Associations of human leukocyte antigen DRB with resistance or susceptibility to HIV-1 infection in the Pumwani Sex Worker Cohort

Objective: A group of commercial sex workers in the Pumwani Sex Worker Cohort, established in 1985 in Nairobi, Kenya, remain HIV-1 uninfected despite heavy exposure to HIV-1 through active sex work. Previous studies showed that this resistance is associated with a strong CD4+ T-cell response, which suggested that human leukocyte antigen class II antigens are important in resistance/susceptibility to HIV-1 infection. DRB1 is the most polymorphic locus among class II genes and forms haplotypes with DRB3, DRB4 and DRB5. The aim of this study is to investigate the role of DRB alleles/haplotypes on resistance/susceptibility to HIV-1 infection.

Design: In total, 1090 women enrolled in the Pumwani cohort were genotyped for DRB1, DRB3, DRB4 and DRB5 using a high-resolution sequence-based method. Allele/haplotype frequencies were compared between HIV-positive women and women who have remained HIV negative for more than 3 years despite frequent exposure.

Methods: Human leukocyte antigen DRB genes were amplified, sequenced and genotyped using a two-step sequence-based method. Allele/haplotype frequencies were determined using PyPop32-0.6.0. Statistical analysis was conducted using SPSS 11.0 for Windows.

Results: Three DRB1 alleles were associated with resistance: DRB1*010101 (P = 0.016; odd ratio (OR): 2.55; 95% confidence interval (CI): 1.16-5.61), DRB1*010201 (P = 0.019; OR: 1.86; 95% CI: 1.10-3.15), and DRB1*1102 (P = 0.025; OR: 1.72; 95% CI: 1.07-2.78). DRB1*030201 (P = 0.038; OR: 0.48; 95% CI: 0.23-0.98), DRB1*070101 (P = 0.035; OR: 0.54; 95% CI: 0.30-0.97), DRB1*1503 (P = 0.0004; OR: 0.34; 95% CI: 0.19-0.64), and DRB5*010101 (P = 0.001; OR: 0.37; 95% CI: 0.20-0.67) were associated with susceptibility. The haplotype DRB1*1102-DRB3*020201 was associated with HIV-1 resistance (P = 0.041; OR: 1.68; 95% CI: 1.02-2.78), whereas the haplotypes DRB1*070101-DRB4*01010101 (P = 0.041; OR: 0.52; 95% CI: 0.28-0.98) and DRB1*1503-DRB5*01010101 (P = 0.0002; OR: 0.30; 95% CI: 0.15-0.58) were associated with susceptibility. These associations with resistance/susceptibility to HIV-1 were independent of previously reported alleles HLA-DRB1*01 and HLA-A*2301.

Conclusion: Our findings indicate that human leukocyte antigen DRB-specific CD4+ T-cell responses are an important factor in resistance/susceptibility to HIV-1 infection.

Introduction

The human immunodeficiency virus type 1 (HIV-1) is one of the most devastating viruses known; more than 25 million people have died of HIV/AIDS since the first case of AIDS was identified in 1981 (AIDS Epidemic Update, 2005, http://www.unaids.org/epi/2005/doc/report_pdf.asp). Presently, no cure or vaccine is available, and current prevention strategies have not been able to stop the epidemic in developing countries. Furthermore, modern treatments are either too expensive or inaccessible to many infected individuals. Effective vaccines are required to stop the HIV/AIDS pandemic. Developing vaccines against HIV-1 is difficult due to the rapid mutation of HIV-1 and an incomplete understanding of protective immunity against the virus.

For more than 20 years, HIV-1 infection rates have been closely monitored in women enrolled in the Pumwani Sex Worker Cohort in Nairobi, Kenya. A subgroup of women in the cohort remain seronegative and PCR negative despite heavy exposure to HIV-1 ^[1]. Previous studies showed that this apparent resistance to HIV-1 infection is associated with several human leukocyte antigen (HLA) alleles ^[2-5], HIV-1-specific CD8+ and CD4+ T-cell responses ^[6-10], as well as mucosal-neutralizing antibody to HIV-1 ^[11,12]. The strong CD4+ T-cell responses of HIV-1-resistant women to p24 antigen indicate that HLA class II antigens are important, as they are directly involved in the initiation of CD4+ T-cell responses ^[8]. Among HLA class II antigens, DRB is the most polymorphic locus with 619 functional alleles reported as of November 2007 (full list of HLA class II alleles assigned as of July 2006, http://www.anthonynolan.org.uk/HIG/lists/class2list.html). The DRB1 gene forms haplotypes with the other three functional DRB genes including DRB3, DRB4 and DRB5, or none at all. In a previous study, the association of the DRB1*01 allele with resistance to HIV-1 seroconversion was identified in a subgroup of women in the Pumwani Sex Worker Cohort ^[2]. However, the effects of other DRB alleles and DRB haplotypes on HIV-1 seroconversion have not been thoroughly investigated. To comprehensively study the effect of DRB alleles and haplotypes on resistance and susceptibility to HIV-1 infection in the Pumwani Sex Worker Cohort, we conducted a high-resolution sequence-based DRB genotyping of 1090 women. We have identified DRB alleles, genotypes and haplotypes that were significantly associated with either resistance or susceptibility to HIV-1 infection. Our findings provide additional support that HLA-DRB-specific CD4+ T-cell responses are important in resistance to HIV-1 infection.

Participants and methods

Study population

The study population comprised of women enrolled in the Pumwani Sex Worker Cohort, an observational cohort study of risk factors for HIV-1 infection and the immunobiology of HIV-1, established in Nairobi, Kenya in 1985. Cohort design and follow-up have been described elsewhere ^[1]. In brief, samples of peripheral blood were taken from women at cohort entry for HIV-1 PCR and serological analyses. Testing was repeated at 6-month intervals. In total, 1090 women were included in this study, of which 332 were HIV negative and 580 were HIV positive at the time of enrolment, whereas 171 entered the cohort with HIV negative status but subsequently seroconverted at a later time. The HIV status of seven women is currently unknown. Women were classified as HIV-1 resistant on the basis of the criteria that they remain HIV seronegative and PCR negative in the cohort for at least 3 years while active in sex work and still negative at the time of this study. The women who were classified as resistant were all enrolled in the cohort before 1999 with an average follow-up time of 9.6 ± 4.3 years. The HIV-negative women who enrolled later and had a shorter follow-up time were classified as negative and were not included in the comparison between resistant women and positive controls. Informed consent was obtained from all women enrolled in the study. The ethics committees of the University of Manitoba and the University of Nairobi have approved this study.

Human leukocyte antigen DRB genotyping

DNA samples were isolated from whole blood or peripheral blood mononuclear cells (PBMCs) using QIAamp DNA Mini Kit (QIAGEN Inc., Mississauga, Ontario, Canada). A two-step sequence-based genotyping method was used for DRB genotyping ^[13]. DRB genes were amplified using a pair of generic primers that amplify all the DRB genes within an individual. The PCR products were then sequenced and DRB subgroups were assigned using computer software CodonExpress^IM. A second round of PCR reactions was then performed using subgroup-specific PCR primers based on the result from the first round of genotyping. The PCR products were sequenced and allele assignment was performed using computer software CodonExpress™ based on a taxonomy-based sequence analysis method ^[13].

Statistical analysis

HLA-DRB allele and haplotype frequencies, as well as conformity to Hardy-Weinberg equilibrium, were determined using PyPop 32-0.6.0 and association analysis was conducted with SPSS 11.0 for Windows (SPSS Inc., Chicago, Illinois, USA). Standard univariate methods including the χ²-test (P value) as well as crosstab analysis [odds ratio (OR), 95% confidence interval (CI)] were used to determine the relationship between binary outcomes and explanatory variables. According to the Horvitz-Thompson theory, observations have to be weighted inversely to the probability of being included in a sample to reach an unbiased estimate of what is being investigated ^[14]. We generated a weighted variable using logistic regression taking into account patient enrollment and samples being genotyped. We used this parameter to adjust for crosstab analysis. The sample sizes (673 positive women vs. 145 resistant women) were adequate to detect, with 80% power, and a 5% two-tailed significance level (unadjusted for multiple comparisons), a difference between the two groups with an OR of two in the presence of a certain characteristic (e.g., presence of a haplotype) when the prevalence in the (lower) resistant group was 10%. P values were adjusted for multiple comparisons by means of the Einot and Gabriel method (significant at the a < 0.05 level) using a modified SPSS syntax written by David Nichols of SPSS. Allele and haplotype associations that were significant at the a < 0.05 level were used in multivariate analysis. P values are reported as both uncorrected and corrected for multiple comparisons. Alleles that associated with an increased or decreased risk of HIV-1 infection were analyzed for correlations with previously identified alleles associated with resistance or susceptibility to HIV-1 infection using binary logistic and Cox regression analysis as well as layered (stratified) crosstab analysis. Kaplan-Meier plots with the log-rank test were used to examine the time to seroconversion. Only women who were HIV-1 negative at cohort entry and enrolled before 2002 were included in the survival analysis.

Results

Allele frequency distributions of DRB1, DRB3, DRB4 and DRB5 in the Pumwani Sex Worker Cohort

Thirty-eight DRB1 alleles were identified in the Pumwani Sex Worker Cohort. Eight alleles were detected at a frequency of greater than 5%. The three most frequently observed alleles were DRB1*110102 (11.61%), DRB1*130201 (10.69%) and DRB1*1503 (10.23%). Five DRB3 alleles were identified in this cohort, three of which were detected at a frequency of greater than 5%. The most frequently observed alleles were DRB3*020201 (31.65%), DRB3*030101 (14.91%) and DRB3*010101 (10.73%). The only DRB4 allele identified in this population was DRB4*010101, which was detected at a frequency of 10.50%. Only one DRB5 allele, DRB5*010101, was identified in this population at a frequency of 10.50%. We compared the DRB1 allele frequencies identified in the Pumwani cohort with previously reported allele frequencies (dbMHC, http://www.ncbi.nlm.nih.gov/projects/mhc/ihwg.cgi) for populations from sub-Saharan Africa, Europe, and south-east Asia. In general, the allele frequencies of the Pumwani Sex Worker Cohort were very similar to that of populations from sub-Saharan Africa and several notable differences in allele frequencies were observed between the Pumwani cohort and European and south-east Asian populations (Table 1). The DRB1 and DRB3 allele frequencies observed in this cohort did not deviate significantly from Hardy-Weinberg equilibrium. Testing DRB4 and DRB5 for Hardy-Weinberg equilibrium is irrelevant, as no variation was detected in DRB4 and DRB5 allele frequencies within the cohort.

Table 1

Major human leukocyte antigen DRB haplotypes identified in the Pumwani Sex Worker Cohort

Fifty DRB1-DRB3 haplotypes were identified in this cohort. Five of these haplotypes were detected at a frequency of greater than 5%. The most frequently observed haplotypes were DRB1*110102-DRB3*020201 (14.16%), DRB1*130201-DRB3*030101 (13.45%) and DRB1*1102-DRB3*020201 (8.30%). Seven DRB1-DRB4 haplotypes were identified in this cohort. Only the DRB1*070101-DRB4*01010101 haplotype (8.84%) was detected at a frequency greater than 5%. Three DRB1-DRB5 haplotypes were identified in this cohort and the haplotype DRB1*1503-DRB5*010101 (12.98%) was the only one detected at a frequency greater than 5%.

Distributions of DRB1, DRB3, DRB4 and DRB5 alleles and haplotypes in HIV-1-positive women and HIV-1-negative women in the Pumwani Sex Worker Cohort

Several notable differences in DRB1 allele frequencies between HIV-1-positive women and HIV-1-resistant women were observed in the Pumwani Sex Worker Cohort (Table 2). The frequencies of DRB1*010101, DRB1*010201 and DRB1*1102 were significantly higher in HIV-1-resistant women in comparison with HIV-1-positive women. In contrast, the frequencies of DRB1*030201, DRB1*070101 and DRB1*1503 were significantly higher in HIV-1-positive women. No significant differences in DRB3 and DRB4 allele frequencies were observed between HIV-1-positive women and HIV-1-resistant women. The frequency of DRB5*010101, the only DRB5 allele detected in this cohort, was significantly higher in the HIV-1-positive group.

Table 2

We observed many significant differences in the distribution of HLA-DRB haplotypes between HIV-1-positive women and HIV-1-resistant women in the Pumwani Sex Worker Cohort (Table 3). The haplotypes DRB1*120101-DRB3*030101, DRB1*130101-DRB3*010101, DRB1*130201-DRB3*010101 and DRB1*130301-DRB3*030101 were only detected in HIV-1-positive women. The frequencies of DRB1*030201-DRB3*010101, DRB1*070101-DRB4*01010101 and DRB1*1503-DRB5*010101 were significantly higher in HIV-1-positive women compared with HIV-1-resistant women, whereas the frequency of DRB1*1102-DRB3*020201 was significantly higher in the HIV-1-resistant group. These results are exploratory and should be verified through future studies.

Table 3

DRB alleles and haplotypes found in two or less individuals were assigned as rare alleles or haplotypes. We compared the frequencies of rare alleles and rare haplotypes between the HIV-1-positive group and HIV-1-resistant group and did not observe any significant difference in distribution between the two groups (results not shown).

Association of human leukocyte antigen DRB genotypes and haplotypes with resistance or susceptibility to HIV-1 infection

In order to identify HLA-DRB genotypes that were associated with either resistance or susceptibility to HIV-1 infection, we conducted weighted crosstab analysis (Table 4). We confirmed the association of the DRB1*01 genotype with resistance to HIV-1 infection, as previously identified in a subpopulation of the Pumwani Sex Worker Cohort ^[2]. Women who have the DRB1*01 genotype are two times more likely to be HIV-1 negative than those without the genotype. Two DRB1*01 alleles, DRB1*010101 and DRB1*010201, were identified in this population, both were significantly associated with a decreased risk of HIV-1 infection. We also identified five novel associations of HLA-DRB alleles with an increased or decreased risk of HIV-1 infection in this cohort. DRB1*1102, the fifth most prevalent DRB1 allele in this population (7.06%), was significantly associated with a decreased risk of HIV-1 infection. Three DRB1 alleles, DRB1*030201, DRB1*070101 and DRB1*1503, were significantly associated with an increased risk of HIV-1 infection. As expected, DRB5*010101, which was in linkage disequilibrium with DRB1*1503, was also associated with an increased risk of HIV-1 infection. After adjusting for multiple comparisons, only the associations of DRB1*01, DRB1*1503 and DRB5*010101 remained significant.

Table 4

Three HLA-DRB haplotypes were associated with resistance or susceptibility to HIV-1 infection (Table 4). DRB1*1102-DRB3*020201 was significantly associated with resistance to HIV-1 infection (Table 4). DRB1*070101-DRB4*010101 and DRB1*1503-DRB5*010101 were significantly associated with an increased risk of HIV-1 infection. Each of these haplotypes contains alleles that were associated with either resistance or susceptibility to HIV-1 infection individually. The associations of the DRB1*1503-DRB5*010101 and DRB1*01 haplotypes remained significant after adjusting for multiple comparisons.

Association of human leukocyte antigen DRB genotypes and haplotypes with reduced or increased risk of HIV-1 seroconversion

Kaplan-Meier survival analysis identified DRB genotypes and haplotypes that possibly influence seroconversion (Fig. 1). Women with either the DRB1*01 (including DRB1*010101 and DRB1*010201), DRB1*010201 and DRB1*1102 genotypes seroconverted at a significantly slower rate than those without the genotypes (Fig. 1a-c). DRB1*1503, DRB5*010101 and DRB1*1503-DRB5*010101 were consistently associated with an increased rate of seroconversion (Fig. 1d-f). Women with either the DRB1*010101 genotype or DRB1*1102-DRB3*020101 haplotype trended toward a decreased rate of seroconversion (Fig. 1g and h). In contrast, the DRB1*030201 and DRB1*070101 genotypes, as well as the DRB1*070101-DRB4*01010101 haplotype, trended toward an increased rate of seroconversion (Fig. 1i-k). The results from survival analysis were consistent with those from crosstab analysis.

Fig. 1

Multivariate analysis

We conducted binary logistic regression analysis to determine if the associations of the alleles identified in this study were independent of previously reported alleles, such as the resistant allele DRB1*01 and the susceptible allele A*2301 ^[2] (Table 5). Results showed that DRB1*01 and DRB1*1102 were independently associated with resistance. A more significant P value and increased OR for individuals that have both DRB1*01 and DRB1*1102 (P = 0.003; OR: 7.24; 95% CI: 1.67-31.4) suggested that a synergistic effect exists between these two alleles. Similarly, the associations of DRB1*030201, DRB1*070101 and the haplotype DRB1*1503-DRB5*010101 with susceptibility to HIV-1 infection were independent of the previously reported allele A*2301 ^[2]. The associations of these alleles or haplotypes with susceptibility to HIV-1 infection were also independent of each other. In the absence of the DRB1*1503-DRB5*010101 haplotype, A*2301 (P = 0.162; OR: 0.61; 95% CI: 0.30-1.23) was no longer significantly associated with susceptibility, indicating that the susceptibility conferred by A*2301 may be due to the presence of the DRB1*1503-DRB5*010101 haplotype.

Table 5

Cox regression analysis was performed to determine if the alleles associated with a significantly increased or decreased rate of seroconversion associated independently of the resistant allele DRB1*01 and the susceptible allele A*2301 ^[2] (Table 5). We found that both DRB1*01 and DRB1*1102 were independently associated with a decreased rate of seroconversion. The associations of DRB1*030201, DRB1*070101 and DRB1*1503-DRB5*010101 with an increased rate of seroconversion are independent of the previously reported allele A*2301 ^[2]. The results obtained in Cox regression analysis are consistent with that of Kaplan-Meier survival analysis.

Discussion

HIV-1 resistance observed in the Pumwani Sex Worker Cohort and other highly exposed persistently seronegative (HEPS) cohorts is one of the best examples of natural protective immunity to HIV-1 in humans. The variability of HLA plays an important part in different immune responses among individuals. HLA class II molecules are directly involved in the initiation of the CD4+ T-cell immune response via antigen binding and presentation. Our study aims to identify differences in HLA-DRB, the most variable and highly expressed HLA class II locus, between women who have become HIV-1 infected and those who remain HIV-1 negative despite frequent exposure. By high-resolution genotyping of 1090 women in this cohort, we identified several DRB1 alleles and haplotypes that are associated with either resistance or susceptibility to HIV-1 infection. In addition to confirming the previously reported association of DRB1*01 with resistance to HIV infection, we have also identified five novel associations.

The results from studying HIV-1 resistance and susceptibility in the Pumwani cohort could have broad applications in other sub-Saharan African countries. The HLA-DRB allele distribution is very similar between the Pumwani cohort and other sub-Saharan African populations; therefore, it is likely that alleles conveying resistance or susceptibility in the Pumwani cohort might have a similar effect in other sub-Saharan African populations. As mentioned, DRB1*01, associated with resistance in the Pumwani cohort, was also associated with the HIV-negative group in a study from Botswana ^[3]. DRB1*1503, which conferred susceptibility in the Pumwani cohort, was also associated with increased HIV-1 seroconversion in Zambian discordant couples ^[15]. A recent report showed that DR2 (DRB1*1516-DRB5 haplotype) was associated with susceptibility to HIV-1 infection in a South Indian cohort ^[16]; therefore, the findings from the Pumwani cohort might also be applicable to other populations such as the European or south-east Asian populations, which have different allele compositions and frequency distributions.

The importance of these alleles in the immune response is highlighted by previous reports that have identified these alleles as being associated with allergies, infectious and autoimmune diseases. DRB1*15 is associated with susceptibility to a variety of autoimmune diseases such as aplastic anemia ^[17], multiple sclerosis ^[18] and rheumatoid arthritis ^[19]. This suggests that individuals with this HLA genotype tend to have a higher level of immune activation. Higher immune activation could lead to higher rates of HIV-1 transmission due to an increased number of activated CD4+ T cells. The prevalence of autoimmune diseases in the Pumwani cohort has not been examined. Follow-up studies are necessary to determine the relationship among autoimmunity, DRB1*1503 and susceptibility to HIV-1 infection. Although DRB1*01 and DRB1*1503 were associated with different outcomes of HIV-1 infection, both alleles were correlated with susceptibility to autoimmune diseases in non-African populations ^[16-24]. Factors such as immune stimulus, different ethnic backgrounds and environment may account for differences seen in allele associations between these populations. The complexity of anti-HIV-1 immunity further emphasizes the importance of comprehensive studies of large well characterized populations.

Characteristics of HLA molecules, such as epitope binding affinity, can affect the level of immune response generated. DRB1*010101 and DRB1*010201 were associated with resistance in the Pumwani cohort. These two alleles are very similar in peptide sequence and it has been shown that they are able to present similar antigenic peptides, but with varying binding affinities ^[22]. Variability in binding affinity by these two alleles has been shown to lead to differential T-cell proliferation ^[23]. This may lead to varying levels of resistance for DRB1*010101 and DRB1*010201. Further investigation is necessary to determine the binding affinities of these alleles for HIV-1 antigenic peptides and their influence on T-cell proliferation within this cohort.

An important feature of HLA class I alleles that determines their influence on infectious disease is their ability to tolerate epitope mutations. Resistant alleles may have a higher tolerance for these mutations or may recognize epitopes in the conserved region of the pathogen so that infectious pathogens are less likely to escape. Presentation of a nonconserved epitope from a nonessential region of a viral protein would increase the likelihood of immune evasion by an escape mutant. Studies have demonstrated this effect for HLA class I protective and susceptible alleles in HIV-1 disease progression ^[24-29]. It would be interesting to see if this effect observed for HLA class I alleles is also observed for HLA class II. Additional studies are required to characterize the epitopes and observe the effect of mutations on epitope recognition by HLA-DRB alleles and subsequent immune responses.

Linkage disequilibrium with unidentified resistant genes may be responsible for the associations observed. It is important to determine whether or not the alleles and haplotypes identified in this study confer resistance or susceptibility independently of other factors.

We have shown the stronger associations with resistance to HIV-1 infection when both DRB1*01 and DRB1*1102 are present. This additive effect may also exist between HLA-DRB alleles and other HLA class II and HLA class I genes. It is necessary to conduct further studies to characterize the influence of HLA class II haplotypes and HLA class I and II haplotypes on resistance to HIV-1 infection.

The advantage of possessing a rare HLA class I supertype in HIV-1 disease progression has been previously reported ^[30]. HIV is more likely to adapt to the most common HLA types in a given population. Individuals carrying rare HLA class I alleles would have an advantage ^[30]. This phenomenon, however, has not been observed for HLA-DRB in the Pumwani cohort. HLA-DRB alleles that were associated with resistance in the Pumwani cohort are quite common in the population. The rare allele advantages in class I versus common DRB allele associations with resistance may reflect different roles for CD8+ and CD4+ T cells in anti-HIV immune responses. HLA class II is directly involved in the initiation of the CD4+ T-cell immune response. This response is mainly cytokine based and is implicated in the proliferation of activated CD8+ T cells as well as the enhancement of the overall CD8+ T-cell response ^[31-33]. The CD4+ T-cell immune response is also directly involved in the activation of antigen-presenting cells, dendritic cells and B cells. The CD8+ T-cell response is mainly responsible for controlling viral spread during the acute and chronic phases of infection ^[34-37]. Currently, we know very little about why these DRB alleles are associated with resistance. Further investigation is required to elucidate these mechanisms and enhance the understanding of HLA-DRB based resistance to HIV-1 infection.

Acknowledgements

This work was supported by the National Institutes of Health and the Canadian Institutes of Health Research and the Bill and Melinda Gates Foundation. We thank Gary Van Domeselaar for assisting in data conversion; Tony Kariri for his dedication in maintaining the databases of both cohorts at the University of Nairobi and Bing-hua Liang for managing the database at the University of Manitoba; Harold Peters, John Rutherford, and Leslie Slaney for laboratory support. We thank the staff of the Majengo clinic, Jane Njoki, Jane Makene, Elizabeth Bwibo, Edith Amatiwa; and the women of the Pumwani Sex Worker Cohort for their continued participation and support. F.A.P. is a Canadian Institutes of Health Research Senior Investigator and is currently a Tier I CIHR Canada Research Chair.

Sponsorship: Funding for this work has been provided by the Bill and Melinda Gates Foundation (Gates Grant #37873), Canadian Institutes of Health Research (BMG 77515) and the National Institutes of Health (R01 AI56980).

This study was presented at a Poster Presentation, International Center for Infectious Disease Research and Innovation Retreat, November 2005, Winnipeg, Manitoba, Canada, Poster Presentation, International AIDS Conference, August 2006, Toronto, Ontario, Canada.

Contributors: P.A.L. provided the data generation, analysis and interpretation, as well as drafting of the manuscript. J.D.H. provided data generation, analysis and interpretation, as well as drafting of the manuscript. M.L. provided data generation, analysis and interpretation, as well as editing. N.J.D.N. provided data analysis and interpretation, as well as editing. T.B. provided data generation and editing. J.K. maintained the Pumwani sex worker cohort and was involved in the acquisition of data and editing of the manuscript. C.W. maintained the Pumwani Sex Worker Cohort and was involved in the acquisition of data and editing of the manuscript. E.N.N. established the Pumwani Sex Worker Cohort and was involved in the acquisition of data. F.A.P. and M.L. conceived and designed the study. F.A.P., the overall principal investigator, secured funding for the study, and established and maintained the Pumwani Sex Worker Cohort and was involved in the acquisition of data and editing of the manuscript.

All authors reviewed and approved the final version of the paper.

There is no conflict of interests.