogram and pulmonary artery catheter among cardiovascular anesthesiologists Anesth Analg 2002 94 1065 1071 11973164 10.1097/00000539-200205000-00003
Perel A The value of functional hemodynamic parameters in hemodynamic monitoring of ventilated patients Anaesthesist 2003 52 1003 1004 14992085 10.1007/s00101-003-0609-5
Reuter DA Felbinger TW Moerstedt K Weis F Schmidt C Kilger E Goetz AE Intrathoracic blood volume index measured by thermodilution for preload monitoring after cardiac surgery J Cardiothorac Vasc Anesth 2002 16 191 195 11957169 10.1053/jcan.2002.31064
Wiesenack C Prasser C Keyl C Rödig G Assessment of intrathoracic blood volume as an indicator of cardiac preload: single transpulmonary thermodilution technique versus assessment of pressure preload parameters derived from a pulmonary artery catheter J Cardiothorac Vasc Anesth 2001 15 584 588 11687999 10.1053/jcan.2001.26536
Tavernier B Makhotine O Lebuffe G Dupont J Scherpereel P Systolic pressure variation as a guide to fluid therapy in patients with sepsis-induced hypotension Anesthesiology 1998 89 1313 1321 9856704 10.1097/00000542-199812000-00007
Tousignant CP Walsh F Mazer CD The use of transesophageal echocardiography for preload assessment in critically ill patients Anesth Analg 2000 90 351 355 10648320 10.1097/00000539-200002000-00021
Perel A Assessing fluid responsiveness by the systolic pressure variation in mechanically ventilated patients Anesthesiology 1998 89 1309 1310 9856702 10.1097/00000542-199812000-00005
Michard F Boussat S Chemla D Anguel N Mercat A Lecarpentier Y Richard C Pinsky MR Teboul JL Relation between respiratory changes in arterial pulse pressure and fluid responsiveness in septic patients with acute circulatory failure Am J Respir Crit Care Med 2000 162 134 138 10903232
Berkenstadt H Margalit N Hadani M Friedman Z Segal E Villa Y Perel A Stroke volume variation as a predictor of fluid responsiveness in patients undergoing brain surgery Anesth Analg 2001 92 984 989 11273937
Reuter DA Felbinger TW Schmidt C Kilger E Goedje O Lamm P Goetz AE Stroke volume variations for assessment of cardiac responsiveness to volume loading in mechanically ventilated patients after cardiac surgery Intensive Care Med 2002 28 392 398 11967591 10.1007/s00134-002-1211-z
Reuter DA Kirchner A Felbinger TW Weis FC Kilger E Lamm P Goetz AE Usefulness of left ventricular stroke volume variation to assess fluid responsiveness in patients with reduced cardiac function Crit Care Med 2003 31 1399 1404 12771609 10.1097/01.CCM.0000059442.37548.E1
Marx G Cope T McCrossan L Swaraj S Cowan C Mostafa SM Wenstone R Leuwer M Assessing fluid responsiveness by stroke volume variation in mechanically ventilated patients with severe sepsis Eur J Anaesthesiol 2004 21 132 138 14977345 10.1017/S0265021504002091
Pinsky MR Probing the limits of arterial pulse contour analysis to predict preload responsiveness Anesth Analg 2003 96 1245 1247 12707113 10.1213/01.ANE.0000055821.40075.38
Wiesenack C Prasser C Rodig G Keyl C Stroke volume variation as an indicator of fluid responsiveness using pulse contour analysis in mechanically ventilated patients Anesth Analg 2003 96 1254 1257 12707116 10.1213/01.ANE.0000053237.29264.01
Cheatham ML Nelson LD Chang MC Safcsak K Right ventricular end-diastolic volume index as a predictor of preload status in patients on positive end-expiratory pressure Crit Care Med 1998 26 1801 1806 9824070
Diebel LN Wilson RF Tagett MG Kline RA End-diastolic volume. A better indicator of preload in the critically ill Arch Surg 1992 127 817 821 1524482
Durham R Neunaber K Vogler G Shapiro M Mazuski J Right ventricular end-diastolic volume as a measure of preload J Trauma 1995 39 218 223 7674388
Reuse C Vincent JL Pinsky MR Measurements of right ventricular volumes during fluid challenge Chest 1990 98 1450 1455 2245688
Michard F Teboul JL Predicting fluid responsiveness in ICU patients Chest 2002 121 2000 2008 12065368 10.1378/chest.121.6.2000
Wagner JG Leatherman JW Right ventricular end-diastolic volume as a predictor of the hemodynamic response to a fluid challenge Chest 1998 113 1048 1054 9554646
Yelderman M Quinn MD McKown RC Thermal safety of a filamented pulmonary artery catheter J Clin Monit 1992 8 147 149 1583553
Lazor MA Pierce ET Stanley GD Cass JL Halpern EF Bode RH Jr Evaluation of the accuracy and response time of STAT-mode continuous cardiac output J Cardiothorac Vasc Anesth 1997 11 432 436 9187990 10.1016/S1053-0770(97)90050-1
Groeneveld AB Berendsen RR Schneider AJ Pneumatikos IA Stokkel LA Thijs LG Effect of the mechanical ventilatory cycle on thermodilution right ventricular volumes and cardiac output J Appl Physiol 2000 89 89 96 10904039
Hoeper MM Tongers J Leppert A Baus S Maier R Lotz J Evaluation of right ventricular performance with a right ventricular ejection fraction thermodilution catheter and MRI in patients with pulmonary hypertension Chest 2001 120 502 507 11502650 10.1378/chest.120.2.502
Kraut EJ Owings JT Anderson JT Hanowell L Moore P Right ventricular volumes overestimate left ventricular preload in critically ill patients J Trauma 1997 42 839 845 9191665
Yu M Takiguchi S Takanishi D Myers S McNamara JJ Evaluation of the clinical usefullness of thermodilution volumetric catheters Crit Care Med 1995 23 681 686 7712758 10.1097/00003246-199504000-00016
Luecke T Roth H Herrmann P Joachim A Weisser G Pelosi P Quintel M Assessment of cardiac preload and left ventricular function under increasing levels of positive end-expiratory pressure Intensive Care Med 2004 30 119 126 12955175 10.1007/s00134-003-1993-7
Nelson LD Safcsak K Cheatham ML Block EF Mathematical coupling does not explain the relationship between right ventricular end-diastolic volume and cardiac output Crit Care Med 2001 29 940 943 11378601 10.1097/00003246-200105000-00006
Michard F Reuter DA Assessing cardiac preload or fluid responsiveness? It depends on the question we want to answer Intensive Care Med 2003 29 1396 12827236 10.1007/s00134-003-1846-4
Bendjelid K Romand JA Fluid responsiveness in mechanically ventilated patients: a review of indices used in intensive care Intensive Care Med 2003 29 352 360 12536268 10.1007/s00134-003-1777-0
Bouchard MJ Denault A Couture P Guertin MC Babin D Ouellet P Carrier M Tardif JC Poor correlation between hemodynamic and echocardiographic indexes of left ventricular performance in the operating room and intensive care unit Crit Care Med 2004 32 644 648 15090941 10.1097/01.CCM.0000108877.92124.DF
| 15987394 | PMC1175882 | CC BY | 2021-01-04 16:04:53 | no | Crit Care. 2005 Apr 1; 9(3):R226-R233 | utf-8 | Crit Care | 2,005 | 10.1186/cc3503 | oa_comm |
==== Front
Crit CareCritical Care1364-85351466-609XBioMed Central London cc35041598739310.1186/cc3504ResearchDaily enteral feeding practice on the ICU: attainment of goals and interfering factors Binnekade JM [email protected] R [email protected] P [email protected] EMH [email protected] Hann RJ [email protected] Research Nurse, Department of Intensive Care, Academic Medical Center, University of Amsterdam, Amsterdam, The Netherlands2 Intensivist, Department of Intensive Care, Academic Medical Center, University of Amsterdam, Amsterdam, The Netherlands3 Dietician, Department of Dietetics, Academic Medical Center, University of Amsterdam, Amsterdam, The Netherlands4 Gastroenterologist, Department of Gastroenterology, Academic Medical Center, University of Amsterdam, Amsterdam, The Netherlands5 Clinical Epidemiologist, Department of Clinical Epidemiology and Biostatistics, Academic Medical Center, University of Amsterdam, Amsterdam, The Netherlands2005 22 3 2005 9 3 R218 R225 8 12 2004 19 1 2005 31 1 2005 21 2 2005 Copyright © 2005 Binnekade et al.; licensee BioMed Central Ltd.This is an Open Access article distributed under the terms of the Creative Commons Attribution License (), which permits unrestricted use, distribution, and reproduction in any medium, provided the original work is properly cited.
Background
The purpose of this study was to evaluate the daily feeding practice of enterally fed patients in an intensive care unit (ICU) and to study the impact of preset factors in reaching predefined optimal nutritional goals.
Methods
The feeding practice of all ICU patients receiving enteral nutrition for at least 48 hours was recorded during a 1-year period. Actual intake was expressed as the percentage of the prescribed volume of formula (a success is defined as 90% or more). Prescribed volume (optimal intake) was guided by protocol but adjusted to individual patient conditions by the intensivist. The potential barriers to the success of feeding were assessed by multivariate analysis.
Results
Four-hundred-and-three eligible patients had a total of 3,526 records of feeding days. The desired intake was successful in 52% (1,842 of 3,526) of feeding days. The percentage of successful feeding days increased from 39% (124 of 316) on day 1 to 51% (112 of 218) on day 5. Average ideal protein intake was 54% (95% confidence interval (CI) 52 to 55), energy intake was 66% (95% CI 65 to 68) and volume 75% (95% CI 74 to 76). Factors impeding successful nutrition were the use of the feeding tube to deliver contrast, the need for prokinetic drugs, a high Therapeutic Intervention Score System category and elective admissions.
Conclusion
The records revealed an unsatisfactory feeding process. A better use of relative successful volume intake, namely increasing the energy and protein density, could enhance the nutritional yield. Factors such as an improper use of tubes and feeding intolerance were related to failure. Meticulous recording of intake and interfering factors helps to uncover inadequacies in ICU feeding practice.
See related commentary
==== Body
Introduction
Protein energy malnutrition is a major problem in severely ill hypercatabolic patients in the intensive care unit (ICU) [1]. Early initiation of enteral nutrition has proved to be beneficial, with significant positive effects on septic complications, and has been shown to improve the outcome when compared with parenteral nutrition. Enteral nutrition guarantees the preservation of gut mass and prevents increased gut permeability to bacteria and toxins [2-5]. In addition, the gut-associated lymphoid tissue is better maintained [4].
Over the years, enteral nutrition has improved with regard to techniques, materials and composition, and has gained popularity because of its lower cost and lower rate of complications compared with parenteral nutrition. This is also reflected in our intensive care by an increased use of enteral nutrition from 16.7% of total patient days in 1992 to 53.8% in 2001, and a slightly decreased use of parenteral nutrition, from 19% of total patient days in 1992 to 14% in 2001.
Although this large increase in enteral feeding days has to be considered a step forward, these figures do not show the actual intake of energy and nutrients per patient; that is, the adequacy of feeding. Despite the attention given to the practice of enteral nutrition in daily rounds by intensivists and ICU nurses, we were not adequately and accurately informed as to the adequacy of our feeding practice [6]. Confronted with a growing number of enterally fed patients we decided to develop a daily record, aimed at obtaining a continuous and long-term overall insight into the volume, energy content and amount of proteins administered to and actually received by the patient. The objective of this study was to evaluate the success of enteral nutrition in our ICU and to report the influence of factors presumed to interfere and, being part of the record, to achieve an optimal nutritional intake.
Materials and methods
Setting
The study was conducted in a 30-bed intensive care unit with access to patients of all specialties at the Academic Medical Center in Amsterdam, a tertiary care university teaching hospital with 1,000 beds.
Feeding process
Standard feeding practice involved the continuous administration of enteral feeding solutions over 24 hours. Although a standard feeding protocol was in use (see Additional file 1) the flow rate was often adjusted according to the understanding of the intensivist.
Patients started feeding at 500 ml per day with a build-up of 500 ml per 24 hours until the individually determined intake in terms of volume, proteins and calories was reached. Given an uneventful course a patient would achieve an intake of 2,000 ml within 5 days. However, to compensate for interruptions of feeding, the intake was targeted at a 20% higher volume. The optimal feeding target of 2,000 kcal per 24 hours therefore became 2,400 kcal per 24 hours after adjustment.
Data collection
Patients admitted to the ICU and receiving enteral nutrition for at least 48 hours were eligible. The study duration for each patient was limited to 30 days. In this retrospective database study we extracted the daily records of enterally fed patients over a period of 1 year. Records containing a single oral-nutrition or total parenteral-nutrition day, or records that lacked a prescription of desired intake, were excluded from the analysis.
Feeding factors assumed to interfere with enteral nutrition and noted in the record were as follows: first, the type of feeding tube (gastric tube, duodenal tube, percutaneous endoscopic gastrostomy, or needle catheter jejunostomy (NCJ)); second, the type of formula with different energy content (100 to 204 kcal/100 ml) and protein content (4 to 7 g/100 ml) and normal or predigested semi-elemental form; third, gastric retention; fourth, therapeutic interventions (mechanical ventilation, endotracheal tube in situ, extubation/intubation, spontaneous respiration, tracheostomy, continuous veno-venous haemofiltration, prone position, and preparation for computed tomography scan); and fifth, medication (lactulose, cisapride, midazolam–morphine, morphine, propofol, vasopressors, inotropics and pantoprazol).
The feeding record was coupled to other databases to extract data on gender, age, length of stay and referral specialty in the ICU, the Acute Physiology and Chronic Health Evaluation score (APACHE II) and the therapy intensity with the Therapeutic Intervention Score System (TISS). The TISS scores were calculated for each patient and subdivided into four categories, classifying the patient's need for ICU care: in category 1 the score was less than 10 points (no need for ICU care); 2, a score of 10 or more to less than 20 points (physiologically stable condition with prophylactic overnight observation); 3 a score of more than 20 to less than 40 points (physiologically stable but requiring intensive nursing and monitoring); and 4, a score of more than 40 points (unstable condition requiring intensive physician and nursing care) [7]. The APACHE II [8] was scored upon admission (within 24 hours); APACHE II scores could range from 0 to 71, with higher scores indicating a more severe illness.
Reliability of the record
As the record had to be filled in by several staff members, its reliability had to be tested. An interobserver study was performed between two regular keepers of the record (a dietician and an intensivist). Nursing charts of 42 feeding days for 14 patients (3 days per patient) were evaluated by three different observers and the data were entered into the record.
Analyses
Descriptive statistics were used to characterize patients. Successful intake was defined as a patient's receiving 90% or more of the prescribed amount of feeding. The difference between the prescribed amount and the tube feeding actually administered was expressed as a percentage and its associated 95% confidence interval (95% CI). The percentages (95% CIs) of realized energy and protein needs were based on a ideal 30 kcal per kg of body weight [9] and 1.5 g of protein per kg of body weight [10,11], respectively. The volume, energy and protein intake were stratified by type of formula and arranged by type of enteral route (total of 28 strata). Patients with zero intake but having a feeding prescription remained in the analysis.
Univariate analysis was performed to assess determinants of successful intake with regard to patients and feeding factors. Before inclusion into the model the independence of these explanatory variables had to be determined. The most common value of the categories (referral specialty, type of feeding tube and type of formula) were used as reference category (odds ratio of 1). Each category of the predictor variable was then compared with the reference category for categorical variables.
Significant variables in the univariate analysis (P ≤ 0.10) from patients and feeding factors were forced into the multivariate logistic regression model (enter method).
The results of the univariate analysis were also compared for the data set of the complete feeding period and a data set of the first three feeding days. Significant differences might show influences of a skewed duration of feeding.
Statistical uncertainty was expressed as 95% CI. Data were analyzed in SPSS version 11.5.
Results
In 2001, 1,479 patients were admitted to the ICU. After the removal of elective admissions with a limited stay of less than 48 hours, the crude data set contained 5,859 feeding days. Because the analysis was limited to 30 days of ICU stay, 5,017 days remained. The removal of feeding days with one single day of oral or total parenteral nutrition and the removal of patients who did not receive a prescription for enteral feeding resulted in 3,526 days to be analyzed in 403 patients.
There was a significant difference between neurosurgery and the other specialties for length of ICU stay. APACHE II scores did not differ between medical and neurosurgery patients. Medical patients had the highest APACHE II score, significantly higher than those of surgical and cardiac surgical patients (Table 1).
Reliability of the record
The test of reliability showed an intra-class correlation (two-way random model) of 0.98 (95% CI 0.96 to 0.99).
Success of enteral nutrition
During the build-up phase of feeding, the number of successful feeding days increased from 39% (124 of 316) on day 1 to 51% (112 of 218) on day 5. At discharge from the ICU only 4% (14 of 371) of patients received 100 ml/hour or more enteral nutrition. Twenty-five percent (93 of 371) of patients left the ICU with an intake of 80 ml/hour, whereas 71% (264 of 371) of patients received 60 ml/hour or less. Thirty-three patients stayed for longer than 30 days in the ICU; food intake on discharge was therefore not analyzed.
The percentage of successful intake and ideal energy and ideal protein calculated for each type of formula and for each type of enteral route showed an overall picture of deficiency. Of the 28 strata, 21 were analyzable (Table 2). Ten strata showed the highest percentage for volume, another ten for energy and only one for protein. In eight strata protein turned out to be less important than volume, whereas in seven strata protein was less important than the percentage of energy (Table 2).
Factors interfering with successful administration of enteral feed
Tube location
The percentage of days with successful feeding was smallest for gastric tubes and greatest for duodenal/jejunal tubes (Table 3). The NCJ had significantly more successful feeding days than the duodenal tube; the difference was 19% (95% CI 27 to 10) (Table 3).
Gastric retention
Patients fed by duodenal tube had the highest gastric retention, with a mean of 558 (95% CI 523 to 593) ml/24 hours. The mean gastric retention among patients with a gastric tube was 159 (148 to 170) ml/24 hours. Of these, a mean of 121 (110 to 132) ml gastric retention over a 24-hour period was discarded by the nurse instead of being given back to the patient. On the assumption that the gastric retention in patients with a gastric tube contained mainly tube feeding, the amount of nutrition delivered would decline to a mean of 1,066 (1,034 to 1,097) ml/24 hours. In this scenario, the removal of gastric retention fluids caused a decline in the percentage of successful feeding of 6% (95% CI 4 to 10%) to a 42% (1,024 of 2,455) success rate. Although the protocol dictated that gastric retention volumes of less than 200 ml in 6 hours had to be given back, 34% (266 of 791) of gastric retention volumes of less than 200 ml/hour were discarded.
TISS scores
Category 3 TISS scores were present on 66% (2,349 of 3,577) and category 4 TISS scores on 31% (1,106 of 3,577) of patient days. Among category 3 TISS patients the success rate of feeding (at least 90% intake) was 55% (1,285 of 2,349) in comparison with a 45% (498 of 1,106) success rate of feeding among category 4 patient days; this is a difference of 10% (95% CI 6 to 13%).
Multivariate analysis
Because of significant collinearity between mechanical ventilation and other variables, such as endotracheal tube, extubation, intubation, spontaneous respiration and tracheostomy, only mechanical ventilation was included in the analysis.
A comparison of the results of the univariate analysis between the complete data set and a subset of the first three feeding days did not reveal any important differences.
Univariate analysis of 32 potential determinants of successful intake revealed 12 significant variables (P ≤ 0.10; not presented). The subsequent multivariate logistics regression analysis resulted in 11 significant variables (P ≤ 0.05) (Table 4).
Both the NCJ and semi-elemental formula showed the odds ratios as to successful feeding, 3.32 and 3.02, respectively, both to be interpreted against the reference, i.e. the gastric tube and standard feeding formula (Table 4). In addition, a gastric retention of less than 200 ml and a length of stay above the median was related to improved success of feeding. Of the remaining interventions, the administration of contrast via the tube, the need for prokinetic drugs, TISS and elective admission showed an adverse effect on the success of feeding (Table 4).
Discussion
With the use of a meticulous, daily record of the ICU feeding practice we evaluated the feasibility of prescribed enteral feeding for a 1-year period. The prescribed nutritional volume turns out to be hardly feasible in the patients involved in our study. When actual intake is compared with ideal energy and protein needs, protein shows the largest overall deficit. Current feeding practice (including the 5-day build-up schedule for enteral nutrition) fails to provide ICU patients with adequate nutrition.
Other studies found comparably bad results. A prospective cohort study among 99 ICU patients found that only the half of patients achieved tolerance of the feeding regime (90% of estimated energy for more than 48 hours) [12].
Better results were found in a multicenter prospective study that followed 193 patients during 1,929 patient days. An average of 76% of the prescribed feed was delivered to the patient. They also concluded that using well-defined protocols significantly improved the intake [13].
A prospective study in ICUs and coronary care units revealed that barely one-half of the 44 patients studied met their caloric requirements because of underordering by physicians and reduced delivery arising from frequent and inappropriate cessation of feeding [14].
Another prospective study found also a low caloric intake in 51 enterally fed ICU patients for whom 78% of the mean caloric amount required was prescribed and 71% was actually delivered [15].
An audit of 40 ICU patients for which the ideal feeding target was calculated by the Harris–Benedict equation. Patients received only 51% of these energy requirements during the 7-day study period [16].
A cross-sectional survey of 66 responding dieticians of ICUs revealed that among patients receiving enteral nutrition only 58% met their prescribed energy and protein needs [17].
Although we were aware of these studies, we did not expect this result until we kept these records. Despite having at our disposal an enteral feeding protocol and despite daily bedside consultations with the intensivist, nurse and dietician, only 50% of the enterally fed patients achieved a successful intake at the end of a 5-day feeding build-up scheme. Although a further improvement in intake occurred as the ICU stay was prolonged, the overall success per feeding day remained low during the ICU stay. Apparently, implementation of a protocol, once it has been set out and accepted, is difficult and needs more attention [13,18,19].
The feeding with a NCJ resulted in odds ratios that favor this enteral route over the gastric tube. In addition semi-elemental formula seemed to be three times better than standard formula (Table 4). In part, this might have been confounded by the use of either duodenal tubes or NCJ, because the NCJ showed the fewest problems in use. Because of this and because it concerned a small group of patients, we cannot unambiguously recommend semi-elemental formula although others have done so [20,21].
Disordered upper gastrointestinal tract motility frequently occurs in ICU patients [22], yet the gastric tube remains the first and simplest choice and the easiest way of starting enteral nutrition. This does not detract from the significant number of patients who have to be switched to a duodenal tube because of persisting gastric retention. We also found that nurses tended to overestimate gastric retention as a risk factor and, more importantly, violated the protocol by discarding a gastric retention volume of less than 200 ml over 6 hours. This behavior might be the result of a misplaced ambition to achieve safer care. Although the measurement of gastric retention is an important tool for guaranteeing safe enteral feeding, no difference is reported between gastric tube and duodenal tube use among ICU patients in terms of aspiration and nosocomial pneumonia. Moreover, the insignificant role of gastric retention levels of up to 250 ml has been reported [23-26].
Using the feeding tube to administer contrast for a CT scan precludes the use of the tube for administering nutrition. In general, a high therapy intensity reflected by a high TISS score indicated a more difficult feeding practice because the subject was more critically ill. This might also reflect the lower priority given in the care routine for optimal continuation of the feeding process in comparison with the efforts taken to support patients in need of ventilation and assisted circulation.
Improvement of nutritional intake can be achieved by implementing simple rules, such as limiting the interruption of enteral nutrition because of diagnostic or therapeutic interventions, a quick replacement of accidentally removed tubes, and giving back gastric retention of less than 250 ml [14,27,28].
Whereas a high TISS score did seem to interfere with the administration of enteral nutrition, the severity of illness did not. It took several days for 50% of the patients to achieve an optimal intake, which to some extent might reflect the unstable physical condition of the ICU patient. This is also shown by the relationship between success of feeding and prolonged ICU stay.
A limitation of this study is that we did not collect or analyze a nutritional anamnesis or patient outcome data. We have focused on measurable aspects of feeding practice. It will be worthwhile to expand the continuous recording to include a (nutritional) anamnesis of the patient. Improving the information load of this record would also require more information about outcome.
Conclusion
Evaluation of feeding practices has revealed otherwise unnoticed, yet disappointing, results. Although the recording process in itself does not improve feeding practice it might lead to the recognition that the patient is underfed while being fed and that ways have to be found to improve feeding practice, namely by implementing protocols for feeding and gastric retention measurements.
Key messages
• A long-term recording of the ICU nutritional intake revealed an unsatisfactory enteral feeding process.
• Factors such as an improper use of tubes and feeding intolerance related to failure of nutritional intake.
• Better use of relative successful volume intake by increasing energy and protein density could enhance the nutritional yield.
Abbreviations
APACHE = Acute Physiology and Chronic Health Evaluation; CI = confidence interval; CT = computed tomography; ICU = intensive care unit; NCJ = needle catheter jejunostomy; TISS = Therapeutic Intervention Score System.
Competing interests
The author(s) declare that they have no competing interests.
Authors' contributions
JMB built the database, analyzed the data and wrote the article. RT performed the data collection, performed the interobserver study and co-wrote the article. PB performed data collection and participated in the interobserver study. EMHMV supervised the writing of the article and co-wrote the article. RJH supervised the statistical analysis and the final draft of the article. All authors read and approved the final manuscript.
Supplementary Material
Additional File 1
A pdf file containing a table listing the protocol outline.
Click here for file
Figures and Tables
Table 1 Characteristics of ICU patients recorded in the 1-year enteral feeding practice (n = 404)
Characteristic Value
Male patients, % 63 (256 of 404)
Age, years (mean ± SD) 60 ± 17
Length of stay, days (median and IQ)
Overall 8 (5/16)
Medical (n = 112) 9 (6/18)
Surgical (n = 117) 8 (5/16)
Cardiac surgery (n = 122) 7 (5/14)
Neurosurgery (n = 53) 10 (6/16)
APACHE II score (mean ± SD)
Overall 18 ±7
Medical 23 ± 7
Surgical 17 ± 6
Cardiac surgery 16 ± 5
Neuro(surgery) 20 ± 6
IQ, interquartile range.
Table 2 Percentage of volume of formula actually delivered (intake/prescribed)
Type of formula (kcal/protein) Enteral route n Successful feeding days, % (95% CI)
Volume Energy Protein
Standard (100/4) Gastric tube 1,309 66 (64–68) 47 (45–50) 37 (35–38)
(n = 1,760) PEG 11 87 (67–107) 86 (44–128) 51 (33–70)
Duodenal tube 392 81 (79–84) 71 (66–75) 52 (50–55)
NCJ 48 85 (76–93) 74 (64–84) 58 (50–66)
Energy+ (150/6) Gastric tube 216 79 (75–83) 95 (89–101) 78 (73–83)
(n = 359) PEG - - - -
Duodenal tube 143 81 (77–86) 102 (96–109) 82 (77–86)
NCJ - - - -
Energy+/Protein+ (204/7) Gastric tube 155 79 (74–83) 82 (74–90) 59 (53–65)
(n = 80) PEG - - - -
Duodenal tube 99 81 (76–87) 85 (76–93) 60 (54–66)
NCJ 14 90 (82–99) 70 (52–84) 91 (68–114)
Fiber+ (106/4.1) Gastric tube 47 71 (61–81) 67 (56–78) 52 (43–61)
(n = 69) PEG - - - -
Duodenal tube 22 71 (56–86) 47 (35–58) 37 (29–45)
NCJ - - - -
Immunologically active (100/5.6) Gastric tube 333 72 (69–76) 51 (46–55) 54 (50–58)
(n = 441) PEG 1 - - -
Duodenal tube 95 80 (74–86) 73 (63–83) 79 (70–89)
NCJ 12 78 (59–97) 66 (48–83) 73 (54–92)
Semi-elemental (100/4) Gastric tube 14 89 (73–105) 97 (68–126) 58 (42–74)
(n = 90) PEG - - - -
Duodenal tube 60 91 (86–96) 63 (54–171) 52 (46–57)
NCJ 16 88 (69–106) 90 (65–116) 71 (51–90)
Standard/Energy+ (125/5) Gastric tube 350 82 (79–85) 85 (81–89) 68 (65–72)
(n = 539) PEG 9 75 (50–101) 77 (52–102) 62 (42–81)
Duodenal tube 168 85 (81–89) 90 (85–95) 74 (70–79)
NCJ 12 89 (71–107) 138 (110–166) 110 (88–133)
Results are percentages of realized 'ideal' energy intake (30 kcal/kg body weight) and percentage of realized 'ideal' protein intake (1.5 g/kg) stratified by tube and by type of formula. CI, confidence interval; NCJ, needle catheter jejunostomy; PEG, percutaneous endoscopic gastrostomy.
Table 3 Days of successful intake divided by feeding route
Enteral route Patients Days of successful intake, %
Gastric tube 383 49 (1,188 of 2,424)
Duodenal tube [previous gastric tube]* [116] 58 (564 of 979)
Needle catheter jejunostomy 17 76 (78 of 102)
Percutaneous endoscopic gastrostomy 3 57 (12 of 21)
Overall success of feeding intake 403 52 (1,842 of 3,526)
*All patients with a duodenal feeding tube were previously fed by gastric tube.
Table 4 Multivariate analysis of patient and feeding factors for the success of feeding intake
Feeding and patient factors OR 95% CI
Duodenal tube* 1.44 1.21–1.73
Needle catheter jejunostomy* 3.32 2.05–5.38
Semi-elemental formula (100/4)† 3.02 1.75–5.21
Standard/Energy+ formula (125/5)† 1.62 1.31–1.99
Mechanical ventilation 1.63 1.27–2.09
Contrast via tube to prepare for CT scan 0.34 0.22–0.55
Gastric retention (<200 ml/>200 ml) 1.51 1.29–1.78
Cisapride 0.83 0.71–0.96
TISS category per patient day 0.84 0.74–0.96
Elective admission 0.81 0.69–0.94
Length of stay in intensive care unit 1.53 1.28–1.82
Success of feeding intake was defined as feeding of more than 90% of the prescribed formula. CI, confidence interval; CT, computed tomography; OR, odds ratio; TISS, Therapeutic Intervention Score System. An odds ratio of more than 1 indicates improved success of feeding. *Compared with reference gastric tube (odds ratio of 1). †Compared with the reference standard feeding (odds ratio of 1); numbers in parentheses are kcal/protein.
==== Refs
Jolliet P Pichard C Biolo G Chiolero R Grimble G Leverve X Nitenberg G Novak I Planas M Preiser JC Enteral nutrition in intensive care patients: a practical approach. Working Group on Nutrition and Metabolism, ESICM. European Society of Intensive Care Medicine Intensive Care Med 1998 24 848 859 9757932 10.1007/s001340050677
Kompan L Kremzar B Gadzijev E Prosek M Effects of early enteral nutrition on intestinal permeability and the development of multiple organ failure after multiple injury Intensive Care Med 1999 25 157 161 10193541 10.1007/s001340050809
Marik PE Zaloga GP Early enteral nutrition in acutely ill patients: a systematic review Crit Care Med 2001 29 2264 2270 11801821 10.1097/00003246-200112000-00005
Minard G Kudsk KA Is early feeding beneficial? How early is early? New Horiz 1994 2 156 163 7922440
Perez J Dellinger RP Other supportive therapies in sepsis Intensive Care Med 2001 27 Suppl 1 S116 S127 11307367
Jonkers CF Prins F Van Kempen A Tepaske R Sauerwein HP Towards implementation of optimum nutrition and better clinical nutrition support Clin Nutr 2001 20 361 366 11478835 10.1054/clnu.2001.0470
Cullen DJ Civetta JM Briggs BA Ferrara LC Therapeutic intervention scoring system: a method for quantitative comparison of patient care Crit Care Med 1974 2 57 60 4832281
Knaus WA Draper EA Wagner DP Zimmerman JE APACHE II: a severity of disease classification system Crit Care Med 1985 13 818 829 3928249
Stroud M Duncan H Nightingale J Guidelines for enteral feeding in adult hospital patients Gut 2003 52 Suppl 7 vii1 vii12 14612488
Shaw JH Wildbore M Wolfe RR Whole body protein kinetics in severely septic patients. The response to glucose infusion and total parenteral nutrition Ann Surg 1987 205 288 294 3103555
Ishibashi N Plank LD Sando K Hill GL Optimal protein requirements during the first 2 weeks after the onset of critical illness Crit Care Med 1998 26 1529 1535 9751589 10.1097/00003246-199809000-00020
Heyland D Cook DJ Winder B Brylowski L Van deMark H Guyatt G Enteral nutrition in the critically ill patient: a prospective survey Crit Care Med 1995 23 1055 1060 7774216 10.1097/00003246-199506000-00010
Adam S Batson S A study of problems associated with the delivery of enteral feed in critically ill patients in five ICUs in the UK Intensive Care Med 1997 23 261 266 9083227 10.1007/s001340050326
McClave SA Sexton LK Spain DA Adams JL Owens NA Sullins MB Blandford BS Snider HL Enteral tube feeding in the intensive care unit: factors impeding adequate delivery Crit Care Med 1999 27 1252 1256 10446815 10.1097/00003246-199907000-00003
De Jonghe B Appere-De-Vechi C Fournier M Tran B Merrer J Melchior JC Outin H A prospective survey of nutritional support practices in intensive care unit patients: what is prescribed? What is delivered? Crit Care Med 2001 29 8 12 11176150 10.1097/00003246-200101000-00002
De B Chapman M Fraser R Finnis M De Keulenaer B Liberalli D Satanek M Enteral nutrition in the critically ill: a prospective survey in an Australian intensive care unit Anaesth Intensive Care 2001 29 619 622 11771607
Heyland DK Schroter-Noppe D Drover JW Jain M Keefe L Dhaliwal R Day A Nutrition support in the critical care setting: current practice in canadian ICUs – opportunities for improvement? JPEN J Parenter Enteral Nutr 2003 27 74 83 12549603
Parker D Lawton R Judging the use of clinical protocols by fellow professionals Soc Sci Med 2000 51 669 677 10975227 10.1016/S0277-9536(00)00013-7
Krishnan JA Parce PB Martinez A Diette GB Brower RG Caloric intake in medical ICU patients: consistency of care with guidelines and relationship to clinical outcomes Chest 2003 124 297 305 12853537 10.1378/chest.124.1.297
Meredith JW Ditesheim JA Zaloga GP Visceral protein levels in trauma patients are greater with peptide diet than with intact protein diet J Trauma 1990 30 825 828 2116533
Ziegler F Ollivier JM Cynober L Masini JP Coudray-Lucas C Levy E Giboudeau J Efficiency of enteral nitrogen support in surgical patients: small peptides v non-degraded proteins Gut 1990 31 1277 1283 2123819
Ritz MA Fraser R Tam W Dent J Impacts and patterns of disturbed gastrointestinal function in critically ill patients Am J Gastroenterol 2000 95 3044 3052 11095317 10.1016/S0002-9270(00)01969-9
Montejo JC Grau T Acosta J Ruiz-Santana S Planas M Garcia-De-Lorenzo A Mesejo A Cervera M Sanchez-Alvarez C Nunez-Ruiz R Multicenter, prospective, randomized, single-blind study comparing the efficacy and gastrointestinal complications of early jejunal feeding with early gastric feeding in critically ill patients Crit Care Med 2002 30 796 800 11940748 10.1097/00003246-200204000-00013
Esparza J Boivin MA Hartshorne MF Levy H Equal aspiration rates in gastrically and transpylorically fed critically ill patients Intensive Care Med 2001 27 660 664 11398691 10.1007/s001340100880
Heyland DK Drover JW MacDonald S Novak F Lam M Effect of postpyloric feeding on gastroesophageal regurgitation and pulmonary microaspiration: results of a randomized controlled trial Crit Care Med 2001 29 1495 1501 11505114 10.1097/00003246-200108000-00001
Montecalvo MA Steger KA Farber HW Smith BF Dennis RC Fitzpatrick GF Pollack SD Korsberg TZ Birkett DH Hirsch EF Nutritional outcome and pneumonia in critical care patients randomized to gastric versus jejunal tube feedings. The Critical Care Research Team Crit Care Med 1992 20 1377 1387 1395657
Mallampalli A McClave SA Snider HL Defining tolerance to enteral feeding in the intensive care unit Clin Nutr 2000 19 213 215 10952791 10.1054/clnu.2000.0137
Spain DA McClave SA Sexton LK Adams JL Blanford BS Sullins ME Owens NA Snider HL Infusion protocol improves delivery of enteral tube feeding in the critical care unit JPEN J Parenter Enteral Nutr 1999 23 288 292 10485441
| 15987393 | PMC1175883 | CC BY | 2021-01-04 16:04:52 | no | Crit Care. 2005 Mar 22; 9(3):R218-R225 | utf-8 | Crit Care | 2,005 | 10.1186/cc3504 | oa_comm |
==== Front
Crit CareCritical Care1364-85351466-609XBioMed Central London cc35081598739610.1186/cc3508ResearchEffect of ventilator-associated tracheobronchitis on outcome in patients without chronic respiratory failure: a case–control study Nseir Saad [email protected] Pompeo Christophe 2Soubrier Stéphane 1Lenci Hélène 3Delour Pierre 3Onimus Thierry 1Saulnier Fabienne 1Mathieu Daniel 3Durocher Alain 11 Intensive Care Unit, Calmette Hospital, Regional University Centre, and Medical Assessment Laboratory, EA 3614, Lille II University, Lille, France2 Medical Assessment Laboratory, EA 3614, Lille II University, Lille, France3 Intensive Care Unit, Calmette Hospital, Regional University Centre, Lille, France2005 31 3 2005 9 3 R238 R245 26 10 2004 9 2 2005 16 2 2005 24 2 2005 Copyright © 2005 Nseir 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
Our objective was to determine the effect of ventilator-associated tracheobronchitis (VAT) on outcome in patients without chronic respiratory failure.
Methods
This was a retrospective observational matched study, conducted in a 30-bed intensive care unit (ICU). All immunocompetent, nontrauma, ventilated patients without chronic respiratory failure admitted over a 6.5-year period were included. Data were collected prospectively. Patients with nosocomial pneumonia, either before or after VAT, were excluded. Only first episodes of VAT occurring more than 48 hours after initiation of mechanical ventilation were studied. Six criteria were used to match cases with controls, including duration of mechanical ventilation before VAT. Cases were compared with controls using McNemar's test and Wilcoxon signed-rank test for qualitative and quantitative variables, respectively. Variables associated with a duration of mechanical ventilation longer than median were identified using univariate and multivariate analyses.
Results
Using the six criteria, it was possible to match 55 (87%) of the VAT patients (cases) with non-VAT patients (controls). Pseudomonas aeruginosa was the most frequently isolated bacteria (34%). Although mortality rates were similar between cases and controls (29% versus 36%; P = 0.29), the median duration of mechanical ventilation (17 days [range 3–95 days] versus 8 [3–61 days]; P < 0.001) and ICU stay (24 days [range 5–95 days] versus 12 [4–74] days; P < 0.001) were longer in cases than in controls. Renal failure (odds ratio [OR] = 4.9, 95% confidence interval [CI] = 1.6–14.6; P = 0.004), tracheostomy (OR = 4, 95% CI = 1.1–14.5; P = 0.032), and VAT (OR = 3.5, 95% CI = 1.5–8.3; P = 0.004) were independently associated with duration of mechanical ventilation longer than median.
Conclusion
VAT is associated with longer durations of mechanical ventilation and ICU stay in patients not suffering from chronic respiratory failure.
See related commentary
==== Body
Introduction
Nosocomial lower respiratory tract infections are the most common nosocomial infections in the intensive care unit (ICU) [1]. Although several studies have investigated nosocomial pneumonia, few evaluated ventilator-associated tracheobronchitis (VAT).
VAT is a common nosocomial infection among mechanically ventilated patients. VAT rates of 3.7–10.6% have been reported in the literature [2-4]. In a previous descriptive prospective cohort study conducted in 2128 patients [4], our group demonstrated that VAT was associated with increased durations of mechanical ventilation and ICU stay. However, two major limitations of the study prevented us from drawing definite conclusions: absence of adjustment for duration of mechanical ventilation before the occurrence of VAT; and inclusion of patients with and patients without chronic respiratory failure. Therefore, we performed a retrospective case–control study to assess the effect of VAT on outcomes in patients without chronic respiratory failure.
Methods
This retrospective case–control study was conducted in our 30-bed ICU from March 1993 to September 1999. Because it was observational, institutional review board approval was not required, which is in accordance with institutional review board regulations.
All immunocompetent, nontrauma patients without chronic respiratory failure who were intubated and ventilated for more than 48 hours were eligible. Patients with chronic respiratory failure, trauma patients, patients who were not ventilated or ventilated for less than 48 hours, patients who received only noninvasive pressure ventilation, patients with tracheostomy at ICU admission and immunocompromised patients were not eligible. Patients who developed nosocomial pneumonia, before or after the occurrence of VAT, were excluded. The patients included in the present study were also included in our previous prospective observational study of VAT [4], representing 5% of the 2128 patients included in the previous study.
Patients were intubated via either the oral or the nasal route, according to clinical status and preference of the physician in charge. The oropharyngeal cavity was cleaned four times daily with chlorhexidine solution. Continuous subglottic suctioning was not utilized. The ventilator circuit was not changed routinely. In all patients a heat–moisture exchanger was positioned between the Y-piece and the patient; the heat–moisture exchangers were changed every 48 hours, or more frequently if they were visibly soiled. No patient received inhaled antibiotics. Patients were kept in a semirecumbent position during most of their period of mechanical ventilation. Sedation and weaning procedures were done at the discretion of the physician in charge. No systematic stress ulcer prophylaxis and no selective digestive decontamination was given. Tracheal aspiration was performed by nurses every 3 hours and whenever necessary.
Throughout the study, endotracheal aspirates for quantitative bacterial cultures were obtained routinely on admission, weekly thereafter, and whenever VAT or ventilator-associated pneumonia (VAP) was suspected. Antimicrobial therapy for VAT was at discretion of the physician in charge.
All data were collected prospectively. VAT episodes were identified by prospective surveillance of nosocomial infections. Only first episodes of VAT occurring more than 48 hours after initiation of mechanical ventilation were included. 'Cases' are VAT patients, and 'controls' are patients without VAT. Tracheobronchitis was defined using all of the following criteria: fever (>38°C) with no other recognizable cause; new or increased sputum production; positive (≥ 106 colony-forming units/ml) endotracheal aspirate culture [5], yielding a new bacteria; and no radiographic evidence of nosocomial pneumonia. In patients with abnormal chest radiograph at admission, the absence of new or progressive radiographic infiltrates was required. To define nosocomial pneumonia, a second set of criteria developed by the US Centers for Disease Control and Prevention was used [6]. Other nosocomial infections were defined using the Centers for Disease Control and Prevention criteria [6].
Antimicrobial therapy was deemed adequate when at least one antibiotic active in vitro on all organisms causing VAT was administrated at an appropriate dosage within the first 48 hours after VAT was identified. Chronic respiratory failure was defined by the presence of chronic obstructive pulmonary disease [7] or chronic restrictive pulmonary disease diagnosed on the basis of history, physical examination, chest radiography and respiratory function tests. Immunosupression was defined as the presence of neutropenia (leucocyte count <1000/μL or neutrophils <500/μL), long-term corticosteroid therapy (≥ 0.5 mg/kg per day for more than 1 month), or HIV infection (CD4+ cell count <50/μL for the previous 6 months). Multidrug-resistant bacteria were defined as methicillin-resistant Staphylococcus aureus, ceftazidime or imipenem-resistant Pseudomonas aeruginosa, Acinetobacter baumannii, extended-spectrum β-lactamase-producing Gram-negative bacilli, and Stenotrophomonas maltophilia. Prior antibiotic treatment was defined as any antibiotic treatment over the 2 weeks preceding ICU admission. Outcomes evaluated included ICU mortality, and durations of mechanical ventilation and ICU stay.
Each case patient was matched to one control patients according to all the following criteria: duration of mechanical ventilation before VAT occurrence (a control patient had to have been mechanically ventilated for at least as long as a case patient had before they developed VAT); primary diagnosis for admission; category of admission (medical/surgical); Simplified Acute Physiology Score II on admission (± 5 points) [8]; age (± 5 years); and date of admission (when more than one potential control was well matched to a case).
Statistical analysis
SPSS software (SPSS Institute Inc., Chicago, IL, USA) was used to analyze the data. Cases were compared with controls using McNemar's test for qualitative variables, and Wilcoxon's signed-rank test for quantitative variables.
Because the distribution of duration of mechanical ventilation was skewed, we first determined the median duration of mechanical ventilation in cases and controls, and then we performed univariate and multivariate analyses to identify those variables associated with duration of mechanical ventilation longer than median. The following variables were included in univariate analysis: age, sex, Simplified Acute Physiology Score II on admission, transfer from other wards, diabetes mellitus, primary diagnosis for admission, organ failures [9], antibiotic use, tracheostomy, VAT related to multidrug-resistant bacteria, and VAT. A stepwise logistic regression, including significant (P < 0.05) variables, was used to determine which variables were independently associated with duration of mechanical ventilation longer than median.
In order to determine the impact of antibiotic administration on VAT patient outcome, case patients receiving adequate antibiotic treatment were compared with those who received inadequate antibiotic treatment.
Proportions were compared using the χ2 test or the Fisher's exact test where appropriate; continuous variables were compared using the Mann–Whitney U-test.
Results
A total of 928 patients were eligible, 136 (14%) of whom were excluded because they developed nosocomial pneumonia before VAT. Seventy (8%) first episodes of VAT were diagnosed in the 792 remaining patients. Seven of the 70 patients (10%) were excluded because they subsequently developed nosocomial pneumonia. Using the six criteria outlined above (see Methods), it was possible to match 55 (87%) of the VAT patients without prior or subsequent nosocomial pneumonia (cases) with non-VAT patients (controls; Fig. 1).
Before ICU admission and during the ICU stay, cases received antibiotics more frequently than did controls. During the ICU stay tracheostomy was performed more frequently in cases than in controls. Other patient characteristics were similar between case and control patients (Table 1). The mean period between ICU admission and development of VAT was 11 ± 8 days (median 8 [range 3–47] days). The mean period between starting mechanical ventilation and development of VAT was 10 ± 9 days (median 7 [range 3–47] days).
A total of 86 micro-organisms were isolated in the 55 VAT episodes. The more frequently isolated bacteria were P aeruginosa (34%), A baumannii (18%) and methicillin-resistant S aureus (11%). Thirty (54%) VAT episodes were polymicrobial, and 31 (56%) were related to multidrug-resistant bacteria (Table 2).
Although the durations of mechanical ventilation and ICU stay were significantly longer in cases than in controls, no significant difference was found in mortality rate between case and control patients (Table 3). No significant difference in outcome was found between VAT patients who received adequate antibiotic treatment and those who received inadequate antibiotic treatment (Table 4). In cases with multidrug-resistant bacteria compared with cases with other bacteria, we observed similar durations of mechanical ventilation (23 ± 17 days versus 18 ± 13 days; P = 0.869), similar lengths of ICU stay (29 ± 14 versus 29 ± 18 days; P = 0.166) and similar ICU mortality rates (10/31 [32%] versus 6/24 [25%]; P = 0.359).
The results of univariate and multivariate analyses are presented in Table 5.
Discussion
The results of this study demonstrate that VAT is associated with increased duration of mechanical ventilation and ICU stay in immunocompetent nontrauma patients without chronic respiratory failure.
Tracheobronchitis is characterized by lower respiratory tract inflammation and increased sputum production. These factors may generate weaning difficulties, resulting in longer duration of mechanical ventilation. Extubation failure and difficult weaning have been reported to be associated with increased sputum volume in mechanically ventilated patients [10].
Previous studies [4,11] highlighted the link between tracheobronchitis and prolonged duration of mechanical ventilation, but these studies did not adjust for confounding factors; in particular, they did not adjust for duration of mechanical ventilation before development of VAT. Thus, based on those studies VAT could be considered a cause or a consequence of prolonged mechanical ventilation. The present case–control study, in which we adjusted for several confounding factors, is to our knowledge the first to demonstrate that VAT is independently associated with longer duration of mechanical ventilation in patients without chronic respiratory failure. However, an interventional randomized study is needed to confirm our findings.
In this study, duration of ICU stay was significantly longer in cases than in controls. However, mortality rates were similar between the two groups. In contrast, a recent prospective observational study [3], conducted in patients who had undergone heart surgery, found significantly higher mortality rates in patients with VAT than in noncolonized patients (20.7% versus 1.6%), and no significant difference in ICU and hospital lengths of stay between the two groups (12 days versus 5 days, and 20 days versus 13 days, respectively). However, the number of patients with VAT included in that study was small (n = 29). In addition, VAT patients who developed subsequent VAP were not excluded. Moreover, no adjustment was made for confounding factors.
VAT is probably an intermediate process between lower respiratory tract colonization and VAP. The diagnosis of VAT may be difficult in patients with chest radiographic abnormalities at ICU admission. However, recent guidelines recommend using new chest radiograph infiltrates as a criterion for diagnosis of VAP [12]. On the other hand, VAT is also difficult to differentiate from colonization. However, only new bacteria were taken into account in the present study. Moreover, we used quantitative tracheal aspirates to diagnose VAT, with a high threshold at 106 colony-forming units/ml.
The high proportion of multidrug-resistant bacteria in patients with VAT may be accounted for by the following factors: 87% of these patients were transferred from other wards; 72% of patients with VAT received antibiotics before ICU admission; and there was a long mean period between ICU admission and VAT development. These factors are well known to be associated with the emergence of multidrug-resistant bacteria in ICU patients [13].
Whether antibiotics should be administered to patients with VAT is actually a subject of debate. Clinical practice with respect to antibiotic treatment in patients with VAT varies widely between ICU physicians. Whereas some physicians do not treat this infection, considering it to be simple colonization, others routinely treat patients with VAT or only those patients with weaning difficulties and/or underlying disease [11,14,15]. In the present study only 21% of patients with VAT received antibiotics to treat this infection. No significant difference in outcome was found between patients who received adequate antimicrobial treatment and those with inadequate antimicrobial treatment. However, our findings are limited by the small number of VAT patients who received adequate antibiotic treatment. Antibiotic treatment could eradicate respiratory bacterial load and decrease sputum production. In a prospective study conducted in long-term mechanically ventilated patients with chronic bacterial colonization, Palmer and coworkers [11] observed a significant decrease in tracheal secretion volume, inflammatory cells and soluble intercellular adhesion molecule-1 burden in those patients who received antibiotics. Nevertheless, excessive antibiotic usage is associated with subsequent emergence of multidrug-resistant bacteria and causes measurable harm in ICU patients [16,17]. Therefore, further randomized studies are warranted to determine whether patients with VAT should be treated with antibiotics [18].
Recent guidelines on appropriate antibiotic use for treatment of acute respiratory tract infections in adults [19] indicate that antibiotic treatment of uncomplicated acute bronchitis should not be routinely applied. This recommendation is based on several randomized controlled studies [20-25] and recent meta-analyses [26-30]; all studies reported no impact of antibiotic treatment on illness duration, activity limitation, or work loss, and all concluded that routine antibiotic treatment of adults with acute bronchitis is not justified. However, all of those studies were conducted in healthy adults. To our knowledge, no randomized controlled study has been reported in mechanically ventilated patients with nosocomial tracheobronchitis.
Our study has several limitations. First, the study was a retrospective analysis of prospectively collected data. Second, our study was performed in a single ICU, and the results may not be applicable to patients in other ICUs. Third, some of the trends observed in the study might have reached statistical significance if the study sample had been larger. Forth, over the long period of study, some changes in case-mix, medical and nursing practices, workload and workforce might have occurred. However, VAT was independently associated with longer than median duration of mechanical ventilation in case and control patients during the study period. Finally, that patients with VAT who subsequently developed VAP were excluded probably overlooked an important consequence of VAT. However, VAP is associated with increased morbidity and mortality, and so exclusion of these patients allowed us to assess the true impact of VAT on outcome [31].
Conclusion
VAT is associated with increased duration of mechanical ventilation and ICU stay in immunocompetent nontrauma patients without chronic respiratory failure. Further studies are required to confirm our results and to evaluate the impact of antibiotic treatment on outcomes of patients with VAT.
Key messages
• VAT is associated with increased duration of mechanical ventilation and ICU stay in immunocompetent nontrauma patients without chronic respiratory failure.
• There was no significant difference in outcome between VAT patients who received adequate antibiotic treatment and those who received inadequate antibiotic treatment.
• Further studies are needed to evaluate the impact of antibiotic treatment on outcomes in patients with VAT.
Abbreviations
ICU = intensive care unit; VAP = ventilator-associated pneumonia; VAT = ventilator-associated tracheobronchitis.
Competing interests
The author(s) declare that they have no competing interests.
Acknowledgements
The results of this study were presented in part at the 100th ATS International Conference (2004; Orlando, FL, USA).
Figures and Tables
Figure 1 Study profile. VAT, ventilator-associated tracheobronchitis.
Table 1 Patient characteristics
Parameter/characteristic Cases (n = 55) Controls (n = 55)
At admission
Age (years; mean ± SD] 59.9 ± 18.2 60.3 ± 17.5
Male (n [%]) 31 (56) 33 (60)
SAPS II (mean ± SD) 37.2 ± 15.3 37.4 ± 14.7
Transfer from other wards (n [%]) 48 (87) 49 (89)
Diabetes mellitus (n [%]) 12 (21) 13 (23)
Prior antibiotic treatment* (n [%]) 40 (72) 15 (27)
Admission category (n [%])
Medical 39 (70) 39 (70)
Surgical 16 (29) 16 (29)
Primary diagnosis for admission (n [%])
Community-acquired pneumonia 12 (21) 12 (21)
Cellulitis 12 (21) 12 (21)
Septic shock 10 (18) 10 (18)
Congestive heart failure 7 (12) 7 (12)
Peritonitis 4 (7) 4 (7)
Acute respiratory distress syndrome 4 (7) 4 (7)
Other 6 (10) 6 (10)
Organ failure (n [%])
Cardiac 11 (20) 17 (30)
Respiratory 38 (69) 41 (74)
Renal 14 (25) 10 (18)
Neurologic 12 (21) 19 (34)
Digestive 3 (5) 5 (9)
During hospitalization
Tracheostomy† (n [%]) 12 (21) 5 (9)
Antibiotic treatment‡
n (%) 55 (100) 43 (78)
Duration (days; mean ± SD) 13.3 ± 11.5 5.8 ± 9.2
Antibiotic days/1000 ICU-days 485 330
*P = 0.006, †P = 0.056 and ‡P < 0.001 (cases/controls) by univariate analysis. ICU, intensive care unit; SAPS, simplified acute physiology score; SD, standard deviation.
Table 2 Bacteria associated with 55 episodes of ventilator-associated tracheobronchitis
Micro-organism n (%)
Gram-negative 66 (76)
Pseudomonas aeruginosa 30 (34)
Acinetobacter baumannii 16 (18)
Serratia spp. 6 (6)
Enterobacter spp. 4 (4)
Escherichia coli 4 (4)
Stenotrophomonas maltophilia 3 (3)
Klebsiella spp. 3 (3)
Gram-positive 20 (23)
MRSA 10 (11)
MSSA 5 (5)
Streptococcus pneumoniae 5 (5)
MRSA, methicillin-resistant Staphylococcus aureus; MSSA, methicillin-sensitive Staphylococcus aureus.
Table 3 Outcomes of patients with (cases) and patients without (controls) ventilator-associated tracheobronchitis
Outcome Cases (n = 55) Controls (n = 55) Pa
Duration of mechanical ventilation (days)
Median (range) 17.0 (3–95) 8.0 (3–61) <0.001
Mean ± SD 21.6 ± 16.0 13.3 ± 13.1
Length of ICU stay (days)
Median (range) 24.5 (5–95) 12.0 (4–74) <0.001
Mean ± SD 28.0 ± 15.7 17.6 ± 16.6
ICU mortality (n [%]) 16 (29) 20 (36) 0.294
aResults by univariate analysis. ICU, intensive care unit; SD, standard deviation.
Table 4 Impact of antibiotic treatment on outcomes of patients with ventilator-associated tracheobronchitis
Outcome Antibiotic treatment Pa
Adequate (n = 12) Inadequate (n = 43)
Duration of mechanical ventilation (days)
Median (range) 17.0 (3–95) 18.5 (3–58) 0.833
Mean ± SD 18.8 ± 9.7 22.3 ± 17.2
Length of ICU stay (days)
Median (range) 22.0 (5–95) 25.0 (6–62) 0.344
Mean ± SD 24.8 ± 14.5 30.5 ± 16.8
ICU mortality (n [%]) 5 (41) 11 (25) 0.178
Inadequate antibiotic treatment was given for infectious diseases other than ventilator-associated tracheobronchitis. aResults by univariate analysis. ICU, intensive care unit; SD, standard deviation.
Table 5 Factors associated with duration of mechanical ventilation longer than median (14 days) in patients with (cases) and without (controls) ventilator-associated tracheobronchitis
Factor Univariate analysis Multivariate analysis
Number of patients (n = 110) Number of patients with MV duration ≥ 14 days (%) P OR (95% CI) P
Renal failure on ICU admission
Yes 24 18 (75) 0.002 4.9 (1.6–14.6) 0.004
No 86 34 (39)
Tracheostomy
Yes 17 13 (76) 0.009 4.0 (1.1–14.5) 0.032
No 93 39 (41)
VAT related to multidrug-resistant bacteria
Yes 31 22 (70) 0.002 - -
No 79 30 (37)
VAT
Yes 55 35 (63) 0.001 3.5 (1.5–8.3) 0.004
No 55 17 (30)
CI, confidence interval; ICU, intensive care unit; MV, mechanical ventilation; OR, odds ratio; VAT, ventilator-associated tracheobronchitis.
==== Refs
Vincent JL Bihari DJ Suter PM Bruining HA White J Nicolas-Chanoin MH Wolff M Spencer RC Hemmer M The prevalence of nosocomial infection in intensive care units in Europe. Results of the European Prevalence of Infection in Intensive Care (EPIC) Study. EPIC International Advisory Committee JAMA 1995 274 639 644 7637145 10.1001/jama.274.8.639
Rello J Ausina V Castella J Net A Prats G Nosocomial respiratory tract infections in multiple trauma patients. Influence of level of consciousness with implications for therapy Chest 1992 102 525 529 1643942
Bouza E Perez A Munoz P Jesus Perez M Rincon C Sanchez C Martin-Rabadan P Riesgo M Ventilator-associated pneumonia after heart surgery: a prospective analysis and the value of surveillance Crit Care Med 2003 31 1964 1970 12847390 10.1097/01.ccm.0000084807.15352.93
Nseir S Di Pompeo C Pronnier P Beague S Onimus T Saulnier F Grandbastien B Mathieu D Delvallez-Roussel M Durocher A Nosocomial tracheobronchitis in mechanically ventilated patients: incidence, aetiology and outcome Eur Respir J 2002 20 1483 1489 12503708 10.1183/09031936.02.00012902
Marquette CH Georges H Wallet F Ramon P Saulnier F Neviere R Mathieu D Rime A Tonnel AB Diagnostic efficiency of endotracheal aspirates with quantitative bacterial cultures in intubated patients with suspected pneumonia. Comparison with the protected specimen brush Am Rev Respir Dis 1993 148 138 144 8317789
Garner JS Jarvis WR Emori TG Horan TC Hughes JM CDC definitions for nosocomial infections, 1988 Am J Infect Control 1988 16 128 140 2841893
Anonymous Standards for the diagnosis and care of patients with chronic obstructive pulmonary disease (COPD) and asthma. This official statement of the American Thoracic Society was adopted by the ATS Board of Directors, November 1986 Am Rev Respir Dis 1987 136 225 244 3605835
Le Gall JR Lemeshow S Saulnier F A new Simplified Acute Physiology Score (SAPS II) based on a European/North American multicenter study JAMA 1993 270 2957 2963 8254858 10.1001/jama.270.24.2957
Knaus WA Draper EA Wagner DP Zimmerman JE Prognosis in acute organ-system failure Ann Surg 1985 202 685 693 4073980
Epstein SK Decision to extubate Intensive Care Med 2002 28 535 546 12029399 10.1007/s00134-002-1268-8
Palmer LB Smaldone GC Simon S O'Riordan T Morra L Tracheal aspirates in long-term mechanically ventilated patients. A human model of gram-negative infection and airway inflammation Chest 1995 108 1326 1332 7587436
Cohen J Brun-Buisson C Torres A Jorgensen J Diagnosis of infection in sepsis: an evidence-based review Crit Care Med 2004 32 S466 S494 15542957 10.1097/01.CCM.0000145917.89975.F5
Safdar N Maki DG The commonality of risk factors for nosocomial colonization and infection with antimicrobial-resistant Staphylococcus aureus, enterococcus, gram-negative bacilli, Clostridium difficile, and Candida Ann Intern Med 2002 136 834 844 12044132
Ahmed QA Niederman MS Respiratory infection in the chronically critically ill patient. Ventilator-associated pneumonia and tracheobronchitis Clin Chest Med 2001 22 71 85 11315460
Hamer DH Treatment of nosocomial pneumonia and tracheobronchitis caused by multidrug-resistant Pseudomonas aeruginosa with aerosolized colistin Am J Respir Crit Care Med 2000 162 328 330 10903263
Kollef MH Fraser VJ Antibiotic resistance in the intensive care unit Ann Intern Med 2001 134 298 314 11182841
Nseir S Di Pompeo C Soubrier S Delour P Lenci H Roussel-Delvallez M Onimus T Saulnier F Mathieu D Durocher A First-generation fluoroquinolone use and subsequent emergence of multiple drug-resistant bacteria in the intensive care unit Crit Care Med 2005 33 283 289 15699829 10.1097/01.CCM.0000152230.53473.A1
Anonymous Guidelines for the management of adults with hospital-acquired, ventilator-associated, and healthcare-associated pneumonia Am J Respir Crit Care Med 2005 171 388 416 15699079 10.1164/rccm.200405-644ST
Snow V Mottur-Pilson C Gonzales R Principles of appropriate antibiotic use for treatment of acute bronchitis in adults Ann Intern Med 2001 134 518 520 11255531
King DE Williams WC Bishop L Shechter A Effectiveness of erythromycin in the treatment of acute bronchitis J Fam Pract 1996 42 601 605 8656171
Verheij TJ Hermans J Mulder JD Effects of doxycycline in patients with acute cough and purulent sputum: a double blind placebo controlled trial Br J Gen Pract 1994 44 400 404 8790652
Williamson HA Jr A randomized, controlled trial of doxycycline in the treatment of acute bronchitis J Fam Pract 1984 19 481 486 6384419
Brickfield FX Carter WH Johnson RE Erythromycin in the treatment of acute bronchitis in a community practice J Fam Pract 1986 23 119 122 3525736
Franks P Gleiner JA The treatment of acute bronchitis with trimethoprim and sulfamethoxazole J Fam Pract 1984 19 185 190 6611385
Dunlay J Reinhardt R Roi LD A placebo-controlled, double-blind trial of erythromycin in adults with acute bronchitis J Fam Pract 1987 25 137 141 3302093
Orr PH Scherer K Macdonald A Moffatt ME Randomized placebo-controlled trials of antibiotics for acute bronchitis: a critical review of the literature J Fam Pract 1993 36 507 512 8482934
MacKay DN Treatment of acute bronchitis in adults without underlying lung disease J Gen Intern Med 1996 11 557 562 8905509
Fahey T Stocks N Thomas T Quantitative systematic review of randomised controlled trials comparing antibiotic with placebo for acute cough in adults BMJ 1998 316 906 910 9552842
Smucny JJ Becker LA Glazier RH McIsaac W Are antibiotics effective treatment for acute bronchitis? A meta-analysis J Fam Pract 1998 47 453 460 9866671
Bent S Saint S Vittinghoff E Grady D Antibiotics in acute bronchitis: a meta-analysis Am J Med 1999 107 62 67 10403354 10.1016/S0002-9343(99)00167-9
Ioanas M Ewig S Torres A. Treatment failures in patients with ventilator-associated pneumonia Infect Dis Clin North Am 2003 17 753 771 15008597 10.1016/S0891-5520(03)00070-9
| 15987396 | PMC1175884 | CC BY | 2021-01-04 16:04:53 | no | Crit Care. 2005 Mar 31; 9(3):R238-R245 | utf-8 | Crit Care | 2,005 | 10.1186/cc3508 | oa_comm |
==== Front
Crit CareCritical Care1364-85351466-609XBioMed Central London cc35111598739510.1186/cc3511ResearchDoes fluid loading influence measurements of intestinal permeability? Parviainen Ilkka [email protected] Jukka [email protected] Stephan M [email protected] Consultant, Department of Anesthesiology and Intensive Care, Kuopio University Hospital, Kuopio, Finland2 Professor, Department of Intensive Care Medicine, University Hospital Bern, Bern, Switzerland3 Consultant, Department of Intensive Care Medicine, University Hospital Bern, Bern, Switzerland2005 21 3 2005 9 3 R234 R237 9 2 2005 24 2 2005 25 2 2005 3 3 2004 Copyright © 2005 Parviainen 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
Urinary recovery of enterally administered probes is used as a clinical test of intestinal mucosal permeability. Recently, evidence has been provided that the recovery of some but not all sugar probes is dependent on the amount of diuresis and renal function. The aim of this study was to assess the effect of fluid loading on the urinary recovery of sugar probes in healthy volunteers.
Methods
In a cross-over study, 10 healthy volunteers ingested 100 ml of a solution containing 0.2 g of 3-O-methyl-D-glucose (3-OMG), 0.5 g of D-xylose, 1.0 g of L-rhamnose, and 5.0 g of lactulose on two different days. The volunteers were randomized to receive either 2 litres of Ringer acetate or no fluid during the following 3 hours. The sugar concentrations were measured in 5-hour urine samples period.
Results
Fluid loading increased urine production and urinary recovery of xylose. Fluid loading did not influence the urinary recovery of 3-OMG, L-rhamnose, or lactulose. Neither the lactulose/rhamnose ratio nor the 3-OMG/rhamnose ratio changed.
Conclusion
Fluid loading increases mediated carbohydrate transport but not the lactulose/rhamnose ratio, after oral sugar administration in healthy volunteers. It remains to be determined whether sugar probes are handled differently in response to fluids in patients with organ dysfunctions.
==== Body
Introduction
Mucosal permeability is one of the few functions of the gastro-intestinal tract that can be quantified in the clinical setting. Increased intestinal permeability in critically ill patients has been associated with endotoxemia and remote organ failure [1]. Mucosal intestinal permeability can be assessed noninvasively by measuring urinary excretion of orally administered test substances. Typically, this involves estimation of the urinary recovery of single or multiple probes administered orally. Quantifying the absorption of two sugars of different sizes offers advantages compared with the use of single probes: urinary recovery of these probes expressed as a ratio is particularly sensitive because this can reflect the contrasting effects of decreased absorption of monosaccharides, such as rhamnose or mannitol, owing to the reduced surface area and increased permeability for larger disaccharides, such as lactulose or cellobiose, owing to the opening of intracellular pathways. A further advantage of this ratio is the elimination of errors due to non-mucosal factors, because variables such as rate of gastric emptying, intestinal transit, impairment of renal function, and completeness of urinary collection should affect both sugars similarly [2].
In experimental animals, fluid loading increased the urinary recovery of intravenously administered lactulose but not rhamnose, and thereby increased the lactulose/rhamnose (L/R) ratio [3], suggesting changes in tissue distribution or altered renal handling of the sugar probes. In critically ill patients with organ dysfunction, recoveries of cellobiose, sucrose, and mannitol were positively related to urinary volume [4]. If fluid loading and the related increase in diuresis effectively alters the recovery of some or all sugar probes in humans, this method might no longer be considered a valid test for the assessment of intestinal permeability. The aim of this study was to test whether intravenous fluid loading increases the urinary recovery of lactulose or other orally administered sugars in healthy volunteers.
Methods
Ten healthy volunteers (four female, six male) with a median age of 32 years (range 23 to 42 years) were studied on the morning after an overnight fast. The subjects were not allowed to consume alcoholic beverages or to use drugs the day before the testing days. The study was approved by the Ethics Committee of Kuopio University Hospital. The research was performed in accordance with the Declaration of Helsinki. All volunteers gave written informed consent to participate in the study. In a cross-over design, the volunteers ingested 100 ml of a solution containing 0.2 g of 3-O-methyl-D-glucose (3-OMG), 0.5 g of D-xylose, 1.0 g of L-rhamnose, and 5.0 g of lactulose on two different days (Fig. 1). The volunteers were randomized to receive either 2 litres of Ringer acetate or no fluid during the following 3 hours. Before the ingestion of the sugar probes, the volunteers voided their urinary bladders. The sugar concentrations were measured at the end of the 5-hour urine sample period. The experiment was repeated after 4 to 6 days. Urinary recovery of the sugar probes assesses active (OMG) and passive (D-xylose) carrier-mediated, and nonmediated (L-rhamnose) transcellular absorption [2,5]. Lactulose permeates enterocytes through intercellular tight junctions [6]. The L/R ratio was calculated as an indicator of gut permeability. Urine samples were stored at -20°C until analysed by high-pressure liquid chromatography with pulsed amperometric detection [7]. The urinary recovery of each sugar probe was expressed as a percentage of the orally administered dose.
The data were analysed with Wilcoxon signed-rank test. Data are presented as medians and ranges.
Results
Fluid loading increased urine production during the 5-hour collection period (Table 1, Fig. 2). Fluid loading did not influence the urinary recovery of 3-OMG, L-rhamnose, or lactulose. Neither the L/R ratio (Fig. 2) nor the 3-OMG/rhamnose ratio changed during fluid loading. The urinary recovery of xylose increased during fluid loading (Fig. 2).
Discussion
Altered intestinal permeability has frequently been reported in patients requiring intensive care after major surgery and in sepsis. The mechanisms of these alterations are not entirely defined. In addition to permeability of intestinal epithelium, both pre-mucosal and postmucosal factors can affect the urinary excretion of enterally administered sugar probes [3,8]. Fluid loading has been reported to modify the excretion of sugar probes in rats [3]. Fluid loading is a common intervention in patients in intensive care. Altered excretion of sugar probes by fluid loading would invalidate the measurement of intestinal permeability in patients requiring rapid intravenous infusions of fluids.
We studied the effect of a postmucosal factor, fluid loading, on the urinary excretion of enterally administered sugar probes in healthy volunteers. In our study, only the urinary recovery of xylose increased after fluid loading. These results suggest that fluid loading does not affect measures of intestinal permeability, assessed by the L/R ratio, but increases mediated carbohydrate transport in healthy volunteers, measured by xylose recovery.
In contrast with our findings, fluid loading has been shown to increase the L/R ratio, caused by increased urinary lactulose excretion, in both control and endotoxemic rats [3]. Increased urinary lactulose recovery was observed both after intragastric and intravenous administration of the sugar probes. Together with the increased urine flow this suggests changes in the renal handling of lactulose during fluid loading [3]. In our study, despite increased urine flow during fluid administration, lactulose recovery did not increase. Proposed renal mechanisms for increased urinary lactulose recovery are a decrease in net proximal tubular Na+ reabsorption and decreased pericapillary osmotic pressure with saline infusion, which might enhance the net movement of lactulose from the peritubular capillary into the lumen [9,10]. Systemic and renal handling of lactulose might be different in rats and humans. Nevertheless, we found a similar recovery rate of lactulose, rhamnose, and 3-OMG within a comparable urine collection period (5 hours versus 6 hours). In our study, the L/R ratio is also comparable with the ratios from other studies in healthy subjects [6,11,12].
In rats, Hallemeesch and colleagues [3] used double the amount of the daily fluid intake for fluid loading within 8 hours. This compares well with the 2 litres we infused during 3 hours in our volunteers. It is unlikely that the method of fluid administration (twice subcutaneously in the rats versus continually intravenously in human volunteers) influenced the results. Saline was used in rats, whereas we used Ringer acetate for fluid loading. The amounts of sodium and chloride are higher in saline, which might have influenced the different results from the two studies.
It has been shown in volunteers that renal clearance of lactulose but not of rhamnose is dependent on the intravenously administered quantity of the respective sugar [12]: urinary lactulose recovery was significantly lower after intravenous administration of a high dose in comparison with the regular dose. Although lactulose cannot enter the cells and therefore has an extracellular distribution, the monosaccharides might be present both intracellularly and extracellularly. Because the infusion of saline and Ringer solution increases the extracellular space, it is conceivable that crystalloid infusion might affect blood lactulose concentrations more than blood concentrations of the other sugars. Our results suggest that clinically relevant amounts of crystalloid infusion do not induce major alterations in concentrations of lactulose in the plasma.
Xylose and 3-OMG are used to measure the mediated transport of carbohydrates and to assess the intestinal absorptive capacity. In animals, D-xylose is transported across membranes by a Na-dependent transport [13]. However, the precise mechanism for D-xylose transport across human epithelial cells might differ from that described in animals [14]. Two possible routes are suggested: Na-dependent carrier-mediated entry and a paracellular shunt pathway [14]. In our study, the recovery of xylose increased during fluid loading. There was also a minor, but nonsignificant, increase in urinary recovery of 3-OMG. It is possible that fluid loading increases absorptive capacity by increasing cardiac output and intestinal mucosal perfusion.
In critically ill patients, sugar absorption tests as a measure of gastrointestinal permeability have been challenged. In patients with multiple organ failure, urinary flow and creatinine clearance were the most important determinants of urinary sugar recovery [4]. It was observed that renal dysfunction limits the excretion of cellobiose and sucrose but not that of mannitol. In addition, the recovery of disaccharides and mannitol decreased when urinary flow was lower. The authors concluded that a differential sugar absorption test is not a reliable measure of gastrointestinal permeability in patients with multiple organ failure. Because the relation between urinary recovery of the sugar probes and renal function and diuresis was analysed only post hoc, it cannot be excluded that renal dysfunction coexists with maintained gastrointestinal permeability in critically ill patients.
Conclusion
Our results suggest that fluid loading despite increasing mediated carbohydrate transport does not affect the L/R ratio after oral sugar administration in healthy volunteers. However, further studies on intestinal and renal sugar handling in the assessment of gastrointestinal permeability are needed in critically ill patients.
Key messages
• Fluid loading increases mediated intestinal carbohydrate transport in healthy volunteers.
• Fluid loading does not affect intestinal permeability as assessed by the lactulose/rhamnose ratio in healthy volunteers.
Abbreviations
3-OMG = 3-O-methyl-D-glucose; L/R ratio = lactulose/rhamnose ratio.
Competing interests
The author(s) declare that they have no competing interests.
Authors' contributions
IP analysed data and wrote the final manuscript. JT designed the study and participated in the drafting of the manuscript. SJ recruited subjects into the study, managed them and participated in the drafting of the manuscript. All authors read and approved the final manuscript.
Figures and Tables
Figure 1 Experimental procedure for oral administration of sugar probes and fluid loading. 1, oral sugar administration; 2, 3 hours of fluid loading; 3, 5 hours of urine collection; 4, washout period from 4 to 6 days.
Figure 2 Urine output, L/R ratio and urinary recovery of xylose in individual volunteers.
Table 1 Urine volumes, urinary recoveries of sugar probes, L/R ratio and 3-OMG/R ratio.
Parameter Without fluid load With fluid load Wilcoxon SR test
Urine volume, ml 300 (10–500) 800 (400–1600) 0.005
3-OMG, % 44 (29–63) 51 (40–77) 0.074
Rhamnose, % 7 (4–14) 11 (3–22) 0.169
Xylose, % 25 (16–42) 34 (16–50) 0.017
Lactulose, % 0.2 (0.1–0.3) 0.3 (0.0–0.6) 0.074
L/R ratio 0.03 (0.01–0.05) 0.03 (0.00–0.17) 0.878
3-OMG/R ratio 6.1 (3.3–9.6) 4.8 (3.0–14.5) 0.878
Values are median (range). 3-OMG, 3-O-methyl-D-glucose; 3-OMG/R, 3-O-methyl-D-glucose/rhamnose; L/R, lactulose/rhamnose; SR, signed-rank.
==== Refs
Ammori BJ Leeder PC King RF Barclay GR Martin IG Larvin M McMahon MJ Early increase in intestinal permeability in patients with severe acute pancreatitis: correlation with endotoxemia, organ failure, and mortality J Gastrointest Surg 1999 3 252 262 10481118 10.1016/S1091-255X(99)80067-5
Bjarnason I MacPherson A Hollander D Intestinal permeability: an overview Gastroenterology 1995 108 1566 1581 7729650
Hallemeesch MM Lamers WH Soeters PB Deutz NEP Increased lactulose/rhamnose ratio during fluid load is caused by increased urinary lactulose excretion Am J Physiol Gastrointest Liver Physiol 2000 278 G83 G88 10644565
Oudemans-van Straaten H van der Voort PHJ Hoek FJ Bosman RJ van der Spoel JI Zanstra DF Pitfalls in gastrointestinal permeability measurement in ICU patients with multiple organ failure using differential sugar absorption Intensive Care Med 2002 28 130 138 11907655 10.1007/s00134-001-1140-2
Menzies IS Skadhauge E, Heintze K Transmucosal passage of inert molecules in health and disease Intestinal Absorption and Secretion (Falk Symposium 36) 1984 Lancaster: MTP 527 543
Menzies IS Laker MF Pounder R Bull J Heyer S Wheeler PG Creamer P Abnormal intestinal permeability to sugars in villus atrophy Lancet 1979 ii 1107 1109 10.1016/S0140-6736(79)92507-8
Sörensen SH Proud FJ Adam A Rutgers HC Batt RM A novel HPLC method for the simultaneous quantification of monosaccharides and disaccharides used in test of intestinal function and permeability Clin Chim Acta 1993 221 115 125 8149629 10.1016/0009-8981(93)90026-Z
Fink M Clinical tests of gastrointestinal permeability that rely on the urinary recovery of enterally administered probes can yield invalid results in critically ill patients Intensive Care Med 2002 28 103 104 11907651 10.1007/s00134-001-1191-4
Andersen LJ Jensen TU Bestle MH Bie P Isotonic and hypertonic sodium loading in supine humans Acta Physiol Scand 1999 166 23 30 10372975 10.1046/j.1365-201x.1999.00528.x
Rose BD Clinical Physiology of Acid-Base and Electrolyte Disorders 1977 Tokyo: McGraw-Hill Kogakusha
Noone C Menzies IS Banatvala JE Scopes JW Intestinal permeability and lactose hydrolysis in human rotaviral gastroenteritis assessed simultaneously by non-invasive differential sugar permeation Eur J Clin Invest 1986 16 217 225 3089818
Van Nieuwenhoven MA de Swart EAM van Eijk HM Deutz NEP Brouns F Brummers R-JM Effects of pre- and post-absorptive factors on the lactulose/rhamnose gut permeability test Clin Sci 2000 98 349 353 10677394 10.1042/CS19990274
Alvarado F D-xylose active transport in the hamster small intestine Biochim Biophys Acta 1966 112 292 5942959
Heyman M Desjeux J-F Grasset E Dumontier A-M Lestradet H Relationship between transport of D-xylose and other monosaccharides in jejunal mucosa of children Gastroenterology 1981 80 758 762 7202947
| 15987395 | PMC1175885 | CC BY | 2021-01-04 16:04:52 | no | Crit Care. 2005 Mar 21; 9(3):R234-R237 | utf-8 | Crit Care | 2,005 | 10.1186/cc3511 | oa_comm |
==== Front
Crit CareCritical Care1364-85351466-609XBioMed Central London cc35121598739710.1186/cc3512ResearchUse of intranasal mupirocin to prevent methicillin-resistant Staphylococcus aureus infection in intensive care units Muller Arno [email protected] Daniel [email protected] Alexandre 3Belle Evelyne 4Cappelier Gilles 5Bertrand Xavier [email protected] Student, Service d'Hygiène hospitalière et d'Epidémiologie moléculaire, Centre Hospitalier Universitaire Jean Minjoz, Besançon, France2 Head of Department, Service d'Hygiène hospitalière et d'Epidémiologie moléculaire, Centre Hospitalier Universitaire Jean Minjoz, Besançon, France3 House Officer, Service de Réanimation médicale Centre Hospitalier Universitaire Jean Minjoz, Besançon, France4 Clinician, Service de Réanimation médicale Centre Hospitalier Universitaire Jean Minjoz, Besançon, France5 Head of Department, Service de Réanimation médicale Centre Hospitalier Universitaire Jean Minjoz, Besançon, France6 Clincian, Service d'Hygiène hospitalière et d'Epidémiologie moléculaire, Centre Hospitalier Universitaire Jean Minjoz, Besançon, France2005 31 3 2005 9 3 R246 R250 13 1 2005 10 2 2005 22 2 2005 3 3 2005 Copyright © 2005 Muller 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
Methicillin-resistant Staphylococcus aureus (MRSA) causes severe morbidity and mortality in intensive care units (ICUs) worldwide. The purpose of this study was to determine whether intranasal mupirocin prophylaxis is useful to prevent ICU-acquired infections with MRSA.
Materials and methods
We conducted a 4-year observational retrospective study in a 15-bed adult medical ICU. During the first 2-year period mupirocin ointment was included in the MRSA control programme; during the second, mupirocin was not used. The main endpoint was the number of endogenous ICU-acquired infections with MRSA.
Results
The number of endogenous acquired infections was significantly higher during the second period than during the first (11 versus 1; P = 0.02), although there was no significant difference in the total number of patients infected with MRSA between the two periods. We also observed that nasal MRSA decolonisation was significantly higher in the mupirocin period than in mupirocin-free period (P = 0.002).
Conclusion
Our findings suggest that intranasal mupirocin can prevent endogenous acquired MRSA infection in an ICU. Further double-blind, randomised, placebo-controlled studies are needed to demonstrate its cost-effectiveness and its impact on resistance.
See related commentary
==== Body
Introduction
Over the past four decades, methicillin-resistant Staphylococcus aureus (MRSA) has spread throughout the world and become highly endemic in many geographic areas. This pathogen causes severe morbidity and mortality in hospitals worldwide [1-3]. In France, 30 to 40% of S. aureus strains are methicillin-resistant and the median incidence of MRSA in clinical specimens is about 0.8 per 1000 patient-days in acute care facilities and 3.42 in intensive care units (ICUs) [4]. French intensive care experts had recommended that barrier precautions should be implemented for patients colonised or infected with MRSA, but there is no consensus about an MRSA screening programme and MRSA nasal decolonisation. Systematic MRSA screening on admission and preventive isolation have been shown to be cost-effective [5-7] but the usefulness of mupirocin-based nasal decolonisation remains a matter of debate. Indeed, mupirocin has emerged as the topical antibacterial agent of choice for the elimination of S. aureus nasal carriage. In hospitals in which MRSA is epidemic, the intra-nasal administration of mupirocin to both patients and personnel colonised with MRSA is considered to be appropriate [8]. Furthermore, mupirocin has been used successfully to prevent staphylococcal infections in surgical and haemodialysis patients [9]. Although the available literature does not support routine use of intranasal mupirocin to prevent subsequent infections, there may be a role in some selected cases, such as those involving critically ill patients [9]. Thus, it was reported that intranasal mupirocin could decrease the occurrence of S. aureus pneumonia in ICU patients [10]. In 1994, an MRSA control programme was implemented in all high-risk units of our hospital, including adult ICUs. Thus, all patients were screened for MRSA on admission and during hospitalisation; MRSA-positive patients were kept in isolation and positive patients were prescribed nasal mupirocin ointment. In June 2001, clinicians from the medical ICU decided to stop using mupirocin for MRSA nasal decolonisation. We conducted a retrospective study to determine whether intranasal mupirocin prophylaxis is useful in preventing ICU-acquired infections with MRSA.
Materials and methods
Setting and study period
The 15-bed medical ICU of Besançon hospital admits about 450 to 500 patients per year, giving a mean of 5,000 patient-days per year. All patients admitted between 1 June 1999 and 31 May 2003 were included in this retrospective study. Mean ICU length of stay, mean gravity scores (Simplified Acute Physiology Score, SAPSII) and fatality rates were calculated monthly [11]. Ethical approval for this study was granted by the ethical committee of the hospital.
MRSA control programme
The MRSA control strategy was based on screening nasal fluid samples from all patients for MRSA on admission and once a week during hospitalisation. When the screening test was positive for MRSA, patients were given nasal mupirocin for 5 days, even if other body sites were colonised with MRSA. Special precautions were taken to prevent cross-contamination, including the use of disposable gowns and gloves, the use of an alcohol rub for hand hygiene, and the implementation of strict environmental hygiene measures. The MRSA status of the patient was written on the door of their room and in their medical chart. The patient was removed from isolation when two consecutive screening tests were negative and no clinical samples tested positive for MRSA. The programme did not include restrictions on antibiotic use. This programme was applied in our medical ICU until June 2001. At that date, nasal mupirocin ointment was stopped.
Microbiological techniques
The screening programme involved the collection of nasal samples from each patient and of tracheal aspirates from ventilated patients. Clinical diagnostic samples were obtained as requested by the physician in charge of the patient. Screening samples were used to inoculate both Mueller–Hinton agar and Mueller–Hinton agar supplemented with 10 mg/l tobramycin because 20% of all MRSA isolated in our hospital are tobramycin-susceptible [12,13]. Plates were examined for staphylococci after 24 hours at 37°C. Identification of S. aureus was initially based on colony aspect and the detection of both clumping factor and protein A with the Pastorex Staph-Plus latex agglutination test (Bio-Rad, Marnes la Coquette, France). Complementary tests, such as the coagulase and DNase tests, were performed if necessary. S. aureus strains that grew both on plates with and on those without tobramycin were considered to be MRSA. Indeed, a continuous surveillance of antimicrobial resistance among clinical samples showed that more than 98% of methicillin-susceptible Staphylococcus aureus were susceptible to tobramycin. Those that grew only on tobramycin-free plates were tested for oxacillin resistance. This antibiotic susceptibility was determined with the disk diffusion technique by incubating with 5 μg oxacillin disks for 48 hours at 30°C, as recommended by the Antibiogram Committee of the French Society for Microbiology [14]. S. aureus strains with oxacillin inhibition zones of less than 20 mm (corresponding to a minimum inhibitory concentration of more than 2 mg/l) were classified as MRSA. About 75% of the positive MRSA screening results were obtained within 24 hours; the others (25%) were obtained within 48 hours.
MRSA definitions
The following definitions were used: case, a patient from whom MRSA was recovered from any site irrespective of the sample type (screening or clinical); carrier, a patient from whom MRSA was recovered from screening samples or from clinical samples without signs of infection; and infected patient, a patient from whom MRSA was recovered from clinical samples with signs of infection according to the definitions of the Centers for Disease Control and Prevention [15]. MRSA was considered to be imported if the patients tested positive within 48 hours of admission in the ICU and to be acquired MRSA if they tested negative for the first 48 hours after admission. Infections were considered to be endogenous if a screening sample was positive before the clinical sample, or exogenous otherwise. MRSA was considered to have been eradicated when two consecutive screening tests were negative.
Antibiotic use
The quantities of each antimicrobial delivered to the ICU were determined from the pharmacy information system. Grams and international units of antimicrobials were further converted into defined daily doses (DDD) in accordance with the recommendations of the World Health Organization [16]. The amount of each class of antibiotics used is expressed in DDD per 1,000 days of hospitalisation.
Statistical analysis
The aim of the study was to compare ICU-acquired infections with MRSA before (period 1, June 1999 to May 2001) and after (period 2, June 2001 to May 2003) the use of intranasal mupirocin stopped. Analysis was performed by using the χ2 test and Fisher's exact test for categorical variables and the Mann–Whitney test for continuous variables. P ≤ 0.05 was considered to be significant. Statistical analysis was performed with Epi-info 6.0 (Centers for Disease Control and Prevention, Atlanta, GA) and R (The R Project for Statistical Computing, ) software.
Results
During the first study period with mupirocin use, 912 patients were admitted to the medical ICU, 775 were screened for MRSA (84.5%), 38 were MRSA carriers and 9 were infected with MRSA. During the second period without mupirocin use, 987 patients were admitted, 819 were screened (83.0%), 49 were MRSA carriers and 17 were infected with MRSA (Table 1). The number of endogenous acquired infections was significantly higher during the second period than during the first (11 versus 1; P = 0.006, relative risk = 8.53 (CI95% [1.15–63.21])) although there was no significant difference in the total number of patients infected with MRSA between the two periods. The 12 endogenous acquired infections detected are described in Table 2. In the second period, the delay between the detection of MRSA carriage and the date of the first positive clinical sample was sufficient to implement nasal mupirocin ointment for 7 of the 11 infected patients. Three patients died during the 2 weeks after the MRSA bacteraemia or pneumonia.
We next evaluated the efficiency of nasal mupirocin for MRSA eradication. The rate of nasal MRSA decolonisation among the patients hospitalised for more than 14 days after the discovery of MRSA carriage (delay necessary to evaluate the efficacy of mupirocin) reached 86.4% in period 1 and just 46.4% in period 2 (P = 0.002). Three of the 19 patients who were successfully decolonised during the first period received concomitant treatment with antimicrobials effective against MRSA (namely vancomycin).
The total amount of antibiotics used remained stable during the study period (1,700 DDD per 1,000 hospitalisation days) and the use of glycopeptides did not significantly vary (104 DDD per 1,000 hospitalisation days in the first period and 91 in the second period).
Discussion
Our observational study suggests that nasal mupirocin can effectively prevent the occurrence of endogenous acquired MRSA infections in ICUs. The increase in endogenous ICU-acquired MRSA infections in the second period occurred although different indicators such as mean length of stay, mean gravity scores of fatality rates did not significantly vary between the periods (Table 1). In 7 of the 11 MRSA infections observed in the period 2, the delay between the two samples was sufficient to implement mupirocin treatment; that is, at least 5 days. Our findings stand in contrast to double-blind randomised, placebo-controlled trials that included patients hospitalised in different type of unit [17]. However, in this study, the number of infections was three in the mupirocin group and seven in the placebo group. It is likely that the enrolment of more than 98 patients would show a significant effect of mupirocin in preventing subsequent MRSA infection. Moreover, the only randomised study applied to ICU patients also concluded that mupirocin was effective in reducing the occurrence of S. aureus pneumonia [10].
Some limitations of our study have to be addressed. First, we compared historical groups, whereas randomised, double-blind placebo controlled trials are more powerful. However, MRSA infection remains a rare event in our medical ICU and a very large study period would be needed to show that a programme including mupirocin use is beneficial. We simply reported the evolution of the infections with MRSA between two periods during which only one element had changed: the use of mupirocin for nasal MRSA decolonisation.
Second, we have not evaluated mupirocin resistance in this study. Low-level mupirocin resistance has been identified as a risk factor for a failure to eradicate MRSA [18], and numerous authors have reported the emergence of mupirocin resistance in settings where mupirocin is commonly used [19]. Concerns over the development of resistance have dissuaded many hospitals from using mupirocin in this manner, and recent recommendations from the guidelines of the Society for Healthcare Epidemiology of America stated that 'widespread use, prolonged use or both of decolonization therapy should be avoided' [20]. Others studies have reported that relapse of MRSA carriage was not associated with the development of resistance to mupirocin [21,22].
The body of the literature currently does not support routine intranasal mupirocin prophylaxis to all inpatients to decrease the rate of clinical infection. However, there might be as yet unidentified patient populations that could benefit. One can speculate that a significant effect should be seen for patients at high risk of infection, such as patients admitted to ICUs [23]. Indeed, our data showed that the infectious risk in our ICUs is high; during period 2 (mupirocin-free), 25.6% of carriers became infected (Table 1). A recent meta-analysis reported a significant increase in mortality associated with MRSA infection (odds ratio 1.93) and most studies depicted a crude mortality rate between 20% and 40% [3].
Our results concerning MRSA nasal eradication are in agreement with those observed by several authors in different settings, Kalmeijer and colleagues reported that MRSA was eradicated from 83.5% of patients admitted for orthopaedic surgery [24], and Mody and colleagues also showed that treatment with mupirocin had decolonised 93% of residents in long-term care facilities [21]. However, they differed from those obtained in other randomised trials. This difference was probably due to the status of the patients included, regardless of the type of setting.
Conclusion
Our results suggest that the use of mupirocin to prevent MRSA infections in ICUs has to be evaluated. Further double-blind, randomised, placebo-controlled clinical trials are needed to demonstrate its cost-effectiveness and its impact on mupirocin resistance.
Key messages
• MRSA causes severe morbidity and mortality in ICUs worldwide.
• Mupirocin is regarded as the topical antibacterial agent of choice for the elimination of S. aureus nasal carriage.
• The body of the literature does not currently support routine intranasal mupirocin prophylaxis.
• Our observational study suggests that nasal mupirocin can effectively prevent the occurrence of endogenous acquired MRSA infections in ICUs.
• Further double-blind, randomised, placebo-controlled clinical trials are needed to confirm our findings in ICUs.
Abbreviations
DDD = defined daily doses; ICU = intensive care unit; MRSA = methicillin-resistant Staphylococcus aureus.
Competing interests
The author(s) declare that they have no competing interests.
Authors' contributions
AM collected the data and performed the statistical analysis. AP recorded the clinical information. XB and DT composed the writing committee. EB and GC supervised AP for the record of clinical information. All authors read and approved the final manuscript.
Figures and Tables
Table 1 Description of the medical intensive care unit of Besançon hospital, June 1999 to May 2003
Parameter Period 1 (June 1999 to May 2001) Period 2 (June 2001 to May 2003) P
No. of patients admitted 912 987
Patients screened for MRSA (% of patients admitted) 775 (84.5%) 819 (83.0%) 0.26
Mean length of stay, days 11.44 9.73 0.02
Mean gravity score, SAPS II 46.32 46.44 0.91
Rate of death, % 32.73 28.43 0.07
MRSA cases 46 55 0.59e
Carriers of MRSAa (then infected) 38 (1) 49 (11) 0.35g
Infected with MRSAb 9 17 0.24e
Imported MRSAc 6 4 0.66e
Exogenous acquired MRSAc,d 2 2 1.00f
Endogenous acquired MRSAc,d 1 11 0.006g
aPatients from whom methicillin-resistant Staphylococcus aureus (MRSA) was recovered from screening samples or from clinical samples without signs of infection. bPatients from whom MRSA was recovered from clinical samples with signs of infection in accordance with the definitions of the Centers for Disease Control and Prevention. cDefined as imported if they tested positive within 48 hours of admission, and as acquired if they tested negative for the first 48 hours after admission. dDefined as endogenous if a screening sample was positive before the clinical sample, and as exogenous otherwise. eIn comparison with admitted patients. fIn comparison with screened patients. gIn comparison with carriers of MRSA. SAPS II, Simplified Acute Physiology Score, SAPS II.
Table 2 Characteristics of endogenous invasive MRSA infections
Patient no. Type of infection Period Date of infection Delay between admission and infection (days) Delay between positive MRSA screening and infection (days) Antibiotic treatment (Y/N) Death (Y/N) Time between infection and death (days)
1 Pneumonia P1 4 Sep 2000 26 7 Y (vancomycin) N
2 Bacteraemia P2 14 Aug 2001 8 7 Y (vancomycin) Y 15
3 Pneumonia P2 29 Aug 2001 37 6 Y (vancomycin) N
4 Pneumonia P2 29 Aug 2001 13 2 N Y 0
5 Endocarditis P2 10 Oct 2001 6 6 Y (vancomycin) N
6 Bacteraemia P2 24 Dec 2001 18 13 Y (vancomycin) N
7 Bacteraemia P2 16 Feb 2002 8 5 Y (vancomycin) N
8 Pneumonia P2 1 Apr 2002 3 2 Y (vancomycin) N
9 Pneumonia P2 6 Jun 2002 13 13 Y (vancomycin) N
10 Bacteraemia P2 22 Jun 2002 26 26 Y (vancomycin) N
11 Pneumonia P2 6 Feb 2003 13 3 Y (linezolid) Y 7
12 Pneumonia P2 24 Mar 2003 7 5 Y (vancomycin) N
MRSA, methicillin-resistant Staphylococcus aureus; P1, June 1999 to May 2001; P2, June 2001 to May 2003.
==== Refs
Aubry-Damon H Legrand P Brun-Buisson C Astier A Soussy CJ Leclercq R Reemergence of gentamicin-susceptible strains of methicillin-resistant Staphylococcus aureus : roles of an infection control program and changes in aminoglycoside use Clin Infect Dis 1997 25 647 653 9314454
Bertrand X Thouverez M Talon D Antibiotic susceptibility and genotypic characterization of methicillin-resistant Staphylococcus aureus strains in eastern France J Hosp Infect 2000 46 280 287 11170759 10.1053/jhin.2000.0841
Cosgrove SE Sakoulas G Perencevich EN Schwaber MJ Karchmer AW Carmeli Y Comparison of mortality associated with methicillin-resistant and methicillin-susceptible Staphylococcus aureus bacteremia: a meta-analysis Clin Infect Dis 2003 36 53 59 12491202 10.1086/345476
Albertini MT Benoit C Berardi L Berrouane Y Boisivon A Cahen P Cattoen C Costa Y Darchis P Deliere E Surveillance of methicillin-resistant Staphylococcus aureus (MRSA) and Enterobacteriaceae producing extended-spectrum beta-lactamase (ESBLE) in Northern France: a five-year multicentre incidence study J Hosp Infect 2002 52 107 113 12398076 10.1053/jhin.2002.1286
Girou E Pujade G Legrand P Cizeau F Brun-Buisson C Selective screening of carriers for control of methicillin-resistant Staphylococcus aureus (MRSA) in high-risk hospital areas with a high level of endemic MRSA Clin Infect Dis 1998 27 543 550 9770155
Lucet JC Chevret S Durand-Zaleski I Chastang C Regnier B Prevalence and risk factors for carriage of methicillin-resistant Staphylococcus aureus at admission to the intensive care unit: results of a multicenter study Arch Intern Med 2003 163 181 188 12546608 10.1001/archinte.163.2.181
Chaix C Durand-Zaleski I Alberti C Brun-Buisson C Control of endemic methicillin-resistant Staphylococcus aureus : a cost–benefit analysis in an intensive care unit JAMA 1999 282 1745 1751 10568647 10.1001/jama.282.18.1745
Doebbeling BN Breneman DL Neu HC Aly R Yangco BG Holley HP JrMarsh RJ Pfaller MA McGowan JE JrScully BE Elimination of Staphylococcus aureus nasal carriage in health care workers: analysis of six clinical trials with calcium mupirocin ointment. The Mupirocin Collaborative Study Group Clin Infect Dis 1993 17 466 474 8218691
Laupland KB Conly JM Treatment of Staphylococcus aureus colonization and prophylaxis for infection with topical intranasal mupirocin: an evidence-based review Clin Infect Dis 2003 37 933 938 13130405 10.1086/377735
Nardi G Di Silvestre AD De Monte A Massarutti D Proietti A Grazia Troncon M Lesa L Zussino M Reduction in gram-positive pneumonia and antibiotic consumption following the use of a SDD protocol including nasal and oral mupirocin Eur J Emerg Med 2001 8 203 214 11587466 10.1097/00063110-200109000-00008
Le Gall JR Lemeshow S Saulnier F A new Simplified Acute Physiology Score (SAPSII) based on a European/North American mutlicenter study JAMA 1993 270 2957 2963 8254858 10.1001/jama.270.24.2957
Talon D Deliere E Bertrand X Characterization of methicillin-resistant Staphylococcus aureus strains susceptible to tobramycin Int J Antimicrob Agents 2002 20 174 179 12385695 10.1016/S0924-8579(02)00173-5
Thouverez M Muller A Hocquet D Talon D Bertrand X Relationship between molecular epidemiology and antibiotic susceptibility of methicillin-resistant Staphylococcus aureus (MRSA) in a French teaching hospital J Med Microbiol 2003 52 801 806 12909658 10.1099/jmm.0.05252-0
Soussy CJ Carret G Cavallo JD Chardon H Chidiac C Choutet P Courvalin P Dabernat H Drugeon H Dubreuil L Antibiogram Committee of the French Microbiology Society. Report 2000–2001 [in French] Pathol Biol (Paris) 2000 48 832 871 11141919
Garner JS Jarvis WR Emori TG Horan TC Hughes JM CDC definitions for nosocomial infections, 1988 Am J Infect Control 1988 16 128 140 2841893
Natsch S Hekster YA de Jong R Heerdink ER Herings RM van der Meer JW Application of the ATC/DDD methodology to monitor antibiotic drug use Eur J Clin Microbiol Infect Dis 1998 17 20 24 9512177 10.1007/s100960050160
Harbarth S Dharan S Liassine N Herrault P Auckenthaler R Pittet D Randomized, placebo-controlled, double-blind trial to evaluate the efficacy of mupirocin for eradicating carriage of methicillin-resistant Staphylococcus aureus Antimicrob Agents Chemother 1999 43 1412 1416 10348762
Harbarth S Liassine N Dharan S Herrault P Auckenthaler R Pittet D Risk factors for persistent carriage of methicillin-resistant Staphylococcus aureus Clin Infect Dis 2000 31 1380 1385 11096006 10.1086/317484
Schmitz FJ Lindenlauf E Hofmann B Fluit AC Verhoef J Heinz HP Jones ME The prevalence of low- and high-level mupirocin resistance in staphylococci from 19 European hospitals J Antimicrob Chemother 1998 42 489 495 9818748 10.1093/jac/42.4.489
LeDell K Muto CA Jarvis WR Farr BM SHEA guideline for preventing nosocomial transmission of multidrug-resistant strains of Staphylococcus aureus and Enterococcus Infect Control Hosp Epidemiol 2003 24 639 641 14510243
Mody L Kauffman CA McNeil SA Galecki AT Bradley SF Mupirocin-based decolonization of Staphylococcus aureus carriers in residents of 2 long-term care facilities: a randomized, double-blind, placebo-controlled trial Clin Infect Dis 2003 37 1467 1474 14614669 10.1086/379325
Tomic V Svetina Sorli P Trinkaus D Sorli J Widmer AF Trampuz A Comprehensive strategy to prevent nosocomial spread of methicillin-resistant Staphylococcus aureus in a highly endemic setting Arch Intern Med 2004 164 2038 2043 15477440 10.1001/archinte.164.18.2038
Campbell W Hendrix E Schwalbe R Fattom A Edelman R Head-injured patients who are nasal carriers of Staphylococcus aureus are at high risk for Staphylococcus aureus pneumonia Crit Care Med 1999 27 798 801 10321672 10.1097/00003246-199904000-00039
Kalmeijer MD Coertjens H van Nieuwland-Bollen PM Bogaers-Hofman D de Baere GA Stuurman A van Belkum A Kluytmans JA Surgical site infections in orthopedic surgery: the effect of mupirocin nasal ointment in a double-blind, randomized, placebo-controlled study Clin Infect Dis 2002 35 353 358 12145715 10.1086/341025
| 15987397 | PMC1175886 | CC BY | 2021-01-04 16:04:52 | no | Crit Care. 2005 Mar 31; 9(3):R246-R250 | utf-8 | Crit Care | 2,005 | 10.1186/cc3512 | oa_comm |
==== Front
Crit CareCritical Care1364-85351466-609XBioMed Central London cc35131598739810.1186/cc3513ResearchInfluence of support on intra-abdominal pressure, hepatic kinetics of indocyanine green and extravascular lung water during prone positioning in patients with ARDS: a randomized crossover study Michelet Pierre [email protected] Antoine [email protected] Marc [email protected] Jean-Marie [email protected] Jean-Pierre [email protected] Laurent [email protected] Service de Réanimation Chirurgicale, Hôpitaux Sud, Marseille, France2005 31 3 2005 9 3 R251 R257 19 11 2004 26 1 2005 21 2 2005 7 3 2005 Copyright © 2005 Michelet 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
Prone positioning (PP) on an air-cushioned mattress is associated with a limited increase in intra-abdominal pressure (IAP) and an absence of organ dysfunction. The respective influence of posture by itself and the type of mattress on these limited modifications during the PP procedure remains unclear. The aim of this study was to evaluate whether the type of support modifies IAP, extravascular lung water (EVLW) and the plasma disappearance rate of indocyanine green (PDRICG) during PP.
Methods
A prospective, randomized, crossover study of 20 patients with acute respiratory distress syndrome (ARDS) was conducted in a medical intensive care unit in a teaching hospital. Measurements were made at baseline and repeated after 1 and 6 hours of two randomized periods of 6 hours of PP with one of two support types: conventional foam mattress or air-cushioned mattress.
Results
After logarithmic transformation of the data, an analysis of variance (ANOVA) showed that IAP and PDRICG were significantly influenced by the type of support during PP with an increase in IAP (P < 0.05 by ANOVA) and a decrease in PDRICG on the foam mattress (P < 0.05 by ANOVA). Conversely, the measurements of EVLW did not show significant modification between the two supports whatever the posture. The ratio of the arterial oxygen tension to the fraction of inspired oxygen significantly increased in PP (P < 0.0001 by ANOVA) without any influence of the support.
Conclusion
In comparison with a conventional foam mattress, the use of an air-cushioned mattress limited the increase in IAP and prevented the decrease in PDRICG related to PP in patients with ARDS. Conversely, the type of support did not influence EVLW or oxygenation.
See related commentary
==== Body
Introduction
Prone positioning (PP) improves arterial oxygenation in 50 to 75% of patients presenting with acute respiratory distress syndrome (ARDS) [1]. Although this postural treatment is currently considered simple and safe [1], the restriction of abdominal movements during PP is associated with an increase in intra-abdominal pressure (IAP) [2-5] with potential adverse effects on hemodynamic status and splanchnic perfusion [6-8]. Several studies have evaluated the clinical implications of this side effect of PP [4,5,9]. They reported significant but limited increases in IAP without impairment of cardiopulmonary, renal or hepatosplanchnic functions during short periods of PP. However, all patients included in these studies were placed on air-cushioned beds, which might have reduced the restriction of abdominal movement during PP [4,5,9] compared with a conventional foam mattress. Indeed, the air-cushioned mattress reduced interface pressure to a greater extent than the foam mattress [10,11]. Moreover, the addition of pillows under the thorax and the pelvis did not produce a decrease in IAP during the PP procedure with a foam mattress [12]. Because no direct comparison has been made between different supports, the respective influences of the type of support and the posture itself on IAP and its related potential adverse effects remain unclear. Therefore, in the perspective of standardization, the question of the interest of a special device before the institution of PP should be clarified [1].
The aim of the present study was to investigate whether the evolution of IAP, liver function assessed by the plasma disappearance rate of indocyanine green (PDRICG) [5] and extra-vascular lung water is related to the type of support during PP. We therefore prospectively compared, in a population of medical-ARDS patients, the effects of an air-cushioned mattress and a conventional foam mattress during PP.
Methods
Patients
Twenty consecutive patients with ARDS were included and turned prone in the medical intensive care unit of Sainte Marguerite University Hospital in Marseille, France. Patients were prospectively included in this study after obtaining written informed consent from the next of kin. The study design was approved by the Comité Consultatif de Protection des Personnes dans la Recherche Biomédicale of Marseille. ARDS was defined in accordance with the recommendations of the American–European Consensus Conference [13]. Patients with unstable cardiovascular function, cerebral injury or unstable spinal fractures, patients subjected to major abdominal surgery and patients with a history of neuromuscular disease were excluded.
Sedation, catheters and ventilation
All patients were sedated and paralyzed by the continuous infusion of sufentanil, midazolam and cisatracurium throughout the study period and were ventilated with conventional volume-controlled mechanical ventilation (7200 series or 840, Mallinckrodt Puritan Bennett, Carlsbad, CA, USA).
Respiratory and hemodynamic status was stable for 12 hours before inclusion. When patient received vasoactive drugs, the rate of infusion was kept stable throughout the study. On inclusion into the study, the mean tidal volume was 6.9 ± 1.9 ml/kg, the mean respiratory rate was 20 ± 4 cycles/min (the respiratory rate was adjusted to maintain a constant minute ventilation throughout the study period) and the positive end-expiratory pressure was 11.3 ± 2.0 cmH2O. The selection of the appropriate PEEP level was performed by increasing PEEP in steps of 2 cmH2O. A blood gas analysis was performed after a 30 min period of stabilization of blood oxygen saturation. Finally, the lower level of PEEP giving the greater improvement of oxygenation was chosen. The levels of PEEP, the tidal volume and a fraction of inspired oxygen (FiO2) of 0.8 were maintained constant throughout the study period.
A pulmonary artery catheter (Baxter Healthcare Corporation, Irvine, CA, USA) was placed in all patients. It was inserted percutaneously through the right jugular or left subclavian vein and positioned with the distal port in the pulmonary artery and the proximal port in the right atrium. For measurements of the PDRICG and extravascular lung water (EVLW), a 5F catheter was placed into the left femoral artery, together with a thermistor-tipped fiber-optic catheter (Pulsiocath, 4F, FT, PV-2024-L; Pulsion Medical System, Munich, Germany) which was advanced into the descending aorta.
Support surfaces
The control surface was a three-piece molded foam mattress (APLOT®; Asklé, Nîmes, France). The specialist air mattress was a dynamic alternating cells design, with automatic adjustment for patient weight (ProNimbus®; Huntleigh Healthcare, Luton, UK).
Measurements
Hemodynamic parameters
Routine invasive hemodynamic monitoring included arterial and pulmonary artery thermodilution catheters. Systemic and pulmonary arterial pressures, pulmonary artery occlusion pressure and right atrial pressure were measured at end-expiration. The midaxillary line was taken as the zero reference point in the supine and prone positions. Cardiac index, venous admixture and pulmonary vascular resistance were calculated from conventional formulas.
Blood gas analysis
Systemic and pulmonary arterial blood samples were withdrawn simultaneously within 3 min after the measurement of cardiac output. Arterial pH, arterial oxygen tension (PaO2) and arterial CO2 tension were measured with a blood gas analyzer (278-blood gas system; Ciba Corning, Medfield, MA, USA). Arterial and mixed venous oxygen saturation (SaO2 and SvO2, respectively) were measured with a calibrated hemoximeter (270-CO-oxymeter; Ciba Corning).
Respiratory parameters
The respiratory parameters measured were compliance of the respiratory system, exhaled tidal volume, peak inspiratory pressure, mean inspiratory pressure, and respiratory rate. Quasi-static compliance of the total respiratory system was obtained by dividing the tidal volume by the difference between plateau pressure and the total PEEP according to the method described by Gattinoni and colleagues [14].
Thoracic volumes, EVLW and hepatosplanchnic perfusion measurements
The transpulmonary indicator dilution technique was used to determine the intrathoracic blood volume, the EVLW and the PDRICG [15]. PDRICG is derived from the half-life of indocyanine green and reflects the percentage of the initial plasma dye level eliminated by the liver [16]. A thermistor-tipped fiber-optic catheter placed in the descending aorta detected the dye and temperature dilution curves. The EVLW and the PDRICG were automatically calculated by a computer (Pulsion Cold Z-021; Pulsion Medical System) from the average of three measurements [16].
Measurement of IAP
IAP was measured with a transurethral bladder catheter [17]. Normal saline (100 ml) was infused through the urinary catheter into the bladder. The catheter was then clamped and the IAP was recorded by a pressure transducer as mean pressure at end-expiration. Zero was set at the level of the pubis in both positions.
Protocol
Baseline measurements were performed in the supine position after 1 hour of steady-state conventional mechanical ventilation. Then the following two periods of PP were randomized: 6 hours of PP on the moulded foam mattress, and 6 hours of PP on the air-cushioned mattress. A period of 18 h in the supine position separated the two periods in the prone position. Each patient was his or her own control. Measurements were achieved in the supine position, after 1 and 6 hours of PP.
Change in position was performed manually by three nurses and two staff members. In the prone position, the arms were laid parallel to the body. Care was taken to avoid eye damage and any non-physiological movements of the limbs during posture changes. Whatever type of support was used, no pillow was used to support regions such as the chest or pelvis.
Measurements were performed before, after 1 hour and after 6 hours of each period of PP.
Statistical analysis
Statistical calculation was performed with the Sigma Stat 3.0 package (SPSS Inc., Chicago, IL, USA). Distribution was checked. Data were expressed as mean ± SD if the distribution was normal and as medians and interquartile range if the distribution was not normal. Significant differences were analyzed by general factorial analysis of variance (ANOVA) with a prior logarithmic transformation when required (non-parametric distribution). For intra-group changes, Tukey's test for multiple comparisons was applied to compare the variations between supports and different times of the study. For serum transaminases and creatinine, a Mann–Whitney U-test was used to compare the values before and after the protocol. When a correlation was calculated, Pearson's coefficient of correlation was used. When distribution was not normal, Spearman's rank correlation was used.
P < 0.05 was considered significant.
Results
Characteristics of the population are listed in Table 1. On admission, the mean Simplified Acute Physiologic Score II (SAPS II) score was 52 ± 12 and the severity of ARDS was assessed by a lung injury score of more than 2.5 in all patients (3.1 ± 0.3). The mortality rate for the 20 patients (15 men, 5 women; age 53 ± 12) was 25%. Among the patients enrolled in the study, six patients received norepinephrine (noradrenaline) (dose 0.3 ± 0.2 μg/kg per min) and one patient received epinephrine (adrenaline) (0.4 μg/kg per min) on inclusion. No modification of the infusion rate of norepinephrine and epinephrine and no fluid expansion were undertaken during the study period.
Effects of prone position and support on IAP, hepatosplanchnic function and EVLW
A logarithmic transformation of the data led to normally distributed values of IAP and PDRICG. We therefore performed a two-way ANOVA that showed an increase in IAP after 1 and 6 hours of PP on a foam mattress in comparison with baseline values (P < 0.01 by Tukey's test; Fig. 1) whereas it remained unchanged when patients were positioned on a specialist air mattress. IAP was higher on a foam mattress after 6 hours of PP in comparison with a specialist air mattress (P < 0.05 by Tukey's test; Fig. 1). PDRICG decreased after 1 and 6 hours of PP on a foam mattress in comparison with baseline values (P < 0.05 by Tukey's test, Table 2 and Fig. 2) whereas it remained unchanged when patients were positioned on a specialist air mattress.
There was no correlation between changes in IAP and PDRICG. Furthermore, the analysis of hepatic and renal variables did not show any statistical variation after the PP procedure period (Table 3).
There was no modification in EVLW and intrathoracic blood volume related to posture or support changes (Table 2).
Effects of prone position and support on gas exchange (Table 2)
PP induced an increase in the PaO2/FiO2 ratio (P < 0.001 by ANOVA) regardless of the type of support. There was no correlation between evolution in PaO2 and changes in IAP. PP reduced the true pulmonary shunt (P < 0.05 by ANOVA) without any influence of the kind of support.
Effects of prone position and support on hemodynamic and respiratory parameters (Table 2)
ANOVA showed that CVP, mean pulmonary arterial pressure and pulmonary artery occlusion pressure increased significantly in PP (P < 0.001) without any influence of the kind of support. Although PP induced a significant reduction in the static compliance of the total respiratory system (P = 0.007 by ANOVA), these modifications were not influenced by the kind of support. Other hemodynamic or respiratory parameters were not affected by prone position or type of support.
Discussion
The results of this study indicate that the use of an air-cushioned mattress for the PP procedure limited the increase of IAP and prevented a decrease in PDRICG related to PP. Nevertheless, these modifications were not associated with differences between supports for EVLW, oxygenation or cardiovascular parameters.
Even if the routine use of PP did not improve the survival of patients with ARDS in a recent multicenter randomized trial [18], it could be considered useful in the more hypoxemic patients. The improvement in oxygenation produced by PP is often related to an increase in aerated lung tissue with a decrease in venous admixture and a decrease in thoraco-abdominal compliance [19]. On the basis of these considerations, one can presume that the more the thoraco-abdominal wall is restricted during PP, the greater the gain in arterial oxygenation that should be obtained [4]. Nevertheless, in different settings, turning patients prone has been reported to induce an increase in IAP with potential side effects on cardiopulmonary, renal and hepatosplanchnic function [7,8]. The impaired hepatosplanchnic perfusion could lead to the development of multiple system organ failure [7,20,21], increasing the mortality rate of patients with ARDS [22,23].
Several studies were therefore designed to investigate the effects of PP on IAP modifications and clinical consequences [4,5,9]. The cumulative results of these studies indicate that, despite a small increase in IAP, PP improves arterial oxygenation without affecting cardiopulmonary, renal or hepatosplanchnic function. Nevertheless, the constant use of air-cushioned beds during PP in these studies could have contributed to limiting the effects of PP on IAP [5]. Indeed, this specialist air mattress presents a dynamic alternating-cells design with automatic adjustment for patient weight and a redistribution in pressure from the heavy parts (abdomen) to other areas. This support has been extensively reported as providing the lowest interface pressure in comparison with other supports, including foam mattresses [10,11]. Conversely, the use of a foam rubber mattress has recently been associated with an increase in IAP during PP without further reduction when pillows were added under the thorax and the pelvis [12].
Similarly, in our study, the use of foam rubber mattresses during PP induced a significant increase in IAP with a concomitant reduction in PDRICG, whereas air-cushion mattresses did not do so. Although these findings seem to indicate a superiority of the air-cushion support in terms of abdomen release, the clinical consequences of such modifications should be interpreted with regard to their duration. Indeed, despite an increase in IAP reaching 15 mmHg after 6 hours of PP on a foam mattress, the differences in liver function evaluated by PDRICG were limited during the PP procedure and did not induce significant modifications in hepatic biological variables during the period after PP. Conversely, although previous results have shown that a higher level must be reached to induce clinical modifications [24,25], a recent multicenter epidemiological study has reported the detrimental influence of a prolonged increase in IAP even for such moderate levels as 12 mmHg [26].
Another consequence of an increase in IAP during PP could be the modification of EVLW. Indeed, a decrease in EVLW has been proposed as an alternative mechanism that might account for the benefit related to PP [27]. Furthermore, an increase in IAP has recently been reported to worsen pulmonary edema in a lung injury induced by oleic acid [28]. Despite a PP-related increase in IAP with the foam mattress, our results did not demonstrate a significant modification in EVLW. The basal levels and the limited differences in IAP between the two supports probably explained the lack of influence of support on pulmonary edema assessed by EVLW in our study. An alternative, although not mutually exclusive, explanation for the observed evolution in EVLW could result from the underestimation of edema by this technique, because several areas of the lung are underperfused during ARDS [29].
Because the first objective of PP is still the improvement of ARDS-related hypoxemia, the potential influence of the support on oxygenation evolution represents another subject of concern [19,30,31]. Our results confirm previous data on PP-related oxygenation improvement but do not report any difference between the supports. The fact that the type of support does not appreciably influence the effects of PP on gas exchange has been deduced from previously published studies. Although Pelosi and colleagues reported an improvement in PaO2 that was correlated with a decrease in thoraco-abdominal compliance [19], these results were interpreted as being principally related to the modification of the chest-wall component of the thoraco-abdominal compliance during PP, without significant intervention of the abdominal part. Furthermore, Colmenero-Ruiz and colleagues, in an experimental study, showed that abdomen release in PP does not induce significant modification in oxygenation [31].
However, it should be noted that all patients included in the present study presented a medical etiology for their ARDS with an IAP in the supine position that was expected to be lower than in a surgical patient population. Consequently, our results must be interpreted with regard to our selected population because the effects of different support during PP could be different in patients with increased IAP (especially in surgical patients). Indeed, Gattinoni and colleagues [32] reported an IAP ranging from 6 mmHg in patients with ARDS of pulmonary cause to 16 mmHg in patients presenting with ARDS of extrapulmonary cause. The basal level of IAP and the magnitude of changes during PP for surgical patients could induce more relevant results than in our study and imply the need for further investigation in this population.
Our results suggest that the limited modifications in cardiovascular, renal or hepatosplanchnic function observed during PP are probably not related to the type of support but most probably to the relative harmlessness of this postural technique because patients do not present abdominal hypertension before prone positioning.
Conclusion
Consequently, the use of an air-cushion mattress for PP seems to be unnecessary in a standardized protocol in medical patients. However, the use of an air-cushion mattress is still of particular interest in reducing the incidence of pressure ulcers when prolonged periods of PP are needed and in facilitating the PP procedure for tracheostomized patients. Nevertheless, because the duration of PP period was limited to 6 hours in our protocol, a specific comparison between the two supports regarding skin lesions was not made in the present study and will require further evaluation.
Key messages
• In a population of medical-ARDS patients, the evolution of IAP and liver function during prone positioning is partly related to the type of support.
• The use of an air-cushion mattress do not influence the oxygenation and EVLW evolution during prone positioning.
Abbreviations
ANOVA = analysis of variance; ARDS = acute respiratory distress syndrome; EVLW = extravascular lung water; FiO2 = fraction of inspired oxygen; IAP = intra-abdominal pressure; PaO2 = arterial oxygen tension; PDRICG = plasma disappearance rate of indocyanine green; PEEP = positive end-expiratory pressure; PP = prone positioning.
Competing interests
The author(s) declare that they have no competing interests.
Authors' contributions
PM, AR and LP were the principal investigators and led the conceptual design of the design and the manuscript preparation. MG made contributions to the acquisition of data and to the analysis and interpretation of data. JMS, JPA and LP performed a first manuscript revision and gave final approval of the version to be submitted. All authors read and approved the final manuscript.
Figures and Tables
Figure 1 Effects of both supports on intra-abdominal pressure (IAP). The different durations of study are reported along the X-axis: SP, PP 1 H and PP 6 H represent baseline in supine position, prone position at 1 hour and prone position at 6 hours, respectively. Pale gray boxes, foam mattress; dark gray boxes, air-cushion mattress. Each box plot represents the median, the 25th and 75th centiles, and the largest and smallest values that are not outliers. Outliers are represented as filled circles. P < 0.01 and P < 0.05 by Tukey's post-hoc test.
Figure 2 Effects of both supports on the plasma disappearance rate (PDR) of indocyanine green. The different durations of study are reported along the x-axis: SP, PP 1 H and PP 6 H represent baseline in supine position, prone position at 1 hour and prone position at 6 hours, respectively. Pale gray box, foam mattress; dark gray box, air-cushion mattress. Each box plot represents the median, the 25th and 75th centiles, and the largest and smallest values that are not outliers. Outliers are represented as filled circles. P < 0.05 by Tukey's post-hoc test
Table 1 Characteristics of the population
Patient no. Age (years) Sex BMI LIS SAPS II ARDS duration (days) Causes of ARDS Outcome
1 51 M 42 3.0 39 4 Bacterial pneumonia Alive
2 39 M 27 3.0 60 1 Aspiration pneumonia Alive
3 65 M 33 3.3 46 1 Cytomegalovirus pneumonia Alive
4 64 M 28 2.7 71 2 Bacterial pneumonia Alive
5 50 F 23 2.7 49 7 Septic shock Died
6 44 M 24 3.3 45 1 Bacterial pneumonia Alive
7 71 M 26 3.6 51 7 Bacterial pneumonia Died
8 37 M 34 3.2 30 3 Lung contusion Alive
9 45 M 25 3.3 62 1 Bacterial pneumonia Died
10 37 M 25 3.2 74 6 Community-acquired pneumonia Alive
11 52 F 37 2.7 52 4 Aspiration pneumonia Alive
12 49 M 25 2.7 40 2 Bacterial pneumonia Alive
13 70 M 28 3.2 57 4 Bacterial pneumonia Alive
14 48 F 31 3.5 72 1 Aspiration pneumonia Alive
15 60 F 27 3.2 58 2 Septic shock Alive
16 48 M 29 3.7 44 7 Bacterial pneumonia Alive
17 73 M 38 3.3 62 4 Aspiration pneumonia Died
18 56 M 25 2.7 56 3 Community-acquired pneumonia Alive
19 34 F 28 2.5 32 3 Bacterial pneumonia Alive
20 65 M 23 3.3 48 2 Aspiration pneumonia Died
ARDS duration, duration of acute respiratory distress syndrome before inclusion BMI, body mass index; LIS, lung severity score; SAPS II, Simplified Acute Physiologic Score II
Table 2 Respiratory and hemodynamic parameters
Parameter Foam mattress Specialist mattress ANOVA
Baseline PP, 1 h PP, 6 h Baseline PP, 1 h PP, 6 h Time Group Interaction
PaO2/FiO2 142 ± 63 178 ± 73 217 ± 69* 143 ± 44 206 ± 127* 243 ± 80* P < 0.001 NS NS
PaO2 (mmHg) 45 ± 9 44 ± 9 44 ± 1 0 45 ± 6 46 ± 10 44 ± 11 NS NS NS
QVA/QT (%) 37 ± 8 37 ± 9 35 ± 9* 37 ± 8 36 ± 10 34 ± 10* P < 0.05 NS NS
Paw (cmH2O) 18 ± 3.5 18.2 ± 3 18 ± 3 19 ± 4 19 ± 4 18.4 ± 4 NS NS NS
Cst (ml/cmH2O) 38.4 ± 10 35.4 ± 8* 37.1 ± 9 38.6 ± 11 35.8 ± 8* 37 ± 8 P = 0.007 NS NS
MPAP (mmHg) 27 ± 4 32 ± 7 30 ± 6 27 ± 4 31 ± 7 32 ± 7 NS NS NS
PAOP (mmHg) 12.2 ± 3.0 16.7 ± 4.0* 15.7 ± 4.0* 12.5 ± 5.0 15.8 ± 4.0* 15.9 ± 4.0* P = 0.02 NS NS
CVP (mmHg) 9.6 ± 4.1 13.7 ± 4 13.1 ± 5 9.4 ± 6 13 ± 5.0 13.5 ± 5.0 P = 0.02 NS NS
CI (l/min per m2) 4.3 ± 1.1 4.3 ± 1.2 4 ± 0.9 4.5 ± 0.9 4.1 ± 1.3 4.1 ± 0.9 NS NS NS
EVLWI (ml/kg per m2) 11.2 ± 6.0 11.1 ± 6.2 10 ± 6.5 11.5 ± 7.3 9.5 ± 6.2 10.6 ± 6.0 NS NS NS
ITBV (ml) 1140 ± 475 1100 ± 535 942 ± 330 1165 ± 435 972 ± 205 985 ± 225 NS NS NS
Values are expressed as mean ± SD.
ANOVA, analysis of variance; CI, cardiac index; Cst, respiratory static compliance; CVP, central venous pressure; EVLWI, indexed extravascular lung water; FiO2, fraction of inspired oxygen; ITBV, intrathoracic blood volume; MPAP, mean pulmonary arterial pressure; PaCO2, arterial CO2 tension; PaO2, arterial oxygen tension; PAOP, pulmonary artery occlusion pressure; Paw, plateau airway pressure; PP, prone positioning; QVA/QT, venous admixture. *P < 0.05 compared with baseline by Tukey's post-hoc test.
Table 3 Hepatic and renal variables
Variable On inclusion End of protocol
Bilirubin (μmol/l) 14 ± 9 21 ± 17
ASAT (IU/l) 30 [20–40] 27 [18–52]
ALAT (IU/l) 22 [15–34] 24 [17–49]
Creatinine (μmol/l) 72 [58–100] 75 [58–91]
Prothrombin (%) 63 ± 9 61 ± 8
ASAT, aspartate aminotransferase; ALAT, alanine aminotransferase. For bilirubin and prothrombin, data are expressed as mean ± SD. For transaminases and creatinine, data are expressed as median [interquartile range].
==== Refs
Pelosi P Brazzi L Gattinoni L Prone position in acute respiratory distress syndrome Eur Respir J 2002 20 1017 1028 12412699 10.1183/09031936.02.00401702
Kiefer P Nunes S Kosonen P Takala J Effect of positive end-expiratory pressure on splanchnic perfusion in acute lung injury Intensive Care Med 2000 26 376 383 10872128 10.1007/s001340051170
Kiefer P Morin A Putzke C Wiedeck H Georgieff M Radermacher P Influence of prone position on gastric mucosal-arterial PCO2 gradients Intensive Care Med 2001 27 1227 1230 11534573 10.1007/s001340100999
Hering R Wrigge H Vorwerk R Brensing KA Schroder S Zinserling J Hoeft A Spiegel TV Putensen C The effects of prone positioning on intraabdominal pressure and cardiovascular and renal function in patients with acute lung injury Anesth Analg 2001 92 1226 1231 11323351
Hering R Vorwerk R Wrigge H Zinserling J Schroder S von Spiegel T Hoeft A Putensen C Prone positioning, systemic hemodynamics, hepatic indocyanine green kinetics, and gastric intramucosal energy balance in patients with acute lung injury Intensive Care Med 2002 28 53 58 11819000 10.1007/s00134-001-1166-5
Takata M Wise R Robotham J Effects of abdominal pressure on venous return: abdominal vascular zone conditions J Appl Physiol 1990 69 1961 1972 2076989
Malbrain M Bakajika D Abdominal pressure in the critically ill: measurement and clinical relevance Intensive Care Med 1999 25 1453 8 10702030 10.1007/s001340051098
Diebel L Wilson R Dulchavsky S Saxe J Effect of increased intra-abdominal pressure on hepatic arterial, portal venous, and hepatic microcirculatory blood flow J Trauma 1992 33 279 282 1507294
Matejovic M Rokyta R Radermacher P Krouzecky A Sramek V Novak I Effect of prone position on hepato-splanchnic hemodynamics in acute lung injury Intensive Care Med 2002 28 1750 1755 12447518 10.1007/s00134-002-1524-y
Shelton F Barnett R Meyer E Full-body interface pressure testing as a method for performance evaluation of clinical support surfaces Appl Ergon 1998 29 491 497 9796795 10.1016/S0003-6870(97)00069-0
Ferrell BA Osterweil D Christenson P A randomized trial of low-air-loss beds for treatment of pressure ulcers JAMA 1993 269 494 497 8338511 10.1001/jama.269.4.494
Chiumello D Cressoni M De Grandis E Landi L Racagni M D'Adda A Gattinoni L The chest-abdomen support in prone position Intensive Care Med 2004 30 Suppl 1 S183
Bernard G Artigas A Brigham K Carlet J Falke K Hudson L Lamy M LeGall J Morris A Spragg R Report of the American-European Consensus conference on acute respiratory distress syndrome: definitions, mechanisms, relevant outcomes, and clinical trial coordination. Consensus Committee J Crit Care 1994 9 72 81 8199655 10.1016/0883-9441(94)90033-7
Gattinoni L Mascheroni D Basilico E Foti G Pesenti A Avalli L Volume/pressure curve of total respiratory system in paralysed patients: artefacts and correction factors Intensive Care Med 1987 13 19 25 3558932 10.1007/BF00263552
Mihm F Feeley T Rosenthal M Lewis F Measurement of extravascular lung water in dogs using the thermal-green dye indicator dilution method Anesthesiology 1982 57 116 122 7046519
Hoeft A Vincent J Transpulmonary indicator dilution: an alternative approach for hemodynamic monitoring Yearbook of Intensive Care and Emergency Medicine 1995 Berlin: Springer 593 605
Iberti T Lieber C Benjamin E Determination of intra-abdominal pressure using a transurethral bladder catheter: clinical validation of the technique Anesthesiology 1989 70 47 50 2912315
Gattinoni L Tognoni G Pesenti A Taccone P Mascheroni D Labarta V Malacrida R Di Giulio P Fumagalli R Pelosi P Effect of prone positioning on the survival of patients with acute respiratory failure N Engl J Med 2001 345 568 573 11529210 10.1056/NEJMoa010043
Pelosi P Tubiolo D Mascheroni D Vicardi P Crotti S Valenza F Gattinoni L Effects of the prone position on respiratory mechanics and gas exchange during acute lung injury Am J Respir Crit Care Med 1998 157 387 393 9476848
Doig C Sutherland L Sandham J Fick G Verhoef M Meddings J Increased intestinal permeability is associated with the development of multiple organ dysfunction syndrome in critically ill ICU patients Am J Respir Crit Care Med 1998 158 444 451 9700119
Deitch E Role of the gut lymphatic system in multiple organ failure Curr Opin Crit Care 2001 7 92 98 11373517 10.1097/00075198-200104000-00007
Montgomery A Stager M Carrico C Hudson L Causes of mortality in patients with the adult respiratory distress syndrome Am Rev Respir Dis 1985 132 485 489 4037521
Slutsky A Tremblay L Multiple system organ failure. Is mechanical ventilation a contributing factor? Am J Respir Crit Care Med 1998 157 1721 1725 9620897
Bradley S Bradley G The effect of increased intra-abdominal pressure on renal function in man J Clin Invest 1947 26 1010 1022
Diebel L Wilson R Dulchavsky S Saxe J Effect of increased intra-abdominal pressure on hepatic arterial, portal venous, and hepatic microcirculatory blood flow J Trauma 1992 33 279 283 1507294
Malbrain M Chiumello D Pelosi P Wilmer A Brienza N Malcangi V Bihari D Innes R Cohen J Singer P Prevalence of intra-abdominal hypertension in critically ill patients: a multicentre epidemiological study Intensive Care Med 2004 30 822 829 14758472 10.1007/s00134-004-2169-9
McAuley DF Giles S Fichter H Perkins GD Gao F What is the optimal duration of ventilation in the prone position in acute lung injury and acute respiratory distress syndrome? Intensive Care Med 2002 28 414 418 11967594 10.1007/s00134-002-1248-z
Quintel M Pelosi P Caironi P Meinhardt J Luecke T Herrmann P Taccone P Rylander C Valenza F Carlesso E An increase of abdominal pressure increases pulmonary edema in oleic acid-induced lung injury Am J Respir Crit Care Med 2004 169 534 541 14670801 10.1164/rccm.200209-1060OC
Roch A Michelet P Lambert D Delliaux S Saby C Perrin G Ghez O Bregeon F Thomas P Carpentier J Accuracy of the double indicator method for measurement of extravascular lung water depends on the type of acute lung injury Crit Care Med 2004 32 811 817 15090967 10.1097/01.CCM.0000114831.59185.02
Vollman K Bander J Improved oxygenation utilizing a prone positioner in patients with acute respiratory distress syndrome Intensive Care Med 1996 22 1105 1111 8923079 10.1007/s001340050222
Colmenero-Ruiz M Pola-Gallego de Guzman D Jimenez-Quintana MM Fernandez-Mondejar E Abdomen release in prone position does not improve oxygenation in an experimental model of acute lung injury Intensive Care Med 2001 27 566 573 11355127 10.1007/s001340100858
Gattinoni L Pelosi P Suter PM Pedoto A Vercesi P Lissoni A Acute respiratory distress syndrome caused by pulmonary and extrapulmonary disease. Different syndromes? Am J Respir Crit Care Med 1998 158 3 11 9655699
| 15987398 | PMC1175887 | CC BY | 2021-01-04 16:04:52 | no | Crit Care. 2005 Mar 31; 9(3):R251-R257 | utf-8 | Crit Care | 2,005 | 10.1186/cc3513 | oa_comm |
==== Front
Crit CareCritical Care1364-85351466-609XBioMed Central London cc35141598739910.1186/cc3514ResearchWorldwide research productivity in critical care medicine Michalopoulos Argyris 1Bliziotis Ioannis A 2Rizos Michael 3Falagas Matthew E [email protected] Director, Intensive Care Unit, Henry Dunant Hospital, Athens, Greece2 Research fellow, Alfa Health Care, Athens, Greece3 Attending Physician, Intensive Care Unit, Henry Dunant Hospital, Athens, Greece4 President, Board of Trustees, Alfa Institute of Biomedical Sciences (AIBS), Athens, Greece, and Adjunct Assistant Professor of Medicine, Department of Medicine, Tufts University School of Medicine, Boston, Massachusetts, USA2005 4 4 2005 9 3 R258 R265 25 1 2005 22 2 2005 1 3 2005 7 3 2005 Copyright © 2005 Michalopoulos 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
The number of publications and the impact factor of journals are accepted estimates of the quantity and quality of research productivity. The objective of the present study was to assess the worldwide scientific contribution in the field of critical care medicine.
Method
All research studies published between 1995 and 2003 in medical journals that were listed in the 2003 Science Citation Index (SCI®) of Journal Citation Reports under the subheading 'critical care' and also indexed in the PubMed database were reviewed in order to identify their geographical origin.
Results
Of 22,976 critical care publications in 14 medical journals, 17,630 originated from Western Europe and the USA (76.7%). A significant increase in the number of publications originated from Western European countries during the last 5 years of the study period was noticed. Scientific publications in critical care medicine increased significantly (25%) from 1995 to 2003, which was accompanied by an increase in the impact factor of the corresponding journals (47.4%). Canada and Japan had the better performance, based on the impact factor of journals.
Conclusion
Significant scientific progress in critical care research took place during the period of study (1995–2003). Leaders of research productivity (in terms of absolute numbers) were Western Europe and the USA. Publications originating from Western European countries increased significantly in quantity and quality over the study period. Articles originating from Canada, Japan, and the USA had the highest mean impact factor.. Canada was the leader in productivity when adjustments for gross domestic product and population were made.
==== Body
Introduction
Critical care is an integral part of hospitals, consuming an important proportion of all beds and of the hospital budget. Critical care medicine is thought to account for 1% of the gross domestic product (GDP) in the USA and has been implicated in a disproportionate amount of the increase in hospital costs [1-3]. In a recently reported study conducted over a 16-year period, Halpern and coworkers [4] demonstrated that, 'Critical Care Medicine is increasingly used and prominent in a shrinking U.S. hospital system'. In addition, during recent decades there was increasing utilization of intensive care unit (ICU) resources by the elderly. Although adults younger than 65 years accounted for 37 ICU days/year per 1000 population, patients aged 65–85 years incurred five to six times that rate [5].
Intensive care medicine is a unique discipline. It is practised by physicians from several primary specialties, all having special training in emergency and critical care medicine. Research is an important and special field that intensivists all over the world deal with, in addition to their daily clinical practice. Publications represent a central part of the research process.
The objective of this bibliometric analysis was to examine the geographical origin of biomedical publications in the area of critical care medicine. We also examined the quantity and quality of these publications from around the world.
Methods
Journals
All journals in the 'critical care medicine' category of the Journal Citation Reports (JCR) database, according to the Institute for Scientific Information [6], within the period 1995–2003 were included. To identify whether these journals were included in Index Medicus, we performed a detailed computerized search for each journal in PubMed's database for every year of the study period [7]. Journals included in the 'critical care medicine' category of the JCR database but not indexed in Index Medicus were excluded from the study. We also excluded medical journals referring to emergency medicine and all nursing journals dealing with critical care issues. We did not include articles published before 1995 because the full address of the authors was frequently not registered in PubMed prior to this year. Furthermore, the JCR database provided available data up to the year 2003 at the time of our data collection.
To quantify research productivity, the number of published articles was considered an index of quantity. The mean impact factor of the published articles was considered an indicator of quality. Finally, the product of the number of articles published in a journal multiplied by the impact factor of the journal, for each year studied, was considered a combined indicator of the quantity and quality of research productivity. The sum of the above products from all journals, for each world region within a year, was considered a 'total product' for that region.
Search procedures
A phrase consisting of four parts joined together by the so-called Boolean operators (i.e. AND, OR, and NOT) was used in our search of the PubMed database. Each search was limited to a specific year using the 'Limits' option, which is incorporated into the search engine. We only analyzed data from original articles and reviews, excluding publication types such as letters, editorials and news reports. This was accomplished by selecting publications of type 'journal article [pt]' in the search field of the database ('pt' means publication type). For example, in order to search for articles published in 'intensive care medicine' originating from Europe, the following phrase was used (where 'AD' means 'address'): intensive care medicine [journal] AND journal article [pt] AND (Andorra [AD] OR Austria [AD] OR ... OR Wales [AD]) NOT (Australia [AD] OR Canada [AD] OR ...). Included were all countries from each region cited in the first pair of parenthesis of the search phrase. The second pair of parenthesis, following 'NOT', contained countries to be excluded in order to avoid double counting.
The results of our search (the number of articles produced by each world region in a specific journal within a year) were summed. We confirmed our findings by summing the number of articles retrieved in our search for all different world regions in a specific journal and comparing the sum with the actual total number of articles published in the same journal for a specific year. The total number was obtained from PubMed without using address limits. Using this methodology, we were able to cross examine those articles for which the originating location was either missing or not retrieved in our search. This scenario did occur occasionally, where articles had no registered address or only the affiliated institution or the city (and not the country) was recorded.
If fewer than 5% of the total articles from a specific journal during 1 year had missing/unretrieved addresses, we considered the number of articles retrieved from the search sufficient. On the other hand, if more than 5% of the total articles from a specific journal during 1 year had missing addresses, we performed additional searches for the author's address by checking other articles from the same author within the same year. In order to include addresses for which only cities or areas were registered, we expanded our search criteria, including search phrases with large cities or capitals (e.g. Munchen, London or Moscow) and all of the individual states of the USA.
Using this retrieval system we identified a few addresses that were double-counted in two different regions. For instance, if 'Beth Israel' – the name of several hospitals in the USA – appears in the address field of the article, then this individual article could be counted, for example, in both USA and Asia. To avoid such problems, a large number of initial search results was manually checked and exclusion criteria were added in the second parenthesis of this search string; for example, when searching for Asia, we added the following: NOT (Beth Israel [AD] OR USA [AD]). Two investigators from our team performed independent searches to further strengthen our methodological validity. In cases of disagreement between the two investigators, the findings were discussed at meetings including all authors and final decisions were based on majority consensus.
World regions
For the purposes of the present study, the world was divided into nine regions based on a combination of geographic, economic and scientific criteria [8].: Western Europe, USA, Japan, Canada, Asia, Eastern Europe, Oceania, Latin America and the Caribbean, and Africa. All former socialist countries of Europe and Turkey were included in the category of Eastern Europe. Greenland was designated Western Europe. Japan was studied as a separate region relative to the rest of Asia. Puerto Rico and the Virgin Islands were included within the USA region.
Relationships of research productivity with economic and scientific resources
The relevant 'World Development Indicators' from the online databases of the World Bank were used for further evaluation of the association between research productivity of each region and other factors [9]. The research productivity of different world regions (estimated by the 'total product') was evaluated in relation to the total population, GDP in standard 1995 US dollars and gross national income (GNI) per capita (Atlas method). Data analysis was performed using statistical software SPSS 10.0., SPSS Inc., 233 S. Wacker Drive, Chicago, Illinois 60606, USA.
Results
Of 26 journals related directly to the field of critical care medicine, 16 were listed in the 2003 Science Citation Index (SCI®) of the JCR database under the subheading 'critical care'. Of these, 14 were also indexed in the PubMed database. The titles of these medical journals are presented in Table 1.
A total of 23,403 articles published in journals included in the 'critical care' category of the JCR database and indexed in PubMed within the period 1995–2003 were evaluated in the study. We were able to retrieve 98.2% of all articles (22,976 articles) and categorize them according to the country of origin, based on the methodology described above. Table 2 shows the number of studies originating from each world area/year within the period 1995–2003. In addition, the total number of publications by world region and the relative contribution of each region to the total production of articles, for all journals retrieved, are also presented. The majority of articles published between 1995 and 2003 originated from Western Europe and the USA (76.7%). More articles originated from Western Europe than from USA during the last 6 years of the study period. The USA ranks second, except in years 1995–1997, when production from the USA exceeded that from Western Europe. Asia (excluding Japan) ranks third, Canada fourth, Oceania fifth, and Japan sixth. Eastern Europe, Central and Latin America, and Africa made little contribution in critical care research within this period. A significant increase in the number of publications originating from Western European countries during the last 5 years of the study period was noticed.
Although more articles originated from Western European countries than from the other world regions, the mean impact factor for articles from Western Europe over the study period was lower than the mean impact factors for articles originating from the USA, Canada and Japan (Table 3). Among the regions studied, publications from Eastern Europe had the lowest mean impact factor.
Table 4 presents the 'total product' (summation of [number of published articles in a journal × the impact factor of that journal] for all journals included) for each world region. The USA had the greatest total product, Western Europe ranked second and was followed by Canada, Japan, Asia (excluding Japan) and Oceania. Eastern Europe, Central and Latin America, and Africa made little contribution. Western Europe and the USA exhibited the most significant relative growth in the total product of medical research in the field of critical care over the period 1995–2003, followed by Canada Japan, Asia and Oceania. All other regions exhibited minimal growth in research productivity.
Table 5 presents the total product adjusted for regional population as well as the GDP of the studied regions. Canada ranks first among the world areas with respect to production adjusted for both variables. Oceania also ranks high (second when production was adjusted for GDP and third when it was adjusted for population). USA outweighs Western Europe for both adjustments.
In Fig. 1 we present the association between the GDP in trillions of 1995 US dollars and the 'total product' of research for each region. For all regions there is a positive association between GDP and total product. Publication performance of Canada, Oceania, Western European countries and the USA was better in relation to quantity and quality of articles as compared with the other regions.
Figure 2 shows the association between GNI per capita and the total product of research adjusted for population size for each region. The regions are clustered into three groups. Africa, Asia, Eastern Europe, and Central and Latin America comprise the first group, in which both GNI per capita and total product adjusted for population size are very low. The second group consists of Canada, USA, Oceania and Western Europe, in which the greater the GNI per capita, the greater the population-adjusted total product. Japan stood out as an example of high GNI per capita associated with relatively lower population adjusted total product.
Discussion
Following our evaluation of worldwide trends in research productivity in the field of critical care medicine research over a 9-year period (1995–2003), we conclude that Western Europe produces the most reports on critical care medicine. Western Europe is the only region around the world exhibiting a significant absolute increase in research productivity over the period studied. Nevertheless, although USA produced fewer publications than did Western Europe in this field, the mean impact factor of the published articles from the USA was higher (3.01 versus 2.60). It is remarkable that publications in journals with higher mean impact factors originated from Canada, Japan and Latin America. Although the value of the impact factor as a tool for assessing the quality of a medical journal is controversial, publications in critical care journals from all world areas showed a significant increase in their mean impact factor over this period. However, it should be noted that the average impact factors for anaesthesia and critical care journals, as well as those for other biomedical journals, have tended to increase over recent years for several reasons [10,11].
Scientific publications in critical care medicine increased significantly (25%) from 1995 to 2003, which was accompanied by increased impact factors for these journals (47.4%). Subsequently, the product of the number of published articles multiplied by the impact factor of each journal ('total product') – a combined indicator of research productivity – also increased within the study period. Western Europe and USA together produced 76.7% of the total number of articles published in the field of critical care medicine. These two world regions were superior to all others in terms of total research productivity. It is clear that scientific productivity from these two world regions in this new discipline has increased exponentially over the period of study. In contrast, the contributions of other world regions to research productivity were low, especially those from low income areas such as Eastern Europe, Latin America and the Caribbean, and Africa. This might be because critical care medicine was mainly developed in countries with vigorous economies, because the cost of hospitalization in ICUs is high. It should be emphasized that several factors, including resources, interest in research and language barriers, influence research productivity by various areas of the world.
When total product was adjusted for GDP and/or regional population, Canada ranked first and Oceania ranked very high. Thus, these two regions are clearly among world leaders in research in this field, but because of their relatively small populations (and consequently relatively small GDPs) their absolute number of publications is small. In two previous studies that we conducted using the same methodology, one in the 'Cardiac and cardiovascular systems' category [12] and one in the 'Microbiology' category [13] of the JCR database, the results were similar to those of the present study. Again, Western Europe and USA were the leaders in terms of absolute number of papers, whereas Canada and Oceania were in the top positions when adjustments for GDP and population were made.
Although intensivists from the USA led the research in critical care medicine, their colleagues from Western European countries made greater contribution during the last 6 years of study. It is noteworthy that North America and Canada performed better than Europe in terms of mean impact factor of publications. Similar findings were reported for fields other than critical care (i.e. cardiology, clinical cancer, microbiology and radiology) [12-16].
We should like to acknowledge several limitations of this study. First, we used JCR criteria for inclusion of medical journals in the present study. Articles published in non-JCR cited journals were not included, but we do recognize that they contribute to scientific production [17]. This pertains in particular to originating regions in which English is not the native language (i.e. Eastern Europe and Japan), where researchers tend to publish their findings in regional journals of their own language [18]. We also used Medline, which is an easily accessible and widely used database. It should also be emphasized that, in Medline, only the address of the first author is presented; that a study might be the result of multinational collaboration is not taken into account. Furthermore, it is known that there are many medical journals on critical care medicine from all over the world in languages that are not indexed.
In addition, one should take into account that the impact factor, as an index of quality of scientific research, has often been criticized [19,20]. Impact factors change every 12 months, and so they are not very responsive to change. However, the impact factor is yet to be replaced by another internationally accepted method [21]. Furthermore, the division of the world into regions could be done in several different ways, based on various criteria (e.g. Canada could be grouped with the USA, and Japan could be studied together with the rest of Asia). We believe that our categorization takes into account geographic, economic and, most importantly, scientific criteria (i.e. Canada and Japan represent powerful scientific world regions on their own).
Finally, when interpreting the results, one should take into account the fact that many articles regarding critical care medicine are published in journals other than those included in the 'critical care medicine' category. Furthermore, a proportion of articles published in journals included in the 'critical care medicine' category of the Science Citation Index are related to non-critical-care topics. However, it seems that there is no systematic bias in the analysis of these data, because there is no specific reason to publish articles on this subject in journals included in other JCR categories, especially from specific world regions, and neither is there any reason for non-critical-care articles to originate mainly from certain regions.
Conclusion
We took a global view of worldwide trends in research productivity in the field of critical care medicine over a 9-year period. It is notable that Western Europe and USA ranked top in terms of quantity and quality of published articles in absolute numbers, whereas Canada was the leader in productivity when adjustments for GDP and population were made. As expected, developed world regions ranked first in quantity and quality of published articles, and had greater productivity adjusted for population.
Our data may be used to compare the productivity of different world regions with diverse economic status and priorities for funding different social needs. The World Health Organization, the World Bank, other United Nations organizations and national governments should encourage biomedical research in less developed parts of the world. Intensivists all over the world must acquire and maintain the necessary skills to provide state-of-the art clinical care for critically ill patients so that they may confront life-threatening disease, improve patient outcomes, optimize the use of limited ICU resources and, in parallel, advance the theory and practice of critical care medicine. The quality of care provided in ICUs worldwide has improved over the past decade. Nevertheless, disorders such as adult respiratory distress syndrome, sepsis and ICU-acquired infections remain foci of interest, and are difficult to manage and associated with high mortality rates. Consequently, further research studies on several fields are urgently needed.
Competing Interests
The author(s) declare that they have no competing interests.
Key messages
• Leaders of research productivity in critical care medicine, in terms of absolute numbers of published papers during the study period (1995–2003), were Western Europe and the USA.
• Articles originating from Canada, Japan, and the USA had the highest mean impact factor.
• Canada was the leader in productivity when adjustments for gross domestic product and population were made.
Abbreviations
GDP = gross domestic product; GNI = gross national income; ICU = intensive care unit; JCR = Journal Citation Reports.
Authors' contributions
AM and MEF conceived the study. IAB and MR collected data. All authors contributed to the writing and preparation of the manuscript.
Figures and Tables
Figure 1 'Total product' of research productivity. Scatter plot depicting the relationship of the annual total product of research productivity (number of articles published × their impact factor [IF]) in critical care medicine, for different world regions, for the period 1995–2003, with the gross domestic product (GDP) in trillions of 1995 US dollars.
Figure 2 Relation of 'total product' adjusted for population to GNIPC. Scatter plot depicting the relationship of the annual "total product" of research productivity for different world regions (number of articles published multiplied by their impact factor) in Critical Care Medicine, with the population of a region and the gross national income per capita (GNIPC)
Table 1 Summary of journals related to critical care medicine included in the study
Abbreviated journal title Impact factor (2003)
Am J Resp Crit Care 8.87
Crit Care Med 4.19
Intensive Care Med 2.97
J Neurotrauma 2.58
Shock 2.54
Crit Care 1.91
Crit Care Clin 1.48
J Trauma 1.42
Resuscitation 1.37
J Crit Care 1.29
Burns 1.12
Anaesth Intens Care 0.77
Injury 0.51
Anasth Intensiv Notf 0.31
Table 2 Number of articles published in critical care medicine: 1995–2003
World area Year(s)
1995 1996 1997 1998 1999 2000 2001 2002 2003 1995–2003
Western Europe 720 (32.7) 829 (35.2) 801 (36.3) 870 (38.6) 1037 (41.3) 1324 (43.2) 1223 (43.1) 1125 (40.2) 1147 (41.7) 9076 (39.5)
USA 982 (44.6) 978 (41.5) 853 (38.6) 829 (36.8) 904 (36.0) 1088 (35.5) 992 (35.0) 972 (34.8) 956 (34.8) 8554 (37.2)
Asia (excluding Japan) 114 (5.2) 128 (5.4) 132 (6.0) 134 (6.0) 138 (5.5) 154 (5.0) 169 (6.0) 193 (6.9) 186 (6.8) 1348 (5.9)
Canada 113 (5.1) 144 (6.1) 126 (5.7) 126 (5.6) 139 (5.5) 134 (4.4) 123 (4.3) 129 (4.6) 130 (4.7) 1164 (5.1)
Oceania 120 (5.5) 116 (4.9) 128 (5.8) 123 (5.5) 111 (4.4) 120 (3.9) 128 (4.5) 140 (5.0) 130 (4.7) 1116 (4.9)
Japan 94 (4.3) 92 (3.9) 103 (4.7) 107 (4.8) 110 (4.4) 173 (5.6) 120 (4.2) 126 (4.5) 101 (3.7) 1026 (4.5)
Eastern Europe 24 (1.1) 39 (1.7) 29 (1.3) 28 (1.2) 35 (1.4) 30 (1.0) 33 (1.2) 61 (2.2) 46 (1.7) 325 (1.4)
Latin America and the Caribbean 8 (0.4) 13 (0.6) 16 (0.7) 19 (0.8) 21 (0.8) 29 (0.9) 31 (1.1) 35 (1.3) 34 (1.2) 206 (0.9)
Africa 25 (1.1) 19 (0.8) 20 (0.9) 16 (0.7) 13 (0.5) 15 (0.5) 17 (0.6) 16 (0.6) 20 (0.7) 161 (0.7)
Total 2200 (100) 2358 (100) 2208 (100) 2252 (100) 2508 (100) 3067 (100) 2836 (100) 2797 (100) 2750 (100) 22976 (100)
Shown are the numbers of articles published in journals included in the 'critical care medicine' category of the Journal Citation Report database and indexed by PubMed, from different world regions, for the period 1995–2003. Values are expressed as number of articles (%) within a calendar year.
Table 3 Mean impact factors of articles published in critical care medicine: 1995–2003
World area Year
1995 1996 1997 1998 1999 2000 2001 2002 2003 1995–2003
Western Europe 1.95 2.64 2.43 2.80 2.67 2.80 2.58 2.39 2.91 2.60
USA 2.30 3.01 2.80 2.98 3.16 3.21 3.09 3.07 3.39 3.01
Asia (excluding Japan) 1.10 1.49 1.44 1.57 1.72 2.06 1.59 1.76 2.19 1.70
Canada 2.79 3.66 3.38 3.62 4.10 3.92 3.77 3.71 4.20 3.70
Oceania 1.53 1.53 1.74 1.90 1.87 2.46 2.50 2.21 2.31 2.01
Japan 2.65 3.26 2.99 2.82 3.33 3.47 3.00 3.13 3.93 3.19
Eastern Europe 0.86 1.46 1.09 1.56 1.44 2.00 1.83 1.43 1.99 1.55
Latin America and the Caribbean 2.42 3.30 2.23 2.85 2.68 2.65 2.57 2.23 3.11 2.66
Africa 0.98 2.29 1.75 1.52 2.06 1.16 1.26 1.08 2.11 1.57
Mean (for all regions) 2.09 2.74 2.53 2.76 2.85 2.97 2.75 2.64 3.08 2.73
Shown are the mean impact factors of articles published in journals included in the 'critical care medicine' category of the Journal Citation Report database and indexed by PubMed, from different world regions, for the period 1995–2003.
Table 4 'Total product' of articles published in 'critical care medicine': 1995–2003
World area Year
1995 1996 1997 1998 1999 2000 2001 2002 2003 1995–2003
Western Europe 1401 2186 1944 2434 2772 3713 3158 2694 3340 23,642
USA 2263 2941 2390 2468 2860 3492 3068 2989 3244 25,715
Asia (excluding Japan) 125 191 190 211 237 317 269 341 407 2288
Canada 316 526 426 456 570 525 464 479 546 4308
Oceania 184 178 222 233 207 295 320 310 300 2249
Japan 249 300 308 302 367 600 360 394 397 3276
Eastern Europe 21 57 32 44 50 60 61 87 91 502
Latin America and the Caribbean 19 43 36 54 56 77 80 78 106 549
Africa 24 44 35 24 27 17 21 17 42 252
Total 4602 6464 5583 6226 7147 9096 7800 7389 8473 62781
Shown are the total products of articles (number of articles published × their impact factor) published in journals included in the 'critical care medicine' category of the Journal Citation Report database and indexed by PubMed, from different world regions, for the period 1995–2003.
Table 5 Product of number of articles and impact factor, adjusted for population and GDP
World Areas Number of articles multiplied by their impact factor/population of the area (in millions) Number of articles multiplied by their impact factor/GDP of the area (in hundreds of billions 1995 US dollars)
Western Europe 6.74 26.3
USA 10.10 33.5
Asia (excluding Japan) 0.07 7.1
Canada 15.67 71.5
Oceania 8.15 47.9
Japan 2.88 6.5
Eastern Europe 0.13 5.1
Latin America and the Caribbean 0.12 3.2
Africa 0.04 5.2
Shown are the numbers of articles in journals included in the 'critical care medicine' category of the Journal Citation Report database multiplied by their impact factors, adjusted for population and gross domestic product (GDP).
==== Refs
Gipe BT Financing critical care medicine in 2010 New Horiz 1999 7 184 197
Halpern N Bettes L Greenstein R Federal and nationwide intensive care units and healthcare costs: 1986–1992 Crit Care Med 1994 22 2001 2007 7988140
Chalfin DB Cohen IL Lanken PN The economics and cost-effectiveness of critical care medicine Intensive Care Med 1995 21 952 961 8636530
Halpern NA Pastores SM Greenstein RJ Critical care medicine in the United States 1985–2000: an analysis of bed numbers, use, and costs Crit Care Med 2004 32 1254 1259 15187502 10.1097/01.CCM.0000128577.31689.4C
Rainey T Shapiro MJ Critical care medicine for the 21st century Crit Care Med 2001 29 436 437 11246327 10.1097/00003246-200102000-00040
Institute for Scientific Information SCI: Science Citation Index – Journal Citation Reports, 1996–2000 2004 Philadelphia: The Institute for Scientific Information
National Library of Medicine Index Medicus Database (PubMed) 2004 Bethesda, MA: National Library of Medicine
United Nations United Nations Statistical Yearbook, 42nd issue, CD-Rom Edition 2004 New York, NY: United Nations
World Bank World Development Indicators 2002, CD-ROM Edition 2004 World Bank; Washington, DC, USA
Boldt J Haisch G Maleck WH Changes in the impact factor of anesthesia/critical care journals within the past 10 years Acta Anaesthesiol Scand 2000 44 842 849 10939697 10.1034/j.1399-6576.2000.440710.x
Jemec GBE Impact factors of dermatological journals for 1991–2000 BMC Dermatol 2001 1 7 11710969 10.1186/1471-5945-1-7
Rosmarakis ES Vergidis PI Soteriades ES Paraschakis K Papastamataki PA Falagas ME Estimates of global production in cardiovascular diseases research Intern J Cardiol 2005
Vergidis PI Karavasiou AI Paraschakis K Bliziotis I Falagas ME A bibliometric analysis of global trends of research productivity in microbiology Eur J Clin Microbiol Infect Dis 2005
De Jong JW Schaper W The international rank order of clinical cardiology Eur Heart J 1996 17 35 42 8682128
Grossi F Belvedere O Rosso R Geography of clinical cancer research publications from 1995 to 1999 Eur J Cancer 2003 39 106 111 12504666 10.1016/S0959-8049(02)00239-3
Mela GS Martinoli C Poggi E Derchi LE Radiological research in Europe: a bibliometric study Eur Radiol 2003 13 657 662 12664100
Winkmann G Schweim HG Medical-bioscientific databanks and the Impact Factor Dtsch Med Wochenschr 2000 125 1133 1141 11147369 10.1055/s-2000-7581
Coates R Sturgeon B Bohannan J Pasini E Language and publication in 'cardiovascular research' articles Cardiovasc Res 2002 53 279 285 11827675 10.1016/S0008-6363(01)00530-2
Neuberger J Counsell C Impact factors: uses and abuses Eur J Gastroenterol Hepatol 2002 14 209 211 11953682 10.1097/00042737-200203000-00001
Whitehouse GH Impact factors: facts and myths Eur Radiol 2002 12 715 717 11960216 10.1007/s00330-001-1212-2
Luukkonen T Bibliometrics and evaluation of research performance Ann Med 1990 22 145 150 2393549
| 15987399 | PMC1175888 | CC BY | 2021-01-04 16:04:52 | no | Crit Care. 2005 Apr 4; 9(3):R258-R265 | utf-8 | Crit Care | 2,005 | 10.1186/cc3514 | oa_comm |
==== Front
Crit CareCritical Care1364-85351466-609XBioMed Central London cc35171598740010.1186/cc3517ResearchIn vivo validation of the adequacy calculator for continuous renal replacement therapies Ricci Zaccaria [email protected] Gabriella 2Bonello Monica 3Pisitkun Tirak 4Bolgan Irene 5D'Amico Giuseppe 6Dan Maurizio 7Piccinni Pasquale 7Ronco Claudio 81 Consultant, Department of Intensive Care, Policlinico Umberto I, Rome, Italy2 Research fellow, Department of Nephrology, St Bortolo Hospital, Vicenza, Italy3 Specialist registrar, Department of Nephrology, St Bortolo Hospital, Vicenza, Italy4 Research fellow, Department of Nephrology, St Bortolo Hospital, Vicenza, Italy5 Statistician, Department of Nephrology, St Bortolo Hospital, Vicenza, Italy6 Research fellow, Department of Intensive Care, Policlinico Umberto I, Rome, Italy7 Head, Department of Intensive Care, St Bortolo Hospital, Vicenza, Italy8 Head, Department of Nephrology, St Bortolo Hospital, Vicenza, Italy2005 7 4 2005 9 3 R266 R273 23 9 2004 19 10 2004 22 2 2005 14 3 2005 Copyright © 2005 Ricci 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
The study was conducted to validate in vivo the Adequacy Calculator, a Microsoft Excel-based program, designed to assess the prescription and delivery of renal replacement therapy in the critical care setting.
Methods
The design was a prospective cohort study, set in two intensive care units of teaching hospitals. The participants were 30 consecutive critically ill patients with acute renal failure treated with 106 continuous renal replacement therapies (CRRT). Urea clearance computation was performed with the Adequacy Calculator (KCALC). Simultaneous blood and effluent urea samples were collected to measure the effectively delivered urea clearance (KDEL) at the beginning of each treatment and, during 73 treatments, between the 18th and 24th treatment hour. The correlation between 179 computed and 179 measured clearances was assessed. Fractional clearances for urea were calculated as spKt/V (where sp represents single pool, K is clearance, t is time, and V is urea volume of distribution) obtained from software prescription and compared with the delivered spKt/V obtained from empirical data.
Results
We found that the value of clearance predicted by the calculator was strongly correlated with the value obtained from computation on blood and dialysate determination (r = 0.97) during the first 24 treatment hours, regardless of the renal replacement modality used. The delivered spKt/V (1.25) was less than prescribed (1.4) from the Adequacy Calculator by 10.7%, owing to therapy downtime.
Conclusion
The Adequacy Calculator is a simple tool for prescribing CRRT and for predicting the delivered dose. The calculator might be a helpful tool for standardizing therapy and for comparing disparate treatments, making it possible to perform large multi-centre studies on CRRT.
==== Body
Introduction
Acute renal failure (ARF), as a component of the multiple organ failure syndrome, affects morbidity and mortality in critically ill patients [1]. This is still the case even though several aspects of medical care and applied technology in ARF were improved. Much development in renal replacement therapy (RRT) is ongoing, concerning new techniques, new membranes, and new integrated equipment. However, it is still unclear whether a correlation between treatment dose and outcome exists and no consensus has been reached on how much treatment is adequate [2,3]. A long-term, large-scale, multi-center study to determine how the outcome of critically ill patients is affected by RRT dose and modality (intermittent or continuous, diffusive or convective) is still lacking [4,5]. This is in part due to the complexity of data collection and to the variety of existing standards in RRT prescription and dose evaluation.
We tested a computer program called 'Adequacy Calculator for ARF', a simple and manageable tool designed to prescribe RRT dose and to collect information about the quantity of delivered treatments. Pisitkun and colleagues [6] have described this Microsoft excel based program and its algorithms in a previous paper. Once the required parameters are entered, it calculates urea clearance and fractional clearance, spKt/V (sp = single pool; K = clearance, t = time, V = urea volume of distribution) for each continuous RRT (CRRT) modality.
Materials and methods
We prospectively collected data from 106 consecutive continuous renal replacement treatments administered to 30 patients with acute renal failure in the intensive care unit of St Bortolo Hospital and Policlinico Umberto I in the period from March 2003 to November 2004. The decision to start and to withdraw RRT, anticoagulation and prescription of net ultrafiltration rate were left to institutional protocols (Table 1). Treatments were delivered at different modalities and machine settings depending on the preference of the prescribing physician, but a final spKt/V of 1.4 had to be prescribed by means of the Adequacy Calculator. The plasma filtration fraction, in the case of postfilter reinfusion of replacement solution, was kept below 20%. By protocol, filters were changed after 24 hours of treatment, or earlier if clotting occurred. The available membranes were Diacap α (1.2 m2, polysulphone, B Braun) and Aquamax HF 12 (1.2 m2; polyethersulphone; Edwards Lifescience) for 59 and 47 treatments respectively. Sixty-four treatments were performed with bicarbonate-buffered replacement and dialysate fluids, and 42 with lactate-buffered fluids. Daily operative treatment times and downtimes were reported. Intermittent treatments were excluded from the analysis.
The Adequacy Calculator estimated urea clearance (KCALC) for each different modality and machine setting (Additional file 1). The calculator estimation is founded on the assumption that urea sieving coefficient is equal to unity for convective therapies; at the same time the calculator assumes that complete saturation of spent dialysate occurs under continuous veno-venous hemodialysis (CVVHD) conditions.
To correlate KCALC with effectively delivered instantaneous urea clearance (KDEL), simultaneous samples from prefilter blood and effluent were collected during each treatment, to measure urea concentration; 106 blood and 106 effluent samples were withdrawn during the first hour from the start of RRT; 73 blood-effluent samples were withdrawn between the 18th and 24th hours (in 33 cases treatment was stopped before the 18th hour). KDEL was calculated as described in Additional file 1. One hundred and six KDEL values at treatment start (T0) and 73 values after 18 to 24 hours of treatment (T18) were correlated with 106 and 73 KCALC values obtained for the same treatments.
The calculator prescribed spKt/VCALC after KCALC, the expected treatment time and patient's body weight (for assessment of urea volume of distribution, V; Fig. 1) had been entered. spKt/VDEL was calculated from KDEL, V and effective operative treatment times (Additional file 1).
Statistical analysis
Statistical analysis was performed with the SPSS 11.5 software package. Data are reported as means ± standard deviation (SD). Correlations between estimated and measured urea clearance were performed with the Pearson correlation coefficients (r). spKt/V, KCALC and KDEL have no normal distribution, so we used a Mann-Whitney test (between two samples) or a Kruskal-Wallis test (between three or more samples) to indicate whether groups had different locations. P < 0.05 was considered statistically significant.
Results
A total of 106 RRTs administered to 30 patients were analysed with the Adequacy Calculator. An average of 3.5 treatment days was examined for each patient. Nineteen post-dilution continuous veno-venous hemofiltrations (CVVHs), 23 pre-dilution CVVHs, 23 CVVHDs and 41 post-dilution continuous veno-venous hemodiafiltrations (CVVHDF) were prescribed. The duration of each treatment was 17 ± 6 hours. The daily operative treatment time was 20 ± 3 hours with a downtime of 3 ± 2 hours. Thirty-three treatments lasted less than 18 hours (16 CVVH and 17 CVVHD); 73 treatments lasted more than 18 hours (26 CVVH, 6 CVVHD and 41 CVVHDF). Examined clearances ranged from 15 ml min-1 to 100 ml min-1 (Table 2), this wide range being explained by variability in patients' weights and prescribed treatment times: because the prescribed spKt/VCALC was maintained at a constant 1.4, a 35 kg patient treated for 24 hours with a KCALC of 20 ml min-1 obtained the same fractional clearance as a 98 kg patient dialyzed for 12 hours with a KCALC of 100 ml min-1.
The difference between KDEL and KCALC was -1.75 ± 5.9 ml min-1. Applying a Pearson correlation we obtained r = 0.97; a significant (P = 0.022) decrease in calculator accuracy in predicting effectively delivered clearance was obtained when data from the KCALC < 60 ml min-1 subgroup (r = 0.95) were compared with data from the KCALC > 60 ml min-1 subgroup (r = 0.89). A Bland-Altman analysis (Fig. 2) confirmed high correlation: this result was particularly evident up to an average clearance ([KDEL + KCALC]/2) of 60 ml min-1, with the KDEL - KCALC difference never exceeding a standard deviation of 5.9 ml min-1, whereas for [KDEL + KCALC]/2 > 60 ml min-1, the KDEL - KCALC difference tended to increase. However, we found that 155 of 179 (87%) KDEL values fell within a ± 15% KCALC error: in 5 cases the calculator underestimated, and in 19 overestimated, the delivered clearance. No significant KDEL - KCALC difference was observed when T0 and T18 clearances were analysed (P = 0.54) and no significant difference (P = 0.394) was observed when KCALC > 60 ml min-1 in the T0 subgroup and KCALC > 60 ml min-1 in the T18 subgroup were analyzed: calculator accuracy was not affected by filter lifespan (Table 3).
After analysis of each modality group, correlations were still high: rCVVHpre = 0.96, rCVVHpost = 0.96, rCVVHD = 0.97 and rCVVHDF = 0.98 (Fig. 3), with no significant difference between groups (P = 0.099) (Table 3).
Membrane type did not affect the KDEL - KCALC correlation: r obtained for Diacap M and Aquamax HF 12 were 0.96 and 0.97 respectively (P = 0.1).
The average spKt/VDEL obtained during our treatments was 1.25 ± 0.6; the delivered/prescribed ratio was 0.89 (Table 3); the delivered fractional clearance was significantly less than the prescribed spKt/VCALC of 1.4 (P = 0.045).
Discussion
Ideal marker molecules and performance parameters to compare treatment dose in different techniques are difficult to establish. In spite of its moderate toxicity, urea is currently used as a marker of RRT adequacy because it is easily measurable and, representing the end of protein metabolism, its accumulation during kidney failure defines the requirement for dialysis while its elimination defines the efficiency of treatment. Because urea is equally distributed at steady state in body water compartments, its volume of distribution (V) equals total body water. Urea is therefore a surrogate of the low-molecular-mass toxins. In chronic hemodialysis, the treatment dose of RRT is defined as a fractional clearance, Kt/V, where K is the instantaneous clearance, t is treatment time and V is the volume of distribution of the marker molecule. This is a dimensionless parameter that represents the efficacy of treatments, and allows comparison between different therapies and among different patients. In fact, different instantaneous clearances, representing treatment efficiency, can yield comparable results in terms of efficacy only if correlated with treatment time and the patient's total body water. A Kt/V value of 1.2 is an established maker of adequacy that has been shown to be correlated with morbidity and mortality in patients with end-stage kidney disease [7-11]. Kt/V has not yet been validated as a marker of adequacy in patients with acute renal failure, but it seems that a good rationale exists for its use in continuous therapies. Theoretically, in its original conception, clearance was thought to evaluate renal function of disparate individuals whose kidneys were operating 24 hours per day and blood levels were at steady state. Similarly, after some days of CRRT, patients' urea levels approach a real steady state (never obtained with intermittent dialysis) and post-dialysis rebound is not present. It is thus reasonable to consider urea distribution volume as in a single-pool kinetic model (spKt/V).
Recently, Brause and colleagues [12] stated that spKt/V is a valuable tool for evaluating continuous hemofiltration, and higher values (0.8 versus 0.53) were correlated to improve uremia control and acid–base balance. Ronco and colleagues [2] showed an improved outcome with postdilution hemofiltration delivered at 35 ml h-1 kg-1 in a 450-patient population. Setting a spKt/V threshold that could guide clinicians towards adequate treatments, we should possibly meet the target of 35 ml h-1 kg-1, which, delivered as a 24-hour treatment, may translate into a spKt/V of 1.4 independently of the RRT modality.
We found that the Adequacy Calculator was able to predict the delivered urea clearance accurately, regardless of which CRRT modality was selected; the correlation between prediction and effective delivery remained high over a time range of 24 hours. When clearances above 60 ml min-1 were prescribed, the calculator showed a tendency to overestimate effective clearances: this overestimation remained generally within an error of 15%.
Considering our results and the dissociation between treatment delivery and calculator estimation when high clearances are involved, as could occur with low-efficiency extended dialysis or high-volume hemofiltration, a slight correction to prevent the overestimation of effective treatment delivery is strongly advised. Nevertheless, even in the presence of an error of up to 15%, which is unlikely to occur, the delivered Kt/V in 24 hours will always approach the target value of 1.2.
The use of the calculator allowed us to strictly monitor our treatments during the study period and described an average 10.7% (P < 0.05) decrease in delivery of therapy in comparison with prescribed dose. The differences between prescribed and delivered dose in critically ill patients with ARF undergoing intermittent hemodialysis were analyzed by Evanson and colleagues [13]; they found that only 30% of dialysis delivered a Kt/V of 1.2; high patient weight, male sex and low blood flow were the limiting factors affecting RRT administration. In our population, this decrease in delivery was sometimes due to overestimation of KCALC by the calculator, and, more often, to operative treatment time, which was often shorter than the prescribed treatment time (during bag substitution and filter change the treatment was not administered). Our observation is consistent with a recent large retrospective analysis [14]. In this setting, when a 'standardized' downtime is foreseen, treatment prescription might be adjusted to correct for the time of zero clearance.
However, all these considerations must be seen in the light of an absolute lack of any previous attempt to adjust treatment dose to specific target levels. Furthermore, a clear understanding of adequate levels of renal replacement therapy has yet to be achieved. In this state of absence of information and of wide ignorance of the field, the calculator might have the merit of placing the issue of treatment dose among the priorities of critical care nephrology: a dose prescription should be made before embarking on an extracorporeal blood purification technique, and the delivered treatment dose should be monitored.
The limitations of this study are as follows. A subgroup analysis of net ultrafiltration (UF) prescription, daily treatment length and downtime difference within different modalities was not performed: in our opinion these factors do not affect Adequacy Calculator accuracy. Slight subgroup disparities in KCALC prescription within different modalities were present because prescribing physicians were not asked to modify their usually preferred modality. The effect of different blood pump flow rates on error in KCALC was not evaluated: higher blood flow rates might have decreased some KCALC - KDEL differences, especially when high-volume treatments were used. The observational nature of our study did not allow us to analyse all possible prescriptions systematically: a dedicated study should be performed. Finally, partial thromboplastin time, prothrombin time, platelet levels, anticoagulation and administration of drotrecogin alfa were not taken into consideration; however, our study showed that, during a period of 24 hours, urea sieving coefficient and clearance were not significantly affected by treatment duration and, indirectly, by progressive filter clogging. In our experience, anticoagulation parameters affect the lifespan of membranes in the first 24 hours but do not affect urea clearance.
Conclusion
We assume that by using simple CRRT parameters and the Adequacy Calculator it is possible to simply prescribe and closely monitor the dose of different continuous therapies. This tool might help in future prospective studies to correlate different dose prescriptions with different clinical outcomes.
Key messages
• The Adequacy Calculator is a Microsoft Excel-based program, designed to assess the prescription and delivery of renal replacement therapy in the critical care setting.
• A prospective study was performed in order to evaluate correlation between calculated and measured clearances.
• The value of clearance predicted by the calculator was strongly correlated with the value obtained from determination on blood and dialysate: the Adequacy Calculator is a reliable tool for prescribing CRRT and for predicting the delivered dose.
Abbreviations
ARF = acute renal failure; CRRT = continuous renal replacement therapy; CVVH = continuous veno-venous hemofiltration; CVVHD = continuous veno-venous hemodialysis; CVVHDF = continuous veno-venous hemodiafiltration; K = clearance; KCALC = calculator-estimated urea clearance; KDEL = delivered clearance evaluated from urea concentrations on simultaneous blood and effluent samples; RRT = renal replacement therapy; spKt/V = single pool fractional clearance for urea;t = time; V = urea volume of distribution.
Competing interests
The author(s) declare that they have no competing interests.
Authors' contributions
ZR designed the study, participated in data collection and drafted the paper. MB, GS, EA and GD participated in data collection. IB provided statistical expertise. MD and PP revised the article. CR designed the study and participated in data interpretation. All authors read and approved the final manuscript.
Supplementary Material
Additional File 1
A pdf file containing Adequacy Calculator algorithms for urea clearance and single pool fractional clearance computation is provided.
Click here for file
Figures and Tables
Figure 1 An Adequacy Calculator worksheet: continuous veno-venous hemodiafiltration (CVVHDF) is delivered in a 70 kg patient. Post-dilution mode is selected, machine settings and prescribed treatment time per day are entered on the upper left panel: estimated urea clearance (KCALC) and 'daily Kt/V' (spKt/VCALC) are displayed on the right. In the lower left panel it is possible to obtain KDEL measure after entering prefilter blood (Cbi) and effluent (Cdo) urea concentration: in this case, when operative times are entered, 'daily Kt/V' cell displays effectively delivered fractional clearance (spKt/VDEL).
Figure 2 Bland–Altman correlation between urea clearance obtained by two methods: urea clearance calculated with the described software (KCALC) and urea clearance obtained by direct measure on prefilter blood and effluent samples (KDEL). It is possible to distinguish the correlations between KCALC and KDEL, at therapy start (T0) and after 18 to 24 hours of uninterrupted therapy (T18).
Figure 3 Bland–Altman analysis. The same data as above are used; here it is possible to distinguish between different modalities. Parallel lines indicate standard deviation. CVVH, continuous veno-venous hemofiltration; CVVHD, continuous veno-venous hemodialysis; CVVHDF, continuous veno-venous hemodiafiltration; KCALC, calculator-estimated urea clearance; KDEL, delivered clearance evaluated from urea concentrations on simultaneous blood and effluent samples.
Table 1 Characteristics of patients
Parameter Value (mean ± SD; range)
Total no. of patients 30
Age (years) 58 ± 12
Weight (kg) 73.7 ± 15.7 (35–105)
SAPS II score 38 ± 12
No. of evaluated treatments 106
Examined treatment days 3.5 ± 1.1
Prescribed spKt/V 1.4
Diagnosis
Sepsis/septic shock 8
Bowel perforation 3
Aortic aneurysm repair 3
Pneumonia 5
Hemorrhagic shock 2
Pancreatitis 1
Multiple trauma 5
Cardiogenic shock 3
Urine output at RRT start (ml per 24 h) 150 ± 70
Urea levels at RRT start (mg dl-1) 124 ± 55
Urea levels after 24 h from RRT start (mg dl-1) 98 ± 50
Urea levels at RRT withold 68 ± 53
Creatinine levels at RRT start (mg dl-1) 2.9 ± 0.9
Creatinine levels after 24 h from RRT start (mg dl-1) 2.4 ± 1.25
Creatinine levels at RRT withold 1.8 ± 0.6
No. of patients receiving heparin < 500 U h-1 12
No. of patients receiving heparin > 500 U h-1 11
No. of patients receiving no anticoagulation 7
RRT, renal replacement therapy; SAPS II, Systems Approach Problem Solver II.
Table 2 Treatments characteristics
Parameter Value
No. of treatments 106
Total no. of examined KCALC/KDEL(ml min-1) 179 (106 T0; 73 T18)
CVVHpost 38 (19 T0; 19 T18)
CVVHpre 30 (23 T0; 7 T18)
CVVHD 29 (23 T0; 6 T18)
CVVHDF 82 (42 T0; 42 T18)
Total prescribed KCALC (ml min-1) 48.6 ± 24; 18–100
CVVHpost 35.6 ± 16; 16–66
CVVHpre 48.8 ± 19; 15–83
CVVHD 56.2 ± 28; 20–100
CVVHDFa 52.2 ± 24; 23–100
Total prescribed net UF (ml h-1) 250 ± 100; 0–500
Treatment hours per day 20 ± 3; 8.5–24
Treatment hours per run 17 ± 6; 5–24
Downtime (hours) 3 ± 2
aDuring continuous veno-venous hemodiafiltration (CVVHDF) modality, prescribed clearance was delivered with even hemofiltration and hemodialysis flow rates. Where errors are given, results are means ± SD; ranges follow a semicolon. CVVH, continuous veno-venous hemofiltration; CVVHD, continuous veno-venous hemodialysis; KCALC, calculator-estimated urea clearance; KDEL, delivered clearance evaluated from urea concentrations on simultaneous blood and effluent samples; T0, at therapy start; T18, after 18 to 24 hours of uninterrupted therapy.
Table 3 Calculated-delivered urea clearance correlation
Parameter KCALC - KDEL (ml min-1) r
Total -1.7 ± 5.9 0.97
CVVHpost -2 ± 4.3 0.96
CVVHpre 1.4 ± 6 0.96
CVVHD -4.5 ± 7.7 0.97
CVVHDF -1.8 ± 5.3 0.98
Subgroups
KCALC < 60 ml min-1 0.05 ± 3.3 0.95
KCALC > 60 ml min-1 -5.2 ± 8.2a 0.89
KCALC at T0 -1.04 ± 6.3 0.97
KCALC at T18 -2.8 ± 5.2 0.96
KCALC < 60 ml min-1 at T0 0.4 ± 3.6 0.95
KCALC > 60 ml min-1 at T0 -4.8 ± 9.7a 0.87
KCALC < 60 ml min-1 at T18 -0.6 ± 2.9 0.94
KCALC > 60 ml min-1 at T18 -5.5 ± 6.5a 0.89
Delivered spKt/V 1.25 ± 0.6
Delivered/prescribed 0.89
aP < 0.05 (referred to total KCALC - KDEL difference).
Where errors are given, results are means ± SD. CVVH, continuous veno-venous hemofiltration; CVVHD, continuous veno-venous hemodialysis; CVVHDF, continuous veno-venous hemodiafiltration; KCALC, calculator-estimated urea clearance; KDEL, delivered clearance evaluated from urea concentrations on simultaneous blood and effluent samples; T0, therapy start; T18, 18 to 24 hours of uninterrupted therapy.
==== Refs
Brivet F Kleinknecht D Loriat P Landais P The French Study Group on Acute Renal Failure: acute renal failure in intensive care units – causes, outcome, and prognostic factors on hospital mortality: a prospective, multicenter study Crit Care Med 1996 24 192 198 8605788 10.1097/00003246-199602000-00003
Ronco C Bellomo R Homel P Brendolan A Dan M Piccinni P La Greca G Effect of different doses in continuous veno-venous haemofiltration on outcomes of acute renal failure: a prospective randomised trial Lancet 2000 356 26 30 10892761 10.1016/S0140-6736(00)02430-2
Gotch FA Daily hemodialysis is a complex therapy with unproven benefits Blood Purif 2001 19 211 216 11150812 10.1159/000046943
Ronco C Bellomo R Continuous renal replacement therapy: evolution in technology and current nomenclature Kidney Int 1998 66 Suppl S160 S164
Clark WR Ronco C Renal replacement therapy in acute renal failure: solute removal mechanisms and dose quantification Kidney Int 1998 66 Suppl S133 S137
Pisitkun T Tiranathanagul K Poulin S Bonello M Salvatori G D'Intini V Ricci Z Bellomo R Ronco C A practical tool for determining the adequacy of renal replacement therapy in acute renal failure patients Contrib Nephrol 2004 144 329 349 15264421
Parker T Husni L Huang W Lew N Lowrie EG Survival of hemodialysis patients in the United States is improved with a greater quantity of dialysis Am J Kidney Dis 1994 23 661 669 8172208
NKF/DOQI Clinical practice guidelines for haemodialysis adequacy: updater 2000 Am J Kidney Dis 2001 37 Suppl 1 S7 S64
Owen WF JrChertow GM Lazarus JM Lowrie EG Dose of hemodialysis and survival: differences by race and sex JAMA 1998 280 1764 1768 9842952 10.1001/jama.280.20.1764
Daugirdas JT Depner TA Gotch FA Greene T Keshaviah P Levin NW Schulman G Comparison of methods to predict equilibrated Kt/V in the HEMO pilot study Kidney Int 1997 52 1395 1405 9350665
Gotch F Sargent J A mechanistic analysis of the National Cooperative Dialysis Study (NCDS) Kidney Int 1985 28 526 534 3934452
Brause M Neumann A Schumacher T Grabensee B Heering P Effect of filtration volume of continuous venovenous hemofiltration in the treatment of patients with acute renal failure in intensive care units Crit Care Med 2003 31 841 846 12626994 10.1097/01.CCM.0000054866.45509.D0
Evanson JA Himmelfarb J Wingard R Knights S Shyr Y Schulman G Ikizler TA Hakim RM Prescribed versus delivered dialysis in acute renal failure patients Am J Kidney Dis 1998 32 731 738 9820441
Venkataraman R Kellum JA Palevsky P Dosing patterns for CRRT at a large academic medical center in the United States J Crit Care 2002 17 246 250 12501152 10.1053/jcrc.2002.36757
| 15987400 | PMC1175890 | CC BY | 2021-01-04 16:04:52 | no | Crit Care. 2005 Apr 7; 9(3):R266-R273 | utf-8 | Crit Care | 2,005 | 10.1186/cc3517 | oa_comm |
==== Front
Crit CareCritical Care1364-85351466-609XBioMed Central London cc35201598740110.1186/cc3520ResearchHigh-frequency oscillatory ventilation in children: a single-center experience of 53 cases Slee-Wijffels Fieke YAM [email protected] der Vaart Klara RM [email protected] Jos WR [email protected] Dick G [email protected]ötz Frans B [email protected] Pediatrician, Department of Pediatric Intensive Care, VU Medical Center, Amsterdam, The Netherlands2 PhD Student, Department of Pediatric Intensive Care, VU Medical Center, Amsterdam, The Netherlands3 Epidemiologist, Department of Clinical Epidemiology and Biostatistics, VU Medical Center, Amsterdam, The Netherlands4 Pediatric Intensivist, Department of Pediatric Intensive Care, VU Medical Center, Amsterdam, The Netherlands5 Pediatric Intensivist, Department of Pediatric Intensive Care, VU Medical Center, Amsterdam, The Netherlands2005 8 4 2005 9 3 R274 R279 6 2 2005 2 3 2005 4 3 2005 15 3 2005 Copyright © 2005 Slee-Wijffels 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 cited.
Introduction
The present article reports our experience with high-frequency oscillatory ventilation (HFOV) in pediatric patients who deteriorated on conventional mechanical ventilation.
Methods
The chart records of 53 consecutively HFOV-treated patients from 1 January 1998 to 1 April 2004 were retrospectively analyzed. The parameters of demographic data, cause of respiratory insufficiency, Pediatric Index of Mortality score, oxygenation index and PaCO2 were recorded and calculated at various time points before and after the start of HFOV, along with patient outcome and cause of death.
Results
The overall survival rate was 64%. We observed remarkable differences in outcome depending on the cause of respiratory insufficiency; survival was 56% in patients with diffuse alveolar disease (DAD) and was 88% in patients with small airway disease (SAD). The oxygenation index was significantly higher before and during HFOV in DAD patients than in SAD patients. The PaCO2 prior to HFOV was higher in SAD patients compared with DAD patients and returned to normal values after the initiation of HFOV.
Conclusion
HFOV rescue therapy was associated with a high survival percentage in a selected group of children. Patients with DAD primarily had oxygenation failure. Future studies are necessary to evaluate whether the outcome in this group of patients may be improved if HFOV is applied earlier in the course of disease. Patients with SAD primarily had severe hypercapnia and HFOV therapy was very effective in achieving adequate ventilation.
See related commentary
==== Body
Introduction
High-frequency oscillatory ventilation (HFOV) is, from a theoretical point of view, an ideal method of ventilation to minimize ventilator-associated lung injury. HFOV avoids high peak inspiratory pressures, thus preventing end-inspiratory overdistension, and it avoids repetitive recruitment and de-recruitment of the unstable lung alveoli, thus preventing end-expiratory collapse [1-3]. Despite these factors, HFOV is primarily used as a rescue therapy in pediatric patients with diffuse alveolar disease (DAD), and the reported survival varies between 18% and 67% [4-15].
We have used HFOV as a rescue therapy in our pediatric intensive care unit since 1995. In addition, in contrast to most other centers, we also apply HFOV as a rescue therapy in children with small airway disease (SAD). The purpose of the present article is to report our HFOV experience with 53 consecutively treated pediatric patients who deteriorated on conventional mechanical ventilation (CMV). In addition, we considered whether the cause of respiratory insufficiency had an effect on outcome.
Patients and methods
Our pediatric intensive care unit is a nine-bed combined medical and surgical intensive care unit, staffed by trained pediatric intensivists. The chart records of all HFOV-treated children between 1 January 1998 and 1 April 2004 were retrospectively analyzed. During this period a median of 356 patients (range, 326–395 patients) were admitted per year. At the time of the study, it was not institutional policy to require ethical committee approval for a retrospective review of this nature.
The following demographic data were recorded: sex, age, weight, cause of respiratory insufficiency, time on CMV prior to HFOV, and Pediatric Index of Mortality score. The oxygenation index (OI) was calculated 24, 12 and 6 hours before transition to HFOV and at 1, 6, 12, 24 and 48 hours after the institution of HFOV. The outcomes included survival at pediatric intensive care unit discharge, the total number of ventilation days (CMV and HFOV), and the change in the OI and PaCO2 before and during HFOV. The OI was defined as: 100 × mean airway pressure × (FiO2 / PaO2) [cmH2O/mmHg].
All patients with severe respiratory failure are initially managed with CMV. We use an open lung ventilation strategy that is a volume-targeted pressure-limited strategy, aimed at adequate oxygenation and ventilation with limited pressures (plateau pressures <30–35 cmH2O and tidal volumes of 8–10 ml/kg bodyweight) with, when indicated, permissive hypercapnia (pH >7.25) and optimal positive end-expiratory pressure to achieve a goal of FiO2 <0.6 with a minimum oxygen saturation of 90% (PaO2 >60 mmHg). We do not use exogenous surfactant to improve gas exchange in our pediatric intensive care unit, and prone positioning is considered occasionally. In general, we try to avoid the use of neuromuscular blockade agents except in patients with small airway disease with refractory acidosis.
The reason for converting to HFOV in these patients was persistent oxygenation failure or ventilation failure, based on one or both of the following criteria: intractable respiratory failure with an OI >13 demonstrated by two consecutive blood gas measurements over at least a 6-hour period, or a plateau pressure exceeding 30 cmH2O despite the use of permissive hypercapnia for at least 2 hours. However, this treatment was not protocolized and the decision to start HFOV was, at times, based on clinical discretion. Former prematurity with residual bronchopulmonary dysplasia or obstructive airway disease with clinical evidence of increased expiratory resistance or hyperinflation on chest X-ray were not considered a contraindication for HFOV. HFOV was performed using the SensorMedics 3100A or 3100B (Yorba Linda, CA, USA).
Depending on the lung function and chest X-ray characteristics during CMV, patients are classified either as having DAD or SAD. DAD patients primarily had oxygenation disturbances necessitating high plateau pressures and a chest X-ray with bilateral diffuse whitening, whereas SAD patients primarily had ventilation disturbances, with increased airway resistance and prolonged time constants and a chest X-ray with hyperinflation. We use different HFOV strategies depending on the underlying disease [6].
The 'high-volume' or 'open-lung' strategy for DAD
The initial continuous distending pressure (CDP) is set 4 cm above the mean airway pressure used during CMV. Our oxygenation goal is to reach an adequate PaO2 (>60 mmHg) with FiO2 <0.4. Thereafter, CDP is weaned once the patient achieves FiO2 <0.4. When hypoxemia persists with adequate circulation and with no radiographic signs of lung overinflation, CDP is increased further until the oxygenation targets are reached and is subsequently rapidly weaned. The pressure amplitude of oscillation is initially set to achieve chest wall vibration to the level of the mid-thigh. The pressure amplitude of oscillation and the frequency are sequentially adjusted to achieve a PaCO2 within the target range and to maintain a pH >7.25. In children weighing <10 kg we used a frequency of 10 Hz, in children weighing >10 kg we used a frequency of 8 Hz. The frequency is decreased with persistent respiratory acidosis despite maximization of the pressure amplitude of oscillation.
The 'open-airway' strategy for SAD
In patients with SAD we used the same initial settings as already described in the 'open-lung' strategy, but high CDP is now used to open up the small airways, allowing oscillations to move freely in and out of the alveolus. The CDP must be applied carefully; if the airways are opened up, compliant alveoli can be faced with high pressures. Every incremental change should be followed by PaCO2 determination to see at which CDP the airways are opened and the PaCO2 decreases. When the airways are open, the lowest possible CDP and pressure amplitude of oscillation are sought to minimize the risk of overdistension. Overdistension is suspected if the circulation becomes compromised and if this can be restored by lowering the CDP. The degree of lung hyperinflation on chest X-ray is not used to modify CDP.
All patients are sedated during HFOV. Patients are either weaned to continuous positive airway pressure or weaned to CMV when CDP <20 cmH2O on FiO2 <0.4 and endotracheal suctioning is well tolerated.
Statistical analysis
Baseline characteristics for survivors and nonsurvivors were compared with nonparametric Mann–Whitney tests for continuous variables and with chi-square tests or Fisher exact tests for dichotomous variables. The development over time in the OI and PaCO2 between groups of patients was analyzed with generalized estimating equations [16].
Generalized estimating equation analysis is an extended linear regression analysis taking into account the fact that the same patients are measured over time. The advantage of generalized estimating equation analysis (for instance, compared with a repeated-measures analysis of variance) is that each patient is part of the analysis, irrespective of the number of repeated measurements performed for that patient; that is, missing data and an unequal number of measurements between patients are allowed.
Time was added to the generalized estimating equation analysis as a categorical variable (i.e. represented by dummy variables) in order to estimate the development over time as accurately as possible. Five patients, after being switched from HFOV to CMV, had another HFOV run (two nonsurvivors, three survivors). This second run is not used in the analysis. The significance level for all tests was set at P <0.05. All statistical analyses were performed with STATA (version 7; Stata Corp LP, College Station, Texas, USA).
Results
During the study period 52 children were treated with HFOV after failure on CMV. One patient was excluded from the analysis because differentiated HFOV and CMV for independent lung ventilation was applied [17]. One patient underwent three HFOV runs on different occasions. Thus 51 children (53 HFOV runs) composed the final study sample.
The overall survival rate was 32/53 (64%). The demographics of the surviving and nonsurviving patients are presented in Table 1. We observed that nine patients (47%) died during HFOV rescue therapy. A remarkable difference in outcome between DAD patients and SAD patients was observed; 18 of 32 (56%) DAD patients and 15 of 17 (88%) SAD patients survived. We therefore compared the course of the OI and PaCO2 between these two groups of patients.
The DAD patients had a significantly higher OI at the time of transition than the SAD patients (Fig. 1). The observed rise in the OI in the first hour after transition to HFOV in both groups is due to the applied higher CDP when compared with the mean airway pressure during CMV. The OI was higher, but not significantly, in the nonsurvivors in the DAD group before the start of HFOV, and after the initiation of HFOV it became significantly higher (Fig. 2). The SAD patients had a higher (66.9 ± 27.9 mmHg), but not significant, PaCO2 before transition to HFOV than the DAD patients (55.2 ± 23.7 mmHg). The PaCO2 rapidly decreased after transition to HFOV (Fig. 1). The mean PaCO2 values 1 hour after the start of HFOV were 51.6 ± 15.5 mmHg in the SAD group and 55.4 ± 39.2 mmHg in the DAD group, respectively.
Discussion
The overall survival rate was 64% in patients where adequate oxygenation or ventilation could not be achieved with CMV. We observed remarkable differences in outcome depending on the cause of respiratory insufficiency, indicating that a different disease process carries a different prognosis and outcome. In patients with DAD the survival rate was 56%, and this rate was 88% in patients with SAD. The OI was significantly higher in DAD patients than in SAD patients, whereas the PaCO2 prior to HFOV was higher in SAD patients than in DAD patients.
Only one prospective study and a few retrospective observational studies report the outcome in pediatric patients treated with HFOV [4-15]. Mortality rates vary between 18% and 67%. There are several reasons to explain this difference. First, the numbers of patients included in the studies were very small, ranging from four to 35 patients, so even the death of one patient could substantially alter the mortality rate. Second, mortality rates can be affected by the underlying cause of respiratory insufficiency. Most studies use HFOV as a rescue therapy only in children showing signs of DAD. This in contrast to our study, and we observed remarkable differences in outcome depending on the cause of respiratory insufficiency. Third, it is not evidently clear in the reports from the previous studies whether all nonsurviving patients died of pulmonary causes or because of other reasons. Finally, the experience with HFOV differs between studies and hospitals, which could have had an influence on the mortality rates reported. The existence of a learning curve for new technologies, as for the use of HFOV, has been widely acknowledged in the past.
Most rescue HFOV therapies are applied in patients with DAD. It is suggested that an OI >13 may serve as an indication for HFOV rescue therapy. When reviewing previous studies, however, the actual OI at the time of transition varies widely from 10 to 45.9 [4-6]. A large survey among 14 centers including 232 pediatric patients also revealed a mean OI >27.1 before initiation of HFOV [18]. We started HFOV at a median OI of 18 in the survivors and a median OI of 28 in the nonsurvivors (Fig. 1), suggesting that we may have started HFOV rescue therapy too late. However, the OI values 6 hours before transition were comparable between survivors and nonsurvivors (Fig. 2).
Most studies have focused on the OI as a predictor of mortality after switching to HFOV. Sarnaik and colleagues proposed that those patients with an initial OI >20 who did not have a reduction of at least 20% in OI by 6 hours on HFOV can be predicted to die [8]. We think it is more important to identify early those patients who are at risk by prospectively recording the OI at small time intervals. This may serve to switch these patients to HFOV therapy before achieving OI >20 (Fig. 2). It remains uncertain whether this will result in an improved survival. It is therefore necessary to perform a large prospective multicenter trial to evaluate whether outcome in patients with DAD may be improved if HFOV is applied earlier in the course of the disease.
The use of HFOV in children with SAD is limited to a few case reports and is usually avoided because of the assumption of an associated increased risk of dynamic air trapping with this condition [19,20]. The reason for converting to HFOV in patients with SAD was primarily hypercapnia. HFOV therapy was very effective in achieving rapid adequate ventilation, resulting in an 88% survival. Our results suggest that HFOV is safe but it remains very important to apply the adequate HFOV strategy in this group of patients. HFOV is used to open up and stent the small airways ('open airway' – a concept in analogy to the 'open lung' concept) to provide adequate ventilation, which is in sharp contrast with the application of CDP to provide optimal oxygenation. The airway diameter remains stable and oscillations can move freely in and out of the alveoli, providing an adequate ventilation – particularly since expiration during HFOV is active.
In conclusion, despite the retrospective nature of this study creating several limitations, we observed that HFOV rescue therapy was associated with a high survival percentage in a selected group of children where CMV failed. Future studies are necessary to evaluate whether the outcome in patients with DAD may be improved if HFOV is applied earlier in the course of disease. HFOV rescue therapy in patients with SAD can be considered in refractory hypercapnia.
Key messages
- HFOV rescue therapy was associated with a high survival percentage (64%) in a selected group of children.
- A remarkable difference in outcome was observed depending on the cause of respiratory insufficiency, indicating that a different disease process carries a different prognosis and outcome.
- In patients with diffuse alveolar disease the survival rate was 56%, and this rate was 88% in patients with small airway disease.
- The oxygenation index prior to HFOV was significantly higher in diffuse alveolar disease patients than in small airway disease patients.
HFOV rescue therapy in patients with small airway disease can be considered in refractory hypercapnia.
Abbreviations
CDP = continuous distending pressure; CMV = conventional mechanical ventilation; DAD = diffuse alveolar disease; HFOV = high-frequency oscillatory ventilation; OI = oxygenation index; SAD = small airway disease.
Competing interests
The author(s) declare that they have no competing interests.
Authors' contributions
FYAMS-W carried out the data collection and drafted the manuscript. KRMvdV carried out the data collection and drafted the manuscript. JWRT performed the statistical analysis. DGM participated in the study design and helped to draft the manuscript. FBP conceived of the study and participated in its design and coordination, and helped to draft the manuscript. All authors read and approved the final manuscript.
Figures and Tables
Figure 1 The oxygenation index (OI) before and during high-frequency oscillatory ventilation (HFOV) in patients with diffuse alveolar disease (DAD) (●) and in patients with small airway disease (SAD) (■). The OI became significantly higher 6 hours prior to HFOV therapy and remained higher. The observed rise in the OI in the first hour after transition to HFOV in both groups is due to the applied higher CDP when compared with the mean airway pressure during conventional mechanical ventilation. The SAD patients had a higher, but not significant, PaCO2 before transition to HFOV than the DAD patients. PaCO2 returned to normal values after transition to HFOV. * P < 0.05.
Figure 2 The oxygenation index (OI) was higher in the nonsurvivors (solid line) compared with the survivors (dash line) in the diffuse alveolar disease group before the start of high-frequency oscillatory ventilation (HFOV). The OI became significant after the start of HFOV. * P < 0.05.
Table 1 Patient demographics
Parameter Survivors Nonsurvivors P value
Number of patients 34 19
Age (months)a 9.5 (0–158) 14.0 (0–169) Not significant
Weight (kg)a 6.7 (2.6–30.0) 10.0 (1.7–86.0) Not significant
Male 20 (58.8%) 7 (36.8%) Not significant
Pediatric Index of Mortality score (%)a 2.8 (0.2–54.5) 3.9 (0.6–97.7) Not significant
Cause of respiratory insufficiency
Diffuse alveolar disease 18 (52.9%) 14 (73.7%)
Acute respiratory distree syndrome 10 (29.4%) 7 (36.8%)
Pneumonia 8 (23.6%) 6 (31.6%)
Aspiration 0 1 (5.3%)
Small airway disease 15 (44.2%) 2 (10.5%)
Bronchiolitis 15 (44.1%) 2 (10.5%)
Different 1 (2.9%) 3 (15.8%)
Duration of conventional mechanical ventilation before transition (hours)a 29.5 (0–690) 63.0 (2–473) Not significant
Duration of high-frequency oscillatory ventilation (hours)a 214 (1–648) 177 (9–845) Not significant
Duration of conventional mechanical ventilation after high-frequency oscillatory ventilation (hours)a 66 (0–1218) 8 (0–427) Not significant
Number of pediatric intensive care unit daysa 23 (7–47) 22 (3–50) Not significant
aData presented as median (range).
==== Refs
Froese AB McCulloch PR Sugiura M Vaclavik S Possmayer F Moller F Optimizing alveolar expansion prolongs the effectiveness of exogenous surfactant therapy in the adult rabbit Am Rev Respir Dis 1993 148 569 577 8368625
Froese AB High-frequency oscillatory ventilation for adult respiratory distress syndrome: let's get it right this time! Crit Care Med 1997 25 906 908 9201040 10.1097/00003246-199706000-00004
Venegas JG Fredberg JJ Understanding the pressure cost of ventilation: why does high-frequency ventilation work? Crit Care Med 1994 22 S49 S57 8070270
Arnold JH Hanson JH Toro-Figuero LO Gutierrez J Berens RJ Anglin DL Prospective, randomized comparison of high-frequency oscillatory ventilation and conventional mechanical ventilation in pediatric respiratory failure Crit Care Med 1994 22 1530 1539 7924362
Anton N Joffe KM Joffe AR Inability to predict outcome of acute respiratory distress syndrome in children when using high frequency oscillation Intensive Care Med 2003 29 1763 1769 12923619 10.1007/s00134-003-1928-3
Duval EL Markhorst DG Gemke RJ van Vught AJ High-frequency oscillatory ventilation in pediatric patients Neth J Med 2000 56 177 185 10781709 10.1016/S0300-2977(00)00007-3
Tang JR Yau KI Shih HH High-frequency oscillatory ventilation for infants and children with adult respiratory distress syndrome Zhonghua Min Guo Xiao Er Ke Yi Xue Hui Za Zhi 1997 38 137 144 9151467
Sarnaik AP Meert KL Pappas MD Simpson PM Lieh-lai MW Heidemann SM Predicting outcome in children with severe acute respiratory failure treated with high-frequency ventilation Crit Care Med 1996 24 1396 1402 8706497 10.1097/00003246-199608000-00020
Fedora M Klimovic M Seda M Dominik P Nekvasil R Effect of early intervention of high-frequency oscillatory ventilation on the outcome in pediatric acute respiratory distress syndrome Bratisl Lek Listy 2000 101 8 13 10824405
Rosenberg RB Broner CW Peters KJ Anglin DL High-frequency ventilation for acute pediatric respiratory failure Chest 1993 104 1216 1221 8404196
Lochindarat S Srisan P Jatanachai P Factors effecting the outcome of acute respiratory distress syndrome in pediatric patients treated with high frequency oscillatory ventilation J Med Assoc Thai 2003 86 S618 S627 14700158
McDougall PN Loughnan PM Campbell NT Hochmann M Timms BJ Butt WW High frequency oscillation in newborn infants with respiratory failure J Paediatr Child Health 1995 31 292 296 7576885
Watkins SJ Peters MJ Tasker RC One hundred courses of high frequence oscillatory ventilation: what have we learned? Eur J Pediatr 2000 159 134 10653350 10.1007/s004310050033
MacIntyre NR High-frequency jet ventilation Respir Care Clin N Am 2001 7 599 610 11926758
Cox CE Carson SS Ely EW Govert JA Garrett JM Brower RG Morris DG Abraham E Donnabella V Spevetz A Hall JB Effectiveness of medical resident education in mechanical ventilation Am J Respir Crit Care Med 2003 167 32 38 12406827 10.1164/rccm.200206-624OC
Twisk JWR Applied Longitudinal Data Analysis for Epidemiology A Practical Guide 2003 Cambridge: Cambridge University Press
Plötz FB Hassing MB Sibarani-Ponsen RD Markhorst DG Differentiated HFO and CMV for independent lung ventilation in a pediatric patient [Letter] Intensive Care Med 2003 29 1855 14534775 10.1007/s00134-003-1949-y
Arnold JH Anas NG Luckett P Cheifetz IM Reyes G Newth CJL Kocis KC Heidemann SM Hanson JH Brogan TV Bohn DJ High-frequency oscillatory ventilation in pediatric respiratory failure: a multicenter experience Crit Care Med 2000 28 3913 3919 11153635 10.1097/00003246-200012000-00031
Duval EL Leroy PL Gemke RJ van Vught AJ High-frequency oscillatory ventilation in RSV bronchiolitis patients Respir Med 1999 93 435 440 10464828 10.1053/rmed.1999.0578
Duval EL van Vught AJ Status asthmaticus treated by high-frequency oscillatory ventilation Pediatr Pulmonol 2000 30 350 353 11015138 10.1002/1099-0496(200010)30:4<350::AID-PPUL13>3.0.CO;2-2
| 15987401 | PMC1175892 | CC BY | 2021-01-04 16:04:52 | no | Crit Care. 2005 Apr 8; 9(3):R274-R279 | utf-8 | Crit Care | 2,005 | 10.1186/cc3520 | oa_comm |
==== Front
Crit CareCritical Care1364-85351466-609XBioMed Central London cc35221598740310.1186/cc3522ResearchAdhesion of the probiotic bacterium Lactobacillus plantarum 299v onto the gut mucosa in critically ill patients: a randomised open trial Klarin Bengt [email protected] Marie-Louise [email protected] Göran [email protected] Anders [email protected] Bengt [email protected] Consultant, Assistant Professor Department of Anaesthesiology & Intensive Care, University Hospital, Lund, Sweden2 Research manager, Probi AB, Ideon, Lund, Sweden3 Professor, Laboratory of Food Hygiene, Lund University, Lund, Sweden4 Professor, Department of Anaesthesiology, Aalborg University Hospital, Aalborg, Denmark5 Professor, Department of Surgery, University Hospital, Malmö, Sweden2005 28 4 2005 9 3 R285 R293 22 11 2004 9 2 2005 3 3 2005 18 3 2005 Copyright © 2005 Klarin 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 cited.
Introduction
To achieve any possible positive effect on the intestinal mucosa cells it is important that probiotics adhere tightly onto the intestinal mucosa. It has been shown in healthy volunteers that Lactobacillus plantarum 299v (Lp 299v) (DSM 9843), a probiotic bacterium, given orally in a fermented oatmeal formula adheres onto the intestinal mucosa, but whether this also occurs in critically ill patients is unknown.
Methods
After randomisation, nine enterally fed, critically ill patients treated with broad-spectrum antibiotics received an oatmeal formula fermented with Lp 299v throughout their stay in the intensive care unit; eight patients served as controls. Biopsies of the rectal mucosa were made at admission and then twice a week, and the biopsies were analysed blindly.
Results
Four patients in the control group were colonised with Lp 299v at admission but thereafter all their biopsies were negative (Lp 299v is an ingredient in a common functional food, ProViva®, in Sweden). Of the treated patients none was colonised at admission but three patients had Lp 299v adhered on the mucosa from the second or third biopsy and in the following samples.
Conclusion
This study shows that Lp 299v could survive the passage from the stomach to the rectum and was able adhere onto the rectal mucosa also in critically ill, antibiotic-treated patients.
==== Body
Introduction
In critical illness, the intestine has been indicted as a source of pathogens sustaining the inflammatory response initiating or maintaining multiple organ failure. Various interventions have therefore been proposed to limit the growth of putatively causative pathogens in the gut; for example, selective intraluminal eradication of facultative aerobic Gram-negative bacteria – selective digestive decontamination. Indeed, selective digestive decontamination reduces the infection rate, especially in the respiratory tract [1]. Although a meta-analysis [2] and a recent study in critically ill patients [3] suggest a decreased mortality using selective digestive decontamination, there is a risk of emergence of multiresistant bacteria by the high antibiotic load.
Another method, potentially more beneficial for the microbiological environment, to reduce growth of pathogens in the gut is the administration of probiotics – lactobacilli and bifidobacteria [4]. Intestinal permeability is increased during critical illness, particularly after burns, major trauma and sepsis [5-7], and bacterial translocation has been demonstrated in patients with bowel obstruction [8,9]. The administration of probiotic Lactobacillus strains in animal experiments has been associated with reduced bacterial translocation and intestinal inflammation [10,11].
The strain Lactobacillus plantarum 299v (Lp 299v) has excellent adherence characteristics using the mannose binding sites on the mucosal cells [12]. In fact, in healthy volunteers oral administration of Lp 299v produced adherence onto and colonisation of the rectal mucosa and remained viable, verified by biopsies, for more than 11 days after end of administration [13]. The positive effects might be due to the lactobacilli fermenting nutritional carbohydrates and fibres to the preferred substrates for enterocytes – the short chain fatty acids. However, the mannose binding adhesion of Lp 299v [12] and the ability for Lp 299v to adhere to the intestinal mucosa are a possible basis for exclusion of other bacteria from adhering, thus preventing translocation. Furthermore, Lp 299v has been shown to stimulate the mucin-production in HT-29 cells [14,15]. To have beneficial effects, however, the lactobacilli should survive and adhere to the gut wall in sufficient numbers. Lp 299v is sensitive to several of the commonly used antibiotics (e.g. ampicillin, erythromycin, clindamycin, and trimethoprim/sulphamethoxaxol). In addition, the decreased gut motility often seen in critical illness might influence the transport of Lp 299v down to the lower gastrointestinal tract. Whether Lp 299v survives and adheres to the mucosa in the lower gastrointestinal tract in critically ill patients is therefore uncertain.
The primary aim of this pilot study was to examine this survival and adherence by obtaining rectal biopsies from critically ill, antibiotic-treated patients given Lp 299v enterally. The secondary aims were to evaluate the influence on the main groups of bacteria in the gut and explore the side effects of the treatment and to evaluate how the given product was tolerated when given to critically ill patients.
Materials and methods
The present study was approved by the Human Ethics Committee at Lund University and was performed in compliance with the Helsinki Declaration. Informed consent was obtained from the patient or from the next of kin. The study was performed in the general intensive care unit (ICU) (nine beds) at Lund University Hospital.
The inclusion criteria were that the patient should be 18 years or older, should be critically ill (defined by a presumed need of intensive care for 3 days or more), should tolerate enteral feeding, should have no significant coagulation disorder or thrombocytopenia, and should have an indication for broad-spectrum antibiotics.
After inclusion (which was made within 12 hours after admission), randomisation was performed with sealed envelopes. Enteral nutrition was started within 24 hours after admission to the ICU. Nine patients (treatment group) were given the test solution in addition to the enteral formula, and eight patients (controls) received the enteral formula alone (Nutrodrip Standard, Nutrodrip Fiber, or Impact; Novartis AG, Basel, Switzerland)
The test solution consisted of a fermented oatmeal formula containing 109 colony-forming units/ml Lp 299v (Probi AB, Lund, Sweden and Skånemejerier AB, Malmö, Sweden). The formula was given through a nasogastric catheter every 6 hours. The two first patients in the treatment group were given 50 ml portions throughout their study period but, due to bowel distension, the dose was adjusted in the other six patients to 50 ml test solution every 6 hours for 3 days and then 25 ml every 6 hours throughout the rest of their stay in the ICU.
All patients received prokinetic agents – metoclopramid (Primperan; Sanofi, Paris, France), cisapride (Prepulsid; Janssen-Cilag, Beerse, Belgium and sodium picosulphate (Laxoberal; Boehringer Ingelheim, Ingelheim, Germany).
Biopsies from the rectal mucosa were taken in both groups on the admission day and thereafter twice a week. The first biopsy from patients in the treatment group was taken before the administration of bacteria. Administration of enteral nutrition was started as soon as the patients' circulatory and respiratory functions had been stabilised and in all patients before 24 hours after admission. Biopsies were sent blinded for analysis to the laboratory.
Analysis of the biopsies
The pieces of tissue were washed three times in a solution (0.9% NaCl, 0.1% peptone, 0.1% Tween, and 0.02% cysteine) before dilution and inoculation. Viable counts were obtained from Rogosa agar (Oxoid; Basingstoke, Hampshire, England) incubated anaerobically at 37°C for 3 days for the enumeration of lactobacilli, from Violet Red Bile Glucose agar (Oxoid) incubated aerobically at 37°C for 24 hours for the enumeration of Enterobacteriaceae, and from perfringens agar base (Oxoid) + TSC selective supplement (Oxoid) incubated anaerobically at 37°C for 3 days (sulphite reducing clostridia). Colonies suspected to be Lp 299v on the Rogosa agar plates (large, creamy, white–yellowish and somewhat irregular) were counted. Representative colonies were picked, purified on Rogosa agar and were identified by Randomly Amplified Polymorphic DNA typing [16].
Clinical routine cultures
Specimens from blood, urine and tracheal secretion, from wounds and from other relevant locations were sent for culture weekly or when clinically indicated. Tips from central venous catheters and occasionally, on suspicion of infection, arterial lines were sent for culture at removal.
The specimens were cultured and analysed at the Department of Clinical Microbiology, Lund University Hospital, according to clinical routines.
Chemistry
Blood gases were analysed in the ICU and other routine experiments were performed at the Clinical Chemistry Laboratory, Lund University Hospital.
Statistics
The proportions of conversion of bacterial adherence to the mucosa were analysed with the chi-square test (2 × 3 table) (Statview; SAS institute Inc., Cary, NC, USA). Differences in chemistry and bacterial counts of the main groups of bacteria were analysed with the Student t test (Statistica 6.0; Statsoft, Tulsa, OK, USA). P < 0.05 was considered significant. The results are presented as the median and range unless otherwise indicated.
Results
All patients tolerated total or partial enteral feeding, and from day 2 the patients received at least 25% of the calculated daily nutritional needs via the enteral route. Supplementary nutrition was given parenterally.
Patients in the treatment group were older than the controls (median 70.9 [38–85] years versus 57.5 [34–76] years). There were no differences in the Acute Pathophysiology and Chronic Health Evaluation II score (17 [13–29] and 19 [14–36] for the treatment and control groups, respectively) in the days on a ventilator, in the median length of stay in the ICU (12 [4–37] days versus 11 [4–49] days), in hospital mortality (two patients died in each group) or in 6-month mortality (all patients discharged from the hospital survived) between the groups (Table 1).
All the patients were treated with broad-spectrum antibiotics, mainly imipenem and cefuroxime (Table 2), in consensus with a consultant physician from the Department of Infectious Diseases and according to results from previous cultures. In two patients, one from each group (patients 7 and 9), only one biopsy (before the start of the treatment) was obtained due to short stay; hence, these patients were excluded from the study. The calculations are thus based on eight patients in the treatment group and seven patients in the control group.
C-reactive protein was similar in the two groups throughout the study. The leukocyte count tended initially to be higher in the treatment group, but after day 5 the leukocyte count was lower in the treatment group (P = 0.036 on day 6). There was no difference in the other routine chemistry.
After the adjustment of the dose of the test solution the enteral solutions were well tolerated. There was no difference in the incidence of diarrhoea or gas bloating between the two groups.
Cultures of biopsies and colonisation of Lp 299v
There was no significant bleeding or other side-effects after the biopsies in any patient.
The number of analyses of biopsies in the treatment group and in the control group were two analyses in six patients (three patients and three patients, respectively), three analyses in four patients (two patients and two patients, respectively), four analyses in two patients (one patient and one patient, respectively) and five analyses in three patients (two patients and one patient, respectively). There was a difference (P = 0.029) of bacterial conversion in the biopsies between the groups. At the start of the study, four out of seven control patients were positive for Lp 299v on the first biopsy but Lp 299v was not detectable in subsequent biopsies. In the treatment group, no patient was positive at admission, but two patients converted to positive culture for Lp 299v on the second biopsy and a third patient converted from the third biopsy. The successive tests remained positive in these three patients.
All patients received two or more doses of antibiotics before inclusion and the first biopsy. Five patients had been treated with antibiotics for more than 24 hours (3 days–3 weeks) before ICU admission. The antibiotics used before and during the study and the findings of Lp 299v from the biopsies are depicted in Table 2.
The numbers of Lactobacillus increased in treated patients while there was a tendency for a reduction in the controls (P = 0.061) (samples from the second biopsies). We could not discern any statistical differences between the groups regarding Enterobacteriaceae or sulphite reducing clostridia (Fig. 1), although the mean values of Enterobacteriaceae increased in the control group and decreased in the treatment group (P = 0.27 comparing samples from the second round of samples).
From the 15 patients who completed the study, 240 cultures were performed from inclusion until 36 hours after transfer to other units. Fifty-eight (24%) of these cultures were positive (Table 3). In blood, five out of 32 cultures showed bacterial growth in the control group whereas none of 30 cultures in the treatment group had bacterial growth. In the treatment group blood cultures were taken from five out of the eight patients, and blood cultures were taken from five of seven patients in the control group. The positive cultures came from three patients. In patient 3 we found two different strains of coagulase-negative Staphylococcus. The samples were taken the same day but at different occasions. In patient 8 different enteric bacteria were found on two occasions, days apart. The fifth finding was a coagulase-negative Staphylococcus from patient 11. Findings were more equal in cultures from other sites.
The species found from the blood, the catheter tips, the tracheal secretions, and the urine results are presented in Table 4.
Discussion
This pilot study shows that Lp 299v administered to critically ill, antibiotic-treated patients can survive and colonise the gut mucosa, and that repeated administration of the bacteria is necessary to obtain this effect.
The commercial market for probiotics today is worth about €6 billion, and the European Union has invested more than €15 million in studies of probiotics, but very few results have so far emerged [17]. Probiotics have been proposed to be beneficial for the gut as well as to decrease the risk of superinfections and the development of gastrointestinal malignancies, and to have positive effects on the immune system. However, although animal experiments have shown some beneficial effects [10,11,18], very little is proven in humans. One reason for this could be that some of the proposed probiotics have no effect; even if the bacterium is 'friendly' or harmless but it does not adhere closely to the intestinal mucosa, it is probably not beneficial for the mucosal cells.
Manipulation of the gut flora by stimulating certain species, as opposed to the prevalent therapy today of suppression with antibiotics, may be a possible measure to prevent or reduce the frequency of secondary infections in severely ill patients.
Lactobacillus is an important component of the mucosa-associated flora in humans, but it is not the predominating genus on the colonic mucosa. Other genera are present at the same level or at higher levels [18-20]. Lactobacilli have been claimed to have several therapeutic functions; for example, to prevent diarrhoea, to reduce translocation and to exert immune modulation. Lp 299v is obtained from human colonic mucosa, and this particular strain possesses an excellent ability to establish itself and to adhere to the mucosa [12,13,21]. This is the first time it has been shown that a bacteria like this can be established on the gastrointestinal tract mucosa in critically ill patients.
We have previously shown that Lp 299v does adhere to the mucosa in about 40% of healthy volunteers [13]. In a study on healthy volunteers where 19 different strains of Lactobacillus were given in fermented oatmeal soup, only five strains were retrieved from any of the 13 participants either from jejunal or rectal mucosal biopsies [13]. Biopsies were taken before administration and on day 1 and day 11 after administration had ended. On day 1 post treatment, Lp 299v or Lactobacillus plantarum 299 (similar to Lp 299v and hence analysed as the pair) was found on rectal biopsies from four of the 13 volunteers and, remarkably, on biopsies from six participants on day 11 post treatment. By comparing this with our results where three out of eight treated patients turned from negative to positive on these cultures for Lp 299v, we conclude that the frequency of establishment is about the same as in healthy non-antibiotic-treated volunteers. Why all volunteers or patients did not convert to detectable levels (2 × 103/g tissue) probably has multifactorial explanations, including genetic factors and original microbiotic flora.
In the present pilot study on critically ill patients, however, antibiotics did not seem to be an important factor in preventing survival and mucosal adherence of Lp 299v when distributed enterally.
Our study was not powered to analyse gastrointestinal or systemic effects but there is a demand for such studies because probiotics are now routinely used in many ICUs without any strong scientific proof of beneficial effects. There are, however, some small studies indicating positive effects. In a study by Oláh and colleagues, 22 patients with acute pancreatitis were given Lactobacillus plantarum 299 and 23 patients were given only the oatmeal formula (with heat-inactivated bacteria) [22]. The authors found a significant decrease in episodes of sepsis and pancreatic abscesses in the treated patients.
Rayes and colleagues randomised 95 liver transplantation recipients into three groups, all feed enterally [23]. One group received standard enteral formula plus selective bowel decontamination, a second group received fibre-containing formula plus Lactobacillus plantarum 299, and the third group received the same regimen as the second group but the lactobacilli had been heat-killed. The infection rate was reduced by 35% in the group given active bacilli compared with the group given standard formula or heat-killed bacteria. On the other hand, in another study by the same research group there was no difference in the infection rate between surgical patients that received active Lactobacillus plantarum 299 and patients who received heat-killed lactobacilli [24].
In addition, two studies by McNaught and colleagues have not shown any positive effect of probiotics in patients undergoing major surgery [25,26]. It should be pointed out, however, that the amount of bacteria administered in the three latter studies was probably inadequate; the daily doses of bacteria were only 5–10% of the daily dose administered in our study. Which dose is sufficient and whether probiotics have any positive effects in critically ill patients are thus still inconclusive factors.
The increase of lactobacilli on the rectal mucosa is most probably due to the administration of relatively large numbers of the study bacteria. All other changes that occurred in the amount of bacteria were not statistically significant. It is possible, however, that this is only due to the low power of the study and does not indicate a biological fact. Mean values of Enterobacteriaceae showed dispersing values for treated patients and control patients, and this might imply that the enterally added Lactobacillus changes the gut milieu so that the growth of pathogenic bacteria is inhibited.
Interestingly, the result from other cultures showed no growth of bacteria in blood cultures from the treated patients in contrast to the control group showing 15% positive cultures. This could indicate an effect of Lp 299v on the mucosal barrier, or on the immune system, as shown in the studies on Lactobacillus plantarum 299 on pancreatitis transplant patients and liver transplant patients [23,24].
Our study has several limitations. First, only a few patients were included. We wanted to study as low a number of patients as possible, due to the inherent risks with rectal biopsies, but still wanted to be able to assess whether adherence of Lp 299v could occur in critical illness. An experienced surgeon performed the biopsies and we used very strict inclusion criteria in order to increase the safety of the procedure and to prevent harmful side-effects. Indeed, we had no complications.
Second, four patients in the control group already had growth of Lp 299v on rectal biopsies when entering the study. This is most probably due to the fact that this bacteria is commercially available as part of a probiotic fruit beverage (made from the same base as our study product) in Sweden and is widely consumed by the population. In addition, since the organism used was originally harvested from human mucosa [27], our findings might be explained by the natural occurrence of the bacteria. The bacteria, however, were not identified on the subsequent biopsies in these patients, suggesting that regular administration is necessary to maintain the adhesion onto the mucosa.
Third, the statistics used could be questioned. Nevertheless, there is no reasonable explanation for the conversion from no adherence to adherence of the Lp 299v onto the mucosa other than the enteral administration of this strain per se.
Finally, in the patients in whom we did not find any bacterial adhesion on the rectal mucosa, we cannot exclude that that the bacteria adhered onto the mucosa at other parts of the gastrointestinal tract.
Conclusion
In conclusion, this pilot study shows that enteral administration of Lp 299v is feasible in the intensive care setting. The study also shows that this bacterium can survive transport in the gastrointestinal tract and seems to colonise the gut mucosa, as assessed from rectal biopsies, in critically ill patients treated with broad-spectrum antibiotics.
Key messages
• The probiotic bacteria Lactobacillus plantarum 299v, given enterally to critically ill patients on antibiotic therapy survives the passage through the gastrointestinal tract and has the ability to colonize the rectal mucosa
• It is necessary to administer Lp 299v daily when patients are on antibiotic therapy.
• We saw no adverse effects and the study product containing oatmeal soup was well tolerated.
• Administration increases the number of lactobacilli and reduces the number of Enterobacteriaceae.
• The absence of positive cultures in the treatment group indicates that Lp 299v may have an effect on the mucosal barrier or even have a positive impact on the immune system.
Abbreviations
ICU = intensive care unit; Lp 299v = Lactobaccilus plantarum 299v.
Competing interests
BJ, GM and M-LJ are shareholders in Probi AB. Probi AB provided the study product.
Authors' contributions
BK, the primary investigator, was active in study planning, performed all beside work apart from the biopsies, handled the primary data and some of the statistical work, and prepared and finalised the manuscript together with GM, AL and BJ. M-LJ was active in the planning and practical performance of the study, and performed some of the statistical analysis. AL was involved in the study layout, performed some of the statistical analysis and was active in preparing the manuscript. GM contributed to analyses of the results from the bacterial cultures and to finalising the manuscript. BJ participated actively in the planning of the study and in the preparation of the manuscript.
Acknowledgements
Lars Hansson, MD, PhD, at the Department of Surgery, University Hospital, Lund, Sweden performed most of the biopsies, and on rare occasions (when Dr Hansson was not on duty) two other experienced consultants in the Department of Surgery assisted. The study was supported by grants from the Swedish Medical Research Council No K00-72X-11616-05C, Påhlssons Stiftelse, Malmö University Hospital, Einar och Inga Nilssons Stiftelse, and Julins Stiftelse.
Figures and Tables
Figure 1 Changes of bacterial counts from rectal biopsies (means): comparisons with the initial sample. The Enterobacteriaceae (Ent) species show a 10-fold increase in mean values in the control (-C) group while Lactobacillus (Lac) decrease 10-fold. In contrast, in the treatment group (-Lp) Lactobacillus (Lac) increase and Enterobacteriaceae decrease. Sulphite reducing clostridia (Cl) decrease in the control group. cfu, colony-forming units.
Table 1 Patient characteristics
Patient Age (years), gender Diagnosis at admission APACHE II score Length of stay in ICU (days)
Treatment group
2 38, female Pneumonia 13 14
4 63, male Gun shot wound 15 10
5 52, female Respiratory insufficiency 15 15
10 69, female Pancreatitis 17 37
12 84, male Pneumonia 24 4
14* 84, female Pneumonia 23 10
15† 72, male Respiratory insufficiency 29 20
17 77, female Sepsis 17 4
Control group
1 33, male Multi-trauma 14 5
3 57, female Pancreatitis 19 20
6 57, male Pneumonia 15 11
8† 61, male Septic arthritis 24 49
11† 60, male Retropharyngeal abscess 19 19
13 76, male Respiratory insufficiency 36 4
16 56, female Sepsis 16 7
APACHE, Acute Pathophysiology and Chronic Health Evaluation. *Died in the hospital after the intensive care unit (ICU). †Died in the ICU.
Table 2 Identification of Lactobacillus plantarum 299v (Lp 299v) from biopsies and the antibiotics used
Patient Lp 299v, first biopsy Lp 299v, later biopsies Antibiotics prior to ICU admission (≤ 12 days if not specified) Antibiotics in ICU before first biopsy Antibiotics in ICU (during biopsy period)
Treatment group
2 No Yes Erythromycin Erythromycin + imipenem Erythromycin + imipenem
4 No No Cefuroxime Cefuroxime 1 Imipenem, 2 +metronidazol
5 No No Cefadroxile, 10 days Cefadroxile 1 Cefuroxime, 2 meropenem
10 No No Cefuroxime, 3 days Imipenem 1 Imipenem, 2 +metronidazol
12 No No No antibiotics Imipenem Imipenem
14 No Yes 1 Metronidazol + cefotaxime/cefuroxime, 2 -metronidazol, 3 -cefotaxime/cefuroxime; + imipenem; 12 days in total Imipenem Imipenem
15 No Yes Ciprofloxacin + two doses metronidazol (rectally) Ceftazidime 1 Ceftazidime, 2 +metronidazol
17 No No Cefuroxime Imipenem Imipenem
Control group
1 No No Cloxacillin 1 Cloxacillin, 2 cefuroxime 1 Cefuroxime, 2 +metronidazol
3 Yes No Imipenem Imipenem 1 Imipenem, 2 +metronidazol
6 Yes No 1 Penicillin G, 2 erythromycin, 3 +netilmicin, 4 cefotaxime (-netilmicin, -erythromycin), 5 erythromycin, 6 imipenem; 3 weeks in total Imipenem Imipenem
8 Yes No Penicillin G Imipenem 1 Imipenem, 2 +clindamycin, 3 -clindamycin, +metronidazol, 4 vancomycin+ ciprofloxacin
11 No No Metronidazol and cefuroxime Metronidazol and cefuroxime 1 metronidazol + cefuroxime, 2 +isoniazid, 3 +rifampicin, 4 -(1, 2, 3), +imipenem
13 Yes No 1 PenicillinV, 2 cefuroxime; 6 days in total Cefuroxime Cefuroxime
16 No No Cefuroxime Cefuroxime 1 Cefuroxime, 2 penicillin G
Figures indicate the order in which antibiotics were been given (and changed). +, added medication; -, withdrawn medication. ICU, intensive care unit.
Table 3 Number of cultures
Type of culture Control group Treatment group Fisher's exact test
n Positive n Number of patients with positive cultures n Positive n Number of patients with positive cultures
All 122 25 5/7 118 33 6/8 NS
Blood 32 5 3/7 (3/5) 30 0 0/8 (0/5) NS
Catheter tips 22 4 3/7 (3/4) 22 4 3/8 (3/6) NS
Tracheal secretions 14 6 2/7 (2/6) 15 6 5/8 (5/6) NS
Urine 19 1 1/7 (1/7) 18 4 2/8 2(/6) NS
Figures in parentheses show the number of patients with positive cultures in relation to the number of patients from whom the respective type of culture were taken. In the treated group, five cultures were positive in the control group while no positive cultures were found in the treatment group. Due to the small numbers of patients (we performed statistics as participating patients and not as independent cultures), a significant difference was not reached (NS, not significant).
Table 4 Species found at different locations
Location Control group Lactobacillus plantarum 299v group
Blood Coagulase-negative Staphylococcus, 3 None
Enterococcus faecalis, 1
Pseudomonas aeruginosa, 1
Catheter tips Coagulase-negative Staphylococcus, 3 Coagulase-negative Staphylococcus, 3
Enterococcus faecium, 1 Morganella morgani, 1
Enterobacter cloacae, 1 Enterococcus faecalis, 2 (1 scarce)
Tracheal secretions Klebsiella pneumoniae, 3 Escherichia coli, 2
Pseudomonas aeruginosa, 1 Morganella morgani, 1
Enterococcus faecalis, 2 Pseudomonas aeruginosa, 1
Enterococcus faecium, 1
Enterobacter cloacae, 1
Candida albicans (scarce), 1
Candida kefyr, 1
Urine Pseudomonas aeruginosa, 1 Candida albicans (scarce), 2
Enterococcus faecalis, 1 Candida tropicalis (samples from one patient, same day but separated in time), 2
Main differences between the treatment and control groups are, besides no positive blood cultures, the more abundant findings of fungi. The growth of fungi in the treatment group (urine and tracheal secretions) might be due to less bacteria giving better conditions for the culturing of fungi.
==== Refs
van Nieuwenhoven CA Buskens E van Tiel FH Bonten MJ Relationship between methodological trial quality and the effects of selective digestive decontamination on pneumonia and mortality in critically ill patients JAMA 2001 286 335 340 11466100 10.1001/jama.286.3.335
Selective Decontamination of the Digestive Tract Trialists' Collaborative Group Meta-analysis of randomised controlled studies of selective decontamination of the digestive tract BMJ 1993 307 525 532 8400971
de Jonge E Schultz MJ Spanjaard L Bossuyt PMM Vroom MB Dankert J Kesecioglu J Effects of selective decontamination of digestive tract on mortality and acquisition of resistant bacteria in intensive care: a randomised controlled trial Lancet 2003 362 1011 1016 14522530 10.1016/S0140-6736(03)14409-1
Holzapfel WH Haberer P Snel J Schillinger U Huis in't Veld JH Overview of gut flora and probiotics Int J Food Microbiol 1998 41 85 101 9704859 10.1016/S0168-1605(98)00044-0
Harris CE Griffiths RD Freestone N Billington D Atherton ST Macmillan RR Intestinal permeability in the critically ill Intensive Care Med 1992 18 38 41 1578045
O'Boyle CJ MacFie J Mitchell CJ Johnstone D Sagar PM Sedman PC Microbiology of bacterial translocation in humans Gut 1998 42 29 35 9505882
Hernandez G Velasco D Waintre C Castillo L Bugedo G Maiz A Lopez F Guzman S Vargas C Gut mucosal atrophy after a short enteral fasting period in critically ill patients J Crit Care 1999 14 73 77 10382787 10.1016/S0883-9441(99)90017-5
Deitch EA Simple intestinal obstruction causes bacterial translocation in man Arch Surg 1989 124 699 701 2730322
Sedman PC Macfie J Sagar J Mitchell CJ May J Mancey-Jones B Johnstone D The prevalence of gut translocation in humans Gastroenterology 1994 107 643 649 8076751
Mao Y Nobaek S Kasravi B Adawi D Stenram U Molin G Jeppsson G The effects of Lactobacillus strains and oat fiber on methotrexate induced enterocolitis in rat Gastroenterology 1996 111 334 344 8690198
Adawi D Kasravi B Molin G Jeppsson B Effect of Lactobacillus supplementation with and without arginine on liver damage and bacterial translocation in an acute liver injury model in the rat Hepatology 1997 25 642 647 9049212 10.1002/hep.510250325
Adlerberth I Ahrné S Johansson M-L Molin G Hansson LÅ Wold AE A mannose-specific adherence mechanism in Lactobacillus plantarum conferring to the human colonic cell line HT-29 Appl Environ Microbiol 1996 62 2244 2251 8779562
Johansson M-L Molin G Jeppsson B Nobaek S Ahrné S Bengmark S Administration of different Lactobacillus strains in fermented oatmeal soup: in vivo colonization of human intestinal mucosa and effect on the indigenous flora Appl Environ Microbiol 1993 59 15 20 8439146
Mack DR Michail S Wei S McDougall L Hollingsworth MA Probiotics inhibit enteropathogenic E. Coli adherence in vitro by inducing intestinal mucin gene expression Am J Physiol 1999 276 G941 G950 10198338
Mack DR Ahrné S Hyde L Wei S Hollingsworth MA Extracellular MUC3 mucin secretion follows adherence of Lactobacillus strains to intestinal epithelial cells in vitro Gut 2003 52 827 833 12740338 10.1136/gut.52.6.827
Johansson M-L Quednau M Molin G Ahrné S Randomly Amplified Polymorphic DNA (RAPD) for rapid typing of Lactobacillus plantarum strains Lett Appl Microbiol 1995 21 155 159 7576499
Abbot A Gut reaction Nature 2004 427 284 286 14737139 10.1038/427284a
Mangell P Nejdfors P Wang M Ahrné S Weström B Thorlacius H Jeppsson B Lactobacillus plantarum 299V inhibits Escherichia coli-induced intestinal permeability Dig Dis Sci 2002 47 511 516 11911334 10.1023/A:1017947531536
Majamaa H Isolauri E Saxelin M Vesikari T Lactic acid bacteria in the treatment of acute rotavirus gastroenteritis J Pediatr Gastroenterol Nutr 1995 20 3 333 338 7608829
Malin M Suomalainen H Saxelin M Isolauri E Promotion of IgA immune response in patients with Crohn's disease by oral bacteria therapy with Lactobacillus GG Ann Nutr Metab 1996 40 137 145 8862696
Johansson M-L Nobaek S Berggren A Nyman M Björck I Ahrné S Jeppsson B Molin G Survival of Lactobacillus plantarum DSM 9843 (299V), and effect on the short–chain fatty acid content in feces after ingestion of a rose–hip drink with fermented oats Int J Food Microbiol 1998 42 29 38 9706795 10.1016/S0168-1605(98)00055-5
Oláh A Belágyi T Issekutz Á Gamal ME Bengmark S Randomized clinical trial of specific lactobacillus and fibre supplement to early enteral nutrition in patients with acute pancreatitis Br J Surg 2002 89 1103 1107 12190674 10.1046/j.1365-2168.2002.02189.x
Rayes N Seehofer D Hansen S Boucsein K Muller AR Serke S Bengmark S Neuhaus P Early enteral supply of Lactobacillus and fiber versus selective bowel decontamination: a controlled trial in liver transplant recipients Transplantation 2002 74 123 128 12134110 10.1097/00007890-200207150-00021
Rayes N Hansen S Seehofer D Muller AR Serke S Bengmark S Neuhaus P Early enteral supply of fiber and Lactobacilli versus conventional nutrition: a controlled trial in patients with major abdominal surgery Nutrition 2002 18 609 615 12093440 10.1016/S0899-9007(02)00811-0
McNaught CE Woodcock NP MacFie J Mitchell CJ A prospective randomised study of the probiotic Lactobacillus plantarum 299V on indices of gut barrier function in elective surgical patients Gut 2002 51 827 831 12427785 10.1136/gut.51.6.827
Anderson AD McNaught CE Jain PK MacFie J Randomised clinical trial of synbiotic therapy in elective surgical patients Gut 2004 53 241 245 14724157 10.1136/gut.2003.024620
Molin G Jeppsson B Ahrné S Johansson M-L Nobaek S Ståhl M Bengmark S Numerical taxonomy of Lactobacillus spp. associated with healthy and diseased mucosa of the human intestines J Appl Bacteriol 1993 74 314 323 8468264
| 15987403 | PMC1175894 | CC BY | 2021-01-04 16:04:53 | no | Crit Care. 2005 Apr 28; 9(3):R285-R293 | utf-8 | Crit Care | 2,005 | 10.1186/cc3522 | oa_comm |
==== Front
Crit CareCritical Care1364-85351466-609XBioMed Central London cc35281598740210.1186/cc3528ResearchFatality after deliberate ingestion of the pesticide rotenone: a case report Wood David Michael [email protected] Hadi [email protected] Peter [email protected] Paul Ivor [email protected] Alison Linda [email protected] Specialist Registrar in General Medicine and Clinical Pharmacology, Department of Pharmacology and Clinical Pharmacology, St George's Hospital Medical School, London, UK2 Consultant in Anaesthetics and Intensive Care Medicine, Kingston Hospital, Kingston, Surrey, UK3 Head of Clinical & Forensic Toxicology Section, Medical Toxicology Laboratory, Guy's and St. Thomas' NHS Foundation Trust, London, UK4 Consultant Clinical Toxicologist, National Poisons Information Service (London), Guy's and St. Thomas' NHS Foundation Trust, London, UK5 Director and Clinical Toxicologist, National Poisons Information Service (London), Guy's and St. Thomas' NHS Foundation Trust, London, UK2005 29 4 2005 9 3 R280 R284 11 1 2005 14 2 2005 29 3 2005 5 4 2005 Copyright © 2005 Wood 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.
Rotenone is a pesticide derived from the roots of plants from the Leguminosae family. Poisoning following deliberate ingestion of these plant roots has commonly been reported in Papua New Guinea. However, poisoning with commercially available rotenone in humans has been reported only once previously following accidental ingestion in a 3.5-year-old child. Therefore, the optimal management of rotenone poisoning is not known. After deliberate ingestion of up to 200 ml of a commercially available 0.8% rotenone solution, a 47-year-old female on regular metformin presented with a reduced level of consciousness, metabolic acidosis and respiratory compromise. Metformin was not detected in premortem blood samples obtained. Despite intensive supportive management, admission to an intensive care unit, and empirical use of N-acetylcysteine and antioxidant therapy, she did not survive. Poisoning with rotenone is uncommon but is potentially fatal because this agent inhibits the mitochondrial respiratory chain. In vitro cell studies have shown that rotenone-induced toxicity is reduced by the use of N-acetylcysteine, antioxidants and potassium channel openers. However, no animal studies have been reported that confirm these findings, and there are no previous reports of attempted use of these agents in patients with acute rotenone-induced toxicity.
==== Body
Introduction
Rotenone is a botanical pesticide derived from the roots of species of plants from the family Leguminosae. Most commercially available preparations are derived from the species Derris elliptica, Derris mallaccensis, Lonchocarpus utilis and Lonchocarpus urucu. It has pesticide activity against a wide variety of insects and arachnids encountered in both domestic and commercial horticulture on bush and vine fruits, fruit trees, shade trees, flowers, shrubs and vegetables [1]. Ingestion of naturally occurring rotenone was previously commonly reported as a method of deliberate suicide in natives of New Ireland in Papua New Guinea, who were seen to eat the roots of plants known to contain rotenone prior to their death [2].
There has been only one other reported fatality, in a 3.5-year-old girl, following ingestion of commercially available rotenone solution [3]. Because the deliberate ingestion of commercially available rotenone in humans is uncommon, the optimal management of rotenone poisoning is not currently known. We report here a case of a fatality following ingestion of commercially available rotenone that did not respond to maximal supportive care and treatment with N-acetylcysteine (NAC) and other antioxidants.
Case report
A 47-year-old woman weighing 64 kg and with known type 2 diabetes mellitus managed with metformin (500 mg three times daily) presented after she had ingested up to 200 ml from a bottle of 0.8% rotenone solution (Bio Liquid Derris Plus™; PBI Home & Garden Limited, Waltham Cross, UK). The maximum dose ingested would therefore have been 1.6 g, equating to 25 mg/kg. She was bought to the emergency department having been found collapsed and unconscious at home by her family, with a history of vomiting.
Her initial Glasgow Coma Scale (GCS) score was 3/15 and she was therefore intubated and ventilated; she had no requirement for sedative or paralyzing agents. After intubation, her blood pressure was 93/52 mmHg and heart rate 87 beats/min. Her electrocardiogram showed sinus rhythm. Baseline blood investigations showed normal renal function, but liver dysfunction with an elevated alanine transaminase of 233 IU/l. Arterial blood gases showed a severe metabolic acidosis (pH 7.09, arterial O2 tension 24 kPa, arterial CO2 tension 4.3 kPa, HCO3 10.3 mmol/l, base excess -19 and lactate 13 mmol/l). She was therefore commenced on continuous venovenous haemodialysis with lactate-free dialysate and infusion of 50 ml/hour sodium bicarbonate for management of her metabolic acidosis. Computed tomography scanning showed no intracranial abnormalities to explain her reduced GCS. She was transferred to the intensive care unit, and the National Poisons Information Service (London) was contacted for further advice on management.
A brown watery fluid (30 ml), similar to commercially available rotenone in smell and consistency, was noticed via nasogastric aspiration 6 hours after ingestion. Because previous in vitro studies have demonstrated a benefit of NAC and antioxidants in preventing rotenone toxicity in human cell lines [4-6], she was treated with intravenous NAC (standard Prescott protocol 1979) and other antioxidants (including multivitamins, 5 ml/day ketovite [orally] and 125 mg zinc sulphate three times daily [orally]) empirically. Also, 200 mg/day iron was administered intravenously; this has been shown to activate ATP-dependent potassium channels, which is protective in rotenone-induced toxicity [7].
She remained hypotensive despite fluid resuscitation, and her cardiac index on oesophageal Doppler studies was 7.7 l/min per m2 with a stroke volume of 97 ml and systemic vascular resistance index of 470, indicating the presence of a vasodilated high output state. She was therefore commenced on a noradrenaline (norepinephrine) infusion in order to maintain her blood pressure, with a maximum dose 0.35 μg/kg per hour. Repeat cardiac studies using pulse contour cardiac output 12 hours after the initial presentation revealed a worsening cardiac index (2.5 l/min per m2), with a reduced stroke volume of 30 ml and an increased systemic vascular resistance index of 2359, indicating the presence of a vasoconstricted low output state. She was therefore started on dobutamine (maximum dose 10 μg/kg per h) and weaned from the noradrenaline infusion.
After initial stabilization of her clinical state with maximal supportive care, she then started to deteriorate clinically, with signs of cardiovascular collapse and no signs of obvious neurological recovery. At 48 hours after admission she suffered an asystolic cardiac arrest, which did not respond to cardiopulmonary resuscitation. At postmortem there were signs of multiorgan failure, with pulmonary oedema and congestion of the heart, spleen and kidneys. The liver was icteric with centrilobular necrosis and general disintegration.
Results
Samples of serum were obtained at the time of admission and analyzed locally and by the Medical Toxicology Laboratory in London. There is currently no available method for the quantification of rotenone concentrations, and we were unable to identify a suitable technique for rotenone analysis. However, analysis of serum samples failed to detect the presence of other drugs, including drugs of abuse, alcohol, barbiturates, anticonvulsants, or tricyclic antidepressants. Importantly, there was no detectable metformin in the serum samples that could have accounted for the patient's lactic acidosis.
Discussion
Poisoning with the plant-derived pesticide rotenone is uncommon and potentially fatal. In the case presented here, the patient presented after ingestion of up to 200 ml of a 0.8% (1.6 g) commercially available rotenone solution (Bio Liquid Derris Plus™) with a reduced GCS score and significant metabolic acidosis. Despite meticulous supportive care in the intensive care unit and treatment with NAC, and empirical use of antioxidants (oral multivitamins, oral zinc sulphate and intravenous iron), she did not survive.
Rotenone is a pesticide derived from the roots of members of the Leguminosae family of plants. The roots of these plants were used for many years by the Chinese because of their pesticidal actions [1]. It was first extracted from these plant roots in 1895 and was patented for use as a pesticide in the UK in 1912, although the chemical structure of rotenone was not determined until 1932. Rotenone has been demonstrated to have activity against a wide variety of insects and arachnids, and against vertebrate fishes [1,8]. Commercially available rotenone is limited mainly to domestic use because it rapidly decomposes when exposed to sunlight and the duration of its biological activity is approximately 1 week after use.
Authors have previously reported that ingestion of plant roots known to contain rotenone was common among individuals attempting suicide in the New Ireland region of Papua New Guinea [2]. Although no assays to confirm rotenone ingestion were available, individuals were often seen to have been eating the plant roots before their death, or the chewed roots would be found in close proximity to a deceased individual. The deliberate or accidental ingestion of commercially available rotenone is uncommon, and there has been only one other case report of a fatality, in a 3.5-year-old girl, following accidental ingestion of rotenone [3].
The toxicity of rotenone in animal studies is variable. The 50% lethal dose (LD50; i.e. the dose required to kill 50% of the population studied) varied from 13 to 130 mg/kg in guinea pigs [9-11] and from 25 to 132 mg/kg in rats [9,10,12,13] to 1500 mg/kg in rabbits [11]. In humans the minimum LD is not known, but death occurred in a 3.5-year-old child who had ingested 40 mg/kg rotenone solution [3]. Some of the differences seen in the fatal doses of rotenone may reflect differences in the preparations that were used or ingested, in addition to species differences in toxicity [14]. Higher doses are required for water-based than for fat-based preparations, because rotenone is very poorly soluble in water. Additionally, in animals given rotenone by subcutaneous, intravenous, or intraperitoneal routes, the LD50 was much lower and this probably reflects the rapid first pass metabolism of rotenone by the liver [11].
In animal studies, classical signs following acute ingestion of rotenone include initial respiratory stimulation, followed by significant respiratory depression and respiratory arrest [1]. Death occurs in the first 30 min in roughly half of animals given intraperitoneal rotenone [10] and within 2 days in animals that ingested oral rotenone [12]. Other features in animal studies of rotenone toxicity include vomiting, incoordination, convulsions and muscular tremors. Postmortem studies in the animals that died also demonstrated pulmonary congestion [10,15] and gastrointestinal irritation [10,11]. Individuals who were known to have ingested plant roots in New Ireland were reported to suffer from profound vomiting, dilated pupils and feeble pulse before death, and autopsies in fatal cases showed acute congestive heart failure [2]. In the previous reported accidental overdose, the child suffered from vomiting, severe metabolic acidosis (pH 6.76), drowsiness, coma and respiratory depression leading to respiratory arrest [3]. Following death, postmortem studies showed anoxic damage to the brain, lungs and heart, with an associated sero-haemorrhagic pleural effusion, acute tubular necrosis and significant gastrointestinal irritation and haemorrhage. In the case reported here, the patient presented to hospital with a significant reduction in level of consciousness, associated respiratory depression and severe metabolic acidosis, and at postmortem there were signs of multiorgan failure and significant liver damage.
The toxicity of rotenone has been more widely investigated in neuroblastoma cell lines and rotenone-induced animal models of Parkinson's disease. In rats given systemic rotenone for up to 20 days, 80% exhibited systemic toxicity with autopsy findings of severe liver necrosis [16]. In models of acute toxicity, rotenone was shown to cause both dose and time dependent reductions in neuroblastoma cell line viability [4,17-19]. Mortality in rats given subcutaneous rotenone was proportional to the doses administered, with 0% mortality with 10 mg/kg increasing to 40% mortality with 15 mg/kg [20]. Rotenone is known to be a potent inhibitor of complex I of the mitochondrial respiratory chain in all cell types, by inhibiting the function of mitochondrial NADPH (nicotinamide adenine dinucleotide phosphate, reduced form) dehydrogenase activity [5,6], therefore leading to a decrease in aerobic metabolism and development of a lactic acidosis. This inhibition of the mitochondrial respiratory chain leads to an increase in the production of hydrogen peroxide, and superoxide and oxygen radical species [6,21]. It is thought that the production of these oxidant species is the mechanism by which rotenone exerts its acute toxic effects, leading to fragmentation of DNA [6] and lipid peroxidation [6], increased lactate dehydrogenase release [4] and an increase protein carbonyl concentration, which is a marker of apoptotic cell death [5].
Inhibition of NADPH dehydrogenase within cells leads to a deficiency in the conversion of oxidized NADP+ to reduced NADPH. This NADPH then is utilized by glutathione reductase to act as an electron donor to convert oxidized glutathione to reduced glutathione. Reduced glutathione acts within cells as an antioxidant, reducing cellular damage caused by oxidant molecules. In acute toxicity studies in the SH-SY5Y neuroblastoma cell line, rotenone concentrations of 5 μmol/l resulted in significant increases in oxidized glutathione concentrations and reductions in reduced glutathione concentrations [6]. Pretreatment of neuroblastoma cell lines with 100 μmol/l NAC prior to rotenone exposure resulted in decreased markers of oxidant stress, and levels of production of both hydrogen peroxide and superoxide oxygen radicals were reduced by approximately 80% compared with those cells exposed to rotenone alone [6]. Similarly, with pretreatment with NAC before rotenone exposure, there was reduction in DNA fragmentation, lactate dehydrogenase release and other markers of apoptotic cell death [4,6]. Additionally, overall neuroblastoma cell death was significantly reduced by pretreatment with 100 μmol/l and 500 μmol/l NAC prior to rotenone exposure [4,6].
As well as the glutathione hypothesis for hepatotoxicity in rotenone poisoning, in any patient who is hypotensive for a period of several hours, the possibility that this hypotension has caused 'shock liver' must be considered. Shocked liver is best avoided by maintenance of an adequate mean arterial blood pressure, but any hypotension should be controlled as aggressively and as soon as possible, and oxygenation maximized for recovery to take place. Initial oesophageal Doppler studies in this patient indicated a high cardiac index and vasodilatation. Despite subsequent administration of NAC (a hyperosmolar solution that is associated with vasodilatation and increased cardiac index [22]), the patient's cardiac index, as measured by pulse contour cardiac output, fell and she became vasoconstricted. This was a very poor prognostic sign indeed.
In addition to the previous studies using NAC to replenish glutathione, other antioxidants have been tested in an attempt to prevent rotenone-induced toxicity. In the neuroblastoma cell line SH-SY5Y, preincubation with the plant flavanoid fraxetin produced comparable reductions in rotenone-induced cell death and other markers of rotenone-induced cellular damage to NAC [4,6]. Also, other antioxidants such as α-tocopherol and coenzyme Q10 have been shown to reduce rotenone-induced cell death [5]. Potassium channel opening drugs such as iptakalin [23] and diazoxide [24] also appear to reduce the toxicity of rotenone in cell models. It is proposed that potassium channel opening drugs cause prolonged hyperpolarization of cells, therefore leading to cellular protection. However, the exact mechanism of cellular protection is unclear, because some potassium channel opening drugs, such as glibenclamide, are only partly protective in preventing rotenone-induced toxicity [23]. Other drugs shown to reduce rotenone toxicity in neuroblastoma cell models include pranipexole [25] and prostaglandin A1 [19].
However, no studies have been conducted in whole animal models of rotenone-induced toxicity to confirm whether the use of NAC, antioxidants and potassium channel opening drugs reduce the toxicity of rotenone, as shown in neuroblastoma cell line studies. Additionally, there have been no other reported cases of attempted use of NAC or other potentially beneficial antioxidants in the management of acute rotenone exposure or toxicity in humans. This patient had established toxicity on presentation (coma, hypotension and severe metabolic acidosis), and so it is difficult to predict, based on this case, whether earlier intervention with agents such as NAC would have an impact on outcome.
Conclusion
We describe here a case of a fatality following severe rotenone poisoning. The patient presented with a reduced GCS score and severe metabolic acidosis that did not respond to meticulous supportive care, treatment with NAC and empirical use of antioxidants. Other toxicological causes of the metabolic acidosis, for example metformin toxicity, were excluded. The optimal management of patients who present following rotenone ingestion is still not known, but future use of NAC and antioxidants shown in cellular models to reduce rotenone toxicity may help to improve survival, although their use is cautioned in patients with haemodynamic compromise.
Key messages
• Reports of rotenone toxicity in humans are rare and consequently the optimal management of rotenone toxicity is not known.
• In vitro cell line studies have suggested that rotenone toxicity can be revented by the use of N-acteyl cysteine, anti-oxidants and potassium channel openers. There have been no animal studies to confirm these observations.
• In this reported case, the use of N-acteyl cysteine, antioxidants and potassium channel openers did not alter the outcome, although the patient had features of established toxicity on presentation.
• Future use of NAC and antioxidants may help to improve survival in patients with rotenone toxicity.
Abbreviations
GCS = Glasgow Coma Scale; LD = lethal dose; NAC = N-acetylcysteine; NADPH = nicotinamide adenine dinucleotide phosphate, reduced form.
Authors' contributions
HA was in charge of the patient's immediate care and management. DMW, PID and ALJ were involved in providing specialist advice concerning the management of this patient. PS undertook the serum drug and toxicological analyses. DMW was responsible for drafting the first draft of the manuscript and all authors read and approved the final manuscript.
==== Refs
Ray DE Hayes WJ Jr, Laws ER Jr Pesticides derived from plants and other organisms Handbook of Pesticide Toxicology 1991 New York, NY: Academic Press 2 3
Holland EA Suicide by ingestion of derris root sp. in New Ireland Trans R Soc Trop Med Hyg 1938 32 293 294 10.1016/S0035-9203(38)90078-1
De Wilde AR Heyndrickx A Carton D A case of fatal rotenone poisoning in a child J Forensic Sci 1986 31 1492 1498 3783116
Molina-Jimenez MF Sanchez-Reus MI Benedi J Effect of fraxetin and myricetin on rotenone-induced cytotoxicity in SH-SY5Y cells: comparison with N-acetyl cysteine Eur J Pharmacol 2003 472 81 87 12860476 10.1016/S0014-2999(03)01902-2
Sherer TB Betarbet R Testa CM Seo BB Richardson JR Kim JH Miller GW Yagi T Matsuno-Yagi A Greenamyre JT Mechanism of toxicity in rotenone models of Parkinson's disease J Neurosci 2003 23 10756 10764 14645467
Molina-Jimenez MF Sanchez-Reus MI Andres D Cascales M Benedi J Neuroprotective effect of fraxetin and myricetin against rotenone-induced apoptosis in neuroblastoma cells Brain Res 2004 1009 9 16 15120578 10.1016/j.brainres.2004.02.065
Tai KK Truong DD Activation of adenosine triphosphate-sensitive potassium channels confers protection against rotenone-induced cell death: therapeutic implications for Parkinson's disease J Neurosci Res 2002 69 559 566 12210849 10.1002/jnr.10309
Gilmore RG Hastings PA Kulczycki GR Jennison BL Crystalline rotenone as a selective fish toxin Florida Scientist 1981 44 193 203
Lightbody HD Matthews JA Toxicology of rotenone Ind Eng Chem 1936 28 809 811 10.1021/ie50319a014
Shimkin MB Anderson HH Acute toxicities of rotenone and mixed pyrethrins in mammals Proc Soc Exp Biol Med 1936 34 135 138
Haag HB Toxicological studies of Derris elliptica and its constituents. I. Rotenone J Pharmacol Exp Ther 1931 43 193 208
Lehman AJ Chemicals in foods: a report of the association of food and drug officials on current developments. Part II. Pesticides. Section I. Introduction Q Bull Assoc Food Drug Off 1951 15 122 123
Santi R T'Oth CE Toxicology of rotenone Farmaco [Sci] 1965 20 270 279 14333191
Cutkomp LK Toxicity of rotenone in animals Soap Sanit Chem 1938 19 107 123
Ambrose AM Haag HB Toxicological study of derris Ind Eng Chem Ind Ed 1936 28 815 821 10.1021/ie50319a017
Lapointe N St-Hilaire M Martinoli MG Blanchet J Gould P Rouillard C Cicchetti F Rotenone induces non-specific central nervous system and systemic toxicity FASEB J 2004 18 717 719 14766796
Seaton TA Cooper JM Schapira AH Free radical scavengers protect dopaminergic cell lines from apoptosis induced by complex I inhibitors Brain Res 1997 777 110 118 9449419 10.1016/S0006-8993(97)01034-2
Kitamura Y Inden M Miyamura A Kakimura J Taniguchi T Shimohama S Possible involvement of both mitochondria- and endoplasmic reticulum-dependent caspase pathways in rotenone-induced apoptosis in human neuroblastoma SH-SY5Y cells Neurosci Lett 2002 333 25 28 12401552 10.1016/S0304-3940(02)00964-3
Wang X Qin ZH Leng Y Wang Y Jin X Chase TN Bennett MC Prostaglandin A1 inhibits rotenone-induced apoptosis in SH-SY5Y cells J Neurochem 2002 83 1094 1102 12437580 10.1046/j.1471-4159.2002.01224.x
Antkiewicz-Michaluk L Karolewicz B Romanska I Michaluk J Bojarski AJ Vetulani J 1-methyl-1,2,3,4-tetrahydroisoquinoline protects against rotenone-induced mortality and biochemical changes in rat brain Eur J Pharmacol 2003 466 263 269 12694809 10.1016/S0014-2999(03)01565-6
Sakka N Sawada H Izumi Y Kume T Katsuki H Kaneko S Shimohama S Akaike A Dopamine is involved in selectivity of dopaminergic neuronal death by rotenone NeuroReport 2003 14 2425 2428 14663204 10.1097/00001756-200312190-00027
Jones AL Mechanism of action and value of N-acetylcysteine in the treatment of early and late acetaminophen poisoning: a critical review J Toxicol Clin Toxicol 1998 36 277 285 9711192
Yang Y Liu X Ding JH Sun J Long Y Wang F Yao HH Hu G Effects of iptakalim on rotenone-induced cytotoxicity and dopamine release from PC12 cells Neurosci Lett 2004 366 53 57 15265589 10.1016/j.neulet.2004.05.009
Tai KK McCrossan ZA Abbott GW Activation of mitochondrial ATP-sensitive potassium channels increases cell viability against rotenone-induced cell death J Neurochem 2003 84 1193 1200 12603842 10.1046/j.1471-4159.2003.01625.x
Schapira AH Dopamine agonists and neuroprotection in Parkinson's disease Eur J Neurol 2002 7 14 12464116 10.1046/j.1468-1331.9.s3.9.x
Prescott LF Illingworth RN Critchley JA Proudfoot AT Intravenous N-acetyl cysteine: the treatment of choice for paracetamol poisoning Br Med J 1979 2 1097 1100 519312
| 15987402 | PMC1175899 | CC BY | 2021-01-04 16:04:52 | no | Crit Care. 2005 Apr 29; 9(3):R280-R284 | utf-8 | Crit Care | 2,005 | 10.1186/cc3528 | oa_comm |
==== Front
Crit CareCritical Care1364-85351466-609XBioMed Central London cc30211577405210.1186/cc3021ResearchIncreased blood flow prevents intramucosal acidosis in sheep endotoxemia: a controlled study Dubin Arnaldo [email protected] Gastón [email protected] Bernardo [email protected] Mario O [email protected] Juan P [email protected]án Marcelo [email protected] Vanina S Kanoore [email protected] Héctor S [email protected] Julio C [email protected] Graciela [email protected] Elisa [email protected] Medical Director, Intensive Care Unit, Sanatorio Otamendi y Miroli, Buenos Aires Argentina2 Staff Physician, Intensive Care Unit, Clinicas Bazterrica y Santa Isabel, Buenos Aires, Argentina3 Medical Director, Intensive Care Unit, Hospital Posadas, Buenos Aires, Argentina4 Research Fellow, Cátedra de Farmacología, Facultad de Ciencias Médicas, Universidad Nacional de La Plata, Argentina5 Medical Director, Renal Transplantation Unit, CRAI Sur, CUCAIBA, Argentina6 Staff Physician, Intensive Care Unit, Hospital San Martin de la Plata, Argentina7 Staff Physician, Clinical Chemistry Laboratory, Hospital San Martin de La Plata, Argentina8 Medical Director, Intensive Care Unit, Hospital San Martin de la Plata, Argentina2005 11 1 2005 9 2 R66 R73 23 9 2004 13 10 2004 21 11 2004 22 11 2004 Copyright © 2005 Dubin 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
Increased intramucosal–arterial carbon dioxide tension (PCO2) difference (ΔPCO2) is common in experimental endotoxemia. However, its meaning remains controversial because it has been ascribed to hypoperfusion of intestinal villi or to cytopathic hypoxia. Our hypothesis was that increased blood flow could prevent the increase in ΔPCO2.
Methods
In 19 anesthetized and mechanically ventilated sheep, we measured cardiac output, superior mesenteric blood flow, lactate, gases, hemoglobin and oxygen saturations in arterial, mixed venous and mesenteric venous blood, and ileal intramucosal PCO2 by saline tonometry. Intestinal oxygen transport and consumption were calculated. After basal measurements, sheep were assigned to the following groups, for 120 min: (1) sham (n = 6), (2) normal blood flow (n = 7) and (3) increased blood flow (n = 6). Escherichia coli lipopolysaccharide (5 μg/kg) was injected in the last two groups. Saline solution was used to maintain blood flood at basal levels in the sham and normal blood flow groups, or to increase it to about 50% of basal in the increased blood flow group.
Results
In the normal blood flow group, systemic and intestinal oxygen transport and consumption were preserved, but ΔPCO2 increased (basal versus 120 min endotoxemia, 7 ± 4 versus 19 ± 4 mmHg; P < 0.001) and metabolic acidosis with a high anion gap ensued (arterial pH 7.39 versus 7.35; anion gap 15 ± 3 versus 18 ± 2 mmol/l; P < 0.001 for both). Increased blood flow prevented the elevation in ΔPCO2 (5 ± 7 versus 9 ± 6 mmHg; P = not significant). However, anion-gap metabolic acidosis was deeper (7.42 versus 7.25; 16 ± 3 versus 22 ± 3 mmol/l; P < 0.001 for both).
Conclusions
In this model of endotoxemia, intramucosal acidosis was corrected by increased blood flow and so might follow tissue hypoperfusion. In contrast, anion-gap metabolic acidosis was left uncorrected and even worsened with aggressive volume expansion. These results point to different mechanisms generating both alterations.
Carbon dioxideoxygen consumptionblood flowendotoxemiametabolic acidosisSee related commentary
==== Body
Introduction
Rapid resolution of tissue hypoxia is the cornerstone of the treatment of sepsis and septic shock [1]. Patients who spontaneously develop high oxygen transport have better outcomes [2]. In experimental models of sepsis, animals with spontaneous elevation of oxygen transport present improved survival [3]. In addition, mortality from sepsis and septic shock could be reduced by early goal-directed therapy [4].
The intramucosal minus arterial carbon dioxide tension (PCO2) gradient (ΔPCO2) is considered a sensitive marker of regional gut perfusion [5] and is frequently found in human sepsis and in experimental endotoxemia. Because intramucosal acidosis can appear with normal or increased blood flow, it has been ascribed to a defect in cellular metabolism, namely cytopathic hypoxia [6]. It has also been related to decreased perfusion of villi [7]. Vasodilators might correct these microcirculatory deficits [8-10], but volume expansion or inotropic drugs have often failed to reverse intramucosal acidosis [11-14].
Our goal was to evaluate the effects of supranormal elevations of blood flow on oxygen transport and tissue oxygenation in a sheep model of endotoxemia. Our hypothesis was that increased blood flow could prevent the increase in ΔPCO2 and improve systemic metabolic acidosis.
Methods
Surgical preparation
Nineteen sheep were anesthetized with 30 mg/kg sodium pentobarbital, then intubated and mechanically ventilated (Dual Phase Control Respirator Pump Ventilator; Harvard Apparatus, South Natick, MA, USA) with a tidal volume of 15 ml/kg, a fraction of inspired oxygen (FIO2) of 0.21 and positive end-expiratory pressure adjusted to maintain O2 arterial saturation at more than 90%. The respiratory rate was set to keep the end-tidal PCO2 at 35 mmHg. Neuromuscular blockade was performed with intravenous pancuronium bromide (0.06 mg/kg). Additional pentobarbital boluses (1 mg/kg per hour) were administered as required.
Catheters were advanced through the left femoral vein to administer fluids and drugs, and through the left femoral artery to measure blood pressure and to obtain blood gases. A pulmonary artery catheter was inserted through right external jugular vein (Flow-directed thermodilution fiberoptic pulmonary artery catheter; Abbott Critical Care Systems, Mountain View, CA, USA).
An orogastric tube was inserted to allow drainage of gastric contents. A midline laparotomy and splenectomy were then performed. An electromagnetic flow probe was placed around the superior mesenteric artery to measure intestinal blood flow. A catheter was placed in the mesenteric vein through a small vein proximal to the gut to draw blood gases. A tonometer was inserted through a small ileotomy to measure intramucosal PCO2. Lastly, after careful hemostasis, the abdominal wall incision was closed.
Measurements and derived calculations
Arterial, systemic, pulmonary and central venous pressures were measured with corresponding transducers (Statham P23 AA; Statham, Halo Rey, Puerto Rico). Cardiac output was measured by thermodilution with 5 ml of saline solution (HP OmniCare Model 24 A 10; Hewlett Packard, Andover, MA, USA) at 0°C. An average of three measurements taken randomly during the respiratory cycle were considered and were normalized to body weight to yield Q. Intestinal blood flow was measured by the electromagnetic method (Spectramed Blood Flowmeter model SP 2202 B; Spectramed Inc., Oxnard, CA, USA) with in vitro calibrated transducers 5–7 mm in diameter (Blood Flowmeter Transducer; Spectramed Inc.). Occlusive zero was controlled before and after each experiment. Non-occlusive zero was corrected before each measurement. Superior mesenteric blood flow was normalized to gut weight (Qintestinal).
Arterial, mixed venous and mesenteric venous partial pressure of oxygen (PO2), PCO2 and pH were measured with a blood gas analyzer (ABL 5; Radiometer, Copenhagen, Denmark), and hemoglobin and oxygen saturation were measured with a co-oximeter calibrated for sheep blood (OSM 3; Radiometer). Arterial, mixed venous and mesenteric venous contents (CaO2, CvO2 and CvmO2, respectively) were calculated as (Hb × 1.34 × O2 saturation) + (PO2 × 0.0031). Systemic and intestinal oxygen transport and oxygen consumption (DO2, VO2, DO2i and VO2i, respectively) were calculated as DO2 = Q × CaO2; VO2 = Q × (CaO2 - CvO2); DO2i = Qintestinal × CaO2, and VO2i = Qintestinal × (CaO2 - CvmO2).
Intramucosal PCO2 was measured with a tonometer [15] (TRIP Sigmoid Catheter; Tonometrics, Inc., Worcester, MA, USA) filled with 2.5 ml of saline solution; 1.0 ml was discarded after an equilibration period of 30 min and PCO2 was measured in the remaining 1.5 ml. Its value was corrected to the corresponding equilibration period and was used to calculate ΔPCO2.
Mixed venous–arterial and mesenteric venous–arterial PCO2 differences were also calculated. Arterial, mixed venous and mesenteric venous CO2 contents (CCO2) and their differences were calculated with Douglas's algorithm [16]. Systemic and intestinal CO2 production (VCO2 and VCO2i, respectively) were calculated as VCO2 = Q × mixed venoarterial CCO2, and VCO2i = Qintestinal × mesenteric venoarterial CCO2. Global blood capacity for transporting CO2 was evaluated as the ratio between venoarterial CCO2 and PCO2 differences (Ra-v). This index has been used to evaluate the amount of CO2 transported by the blood in relation to the venoarterial gradient of PCO2 [17].
Lactate, sodium, potassium, chloride and serum total proteins were measured with an automatic analyzer every 60 min (Automatic Analyzer Hitachi 912; Boehringer Mannheim Corporation, Indianapolis, IN, USA). Anion gap was calculated as ([Na+] + [K+]) - ([Cl-] + [HCO3-]). Anion gap was corrected for changes in plasma protein concentration [18].
Experimental procedure
Basal measurements were taken after a stabilization period longer than 30 min. Then animals were assigned to the following groups: (1) sham group (n = 6), consisting of sheep receiving 100 ml of saline in 10 min, followed by an infusion necessary to keep intestinal blood flow at basal levels; (2) normal blood flow group (n = 7), consisting of sheep receiving 5 μg/kg Escherichia coli lipopolysaccharide dissolved in 100 ml of saline in 10 min, and then saline infusion so as to maintain intestinal blood flow at basal levels; and (3) increased blood flow group (n = 6), consisting of sheep receiving 5 μg/kg Escherichia coli lipopolysaccharide dissolved in 100 ml of saline in 10 min, followed by saline infusion so as to increase intestinal blood flow by 50% from basal levels.
FIO2 was increased to 0.50 in endotoxemic sheep to avoid deep hypoxemia.
Measurements were performed at 30 min intervals for 120 min from the start of endotoxin administration.
At the end of the experiment, the animals were killed with an additional dose of pentobarbital and a KCl bolus. A catheter was inserted in the superior mesenteric artery and Indian ink was instilled through it. Dyed intestinal segments were dissected, washed and weighed for the calculation of gut indexes.
The local Animal Care Committee approved the study. Care of animals was in accordance with National Institute of Health guidelines.
Statistical analysis
Data were assessed for normality and expressed as means ± SD. Differences within groups were analyzed with a repeated-measures analysis of variance and Dunnett's multiple comparisons test to compare each time point with basal. One-time comparisons between groups were tested with a one-way analysis of variance and a Newman–Keuls multiple comparison test.
Results
Hemodynamic and oxygen transport effects
Sham, normal blood flow and increased blood flow groups received 10 ± 6, 24 ± 9 and 91 ± 38 ml/kg per hour, respectively, of normal saline solution (P < 0.05) to achieve resuscitation goals. Variations of intestinal blood flow from basal values, at the end of the experiment, were 8 ± 5%, – 1 ± 22% and 60 ± 22%, respectively (P < 0.05). As expected, the increased blood flow group had higher central venous and pulmonary wedge pressures, intestinal blood flow, cardiac output and systemic oxygen transport than the normal blood flow group. The increased blood flow group had also higher intestinal oxygen consumption (Table 1).
Metabolic effects
Metabolic acidosis developed in both groups with endotoxemia, but was greater in the increased blood flow group because of hyperchloremia and an increased anion gap (Table 2 and Fig. 1). These variables did not change in the sham group. Lactate levels remained stable in the three groups (Table 2).
Effects on ΔPCO2 and its determinants
ΔPCO2 increased in the normal blood flow group and remained unchanged in the increased blood flow and sham groups (Fig. 2). Systemic and intestinal venoarterial PCO2 differences were also higher in the normal blood flow group than in the others (Table 3). Systemic and intestinal Ra-v were lower in both endotoxemic groups.
Discussion
The main finding of this study was that increased blood flow prevented the development of intramucosal acidosis. However, anion-gap metabolic acidosis was larger in hyperresuscitated animals. These results underscore the different underlying mechanisms of each type of acidosis.
The experimental model of endotoxemia
We used a short-term infusion of endotoxin followed by saline expansion to induce a state of normodynamic shock, with preserved cardiac output and intestinal blood flow [19,20]. A state of normodynamic shock was chosen as a control group to avoid CO2 accumulation caused by macrovascular hypoperfusion. We found that intramucosal acidosis and systemic metabolic acidosis occurred, in spite of stable systemic and gut oxygen transports and consumptions.
The reason for increased intestinal ΔPCO2 in sepsis remains controversial [21]. It might reflect hypoperfusion, but has also been found in normodynamic states [22]. Vallet and colleagues studied endotoxemic dogs with low blood flow, resuscitated with dextran. Gut flow was increased and oxygen transport normalized, but oxygen uptake and mucosal PO2 and pH remained low, results that were ascribed to flow redistribution from mucosal to serosal layers [13]. Conversely, Revelly and colleagues described flow redistribution from serosa to mucosa induced by endotoxin [23]. VanderMeer and colleagues found that intramucosal acidosis developed despite preserved blood flow and tissue PO2 in endotoxemic pigs, attributed to changes in energetic metabolism [24]. Thus, the concept of 'cytopathic hypoxia' was introduced [6].
However, cytopathic hypoxia and increased anaerobic CO2 production might not be the sole explanation for the increase in ΔPCO2. Vallet and colleagues [25] and Dubin and colleagues [26] recently showed that hypoperfusion is a key factor in the development of venous and tissue hypercarbia. In addition, Tugtekin and colleagues showed an association between increased ΔPCO2 and diminished villi microcirculation [7].
This body of information suggests that intramucosal acidosis in sepsis is due mainly to microcirculatory alterations, even though cardiac output and regional flows might remain unchanged. Disturbed energetic metabolism might be present in sepsis, but it does not explain intramucosal acidosis. However, it might be a reasonable explanation for the development of systemic metabolic acidosis in our experiments. Increased anion-gap metabolic acidosis appeared despite preserved oxygen metabolism. As described previously, metabolic acidosis was not explained by elevations of lactate but by increases in unmeasured anions whose source and identification are still unknown [27,28].
Effects of saline solution expansion on intramucosal acidosis
Increased blood flow by volume expansion prevented ΔPCO2 elevation. PCO2 gradients, venoarterial and tissue-arterial PCO2 differences are the result of interactions between CO2 production, blood capacity to transport CO2 and blood flow to tissues. We and others have previously shown that ΔPCO2 fails to reflect tissue hypoxia when blood flow is preserved [25,26,29]. Our results suggest that intramucosal acidosis is related mainly to local hypoperfusion, because the only difference between our groups, in terms of PCO2 difference determinants, was the level of blood flow. We can speculate that volume expansion might improve microcirculation and, subsequently, CO2 clearance. However, intramucosal acidosis might be corrected by the inhibition of inducible nitric oxide synthase and without microcirculatory recruitment [30]. Improvement of cellular metabolism and/or redistribution of blood flow from the mucosa to other layers have been proposed as underlying mechanisms. We cannot exclude the possibility that increases in blood flow might decrease tissue hypoxia and anaerobically generated CO2. Intestinal VO2 increased after elevation of O2 transport in the increased blood flow group, suggesting unmet needs in the normal blood flow group. Flow might have been inadequate in the face of increased metabolic requirements caused by endotoxemia [31].
Despite this apparent dependence on intestinal oxygen supply, CO2 production remained stable. Possible reasons are error propagation in the VO2 and VCO2 calculations, or an increase in VO2 due to non-metabolic processes, such as the production of inflammatory reactants and reactive oxygen species [32].
Other investigators have reported that volume expansion could not correct intramucosal acidosis, in both clinical and experimental settings [11,13,14]. Differences in the level of attained blood flow, timing of expansion or the type of injury might account for these findings opposite to ours.
Potential limitations of our study are related to the errors of saline tonometry, such as inadequate equilibration time, deadspace effect and underestimation of PCO2 by blood gas analyzers [33,34].
Effects of saline solution expansion on metabolic acidosis
Metabolic acidosis was a prominent finding in our study. Expansion with large volumes of saline predictably produced hyperchloremic metabolic acidosis [35]. In addition, metabolic acidosis arose as a result of unmeasured anions. Previous research has shown that during streptococcal infusion in pigs, metabolic acidosis decreased, but did not disappear, when oxygen transport was supported with dextran and red blood cells [36].
The reason for augmented unmeasured anions in the increased blood flow group is unclear. Possible causes are washout of tissue acids by high blood flow, or an impairment of oxygenation caused by tissue edema. Nevertheless, Gow and colleagues have shown that oxygen extraction is already altered in septic animals, so increased diffusion distances would not be relevant [37].
In addition, hyperchloremic acidosis might induce an inflammatory response, cellular dysfunction and apoptosis, and increased mortality in experimental septic shock [38-41]. In this way, a deleterious effect of acidosis on cellular function with the subsequent production of unknown anions might be operative.
Conclusions
Despite preserved blood flow and oxygen transport, intramucosal acidosis developed in endotoxemic sheep. Volume expansion prevented the increase in ΔPCO2, implying that intramucosal acidosis is related mainly to local hypoperfusion. Despite aggressive expansion, anion-gap metabolic acidosis worsened, which suggests an effect on cellular metabolism.
Key messages
• Intramucosal acidosis developed in endotoxemic sheep, despite preserved blood flow and oxygen transport.
• Increased blood flow prevented elevation in ΔPCO2, suggesting that intramucosal acidosis is mainly related to local hypoperfusion. However, anion-gap metabolic acidosis was higher, pointing to a possible effect on cellular mechanism.
Abbreviations
CaO2 = arterial oxygen content; CCO2 = CO2 content; CvmO2 = mesenteric venous oxygen content; CvO2 = mixed venous oxygen content; DO2 = systemic oxygen transport; DO2i = intestinal oxygen transport; ΔPCO2 = intramucosal minus arterial PCO2 gradient; FIO2 = fraction of inspired oxygen; PCO2 = carbon dioxide tension; PO2 = partial pressure of oxygen; Q = cardiac output; Qintestinal = intestinal blood flow; Ra-v = global blood capacity for transporting CO2; VCO2 = systemic CO2 production; VCO2i = intestinal CO2 production; VO2 = systemic oxygen consumption; VO2i = intestinal oxygen consumption.
Competing interests
The author(s) declare that they have no competing interests.
Authors' contributions
AD was responsible for the study concept and design, the analysis and interpretation of data, and drafting of the manuscript. GM, MOP, VSKE and HSC performed the acquisition of data and contributed to the draft of the manuscript. BM and GE conducted the blood determinations and contributed to the draft of the manuscript. MB and JPS performed the surgical preparation and contributed to the discussion. EE helped in the draft of the manuscript and made a critical revision for important intellectual content. All authors read and approved the final manuscript.
Figures and Tables
Figure 1 Behavior of the anion gap in the sham, normal and increased blood flow groups. A higher degree of anion-gap metabolic acidosis developed in the increased blood flow group than in the normal blood flow group. The anion gap was unchanged in the sham group. 60' and 120' refer to 60 and 120 min, respectively.
Figure 2 Behavior of intramucosal – arterial PCO2 difference in the sham, normal and increased blood flow groups. Intramucosal acidosis developed in the normal blood flow group and was prevented in the increased blood flow group. Intramucosal – arterial PCO2 difference was unchanged in the sham group. 30', 60', 90' and 120' refer to 30, 60, 90 and 120 min, respectively.
Table 1 Systemic and intestinal hemodynamic and oxygen transport parameters in sham, normal and increased blood flow groups
Parameter Group Basal Endotoxemia
30 min 60 min 90 min 120 min
Mean arterial pressure (mmHg) Sham 81 ± 10 85 ± 15 88 ± 15 91 ± 16 92 ± 19
Normal 93 ± 19 89 ± 25 83 ± 23 91 ± 32 94 ± 26
Increased 90 ± 17 98 ± 17 89 ± 18 89 ± 21 99 ± 17
Mean pulmonary arterial pressure (mmHg) Sham 16 ± 3 15 ± 3 16 ± 3 15 ± 4 16 ± 4
Normal 15 ± 5 34 ± 9*† 26 ± 8*† 25 ± 7*† 24 ± 6*†
Increased 20 ± 4 35 ± 10*† 31 ± 4*† 34 ± 6*†‡ 35 ± 6*†‡
Pulmonary wedge pressure (mmHg) Sham 5 ± 2 5 ± 2 5 ± 1 5 ± 2 5 ± 2
Normal 5 ± 2 11 ± 4*† 8 ± 2*† 8 ± 3*† 8 ± 4
Increased 6 ± 1 11 ± 4*† 13 ± 6*† 12 ± 3*† 14 ± 5*†‡
Central venous pressure (mmHg) Sham 5 ± 5 5 ± 3 6 ± 5 5 ± 4 5 ± 4
Normal 4 ± 2 5 ± 3 6 ± 2 6 ± 2 5 ± 3
Increased 4 ± 2 8 ± 3 9 ± 5* 10 ± 4*†‡ 11 ± 4*†‡
Cardiac output (ml/kg per min) Sham 134 ± 30 148 ± 36 153 ± 37 144 ± 33 151 ± 41
Normal 139 ± 43 117 ± 27 135 ± 38 149 ± 42 142 ± 34
Increased 157 ± 51 221 ± 64*†‡ 257 ± 67*†‡ 276 ± 84*†‡ 290 ± 91*†‡
Superior mesenteric artery blood flow (ml/min per g) Sham 498 ± 107 568 ± 126* 551 ± 126* 548 ± 134* 539 ± 131*
Normal 553 ± 184 514 ± 152 566 ± 161 573 ± 145 529 ± 169
Increased 578 ± 206 803 ± 226*‡ 794 ± 209*†‡ 863 ± 326*‡ 923 ± 370*†‡
Increased 362 ± 116 437 ± 75†‡ 286 ± 53 336 ± 102 295 ± 75
Systemic oxygen transport (ml/min per kg) Sham 16.2 ± 4.5 18.0 ± 5.6* 19.0 ± 6.2* 17.8 ± 5.3 18.8 ± 6.1*
Normal 16.4 ± 6.6 13.3 ± 4.9 14.0 ± 4.8 16.4 ± 6.4 15.8 ± 5.7
Increased 17.2 ± 4.0 23.0 ± 5.5*‡ 25.5 ± 6.7*‡ 26.0 ± 8.4*‡ 26.9 ± 9.9*‡
Systemic oxygen consumption (ml/min per kg) Sham 6.4 ± 0.8 6.4 ± 1.1 6.8 ± 1.3 6.6 ± 1.2 7.2 ± 1.3
Normal 6.4 ± 1.2 5.3 ± 1.2* 5.8 ± 1.6* 6.0 ± 1.5 6.5 ± 1.4
Increased 7.6 ± 0.9 7.6 ± 2.0‡ 7.3 ± 2.1 7.4 ± 2.2 8.3 ± 3.2
Intestinal oxygen transport (ml/min per kg) Sham 62.3 ± 22.2 71.4 ± 24.8* 70.8 ± 25.1* 69.9 ± 24.6* 69.1 ± 24.0*
Normal 64.0 ± 22.6 56.1 ± 19.3 57.0 ± 15.8 60.8 ± 18.4 56.5 ± 17.0
Increased 64.3 ± 16.7 86.4 ± 19.1*‡ 81.4 ± 22.1* 82.2 ± 23.5* 87.1 ± 23.6*‡
Intestinal oxygen consumption (ml/min per kg) Sham 21.7 ± 4.0 21.1 ± 3.7 22.0 ± 3.2 22.7 ± 4.2 21.8 ± 4.7
Normal 21.2 ± 4.1 22.1 ± 6.5 22.7 ± 8.9 22.6 ± 7.8 22.4 ± 9.0
Increased 29.3 ± 9.7 28.9 ± 9.3 32.5 ± 13.0 29.8 ± 9.4 37.2 ± 12.3†‡
* P < 0.05 versus basal. † P < 0.05 versus sham. ‡ P < 0.05 versus normal. Sham, sham group; normal, normal blood flow group; increased, increased blood flow group.
Table 2 Arterial hemoglobin, acid-base and metabolic parameters in sham, normal and increased blood flow groups
Parameter Group Basal Endotoxemia
30 min 60 min 90 min 120 min
Hemoglobin (g/l) Sham 9.6 ± 2.4 9.7 ± 2.7 9.9 ± 2.3 9.8 ± 2.2 9.9 ± 2.2
Normal 9.1 ± 2.3 9.0 ± 2.4 8.4 ± 2.0* 8.1 ± 2.2* 8.3 ± 2.4*
Increased 8.9 ± 2.2 8.2 ± 2.3* 7.8 ± 2.4* 7.6 ± 2.5* 7.7 ± 2.5*
pH Sham 7.44 ± 0.03 7.45 ± 0.02 7.45 ± 0.03 7.47 ± 0.02 7.47 ± 0.03
Normal 7.39 ± 0.07 7.34 ± 0.08*† 7.31 ± 0.05*† 7.34 ± 0.05*† 7.35 ± 0.06*†
Increased 7.42 ± 0.04 7.35 ± 0.05*† 7.31 ± 0.05*† 7.28 ± 0.08*† 7.25 ± 0.08*†‡
PCO2 (mmHg) Sham 35 ± 3 34 ± 3 34 ± 3 33 ± 3 34 ± 4
Normal 35 ± 4 38 ± 6* 41 ± 7* 37 ± 6 35 ± 6
Increased 34 ± 2 36 ± 5 34 ± 3 34 ± 5 37 ± 6
PO2 (mmHg) Sham 85 ± 13 88 ± 18 86 ± 16 88 ± 17 84 ± 15
Normal 87 ± 16 119 ± 59 105 ± 39 123 ± 20*† 134 ± 43*†
Increased 90 ± 23 150 ± 48*† 132 ± 21*† 101 ± 20 99 ± 31
[HCO3-] (mmol/l) Sham 24 ± 2 24 ± 3 24 ± 3 24 ± 3 24 ± 3
Normal 21 ± 2 21 ± 2 20 ± 2† 20 ± 2*† 19 ± 2*†
Increased 22 ± 3 20 ± 2*† 17 ± 3*† 16 ± 3*†‡ 16 ± 2*†‡
Base excess (mmol/l) Sham 1 ± 3 1 ± 3 1 ± 3 2 ± 3 2 ± 3
Normal t2 ± 4 t5 ± 3*† t5 ± 2*† t5 ± 3*† t5 ± 3*†
Increased t1 ± 4 t4 ± 3*† t8 ± 4*† t10 ± 4*† t10 ± 3*†‡
[Cl-]/[Na+] Sham 0.76 ± 0.02 0.76 ± 0.03 0.76 ± 0.03
Normal 0.76 ± 0.01 0.77 ± 0.02 0.77 ± 0.01
Increased 0.76 ± 0.02 0.78 ± 0.02* 0.80 ± 0.02*†‡
Lactate (mmol/l) Sham 2.1 ± 0.7 2.0 ± 0.7 1.8 ± 0.6
Normal 1.7 ± 0.8 1.9 ± 0.7 2.2 ± 1.1
Increased 2.2 ± 1.6 1.7 ± 1.1 1.9 ± 1.1
* P < 0.05 versus basal. † P < 0.05 versus sham. ‡ P < 0.05 versus normal. Sham, sham group; normal, normal blood flow group; increased, increased blood flow group.
Table 3 Systemic and intestinal CO2-derived parameters in sham, normal and increased blood flow groups
Parameter Group Basal Endotoxemia
30 min 60 min 90 min 120 min
Mixed venous – arterial PCO2 (mmHg) Sham 6 ± 2 6 ± 2 6 ± 2 6 ± 2 5 ± 2
Normal 7 ± 2 8 ± 2 7 ± 2 8 ± 3 8 ± 3†
Increased 6 ± 2 6 ± 3 7 ± 5 7 ± 4 4 ± 1‡
Mesenteric venous – arterial PCO2 (mmHg) Sham 6 ± 2 5 ± 2 5 ± 2 6 ± 2 5 ± 2
Normal 7 ± 2 8 ± 2 8 ± 3 10 ± 4 10 ± 2*†
Increased 8 ± 3 6 ± 2 8 ± 4 8 ± 3 6 ± 1*‡
Intramucosal – arterial PCO2 (mmHg) Sham 4 ± 4 5 ± 8 5 ± 8 5 ± 8 6 ± 9
Normal 7 ± 4 6 ± 5 12 ± 5 15 ± 6*‡ 19 ± 4*‡
Increased 5 ± 7 2 ± 9 7 ± 7 12 ± 8 9 ± 6†
Systemic VCO2 (ml/min per kg) Sham 5.2 ± 1.9 4.5 ± 1.2 4.0 ± 1.5 4.7 ± 1.2 4.6 ± 1.8
Normal 6.0 ± 2.4 4.9 ± 1.4 4.9 ± 1.7 5.0 ± 1.3 5.0 ± 1.7
Increased 6.5 ± 2.5 4.8 ± 2.4 6.1 ± 2.8 5.8 ± 2.3 5.8 ± 4.7
Intestinal VCO2 (ml/min per kg) Sham 36.7 ± 10.9 38.1 ± 11.3 34.0 ± 8.8 43.2 ± 10.6 36.7 ± 5.6
Normal 37.7 ± 10.9 35.3 ± 11.6 37.2 ± 13.7 41.8 ± 20.3 36.7 ± 16.2
Increased 36.5 ± 21.8 35.3 ± 14.6 27.4 ± 9.4 35.8 ± 12.9 34.0 ± 7.4
Mixed venous blood capacity for transporting CO2 (ml/100 ml per mmHg) Sham 0.67 ± 0.12 0.59 ± 0.40 0.51 ± 0.11 0.61 ± 0.21 0.61 ± 0.13
Normal 0.62 ± 0.12 0.49 ± 0.12* 0.55 ± 0.04* 0.47 ± 0.09* 0.44 ± 0.09*†
Increased 0.67 ± 0.24 0.38 ± 0.27* 0.42 ± 0.24* 0.45 ± 0.19* 0.48 ± 0.12*†
Mesenteric venous blood capacity for transporting CO2 (ml/100 ml per mmHg) Sham 1.14 ± 0.24 1.15 ± 0.32 1.22 ± 0.29 1.37 ± 0.22 1.28 ± 0.08
Normal 1.04 ± 0.22 0.99 ± 0.38 0.86 ± 0.24† 0.78 ± 0.33*† 0.76 ± 0.24*†
Increased 1.17 ± 0.45 0.85 ± 0.29 0.66 ± 0.27† 0.81 ± 0.19† 0.69 ± 0.18*†
* P < 0.05 versus basal. † P < 0.05 versus sham. ‡ P < 0.05 versus normal. Sham, sham group; normal, normal blood flow group; increased, increased blood flow group.
==== Refs
Natanson C Hoffman WD Suffredini AF Eichacker PQ Danner RL Selected treatment strategies for septic shock based on proposed mechanisms of pathogenesis Ann Intern Med 1994 120 771 783 8147551
Shoemaker WC Montgomery ES Kaplan E Elwyn DH Physiologic patterns in surviving and nonsurviving shock patients Arch Surg 1973 106 630 636 4701410
Pittet JF Pastor CM Morel DR Spontaneous high systemic oxygen delivery increases survival rate in awake sheep during sustained endotoxemia Crit Care Med 2000 28 496 503 10708190 10.1097/00003246-200002000-00035
Rivers E Nguyen B Hvastad S Ressler J Muzzin A Knoblich B Peterson E Tomlanovich M for the Early Goal-directed Therapy Collaborative Group Early goal-directed therapy in the treatment of severe sepsis and septic shock N Engl J Med 2001 345 1368 1377 11794169 10.1056/NEJMoa010307
Dubin A Estenssoro E Murias G Canales H Sottile P Badie J Barán M Pálizas F Laporte M Rivas Díaz M Effects of hemorrhage on gastrointestinal oxygenation Intensive Care Med 2001 27 1931 1936 11797030 10.1007/s00134-001-1138-9
Fink M Mitochondrial dysfunction as mechanism contributing to organ dysfunction in sepsis Crit Care Clin 2001 17 219 237 11219231
Tugtekin IF Radermacher P Theisen M Matejovic M Stehr A Ploner F Matura K Ince C Georgieff M Trager K Increased ileal-mucosal-arterial PCO2 gap is associated with impaired villus microcirculation in endotoxic pigs Intensive Care Med 2001 27 757 766 11398705 10.1007/s001340100871
De Backer D Creteur J Preiser JC Dubois MC Vincent JL Microvascular blood flow is altered in patients with sepsis Am J Respir Crit Care Med 2002 166 98 104 12091178 10.1164/rccm.200109-016OC
Spronk PE Ince C Gardien MJ Mathura KR Oudemans-van Straaten HM Zandstra DF Nitroglicerin in septic shock after intravascular volume resuscitation Lancet 2002 360 1395 1396 12423989 10.1016/S0140-6736(02)11393-6
Spronk PE Zandstra DF Ince C Sepsis is a disease of the microcirculation Crit Care 2004 8 462 468 15566617 10.1186/cc2894
Forrest DM Baigorri F Chittock DR Spinelli JJ Russell JA Volume expansion using pentastarch does not change gastric-arterial CO2 gradient or gastric intramucosal pH in patients who have sepsis syndrome Crit Care Med 2000 28 2254 2258 10921549 10.1097/00003246-200007000-00012
Mark P Mohedin M The contrasting effects of dopamine and norepinephrine on systemic and splanchnic oxygen utilization in hyperdynamic sepsis JAMA 1994 272 1354 1357 7933396 10.1001/jama.272.17.1354
Vallet B Lund N Curtis SE Kelly D Cain SM Gut and muscle tissue PO2 in endotoxemic dogs during shock and resuscitation J Appl Physiol 1994 76 793 800 8175591
Lagoa CE de Figueiredo LFP Cruz RJ Silva E Rocha e Silva M Effects of volume resuscitation on splanchnic perfusion in canine model of severe sepsis induced by live Escherichia coli infusion Crit Care 2004 8 R221 R228 15312221 10.1186/cc2871
Taylor DE Gutierrez G Tonometry. A review of clinical studies Crit Care Clin 1996 12 1007 1018 8902381
Douglas AR Jones LN Reed JW Calculation of whole blood CO2 content J Appl Physiol 1988 65 473 477 3136136
Cavaliere F Antonelli M Arcangeli A Conti G Pennisi MA Proietti R Effects of acid-base abnormalities on blood capacity of transporting CO2: adverse effect of metabolic acidosis Intensive Care Med 2002 28 609 615 12029410 10.1007/s00134-002-1259-9
Constable PD Total weak acid concentration and effective dissociation constant of nonvolatile buffers in human plasma J Appl Physiol 2001 91 1364 1371 11509537
Fink MP Heard SO Laboratory models of sepsis and septic shock J Surg Res 1990 49 186 196 2199735 10.1016/0022-4804(90)90260-9
Traber DL Flynn JT Herndon DN Redl H Schlag G Traber LD Comparison of cardiopulmonary responses to single bolus and continuous infusion of endotoxin in an ovine model Circ Shock 1989 27 123 138 2650914
Vallet B Gut oxygenation in sepsis: still a matter of controversy? Crit Care 2002 6 282 283 12225596 10.1186/cc1508
Antonsson JB Engstrom L Rasmussen I Wollert S Haglund UH Changes in gut intramucosal pH and gut oxygen extraction ratio in a porcine model of peritonitis and hemorrhage Crit Care Med 1995 23 1872 1881 7587264 10.1097/00003246-199511000-00014
Revelly JP Ayuse T Brienza N Fessler HE Robotham JL Endotoxic shock alters distribution of blood flow within the intestinal wall Crit Care Med 1996 24 1345 1351 8706490 10.1097/00003246-199608000-00013
VanderMeer TJ Wang H Fink MP Endotoxemia causes ileal mucosal acidosis in the absence of mucosal hypoxia in a normodynamic porcine model of septic shock Crit Care Med 1995 23 1217 1226 7600830 10.1097/00003246-199507000-00011
Vallet B Teboul JL Cain S Curtis S Venoarterial CO2 difference during regional ischemic or hypoxic hypoxia J Appl Physiol 2000 89 1317 1321 11007564
Dubin A Murias G Estenssoro E Canales H Badie J Pozo M Sottile JP Baran M Palizas F Laporte M Intramucosal-arterial PCO2 gap (ΔPCO2) fails to increase during hypoxic hypoxia Crit Care 2002 6 514 520 12493073 10.1186/cc1813
Mecher C Rackow EC Astiz ME Weil MH Unaccounted for anion in metabolic acidosis during severe sepsis in humans Crit Care Med 1991 19 705 711 2026034
Rackow EC Mecher C Astiz ME Goldstein C McKee D Weil MH Unmeasured anion during severe sepsis with metabolic acidosis Circ Shock 1990 30 107 115 2311201
Gutierrez G A mathematical model of tissue-blood carbon dioxide exchange during hypoxia Am J Respir Crit Care Med 2004 169 525 533 14656752 10.1164/rccm.200305-702OC
Pittner A Nalos M Asfar P Yang Y Ince C Georgieff M Bruckner UB Radermacher P Froba G Mechanisms of inducible nitric oxide synthase (iNOS) inhibition-related improvement of gut mucosal acidosis during hyperdynamic porcine endotoxemia Intensive Care Med 2003 29 312 316 12594592
Jakob SM Splanchnic ischaemia Crit Care 2002 6 306 312 12225604 10.1186/cc1515
Taylor DE Piantadosi CA Oxidative metabolism in sepsis and sepsis syndrome J Crit Care 1995 10 122 135 7496449 10.1016/0883-9441(95)90003-9
Oud L Kruse JA Poor in vivo reproducibility of gastric intramucosal pH determined by saline-filled balloon tonometry J Crit Care 1996 11 144 150 8891966 10.1016/S0883-9441(96)90011-8
Steverink PJGM Kolkman JJ Groeneveld ABJ De Vries JW Catheter deadspace: a source of error during tonometry Br J Anaesth 1998 80 337 341 9623434
Kellum JA Saline-induced hyperchloremic metabolic acidosis Crit Care Med 2002 30 259 260 11902280
Rudinsky BF Meadow WL Relationship between oxygen delivery and metabolic acidosis during sepsis in piglets Crit Care Med 1992 20 831 839 1597039
Gow KW Phang PT Tebbutt-Speirs SM English JC Allard MF Goddard CM Walley KR Effect of crystalloid administration on oxygen extraction in endotoxemic pigs J Appl Physiol 1998 85 1667 1675 9804568
Kellum JA Song M Li J Lactic and hydrochloric acids induce different patterns of inflammatory response in LPS-stimulated RAW 264.7 cells Am J Physiol Regul Integr Comp Physiol 2004 286 R686 R692 14695114
Thatte HS Rhee JH Zagarins SE Treanor PR Birjiniuk V Crittenden MD Khuri SF Acidosis-induced apoptosis in human and porcine heart Ann Thorac Surg 2004 77 1376 1383 15063270 10.1016/j.athoracsur.2003.07.047
Baylor AE 3rdDiebel LN Liberati DM Dulchavsky SA Brown WJ Diglio CA The synergistic effects of hypoxia/reoxygenation or tissue acidosis and bacteria on intestinal epithelial cell apoptosis J Trauma 2003 55 241 247 12913632
Kellum JA Fluid resuscitation and hyperchloremic acidosis in experimental sepsis: improved short-term survival and acid-base balance with Hextend compared with saline Crit Care Med 2002 30 300 305 11889298 10.1097/00003246-200202000-00006
| 15774052 | PMC1175914 | CC BY | 2021-01-04 16:04:51 | no | Crit Care. 2005 Jan 11; 9(2):R66-R73 | utf-8 | Crit Care | 2,005 | 10.1186/cc3021 | oa_comm |
==== Front
Crit CareCritical Care1364-85351466-609XBioMed Central London cc30231577405110.1186/cc3023ResearchIntensive care unit-acquired urinary tract infections in a regional critical care system Laupland Kevin B [email protected] Sean M [email protected] Daniel B [email protected] Andrew W [email protected] Terry [email protected] Deirdre L [email protected] Assistant Professor, Departments of Critical Care Medicine, Pathology and Laboratory Medicine, and Community Health Services, Center for Anti-microbial Resistance, Calgary Health Region, Calgary Laboratory Services, and the University of Calgary, Calgary, Alberta, Canada2 Fellow, Departments of Medicine and Community Health Services, Calgary Health Region, Calgary Laboratory Services, and the University of Calgary, Calgary, Alberta, Canada3 Associate Professor, Departments of Pathology and Laboratory Medicine, and Medicine, Calgary Health Region, Calgary Laboratory Services, and the University of Calgary, Calgary, Alberta, Canada4 Clinical Assistant Professor, Departments of Critical Care Medicine, and Surgery, Calgary Health Region and the University of Calgary, Calgary, Alberta, Canada5 Analyst, Center for Anti-microbial Resistance, Calgary Health Region, Calgary Laboratory Services, and the University of Calgary, Calgary, Alberta, Canada6 Professor, Departments of pathology and Laboratory Medicine, and Medicine, Calgary Health Region, Calgary Laboratory services, and the University of Calgary, Calgary, Alberta, Canada2005 6 1 2005 9 2 R60 R65 12 11 2004 23 11 2004 Copyright © 2004 Laupland 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
Few studies have evaluated urinary tract infections (UTIs) specifically acquired within intensive care units (ICUs), and the effect of such infections on patient outcome is unclear. The objectives of this study were to describe the occurrence, microbiology, and risk factors for acquiring UTIs in the ICU and to determine whether these infections independently increase mortality.
Methods
A surveillance cohort study was conducted among all adults admitted to multi-system and cardiovascular surgery ICUs in the Calgary Health Region (CHR, population about 1 million) between 1 January 2000 and 31 December 2002.
Results
During the 3 years, 4465 patients were admitted 4915 times to a CHR ICU for 48 hours or more. A total of 356 ICU-acquired UTIs (defined as at least 105 colony-forming units/ml of one or two organisms 48 hours or more after ICU admission) occurred among 290 (6.5%) patients, yielding an overall incidence density of ICU-acquired UTIs of 9.6 per 1000 ICU days. Four bacteremic/fungemic ICU-acquired UTIs occurred (0.1 per 1000 ICU days). Development of an ICU-acquired UTI was more common in women (relative risk [RR] 1.58; 95% confidence interval [CI] 1.43–1.75; P < 0.0001) and in medical (9%) compared with non-cardiac surgical (6%), and cardiac surgical patients (2%). The most common organisms isolated were Escherichia coli (23%), Candida albicans (20%), and Enterococcus species (15%). Antibiotic-resistant organisms were identified among 14% isolates. Although development of an ICU-acquired UTI was associated with significantly higher crude in-hospital mortality (86/290 [30%] vs. 862/4167 [21%]; RR = 1.43; 95% CI 1.19–1.73; P < 0.001); an ICU-acquired UTI was not an independent predictor for death.
Conclusions
Development of an ICU-acquired UTI is common in critically ill patients. Although a marker of increased morbidity associated with critical illness, it is not a significant attributable cause of mortality.
incidenceintensive care unitmortalityurinary tract infection
==== Body
Introduction
Infection of the urinary tract (UTI) is the most common hospital-acquired infection in North America and is among the most frequent nosocomial infections in critically ill patients [1-4]. Nosocomial UTIs have been associated with a threefold increased risk for mortality in hospital-based studies, with estimates of more than 50,000 excess deaths occurring per year in the USA as a result of these infections [5]. Furthermore, in several studies nosocomial UTIs have been associated with increased length of hospital stay and cost [6,7]. Despite their importance, there have only been very limited studies focused on nosocomial UTIs in the critically ill. Richards and colleagues reported on intensive care unit (ICU) nosocomial infections in the National Nosocomial Infections Surveillance System (NNIS) database and found that UTI was responsible for 20–30% of nosocomial infections in medical/surgical ICUs [1,8]. Finklestein and colleagues determined an incidence of 10–14 UTI per 1000 catheter days among 337 patients in a single Israeli ICU [9]. Rosser and colleagues retrospectively reviewed 126 trauma ICU patients with sepsis and found that increased length of stay, length of catheterization, and age (more than 60 years) were independent factors associated with the development of nosocomial UTI [10]. These studies were limited in part either as a result of being conducted in specialized ICUs or critically ill patient subsets, by small sample size, or by limited assessment of outcome.
We previously conducted a study of all patients admitted to multidisciplinary ICUs in the Calgary Health Region (CHR) during a 1 year period and found that increased length of stay and female gender were independently associated with the acquisition of these infections [2]. However, this study was limited by exclusion of many cardiovascular surgery patients and like other investigations had insufficient statistical power to detect a clinically important increased risk for mortality associated with ICU-acquired UTIs [11]. We therefore conducted further surveillance among all critically ill adults in the CHR to better delineate the occurrence, microbiology, and risk factors for acquiring ICU-acquired UTIs and determine whether these infections increase the risk for mortality.
Methods
Study population
The CHR administers all acute hospital care to the residents of the cities of Calgary and Airdrie and several large adjacent regions (population about 1 million). All ICUs within the CHR are closed units staffed by fully trained intensivists and are administered by the Department of Critical Care Medicine, University of Calgary, and CHR. These include a 14-bed cardiovascular surgery ICU (CVICU) and multidisciplinary ICUs (total 44 beds) at each of the three adult acute care centers in the CHR. All patients 18 years of age and older admitted to an adult multidisciplinary ICU or the CVICU in the CHR for at least 48 hours during 1 January 2000 and 31 December 2002 were included in the study. The Conjoint Health Research Ethics Review Board at the University of Calgary approved the study.
Protocol
The study used a cohort design that linked data from regional administrative, critical care, and microbiology databases. Demographic, clinical, basic laboratory, and outcome data were obtained from all patients admitted to ICUs in the CHR using the ICU Tracer database [12]. Calgary Laboratory Services, a region-based laboratory that handles all routine bacterial specimens from CHR patients, identified all relevant positive culture results. Data from the source databases were linked on the basis of unique hospital numbers using Access 2002 (Microsoft Corp., Redmond, WA).
Definitions
An ICU-acquired UTI was defined using a modification of the criteria of Costel and colleagues as those patients with a positive urine culture (at least 100,000 colony-forming units/ml of one or two organisms) first identified on ICU day 3 (48 hours) or later [2,13]. Patients with positive urine cultures within 48 hours of ICU discharge were also considered to have ICU-acquired UTIs. A bacteremic/fungemic UTI was defined as a UTI with a concomitantly positive blood culture with the same organism within a 48 hour period [2]. A surgical patient was any patient recorded as having an operative diagnosis or admitted from the trauma ward, post cardiac surgery care unit, or directly from the operating room. Severity of illness and intensity of care at admission were assessed using the Acute Physiology and Chronic Health Evaluation II (APACHE II) and the Therapeutic Intervention Scoring System (TISS) scores, respectively [14,15]. Shock was deemed to be present if a vasopressor infusion was required. Laboratory testing was performed in accordance with standard guidelines as described previously [2] with the exception that, as of June 2001, only cultures either positive by screening by an ATPase–luciferase assay or by specific physician request were cultured [16]. Antimicrobial-resistant organisms were defined as methicillin-resistant Staphylococcus aureus, vancomycin-resistant Enterococcus faecalis or faecium, or any Gram-negative organisms resistant to one or more of ciprofloxacin, tobramycin/gentamicin, ceftazidime, piperacillin, or imipenem.
Statistical analysis
Analysis was performed with Stata version 8.0 (Stata Corp, College Station, TX). The occurrence of ICU-acquired UTI was expressed as (1) the cumulative incidence of patients with at least one UTI episode per admittance to ICU, and (2) the incidence density based on the number of ICU-acquired UTI episodes per total patient days of ICU stay. Normally or near-normally distributed variables were reported as means and standard deviations, and non-normally distributed variables as medians with inter-quartile ranges (IQRs). Means were compared with Student's t-test and medians with the Mann–Whitney U-test. Differences in proportions between categorical data were assessed with the χ2 or Fisher's exact test. A multivariable logistic regression model was developed to identify independent risk factors for mortality associated with these infections, with the use of backward stepwise variable elimination. Variables included those identified in our previous study [2] and those found to be significant to the P ≤ 0.1 level in univariate analysis. Final model discrimination was assessed by using the area under the receiver operator characteristic curve and calibration by using the Hosmer–Lemeshow goodness-of-fit test. P < 0.05 was considered significant for all comparisons unless otherwise stated.
Results
Demographics
During the 3 years of the study, 4465 patients were admitted 4915 times to a CHR ICU for 48 hours or more. Twenty-five percent (1099) of admissions were to the CVICU. Sixty-one percent (2709) of the patients were male, the mean age was 61.2 ± 17.4 years, and the mean APACHE II scores were 26.1 ± 8.3 points. In all, 1975 (45%) were classified as medical patients.
Incidence of ICU-acquired UTI
A total of 356 ICU-acquired UTIs occurred among 290 (7%) patients during surveillance. Three hundred and three ICU-acquired UTIs were on first ICU admission episodes (that is, 13 patients fulfilled criteria for a second UTI during their first ICU stay) and 43 were on second, 9 on third, and 1 on fifth ICU admission episodes. The overall incidence density of ICU-acquired UTI was 9.6 per 1000 ICU days. Only four ICU-acquired UTIs were associated with a positive blood culture with the same organism for an overall incidence density of bacteremic/fungemic ICU-acquired UTI of 0.1 per 1000 ICU days. The overall incidence of ICU-acquired UTI was significantly (P ≤ 0.01) higher in the year 2000 (143/1531; 9%) than in 2001 (112/1651; 7%) or 2002 (101/1733; 6%).
Factors associated with the development of an ICU-acquired UTI
Several factors present at admission to ICU were associated with increased incidence of ICU-acquired UTI. Women (174/1755), in comparison with men (116/2709), were at significantly increased risk (relative risk [RR] = 1.58; 95% confidence interval [CI] 1.43–1.75; P < 0.0001) for development of an ICU-acquired UTI. A significantly different rate of development of these infections was observed among admission categories, with an incidence of 9% (181/1975) in medical patients, 6% (89/1391) in non-cardiovascular surgical patients, and 2% (20/1099) in cardiovascular surgical patients (P < 0.001 overall; and P < 0.005 for each pairwise comparison). No differences were observed between patients who developed an ICU-acquired UTI with regard to either mean age or APACHE II score, although patients who developed an ICU-acquired UTI had lower mean admission Therapeutic Intervention Scoring System scores than those patients who did not develop one of these infections (41.6 ± 15.1 versus 45.1 ± 17.3 points; P < 0.01).
A significant association between ICU length of stay and development of an ICU-acquired UTI was observed. The median length of ICU stay among patients with ICU-acquired UTI was 12.0 (IQR 5.7–21.0) days compared with 4.1 (IQR 2.8–7.5) days for those without (P < 0.0001). Similarly an increased overall median hospital length of stay was associated with development of an ICU-acquired UTI (30 days, IQR 16–62; 16 days, IQR 9–29; P < 0.0001).
Microbiology
The median time from ICU admission to development of a first UTI was 7.0 (IQR 4.2–12.1) days. Most (337/356; 95%) of the UTIs were monomicrobial infections but in 19 cases two organisms were identified simultaneously at 108 colony-forming units/l or more. The organisms causing ICU-acquired UTI are shown in Table 1. Antibiotic-resistant organisms were identified in 14% (53/375) of isolates. These organisms were Escherichia coli in 29 (55%), Pseudomonas aeruginosa in 12 (23%), Klebsiella species in 5 (9%), and other Gram-negative enterics in 7 (13%); none were vancomycin-resistant Enterococcus faecalis or faecium or methicillin-resistant Staphylococcus aureus. Among these antibiotic-resistant organisms, resistance occurred to ciprofloxacin in 33% (17/52), gentamicin in 21% (11/52), tobramycin in 9% (5/53), ceftazidime in 13% (7/53), piperacillin in 61% (31/51), and piperacillin/tazobactam in 12% (6/51). Imipenem resistance was identified in one of seven isolates of Pseudomonas aeruginosa tested (susceptibility testing for carbapenems only started routinely in 2002). Resistance to two different classes of antimicrobials occurred in 13 isolates, and 3 isolates were resistant to three different classes.
Mortality
Development of an ICU-acquired UTI was associated with a significantly higher crude ICU-related mortality (52/290 [18%] versus 519/4175 [12%]; RR = 1.44; 95% CI 1.11–1.89; P = 0.01) and overall in-hospital mortality (86/290 [30%] vs. 862/4167 [21%]; RR = 1.43; 95% CI 1.19–1.73; P < 0.001)] than those who did not develop this infection. A multivariable logistic regression model (n = 4434) that had both good discrimination (area under receiver operator characteristic curve = 0.75) and calibration (goodness-of-fit P = 0.08) was developed to assess risk factors for in-hospital death. After controlling for other significant covariates, ICU-acquired UTI was not independently associated with death, as shown in Table 2.
Discussion
We observed an incidence density of ICU-acquired UTI of 9.6 per 1000 ICU days that is comparable to that observed in other studies that evaluated nosocomial UTIs in ICUs [2,9,10,17]. However, an important strength of this study is that all patients admitted to adult ICUs (both academic-based and community-based) in a large region were included. As a result, this study should be representative of many critically ill populations at large and the results more widely generalizable. Previous studies have been limited to single specialized medical, surgical, or combined medical–surgical ICUs [9,10] or in series of selected ICUs participating in surveillance systems [1,8]. Our previous study, which included all multidisciplinary ICUs, was limited in part because we failed to include many cardiovascular surgical patients [2]. As demonstrated by our observation of a significant difference in risk of acquiring ICU-acquired UTI between cardiac surgical, non-cardiac surgical, and medical patients, care must be paid to patient 'case-mix' in comparing between studies of these infections. It is noteworthy that we did not exclude non-residents of the CHR in this study despite the fact that we have previously argued for such a practice [18]. Given that we did not observe any significant rate differences among CHR residents and non-residents (data not shown), that the population at risk was restricted to those admitted to CHR ICU (and not the entire base population of the CHR), and that the mortality outcome for those with a ICU-acquired UTI was not related to residency status, we pooled our entire patient cohort for analysis.
There are several possible explanations for our important observation of a lower rate of ICU-acquired UTIs in the latter 2 years of the study. The first possibility is that heightened awareness from our first report [2] or concomitant preventive efforts (such as a large regional quality improvement initiative to reduce ventilator-associated pneumonia) with increased attention to the use of medical devices and attention to hand washing among staff could have had a role. Anecdotally, we feel it is unlikely to be related to a decreased use of urinary catheters because we estimate that nearly all (more than 90%) of our patients ill enough to require ICU admission for 2 or more days have an indwelling urinary catheter. A second possibility for the reduced rate in the latter years of the study is that there might have been increased use of systemic antimicrobials active against urinary pathogens. This is only speculative because we do not have actual data to support this possibility. A third consideration is that physicians less frequently ordered urine cultures in the second and third years of the study such that the overall culture positivity rate was less. Unlike in our first study, in which we collected data on negative cultures [2], we did not have access to such results in the present study and are therefore unable to assess this. The fourth possibility, and the one that we suspect might be the most important reason for the reduced rate of ICU-acquired UTIs being diagnosed in the latter part of the study, is that we changed our laboratory testing practice in June 2001. At that point a bacteriuria screening assay was implemented regionally after demonstrating its utility in outpatients [16]. Since that time only urine samples either positive by that assay or by specific physician request are cultured. On the basis of the sensitivity of the assay of 86%, it is expected that a 10–15% reduction in the rate of culture positivity would occur with its implementation. However, a more important influence is that this assay does not detect yeast because it is based on the specific detection of bacterial ATP [16,19]. Candida species are among the most important causes of ICU-acquired UTI and a reduced rate of ICU-acquired UTI is expected if these organisms are not routinely cultured. We are currently planning a study to evaluate the optimal laboratory means of identifying ICU-acquired UTIs in our region.
The most clinically important and novel finding of this study was that ICU-acquired UTIs do not independently increase the risk for death among patients admitted to ICUs. Unlike in all previous studies potentially able to investigate this question, the present study was adequately powered to detect a clinically significant increased mortality risk [2,9,10]. Although we did observe that these infections increased the crude mortality risk, once confounding for measures of severity of disease, diagnostic category, and length of ICU stay were controlled for, ICU-acquired UTI was not significantly associated with death. This contrasts with the findings of Platt and colleagues, which showed that nosocomial UTIs were associated with a significant attributable mortality in a general hospital population [5]. It is clinically important to ascertain whether ICU-acquired UTIs are associated with attributable mortality because there may be implications for treatment. Although practice variation among different intensivists and between ICUs in our region probably exists, we commonly withhold antimicrobial therapy for bacteriuria or funguria in the absence of an associated clinical infective syndrome. Although a randomized, prospective, clinical trial is required to address optimal practice, our current observations of a low rate of bacteremic/fungemic ICU-acquired UTI and lack of attributable mortality suggests that a clinical judgment-based approach to treatment may be reasonable.
Conclusion
We present the results of a large observational cohort study that confirms that critical illness is commonly complicated by the development of a nosocomial UTI. Although these infections are crudely associated with death they are not associated with a significantly increased attributable mortality. Further studies are needed for better definition of the potential adverse effect of these infections on patient morbidity and cost to the healthcare system.
Key messages
• A surveillance cohort study was conducted among all adults admitted to multi-system and cardiovascular surgery ICUs in the Calgary Health Region (population about 1 million) during a 3 year period.
• ICU-acquired UTI commonly (7%) complicated the course of patients admitted to ICU for greater than 48 hours; women and medical patients were at highest risk.
• Development of an ICU-acquired UTI was not an independent risk factor for in-hospital mortality.
Abbreviations
APACHE = Acute Physiology and Chronic Health Evaluation; CHR = Calgary Health Region; CI = confidence interval; CVICU = cardiovascular surgery intensive care unit; ICU = intensive care unit; IQR = inter-quartile range; RR = relative risk; UTI = urinary tract infection.
Competing interests
The author(s) declare that they have no competing interests.
Authors' contributions
KBL conceived and designed the study, conducted the principal analysis, and drafted the manuscript. SMB participated in the study design and revision of the manuscript. DBG, TR, and DLC contributed to data collection and manuscript revision. AWK assisted with analysis and manuscript revision. All authors read and approved the final manuscript.
Acknowledgements
This study was supported in part by a grant from the Canadian Intensive Care Foundation.
Figures and Tables
Table 1 Microbial etiologies of ICU-acquired urinary tract infections, Calgary, Alberta, Canada
Organism Occurrence (%)a
Escherichia coli 87 (23)
Candida albicans 74 (20)
Enterococcus species 57 (15)
Pseudomonas aeruginosa 34 (10)
Candida glabrata 20 (5)
Klebsiella species 20 (5)
Coagulase-negative staphylococci 18 (5)
Proteus mirabilis 17 (5)
Candida otherb 16 (4)
Enterobacter species 10 (3)
Citrobacter species 5 (1)
Staphylococcus aureus 4 (1)
Streptococcus species 3 (1)
Otherc 10 (3)
aThe number of isolates (n = 375) exceeds the number of patients with ICU-acquired UTI because some patients had more than one UTI episode or had polymicrobial infection.
bCandida krusei (1), C. lusitaniae (1), C. tropicalis (4), Candida not speciated (10).
cIncludes Lactobacillus species (1), Proteus penneri (1), Serratia marcescens (2), Morganella morganii (2), Providencia rettgeri (1), Hafnia alvei (2) and Staphylococcus saphrophyticus (1).
ICU, intensive care unit; UTI, urinary tract infection.
Table 2 Multivariable logistic modeling of variables associated with in-hospital death among patients admitted for at least 48 hours to a Calgary Health Region intensive care unit, 2000–2002
Variable Odds ratio (95% CI) P
Admission APACHE II score (per point) 1.09 (1.08–1.10) <0.001
Logarithm of ICU length of stay (per day) 1.31 (1.18–1.45) <0.001
Non-cardiac surgical patienta 0.78 (0.66–0.92) <0.01
Cardiac surgical patienta 0.14 (0.11–0.18) <0.001
ICU-acquired urinary tract infection 1.02 (0.76–1.37) 0.9
aCompared with a reference group of medical patients.
APACHE II, Acute Physiology and Chronic Health Evaluation II; CI, confidence interval; ICU, intensive care unit.
==== Refs
Richards MJ Edwards JR Culver DH Gaynes RP Nosocomial infections in combined medical-surgical intensive care units in the United States Infect Control Hosp Epidemiol 2000 21 510 515 10968716
Laupland KB Zygun DA Davies HD Church DL Louie TJ Doig CJ Incidence and risk factors for acquiring nosocomial urinary tract infection in the critically ill J Crit Care 2002 17 50 57 12040549 10.1053/jcrc.2002.33029
Haley RW Culver DH White JW Morgan WM Emori TG The nationwide nosocomial infection rate. A new need for vital statistics Am J Epidemiol 1985 121 159 167 4014113
Erbay H Yalcin AN Serin S Turgut H Tomatir E Cetin B Zencir M Nosocomial infections in intensive care unit in a Turkish university hospital: a 2-year survey Intensive Care Med 2003 29 1482 1488 12898002 10.1007/s00134-003-1788-x
Platt R Polk BF Murdock B Rosner B Mortality associated with nosocomial urinary-tract infection N Engl J Med 1982 307 637 642 7110215
Centers for Disease Control Public health focus: Surveillance, prevention, and control of nosocomial infections MMWR 1992 41 783 787
Givens CD Wenzel RP Catheter-associated urinary tract infections in surgical patients: a controlled study on the excess morbidity and costs J Urol 1980 124 646 648 7452793
Richards MJ Edwards JR Culver DH Gaynes RP Nosocomial infections in medical intensive care units in the United States. National Nosocomial Infections Surveillance System Crit Care Med 1999 27 887 892 10362409 10.1097/00003246-199905000-00020
Finkelstein R Rabino G Kassis I Mahamid I Device-associated, device-day infection rates in an Israeli adult general intensive care unit J Hosp Infect 2000 44 200 205 10706803 10.1053/jhin.1999.0682
Rosser CJ Bare RL Meredith JW Urinary tract infections in the critically ill patient with a urinary catheter Am J Surg 1999 177 287 290 10326844 10.1016/S0002-9610(99)00048-3
Fagon JY Novara A Stephan F Girou E Safar M Mortality attributable to nosocomial infections in the ICU Infect Control Hosp Epidemiol 1994 15 428 434 7963432
Doig CJ Zygun DA Fick GH Laupland KB Boiteau PJ Shahpori R Rosenal T Sandham JD Study of clinical course of organ dysfunction in intensive care Crit Care Med 2004 32 384 390 14758152 10.1097/01.CCM.0000108881.14082.10
Costel EE Mitchell S Kaiser AB Abbreviated surveillance of nosocomial urinary tract infections: a new approach Infect Control 1985 6 11 13 3843988
Knaus W Draper E Wagner D Zimmerman J APACHE II: A severity of disease classification system Crit Care Med 1985 13 818 829 3928249
Cullen D Civetta J Briggs B Ferrara L Therapuetic interventions scoring system: a method of quantitative comparision of patient care Crit Care Med 1974 2 57 60 4832281
Semeniuk H Noonan J Gill H Church D Evaluation of the Coral UTI Screen system for rapid automated screening of significant bacteriuria in a regional centralized laboratory Diagn Microbiol Infect Dis 2002 44 7 10 12376024 10.1016/S0732-8893(02)00424-8
Martinez OV Civetta JM Anderson K Roger S Murtha M Malinin TI Bacteriuria in the catheterized surgical intensive care patient Crit Care Med 1986 14 188 191 3943334
Laupland KB Population-Based Epidemiology of Intensive Care: Critical Importance of Ascertainment of Residency Status Critical Care 2004 8 R431 R436 15566588 10.1186/cc2947
Laupland KB Church DL Gregson DB Evaluation of a rapid bacterial ATP assay for screening BAL samples from ICU patients submitted for quantitative bacterial cultures Diagn Microbiol Infect Dis 2003 47 465 469 14596964 10.1016/S0732-8893(03)00151-2
| 15774051 | PMC1175915 | CC BY | 2021-01-04 16:04:51 | no | Crit Care. 2005 Jan 6; 9(2):R60-R65 | utf-8 | Crit Care | 2,005 | 10.1186/cc3023 | oa_comm |
==== Front
Crit CareCritical Care1364-85351466-609XBioMed Central London cc30251577405310.1186/cc3025ResearchExtravascular lung water in patients with severe sepsis: a prospective cohort study Martin Greg S [email protected] Stephanie [email protected] Meredith [email protected] Marc [email protected] Director, Medical and Coronary Intensive Care Units, Grady Memorial Hospital, and Assistant Professor of Medicine, Division of Pulmonary, Allergy and Critical Care, Emory University School of Medicine, Atlanta, Georgia, USA2 Grady Memorial Hospital, Emory University School of Medicine, Division of Pulmonary and Critical Care Medicine, Atlanta, Georgia, USA3 Senior Research Coordinator, Division of Pulmonary, Allergy and Critical Care, Emory University School of Medicine, Atlanta, Georgia, USA4 Section Head, Pulmonary Allergy and Critical Care Medicine, Grady Memorial Hospital, and Associate Professor of Medicine, Division of Pulmonary, Allergy and Critical Care, Emory University School of Medicine, Atlanta, Georgia, USA2005 11 1 2005 9 2 R74 R82 11 11 2004 23 11 2004 Copyright © 2005 Martin 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 cited.
Introduction
Few investigations have prospectively examined extravascular lung water (EVLW) in patients with severe sepsis. We sought to determine whether EVLW may contribute to lung injury in these patients by quantifying the relationship of EVLW to parameters of lung injury, to determine the effects of chronic alcohol abuse on EVLW, and to determine whether EVLW may be a useful tool in the diagnosis of acute respiratory distress syndrome (ARDS).
Methods
The present prospective cohort study was conducted in consecutive patients with severe sepsis from a medical intensive care unit in an urban university teaching hospital. In each patient, transpulmonary thermodilution was used to measure cardiovascular hemodynamics and EVLW for 7 days via an arterial catheter placed within 72 hours of meeting criteria for severe sepsis.
Results
A total of 29 patients were studied. Twenty-five of the 29 patients (86%) were mechanically ventilated, 15 of the 29 patients (52%) developed ARDS, and overall 28-day mortality was 41%. Eight out of 14 patients (57%) with non-ARDS severe sepsis had high EVLW with significantly greater hypoxemia than did those patient with low EVLW (mean arterial oxygen tension/fractional inspired oxygen ratio 230.7 ± 36.1 mmHg versus 341.2 ± 92.8 mmHg; P < 0.001). Four out of 15 patients with severe sepsis with ARDS maintained a low EVLW and had better 28-day survival than did ARDS patients with high EVLW (100% versus 36%; P = 0.03). ARDS patients with a history of chronic alcohol abuse had greater EVLW than did nonalcoholic patients (19.9 ml/kg versus 8.7 ml/kg; P < 0.0001). The arterial oxygen tension/fractional inspired oxygen ratio, lung injury score, and chest radiograph scores correlated with EVLW (r2 = 0.27, r2 = 0.18, and r2 = 0.28, respectively; all P < 0.0001).
Conclusions
More than half of the patients with severe sepsis but without ARDS had increased EVLW, possibly representing subclinical lung injury. Chronic alcohol abuse was associated with increased EVLW, whereas lower EVLW was associated with survival. EVLW correlated moderately with the severity of lung injury but did not account for all respiratory derangements. EVLW may improve both risk stratification and management of patients with severe sepsis.
See related commentary
==== Body
Introduction
Severe sepsis is a common syndrome among hospitalized patients, occurring at a rate of 250,000–750,000 cases/year in the USA [1,2]. It is defined as pathologic infection accompanied by a spectrum of physiologic abnormalities, originally described as systemic inflammatory response syndrome criteria in combination with acute organ dysfunction [3,4]. Sepsis is associated with high death rates, killing 30–50% of those severely afflicted [1,5], and is the leading cause of death among patients in noncoronary intensive care units (ICUs) in the USA [6]. According to the annual report from the National Center for Health Statistics [7], sepsis has risen to being the 10th leading cause of death overall in the USA.
Respiratory failure is among the most frequent complications of severe sepsis, occurring in nearly 85% of cases [5]. The mechanisms of acute lung failure in sepsis are complex and incompletely understood [8]. The hallmark of sepsis is increased capillary permeability, which manifests in the lungs as altered alveolar–capillary barrier function and is characterized by accumulation of extravascular lung water (EVLW). However, there is a paucity of data regarding EVLW in patients with severe sepsis.
The most severe form of lung failure, acute respiratory distress syndrome (ARDS), occurs in 40% of patients with sepsis [9]. As with sepsis, ARDS is a heterogeneous clinical syndrome. Recognition of ARDS relies upon a clinical definition, which was standardized in 1994 by the American–European Consensus Conference (AECC) [9]. These criteria comprise a constellation of clinical and radiographic findings that are associated with varying degrees of reliability [10]. No previous diagnostic criteria for ARDS have included measures of EVLW.
A variety of pre-existing comorbid conditions may alter the incidence and severity of ARDS. Chronic alcohol abuse is independently associated with a doubling in risk for developing ARDS, and once ARDS has developed it is associated with a nearly twofold risk for dying [11]. Similarly, chronic alcohol abuse is associated with more severe organ dysfunction in patients with septic shock [12]. Animal models of chronic alcohol abuse confirm the presence of steady-state abnormalities in alveolar–capillary permeability [13]. Initial findings in humans with chronic alcohol abuse suggest that alveolar–capillary barrier function is persistently altered [14].
We hypothesized that acute respiratory failure accompanying severe sepsis relates to subclinical abnormalities in capillary permeability. If this is true, then these abnormalities would be clinically apparent in the accumulation of EVLW across a broad population of patients with severe sepsis. We conducted the largest prospective evaluation to date of EVLW among critically ill patients with severe sepsis. We also evaluated the heterogeneity of EVLW in those patients who developed ARDS and the impact that chronic alcohol abuse had on the accumulation of EVLW and respective outcomes.
Methods
This study was reviewed and approved by the Institutional Review Board of Emory University School of Medicine. All patients admitted to the Medical ICU at Grady Memorial Hospital between July 2001 and March 2002 were screened for eligibility. Included patients met standard published criteria for severe sepsis [15]. The exclusion criteria were pregnancy, contraindication to femoral artery catheterization, age less than 18 years, and inability to obtain informed consent from the patient or surrogate. All eligible patients were enrolled within 72 hours of meeting criteria for severe sepsis. Patient management decisions, including the type and amount of volume resuscitation, were at the discretion of the primary intensive care physician.
At the time of enrollment, patient-specific data were obtained, including demographic data, past medical and social history, source of sepsis, and Acute Physiology and Chronic Health Evaluation (APACHE) II score [16]. A 5-F arterial catheter (Pulsiocath PV2015; Pulsion Medical Systems, Munich, Germany) was placed in the descending aorta via the femoral artery using the Seldinger technique. The arterial catheter and a standard central venous catheter were connected to pressure transducers and to an integrated bedside monitor (PiCCO; Pulsion Medical Systems). Continuous cardiac output (CO) calibration and EVLW measurements were obtained by triplicate central venous injections of 15–20 ml iced 0.9% saline solution. CO calibrations and determination of EVLW were performed immediately after catheter insertion and at least every 24 hours for 7 days. The catheter system was discontinued before 7 days had elapsed in the event of patient death or transfer from the ICU.
The PiCCO catheter system uses a single thermal indicator technique to determine EVLW, CO, and volumetric parameters. The bolus thermodilution CO is used to determine the patient's aortic impedance, which is used to calibrate the continuous CO [17,18]. CO is calculated using the Stewart–Hamilton method from thermodilution curves measured in the descending aorta, with accuracy comparable with that of pulmonary artery thermodilution [17-21]. The volume of distribution of the thermal indicator represents the intrathoracic thermal volume (ITTV), where ITTV (ml) = CO × mean transit time of the thermal indicator [22,23]. The pulmonary thermal volume (PTV) is given by PTV (ml) = CO × τ, where τ is exponential decay time of the thermodilution curve [24]. Global end-diastolic volume (GEDV), the combined end-diastolic volumes of all cardiac chambers, is given by ITTV – PTV (ml). This permits calculation of intrathoracic blood volume (ITBV) from the linear relationship with GEDV [22,25]: ITBV = 1.25 × GEDV – 28.4 (ml). EVLW is the difference between the thermal indicator distribution in the chest (ITTV) and the blood volume of the chest (ITBV) [22,25-29]: EVLW = ITTV – ITBV (ml).
Outcome variables
Parameters were indexed to total body surface area or to body weight in order to facilitate comparisons (e.g. EVLW refers to EVLWI). Patients were considered to have elevated EVLW if any measurement was greater than 10 ml/kg, based on previous studies examining the range of EVLW measurements in control patients with no clinical evidence of lung abnormalities [30,31]. Patients were followed for 28 days from enrollment to determine the occurrence of ARDS and death. ARDS was deemed to be present when the AECC criteria [9] were met within 7 days of developing severe sepsis. These criteria are as follows: acute onset of hypoxemia (arterial oxygen tension [PaO2]/fractional inspired oxygen [FiO2] ratio <200 mmHg) with bilateral infiltrates on chest radiograph and pulmonary artery occlusion pressure ≤ 18 mmHg or no evidence of left atrial hypertension. The severity of ARDS was quantitated using the Lung Injury Score (LIS) [32]. In addition, chest radiograph score (number of quadrants with >50% involvement with an alveolar filling process), PaO2/FiO2 ratio, and ventilator settings were recorded daily. The lung permeability index was calculated as the ratio of EVLW to ITBV, which was previously shown to reflect permeability of the alveolar–capillary barrier [23,33]. Patients were considered to have a history of chronic alcohol abuse if they had a history of alcohol abuse in their medical records or had a score of at least 3 on the Short Michigan Alcohol Assessment Test [34].
Statistical analysis
Data are presented as mean ± standard deviation, or as median (interquartile range [IQR]), depending on the distribution normality of the variable. Continuous variable measurements were compared using two-sample t-tests or Mann–Whitney U-tests for normally or non-normally distributed data, respectively. Multiple longitudinal comparisons were made by repeated measures analysis of variance (ANOVA) with time as a covariate. The χ2 statistic was used to compare frequency proportions. Modeling by least squares linear regression for continuous outcome variables and maximum likelihood logistic regression for dichotomous outcome variables was used to assess individual effects while adjusting for individually significant covariables. Statistical analysis was performed using NCSS 2001 software (NCSS, Inc., Kaysville, UT, USA) and all statistical tests were two-sided. P = 0.05 was considered statistically significant and P > 0.20 is reported as not significant.
Results
Severe sepsis study population
Twenty-nine patients with severe sepsis were enrolled at a median of 1 day after development of organ dysfunction requiring ICU admission. Demographic and physiologic characteristics are presented in Table 1. For 17 patients there were complete data for all 7 days; the study was terminated early because of patient death (n = 5) or transfer out of the ICU (n = 7) in the remaining 12 patients. The sources of sepsis were pneumonia (n = 16), intra-abdominal infection (n = 6), primary bloodstream infection (n = 4), and urosepsis (n = 3). The incidence of ARDS, according to the AECC definition, was 52% (15/29). Chronic alcohol abuse was present in 13 out of 29 patients (45%). The overall 28-day mortality was 41% (12/29).
At the time of enrollment, the median EVLW for all patients was 8.5 ml/kg (IQR 5.1–15.8 ml/kg). The mean PaO2/FiO2 ratio was 222.3 ± 149.8 mmHg and LIS was 1.80 ± 1.34; the median chest radiograph score was 2.0 (IQR 1.0–3.0). The mean baseline GEDV index (normal: 680–800 ml/m2) was 681 ml/m2 and the mean systemic vascular resistance index (normal: 1800–2500 dyn·s/cm5 per m2) was 1528 ± 562 dyn·s/cm5 per m2. Fluid balance (net intake/output) was consistently positive, with a cumulative mean during the study period of 8932 ± 9527 ml. The cumulative median EVLW for all patients over time was 9.0 ml/kg (IQR 6.5–15.2 ml/kg) and the mean change in EVLW from the beginning of the study period to the end was -1.1 ± 4.4 ml/kg. EVLW was greater in nonsurvivors than in survivors from severe sepsis (14 ml/kg [IQR 7.4–20 ml/kg] versus 8.0 ml/kg [IQR 5.9–11.2 ml/kg]; P < 0.001), and death was associated with greater EVLW over time (Fig. 1a; ANOVA P < 0.001). There were no significant longitudinal differences in oxygenation between survivors and nonsurvivors (Fig. 1b).
Correlates with extravascular lung water
We examined the relationship between measures of lung injury and EVLW. Using the PaO2/FiO2 ratio as a measure of oxygenation, we found a statistically significant but moderate correlation with EVLW (r2 = 0.27; P < 0.0001; Fig. 2a). Similar relationships were observed between EVLW and the chest radiograph score (r2 = 0.28) and the LIS (r2 = 0.18; both P < 0.0001). There was a significant correlation between the highest EVLW and lowest PaO2/FiO2 ratio (r2 = 0.32; P = 0.003), which was greater in nonsurvivors (r2 = 0.60; P = 0.005; Fig. 2b) than in survivors (r2 = 0.13; P = 0.20). There was a poor correlation between EVLW and GEDV index (r2 = 0.11; P < 0.001) and no correlation between EVLW and either daily or cumulative fluid balance.
Severe sepsis without acute respiratory distress syndrome
The baseline characteristics and physiology of the patients with severe sepsis without ARDS are presented in Table 1; there were no differences in fluid balance or hydrostatic pressure (GEDV index) between this subgroup and all severe sepsis patients combined. The median EVLW for the 14 non-ARDS severe sepsis patients was 7.7 ml/kg (IQR 5.0–10.2 ml/kg), but it was above normal in 57% of patients (8/14; Table 2). The median EVLW for non-ARDS patients with increased EVLW was 12.0 ml/kg (IQR 11.0–14.0 ml/kg), as compared with a median of 6.3 ml/kg (IQR 4.3–8.0 ml/kg) for patients with low EVLW (P < 0.001). Non-ARDS patients with a high EVLW were significantly more hypoxic than those with a low EVLW (mean PaO2/FiO2 ratio 230.7 ± 36.1 mmHg versus 341.2 ± 92.8 mmHg; P < 0.001). Calculated LIS values (mean 0.8 ± 0.7 versus 0.6 ± 0.8) and chest radiograph scores (median 2 [IQR 0–2] versus 1 [IQR 0–1]) were not significantly different between the two groups. A statistically insignificant increase in mortality was observed in non-ARDS patients with high EVLW (50% versus 17%; P = 0.20).
Severe sepsis with acute respiratory distress syndrome
Baseline characteristics and physiology for severe sepsis patients who developed AECC-defined ARDS (n = 15) were similar to those for the non-ARDS patients, with the exception of greater EVLW (Table 1) and increased measures of lung permeability (lung permeability index [EVLW/ITBV ratio] 1.18 ± 0.45 versus 0.60 ± 0.31; P < 0.001). Fluid balance and hydrostatic pressures were not different at baseline or longitudinally from those in non-ARDS patients, and did not correlate with the development of ARDS. GEDV index correlated weakly with EVLW (r2 = 0.17; P < 0.001) whereas fluid balance did not correlate. Differences according to EVLW for the ARDS patients are presented in Table 2. The median EVLW for ARDS patients was 12.0 ml/kg (IQR 7.8–17.7 ml/kg) and the diagnosis of ARDS was associated with increased EVLW over time compared with non-ARDS patients (repeated measures ANOVA, P < 0.001).
Of the ARDS patients, only 73% (11/15) had any evidence of increased EVLW during the study period. The median EVLW for ARDS patients with low EVLW patients was 7.0 ml/kg (IQR 6.0–8.3 ml/kg), as compared with 16.9 ml/kg (IQR 14.8–22.3 ml/kg) for the high EVLW ARDS patients (P < 0.001). Cumulative mean oxygenation during the study period was worse among high EVLW ARDS patients (PaO2/FiO2 ratio 135.4 ± 60.4 versus 197.0 ± 106.7 mmHg; P = 0.001). Cumulative mean chest radiograph scores (4 [IQR 4–4] versus 3 [IQR 2–4]; P = 0.002) and LIS (2.8 ± 1.1 versus 2.1 ± 0.7; P = 0.002) were similarly worse in high EVLW ARDS patients.
There was significantly reduced mortality among the 27% of ARDS patients with consistently low EVLW as compared with the ARDS patients with high EVLW (0/4 versus 7/11; P = 0.03). The high EVLW group had a significantly greater APACHE II score than did the low EVLW group (25.9 ± 6.3 versus 18.5 ± 3.3; P = 0.05), although differences in APACHE II score accounted for under 10% of the differences in EVLW by univariate regression analysis. If EVLW were substituted for bilateral radiographic infiltrates in the AECC diagnostic criteria, then three additional patients would have been diagnosed with ARDS, increasing the incidence by 20%.
Chronic alcohol abuse
Chronic alcohol abuse was present in 45% (13/29) of the severe sepsis patients, including 33% (5/15) of ARDS patients (Table 3). Patients with alcohol abuse had no evidence of cirrhosis or ascites. Hydrostatic pressures and serum albumin levels were not different from those in nonalcoholic patients. The lung permeability index was increased in ARDS patients with chronic alcohol abuse as compared with nonalcoholic ARDS patients (1.73 ± 0.33 versus 1.20 ± 0.47; P = 0.04). Net fluid intake was greater in the 24 hours before enrollment in alcoholic patients with ARDS (Table 3), although cumulative fluid balance during the study period was not different (10683 ± 10247 ml versus 7415 ± 8929 ml; not significant). Adjustment for baseline differences in fluid balance by linear regression revealed that alcohol abuse independently predicts greater EVLW by an average of 9.3 ml/kg in ARDS patients (P < 0.001).
All five ARDS patients with a history of chronic alcohol abuse had increased EVLW. Among ARDS patients, the chronic alcoholic patients' median EVLW over the course of the study was significantly elevated as compared with that in nonalcoholic patients (19.9 [IQR 16.0–28.5] ml/kg versus 8.7 [IQR 7.7–11.0] ml/kg; P < 0.0001); a similar relationship existed for non-ARDS patients (median alcoholic EVLW 8.7 [IQR 5.0–10.3] ml/kg versus 7.0 [IQR 5.0–8.0] ml/kg; P = 0.04). The relative risk for high EVLW was 2.4 times greater in ARDS patients with chronic alcohol abuse (P = 0.03). Using a repeated measures ANOVA, chronic alcohol abuse was associated with higher EVLW over the 7-day study duration among all patients (P = 0.04) and the subset of ARDS patients (P < 0.001). Mortality was 54% (7/13) for chronic alcoholic patients versus 31% (5/16) for nonalcoholic patients (not significant).
Discussion
Among severe sepsis patients without clinical ARDS, more than half manifest abnormal quantities of EVLW. Despite not meeting the consensus conference definition for ARDS, the amount of EVLW correlated with measures of lung injury (PaO2/FiO2 ratio, LIS, and chest radiograph score). Half of these patients were adequately hypoxemic to diagnose ARDS by the AECC criteria, but they did not exhibit the necessary bilateral radiographic infiltrates. Furthermore, 27% of the patients fulfilling the clinical consensus conference criteria for ARDS never had elevated EVLW, and these patients had improved survival as compared with ARDS patients with increased EVLW. These data support the hypothesis that EVLW varies substantially among patients with severe sepsis, and thus it may contribute to the high frequency of respiratory dysfunction. In addition, we found that severe sepsis patients with a history of chronic alcohol abuse had significantly greater EVLW than did nonalcoholic patients. This relationship was strengthened by the presence of ARDS, thus demonstrating the importance of comorbid disease for the risk and severity of ARDS.
Our findings have both diagnostic and prognostic implications for patients with severe sepsis. EVLW parallels the common clinical pathway and represents the physiologic derangements of ARDS, but it is not included in the AECC definition. Given that accumulation of lung water is one of the hallmarks of ARDS, the fact that 57% of severe sepsis patients without clinical ARDS have increased EVLW suggests that these patients have an unrecognized form of lung injury. Thus, despite the presence of hypoxemia, the AECC definition for ARDS may be insensitive to more subtle forms of ARDS because of variability in interpretation of chest radiograph [35] and the greater sensitivity of EVLW measures for detecting pulmonary edema [36,37]. Similar concerns have been voiced about the specificity of the definition [10], highlighting the need for an accurate early diagnostic marker when the diagnosis may be uncertain and therapeutic interventions may be most critical.
EVLW additionally serves as a prognostic marker for patients with ARDS. Previous studies have estimated EVLW in states of respiratory failure and/or ARDS with conflicting outcome results [38-42]. Modern studies including strictly defined ARDS patients corroborate an effect on mortality, particularly if changes in EVLW are considered over time [38]. However, historical methods of estimating EVLW have been complex, clinically difficult, and poorly reproducible [36,43-46]. The most common method of estimating EVLW continues to be with chest radiography, despite being imprecise and highly variable [36,37,47]. Given the ready availability and relative simplicity of EVLW measures compared with past methods, additional clinical trials are warranted to compare EVLW as a prognostic marker with other modern standards, such as pulmonary dead space [48].
The implications of EVLW measurements for severe sepsis patients with a history of chronic alcohol abuse may be even greater. The rate of development of ARDS among critically ill chronic alcoholic individuals is twice that in nonalcoholic individuals [11]; the risk is even higher among chronic alcoholic patients with severe sepsis (relative risk = 2.43, 95% confidence interval = 1.55–3.86). [12] The underlying mechanisms for increased ARDS susceptibility in chronic alcoholic individuals involve permeability defects, in which animal models of alcoholism have shown altered alveolar–capillary membrane permeability [13]. The mechanism for this alteration arises from perturbations in glutathione homeostasis, with otherwise healthy chronic alcoholic individuals having reduced levels of glutathione in their alveolar epithelial lining fluid [49] and apparent increased permeability to proteins [14]. The present report is the first to show an exaggerated increase in EVLW among chronic alcoholic ARDS patients, correlated with measures reported to indicate lung capillary permeability (lung permeability index), supporting the hypothesis that an ineffective permeability barrier may predispose susceptible alcoholic patients to heightened development of ARDS.
This study has several limitations. The size of the study prevents absolute conclusions from being drawn regarding EVLW in patients with severe sepsis, although these results stand as the largest prospective evaluation of EVLW in patients with severe sepsis. The transpulmonary thermodilution technique employed for measuring EVLW has been well validated in critically ill patients [22,25,38,50] despite prior concerns that severe ventilation–perfusion mismatch may preclude access to the complete pulmonary vascular bed [51]. All chest radiographs were interpreted by a single experienced critical care physician to reduce variability in interpretation of chest radiographs [35]. The apparent insensitivity of the consensus ARDS definition may be improved with consideration of less severe forms of lung injury, although this is operationally differentiated by the severity of hypoxemia rather than the discrepant factor in our study, namely evidence of pulmonary edema on chest radiograph. The finding that ARDS patients with higher EVLW have increased mortality, as well as the finding of no difference in mortality among severe sepsis patients stratified by EVLW, may be due to statistical power or inherent heterogeneity in the sepsis and ARDS patient populations (beyond such identified disparities as baseline fluid balance).
Conclusion
Lung water accumulates abnormally in a substantial fraction of severe sepsis patients without recognized respiratory complications. These subtle abnormalities of pulmonary function may represent subclinical lung injury, which are undetectable by standard techniques and current clinical definitions. Furthermore, EVLW has prognostic implications for patients with severe sepsis and ARDS, and correlates with the severity of lung injury. More importantly, EVLW is highly prognostic for critically ill patients with chronic alcohol abuse, presumably representing intrinsic altered alveolar–capillary integrity. Further investigation is required to confirm these findings and to determine the utility of EVLW as a diagnostic or prognostic marker in patients with severe sepsis.
Key messages
• The majority of severe sepsis patients have increased amounts of EVLW, including those who do not meet clinical criteria defining ARDS.
• Increased EVLW is associated with worse survival in patients with severe sepsis, whereas the minority of ARDS patients with normal amounts of EVLW have greater chances of survival.
• Chronic alcohol abuse is associated with increased quantities of EVLW, presumably reflecting inherent alveolar–capillary barrier dysfunction.
• Measurements of EVLW may serve to risk stratify severe sepsis patients and to improve patient management.
Abbreviations
AECC = American–European Consensus Conference; ANOVA = analysis of variance; APACHE = Acute Physiology and Chronic Health Evaluation; ARDS = acute respiratory distress syndrome; CO = cardiac output; EVLW = extravascular lung water; FiO2 = fractional inspired oxygen; GEDV = global end-diastolic volume; ICU = intensive care unit; IQR = interquartile range; ITBV= intrathoracic blood volume; ITTV = intrathoracic thermal volume; LIS = Lung Injury Score; PaO2 = arterial oxygen tension; PTV = pulmonary thermal volume.
Competing interests
The author(s) declare that they have no competing interests.
Authors' contributions
GM was involved in the study concept and design; collection, analysis and interpretation of the data; provision of study materials and patients; statistical expertise; obtaining funding; and drafting, revision, and approval of the manuscript. SE was involved in the collection, analysis, and interpretation of the data; provision of study materials and patients; and drafting, revision, and approval of the manuscript. MM (Mealer) was involved in the collection, analysis, and interpretation of the data; provision of study materials and patients; and approval of the manuscript. MM (Moss) was involved in study concept and design; collection, analysis, and interpretation of the data; provision of study materials and patients; statistical expertise; and drafting, revision, and approval of the manuscript.
Acknowledgements
We gratefully acknowledge the contribution and support of the patients and families requiring intensive care, Ms Leslie Rogin, RN, and Mrs Dana Johnson, without whom this project would not have been possible.
Support was provided by the US National Institutes of Health (Dr Martin: HL K23-67739; Dr Moss: AA R01-11660) and the Oak Ridge Associated Universities (Ralph E Powe Award to Dr Martin).
Figures and Tables
Figure 1 Longitudinal measures of (a) extravascular lung water (EVLW) and (b) oxygenation (arterial oxygen tension [PaO2]/inspired fractional oxygen [FiO2] ratio) in patients with severe sepsis, stratified by survival. Vertical bars indicate standard errors. *Significant between-group differences at the marked time points; P < 0.001 for EVLW differences over time, by analysis of variance.
Figure 2 Scatter plot showing the relationship between (a) oxygenation (arterial oxygen tension [PaO2]/inspired fractional oxygen [FiO2] ratio) and extravascular lung water (EVLW) in all patients (R2 by linear regression = 0.27; P < 0.001), and (b) between minimum PaO2/FiO2 ratio and maximum EVLW in nonsurvivors (R2 = 0.60; P = 0.005).
Table 1 Baseline demographic characteristics and outcomes for all patients with severe sepsis
ARDS Non-ARDS Total
Patients (n) 15 14 29
Baseline characteristics
Age (years) 44 ± 12 57 ± 14 51 ± 15*
Sex (male) 5/15 5/14 10/29
Race (African American/Caucasian/other) 12/3/0 12/0/2 24/3/2
Chronic heart disease 1/15 2/14 3/29
Chronic obstructive lung disease 1/15 1/15 2/29
Chronic renal disease (serum creatinine ≥ 2.0 mg/dl) 1/15 3/14 4/29
HIV infection 7/15 4/14 11/29
Chronic alcohol abuse 5/15 8/14 13/29
Bacteremia 7/15 5/14 12/29
APACHE II score (day 1) 23.9 ± 6.5 27.7 ± 8.3 25.7 ± 7.5
Mechanically ventilated 14/15 11/14 25/29
Physiology at enrollment
White blood cell count (cells × 103/ml) 20.4 (10.8–30.0) 13.6 (3.8–23.4) 17.2 (8.9–25.5)
Prior 24-hour fluid balance (ml) 2506 ± 2583 4993 ± 3625 3523 ± 3223
Shock (vasopressor requirement) 12/15 10/14 22/29
Cardiac index (l/min per m2) 3.8 ± 1.0 4.3 ± 1.4 4.0 ± 1.2
GEDV index (ml/m2) 648 ± 184 719 ± 119 681 ± 159
SVRI (dyn·s/cm5 per m2) 1408 ± 441 1668 ± 669 1528 ± 562
EVLW index (ml/kg) 15.0 (9.0–16.7) 7.0 (5.0–8.2) 8.5 (5.1–15.8)*
Outcome
ICU length of stay (days) 14.0 (8.5–21.0) 10.0 (4.8–16.3) 13.0 (7.0–17.0)
Hospital length of stay (days) 26.0 (12.3–42.3) 19.0 (16.0–31.0) 19.0 (13.0–36.0)
28-day mortality 7/15 5/14 12/29
Values are expressed as mean ± standard deviation, frequency (%), or median (interquartile range), unless otherwise noted.
* P for difference between groups < 0.05. APACHE, Acute Physiology and Chronic Health Evaluation; ARDS, acute respiratory distress syndrome; EVLW, extravascular lung water; GEDV, global end-diastolic volume; ICU, intensive care unit; SVRI, systemic vascular resistance index.
Table 2 Demographics, baseline patient characteristics and outcomes of ARDS and non-ARDS patients stratified by lung water content
Non-ARDS ARDS
Low EVLW (n = 6) High EVLW (n = 8) P Low EVLW (n = 4) High EVLW (n = 11) P
EVLW (ml/kg) 5.0 (4.8–8.1) 7.7 (5.0–8.7) NS 9.0 (8.0–9.8) 16.0 (13.0–20.3) 0.07
Source (n [% pulmonary]) 2/6 (33%) 3/8 (38%) NS 3/4 (75%) 8/11 (73%) NS
APACHE II score 27.8 ± 2.3 27.6 ± 11.6 NS 18.5 ± 3.3 25.9 ± 6.3 0.05
PaO2/FiO2 index (mmHg) 424.0 ± 51.7 244.3 ± 133.9 0.02 91.5 ± 24.6 144.9 ± 89.9 NS
LIS 0.8 ± 0.2 0.9 ± 0.9 NS 2.6 ± 1.7 2.7 ± 1.1 NS
CXR score 0.5 (0–1) 1 (0–2) NS 3 (2–4) 4 (3–4) NS
GEDV index (ml/m2) 750.5 ± 133.6 691.3 ± 108.1 NS 581.5 ± 131.4 671.8 ± 200.0 NS
Fluid balance (prior 24 hours; ml) 5256 ± 4918 4664 ± 1591 NS 2373 ± 776 2545 ± 2959 NS
Chronic alcohol abuse (n [%]) 4/6 (67%) 4/8 (50%) NS 0/4 (0%) 5/11 (45%) 0.09
28-Day mortality (n [%]) 1/6 (17%) 4/8 (50%) 0.20 0/4 (0%) 7/11 (64%) 0.03
Extravascular lung water (EVLW), Acute Physiology and Chronic Health Evaluation (APACHE) II, arterial oxygen tension (PaO2)/fractional inspired oxygen (FiO2) ratio, Lung Injury Score (LIS), and chest radiographic (CXR) score are initial values calculated from study day 1. Values are expressed as means ± standard deviation or as median (interquartile range). ARDS, acute respiratory distress syndrome.
Table 3 Demographics, baseline patient characteristics and outcomes of ARDS and non-ARDS patients stratified by history of chronic alcohol abuse
Non-ARDS ARDS
Chronic alcohol abuse (n = 8) Nonalcoholic (n = 6) P Chronic alcohol abuse (n = 8) Nonalcoholic (n = 6) P
EVLW (ml/kg) 7.0 (5.0–8.0) 6.5 (4.8–8.8) NS 20.3 (16.0–26.7) 9.5 (7.1–15.4) 0.009
Source (n [% pulmonary]) 2/8 (25%) 1/6 (17%) NS 5/5 (100%) 7/10 (70%) 0.17
APACHE II score 28.7 ± 9.3 26.5 ± 7.7 NS 27.0 ± 6.7 22.4 ± 6.1 NS
PaO2/FiO2 index (mmHg) 369.7 ± 140.8 215.7 ± 116.7 NS 96.7 ± 21.7 148.0 ± 95.1 NS
LIS 0.9 ± 0.6 0.8 ± 0.7 NS 3.2 ± 0.5 2.4 ± 1.4 NS
CXR score 0.5 (0.0–1.3) 1.0 (0.5–1.5) NS 4.0 (2.5–4.0) 3.5 (2.5–4.0) NS
GEDV index (ml/m2) 750.3 ± 142.7 681.7 ± 81.6 NS 714.7 ± 149.7 614.2 ± 197.9 NS
Fluid balance (prior 24 hours; ml) 4703 ± 3310 6004 ± 6064 NS 4758 ± 3240 1504 ± 1550 0.03
28-Day mortality (n [%]) 4/8 (50%) 1/6 (17%) 0.19 3/5 (60%) 4/10 (40%) NS
Extravascular lung water (EVLW), Acute Physiology and Chronic Health Evaluation (APACHE) II, arterial oxygen tension (PaO2)/fractional inspired oxygen (FiO2) ratio, Lung Injury Score (LIS), and chest radiographic (CXR) score are initial values calculated from study day 1. Values are expressed as means ± standard deviation or as median (interquartile range). ARDS, acute respiratory distress syndrome.
==== Refs
Martin GS Mannino DM Eaton S Moss M The epidemiology of sepsis in the United States from 1979 through 2000 N Engl J Med 2003 348 1546 1554 12700374 10.1056/NEJMoa022139
Angus DC Linde-Zwirble WT Lidicker J Clermont G Carcillo J Pinsky MR Epidemiology of severe sepsis in the United States: analysis of incidence, outcome, and associated costs of care Crit Care Med 2001 29 1303 1310 11445675 10.1097/00003246-200107000-00002
Bone RC Balk RA Cerra FB Dellinger RP Fein AM Knaus WA Schein RM Sibbald WJ Definitions for sepsis and organ failure and guidelines for the use of innovative therapies in sepsis. The ACCP/SCCM Consensus Conference Committee. American College of Chest Physicians/Society of Critical Care Medicine Chest 1992 101 1644 1655 1303622
Levy MM Fink MP Marshall JC Abraham E Angus D Cook D Cohen J Opal SM Vincent JL Ramsay G 2001 SCCM/ESICM/ACCP/ATS/SIS International Sepsis Definitions Conference Crit Care Med 2003 31 1250 1256 12682500 10.1097/01.CCM.0000050454.01978.3B
Wheeler AP Bernard GR Treating patients with severe sepsis N Engl J Med 1999 340 207 214 9895401 10.1056/NEJM199901213400307
Parrillo JE Parker MM Natanson C Suffredini AF Danner RL Cunnion RE Ognibene FP Septic shock in humans. Advances in the understanding of pathogenesis, cardiovascular dysfunction, and therapy Ann Intern Med 1990 113 227 242 2197912
Arias E Anderson RN Kung HC Murphy SL Kochanek KD Deaths: final data for 2001 Natl Vital Stat Rep 2003 52 1 115 14570230
Martin GS Bernard GR Airway and lung dysfunction in sepsis Intensive Care Med 2001 S63 S79 11307371
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. Definitions, mechanisms, relevant outcomes, and clinical trial coordination Am J Respir Crit Care Med 1994 149 818 824 7509706
Moss M Goodman PL Heinig M Barkin S Ackerson L Parsons PE Establishing the relative accuracy of three new definitions of the adult respiratory distress syndrome Crit Care Med 1995 23 1629 1637 7587227 10.1097/00003246-199510000-00006
Moss M Bucher B Moore FA Moore EE Parsons PE The role of chronic alcohol abuse in the development of acute respiratory distress syndrome in adults JAMA 1996 275 50 54 8531287 10.1001/jama.275.1.50
Moss M Steinberg KP Guidot DM Duhon GF Treece P Wolken R Hudson LD Parsons PE The effect of chronic alcohol abuse on the incidence of ARDS and the severity of the multiple organ dysfunction syndrome in adults with septic shock Chest 1999 116 97S 98S 10424617 10.1378/chest.116.suppl_1.97S
Guidot DM Modelska K Lois M Jain L Moss IM Pittet JF Brown LA Ethanol ingestion via glutathione depletion impairs alveolar epithelial barrier function in rats Am J Physiol Lung Cell Mol Physiol 2000 279 L127 L135 10893211
Burnham EL Brown LA Halls L Moss M Effects of chronic alcohol abuse on alveolar epithelial barrier function and glutathione homeostasis Alcohol Clin Exp Res 2003 27 1167 1172 12878924 10.1097/01.ALC.0000075821.34270.98
Bernard GR Vincent JL Laterre PF LaRosa SP Dhainaut JF Lopez-Rodriguez A Steingrub JS Garber GE Helterbrand JD Ely EW Efficacy and safety of recombinant human activated protein C for severe sepsis N Engl J Med 2001 344 699 709 11236773 10.1056/NEJM200103083441001
Knaus WA Draper EA Wagner DP Zimmerman JE APACHE II: a severity of disease classification system Crit Care Med 1985 13 818 829 3928249
Goedje O Hoeke K Lichtwarck-Aschoff M Faltchauser A Lamm P Reichart B Continuous cardiac output by femoral arterial thermodilution calibrated pulse contour analysis: comparison with pulmonary arterial thermodilution Crit Care Med 1999 27 2407 2412 10579256 10.1097/00003246-199911000-00014
Sakka SG Reinhart K Meier-Hellmann A Comparison of pulmonary artery and arterial thermodilution cardiac output in critically ill patients Intensive Care Med 1999 25 843 846 10447543 10.1007/s001340050962
Lichtwarck-Aschoff M Zeravik J Pfeiffer UJ Intrathoracic blood volume accurately reflects circulatory volume status in critically ill patients with mechanical ventilation Intensive Care Med 1992 18 142 147 1644961
Godje O Peyerl M Seebauer T Dewald O Reichart B Reproducibility of double indicator dilution measurements of intrathoracic blood volume compartments, extravascular lung water, and liver function Chest 1998 113 1070 1077 9554649
Sakka SG Reinhart K Wegscheider K Meier-Hellmann A Is the placement of a pulmonary artery catheter still justified solely for the measurement of cardiac output? J Cardiothorac Vasc Anesth 2000 14 119 124 10794326 10.1016/S1053-0770(00)90002-8
Sakka SG Ruhl CC Pfeiffer UJ Beale R McLuckie A Reinhart K Meier-Hellmann A Assessment of cardiac preload and extravascular lung water by single transpulmonary thermodilution Intensive Care Med 2000 26 180 187 10784306 10.1007/s001340050043
Holm C Tegeler J Mayr M Pfeiffer U Henckel vD Muhlbauer W Effect of crystalloid resuscitation and inhalation injury on extravascular lung water: clinical implications Chest 2002 121 1956 1962 12065363 10.1378/chest.121.6.1956
Newman EV Merrell M Genecin A Monge C Milnor WR McKeever WP The dye dilution method for describing the central circulation Circulation 1951 4 735 746 14870285
Neumann P Extravascular lung water and intrathoracic blood volume: double versus single indicator dilution technique Intensive Care Med 1999 25 216 219 10193551 10.1007/s001340050819
Elings VB Lewis FR A single indicator technique to estimate extravascular lung water J Surg Res 1982 33 375 385 6752580 10.1016/0022-4804(82)90052-X
Risberg B Osburn K Pilgreen K Wax SD Webb WR Lung thermal volume as an indicator of pulmonary extravascular water Eur Surg Res 1982 14 245 251 7117330
Sturm JA Lewis FR, Pfeiffer UJ Development and significance of lung water measurement in clinical and experimental practice Practical Applications of Fiberoptics in Critical Care Monitoring 1990 Berlin, Germany: Springer-Verlag 129 139
Pfeiffer UJ Backus G Blumel G Eckart J Muller P Winkler P Zeravik J Zimmermann GJ Lewis FR, Pfeiffer UJ A fiberoptics-based system for integrated monitoring of cardiac output, intrathoracic blood volume, extravascular lung water, O2 saturation, and a-v differences Practical Applications of Fiberoptics in Critical Care Monitoring 1990 Berlin, Germany: Springer-Verlag 114 125
Elings VB Lewis FR A single indicator technique to estimate extravascular lung water J Surg Res 1982 33 375 385 6752580 10.1016/0022-4804(82)90052-X
Mihm FG Feeley TW Rosenthal MH Lewis F Measurement of extravascular lung water in dogs using the thermal-green dye indicator dilution method Anesthesiology 1982 57 116 122 7046519
Murray JF Matthay MA Luce JM Flick MR An expanded definition of the adult respiratory distress syndrome Am Rev Respir Dis 1988 138 720 723 3202424
Honore PM Jacquet LM Beale RJ Renauld JC Valadi D Noirhomme P Goenen M Effects of normothermia versus hypothermia on extravascular lung water and serum cytokines during cardiopulmonary bypass: a randomized, controlled trial Crit Care Med 2001 29 1903 1909 11588449 10.1097/00003246-200110000-00009
Selzer ML Vinokur A van Rooijen L A self-administered Short Michigan Alcoholism Screening Test (SMAST) J Stud Alcohol 1975 36 117 126 238068
Rubenfeld GD Caldwell E Granton J Hudson LD Matthay MA Interobserver variability in applying a radiographic definition for ARDS Chest 1999 116 1347 1353 10559098 10.1378/chest.116.5.1347
Baudendistel L Shields JB Kaminski DL Comparison of double indicator thermodilution measurements of extravascular lung water (EVLW) with radiographic estimation of lung water in trauma patients J Trauma 1982 22 983 988 7143511
Halperin BD Feeley TW Mihm FG Chiles C Guthaner DF Blank NE Evaluation of the portable chest roentgenogram for quantitating extravascular lung water in critically ill adults Chest 1985 88 649 652 3902385
Sakka SG Klein M Reinhart K Meier-Hellmann A Prognostic value of extravascular lung water in critically ill patients Chest 2002 122 2080 2086 12475851 10.1378/chest.122.6.2080
Davey-Quinn A Gedney JA Whiteley SM Bellamy MC Extravascular lung water and acute respiratory distress syndrome: oxygenation and outcome Anaesth Intensive Care 1999 27 357 362 10470388
Eisenberg PR Hansbrough JR Anderson D Schuster DP A prospective study of lung water measurements during patient management in an intensive care unit Am Rev Respir Dis 1987 136 662 668 3307570
Brigham KL Kariman K Harris TR Snapper JR Bernard GR Young SL Correlation of oxygenation with vascular permeability-surface area but not with lung water in humans with acute respiratory failure and pulmonary edema J Clin Invest 1983 72 339 349 6874950
Sivak ED Richmond BJ O'Donavan PB Borkowski GP Value of extravascular lung water measurement vs portable chest x-ray in the management of pulmonary edema Crit Care Med 1983 11 498 501 6345087
Lewis FR Elings VB Sturm JA Bedside measurement of lung water J Surg Res 1979 27 250 261 384088 10.1016/0022-4804(79)90138-0
Lewis FR Elings VB Hill SL Christensen JM The measurement of extravascular lung water by thermal-green dye indicator dilution Ann N Y Acad Sci 1982 384 394 410 7046565
Mihm FG Feeley TW Jamieson SW Thermal dye double indicator dilution measurement of lung water in man: comparison with gravimetric measurements Thorax 1987 42 72 76 3616974
Velazquez M Haller J Amundsen T Schuster DP Regional lung water measurements with PET: accuracy, reproducibility, and linearity J Nucl Med 1991 32 719 725 2013812
Staub NC Clinical use of lung water measurements. Report of a workshop Chest 1986 90 588 594 3530651
Nuckton TJ Alonso JA Kallet RH Daniel BM Pittet JF Eisner MD Matthay MA Pulmonary dead-space fraction as a risk factor for death in the acute respiratory distress syndrome N Engl J Med 2002 346 1281 1286 11973365 10.1056/NEJMoa012835
Moss M Guidot DM Wong-Lambertina M Ten Hoor T Perez RL Brown LA The effects of chronic alcohol abuse on pulmonary glutathione homeostasis Am J Respir Crit Care Med 2000 161 414 419 10673179
Elings VB Lewis FR A single indicator technique to estimate extravascular lung water J Surg Res 1982 33 375 385 6752580 10.1016/0022-4804(82)90052-X
Matthay MA Clinical measurement of pulmonary edema Chest 2002 122 1877 1879 12475816 10.1378/chest.122.6.1877
| 15774053 | PMC1175916 | CC BY | 2021-01-04 16:04:52 | no | Crit Care. 2005 Jan 11; 9(2):R74-R82 | utf-8 | Crit Care | 2,005 | 10.1186/cc3025 | oa_comm |
==== Front
Crit CareCritical Care1364-85351466-609XBioMed Central London cc30261577405610.1186/cc3026ResearchPatients' recollections of experiences in the intensive care unit may affect their quality of life Granja Cristina [email protected] Alice [email protected] Sara [email protected] Claudia [email protected] Altamiro [email protected] António [email protected] the JMIP Study Group 1 Intensivist, Consultant in Anesthesiology, Medical Intensive Care Unit, Hospital Pedro Hispano, Matosinhos, Portugal2 Consultant in Psychiatry, Department of Psychiatry, Hospital Geral de Santo Antonio, Oporto, Portugal3 Research Assistant, Department of Biostatistics and Medical Informatics, Faculty of Medicine, University of Oporto, Oporto, Portugal4 Professor and Head of Department, Department of Biostatistics and Medical Informatics, Faculty of Medicine University of Oporto, Oporto, Portugal5 Consultant in Internal Medicine, Head of Department of Intensive Care, Intensive Care Unit, Hospital Geral de Santo António, Oporto, Portugal2005 31 1 2005 9 2 R96 R109 3 8 2004 16 9 2004 22 11 2004 24 11 2004 Copyright © 2005 Granja 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
We wished to obtain the experiences felt by patients during their ICU stay using an original questionnaire and to correlate the memories of those experiences with health-related quality of life (HR-QOL).
Methods
We conducted a prospective study in 10 Portuguese intensive care units (ICUs). Six months after ICU discharge, an original questionnaire on experiences of patients during their ICU stay, the recollection questionnaire, was delivered. HR-QOL was evaluated simultaneously, with the EQ-5D questionnaire. Between 1 September 2002 and 31 March 2003 1433 adult patients were admitted. ICU and hospital mortalities were 21% and 28%, respectively. Six months after ICU discharge, 464 patients completed the recollection questionnaire.
Results
Thirty-eight percent of the patients stated they did not remember any moment of their ICU stay. The ICU environment was described as friendly and calm by 93% of the patients. Sleep was described as being good and enough by 73%. The experiences reported as being more stressful were tracheal tube aspiration (81%), nose tube (75%), family worries (71%) and pain (64%). Of respondents, 51% experienced dreams and nightmares during their ICU stay; of these, 14% stated that those dreams and nightmares disturb their present daily life and they exhibit a worse HR-QOL. Forty-one percent of patients reported current sleep disturbances, 38% difficulties in concentrating in current daily activities and 36% difficulties in remembering recent events. More than half of the patients reported more fatigue than before the ICU stay. Multiple and linear regression analysis showed that older age, longer ICU stay, higher Simplified Acute Physiology Score II, non-scheduled surgery and multiple trauma diagnostic categories, present sleep disturbances, daily disturbances by dreams and nightmares, difficulties in concentrating and difficulties in remembering recent events were independent predictors of worse HR-QOL. Multicollinearity analysis showed that, with the exception of the correlation between admission diagnostic categories and length of ICU stay (0.47), all other correlations between the independent variables and coefficient estimates included in the regression models were weak (below 0.30).
Conclusion
This study suggests that neuropsychological consequences of critical illness, in particular the recollection of ICU experiences, may influence subsequent HR-QOL.
critical illnessfollow-uphealth-related quality of lifeintensive careneuropsychological sequelaeoutcome
==== Body
Introduction
Patients admitted to an intensive care unit (ICU) generally present an unexpected life-threatening condition, with the exception of those admitted after scheduled surgery. These patients will remain in their critical condition for various lengths of time and will need several types of life support, such as ventilation, cardiovascular or renal support. They will also receive various types of sedatives and analgesics to ensure compliance with ventilation and to induce some comfort. As the event that takes these critical patients to the ICU was unexpected, most patients will not be aware of their condition until late in their ICU stay and some of them only after their discharge to the ward. However, during their ICU stay they continue to have an emotional life, in a mixture of dreams, delusions and emotional experiences related to real events.
Although various degrees of anxiety or depression that might delay and impair their recovery have been described in critical illness survivors [1-4], little is known about this and other neuropsychological sequelae of critical illness; cognitive impairment and memory disturbances are those more frequently described [1-5]. Post-traumatic stress disorder (PTSD) [6] and PTSD-related symptoms (PTSS) [2] have also been described as possible events occurring after critical illness.
Although functional sequelae seem to depend more on previous health state and on the existence of co-morbidities and on the aggressiveness suffered during the critical illness period, neuropsychological sequelae depend not only on the aggressiveness of the acute event but also on the ability of patients to deal with the memories they retain from that period [1-3]. These memories may be of two kinds: factual memories and delusional memories, which include nightmares, hallucinations, paranoid delusions and dreams [2]. Recall of delusional memories but not of factual memories has been associated with the development of acute PTSS [2].
Several studies have sought to identify factors that can function as stressors during an ICU stay, with the aim of preventing or decreasing them [7-10].
This study has two aims: to recollect the experiences felt by patients during their ICU stay, by using an original questionnaire, and to correlate the memories of those experiences with health-related quality of life (HR-QOL).
Methods
This study is part of a multicentre study on the quality of life after intensive care, involving 10 Portuguese ICUs; these are listed in Additional file 1 and have been named the Jornadas de Medicina Intensiva da Primavera (JMIP) Study Group.
Patients
The study addressed all adult patients (aged 18 years or more) admitted to the 10 ICUs. Background variables included patient's gender, age, main activity and previous health state. On the basis of individual clinical registries and on direct questioning from patients for whom a follow-up consultation was continuing, the previous health state was evaluated according to three categories: healthy, chronic non-disabling diseases (that is, able to perform work or normal daily activities) and chronic disabling diseases (that is, unable to work or to undertake normal daily activities). Each participating physician in each ICU classified all patients into one of these three categories. ICU variables included the severity of disease at admission as evaluated by Simplified Acute Physiology Score II (SAPS II), the length of stay and the admission diagnostic category (medical, scheduled surgery, non-scheduled surgery or multiple trauma).
Methods
The first author developed an original questionnaire to recollect experiences lived by survivors of critical illness, which was called the recollection questionnaire (see Additional file 2) and was based on previous personal experience with an ICU follow-up clinic [11-14] and previous studies on this subject [2,7,8]. The questions were extensively applied over several years by the first author and changes were made over time to achieve the best possible understanding from the patient about each proposed question. The questionnaire was therefore developed after a succession of small pilot and qualitative studies.
The recollection questionnaire comprises 14 questions relating to memories retained by the patients, the environment in the ICU, the relationship with health care professionals, dreams, nightmares, sleep disturbances, difficulties in concentrating and in remembering recent events, fatigue and being able to return to their previous level of activity. Direct questions on memories were made either on real experiences of patients in the ICU or on dreams and nightmares experienced by them. There was no formal division between factual memories and delusional memories. Hallucinations or paranoid delusions were not specifically looked for. One of the questions (number 11) comprises 25 items related to the recollection of experiences lived in the ICU, such as tracheal suctioning, needle punctures, pain, sleep, and dependence on the ventilator. These items can be classified in one of five categories: 0 ('I don't remember'), 1 ('It was not hard'), 2 ('It was indifferent'), 3 ('It was hard but necessary'), 4 ('It was very hard'), and 5 ('It was awful').
HR-QOL was measured with a generic questionnaire (EQ-5D) [15,16] and a specific critical care questionnaire [17]. For the purpose of this study, only data of the generic questionnaire will be reported. EQ-5D is a generic instrument designed to measure health outcome that was developed at the European level [15,16]. The EuroQol Group originally developed the Portuguese version of EQ-5D in 1998 (EuroQol Group Newsletter, January 2000). EQ-5D was applied as reported previously [11].
At 6 months after discharge from ICU, all recollection questionnaires were sent by mail. For practical reasons all patients completed their questionnaires at home. In five ICUs questionnaires were returned by mail and in the other five they were returned directly by hand when patients came to the follow-up consultation.
Informed consent was obtained from all patients at the time of the follow-up consultation, where applicable. Also, because questionnaires were sent by mail, a letter containing detailed information on the aims of the study accompanied them. Thus, because consent was implicit in answering the questionnaire, the need for additional informed consent was waived. A hospital Ethics Committee approved this observational study.
Descriptive analyses of background variables (gender, age, main activity and previous health state), ICU variables (SAPS II, length of ICU stay and admission diagnostic category) and questionnaire variables were presented. Categorical variables were described as absolute frequencies (n) and relative frequencies (%); median and centiles were used for continuous variables. The Pearson test, linear-by-linear test and Mann–Whitney test were used for comparisons.
Multiple logistic regression was performed with the five dimensions of the EQ-5D questionnaire as dependent variables (categorised as not having problems or having problems) and background, ICU and recollection questionnaire variables as independent variables. The stepwise Forward method was used with an entry criterion of P < 0.05 and a removal criterion of P < 0.1. To analyse possible multicollinearity between the variables studied, Spearman correlation coefficients between the variables and regression coefficient estimates correlation matrices were analysed.
Differences were considered statistically significant at P < 0.05. SPSS® 12.0 was used for statistical analysis.
Results
Between 1 September 2002 and 31 March 2003 there were 1433 admissions. Nineteen patients were excluded because they were less than 18 years old. Two hundred and ninety-seven (21%) died in the ICU and a further 95 patients died in the ward (28% in-hospital mortality rate). At 6 months, six patients were still in the hospital. One hundred and five patients died after hospital discharge but before the evaluation at 6 months, at which point there were 911 survivors, 445 (49%) of them being non-respondents. Four hundred and sixty-four patients completed the recollection questionnaire (Fig. 1).
There were no differences between respondents and non-respondents in background and ICU variables, except for admission diagnostic category, for which non-scheduled surgery and multiple trauma survivors answered significantly less (Table 1).
Background variables
Of the 464 respondents included in the study, 61% were male, the median age was 58 years, 49% were retired, 39% were previously healthy, 44% had previous chronic non-disabling disease and 17% had previous chronic disabling disease (Table 1).
ICU variables
The median SAPS II on admission was 31, the median length of ICU stay was 4 days and admission diagnostic categories in ICU included medical reasons in 46% of the patients, scheduled surgery in 32%, non-scheduled surgery in 13% and multiple trauma in 9% (Table 1).
With the exception of gender and length of ICU stay, which exhibited non-significant differences, there was significant variability between the 10 ICUs: the minimum median age was 44 years and the maximum was 68 (P = 0.016), those reporting their main activity as employed varied between 12% and 50% (P = 0.011), previous health state varied between 6% and 66% previously healthy (P = 0.001), median SAPS II exhibited a minimum of 26 and a maximum of 39 (P = 0.004), and diagnostic categories varied for medical admissions between 25% and 71% (P < 0.001; Table 2).
Recollection questionnaire variables
There was also significant variability between the 10 ICUs in the answers to the recollection questionnaire, as follows: of the 464 respondents, 23% stated that they had amnesia about hospital admission (range 6–42%), and 45% stated that they had amnesia about ICU admission (range 21–68%). Moreover, when asked about remembering some moment during their ICU stay (question 3), 38% (range 20–55%) stated that they had amnesia about the whole ICU stay. For purposes of data analysis these 38% of patients will be assumed to be those who had amnesia about the ICU stay (Table 3).
Of those who remembered (n = 236; question 3), the ICU environment was described as friendly and calm by 93% (range 63–100%) of the patients. Confidence in ICU physicians and ICU nurses was described as being excellent by 94% (range 82–100%) and good by 96% (range 81–100%) of the patients. Sleep in the ICU was described as excessive by 11% (range 0–20%) of the patients, enough and restoring by 62% (range 31–82%) of the patients, and insufficient by 27% (range 0–56%) (Table 3).
When asked about their own perception of their quality of life, 40% (range 10–82%) considered that it had improved, 31% (range 0–84%) that it remained the same, 20% (range 0–31%) that it worsened, 1% (range 0–6%) would have preferred to die and 8% (range 0–19%) did not know how to answer (Table 3). Patients who considered that they had improved or remained the same as before the ICU stay exhibited significantly fewer problems in all dimensions of the EQ-5D, and a significantly higher EQ-VAS and EQ Index (data not shown).
Eighty percent of patients had never before been admitted to an ICU. Being previously admitted to an ICU was significantly associated with being retired, previous chronic disease, medical diagnostic categories, and a report of problems in the anxiety/depression dimension (data not shown).
Concerning the 25 items in question 11 (see Additional file 2), where patients were asked to classify experiences according to the degree of stress provoked, to simplify the analysis we combined those items classified as 1 and 2 as being not stressful and those classified as 3, 4 and 5 as being stressful. Table 4 shows the recollection of experiences reported as being more stressful (that is, difficult to endure): tracheal tube aspiration (81%), nose tube (75%), family worries (71%), pain (64%), immobilisation in bed (64%), fear of dying or uncertainty about the future (64%), daily needle punctures (61%), difficulties in communication (59%), machine (ventilator) dependence (58%), general discomfort (58%), bladder tube (56%) and noisy and non-sleeping nights (54%).
Comparing background, ICU and EQ-5D variables between those who remembered some moment in the ICU (62%) and those with amnesia (38%), we found that those remembering some moment in the ICU exhibited significantly fewer problems in the mobility, self-care and usual activities dimensions, had significantly higher EQ-VAS and EQ Index and stated themselves to be better in a significantly higher percentage, although those who exhibited amnesia were also significantly more severely ill and stayed significantly longer in the ICU (data not shown).
Fifty-four percent of patients who were not retired were unable to return to their previous level of activity, and 51% of those who were retired were also not able to return to their previous level of activity (Table 3).
From all respondents, 41% experienced dreams and 30% experienced nightmares during their ICU stay (Table 3). Combining the patients with these experiences, we found no significant differences between background and ICU variables in those who did not experience dreams and nightmares, but those who experienced dreams and nightmares reported significantly more problems in the pain/discomfort and anxiety/depression dimensions (data not shown). Fourteen percent (n = 23) of these respondents stated that those dreams and nightmares disturb their current daily life (that is, at 6 months after ICU discharge). Although not exhibiting statistically significant differences in the background and ICU variables, they reported significantly more moderate to extreme problems in the pain/discomfort dimension (91% versus 55%) and in the anxiety/depression dimension (77% versus 51%). They also exhibited a statistically significantly lower EQ-VAS and EQ Index (Table 5).
Forty-one percent of the patients reported current sleep disturbances (Table 3). Sleep disturbances were significantly associated with female gender, older age, being retired and a worse HR-QOL in all the dimensions of the EQ-5D, including a significantly worse EQ-VAS and EQ Index (data not shown).
Thirty-eight percent of patients reported difficulties with concentrating in present daily activities (Table 3), and these were significantly associated with being retired and a worse HR-QOL in all dimensions of the EQ-5D, including EQ-VAS and EQ Index (data not shown).
Thirty-six percent of patients reported difficulties in remembering recent events (Table 3), and these were significantly associated with being retired, severity of disease at ICU admission and a worse HR-QOL in all dimensions of the EQ-5D including EQ-VAS and EQ Index (data not shown).
Fifty-seven percent of patients reported more fatigue at 6 months than before the ICU stay (Table 3), and these exhibited a significantly worse HR-QOL in all dimensions of the EQ-5D, including a significantly worse EQ-VAS and EQ Index, although there were no significant differences in the background and ICU variables (data not shown). Fatigue was significantly associated with the ability to return to their previous level of activity. These patients exhibited a significantly small rate of return to their previous level of activity, both those who were employed and even those who were retired (data not shown).
With multiple logistic regression analysis we found that older age, longer ICU stay, higher SAPS II, non-scheduled surgery and trauma admission diagnostic categories were, as expected, independent predictors of the report of problems in the dimensions of the EQ-5D (Table 6). It was also found that current sleep disturbances, current dreams and nightmares that disturb daily life, difficulties in concentrating and difficulties in remembering recent events were all independent predictors of the report of problems in the dimensions of the EQ-5D (Table 6).
Multiple linear regression analysis of EQ-VAS and EQ Index showed that older age, higher SAPS II, having dreams and nightmares that disturb daily life, difficulties in concentrating and difficulties in remembering recent events were significantly associated with a lower EQ-VAS and EQ Index (data not shown).
Multicollinearity analysis showed that, with the exception of the correlation between admission diagnostic categories and length of ICU stay (0.47), all other correlations between the independent variables and coefficient estimates included in the five regression models were weak (below 0.30; data not shown).
Discussion
In this study, nearly a half of the patients did not remember the moment of their admission to the ICU, although this percentage fell to 38% when they were asked whether they remembered some moment in their ICU stay. This agrees with previous studies in which 21–30% of patients exhibited amnesia about their ICU stay [8,9]. We found that amnesia was associated with a worse HR-QOL; however, that association was no longer significant in multiple regression analysis. A previous study by Jones and colleagues [2] has suggested that memories of factual events may protect against subsequent PTSS, whereas delusional memories were associated with more anxiety/depression. Results from the present study might suggest the same protective effect of remembering the ICU stay.
Nearly half of the survivors reported dreams and nightmares during their ICU stay and a smaller percentage of these patients (14%) reported still being disturbed by them at 6 months after ICU discharge. These patients exhibited a significantly worse HR-QOL, particularly in the pain/discomfort and anxiety/depression dimensions. In addition, the report of current disturbance to their daily life by those dreams and nightmares might suggest PTSS.
Results from multiple logistic and linear regression analyses showed that current sleep disturbances, difficulties in concentrating and difficulties in remembering recent events at 6 months after ICU discharge were all significantly associated with a worse HR-QOL, indicating a common platform of neuropsychological sequelae in survivors of critical illness involving cognitive problems, memory disturbances and anxiety/depression disturbances; this finding has been described in previous studies [2,3,5,6,18,19]. Multicollinearity analysis suggested that these items might, in fact, be independent predictors of a worse HR-QOL. In a previous study in survivors of cardiac arrest from our ICU, at the follow-up evaluation we found that about half of the survivors exhibited cognitive dysfunction, including memory deficits and problems in executive functions [20], which drew our attention to the need for neurocognitive evaluation of survivors of intensive care.
Tracheal tube aspiration, nose tube, family worries and pain were the ICU experiences described as being more stressful. Neuropsychological consequences in ICU survivors have been described as being related either to environmental factors (characteristic of the ICU, which can lead to an overwhelming of sensory stimuli) or factors related to memory problems (namely delusional memories and amnesia) [2,3,5,18,19]. These findings should suggest a need not only to review our concepts of optimal analgesia and sedation but also to evolve strategies to reinforce and help maintain factual memories, such as dialogue with the patients, explanation of all procedures, maintenance of the day/night cycle, minimisation of sensory stimuli and minimisation of noise and lights. Although for some patients the noise of alarms and seeing staff around them may be reassuring, trying to make the ICU a quiet place, at least during the night, seemed to us a good strategy.
About 65% of those patients who reported more fatigue at 6 months than before their ICU stay did not return to their previous level of activity/employment. This reinforces the fact that these consequences can have an independent effect on the ability of patients to return to work, and thus have a socio-economic impact [1].
Patients' own perception of quality of life significantly correlates with all the domains of EQ-5D, a finding similar to that of Eddleston and colleagues with the Short-Form Health Survey (SF-36) [3], which indicates the usefulness of HR-QOL generic instruments on this population.
The use of specific measurement tools for cognitive disturbances, for post-traumatic stress-related symptoms and for anxiety/depression (not done in this study) would overcome some limitations regarding the identification of these specific sequelae. Kapfhammer and colleagues [21] recently published a study in which they used specific tools to look for psychiatric morbidity and its influence on the HR-QOL of survivors of acute respiratory distress syndrome. The authors established a significant association between the diagnosis of PTSD at follow-up and more unfavorable values in the most important psychosocial dimensions of SF-36.
This study presents some other limitations, as follows.
1. There was a relatively high non-response rate (49%); however, we did not find any statistically significant differences with regard to background and ICU variables between respondents and non-respondents, including previous health state and severity of illness at ICU admission. Thus, other factors might partly contribute to the non-response rate, such as a significant proportion of functional illiteracy. Furthermore, for most of the 10 ICUs, follow-up consultations were something completely new in the evaluation of patients, which might also partly contribute to the relatively high non-response rate.
2. The recollection questionnaire was not formally assessed for its face or content validity. Although the questionnaire was developed after a succession of small pilot and qualitative studies, as stated above, we acknowledge the potential limitation caused by the lack of more formal reproducibility and validity studies.
3. As the multicentre study followed a continuing study in the first author's ICU, we did not apply a standardised tool that was meanwhile developed by Jones and coworkers (Intensive Care Unit Memory tool) [22] as our study progressed. This standardised tool was subsequently applied by Capuzzo and coworkers and was recently published [23].
4. We were unable to collect information regarding either restraint protocols or sedation protocols in the different ICUs.
5. Because specific tools for the evaluation of anxiety, depression or PTSS were not used, we could not establish further findings with regard to not only their own characterisation but also their role on the neuropsychological consequences of intensive care and on their relationships with HR-QOL at 6 months after ICU discharge.
Four main findings may be drawn from this study, as follows.
1. As a multicentre study, it enabled us to understand a core of problems common to all our ICUs, which should draw our attention to specific neuropsychological sequelae from illness requiring critical care.
2. The study contributed to identifying which experiences were reported as responsible for more stress during their ICU stay, a crucial issue in trying to identify and reduce stress factors. Tracheal tube aspiration, nose tube, pain and immobilisation in bed were stressors notably common to the experiences previously described in other studies [7-10,23,24]. In addition, family worries were the third factor identified as responsible for stress in our patients. This can be explained by the traditionally strong family ties in the Portuguese culture. Pain came in fourth place in the ranking of stressors. This finding, together with the need to preserve factual memories [2,18], should encourage revision of analgesia/sedation strategies in accordance with more recent guidelines [25].
3. Amnesia about ICU stay, sleep disturbances at 6 months after ICU stay, and memory and cognitive disturbances were associated with a worse HR-QOL, indicating not only specific neuropsychological sequelae but also their influence on subsequent HR-QOL.
4. About 15% of patients reported dreams and nightmares during their ICU stay, and these patients also exhibited a worse HR-QOL at 6 months after ICU discharge, as measured by EQ-5D. Although we did not look for hallucinations or paranoid delusions, this finding is in accordance with findings of Jones and colleagues [2], linking delusional memories with the development of PTSS and a worse HR-QOL. The association between current memory disturbances, cognitive disturbances, sleep disturbances and subsequent quality of life may be one of the key messages from this study.
Conclusion
This study suggests that neuropsychological consequences of critical illness might affect subsequent HR-QOL, which should direct our attention to these consequences and encourage further research.
Key messages
• The study contributed to identifying which experiences were reported as responsible for more stress during their ICU stay: tracheal tube aspiration, nose tube, family worries, pain, immobilization in bed, and fear of dying/uncertainty in the future, were the most frequent stress factors reported by patients.
• The association between current memory disturbances, cognitive disturbances, sleep disturbances and subsequent quality of life may be one of the key messages from this study.
• This study suggests that neuropsychological consequences of critical illness might affect subsequent HR-QOL, which should direct our attention to these consequences and encourage further research.
Abbreviations
HR-QOL = health-related quality of life; ICU = intensive care unit; LOS = length of ICU stay; PTSD = post-traumatic stress disorder; PTSS = PTSD-related symptoms; SAPS = Simplified Acute Physiology Score.
Competing interests
The author(s) declare that they have no competing interests.
Authors' contributions
CG created and designed the study and was responsible for the final manuscript. AL and SM advised for the search of neuropsychological consequences in critical patients and contributed to the final interpretation of these consequences on the final manuscript. CD undertook the statistical analysis. ACP conducted the statistical analysis and wrote the final manuscript. AC contributed to the design and the coordination of the study. All authors read and approved the final manuscript.
Supplementary Material
Additional File 1
List of the ICUs participating in the JMIP Study Group.
Click here for file
Additional File 2
Table containing the recollection questionnaire.
Click here for file
Acknowledgements
We thank Luís Filipe Azevedo for his invaluable help in the revision of this manuscript. This study was published as an abstract in the supplement of December 2004 from Critical Care Medicine and will be partly presented at the 34th SCCM Congress in January 2005, in Phoenix, Arizona, USA.
Figures and Tables
Figure 1 Patients included in and excluded from the study. Survival and recollection questionnaire response rates.
Table 1 Comparison of background and intensive care unit variables between respondents and non-respondents
Variable Total (n = 909) Respondents (n = 464) Non-respondents (n = 445) P
Background data
Sex, n (%)
Male 535 (59) 281 (39) 254 (42) 0.286a
Female 374 (41) 183 (61) 254 (57)
Median age, years (P25–P75) 59 (42–70) 58 (43–69) 60 (41–72) 0.212b
Main activity, n (%)
Employed 128 (29) 128 (29) -
Retired 216 (49) 216 (49) -
Housework/student/seeking work 58 (12) 58 (12) -
Other 44 (10) 44 (10) -
Previous health state, n (%)
Healthy 371 (41) 182 (39) 189 (42) 0.228a
Chronic non-disabling disease 403 (44) 203 (44) 200 (45)
Chronic disabling disease 135 (15) 78 (17) 57 (13)
ICU variables
Median SAPS II at admission (P25–P75) 32 (22–42) 31 (22–41) 33 (22–43) 0.209b
Median days in ICU (P25–P75) 4 (2–10) 4 (2–10) 5 (2–10) 0.297b
Admission category, n (%)
Medical 417 (46) 214 (46) 203 (46) 0.011a
Scheduled surgery 247 (27) 144 (32) 103 (23)
Non-scheduled surgery 138 (15) 62 (13) 76 (17)
Multiple trauma 106 (12) 44 (9) 62 (14)
a Pearson χ2. b Mann–Whitney test. ICU, intensive care unit; SAPS, Simplified Acute Physiology Score. P25 and P75 are the 25th and 75th centiles.
Table 2 Background and intensive care unit variables from the 10 intensive care units
Variable Total (n = 464) ICU 1 (n = 39) ICU 2 (n = 74) ICU 3 (n = 38) ICU 4 (n = 66) ICU 5 (n = 44) ICU 6 (n = 32) ICU 7 (n = 59) ICU 8 (n = 16) ICU 9 (n = 54) ICU 10 (n = 42) P
Background data
Sex, n (%)
Male 281 (61) 25 (64) 46 (62) 22 (58) 37 (56) 30 (62) 19 (59) 32 (54) 10 (62) 34 (63) 26 (62) 0.995a
Female 183 (39) 14 (36) 28 (38) 16 (42) 29 (44) 14 (32) 13 (41) 27 (46) 6 (38) 20 (37) 16 (38)
Median age, years (P25–P75) 58 (43–69) 44 (32–55) 55 (44–69) 65 (56–71) 53 (39–56) 59 (41–68) 55 (28–67) 62 (48–73) 68 (54–77) 62 (47–73) 57 (43–69) 0.016b
Main activity, n (%)
Employed 128 (29) 14 (38) 19 (26) 6 (17) 28 (45) 11 (26) 8 (27) 12 (21) 2 (12) 7 (13) 21 (50) 0.011a
Retired 216 (48) 12 (32) 36 (49) 19 (54) 24 (39) 24 (56) 12 (40) 30 (53) 12 (75) 33 (65) 14 (33)
Housework/student/seeking work 58 (13) 6 (16) 9 (12) 7 (20) 5 (8) 3 (7) 7 (23) 8 (14) 2 (12) 5 (10) 6 (14)
Other 44 (10) 5 (14) 9 (12) 3 (9) 5 (8) 5 (11) 3 (10) 7 (12) 0 (0) 6 (12) 1 (2)
Previous health state, n (%)
Healthy 183 (39) 19 (49) 16 (22) 10 (26) 28 (42) 19 (43) 21 (66) 31 (52) 1 (6) 15 (28) 23 (55) <0.001a
Chronic non-disabling disease 203 (44) 13 (33) 51 (69) 23 (61) 19 (29) 12 (27) 11 (34) 20 (34) 11 (69) 32 (59) 11 (26)
Chronic disabling disease 78 (17) 7 (18) 7 (9) 5 (13) 19 (29) 13 (30) 0 (0) 8 (14) 4 (25) 7 (13) 8 (19)
ICU variables
Median SAPS II at admission (P25–P75) 31 (22–41) 31 (17–40) 30 (17–39) 31 (24–46) 26 (19–35) 31 (18–42) 38 (27–44) 30 (22–37) 28 (21–47) 31(23–42) 39 (29–52) 0.004b
Median days in ICU (P25–P75) 4 (2–10) 7 (2–11) 1 (1–3) 5 (1–9) 5 (2–10) 4 (2–8) 10 (6–17) 4 (1–11) 2 (1–6) 3 (1–9) 7 (3–12) 0.434b
Admission category, n (%)
Medical 214 (46) 16 (41) 24 (32) 27 (71) 28 (42) 19 (43) 21 (66) 26 (44) 4 (25) 31 (57) 18 (43) <0.001a
Scheduled surgery 144 (32) 3 (8) 42 (57) 10 (26) 23 (35) 18 (41) 2 (6) 17 (29) 10 (63) 10 (19) 9 (21)
Non-scheduled surgery 62 (13) 9 (23) 6 (8) 1 (3) 13 (20) 7 (16) 3 (9) 12 (20) 1 (6) 5 (9) 5 (12)
Multiple trauma 44 (9) 11 (28) 2 (3) 0 (0) 2 (3) 0 (0) 6 (19) 4 (7) 1 (6) 8 (15) 10 (24)
a Pearson χ2. b Mann–Whitney test. ICU, intensive care unit; SAPS, Simplified Acute Physiology Score. P25 and P75 are the 25th and 75th centiles.
Table 3 Results from the recollection questionnaire
Question Number (% of responses)
Do you remember your admission to the hospital?
Yes 319 (77)
Do you remember your admission to the intensive care unit (ICU)?
Yes 230 (55)
Regarding what you saw and felt during your ICU stay:
I prefer not to remember 52 (14)
I don't remember anything 143 (38)
I don't mind remembering 137 (36)
I what to remember everything 34 (9)
None of them 13 (3)
How would you describe the environment in the ICU?
Friendly and calm 189 (93)
Chaotic and terrifying 4 (2)
Hostile and tense 11 (5)
Your confidence in doctors was:
Excellent 122 (53)
Good 93 (41)
Sufficient 12 (5)
Bad 2 (1)
Your confidence in nurses was:
Excellent 118 (52)
Good 100 (44)
Sufficient 8 (3)
Bad 2 (1)
How do you classify your sleep during ICU stay?
Excessive 24 (11)
Enough and restoring 140 (62)
Insufficient 62 (27)
At 6 months after ICU stay your quality of life:
Has improved 162 (40)
Is the same 125 (31)
Is worse 81 (20)
I would prefer to have died 6 (1)
I don't know how to answer 34 (8)
Had you been previously admitted to an ICU?
Once for the same reason 33 (8)
Twice or more for the same reason 12 (3)
Once for a different reason 38 (9)
Never been admitted to an ICU before 320 (80)
If you are not retired, have you returned to your previous activity?
No 101 (54)
If not: Because of ICU stay 42(47)
If you are retired, have you returned to your previous activity?
No 91 (51)
If not: Because of ICU stay 33 (42)
Have you had many dreams during the ICU stay?
Yes 139 (41)
Did you have many nightmares during the ICU stay?
Yes 98 (30)
Currently, do you remember those dreams and nightmares?
Yes 83 (49)
Currently, do you think that those dreams and nightmares disturb your daily life?
Yes 23 (14)
Currently, do you have sleep disturbances?
Yes 153 (41)
Currently, do you have difficulties in concentrating?
Yes 139 (38)
Currently, do you have difficulties in remembering recent events?
Yes 136 (36)
Currently, do you feel more fatigue than before the ICU stay?
Yes 199 (57)
Table 4 Recollection of stressful experiences in the intensive care unit, according to the classification defined in the recollection questionnaire
Remembera
Experience n With stress, n (%) Without stress, n (%) Amnesiab n (%)
Daily needle punctures 362 114 (61) 72 (39) 176 (49)
Tracheal tube aspiration 326 113 (81) 26 (19) 187 (57)
Nose tube 343 127 (75) 42 (25) 174 (51)
Bladder tube 330 90 (56) 72 (44) 166 (50)
Noise from conversation 351 31 (17) 146 (83) 174 (50)
Noise from engines and ventilators 360 63 (32) 132 (68) 165 (46)
Pain 360 121 (64) 69 (36) 170 (47)
Bedridden 347 129 (64) 73 (36) 145 (42)
Music in the intensive care unit 339 14 (12) 100 (88) 225 (66)
Comments from doctors and nurses 351 20 (13) 128 (87) 203 (58)
Noisy and bad sleeping nights 349 83 (54) 71 (46) 195 (56)
Ventilator dependence 343 93 (58) 68 (42) 182 (53)
Dependence on doctors and nurses 347 71 (39) 110 (61) 166 (48)
Lack of privacy in hygiene 347 79 (43) 103 (57) 165 (48)
Communication difficulties 349 111 (59) 78 (41) 160 (46)
Brightness from artificial lights 348 56 (33) 116 (67) 176 (51)
Fear of being disconnected from the ventilator 321 41 (41) 58 (59) 222 (69)
General discomfort 340 98 (58) 71 (42) 171 (50)
Fear of dying, uncertain of the future 353 110 (64) 62 (36) 181 (51)
Medical round near the patient's bed 346 13 (7) 163 (93) 170 (49)
Fear of medical procedures 342 35 (20) 139 (80) 168 (49)
Losing time orientation 348 56 (37) 94 (63) 198 (57)
Family worries 352 129 (71) 53 (29) 170 (48)
Economic worries 339 59 (38) 95 (62) 185 (55)
a Refers to patients who remembered their stay in the intensive care unit.
b Refers to all respondents.
Table 5 Comparison of background, intensive care unit and EQ-5D variables between those who stated that dreams and nightmares from the intensive care unit currently disturbed their daily life and those who did not
Variable Disturbance by dreams and nightmares
Total (n = 169) No (n = 146) Yes (n = 23) P
Background data
Sex, n (%)
Male 95 (56) 82 (56) 13 (56) 0.974a
Female 754 (44) 64 (44) 10 (44)
Median age (P25–P75) 52 (41–67) 51 (40–67) 57 (45–66) 0.4342
Main activity, n (%)
Employed 57 (35) 50 (35) 7 (32) 0.654a
Retired 70 (42) 60 (42) 10 (45)
Housework/student/seeking work 20 (12) 15 (10) 5 (23)
Other 17 (11) 18 (13) 0 (0)
Previous health state, n (%)
Healthy 78 (46) 68 (47) 10 (43) 0.798a
Chronic non-disabling disease 58 (34) 48 (33) 10 (43)
Chronic disabling disease 33 (20) 30 (20) 3 (14)
ICU variables
Median SAPS II at admission (P25–P75) 31 (22–40) 31 (22–40) 26 (22–35) 0.208b
Median ICU days (P25–P75) 5 (2–11) 6 (2–11) 3 (1–7) 0.071b
Admission category, n (%)
Medical 82 (49) 73 (50) 9 (39) 0.450a
Scheduled surgery 46 (27) 38 (26) 8 (35)
Non-scheduled surgery 24 (14) 22 (15) 2 (9)
Multiple trauma 17 (10) 13 (9) 4 (17)
EQ-5D variables
Mobility, n (%)
N: I have no problems in walking about 90 (54) 83 (58) 7 (32) 0.042c
M: I have some problems in walking about 74 (45) 59 (41) 15 (68)
E: I am confined to bed 2 (1) 2 (1) 0 (0)
Self-care, n (%)
N: I have no problems with self-care 111 (67) 101 (70) 10 (45) 0.084c
M: I have some problems washing or dressing myself 41 (25) 31 (22) 10 (45)
E: I am unable to wash or dress myself 14 (8) 12 (8) 2 (10)
Usual activities, n (%)
N: I have no problems with performing my usual activities 61 (37) 56 (39) 5 (24) 0.183c
M: I have some problems with performing my usual activities 75 (46) 64 (44) 11 (52)
E: I am unable to perform my usual activities 29 (18) 24 (17) 5 (24)
Pain/discomfort, n (%)
N: I have no pain or discomfort 67 (40) 65 (45) 2 (9) <0.001c
M: I have moderate pain or discomfort 81 (49) 68 (47) 13 (59)
E: I have extreme pain or discomfort 18 (11) 11 (8) 7 (32)
Anxiety/depression, n (%)
N: I am not anxious or depressed 74 (45) 69 (49) 5 (23) 0.009c
M: I am moderately anxious or depressed 63 (38) 53 (37) 10 (45)
E: I am extremely anxious or depressed 27 (17) 20 (14) 7 (32)
Perceived current health state
Health state today compared with 12 months ago, n (%)
Better 78 (47) 70 (49) 8 (36) 0.119c
The same 51 (31) 45 (31) 6 (27)
Worse 37 (22) 29 (20) 8 (36)
Median EQ-VAS on a 100% scale (P25–P75) 65 (50–80) 70 (50–81) 50 (40–60) 0.001b
Median EQ Index (P25–P75) 67 (49–91) 72 (50–91) 45 (35–67) 0.002b
a Pearson χ2. b Mann–Whitney test. c Linear-by-linear association. ICU, intensive care unit; SAPS, Simplified Acute Physiology Score. P25 and P75 are the 25th and 75th centiles.
Table 6 Results from five regression models
Variable OR 95% CI
Mobility
Age 1.03 1.01–1.05
LOS 1.07 1.02–1.11
SAPS II 1.01 0.99–1.03
Difficulties in concentrating
No 1.00
Yes 1.79 0.965–3.33
Difficulties in remembering recent events
No 1.00
Yes 2.14 1.14–4.01
Self-care
Age 1.02 1.00–1.04
LOS 1.04 1.00–1.08
SAPS II 1.02 1.00–1.04
Dreams and nightmares disturb your daily life
No 1.00
Yes 3.32 1.09–10.08
Difficulties in concentrating
No 1.00
Yes 3.55 1.99–6.35
Usual activities
Age 1.04 1.02–1.05
LOS 1.08 1.03–1.13
Difficulties in concentrating
No 1.00
Yes 6.27 3.29–11.91
Pain/discomfort
Age 1.03 1.01–1.05
Admission category
Scheduled surgery 1.00
Non-scheduled surgery 3.90 1.51–10.08
Medical 1.72 0.93–3.16
Multiple trauma 5.57 1.83–16.91
Dreams and nightmares disturb your daily life
No 1.00
Yes 11.39 1.39–93.33
Sleep disturbances
No 1.00
Yes 2.54 1.46–4.42
Anxiety/depression
Dreams and nightmares disturb your daily life 1.00
No 4.91 1.00–23.95
Yes
Sleep disturbances
No 1.00
Yes 2.49 1.38–4.51
Difficulties in concentrating
No 1.00
Yes 2.53 1.32–4.83
Difficulties remembering recent events
No 1.00
Yes 2.58 1.37–4.86
Dependent variables were the five dimensions of the EQ-5D questionnaire. Independent variables were all background and intensive care unit variables, and questions 12, 13 and 14 from the recollection questionnaire.
CI, confidence interval; LOS, length of ICU stay; OR, odds ratio; SAPS, Simplified Acute Physiology Score.
==== Refs
Sukantarat KT Brett S Angus D, Carlet J The neuropsychological consequences of intensive care Surviving Intensive Care Update in Intensive Care and Emergency Medicine 2003 39 Heidelberg: Springer 51 61
Jones C Griffiths RD Humphris G Skirrow PM Memory, delusions, and the development of acute posttraumatic stress disorder-related symptoms after intensive care Crit Care Med 2001 29 573 580 11373423 10.1097/00003246-200103000-00019
Eddleston JM White P Guthrie E Survival, morbidity and quality of life after discharge from intensive care Crit Care Med 2000 28 2293 2299 10921555 10.1097/00003246-200007000-00018
Jackson JC Hart RP Gordon SM Shintani A Truman B May L Ely EW Six-month neuropsychological outcome of medical intensive care unit patients Crit Care Med 2003 31 1226 1234 12682497 10.1097/01.CCM.0000059996.30263.94
Hopkins RO Weaver LK Pope D Orme JF JrBigler ED Larson-Lohr V Neuropsychological sequelae and impaired health status in survivors of severe acute respiratory distress syndrome Am J Respir Crit Care Med 1999 160 50 56 10390379
Schelling G Stoll C Haller M Briegel J Manert W Hummel T Lenhart A Heyduck M Polasek J Preub U Health-related quality of life and posttraumatic stress disorder in survivors of the acute respiratory distress syndrome Crit Care Med 1998 26 651 659 9559601 10.1097/00003246-199804000-00011
Turner JS Briggs SJ Springhorn HE Potgieter PD Patient's recollection of intensive care unit experience Crit Care Med 1990 18 966 968 2394120
Pennock BE Crawshaw Maher T Price T Kaplan PD Distressful events in the ICU as perceived by patients recovering from coronary artery bypass surgery Heart Lung 1994 23 323 327 7960858
Rotondi AJ Chelluri L Sirio C Mendelsohn A Schulz R Belle S Im K Doahoe M Pinsky MR Patient's recollection of stressful experiences while receiving prolonged mechanical ventilation in an intensive care unit Crit Care Med 2002 30 746 752 11940739 10.1097/00003246-200204000-00004
Fonte Pinto Novaes MA Knobel E Bork AM Nogueira-Martins LA Bosi Ferraz M Stressors in ICU: perception of the patient, relatives and health care team Intensive Care Med 1999 25 1421 1426 10660851 10.1007/s001340051091
Granja C Teixeira-Pinto A Costa-Pereira A Quality of life after intensive care – evaluation with EQ-5D Intensive Care Med 2002 28 898 907 12122528 10.1007/s00134-002-1345-z
Granja C Cabral G Teixeira-Pinto A Costa-Pereira A Quality of life 6 months after cardiac arrest Resuscitation 2002 55 37 44 12297352 10.1016/S0300-9572(02)00203-4
Granja C Morujão E Costa-Pereira A Quality of life in acute respiratory distress syndrome may be no worst than in other ICU survivors Intensive Care Med 2003 29 1744 1750 12774161 10.1007/s00134-003-1808-x
Granja C Dias C Costa-Pereira A Sarmento A Quality of life of survivors from severe sepsis and septic shock may be similar to that of others who survive critical illness Crit Care 2004 8 R91 R98 (DOI 10.1186/cc2818). 15025783 10.1186/cc2818
The EuroQol® Group EuroQol® – a new facility for the measurement of health-related quality of life Health Policy 1990 16 199 208 10109801 10.1016/0168-8510(90)90421-9
Brooks R with the EuroQol Group EuroQol: the current state of play Health Policy 1996 37 53 72 10158943 10.1016/0168-8510(96)00822-6
Rivera Fernandez R Sanchez Cruz JJ Vazquez Mata G Validation of a quality of life questionnaire for critically ill patients Intensive Care Med 1996 22 1034 1042 8923066 10.1007/s001340050209
Jones C Griffiths RD, Jones C Acute psychological problems Intensive Care Aftercare 2002 Oxford: Butterworth-Heinemann 19 26
Skirrow P Griffiths RD, Jones C Delusional memories of ICU Intensive Care Aftercare 2002 Oxford: Butterworth-Heinemann 27 35
Nunes B Pais J Garcia R Duarte Z Granja C Silva MC Cardiac arrest: long-term cognitive and imaging analysis Resuscitation 2003 57 287 297 12804806 10.1016/S0300-9572(03)00033-9
Kapfhammer HP Rothenhausler HB Krauseneck T Stoll C Schelling G Posttraumatic stress disorder and health-related quality of life in long-term survivors of acute respiratory distress syndrome Am J Psychiatry 2004 161 45 52 14702249 10.1176/appi.ajp.161.1.45
Jones C Humphries G Griffiths RD Preliminary validation of the ICUM tool for assessing memory of the intensive care experience Clin Intensive Care 2000 11 251 253
Capuzzo M Valpondi V Cingolani E De Luca S Gianstefani G Grassi L Alvisi R Application of the Italian version of the Intensive Care Unit Memory tool in the clinical setting Crit Care 2004 8 R48 R55 14975055 10.1186/cc2416
Van de Leur JP Van der Schans CP Loef BG Deelman BG Geertzen JHB Zwaveling JH Discomfort and factual recollection in intensive care unit patients Crit Care 2004 8 R467 R473 15566593 10.1186/cc2976
Sedation and Analgesia Task Force Clinical practice guidelines for the sustained use of sedatives and analgesics in the critically ill adult Crit Care Med 2002 30 119 141 11902253
| 15774056 | PMC1175917 | CC BY | 2021-01-04 16:04:51 | no | Crit Care. 2005 Jan 31; 9(2):R96-R109 | utf-8 | Crit Care | 2,005 | 10.1186/cc3026 | oa_comm |
==== Front
Crit CareCritical Care1364-85351466-609XBioMed Central London cc30301577405410.1186/cc3030ResearchImpact of a nurses' protocol-directed weaning procedure on outcomes in patients undergoing mechanical ventilation for longer than 48 hours: a prospective cohort study with a matched historical control group Tonnelier Jean-Marie [email protected] Gwenaël [email protected] Gal Grégoire [email protected] Christophe [email protected] Anne [email protected] Jean-Michel [email protected]'Her Erwan [email protected] Réanimation Médicale, Centre Hospitalier Universitaire de la Cavale Blanche, Brest, France2005 17 1 2005 9 2 R83 R89 20 11 2003 23 4 2004 18 11 2004 25 11 2004 Copyright © 2005 Tonnelier 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 cited.
Introduction
The aim of the study was to determine whether the use of a nurses' protocol-directed weaning procedure, based on the French intensive care society (SRLF) consensus recommendations, was associated with reductions in the duration of mechanical ventilation and intensive care unit (ICU) length of stay in patients requiring more than 48 hours of mechanical ventilation.
Methods
This prospective study was conducted in a university hospital ICU from January 2002 through to February 2003. A total of 104 patients who had been ventilated for more than 48 hours and were weaned from mechanical ventilation using a nurses' protocol-directed procedure (cases) were compared with a 1:1 matched historical control group who underwent conventional physician-directed weaning (between 1999 and 2001). Duration of ventilation and length of ICU stay, rate of unsuccessful extubation and rate of ventilator-associated pneumonia were compared between cases and controls.
Results
The duration of mechanical ventilation (16.6 ± 13 days versus 22.5 ± 21 days; P = 0.02) and ICU length of stay (21.6 ± 14.3 days versus 27.6 ± 21.7 days; P = 0.02) were lower among patients who underwent the nurses' protocol-directed weaning than among control individuals. Ventilator-associated pneumonia, ventilator discontinuation failure rates and ICU mortality were similar between the two groups.
Discussion
Application of the nurses' protocol-directed weaning procedure described here is safe and promotes significant outcome benefits in patients who require more than 48 hours of mechanical ventilation.
intensive care unitmechanical ventilationprotocol-directed weaningSee related commentary
==== Body
Introduction
The duration of weaning from mechanical ventilation (MV) represents a large proportion of the overall ventilation period [1]. The time from initiation of weaning to successful endotracheal extubation may account for as much as 40% of the overall ventilatory time [1]. A great number of studies have demonstrated that prompt recognition of reversal of respiratory failure using standardized procedures and daily screening may shorten the overall duration of MV [2-5]. However, those studies were mainly conducted by respiratory therapists in North American intensive care units (ICUs), whereas in Europe the respiratory therapist's roles are mainly assumed by nurses and physicians. Despite the accumulating evidence to support their routine use, the value of standardized weaning procedures continues to be of clinical and academic interest [6].
In October 2001, the French intensive care society (Société de Réanimation de Langue Francaise) organized a consensus conference to develop recommendations for MV weaning in the ICU [7]. The aim of the present prospective cohort study, which included a matched historical control group, was to determine whether a routine nurses' protocol-directed weaning procedure based on these recommendations and daily screening were efficient in terms of MV duration and ICU length of stay (LOS) in patients who required more than 48 hours of MV.
Methods
Study design
This prospective study was conducted in the 12-bed ICU of an 800-bed teaching hospital from January 2002 to February 2003. We employed a matched control group (1:1 matching) of patients identified from a historical database.
Selection of patients and controls
Screened patients were those who required more than 48 hours of MV and who satisfied eligibility criteria (see below) for a spontaneous breathing trial (SBT). Exclusion criteria were tracheostomy before ICU admission or within the first 48 hours of ICU management, and age under 18 years.
A computer-generated list of potential control individuals was obtained from a historical database of patients attending the same ICU (977 patients from 1999 to 2001). Controls were selected based on matching in terms of the following criteria: age (±5 years), sex, Simplified Acute Physiology Score (SAPS) II (±5 points; calculated within the first 24 hours of ICU admission), and admission diagnosis. The list of potential controls was reviewed for the best possible match, giving a ranking priority to SAPS II, followed by age, sex and then diagnosis.
Study protocol
The study protocol was written according to the final recommendations of the Société de Réanimation de Langue Francaise's consensus conference on weaning. Patient eligibility for the weaning procedure was identified by daily screening by nurses. Screening was deemed to start immediately after ICU admission. The eligibility criteria for a SBT were the following: fractional inspired oxygen <50%; positive end-expiratory pressure <5 cmH2O; no vasopressor infusion; no sedative agent infusion; and response to simple orders. Physician approval for initiation of SBT was not required. A planned SBT duration of 90 min was employed, and SBTs were always performed using a T-piece. The SBT was terminated before 90 min had elapsed and considered a failure if any of the following criteria were satisfied: pulse oximetry <90%, a respiratory rate >35 breaths/min, a heart rate or a systolic arterial pressure variation >20%, or occurrence of patient agitation. All of these criteria for failure were specifically recorded. A SBT was considered to be successful when the patient could breathe spontaneously for 90 min. Physicians were asked to approve discontinuation of MV following a successful SBT. Extubation was therefore performed if cough was subjectively considered efficient, and if a leak test was considered positive (inspiratory and/or expiratory air leaks after cuff deflation). If the SBT was not well tolerated, then the failure criteria were specifically recorded and the patient returned to their prior ventilator settings and mode. Such patients were then subjected to screening the following day (Fig. 1).
Pre-protocol weaning management (historical controls)
Before establishment of the nurse's protocol-directed weaning procedure, routine weaning from MV was 'physician directed', as in most European ICUs. Physician's individual preferences determined the mode of weaning (volume assisted controlled, pressure support ventilation, or T-piece). Weaning criteria were not monitored daily. The decision to extubate after a successful weaning procedure was also physician directed.
Definitions
All definitions were selected a priori before study analysis. Admission diagnosis was divided into four main groups: medicine, surgery, neurosurgery and acute exacerbations in chronic obstructive pulmonary disease. Ventilator-associated pneumonia was defined as the initiation of antibiotics and at least two of the following criteria: positive protected bronchoscopy cultures; fever or rising leucocyte count; and characteristic chest radiograph findings. The duration of ventilation was calculated as follows: (day of extubation + 1) - (day of intubation). Successful discontinuation of MV was defined as continuous independence from ventilator support for a period of at least 48 hours. Unsuccessful MV discontinuation was defined as need for noninvasive ventilation and/or reintubation within a 48 hour period. Criteria to initiate noninvasive ventilation following extubation were occurrence of clinical signs of acute respiratory failure, with or without hypercapnia.
Outcomes
The overall duration of MV and the ICU LOS were considered primary outcomes. Secondary outcomes were the overall incidence of ventilator-associated pneumonia, unsuccessful extubation rates and ICU mortality.
Statistical analysis
Categorical variables were expressed as percentage and continuous variables as mean ± standard deviation. P < 0.05 was considered statistically significant. Percentages were compared using χ2 tests, and means using Student's t-test. Kaplan–Meier curves were used to determine the probability of remaining ventilated during the overall ICU LOS; curves were compared using the log-rank test. When patients died before discontinuation of MV, matched data were censored within the analysis.
Results
Physiological variables
During the prospective study period 392 patients were admitted to our ICU, of whom 384 patients required ventilatory support. A total of 297 patients were mechanically ventilated through an endotracheal tube, of whom 204 required MV for longer than 48 hours. Among these 204 eligible patients, 100 were excluded from analysis because they died prior to initiation of any weaning procedure (85 patients) or because they were tracheostomized before admission or within the first 48-hour period (15 patients; Fig. 2). After 1:1 matching, 208 patients were finally included in the analysis: 104 patients in the prospective routine nurse's protocol-directed weaning procedure (cases) and 104 patients in the standard physician-directed weaning procedure (controls). Matching was successful for all parameters, and patient demographic variables were similar within groups (age 56 ± 18 years, sex ratio [male/female] 3/2; SAPS II = 49 ± 18). Admission diagnoses were similarly distributed within groups (Table 1).
Duration of mechanical ventilation and intensive care unit length of stay
All patients were under volume-assisted controlled or pressure support ventilation before entering the weaning procedure. The overall MV duration was 16.6 ± 13 days within cases, and 22.5 ± 21 days within controls (P = 0.02); this difference between groups is illustrated by Kaplan–Meier analysis in Fig. 3. The ICU LOS was 21.6 ± 14.3 days within cases and 27.6 ± 21.7 days within controls (P = 0.02). In subgroup analysis, the duration of MV and ICU LOS were shorter for medical patients (Table 2).
Outcomes
No significant differences in unsuccessful MV discontinuation rates were observed between groups (31% for cases versus 35% for controls; P = 0.81). Mortality was similar between the two groups (7% for cases versus 5% for controls; P = 0.92). A positive trend toward a decrease in ventilator-associated pneumonia rate was observed for the cases (20.2% versus 31%; P = 0.12; Table 3).
Discussion
The present study was designed to determine the clinical benefit of a nurse's protocol-directed weaning procedure in a broad range of ICU patients (including both medical and surgical patients) who had not been disconnected from the ventilator after 48 hours of MV. Our findings demonstrate the effectiveness of such a procedure performed on a routine basis. The nurse's protocol-directed weaning procedure reduced the duration of MV and the overall ICU LOS in patients who remained dependent on MV after a 48-hour period, without any increase in adverse events or in rate of unsuccessful extubation. Easily implemented, this protocol-directed weaning procedure synthesizes an approach that incorporates the following: daily screening by nursing staff and a single daily 90 min SBT with T-piece trial in selected patients. The criteria used in this procedure to screen patients for readiness for the weaning trial were very simple clinical criteria.
Some authors have demonstrated a key role for daily screening using predetermined criteria during the weaning process [2,4]. In both of those studies, patients who satisfied the criteria were subjected to a 2 hour trial of CPAP while they remained attached to the mechanical ventilator circuit. How exactly a SBT should be performed remains subject to debate, and its optimal duration is not known, although there are data suggesting that it may be shortened [8-11]. For the routine procedure, we arbitrarily chose a 90 min duration and to perform the trial while the patient was disconnected from the ventilator, without adding any positive end-expiratory pressure.
The mean MV duration and ICU LOS found in this study differ significantly from those of previous studies [2,3,12]. In the study by Ely and coworkers [2] patients received MV for a median of 4.5 days, and in the study by Esteban and coworkers [10] patients were ventilated for a median of 5 days. Moreover, in the study by Ely and colleagues the ICU LOS was similar between groups. The main reasons for such a difference between our study and the previous ones are the higher mean SAPS II in our patients (49 ± 18 in both groups) and the selection of patients who survived and were not weaned from MV after a 48 hour period (i.e. patients whose conditions probably remained unstable for a long period and/or those who may be considered 'difficult to wean'). Therefore, we believe that we selected patients who may derive particular benefit from a decrease in the duration of MV.
The unsuccessful extubation rate also appears rather high in our patients as compared with previous studies (31% within cases versus 35% within controls) [10,13-15]. However, this finding must be interpreted with caution because we considered unsuccessful extubation to be the need for any form of ventilatory assistance within a 48 hour period (either delivered noninvasively via face mask or invasively via an endotracheal tube), whereas in the vast majority of other studies unsuccessful extubation was considered just as need for endotracheal intubation. If we consider the reintubation rate alone, the incidence of extubation failure decreases to 21% in cases and 18% in controls. However, one may observe that, apart from earlier extubation, the failure rates were similar between the two groups (i.e. patients were not withdrawn too early, as compared with the 'standard' procedure).
Apart from the rather high mean SAPS II levels, the overall mortality rate in our study appears rather low in both groups (5%). However, in accordance with the analysis protocol we excluded patients who died before the predetermined weaning criteria were satisfied, and so this finding cannot be considered to be an adequate reflection of overall ICU mortality rate.
Case series with historical controls have advantages over randomized trials when the effects of routine daily procedures are studied. Indeed, it is easier for a broad range of nursing staff to institute a protocol throughout an entire ICU population than it is to limit it to specific monitored patients. Moreover, there is no risk for crossover effect, by which the staff may modify their behaviour toward control patients and thus blur the distinction between groups.
On the other hand the nonrandomized design of our study is a major limitation. First, only randomization can ensure comparability between cases and controls. Even if cases and controls were matched for age, sex, SAPS II and diagnosis on admission, selection bias cannot be fully excluded and might account for part of our results. Second, evolution of medical care during the study period might have influenced our results [16]. However, the use of sedatives and paralytics, and routine patient positioning and ventilatory settings did not fundamentally change over the period of interest (1999–2003). For example, patients with acute respiratory distress syndrome have been ventilated using a small tidal volume and an 'open lung approach' since publication of the study by Amato and coworkers [17].
Conclusion
This study demonstrates that a routine and simply applied nurse's protocol-directed weaning procedure was safe and promoted major clinical and outcome benefits for patients requiring MV over a 48 hour period. These benefits were obtained without increasing the average number of complications or the rate of unsuccessful extubation.
Key messages
• A simply applied nurse's protocol-directed weaning procedure was safe and promoted major clinical and outcome benefits for patients.
• Protocol-directed weaning procedure should be used in all ICUs.
Abbreviations
ICU = intensive care unit; LOS = length of stay; MV = mechanical ventilation; SAPS = Simplified Acute Physiology Score; SBT = spontaneous breathing trial.
Competing interests
The author(s) declare that they have no competing interests.
Authors' contributions
JMT drafted the manuscript and participated in the design of the study. GP conceived the study and helped to draft the manuscript. GLG performed the statistical analysis. CGG collected data. AR colected data. JMB participated in the design of the study. EL participated in its design and coordination, and helped to draft the manuscript. All authors read and approved the final manuscript.
Figures and Tables
Figure 1 Mechanical ventilation weaning protocol. Daily nurse screening identified patients eligible for weaning. A spontaneous breathing trial was considered to be successful when the patient could breathe spontaneously for 90 min without clinical intolerance. For such patients, physicians were then asked to approve discontinuation of mechanical ventilation. If the spontaneous breathing trial was not tolerated, then the patient was returned to their prior ventilator settings and screened the day after. FiO2, fractional inspired oxygen; PEEP, positive end-expiratory pressure; SpO2, pulse oximetry.
Figure 2 Case selection. ICU, intensive care unit; NIV, noninvasive ventilation.
Figure 3 Kaplan–Meier curves of the risk for remaining mechanically ventilated in protocol-directed (cases) versus physician-directed weaning groups (controls). The protocol-directed weaning procedure allowed reduction in the overall duration of mechanical ventilation, whatever the patient's diagnosis. The overall mechanical ventilation duration was 16.6 ± 13 days in cases and 22.5 ± 21 days in controls (P = 0.02).
Table 1 Baseline characteristics in the study patients
Physiological parameters Protocol-directed weaning group (cases) Physician-directed weaning group (controls)
Patients (n) 104 104
Male sex (n [%]) 62 (60) 62 (60)
Age (years; mean ± SD) 57 ± 18 56 ± 18
SAPS II (mean ± SD) 49 ± 18 49 ± 18
Admission diagnosis (n [%])
Medicine 54 (52) 63 (61)
Surgery 25 (24) 20 (19)
Neurosurgery 19 (18) 16 (15)
COPD 6 (6) 5 (5)
Matching was successful for all parameters, and patients' physiological parameters were similar between groups. Admission diagnoses were similarly distributed between groups. COPD, chronic pulmonary obstructive disease; SAPS, Simplified Acute Physiology Score; SD, standard deviation.
Table 2 Outcome comparison between the study groups, according to admission diagnosis
Outcomes Protocol-directed weaning group (cases) Physician-directed weaning group (controls) P
Mechanical ventilation duration (days)
Medicine 15.3 ± 13 23 ± 22.6 0.02
Surgery 19.1 ± 15.2 17.4 ± 13.7 0.72
Neurosurgery 18.9 ± 12.8 26.3 ± 23.1 0.29
COPD 14.1 ± 14.6 24.9 ± 24 0.33
Combined 16.6 ± 13 22.5 ± 21 0.02
ICU length of stay (days)
Medicine 20.1 ± 13.7 27.7 ± 23 0.02
Surgery 25 ± 17 23.3 ± 15.4 0.72
Neurosurgery 23 ± 13 34 ± 24 0.09
COPD 20.4 ± 14.6 28.5 ± 25 0.47
Combined 21.6 ± 14.3 27.6 ± 21.7 0.02
Values are expressed as mean ± standard deviation. COPD, chronic pulmonary obstructive disease; ICU, intensive care unit.
Table 3 Comparison of complications between study groups
Complications Protocol-directed weaning group Physician-directed weaning group P
Unsuccessful MV discontinuation 33 (31) 37 (35) 0.81
Reintubation within 48 hours 22 (21) 19 (18) 0.33
NIV for postextubation respiratory distress 22 (21) 26 (25) 0.15
Ventilator-associated pneumonia 21 (20) 33 (31) 0.12
ICU mortality 7 (7) 5 (5) 0.92
Values are expressed as number (%). No significant differences between groups were observed. ICU, intensive care unit; MV, mechanical ventilation; NIV, noninvasive ventilation.
==== Refs
Esteban A Alia I Ibanez J Benito S Tobin MJ Modes of mechanical ventilation and weaning. A national survey of Spanish hospitals. The Spanish Lung Failure Collaborative Group Chest 1994 106 1188 1193 7924494
Ely EW Baker AM Dunagan DP Burke HL Smith AC Kelly PT Johnson MM Browder RW Bowton DL Haponik EF Effect on the duration of mechanical ventilation of identifying patients capable of breathing spontaneously N Engl J Med 1996 335 1864 1869 8948561 10.1056/NEJM199612193352502
Horst HM Mouro D Hall-Jenssens RA Pamukov N Decrease in ventilation time with a standardized weaning process Arch Surg 1998 133 483 488 discussion 488–489 9605909 10.1001/archsurg.133.5.483
Kollef MH Shapiro SD Silver P St John RE Prentice D Sauer S Ahrens TS Shannon W Baker-Clinkscale D A randomized, controlled trial of protocol-directed versus physician-directed weaning from mechanical ventilation Crit Care Med 1997 25 567 574 9142019 10.1097/00003246-199704000-00004
Marelich GP Murin S Battistella F Inciardi J Vierra T Roby M Protocol weaning of mechanical ventilation in medical and surgical patients by respiratory care practitioners and nurses Chest 2000 118 459 467 10936141 10.1378/chest.118.2.459
MacIntyre N Evidence-based guidelines for weaning and discontinuing ventilatory support: a collective task force facilitated by the American College of Chest Physicians; the American Association for Respiratory Care; and the American College of Critical Care Medicine Chest 2001 120 375S 396S 11742959 10.1378/chest.120.6_suppl.375S
Richard C Beydon L Cantagrel S Cuvelier A Fauroux B Garo B Holzapfel L Lesieur O Levraut J Maury E 21st Consensus Conference on Intensive Care and Emergency Medicine: mechanical ventilation weaning [in French] Réanimation 2001 10 697 698 10.1016/S1164-6756(00)00080-3
Esteban A Frutos F Tobin MJ Alia I Solsona JF Valverdu I Fernandez R de la Cal MA Benito S Tomas R A comparison of four methods of weaning patients from mechanical ventilation. Spanish Lung Failure Collaborative Group N Engl J Med 1995 332 345 350 7823995 10.1056/NEJM199502093320601
Brochard L Rauss A Benito S Conti G Mancebo J Rekik N Gasparetto A Lemaire F Comparison of three methods of gradual withdrawal from ventilatory support during weaning from mechanical ventilation Am J Respir Crit Care Med 1994 150 896 903 7921460
Esteban A Alia I Tobin MJ Gil A Gordo F Vallverdu I Blanch L Bonet A Vazquez A de Pablo R Effect of spontaneous breathing trial duration on outcome of attempts to discontinue mechanical ventilation. Spanish Lung Failure Collaborative Group Am J Respir Crit Care Med 1999 159 512 518 9927366
Vallverdu I Calaf N Subirana M Net A Benito S Mancebo J Clinical characteristics, respiratory functional parameters, and outcome of a two-hour T-piece trial in patients weaning from mechanical ventilation Am J Respir Crit Care Med 1998 158 1855 1862 9847278
Marelich GP Murin S Battistella F Inciardi J Vierra T Roby M Protocol weaning of mechanical ventilation in medical and surgical patients by respiratory care practitioners and nurses: effect on weaning time and incidence of ventilator-associated pneumonia Chest 2000 118 459 467 10936141 10.1378/chest.118.2.459
Epstein SK Ciubotaru RL Wong JB Effect of failed extubation on the outcome of mechanical ventilation Chest 1997 112 186 192 9228375
Esteban A Alia I Gordo F Fernandez R Solsona JF Vallverdu I Macias S Allegue JM Blanco J Carriedo D Extubation outcome after spontaneous breathing trials with T-tube or pressure support ventilation. The Spanish Lung Failure Collaborative Group Am J Respir Crit Care Med 1997 156 459 465 9279224
Chan PK Fischer S Stewart TE Hallett DC Hynes-Gay P Lapinsky SE MacDonald R Mehta S Practising evidence-based medicine: the design and implementation of a multidisciplinary team-driven extubation protocol Crit Care 2001 5 349 354 11737924 10.1186/cc1068
Sacks H Chalmers TC Smith H Jr Randomized versus historical controls for clinical trials Am J Med 1982 72 233 240 7058834 10.1016/0002-9343(82)90815-4
Amato MB Barbas CS Medeiros DM Magaldi RB Schettino GP Lorenzi-Filho G Kairalla RA Deheinzelin D Munoz C Oliveira R Effect of a protective-ventilation strategy on mortality in the acute respiratory distress syndrome N Engl J Med 1998 338 347 354 9449727 10.1056/NEJM199802053380602
| 15774054 | PMC1175918 | CC BY | 2021-01-04 16:04:52 | no | Crit Care. 2005 Jan 17; 9(2):R83-R89 | utf-8 | Crit Care | 2,005 | 10.1186/cc3030 | oa_comm |
==== Front
Crit CareCritical Care1364-85351466-609XBioMed Central London cc30331577404210.1186/cc3033ResearchMinimal instructions improve the performance of laypersons in the use of semiautomatic and automatic external defibrillators Beckers Stefan [email protected] Michael [email protected] Johannes [email protected] Matthias [email protected] Ralf [email protected] Rolf [email protected] Resident, Department of Anaesthesiology, University Hospital Aachen, Aachen, Germany2 Medical Student, Department of Anaesthesiology, University Hospital Aachen, Aachen, Germany3 Professor, Department of Anaesthesiology, University Hospital Aachen, Aachen, Germany4 Professor and Chairman, Department of Anaesthesiology, University Hospital Aachen, Aachen, Germany2005 31 1 2005 9 2 R110 R116 14 9 2004 13 10 2004 1 11 2004 30 11 2004 Copyright © 2005 Beckers 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 cited.
Introduction
There is evidence that use of automated external defibrillators (AEDs) by laypersons improves rates of survival from cardiac arrest, but there is no consensus on the optimal content and duration of training for this purpose. In this study we examined the use of semiautomatic or automatic AEDs by laypersons who had received no training (intuitive use) and the effects of minimal general theoretical instructions on their performance.
Methods
In a mock cardiac arrest scenario, 236 first year medical students who had not previously attended any preclinical courses were evaluated in their first study week, before and after receiving prespecified instructions (15 min) once. The primary end-point was the time to first shock for each time point; secondary end-points were correct electrode pad positioning, safety of the procedure and the subjective feelings of the students.
Results
The mean time to shock for both AED types was 81.2 ± 19.2 s (range 45–178 s). Correct pad placement was observed in 85.6% and adequate safety in 94.1%. The time to shock after instruction decreased significantly to 56.8 ± 9.9 s (range 35–95 s; P ≤ 0.01), with correct electrode placement in 92.8% and adequate safety in 97%. The students were significantly quicker at both evaluations using the semiautomatic device than with the automatic AED (first evaluation: 77.5 ± 20.5 s versus 85.2 ± 17 s, P ≤ 0.01; second evaluation: 55 ± 10.3 s versus 59.6 ± 9.6 s, P ≤ 0.01).
Conclusion
Untrained laypersons can use semiautomatic and automatic AEDs sufficiently quickly and without instruction. After one use and minimal instructions, improvements in practical performance were significant. All tested laypersons were able to deliver the first shock in under 1 min.
automated external defibrillatorcardiopulmonary resuscitationdefibrillationlaypersonintuitiveSee related commentary
==== Body
Introduction
Mortality from sudden cardiac death is up to 375,000 patients per year in Europe [1] and in the vast majority of cases it is caused by ventricular fibrillation [2]. To increase survival rates, the period between developing ventricular fibrillation and the first defibrillation must be as short as possible. Early defibrillation, done during the first minute of the event, is successful in 85% of cases. Each additional minute without treatment reduces the survival rate by a further 10% [3]. Therefore, early defibrillation must be implemented into the chain of survival [4], and to this end the development of programmes for nonmedical lay responders is recommended and supported by many international societies. For years, the American Heart Association has postulated inclusion of AED use in basic life support (BLS) training [4,5]. Furthermore, first responders may operate an AED without having any background knowledge about the instrument. Previous studies have shown that even children can handle an AED confidently and effectively [6].
There is no consensus as yet regarding time frames for specific training programmes, but for organizational reasons and for further implementation of public access defibrillation (PAD) programmes in the future, it is necessary that this period be defined. It remains unclear how lay users should be instructed to perform safe and effective defibrillation. The aim of the present study was to evaluate the intuitive use (i.e. without training) of AEDs, both in fully automatic and in semiautomatic modes, and to study the effect of brief and well directed theoretical instruction.
Methods
Participants
A total of 236 first year medical students were tested during their first 2 weeks at the medical school of the University of Aachen. All students were informed that their performance would be evaluated and used for scientific study. No personal data were collected. Furthermore, no damage to anyone's health was anticipated because the AED uses no current. Therefore, the institutional ethical committee waived the need to obtain informed consent from each participant. None of the students were prompted or prepared in any way before the study.
Equipment
The Medtronic Physio-Control LifePak™ CR-T AED trainer (Medtronic Physio-Control LifePak™; Medtronic, Düsseldorf, Germany) provides the necessary interface for demonstrating practical skills during a simulated cardiac arrest, and was used rather than the original Medtronic Physio-Control LifePak™ CR Plus. No current is applied by the training device.
After opening the lid a red handle must be pulled, which then releases self-adhesive electrode pads with integrated cables connected to the device (Fig. 1). Voice prompts (Table 1) and an illustrated reference card inside the opened lid guides users in a step-by-step manner. No text prompts are displayed on the screen. After turning the device on and positioning the electrodes properly, the analyzing process of the AED starts automatically and is finished after 10 s in both types of AED. In the semiautomatic mode it takes 18 s from the beginning of the analyzing process until the device is charged, and an alarm tone sounds until the shock button is pressed. In the automatic mode the shock is delivered automatically after 21 s and the charge is calculated from the analysis of heart rhythm over this period [7].
Study protocol
In a mock cardiac arrest scenario, the students were evaluated on a manikin (ResusciAnne®; Laerdal, Stavanger, Norway). After randomization, 118 students were tested on an AED in automatic mode, and 118 were tested on a semiautomatic AED. The device was kept in its usual standby mode. The manikin was positioned supine and dressed in a zippered jacket. Three physicians skilled in providing and teaching advanced life support (certified instructors of the European Resuscitation Council) were present and recorded data while each student operated the AED. Each student was tested individually and was unable to observe the performance of other participants. They were read the following text: 'This patient is unconscious, not breathing and has no signs of circulation. The device in front of you may help to restore spontaneous circulation.'
The procedure ended when the first shock was delivered or no shock could be given in 240 s. Placement of the electrode pads was accepted as correct if the left pad covered at least 50% of an area circumscribed by the nipple line superiorly, costal margin inferiorly, mid-clavicular line medially and mid-axillary line laterally. The right pad was required to cover at least 50% of an area circumscribed by the clavicle superiorly, nipple line inferiorly, anterior axillary line laterally and right sternal margin medially. Application of the AED was considered to be safe if the student remained clear of the manikin during delivery of the shock. If a technical problem occurred, the student damaged the AED, started with conventional cardiopulmonary resuscitation, or had language difficulties, then this was classified as 'any other problem'.
After having completed the tests, each student completed a standardized questionnaire to evaluate whether they had any experience with an AED before the study or whether they had any medical education (e.g. nurse, paramedic etc.). After a period of 1 week all test candidates were assigned the same type of device they had used in their first test and were re-evaluated in the same scenario. During this week they attended a short lecture (15 min) emphasizing the following core objectives: importance of sudden cardiac death and of defibrillation in this context; importance of 'time to shock' to return of spontaneous circulation and success of resuscitation over time; importance of correct electrode pad positioning; safety aspects when using an AED; general procedure for defibrillation devices (e.g. analysis, voice prompts); general AED algorithm following guidelines; and special instructions for slim and overweight victims.
There were no practical training sessions available between the two evaluations and no specific information on the tested AED devices was given.
Data analysis
Data are expressed as means ± standard deviation. P ≤ 0.05 was considered statistically significant. Statistical software SPSS version 11.0 (SPSS Inc., Chicago, IL, USA) was used.
Primary end-points
The primary end-point was to determine the time from the beginning of the scenario to first shock. Using a t-test, differences in time to shock between the first and second evaluations were calculated, as well as between the semiautomatic and the automatic devices for each time point.
Secondary end-points
The secondary end-points were chosen to assess correct electrode pad positioning and the safety of the procedure, as well as previous medical knowledge. Data were compared in a proportional manner and tested for significant differences using the McNemar test.
Results
The mean age of the study population was 20.7 ± 2.9 years (range 18–42 years). Of the 236 students included, 28 (11.9%) had a history of medical education (16 emergency medicine technicians and paramedics, and 12 nurses).
Time to defibrillation, electrode pad positioning and safety
In the first evaluation the time to shock for both devices was 81.2 ± 19.2 s (range 45–178 s). The pads were positioned correctly by 85.6% of the students. Shock was administered safely by 94.1%. In the second evaluation the time to first defibrillation decreased significantly to 56.8 ± 9.9 s (range 35–95 s; P ≤ 0.01). The electrodes were correctly placed in 92.8% of cases, and shock was administered safely in 97% of cases.
Table 2 summarizes these variables by type of AED. When comparing time to first shock between semiautomatic and automatic AEDs, the students were significantly faster in both evaluations using the semiautomatic device (first evaluation: 77.5 ± 20.5 s versus 85.2 ± 17 s, P ≤ 0.01; second evaluation: 55 ± 10.3 s versus 59.6 ± 9.6 s, P ≤ 0.01).
In the second evaluation 113 out of 118 (95.8%) students were able to deliver a shock safely and none failed in the semiautomatic group. In the automatic group 115 of 118 (97.5%) were able to deliver a shock, but three students failed.
Students with pre-existing medical education were significant faster at both times (first evaluation: 73.0 ± 17.1 s versus 83.0 ± 19.1 s, P ≤ 0.01; second evaluation: 51.8 ± 9.2 s versus 58.3 ± 10.1 s, P ≤ 0.01). All other findings are summarized in Table 2.
Discussion
This study represents the first comparison in laypersons of the use of fully automatic devices with that of semiautomatic devices, including the largest study group yet reported. The improvements with both devices, in terms of time to first shock, between initial use without instruction and use following the described 15-min theoretical instruction were significant.
Since the first clinical use of AEDs in the early 1980s [8], developments in technology have led to initiatives by health and governmental organizations to develop PAD programmes [9]. Various studies [10-13] have shown improvements in rates of survival from out-of-hospital cardiac arrest where nonmedical personnel were trained in PAD programmes. However, only a few studies described the performance of laypersons, but even these individuals were initially instructed before evaluation [6,14]. In a cross-over design, Eames and coworkers [15] compared the use of three different AEDs by nearly untrained laypersons (n = 24), but information had been given concerning the application of a shock, following instructions for the device and the impact that time to defibrillation has on outcome. To our knowledge, the present study is the first to describe the use of AEDs without any instructions before first use. It is noteworthy that, even without instruction, 226 out of 236 participants (95.8%) were able to deliver a shock.
Safety aspects associated with automatic mode have been considered and critically discussed, but the question of whether it is better not to administer an advised shock in the case of proven ventricular fibrillation or to have a shock delivered automatically with a delay is rhetorical. Surveying safety aspects of the tested AED, we found that 92.4% of students were able to deliver a shock safely in semiautomatic mode and 95.8% in automatic mode during testing without prior instruction. After theoretical instruction, these rates increased to 95.8% and 97.5%, respectively. Eames and coworkers [15] found that all individuals stood clear while delivering the shock but, as mentioned above, only 24 subjects were tested; it follows that possible reluctance to adhere to safety procedures might not have been detected in that investigation. Fromm and Varon [16] found that still 10 months after initial training, the 'simplicity of use of the particular AED' was the core issue determining safety. The important benefit of devices programmed in automatic mode is that they relieve the layperson of decision making in an unfamiliar and stressful situation.
Contrary to expectations were our findings regarding electrode pad placement. There was an anticipated and significant improvement in the automatic group, but only a trend was observed in the semiautomatic group. It is inexplicable why, after instruction, 9.3% (11 students) still could not achieve correct pad positioning. This is in contrast to the study by Gundry and coworkers [6], in which all children were able to position the pad in the required area, whereas Eames and coworkers [15] observed 20.9% incorrect electrode placement with the LifePak CR Plus. With the Philips/Laerdal Heartstart1 the result was only 4.2%, and the Zoll AEDPlus had the worst result, with 41.6% incorrect pad positioning. In some cases, confusing descriptions or drawings might have caused the incorrect positioning of the adhesive pad electrodes in the present study. Overall, this supports the conclusion of Eames and coworkers that simple devices should be developed with clear visual instructions, and it reiterates that design, construction and visual aids have an impact on user performance. This statement was confirmed by our observation that even in the second evaluation, in the automatic group three students were unable to deliver a shock. In these three cases the students were confused by the voice prompts of the automatic device, and while trying to push the shock button they turned the device off. Other detected problems in both testing sessions occurred mainly as a result of language problems, but they were reduced after instruction. In general, none of the participants appeared to be apprehensive about operating the AED because none of them refused to participate in the study or to apply the device to the manikin.
The significant difference in time to shock before and after instruction between semiautomatic AED and the automatic device is a possible effect of the software version used. However, the programmed delay of 3 s to delivery of shock in the automatic device does not adequately explain this finding. Changing the timing of voice prompts and development of clearer instructions may lead to different results. In general, however, the voice prompts that lead to the best results remains a matter for discussion.
The studies published thus far led to the statement from the American Heart Association and the Resuscitation Council UK 'not to specify the nature of content or duration of BLS plus AED programs due to the lack of current evidence on which to base any such guidance' [17]. As yet there is no consensus regarding the optimal duration of specific training programmes. It will be difficult to achieve that perfect performance of certain skills that indicates successful training of laypersons. Especially for organizational reasons, it is fundamental to define time frames of course concepts. We endorse the assertion by Gundry and coworkers [6] and Moule and Albarran [17] that simplified training programmes should be developed, exploiting the potential of multimedia technology, along with adequate teaching and learning materials.
Various concepts have been described [10,11,17-21], but no data exist regarding how best to train and what the optimal duration of training is to achieve the best outcomes. Moule and Albarran [17] recently stated that the duration and most effective methods for teaching professionals and laypersons remain undefined. For this reason, no recommendation can yet be given. The implementation of PAD programmes in the future will depend mainly on the willingness of the public to participate in AED or cardiopulmonary resuscitation courses. The more time required to achieve the necessary skills, the less people will feel able to participate voluntarily. Furthermore, training sessions must be as precise and short as possible for organizational reasons; ideally, it should be possible for even a small number of instructors to reach a large group of trainees in minimal time.
Limitations of the study
The groups evaluated here are not representative of the general population with respect to sex (male 35% [n = 83], female 65% [n = 153]) and age, but the two groups are comparable (Table 2). In addition, the students were not chosen by random; nevertheless, they do represent young and inexperienced laypersons with respect to medical issues because, in Germany, students begin medical school directly after graduation from secondary school, without any specific preparation.
As considered by other studies [16,22], the participants might not have been free from external or internal motivations because of the fact that they were going into medicine. However, at this stage they are at best minimally trained and are not representative of the health care professional community. Furthermore, this internal motivation could have influenced their knowledge of theoretical issues concerning defibrillation within the evaluation period, but it is unlikely that there would have been a significant improvement in practical performance after, for instance, a web search.
Finally, no manikin used to represent an unconscious, breathless and pulseless victim can simulate a human perfectly. Because of this limitation, it is debatable whether benefits obtained in a simulated representation of a complex situation can be realized in clinical practice.
Conclusions
Untrained laypersons are able to use AEDs quickly and safely. The observation that measures of practical performance (i.e. time to first shock, accuracy of electrode pad placement and safety) were significantly improved after minimal theoretical instruction and one use, but without technical instructions in the use of the specific device, is supportive of widespread implementation of PAD programmes wherever possible. Moreover, enhanced acceptance of AEDs and the increased likelihood that AEDs will be used following directed 'public information' (e.g. television campaigns or other extensive publicly available media) is of great importance. Core issues (e.g. the significance of sudden cardiac death and the importance of defibrillation in this context) should be at the forefront of new educational changes; some suggestions in this regard were made in the present study.
Taken together, our findings support previous recommendations to develop features that can be made available in all AEDs. Sophisticated devices with simple instructions – visual or vocal – should be implemented in PAD programmes. Further developments should aim at simplifying the application of electrodes and achieving consistency in the instructions given by the different manufacturers. Value must be attached to giving general instructions and information about features that are common to all devices; describing the specific details of a device does not appear to be essential, as was assumed.
In our opinion, one of the most remarkable findings is that all tested laypersons were able to deliver a shock in less than 1 min after minimal instructions had been given, regardless of whether automatic or semiautomatic mode was used. Despite the limitations of the study, we conclude that only minimal background knowledge is needed for laypersons to use an AED safely and quickly, and that further implementation of AEDs for use by minimally trained persons without any medical training is possible. We believe that keeping instructions for laypersons as simple as possible will lead to greater acceptance and motivation, and will further facilitate PAD programmes. Time spent training to acquire necessary cardiopulmonary resuscitation skills within the BLS algorithm can be saved by focusing AED instructions in this way. Further studies are warranted to determine whether skills are retained over the long term.
Key messages
• This first observation in 'fully' automatic devices confirms that this type of AED can be used safely and effectively by lay responders.
• All tested laypersons were able to deliver a shock in less than 1 min after minimal instructions, regardless of whether automatic or semiautomatic mode was used.
• In future value must be attached to general instruction and similarities; describing specific details of available devices is not essential.
• Previous recommendations to develop features that can be made available in all AEDs are supported by our findings.
Abbreviations
AED = automated external defibrillator; BLS = basic life support; PAD = public access defibrillation.
Competing interests
The author(s) declare that they have no competing interests.
Authors' contributions
SB had conceived the study. SB, MF, JB, RK and RR designed the study protocol. Testing was performed by SB, MF, JB and MD. Statistical analysis was done by MF and MD. SB, MF, JB, RK and RR wrote and reviewed the manuscript before submission. All authors read and approved the final manuscript.
Acknowledgements
We thank all first year students of the medical faculty, University Aachen, Germany, for participating in this study and Medtronic Physio-Control Germany Corp for loaning the AED trainer and electrode pads to the investigators of the study.
Part of this work was presented at the 24th International Symposium on Intensive Care and Emergency Medicine (ISICEM), Brussel, Belgium; 30 March 30 2004 [23].
Figures and Tables
Figure 1 Evaluated automated external defibrillator: (left) automatic mode and (right) semiautomatic mode. Weight: 2.1 kg; physical dimensions: 10 × 20 × 24 cm.
Table 1 Voice prompts of the automated external defibrillator during the simulated cardiac arrest scenario
Automatic Semiautomatic Tones
Call for help now After the AED lid opens, two beeping tones sound. The voice prompts will sound following the beeping tones
Remove clothing from chest
Pull red handle to open bag
Peel each pad off blue plastic
Apply pads to exposed chest
Do not touch patient – evaluating heart rhythm Two beeping tones sound to simulate heart rhythm analysis
Stand by – preparing to shock
Everyone clear
Press flashing button Semiautomatic model only; an alarm tone sounds until the shock button is pressed
Do not touch patient – delivering shock Automatic model only; an alarm tone sounds until shock is delivered automatically
Shock delivered
Voice prompts that are not used
No shock advised
Shock not delivered
Check for pulse; if no pulse start CPR
Check for breathing; if not breathing start CPR
Check for signs of circulation; if no signs of circulation start CPR
Continue care
Check pads for good contact
Motion detected
Stop motion
AED, automated external defibrillator; CPR, cardiopulmonary resuscitation. Data from Medtronic [7].
Table 2 Time to first shock, correct electrode pad positioning and safety aspects before and after brief general instruction in defibrillation
Device
Semiautomatic Automatic
n 118 118
Age (mean ± SD) 21.1 ± 3.3 20.4 ± 2.3
Male (n [%]) 45 (38.1%) 38 (32.2%)
First evaluation
Time to shock (s; mean ± SD) 77 ± 20.4*† 85 ± 17.2†
Not able to deliver shock (n [%]) 6 (5.1%)‡ 4 (3.4%)
Incorrect pad positioning (n [%]) 15 (12.7%) 19 (16.1%)‡
Safe shock (n [%]) 109 (92.4%) 113 (95.8%)
Any other problemsa (n [%]) 18 (15.3%) 12 (10.2%)
Second evaluation
Time to shock (s; mean ± SD) 55 ± 10.3* 59 ± 9.1
Not able to deliver shock (n [%]) 0 (0%) 3 (2.5%)
Incorrect pad positioning (n [%]) 11 (9.3%) 3 (2.5%)
Safe shock (n [%]) 113 (95.8%) 115 (97.5%)
Any other problemsa (n [%]) 3 (2.5%) 2 (1.7%)
Comparison of subjects using semiautomatic and automatic devices at different evaluations. aAny other problems as described in the study protocol. *P < 0.05, versus automatic device (t-test). †P < 0.05, versus second evaluation (t-test). ‡P < 0.05, versus second evaluation (McNemar test). SD, standard deviation.
==== Refs
The Hypothermia After Cardiac Arrest Study Group Mild therapeutic hypothermia to improve the neurological outcome after cardiac arrest N Engl J Med 2002 346 549 556 11856793 10.1056/NEJMoa012689
Weaver DW Considerations for improving survival from out-of-hospital cardiac arrest Ann Emerg Med 1986 15 1181 1186 3752649
Advanced Life Support Working Party of the ERC Guidelines for adult advanced cardiac life support Resuscitation 1992 24 111 121 1335602 10.1016/0300-9572(92)90016-6
The American Heart Association in collaboration with the International Liaison Committee on Resuscitation Part 4: the automated external defibrillator: key link in the chain of survival. European Resuscitation Council Resuscitation 2000 46 73 91 10978789 10.1016/S0300-9572(00)00272-0
The American Heart Association in collaboration with the International Liaison Committee on Resuscitation Guidelines 2000 for Cardiopulmonary Resuscitation and Emergency Cardiovascular Care. An international consensus on science Circulation 2000 102 1 384 10880404
Gundry JW Comess KA DeRook FA Jorgenson D Bardy GH Comparison of naive sixth-grade children with trained professionals in the use of an automated external defibrillator Circulation 1999 100 1703 1707 10525489
Medtronic Physio-Control LifePak Corp RU LifePak CR-T AED-Trainer Operating Instructions 2002 Minneapolis, MN: Medtronic Physio-Control LifePak Corp
Weaver DW Copass MK Hill DL Fahrenbruch C Hallstrom AP Cobb LA Cardiac arrest treated with a new automatic external defibrillator by out-of-hospital first responders Am J Cardiol 1986 57 1017 1021 3706154 10.1016/0002-9149(86)90667-3
Davies CS Colquhoun M Graham S Evans T Chamberlain D Defibrillators in public places: introduction of a national scheme for public access defibrillation in England Resuscitation 2002 52 13 21 11801344 10.1016/S0300-9572(01)00439-7
Page RL Joglar JA Kowal RC Zagrodzky JD Nelson LL Ramaswamy K Barbera SJ Hamdan MH McKenas DK Use of automated external defibrillators by a U.S. airline N Engl J Med 2000 343 1210 1216 11071671 10.1056/NEJM200010263431702
Valenzuela TD Roe DJ Nichol G Clark LL Spaite DW Hardman RG Outcomes of rapid defibrillation by security officers after cardiac arrest in casinos N Engl J Med 2000 343 1206 1209 11071670 10.1056/NEJM200010263431701
White RD Asplin BR Bugliosi TF Hankins DG High discharge survival rate after out-of-hospital ventricular fibrillation with rapid defibrillation by police and paramedics Ann Emerg Med 1996 28 480 485 8909267
Ross P Nolan J Hill E Dawson J Whimster F Skinner D The use of AEDs by police officers in the City of London. Automated external defibrillators Resuscitation 2001 50 141 146 11719140 10.1016/S0300-9572(01)00343-4
Moore JE Eisenberg MS Cummins RO Hallstrom A Litwin P Carter W Lay person use of automatic external defibrillation Ann Emerg Med 1987 16 669 672 3578973
Eames P Larsen PD Galletly DC Comparison of ease of use of three automated external defibrillators by untrained lay people Resuscitation 2003 58 25 30 12867306 10.1016/S0300-9572(03)00103-5
Fromm RE JrVaron J Automated external versus blind manual defibrillation by untrained lay rescuers Resuscitation 1997 33 219 221 9044494 10.1016/S0300-9572(96)01036-2
Moule P Albarran JW Automated external defibrillation as part BLS: implications for education and practice Resuscitation 2002 54 223 230 12204454 10.1016/S0300-9572(02)00150-8
White RD Vukow L Buglosi T Early defibrillation by police: initial experience with measurement with time intervals and patient outcome Ann Emerg Med 1994 23 1009 1013 8185091
Stults KR Brown D Schug V Bean JA Prehospital defibrillation performed by emergency medical technicians in rural communities N Engl J Med 1984 310 219 223 6361562
Stults KR Brown DD Kerber RE Efficacy of an automated external defibrillator in the management of out-of-hospital cardiac arrest: validation of the diagnostic algorithm and initial clinical experience in a rural environment Circulation 1986 73 701 709 3512123
Mols P Beaucarne E Bruyninx J Labruyere JP De Myttenaere L Naeije N Watteeuw G Verset D Flamand JP Early defibrillation by EMTs: the Brussels experience Resuscitation 1994 27 129 136 8029534 10.1016/0300-9572(94)90005-1
Domanovits H Meron G Sterz F Kofler J Oschatz E Holzer M Mullner M Laggner AN Successful automatic external operation by people trained only in basic life support in a simulated cardiac arrest situation Resuscitation 1998 38 47 50 9783510 10.1016/S0300-9572(98)00114-2
Beckers S Fries M Bickenbach J Derwall M Kuhlen R Rossaint R Comparison of automatic vs semiautomatic automated external defibrillators used by laypersons [abstract] Critical Care 2004 P293 10.1186/cc2760
| 15774042 | PMC1175919 | CC BY | 2021-01-04 16:04:50 | no | Crit Care. 2005 Jan 31; 9(2):R110-R116 | utf-8 | Crit Care | 2,005 | 10.1186/cc3033 | oa_comm |
==== Front
Crit CareCritical Care1364-85351466-609XBioMed Central London cc30341577405510.1186/cc3034ResearchMyoglobin clearance by super high-flux hemofiltration in a case of severe rhabdomyolysis: a case report Naka Toshio 1Jones Daryl 2Baldwin Ian 3Fealy Nigel 4Bates Samantha 5Goehl Hermann 6Morgera Stanislao 7Neumayer Hans H 8Bellomo Rinaldo [email protected] Research Fellow, Department of Intensive Care and Department of Medicine, Melbourne University, Austin Hospital, Melbourne, Australia2 Registrar, Department of Intensive Care and Department of Medicine, Melbourne University, Austin Hospital, Melbourne, Australia3 Nurse Educator, Department of Intensive Care and Department of Medicine, Melbourne University, Austin Hospital, Melbourne, Australia and PhD Candidate, Latrobe University, Bundoora, Melbourne, Australia4 Nurse Educator, Department of Intensive Care and Department of Medicine, Melbourne University, Austin Hospital, Melbourne, Australia5 Research Nurse, Department of Intensive Care and Department of Medicine, Melbourne University, Austin Hospital, Melbourne, Australia6 Chief Engineer, Gambro Dialysatoren GmbH & Co., KG, Hechingen, Germany7 Staff Specialist, Department of Nephrology, Charitè, Humboldt, University of Berlin, Germany8 Director, Department of Nephrology, Charitè, Humboldt, University of Berlin, Germany9 Director of Research, Department of Intensive Care and Department of Medicine, Melbourne University, Austin Hospital, Melbourne, Australia2005 21 1 2005 9 2 R90 R95 4 10 2004 15 11 2004 25 11 2004 1 12 2004 Copyright © 2005 Naka 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.
Objective
To test the ability of a novel super high-flux (SHF) membrane with a larger pore size to clear myoglobin from serum.
Setting
The intensive care unit of a university teaching hospital.
Subject
A patient with serotonin syndrome complicated by severe rhabodomyolysis and oliguric acute renal failure
Method
Initially continuous veno-venous hemofiltration was performed at 2 l/hour ultrafiltration (UF) with a standard polysulphone 1.4 m2 membrane (cutoff point, 20 kDa), followed by continuous veno-venous hemofiltration with a SHF membrane (cutoff point, 100 kDa) at 2 l/hour UF, then at 3 l/hour UF and then at 4 l/hour UF, in an attempt to clear myoglobin.
Results
The myoglobin concentration in the ultrafiltrate at 2 l/hour exchange was at least five times greater with the SHF membrane than with the conventional membrane (>100,000 μg/l versus 23,003 μg/l). The sieving coefficients with the SHF membrane at 3 l/hour UF and 4 l/hour UF were 72.2% and 68.8%, respectively. The amount of myoglobin removed with the conventional membrane was 1.1 g/day compared with 4.4–5.1 g/day for the SHF membrane. The SHF membrane achieved a clearance of up to 56.4 l/day, and achieved a reduction in serum myoglobin concentration from >100,000 μg/l to 16,542 μg/l in 48 hours.
Conclusions
SHF hemofiltration achieved a much greater clearance of myoglobin than conventional hemofiltration, and it may provide a potential modality for the treatment of myoglobinuric acute renal failure.
See related commentary
==== Body
Introduction
Rhabdomyolysis and myoglobinuria are responsible for approximately 5% of all causes of acute renal failure (ARF) in the USA [1]. The cause of rhabdomyolysis is often multifactorial [2], and approximately 8–20% of such patients develop myoglobinuric ARF [3]. The myoglobin released during rhabdomyolysis is thought to cause renal dysfunction by producing renal tubular obstruction, lipid peroxidation within the tubular cells, and renal vasoconstriction [1]. Intravascular volume expansion, urinary alkalinization, and forced diuresis are currently used as renal-protective measures [4]. These treatments are not useful in the context of severe oliguria, and may also cause pulmonary edema. In the setting of oliguria, removal of myoglobin using blood purification techniques may be advantageous.
Previous attempts to remove myoglobin using plasma exchange [5,6], intermittent hemodialysis [7], and continuous renal replacement therapies have unfortunately met with limited success [6-11].
We have recently described a novel super high-flux (SHF) membrane with a pore cutoff size of 100 kDa, which has the ability to dialyze inflammatory cytokines and β2-microglobulin [12-15]. We now describe the use of SHF hemofiltration in a patient with severe rhabdomyolysis and oliguric ARF secondary to serotonin syndrome, and report on the kinetics of myoglobin clearance using both conventional and SHF hemofiltration.
Case report
A 53-year-old female presented with complications of a polypharmacy overdose. Her past medical history included depression, hyperlipidemia, and anorexia nervosa. There was a previous history of amphetamine abuse as well as two prior episodes of polypharmacy overdose. The patient was discovered at home with a Glasgow Coma Score of 11 and was noted to be hypotensive and tachycardic. Soon after transfer to hospital, the patient became unconscious (Glasgow Coma Score, 3) and was intubated. Empty bottles of paroxetine, moclobemide, pravastatin, and alprazolam were found in her house. The patient was not known to be on neuroleptic antipsychotic medications, and had not used amphetamine for at least 1 year.
The patient subsequently developed features of serotonin syndrome [16,17], with progressive hypotension despite a central venous pressure of 10 mmHg and commencement of norepinephrine to a maximum dose of 2.5 μg/kg/min, as well as 10 l fluid resuscitation on the first day of admission. Additional features of serotonin syndrome included fever (42°C) and generalized hypertonicity, which was complicated by severe rhabdomyolyis (peak creatine kinase level, 109,000 IU/l). Furthermore, there was evidence of coagulopathy, with severe bleeding from puncture sites and the gastrointestinal tract. Investigations confirmed the presence of disseminated intravascular coagulopathy.
The patient developed profound metabolic derangement with lactic acidemia (lactate peak, 11.7 mmol/l; pH nadir, 7.14), oliguric ARF, severe hyperkalemia (K+, 9.1 mmol/l at presentation), hypocalcemia, hyperphosphatemia, and hypomagnesemia, and with the subsequent development of ischemic hepatitis (alanine aminotransferase, 4048 U/l; alkaline phosphatase, 92 U/l). On the first intensive care unit day, continuous veno-venous hemofiltration (CVVH) with a conventional hemofilter was employed for the management of oliguric ARF and hyperkalemia.
Treatment with fresh frozen plasma, cryoprecipitate, platelet and red blood cells was required. Specific treatment of serotonin syndrome was based on therapies described elsewhere [16,17], and included active cooling, cyproheptadine, a midazolam infusion, and neuromuscular blockade with vecuronium.
Conventional CVVH and SHF CVVH
A 13 Fr dual-lumen catheter (Niagara, Bard, Toronto, Canada) was inserted into a femoral vein and CVVH was performed using a Kimal Hygieia Plus machine (Medtel, Perth, Australia). The blood flow was set at 200 ml/min and the ultrafiltration (UF) rate was set at 2 l/hour. A bicarbonate-based commercial replacement fluid (Hemosol; Gambro, Sydney, Australia) was administered at the pre-filter site. Initially a polysulphone 1.4 m2 membrane (molecular cutoff, 20 kDa) (APS; ASAHI-medical, Tokyo, Japan) was used for conventional CVVH. On the following day, the hemofilter was changed to a novel SHF membrane (molecular cutoff point, 100 kDa) (Polyflux P2SH; Gambro Dialysatoren, Hechingen, Germany). The UF volume rate was commenced at 2 l/hour and sequentially increased to 4 l/hour over an 8-hour period, before being maintained at 2 l/hour.
Measurement and calculations
Blood samples were collected at pre-filter and post-filter sites, and the serum was separated immediately. Ultrafiltrate was collected simultaneously and all specimens were frozen at -70°C until measurements were performed. Myoglobin was measured by a microparticle enzyme immunoassay (AxSYM system; Abbot Laboratories, Perth, Australia). Using this AxSYM system assay, the normal lower limit for serum myoglobin is 150 μg/l and the upper limit of testing is 100,000 μg/l.
The sieving coefficient (SC) was calculated using the formula: SC = 2Cuf/(Cpre + Cpost), where Cuf represents the concentration of myoglobin in the ultrafiltrate, and Cpre and Cpost are the pre-filter and post-filter concentrations of myoglobin, respectively.
The amount of removed myoglobin was calculated as Cuf × Vuf, where Cuf represents the concentration of myoglobin in the ultrafiltrate and Vuf represents the UF volume per unit of time. The clearance of myoglobin was calculated as Cuf × Vuf/Cpre.
Results
The serum myoglobin concentration was >100,000 μg/l at the initiation of conventional CVVH, and remained so at commencement of SHF CVVH on the following day (Table 1). The myoglobin concentration in the ultrafiltrate at 2 l/hour UF exchange was at least five times greater with the SHF membrane than with the conventional membrane (>100,000 μg/l versus 23,003 μg/l).
The SCs with SHF hemofiltration at 3 l/hour UF and 4 l/hour UF exchange rates were 72.2% and 68.8%, respectively, resulting in the removal of 4.3–5.1 g myoglobin/day and in myoglobin clearances of up to 56.4 l/day. Treatment with SHF hemofiltration resulted in a reduction of the serum myoglobin concentration from >100,000 μg/l to 16,542 μg/l in 48 hours (Fig. 1). There was a concurrent reduction in the degree of pigmentation of the UF fluid over this period (Fig. 2).
Discussion
In the case described in the present study, SHF CVVH achieved myoglobin clearance values significantly greater than control treatment with standard high-flux hemofiltration, and values greater than those previously reported in the literature [8-11]. Clearance of myoglobin using SHF CVVH was greater than the total body water over a 24-hour period.
Multiple modalities of renal replacement therapy have been used in the past in an attempt to clear myoglobin. However, these modalities have shown limited efficiency [8-11] (Table 2). We have previously presented the results of myoglobin clearances in three critically ill patients with rhabdomyolysis treated with continuous hemodiafiltration, which was associated with a mean daily myoglobin removal of 1.8 g/day with a myoglobin clearance of 4.6 ml/min [8].
Continuous arterio-venous hemofiltration and CVVH have been used to investigate myoglobin clearance in a swine model of myoglobinuric ARF. The use of continuous arterio-venous hemofiltration and a Hospal AN 69S hemofilter resulted in myoglobin clearances of 2.05 ± 1.48 l/day, with 410 ± 234 mg myoglobin excreted in the ultrafiltrate over the 6-hour period [9].
CVVH has more recently been used in a single case of myoglobin-induced renal failure [11]. Using a Hospal AN69 Multiflow-100 hemofilter, the SC for myoglobin was reported at 40–60%. The corresponding clearance of myoglobin over this period was reported to be up to 22 ml/min (32 l/day).
In the present study, the use of CVVH with a novel SHF hemofilter and UF rates of 3–4 l/min was associated with a SC of 69–72% and a clearance of up to 56.4 l/day. This was approximately five times greater than the clearance achieved with conventional hemofiltration (control) in the same patient. These results are in keeping with the known difference in the pore size of the two filters studied (polysulfone cutoff point, 20 kDa versus polyamide cut-off point, 100 kDa) in relation to the molecular weight of myoglobin (17 kDa), and are also consistent with observations that SHF CVVH can achieve high clearances of other middle molecules such as β2-microglobulin and cytokines [12-15].
Two previous studies utilizing CVVH and high-flux hemofiltration membranes have documented 24-hour myoglobin clearances of approximately 4.3 g [18,19]. The SC using a Gambro polyflux 11S membrane was only 37%, however, and neither of these studies reported on myoglobin clearance or the cutoff size of the pores of the membrane used.
A potential adverse effect of SHF (high cutoff point) hemofiltration is the loss of serum albumin (69 kDa). The SC for albumin with SHF hemofiltration has been shown to be at 0.11 at 1 l/hour exchange [12], so that a patient with a plasma albumin concentration of 20 g/l would be expected to lose 2.2 g/hour albumin. Over a 24-hour treatment period, this might necessitate albumin replacement using 200 ml of 20% albumin. Our patient received volume expansion with albumin (100 g over a 24-hour period) as part of her treatment, and her albumin level increased from 21 to 32 g/l (Table 3) despite SHF hemofiltration. Nonetheless, there is a possibility for losses of albumin and, perhaps, for losses of clotting factors or protein-bound drugs such as fentanyl and midazolam. We thus consider that monitoring of albumin levels and the clotting status, and adjustment of the drug dosage might be important when treating patients with SHF hemofiltration.
Conclusion
The current case offers 'proof of concept' that SHF CVVH can be employed to clear myoglobin effectively in patients with rhabdomolysis and ARF. The technique may be of greatest advantage in patients with oliguric renal failure and in those already requiring renal replacement therapy. It remains to be determined whether this technique is capable of altering renal recovery or mortality in patients with rhabdomyolysis and myoglobinuria-related ARF.
Key messages
• New super high-flux membranes when used in hemofiltration mode can achieve the highest clearances ever reported for myoglobin and may be useful in patients with severe rhabdomyolysis
• New super high-flux membranes when used in hemofiltration result in some albumin losses. These losses, however, decrease in time and are not vary large
Abbreviations
ARF = acute renal failure; CVVH = continuous veno-venous hemofiltration; SC = sieving coefficient; SHF = super high-flux; UF = ultrafiltration.
Competing interests
The author(s) declare that they have no competing interests.
Authors' contributions
TN, DJ, IB, NF, and SB conducted the study and collected the data, and helped write the paper. HG, SM, HHN, and RB helped develop and test the membrane, and also helped write the paper.
Figures and Tables
Figure 1 Progressive reduction of serum myoglobin concentration over a 48-hour period using super high-flux (SHF) continuous veno-venous hemofiltration (CVVH). The serum concentration of myoglobin remained >100,000 μg/l until commencement of SHF CVVH.
Figure 2 Images of ultrafiltration fluid demonstrating progressive reduction in pigmentation over a 3-day period concurrent with treatment using super high-flux continuous veno-venous hemofiltration. ICU, intensive care unit.
Table 1 Sieving coefficient and clearance of myoglobin using conventional and super high-flux (SHF) continuous veno-venous hemofiltration in a patient with severe rhabdomyolysis secondary to serotonin syndrome
Filter type Ultrafiltration rate (l/hour) Myoglobin concentration (μg/l) Sieving coefficient (%) Myoglobin removal (g/day) Myoglobin clearance (ml/min)
Pre-filter Post-filter Ultrafiltrate
Conventional (polysulphone) 2 >100,000 >100,000 23003 <0.23 1.1 <8
SHF 2 >100,000 >100,000 >100,000 Unable to calculate >4.8 Unable to calculate
SHF 3 100,000 68,776 60,912 0.722 4.4 30.5
SHF 4 91,058 64,587 53,527 0.688 5.1 39.2
Table 2 Comparison of studies of renal replacement therapy (RRT) in the management of myoglobinuric acute renal failure
Reference Subject(s) Mode of RRT Filter used Sieving coefficient (%) Myoglobin removal (g/day) Clearance (ml/min)
Present study Single patient with ARF due to rhabdomyolysis CVVH, 2 l exchange Polysulphone (APS) <0.23 1.1 <8
SHF CVVH, 3–4 l exchange Gambro Polyflux P2SH 0.69–0.72 4.3–5.1 30.5–39.2
[8] Three patients with myoglobinuric ARF Continuous hemodiafiltration Hospal AN 69S 0.21* 1.8 4.6
[9] Swine model of myoglobinuric ARF CAVH Hospal AN 69S 0.15 1.64 (0.41 g/6 hours) 1.42 ± 1.03
[10] Swine model CVVH Hospal AN 69HF 0.36–0.55 2.3–3.8 (0.95 g/6 hours) 11.2–25.2
[11] Single patient with ARF due to rhabdomyolysis CVVH Hospal AN69 Multiflow 100 0.40–0.60 1.05 (0.7 g/16 hours) 14–22
[19] Patient with malignant hyperthermia and rhabdomyolysis CVVH Gambro polyflux 11S 0.37 4.3 11*
ARF, acute renal failure; CAVH, continuous arterio-venous haemofiltration; CVVH, continuous veno-venous haemofiltration; SHF, super high-flux. * Calculated from the CUF, VUF, and Cpre values provided in these studies
Table 3 Summary of laboratory investigations and blood product administration
Day 1 Day 2 Day 3 Day 4 Day 5
Na+ 144 144 137 131 129
K+ 9.1 4.2 3.9 4.4 4.6
Cl- 109 107 98 99 95
HCO3 20 18 19 19 13
Urea 9 10.3 6.5 4.3 3.7
Creatinine 0.127 0.214 0.111 0.118 0.107
Ca2+ 1.64 1.13 2.3 2.25 2.26
PO4 2.1 3.22 1.8 1.24 1.6
Albumin 21 29 32 29 31
Creatine kinase 5199 90,000 71,120 85,426 87,170
Lactate 6.42 9.13 7.37 8.02 13.21
Hemoglobin 106 8.6 82 86 73
White cell count 4.8 5 2 4.7 2.5
Platelets 173 72 31 64 112
International normalized ratio for prothrombin time 1.5 4.6 1.9 3.5 5.4
Activated partial thromboplastin time 32 143 44 64 83
Packed red blood cell units 2 3 0 0 0
Platelets 5 5 5 0 0
Fresh frozen plasma 0 26 2 5 0
Cryoprecipitate 10 14 3 0 0
Albumin (20% solution) (ml) 200 500 100 200 0
Norepinephrine dose (μg/min) 3 6 25 30 100
==== Refs
Zager RA Rhabdomyolysis and myohaemoglobinuric acute renal failure Kidney Int 1996 49 314 326 8821813
Knochel JP Mechanism of rhabdomyolysis Curr Opin Rheumatol 1993 5 725 731 8117534
Bywaters EG Beall D Crush injuries with impairment of renal function Br Med J 1941 1 427 432
Better OS Stein JH Early management of shock and prophylaxis of acute renal failure in traumatic rhabdomyolysis N Engl J Med 1990 322 825 829 2407958
Kuroda M Katsuki K Uehara H Kita T Asaka S Miyazaki R Akiyama T Tofuku Y Takeda R Successful treatment of fulminating complications associated with extensive rhabdomyolysis by plasma exchange Artif Organs 1981 5 372 378 7325879
Cornelissen JJ Hoanstra W Haarman HJ Derksen RH Plasma exchange in rhabdomyolysis Intensive Care Med 1989 15 528 529 2607040 10.1007/BF00273565
Hart PM Feinfeld DA Briscoe AM Nurse HM Hotchkiss JL Thomson GE The effect of renal failure and hemodialysis on serum and urine myoglobin Clin Nephrol 1982 18 141 143 7140026
Bellomo R Daskalakis M Parkin G Boyce N Myoglobin clearance during acute continuous hemodiafiltration Intensive Care Med 1991 17 509 1797903
Nicolau DP Feng YJ Wu AH Bernstein SP Nightingale CH Evaluation of myoglobin clearance during continuous hemofiltration in a swine model of acute renal failure Int J Artif Organs 1996 19 578 581 8946233
Nicolau D Feng YS Wu AHB Bernstein SP Nightingale CH Myoglobin clearance during continuous veno-venous hemofiltration with or without dialysis Int J Artif Organs 1998 21 205 209 9649061
Amyot SL Leblanc M Thibeault Y Geadah D Cardinal J Myoglobin clearance and removal during continuous venovenous hemofiltration Intensive Care Med 1999 25 1169 1172 10551978 10.1007/s001340051031
Uchino S Bellomo R Goldsmith D Davenport P Cole L Baldwin I Panagiotopoulos S Tipping P Super high flux hemofiltration: a new technique for cytokine removal Intensive Care Med 2002 28 651 655 12029417 10.1007/s00134-002-1495-z
Uchino S Bellomo R Morimatsu H Goldsmith D Davenport P Cole L Baldwin I Panagiotopoulos S Tipping P Morgera S Cytokine dialysis: an ex vivo study ASAIO J 2002 48 650 653 12455777 10.1097/00002480-200211000-00013
Lee WC Uchino S Fealy N Baldwin I Panagiotopoulos S Goehl H Morgera S Neumayer HH Bellomo R Beta2-microgrobulin clearance with super high flux hemodialysis: an ex vivo study Int J Artif Organs 2003 26 723 727 14521169
Lee WC Uchino S Fealy N Baldwin I Panagiotopoulos S Goehl H Morgera S Neumayer HH Bellomo R Super high flux hemodialysis at high dialysate flows: an ex vivo assessment Int J Artif Organs 2004 27 24 28 14984180
Martin TG Serotonin syndrome Ann Emerg Med 1996 28 520 526 8909274
Mills KC Serotonin syndrome. A clinical update Crit Care Clin 1997 13 763 783 9330840
Bastani A Frenchie D Significant myoglobin removal during continuous veno-venous haemofiltration using F80 membrane Nephrol Dial Transplant 1997 12 2035 2036 9306377 10.1093/ndt/12.9.2035
Schenk MR Beck DH Nolte M Wolfgang J Continuous veno-venous hemofiltration for the immediate management of massive rhabdomyolysis after fulminant malignant hyperthermia in a body builder Anesthesiology 2001 94 1139 1141 11465608 10.1097/00000542-200106000-00031
| 15774055 | PMC1175920 | CC BY | 2021-01-04 16:04:50 | no | Crit Care. 2005 Jan 21; 9(2):R90-R95 | utf-8 | Crit Care | 2,005 | 10.1186/cc3034 | oa_comm |
==== Front
Crit CareCritical Care1364-85351466-609XBioMed Central London cc30351577404310.1186/cc3035ResearchPractice of sedation and analgesia in German intensive care units: results of a national survey Martin Jörg [email protected] Axel 2Franck Martin 3Wernecke Klaus D 4Fischer Matthias 5Spies Claudia 61 Senior physican, Department of Anesthesiology, Intensive Care Medicine and Pain Therapy, Hospital am Eichert, Göppingen, Germany2 Assistant physician, Department of Anesthesiology, Intensive Care Medicine and Pain Therapy, Hospital am Eichert, Göppingen, Germany3 Assistant physician, Department of Anesthesiology and Intensive Care Medicine, University Hospital Charité, Berlin, Germany4 Chairman, Institute of Medical Biometrics, University Hospital Charité, Berlin, Germany5 Chairman, Department of Anesthesiology, Intensive Care Medicine and Pain Therapy, Hospital am Eichert, Göppingen, Germany6 Professor of Anesthesiology and Chairman, Department of Anesthesiology and Intensive Care Medicine, University Hospital Charité, Berlin, Germany2005 26 1 2005 9 2 R117 R123 21 11 2004 2 12 2004 Copyright © 2005 Martin 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
Sedation and analgesia are provided by using different agents and techniques in different countries. The goal is to achieve early spontaneous breathing and to obtain an awake and cooperative pain-free patient. It was the aim of this study to conduct a survey of the agents and techniques used for analgesia and sedation in intensive care units in Germany.
Methods
A survey was sent by mail to 261 hospitals in Germany. The anesthesiologists running the intensive care unit were asked to fill in the structured questionnaire about their use of sedation and analgesia.
Results
A total of 220 (84%) questionnaires were completed and returned. The RAMSAY sedation scale was used in 8% of the hospitals. A written policy was available in 21% of hospitals. For short-term sedation in most hospitals, propofol was used in combination with sufentanil or fentanyl. For long-term sedation, midazolam/fentanyl was preferred. Clonidine was a common part of up to two-thirds of the regimens. Epidural analgesia was used in up to 68%. Neuromuscular blocking agents were no longer used.
Conclusion
In contrast to the US 'Clinical practice guidelines for the sustained use of sedatives and analgesics in the critically ill adult', our survey showed that in Germany different agents, and frequently neuroaxial techniques, were used.
==== Body
Introduction
Critical care therapies such as ventilation, invasive procedures or other measures inducing pain or stress require analgesia and sedation of the patient. Adequate analgesia and sedation is supposed to prevent stress-induced reactions such as hypermetabolism, sodium and water retention, hypertension, tachycardia and altered wound healing [1-3] and to optimize patient comfort. Whipple and colleagues [4] pointed out that 70% of the patients in an intensive care unit (ICU) indicate pain as the worst recollection, although 70–90% of the nurses and physicians taking care of them claimed their patients to be pain free. If sedation is too deep it can have negative side effects [5-7] such as increased risk of pneumonia, venous thrombosis, bowel motility problems, hypotension and a prolonged stay in the ICU, resulting in increased costs [8-10]. The requirements for ideal analgesia and sedation are the ability to sedate the patient deeply for necessary procedures, but with medication of short duration so that the patient can be quickly responsive and cooperative [11].
Goal-oriented sedation [5,6,12,13] complies with the establishment of a modern ventilation regimen to allow early spontaneous breathing [14]. This is shown in the use of short-acting agents for analgesia and sedation, as was demonstrated in surveys from the UK [15] and Denmark [16].
Soliman and colleagues [17] conducted a survey in 16 European countries about the current practice of sedation. They found distinct differences between countries in the practice of analgesia and sedation: 75% of ventilated patients in the UK received medication for analgesia and sedation continuously, whereas only 30% of Italian ventilated patients did so. The most commonly used medication for continuous sedation in Europe is propofol and midazolam, whereas in the USA [18] midazolam and propofol is administered as the preferred medication for short-term sedation and lorazepam for long-term sedation. Analgesic agents differ broadly between countries, and neuroaxial techniques are not often reported [15-18].
The goal of the present survey was to ask for the current practice of analgesia and sedation in German ICUs.
Methods
Data collection
From an address database of the German Society for Anesthesiology and Intensive Care Medicine (Deutsche Gesellschaft für Anästhesiologie und Intensivmedizin [DGAI]) with a total of 808 addresses of ICUs (45 university hospitals and 763 general hospitals) a simple random sample of one-third of the addresses (254 general hospitals and 15 university ICUs) were selected and written to. For eight hospitals the letter was undeliverable. It was written up to four times to the hospitals during May 2002 to October 2002. The return rate was 84% (220 of 261). All data were included in the analysis. At this return rate the 'non-responder bias' can be neglected [19]. The hospitals included in the analysis were 206 general hospitals and 14 university hospitals. Numbers of hospital beds varied between less than 300 beds in 71 hospitals, 300–499 in 88 hospitals, 500–1000 in 43 hospitals, and more than 1000 beds in 14 hospitals; 2 hospitals did not answer. The questionnaire itself is provided in Additional file 1.
Duration of sedation
The periods of sedation were clustered in accordance with the American guidelines [18] into the following groups: duration of sedation less than 24 hours, 24–72 hours and more than 72 hours. In addition we asked about the duration of sedation during weaning from ventilation.
Statistics
The data were collected in a Microsoft Access 97 database and analyzed with the programs Microsoft Excel 9.0 and SPSS for Windows (SPSS Inc., version 10.07). Univariate statistitical analyses were calculated depending on the scaling of the data with the Mann–Whitney U-test or the χ2 test. If multiple tests in multigroup comparison were necessary, we used the Bonferroni–Holm sequential rejective multiple test procedure [20].
Significant two-sided differences were defined as P < 0.05.
Results
General data
General hospitals were equipped with a median of 9 intensive care beds (university hospitals 14), the median number of patients in the ICU was 930 (university hospitals 1481) and the resulting median nursing care days per year were 2565 (university hospitals 4950).
Procedure instructions for analgesia and sedation
Oral procedure instructions (departmental common consensus) existed in 43% of the hospitals.
A standard operating procedure set out in writing existed in 21% of the hospitals.
Use of sedation scales
Sedation monitoring was done by 30% of the hospitals. It was noticeable that only 8% of the hospitals provided data for the question about which sedation scale was used. The Ramsay scale [21] was named exclusively as the sedation scale used.
Medication costs
Sixty-two percent of all hospitals answered 'yes' when asked about considering costs in their choice of medication. The analysis showed that there were no significant differences in the choice of agents compared with the hospitals that said 'no' to this question (P = 0.758).
Choice of agents depending on the duration of sedation
Ninety-two percent of the hospitals stated that they selected the medication depending on the expected duration of analgesia and sedation.
Day–night rhythm
Eighty-one percent of the hospitals tried to maintain a day–night rhythm in their patients.
Neuromuscular blockade
Neuromuscular blockade no longer had a role in ICUs run by anesthesiologists.
Withdrawal/transitional syndrome
Questioned about the frequency of withdrawal/transitional syndrome, the hospitals stated an average rate of 20%.
Sedative agents
For sedation up to 24 hours, propofol was used significantly more often (81%) as a continuous agent than midazolam (45%, P < 0.05). For sedation between 24 and 72 hours, midazolam was used significantly more often (77%) than propofol (56%). For sedation longer than 72 hours, all hospitals preferred midazolam (90%) as a sedative and only 25% of the hospitals were using propofol. During weaning, 92% of the hospitals used propofol for sedation, and 34% used midazolam.
Analgesic agents
For analgesia up to 24 hours, sufentanil (35%) and fentanyl (40%) were used most often. There was no significant difference in the use of these two agents. Thirty-eight percent of the hospitals saw an indication for the use of piritramid, non-steroidal anti-inflammatory drugs (NSAIDs) were used by 27%, morphine was used by 9%, alfentanil by 2%, remifentanil by 6%, and hydromorphone was not an option in our questionnaire. It was not possible to derive from the data how often opioids and NSAIDs were used as additional agents for analgesia or as alternative drugs. For analgesic agents between 24 and 72 hours, fentanyl (55%) and sufentanil (53%) were used by most of the hospitals. Morphine was used by 5%, piritramid by 16%, alfentanil by 2% and remifentanil by 2%. Twelve percent of the hospitals saw an indication for the use of NSAIDs. For analgesia of more than 72 hours, fentanyl was used in 64% of hospitals, significantly more often than sufentanil (44%) was used. The use of piritramid occurred in 9% of all hospitals, morphine was used by 7%, alfentanil by 1% and remifentanil by 1%. Ten percent of all hospitals were using NSAIDs. For analgesia during weaning from ventilation, 39% of the hospitals used fentanyl and 42% used sufentanil. No significant difference was shown between the usage of these two agents. In all hospitals piritramid was used by 25%, morphine by 9%, alfentanil by 2%, remifentanil by 6% and NSAIDs by 14% in this phase.
Adjuvant techniques for analgesia and sedation
Clonidine was used by 35% of the hospitals for sedation of less than 24 hours; 7% of the respondent hospitals decided on ketamine (S). Haloperidol was not a selectable option in our questionnaire. For sedation between 24 and 72 hours, clonidine was used in 48% of the hospitals, and ketamine (S) by 20%. For sedation longer than 72 hours, clonidine was used by 56% of all the hospitals, and 19% were using ketamine (S). During weaning from ventilation, the α2 agonist clonidine was used by 62% of all hospitals, and 5% used ketamine (S) in this phase.
Regional anesthetic techniques
A central neuroaxial block with an epidural catheter was used by 68% of all hospitals as a routine for analgesia up to 24 hours. Peripheral blocks were used by 15% of all hospitals. As a regional anesthetic technique for analgesia between 24 and 72 hours, epidural analgesia was used by 60% of all hospitals, and peripheral blocks were used by 13%. For analgesia and sedation longer than 72 hours, epidural analgesia was used by 46% of all hospitals, and peripheral blocks were used by 11%. Epidural analgesia was used by 43% of all hospitals during ventilator weaning, and peripheral blocks were used by 7% of all hospitals.
The use of the different agents and techniques is summarized in Table 1.
Discussion
The most important result is the use of different agents according to the expected length of analgesia and sedation. In the American guidelines [18] for short-term sedation only propofol is recommended, and for long-term sedation midazolam and lorazepam are recommended. In our survey the most commonly used agent for sedation up to 24 hours and during weaning from ventilation was propofol. Midazolam was used mainly for sedation longer than 72 hours. Lorazepam was not used by any department. This is due mainly to handling (2 mg ampoule) and costs, because lorazepam is one of the more expensive agents in Germany for sedation.
Whereas the American guidelines recommend fentanyl, hydromorphone and morphine for analgesia in all phases [18], in our survey fentanyl and sufentanil were used most often for analgesia for up to 24 hours, between 24 and 72 hours and for weaning from ventilation. In addition, NSAIDs were used preferentially in short-term sedation (less than 24 hours). For longer than 72 hours, fentanyl was preferred for analgesia. Whereas in the British survey [15] from the year 2000 alfentanil was very often used, this opioid did not have a role in German ICUs. In the Danish survey the preferred drugs for analgesia were morphine (94%), fentanyl (76%) and sufentanil (43%) [16]. Our survey shows that morphine did not have a major role in analgesia and sedation in German ICUs. Specifically in Germany, piritramide is a frequently used agent for postoperative analgesia. The reason may be that in Germany some anesthesiologists claim to achieve a lower incidence of nausea and vomiting with piritramid than with morphine [22].
Noticeable was the widespread use of central neuroaxial techniques in analgesia for up to 24 hours. Brodner and colleagues [23] and Beattie and colleagues [24] showed that the perioperative use of epidural analgesia leads to a shortened length of stay in the ICU and also a decrease in cardiac events.
Clonidine as an adjuvant for sedation was used in our survey in a high percentage in all phases, whereas haloperidol, which is recommended in the American guidelines, was not a selectable option in our questionnaire. Most often clonidine was used in the phases longer than 72 hours and during weaning from ventilation. A reasonable use (with regard to time of ventilation and ICU stay) of this agent during weaning was shown by Walz and colleagues [25]. Bohrer and colleagues showed [26] that with clonidine the requirements for opioids and sedation may be reduced.
Ketamine (S) was preferred as an adjuvant in the phases of sedation longer than 24 hours. There have been few studies for long-term sedation with ketamine, as Ostermann and colleagues [11] showed in their review. One of the reasons for the use of ketamine is the lower negative influence on bowel motility than with opioids [27].
In our survey 43% of the hospitals stated that they had have established an oral policy for analgesia and sedation. A procedure in writing was used in 21%. In the survey by Murdoch and colleagues [15] 43% of the British ICUs stated that they had procedures in writing for analgesia and sedation, and 51% had a defined oral policy. In addition, in the 1987 survey of British ICUs by Bion and colleagues [28], 40% stated that they had established a formal procedure. In other surveys it was shown that with the use of standard operating procedures a decrease in the durations of sedation and ventilation, and with this a reduction of costs, is possible [29,30]. Mascia and colleagues [31] showed that the use of written standard operating procedures decreases the duration of ventilation, the length of stay in the ICU and the overall hospital stay.
A sedation scale for the monitoring of analgesia and sedation was used by 31% of the hospitals questioned; 8% stated that they used the Ramsay sedation scale [21] for monitoring sedation. In the survey by Soliman and colleagues [17], 49% of the German hospitals answered that they used the Ramsay sedation scale [19]. In English hospitals in the survey by Murdoch and colleagues [15], 60% were using a sedation scale. In the survey of Danish ICUs [16] from the year 1996/1997, 16% of the hospitals answered that they were using a sedation scale.
More recent studies showed that close monitoring with the help of a scoring system can lead to a decrease in the length of ICU stay and in the length of hospital stay [32].
Nearly all hospitals in our survey stated that they paid attention to cost in their choice of medication. However, the survey showed that there were no significant differences in the use of medications between the hospitals that answered yes and those that answered no. Murdoch and colleagues [15] came to the same conclusion in their survey of English ICUs. More expensive agents may be useful with regard to overall costs because the length of stay in the ICU may be reduced, as was shown by Barrientos-Vega and colleagues [33] and Dahaba and colleagues [34].
Questioned on whether the expected length of sedation had a role in selecting the medication, 92% of the hospitals agreed. The analysis showed that for short-term sedation agents were also used that had a long context-sensitive half-time (fentanyl 45%, midazolam 40%) [35]. Nearly all ICUs tried to maintain a day–night rhythm, although only few studies exist [36,37] that have shown advantages of it for the patients. The Danish survey [16] yielded almost the same results.
In our survey the use of neuromuscular blocking agents had almost disappeared, confirming the results of the European [17], British [15] and Danish [16] surveys of the routine use of neuromuscular blocking agents in intensive care medicine.
The incidence of withdrawal in long-term sedation is 60–80% [38,39]. In our survey, values between 20% and 25% were stated, which is explained by the fact that all patients, even short-term patients, were included.
In our survey the return rate was 84%. Christensen and colleagues [16] in Denmark and Murdoch and colleagues [15] in the UK achieved similar return rates (92.5% and 79%, respectively). In a pan-European questionnaire about the practice of analgesia and sedation by Soliman and colleagues [17] the return rate was 20%.
One of the problems of this survey was the limitation to ICUs run by the department of anesthesiology. We do not have data on whether the patients were mainly postoperative and trauma patients, or whether the ICUs also routinely took care of patients from the department of internal medicine. Hack and colleagues showed in their survey [40] that the most of the interdisciplinary ICUs in general hospitals in Germany are run by the department of anesthesiology [40].
Conclusion
In German anesthesiological ICUs the main short-acting agent used for sedation was propofol, and the benzodiazepine midazolam was used for long-term sedation. For analgesia the opioids fentanyl and sufentanil were used. A very large proportion of hospitals used epidural analgesia. In addition, clonidine was very often used as an adjuvant agent. Only a small proportion of hospitals had established a sedation protocol in writing or a scoring system for the monitoring of analgesia and sedation, although numerous publications showed that the consistent use of these methods can lead to a decrease in ventilator time and length of ICU stay [29-31].
Key messages
• Very little sedation monitoring is used in German intensive care units.
• Only few intensive care units use guidelines for analgesia and sedation.
• Neuroaxial techniques are commonly applied.
Abbreviations
NSAIDs = non-steroidal anti-inflammatory drugs; ICU = intensive care unit.
Competing interests
The author(s) declare that they have no competing interests.
Authors' contributions
JM made substantial contributions to the conception, design, analysis and interpretation of data. AP performed the acquisition, analysis and interpretation of data. MF was involved in drafting the article and revising it critically for important intellectual content. KDW participated in the design of the study and performed the statistical analysis. MF participated in the design and coordination and helped to draft the manuscript. CS made substantial contributions to the conception, design, analysis and interpretation of data. All authors read and approved the final manuscript.
Supplementary Material
Additional File 1
A pdf file containing the questionnaire.
Click here for file
Acknowledgement
We thank Hilge Otter (Charité Berlin) for her support during the data collection.
Figures and Tables
Figure 1 The most commonly used sedative agents for the different sedation periods. *Differences between midazolam and propofol in a phase (χ2 test, P < 0.05). The values were tested with the χ2 test (P < 0.05) and multiple differences with the Bonferroni–Holm multiple test procedure. (Propofol: less than 24 hours versus 24–72 hours versus more than 72 hours versus weaning, all P < 0.0001. Midazolam: less than 24 hours versus 24–72 hours versus more than 72 hours versus weaning, all P < 0.0001.)
Figure 2 The most commonly used analgesic agents during the different sedation periods. *Differences between fentanyl and sufentanil in a phase (χ2 test, P < 0.05). The values were tested with the χ2 test (P < 0.05) and multiple differences with the Bonferroni–Holm multiple test procedure. (Fentanyl: less than 24 hours versus 24–72 hours versus more than 72 hours versus weaning, all P < 0.01. Sufentanil: less than 24 hours versus 24–72 hours; less than 24 hours versus more than 72 hours; 24–72 hours versus more than 72 hours and 24–72 hours versus weaning, all P = 0.01; less than 24 hours versus weaning, P = 0.04; more than 72 hours versus weaning, P = 0.55.)
Figure 3 Epidural analgesia, piritramid and NSAIDs in the different phases of analgesia and sedation. The values were tested with the χ2 test (P < 0.05) and multiple differences with the Bonferroni–Holm multiple test procedure. NSAIDs, non-steroidal anti-inflammatory drugs. (Epidural: less than 24 hours versus more than 72 hours; less than 24 hours versus weaning; 24–72 hours versus more than 72 hours and 24–72 hours versus weaning, all P < 0.01; less than 24 hours versus 24–72 hours, P = 0.015; more than 72 hours versus weaning, P = 0.37. Piritramid: less than 24 hours versus 24–72 hours versus more than 72 hours versus weaning, all P < 0.01. NSAIDs: less than 24 hours versus 24–72 hours, less than 24 hours versus more than 72 hours and less than 24 hours versus weaning, all P < 0.0001; 24–72 hours versus more than 72 hours, P = 0.14; 24–72 hours versus weaning, P = 0.67; more than 72 hours versus weaning, P = 0.087.)
Figure 4 The most commonly used adjunct techniques in the different phases of analgesia and sedation. The values were tested with the χ2 test (P < 0.05) and multiple differences with the Bonferroni–Holm multiple test procedure. (Clonidine: less than 24 hours versus more than 72 hours, less than 24 hours versus weaning and 24–72 hours versus weaning, all P < 0.01; 24 hours versus 24–72 hours, P = 0.23; 24–72 hours versus more than 72 hours, P < 0.017; more than 72 hours versus weaning, P = 0.067. Ketamine (S): less than 24 hours versus 24–72 hours, less than 24 hours versus more than 72 hours, 24–72 hours versus weaning and more than 72 hours versus weaning, all P < 0.0001; less than 24 hours versus weaning, P = 0.22; 24–72 hours versus more than 72 hours, P = 0.087.)
Table 1 Comparison of the used agents and techniques for analgesia and sedation
Agent/technique Percentage (95% confidence interval)
< 24 h 24–72 h >72 h Weaning
Midazolam 45.9 (36.1–55.6) 77.3 (71.0–83.6) 90.5 (86.4–94.5) 34.1 (23.4–44.8)
Diazepam 0.0 2.7 (- 10.3–15.8) 3.2 (- 9.8–16.2) 2.3 (- 10.8–15.3)
Propofol 81.4 (75.7–87.1) 55.9 (47.2–64.7) 26.4 (15.0–37.7) 72.3 (65.3–79.3)
Methohexital 1.4 (- 11.8–14.5) 2.7 (- 10.3–15.8) 4.1 (- 8.9–17.0) 2.3 (- 10.8–15.3)
GHBA 4.5 (8.4–17.5) 7.7 (- 5.0–20.4) 10.9 (- 1.5–23.4) 9.6 (- 3.0–22.1)
Remifentanil 5.9 (- 6.9–18.7) 2.3 (- 10.8–15.3) 1.4 (- 11.8–14.5) 5.9 (- 6.9–18.7)
Alfentanil 1.8 (- 11.3–14.9) 2.3 (- 10.8–15.3) 0.9 (- 12.3–14.1) 1.8 (- 11.3–14.9)
Fentanyl 40.0 (29.8–50.3) 55.9 (47.1–64.7) 65.0 (57.2–72.8) 30.0 (19.0–41.0)
Sufentanil 35.0 (24.4–45.7) 47.7 (38.2–57.3) 43.6 (33.7–53.5) 41.8 (31.8–51.9)
Piritramid 38.2 (27.8–48.6) 15.5 (3.3–27.6) 9.1 (- 3.5–21.7) 25.5 (14.1–36.9)
Morphine 8.6 (- 4.0–21.3) 4.5 (- 8.4–17.5) 7.3 (- 5.5–20.0) 8.6 (- 4.0–21.3)
PCA 25.5 (14.0–36.9) 15.5 (3.3–27.6) 9.5 (- 3.0–22.1) 12.3 (- 0.1–24.7)
Ketamine (S) 6.8 (- 5.9–19.6) 20.0 (8.2–31.8) 19.1 (13.6–36.4) 5.0 (- 7.9–17.9)
Clonidine 35.9 (25.3–46.5) 48.2 (38.7–57.7) 56.4 (47.7–65.1) 62.7 (54.6–70.8)
NSAIDs 26.8 (15.6–38.2) 13.2 (0.8–25.5) 10.5 (- 2.1–23.0) 14.1 (1.8–26.3)
NMBAs 3.6 (- 9.3–16.6) 0.0 0.0 0.0
PNB 15.5 (3.3–27.6) 12.7 (0.4–25.1) 10.0 (- 2.5–22.5) 7.7 (- 5.0–20.4)
PCEA 7.3 (- 5.5–20.0) 5.9 (- 6.9–18.7) 4.1 (- 8.9–17.0) 4.6 (- 8.4–17.5)
Epidural 68.2 (60.8–75.7) 59.1 (50.7–67.6) 45.9 (36.2–55.6) 42.3 (32.2–52.3)
GHBA, γ-hydroxybutyric acid; NMBAs, neuromuscular blocking agents; NSAIDs, non-steroidal anti-inflammatory drugs; PCA, patient-controlled analgesia; PCEA, patient-controlled epidural analgesia; PNB, peripheral nerve block.
==== Refs
Koepke JP Effect of environmental stress on neural control of renal function Miner Electrolyte Metab 1989 15 83 87 2644527
Bonica JJ Importance of effective pain control Acta Anaesthesiol Scand Suppl 1987 85 1 16 3310496
Lewis KS Whipple JK Michael KA Quebbeman EJ Effect of analgesic treatment on the physiological consequences of acute pain Am J Hosp Pharm 1994 51 1539 1554 8092155
Whipple JK Lewis KS Quebbeman EJ Wolff M Gottlieb MS Medicus-Bringa M Hartnett KR Graf M Ausman RK Analysis of pain management in critically ill patients Pharmacotherapy 1995 15 592 599 8570431
Merriman HM The techniques used to sedate ventilated patients. A survey of methods used in 34 ICUs in Great Britain Intensive Care Med 1981 7 217 224 6456287
Gast PH Fisher A Sear JW Intensive care sedation now Lancet 1984 2 863 864 6148586 10.1016/S0140-6736(84)90892-4
Miller-Jones CMH Williams JH Sedation for ventilation. A retrospective study to ventilated patients Anaesthesia 1980 35 1104 1106 7004256
Burns AM Shelly MP Park GR The use of sedative agents in critically ill patients Drugs 1992 43 507 515 1377117
Kollef MH Levy NT Ahrens TS Schaiff R Prentice D Sherman G The use of continuous i.v. sedation is associated with prolongation of mechanical ventilation Chest 1998 114 541 548 9726743
Durbin CG Jr Sedation in the critically ill patient New Horiz 1994 2 64 74 7922431
Ostermann ME Keenan SP Seiferling RA Sibbald WJ Sedation in the intensive care unit: a systematic review JAMA 2000 283 1451 1459 10732935 10.1001/jama.283.11.1451
Tonner PH Weiler N Paris A Scholz J Sedation and analgesia in the intensive care unit Curr Opin Anaesthesiol 2003 16 113 121 10.1097/00001503-200304000-00003
Tung A Rosenthal M Patients requiring sedation Crit Care Clin 1995 11 791 802 8535979
Putensen C Zech S Wrigge H Zinserling J Stuber F Von Spiegel T Mutz N Long-term effects of spontaneous breathing during ventilatory support in patients with acute lung injury Am J Respir Crit Care Med 2001 164 43 49 11435237
Murdoch S Cohen A Intensive care sedation: a review of current British practice Intensive Care Med 2000 26 922 928 10990107 10.1007/s001340051282
Christensen B Thunedborg L Use of sedatives, analgesics and neuromuscular blocking agents in Danish ICUs 1996/97. A national survey Intensive Care Med 1999 25 186 191 10193546 10.1007/s001340050814
Soliman H Melot C Vincent J Sedative and analgesic practice in the intensive care unit: the results of a European survey Br J Anaesth 2001 87 186 192 11493487 10.1093/bja/87.2.186
Jacobi J Fraser GL Coursin DB Riker RR Fontaine D Wittbrodt ET Chalfin DB Masica MF Bjerke HS Coplin WM Clinical practice guidelines for the sustained use of sedatives and analgesics in the critically ill adult Crit Care Med 2002 30 119 141 11902253 10.1097/00003246-200201000-00020
National Center for Education Statistics Standard 2-2
Holm S A simple sequentially rejective multiple test procedure Scand J Stat 1979 6 65 70
Ramsay MA Savege TM Simpson BR Goodwin R Controlled sedation with alphaxalone-alphadolone BMJ 1974 2 656 659 4835444
Breitfeld C Peters J Vockel T Lorenz C Eikermann M Emetic effects of morphine and piritramide Br J Anaesth 2003 91 218 223 12878621 10.1093/bja/aeg165
Brodner G Van Aken H Hertle L Fobker M Von Eckardstein A Goeters C Buerkle H Harks A Kehlet H Multimodal perioperative management – combining thoracic epidural analgesia, forced mobilization, and oral nutrition – reduces hormonal and metabolic stress and improves convalescence after major urologic surgery Anesth Analg 2001 92 1594 1600 11375853
Beattie WS Badner NH Choi P Epidural analgesia reduces postoperative myocardial infarction: a meta-analysis Anesth Analg 2001 93 853 858 11574345 10.1097/00000539-200110000-00010
Walz M Mollenhoff G Muhr G Verkürzung der Weaningphase nach maschineller Beatmung durch kombinierte Gabe von Clonidin und Sufentanil Chirurg 1999 70 66 73 10068833 10.1007/s001040050608
Bohrer H Bach A Layer M Werning P Clonidine as a sedative adjunct in intensive care Intensive Care Med 1990 16 265 266 2358560
Zielmann S Grote R Auswirkungen der Langzeitsedierung auf die intestinale Funktion Anaesthesist 1995 44 Suppl 3 S549 S558 8592966
Bion JF Ledingham IM Sedation in intensive care – a postal survey Intensive Care Med 1987 13 215 216 3584654
Brattebo G Hofoss D Flaatten H Muri AK Gjerde S Plsek PE Effect of a scoring system and protocol for sedation on duration of patients' need for ventilator support in a surgical intensive care unit BMJ 2002 324 1386 1389 12052813 10.1136/bmj.324.7350.1386
MacLaren R Plamondon JM Ramsay KB Rocker GM Patrick WD Hall RI A prospective evaluation of empiric versus protocol-based sedation and analgesia Pharmacotherapy 2000 20 662 672 10853622 10.1592/phco.20.7.662.35172
Mascia MF Koch M Medicis JJ Pharmacoeconomic impact of rational use guidelines on the provision of analgesia, sedation, and neuromuscular blockade in critical care Crit Care Med 2000 28 2300 2306 10921556 10.1097/00003246-200007000-00019
Kress JP Pohlman AS O'Connor M Hall JB Daily interruption of sedative infusions in critically ill patients undergoing mechanical ventilation N Engl J Med 2000 342 1471 1477 10816184 10.1056/NEJM200005183422002
Barrientos-Vega R Sanchez-Soria MM Morales-Garcia C Cuena-Boy R Castellano-Hernandez M Pharmacoeconomic assessment of propofol 2% used for prolonged sedation Crit Care Med 2001 29 317 322 11246312 10.1097/00003246-200102000-00018
Dahaba AA Grabner T Rehak PH List WF Metzler H Remifentanil versus morphine analgesia and sedation for mechanically ventilated critically ill patients: a randomized double blind study Anesthesiology 2004 101 640 646 15329588 10.1097/00000542-200409000-00012
Hughes MA Glass PS Jacobs JR Context-sensitive half-time in multicompartment pharmacokinetic models for intravenous anesthetic drugs Anesthesiology 1992 76 334 341 1539843
Meyer TJ Eveloff SE Bauer MS Schwartz WA Hill NS Millman RP Adverse environmental conditions in the respiratory and medical ICU settings Chest 1994 105 1211 1216 8162751
Walder B Francioli D Meyer JJ Lancon M Romand JA Effects of guidelines implementation in a surgical intensive care unit to control nighttime light and noise levels Crit Care Med 2000 28 2242 2247 10921547 10.1097/00003246-200007000-00010
Ely EW Gautam S Margolin R Francis J May L Speroff T Truman B Dittus R Bernard R Inouye SK The impact of delirium in the intensive care unit on hospital length of stay Intensive Care Med 2001 27 1892 1900 11797025 10.1007/s00134-001-1132-2
Tobias JD Tolerance, withdrawal, and physical dependency after long-term sedation and analgesia of children in the pediatric intensive care unit Crit Care Med 2000 28 2122 2132 10890677 10.1097/00003246-200006000-00079
Hack G Götz E Sorgatz H van Eimeren W Wulff A Umfrage zur Situation der Anästhesiologie in Deutschland Anästh Intensivmed 2000 41 535 541
| 15774043 | PMC1175921 | CC BY | 2021-01-04 16:04:50 | no | Crit Care. 2005 Jan 26; 9(2):R117-R123 | utf-8 | Crit Care | 2,005 | 10.1186/cc3035 | oa_comm |
==== Front
Crit CareCritical Care1364-85351466-609XBioMed Central London cc30381577404510.1186/cc3038ResearchPulmonary capillary pressure in pulmonary hypertension Souza Rogerio [email protected] Marcelo Britto Passos [email protected] Sergio Eduardo [email protected] Daniel [email protected] Carmen Silvia Valente [email protected] Guilherme Paula Pinto [email protected] Carlos Roberto Ribeiro [email protected] Pulmonary Division, Respiratory ICU – Heart Institute (InCor), University of São Paulo Medical School, São Paulo, Brazil2005 11 2 2005 9 2 R132 R138 2 9 2004 2 11 2004 22 11 2004 7 12 2004 Copyright © 2005 Souza 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 cited.
Introduction
Pulmonary capillary pressure (PCP), together with the time constants of the various vascular compartments, define the dynamics of the pulmonary vascular system. Our objective in the present study was to estimate PCPs and time constants of the vascular system in patients with idiopathic pulmonary arterial hypertension (IPAH), and compare them with these measures in patients with acute respiratory distress syndrome (ARDS).
Methods
We conducted the study in two groups of patients with pulmonary hypertension: 12 patients with IPAH and 11 with ARDS. Four methods were used to estimate the PCP based on monoexponential and biexponential fitting of pulmonary artery pressure decay curves.
Results
PCPs in the IPAH group were considerably greater than those in the ARDS group. The PCPs measured using the four methods also differed significantly, suggesting that each method measures the pressure at a different site in the pulmonary circulation. The time constant for the slow component of the biexponential fit in the IPAH group was significantly longer than that in the ARDS group.
Conclusion
The PCP in IPAH patients is greater than normal but methodological limitations related to the occlusion technique may limit interpretation of these data in isolation. Different disease processes may result in different times for arterial emptying, with resulting implications for the methods available for estimating PCP.
See related commentary
==== Body
Introduction
Pulmonary capillary pressure (PCP) is the major force determining fluid filtration from pulmonary capillaries into the interstitium, and thus it is the major determinant of oedema formation [1,2]. Measurement of PCP is therefore of clinical importance. However, equally important is the methodological difficulty in measuring it. Many methods for estimating PCP have been described, including the Gaar equation [1], the osmometric method [3] and others [4-7]. Because of the inaccuracy of the Gaar equation and because the other methods are not suitable for clinical application, pulmonary artery occlusion is currently the most frequently used method for estimating PCP in a broad range of clinical and/or experimental conditions [8].
The pulmonary artery occlusion method is based on the assumption that one can determine the pulmonary capillaries' emptying pattern from the decaying pulmonary arterial occlusion pressure waveform. However, this method does not allow reliable visualization of the two separate emptying phases of the arteries and capillaries in patients, and thus it obscures the point at which the capillary pressure should be derived. In order to obtain a better estimate of PCP, some investigators have represented the pulmonary circulation as an electrical circuit model and used various mathematical approaches to analyze the pressure decay after balloon occlusion. The complexity of the circuit reflects whether the initial decrease in the postocclusion pressure is linear or nonlinear.
The simplest circuit model includes an arterial resistor but no arterial capacitor. However, a three-compartment model proposed by Baconnier and coworkers [9] includes two resistors, representing arterial and venous pulmonary resistance, interposed among three capacitors in series. The latter represent arterial, capillary and venous capacitance compartments, respectively. Under the usual conditions, the capillary compartment would be the dominant capacitor whereas arterial resistance would be the major resistor.
The two primary mathematical approaches to analyzing the pressure decay after balloon occlusion consist of a monoexponential curve, which is fitted from 200 to 300 ms after the instant of occlusion, and a biexponential curve, which is fitted beginning at the instant of occlusion [10-14]. The monoexponential curve corresponds to the emptying of the capillary capacitance compartment, with only one time constant; the biexponential curve theoretically represents the complex emptying of two capacitance compartments (arterial plus capillary). Thus, the main difference between these two types of analysis is whether the arterial capacitance is computed. A useful feature of biexponential modelling is the opportunity to study emptying rates (time constants) of two compartments, which reflect the dynamics of the system and the relationship between capacitance and resistance downstream of each compartment.
Use of the pulmonary artery occlusion technique has identified high levels of PCP in patients with pulmonary arterial hypertension [15,16], even though lung oedema is not a feature of this clinical condition. However, no studies of the time constants of the various compartments of the pulmonary circulation in this condition have yet been conducted; such studies would clarify the mechanisms that lead to these PCP levels. The main objective of the present study was to compare estimates of PCP in patients with idiopathic pulmonary arterial hypertension (IPAH) obtained through different methods; all of these methods assumed the three-compartment model of pulmonary circulation, but each fits the pulmonary artery pressure (PAP) decay to the rather different algorithms proposed in the literature [8,11,17]. Because a 'gold standard' for PCP measurement at the bedside does not exist, we compared estimates of PCP in patients with IPAH versus well described estimates of PCP in patients with acute respiratory distress syndrome (ARDS) [18,19]. We hypothesized that, independent of the precise correspondence between the three-compartment model and the real pulmonary circulation, comparing the emptying rates of the capacitance territories in these two different pathological conditions should reveal important aspects of the pulmonary circulation.
Methods
Patients
We studied two groups of patients with pulmonary hypertension (defined as a mean PAP greater than 25 mmHg): patients with IPAH and patients with ARDS. All patients provided informed consent, and our institutional ethics committee approved the study protocol.
Idiopathic pulmonary arterial hypertension
Twelve patients with IPAH, according to the World Health Organization World Symposium definition [20], were included. All patients were breathing spontaneously. The data used in the present study were collected at the time of acute vasodilator test, before treatment with vasoactive drugs for IPAH.
Acute respiratory distress syndrome
Eleven patients diagnosed with ARDS, according to criteria defined by the American–European Consensus Conference on ARDS [21], were included. These patients had a mean PAP greater than 25 mmHg and were undergoing mechanical ventilation.
Haemodynamic monitoring
A 7-F pulmonary artery catheter (Baxter Healthcare Corporation, Irvine, CA, USA) was introduced in all patients. The catheter position was verified by comparing the variation in pulmonary diastolic pressure during the respiratory cycle with the variation in wedge pressure at the same time [22]. Cardiac output was measured using the standard thermodilution technique.
The ports of each lumen were connected to transducers (HP1290C; Hewlett-Packard, Waltham, MA, USA) and pressure modules (M1006A; Hewlett-Packard) were connected to a Hewlett-Packard monitor (M1176-A; Hewlett-Packard). This monitor was previously modified with a continuous analogue/digital data output that allowed us to record the pressure curves on a personal computer, at 200 Hz. The acquisition and analysis of the decay curves were based on customized LabVIEW software (National Instruments, Austin, TX, USA).
Ventilatory parameters
Haemodynamic data were collected in all patients under mechanical ventilation during a period of standard ventilation, with the following parameters/settings: pressure control ventilation, positive end-expiratory pressure 5 cmH2O, tidal volume 8 ml/kg, inspired fractional oxygen 1, inspiratory time 1 s, and respiratory rate 10 breaths/min. An expiratory pause of 10 s was imposed during data acquisition.
During data acquisition, patients breathing spontaneously were instructed to maintain a relaxed expiratory pause of at least 10 s in order to minimize artifacts caused by respiratory variations in the intrathoracic pressure.
Pulmonary artery pressure curves
In order to obtain two optimal curves, we obtained five PAP curves from each patient. We defined optimal curves to be those with at least 10 s without respiratory oscillations after balloon occlusion and with occurrence of balloon occlusion during the fast rising phase of ventricular systole; these parameters allowed us to determine the precise time of occlusion (Figs 1 and 2).
Curve fitting
We used two methods (monoexponential and biexponential) for curve fitting. The monoexponential method considers only the situation at 200 ms after balloon occlusion, and then extrapolates the fit back to the time of balloon occlusion or to some other time point. In contrast, the biexponential method considers all time points following balloon occlusion. We used a customized routine for curve fitting, employing the algorithm proposed by Foss [23].
Wedge pressure was considered to be the average pressure obtained after a steady state of at least 1 s. Analysis of the fitted curve allowed us to calculate the time constant of pressure decay. We were able to calculate the single time constant for the monoexponential curve and the two time constants, for the fast and slow components, for the biexponential curve.
Pulmonary capillary pressure algorithms
We used the four most commonly employed algorithms to represent the PCP (Figs 1 and 2): mono 0, the value obtained by extrapolating the monoexponential fit back to the time of balloon occlusion; mono 150, the value obtained by extrapolating the monoexponential fit back to 150 ms after balloon occlusion; BI 0, the value obtained by extrapolating the slow component of the biexponential fit back to the time of balloon occlusion; and BI 150, the value obtained 150 ms after balloon occlusion in the biexponential fitted curve.
Statistical analysis
We used repeated measures analysis of variance (two way) to compare PCP values between the groups, and repeated measures analysis of variance (one way) to compare the other haemodynamic data between groups.
Results
Pulmonary haemodynamic data for all the patients, as well as time constants calculated from both mathematical analyses, are shown in Table 1. Patients with IPAH had a significantly lower cardiac index. These results were expected because IPAH is frequently associated with severe cardiac dysfunction whereas ARDS is often associated with hyperdynamic states [24].
Although patients with IPAH had a very high mean PAP, those with ARDS exhibited only mild pulmonary hypertension. The PCP values for both groups, estimated using all of the algorithms, are shown in Fig. 3. The PCP values in the IPAH group were significantly higher than those in the ARDS group (P < 0.001). The different algorithms also yielded PCP values that significantly differed within each group (P < 0.02).
The time constant obtained from the monoexponential fit and the time constant obtained for the slow component of the biexponential fit were much higher in the IPAH than in the ARDS group (P < 0.001; Table 1). No difference between groups was found in the time constant for the fast component. Comparison of PAP decay curves from the two groups (Fig. 4) illustrates effect of these findings – the time required to reach a steady state was longer in IPAH group.
Discussion
The major finding of the present study was that the IPAH group had a significantly higher PCP than did the ARDS group, independent of algorithm used for its estimation; this was accompanied by a marked increase in the time constant for the slow decay.
The PCP is a key determinant of the pathophysiology of the cardiopulmonary system. Commonly, the PCP is considered to be close to the wedge pressure; although this assumption is quite true under normal circumstances, it can lead to incorrect interpretations under pathological conditions [25]. In our study, in agreement with data previously reported by Kafi [15] and Fesler [16] and their colleagues, we found that the effective filtration pressure can be dramatically underestimated by this assumption. Those studies showed high levels of PCP (>30 mmHg) in patients with pulmonary hypertension.
Under acute conditions, PCP levels greater than 30 mmHg can lead to movement of fluid from the capillaries into the interstitium and alveolar spaces, once this pressure exceeds the blood oncotic pressure. However, under chronic conditions an increased lymphatic drainage capability can prevent PCP levels as high as 40 mmHg from causing lung oedema [26]. However, before the high levels of PCP found in our study may be considered valid, we must account for some methodological issues. Hakim and Kelly [27] suggested that the arterial occlusion technique measures the pressure in vessels with diameters that range from 50 to 900 μm, which usually encompass the major territory of the pulmonary blood volume. However, the algorithms used to estimate PCP in our two groups yielded significantly different PCP values in each group; this suggests that each method reflects the pressure at different arterial sites, or perhaps the pressure in territories with different vascular diameters.
Considering the same three-compartmental model, we used both monoexponential and biexponential fitting processes to analyze PCP. Some methodological approaches, such as use of the monoexponential curve fitting process starting 200 ms after occlusion and the use of back extrapolation (looking back at a sampling time corresponding to 150 ms after occlusion), were taken from previous work by Gilbert and Hakim [28]. Both procedures assume that the influence of the fast arteriolar emptying process is negligible after a few milliseconds, allowing study of the emptying of an almost isolated capillary. However, in testing the arterial occlusion method under vasodilating and vasoconstricting conditions, Pellett and coworkers [29] demonstrated that 150 ms was not the optimal time at which to determine PCP in dogs with intact lungs.
Another methodological issue that must be considered when using the arterial occlusion technique is determination of the time of balloon occlusion, because mathematical modelling is based on this time point being the start of curve fitting. In experimental models the double occlusion technique permits perfect recognition of the time of occlusion; however, at the bedside the precise time point at which balloon inflation takes place can only be identified by a clear modification in the PAP curve. In our analysis, we accepted only those curves in which a clear cut in the systolic limb of the PAP curve was identifiable. Holloway [30] and Nunes [31] and their coworkers identified a difference between the PCP estimated using this method and that estimated using superimposed occluded and nonoccluded curves. Although significant, this possible difference does not alter the interpretation of our data because of the magnitude of PCP levels and time constants found in our study.
The significant changes in microvascular dynamics observed in individual disease states mean that the optimal time for estimating the PCP may differ between them. This is supported by the finding of different time constants for the exponential decays in our patients with IPAH as compared with those with ARDS, especially considering the constants for the slowly emptying compartments. All of the fitting methods tested in the present study assumed three basic hypotheses: the fast emptying of arteriolar territories into the capillary territory occurs in a few milliseconds; the time constant for the slow decay (corresponding to capillary emptying) is several times longer than that for arterial emptying; and the dominant capacitance in the pulmonary circulation is the capillary network, which therefore corresponds to the slowest emptying observed after occlusion.
Nevertheless, in settings involving pulmonary arterial hypertension, many considerations oppose these basic hypotheses. The arteriolar territory is heterogeneously abnormal, and therefore longer periods may be required for complete emptying, which also changes the relation with capillary emptying. The capillary territory still has a large cross-sectional area but a shorter length and decreased compliance; consequently, it may not represent the dominant capacitance in the pulmonary circulation.
It is likely that none of the proposed sampling times derived from experimental studies is valid for studies in humans, which are mainly conducted in the setting of pulmonary arterial hypertension. Because of slower arterial emptying, the time needed to estimate the PCP should be longer. The time constants for the fast compartments estimated in the present study were around 0.25 s, suggesting that at least 0.75 s (for 95% emptying) should be allowed for this fast arteriolar emptying. As such, the increased PCP levels reported could reflect a precapillary territory pressure, leading to erroneous interpretation of the data. Probably an individual sample time, based on the fast compartment time constant, should be used for PCP estimation, but this requires confirmation in experimental models of pulmonary hypertension.
Biexponential curve fitting appears to be preferable in this setting because it does not assume any fixed time constant for the fast compartment, but estimates it instead. However, one of the basic requirements for realistic bicompartmental fitting is that there are two exponential decay processes with very different time constants, preferentially differing by an order of magnitude. We are not sure that this condition can be met in patients with IPAH or ARDS (Table 1).
With regard to the high PCP values found in our patients, we do not believe that artifacts caused by non-instantaneous occlusion could account for our results or for the variability in results achieved with different methods. According to a study conducted by Fesler and coworkers [16], these high PCP levels could be explained by increased venous resistance. There is increasing recognition of venous involvement in the pathophysiology of pulmonary hypertension. However, the methodological limitations described above do not allow validation of the assumption that increased venous resistance is the only cause of high PCP levels in patients with pulmonary hypertension.
Conclusion
We conclude that analysis of PAP decay curves permits a better understanding of the pulmonary microvasculature. However, analysis of these curves' time constants suggests that the 150–200 ms allotted for fast arteriolar emptying may be insufficient under pathological conditions. Whereas the mean PCP measured using the artery occlusion technique was greater than normal in our patients with IPAH, the methodological limitations related to this technique may limit the interpretation of these data in isolation.
Key messages
• PCP is elevated in IPAH, although its interpretation must take into account the methodological limitations of measurement using the arterial occlusion technique.
• The time constants of pulmonary artery emptying may differ according to the disease process.
• The time constants may be useful for increasing the accuracy of PCP measurement using the arterial occlusion technique.
Abbreviations
ARDS = acute respiratory distress syndrome; IPAH = idiopathic pulmonary arterial hypertension; PAP = pulmonary artery pressure; PCP = pulmonary capillary pressure.
Competing interests
The author(s) declare that they have no competing interests.
Authors' contributions
RS conducted patient monitoring and data analysis, participated in statistical analysis and drafted the manuscript. MBPA conceived the study, and participated in its design and statistical analysis. SED conducted patient monitoring and data analysis. DD participated in the design of the study and statistical analysis, and helped to draft the manuscript. CSVB participated in data analysis. GPPS participated in study coordination and data analysis. CRRC participated in the design and coordination of the study, and helped to draft the manuscript. All authors read and approved the final manuscript.
Acknowledgements
We are indebited to Professor Marc Hunbert and Dr Carlos Jardim for their support and helpful comments during the preparation and review of this manuscript.
Figures and Tables
Figure 1 Monoexponential curve fitting for estimation of pulmonary capillary pressure (PCP). MONO 0, PCP obtained by extrapolation back to the time of occlusion; MONO 150, PCP obtained 150 ms after balloon occlusion.
Figure 2 Biexponential curve fitting for estimation of pulmonary capillary pressure (PCP). BI 0, PCP obtained by extrapolation of the slow component back to the time of occlusion; BI 150, PCP obtained 150 ms after balloon occlusion.
Figure 3 Estimates of pulmonary capillary pressure (PCP) using four different methods in patients with idiopathic pulmonary arterial hypertension (IPAH) and acute respiratory distress syndrome (ARDS). Significance for difference between IPAH and ARDS groups: P < 0.001. Significance for difference within groups: P < 0.02. BI 0: PCP obtained by extrapolation of the slow component back to the time of occlusion; BI 150, PCP obtained 150 ms after balloon occlusion; MONO 0: PCP obtained by extrapolation back to the time of occlusion; MONO 150: PCP obtained 150 ms after balloon occlusion.
Figure 4 Pulmonary artery pressure decay curve in a patient with acute respiratory distress syndrome (ARDS; gray line) as compared with that in a patient with idiopathic pulmonary arterial hypertension (IPAH; black line). The time necessary to reach a steady state is longer in the patient with IPAH.
Table 1 Haemodynamic data
Group n CI (l/min per m2) Pwedge (mmHg) Mean PAP (mmHg) PVRI (dyn·s/cm5 per m2) Tc mono Tc bi-fast Tc bi-slow
IPAH 12 1.60 ± 0.33* 15.1 ± 4.0 85.3 ± 19.2* 3588.4 ± 1115.1* 1.51 ± 0.23* 0.28 ± 0.05 2.15 ± 0.41*
ARDS 11 4.23 ± 0.94 12.5 ± 2.1 29.0 ± 3.9 307.4 ± 67.5 0.70 ± 0.32 0.25 ± 0.13 1.30 ± 0.43
Values are expressed as mean ± standard deviation. *P ≤ 0.001 – comparison between IPAH and ARDS patients. ARDS, acute respiratory distress syndrome; CI, cardiac index; IPAH, idiopathic pulmonary arterial hypertension; PAP, pulmonary artery pressure; PVRI, pulmonary vascular resistance index; Pwedge, pulmonary artery occlusion pressure; Tc bi fast, time constant of the fast component of the biexponential curve(s); Tc bi slow, time constant of the slow component of the biexponential curve(s); Tc mono, time constant of the monoexponential curve(s).
==== Refs
Gaar KA JrTaylor AE Owens LJ Guyton AC Pulmonary capillary pressure and filtration coefficient in the isolated perfused lung Am J Physiol 1967 213 910 914 6051189
Grimbert FA Effective pulmonary capillary pressure Eur Respir J 1988 1 297 301 3294035
Agostoni E Piiper J Capillary pressure and distribution of vascular resistance in isolated lung Am J Physiol 1962 202 1033 1036 13859627
Bhattacharya J Staub NC Direct measurement of microvascular pressures in the isolated perfused dog lung Science 1980 210 327 328 7423192
Hakim TS Michel RP Chang HK Partitioning of pulmonary vascular resistance in dogs by arterial and venous occlusion J Appl Physiol 1982 52 710 715 7068486
Hakim TS Maarek JM Chang HK Estimation of pulmonary capillary pressure in intact dog lungs using the arterial occlusion technique Am Rev Respir Dis 1989 140 217 224 2751168
Hakim TS Kelly S Occlusion pressures vs. micropipette pressures in the pulmonary circulation J Appl Physiol 1989 67 1277 1285 2793720
Collee GG Lynch KE Hill RD Zapol WM Bedside measurement of pulmonary capillary pressure in patients with acute respiratory failure Anesthesiology 1987 66 614 620 3578875
Baconnier PF Eberhard A Grimbert FA Theoretical analysis of occlusion techniques for measuring pulmonary capillary pressure J Appl Physiol 1992 73 1351 1359 1447079
Corboz M Sanou S Grimbert F Capillary pressure estimates from arterial and venous occlusion in intact dog lung Eur Respir J 1995 8 1122 1129 7589396 10.1183/09031936.95.08071122
Dawson CA Bronikowski TA Linehan JH Haworth ST Rickaby DA On the estimation of pulmonary capillary pressure from arterial occlusion Am Rev Respir Dis 1989 140 1228 1236 2817586
Wakerlin GE JrFinn JC Siegel LC Benson GV Flavin TF Pearl RG Pulmonary capillary pressure measurement from pulmonary artery occlusion pressure decay profile analysis in sheep Anesth Analg 1995 81 17 23 7598249 10.1097/00000539-199507000-00004
Yamada Y Suzukawa M Chinzei M Chinzei T Kawahara N Suwa K Numata K Phasic capillary pressure determined by arterial occlusion in intact dog lung lobes J Appl Physiol 1989 67 2205 2211 2606825
Cope DK Allison RC Parmentier JL Miller JN Taylor AE Measurement of effective pulmonary capillary pressure using the pressure profile after pulmonary artery occlusion Crit Care Med 1986 14 16 22 3940751
Kafi SA Melot C Vachiery JL Brimioulle S Naeije R Partitioning of pulmonary vascular resistance in primary pulmonary hypertension J Am Coll Cardiol 1998 31 1372 1376 9581736 10.1016/S0735-1097(98)00091-6
Fesler P Pagnamenta A Vachiery JL Brimioulle S Abdel Kafi S Boonstra A Delcroix M Channick RN Rubin LJ Naeije R Single arterial occlusion to locate resistance in patients with pulmonary hypertension Eur Respir J 2003 21 31 36 12570105 10.1183/09031936.03.00054202
Maarek J Hakim T Chang H Analysis of pulmonary arterial pressure profile after occlusion of pulsatile blood flow J Appl Physiol 1990 68 761 769 2318783
Rossetti M Guénard H Gabinski C Effects of nitric oxide inhalation on pulmonary serial vascular resistances in ARDS Am J Respir Crit Care Med 1996 154 1375 1381 8912751
Jones R Zapol WM Reid L Pulmonary artery remodeling and pulmonary hypertension after exposure to hyperoxia for 7 days. A morphometric and hemodynamic study Am J Pathol 1984 117 273 285 6238536
Rich S Primary pulmonary hypertension Executive Summary from the World Symposium: Primary Pulmonary Hypertension 1998 Geneva: World Health Organization
Bernard GR Artigas A Brigham KL Carlet J Falke K Hudson L Lamy M Legall JR Morris A Spragg R Report of the American–European Consensus Conference on ARDS, part 1: Definitions, mechanisms, relevant outcomes, and clinical trial coordination Am J Respir Crit Care Med 1994 149 818 824 7509706
Teboul JL Andrivet P Ansquer M Besbes M Rekik N Lemaire F Brun-Buisson C A bedside index assessing the reliability of pulmonary artery occlusion pressure measurements during mechanical ventilation with positive end-expiratory pressure J Crit Care 1992 7 22 29 10.1016/0883-9441(92)90005-R
Foss SD A method of exponential curve fitting by numerical integration Biometrics 1970 26 815 821
Carvalho CR Barbas CS Medeiros DM Magaldi RB Lorenzi Filho G Kairalla RA Deheinzelin D Munhoz C Kaufmann M Ferreira M Temporal hemodynamic effects of permissive hypercapnia associated with ideal PEEP in ARDS Am J Respir Crit Care Med 1997 156 1458 1466 9372661
Radermacher P Santak B Wust HJ Tarnow J Falke KJ Prostacyclin for the treatment of pulmonary hypertension in the adult respiratory distress syndrome: effects on pulmonary capillary pressure and ventilation-perfusion distributions Anesthesiology 1990 72 238 244 2105674
Guyton AC Hall JE Guyton AC, Hall JE Pulmonary circulation, pulmonary edema, pleural fluid Textbook of Medical Physiology 2000 Philadelphia, PA: WB Saunders Co 444 452
Hakim T Kelly S Occlusion pressures vs. micropipette pressures in the pulmonary circulation J Appl Physiol 1989 67 1277 1285 2793720
Gilbert E Hakim TS Derivation of pulmonary capillary pressure from arterial occlusion in intact conditions Crit Care Med 1994 22 986 993 8205832
Pellett AA Johnson RW Morrison GG Champagne MS deBoisblanc BP Levitzky MG A comparison of pulmonary arterial occlusion algorithms for estimation of pulmonary capillary pressure Am J Respir Crit Care Med 1999 160 162 168 10390395
Holloway H Perry M Downey J Parker J Taylor A Estimation of effective pulmonary capillary pressure in intact lungs J Appl Physiol 1983 54 846 851 6841232
Nunes S Ruokonen E Takala J Pulmonary capillary pressures during the acute respiratory distress syndrome Intensive Care Med 2003 29 2174 2179 14586495 10.1007/s00134-003-2036-0
| 15774045 | PMC1175923 | CC BY | 2021-01-04 16:04:51 | no | Crit Care. 2005 Feb 11; 9(2):R132-R138 | utf-8 | Crit Care | 2,005 | 10.1186/cc3038 | oa_comm |
==== Front
Crit CareCritical Care1364-85351466-609XBioMed Central London cc30421577404410.1186/cc3042ResearchCoronary artery bypass surgery and longitudinal evaluation of the autonomic cardiovascular function Soares Pedro Paulo S [email protected] Adalgiza M 2Cravo Sérgio LD 3Nóbrega Antonio Claudio L [email protected] Research Associate, Department of Physiology and Pharmacology, Universidade Federal Fluminense, Niterói, RJ, Brazil2 Physical Therapy Master Program, Centro Universitário do Triângulo Mineiro, Uberlândia, MG, Brazil3 Associate Professor, Department of Physiology, Universidade Federal de São Paulo, São Paulo, SP, Brazil4 Professor, Department of Physiology and Pharmacology, Universidade Federal Fluminense, Niterói, RJ, Brazil2005 26 1 2005 9 2 R124 R131 29 10 2004 8 12 2004 10 12 2004 15 12 2004 Copyright © 2005 Soares 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
Imbalance in autonomic cardiovascular function increases the risk for sudden death in patients with coronary artery disease (CAD), but the time course of the impact of coronary artery bypass grafting (CABG) on autonomic function has been little studied. Thus, the purpose of the present study was to determine the effects of the CABG on the cardiovascular autonomic function.
Methods
Patients undergoing CABG (n = 13) and two matched control groups (patients with CAD who refused surgical treatment [n = 9], and healthy volunteers [n = 9]) underwent a prospective longitudinal study consisting of autonomic evaluation before and after (3, 6, 15, 30, 60, and 90 days) surgery, including measurement of heart rate variability (HRV), respiratory sinus arrhythmia (RSA), and Valsalva maneuver.
Results
After CABG there was a decrease in, and a later recovery of, (1) the HRV in the time domain and in the frequency domain, (2) RSA, and (3) Valsalva maneuver.
Conclusions
CABG caused an impairment, reversible after 60 days, of cardiovascular autonomic function, with a maximal decrease on about the sixth day after surgery.
==== Body
Introduction
Imbalance in autonomic cardiovascular function has been shown to increase the risk for ventricular arrhythmias and sudden death in patients with coronary artery disease (CAD) and after myocardial infarction [1,2]. Under these conditions there is an increased sympathetic adrenergic tone and reduced parasympathetic activity [3], a combination that causes augmented ventricular workload and oxygen demand, increasing the occurrence of ischemic events, and causes modification of the ionic currents across the cellular membrane, leading to direct electrical instability of myocytes.
The autonomic cardiovascular function has been traditionally evaluated by bedside tests such as respiratory sinus arrhythmia (RSA) and Valsalva maneuver, which measure the blood pressure and heart responses to standard stimuli [4,5]. In the past two decades, quantification of heart rate variability (HRV) has been used as an indicator of the autonomic control of sinus rate [6], providing independent predictive power for sudden death and all-cause mortality in CAD [1,2]. The signals necessary for HRV analysis can be obtained from electrocardiogram tracings recorded from short (15 min) resting periods to an entire day (24 hours) with multiple moments of physical activity. The HRV profile can be expressed both in the time domain, by measures of the variations in the R–R interval durations, and in the frequency domain, through spectral analysis [6].
Myocardial revascularization by coronary artery bypass grafting surgery (CABG) is an effective measure for reducing the symptoms and mortality in patients with unstable or severe CAD [7,8]. Despite the increasing importance of autonomic cardiac function for risk stratification in heart disease and, in contrast, the positive clinical outcome of CABG, there have been only few longitudinal studies investigating the impact of CABG on cardiac autonomic function. Although previous studies have indicated a diminished autonomic function after CABG [9-13], they could not describe properly the profile of cardiac autonomic function after surgery because they either lacked higher resolution in the first weeks after CABG or did not follow the patients up long enough to detect full recovery of cardiac autonomic indexes after CABG. The purpose of this study was therefore to evaluate, noninvasively and repeatedly, the cardiovascular autonomic function of patients with CAD during the initial three months after CABG, starting with a pre-surgery evaluation and repeated 3, 6, 15, 30, 60 and 90 days after surgery. Results obtained were compared with those from two control groups: matched patients with CAD but not undergoing CABG, and matched healthy volunteers.
Methods
Subjects
Thirteen patients (five female, eight male) undergoing CABG were recruited from the Procordis Hospital, Niterói, RJ, Brazil, fulfilling the following inclusion criteria: subjects of either sex, with diagnosed CAD (including laboratory tests, such as electrocardiogram, stress test, or coronary angiogram) and clinical indication for CABG. The exclusion criteria were as follows: presence of diabetes mellitus, congestive heart failure, previous cardiac surgery, recent (less than 6 months) myocardial infarction, implanted cardiac pacemaker, presence of atrial fibrillation, use of intraortic balloon, use of mechanical ventilation for more than 24 hours after surgery, myocardial infarction after surgery, and any other condition that could affect the autonomic function.
A control group (CAD; n = 9; five females, four males) for the surgical intervention was selected and paired by age and number of coronary vessels compromised (defined angiographically as more than 50% lumen obstruction). These patients also had clinical indication for CABG but refused to undergo surgery.
A further control group (healthy; n = 9; four females, five males) was included to evaluate the effect of CAD on autonomic function. The volunteers of this second control group were paired by age and the inclusion criteria for this group were to be asymptomatic and to have physiological responses to an exercise test without myocardial ischemia. The present study therefore employed two different groups as controls, one with patients with CABG who refused to undergo surgery and another consisting of healthy adults.
Procedures
All subjects were informed with details about the protocol, which had been approved by the Institutional Review Board, and signed an informed consent form to participate in the study. They were instructed not to exercise or to take alcoholic or caffeinated drinks on the experimental days. All subjects were trained to perform autonomic tests with acceptable technical quality before actual measurements were made. Heart rate was recorded with a digital system for telemetry consisting of a transmitter placed on the subject's chest and a receptor and interface connected to a personal computer (Polar Vantage; Polar Electro Oy, Kempele, Finland). This system detects ventricular depolarization, corresponding to the R wave on the electrocardiogram, with a sampling rate of 500 Hz and a temporal resolution of 1 ms [14], and has been validated previously against standard Holter electrocardiography [15].
The patients in the CABG group were evaluated on the day preceding surgery and 3, 6, 15, 30, 60, and 90 days after surgery. The patients in the CAD group were tested four times in 3 months (0, 30, 60, and 90 days) and the volunteers in the healthy group were tested once.
Respiratory sinus arrhythmia
After resting in the supine position for 15 min, the subjects were trained to perform respiratory cycles lasting 10–12 s, changing lung volume from maximal expiration (residual volume) to maximal inspiration (total lung capacity). The heart rate response to the RSA test was measured by the expiration/inspiration index (E/I index), calculated as the ratio between the longest R–R interval during expiration and the shortest R–R interval during inspiration.
Valsalva maneuver
Shortly after resting for 15 min in sitting position, the subjects blew into a closed system connected to an aneroid manometer, exerting an expiratory pressure of 40 mmHg for 15 s, followed by expiration and spontaneous breathing. The Valsalva index was calculated by the ratio between the longest R–R interval after releasing the pressure and the shortest R–R interval during the forced expiratory pressure.
Heart rate variability
The subjects rested for 10 min after the autonomic tests in the supine position and started a 15 min period of paced breathing at 18 cycles/min (0.30 Hz) and spontaneous tidal volume. The respiratory rate was kept constant with the aid of a metronome.
The original series of R–R intervals were filtered by a semi-automatic method [6], keeping the original time reference. The following parameters of HRV in the time domain were calculated: the standard deviation of all R–R intervals; the standard deviation of the averages of the R–R intervals calculated in 5 min segments; the percentage number of pairs of adjacent R–R intervals differing by more than 50 ms; and the square root of the mean of squares of differences between adjacent R–R intervals (RMSSD).
For frequency-domain analysis, the filtered R–R series were subjected to a cubic spline interpolation and decimated to be equally spaced in time. The decimated time series were used to calculate the power spectrum, calculated by the fast Fourier transform. The following parameters were analyzed: total power, power in the low-frequency (LF, 0.08–0.15 Hz) and high-frequency (HF, 0.15–0.40 Hz) bands, and the LF/HF ratio.
Respiratory function
Forced expiratory vital capacity was measured (Spirodoc; Mir, Rome, Italy) before the autonomic tests. The expiratory maneuvers were performed after full inspiration (total lung capacity) in the seated position and repeated three times, taking the best result for analysis. Another series of forced expiratory maneuvers after full inspiration were used to measure the patient's peak expiratory pressure by manovacuometry (GEAR 07-01, Instrumentation Inc., Springfield, IL, USA). For this evaluation, the subjects were instructed to produce the effort with the chest and abdominal muscles and to keep the expiratory pressure for 3 s. The best result of three consecutive maneuvers was used for analysis.
Statistical analysis
The patients' characteristics were compared by one-way analysis of variance (ANOVA) for the continuous variables or by Fisher's exact test for the proportions. Data from the autonomic evaluations were compared by a two-way ANOVA for repeated measures in which group (CABG, CAD, or healthy) and time were the main factors. The Bonferroni test was used for post-hoc pairwise analysis. Student's t-test was used to compare the results from patients with or without previous myocardial infarction within the CABG and CAD groups. The minimum necessary sample size was determined by setting the statistical power to 0.8 and the alpha error to 0.5. On the basis of the results of the R–R interval from pilot experiments and previous experience with patients undergoing CABG, we considered the minimum detectable difference to be 80 ms and the expected standard deviation of residuals as 50 ms for the ANOVA with three groups, yielding a sample size of nine subjects. All statistical procedures were performed on the SigmaStat® (Jandel Scientific, San Rafael, CA, USA) and statistical significance was accepted at P < 0.05.
Results
The results of the autonomic evaluation in patients with or without previous myocardial infarction in the CABG group (n = 5 and n = 8, respectively) and in the CAD group (n = 2 and n = 7, respectively) were similar. The data were therefore pooled within each group. The patients' characteristics are presented in Table 1. Note that gender proportion, age, body mass index and drugs used were not statistically different between groups.
Respiratory sinus arrhythmia and Valsalva maneuver
The indexes obtained during Valsalva maneuver and RSA presented a similar time profile in the CABG group. This profile was characterized by a marked decrease 3 days after surgery, followed by a significant increase after 15 days and a recovery to values similar to those observed before surgery after 30 days. Valsalva and RSA indexes in the CAD group were comparable to those observed in the CABG before surgery, and they did not change during the 90 days afterwards. Results for the healthy group were higher than those for the CABG and CAD groups (Fig. 1).
Heart rate variability
The time-domain indexes of HRV decreased after CABG, returning to pre-surgery values by 30 or 60 days after surgery (Table 2). The values for the CAD group were similar to those from the CABG pre-surgery evaluation, whereas all values in the healthy group were higher than those for the CAD and CABG groups, except for the mean R–R interval and RMSSD.
Spectral analysis of HRV showed a similar change for total power when compared with the time-domain indexes; that is, a decrease in the first evaluation after CABG, a relative increase at 15 days, and a return to pre-surgery values at 60 days. The CAD and CABG groups presented similar values for this variable at the pre-surgery evaluation and after recovery of the CABG patients at 60 days. Total power for the healthy group was higher than those for the CAD and CABG groups (Fig. 2). HF power was similar in CABG and CAD patients before surgery, decreased after the intervention, and recovered after 30 days. LF power and the LF/HF ratio showed increases in the first days after surgery, followed by gradual decrements towards the pre-surgery values (Fig. 2). The results of the CAD group remained stable throughout the 90 days of evaluation.
Respiratory function
Forced expiratory vital capacity and peak expiratory pressure were similar in the healthy, CAD and CABG groups before surgery and remained constant during the 90 days of follow-up in the CAD group. In the CABG group, both forced expiratory vital capacity and peak expiratory pressure decreased after surgery and returned to pre-intervention values by 15 days afterwards.
Discussion
The present results have shown that CABG is followed by a depression of autonomic cardiac modulation, as demonstrated by reduced indices of conventional autonomic bedside tests and decreased HRV. The impaired autonomic modulation reached the lowest level 3–6 days after surgery, returning to pre-surgery values at about postoperative day (POD) 30–60. Although previous studies have shown similar trends [12,13], the present design employed an original approach, combining higher temporal resolution during the first month (PODs 3, 6, 15 and 31) with a follow-up long enough (POD 90) to detect the recovery of autonomic function to pre-surgery values.
Adequate control groups have been employed previously to evaluate the specific effect of CABG. For example, Hogue and colleagues [13] showed that patients undergoing CABG presented HRV indices 40–50% lower than patients undergoing nonthoracic vascular surgery, an effect that persisted for 5 days. Thus, the CABG itself reduced HRV. The work by Bronner and colleagues [11] employed patients undergoing aortic valve replacement as a specific control for patients undergoing cardiac surgery due to myocardial ischemia. These authors found that HRV decreased to similar proportions in the two groups, suggesting that the factors common to both surgical procedures, such as the cardiopulmonary bypass with cardioplegia and mechanical manipulation of the heart, were responsible for impairment of the autonomic function. In the present study, patients with CAD in similar clinical conditions to the CABG patients but who refused surgery were studied as a control for the surgical procedure group. As expected, the results from the autonomic tests and HRV variability before surgery were similar in the two groups and decreased after surgery, but did not change in the CAD group during 90 days. In addition, both groups have shown lower values of autonomic cardiac indexes in comparison with paired healthy subjects, emphasizing the effects of CAD on autonomic function. The study by Bauernschmitt and colleagues [9] also compared the results from patients undergoing CABG against those obtained from healthy volunteers, and showed that parasympathetic function was impaired 20 hours after surgery, but longer periods were not evaluated. The value of HRV as a tool for investigating autonomic function was also evaluated by Carpeggiani and colleagues [16], who showed that the LF spectral component has independent prognostic value early after acute myocardial infarction to predict in-hospital complications.
Mechanisms of autonomic alterations in CABG
Various factors related to the CABG surgery could be involved in the impairment of cardiac autonomic modulation. Induction of anesthesia with fentanyl–diazepam–pancuronium have been demonstrated to decrease HRV, particularly the HF component, therefore increasing the LF/HF ratio [17]. These results suggest that this combination of drugs decreased vagal modulation and increased cardiac adrenergic activity (see below). In addition, Hogue and colleagues [13] have shown that induction of anesthesia before cardiopulmonary bypass surgery decreased HRV in comparison with the preoperative day. However, patients undergoing CABG presented a further decrease in HRV indices, revealing that the surgery had an effect besides that of anesthesia on autonomic function.
Several mechanisms have been suggested to impair autonomic function after CABG, such as mechanical destructions of autonomic fibers caused by aortic clamping [18,19], although an experiment in a canine model did not show a significant effect [20]. Extracorporal circulation could be theoretically involved as mechanism, but a recent study by Demirel and colleagues has not found a correlation between the duration of extracorporal circulation and the magnitude of autonomic function [12]. In addition, other factors involved with the surgical procedure, including cardioplegia and hypothermia, have been speculated to cause autonomic dysfunction [20] but have not been evaluated in a systematic manner. More recently, autonomic modulation of heart rate has been shown to be reduced by systemic inflammation [21], a condition known to occur after CABG [22].
Clinical implications
The reduction of bedside tests and the time-domain indexes of HRV observed after CABG indicate an overall impairment in the autonomic modulation of heart rate. However, these indexes cannot distinguish between sympathetic adrenergic and parasympathetic function. Conversely, analysis of the power spectrum in the frequency domain has been used to provide insights into the relative contributions of the two branches of the autonomic nervous system to the global behavior of HRV [6]. The HF component is determined by vagal modulation, whereas the LF component carries the influence of both sympathetic adrenergic and parasympathetic modulation, and the ratio between LF and HF components should indicate sympathetic modulation, or the reciprocal relation of LF and HF components as a marker of the state of the sympathovagal balance [6]. In the present study, RSA, Valsalva maneuver, and time-domain indices of HRV showed patterns similar to those observed in the HF power; that is, nadir values at PODs 3–6 and recovery at about POD 30. It therefore seems that the HR response to the bedside tests and the time-domain indices of HRV have parasympathetic modulation as the major mechanism. In contrast, the LF component and the LF/HF ratio presented the exact opposite behavior, showing an increase after CABG with peak values at POD 6, returning to pre-surgery values at about POD 30. This should be interpreted as a reversible increase in the sympathetic drive after CABG and also suggests that, at least in these patients, the LF/HF ratio is linked mainly to adrenergic modulation.
The present results may have practical implications. Although a previous report by McHugh and colleagues did not find a relationship between HRV and cardiovascular instability in the intensive care unit after cardiac surgery [23], altered heart rate dynamics has been shown to be related to myocardial ischemic episodes in patents after CABG [23], suggesting that the autonomic nervous system has an important role in the pathogenesis of myocardial ischemia in the post-operative phase of CABG. Sympathetic activation increases cardiac oxygen demand, causing myocardial ischemia in patients with coronary obstruction. In addition, norepinephrine (noradrenaline) enhances myocardial electrical excitability and can trigger ectopic foci and ventricular arrhythmias [3]. In contrast, vagal stimulation decreases myocardial work and oxygen demand and is known to protect against cardiac arrhythmias acting on electrophysiological properties of the myocardium, such as increasing fibrillatory threshold [24,25]. Patients recovering from CABG might therefore be at augmented risk for myocardial ischemia and arrhythmias at about POD 3, when adrenergic influence is higher and vagal modulation is at its lowest. It is worth noticing that RSA and Valsalva maneuver, simple autonomic bedside tests, were able to show the impairment and recovery of autonomic function after CABG in a similar manner to that of HRV, which demands relatively more complex analytical methods. This might facilitate the application of autonomic evaluation after CABG in the clinical setting of a larger number of centers.
The results of the HF component are particularly interesting in the clinical perspective. The values for CABG patients were lower than those for healthy controls and quite similar to patients with CAD. After surgery there was a progressive increase in the HF values such that after 90 days these values were higher than those observed in CAD patients and identical to those of the healthy group. This suggests that CABG was able to induce an increase in cardiac vagal modulation in comparison with pre-surgery values, with potential protective action, because there is specific interest in measures capable of correcting parasympathetic dysfunction [26]. Similar results were obtained by Bellwon and colleagues [10]. Future studies should be conducted to establish whether autonomic evaluation is useful in predicting clinical outcome in patients after CABG and whether correction of vagal modulation can be considered as a mechanism for protection in patients undergoing CABG.
Limitations of study
Because the present study involved only patients with normal ventricular function (ejection fraction more than 50%), we cannot extrapolate the results to other patients with poor ventricular function. In addition, the relatively small sample size, although sufficient to detect group differences, prevented us from exploring a potential difference between patients with coronary lesions affecting diverse ventricular walls. This would be an interesting aspect for exploration in future studies, because a more pronounced autonomic impairment has previously been shown in patients with myocardial ischemia on the anterior wall in comparison with those showing inferior wall ischemia [27]. Given that several pharmacological agents can interfere with autonomic function, the apparent more prevalent use of β-blockers in the CABG group could represent a confounding effect. Nevertheless, the proportion of drugs used was not statistically different, and the core analysis of the study was the longitudinal evaluation of autonomic function of each group, in which each subject served as his or her own control. Another limitation of the present study is that it cannot be determined which factors related to CABG were responsible for the autonomic changes (see above). Despite the underlying mechanism, it was quite clear that CABG and/or related factors caused reversible autonomic dysfunction with potential clinical impact.
Conclusion
The results of this study show a decrease in the cardiovascular autonomic function indexes after CABG. The lowest values occurred 3–6 days after surgery with a gradual recovery, returning to preoperative values by POD 60. Future investigations should verify whether these results have practical implications for predicting clinical outcome; that is, whether the evaluation of autonomic function can detect patients with a higher risk after CABG.
Key messages
• Cardiovascular autonomic function was impaired after CABG, as identified by noninvasive bedside tests.
• Autonomic function reached lowest values 3–6 days after CABG and returned to pre-surgery values after 60 days.
Abbreviations
ANOVA = analysis of variance; CABG = coronary artery bypass grafting; CAD = coronary artery disease; HF = high-frequency; HRV = heart rate variability; LF = low-frequency; POD = postoperative day; RMSSD = square root of the mean of squares of differences between adjacent R–R intervals; RSA = respiratory sinus arrhythmia.
Competing interests
The author(s) declare that they have no competing interests.
Authors' contributions
ACLN conceived and designed the study, assisted with analysis and revised the manuscript. PPSS conducted the principal analysis and revised the manuscript. AMM participated in the study design, conducted data collection, assisted with analysis and drafted the manuscript. SLDC contributed to study design and manuscript revision. All authors read and approved the final manuscript.
Acknowledgements
We are grateful for the valuable assistance of the medical staff from Hospital Procordis, Niterói, RJ, Brazil, especially Dr Claudio Catharina. This study was performed at Universidade Federal Fluminense and was supported by the Conselho Nacional de Desenvolvimento Científico e Tecnológico, Brazil (CNPq; 520660/95-1) and Fundação de Amparo de Pesquisa do Estado do Rio de Janeiro, Brazil (FAPERJ; E-26/170037/96-APQ1)
Figures and Tables
Figure 1 Longitudinal evaluation of autonomic cardiovascular tests in patients undergoing coronary artery bypass grafting surgery. Longitudinal autonomic cardiovascular evaluation (Valsalva maneuver and respiratory sinus arrhythmia) in the three groups: CABG, patients who underwent coronary artery bypass grafting; CAD, patients with coronary artery disease who did not undergo surgery; healthy, control subjects without CAD. *P < 0.05 versus CABG preoperative value; †P < 0.05 versus CABG day 3; ‡P < 0.05 versus CABG and CAD.
Figure 2 Longitudinal evaluation of heart rate variability in patients undergoing coronary artery bypass grafting surgery. Longitudinal results of spectral analysis of heart rate variability in the three groups: CABG, patients who underwent coronary artery bypass grafting; CAD, patients with coronary artery disease who did not undergo surgery; healthy, control subjects without CAD. Total power (a) and low-frequency power (b) indicate both adrenergic and parasympathetic modulation of heart rate; high-frequency power (c) represents the parasympathetic component; and the low-frequency/high-frequency ratio (d) represents autonomic balance that modulates heart rate. *P < 0.05 versus CABG preoperative value; †P < 0.05 versus CABG day 3 or 6; ‡P < 0.05 versus CABG and CAD, except high frequency, for which ‡P < 0.05 versus CAD.
Table 1 Demographic characteristics of patients and control subjects
Characteristic CABG (n = 13) CAD (n = 9) Healthy (n = 9)
Sex (males/females) 8/5 5/4 4/5
Age (years) 64 ± 2 64 ± 2 63 ± 2
Body mass index (kg/m2) 27.6 ± 0.6 27.4 ± 0.7 26.5 ± 0.9
Drug used
ACE inhibitors/AngioII-blockers 5 6 -
Digital 3 0 -
Anti-arrhythmic 1 0 -
β-blockers 6 1 -
Calcium-channel antagonists 4 1 -
AAS 7 3 -
Values are means ± SEM. AAS, acetylsalicylic acid; ACE, angiotensin-converting enzyme; CABG, patients who underwent coronary artery bypass grafting; CAD, patients with coronary artery disease who did not undergo surgery; healthy, control subjects without CAD. The three groups were similar with regard to all variables (P > 0.05).
Table 2 Time-domain indexes of heart rate variability before and at various intervals after coronary artery bypass grafting
Variable Group Preoperative After surgery
3 days 6 days 15 days 30 days 60 days 90 days
Mean R–R (ms) CABG 903 ± 40 761 ± 45* 781 ± 42* 790 ± 36* 802 ± 26* 801 ± 36 802 ± 35
CAD 856 ± 49 - - - 842 ± 38 816 ± 33 835 ± 47
Healthy 860 ± 36 - - - - - -
SDNN (ms) CABG 35 ± 3 16 ± 2* 16 ± 2* 19 ± 2* 25 ± 2* 32 ± 3 27 ± 3
CAD 31 ± 4 - - - 29 ± 3 30 ± 3 31 ± 2
Healthy 42 ± 5† - - - - - -
SDANN (ms) CABG 16 ± 3 6 ± 1* 6 ± 1* 9 ± 1* 15 ± 3 18 ± 3 14 ± 2
CAD 21 ± 4 - - - 19 ± 3 20 ± 3 20 ± 3
Healthy 39 ± 7† - - - - - -
RMSSD (ms) CABG 24 ± 4 10 ± 2* 10 ± 1* 12 ± 1* 19 ± 4 21 ± 3 19 ± 2
CAD 22 ± 3 - - - 18 ± 3 19 ± 2 20 ± 4
Healthy 27 ± 3 - - - - - -
pNN50 (%) CABG 8 ± 0 3 ± 0* 3 ± 0* 3 ± 0* 4 ± 0* 7 ± 0 6 ± 0
CAD 4 ± 0 - - - 3 ± 0 4 ± 0 3 ± 0
Healthy 11 ± 0b - - - - - -
Values are means ± SEM. *P < 0.05 versus CABG preoperative value; †P < 0.05 versus CABG and CAD. CABG, patients who underwent coronary artery bypass grafting; CAD, patients with coronary artery disease who did not undergo surgery; healthy, control subjects without CAD; pNN50, percentage number of pairs of adjacent R–R intervals differing by more than 50 ms; RMSSD, square root of the mean of squares of differences between adjacent R–R intervals; SDANN, standard deviation of the averages of the R–R intervals calculated in 5-minute segments; SDNN, standard deviation of all R–R intervals.
==== Refs
Kleiger RE Miller JP Bigger JT JrMoss AJ Decreased heart rate variability and its association with increased mortality after acute myocardial infarction Am J Cardiol 1987 59 256 262 3812275 10.1016/0002-9149(87)90795-8
La Rovere MT Bigger JT JrMarcus FI Mortara A Schwartz PJ Baroreflex sensitivity and heart-rate variability in prediction of total cardiac mortality after myocardial infarction. ATRAMI (Autonomic Tone and Reflexes After Myocardial Infarction) investigators Lancet 1998 351 478 484 9482439 10.1016/S0140-6736(97)11144-8
Schwartz PJ La Rovere MT Vanoli E Autonomic nervous system and sudden cardiac death. Experimental basis and clinical observations for post-myocardial infarction risk stratification Circulation 1992 85 1 Suppl I77 I91 1728509
Hrushesky WJ Fader D Schmitt O Gilbertsen V The respiratory sinus arrhythmia: a measure of cardiac age Science 1984 224 1001 1004 6372092
Korner PI Tonkin AM Uther JB Reflex and mechanical circulatory effects of graded Valsalva maneuvers in normal man J Appl Physiol 1976 40 434 440 931859
Heart rate variability: standards of measurement, physiological interpretation and clinical use. Task Force of the European Society of Cardiology and the North American Society of Pacing and Electrophysiology Circulation 1996 93 1043 1065 8598068
Eagle KA Guyton RA Davidoff R Ewy GA Fonger J Gardner TJ Gott JP Herrmann HC Marlow RA Nugent W ACC/AHA guidelines for coronary artery bypass graft surgery: executive summary and recommendations. A report of the American College of Cardiology/American Heart Association Task Force on Practice Guidelines (Committee to revise the 1991 guidelines for coronary artery bypass graft surgery) Circulation 1999 100 1464 1480 10500052
European Coronary Surgery Study Group Long-term results of prospective randomised study of coronary artery bypass surgery in stable angina pectoris Lancet 1982 2 1173 1180 6128492
Bauernschmitt R Malberg H Wessel N Kopp B Schirmbeck EU Lange R Impairment of cardiovascular autonomic control in patients early after cardiac surgery Eur J Cardiothorac Surg 2004 25 320 326 15019655 10.1016/j.ejcts.2003.12.019
Bellwon J Siebert J Rogowski J Szulc J Ciecwierz D Deptulski T Narkiewicz M Rynkiewicz A Heart rate power spectral analysis in patients before and 6 weeks after coronary artery bypass grafting Clin Sci (Lond) 1996 91 Suppl 19 21 8813817
Bronner F Douchet MP Quiring E Charpentier A Vi-Fane R Eisenmann B Chauvin M Brechenmacher C Variability of heart rate after heart surgery under extracorporeal circulation: aortocoronary bypass or aortic valve replacement. [In French] Ann Cardiol Angeiol (Paris) 1998 47 549 554 9809138
Demirel S Akkaya V Oflaz H Tukek T Erk O Heart rate variability after coronary artery bypass graft surgery: a prospective 3-year follow-up study Ann Noninvasive Electrocardiol 2002 7 247 250 12167187
Hogue CW JrStein PK Apostolidou I Lappas DG Kleiger RE Alterations in temporal patterns of heart rate variability after coronary artery bypass graft surgery Anesthesiology 1994 81 1356 1364 7992903
Ruha A Sallinen S Nissila S A real-time microprocessor QRS detector system with a 1-ms timing accuracy for the measurement of ambulatory HRV IEEE Trans Biomed Eng 1997 44 159 167 9216129 10.1109/10.554762
Loimaala A Sievanen H Laukkanen R Parkka J Vuori I Huikuri H Accuracy of a novel real-time microprocessor QRS detector for heart rate variability assessment Clin Physiol 1999 19 84 88 10068870 10.1046/j.1365-2281.1999.00152.x
Carpeggiani C L'Abbate A Landi P Michelassi C Raciti M Macerata A Emdin M Early assessment of heart rate variability is predictive of in-hospital death and major complications after acute myocardial infarction Int J Cardiol 2004 96 361 368 15301888 10.1016/j.ijcard.2003.07.023
Komatsu T Kimura T Sanchala V Shibutani K Shimada Y Effects of fentanyl-diazepam-pancuronium anesthesia on heart rate variability: a spectral analysis J Cardiothorac Vasc Anesth 1992 6 444 448 1498300 10.1016/1053-0770(92)90011-U
Inoue H Skale BT Zipes DP Effects of ischemia on cardiac afferent sympathetic and vagal reflexes in dog Am J Physiol 1988 255 H26 H35 3394822
Kammerling JJ Green FJ Watanabe AM Inoue H Barber MJ Henry DP Zipes DP Denervation supersensitivity of refractoriness in noninfarcted areas apical to transmural myocardial infarction Circulation 1987 76 383 393 3038369
Murphy DA Armour JA Influences of cardiopulmonary bypass, temperature, cardioplegia, and topical hypothermia on cardiac innervation J Thorac Cardiovasc Surg 1992 103 1192 1199 1597985
Sajadieh A Nielsen OW Rasmussen V Hein HO Abedini S Hansen JF Increased heart rate and reduced heart-rate variability are associated with subclinical inflammation in middle-aged and elderly subjects with no apparent heart disease Eur Heart J 2004 25 363 370 15033247 10.1016/j.ehj.2003.12.003
Furtado de Mendonca-Filho HT Gomes RV Campos LA Tura B Nunes EM Gomes R Bozza F Bozza PT Castro-Faria-Neto HC Circulating levels of macrophage migration inhibitory factor are associated with mild pulmonary dysfunction after cardiopulmonary bypass Shock 2004 22 533 537 15545824 10.1097/01.shk.0000142817.84070.df
McHugh GJ Sleigh JW Bo H Henderson JD Heart rate variability following cardiac surgery fails to predict short-term cardiovascular instability Anaesth Intensive Care 1997 25 621 626 9452842
Kolman BS Verrier RL Lown B The effect of vagus nerve stimulation upon vulnerability of the canine ventricle: role of sympathetic-parasympathetic interactions Circulation 1975 52 578 585 239801
Vanoli E De Ferrari GM Stramba-Badiale M Hull SS JrForeman RD Schwartz PJ Vagal stimulation and prevention of sudden death in conscious dogs with a healed myocardial infarction Circ Res 1991 68 1471 1481 2019002
Nobrega AC Teixeira De Castro RR Parasympathetic dysfunction as a risk factor in myocardial infarction: what is the treatment? Am Heart J 2000 140 E23 11011320
Pipilis A Flather M Ormerod O Sleight P Heart rate variability in acute myocardial infarction and its association with infarct site and clinical course Am J Cardiol 1991 67 1137 1139 2024606 10.1016/0002-9149(91)90880-T
| 15774044 | PMC1175925 | CC BY | 2021-01-04 16:04:52 | no | Crit Care. 2005 Jan 26; 9(2):R124-R131 | utf-8 | Crit Care | 2,005 | 10.1186/cc3042 | oa_comm |
==== Front
Crit CareCritical Care1364-85351466-609XBioMed Central London cc30441577404610.1186/cc3044ResearchSerum cystatin C concentration as a marker of acute renal dysfunction in critically ill patients Villa Patricia [email protected]énez Manuel [email protected] Maria-Cruz [email protected] Jesus [email protected] Pilar 21 Intensive Care Unit, Hospital Universitario La Paz, Madrid, Spain2 Biochemistry Unit, Hospital Universitario La Paz, Madrid, Spain2005 7 2 2005 9 2 R139 R143 4 6 2004 26 7 2004 25 10 2004 17 12 2004 Copyright © 2005 Villa 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
In critically ill patients sudden changes in glomerular filtration rate (GFR) are not instantly followed by parallel changes in serum creatinine. The aim of the present study was to analyze the utility of serum cystatin C as a marker of renal function in these patients.
Methods
Serum creatinine, serum cystatin C and 24-hour creatinine clearance (CCr) were determined in 50 critically ill patients (age 21–86 years; mean Acute Physiology and Chronic Health Evaluation II score 20 ± 9). They did not have chronic renal failure but were at risk for developing renal dysfunction. Serum cystatin C was measured using particle enhanced immunonephelometry. Twenty-four-hour body surface adjusted CCr was used as a control because it is the 'gold standard' for determining GFR.
Results
Serum creatinine, serum cystatin C and CCr (mean ± standard deviation [range]) were 1.00 ± 0.85 mg/dl (0.40–5.61 mg/dl), 1.19 ± 0.79 mg/l (0.49–4.70 mg/l), and 92.74 ± 52.74 ml/min per 1.73 m2 (8.17–233.21 ml/min per 1.73 m2), respectively. Our data showed that serum cystatin C correlated better with GFR than did creatinine (1/cystatin C versus CCr: r = 0.832, P < 0.001; 1/creatinine versus CCr: r = 0.426, P = 0.002). Cystatin C was diagnostically superior to creatinine (area under the curve [AUC] for cystatin C 0.927, 95% confidence interval 86.1–99.4; AUC for creatinine 0.694, 95% confidence interval 54.1–84.6). Half of the patients had acute renal dysfunction. Only five (20%) of these 25 patients had elevated serum creatinine, whereas 76% had elevated serum cystatin C levels (P = 0.032).
Conclusion
Cystatin C is an accurate marker of subtle changes in GFR, and it may be superior to creatinine when assessing this parameter in clinical practice in critically ill patients.
==== Body
Introduction
Glomerular filtration rate (GFR) is considered the best marker of renal function, and serum creatinine is the most commonly used biochemical parameter to estimate GFR in routine practice. However, there are some shortcomings to the use of this parameter. Factors such as muscle mass and protein intake can influence serum creatinine, leading to an inaccurate estimation of GFR. Normal serum creatinine may be observed in individuals with significantly impaired GFR [1,2]. Moreover, in unstable, critically ill patients, acute changes in renal function can make real-time evaluation of GFR using serum creatinine difficult.
Cystatin C is a nonglycosylated protein that belongs to the cysteine protease inhibitors, cystatin superfamily [3]. These proteins play an important role in the regulation of proteolytic damage to the cysteine proteases. Cystatin C is produced at a constant rate by nucleated cells [4]. It is found in relatively high concentrations in many body fluids, especially in the seminal fluid, cerebrospinal fluid and synovial fluid [5]. Its low molecular weight (13.3 kDa) and positive charge at physiological pH levels facilitate its glomerular filtration. Subsequently, it is reabsorbed and almost completely catabolized in the proximal renal tubule [6,7]. Therefore, because of its constant rate of production, its serum concentration is determined by glomerular filtration [8-11]. Moreover, its concentration is not influenced by infections, liver diseases, or inflammatory diseases. Use of serum cystatin C as a marker of GFR is well documented, and some authors have suggested that it may be more accurate than serum creatinine for this purpose [12-19].
The difficulties associated with monitoring and evaluating GFR in critically ill individuals are well known. Thus far no studies evaluating serum cystatin C as a marker of GFR in these patients have been reported. The aim of the present study were to determine the accuracy of serum cystatin C concentration as a marker of GFR in critically ill individuals.
Methods
Fifty patients, aged 21–86 years (mean 54 years), who were admitted to the intensive care unit at the Hospital Universitario La Paz in Madrid, Spain between January and September 2001, were included in the study. All patients were at risk for developing renal failure (haemodynamically unstable patients, septic patients, individuals receiving nephrotoxic drugs and others). Patients receiving corticoid therapy or with thyroid diseases were excluded. The patients' demographic characteristics and clinical conditions are summarized in Table 1.
A serum sample was drawn from each patient in the morning (between 07:00 and 10:00) to determine serum creatinine and serum cystatin C. A 24-hour urine sample was obtained just before the serum sample to calculate the creatinine clearance (CCr) using the following formula: CCr (ml/min) = (urine volume × urine creatinine)/(serum creatinine × 1440).
Serum creatinine values were obtained according to standard laboratory methods. CCr was adjusted to body surface (ml/min per 1.73 m2). Cystatin C values were obtained using particle enhanced immunonephelometry [10]. Normal serum creatinine values range from 0.6 to 1.3 mg/dl, and normal serum cystatin C values range from 0.6 to 1 mg/l. Renal dysfunction was defined as CCr below 80 ml/min per 1.73 m2.
Statistical analysis
The data are expressed in mean ± standard deviation (range). Correlations between quantitative data were determined using Pearson's test. P < 0.05 was considered statistically significant. The diagnostic value of serum cystatin C and serum creatinine for identifying renal dysfunction was evaluated using receiver operating characteristic curve analysis, and the data are expressed as area under the curve (AUC; 95% confidence interval). For statistical analysis, the SPSS R 9.0 (SPSS Inc., Chicago, IL, USA) program was used.
Results
The mean serum creatinine concentration was 1.00 ± 0.85 mg/dl (0.40–5.61 mg/dl) and the mean serum cystatin C concentration was 1.19 ± 0.79 mg/l (0.49–4.70 mg/l). The mean CCr adjusted for the body surface was 92.74 ± 52.74 ml/min per 1.73 m2 (8.17–233.21 ml/min per 1.73 m2).
A decline in CCr was followed by an increase in levels of serum creatinine and serum cystatin C (Fig. 1). The inverse of the serum cystatin C and serum creatinine levels were plotted against CCr to determine the relationships of those parameters to this marker of renal function (Fig. 2a,b). There were significant correlations between CCr and 1/serum creatinine (r = 0.426, P = 0.002) and between CCr and 1/serum cystatin C (r = 0.832, P < 0.001).
Twenty-five out of the 50 patients enrolled in the study had renal dysfunction (CCr <80 ml/min per 1.73 m2). Five (20%) of these 25 patients with renal dysfunction had elevated serum creatinine concentrations, whereas 19 (76%) of them had elevated serum cystatin C levels at the time of renal dysfunction (P = 0.032). On the other hand, serum creatinine levels were within normal ranges in all patients with normal CCr (>80 ml/min per 1.73 m2) whereas 23 (92%) of them had normal concentrations of serum cystatin C.
Nonparametric receiver operating characteristic plots of sensitivity and specificity of serum creatinine and cystatin C for detecting renal dysfunction are shown in Fig. 3. The AUC for serum creatinine was 0.694 (95% confidence interval 54.1–84.6) and the AUC for serum cystatin C was 0.927 (95% confidence interval 86.1–99.4).
Discussion
Monitoring renal function is extremely important in the management of critically ill patients. GFR, which can be measured by determining the clearance of various substances, is the 'gold standard' parameter for monitoring renal function. The ideal endogenous marker would be characterized by stable production rate, stable circulating levels (unaffected by pathological changes), lack of protein binding, free glomerular filtration, and lack of reabsorption or secretion; to date, no such marker has yet been identified. Some substances such as creatinine, urea, β2-microglobulin and retinol-binding protein have been used as endogenous markers of GFR, by measuring either their plasma levels or their renal clearance. Among them, the most useful markers for assessing GFR are serum creatinine and renal CCr. This is secondary to their correlations with the renal clearance of some exogenous substances (inulin, creatinine-EDTA, iothalamate) that are considered 'gold standards' for determining GFR.
Creatinine production changes significantly according to the muscle mass of the body and dietetic factors. It is filtered by the glomeruli, but it is also secreted by the renal tubules. This tubular secretion contributes approximately 20% of the total creatinine excretion by the kidney, and it can increase as GFR decreases. All of these factors explain why serum creatinine concentration may not be a good parameter for accurate determination of GFR, especially at lower rates [1].
Cystatin C production in the body is a stable process that is not influenced by renal conditions, increased protein catabolism, or dietetic factors. Moreover, it does not change with age or muscle mass like creatinine does. Its biochemical characteristics allow free filtration in the renal glomerulus, and subsequent metabolism and reabsorption by the proximal tubule. For these reasons, serum cystatin C has been suggested to be an ideal endogenous marker of GFR [12-19].
Most studies conducted to evaluate whether there is a role for serum cystatin C in determining GFR involved measurement of the clearance of exogenous substances such as creatinine-EDTA [14,20-22], inulin [15,23], Tc-DTPA [24,25] and I-iothalamate [16,26,27]. Nevertheless, CCr is still the most reliable marker for determining GFR on a routine basis, and multiple studies have used CCr as a control for evaluating the role for serum cystatin C as a measure of GFR [28-30]. It is also a simple and cheap test, and, as mentioned above, its accuracy is sufficient for determining GFR. However, measurement of CCr can yield erroneous findings in many situations, particularly when poor urine collection technique is employed. That the present study was conducted in critically ill individuals, all of whom had a bladder catheter in place, makes such errors less likely.
Previous studies [14-16,20-22,25,27] have found a wide range of correlations between 1/serum creatinine and clearance of exogenous substances (r = 0.50–0.89). In this study we found a correlation coefficient of 0.426 (P = 0.002) between CCr and 1/serum creatinine. This difference in correlation rates among studies may be explained better by the characteristics of the patients than by the methods used. Most of the studies in the literature were performed in individuals who were in a stable clinical condition (healthy individuals, patients with various renal diseases, and oncological patients undergoing chemotherapy). GFR in critically ill individuals can change rapidly because of, for example, renal hypoperfusion secondary to shock or the use of nephrotoxic agents. Despite this, it is not uncommon to see changes in the serum creatinine for up to several days until the stabilization phase is reached. This may also explain the poor diagnostic usefulness of serum creatinine as seen in our study (AUC 0.694) compared with that in other studies [25]. Only five out of 25 (20%) of the individuals enrolled in our study who developed renal dysfunction exhibited high serum creatinine levels at the time when CCr was tested. The delay that usually exists between the decline in GFR and that in serum creatinine makes the latter test poorly reliable for making therapeutic decisions in critically ill patients, such as a decision to change nephrotoxic agents or to increase renal perfusion.
We found a strong correlation between serum cystatin C concentrations and CCr in this study (r = 0.832, P < 0.001). This is similar to findings reported by other investigators (r = 0.73–0.91) [14-16,20-22,25,27]. The diagnostic utility of cystatin C seen in our study (AUC = 0.927) is similar to that previously reported by other investigators [25]. The fact that most of our patients (76%) with acute renal dysfunction had high serum cystatin C levels at the time of CCr evaluation demonstrates that cystatin C is a good marker for application in real time, and suggests that serum cystatin C is a better marker of GFR than is serum creatinine in unstable, critically ill patients (20% of patients with acute renal dysfunction had high serum cystatin C level).
Conclusion
In the present study we evaluated and compared serum creatinine and serum cystatin C as markers of GFR in unstable, critically ill patients. Our data indicate that serum cystatin C is a good real-time marker of GFR in such patients. If this finding is subsequently confirmed, then the simplicity of serum cystatin C detection and its reasonable cost suggest that this test may soon replace CCr as the biochemical marker of choice for monitoring GFR in a routine practice.
Key messages
• In this study, serum cystatin C was found to be a good marker of GFR.
• Serum cystatin C was better at detecting changes in GFR than was serum creatinine in critically ill patients.
• Determination of serum cystatin C levels is useful in the management of critically ill patients.
Abbreviations
AUC = area under the curve; CCr = creatinine clearance; GFR = glomerular filtration rate.
Competing interests
The author(s) declare that they have no competing interests.
Authors' contributions
PV managed patients, recruited them into the study and participated in the drafting of the manuscript. MJ conceived the study and participated in its design and coordination. MCS and JM were subinvestigators of the study; their principal role was to recruit patients. PC analyzed the samples.
Acknowledgements
We thank Concepción Madero, MD, from the Statistical Service of Hospital Universitario La Paz for her assistance with the analysis of the data. We thank Cynthia McCoig MD for reviewing the manuscript.
Figures and Tables
Figure 1 Relationships of (a) serum creatinine and (b) serum cystatin C to creatinine clearance (CCr).
Figure 2 The (a) inverse of serum creatinine (1/creatinine) and the (b) inverse of cystatin C (1/cystatin c) were plotted against creatinine clearance (CCr) for each of patient (P = 0.002 versus P < 0.001).
Figure 3 Nonparametric receiver operating characteristic plots of sensitivity and specificity of serum creatinine and cystatin C. Area under the curve (95% confidence interval): creatinine 0.694 (54.1–84.6) and cystatin C 0.927 (86.1–99.4).
Table 1 Demographic characteristics and clinical conditions of 50 critically ill patients at risk for developing acute renal dysfunction
Parameter Details
Age (years; mean [range]) 54 (21–86)
Sex (male/female) 34/16
APACHE II score (mean ± SD) 20 ± 9
Disease
Multiple trauma 10/50 (20%)
Pulmonary disease 10/50 (20%)
Neurological disease 9/50 (18%)
Sepsis/septic shock 9/50 (18%)
Cardiological disease 7/50 (14%)
Postsurgical high risk 5/50 (10%)
APACHE, Acute Physiology and Chronic Health Evaluation; SD, standard deviation.
==== Refs
Levey AS Perrone RD Madias NE Serum creatinine and renal function Annu Rev Med 1988 39 465 490 3285786 10.1146/annurev.me.39.020188.002341
Shemesh O Golbetz H Kriss JP Myers BD Limitations of creatinine as filtration marker in glomerulopathic patients Kidney Int 1985 28 830 836 2418254
Perrone RD Madias NE Levey AS Serum creatinine as an index of function renal: new insights into old concepts Clin Chem 1992 38 1933 1953 1394976
Abrahamson M Olafsson I Palsdottir A Ulvsback M Lundwall A Jensson O Grubb A Structure and expression of the human cystatin C gene Biochem J 1990 268 287 294 2363674
Abrahasom M Barret AJ Salveson G Grubb A Isolation of six cysteine protease inhibitors from human urine J Biol Chem 1986 261 11282 11289 3488317
Grubb A Diagnostic value of analysis of cystatin C and protein HC in biological fluids Clin Nephrol 1992 S20 S27 1284235
Tenstad O Roald AB Grubb A Aukland K Renal handling of radiolabelled human cystatin C in the rat Scand J Clin Lab Invest 1996 56 409 414 8869663
Nilsson-Ehle P Grubb A New markers for the determination of GFR: iohexol clearance and cystatin C serum concentration Kidney Int Suppl 1994 47 S17 S19 7869664
Kyhse-Andersen J Schmidt C Nordin G Andersson B Nilsson-Ehle P Lindstrom V Grubb A Serum cystatin C, determined by a rapid, automated particle-enhanced turbidimetric method, is a better marker than serum creatinine for glomerular filtration rate Clin Chem 1994 40 1921 1926 7923773
Finney H Newman DJ Gruber W Merle P Price CP Initial evaluation of cystatin C measurement by partcle-enhanced inmunonephelometry on Behring nephelometer system (BNA, BN II) Clin Chem 1997 43 1016 1022 9191555
Pergande M Jung K Sandwich enzyme inmunoassay of cystatin C in serum with commercially avaible antibodies Clin Chem 1993 39 1885 1890 8375065
Herget-Rosenthal S Trabold S Pietruck F Holtmann M Philipp T Kribben A Cystatin C Efficacy as screening test for reduced glomerular filtration rate Am J Nephrol 2000 20 97 102 10773608 10.1159/000013564
Jung K Jung M Cystatin C: a promising marker of glomerular filtration rate to replace creatinine Nephron 1995 70 370 371 7477630
Newman DJ Thakkar H Hedward RG Wilkie M White T Grubb A Price CP Serum cystatin C measured by automated inmunoassay: a more sensititive marker of changes in GFR than serum creatinine Kidney Int 1995 47 312 318 7731163
Stickle D Cole B Hock K Hruska KA Scott MG Correlation of plasma concentrations of cystatin C and creatinine to inulin clearance in pediatric population Clin Chem 1998 44 1334 1338 9625061
Risch L Blumberg A Huber AR Assessment of function renal in renal transplant patiens using cystatin C. A comparison to other renal function markers and estimates Ren Fail 2001 23 439 448 11499559 10.1081/JDI-100104727
Le Bricon T Thervet E Benlakehal M Bousquet B Legendre C Erlich D Changes in plasma cystatin C after renal transplantation and acute rejection in adults Clin Chem 1999 45 2243 2249 10585359
Randers E Erlandsen EJ Serum cyistatin C as as endogenous marker of the renal function Clin Chem Lab Med 1999 37 389 395 10369108 10.1515/CCLM.1999.064
Finney H Newman DJ Price CP Adult reference for serum cystatin C, creatinine and predicted creatinine clearance Ann Clin Biochem 2000 37 49 59 10672373 10.1258/0004563001901524
Grubb A Simonsen O Sturfelt G Trudsson L Thysell H Serum concentration of cystatin C, factor D and β2-microglobulin as a measure of glomerular filtrate rate Acta Med Scand 1985 218 499 503 3911736
Simonsen O Grubb A Thysell H The blood serum concentration of cystatin D (gamma-trace) as a measure of the glomerular filtration rate Scand J Clin Lab Invest 1985 45 97 101 3923607
Bökenkamp A Domanetzki M Zinck R Schumann G Byrd D Brodehl J Cystatin C: a new marker of glomerular filtration rate in children independent of age and height Pediatrics 1998 101 875 881 9565418 10.1542/peds.101.5.875
Fliser D Ritz E Serum cystatin C concentration as a marker of renal dysfunction in the elderly Am J Kidney Dis 2001 37 79 83 11136171
Randers E Erlandsen EJ Pedersen OL Hasling C Danielsen H Serum cystatin C as an endogenous parameter of the renal function in patients with normal to moderately impaired kidney function Clin Nephrol 2000 54 203 209 11020018
Randers E Kristensen JH Erlandsen EJ Danielsen H Serum cystatin C as a marker of the renal function Scand J Clin Lab Invest 1998 58 585 592 9890342 10.1080/00365519850186210
Coll E Botey A Alvarez L Poch E Quintó Ll Saurina A Vera M Piera C Darnell A Serum cystanin C as a new marker for noninvasive estimation of glomerular filtration rate and as a marker for early renal impairment Am J Kidney Dis 2000 36 29 34 10873868
Risch L Blumberg A Huber A Rapid and accurate assessment of glomerular filtration rate in patients with renal transplants using serum cystatin C Nephrol Dial Transplant 1999 14 1991 1996 10462282 10.1093/ndt/14.8.1991
Herget-Rosenthal S Trabold S Huesing J Heemann U Philipp T Kribben A Cystatin C: an accurate marker of glomerular filtration rate after renal transplantion? Transpl Int 2000 13 285 289 10959481 10.1007/s001470050703
Paskalev E Lambreva L Simeonov P Koicheva N Beleva B Genova M Marcovska R Nashkov A Serum cystatin C in renal transplant patients Clin Chim Acta 2001 310 53 56 11485755 10.1016/S0009-8981(01)00522-8
Tian S Kusano E Ohara T Tabei K Itoh Y Kawai T Asano Y Cystatin C measurement and its practical use in patients with various renal diseases Clin Nephrol 1997 48 104 108 9285147
| 15774046 | PMC1175926 | CC BY | 2021-01-04 16:04:51 | no | Crit Care. 2005 Feb 7; 9(2):R139-R143 | utf-8 | Crit Care | 2,005 | 10.1186/cc3044 | oa_comm |
==== Front
Crit CareCritical Care1364-85351466-609XBioMed Central London cc30471577404710.1186/cc3047ResearchInitial distribution volume of glucose can be approximated using a conventional glucose analyzer in the intensive care unit Ishihara Hironori [email protected] Hitomi 2Okawa Hirobumi 3Takase Hajime 4Tsubo Toshihito 5Hirota Kazuyoshi 61 Department of Anesthesiology, University of Hirosaki School of Medicine, Hirosaki-Shi, Japan2 Department of Anesthesiology, University of Hirosaki School of Medicine, Hirosaki-Shi, Japan3 Intensive Care Unit, University of Hirosaki Hospital, Hirosaki-Shi, Japan4 Department of Anesthesiology, University of Hirosaki School of Medicine, Hirosaki-Shi, Japan5 Intensive Care Unit, University of Hirosaki Hospital, Hirosaki-Shi, Japan6 Department of Anesthesiology, University of Hirosaki School of Medicine, Hirosaki-Shi, Japan2005 11 2 2005 9 2 R144 R149 13 12 2004 6 1 2005 Copyright © 2005 Ishihara 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
We previously reported that initial distribution volume of glucose (IDVG) reflects central extracellular fluid volume, and that IDVG may represent an indirect measure of cardiac preload that is independent of the plasma glucose values present before glucose injection or infusion of insulin and/or vasoactive drugs. The original IDVG measurement requires an accurate glucose analyzer and repeated arterial blood sampling over a period of 7 min after glucose injection. The purpose of the present study was to compare approximated IDVG, derived from just two blood samples, versus original IDVG, and to test whether approximated IDVG is an acceptable alternative measure of IDVG in the intensive care unit.
Methods
A total of 50 consecutive intensive care unit patients were included, and the first IDVG determination in each patient was analyzed. Glucose (5 g) was injected through the central venous line to calculate IDVG. Original IDVG was calculated using a one-compartment model from serial incremental arterial plasma glucose concentrations above preinjection using a reference glucose analyzer. Approximated IDVG was calculated from glucose concentrations in both plasma and whole blood, using a combined blood gas and glucose analyzer, drawn at two time points: immediately before glucose injection and 3 min after injection. Subsequently, each approximated IDVG was calculated using a formula we proposed previously.
Results
The difference (mean ± standard deviation) between approximated IDVG calculated from plasma samples and original IDVG was -0.05 ± 0.54 l, and the difference between approximated IDVG calculated from whole blood samples and original IDVG was -0.04 ± 0.61 l. There was a linear correlation between approximated and original IDVG (r2 = 0.92 for plasma samples, and r2 = 0.89 for whole blood samples).
Conclusion
Our findings demonstrate that there was good correlation between each approximated IDVG and original IDVG, although the two measures are not interchangeable. This suggests that approximated IDVG is clinically acceptable as an alternative calculation of IDVG, although approximated and original IDVGs are not equivalent; plasma rather than whole blood measurements are preferable.
distribution volumeglucosemeasurement techniquesplasmawhole bloodSee related commentary
==== Body
Introduction
We previously proposed initial distribution volume of glucose (IDVG), determined using injection of a small amount of glucose (5 g), as a measure of central extracellular fluid volume status [1-3]. Neither the plasma glucose values present before glucose injection nor infusion of insulin and/or vasoactive drugs had any apparent effect on IDVG calculation [1-3]. IDVG has been demonstrated to correlate well with cardiac output in various critically ill conditions in the absence of congestive heart failure [1,4]. We [5] and Gabbanelli and coworkers [6] recently showed that IDVG, rather than cardiac filling pressures, is clinically relevant as an indirect measure of cardiac preload, based on the close correlation between IDVG and intrathoracic blood volume, even though glucose administered intravenously distributes rapidly not only through the intravascular compartment but also through the extravascular space. Measurement of IDVG can be repeated at 30 min intervals [7,8]. Our original method for IDVG measurement requires repeated arterial blood samplings over 7 min after glucose injection. However, we have proposed that IDVG may be approximated using just two plasma samples, drawn immediately before injection and 3 min after injection [9]. In this manner, IDVG could be simply and rapidly assessed in the intensive care unit (ICU) if an accurate glucose analyzer were readily available.
Rapid and relatively accurate blood glucose measurement has become possible using combined blood gas and glucose analyzers. Many ICUs have this type of glucose analyzer, which would permit routine use of approximated IDVG as a measure of fluid volume in those units, provided that plasma or whole blood glucose concentrations measured using these devices are suitable for IDVG determination.
In the present study we compared approximated IDVG (calculated from plasma or whole blood samples using a combined blood gas and glucose analyzer) with original IDVG (measured using a laboratory reference method), and examined whether approximated IDVG is a clinically acceptable alternative measure of IDVG.
Methods
The research protocol was approved by the Ethics Committee of the University of Hirosaki. Patients or their relatives gave informed consent. A total of 50 patients admitted to the general ICU of the University of Hirosaki Hospital between July and September 2004 were included in this prospective study (Table 1). Although patients may undergo several fluid volume determinations during their stay in the ICU, the present study considered only the first IDVG measurement in each patient during their stay in the ICU. We included 40 surgical patients who had undergone cardiac surgery, mostly coronary artery bypass grafting and aortic arch replacement (n = 23), major abdominal surgery such as bowel resection and oesophagectomy (n = 5), laryngectomy (n = 4), hip joint surgery (n = 4), thoracic surgery (n = 2), large vessel surgery (n = 1), or spine surgery (n = 1). The remaining 10 patients had nonsurgical pathology such as cardiac failure (n = 2), respiratory failure (n = 2), chest trauma (n = 2), renal failure (n = 1), water intoxication (n = 1), tetanus (n = 1) and heat stroke (n = 1).
To calculate IDVG 10 ml of 50% glucose solution (5 g) was injected through the central venous line, as reported previously [1-3]. Blood samples were obtained through a radial artery catheter immediately before and 3, 4, 5 and 7 min after injection. Each 2 ml blood sample was collected in a heparinized syringe. Both plasma and whole blood glucose concentrations were measured using a combined blood gas and glucose analyzer (EML100 Electrolyte Metabolite Laboratory; Radiometer, Copenhagen, Denmark) from two blood samples: one drawn immediately before glucose injection and one 3 min after injection. Other than automatic regular calibration, the analyzer was not calibrated. Plasma glucose concentrations in all blood samples were also measured using amperometry by glucose oxidase immobilized membrane–H2O2 electrode (glucose analyzer GA-1150; Arkray Co., Ltd, Kyoto, Japan) as the reference. The interassay coefficients of variation were 2.6% for the former and 0.3% for the latter at a glucose concentration of 150 mg/100 ml (n = 6). Original IDVG (the reference) was calculated from the plasma decay curve with a one-compartment model from plasma values increased above preinjection levels between 3 and 7 min postinjection, as described in our previous reports [1-5]. Akaike's information criterion (AIC) [10] for the original IDVG curve was examined, as described previously [1-5], to evaluate the exponential term of the pharmacokinetic model. The lower the AIC value, the better the fit between observed data and the plasma glucose decay curve.
Approximated IDVG was calculated from the increase in either plasma or whole blood glucose concentration above the preinjection level at 3 min after glucose injection using a combined blood gas and glucose analyzer, as described above. In addition, we calculated approximated IDVG from the increase in plasma values above baseline at 3 min after glucose injection determined using the reference glucose analyzer Each approximated IDVG was calculated according to the following formula (proposed by us [9]; Table 2): approximated IDVG (l) = 24.4 × exp(-0.03 × Δgl) + 2.7. (Δgl is the increase in glucose concentration above the preinjection level at 3 min after injection.)
Data are expressed as mean ± standard deviation (SD). Bland–Altman plots were used to compare the bias (the mean of the differences) and precision (SD of bias) between measurements. In addition, regression analysis or a t-test was performed in the comparison of two paired variables. P < 0.05 was considered statistically significant.
Results
Glucose concentrations and other variables for approximated IDVGs are summarized in Table 3. Glucose concentrations in plasma were higher than in whole blood by an average of 2 ± 3 mg/100 ml (n = 100; P < 0.001). The mean haematocrit was 30.3 ± 5.5%, and the total plasma protein concentration was 5.1 ± 0.7 g/100 ml. Neither haematocrit nor total plasma protein concentration were correlated with differences in glucose values between plasma and whole blood samples.
Because the AIC value for original IDVG was -24.8 ± 5.5, convergence was assumed in each glucose decay curve in the present study, as was observed in previous reports [1-5]. The mean original IDVG was 7.44 ± 1.83 l and the rate of disappearance of glucose from plasma was 0.069 ± 0.018 min.
Bland–Altman plots of the differences between each approximated IDVG and original IDVG are shown in Fig. 1. There was a close correlation between each approximated IDVG and original IDVG (reference plasma values: n = 50, r2 = 0.94, P < 0.0001; plasma values from the combined blood gas and glucose analyzer: n = 50, r2 = 0.92, P < 0.0001; whole blood values from the combined blood gas and glucose analyzer: n = 50, r2 = 0.89, P < 0.0001).
Discussion
Although bedside reflectance glucometers rarely overestimate or underestimate the 'true' glucose concentration by more than 40 mg/100 ml (2.2 mmol/l) [11], this margin of error is too great for measurement of IDVG. In addition, plasma protein concentrations, haematocrit and body temperature, as well as blood oxygen tension, may influence measurements from such devices significantly [12-14]. Accordingly, bedside glucometers were not used in our measurement of IDVG. Instead, we used a conventional but more accurate glucose analyzer, specifically a combined blood gas and glucose analyzer.
We demonstrated that approximated IDVG, calculated from either plasma or whole blood values using a conventional glucose analyzer, is not markedly different from original IDVG, with the two measures correlating closely. We recently reported that repeated IDVG measurements, done at an interval of 30 min, differ by 0.08 ± 0.32 l in haemodynamically stable patients [8]. Based on this finding the limits of clinical agreement for IDVG measurement can be set at ± 0.4 l, although the limits within which the two methods were considered to be interchangeable were set at ± 0.5 l/min for measurement of cardiac output [15]. Although the difference between approximated and original IDVG in the present study was not particularly great, it extended beyond the limits of agreement. Our previous study [9] also showed that the difference between approximated and original IDVG was 0.03 ± 0.43 l in 150 paired data using the same reference plasma glucose measurement system, again indicating that the methods are not interchangeable. However, bearing in mind the close correlation between the two measures and the clinically applicable procedure for measurement of approximated IDVG, the latter – measured using a conventional glucose analyzer (but not a bedside reflectance glucometer) – may be useful in the ICU.
We previously proposed [1-3] that IDVG represents central extracellular fluid volume status, including plasma volume and the interstitial fluid volume of highly perfused organs such as brain, heart, lungs, liver and kidneys, without modification of glucose metabolism and regardless of the presence or absence of peripheral oedema. Glucose rapidly traverses the red cell membrane by facilitated diffusion without requiring energy or insulin [16]. Because the mass concentration of water in plasma is 0.93 kg H2O/l and that in red cells is 0.71 kg H2O/l, whole blood has a mass concentration of water of approximately 0.84 kg H2O/l. Although the molality of glucose in plasma (mmol/kg H2O) is equal in red cells, the glucose concentration in plasma (mmol/l) is greater than in either red cells or whole blood, depending on the haematocrit of the blood sample [16]. There was no significant correlation between haematocrit and the difference between paired plasma and whole blood glucose data in the present study (r2 = 0.004), but the plasma glucose value was significantly greater than that in whole blood. However, the impact of this difference on incremental values would be less significant than that on absolute values. Thus, we may approximate IDVG from two whole blood glucose measurements, even measurements determined using a conventional glucose analyzer (but not a bedside reflectance glucometer). However, we believe that plasma glucose measurement is superior to whole blood glucose measurement, based on the bias and precision of the present data as well as by recommendation of plasma glucose rather than whole blood measurement, since the former is routinely used as the reference method [17].
Furthermore, a 5–10% decrease in whole blood glucose concentrations was observed during the first hour after sampling in routine conditions [18]. Whatever the calculation, it is important that all procedures be performed using proper technique and with an accurate sampling time.
The turnaround time for approximated IDVG measurement from the first blood sample to completion of the calculation is about 5 min in our ICU. In our experience, gained in more than 3500 determinations of original IDVG, it can be measured during routine fluid management, and it is not necessary to stabilize plasma glucose concentrations, provided that the infusion rate of glucose for routine fluid management remains unchanged before and during the measurement procedure. We observed a continuous decline in plasma glucose concentration over 60 min after injection, although plasma glucose concentrations at 60 min postinjection remained slightly elevated as compared with the preinjection value [8]. Hence, IDVG measurement will not induce a continued hyperglycaemic state, even in critically ill patients. However, Diaz-Parejo and coworkers [19] suggested that transient moderate hyperglycaemia had no adverse effect on outcome in patients with severe traumatic brain lesions and stroke. Therefore, we should be more concerned about normalization in basal plasma glucose concentration than about transient hyperglycaemia in these patients.
Gabbanelli and coworkers [6] utilized plasma glucose values, measured using a glucose analyzer similar to that used in the present study, to approximate IDVG based on the formula we proposed [9]. In accordance with our findings and corroborating our previous suggestions [1,3-5], those investigators found that approximated IDVG correlated well with both cardiac output and intrathoracic blood volume. Accordingly, either original or approximated IDVG is useful as an indirect measure of cardiac preload. Based on our clinical experience, normal IDVG is approximately 120 ml/kg, apparently high IDVG is above 140 ml/kg and apparently low IDVG is less than 100 ml/kg in the presence or absence of cardiac pathology or peripheral oedema. However, further detailed studies are required to determine the IDVG that are critical in terms of decision making regarding fluid management in different underlying pathologies.
Conclusion
We calculated approximated IDVG from plasma and whole blood glucose concentrations measured using a combined blood gas and glucose analyzer. The results indicate that either calculation of approximated IDVG exhibits a close linear correlation with original IDVG measured using a reference glucose analyzer, although they are not interchangeable. Our findings suggest that approximated IDVG is clinically relevant because it may be used for point-of-care testing to assess fluid volume.
Key messages
• IDVG has been proposed to be an indirect measure of cardiac preload without significant modification of glucose metabolism, but requiring repeated arterial blood samplings over 7 min after injection of glucose 5 g.
• Approximated IDVG derived from just two blood samples using a conventional glucose analyzer in the ICU is clinically acceptable as an alternative calculation of IDVG, although approximated and original IDVGs are not equivalent.
Abbreviations
AIC = Akaike's information criterion; ICU = intensive care unit; IDVG = initial distribution volume of glucose; SD = standard deviation.
Competing interests
The author(s) declare that they have no competing interests.
Authors' contributions
HI designed the study, performed statistical analysis and drafted the manuscript. HN, HO and TT collected data from the patients and performed calculations. KH designed the study and evaluated the data. All authors read and approved the final manuscript.
Acknowledgements
The authors thank Professor AH Giesecke Jr (Dallas, Texas, USA) and Professor D Grimaud (Nice, France) for continued support of the study.
Figures and Tables
Figure 1 Bland–Altman plots of the differences between each approximated IDVG and original IDVG. Approximated IDVG was calculated from a formula using the increased glucose concentration above baseline at 3 min after injection of glucose [9]. Shown are the reference plasma glucose measurement (left), a conventional plasma glucose measurement (middle) and a conventional whole blood measurement (right). Solid lines represent the mean difference, and dashed lines represent the 95% confidence interval.
Table 1 Patient demographics
Characteristic/parameter Value
Number of patients 50
Sex (male/female) 36/14
Age (years) 62 ± 12 (34–79)
Body weight (kg) 59.1 ± 12.9 (37.8–102)
Body surface area (m2) 1.60 ± 0.19 (1.27–2.14)
Number of patients receiving mechanical ventilation 26
Number of patients receiving catecholaminesa 14
Number of patients receiving insulinb 7
Number of patients receiving on continuous haemodiafiltration 4
Values are presented as mean ± standard deviation (range) or as number of patients.
aCatecholamines: an infusion of dopamine, dobutamine, noradrenaline, or adrenaline.
bInsulin: continuous insulin infusion.
Table 2 Approximated initial distribution volume of glucose using the incremental glucose level at 3 min postinjection
Δgl 3 min (mg/100 ml) IDVG (l) Δgl 3 min (mg/100 ml) IDVG (l) Δgl 3 min (mg/100 ml) IDVG (l)
31 12.3 61 6.6 91 4.3
32 12.0 62 6.5 92 4.2
33 11.8 63 6.4 93 4.2
34 11.5 64 6.3 94 4.2
35 11.2 65 6.2 95 4.1
36 11.0 66 6.1 96 4.1
37 10.7 67 6.0 97 4.0
38 10.5 68 5.9 98 4.0
39 10.3 69 5.8 99 4.0
40 10.0 70 5.7 100 3.9
41 9.8 71 5.6 101 3.9
42 9.6 72 5.5 102 3.8
43 9.4 73 5.4 103 3.8
44 9.2 74 5.4 104 3.8
45 9.0 75 5.3 105 3.7
46 8.8 76 5.2 106 3.7
47 8.7 77 5.1 107 3.7
48 8.5 78 5.1 108 3.7
49 8.3 79 5.0 109 3.6
50 8.1 80 4.9 110 3.6
51 8.0 81 4.8 111 3.6
52 7.8 82 4.8 112 3.5
53 7.7 83 4.7 113 3.5
54 7.5 84 4.7 114 3.5
55 7.4 85 4.6 115 3.5
56 7.2 86 4.5 116 3.5
57 7.1 87 4.5 117 3.4
58 7.0 88 4.4 118 3.4
59 6.9 89 4.4 119 3.4
60 6.7 90 4.3 120 3.4
Each initial distribution volume of glucose (IDVG) was calculated using a formula we previously proposed [9]. Δgl 3 min, increase in glucose concentration above the preinjection level at 3 min after injection.
Table 3 Glucose values and approximated initial distribution volume of glucose
Variable Referencea Plasmab Whole bloodb
Preinjection glucose (mg/100 ml) 158 ± 42 155 ± 41 153 ± 40
Incremental glucose(mg/100 ml)c 58 ± 11 57 ± 12 57 ± 11
Difference in glucose (mg/100 ml)d - -2 ± 3 -3 ± 4
Approximated IDVG (l) 7.26 ± 1.73 7.38 ± 1.8 7.40 ± 1.65
Difference from original IDVG (l) -0.17 ± 0.47 -0.05 ± 0.54 -0.04 ± 0.62
Values are presented as mean ± standard deviation. aFrom plasma glucose values using the same glucose analyzer for original IDVG. bUsing a conventional glucose analyzer (combined blood gas and glucose analyzer) for approximated IDVG. cThe incremental glucose value at 3 min after glucose injection. dDifference in glucose values in either plasma or whole blood from the reference plasma value. IDVG, initial distribution volume of glucose.
==== Refs
Ishihara H Suzuki A Okawa H Sakai I Tsubo T Matsuki A The initial distribution volume of glucose rather than indocyanine green derived plasma volume is correlated with cardiac output following major surgery Intensive Care Med 2000 26 1441 1448 11126254 10.1007/s001340000653
Ishihara H Matsui A Muraoka M Tanabe T Tsubo T Matsuki A Detection of capillary leakage by the indocyanine green and glucose dilutions in septic patients Crit Care Med 2000 28 620 626 10752804 10.1097/00003246-200003000-00003
Ishihara H Suzuki A Okawa H Ebina T Tsubo T Matsuki A Comparison of the initial distribution volume of glucose and plasma volume in thoracic fluid-accumulated patients Crit Care Med 2001 29 1532 1538 11505121 10.1097/00003246-200108000-00006
Ishihara H Shimodate Y Koh H Isozaki K Tsubo T Matsuki A The initial distribution of glucose and cardiac output in the critically ill Can J Anaesth 1993 40 28 31 8425240
Nakamura H Ishihara H Okawa H Yatsu Y Tsubo T Matsuki A Initial distribution volume of glucose is correlated with intrathoracic blood volume in hypovolemia and following volume loading in dogs Eur J Anaesthesiol 2005
Gabbanelli V Pantanetti S Donati A Montozzi A Carbini C Pelaia P Initial distribution volume of glucose as noninvasive indicator of cardiac preload: comparison with intrathoracic blood volume Intensive Care Med 2004 30 2067 2073 15448888 10.1007/s00134-004-2421-3
Mi W Ishihara H Sakai T Matsuki A Possible overestimation of indocyanine green-derived plasma volume early after induction of anesthesia with propofol/fentanyl Anesth Analg 2003 97 1421 1427 14570660 10.1213/01.ANE.0000084361.12884.D1
Rose BO Ishihara H Okawa H Panning B Piepenbrock S Matsuki A Repeatability of measurements of the initial distribution volume of glucose in haemodynamically stable patients J Clin Pharm Ther 2004 29 317 323 15271098 10.1111/j.1365-2710.2004.00565.x
Hirota K Ishihara H Tsubo T Matsuki A Estimation of the initial distribution volume of glucose by an incremental plasma glucose level at 3 min after i.v. glucose in humans Br J Clin Pharmacol 1999 47 361 364 10233198 10.1046/j.1365-2125.1999.00889.x
Akaike H A new look at the statistical model identification IEEE Trans Automat Control 1974 AC-19 716 723 10.1109/TAC.1974.1100705
Ray JG Hamielec C Mastracci T Pilot study of the accuracy of bedside glucometry in the intensive care unit Crit Care Med 2001 29 2205 2207 11700424 10.1097/00003246-200111000-00025
Maser RE Butler MA Decherney GS Use of arterial blood with bedside glucose reflectance meters in an intensive care unit: are they accurate? Crit Care Med 1994 22 595 599 8143469
Karcher RE Ingram RL Kiechle FL Sykes E Comparison of the HomoCue berta-glucose photometer and reflotron for open heart surgery Am J Clin Pathol 1993 100 130 134 8356945
Kurahashi K Maruta H Usuda Y Ohtsuka M Influence of blood sample oxygen tension on blood glucose concentration measured using an enzyme-electrode method Crit Care Med 1997 25 231 235 9034256 10.1097/00003246-199702000-00006
Zöller C Goetz AE Weis M Mörstedt K Pichler B Lamm P Kelger E Haller M Continuous cardiac output measurements do not agree with conventional bolus thermodilution cardiac output determination Can J Anaesth 2001 48 1143 1147 11744592
Fogh-Andersen N Wimberley PD Thode J Siggard-Andersen O Direct reading glucose electrodes detect the molality of glucose in plasma and whole blood Clin Chim Acta 1990 189 33 38 2383919 10.1016/0009-8981(90)90232-H
Kuwa K Nakayama T Hoshino T Tominaga M Relationships of glucose concentrations in capillary whole blood, venous whole blood and venous plasma Clin Chim Acta 2001 307 187 192 11369356 10.1016/S0009-8981(01)00426-0
Savolainen K Vitala A Puhakainen E Väisänen M Problems with the use of whole blood as a sample material in novel direct glucose analysers Scand J Clin Lab Invest 1990 50 221 223 2339284
Diaz-Parejo P Stahl N Xu W Reinstrup P Ungerstedt U Nodstrom CH Cerebral energy metabolism during transient hyperglycaemia in patients with severe brain trauma Intensive Care Med 2003 29 544 550 12655390
| 15774047 | PMC1175927 | CC BY | 2021-01-04 16:04:51 | no | Crit Care. 2005 Feb 11; 9(2):R144-R149 | utf-8 | Crit Care | 2,005 | 10.1186/cc3047 | oa_comm |
==== Front
Crit CareCritical Care1364-85351466-609XBioMed Central London cc30501577404910.1186/cc3050ResearchMicrovascular permeability during experimental human endotoxemia: an open intervention study van Eijk Lucas TGJ [email protected] Peter [email protected] Paul [email protected] Wim van den [email protected] Martijn PWJM [email protected] der Hoeven Johannes G [email protected] Departments of Intensive Care Medicine and Pharmacology-Toxicology, Radboud University Nijmegen Medical Centre, Nijmegen, The Netherlands2 Department of Pharmacology-Toxicology, Radboud University Nijmegen Medical Centre, Nijmegen, The Netherlands3 Department of Nuclear Medicine, Radboud University Nijmegen Medical Centre, Nijmegen, The Netherlands4 Department of Intensive Care Medicine and Nijmegen UniversityCenter for Infectious Diseases, Radboud University Nijmegen Medical Centre, Nijmegen, The Netherlands2005 21 2 2005 9 2 R157 R164 6 8 2004 9 12 2004 16 12 2004 10 1 2005 Copyright © 2005 van Eijk 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
Septic shock is associated with increased microvascular permeability. As a model for study of the pathophysiology of sepsis, endotoxin administration to humans has facilitated research into inflammation, coagulation and cardiovascular effects. The present study was undertaken to determine whether endotoxin administration to human volunteers can be used as a model to study the sepsis-associated increase in microvascular permeability.
Methods
In an open intervention study conducted in a university medical centre, 16 healthy volunteers were evaluated in the research unit of the intensive care unit. Eight were administered endotoxin intravenously (2 ng/kg Escherichia coli O113) and eight served as control individuals. Microvascular permeability was assessed before and 5 hours after the administration of endotoxin (n = 8) or placebo (n = 8) by three different methods: transcapillary escape rate of I125-albumin; venous occlusion strain-gauge plethysmography to determine the filtration capacity; and bioelectrical impedance analysis to determine the extracellular and total body water.
Results
Administration of endotoxin resulted in the expected increases in proinflammatory cytokines, temperature, flu-like symptoms and cardiovascular changes. All changes were significantly different from those in the control group. In the endotoxin group all microvascular permeability parameters remained unchanged from baseline: transcapillary escape rate of I125-albumin changed from 7.2 ± 0.6 to 7.7 ± 0.9%/hour; filtration capacity changed from 5.0 ± 0.3 to 4.2 ± 0.4 ml/min per 100 ml mmHg × 10-3; and extracellular/total body water changed from 0.42 ± 0.01 to 0.40 ± 0.01 l/l (all differences not significant).
Conclusion
Although experimental human endotoxaemia is frequently used as a model to study sepsis-associated pathophysiology, an endotoxin-induced increase in microvascular permeability in vivo could not be detected using three different methods. Endotoxin administration to human volunteers is not suitable as a model in which to study changes in microvascular permeability.
See related commentary
==== Body
Introduction
Sepsis is the leading cause of mortality in noncardiac intensive care units, resulting in an estimated mortality of 200,000 patients per year in the USA alone [1]. Sepsis is notably characterized by an increase in microvascular permeability, which accounts for the extravasation of macromolecules and fluid from the plasma to the tissues. The impaired diffusion of oxygen to cells as a result of the extracellular oedema appears to be a critical factor in the development of multiple organ failure [2,3]. Few studies have been conducted in humans to examine the mechanism that underlies the sepsis-associated increase in microvascular permeability.
Endotoxin is among the principal bacterial components that interacts with the host during Gram-negative sepsis [4]. Administration of endotoxin to humans is an appropriate model in which to investigate acute inflammatory responses (activation of cytokines and coagulation pathways) and to evaluate novel therapeutic interventions [5]. In vitro, exposure of human endothelial cells to endotoxin induces an increase in permeability [6], and in vivo an increase in microvascular permeability is among the major manifestations observed in animal endotoxaemia [7-12]. In humans, microvascular permeability can be assessed by plasma disappearance of a tracer (e.g. I125-albumin), changes in tissue volume caused by an imposed hydrostatic pressure and changes in bio-impedance. These methods were validated for the detection of a modest increase in microvascular permeability in patients with various diseases [13-17] and, more relevant to our study, patients with sepsis or septic shock [18-20]. In septic patients, transcapillary escape rate of albumin varies from 6.7%/hour [21] to 13.4%/hour [18], whereas permeability measured using venous congestion plethysmography (VCP) ranged from 6.1 ml/min per 100 ml mmHg × 10-3 [19] to 9.3 ml/min per 100 ml mmHg × 10-3 [22].
The present study was undertaken to determine whether endotoxin administration to human volunteers can be employed as a model in which to study the sepsis-associated increase in microvascular permeability.
Materials and methods
Subjects
After approval had been granted by the local ethics committee, 16 nonsmoking individuals gave written informed consent to participate in the study. Those who were taking prescription drugs or asprin or other nonsteroidal anti-inflammatory drugs were excluded (except for oral anticontraceptives). Screening of the participants before the test revealed no abnormalities in medical history or physical examination. Routine laboratory tests and electrocardiograms were normal. All participants were HIV and hepatitis B negative. They had not suffered a febrile illness within the 2 weeks preceding the study. Ten hours before the experiment, the participants refrained from consuming caffeine, alcohol and food.
Study design and procedures
Heart rate was continuously monitored using a three-lead electrogradiograph. An intra-arterial catheter in the radial artery permitted arterial blood sampling as well as continuous monitoring of blood pressure throughout the experiment. Forearm blood flow was measured in both arms using VCP, as described previously [23]. All participant received an intravenous infusion of a glucose/saline solution (2.5% glucose, 0.45% saline; 75 cm3/hour) via a cannula in an antecubital vein. At baseline, purified lipopolysaccharide (LPS) prepared from Escherichia coli O113 was injected intravenously (2 ng/kg) over 1–2 min in eight individuals, followed by 5 ml normal saline to ensure complete delivery. Another eight served as control individuals and received NaCl 0.9% instead of endotoxin in an equivalent volume. Because of obvious symptomatic changes after infusion of endotoxin, neither the volunteers nor the staff members were blinded to the study protocol.
The course over time of temperature, C-reactive protein, and plasma levels of tumour necrosis factor (TNF)-α and interleukin (IL)-1β [24] were monitored to confirm the inflammatory effects of endotoxin administration.
Transcapillary escape rate of I125albumin
Microvascular permeability determined by the transcapillary escape rate of I125-albumin (TER-alb) was measured at baseline and 5 hours after endotoxin administration, when haemodynamic changes are at their maximum [25]. I125 labelled albumin solution of 2 μCi (baseline) and 6 μCi (at 5 hours) in 5 cm3 normal saline were given as an intravenous bolus injection followed by 5 cm3 normal saline. The second dose is higher to overcome the background signal of the first dose. Arterial blood samples were drawn at baseline, and at 5, 10, 15, 20, 30, 45 and 60 min. Plasma radioactivity was measured in each sample using a scintillation detector (automatic γ-counter; 1480 Wizard 3", Wallac, Turku, Finland).
Venous congestion plethysmography
Microvascular permeability was also determined by VCP, in accordance with methods fully described previously [26,27]. Microvascular filtration capacity (Kf) – an index of vascular permeability – was measured using a protocol in which a series of eight small (10 mmHg) cumulative pressure steps were applied to venous congestion cuffs placed around both upper arms. Kf was estimated from alterations in forearm circumference due to the pressures applied, using the Filtrass strain gauge plethysmograph (Filtrass Angio, DOMED, Munich, Germany) [27]. Using this system, no change in the recorded signal is observed until ambient venous pressure in the arm is exceeded. At congestion cuff pressures greater than this value, each additional pressure increment causes a change in forearm volume that is attributed to vascular filling. When the congestion cuff pressure exceeds the isovolumetric venous pressure, a steady state change in volume is observed, reflecting fluid filtration. Kf reflects the product of the area available for fluid filtration and the permeability per unit surface area. Computer-based analysis enables differentiation between volume and filtration responses [28]. The value of Kf is determined by linear regression of the fluid filtration as a function of the cuff pressure. The slope of this relationship is Kf and the units are expressed as KfU (ml/min per 100 ml mmHg × 10-3) [28]. The files were recorded and saved for subsequent offline analysis. Kf measurements were conducted before, and 4.5 hours and 22 hours after the administration of endotoxin or normal saline.
Bioelectrical impedance analysis
In septic patients, fluid shifts from intracellular water to extracellular water (ECW) and an increase in total body water (TBW) occur because of an altered cellular membrane function, resulting in the formation of oedema. Bioelectrical impedance analysis (BIA) can estimate body composition parameters and has been used to estimate body water distribution and cellular membrane function in healthy individuals [29] and intensive care patients [20,30-33]. The principles of bioelectrical impedance postulate that resistance (R) is the opposition of TBW and electrolytes to the flow of an alternating current of low amplitude (800 μA) and high frequency (50 kHz). Reactance is the capacitance produced by tissue interfaces and cell membranes. An increase in microvascular permeability and an altered membrane function result in the formation of oedema, which decreases the resistance and reactance to an alternating electric current throughout the body. ECW will increase in relation to TBW, and reactance/resistance will decrease. BIA was performed using a body composition analyzer (Akern Srl, Florence, Italy). This device employs four-electrode polarization and measures the resistance and reactance of a conductor to application of an alternating electric current of 800 μA and 50 kHz. All measurements were made with the patient supine, with their arms relaxed at their sides but not touching their body, and with their thighs slightly separated. Electrodes were placed on the dorsal surface of the skin of the wrist and ankle, with the detector electrodes applied along the articulation bisecting line of both joints. BIA was performed at baseline and 4, 6, 8 and 22 hours after endotoxin administration.
Drugs and solutions
All solutions were freshly prepared on the day of the experiment. Endotoxin from Escherichia coli (batch 0:113, lot G2B274) was obtained from US Pharmacopia Convention (Rockville, MD, USA) and dissolved in normal saline 0.9% to a concentration of 200 EU/ml (0.1 ml/kg). I125-albumin (Iodinated [125I] Human Serum Albumin; code IM 17 P) was obtained from Amersham International (Amersham, UK).
Data analysis, calculations and statistics
Power analysis was based on clinically relevant changes in TER-Alb. In a previous study using the TER-alb method, we found a standard deviation ranging from 1.5% to 2.5%. An increase in transcapillary escape rate of 2.5% was considered clinically relevant. With an estimated standard deviation of 2% and α = 0.05, we calculated that a sample size of seven individuals per group would be needed to achieve a power of 95%. Therefore, eight individuals per group were included.
TER-alb was calculated and expressed as the percentage disappearance per hour. Fluid filtration capacity (Kf) was determined by venous occlusion plethysmography in both forearms and averaged. The mean Kf was used for further calculations. A change in the ratio of ECW/TBW was taken to give an impression of microvascular permeability, using BIA.
Student's t-tests or analysis of variance with repeated measures were used for the assessment of the effects of endotoxin on microvascular permeability parameters. All data are expressed as mean ± standard error of the mean of n experiments unless otherwise stated. P < 0.05 was considered statistically significant.
Results
Demographic characteristics of the participants are presented in Table 1. There were no significant differences between the groups.
Changes in clinical and inflammatory parameters
The first flu-like symptoms (headache, nausea, chills) occurred in the endotoxin-treated group between 55 and 90 min after LPS injection. Body temperature started to rise 1 hour after endotoxin administration to a maximum of 38.7 ± 0.3°C at 4 hours versus 36.9 ± 0.2°C in the control group (P < 0.001). At 8 hours all clinical symptoms had declined to control values. The clinical onset of inflammation was accompanied by a sudden rise in TNF-α plasma levels at 60 min (373 ± 71 pg/ml), which reached its zenith at 90 min (856 ± 158 pg/ml), closely followed by a rise in IL-1β that was maximal at 120 min (23.9 ± 2.2 pg/ml). C-reactive protein increased from under 5 mg/ml at baseline to 22.3 ± 1.4 mg/ml at 12 hours after endotoxin administration and reached its maximum at 22 hours (38.9 ± 3.0 mg/ml). In the control individuals no elevations in temperature (from 36.9 ± 0.1 to 37.0 ± 0.1°C), clinical symptoms, cytokine levels (TNF-α <8 pg/ml, IL-1β <8 pg/ml) or C-reative protein (<5 mg/ml) were observed (Fig. 1).
Changes in haemodynamic parameters
Figure 2 shows the course of heart rate, mean arterial pressure and forearm blood flow in the endotoxin and control group. In the control group the mean arterial blood pressure decreased from 88 to 80 mmHg at 6 hours (P = 0.035); the blood pressure decreased significantly more in the individuals administered LPS (from 96 ± 3 mmHg to 79 ± 4 mmHg at 6 hours, P < 0.0001; difference from control individuals: P = 0.002). Heart rate remained unchanged in the control group (from 66 ± 4 to 65 ± 2 beats/min; not significant) and increased from 63 ± 3 to 91 ± 3 beats/min at 6 hours in the LPS group (P < 0.0001). Forearm blood flow increased from 3.7 ± 0.6 to 6.8 ± 1.1 ml/min per dl at 6 hours (P = 0.018) in the endotoxin group, but remained unchanged in the control group (3.8 ± 0.8 versus 4.4 ± 0.9 ml/min per dl; not significant).
Changes in microvascular permeability parameters
In neither the endotoxin group nor the control group were significant alterations in microvascular permeability parameters detected. In the endotoxin group TER-alb was 7.2 ± 0.6%/hour before and 7.7 ± 0.9%/hour at 4.5 hours after endotoxin administration (not significant); Kf remained unchanged (from 5.0 ± 0.3 to 4.2 ± 0.4 ml/min per 100 ml mmHg × 10-3; not significant); and ECW/TBW, as measured by BIA, did not change (from 0.42 ± 0.01 l/l to 0.40 ± 0.01 l/l; not significant). Also, no significant changes in microvascular permeability parameters were found in the control group (all not significant: TER-alb from 9.08 ± 1.28 to 10.38 ± 0.63%/hour; Kf from 4.14 ± 0.42 to 5.17 ± 0.39 ml/min per 100 ml mmHg × 10-3; and ECW/TBW from 0.43 ± 0.01 l/l to 0.42 ± 0.01 l/l). The effect of endotoxin on microvascular parameters is shown in Fig. 3.
Discussion
Although administration of endotoxin to human volunteers has facilitated sepsis-associated research, the present study demonstrates that human experimental endotoxaemia is not a suitable model in which to study sepsis-induced changes in microvascular permeability. In a negative study the first issue to address is methodology. We conducted the present study with all three methods that are available for human in vivo experiments. Differences in microvascular permeability have been detected in various other diseases with these methods [13-17]. In septic patients an increase in Kf was demonstrated with TER-alb [18], VCP [19] and BIA [20]. In view of the ability of these methods to detect differences in microvascular permeability and the consistently negative findings of all three methods used in this endotoxin study, we believe our results are valid.
There are several possible reasons for our negative findings. First, the inflammatory stimulus might not have been sufficiently powerful. Endotoxin is known to stimulate the immune system in a dose-dependent manner [25]. Indeed, a marked increase in permeability in vivo has previously been shown in, for example, cats after intravenous administration of 1 mg/kg endotoxin [10]. On one occasion, an autointoxication with 1 mg of Salmonella endotoxin resulted in profound vasodilatory shock and a 15 l cumulative fluid balance over 72 hours in a laboratory worker [34]. This demonstrates unequivocally that high doses of endotoxin can cause shock and vascular leakage. In human volunteers an endotoxin concentration of 4 ng/kg is considered the maximal tolerable dose. The concentration of 2 ng/kg is widely applied and results in systemic inflammation, activation of coagulation pathways and distinct haemodynamic changes. Although the rise in proinflammatory cytokines is dose dependent, studies that used 4 ng/kg LPS found changes in clinical parameters similar to those reported here (e.g. rise in body temperature and fall in blood pressure) [35]. In the individuals included in the present study (who received 2 ng/kg) the flu-like symptoms, rise in body temperature, rise in heart rate, fall in blood pressure and rise in C-reactive protein were considerable; we therefore believe that the inflammatory stimulus was adequate. Also, the TNF-α and IL-1β concentrations in these individuals exceeded considerably the threshold levels of 50 pg/ml and 20 pg/ml, respectively, that are necessary to increase permeability significantly in vitro [6].
Naturally, it remains difficult for many reasons to compare an in vitro study in endothelial cells of large vessels with our in vivo experiment. The human endotoxaemia model is currently the only available in vivo human model that mimics Gram-negative sepsis. Whereas in experimental endotoxaemia the stimulus is restricted to LPS, other (non-LPS) bacterial components are also of importance for the induction of cytokines and the inflammatory response [36] and possibly the induction of vascular leakage. These differences could represent the reason why therapies directed at endotoxaemia itself are not of benefit in patients with septic shock [37]. However, as a model, the changes in haemodynamics that occur during human endotoxaemia are similar to those observed in septic shock, and suggest that endotoxin is a major mediator of the cardiovascular dysfunction that occurs in this condition [35].
A second possible reason for our negative findings is that not only the peak concentration of cytokines but also the duration of the increased level of the inflammatory mediators may be important in the pathophysiology of oedema formation in sepsis. The stimulus caused by a single bolus injection of endotoxin may be too short to induce an increase in microvascular permeability. The induction of capillary leakage in vitro was accomplished after incubation with endotoxin or cytokines for 6 hours [6]. Also, in pre-eclampsia a sustained rise in plasma cytokines is associated with an increase in microvascular permeability, suggesting a causal relationship [38]. However, although in some cases of sepsis in humans (e.g. meningococcal disease) elevated serum levels of TNF-α have been found in up to 90% of patients [39], several other clinical studies in septic patients reported only minimally elevated or undetectable levels of TNF-α [40,41]. Because these patients exhibit an overt increase in microvascular permeability, sustained high cytokine levels are apparently not mandatory for the development of oedema.
A third reason is that the timing of the measurements might not have been optimal for the detection of changes in permeability. In previous studies maximal changes in haemodynamic parameters were found between 2 and 6 hours after administration of endotoxin [35]. Because these vascular changes can partly be accounted for by endothelial dysfunction [42], we opted to measure microvascular permeability in the same time window. The possibility that an increase in permeability occurred outside the time window of interest appears unlikely because BIA was unchanged at five time points during the experiment, and Kf was also unaltered at 22 hours after endotoxin administration. Timing may be of critical importance because an accelerated plasma efflux of albumin was only observed during the early phase of sepsis in rats [43]. Also, late-acting cytokines (e.g. high mobility group protein 1) remain elevated for 16–32 hours after the administration of endotoxin and may play a role in the capillary leak found in septic patients. This and possibly other mediators were not measured during our experiment. Again, BIA and VCP measurements after 22 hours did not reveal an increase in vascular permeability in our experiments, suggesting that a possible late increase in vascular permeability was not missed.
Finally, oedema formation may differ from tissue to tissue and from organ to organ. In human endotoxaemia increases in intestinal permeability [44] and alveolar epithelial permeability [45] were previously demonstrated. In contrast, human endotoxaemia did not induce an increase in the ocular blood–aqueous barrier [46]. With the TER-alb and BIA whole body permeability is assessed, whereas the Filtrass strain gauge plethysmograph focuses on the forearms. An increase in microvascular permeability in, for example, the lungs was not specifically assessed, but if it was present it was insufficient to affect whole body permeability. Administration of iodated albumin as a measure of capillary leak may vary with hydration status, and albumin molecules might be too large to be useful as a sensitive permeability marker. However, these problems are overcome with the use of VCP. We believe that fluid loading would not have altered transcappilary leakage, because with the VCP method a venous occlusion pressure is applied to the forearms, so that vascular permeability is measured independent of the volume status of the subject. The suggestion that permeability might have been increased for smaller molecules than albumin can be ruled out for the same reason.
In summary, we do not believe that the methods used, the timing of the permeability measurements, or the absolute maximal cytokine concentrations can account for the observed lack of effect of endotoxin on microvascular permeability in humans. However, the short duration of cytokine increase possibly played a role.
Conclusion
Although endotoxin administration to humans has proven to be a valuable model for studying systemic inflammation and coagulation, this model cannot be used to investigate the pathophysiological mechanisms that underlie capillary leakage in sepsis or to evaluate pharmacological interventions aimed at attenuating the increase in microvascular permeability.
Key messages
• Endotoxin administration to humans is a valuable model in which to investigate inflammatory and haemodynamic mechanisms in sepsis.
• Endotoxin administration to humans does not affect microvascular permeability measured using TER-alb, VCP and BIA.
• Endotoxin administration can not be used as a model to study the pathopysiological mechanisms that underlie capillary leakage in sepsis, or to evaluate the pharmacological interventions aimed at restoring normal microvascular permeability.
Abbreviations
BIA = bioelectrical impedance analysis; ECW = extracellular water; IL = interleukin; Kf = filtration capacity; LPS = lipopolysaccharide; TER-alb = transcapillary escape rate of I125-albumin; TNF = tumour necrosis factor; TBW = total body water; VCP = venous congestion plethysmography.
Competing interests
The author(s) declare that they have no competing interests.
Authors' contributions
LTGJvE (medical student) carried out the experiments, performed the statistical analysis and drafted the manuscript. PP conceived the study, and supervised the experiments and writing of the paper. PS participated in the design of the study and corrected the manuscript. WvdB administrated the Alb125 to the participants and measured the plasma radioactivity. MPWJMB (research nurse) assisted with the coordination and practical conduction of the experiments. JGvdH participated in the design of the study and corrected the manuscript. All authors read and approved the final manuscript.
Acknowledgement
PP is a recipient of a Clinical Fellowship grant of the Netherlands Organisation for Scientific Research (ZonMw).
Figures and Tables
Figure 1 Changes in inflammatory parameters. Administration of endotoxin (n = 8; 2 ng/kg) resulted in a marked increase in tumour necrosis factor (TNF)-α (closed squares, left axis) and IL-1β (open squares, right axis). In control individuals cytokine levels remained below the detection limit (n = 8; data not shown). Cytokine release was associated with fever and an increase in C-reactive protein (CRP; endotoxin group, closed squares; control group, open circles). Data are expressed as means ± standard error of the mean. The P values in the figure refer to the difference between endotoxin and control groups as analyzed using analysis of variance with repeated measures over the complete curve.
Figure 2 Changes in haemodynamic parameters. Administration of endotoxin (2 ng/kg; n = 8; closed squares) resulted in a significant increase in heart rate (HR; measured using electrocardiography; P < 0.0001), a significant decrease in mean arterial pressure (MAP; measured intra-arterially; P < 0.0001) and a significant increase in forearm blood flow (FBF; measured using venous occlusion plethysmography; P = 0.018). HR and FBF did not change significantly in the control group (open circles; n = 8), whereas MAP decreased (P = 0.035). MAP decreased significantly more in the endotoxin group than in the control group (P = 0.002). These changes demonstrate that endotoxin induces a vasodilatory state. Data are expressed as means ± standard error of the mean. The P values in the figure refer to the difference between endotoxin and control group as analyzed using analysis of variance with repeated measures over the complete curve.
Figure 3 Changes in microvascular permeability parameters. Microvascular permeability parameters were measured using transcapillary escape rate of I125-albumin (TER-alb), venous congestion plethysmography (VCP) and bioelectrical impedance analysis (BIA). There were no changes in microvascular permeability as measured using all three parameters in either the endotoxin group (n = 8; 2 ng/kg; closed squares) or in the control group (n = 8; open circles). Data are expressed as means ± standard error of the mean. ECW, extracellular water; TBW, total body water.
Table 1 Demographic characteristics of the participants
Parameter Endotoxin group Control group
n (male/female) 8 (4/4) 8 (4/4)
Age (years) 23.9 ± 1.0 22.5 ± 0.8
Length (cm) 176 ± 5 183 ± 3
Weight (kg) 71.8 ± 4.9 70.1 ± 3.6
BMI (kg/m2) 23.0 ± 0.7 20.9 ± 0.8
SBP/DBP (mmHg) 127 ± 2/80 ± 3 119 ± 3/73 ± 3
Forearm volume (ml) 1019 ± 95 931 ± 67
Data are expressed as mean ± standard deviation. There were no significant differences between the groups. BMI, body mass index; DBP, diastolic blood pressure; SBP, systolic blood pressure.
==== Refs
Angus DC Linde-Zwirble WT Lidicker J Clermont G Carcillo J Pinsky MR Epidemiology of severe sepsis in the United States: analysis of incidence, outcome, and associated costs of care Crit Care Med 2001 29 1303 1310 11445675 10.1097/00003246-200107000-00002
Fietsam R JrVillalba M Glover JL Clark K Intra-abdominal compartment syndrome as a complication of ruptured abdominal aortic aneurysm repair Am Surg 1989 55 396 402 2729780
Stone HH Fulenwider JT Renal decapsulation in the prevention of post-ischemic oliguria Ann Surg 1977 186 343 355 407854
van Deventer SJ Buller HR ten Cate JW Sturk A Pauw W Endotoxaemia: an early predictor of septicaemia in febrile patients Lancet 1988 1 605 609 2894546 10.1016/S0140-6736(88)91412-2
Martich GD Boujoukos AJ Suffredini AF Response of man to endotoxin Immunobiology 1993 187 403 416 8330905
Nooteboom A Van Der Linden CJ Hendriks T Tumor necrosis factor-alpha and interleukin-1beta mediate endothelial permeability induced by lipopolysaccharide-stimulated whole blood Crit Care Med 2002 30 2063 2068 12352042 10.1097/00003246-200209000-00019
Balsa D Merlos M Giral M Ferrando R Garcia-Rafanell J Forn J Effect of endotoxin and platelet-activating factor on rat vascular permeability: role of vasoactive mediators J Lipid Mediat Cell Signal 1997 17 31 45 9302653 10.1016/S0929-7855(97)00019-9
Green K Paterson CA Cheeks L Slagle T Jay WM Aziz MZ Ocular blood flow and vascular permeability in endotoxin-induced inflammation Ophthalmic Res 1990 22 287 294 2090983
Hinder F Booke M Traber LD Traber DL Nitric oxide and endothelial permeability J Appl Physiol 1997 83 1941 1946 9390966
Holbeck S Grande PO Endotoxin increases both protein and fluid microvascular permeability in cat skeletal muscle Crit Care Med 2003 31 560 565 12576966 10.1097/01.CCM.0000048620.88344.70
Laszlo F Whittle BJ Evans SM Moncada S Association of microvascular leakage with induction of nitric oxide synthase: effects of nitric oxide synthase inhibitors in various organs Eur J Pharmacol 1995 283 47 53 7498320 10.1016/0014-2999(95)00281-O
Sakai I Ishihara H Iwakawa T Suzuki A Matsuki A Ratio of indocyanine green and glucose dilutions detects capillary protein leakage following endotoxin injection in dogs Br J Anaesth 1998 81 193 197 9813522
Brown MA Zammit VC Lowe SA Capillary permeability and extracellular fluid volumes in pregnancy-induced hypertension Clin Sci (Lond) 1989 77 599 604 2691173
Morita H Takeuchi K Funakoshi T Mizutori M Maruo T Potential use of bioelectrical impedance analysis in the assessment of edema in pregnancy Clin Exp Obstet Gynecol 1999 26 151 154 10668141
Vervoort G Lutterman JA Smits P Berden JH Wetzels JF Transcapillary escape rate of albumin is increased and related to haemodynamic changes in normo-albuminuric type 1 diabetic patients J Hypertens 1999 17 1911 1916 10703889 10.1097/00004872-199917121-00021
Bedogni G Polito C Severi S Strano CG Manzieri AM Alessio M Iovene A Battistini N Altered body water distribution in subjects with juvenile rheumatoid arthritis and its effects on the measurement of water compartments from bioelectric impedance Eur J Clin Nutr 1996 50 335 339 8793411
Pedrinelli R Dell'Omo G Bandinelli S Penno G Mariani M Transvascular albumin leakage and forearm vasodilatation to acetylcholine in essential hypertension Am J Hypertens 2000 13 256 261 10777029 10.1016/S0895-7061(00)00250-8
Fleck A Raines G Hawker F Trotter J Wallace PI Ledingham IM Calman KC Increased vascular permeability: a major cause of hypoalbuminaemia in disease and injury Lancet 1985 1 781 784 2858667 10.1016/S0140-6736(85)91447-3
Christ F Gamble J Gartside IB Kox WJ Increased microvascular water permeability in patients with septic shock, assessed with venous congestion plethysmography (VCP) Intensive Care Med 1998 24 18 27 9503218 10.1007/s001340050509
Marx G Vangerow B Burczyk C Gratz KF Maassen N Cobas MM Leuwer M Kuse E Rueckholdt H Evaluation of noninvasive determinants for capillary leakage syndrome in septic shock patients Intensive Care Med 2000 26 1252 1258 11089750 10.1007/s001340000601
Margarson MP Soni NC Effects of albumin supplementation on microvascular permeability in septic patients J Appl Physiol 2002 92 2139 2145 11960967 10.1063/1.1495889
Bethell DB Gamble J Pham PL Nguyen MD Tran TH Ha TH Tran TN Dong TH Gartside IB White NJ Noninvasive measurement of microvascular leakage in patients with dengue hemorrhagic fever Clin Infect Dis 2001 32 243 253 11170914 10.1086/318453
Whitney RJ The measurement of volume changes in human limbs J Physiol 1953 121 1 27 13085295
Prabhakar U Eirikis E Davis HM Simultaneous quantification of proinflammatory cytokines in human plasma using the LabMAP assay J Immunol Methods 2002 260 207 218 11792390 10.1016/S0022-1759(01)00543-9
Suffredini AF Hochstein HD McMahon FG Dose-related inflammatory effects of intravenous endotoxin in humans: evaluation of a new clinical lot of Escherichia coli O:113 endotoxin J Infect Dis 1999 179 1278 1282 10191237 10.1086/314717
Christ F Bauer A Brugger D Niklas M Gartside IB Gamble J Description and validation of a novel liquid metal-free device for venous congestion plethysmography J Appl Physiol 2000 89 1577 1583 11007598
Gamble J Gartside IB Christ F A reassessment of mercury in silastic strain gauge plethysmography for microvascular permeability assessment in man J Physiol 1993 464 407 422 8229810
Gamble J Bethell D Day NP Loc PP Phu NH Gartside IB Farrar JF White NJ Age-related changes in microvascular permeability: a significant factor in the susceptibility of children to shock? Clin Sci (Lond) 2000 98 211 216 10657278
Van Loan MD Bioelectrical impedance analysis to determine fat-free mass, total body water and body fat Sports Med 1990 10 205 217 2247723
Chiolero RL Gay LJ Cotting J Gurtner C Schutz Y Assessment of changes in body water by bioimpedance in acutely ill surgical patients Intensive Care Med 1992 18 322 326 1469158
Mattar JA Application of total body bioimpedance to the critically ill patient. Brazilian Group for Bioimpedance Study New Horiz 1996 4 493 503 8968982
Roos AN Westendorp RG Frolich M Meinders AE Weight changes in critically ill patients evaluated by fluid balances and impedance measurements Crit Care Med 1993 21 871 877 8504655
Scheltinga MR Jacobs DO Kimbrough TD Wilmore DW Identifying body fluid distribution by measuring electrical impedance J Trauma 1992 33 665 670 1464913
Taveira da Silva AM Kaulbach HC Chuidian FS Lambert DR Suffredini AF Danner RL Brief report: shock and multiple-organ dysfunction after self-administration of Salmonella endotoxin N Engl J Med 1993 328 1457 1460 8479465 10.1056/NEJM199305203282005
Suffredini AF Fromm RE Parker MM Brenner M Kovacs JA Wesley RA Parrillo JE The cardiovascular response of normal humans to the administration of endotoxin N Engl J Med 1989 321 280 287 2664516
Sprong T Stikkelbroeck N van der LP Steeghs L van Alphen L Klein N Netea MG van der Meer JW van Deuren M Contributions of Neisseria meningitidis LPS and non-LPS to proinflammatory cytokine response J Leukoc Biol 2001 70 283 288 11493621
Corriveau CC Danner RL Endotoxin as a therapeutic target in septic shock Infect Agents Dis 1993 2 35 43 8162352
Anim-Nyame N Gamble J Sooranna SR Johnson MR Steer PJ Microvascular permeability is related to circulating levels of tumour necrosis factor-alpha in pre-eclampsia Cardiovasc Res 2003 58 162 169 12667958 10.1016/S0008-6363(02)00844-1
Girardin E Grau GE Dayer JM Roux-Lombard P Lambert PH Tumor necrosis factor and interleukin-1 in the serum of children with severe infectious purpura N Engl J Med 1988 319 397 400 3135497
Oberholzer A Oberholzer C Moldawer LL Cytokine signaling: regulation of the immune response in normal and critically ill states Crit Care Med 2000 28 N3 N12 10807312 10.1097/00003246-200004001-00002
Pruitt JH Welborn MB Edwards PD Harward TR Seeger JW Martin TD Smith C Kenney JA Wesdorp RI Meijer S Cuesta MA Abouhanze A Increased soluble interleukin-1 type ii receptor concentrations in postoperative patients and in patients with sepsis syndrome Blood 1996 87 3282 3288 8605344
Pleiner J Mittermayer F Schaller G MacAllister RJ Wolzt M High doses of vitamin c reverse Escherichia coli endotoxin-induced hyporeactivity to acetylcholine in the human forearm Circulation 2002 106 1460 1464 12234948 10.1161/01.CIR.0000030184.70207.FF
Ruot B Papet I Bechereau F Denis P Buffiere C Gimonet J Glomot F Elyousfi M Breuille D Obled C Increased albumin plasma efflux contributes to hypoalbuminemia only during early phase of sepsis in rats Am J Physiol Regul Integr Comp Physiol 2003 284 R707 R713 12571074
O'Dwyer ST Michie HR Ziegler TR Revhaug A Smith RJ Wilmore DW A single dose of endotoxin increases intestinal permeability in healthy humans Arch Surg 1988 123 1459 1464 3142442
Suffredini AF Shelhamer JH Neumann RD Brenner M Baltaro RJ Parrillo JE Pulmonary and oxygen transport effects of intravenously administered endotoxin in normal humans Am Rev Respir Dis 1992 145 1398 1403 1596009
Herman DC Suffredini AF Parrillo JE Palestine AG Ocular permeability after systemic administration of endotoxin in humans Curr Eye Res 1991 10 121 126 2036803
| 15774049 | PMC1175929 | CC BY | 2021-01-04 16:04:51 | no | Crit Care. 2005 Feb 21; 9(2):R157-R164 | utf-8 | Crit Care | 2,005 | 10.1186/cc3050 | oa_comm |
==== Front
Crit CareCritical Care1364-85351466-609XBioMed Central London cc30541577404810.1186/cc3054ResearchComparison between logistic regression and neural networks to predict death in patients with suspected sepsis in the emergency room Jaimes Fabián [email protected] Jorge [email protected] Diego [email protected]ínez Carlos [email protected] Associate Professor, Department of Internal Medicine and Escuela de Investigaciones Médicas Aplicadas (EIMA – GRAEPI), School of Medicine, Universidad de Antioquia, Medellín, Colombia2 Chairman, Department of Physiology, Universidad de Antioquia, Medellín, Colombia3 Assistant Professor, Department of Physiology, Universidad de Antioquia, Medellín, Colombia4 Assistant Physician, Department of Internal Medicine, Division of Pulmonary and Critical Care Medicine, Fundación Santa Fe de Bogotá, Bogotá, Colombia2005 17 2 2005 9 2 R150 R156 5 10 2004 1 12 2004 17 12 2004 13 1 2005 Copyright © 2005 Jaimes 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
Neural networks are new methodological tools based on nonlinear models. They appear to be better at prediction and classification in biological systems than do traditional strategies such as logistic regression. This paper provides a practical example that contrasts both approaches within the setting of suspected sepsis in the emergency room.
Methods
The study population comprised patients with suspected bacterial infection as their main diagnosis for admission to the emergency room at two University-based hospitals. Mortality within the first 28 days from admission was predicted using logistic regression with the following variables: age, immunosuppressive systemic disease, general systemic disease, Shock Index, temperature, respiratory rate, Glasgow Coma Scale score, leucocyte counts, platelet counts and creatinine. Also, with the same input and output variables, a probabilistic neural network was trained with an adaptive genetic algorithm. The network had three neurone layers: 10 neurones in the input layer, 368 in the hidden layer and two in the output layer. Calibration was measured using the Hosmer-Lemeshow goodness-of-fit test and discrimination was determined using receiver operating characteristic curves.
Results
A total of 533 patients were recruited and overall 28-day mortality was 19%. The factors chosen by logistic regression (with their score in parentheses) were as follows: immunosuppressive systemic disease or general systemic disease (2), respiratory rate 24–33 breaths/min (1), respiratory rate ≥ 34 breaths/min (3), Glasgow Come Scale score ≤12 (3), Shock Index ≥ 1.5 (2) and temperature <38°C (2). The network included all variables and there were no significant differences in predictive ability between the approaches. The areas under the receiver operating characteristic curves were 0.7517 and 0.8782 for the logistic model and the neural network, respectively (P = 0.037).
Conclusion
A predictive model would be an extremely useful tool in the setting of suspected sepsis in the emergency room. It could serve both as a guideline in medical decision-making and as a simple way to select or stratify patients in clinical research. Our proposed model and the specific development method – either logistic regression or neural networks – must be evaluated and validated in an independent population.
See related commentary
==== Body
Introduction
Sepsis is the second leading cause of death among patients in noncoronary intensive care units (ICUs) and is the 10th leading cause of death overall in the USA [1]. Despite new and complex therapies, the incidence of sepsis has increased annually at a constant rate over the past 20 years, and there have been no substantial changes in the associated mortality [2].
A tool that could stratify the severity of sepsis from the initial stages in the clinical course would enhance our understanding of this disorder and its management. A simple system designed to estimate the probability of death would represent the basis for improved diagnosis, prognostication and treatment. Specifically, such a model, in the setting of the emergency room (ER), could guide decisions regarding ICU admission or whether a particular type of therapy should be instituted. The strategy may be developed from the definitions proposed by the American College of Chest Physicians/Society of Critical Care Medicine in 1992 [3]. These definitions include a generalized process with clinical findings that may represent an initial phase during the sepsis phenomenon – the systemic inflammatory response syndrome (SIRS). Although the natural history seems to reflect a continuum through different stages of an inflammatory response, from SIRS to septic shock [4], an unequivocal linear sequence of events is far from clinically apparent. Thus, classical analytical models, such as logistic regression, are limited in terms of their ability to elucidate the interplay that underlies the sepsis phenomenon.
Advances in statistical methods have supplied the tools necessary to model complex nonlinear relationships among many variables relevant to biological systems. Artificial neural networks (ANNs) are computer programs that simulate some of the higher level functions of the human brain. As in the brain, there are neurones and synapses, with various synaptic connection strengths – called 'weights' – for each connected pair of neurones. However, unlike the brain but similar to many computer programs, there is a specific set of input and output neurones for each problem and each net. These input and output neurones correspond to the inputs to and outputs from a traditional computer program. The other, termed 'hidden' neurones, along with the synapses and weights, correspond to the instructions in a traditional program. Use of ANNs as clinical prediction models has been explored in many areas of medicine, including nephrology [5], microbiology [6], radiology [7] and neurology [8]. Thus far, however, we are unaware of their use in sepsis. In this study we present a practical example that contrasts the abilities of logistic regression and neural networks to predict death in patients admitted to the ER with suspected sepsis as their main cause of hospitalization.
Materials and methods
Study design
In this longitudinal cohort study, patients were recruited between August 1998 and March 1999. Starting from admission to the ER, the patients were followed for 28 days or until death.
Setting
The patients were admitted to the ERs of two reference hospitals: the Hospital Universitario San Vicente de Paúl and the Hospital General de Medellín. Hospital Universitario San Vicente de Paúl is a 550-bed, fourth level university hospital, and is a referral centre for a region including approximately 3 million habitants. Hospital General de Medellín is a 300-bed, third level teaching hospital, and is a referral centre for the metropolitan area. Both are located in Medellín, Colombia.
Participants
We included patients aged 15 years or older with any suspected or confirmed bacterial infection as their admission diagnosis and at least one of the following SIRS criteria: temperature >38°C or <36°C; and leucocyte count >12000/mm3, <4000/mm3, or >10% immature forms (bands). We excluded eligible participants if they, their relatives, or their doctors refused to provide consent to participate in the study, or if they died or were discharged before 24 hours. Ethics committees of both hospitals had previously approved the protocol, and patients or their legal representatives signed an informed consent form.
Measurements
The primary outcome variable was mortality within the first 28 days after admission to the ER. For those patients who were discharged before day 28, an evaluation of their vital status was conducted in the outpatient control centre or by phone if a personal interview was not possible. Independent variables recorded at admission were as follows: age, immunosuppressive systemic disease (ISD; i.e. any of cancer, chemotherapy, steroid use or AIDS), general systemic disease (GSD; i.e. any of cardiac failure, diabetes, renal failure, chronic obstructive lung disease, or cirrhosis), Shock Index (heart rate/systolic arterial pressure), body temperature, respiratory rate, Glasgow Coma Scale (GCS) score, leucocyte count, platelet count and creatinine blood level. Research assistants in the ER collected clinical variables at admission in a standardized manner. Laboratory variables were analyzed using standard quality control procedures at the participating institutions. Missing data for continuous variables were estimated with simple imputations using the median nonmissing value. In total, estimation procedures were performed in 2.6% (14 simple records) of baseline values.
Data analysis and management
The procedure for the logistic model has been described in detail elsewhere [9]. Briefly, we conducted univariate logistic regression analysis for each candidate variable, with P < 0.25 being the criterion for acceptance in the model. Collinearity was checked with a matrix of correlations, using the Spearman rank correlation coefficient between independent variables. We chose a conservative strategy, with r ≥ 0.4 in at least one correlation as the criterion for multicollinearity. Logistic model assumptions (i.e. no interaction terms and a linear relationship between the logit and the continuous covariates) were verified. Then, a logistic regression analysis, employing a forward stepwise inclusion method, was developed using a P value of 0.05 at entry. This automatic procedure was contrasted with a backward elimination method and with a full model that included all of the candidate variables, in order to confirm the validity and stability of our results. For continuous variables, the cutoff points for changes in the probability of death were explored with locally weighted regression analysis and the lowess procedure [10]. According to the cutoff points detected, dummy variables were constructed and a new logistic regression model was fitted with those variables. In order to obtain the simplest score with the same scale within and between ranges of physiological variables and co-morbid conditions, the regression coefficients were all divided by the lowest one, and then rounded off to the nearest whole number, as the weight reflecting 'risk' for death for each variable. In defining the severity levels by the size of the coefficients, comparable severity levels within variables or conditions were grouped together. The global score for every patient in the cohort was calculated and a new logistic regression equation with the score as independent variable was fitted.
The model calibration – observed mortality versus that predicted with the score – was evaluated using the Hosmer-Lemeshow goodness-of-fit test. The test result, under a χ2 distribution, provides a P value in which higher values (P > 0.05) indicate nonsignificant differences between observed and predicted mortality. The discriminatory ability – the capacity of the model to separate survivors from nonsurvivors, with 1.0 and 0.5 meaning perfect and random discrimination, respectively – was determined using receiver operating characteristic (ROC) curve analysis. Internal validation was done with 2000 bootstrap replications of the model. All statistical analyses were performed with Stata Statistical Software, Release 7.0 (Stata Corporation, College Station, TX, USA).
Using the same input and output variables, a probabilistic neural network was trained using an adaptive genetic algorithm (NeuroShell©; Ward Systems Group Inc., Frederick, MD, USA). The network has three neurone layers, with 10 neurones in the input layer, 368 in the hidden layer and two in the output layer, the latter indicating death versus survival. Of the cohort 75% was used to train the network and 25% was used in testing. The training criterion was that 20 generations had elapsed without changes in the minimum error. The general performance of the neural network was evaluated using the ROC curve and the Hosmer-Lemeshow goodness-of-fit test. The difference between the two ROC curves – logistic regression and neural network – was tested using the Wilcoxon statistic based on pairwise comparisons [11].
Results
A total of 542 potentially eligible participants were admitted during the study period. Nine were excluded because of death (n = 5) or discharge (n = 4) during the first 24 hours. The final study population therefore included 533 patients, 55% (n = 293) of whom were male. Their age (mean ± standard deviation) was 48 ± 21 years, and their median hospital stay was 8 days (interquartile range 4–15 days). Overall 28-day mortality was 19% (n = 101), and 14% (n = 75) of the cohort was admitted to ICU.
The most common diagnoses suspected at admission were community-acquired pneumonia (recorded in 36% of patients), followed by soft tissue infection (17%), intra-abdominal infection (12%), urinary tract infection (11%) and others (11%); sepsis of undetermined source was recorded in 13% patients. The major pre-existing conditions related to admission were trauma or surgery more than 24 hours before admission (21%), chronic obstructive pulmonary disease (12%), diabetes (13%) and miscellaneous others (9%). Of the patients, 45% were free of associated diseases.
A total of 283 (53%) out of 533 cases of clinically suspected bacterial infection were microbiologically confirmed, 113 of which (40%) grew on blood samples. The rate of positive blood cultures among the total requested was 27%, and the most frequently isolated micro-organisms were Escherichia coli (19%), Staphylococcus aureus (16%), Streptococcus pneumoniae (13%), Staphylococcus coagulase negative (13%), Klebsiella pneumoniae (9%), Enterobacter spp. (6%), Enterococcus spp. (4%), Streptococcus pyogenes (3%), nonfermenting Gram-negative bacilli (3%) and others (14%).
After conducting univariate analysis for the logistic regression, leucocyte count was considered ineligible for inclusion in the model (P = 0.893). The evaluation of collinearity was carried out for all variables using the Spearman correlation coefficient. A significant correlation (r = 0.44) was found between age and GSD (P = 0.0000). Similar correlations, but to a lesser degree, were found between age and Shock Index (r = 0.1453; P = 0.0008) and between age and temperature (r = 0.1940; P = 0.0000). Therefore, age was excluded from the predictor variables. A multiple logistic regression model was applied to the overall 28-day mortality, taking into account GSD, ISD, Shock Index, respiratory rate, temperature, GCS score, creatinine and platelet count as predictive variables. This model allowed us to discard the latter two variables because they were statistically nonsignificant. For the variables respiratory rate, temperature, Shock Index and GCS score, the cutoff points for changes in the probability of death were sought by locally weighted regression. The results are shown in Table 1.
With the previous values, 12 dummy variables were constructed considering the first level (1) as the reference value. These new variables, in conjunction with the two nominal variables previously involved (GSD and ISD), were fitted in a new logistic regression model for prediction of mortality. After dividing and rounding off coefficients to the nearest whole number, some levels and variables were bound together, namely co-morbid conditions, GCS score, Shock Index and body temperature. The final meaningful variables are summarized in Table 2 according to their levels and relative weights.
In this way the final scale of severity was a range between 0 and 12. With these data, the score for each patient in the cohort was calculated, and a model that provides an estimate of severity, defined as the probability of 28-day mortality, was obtained. The Hosmer-Lemeshow goodness-of-fit test yielded a value of 7.54 (P = 0.5807). By ROC curve analysis for discriminative capacity, the area under the curve was 0.7517. The bootstrapped coefficients for 2000 replications exhibited standard errors of under 10% of those observed in the model, and the values for the Hosmer-Lemeshow goodness-of-fit test and the area under the ROC curve in this set were 8.96 (P = 0.4321) and 0.7119, respectively.
The neural network included all of the independent variables. Their weight, by the smoothing factor, ranged from 2.65 for temperature to 0.34 for ISD. The Hosmer-Lemeshow goodness-of-fit test yielded a value of 8.03 (P = 0.475), and the area under the ROC curve was 0.8782. The difference between ROC curves was statistically significant according to the Wilcoxon statistic based on pairwise comparisons (P = 0.037). Figure 1 shows the comparison of observed and predicted deaths with both methods.
Discussion
The present study shows that it is possible to obtain a simple indicator of the risk for death under clinical conditions compatible with severe infections. The system uses variables taken from the initial clinical interview and physical examination, all of which are available at the moment of admission to the ER. This suggests that it is possible to develop a reproducible and transportable predictive instrument in patients with signs indicative of sepsis. However, the model must be specifically tested in an independent population with a larger sample size. The main determinants of mortality reflect two acknowledged host factors, namely co-morbid conditions and the type of individual biological response, the latter being determined from clinical findings such as vital signs and GCS score.
The use of ANNs in the setting of sepsis has not been explored. However, with regard to overall mortality in ICUs, two recent studies compared hospital outcome prediction using neural networks versus logistic regression [12,13]. Clermont and coworkers [12] designed a prospective cohort study including 1647 patients admitted to seven ICUs at a tertiary care centre. The predictor variables considered were age and the acute physiology variables of the Acute Physiology and Chronic Health Evaluation (APACHE) III score. They constructed logistic regression and ANN models for a random set of 1200 admissions (development set), and used the remaining 447 admissions as the validation set. Then, model construction was repeated on progressively smaller development sets (800, 400 and 200 admissions) and re-tested in the original validation set. As the size of the development set sample decreased, the performance of the model on the validation set deteriorated rapidly, although the ANNs retained marginally better fit than logistic regression, as measured using the Hosmer-Lemeshow test, at 800 admissions. At under 800 admissions, however, the fit was poor with both approaches. The authors concluded that both ANN and logistic regression have similar performance with appropriate sample size, and share the same limitations with development sets on small samples.
Nimgaonkar and coworkers [13] compared the performance of the APACHE II score with that of a neural network in a medical-neurological ICU at a university hospital in Mumbai, India. A total of 2062 consecutive admissions between 1996 and 1998 were evaluated. Data from 2962 patients were used to train the neural network and data from the remaining 1000 patients were used to test the model and compare it with the APACHE II score. There were 337 deaths in these 1000 patients; APACHE II predicted 246 deaths whereas the neural network predicted 336 deaths. Calibration, as assessed using the Hosmer-Lemeshow statistic, was better with the neural network than with APACHE II score, and so was discrimination. As probable explanations for this apparent superiority of the ANN, the authors suggested differences in demographic characteristics and case-mix of patients in Indian ICUs. These specific features were certainly not accounted for in the original Western cohorts used to develop and validate the APACHE score.
In our research, both logistic regression and neural network models did a good job of predicting death. Although there was a statistically significant difference in discrimination as measured by ROC curve in favour of the neural network, the clinical meaning of this difference is not clear. A prediction model cannot be both perfectly reliable (i.e. calibrated) and perfectly discriminatory. According to Diamond [14], 'A model that maximizes discrimination does so at the expense of reliability ... On the other hand, a model that maximizes reliability does so at the expense of discrimination, and thereby trades categorical confidence for quantitative meaning.'
One of the advantages of neural network analysis is that there are few assumptions that must be verified before the models can be constructed; also, ANNs are able to model complex nonlinear relationships between independent and dependent variables, and so they allow the inclusion of a large number of variables. The comparison method is supposed to constrain the neural network analysis by limiting the number of potential predictor variables to the same set of predictor variables used in the logistic regression analysis. However, in this practical example, our network was able to use all of the 10 initial variables in its modelling, whereas logistic regression excluded four variables in the final model. Nevertheless, the predictive ability was almost the same with both approaches. A limitation of ANNs in the setting of aetiological research is that standardized coefficients and/or odds ratios corresponding to each variable cannot be calculated and presented as they can in regression models. This lack of interpretability at the level of individual predictors is one of the most criticized features of neural network models [15]. Furthermore, neural network models require sophisticated software, and the computer resources involved in training and testing neural networks can be substantial.
Our work has some limitations. First, the sample size – specifically the number of outcomes (101 deaths) – limit the number of potential predictor variables. As a rule of thumb, no more that 10 outcome events for each independent variable are permissible if over-fitting or under-fitting problems are to be avoided [16]. We tried to overcome this limitation by considering just those variables that are more likely to be related to mortality from a clinical point of view. However, as is usual in any observational study, residual confounding or unmeasured factors may compromise the scope or precision of the model. Second, external validity was tested neither for logistic regression nor for the ANN. Furthermore, the small sample size prevented a comprehensive split-sample validation with any strategy. Determination of the applicability and usefulness of any predictive model requires independent and external validation in a population that is intrinsically different from the development sample [17]. Therefore, both the proposed score and the neural network merit a new cohort study before any potential clinical use can be considered.
Conclusion
A predictive model would be an extremely useful tool in the setting of suspected sepsis in the ER. It could serve both as a guideline in medical decision-making regarding ICU admission or specific therapies, and as a simple way to select or stratify patients for clinical research. Our proposed model and the specific development method – either logistic regression or neural networks – must be evaluated and validated in an independent population. Further research is required to determine whether there are practical or clinical advantages to one approach over the other. As a general concept, we agree with Tu [15] that logistic regression remains the best choice when the primary goal of model development is to examine possible causal relationships among variables, but that some form of hybrid technique incorporating the best features of both approaches might lead to the development of optimal prediction models.
Key messages
- Simple clinical variables were useful in predicting death in patients with suspected sepsis in the ER.
- Logistic regression and ANNs were equivalent in terms of predictive ability.
- Discriminative ability, as measured using ROC curve analysis, was better with the ANN.
- Further research is required to validate the model and to determine whether there are practical or clinical advantages to one approach over the other.
Abbreviations
ANN = artificial neural network; APACHE = Acute Physiology and Chronic Health Evaluation; ER = emergency room; GCS = Glasgow Coma Scale; GSD = general systemic disease; ICU = intensive care unit; ISD = immunosuppressive systemic disease; ROC = receiver operating characteristic; SIRS = systemic inflammatory response syndrome.
Competing interests
The author(s) declare that they have no competing interests.
Authors' contributions
FJ conceived the study, participated in its design and coordination, performed the statistical analysis for logistic regression, and drafted the manuscript. CM participated in the design and coordination of the study, and contributed to the statistical analysis. JF and DA participated in the design of the study and performed the procedures for the neural network analysis. All authors read and approved the final manuscript.
Acknowledgements
We are indebted to the staff of emergency services at Hospital Universitario San Vicente de Paul and Hospital General de Medellín for their collaboration. We appreciated helpful suggestions from three anonymous referees. The research was partially supported by a grant 'Comité para el desarrollo de la Investigacion (CODI) – Universidad de Antioquia'.
Figures and Tables
Figure 1 Observed and predicted deaths with logistic regression and neural network in patients with suspected sepsis admitted to the emergency room. There were no patients with scores 11 or 12 in the cohort.
Table 1 Cutoff points on continuous variables for changes in the probability of death according to locally weighted regression
Variable Cutoff points
1 2 3
Respiratory rate (breaths/min) <24 24–33 ≥ 34
GCS score >12 8–12 <8
Temperature (°C) >38 36.6–38 ≤ 36.5
Shock Index <1 1–1.4 ≥ 1.5
GCS, Glasgow Coma Scale.
Table 2 Level of variables and relative weight according to their score
Variable Level of variable Score
GSD or ISDa Presence of GSD or ISD 2
Respiratory rate Rate >34 breaths/min 3
Respiratory rate Rate 24–33 breaths/min 1
GCS score Score <12 3
Temperature <38°C 2
Shock Index ≥ 1.5 2
aSee text for definitions of general systemic disease (GSD) and immunosuppressive systemic disease (ISD). GCS, Glasgow Coma Scale.
==== Refs
Martin GS Mannino DM Eaton S Moss M The epidemiology of sepsis in the United Sates from 1979 through 2000 N Engl J Med 2003 348 1546 1554 12700374 10.1056/NEJMoa022139
Moss M Martin GS A global perspective on the epidemiology of sepsis Intensive Care Med 2004 30 527 529 14985955 10.1007/s00134-004-2182-z
American College of Chest Physicians/Society of Critical Care Medicine Consensus Conference Definitions for sepsis and organ failure and guidelines for the use of innovative therapies in sepsis Crit Care Med 1992 20 864 874 1597042
Rangel-Frausto S Pittet D Costignan M Hwang T Davis CS Wenzel RP The natural history of the systemic inflammatory response syndrome JAMA 1995 273 117 123 7799491 10.1001/jama.273.2.117
Gabutti L Burnier M Mombelli G Male F Pellegrini L Marone C Usefulness of artificial neural networks to predict follow-up dietary protein intake in hemodialysis patients Kidney Int 2004 66 399 407 15200449 10.1111/j.1523-1755.2004.00744.x
Maiellaro PA Cozzolongo R Marino P Artificial neural networks for the prediction of response to interferon plus ribavirin treatment in patients with chronic hepatitis C Curr Pharm Des 2004 10 2101 2109 15279549
Lim WK Er MJ Classification of mammographic masses using generalized dynamic fuzzy neural networks Med Phys 2004 31 1288 1295 15191321 10.1118/1.1708643
Loukas C Brown P Online prediction of self-paced hand-movements from subthalamic activity using neural networks in Parkinson's disease J Neurosci Methods 2004 137 193 205 15262061 10.1016/j.jneumeth.2004.02.017
Jaimes F Garcés J Cuervo J Ramírez F Ramírez J Quintero C Vargas A Zapata L Ochoa J Yepes M Prognostic factors in systemic inflammatory response syndrome (SIRS). Development of a severity index [in Spanish] Acta Medica Colombiana 2001 26 149 157
Loader C Local Regression and Likelihood 1999 New York: Springer-Verlag
Hanley JA McNeil BJ The meaning and use of the area under a receiver operating charasteristic (ROC) curve Radiology 1982 143 29 36 7063747
Clermont G Angus D DiRusso S Griffin M Linde-Zwirble W Predicting hospital mortality for patients in the intensive care unit: a comparison of artificial neural networks with logistic regression models Crit Care Med 2001 29 291 296 11246308 10.1097/00003246-200102000-00012
Nimgaonkar A Karnad D Sudarshan S Ohno-Machado L Kohane I Prediction of mortality in an Indian intensive care unit. Comparison between APACHE II and artificial neural networks Intensive Care Med 2004 30 248 253 14727015 10.1007/s00134-003-2105-4
Diamond G What price perfection? Calibration and discrimination of clinical prediction models J Clin Epidemiol 1992 45 85 89 1738016 10.1016/0895-4356(92)90192-P
Tu J Advantages and disadvantages of using artificial neural networks versus logistic regression for predicting medical outcomes J Clin Epidemiol 1996 49 1225 1231 8892489 10.1016/S0895-4356(96)00002-9
Concato J Feinstein AR Holford TR The risk of determining risk with multivariable models Ann Intern Med 1993 118 201 210 8417638
Justice AC Covinsky KE Berlin JA Assessing the generalizability of prognostic information Ann Intern Med 1999 130 515 524 10075620
| 15774048 | PMC1175932 | CC BY | 2021-01-04 16:04:52 | no | Crit Care. 2005 Feb 17; 9(2):R150-R156 | utf-8 | Crit Care | 2,005 | 10.1186/cc3054 | oa_comm |
==== Front
Crit CareCritical Care1364-85351466-609XBioMed Central London cc30581577405010.1186/cc3058ResearchUneven distribution of ventilation in acute respiratory distress syndrome Rylander Christian [email protected]én Ulf 2Rossi-Norrlund Rauni 3Herrmann Peter 4Quintel Michael 5Bake Björn 61 Department of Anaesthesiology and Intensive Care, Sahlgrenska University Hospital, Göteborg, Sweden2 Professor, The Sahlgrenska Academy at Göteborg University, Department of Radiology, Sahlgrenska University Hospital, Göteborg, Sweden3 The Sahlgrenska Academy at Göteborg University, Department of Radiology, Sahlgrenska University Hospital, Göteborg, Sweden4 Engineer, Department of Anaesthesiology II – Intensive Care Medicine, Z.A.R.I., University Hospital Gottingen, Gottingen, Germany5 Professor, Department of Anaesthesiology II – Intensive Care Medicine, Z.A.R.I., University Hospital Gottingen, Gottingen, Germany6 Professor, The Sahlgrenska Academy at Göteborg University, Department of Pulmonary Medicine, Sahlgrenska University Hospital, Göteborg, Sweden2005 21 2 2005 9 2 R165 R171 5 8 2004 12 10 2004 20 12 2004 17 1 2005 Copyright © 2005 Rylander 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
The aim of this study was to assess the volume of gas being poorly ventilated or non-ventilated within the lungs of patients treated with mechanical ventilation and suffering from acute respiratory distress syndrome (ARDS).
Methods
A prospective, descriptive study was performed of 25 sedated and paralysed ARDS patients, mechanically ventilated with a positive end-expiratory pressure (PEEP) of 5 cmH2O in a multidisciplinary intensive care unit of a tertiary university hospital. The volume of poorly ventilated or non-ventilated gas was assumed to correspond to a difference between the ventilated gas volume, determined as the end-expiratory lung volume by rebreathing of sulphur hexafluoride (EELVSF6), and the total gas volume, calculated from computed tomography images in the end-expiratory position (EELVCT). The methods used were validated by similar measurements in 20 healthy subjects in whom no poorly ventilated or non-ventilated gas is expected to be found.
Results
EELVSF6 was 66% of EELVCT, corresponding to a mean difference of 0.71 litre. EELVSF6 and EELVCT were significantly correlated (r2 = 0.72; P < 0.001). In the healthy subjects, the two methods yielded almost identical results.
Conclusion
About one-third of the total pulmonary gas volume seems poorly ventilated or non-ventilated in sedated and paralysed ARDS patients when mechanically ventilated with a PEEP of 5 cmH2O. Uneven distribution of ventilation due to airway closure and/or obstruction is likely to be involved.
==== Body
Introduction
Decreased functional residual capacity (FRC) and increased pulmonary resistance are hallmarks of acute respiratory distress syndrome (ARDS) [1]. Pathophysiological mechanisms include alveolar flooding and/or collapse, which contribute to shunting of blood and to hypoxaemia [2]. Whether true alveolar collapse or intraluminar oedema with increased impedance dominates is a matter of debate [3]. Furthermore, the expiratory flow limitation observed in ARDS patients has been attributed to the closure of small airways [4]. Pulmonary gas distal to such an airway closure/obstruction may be poorly ventilated or non-ventilated. If so, it might not be included in FRC measurements based on tracer gas dilution. The end-expiratory lung volume determined by tracer gas dilution is termed 'ventilated gas volume' in this paper. Other techniques such as body plethysmography and radiographical methods [5] determine the total end-expiratory volume of pulmonary gas, irrespective of whether it is well ventilated, poorly ventilated or non-ventilated. This volume is termed 'total gas volume' in this report. A difference between the ventilated gas volume and the total gas volume can be interpreted as a volume of gas being poorly ventilated or non-ventilated. This difference is obvious in patients with chronic obstructive airway disease in whom FRC determined by gas dilution might be considerably lower than FRC determined by body plethysmography [6]. However, in mechanically ventilated ARDS patients the volume of poorly ventilated or non-ventilated gas seems not to have been studied in detail.
The aim of the present study was therefore to assess the volume of poorly ventilated or non-ventilated gas in mechanically ventilated ARDS patients, assuming the difference between the ventilated gas volume and the total gas volume to represent poorly ventilated or non-ventilated gas. To validate the methods involved, similar measurements were performed in young healthy subjects in whom no poorly ventilated or non-ventilated gas is expected to be found.
Materials and methods
Ethical approval
The study was approved by the local ethics committee and conducted in accordance with the Helsinki Declaration. Informed consent was obtained from the next-of-kin of the patients and directly from the healthy subjects.
Patients
Twenty-five sedated and mechanically ventilated patients were included from a mixed-adult intensive care unit. The criterion for selection was the eligible ARDS patient [7] having spent the longest time on mechanical ventilation at the time of the once-weekly available opportunity for computed tomography (CT). Patients were eligible for the study only if their arterial oxygenation was stable and between 10 and 26 kPa during mechanical ventilation with the following parameters: fraction of inspired oxygen 0.5; constant flow volume-controlled mode; tidal volume 8 to 10 ml/kg; positive end-expiratory pressure (PEEP) 5 cmH2O. Chronic obstructive pulmonary disease was not an exclusion criterion but was present only in one patient (no. 13). Clinical data are given in Table 1. Twenty healthy non-smoking students independent of the investigating institutions were enrolled and interviewed to rule out any history of tobacco use or obstructive lung disease. Anthropometric data for both groups are given in Table 2.
Measurements
The ventilated gas volume was determined in both groups by a gas dilution technique using rebreathing of sulphur hexafluoride. End-expiratory measurements in the ventilated patients were made at a PEEP of 5 cmH2O (EELVSF6) and measurements in the spontaneously breathing healthy subjects were made at the FRC level (FRCSF6). A prototype system (AMIS 2001; Innovision A/S, Odense, Denmark) equipped with a photoacoustic and magnetoacoustic multigas analyser [8] was used. The accuracy of the analyser was checked by comparison with mass spectrometry (AMIS 2000; Innovision A/S) before and after the series of experiments. Before each measurement, the ambient temperature and pressure were registered and correct readings from the gas analyser were verified by supplying room air and the undiluted tracer gas mixture to the gas inlet. The gas sampling rate was 120 ml/min. The rebreathing unit consisted of a bag-in-box system in which the flexible rubber bellows could be manually ventilated by a piston fitted through the distal short end of the cylinder. For operation, the unit was instantly switched into the patient circuit by a pneumatic slide valve without disconnection. The bellows was initially filled with 1.2 litres (ambient temperature and pressure, dry) of a gas mixture of 1.0% SF6 in 5.0% nitrous oxide (N2O) and oxygen (bal; medical grade). The presence of N2O was due to the circulatory monitoring function of the multimodal monitoring system. The SF6 concentration was continuously plotted during 30 s of ventilation at a frequency of 20 breaths per minute (Fig. 1). Allowing for the tubing dead space (101 ml in the subjects, 107 ml in the patients), the ventilated gas volume was calculated from a formula based on standard gas dilution principles for FRC measurements:
where Pb is the barometric pressure in torr, T is the ambient absolute temperature and SF6i and SF6e are the initial and equilibrated concentration of SF6 (standard temperature and pressure, dry), respectively, and 1.2 is the bellows volume. FRC symbolises both FRCSF6 in the young healthy subjects and EELVSF6 in the ventilated patients.
The total gas volume was calculated from CT images reconstructed from a scan lasting about 20 s in a high-speed scanner (GE High Speed CT/i; General Electric Medical Systems, Milwaukee, WI, USA). End-expiratory measurements in the patients were made in apnoea at PEEP 5 cmH2O (EELVCT) and measurements in the healthy subjects were made in apnoea at the FRC level (FRCCT). The following exposure parameters were used: 120 kV; 170 mA; rotation time 1.0 s; collimation 1 mm and a matrix of 512 × 512, yielding voxel volumes of 0.25 to 0.49 mm3 depending on the field of view. An initial topogram defined the limits of the lungs, and the first and last scanning levels were positioned at the apical and caudal extremes, respectively. In between, eight more scanning levels were evenly dispersed, making a total of 10 consecutive single exposures with a distance between the scans of 18 to 25 mm, depending on thoracic dimensions. The total effective radiation dose was estimated to equal one standard chest X-ray examination, yielding an average absorbed radiation of 0.8 mGy to the breasts of female subjects. Within each image, the lungs were manually delineated from the thoracic wall in a single region of interest. Within the region of interest, the voxels with attenuation values between – 1,000 and 0 Hounsfield units (HU) were automatically selected for analysis by software (MALUNA 2.02; Peter Herrmann, Mannheim, Germany) on a personal computer, and their gas volume (V) was calculated from the formula [9]
where Vvox is the single-voxel volume of n voxels within the slice. The total gas volume was calculated by interpolating for the volume of gas in the lung tissue between the 10 scan levels by the method of Kvist [10] with the modified formula
where V1 and V2 are the gas volumes of two adjacent slices with the thickness t, separated by the centre distance d. FRC symbolises both FRCCT in the young healthy subjects and EELVCT in the ventilated patients.
During the measurements, the sedated patients were temporarily paralysed and ventilated by means of a mobile ventilator (Servo 900 C; Siemens, Solna, Sweden) with the settings described above. The end-expiratory position was achieved by activation of the expiratory hold function on the ventilator. The patient was then either ventilated from the rebreathing circuit or CT scanned in maintained apnoea. The rebreathing procedure was performed in duplicate before and after a single CT exposure.
Before the supine measurements, the nose-clipped, supine and relaxed healthy subjects breathed room air through a mouthpiece connected to the rebreathing system through a three-way valve. At the FRC level, the valve was either switched into the rebreathing system for gas dilution by spontaneous breathing or was closed during the CT examination. The rebreathing procedure was performed in duplicate before and after a single CT exposure.
Statistical analysis
Data are presented as means ± standard deviation if not specified otherwise. The level of significance was defined as P < 0.05. The coefficient of variation (CV) for paired measurements was calculated as the standard deviation of the differences divided by the mean of all measurements [11]. Calculations were performed with the software package Statistica 6.0 (StatSoft Inc., Tulsa, OK, USA) on a personal computer.
Results
In the ARDS patients, EELVSF6 was 66 ± 14% of EELVCT. EELVSF6 was found systematically lower than EELVCT except in one patient (no. 19), in whom they were similar. The mean difference, corresponding to the poorly ventilated or non-ventilated gas volume, was 0.71 ± 0.47 litre. The magnitude of the poorly ventilated or non-ventilated gas volume was not correlated with age or ventilator days. Mean results are given in Table 3. EELVSF6 and EELVCT were significantly correlated (r = 0.85; P < 0.001) (Fig. 2). The CV of duplicate EELVSF6 measurements was 5.6%.
In the supine healthy subjects, FRCSF6 was 99 ± 9% of FRCCT, and they were closely correlated (r = 0.91; P < 0.001) (Fig. 3). The differences did not depend on the magnitude of FRC (Fig. 4). The CV of duplicate FRCSF6 measurements was 3.1%.
Discussion
This study shows that there is a considerable volume of poorly ventilated or non-ventilated gas present in the lungs of sedated and paralysed ARDS patients when mechanically ventilated with a PEEP of 5 cmH2O.
We assumed that the difference between the ventilated gas volume determined by gas dilution and the total gas volume calculated from CT corresponds to a poorly ventilated or non-ventilated gas volume. The methods used to determine these volumes were validated by comparison of similar measurements in young healthy subjects, in whom they should yield similar results because these lungs are homogeneously ventilated with no obstruction and no airway closure. Indeed, almost identical results were obtained in the young healthy subjects. Furthermore, the CV of duplicate measurements in the healthy subjects indicated a good repeatability. The FRCSF6 values might seem somewhat low compared with predicted FRC values based on a mixed adult population (Table 3), but normal FRC values in supine young subjects are rare and the predictions therefore remain uncertain. The CT interpolation technique has been validated previously for heterogeneously scattered tissue [12] and should be precise enough with 10 scans evenly distributed over the lungs. In summary, we consider that the two methods used were adequate and that the difference between their results in the ARDS patients can be assumed to correspond to a poorly ventilated or non-ventilated gas volume.
The most likely pathophysiological mechanism associated with this volume is airway closure and/or obstruction. Further contribution from atelectasis formation during the inspiratory hold is unlikely with the fraction of inspiratory oxygen used [13]. However, the deep sedation and paralysis of the ARDS patients might have contributed to poor ventilation in the dependent parts of the lungs [14]. The lung injury is unevenly distributed in ARDS [15], which causes an uneven distribution of ventilation including overdistension of non-dependent regions. By definition, open but non-compliant lung units are poorly ventilated or non-ventilated, but this seems unlikely to be of any importance in ARDS patients during ventilation with a PEEP of 5 cmH2O.
The PEEP level applied in the present study was chosen to be clinically relevant [16] but it does not effectively counteract expiratory derecruitment of lung units. In a study of 10 ARDS patients, mechanically ventilated with zero end-expiratory pressure (ZEEP), Koutsoukou and colleagues determined an intrinsic PEEP of 4.1 ± 2.4 cmH2O, and expiratory flow limitation was demonstrated in eight of them [4]. These results suggest the presence of airway closure and/or obstruction at the FRC level in ARDS. In contrast, when closed circuit helium rebreathing and CT were recently compared in a group of 21 ARDS patients, mechanically ventilated with a PEEP of 12 ± 5 cmH2O, similar EELVs were found [17]. This finding indicates that there is no airway closure and/or obstruction when a PEEP of 12 cmH2O is applied. Indeed, it was recently also shown that the intrinsic PEEP and the expiratory flow limitation present at ZEEP can be eliminated by a PEEP of 10 cmH2O [18]. In summary, those studies and the present results indicate that airway closure and/or obstruction occurs at low levels of PEEP or ZEEP and that the distal gas volume is recruitable for more effective ventilation by a moderate increase in PEEP. Accordingly, increasing PEEP from 0 to 15 cmH2O has been shown in a study of pulmonary mechanics to increase pulmonary compliance in some patients, which was associated with the recruitment of lung units with preserved normal compliance [19]. Furthermore, low compliance during the initial phase of inspiration has been attributed to non-collapsed but slowly ventilated lung units, in which the ventilation can be increased by increased PEEP [20]. The gas content of such non-collapsed but poorly ventilated lung units may correspond to the volume of poorly ventilated or non-ventilated gas demonstrated in the present study.
Substantially elevated pressure in the airways is associated with signs of parenchymal overdistension [21]. CT studies have shown that this effect is located to non-dependent well-aerated lung units that become overdistended by the airway pressure required to inflate compressed dependent lung units [22]. Overdistension associated with increased airway pressure seems to be less pronounced when the parenchyma is diffusely affected without regional atelectasis [23], as in our patients. Possibly, the poorly ventilated or non-ventilated gas volume in this type of diffuse ARDS might reflect gas contained in lung units distal to airway closure and/or obstruction. The recruitment of such gas-containing lung units, excluded from effective ventilation by partial compression or oedema, can be expected to require a smaller elevation of transmural pressure than that needed to inflate completely collapsed lung units. If the volume of poorly ventilated or non-ventilated gas is small or non-existent, a moderately raised airway pressure might be ineffective for recruitment and merely contribute to the risk of overdistension.
Conclusion
We conclude that about one-third of the total gas volume is poorly ventilated or non-ventilated in the lungs of sedated and paralysed ARDS patients when mechanically ventilated with a PEEP of 5 cmH2O. This indicates uneven distribution of ventilation due to the presence of small-airway closure and/or obstruction at this PEEP level. Such a poorly ventilated or non-ventilated gas volume might be recruited for more effective ventilation by an increase in airway pressure that is less than the inflation pressure of completely collapsed lung units.
Key messages
• This study demonstrates uneven distribution of ventilation in 25 sedated and ventilated ARDS patients by comparing the total end-expiratory gas end volume calculated from computed tomography and the ventilated gas volume measured by inert gas rebreathing.
• The poorly ventilated or non-ventilated volume distal to the possible airway closure and/or obstruction might be recruited for more effective ventilation by an increase in airway pressure that is less than the inflation pressure of completely collapsed lung units.
Abbreviations
ARDS = acute respiratory distress syndrome; CT = computed tomography; CV = coefficient of variation; EELVCT = total gas volume calculated from computed tomography images in the end-expiratory position; EELVSF6 = end-expiratory lung volume measured by rebreathing of sulphur hexafluoride; FRC = functional residual capacity; FRCCT = FRC calculated from computed tomography scans; FRCSF6 = FRC measured by sulfur hexafluoride rebreathing; HU = Hounsfield unit; PEEP = positive end-expiratory pressure; ZEEP = zero end-expiratory pressure.
Competing interests
The author(s) declare that they have no competing interests.
Authors' contributions
CR, UT and BB conceived the study and designed the protocol. UT, MQ and PH defined the radiographical image analysis. CR and RRN performed measurements. CR, UT and BB wrote and revised the manuscript, which was reviewed and approved by all authors before final submission.
Acknowledgements
The results of this study were in part presented at the ESICM meeting in Rome in 1999. The study was supported by departmental funding and by grants from the Gothenburg Medical Association. The inert gas system (AMIS 2001) with consumables was made available by Innovision A/S, Odense, Denmark.
Figures and Tables
Figure 1 Concentration of the tracer gas sulphur hexafluoride (SF6) plotted during 30 s of rebreathing in a supine healthy subject.
Figure 2 Linear regression between EELV measurements by rebreathing of sulphur hexafluoride (EELVSF6) and by computed tomography (EELVCT) obtained in 25 ARDS patients. The dotted line is the regression line EELVSF6 = 0.4EELVCT + 0.3 (r2 = 0.72; P < 0.001).
Figure 3 Linear regression between FRC measurements by rebreathing of sulphur hexafluoride (FRCSF6) and by computed tomography (FRCCT) in 20 healthy subjects. The dotted line is the regression line: EELVSF6 = 0.9FRCCT + 0.1 (r2 = 0.83; P < 0.001).
Figure 4 Bland-Altman plot [24] of supine functional residual capacity measured by rebreathing of sulphur hexafluoride (FRCSF6) and by computed tomography (FRCCT) in 20 healthy subjects. The individual differences of paired measurements (y axis) did not depend on the magnitude of their average values (x axis). The mean difference (solid line; dotted lines represent the mean ± 2SD) was small.
Table 1 Clinical data of the ARDS patients
Patient Age (years) Diagnosis Ventilator days Discharge status
1 52 Bilateral pneumonia 4 S
2 44 Sepsis with MOF 63 S
3 79 Postoperative MOF 10 NS
4 37 Bilateral pneumonia 40 S
5 85 Postoperative MOF 3 NS
6 18 Postoperative ARDS 2 S
7 54 Bilateral pneumonia 5 S
8 79 Postoperative ARDS 2 S
9 62 Bilateral pneumonia 5 NS
10 60 Aspiration 3 NS
11 46 Postoperative ARDS 3 S
12 44 Bilateral pneumonia 6 S
13 76 Bilateral pneumonia 13 S
14 22 Multiple trauma 2 S
15 78 Bilateral pneumonia 4 NS
16 58 Postoperative sepsis 3 NS
17 29 Multiple trauma 10 S
18 63 Pneumonia 4 S
19 53 Postoperative sepsis 26 S
20 71 Postoperative ARDS 2 S
21 31 Multiple trauma 3 S
22 22 Multiple trauma 5 S
23 70 Bilateral pneumonia 12 NS
24 82 Bilateral pneumonia 13 NS
25 20 Bilateral pneumonia 7 S
Mean 53 - 10 8/25 NS
Ventilator days were calculated on the day of study. ARDS, acute respiratory distress syndrome; MOF, multi-organ failure; NS, non-survivor; S, survivor.
Table 2 Anthropometric data
Group n Age (years) Sex (M/F) Height (cm) BMI (kg/m2)
ARDS patients 25 53 (18–85) 13/12 174 (165–195) 25 (17–30)
Healthy subjects 20 24 (19–28) 8/12 173 (161–192) 22 (18–25)
Data are given as mean and range except for number and gender. ARDS, acute respiratory distress syndrome; BMI, body mass index.
Table 3 Lung volumes
Group Supine EELV or FRC (litres)
SF6 CT
ARDS patients 1.2 ± 0.4 1.9 ± 0.8
Healthy subjects 1.7 ± 0.3 1.8 ± 0.3
(78% of predicted) (80% of predicted)
End-expiratory lung volume (EELV) in the acute respiratory distress syndrome (ARDS) patients and functional residual capacity (FRC) in the healthy subjects were measured by rebreathing of sulphur hexafluoride (SF6) and computed tomography (CT), respectively. Predicted normal FRC values are from [25].
==== Refs
Ramachandran PR Fairley HB Changes in functional residual capacity during respiratory failure Can Anaesth Soc J 1970 17 359 369 5429679
Ware LB Matthay MA The acute respiratory distress syndrome N Engl J Med 2000 342 1334 1349 10793167 10.1056/NEJM200005043421806
Hubmayr RD Perspective on lung injury and recruitment: a skeptical look at the opening and collapse story Am J Respir Crit Care Med 2002 165 1647 1653 12070067 10.1164/rccm.2001080-01CP
Koutsoukou A Armaganidis A Stavrakaki-Kallergi C Vassilakopoulos T Lymberis A Roussos C Milic-Emili J Expiratory flow limitation and intrinsic positive end-expiratory pressure at zero positive end-expiratory pressure in patients with adult respiratory distress syndrome Am J Respir Crit Care Med 2000 161 1590 1596 10806160
Kendrick AH Comparison of methods of measuring static lung volumes Monaldi Arch Chest Dis 1996 51 431 439 9009635
Rodenstein DO Stanescu DC Reassessment of lung volume measurement by helium dilution and by body plethysmography in chronic air-flow obstruction Am Rev Respir Dis 1982 126 1040 1044 7181224
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. Definitions, mechanisms, relevant outcomes, and clinical trial coordination Am J Respir Crit Care Med 1994 149 818 824 7509706
Clemensen P Christensen P Norsk P Gronlund J A modified photo- and magnetoacoustic multigas analyzer applied in gas exchange measurements J Appl Physiol 1994 76 2832 2839 7928918
Gattinoni L Pesenti A Avalli L Rossi F Bombino M Pressure-volume curve of total respiratory system in acute respiratory failure. Computed tomographic scan study Am Rev Respir Dis 1987 136 730 736 3307572
Kvist H Sjostrom L Tylen U Adipose tissue volume determinations in women by computed tomography: technical considerations Int J Obes 1986 10 53 67 3710689
Hankinson JL Stocks J Peslin R Reproducibility of lung volume measurements Eur Respir J 1998 11 787 790 9596139
Kvist H Adipose tissue volume determinations by computed tomography Dissertation 1988 Göteborg University, Sweden
Edmark L Kostova-Aherdan K Enlund M Hedenstierna G Optimal oxygen concentration during induction of general anesthesia Anesthesiology 2003 98 28 33 12502975 10.1097/00000542-200301000-00008
Frerichs I Hahn G Golisch W Kurpitz M Burchardi H Hellige G Monitoring perioperative changes in distribution of pulmonary ventilation by functional electrical impedance tomography Acta Anaesthesiol Scand 1998 42 721 726 9689281
Kallet RH Katz JA Respiratory system mechanics in acute respiratory distress syndrome Respir Care Clin N Am 2003 9 297 319 14690068
Esteban A Anzueto A Alia I Gordo F Apezteguia C Palizas F Cide D Goldwaser R Soto L Bugedo G How is mechanical ventilation employed in the intensive care unit? An international utilization review Am J Respir Crit Care Med 2000 161 1450 1458 10806138
Patroniti N Bellani G Manfio A Maggioni E Giuffrida A Foti G Pesenti A Lung volume in mechanically ventilated patients: measurement by simplified helium dilution compared to quantitative CT scan Intensive Care Med 2004 30 282 289 14714108 10.1007/s00134-003-2109-0
Koutsoukou A Bekos B Sotiropoulou C Koulouris NG Roussos C Milic-Emili J Effects of positive end-expiratory pressure on gas exchange and expiratory flow limitation in adult respiratory distress syndrome Crit Care Med 2002 30 1941 1949 12352025 10.1097/00003246-200209000-00001
Ranieri VM Eissa NT Corbeil C Chasse M Braidy J Matar N Milic-Emili J Effects of positive end-expiratory pressure on alveolar recruitment and gas exchange in patients with the adult respiratory distress syndrome Am Rev Respir Dis 1991 144 544 551 1892293
Vieillard-Baron A Prin S Chergui K Page B Beauchet A Jardin F Early patterns of static pressure–volume loops in ARDS and their relations with PEEP-induced recruitment Intensive Care Med 2003 29 1929 1935 12923622 10.1007/s00134-003-1938-1
Vieira SR Puybasset L Richecoeur J Lu Q Cluzel P Gusman PB Coriat P Rouby JJ A lung computed tomographic assessment of positive end-expiratory pressure-induced lung overdistension Am J Respir Crit Care Med 1998 158 1571 1577 9817710
Rouby JJ Puybasset L Nieszkowska A Lu Q Acute respiratory distress syndrome: lessons from computed tomography of the whole lung Crit Care Med 2003 31 4 Suppl S285 295 12682454 10.1097/01.CCM.0000057905.74813.BC
Puybasset L Gusman P Muller JC Cluzel P Coriat P Rouby JJ Regional distribution of gas and tissue in acute respiratory distress syndrome. III. Consequences for the effects of positive end-expiratory pressure. CT Scan ARDS Study Group. Adult Respiratory Distress Syndrome Intensive Care Med 2000 26 1215 1227 11089745 10.1007/s001340051340
Bland JM Altman DG Statistical methods for assessing agreement between two methods of clinical measurement Lancet 1986 1 307 310 2868172
Ibanez J Raurich JM Normal values of functional residual capacity in the sitting and supine positions Intensive Care Med 1982 8 173 177 7119270
| 15774050 | PMC1175934 | CC BY | 2021-01-04 16:04:52 | no | Crit Care. 2005 Feb 21; 9(2):R165-R171 | utf-8 | Crit Care | 2,005 | 10.1186/cc3058 | oa_comm |
==== Front
Genome BiolGenome Biology1465-69061465-6914BioMed Central London gb-2005-6-5-r401589286810.1186/gb-2005-6-5-r40ResearchConsolidating the set of known human protein-protein interactions in preparation for large-scale mapping of the human interactome Ramani Arun K [email protected] Razvan C [email protected] Raymond J [email protected] Edward M [email protected] Center for Systems and Synthetic Biology and Institute for Cellular and Molecular Biology, University of Texas, Austin, TX 78712, USA2 Department of Computer Sciences, University of Texas, Austin, TX 78712, USA3 Department of Chemistry and Biochemistry, University of Texas, Austin, TX 78712, USA2005 15 4 2005 6 5 R40 R40 20 12 2004 9 2 2005 11 3 2005 Copyright © 2005 Marcotte et al.; licensee BioMed Central Ltd.This is an Open Access article distributed under the terms of the Creative Commons Attribution License (), which permits unrestricted use, distribution, and reproduction in any medium, provided the original work is properly cited.
In order to consolidate the known human proteins interactions two tests were developed to measure the relative accuracy of the available interaction data. In addition, 6,580 interactions among 3,737 human proteins were recovered from Medline abstracts and combined with existing interaction data to obtain a network of 31,609 interactions among 7,748 human proteins, accurate to the same degree as the existing data sets.
Background
Extensive protein interaction maps are being constructed for yeast, worm, and fly to ask how the proteins organize into pathways and systems, but no such genome-wide interaction map yet exists for the set of human proteins. To prepare for studies in humans, we wished to establish tests for the accuracy of future interaction assays and to consolidate the known interactions among human proteins.
Results
We established two tests of the accuracy of human protein interaction datasets and measured the relative accuracy of the available data. We then developed and applied natural language processing and literature-mining algorithms to recover from Medline abstracts 6,580 interactions among 3,737 human proteins. A three-part algorithm was used: first, human protein names were identified in Medline abstracts using a discriminator based on conditional random fields, then interactions were identified by the co-occurrence of protein names across the set of Medline abstracts, filtering the interactions with a Bayesian classifier to enrich for legitimate physical interactions. These mined interactions were combined with existing interaction data to obtain a network of 31,609 interactions among 7,748 human proteins, accurate to the same degree as the existing datasets.
Conclusion
These interactions and the accuracy benchmarks will aid interpretation of current functional genomics data and provide a basis for determining the quality of future large-scale human protein interaction assays. Projecting from the approximately 15 interactions per protein in the best-sampled interaction set to the estimated 25,000 human genes implies more than 375,000 interactions in the complete human protein interaction network. This set therefore represents no more than 10% of the complete network.
==== Body
Background
The past few years have seen a tremendous development of functional genomics technologies. In particular, the yeast proteome has been the subject of considerable effort, including genome-wide protein interaction assays using yeast two-hybrid technology [1,2], affinity chromatography/mass spectrometry [3,4], synthetic lethal assays [5,6], and genome context methods [7-10]. Success in these areas, even given the limited accuracy of these technologies [11-15], has led to the application of the yeast two-hybrid method for the fly [16] and the worm proteomes [17], providing initial steps toward maps of the fly and worm interactomes.
Only minimal progress has been made with respect to the human proteome. The existing protein interaction data are largely composed of small-scale experiments collected in the BIND [18] and DIP [19] databases, as well as a set of approximately 12,000 interactions recovered by manual curation from Medline articles [20] and interactions transferred from other organisms on the basis of orthology [21]. The Reactome database [22] has around 11,000 interactions [23] that have been manually entered from articles focusing on core cellular pathways. Large-scale interaction assays among human proteins have yet to be performed, although a medium-scale map was created for the purified TNFα/NFκB protein complex [24] and the proteins involved in the human Smad signaling pathway [25]. This situation is in stark contrast to the abundant data available for yeast and calls for the application of high-throughput interaction assays for mapping the human protein interaction network.
One lesson from the yeast interactome research is clear: it is critical that such upcoming interaction assays be accompanied by measured error rates, without which the utility and interpretability of the data is jeopardized. To establish a basis for future interaction mapping we sought to consolidate existing human protein interaction data and to establish quantitative tests of data accuracy. We also sought to use data-mining approaches to extract additional known interactions from Medline abstracts to add to the existing interactions.
Most of the current biological knowledge can be retrieved from the Medline database, which now has records from more than 4,800 journals accounting for around 15 million articles. These citations contain thousands of experimentally recorded protein interactions. However, retrieving these data manually is made difficult by the large number of articles, all lacking formal structure. Automated extraction of information would be preferable, and therefore, mining data from Medline abstracts is a growing field [26-29].
In this paper, we present two quantitative tests (benchmarks) of the accuracy of large-scale human protein interaction assays, test the existing sets of interaction data for their relative accuracy, then apply these benchmarks in order to recover protein interactions from the approximately 750,000 Medline abstracts that concern human biology, resulting in a set of 6,580 interactions between 3,737 proteins of accuracy comparable to manual extraction. Combination of the interaction data creates a consolidated set of 31,609 interactions between 7,748 human proteins. On the basis of this initial set of interactions, we estimate the scale of the human interactome.
Results
Assembling existing public protein interaction data
We first gathered the existing human protein interaction datasets (summarized in Table 1), representing the current status of the human interactome. This required unification of the interactions under a shared naming and annotation convention. For this purpose, we mapped each interacting protein to LocusLink (now EntrezGene) identification numbers and retained only unique interactions (that is, for two proteins A and B, we retain only A-B or B-A, not both. We have chosen to omit self-interactions, A-A or B-B, for technical reasons, as their quality cannot be assessed on the functional benchmark we develop). In most cases, a small loss of proteins occurs in the conversion between the different gene identifiers (for example, converting from the NCBI 'gi' codes in BIND to LocusLink identifiers). In the case of the Human Protein Reference Database (HPRD), this processing resulted in a significant reduction in the number of interactions from 12,013 total interactions to 6,054 unique, non-self interactions, largely due to the fact that HPRD often records both A-B and B-A interactions, as well as a large number of self interactions, and indexes genes by their common names rather than conventional database entries, often resulting in multiple entries for different synonyms.
Although the interactions from these datasets are in principle derived from the same source (Medline), the sets are quite disjoint (Figure 1), implying either that the sets are biased for different classes of interactions, or that the actual number of interactions in Medline is quite large. We suspect both reasons. It is clear that each dataset has a different explicit focus (Reactome towards core cellular machinery, HPRD towards disease-linked genes, and BIND more randomly distributed). Due to these biases, it is likely that many interactions from Medline are still excluded from these datasets. The maximal overlap between interaction datasets is seen for BIND: 25% of these interactions are also in HPRD or Reactome; only 1% of Reactome interactions are in HPRD or BIND. An additional 9,283 (or around 60,000 at lower confidence) interactions are available from orthologous transfer of interactions from large-scale screens in other organisms (orthology-core and orthology-all) [21].
Benchmarking of protein interaction data
To measure the relative accuracy of each protein interaction dataset, we established two benchmarks of interaction accuracy, one based on shared protein function and the other based on previously known interactions. First, we constructed a benchmark in which we tested the extent to which interaction partners in a dataset shared annotation, a measure previously shown to correlate with the accuracy of functional genomics datasets [13,14,21]. We used the functional annotations listed in the Kyoto Encyclopedia of Genes and Genomes (KEGG) [30] and Gene Ontology (GO) [31] annotation databases. These databases provide specific pathway and biological process annotations for approximately 7,500 human genes, assigning human genes into 155 KEGG pathways (at the lowest level of KEGG) and 1,356 GO pathways (at level 8 of the GO biological process annotation). KEGG and GO annotations were combined into a single composite functional annotation set, which was then split into independent testing and training sets by randomly assigning annotated genes into the two categories (3,792 and 3,809 annotated genes respectively). For the second benchmark based on known physical interactions, we assembled the human protein interactions from Reactome and BIND, a set of 11,425 interactions between 1,710 proteins. Each benchmark therefore consists of a set of binary relations between proteins, either based on proteins sharing annotation or physically interacting. Generally speaking, we expect more accurate protein interaction datasets to be more enriched in these protein pairs. More specifically, we expect true physical interactions to score highly on both tests, while non-physical or indirect associations, such as genetic associations, should score highly on the functional, but not the physical interaction, test.
For both benchmarks, the scoring scheme for measuring interaction set accuracy is in the form of a log odds ratio of gene pairs either sharing annotations or physically interacting. To evaluate a dataset, we calculate a log likelihood ratio (LLR) as:
where P(D|I) and P(D|~I) are the probability of observing the data (D) conditioned on the genes sharing benchmark associations (I) and not sharing benchmark associations (~I). By Bayes theorem, this equation can be rewritten as:
where P(I|D) and P(~I|D) are the frequencies of interactions observed in the given dataset (D) between annotated genes sharing benchmark associations (I) and not sharing associations (~I), respectively, while P(I) and P(~I) represent the prior expectations (the total frequencies of all benchmark genes sharing the same associations and not sharing associations, respectively). This latter version of the equation is simpler to compute. A score of zero indicates interaction partners in the dataset being tested are no more likely than random to belong to the same pathway or to interact; higher scores indicate a more accurate dataset.
Among the literature-derived interactions (Reactome, BIND, HPRD), a total of 17,098 unique interactions occur in the public datasets. Testing the existing protein interaction data on the function benchmark reveals that Reactome has the highest accuracy (LLR = 3.8), followed by BIND (LLR = 2.9), HPRD (LLR = 2.1), core orthology-inferred interactions (LLR = 2.1) and the non-core orthology-inferred interaction (LLR = 1.1). The two most accurate datasets, Reactome and BIND, form the basis of the protein interaction-based benchmark. Testing the remaining datasets on this benchmark (that is, for their consistency with these accurate protein interaction datasets) reveals a similar ranking in the remaining data. Core orthology-inferred interactions are the most accurate (LLR = 5.0), followed by HPRD (LLR = 3.7) and non-core orthology inferred interactions (LLR = 3.7).
Recognizing protein names with a conditional random field (CRF) algorithm
To expand the list of human interactions, we turned to literature mining. We adopted the strategy of separately identifying the protein names in the abstracts and then matching up the interacting protein partners. This process was made difficult by the fact that unlike other organisms, such as yeast or Escherichia coli, the human genes have no standardized naming convention, and thus present one of the hardest sets of gene/protein names to extract. For example, human proteins may be named with typical English words, such as 'light', 'map', 'complement', and 'Sonic Hedgehog'. Names may be alphanumeric, may include Greek or Roman letters, may be case sensitive, and may be composed of multiple words. Names are frequently sub-strings of each other, such as 'epidermal growth factor' and 'epidermal growth factor receptor', which refer to two distinct proteins. It is therefore necessary that an information-extraction algorithm be specifically trained to extract gene and protein names accurately.
We developed an algorithm capable of distinguishing human protein names from similar words on the basis of their context in the sentence. Building on our previous work in this area [32], we developed a classification algorithm that accurately recognized human protein names in Medline abstracts. The performance of the protein name 'tagger' on a set of human-labeled test abstracts is plotted in Figure 2. The accuracy of the algorithm was measured as its precision (the fraction of correct protein names identified among all identified names) and its recall (the fraction of correctly identified protein names among all possible correct protein names) on a set of 200 publicly available hand-tagged abstracts [33] as well as on 750 Medline abstracts with hand-labeled human protein names (comparable results; data not shown). The algorithm, termed the CRF algorithm due to its use of conditional random fields, significantly out-performs the picking of exact protein names from a dictionary ('dictionary only') by taking into account the words' parts of speech and the context in which they appear. The CRF algorithm also outperforms the other name recognition algorithms available in the public domain [32,34,35]. To prepare for extracting protein interactions, the names of human proteins were identified using the CRF algorithm in the complete set of 753,459 Medline abstracts citing the word 'human'.
Extracting functional interactions via co-citation analysis
In order to establish which interactions occurred between the proteins identified in the Medline abstracts, we used a two-step strategy: measure co-citation of protein names, then enrich these pairs for physical interactions using a Bayesian filter. First, we counted the number of abstracts citing a pair of proteins, and then calculated the probability of co-citation under a random model. Figure 3a shows the performance of the co-citation algorithm, plotting the probability of being co-cited by random chance against the accuracy, calculated as a log likelihood score based on the functional annotation training benchmark. Empirically, we find the co-citation probability has a hyperbolic relationship with the accuracy on this benchmark, with protein pairs co-cited with low random probability scoring high on the benchmark.
The co-citation algorithm is remarkably robust to variations in the minimal accuracy with which the protein names are identified by the CRF algorithm (Figure 3b). This robustness is presumably due to the fact that co-citation requires proteins to be named repeatedly across many abstracts, thereby tolerating occasional errors in the name extraction process. With a threshold on the estimated extraction probability of 80% (as computed by the CRF model) in the protein name identification, around 15,000 interactions are extracted with the co-citation approach that score comparably or better on the independent functional annotation test benchmark than the manually extracted interactions from HPRD, which serves to establish a minimal threshold for our mined interactions.
However, it is clear that proteins are co-cited for many reasons other than physical interactions. We therefore tried to enrich specifically for physical interactions by applying a secondary filter: We applied a Bayesian classifier to measure the likelihood of the abstracts citing the protein pairs to discuss physical protein-protein interactions. The classifier [36] scores each of the co-citing abstracts according to the usage frequency of words relevant to physical protein interactions. Interactions extracted by co-citation and filtered using the Bayesian estimator compare favorably with the other interaction datasets on the functional annotation test benchmark (Figure 4a). Testing the accuracy of these extracted protein pairs on the physical interaction benchmark (Figure 4b) reveals that the co-cited proteins scored high by this classifier are indeed strongly enriched for physical interactions.
Taking as a minimally acceptable level of accuracy the interactions hand-entered from Medline (HPRD), our co-citation/Bayesian classifier analysis yields 6,580 interactions between 3,737 proteins. By combining these interactions with the 26,280 interactions from other sources, we obtained a final set of 31,609 interactions between 7,748 human proteins. In this, we have chosen not to include the complete set of orthology-derived interactions due to their lower performance on the annotation benchmark, although these will ultimately be quite useful when supported by future data. Table 2 shows the contributions from each of the datasets at this threshold and a comparison of the overlap of interactions in each of the datasets is depicted as a Venn diagram in Figure 5. The Venn diagram indicates small overlap among the various datasets, with less than 0.2% of the interactions represented in all datasets. Nonetheless, this network of interactions represents the current state of the human interactome at a reasonable level of accuracy.
The ID-Serve database of annotation and interactions
We have incorporated the results of this analysis into a web-based server [37], which can be queried for interactions of specific proteins. Genes are cross-listed under a variety of naming conventions, including LocusLink/EntrezGene, RefSeq, and Swiss-Prot, and are accompanied by links to other databases and GO and KEGG functional annotations. Protein interactions derived from the co-citation/Bayesian analysis are hyperlinked to the co-citing Medline abstracts, where they can be directly manually verified.
Discussion
Features of the network
In order to study the features of the network, we visualized the complete network of protein interactions in Figure 6. On superimposing a histogram of the density of interactions on the plot, we see that there is considerable clustering of proteins in the network, represented as peaks in the histogram. A closer look reveals that these regions correspond to proteins involved with the ribosome, spliceosome, proteasome, replication, transcription and the immune components.
A quantitative analysis of the network clustering and connectivity distribution (reviewed in Barabasi and Oltvai [38]) is presented in Table 2. The clustering coefficient (<C>) captures the modularity of the network. A comparison of our final network (<C> = 0.24) with 10 randomly generated networks with the same number of interactions and proteins (<C> = 9 × 10-3 ± 3 × 10-5) shows the clustering in the human protein interaction network is considerably above that expected at random, in spite of the incompleteness of the network. The 'degree' of the network is defined as the average number of links per protein and captures the connectivity of the network. Except for Reactome, each of the datasets indicated in Table 2 show low connectivity. The combined network is intermediate in both connectivity and modularity. Projecting from the approximately 15 interactions per protein in the best sampled interaction dataset (Reactome) to the 25,000 or so estimated in the human genome [39] implies more than 375,000 interactions in the complete human protein interaction network. Note that any overestimates in the average number of interactions per protein will be counterbalanced by the effect of alternative splicing in increasing the number of actual proteins, making this estimate at least a reasonable ballpark estimate. The current set of interactions therefore represents no more than 10% of the complete network.
Advantages of the log likelihood benchmarks
A good accuracy measure is of tremendous importance, impacting on the reliability of all downstream analysis. The log likelihood analysis eases comparison and assessment of diverse datasets. The score indicates the probability that the identified interactions are correct based on enrichment of positive interactions over background expectations. Note that this approach is distinct from simply measuring the intersection with the benchmark associations - because enrichment of positive to negative associations is measured, rather than just recovery of positive associations, even datasets with small intersections to the benchmark set can be evaluated for accuracy. Note also that the benchmarks themselves are not likely to be 100% correct - protein annotations are subjectively assigned, many proteins belong to multiple pathways, and even hand-curated protein interaction data can be mis-entered. Nonetheless, the log likelihood framework is tolerant of errors and merely requires that the benchmark data are generally correct among true interaction partners. Figure 4a shows the accuracy of each of the datasets. While the existing datasets have a single accuracy value, the mined interactions can be adjusted for accuracy based on the CRF threshold and the co-citation probabilities. New datasets can be incorporated using the log likelihood scoring scheme, and the ultimate strength of these benchmarks will be their utility in integrating data from diverse experiments [14].
Shortcomings and strengths of literature mining via the co-citation/Bayesian classifier approach
From our previous work [32], we realized that directly identifying protein interactions would be a difficult task if we were unable to differentiate proteins and genes from the rest of the text. We therefore concentrated on building protein name extractors and interaction extractors in parallel so that the results of the former analysis could be fed into the latter.
Crucial to this process was the creation of a high-quality dictionary of human protein names and synonyms with mappings back to database entries. We therefore decided to start by creating a set of unambiguous gene names along with their synonyms that could all be mapped to a single unified gene identifier (LocusLink identifiers, now maintained through EntrezGene). The dictionary had to have very few spurious entries to ensure minimal false positives. The resulting ID-Serve database captures the various identifiers for a given gene and creates a repository for the retrieval of these genes along with their mined interactions. Building on this dictionary, the CRF algorithm then analyzed the context in which likely protein names appeared in order to identify the protein names more accurately. In the approach we describe, protein interaction partners are identified from among these protein names by a filtered version of co-citation.
The co-citation approach [14,26,40] calculates the random probability of co-occurrence of two protein names. The assumption is that if the co-citation is statistically unlikely under the random model, then there is a true underlying reason for the proteins to be co-cited - that is, they are interacting at either the functional, pathway level, or are co-localized or physically interact. The method has both advantages and disadvantages. It does not extract all interactions, but only those with statistically significant co-citations. By using the Bayesian estimator [36] we enrich further for physical interactions, but at the expense of coverage. Among the disadvantages are that the algorithm enriches for certain types of errors (for example, 'A does not interact with B', dictionary errors leading to synonyms being wrongly enriched, and so on). However, we feel the advantages outweigh the disadvantages: In particular, the probabilistic ranking, combined with the Bayesian filter, minimizes systematic errors, and at the left side of Figure 4b, it can be seen that errors in the co-citation data are no more extensive than errors introduced in transferring annotation from other organisms, or those errors introduced by human curators reading Medline abstracts. The method is easily applied, and currently outperforms other publicly available protein interaction extraction algorithms [34,35]. Finally, the precise nature of the interaction can be directly checked from the linked Medline abstracts. Thus, the mined interactions will be ideal for manual validation by curators of protein interaction databases (for example, DIP and BIND).
Conclusion
In conclusion, to prepare for attempts to map the set of human protein interactions we sought to consolidate known interactions and to establish measures of accuracy that are useful for the evaluation and integration of upcoming datasets. We established two benchmarks for assessing the quality of large-scale human protein interaction datasets, providing quantitative measures useful for the testing and integration of interaction data. Using these benchmarks, along with available and mined interactions, we assembled an integrated dataset of 31,609 interactions between 7,748 human proteins, forming a framework for the interpretation of human functional genomics data. These data are collected in the ID-Serve database [37], which can be queried for protein interactions and their corresponding Medline citations. We estimate these interactions form less than 10% of the human interactome, setting the stage for future efforts to map the complete human network of protein interactions.
Materials and methods
Identification of human protein names and interactions in Medline abstracts
The training datasets used for the literature mining are as in [32]. The dictionary of human protein names was assembled from the LocusLink and Swiss-Prot databases by manually curating the gene names and synonyms (87,723 synonyms between 18,879 unique gene names) to remove genes that were referred to as 'hypothetical' or 'probable' and to omit entries that referred to more than one protein identifier. From the Medline database of approximately 11 million abstracts (1951-2002) we retrieved 753,459 abstracts containing the word 'human' either in the title or the text to use as our corpus for extracting protein interactions.
We have previously described [32] effective protein and gene name tagging using an algorithm based on maximum entropy. Conditional random fields (CRF) [41] are new types of probabilistic models that preserve all the advantages of maximum entropy models and at the same time avoid the label bias problem by allowing a sequence of tagging decisions to compete against each other in a global probabilistic model. In this paper, we show that CRF outperforms our best previous maximum entropy tagger.
In both training and testing the CRF protein-name tagger, the corresponding Medline abstracts were processed as follows: text was tokenized using white space as delimiters and treating all punctuation marks as separate tokens. The text was segmented into sentences, and part-of-speech tags were assigned to each token using Brill's tagger [42]. For each token in each sentence, a vector of binary features was generated using the feature templates employed by the maximum entropy approach described in [32]. Each feature occurring in the training data was associated with a parameter in the CRF model. We used the CRF implementation from McCallum [43]. To train the CRF's parameters, we used 750 Medline abstracts manually annotated for protein names [32]. We then tagged predicted protein names in the entire set of 753,459 Medline abstracts using the version of the CRF algorithm that utilizes the dictionary as part of the learned model (Figure 2), and in this way linked each tagged name to a dictionary entry. The Medline abstracts with marked-up protein names are available on request.
The model assigns each candidate phrase a probability of being a protein name. We selected all names scoring higher than a given threshold (testing thresholds between 40% and 95%), retaining the proteins' LocusLink identifiers along with the PubMed identifiers (PMID) of the associated abstracts. The significance of co-citation of two protein names across a set of Medline abstracts was calculated from the hypergeometric distribution [14,26] as:
,
where:
and N equals the total number of abstracts, n of which cite the first protein, m cite the second protein, and l cite both.
The top-scoring 15,000 co-cited protein pairs were then re-ranked according to the tendency of the co-citing abstracts to discuss protein-protein interactions. Specifically, the likelihood of a co-citing abstract to discuss physical protein interactions was evaluated using the naive Bayesian classifier as described in [36], which scores Medline abstracts according to usage frequencies of discriminating words relating to protein-protein interactions. For each co-cited protein pair, we calculated the average of the scores of the co-citing Medline abstracts, then re-ranked the co-cited protein pairs by these average scores.
Analysis of network properties
We evaluated the clustering of genes in an interaction network [38] by calculating the average clustering coefficient (<C>) of the N genes as:
where Ci is the clustering coefficient of gene i, evaluated over the set of genes with at least two interactions and measured as the number of links, n, among the gene's k neighbors, divided by the number of maximum possible linkages, k(k -1)/2.
Construction of the functional annotation benchmark
The specific GO and KEGG annotations for the functional benchmarks were downloaded from the Gene Ontology database [44] and the KEGG database [45]. Within the GO process annotation hierarchy (more strictly, a directed acyclic graph (DAG)), the number of distinct annotation terms is maximal at level 8, where the level is defined as the number of nestings from the root node (level 1), as given in the Gene Ontology DAG file [44]. KEGG functional annotations were constructed as the sets of numerical codes for the KEGG pathway diagrams associated with each gene. The functional annotation benchmark is composed of all pairs of human genes sharing annotation from either source (KEGG or GO). For training and testing sets, annotated genes were randomly assigned into two categories and associations were only considered between genes of the same category.
The ID-Serve database
ID-Serve is a relational mySQL database of human proteins created to simplify comparison of datasets with differing protein identifiers. The database maps 42,232 LocusLink (now EntrezGene) identifiers to their corresponding Genecard, Swiss-Prot, Ensembl, OMIM, Unigene, NCBI GI codes and Accession numbers and to the GO and KEGG pathway annotations. Protein interaction data can be retrieved from ID-Serve, with co-citation derived interactions hyperlinked to the supporting Medline abstracts.
Additional data files
The following additional data relevant to the analysis, training and testing carried out in this work are available with the online version of this paper and can also be obtained from the ID-Serve database [37]. Additional data files 1 and 2 contain tables of protein 'tagger' training sets. Additional data file 3 contains a dictionary of human protein names and synonyms indexed to LocusLink identifiers. Additional data file 4 contains the final set of 31,609 protein interactions between 7,748 proteins derived from this analysis. Additional data file 5 contains the final set of co-citation/Bayesian classifier-derived interactions with the PubMed identifiers of co-citing abstracts. Additional data file 6 contains the benchmark training set of functional annotations. Additional data file 7 contains the benchmark test set of functional annotations. Additional data file 8 contains the benchmark set of physical interactions. Additional data file 9 contains the discriminating word list used by the Bayesian classifier to estimate the likelihood of Medline abstracts to discuss protein interactions.
Supplementary Material
Additional File 1
Training set of 200 Medline abstracts with all occurrences of protein names tagged
Click here for file
Additional File 2
Training set of 750 Medline abstracts with all occurrences of protein names tagged
Click here for file
Additional File 3
Dictionary of human protein names and synonyms indexed to LocusLink identifiers
Click here for file
Additional File 4
Final set of 31,609 protein interactions between 7,748 proteins derived from this analysis
Click here for file
Additional File 5
Final set of co-citation/Bayesian classifier-derived interactions with the PubMed identifiers of co-citing abstracts
Click here for file
Additional File 6
Benchmark training set of functional annotations
Click here for file
Additional File 7
Benchmark test set of functional annotations
Click here for file
Additional File 8
Benchmark set of physical interactions
Click here for file
Additional File 9
Discriminating word list used by the Bayesian classifier to estimate the likelihood of Medline abstracts to discuss protein interactions
Click here for file
Acknowledgements
We thank Insuk Lee for critical comments and Zack Simpson for critical comments and help with network visualization. We also thank Ewan Birney's group at the European Bioinformatics Institute for providing us with the interaction data from Reactome. This work was supported by grants from the NSF. (IIS-0325116, EIA-0219061), NIH. (GM06779-01), Welch (F1515), and a Packard Fellowship (E.M.M.).
Figures and Tables
Figure 1 Overlap between existing human protein interaction sets. A Venn diagram shows the overlap is small among the existing, publicly available human protein interaction datasets (specifically, Reactome, BIND, and HPRD protein interaction data). The small overlap (< 0.1% in common in all three datasets) implies that the number of protein interactions described in the literature is actually quite large and that the individual datasets carry specific biases.
Figure 2 Comparison of precision and accuracy of the algorithms. The conditional random fields (CRF) algorithm considerably outperforms other approaches for identifying human protein names in Medline abstracts, such as the simple matching of words to a dictionary of protein names, as well as the other available protein name-tagging algorithms in [32], Kex [34] and Abgene [35]. The tests are performed on 200 manually annotated Medline abstracts [33]. The precision (the number of correct protein names among all identified names) in identifying proteins is plotted against the recall (the number of correct protein names among all possible correct protein names). Higher scores on both precision and recall are preferable; however, for this purpose, we seek to maximize precision and can tolerate lower recall.
Figure 3 The performance of the co-citation algorithm at identifying protein interactions. (a) The probabilistic score effectively ranks co-cited proteins by their tendency to participate in the same pathway, as measured on the functional annotation training benchmark. As the probability of random co-citation decreases, the functional relatedness of the co-cited proteins increases. This tendency is robust to changes in the CRF confidence threshold chosen (data not shown). Each point represents 3,000 protein pairs. (b) An examination of the number of protein pairs identified at different CRF thresholds (0.8, 0.6, and 0.4) shows that the recall of the method is increased with lowered thresholds. Re-ranking the 15,000 top-scoring protein pairs (CRF threshold = 0.8) by the tendency of the abstracts to discuss physical protein interactions shows their consistent performance in the annotation benchmark.
Figure 4 A comparison of the available human protein interaction data on the two benchmarks. (a) An examination of the initial performance of the datasets on the functional annotation test benchmark reveals the relative quality of each dataset. The interactions extracted using co-citation analysis filtered by the Bayesian estimator show a robust behavior in terms of their scores. (b) Comparison of the performance of the interactions retrieved from the co-citation analysis after incorporating the Bayesian filter and the interactions from HPRD and orthology transfer, as assessed on the physical interaction benchmark. The Bayesian filter effectively ranks the co-citation-derived interactions in terms of their correspondence to physical protein interactions.
Figure 5 Comparison of extracted interactions with existing interactions. A comparison of interactions inferred from orthology [21] and those recovered by co-citation with the other existing human protein interaction datasets reveals that the overlap is small. The trend implies that the different methods are sampling relatively exclusive sets of interactions although, with the exception of the orthology-derived interactions, they are all derived directly from the primary biological literature.
Figure 6 Visualization of the final consolidated network of protein interactions. A view of the composite interaction network (31,609 interactions between 7,748 proteins). Of these, 6,706 proteins (87%) are connected by at least one interaction into the central, connected network component. The modularity in the network can be seen in the superimposed three-dimensional visualization, a histogram in which higher peaks correspond to larger numbers of edges per unit area. The network coordinates were generated by LGL [46] and visualized with Zlab by Zack Simpson.
Table 1 The initial list of the interactions and proteins represented in each of the existing human protein interaction datasets with total interactions, unique self-interactions and unique non-self interactions
Dataset Version Total interactions (number of proteins) Unique self (A-A) interactions (number of proteins) Unique (A-B) interactions (number of proteins)
Reactome 08/03/04 12,497 (6,257) 160 (160) 12,336 (807)
BIND 08/03/04 6,212 (5,412) 549 (549) 5,663 (4,762)
HPRD* 04/12/04 12,013 (4,122) 3,028 (3,028) 6,054 (2,747)
Orthology transfer (all) 03/31/04 71,497 (6,257) 373 (373) 71,124 (6,228)
Orthology transfer (core) 03/31/04 11,488 (3,918) 206 (206) 11,282 (3,863)
*Difficult to measure: HPRD records genes by their names, leading occasionally to entries for the same gene under different synonyms. The numbers reported are after mapping to LocusLink.
Table 2 A comparison of the contributions of each dataset to the composite human protein interaction map, with network properties of each of the datasets
Dataset Version Number of interactions Number of proteins Clustering <C> Connectivity <#interactions/protein>
Reactome 08/03/04 9,987 619 0.74 15.4
BIND 08/03/04 1,536 1,212 0.1 1.3
HPRD 04/12/04 6,054 2,747 0.09 2.2
Orthology inferred (core) 03/31/04 9,283 3,469 0.13 2.7
Co-citation This paper 6,580 3,737 0.3 1.8
Total This paper 31,609 7,748 0.24 4.1
An analysis of network features (clustering coefficient [38] and degree of connectivity) of each of the datasets indicates low degree (<k>) for all except Reactome, which is by far the most densely sampled protein interaction dataset. The final combined network is modular in structure and shows extensive, non-random clustering of proteins as compared to randomly generated networks with equal numbers of proteins and interactions (<C> = 9 × 10-3 ± -3 × 10-5; average of 10 trials).
==== Refs
Ito T Chiba T Ozawa R Yoshida M Hattori M Sakaki Y A comprehensive two-hybrid analysis to explore the yeast protein interactome. Proc Natl Acad Sci USA 2001 98 4569 4574 11283351 10.1073/pnas.061034498
Uetz P Giot L Cagney G Mansfield TA Judson RS Knight JR Lockshon D Narayan V Srinivasan M Pochart P A comprehensive analysis of protein-protein interactions in Saccharomyces cerevisiae. Nature 2000 403 623 627 10688190 10.1038/35001009
Gavin AC Bosche M Krause R Grandi P Marzioch M Bauer A Schultz J Rick JM Michon AM Cruciat CM Functional organization of the yeast proteome by systematic analysis of protein complexes. Nature 2002 415 141 147 11805826 10.1038/415141a
Ho Y Gruhler A Heilbut A Bader GD Moore L Adams SL Millar A Taylor P Bennett K Boutilier K Systematic identification of protein complexes in Saccharomyces cerevisiae by mass spectrometry. Nature 2002 415 180 183 11805837 10.1038/415180a
Tong AH Evangelista M Parsons AB Xu H Bader GD Page N Robinson M Raghibizadeh S Hogue CW Bussey H Systematic genetic analysis with ordered arrays of yeast deletion mutants. Science 2001 294 2364 2368 11743205 10.1126/science.1065810
Tong AH Lesage G Bader GD Ding H Xu H Xin X Young J Berriz GF Brost RL Chang M Global mapping of the yeast genetic interaction network. Science 2004 303 808 813 14764870 10.1126/science.1091317
Gabaldon T Huynen MA Prediction of protein function and pathways in the genome era. Cell Mol Life Sci 2004 61 930 944 15095013 10.1007/s00018-003-3387-y
Eisenberg D Marcotte EM Xenarios I Yeates TO Protein function in the post-genomic era. Nature 2000 405 823 826 10866208 10.1038/35015694
Huynen MA Snel B von Mering C Bork P Function prediction and protein networks. Curr Opin Cell Biol 2003 15 191 198 12648675 10.1016/S0955-0674(03)00009-7
Mellor JC Yanai I Clodfelter KH Mintseris J DeLisi C Predictome: a database of putative functional links between proteins. Nucleic Acids Res 2002 30 306 309 11752322 10.1093/nar/30.1.306
Jansen R Yu H Greenbaum D Kluger Y Krogan NJ Chung S Emili A Snyder M Greenblatt JF Gerstein M A Bayesian networks approach for predicting protein-protein interactions from genomic data. Science 2003 302 449 453 14564010 10.1126/science.1087361
Deane CM Salwinski L Xenarios I Eisenberg D Protein interactions: two methods for assessment of the reliability of high throughput observations. Mol Cell Proteomics 2002 1 349 356 12118076 10.1074/mcp.M100037-MCP200
von Mering C Krause R Snel B Cornell M Oliver SG Fields S Bork P Comparative assessment of large-scale datasets of protein-protein interactions. Nature 2002 417 399 403 12000970 10.1038/nature750
Lee I Date SV Adai AT Marcotte EM A probabilistic functional network of yeast genes. Science 2004 306 1555 1558 15567862 10.1126/science.1099511
Mrowka R Patzak A Herzel H Is there a bias in proteome research? Genome Res 2001 11 1971 1973 11731485 10.1101/gr.206701
Giot L Bader JS Brouwer C Chaudhuri A Kuang B Li Y Hao YL Ooi CE Godwin B Vitols E A protein interaction map of Drosophila melanogaster. Science 2003 302 1727 1736 14605208 10.1126/science.1090289
Li S Armstrong CM Bertin N Ge H Milstein S Boxem M Vidalain PO Han JD Chesneau A Hao T A map of the interactome network of the metazoan C. elegans. Science 2004 303 540 543 14704431 10.1126/science.1091403
Bader GD Betel D Hogue CW BIND: the Biomolecular Interaction Network Database. Nucleic Acids Res 2003 31 248 250 12519993 10.1093/nar/gkg056
Xenarios I Salwinski L Duan XJ Higney P Kim SM Eisenberg D DIP, the Database of Interacting Proteins: a research tool for studying cellular networks of protein interactions. Nucleic Acids Res 2002 30 303 305 11752321 10.1093/nar/30.1.303
Peri S Navarro JD Kristiansen TZ Amanchy R Surendranath V Muthusamy B Gandhi TK Chandrika KN Deshpande N Suresh S Human protein reference database as a discovery resource for proteomics. Nucleic Acids Res 2004 32 Database D497 501 14681466 10.1093/nar/gkh070
Lehner B Fraser AG A first-draft human protein-interaction map. Genome Biol 2004 5 R63 15345047 10.1186/gb-2004-5-9-r63
Joshi-Tope G Gillespie M Vastrik I D'Eustachio P Schmidt E de Bono B Jassal B Gopinath GR Wu GR Matthews L Reactome: a knowledgebase of biological pathways. Nucleic Acids Res 2005 33 Database D428 432 15608231 10.1093/nar/gki072
Reactome database
Bouwmeester T Bauch A Ruffner H Angrand PO Bergamini G Croughton K Cruciat C Eberhard D Gagneur J Ghidelli S A physical and functional map of the human TNF-alpha/NF-kappa B signal transduction pathway. Nat Cell Biol 2004 6 97 105 14743216 10.1038/ncb1086
Colland F Jacq X Trouplin V Mougin C Groizeleau C Hamburger A Meil A Wojcik J Legrain P Gauthier JM Functional proteomics mapping of a human signaling pathway. Genome Res 2004 14 1324 1332 15231748 10.1101/gr.2334104
Jenssen TK Laegreid A Komorowski J Hovig E A literature network of human genes for high-throughput analysis of gene expression. Nat Genet 2001 28 21 28 11326270 10.1038/88213
Rzhetsky A Iossifov I Koike T Krauthammer M Kra P Morris M Yu H Duboue PA Weng W Wilbur WJ GeneWays: a system for extracting, analyzing, visualizing, and integrating molecular pathway data. J Biomed Inform 2004 37 43 53 15016385 10.1016/j.jbi.2003.10.001
Liu H Wong L Data mining tools for biological sequences. J Bioinform Comput Biol 2003 1 139 167 15290785 10.1142/S0219720003000216
Hirschman L Park JC Tsujii J Wong L Wu CH Accomplishments and challenges in literature data mining for biology. Bioinformatics 2002 18 1553 1561 12490438 10.1093/bioinformatics/18.12.1553
Kanehisa M Goto S Kawashima S Okuno Y Hattori M The KEGG resource for deciphering the genome. Nucleic Acids Res 2004 32 Database D277 280 14681412 10.1093/nar/gkh063
Ashburner M Ball CA Blake JA Botstein D Butler H Cherry JM Davis AP Dolinski K Dwight SS Eppig JT Gene ontology: tool for the unification of biology. The Gene Ontology Consortium. Nat Genet 2000 25 25 29 10802651 10.1038/75556
Bunescu R Ge R Kate R Marcotte EM Mooney RJ Ramani AK Wong YW Comparative experiments on learning information extractors for proteins and their interactions. Artificial Intell Med 2005 doi:10.1016/j.artmed.2004.07.016
Franzen K Eriksson G Olsson F Asker L Liden P Coster J Protein names and how to find them. Int J Med Inform 2002 67 49 61 12460631 10.1016/S1386-5056(02)00052-7
Fukuda K Tamura A Tsunoda T Takagi T Toward information extraction: identifying protein names from biological papers. Pac Symp Biocomput 1998 707 718 9697224
Tanabe L Wilbur WJ Tagging gene and protein names in biomedical text. Bioinformatics 2002 18 1124 1132 12176836 10.1093/bioinformatics/18.8.1124
Marcotte EM Xenarios I Eisenberg D Mining literature for protein-protein interactions. Bioinformatics 2001 17 359 363 11301305 10.1093/bioinformatics/17.4.359
ID-Serve
Barabasi AL Oltvai ZN Network biology: understanding the cell's functional organization. Nat Rev Genet 2004 5 101 113 14735121 10.1038/nrg1272
International Human Genome Sequencing Consortium Finishing the euchromatic sequence of the human genome. Nature 2004 431 931 945 15496913 10.1038/nature03001
Stapley BJ Benoit G Biobibliometrics: information retrieval and visualization from co-occurrences of gene names in Medline abstracts. Pac Symp Biocomput 2000 529 540 10902200
Lafferty J McCallum A Pereira F Danyluk A Conditional Random Fields: Probabilistic models for segmenting and labeling sequence data. Proc 18th Int Conf Machine Learning (ICML 2001) 2001 San Francisco: Morgan Kaufman
Brill E. Transformation-based error driven learning and natural language processing: A case study in parts of speech tagging. Comput Linguistics 1995 21 543 565
McCallum AK MALLET: A Machine Learning for Language Toolkit 2002
Gene Ontology database
KEGG Encyclopedia
Adai AT Date SV Wieland S Marcotte EM LGL: creating a map of protein function with an algorithm for visualizing very large biological networks. J Mol Biol 2004 340 179 190 15184029 10.1016/j.jmb.2004.04.047
| 15892868 | PMC1175952 | CC BY | 2021-01-04 16:05:39 | no | Genome Biol. 2005 Apr 15; 6(5):R40 | utf-8 | Genome Biol | 2,005 | 10.1186/gb-2005-6-5-r40 | oa_comm |
==== Front
Genome BiolGenome Biology1465-69061465-6914BioMed Central London gb-2005-6-5-r411589286910.1186/gb-2005-6-5-r41ResearchGenome-scale evidence of the nematode-arthropod clade Dopazo Hernán [email protected] Joaquín [email protected] Pharmacogenomics and Comparative Genomics Unit, Bioinformatics Department, Centro de Investigación Príncipe Felipe, Autopista del Saler 16, 46013 Valencia, Spain2 Functional Genomics Unit, Bioinformatics Department, Centro de Investigación Príncipe Felipe, Autopista del Saler 16, 46013 Valencia, Spain3 Functional Genomics Node, INB, Centro de Investigación Príncipe Felipe, Autopista del Saler 16, 46013 Valencia, Spain2005 28 4 2005 6 5 R41 R41 7 3 2005 6 4 2005 Copyright © 2005 Dopazo and Dopazo; licensee BioMed Central Ltd.This is an Open Access article distributed under the 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 most extensive phylogenetic analysis carried out to date, including 11 complete genomes, is shown to support the Ecdysozoa hypothesis in the open-ended debate of the Coelomata-Ecdysozoa evolutionary problem.
Background
The issue of whether coelomates form a single clade, the Coelomata, or whether all animals that moult an exoskeleton (such as the coelomate arthropods and the pseudocoelomate nematodes) form a distinct clade, the Ecdysozoa, is the most puzzling issue in animal systematics and a major open-ended subject in evolutionary biology. Previous single-gene and genome-scale analyses designed to resolve the issue have produced contradictory results. Here we present the first genome-scale phylogenetic evidence that strongly supports the Ecdysozoa hypothesis.
Results
Through the most extensive phylogenetic analysis carried out to date, the complete genomes of 11 eukaryotic species have been analyzed in order to find homologous sequences derived from 18 human chromosomes. Phylogenetic analysis of datasets showing an increased adjustment to equal evolutionary rates between nematode and arthropod sequences produced a gradual change from support for Coelomata to support for Ecdysozoa. Transition between topologies occurred when fast-evolving sequences of Caenorhabditis elegans were removed. When chordate, nematode and arthropod sequences were constrained to fit equal evolutionary rates, the Ecdysozoa topology was statistically accepted whereas Coelomata was rejected.
Conclusions
The reliability of a monophyletic group clustering arthropods and nematodes was unequivocally accepted in datasets where traces of the long-branch attraction effect were removed. This is the first phylogenomic evidence to strongly support the 'moulting clade' hypothesis.
==== Body
Background
Understanding the evolution of the great diversity of life is a major goal in biology. Despite decades of effort by systematists, evolutionary relationships between major groups of animals still remain unresolved. The inability to cluster taxa in monophyletic groups was originally due to the lack of morphological synapomorphies among phyla. An alternative solution came from embryology, and animal systematics relied on criteria based on increasing complexity of body plan [1]. Thus, the traditional metazoan phylogeny clusters animals from the simplest basal forms with loose tissue organization (for example, sponges) to those having two germ layers (dipoblastic animals, for example cnidarians), and those developing from three germ layers (triploblastic animals, such as the Bilateria - animals with bilateral symmetry). Bilateral animals were ordered into those lacking a coelom (the acoelomates, such as platyhelminths), those with a false coelom (the pseudocoelomates, such as nematodes), and, finally, those animals with a true coelom (the Coelomata, such as the arthropods and chordates). This comparative developmental theory of animal evolution dominated animal systematics for more than 50 years [2].
Subsequently, molecular systematic studies based on small subunit ribosomal RNA (18S rRNA) sequences began to undermine this scenario [1]. Put briefly, the new animal phylogeny suggested that clades such as acoelomates and pseudocoelomates are artificial systematic groups. Moreover, although the coelomate designation still remains, this clade now contains two new lineages: the lophotrochozoa and the Ecdysozoa [3]. The 'Ecdysozoa hypothesis' postulated that all phyla composed of animals that grow by moulting a cuticular exoskeleton (such as arthropods and nematodes) originate from a common ancestor, thus forming a distinct clade. Thus, under the Ecdysozoa hypothesis arthropods are genetically more closely related to nematodes than to chordates. Under the 'Coelomata hypothesis' of animal evolution, however, arthropods are more closely related to chordates than to nematodes.
At the heart of this systematic debate, a technical discussion emerged surrounding the long-branch attraction effect (LBAE), taxon sampling, and the number of characters used. Subsequent molecular and morphological studies have been carried out, but the controversy remains unresolved and is presented as a multifurcation [4]. Although the use of different single-gene sequences supported the Ecdysozoa hypothesis [5-11], the analysis of dozens to hundreds of concatenated sequences supported the Coelomata clade [12-15]. Indeed, with an element of caution, we favored the Coelomata hypothesis in a previous whole-genome study designed to determine the number of characters needed to obtain a reliable topology [16]. The gene-based Ecdysozoa versus genome-scale Coelomata alternative hypotheses were recently challenged by two phylogenomics studies that partly supported the Ecdysozoa clade [17] and a paraphyletic Coelomata group [18]. Although it is generally accepted that phylogenetic analysis of whole genomes has begun to supplement (and in some cases improve on) phylogenetic studies previously carried out with one or a few genes [19], all genome-wide phylogenetic studies have failed to support the proposed new animal phylogeny.
Here we present the first phylogenomic evidence that strongly supports the Ecdysozoa hypothesis and at the same time demonstrates that the LBAE biases the position of Caenorhabditis elegans in the phylogenetic tree. We show that by using a large number of characters and choosing a phylogenetic weighted scheme of outgroups to test the constancy of evolutionary rates, the new animal phylogeny can be statistically supported. Moreover, we show that both the Coelomata and the Ecdysozoa hypotheses can be supported with the highest statistical confidence when genomic datasets are ordered according to a gradually increased adjustment to equal evolutionary rates between C. elegans and Drosophila melanogaster sequences. In between, neither Ecdysozoa nor Coelomata were sufficiently supported. To our knowledge, this is the most extensive phylogenomic analysis carried out to date in the number of characters and the number of eukaryotic species involved.
Results
Dataset properties
Sequences homologous to human exon sequences were derived from filtering tblastn search results on 11 complete eukaryotic genomes. Because the most-criticized issue in resolving the Ecdysozoa-Coelomata problem seems to be the LBAE produced by the nematode species, we decided to rearrange homologous sequences in a series of nested datasets that gradually reduced LBAE. Aligned homologous sequences were arranged in eight datasets (Di) and concatenated in their corresponding matrices (Mi) (see Materials and methods), such that as suffix i increases, datasets and matrices comprise a smaller number of homologous sequences showing more similar relative branch lengths (RBL) between C. elegans (LCe) and D. melanogaster (LDm) (Figure 1). RBL are relative human distances.
To quantify the effect on the RBL of C. elegans of concatenating alternative homologous sequences, maximum likelihood (ML) estimates of branch length were obtained using the star-like unrooted tree transformation for each dataset (see Materials and methods). Figure 2a shows that the RBL of C. elegans over D. melanogaster decreased by approximately 30% continuously from dataset D1 to D8. To test whether the gradual decrease in C. elegans branch length was enough to produce statistical confidence on equal evolutionary rates between the nematode and the arthropod sequences, relative rate tests using two outgroup schemes were assayed on concatenated sequences (see Materials and methods). Figure 2b shows that using Saccharomyces cerevisae as the unique outgroup species (OUG1), all the individual tests on the eight matrices failed to detect statistical deviations (at the 5% level family-wise) between sequences. Only when the phylogenetically weighted scheme of outgroup species (OUG2) was used did the relative rate test detect significant deviation of clock behavior from D1 to D5 datasets. We are therefore confident that the arthropod and nematode concatenated sequences of the M6, M7, and M8 matrices meet the desired clock-like conditions to test the Coelomata and Ecdysozoa hypotheses and exclude any artifacts derived from a possible LBAE. This result supports previous work suggesting that the genetic distance between ingroup and outgroup modifies the power of the relative rate test [20].
To test whether concatenated matrices carry sufficient phylogenetic signal, the ML mapping method was used. The compound posterior probability point (P) for all the possible quartets of each Mi matrix could be placed, with almost equivalent values (approximately 33%), inside the corner areas of the equilateral triangle probability surface (see Additional data file 1). Thus, concatenated matrices derived from selecting a different number of homologous sequences contained sufficient phylogenetic signal to represent topologies as strictly bifurcating trees. Finally, using the Akaike information criterion (AIC) [21], the statistical test of the best-fit model of sequence evolution for each dataset was selected from six different alternatives (see Materials and methods). As all the models are not nested and share the same number of parameters, the best one was that with the greatest log likelihood result. The WAG amino-acid replacement matrix [22] adjusted for frequencies (+F), rate heterogeneity (+Γ) and invariable sites (+I) was the best evolutionary model chosen for all the datasets. Moreover, model-fit-data values followed the same inequality independently of the dataset (WAG [22] > VT [23] > BLOSUM62 [24] > JTT [25] > PAM [26] > mtREV24 [27]), suggesting that the best models were those that consider more distantly related amino-acid sequences.
The clade Coelomata disappears under clock conditions
Distance and ML phylogenetic methods were used on all the datasets (see Materials and methods). Figure 3 shows phylogenetic reconstructions and statistical support for the two extreme conditions of the nested datasets. Whereas the M1 matrix supported the Coelomata tree with the highest statistical confidence, M8 showed the same result for the Ecdysozoa tree. Thus, by decreasing the RBL of C. elegans, the statistical support switched from the Coelomata to the Ecdysozoa hypothesis. Figure 4 shows that, whichever phylogenetic method was used, C. elegans bootstrap support between datasets and topologies changed in agreement with the gradual RBL decrement. Specifically, using M1 and M8 (the matrices showing the most extreme evolutionary rate conditions for C. elegans and D. melanogaster sequences - from a clock-absent to the most adjusted behavior), the statistical support moved from Coelomata to Ecdysozoa. The same occurred with M2 and M7. Alternatively, using M3 and M6, only one of the two distance and ML methods (Figure 4a,b) provided sufficient support (90% or more) to the hypothesis. Finally, using M4 and M5, only one distance method supported Coelomata and Ecdysozoa with confidence. Given that datasets differed principally in the RBL of C. elegans over D. melanogaster, the gradual change in topology strongly favors an LBAE between C. elegans and the more basal species. To test whether a paired-sites test [28] supports the bootstrap conclusions, Shimodaira-Hasegawa (SH) and expected-likelihood weight (ELW) tests were evaluated on the datasets (see Materials and methods).
Figure 5 shows the assessment of paired-sites tests for the two competing trees on all the datasets. Paired-sites tests supporting topologies (p > 0.05) changed almost gradually on datasets. Figure 5a and 5b show that the SH test is more conservative than the ELW [29]. Using matrices M1 and M2, both tests strongly rejected the Ecdysozoa hypothesis, whereas M6, M7, and M8 rejected the Coelomata tree. Interestingly, datasets between them did not reject any topology with sufficient statistical evidence. We can conclude that by decreasing the RBL of C. elegans over D. melanogaster by around 13% (Figure 2a) the LBAE favoring the Coelomata hypothesis disappears and we can confirm that under strict conditions of clock-like behavior, the Coelomata hypothesis was strongly rejected by paired-sites tests and bootstrap support.
To test if the shortness of the evolutionary distances between C. elegans and D. melanogaster resulting from the above filtering method biased topology over the common ancestry of arthropods and nematodes, we searched for chordate, arthropod, and nematode sequences showing clock-like behavior between them. To increase the probability of finding sequences to fit the criteria, we focused on sequences from the most closely related chordate to the molting species, that is, the ascidian Ciona intestinalis. Only 14 exon sequences met the above criteria. A relative rate test showed that the probability of a perfect clock-like behavior was p = 0.515 for C. elegans and D. melanogaster, p = 0.308 for C. intestinalis and D. melanogaster and p = 0.712 for C. intestinalis and C. elegans. The ML mapping method showed that the concatenation of all the 810 characters carried sufficient phylogenetic signal in the matrix to represent a strictly bifurcating tree (see Additional data file 2). Despite the reduced number of characters, phylogenetic analysis showed significant support for the Ecdysozoa hypothesis. Using distance and ML methods, bootstrap values reached 97%. Moreover, the Ecdysozoa hypothesis was accepted with a probability of p = 1.00 and p = 0.997 when SH and ELW paired-sites tests, respectively, were performed. Conversely, the Coelomata hypothesis was rejected at p = 0.006 and p = 0.0023, respectively.
The clade Coelomata disappears by removing fast-evolving sequences of C. elegans
In order to discard a probable biased selection of exon sequences favoring the Ecdysozoa hypothesis, two additional matrices were built by removing from the original dataset (D1) the exons in which the C. elegans sequences evolved at a faster rate. Figure 6 shows that by removing the fastest 15% of total exon sequences the reliability of the Coelomata hypothesis is reduced from 100% to 78%. Moreover, when the fastest 30% of all exons were removed, the topology changes to Ecdysozoa with 90% confidence level. The change in topology in parallel with the reduction of the C. elegans branch length points to the LBAE as the main obstacle to obtaining the true phylogenetic relationship between chordates, arthropods and nematodes. We conclude that the Ecdysozoa hypothesis does not depend on adjusting a particular set of homologous exon sequences to clock-like behavior.
Discussion
There are many reasons why the Coelomata-Ecdysozoa problem should be considered the most puzzling problem in animal systematics and a major open-ended subject in evolutionary biology. The monophyly of the Ecdysozoa group, strongly championed by the evo-devo community [30], was originally deduced, and continually recovered, through the analysis of different single-gene sequences [3,5,6,8-11], sometimes in combination with morphological characters [7]. There is need for caution, however, as previous studies had shown that individual genes are not sufficient to estimate the correct genome phylogeny [19,31]. Furthermore, the reliability of some of the phylogenetic markers used to derive Ecdysozoa has been seriously questioned [32,33]. Those that consider the Ecdysozoa hypothesis as more plausible insist that the Coelomata topology is an artifact of LBAE, derived from the fact that nematode genomes, particularly that of C. elegans, evolve at higher rates [3], and are consequently displaced to a more basal position.
On the other hand, as phylogenetic reconstruction assumes that sampled data are representative of the whole genome from which they are drawn [34], there is increasing agreement to consider genome-scale analysis more accurate than single-gene analysis when deciding between conflicting topologies [19,31]. Conflict derives from the fact that all previous genome-wide phylogenetic attempts to test the hypothesis have failed to confirm the 'moulting group' - the Ecdysozoa - as a clade. All phylogenomic analyses carried out to date favor the Coelomata hypothesis with the highest statistical support [12-16]. Furthermore, the Coelomata tree has shown to be robust to criticism deriving from LBAE [12,14-16] and nematode species inclusion [14]. Those that consider the Coelomata hypothesis to be more appropriate insist that longer sequences, rather than extensive taxon sampling [35], will more effectively improve the accuracy of phylogenetic inference [14,15,36,37], and emphasize that an inevitable trade-off exists between the number of characters and the number of species used in the study [15].
We show here that by using the fast-evolving nematode C. elegans the Ecdysozoa can be recovered using genome-scale phylogenetic analysis. Our analysis has been performed over the largest number of eukaryotic genomes and over the largest number of amino-acid residues ever used to test the hypothesis. The major differences from previous genomic approaches are threefold. First, we used a large number of short conserved sequences (around 50 amino acids long) derived from human homologous exon sequences. Only exon sequences derived from eight genes, out of a total of around 100 analyzed by Blair et al. [14], were used in our analysis. The remaining genes contained in the 18 human chromosomes did not pass the BLAST filters applied in the analysis. Second, we arranged the dataset such that the sequences, including those evolving faster or slower, were included if they met the condition of equal rate of change between two (C. elegans and D. melanogaster) or three species (C. intestinalis, D. melanogaster and C. elegans). Third, we used a large number of characters (amino-acid residues) and a weighted distant outgroup species to enhance the power of the relative rate test [20].
As discussed in our previous paper [16], by including or excluding certain human homologous exon sequences, we reduced the problem of LBAE and added a probable bias favoring Coelomata. The present work confirms that this bias exists. The concatenation and the posterior phylogenetic analysis of the sequences shared by the eukaryotes used in this analysis provide a viable solution to the ancestor-descendant relationships of animal species once the LBAE is removed.
Conclusions
Acceptance of the new animal phylogeny and the Ecdysozoa hypothesis would provide a new scheme to understand the Cambrian explosion [38,39] and the origin of metazoan body plans [9,30] and consequently would set a new phylogenetic framework for comparative genomics [40]. We have shown how phylogenetic reconstruction based on whole-genome sequences has the potential to solve one of the most controversial hypotheses in animal evolution: the reliability of the Ecdysozoa clade.
Materials and methods
Dataset collection
Complete genome sequences from Plasmodium falciparum [41], Arabidopsis thaliana [42], Oryza sativa [43], Saccharomyces cerevisae [44], Caenorhabditis elegans [45], Anopheles gambiae [46], Drosophila melanogaster [47], Ciona intestinalis [48], Fugu rubripes [49], Mus musculus [50] and Homo sapiens [51] were downloaded and formatted to run local BLAST [52]. Amino-acid sequences corresponding to all the gene exons in a sample of 18 human chromosome including 6-18, 20-22, X and Y (approximately 14,000 genes and 140,000 exons), were obtained from the Ensembl database project [53]. Human paralogous exons were excluded by running local blastp [52] on a human exon database built ad hoc. Only the best of those sequences, with more than a single hit with a fraction of aligned and conserved amino-acid sequence ≥ 95% and ≥ 90% respectively, were retained to find homologous sequences in the other eukaryotic species (threshold values based on a previous human paralogous study [54]). We used tblastn [52] that searches a query amino-acid sequence on the six translation frames of the target sequence to search for homology in the complete genome databases of the species mentioned above. Exons less than 22 amino acids were removed from the analysis. Each best hit of tblastn was filtered by means of a threshold e-value (≤ 1e-03) and a threshold proportion of the query over the subject sequence length (≥ 75%). Only those exons that pass through all the species filter conditions were selected as the final dataset of human exon homologous sequences. All the exon homologous sequences were aligned using Clustal W [55] with default parameters. The total number of homologous sequences, derived from 18 human chromosomes, corresponds to 1,192 exons selected from 610 known genes, adding up to more than 55,500 amino-acid characters.
To arrange homologous sequences in different datasets, pairwise distances between sequences were extracted using the PROTDIST program (Kimura option) of the PHYLIP package [56]. Distances between C. elegans, D. melanogaster and H. sapiens were transformed into branch lengths in a star-like unrooted tree (la = (dab + dac - dbc)/2, where la is the length of the branch leading to a and dab, dac, dbc are the distances between a and b, a and c, and b and c, respectively). It is important to emphasize that we are not considering that the phylogenetic relationships of C. elegans, D. melanogaster and H. sapiens is a star topology. We used this exact equation for determining the branch lengths of the three species, because the unique way to arrange three species in a phylogenetic tree is a star topology. We consider C. elegans, D. melanogaster and H. sapiens to be members of the ingroup and P. falciparum, A. thaliana, O. sativa and S. cerevisae as the outgroup species at the moment to root the phylogenetic tree. Homologous exon sequences were arranged in eight datasets according to their pertinence to more inclusive areas surrounding the straight line representing identical relative branch lengths (RBLs) of C. elegans (LCe = lCe/lHs) and D. melanogaster (LDm = lDm/lHs). The Di dataset clusters all the homologous exon alignments where LDm - δi ≤ LCe ≤ LDm + δi, where i is an integer ranging from 2 to 7 and δi = 5.0, 3.0,2.5,2.0,15,1.0,0.5. The D1 dataset contains all the exon homologous sequences without the constraints of evolutionary rates. Exons with negative or undefined normalized distances (lHs = 0) were excluded from the analysis. All the aligned homologous exon sequences of the Di dataset were concatenated in the Mi matrix. Three additional matrices were derived from D1: two by removing exons containing LCe ≥ and LCe ≥ , and the last one by adjusting the sequences of C. intestinalis, D. melanogaster and C. elegans to clock-like behavior.
Phylogenetic methods
The relative rate test was performed at the 5% statistical level by means of the RRTree program [57] using outgroups with one (S. cerevisae; OUG1) or more species (S. cerevisae, A. thaliana, O. sativa and P. falciparum; OUG2). In the latter case, an explicit weighted phylogenetic scheme was chosen (1/2 S. cerevisae, ((1/8 A. thaliana, 1/8 O. sativa), 1/4 P. falciparum)). Given that three ingroups were set for all analyses (the chordates H. sapiens, M. musculus, F. rubripes, and C. intestinalis; the arthropods Anopheles gambiae and Drosophila melanogaster; and the nematode C. elegans), the threshold value was corrected for multiple testing to 5/3 = 1.7%. TREE-PUZZLE [58] was used to evaluate six alternative evolutionary models adjusted for frequencies (+F), site rate variation (+Γ distribution with two rates) and a proportion of invariable sites (+I), to estimate the amount of evolutionary information of datasets by the likelihood-mapping method [59], to derive the maximum likelihood (ML) trees using the quartet-puzzling algorithm, to set the ML pairwise sequence distances, and to test alternative topologies using SH [60] and ELW [29] tests. The PROML (JTT+f) program of the PHYLIP package [56] was used to estimate ML trees derived from the stepwise addition algorithm. Distance methods of phylogenetic reconstruction were performed using PROTDIST (JTT, Kimura options), NEIGHBOR (neighbor-joining (NJ) [61]) and least squares (LS) [62] algorithms, and CONSENSE (50% majority-consensus rule option) programs on 100 bootstrap replications using PHYLIP.
Additional data files
The following additional data files are available with the online version of this paper. Additional data file 1 contains a figure showing ML puzzle mapping of the Mi matrices.
Additional data file 2 contains a figure showing ML puzzle mapping of the matrix derived from chordate, arthropod and nematode sequences showing clock-like behavior. Additional data file 3 contains the matrices.
Supplementary Material
Additional File 1
ML puzzle mapping of the Mi matrices. Maximum likelihood mapping results for each one of the Mi concatenated matrices. From the first row and from left to right, M1 to M2 until the fourth row, M7 to M8.
Click here for file
Additional File 2
ML puzzle mapping of the matrix derived from chordate, arthropod and nematode sequences showing clock-like behavior. ML mapping of the concatenated matrix derived from constraining sequences to 3 clocks-like behavior.
Click here for file
Additional File 3
Matrices. The full set of matrices (phylip format) used in the phylogenetic analyzes.
Click here for file
Acknowledgements
We thank especially Javier Santoyo and the Bioinformatics department members at the Centro de Investigación Príncipe Felipe. We thank J. Castresana, D. Posada and R. Zardoya for comments and suggestions, and M. Robinson-Rechavi for updating the code of the RRTree software. Special thanks goes to Amanda Wren for her revision of the English. H.D. acknowledges the support of Fundación Carolina and Fundación la Caixa.
Figures and Tables
Figure 1 Description of the dataset. Di datasets are arranged according to a gradual decrease in the parameter δ. δ controls the inclusion of each homologous exon sequence in the dataset by defining margins above and below (y = x ± δ) a diagonal line (y = x) that constrains clock-like behavior in the evolution of C. elegans and D. melanogaster sequences. LCe and LDm are the respective relative branch lengths of C. elegans and D. melanogaster using H. sapiens as reference. Comma-separated values represent the number of homologous sequences and characters aligned in the Mi concatenated matrix. Di contains all the sequences without any constraint of evolutionary rates. Dotted black and red lines represent mean , and median values, respectively.
Figure 2 Relative rate test. (a) Relative C. elegans branch lengths derived from each one of the eight Mi matrices. Maximum likelihood estimates are expressed as relative distance units of D. melanogaster. (b) Relative rate test probability values evaluated at the 5% level family-wise (red line 1.7%). OUG1, S. cerevisae; OUG2, phylogenetic weighted scheme using S. cerevisae, A. thaliana, O. sativa and P. falciparum as outgroup species.
Figure 3 Phylogenetic trees. Trees derived from M1 and M8 datasets, respectively support (a) the Coelomata and (b) the Ecdysozoa hypothesis. From left to right or top to bottom, values besides nodes show the maximum likelihood reliability values of the quartet-puzzling tree and bootstrap values using maximum likelihood, least squares, and neighbor-joining methods, respectively. Values in red show the support for (a) Coelomata and (b) Ecdysozoa nodes. Red branches display distances between C. elegans and D. melanogaster. Smaller trees are minimal representations of both hypothesis.
Figure 4 Bootstrap and reliability support for alternative topologies. Bootstrap and reliability support (50% majority consensus rule) for Coelomata (C) and Ecdysozoa (E) hypotheses derived from each one of the eight Mi matrices. (a) Distance methods. LS, least squares; NJ, neighbor joining. (b) Maximum likelihood, using PHYLIP (ph) and PUZZLE (pz). Highly supported trees were considered those with values above 90% (dotted red line).
Figure 5 Paired-sites tests. p-values inferred from paired-sites tests considering Coelomata (C) and Ecdysozoa (E) hypotheses at the 5% level (red line) for all the datasets. (a) Shimodaira-Hasegawa test (SH); (b) expected-likelihood weight method (ELW).
Figure 6 Removing fast-evolving sequences. Exon sequences of C. elegans showing LCe ≥ = 4.06 represent 15% of the total exon. When these faster exons were removed (above blue line), support for the Coelomata topology was reduced from the original 100% to 85%. Furthermore, when 28% of the faster exons were deleted (red line), Ecdysozoa is recovered with 90% statistical support. This suggests that LBAE is the main problem in obtaining the Ecdysozoa tree. Blue line, = 4.06; red line, = 2.66.
==== Refs
Adoutte A Balavoine G Lartillot N de Rosa R Animal evolution. The end of the intermediate taxa? Trends Genet 1999 15 104 108 10203807 10.1016/S0168-9525(98)01671-0
Raff RR The Shape of Life Genes, Development and the Evolution of Animal Form 1996 Chicago: The University of Chicago Press
Aguinaldo AM Turbeville JM Linford LS Rivera MC Garey JR Raff RA Lake JA Evidence for a clade of nematodes, arthropods and other moulting animals. Nature 1997 387 489 493 9168109 10.1038/387489a0
Hedges SB The origin and evolution of model organisms. Nat Rev Genet 2002 3 838 849 12415314 10.1038/nrg929
Mallatt J Winchell CJ Testing the new animal phylogeny: first use of combined large-subunit and small-subunit rRNA gene sequences to classify the protostomes. Mol Biol Evol 2002 19 289 301 11861888
Ruiz-Trillo I Paps J Loukota M Ribera C Jondelius U Baguna J Riutort M A phylogenetic analysis of myosin heavy chain type II sequences corroborates that Acoela and Nemertodermatida are basal bilaterians. Proc Natl Acad Sci USA 2002 99 11246 11251 12177440 10.1073/pnas.172390199
Peterson KJ Eernisse DJ Animal phylogeny and the ancestry of bilaterians: inferences from morphology and 18S rDNA gene sequences. Evol Dev 2001 3 170 205 11440251 10.1046/j.1525-142x.2001.003003170.x
Manuel M Kruse M Muller WE Le Parco Y The comparison of beta-thymosin homologues among metazoa supports an arthropod-nematode clade. J Mol Evol 2000 51 378 381 11040289
de Rosa R Grenier JK Andreeva T Cook CE Adoutte A Akam M Carrol SB Balavoine G Hox genes in brachiopods and priapulids and protostome evolution. Nature 1999 399 772 776 10391241 10.1038/21631
Mallatt JM Garey JR Shultz JW Ecdysozoan phylogeny and Bayesian inference: first use of nearly complete 28S and 18S rRNA gene sequences to classify the arthropods and their kin. Mol Phylogenet Evol 2004 31 178 191 15019618 10.1016/j.ympev.2003.07.013
Anderson FE Cordoba AJ Thollesson M Bilaterian phylogeny based on analyzes of a region of the sodium-potassium ATPase beta-subunit gene. J Mol Evol 2004 58 252 268 15045481 10.1007/s00239-003-2548-9
Mushegian AR Garey JR Martin J Liu LX Large-scale taxonomic profiling of eukaryotic model organisms: a comparison of orthologous proteins encoded by the human, fly, nematode, and yeast genomes. Genome Res 1998 8 590 598 9647634
Hausdorf B Early evolution of the bilateria. Syst Biol 2000 49 130 142 12116476 10.1080/10635150050207438
Blair JE Ikeo K Gojobori T Hedges SB The evolutionary position of nematodes. BMC Evol Biol 2002 2 7 11985779 10.1186/1471-2148-2-7
Wolf YI Rogozin IB Koonin EV Coelomata and not Ecdysozoa: evidence from genome-wide phylogenetic analysis. Genome Res 2004 14 29 36 14707168 10.1101/gr.1347404
Dopazo H Santoyo J Dopazo J Phylogenomics and the number of characters required for obtaining an accurate phylogeny of eukaryote model species. Bioinformatics 2004 20 Suppl 1 I116 I121 15262789 10.1093/bioinformatics/bth902
Copley RR Aloy P Russell RB Telford MJ Systematic searches for molecular synapomorphies in model metazoan genomes give some support for Ecdysozoa after accounting for the idiosyncrasies of Caenorhabditis elegans. Evol Dev 2004 6 164 169 15099303 10.1111/j.1525-142X.2004.04021.x
Philippe H Snell EA Bapteste E Lopez P Holland PW Casane D Phylogenomics of eukaryotes: the impact of missing data on large alignments. Mol Biol Evol 2004 21 1740 1752 15175415 10.1093/molbev/msh182
Rokas A Williams BL King N Carroll SB Genome-scale approaches to resolving incongruence in molecular phylogenies. Nature 2003 425 798 804 14574403 10.1038/nature02053
Bromham L Penny D Rambaut A Hendy MD The power of relative rates tests depends on the data. J Mol Evol 2000 50 296 301 10754073
Kullback S Leibler RA On information and sufficiency. Annls Math Stat 1951 22 79 86
Whelan S Goldman N A general empirical model of protein evolution derived from multiple protein families using a maximum-likelihood approach. Mol Biol Evol 2001 18 691 699 11319253
Muller T Vingron M Modeling amino acid replacement. J Comput Biol 2000 7 761 776 11382360 10.1089/10665270050514918
Henikoff S Henikoff JG Amino acid substitution matrices from protein blocks. Proc Natl Acad Sci USA 1992 89 10915 10919 1438297
Jones DT Taylor WR Thornton JM The rapid generation of mutation data matrices from protein sequences. Comput Appl Biosci 1992 8 275 282 1633570
Dayhoff MO Schwartz RM Orcutt BC Dayhoff MO A model of evolutionary change in proteins. Atlas of Protein Sequence and Structure 1978 5 Washington DC: National Biomedical Research Foundation 345 358
Adachi J Hasegawa M Model of amino acid substitution in proteins encoded by mitochondrial DNA. J Mol Evol 1996 42 459 468 8642615
Felsenstein J Inferring Phylogenies 2004 Sunderland, MA: Sinauer
Strimmer K Rambaut A Inferring confidence sets of possibly misspecified gene trees. Proc Biol Sci 2002 269 137 142 11798428 10.1098/rspb.2001.1862
Carrol SB Grenier JK Weatherbee SD From DNA to Diversity Molecular Genetics and the Evolution of Animal Design 2001 Malden, MA: Blackwell Science
Cummings MP Otto SP Wakeley J Sampling properties of DNA sequence data in phylogenetic analysis. Mol Biol Evol 1995 12 814 822 7476127
Hasegawa M Hashimoto T Ribosomal RNA trees misleading? Nature 1993 361 23 8421491 10.1038/361023b0
Abouheif E Zardoya R Meyer A Limitations of metazoan 18S rRNA sequence data: implications for reconstructing a phylogeny of the animal kingdom and inferring the reality of the Cambrian explosion. J Mol Evol 1998 47 394 405 9767685
Martin MJ Gonzalez-Candelas F Sobrino F Dopazo J A method for determining the position and size of optimal sequence regions for phylogenetic analysis. J Mol Evol 1995 41 1128 1138 8587110 10.1007/BF00173194
Hillis DM Pollock DD McGuire JA Zwickl DJ Is sparse taxon sampling a problem for phylogenetic inference? Syst Biol 2003 52 124 126 12554446
Rosenberg MS Kumar S Incomplete taxon sampling is not a problem for phylogenetic inference. Proc Natl Acad Sci USA 2001 98 10751 10756 11526218 10.1073/pnas.191248498
Rosenberg MS Kumar S Taxon sampling, bioinformatics, and phylogenomics. Syst Biol 2003 52 119 124 12554445
Balavoine G Adoutte A One or three Cambrian radiations? Science 1998 4280 397 398 10.1126/science.280.5362.397
Conway Morris S The Cambrian "explosion": slow-fuse or megatonnage. Proc Natl Acad Sci USA 2000 97 4426 4429 10781036 10.1073/pnas.97.9.4426
Eisen JA Fraser CM Phylogenomics: intersection of evolution and genomics. Science 2003 300 1706 1707 12805538 10.1126/science.1086292
Gardner MJ Hall N Fung E White O Berriman M Hyman RW Carlton JM Pain A Nelson KE Bowman S Genome sequence of the human malaria parasite Plasmodium falciparum. Nature 2002 419 498 511 12368864 10.1038/nature01097
Arabidopsis Genome Initiative Analysis of the genome sequence of the flowering plant Arabidopsis thaliana. Nature 2000 408 796 815 11130711 10.1038/35048692
Yu J Hu S Wang J Wong GK Li S Liu B Deng Y Dai L Zhou Y Zhang X A draft sequence of the rice genome (Oryza sativa L. ssp. indica). Science 2002 296 79 92 11935017 10.1126/science.1068037
Goffeau A The yeast genome directory. Nature 1997 387 Suppl 5
C. elegans Sequencing Consortium Genome sequence of the nematode C. elegans: a platform for investigating biology. Science 1998 282 2012 2018 9851916 10.1126/science.282.5396.2012
Holt RA Subramanian GM Halpern A Sutton GG Charlab R Nusskern DR Wincker P Clark AG Ribeiro JM Wides R The genome sequence of the malaria mosquito Anopheles gambiae. Science 2002 298 129 149 12364791 10.1126/science.1076181
Adams MD Celniker SE Holt RA Evans CA Gocayne JD Amanatides PG Scherer SE Li PW Hoskins RA Galle RF The genome sequence of Drosophila melanogaster. Science 2000 287 2185 2195 10731132 10.1126/science.287.5461.2185
Dehal P Satou Y Campbell RK Chapman J Degnan B De Tomaso A Davidson B Di Gregorio A Gelpke M Goodstein DM The draft genome of Ciona intestinalis : insights into chordate and vertebrate origins. Science 2002 298 2157 2167 12481130 10.1126/science.1080049
Aparicio S Chapman J Stupka E Putnam N Chia JM Dehal P Christoffels A Rash S Hoon S Smit A Whole-genome shotgun assembly and analysis of the genome of Fugu rubripes. Science 2002 297 1301 1310 12142439 10.1126/science.1072104
Waterston RH Lindblad-Toh K Birney E Rogers J Abril JF Agarwal P Agarwala R Ainscough R Alexandersson M An P Initial sequencing and comparative analysis of the mouse genome. Nature 2002 420 520 562 12466850 10.1038/nature01262
Lander ES Linton LM Birren B Nusbaum C Zody MC Baldwin J Devon K Dewar K Doyle M FitzHugh W Initial sequencing and analysis of the human genome. Nature 2001 409 860 921 11237011 10.1038/35057062
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
Birney E Andrews D Bevan P Caccamo M Cameron G Chen Y Clarke L Coates G Cox T Cuff J Ensembl 2004. Nucleic Acids Res 2004 32 Database issue D468 D470 14681459 10.1093/nar/gkh038
Bailey JA Gu Z Clark RA Reinert K Samonte RV Schwartz S Adams MD Myers EW Li PW Eichler EE Recent segmental duplications in the human genome. Science 2002 297 1003 1007 12169732 10.1126/science.1072047
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
Felsenstein J PHYLIP (Phylogeny Inference Package) version 36a3 2002 Seattle, WA: Department of Genome Sciences, University of Washington
Robinson-Rechavi M Huchon D RRTree: relative-rate tests between groups of sequences on a phylogenetic tree. Bioinformatics 2000 16 296 297 10869026 10.1093/bioinformatics/16.3.296
Schmidt HA Strimmer K Vingron M von Haeseler A TREE-PUZZLE: maximum likelihood phylogenetic analysis using quartets and parallel computing. Bioinformatics 2002 18 502 504 11934758 10.1093/bioinformatics/18.3.502
Strimmer K von Haeseler A Likelihood-mapping: a simple method to visualize phylogenetic content of a sequence alignment. Proc Natl Acad Sci USA 1997 94 6815 6819 9192648 10.1073/pnas.94.13.6815
Shimodaira H Hasegawa M Multiple comparisons of log-likelihoods with applications to phylogenetic inference. Mol Biol Evol 1999 16 1114 1116
Saitou N Nei M The neighbor-joining method: a new method for reconstructing phylogenetic trees. Mol Biol Evol 1987 4 406 425 3447015
Fitch WM Margoliash E Construction of phylogenetic trees: a method based on mutation distances as estimated from cytochrome c sequences is of general applicability. Science 1967 155 279 284 5334057
| 15892869 | PMC1175953 | CC BY | 2021-01-04 16:05:38 | no | Genome Biol. 2005 Apr 28; 6(5):R41 | utf-8 | Genome Biol | 2,005 | 10.1186/gb-2005-6-5-r41 | oa_comm |
==== Front
Genome BiolGenome Biology1465-69061465-6914BioMed Central London gb-2005-6-5-r421589287010.1186/gb-2005-6-5-r42ResearchNanoarchaea: representatives of a novel archaeal phylum or a fast-evolving euryarchaeal lineage related to Thermococcales? Brochier Celine [email protected] Simonetta [email protected] Yvan [email protected] Fabrice [email protected] Patrick [email protected] EA EGEE (Evolution, Génomique, Environnement) Université Aix-Marseille I, Centre Saint-Charles, 3 Place Victor Hugo, 13331 Marseille, Cedex 3, France2 Unite Biologie Moléculaire du Gène chez les Extremophiles, Institut Pasteur, 25 rue du Dr Roux, 75724 Paris Cedex 15, France3 Institut de Génétique et Microbiologie, UMR CNRS 8621, Université Paris-Sud, 91405 Orsay, France2005 14 4 2005 6 5 R42 R42 3 12 2004 10 2 2005 9 3 2005 Copyright © 2005 Brochier 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.
An analysis of the position of Nanoarcheum equitans in the archaeal phylogeny using a large dataset of concatenated ribosomal proteins from 25 archaeal genomes suggests that N. equitans is likely to be the representative of a fast-evolving euryarchaeal lineage.
Background
Cultivable archaeal species are assigned to two phyla - the Crenarchaeota and the Euryarchaeota - by a number of important genetic differences, and this ancient split is strongly supported by phylogenetic analysis. The recently described hyperthermophile Nanoarchaeum equitans, harboring the smallest cellular genome ever sequenced (480 kb), has been suggested as the representative of a new phylum - the Nanoarchaeota - that would have diverged before the Crenarchaeota/Euryarchaeota split. Confirming the phylogenetic position of N. equitans is thus crucial for deciphering the history of the archaeal domain.
Results
We tested the placement of N. equitans in the archaeal phylogeny using a large dataset of concatenated ribosomal proteins from 25 archaeal genomes. We indicate that the placement of N. equitans in archaeal phylogenies on the basis of ribosomal protein concatenation may be strongly biased by the coupled effect of its above-average evolutionary rate and lateral gene transfers. Indeed, we show that different subsets of ribosomal proteins harbor a conflicting phylogenetic signal for the placement of N. equitans. A BLASTP-based survey of the phylogenetic pattern of all open reading frames (ORFs) in the genome of N. equitans revealed a surprisingly high fraction of close hits with Euryarchaeota, notably Thermococcales. Strikingly, a specific affinity of N. equitans and Thermococcales was strongly supported by phylogenies based on a subset of ribosomal proteins, and on a number of unrelated molecular markers.
Conclusion
We suggest that N. equitans may more probably be the representative of a fast-evolving euryarchaeal lineage (possibly related to Thermococcales) than the representative of a novel and early diverging archaeal phylum.
==== Body
Background
Despite a ubiquitous distribution [1] and a diversity that may parallel that of the Bacteria (for a recent review see [2]), the Archaea still remain the most unexplored of life's domains. Whereas 21 different phyla are identified in the Bacteria (National Center for Biotechnology Information (NCBI) Taxonomy Database, as of October 2004 [3]), known cultivable archaeal species fall into only two distinct phyla - the Crenarchaeota and the Euryarchaeota [4] - on the basis of small subunit rRNA (SSU rRNA) (NCBI Taxonomy Database, as of October 2004 [3]). A number of non-cultivated species that do not group with either Crenarchaeota or Euryarchaeota have been tentatively assigned to a third phylum, the Korarchaeota [5]. However, this group may be artefactual, as well as that formed by other environmental 16S rRNA sequences [2].
The Crenarchaeota/Euryarchaeota divide indicated by SSU rRNA phylogenies is strongly supported by comparative genomics, as a number of genes present in euryarchaeal genomes are missing altogether in crenarchaeal ones and vice versa. These differences are not trivial, as they involve key proteins involved in DNA replication, chromosome structure and replication. For example, the Crenarchaeota lack both DNA polymerases of the D family and eukaryotic-like histones, which are present in the Euryarchaeota [6,7]. Similarly, replication protein RPA and cell-division protein FtsZ remain exclusive to the Euryarchaeota [8], while only the Crenarchaeota harbor the ribosomal protein S30 (COG4919). This suggests that members of these two archaeal subdomains may employ critically different molecular strategies for key cellular processes. The distinctiveness of the phyla Euryarchaeaota and Crenarchaeota is further strengthened by phylogenetic analysis ([9,10] and this work) and is likely to remain unaffected even when additional cultivable species will be defined. Such a dramatic split is intriguing as it may be more profound than that separating the different bacterial phyla and leaves open different scenarios for the origin of these important differences during early archaeal evolution.
Karl Stetter and his colleagues recently described a novel archaeal species - Nanoarchaeum equitans - representing the smallest known living cell [11]. This tiny hyperthermophile grows and divides at the surface of crenarchaeal Ignicoccus species and cannot be cultivated independently, indicating an obligate symbiotic, and possibly parasitic, life style [12]. Sequencing of the N. equitans genome revealed the smallest cellular genome presently known (480 kb) and raised fascinating questions regarding the origin and evolution of this archaeon [13]. Indeed, in contrast to typical genomes from parasitic/symbiotic microbes [14-16], that of N. equitans does not show any evidence of decaying genes and contains a full complement of tightly packed genes encoding informational proteins [13]. This suggests that the establishment of the dependence-relationship between N. equitans and Ignicoccus is probably very ancient. In a phylogeny of 14 archaeal taxa based on a concatenation of 35 ribosomal proteins and rooted by eukaryotic sequences, N. equitans emerged as the first archaeal lineage, that is, before the divergence of the two main archaeal phyla, the Euryarchaeota and the Crenarchaeota [13]. This is consistent with the early emergence of N. equitans in a phylogeny based on SSU rRNA [12], and with the proposal that N. equitans should be considered as the representative of a novel and very ancient archaeal phylum, the Nanoarchaeota [11].
Testing the phylogenetic position of N. equitans is thus crucial to deciphering the history of the archaeal domain. For instance, if the divergence of this lineage indeed preceded the divergence of Euryarchaeota and Crenarchaeota, features common to N. equitans and any other archaeal taxa could probably be considered as ancestral characters (provided that lateral gene transfers (LGTs) are excluded). For example, the most parsimonious interpretation for the presence in the genome of N. equitans of all those genes that are otherwise found in the Euryarchaeota only [13] is that all these proteins were present in the last archaeal ancestor and were subsequently lost in the Crenarchaeota. However, the hypothesis of an early divergence of the Nanoarchaeota should be treated with caution. There are now several examples in which fast-evolving taxa are mistakenly assigned to early branches because of a long branch attraction (LBA) artifact due to their high evolutionary rates [17], especially when a distant outgroup is used [18-21]. Similarly, since adaptation to a symbiotic or parasitic life style may have accelerated its evolutionary rate, the basal position of N. equitans in phylogenetic analyses using distant eukaryotic sequences as the outgroup [13] may be strongly affected by LBA.
We tested the position of N. equitans in the archaeal phylogeny by using a dataset of concatenated ribosomal proteins larger than that used by Waters and colleagues [13], a much broader taxonomic sampling, and without including any outgroup in order to reduce LBA. By applying phylogenetic approaches that accurately handle reconstruction biases, we show that the early emergence of N. equitans observed in previous analyses probably resulted from an LBA artifact due to the fast evolutionary rate of this archaeon, possibly worsened by LGT affecting a fraction of its ribosomal proteins. Indeed, the phylogenies based on our new ribosomal protein dataset and on additional single genes suggest that N. equitans is more likely to be a very divergent euryarchaeon - possibly a sister lineage of Thermococcales - than a new and ancestral archaeal phylum. This is consistent with further evidence gathered from close BLAST hits analyses on the whole genome complement of this taxon.
Results and discussion
Phylogenetic analysis of concatenated ribosomal proteins
Fifty ribosomal proteins having a sufficient taxonomic sampling and for which no LGT were evidenced in previous analyses (see Materials and methods and Table 1) [9,10] were concatenated into a large dataset (F1 dataset) comprising 6,384 positions and 25 archaeal taxa. The datasets contained 18 taxa previously used for the study of archaeal phylogeny based on ribosomal proteins [10] plus seven new taxa: the Thermococcale Thermococcus gammatolerans, the Methanomicrobiale Methanogenium frigidum, the Methanosarcinales Methanococcoides burtonii, Methanosarcina mazei and Methanosarcina acetivorans, the halobacterium Haloferax volcanii and N. equitans. Exhaustive maximum likelihood searches were performed with a Jones Taylor Thornton (JTT) model and limited constraints on indisputable nodes as recovered in unconstrained maximum likelihood and neighbor-joining analyses (data not shown) and in previous work [10].
The corresponding maximum likelihood unrooted tree is shown in Figure 1a. The monophyly of the two main archaeal domains, Crenarchaeota and Euryarchaeota, was recovered and supported by high bootstrap values (BV) (100% and 98%, respectively). Within the Euryarchaeota, the basal branching of Thermococcales (including T. gammatolerans) was also recovered (BV = 84%) as was the group comprising Methanobacteriales and Methanococcales (BV = 64%), and a well sustained group (BV = 96%) comprising Thermoplasmatales, Archaeoglobales, Halobacteriales (including H. volcanii) and Methanomicrobia (including the three new members of the Methanosarcinales M. acetivorans, M. mazei, M. burtonii and the Methanomicrobiale M. frigidum). N. equitans emerged as a separate branch distinct from those leading to Crenarchaeota and Euryarchaeota, in agreement with the rooted phylogeny of Waters and colleagues [13]. However, in our analysis the branch leading to N. equitans was relatively long, suggesting a possible above-average substitution rate with respect to the other taxa in the dataset that may affect its correct placement. Consequently, in order to identify the origin of possible biases in our global analysis, we analyzed two additional fusion datasets, one including the 27 proteins of the F1 dataset belonging to the large ribosomal subunit (F2 dataset) and one including the 23 proteins of the F1 dataset belonging to the small ribosomal subunit (F3 dataset).
The F2 tree (Additional data file 1A) was highly consistent with the F1 tree (Figure 1a) including the placement of N. equitans on a separate branch with respect to the other two archaeal domains. In contrast, in the F3 tree (Additional data file 1B), N. equitans emerged within the Euryarchaeota with a high statistical confidence (BV = 98%) and was supported - albeit weakly - as sister group of the Thermococcales (BV = 54%). This indicates that the components of the two ribosomal subunits may harbor a conflicting signal for the placement of N. equitans. Such incongruence was unexpected and led us to question the reliability of global ribosomal protein fusions in the assignment of the correct phylogenetic position of N. equitans in the archaeal phylogeny.
Phylogenetic analyses of individual ribosomal proteins
To further characterize the conflicting phylogenetic signal for the placement of N. equitans in our concatenated analyses, we investigated its position in individual trees obtained by both unconstrained maximum likelihood and Bayesian analysis of each of the 50 ribosomal proteins. The topologies of these trees were consistent overall, despite the weakness of the phylogenetic signal contained in individual ribosomal proteins, often of small size. N. equitans generally displayed above-average branch lengths in these phylogenies, reinforcing the idea that LBA may strongly bias its placement in the global fusion trees. Moreover, N. equitans showed a highly unstable position (Table 1). In fact, it emerged as a separate branch distinct from the crenarchaeal and euryarchaeal domains (as in the F1 and F2 trees, Additional data file 1A), in only seven ribosomal protein phylogenies.
This is at odds with the indication of N. equitans as the representative of a novel archaeal domain, as Euryarchaeota and Crenarchaeota were generally well segregated in these individual phylogenies (data not shown). In contrast, as many as 33 ribosomal proteins supported the inclusion of N. equitans within the Euryarchaeota, 13 of which indicated a sister grouping with Thermococcales, similarly to the small ribosomal subunit protein tree (F3, Additional data file 1B). This striking affiliation may be explained by the occurrence of massive LGT involving these proteins between N. equitans and other euryarchaeal lineages. However, as no specific ecological reasons may especially favor such exchanges, this would rather indicate N. equitans as a euryarchaeal phylum rather than a novel archaeal domain. Conversely, LGT could easily explain the grouping of N. equitans with Crenarchaeota in the individual trees of nine ribosomal proteins (Table 1), as the genes coding for these proteins in N. equitans may have been acquired from its crenarchaeal host Ignicoccus species. If confirmed by future analyses, especially once the complete genome sequence of the Ignicoccus species is available, this would be the first report of numerous LGTs involving ribosomal proteins between two archaeal species.
It is worth noting that five of the nine proteins grouping N. equitans with Crenarchaeota belong to the large ribosomal subunit, and may introduce a strong bias for the basal position of N. equitans in the F2 tree (Additional data file 1A), as well as in the F1 tree (Figure 1a). To test this, we constructed a fourth dataset (F4 dataset) by removing these nine ribosomal proteins from the F1 dataset, and the resulting maximum likelihood tree is shown in Figure 1b. Strikingly, the F4 tree was highly consistent with the F1 tree, except for the position of N. equitans, which was strongly assigned to Euryarchaeota (BV = 100%) and branched off as a sister lineage of Thermococcales (BV = 60%), similarly to the small ribosomal subunit protein tree (F3, Additional data file 1B). Importantly, this placement is not likely to be the result of an LBA between the branch leading to N. equitans and that leading to Thermococcales, since the latter was rather short (Figure 1b). Our results strongly suggest that the basal position of N. equitans observed in our global ribosomal protein fusion analysis (Figure 1a) and in others [13] could resulted from the combination of conflicting phylogenetic signal from different subsets of ribosomal proteins (Table 1), either due to LGT and/or to LBA given the relatively fast evolutionary rates displayed by this taxon. Instead, once these biases are reduced, N. equitans shows a weak but specific affinity to Thermococcales (Figure 1b) that may represent its genuine placement in the archaeal phylogeny.
Phylogenetic pattern of N. equitans protein complement
We investigated whether the difficulty of assigning the ribosomal proteins of N. equitans to a clear phylogenetic status reflected a general characteristic of the whole protein complement of this taxon. With this aim, we performed a complete survey of all 563 open reading frames (ORFs) encoded in the N. equitans genome by BLASTP searches against all other available complete archaeal genomes (including T. gammatolerans). Although a close hit does not always correspond to the nearest phylogenetic neighbor [22], a genome-scale analysis of the distribution of such hits can highlight interesting patterns. We have chosen not to extend this analysis further by automated molecular phylogeny reconstructions because we reckon that such an approach is highly prone to error. Indeed, dataset assembly is strictly dependent on human judgment at critical steps such as choice of homologs and alignment editing.
The distribution of close hits for the N. equitans ORFs according to an E-value cutoff of 10-4 is shown in Figure 2a. Thresholds between 10-2 and 10-10 either increased or decreased the proportion of N. equitans-specific genes, but did not significantly change the relative distribution of close BLAST hits between archaeal groups (data not shown). A third of the N. equitans ORFs appeared to have no homologs in other archaea (gray section in Figure 2a), consistent with a previous analysis [13]. However, the remaining ORFs displayed many more close hits with different euryarchaeal lineages (56%) than with crenarchaeal ones (12%) (Figure 2a). Strikingly, nearly half of the euryarchaeal close hits (approximately 25% of the N. equitans ORFs) were represented by Thermococcales (green section in Figure 2a).
To identify possible biases introduced by LGT, we determined the global distribution of the second, third and fourth close BLAST hits (Figure 2b). Fifty percent of N. equitans close hits were indeed represented exclusively by members of different euryarchaeal phyla (green section in Figure 2b), and this proportion was even higher when we included ORFs with a crenarchaeon as close hit, but euryarchaeal species as next three close hits, suggesting possible Euryarchaeota-to-Crenarchaeota LGT (pale-green section in Figure 2b). Such a high fraction of close hits with the Euryarchaeota may be due to the effect of overall higher evolutionary rates in Crenarchaeota, although this has never been proposed. This unexpected high proportion of best close hits with Euryarchaeota - and notably Thermococcales - for the proteins of N. equitans is strikingly consistent with the phylogenetic analyses of individual (Table 1) and concatenated (Figure 1b and Additional data file 1B) ribosomal proteins, further suggesting that N. equitans may be a divergent euryarchaeon related to Thermococcales.
Additional single-gene phylogenies
To test further the phylogenetic position of N. equitans, we performed single-gene analyses by both maximum likelihood and Bayesian approaches of additional proteins known to be potential good molecular markers. Two unrooted archaeal maximum likelihood trees based on the elongation factors EF-1α and EF-2 are shown in Figure 3a and 3b, respectively. Strikingly, both trees strongly placed N. equitans within the Euryarchaeota (BV = 100% and a posterior probability (PP) of 1.00), and specifically as a sister-group of Thermococcales (BV = 79%, and PP = 1.00 and BV = 64% and PP = 1.00 in EF-1α and EF-2 trees, respectively), consistently with the F3 and F4 trees (Additional data file 1B and Figure 1b, respectively). The inclusion of N. equitans within the Euryarchaeota in the phylogeny based on EF-1α is further supported by an insertion/deletion (indel)-containing region that displays identical structure in N. equitans and several euryarchaeal lineages including Thermococcales (data not shown). These results may be interpreted by positing the concerted LGT of EF-1α and EF-2 from Thermococcales to N. equitans, since the two factors are part of the same macromolecular complex.
Thus, we analyzed additional markers involved in different molecular functions, such as the A subunit of topoisomerase VI, a type IIB DNA topoisomerase involved in DNA replication and whose phylogeny is highly consistent with that based on 16S rRNA [23]. The resulting tree (Figure 3c) was largely congruent with the previous ones, and once more placed N. equitans as sister-group of Thermococcales (BV = 98%, PP = 1.00), within the Euryarchaeota (BP = 100%, PP = 1.00). Finally, we investigated the position of N. equitans in an archaeal phylogeny based on reverse gyrase, a key enzyme composed of two domains, a helicase and a topoisomerase [24] and specific to thermophiles, where it catalyzes DNA positive supercoiling [25]. In N. equitans the gene encoding reverse gyrase is split into two noncontiguous coding sequences encoding the helicase and topoisomerase functions, respectively [13]. This has been taken as evidence for an ancestral nature of the reverse gyrase gene of N. equitans, consistent with the supposedly early emergence of this taxon [13]. However, the phylogeny of reverse gyrase (Figure 3d) supports a late branching of N. equitans, and surprisingly once more grouped with Thermococcales (BV = 60% and PP = 1.00). This suggests that the fission of the reverse gyrase gene in N. equitans probably resulted from a secondary event. Indeed, a high number of split genes appear to be a general feature of the N. equitans genome [13], as well as of those of fast-evolving archaeal taxa, such as Methanopyrus kandleri [26].
Conclusion
The description of N. equitans by Huber and colleagues little more than two years ago marked an important step in our knowledge of the diversity and evolution of the Archaea, still the most unexplored of life's three domains. Indeed, N. equitans represents an example of symbiotic/parasitic life style between two archaeal species that is unprecedented [11,12]. The exceptionality of this archaeon was confirmed by the sequencing of its genome, which combines a minimal size close to the theoretical limits of a living cell with a stability not observed in other highly reduced genomes [13].
Despite all these characters indicating N. equitans as the member of a highly divergent lineage, we feel that its assignment to a novel archaeal phylum - the Nanoarchaeota - other than the well established Euryarchaeota and Crenarchaeota may be premature. Indeed, the distinctiveness of the N. equitans SSU rRNA primary structure may be an idiosyncrasy of this taxon due to a unique combination of adaptation to hyperthermophily and genome reduction. Our phylogenetic analyses of ribosomal proteins consistently show that N. equitans does not behave like the Euryarchaeota or the Crenarchaeota, which generally form clearly distinct branches in the archaeal tree, but shows instead a highly unstable placement. Similarly, the suggestion that N. equitans may represent an ancient divergence in the archaeal domain is far from being settled. In fact, the branching point of N. equitans is largely unresolved in the SSU rRNA phylogeny [12], and its basal placement in a recent tree of a ribosomal protein concatenation may be biased by the attraction of the long branches leading to N. equitans and to the eukaryotic sequences used as the outgroup [13]. Indeed, our unrooted phylogenies underline the above-average evolutionary rate of N. equitans and warn against the unreliability of global ribosomal protein fusions in assessing the correct placement of this taxon, because of LBA. Moreover, an additional bias may be introduced by LGT, as we suggest that a substantial fraction of N. equitans ribosomal proteins may have been exchanged with its crenarchaeal host. Our results indeed indicate an unsuspected close affinity of N. equitans with the Euryarchaeota, and notably with Thermococcales. This evidence is strongly reinforced by the specific and strong affinity of N. equitans with Thermococcales in trees of diverse molecular markers that do not lie in close proximity in the N. equitans genome, and on close BLAST hit analyses on the whole genome complement of this taxon. To explain all these findings, the most parsimonious explanation would be that N. equitans is a highly divergent euryarchaeal lineage possibly related to Thermococcales.
The hypothesis of nanoarchaea being a euryarchaeal lineage has important implications for our understanding of archaeal evolution, as characters in common between N. equitans and Euryarchaeota could be more easily considered as synapomorphies of the group rather than ancestral traits that would have been lost in the branch leading to Crenarchaeota. The characterization and genomic analysis of additional nanoarchaeal species will be necessary to confirm a specific affinity to Thermococcales, and to shed further light on the evolution of this intriguing group of archaea.
Materials and methods
Sequence retrieval and dataset construction
We updated a dataset of 62 ribosomal proteins from previous work [9,10]. In addition to N. equitans [11], we included six new taxa: two Methanosarcinales (Methanosarcina mazei [27] and Methanosarcina acetivorans [28]) whose complete genomes have been recently made available in public databases [29,30], and four other archaeal species whose genome sequencing is under way, that is, the Methanomicrobiale Methanogenium frigidum [31], the Methanosarcinale Methanococcoides burtonii [32], the Halobacteriale Haloferax volcanii [33], and the Thermococcale Thermococcus gammatolerans [34] (Y.Z. and F.C., unpublished work). Sequences were retrieved using BLASTP [35] at NCBI for N. equitans, M. acetivorans and M. mazei, and by TBLASTN [35] at the genome-sequencing website for H. volcanii [36], and at the draft genome analysis website [37] for M. burtonii [38] and for M. frigidum [38]. Unlike Waters and colleagues [13], and like our previous studies [9,10], we did not include any eukaryotic outgroup, in order to prevent LBA. Novel sequences were manually added to previous alignments [39] and ambiguous regions were removed.
Single alignment datasets were constructed for each of the 62 ribosomal proteins. From these, four concatenated datasets were constructed: one including 50 ribosomal proteins for which no LGT was evidenced in previous analyses and had a sufficient taxonomic sampling (at least 21 taxa) (F1 dataset); one including the 27 proteins from the F1 dataset belonging to the large ribosomal subunit (F2 dataset); one including the 23 proteins from the F1 dataset belonging to the small ribosomal subunit (F3 dataset); and one corresponding to the F1 dataset excluding nine ribosomal proteins supporting a close relationship between N. equitans and the Crenarchaeota (see Results and discussion) (F4 dataset). Four additional single alignment datasets were similarly constructed for the two elongation factors EF-1α and EF-2, the A subunit of topoisomerase VI (TopoVIa), and reverse gyrase.
Phylogenetic analyses
To handle rate variation among sites, maximum likelihood-distance matrices (JTT model with a Gamma-law and eight discrete classes) were computed with TREE-PUZZLE [40] and used for neighbor-joining tree reconstruction by the NEIGHBOR program of the PHYLIP package [41]. Unconstrained maximum likelihood trees were computed using PHYML and the same parameters [42]. Bayesian phylogenetic trees were constructed using MrBayes [43] with a mixed model of amino-acid substitution and a Gamma-law (eight discrete classes). MrBayes was run with four chains for 1 million generations and trees were sampled every 100 generations. Exhaustive maximum likelihood searches were performed using the PROTML program of the MOLPHY package [44] with a JTT model and limited constraints on indisputable nodes as recovered in unconstrained maximum likelihood and neighbor-joining analyses and previous work [10]. Branch lengths and likelihoods for the 2,000 top-ranking topologies were computed using a JTT model including a Gamma-law and eight discrete classes with TREE-PUZZLE [40]. Bootstrap analyses were performed on 1,000 replicates using PUZZLEBOOT [45] and extended majority rule consensus trees were inferred with CONSENSE from the PHYLIP package [46]. All datasets and corresponding phylogenetic trees are available on request from C.B.
Close BLAST hit analyses
All the ORFs of the N. equitans genome were retrieved from NCBI. For each ORF a BLASTP search was performed locally on a database of complete archaeal genomes including T. gammatolerans. Different distributions of close BLAST hits were manually established with E-value threshold cutoffs ranging from 10-2 to 10-10. The same criteria were used to establish additional distributions including information from the next three close-hit representatives of different phyla. For example, when the first six close hits were represented by T. gammatolerans, Pyrococcus abyssi, P. horikoshii, P. furiosus, M. kandleri and Sulfolobus solfataricus, we considered as three first close BLAST hits Thermococcales, Methanopyrales and Sulfolobales.
Additional data files
Additional data are available with the online version of this article. Additional data file 1 contains a figure showing unrooted unconstrained maximum likelihood trees computed by PHYML from a concatenation of large subunit and small subunit ribosomal proteins.
Supplementary Material
Additional File 1
Numbers at nodes are bootstrap values. Scale bars represent the number of changes per position for a unit branch length
Click here for file
Acknowledgements
We thank Eric Armanet and Gael Stefan for allowing part of calculations on their computers. We thank also Shiladitya DasSarma and the members of the University of Scranton, PA, for the sequences of H. volcanii freely available by BLAST [36].
Figures and Tables
Figure 1 Unrooted maximum likelihood trees from exhaustive searches based on the F1 and the F2 datasets. (a) F1 dataset; (b) F2 dataset. Numbers at nodes are bootstrap values. Scale bars represent the number of changes per position for a unit branch length. Asterisks indicate constrained nodes.
Figure 2 Distribution of close BLASTP hits. Hits are displayed as (a) per lineage and (b) per archaeal domain of the 563 ORFs of the N. equitans genome with a threshold of 10-4.
Figure 3 Phylogenetic trees for elongation factors EF-1α and EF-2, subunit A of topoisomerase VI and reverse gyrase. Unconstrained unrooted maximum likelihood trees of (a) elongation factor EF-1α, (b) elongation factor EF-2, (c) subunit A of topoisomerase VI, and (d) Bayesian tree of reverse gyrase. Bold numbers at nodes are bootstrap values; the other numbers are the Bayesian posterior probabilities. Scale bars represent the number of changes per position for a unit branch length.
Table 1 Position of Nanoarchaeum equitans in maximum likelihood and Bayesian phylogenies of individual ribosomal proteins
Position of N. equitans Proteins Total
Basal position L3, L10, L11, L31e, S5, S19e, S24e 7
Within Crenarchaeota and sister group to: 9
Sulfolobales L16, L18e, L23 3
Aeropyrum pernix S17e 1
Pyrobaculum aerophilum L6, L20a, L29, S6e 4
Other S10 1
Within Euryarchaeota and sister group of: 33
Thermococcales L1, L2, L14, L15, L21e, L24, L32e, L37e, S3, S7, S17, S19, S28e 13
Methanopyrus kandleri L4, L13, S13 3
Methanococcales L18 1
Methanothermobacter thermautotrophicus S4, S11 2
Archaeoglobus fulgidus S8e, S9 2
Thermoplasmatales L22, L30, S2, S3ae, S15 5
Methanomicrobiales S8 1
Halobacteriales S27a 1
Other L5, L19e, L24e, S4e, S27e 6
Absent in N. equitans L39e 1
==== Refs
Karner MB DeLong EF Karl DM Archaeal dominance in the mesopelagic zone of the Pacific Ocean. Nature 2001 409 507 510 11206545 10.1038/35054051
Forterre P Brochier C Philippe H Evolution of the Archaea. Theor Popul Biol 2002 6 409 422 10.1006/tpbi.2002.1592
NCBI Taxonomy Database
Woese CR Kandler O Wheelis ML Towards a natural system of organisms: proposal for the domains Archaea, Bacteria, and Eucarya. Proc Natl Acad Sci USA 1990 87 4576 4579 2112744
Barns SM Delwiche CF Palmer JD Pace NR Perspectives on archaeal diversity, thermophily and monophyly from environmental rRNA sequences. Proc Natl Acad Sci USA 1996 93 9188 9193 8799176 10.1073/pnas.93.17.9188
Uemori T Sato Y Kato I Doi H Ishino Y A novel DNA polymerase in the hyperthermophilic archaeon, Pyrococcus furiosus : gene cloning, expression, and characterization. Genes Cells 1997 2 499 512 9348040 10.1046/j.1365-2443.1997.1380336.x
Bell SD Jackson SP Mechanism and regulation of transcription in archaea. Curr Opin Microbiol 2001 4 208 13 11282478 10.1016/S1369-5274(00)00190-9
Myllykallio H Lopez P Lopez-Garcia P Heilig R Saurin W Zivanovic Y Philippe H Forterre P Bacterial mode of replication with eukaryotic-like machinery in a hyperthermophilic archaeon. Science 2000 288 2212 2215 10864870 10.1126/science.288.5474.2212
Matte-Tailliez O Brochier C Forterre P Philippe H Archaeal phylogeny based on ribosomal proteins. Mol Biol Evol 2002 19 631 639 11961097
Brochier C Forterre P Gribaldo S Archaeal phylogeny based on proteins of the transcription and translation machineries: tackling the Methanopyrus kandleri paradox. Genome Biol 2004 5 R17 15003120 10.1186/gb-2004-5-3-r17
Huber H Hohn MJ Rachel R Fuchs T Wimmer VC Stetter KO A new phylum of Archaea represented by a nanosized hyperthermophilic symbiont. Nature 2002 417 63 67 11986665 10.1038/417063a
Huber H Hohn MJ Stetter KO Rachel R The phylum Nanoarchaeota: present knowledge and future perspectives of a unique form of life. Res Microbiol 2003 154 165 171 12706504 10.1016/S0923-2508(03)00035-4
Waters E Hohn MJ Ahel I Graham DE Adams MD Barnstead M Beeson KY Bibbs L Bolanos R Keller M The genome of Nanoarchaeum equitans : insights into early archaeal evolution and derived parasitism. Proc Natl Acad Sci USA 2003 100 12984 12988 14566062 10.1073/pnas.1735403100
Silva FJ Latorre A Moya A Genome size reduction through multiple events of gene disintegration in Buchnera APS. Trends Genet 2001 17 615 618 11672844 10.1016/S0168-9525(01)02483-0
Moran NA Tracing the evolution of gene loss in obligate bacterial symbionts. Curr Opin Microbiol 2003 6 512 518 14572545 10.1016/j.mib.2003.08.001
Andersson JO Andersson SG Genome degradation is an ongoing process in Rickettsia. Mol Biol Evol 1999 16 1178 1191 10486973
Felsenstein J Cases in which parsimony or compatibility methods will be positively misleading. Syst Zool 1978 27 401 410
Hirt RP Logsdon JM JrHealy B Dorey MW Doolittle WF Embley TM Microsporidia are related to fungi: evidence from the largest subunit of RNA polymerase II and other proteins. Proc Natl Acad Sci USA 1999 96 580 585 9892676 10.1073/pnas.96.2.580
Dacks JB Marinets A Ford Doolittle W Cavalier-Smith T Logsdon JM Jr Analyses of RNA polymerase II genes from free-living protists: phylogeny, long branch attraction, and the eukaryotic big bang. Mol Biol Evol 2002 19 830 840 12032239
Philippe H Lopez P Brinkmann H Budin K Germot A Laurent J Moreira D Müller M Le Guyader H Early branching or fast evolving eukaryotes? An answer based on slowly evolving positions. Phil Trans R Soc Lond B Biol Sci 2000 267 1213 1221
Gribaldo S Philippe H Ancient phylogenetic relationships. Theor Popul Biol 2002 61 391 408 12167360 10.1006/tpbi.2002.1593
Koski LB Golding GB The closest BLAST hit is often not the nearest neighbor. J Mol Evol 2001 52 540 542 11443357
Gadelle D Filee J Buhler C Forterre P Phylogenomics of type II DNA topoisomerases. BioEssays 2003 25 232 242 12596227 10.1002/bies.10245
Krah R Kozyavkin SA Slesarev AI Gellert M A two-subunit type I DNA topoisomerase (reverse gyrase) from an extreme hyperthermophile. Proc Natl Acad Sci USA 1996 93 106 110 8552584 10.1073/pnas.93.1.106
Forterre P A hot story from comparative genomics: reverse gyrase is the only hyperthermophile-specific protein. Trends Genet 2002 18 236 237 12047940 10.1016/S0168-9525(02)02650-1
Slesarev AI Mezhevaya KV Makarova KS Polushin NN Shcherbinina OV Shakhova VV Belova GI Aravind L Natale DA Rogozin IB The complete genome of hyperthermophile Methanopyrus kandleri AV19 and monophyly of archaeal methanogens. Proc Natl Acad Sci USA 2002 99 4644 4649 11930014 10.1073/pnas.032671499
Mah RA Isolation and characterization of Methanococcus mazei. Curr Microbiol 1980 3 321 325
Sowers KR Baron SF Ferry JG Methanosarcina acetivorans sp. nov., an acetotrophic methane-producing bacterium isolated from marine sediments. Appl Environ Microbiol 1984 47 971 978 16346552
Deppenmeier U Johann A Hartsch T Merkl R Schmitz RA Martinez-Arias R Henne A Wiezer A Baumer S Jacobi C The genome of Methanosarcina mazei : evidence for lateral gene transfer between bacteria and archaea. J Mol Microbiol Biotechnol 2002 4 453 461 12125824
Galagan JE Nusbaum C Roy A Endrizzi MG Macdonald P FitzHugh W Calvo S Engels R Smirnov S Atnoor D The genome of M. acetivorans reveals extensive metabolic and physiological diversity. Genome Res 2002 12 532 542 11932238 10.1101/gr.223902
Franzmann PD Liu Y Balkwill DL Aldrich HC Conway de Macario E Boone DR Methanogenium frigidum sp. nov., a psychrophilic, H2-using methanogen from Ace Lake, Antarctica. Int J Syst Bacteriol 1997 47 1068 1072 9336907
Franzmann PD Springer N Ludwig W Conway de Macario E Rohde M A methanogenic archaeon from Ace Lake, Antarctica: Methanococcoides burtonii sp. nov. Syst Appl Microbiol 1992 15 573 581
Torreblanca M Rodriguez-Valera F Juez G Ventosa A Kamekura M Kates M Classification of non-alkaliphilic halobacteria based on numerical taxonomy and polar lipid composition, and description of Haloarcula gen. nov and Haloferax gen. nov. Syst Appl Microbiol 1986 8 89 99
Jolivet E L'Haridon S Corre E Forterre P Prieur D Thermococcus gammatolerans sp. nov., a hyperthermophilic archaeon from a deep-sea hydrothermal vent that resists ionizing radiation. Int J Syst Evol Microbiol 2003 53 847 851 12807211 10.1099/ijs.0.02503-0
Altschul SF Gish W Miller W Myers EW Lipman DJ Basic local alignment search tool. J Mol Biol 1990 215 403 410 2231712 10.1006/jmbi.1990.9999
Haloferax volcanii genome web site
Draft genome analysis of Methanogenium frigidum and Methanococcoides burtonii
Saunders NF Thomas T Curmi PM Mattick JS Kuczek E Slade R Davis J Franzmann PD Boone D Rusterholtz K Mechanisms of thermal adaptation revealed from the genomes of the Antarctic Archaea Methanogenium frigidum and Methanococcoides burtonii. Genome Res 2003 13 1580 1588 12805271 10.1101/gr.1180903
Adachi J Hasegawa M Phylogeny of whales: dependence of the inference on species sampling. Mol Biol Evol 1995 12 177 179 7877493
Schmidt HA Strimmer K Vingron M von Haeseler A TREE-PUZZLE: maximum likelihood phylogenetic analysis using quartets and parallel computing. Bioinformatics 2002 18 502 504 11934758 10.1093/bioinformatics/18.3.502
Felsenstein J Phylogeny Inference Package (Version 3.2). Cladistics 1989 5 164 166
Guindon S Gascuel O A simple, fast, and accurate algorithm to estimate large phylogenies by maximum likelihood. Syst Biol 2003 52 696 704 14530136 10.1080/10635150390235520
Rönner S Liesack W Wolters J Stackebrandt E Cloning and sequencing of a large fragment of the ATPD gene of Pirellula marine - a contribution to the phylogeny of Planctomycetales. Endocyt Cell Res 1991 7 219 229
Adachi J Hasegawa M MOLPHY version 2.3: programs for molecular phylogenetics based on maximum likelihood. Comput Sci Monogr 1996 28 1 150
Holder ME Roger AJ A shell-script program called "puzzleboot" that allows the analysis of multiple data sets with PUZZLE even though PUZZLE lacks the "M" option of many PHYLIP programs 2002
J Felsenstein PHYLIP (Phylogeny Inference Package) version 3.6 2004
| 15892870 | PMC1175954 | CC BY | 2021-01-04 16:05:39 | no | Genome Biol. 2005 Apr 14; 6(5):R42 | utf-8 | Genome Biol | 2,005 | 10.1186/gb-2005-6-5-r42 | oa_comm |
==== Front
Genome BiolGenome Biology1465-69061465-6914BioMed Central London gb-2005-6-5-r431589287110.1186/gb-2005-6-5-r43ResearchDissection of a DNA-damage-induced transcriptional network using a combination of microarrays, RNA interference and computational promoter analysis Elkon Ran [email protected] Sharon [email protected] Yaniv [email protected] Chaim [email protected] Tamar [email protected] Ninette [email protected] Gideon [email protected] Ron [email protected] Yosef [email protected] The David and Inez Myers Laboratory for Genetic Research, Department of Human Genetics, Sackler School of Medicine, Tel Aviv University, Tel Aviv, 69978, Israel2 School of Computer Science, The Chaim Sheba Medical Center and Sackler School of Medicine, Tel Aviv University, Tel Aviv, 69978, Israel3 Department of Pediatric Hemato-Oncology and Functional Genomics, The Chaim Sheba Medical Center and Sackler School of Medicine, Tel Aviv University, Tel Aviv, 69978, Israel2005 13 4 2005 6 5 R43 R43 29 12 2004 3 2 2005 8 3 2005 Copyright © 2005 Elkon 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.
Microarray and RNAi technologies were applied to dissect a transcriptional network induced by DNA damage in human cells, revealing that two pivotal stress-induced transcription factors (NFκB and p53) mediated most of the damage-induced gene activation while a major transducer of the cellular responses to double strand breaks (ATM) was required for the activation of both pathways.
Background
Gene-expression microarrays and RNA interferences (RNAi) are among the most prominent techniques in functional genomics. The combination of the two holds promise for systematic, large-scale dissection of transcriptional networks. Recent studies, however, raise the concern that nonspecific responses to small interfering RNAs (siRNAs) might obscure the consequences of silencing the gene of interest, throwing into question the ability of this experimental strategy to achieve precise network dissections.
Results
We used microarrays and RNAi to dissect a transcriptional network induced by DNA damage in a human cellular system. We recorded expression profiles with and without exposure of the cells to a radiomimetic drug that induces DNA double-strand breaks (DSBs). Profiles were measured in control cells and in cells knocked-down for the Rel-A subunit of NFκB and for p53, two pivotal stress-induced transcription factors, and for the protein kinase ATM, the major transducer of the cellular responses to DSBs. We observed that NFκB and p53 mediated most of the damage-induced gene activation; that they controlled the activation of largely disjoint sets of genes; and that ATM was required for the activation of both pathways. Applying computational promoter analysis, we demonstrated that the dissection of the network into ATM/NFκB and ATM/p53-mediated arms was highly accurate.
Conclusions
Our results demonstrate that the combined experimental strategy of expression arrays and RNAi is indeed a powerful method for the dissection of complex transcriptional networks, and that computational promoter analysis can provide a strong complementary means for assessing the accuracy of this dissection.
==== Body
Background
With completion of the sequencing of the human genome and those of many other organisms, research is shifting to functional genomics, that is, to gaining system-level understanding of the mechanisms by which gene products interact and regulate each other to produce coherent and coordinated physiological processes during normal development and in response to homeostatic challenges. Great progress has been made in the delineation of transcriptional regulatory networks [1-4], thanks to the maturation of gene-expression microarrays and the development of advanced computational approaches for analysis of the volumes of data generated by this technology. Another technological breakthrough that greatly enhances the ability to manipulate and characterize gene function in mammalian cells is the use of RNA interference (RNAi) for targeted silencing of specific genes [5-7]. The combination of global gene-expression profiling and RNAi-mediated silencing of key regulatory genes appears to offer a powerful tool for systematic dissection of transcriptional networks. However, recent studies pointed out that applying RNAi to mammalian cells triggers some nonspecific pathways [8-10] and affects an unpredicted number of off-targets [11] in addition to knocking-down the target of interest. This raises concern that nonspecific responses to small interfering RNAs (siRNA) might obscure the consequences of silencing the target of interest.
In this work, focusing on a DNA-damage-induced transcriptional network as a test case, we established human cells stably knocked-down for one of the major activators of the network, the protein kinase ATM (a gene that is mutated in the disease ataxia-telangiectasia), and for two key transcription factors that function downstream to it, NFκB and p53. Comparing gene-expression profiles measured in these cellular systems with and without exposure to a DNA damaging agent, we observed that NFκB and p53 mediated most of the damage-induced gene activation; that they controlled the activation of largely disjoint sets of genes; and that ATM was required for the activation of both pathways. Applying statistical tests coupled with computational promoter analysis, we demonstrated that the dissection of the damage-induced network into ATM/ NFκB - and ATM/p53-mediated arms was highly accurate. Thus, we show that this combined strategy is indeed a powerful method for the dissection of complex transcriptional networks.
Results
We established human cellular systems stably knocked-down for the ATM protein kinase, for the Rel-A subunit of NFκB, and for p53. Stable knock-down of the proteins was obtained by infecting HEK 293 cells with retroviral vectors expressing the corresponding short hairpin RNAs (shRNAs). Efficient reduction of protein levels was confirmed using western blotting analysis (Figure 1). Controls for our experiments were uninfected cells and cells infected with a vector carrying siRNA against lacZ, which has no significant homology to any human gene. Using Affymetrix Human Focus GeneChip arrays, we recorded gene-expression profiles in these cellular systems before and 4 hours after exposure to neocarzinostatin (NCS), an enediyne antitumor antibiotic that intercalates into the DNA and induces double-strand breaks (DSBs) [12]. Our dataset contains profile measurements for ten conditions: five cellular systems (two controls - uninfected cells and cells expressing siRNA against the bacterial enzyme LacZ - and cells knocked-down for Rel-A, p53 and ATM), each probed at two time points: without treatment and 4 hours after exposure to NCS. Each condition was measured in independent triplicates. Expression levels were computed using the Robust Multi-array Average (RMA) method [13] (see Materials and methods).
As a first step in our data analysis we searched for nonspecific responses to siRNA expression. We scanned the dataset for genes that were either consistently up- or downregulated in all four cells expressing siRNAs compared with their basal level in the uninfected control, all before exposure to NCS. We observed a subtle but statistically significant response to viral infection/siRNA expression. Very few genes were consistently responsive when a cutoff of 1.5-fold change was set, but lowering the threshold to 1.3-fold resulted in 20 consistently upregulated and 75 consistently downregulated genes in the infected cells (Additional data file 3). The threshold is low, but the number of genes that showed consistent response is significantly higher than expected by chance (in 1,000 datasets with randomly permutated entries for each gene, an average of 0.1 and 0.2 consistently up- and downregulated genes, respectively, were found). The set of consistently upregulated genes contained mainly genes involved in different aspects of cellular metabolism (Additional data file 2). The consistently downregulated genes included metabolic genes and genes that function in control of cell growth, signal transduction and stress responses (Additional data file 2). In contrast to some reports [8,10], we did not observe induction of the interferon pathway following the introduction of siRNA into the cells.
Next, we searched the dataset for genes that responded to the NCS treatment in the control uninfected cells and whose response was not disturbed by the introduction of siRNA into the cells: namely, genes that responded to the treatment in a coherent manner in the uninfected and the LacZ control cells. This damage-induced gene set (additional data file 4) contained 112 genes that were induced in both controls and met our criterion (see Materials and methods). Only seven genes met an analogous criterion for repression in response to NCS treatment; six of them are related to mitosis, presumably reflecting the activation of cell-cycle checkpoints in response to DNA damage (see Additional data file 4).
We divided the expression level of each damage-induced gene at the 4-hour time point by its level in untreated cells in the same cellular system, and subjected the data to hierarchical cluster analysis. The damage-induced gene set was found to fall into four major response patterns (Figure 2): Cluster 1 contained 26 damage-induced genes whose response was strongly reduced in the absence of ATM and Rel-A, and only partially affected by the absence of p53. Cluster 2 contained 11 genes whose response was abolished in the absence of ATM and p53, but augmented in the absence of Rel-A, suggesting some negative regulatory effect for NFκB on their expression. Cluster 3 contained 46 genes whose response was markedly attenuated in the absence of ATM and p53, and not substantially affected by the absence of Rel-A. Cluster 4 contained 12 genes whose induction was strongly reduced in the absence of p53, partially affected by the absence of ATM, and not affected by the absence of Rel-A.
This analysis shows the following. First, the transcriptional network induced on exposure to NCS in these cells is almost completely mediated by NFκB and p53, and these two transcription factors induce nearly disjoint sets of genes: the former controls the induction of cluster 1 genes, the latter controls the induction of the genes in clusters 2-4. Second, ATM is required for the activation of a major part of the damage-induced transcriptional program, comprising both the NFκB and p53 response arms (the activation of clusters 1-3 genes is ATM-dependent). Third, there is some cross-talk between the NFκB and p53 pathways: the absence of p53 partially reduces the induction of the NFκB arm (cluster 1), suggesting a positive effect of p53 on the induction of the NFκB mediated response; and the absence of Rel-A leads to increased activation of a subset of the p53-mediated arm (cluster 2), pointing to a negative regulatory role for NFκB in the induction of these genes.
The cluster analysis identified transcriptional responses mediated by both ATM/NFκB and ATM/p53. We sought to demonstrate that this dissection of the ATM-mediated transcriptional network induced by DNA damage is precise and cannot reasonably be ascribed to some nonspecific or off-target effects. To this end, we examined the effect of knocking-down Rel-A and p53 on several of their respective known direct targets that were included in the damage-induced genes set. Table 1a shows that knocking-down Rel-A and ATM significantly blocked the induction of known NFκB target genes, whereas knocking-down p53 had a much milder effect on their induction. Table 1b shows that knocking-down p53 and ATM specifically blocked the induction of known p53 target genes, whereas knocking-down Rel-A did not disrupt their induction (and even augmented it for some genes). Results of quantitative real-time reverse transcription PCR (RT-PCR), performed to validate the microarray results for these genes, were in good agreement with the microarray data in most cases; the magnitudes of induction differed between the two experimental systems, but the dependency of transcriptional induction on the various regulators was similar for 10 out of 13 genes examined.
To confirm the accuracy of the network dissection obtained by our experimental setup, we applied the PRIMA tool to our dataset. PRIMA, a computational promoter analysis tool recently developed by us [14], identifies transcription factors whose binding-site signatures are significantly more prevalent in a given set of promoters than expected by chance (see Materials and methods). In particular, promoters of genes assigned to cluster 1, which represents an ATM/NFκB-dependent response, were specifically and highly significantly enriched for the binding site signature of NFκB (Table 2), whereas p53-dependent clusters 3 and 4 were specifically enriched for the binding site of ATF2. ATF2 regulates transcription after heterodimerization with either ATF3 or c-Jun [15].
Notably, in our dataset the induction of both ATF3 and c-Jun was p53-dependent (Table 1b); hence the enrichment for this signature probably reflects a second wave of transcriptional regulation controlled by these transcription factors, whose induction is mediated by p53. This agrees with other studies that reported a p53-dependent activation of ATF3 in response to DNA damage [16,17]. PRIMA did not identify enrichment for the p53-binding-site signature in the p53-dependent clusters. It is possible that PRIMA is not sensitive enough to detect p53 enrichments because of the complex nature of the binding sites for p53 [18] or their relatively long distance from the transcription start sites (many experimentally validated p53-binding sites are located outside the promoter region included in PRIMA analysis). However, using the same parameters, PRIMA did identify significant enrichment for p53-binding signature in several other microarray datasets that we analyzed (data not shown). We therefore believe that p53 signature is not over-represented in these clusters, suggesting that p53 in the cells we used exerts its direct effect on a limited number of target genes, which are then further expanded into a wider network of transcriptional responses mediated mainly by ATF/Jun.
Discussion
The fine dissection of complex transcriptional responses has been a long-standing challenge in the signal transduction field. External and internal stimuli may activate complex networks whose analysis by traditional biochemistry can be daunting. High-throughput methods developed for functional genomics combined with powerful computational tools hold promise for deciphering such networks. The DNA damage response is an appropriate target for such an analysis. This highly branched signaling network spans numerous aspects of cellular metabolism and involves a vigorous wave of gene transcription across the genome.
In this study we have demonstrated the combined use of RNAi and microarray technologies and a recently developed computational tool to dissect the ATM-dependent transcriptional response following the induction of DSBs in DNA. RNAi technology has recently revolutionized biological research, but questions have been raised about the specificity of RNAi-mediated gene repression [8-11]. One way to filter out off-target effects is to use several different siRNA sequences against the same target on the assumption that completely different siRNAs will not induce the same off-target effects [7,11]. Following this logic, dissection of a signaling pathway that is mediated by several regulators using independent targeting of these regulators should similarly boost confidence. In this case, overlapping sets of genes whose expression is attenuated by knocking down different regulators are unlikely to be a result of off-target effects. It is also important to show that the observed effects are not a general consequence of the expression of siRNAs in the cells.
Our general goal is to dissect the DNA damage-induced transcriptional response in various cell types and tissues. In this study we focused on two arms of the this network whose induction is specifically mediated by the ATM/NFκB and the ATM/p53 regulators. First, we identified a set of genes whose induction in response to DNA damage was abrogated in cells knocked-down for two different components of the damage-induced signaling pathway, ATM and the Rel-A subunit of NFκB. Importantly, the induction of these genes was not disrupted in cells expressing siRNA against LacZ and was only mildly attenuated in cells knocked-down for p53, indicating that the loss of induction was not a general nonspecific consequence of siRNA expression. Moreover, computational promoter analysis showed that the set of promoters of these genes was highly and specifically enriched for the binding site signature of NFκB, providing independent evidence of the accuracy of this analysis. We then identified a set of genes whose induction in response to DNA damage was significantly abrogated in cells knocked-down for ATM and p53, but not in cells knocked-down for the Rel-A subunit of NFκB, or in the LacZ control. Again, it is unlikely this dissection of the ATM/p53-mediated arm can be ascribed to nonspecific or off-targets effects. According to computational promoter analysis, this set was highly enriched for the binding signature of ATF2/ATF3/Jun, a secondary transcriptional pathway whose induction was indeed p53-dependent in our data. This observation is in agreement with several studies reporting p53-dependent activation of this transcriptional pathway in response to DNA damage [16,17]. However, evidence suggests that p53-dependence of the induction of the ATF2/ATF3/Jun pathway depends on the cellular context, the type of DNA lesion, or the extent of damage, as p53-independent induction of this pathway was observed in other studies [19,20].
Evidence suggests that the sets of genes regulated by specific transcription factors depend on cell type and tissue context (see [21,22]). We are currently extending the analysis to various types of cell lines treated with a variety of DNA-damaging agents. Initial results indicate a marked cell-type specificity of the transcriptional response to DNA damage. The strategy presented here holds promise for disclosing and better understanding of this specificity.
Conclusions
Our analysis demonstrates that the combination of RNAi-targeting of key regulators, gene-expression profiling using microarrays, and computational promoter analysis is an informative method for the dissection of transcriptional networks in mammalian cellular systems despite the potential nonspecific and off-target effects of the RNAi technology. Targeting the primary activator of a DNA damage response network, the ATM protein kinase, and two key transcription factors that function downstream to it, p53 and NFκB, we showed that while the upstream regulator was indeed required for the induction of much of the network, the two downstream regulators mediated the activation of largely disjoint sets of genes. Thus, we dissected the network into two major arms. Statistical tests coupled with computational promoter analysis showed that this dissection was highly accurate.
Materials and methods
Establishment of siRNA knocked-down cellular systems
The following DNA fragments expressing shRNAs were cloned in the pSUPER retroviral vector [23,24], specifically designed to express siRNAs:
ATM_I (7218) 5'-GATCCCCCTGGTTAGCAGAAACGTGCTTCAAGAGAGCA CGTTTCTGCTAACCAGTTTTTGGAAA-'3.
ATM_II (p480): 5'-GATCCCCGATACCAGATCCTTGGAGATTCAAGAG ATCTCCAAGGATCTGGTATCTTTTTGGAAA-3', a generous gift from R. Agami. (ATM level was knocked-down using a combination of two different siRNAs.)
Rel_A: 5'-GATCCCCGAAGAGTCCTTTCAGCGGATTCAAGAGATCCGCTGAAAG GACTCTTCTTTTTGGAAA -3'.
p53: 5'-GATCCCCGACTCCAGTGGTAATCTACTTCAAGAGAGTAGATTACCACTG GAGTCTTTTTGGAAA-'3 (previously described in Brummelkamp et al. [24]).
LacZ: 5'-GATCCCCAAGGCCAGACGCGAATTATTTCAAGAGAATAATTCGCGTCT GGCCTTTTTTTGGAAA-3'.
HEK293 cells were transfected with ecotropic receptor expressing vector, infected with packaged viral particles, and selected with puromycin or hygromycin. Once stabilized, the cells were grown without selection.
Sample preparation and microarray hybridization
Cells were treated for 4 h with 200 ng/ml of NCS. Total RNA was isolated using TRIzol reagent (Life Technologies) and treated with DNase I (DNA free, Ambion). RNA was then purified using PLG tubes (Eppendorf), phenol/chloroform extracted, ethanol-precipitated and quantitated. The integrity of the RNA and the absence of contaminating genomic DNA were examined using gel electrophoresis. Expression profiles were recorded using Affymetrix Human Focus GeneChip arrays, which represent some 8,500 well annotated genes. Targets for hybridization to the microarrays were prepared using standard methods according to the manufacturer's instructions. Hybridization and scanning were performed as recommended by the manufacturer. All samples were probed in independent triplicates.
Computation of gene expression levels from microarray signals
Expression levels were computed using the RMA method [13] that was run from the BioConductor package [25]. The dataset was submitted to the Gene Expression Omnibus database [26] with accession number GSE1676. We preferred to use RMA over Affymetrix' MAS5 for two reasons. First, several studies have indicated that the mismatch signals are correlated with the mRNA concentration of their corresponding gene; that is, they themselves contain information on the expression level of the genes. Hence, subtracting their signals from the perfect-match ones, as MAS5 does, may add noise to the measurement and therefore be counterproductive [13]. RMA ignores the mismatch probes and computes expression levels based only on perfect match signals. When we examined the mismatch probe signals for several genes activated by the NCS treatment, we found that these signals indeed increased, in a manner correlated with the increase exhibited by their corresponding perfect-match signals (Additional data file 1). Second, whereas MAS5 uses global scaling to normalize between arrays, RMA applies the quantile normalization that was demonstrated to perform better [27]. Comparison of expression levels computed by MAS5 and RMA showed that RMA reduced noise between replicates (Additional data file 1), as well as the range of fold-changes in gene expression after the treatment (Additional data file 2).
Probe sets that received 'Absent' calls in all chips were filtered out, leaving 6,002 probe sets for subsequent steps of the data analysis. Averaging expression levels over replicates, our dataset contained measurements for ten conditions: five cellular systems (uninfected and the LacZ control cells and cells knocked-down for Rel-A, p53 and ATM), each probed at two time points: without treatment and 4 h after exposure to NCS.
Definition of the damage-responding gene set
We defined the damage-responding gene set as all genes whose expression levels changed by at least 1.5-fold in one control (either the uninfected or the LacZ-infected cells), and at least 1.4-fold in the same direction in the other control. A total of 112 genes that were induced in both controls met this criterion and are referred to as the damage-induced gene set (Additional data file 4). Only seven genes met an analogous criterion for repression in response to NCS treatment (Additional data file 4). We chose thresholds of 1.5 and 1.4 - lower than those usually used in microarray analysis - because the RMA method significantly narrows the distribution of expression levels and of the fold changes compared to Affymetrix' MAS5 package (Additional data files 1 and 2). Although the thresholds are low, the expected false-positive rate in our damage-induced gene set is low: not a single gene passed this criterion when it was applied to expression levels measured 30 min after exposure of the cells to NCS (data not shown). In addition, this number is significantly higher than expected at random: in 1,000 datasets with randomly permuted entries for each gene, the average number of genes that met this criterion was 14.1.
Cluster analysis
For each of the 112 damage-induced genes, induction fold-change of expression level after NCS treatment was computed in the control uninfected cells and in the cells knocked-down for Rel-A, p53 and ATM. The expression level of each damage-induced gene at the 4-h time point was divided by its level at the 0 time point in the same cellular system, yielding a 112 × 4 data matrix, with rows corresponding to genes. We normalized each row to mean = 0 and standarad deviation (SD) = 1, and subjected the normalized matrix to average-linkage hierarchical clustering using the EXPANDER package for microarray data analysis [28,29].
GO functional gene annotations
The gene ontology (GO) annotations of the genes were extracted using the DAVID utility [30].
Computational promoter analysis
Computational promoter analysis was done using PRIMA software, described in detail in Elkon et al. [14] and available at [31]. In brief, given target and background sets of promoters, PRIMA performs statistical tests aimed at identifying transcription factors whose binding sites are significantly more abundant in the target set than in the background set. PRIMA uses position weight matrices (PWMs) as models for regulatory sites that are bound by transcription factors. PWMs that represent human or mouse transcription-factor-binding sites were obtained from the TRANSFAC database [32]. The four gene clusters were used as target sets, and the entire collection of genes present on the chip (after filtering out those that got Absent calls in all chips) served as the background set in PRIMA tests. Putative promoter sequences corresponding to all known human genes were extracted from the human genome (Ensembl, version 19, Feb 2004), using a Perl script based on the application programming interface provided by the Ensembl project [33]. PRIMA tests were confined to 800 bp upstream to the putative genes' transcription start sites. Repetitive elements were masked out. Both strands were scanned.
Quantitative real-time RT-PCR
Five micrograms of total RNA were used for cDNA synthesis by oligo(dT) and SuperScript II RNase H- reverse transcriptase (Life Technologies). Quantitative real-time PCR using SYBR Green PCR master mix (Applied Biosystems) was performed with ABI PRISM 7900HT sequence detection system (Applied Biosystems). The comparative Ct method was used for quantification of transcripts according to the manufacturer's protocol. Measurement of ΔCt was performed in triplicate. We used glyceraldehyde-3-phosphate dehydrogenase (GAPDH) as the control gene for normalization. Primer pairs used in this study are given in Additional data file 2.
Additional data files
The following additional data are available with the online version of this paper. Additional data file 1 contains two figures showing the microarray results and their analysis. Additional data file 2 contains tables showing GO categories of affected genes, comparison between MAS5 and RMA computation of expression levels, primers used for real-time RT-PCR and the sequences of the shRNAs use in this study. Additional data file 3 contains a table listing genes whose expression was affected by infection of the cells with the shRNA-expressing retroviral vectors. Additional data file 4 contains a table listing the genes induced in both controls in in response to NCS treatment, and their assignment into the four clusters.
Supplementary Material
Additional File 1
Two figures showing the microarray results and their analysis. Supplementary Figure 1. Perfect-match (PM) and mismatch (MM) probe signals measured prior to and 4 hours after treatment with NCS. These signals are shown for four genes that were induced by the NCS treatment. As can be seen, mismatch signals were increased as well, pointing that they too contain information on gene expression level. Supplementary Figure 2. Comparison between RMA and MAS 5 computed signals. M vs. A plots (as introduced by Speed's lab ) based on expression levels that were computed by MAS5 or RMA for comparison between: (i) two replicated chips (C0a vs. C0b) (ii) post-treatment vs. pre-treatment chips (C0a vs. C4a), and (iii) same as (ii) but expression levels were averaged on triplicate chips at both time points. In all comparisons, the fold induction distributions (represented by the Y-axis) were markedly narrower when expression levels were computed by RMA. Distributions based on MAS5 were especially noisy in the low intensity genes.
Click here for file
Additional File 2
Tables showing GO categories of affected genes, comparison between MAS5 and RMA computation of expression levels, primers used for real-time RT-PCR and the sequences of the shRNAs use in this study. Supplementary Table B. GO categories of the genes that were upregulated in response to infection of the cells with shRNA-expressing retroviral vectors. Supplementary Table C. GO categories of the genes that were downregulated in response to infection of the cells with the shRNA-expressing retroviral vectors. Supplementary Table E. Comparison between MAS 5 and RMA computation of expression levels. Supplementary Table F. Primers used for quantitative real-time RT-PCR assays. Supplementary Table G. Sequences of shRNAs used in this study.
Click here for file
Additional File 3
A table listing genes whose expression was affected by infection of the cells with the shRNA-expressing retroviral vectors. Supplementary Table A. Genes whose expression was affected by infection of the cells with the shRNA-expressing retroviral vectors.
Click here for file
Additional File 4
A table listing the genes induced in both controls in in response to NCS treatment, and their assignment into the four clusters. Supplementary Table D. List of the 112 genes that were induced in both controls in response to NCS treatment, and their assignment into the four clusters.
Click here for file
Acknowledgements
We thank the Arison family for their donation to the Center of DNA Microarrays in Pediatric Oncology, Chaim Sheba Medical Center, and R. Agami for the p480 construct. R. Elkon is a Joseph Sassoon Fellow. G.R. holds the Djerassi Chair in Oncology and Y.S. holds the David and Inez Myers Chair in Cancer Genetics at the Sackler School of Medicine. This work was supported by research grants from the A-T Children's Project, the A-T Medical Research Foundation, and the Ministry of Science and Technology, Israel. This work was carried out in partial fulfillment of the requirements for the Ph.D. degree of R. Elkon.
Figures and Tables
Figure 1 Western blotting analysis showing the reduction in protein levels encoded by mRNAs that were targeted by siRNAs. α-Tubulin was used as a loading control.
Figure 2 The four major expression patterns in the damage-induced gene set revealed by cluster analysis. For each of the 112 damage-induced genes, the fold change in expression level 4 h after NCS treatment was computed in uninfected cells and in the cells knocked-down for Rel-A, p53 and ATM, yielding a 112 × 4 data matrix, with the rows corresponding to genes. This matrix was subjected to hierarchical clustering after normalizing the rows to have mean = 0 and SD = 1. The heat map visually represents the normalized matrix after being clustered. Red, green and black entries represent above-, below- and near-average fold change of induction, respectively. Four prominent expression patterns are evident. Cluster 1 represents genes whose induction is strongly attenuated in cells knocked-down for Rel-A and ATM (compared to the response in the control uninfected cells), and only partially attenuated in cells knocked-down for p53. Cluster 2 represents genes whose response is attenuated in cells knocked-down for p53 and ATM, but augmented in cells knocked-down for Rel-A. Cluster 3 represents genes whose response is attenuated in cells knocked-down for p53 and ATM, but not affected by knocking-down Rel-A. Cluster 4 represents genes whose response is markedly attenuated in cells knocked-down for p53, and only partially attenuated in cells knocked-down for ATM.
Table 1 Fold change in gene expression after 4 h exposure to NCS as measured by microarrays and by quantitative real-time RT-PCR
Gene Affy_ID Fold induction microarray Fold induction RT-PCR
C LacZ Rel-A (NFκB) p53 ATM C Rel-A (NFκB) p53 ATM
(a) Known direct targets of NFκB
TNFAIP3 202644_s_at 8.28 5.34 1.15 3.02 1.19 9.5 1.1 9.5 0.9
RELB 205205_at 3.7 2.89 0.82 2.95 0.91 15.7 6.0 21.3 2.5
TNFRSF9 207536_s_at 4.01 3.5 1.1 2.08 1.21 14.3 3.5 11.0 1.4
NFKBIA 201502_s_at 4.61 5.4 1.26 2.67 1.02 4.2 1.7 4.5 1.2
CD83 204440_at 3.46 2.99 1.0 1.73 1.06 6.5 1.0 5.7 1.3
IER3 201631_s_at 4.44 5.12 1.43 2.35 1.44 6.6 1.8 3.4 1.8
(b) Known direct targets of p53
ATF3 202672_s_at 3.44 3.74 7.03 1.54 1.47 5.2 5.9 1.6 1.6
EGR1 201694_s_at 2.78 1.77 6.77 1.04 1.02 4.4 13.4 0.7 2.4
JUN* 213281_at 2.01 1.45 2.71 1.36 1.25 6.6 3.9 0.64 2.5
FOS 209189_at 1.72 1.42 2.22 1.07 1.22 3.4 13.1 3.4 1.9
ETR101* 202081_at 1.97 2 2.6 1.06 1.13 2.0 3.0 1.4 1.4
GADD45A 203725_at 2.36 2.07 2.00 1.07 1.22 1.8 2.3 1.8 1.3
DUSP1 201041_s_at 2.06 2.57 3.45 1.11 1.22 2.2 4.5 2.0 1.9
*These genes are not reported as direct targets of p53 but are known to be functionally related to p53.
Table 2 Significantly enriched transcription factor binding site signatures in promoters of co-clustered genes
Cluster Number of genes* Dependence of gene induction† Binding-site enrichment‡
ATM Rel-A (NFκB) p53 NFκB (M00054) ATF2 (M00179)
1 26 ++ ++ + 9.7 (6.0 × 10-12) -
3 46 ++ - ++ - 2.9 (2.7 × 10-5)
4 12 + - ++ - 6.6 (3.6 × 10-6)
*Number of genes with promoter sequence data. †Strong attenuation in induction of the cluster's genes in the respective cells is denoted by ++; partial attenuation is denoted by +; and no attenuation by -. ‡The ratio between transcription-factor hit prevalence in the cluster and in the background sets of promoters, and its p-value (accession numbers for transcription-factor binding site models are from TRANSFAC DB).
==== Refs
Lee TI Rinaldi NJ Robert F Odom DT Bar-Joseph Z Gerber GK Hannett NM Harbison CT Thompson CM Simon I Transcriptional regulatory networks in Saccharomyces cerevisiae. Science 2002 298 799 804 12399584 10.1126/science.1075090
Pilpel Y Sudarsanam P Church GM Identifying regulatory networks by combinatorial analysis of promoter elements. Nat Genet 2001 29 153 159 11547334 10.1038/ng724
Segal E Yelensky R Koller D Genome-wide discovery of transcriptional modules from DNA sequence and gene expression. Bioinformatics 2003 19 Suppl 1 i273 i282 12855470 10.1093/bioinformatics/btg1038
Tavazoie S Hughes JD Campbell MJ Cho RJ Church GM Systematic determination of genetic network architecture. Nat Genet 1999 22 281 285 10391217 10.1038/10343
Hannon GJ RNA interference. Nature 2002 418 244 251 12110901 10.1038/418244a
Dykxhoorn DM Novina CD Sharp PA Killing the messenger: short RNAs that silence gene expression. Nat Rev Mol Cell Biol 2003 4 457 467 12778125 10.1038/nrm1129
Hannon GJ Rossi JJ Unlocking the potential of the human genome with RNA interference. Nature 2004 431 371 378 15372045 10.1038/nature02870
Bridge AJ Pebernard S Ducraux A Nicoulaz AL Iggo R Induction of an interferon response by RNAi vectors in mammalian cells. Nat Genet 2003 34 263 264 12796781 10.1038/ng1173
Persengiev SP Zhu X Green MR Nonspecific, concentration-dependent stimulation and repression of mammalian gene expression by small interfering RNAs (siRNAs). RNA 2004 10 12 18 14681580 10.1261/rna5160904
Sledz CA Holko M de Veer MJ Silverman RH Williams BR Activation of the interferon system by short-interfering RNAs. Nat Cell Biol 2003 5 834 839 12942087 10.1038/ncb1038
Jackson AL Bartz SR Schelter J Kobayashi SV Burchard J Mao M Li B Cavet G Linsley PS Expression profiling reveals off-target gene regulation by RNAi. Nat Biotechnol 2003 21 635 637 12754523 10.1038/nbt831
Povirk LF DNA damage and mutagenesis by radiomimetic DNA-cleaving agents: bleomycin, neocarzinostatin and other enediynes. Mutat Res 1996 355 71 89 8781578
Irizarry RA Hobbs B Collin F Beazer-Barclay YD Antonellis KJ Scherf U Speed TP Exploration, normalization, and summaries of high density oligonucleotide array probe level data. Biostatistics 2003 4 249 264 12925520 10.1093/biostatistics/4.2.249
Elkon R Linhart C Sharan R Shamir R Shiloh Y Genome-wide in silico identification of transcriptional regulators controlling the cell cycle in human cells. Genome Res 2003 13 773 780 12727897 10.1101/gr.947203
van Dam H Castellazzi M Distinct roles of Jun: Fos and Jun: ATF dimers in oncogenesis. Oncogene 2001 20 2453 2464 11402340 10.1038/sj.onc.1204239
Fan F Jin S Amundson SA Tong T Fan W Zhao H Zhu X Mazzacurati L Li X Petrik KL ATF3 induction following DNA damage is regulated by distinct signaling pathways and over-expression of ATF3 protein suppresses cells growth. Oncogene 2002 21 7488 7496 12386811 10.1038/sj.onc.1205896
Zhang C Gao C Kawauchi J Hashimoto Y Tsuchida N Kitajima S Transcriptional activation of the human stress-inducible transcriptional repressor ATF3 gene promoter by p53. Biochem Biophys Res Commun 2002 297 1302 1310 12372430 10.1016/S0006-291X(02)02382-3
Hoh J Jin S Parrado T Edington J Levine AJ Ott J The p53MH algorithm and its application in detecting p53-responsive genes. Proc Natl Acad Sci USA 2002 99 8467 8472 12077306 10.1073/pnas.132268899
Hayakawa J Depatie C Ohmichi M Mercola D The activation of c-Jun NH2-terminal kinase (JNK) by DNA-damaging agents serves to promote drug resistance via activating transcription factor 2 (ATF2)-dependent enhanced DNA repair. J Biol Chem 2003 278 20582 20592 12663670 10.1074/jbc.M210992200
Kool J Hamdi M Cornelissen-Steijger P van der Eb AJ Terleth C van Dam H Induction of ATF3 by ionizing radiation is mediated via a signaling pathway that includes ATM, Nibrin1, stress-induced MAPkinases and ATF-2. Oncogene 2003 22 4235 4242 12833146 10.1038/sj.onc.1206611
Odom DT Zizlsperger N Gordon DB Bell GW Rinaldi NJ Murray HL Volkert TL Schreiber J Rolfe PA Gifford DK Control of pancreas and liver gene expression by HNF transcription factors. Science 2004 303 1378 1381 14988562 10.1126/science.1089769
Coates PJ Lorimore SA Wright EG Cell and tissue responses to genotoxic stress. J Pathol 2005 205 221 235 15643669 10.1002/path.1701
Brummelkamp TR Bernards R Agami R A system for stable expression of short interfering RNAs in mammalian cells. Science 2002 296 550 553 11910072 10.1126/science.1068999
Brummelkamp TR Bernards R Agami R Stable suppression of tumorigenicity by virus-mediated RNA interference. Cancer Cell 2002 2 243 247 12242156 10.1016/S1535-6108(02)00122-8
BioConductor
Gene Expression Omnibus (GEO)
Bolstad BM Irizarry RA Astrand 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
Sharan R Maron-Katz A Shamir R CLICK and EXPANDER: a system for clustering and visualizing gene expression data. Bioinformatics 2003 19 1787 1799 14512350 10.1093/bioinformatics/btg232
EXPANDER
Dennis G JrSherman BT Hosack DA Yang J Gao W Lane HC Lempicki RA DAVID: Database for Annotation, Visualization, and Integrated Discovery. Genome Biol 2003 4 P3 12734009 10.1186/gb-2003-4-5-p3
PRIMA
Matys V Fricke E Geffers R Gossling E Haubrock M Hehl R Hornischer K Karas D Kel AE Kel-Margoulis OV TRANSFAC: transcriptional regulation, from patterns to profiles. Nucleic Acids Res 2003 31 374 378 12520026 10.1093/nar/gkg108
Birney E Andrews TD Bevan P Caccamo M Chen Y Clarke L Coates G Cuff J Curwen V Cutts T An overview of Ensembl. Genome Res 2004 14 925 928 15078858 10.1101/gr.1860604
| 15892871 | PMC1175955 | CC BY | 2021-01-04 16:05:39 | no | Genome Biol. 2005 Apr 13; 6(5):R43 | utf-8 | Genome Biol | 2,005 | 10.1186/gb-2005-6-5-r43 | oa_comm |
==== Front
Genome BiolGenome Biology1465-69061465-6914BioMed Central London gb-2005-6-5-r441589287210.1186/gb-2005-6-5-r44MethodThe Sequence Ontology: a tool for the unification of genome annotations Eilbeck Karen [email protected] Suzanna E [email protected] Christopher J [email protected] Mark [email protected] Lincoln [email protected] Richard [email protected] Michael [email protected] Department of Molecular and Cellular Biology, Life Sciences Addition, University of California, Berkeley, CA 94729-3200, USA2 Howard Hughes Memorial Institute, Department of Molecular and Cellular Biology, Life Sciences Addition, University of California, Berkeley, CA 94729-3200, USA3 Cold Spring Harbor Laboratory, 1 Bungtown Road, Cold Spring Harbor, New York 11724, USA4 Sanger Institute, Wellcome Trust Genome Campus, Hinxton, Cambridgeshire, CB10 1SA, UK5 Department of Genetics, University of Cambridge, Downing Street, Cambridge, CB2 3EH, UK2005 29 4 2005 6 5 R44 R44 4 10 2004 1 2 2005 30 3 2005 Copyright © 2005 Eilbeck 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 goal of the Sequence Ontology (SO) project is to produce a structured controlled vocabulary with a common set of terms and definitions for parts of a genomic annotation, and to describe the relationships among them. Details of SO construction, design and use, particularly with regard to part-whole relationships are discussed and the practical utility of SO is demonstrated for a set of genome annotations from Drosophila melanogaster.
The Sequence Ontology (SO) is a structured controlled vocabulary for the parts of a genomic annotation. SO provides a common set of terms and definitions that will facilitate the exchange, analysis and management of genomic data. Because SO treats part-whole relationships rigorously, data described with it can become substrates for automated reasoning, and instances of sequence features described by the SO can be subjected to a group of logical operations termed extensional mereology operators.
==== Body
Background
Why a sequence ontology is needed
Genomic annotations are the focal point of sequencing, bioinformatics analysis, and molecular biology. They are the means by which we attach what we know about a genome to its sequence. Unfortunately, biological terminology is notoriously ambiguous; the same word is often used to describe more than one thing and there are many dialects. For example, does a coding sequence (CDS) contain the stop codon or is the stop codon part of the 3'-untranslated region (3' UTR)? There really is no right or wrong answer to such questions, but consistency is crucial when attempting to compare annotations from different sources, or even when comparing annotations performed by the same group over an extended period of time.
At present, GenBank [1] houses 220 viral genomes, 152 bacterial genomes, 20 eukaryotic genomes and 18 archeal genomes. Other centers such as The Institute for Genomic Research (TIGR) [2] and the Joint Genome Institute (JGI) [3] also maintain and distribute annotations, as do many model organism databases such as FlyBase [4], WormBase [5], The Arabidopsis Information Resource (TAIR) [6] and the Saccharomyces Genome Database (SGD) [7]. Each of these groups has their own databases and many use their own data model to describe their annotations. There is no single place at which all sets of genome annotations can be found, and several sets are informally mirrored in multiple locations, leading to location-specific version differences. This can make it hazardous to exchange, combine and compare annotation data. Clearly, if genomic annotations were always described using the same language, then comparative analysis of the wealth of information distributed by these institutions would be enormously simplified: Hence the Sequence Ontology (SO) project. SO began 2 years ago, when a group of scientists and developers from the model organism databases - FlyBase, WormBase, Ensembl, SGD and MGI - came together to collect and unify the terms they used in their sequence annotation.
The Goal of the SO is to provide a standardized set of terms and relationships with which to describe genomic annotations and provide the structure necessary for automated reasoning over their contents, thereby facilitating data exchange and comparative analyses of annotations. SO is a sister project to the Gene Ontology (GO) [8] and is part of the Open Biomedical Ontologies (OBO) project [9]. The scope of the SO project is the description of the features and properties of biological sequence. The features can be located in base coordinates, such as gene and intron, and the properties of these features describe an attribute of the feature; for example, a gene may be maternally_imprinted.
SO terminology and format
Like other ontologies, SO consists of a controlled vocabulary of terms or concepts and a restricted set of relationships between those terms. While the concepts and relationships of the sequence ontology make it possible to describe precisely the features of a genomic annotation, discussions of them can lead to much lexical confusion, as some of the terms used by SO are also common words; thus we begin our description of SO with a discussion of its naming conventions, and adhere to these rules throughout this document.
Wherever possible, the terms used by SO to describe the parts of an annotation are those commonly used in the genomics community. In some cases, however, we have altered these terms in order to render them more computer-friendly so that users can create software classes and variables named after them. Thus, term names do not include spaces; instead, underscores are used to separate the words in phrases. Numbers are spelled out in full, for example five_prime_UTR, except in cases where the number is part of the accepted name. If the commonly used name begins with a number, such as 28S RNA, the stem is moved to the front - for example, RNA_28S. Symbols are spelled out in full where appropriate, for example, prime, plus, minus; as are Greek letters. Periods, points, slashes, hyphens, and brackets are not allowed. If there is a common abbreviation it is used as the term name, and case is always lower except when the term is an acronym, for example, UTR and CDS. Where there are differences in the accepted spelling between English and US usage, the US form is used.
Synonyms are used to record the variant term names that have the same meaning as the term. They are used to facilitate searching of the ontology. There is no limit to the number of synonyms a term can have, nor do they adhere to SO naming conventions. They are, however, still lowercase except when they are acronyms.
Throughout the remainder of this document, the terms from SO are highlighted in italics and the names of relationships between the terms are shown in bold. The terms are always depicted exactly as they appear in the ontology. The names of EM operators are underlined.
SO, SOFA, and the feature table
To facilitate the use of SO for the markup of gene annotation data, a subset of terms from SO consisting of some of those terms that can be located onto sequence has been selected; this condensed version of SO is especially well suited for labeling the outputs of automated or semi-automated sequence annotation pipelines. This subset is known as the Sequence Ontology Feature Annotation, or SOFA.
SO, like GO, is an 'open source' ontology. New terms, definitions, and their location within the ontology are proposed, debated, and approved or rejected by an open group of individuals via a mailing list. SO is maintained in OBO format and the current version can be downloaded from the CVS repository of the SO website [10]. For development purposes, SOFA was stabilized and released (in May 2004) for at least 12 months to allow development of software and formats. SO is a directed acyclic graph (DAG), and can be viewed using the editor for OBO files, OBO-Edit [11].
The terms describing sequence features in SO and SOFA are richer than those of the Feature Table [12] of the three large genome databanks: GenBank [1], EMBL [13] and the DNA Data Bank of Japan (DDBJ) [14]. The Feature Table is a controlled vocabulary of terms describing sequence features and is used to describe the annotations distributed by these data banks. The Feature Table does provide a grouping of its terms for annotation purposes, based on the degree of specificity of the term. The relationships between the terms are not formalized; thus the interpretation of these relationships is left to the user to infer, and, more critically, must be hard-coded into software applications. Most of the terms in the Feature Table map directly to terms in SO, although the term names may have been changed to fit SO naming conventions. In general, SO contains a more extensive set of features for detailed annotation. There are currently 171 locatable sequence features in SOFA compared to 65 of the Feature Table. There are 11 terms in the Feature Table that are not included in SO. These terms fall into two categories: remarks and immunological features, both of which have been handled slightly differently in SO. A mapping between SO and the Feature Table is available from the SO website [10].
Database schemas, file formats and SO
SO is not a database schema, nor is it a file format; it is an ontology. As such, SO transcends any particular database schema or file format. This means it can be used equally well as an external data-exchange format or internally as an integral component of a database.
The simplest way to use SO is to label data destined for redistribution with SO terms and to make sure that the data adhere to the SO definition of the data type. Accordingly, SO provides a human-readable definition for each term that concisely states its biological meaning. Usually the definitions are drawn from standard authoritative sources such as The Molecular Biology of the Cell [15], and each definition contains a reference to its source. Defining each term in such a way is important as it aids communication and minimizes confusion and disputes as to just what data should consist of. For example, the term CDS is defined as a contiguous RNA sequence which begins with, and includes, a start codon and ends with, and includes, a stop codon. According to SO, the sequence of a three_prime_utr does not contain the stop_codon - and files with such sequences are SO-compliant; files of three_prime_utr containing stop_codons are not. This is a trivial example, illustrating one of the simplest use cases, but it does demonstrate the power of SO to put an end to needless negotiations between parties as to the details of a data exchange. This aspect of SO is especially well suited for use with the generic feature format (GFF) [16]. Indeed, the latest version, GFF3, uses SO terms and definitions to standardize the feature type described in each row of a file and SO terms as optional attributes to a feature.
SO can also be employed in a much more sophisticated manner within a database. CHADO [17] is a modular relational database schema for integrating molecular and genetic data and is part of the Generic Model Organism Database project (GMOD) [18], currently used by both FlyBase and TIGR. The CHADO relational schema is extremely flexible, and is centered on genomic features and their relationships, both of which are described using SO terms. This use of SO ensures that software that queries, populates and exports data from different CHADO databases is interoperable, and thus greatly facilitates large-scale comparisons of even very complex genomics data.
Like GFF3, Chaos-XML [19] is a file format that uses SO to label and structure data, but it is more intimately tied to the CHADO project than is GFF3. Chaos-XML is a hierarchical XML mapping of the CHADO relational schema. Annotations are represented as an ontology-typed feature graph. The central concept of Chaos-XML is the sequence-feature, which is any sequence entity typed by SO. The features are interconnected via feature relationship elements, whereby each relationship connects a subject feature and an object feature. Features are located via featureloc elements which use interbase (zero-based) coordinates. Chaos-XML and CHADO are richer models than GFF3 in that feature_relationships are typed, and a more sophisticated location model is used. Chaos-XML is the substrate of a suite of programs called Comparative Genomics Library (CGL), pronounced 'seagull' [20], which we have used for the analyses presented in our Results section.
The basic types in SOFA, from which other types are defined, are region and junction, equivalent to the concepts of interiors and boundaries defined in the field of topological relationships [21]. A region is a length of sequence such as an exon or a transposable_element. A junction is the space between two bases, such as an insertion_site. Building on these basic data types, SOFA can be used to describe a wide range of sequence features. Raw sequence features such as assembly components are captured by terms like contig and read. Analysis features, defined by the results of sequence-analysis programs such as BLAST [22] are captured by terms such as nucleotide_match. Gene models can be defined on the sequence using terms like gene, exon and CDS. Variation in sequence is captured by subtypes of the term sequence_variant. These terms have multiple parentages with either region or junction. SOFA (and SO) can also be used to describe many other sequence features, for example, repeat, reagent, remark. Thus, SOFA together with GFF3 or Chaos-XML provide an easy means by which parties can describe, standardize, and document the data they distribute and exchange.
The SO and SOFA controlled vocabularies can be used for de novo annotation. Several groups including SGD and FlyBase now use either SO or SOFA terms in their annotation efforts. SO is not restricted to new annotations, however, and may be applied to existing annotations. For example, annotations from GenBank may be converted into SO-compliant formats using Bioperl [23] (see Materials and methods).
SO relationships
One essential difference between a controlled vocabulary, such as the Feature Table, and an ontology is that an ontology is not merely a collection of predefined terms that are used to describe data. Ontologies also formally specify the relationships between their terms. Labeling data with terms from an ontology makes the data a substrate for software capable of logical inference. The information necessary for making logical inferences about data resides in the class designations of the relationships that unite terms within SO. We detail this aspect of the ontology below. For purposes of reference, a section of SO illustrating the various relationships between some of its terms is shown in Figure 1.
Currently, SO uses three basic kinds of relationship between its terms: kind_of, derives_from, and part_of. These relationships are defined in the OBO relationship types ontology [24]. kind_of relationships specify what something 'is'. For example, an mRNA is a kind_of transcript. Likewise an enhancer is a kind_of regulatory_region. kind_of relationships are valid in only one direction. Hence, a regulatory_region is not a kind_of enhancer. One consequence of the directional nature of kind_of relationships is that their transitivity is hierarchical - inferences as to what something 'is' proceed from the leaves towards the root of the ontology. For example, an mRNA is a kind_of processed_transcript AND a processed_transcript is a kind_of transcript. Thus, an mRNA is a kind_of transcript. kind_of relationships are synonymous with is_a relationships. We adopted the 'kind_of' notation to avoid the lexical confusion often encountered when describing relationships, as the phrase 'is a' is often used in conjunction with another relationships in English - for example 'is a part_of'.
SO uses the term derives_from to denote relationships of process between two terms. For example, an EST derives_from an mRNA. derives_from relationships imply an inverse relationship; derives. Note that although a polypeptide derives_from an mRNA, a polypeptide cannot be derived from an ncRNA (non-coding RNA), because no derives_from relationship unites these two terms in the ontology. This fact illustrates another important aspect of how SO handles relationships: children always inherit from parents but never from siblings. An ncRNA is a kind_of transcript as is an mRNA. Labeling something as a transcript implies that it could possibly produce a polypeptide; labeling that same entity with the more specific term ncRNA rules that possibility out. Thus, a file that contained ncRNAs and their polypeptides would be semantically invalid.
part_of relationships pertain to meronomies; that is to say 'part-whole' relationships. An exon, for example, is a part_of a transcript. part_of relationships are not valid in both directions. In other words, while an exon is a part_of a transcript, a transcript is not a part_of an exon. Instead, we say a transcript has_part exon. SO does not explicitly denote whole-part relationships, as every part_of relationship logically implies the inverse has_part relationship between the two terms.
Transitivity is a more complicated issue with regards to part-whole relationships than it is for the other relationships in SO. In general, part_of relationships are transitive - an exon is a part_of a gene, because an exon is a part_of a transcript, and a transcript is a part_of a gene. Not every chain of part-whole relationships, however, obeys the principle of transitivity. This is because parts can be combined to make wholes according to different organizing principles. Winston et al. [25] have described six different subclasses of the part-whole relationship, based on the following three properties: configuration, whether the parts have a structural or functional role with respect to one another or the whole they form; substance, whether the part is made of the same stuff as the whole (homomerous or heteromerous); and invariance, whether the part can be separated from the whole. These six relations and their associated part_of subclasses are detailed in Table 1.
Winston et al. [25] argue that there is transitivity across a series of part_of relationships only if they all belong to the same subclass. In other words, an exon can only be part_of a gene, if an exon is a component_part_of a transcript, and a transcript is component_part_of a gene. If, however, the two statements contain different types of part_of relationship, then transitivity does not hold.
By addressing the vague English term 'part of' in this way, Winston et al. solve many of the problems associated with reasoning across part_of relationships; thus, we are adopting their approach with SO. The parts contained in the sequence ontology are mostly of the type component_part_of such as exon is a part_of transcript, although there are a few occurrences of member_part_of such as read is a part_of contig.
SO's relationships facilitate software design and bioinformatics research
Genomic annotations are substrates for a multitude of software applications. Annotations, for example, are rendered by graphical viewers, or, as another example, their features are searched and queried for purposes of data validation and genomics research. Using an ontology for sequence annotation purposes offers many advantages over the traditional Feature Table approach. Because controlled vocabularies do not specify the relationships that obtain between their terms, using the Feature Table has meant that relationships between features have had to be hard-coded in software applications themselves; consequently, adding a new term to the Feature Table and/or changing the details of the relationships that obtain between its terms has meant revising every software application that made use of the Feature Table. Ontologies mitigate this problem as all of the knowledge about terms and their relationships to one another is contained in the ontology, not the software.
SO-compliant software need only be provided with an updated version of the ontology, and everything else will follow automatically. This is because SO-compliant software need not hard-code the fact that a tRNA is a kind_of transcript; it need merely know that kind_of relationships are transitive and hierarchical and be capable of internally navigating the network of relationships specified by the ontology (see Figure 1) in order to logically infer this fact. This means that every time a new form of ncRNA is discovered, and added to SO, all SO-compliant software applications will automatically be able to infer that any data labeled with that new term is a kind_of transcript. This means that existing graphical viewers will render those data with the appropriate transcript glyph, and validation and query tools will automatically deal with this new data-type in a coherent fashion. Placing the biological knowledge in the ontology rather than in the software means that the ontology and the software that uses it can be developed, revised, and extended independently of one another. Thus ontologies offer the bioinformatics programming community significant opportunities as regards software design and the speed of the development cycle. Using an ontology does, however, mean that software applications must meet certain professional standards; namely, they must be capable of parsing an OBO file and navigating the network of relationships that constitute the ontology, but these are minimal hurdles.
SO facilitates bioinformatics research in ways that reach far beyond its utility as regards software design. For example, SO's kind_of relationships provide a subsumption hierarchy, or classification system for its terms. This added depth of knowledge greatly improves the searching and querying capabilities of software using SO. The ontology's higher-level terms may be used to query via inference, even if they are never used for annotation. We recommend that annotators label their data using terms corresponding to terminal nodes in the ontology. Transcripts, for example, might be annotated using terms such as mRNA, tRNA, and rRNA (see Figure 1). Note that doing so means that if, for example, non-coding RNA sequences are required for some subsequent analysis, then SO-compliant software tools can locate annotations labelled with the subtypes of ncRNA, and retrieve tRNAs and rRNAs to the exclusion of mRNAs, even though these data have not been explicitly labelled with the term ncRNA. Thus, many analyses become easy, for example, how many ncRNAs are annotated in H. sapiens? Of these what percent have more than one exon? Are any maternally imprinted? Moreover, using SO as part of a database schema ensures that such questions 'mean' the same thing in different databases.
SO also greatly facilitates the automatic validation of annotation data, as the relationships implied by an annotation can be compared to the allowable relationships specified in the ontology. For example, an annotation that asserts an intron to be part_of an mRNA would be invalid, as this relationship is not specified in the ontology (Figure 1). On the other hand, an annotation that asserted that an UTR sequence was part_of mRNA would be valid (Figure 1). This makes possible better quality control of annotation data, and makes it possible to check existing annotations for such errors when converting them to a SO-compliant format such as GFF3.
To summarize, by identifying the set of relationships between terms that are possible, we are also specifying the inferences that can be drawn from these relationships: that is, the software operations that can be carried out over the data. As a consequence, software is easier to maintain, SO can easily be extended to embrace new biological knowledge, quality controls can be readily implemented, and software to mine data can be written so as to be very flexible.
EM operators and SO
SO also enables some modes of analyses of genomics data that are completely new to the field. One such class of analyses involves the use of extensional mereology (EM) operators to ask questions about gene parts. Although new to genomics, EM operators are well known in the field of ontology, where they provide a basis for asking and answering questions pertaining to how parts are distributed within and among different wholes (reviewed in [26,27]). These operators are usually applied to studies of how parts are shared between complex wholes - such as different models of automobiles or personal computers - for the purpose of optimizing manufacturing procedures. Below we explain how these same operators can be applied to the analyses of genomics data. Although these operators, difference and overlap, share the same name as topological operators, they are different as they function on the parts of an object, not on its geometric coordinate space. The topological operators, regarding the coincidence of edges and interiors - equality, overlap, disjointedness, containment and coverage of spatial analysis [21] - may also be applied to biological sequence.
EM is a formal theory of parts: it defines the properties of the part_of relationship and then provides a set of operations (Table 2) that can be applied to those parts. These operators are akin to those of set theory, but whereas set theory makes use of an object's kind_of relationships, EM operators function on an object's part_of relationships. Only wholes and their 'proper parts' are legitimate substrates for EM operations. Proper parts are those parts that satisfy three self-evident criteria: first, nothing is a proper part of itself (a proper part is part of but not identical to the individual or whole); second, if A is a proper part of B then the B is not a part of A; third, if A is a part of B and B is a part of C then A is a part of C.
Note that the third criterion of proper parts is that they obey the rule of transitivity. As we discussed earlier, not all part_of relationships are transitive. Accordingly, we have restricted our analyses (see Results and discussion) to component parts (Table 2).
Figure 2 illustrates the effects of applying EM operations to analyze the relationships 'transcript is a part_of gene' and 'exon is a part_of transcript'. The EM operations overlap and disjoint pertain to relationships between transcripts, whereas difference and binary product pertain to exons. Two transcripts overlap if they share one or more exon in common. Two transcripts are disjoint if they do not share any exons in common. The exons shared between two overlapping transcripts are the binary product of the two transcripts, and the exons not shared in common comprise the difference between the two transcripts. The binary sum of two transcripts is simply the sum of their parts.
One key feature of EM operations is that they operate in 'identifier space' rather than 'coordinate space'. Two transcripts overlap only if they share a part in common rather than if their genomic coordinates overlap. Thus, two transcripts may be disjoint even if their exons partially overlap one another. This is one way in which EM analyses differ from standard bioinformatics analyses, and it has some interesting repercussions. This is particularly so with regard to modes of alternative splicing, as each of the EM operations suggests a distinct category by means of which two alternatively spliced transcripts can be related to one another. We further explore the potential of these operations to classify alternative transcripts and their exons below.
Results and discussion
As part of a pilot project to evaluate the practical utility of SO as a tool for data management and analysis, we have used SO to name and enumerate the parts of every protein-coding annotation in the D. melanogaster genome. Doing so has allowed us to compare annotations with respect to their parts, for example, number of exons, amount of UTR sequence, and so on.
These data afford many potential analyses, but as our motivation was primarily to demonstrate the practical utility of SO as a tool for data management, rather than comparative genomics per se, we have focused more on what exon-transcript-gene part-whole relationships have to say about the annotations themselves, than what the annotations have to say about the biology of the genome. Accordingly, we have used EM-operators to characterize the annotations with respect to their parts, especially with regard to alternative splicing. The current version of FlyBase (5 August, 2004) contained 13,539 genes, (of which 10,653 have a single transcript and 2,886 are alternatively spliced), 18,735 transcripts and 61,853 exons.
An EM-based scheme for classifying alternatively spliced genes
As we had characterized the parts of the annotations using SO, we were able to employ the EM operators over these parts. This proved to be a natural way to explore the relative complexity of alternative splicing, as the alternatively spliced transcripts have different combinations of parts: that is, exons. We grouped alternatively spliced transcripts into two classes. An alternatively spliced gene will contain overlapping transcripts if at least one of its exons is shared between two of its transcripts, and will have disjoint transcripts if one of its transcripts shares no exons in common with any other transcript of that gene. For the purposes of this analysis, we further classified disjoint transcripts as sequence-disjoint and parts-disjoint. We term two disjoint transcripts sequence-disjoint if none of their exons shares any sequence in common with one another; and parts-disjoint if one or more of their exons overlap on the chromosome but have different exon boundaries. Note that the three operations are pairwise, and thus not mutually exclusive. To see why this is, imagine a gene having three transcripts, A, B, and C. Obviously, transcript A can be disjoint with respect to B, but overlap with respect to C. Thus, we can speak of a gene as having both disjoint and overlapping transcripts.
The relative numbers of disjoint and overlapping transcripts in a genome says something about the relative complexity of alternative splicing in that genome. A gene may have any combination of these types of disjoint and overlapping transcripts, so we created a labeling system consisting of the seven possible combinations. We did this by asking three EM-based questions about the relationships between pairs of a gene's transcripts: How many pairs are there of sequence-disjoint transcripts? How many pairs are there of parts-disjoint transcripts? How many pairs are there of overlapping transcripts? Doing so allowed us to place that gene into one of seven classes with regards to the properties of its alternatively spliced transcripts. We also kept track of the number of times each of the three relationships held true for each pair combination. For example, a gene having two transcripts that are parts-disjoint with respect to one another would be labeled 0:1:0. Keeping track of the number of transcript pairs falling into each class provides an easy means to prioritize them for manual review. These results are summarized in Figure 3.
Of the alternatively spliced fly genes, none has a sequence-disjoint transcript, 275 have parts-disjoint transcripts, and 2,664 have overlapping transcripts, and 53 have both parts-disjoint and overlapping transcripts. The percentage of D. melanogaster genes in each category is shown in Table 3. Most alternatively spliced genes contain at least one pair of overlapping transcripts. These data also have something to say about the ways in which research and management issues are intertwined with one another with respect to genome annotation, as some aspects of these data are clearly attributable to annotation practice. The lack of any sequence-disjoint transcripts in D. melanogaster, for example, is due to annotation practice; in fact, current FlyBase annotation practices forbid their creation, the reason being that any evidence for such transcripts is evidence for a new gene [28]. This is not true for all genomic annotations. Annotations converted from the genomes division of GenBank to a SO-compliant form, were subjected to EM analysis, and inspection of the corresponding gene-centric annotations provided by Entrez Gene [29] revealed examples of genes that fall into each of the seven categories. Some of these annotations are shown in Figure 3.
The frequencies of genes that fall into each of the seven classes shown in Table 3 provides a concise summary of genome-wide trends in alternative splicing in the fly. This EM-based classification schema, when applied to many model organisms, from many original sources, makes very apparent the magnitude of the practical challenges that surround decentralized annotation, and the distribution and redistribution of annotations. Certainly, they highlight the need for data-management tools such as SO to assist the community in enforcing biological constraints and annotation standards. Only then will comparative genomic analyses show their full power.
Exons as alternative parts of transcripts
EM-operators can also be used to classify the exons of alternatively spliced genes. Exons shared between two transcripts comprise the binary product of the two transcripts; whereas those exons present in only one of the transcripts constitute their difference (see Table 2 and Figure 2 for more information). These basic facts suggest a very simple, three-part classification system. If an exon is the difference between all other transcripts, then it is only in one transcript; we term these UNIQUE exons. If an exon is the difference of some transcripts, and the binary product of others, it is in a fraction of transcripts; we term these SOMETIMES_FOUND exons. And, if an exon is the binary product of all combinations of transcripts, then it must be in all transcripts; we term such exons ALWAYS_FOUND exons. Classifying exons in this way allows us to look more closely at alternative splicing from the exon's perspective.
As can be seen from Table 4, despite the low frequency of alternatively spliced genes, a large fraction of their exons are associated with alternatively spliced transcripts - almost 39%. A sizable proportion of SOMETIMES_FOUND and ALWAYS_FOUND exons are coding exons in some of the transcripts and entirely untranslated exons in others. In some cases, this is due to actual biology: some transcripts in D. melanogaster are known to produce more than one protein (see, for example [30]). In other cases, this situation appears to be a result of best attempts on the part of annotators to interpret ambiguous supporting evidence; in yet others the supporting data sometimes unambiguously points to patterns of alternative splicing that would seem to produce transcripts destined for nonsense-mediated decay [31]. Whatever the underlying cause, these exons, like the N:0:0 class annotations, should be subjected to further investigation.
To investigate these conclusions in more detail, we further examined each exon with respect to its EM-based class and its coding and untranslated portions. These results are shown Figure 4, and naturally extend the analyses presented in Table 4. First, regardless of exon class, most entirely untranslated exons are 5-prime exons; the lower frequency of 3-prime untranslated exons is perhaps due to nonsense-mediated decay [31], as the presence of splice junctions in a processed transcript downstream of its stop codon are believed to target that transcript for degradation. A second point made clear by the data in Table 4 is that alternatively spliced genes of D. melanogaster are highly enriched for 5-prime untranslated exons compared with single-transcript genes. Most of these exons belong to ALWAYS_FOUND; thus, there seems to be a strong tendency in D. melanogaster for alternative transcripts to begin with a unique 5' UTR region. This fact suggests that alternative transcription in the fly may, in many cases, be a consequence of alternative-promoter usage and perhaps tissue-specific transcription start sites. The high percentage of untranslated 5-prime UNIQUE exons in D. melanogaster may also be a consequence of the large numbers of 5' ESTs that have been sequenced in the fly [32].
Figure 4 also shows that most (> 95%) D. melanogaster ALWAYS_FOUND exons are coding. This makes sense, as it seems likely that one reason for an exon's inclusion in every one of a gene's alternative transcripts is that it encodes a portion of the protein essential for its function(s).
As with our previous analyses of alternative transcripts, our analyses of alternatively transcribed exons also illustrate the ways in which basic biology and annotation-management issues intersect one another. The fact that most ALWAYS_FOUND exons are entirely coding, for example, may have something important to say about which parts of a protein are essential for its function(s). Whereas the over-abundance of un-translated UNIQUE exons probably has more to say about the resources available to, and the protocols used by, the annotation project than it does about biology. Such considerations make it clear that the evidence used to produce an annotation is an essential part of the annotation. In this regard SO has much to offer, as it provides a rational means by which to manage annotation evidence in the context of gene-parts and the relations between those parts.
Conclusion
We have sought to provide an introduction to the SO and justify why its use to unify genomic annotations is beneficial to the model organism community. We illustrate some of the ways in which SO can be used to analyze and manage annotations. Relationships are an essential component of SO, and understanding their role within the ontology is a basic prerequisite for using SO in an intelligent fashion. Much of this paper revolves around the part_of relationship because SO is largely a meronomy - a particular kind of ontology concerned with the relationships of parts to wholes. Extensional mereology (EM) is an area that is largely new to bioinformatics for which there are several excellent reference works available [26,27,33], and even a cursory examination of these texts will make it clear that EM has much to offer bioinformatics.
Using all of the relationships in SO allows us to automatically draw logical conclusions about data that has been labelled with SO terms and thereby provide useful insights into the underlying annotations. We have shown how SO, together with the EM-based operations it enables, can be used to standardize, analyze, and manage genome annotations.
Given any standardized set of genome annotations described with SO these annotations can then be rigorously characterized. For our pilot analyses, we focused on alternatively transcribed genes and their exons, and explored the potential of EM-operators to classify and characterize them. We believe that the results of these analyses support two principle conclusions. First, EM-based classification schemes are simple to implement, and second, they capture important trends in the data and provide a concise, natural, and meaningful overview of annotations in these genomes.
One criticism that might be justifiably leveled against the SO- and EM-based analyses presented here is that they are too formal, and that simpler approaches could have accomplished the same ends. As our discussion of part_of relationships made clear, however, reasoning across diverse types of parts is a complicated process; ad-hoc approaches will not suffice where the data are complex. The more formal approach afforded by SO means that analyses can be easily be extended beyond the domain of transcripts and exons to include many other gene parts and relationships as well - including evidence. It seems clear that over the next few years both the number and complexity of annotations will increase, especially with regard to the diversity of their parts. Drawing valid conclusions from comparisons of these annotations will prove challenging. That SO has much to offer such analyses is indisputable.
SO and SOFA provide the model organism community with a means to unify the semantics of sequence annotation. This facilitates communication within a group and between different model organism groups. Adopting SO terminology to type the features and properties of sequence will provide both the group and the community the advantages of a common vocabulary, to use for sharing and querying data and for automated reasoning over large amounts of sequence data.
Materials and methods
SO and SOFA have been built and are maintained using the ontology-editing tool OBO-Edit. The ontologies are available at [34].
The FlyBase D. melanogaster [35] data was derived from the GadFly [36] relational database and converted to Chaos-XML using the Bio-chaos tools. The features were annotated to the deepest concept in the ontology possible, given the available information. For example, the degree of information in annotations was sufficiently deep to describe the transcript features with the type of RNA such as mRNA, or tRNA. It was therefore possible to restrict the analysis to given types of transcript. CGL tools were used to validate each of the annotations, iterate through the genes and query the features. EM-operators were applied to the part features of genes.
Other organism data was derived from the genomes section of GenBank [37]. GenBank flat files were converted to SO-compliant Chaos-XML using the script cx-genbank2chaos.pl (available from [19]) and BioPerl [23]. The BioPerl GenBank parser, Bio::SeqIO::genbank was used to convert GenBank flat files to Bioperl SeqFeature objects. Feature_relationships between these objects were inferred from location information using the Bioperl Bio::SeqFeature::Tools::Unflattener code. GenBank Feature Table types were converted to SO terms using the Bio::SeqFeature::Tools::TypeMapper class, which contains a hardcoded mapping for the subset of the GenBank Feature Table which is currently used in the genomes section of GenBank. The same Perl class was used to type the feature_relationships according to SO relationship types. The EM analysis was performed over the Chaos-XML annotations using the CGL suite of modules to iterate over the parts of each gene.
Figures and Tables
Figure 1 A section of the Sequence Ontology showing how terms and relationships are used together to describe knowledge about sequence. The kind_of relationships are depicted using arrows labeled with 'i', the part_of relationships use arrows with 'P' and the derives_from relationships with 'd'. By tracing the arrows that connect the terms, different logical inferences can be made regarding what a term 'is' and what are its allowable parts. For example, an exon is a part_of a transcript, a tRNA is a kind_of ncRNA which is a kind_of processed_transcript.
Figure 2 Using EM operations to characterize alternatively spliced transcripts and their exons. The EM operations overlap and disjoint can be used to characterize pair-wise relationships between alternative transcripts. Binary product and difference, on the other hand, pertain to exons shared, or not-shared between two alternative transcripts.
Figure 3 Examples of alternatively spliced genes from Entrez Gene at the NCBI. Of the seven classes of alternatively spliced genes, some classes are more likely to indicate annotation problems than others - particularly those genes having one or more sequence-disjoint transcripts. Parts-disjoint transcripts, on the other hand, are more suggestive of complex biology. Alternatively spliced genes having only overlapping transcripts (0:0:N) comprise the vast majority of instances.
Figure 4 A series of Venn diagrams showing the relationship between exon class and coding potential. An exon may be fully protein coding, partially protein coding, or be fully UTR. An exon may be a part_of a single transcript gene (single-transcript genes), be a part_of either one (UNIQUE exons), all (ALWAYS_FOUND exons), or a fraction (SOMETIMES_FOUND exons) of transcripts in an alternatively transcribed gene.
Table 1 Six subclasses of part-whole relationships
Part_of subtype Whole Properties of relationship Example
component_part_of integral object Functional/heteromerous/separable A leg is a part_of a body.
A regulatory_region is a part_of a gene.
portion_part_of mass Not functional/homomerous/separable A slice is a part_of a cake.
A restriction_fragment is part_of a chromosome.
stuff_part_of object Not functional/heteromerous/not separable Carbon is a part_of a chromosome.
member_part_of collection Not functional/heteromerous/separable A sheep is a part_of a flock.
A read is a part_of a contig.
place_part_of area Not functional/homomerous/not separable England is a part_of Britain.
feature_part_of activity Functional/heteromerous/not separable Inhaling is a part_of breathing.
Translation is part_of protein synthesis.
Column 1 gives the name of the subclass; column 2, the class or 'whole' to which such parts belong; column 3, the essential properties that define that particular part-whole relationship; and column 4 provides examples. Of the six classes only two - component_part_of and member_part_of occur in SO.
Table 2 The EM operators
EM operation Definition
Overlap (x ○ y) x and y overlap if they have a part in common.
Disjoint (x ι y) x and y are disjoint if they share no parts in common.
Binary product (x . y) The parts that x and y share in common.
Difference (x - y) The largest portion of x which has no part in common with y.
Binary sum (x + y) The set consisting of individuals x and y.
In each case x and y refer to two wholes. The first two operators are Boolean and pertain to whether two wholes share any parts in common; whereas the remainder return either the parts, or, in the case of binary sum, the wholes, that satisfy the operation.
Table 3 Percentage of each of the seven EM-based classes among the alternatively spliced genes in the D. melanogaster genome
Class D. melanogaster
N:0:0 0%
N:N:0 0%
N:0:N 0%
N:N:N 0%
0:N:0 7.70%
0:N:N 1.83%
0:0:N 90.47%
The number of genes with one or more pairs of sequence-disjoint transcripts, no pairs of parts-disjoint transcripts, and no pairs of overlapping transcripts - denoted as N:0:0 - is given in the first row. Row 2 gives the number of genes having both sequence-disjoint and parts-disjoint transcripts, but no overlapping transcripts - these are N:N:0 genes. Rows 3 to 7 detail the counts for each of the remaining possible classes.
Table 4 Summary of the types of exons present in each of the genomes and their functions
Exon part of gene with single transcript Exon part of one transcript of alternatively spliced gene (UNIQUE) Exon part of fraction of alternatively spliced transcripts (SOMETIMES_FOUND) Exon part of all of the transcripts of alternatively spliced gene (ALWAYS_FOUND)
Percentage of all exons 60.1% 16.1% 5.2% 18.6%
Coding 94.5% 68% 73% 93%
Non-coding 4.5% 32% 19% 3.5%
Coding/non-coding - - 8% 3.5%
Exons of alternatively spliced genes were divided into three categories based on the binary product and difference operations. UNIQUE exons (column 2) occur in only a single transcript; SOMETIMES_FOUND exons (column 3) occur in some, but not all of a gene's alternatively spliced transcripts. ALWAYS_FOUND exons occur in every alternative transcript. The table rows show the breakdown of each exon class with respect to function, i.e., coding exons are those that consist at least partially of translated nucleotides, whereas non-coding exons consist entirely of UTR sequence. In some genes, an exon may be coding in one transcript and non-coding in another, depending on the annotated start and stop codons and the phase of the upstream intron; these exons are denoted as coding/non-coding exons. For reference purposes, the breakdown of exons in single-transcript genes is shown in column 1.
==== Refs
Genbank
The Institute for Genome Research
Joint Genome Institute
Misra S Crosby MA Mungall CJ Matthews BB Campbell KS Hradecky P Huang Y Kamiker JS Millburn GH Prochnik SE Annotation of the Drosophila melanogaster euchromatic genome: a systematic review. Genome Biol 2002 3 research0083.1 0083.22 12537572 10.1186/gb-2002-3-12-research0083
Stein L Sternberg P Durbin R Thierry-Mieg J Spieth J WormBase: network access to the genome and biology of Caenorhabditis elegans. Nucleic Acids Res 2001 29 82 86 11125056 10.1093/nar/29.1.82
Rhee SY Beavis W Berardini TZ Chen G Dixon D Doyle A Garcia-Hernandez M Huala E Lander G Montoya M The Arabidopsis Information Resource (TAIR): a model organism database providing a centralized, curated gateway to Arabidopsis biology, research materials and community. Nucleic Acids Res 2003 31 224 228 12519987 10.1093/nar/gkg076
Dwight SS Balakrishnan R Christie KR Costanzo MC Dolinski K Engel SR Feierbach B Fisk DG Hirschman J Hong EL Saccharomyces genome database: underlying principles and organization. Brief Bioinform 2004 5 9 22 15153302 10.1186/1471-2105-5-9
Gene Ontology Consortium Creating the gene ontology resource: design and implementation. Genome Res 2001 11 1425 1433 11483584 10.1101/gr.180801
Open Biomedical Ontologies
The Sequence Ontology
OBO-Edit
DDBJ/EMBL/GenBank Feature Table documentation
Kulikova T Aldebert P Althorpe A The EMBL Nucleotide Sequence Database. Nucleic Acids Res 2004 32 D27 D30 14681351 10.1093/nar/gkh120
Miyazaki S Sugawara H Ikeo K Gojobori T Tateno Y DDBJ in the stream of various biological data. Nucleic Acids Res 2004 32 D31 D34 14681352 10.1093/nar/gkh127
Alberts B Johnson A Lewis J Raff M Roberts K Walter P Molecular Biology of the Cell 2002 4 New York: Garland
Generic Feature Format
Chado schema
Generic Model Organism Database
Chaos-XML
Comparative Genomics Library
Egenhofer MJ A formal definition of binary topological relationships. Lecture Notes Comp Sci 1989 367 457 472
Altschul SF Gish W Miller W Myers EW Lipman DJ Basic local alignment search tool. J Mol Biol 1990 215 403 410 2231712 10.1006/jmbi.1990.9999
Stajich JE Block D Boulez K Brenner SE Chervitz SA Dagdigian C Fuellen G Gilbert JG Korf I Lapp H The Bioperl toolkit: Perl modules for the life sciences. Genome Res 2002 12 1611 1618 12368254 10.1101/gr.361602
Smith B Ceusters W Köhler J Kumar A Lomax J Mungall CJ Neuhaus F Rector A Rosse C Relations in biological ontologies. Genome Biol 2005
Winston M Chaffin R Herrmann A taxonomy of part-whole relations. Cog Sci 1987 11 417 444 10.1016/S0364-0213(87)80015-0
Simons P Parts - A Study in Ontology 1987 Oxford: Clarendon Press
Husserl E Logical Investigations 1970 II London: Routledge & Keagan Paul
Flybase Re-annotation guideline
Entrez Gene
Hanke PD Storti RV The Drosophila melanogaster tropomyosin II gene produces multiple proteins by the use of alternate tissue specific promoters and alternate splicing. Mol Cell Biol 1988 8 3591 3602 2851721
Lewis BP Green RE Brenner SE Evidence for the widespread coupling of alternative splicing and nonsense-mediated mRNA decay in humans. Proc Natl Acad Sci USA 2003 100 189 192 12502788 10.1073/pnas.0136770100
Celniker CE Rubin GE The Drosophila melanogaster genome. Annu Rev Genomics Hum Genet 2003 4 89 117 14527298 10.1146/annurev.genom.4.070802.110323
Cruse DA Lexical Semantics 1986 Cambridge, UK: Cambridge University Press
Sequence Ontology
FlyBase release 3.2
Mungall CJ Misra S Berman BP Carlson J Frise E Harris N Marshall B Shu S Kaminker JS Prochnik SE An integrated computational pipeline and database to support whole-genome sequence annotation. Genome Biol 2002 3 research0081.1 0081.11 12537570 10.1186/gb-2002-3-12-research0081
Genomes Division of GenBank
| 15892872 | PMC1175956 | CC BY | 2021-01-04 16:05:39 | no | Genome Biol. 2005 Apr 29; 6(5):R44 | utf-8 | Genome Biol | 2,005 | 10.1186/gb-2005-6-5-r44 | oa_comm |
==== Front
Genome BiolGenome Biology1465-69061465-6914BioMed Central London gb-2005-6-5-r451589287310.1186/gb-2005-6-5-r45MethodSynthetic lethal analysis of Caenorhabditis elegans posterior embryonic patterning genes identifies conserved genetic interactions Baugh L Ryan [email protected] Joanne C 1Hill Andrew A 2Slonim Donna K 24Brown Eugene L 2Hunter Craig P [email protected] Department of Molecular and Cellular Biology, Harvard University, Cambridge, MA 02138, USA2 Department of Genomics, Wyeth Research, Cambridge, MA 02140, USA3 Current address: Biology Division, California Institute of Technology, Pasadena, CA 911254 Current address: Department of Computer Science, Tufts University, Medford, MA 021552005 11 4 2005 6 5 R45 R45 21 10 2004 24 12 2004 9 3 2005 Copyright © 2005 Baugh et al.; licensee BioMed Central LtdThis is an Open Access article distributed under the terms of the Creative Commons Attribution License (), which permits unrestricted use, distribution, and reproduction in any medium, provided the original work is properly cited.
To identify interactions among genes implicated in posterior patterning of the C. elegans embryo, synthetic lethality following RNA interference of 22 genes was measured in 15 mutant strains. A pair of homologous T-Box transcription factors was found to interact in both C. elegans and C. briggsae, indicating that their compensatory function is conserved.
Phenotypic robustness is evidenced when single-gene mutations do not result in an obvious phenotype. It has been suggested that such phenotypic stability results from 'buffering' activities of homologous genes as well as non-homologous genes acting in parallel pathways. One approach to characterizing mechanisms of phenotypic robustness is to identify genetic interactions, specifically, double mutants where buffering is compromised. To identify interactions among genes implicated in posterior patterning of the Caenorhabditis elegans embryo, we measured synthetic lethality following RNA interference of 22 genes in 15 mutant strains. A pair of homologous T-box transcription factors (tbx-8 and tbx-9) is found to interact in both C. elegans and C. briggsae, indicating that their compensatory function is conserved. Furthermore, a muscle module is defined by transitive interactions between the MyoD homolog hlh-1, another basic helix-loop-helix transcription factor, hnd-1, and the MADS-box transcription factor unc-120. Genetic interactions within a homologous set of genes involved in vertebrate myogenesis indicate broad conservation of the muscle module and suggest that other genetic modules identified in C. elegans will be conserved.
==== Body
Background
Forward and reverse genetic screens in flies and worms indicate that most genes are not essential to the development or viability of the organism [1-4]. Two primary explanations for such phenotypic robustness to mutation have been offered: homologous gene products may directly compensate for one another's function, or indirect compensation of function may occur through regulatory networks via non-homologous genes acting in alternative pathways or feedback mechanisms [5,6]. In both Saccharomyces cerevisiae and C. elegans, genes with at least one homolog are less likely than unique genes to have a loss-of-function phenotype [7,8]. However, homology accounts for no more than two thirds of the observed phenotypic robustness to mutation in S. cerevisiae and even less in C. elegans, indicating a significant role of the regulatory network in genetic buffering [7,8]. By identifying gene disruptions that are viable in wild type but lethal in a specific mutant background, synthetic lethal screens can shed light on how regulatory networks buffer gene function [9]. Although genetic interactions are being mapped on a genome-wide scale in yeast [9], no such efforts have been reported for an animal system. We report here the use of existing mutants and RNA interference (RNAi) to assemble a synthetic lethal matrix in C. elegans. Our aim was to build on prior knowledge by focusing on a characterized set of genes likely to have interactions among them.
The C blastomere is one of five somatic founder blastomeres in the C. elegans embryo. Each founder blastomere produces a characteristic set of cell types via an invariant lineage; the C blastomere predominantly gives rise to muscle and epidermal cells [10]. Maternal and zygotic activities of the homeodomain protein PAL-1 specify the identity and maintain the development of the C-blastomere lineage [11,12]. We have identified genes whose expression depends either directly or indirectly on PAL-1 function ('PAL-1 targets') and shown that they affect specification, differentiation, and morphogenesis of C-lineage cells. PAL-1 targets include many transcription factors and based on their temporal and spatial expression patterns in wild type, we have proposed a model for their regulatory relationships [13]. However, RNAi of most PAL-1 targets does not result in a detectable phenotype [13]. Because the regulatory network specified by PAL-1 appears to function autonomously while controlling multiple, discrete developmental processes (that is, specification and morphogenesis of muscle and epidermal cells) [11,13,14], we reasoned that the network is organized in a modular fashion and that PAL-1 targets may have overlapping or compensatory functions. We set out to measure synthetic lethality for as many pairs of PAL-1 targets as possible, and we included three additional genes encoding transcriptional regulators (lin-26, unc-62, and ceh-40) whose expression patterns and phenotypes suggest that they may have genetic interactions with PAL-1 or its targets [15,16].
Results and discussion
A synthetic lethal matrix
We began our analysis with 26 genes, 23 PAL-1 targets and three genes likely to have genetic interactions with PAL-1 targets. Mutations in 15 of these genes have been isolated, and appropriate strains were obtained from the Caenorhabditis Genetics Center (Table 1). To screen for synthetic lethal interactions, RNAi of the 13 nonessential genes was performed and embryonic lethality scored in each of the mutant strains as well as wild type to generate a synthetic lethal matrix (Figure 1 and Additional data file 1). However, RNAi of a given gene could not be used if it results in 100% lethality in wild type (pal-1, elt-1, unc-62, and lin-26), and so the matrix is not entirely symmetric. To reduce the penetrance of lethality following RNAi for these four genes we used a lower concentration of double-stranded RNA (dsRNA), but the results were variable (data not shown), and we decided that it was an unreliable approach. We also tried to assemble the matrix using exclusively RNAi, but we found that soaking worms in two different dsRNA molecules targeting two different genes at once to be unacceptably inefficient (Additional data file 2).
Given the effects of mutation alone without RNAi, and the effects of RNAi in wild type, it may not be obvious from the raw data which disruptions result in significantly more lethality when paired than when alone. Measurements of lethality were thus duplicated and a t-test was used to assign statistical significance to each of the interactions tested (Figure 1 and Additional data file 1). Seven genetic interactions were identified (P < 0.001) among six genes out of 195 total interactions tested. In addition to the genes shown in Figure 1, RNAi of the remaining genes for which no mutant was available was performed in all 16 strains. Because synthetic lethality was measured only a single time for each of these possible interactions, the data could not be subjected to statistical analysis. However, RNAi of these additional genes did not appear to elevate lethality in any of the genetic backgrounds tested (data not shown).
Conservation of a genetic interaction
The T-box family of proteins regulates diverse aspects of animal development, including cell fate specification, differentiation, and morphogenesis [17]. The C. elegans genome contains 21 of these animal-specific genes, compared to 32 in C. briggsae and eight in D. melanogaster [18,19], suggesting duplication or retention within nematodes. tbx-37 and tbx-38 have been shown to redundantly control embryonic patterning of the C. elegans AB lineage [20], and our systematic approach identified tbx-8 and tbx-9 as a synthetic lethal pair. tbx-8 and tbx-9 are expressed in nearly identical temporal and spatial patterns in the early embryo, with the strongest expression detected in dorsal epidermal cells and posterior expression depending on pal-1 function [13,21,22]. tbx-8(ok656), tbx-8(RNAi), and tbx-9(RNAi) each result in a low frequency of embryonic lethality (less than 5%) with 1-10% of hatching larvae displaying posterior morphological defects (Figure 2) [21,22]. However, 50% of tbx-8(ok656); tbx-9(RNAi) embryos fail to hatch and those that do display severe morphological defects and die as larvae (Figure 2). Phenotypic analysis demonstrates that tbx-8 and tbx-9 are required for intercalation of dorsal epidermal cells thus explaining the morphological defects observed following their disruption [22].
The overlapping function of tbx-8 and tbx-9 is not the residual result of a recent gene duplication event. The C. briggsae genome contains a pair of syntenic orthologs (by best reciprocal blastp match) [18], and as in C. elegans, the C. briggsae orthologs have overlapping function during morphogenesis (Table 2), indicating that the functional relationship between this pair of genes has been selectively maintained for over 100 million years. This observation suggests that the overlapping functions of other pairs of T-box genes, which are likely to be common throughout the Metazoa, may also be conserved.
A muscle differentiation module
It has been suggested that modules of genes comprise units of biological function [23]. Because they operate in concerted fashion, genes in a module are expected to have increased likelihood of being co-expressed and sharing physical and genetic interactions relative to a random set of genes. Clustering a large synthetic lethal matrix in S. cerevisiae identified multiple functional modules [9], and we identify a module in our matrix around the hlh-1 gene (Figure 1). We detect five genetic interactions (P < 0.001) out of six tested among hlh-1, hnd-1, and unc-120. All three genes encode transcription factors expressed exclusively in the muscle progenitor cells of the early embryo [13,24,25]. Disruption of function by mutation or RNAi of any one of the three genes results in a low frequency of embryonic lethality with embryos arresting paralyzed at the two-fold stage (Pat) (Figure 3). Embryos require muscle contraction to elongate and pass from the two-fold to three-fold embryonic stage, and the Pat phenotype results from a complete failure of muscle differentiation [26,27]. A large fraction of animals that do hatch after disruption of hlh-1, unc-120, or hnd-1 show a less severe phenotype (uncoordinated, dumpy larvae) likely reflecting partial muscle function [26,27] (Figure 3). However, disruption of function of any two of the three genes significantly elevates the frequency of Pat embryos, with disruption of hlh-1 and unc-120 resulting in 100% Pat embryos (Figure 3). The fact that hlh-1, unc-120, and hnd-1 have identical individual and synthetic phenotypes, in addition to their identical early embryonic expression patterns, indicate that they comprise a muscle-differentiation module. Although fewer genes define this module than those identified in larger screens in yeast [9], the specificity of its defining phenotype indicates that hlh-1, unc-120, and hnd-1 function in concert, and it is likely that future synthetic screens will identify additional members of the module.
If both RNAi and genetic disruptions of gene function are null, the synthetic lethal matrix should be symmetric. However, we can only test the pair of reciprocal interactions when mutations of both genes are available and RNAi of neither gene results in 100% lethality in wild type. Although there is some symmetry in the matrix around hlh-1 (Figure 1), the interactions detected are not all reciprocal (Figure 3). The efficiency of RNAi varies from gene to gene and often fails to result in null phenotypes, suggesting a significant number of false negatives among the interactions tested and explaining the imperfect symmetry among the interactions detected. The pattern of interactions defining the muscle module suggests that hlh-1 is the most essential of the three genes, or that its function is the most potent, and hnd-1 the least (Figure 3), a model that accounts for the reported lack of genetic interaction between heterozygous hlh-1 and homozygous hnd-1 mutations [25].
The functional importance of the muscle module we identified is underscored by its conservation. Well characterized genetic interactions exist among the vertebrate homologs of hlh-1 - the myogenic basic helix-loop-helix (bHLH) regulatory factors MyoD, myf5, myogenin, and MRF4 [28]. In addition, interactions between vertebrate myogenic bHLH genes and homologs of unc-120 (the MEF2 group of MADS-box regulators) have been described [29]. Interactions among and between vertebrate myogenic bHLH and MADS-box regulators is likely to be the result of cooperative activation of common target genes as well as their ability to activate each other transcriptionally [28-30]. It is possible that hlh-1, hnd-1, and unc-120, like their vertebrate counterparts, display modular genetic interactions because they share common target genes, and they may also activate each other transcriptionally.
Conclusions
We show that the facility of RNAi in C. elegans can be used to screen a large number of genes for synthetic phenotypes. Using an unbiased experimental approach we identified both easy-to-predict genetic interactions between similar co-expressed genes and difficult-to-predict genetic interactions between non-homologous genes. We also used reverse genetics in C. briggsae to show that genetic interaction between a pair of co-expressed homologous transcription factors (tbx-8 and tbx-9) is conserved, suggesting an adaptive role of overlapping function as opposed to the transient result of gene duplication. Furthermore, we identify a genetic module defined by reciprocal interactions among three transcription factors that is required for differentiation of muscle. A similar network among homologous genes is conserved in vertebrates, suggesting that a single master regulator cannot robustly control muscle development. Given the wealth of existing mutants and the ease of RNAi in C. elegans, our approach can be scaled up to identify the core set of metazoan developmental genetic interactions.
Materials and methods
DNA templates for in vitro transcription were prepared by amplification of 100-1000 base pair (bp) sequences from cDNA by PCR. PCR primers were designed to amplify unique sequences in order to ensure specificity of RNAi (Additional data file 3), and a minimal T7 promoter sequence was added to the 5' end of both primers. PCR DNA was purified on a DNA Clean and Concentrator-5 column (Zymo Research), and a high-yield in vitro transcription kit (Promega) was used to produce dsRNA. The dsRNA was reannealed by heating to 90°C and then cooling 6°C/min to 25°C before organic extraction with a 1:1 mixture of phenol and chloroform. Following ethanol precipitation, dsRNA was dissolved in Soaking Buffer [31] at high concentration, which was then measured by spectrophotometer and diluted to 5 mg/ml.
Eight to twelve larval stage 4 (L4) worms were soaked in 0.75 μl 5 mg/ml dsRNA for 24 h at 20°C and then transferred to a recovery plate with Escherichia coli strain OP50 for 12 h at 25°C and then transferred to a score plate with OP50. After 12 h at 25°C on the score plate, adults were removed, and after another 24 h at 25°C, embryos and larvae were counted under a dissecting microscope (with the exception of unc-120(st364) which is temperature-sensitive and was therefore scored at 15°C). For strains carrying viable alleles, embryonic lethality was scored for the homozygous progeny of homozygous parents. For strains carrying 100% lethal alleles, embryonic lethality was scored for the mixed progeny of either heterozygous or hemizygous parents. For RNAi in C. briggsae, 1 mg/ml dsRNA was injected into the germline of young wild-type adults (strain AF16), and the worms were rescued and scored in the same way as the soaked C. elegans.
For statistical analysis of the data, percent embryonic lethality was converted to survival (1 - % embryonic lethality). All survival values were normalized by the survival of N2 (wild type) in Soaking Buffer (no dsRNA). For each genetic interaction tested, the null hypothesis used is that the survival of the double disruption (mutation and RNAi) equals the product of survivals for each single disruption (mutation alone × RNAi alone). The pairs of duplicate survival measurements for each single disruption are multiplied to generate four values for the expected survival of the double disruption, and a t-test is used to determine if the observed survival (n = 2) is significantly different from expected (n = 4). For example, if two genes function independently, and the survival following disruption of each alone is 90%, then our model expects 81% survival following double disruption since the 90% of individuals surviving one disruption have a 10% chance of dying from the other disruption. If survival is significantly different from the expected 81% then we conclude that the two genes interact.
For microscopy, embryos were mounted on 2% agar pads under coverslips and allowed to develop in a humid box for 12-15 h at 25°C. A Zeiss Axiophot with 100× differential interference contrast optics was used and digital images captured.
Additional data files
The following additional data are available with the online version of this paper. Additional data file 1 contains replicate embryonic lethality measurements and corresponding P-values for synthetic lethality. Additional data file 2 contains data showing that double RNAi by soaking worms in dsRNA for two genes at once is an inefficient means of synthetic genetic analysis. Additional data file 3 lists primer sequences used in this study to amplify coding sequences from cDNA for use as templates for dsRNA.
Supplementary Material
Additional File 1
Replicate embryonic lethality measurements and corresponding P-values for synthetic lethality. Replicate embryonic lethality measurements and corresponding P-values for synthetic lethality.
Click here for file
Additional File 2
Double RNAi by soaking worms in dsRNA for two genes at once is an inefficient means of synthetic genetic analysis. Percent embryonic lethality measured by both double RNAi and RNAi in mutant backgrounds is presented for comparison. Each column corresponds to RNAi of a given gene, and each row corresponds to either RNAi or a mutant allele of a given gene. Each value corresponds to a specific combination of perturbations, either RNAi of two genes or RNAi and mutation, with the controls as exceptions. A genetic interaction is detected for the highly homologous genes tbx-8 and tbx-9 with both methods, but none of the strong interactions between the three relatively non-homologous myogenic regulators (hlh-1, unc-120, and hnd-1) are detected by double RNAi. Each value presented is the average of two independent experiments; at least 100 progeny were counted for each experiment. dsRNA concentration was 5 mg/ml for all experiments. NA stands for not applicable.
Click here for file
Additional File 3
A table listing primer sequences used in this study to amplify coding sequences from cDNA for use as templates for dsRNA. A minimal T7 promoter sequence (TAATACGACTCACTATAGGG) was added to the 5' end of each of the gene-specific sequences listed in the table so that PCR products could be used as in vitro transcription templates to make dsRNA.
Click here for file
Acknowledgements
We thank the Kimble lab for providing strain JK3276, and the Caenorhabditis Genetics Center for providing the other strains used in this work. This work was funded by an NIH grant GM64429 to C.P.H.
Figures and Tables
Figure 1 A synthetic lethal matrix of genes implicated in posterior patterning. (a) A heat map of survival (1 - % embryonic lethality) for RNAi of 13 genes (and a soaking buffer control) in 16 different genetic backgrounds. Values plotted are the average of two measurements (see Additional data file 1 for raw data). (b) A heat map of P-values corresponding to each test of synthetic lethality. A t-test was used with the null hypothesis that the survival of a given combination of RNAi and mutation is equal to the product of survivals of the RNAi in wild type and the mutation without RNAi (P-values can be found in Additional data file 1). A dashed gray line outlines the square (symmetric) part of the matrix and its diagonal in (a) and (b). RNAi of genes resulting in 100% embryonic lethality in wild type are not included in the matrix (pal-1, elt-1, lin-26, and unc-62). Mutant genotypes are listed in Table 1.
Figure 2 Synthetic lethal phenotypes of the homologous transcription factors tbx-8 and tbx-9. Recently hatched L1 larvae or an arrested embryo are shown for (a) wild-type, (b) tbx-9(RNAi), (c) tbx-8(ok656), and (d, e) tbx-8(ok656); tbx-9(RNAi). Panels (b) to (e) reflect the increasing severity of the phenotype, with (b) and (c) providing examples of the posterior morphological defects observed at low frequency after disruption of tbx-8 or tbx-9. (d, e) Examples of the severe morphological defects observed at high frequency after disruption of both tbx-8 and tbx-9; (d) shows an individual that has hatched but will arrest as a larva, and (e) shows an embryo that has arrested before hatching (embryonic lethal). The anterior is cropped in (b) and (c). The scale bar in (a) equals 10 μm and applies to (a-e).
Figure 3 Modular genetic interactions between three transcription factors controlling differentiation of muscle. A recently hatched L1 larva is shown for (a) wild type and (b) hlh-1(cc561), and (c) an arrested Pat embryo is shown for hlh-1(cc561); unc-120(RNAi). The scale bar in (a) equals 10 μm and applies to (a-c). Mutation of hlh-1, hnd-1 or unc-120 alone results in dumpy, uncoordinated larvae and Pat embryos (as in (b) and (c)) at low frequency. Disruption of function of any two of these three genes by mutation and RNAi significantly elevates both frequencies so that in the most potent combinations 100% Pat embryos result. The Pat phenotype has been shown to result specifically from the disruption of muscle differentiation and function [26,27]. (d) A summary of the proportion hatching (green) and proportion arresting as Pat embryos (black) for each single mutant and each of the six genetic interactions tested.
Table 1 Genes included in the synthetic lethality matrix
ORF Gene Identification Genotype used in this study
F17A2.5 ceh-40 Homeodomain protein (Extradenticle ortholog) ceh-40(ok740) X
K10B4.6 cwn-1 Putative Wnt ligand cwn-1(ok546) II
D1081.2 unc-120 MADS domain TF unc-120(st364) I
B0304.1 hlh-1 bHLH TF hlh-1(cc561) II
C44C10.8 hnd-1 Hand bHLH TF hnd-1(q740) X
F35G12.6 mab-21 Highly conserved novel protein mab-21(bx53)III; him-5(e1490)V
F11C1.6 nhr-25 Nuclear hormone receptor +/szT1 [lon-2(e678)] I; nhr-25(jm2389)/szT1 X
Y75B8A.2 nob-1 Homeodomain TF (posterior Hox paralog) nob-1(ct223) dpy-18(e364) unc-25(e156) III; eDp6 (III;f)
K02B9.4 elt-3 GATA TF elt-3(gk121) X
T07C4.2 tbx-8 T-box TF (Brachyury) tbx-8(ok656) III
M142.4 vab-7 Homeobox TF (even-skipped subfamily) vab-7(e1562) III
T07C4.6 tbx-9 T-box TF (Brachyury) Mutant not available
C55C2.1 Zinc-finger TF Mutant not available
C38D4.6 pal-1 Homeobox TF (cad subfamily) pal-1(ct224) dpy-17(e164) ncl-1(e1865) unc-36(e251) III; sDp3(f:III); lin-2(e1309) X
W09C2.1 elt-1 GATA TF elt-1(zu180) unc-43(e408)/unc-24(e138) dpy-20(e1282) IV
F18A1.2 lin-26 Zinc-finger TF lin-26(n156) II
T28F12.2 unc-62 Meis-class homeodomain (Homothorax ortholog) unc-62(e644) V
C28C12.7 spp-10 Predicted prosaposin Mutant not available
R02D3.1 Dehydrogenase Mutant not available
C09D4.2 Uncharacterized Mutant not available
R07C3.11 Uncharacterized Mutant not available
T22B7.3 Uncharacterized Mutant not available
C46H11.2 Uncharacterized Mutant not available
ZK1307.1 Uncharacterized Mutant not available
C01G6.1 aqp-2 Aquaporin Mutant not available
T27D12.1 Sodium/phosphate transporter Mutant not available
Most of the genes were included in the matrix because they had been genetically determined to be PAL-1 targets (that is, their expression either directly or indirectly depends on PAL-1 function) and were therefore presumed to have genetic interactions among them. Three additional genes encoding transcriptional regulators (lin-26, ceh-40, and unc-62) were included as their phenotypes and expression patterns suggest that they may have genetic interactions with PAL-1 targets. TF, transcription factor.
Table 2 Functional redundancy of tbx-8 and tbx-9 is conserved in C. briggsae
Genotype C. elegans C. briggsae
Embryonic lethality Morphological defect Wild type larvae Embryonic lethality Morphological defect Wild type larvae
tbx-8(RNAi) 1% 1% 98% 4% 1% 96%
tbx-9(RNAi) 1% 1% 98% 3% 2% 95%
tbx-8(RNAi);tbx-9(RNAi) 48% 52% 0% 39% 46% 15%
RNAi was used for both genes in both species as no mutations are available for C. briggsae.
==== Refs
Nusslein-Volhard C Of flies and fishes. Science 1994 266 572 574 7939708
Riddle DL C elegans II 1997 Plainview, NY: Cold Spring Harbor Laboratory Press
Boutros M Kiger AA Armknecht S Kerr K Hild M Koch B Haas SA Consortium HF Paro R Perrimon N Genome-wide RNAi analysis of growth and viability in Drosophila cells. Science 2004 303 832 835 14764878 10.1126/science.1091266
Kamath RS Ahringer J Genome-wide RNAi screening in Caenorhabditis elegans. Methods 2003 30 313 321 12828945 10.1016/S1046-2023(03)00050-1
Hartman JLt Garvik B Hartwell L Principles for the buffering of genetic variation. Science 2001 291 1001 1004 11232561 10.1126/science.291.5506.1001
Wagner A Robustness against mutations in genetic networks of yeast. Nat Genet 2000 24 355 361 10742097 10.1038/74174
Gu Z Steinmetz LM Gu X Scharfe C Davis RW Li WH Role of duplicate genes in genetic robustness against null mutations. Nature 2003 421 63 66 12511954 10.1038/nature01198
Conant GC Wagner A Duplicate genes and robustness to transient gene knock-downs in Caenorhabditis elegans. Proc R Soc Lond B Biol Sci 2004 271 89 96 15002776 10.1098/rspb.2003.2560
Tong AH Lesage G Bader GD Ding H Xu H Xin X Young J Berriz GF Brost RL Chang M Global mapping of the yeast genetic interaction network. Science 2004 303 808 813 14764870 10.1126/science.1091317
Sulston JE Schierenberg E White JG Thomson JN The embryonic cell lineage of the nematode Caenorhabditis elegans. Dev Biol 1983 100 64 119 6684600 10.1016/0012-1606(83)90201-4
Hunter CP Kenyon C Spatial and temporal controls target pal-1 blastomere-specification activity to a single blastomere lineage in C. elegans embryos. Cell 1996 87 217 226 8861906 10.1016/S0092-8674(00)81340-9
Edgar LG Carr S Wang H Wood WB Zygotic expression of the caudal homolog pal-1 is required for posterior patterning in Caenorhabditis elegans embryogenesis. Dev Biol 2001 229 71 88 11133155 10.1006/dbio.2000.9977
Baugh LR Hill AA Claggett JM Hill-Harfe K Wen JC Slonim DK Brown EL Hunter CP The homeodomain protein PAL-1 specifies a lineage-specific regulatory network in the C. elegans embryo. Development 2005 132 1843 1845 15772128 10.1242/dev.01782
Draper BW Mello CC Bowerman B Hardin J Priess JR MEX-3 is a KH domain protein that regulates blastomere identity in early C. elegans embryos. Cell 1996 87 205 216 8861905 10.1016/S0092-8674(00)81339-2
Labouesse M Hartwieg E Horvitz HR The Caenorhabditis elegans LIN-26 protein is required to specify and/or maintain all non-neuronal ectodermal cell fates. Development 1996 122 2579 2588 8787733
Van Auken K Weaver D Robertson B Sundaram M Saldi T Edgar L Elling U Lee M Boese Q Wood WB Roles of the Homothorax/Meis/Prep homolog UNC-62 and the Exd/Pbx homologs CEH-20 and CEH-40 in C. elegans embryogenesis. Development 2002 129 5255 5268 12399316
Showell C Binder O Conlon FL T-box genes in early embryogenesis. Dev Dyn 2004 229 201 218 14699590 10.1002/dvdy.10480
Stein LD Bao Z Blasiar D Blumenthal T Brent MR Chen N Chinwalla A Clarke L Clee C Coghlan A The genome sequence of Caenorhabditis briggsae: a platform for comparative genomics. PLoS Biol 2003 1 E45 14624247 10.1371/journal.pbio.0000045
Riechmann JL Heard J Martin G Reuber L Jiang C Keddie J Adam L Pineda O Ratcliffe OJ Samaha RR Arabidropsis transcription factors: genome-wide comparative analysis among eukaryotes Science 2000 290 2105 2110 11118137 10.1126/science.290.5499.2105
Good K Ciosk R Nance J Neves A Hill RJ Priess JR The T-box transcription factors TBX-37 and TBX-38 link GLP-1/Notch signaling to mesoderm induction in C. elegans embryos. Development 2004 131 1967 1978 15056620 10.1242/dev.01088
Andachi Y Caenorhabditis elegans T-box genes tbx-9 and tbx-8 are required for formation of hypodermis and body-wall muscle in embryogenesis. Genes Cells 2004 9 331 344 15066124 10.1111/j.1356-9597.2004.00725.x
Pocock R Ahringer J Mitsch M Maxwell S Woollard A A regulatory network of T-box genes and the even-skipped homologue vab-7 controls patterning and morphogenesis in C. elegans. Development 2004 131 2373 2385 15102704 10.1242/dev.01110
Hartwell LH Hopfield JJ Leibler S Murray AW From molecular to modular cell biology. Nature 1999 402 Suppl C47 C52 10591225 10.1038/35011540
Krause M Fire A Harrison SW Priess J Weintraub H CeMyoD accumulation defines the body wall muscle cell fate during C. elegans embryogenesis. Cell 1990 63 907 919 2175254 10.1016/0092-8674(90)90494-Y
Mathies LD Henderson ST Kimble J The C. elegans Hand gene controls embryogenesis and early gonadogenesis. Development 2003 130 2881 2892 12756172 10.1242/dev.00483
Hresko MC Williams BD Waterston RH Assembly of body wall muscle and muscle cell attachment structures in Caenorhabditis elegans. J Cell Biol 1994 124 491 506 8106548 10.1083/jcb.124.4.491
Williams BD Waterston RH Genes critical for muscle development and function in Caenorhabditis elegans identified through lethal mutations. J Cell Biol 1994 124 475 490 8106547 10.1083/jcb.124.4.475
Weintraub H The MyoD family and myogenesis: redundancy, networks, and thresholds. Cell 1993 75 1241 1244 8269506 10.1016/0092-8674(93)90610-3
Yun K Wold B Skeletal muscle determination and differentiation: story of a core regulatory network and its context. Curr Opin Cell Biol 1996 8 877 889 8939680 10.1016/S0955-0674(96)80091-3
Molkentin JD Olson EN Combinatorial control of muscle development by basic helix-loop-helix and MADS-box transcription factors. Proc Natl Acad Sci USA 1996 93 9366 9373 8790335 10.1073/pnas.93.18.9366
Maeda I Kohara Y Yamamoto M Sugimoto A Large-scale analysis of gene function in Caenorhabditis elegans by high-throughput RNAi. Curr Biol 2001 11 171 176 11231151 10.1016/S0960-9822(01)00052-5
| 15892873 | PMC1175957 | CC BY | 2021-01-04 16:05:39 | no | Genome Biol. 2005 Apr 11; 6(5):R45 | utf-8 | Genome Biol | 2,005 | 10.1186/gb-2005-6-5-r45 | oa_comm |
==== Front
Genome BiolGenome Biology1465-69061465-6914BioMed Central London gb-2005-6-5-r461589287410.1186/gb-2005-6-5-r46MethodRelations in biomedical ontologies Smith Barry [email protected] Werner [email protected] Bert [email protected]öhler Jacob [email protected] Anand [email protected] Jane [email protected] Chris [email protected] Fabian [email protected] Alan L [email protected] Cornelius [email protected] Institute for Formal Ontology and Medical Information Science, Saarland University, D-66041 Saarbrücken, Germany2 Department of Philosophy, University at Buffalo, Buffalo, NY 14260, USA3 European Centre for Ontological Research, Saarland University, D-66041 Saarbrücken, Germany4 Department of Genetics, University of Leipzig, D-04103 Leipzig, Germany5 Rothamsted Research, Harpenden, AL5 2JQ, UK6 European Bioinformatics Institute, Hinxton, CB10 1SD, UK7 HHMI, Department of Molecular and Cellular Biology, University of California, Berkeley, CA 94729, USA8 Department of Computer Science, University of Manchester, M13 9PL, UK9 Department of Biological Structure, University of Washington, Seattle, WA 98195, USA2005 28 4 2005 6 5 R46 R46 28 10 2004 3 2 2005 31 3 2005 Copyright © 2005 Smith 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.
To enhance the treatment of relations in biomedical ontologies we advance a methodology for providing consistent and unambiguous formal definitions of the relational expressions used in such ontologies in a way designed to assist developers and users in avoiding errors in coding and annotation.
To enhance the treatment of relations in biomedical ontologies we advance a methodology for providing consistent and unambiguous formal definitions of the relational expressions used in such ontologies in a way designed to assist developers and users in avoiding errors in coding and annotation. The resulting Relation Ontology can promote interoperability of ontologies and support new types of automated reasoning about the spatial and temporal dimensions of biological and medical phenomena.
==== Body
Background
Controlled vocabularies in bioinformatics
The background to this paper is the now widespread recognition that many existing biological and medical ontologies (or 'controlled vocabularies') can be improved by adopting tools and methods that bring a greater degree of logical and ontological rigor. We describe one endeavor along these lines, which is part of the current reform efforts of the Open Biomedical Ontologies (OBO) consortium [1,2] and which has implications for ontology construction in the life sciences generally.
The OBO ontology library [1] is a repository of controlled vocabularies developed for shared use across different biological and medical domains. Thus the Gene Ontology (GO) [3,4] consists of three controlled vocabularies (for cellular components, molecular functions, and biological processes) designed to be used in annotations of genes or gene products. Some ontologies in the library - for example the Cell and Sequence Ontologies, as well as the GO itself - contain terms which can be used in annotations applying to all organisms. Others, especially OBO's range of anatomy ontologies, contain terms applying to specific taxonomic groups such as fly, fungus, yeast, or zebrafish.
Controlled vocabularies can be conceived as graph-theoretical structures consisting on the one hand of terms (which form the nodes of each corresponding graph) linked together by means of edges called relations. The ontologies in the OBO library are organized in this way by means of different types of relations. OBO's Mouse Anatomy ontology, for example, uses just one type of edge, labeled part_of. The GO currently uses two, labeled is_a and part_of. The Drosophila Anatomy ontology includes also a develops_from link. Other OBO ontologies include further links, for example (in the Sequence Ontology) position_of and disjoint_from. The National Cancer Institute (NCI) Thesaurus adds many additional links, including has_location for anatomical structures and different part_of relations for structures and for processes.
The problem is that when OBO and similar ontologies incorporate such relations they typically do so in informal ways, often providing no definitions at all, so that the logical interconnections between the various relations employed are unclear, and even the relations is_a and part_of are not always used in consistent fashion both within and between ontologies. Our task in what follows is to rectify these defects, drawing on the requirements analysis presented in [5].
Of the criteria that ontologies must currently satisfy if they are to be included in the OBO library, the most important for our purposes are: first, inclusion of textual definitions or descriptions designed to ensure that the precise meanings of terms as used within particular ontologies will be clear to a human reader; second, employment of a standard syntax, such as the OWL or OBO flatfile syntax; third, orthogonality to the other ontologies already included in the library. These criteria are designed to support the integration of OBO ontologies, above all by ensuring the compatibility of ontologies pertaining to an identical subject matter. OBO has now added a fourth criterion to assist in achieving such compatibility, namely that the relations (edges) used to connect terms in OBO ontologies should be applied in ways consistent with their definitions as set forth in this paper.
The Relation Ontology offered here is designed to put flesh on this criterion. How, exactly, should part_of or located_in be defined in order to ensure maximally reliable curation of each single ontology while at the same time guaranteeing maximal leverage in building a solid base for life-science knowledge integration in general? We describe a rigorous methodology for providing an answer to this question and illustrate its use in the construction of an easily extendible list of ten relations of a type familiar to those working in the bio-ontological field. This list forms the core of the new OBO Relation Ontology. What is distinctive about our methodology is that, while the relations are each provided with rigorous formal definitions, these definitions can at the same time be formulated in such a way that the underlying technical details remain invisible to ontology authors and curators.
Shortcomings of biomedical ontologies
While considerable effort has been invested in the formulation and definition of terms in biomedical ontologies, too little attention has been paid in the ontological literature to the associated relations. A number of characteristic types of shortcomings of controlled vocabularies can be traced back especially to the neglect of issues of formal structure in the treatment of relations [5-10]. To take just one example, the pre-2004 versions of GO allowed at least three different readings of the expression 'part of' as representing simultaneously: inclusion relations between vocabularies; a relation of possible parthood between biological entities; a relation of necessary parthood between biological entities. As was shown in [6], this coexistence of conflicting readings meant that three of the four rules given in the then effective documentation for reasoning with GO's hierarchies were logically incorrect.
Another characteristic family of problems turns on the paucity of resources for expressing relations in ontologies like GO. For example, because GO has no direct means of asserting location relations, it must capture such relations indirectly by constructing new terms involving syntactic operators such as 'site of', 'within', 'extrinsic to', 'space', 'region', and so on. It then simulates assertions of location by means of 'is_a' and 'part_of' statements involving such composites, for example in:
extracellular region is_a cellular component
extrinsic to membrane part_of membrane
both of which are erroneous. Additional problems arise from the fact that GO's extracellular region and extracellular space are both specified in their definitions as referring to the space (how large a space?) external to the outermost structure of a cell.
Another type of problem turns on the failure to distinguish relational expressions which, though closely related in meaning, are revealed to be crucially distinct when explicated in the formally precise way that is demanded by computer implementations. An example is provided by the simultaneous use in OBO's Cell Ontology of both derives_from and develops_from while no clear distinction is drawn between the two [11]. This problem is resolved in the treatment of derivation and transformation below, and has been correspondingly corrected in versions 1.14 and later of the Cell Ontology.
Efforts to improve GO from the standpoint of increased formal rigor have thus far been concentrated on re-expressing the existing GO schema in a description logic (DL) framework. This has allowed the use of a DL-reasoner that can identify certain kinds of errors and omissions, which have been corrected in later versions of GO [12]. DLs, however, can do no more than guarantee consistent reasoning according to the definitions provided to them. If the latter are themselves problematic, then a DL can do very little to identify or resolve the problems which result. Here, accordingly, we take a more radical approach, which consists in re-examining the basic definitions of the relations used in GO and in related ontologies in an attempt to arrive at a methodology which will lead to the construction of ontologies which are more fundamentally sound and thus more secure against errors and more amenable to the use of powerful reasoning tools. This approach is designed also to be maximally helpful to biologists by avoiding the problems which arise by virtue of the fact that the syntax favored in the DL-community is of a type which can normally be understood only by DL-specialists.
A theory of classes and instances
The relations in biological ontologies connect classes as their relata. The term 'class' here is used to refer to what is general in reality, or in other words to what, in the knowledge-representation literature, is typically (and often somewhat confusingly [13]) referred to under the heading 'concept' and in the literature of philosophical ontology under the headings 'universal', 'type' or 'kind'. Biological classes are in first approximation those classes which have been implicitly sanctioned through usage of the corresponding general terms in the biological literature, for example cell or fat body development.
Our task is to develop a suite of coherently defined bio-ontological relations that is sufficiently compact to be easily learned and applied, yet sufficiently broad in scope to capture a wide range of the relations currently coded in standard biomedical ontologies. Unfortunately the realization of this task is not a trivial matter. This is because, while the terms in biomedical ontologies refer exclusively to classes - to what is general in reality - we cannot define what it means for one class to stand to another, for example in the part_of relation, without taking the corresponding instances into account [6]. Here the term 'instance' refers to what is particular in reality, to what are otherwise called 'tokens' or 'individuals' - entities (including processes) which exist in space and time and stand to each other in a variety of instance-level relations. Thus we cannot make sense of what it means to say cell nucleus part_of cell unless we realize that this is a statement to the effect that each instance of the class cell nucleus stands in an instance-level part relation to some corresponding instance of the class cell.
This dependence of class-relations on relations among corresponding instances has long been recognized by logicians, including those working in the field of description logics, where the (all - some) form of definition we utilize below has been basic to the formalism from the start [14]. Definitions of this type were incorporated also into the DL-based GALEN medical ontology [15], though the significance of such definitions, and more generally of the role of instances in defining class relations, has still not been appreciated in many user communities.
It is also characteristically not realized that talk of classes involves in every case a more-or-less explicit reference to corresponding instances. When we assert that one class stands in an is_a relation to another (that is, that the first is a subtype of the second), for example, that glucose metabolism is_a carbohydrate metabolism, then we are stating that instances of the first class are ipso facto instances of the second. When we are dealing exclusively with is_a relations there is little reason to take explicit notice of this two-sided nature of ontological relations. When, however, we move to ontological relations of other types, then it becomes indispensable, if many characteristic families of errors are to be avoided, that the implicit reference to instances be taken carefully into account.
Types of relations
We focus here exclusively on genuinely ontological relations, which we take to mean relations that obtain between entities in reality, independently of our ways of gaining knowledge about such entities (and thus of our experimental methods) and independently of our ways of representing or processing such knowledge in computers. A relation like annotates is not ontological in this sense, as it links classes not to other classes in nature but rather to terms in a vocabulary that we ourselves have constructed. We focus also on general-purpose relations - relations which can be employed, in principle, in all biological ontologies - rather than on those specific relations (such as genome_of or sequence_of employed by OBO's Sequence Ontology) which apply only to biological entities of certain kinds. The latter will, however, need to be defined in due course in accordance with the methodology advanced here.
The ontologies in OBO are designed to serve as controlled vocabularies for expressing the results of biological science. Sentences of the form 'A relation B' (where 'A' and 'B' are terms in a biological ontology and 'relation' stands in for 'part_of' or some similar expression) can thus be conceived as expressing general statements about the corresponding biological classes or types. Assertions about corresponding instances or tokens (for example about the mass of this particular specimen in this particular Petri dish), while indispensable to biological research, do not belong to the general statements of biological science and thus they fall outside the scope of OBO and similar ontologies as these are presented to the user as finished products.
Yet such assertions are still relevant to ontologies. For it turns out that it is only by means of a detour through instances that the definitions and rules for coding relations between classes can be formulated in an intuitive and unambiguous - and thus reliably applicable - way.
We can distinguish, in fact, the following three kinds of binary relations:
<class, class>: for example, the is_a relation obtaining between the class SWR1 complex and the class chromatin remodeling complex, or between the class exocytosis and the class secretion;
<instance, class>: for example, the relation instance_of obtaining between this particular vesicle membrane and the class vesicle membrane, or between this particular instance of mitosis and the class mitosis;
<instance, instance>: for example, the relation of instance-level parthood (called part_of in what follows), obtaining between this particular vesicle membrane and the endomembrane system in the corresponding cell, or between this particular M phase of some mitotic cell cycle and the entire cell cycle of the particular cell involved.
Here classes and the relations between them are represented in italic; all other relations are picked out in bold.
Continuants and processes
The terms 'continuant' and 'process' are generalizations of GO's 'cellular component' and 'biological process' but applied to entities at all levels of granularity, from molecule to whole organism. Continuants are those entities which endure, or continue to exist, through time while undergoing different sorts of changes, including changes of place. Processes are entities that unfold themselves in successive temporal phases [16]. The terms 'continuant' and 'process' thus correspond to what, in the literature of philosophical ontology, are known respectively as 'things' (objects, endurants) and 'occurrents' (activities, events, perdurants) respectively. A continuant is what changes; a process is the change itself. The continuant classes relevant to biological ontologies include molecule, cell, membrane, organ; the process classes include ion transport, cell division, fat body development, breathing.
To formulate precise definitions of the <class, class> relations which form the target of ontology construction in biology we will need to employ a vocabulary that allows reference both to classes and to instances. For this we take advantage of the machinery of logic, and more specifically of the standard device of variables and quantifiers [17], using different sorts of variables to range across the classes and instances of continuants and processes, spatial regions and temporal instants, respectively. For the sake of intelligibility we use a semi-formal syntax, which can, however, be translated in a simple way into standard logical notation.
We use variables of the following sorts:
C, C1, ... to range over continuant classes;
P, P1, ... to range over process classes;
c, c1, ... to range over continuant instances;
p, p1, ... to range over process instances;
r, r1, ... to range over three-dimensional spatial regions;
t, t1, ... to range over instants of time.
In an expanded version of our formal machinery we will need also to incorporate further variables, ranging for example over temporal intervals, biological functions, attributes and values.
Note that continuants and processes form non-overlapping categories. This means in particular that no subtype or parthood relations cross the continuant-process divide. The tripartite structure of the GO recognizes this categorical exclusivity and extends it to functions also.
Continuants can be material (a mitochondrion, a cell, a membrane), or immaterial (a cavity, a conduit, an orifice), and this, too, is an exclusive divide. Immaterial continuants have much in common with spatial regions [18]. They are distinguished therefrom, however, in that they are parts of organisms, which means that, like material continuants, they move from one spatial region to another with the movements of their hosts.
The three-dimensional continuants that are our primary focus here typically have a top and a bottom, an anterior and a posterior, an interior and an exterior. Processes, in contrast, have a beginning, a middle and an end. Processes, but not continuants, can thus be partitioned along the time axis, so that, for example, your youth and your adulthood are temporal parts of that biological process which is your life.
As child and adult are continuants, so youth and adulthood are processes. We are thus clearly dealing here with two complementary - space-focused and time-focused - views of the same underlying subject matter, with determinate logical and ontological connections between them [16]. The framework advanced below allows us to capture these connections by incorporating reference to spatial regions and to temporal instants, both of which can be thought of as special kinds of instances.
We shall also need to distinguish two kinds of instance-level relations: those (applying to continuants) whose representations must involve a temporal index, and those (applying to processes) which do not. Note that the drawing of this distinction is still perfectly consistent with the fact that processes themselves occur in time, and that processes may be built out of successive subprocesses instantiating distinct classes.
Primitive instance-level relations
We cannot, on pain of infinite regress, define all relations, and this means that some relations must be accepted as primitive. The relations selected for this purpose should be self-explanatory and they should as far as possible be domain-neutral, which means that they should apply to entities in all regions of being and not just to those in the domain of biology.
Our choice of primitive relations is as follows:
c instance_of C at t - a primitive relation between a continuant instance and a class which it instantiates at a specific time
p instance_of P - a primitive relation between a process instance and a class which it instantiates holding independently of time
c part_of c1 at t - a primitive relation between two continuant instances and a time at which the one is part of the other
p part_of p1, r part_of r1 - a primitive relation of parthood, holding independently of time, either between process instances (one a subprocess of the other), or between spatial regions (one a subregion of the other)
c located_in r at t - a primitive relation between a continuant instance, a spatial region which it occupies, and a time
r adjacent_to r1 - a primitive relation of proximity between two disjoint continuants
t earlier t1 - a primitive relation between two times
c derives_from c1 - a primitive relation involving two distinct material continuants c and c1
p has_participant c at t - a primitive relation between a process, a continuant, and a time
p has_agent c at t - a primitive relation between a process, a continuant and a time at which the continuant is causally active in the process
This list includes only those <instance-instance> relations, together with one <instance-class> relation, which are needed for defining the <class, class> relations which are our principal target in this paper. The items on the list have been selected because they enjoy a high degree of intelligibility to the human authors and curators of biological ontologies. For purposes of supporting computer applications, however, the meanings of the corresponding relational expressions must be specified formally via axioms, for example in the case of 'part_of' by axioms of mereology (the theory of part and whole: see below), and in the case of 'earlier' by axioms governing a linear order [17]. The relation located_in will satisfy axioms to the effect that for every continuant there is some region in which it is located; instance_of will satisfy axioms to the effect that all classes have (at some stage in their existence) instances, and that all instances are instances of some class.
The formal machinery for reasoning with such axioms is in place, and a comprehensive set of axioms is being compiled. For the typical human user of biological ontologies, however, the listed primitive relations and associated axioms are designed to work invisibly behind the scenes. That is, they serve as part of the background framework that guides the construction and maintenance of such ontologies.
Results
Methodology
We employed a multi-stage methodology for the selection of the relations to be included in this ontology and for the formulation of corresponding definitions. First, a sample of researchers involved in ontology construction in the life sciences, representing different groups and including the co-authors of this paper, was asked to prepare lists of principal relations in light of their own specific experience but focusing on relations which would be: 'ontological' in the sense introduced above; 'general-purpose' in the sense that they apply across all biological domains; and also such as to manifest a high degree of universality (in the sense explained in the section 'Types of relational assertions' below). The submitted lists manifested a significant degree of overlap, which allowed us to prepare a core list in whose terms a large number of the remaining relations on the list could be simply defined.
A further constraint on the process was the goal of providing a simple formal definition for each included <class-class> relation. Those relations for which an appropriate simple definition could not be agreed upon were not included in this interim list. This includes most conspicuously relations involving analogs of the GO notion of molecular function. The relation has_agent was, however, included in light of a common understanding that the notion of agency would be involved in whatever candidate definition of function in biology is eventually accepted for use in OBO. This further constraint was chosen in light of the fact that our capacity to provide simple formal definitions - definitions which will at one and the same time be intelligible to ontology authors and curators and also able to support logic-based tools for automatic reasoning and consistency-checking - is the primary rationale for the methodology here advanced.
The two relations is_a and part_of were unproblematic candidates for inclusion in the resulting list (though providing simple definitions even for these relations was not, as we shall see, a simple matter). Is_a and part_of have established themselves as foundational to current ontologies. They have a central role in almost all domain ontologies, including the Foundational Model of Anatomy (FMA) [19,20], GO and other ontologies in OBO, as well as in influential top-level ontologies such as DOLCE [21] and in digitalized lexical resources such as WordNet [22].
In preparing our sample lists we drew on representatives not only of the OBO consortium but also of GALEN and the FMA (itself a candidate for inclusion in OBO). Our temporal relations draw on existing OBO practice (where transformation_of is a generalization of the develops_from relation used in OBO's cell and anatomy ontologies) and our participation relations draw on current work addressing the need to provide relations that link entities in different ontologies (for example entities in GO's process, function and component ontologies) and on an evolving Physiology Reference Ontology that is being developed in conjunction with the FMA [23], from which our spatial relations were extracted.
The OBO Relation Ontology
The first proposed version of the OBO Relation Ontology is shown in Table 1. We shall deal here with each of the ten relations listed in Table 1 in turn, providing rigorous yet easily understandable definitions.
Is_a
It is commonly assumed in the literature of knowledge representation that the relation is_a (meaning 'is a subtype of') can be identified with the subset or set inclusion relation with which we are familiar from mathematical set theory [17]. Instance_of functions on this reading as a counterpart of the usual set-theoretic membership relation, yielding a definition of A is_a B along the lines of: for all x, if x instance_of A, then x instance_of B. Unfortunately, this reading provides at best a necessary condition for the truth of A is_a B. It falls short of providing a sufficient condition for two reasons. The first is because it admits cases of contingent inclusion such as: bacterium in 90 mm × 18 mm glass Petri dish is_a bacterium, and the second is because it fails to take account of time, so that when applied to classes of continuants it yields false positives such as adult is_a child (because every instance of adult was at some time an instance of child).
We resolve the first problem by admitting as is_a links only assertions that reflect truths of biological science - assertions involving genuine biological class names (such as 'enzyme' or 'apoptosis') rather than, for example, commercial or indexical names (such as 'bacterium in this Petri dish'). The second problem we resolve by exploiting our machinery for taking account of time in the assertion of is_a relations involving continuants.
We can then define:
C is_a C1 = [definition] for all c, t, if c instance_of C at t then c instance_of C1 at t.
P is_a P1 = [definition] for all p, if p instance_of P then p instance_of P1.
Note how the device of logical quantifiers (for all ..., for some ...) allows us to refer to instances 'in general' - which means without the need to call on the proper names or indexical expressions (such as 'this' or 'here') which we use when referring to instances 'in specific'. Note also how instantiation for continuants involves a temporal argument. This reflects the fact that continuants, but not processes, can instantiate different classes in the course of their existence and yet preserve their identity.
For simplicity of expression we shall henceforth write 'Cct' and 'Pp', as abbreviations for: 'c instance_of C at t ' and 'p instance_of P ', respectively.
Part_of
Parthood as a relation between instances. The primitive instance-level relation p part_of p1 is illustrated in assertions such as: this instance of rhodopsin mediated phototransduction part_of this instance of visual perception.
This relation satisfies at least the following standard axioms of mereology: reflexivity (for all p, p part_of p); anti-symmetry (for all p, p1, if p part_of p1 and p1 part_of p then p and p1 are identical); and transitivity (for all p, p1, p2, if p part_of p1 and p1 part_of p2, then p part_of p2). Analogous axioms hold also for parthood as a relation between spatial regions.
For parthood as a relation between continuants, these axioms need to be modified to take account of the incorporation of a temporal argument. Thus for example the axiom of transitivity for continuants will assert that if c part_of c1 at t and c1 part_of c2 at t, then also c part_of c2 at t.
Parthood as a relation between classes. To define part_of as a relation between classes we again need to distinguish the two cases of continuants and processes, even though the explicit reference to instants of time now falls away. For continuants, we have C part_of C1 if and only if any instance of C at any time is an instance-level part of some instance of C1 at that time, as for example in: cell nucleus part_ of cell.
Formally:
C part_of C1 = [definition] for all c, t, if Cct then there is some c1 such that C1c1t and c part_of c1 at t.
Note the 'all-some' structure of this definition, a structure which will recur in almost all the relations treated here.
C part_of C1 defines a relational property of permanent parthood for Cs. It tells us that Cs, whenever they exist, exist as parts of C1s. We can also define in the obvious way C temporary_part_of C1 (every C exists at some time in its existence as part of some C1) and also C initial_part_of C1 (every C is such that it begins to exist as part of some instance of C1).
For processes, we have by analogy, P part_of P1 if and only if any instance of P is an instance-level part of some instance of P1, as for example in: M phase part_of cell cycle or neuroblast cell fate determination part_of neurogenesis. Formally:
P part_of P1 = [definition] for all p, if Pp then there is some p1 such that: P1p1 and p part_of p1.
An assertion to the effect that P part_of P1 thus tells us that Ps in general are in every case such as to exist as parts of P1s. P1s themselves, however, may exist without having Ps as parts (consider: menopause part_of aging).
Note that part_of is in fact two relations, one linking classes of continuants, the other linking classes of processes. While both of the mentioned relations are transitive, this does not mean that part_of relations could be inferred which would cross the continuant-process divide.
Located_in
Location as a relation between instances. The primitive instance-level relation c located_in r at t reflects the fact that each continuant is at any given time associated with exactly one spatial region, namely its exact location [24]. Following [25] we can use this relation to define a further instance-level location relation - not between a continuant and the region which it exactly occupies, but rather between one continuant and another. c is located in c1, in this sense, whenever the spatial region occupied by c is part_of the spatial region occupied by c1. Formally:
c located_in c1 at t = [definition] for some r, r1, c located_in r at t and c1 located_in r1 at t and r part_of r1.
Note that this relation comprehends both the relation of exact location between one continuant and another which obtains when r and r1 are identical (for example, when a portion of fluid exactly fills a cavity), as well as those sorts of inexact location relations which obtain, for example, between brain and head or between ovum and uterus.
Location as a relation between classes. To define location as a relation between classes - represented by sentences such as ribosome located_in cytoplasm, intracellular located_in cell - we now set:
C located_in C1 = [definition] for all c, t, if Cct then there is some c1 such that C1c1t and c located_in c1 at t.
Note that C located_in C1 is an assertion about Cs in general, which does not tell us anything about C1s in general (for example, that they have Cs located in them).
Contained_in
If c part_of c1 at t then we have also, by our definition and by the axioms of mereology applied to spatial regions, c located_in c1 at t. Thus, many examples of instance-level location relations for continuants are in fact cases of instance-level parthood. For material continuants location and parthood coincide. Containment is location not involving parthood, and arises only where some immaterial continuant is involved. To understand this relation, we first define overlap for continuants as follows:
C1 overlap c2 at t = [definition] for some c, c part_of c1 at t and c part_of c2 at t.
The containment relation on the instance level can then be defined as follows:
c contained_in c1 at t = [definition] c located_in c1 at t and not c overlap c1 at t.
On the class level this yields:
C contained_in C1 = [definition] for all c, t, if Cct then there is some c1 such that: C1c1t and c contained_in c1 at t.
Containment obtains in each case between material and immaterial continuants, for instance: lung contained_in thoracic cavity; bladder contained_in pelvic cavity. Hence containment is not a transitive relation.
Adjacent_to
We can define additional spatial relations by appealing to the primitive adjacent_to, a relation of proximity between disjoint continuants. Adjacent_to satisfies some of the axioms governing the relation referred to in the literature of qualitative topology as 'external connectedness' [26]. Analogs of other mereotopological relations (qualitative relations between spatial regions involving parthood, boundary and connectedness) (Figure 1) can also be defined, and these too can be applied to the material and immaterial continuants which occupy such regions on the instance level.
We define overlap for spatial regions as follows:
r1 overlap r2 = [definition] for some r, r part_of r1 and r part_of r2.
We then assert axiomatically that r1 adjacent_to r2 implies not r1 overlap r2
We can then define the counterpart relation of adjacency between classes as follows:
C adjacent_to C1 = [definition] for all c, t, if Cct, there is some c1 such that: C1c1t and c adjacent_to c1 at t.
Note that adjacent_to as thus defined is not a symmetric relation, in contrast to its instance-level counterpart. For it can be the case that Cs are in general such as to be adjacent to instances of C1 while no analogous statement holds for C1s in general in relation to instances of C. Examples are:
nuclear membrane adjacent_to cytoplasm
seminal vesicle adjacent_to urinary bladder
ovary adjacent_to parietal pelvic peritoneum.
We can, however, very simply define a symmetric relation of co-adjacency on the class level as follows:
C1 co-adjacent_to C2 = [definition] C1 adjacent_to C2 and C2 adjacent_to C1.
Examples are:
inner layer of plasma membrane co-adjacent_to outer layer of plasma membrane
right pulmonary artery co-adjacent_to right principal bronchus
urinary bladder of female co-adjacent_to parietal peritoneum of female pelvis.
Transformation_of
When an embryonic oenocyte (a type of insect cell) is transformed into a larval oenocyte, one and the same continuant entity preserves its identity while instantiating distinct classes at distinct times. The class-level relation transformation_of obtains between continuant classes C and C1 wherever each instance of the class C is such as to have existed at some earlier time as an instance of the distinct class C1 (see Figure 2). This relation is illustrated first of all at the molecular level of granularity by the relation between mature RNA and the pre-RNA from which it is processed, or between (UV-induced) thymine-dimer and thymine dinucleotide. At coarser levels of granularity it is illustrated by the transformations involved in the creation of red blood cells, for example, from reticulocyte to erythrocyte, and by processes of development, for example, from larva to pupa, or from (post-gastrular) embryo to fetus [27] or from child to adult. It is also manifest in pathological transformations, for example, of normal colon into carcinomatous colon. In each such case, one and the same continuant entity instantiates distinct classes at different times in virtue of phenotypic changes.
As definition for this relation we offer:
C transformation_of C1 = [definition] C and C1 for all c, t, if Cct, then there is some t1 such that C1ct1, and t1 earlier t, and there is no t2 such that Cct2 and C1ct2.
That is to say, the class C is a transformation of the class C1 if and only if every instance c of C is at some earlier time an instance of C1, and there is no time at which it is an instance of both C and C1. (The final clause, which asserts that C and C1 do not share instances at a time, is inserted in order to rule out, for example, adult human transformation_of human.)
Note that C transformation_of C1 is a statement about Cs in general. It does not tell us of C1s in general that each gives rise to some C which stands to it in a transformation_of relation.
Derives_from
Derivation as a relation between instances. The temporal relation of derivation is more complex. Transformation, on the instance level, is just the relation of identity: each adult is identical to some child existing at some earlier time. Derivation on the instance-level is a relation holding between non-identicals. More precisely, it holds between distinct material continuants when one succeeds the other across a temporal divide in such a way that at least a biologically significant portion of the matter of the earlier continuant is inherited by the later. Thus we will have axioms to the effect that from c derives_from c1 we can infer that c and c1 are not identical and that there is some instant of time t such that c1 exists only prior to and c only subsequent to t. We will also be able to infer that the spatial region occupied by c as it begins to exist at t overlaps with the spatial region occupied by c1 as it ceases to exist in the same instant.
Three simple kinds of instance-level derivation can then be distinguished (Figure 3): first, the succession of one single continuant by another single continuant across a temporal threshold (for example, this blastocyst derives from this zygote); second, the fusion of two or more continuants into one continuant (for example, this zygote derives from this sperm and from this ovum); and third, the fission of an earlier single continuant to create a plurality of later continuants (for example, these promyelocytes derive from this myeoloblast). In all cases we have two continuants c and c1 which are such that c begins to exist at the same instant of time at which c1 ceases to exist, and at least a significant portion of the matter of c1 is inherited by its successor c.
Derivation of the first type is still essentially weaker than transformation, for the latter involves the identity of the continuant instances existing on either side of the relevant temporal divide. In derivation of the second type, the successor continuant takes the bulk of its matter from a plurality of precursors, where in cases of the third type, the bulk of the matter of a single precursor continuant is shared among a plurality of successors. We can also represent more complex cases where transformation and an analog of derivation are combined, for example in the case of budding in yeast [27], where one continuant continues to exist identically through a process wherein a second continuant floats free from its host; or in absorption, where one continuant continues to exist identically through a process wherein it absorbs another continuant, for example through digestion.
Derivation as a relation between classes. To avoid troubling counter-examples, the relation of derivation we are seeking on the class level must be defined in two steps. First, the class-level counterpart of the relation of derivation on the instance level is identified as a relation of immediate derivation:
C derives_immediately_from C1 = [definition] for all c, t, if Cct, then there is some c1,t1, such that: t1 earlier t and C1c1t1 and c derives_from c1.
The more general class level derivation relation must then be defined in terms of chains of immediate derivation relations, as follows:
C derives_from C1 = [definition] there is some sequence C = Ck, Ck-1, ..., C2, C1, such that for each Ci (1 ≤ i < k), Ci+1 derives_immediately_from Ci.
In this way we can represent cases of derivation involved in the formation of lineages where there occurs a sequence of cell divisions or speciation events.
Preceded_by
With the primitive relations has_participant and earlier at our disposal we can define the instance-level relation p occurring_at t as follows:
p occurring_at t = [definition] for some c, p has_participant c at t.
We can then define:
c exists_at t = [definition] for some p, p has_participant c at t
p preceded_by p1 = [definition] for all t, t1, if p occurring_at t and p1 occurring_at t1, then t1 earlier t
t first_instant p = [definition] p occurring_at t and for all t1, if t1 earlier t, then not p occurring_at t1
t last_instant p = [definition] p occurring_at t and for all t1, if t earlier t1, then not p occurring_at t1
p immediately_preceded_by p1 = [definition] for some t, t first_instant p and t last_instant p1.
At the class level we have:
P preceded_by P1 = [definition] for all p, if Pp then there is some p1 such that P1p1and p preceded_by p1.
An example is: translation preceded_by transcription; aging preceded_by development (not however death preceded_by aging). Where derives_from links classes of continuants, preceded_by links classes of processes. Clearly, however, these two relations are not independent of each other. Thus if cells of type C1 derive_from cells of type C, then any cell division involving an instance of C1 in a given lineage is preceded_by cellular processes involving an instance of C.
The assertion P preceded_by P1 tells us something about Ps in general: that is, it tells us something about what happened earlier, given what we know about what happened later. Thus it does not provide information pointing in the opposite direction, concerning instances of P1 in general; that is, that each is such as to be succeeded by some instance of P. Note that an assertion to the effect that P preceded_by P1 is rather weak; it tells us little about the relations between the underlying instances in virtue of which the preceded_by relation obtains. Typically we will be interested in stronger relations, for example in the relation immediately_preceded_by, or in relations which combine preceded_by with a condition to the effect that the corresponding instances of P and P1 share participants, or that their participants are connected by relations of derivation, or (as a first step along the road to a treatment of causality) that the one process in some way affects (for example, initiates or regulates) the other.
Has_participant
Has_participant is a primitive instance-level relation between a process, a continuant, and a time at which the continuant participates in some way in the process. The relation obtains, for example, when this particular process of oxygen exchange across this particular alveolar membrane has_participant this particular sample of hemoglobin at this particular time.
To define the class-level counterpart of the participation relation we set:
P has_participant C = [definition] for all p, if Pp then there is some c, t such that Cct and p has_participant c at t.
Examples are:
cell transport has_participant cell
death has_participant organism
breathing has_participant thorax.
Once again, P has_participant C provides information only about Ps in general (that is, that they require instances of C as bearers).
Has_agent
Special types of participation can be distinguished according to whether a continuant is agent or patient in a process (for a survey see [28].) Here we focus on the factor of agency, which is involved, for example, when an adult engages in adult walking behavior. It is not involved when the same adult is the victim of an infection. Synonyms of 'is agent in' include: 'actively participates in', 'does', 'executes', 'performs', and so forth.
We introduce the primitive instance-level relation has_agent, which obtains between a process, a continuant and a time whenever the continuant is a participant in the process and is at the same time directly causally responsible for its occurrence. Thus we have an axiom to the effect that agency implies participation: for all p, c, t, if p has_agent c at t, then p has_participant c at t. In addition we will have axioms to the effect that only material continuants can fill the agent role, that if c fills the agent role at t, then c must have existed at times earlier than t, that it must exercise its agent role for an interval of time including t, and so on.
We can then define the class-level relation has_agent by stipulating:
P has_agent C = [definition] for all p, if Pp then there is some c, t such that Cct and p has_agent c at t
This relation gives us the means to capture the directionality (the from-to) nature of biological processes such as signaling, transcription, and expression, via assertions, for example, to the effect that in an interaction between molecules of types m1 and m2 it is molecules of the first type that play the role of agent.
One privileged type of agency consists in the realization of a biological function. To say that a continuant has a function is to assert, in first approximation, that it is predisposed (has the potential, the casual power) to cause (to realize as agent) a process of a certain type. Thus to say that your heart has the function: to pump blood is to assert that your heart is predisposed to realize as agent a process of the type pumping blood [29]. Regulation, promotion, inhibition, suppression, activation, and so forth, are among the varieties of agency that fall under this heading.
On the other hand, many processes - such as metabolic reactions involving enzymes, cofactors, and metabolites - involve no clear factor of agent participation, but rather require more nuanced classifications of the roles of participants - as acceptors or donors, for example. Hence the has_agent relation should be used in curation with special care. It should be borne in mind in this connection that agency is in every case a matter of the imposition of direct causal influence of a continuant in a process (a constraint that is designed to rule out inheritance of agency along causal chains), and also that (by our definition) only continuants can be agents. Where biologists describe processes as agents, for example, in talking about the effects of diffusion in development and differentiation, such phenomena are of a type that call for an expansion of our proposed Relation Ontology in the direction, again, of a treatment of the factor of causality.
Discussion
The logic of biological relations
Inverse and reciprocal relations
The inverse of a relation R is defined as that relation which obtains between each pair of relata of R when taken in reverse order. Inverses can be unproblematically defined for all instance-level relations. What, then, of inverses for class-level relations? The inverse relation for is_a can be defined trivially as follows:
A has_subclass B = [definition] B is_a A.
For the remaining class-level relations on our list, in contrast, the issue of corresponding inverses is more problematic [7]. Thus, while we have the true relational assertion human testis part_of human - which means that all instances of human testis are part of instances of some human - there is no corresponding true relational assertion linking instances of human to instances of human testis as their parts. For these remaining relations we need to work not with inverses but rather with what, following GALEN, we can call reciprocal relations. These are defined using the same family of instance-level primitives we introduced earlier. As reciprocal relations for the two varieties of part_of we have:
C has_part C1 = [definition] for all c, t, if Cct then there is some c1 such that C1c1t and c1 part_of c at t
P has_part P1 = [definition] for all p, if Pp then there is some p1 such that P1p1 and p1 part_of p
Note that from A part_of B we cannot infer that B has_ part A; similarly, from A has_ part B we cannot infer that B part_of A. Thus cell nucleus part_of cell, but not cell has_part cell nucleus; running has_ part breathing, but not breathing part_of running. A third significant relation conjoining part_of and has_part can be defined as [6,30]:
C integral_part_of C1 = [definition] C part_of C1 and C1 has_part C.
For contained_in we have similarly the reciprocal relation:
C contains C1 = [definition] for all C, t, if Cct then there is some c1 such that: C1c1t and c located_in c at t
For participation we can usefully define two alternative reciprocal relations:
C sometimes_ participates_in P = [definition] for all c there is some t and some p such that Cct and Pp and p has_participant c at t
C always_participates_in P = [definition] for all c, t, if Cct then there is some p such that Pp and p has_participant c at t
We can also define, for example, what it is for continuants of a given type to participate at every stage in a process of a given type. Thus if a sperm participates in the penetration of an ovum, then it does so throughout the penetration.
Types of relational assertions
In light of the above, we can now observe certain differences in what we might call the relative universality of class-level relational assertions. There are many cases, above all involving is_a relations, where relational assertions hold with a maximal degree of universality, which means that they hold for every instance of the classes in question because they are a matter of analytic connections, that is, connections resting on the compositional nature of the class terms involved [10], as, for example, in: eukaryotic cell is_a cell, or adult walking behavior has_participant adult. (Contrast, adult participates_in adult walking behavior.)
There are also other kinds of statements enjoying a high degree of universality, for example: penetration of ovum has_participant sperm. The first of our two corresponding reciprocal statements - sperm participates_in penetration of ovum - is in contrast true only in relation to certain isolated instances of sperm, and the second of our reciprocal statements - sperm always_participates_in penetration of ovum - is true in relation to no instances at all.
It then seems reasonable to insist that biomedical ontologies should reflect those sorts of biological assertions that enjoy a high degree of universality (typically assertions involving just one of each pair of reciprocal relations).
Tools for ontology curation
We hope that, by providing clear and unambiguous specifications of what the class-level relational expressions used in biological ontologies mean, our formal definitions will assist curators engaged in ontology creation and maintenance. The corresponding definitions are summarized in Table 2, which also contains representative examples for each of the relations distinguished.
Our definitions are designed to ensure that the corresponding general-purpose relational expressions are used in a uniform way in all biological ontologies. In this way we shall be in a position to contribute to the realization of the goal of bringing about a high degree of interoperability even where ontologies are produced by different groups and for different purposes. These definitions are designed also to enable the automatic detection of errors in biomedical ontologies, for example by allowing the construction of extensions of OBO-Edit and similar tools with the facility to test whether given relations are employed in an ontology in such a way as to involve relata of the appropriate types [31] or in such a way as to have the formal characteristics, such as transitivity or reflexivity, dictated by the definitions (Table 3). The framework can also support reasoning applications designed to enable the automated derivation of information from existing bodies of knowledge - for example to infer the parts of a given cell continuant via the traversal of a part_of hierarchy - including instance-based knowledge derived from the clinical record.
Conclusion
The Relation Ontology outlined above arose through collaboration between formal ontologists and biologists in the OBO, FMA and GALEN research groups and also incorporates suggestions from a number of other authors and curators of biomedical ontologies. It is designed to be large enough to overcome some of the problems arising in GO and similar systems as a result of the paucity of resources available hitherto for expressing relations between the classes in such ontologies [32]. It is this paucity of resources, above all, which gives rise to cases of multiple inheritance in GO as presently constructed, and we note here that multiple inheritance often goes hand in hand with errors in ontology construction not least because it encourages a relaxed reading of is_a (often a reading which involves the assertion of is_a relations which erroneously cross the divide between different ontological categories) [5,33]. Our present framework can contribute to error resolution not only by dictating a common interpretation of is_a which can serve as orientation for ontology authors and curators in their future work, but also by providing richer resources for the assertion of class-class relations within and between ontologies in such a way that the appeal to contrived and error-prone is_a relations can be more easily avoided.
At the same time our suite of relations has been designed to be sufficiently small to attract wide acceptance in a range of different types of life-science communities. Where the latter use further, general-purpose or domain-specific relations of their own, we plan in due course to subject such relations to the same kind of analysis as presented here in order to preserve interoperability. The Relation Ontology has been incorporated into the OBO ontology library [34] and curators of the GO and FMA ontologies and also of the ChEBI chemical entities vocabulary [35] are already applying the relevant parts of the ontology in their work. The ontology has already been used to find errors not only in GO but also in SNOMED [36]. It is also being applied systematically in evaluations of the NCI Thesaurus [37] and the UMLS (Unified Medical Language System) Semantic Network of the National Library of Medicine. We are currently testing methodologies to obtain reliable quantitative evaluations of the utility of the proposed framework for purposes of ontology authoring and also for use in annotation and reasoning. We are also testing ways in which the framework can be expanded through the admission of pre-coordinated disjunctions (for example: either derivation or transformation), which can allow the coding of information in those cases where the precise nature of the relations involved is insufficiently clear to allow unique assignment.
The Relation Ontology will be evaluated on two levels. First, on whether it succeeds in preventing those characteristic kinds of errors which have been associated with a poor treatment of relations in biomedical ontologies in the past. Second, and more important, on whether it helps to achieve greater interoperability of biomedical ontologies and thus to improve reasoning about biological phenomena.
Acknowledgements
Work on this paper was carried out under the auspices of the Wolfgang Paul Program of the Alexander von Humboldt Foundation, the EU Network of Excellence in Medical Informatics and Semantic Data Mining, the Project 'Forms of Life' sponsored by the Volkswagen Foundation, and the DARPA Virtual Soldier Project. Thanks go to Michael Ashburner, Fabrice Correia, Maureen Donnelly, Kai Hauser, Win Hyde, Ingvar Johansson, Janet Kelso, Suzanna Lewis, Katherine Munn, Maria Reicher, Alan Ruttenberg, Mark Scala, Stefan Schulz, Neil Williams, Lina Yip, Sumi Yoshikawa, and anonymous referees for valuable comments.
Figures and Tables
Figure 1 Standard mereotopological relations between spatial regions.
Figure 2 Transformation.
Figure 3 Three simple cases of derivation. (a) Continuation; (b) fusion; (c) fission.
Table 1 First version of the OBO Relation Ontology
Foundational relations
is_a
part_of
Spatial relations (connecting one entity to another in terms of relations between the spatial regions they occupy)
located_in
contained_in
adjacent_to
Temporal relations (connecting entities existing at different times)
transformation_of
derives_from
preceded_by
Participation relations (connecting processes to their bearers)
has_participant
has_agent
Table 2 Definitions and examples of class-level relations
Relations and relata Definitions Examples
C is_a C1; Cs and C1s are continuants Every C at any time is at the same time a C1 myelin is_a lipoprotein
serotonin is_a biogenic amine
mitochondrion is_a membranous cytoplasmic organelle
protein kinase is_a kinase
DNA is_a nucleic acid
P is_a P1; Ps and P1s are processes Every P is a P1 endomitosos is_a DNA replication
catabolic process is_a metabolic process
photosynthesis is_a physiological process
gonad development is_a organogenesis
intracellular signaling cascade is_a signal transduction
C part_of C1; Cs and C1s are continuants Every C at any time is part of some C1 at the same time mitochondrial matrix part_of mitochondrion
microtubule part_of cytoskeleton
nuclear pore complex part_of nuclear membrane
nucleoplasm part_of nucleus
promotor part_of gene
P part_of P1; Ps and P1s are processes Every P is part of some P1 gastrulation part_of embryonic development
cystoblast cell division part_of germ cell development
cytokinesis part_of cell proliferation
transcription part_of gene expression
neurotransmitter release part_of synaptic transmission
C located_in C1; Cs and C1s are continuants Every C at any given time occupies a spatial region which is part of the region occupied by some C1 at the same time 66s pre-ribosome located_in nucleolus
intron located_in gene
nucleolus located_in nucleus
membrane receptor located_in cell membrane
chlorophyll located_in thylakoid
C contained_in C1; Cs are material continuants, C1s are immaterial continuants (holes, cavities) Every C at any given time is located in but shares no parts in common with some C1 at the same time thoracic aorta contained_in posterior mediastinal cavity
cytosol contained_in cell compartment space
thylakoid contained_in chloroplast membrane
synaptic vesicle contained_in neuron
C adjacent_to C1; Cs and C1s are continuants Every C at any time is proximate to some C1 at the same time Golgi apparatus adjacent_to endoplasmic reticulum
intron adjacent_to exon
cell wall adjacent_to cytoplasm
periplasm adjacent_to plasma membrane
presynaptic membrane adjacent_to synaptic cleft
C transformation_of C1; Cs and C1s are material continuants Every C at any time is identical with some C1 at some earlier time facultative heterochromatin transformation_of euchromatin
mature mRNA transformation_of pre-mRNA
hemosiderin transformation_of hemoglobin
red blood cell transformation_of reticulocyte
fetus transformation_of embryo
C derives_from C1; Cs and C1s are material continuants Every C is such that in the first moment of its existence it occupies a spatial region which overlaps the spatial region occupied by some C1 in the last moment of its existence plasma cell derives_from B lymphocyte
fatty acid derives_from triglyceride
triple oxygen molecule derives_from oxygen molecule
Barr body derives_from X-chromosome
mammal derives_from gamete
P preceded_by P1; Ps and P1s are processes Every P is such that there is some earlier P1 translation preceded_by transcription
meiosis preceded_by chromosome duplication
cytokinesis preceded_by DNA replication
apoptotic cell death preceded_by nuclear chromatin degradation
digestion preceded_by ingestion
P has_participant C; Ps are processes, Cs are continuants Every P involves some C as participant mitochondrial acetylCoA formation has_participant pyruvate dehydrogenase complex
translation has_participant amino acid
photosynthesis has_participant chlorophyll
apoptosis has_participant cell
cell division has_participant chromosome
P has_agent C; Ps are processes, Cs are material continuants Every P involves some C as agent (the C is involved in and is causally responsible for the P) gene expression has_agent RNA polymerase
signal transduction has_agent receptor
pathogenesis has_agent pathogen
transcription has_agent RNA polymerase
translation has_agent ribosome
Table 3 Some properties of the relations in the OBO Relation Ontology
Relation Transitive Symmetric Reflexive Antisymmetric
is_a + - + +
part_of + - + +
located_in + - + -
contained_in - - - -
adjacent_to - - - -
transformation_of + - - -
derives_ from + - - -
preceded_by + - - -
has_participant - - - -
has_agent - - - -
==== Refs
OBO: Open Biomedical Ontologies
Mungall C OBOL: integrating language and meaning in bio-ontologies. Comp Funct Genomics 2004 5 509 520 10.1002/cfg.435
Gene Ontology Consortium Creating the Gene Ontology resource: design and implementation. Genome Res 2001 11 1425 1433 11483584 10.1101/gr.180801
Bada M Stevens R Goble C Gil Y Ashburner M Blake JA Cherry JM Harris M Lewis S A short study on the success of the GeneOntology. J Web Semantics 2004 1 235 240
Smith B Köhler J Kumar A On the application of formal principles to life science data: a case study in the Gene Ontology. DILS 2004: Data Integration in the Life Sciences Lecture Notes in Computer Science 2994 2004 124 139
Smith B Rosse C The role of foundational relations in the alignment of biomedical ontologies. Proceedings Medinf 2004 2004 Amsterdam: IOS Press 444 448
Smith B Kumar A On controlled vocabularies in bioinformatics: a case study in the Gene Ontology. BioSilico: Inform Technol Drug Discovery 2004 2 246 252 10.1016/S1741-8364(04)02424-2
Smith B Williams J Schulze-Kremer S The ontology of the Gene Ontology. Proc AMIA Symp 2003 609 13 14728245
Ogren PV Cohen KB Acquaah-Mensah GK Eberlein J Hunter L The compositional structure of Gene Ontology terms. Pac Symp Biocomput 2004 214 225 14992505
Ogren P Bretonnel Cohen K Hunter L Implications of compositionality in the Gene Ontology for its curation and usage. Pac Symp Biocomput 2005 174 185 15759624
Bard J Rhee SY Ashburner M An ontology for cell types. Genome Biol 2005 6 R21 15693950 10.1186/gb-2005-6-2-r21
Wroe C Stevens R Goble CA Ashburner M An evolutionary methodology to migrate the Gene Ontology to a Description Logic environment using DAML+OIL. Pac Symp Biocomput 2003 624 635 12603063
Smith B Beyond concepts: ontology as reality representation. Formal Ontology and Information Systems 2004 2004 Amsterdam: IOS Press 73 84
Levesque HJ Brachman RJ A fundamental tradeoff in knowledge representation and reasoning. Readings in Knowledge Representation 1985 San Francisco: Morgan Kaufman 41 70
Rogers J Rector AL The GALEN ontology. Medical Informatics Europe 1996 1996 Amsterdam: IOS Press 174 178
Grenon P Smith B Goldberg L Biodynamic ontology: applying BFO in the biomedical domain. Ontologies in Medicine 2004 Amsterdam: IOS Press 20 38
Stoll R Set Theory and Logic 1979 New York: Dover Publications
Casati R Varzi AC Holes and Other Superficialities 1994 Cambridge, MA: MIT Press
Rosse C Mejino JLV Jr A reference ontology for bioinformatics: the Foundational Model of Anatomy. J Biomed Inform 2003 36 478 500 14759820 10.1016/j.jbi.2003.11.007
Rogers J Rector AL GALEN's model of parts and wholes: experience and comparisons. Proceedings AMIA Symposium 2000 2000 Bethesda, MD: American Medical Informatics Association 819 823
Gangemi A Guarino N Masolo C Oltramari A Sweetening WordNet with DOLCE. AI Magazine 2003 24 13 24
Fellbaum C Ed Wordnet An Electronic Lexical Database 1998 Cambridge, MA: MIT Press
Cook DL Mejino JLV JrRosse C Evolution of a Foundational Model of Physiology: symbolic representation for functional bioinformatics. Proceedings MedInfo 2004 2004 Amsterdam: IOS Press 336 340
Bittner T Axioms for parthood and containment relations in bio-ontologies. KR-MED 2004: Workshop on Formal Biomedical Knowledge Representation 2004 Aachen: University of Aachen 4 11
Donnelly M Layered mereotopology. Proceedings 18th Joint International Conference on Artificial Intelligence 2003 San Francisco: Morgan Kaufman 1269 1274
Smith B Mereotopology: a theory of parts and boundaries. Data Knowledge Eng 1996 20 287 303 10.1016/S0169-023X(96)00015-8
Smith B Brogaard B Sixteen days. J Med Philos 2003 28 45 78 12715281 10.1076/jmep.28.1.45.14172
Smith B Grenon P The cornucopia of formal-ontological relations. Dialectica 2004 58 279 296
Johansson I Smith B Munn K Tsikolia N Elsner K Ernst D Siebert D Functional anatomy: a taxonomic proposal. Acta Biotheoret 2005
Schulz S Hahn U Towards a computational paradigm for biomedical structure. KR-MED 2004: Workshop on Formal Biomedical Knowledge Representation 2004 Aachen: University of Aachen 63 71
dos Santos MC Dhaen C Fielding M Ceusters W Philosophical scrutiny for run-time support of application ontology development. Formal Ontology and Information Systems 2004 Amsterdam: IOS Press 342 352
Kumar A Smith B Borgelt C Dependence relationships between Gene Ontology terms based on TIGR gene product annotations. Proceedings CompuTerm 2004 2004 Geneva: COLING 31 38
Bouaud J Bachimont B Charlet J Zweigenbaum P Acquisition and structuring of an ontology within conceptual graphs. Proceedings 2nd International Conference on Conceptual Structures: Workshop on Knowledge Acquisition using Conceptual Graph Theory Lecture Notes Computer Sci 1994 835 1 25
OBO Relationship Ontology
ChEBI: Chemical Entities of Biological Interest
Ceusters W Smith B Kumar A Dhaen C Ontology-based error detection in SNOMED-CT. Proceedings Medinfo 2004 2004 Amsterdam: IOS Press 482 486
Ceusters W Smith B Goldberg L A terminological and ontological analysis of the NCI Thesaurus. Meth Inform Medicine 2005
| 15892874 | PMC1175958 | CC BY | 2021-01-04 16:05:38 | no | Genome Biol. 2005 Apr 28; 6(5):R46 | utf-8 | Genome Biol | 2,005 | 10.1186/gb-2005-6-5-r46 | oa_comm |
==== Front
Genome BiolGenome Biology1465-69061465-6914BioMed Central London gb-2005-6-5-r461589287410.1186/gb-2005-6-5-r46MethodRelations in biomedical ontologies Smith Barry [email protected] Werner [email protected] Bert [email protected]öhler Jacob [email protected] Anand [email protected] Jane [email protected] Chris [email protected] Fabian [email protected] Alan L [email protected] Cornelius [email protected] Institute for Formal Ontology and Medical Information Science, Saarland University, D-66041 Saarbrücken, Germany2 Department of Philosophy, University at Buffalo, Buffalo, NY 14260, USA3 European Centre for Ontological Research, Saarland University, D-66041 Saarbrücken, Germany4 Department of Genetics, University of Leipzig, D-04103 Leipzig, Germany5 Rothamsted Research, Harpenden, AL5 2JQ, UK6 European Bioinformatics Institute, Hinxton, CB10 1SD, UK7 HHMI, Department of Molecular and Cellular Biology, University of California, Berkeley, CA 94729, USA8 Department of Computer Science, University of Manchester, M13 9PL, UK9 Department of Biological Structure, University of Washington, Seattle, WA 98195, USA2005 28 4 2005 6 5 R46 R46 28 10 2004 3 2 2005 31 3 2005 Copyright © 2005 Smith 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.
To enhance the treatment of relations in biomedical ontologies we advance a methodology for providing consistent and unambiguous formal definitions of the relational expressions used in such ontologies in a way designed to assist developers and users in avoiding errors in coding and annotation.
To enhance the treatment of relations in biomedical ontologies we advance a methodology for providing consistent and unambiguous formal definitions of the relational expressions used in such ontologies in a way designed to assist developers and users in avoiding errors in coding and annotation. The resulting Relation Ontology can promote interoperability of ontologies and support new types of automated reasoning about the spatial and temporal dimensions of biological and medical phenomena.
==== Body
Background
Controlled vocabularies in bioinformatics
The background to this paper is the now widespread recognition that many existing biological and medical ontologies (or 'controlled vocabularies') can be improved by adopting tools and methods that bring a greater degree of logical and ontological rigor. We describe one endeavor along these lines, which is part of the current reform efforts of the Open Biomedical Ontologies (OBO) consortium [1,2] and which has implications for ontology construction in the life sciences generally.
The OBO ontology library [1] is a repository of controlled vocabularies developed for shared use across different biological and medical domains. Thus the Gene Ontology (GO) [3,4] consists of three controlled vocabularies (for cellular components, molecular functions, and biological processes) designed to be used in annotations of genes or gene products. Some ontologies in the library - for example the Cell and Sequence Ontologies, as well as the GO itself - contain terms which can be used in annotations applying to all organisms. Others, especially OBO's range of anatomy ontologies, contain terms applying to specific taxonomic groups such as fly, fungus, yeast, or zebrafish.
Controlled vocabularies can be conceived as graph-theoretical structures consisting on the one hand of terms (which form the nodes of each corresponding graph) linked together by means of edges called relations. The ontologies in the OBO library are organized in this way by means of different types of relations. OBO's Mouse Anatomy ontology, for example, uses just one type of edge, labeled part_of. The GO currently uses two, labeled is_a and part_of. The Drosophila Anatomy ontology includes also a develops_from link. Other OBO ontologies include further links, for example (in the Sequence Ontology) position_of and disjoint_from. The National Cancer Institute (NCI) Thesaurus adds many additional links, including has_location for anatomical structures and different part_of relations for structures and for processes.
The problem is that when OBO and similar ontologies incorporate such relations they typically do so in informal ways, often providing no definitions at all, so that the logical interconnections between the various relations employed are unclear, and even the relations is_a and part_of are not always used in consistent fashion both within and between ontologies. Our task in what follows is to rectify these defects, drawing on the requirements analysis presented in [5].
Of the criteria that ontologies must currently satisfy if they are to be included in the OBO library, the most important for our purposes are: first, inclusion of textual definitions or descriptions designed to ensure that the precise meanings of terms as used within particular ontologies will be clear to a human reader; second, employment of a standard syntax, such as the OWL or OBO flatfile syntax; third, orthogonality to the other ontologies already included in the library. These criteria are designed to support the integration of OBO ontologies, above all by ensuring the compatibility of ontologies pertaining to an identical subject matter. OBO has now added a fourth criterion to assist in achieving such compatibility, namely that the relations (edges) used to connect terms in OBO ontologies should be applied in ways consistent with their definitions as set forth in this paper.
The Relation Ontology offered here is designed to put flesh on this criterion. How, exactly, should part_of or located_in be defined in order to ensure maximally reliable curation of each single ontology while at the same time guaranteeing maximal leverage in building a solid base for life-science knowledge integration in general? We describe a rigorous methodology for providing an answer to this question and illustrate its use in the construction of an easily extendible list of ten relations of a type familiar to those working in the bio-ontological field. This list forms the core of the new OBO Relation Ontology. What is distinctive about our methodology is that, while the relations are each provided with rigorous formal definitions, these definitions can at the same time be formulated in such a way that the underlying technical details remain invisible to ontology authors and curators.
Shortcomings of biomedical ontologies
While considerable effort has been invested in the formulation and definition of terms in biomedical ontologies, too little attention has been paid in the ontological literature to the associated relations. A number of characteristic types of shortcomings of controlled vocabularies can be traced back especially to the neglect of issues of formal structure in the treatment of relations [5-10]. To take just one example, the pre-2004 versions of GO allowed at least three different readings of the expression 'part of' as representing simultaneously: inclusion relations between vocabularies; a relation of possible parthood between biological entities; a relation of necessary parthood between biological entities. As was shown in [6], this coexistence of conflicting readings meant that three of the four rules given in the then effective documentation for reasoning with GO's hierarchies were logically incorrect.
Another characteristic family of problems turns on the paucity of resources for expressing relations in ontologies like GO. For example, because GO has no direct means of asserting location relations, it must capture such relations indirectly by constructing new terms involving syntactic operators such as 'site of', 'within', 'extrinsic to', 'space', 'region', and so on. It then simulates assertions of location by means of 'is_a' and 'part_of' statements involving such composites, for example in:
extracellular region is_a cellular component
extrinsic to membrane part_of membrane
both of which are erroneous. Additional problems arise from the fact that GO's extracellular region and extracellular space are both specified in their definitions as referring to the space (how large a space?) external to the outermost structure of a cell.
Another type of problem turns on the failure to distinguish relational expressions which, though closely related in meaning, are revealed to be crucially distinct when explicated in the formally precise way that is demanded by computer implementations. An example is provided by the simultaneous use in OBO's Cell Ontology of both derives_from and develops_from while no clear distinction is drawn between the two [11]. This problem is resolved in the treatment of derivation and transformation below, and has been correspondingly corrected in versions 1.14 and later of the Cell Ontology.
Efforts to improve GO from the standpoint of increased formal rigor have thus far been concentrated on re-expressing the existing GO schema in a description logic (DL) framework. This has allowed the use of a DL-reasoner that can identify certain kinds of errors and omissions, which have been corrected in later versions of GO [12]. DLs, however, can do no more than guarantee consistent reasoning according to the definitions provided to them. If the latter are themselves problematic, then a DL can do very little to identify or resolve the problems which result. Here, accordingly, we take a more radical approach, which consists in re-examining the basic definitions of the relations used in GO and in related ontologies in an attempt to arrive at a methodology which will lead to the construction of ontologies which are more fundamentally sound and thus more secure against errors and more amenable to the use of powerful reasoning tools. This approach is designed also to be maximally helpful to biologists by avoiding the problems which arise by virtue of the fact that the syntax favored in the DL-community is of a type which can normally be understood only by DL-specialists.
A theory of classes and instances
The relations in biological ontologies connect classes as their relata. The term 'class' here is used to refer to what is general in reality, or in other words to what, in the knowledge-representation literature, is typically (and often somewhat confusingly [13]) referred to under the heading 'concept' and in the literature of philosophical ontology under the headings 'universal', 'type' or 'kind'. Biological classes are in first approximation those classes which have been implicitly sanctioned through usage of the corresponding general terms in the biological literature, for example cell or fat body development.
Our task is to develop a suite of coherently defined bio-ontological relations that is sufficiently compact to be easily learned and applied, yet sufficiently broad in scope to capture a wide range of the relations currently coded in standard biomedical ontologies. Unfortunately the realization of this task is not a trivial matter. This is because, while the terms in biomedical ontologies refer exclusively to classes - to what is general in reality - we cannot define what it means for one class to stand to another, for example in the part_of relation, without taking the corresponding instances into account [6]. Here the term 'instance' refers to what is particular in reality, to what are otherwise called 'tokens' or 'individuals' - entities (including processes) which exist in space and time and stand to each other in a variety of instance-level relations. Thus we cannot make sense of what it means to say cell nucleus part_of cell unless we realize that this is a statement to the effect that each instance of the class cell nucleus stands in an instance-level part relation to some corresponding instance of the class cell.
This dependence of class-relations on relations among corresponding instances has long been recognized by logicians, including those working in the field of description logics, where the (all - some) form of definition we utilize below has been basic to the formalism from the start [14]. Definitions of this type were incorporated also into the DL-based GALEN medical ontology [15], though the significance of such definitions, and more generally of the role of instances in defining class relations, has still not been appreciated in many user communities.
It is also characteristically not realized that talk of classes involves in every case a more-or-less explicit reference to corresponding instances. When we assert that one class stands in an is_a relation to another (that is, that the first is a subtype of the second), for example, that glucose metabolism is_a carbohydrate metabolism, then we are stating that instances of the first class are ipso facto instances of the second. When we are dealing exclusively with is_a relations there is little reason to take explicit notice of this two-sided nature of ontological relations. When, however, we move to ontological relations of other types, then it becomes indispensable, if many characteristic families of errors are to be avoided, that the implicit reference to instances be taken carefully into account.
Types of relations
We focus here exclusively on genuinely ontological relations, which we take to mean relations that obtain between entities in reality, independently of our ways of gaining knowledge about such entities (and thus of our experimental methods) and independently of our ways of representing or processing such knowledge in computers. A relation like annotates is not ontological in this sense, as it links classes not to other classes in nature but rather to terms in a vocabulary that we ourselves have constructed. We focus also on general-purpose relations - relations which can be employed, in principle, in all biological ontologies - rather than on those specific relations (such as genome_of or sequence_of employed by OBO's Sequence Ontology) which apply only to biological entities of certain kinds. The latter will, however, need to be defined in due course in accordance with the methodology advanced here.
The ontologies in OBO are designed to serve as controlled vocabularies for expressing the results of biological science. Sentences of the form 'A relation B' (where 'A' and 'B' are terms in a biological ontology and 'relation' stands in for 'part_of' or some similar expression) can thus be conceived as expressing general statements about the corresponding biological classes or types. Assertions about corresponding instances or tokens (for example about the mass of this particular specimen in this particular Petri dish), while indispensable to biological research, do not belong to the general statements of biological science and thus they fall outside the scope of OBO and similar ontologies as these are presented to the user as finished products.
Yet such assertions are still relevant to ontologies. For it turns out that it is only by means of a detour through instances that the definitions and rules for coding relations between classes can be formulated in an intuitive and unambiguous - and thus reliably applicable - way.
We can distinguish, in fact, the following three kinds of binary relations:
<class, class>: for example, the is_a relation obtaining between the class SWR1 complex and the class chromatin remodeling complex, or between the class exocytosis and the class secretion;
<instance, class>: for example, the relation instance_of obtaining between this particular vesicle membrane and the class vesicle membrane, or between this particular instance of mitosis and the class mitosis;
<instance, instance>: for example, the relation of instance-level parthood (called part_of in what follows), obtaining between this particular vesicle membrane and the endomembrane system in the corresponding cell, or between this particular M phase of some mitotic cell cycle and the entire cell cycle of the particular cell involved.
Here classes and the relations between them are represented in italic; all other relations are picked out in bold.
Continuants and processes
The terms 'continuant' and 'process' are generalizations of GO's 'cellular component' and 'biological process' but applied to entities at all levels of granularity, from molecule to whole organism. Continuants are those entities which endure, or continue to exist, through time while undergoing different sorts of changes, including changes of place. Processes are entities that unfold themselves in successive temporal phases [16]. The terms 'continuant' and 'process' thus correspond to what, in the literature of philosophical ontology, are known respectively as 'things' (objects, endurants) and 'occurrents' (activities, events, perdurants) respectively. A continuant is what changes; a process is the change itself. The continuant classes relevant to biological ontologies include molecule, cell, membrane, organ; the process classes include ion transport, cell division, fat body development, breathing.
To formulate precise definitions of the <class, class> relations which form the target of ontology construction in biology we will need to employ a vocabulary that allows reference both to classes and to instances. For this we take advantage of the machinery of logic, and more specifically of the standard device of variables and quantifiers [17], using different sorts of variables to range across the classes and instances of continuants and processes, spatial regions and temporal instants, respectively. For the sake of intelligibility we use a semi-formal syntax, which can, however, be translated in a simple way into standard logical notation.
We use variables of the following sorts:
C, C1, ... to range over continuant classes;
P, P1, ... to range over process classes;
c, c1, ... to range over continuant instances;
p, p1, ... to range over process instances;
r, r1, ... to range over three-dimensional spatial regions;
t, t1, ... to range over instants of time.
In an expanded version of our formal machinery we will need also to incorporate further variables, ranging for example over temporal intervals, biological functions, attributes and values.
Note that continuants and processes form non-overlapping categories. This means in particular that no subtype or parthood relations cross the continuant-process divide. The tripartite structure of the GO recognizes this categorical exclusivity and extends it to functions also.
Continuants can be material (a mitochondrion, a cell, a membrane), or immaterial (a cavity, a conduit, an orifice), and this, too, is an exclusive divide. Immaterial continuants have much in common with spatial regions [18]. They are distinguished therefrom, however, in that they are parts of organisms, which means that, like material continuants, they move from one spatial region to another with the movements of their hosts.
The three-dimensional continuants that are our primary focus here typically have a top and a bottom, an anterior and a posterior, an interior and an exterior. Processes, in contrast, have a beginning, a middle and an end. Processes, but not continuants, can thus be partitioned along the time axis, so that, for example, your youth and your adulthood are temporal parts of that biological process which is your life.
As child and adult are continuants, so youth and adulthood are processes. We are thus clearly dealing here with two complementary - space-focused and time-focused - views of the same underlying subject matter, with determinate logical and ontological connections between them [16]. The framework advanced below allows us to capture these connections by incorporating reference to spatial regions and to temporal instants, both of which can be thought of as special kinds of instances.
We shall also need to distinguish two kinds of instance-level relations: those (applying to continuants) whose representations must involve a temporal index, and those (applying to processes) which do not. Note that the drawing of this distinction is still perfectly consistent with the fact that processes themselves occur in time, and that processes may be built out of successive subprocesses instantiating distinct classes.
Primitive instance-level relations
We cannot, on pain of infinite regress, define all relations, and this means that some relations must be accepted as primitive. The relations selected for this purpose should be self-explanatory and they should as far as possible be domain-neutral, which means that they should apply to entities in all regions of being and not just to those in the domain of biology.
Our choice of primitive relations is as follows:
c instance_of C at t - a primitive relation between a continuant instance and a class which it instantiates at a specific time
p instance_of P - a primitive relation between a process instance and a class which it instantiates holding independently of time
c part_of c1 at t - a primitive relation between two continuant instances and a time at which the one is part of the other
p part_of p1, r part_of r1 - a primitive relation of parthood, holding independently of time, either between process instances (one a subprocess of the other), or between spatial regions (one a subregion of the other)
c located_in r at t - a primitive relation between a continuant instance, a spatial region which it occupies, and a time
r adjacent_to r1 - a primitive relation of proximity between two disjoint continuants
t earlier t1 - a primitive relation between two times
c derives_from c1 - a primitive relation involving two distinct material continuants c and c1
p has_participant c at t - a primitive relation between a process, a continuant, and a time
p has_agent c at t - a primitive relation between a process, a continuant and a time at which the continuant is causally active in the process
This list includes only those <instance-instance> relations, together with one <instance-class> relation, which are needed for defining the <class, class> relations which are our principal target in this paper. The items on the list have been selected because they enjoy a high degree of intelligibility to the human authors and curators of biological ontologies. For purposes of supporting computer applications, however, the meanings of the corresponding relational expressions must be specified formally via axioms, for example in the case of 'part_of' by axioms of mereology (the theory of part and whole: see below), and in the case of 'earlier' by axioms governing a linear order [17]. The relation located_in will satisfy axioms to the effect that for every continuant there is some region in which it is located; instance_of will satisfy axioms to the effect that all classes have (at some stage in their existence) instances, and that all instances are instances of some class.
The formal machinery for reasoning with such axioms is in place, and a comprehensive set of axioms is being compiled. For the typical human user of biological ontologies, however, the listed primitive relations and associated axioms are designed to work invisibly behind the scenes. That is, they serve as part of the background framework that guides the construction and maintenance of such ontologies.
Results
Methodology
We employed a multi-stage methodology for the selection of the relations to be included in this ontology and for the formulation of corresponding definitions. First, a sample of researchers involved in ontology construction in the life sciences, representing different groups and including the co-authors of this paper, was asked to prepare lists of principal relations in light of their own specific experience but focusing on relations which would be: 'ontological' in the sense introduced above; 'general-purpose' in the sense that they apply across all biological domains; and also such as to manifest a high degree of universality (in the sense explained in the section 'Types of relational assertions' below). The submitted lists manifested a significant degree of overlap, which allowed us to prepare a core list in whose terms a large number of the remaining relations on the list could be simply defined.
A further constraint on the process was the goal of providing a simple formal definition for each included <class-class> relation. Those relations for which an appropriate simple definition could not be agreed upon were not included in this interim list. This includes most conspicuously relations involving analogs of the GO notion of molecular function. The relation has_agent was, however, included in light of a common understanding that the notion of agency would be involved in whatever candidate definition of function in biology is eventually accepted for use in OBO. This further constraint was chosen in light of the fact that our capacity to provide simple formal definitions - definitions which will at one and the same time be intelligible to ontology authors and curators and also able to support logic-based tools for automatic reasoning and consistency-checking - is the primary rationale for the methodology here advanced.
The two relations is_a and part_of were unproblematic candidates for inclusion in the resulting list (though providing simple definitions even for these relations was not, as we shall see, a simple matter). Is_a and part_of have established themselves as foundational to current ontologies. They have a central role in almost all domain ontologies, including the Foundational Model of Anatomy (FMA) [19,20], GO and other ontologies in OBO, as well as in influential top-level ontologies such as DOLCE [21] and in digitalized lexical resources such as WordNet [22].
In preparing our sample lists we drew on representatives not only of the OBO consortium but also of GALEN and the FMA (itself a candidate for inclusion in OBO). Our temporal relations draw on existing OBO practice (where transformation_of is a generalization of the develops_from relation used in OBO's cell and anatomy ontologies) and our participation relations draw on current work addressing the need to provide relations that link entities in different ontologies (for example entities in GO's process, function and component ontologies) and on an evolving Physiology Reference Ontology that is being developed in conjunction with the FMA [23], from which our spatial relations were extracted.
The OBO Relation Ontology
The first proposed version of the OBO Relation Ontology is shown in Table 1. We shall deal here with each of the ten relations listed in Table 1 in turn, providing rigorous yet easily understandable definitions.
Is_a
It is commonly assumed in the literature of knowledge representation that the relation is_a (meaning 'is a subtype of') can be identified with the subset or set inclusion relation with which we are familiar from mathematical set theory [17]. Instance_of functions on this reading as a counterpart of the usual set-theoretic membership relation, yielding a definition of A is_a B along the lines of: for all x, if x instance_of A, then x instance_of B. Unfortunately, this reading provides at best a necessary condition for the truth of A is_a B. It falls short of providing a sufficient condition for two reasons. The first is because it admits cases of contingent inclusion such as: bacterium in 90 mm × 18 mm glass Petri dish is_a bacterium, and the second is because it fails to take account of time, so that when applied to classes of continuants it yields false positives such as adult is_a child (because every instance of adult was at some time an instance of child).
We resolve the first problem by admitting as is_a links only assertions that reflect truths of biological science - assertions involving genuine biological class names (such as 'enzyme' or 'apoptosis') rather than, for example, commercial or indexical names (such as 'bacterium in this Petri dish'). The second problem we resolve by exploiting our machinery for taking account of time in the assertion of is_a relations involving continuants.
We can then define:
C is_a C1 = [definition] for all c, t, if c instance_of C at t then c instance_of C1 at t.
P is_a P1 = [definition] for all p, if p instance_of P then p instance_of P1.
Note how the device of logical quantifiers (for all ..., for some ...) allows us to refer to instances 'in general' - which means without the need to call on the proper names or indexical expressions (such as 'this' or 'here') which we use when referring to instances 'in specific'. Note also how instantiation for continuants involves a temporal argument. This reflects the fact that continuants, but not processes, can instantiate different classes in the course of their existence and yet preserve their identity.
For simplicity of expression we shall henceforth write 'Cct' and 'Pp', as abbreviations for: 'c instance_of C at t ' and 'p instance_of P ', respectively.
Part_of
Parthood as a relation between instances. The primitive instance-level relation p part_of p1 is illustrated in assertions such as: this instance of rhodopsin mediated phototransduction part_of this instance of visual perception.
This relation satisfies at least the following standard axioms of mereology: reflexivity (for all p, p part_of p); anti-symmetry (for all p, p1, if p part_of p1 and p1 part_of p then p and p1 are identical); and transitivity (for all p, p1, p2, if p part_of p1 and p1 part_of p2, then p part_of p2). Analogous axioms hold also for parthood as a relation between spatial regions.
For parthood as a relation between continuants, these axioms need to be modified to take account of the incorporation of a temporal argument. Thus for example the axiom of transitivity for continuants will assert that if c part_of c1 at t and c1 part_of c2 at t, then also c part_of c2 at t.
Parthood as a relation between classes. To define part_of as a relation between classes we again need to distinguish the two cases of continuants and processes, even though the explicit reference to instants of time now falls away. For continuants, we have C part_of C1 if and only if any instance of C at any time is an instance-level part of some instance of C1 at that time, as for example in: cell nucleus part_ of cell.
Formally:
C part_of C1 = [definition] for all c, t, if Cct then there is some c1 such that C1c1t and c part_of c1 at t.
Note the 'all-some' structure of this definition, a structure which will recur in almost all the relations treated here.
C part_of C1 defines a relational property of permanent parthood for Cs. It tells us that Cs, whenever they exist, exist as parts of C1s. We can also define in the obvious way C temporary_part_of C1 (every C exists at some time in its existence as part of some C1) and also C initial_part_of C1 (every C is such that it begins to exist as part of some instance of C1).
For processes, we have by analogy, P part_of P1 if and only if any instance of P is an instance-level part of some instance of P1, as for example in: M phase part_of cell cycle or neuroblast cell fate determination part_of neurogenesis. Formally:
P part_of P1 = [definition] for all p, if Pp then there is some p1 such that: P1p1 and p part_of p1.
An assertion to the effect that P part_of P1 thus tells us that Ps in general are in every case such as to exist as parts of P1s. P1s themselves, however, may exist without having Ps as parts (consider: menopause part_of aging).
Note that part_of is in fact two relations, one linking classes of continuants, the other linking classes of processes. While both of the mentioned relations are transitive, this does not mean that part_of relations could be inferred which would cross the continuant-process divide.
Located_in
Location as a relation between instances. The primitive instance-level relation c located_in r at t reflects the fact that each continuant is at any given time associated with exactly one spatial region, namely its exact location [24]. Following [25] we can use this relation to define a further instance-level location relation - not between a continuant and the region which it exactly occupies, but rather between one continuant and another. c is located in c1, in this sense, whenever the spatial region occupied by c is part_of the spatial region occupied by c1. Formally:
c located_in c1 at t = [definition] for some r, r1, c located_in r at t and c1 located_in r1 at t and r part_of r1.
Note that this relation comprehends both the relation of exact location between one continuant and another which obtains when r and r1 are identical (for example, when a portion of fluid exactly fills a cavity), as well as those sorts of inexact location relations which obtain, for example, between brain and head or between ovum and uterus.
Location as a relation between classes. To define location as a relation between classes - represented by sentences such as ribosome located_in cytoplasm, intracellular located_in cell - we now set:
C located_in C1 = [definition] for all c, t, if Cct then there is some c1 such that C1c1t and c located_in c1 at t.
Note that C located_in C1 is an assertion about Cs in general, which does not tell us anything about C1s in general (for example, that they have Cs located in them).
Contained_in
If c part_of c1 at t then we have also, by our definition and by the axioms of mereology applied to spatial regions, c located_in c1 at t. Thus, many examples of instance-level location relations for continuants are in fact cases of instance-level parthood. For material continuants location and parthood coincide. Containment is location not involving parthood, and arises only where some immaterial continuant is involved. To understand this relation, we first define overlap for continuants as follows:
C1 overlap c2 at t = [definition] for some c, c part_of c1 at t and c part_of c2 at t.
The containment relation on the instance level can then be defined as follows:
c contained_in c1 at t = [definition] c located_in c1 at t and not c overlap c1 at t.
On the class level this yields:
C contained_in C1 = [definition] for all c, t, if Cct then there is some c1 such that: C1c1t and c contained_in c1 at t.
Containment obtains in each case between material and immaterial continuants, for instance: lung contained_in thoracic cavity; bladder contained_in pelvic cavity. Hence containment is not a transitive relation.
Adjacent_to
We can define additional spatial relations by appealing to the primitive adjacent_to, a relation of proximity between disjoint continuants. Adjacent_to satisfies some of the axioms governing the relation referred to in the literature of qualitative topology as 'external connectedness' [26]. Analogs of other mereotopological relations (qualitative relations between spatial regions involving parthood, boundary and connectedness) (Figure 1) can also be defined, and these too can be applied to the material and immaterial continuants which occupy such regions on the instance level.
We define overlap for spatial regions as follows:
r1 overlap r2 = [definition] for some r, r part_of r1 and r part_of r2.
We then assert axiomatically that r1 adjacent_to r2 implies not r1 overlap r2
We can then define the counterpart relation of adjacency between classes as follows:
C adjacent_to C1 = [definition] for all c, t, if Cct, there is some c1 such that: C1c1t and c adjacent_to c1 at t.
Note that adjacent_to as thus defined is not a symmetric relation, in contrast to its instance-level counterpart. For it can be the case that Cs are in general such as to be adjacent to instances of C1 while no analogous statement holds for C1s in general in relation to instances of C. Examples are:
nuclear membrane adjacent_to cytoplasm
seminal vesicle adjacent_to urinary bladder
ovary adjacent_to parietal pelvic peritoneum.
We can, however, very simply define a symmetric relation of co-adjacency on the class level as follows:
C1 co-adjacent_to C2 = [definition] C1 adjacent_to C2 and C2 adjacent_to C1.
Examples are:
inner layer of plasma membrane co-adjacent_to outer layer of plasma membrane
right pulmonary artery co-adjacent_to right principal bronchus
urinary bladder of female co-adjacent_to parietal peritoneum of female pelvis.
Transformation_of
When an embryonic oenocyte (a type of insect cell) is transformed into a larval oenocyte, one and the same continuant entity preserves its identity while instantiating distinct classes at distinct times. The class-level relation transformation_of obtains between continuant classes C and C1 wherever each instance of the class C is such as to have existed at some earlier time as an instance of the distinct class C1 (see Figure 2). This relation is illustrated first of all at the molecular level of granularity by the relation between mature RNA and the pre-RNA from which it is processed, or between (UV-induced) thymine-dimer and thymine dinucleotide. At coarser levels of granularity it is illustrated by the transformations involved in the creation of red blood cells, for example, from reticulocyte to erythrocyte, and by processes of development, for example, from larva to pupa, or from (post-gastrular) embryo to fetus [27] or from child to adult. It is also manifest in pathological transformations, for example, of normal colon into carcinomatous colon. In each such case, one and the same continuant entity instantiates distinct classes at different times in virtue of phenotypic changes.
As definition for this relation we offer:
C transformation_of C1 = [definition] C and C1 for all c, t, if Cct, then there is some t1 such that C1ct1, and t1 earlier t, and there is no t2 such that Cct2 and C1ct2.
That is to say, the class C is a transformation of the class C1 if and only if every instance c of C is at some earlier time an instance of C1, and there is no time at which it is an instance of both C and C1. (The final clause, which asserts that C and C1 do not share instances at a time, is inserted in order to rule out, for example, adult human transformation_of human.)
Note that C transformation_of C1 is a statement about Cs in general. It does not tell us of C1s in general that each gives rise to some C which stands to it in a transformation_of relation.
Derives_from
Derivation as a relation between instances. The temporal relation of derivation is more complex. Transformation, on the instance level, is just the relation of identity: each adult is identical to some child existing at some earlier time. Derivation on the instance-level is a relation holding between non-identicals. More precisely, it holds between distinct material continuants when one succeeds the other across a temporal divide in such a way that at least a biologically significant portion of the matter of the earlier continuant is inherited by the later. Thus we will have axioms to the effect that from c derives_from c1 we can infer that c and c1 are not identical and that there is some instant of time t such that c1 exists only prior to and c only subsequent to t. We will also be able to infer that the spatial region occupied by c as it begins to exist at t overlaps with the spatial region occupied by c1 as it ceases to exist in the same instant.
Three simple kinds of instance-level derivation can then be distinguished (Figure 3): first, the succession of one single continuant by another single continuant across a temporal threshold (for example, this blastocyst derives from this zygote); second, the fusion of two or more continuants into one continuant (for example, this zygote derives from this sperm and from this ovum); and third, the fission of an earlier single continuant to create a plurality of later continuants (for example, these promyelocytes derive from this myeoloblast). In all cases we have two continuants c and c1 which are such that c begins to exist at the same instant of time at which c1 ceases to exist, and at least a significant portion of the matter of c1 is inherited by its successor c.
Derivation of the first type is still essentially weaker than transformation, for the latter involves the identity of the continuant instances existing on either side of the relevant temporal divide. In derivation of the second type, the successor continuant takes the bulk of its matter from a plurality of precursors, where in cases of the third type, the bulk of the matter of a single precursor continuant is shared among a plurality of successors. We can also represent more complex cases where transformation and an analog of derivation are combined, for example in the case of budding in yeast [27], where one continuant continues to exist identically through a process wherein a second continuant floats free from its host; or in absorption, where one continuant continues to exist identically through a process wherein it absorbs another continuant, for example through digestion.
Derivation as a relation between classes. To avoid troubling counter-examples, the relation of derivation we are seeking on the class level must be defined in two steps. First, the class-level counterpart of the relation of derivation on the instance level is identified as a relation of immediate derivation:
C derives_immediately_from C1 = [definition] for all c, t, if Cct, then there is some c1,t1, such that: t1 earlier t and C1c1t1 and c derives_from c1.
The more general class level derivation relation must then be defined in terms of chains of immediate derivation relations, as follows:
C derives_from C1 = [definition] there is some sequence C = Ck, Ck-1, ..., C2, C1, such that for each Ci (1 ≤ i < k), Ci+1 derives_immediately_from Ci.
In this way we can represent cases of derivation involved in the formation of lineages where there occurs a sequence of cell divisions or speciation events.
Preceded_by
With the primitive relations has_participant and earlier at our disposal we can define the instance-level relation p occurring_at t as follows:
p occurring_at t = [definition] for some c, p has_participant c at t.
We can then define:
c exists_at t = [definition] for some p, p has_participant c at t
p preceded_by p1 = [definition] for all t, t1, if p occurring_at t and p1 occurring_at t1, then t1 earlier t
t first_instant p = [definition] p occurring_at t and for all t1, if t1 earlier t, then not p occurring_at t1
t last_instant p = [definition] p occurring_at t and for all t1, if t earlier t1, then not p occurring_at t1
p immediately_preceded_by p1 = [definition] for some t, t first_instant p and t last_instant p1.
At the class level we have:
P preceded_by P1 = [definition] for all p, if Pp then there is some p1 such that P1p1and p preceded_by p1.
An example is: translation preceded_by transcription; aging preceded_by development (not however death preceded_by aging). Where derives_from links classes of continuants, preceded_by links classes of processes. Clearly, however, these two relations are not independent of each other. Thus if cells of type C1 derive_from cells of type C, then any cell division involving an instance of C1 in a given lineage is preceded_by cellular processes involving an instance of C.
The assertion P preceded_by P1 tells us something about Ps in general: that is, it tells us something about what happened earlier, given what we know about what happened later. Thus it does not provide information pointing in the opposite direction, concerning instances of P1 in general; that is, that each is such as to be succeeded by some instance of P. Note that an assertion to the effect that P preceded_by P1 is rather weak; it tells us little about the relations between the underlying instances in virtue of which the preceded_by relation obtains. Typically we will be interested in stronger relations, for example in the relation immediately_preceded_by, or in relations which combine preceded_by with a condition to the effect that the corresponding instances of P and P1 share participants, or that their participants are connected by relations of derivation, or (as a first step along the road to a treatment of causality) that the one process in some way affects (for example, initiates or regulates) the other.
Has_participant
Has_participant is a primitive instance-level relation between a process, a continuant, and a time at which the continuant participates in some way in the process. The relation obtains, for example, when this particular process of oxygen exchange across this particular alveolar membrane has_participant this particular sample of hemoglobin at this particular time.
To define the class-level counterpart of the participation relation we set:
P has_participant C = [definition] for all p, if Pp then there is some c, t such that Cct and p has_participant c at t.
Examples are:
cell transport has_participant cell
death has_participant organism
breathing has_participant thorax.
Once again, P has_participant C provides information only about Ps in general (that is, that they require instances of C as bearers).
Has_agent
Special types of participation can be distinguished according to whether a continuant is agent or patient in a process (for a survey see [28].) Here we focus on the factor of agency, which is involved, for example, when an adult engages in adult walking behavior. It is not involved when the same adult is the victim of an infection. Synonyms of 'is agent in' include: 'actively participates in', 'does', 'executes', 'performs', and so forth.
We introduce the primitive instance-level relation has_agent, which obtains between a process, a continuant and a time whenever the continuant is a participant in the process and is at the same time directly causally responsible for its occurrence. Thus we have an axiom to the effect that agency implies participation: for all p, c, t, if p has_agent c at t, then p has_participant c at t. In addition we will have axioms to the effect that only material continuants can fill the agent role, that if c fills the agent role at t, then c must have existed at times earlier than t, that it must exercise its agent role for an interval of time including t, and so on.
We can then define the class-level relation has_agent by stipulating:
P has_agent C = [definition] for all p, if Pp then there is some c, t such that Cct and p has_agent c at t
This relation gives us the means to capture the directionality (the from-to) nature of biological processes such as signaling, transcription, and expression, via assertions, for example, to the effect that in an interaction between molecules of types m1 and m2 it is molecules of the first type that play the role of agent.
One privileged type of agency consists in the realization of a biological function. To say that a continuant has a function is to assert, in first approximation, that it is predisposed (has the potential, the casual power) to cause (to realize as agent) a process of a certain type. Thus to say that your heart has the function: to pump blood is to assert that your heart is predisposed to realize as agent a process of the type pumping blood [29]. Regulation, promotion, inhibition, suppression, activation, and so forth, are among the varieties of agency that fall under this heading.
On the other hand, many processes - such as metabolic reactions involving enzymes, cofactors, and metabolites - involve no clear factor of agent participation, but rather require more nuanced classifications of the roles of participants - as acceptors or donors, for example. Hence the has_agent relation should be used in curation with special care. It should be borne in mind in this connection that agency is in every case a matter of the imposition of direct causal influence of a continuant in a process (a constraint that is designed to rule out inheritance of agency along causal chains), and also that (by our definition) only continuants can be agents. Where biologists describe processes as agents, for example, in talking about the effects of diffusion in development and differentiation, such phenomena are of a type that call for an expansion of our proposed Relation Ontology in the direction, again, of a treatment of the factor of causality.
Discussion
The logic of biological relations
Inverse and reciprocal relations
The inverse of a relation R is defined as that relation which obtains between each pair of relata of R when taken in reverse order. Inverses can be unproblematically defined for all instance-level relations. What, then, of inverses for class-level relations? The inverse relation for is_a can be defined trivially as follows:
A has_subclass B = [definition] B is_a A.
For the remaining class-level relations on our list, in contrast, the issue of corresponding inverses is more problematic [7]. Thus, while we have the true relational assertion human testis part_of human - which means that all instances of human testis are part of instances of some human - there is no corresponding true relational assertion linking instances of human to instances of human testis as their parts. For these remaining relations we need to work not with inverses but rather with what, following GALEN, we can call reciprocal relations. These are defined using the same family of instance-level primitives we introduced earlier. As reciprocal relations for the two varieties of part_of we have:
C has_part C1 = [definition] for all c, t, if Cct then there is some c1 such that C1c1t and c1 part_of c at t
P has_part P1 = [definition] for all p, if Pp then there is some p1 such that P1p1 and p1 part_of p
Note that from A part_of B we cannot infer that B has_ part A; similarly, from A has_ part B we cannot infer that B part_of A. Thus cell nucleus part_of cell, but not cell has_part cell nucleus; running has_ part breathing, but not breathing part_of running. A third significant relation conjoining part_of and has_part can be defined as [6,30]:
C integral_part_of C1 = [definition] C part_of C1 and C1 has_part C.
For contained_in we have similarly the reciprocal relation:
C contains C1 = [definition] for all C, t, if Cct then there is some c1 such that: C1c1t and c located_in c at t
For participation we can usefully define two alternative reciprocal relations:
C sometimes_ participates_in P = [definition] for all c there is some t and some p such that Cct and Pp and p has_participant c at t
C always_participates_in P = [definition] for all c, t, if Cct then there is some p such that Pp and p has_participant c at t
We can also define, for example, what it is for continuants of a given type to participate at every stage in a process of a given type. Thus if a sperm participates in the penetration of an ovum, then it does so throughout the penetration.
Types of relational assertions
In light of the above, we can now observe certain differences in what we might call the relative universality of class-level relational assertions. There are many cases, above all involving is_a relations, where relational assertions hold with a maximal degree of universality, which means that they hold for every instance of the classes in question because they are a matter of analytic connections, that is, connections resting on the compositional nature of the class terms involved [10], as, for example, in: eukaryotic cell is_a cell, or adult walking behavior has_participant adult. (Contrast, adult participates_in adult walking behavior.)
There are also other kinds of statements enjoying a high degree of universality, for example: penetration of ovum has_participant sperm. The first of our two corresponding reciprocal statements - sperm participates_in penetration of ovum - is in contrast true only in relation to certain isolated instances of sperm, and the second of our reciprocal statements - sperm always_participates_in penetration of ovum - is true in relation to no instances at all.
It then seems reasonable to insist that biomedical ontologies should reflect those sorts of biological assertions that enjoy a high degree of universality (typically assertions involving just one of each pair of reciprocal relations).
Tools for ontology curation
We hope that, by providing clear and unambiguous specifications of what the class-level relational expressions used in biological ontologies mean, our formal definitions will assist curators engaged in ontology creation and maintenance. The corresponding definitions are summarized in Table 2, which also contains representative examples for each of the relations distinguished.
Our definitions are designed to ensure that the corresponding general-purpose relational expressions are used in a uniform way in all biological ontologies. In this way we shall be in a position to contribute to the realization of the goal of bringing about a high degree of interoperability even where ontologies are produced by different groups and for different purposes. These definitions are designed also to enable the automatic detection of errors in biomedical ontologies, for example by allowing the construction of extensions of OBO-Edit and similar tools with the facility to test whether given relations are employed in an ontology in such a way as to involve relata of the appropriate types [31] or in such a way as to have the formal characteristics, such as transitivity or reflexivity, dictated by the definitions (Table 3). The framework can also support reasoning applications designed to enable the automated derivation of information from existing bodies of knowledge - for example to infer the parts of a given cell continuant via the traversal of a part_of hierarchy - including instance-based knowledge derived from the clinical record.
Conclusion
The Relation Ontology outlined above arose through collaboration between formal ontologists and biologists in the OBO, FMA and GALEN research groups and also incorporates suggestions from a number of other authors and curators of biomedical ontologies. It is designed to be large enough to overcome some of the problems arising in GO and similar systems as a result of the paucity of resources available hitherto for expressing relations between the classes in such ontologies [32]. It is this paucity of resources, above all, which gives rise to cases of multiple inheritance in GO as presently constructed, and we note here that multiple inheritance often goes hand in hand with errors in ontology construction not least because it encourages a relaxed reading of is_a (often a reading which involves the assertion of is_a relations which erroneously cross the divide between different ontological categories) [5,33]. Our present framework can contribute to error resolution not only by dictating a common interpretation of is_a which can serve as orientation for ontology authors and curators in their future work, but also by providing richer resources for the assertion of class-class relations within and between ontologies in such a way that the appeal to contrived and error-prone is_a relations can be more easily avoided.
At the same time our suite of relations has been designed to be sufficiently small to attract wide acceptance in a range of different types of life-science communities. Where the latter use further, general-purpose or domain-specific relations of their own, we plan in due course to subject such relations to the same kind of analysis as presented here in order to preserve interoperability. The Relation Ontology has been incorporated into the OBO ontology library [34] and curators of the GO and FMA ontologies and also of the ChEBI chemical entities vocabulary [35] are already applying the relevant parts of the ontology in their work. The ontology has already been used to find errors not only in GO but also in SNOMED [36]. It is also being applied systematically in evaluations of the NCI Thesaurus [37] and the UMLS (Unified Medical Language System) Semantic Network of the National Library of Medicine. We are currently testing methodologies to obtain reliable quantitative evaluations of the utility of the proposed framework for purposes of ontology authoring and also for use in annotation and reasoning. We are also testing ways in which the framework can be expanded through the admission of pre-coordinated disjunctions (for example: either derivation or transformation), which can allow the coding of information in those cases where the precise nature of the relations involved is insufficiently clear to allow unique assignment.
The Relation Ontology will be evaluated on two levels. First, on whether it succeeds in preventing those characteristic kinds of errors which have been associated with a poor treatment of relations in biomedical ontologies in the past. Second, and more important, on whether it helps to achieve greater interoperability of biomedical ontologies and thus to improve reasoning about biological phenomena.
Acknowledgements
Work on this paper was carried out under the auspices of the Wolfgang Paul Program of the Alexander von Humboldt Foundation, the EU Network of Excellence in Medical Informatics and Semantic Data Mining, the Project 'Forms of Life' sponsored by the Volkswagen Foundation, and the DARPA Virtual Soldier Project. Thanks go to Michael Ashburner, Fabrice Correia, Maureen Donnelly, Kai Hauser, Win Hyde, Ingvar Johansson, Janet Kelso, Suzanna Lewis, Katherine Munn, Maria Reicher, Alan Ruttenberg, Mark Scala, Stefan Schulz, Neil Williams, Lina Yip, Sumi Yoshikawa, and anonymous referees for valuable comments.
Figures and Tables
Figure 1 Standard mereotopological relations between spatial regions.
Figure 2 Transformation.
Figure 3 Three simple cases of derivation. (a) Continuation; (b) fusion; (c) fission.
Table 1 First version of the OBO Relation Ontology
Foundational relations
is_a
part_of
Spatial relations (connecting one entity to another in terms of relations between the spatial regions they occupy)
located_in
contained_in
adjacent_to
Temporal relations (connecting entities existing at different times)
transformation_of
derives_from
preceded_by
Participation relations (connecting processes to their bearers)
has_participant
has_agent
Table 2 Definitions and examples of class-level relations
Relations and relata Definitions Examples
C is_a C1; Cs and C1s are continuants Every C at any time is at the same time a C1 myelin is_a lipoprotein
serotonin is_a biogenic amine
mitochondrion is_a membranous cytoplasmic organelle
protein kinase is_a kinase
DNA is_a nucleic acid
P is_a P1; Ps and P1s are processes Every P is a P1 endomitosos is_a DNA replication
catabolic process is_a metabolic process
photosynthesis is_a physiological process
gonad development is_a organogenesis
intracellular signaling cascade is_a signal transduction
C part_of C1; Cs and C1s are continuants Every C at any time is part of some C1 at the same time mitochondrial matrix part_of mitochondrion
microtubule part_of cytoskeleton
nuclear pore complex part_of nuclear membrane
nucleoplasm part_of nucleus
promotor part_of gene
P part_of P1; Ps and P1s are processes Every P is part of some P1 gastrulation part_of embryonic development
cystoblast cell division part_of germ cell development
cytokinesis part_of cell proliferation
transcription part_of gene expression
neurotransmitter release part_of synaptic transmission
C located_in C1; Cs and C1s are continuants Every C at any given time occupies a spatial region which is part of the region occupied by some C1 at the same time 66s pre-ribosome located_in nucleolus
intron located_in gene
nucleolus located_in nucleus
membrane receptor located_in cell membrane
chlorophyll located_in thylakoid
C contained_in C1; Cs are material continuants, C1s are immaterial continuants (holes, cavities) Every C at any given time is located in but shares no parts in common with some C1 at the same time thoracic aorta contained_in posterior mediastinal cavity
cytosol contained_in cell compartment space
thylakoid contained_in chloroplast membrane
synaptic vesicle contained_in neuron
C adjacent_to C1; Cs and C1s are continuants Every C at any time is proximate to some C1 at the same time Golgi apparatus adjacent_to endoplasmic reticulum
intron adjacent_to exon
cell wall adjacent_to cytoplasm
periplasm adjacent_to plasma membrane
presynaptic membrane adjacent_to synaptic cleft
C transformation_of C1; Cs and C1s are material continuants Every C at any time is identical with some C1 at some earlier time facultative heterochromatin transformation_of euchromatin
mature mRNA transformation_of pre-mRNA
hemosiderin transformation_of hemoglobin
red blood cell transformation_of reticulocyte
fetus transformation_of embryo
C derives_from C1; Cs and C1s are material continuants Every C is such that in the first moment of its existence it occupies a spatial region which overlaps the spatial region occupied by some C1 in the last moment of its existence plasma cell derives_from B lymphocyte
fatty acid derives_from triglyceride
triple oxygen molecule derives_from oxygen molecule
Barr body derives_from X-chromosome
mammal derives_from gamete
P preceded_by P1; Ps and P1s are processes Every P is such that there is some earlier P1 translation preceded_by transcription
meiosis preceded_by chromosome duplication
cytokinesis preceded_by DNA replication
apoptotic cell death preceded_by nuclear chromatin degradation
digestion preceded_by ingestion
P has_participant C; Ps are processes, Cs are continuants Every P involves some C as participant mitochondrial acetylCoA formation has_participant pyruvate dehydrogenase complex
translation has_participant amino acid
photosynthesis has_participant chlorophyll
apoptosis has_participant cell
cell division has_participant chromosome
P has_agent C; Ps are processes, Cs are material continuants Every P involves some C as agent (the C is involved in and is causally responsible for the P) gene expression has_agent RNA polymerase
signal transduction has_agent receptor
pathogenesis has_agent pathogen
transcription has_agent RNA polymerase
translation has_agent ribosome
Table 3 Some properties of the relations in the OBO Relation Ontology
Relation Transitive Symmetric Reflexive Antisymmetric
is_a + - + +
part_of + - + +
located_in + - + -
contained_in - - - -
adjacent_to - - - -
transformation_of + - - -
derives_ from + - - -
preceded_by + - - -
has_participant - - - -
has_agent - - - -
==== Refs
OBO: Open Biomedical Ontologies
Mungall C OBOL: integrating language and meaning in bio-ontologies. Comp Funct Genomics 2004 5 509 520 10.1002/cfg.435
Gene Ontology Consortium Creating the Gene Ontology resource: design and implementation. Genome Res 2001 11 1425 1433 11483584 10.1101/gr.180801
Bada M Stevens R Goble C Gil Y Ashburner M Blake JA Cherry JM Harris M Lewis S A short study on the success of the GeneOntology. J Web Semantics 2004 1 235 240
Smith B Köhler J Kumar A On the application of formal principles to life science data: a case study in the Gene Ontology. DILS 2004: Data Integration in the Life Sciences Lecture Notes in Computer Science 2994 2004 124 139
Smith B Rosse C The role of foundational relations in the alignment of biomedical ontologies. Proceedings Medinf 2004 2004 Amsterdam: IOS Press 444 448
Smith B Kumar A On controlled vocabularies in bioinformatics: a case study in the Gene Ontology. BioSilico: Inform Technol Drug Discovery 2004 2 246 252 10.1016/S1741-8364(04)02424-2
Smith B Williams J Schulze-Kremer S The ontology of the Gene Ontology. Proc AMIA Symp 2003 609 13 14728245
Ogren PV Cohen KB Acquaah-Mensah GK Eberlein J Hunter L The compositional structure of Gene Ontology terms. Pac Symp Biocomput 2004 214 225 14992505
Ogren P Bretonnel Cohen K Hunter L Implications of compositionality in the Gene Ontology for its curation and usage. Pac Symp Biocomput 2005 174 185 15759624
Bard J Rhee SY Ashburner M An ontology for cell types. Genome Biol 2005 6 R21 15693950 10.1186/gb-2005-6-2-r21
Wroe C Stevens R Goble CA Ashburner M An evolutionary methodology to migrate the Gene Ontology to a Description Logic environment using DAML+OIL. Pac Symp Biocomput 2003 624 635 12603063
Smith B Beyond concepts: ontology as reality representation. Formal Ontology and Information Systems 2004 2004 Amsterdam: IOS Press 73 84
Levesque HJ Brachman RJ A fundamental tradeoff in knowledge representation and reasoning. Readings in Knowledge Representation 1985 San Francisco: Morgan Kaufman 41 70
Rogers J Rector AL The GALEN ontology. Medical Informatics Europe 1996 1996 Amsterdam: IOS Press 174 178
Grenon P Smith B Goldberg L Biodynamic ontology: applying BFO in the biomedical domain. Ontologies in Medicine 2004 Amsterdam: IOS Press 20 38
Stoll R Set Theory and Logic 1979 New York: Dover Publications
Casati R Varzi AC Holes and Other Superficialities 1994 Cambridge, MA: MIT Press
Rosse C Mejino JLV Jr A reference ontology for bioinformatics: the Foundational Model of Anatomy. J Biomed Inform 2003 36 478 500 14759820 10.1016/j.jbi.2003.11.007
Rogers J Rector AL GALEN's model of parts and wholes: experience and comparisons. Proceedings AMIA Symposium 2000 2000 Bethesda, MD: American Medical Informatics Association 819 823
Gangemi A Guarino N Masolo C Oltramari A Sweetening WordNet with DOLCE. AI Magazine 2003 24 13 24
Fellbaum C Ed Wordnet An Electronic Lexical Database 1998 Cambridge, MA: MIT Press
Cook DL Mejino JLV JrRosse C Evolution of a Foundational Model of Physiology: symbolic representation for functional bioinformatics. Proceedings MedInfo 2004 2004 Amsterdam: IOS Press 336 340
Bittner T Axioms for parthood and containment relations in bio-ontologies. KR-MED 2004: Workshop on Formal Biomedical Knowledge Representation 2004 Aachen: University of Aachen 4 11
Donnelly M Layered mereotopology. Proceedings 18th Joint International Conference on Artificial Intelligence 2003 San Francisco: Morgan Kaufman 1269 1274
Smith B Mereotopology: a theory of parts and boundaries. Data Knowledge Eng 1996 20 287 303 10.1016/S0169-023X(96)00015-8
Smith B Brogaard B Sixteen days. J Med Philos 2003 28 45 78 12715281 10.1076/jmep.28.1.45.14172
Smith B Grenon P The cornucopia of formal-ontological relations. Dialectica 2004 58 279 296
Johansson I Smith B Munn K Tsikolia N Elsner K Ernst D Siebert D Functional anatomy: a taxonomic proposal. Acta Biotheoret 2005
Schulz S Hahn U Towards a computational paradigm for biomedical structure. KR-MED 2004: Workshop on Formal Biomedical Knowledge Representation 2004 Aachen: University of Aachen 63 71
dos Santos MC Dhaen C Fielding M Ceusters W Philosophical scrutiny for run-time support of application ontology development. Formal Ontology and Information Systems 2004 Amsterdam: IOS Press 342 352
Kumar A Smith B Borgelt C Dependence relationships between Gene Ontology terms based on TIGR gene product annotations. Proceedings CompuTerm 2004 2004 Geneva: COLING 31 38
Bouaud J Bachimont B Charlet J Zweigenbaum P Acquisition and structuring of an ontology within conceptual graphs. Proceedings 2nd International Conference on Conceptual Structures: Workshop on Knowledge Acquisition using Conceptual Graph Theory Lecture Notes Computer Sci 1994 835 1 25
OBO Relationship Ontology
ChEBI: Chemical Entities of Biological Interest
Ceusters W Smith B Kumar A Dhaen C Ontology-based error detection in SNOMED-CT. Proceedings Medinfo 2004 2004 Amsterdam: IOS Press 482 486
Ceusters W Smith B Goldberg L A terminological and ontological analysis of the NCI Thesaurus. Meth Inform Medicine 2005
| 15892875 | PMC1175959 | CC BY | 2021-01-04 16:05:39 | no | Genome Biol. 2005 May 3; 6(5):R47 | latin-1 | Genome Biol | 2,005 | 10.1186/gb-2005-6-5-r47 | oa_comm |
==== Front
Genome BiolGenome Biology1465-69061465-6914BioMed Central London gb-2005-6-6-r481596080010.1186/gb-2005-6-6-r48ResearchMechanisms of aging in senescence-accelerated mice Carter Todd A [email protected] Jennifer A [email protected] Shigeo 2Fuchs Sebastian 1Helton Robert 1Swaroop Anand 23Lockhart David J 4Barlow Carrolee [email protected] The Salk Institute for Biological Studies, La Jolla, CA 92037, USA2 Department of Ophthalmology and Visual Sciences, University of Michigan, Ann Arbor, MI 48105, USA3 Department of Human Genetics, University of Michigan, Ann Arbor, MI 48105, USA4 Ambit Biosciences, San Diego CA 92121, USA5 Current address: BrainCells Inc., 10835 Road to the Cure, San Diego, CA 92121, USA2005 1 6 2005 6 6 R48 R48 16 12 2004 9 3 2005 5 5 2005 Copyright © 2005 Carter 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.
Gene-expression analysis and polymorphism screening to study molecular senescence of the retina and hippocampus in two rare inbred mouse models of accelerated neurological senescence as well as a related and an unrelated normal strain showed that most age-related gene expression changes were strain-specific.
Background
Progressive neurological dysfunction is a key aspect of human aging. Because of underlying differences in the aging of mice and humans, useful mouse models have been difficult to obtain and study. We have used gene-expression analysis and polymorphism screening to study molecular senescence of the retina and hippocampus in two rare inbred mouse models of accelerated neurological senescence (SAMP8 and SAMP10) that closely mimic human neurological aging, and in a related normal strain (SAMR1) and an unrelated normal strain (C57BL/6J).
Results
The majority of age-related gene expression changes were strain-specific, with only a few common pathways found for normal and accelerated neurological aging. Polymorphism screening led to the identification of mutations that could have a direct impact on important disease processes, including a mutation in a fibroblast growth factor gene, Fgf1, and a mutation in and ectopic expression of the gene for the chemokine CCL19, which is involved in the inflammatory response.
Conclusion
We show that combining the study of inbred mouse strains with interesting traits and gene-expression profiling can lead to the discovery of genes important for complex phenotypes. Furthermore, full-genome polymorphism detection, sequencing and gene-expression profiling of inbred mouse strains with interesting phenotypic differences may provide unique insights into the molecular genetics of late-manifesting complex diseases.
==== Body
Background
Aging is defined by an increase in the probability of death over time associated with characteristic changes in phenotype [1]. Changes in the global control of transcription have been directly implicated in the aging process in yeast, and increased histone deacetylation activity (a process involved in chromatin silencing) results in extended life span in Caenorhabditis elegans [2-4]. Genomic instability has also been implicated as a causative agent in cellular senescence in mammals. This relationship between genomic instability and aging in mammals is supported by work demonstrating a correlation between senescence and the loss of ribosomal DNA, increases in chromosomal abnormalities and telomere shortening [1,5-7]. In addition, certain mutations in humans can accelerate aging-specific events, resulting in progeric diseases that include Hutchinson-Gilford syndrome, Werner syndrome, Cockayne syndrome and xeroderma pigmentosum [8-10]. Except for Hutchinson-Gilford syndrome, each of these disorders results from mutations in DNA repair proteins, suggesting that a stochastic build-up of errors in DNA could form the basis for some common traits of aging. Recent studies have indicated that Hutchinson-Gilford syndrome is caused by specific mutations in lamin A, a gene involved in structural integrity of the nuclear membrane [11,12]. Interestingly, some genetic disorders that exhibit aspects of accelerated senescence also demonstrate genomic instability, including several mentioned above as well as ataxia-telangiectasia and Bloom's syndrome [13-17].
While single-gene progerias can provide insight into age-related processes, most patients exhibit only a subset of the phenotypes associated with aging. Thus, the process may be fundamentally different from normal aging, which involves multiple events and tissues. To complement studies of single-gene progerias and other models of mammalian senescence, we have chosen to study a more complex model of aging: the senescence-accelerated mouse (SAM) strains. The senescence-accelerated mice are a collection of inbred mouse strains developed as models of accelerated aging, and include nine short-lived, senescence-accelerated mouse prone strains (SAMP) and three longer lived control strains designated senescence-accelerated mouse resistant (SAMR) [18]. The SAMP strains exhibit several features that make them interesting models of human aging, including age-associated early onset of senile amyloidosis, degenerative arthropathy, cataracts, osteoporosis and osteoarthritis, reduced fecundity and early loss of fertility [18-20]. Mapping studies have been limited to microsatellite haplotype analyses characterizing the genetic relationships between the SAM strains [21]. In addition, there is currently no genome sequence available for these strains, making it difficult to use comparative genomics to identify genetic differences responsible for their phenotypic differences. Furthermore, the strains involved in these studies require continual trait-based selection to maintain the phenotype. As standard quantitative trait locus mapping approaches would be extremely difficult with such strains, we sought to test the hypothesis that gene-expression profiling combined with candidate gene sequencing would lead to the identification of mutations and/or expression changes that track with the strain-specific phenotypes, thereby allowing us to identify relevant pathways and generate candidate genes for future experiments.
Our study focused on the identification of genes involved in neurological aspects of aging, using two SAMP strains: SAMP8/Ta (S8) and SAMP10//Ta (S10), and two control strains: the related SAMR strain SAMR1TA (SR1) and a commonly used inbred mouse strain C57BL/6J (B6J). The S8 and S10 strains exhibit age-related behavioral and neuropathological phenotypes, in addition to osteoporosis and premature loss of fertility, that make them particularly useful models of human neurological aging [22-25]. These phenotypes include deficits in learning and memory, emotional disorders and abnormal circadian rhythms [18,26]. S8 mice also develop a severe age-related impairment in acquisition and retention of the passive avoidance response, as well as a reduced-anxiety behavior [23,27]. Old S10 mice exhibit behavioral depression on tail suspension and forced swimming tests [23]. A unique pathological feature of senescence in S10 mice is an age-related atrophy of the brain [28]. Neuron shrinkage and degeneration in S10 mice result in progressive decrease of mean brain weight beginning at 4 months of age [28]. In addition to the neurobehavioral and physiological phenotypes, S8 mice demonstrate an age-related degeneration of the retinal pigment epithelium-Bruch's membrane-choriocapillaris complex, and a degeneration of receptor cells and ganglion neurons in the retina suggestive of age-related macular degeneration in humans [29].
The S8 and S10 strains are also interesting in that although inbred, trait-based selection is necessary to maintain the phenotype of the age-associated disorders over generations [30]. Thus, while the phenotypes are heritable, they are clearly part of a complex trait that probably involves the interaction of multiple genes and/or alleles, suggesting that these strains may better model the processes associated with mammalian aging than single-gene progeria models.
To explore the events involved in the molecular senescence of the mammalian brain, we established and aged the colonies, verified the phenotypes, and performed gene-expression analysis of the retina and hippocampus in S8 and S10 mice using oligonucleotide microarrays [31]. As a control, we studied the SAMR strain SR1. All SAM strains were originally derived from the AKR/J inbred mouse strain, but SR1 demonstrates a longer life span and lacks the accelerated senescence that is a hallmark of the SAMP strains. Furthermore, we also analyzed the gene-expression data in the context of an additional, unrelated inbred mouse strain, B6J, to distinguish strain-background specific (AKR-specific) from more general changes of the aging transcriptome. Finally, we took advantage of a focused polymorphism screen to identify two genes harboring mutations in the SAMP strains that may play important roles in their accelerated-aging phenotypes.
Results
Verification of phenotype in SAM Strains
As accelerated aging of the senescence-prone mouse strains is a complex phenotype, we monitored the life span, pathology, fecundity and learning and memory behavior of all generations of mice to validate the accelerated-senescence phenotype in our facility and employed retrospective pedigree selection on S8 and S10 mice as previously described [30] (see Additional data file 2). As expected, both S8 and S10 strains had increased mortality and morbidity with age relative to SR1 and B6 mice (Figure 1a; B6 data from Pugh et al. [32]). S8 mice had the shortest life span, with a median life span 39% shorter than that of SR1. S10 mice showed an 18% decrease in median life span. These data are consistent with the reported life spans of SAM strains reared under conventional conditions, where the median survival time of all SAMP strains (including not only S8 and S10 but also other accelerated-senescence strains) is reported to be 40% less than that of SAMR strains [20]. Whereas the single greatest cause of death in the SAMR1 strain was cancer (consistent with the original AKR/J strain from which the SAM strains were developed), the SAMP10 and SAMP8 mice showed a decrease in cancer-related deaths and an increase in death due to infection or a wasting syndrome consistent with neurological dysfunction. Finally, an analysis of litter size also showed a significant decrease in fecundity in both S8 and S10 mice as compared with SR, consistent with accelerated aging of the reproductive organs (data not shown) [33].
Most relevant to the studies performed in our laboratory, a progressive deterioration in learning and/or memory performance has been reported in S8 and S10 mice [22,23]. To confirm and elucidate these phenotypes, behavioral analysis of our colonies was performed using a single-trial passive avoidance paradigm, in which shorter latency to entering a darkened chamber indicates a lower retention (memory) of a previous foot shock. This test was performed on both younger (average age of 16 weeks) and older (average age of 81 weeks) mice (Figure 1b). At 16 weeks of age, S8 mice demonstrate an average latency 52.1 seconds shorter than SR1 and 16 week-old S10 mice demonstrate an average latency 56.7 seconds shorter than SR1. These differences between the SAMP strains and the SAMR strain are significant using a two-tailed Student's t-test (p < 0.003 for SAMP8 versus SAMR1 and p < 0.002 for SAMP10 versus SAMR1). At 81 weeks of age, the retention deficits for S8 and S10 mice have worsened, with S8 mice showing a 119% decrease and S10 mice showing a 52% decrease in performance, while SR1 mice show only a 14% decrease. Consistent with previously published results, the SAMP mice exhibit a severe age-related decline in learning and memory relative to control mice. Therefore, within the cohort studied for phenotype, RNA profiling, DNA sequencing and in situ hybridization, the animals showed a consistent phenotype within the colony.
Neurological gene-expression profile of aging is unique among strains
The anatomical and behavioral analyses of the SAM strains are consistent with age-related deficits in hippocampal-mediated processes that are accelerated in S8 and S10. In addition, studies of the retina also suggest a retinal degeneration phenotype specific to S8 mice that may mimic age-related declines in retinal function in humans [29]. The hippocampus and retina of old (16 month-old S8, S10, SR and 21 month-old B6J) and young (3 month-old) mice were subjected to gene-expression analysis studies using Affymetrix oligonucleotide microarrays (see Lipshutz et al. [31], Sandberg et al. [34], Caceres et al. [35], and Materials and methods for details). As B6J mice exhibit a longer life-span than SR1 mice, we sacrificed B6J mice at 21 months of age, approximately at the same 95% survival point as seen for 16 month-old SR1 mice (see Figure 1a). To ensure that the analysis method used minimized false positives and maximized reliability of the results, the number of independent replicate samples needed was determined by the variation inherent in the samples (see Additional data file 3). In the case of younger animals, two independent samples for each time point and tissue were required. For the 16 month-old S8, S10 and SR hippocampus samples, three samples were used. For retinal samples, four retinas from two mice were pooled to obtain sufficient RNA for each sample. Reproducibility was measured using the Pearson correlation coefficient based on the signal intensities of all genes on the array between replicate samples (perfect correlation = 1.0), and the average correlation coefficients of the replicates in each condition were as follows: young hippocampus, 0.9914; young retina, 0.9946; old retina, 0.9949; old B6J hippocampus, 0.9922, and old S8, S10 and SR hippocampus, 0.9695 (see Additional data file 3 for all replicate correlation coefficients). Representative correlation plots are shown in Additional data file 5. Whereas most of the replicates demonstrated a high reproducibility (> 0.99 correlation coefficient), there was greater variability seen in the old S8, S10 and SR hippocampus replicates (as indicated by lower correlation coefficients). As a result, in these cases additional samples were prepared and analyzed. The number of genes differentially expressed between replicates was used as an estimation of the false positive rate. In all cases very few genes were identified as differentially expressed between replicates, indicating a very low expected false-positive rate in the experimental analyses (Additional data file 3). All data and analysis tools used in this publication are available at [36].
Several types of analyses were performed. First, we characterized gene-expression profiles of aging within each strain. Pairwise comparisons between each tissue sample for young and old mice of the same strain were performed. A given mRNA transcript was considered differentially expressed in a comparison of any two samples if it met the following criteria: a Wilcoxon signed rank test (relative) (WSRR) p-value of p ≤ 0.01 and increase fraction ≥ 0.7; or p ≤ 0.0316 and increase fraction ≥ 0.8; or p ≤ 0.01 and increase fraction ≤ 0.3; or p ≤ 0.0316 and increase fraction ≤ 0.2. A fold change of 1.5 or greater and an average difference change in signal of 30 or more was also required. A gene was considered differentially expressed between conditions (that is, old S8 retina versus young S8 retina) only if it met the above criteria in more than 70% of the pairwise comparisons (3/4 or 4/4 comparisons), and carried a statistically significant absolute call of 'Present' (P) or 'Marginal' (M) in at least one sample (see Materials and methods for more detail).
Subsequently, those genes found to be differentially expressed by the strict criteria described above were examined in all other strains and were considered differentially expressed in another strain if the expression change during aging was significant to an average (WSRR) p-value = 0.05. Finally, the genes that were differentially expressed during aging in each strain were clustered into heat-map views based on their expression patterns, allowing us to examine similarities and differences in transcriptional aging between strains (Figure 2).
Unexpectedly, each strain showed a remarkably unique profile of aging. In the aging hippocampus, only a single gene out of a total 115 (complement component 4 (C4)) changed with age in all four strains (Figure 2a). Seven genes increased in B6J and at least one SAM strain hippocampus. Finally, two genes were downregulated with age in all three SAM strains, but did not change in B6J. The vast majority of changes during aging (75/115 or 65%) were unique to one of the four strains. Interestingly, the genetic background of the animals played an important part in the similarity of the profiles, as related SAM strains exhibited patterns of gene-expression change more similar to one another than they did to B6J, in spite of the fact that SR1 and B6J both demonstrate a 'normal' phenotype, lacking the accelerated neurological pathology seen in S8 and S10 mice.
To determine if these observations extended to other central nervous system (CNS) tissues, similar analyses of the retina were performed (Figure 2b). As seen for the hippocampal data, only a single gene (AI845165, similar to the phosphatidylserine decarboxylase gene) changed with age in all SAM strains and B6J. Also like the hippocampus, the majority (30/46 or 65%) of gene expression changes in the retina were unique to a single strain, again indicating that neuronal tissues of different strains can exhibit dramatically different transcriptional responses to aging.
Strain differences in gene expression
The analysis of the aging retina and hippocampus demonstrated that interesting and specific transcriptional events occurred within the hippocampus and retina of each strain with age. The results suggest that differences in expression levels of important genes between the senescence-prone and -resistant strains could play an important role in mediating the age-related differences observed between these strains. One hypothesis suggests that differences in expression levels of important genes between the senescence-prone and -resistant strains could play an important role in mediating the age-related differences observed between these strains.
To identify such differences, we compared gene expression results for the senescence-accelerated S8 and S10 strains to the closely related, yet disease-free, control SR1 strain at both young and old time points (Figure 3). Gene-expression differences were identified using similar analyses to those described above, but comparing young and old S8 and S10 with SR1 (young 'prone' versus young 'resistant' or old 'prone' versus old 'resistant') (see Materials and methods and Additional data file 1 for more analysis methodology). A total of 124 genes were identified as differentially expressed in the hippocampus between strains in either young or old animals (Figure 3a). A similar analysis of the retina yielded 118 genes that differed between the senescence-prone and -resistant strains (Figure 3b).
The genes differentially expressed between the senescence-prone and -resistant strains could be early markers for aging, or may establish the foundation for accelerated senescence in the S8 and S10 strains. Several of these genes fell into interesting Gene Ontology (GO) categories, as determined by gene enrichment analysis using the GO Tree Machine [37,38]. In the hippocampus, these included genes involved in learning and behavior (phosphodiesterase 1B; Ca2+-calmodulin dependent, protein kinase C-gamma; and preproenkephalin 1), and genes involved in the heat-shock response (heat-shock 70 kD protein 5 (glucose-regulated protein); heat-shock protein 1B; and heat-shock protein 2). In the retina, genes fell into categories involved in the perception of light (ATP-binding cassette, subfamily A (ABC1), member 4; prominin 1 phosphodiesterase 6A, cGMP-specific, rod, alpha; and retinal G-protein-coupled receptor), chloride transport (chloride channel 4-2; gamma-aminobutyric acid (GABA-A) receptor, subunit beta 3; solute carrier family 12, member 2; and chloride intracellular channel 4 (mitochondrial)) and lipid metabolism (ATP-binding cassette, subfamily A (ABC1), member 1; ATP-binding cassette, subfamily A (ABC1), member 4; glycerol kinase; peroxisome proliferator activated receptor alpha; prostaglandin D2 synthase (brain); retinol binding protein 1, cellular; sterol-C5-desaturase (fungal ERG3, delta-5-desaturase) homolog; and sterol-C4-methyl oxidase-like).
To establish the real-world performance of both the analytical methodologies and experimental procedures used to identify gene expression changes, ten genes were chosen from the hippocampal analysis for quantitative reverse transcription PCR (qRT-PCR) verification using independent samples (from mice not used in the microarray analysis) (indicated with ‡ in Figure 3). Of the ten genes assayed, the expression changes for eight genes were confirmed with a change of 1.3-fold or greater: intracisternal-A particles (Iap), upstream transcription factor 1 (Usf1), potassium voltage-gated channel, subfamily Q, member 2 (Kcnq2), chemokine (C-C motif) ligand 19 (Ccl19), erythroid differentiation regulator (edr), caspase 9 (Casp9), chemokine (C-C motif) ligand 27 (Ccl27), complement and component 4 (C4). Of the remainder, one showed a similar trend (1.2-fold change, chromobox homolog 3 (Drosophila HP1 gamma) (Cbx3)) and one (ATP-binding cassette, subfamily D (ALD), member 3 (Abcd3)), showed no gene-expression change, and thus represents a possible false positive.
To cross-validate the gene-expression results, a method besides qRT-PCR was used to examine the expression levels of two genes identified as differentially expressed. In situ hybridization was performed on S8, S10, and SR1 mice for the regulator of G-protein signaling 5 (Rgs5) and Iap. Gene-expression profiling showed that Rgs5 was more highly expressed in the hippocampus and retina of S8 than SR1. In situ hybridization confirmed that Rgs5 was more abundant in the hippocampus and retina of S8 mice than SR1, and also revealed an increased level of Rgs5 in the S8 cerebellum (Figure 4a). The other transcript studied in this manner, Iap had a two- to five-fold higher expression level in both S8 and S10 relative to SR1 mice in the hippocampus and retina. In situ hybridization clearly showed an increased signal in S8 and S10 hippocampus and retina relative to SR1. In contrast to Rgs5, no difference was seen in the cerebellum (Figure 4b). For both transcripts, the in situ hybridization results were correlated with the microarray analyses, indicating a high degree of confidence in those results. Additionally, the pattern of expression observed in the in situ experiments suggests that the higher signal resulted from increased transcript levels in cells that normally express the gene, not ectopic expression in unusual cell types.
Using both qRT-PCR and in situ hybridization, 10/11 genes identified as differentially expressed using Affymetrix oligonucleotide microarrays were confirmed using independent samples and independent methods.
A cluster of genes on chromosome 4, including Ccl19, is differentially expressed between S8 and SR1 mice
Because the gene-expression differences between the accelerated-senescence prone and resistant strains are consistent in multiple independent animals, we sought to identify genetic differences between the strains that might mediate these expression level differences. To identify patterns of expression related to gene position, we looked for correlations between gene location and expression difference. This analysis revealed an interesting region on chromosome 4 of S8 mice harboring multiple genes that were more highly expressed in S8 than SR1 (Figure 5). The RNA levels for six genes in retina or hippocampus were higher in S8 than SR1, representing 21% (6/29) of the S8-SR1 specific hippocampal gene expression differences and 26% (5/19) of the retinal differences. The identified genes were Ccl19, Ccl27, dynactin 3 (Dctn3), opioid receptor, sigma 1 (Oprs1), galactose-1-phosphate uridyl transferase (Galt), and 2810432D09Rik, all of which are within less than 100 kb of each other on chromosome 4 (based on the Celera mouse genome database). Ccl19 was not formally localized to this region but has been shown to be located near Ccl27 [39,40]. These genes are more highly expressed by a factor of 1.7 to 7.4 in S8 relative to SR1 mice. The physical clustering of differentially expressed genes may indicate involvement of a large-scale chromosomal regulatory mechanism.
To investigate this cluster of differentially expressed genes, we pursued one of them in more detail: Ccl19. Our gene-expression studies indicated an increased level of mRNA in the S8 hippocampus relative to SR1. To examine Ccl19 expression in the SAM strains more closely, northern analysis was performed on spleen and hippocampus RNA (Figure 6a). Whereas a consistent band was detected in the spleen of all three SAM strains (Figure 6a, lower band), a band was detected only in the hippocampus of S8 mice (consistent with the gene-expression data). Interestingly, the transcript found in the S8 hippocampus was larger than that seen in the spleen, suggesting either tissue-specific alternative splicing from a single gene locus or the presence of a novel expressed locus.
Motivated by reports that some strains of mice can harbor nearby pseudogenes of Ccl19 [39,40] and the different size of the transcript identified in the S8 hippocampus, sequence analysis was performed on Ccl19 cDNA from both hippocampus and spleen of S8, S10 and SR1 mice. While no bands were identified by northern analysis in the hippocampus of S10 or SR1, fragments were obtained from these tissues using the more sensitive method of reverse-transcriptase PCR. Sequencing of fragments amplified from cDNA revealed an altered coding sequence for the S8 hippocampus transcript relative to all other transcripts, including that found in S8 spleen. The predicted amino-acid sequence of the transcript unique to the S8 hippocampus had two mutations relative to the canonical Ccl19 sequence: a point mutation eliminating the canonical start ATG of Ccl19 and a substitution mutation resulting in a novel methionine 47 residues further downstream in the S8 hippocampus transcript (Figure 6b). Interestingly, while we found no prior description of the novel methionine, the mutation in the canonical start ATG has been previously described in unexpressed Ccl19 pseudogenes found in other strains of mice [39,40]. It is possible that the novel downstream ATG identified in the Ccl19 transcript from the S8 hippocampus may provide a compensatory in-frame start site allowing expression of a truncated protein. No differences within the coding sequence were identified that could result in the larger transcript observed in the S8 hippocampus, suggesting that the longer mRNA results from additional 5' or 3' untranslated region (UTR) sequence.
Because Ccl19 was one of several up-regulated genes located in proximity to one another and because sequence differences between S8 Ccl19 hippocampus and spleen mRNA suggested expression from at least two distinct genes, we hypothesized that a genomic duplication encompassing Ccl19 and surrounding genes was present within the S8 genome. Indeed, Southern analysis using a Ccl19 probe demonstrated a two-fold increase in signal intensity in S8 mice relative to S10 and SR1, consistent with such a duplication (Figure 6c, d). The Southern analysis and sequence information from S8 mice are consistent with a duplication of a block of genes on chromosome 4, resulting in increased expression.
Fgf1 is mutated in S10 mice
As no large-scale genomic sequencing has been reported for either S10 or SR1, we used an algorithm developed in our laboratory that takes advantage of the fact that Affymetrix GeneChips use a series of oligonucleotides that span up to hundreds of bases of a given gene to detect potential sequence variations between the strains (J.A.G., M.A. Zapala, C.B. and D.J.L., unpublished data, see Materials and methods). These oligonucleotides (called probes) yield distinct patterns of intensity for each gene. Sequence differences can be detected based on differences in the hybridization pattern across the set of probes between samples. We compared the underlying patterns of signal intensity between the SAM strains to identify genes that may harbor sequence differences between strains [35] (J.A.G., M.A. Zapala, C.B.and D.J.L., unpublished data, see Materials and methods). These oligomers (probe pairs, 11-20 per gene) yield distinct patterns of intensity for each gene. The probe pairs are sensitive enough that appropriately positioned single base differences between the probe pair and the detected RNA can significantly change the signal intensity, and thus produce different patterns between slightly different sequences. We compared these underlying patterns of signal intensity between the SAM strains to identify genes harboring candidate sequence differences between strains [35] (J.A.G., M.A. Zapala, C.B. and D.J.L., unpublished data, see Materials and methods). Using a threshold p-value of p < 0.000001 (calculated from a two-tailed Student's t-test (unpaired, equal variance)), 20 transcripts were predicted to harbor sequence differences (possibly including nucleotide substitutions, splice differences and/or deletions/insertions) in S8, 36 genes in S10, and 17 genes in both S8 and S10 relative to SR1 (see Additional data file 4).
We have found previously that genes containing at least two predicted polymorphisms (even in the 3' UTR of a gene) often contain additional sequence variations (data not shown). Therefore, we sequenced the coding regions of several genes containing predicted sequence differences between S10 and SR1 that are also known to be involved in important cellular pathways. This led to the identification of mutations in the fibroblast growth factor 1 (Fgf1) gene in S10 mice. Sequencing of the Fgf1 transcript confirmed the predicted polymorphisms in the 3' UTR in S10 (T-C at base pair 2,190 and C-T at base pair 2,931, reference sequence: AF067197). Of particular importance, further sequencing into the coding region of Fgf1 revealed a 15-nucleotide insertion that alters the coding sequence and is expected to result in a truncated protein lacking approximately 45% of the conserved Fgf1 domain (Figure 7a). To confirm that the mutation affected the protein, western blotting was performed. Western analysis on brain extracts from S8, S10 and SR1 mice using a carboxy-terminal antibody confirmed the absence of normal FGF1 protein in the brain of S10 mice (Figure 7b). As fibroblast growth factors have been linked to specific roles in the central nervous system [41,42], it is tempting to speculate that the mutation in this gene plays a role in the senescence-associated neurodegeneration seen in S10 mice.
Discussion
Animal models for the study of complex diseases that are late-manifesting are difficult to create. This poses a major challenge when attempting to study human diseases that occur during the process of aging. This problem is exaggerated when the disease process may emerge not as a discrete phenomenon but rather as a constellation of processes that cascade and ultimately lead to the complex disease. The rare inbred mouse strains generated by careful breeding and selection known as the senescence-accelerated prone mice afforded us the opportunity to study a complex cascade that ultimately leads to progressive neurological decline during the aging process. Although these animals are difficult to maintain and little is known regarding the underlying genetics that predispose these mice to accelerated senescence, their extraordinary phenotype motivated us to study these mice using transcriptional profiling and targeted polymorphism screening to begin to dissect the pathways that lead to the disease phenotype and to compare these changes with those that occur during the 'normal' neurological aging process in two control strains.
Our studies suggest a role for the identified genes and pathways in the neuropathological phenotypes seen in S8 and S10 mice and, at a minimum, provide a set of likely candidate genes and mutations for further study. Importantly, this work demonstrates that it is possible to use large-scale gene expression profiling to identify genotypic differences between strains and link them to a phenotype. We sought to test whether gene expression profiling combined with follow-up of specific genes could be exploited to identify candidate pathways involved in aging processes. We found that we could identify genes likely to be involved in the aging process in multiple mouse strains, and also apply genomics to the study of inbred mice to proceed beyond the search for general patterns of gene expression toward the identification of specific mutations in the accelerated-aging models.
From genomics to specific genes
The combination of gene-expression profiling and the use of inbred mouse strains revealed the presence of a block of differentially regulated genes on chromosome 4 in S8 mice. Follow-up studies confirmed a duplication event containing at least one gene in the region, Ccl19. Given the important role that inflammatory processes could have in CNS aging, both Ccl19 and the closely linked Ccl27 could play primary roles in the neurological phenotypes seen in the S8 strain. Our results suggest that a large-scale duplication event of a region of chromosome 4 results in increased expression of multiple genes in the brains of S8 mice.
We have also identified a potentially important mutation in the growth factor gene Fgf1 in S10 that results in a nonsense codon in the predicted amino-acid sequence, and western analysis confirmed that normal FGF1 protein is not detectable in the brains of these animals. Interestingly, FGF1 has been shown to protect neurons from excitotoxic stress, and low FGF1 mRNA levels have been implicated in Alzheimer's disease [43-46]. The absence of intact FGF1 in the S10 mice could be critical in the age-related neurodegeneration and brain atrophy observed in the S10 strain.
The opportunistic polymorphism screening approach we pursued led to multiple candidate genes. It is exciting to postulate that future availability of more complete sequence and/or high density SNP information will make possible the combination of genomic information with full genome expression profiling to delineate specific mutations and expression differences that point to a host of intriguing candidates for complex traits.
General aspects of aging
While the majority of genes found to be differentially expressed in the hippocampus and retina were specific to one or two strains, two genes in particular stood out as potentially 'universal' markers for molecular processes involved in aging. In the hippocampus, C4 was upregulated in old S8, S10, SR1 and B6J mice. C4 was also upregulated in the aging neocortex and cerebellum of B6J, suggesting that the complement cascade plays a common role in aging in multiple inbred mouse strains and brain regions [47]. Other components of the complement cascade were also upregulated in the Lee et al. study [47], including C1q A-chain, C1q B-chain, and C1q C-chain. Examination of these genes in our data revealed that, while the changes were not significant enough to meet our most stringent criteria, the C1q A-chain, C1q B-chain transcripts were elevated approximately 1.3-fold in aged S8, S10 and SR1 hippocampus (S8 p-value = 0.05 and 0.08 for C1q A- and B-chains, respectively; S10 p-value = 0.03 and 0.05 for C1q A- and B-chain, respectively; and SR1 p-value = 0.03 and 0.1 for C1q A- and B-chains, respectively). Increases in immune-response genes in the aging brain have been found in several different mouse strains and there is a growing body of evidence supporting the importance of chronic inflammation in mammalian aging [47-49]. Intriguingly, such an immune/inflammatory response may play a particularly important role in the nervous system, as similar increases were not seen in gene expression studies of aging muscle or fibroblasts [50,51]. Other studies in B6J mice also show involvement of inflammatory pathways in the aging retina ([52] and S.Y., unpublished work). One study has also reported increased levels of immune-related transcripts with age in the liver of another strain (C3B10RF1), suggesting that there may be other tissues with age-related immune responses [53]. Importantly, studies of caloric restriction, the only known method of life-extension in mammals, show that restricting food intake results in the downregulation of genes involved in inflammation [47,49].
Another interesting gene identified was the mouse homolog of phosphatidylserine decarboxylase (PSDC), a transcript up-regulated in the aging retina of all strains and the aging hippocampus of S8, SR1 and B6J. This gene encodes an enzyme localized to the inner mitochondrial membrane that is involved in phospholipid biosynthesis. It is possible that the upregulation of PSDC may provide a mechanism to compensate for oxidative damage to membranes in the CNS [54]. Finally, a recent report examining gene regulation and the aging human brain identified many genes in the same pathways that we have identified here, including genes involved in stress response, inflammation and vesicular transport [55].
Strain-specific transcriptional response to aging
In addition to the identification of genes and mutations potentially involved in aging, our results demonstrated some unexpected findings. Surprisingly, we found that different inbred strains of mice demonstrated strikingly distinct aging patterns (Figure 2). Of the many genes differentially expressed, very few were in common between the three SAM strains and B6J. More age-related gene expression changes were found in common between the SAM strains than the B6J, most likely reflecting the close genetic relationship between the AKR/J-derived SAM strains, but even within the SAM mice, there were significant differences between the strains in the patterns of expression changes. This suggests that the use of any single inbred mouse strain to study transcriptional events associated with mammalian aging may be misleading or incomplete and that it is best to use multiple strains in order to model general mammalian aging processes.
The two strains demonstrating accelerated aging, S8 and S10, shared many differences relative to SR1, including increased expression levels of stress-response genes such as those for the heat-shock proteins Hsp70-1 and Hspa5, and for X-box binding protein 1 (Xbp1). The senescence-prone mice also demonstrated differential expression of some genes known to be involved in human diseases: these include transthyretin (Ttr) and Kcnq2, both of which show decreased levels of expression in S8 and S10.
Another interesting gene-expression difference between the strains was the increase in Iap mRNA levels observed in S8 and S10 hippocampus and retina relative to SR1. Iap sequences are repetitive transposable elements present in approximately 1,000 copies per mouse genome, and transcript levels have been reported to increase with age in the liver of some strains [56-58]. This increase has been shown to be associated with demethylation of an Iap promoter, suggesting a failure of repression mechanisms to control retrotransposon expression in the aging mouse. In the case of S8 and S10, Iap expression was elevated two- to five-fold in young mice relative to SR1, and remained elevated throughout the life span of the mouse.
Interestingly, a global evaluation of transcriptional changes also revealed a difference in expression trends between S8 and the other strains. In the aging hippocampus, 61% (40/66) of differentially expressed genes were upregulated in SR1, 69% (29/42) were upregulated in S10 and 82% (14/17) were upregulated in B6J. These values are consistent with trends seen in previous studies of other regions of the B6J brain [50]. In contrast, only 46% (18/39) of genes were upregulated with age in the S8 hippocampus. These same trends were seen in the aging retina, with S10, SR1 and B6J showing upregulation of at least 80% of differentially expressed genes (22/24, 15/16 and 8/10, respectively). The aging S8 retina showed only 24% (5/21) of genes up-regulated. This general difference in age-related gene expression levels supports a global difference in transcriptional regulation in S8 neuronal tissues. Therefore, while the neurological phenotypes of S8 and S10 mice are somewhat similar, S8 mice may have a very abnormal transcriptional response to the aging process. One hypothesis suggested by the data is that certain events during aging trigger the upregulation of important genes. This response occurs in the SR1, S10 and B6J mice, but is abnormal in S8 mice. Thus, it is possible that S10 mice exhibit an acceleration of certain normal molecular responses to aging, while S8 mice exhibit a malfunction of normal transcriptional responses.
Conclusion
Aging is a complex process involving multiple tissues and events. Many genes and pathways are implicated from our gene-expression data, with some very interesting candidates implicated in the pathology of S8 and S10 mice. Elucidation of the precise roles these candidate genes play remains difficult, but the simple identification of such candidates provides the opportunity for hypothesis formation and testing to further characterize their involvement in the aging process. Our studies suggest that careful analysis of multiple strains, multiple tissues, and the integration of gene expression data generated from multiple laboratories will be important for deciphering the molecular biology of the aging process in mammals. Ultimately, combining the knowledge gained from whole-genome sequencing of multiple strains with gene-expression analyses and careful phenotypic comparison is likely to provide great insight into the aging process.
Materials and methods
Animal handling
All animal procedures were performed according to protocols approved by The Salk Institute for Biological Studies Animal Care and Use Committee. C57BL/6J mice were obtained from Jackson Laboratories and were housed for one week before dissection. The senescence-accelerated mouse (SAM) strains were developed at Kyoto University from a colony of AKR/J mice that were selectively bred based on senescence, life span and pathologic phenotypes and were generously provided to our laboratory [18,20,59]. Founder mating pairs of S8/Ta (F-105), S10//Ta (F-99) and SR1TA (F-99) mice raised under specific-pathogen-free conditions were obtained from Kyoto University as approved by the SAM council. Animals were maintained in a specific-pathogen-free environment under standard 12-h light-dark cycles (lights on 6:00 AM - 6:00 PM) with food and water provided ad libitum. Fertility of the mice was monitored by plug-checks during matings, and fecundity was determined by recording the number of offspring. Necropsies were performed on mice that were used for dissection, euthanized, or found dead. All mice that died of natural causes were used in the pathology studies. 237 S8 mice, 169 S10 mice, and 189 SR1 mice were used for the life span analysis. Cause of death was determined by necropsy and classified as 'cancer,' 'infection,' or 'no cancer or infection.'
Single-trial passive avoidance
Behavioral analyses were performed using a single trial in a two-compartment step-through passive avoidance apparatus [23]. During acquisition, a 0.5 mA current was applied to a floor grid for 3 sec upon animal entry into a darkened chamber. Retention was measured 24 h later as time to enter the dark chamber, up to a maximum of 300 sec. Male and female mice were tested at two time points (young time point: average age of 16 weeks; old time point: average age of 81 weeks).
Tissue collection
Mice were sacrificed between 15:00 and 17:00 by cervical dislocation. Retinal samples were dissected first and frozen on dry ice. The brain was then removed and the hippocampus rapidly excised, followed by the removal of the spleen from the body. All samples were rapidly frozen on dry ice and stored at -80°C. Total RNA was prepared using Trizol Reagent (Invitrogen) according to the manufacturer's recommended protocol [34].
Gene-expression profiling
Gene-expression profiling was performed on 3 month-old (young), 16 month-old (old) S8, S10 and SR1 mice, and 3 month-old and 21 month-old B6J mice. Two independent samples for each time point were used in gene-expression profiling for each strain. Because of greater replicate variability, three samples were used for hippocampus of 16-month SAM mice. A 10.0 μg sample of total RNA was used to generate labeled cRNA for each sample according to recommended protocols (Affymetrix). RNA from multiple animals was not pooled, except in the case of retina, where the retinas of two mice were combined to generate sufficient total RNA. Hybridizations were performed on MG_U74Av2 Affymetrix GeneChips for 16 h at 50°C at a final cRNA concentration of 0.66 μg/μl. Data were analyzed using the TeraGenomics relational database [36]. To identify differentially expressed genes between any two samples, a Wilcoxon signed-rank test (WSRR) p-value was calculated using the probe-pair difference values for a given gene between the two samples. A combined set of criteria, WSRR p-value of p ≤ 0.01 and increase fraction ≥ 0.7, or p ≤ 0.01 and increase fraction ≤ 0.3, was used to detect genes that showed an increase or decrease in expression level. To detect less robust but still potentially interesting gene expression changes, a less stringent set of criteria, p ≤ 0.0316 and increase fraction ≥ 0.8, or p ≤ 0.0316 and increase fraction ≤ 0.2, was used to identify genes as showing a 'marginal' increase or decrease. Additionally, a fold change of 1.5 or greater and a change in signal ≥ 30 was also required. The above analysis was performed on all possible pairwise comparisons of the samples used (that is, two independent young S8 retinal samples and two independent old S8 retinal samples resulted in four pairwise comparisons). A gene was considered differentially expressed (for example, old S8 retina vs young S8 retina) only if it met the above criteria in at least 70% of the pairwise comparisons (3/4 or 4/4 comparisons). Finally, the transcript had to meet the following criteria resulting in an absolute call of P (present, indicating the transcript reached reliably detectable levels) in at least one sample: p ≤ 0.0316 and positive fraction ≥ 0.6, or p ≤ 0.1 and positive fraction ≥ 0.75.
For each analysis (old vs young or strain 1 vs strain 2), all possible pairwise comparisons were generated (that is, four files resulting from a comparison of two S8 hippocampus samples to two SR, or six files resulting from a comparison of three old S10 hippocampus samples to two young S10 hippocampus samples) and used in the analyses. This analysis methodology has been described in more detail in previous publications [34,61].
The data and analysis tools are freely available at [36].
Polymorphism prediction
Candidate genes harboring predicted polymorphisms were identified using an algorithm developed by our laboratory (J.A.G., M.A. Zapala, C.B. and D.J.L., unpublished work). Briefly, the algorithm works as follows: first, for the selected probe sets, the individual hybridization intensity values are extracted and the difference between the perfect match and the mismatch (PM-MM) intensities is calculated for each probe pair for each sample, excluding probe sets from samples that do not meet certain pattern quality measures. The PM-MM values for each of the probe sets for each sample are globally scaled (by a factor derived from the standard deviation across the multi-probe pattern obtained in each experiment) to compensate for gene-expression differences. Next, the scaled values for each sample group are averaged, and an average and a standard deviation are calculated for each probe pair in a probe set. A threshold, an empirical measure of significance, was computed for each probe pair (PP) as follows:
The algorithm was written in structured query language (SQL) using Queryman, a Teradata specific compiler [61] to run on the Teradata relational database.
Real-time quantitative RT-PCR
RNA samples from two to three mice for each strain at each time-point were used to verify the gene-expression differences. RNA used for quantitative PCR (Q-PCR) was independent from the RNA used in the Affymetrix microarray experiments. Standard protocols were used for the generation of cDNA from RNA following elimination of genomic DNA contamination using DNA-free (Ambion). Oligonucleotide primers were designed using Applied Biosystems Primer Express software v. 1.5. ABI's SYBR Green PCR Master Mix was used for the Q-PCR reactions, which were then run on the ABI Prism 7700 sequence detection system. All Q-PCR data analysis was normalized to peptidylprolyl isomerase B (cyclophilin) levels as an internal control.
In situ hybridization
Templates for probes were synthesized by PCR, subcloned into TOPO-2 (Invitrogen), linearized with EcoRV, and transcribed in vitro using digoxigenin (DIG)-labeled uridine triphosphate (Roche Biomedical Systems) according to the manufacturer's protocols. For hippocampal and cerebellar tissue preparation, mice were sacrificed by rapid cervical dislocation, the brains removed and embedded in OCT on dry ice, and sliced in 10 μm sections using a cryostat. For retinal tissue preparation, animals were perfused using 4% paraformaldehyde. The eyes were then enucleated, fixed in 4% paraformaldehyde for 24 h and embedded in paraffin and sliced. Retinal sections were rehydrated and treated with proteinase K. Subsequently, both brain and retinal sections were refixed in 4% paraformaldehyde for 10 min, washed in PBS and acetylated. Hybridization was performed overnight at 60°C using a probe concentration of 0.5-1 μg/ml, and detection was performed with anti-digoxigenin alkaline phosphatase-conjugated Fab fragments (Roche Biomedical Systems) followed by staining with NBT and BCIP (Roche Biomedical Systems). Images were collected with a Nikon microscope attached to a charge-coupled device camera and software (Media Cybernetics), using electronically acquired composite images.
Polymorphism identification
Total RNA was extracted from hippocampus and spleen using methods described above and contaminating genomic DNA was removed using Ambion's DNA-free according to the manufacturer's recommended procedures. Primers for the amplification of regions containing putative sequence differences were designed (length = 17-26 bp; Tm = 55-60°C). Whenever possible, primers were designed to include the complete Affymetrix target sequence. All sequencing was performed by the Salk Institute Sequencing Core.
Northern analysis
Total RNA (10.0 μg for each sample) was incubated at 50°C for 1 h in the presence of glyoxal, and electrophoresed in 1× DEPC-treated BPTE buffer in a 1% agarose gel. The gel was blotted onto Hybond-N+ membrane overnight. A fragment of the Ccl19 mRNA was amplified using RT-PCR from hippocampus total RNA. Random primed labeling was used to radioactively label the fragment with 32P and hybridization to the membrane was performed overnight in Church buffer at 65°C. Four washes of 15-20 min were performed to a maximal stringency of 0.1× SSC, 0.1% SDS at 65°C. Visualization was performed on a Molecular Dynamics Phosphorimaging system.
Southern analysis
Genomic DNA (10.0 μg for each sample) was digested at37°C using BamHI or EcoRI. The digested samples were electrophoresed in 1× TAE buffer on a 0.8% agarose gel overnight at 30 V. The gel was then washed in 0.2 M HCl for 30 min., rinsed quickly in distilled water twice, placed in denaturing buffer (1.5 M NaCl, 0.5 M NaOH) for 30 min, neutralization buffer (1.5 M NaCl, 0.5 M Tris pH 7.5) for 30 min, and 20× SSC for 2 min. The gel was then blotted onto a Nytran supercharge filter (Schleicher & Schuell) in 20× SSC overnight and UV cross-linked. To create the Ccl19 probe, a fragment of the Ccl19 mRNA was amplified using RT-PCR from SR cortex total RNA (primers: GCGGGCTCACTGG-GGCACAC, TGGGAAGGTCCAGAGAACCAG). Random primed labeling was used to radioactively label the fragment with 32P and hybridization to the membrane was performed overnight in Church buffer at 65°C. Four washes of 15-20 min were performed to a maximal stringency of 0.1× SSC, 0.1% SDS at 55°C. Visualization was performed on a Molecular Dynamics Phosphorimaging screen. A SacI-EcoRI 795-bp fragment of Grik1 was subsequently used as a control probe following the procedures described for Ccl19.
Immunoprecipitation and western blot
Cortices for two mice of each strain were homogenized in NP-40 lysis Buffer (20 mM Tris-HCl pH 8.0, 137 mM NaCl, 10% glycerol, 1% Nonidet P-40, 1 mM EDTA, protease inhibitors (aprotinin + leupeptin), 1 mM phenyl methyl sulfonyl fluoride (PMSF)) at 4°C. The samples were then centrifuged and the supernatant was removed and stored. The protein supernatant (1 mg) was pre-cleared with Protein-G Sepharose slurry for 30 min, then removed and incubated with the Protein-G Sepharose for 2 h with the anti-FGF1 antibody raised against the carboxy terminus of human FGF1 (Santa Cruz Biotechnology, 1/100 concentration). The Protein-G Sepharose beads were then removed, washed, and the protein was liberated upon incubation with SDS buffer at 70°C for 10 min. The supernatant was then run out on a Novex 12% Bis-Tris gel in MOPS buffer, blotted, and visualized with a secondary horseradish peroxidase-conjugate antibody and detected using the ECL Plus kit (Amersham Biosciences).
Additional data files
The following additional data are available with the online version of this paper. Additional data file 1 contains four sections explaining the animal handling and gene expression methodologies. Additional data file 5 contains four correlation plots of replicate samples. Additional data file 2 contains a summary of all mice used for breeding. Additional data file 3 contains a summary of the correlations between all replicate samples and the empirically determined estimated false-positive rates. Additional data file 4 contains a list of the genes predicted to harbor polymorphisms between the S8, S10 and SR1 strains.
Supplementary Material
Additional data file 1
Animal handling and gene expression methodologies.
Click here for file
Additional data file 2
A table containing a summary of all mice used for breeding
Click here for file
Additional data file 3
A table containing a summary of the correlations between all replicate samples and the empirically determined estimated false positive rates.
Click here for file
Additional data file 4
A table listing the genes predicted to harbor polymorphisms between the S8, S10 and SR1 strains.
Click here for file
Additional data file 5
Correlation plots for independent replicate measurements. Examples shown are representative of the gene expression data quality obtained throughout the study. The signal for every gene on the microarray is graphed on a scatter plot for two replicate samples (that is, samples of the same strain, same tissue and same age but different animals). The Pearson correlation coefficient (R) is given for each comparison, ranging from 0.9752 to 0.9958. (A) Correlation plot for S8: 609-4HpOld and S8: 609-3HpOld showing the high degree of correlation between the two independent replicates. (B) Correlation plot for one of the same samples as in (A) (S8: 609-3HpOld) and a third replicate sample (S8: 674-2HpOld) showing a larger variability between replicates. (C) Correlation plot for S10: 704-3+704-4 Retina Old and S10: 15-4+705-2 Retina Old. (D) Correlation plot for SR1: 1349-1+1349-2 Retina Young and SR1: 1348-1+1348-2.
Click here for file
Acknowledgements
We would like to thank Dan Lockhart and Matt Zapala for their work on software tools for data analysis; Information Management Consultants, Inc. (IMC, Reston, VA) for donation and programming of the Teradata database; Ling Ouyang for cDNA array hybridizations not described herein, Matt Hemming and Mark Latronica for assistance with behavioral analyses, Thomas D. Pugh for lifespan data on B6 mice, T.K. Booker for providing the Grik1 probe, and Jo A. Del Rio and Richard Tennant for additional assistance. We would also like to thank Patrick Zarrinkar and Lisa Wodicka for assistance and guidance in Affymetrix GeneChip experimentation and data analysis. T.A.C. was supported by the J. Aaron Charitable Foundation and NIH training grant CA09370. J.A.G was supported by the National Defense Science and Engineering Graduate Fellowship, a generous gift from the Lewin family and the Sprint Corporation, a grant from the Legler-Benbough Foundation. Additional funding was provided by DOD grant DAMD17-99-1-9561, the NIH grant 5-RO1_NS039601-04 and the Frederick B. Rentschler Developmental Chair to CB. SY and AS were supported by grants from Macula Vision Research Foundation (West Conshohocken, PA), Elmer and Sylvia Sramek Charitable Foundation (Chicago, IL), and National Institutes of Health (EY11115 administrative supplements).
Figures and Tables
Figure 1 S8 and S10 mice exhibit accelerated-senescence phenotypes. (a) Proportion of surviving S8 mice (blue squares, n = 237), S10 mice (purple diamonds, n = 169), SR1 mice (green triangles, n = 189) and B6 mice (black crosses, n = 75, data from Pugh et al. [32]). Using a Kaplan-Meier survival analysis, the S8 and S10 survival profiles are significantly different from SR1 (p < 0.0001 and p < 0.0015, respectively) by the Mantel-Cox log rank test. Arrows and lines indicate the two ages (3 months and 16 months) at which mice were dissected and used for the gene expression experiments in the SAM strains. An arrow also indicates the age at which the old B6J mice were sacrificed for gene expression analysis. (b) Retention of foot shock is indicated as latency to entry into the dark chamber (day 2-day 1) in a passive avoidance paradigm. Difference in latency to entry is shown on the y-axis and age on the x-axis at a young time point and an old time point for S8 (blue: young n = 43, old n = 8), S10 (purple: young n = 31, old n = 7) and SR1 (green: young n = 41, old n = 27) mice. Error bars indicate standard error. S8 and S10 mice show decreased latency to entry relative to SR1 at both young (two-tailed Student's t-test: p < 0.003 and p < 0.002, respectively) and old (p < 0.05, S8) time points.
Figure 2 Strain-specific aging gene-expression profiles. (a) Heat-map view of differentially expressed genes between 3- and 16-month-old S8, S10, SR1 and B6J hippocampus. Fold change of old relative to young hippocampus is numerically indicated within each box. Bright red (increase in expression level) and bright green (decrease in expression level) indicate genes that met the most stringent criteria for an expression-level change, including an average WSRR p < 0.01. A gene is colored orange or light green if its expression increased or decreased with an average WSRR p < 0.05. Box 1, genes changed in common with three or more strains; box 2, genes changed in only two strains; box 3, genes with changes unique to a single strain; and box 4, genes with inconsistent patterns of change. (b) Heat-map view of differentially expressed genes between 3-month-old and 16-month-old S8, S10, SR1 and B6J retinas. The analysis was performed as described for the hippocampus.
Figure 3 Inter-strain, age-independent gene-expression changes. (a) A heat-map view of genes differentially expressed in hippocampus with either S8 compared to SR1, or S10 compared to SR1 is shown. Columns indicate fold change in gene expression between young S8 and SR1, young S10 and SR1, old S8 and SR1 and old S10 and SR1, respectively. Bright red and bright green indicate genes that met the most stringent criteria for an increased or decreased expression level change, including a WSRR p < 0.01. A gene is colored orange or light green if it was increased or decreased with an average WSRR p < 0.05 in all comparison files. Boxes, numbers and examples of genes with the pattern of change are as follows: group 1, genes differentially expressed between both young and old S8 and S10 mice compared to SR1, example gene is melanoma antigen, 80 kDa; group 2, genes differentially expressed in 3/4 comparisons of young and old S8 and S10 compared to SR1, example gene is decorin; group 3, genes differentially expressed in either young and old S8 versus SR1 or young and old S10 versus SR1, example gene is heat shock 70 kD protein 5; group 4, genes differentially expressed in any single comparison group of S8 or S10 to SR1, example gene is protein phosphatase 1-1A (differences unique to a single age and strain); group 5, genes with inconsistent expression differences, example genes are RAN binding protein 9, glutathione S-transferase a4 and kinesin family member 5b. Genes differentially expressed in both retina and hippocampus are indicated by an asterisk. Gene-expression differences confirmed by qRT-PCR are indicated by the symbol ‡. Line graphs in the last column represent the signal intensity on the y-axis and the time point of collection on the x-axis. S8 is indicated by a red line, S10 by a blue line, SR1 by a green line and B6J by a black line. (b) Heat-map view and line graphs of exemplars from each group representing differentially expressed genes in the retina between S8 versus SR1 and S10 versus SR1 mice is shown. Analysis and representation of the data is similar as described for (a).
Figure 4 In situ hybridization of Rgs5 and Iap. (a) In situ hybridization of an Rgs5 probe to SR and S8 hippocampus (A, D), cerebellum (B, E) and retina (C, F) is shown. Increased levels of expression can be seen in all three tissues of S8 mice relative to SR1 mice. (b) In situ hybridization of an Iap probe to SR, S8 and S10 hippocampus (A, D, G), cerebellum (B, E, H) and retina (C, F, I) are shown. Increased expression levels of Iap can be seen in the hippocampus and retina of S8 and S10 mice relative to SR1 mice.
Figure 5 S8-specific cluster of differentially expressed genes on chromosome 4. The location and relative expression levels of a cluster of differentially expressed genes on chromosome 4 containing a cluster of differentially expressed genes in S8 mice are depicted. FC, fold change, represents the average fold change in S8 relative to SR hippocampus (Hp) and retina (Rt). Those gene-expression changes significant to a p-value ≤ 0.05 (WSRR) are indicated and those genes that were not significantly different are marked as nonsignificant (ns). Chromosomal map position is given in megabases, and both the Affymetrix ID and LocusLink gene symbol are specified. The genes demonstrating upregulation in S8 mice are highlighted in gray.
Figure 6 Ccl19 is abnormally expressed in S8 hippocampus and is duplicated. (a) A northern blot of total RNA from spleen and hippocampus after hybridization with a Ccl19-specific probe is shown. An abnormally large Ccl19 transcript is detected only in the hippocampus of S8 mice, while the spleen of S8, S10 and SR1 all show normal Ccl19 expression. (b) Predicted amino-acid sequence of CCL19 from the hippocampal cDNA sequence of S8 mice reveals a putative truncation at the amino terminus of the protein relative to that of S8 spleen, and SR1 spleen and hippocampus. Arrows indicate the sites of mutations in the protein found in the S8 hippocampus. The region of the protein deleted by the mutations is highlighted in yellow. A box surrounds the first two conserved cysteines, which are adjacent and conserved in all β-chemokines. Colored boxes indicate the location of the signal peptide and the SCY-domain. (c) A Southern blot of S8, S10 and SR1 genomic DNA digested with EcoRI. The upper panel shows the signal from hybridization with a Ccl19 probe, and the lower panel shows the signal of the same blot hybridized with a control probe demonstrating relative DNA loading. The average Ccl19 signal intensity of each S8 lane is 1.9-fold greater than in SR1 and S10 when normalized to the control probe (p < 0.005 with a two-tailed Student's t-test). (d) A Southern blot of genomic DNA digested with BamHI. In this case, the lower panel shows the signal from hybridization with a Ccl19 probe, and the upper panel shows the signal of the same blot hybridized with a control probe demonstrating relative DNA loading. The average Ccl19 signal intensity seen in S8 is 2.6-fold greater than in SR1 and S10 when normalized to the control probe (p < 0.003 with a two-tailed Student's t-test).
Figure 7 Absence of wild-type FGF1 in S10 cortex. (a) Predicted amino-acid sequence of FGF1 protein from SR1 and S10 hippocampus reveals a putative truncation of the S10 protein sequence at the carboxy terminus. The reference FGF1 protein sequence is shown at the top. The site of the 15-bp insertion is indicated by an arrow and the resulting amino-acid changes and the truncation predicted in S10 FGF1 are highlighted in yellow. The nuclear localization signal (NLS), the FGF core domain and the heparin-binding motif are indicated by boxes. (b) A western blot of S8, S10 and SR1 cortices probed with an antibody specific for the carboxy terminus of FGF1 is shown. This region is predicted to be deleted in S10 mice. Wild-type mouse FGF1 is 155 amino acids and 17.4 kD. As predicted, wild-type FGF1 is detected in the cortices of S8 and SR1, but not S10, consistent with the predicted truncation of the protein in that strain of mice.
==== Refs
Johnson FB Sinclair DA Guarente L Molecular biology of aging. Cell 1999 96 291 302 9988222 10.1016/S0092-8674(00)80567-X
Smeal T Claus J Kennedy B Cole F Guarente L Loss of transcriptional silencing causes sterility in old mother cells of S. cerevisiae. Cell 1996 84 633 642 8598049 10.1016/S0092-8674(00)81038-7
Kaeberlein M McVey M Guarente L The SIR2/3/4 complex and SIR2 alone promote longevity in Saccharomyces cerevisiae by two different mechanisms. Genes Dev 1999 13 2570 2580 10521401 10.1101/gad.13.19.2570
Tissenbaum HA Guarente L Increased dosage of a sir-2 gene extends life span in Caenorhabditis elegans. Nature 2001 410 227 230 11242085 10.1038/35065638
Strehler BL Genetic instability as the primary cause of human aging. Exp Gerontol 1986 21 283 319 3545872 10.1016/0531-5565(86)90038-0
Martin GM Oshima J Lessons from human progeroid syndromes. Nature 2000 408 263 266 11089984 10.1038/35041705
Bodnar AG Ouellette M Frolkis M Holt SE Chiu CP Morin GB Harley CB Shay JW Lichtsteiner S Wright WE Extension of life-span by introduction of telomerase into normal human cells. Science 1998 279 349 352 9454332 10.1126/science.279.5349.349
Delgado Luengo W Rojas Martinez A Ortiz Lopez R Martinez Basalo C Rojas-Atencio A Quintero M Borjas L Morales-Machin A Gonzalez Ferrer S Pineda Bernal L Del(1)(q23) in a patient with Hutchinson-Gilford progeria. Am J Med Genet 2002 113 298 301 12439901 10.1002/ajmg.10753
Yu CE Oshima J Fu YH Wijsman EM Hisama F Alisch R Matthews S Nakura J Miki T Ouais S Positional cloning of the Werner's syndrome gene. Science 1996 272 258 262 8602509
Rapin I Lindenbaum Y Dickson DW Kraemer KH Robbins JH Cockayne syndrome and xeroderma pigmentosum. Neurology 2000 55 1442 1449 11185579
Eriksson M Brown WT Gordon LB Glynn MW Singer J Scott L Erdos MR Robbins CM Moses TY Berglund P Recurrent de novo point mutations in lamin A cause Hutchinson-Gilford progeria syndrome. Nature 2003 423 293 298 12714972 10.1038/nature01629
Mounkes LC Kozlov S Hernandez L Sullivan T Stewart CL A progeroid syndrome in mice is caused by defects in A-type lamins. Nature 2003 423 298 301 12748643 10.1038/nature01631
Barlow C Eckhaus MA Schaffer AA Wynshaw-Boris A Atm haploinsufficiency results in increased sensitivity to sublethal doses of ionizing radiation in mice. Nat Genet 1999 21 359 360 10192382 10.1038/7684
Hande MP Balajee AS Tchirkov A Wynshaw-Boris A Lansdorp PM Extra-chromosomal telomeric DNA in cells from Atm(-/-) mice and patients with ataxia-telangiectasia. Hum Mol Genet 2001 10 519 528 11181576 10.1093/hmg/10.5.519
Gaymes TJ North PS Brady N Hickson ID Mufti GJ Rassool FV Increased error-prone nonhomologous DNA end-joining - a proposed mechanism of chromosomal instability in Bloom's syndrome. Oncogene 2002 21 2525 2533 11971187 10.1038/sj.onc.1205331
Orren DK Theodore S Machwe A The Werner syndrome helicase/exonuclease (WRN) disrupts and degrades D-loops in vitro Biochemistry 2002 41 13483 13488 12427008 10.1021/bi0266986
Brosh RM JrDriscoll HC Dianov GL Sommers JA Biochemical characterization of the WRN-FEN-1 functional interaction. Biochemistry 2002 41 12204 12216 12356323 10.1021/bi026031j
Takeda T Hosokawa M Higuchi K Senescence-accelerated mouse (SAM): a novel murine model of senescence. Exp Gerontol 1997 32 105 109 9088907 10.1016/S0531-5565(96)00036-8
Higuchi K Matsumura A Honma A Toda K Takeshita S Matsushita M Yonezu T Hosokawa M Takeda T Age-related changes of serum apoprotein SASSAM, apoprotein A-I and low-density lipoprotein levels in senescence accelerated mouse (SAM). Mech Ageing Dev 1984 26 311 326 6434885 10.1016/0047-6374(84)90103-9
Takeda T Senescence-accelerated mouse (SAM): a biogerontological resource in aging research. Neurobiol Aging 1999 20 105 110 10537019 10.1016/S0197-4580(99)00008-1
Xia C Higuchi K Shimizu M Matsushita T Kogishi K Wang J Chiba T Festing MF Hosokawa M Genetic typing of the senescence-accelerated mouse (SAM) strains with microsatellite markers. Mamm Genome 1999 10 235 238 10051317 10.1007/s003359900979
Miyamoto M Kiyota Y Yamazaki N Nagaoka A Matsuo T Nagawa Y Takeda T Age-related changes in learning and memory in the senescence-accelerated mouse (SAM). Physiol Behav 1986 38 399 406 3786521 10.1016/0031-9384(86)90112-5
Miyamoto M Characteristics of age-related behavioral changes in senescence-accelerated mouse SAMP8 and SAMP10. Exp Gerontol 1997 32 139 148 9088911 10.1016/S0531-5565(96)00061-7
Shimada A Ohta A Akiguchi I Takeda T Inbred SAM-P/10 as a mouse model of spontaneous, inherited brain atrophy. J Neuropathol Exp Neurol 1992 51 440 450 1619443
Yagi H Katoh S Akiguchi I Takeda T Age-related deterioration of ability of acquisition in memory and learning in senescence accelerated mouse: SAM-P/8 as an animal model of disturbances in recent memory. Brain Res 1988 474 86 93 3214716 10.1016/0006-8993(88)90671-3
Kawamata T Akiguchi I Yagi H Irino M Sugiyama H Akiyama H Shimada A Takemura M Ueno M Kitabayashi T Neuropathological studies on strains of senescence-accelerated mice (SAM) with age-related deficits in learning and memory. Exp Gerontol 1997 32 161 169 9088913 10.1016/S0531-5565(96)00063-0
Miyamoto M Kiyota Y Nishiyama M Nagaoka A Senescence-accelerated mouse (SAM): age-related reduced anxiety-like behavior in the SAM-P/8 strain. Physiol Behav 1992 51 979 985 1615059 10.1016/0031-9384(92)90081-C
Shimada A Hosokawa M Ohta A Akiguchi I Takeda T Localization of atrophy-prone areas in the aging mouse brain: comparison between the brain atrophy model SAM-P/10 and the normal control SAM-R/1. Neuroscience 1994 59 859 869 8058124 10.1016/0306-4522(94)90290-9
Hosokawa M Ueno M Aging of blood-brain barrier and neuronal cells of eye and ear in SAM mice. Neurobiol Aging 1999 20 117 123 10537021 10.1016/S0197-4580(99)00029-9
Hosokawa M Abe T Higuchi K Shimakawa K Omori Y Matsushita T Kogishi K Deguchi E Kishimoto Y Yasuoka K Takeda T Management and design of the maintenance of SAM mouse strains: an animal model for accelerated senescence and age-associated disorders. Exp Gerontol 1997 32 111 116 9088908 10.1016/S0531-5565(96)00078-2
Lipshutz RJ Fodor SP Gingeras TR Lockhart DJ High density synthetic oligonucleotide arrays. Nat Genet 1999 21 20 24 9915496 10.1038/4447
Pugh TD Oberley TD Weindruch R Dietary intervention at middle age: caloric restriction but not dehydroepiandrosterone sulfate increases life span and lifetime cancer incidence in mice. Cancer Res 1999 59 1642 1648 10197641
Gosden RG Laing SC Flurkey K Finch CE Graafian follicle growth and replacement in anovulatory ovaries of ageing C57BL/6J mice. J Reprod Fertil 1983 69 453 462 6631813
Sandberg R Yasuda R Pankratz DG Carter TA Del Rio JA Wodicka L Mayford M Lockhart DJ Barlow C Regional and strain-specific gene expression mapping in the adult mouse brain. Proc Natl Acad Sci USA 2000 97 11038 11043 11005875 10.1073/pnas.97.20.11038
Caceres M Lachuer J Zapala MA Redmond JC Kudo L Geschwind DH Lockhart DJ Preuss TM Barlow C Elevated gene expression levels distinguish human from non-human primate brains. Proc Natl Acad Sci USA 2003 100 13030 13035 14557539 10.1073/pnas.2135499100
TeraGenomics
GOTM: Gene Ontology Tree Machine
Zhang B Schmoyer D Kirov S Snoddy J GOTree Machine (GOTM): a web-based platform for interpreting sets of interesting genes using Gene Ontology hierarchies. BMC Bioinformatics 2004 5 16 14975175 10.1186/1471-2105-5-16
Luther SA Tang HL Hyman PL Farr AG Cyster JG Coexpression of the chemokines ELC and SLC by T zone stromal cells and deletion of the ELC gene in the plt/plt mouse. Proc Natl Acad Sci USA 2000 97 12694 12699 11070085 10.1073/pnas.97.23.12694
Nakano H Gunn MD Gene duplications at the chemokine locus on mouse chromosome 4: multiple strain-specific haplotypes and the deletion of secondary lymphoid-organ chemokine and EBI-1 ligand chemokine genes in the plt mutation. J Immunol 2001 166 361 369 11123313
Eckenstein FP Fibroblast growth factors in the nervous system. J Neurobiol 1994 25 1467 1480 7852998 10.1002/neu.480251112
Hashimoto M Sagara Y Langford D Everall IP Mallory M Everson A Digicaylioglu M Masliah E Fibroblast growth factor 1 regulates signaling via the glycogen synthase kinase-3beta pathway. Implications for neuroprotection. J Biol Chem 2002 277 32985 32991 12095987 10.1074/jbc.M202803200
Mitani A Oomura Y Yanase H Kataoka K Acidic fibroblast growth factor delays in vitro ischemia-induced intracellular calcium elevation in gerbil hippocampal slices: a sign of neuroprotection. Neurochem Int 1992 21 337 341 1284622 10.1016/0197-0186(92)90184-S
Thorns V Masliah E Evidence for neuroprotective effects of acidic fibroblast growth factor in Alzheimer disease. J Neuropathol Exp Neurol 1999 58 296 306 10197821
Thorns V Licastro F Masliah E Locally reduced levels of acidic FGF lead to decreased expression of 28-kda calbindin and contribute to the selective vulnerability of the neurons in the entorhinal cortex in Alzheimer's disease. Neuropathology 2001 21 203 211 11666017 10.1046/j.1440-1789.2001.00399.x
Hossain MA Russell JC Gomez R Laterra J Neuroprotection by scatter factor/hepatocyte growth factor and FGF-1 in cerebellar granule neurons is phosphatidylinositol 3-kinase/akt-dependent and MAPK/CREB-independent. J Neurochem 2002 81 365 378 12064484 10.1046/j.1471-4159.2002.00837.x
Lee EY Lee SY Lee TS Chi JG Choi W Suh YH Ultrastructural changes in microvessel with age in the hippocampus of senescence-accelerated mouse (SAM)-P/10. Exp Aging Res 2000 26 3 14 10689553 10.1080/036107300243650
Chung HY Kim HJ Kim JW Yu BP The inflammation hypothesis of aging: molecular modulation by calorie restriction. Ann NY Acad Sci 2001 928 327 335 11795524
Prolla TA DNA microarray analysis of the aging brain. Chem Senses 2002 27 299 306 11923192 10.1093/chemse/27.3.299
Lee CK Klopp RG Weindruch R Prolla TA Gene expression profile of aging and its retardation by caloric restriction. Science 1999 285 1390 1393 10464095 10.1126/science.285.5432.1390
Ly DH Lockhart DJ Lerner RA Schultz PG Mitotic misregulation and human aging. Science 2000 287 2486 2492 10741968 10.1126/science.287.5462.2486
Yoshida S Yashar BM Hiriyanna S Swaroop A Microarray analysis of gene expression in the aging human retina. Invest Ophthalmol Vis Sci 2002 43 2554 2560 12147584
Cao SX Dhahbi JM Mote PL Spindler SR Genomic profiling of short- and long-term caloric restriction effects in the liver of aging mice. Proc Natl Acad Sci USA 2001 98 10630 10635 11535822 10.1073/pnas.191313598
Salvador GA Lopez FM Giusto NM Age-related changes in central nervous system phosphatidylserine decarboxylase activity. J Neurosci Res 2002 70 283 289 12391587 10.1002/jnr.10385
Lu T Pan Y Kao SY Li C Kohane I Chan J Yankner BA Gene regulation and DNA damage in the ageing human brain. Nature 2004 429 883 891 15190254 10.1038/nature02661
Lueders KK Kuff EL Sequences associated with intracisternal A particles are reiterated in the mouse genome. Cell 1977 12 963 972 597866 10.1016/0092-8674(77)90161-1
Dupressoir A Puech A Heidmann T IAP retrotransposons in the mouse liver as reporters of ageing. Biochim Biophys Acta 1995 1264 397 402 8547329
Puech A Dupressoir A Loireau MP Mattei MG Heidmann T Characterization of two age-induced intracisternal A-particle-related transcripts in the mouse liver. Transcriptional read-through into an open reading frame with similarities to the yeast ccr4 transcription factor. J Biol Chem 1997 272 5995 6003 9038221 10.1074/jbc.272.9.5995
Hosokawa M Kasai R Higuchi K Takeshita S Shimizu K Hamamoto H Honma A Irino M Toda K Matsumura A Grading score system: a method for evaluation of the degree of senescence in senescence accelerated mouse (SAM). Mech Ageing Dev 1984 26 91 102 6748759 10.1016/0047-6374(84)90168-4
Carter TA Del Rio JA Greenhall JA Latronica ML Lockhart DJ Barlow C Chipping away at complex behavior: transcriptome/phenotype correlations in the mouse brain. Physiol Behav 2001 73 849 857 11566218 10.1016/S0031-9384(01)00522-4
Queryman: Teradata technical documents
| 15960800 | PMC1175968 | CC BY | 2021-01-04 16:05:39 | no | Genome Biol. 2005 Jun 1; 6(6):R48 | utf-8 | Genome Biol | 2,005 | 10.1186/gb-2005-6-6-r48 | oa_comm |
==== Front
Genome BiolGenome Biology1465-69061465-6914BioMed Central London gb-2005-6-6-r491596080110.1186/gb-2005-6-6-r49ResearchLarge-scale 13C-flux analysis reveals mechanistic principles of metabolic network robustness to null mutations in yeast Blank Lars M [email protected] Lars [email protected] Uwe [email protected] Institute of Biotechnology, ETH Zürich, 8093 Zürich, Switzerland2005 17 5 2005 6 6 R49 R49 1 2 2005 8 3 2005 6 4 2005 Copyright © 2005 Blank 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.
Genome-scale 13C-flux analysis in Saccharomyces cerevisiae revealed that the apparent dispensability of knockout mutants with metabolic function can be explained by gene inactivity under a particular condition, by network redundancy through duplicated genes or by alternative pathways.
Background
Quantification of intracellular metabolite fluxes by 13C-tracer experiments is maturing into a routine higher-throughput analysis. The question now arises as to which mutants should be analyzed. Here we identify key experiments in a systems biology approach with a genome-scale model of Saccharomyces cerevisiae metabolism, thereby reducing the workload for experimental network analyses and functional genomics.
Results
Genome-scale 13C flux analysis revealed that about half of the 745 biochemical reactions were active during growth on glucose, but that alternative pathways exist for only 51 gene-encoded reactions with significant flux. These flexible reactions identified in silico are key targets for experimental flux analysis, and we present the first large-scale metabolic flux data for yeast, covering half of these mutants during growth on glucose. The metabolic lesions were often counteracted by flux rerouting, but knockout of cofactor-dependent reactions, as in the adh1, ald6, cox5A, fum1, mdh1, pda1, and zwf1 mutations, caused flux responses in more distant parts of the network. By integrating computational analyses, flux data, and physiological phenotypes of all mutants in active reactions, we quantified the relative importance of 'genetic buffering' through alternative pathways and network redundancy through duplicate genes for genetic robustness of the network.
Conclusions
The apparent dispensability of knockout mutants with metabolic function is explained by gene inactivity under a particular condition in about half of the cases. For the remaining 207 viable mutants of active reactions, network redundancy through duplicate genes was the major (75%) and alternative pathways the minor (25%) molecular mechanism of genetic network robustness in S. cerevisiae.
==== Body
Background
The availability of annotated genomes and accumulated biochemical evidence for individual enzymes triggered the reconstruction of stoichiometric reaction models for network-based pathway analysis [1,2]. For many microbes, such network models are available at the genome scale, providing a largely comprehensive metabolic skeleton by interconnecting all known reactions in a given organism [3,4]. Thus, network properties such as optimal performance, flexibility to cope with ever-changing environmental conditions, and enzyme dispensability (also referred to as robustness or genetic robustness [5,6]) become mathematically tractable. These computational advances are matched with post-genomic advances in experimental methods that assess the cell's molecular make-up at the level of mRNA, protein, or metabolite concentrations. As the functional complement to these compositional data, quantification of intracellular in vivo reaction rates or molecular fluxes has been a focal point of method development in the realm of metabolism [7-9]. Recent progress in increasing the throughput of stable-isotope-based flux analyses [8,10,11] has allowed the quantification of flux responses to more than just a few intuitively chosen genetic or environmental perturbations [12-14]. Now that flux quantification in hundreds of null mutants under a particular condition is feasible in principle, the question arises of which mutants should be analyzed.
As perhaps the most widely used model eukaryote, the yeast Saccharomyces cerevisiae features a metabolic network of about 1,200 reactions that represent about 750 biochemically distinct reactions [3,15]. Is it necessary to quantify flux responses to null mutations in all reactions for a comprehensive view of the metabolic capabilities under a given condition? To address this question, we used a recently modified version (iLL672; L Kuepfer, U Sauer and LM Blank, unpublished work) of the original iFF708 genome-scale model published by Förster et al. [3]. On the basis of this model, we estimated the genome-scale flux distribution in wild-type S. cerevisiae from 13C-tracer experiments, to identify the 339 biochemical reactions that were active during growth on glucose. Yeast metabolism has the potential flexibility to use alternative pathways for 105 of these active reactions. For a major fraction of the potentially flexible reactions that catalyze significant flux, we then constructed prototrophic knockout mutants to elucidate whether or not the alternative pathway was used upon experimental knockout; that is, whether it contributes to the genetic robustness of the network [5,6]. For the purpose of this work, robustness is defined as the ability to proliferate on glucose as the sole carbon source upon knockout of a single gene with metabolic function.
Results
Identification of flexible reactions in yeast metabolism
To identify all potentially flexible reactions in yeast glucose metabolism that were active under a given condition, we used the recently reconciled metabolic network model iLL672 with 1,038 reactions (encoded by 672 genes) that represent 745 biochemically distinct reactions (L Kuepfer, U Sauer and LM Blank, unpublished work), which was based on the genome-scale S. cerevisiae model iFF708 [3]. The main modifications to the original model include elimination of dead-end reactions and a new formulation of cell growth. It should be noted that none of the results below critically depended on the network model, but the reconciliation of iLL672 enabled a more accurate discrimination between lethal and viable reactions than iFF708, as was validated by large-scale growth experiments (L Kuepfer, U Sauer and LM Blank, unpublished work).
First, we identified all reactions active in wild-type glucose metabolism by genome-scale flux analysis. For this purpose, we determined the wild-type flux distribution in central metabolism from a stable isotope batch experiment with 20% [U-13C] and 80% unlabeled glucose. This flux solution was then mapped to the genome scale by using minimization of the Euclidean norm of fluxes as the objective function. In total, 339 of the 745 biochemical reactions were active during growth on glucose alone (Figure 1 and Additional data file 1), which agrees qualitatively with the estimate of Papp et al. [16]. Most active reactions (234) were essential: 155 are encoded by singleton genes, 64 by two or more duplicate genes and 15 by yet unknown genes (Figure 1; Additional data file 1). In the entire network, only the remaining 105 reactions (30 encoded by yet unknown genes) were active and potentially flexible in the sense that they may be bypassed via alternative pathways (Figure 1). As fluxes in the peripheral reactions were typically below 0.1% of the glucose uptake rate (see Additional data file 1), we focused on the 51 gene-encoded flexible reactions that catalyzed a flux of at least 0.1%. These 51 reactions were encoded by 75 genes (43 duplicates, 23 singletons and 9 multiprotein complexes).
Physiological fitness of mutants deleted in flexible reactions
In 38 of these genes, which encoded 28 of the 51 potentially flexible and highly active reactions, we constructed prototrophic deletion mutants by homologous recombination [17] in the physiological model strain CEN.PK [18] (Figure 2). The prototrophic background was chosen to minimize potential problems of amino-acid supplementation for quantitative analysis [19]. These 38 experimental knockouts were in the pentose phosphate (PP) pathway, tricarboxylic acid (TCA) cycle, glyoxylate cycle, polysaccharide synthesis, mitochondrial transporters, and by-product formation (Figure 2, Table 1). Genetically, the knockouts encompass 14 singleton and 24 duplicate genes, including six gene families of which all members were deleted.
With the exception of gnd1, all 38 mutants grew with glucose as the sole carbon source. The lethal phenotype of the gnd1 mutant is consistent with a previous report [20] and is similar to the gndA mutant in Bacillus subtilis [21]. As in B. subtilis, we could select gnd1 suppressor mutants on glucose (data not shown). To assess the quantitative contribution of each gene to the organism's fitness, we determined maximum specific growth rates in minimal and complex medium using a well-aerated microtiter plate system [22]. Mutant fitness was then expressed as the normalized growth rate, relative to the growth rate of the reference strain (Table 1). In contrast to the previously reported competitive fitness [20,23,24], the fitness determined here is a quantitative physiological value.
In complex YPD medium, physiological fitness in the 38 viable haploid mutants was generally in qualitative agreement with the competitive fitness [20]. Quantitatively, however, our data seem to allow a better discrimination (Table 1), and significant differences between physiological and competitive fitness were seen in the adh1, fum1, and gpd1 mutants. Only threemutants - adh1, fum1, and gly1 - exhibited a fitness defect of 20% or greater (Table 2). gly1 lacks threonine aldolase, which catalyzes cleavage of threonine to glycine [25], hence its phenotype remains unexplained because glycine was present in the YPD medium.
In general, growth on the single substrate reduced the metabolic flexibility, as a much greater proportion of mutants exhibited significant fitness defects (Table 2). Major fitness defects were prominent in mutants of the PP pathway (gnd1, rpe1, sol3, and zwf1), which indicates an increased demand of NADPH for biosynthesis. Fitness of the fum1 mutant was clearly lower than that of other TCA-cycle mutants, for which duplicate genes exist. The strong phenotype of the fum1 mutant was somewhat unexpected because the flux through the TCA cycle is generally low or absent in glucose batch cultures of S. cerevisiae [13,14,26,27].
Intracellular carbon flux redistribution in response to gene deletions
While physiological data quantify the fitness defect, they cannot differentiate between intracellular mechanisms that bring about robustness to the deletion. To identify how carbon flux was redistributed around a metabolic lesion, we used metabolic flux analysis based on 13C-glucose experiments [8,9]. In contrast to in vitro enzyme activities and expression data, 13C-flux analysis provides direct evidence for such in vivo flux rerouting or its absence. The flux protocol consists of two distinct steps: first, analytical identification of seven independent metabolic flux ratios with probabilistic equations from the 13C distribution in proteinogenic amino acids [12,28,29]; and second, estimation of absolute fluxes (in vivo reaction rates) from physiological data and the flux ratios as constraints [10,30]. The relative distribution of intracellular fluxes was rather invariant in the 37 mutants, with the fraction of mitochondrial oxaloacetate derived through the TCA cycle flux and the fraction of mitochondrial pyruvate originating from malate as prominent exceptions (Figure 3).
From the experimentally determined uptake/production rates and the flux ratios as constraints (Additional data file 3), absolute intracellular fluxes were calculated using a compartmentalized stoichiometric model that consists of 35 reactions and 30 metabolites (Additional data file 2). This flux model comprised mostly the reactions of central carbon metabolism that were most relevant to the genetic changes introduced. It should be noted that the deleted reactions, with the exception of pyruvate dehydrogenase (PDA1), were not omitted from the network model; thus the calculated absence of flux through a given reaction was independently verified from the 13C-labeling data. In contrast to the relative distribution of intracellular fluxes, absolute reaction rates varied significantly in the mutants. With the exception of the flux through the TCA cycle (Figure 4f) and the gluconeogenic PEP carboxykinase (Figure 4d), all other fluxes generally increased with increasing glucose uptake rates (Figure 4). Eleven of the 37 mutants, however, exhibited specific flux responses that deviated from this general trend (Table 2, Figure 4).
Specific flux responses in singleton and duplicate gene knockouts
Specific flux responses were more prominent among the singleton mutants (Table 2, Figure 4). Although the TCA cycle flux through the NAD+-dependent fumarase reaction from fumarate to malate was already very low in the reference strain (Figures 3, 4f), the fum1 mutant exhibited a pronounced phenotype with altered redox metabolism and significant glycerol production (Figure 5). Inactivation of the mitochondrial pyruvate dehydrogenase complex in the pda1 mutant was bypassed by the import of cytosolic acetyl-CoA into the mitochondria. Inactivation of the oxidative PP pathway branch in the zwf1 mutant was compensated by a reversed flux in the non-oxidative PP pathway to provide the biomass precursors pentose 5-phosphate and erythrose 4-phosphate (Figure 5). Because the primary role of the PP pathway on glucose is generation of NADPH, NADP+-dependent mitochondrial malic enzyme flux was significantly increased in the zwf1 mutant. This NADPH compensation by malic enzyme was also suggested recently from co-feeding experiments [31].
In contrast to singletons, deletion of flexible duplicate genes could be compensated by either alternative pathways or isoenzymes. In most cases, however, the isoenzymes were used because no flux alteration was detected, with the a dh1, ald6, cox5A, and mdh1 mutants as exceptions (Table 2). Deletion of the major acetate-producing acetaldehyde dehydrogenase, the cytoplasmic ALD6 [32], significantly reduced acetate formation. The primary effect of the deletion was the strongly reduced glucose-uptake rate (Figure 4). Although a major source of NADPH was inactivated in this mutant [33], the PP pathway flux was not increased, but was even lower than in the reference strain (Figure 6). This indicates that the strongly decreased fitness of the ald6 mutant (Table 1) could result from NADPH starvation - that is, a suboptimal rate of NADP+ reduction. Consistent with this, we estimated that the NADPH requirement exceeded the combined NADPH formation from the oxidative PP pathway and malic enzyme by 70%, indicating that an as-yet-unidentified reaction(s) substitutes for the remaining NADPH production. Candidates are the mitochondrial acetaldehyde dehydrogenase Ald4p [34], which can use either NAD+ or NADP+ as redox cofactors or the mitochondrial NADH kinase Pos5p [35]. Deletion of the cytochrome c oxidase subunit Va COX5A in the mitochondrial respiratory chain increased glycerol production, which serves as means to reoxidize NADH (Figures 4b, 6). Because this mutant lacks functional mitochondria [36], glycerol production was driven by the limited NADH reoxidation through residual NADH oxidase activity in the electron-transport chain. Thus, robustness was brought about by using an alternative NADH sink. Considering that the flux through the mitochondrial malate dehydrogenase Mdh1 was already very low in the reference strain, the fitness defect of the mdh1 was surprising. Akin to the fum1 and ald6 mutants, the significantly reduced fitness of mdh1 may be explained by the imbalance between the TCA cycle and glucose catabolism (Figure 4f). Generally, the TCA cycle flux increases with decreasing glucose uptake rates [29], but remains non-proportionally low (absent) in the fum1, ald6, and mdh1 mutants (Figure 4f). The cytosolic and peroxisomal duplicate genes MDH2 and MDH3, respectively, did not compensate for the mitochondrial lesion, which is consistent with the observed lethal phenotype of the mdh1 mutant when grown on acetate [37].
Genetic network robustness
The above flux results reveal that knockouts of flexible reactions are bypassed through alternative pathways in about one third of the cases and through isoenzymes in the other two thirds. Does this reflect the relative contribution of alternative pathways and duplicate genes to genetic network robustness? [5] To address this question quantitatively for glucose metabolism, we grew the 196 duplicate (encoding 87 reactions) and 171 singleton (encoding 207 reactions) knockout mutants of all 294 gene-encoded active reactions on glucose plates.
In the 47 viable singleton knockouts, flux rerouting through an alternative pathway ensures survival, which was directly verified by flux data in 10 cases (Figure 4, Table 3 and Additional data file 3). Of the 196 experimental duplicate mutants, 180 grew on sole glucose, while 16 of the mutations were lethal. As these 16 duplicates obviously did not contribute to genetic robustness, their entire families (36 genes) were subtracted from the 150 duplicate-encoded essential reactions (Figure 1). For the remaining 114 duplicate genes we have strong evidence for network redundancy as the underlying mechanism of robustness, because they encode essential reactions (as determined in silico) and each of the experimental knockouts was viable (Figure 7). For the 46 duplicate genes that encode flexible reactions (Figure 1), both compensation by duplicates and/or alternative pathways might ensure proliferation. Where available, these mutants were classified according to their flux distribution; that is, of the 24 experimental duplicate mutants analyzed, four used alternative pathways and 20 an isoenzyme (Figure 4, Table 3 and Additional data file 3). In total we analyzed all 367 experimental mutants that encode the 294 active reactions of glucose metabolism, 140 of which were lethal and 227 viable. For the vast majority of the viable mutants, we identified the molecular mechanism that brought robustness to the knockout about: about 25% were alternative pathways and 75% duplicate genes (Figure 7).
Discussion
Using an integrated computational and experimental approach, we show here that metabolic flexibility to knockout mutations is restricted to a relatively small set of biochemical reactions. About a third of all active reactions under the particular condition investigated may be bypassed by alternative pathways, of which about 30% support only negligible fluxes. The occurrence of flexible reactions might be even lower in prokaryotes, because several alternative pathways involved inter-compartmental transport. In general, the number of flexible reactions will differ substantially between species, with free-living yeast and fungi at the upper end of the scale, and intracellular pathogens with highly reduced genomes at the lower end.
Using flux balance (FBA) [1,2], elementary flux mode [38,39], or similar analyses [40], all in silico flexible reactions can be precisely identified. Hence, experimental analysis of intracellular flux responses to metabolic gene deletions can be limited to these potentially flexible mutants, rather than having to analyze the entire mutant collection. Using the systems biological approach described here, the true in vivo capability of metabolic network operation can be mapped with a reasonable workload. As the knowledge base on intracellular flux responses increases, a handful of flux experiments in computationally identified mutants will probably suffice to identify the in vivo network capability under a given condition. At the next level, such flux analyses will also include mutants affected in regulatory genes that modulate the network composition. Although not covered in the stoichiometric models employed for flux-balance analysis, several recently discussed computational approaches [39,41,42] may aid in identifying the most relevant regulatory mutants for in vivo flux quantification.
Consistent with the notion that metabolic networks undergo minimal flux redistributions with respect to the metabolic state of the parent [40], deletion of flexible singleton genes was mostly counteracted by local flux rerouting, for example, the lsc1,mae1, and oac1 mutants (see Additional data file 3). Deletions in redox cofactor-dependent singleton or duplicate reactions such as those mediated by adh1, ald6, cox5A, pda1, and zwf1, however, affected flux alterations in more distant reactions. While the relative flux distribution (in % values) was perturbed only very little in these mutants, the absolute magnitude of fluxes (in vivo reaction velocities) varied dramatically. In particular, knockout of fum1, whose encoded protein catalyzes only a rather small flux, led to an unexpectedly strong phenotype with about a 50% reduction in glucose-uptake rate. Although unexpected, this finding was qualitatively consistent with results from a recent genetic footprinting study [43], which also showed a significant fitness defect in this mutant. It was speculated that intramitochondrial shortage of amino acids such as aspartate and glutamate causes a lack of respiratory chain components, which leads to a petite-like phenotype [44]. Another key mutant was pda1, whose knockout caused a substantial import of acetyl-CoA into the mitochondria; the mechanism for this remains elusive because the carnitine auxotrophic CEN.PK strain does not use the carnitine shuttle [45]. As a consequence, a twofold overproduction of NADPH was estimated, which suggests that the NAD+-dependent acetaldehyde dehydrogenases instead of the NADP+-dependent ALD6 were active to balance NADPH formation/consumption. Consistent with this, the flux through the NADPH-producing PP pathway was significantly lower in this mutant. The strongly altered redox metabolism in pda1 is further evidenced by the substantial secretion of glycerol and succinate (Figure 5).
The metabolic flexibility to cope with metabolic lesions is generally known as genetic robustness [5], a concept that is used to explain the seemingly surprising number of phenotypically silent deletion mutations: only about 1,100 knockouts of the 5,700 genes are lethal in haploid S. cerevisiae [23,46]. The causes and evolution of gene dispensability have been investigated in several theoretical analyses of pre-existing data, but the issue remains controversial [5,6,16,47-49]. For metabolic networks, our flux data differentiate between the relative contributions of three mechanisms to the apparent genetic robustness: inactive, and thus dispensable, genes; 'genetic buffering' through alternative reactions; and functional complementation from duplicate genes ('redundancy').
Conclusions
In qualitative agreement with a recent estimate [16], genome-scale flux analysis revealed that about half of the available reactions (45% of the known metabolic genes) were not required for growth on glucose (Figure 1). Hence, deletion of these genes would not affect the growth phenotype on this substrate, making inactive reactions the primary reason for the apparent dispensability of genes with metabolic function. It should be noted that this apparent gene dispensability is a somewhat artificial classification that does not contribute to genetic robustness because most of these genes encode metabolic functions that are only relevant under conditions different from the one tested. The most important mechanism of true genetic robustness in yeast glucose metabolism was duplicate genes (Figure 7), the majority of which encoded essential reactions with no alternative pathway. Alternative pathways, contributed about 25% to genetic robustness by carbon flux rerouting. This leaves redundancy as the major, and modularity the minor, cause [50] of metabolic network robustness to single-gene deletions during growth on glucose.
Materials and methods
Yeast strains
All prototrophic S. cerevisiae deletion mutants were constructed in the haploid, CEN.PK113-7D (Mata MAL2-8c SUC2) background with the homolog flanking region approach [17] (Table 1). Briefly, genomic DNA was isolated from the corresponding amino-acid auxotrophic mutants [23]. The kanMX4 cassettes of each mutant were amplified by PCR with primers located about 500 bp upstream and downstream of the deleted genes. The PCR reaction mixture was directly used for transformation and integrants were selected on YPD plates with 300 μg/ml geneticin. Correct cassette insertion was confirmed by overlapping PCR using either primer KanB (5'-CTGCAGCGAGGAGCCGTAAT-3') or KanC (5'-TGATTTTGATGACGAGCGTAA-3') primers in combination with one gene-specific primer. The reference strain was CEN.PK 113-7D with a deletion of the switching endonuclease, which was shown to be phenotypically neutral in chemostat competition experiments [51] and is commonly used as reference [52,53].
Media and growth conditions
The composition of the yeast minimal medium (MM) was [54]: 5 g (NH4)2SO4, 3 g KH2PO4, 0.5 g MgSO4·7H2O, 4.5 mg ZnSO4·7H2O, 0.3 mg CoCl2·6H2O, 1.0 mg MnCl2·4H2O, 0.3 mg CuSO4·5H2O, 4.5 mg CaCl2·2H2O, 3.0 mg FeSO4·7H2O, 0.4 mg NaMoO4·2H2O, 1.0 mg H3BO3, 0.1 mg KI, 15 mg EDTA, 0.05 mg biotin, 1.0 mg calcium pantothenate, 1.0 mg nicotinic acid, 25 mg inositol, 1.0 mg pyridoxine, 0.2 mg p-aminobenzoic acid, and 1.0 mg thiamine. The medium was buffered at pH 5.0 with 100 mM KH-phthalate to reduce pH changes throughout the growth experiments to 0.05. Filter-sterilized glucose and geneticin (300 μg/ml) were added freshly to the media. Batch growth experiments (1.2 ml) were carried out in deep-well plates (System Duetz, Kühner AG, Switzerland) using an orbital shaker with 5 cm amplitude at 300 rpm to allow optimal mixing [22].
Qualitative testing of mutant growth on glucose was done on agar plates. For this purpose, we used the haploid yeast mutant library in the BY4741 strain (MATa his3Δ1 leu2Δ0 met15Δ0 ura3Δ0) [23]. The composition of the yeast minimal medium for the plate growth assay was as described above [54] with the exception of 20 g/l agar for solidification. Glucose was added to a final concentration of 20 g/l. Strain auxotrophies were complemented with 20 mg/l histidine, uracil, methionine, lysine and 60 mg/l leucine. The plates were incubated at 30°C for 3 days before scoring of the growth phenotype and further incubated for 1 week to score slow-growth phenotype mutants.
Analytical procedures and 13C-labeling experiments
Cell growth was monitored by following optical density changes at a wavelength of 600nm (OD600). Aliquots for extracellular metabolite analysis were centrifuged at 14,000 rpm in an Eppendorf tabletop centrifuge to remove cells. Glucose, acetate, ethanol and glycerol concentrations in the supernatant were determined with commercial enzymatic kits (Scil Diagnostics, Germany). Organic acids were quantified by high-pressure liquid chromatography (HPLC) using a Supelcogel C8 (4.6 by 250 mm) ion-exclusion column. The column was eluted at 30°C with 2% sulfuric acid at a flow rate of 0.3 ml/min. The organic acids were detected using a PerkinElmer UV detector (Series 2000) at a wavelength of 210 nm. The physiological parameters maximum specific growth rate, biomass yield on glucose, and specific glucose consumption rate were calculated during the exponential growth phase.
All labeling experiments were carried out in batch cultures assuming pseudosteady-state conditions during the exponential growth phase [12,55]. 13C-labeling of proteinogenic amino acids was achieved either by growth on 5 g/l glucose as a mixture of 80% (w/w) unlabeled and 20% (w/w) uniformly labelled [U-13C]glucose (13C > 99%; Martek Biosciences, Columbia, MD) or 100% [1-13C]glucose (> 99%; Omicron Biochemicals, South Bend, IN). Cells from overnight cultures were harvested by centrifugation and washed using sugar-free MM to remove residual unlabeled carbon sources. Cultures were routinely inoculated to an maximum OD600 of 0.03 and harvested by centrifugation at an OD600 ≤ 1. Residual medium was removed by washing the pellet with water. Cell protein was hydrolyzed for 24 h at 105°C in 6 M HCL and dried in a heating block at 85°C for 6 h. The free amino acids were derivatized at 85°C for 1 h using 15 μl dimethylformamide and 15 μl N-(tert-butyldimethylsilyl)-N-methyl-trifluoroacetamide [10]. Gas chromatography-mass spectrometry (GC-MS) analysis was carried out as reported [12] using a series 8000 GC in combination with an MD800 mass spectrometer (Fisons Instruments, Beverly, MA).
Metabolic flux ratio analysis
The recorded MS spectra include the distribution of mass isotopomers in 1-5 fragments of alanine, aspartate, glutamate, glycine, isoleucine, leucine, phenylalanine, proline, serine, threonine, tyrosine, and valine. For each amino-acid fragment α, a mass isotopomer distribution vector (MDV) was assigned:
with m0 being the fractional abundance of the lowest mass and mi>0 the abundances of molecules with higher masses. The MDVα values were corrected for naturally occurring stable isotopes [12] to obtain the exclusive mass isotopomer distributions of the carbon skeletons. The corrected MDVα were used to calculate the amino acids (MDVAA) and metabolites (MDVM) mass distribution vectors. Ratios of converging intracellular fluxes to a given metabolite were calculated from the MDVM as described previously [12,29].
In addition, the relative contribution of the PP pathway was quantified from [1-13C]glucose experiments by tracking the positional 13C-labeling [10,56]. The expected labeling pattern of triose phosphates or serine, which is derived exclusively through glycolysis, is 50% 13C-label in the C1 positions. Hence, the fraction of serine derived through the pentose phosphate (PP) pathway can be derived according to Equation 2 [12]:
where GLU3unlabeled is an unlabeled three-carbon fragment from a source molecule of glucose. The remaining fraction of serine must then be derived through glycolysis. This flux ratio was not corrected for the potential withdrawal of 13C-label in dihydroxyacetone-phosphate-based biomass synthesis (such as phospholipids) and glycerol formation [21], because the influence was negligible under the condition used. The largest effect was found in the mutant with the highest specific glycerol formation rate (cox5A), where the estimated relative flux through the PP pathway would decrease from 12 ± 1% to 9 ± 1%.
13C-constrained flux analysis
Absolute values of intracellular fluxes were calculated with a flux model comprising all the major pathways of yeast central carbon metabolism (Additional data file 2). Deleted reactions were not omitted from the mutant models; thus the mutations were independently verified from the 13C data. The stoichiometric matrix of 34 linear equations and 30 metabolites has an infinite condition number [57]; it is thus underdetermined, and has a solution space with an infinite number of different flux vectors that fulfill the constraints from determined uptake and production rates. To uniquely solve the system for fluxes (ν), a set of linearly independent equations that quantify flux ratios (FlRs) were used to obtain eight constraints on the relative flux distribution from METAFoR analysis (see Additional data file 2).
The fraction of cytosolic oxaloacetate originating from cytosolic pyruvate is given by:
The fraction of mitochondrial oxaloacetate derived through anaplerosis is given by:
The fraction of PEP originating from cytosolic oxaloacetate is given by:
The fraction of serine derived through glycolysis is given by:
The upper and lower bounds for mitochondrial pyruvate derived through the malic enzyme (from mitochondrial malate) are given by:
The contribution of glycine to serine biosynthesis is given by:
and, finally, the contribution of serine to glycine biosynthesis is given by:
The stoichiometric matrix including Equations 3-10 has a condition number of 31, implying that the model is numerically robust [57]. Error minimization was carried out as described by Fischer et al. [10]. Balanced NADPH production and consumption were not added as additional constraints. In general, NADPH production was constrained by Equations 3 and 7/8, which estimate the relative use of the PP pathway and malic enzyme, respectively. As an additional source of NADPH, the flux through the NADPH-dependent acetaldehyde dehydrogenase [33] was estimated from the acetate production rate and the biomass requirement for cytosolic acetyl-CoA. Deviation of the NADPH production estimated in this way from the consumption for biosynthesis was generally below ± 20%, suggesting that the model assumptions and the experimental data are highly consistent. All extreme flux patterns were independently verified in 30-ml cultures (data not shown).
Genome-scale flux analysis
We used the experimentally determined in vivo flux data (νexp) to constrain the purely stoichiometric solution space of model iLL672 to obtain an experimentally validated genome-scale wild-type flux solution νWT. For glucose minimal medium, we constrained the model iLL672 with 30 fluxes that were derived from 13C-labeling experiments [8]. In particular, we used 13C-constrained flux analysis [58] for GC-MS-detected mass isotope distributions in proteinogenic amino acids from a 20% [U-13C] glucose experiment and a compartmentalized yeast model [29]. These experimental data were to be kept within an accuracy δ of ± 10% when mapping the determined central metabolic fluxes to the genome-scale reference flux solution. To overcome mathematical artifacts such as futile cycling (that is, a closed loop of fluxes that bring no net change), the original linear programming problem was modified. A minimization of the Euclidean norm of fluxes was chosen as the objective function such that (s.t.) the mass balance equations hold:
with j as the set of experimentally determined fluxes. Reactions were categorized as flexible when fulfilling the following criteria: the reactions carried a non-zero flux; and the reaction was not essential for growth.
In silico phenotyping of duplicate gene families
Phenotype predictions of deletion mutants were analyzed computationally with FBA [3,59]. Assuming steady-state growth, mass balances were put up for each intracellular metabolite Mi (1 × n) that have to be fulfilled, when multiplied with the overall flux vector ν (n × 1):
Mi·ν = 0. (12)
The entity of all m metabolite mass balances yields the stoichiometric matrix S (m × n), where:
S·ν = 0. (13)
To pick one solution out of the overall solution space formed by the stoichiometric constraints, FBA generally assumes maximization of biomass growth μ as the global cellular goal [3,59]. Thus, the search for a single flux distribution ν results in a linear programming (LP) problem:
where i = 1,..,M and νlb,i and νub,i correspond to upper and lower bounds of a specific reaction i. Gene knockout mutants can be simulated easily in silico by setting the deleted reactions to zero. All LPproblems were solved using the open-source GNU linear programming kit [60].
Additional data files
The following additional data are available with the online version of this paper. The classification of reactions according to Figure 1 is presented in Additional data file 1. The flux analysis model is defined in Additional data file 2. The physiological data, flux ratios and the calculated flux distributions are presented in Additional data file 3.
Supplementary Material
Additional File 1
Classification of reactions according to Figure 1. Classification of reactions according to Figure 1
Click here for file
Additional File 2
The flux analysis model. The flux analysis model
Click here for file
Additional File 3
Physiological data, flux ratios and the calculated flux distributions. Physiological data, flux ratios and the calculated flux distributions
Click here for file
Acknowledgements
We are grateful to Eckhard Boles for providing the mae1 mutant. LarsM. Blank gratefully acknowledges financial support by the Deutsche Akademie der Naturforscher Leopoldina (BMBF-LPD/8-78).
Figures and Tables
Figure 1 Genome-wide proportion of active, essential and flexible metabolic reactions during growth of S. cerevisiae (iLL672) on glucose. Flexible reactions are defined as having a non-zero flux but are not essential for growth. The number of genes that encode biochemical reactions is given in parentheses.
Figure 2 Central carbon metabolism of S. cerevisiae during aerobic growth on glucose. Gene names in boxes are given for reactions that were identified as flexible by flux balance analysis. Dark gray boxes indicate mutants, for which the carbon flux distribution was determined by 13C-tracer experiments. Dots indicate that the gene is part of a protein complex. Arrowheads indicate reaction reversibility. Extracellular substrates and products are capitalized. C1, one-carbon unit from C1 metabolism.
Figure 3 The distribution of six independently determined metabolic flux ratios in 37 deletion mutants during growth on glucose. In each case, the median of the distribution is indicated by a vertical line, the 25th percentile by the grey box and the 90th percentile by the horizontal line. Data points outside the 90th percentile are indicated by dots. The reference strain is indicated by the open circle.
Figure 4 Absolute metabolic fluxes in the 37 flexible mutants as a function of glucose uptake rate or selected intracellular fluxes. (a-f) Glucose uptake rate; (g,h) selected intracellular fluxes. The linear regression of the distribution and the 99% prediction interval are indicated by the solid and dashed lines, respectively. Mutants with significant changes in the carbon-flux distribution are indicated. The reference strain is indicated by an open circle. Extreme flux patterns were verified in 30-ml shake flask cultures (data not shown).
Figure 5 Relative distributions of absolute carbon fluxes in the S. cerevisiae reference strain (Ref) and the singleton gene mutants fum1, pda1 and zwf. All fluxes are normalized to the specific glucose uptake rate, which is shown in the top inset, and are given in the same order in each box. Reactions encoded by deleted genes are shown on a black background, but were not removed from the flux model (except for PDA1). The NADPH balance that is based on the quantified fluxes and the known cofactor specificities is given as a synthetic transhydrogenase flux. In general, the 95% confidence intervals were between 5 and 10% for the major fluxes. Larger confidence intervals were estimated for reactions with low flux such as malic enzyme and PEP carboxykinase. Flux distributions were verified in 30-ml shake flask cultures (data not shown). C1, one-carbon unit from C1 metabolism; P5P, pentose 5-phosphates.
Figure 6 Relative distributions of absolute carbon fluxes in the S. cerevisiae reference strain and the duplicate gene mutants ald6, cox5A and mdh1. All fluxes are normalized to the specific glucose uptake rate, which is shown in the top inset, and are given in the same order in each box. Reactions encoded by deleted genes are shown on a black background, but were not removed from the flux model. The NADPH balance that is based on the fluxes and the known cofactor specificities is given as a synthetic transhydrogenase. In general, the 95% confidence intervals were between 5 and 10% for the major fluxes. Larger confidence intervals were estimated for reactions with low flux such as malic enzyme and PEP carboxykinase. Flux distributions were verified in 30-ml shake flask cultures (data not shown). C1, one-carbon unit from C1 metabolism; P5P, pentose 5-phosphates.
Figure 7 The mechanistic basis of gene dispensability in all active reactions during glucose metabolism of S. cerevisiae. The mechanism was mostly identified from the phenotype on glucose plates. For 10 of the alternative pathways and for 20 duplicates encoding flexible reactions, the results were confirmed by 13C-flux analysis. For 22 duplicate genes the data are not sufficient to distinguish between both mechanisms and they are labeled as not analyzed.
Table 1 Fitness of mutants with deletions in flexible central metabolic reactions
Physiological fitness* Competitive fitness† Physiological fitness Competitive fitness
Mutants MM YPD YPD Mutants MM YPD YPD
Reference strain 1 1 1
adh1/YOL086C 0.47 0.57 0.79 mdh2/YOL126C 0.89 0.98 1.01
adh3/YMR083W 0.92 0.87 0.98 mdh3/YDL078C 1.00 0.96 1.01
ald5/YER073W 1.02 0.94 1 mls1/YNL117W 1 0.98 1
ald6/YPL061W 0.34 0.87 0.9 oac1/YKL120W 0.71 0.94 1.01
cox5A/YNL052W 0.63 0.91 1 pck1/YKR097W 1 0.96 1
ctp1/YBR291C 0.91 1 0.97 pda1/YER178W 0.41 0.98 1
dal7/YIR031C 0.94 0.85 1 pgm1/YKL127W 0.82 0.94 1
fum1/YPL262W 0.52 0.62 0.93 pgm2/YMR105C 0.90 1 1
gnd1/YHR183W 0 0.87 1.01 rpe1/YJL121C 0.33 0.94 0.88
gnd2/YGR256W 0.83 0.98 1 sdh1/YKL148C 0.72 0.94 1
gcv2/YMR189W 0.92 0.94 1 ser33/YIL074C 0.92 0.94 1.01
gly1/YEL046C 0.79 0.74 0.87 sfc1/YJR095W 0.84 0.96 1.01
gpd1/YDL022W 1 0.98 0.84 sol1/YNR034W 0.91 1 1.02
icl1/YER065C 1 1 1 sol2/YCRX13W 0.99 0.98 1
idp1/YDL066W 0.92 0.94 1.03 sol3/ YHR163W 0.71 0.94 1
idp2/YLR174W 0.86 0.96 0.95 sol4/ YGR248W 0.95 0.91 1.01
lsc1/YOR142W 1.05 0.93 1 tal1/ YLR354C 0.89 0.94 1
mae1/YKL029C 1.01 0.96 1 YGR043C 0.92 0.87 1.02
mdh1/YKL085W 0.72 0.91 1 zwf1/YNL241C 0.38 0.96 ND
*Physiological fitness is defined as the maximal specific growth rate of a mutant normalized to the reference strain CEN.PK 113-7D ho::kanMX4. The average from triplicate experiments is shown. The standard deviation was generally below 0.05. †From Steinmetz et al. [20]. ND, not detected.
Table 2 Overview of mutants with a fitness defect of at least 20% or altered flux distribution
Mutants Fitness defect in YPD Fitness defect in MM Altered intracellular flux distribution*
Total number of mutants 3 of 38 12 (+1)† of 38 11 of 38
Singleton genes fum1 gly1 fum1 pda1 fum1 pda1
gly1 rpe1 lsc1 rpe1
oac1 zwf1 mae1 zwf1
oac1
Duplicate genes adh1 adh1 sdh1 adh1 cox5A
ald6 sol3 ald6 mdh1
cox5A (gnd1)
mdh1
*See Figures 5 and 6. †Lethal mutations are given in parentheses.
Table 3 Overview of mechanisms of metabolic flexibility that confer robustness to central metabolic deletions
Duplicate gene* Duplicate gene and alternative pathway† Alternative pathway‡ None
ADH3, ALD5, DAL7, GPD1,ICL1, IDP1, IDP2, MDH2, MDH3, MLS1, PGM1, PGM2, SDH1, SER33,SOL1, SOL2, SOL3, SOL4, TAL1, YGR043c ADH1, ALD6, COX5A,MDH1 FUM1, GLY1, LSC1, MAE1, MDH1, OAC1, PCK1, PDA1, RPE1, ZWF1 CTP1, GCV2, GND1§, GND2, SFC1
*Wild-type-like flux distribution. †Altered flux distribution, but some residual flux through the reaction was observed. ‡Altered flux distribution, but no residual flux through the reaction was observed. § Lethal, probably because of a non-stoichiometric effect.
==== Refs
Papin JA Stelling J Price ND Klamt S Schuster S Palsson BO Comparison of network-based pathway analysis methods. Trends Biotechnol 2004 22 400 405 15283984 10.1016/j.tibtech.2004.06.010
Price ND Reed JL Palsson BO Genome-scale models of microbial cells: evaluating the consequences of constraints. Nat Rev Microbiol 2004 2 886 897 15494745 10.1038/nrmicro1023
Förster 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 10.1101/gr.234503
Reed JL Palsson BO Thirteen years of building constraint-based in silico models of Escherichia coli. J Bacteriol 2003 185 2692 2699 12700248 10.1128/JB.185.9.2692-2699.2003
Gu X Evolution of duplicate genes versus genetic robustness against null mutations. Trends Genet 2003 19 354 356 12850437 10.1016/S0168-9525(03)00139-2
Hartman JLT Garvik B Hartwell L Principles for the buffering of genetic variation. Science 2001 291 1001 1004 11232561 10.1126/science.291.5506.1001
Hellerstein MK In vivo measurement of fluxes through metabolic pathways: the missing link in functional genomics and pharmaceutical research. Annu Rev Nutr 2003 23 379 402 12704218 10.1146/annurev.nutr.23.011702.073045
Sauer U High-throughput phenomics: experimental methods for mapping fluxomes. Curr Opin Biotechnol 2004 15 58 63 15102468 10.1016/j.copbio.2003.11.001
Wiechert W 13C metabolic flux analysis. Metab Eng 2001 3 195 206 11461141 10.1006/mben.2001.0187
Fischer E Zamboni N Sauer U High-throughput metabolic flux analysis based on gas chromatography-mass spectrometry derived 13C constraints. Anal Biochem 2004 325 308 316 14751266 10.1016/j.ab.2003.10.036
Zamboni N Sauer U Model-independent fluxome profiling from 2H and 13C experiments for metabolic variant discrimination. Genome Biol 2004 5 R99 15575973 10.1186/gb-2004-5-12-r99
Fischer E Sauer U Metabolic flux profiling of Escherichia coli mutants in central carbon metabolism using GC-MS. Eur J Biochem 2003 270 880 891 12603321 10.1046/j.1432-1033.2003.03448.x
Blank LM Sauer U TCA cycle activity in Saccharomyces cerevisiae is a function of the environmentally determined specific growth and glucose uptake rates. Microbiology 2004 150 1085 1093 15073318 10.1099/mic.0.26845-0
Blank LM Lehmbeck F Sauer U Metabolic flux and network analysis in fourteen hemiascomycetous yeasts. FEMS Yeast Res 2005 5 545 558 15780654 10.1016/j.femsyr.2004.09.008
Duarte NC Herrgard MJ Palsson BO Reconstruction and validation of Saccharomyces cerevisiae iND750, a fully compartmentalized genome-scale metabolic model. Genome Res 2004 14 1298 1309 15197165 10.1101/gr.2250904
Papp B Pal C Hurst LD Metabolic network analysis of the causes and evolution of enzyme dispensability in yeast. Nature 2004 429 661 664 15190353 10.1038/nature02636
Wach A Brachat A Pohlmann R Philippsen P New heterologous modules for classical or PCR-based gene disruptions in Saccharomyces cerevisiae. Yeast 1994 10 1793 1808 7747518
van Dijken JP Bauer J Brambilla L Duboc P Francois JM Gancedo C Giuseppin ML Heijnen JJ Hoare M Lange HC An interlaboratory comparison of physiological and genetic properties of four Saccharomyces cerevisiae strains. Enzyme Microb Technol 2000 26 706 714 10862876 10.1016/S0141-0229(00)00162-9
Pronk JT Auxotrophic yeast strains in fundamental and applied research. Appl Environ Microbiol 2002 68 2095 2100 11976076 10.1128/AEM.68.5.2095-2100.2002
Steinmetz LM Scharfe C Deutschbauer AM Mokranjac D Herman ZS Jones T Chu AM Giaever G Prokisch H Oefner PJ Systematic screen for human disease genes in yeast. Nat Genet 2002 31 400 404 12134146
Zamboni N Fischer E Laudert D Aymerich S Hohmann HP Sauer U The Bacillus subtilis yqjI gene encodes the NADP+-dependent 6-P-gluconate dehydrogenase in the pentose phosphate pathway. J Bacteriol 2004 186 4528 4534 15231785 10.1128/JB.186.14.4528-4534.2004
Duetz WA Ruedi L Hermann R O'Connor K Buchs J Witholt B Methods for intense aeration, growth, storage, and replication of bacterial strains in microtiter plates. Appl Environ Microbiol 2000 66 2641 2646 10831450 10.1128/AEM.66.6.2641-2646.2000
Giaever G Chu AM Ni L Connelly C Riles L Veronneau S Dow S Lucau-Danila A Anderson K Andre B Functional profiling of the Saccharomyces cerevisiae genome. Nature 2002 418 387 391 12140549 10.1038/nature00935
Winzeler EA Liang H Shoemaker DD Davis RW Functional analysis of the yeast genome by precise deletion and parallel phenotypic characterization. Novartis Found Symp 2000 229 105 109 discussion 109-111. 11084935
Monschau N Stahmann KP Sahm H McNeil JB Bognar AL Identification of Saccharomyces cerevisiae GLY1 as a threonine aldolase: a key enzyme in glycine biosynthesis. FEMS Microbiol Lett 1997 150 55 60 9163906 10.1016/S0378-1097(97)00096-7
Gombert AK Moreira dos Santos M Christensen B Nielsen J Network identification and flux quantification in the central metabolism of Saccharomyces cerevisiae under different conditions of glucose repression. J Bacteriol 2001 183 1441 1451 11157958 10.1128/JB.183.4.1441-1451.2001
Maaheimo H Fiaux J Cakar ZP Bailey JE Sauer U Szyperski T Central carbon metabolism of Saccharomyces cerevisiae explored by biosynthetic fractional 13C labeling of common amino acids. Eur J Biochem 2001 268 2464 2479 11298766 10.1046/j.1432-1327.2001.02126.x
Szyperski T Biosynthetically directed fractional 13C-labeling of proteinogenic amino acids. An efficient analytical tool to investigate intermediary metabolism. Eur J Biochem 1995 232 433 448 7556192
Blank LM Sauer U TCA cycle activity in Saccharomyces cerevisiae is a function of the environmentally determined specific growth and glucose uptake rates. Microbiology 2004 150 1085 1093 15073318 10.1099/mic.0.26845-0
Sauer U Hatzimanikatis V Bailey JE Hochuli M Szyperski T Wüthrich K Metabolic fluxes in riboflavin-producing Bacillus subtilis. Nat Biotechnol 1997 15 448 452 9131624 10.1038/nbt0597-448
Dos Santos MM Gombert AK Christensen B Olsson L Nielsen J Identification of in vivo enzyme activities in the cometabolism of glucose and acetate by Saccharomyces cerevisiae by using (13)C-labeled substrates. Eukaryotic Cell 2003 2 599 608 12796305 10.1128/EC.2.3.599-608.2003
Meaden PG Dickinson FM Mifsud A Tessier W Westwater J Bussey H Midgley M The ALD6 gene of Saccharomyces cerevisiae encodes a cytosolic, Mg(2+)-activated acetaldehyde dehydrogenase. Yeast 1997 13 1319 1327 9392076 10.1002/(SICI)1097-0061(199711)13:14<1319::AID-YEA183>3.0.CO;2-T
Grabowska D Chelstowska A The ALD6 gene product is indispensable for providing NADPH in yeast cells lacking glucose-6-phosphate dehydrogenase activity. J Biol Chem 2003 278 13984 13988 12584194 10.1074/jbc.M210076200
Boubekeur S Bunoust O Camougrand N Castroviejo M Rigoulet M Guerin B A mitochondrial pyruvate dehydrogenase bypass in the yeast Saccharomyces cerevisiae. J Biol Chem 1999 274 21044 21048 10409655 10.1074/jbc.274.30.21044
Outten CE Culotta VC A novel NADH kinase is the mitochondrial source of NADPH in Saccharomyces cerevisiae. EMBO J 2003 22 2015 2024 12727869 10.1093/emboj/cdg211
McEwen JE Cumsky MG Ko C Power SD Poyton RO Mitochondrial membrane biogenesis: characterization and use of pet mutants to clone the nuclear gene coding for subunit V of yeast cytochrome c oxidase. J Cell Biochem 1984 24 229 242 6330135 10.1002/jcb.240240305
McAlister-Henn L Thompson LM Isolation and expression of the gene encoding yeast mitochondrial malate dehydrogenase. J Bacteriol 1987 169 5157 5166 3312168
Schuster S Fell DA Dandekar T A general definition of metabolic pathways useful for systematic organization and analysis of complex metabolic networks. Nat Biotechnol 2000 18 326 332 10700151 10.1038/73786
Stelling J Klamt S Bettenbrock K Schuster S Gilles ED Metabolic network structure determines key aspects of functionality and regulation. Nature 2002 420 190 193 12432396 10.1038/nature01166
Segre D Vitkup D Church GM Analysis of optimality in natural and perturbed metabolic networks. Proc Natl Acad Sci USA 2002 99 15112 15117 12415116 10.1073/pnas.232349399
Covert MW Knight EM Reed JL Herrgard MJ Palsson BO Integrating high-throughput and computational data elucidates bacterial networks. Nature 2004 429 92 96 15129285 10.1038/nature02456
Segre D The regulatory software of cellular metabolism. Trends Biotechnol 2004 22 261 265 15158051 10.1016/j.tibtech.2004.04.013
Dunn B Ferea T Spellman P Schwarz J Terraciano J Troyanovich J Walker S Greene J Shaw K DiDomenico B Genetic footprinting: a functional analysis of the S. cerevisiae genome.
Wu M Tzagoloff A Mitochondrial and cytoplasmic fumarases in Saccharomyces cerevisiae are encoded by a single nuclear gene FUM1. J Biol Chem 1987 262 12275 12282 3040736
Van Maris AJ Luttik MA Winkler AA Van Dijken JP Pronk JT Overproduction of threonine aldolase circumvents the biosynthetic role of pyruvate decarboxylase in glucose-limited chemostat cultures of Saccharomyces cerevisiae. Appl Environ Microbiol 2003 69 2094 2099 12676688 10.1128/AEM.69.4.2094-2099.2003
Winzeler EA Shoemaker DD Astromoff A Liang H Anderson K Andre B Bangham R Benito R Boeke JD Bussey H Functional characterization of the S. cerevisiae genome by gene deletion and parallel analysis. Science 1999 285 901 906 10436161 10.1126/science.285.5429.901
Nowak MA Boerlijst MC Cooke J Smith JM Evolution of genetic redundancy. Nature 1997 388 167 171 9217155 10.1038/40618
Wagner A Robustness against mutations in genetic networks of yeast. Nat Genet 2000 24 355 361 10742097 10.1038/74174
Gu Z Steinmetz LM Gu X Scharfe C Davis RW Li WH Role of duplicate genes in genetic robustness against null mutations. Nature 2003 421 63 66 12511954 10.1038/nature01198
Stelling J Sauer U Szallasi Z Doyle FJ 3rdDoyle J Robustness of cellular functions. Cell 2004 118 675 685 15369668 10.1016/j.cell.2004.09.008
Baganz F Hayes A Marren D Gardner DC Oliver SG Suitability of replacement markers for functional analysis studies in Saccharomyces cerevisiae. Yeast 1997 13 1563 1573 9509575 10.1002/(SICI)1097-0061(199712)13:16<1563::AID-YEA240>3.3.CO;2-Y
Allen J Davey HM Broadhurst D Heald JK Rowland JJ Oliver SG Kell DB High-throughput classification of yeast mutants for functional genomics using metabolic footprinting. Nat Biotechnol 2003 21 692 696 12740584 10.1038/nbt823
Raamsdonk LM Teusink B Broadhurst D Zhang N Hayes A Walsh MC Berden JA Brindle KM Kell DB Rowland JJ A functional genomics strategy that uses metabolome data to reveal the phenotype of silent mutations. Nat Biotechnol 2001 19 45 50 11135551 10.1038/83496
Verduyn C Postma E Scheffers WA Van Dijken JP Effect of benzoic acid on metabolic fluxes in yeasts: a continuous-culture study on the regulation of respiration and alcoholic fermentation. Yeast 1992 8 501 517 1523884
Sauer U Lasko DR Fiaux J Hochuli M Glaser R Szyperski T Wüthrich K Bailey JE Metabolic flux ratio analysis of genetic and environmental modulations of Escherichia coli central carbon metabolism. J Bacteriol 1999 181 6679 6688 10542169
Christensen B Christiansen T Gombert AK Thykaer J Nielsen J Simple and robust method for estimation of the split between the oxidative pentose phosphate pathway and the Embden-Meyerhof-Parnas pathway in microorganisms. Biotechnol Bioeng 2001 74 517 523 11494219 10.1002/bit.1143
Nissen TL Schulze U Nielsen J Villadsen J Flux distributions in anaerobic, glucose-limited continuous cultures of Saccharomyces cerevisiae. Microbiology 1997 143 203 218 9025295
Fischer E Zamboni N Sauer U High-throughput metabolic flux analysis based on gas chromatography-mass spectrometry derived 13C constraints. Anal Biochem 2004 325 308 316 14751266 10.1016/j.ab.2003.10.036
Edwards JS Palsson BO The Escherichia coli MG1655 in silico metabolic genotype: its definition, characteristics, and capabilities. Proc Natl Acad Sci USA 2000 97 5528 5533 10805808 10.1073/pnas.97.10.5528
Makhorin A GNU Linear Programming Kit 2001 Moscow, Russia: Moscow Aviation Institute
Yeast Deletion Project and Proteomics of Mitochondria Database
| 15960801 | PMC1175969 | CC BY | 2021-01-04 16:05:40 | no | Genome Biol. 2005 May 17; 6(6):R49 | utf-8 | Genome Biol | 2,005 | 10.1186/gb-2005-6-6-r49 | oa_comm |
==== Front
Genome BiolGenome Biology1465-69061465-6914BioMed Central London gb-2005-6-6-r501596080210.1186/gb-2005-6-6-r50ResearchThe rarity of gene shuffling in conserved genes Conant Gavin C [email protected] Andreas [email protected] Department of Genetics, Smurfit Institute, University of Dublin, Trinity College, Dublin 2, Ireland2 Department of Biology, The University of New Mexico, Albuquerque, NM 87131-0001, USA2005 9 5 2005 6 6 R50 R50 31 1 2005 23 3 2005 13 4 2005 Copyright © 2005 Conant and Wagner; licensee BioMed Central Ltd.This is an Open Access article distributed under the 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 incidence of gene shuffling is estimated in conserved genes in 10 organisms from the three domains of life. Successful gene shuffling is found to be very rare among such conserved genes. This suggests that gene shuffling may not be a major force in reshaping the core genomes of eukaryotes.
Background
Among three sources of evolutionary innovation in gene function - point mutations, gene duplications, and gene shuffling (recombination between dissimilar genes) - gene shuffling is the most potent one. However, surprisingly little is known about its incidence on a genome-wide scale.
Results
We have studied shuffling in genes that are conserved between distantly related species. Specifically, we estimated the incidence of gene shuffling in ten organisms from the three domains of life: eukaryotes, eubacteria, and archaea, considering only genes showing significant sequence similarity in pairwise genome comparisons. We found that successful gene shuffling is very rare among such conserved genes. For example, we could detect only 48 successful gene-shuffling events in the genome of the fruit fly Drosophila melanogaster which have occurred since its common ancestor with the worm Caenorhabditis elegans more than half a billion years ago.
Conclusion
The incidence of gene shuffling is roughly an order of magnitude smaller than the incidence of single-gene duplication in eukaryotes, but it can approach or even exceed the gene-duplication rate in prokaryotes. If true in general, this pattern suggests that gene shuffling may not be a major force in reshaping the core genomes of eukaryotes. Our results also cast doubt on the notion that introns facilitate gene shuffling, both because prokaryotes show an appreciable incidence of gene shuffling despite their lack of introns and because we find no statistical association between exon-intron boundaries and recombined domains in the two multicellular genomes we studied.
==== Body
Background
How do genes with new functions originate? This remains one of the most intriguing open questions in evolutionary genetics. Three principal mechanisms can create genes of novel function: point mutations and small insertions or deletions in existing genes; duplication of entire genes or domains within genes, in combination with mutations that cause functional divergence of the duplicates [1-3]; and recombination between dissimilar genes to create new recombinant genes (see, for example [4,5]). We here choose to call only this kind of recombination gene shuffling, excluding, for example, duplication of domains within a gene. In such a gene shuffling event, the parental genes may be either destroyed or preserved [6]. Gene shuffling is clearly the most potent of the three causes of functional innovation because it can generate new genes with a structure drastically different from that of either parental gene. Laboratory evolution studies show that gene shuffling allows new gene functions to arise at rates of orders of magnitudes higher than point mutations [7,8].
Much is known about rates of point mutations [9] and of gene duplications [10,11]. In contrast, the rate at which gene shuffling occurs is relatively unexplored, despite the importance of shuffling for functional innovation. To be sure, anecdotal evidence suggests that successful gene shuffling occurs and that it creates genes with new functions [4]. In particular, proteins are often mosaics of domains that are characterized by sequence and structural similarity [12-19]. Many domains occur in multiple proteins of different functions, suggesting that new proteins can arise through the combination of domains of other proteins, a process requiring recombination. In addition, many studies have systematically identified one subclass of gene-recombination events - gene fusions [20-24]. These studies count gene fusion events in a genome of interest relative to multiple, often very distantly related, species. Because fused genes often have similar functions, identification of fusion events can aid in inferring gene functions. Here we address a question that goes beyond the above studies: how frequent is gene shuffling in comparison with other forces of genome change, such as gene duplication? This problem is difficult because of the many possible outcomes of recombination events. These outcomes fall into three principal categories, gene fusions, domain deletions, and domain insertions (Figure 1a). To identify these outcomes systematically on a genomic scale is computationally intensive, which has limited our analyses to a modest number of genomes (Table 1).
One can identify gene-shuffling events either from protein sequence information or from information about protein structure. Structure-based approaches [12-15] have the advantage of being able to detect recombination events where sequence similarity between a recombination product and its parents has eroded beyond recognition. However, because two very distantly related structural domains can also have arisen through convergent evolution [25,26], identifying common ancestry of two domains based on structure alone can be problematic. As a further limitation, structure-based approaches can only identify recombination events that respect the boundaries of protein domains, whereas some successful recombination events may occur within domains [27-29]. In addition, structural information is not available for all genes. For example, the Pfam database of protein domains [30] contains no structural information for more than 40% of proteins in budding yeast (Saccharomyces cerevisiae). Structure-based approaches may thus miss many shuffled genes. Because of these issues we chose a sequence-based approach which allows us to search for shuffling events without making restrictive assumptions regarding their nature. Essentially, our search imposes no restrictions on shuffling except that it must merge in a single gene two protein-coding sequences that were previously a part of two different genes. We thus avoid assuming that shuffling occurs only at domain boundaries or with certain recombination mechanisms without precluding either possibility. Our analysis can also account for gene-duplication events in either parental or recombined genes.
We here identify gene-shuffling events that have occurred in a 'test' species T since its divergence from a reference species R1. A gene in the test genome whose parts match more than one gene in the reference genome is a candidate for a gene-shuffling event that has occurred since the common ancestor of the two genomes. Our analysis also uses a third genome (reference genome R2) to prevent gene fission or gene loss in the reference genome R1 from resulting in spurious identification of gene shuffling events. Because R2 is an outgroup relative to T and R1, it allows us to detect such events in R1 (see Figure 1b). Like any comparative sequence-based approach, our analysis depends on detectable sequence similarity among genes. In other words, our analysis excludes rapidly evolving genes.
Results
Little gene shuffling in closely related genomes
Mosaic proteins are not rare in most genomes, which suggests that successful gene shuffling might be frequent on an evolutionary timescale. We thus searched four closely related genomes for shuffled genes. These genomes fit three essential criteria for this analysis: close taxonomic spacing; availability of complete genome sequence; and, most important, reliable gene identification. (Gene identification is notoriously unreliable in higher organisms because of their complex gene structure.) These species were the four yeasts Saccharomyces cerevisiae, S. paradoxus, S. bayanus and S. mikatae [31], which diverged from their common ancestor between 5 and 20 million years ago (Mya) [31]. We found multiple candidate genes for shuffling in different T-R1 pairs of these four species. However, almost all of these candidates proved spurious for a variety of reasons: First, some of them occurred in two or more species in a manner inconsistent with these species' phylogeny, or they matched more closely a single reference species gene than their two putative parents. Both observations make recent recombination an unparsimonious explanation for a gene's origin (Figure 1b). Second, the putative shuffled domains in some candidate genes had a synonymous, or silent, nucleotide divergence from their parental domains that differed by a factor of two or more. However, the recombined parts of a shuffled gene should show equal sequence divergence to their respective parental genes, because they have identical divergence times (namely the time since T and R1 shared a common ancestor). We used silent nucleotide substitutions as an indicator of sequence divergence because such substitutions are under little or no selection and thus accumulate in an approximately clock-like fashion [32]. Use of amino-acid changing (nonsynonymous) substitutions (Ka) as an indicator led to similar conclusions. After exclusion of all such spurious genes, only two potential shuffled genes remained in our analysis, which indicates a low incidence of gene shuffling.
Shuffled genes in distantly related genomes
Because our analysis of yeast genomes suggests that gene shuffling may be rarer than one might expect, the need arises to study more distantly related genomes. This raises two principal problems. First, such an analysis will miss events where either parental or shuffled genes have diverged beyond sequence recognition since two genomes shared a common ancestor. We thus emphasize that our analysis applies only to 'core' genomes: genes so well conserved that their homology even among distantly related species is beyond doubt. The incidence of shuffling among more rapidly evolving genes may be different and cannot be estimated with this approach. In this regard, we also note that our analysis cannot simply use multiple outgroups for a given test genome [20-24] to solve this problem, because doing so has the potential to misestimate shuffling rates by making wrong assumptions about the most parsimonious placement of such events (especially among prokaryotes, where horizontal transfer of shuffled genes may occur). For the remainder of our analysis, we chose ten distantly related genomes (Table 1) that best met the joint requirements of well known phylogenetic relationships and reliably annotated genome sequences (which is often problematic for the higher eukaryotes).
In addition to raising problems, the comparison of distantly related genomes also has one advantage: such genomes are more likely to be annotated independently from each other than are closely related genomes. In a group of closely related genomes, the first sequenced genome may often be used as a guidepost to annotate the other genomes, which may lead to errors (for instance, by misidentifying a shuffled region as an intron).
The number of shuffled genes we found is modest even for anciently diverged species pairs. For example, only 82 gene-shuffling events among the 5,800 genes considered (Table 2) may have been preserved in Caenorhabditis elegans since its common ancestor with Drosophila melanogaster, which lived around 600 Mya [33]. Similarly, only four surviving recombination events (out of 2,300 genes) may have occurred in the budding yeast Saccharomyces cerevisiae since its split from the fission yeast Schizosaccharomyces pombe more than 300 Mya [34]. We emphasize that all these numbers refer to shuffled genes that have 'survived': extant genome sequences alone are insufficient for estimating the frequency of the recombination events themselves, since the products of these events often will not become fixed in populations.
One further observation indicates the rarity of gene shuffling: most shuffled genes contain at least one domain of low sequence similarity to a parental gene. The above analysis is based on identifying sequence domains as homologous in a parental and recombined gene if they show more than 35% amino-acid sequence identity. Increasing this identity threshold to 40% can reduce the number of candidate shuffled genes dramatically (see Table 2). For instance, it removes 28 of 48 shuffled fruit fly genes and half of the shuffled fission yeast genes. This observation underscores that shuffling is rare among highly conserved genes: otherwise we would see higher sequence similarities among parental/recombined domain pairs.
Figure 2 shows representative examples of shuffled genes, illustrating some of the types of recombination diagrammed in Figure 1a. For example, Figure 2b shows the budding yeast his4 gene, which is involved in histidine biosynthesis. This (apparent fusion) gene appears to combine the functions of the two fission yeast genes his7 (a phosphoribosyl-AMP cyclohydrolase) and his2 (a histidinol dehydrogenase) [35]. Figure 2c shows the fruit fly gene Aats-tyr, a tyrosyl-tRNA synthetase [36]. This gene is a likely recombination product of a predicted worm methionyl-tRNA synthetase gene mrs-1 [37] and a second worm gene Y105E8A.19 of unknown function. A list of all shuffled genes identified in these ten genomes is available in Additional data file 1.
Gene shuffling and structural domains
Because our approach is based on sequence domains, we wished to find out whether the recombined regions of shuffled genes match structural protein domains. If so, this would indicate that successful recombination events - events preserved in the evolutionary record - occur mostly at structural domain boundaries. To address this question, we used the Pfam database [30,38] of protein domains to identify domains in our shuffled genes that were significant at E ≤ 10-5. These Pfam domains were compared to the sequence alignments that we used to identify shuffled genes in the first place. As Figure 3 shows, the boundaries of recombined sequence domains and Pfam structural domains tend to coincide (P < 0.001 using a domain randomization approach, see Materials and methods). However, Figure 3 also suggests that not all successful recombination events occur at structural domain boundaries. Experimental and computational work on individual proteins [27] supports the notion that successful recombination occurs preferentially, but not exclusively, at structural domain boundaries.
Gene shuffling and exon-intron structure
The exon-shuffling/introns-early hypothesis [39-41] predicts that exon-intron boundaries delimit functional domains and hence that recombination events that preserve exons would be more likely to yield functional recombinant proteins. Long introns also increase the probability of a DNA-level recombination event preserving exons (since in this case the number of possible DNA-level recombination events leading to the same recombinant protein may be quite large), a further reason to expect an association of shuffling boundaries and exon boundaries. The two multicellular eukaryotes (Drosophila and C. elegans) have a sufficient number of introns to allow us to test this prediction by comparing the boundaries of recombined sequence domains to the positions of introns in the sequences in question. However, contrary to these expectations, we found no tendency for our shuffling boundaries to associate with exon-intron boundaries (P > 0.1, domain randomization test; see Materials and methods).
The incidence of gene shuffling
We cannot estimate the incidence of gene shuffling in absolute (geological) time, because divergence dates for most of our test species are unknown or highly uncertain. In addition, the rarity of gene-shuffling events further complicates such estimates. However, we can obtain order-of-magnitude estimates of the incidence of gene shuffling relative to the incidence of other mutational events important in genome evolution. One such event is gene duplication, whose incidence has been estimated previously [10,11].
To compare the incidence of gene duplication to that of gene shuffling, we cannot rely on the silent nucleotide divergence among duplicate genes to estimate the rate of duplication, as is commonly done [10,11], because several of our study genomes are very distantly related. We thus estimated the rate at which gene duplications occurred in a test species T since its common ancestor with R1 using the following approach. We identified, for each test species T, all genes that had only a single homolog in the reference species R1. We denote the number of these reference species genes as g. Second, for each of these genes i we determined the number ni of test species genes homologous to gene i. If this number ni is greater than 1, then the test species homolog of gene i underwent one or more duplications since the common ancestor of T and R1. We estimated the (minimal) number of duplication events necessary to establish a gene family of size ni as . The total estimated number d of gene duplications for the g reference species genes then calculates as the sum . Values for g and d are shown for each reference species in Table 2. One can view the ratio d/g as the per-gene incidence of gene duplication.
We then used this ratio to estimate the ratio of gene-shuffling events per gene duplication event (Table 1). To do so, we first had to estimate the number of gene-shuffling events per gene, which we obtained by dividing the number s of gene-shuffling events in a test species T (Table 1) by the average number h of genes in T or R1 with detectable sequence similarity to genes in the other genome (Table 2). This approach of estimating the number of gene-shuffling events per gene compensates for the reduced ability to recognize gene homology in distantly related genomes. The ratio of shuffling events per duplication can then be calculated as (s/h)/(d/g). Figure 4a compares this ratio for the organisms we studied. The bacteria analyzed share with the archaeans a high incidence of gene shuffling relative to duplication, while the eukaryotes show a much lower incidence. The Bacillus species (B. anthracis and B. cereus) have a much higher relative incidence of gene shuffling than any other species pair we studied.
Other mutations useful to calibrate the incidence of gene shuffling are nonsynonymous (amino-acid replacement) and synonymous (silent) mutations on DNA. Synonymous substitutions are an indicator of divergence time between two genes or species because they are subject to few evolutionary constraints and thus may change at an approximately constant (neutral) rate [32]. We estimated the incidence of gene shuffling relative to synonymous substitutions by first determining the average fraction, Ks, of synonymous nucleotide changes per synonymous nucleotide site for 100 orthologous genes in a T-R1 species pair. We then simply divided the number of gene-shuffling events per gene (s/h) by this average Ks (Figure 4b). The evolutionary distance of two of our species pairs (E. coli vs Salmonella and B. anthracis vs B. cereus was sufficiently low to allow us to directly calculate the average synonymous divergence for 100 pairs of randomly selected single-copy orthologs in the test and reference species (see Materials and methods). For the other species pairs, most synonymous sites are saturated with substitutions [32]. In these cases, we thus extrapolated the value of Ks between R1 and T from that observed between either of these species and a third, closely related species (see Materials and methods for details). We emphasize that this procedure would be unsuitable to make evolutionary inferences for any one gene, because it introduces considerable uncertainty into our estimates. It is, however, adequate to identify the approximate, genome-wide patterns we are concerned with.
Finally, we also estimated, completely analogously, the number of gene-shuffling events per unit amino-acid divergence (Ka = 1). These results are summarized in Table 1 and Figure 4c. The incidence of gene shuffling relative to silent and amino-acid divergence varies less systematically among the domains of life than that of gene shuffling relative to duplication. However, it is again apparent from these analyses that successful gene shuffling is very rare for conserved genes. For some species, crude estimates of the absolute geological time needed for two sequences to accumulate a pairwise divergence of one silent nucleotide substitution per silent site are available. In the fruit fly this amount of time is approximately equal to 64 million years [32]. During this period of time, we would expect only 5,864 × 2.8 × 10-3 = 16 gene shuffling events to occur (Tables 1 and 2; 5,864 is the average number of fruit fly and worm genes in our core gene set). By way of comparison, even using our very conservative method of counting duplicate genes, we would expect 146 gene duplications in this period. Similarly, in the yeast S. cerevisiae, where Ks = 1 synonymous substitutions accumulate every 100 million years [42], one would expect three shuffling events during this period of time (2,365 × 1.3 × 10-3), as compared to 200 gene duplications. We emphasize that these are order-of-magnitude estimates that mainly serve to underscore the rarity of successful gene shuffling.
A multidomain protein may include both distinct and repeated structural domains [13]. Multidomain proteins with repeated domains raise a special problem for identifying gene-shuffling events: a shuffling event followed by domain duplication might lead us to miss a shuffled gene because our local alignments of that gene to its parental genes would only include one copy of the duplicated domain and hence might reveal less than the 50% of alignable amino-acid residues we require (see Materials and methods). To assess whether this problem would substantially bias our results, we examined candidate shuffled genes that had been excluded by our criterion (that is, those having between 10% and 50% of their residues alignable). We asked whether a failure to account for domain duplication was responsible for their exclusion. After adding potentially duplicated domains to the aligned regions of these genes, we found that only a handful of them (two genes in Drosophila and three in C. elegans) met our 50% alignability threshold. Failure to account for domain duplications internal to a gene is thus not the reason for our low estimates of the incidence of gene shuffling.
Several lines of evidence show that successful gene shuffling is very rare for genes conserved between the distantly related genomes we studied. For the single-celled yeasts - currently the only group of very closely related eukaryotes with sufficiently reliable genome annotation - shuffling appears rare in the genome as a whole. In most of the genomes we analyzed, gene shuffling is much rarer than other important kinds of mutations affecting gene structure, such as gene duplication. For example, in the time that it takes to accumulate Ks = 0.01 synonymous substitutions per synonymous site, other research indicates that ten fruit fly genes and 164 worm genes undergo duplication [10]. In contrast, each lineage has only a 50% chance of undergoing a successful gene shuffling event in the same amount of time (if one assumes our estimates can be applied to the entire genome). We note that our estimates of duplication rates are more conservative than those of others [10], partly because we limit ourselves to single-gene duplications. The fact that we still see a lower incidence of shuffling than duplication (Figure 4) is thus all the more remarkable.
Discussion
The rarity of gene shuffling relative to gene duplication has a simple potential explanation. A gene duplication creates a copy of a gene while preserving an original that is able to exercise its function. In contrast, unless a recombination event is accompanied by gene duplication, the original (parental) genes disappear in the event. An organism may survive a recombination event only if neither parental gene was essential to its survival and reproduction or if the recombinant gene(s) can carry out the function of both parental genes. This rarity of successful gene shuffling stands in stark contrast to the frequency of DNA recombination itself, which is a ubiquitous process accompanying DNA replication and repair. This suggests that the vast majority of recombined genes have deleterious effects on the organism, which may be particularly true for the highly conserved genes examined here.
We emphasize that the rarity of gene shuffling we find is not in contradiction with earlier studies that have identified multiple gene fusions - a special, simple case of gene shuffling - in fully sequenced genomes [20-24]. These studies identified prokaryotic gene fusion events in one test genome relative to multiple, often very distantly related, reference genomes. Any such approach may find many fusion events even if such events are rare. Our data also do not rule out the possibility that shuffling played an important role in forming the conserved eukaryotic core genome, because the pertinent gene-shuffling events would have occurred before the divergence of the eukaryotic species pairs we examined. (The identification of such ancient shuffling events may require an approach based on protein structure.)
Furthermore, our results are not in contradiction with anecdotal evidence for the abundance of gene-shuffling events in some functional categories of genes [32]. The reason is that our results pertain to the average incidence of gene shuffling among conserved genes. Some genes may be shuffled at a much greater rate. Indeed, structural studies of multidomain proteins tend to find a few domains which co-occur with a wide variety of other domains (indicating the common shuffling of such promiscuous domains), whereas many other domains co-occur with only one or a few other domains (rare shuffling) [43]. Similarly, a lack of reliable genome annotation made it impossible to reliably identify gene-shuffling events in vertebrate genomes, where gene shuffling may be more frequent overall [44].
Perhaps the central caveat to our results regards sources of ascertainment bias. The comparison of distantly related genomes alone introduces a powerful source of ascertainment bias: we can only analyze gene-shuffling events for genes that have been sufficiently conserved to be recognizable in both genomes. However, shuffling might be more common among rapidly evolving genes. An additional possible source of bias is that after a successful gene-shuffling event the rate of amino-acid substitutions may be elevated as a result of directional selection on the newly created gene. Such a bias would cause us to underestimate shuffling frequencies in distantly related species even for conserved genes. Nonetheless, our results from the four closely related genomes argue against such a bias, because shuffling also appears rare in these genomes.
Another caveat is that our ability to identify successful gene-shuffling events depends on the continued presence of both parental genes in the reference genome. Genomes, however, occasionally lose genes. For instance, recent work has suggested that S. cerevisiae has lost roughly 10% of its genes since its last common ancestor with S. pombe [45]. If gene loss in other organisms occurs at comparable rates, our approach may slightly underestimate the number of recombination events in a lineage. However, note that gene loss affects our estimates of gene shuffling and gene duplication in similar ways, thus compensating for any such bias.
We used a second reference genome R2 to be able to exclude gene-fission events in reference genome R1. Such events can lead to misidentification of recombination in the test species and have been documented in several organisms [20]. Unfortunately, this approach fails if the same recombination event occurred twice, once in the lineage leading to reference species R2 and once in the lineage leading to test species T. Such a case of parallel evolution or homoplasy would lead us to misidentify a recombination event in T as a gene-fission event in R1. However, because successful gene shuffling is very rare in general, and because the required recombination event would have to occur at exactly the same position twice, this possibility is probably not a major confounding factor in our analysis.
A fourth caveat lies in the possibility that some of our recombinant genes may result from two independent recombination events. Our algorithm can identify such genes, but given the high sequence divergence of recombinant domains it may often be impossible to resolve the order of the individual recombination events. The generally small number of recombinant proteins implies that genes produced by two or more recombination events would be extremely rare. Indeed, among 203 identified recombinant genes, a mere 16 show matches to more than two parental genes, making these the only cases with indications that more than one recombination process was involved in their creation.
Finally, our approach to estimating the rate of gene duplication identifies only duplications of single-copy genes in the reference species. Multicopy genes may undergo duplication more frequently. We may thus have underestimated the number of gene duplications. As a result, the incidence of gene shuffling relative to gene duplication may be even lower than indicated by our estimates.
Recombination and introns
Our findings speak to a long-standing debate in molecular evolution, a debate that revolves around the origin of introns. Introns are stretches of DNA that do not code for proteins and that separate exons, the protein-coding regions of genes. According to one point of view, introns originated early in the evolution of life, perhaps as early as the common ancestor of prokaryotes and eukaryotes [39-41]. According to this perspective, introns may have acted as spacers between exons and thus greatly facilitated recombination among exons to create new proteins. The opposite point of view is that introns arose late in life's evolution, perhaps as late as eukaryotes themselves [28,29] and thus had no role in gene shuffling earlier in life's history. Genes in two of our test genomes have a sufficient number of introns to test the hypothesis that introns facilitate gene shuffling. Neither of these genomes showed an association between gene-shuffling boundaries and exon position. In addition, neither of these genomes showed an elevated incidence of gene shuffling. Although based on a small number of genomes, this finding casts doubt on the importance of introns for gene shuffling, and it suggests that other aspects of genome architecture may be more important. One potential example is the organization of functionally related prokaryotic genes into operons. The close proximity of such genes may facilitate their reorganization and the generation of new functions, whether through simple fusion or fission or through more radical change.
Natural selection or drift?
A nonhomologous recombination event that gives rise to a shuffled gene occurs in only one individual of a potentially large population. Does a shuffled gene typically rise to high frequency and become fixed through natural selection or genetic drift? To answer this question, one could in principle study the relationship between the rate at which fixed shuffled genes arise and population size (taking account of differences in nonhomologous recombination rates among species). Three possibilities exist in principle. First, there may be no relation between population size and the rate at which fixed shuffled genes arise. This would be the case if most gene-shuffling events are strictly neutral [46] or if they have very large beneficial effects. Second, there may be a positive relation between the rate at which fixed shuffled genes arise and population size. This would be the case if most shuffling events are mildly beneficial. The reason is that in this case selection favoring the fixation of a shuffled gene has to overcome the effects of genetic drift, which are weakest in large populations. Finally and perhaps most likely, there may be a negative association between population size and the rate at which fixed shuffled genes arise. This would be the case if most shuffling events are mildly deleterious. Such a negative association has been observed for several indicators of genome structure such as genome size, transposable element load, and rates of preservation of duplicated material [47].
Unfortunately, insufficient data are available to distinguish rigorously between these possibilities. First, estimates of effective population sizes Ne, based on estimates of Neμ [47] and the mutation rate μ [9], exist only for three of our five pairs of genomes (E. coli-Salmonella, S. cerevisiae-S. pombe and D. melanogaster-C. elegans). Second, we have insufficient information on recombination rates (whose variation among genomes needs to be taken into account). Specifically, although estimates of homologous recombination rates are available for a few of our organisms [48-51], gene shuffling occurs strictly by nonhomologous recombination, whose rate need not have a simple relationship with the homologous recombination rate. A third difficulty is that recombination rates and mutation rates are conventionally measured per cycle of DNA replication, whereas we would require per-year estimates as well as estimates of absolute divergence times between our taxa of interest to make appropriate comparisons.
Despite such insufficient data, we can make the qualitative observation that the observed incidence of shuffling does not follow a simple pattern: For example, S. cerevisiae has a relatively high effective population size (Neμ is approximately half of that for E. coli [47] while μ is actually higher than that of E. coli [9]) and a high homologous recombination rate compared to C. elegans or E. coli [48-51], and yet it shows the lowest incident of gene shuffling of any of our taxa. In the slightly deleterious scenario above, we would instead expect yeast to show an incidence of shuffling greater than that of E. coli, while in the slightly beneficial scenario we would expect it to show an incidence greater than that of the multicellular eukaryotes.
A second qualitative observation is that the incidence of gene shuffling is not elevated in higher organisms relative to the rate of nucleotide substitutions. (The higher incidence of gene shuffling relative to gene duplication in prokaryotes from Figure 4a may be a consequence of the lower rate of gene duplication in these taxa.) This is consistent with the hypothesis that the fate of most shuffled genes is driven by natural selection rather than genetic drift. In other words, most shuffling events may not be neutral. This is again plausible if one considers that most gene-shuffling events change a gene's structure drastically. A corollary of this hypothesis is that preserved shuffled genes have been preserved for a reason - the benefit they confer to an organism. While rare in number, shuffled genes may thus be of great importance in organismal evolution.
Our analysis of gene shuffling has left many open questions, most notably about the association between the rate of sequence evolution and the rate of gene shuffling. To arrive at firm answers for this and other questions, we must be able to study shuffling rates not only for conserved proteins but also for rapidly evolving proteins. Such studies will require closely related genome sequences with reliable gene identification derived independently for each genome.
Materials and methods
Identifying shuffled genes
Our method identifies shuffled genes in a test genome (T) relative to a reference genome (R1). Table 1 shows the ten test genomes - two archaeal, six prokaryotic, and four eukaryotic genomes - we used in this analysis. Every pair of genomes R1-T occurs twice in Table 1, because one of two genomes can be used either as the test or the reference genome. To exclude spurious recombination events that reflect gene loss or fission in R1, the method also employs a second reference genome, R2. The two archaeans in our analysis were Pyrococcus horikoshii [52] and Methanocaldococcus jannaschii [53]. The R2 species for these archaeans was Archaeoglobus fulgidus [54]. The bacterial genomes we analyzed were those of Escherichia coli [55], Salmonella enterica [56], Bacillus anthracis [57], and Bacillus cereus [58]. The reference species R2 were Bacillus subtilis [59] for the B. anthracis-B. cereus comparison and Haemophilus influenzae [60] for the E. coli-Salmonella comparison. Our four eukaryotic genomes were budding yeast Saccharomyces cerevisiae [61], fission yeast Schizosaccharomyces pombe [62], nematode worm Caenorhabditis elegans [63] and fruit fly Drosophila melanogaster [64]. We used the genome of Neurospora crassa [65] as the R2 genome for all these eukaryotes.
To identify sequence homology between all genes in these genomes we used the Washington University implementation of gapped BLASTP [66,67], followed by exact pairwise local alignment using the Smith-Waterman algorithm [68] with a gap-opening penalty of 10 and a gap-extension penalty of 2, and the BLOSUM 62 scoring matrix [69]. We excluded from further analysis all gene pairs with BLAST E-values greater than 10-6, fewer than 50 aligned amino acids, amino-acid identity in the alignment of less than 35%, or alignments consisting of more than 50% low-complexity sequences as determined by the SEG program [70,71].
The requirement of 35% sequence identity may appear to bias our estimates of shuffling incidence between distantly related taxa. However, because we calculate these values relative to the total number of genes with the same (35%) degree of sequence identity between the test and reference genome (h), this bias is most likely to be small.
The result of this procedure is a list of partially or fully matching genes in the two species T and R1. We used this list to identify shuffled genes in the test genome T. Specifically, for each gene in the test genome T we searched for pairs of genes in the reference genome R1 that matched the test species gene, but in nonoverlapping or minimally overlapping regions. (To account for edge errors in local alignments, we allowed regions to overlap by a maximum of 20 residues). After having identified any such gene, we verified that it did not also have full-length homologs in the reference genome, because otherwise gene shuffling would not be the most parsimonious explanation of the gene's origin. We developed a special-purpose algorithm for this search [72], which identifies, for any one gene, the combination of local alignments to genes in the reference genome that covers the maximum number of residues in the shuffled gene. This algorithm can identify shuffled genes (genes to which two or more reference species genes contributed), but it will also return only a single alignment if this alignment is longer than any combination of non-overlapping alignments.
Three criteria for validating shuffling events
We used three additional criteria to validate candidates for shuffled genes. First, we computed the proportion of a shuffled gene's amino-acid residues that could be aligned to its (parental) reference species genes. If this proportion is small, a gene may be too highly diverged for us to confidently ascertain that it is a recombination product. We excluded genes where this proportion was smaller than 50%. This requirement may appear restrictive, but additional analyses show that our conclusions hold even if it is completely eliminated. For example, eliminating this criterion increases the number of shuffled genes by a factor ranging from 1 (no increase, E. coli) to 4.2 (C. elegans), but the eukaryotes surveyed still show an incidence of shuffling smaller than the duplication rate, while the prokaryotes show similar frequencies of shuffling and duplication. We have maintained the 50% requirement throughout our main analysis to err on the side of caution: Putative shuffled genes with very short alignable regions to a parental gene are more likely to be false positives. They also do not belong in the set of genes conserved between T and R1, which is our focus here.
To motivate our second validation criterion, we note that in the eukaryotic test genomes some shuffled genes had undergone duplication. We identified gene duplicates as gene pairs with amino-acid divergences Ka< 1 using a previously described and publicly available tool [73]. We counted each gene family of shuffled genes only once to avoid double-counting duplicates of shuffled genes.
A third indicator of true recombination is the divergence of different sequence domains within a putative shuffled gene. The recombined parts of a shuffled gene should show equal sequence divergence from their respective parental genes, if these parts have diverged in a clock-like fashion. The two principal indicators of DNA sequence divergence are the number of silent nucleotide substitutions at synonymous sites (Ks) and the number of non-synonymous substitutions at amino-acid replacement sites (Ka) [32]. We used the methods of Muse and Gaut and Goldman and Yang [74,75] to estimate these divergence indicators for our putative shuffled genes. Silent substitutions are subject to much weaker selection pressures than amino-acid replacement substitutions, and their rate of accumulation is thus more clock-like [32]. For our analysis of four closely related yeast species, we calculated the distance (Ka or Ks) between each sequence domain of a shuffled gene and its counterpart parental gene. We excluded candidate shuffled genes from further analysis if these distances (Ka or Ks) differed by more than a factor two between the different domains. For our highly divergent species (Table 1) we could not use synonymous divergence Ks because most domain pairs had become saturated with silent substitutions. The (unavoidable) disadvantage of using only amino-acid divergence Ka is that natural selection on amino-acid changes can cause non-clock-like evolution of domains. However, even with this caveat, our approach identifies true recombination events. Otherwise, our shuffled genes would not show the highly significant statistical association in amino-acid divergence Ka we observe between sequence domain pairs (Figure 5; Pearson's r 0.47, Spearman's s 0.45, P < 0.0001 for both). Note that we did not exclude shuffled genes from our analysis of distantly related genomes based on unequal Ka estimates.
Detecting domain duplication in putative shuffled genes
Internal domain duplications could cause a shuffled gene to fail the first of the three test criteria above - 50% alignable residues with its parental genes - which would cause us to miss such genes. To assess how serious a problem such missed genes might be, we relaxed the first of the above test criteria, requiring only 10% alignable residues between a test gene and two or more potential parental genes. For test species genes that met this criterion, we then excised the portion of the test species gene that aligned to a reference gene, and recomputed a local alignment between this trimmed gene and the reference gene. If the test-species gene contained internal duplications of the reference gene, the trimmed sequence should still align to the reference gene. We added any new alignments found in this way to the original alignment combination and assessed whether the resulting combination of alignments met the required threshold of 50% alignable residues. For alignments that met our other prescreening criteria for candidate shuffled genes (50 amino acids in length and > 35% amino-acid identity) we iterated this excision procedure to determine the total number of repeated domains. This analysis identified only five additional shuffled genes (see above) and led us to conclude that internal duplications are not a major confounding factor in our analysis.
Estimating relative frequencies of gene shuffling
We estimated the incidence of gene shuffling relative to two other frequent kinds of evolutionary events - gene duplications and nucleotide substitutions. To do so, we first needed to account for differences in genome size among our study species. We thus estimated the number of gene shuffling events per gene. This estimate poses a problem that stems from the different (and highly uncertain) times since common ancestry of our different T-R1 species pairs. The longer the time since common ancestry, the fewer genes (shuffled or not) with recognizable sequence similarity two species will share. Thus, when simply dividing the number of shuffled genes s by the total number of genes in a test genome, one may wrongly estimate the number of gene-shuffling events per gene. To account for this problem, we divided the total number of gene-shuffling events s by the number of recognizable homologs h shared between species T and R1 to obtain the number of gene shuffling events per gene, s/h. To obtain h itself, we determined the total number of genes in T with at least one homolog - using the criteria outlined earlier - in the genome of R1, the total number of genes in R1 with at least one homolog in T, and averaged these two numbers.
We then related this number of gene shuffling events per gene to the number of gene duplications in T since its common ancestor with R1. Because our test and reference genomes are only distantly related, we could not rely on the silent nucleotide divergence among duplicate genes to estimate the rate of duplication, as is commonly done [10,11]. We thus used the following, alternative, method. First, for each test species T, we identified all genes which match only a single gene in R1 at a BLAST threshold of 10-6 and had 70% or more of their sequences aligned to that gene. The number of unique reference species genes matching one or more test species genes gives a baseline number g of genes before duplication. Second, for each such unique reference species gene i we identified duplicate pairs of test species genes that showed a pairwise amino-acid divergence Ka which was less than either gene's amino-acid divergence from the putative ortholog i. For many genes i, more than one pair of test species genes fulfilled these criteria. Such genes represent families of ni duplicates of gene i. We estimated the (minimal) number of duplication events necessary to establish such a gene family of size ni as . The total estimated number d of gene duplications for the g reference species genes then calculates as the sum . Genes in very large families are more likely to undergo duplication than genes in smaller families [76]. For this reason, we excluded reference species genes i with more than ten duplicate genes from this analysis. Including such genes would tend to increase our estimated rate of gene duplication, meaning that the results shown in Figure 4a are conservative estimates of the excess of duplicates relative to shuffled genes. To estimate the rate of gene shuffling events relative to gene duplication events, we then divided the number of gene shuffling events per gene (s/h, obtained above) by the number d/g of gene duplication events per gene.
We also estimated the number of gene-shuffling events per unit of silent substitutions Ks that accumulate in a gene. Two of our species pairs (B. anthracis-B. cereus and E. coli-S. enterica) allowed us to estimate synonymous divergence Ks directly. For these two species pairs, we first identified 100 pairs of single-copy genes in each genome that are unambiguous orthologs [32,77]. We then divided the number of shuffled genes per gene by the average synonymous divergence Ks of the orthologs with unsaturated synonymous divergence (> 97 for both species pairs) to obtain an estimate of the number of gene-shuffling events per unit change in Ks.
For the three other genome pairs, this approach was not feasible because of their mostly saturated synonymous divergence. We thus had to estimate the average synonymous divergence Ks between T and R1 by extrapolating from the synonymous divergence between T and a more closely related species. This approach relies on previous work [10] which indicates that the genome-wide average ratio Ka/Ks of amino-acid divergence to synonymous divergence approaches an asymptotic value for large numbers of distantly related genes. For each of the three T-R1 species pairs, our approach to estimating Ks in this way consists of two steps. We first estimated the average amino-acid divergence Ka between 100 randomly chosen unique single-copy orthologs of a T-R1 species pair. For the second step, we first identified an organism C with fully sequenced genome that is closely related to either T or R1. The genome of C should be sufficiently closely related to estimate Ks reliably, but sufficiently distantly related to reliably estimate the asymptotic ratio of amino-acid to silent divergence. This organism was Saccharomyces paradoxus for S. cerevisiae; Pyrococcus furiosus [78] for P. horikoshii; and Caenorhabditis briggsae [79] for C. elegans. For each of these closely related genome pairs, we chose at random 100 single-copy gene pairs that were unambiguous orthologs. We calculated the average ratio Kac/Ksc for these orthologs. We then used this value to extrapolate the average fraction Ks of synonymous substitutions between genes in T and R1 as Ks = Ka/(Kac/Ksc). This is the estimated average synonymous substitution rate between a T-R1 species pair. We then related the rate of gene shuffling to this extrapolation of Ks. Specifically, we estimated the number of shuffling events per gene per one Ks as (s/h)/(Ks/2). (The reason for dividing the average Ks by 2 is that our approach estimates the number of gene-shuffling events only for one of the two species of a T-R1 species pair.) We are well aware of the shortcomings of this approach, which averages heterogeneous substitution rates and assumes that the ratio of amino acid to silent divergence is constant within the taxonomic group considered. However, we emphasize that we use the approach here only to arrive at order-of-magnitude estimates of the incidence of gene shuffling. We also note that although we have not explicitly taken codon usage bias into account, the use of codon position-specific nucleotide frequencies (which partially account for such a bias, at a cost of larger estimate variances) increased all of our estimated average Ks values without changing the patterns seen in Figure 4. Thus, the values in Figure 4 are conservative in the sense that the actual incidence of shuffling relative to Ks may be lower than shown.
Third, and finally, we also estimated the number of gene-shuffling events per unit of amino-acid replacement substitutions Ka that accumulate in a gene. To do so, we divided s/h as above by one-half the average amino-acid divergence (Ka/2) of 100 unambiguous orthologs in a T-R1 species pair.
Comparison of identified domains to Pfam database
Because our analysis was sequence and not structure-based, we used the Pfam database of structural protein domains [30,38] to evaluate how well the recombined sequence domains we identified matched structural domains. To do so, we queried all shuffled genes against the Pfam database and retained identified Pfam domains with E ≤ 10-5. We then compared the location of these structural domains to the location of the sequence domains in the shuffled genes. Specifically, we calculated the distance between each alignment domain and its closest Pfam domain. For the starting (As) and ending positions (Ae) of the alignment, and for the starting (Ps) and ending positions (Ps) of the Pfam domains found in the shuffled gene, we calculated the quantity . If D is very small, then the sequence domain and the Pfam structural domain overlap to a large extent. Parametric tests are not appropriate to evaluate the statistical significance of this association. We thus applied a gene-randomization procedure that created new alignment domains within each gene, domains whose starting and ending positions are uniformly distributed but cover the same proportion of the gene as did the original domains. Each randomized gene possessed the same number of simulated domains as observed domains, but with different positions and lengths. We then calculated D for the simulated alignment domains and compared its distribution to the empirically observed values of D. We applied an analogous approach to test whether exon/intron boundaries and shuffling boundaries are associated in our two multicellular eukaryotes (C. elegans and Drosophila). This approach substituted the closest exon for the closest Pfam domain and applied the same randomization procedure.
Additional data files
Additional data file 1, available with the online version of this paper, contains a table listing all shuffled genes included in our analysis of the ten distantly related genomes.
Supplementary Material
Additional File 1
A table listing all shuffled genes included in our analysis of the ten distantly related genomes. A table listing all shuffled genes included in our analysis of the ten distantly related genomes.
Click here for file
Acknowledgements
G.C.C. would like to thank the Department of Energy Computational Science Graduate Fellowship Program of the Office of Scientific Computing and Office of Defense Programs in the Department of Energy under contract DE-FG02-97ER25308, the Bioinformatics Initiative of the Deutsche Forschungsgemeinschaft (DFG), grant BIZ-6/1-2, and Science Foundation Ireland for financial support. A.W. would like to thank the National Institutes of Health for its support through NIH grant GM063882-01 to the University of New Mexico, as well as the Santa Fe Institute and the Institut des Hautes Etudes Scientifique (IHES) for continued support.
Figures and Tables
Figure 1 Identifying gene shuffling. (a) Gene shuffling and how it changes gene structure. The three scenarios of 'domain insertion' represent insertions of domains from gene 2 into gene 1. The reciprocal insertions (gene 1 into gene 2) are not shown. (b) Distinguishing true from spurious recombination events. In a spurious recombination event, reference genome R1 has two separate genes, where both T and R2 have a single, shuffled gene. The most parsimonious explanation for this observation is that the shuffled gene was present in R1 but was lost since R1's divergence from T.
Figure 2 Representative examples of shuffled genes identified. (a) Bacillus anthracis M23/M37 peptidase BA1903, the result of a domain exchange between B. cereus genes BC5234 (12098), a N-acetylmuramoyl-L-alanine amidase and BC1480(08460.1), another M23/M37 peptidase. (b) A fusion of the fission yeast genes his7 (a phosphoribosyl-AMP cyclohydrolase) and his2 (a histidinol dehydrogenase) to produce the budding yeast his4 gene, which is involved in histidine biosynthesis. The budding yeast gene appears to combine the functions of the two fission yeast genes [35]. (c) The fruit fly gene Aats-tyr is a tyrosyl-tRNA synthetase (Flybase annotation) [36]. It is a probable recombination product of a predicted worm methionyl-tRNA synthetase gene mrs-1 (WormBase annotation) [37] and a second worm gene Y105E8A.19 of unknown function. (d) C. elegans gene ceh-20, which encodes a homeodomain protein. This gene appears to be the result of a domain exchange between the Drosophila genes exd (extradenticle, also a homeodomain protein) and Pkg21D (cGMP-dependant protein kinase). (e) E. coli b4343, a hypothetical protein apparently formed via a domain exchange between Salmonella genes STY4850 (annotated as a DEAD-box helicase-related protein) and STY4851 (hypothetical protein). The numbers in the recombinant gene box are amino-acid positions in the protein product, indicating the portion of the protein derived from each of its 'parental' proteins.
Figure 3 Association between recombined sequence domains and Pfam structural domains. The horizontal axis shows the starting and ending positions of the sequence domains in recombined genes (in amino acids, relative to the translation start site of the gene). The vertical axis shows the starting and ending positions of the Pfam domain closest to each recombined sequence domain.
Figure 4 Incidence of gene shuffling relative to various other mutational events. (a) Gene duplication, (b) silent nucleotide substitutions, and (c) amino-acid changing nucleotide substitutions for the species pairs indicated on the horizontal axis. Note the scale breaks on the vertical axes.
Figure 5 Similarity in sequence divergence between regions of shuffled genes. The amino-acid divergences (Ka) of recombined domains to their respective parental counterparts are correlated. One outlying observation (Ka1 = 3.26 and Ka2 = 0.36) is not shown in this plot but was included in the calculation of correlation coefficients. Excluding this observation increases the Pearson correlation coefficient to 0.61 and leaves the Spearman correlation coefficient unchanged (P < 0.0001 for both).
Table 1 Relative abundances of shuffled genes
Organism (T) Reference taxa (R1) Shuffled genes (s) (35%) Shuffling events/duplication Shuffling events/gene/Ks unit Shuffling events/gene/Ka unit
M. jannaschii P. horikoshii 7 0.63 2.7 × 10-3 4.8 × 10-2
P. horikoshii M. jannaschii 7 0.95
B. anthracis B. cereus 21 4.33 4.1 × 10-2 0.92
B. cereus B. anthracis 20 3.16
E. coli S. enterica 1 0.69 1.9 × 10-3 3.0 × 10-2
S. enterica E. coli 5 0.37
S. cerevisiae S. pombe 4 0.015 1.3 × 10-3 1.0 × 10-2
S. pombe S. cerevisiae 8 0.13
D. melanogaster C. elegans 48 0.11 2.8 × 10-3 4.3 × 10-2
C. elegans D. melanogaster 82 0.16
Table 2 Estimating the incidence of gene shuffling
Organism (T) Reference taxa 1 (R1) Reference taxa 2 (R2) Shuffled genes (40%) Sequences with detectable homology (h) Number of duplicates/R1 genes tested Duplication Rate (d/g) Average Ks* Average Ka
M. jannaschii P. horikoshii A. fulgidus 1 661 7/418 0.017 7.7 0.44
P. horikoshii M. jannaschii A. fulgidus 2 661 5/449 0.011 7.7 0.44
B. anthracis B. cereus B. subtilis 17 4,155.5 3/2568 0.0012 0.24 0.01
B. cereus B. anthracis B. subtilis 19 4,155.5 4/2624 0.0015 0.24 0.01
E. coli S. enterica H. influenzae 1 3,183.5 1/2182 0.0005 0.98 0.06
S. enterica E. coli H. influenzae 2 3,183.5 9/2140 0.0042 0.98 0.06
S. cerevisiae S. pombe N. crassa 3 2,365 104/946 0.110 3.9 0.50
S. pombe S. cerevisiae N. crassa 3 2,365 25/955 0.026 3.9 0.50
D. melanogaster C. elegans N. crassa 20 5,864 74/1008 0.073 8.0 0.52
C. elegans D. melanogaster N. crassa 34 5,864 98/1120 0.088 8.0 0.52
*Note that to obtain shuffling events/per gene/Ks = 1.0 (Table 1) we divided the average Ks by 2.
This was done because Ks is a pairwise distance, meaning that it gives the sum of the divergences from the common ancestor to T and from the common ancestor to R1. The same was done for the Ka analysis.
==== Refs
Force A Lynch M Pickett FB Amores A Yan Y Postlethwait J Preservation of duplicate genes by complementary, degenerative mutations. Genetics 1999 151 1531 1545 10101175
Lynch M Force A The probability of duplicate gene preservation by subfunctionalization. Genetics 2000 154 459 473 10629003
Katju V Lynch M The structure and early evolution of recently arisen gene duplicates in the Caenorhabditis elegans genome. Genetics 2003 165 1793 1803 14704166
Long MY Langley CH Natural selection and the origin of jingwei, a chimeric processed functional gene in Drosophila. Science 1993 260 91 95 7682012
Yun S-H Berbee ML Yoder OC Turggeon BG Evolution of the fungal self-fertile reproductive life style from self-sterile ancestors. Proc Natl Acad Sci USA 1999 96 5592 5597 10318929 10.1073/pnas.96.10.5592
Lundin L Gene duplications in early metazoan evolution. Semin Cell Dev Biol 1999 10 523 530 10597636 10.1006/scdb.1999.0333
Powell SK Kaloss MA Pinkstaff A McKee R Burimski I Pensiero M Otto E Stemmer WPC Soong N-W Breeding of retroviruses by DNA shuffling for improved stability and processing yields. Nat Biotechnol 2000 18 1279 1282 11101807 10.1038/82391
Leong SR Chang JCC Ong R Dawes G Stemmer WPC Punnonen J Optimized expression and specific activity of IL-12 by directed molecular evolution. Proc Natl Acad Sci USA 2003 100 1163 1168 12529500 10.1073/pnas.0237327100
Drake JW Charlesworth B Charlesworth D Crow JF Rates of spontaneous mutation. Genetics 1998 148 1667 1686 9560386
Lynch M Conery JS The evolutionary fate and consequences of duplicate genes. Science 2000 290 1151 1155 11073452 10.1126/science.290.5494.1151
Gu Z Cavalcanti A Chen F-C Bouman P Li W-H Extent of gene duplication in the genomes of Drosophila, nematode, and yeast. Mol Biol Evol 2002 19 256 262 11861885
Teichmann SA Park J Chothia C Structural assignments to the Mycoplasma genitalium proteins show extensive gene duplications and domain rearrangements. Proc Natl Acad Sci USA 1998 95 14658 14663 9843945 10.1073/pnas.95.25.14658
Apic G Gough J Teichmann SA Domain combinations in archaeal, eubacterial and eukaryotic proteomes. J Mol Biol 2001 310 311 325 11428892 10.1006/jmbi.2001.4776
Bashton M Chothia C The geometry of domain combinations in proteins. J Mol Biol 2002 315 927 939 11812158 10.1006/jmbi.2001.5288
Li W-H Gu Z Wang H Nekrutenko A Evolutionary analyses of the human genome. Nature 2001 409 847 849 11237007 10.1038/35057039
Koonin EV Wolf YI Karev GP The structure of the protein universe and genome evolution. Nature 2002 420 218 223 12432406 10.1038/nature01256
Wuchty S Scale-free behavior in protein domain networks. Mol Biol Evol 2001 18 1694 1702 11504849
Wolf YI Brenner SE Bash PA Koonin EV Distribution of protein folds in the three superkingdoms of life. Genome Res 1999 9 17 26 9927481
Henikoff S Greene EA Pietrokovski S Bork P Attwood TK Hood L Gene families: the taxonomy of protein paralogs and chimeras. Science 1997 278 609 614 9381171 10.1126/science.278.5338.609
Snel B Bork P Huynen M Genome evolution: gene fusion versus gene fission. Trends Genet 2000 16 9 11 10637623 10.1016/S0168-9525(99)01924-1
Snel B Bork P Huynen M The identification of functional modules from the genomic association of genes. Proc Natl Acad Sci USA 2002 99 5890 5895 11983890 10.1073/pnas.092632599
Enright AJ Iliopoulos I Kyrpides NC Ouzounis CA Protein interaction maps for complete genomes based on gene fusion events. Nature 1999 402 86 90 10573422 10.1038/47056
Marcotte EM Pellegrini M Ng H-L Rice DW Yeates TO Eisenberg D Detecting protein function and protein-protein interactions from genome sequences. Science 1999 285 751 753 10427000 10.1126/science.285.5428.751
Marcotte EM Pellegrini M Thompson MJ Yeates TO Eisenberg D A combined algorithm for genome-wide prediction of protein function. Nature 1999 402 83 86 10573421 10.1038/47048
Rost B Protein structures sustain evolutionary drift. Fold Des 1997 2 S19 S24 9218962
Todd AE Orengo CA Thornton JM Evolution of protein function, from a structural perspective. Curr Opin Chem Biol 1999 3 548 556 10508675 10.1016/S1367-5931(99)00007-1
Voigt CA Martinez C Wang Z-G Mayo SL Arnold FH Protein building blocks preserved by recombination. Nat Struct Biol 2002 9 553 558 12042875
Doolittle WF Genes in pieces: Were they ever together? Nature 1978 272 581 582
Stolzfus A Spencer DF Zuker M Logsdon JM Doolittle WF Testing the exon theory of genes: the evidence from protein structure. Science 1994 265 202 207 8023140
Bateman A Birney E Cerruti L Durbin R Etwiller L Eddy SR Griffiths-Jones S Howe KL Marshall M Sonnhammer EL The Pfam Protein Families Database. Nucleic Acid Res 2002 30 276 280 11752314 10.1093/nar/30.1.276
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 10.1038/nature01644
Li W-H Molecular Evolution 1997 Sunderland, MA: Sinauer
Doolittle RF Feng DF Tsang S Cho G Little E Determining divergence times of the major kingdoms of living organisms with a protein clock. Science 1996 271 470 477 8560259
Sipiczki M Where does fission yeast sit on the tree of life? Genome Biol 2000 1 reviews1011.1 1011.4 11178233 10.1186/gb-2000-1-2-reviews1011
The S. pombe Genome Project
The FlyBase Consortium The FlyBase database of the Drosophila genome projects and community literature. Nucleic Acids Res 2002 30 106 108 11752267 10.1093/nar/30.1.106
Stein L Sternberg P Durbin R Thierry-Mieg J Spieth J WormBase: network access to the genome and biology of Caenorhabditis elegans. Nucleic Acids Res 2001 29 82 86 11125056 10.1093/nar/29.1.82
Protein Families Database of alignments and HMMs
Gilbert W Why genes in pieces? Nature 1978 271 501 622185
Blake CCF Exons - present from the beginning? Nature 1983 306 535 537 6646232 10.1038/306535a0
Dorit RL Schoenbach L Gilbert W How big is the universe of exons? Science 1990 250 1377 1382 2255907
Wagner A How large protein interaction networks evolve. Proc R Soc Lond, Ser B 2003 270 457 466 10.1098/rspb.2002.2269
Bornberg-Bauer E Beaussart F Kummerfeldy S Teichmann S Weiner J 3rd The evolution of domain arrangements in proteins and interaction networks. Cell Mol Life Sci 2005 62 435 445 15719170 10.1007/s00018-004-4416-1
Eichler EE Recent duplication, domain accretion and the dynamic mutation of the human genome. Trends Genet 2001 17 661 669 11672867 10.1016/S0168-9525(01)02492-1
Aravind L Watanabe H Lipman DJ Koonin EV Lineage-specific loss and divergence of functionally linked genes in eukaryotes. Proc Natl Acad Sci USA 2000 97 11319 11324 11016957 10.1073/pnas.200346997
Kimura M The Neutral Theory of Molecular Evolution 1983 Cambridge, UK: Cambridge University Press
Lynch M Conery JS The origins of genome complexity. Science 2003 302 1401 1404 14631042 10.1126/science.1089370
Hedrick PW Thomson G A two-locus neutrality test: applications to humans, E. coli and lodgepole pine. Genetics 1986 112 135 156 3510942
Cherry JM Ball C Weng S Juvik G Schmidt R Adler C Dunn B Dwight S Riles LM Mortimer RK Botstein D Genetic and physical maps of Saccharomyces cerevisiae. Nature 1997 387:(6632 Suppl) 67 73 9169866
Cherry JM Adler C Ball C Chervitz SA Dwight SS Hester ET Jia Y Juvik G Roe T Schroeder M SGD: Saccharomyces Genome Database. Nucleic Acids Res 1998 26 73 80 9399804 10.1093/nar/26.1.73
Barnes TM Kohara Y Coulson A Hekimi S Meotic recombination, noncoding DNA and genome organization in Caenorhabditis elegans. Genetics 1995 141 159 179 8536965
Kawarabayasi Y Sawada M Horikawa H Haikawa Y Hino Y Yamamoto S Sekine M Baba S Kosugi H Hosoyama A Complete sequence and gene organization of the genome of a hyper-thermophilic archaebacterium, Pyrococcus horikoshii OT3. DNA Res 1998 5 55 76 9679194
Bult CJ White O Olsen GJ Zhou LX Fleischmann RD Sutton GG Blake JA Fitzgerald LM Clayton RA Gocayne JD Complete genome sequence of the methanogenic archaeon: Methanococcus jannaschii. Science 1996 273 1058 1073 8688087
Klenk HP Clayton RA Tomb JF White O Nelson KE Ketchum KA Dodson RJ Gwinn M Hickey EK Peterson JD The complete genome sequence of the hyperthermophilic, sulfate-reducing archaeon Archaeoglobus fulgidus. Nature 1997 390 364 370 9389475 10.1038/37052
Blattner FR Plunkett G Bloch CA Perna NT Burland V Riley M Collado-vides J Glasner JD Rode CK Mayhew GF The complete genome sequence of Escherichia-Coli K-12. Science 1997 277 1453 1462 9278503 10.1126/science.277.5331.1453
Parkhill J Dougan G James KD Thomson NR Pickard D Wain J Churcher C Mungall KL Bentley SD Holden MT Complete genome sequence of a multiple drug resistant Salmonella enterica serovar Typhi CT18. Nature 2001 413 848 852 11677608 10.1038/35101607
Read T Peterson S Tourasse N Baillie L Paulsen I Nelson K Tettelin H Fouts D Eisen J Gill S The genome sequence of Bacillus anthracis Ames and comparison to closely related bacteria. Nature 2003 423 81 86 12721629 10.1038/nature01586
Ivanova N Sorokin A Anderson I Galleron N Candelon B Kapatral V Bhattacharyya A Reznik G Mikhailova N Lapidus A Genome sequence of Bacillus cereus and comparative analysis with Bacillus anthracis. Nature 2003 423 87 91 12721630 10.1038/nature01582
Kunst F Ogasawara N Moszer I Albertini AM Alloni G Azevedo V Bertero MG Bessières P Bolotin A Borchert S The complete genome sequence of the Gram-postive bacterium Bacillus subtilis. Nature 1997 390 249 256 9384377 10.1038/36786
Fleischmann RD Adams MD White O Clayton RA Kirkness EF Kerlavage AR Bult CJ Tomb J-F Dougherty BA Merrick JM Whole-genome random sequencing and assembly of Haemophilus influenzae Rd. Science 1995 269 496 512 7542800
Goffeau A Barrell BG Bussey H Davis RW Dujon B Feldmann H Galibert F Hoheisel JD Jacq C Johnston M Life with 6000 genes. Science 1996 274 546 567 8849441 10.1126/science.274.5287.546
Wood V Gwilliam R Rajandream MA Lyne M Lyne R Stewart A Sgouros J Peat N Hayles J Baker S The genome sequence of Schizosaccharomyces pombe. Nature 2002 415 871 880 11859360 10.1038/nature724
The C. elegans Sequencing Consortium Genome sequence of the nematode C. elegans: A platform for investigating biology. Science 1998 282 2012 2018 9851916 10.1126/science.282.5396.2012
Adams MD Celniker SE Holt RA Evans CA Gocayne JD Amanatides PG Scherer SE Li PW Hoskins RA Galle RF The genome sequence of Drosophila melanogaster. Science 2000 287 2185 2195 10731132 10.1126/science.287.5461.2185
Galagan JE Calvo SE Borkovich KA Selker EU Read ND Jaffe D FitzHugh W Ma LJ Smirnov S Purcell S The genome sequence of the filamentous fungus Neurospora crassa. Nature 2003 422 859 868 12712197 10.1038/nature01554
Altschul SF Madden TL Schaffer AA Zhang JH 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
Washington University BLAST Archives
Smith TF Waterman MS Identification of common molecular subsequences. J Mol Biol 1981 147 195 197 7265238 10.1016/0022-2836(81)90087-5
Henikoff S Henikoff JG Amino-acid substitution matrices from protein blocks. Proc Natl Acad Sci USA 1992 89 10915 10919 1438297
SEG Download Site
Wootton JC Federhen S Statistics of local complexity in amino acid sequences and sequence databases. Comput Chem 1994 17 149 163 10.1016/0097-8485(93)85006-X
Conant GC Wagner A A fast algorithm for determining the longest combination of local alignments to a query sequence. BMC Bioinformatics 2004 5 62 15149555 10.1186/1471-2105-5-62
Conant GC Wagner A GenomeHistory: A software tool and its application to fully sequenced genomes. Nucleic Acids Res 2002 30 3378 3386 12140322 10.1093/nar/gkf449
Muse SV Gaut BS A likelihood approach for comparing synonymous and nonsynonymous nucleotide substitution rates, with application to the chloroplast genome. Mol Biol Evol 1994 11 715 724 7968485
Goldman N Yang Z A codon-based model of nucleotide substitution for protein-coding DNA sequences. Mol Biol Evol 1994 11 725 736 7968486
Qian J Luscombe NM Gerstein M Protein family and fold occurrence in genomes: power-law behavior and evolutionary model. J Mol Biol 2001 313 673 681 11697896 10.1006/jmbi.2001.5079
Hahn MW Conant GC Wagner A Molecular evolution in large genetic networks: connectivity does not equal constraint. J Mol Evol 2004 58 203 211 15042341 10.1007/s00239-003-2544-0
Maeder DL Weiss RB Dunn DM Cherry JL Gonzalez JM DiRuggiero J Robb FT Divergence of the hyperthermophilic archaea Pyrococcus furiosus and P. horikoshii inferred from complete genomic sequences. Genetics 1999 152 1299 1305 10430560
Stein LD Bao Z Blasiar D Blumenthal T Brent MR Chen N Chinwalla A Clarke L Clee C Coghlan A The genome sequence of Caenorhabditis briggsae: a platform for comparative genomics. PLoS Biol 2003 1 E45 14624247 10.1371/journal.pbio.0000045
| 15960802 | PMC1175970 | CC BY | 2021-01-04 16:05:39 | no | Genome Biol. 2005 May 9; 6(6):R50 | utf-8 | Genome Biol | 2,005 | 10.1186/gb-2005-6-6-r50 | oa_comm |
==== Front
Genome BiolGenome Biology1465-69061465-6914BioMed Central London gb-2005-6-6-r511596080310.1186/gb-2005-6-6-r51ResearchNatural antisense transcripts with coding capacity in Arabidopsis may have a regulatory role that is not linked to double-stranded RNA degradation Jen Chih-Hung [email protected] Ioannis [email protected] David R [email protected] Peter [email protected] School of Biochemistry and Microbiology, University of Leeds, Leeds LS2 9JT, UK2 Centre for Plant Science, The University of Leeds, Leeds LS2 9JT, UK2005 1 6 2005 6 6 R51 R51 16 12 2004 9 3 2005 5 5 2005 Copyright © 2005 Jen 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.
Transcription data analysis of overlapping gene pairs in Arabidopsis thaliana argues against a predominant RNA degradation effect induced by dsRNA formation. Instead, it suggests alternative roles for dsRNAs such as regulation of alternative splicing in polyadenylation.
Background
Overlapping transcripts in antisense orientation have the potential to form double-stranded RNA (dsRNA), a substrate for a number of different RNA-modification pathways. One prominent route for dsRNA is its breakdown by Dicer enzyme complexes into small RNAs, a pathway that is widely exploited by RNA interference technology to inactivate defined genes in transgenic lines. The significance of this pathway for endogenous gene regulation remains unclear.
Results
We have examined transcription data for overlapping gene pairs in Arabidopsis thaliana. On the basis of an analysis of transcripts with coding regions, we find the majority of overlapping gene pairs to be convergently overlapping pairs (COPs), with the potential for dsRNA formation. In all tissues, COP transcripts are present at a higher frequency compared to the overall gene pool. The probability that both the sense and antisense copy of a COP are co-transcribed matches the theoretical value for coexpression under the assumption that the expression of one partner does not affect the expression of the other. Among COPs, we observe an over-representation of spliced (intron-containing) genes (90%) and of genes with alternatively spliced transcripts. For loci where antisense transcripts overlap with sense transcript introns, we also find a significant bias in favor of alternative splicing and variation of polyadenylation.
Conclusion
The results argue against a predominant RNA degradation effect induced by dsRNA formation. Instead, our data support alternative roles for dsRNAs. They suggest that at least for a subgroup of COPs, antisense expression may induce alternative splicing or polyadenylation.
==== Body
Background
Genome-wide searches in the genomes of several species have identified a surprisingly high proportion of overlapping gene pairs. Depending on the sample sizes analyzed and search criteria, the frequencies for overlapping gene pairs vary between 4% and 9% for the human genome, 1.7%-14% in the murine genome, and up to 22% in the fly genome [1]. The predominant composition of overlapping gene pairs is an antiparallel convergent arrangement [2,3], where sense and antisense genes overlap within their 3' regions. Joint expression of both these genes in the same cell would allow the partly overlapping transcripts to associate as dsRNA molecules, which may interfere with RNA processing, transport, stability or other molecular mechanisms. Convergently overlapping gene pairs (COPs) can therefore provide the source for natural antisense transcripts (NATs) that may act as regulators of the sense gene. In addition to NATs being transcribed from the same locus as the sense transcript (cis-NATs), NATs can be transcribed from a different locus (trans-NATs), as illustrated by a search for overlapping transcripts with coding capacity in the human genome, which identified 87 cis-NATs and 80 trans-NATs [3].
In bacteria, more than 100 NATs are involved in the regulation of a variety of biological functions, including the control of copy number, conjugation and post-segregational killing in plasmids, lysis/lysogeny switches in phages, and transposition frequency in transposons [4]. In eukaryotes, a very detailed characterization of the molecular role of specific NATs has only been achieved for a few examples.
NAT-mediated interference with splicing is illustrated by the alternative processing of mRNAs of the gene for the thyroid hormone receptor ErbAα, which is regulated by an antisense transcript [5]. Overlapping genes can share a bidirectional poly(A) region as demonstrated for the human genes ABHD1 and Sec12 [6]. Several examples document the fact that antisense transcripts can increase sense transcript stability, when dsRNA regions cover the 3' untranslated region (UTR) and possibly mask out target sequences for RNA cleavage [7]. Alternatively, RNA duplex formation can increase transcript sensitivity and induce site-specific cleavage, as shown for the human TYMS mRNA and TRS antisense transcripts [8].
An example of RNA interference (RNAi)-based regulation of an endogenous gene via NATs is the repression of the testis-expressed Stellate gene in Drosophila by paralogous Su(Ste) tandem repeats [9]. Both strands of repressor Su(Ste) repeats are transcribed, producing sense and antisense RNA, most probably as part of a dsRNA-based silencing mechanism, as Stellate silencing is associated with the presence of short Su(Ste) RNAs. Antisense expression can also affect translation, as illustrated by the influence of an antisense transcript on the translation of different isoforms of fibroblast growth factor-2 (FGF2) [10]. In the nucleus, dsRNA can be edited by dsRNA-dependent adenosine deaminases, which convert about 50% of adenosine residues into inosines, leading to the unwinding of the RNA duplex [11]. Inosine-containing RNAs are not translated as they are retained in the nucleus [12].
In mice about 35% of overlapping genes transcribe noncoding RNA. Overlapping genes are scattered around the genome with no apparent bias. Overlaps range from 20 to 3,400 base-pairs (bp) with an average of 372 bp, as far as the quality of transcript annotation allows such predictions. There is some evidence for an over-representation of overlapping genes among specific functional categories, that is, imprinted genes and DNA repair genes [1]. Twenty-two out of 58 known imprinted murine genes are transcribed from both strands. Frequently, one partner transcribes a noncoding RNA. Antisense transcripts may regulate imprinting states of the sense promoter (Kcnq1/Kcnq1ot1) or may induce dsRNA-based gene silencing as proposed for Ifd2R/Air. About 20% of known human DNA repair genes overlap either convergently or divergently in an antiparallel arrangement [1].
Mammalian mRNAs that form sense-antisense pairs frequently exhibit reciprocal expression patterns, but permanent coexpression of sense and antisense transcripts can also occur in some tissues, although it is difficult to prove that both genes are transcribed in the same cell. Coexistence of sense and antisense transcripts may indicate a stabilizing effect of dsRNA, or it may depict cases where RNA duplex formation is impaired as a result of secondary structures, or because sense and antisense transcripts or the enzymes required for duplex formation are separated by compartmentalization [13].
To gain an insight into the existence and role of overlapping antisense pairs in plants, we have screened the Arabidopsis thaliana genome for COPs with sense and antisense genes that encode a protein, and have compared the expression profile of the associated genes.
Results
Overlapping gene pairs in the Arabidopsis genome
A screen of the Arabidopsis genome for protein-coding genes with overlapping orientations identified 1,083 groups containing a total of 2,147 overlapping genes. For 26 groups, the overlap involves three genes, and for 1,057 groups, two genes are arranged as overlapping pairs (Figure 1 and Additional data file 1). The majority of overlapping gene pairs are organized as COPs. The size of the overlapping region for these 956 COPs varies between 1 and 2,820 bp, with an average length of 431 bp. The genes are scattered around all five chromosomes with no obvious clustering bias (data not shown). Among the 1,912 COP genes, we found nine transposable elements. This is in contrast to the presence of 2,372 transposons among the 30,624 Arabidopsis genes. Transposons are therefore strongly under-represented among COPs, therefore transposon-derived antiparallel gene pairs are under heavy selective pressure. Among the 1,912 genes of the 956 COPs, 954 genes have antisense transcripts that extend into the open reading frame (ORF) region of the sense transcript (Figure a-c and Additional data file 2) but the ORF regions of the sense and antisense transcript overlap for only 13 COPs (Figure 2a and Table 1).
To examine the degree of sequence conservation for COPs sense and antisense genes, we used BLAST to search for homologs of each COPs gene. For a subset of 242 genes, we can define 89 homology groups with 2-11 members. The proteins encoded by the sense members of each group are at least 20% identical, with an E-value less than 0.001. An analysis of the degree of sequence conservation among family members showed very low conservation among the coding or promoter regions of the antisense partners of homologous sense genes. With the exception of the largest family, the promoter regions of homologous sense genes are also poorly conserved (Table 2 and Additional data files 4, 5 and 6). With a few exceptions, possibly representing relatively recent duplications, the data indicate that homologous sense genes do not in general have homologous antisense partners.
COPs gene-expression profiles
To analyze the transcriptional activity of COPs sense and antisense genes, we used the GSE636 annotated gene-expression database [14,15], which provides expression data for 1,866 COPs genes in suspension culture, in 7-day old seedlings, in roots and in flowers. If antisense arrangements are predominantly responsible for dsRNA-mediated transcript degradation, we would expect that a significant proportion of COPs genes would be under-represented among the transcript pool. For the total pool of 26,939 genes represented in the GSE636 database, we find a representation of 49.8-53.1% of these genes among the detectable transcripts (Table 3). Of the 1,866 COPs genes represented among this pool, 63.4-67.9% are expressed, which argues against a specific depletion of COPs transcripts in any of the four sample tissues.
This assumption is further supported by the lack of any bias against the joint expression of sense and antisense copies from the same COP. We can calculate a theoretical value for the joint expression of a sense and an antisense member of the same COP based on the representation of the COP genes in the transcriptional pool. If, for example, the probability that a COP gene is expressed in flowers is 67.9%, the probability that any two COP genes are jointly expressed is 67.9% × 67.9%. The expected value of 46.1% matches the observed value of 45.6% determined for the joint expression of both genes of a COP (Table 3). We observe a similar match for the other tissues, which suggests that there is no bias against the joint expression of both COP partner genes. For about 20% of all COPs, both members are jointly expressed in all tissues tested.
We also examined the microarray data provided by the Nottingham Arabidopsis Stock Centre (NASC) [16]. Table 4 compiles 21 Affymetrix ATH1 arrays for seven different Arabidopsis tissues based on three replicates for each assay. The datasets were retrieved by searching for BioSource_ID on [17].
Although the expression probabilities of both the total gene pool and the COPs gene pool differ significantly among individual tissues, the expected and observed values for the coexpression of COPs sense and antisense genes again match, reinforcing the lack of any indication of a transcript degradation mechanism (Table 4). To assess whether transcript degradation depends on a specific experimental condition, we assembled the expression data for all 1,437 microarrays that were available. For each microarray, we calculated the proportion of transcripts that were expressed, both for the total gene pool comprising 22,746 genes and for the 1,596 COPs genes represented in the total pool. A depletion of COPs-specific transcripts for any of the 1,437 microarrays should result in a significant reduction of the expressed COPs pool. We do not find a single case where the COPs genes are under-represented among the transcripts detectable in an array experiment. Compared to the transcriptional activity of the whole gene pool, the transcriptional activity of the COPs gene pool is between 1.008 and 1.485 times higher. This argues against a specific reduction in COPs gene activity under any of the experimental conditions used for the array experiments (Additional data file 7).
Indications for a role of antisense transcripts in sense transcript splicing
Among the 1,912 COPs genes we find a considerable bias for splicing. Of the COPs genes, 1,723 (90.1%) are spliced, while among a total of 30,624 Arabidopsis genes only 21,157 (69.1%) genes are spliced. If antisense transcripts played a role in sense transcript alternative splicing, we would expect an enrichment among COPs genes in alternative splicing. Out of the total pool of 30,624 genes, 7.6% encode more than one transcript, which are all alternatively spliced. This proportion rises to 14.5% among the 1,912 COPs genes (Table 5). This increase is even more pronounced in the rice genome, where it increases from 4.4% to 20.9% COPs.
To assess if the over-representation of multiple transcripts among COPs was linked to a variation in splicing, transcriptional start site (TSS) or polyadenylation, we analyzed the representation of these modifications among the spliced COPs genes. The results showed a positive bias for alternative splicing (Table 5). Interestingly, this bias was restricted to COPs with an antisense transcript that overlaps the intron region of the sense transcript (Table 6). Moreover, among the COPs with antisense transcripts overlapping at least 40 bp of the sense transcript, we also detected a positive bias for a variation of the polyadenylation sites (Table 6), whereas no positive or negative bias was observed for TSS variation (see Additional data file 8).
For 146 genes, the antisense transcript terminates within 10 bp away from the intron-exon boundary of the sense transcript (Figure 3 and Additional data file 3). The proportion of alternatively spliced genes among this group increases to 21.2%. We tested whether these COPs genes were specifically prone to alternative splicing of the final exon but could not find any evidence for this assumption (see Additional data file 8).
Overall, the enhanced likelihood that members of overlapping gene pairs contain introns, the enrichment in genes encoding alternatively spliced transcripts, and the increased frequency of alternatively spliced and variably polyadenylated transcripts when an intron overlaps with an antisense transcript, suggest that, at least for the majority of overlapping gene pairs, the antisense transcript could play a role in the regulation of splicing and/or polyadenylation.
Discussion
We have characterised the organization and expression profiles of 956 convergent overlapping gene pairs of A. thaliana to assess the potential molecular mechanisms associated with this unusual gene organization, which provided the opportunity for dsRNA formation as a result of the annealing of a sense and antisense transcript.
In animal genomes especially, a number of different mechanisms have been described that involve dsRNA formation. dsRNA formation can interfere with biological activities that require binding of RNA or proteins to the transcript [13]. This may include processes such as RNA splicing, editing, transport, degradation or translation. Alternatively, dsRNA could function as a trigger for an RNAi process, providing a target for Dicer enzymes, and be specifically degraded to small interfering RNA (siRNA) molecules [18]. The latter mechanism would lead to mutual destruction of the sense and antisense transcripts, whereas antisense transcript-mediated effects on sense RNA processing would not necessarily alter the primary transcript levels, although they could still influence the potential for primary transcript expression.
As in other species [1,19], the majority of overlapping gene pairs in A. thaliana are arranged in a convergent orientation. Only a very small group of 13 out of 956 COPs show an overlap between the ORFs of the sense and antisense transcripts, which probably reflects the associated evolutionary stress of such an arrangement, as any mutation in the overlapping region would affect both genes. For 50.1% of COPs the sense-antisense overlap does not include any protein-coding region, which makes it unlikely that in this subgroup of COPs the antisense transcript plays a role in regulating the coding region of the sense gene. Antisense transcripts for the members of this group are more likely to jointly use bidirectional poly(A) signals [6] or to regulate transcript stability [7].
We do observe a very high level of joint expression of sense and antisense transcript from overlapping gene pairs. This is in marked contrast to data from Plasmodium falciparum, where only 5% of sense-antisense loci show joint expression [20] and thus support models for a direct regulation of sense transcript by antisense expression via dsRNA degradation. In contrast, the relatively high expression frequency of COPs in Arabidopsis, and the joint presence of sense and antisense transcripts in the same tissue, do not support a dsRNA degradation model. Even a detailed analysis of 1,437 microarrays does not imply that under any conditions or for any specific tissue the COPs gene pool is significantly depleted. While dsRNA-based transcript degradation may occur for some COPs, our data suggest that for the majority of COPs, antisense expression is not linked to transcript degradation pathways.
An interesting observation, which may hint at an alternative interference mechanism between sense and antisense transcripts, is the significant bias for COPs to be spliced, and the enrichment among COPs of alternatively spliced transcripts. These features may indicate a role for antisense transcripts in alternative splicing. Such a mechanism would resemble the effect of antisense expression for the thyroid hormone receptor gene, erbAα, which leads to alternative RNA processing [5]. The bias for antisense transcripts to terminate close to the final intron-exon boundary remains a mystery. One could assume that the termination of the antisense transcript near the final sense intron-exon boundary might reflect a selection for antisense transcripts that interfere with splicing of the last intron. However, we could not observe any positive bias for such events among this COPs gene group.
The assumption that antisense transcripts can interfere with splicing events is further supported by the observation that overlaps between antisense transcripts and a sense intron region generate a bias for alternative splicing and also for polyadenylation variation. This may reflect a linkage between these two mechanisms, which has been demonstrated for animal systems where polyadenylation and the splicing of the final intron especially can be coupled [21].
During the review process of this paper, a similar study by Wang and co-workers [22] was published. The main differences between the two studies are in the sets of overlapping genes considered, and the nature of the experimental evidence of gene expression. While we consider only gene pairs where both partners show evidence of protein-coding capacity, Wang and co-workers also considered cases where at least one transcript may not be protein coding. We base our studies of expression on two large microarray datasets, while Wang and co-workers use data from a massively parallel signature sequencing (MPSS) study. While we find no evidence in the microarray data of exclusive transcription relationships for COPs gene pairs, the MPSS evidence of exclusive transcription of the gene pairs in the Wang study is clear. This apparent contradiction may be explained by differences in the gene sets studied, particularly as expression data is only available for a subset of the genes in each study, or by differences in the quality of the expression data. Nevertheless, the two studies taken together give evidence of various significant biological consequences of gene overlaps, including effects on sense gene splicing or polyadenylation (our study) and coexpression of gene pairs [22].
It is important to remember that in our study, we have concentrated on a specific subgroup of convergently overlapping genes, with both sense and antisense transcripts encoding an ORF. Among this group there may be an over-representation of gene pairs for which sense and antisense transcription jointly regulate the production efficiency of both proteins, for example via the common use of a bidirectional polyadenylation region, or by co-editing of both strands associated as dsRNA. Such mechanisms would require the joint transcription of both genes in the same tissue, and our data do suggest that sense and antisense transcripts are frequently coexpressed. On the other hand, one would assume that co-regulation would preferentially be used for genes encoding proteins that are linked in their biological role. One would therefore expect a high degree of conservation for both proteins among homologous COPs, whereas our data show that COPs with homologous proteins encoded by their sense genes do not show the same conservation for the proteins encoded by their antisense genes.
The selection of sense/antisense transcripts with coding capacity may also be the reason for the lack of an indication of dsRNA-based degradation of COPs. A considerable proportion of overlapping antisense genes are noncoding RNAs [23] or are trans-NATs transcribed from different genetic loci [3]. These overlapping transcript types may contain a much higher proportion of genes regulated by transcript degradation than the COPs analyzed in this study. A final confirmation of the role of dsRNA for individual genes will require a more detailed experimental analysis. Our analysis should, however, provide a useful first step in defining distinct groups of COP genes as a basis for a more detailed molecular characterization.
Conclusion
The Arabidopsis genome contains 956 COPs with coding capacity that have the potential to form dsRNA. In contrast to data from other species, a comparative expression analysis indicates that sense and antisense transcripts of COPs loci can coexist in the same tissue at the same frequency as the transcripts of any other unlinked genes, with no indication of specific degradation of such sense-antisense transcript pairs. This observation does not exclude the presence of dsRNA degradation pathways for individual loci, but it refocuses the attention on alternative roles for natural antisense transcripts in plants, preferentially those that do not lead to an overall change in transcript levels but rather affect transcript processing or localization. In line with this view, we observe a high proportion of intron-containing genes in COPs, in both Arabidopsis and rice, and an enrichment for genes with alternatively spliced transcripts, indicative of a role for some COP antisense transcripts in splicing modification. In addition, we detect a potential link between alternative splicing and poly(A) site variation. This work provides a set of databases for COPs, based on the degree of sense-antisense overlap and expression, which should provide a basis for the selection of individual candidate loci for a detailed molecular analysis of the different dsRNA pathways.
Materials and methods
Analysis of overlapping transcripts
All Arabidopsis genome information, including gene ID (AGI code), transcript orientation, and gene and exon position coordinates of transcripts with coding regions, was downloaded from The Arabidopsis Information Resource (TAIR) ftp website [24]. These data were stored in a MySQL database [25] designed for general genomic analysis, and the overlapping transcript analysis was implemented with custom SQL language queries and Perl scripts using the Perl DBI module. The rice genome data were obtained from [26,27].
Analysis of gene variation
Genes with more than one transcript were further analyzed for variation in TSS, polyadenylation site, and alternative splicing. TSS variation was detected by comparing the starting genomic positions of the first exons of genes with more than one transcript, and variation in the polyadenylation site by comparing the ends of the last exons. Genes with more than one transcript that had identical TSSs and polyadenylation sites, or had different numbers of exons, were considered as alternatively spliced. Genes with more than one transcript that had the same number of exons and variation in TSS and polyadenylation site, underwent comparison of their intron boundaries to detect alternative splicing. To detect alternative splicing in the last intron, alternatively spliced transcripts underwent comparison of the borders of their last intron.
Hypergeometric distribution
p-values for over- or under-representations of genes were calculated as the upper or lower tail of the hypergeometric distribution p(x ≥ X) or p(x ≤ X), respectively, where p(x;N,R,k) = C(R,x)C(N - R,k - x)/C(N,k). Here p refers to the probability that a list of k genes should contain x genes with a particular property (for example, alternative splicing), when the list has been selected randomly without replacement from a set of N genes in which R genes exhibit the same property. C(n,m) is the number of distinct combinations of m objects that can be drawn from a set of n objects. The hypergeometric distribution was calculated with the R package [28].
Microarray data
Arabidopsis microarray data were obtained from two sources. The dataset from Gene Expression Omnibus [14,15] with accession number GSE636 is a collection of microarray experiments using high-density oligonucleotide arrays. It contains transcriptional activity information (detection call only) for the complete set of all protein-coding genes in different tissues. The Affymetrix ATH1 array data were acquired using the Nottingham Arabidopsis Stock Centre (NASC) AffyWatch service [16]. In NASC's datasets, including 1,437 arrays for 93 experimental purposes, the transcription information for each gene consists of detection call and signal value, as calculated from the Affymetrix MAS 5.0 analysis algorithms [29]. The analysis of expression data reported in Results was achieved using a combination of Perl script processing and Microsoft Excel spreadsheet analysis.
Coding region and upstream sequence similarity analysis of homologous COP genes
The COPs genes were clustered on the basis of their protein sequences with a 20% similarity threshold using the program BLASTclust [30]. The similarity of the associated coding and upstream regulatory regions within each cluster were tested by pairwise searches using BLAST2P [31].
Additional data files
The following additional files are available with the online version of this paper. Additional data file 1 is a supplement to Figure 1 listing all overlapping genes with ID, annotation and size of overlapping region. Additional data file 2 is a supplement to Figure 2 classifying 1,912 COPs genes according to their overlapping regions. Additional data file 3 is a supplement to Figure 3, calculating the antisense transcript end position in relation to the sense intron-exon boundary for the 956 COPs pairs. Additional data file 4 is a supplement A to Table 2, listing 242 COPs member genes of 89 COPs families. Additional data file 5 is a supplement B to Table 2, comparing homology among 1 kb promoter regions of COPs family members. Additional data file 6 is a supplement C to Table 2, comparing homology among sense and antisense encoded proteins for members of 89 COPs families. Additional data file 7 is a supplement to Table 4, showing expression analysis of the total gene pool and the COPs gene pool for 1,437 microarray experiments. Additional data file 8 is a supplement to Table 6, including a correlation analysis of alternative splicing, TSS variation and polyadenylation variation for COPs with respect to the termination of the antisense transcript in relation to the sense intron-exon boundary.
Supplementary Material
Additional File 1
Supplement to Figure 1. List of all overlapping genes with ID, annotation and size of overlapping region. List of all overlapping genes with ID, annotation and size of overlapping region
Click here for file
Additional File 2
Supplement to Figure 2. Classification of 1,912 COPs genes according to their overlapping regions. Classification of 1,912 COPs genes according to their overlapping regions
Click here for file
Additional File 3
Supplement to Figure 3. Calculation of the antisense transcript end position in relation to the sense intron-exon boundary for the 956 COPs pairs. Calculation of the antisense transcript end position in relation to the sense intron-exon boundary for the 956 COPs pairs
Click here for file
Additional File 4
Supplement A to Table 2. 242 COPs member genes of 89 COPs families. 242 COPs member genes of 89 COPs families
Click here for file
Additional File 5
Supplement B to Table 2. Comparison of homology among 1 kb promoter regions of COPs family members. Comparison of homology among 1 kb promoter regions of COPs family members
Click here for file
Additional File 6
Supplement C to Table 2. Homology comparison among sense and antisense encoded proteins for members of 89 COPs families. Homology comparison among sense and antisense encoded proteins for members of 89 COPs families
Click here for file
Additional File 7
Supplement to Table 4. Expression analysis of the total gene pool and the COPs gene pool for 1,437 microarray experiments. Expression analysis of the total gene pool and the COPs gene pool for 1,437 microarray experiments
Click here for file
Additional File 8
Supplement to Table 6. Correlation analysis of alternative splicing, TSS variation and polyadenylation variation for COPs with respect to the termination of the antisense transcript in relation to the sense intron-exon boundary. Correlation analysis of alternative splicing, TSS variation and polyadenylation variation for COPs with respect to the termination of the antisense transcript in relation to the sense intron-exon boundary
Click here for file
Acknowledgements
We are grateful to Athanasios Theologis for providing access to the GSE636 database. This work was supported by a European Commission grant to the Epigenome Network of Excellence (503433).
Figures and Tables
Figure 1 A comparison of the arrangements of overlapping gene pairs in Arabidopsis thaliana. A and A' label the start and end of the sense transcript, B' and B label the start and end of the antisense transcript. The total number of genes involved in group 1, 2 and 3 is 2,157, of which 2,147 are unique; the remaining 10 comprise four genes that are members of both group 1 and group 2 pairs, five genes that are members of both group 1 and group 3 pairs, and 1 gene that is a member of both a group 1 and group 3 pair.
Figure 2 The organization of convergent overlapping gene pairs with respect to the protein coding capacity of the sense and antisense transcripts.
Figure 3 Illustration of the distance between the end of the antisense transcript and the last intron-exon boundary of the sense transcript. Negative values refer to a termination of the antisense transcript 5' to the intron-exon boundary.
Table 1 COPs with sense-antisense overlaps within the coding regions
Sense gene ID Annotation Antisense gene ID Annotation ORF overlap (bp)
AT1G08260 DNA-directed DNA polymerase epsilon catalytic subunit, putative AT1G08270 Expressed protein 45
AT1G52010 Mutator-like transposase family AT1G52020 Pseudogene, Ulp1 protease family 44
AT1G52087 Hypothetical protein AT1G52090 Hypothetical protein 72
AT1G68935 Expressed protein AT1G68940 Armadillo/beta-catenin repeat protein-related/U-box domain-containing protein 698
AT2G12855 Gypsy-like retrotransposon family AT2G12860 Gypsy-like retrotransposon family 116
AT2G19330 Leucine-rich repeat family protein AT2G19340 Membrane protein, putative 141
AT3G59940 Kelch repeat-containing F-box family protein AT3G59950 Autophagy 4b (APG4b) 10
AT4G02200 Drought-responsive family protein AT4G02210 Expressed protein 13
AT4G21366 S-locus protein kinase-related AT4G21370 S-locus protein kinase, putative 72
AT4G29830 Transducin family protein/WD-40 repeat family protein AT4G29840 Threonine synthase, chloroplast 587
AT5G18210 Short-chain dehydrogenase/reductase (SDR) family protein AT5G18220 Glycosyl hydrolase family 17 protein 6
AT5G28232 Mutator-like transposase family AT5G28235 Ulp1 protease family protein 29
AT5G48200 Hypothetical protein AT5G48205 Hypothetical protein 334
Table 2 Homology assessment for 89 COPs families that contain 2-11 family members
Number of COPs family Number of family members Sense-gene-encoded proteins with a similarity E-value < 0.001 1 kb sense promoter regions with a similarity E-value < 0.001 Antisense-gene-encoded proteins with a similarity E-value < 0.001 1 kb antisense promoter regions with a similarity E-value < 0.001
1 11 11 7 2 2
2 8 8 2 0 0
3 8 8 0 0 2
4 7 7 0 0 0
5-6 7 7 2 0 0
7 6 6 2 2 0
8-9 5 5 0 0 0
10 4 4 2 0 0
11-12 4 4 0 0 0
13-14 4 4 0 2 0
15 4 4 2 2 0
16-21 3 3 0 0 0
22-72 2 2 0 0 0
73-75 2 2 2 0 0
76-77 2 2 0 2 2
78-89 2 2 0 2 0
The numbers refer to the family members that share sequence similarity of an E-value below 0.001 with at least one other family member. Among the COPs families, the homology is well conserved among sense-gene-encoded proteins, while sequence conservation is rare among antisense-gene-encoded proteins. With the exception of family 1 sense gene promoters, the homology is also poorly conserved among promoter regions of sense and antisense genes.
Table 3 Expression analysis of 1,866 COPs genes based on expression data from the GSE636 annotated gene-expression database
Tissue % of expressed genes among 26,939 Arabidopsis genes % of expressed genes among 1,866 overlapping COPs genes % of COPs with jointly expressed sense and antisense genes (observed value) % of COPs with jointly expressed sense and antisense genes (expected value)
Flowers 52.5% 67.9% 45.6% 46.1%
Roots 51.4% 63.4% 38.5% 40.2%
Suspension culture 53.1% 66.3% 42.7% 44.0%
7 day old seedlings 49.8% 64.1% 40.4% 41.1%
Table 4 Expression of 1,596 COPs genes based on the NASC microarray database
Tissue % of expressed genes among 22,746 Arabidopsis genes % of expressed genes among 1,596 overlapping COPs genes % of COPs with jointly expressed sense and antisense genes (observed value) % of COPs with jointly expressed sense and antisense genes (expected value)
Flowers 62.9% 82.5% 67.7% 68.1%
Pollen 31.7% 36.0% 12.3% 13.0%
Seedlings, green parts 57.3% 76.7% 57.8% 58.8%
Cotyledons 55.7% 75.3% 56.1% 56.7%
Leaves 55.6% 74.8% 54.5% 56.0%
Roots 62.6% 76.9% 58.3% 59.1%
Hypocotyl 62.9% 82.1% 66.8% 67.4%
Table 5 Representation of spliced genes among COPs, and correlation analysis for transcript modifications among these genes
Total genes COPs p-value
COP genes show a strong positive bias for splicing
Total 30,624 1,912
Spliced 21,157 (69.1%) 1,723 (90.1%) 4.7e-113
Spliced COP genes show a positive bias for alternative splicing
Spliced 21,157 1,723
Alternatively spliced 2,331 (11.0%) 268 (15.6%) 1.3e-9
Alternatively spliced COP genes do not show a significant bias for alternative splicing at the last intron, TSS variation or polyadenylation site variation
Alternatively spliced 2,331 268
Last intron alternative splicing 1,662 (71.3%) 195 (72.8%) 0.31
TSS variation 1,424 (61%) 158 (59.0%) 0.24
Polyadenylation site variation 1,019 (43.7%) 107 (39.9%) 0.10
Table 6 Analysis of preferences for alternative splicing and polyadenylation site variation among spliced COPs genes, in dependence of the termination site of the antisense transcript
Spliced COP genes with an antisense transcript not overlapping a sense transcript intron region, show a significant negative bias for alternative splicing
COPs genes COPs with antisense gene ending 3,000-0 bp before the sense I/E boundary p-value
Spliced genes 1,723 1,497
Alternatively spliced genes 268 (15.6%) 217 (14.5%) 0.0018
Spliced COPs genes with an antisense transcript overlapping a sense transcript intron region, show a significant positive bias for alternative splicing
COPs genes COPs with an antisense gene ending 0-3,000 bp behind the sense I/E boundary p-value
Gene with splicing 1,723 226
Alternative splicing 268 (15.6%) 51 (22.6%) 0.0018
COPs genes COPs with an antisense gene ending > 40 bp behind the sense I/E boundary p-value
Gene with splicing 1,723 129
Alternative splicing 268 (15.6%) 35 (27.1%) 0.00032
Alternatively spliced COPs sense genes with an antisense transcript ending more than 40 bp behind their last I/E boundary, show a positive bias for polyadenylation site variation
COPs genes COPs with an antisense gene ending > 40 bp behind the sense I/E boundary p-value
Alternatively spliced 268 35
Polyadenylation site variation 107 (39.9%) 25 (71.4%) 5.5e-05
==== Refs
Boi S Solda G Tenchini ML Shedding light on the dark side of the genome: overlapping genes in higher eukaryotes. Curr Genomics 2004 5 509 524 10.2174/1389202043349020
Fahey ME Moore TF Higgins DG Overlapping antisense transcription in the human genome. Comp Funct Genomics 2002 3 244 253 10.1002/cfg.173
Lehner B Williams G Campbell RD Sanderson CM Antisense transcripts in the human genome. Trends Genet 2002 18 63 65 11818131 10.1016/S0168-9525(02)02598-2
Wagner EGH Flardh K Antisense RNAs everywhere? Trends Genet 2002 18 223 226 12047936 10.1016/S0168-9525(02)02658-6
Hastings ML Milcarek C Martincic K Peterson ML Munroe SH Expression of the thyroid hormone receptor gene, erbAα, in B lymphocytes: alternative mRNA processing is independent of differentiation but correlates with antisense RNA levels. Nucleic Acids Res 1997 25 4296 4300 9336460 10.1093/nar/25.21.4296
Edgar A The gene structure and expression of human ABHD1: overlapping polyadenylation signal sequence with Sec12. BMC Genomics 2003 4 18 12735795 10.1186/1471-2164-4-18
Gray TA Azama K Whitmore K Min A Abe S Nicholls RD Phylogenetic conservation of the Makorin-2 gene, encoding a multiple zinc-finger protein, antisense to the RAF1 proto-oncogene. Genomics 2001 77 119 126 11597136 10.1006/geno.2001.6627
Chu J Dolnick BJ Natural antisense (rTS[α]) RNA induces site-specific cleavage of thymidylate synthase mRNA. Biochim Biophys Acta 2002 1587 183 193 12084460 10.1016/S0925-4439(02)00081-9
Aravin AA Naumova NM Tulin AV Vagin VV Rozovsky YM Gvozdev VA Double-stranded RNA-mediated silencing of genomic tandem repeats and transposable elements in the D. melanogaster germline. Curr Biol 2001 11 1017 1027 11470406 10.1016/S0960-9822(01)00299-8
Li AW Murphy PR Erratum to expression of alternatively spliced FGF-2 antisense RNA transcripts in the central nervous system: regulation of FGF-2 mRNA translation. Mol Cell Endocrinol 2000 170 231 242 10.1016/S0303-7207(00)00440-8
Bass BL Weintraub H An unwinding activity that covalently modifies its double-stranded RNA substrate. Cell 1988 55 1089 1098 3203381 10.1016/0092-8674(88)90253-X
Zhang Z Carmichael GG The fate of dsRNA in the nucleus: a p54nrb-containing complex mediates the nuclear retention of promiscuously A-to-I edited RNAs. Cell 2001 106 465 475 11525732 10.1016/S0092-8674(01)00466-4
Vanhee-Brossollet C Vaquero C Do natural antisense transcripts make sense in eukaryotes? Gene 1998 211 1 9 9573333 10.1016/S0378-1119(98)00093-6
Gene Expression Omnibus
Barrett T Suzek TO Troup DB Wilhite SE Ngau WC Ledoux P Rudnev D Lash AE Fujibuchi W Edgar R NCBI GEO: mining millions of expression profiles - database and tools. Nucleic Acids Res 2005 33 Database D562 D566 15608262
Craigon D James N Okyere J Higgins J Jotham J May S NASCArrays: a repository for microarray data generated by NASC's transcriptomics service. Nucleic Acid Res 2004 32 Database D575 D577 14681484 10.1093/nar/gkh133
NASCA Arrays: Affymetrix ATH1 arrays database
Fire A Xu S Montgomery MK Kostas SA Driver SE Mello CC Potent and specific genetic interference by double-stranded RNA in Caenorhabditis elegans. Nature 1998 391 806 811 9486653 10.1038/35888
Lavorgna G Dahary D Lehner B Sorek R Sanderson CM Casari G In search of antisense. Trends Biochem Sci 2004 29 88 94 15102435 10.1016/j.tibs.2003.12.002
Gunasekera AM Patankar S Schug J Eisen G Kissinger J Roos D Wirth DF Widespread distribution of antisense transcripts in the Plasmodium falciparum genome Mol Biochem Parasitol 2004 136 35 42 15138065 10.1016/j.molbiopara.2004.02.007
Bauren G Belikov S Wieslander L Transcriptional termination in the Balbiani ring 1gene is closely coupled to 3'-end formation and excision of the 3'-terminal intron. Genes Dev 1998 12 2759 2769 9732273
Wang X-J Gaasterland T Chua N-H Genome-wide prediction and identification of cis-natural antisense transcripts in Arabidopsis thaliana. Genome Biol 2005 6 R30 15833117 10.1186/gb-2005-6-4-r30
Kiyosawa H Yamanaka I Osato N Kondo S Hayashizaki Y Antisense transcripts with FANTOM2 clone set and their implications for gene regulation. Genome Res 2003 13 1324 1334 12819130 10.1101/gr.982903
TAIR ftp website
DuBois P MySQL. 2000 Indianapolis, IN: New Riders Publishing
TIGR rice: rice expression database
O. sativa database
Ihaka R Gentlement G R:A language for data analysis and graphics. J Comp Graph Statist 1996 5 299 314
Hubbell E Liu WM Mei R Robust estimators for expression analysis. Bioinformatics 2002 18 1585 1592 12490442 10.1093/bioinformatics/18.12.1585
McGinnis S Madden TL BLAST: at the core of a powerful and diverse set of sequence analysis tools. Nucleic Acids Res 2004 32 W20 W25 15215342 10.1093/nar/gnh003
Yuan J Bush B Elbrecht A Liu Y Zhang T Zhao W Blevins R Enhanced homology searching through genome reading frame predetermination. Bioinformatics 2004 20 1416 1427 14976033 10.1093/bioinformatics/bth115
| 15960803 | PMC1175971 | CC BY | 2021-01-04 16:35:47 | no | Genome Biol. 2005 Jun 1; 6(6):R51 | utf-8 | Genome Biol | 2,005 | 10.1186/gb-2005-6-6-r51 | oa_comm |
==== Front
Genome BiolGenome Biology1465-69061465-6914BioMed Central London gb-2005-6-6-r521596080410.1186/gb-2005-6-6-r52ResearchTiling microarray analysis of rice chromosome 10 to identify the transcriptome and relate its expression to chromosomal architecture Li Lei [email protected] Xiangfeng [email protected] Mian [email protected] Viktor [email protected] Ning [email protected] Zhiyu [email protected] Songgang [email protected] Jun [email protected] Xiping [email protected] Xing Wang [email protected] Department of Molecular, Cellular, and Developmental Biology, Yale University, New Haven, CT 06520, USA2 National Institute of Biological Sciences, Zhongguancun Life Science Park, Beijing 102206, China3 Peking-Yale Joint Research Center of Plant Molecular Genetics and Agrobiotechnology, College of Life Sciences, Peking University, Beijing 100871, China4 Beijing Institute of Genomics, Chinese Academy of Sciences, Beijing 101300, China5 National Center of Crop Design, China Bioway Biotech Group Co., LTD, Beijing 100085, China6 Genome Research Facility, NASA Ames Research Center, MS 239-11, Moffett Field, CA 94035, USA2005 27 5 2005 6 6 R52 R52 14 1 2005 1 4 2005 25 4 2005 Copyright © 2005 Li 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.
A transcriptome analysis of chromosome 10 of 2 rice subspecies identifies 549 new gene models and gives experimental evidence for around 75% of the previously unsupported predicted genes.
Background
Sequencing and annotation of the genome of rice (Oryza sativa) have generated gene models in numbers that top all other fully sequenced species, with many lacking recognizable sequence homology to known genes. Experimental evaluation of these gene models and identification of new models will facilitate rice genome annotation and the application of this knowledge to other more complex cereal genomes.
Results
We report here an analysis of the chromosome 10 transcriptome of the two major rice subspecies, japonica and indica, using oligonucleotide tiling microarrays. This analysis detected expression of approximately three-quarters of the gene models without previous experimental evidence in both subspecies. Cloning and sequence analysis of the previously unsupported models suggests that the predicted gene structure of nearly half of those models needs improvement. Coupled with comparative gene model mapping, the tiling microarray analysis identified 549 new models for the japonica chromosome, representing an 18% increase in the annotated protein-coding capacity. Furthermore, an asymmetric distribution of genome elements along the chromosome was found that coincides with the cytological definition of the heterochromatin and euchromatin domains. The heterochromatin domain appears to associate with distinct chromosome level transcriptional activities under normal and stress conditions.
Conclusion
These results demonstrated the utility of genome tiling microarray in evaluating annotated rice gene models and in identifying novel transcriptional units. The tiling microarray sanalysis further revealed a chromosome-wide transcription pattern that suggests a role for transposable element-enriched heterochromatin in shaping global transcription in response to environmental changes in rice.
==== Body
Background
As one of the most important crop species in the world and a model for the Gramineae family, rice (Oryza sativa) was selected as the first monocotyledonous plant to have its genome completely sequenced. Draft genome sequences of the two major subspecies of rice, indica and japonica, were made available in 2002 [1,2]. These were followed by the advanced sequences of japonica chromosomes 1, 4 and 10 [3-5]. The finish-quality whole-genome sequences of indica and japonica have recently been obtained [6-8].
Available rice sequences have been subjected to extensive annotation using ab initio gene prediction, comparative genomics, and a variety of other methods. These analyses revealed abundant compositional and structural features of the predicted rice genes that deviate from genes in other model organisms. For example, distinctive negative gradients of GC content, codon usage, and amino-acid usage along the direction of transcription were observed in many rice gene models [2,9]. On the other hand, many predicted rice genes that lack significant homology to genes in other organisms also exhibit characteristics such as unusual GC composition and distribution, suggesting that they might not be true genes [10,11]. Furthermore, the abundance and diversity of transposable elements (TEs) within the rice genome that possess a coding capacity pose an additional challenge to accurate annotation of the rice genome [10,12,13].
As such, our understanding of the rice genome is largely limited to the state-of-the-art gene prediction and annotation programs. This is probably best reflected by the lack of a consensus of the estimation of the total gene number in rice [6-8,10,11]. Estimated total gene number based on the draft sequences of japonica and indica ranged widely from 30,000 to 60,000 [1,2]. Finished sequences of chromosome 1, 4 and 10 allowed a more finely tuned estimate that placed the total number of rice genes between 57,000 and 62,500 [3-5]. These estimates included a large number of gene models that contain TE-related open reading frames (ORFs). Excluding the TE-related ORFs could reduce the gene number to about 45,000 [6-8]. Even then, between one third and one half of the predicted genes appear to have no recognizable homologs in the other model plant Arabidopsis thaliana [6-8]. Further, aggressive manual annotations of portions of the finished rice sequence have disqualified many of the low-homology gene models as TE-related or artifacts, arguing that there are no more than 40,000 nonredundant genes in rice [10].
Experimental evidence such as full-length cDNA sequences and expressed sequence tags (ESTs) is critical for evaluation and improvement of the genome annotation [14-16]. Large collections of rice full-length cDNA and ESTs are available [15,17]; however, given the large number of rice genes, current methods for collecting expressed sequences do not provide the necessary depth of coverage. For example, based on high-stringency alignments to EST sequences available at that time, only 24.7% of the 3,471 initially predicted genes of chromosome 10 were matched [5]. Conversely, other experiment-oriented approaches, such as massively parallel signature sequencing [18], are able to provide sufficient coverage of the transcriptome but by their nature are limited in their ability to define gene structures. Thus, it is important to survey the transcriptome using additional experimental means that permit detailed analyses of current gene models and the identification of new models.
Recent studies in several model organisms have demonstrated the utility of tiling microarrays in transcriptome identification [19-27]. Armed with new microarray technologies, it is now possible to prepare high-density oligonucleotide tiling microarrays to interrogate genomic sequences irrespective of their annotations. Consequently, results from these studies indicate that a significant portion of the transcriptome resides outside the predicted coding regions [19-21,24,25]. In addition, these studies show that tiling microarrays are able to improve or correct the predicted gene structures [19,23,26]. Based on considerations of feature density, versatility of modification, and compatibility with our existing conventional microarray facility, the maskless array synthesizer (MAS) platform [24,26,28,29] was chosen for our rice transcriptome analysis.
Here we report the construction and analysis of two independent sets of custom high-density oligonucleotide tiling microarrays with unique 36-mer probe sequences tiled throughout the nonrepetitive sequences of chromosome 10 for both japonica and indica rice. Hybridized with a mixed pool of cDNA targets, these tiling microarrays detected over 80% of the annotated nonredundant gene models in both japonica and indica, and identified a large number of transcriptionally active intergenic regions. These results, coupled with comparative gene model mapping and reverse transcription PCR (RT-PCR) analysis, allowed the first comprehensive identification and analysis of a rice chromosomal transcriptome. These results further revealed an association of chromosome 10 transcriptome regulation with the euchromatin-heterochromatin organization at the chromosomal level.
Results
Rice chromosome 10 oligonucleotide tiling microarrays
Based on recent studies using MAS oligonucleotide tiling microarrays to obtain gene expression and structure information [24,26,28,29], we designed two independent sets of 36-mer probes, with 10-nucleotide intervals, tiled throughout both strands of japonica and indica chromosome 10, respectively. After filtering out those probes that represent sequences with a high copy number or a high degree of complementarity, 750,282 and 838,816 probes were retained to interrogate the entire nonrepetitive sequences of japonica and indica chromosome 10 and were synthesized in two sets of MAS microarrays [24,26,29]. The arrays were hybridized with target cDNA prepared from equal amounts of four selected poly(A)+ RNA populations (the N Arrays), namely, seedling roots, seedling shoots, panicles, and suspension cultured cells of the respective rice subspecies. In addition, a set of japonica arrays was hybridized to shoot poly(A)+ RNA derived from seedlings with a mineral/nutrient disturbance (the S Arrays).
Our MAS microarrays utilize a 'chessboard' design, meaning that each positive feature, which contains an interrogating probe, is surrounded by four negative features and vice versa [24,26]. Given that both positive and negative features contain a linker oligo to which the interrogating probes were synthesized, it was possible to determine signal probes (those that detect an RNA target) using a two-step procedure. After normalization (Figure 1a,b), positive features with fluorescence intensities lower than the mean intensity of the four surrounding negative features were masked. A characteristic bimodal intensity distribution of the remaining positive features was observed for each microarray (Figure 1c). Based on a statistical model to reject noise probes at a 90% confidence (see Materials and methods), signal probes and their normalized fluorescence intensities were determined (Figure 1c). Signal probes were correlated with the transcriptionally active regions (TARs) of the chromosome by alignment of the probes to the chromosomal coordinates (Figure 2). Experimental identification of the transcriptome was then achieved by systematically examining the expression of the annotated gene models and screening for intergenic TARs.
Rice chromosome 10 gene models
Finished sequences have been determined for both japonica and indica chromosome 10 [5-8]. Initial annotation of japonica chromosome 10 produced 3,471 protein-coding gene models [5], which was updated to 3,856 in the release 2 of the Rice Pseudomolecules from The Institute for Genomic Research (TIGR) [8]. Of these, 829 (21.5%) were found to be TE-related models. Eight gene models were mapped to other chromosomes, and were not included in this study. Classification of the 3,019 nonredundant protein-coding gene models was based on alignments to the rice full-length cDNA and ESTs [15,17]. These analyses led to the identification of 935 (31.0%) cDNA-supported gene (CG) and 321 (10.6%) EST-supported gene (EG) models. The remaining 1763 (58.4%) models were classified as unsupported gene (UG) models. This model set is designated TIGR japonica (Table 1, Figure 2 and see Additional data file 1).
For comparison, the so-called BGI japonica gene models were included, whereby the japonica chromosome 10 sequence was independently annotated by the Beijing Genomics Institute (BGI) [6,30]. This model set, generated by the FGENESH output with limited full-length cDNA/EST input, contains 851 TE, 943 CG, 272 EG, and 1,549 UG models (Table 1, Figure 2). To analyze the indica chromosome 10 transcriptome, and for comparative analysis, the BGI indica models were also examined [2,6,30]. Classification of the indica models identified 574 TE, 821 CG, 328 EG, and 1,660 UG models (Table 1, Figure 2 and see Additional data file 2).
Tiling microarray detection of rice chromosome 10 gene models
Analysis of the N arrays detected 2,428 out of 2,809 BGI indica (86.4%), 2,319 out of 2,764 BGI japonica (83.9%), and 2,472 out of 3,019 TIGR japonica (81.9%) nonredundant gene models (Table 1). Although no technical replication was performed, several observations indicate that tiling microarray analysis provides a reliable evaluation of the expression of the gene models. First, consistent with their classification, gene models with previous experimental support (CG and EG) showed a higher detection rate than the unsupported models (Table 1). For example, 93.2% and 90.7% of the TIGR japonica CG and EG models were detected, respectively, whereas only 74.3% of the UG models were (Table 1). Second, supported models (CG and EG) exhibited very similar array detection rates across the three sets of gene models. Because the same cDNA and ESTs were used to classify the three sets of gene models, this result implies a strong correlation between tiling microarray detection and expressed sequences. In supporting of this conclusion, TIGR japonica models with at least one match with rice EST sequences exhibited a 92.7% (1,010 of 1,089) detection rate whereas only 75.7% (1,458 of 1,925) models without a matching EST were detected. Third, examination of signal probe distribution, measured by hybridization rate (HR, see Materials and methods), in the annotated exonic and intronic regions indicates that the tiling microarrays detected transcription predominantly locate in the exons. Across the three annotations, the HRs of both the intronic regions (dashed lines) and exonic regions (solid lines) showed bimodal distributions, with their respective major peaks well separated (Figure 1d). The minor intronic HR peak likely reflects transcriptional activities of exons misidentified as introns or in uncharacterized splice variants. Conversely, the minor exonic HR peak is likely to be due to misinterpretation of introns as exons, or exons or genes not expressed at all in the RNA populations used (Figure 1d).
Analysis of previously unsupported gene models
The relatively poor detection rate for the unsupported models suggests that their expression may be more restricted to specific cell types or developmental stages, thus eluding tiling array detection. Alternatively, some of these UG models might be false and do not represent real genes. For further analysis, gene models were classified as high homology (HH) and low homology (LH) models based on comparison using an expect value of e-7 for predicted protein homology between rice and Arabidopsis [6]. It should be noted that the simple sequence alignment is likely to fail to detect some structural homology. However, this simple division is useful for separating two groups of gene models for expression comparison. For example, in the BGI japonica annotation, there are 589 UG/HH and 960 UG/LH models. By comparison, our tiling microarray detected 495 (84.0%) UG/HH models, but only 707 (73.7%) UG/LH models. Because the UG/LH models lack any previous supporting evidence (either homology or expression), concerns have been raised as to whether they represent real genes [10,11]; therefore, the expression properties of the UG/LH models are of particular interest for further evaluation.
To investigate the possibility that expression of some UG/LH models is restricted to special conditions, we analyzed the S Arrays with regard to UG model expression. Of the gene models in the BGI japonica annotation, 63.4% were detected in seedling shoots under a variety of stress conditions that are known to significantly alter gene expression profiles [31,32]. These included 39 (2 CG/HH, 2 EG/HH, 8 UG/HH, 2 CG/LH, 2 EG/LH and 23 UG/LH) models that eluded detection by the N Arrays. The enrichment of UG/LH models in S Arrays-specific models indicates that some UG/LH models indeed have specialized expression. Though it is entirely possible that additional UG/LH models could be detected under other stress conditions, the small number of UG/LH models specifically detected from the S Arrays (23 of 960, or 2.4%) suggests that specialized expression of UG/LH models alone may not account for the overall low detection rate of the UG/LH models.
In a separate approach to verify UG model annotation, 589 UG models were randomly selected for a high throughput RT-PCR analysis. Overall, 196 (33.3%) of the selected UG models were cloned and sequence-confirmed from the same RNA samples used for the N Arrays (Figure 3a and Additional data file 3). Given that only 62% (49/79) of CG models were successfully cloned and sequence-confirmed in a control experiment, these results suggest that expression of approximately half (33% over 62%) of the UG models can be confirmed in our experimental conditions. Closer inspection of the confirmed UG transcripts showed that only 102 (52%) contain an identical ORF as predicted, whilst 94 (48%) exhibit different ORFs compared to the predictions (Figure 3a,c), suggesting that the gene structure of about half of the UG models need to be corrected or improved. Since the tiling microarrays used in this study have limited ability to pinpoint precise intron-exon junctions, transcript cloning and sequence analysis are still required to verify the annotated gene structures.
Identification and analysis of intergenic TARs
We found that 10.26% and 11.75% of the probes in the japonica and indica N Arrays were considered signal probes, respectively (Figure 1c). Approximately 55% and 15% of these signal probes were found to locate in the intergenic and intronic regions, respectively, of the TIGR japonica, BGI japonica, and BGI indica annotations. These results indicate that, irrespective of different annotations, significant transcriptional activities locate in the annotated intergenic regions. A sliding-window-based approach was used to systematically identify intergenic TARs (see Materials and methods). Through this analysis, 574 and 522 intergenic TARs in indica and japonica were identified from the N Arrays, respectively. In addition, 466 unique intergenic TARs were identified from the S Arrays, bringing the total number of japonica intergenic TARs to 988. These TARs have a cumulative length of approximately 700 Kb or 3% of the chromosome. The average length of the intergenic TARs was about 700 bp (Figure 4a and Additional data file 4).
Several lines of evidence support the idea that the majority of intergenic TARs represent legitimate elements of the rice transcriptome. Sequence analysis revealed that 301 (55.0%) indica and 455 (46.0%) japonica intergenic TARs possess a significant coding capacity (more than 50 amino acids). Selected intergenic TARs were used as probes in RNA gel-blot analysis to confirm expression of these TARs. Overall, 26 out of 34 probes detected a discrete band, with tissue specificity, whereas the rest failed to detect any, suggesting that the majority of the intergenic TARs correspond to in vivo transcripts rather than being caused by cross hybridization (Figure 4b-d). A total of 280 intergenic TARs were selected for further analysis using an RT-PCR strategy designed to clone transcripts containing an intergenic TAR and its entire downstream (3') sequence (see Materials and methods and Additional data file 5). Of the 77 cloned transcripts whose sequences could be unambiguously confirmed, 37 overlap with existing gene models (Figure 3b,d), suggesting they are uncharacterized portions, such as 5' or 3' untranslated regions (UTRs), or splice variants of the neighboring gene models. The rest of the confirmed transcripts (40 out of 77) were located entirely in intergenic regions, suggesting that they likely represent independent novel transcriptional units (Figure 3b,d).
To further characterize the 988 japonica intergenic TARs, they were aligned to the output of the rice gene finder BGF [2,6,30] using the japonica chromosome 10 sequence, and 72 novel gene models were identified (Additional data file 1). Comparison with the cloned intergenic TARs showed that 23 of the 40 cloned novel transcripts (57.5%) were also predicted in the novel BGF models (Figure 3b), indicating that the BGF program was able to detect half of the potential novel genes represented by the intergenic TARs. However, the incomplete nature of the 17 unaccounted transcripts (Figure 3b) made it difficult to unambiguously determine whether they encode proteins.
Tiling microarray-based gene model comparison and integration
The TIGR model set contained 200-250 more gene models than the BGI sets (Table 1). These extra models were evenly distributed into HH and LH models (Figure 5a). The TIGR/HH models showed a similar array-detection rate, while the TIGR/LH models were detected at a lower rate (but of a similar number) in comparison with the two BGI sets (Figure 5a). This result suggests that the extra TIGR/LH models may be of low confidence and need to be further examined. Comparison of the BGI and TIGR japonica models indicates that there were 2323 (84.0%) and 2488 (82.4%) common to each annotation, respectively, based on ORF sequence overlaps (Additional data file 6). Meanwhile, 441 (16.1%) BGI models and 531 (17.6%) TIGR models were regarded as unique to each annotation (Additional data file 6). Naturally, the common models are more reliable, and were consequently enriched with expression- or homology-supported models. For example, only 64.5% of the unique TIGR models were detected by tiling microarrays. However, expression of 363 of the unique BGI models was confirmed by tiling array and/or cDNA and EST alignment, indicating that they are part of the japonica chromosome 10 transcriptome (Figure 5b).
The indica gene models were more evenly distributed along the chromosome, and the number and distribution of array-detected models was similar to that of japonica (Figure 6a-c). Exceptions were noted in certain regions, such as at approximately10 Mb, where indica models showed increased array detection rates. Such a disparity is likely to be caused by the skewed distance between corresponding japonica/indica model pairs (see below). Comparative gene model mapping indicates that 97.6% of the japonica chromosome10 CG/HH models had their counterparts in indica, while 98.3% of the indica CG/HH models were mapped to japonica (Additional data file 6 and data not shown). As the full-length cDNAs were derived from japonica [15], this result suggests that roughly 2% of either genome sequence was erroneous or incomplete, thereby disrupting the integrity of the affected genes such that they could not be recognized. However, only 85.3% and 88.1% of japonica and indica UG/LH models could be mapped to their reciprocal genomes. These results indicate that the unmapped UG models between japonica and indica were common but not recognized in the reciprocal genomes, or subspecies specific, or false predictions. Thus, identification of the first group of models would facilitate a better recognition of the transcriptome of both genomes. Indeed, 2,640 indica models were mapped to japonica chromosome 10 (Additional data file 7). Among those mapped indica models, 114 were detected by tiling array, with corresponding genome sequences that were more than 95% identical to that of japonica chromosome 10, but were not annotated in japonica. These results suggest that the counterparts of these 114 indica models may exist in the japonica chromosome 10 transcriptome (Figure 5b).
To provide a comprehensive representation of the japonica chromosome 10 transcriptome, the 549 new models, including 363 BGI japonica models, 114 BGI indica models, and 72 novel BGF models (see above), were integrated with the TIGR japonica gene models (Figure 5b). The resulting 3,568 nonredundant protein-coding gene models, including the 3,019 TIGR models, represent an 18% increase in the annotated coding capacity of japonica chromosome 10 (Figure 5b). The integrated models included 3005 (84.2%) that were detected by tiling arrays, of which, 1,120 (31.4%) were not previously supported by expression data or homology. Thus, 3,255 (91.2%) models in the integrated set now have at least one piece of supporting evidence (for example, expressed sequences, homology, or tiling microarray) (Figure 5c). Classification of the array-detected and undetected models, based on exon number, homology to Arabidopsis genes, and previous supporting evidence, indicates that detection by our tiling microarray was not biased regarding gene structure and was in general agreement with all other annotation information (Figure 5c). These results demonstrate tiling microarray analysis as a useful platform to validate and incorporate information from multiple sources to fully identify the rice transcriptome.
Heterochromatin-associated regulation of chromosome-wide transcriptional activity
We applied the tiling microarrays to study chromosomal position effects on gene expression. As shown in Figure 6, chromosome-wide gene model distribution and expression suggests that chromosome 10 can be divided into two roughly equal-sized domains, with domain I consisting of the short arm and the proximal end of the long arm, while domain II encompasses the rest of the chromosome. This division was based on transcriptional profiles of the two domains, as revealed by tiling microarray analysis (Figure 6). Domain II had a higher density of nonredundant gene models (Figure 7a). Under normal growth conditions (the N Arrays), it also contained more signal oligos and more array-detected models and thus was more transcriptionally active relative to domain I (Figure 6). Such a distinction between the two domains was further supported by the higher number of CG models in domain II, which are presumably highly expressed (Figure 7b). Interestingly, although only a small number of gene models were specifically detected from the S Arrays (see above), overall transcriptional activity in domain I was elevated under the examined stress conditions (Figure 6d). The activation was observed both at the individual gene model level and in 100 kb windows across domain I (Figure 6d). Such a general derepression of transcription under stress conditions may imply another layer of gene regulation at the chromosomal level in rice.
The observed transcriptional profiles of the two domains were associated with several architectural features of the chromosome. In general, domain I was more enriched with TE and LH models (Figure 7a,c). Domain I also harbored more repetitive sequence, as was evident from the greater number of oligos masked during array design (Figure 6a). To further examine the two domains, colinearity of the CG models in chromosome 10 of japonica and indica rice was calculated. Mapping chromosomal positions of corresponding orthologous CG model pairs along chromosome 10 of japonica (blue) and indica (red) against the sequential orders of the CG pairs resulted in two apparently smooth parallel curves (Figure 8a). This observation indicates that the order of CG models is well preserved between chromosome 10 of japonica and indica rice. However, calculation of the physical distance between corresponding japonica and indica CG models along the chromosome indicated that the positions of the CG models were more skewed in domain I, with many CG models shuffled more than 1 Mb away from their orthologous counterparts in the reciprocal chromosome (Figure 8b).
These results coincide with cytological data showing that domain I is primarily heterochromatin, whereas domain II is primarily euchromatin [5,33]. Although it remains to be seen whether the phenomena mentioned above are general features associated with the division of heterochromatin and euchromatin in rice, these results collectively indicate that the heterochromatic domain of chromosome 10 is more evolutionarily active and compositionally dynamic. Our results further indicate that the genomic characteristics of the heterochromatin domain are associated with its transcriptional activities (Figure 6).
Discussion
Sequencing of the rice genome provides a cornerstone to understand the biology of this agriculturally important crop [1-8,34-36]. A first step in fully realizing the potential of available genome sequence is to understand its coding information and expression; however, current annotated gene models and other functional elements of a genome by and large represent hypotheses that must be experimentally tested and validated. Importantly, approximately 20,000 predicted rice genes exhibit no recognizable sequence homology to genes in other organisms, especially Arabidopsis, the first model plant sequenced [1-8]. The unusual compositional and structural features, as well as the lack of EST coverage for a large number of novel genes, require high-throughput experimental means that are not limited by the current annotations.
Identification of the rice chromosome 10 transcriptome by tiling microarrays
In this study, we developed whole-chromosome oligonucleotide tiling microarrays, and demonstrated their utility in experimentally identifying the transcriptome of both japonica and indica chromosome 10. Because oligonucleotide tiling microarrays provide unbiased end-to-end coverage of the entire chromosome and measure transcriptional activity of gene models from multiple independent probes (Figure 2), they can detect the transcriptome in a comprehensive and unbiased way [19-21,23-25]. The tiling microarray analysis of rice chromosome 10 detected transcription of 86.4% BGI indica (2,428/2,809), 83.9% BGI japonica (2,319/2,764), and 81.9% TIGR japonica (2,472/3,019) gene models (Table 1). Using a set of the least reliable gene models (UG models, see below), RT-PCR analysis revealed disparity in gene structure of close to 50% of these models (Figure 3). These results are consistent with previous assessments of current computational gene finders, which can reliably locate a gene model in the correct chromosome locus, but are less than satisfactory to predict the fine gene structure [37,38].
Based on alignment to rice full-length cDNA and EST sequences, the gene models for both japonica and indica chromosome 10 were classified as UG, EG, and CG models (Table 1, Figure 2). This classification places the gene models in three groups with an ascending order of confidence, because the presence of an expressed sequence provides strong support to the corresponding model. In keeping with this idea, these three classes of gene models were also detected by tiling microarrays in an ascending order (Table 1). This result, together with the high detection rate of CG models, suggests that the chromosome 10 transcriptomes identified by the tiling microarrays are rather exhaustive. In support of this conclusion, tiling array analysis of rice seedlings which had undergone severe stress treatments only identified an additional 39 (less than 1.7% of the total detected) models. These results likely can be attributed to the high sensitivity of the tiling microarrays such that even if activation of certain genes is conditional, the basal level transcripts could still be detected by the tiling microarray.
Therefore, the UG models (particularly UG/LH) that failed to be detected by the tiling microarray need to be more closely inspected (Table 1, Figure 3). We did find that the gene models specifically detected following the stress treatments were enriched with UG/LH models (23/39), suggesting that some UG/LH might be stress responsive and their expression is not readily detectable under normal conditions. It should be noted that though redundant gene models such as those derived from long terminal repeat (LTR) retrotransposons and Pack-MULEs are generally under-represented in the expressed sequence collections [12,39], many are stress responsive and share similar cis-elements with plant defense genes [40]. Thus, it cannot be ruled out that some of the UG/LH models are related to low copy number retrotransposons with unusual structures.
Reasoning that the tiling microarray-detected transcriptome is both exhaustive and reliable, tiling microarray-supported gene models were mapped and integrated. This analysis identified 363 unique BGI japonica, 114 unique BGI indica, and 72 novel models that could be integrated into the TIGR japonica gene model set to comprehensively represent the japonica chromosome 10 transcriptome (Figure 5). Note that the added gene models do not necessarily increase the number of japonica chromosome 10 genes, even if their transcription was detected. As elaborated above, some of these gene models could be unrecognized TEs, uncharacterized UTRs or alternative exons. However, as all these extra gene models are transcribed, their identification will not only better represent the transcriptome, but further examination of these elements will also yield insight into rice genome composition and structure.
Extensive antisense transcription was observed for the rice chromosome 10 gene models. For instance, in a preliminary analysis whereby regions of the antisense strand covering the 3,019 TIGR japonica gene models were examined, excluding those that contain less than three signal oligos, 591 (19.6%) were found to have antisense expression. The proportion of rice gene models showing antisense transcription is consistent with that reported from tiling microarray analyses in Arabidopsis [23] and human [24,25], adding to an increasing body of evidence that indicates antisense transcription as an inherent property of the genomes. However, it should be cautioned that the potential effects of several experimental artifacts such as unintended second-strand synthesis, formation of specific RNA-DNA hybrids, or spurious priming events during target preparation have to be precisely assessed before a final conclusion on the nature and extent of antisense transcription in rice can be drawn.
Transcriptional activities outside the annotated gene models in the form of intergenic TARs, accounted for approximately 3% of the chromosome size (Figure 4a). RNA gel blotting and RT-PCR analyses confirmed only a portion of the selected TARs (Figure 3, 4), suggesting that the unconfirmed TARs could be experimental artifacts or correspond to transcripts of extreme low abundance [21,25,27]. Transcriptome components outside of previously annotated gene models are expected to correspond to: novel genes with unusual sequence composition; under-represented UTRs or exons of splice variants; nonprotein coding RNA transcripts; or uncharacterized transcribed TEs. RT-PCR analysis of selected japonica intergenic TARs suggests that the majority of the TARs belong to the first two groups (Figure 3b). This conclusion is consistent with the observation that the intergenic TARs were slightly enriched in regions of the chromosome with lower gene density (Figure 7d). A preliminary analysis whereby 214 plant miRNAs (including 122 from rice and 92 from Arabidopsis) [41,42] were used in a BLAST search against the intergenic TARs revealed no significant hits, suggesting that the TARs do not contain known plant microRNAs.
We thus focused our efforts on further analyzing the first two groups of TARs. For the current rice annotation, five different gene finders (primarily FGENESH) were used to generate gene models [8]. To annotate the intergenic TARs, we used the relatively new rice gene-finder program BGF [2,6,30], which identified 72 novel gene models (Figure 5). Sequence comparison between the 40 cloned intergenic TAR transcripts and the novel BGF models showed that 23 (57.5%) were predicted (Figure 3b), indicating that the BGF program was able to detect slightly more than half of the novel transcriptional units that might be represented by the intergenic TARs. Extrapolation from these observations suggests that there might be up to 2,000 novel genes yet to be recognized by current rice gene finders; however, the incomplete nature of the cloned transcripts made it difficult to unambiguously determine whether they encode proteins. Thus, it is possible that some of these transcripts may correspond to noncoding RNAs.
Association of chromosomal architecture with transcriptional activity
Eukaryotic genomes contain heterochromatin as cytologically intensely staining nuclear materials that are thought to be composed mainly of noncoding DNA and silent transposons [33,43]. A salient feature of rice chromosome 10 is that its heterochromatin is not limited to the pericentric regions, but includes the entire short arm as well as the proximal portion of the long arm [33]. Comparison of cytological and sequence data suggests that this heterochromatin region is roughly 11-12 Mb in length [5,33]. Although recent genetic and microarray studies in plants have indicated a role for gene regulation by well defined small heterochromatin regions [44-47], virtually no data are available regarding the association of transcriptional activity with large-scale heterochromatin domains in regulating gene expression, chromosome behavior, and genome evolution.
Profiling the transcriptional activities of rice chromosome 10 using tiling microarrays revealed that gene expression in the heterochromatin region is generally low under normal growth conditions (the N Arrays) relative to the euchromatin (Figure 6a-c). Consistent with this observation, gene model distribution showed that the heterochromatin domain is relatively low in CG models but more abundant in UG models (Figure 7b). In support of the cytological data, an enrichment of TE models in the heterochromatin domain is evident (Figure 7a) [5]. Exclusion of the high copy number TEs and repetitive sequences from the tiling microarray analysis might contribute to the lower gene model density in the heterochromatin (Figure 7a-c); however, the generally lower detection rate of gene expression indicates that expression of many non-TE models is also somewhat repressed (Figure 7a-c). Interestingly, when plants were subjected to mineral or nutrient stresses, a general activation of transcription was observed in the heterochromatin (Figure 6d). These results are consistent with findings that heterochromatin stability and heterochromatin-mediated gene silencing can be regulated by development [48,49] or by modulating levels of specific transcription factors [50].
The distribution of TE and non-TE gene models in the heterochromatic and euchromatic regions was a near mirror image (Figure 7a). This result suggests that the heterochromatin and euchromatin may have similar capacities to accommodate protein-coding gene models (TE and non-TE), even though the heterochromatin is enriched with repetitive sequences (Figure 6a) [5]. Furthermore, the heterochromatin is relatively enriched with LH models and low in CG models compared with the euchromatin (Figure 7b, c). Thus it is likely that the differential packaging of genome elements in heterochromatin and euchromatin might enable rice to regulate and coordinate gene expression at the chromosomal level. Although the underlying molecular mechanism of this regulation is currently unknown, DNA methylation, histone modifications, and small interfering RNAs have all been implicated [51-55].
The distance between corresponding japonica and indica CG models along the chromosome was more skewed in the heterochromatin, with many CG genes shuffled more than 1 Mb in physical distance from the location of their orthologous counterparts. In contrast, the gene distance in the euchromatin is largely homogeneous (Figure 8). Previous studies have shown a mosaic organization of grass genomes where conserved sequences are disrupted by nonconserved sequences, and that gene amplification, movement, and activity of retrotransposons account for the bulk of the interspersing nonconserved sequences [56-58]. Thus, these results collectively indicate that the heterochromatin domain is more evolutionarily active and compositionally dynamic. Such a conclusion is in keeping with the genomic stress hypothesis that TEs are involved in host adaptation to environmental changes [39,40,59].
Materials and methods
Plant materials and treatments
Oryza sativa ssp. japonica cv. Nipponbare and Oryza sativa ssp. indica cv. 93-11 were used for all experiments. Seeds were surface-sterilized, imbibed at 37°C for 2 days, and then transferred to MS medium (Invitrogen) solidified with 0.8% (w/v) agar. Seedlings were kept under continuous light at 28°C for seven days before harvest for total RNA isolation. Alternatively, 7-day-old seedlings were transferred to soil and maintained under long-day conditions (16 h light/8 h dark) at 26-28°C in the greenhouse until flowering. Heading and filling stage panicles were then collected from these plants. Suspension-cultured cells were prepared and maintained as previously described [60]. For stress treatment, japonica seedlings were grown for seven days on MS medium under four different conditions: MS medium deprived of nitrogen; MS medium deprived of phosphorus, or supplemented with 150 mM NaCl or 100 μM CdSO4. For RNA isolation, plant materials were frozen in liquid nitrogen and homogenized. Total RNA and mRNA were isolated using the RNeasy Plant Mini kit (Qiagen) and the Oligotex mRNA kit (Qiagen) according to the manufacturer's recommendations, respectively.
MAS microarray design, production, and hybridization
Based on the MAS platform, a minimal tiling strategy was designed to effectively represent the nonrepetitive sequences of rice chromosome 10 [24,26]. Briefly, 36-mer oligonucleotides were designed using an algorithm based on sequence-dependent factors such as length, extent of complementarity, and the overall base composition. Oligos that could form a stem-loop structure with stem length greater than seven bases and those that have an oligo index score greater than 5 were excluded. To calculate the index score for each oligo, the 20 possible consecutive 17-mer sequences within each oligo were searched against the whole genome. The average copy number of the 17-mer sequences was scored as the oligo index. MAS microarray production was performed as previously described [24,26,29] using the sequences of chromosome 10 for japonica and indica rice as were available on 12 April, 2004 [8] and 1 August, 2003 [6,30], respectively. Oligos were synthesized at a density of 389,000 oligos per array in a chessboard design wherein each positive feature, which contains an interrogating oligo, was surrounded by four negative features and vice versa.
The japonica and indica N Arrays both included four individual MAS arrays that contain oligos representing other portions of the genome (other than chromosome 10) not analyzed in the current study. The N Arrays were hybridized to cDNA target mixtures derived in equal amounts from seedling roots, seedling shoots, panicles, and suspension-cultured cells of both japonica (cv. Nipponbare) and indica (cv. 93-11) rice. Additionally, a set of two japonica arrays (S Arrays) were hybridized to targets derived from pooled poly(A)+ RNA isolated from leaves of stress-treated japonica seedlings. Target preparation, array hybridization, and hybridization intensity value acquisition were carried out as previously described [24,26,29,61]. Tiling microarray design and experimental data are available in the National Center for Biotechnology Information (NCBI) Gene Expression Omnibus under series GSE2500.
Chromosome 10 gene model compilation
The japonica (TIGR Rice Pseudomolecule released on 12 April 2004) [8] and indica (released by BGI on 1 August 2003) [6,30] chromosome 10 annotations were used in this study. In addition, the japonica chromosome 10 sequence was annotated using the BGI gene prediction flow to generate the BGI japonica gene model set. All gene models were aligned to a collection of rice full-length cDNA sequences [15] and all available rice EST sequences in GenBank [17] as of 15 April 2004 by the BLAT program [62] using cutoff criteria of 100 bp overlap and 90% identity over the entire length of each match. The predicted genes without matches to cDNA and EST sequences, excluding those with coding capacities of less than 50 amino acids, were classified as UG models.
Determination of gene model expression and identification of intergenic TARs
Hybridization intensity of all positive and all negative features within each array was plotted separately and then scaled to have a peak log2 intensity of 8.0 (Figure 1a,b). Signal and noise probe determination is shown in Figure 1c and discussed in main text. Expression level of a given gene model was represented by the value of hybridization intensity (HI) of this model locus that takes into account two parameters: FI, which is the mean of fluorescence intensity of all signal probes of a given gene model, and hybridization rate (HR), which is defined as the percentage of signal probes over total interrogating probes per kilobase of genomic sequence. HI is calculated using the formula HI = FI + FI × (HRE - HRM) in which HRE is HR of the exon regions whilst HRM is the mean HR of all intron regions. HI value of each model was then compared against a threshold designated as the mean fluorescence intensity plus twice the standard deviation (95% confidence) of all noise probes within each array.
To identify intergenic TARs, HR was calculated in a sliding window of 500 nucleotides across the intergenic regions of chromosome 10 with a bandwidth equal to an interrogating probe. Windows with HR above a threshold of 0.4 were considered positive. Contiguously transcribed regions (TARs) were generated by joining overlapping positive windows that were delineated by the 5' probe of the first window and 3' probe of the last. TARs less than 220 bp (five consecutive probes) long were discarded. The japonica intergenic TARs were first identified using the BGI japonica annotation, followed by comparison with TIGR models. TARs overlapping with TIGR models were masked. Sequences of all retained intergenic TARs were aligned to the BGF gene predictions, and were used to BLASTX search the nonredundant protein database SWISS-PROT. Those BGF-predicted genes that overlap more than 100 bp with the sequence of intergenic TARs on the same strand of DNA were considered positive.
Cloning and verification of UG models and intergenic TARs
Selected UG models were cloned by means of RT-PCR. The PCR products were cloned into the pGEM-T vector (Promega) and sequenced. To clone intergenic TARs with downstream sequence, reverse transcription was performed on mixed poly(A)+ RNA derived from seedling roots, seedling shoots, panicles and suspension-cultured cells of japonica rice using the primer RT-CPK (5'-TGCAGTCTAGCTGGAATGACCTCATTGCAGAAT24). The PCR procedure to clone the TARs was carried out using a cascade of thermal asymmetric interlaced PCR cycles [63,64] that employ three consecutively nested gene-specific primers to pair with primer RT-1 (5'-GCAGTCTAGCTGGAAT), RT-2 (5'-CTGGAATGACCTCATT), and RT-3 (5'-GCTGGAATGACCTCATTGCAGAAT), which anneal to overlapping regions of RT-CPK. Sequences of all the cloned PCR products were aligned back to japonica chromosome 10 using BLAT [62] to confirm their identify and to map their corresponding gene structure. RNA gel-blot analysis of intergenic TARs was conducted as previously described [65].
Integration of japonica chromosome 10 gene models
All japonica chromosome 10 related gene models were sorted, and only those that met certain criteria were retained. The TIGR nonredundant gene models that can be mapped to the japonica chromosome 10 sequence were all retained. The additional models included BGI japonica, BGI indica models mapped to japonica chromosome 10, and tiling array-derived novel BGF models. From these models, those without previous full-length cDNA/EST or tiling microarray support, or those overlapping with TIGR models were discarded. All retained models were aligned back to the japonica chromosome 10 sequences to further confirm their identities and were combined with the TIGR japonica models.
Additional data files
The following additional data files are available with the online verison of this paper. Additional data file 1 contains a table of integrated japonica chromosome 10 nonredundant gene models. Additional data file 2 contains a table of indica chromosome 10 nonredundant gene models. Additional data file 3 contains a table of the sequence analysis of cloned UG models. Additional data file 4 contains japonica chromosome 10 intergenic TARs. Additional data file 5 contains the sequence analysis of cloned intergenic TARs. Additional data file 6 contains a comparison of BGI and TIGR japonica chromosome 10 gene models. Additional data file 7 contains a comparison of BGI indica and japonica chromosome 10 gene models.
Supplementary Material
Additional File 1
Table S1. Integrated japonica chromosome 10 nonredundant gene models. Integrated japonica chromosome 10 nonredundant gene models.
Click here for file
Additional File 2
Table S2: Indica chromosome 10 nonredundant gene models. Indica chromosome 10 nonredundant gene models.
Click here for file
Additional File 3
Table S3: Sequence analysis of cloned UG models. Sequence analysis of cloned UG models.
Click here for file
Additional File 4
Table S4: Japonica chromosome 10 intergenic TARs. Japonica chromosome 10 intergenic TARs.
Click here for file
Additional File 5
Table S5: Sequence analysis of cloned intergenic TARs. Sequence analysis of cloned intergenic TARs.
Click here for file
Additional File 6
Table S6: Comparison of BGI and TIGR japonica chromosome 10 gene models. Comparison of BGI and TIGR japonica chromosome 10 gene models.
Click here for file
Additional File 7
Table S7: Comparison of BGI indica and japonica chromosome 10 gene models. Comparison of BGI indica and japonica chromosome 10 gene models.
Click here for file
Acknowledgements
We thank Jessica Habashi for critical reading of the manuscript. The rice tiling microarray project at Yale University was supported by a grant from the NSF Plant Genome Program (DBI-0421675). The collaborative research effort in China was supported by the 863-rice functional genomics program from the Ministry of Science and Technology of China, and by the National Institute of Biological Sciences at Beijing. L.L. was initially supported by a Yale University Brown postdoctoral fellowship.
Figures and Tables
Figure 1 Processing the rice chromosome 10 tiling microarray hybridization data. (a) Distribution of fluorescence intensity of all positive and negative features of the four indica N Arrays. (b) All eight distributions were scaled to have a uniform intensity peak value at 8 (log2). (c) Mathematic model for determination of signal probes. A bimodal distribution of log2 background-adjusted intensity of all positive features is used to model the noise as a normal distribution by mirroring the distribution of low intensity (< 6 of log2). A cutoff value corresponding to a 90% confidence level to reject noise probes according to the modeled noise distribution is indicated. (d) Distribution of hybridization rate in the exonic and intronic regions of rice chromosome 10. Hybridization rate (HR) is calculated as the ratio of the number of signal probes against the total number of interrogating probes per kilobase of sequence.
Figure 2 Tiling microarray analysis of the rice chromosome 10 transcriptome. (a) Schematic representation of rice chromosome 10. The purple oval denotes the centromere. (b) A region from the long arm of chromosome 10 displaying the three sets of gene models used: BGI indica; TIGR japonica and BGI japonica. The nonredundant protein-coding gene models are aligned to the chromosomal sequences and color-coded on the basis of their classification (see text). (c) Detailed tiling profile of one representative CG model. The model is represented here as block arrows, which point in the direction of transcription. Signal oligos are aligned according to their chromosomal coordinates. The fluorescence intensity value of each signal oligo, capped at 2,500, is depicted as a vertical bar. The shade of the bar represents the oligo index score (see Materials and methods). The red blocks underneath the bars indicate the presence of an interrogating oligo in the microarray.
Figure 3 Cloning and sequence analysis of japonica chromosome 10 UG models and intergenic TARs. (a) Summary of RT-PCR analysis of selected UG models. ORF identical, annotated ORF is the same as determined from the cloned sequence; ORF different, annotated ORF is different from that in the cloned sequence. (b) Summary of RT-PCR analysis of selected intergenic TARs. Gene model, cloned TARs overlapping with TIGR models; BGF prediction, cloned TARs overlapping with BGF predictions; unique, cloned TARs not overlapping with any annotated feature. (c) Representative UG models whose cloned sequences either differ from (OsJN02936) or are the same as (OsJN03072) the annotated ones. (d) Representative intergenic TARs whose cloned sequences either overlap with a TIGR model (OsJN01855) or are completely intergenic (C10_ZN376). Representation of microarray data in this figure is the same as in Figure 2 except that the oligo index is omitted.
Figure 4 Analysis of intergenic TARs of japonica chromosome 10. (a) The 988 japonica chromosome 10 intergenic TARs distributed by length. (b) RNA gel blotting analysis of selected japonica intergenic TARs. Probes for the intergenic TARs shown in this panel were derived from corresponding PCR-amplified TAR sequences from japonica rice genomic DNA. (c) Probes shown in this panel were derived from RT-PCR amplification of the corresponding TARs from poly(A)+ RNA. (d) The rice cDNAs for eIF4A and actin2 were used as loading controls. 5 μg of RNA from the four sources - root, shoot, panicle, and suspension cell culture - that were used for probing tiling microarrays were used for RNA blot analysis here.
Figure 5 Comparison and integration of chromosome 10 gene models. (a) Number of annotated and array-detected high homology (HH) and low homology (LH) models in the BGI indica, BGI japonica, and TIGR japonica annotations. (b) The 549 new gene models were combined with the 3,019 TIGR models. Origins of the new models are shown on the left. Expression support for the TIGR models is shown on the right. Expressed, models matching full-length cDNA/EST; array-detected, models not supported by the expressed sequences but detected by microarray; undetected, models neither supported by expressed sequences nor detected by microarray. (c) Classification of integrated japonica chromosome 10 gene models based on tiling array detection and exon number (left), homology to Arabidopsis genes (middle), and previous expression or homology support to the models (right).
Figure 6 Rice chromosome 10 gene model distribution and expression. (a) Characterization of TIGR nonredundant protein-coding gene models. Model density, array detection rate, number of signal oligos, number of intergenic TARs, and cumulative length (in kilobases) of masked oligos are calculated in 100-kb windows along the length of chromosome 10, and are represented by color-coded vertical bars. A scale representing the physical length of chromosome 10 is shown at the bottom of the panel. The arrowhead delimits the division of domain I and domain II as indicated in the text. Note that the centromere is located at a position around 7 to 8 Mb in chromosome 10. (b) Gene model density and array detection rate of the BGI japonica annotation. (c) Gene model density and array detection rate of the BGI indica annotation. (d) Comparison of the S Arrays and the N Arrays using the BGI japonica annotation. Log2 (S/N) of the hybridization intensity was calculated for individual models (top) and the mean intensity of all models in 100-kb windows along the length of chromosome 10 (bottom).
Figure 7 Chromosome-wide distribution of gene models and chromosomal elements. (a) Distribution of TIGR japonica nonredundant protein-coding gene models (non-TE) and transposable element-related models (TE) in 1-Mb windows across chromosome 10. The division between domain I and II is indicated by the arrowhead. Note that the centromere is located at around 7 to 8 Mb in chromosome 10. (b) Distribution of BGI japonica CG and UG models in 1-Mb windows across chromosome 10. (c) Distribution of BGI japonica HH and LH models in 1-Mb windows across chromosome 10. (d) Numbers of the TIGR japonica nonredundant protein-coding gene models (TIGR Non-TE) and tiling array-detected intergenic TARs in 1-Mb windows across chromosome 10.
Figure 8 Colinearity of the CG models for chromosome 10 in japonica and indica rice. (a) Chromosomal positions of corresponding CG model pairs along chromosome 10 in japonica (blue) and indica (red) rice are plotted against the sequential orders of the CG pairs. (b) Physical distance between corresponding CG pairs is plotted against their sequential orders along the chromosome.
Table 1 Classification and array detection of rice chromosome 10 gene models
Annotation Nonredundant protein-coding gene model TE
Type Annotated Detected Percentage
BGI indica CG 821 784 95.5%
EG 328 290 88.4%
UG 1,660 1,354 81.6%
Total 2,809 2,428 86.4% 574
BGI japonica CG 943 879 93.2%
EG 272 238 87.5%
UG 1,549 1,202 77.6%
Total 2,764 2,319 83.9% 851
TIGR japonica CG 935 871 93.2%
EG 321 291 90.7%
UG 1,763 1,310 74.3%
Total 3,019 2,472 81.9% 829
Rice chromosome 10 protein-coding gene models were divided into TE and nonredundant models based on available annotations. Because of their repetitiveness, expression of TE models was not assessed. The nonredundant models were further divided into CG, EG and UG models based on their alignment to rice full-length cDNAs and ESTs and their expression assessed by tiling microarray analysis.
==== Refs
Goff SA Ricke D Lan TH Presting G Wang RL Dunn M Glazebrook J Sessions A Oeller P Varma H A draft sequence of the rice genome (Oryza sativa L. ssp japonica). Science 2002 296 92 100 11935018 10.1126/science.1068275
Yu J Hu S Wang J Wong GK Li S Liu B Deng Y Dai L Zhou Y Zhang X A draft sequence of the rice genome (Oryza sativa L. ssp. indica). Science 2002 296 79 92 11935017 10.1126/science.1068037
Feng Q Zhang Y Hao P Wang S Fu G Huang Y Li Y Zhu J Liu Y Hu X Sequence and analysis of rice chromosome 4. Nature 2002 420 316 320 12447439 10.1038/nature01183
Sasaki T Matsumoto T Yamamoto K Sakata K Baba T Katayose Y Wu J Niimura Y Cheng Z Nagamura Y The genome sequence and structure of rice chromosome 1. Nature 2002 420 312 316 12447438 10.1038/nature01184
The Rice Chromosome 10 Sequencing Consortium In-depth view of structure, activity, and evolution of rice chromosome 10. Science 2003 300 1566 1569 12791992 10.1126/science.1083523
Yu J Wang J Lin W Li S Li H Zhou J Ni P Dong W Hu S Zeng C The genomes of Oryza sativa: a history of duplications. PLoS Biol 2005 3 e38 15685292 10.1371/journal.pbio.0030038
IRGSP releases the assembled rice genome sequences
TIGR Rice Genome Annotation
Wong GK Wang J Tao L Tan J Zhang J Passey DA Yu J Compositional gradients in Gramineae genes Genome Res 2002 12 851 856 12045139 10.1101/gr.189102
Bennetzen JL Coleman C Liu R Ma J Ramakrishna W Consistent over-estimation of gene number in complex plant genomes. Curr Opin Plant Biol 2004 7 732 736 15491923 10.1016/j.pbi.2004.09.003
Jabbari K Cruveiller S Clay O Le Saux J Bernardi G The new genes of rice: a closer look. Trends Plant Sci 2004 9 281 285 15165559 10.1016/j.tplants.2004.04.006
Jiang N Bao Z Zhang X Eddy SR Wessler SR Pack-MULE transposable elements mediate gene evolution in plants. Nature 2004 431 569 573 15457261 10.1038/nature02953
Juretic N Bureau TE Bruskiewich RM Transposable element annotation of the rice genome. Bioinformatics 2004 20 155 160 14734305 10.1093/bioinformatics/bth019
Hass BJ Volfovsky N Town CD Troukhan M Alexandrov N Feldmann KA Flavell RB White O Salzberg SL Full-length messenger RNA sequences greatly improve genome annotation. Genome Biol 2002 3 research0029.1 0029.12 12093376 10.1186/gb-2002-3-6-research0029
Kikuchi S Satoh K Nagata T Kawagashira N Doi K Kishimoto N Yazaki J Ishikawa M Yamada H Ooka H Collection, mapping, and annotation of over 28,000 cDNA clones from japonica rice. Science 2003 301 376 379 12869764 10.1126/science.1083523
Castelli V Aury JM Jaillon O Wincker P Clepet C Menard M Cruaud C Quetier F Scarpelli C Schachter V Whole genome sequence comparisons and 'full-length' cDNA sequences: a combined approach to evaluate and improve Arabidopsis genome annotation. Genome Res 2004 14 406 413 14993207 10.1101/gr.1515604
NCBI Expressed Sequence Tags Database
Meyers BC Vu TH Tej SS Ghazal H Matvienko M Agrawal V Ning J Haudenschild CD Analysis of the transcriptional complexity of Arabidopsis thaliana by massively parallel signature sequencing. Nat Biotechnol 2004 22 1006 1011 15247925 10.1038/nbt992
Shoemaker DD Schadt EE Armour CD He YD Garrett-Engele P McDonagh PD Loerch PM Leonardson A Lum PY Cavet G Experimental annotation of the human genome using microarray technology. Nature 2001 409 922 927 11237012 10.1038/35057141
Selinger DW Cheung KJ Mei R Johansson EM Richmond CS Blattner FR Lockhart DJ Church GM RNA expression analysis using a 30 base pair resolution Escherichia coli genome array. Nature Biotechnol 2000 18 1262 1268 11101804 10.1038/82367
Kapranov P Cawley SE Drenkow J Bekiranov S Strausberg RL Fodor SP Gingeras TR Large-scale transcriptional activity in chromosomes 21 and 22. Science 2002 296 916 919 11988577 10.1126/science.1068597
Rinn JL Euskirchen G Bertone P Martone R Luscombe NM Hartman S Harrison PM Nelson FK Miller P The transcriptional activity of human chromosome 22. Genes Dev 2003 17 529 540 12600945 10.1101/gad.1055203
Yamada K Lim J Dale JM Chen H Shinn P Palm CJ Southwick AM Wu HC Kim C Nguyen M Empirical analysis of transcriptional activity in the Arabidopsis genome. Science 2003 302 842 846 14593172 10.1126/science.1088305
Bertone P Stolc V Royce TE Rozowsky JS Urban AE Zhu X Tongprasit W Samanta M Weissman S Rinn JL Global identification of human transcribed sequences with genome tiling arrays. Science 2004 306 2242 2246 15539566 10.1126/science.1103388
Kampa D Cheng J Kapranov P Yamanaka M Brubaker S Cawley S Drenkow J Piccolboni A Bekiranov S Helt G Novel RNAs identified from an in-depth analysis of the transcriptome of human chromosomes 21 and 22. Genome Res 2004 14 331 342 14993201 10.1101/gr.2094104
Stolc V Gauhar Z Mason C Halasz G vanBatenburg MF Rifkin SA Hua S Herreman T Tongprasit W Barbano PE A gene expression map for the euchromatic genome of Drosophila melanogaster. Science 2004 306 655 660 15499012 10.1126/science.1101312
Mockler TC Ecker JR Applications of DNA tiling arrays for whole-genome analysis. Genomics 2005 85 1 15 15607417 10.1016/j.ygeno.2004.10.005
Singh-Gasson S Green RD Yue YJ Nelson C Blattner F Sussman MR Cerrina F Maskless fabrication of light-directed oligonucleotide microarrays using a digital micromirror array. Nat Biotechnol 1999 17 974 978 10504697 10.1038/13664
Nuwaysir EF Huang W Albert TJ Singh J Nuwaysir K Pitas A Richmond T Gorski T Berg JP Ballin J Gene expression analysis using oligonucleotide arrays produced by maskless photolithography. Genome Res 2002 12 1749 1755 12421762 10.1101/gr.362402
Zhao WM Wang J He X Huang X Jiao Y Dai M Wei S Fu J Chen Y Ren X BGI-RIS: an integrated information resource and comparative analysis workbench for rice genomics. Nucleic Acids Res 2004 32 D377 D382 14681438 10.1093/nar/gkh085
Wu P Ma L Hou X Wang M Wu Y Liu F Deng XW Phosphate starvation triggers distinct alterations of genome expression in Arabidopsis roots and leaves. Plant Physiol 2003 132 1260 1271 12857808 10.1104/pp.103.021022
Rabbani MA Maruyama K Abe H Khan MA Katsura K Ito Y Yoshiwara K Seki M Shinozaki K Yamaguchi-Shinozaki K Monitoring expression profiles of rice genes under cold, drought, and high-salinity stresses and abscisic acid application using cDNA microarray and RNA gel-blot analyses. Plant Physiol 2003 133 1755 1767 14645724 10.1104/pp.103.025742
Cheng Z Buell CR Wing RA Gu M Jiang J Toward a cytological characterization of the rice genome. Genome Res 2001 11 2133 2141 11731505 10.1101/gr.194601
Gale MD Devos KM Comparative genetics in the grasses. Proc Natl Acad Sci USA 1998 95 1971 1974 9482816 10.1073/pnas.95.5.1971
Shimamoto K Kyozuka J Rice as a model for comparative genomics of plants. Annu Rev Plant Biol 2002 53 399 419 12221982 10.1146/annurev.arplant.53.092401.134447
Rensink WA Buell CR Arabidopsis to rice. Applying knowledge from a weed to enhance our understanding of a crop species. Plant Physiol 2004 135 622 629 15208410 10.1104/pp.104.040170
Mathé C Sagot M-F Schiex T Rouzé P Current methods of gene prediction, their strengths and weaknesses. Nucleic Acids Res 2002 30 4103 4117 12364589 10.1093/nar/gkf543
Zhang MQ Computational prediction of eukaryotic protein-coding genes. Nat Rev Genet 2002 3 698 709 12209144 10.1038/nrg890
Feschotte C Jiang N Wessler SR Plant transposable elements: where genetics meets genomics. Nat Rev Genet 2002 3 329 341 11988759 10.1038/nrg793
Grandbastien MA Activation of plant retrotransposons under stress conditions. Trends Plant Sci 1998 3 181 187 10.1016/S1360-1385(98)01232-1
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 10.1016/j.molcel.2004.05.027
Bedell JA Budiman MA Nunberg A Citek RW Robbins D Jones J Flick E Rohlfing T Fries J Bradford K Sorghum genome sequencing by methylation filtration. PLoS Biol 2005 3 e13 15660154 10.1371/journal.pbio.0030013
Hennig W Heterochromatin. Chromosoma 1999 108 1 9 10199951 10.1007/s004120050346
Hoekenga OA Muszynski MG Cone KC Developmental patterns of chromatin structure and DNA methylation responsible for epigenetic expression of a maize regulatory gene. Genetics 2000 155 1889 1902 10924483
Stam M Belele C Dorweiler JE Chandler VL Differential chromatin structure within a tandem array 100 kb upstream of the maize b1 locus is associated with paramutation. Genes Dev 2002 16 1906 1918 12154122 10.1101/gad.1006702
Kim H Snesrud EC Haas B Cheung F Town CD Quackenbush J Gene expression analyses of Arabidopsis chromosome 2 using a genomic DNA amplicon microarray. Genome Res 2003 13 327 340 12618363 10.1101/gr.552003
Mittelsten Scheid O Afsar K Paszkowski J Formation of stable epialleles and their paramutation-like interaction in tetraploid Arabidopsis thaliana. Nat Genet 2003 34 450 454 12847525 10.1038/ng1210
Preuss D Chromatin silencing and Arabidopsis development: a role for polycomb protein. Plant Cell 1999 11 765 768 10330463 10.1105/tpc.11.5.765
Meyer P Transcriptional transgene silencing and chromatin components. Plant Mol Biol 2000 43 221 234 10999406 10.1023/A:1006483428789
Ahmad K Henikof S Modulation of a transcription factor counteracts heterochromatic gene silencing in Drosophila. Cell 2001 104 839 847 11290322 10.1016/S0092-8674(01)00281-1
Pandey R Muller A Napoli CA Selinger DA Pikaard CS Richards EJ Bender J Mount DW Jorgensen RA Analysis of histone acetyltransferase and histone deacetylase families of Arabidopsis thaliana suggests functional diversification of chromatin modification among multicellular eukaryotes. Nucleic Acids Res 2002 30 5036 5055 12466527 10.1093/nar/gkf660
Reyes JC Hennig L Gruissem W Chromatin-remodeling and memory factors. New regulators of plant development. Plant Physiol 2002 130 1090 1101 12427976 10.1104/pp.006791
Soppe WJ Jasencakova Z Houben A Kakutani T Meister A Huang MS Jacobsen SE Schubert I Fransz PF DNA methylation controls histone H3 lysine 9 methylation and heterochromatin assembly in Arabidopsis. EMBO J 2002 21 6549 6559 12456661 10.1093/emboj/cdf657
Lippman Z Gendrel AV Black M Vaughn MW Dedhia N McCombie WR Lavine K Mittal V May B Kasschau KD Role of transposable elements in heterochromatin and epigenetic control. Nature 2004 430 471 476 15269773 10.1038/nature02651
Lippman Z Martienssen R The role of RNA interference in heterochromatic silencing. Nature 2004 431 364 370 15372044 10.1038/nature02875
Dubcovsky J Ramakrishna W SanMiguel PJ Busso CS Yan L Shiloff BA Bennetzen JL Comparative sequence analysis of colinear barley and rice bacterial artificial chromosomes. Plant Physiol 2001 125 1342 1353 11244114 10.1104/pp.125.3.1342
Song R Llaca V Messing J Mosaic organization of orthologous sequences in grass genomes. Genome Res 2002 12 1549 1555 12368247 10.1101/gr.268302
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 10.1016/S1369-5266(03)00015-3
Wessler SR Plant transposable elements. A hard act to follow. Plant Physiol 2001 125 149 151 11154320 10.1104/pp.125.1.149
Baba A Hasezawa S Syono K Cultivation of rice protoplasts and their transformation mediated by Agrobacterium spheroplasts. Plant Cell Physiol 1986 27 463 471
Ma L Li J Qu L Hager J Chen Z Zhao H Deng XW Light control of Arabidopsis development entails coordinated regulation of genome expression and cellular pathways. Plant Cell 2001 13 2589 2607 11752374 10.1105/tpc.13.12.2589
Kent WJ BLAT-the BLAST-like alignment tool. Genome Res 2002 12 656 664 11932250 10.1101/gr.229202. Article published online before March 2002
Liu YG Mitsukawa N Oosumi T Whittier RF Efficient isolation and mapping of Arabidopsis thaliana T-DNA insert junctions by thermal asymmetric interlaced PCR. Plant J 1995 8 457 463 7550382 10.1046/j.1365-313X.1995.08030457.x
Liu YG Whittier RF Thermal asymmetric interlaced PCR: automatable amplification and sequencing of insert end fragments from P1 and YAC clones for chromosome walking. Genomics 1995 25 674 681 7759102 10.1016/0888-7543(95)80010-J
Li L Zhao Y McCaig BC Wingerd BA Wang J Whalon ME Pichersky E Howe GA The tomato homolog of CORONATINE-INSENSITIVE1 is required for the maternal control of seed maturation, jasmonate-signaled defense responses, and glandular trichome development. Plant Cell 2004 16 126 143 14688297 10.1105/tpc.017954
| 15960804 | PMC1175972 | CC BY | 2021-01-04 16:05:39 | no | Genome Biol. 2005 May 27; 6(6):R52 | utf-8 | Genome Biol | 2,005 | 10.1186/gb-2005-6-6-r52 | oa_comm |
==== Front
Genome BiolGenome Biology1465-69061465-6914BioMed Central London gb-2005-6-6-r531596080510.1186/gb-2005-6-6-r53MethodIdentification of co-regulated transcripts affecting male body size in Drosophila Coffman Cynthia J [email protected] Marta L [email protected] Sergey V [email protected] Laura A [email protected] Lauren M [email protected] Health Services Research and Development Biostatistics Unit, Durham VA Medical Center (152), Durham, NC 27705, USA2 Duke University Medical Center, Department of Biostatistics and Bioinformatics, Durham, NC 27710, USA3 Department of Zoology, University of Florida, Gainesville, FL 32611, USA4 Department Ecology and Evolution, University of California at Davis, Davis, CA 95616, USA5 Department of Agronomy, Purdue University, West Lafayette, IN 47907, USA2005 1 6 2005 6 6 R53 R53 20 1 2005 21 2 2005 9 5 2005 Copyright © 2005 Coffman 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.
Factor analysis is applied to microarray data in order to relate gene networks to complex traits and identifies a factor associated with body size in Drosophila simulans.
Factor analysis is an analytic approach that describes the covariation among a set of genes through the estimation of 'factors', which may be, for example, transcription factors, microRNAs (miRNAs), and so on, by which the genes are co-regulated. Factor analysis gives a direct mechanism by which to relate gene networks to complex traits. Using simulated data, we found that factor analysis clearly identifies the number and structure of factors and outperforms hierarchical cluster analysis. Noise genes, genes that are not correlated with any factor, can be distinguished even when factor structure is complex. Applied to body size in Drosophila simulans, an evolutionarily important complex trait, a factor was directly associated with body size.
==== Body
Background
Unraveling complex traits requires an understanding of how genetic variation results in variation among transcript levels, proteins, and metabolites, and how this variation generates phenotypic variation. These distinct levels in the biological system are interdependent. The ability to model interactions among loci at each of these levels, and relationships between levels, is key to providing insight into complex traits. The promise of genomic and proteomic technology is in capturing variation for thousands of loci simultaneously. This affords an unprecedented opportunity to understand the consequences of genetic variation. Many studies have exploited this ability through the use of mutant analysis applied to whole-genome transcript arrays. Mutant analysis provides insight into the impact of a mutation on a gene network and whole-genome studies of transcription have revealed misexpression due to gene knockouts and have established redundancy and specificity of transcriptional regulation [1]. Cluster analysis has been successfully combined with tests of differential expression to study whole-genome response to mutation in order to develop hypotheses about co-regulation and coordinated expression [2,3].
However, the consequences of such strong perturbations are difficult to apply to pathways in non-mutant individuals. In addition, the mutations chosen usually cause a severe alteration in a single gene, such as a knockout. Natural variants introduce smaller changes in pathways [4] and natural variants may exhibit allelic differences at several loci. Natural variation in the transcriptome as a consequence of genetic variation has been demonstrated [5,6]. Natural genotypes can also be mated in a deliberate manner and the progeny of such matings can be used to estimate the genetic architecture of individual traits [7,8], and to link traits across different levels of the biological system [9,10]. We focus here on providing insight into how coordinated gene expression affects phenotype. Links between transcript abundance and phenotypic variation have been established [11-15]. What is needed now is an analytic approach that allows interpretation of the relationships among transcript levels and modeling of the link between transcript level and complex trait.
Factor analysis is an analytic approach that describes the covariation among a set of genes through the estimation of factors. One may interpret the factor as the mechanism, for example transcription factors, microRNAs (miRNAs), and so on, by which genes are co-regulated. The resulting factor model represents sets of coordinately expressed genes. Genes may participate in multiple factors. Principal components analysis, spectral map analysis and correspondence analysis are alternative multivariate techniques for microarray analysis [16,17] that can all be used in this capacity. Factor analysis, however, provides a convenient representation of the gene network by describing each gene's association with the factor as a load (between -1 and 1), where the strength of the load indicates how much influence the transcript level of that gene has upon the factor. The factor can then be examined for associations with complex traits [18]. Factor analysis is the extension of Sewell Wright's work on the correspondence among traits [19], and as such is perfectly suited for modeling the relationships among transcript levels for a set of crosses. The high dimensionality of genome-wide expression data presents special challenges. This challenge, primarily the ill-conditioned matrices resulting from such studies, has been well described and explicitly acknowledged in much of the literature on the analysis of gene-expression data [20-22]. If thousands of genes belonging to dozens of networks are simultaneously considered as current theory indicates, spurious associations may emerge and/or true associations may be obscured [23,24]. Previous applications of factor analyses to array data [25,26] dealt with this issue by an initial reduction of dimensionality through the use of cluster analysis.
Using simulation studies, we evaluate the utility of factor analysis for identifying covariation in gene-expression data and identifying underlying factors. We compare the performance of factor analysis to hierarchical clustering and tight clustering [27]. We then test the estimation of factors on a set of Drosophila lines for genes involved in the immune pathway. The immune system provides a relatively well understood set of interactions and as such allows a real data check on the applicability of factor analysis to microarray data.
A logical next step is to use factor analysis to relate variation among transcript levels to phenotypic variation, a step not possible in a cluster analysis. For Drosophila, body size is a complex trait where latitudinal clines in body size have been repeatedly demonstrated across ectotherms [28]. In D. subobscura, a body-size cline evolved in 12 decades, thus ranking body size in flies among the fastest-evolving morphological traits ever observed in natural populations [29]. The proximate reasons for these clines are complex, especially given that body size in flies is positively correlated with mating success in males [30-32]. Of further interest are data suggesting that the same genomic regions are involved in adaptation in two of these clines, South America and Australia [33]. In contrast to the immune system, there is little a priori information on how the candidates genes are related to one another. In addition, identification of factors associated with variation in body size in natural populations of Drosophila is a question of great evolutionary interest.
Results
Simulations
In the initial scenario, a sample size of 100 individuals was examined. This sample size is large for a microarray experiment, but is in the low range of the minimum sample size suggested in factor analysis methodology [34]. We simulated a high degree of correlation among genes within a factor (ρ = 0.80), three factors with a manageable number of genes associated within each factor (correlated genes: 30), and some genes not associated with any factor (noise genes: 100). We assume genetically variable lines for which differences in transcript abundance among lines was moderate within each of the three factors [35]. Factor analysis on these data was performed. Factors were identified by examining the eigenvalues of the correlation matrix [23]. The first five eigenvalues were 25.3, 21.3, 18.3, 4.3, and 4.1. The substantial drop between the third and fourth eigenvalues (from 18.3 to 4.3) indicates that three factors (the number simulated) are clearly identified, explaining 34% of the variation. We then set the number of factors in the analysis to three, and estimated factor loadings in order to examine the structure of the factor. All (100%, n = 90) of the correlated genes loaded [36] on the correct factor, with none of the noise genes loading on any factor (see Table 1). Reducing the correlation among factors, and reducing the effect size do not affect the ability of factor analysis to identify the correct underlying structure (Table 1).
Results of a hierarchical cluster analysis found that the three groups of genes clustered together with the noise genes which formed two distinct clusters. However, discriminating the true clusters from the noise clusters was not obvious using standard approaches. Tight clustering [27], where a resampling strategy is used to separate noise genes from signal, on these data was interesting. If the number of clusters is set to the true value of three, all 190 genes are identified as noise. If the number of clusters is set to five, 45 of the 100 noise genes are correctly identified as noise. All of the correlated genes are placed into the correct clusters. The remaining 55 noise genes are placed into clusters.
For the case with lower effect size and lower correlation, the dendrogram resulting from hierarchical cluster analysis is given in Figure 1. As in the factor analysis, the three groups of genes clustered together well, although not perfectly. Once again, however, statistics for determining the appropriate number of clusters did not clearly identify the correct number of clusters. The noise genes also seem to follow discernible clustering patterns. In tight clustering, when the correct number of clusters are specified and the number of extra clusters (k0) is set to 6-7, 23 of the 90 correlated genes are identified as noise and all of the noise genes are correctly identified. Setting the number of clusters higher results in clusters of noise genes. In this simple case, factor analysis clearly outperforms both traditional hierarchical cluster analysis and tight clustering, as it is easily able to discern the correct number of underlying clusters.
We then increased the number of genes from a total of 190 (90 in the three networks and 100 noise genes) to 1,900 (900 in the three factors and 1,000 noise genes). Using factor analysis, we easily identified the correct number of factors and 100% of the genes in each factor loaded on the correct factor (see Table 1). Lowering the correlation among genes in a factor to ρ = 0.4 resulted in the reduction of the explanatory power of the factor analysis. The number of underlying factors was correctly identified although, as expected, the total variation explained by the factors was reduced. Of the correlated genes, 66% loaded on the correct factors and only one noise gene (out of 1,000) was mistakenly placed into a factor. Given the reasonable fractions identified when the number of genes in factors differs by an order of 10 (190 versus 1,900), and the fact that our recovery of the structure was virtually unchanged, it is apparent that the number of genes in a factor does not impact on the ability of factor analysis to recover the factor structure. In contrast, hierarchical cluster analysis performs less well as the number of noise genes increases, with the noise genes increasing in their dispersion among clusters.
In a set of simulations to match our Drosophila experimental design, 10 genotypes with three replicates per genotype for a total of 30 samples (chips) were simulated. Averaging transcript abundance within each genotype removed uninteresting variation and increased resolution (data not shown). We began with three factors of 30 genes each, and 100 noise genes. The maximum difference in transcript abundance among genotypes was large (1, 2 and 3). In this case, the three factors were not clearly identified. To determine whether genes would be correctly placed inside factors, the number of factors was set to three. When difference in transcript abundance among genotypes was lowest, fewer genes were placed in the correct factor (13/30). On the other hand, where difference in transcript abundance among genotypes was highest and 100% of the correlated genes loaded on the correct factor, 29 of the genes from the other two factors and 29 noise genes were erroneously identified as a part of this factor. When correlation among genes within a factor was reduced to ρ = 0.40, the difficulty in identifying genes on the correct factor increased.
We hypothesized that the presence of many noise genes, coincident with a sample size of ten, was responsible for the difficulty in identifying the correct number of gene factors. Without the noise genes, when correlation was ρ = 0.8, each of the underlying factors was clearly identified. All (100%) of the genes were correctly placed in the corresponding factors, although several genes were placed in multiple factors. When correlation among genes in a factor was reduced to ρ = 0.40, the number of factors was underestimated and many of the genes were identified in multiple factors. This indicates a complex interplay between the difference in transcript abundance among genotypes, the sample size, the number of noise genes and the correlation structure in the identification of factors.
We also examined the impact of increasing the number of factors beyond the number of genotypes. We simulated a set of 600 genes belonging to 20 factors, with 30 genes in each factor and 100 noise genes. In an analysis with nine factors (the maximum estimable with ten genotypes), factor 1 seemed to capture the majority of all the genes in the 20 factors. When effect size was large, the majority of the noise genes (58%) were identified by their failure to load highly (greater than 0.7) on any factor, and the majority of genes that were correlated did load highly on at least one factor (96%). Lowering the correlation lowers the ability to identify correlated genes as loading highly. In this case, 30% are identified and approximately the same fraction of noise genes (60%) are identified. Consistent with previous factor analysis theory [23,36], when the number of factors is larger than the sample size the number and composition of the factors cannot be estimated.
We then wanted to determine whether hierarchical cluster analysis would resolve this structure more clearly. We plotted the results in a dendrogram where each of the simulated factors are plotted with a separate color (Figure 2). No clear pattern of clustering was found. However, the clusters do form some 'kernels', so that biological knowledge of pathways could potentially be applied to interpret some of the groupings, as is common practice. However, the clear presence of noise genes throughout the cluster structure clearly makes interpretation difficult in cases where the true structure is unknown.
We then applied tight clustering to these data. When the correct number of clusters is specified, tight clustering does identify the structure of the 20 clusters. Each cluster consists of a subset of genes that belong to that cluster. Notably, it does not erroneously place noise genes into clusters. It performs less well in identifying the correct number of correlated genes failing to place 50% into clusters, instead classifying these correlated genes as noise. When a larger number of clusters (25) is specified, then there are some 'extra' clusters of noise genes and the clusters of genes are themselves not as distinct, that is 40 genes of the 600 are incorrectly grouped and 221 or 37% of the genes which should be in a cluster are designated as noise.
Overall, the simulation results indicate that focusing on a manageable number of possible factors with a measurable amount of difference in transcript abundance among lines can result in successful identification of factor structure, even when the number of genes examined is large relative to the sample size. We also find that the factor loadings can distinguish noise genes even in complex cases, although the factor structure can not be resolved clearly in those cases.
Data analysis
Loci showing evidence for transcript variation
The GeneChip Drosophila Genome Array was used for this study and of the approximately 13,500 genes on the array, 7,886 showed expression on at least one array, and of these, 4,667 showed evidence for variation among genotypes. As we are studying covariation, we restricted our examination to this list of 4,667 loci (see Additional data file 1 [Supplementary Table 1]).
The immune pathway
To provide an assessment of the performance of factor analysis on a set of well characterized genes, FlyBase [37] was queried for candidate genes involved in the immune pathway [38]. We compared the candidate genes to the list of 4,667 transcripts, and 54 genes were identified. Factor analysis on these transcript levels resulted in the identification of three factors (see Figure 3). Notably, the first factor contained all the lysogen genes present in the study, and the second factor contained all cecropins (Table 2). Factor-analysis groups co-regulated genes in a manner consistent with our understanding of the immune pathway. Hierarchical cluster analysis was also performed on this set of genes. As with the factor analysis, hierarchical cluster analysis found that the lysogen genes clustered together, as did the cecropins. However, determining the appropriate number of clusters was problematic and did not lead to any clear interpretation of the appropriate number of clusters. The final factor model for the immune genes included three factors, therefore we examined a k-means clustering analysis with three clusters for comparison. We found that 17/24 (71%) of the genes that loaded high on factor 1 were on the same cluster. This cluster also includes a few genes that loaded on factors 2 and 3. Genes that loaded high on the second and third factor were distributed among the remaining two clusters. The genes in these groups that did not load significantly on any factor were distributed among the three clusters.
Candidate loci for body size
We again queried FlyBase for a list of genes involved in body size determination and found 92 body size candidates in our list of 4,667 transcripts. Four factors were identified (Figure 3). The identification of covarying genes in the same factor was intriguing. Of particular note is the presence of loci that are contained within quantitative trait loci (QTL) for body size on factor 1: Cdk4, trx, akt1, fru, Dr, mask, khc, and InR [33].
In our analysis, we regressed transcript level for individual candidate genes on the body size phenotype. We found 2,892 genes significant at a nominal level of 0.05 (false discovery rate (FDR), 16%, see Table 3). At a more stringent nominal threshold of 0.01, there were 14 candidate genes for body size which showed significant association between transcript abundance and phenotypic variation for male body size (FDR of 7%). Of the QTL candidates, only InR showed significant association with the phenotype.
We then tested the hypothesis that the estimated factors were directly related to phenotypic variation in body size. In our analysis, we regressed the estimated factor (latent variable) on the phenotype body size for each genotype. The regression of factor 1 on body size showed evidence of an association between the factor and the phenotype of body size (P = 0.04, Figure 4a).
Targets of miRNAs
Of 535 putative miRNA targets [39], 203 were contained in our set of 4,667 gene transcripts. Factor analysis resulted in the identification of four gene factors (Table 4). The second factor contained four of the same genes as factor 1 for body size (puc, Eh, mys, bon) with 76 additional genes contained in this factor (loaded at 0.40 or greater). However, this factor was not associated with body size (P = 0.55). While some of the QTL candidates are also putative targets of miRNA regulation, (Cdk4, trx, Dr) these genes did not participate in this factor but were common to the factor identified by the third factor (see Table 4), for which 69 additional genes loaded. This third factor was negatively correlated with body size (P = 0.04, see Figure 4b). (Cdk4, trx, Dr) were not associated with body size in regressions between these individual genes and body size.
Discussion
We applied factor analysis to high-dimensional microarray data. Using simulated data to estimate factors, we found that when correlation among genes is strong, the number of factors and their structure can be estimated, even in the case where genes unrelated to the factor structure (noise genes) are included. We also found that when noise genes were included hierarchical cluster analysis was unable to separate the noise genes from the signal, or to correctly identify the number of clusters. When the number of genes is large relative to the sample size, as is common in array studies, the number of factors and the genes belonging to each factor can still be identified, as long as the number of factors is less than the sample size. In contrast, hierarchical cluster analysis did not identify the number of clusters even when the number of clusters was smaller than the sample size.
We found that if the number of factors is larger than the number of genotypes, the majority (58%) of noise genes still do not load on any factor, while all but 26 of the 600 correlated genes do load on at least one factor. However, the correct association between individual genes and factors is lost. We conclude that while factor analysis is effective at separating the signal from the noise, the structure of the signal is not estimable. This is consistent with reports in the literature [23,36]. Using hierarchical clustering, the number and the structure of clusters is not recovered and noise genes are scattered throughout the cluster structure. Kernels of tightly correlated genes were visible, however, indicating that kernel identification is possible in cases where biological knowledge is present. Any separation of signal from noise is purely serendipitous.
These results are not unexpected as the mathematical properties of gene-expression data or the high dimensionality of the data lead to problems for any analysis. When the number of columns (in this case observations) is less than the number of rows (or variables), the matrix is considered ill conditioned [40]. Ill-conditioned matrices can cause problems for many types of statistical analysis and can lead to overfitting of predictive models, among other problems [41,42]. Some of the effects of the ill-conditioned matrices can be mitigated; however, the problem of an overdetermined system will always exist. An example is in multiple regression, where models with more variables than we have data can not be fit [43].
Tight clustering [27] represents a significant advance over hierarchical clustering in the estimation of cluster structure for microarray data. It provides a reasonable way of identifying most noise genes. In simple cases, however, the algorithm needs to have some flexibility when specifying starting values; that is, more rather than fewer clusters improve the chances of correctly identifying true clusters. The algorithm requires that the number of clusters be specified a priori. In contrast, in many simple cases the number and structure of factors can be recovered precisely using a factor analysis. In the most complex case examined (20 factors, 600 genes and 100 noise genes), when the number of clusters is correctly specified, tight clustering identifies 100% of the noise genes and the 20 clusters with correct genes within each cluster. However, it also incorrectly identifies 50% of genes with signal as noise genes. If too many clusters are specified (25), then the number of genes identified in clusters increases to 63%, although the correct structure is no longer maintained and 20% of the noise genes are incorrectly clustered. In contrast, factor analysis separates the signal from the noise, correctly identifying 96% of the noise genes as noise and 58% of the genes as having signal. In this complex case it is difficult to say whether factor analysis or tight clustering is 'better'. Both of these approaches clearly outperform hierarchical clustering, and they are complementary in their approach. If the primary goal is to separate the correlated genes from the uncorrelated genes then factor analysis performs better than tight clustering. If determining the number and structure of factors (coexpressed genes) is the goal, using both approaches and comparing findings will be reasonable.
Clear network structures can be identified with factor analysis and can then be followed up experimentally. Furthermore, unlike cluster analysis (hierarchical or tight), which provides no summary of the clusters into a single variable, the factor loading values are directly interpretable as the degree of participation of that locus in a factor. In contrast to cluster analysis, the factor loadings for genes give us information on the relative strength of a gene on a factor. We can use factor loadings to identify the most 'significant' genes on a factor and we can use the factor loadings to remove genes that are not contributing to any factor. As such, genes that all load highly on the same factor can be said to be coordinately expressed and putatively co-regulated, and noise genes can be identified as they will not load high on any factor. Unlike clustering, factor analysis allows genes to participate in several factors, thus reflecting biological reality more accurately.
Extensions of factor analysis have been developed and include allowing for nonlinearity [44,45] and the application of Bayesian approaches [46]. Factor analysis is itself a subclass of modeling techniques known as structural equation models [47] and the ongoing theoretical interest in these approaches allows for expansion of consideration to more than the present circumstance.
We focus our Drosophila analyses on a set of genotypes that are mated according to a round-robin design where the parental lines are natural variants. Analyses of such lines allows inferences about the underlying genetic contribution to variation and inferences about variation in pathways that occurs as a result of such natural genetic variation. By using candidate loci from reverse genetic and mutational projects, the importance of these loci in a broad context can be assessed. The factor analysis for the list of genes annotated as body size candidates resulted in the estimation of four factors. Factor 1 is of great interest as several of the genes that load on this factor - Cdk4, trx, akt1, fru, Dr, mask, woc, Khc, and InR - are contained in QTL for body size [33]. This overlap is exciting as the data from this QTL analyses are independent of our factor analysis. The identification of the same set of loci lends weight to the evidence that these loci are involved with the formation of body size in a natural population. Of these loci, only InR is directly correlated with body size. Given our limited knowledge of pathways for body size, it was exciting to note that two of the genes in this factor - trx and M(2)21AB - have been shown to interact [48]. In addition, the early sex-determination cascade genes Sxl, tra, and tra2 are all associated with a different factor (factor 2).
If the expression of multiple genes is regulated by a common underlying factor such as a transcription factor or a miRNA, and this regulation exhibits genetic variation, then we expect that the gene expression among these genes in a set of genotypes from a mating design will also be correlated. Factor analysis on putative targets of miRNA control revealed that putative targets of the same miRNA often occurred on the same factor. The number of targets for each miRNA was small, and so we cannot conclude that the coexpression is greater than we would expect by chance. Nevertheless, this is the first opportunity to examine the correlation structure of these putative targets of miRNAs. We found that of the 203 genes identified as miRNA targets, there was evidence for 188 participating in at least one of the four factors. While four genes on factor 1 for body size (puc, Eh, mys, and bon), and nine additional candidate body size genes (Sh, Abd-B, trx, fng, qm, woc, Dr, and Cdk4) are putative targets of miRNAs [39], the resulting miRNA factors are uncorrelated with the factors for body size. One of the miRNA factors is associated with the body size phenotype.
In summary, factor analysis, a technique developed to discover and model underlying mechanisms in complex social and psychiatric situations, seems to offer a reasonable middle ground for gaining understanding of coordinated gene expression. Our simulations show that when the number of underlying factors is larger than the sample size, it is not straightforward to recover the structure of the simulated data, although signal can be separated from noise. In contrast, target groups, even when the number of genes is large, can be used to identify several underlying regulatory mechanisms. In the case of body size for Drosophila, factor analysis offers an exciting opportunity to estimate gene networks, as relatively little is known about how genes involved in body size work together.
Materials and methods
Simulations
In trying to estimate the structure of gene networks, one of the first questions to be addressed is whether the number of genes contributing to each network affects the ability of the analysis to determine the structure of the network. Accordingly, we varied the number of genes examined in our simulations to explore this process (see Table 1). Gene networks were simulated as a set of correlated expression values. These networks represent genes affected by some common underlying biological process and the correlation structure that results from this biological connectedness is the concrete evidence of the network. In our simulations, for each network, a mean from a gamma distribution (γ) is drawn. A gamma distribution was chosen for the mean values across genes as this distribution has been shown to fit the distribution of transcript levels seen from genes on an array [49]. Genes in the network were simulated according to a multivariate normal distribution with correlation among genes indicated by the parameter (ρ). We looked at two levels of correlation ρ = 0.40 (weak) and ρ = 0.80 (strong). Genetic variation for individual genes within a network was simulated as a linear combination of the multivariate normal mean, a fixed genotypic effect and random noise. Genotypic effects for different lines differed in magnitude. The standard deviation was set to 1, and the standard normal was used, so that differences between lines can be directly interpreted as the size of the effect, that is the effect sizes are absolute and not relative to the magnitude of the gene expression (or amount of variation) for a particular gene. When the maximum difference among genotypes is less than 0.2 the effects are small, when differences are between 0.2 and 0.6 the effects are medium and when effects are larger than 0.8 they are relatively large [35]. Between networks the maximum difference of transcript abundance (effect size) among lines was allowed to vary.
Genes not participating in any network, were also simulated. These genes, uncorrelated to each other, are hereafter referred to as noise genes. For noise genes, the mean value for gene expression was drawn from a gamma distribution and variation from that mean was random and without regard to the genotype.
Concurrently, factor analysis on the data was performed and oblique rotations were used for estimating factor loadings [36]. We used the eigenvalues to determine the number of factors according to standard factor analytic approaches [36]. The plot of the eigenvalues, against the ordinal number for the eigenvalue (SCREE plot) is examined for the last substantial drop in the magnitude of the eigenvalues and a model with the same number of factors as the number of eigenvalues before the last substantial drop are retained [23,36]. Hierarchical cluster analysis was performed using a Ward distance. Tight clustering, a resampling-based approach to cluster analysis, was performed using software provided by [27].
Drosophila lines
Isogenic lines of Drosophila simulans were made from flies caught in Wolfskill orchard [50] and crossed in a round-robin mating design as follows: the ten parental stocks (randomly sampled and independently derived from a large natural reference population) were crossed together in ten combinations to create heterozygous lines such that each parent was present twice, once as a dam and once as a sire. Crosses were (dam × sire): SIM6 × SIM11, SIM11 × SIM36, SIM36 × SIM 37, SIM37 × SIM43, SIM43 × SIM61, SIM61 × SIM77, SIM77 × SIM85, SIM85 × SIM99, SIM99 × SIM105, and SIM105 × SIM6. The resulting heterozygous genotypes were used for the experiment as follows: RNA from males was extracted and labeled 4-7 days post-eclosion [6,50]. Transcript level was estimated the average difference using MAS 5 [51]. Genes without positive signal on at least one array were removed from further consideration. The remaining genes were normalized to the chip median and log transformed. Genes lacking variation among genotypes cannot be meaningfully assessed for covariation. Accordingly, any gene that showed evidence (P ≤ 0.2) for variation in transcript level across lines [50] or evidence for additive genetic effect (P ≤ 0.05) [6] was considered in further analyses. Adult male flies were measured for body size [52].
Drosophila data analysis
Candidate gene lists were developed using FlyBase queries. The miRNA targets were taken from Enright and colleagues [39]. The resulting lists were matched against the set of loci for which we had evidence of genetic variation in transcript abundance among lines.
In the first step of the data analysis, regression of transcript abundance for individual genes that were candidates for body size, or putative miRNA targets, was conducted to test the hypothesis that individual genes transcript levels were associated with body size. The average transcript abundance for each gene i within genotype j (gij) was regressed on the average male body size (Yj) for each genotype j as follows:
Yj = μ + gij + εij. (1)
The mean of all genotypes is μ and the random error is ε. Cluster analysis was performed on the set of immune genes. This was done using a hierarchical cluster analysis on the standardized values of gene expression. The results were plotted in a dendrogram.
Factor analysis on standardized values of gene expression, for each of the three lists (genes in the immune pathway, candidates for body size and putative miRNA targets) was conducted separately and factor loadings were estimated using an oblique rotation. The resulting set of eigenvalues was plotted in a SCREE plot, and the number of factors chosen such that the drop between eigenvalues was apparent, and a reasonable proportion of the variation was explained [36]. Each factor identified represents a gene network. Once the number of factors was identified, factor analysis was repeated for that fixed number of factors and loading values (the correlation between individual genes and the estimated factor structure) were estimated. All factor analyses were conducted in SAS (PROC FACTOR) using standard options, no special coding was required.
The effect of the factor upon the genotype is also estimable in the form of a factor value. Factor values were computed as a linear combination of the factor loading times the standardized value for the gene expression [18,23]. The factor value for genotype j (GNj) is an estimate of the impact of the network upon the genotype and is estimated as
where gij is the expression value for gene i within genotype j and li is the loading value estimated from the factor analysis for gene i across all genotypes. Differences in factor values represent differences between genotypes for the factor.
As all the genotypes are related in a mating design, we expect that differences in networks due to genetic variation which result in phenotypic variation should be detectable. Accordingly, the factor values were then regressed on the phenotype where
Yj = μ + GNj + εj (2)
where GNj is the factor value for genotype j.
Additional data files
The following additional data is available with the online version of this paper. Additional data file 1 lists the median values of gene expression for each of the ten lines (after normalization) for the 4,667 genes found to have some evidence of a line effect. Some basic annotation information from Affymetrix is also provided along with the feature number.
Supplementary Material
Additional File 1
Line 1 through Line 10 medians for the 4667 genes considered in this analysis, after normalization (line1median) and standardization (line1median_st). Some basic annotation is also included as well as the Affymetrix feature number , Fall 2004). The median values of gene expression for each of the ten lines (after normalization) are provided for the 4,667 genes found to have some evidence of a line effect. These are labeled as line1_median-line10_median. Standardized values are also provided and labeled as line1_median _st-line1-_median_st. Some basic annotation information from Affymetrix is also provided along with the feature number. The chromosome number was inferred based upon the cytogenetic map position.
Click here for file
Acknowledgements
This work is supported by US Department of Agriculture-IFAFS N0014-94-1-0318 (L.M.M., C.J.C.), NIH GLUE Grant R24-GM65513 (S.V.N., M.L.W., L.M.M.), NIH-NIAID 5R01AI059111-02 (L.M.M.), the Purdue Agricultural Research Station, the University of Florida Microarray Core Facility, and the UC Davis Microarray Core Facility. We thank Hayden Bosworth (Duke University) for introducing us to factor analysis and Chien-Cheng (George) Tseng for helping us to implement tight clustering.
Figures and Tables
Figure 1 Hierarchical cluster plot of simulation with two genotypes, 100 noise genes, and three factors. ρ = 0.40, effect size = 0.2, 0.4, and 0.6. Blue, noise genes; green, genes from underlying factor 1; red, genes from underlying factor 2; black, genes from underlying factor 3.
Figure 2 Hierarchical cluster plot of simulation with ten genotypes, 100 noise genes, and 20 factors. ρ = 0.4, effect size = 1, 2,...20. Blue, noise genes; other colors represent genes that should cluster together.
Figure 3 SCREE plots. The x-axis is the ordinal number of the eigenvalue and the y-axis is the magnitude of the eigenvalue. The number to the right of the plotted point indicates the cumulative variance explained as each factor is added. The dotted line indicates the cutoff point in the SCREE plot where there is a sharp drop off in the magnitude of the eigenvalues. The number of factors above the dotted line are the number retained for the factor analysis. (a) Body size, 92 genes; four factors are selected. (b) Immune, 53 genes; three factors are selected [36].
Figure 4 Regression plots. (a) A plot of factor 1 from candidate genes for body size on the x-axis and measured male body size on the y-axis. The solid line is the regression of factor 1 on measured male body size with an estimated slope of -0.021y with a standard deviation of 0.008 and is significantly different from zero (P = 0.04). Line crosses: open square 1136; open circle (top left), 611; open triangle 3743; plus sign, 4361; multiplication sign, 6177; open diamond, g785; inverted open triangle 8599; star 99105; solid circle, 1056; open circle (middle), 3637. (b) A plot of factor 3 from candidate genes for miRNA on the x-axis and measured male body size on the y-axis. The solid line is the regression of factor 3 on measured male body size with an estimated slope of -0.020y with a standard deviation of 0.008 and is significantly different from zero (P = 0.04). Symbols as in (a).
Table 1 Gene expression simulations
Number of genotypes Number of factors Number of genes Correlation (ρ) Effect size Factors clearly identified Proportion correct
Noise Each factor
2 3 100 30 0.8 0.2,0.4,0.6 Y 1.00
0.02,0.04,0.06 Y 1.00
0.4 0.2,0.4,0.6 Y 0.84
0.02,0.04,0.06 Y 0.66
2 3 1000 300 0.8 0.2,0.4,0.6 Y 1.00
0.4 0.2,0.4,0.6 Y 0.66
10 3 100 30 0.8 1,2,3 N 0.81
0.4 1,2,3 N 0.64
0 30 0.8 1,2,3 Y 1.00
0.4 1,2,3 N 0.63
10 20 100 30 0.4 1,2,...,20 N -
0.1,0.2,...,2 N -
The number of genotypes simulated is given in the first column. The number of underlying latent factors is given in the second column, followed by the number of genes simulated that are not a part of any underlying factor. The number of genes on each factor is given next, and are simulated as a multivariate normal with pairwise correlation among genes within the factor of ρ. The mean for the first genotype is drawn from a gamma distribution, and the subsequent means were drawn from a multivariate normal, with standard deviation of one such that the maximum difference between the means can be interpreted as the genotypic effect size. Thus, for each underlying factor the simulated genotypic effect is the maximum difference in transcript abundance among genotypes for the first, second, and third factor, respectively. Factors are considered to be clearly identified if there is a substantial drop in the eigenvalues of the correlation matrix, and a reasonable proportion of the total variation is explained. The proportion correct is the proportion of genes correctly identified when setting the number of factors in the factor analysis to be the simulated number of latent factors. For the simulation with 20 latent factors we cannot compute the proportion correctly identified, as there are more simulated factors than possible factors.
Table 2 Factor analysis for candidate genes for immune function
Factor 1 Factor 2 Factor3
Name Load Name Load Name Load
1 dl 0.93 scrib 0.95 AttB 0.86
2 cact 0.86 CecA1 0.88 GNBP2 0.86
3 LysC 0.86 CecA2 0.85 CG16756 0.85
4 LysD 0.84 IM2 -0.84 CG8193 0.83
5 LysB 0.84 PGRP-SA -0.78 Bc 0.74
6 LysE 0.83 CG5140 -0.76 Eip93F -0.73
7 GNBP3 -0.81 IM1 -0.69 ref(2)P 0.72
8 tub 0.77 IM4 -0.69 CG2736 0.69
9 Tl 0.74 CG6214 0.66 CG3829 0.65
10 CG12780 0.74 CecC 0.63 PGRP-SC2 0.60
11 Mpk2 0.73 PGRP-SC1b -0.59 IM1 -0.59
12 PGRP-LE 0.72 CG1643 -0.53 TepIV 0.53
13 Lectin-galC1 -0.72 PGRP-SD 0.52 CG6214 -0.46
14 LysS 0.70 TepIV -0.51 Drs -0.44
15 CG17338 0.67 cact 0.45 GNBP2 0.43
16 PGRP-SC1a 0.59 Anp 0.44 tub 0.43
17 IM4 -0.56 Lectin-galC1 0.43 Nos 0.42
18 Bc -0.53 Tl 0.41 ref(2)P 0.50
19 ref(2)P 0.50
20 PGRP-SC2 0.49
21 ik2 -0.48
22 BEST:GH02921 -0.48
23 CG3066 0.46
24 CG8193 0.45
Factor analysis for candidate genes for immune function. There were 53 candidate genes and a three-factor model was fitted. The genes that loaded with a value greater than 0.40 are listed here. For each factor, the first column is the gene symbol name from [37] and the second column is the loading value for that gene. Genes are considered as loading 'significantly' if the absolute value of the loading value is ≥ 0.40. Genes are considered as loading 'high' if the absolute value of loading value is ≥ 0.70.
Table 3 Factor analysis for candidate genes for body size
Factor 1 Factor 2 Factor 3 Factor 4
Name Load p1 Name Load p1 Name Load p1 Name Load p1
1 Cdk4 0.98 0.21 l(2)gl -0.90 0.29 Jheh3 -0.97 0.02 dpp 0.79 0.25
2 Kr-h1 0.95 0.01 CkIIbeta 0.86 0.67 cdc2c 0.88 0.21 per 0.72 0.03
3 sqh 0.93 0.02 betaTub85D -0.84 0.00 tgo 0.84 0.16 Top1 0.71 0.41
4 trx 0.93 0.26 lilli 0.80 0.09 jar 0.75 0.07 Jheh1 -0.69 0.00
5 babo 0.90 0.00 RpS3 0.78 0.10 Fs(2)Ket 0.75 0.02 wupA 0.65 0.11
6 Akt1 0.88 0.15 Cg25C 0.76 0.13 Sh 0.73 0.11 Fas2 -0.64 0.50
7 fru 0.85 0.13 tra -0.75 0.20 corto 0.72 0.03 Eh -0.62 0.12
8 vg -0.83 0.00 CG17309 0.73 0.00 tok -0.67 0.17 tkv 0.62 0.10
9 fng 0.78 0.56 dnc -0.70 0.04 Jheh2 -0.65 0.00 tra 0.62 0.20
10 RpS13 0.72 0.05 mbt 0.69 0.18 woc 0.63 0.32 sbr 0.61 0.15
11 Dp 0.69 0.36 debcl -0.69 0.08 Pi3K92E 0.57 0.12 Jheh2 -0.60 0.00
12 Mef2 0.68 0.06 RpS6 0.66 0.07 qm -0.56 0.26 puc 0.59 0.01
13 Rac2 0.68 0.00 rut 0.64 0.07 aur 0.55 0.02 Nos 0.56 0.02
14 shot 0.65 0.00 ben 0.62 0.24 dare -0.55 0.18 qm 0.56 0.26
15 puc 0.62 0.01 M(2)21AB -0.59 0.17 Jheh1 -0.55 0.00 Dr -0.56 0.19
16 M(2)21AB 0.61 0.17 bon 0.59 0.11 Nos -0.52 0.02 Eip75B 0.53 0.00
17 Dr 0.60 0.19 l(3)mbt 0.58 0.01 hh 0.51 0.04 Pk61C 0.52 0.23
18 trk -0.58 0.27 Pka-C1 0.57 0.55 Pk61C 0.50 0.65 ftz-f1 0.51 0.08
19 Eip75B 0.58 0.00 Eip63E -0.57 0.15 fru 0.50 0.11 CG11910 -0.49 0.02
20 fru 0.56 0.11 tkv 0.54 0.10 M(2)21AB 0.49 0.17 prod -0.47 0.39
21 mask 0.56 0.07 rok -0.54 0.16 mask 0.49 0.07 ninaE -0.47 0.19
22 woc -0.56 0.32 per 0.52 0.03 how 0.49 0.41 dnc 0.45 0.04
23 Dot -0.53 0.11 Sxl 0.50 0.15 neb 0.48 0.74 robl -0.45 0.38
24 Khc 0.53 0.14 neb 0.48 0.74 Egfr -0.47 0.24 InR 0.44 0.01
25 ade2 0.52 0.01 Eh 0.48 0.12 RpS3 -0.47 0.10 vg 0.42 0.00
26 Tsc1 -0.52 0.04 prod -0.47 0.39 mys -0.46 0.11 Pka-C1 -0.41 0.55
27 Fas2 0.51 0.50 wupA 0.47 0.11 robl -0.44 0.38 Khc 0.41 0.14
28 l(3)mbt -0.50 0.01 shot 0.46 0.00 Kr-h1 -0.42 0.01 corto 0.40 0.03
29 Dfd -0.49 0.19 Pk61C -0.44 0.23 Tor 0.41 0.20
30 mys -0.49 0.11 Ddc -0.43 0.19
31 Sh -0.47 0.11 tra2 -0.42 0.24
32 how 0.47 0.41 InR 0.42 0.01
33 Iswi -0.46 0.13 Pi3K92E 0.42 0.12
34 InR 0.45 0.01 hh 0.40 0.04
35 ben 0.44 0.24 Tor 0.40 0.20
36 neb 0.43 0.74
37 Top1 -0.41 0.41
38 tgo -0.40 0.16
Factor analysis for candidate genes for body size. There were 92 candidate genes and a four factor model was fit. The genes that loaded with a value greater than 0.40 are listed here. For each factor, the first column is the gene symbol name from [37] and the second column is the loading value for that gene. Genes are considered as loading 'significantly' if the absolute value of the loading value is greater than or equal to 0.40. Genes are considered as loading 'high' if the absolute value of the loading value is greater than or equal to 0.70. The third column for each factor is the p-value for the individual gene expression value regression on male body size (p1).
Table 4 Factor analysis for putative targets of miRNAs
Factor 1 Factor 2 Factor 3 Factor 4
Name Load p1 Name Load p1 Name Load p1 Name Load p1
1 kel -0.90 0.05 CG5805 -0.92 0.04 CG7995 0.89 0.01 tws 0.80 0.06
2 CG6327 0.89 0.18 Cirl 0.89 0.23 CG4710 0.89 0.07 CG3689 0.78 0.14
3 CG18812 0.88 0.00 CG9245 0.82 0.05 CG4851 0.87 0.19 CG3811 0.77 0.11
4 Cyp314a1 -0.88 0.06 l(2)03709 0.82 0.13 Rpn6 0.87 0.12 CG11883 0.77 0.03
5 Gclc 0.87 0.01 Eh 0.82 0.12 up -0.84 0.24 CG10809 0.74 0.34
6 sima 0.86 0.11 fz 0.81 0.13 Pkc98E 0.82 0.42 CG11128 0.74 0.17
7 CG9924 0.85 0.17 CG6330 0.80 0.11 CSN4 0.80 0.01 CG3961 0.72 0.15
8 CrebA 0.84 0.00 Atpalpha 0.79 0.38 cpo 0.72 0.41 CG5087 0.72 0.06
9 Mbs -0.84 0.04 Surf4 0.79 0.29 drongo 0.72 0.12 unc-13 -0.70 0.07
10 fax 0.79 0.03 sano 0.79 0.02 fng 0.71 0.56 CG12424 0.70 0.16
11 RhoGEF2 0.78 0.11 CG12424 -0.78 0.40 G-oalpha47A 0.70 0.05 CG13344 0.70 0.03
12 CG9664 -0.78 0.34 CG4911 0.78 0.48 PFgn0025879 0.69 0.26 Sap47 0.69 0.05
13 CG2991 0.77 0.12 JhI-21 -0.77 0.34 Cka -0.69 0.12 CG6325 0.69 0.29
14 CG8602 -0.76 0.04 CG9339 0.77 0.09 Ptp99A 0.68 0.07 CG8475 0.68 0.09
15 BicD 0.76 0.07 CG18604 0.76 0.02 cenG1A 0.67 0.22 CG5625 0.67 0.04
16 CG8954 0.76 0.29 G-oalpha47A -0.76 0.02 Ef1gamma 0.67 0.18 PFgn0025879 0.67 0.26
17 dock 0.75 0.08 CG11266 0.76 0.06 CG4452 0.67 0.03 CG5039 0.67 0.01
18 CG10077 0.75 0.03 CG15628 0.75 0.11 CG3764 0.66 0.05 CG3534 -0.67 0.14
19 CG9381 -0.73 0.33 AP-47 0.73 0.15 CG5853 0.66 0.00 Pkc53E 0.65 0.15
20 Mbs 0.72 0.11 CG14762 0.73 0.05 ed 0.65 0.07 CG17646 0.65 0.13
21 BG:DS04929.1 0.71 0.03 CG4963 0.73 0.11 trx 0.65 0.26 CG11178 0.64 0.07
22 bon 0.70 0.11 Eip71CD 0.68 0.09 Cyp18a1 0.64 0.17 CG1814 -0.63 0.00
23 CG9413 0.70 0.60 CG7956 0.67 0.11 SoxN 0.63 0.03 CG14989 0.62 0.07
24 CG6282 -0.69 0.16 ple 0.66 0.17 eIF-5A 0.63 0.23 Sh 0.61 0.11
25 puc 0.69 0.01 CG9297 0.66 0.03 Atpalpha 0.63 0.02 CG9828 0.59 0.20
26 Mkp3 0.68 0.09 CG11198 0.66 0.01 Dr 0.62 0.19 pdm2 -0.58 0.20
27 CG4841 0.68 0.19 Cyp49a1 0.65 0.00 CG3961 0.62 0.15 CG6803 0.58 0.01
28 CG5886 0.67 0.05 gish 0.65 0.04 lack 0.61 0.08 Ptp99A -0.56 0.07
29 CG9934 0.67 0.42 wdp -0.65 0.01 G-oalpha47A -0.61 0.02 Abd-B 0.55 0.06
30 Rab6 0.66 0.09 Pdi 0.65 0.14 CG16971 0.60 0.17 ana 0.55 0.03
31 CG7492 -0.65 0.29 sdk 0.63 0.34 Cdk4 0.60 0.21 CG6199 -0.54 0.26
32 CanA-14F 0.65 0.01 CG4484 -0.63 0.23 CG18854 0.60 0.06 Nmda1 0.54 0.34
33 CG10338 -0.65 0.03 nmdyn-D7 0.62 0.08 aop 0.60 0.01 CG9265 -0.53 0.13
34 CG8617 -0.65 0.22 scrt -0.61 0.17 tws 0.59 0.00 CG10494 0.51 0.26
35 CG6064 0.64 0.01 mys -0.61 0.11 wdp 0.58 0.01 CG9638 0.50 0.05
36 fkh 0.64 0.13 CG11537 0.61 0.33 CG1441 0.57 0.33 RhoGAPp190 0.50 0.00
37 Cka 0.64 0.12 CG8104 0.60 0.07 Cyp49a1 -0.57 0.18 CG4452 0.48 0.03
38 Mkp3 0.62 0.01 Mbs 0.60 0.11 CG17646 0.57 0.02 CG6554 -0.47 0.38
39 trx 0.62 0.26 CG5853 -0.60 0.00 CG4841 0.57 0.19 CG18375 -0.46 0.00
40 CG13586 0.62 0.25 tsl 0.59 0.00 CrebA 0.55 0.10 woc 0.45 0.32
41 amon -0.61 0.21 UbcD2 0.58 0.21 Ef1alpha100E 0.54 0.08 CG8791 0.45 0.03
42 osp 0.60 0.01 ytr 0.58 0.02 Hr39 0.52 0.21 BicD -0.45 0.07
43 Trn 0.60 0.02 G-oalpha47A 0.58 0.05 woc -0.50 0.32 dco 0.44 0.02
44 CG7283 0.60 0.00 CG3800 0.56 0.52 vri 0.50 0.00 amon 0.44 0.21
45 Eip93F 0.59 0.06 Tsf2 0.56 0.21 CG16953 -0.50 0.10 CG1441 -0.44 0.33
46 CrebA 0.59 0.10 BcDNA:LD23587 0.56 0.20 dco -0.49 0.02 CG9297 -0.43 0.03
47 Ptp4E 0.59 0.79 Nmda1 -0.54 0.34 puc 0.49 0.01 CG13586 -0.43 0.25
48 BcDNA:LD32788 0.58 0.13 Ubc-E2H -0.53 0.00 CG8475 0.48 0.09 CG9664 -0.42 0.34
49 Hr39 0.56 0.21 Ac3 0.52 0.23 CG8602 0.48 0.04 scrt 0.41 0.17
50 ed 0.56 0.07 CG15658 -0.51 0.10 CG12424 -0.47 0.16 Cf2 0.41 0.47
51 CG11099 0.56 0.05 bon 0.50 0.11 Asph 0.47 0.02 insc 0.41 0.14
52 sdk -0.56 0.34 sbb 0.49 0.10 BcDNA:LD32788 0.46 0.13 CG4484 0.40 0.23
53 cenG1A 0.55 0.22 SoxN -0.49 0.03 Cf2 -0.46 0.47 Tsf2 0.40 0.21
54 Vha16 0.54 0.01 aop -0.48 0.01 CG9638 0.46 0.05
55 BcDNA:LD23587 0.54 0.20 CG3764 0.47 0.05 CG14762 -0.46 0.05
56 CG18604 0.54 0.02 CanA-14F 0.47 0.01 CG11266 0.45 0.06
57 Trn 0.52 0.07 CG16953 0.47 0.10 Eip93F -0.44 0.06
58 CG15236 -0.52 0.06 CG15236 -0.47 0.06 CG9413 0.44 0.60
59 fng 0.52 0.56 Trn 0.45 0.02 Abd-B 0.44 0.06
60 did -0.51 0.18 Cka -0.45 0.12 CG6707 0.43 0.75
61 CG11537 0.51 0.33 Ptp4E 0.44 0.79 BcDNA:LD21720 -0.43 0.12
62 CG10494 -0.51 0.26 BG:DS04929.1 0.44 0.03 Cyp49a1 -0.43 0.00
63 CG3534 -0.50 0.14 CG18854 0.44 0.06 CG8791 -0.42 0.03
64 pdm2 -0.50 0.20 vri -0.44 0.00 Sh -0.42 0.11
65 CG8451 -0.50 0.06 CG9924 -0.43 0.17 CG15658 0.42 0.10
66 Eip71CD -0.49 0.09 dco 0.43 0.02 l(2)03709 0.41 0.13
67 gish 0.49 0.04 lack 0.43 0.08 CG6803 0.41 0.01
68 Aef1 -0.49 0.11 CG8791 -0.43 0.03 CG15628 0.41 0.11
69 GLaz 0.48 0.11 CG9664 0.43 0.34 Cyp310a1 -0.41 0.00
70 dco 0.47 0.02 CG5087 -0.43 0.06 BicD -0.41 0.07
71 sbb 0.45 0.10 CG14989 -0.42 0.07 tsl -0.41 0.00
72 BcDNA:LD21720 0.45 0.12 CG6064 0.42 0.01 CG7283 0.40 0.00
73 CG1814 0.44 0.00 CG18375 0.42 0.00
74 CG6199 0.43 0.26 ATPCL 0.41 0.00
75 CG4851 -0.42 0.19 puc -0.41 0.01
76 CG16971 0.41 0.17 CG9638 -0.41 0.05
77 Sap47 0.40 0.05 CG6707 0.41 0.75
78 M(2)21AB 0.40 0.17 CG17646 -0.41 0.02
79 CG4452 0.40 0.03 Atet 0.40 0.08
80 Cf2 -0.40 0.47
Factor analysis for putative targets of miRNAs. There were 203 candidate genes and a four-factor model was fit. The genes that loaded with a value greater than 0.40 are listed here. For each factor, the first column is the gene symbol name from [37] and the second column is the loading value for that gene. Genes are considered as loading 'significantly' if the absolute value of the loading value is ≥ 0.40. Genes are considered as loading 'high' if the absolute value of loading value is ≥ 0.70. The third column for each factor is the the P-value for the individual gene expression value regression on male body size (p1).
==== Refs
Delneri D The use of yeast mutant collections in genome profiling and large-scale functional analysis. Curr Genomics 2004 5 59 65 10.2174/1389202043490032
Singh A McIntyre L Sherman L Microarray analysis of the genome-wide response to iron deficiency and iron reconstitution in the cyanobacterium Synechocystis sp. PCC68030 Plant Physiol 2003 132 1825 1839 12913140 10.1104/pp.103.024018
Caldo R Nettleton D Wise R Interaction-dependent gene expression in Mla-specified response to barley powdery mildew. Plant Cell 2004 16 2514 2528 15319481 10.1105/tpc.104.023382
Stern D Perspective: Evolutionary developmental biology and the problem of variation. Evolution 2000 54 1079 1091 11005278
Gibson G Riley-Berger R Harshman LG Kopp A Vacha S Nuzhdin SV Wayne ML Extensive sex-specific non-additivity in gene expression in Drosophila melanogaster. Genetics 2004 167 179 1799 10.1534/genetics.104.026583
Wayne ML Pan YJ Nuzhdin S McIntyre L Additivity and trans-acting effects on gene expression in male Drosophila simulans. Genetics 2004 168 1413 1420 15579694 10.1534/genetics.104.030973
Falconer Mackay T Introduction to Quantitative Genetics 1996 Harlow, UK: Longman
Lynch M Walsh B Genetics and Analysis of Quantitative Traits 1998 Sunderland, MA: Sinauer
Jansen R Studying complex biological systems using multifactorial perturbation. Nat Rev Genet 2003 4 145 151 12560811 10.1038/nrg996
Mackay T The genetic architecture of quantitative traits: lessons from Drosophila. Curr Opin Genet Dev 2004 14 253 257 15172667 10.1016/j.gde.2004.04.003
Adams M Sekelesky J From sequence to phenotype: reverse genetics in Drosophila melanogaster. Nat Rev Genet 2002 3 189 198 11972156 10.1038/nrg752
Kalidas S Smith D Novel genomic cDNA hybrids produce effective RNA interference in adult Drosophila. Neuron 2002 33 2 177 184 11804566 10.1016/S0896-6273(02)00560-3
Goto A Blandin S Royet J Reichhart J Levashina E Silencing of Toll pathway components by direct injection of double-stranded RNA into Drosophila adult flies. Nucleic Acids Res 2003 31 6619 6623 14602922 10.1093/nar/gkg852
Johnston DS The art and design of genetic screens: Drosophila melanogaster. Nat Rev Genet 2002 3 176 188 11972155 10.1038/nrg751
Hammond S Caudy A Hannon G Post-transcriptional gene silencing by double-stranded RNA. Nat Rev Genet 2001 2 110 119 11253050 10.1038/35052556
Fellenberg K Hauser NC Brors B Neutzner A Hoheisel JD Vingron M Correspondence analysis applied to microarray data. Proc Natl Acad Sci USA 2001 98 10781 10786 11535808 10.1073/pnas.181597298
Wouters L Gohlmann H Bijnens L Kass S Mohlenbergs G Lewi P Graphical exploration of gene expression data: a comparative study of three gene expression methods. Biometrics 2003 59 1131 1139 14969494 10.1111/j.0006-341X.2003.00130.x
Hatcher L A Step-by-Step Approach to Using the SAS System for Factor Analysis and Structural Equation Modeling 1994 Cary, NC: SAS
Wright S The method of path coefficients. Math Stat 1934 5 161 215
Dudoit S Fridlyand J Speed T Comparison of discrimination methods for the classification of tumors using gene expression data. J Am Stat Assoc 2002 97 77 87 10.1198/016214502753479248
Tibshirani R Hastie T Narasimhan B Eisen M Sherlock G Brown P Botstein D Exploratory screening of genes and cluster from microarray experiments. Stat Sin 2002 12 47 59
Parmigiani G Garrett E Irizarry R Zeger S Parmigiani G, Garrett E, Irizarry R, Zeger S The analysis of gene expression data: an overview of methods and software. The Analysis of Gene Expression Data 2003 New York, NY: Springer 1 36
Fabrigar L MacCallum R Wegener D Stahan E Evaluating the use of exploratory factor analysis in psychological research. Psychol Methods 1999 4 272 299 10.1037//1082-989X.4.3.272
Preacher K MacCallum R Exploratory factor analysis in behavior genetics research: factor recovery with small sample sizes. Behav Genet 2002 32 153 161 12036113 10.1023/A:1015210025234
Peterson LE Factor analysis of cluster-specific gene expression levels from cDNA microarrays. Comput Methods Programs Biomed 2002 69 179 188 12204446 10.1016/S0169-2607(01)00189-4
Peterson LE CLUSFAVOR 5.0: hierarchical cluster and principal-component analysis of microarray-based transcriptional profiles. Genome Biol 2002 3 software0002.1 0002.8 12184816 10.1186/gb-2002-3-7-software0002
Tseng G Wong W Tight clustering: a resampling-based approach for identifying stable and tight patterns in data. Biometrics 2005 61 10 16 15737073 10.1111/j.0006-341X.2005.031032.x
Partridge L French V Johnston I, Bennett A Thermal evolution of ectotherm body size: why get big in the cold. Animals and Temperature: Phenotypic and Evolutionary Adaptation 1996 Cambridge, UK; Cambridge University Press 265 296
Huey RB Gilchrist GW Carlson ML Berrigan D Serra L Rapid evolution of a geographic cline in size in an introduced fly. Science 2000 287 308 309 10634786 10.1126/science.287.5451.308
Partridge L Farquhar M Lifetime mating success of male fruitflies Drosophila melanogaster is related to their size. Animal Behav 1983 31 871 877
Partridge L Mackay T Aitken S Male mating success and fertility in Drosophila melanogaster. Genet Res 1985 46 279 285
Partridge L Hoffmann A Jones JS Male size and mating success in Drosophila melanogaster and D. pseudoobscura under field conditions. Animal Behav 1987 35 468 476
Caboli F Kennington W Partridge L QTL mapping reveals a striking coincidence in the positions of genomic regions associated with adaptive variation in body size in parallel clines of Drosophila melanogaster on different continents. Evolution Int J Org Evolution 2003 57 2653 2658 14686541
Gorsuch R Factor Analysis 1983 2 Hillsdale, NJ: Lawrence Erlbaum Associates
Cohen J Statistical Power Analysis for the Behavioral Sciences 1988 2 Hillsdale, NJ: Lawrence Erlbaum Associates
Stevens J Applied Multivariate Statistics for the Social Sciences 1996 Lawrence Erlbaum Associates
FlyBase
Andersson M Sexual Selection 1994 Princeton, NJ: Princeton University Press
Enright A John B Gaul U Tuschl T Sander C Marks D MicroRNA targets in Drosophila. Genome Biol 2003 5 R1 14709173 10.1186/gb-2003-5-1-r1
Williams G Linear Algebra with Applications 2005 Boston, MA: Jones and Bartlett
Olshen A Jain A Deriving quantitative conclusions from microarray expression data. Bioinformatics 2002 18 961 970 12117794 10.1093/bioinformatics/18.7.961
Hastie T Tibshirani R Friedman J The Elements of Statistical Learning: Data Mining, Inference, and Prediction 2001 New York: Springer
Neter J Wasserman W Kutner M Applied Linear Statistical Models 1996 New York, NY: McGraw-Hill/Irwin
McDonald R Factor interaction in nonlinear factor analysis. Br J Math Stat Psychol 1967 20 205 15 5591508
Molenaar PC Boomsma D Application of nonlinear factor analysis to genotype-environment interaction. Behav Genet 1987 17 71 80 3593154 10.1007/BF01066011
Fokoue E Titterington D Mixtures of factor analyses. Bayesian estimation and inference by stochastic simulation. Machine Learning 2003 50 73 94 10.1023/A:1020297828025
Lee S Zhu H Maximum likelihood estimation of nonlinear structural equation models. Psychometrika 2002 67 189 210
Gildea J Lopez R Shearn A A screen for new trithorax group genes identified little imaginal discs, the Drosophila melanogaster homologue of human reinoblastoma binding protein 2. Genetics 2000 156 645 663 11014813
Black M Doerge R Calculation of the minimum number of replicate spots required for detection of significant gene expression fold change in microarray experiments. Bioinformatics 2002 18 1609 1616 12490445 10.1093/bioinformatics/18.12.1609
Nuzhdin S Wayne M Harmon K McIntyre L Common patterns of evolution of gene expression level and protein sequence in Drosophila. Mol Biol Evol 2004 21 1308 1317 15034135 10.1093/molbev/msh128
Affymetrix Affymetrix GeneChip Expression Analysis Technical Manual 2000 Santa Clara, CA: Affymetrix
Wayne M Hackett J Mackay T Quantitative genetics of ovariole number in Drosophila melanogaster. I. Segregating variation. Evolution 1997 51 1156 1163
| 15960805 | PMC1175973 | CC BY | 2021-01-04 16:05:39 | no | Genome Biol. 2005 Jun 1; 6(6):R53 | utf-8 | Genome Biol | 2,005 | 10.1186/gb-2005-6-6-r53 | oa_comm |
Subsets and Splits
No community queries yet
The top public SQL queries from the community will appear here once available.