The small scale spatial dynamics of HIV-1 transmission in Rakai, Uganda

Transcription

1 The small scale spatial dynamics of HIV-1 transmission in Rakai, Uganda by Mary Kathryn Grabowski A dissertation submitted to the Johns Hopkins University in conformity with the requirements for the degree of Doctor of Philosophy Baltimore, MD February 2014

2 Abstract The highest prevalence of viral sexually transmitted infections is found in sub- Saharan Africa where very little is known about the dynamics of viral transmission within local sexual networks. An understanding of these dynamics is essential for the control of sexually transmitted infections within and between communities. Here we analyzed the small scale spatial dynamics of HIV transmission in rural Rakai District, Uganda using data from a cohort of 14,594 individuals residing within 46 communities. We applied spatial clustering statistics, molecular phylogenetics, and egocentric transmission models to quantify the relative contribution of viral introductions into communities versus community and household-based transmission to HIV incidence. We found that the majority of HIV transmissions acquired outside of households are acquired from non-stable sexual partners who reside outside of an individual s community of residence. Our results have critical implications for the design of community randomized trials which assume that the preponderance of HIV transmissions occur within local sexual networks by design. They also suggest that HIV prevention efforts should be implemented at spatial scales broader than the community and target key populations responsible for introductions into communities. We next examined household-based HIV transmission in Rakai, Uganda, focusing on the dynamics of HIV introduction into stable concordant HIV-negative couples. Using data from more than ten thousand couples, we found men introduced HIV into initially uninfected couples more often than did females, though female partners accounted for a substantial proportion of HIV introductions. HIV was introduced into household-based partnerships most frequently by young men and women, the latter of whom were at ii

3 especially high risk from extra-marital sexual partners. This study was conducted before and after the introduction of antiretroviral therapies and rapid scale up of medical male circumcision in Rakai. We found that incidence of HIV introduction into stable partnerships declined significantly with the scale up of HIV prevention services, though the protective benefits were only observed in males. Lastly, this dissertation examined the dynamics of high risk human papilloma virus (HR-HPV) viral load and persistent HR-HPV detection among HIV-infected and uninfected men in Rakai, Uganda. Using data from a randomized clinical trial, we found HR-HPV genotypes with high viral load are more likely to persist among both HIVnegative and HIV-positive men, though persistence is more common among HIVpositive men overall. These results may explain the association between high HR-HPV viral load and HR-HPV transmission and increased levels of HR-HPV detection among individuals co-infected with HIV. Thesis committee Ronald H. Gray (Thesis advisor), Chris Beyrer (Academic adviser), Thomas C. Quinn, and Lawrence Moulton iii

4 Acknowledgements This research was a collaborative effort between the Johns Hopkins Bloomberg School of Public Health, the Rakai Health Sciences Program, and the laboratory of Dr. Thomas Quinn. It was directly overseen by my thesis adviser, Dr. Ronald Gray. Dr. Gray provided me with outstanding mentorship and support as a doctoral student. Dr. Thomas Quinn provided laboratory support and mentorship, and it was through his laboratory that I learned to unify the epidemiological and evolutionary dynamics of HIV. Dr. Andrew Redd and Dr. Oliver Laeyendecker also provided invaluable mentorship in the laboratory. I owe many thanks to Dr. Justin Lessler who provided countless hours of one-onone mentorship in epidemiology and statistics. Justin is also a great and very patient friend who sacrificed much time on my behalf, and for that I am eternally grateful. I also thank Dr. Aaron Tobian. Dr. Tobian provided me with multiple opportunities to work on a variety of research projects as a doctoral student. I thank Dr. Chris Beyrer, my academic adviser. Dr. Beyrer s unwavering commitment to public health and human rights served as great inspiration to me during this process. This research would not have been possible without the Rakai Health Sciences Program and the participants of the Rakai Community Cohort Study. Dr. Serwadda and the Rakai investigators have provided me and the world with an incredible resource. I only hope that these efforts do their years of hard labor some justice. Lastly, I would like to thank my family and friends, especially my parents, Andrew, and Archer. Your love has been the source of my successes and any remaining sanity. iv

5 Table of Contents Abstract..ii Acknowledgements...iv Table of Contents v List of Table...xi List of Figure....xv Chapter 1: Introduction Organization.3 Chapter 2: Background A brief overview of the epidemiology of HIV/AIDS in sub-saharan Africa The HIV epidemic in Uganda: past and present An overview of the molecular biology of HIV and genomic organization Classification Replication Genomic organization The natural history HIV infection Mechanisms of HIV transmission and infection Early stage infection Chronic stage infection...15 v

6 Advanced stage The evolutionary dynamics of HIV within individuals and populations The molecular epidemiology of HIV in sub-saharan Africa Epidemiological insights into the transmission dynamics of HIV in sub- Saharan Africa Probability of HIV transmission per coital act in heterosexual couples Sexual contact networks and HIV spread in sub-saharan Africa References.30 Chapter 3: Molecular tools for studying HIV transmission in sexual networks: A review Abstract Introduction HIV sequence data for phylogenetic analyses Viral sequencing technologies Reconstructing a phylogenetic tree Viral linkage analysis HIV transmission patterns from phylogenetics Conclusions Keynotes...52 vi

7 3.10. References.57 Chapter 4: The role of viral introductions in sustaining community-based HIV epidemics in rural Uganda: evidence from spatial clustering, phylogenetics, and egocentric transmission models Abstract Introduction Materials and Methods Ethics Statement Study population and setting Spatial clustering analyses Viral extractions and HIV-1 subtype assignment Phylogenetic analysis Egocentric transmission model Results Spatial clustering of HIV-seropositive individuals Spatial clustering of HIV-seropositive individuals within households Spatial clustering of HIV-seropositive individuals within communities HIV phylogenetics within and across communities Genetic relatedness of HIV viruses within households...80 vii

8 Genetic relatedness of HIV viruses within and across communities Probable infection from household, community and extracommunity sources Attributable fractions of HIV infections from householdbased transmission Attributable fractions of HIV infections from community, extra-community and unknown sources Sensitivity Analyses Discussion References...93 Chapter 5: HIV transmission dynamics in stable heterosexual couples: A retrospective study of initially concordant HIV-negative couples in Rakai Uganda Abstract Introduction Methods The Rakai Community Cohort Study (RCCS) Identification of stable couples and primary study outcomes Primary study exposures Statistical Analyses viii

9 5.4. Results Self-report of extra-couple sexual partnerships by age and gender Incidence of HIV introduction into initially concordant HIVnegative couples Self-report of extra-couple sexual partnerships among introducing partners Genital ulcer disease and male circumcision as risk factors for HIV introduction Scale-up of HIV treatment and prevention services and risk of HIV introduction Discussion References 165 Chapter 6. High risk human papillomavirus viral load and persistence among heterosexual HIV-negative and HIV-positive men Abstract Introduction Materials and methods Study design and participants HR-HPV Detection, Viral load quantification, and STI Testing Statistical Analysis Results ix

10 6.5. Discussion References.194 Chapter 7. Conclusions.205 Curriculum Vitae x

11 List of Tables Chapter 3. Table 1. Overview of common sequencing technologies for HIV research. 53 Table 2. Summary of participant recruitment and phylogenetic methods for a random sample of 20 phylogenetic studies of HIV-1 transmission clustering published in PubMed Central in Chapter 4. Table 1. Summary statistics for 46 communities (within 11 geographic regions) surveyed in RCCS R Table 2. Characteristics of 95 phylogenetic clusters identified in maximum likelihood phylogenetic analyses (HKY-85) of 915 gag sequences and 1026 env sequences obtained from 1,099 HIV-infected participants in RCCS R Table 3. Descriptive characteristics of HIV seronegative and -incident participants in ego-centric partner analysis (N=9,520) Table 4. Attributable HIV transmissions by the geographic location of sexual partner and gender of newly infected participant (estimated from ego-centric transmission model)..102 Table 5. Probability of HIV-infection by partner type over 18 month interval..103 xi

12 Table S1. Accession numbers for Los Alamos National Laboratory HIV Sequence Database reference sequences used for maximum likelihood and Bayesian phylogenetic analyses Table S2. Summary of HIV sequences from 1,434 HIV-1-seropositive participants in RCCS R Table S3. Summary of HIV sequences obtained from 189 HIV-1-incident participants in RCCS R Table S4. Summary phylogenetic data (HIV subtype, genetic pairwise distance, and phylogenetic clustering results) for the 105 epidemiologically linked incident couples with phylogenetic data in gag and/or env gene regions..120 Table S5. Detailed summary data for each of the 95 phylogenetic clusters identified in maximum likelihood phylogenetic trees (HKY-85 model) Table S6. Sensitivity analyses of phylogenetic clustering results to choice of evolutionary model and bootstrap and genetic distance thresholds..129 Table S7. Numbers of recent sexual partners self-reported by 9,520 HIV-seronegative and -incident participants in egocentric analysis by gender and marital status of the study participant 130 Table S8. Summary of self-reported sexual partner data from 9,520 HIV-seronegative and -incident participants in egocentric analysis by gender of the study participant and geographic location of the sexual partner 131 xii

13 Table S9. Comparison of demographics and sexual behaviors (percent distribution) between RCCS study population (RCCS R13, ) and the surveyed population in the 2011 Ugandan Demographic and Health Survey Chapter 5. Table 1. Univariate and multivariate associations between couple demographics and selfreport of one or more extra-couple, unstable sexual partnerships at baseline (n=4,570) and follow-up study visits (n=12,319) by gender in 4,570 initially HIV-1 seroconcordant negative couples in the RCCS, Table 2. Relative hazard of HIV introduction and self-report of extra-couple sexual partnerships, genital ulcer disease and male-circumcision status 171 Table S1: Comparison of selected baseline characteristics for concordant HIV-1 negative couples with and without follow-up in the RCCS, Table S2: Characteristics of non-stable extra-couple sexual partners of incident male and female HIV-1 cases (at T S and T S-1 visits) and partners of matched HIV-negative controls.175 Table S3. Risk of HIV introduction with higher community coverage of ART in HIV infected persons and MMC in non-muslim men.177 Table S4. Sensitivity of the unadjusted and adjusted pre-post ART and MMC hazard ratios to choice of cutoff Table S5. Sensitivity of the unadjusted and adjusted pre-post ART and MMC hazard ratios to exclusion of a single study visit.178 xiii

14 Chapter 6. Table 1: Associations between qualitative HPV viral load and demographic and clinical factors among HR-HPVs (n=802) detected at baseline Table 2: Persistence of high vs. low viral load baseline HR-HPVs at 6,12, and 24 months follow-up.199 Table 3: Association between viral persistence at 24 months follow-up and HPV viral load at month of initial detection among newly detected HR-HPVs Table S1: Baseline Characteristics of HIV-positive and HIV-negative men by HR-HPV status Table S2. Distribution of HR-HPV genotypes by qualitative HPV viral load at baseline and HIV co-infection status.203 Table S3: Patterns of HPV detection for baseline HR-HPVs with complete follow-up* by qualitative HPV viral load at baseline and HIV co-infection status xiv

15 List of Figures Chapter 2. Figure 1. Community HIV prevalence in the Rakai District, The red histogram shows the distribution of HIV prevalence across 50 Rakai communities, while the HIV point density map on the right shows that the highest levels of HIV infection are found along Lake Victoria and the main roads...6 Figure 2. HIV prevalence in Uganda, A) HIV prevalence in Kampala by age group and urban /rural settings vs. elsewhere in Uganda. B) Declines in HIV prevalence were not observed in other sub-saharan African countries over same time period (taken from Stoneburner et al, Science, 2004) 8 Figure 3. Gene map of the HXB2 HIV-1 reference genome. Numbers in the upper left and low right hand corners of the rectangles correspond to the gene start and end positions of the HXB2 reference genome, respectively (taken from Fields Virology, 5 th edition, 2007) Figure 4. The natural history of HIV-1 infection in the first 1000 days of infection. The natural history of HIV-1 can be divided into a series of stages characterized by the absence (-) and presence (+) of viral antigens and HIV-specific immune responses in the peripheral blood (taken from Cohen et al., NEJM, 2011).. 14 xv

16 Chapter 3. Figure 1. Rakai District, Uganda. (A) Rakai (~2,200 km 2 ), a rural district in southwest Uganda, with population ~450,000 (~700 communities). RCCS R13 study participants (n = 1,085) reported 1,169 sexual partners with primary residence outside the Rakai District, but within Uganda (where disclosed, residential locations of sexual partners are indicated with red dots on the map). Only three sexual partners were reported to be living outside Uganda (two in Tanzania and one in the United Kingdom, not shown). (B) The Rakai district at a higher resolution, with the 11 geographic regions surveyed in RCCS R13 indicated in color. There are two primary highways (Masaka Road to Tanzania and the Trans-African National Highway to Rwanda and the Democratic Republic of the Congo [DR of Congo]) and numerous secondary roads that extend throughout the district..104 Figure 2. Spatial clustering of HIV-seropositive persons within households (0 km) and in geographic windows of 250 m up to 10 km (the first window is m and windows are centered every 50 m starting at 125 m). Spatial clustering analyses show whether HIV prevalence or incidence is elevated within certain distances of other HIVseropositive persons. We define the spatial clustering of HIV-seropositive individuals as τ(d 1,d 2 ), the relative probability that an HIV-seropositive person resides within a distance window, d 1 to d 2, from another HIV-seropositive person compared to the probability that any individual is HIV seropositive in the entire study population. Where spatial clustering exists, values of τ(d 1,d 2 ) exceed one. Shaded areas show the 95% bootstrapped confidence intervals for spatial clustering estimates. (A) The spatial clustering between HIV-seropositive persons (prevalent or incident cases with other prevalent or incident xvi

17 cases; red). (B) The spatial clustering of HIV-seroincident cases with ART-naïve HIVseroprevalent persons (yellow). (C) The spatial clustering of HIV-seroincident cases with other HIV-seroincident cases (blue). (D) A blowup of the area where significant extrahousehold spatial clustering (<500 m) was identified among all HIV-seropositive persons (marked with black box in [A C]). Data are shown only up to 10 km (no significant spatial clustering was observed beyond this distance).105 Figure 3. Maximum likelihood phylogenetic analyses of the HIV-1 gag gene. (A) Boxplots of the intra-subtype gag genetic pairwise distances for epidemiologically linked (Epi linked) incident couples (i.e., at least one member of the couple was an incident case) and for all epidemiologically unlinked incident pairs of individuals in RCCS R13. (B) Boxplots of intra-subtype gag genetic pairwise distances by the geographic distance between the incident pair. (C) A ML phylogenetic tree (radial) of HIV-1 subtype A gag sequences from HIV-seroprevalent (n = 245) and HIV-incident (n = 55) cases, where taxa are colored by the geographic region from which they were isolated. Reference strains (n = 87) are in black. Grey circles indicate nodes with bootstrap support of 70%; black circles indicate intra-household clusters; indicates an intra-household virus also sharing a cluster with at least one other household. Additional radial and rectangular phylogenetic trees for HIV-1 subtypes A, D, and C for gag and env genes are included in Figures S5 S Figure 4. Summary of inferential methods and study results and conclusions. The dotted blue line represents the border of a hypothetical community xvii

18 Figure S1. The geographic scale of RCCS communities. Communities are color-coded according to their RCCS geographic region (see Figure 1 for color key). The means for the average and maximum geographic distances between households within a community (across all communities) are marked with dotted red lines. The size of the dot is proportional to the size of the surveyed population/community size..133 Figure S2. Phylogenetic analyses of gag and env genes for specimens that underwent repeated viral RNA extraction and PCR testing. Repeated viral RNA extractions and PCR testing was performed for a sample of patient specimens for gag (n = 26) (A) and env (n = 46) (B) to assess the reliability of our laboratory methods. Sequences were compared using neighbor-joining trees (1,000 bootstrap replicates). Trees were constructed separately for each gene region using a Tamura-Nei model of nucleotide substitution. Results of the phylogenetic analyses showed that the laboratory methods yielded reliable sequence information: sequences obtained from the same individual always clustered together. 134 Figure S3. Genetic pairwise distances in gag and env genes for epidemiologically linked HIV-infected couples where at least one partner was an HIV-incident case. Figures show only those incident couples who shared a monophyletic clade in a ML tree with 70% or greater bootstrap support. These distributions were used to determine the genetic distance thresholds for phylogenetic cluster analyses. 135 Figure S4. Spatial clustering of HIV-seroprevalent persons on ART with HIVincident cases within households (0 km) and in geographic windows of 250 m up to 10 km (every 50 m beginning at 125 m). Spatial clustering, τ(d 1,d 2 ), shown in black, is xviii

19 the relative probability that an HIV-seroprevalent person on ART resides within a distance range, d 1 to d 2, from an incident case compared to the probability that any individual participant is an incident case. The shaded area is the bootstrapped 95% confidence interval (1,000 iterations)..135 Figure S5. Maximum likelihood tree (radial) of gag HIV-1 subtype D sequences. Color corresponds to geographic region (see Figure 1 key). Reference sequences (n = 57) are in black. Grey circles indicate nodes with bootstrap support of 70%; black circles indicate intra-household clusters; indicates intra-household viruses also sharing a cluster with at least one other household.136 Figure S6. Maximum likelihood tree (rectangular) of gag HIV-1 subtype C sequences. Taxa are labeled using participant gender/geographic region/community/household. Reference sequences (n = 37) are in black, and only bootstrap values 50% are shown. Color corresponds to the geographic region Figure S7. Maximum likelihood tree (radial) of env HIV-1 subtype A sequences. Color corresponds to geographic region (see Figure 1 key). Reference sequences (n = 107) are in black. Grey circles indicate nodes with bootstrap support of 70%; black circles indicate intra-household clusters; indicates intra-household viruses also sharing a cluster with at least one other household Figure S8. Maximum likelihood tree (radial) of env HIV-1 subtype D sequences. Color corresponds to geographic region (see Figure 1 key). Reference sequences (n = 70) are in black. Grey circles indicate nodes with bootstrap support of 70%; black circles indicate intra-household clusters; indicates intra-household viruses also sharing a cluster with at least one other household.139 xix

20 Figure S9. Maximum likelihood tree (rectangular) of env HIV-1 subtype C sequences. Taxa are labeled using participant gender/geographic region/community/household. Reference sequences (n = 37) are in black, and only bootstrap values 50% are shown. Color corresponds to the geographic regions 140 Chapter 5. Figure 1. Probability of self-reporting one or more non-stable sexual partners and incidence of a first HIV introduction event (male or female, female only, and male only) among initially HIV concordant negative couples by age and gender of the male and female partners in the couple. A) Probability of a female self-reporting one or more non-stable extra-couple partners by her age. B) Probability of a male selfreporting one or more non-stable extra-couple partners by his age. Note difference in y- axes between Figures 1A and 1B. C-D) Incidence of any HIV-1 introduction (black), female introduction only (red), and male introduction only (blue) by age of female (C) and age of male (D)..170 Figure 2. Time trends in ART and MMC coverage, consistent condom use with nonstable sexual partners, extra-couple partnership selection, and self-reported genital ulcer disease in the full RCCS cohort (solid lines) and in the incidence cohort of HIV concordant negative couples (dotted lines). A) Percentages of all HIV infected males and females in the full RCCS cohort who were receiving anti-retroviral therapies B) Percentages of non-muslim men in full RCCS and incidence cohorts who were circumcised. C) Percentages of non-stable sexual partnerships in which there was consistent condom use as self-reported by males and females in the full RCCS and xx

21 incidence cohorts. D) Percentage of males in the incidence cohort who reported one more non-stable sexual partnerships. E) Percentage of females in incidence cohort who reported one or more non-stable sexual partnerships. F) Percentages of HIV-negative males and females who self-reported symptoms of genital ulcer disease in the full RCCS and incidence cohorts Figure 3: Incidence of HIV introduction into stable HIV concordant negative couples before and after the scale-up of antiretroviral therapies (ART) and medical male circumcision (MMC) in Rakai. A) Incidence of HIV introduction per 100 coupleyears pre and post ART and MMC scale-up. B) Relative hazard of any introduction (black), female introduction (red), and male (blue) in the post scale-up vs. pre-scale up eras. C) Proportion of virus introduced by males (M+/F-), females (M-/F+), or unknown partner (M+/F+) in pre and post scale-up eras 173 Figure S1. Age of male partner vs. age of female partner for stable sexual partnerships (index/incidence cohort partners) (A) and for extra-couple non-stable sexual partnerships self-reported by B) incidence cohort females and C) incidence cohort males. Colors indicate the HIV-1 status of RCCS participant. The black dotted is the diagonal. The black solid line is the best fit line of male index age vs. female index age using a linear generalized additive model. Similarly the pink line is the best fit line for female index vs. the age of her extra-couple partner and the cyan line the best fit line for index male vs. the age of his extra-couple partner 176 xxi

22 Chapter 6. Figure 1: Viral load of HR-HPV detected at baseline at 6, 12, and 24 months followup visits by baseline HR-HPV viral load and HIV status xxii

23 Chapter 1 Overview The highest prevalence of viral sexually transmitted infections is found in sub- Saharan Africa where very little is known about the dynamics of viral transmission within local, community-based sexual networks. An understanding of viral transmission dynamics is essential for the control of sexually transmitted infections within and between communities in sub-saharan Africa. This dissertation predominantly focuses on the small scale spatial dynamics of human immunodeficiency virus (HIV) transmission in rural Rakai, Uganda. Rakai, bordered by Lake Victoria to the East and Tanzania to the South, was the initial epicenter of the HIV epidemic in Eastern Africa. Today, HIV transmission in Rakai is endemic with an HIV prevalence of 12% and incidence 1.2 per 100 person-years. We use data from the Rakai Community Cohort study (RCCS) for the first two of three research studies included in this dissertation. The RCCS is an ongoing longitudinal study of HIV incidence, sexual behaviors, and health service utilization in the Rakai District, Uganda. The RCCS has been ongoing since 1994, and there have been fifteen survey rounds of data collection (RCCS R1 through RCCS R15) to date. Data for the first two studies were obtained between 1997 and 2011 (RCCS R4 through RCCS R14), during which time antiretroviral therapies (ART) were introduced into Rakai and the prevalence of medical male circumcision (MMC) was dramatically increased among non- Muslim males. The RCCS obtains biological specimens and detailed demographic, sexual behavioral and personal (i.e. egocentric) sexual network data from over 15,000 study 1

24 participants on an annual basis. Approximately 70% of censused adults are surveyed, and of those persons more than 99% agree to HIV testing and counselling. Our first study examined the role of cross-community HIV-1 transmission in sustaining HIV epidemics within 46 rural Rakai communities using RCCS data from RCCS R13 ( ). We applied global spatial clustering statistics, molecular viral phylogenetics, and egocentric transmission models to quantify the spatial scale of personto-person HIV spread. We also estimated the proportion of total incident HIV infections within a community that are introduced from elsewhere (i.e. via sexual partnerships in which persons reside in different communities). Prior to our first study of community-based HIV-1 transmission dynamics, we conducted a literature review on molecular phylogenetic tools to study HIV spread in sexual networks. This review focused on potential biases in molecular epidemiological studies of HIV transmission, including biases from the sampling, molecular methods, and statistical analyses. We specifically emphasized the challenges in identifying HIV transmission chains using viral phylogenetic data obtained from sparsely sampled networks. We also identified areas of future theoretical and empirical work needed for more informative phylogenetic analysis and insight into HIV transmission networks. The second study in this dissertation examined the dynamics of HIV introduction into household-based sexual partnerships in Rakai, Uganda. Using data from more than ten-thousand heterosexual couples in the RCCS, we sought to determine (1) whether male or female partners are more likely to bring HIV-infection into initially HIVuninfected dyads, (2) the proportion of newly introduced virus that results in rapid transmission to non-introducing stable partners, and (3) the validity of self-reported extra- 2

25 couple partnership data among HIV introducing partners. We also (4) examined the impact of scale-up of ART and MMC services on the incidence of HIV introduction into household-based partnerships. Lastly, this dissertation examined the dynamics of human papilloma virus (HPV) viral load and persistent HPV detection among HIV-infected and uninfected men. Though HPV associated diseases in men are relatively rare as compared to in women, men are the primary source of HPV infection to women, who in Uganda have among the highest rates of cervical cancer worldwide. We used data from a randomized clinical trial of MMC in Rakai to study the association of HPV viral load and HPV detection at three study visits over a twenty-four month period Organization This dissertation is organized into seven chapters. The first chapter presents an overview of the dissertation contents and its organization. The second chapter provides relevant background on the biology and epidemiology of HIV with a focus on heterosexual HIV transmission among adults in sub-saharan Africa. The third chapter presents a review of molecular tools to study HIV transmission within sexual networks. This review focuses on the potential biases in molecular epidemiological studies of HIV transmission networks. The next three chapters (4-6) detail the methods and results from our three primary research studies. The fourth chapter summarizes our study on community-based HIV-1 transmission dynamics in Rakai, Uganda using spatial statistics, viral phylogenetics, and egocentric transmission models. The fifth chapter summarizes our findings from a study 3

26 on household-based HIV-1 transmission dynamics among initially HIV-uninfected stable couples. The sixth chapter presents the third and final study on HPV viral load and persistence of HPV in HIV-infected and uninfected men. We close with a discussion of our results from chapters 4-6 in chapter 7, which includes a discussion of the public health significance of the dissertation findings and future studies. 4

27 Chapter 2 Background 2.1. A brief overview of the epidemiology of HIV/AIDS in Sub- Saharan Africa More than two thirds of persons infected with the human immunodeficiency virus (HIV) reside in sub-saharan Africa, where approximately 67% of incident infections occurred globally in Though the first reported cases of HIV were among mobile traders and commercials sex workers in central and eastern countries 2 5, today s African HIV epidemic is generalized, affecting large numbers of men and women from low to high risk demographic and behavioral profiles continent-wide 6,7. The majority of sub- Saharan Africans acquire HIV through heterosexual intercourse 8, though there is growing evidence that men who have sex with men (MSM) and injection drug users (IDU) are also highly affected subgroups Overall, the disease burden is higher among women than men (58% vs 42%) 5,6 ; however, HIV prevalence and risk factors vary markedly across geographic and cultural contexts 14. There is substantial spatial heterogeneity in HIV prevalence throughout sub-saharan Africa. At the largest spatial scale variability is observed between the central, eastern, southern and western regions with the southern countries having the greatest disease burden both within the African continent and worldwide 1,6. Of the ten most HIV-affected countries globally, nine are in Southern Africa 1, including Swaziland which has the highest HIV prevalence in the world: 26% of year olds are HIV infected (95% CI: 5

28 ). In contrast, HIV prevalence among Western African adults is only 2%. HIV prevalence in Eastern Africa falls between levels in Western and Southern Africa with an average HIV prevalence across countries of 5% 1. There is also substantial variation in HIV burden within geographic regions and countries For example, HIV prevalence in Lake Victoria fishing communities in eastern Africa is between 2-4 times greater than the prevalence in neighboring agrarian communities 16. More broadly, communities along major international highways and trading routes tend to have higher HIV prevalence and levels of viral diversity 15, Figure 1 shows substantial variation in community HIV prevalence within the rural Rakai District in southwest Uganda with the highest disease burden being found in communities along the two main trading routes and Lake Victoria. Figure 1. Community HIV prevalence in the Rakai District, The red histogram shows the distribution of HIV prevalence across 50 Rakai communities, while the HIV point density map on the right shows that the highest levels of HIV infection are found along the main trading routes and Lake Victoria. The 2012 UNAIDS report on the global HIV/AIDS epidemic indicated that the number of newly HIV infected persons in sub-saharan Africa declined by 25% since This continent-wide trend obscures large variation in HIV prevention successes 6

29 between and within countries. This same report shows that in some countries in central and eastern Africa, including Uganda, incidence rates have stabilized 1. Disease burden within key populations, including commercial sex workers (CSW) and MSM, continues to rise or remains steady despite reductions in incidence within general populations The HIV epidemic in Uganda: past and present HIV spread rapidly throughout the Lake Victoria basin in the early 1980s and by the middle of that decade virus was detected in all Eastern African countries, including Uganda 5. The first cases of HIV were described by medical doctors in the Rakai District, a rural region in the southwest of Uganda consisting of small agrarian villages, fishing communities on Lake Victoria, and semi-urban trading centers 4. The Rakai doctors called the unknown illness slim disease for the characteristic wasting in what we now know to be the late stages of HIV infection, or acquired immunodeficiency syndrome (AIDS) 4. The Ugandan HIV epidemic was initially concentrated in rural communities. This is in contrast to neighboring epidemics in the Democratic Republic of Congo and Rwanda, where virus first emerged in the city centers 3,24. Phylogenetic analyses suggest that this initial concentration of infection among rural populations may have been driven by civil war in the preceding decade 19. Virus quickly spread from the rural lakeside communities in Uganda via trading routes to commercial centers and from there to rural inland communities 5. In Rakai, individuals residing in the main commercial trading centers and intermediate trading villages had 8.4 (95%CI: ) and 3.8 ( ) times the odds of being HIV-infected, respectively, than persons living in the more isolated agrarian communities inland from the lake 15,25. The Ugandan HIV epidemic peaked in 1991 with 21.1% of women in antenatal clinics testing HIV seropositive nationwide 26. 7

30 Approximately 90% of men and women in Uganda by 1995 had known someone who died from AIDS 26. With more than 1 million persons HIV infected, the Ugandan government initiated a comprehensive response to the epidemic including communitybased reporting of HIV-related deaths and prevention messages focused on abstinence, faithfulness within marriage, and to a lesser extent condom use with non-marital partners (ABC prevention) Rapid declines in high risk sexual behaviors were observed over the latter half of the decade, and by the turn of the century HIV prevalence declined markedly across all age groups, though the reasons for these declines are debated (Figure 2) Figure 2. HIV prevalence in Uganda, A) HIV prevalence in Kampala by age group and urban /rural settings vs. elsewhere in Uganda. B) Declines in HIV prevalence were not observed in other sub-saharan African countries over same time period (taken from Stoneburner et al, Science, 2004). 8

31 In 2004, antiretroviral therapies (ART) were introduced into Uganda through the United States President s Emergency Plan for Aids Relief (PEPFAR) and the Global Fund for AIDS, Tuberculosis and Malaria 30. These medications have significantly prolonged the survival of HIV-infected persons, resulting in a modest increase in the Ugandan HIV prevalence following a decline in the early 2000s. In 2012, HIV prevalence in Uganda was 7.2%. The incidence of new HIV cases in Uganda is stabilized at ~1.0 incident cases per 100 person-years 1. Incidence rates hold steady despite relatively high levels of ART use among clinically-indicated patients (54% coverage in patients with CD4 counts <350 cells/mm 3 ) and circumcision among non-muslim adult males (34%) 1 compared to other countries in sub-saharan Africa. It is unclear whether this lack of decline in HIV incidence is because the epidemic in Uganda is mature (i.e. in an endemic state) or because of inadequate intervention An overview of the molecular biology of HIV and genomic organization Classification HIV belongs to the family of viruses Retroviridae, and within that family it is a member of the lentivirus genus. HIV is approximately µm in diameter and has a lipid envelope bilayer in which the HIV glycoprotein gp120 (env) is interspersed Within the viral envelope, there is a cylindrical capsid containing two copies of the viral genome and various regulatory proteins. The capsid is comprised of several thousand units of the viral protein, p

32 Replication The virus replicates by first binding to the CD4 protein on the surface of target cells via the envelope glycoprotein, gp Primary HIV target cells include CD4+ T- cells and macrophages, though the virus may infect other cell types 33. The virus is initially presented to target cells by dendritic cells in the peripheral tissues and plasma 35. Once the virus binds to the CD4 protein on the target cell, the gp120 and gp41 proteins undergo conformational changes permitting attachment to a second target cell coreceptor, either the CCR5 or CXCR4 chemokine receptors 34. Viruses that bind to CCR5 and CXCR4 co-receptors are referred to as R5 and X4 tropic viruses, respectively 36,37. This second attachment is also mediated by the gp120 protein, which through interaction with the co-receptor facilitates fusion of virus to the plasma membrane and then entry into the cell cytoplasm via gp41 36,37. After virus has entered into the cell, the singlestranded viral RNA is converted to double-stranded DNA by the HIV reverse transcriptase (RT) enzyme 33. This enzyme is highly error-prone (3 x 10-5 mutations/per replication cycle/per year) 38,39. Once viral DNA is formed by RT it is transported across the nuclear membrane and integrated into the host DNA by the viral integrase protein. After integration, viral genes are transcribed into messenger RNA through host RNA Polymerase II. The new viral RNA serves as genomic RNA for future HIV virions and as a template for new viral proteins 33. In the final steps, viral proteins and genomic RNA are transported to the cell membrane where immature viral particles are formed and then released from the cell through a budding process. Following release from the host cell, the protease enzyme 10

33 cleaves the structural proteins resulting in a mature virus particle. In the absence of this final step, the virus is not infectious 33, Genomic organization The HIV genome is ~10,000 base pairs and consists of ten genes encoding for structural, regulatory, and accessory proteins and non-coding regulatory structural elements 33. Figure 3 shows a map of the HIV-1 HXB2 reference genome, where rectangles correspond to open reading frames (of which several overlap). The gag gene, one of the three genes found in all lentiviruses, encodes several key structural proteins including the matrix (p17), capsid (p24), nucleocapsid (p7), and p6 proteins 33,41. The p6 protein is unique to primate lentiviruses; however, its function is unknown. The precursor of the gag structural proteins is the p55 protein. This protein is essential for assembly of progeny virus at the cell membrane prior to budding. Once the viral particle has been released from the cell, the p55 protein is proteolytically cleaved forming the structural proteins that ultimately render viral particles mature and infectious 33,41. The pol gene encodes the protease, reverse transcriptase and integrase enzymes. These genes are initially encoded as a gag/pol poly-protein precursor which is proteolytically cleaved to form p55 and the pol enzymes. The envelope gene encodes the gp160 glycoprotein. This large protein is eventually processed forming the gp120 and gp41 envelope proteins, critical for attachment and entry into target cells. Like gag, the pol and env genes are encoded in all lentiviruses

34 Figure 3. Gene map of the HXB2 HIV-1 reference genome. Numbers in the upper left and low right hand corners of the rectangles correspond to the gene start and end positions of the HXB2 reference genome, respectively (taken from Fields Virology, 5 th edition, 2007) The tat and rev gene products are important in regulation of HIV-1 gene expression in the nucleus 42,43, whereas the vif protein inhibits the host anti-viral protein APOBEC-3G thereby promoting viral infectivity 44. Functions of the vpr gene product (viral protein R) are not well known, though may include targeting of HIV DNA to the cell nucleus, arrest of cell growth and differentiation, and gene regulation. The vpu gene encodes the viral protein U which degrades CD4 in the endoplasmic reticulum and enhances releases of virions from the host cell membrane. The vpu gene is unique to group M HIV-1 33.The nef gene encodes a 27-kd protein with multiple accessory functions that enhance viral spread and maintain high viral load 45. The vpx gene also encodes an accessory protein, whose functions are not well known though they are believed to overlap with those of nef The natural history HIV infection Mechanisms of HIV transmission and infection HIV is transmitted from person to person through the exchange of blood, tissues, and body fluids. Infectious body fluids include the semen, vaginal fluids, and breast milk. 12

35 In sub-saharan Africa, the vast majority of HIV-infected persons acquire virus through sexual intercourse, though a substantial fraction of infections are also passed from HIVinfected mothers to their newborn children during child birth or through breast feeding following child birth 46. This dissertation focuses on those HIV-infections acquired through heterosexual sexual intercourse during adulthood (15-49 years of age). In ~80% of heterosexual transmissions, individuals are infected with a single viral variant at the time of HIV acquisition 47,48. This occurs despite high levels of intra-host viral diversity in the transmitting partner, and consequently is referred to as the HIV transmission bottleneck 48. During the first week of infection, the founding virus establishes infection and replicates in a small foci of resting CD4+ T-cells in the mucosal epithelium before invading the primary and secondary lymphoid tissues 49. Data from stable heterosexual couples in the Rakai Community Cohort Study (RCCS) suggest that the founder virus may be evolutionarily conserved across hosts, such that the initially infecting virus in an index partner is the same viral variant that is transmitted to a subsequent partner 50. Viral co-infection occurs when individuals become HIV-infected with multiple, genetically distinct viruses at the time of transmission. Individuals may also acquire additional viral infections 51. These subsequent new acquisitions are referred to as HIV superinfection events. Cohort analyses in Uganda and Kenya suggest that the incidence of HIV superinfection occurs at the same or lower rate than initial HIV infection 52,53. 13

36 Early stage infection In the earliest stages of HIV-infection, virus spreads from the initial site of infection in the mucosal epithelium to the primary and secondary lymphoid tissues 49. It is in the lymphoid tissues where the viral population rapidly expands, doubling in size every 10 hours over a period of approximately three weeks (21 days) 54. The high rate of viral replication during the early phase of HIV infection results in an exponential increase in HIV viral load, a measure of the density of viral particles in the peripheral blood 55. The period of rapid viral expansion that precedes detectable antibody responses to HIV is specifically denoted as the acute phase of HIV-infection (AHI). Early HIV-infection (EHI), which includes AHI, is more broadly characterized by rapid viral replication and diversification, the early and widespread depletion of the T-cell reservoirs in the lymphoid tissues, and the onset of the adaptive immune response to HIV-infection (Figure 4) 54,56. Figure 4. The natural history of HIV-1 infection in the first 1000 days of infection. The natural history of HIV-1 can be divided into a series of stages characterized by the absence (-) and presence (+) of viral antigens and HIV-specific immune responses in the peripheral blood (taken from Cohen et al., NEJM, 2011) 14

37 The initial B-cell mediated antibody response to HIV infection begins at the time or immediately following peak viral load. These responses are non-neutralizing, and they do not mediate antibody-dependent cell cytotoxicity 57,58. In contrast, the CD8+ T-cell responses mount strong anti-viral responses to the founder virus several days prior to peak viral load. The CD8+ T-cell responses are critical for control of early HIV-infection; however, it is these same responses that drive rapid evolution of the founder virus, resulting in genetically distinct viral populations that are not susceptible to the adaptive immune defenses 54. Between 50-80% HIV-infected patients experience clinical symptoms during the early period of HIV-infection 46. Symptoms last for approximately seven days and are non-specific to HIV. The most commonly reported symptoms include fever, lymphadenopathy, pharyngitis, rash, myalgia, diarrhea, genital ulcer disease, and headache. During early HIV-infection, the CD4+ T-cell count declines as viral load increases 46. More precipitous T-cell declines are associated with more severe symptoms during this period. Median viral load in the early stages of HIV-infection ranges between 10 6 and 10 7 RNA copies/ml Chronic stage infection Following peak viral load and the onset of the adaptive immune responses, HIV replication rates rapidly decelerate and viral load decreases until it reaches a steady setpoint value. The viral load remains relatively constant for a variable period of time thereafter, fluctuating between +/ copies of the viral load set-point After the early phase, CD4+ T-cell counts initially increase and then begin a slow and steady decline. This period of relatively low viral replication accompanied by slow CD4+ T-cell 15

38 decline is defined as the chronic stage of HIV-infection, and it is primarily asymptomatic 46. Median viral load set point is 30,000 copies/ml; however, this value is highly heterogeneous across persons 59. Higher viral load set-points are associated with more rapid disease progression 61, though intermediate values may maximize transmission potential 60. In an analysis of HIV-1 transmission in 112 HIV sero-discordant couples, Hollingsworth et al. showed that the viral load set point was heritable, such that transmitting couples shared similar viral load set points 62. This finding implies that the set point may be associated with evolutionary advantages in the transmitted founder virus Advanced stage In the absence of anti-retroviral therapy, CD4+ T-cells eventually decline to levels where the HIV-infected individual is longer able to control HIV replication or mount adequate immune responses to foreign pathogens and malignancies 46. As a consequence, HIV viral load rapidly increases and the individual becomes susceptible to a wide variety of opportunistic infections and HIV-related cancers, the latter of which are often viral in origin. One of the most common HIV-related cancers is Kaposi s sarcoma (KS), a tumor caused by the human herpes virus 8 (HHV-8). The advanced stage of HIV disease is designated as acquired immune deficiency syndrome (AIDS) and it is the inability of an HIV-infected person to respond to opportunistic infection and cancers during this period that ultimately results in death

39 2.5. The evolutionary dynamics of HIV within individuals and populations After HIV establishes productive infection in the lymphoid tissues, the immune system mounts B and T-cell mediated immunological responses to the virus 33. These adaptive host responses drive evolutionary changes in the virus 63. The rapidity with which HIV evolves in response to intra-host immunologic pressures is one of its defining features, and it is a consequence of three mechanisms: low fidelity of the viral RT and high rates of viral replication and recombination 33. In this latter mechanism, two distinct viruses must infect the same cell in order for a recombination event to occur. Recombination takes place during retroviral transcription of RNA to DNA through a copy-choice mechanism, occurring at multiple locations across the viral genome. The recombination rate of HIV is the highest of any known organism 33,64. Intra-host HIV evolution occurs at different rates across the viral genome, with the envelope gene undergoing the most selection, followed by the pol and gag genes 65. That the env gene is most variable is not surprising given that the envelope protein is on the surface of the viral particle and, hence, more frequently exposed to the antibody and cellmediated defenses. Analysis of serial samples from the same individuals suggests that intra-host HIV evolution occurs at a constant rate of ~1% per year giving rise to ladderlike HIV phylogenies with continual turnover of viral lineages and low diversity at any one point in time 66. When HIV sequences from different individuals are compared, a strikingly different phylogenetic picture is observed. In contrast, inter-host phylogenies are diffuse or starlike, showing little evidence of natural selection 67. This suggests that the processes that govern population-based evolution of virus are determined by non-selective host 17

40 dynamics rather than direct competition between different viruses. For instance, viruses with selectively advantageous mutations may arise later in infection when individuals are less likely to transmit virus or they are passed to individuals with low rates of partner exchange. There is some evidence that HIV-1 subtypes A and C may be more infectious than other viral subtypes 68,69, though it is unclear whether such viruses directly compete with one another on a scale or time frame sufficient to observe the evolutionary impact of competition in populations The molecular epidemiology of HIV in sub-saharan Africa HIV evolves rapidly within and among its human hosts, and it is this rapid evolution that has given rise to the vast heterogeneity in HIV genetic subtypes worldwide 63. Greatest HIV diversity is found in the Democratic Republic of Congo (DRC), where all known non-recombinant HIV-subtypes have been identified. It is here in western Africa where the virus likely originated, first as zoonotic infections in the chimpanzee and sooty-mangabey 70,71. Molecular dating analyses suggest that ancestral retroviruses first crossed from primates to humans at the turn of the twentieth century 72. HIV is genetically subdivided into two major groups: HIV-1 and HIV-2. HIV-1 shares a common genetic ancestor with the chimpanzee simian immunodeficiency virus (SIVcpz), whereas HIV-2 shares ancestry with the sooty-mangabey immunodeficiency virus (SIVsmm) 71. This genetic history reflects the contemporary geographic distribution of HIV-2, which predominately co-circulates in Western Africa with SIVsmm 70. Within the HIV-1 group, there are four viral clades (O, M, N, and P), including the main clade (Clade M) which accounts for over 95% of HIV-1 infections worldwide. Nine HIV-1 18

41 clade M subtypes (A-D, F-H, J, and K) and 45 circulating recombinant forms have been identified. Inter-subtype genetic difference between HIV-1 subtypes ranges from 15-20% in gag and 20-30% in env 33,71. HIV-1 subtypes A and D predominate in Uganda 73,74. The links between spatial distributions in HIV-1 genetic diversity and HIV incidence and prevalence in sub-saharan Africa are unknown. This is particularly true at the community-level, where HIV surveillance has been very limited. A detailed understanding of these relationships could provide critical insight into the mechanisms that sustain HIV-1 transmission within local sexual networks. Grenfell et al. argue that this connection is central to many applied issues, from the evolution of drug resistance and virulence, to vaccine design and the emergence of new diseases" Epidemiological insights into the transmission dynamics of HIV in sub-saharan Africa. There is great heterogeneity in community HIV prevalence and incidence throughout sub-saharan Africa, yet the mechanisms underlying inter-community disparities in HIV spread are poorly understood. Mathematical models offer insight into the patterns and predictors of HIV epidemic dynamics and the potential for HIV prevention interventions to interrupt HIV transmission chains. Models are also a logical framework within which to understand the mechanism through which factors promote or inhibit HIV spread in populations. For instance, the likelihood of an HIV epidemic occurring within a completely susceptible population can be quantified by the basic reproductive number (R 0 ), defined as follows 76 : R 0 = βcd Equation 1 19

42 R 0 is a function of biological and socio-behavioral factors, where β is the probability of HIV transmission per coital act, c is the contact rate per unit time, and D is the duration of HIV infection. Epidemics will occur only when R 0 is greater than 1. Thus, those factors that increase the probability of HIV transmission, contact rate or duration of disease will promote viral spread. By the same rationale, factors that decrease any one of these three factors will reduce the likelihood that an HIV epidemic occurs Probability of HIV transmission per coital act in heterosexual couples The probability that an individual acquires HIV during a given sexual act is determined by factors specific to the transmitting and recipient partners 77. It is arguable that the single most important of these factors is the viral load of the HIV-infected partner at the time of sexual intercourse. In a landmark study of HIV transmission in HIV serodiscordant heterosexual couples in the Rakai District, Quinn et al. showed that higher viral loads correlated with increased likelihood of HIV transmission 78. The Rakai study of viral load and HIV transmission motivated the concept of treatment as prevention (TasP), the hypothesis that anti-retroviral treatment (ART), which lowers HIV viral load, could be used to prevent HIV transmission 79. In the decade following the study by Quinn et al., a series of observational studies in the Rakai population and other study settings showed that ART reduced the probability of HIV transmission by ~90% in heterosexual HIV-discordant couples 80,81. These studies were followed by the landmark HPTN 052 clinical trial, which showed ART reduced the risk of HIV transmission in stable HIV-discordant couples by 96% 82. Scale-up of ART may also reduce HIV spread at a community-level. More recently, Tanser et al. showed that 20

43 scale-up of ART in HIV-infected persons with CD4 counts <350 cells/µl reduced the individual risk of HIV-infection in rural South African communities by up to 40% 83, though it is unclear what proportion of interrupted transmissions were household and community-based. This distinction is important if we wish to conclude that ART reduces inter-household spread of HIV. Two community-randomized TasP trials are now underway in Botswana 84 (Mochudi Prevention Project) and Zambia and South Africa (HPTN 071) 85. Other studies from Rakai showed that stage of HIV-infection in the transmitting partner was also an important determinant of transmission 86,87. Wawer et al. analyzed the rate of HIV transmission in 235 monogamous HIV-discordant couples and showed that transmission rates followed a U-shaped curve, with the highest rates of viral transmission in the earliest ( transmission/coital act; 95%CI: ) and advanced stages of disease ( transmission/coital act; 95%CI: ) and the lowest rate during chronic HIV-infection ( transmission/coital act; 95%CI: ). Though viral load is highly correlated with the natural history of viral infection, Wawer et al. found that that the early and late stages of HIV infection were associated with a higher HIV transmission rate independent of viral load and other potential cofactors of transmission, including self-reported genital ulcer disease (GUD) and age 88. The finding that early HIV transmission is associated with a higher rate of HIV transmission in stable discordant couples prompted an ongoing debate on the role of early stage transmission in sustaining the broader epidemic 89, which includes transmission on sexual networks both comprised of stable couples and individuals who engage in casual or shorter-term sexual partnerships. Knowing the attributable fraction of total HIV 21

44 transmissions that are propagated by the earliest stages of disease is important because the effectiveness of HIV prevention interventions such as TasP may be limited in scenarios where early stage transmission accounts for a greater proportion of incident cases. This is because prevention programs would have limited time to identify, test and treat someone before they transmitted infection to their sexual partners. Many groups have attempted to estimate the proportion of HIV-infections attributable to early stage transmission; however, estimates from these studies are highly variable, ranging from 8% to 75% of total transmissions 89. Factors specific to the HIV-negative partner may also increase their susceptibility to HIV acquisition. Studies in women have shown that pregnancy and use of injectable contraceptives significantly increase female risk for HIV acquisition 90,91. In a systematic review of HIV transmission co-factors, Powers et al. estimated genital ulcer disease in HIV-uninfected partners accounted for an additional 6.0 transmissions per 1000 coital acts (95% CI: ) 77. Other sexually transmitted infections (STI) may also increase susceptibility to HIV-1 infection, including human papilloma virus (HPV). An analysis of HIV transmission in heterosexual men with human papilloma virus suggested that the clearance of HPV promotes an inflammatory response in the genital tract that augments HIV risk 92. Treatment of STI for HIV prevention has been tested in several settings. Two large randomized community trials of STI treatment for HIV prevention, one conducted in Rakai, Uganda and the other in Mwanza, Tanzania provided discordant results 93,94. Wawer et al. found treatment of bacterial STI did not reduce HIV transmission within Rakai communities, where in contrast investigators in Mwanza showed that population- 22

45 based treatment of STI reduced HIV incidence. The reasons for the failure of STI treatment in Rakai and its success in Mwanza are unknown, though hypotheses include differences in treatments used, stage of epidemic, and population mobility 95,96. Another clinical trial of acyclovir treatment of HSV for HIV prevention failed to demonstrate protective effects despite increased risk of HIV infection from HSV 97. Other public health interventions have been shown to reduce the probability of HIV transmission. Three large randomized clinical trials of medical male circumcision (MMC) in Southern and Eastern Africa showed that MMC reduces HIV incidence by 60% , including one trial in Rakai, Uganda. Additionally, male condom use also substantially reduces the likelihood of HIV transmission when condoms are used consistently (~80%) 101 ; however, use of condoms in African settings is infrequent, particularly within stable heterosexual partnerships 102. While treatment for HIV prevention has proved highly efficacious in clinical trial and observational settings, studies of oral ART for HIV prevention in HIV-negative partners (PreP) have met with mixed results in African populations 103,104. Analyses of ART levels in the blood of study participants in two PreP trials conducted in sub-saharan Africa indicated that a lack of efficacy was likely due to poor adherence to the study drug regimen. Other interventions include vaginal microbicides and HIV vaccines , both of which have met with variable success in clinical trials, including two studies in which the interventions may have increased HIV risk 105,

46 Sexual contact networks and HIV spread in sub-saharan Africa The dynamics of the sexual contact network is a critical determinant of the potential for HIV spread in a population, though empirical studies of network structures and properties are uncommon, particularly in sub-saharan Africa 108. This is partly because network analyses rely on highly sensitive information that is in most cases is self-reported. Individuals may be reluctant to disclose sexual partners, particularly if the reporting process serves as a source of shame or may cause harm to that person. Moreover, individuals may not recall the exact number of sexual partners that they have had or the details of those partnerships 108,109. Network models of HIV epidemics are also complex relative to compartmental models, the latter of which assumes random mixing between partners within compartments. Like all epidemiological studies, the accurate reconstruction and study of a sexual network is subjective to sampling and measurement biases. The degree and direction of bias in a network study will depend on the study design, the quality of data that is obtained, and the analysis plan 108,109. Network data is often obtained through egocentric studies in which research participants are first asked the number of sexual partners that they have had in the past year and then to disclose the characteristics of the those partners (i.e. age, occupation) and details of the partnership (i.e. duration, coital frequency, condom use). Egocentric designs do not typically obtain the names of these partners, and so empirical reconstruction of sexual networks from these data is not possible 108,110. Egocentric data can be used to determine degree distributions of networks, sexual mixing between sociodemographic subgroups and risk associated with particular partner types. In contrast, sociometric study designs obtain the names of the sexual partners for all members of the 24

47 study population. In these studies, the positions of individuals within a network are more readily determined. There has been only one sociometric study of an HIV transmission network in sub-saharan Africa. This study was a cross-sectional study of sexual partnerships among year olds in Likoma Island, Malawi 111. Researchers found that the degree distribution in Likoma was highly skewed: there were a large number of smaller sub-components (86% of subcomponents contained 5 or fewer persons) and fewer, but much larger sub-components that contained a significant proportion of sexually active adults. One giant component was identified containing 45.6% and 56% of female and male survey respondents, respectively. Surprisingly, higher HIV prevalence was found in the smaller sub-components. Individuals in the more isolated regions of the network were more likely to be widowed, older, and have partners outside of the island 108. Sociometric studies like the Likoma study are rare, most likely because of the resource and logistical constraints that are associated with their conduct. Moreover, simply providing the name of a sexual partner may not be sufficient to uniquely identify individuals within a population. Likoma Island also represents an unusual scenario in which the study population was isolated from other communities and transportation networks. In most settings, defining a community or geographic network of interest may not be as straightforward. When all individuals in a network cannot be sampled, other study designs can provide information about network structure and increase the probability that the high degree or less frequent nodes within a network are sampled. Snowball and respondent driven sampling (RDS) include two such approaches in which a chain referral sampling 25

48 method is used to recruit study participants. In contrast to snow-ball sampling, RDS adjusts for the non-random nature of the sampling process 108,109,112. The simplest way to characterize a sexual network is by the degree distribution, which is the distribution of the number of sexual partners of each person in the network. The degree for a given individual is a measure of that person s centrality within the network. Other measures of network centrality include closeness, betweeness and Eigen vector centrality 110. These additional measures capture the position and influence of individuals within the network both of which do not necessarily scale with degree. Of these measures, the degree is the most common measure of network centrality and is defined over a specified period of time, often over a one year period or a participant s lifetime. The contact parameter in Equation 1 is a function of the mean (m) and variance (σ) in degree 76. c = m + σ2 m Equation 2 Equation 2 implies that few individuals with many sexual partners disproportionately contribute to epidemic spread. This concept serves as motivation for the core group hypothesis wherein a highly infected subset of the population (i.e. commercial sex workers) serves as the primary source of HIV to the broader sexual network 113. An empirical analysis of a Swedish sexual network showed that network was scale-free, challenging the core group hypothesis for which the expected degree distribution would be bimodal 114. In contrast, the degree distribution of a scale-free network follows a power law distribution in which there are also individuals with intermediate numbers of partners. Thus, in the case of scale-free networks the core-group 26

49 is part of a continuum of sexual behaviors rather than a distinct subgroup in the population. In generalized HIV epidemics, scale-free networks are highly susceptible to the removal of high degree nodes where interventions that target individuals with higher partner degree will interrupt HIV transmission 114,115. The existence of the power law distribution also has important implications for the likelihood that HIV will persist within a network. Specifically, when the power-law exponent falls between the values of two and three, there is no critical threshold for disease spread other than when virus is completely non-infectious. This means that HIV can persist within a network at infinitesimally low transmission probabilities 114, implying interventions which reduce but do not eliminate infectiousness may control but not eradicate disease spread. There have been only several network analyses of sexual partner degree distributions in sub-saharan Africa. Analyses of sexual networks in Burkina Faso and Zimbabwe found that degree-distributions in those settings were scale-free 115,116. Handcok and Jones analyzed degree distributions by gender in the Rakai, District 117. In this analysis, the male degree exhibited scale-free properties though that for the females did not. It is unclear how such gender differences affect epidemic dynamics. A critical limitation of the Handcok and Jones an analysis was it was restricted to data from men and women in stable sexual partnerships. These partnerships represent a biased sample of the Rakai sexual network. Moreover, none of these three network analyses in sub- Saharan Africa (Burkina Faso, Zimbabwe, or Uganda) accounted for potential reporting biases in the numbers of sexual partners of men and women. 27

50 There are important sexual network properties other than the degree distribution that affect HIV epidemic dynamics. Sexual mixing patterns among individuals also limit or enhance epidemic potential 108. Assortative sexual mixing occurs when individuals select sexual partners who share similar characteristics with themselves, such as age, community, sexual behaviors, economic status, race, or occupation. In contrast, disassortative sexual mixing patterns arise when partner-selection is independent of sexual behavioral and socio-demographic factors 108. Disassortative mixing by age, sexual behaviors, and geographic location is associated with increased risk for HIVinfection Scale-free networks fall into a class of networks known as small world networks 121. In the case of small world networks, individuals are connected to one another through only a small number of sexual partners on average. The shorter the average path between any two randomly selected individuals in a network, the higher the clustering coefficient is for that network. Higher levels of community-based partner selection and other forms of assortative mixing enhance local network clustering. As the local clustering coefficient increases, HIV spreads more rapidly resulting in saturation of network subcomponents 115. Sexual migration of HIV-infected individuals (i.e. long distance sexual partnerships) and other forms of disassortative sexual mixing promote viral persistence within the global network Indeed population migration has been long associated with HIV epidemic spread in Africa 125,126. For example, the cyclic migration of miners from between rural communities and mining communities may sustain the HIV epidemic in KwaZulu-Natal, South Africa 127. Similarly, fisherman who work in high prevalence 28

51 communities along Lake Victoria and then return to their home villages may also help sustain HIV epidemics in rural East Africa. Sexual networks are often depicted as static entities; however, the dynamics of partnership formation, dissolution and timing are also critical to HIV spread. Morris and Kretzschmar used graph theory to show that as the number of individuals with overlapping (i.e. concurrent) sexual partnerships in a network increases so does HIV incidence 7. The association between concurrency and HIV incidence has been difficult to demonstrate through empirical studies of HIV transmission. This is because one s risk from concurrency may be through the sexual behaviors of their partner rather than through their own behaviors, implying that a network-based approach is warranted to study such a phenomenon. Tanser et al. studied the association between local levels of male concurrency and individual female HIV risk in KwaZulu-Natal, South Africa and found that sexual concurrency was not associated with incident HIV infection 128. Despite the absence of a network-based study design, the authors concluded that concurrency was not an important driver of HIV transmission in Africa. Moreover, a key assumption in this study was that HIV transmission is sustained through local sexual partnerships; however, this was not demonstrated by the authors despite high levels of HIV-risk related migration in the study population

52 2.8. References 1. UNAIDS. UNAIDS Gloabl Report on the AIDS Epidemic. 2012;. 2. Mann JM, Francis H, Quinn T, et al. Surveillance for AIDS in a central African city. Kinshasa, Zaire. JAMA 1986; 255(23): Mann JM, Nzilambi N, Piot P, et al. HIV infection and associated risk factors in female prostitutes in Kinshasa, Zaire. AIDS (London, England) 1988; 2(4): Serwadda D, Mugerwa RD, Sewankambo NK, et al. Slim disease: a new disease in Uganda and its association with HTLV-III infection. Lancet 1985; 2(8460): Iliffe J. The African AIDS Epidemic: A history. Athens, Ohio: Ohio University Press; Kilmarx PH. Global epidemiology of HIV. Curr Opin HIV AIDS 2009; 4(4): Morris M, Kretzschmar M. Concurrent partnerships and the spread of HIV. AIDS 1997; 11(5): Thoma M, Gray RH, Kiwanuka N, Serwadda D, Wawer M. Unsafe injections and transmission of HIV-1 in sub-saharan Africa. Lancet 2004; 363(9421): Baral S, Scheibe A, Sullivan P, Trapence G, Lambert A, Bekker LG, Beyrer C. Assessing priorities for combination HIV prevention research for men who have sex with men (MSM) in Africa. AIDS Behav 2013; 17 Suppl 1: S Beyrer C, Baral SD, van Griensven F, Goodreau SM, Chariyalertsak S, Wirtz AL, Brookmeyer R. Global epidemiology of HIV infection in men who have sex with men. Lancet 2012; 380(9839): Baral S, Trapence G, Motimedi F, Umar E, Iipinge S, Dausab F, Beyrer C. HIV prevalence, risks for HIV infection, and human rights among men who have sex with men (MSM) in Malawi, Namibia, and Botswana. PLoS One 2009; 4(3): e Johnston LG, Corceal S. Unexpectedly high injection drug use, HIV and hepatitis C prevalence among female sex workers in the Republic of Mauritius. AIDS Behav 2013; 17(2): Johnston LG, Holman A, Dahoma M, et al. HIV risk and the overlap of injecting drug use and high-risk sexual behaviours among men who have sex with men in Zanzibar (Unguja), Tanzania. Int J Drug Policy 2010; 21(6):

53 14. Chen L, Jha P, Stirling B, et al. Sexual risk factors for HIV infection in early and advanced HIV epidemics in sub-saharan Africa: systematic overview of 68 epidemiological studies. PLoS One 2007; 2(10): e Serwadda D, Wawer MJ, Musgrave SD, Sewankambo NK, Kaplan JE, Gray RH. HIV risk factors in three geographic strata of rural Rakai District, Uganda. AIDS 1992; 6(9): Seeley J, Nakiyingi-Miiro J, Kamali A, et al. High HIV incidence and sociobehavioral risk patterns in fishing communities on the shores of Lake Victoria, Uganda. Sex Transm Dis 2012; 39(6): Tanser F, Barnighausen T, Cooke GS, Newell ML. Localized spatial clustering of HIV infections in a widely disseminated rural South African epidemic. Int J Epidemiol 2009; 38(4): Arroyo MA, Sateren WB, Serwadda D, et al. Higher HIV-1 incidence and genetic complexity along main roads in Rakai District, Uganda. J Acquir Immune Defic Syndr 2006; 43(4): Gray RR, Tatem AJ, Lamers S, et al. Spatial phylodynamics of HIV-1 epidemic emergence in east Africa. AIDS 2009; 23(14): F9-F UNAIDS Reference Group on Estimates Modelling and Projections. Identifying populations at reatest risk of HIV infection. Geographic hotspots and key populations.. Geneva, Switzerland Needle R, Fu J, Beyrer C, et al. PEPFAR's evolving HIV prevention approaches for key populations--people who inject drugs, men who have sex with men, and sex workers: progress, challenges, and opportunities. J Acquir Immune Defic Syndr 2012; 60 Suppl 3: S Hankins C. Overview of the current state of the epidemic. Curr HIV/AIDS Rep 2013; 10(2): Beyrer C, Abdool Karim Q. The changing epidemiology of HIV in Curr Opin HIV AIDS 2013; 8(4): Allen S, Lindan C, Serufilira A, et al. Human immunodeficiency virus infection in urban Rwanda. Demographic and behavioral correlates in a representative sample of childbearing women. JAMA 1991; 266(12): Wawer MJ, Sewankambo NK, Berkley S, et al. Incidence of HIV-1 infection in a rural region of Uganda. BMJ 1994; 308(6922):

54 26. Stoneburner RL, Low-Beer D. Population-level HIV declines and behavioral risk avoidance in Uganda. Science 2004; 304(5671): Wawer, M. J., Gray, R. H., Serwadda, D., Namukwaya, Z., Makumbi, F., Sewankambo, N., Li, X., Lutalo, T., Nalugoda, F., and Quinn, T. Declines in HIV prevalence in Rakai, Uganda: Not as simple as ABC. Late Breaker Oral. 12th Conference on Retroviruses and Opportunistic Infections; Feb 22-25; Boston. 28. Murphy EM, Greene ME, Mihailovic A, Olupot-Olupot P. Was the "ABC" approach (abstinence, being faithful, using condoms) responsible for Uganda's decline in HIV? PLoS Med 2006; 3(9): e Green EC, Halperin DT, Nantulya V, Hogle JA. Uganda's HIV prevention success: the role of sexual behavior change and the national response. AIDS Behav 2006; 10(4): 335,46; discussion Reynolds SJ, Makumbi F, Nakigozi G, et al. HIV-1 transmission among HIV-1 discordant couples before and after the introduction of antiretroviral therapy. AIDS 2011; 25(4): Gallo RC, Sarin PS, Gelmann EP, et al. Isolation of human T-cell leukemia virus in acquired immune deficiency syndrome (AIDS). Science 1983; 220(4599): Barre-Sinoussi F, Chermann JC, Rey F, et al. Isolation of a T-lymphotropic retrovirus from a patient at risk for acquired immune deficiency syndrome (AIDS). Science 1983; 220(4599): Freed E.O., Martin MA. Human Immunodeficeincy Viruses: Replication. In: Knipe DM, Howley PM, editors. Fields Virology. 6th ed. Philadelphia, PA: Lippincott Williams & Wilkins. 34. Alkhatib G, Berger EA. HIV coreceptors: from discovery and designation to new paradigms and promise. Eur J Med Res 2007; 12(9): Geijtenbeek TB, van Kooyk Y. DC-SIGN: a novel HIV receptor on DCs that mediates HIV-1 transmission. Curr Top Microbiol Immunol 2003; 276: Berger EA, Murphy PM, Farber JM. Chemokine receptors as HIV-1 coreceptors: roles in viral entry, tropism, and disease. Annu Rev Immunol 1999; 17: Gorry PR, Ancuta P. Coreceptors and HIV-1 pathogenesis. Curr HIV/AIDS Rep 2011; 8(1): Roberts JD, Bebenek K, Kunkel TA. The accuracy of reverse transcriptase from HIV-1. Science 1988; 242(4882):

55 39. Preston BD, Poiesz BJ, Loeb LA. Fidelity of HIV-1 reverse transcriptase. Science 1988; 242(4882): Kohl NE, Emini EA, Schleif WA, et al. Active human immunodeficiency virus protease is required for viral infectivity. Proc Natl Acad Sci U S A 1988; 85(13): Freed EO. HIV-1 gag proteins: diverse functions in the virus life cycle. Virology 1998; 251(1): Malim MH, Bohnlein S, Hauber J, Cullen BR. Functional dissection of the HIV-1 Rev trans-activator--derivation of a trans-dominant repressor of Rev function. Cell 1989; 58(1): Rice AP, Mathews MB. Transcriptional but not translational regulation of HIV-1 by the tat gene product. Nature 1988; 332(6164): Marin M, Rose KM, Kozak SL, Kabat D. HIV-1 Vif protein binds the editing enzyme APOBEC3G and induces its degradation. Nat Med 2003; 9(11): Kirchhoff F, Greenough TC, Brettler DB, Sullivan JL, Desrosiers RC. Brief report: absence of intact nef sequences in a long-term survivor with nonprogressive HIV-1 infection. N Engl J Med 1995; 332(4): Lewthwaite P, Wilkins E. Natural history of HIV/AIDS. Medicine 2009; 37(7): Salazar-Gonzalez JF, Salazar MG, Keele BF, et al. Genetic identity, biological phenotype, and evolutionary pathways of transmitted/founder viruses in acute and early HIV-1 infection. J Exp Med 2009; 206(6): Derdeyn CA, Decker JM, Bibollet-Ruche F, et al. Envelope-constrained neutralization-sensitive HIV-1 after heterosexual transmission. Science 2004; 303(5666): Haase AT. Targeting early infection to prevent HIV-1 mucosal transmission. Nature 2010; 464(7286): Redd AD, Collinson-Streng AN, Chatziandreou N, et al. Previously transmitted HIV-1 strains are preferentially selected during subsequent sexual transmissions. J Infect Dis 2012; 206(9): Redd AD, Quinn TC, Tobian AA. Frequency and implications of HIV superinfection. Lancet Infect Dis 2013; 13(7):

56 52. Ronen K, McCoy CO, Matsen FA, et al. HIV-1 superinfection occurs less frequently than initial infection in a cohort of high-risk Kenyan women. PLoS Pathog 2013; 9(8): e Redd AD, Mullis CE, Serwadda D, et al. The rates of HIV superinfection and primary HIV incidence in a general population in Rakai, Uganda. J Infect Dis 2012; 206(2): Cohen MS, Shaw GM, McMichael AJ, Haynes BF. Acute HIV-1 Infection. N Engl J Med 2011; 364(20): Quinn TC, Wawer MJ, Sewankambo N, et al. Viral load and heterosexual transmission of human immunodeficiency virus type 1. Rakai Project Study Group. N Engl J Med 2000; 342(13): Powers KA, Ghani AC, Miller WC, et al. The role of acute and early HIV infection in the spread of HIV and implications for transmission prevention strategies in Lilongwe, Malawi: a modelling study. Lancet 2011; 378(9787): Tomaras GD, Haynes BF. HIV-1-specific antibody responses during acute and chronic HIV-1 infection. Curr Opin HIV AIDS 2009; 4(5): Tomaras GD, Yates NL, Liu P, et al. Initial B-cell responses to transmitted human immunodeficiency virus type 1: virion-binding immunoglobulin M (IgM) and IgG antibodies followed by plasma anti-gp41 antibodies with ineffective control of initial viremia. J Virol 2008; 82(24): de Wolf F, Spijkerman I, Schellekens PT, et al. AIDS prognosis based on HIV-1 RNA, CD4+ T-cell count and function: markers with reciprocal predictive value over time after seroconversion. AIDS 1997; 11(15): Fraser C, Hollingsworth TD, Chapman R, de Wolf F, Hanage WP. Variation in HIV-1 set-point viral load: epidemiological analysis and an evolutionary hypothesis. Proc Natl Acad Sci U S A 2007; 104(44): Mellors JW, Munoz A, Giorgi JV, et al. Plasma viral load and CD4+ lymphocytes as prognostic markers of HIV-1 infection. Ann Intern Med 1997; 126(12): Hollingsworth TD, Laeyendecker O, Shirreff G, et al. HIV-1 transmitting couples have similar viral load set-points in Rakai, Uganda. PLoS Pathog 2010; 6(5): e Rambaut A, Posada D, Crandall KA, Holmes EC. The causes and consequences of HIV evolution. Nat Rev Genet 2004; 5(1): Burke DS. Recombination in HIV: an important viral evolutionary strategy. Emerg Infect Dis 1997; 3(3):

57 65. Alizon S, Fraser C. Within-host and between-host evolutionary rates across the HIV-1 genome. Retrovirology 2013; 10: 49, Shankarappa R, Margolick JB, Gange SJ, et al. Consistent viral evolutionary changes associated with the progression of human immunodeficiency virus type 1 infection. J Virol 1999; 73(12): Lemey P, Rambaut A, Pybus OG. HIV evolutionary dynamics within and among hosts. AIDS Rev 2006; 8(3): Kiwanuka N, Laeyendecker O, Quinn TC, et al. HIV-1 subtypes and differences in heterosexual HIV transmission among HIV-discordant couples in Rakai, Uganda. AIDS 2009; 23(18): Silveira J, Santos AF, Martinez AM, et al. Heterosexual transmission of human immunodeficiency virus type 1 subtype C in southern Brazil. J Clin Virol 2012; 54(1): Hemelaar J, Gouws E, Ghys PD, Osmanov S. Global and regional distribution of HIV-1 genetic subtypes and recombinants in AIDS 2006; 20(16): W Hemelaar J. The origin and diversity of the HIV-1 pandemic. Trends Mol Med 2012; 18(3): Korber B, Muldoon M, Theiler J, et al. Timing the ancestor of the HIV-1 pandemic strains. Science 2000; 288(5472): Conroy S., Laeyendecker, O., Redd, A.D., Collinson-Streng, A., Kong, X., Makumbi, f., Lutalo, T., Sewankambo, N., Kiwanuka N., Gray, R., Wawer, M., Swerwadda D., Quinn, T.C. Changes in distribution of HIV-1 subtypes D and A in Rakai District, Uganda between 1994 and ;. 74. Collinson-Streng AN, Redd AD, Sewankambo NK, et al. Geographic HIV type 1 subtype distribution in Rakai district, Uganda. AIDS Res Hum Retroviruses 2009; 25(10): Grenfell BT, Pybus OG, Gog JR, Wood JL, Daly JM, Mumford JA, Holmes EC. Unifying the epidemiological and evolutionary dynamics of pathogens. Science 2004; 303(5656): May RM, Anderson RM. Transmission dynamics of HIV infection. Nature 1987; 326(6109): Powers KA, Poole C, Pettifor AE, Cohen MS. Rethinking the heterosexual infectivity of HIV-1: a systematic review and meta-analysis. The Lancet infectious diseases 2008; 8(9):

58 78. Quinn TC, Wawer MJ, Sewankambo N, et al. Viral load and heterosexual transmission of human immunodeficiency virus type 1. Rakai Project Study Group. The New England journal of medicine 2000; 342(13): Boily MC, Masse B, Alsallaq R, Padian NS, Eaton JW, Vesga JF, Hallett TB. HIV treatment as prevention: considerations in the design, conduct, and analysis of cluster randomized controlled trials of combination HIV prevention. PLoS Med 2012; 9(7): e Donnell D, Baeten JM, Kiarie J, et al. Heterosexual HIV-1 transmission after initiation of antiretroviral therapy: a prospective cohort analysis. Lancet 2010; 375(9731): Reynolds SJ, Makumbi F, Nakigozi G, et al. HIV-1 transmission among HIV-1 discordant couples before and after the introduction of antiretroviral therapy. AIDS 2011; 25(4): Cohen MS, Mastro TD, Cates W,Jr. Universal voluntary HIV testing and immediate antiretroviral therapy. Lancet 2009; 373(9669): 1077; author reply Tanser F, Barnighausen T, Grapsa E, Zaidi J, Newell ML. High coverage of ART associated with decline in risk of HIV acquisition in rural KwaZulu-Natal, South Africa. Science 2013; 339(6122): The Mochudi Prevention Project ART protocol. Available at: Population effects of antiretroviral therapy to reduce HIV transmission (POPART): a cluster-randomized trial of the impact of a combination prevention package on population-level HIV incidence in Zambia and South Africa. Available at: Gray RH, Wawer MJ, Brookmeyer R, et al. Probability of HIV-1 transmission per coital act in monogamous, heterosexual, HIV-1-discordant couples in Rakai, Uganda. Lancet 2001; 357(9263): Wawer MJ, Gray RH, Sewankambo NK, et al. Rates of HIV-1 transmission per coital act, by stage of HIV-1 infection, in Rakai, Uganda. The Journal of infectious diseases 2005; 191(9): Wawer MJ, Gray RH, Sewankambo NK, et al. Rates of HIV-1 transmission per coital act, by stage of HIV-1 infection, in Rakai, Uganda. J Infect Dis 2005; 191(9):

59 89. Cohen MS, Dye C, Fraser C, Miller WC, Powers KA, Williams BG. HIV treatment as prevention: debate and commentary--will early infection compromise treatment-asprevention strategies? PLoS Med 2012; 9(7): e Gray RH, Li X, Kigozi G, et al. Increased risk of incident HIV during pregnancy in Rakai, Uganda: a prospective study. Lancet 2005; 366(9492): Heffron R, Donnell D, Rees H, et al. Use of hormonal contraceptives and risk of HIV-1 transmission: a prospective cohort study. Lancet Infect Dis 2012; 12(1): Tobian AA, Grabowski MK, Kigozi G, et al. Human Papillomavirus Clearance Among Males Is Associated With HIV Acquisition and Increased Dendritic Cell Density in the Foreskin. J Infect Dis 2013; 207(11): Wawer MJ, Sewankambo NK, Serwadda D, et al. Control of sexually transmitted diseases for AIDS prevention in Uganda: a randomised community trial. Rakai Project Study Group. Lancet 1999; 353(9152): Grosskurth H, Mosha F, Todd J, et al. Impact of improved treatment of sexually transmitted diseases on HIV infection in rural Tanzania: randomised controlled trial. Lancet 1995; 346(8974): Gray RH, Wawer MJ. Reassessing the hypothesis on STI control for HIV prevention. Lancet 2008; 371(9630): Hayes R, Watson-Jones D, Celum C, van de Wijgert J, Wasserheit J. Treatment of sexually transmitted infections for HIV prevention: end of the road or new beginning? AIDS 2010; 24 Suppl 4: S Celum C, Wald A, Lingappa JR, et al. Acyclovir and transmission of HIV-1 from persons infected with HIV-1 and HSV-2. N Engl J Med 2010; 362(5): Bailey RC, Moses S, Parker CB, et al. Male circumcision for HIV prevention in young men in Kisumu, Kenya: a randomised controlled trial. Lancet 2007; 369(9562): Auvert B, Sobngwi-Tambekou J, Cutler E, Nieuwoudt M, Lissouba P, Puren A, Taljaard D. Effect of male circumcision on the prevalence of high-risk human papillomavirus in young men: results of a randomized controlled trial conducted in orange farm, South Africa. The Journal of infectious diseases 2009; 199(1): Gray RH, Kigozi G, Serwadda D, et al. Male circumcision for HIV prevention in men in Rakai, Uganda: a randomised trial. Lancet 2007; 369(9562): Weller S, Davis K. Condom effectiveness in reducing heterosexual HIV transmission. Cochrane Database Syst Rev 2002; (1)(1): CD

60 102. Lagarde E, Auvert B, Chege J, et al. Condom use and its association with HIV/sexually transmitted diseases in four urban communities of sub-saharan Africa. AIDS 2001; 15 Suppl 4: S Baeten JM, Donnell D, Ndase P, et al. Antiretroviral prophylaxis for HIV prevention in heterosexual men and women. N Engl J Med 2012; 367(5): Van Damme L, Corneli A, Ahmed K, et al. Preexposure prophylaxis for HIV infection among African women. N Engl J Med 2012; 367(5): Van Damme L, Ramjee G, Alary M, et al. Effectiveness of COL-1492, a nonoxynol-9 vaginal gel, on HIV-1 transmission in female sex workers: a randomised controlled trial. Lancet 2002; 360(9338): Buchbinder SP, Mehrotra DV, Duerr A, et al. Efficacy assessment of a cellmediated immunity HIV-1 vaccine (the Step Study): a double-blind, randomised, placebo-controlled, test-of-concept trial. Lancet 2008; 372(9653): Rerks-Ngarm S, Pitisuttithum P, Nitayaphan S, et al. Vaccination with ALVAC and AIDSVAX to prevent HIV-1 infection in Thailand. N Engl J Med 2009; 361(23): Doherty IA, Padian NS, Marlow C, Aral SO. Determinants and consequences of sexual networks as they affect the spread of sexually transmitted infections. J Infect Dis 2005; 191 Suppl 1: S Doherty IA. Sexual networks and sexually transmitted infections: innovations and findings. Curr Opin Infect Dis 2011; 24(1): Luke DA, Harris JK. Network analysis in public health: history, methods, and applications. Annu Rev Public Health 2007; 28: Helleringer S, Kohler HP. Sexual network structure and the spread of HIV in Africa: evidence from Likoma Island, Malawi. AIDS 2007; 21(17): Heckathorn DD. Respondent-driven sampling: a new approach to the study of hidden populations. Social problems 1997; 44(2): Thomas JC, Tucker MJ. The development and use of the concept of a sexually transmitted disease core. J Infect Dis 1996; 174 Suppl 2: S Liljeros F, Edling CR, Amaral LA, Stanley HE, Aberg Y. The web of human sexual contacts. Nature 2001; 411(6840):

61 115. Schneeberger A, Mercer CH, Gregson SA, et al. Scale-free networks and sexually transmitted diseases: a description of observed patterns of sexual contacts in Britain and Zimbabwe. Sex Transm Dis 2004; 31(6): Latora V, Nyamba A, Simpore J, Sylvette B, Diane S, Sylvere B, Musumeci S. Network of sexual contacts and sexually transmitted HIV infection in Burkina Faso. J Med Virol 2006; 78(6): Jones JH, Handcock MS. An assessment of preferential attachment as a mechanism for human sexual network formation. Proc Biol Sci 2003; 270(1520): Gregson S, Adamson S, Papaya S, et al. Impact and process evaluation of integrated community and clinic-based HIV-1 control: a cluster-randomised trial in eastern Zimbabwe. PLoS Med 2007; 4(3): e Boily MC, Lowndes CM, Gregson S. Population-level risk factors for HIV transmission and 'the 4 Cities Study': temporal dynamics and the significance of sexual mixing patterns. AIDS 2002; 16(15): Gregson S, Nyamukapa CA, Garnett GP, et al. Sexual mixing patterns and sexdifferentials in teenage exposure to HIV infection in rural Zimbabwe. Lancet 2002; 359(9321): Watts DJ, Strogatz SH. Collective dynamics of 'small-world' networks. Nature 1998; 393(6684): Read JM, Keeling MJ. Disease evolution on networks: the role of contact structure. Proc Biol Sci 2003; 270(1516): Messinger SM, Ostling A. The consequences of spatial structure for the evolution of pathogen transmission rate and virulence. Am Nat 2009; 174(4): Keeling MJ, Woolhouse ME, Shaw DJ, et al. Dynamics of the 2001 UK foot and mouth epidemic: stochastic dispersal in a heterogeneous landscape. Science 2001; 294(5543): Nunn AJ, Wagner HU, Kamali A, Kengeya-Kayondo JF, Mulder DW. Migration and HIV-1 seroprevalence in a rural Ugandan population. AIDS 1995; 9(5): Pison G, Le Guenno B, Lagarde E, Enel C, Seck C. Seasonal migration: a risk factor for HIV infection in rural Senegal. J Acquir Immune Defic Syndr 1993; 6(2): Lurie M, Harrison A, Wilkinson D, Karim SA. Circular migration and sexual netowrking in rural KwaZulu/Natal: implications for the spread of HIV and other sexually transmitted infections. Health Trans Rev 1997; 7(3):

62 128. Tanser F, Barnighausen T, Hund L, Garnett GP, McGrath N, Newell ML. Effect of concurrent sexual partnerships on rate of new HIV infections in a high-prevalence, rural South African population: a cohort study. Lancet 2011; 378(9787):

63 Chapter 3 Molecular tools for studying HIV transmission in sexual networks: A Review Mary K. Grabowski and Andrew D. Redd 3.1. Abstract Purpose of review: Phylogenetics is frequently used for studies of population-based HIV transmission. The purpose of this review is to highlight current utilities and limitations of phylogenetics in HIV epidemiological research from sample collection through data analysis. Recent findings: Studies of HIV phylogenies can provide critical information about HIV epidemics that are otherwise difficult to obtain trough traditional study design such as transmission of drug resistant virus, mixing between demographic groups, and rapidity of viral spread within populations. However, recent results from empirical and theoretical studies of HIV phylogenies challenge some of the key assumptions and interpretations from phylogenetic studies. Recent findings include lack of transmission bottlenecks in men who have sex with men and injection drug users, evidence for preferential transmission of HIV virus in heterosexual epidemics, and limited evidence that tree topologies correlate with underlying network structures. Other challenges include a lack of a standardized definition for a phylogenetic transmission cluster and biased or sparse sampling of HIV transmission networks. Summary: Phylogenetics is an important tool for HIV research, and offers opportunities to understand critical aspects of the HIV epidemic. Like all epidemiological research, the 41

64 methods used and interpretation of results from phylogenetic studies should be made cautiously with careful consideration Introduction One of the defining features of HIV is its ability to rapidly evolve and persist within individuals despite continued pressure from the cellular and humoral host immune responses 1. This struggle between virus and human has resulted in one of the most genetically diverse pandemics in recorded history 2. The diversity of HIV has been undoubtedly critical to its persistence and spread throughout the world; however, recent developments in genetic sequencing technologies, computational methodologies, and statistics are providing researchers with new tools to utilize viral diversity to combat the global HIV pandemic. Guided by population genetics and epidemiological principles 3, scientists are using viral phylogenetics to improve our understanding of HIV diversity within individuals and populations, generating an unprecedented knowledge of viral dynamics to improve strategies of HIV prevention and treatment of HIV-infected persons 4. This review examines the application of viral phylogenetics to HIV epidemiological research, with a focus on selected theoretical and practical considerations in phylogenetic studies of HIV transmission HIV sequence data for phylogenetic analyses HIV phylogenetics requires the generation of viral sequence. In most cases, these sequences are derived from virions isolated from the peripheral blood of infected persons with unknown duration of disease. However, sexual transmission of HIV, the most 42

65 common route of infection, involves the transmission of virus within the anogenital mucosa, with the highest probability of viral transmission occurring in the earlier and later stages of disease 5. In an analysis of HIV envelope sequences obtained from the serum and vaginal secretions of twelve women in a Kenyan cohort, Kemal et al. found phylogenetically related but distinct viral sub-populations within the blood and genital tract (GT) 6. Similarly, Boeras et al. found distinct, though genetically diverse, viral populations in the GT in chronically infected individuals in HIV serodiscordant relationships, but also showed that only a subset of these GT viruses are transmitted to sexual partners 7. This study and others of heterosexual (HET) HIV transmission further suggested certain GT viruses are preferentially transmitted to sexual partners 8,9, and that these newly acquired founder viruses may be selectively sequestered in the GT of HIVinfected individuals for subsequent transmission 7,10,11. The extent to which intra-host HIV evolutionary processes affect inference from viral phylogenies is unknown. Volz et al.argued that intra-host selection process have limited impact on inferences from viral phylogenies for three reasons: (1) there is a viral bottleneck at the time of transmission, (2) most infections are transmitted during the early stages of infection, and (3) transmission of virus is independent of viral factors 12. While there is strong evidence to support a transmission bottleneck during heterosexual intercourse, studies suggest high multiplicity of HIV infection in MSM and IDUassociated transmissions 13,14. The role of early stage transmission in HIV epidemics is widely debated 15, but intra-host HIV evolution occurs within only weeks following acquisition and the onset of the adaptive immunologic responses 16. Phylogenetic studies of HIV epidemics in MSM have suggested early infection sustains MSM epidemics ; 43

66 however, this early period may not have as substantial of a role in IDU or HET epidemics 20. Lastly, given data supporting the preferential transmission of HIV and association of viral factors with established viral infection 9,11,21, it seems unlikely that the last of the three arguments would hold for HIV phylogenies Viral sequencing technologies Once HIV has been isolated from the blood or other body fluids and then amplified through a single or series of polymerase chain reactions, a viral sequence(s) can be obtained from the resulting viral amplicon. HIV sequencing methodologies include direct Sanger sequencing 22, single genome amplification or cloning (SGA/cloning) 23, and next-generation sequencing (NGS) 24. The preponderance of historical sequences currently available in large HIV sequence databases (i.e. Los Alamos HIV database) were generated using Sanger sequencing of a single gene region. Sanger sequencing results is a single consensus sequence, or bulk sequence, from the post-pcr viral amplicon in which each base in the sequence is the most frequent base at that position across all unique sequences within the amplicon 22. In contrast, SGA/cloning and NGS yield multiple distinct viral sequences, which allows one to quantify and examine viral diversity within samples. The level of viral diversity that can be measured is directly related to the genetic length and depth of sequencing that is obtained 25. In the case of SGS/cloning this is usually done at a level of a few up to multiple dozens of longer sequences from one sample. NGS sequencing strategies are significantly more robust compared to SGA and can generate tens of thousands of shorter sequences from a single sample. However, certain NGS technologies can be error-prone within homopolymer stretches of DNA 26, 44

67 and one of the major challenges in using NGS has been to distinguish systematic and random technical error from true viral diversity Due to these inherent limitations, as well as differences in cost and ease of data analysis, sequencing strategies have different applications (Table 1). One aspect of HIV virology that is particularly important for consideration in sequencing studies of concentrated HIV epidemics, where partner turnover is more frequent and hence reinfection with virus more common, is viral coinfection or superinfection with multiple genetically distinct HIV viruses (reviewed by Redd and Tobian 30 ). When individuals are infected with multiple viruses either at the time of initial infection or later, a bulk sequence will not adequately represent all viruses that an individual has acquired or possibly transmitted 31,32. In these instances SGA and NGS methods can provide a more detailed depiction of intra-host HIV diversity Phylogenies reconstructed from bulk sequences may be subject to greater bias with increasing prevalence of HIV coinfection, though this remains an understudied area in phylogenetics Reconstructing a phylogenetic tree HIV phylogenies are evolutionary trees in which the leaves of the tree are the sampled sequences or taxa, branches are the genetic distance between taxa, and the nodes denote estimated speciation events 36. Reconstruction of HIV phylogenies can be accomplished through a variety of methodological approaches including neighbor joining (NJ), maximum likelihood, and Bayesian methods. These approaches and their strengths and limitations are reviewed in detail by Yang 37. Briefly, NJ methods employ a 45

68 hierarchical cluster algorithm wherein aligned sequences are grouped together based on their genetic similarity 38,39. The NJ algorithm is the most computationally efficient of tree reconstruction approaches in part because the clustering algorithm does not need to search the full tree space, which can be extremely large for datasets with many taxa. Maximum likelihood and Bayesian approaches are relatively more complex because the searched tree-space is often more extensive and users are required a priori to specify model assumptions, including a model of nucleotide substitution (i.e. model of HIV evolution) 36. Despite more limited computational efficiency, maximum likelihood and Bayesian approaches remain popular because user-defined evolutionary models can result in more accurate phylogenetic trees when correctly specified, assumptions are explicit, and model specification is flexible 37. Software has been developed to help analysts determine which evolutionary model is the best for their dataset 40,41. In a maximum likelihood analysis, the phylogenetic tree that maximizes the likelihood among all possible trees given the sequence data is the maximum likelihood, or best phylogenetic tree 42. In comparison, posterior probabilities are estimated in Bayesian analyses, where the posterior is a function of the prior probability of the tree and the tree likelihood 36. Interpretation of Bayesian posterior probabilities are more straightforward than bootstrap values in maximum likelihood analyses; however, node confidence can be higher than expected using Bayesian approaches 43,44. Formal comparison between Bayesian posterior probabilities and ML bootstrap values reveals that posterior probabilities are almost always higher than corresponding bootstrap values 43,44. Thus, comparisons of clade support or phylogenetic criteria incorporating 46

69 clade support values (i.e definitions of HIV transmission clusters) across maximum likelihood and Bayesian analyses should be made with caution. Bayesian trees were first constructed using a molecular clock assuming a constant rate of evolution across tree branches 37. This assumption was then relaxed to allow for rate variation across viral linages and for clock-free analyses 45. MrBayes and BEAST are two software packages frequently used in Bayesian phylogenetics 46,47, though the latter of these packages is preferred for the estimation of time-scaled HIV phylogenies. Both packages utilize Markov Chain Monte Carlo (MCMC) algorithms 48,49. The BEAST software is often used for HIV phylodynamic studies, including the estimation of temporal and geographical epidemic dynamics 47. Critical to the validity of phylodynamic studies is representative sampling over the space and time for which inference is to be made 50,51. A detailed review of viral phylodynamics and its applications can be found elsewhere Viral linkage analysis Viral linkage is the process of using viral sequence data and phylogenetics to determine directionality of HIV transmission within groups of individuals that have an underlying epidemiological linkage and one of the members is presumed to be the source of HIV-infection to the other member(s), such as concordant HIV-positive couples, known sexual partners, or needle-sharing groups 31, The objective of linkage analysis is to confirm or rule out a specific sexual partner or contact as the infection source; therefore, the criteria used to establish phylogenetic associations should be conservative, 47

70 with an emphasis on excluding (specificity) rather than confirming (sensitivity) infection sources. Viral linkage analysis was recently applied in two large HIV clinical trials of acyclovir (Partners in Prevention) and HIV antiretroviral therapy (HPTN-052) for HIV prevention among HIV serodiscordant couples 31,54. Criteria for determining whether transmission occurred from the index HIV-positive to the initially HIV-negative partner differed but were highly conservative in both studies. Strict criteria were used so that viral sources of infection were not erroneously attributed to the index HIV infected partner or the benefits of the interventions overestimated. In the Partners in Prevention study (PIP) a newly HIV-seroconcordant couple (n=151) was considered virologically linked if (1) gag and/or env sequences formed a monophyletic cluster in a maximum likelihood tree and (2) the genetic distance between partner viruses exceeded a probability threshold (50%) using a Bayesian algorithm that was developed by the study investigators. Using these methods the PIP investigators determined that 26.5% of newly acquired infections were obtained from sources other than the known index HIV-infected partner. To determine viral linkage of 38 newly HIV seroconcordant couples in HPTN- 052 clinical trial, Eshelman et al. also used a similar approach to that applied in the PIP study though they analyzed the pol gene and performed 454 pyrosequencing of the gp41 on all couples that could not be initially linked using consensus sequences. In comparison, the investigators found 18.4% of newly HIV-concordant couples in HPTN- 052 were virologically unlinked

71 3.7. HIV transmission patterns from phylogenetics One of the fasting growing applications of HIV phylogenetics is to the study of population- based patterns of HIV transmission. In comparison to viral linkage analyses, the unit of analysis in these studies is the transmission network wherein direct and indirect phylogenetic links between individuals are both of primary interest. In the absence of a fully sampled transmission network and detailed contact histories, direct linkage between two individuals in a phylogenetic tree is not possible 55. Instead, phylogenetics is used to identify groups of persons who share an HIV transmission chain with unknown directionality and contact structure. Identified chains may then be characterized to inform HIV prevention efforts. Epidemic trajectories and key transmission parameters (reproductive number, generation times 56 ) also inform HIV control and these too can be estimated using phylogenetic methods. Broadly speaking, individuals whose viral sequences are closely related because they have low genetic distance and share a most recent common ancestor constitute a phylogenetic cluster, or putative, recent HIV transmission chain. Definitions of phylogenetic clusters vary widely in the HIV literature, though many studies use both clade support and genetic distance cutoffs (Table 2) Rationale for cutoffs is rarely provided; however, these decisions may affect study inferences. For example, Robinson et al. simulated HIV spread and pathogen phylogenies on two different network topologies and showed that cutoffs choices affect the size and distribution of phylogenetic clusters obtained

72 The relationship between phylogenetic tree topologies and the structure of the underlying contact network on which those topologies are generated is an ongoing and critical area in HIV phylogenetic research. Interpretation of phylogenetic cluster distribution as indicators of contact network phenomena is complicated, in part because of the absence of meaningful statistical comparisons between empirical and null cluster distributions. More recently, tree balance statistics have been developed and used to compare phylogenetic clustering patterns against null contact models of random mixing 78. However, Robinson et al. found that cluster distributions and tree balance statistics revealed little of the global network structure from simulation, particularly when networks were dynamic 77. Frost and Volz reached similar conclusions, showing tree shape rarely corresponded to population structure 79. While determining global network structure from HIV phylogenies may be challenging, identifying individuals with related HIV viruses can be done readily from phylogenies and is often very informative. When combined with detailed epidemiological and clinical data, phylogenetic analyses can reveal information of public health relevance including transmission of drug resistant virus and mixing across transmission, demographic and behavioral subgroups 4. The majority of population-based HIV phylogenetic studies have been conducted outside of Africa, where viral transmission is typically concentrated within high risk populations such as commercial sex workers, MSM and IDUs 80. Lewis et al. analyzed over 14,000 viral pol sequences collected from approximately 60% of HIV-infected men who have sex with men (MSM) in Great Britain 17. Using dated HIV phylogenies, the investigators found that a substantial fraction of MSM who phylogenetically clustered did so within large groups ( 10 persons) and over short time periods. In a more recent 50

73 analysis of this same data but using different methodological approaches, Volz et al. confirmed the importance of early HIV-infection in the UK MSM epidemic 18, and investigations of other MSM epidemics have yielded similar results 19. However, varying definitions of what constitutes early infection makes conclusions regarding its role in concentrated epidemics difficult from phylogenetics alone 81. The study of population-based HIV transmission patterns using phylogenetics is also complicated by biased and or incomplete sampling of transmission networks. Sequences are often obtained through convenience or clinic-based recruitment of participants and sampling fractions of the target network are rarely known or specified (Table 2). Moreover, HIV transmission networks are usually under sampled: the proportion of sequences that phylogenetically cluster rarely exceed 30% 4. While limited clustering may reflect inadequate sampling of target networks or some loss of tree resolution with intra-host viral evolution, high numbers of singleton lineages may also indicate important epidemiologic phenomenon driving HIV transmission in populations, such as viral introductions via migration or shorter term travel Conclusions Phylogenetics is a growing and exciting field in the HIV prevention sciences. Like all sciences, phylogenetics has limitations and when used for epidemiological purposes, it is subject to many of the same measurement and sampling biases of traditional epidemiological study designs. From a public health standpoint, HIV phylogenetics is most powerful when combined with detailed clinical and epidemiological data, in which case HIV phylogenies can reveal critical information 51

74 relevant to disease control, including transmission of drug resistant virus, associations between socio-demographic characteristics and viral spread within populations, and the time scales over which HIV epidemics occur. Additional theoretical studies relating HIV phylogenies to network structure and transmission processes are needed Keynotes -When using phylogenetics to analyze HIV epidemiological phenomenon, including transmission network structure or transmisson dynamics, it is essential that potential measurement and sampling biases are considered in the interpretation of study results. -HIV sequencing technology is rapidly changing, and it is critical to understand the advantages and limitations of the different sequencing technologies available prior to performing phylogenetic studies. -There is no consensus on phylogenetic criteria that should be used to establish a putative transmission cluster in population-based studies of HIV transmission networks. Appropriate phylogenetic criteria may depend on the underlying epidemiological and evolutionary dynamics in a given research setting. 52

75 Table 1. Overview of common sequencing technologies for HIV research Technology No. of sequences Advantages Limitations Utility Sanger bulk Single -Low labor -Cannot accurately detect multiple infections -Clinical resistance testing -Rapid turnaround -Medium sequence length -Large epidemiological studies utilizing phylogenetics -Easily analyzed -Limited ability to capture -Subtype screening intra-host viral diversity -Large historical record -Initial linkage analysis between individuals with known or suspected linkage -Lowest overall cost per sample Single genome sequencing (SGS)/Cloning Next generation sequencing (NGS) Technically simple -Viral diversity/ability to detect multiple infections is dependent on number of clones analyzed -More accurate base calls -Labor intensive and higher costs at high copy or sample numbers -Long sequence length -Moderate cost at low copy or sample number More accurately captures intra-host viral diversity including multiple infections -Highly sensitive to contamination -Viral diversity studies of large genetic regions -Full viral genome sequencing -Recombination studies -HIV superinfection and dual infection studies 53

76 -High-throughput -Lowest cost per nucleotide base sequenced -Requires specialized data cleaning and analysis -Expensive upfront equipment and reagent costs -Limited access globally -Epidemiological studies of populations at high-risk for multiple infections (IDU, high-incidence populations) -Detection of minor variants -NGS produces sequence reads that are consolidated to form consensus sequences which can total depending on sequencing efficiency 54

77 Table 2. Summary of participant recruitment and phylogenetic methods for a random sample of 20 phylogenetic studies of HIV-1 transmission clustering published in PubMed Central in 2013 Authors [Ref] Location Predominate Risk group Patient recruitment No. of Participants Method of tree reconstruction Antoniadou et al. 57 Greece HET, IDU, MSM Clinic 98 NJ Bezemer et al. 58 Kenya HET, MSM Convenience 674 ML, Bayesian Phylogenetic HIV transmission cluster definition 85% bootstrap, mean intra-cluster genetic distance 70% bootstrap, mean intra-cluster genetic distance Ruelle et al. 59 Belgium HET, MSM Clinic 55 ML, Bayesian 0.89 posterior probability Frentz et al. 60 Europe, Israel HET, IDU, MSM Clinic 4260 ML 98% bootstrap, mean intra-cluster genetic distance Dennis et al. 83 El Salvador FSW, MSM Respondent driven 119 Bayesian posterior probability=1, mean intra-cluster genetic distance Li et al. 62 China HET Clinic 253 ML 70% bootstrap Yebra et al. 63 Spain HET, IDU, MSM Clinic 1293 Bayesian 90 posterior probability, cluster depth cutoff a Ng et al. 64 Malaysia MSM Clinic 496 ML, Bayesian 2 individuals from same geographic location, >90% bootstrap, posterior probability=1 Feng et al. 65 China HET, IDU, MSM Clinic 75 ML 90% bootstrap Siljic et al. 66 Serbia HET, MSM Clinic 221 ML, Bayesian 90% bootstrap, mean intra-cluster genetic distance, 0.9 posterior probability Yebra et al. 67 Equatorial Guinea b HET Clinic 278 ML, Bayesian 95% bootstrap, 0.95 posterior probability Murillo et al. 68 Central America HET, FSW, MSM Clinic 625 ML S-H test (p-value <0.01) c, patristic distance threshold (25th percentile) d 55

78 Temereanca et al. 69 Romania HET, IDU Clinic 61 ML 0.01 maximum intracluster genetic distance Audelin et al. 70 Denmark HET, IDU, MSM Clinic 1515 NJ, Bayesian 90 % bootstrap, 0.025/ mean/maximum intra-cluster genetic distance, posterior probability=1 Chen et al. 71 China HET, IDU Clinic 308 Neighbor-joining 70% bootstrap Han et al. 72 China MSM Snowball, Clinic, Cross-sectional survey 583 Neighbor-joining 70% bootstrap 96 % bootstrap, 0.10 maximum intra-cluster genetic distance, 0.97 Ivanov et al. 73 Bulgaria HET, IDU, MSM Clinic 125 ML, Bayesian posterior probability Avidor et al. 74 Israel IDU, MSM Clinic 318 Bayesian 0.95 posterior probability Ndiaye et al. 75 Senegal MSM Snowball 109 ML 98% bootstrap Tramuto et al. 76 Sicily HET, MSM Clinic 155 ML 75% bootstrap HET=heterosexuals, IDU= Injection drug users; MSM=Men who have sex with Men, FSW=Female Sex Workers, NJ=Neighbor-joining, ML=Maximum likelihood a Defined as maximum length of time separating the ancestral node and the most recent tip in the cluster; 6 years for non-subtype B viruses and 8 years for subtype B viruses. b Study conducted among Equatorial Guineans attending Spanish clinics c Shimodaira-Hasegawa test d Inferred using PhyloPart software 56

79 3.9. References 1. Lemey P, Rambaut A, Pybus OG. HIV evolutionary dynamics within and among hosts. AIDS Rev 2006; 8(3): Hemelaar J, Gouws E, Ghys PD, Osmanov S. Global and regional distribution of HIV-1 genetic subtypes and recombinants in AIDS 2006; 20(16): W Grenfell BT, Pybus OG, Gog JR, Wood JL, Daly JM, Mumford JA, Holmes EC. Unifying the epidemiological and evolutionary dynamics of pathogens. Science 2004; 303(5656): Brenner B, Wainberg MA, Roger M. Phylogenetic inferences on HIV-1 transmission: implications for the design of prevention and treatment interventions. AIDS 2013; 27(7): Wawer MJ, Gray RH, Sewankambo NK, et al. Rates of HIV-1 transmission per coital act, by stage of HIV-1 infection, in Rakai, Uganda. J Infect Dis 2005; 191(9): Kemal KS, Foley B, Burger H, et al. HIV-1 in genital tract and plasma of women: compartmentalization of viral sequences, coreceptor usage, and glycosylation. Proc Natl Acad Sci U S A 2003; 100(22): Boeras DI, Hraber PT, Hurlston M, et al. Role of donor genital tract HIV-1 diversity in the transmission bottleneck. Proc Natl Acad Sci U S A 2011; 108(46): E Derdeyn CA, Decker JM, Bibollet-Ruche F, et al. Envelope-constrained neutralization-sensitive HIV-1 after heterosexual transmission. Science 2004; 303(5666): Hollingsworth TD, Laeyendecker O, Shirreff G, et al. HIV-1 transmitting couples have similar viral load set-points in Rakai, Uganda. PLoS Pathog 2010; 6(5): e Lythgoe KA, Fraser C. New insights into the evolutionary rate of HIV-1 at the within-host and epidemiological levels. Proc Biol Sci 2012; 279(1741): Redd AD, Collinson-Streng AN, Chatziandreou N, et al. Previously transmitted HIV-1 strains are preferentially selected during subsequent sexual transmissions. J Infect Dis 2012; 206(9): Volz EM, Koelle K, Bedford T. Viral phylodynamics. PLoS Comput Biol 2013; 9(3): e Li H, Bar KJ, Wang S, et al. High Multiplicity Infection by HIV-1 in Men Who Have Sex with Men. PLoS Pathog 2010; 6(5): e

80 14. Bar KJ, Li H, Chamberland A, et al. Wide variation in the multiplicity of HIV-1 infection among injection drug users. J Virol 2010; 84(12): Cohen MS, Dye C, Fraser C, Miller WC, Powers KA, Williams BG. HIV treatment as prevention: debate and commentary--will early infection compromise treatment-asprevention strategies? PLoS Med 2012; 9(7): e Wood N, Bhattacharya T, Keele BF, et al. HIV evolution in early infection: selection pressures, patterns of insertion and deletion, and the impact of APOBEC. PLoS Pathog 2009; 5(5): e Lewis F, Hughes GJ, Rambaut A, Pozniak A, Leigh Brown AJ. Episodic sexual transmission of HIV revealed by molecular phylodynamics. PLoS Med 2008; 5(3): e Volz EM, Koopman JS, Ward MJ, Brown AL, Frost SD. Simple epidemiological dynamics explain phylogenetic clustering of HIV from patients with recent infection. PLoS Comput Biol 2012; 8(6): e Brenner BG, Roger M, Routy JP, et al. High rates of forward transmission events after acute/early HIV-1 infection. J Infect Dis 2007; 195(7): Hughes GJ, Fearnhill E, Dunn D, Lycett SJ, Rambaut A, Leigh Brown AJ, UK HIV Drug Resistance Collaboration. Molecular phylodynamics of the heterosexual HIV epidemic in the United Kingdom. PLoS Pathog 2009; 5(9): e Keele BF, Giorgi EE, Salazar-Gonzalez JF, et al. Identification and characterization of transmitted and early founder virus envelopes in primary HIV-1 infection. Proc Natl Acad Sci U S A 2008; 105(21): Sanger F, Nicklen S, Coulson AR. DNA sequencing with chain-terminating inhibitors. Proc Natl Acad Sci U S A 1977; 74(12): Salazar-Gonzalez JF, Bailes E, Pham KT, et al. Deciphering human immunodeficiency virus type 1 transmission and early envelope diversification by singlegenome amplification and sequencing. J Virol 2008; 82(8): Metzker ML. Sequencing technologies - the next generation. Nat Rev Genet 2010; 11(1): Redd AD, Mullis CE, Serwadda D, et al. The rates of HIV superinfection and primary HIV incidence in a general population in Rakai, Uganda. J Infect Dis 2012; 206(2): Margulies M, Egholm M, Altman WE, et al. Genome sequencing in microfabricated high-density picolitre reactors. Nature 2005; 437(7057):

81 27. Zagordi O, Klein R, Daumer M, Beerenwinkel N. Error correction of nextgeneration sequencing data and reliable estimation of HIV quasispecies. Nucleic Acids Res 2010; 38(21): Beerenwinkel N, Zagordi O. Ultra-deep sequencing for the analysis of viral populations. Curr Opin Virol 2011; 1(5): Zagordi O, Geyrhofer L, Roth V, Beerenwinkel N. Deep sequencing of a genetically heterogeneous sample: local haplotype reconstruction and read error correction. J Comput Biol 2010; 17(3): Redd AD, Quinn TC, Tobian AA. Frequency and implications of HIV superinfection. Lancet Infect Dis 2013; 13(7): Eshleman SH, Hudelson SE, Redd AD, et al. Analysis of genetic linkage of HIV from couples enrolled in the HIV Prevention Trials Network 052 trial. J Infect Dis 2011; 204(12): Redd AD, Collinson-Streng A, Martens C, et al. Identification of HIV superinfection in seroconcordant couples in Rakai, Uganda, by use of next-generation deep sequencing. J Clin Microbiol 2011; 49(8): Kraft CS, Basu D, Hawkins PA, et al. Timing and source of subtype-c HIV-1 superinfection in the newly infected partner of Zambian couples with disparate viruses. Retrovirology 2012; 9: 22, Redd AD, Mullis CE, Serwadda D, et al. The rates of HIV superinfection and primary HIV incidence in a general population in Rakai, Uganda. J Infect Dis 2012; 206(2): Barlow KL, Green J, Clewley JP. Viral genome characterisation by the heteroduplex mobility and heteroduplex tracking assays. Reviews in medical virology 2000; 10(5): Lemey P, Salemi M, VanDamme A, editors. The phylogenetic handbook: a practical approach to phylogenetic analysis and hypothesis testing. Second ed. Cambridge; Yang Z, Rannala B. Molecular phylogenetics: principles and practice. Nat Rev Genet 2012; 13(5): Gascuel O, Steel M. Neighbor-joining revealed. Mol Biol Evol 2006; 23(11): Saitou N, Nei M. The neighbor-joining method: a new method for reconstructing phylogenetic trees. Mol Biol Evol 1987; 4(4):

82 40. Darriba D, Taboada GL, Doallo R, Posada D. jmodeltest 2: more models, new heuristics and parallel computing. Nat Methods 2012; 9(8): Tamura K, Peterson D, Peterson N, Stecher G, Nei M, Kumar S. MEGA5: molecular evolutionary genetics analysis using maximum likelihood, evolutionary distance, and maximum parsimony methods. Mol Biol Evol 2011; 28(10): Felsenstein J. Evolutionary trees from DNA sequences: a maximum likelihood approach. J Mol Evol 1981; 17(6): Alfaro ME, Zoller S, Lutzoni F. Bayes or bootstrap? A simulation study comparing the performance of Bayesian Markov chain Monte Carlo sampling and bootstrapping in assessing phylogenetic confidence. Mol Biol Evol 2003; 20(2): Erixon P, Svennblad B, Britton T, Oxelman B. Reliability of Bayesian posterior probabilities and bootstrap frequencies in phylogentics. Syst Biol 2003; 52(5): Drummond AJ, Ho SY, Phillips MJ, Rambaut A. Relaxed phylogenetics and dating with confidence. PLoS Biol 2006; 4(5): e Ronquist F, Teslenko M, van der Mark P, et al. MrBayes 3.2: efficient Bayesian phylogenetic inference and model choice across a large model space. Syst Biol 2012; 61(3): Drummond AJ, Suchard MA, Xie D, Rambaut A. Bayesian phylogenetics with BEAUti and the BEAST 1.7. Mol Biol Evol 2012; 29(8): Yang Z, Rannala B. Bayesian phylogenetic inference using DNA sequences: a Markov Chain Monte Carlo Method. Mol Biol Evol 1997; 14(7): Mau B, Newton MA, Larget B. Bayesian phylogenetic inference via Markov chain Monte Carlo methods. Biometrics 1999; 55(1): Stack JC, Welch JD, Ferrari MJ, Shapiro BU, Grenfell BT. Protocols for sampling viral sequences to study epidemic dynamics. J R Soc Interface 2010; 7(48): Lemey P, Rambaut A, Drummond AJ, Suchard MA. Bayesian phylogeography finds its roots. PLoS Comput Biol 2009; 5(9): e Nett RJ, Bartschi JL, Ellis GM, et al. Two clusters of HIV-1 infection, rural Idaho, USA, Emerg Infect Dis 2010; 16(11): Leitner T, Escanilla D, Franzen C, Uhlen M, Albert J. Accurate reconstruction of a known HIV-1 transmission history by phylogenetic tree analysis. Proc Natl Acad Sci U S A 1996; 93(20):

83 54. Campbell MS, Mullins JI, Hughes JP, et al. Viral linkage in HIV-1 seroconverters and their partners in an HIV-1 prevention clinical trial. PLoS One 2011; 6(3): e Resik S, Lemey P, Ping LH, et al. Limitations to contact tracing and phylogenetic analysis in establishing HIV type 1 transmission networks in Cuba. AIDS Res Hum Retroviruses 2007; 23(3): Stadler T, Kouyos R, von Wyl V, et al. Estimating the basic reproductive number from viral sequence data. Mol Biol Evol 2012; 29(1): Antoniadou ZA, Kousiappa I, Skoura L, et al. Molecular Epidemiology of HIV-1 Infection in Northern Greece ( ): Evidence of a Transmission Cluster of HIV-1 Subtype A1 Drug Resistant Strains among Men Having Sex with Men. AIDS Res Hum Retroviruses 2013;. 58. Bezemer D, Faria NR, Hassan A, et al. HIV Type 1 Transmission Networks Among Men Having Sex with Men and Heterosexuals in Kenya. AIDS Res Hum Retroviruses 2013;. 59. Ruelle J, Ingels MG, Jnaoui K, et al. Transmission Network of an HIV Type 1 Strain with K103N in Young Belgian Patients from Different Risk Groups. AIDS Res Hum Retroviruses 2013; 29(10): Frentz D, Wensing AM, Albert J, et al. Limited cross-border infections in patients newly diagnosed with HIV in Europe. Retrovirology 2013; 10: 36, Dennis AM, Murillo W, de Maria Hernandez F, et al. Social Network-Based Recruitment Successfully Reveals HIV-1 Transmission Networks Among High-Risk Individuals in El Salvador. J Acquir Immune Defic Syndr 2013; 63(1): Li L, Chen L, Liang S, et al. Subtype CRF01_AE dominate the sexually transmitted human immunodeficiency virus type 1 epidemic in Guangxi, China. J Med Virol 2013; 85(3): Yebra G, Holguin A, Pillay D, Hue S. Phylogenetic and demographic characterization of HIV-1 transmission in Madrid, Spain. Infect Genet Evol 2013; 14: Ng KT, Ong LY, Lim SH, Takebe Y, Kamarulzaman A, Tee KK. Evolutionary history of HIV-1 subtype B and CRF01_AE transmission clusters among men who have sex with men (MSM) in Kuala Lumpur, Malaysia. PLoS One 2013; 8(6): e Feng Y, He X, Hsi JH, et al. The rapidly expanding CRF01_AE epidemic in China is driven by multiple lineages of HIV-1 viruses introduced in the 1990s. AIDS 2013; 27(11):

84 66. Siljic M, Salemovic D, Jevtovic D, et al. Molecular typing of the local HIV-1 epidemic in Serbia. Infect Genet Evol 2013;. 67. Yebra G, de Mulder M, Holguin A. Description of HIV-1 group M molecular epidemiology and drug resistance prevalence in Equatorial Guinea from migrants in Spain. PLoS One 2013; 8(5): e Murillo W, Veras N, Prosperi M, et al. A single early introduction of HIV-1 subtype B into Central America accounts for most current cases. J Virol 2013; 87(13): Temereanca A, Ene L, Mehta S, Manolescu L, Duiculescu D, Ruta S. Transmitted HIV drug resistance in treatment-naive Romanian patients. J Med Virol 2013; 85(7): Audelin AM, Cowan SA, Obel N, Nielsen C, Jorgensen LB, Gerstoft J. Phylogenetics of the Danish HIV epidemic: the role of very late presenters in sustaining the epidemic. J Acquir Immune Defic Syndr 2013; 62(1): Chen M, Yang L, Ma Y, et al. Emerging variability in HIV-1 genetics among recently infected individuals in Yunnan, China. PLoS One 2013; 8(3): e Han X, An M, Zhang M, et al. Identification of 3 Distinct HIV-1 Founding Strains Responsible for Expanding Epidemic Among Men Who Have Sex With Men in 9 Chinese Cities. J Acquir Immune Defic Syndr 2013; 64(1): Ivanov IA, Beshkov D, Shankar A, et al. Detailed molecular epidemiologic characterization of HIV-1 infection in Bulgaria reveals broad diversity and evolving phylodynamics. PLoS One 2013; 8(3): e Avidor B, Turner D, Mor Z, et al. Transmission patterns of HIV-subtypes A/AE versus B: inferring risk-behavior trends and treatment-efficacy limitations from viral genotypic data obtained prior to and during antiretroviral therapy. PLoS One 2013; 8(3): e Ndiaye HD, Tchiakpe E, Vidal N, et al. HIV Type 1 Subtype C Remains the Predominant Subtype in Men Having Sex with Men in Senegal. AIDS Res Hum Retroviruses 2013; 29(9): Tramuto F, Maida CM, Bonura F, Perna AM, Vitale F. Dynamics and molecular evolution of HIV-1 strains in Sicily among antiretroviral naive patients. Infect Genet Evol 2013; 16: Robinson K, Fyson N, Cohen T, Fraser C, Colijn C. How the dynamics and structure of sexual contact networks shape pathogen phylogenies. PLoS Comput Biol 2013; 9(6): e

85 78. Leventhal GE, Kouyos R, Stadler T, et al. Inferring epidemic contact structure from phylogenetic trees. PLoS Comput Biol 2012; 8(3): e Frost SD, Volz EM. Modelling tree shape and structure in viral phylodynamics. Philos Trans R Soc Lond B Biol Sci 2013; 368(1614): UNAIDS. Global HIV/AIDS respone: Epidemic update and health sector progress towards universal access. 2011;. 81. Brown AE, Gifford RJ, Clewley JP, et al. Phylogenetic reconstruction of transmission events from individuals with acute HIV infection: toward more-rigorous epidemiological definitions. J Infect Dis 2009; 199(3): Holmes EC, Ghedin E, Halpin RA, et al. Extensive geographical mixing of 2009 human H1N1 influenza A virus in a single university community. J Virol 2011; 85(14): Dennis AM, Hue S, Hurt CB, Napravnik S, Sebastian J, Pillay D, Eron JJ. Phylogenetic insights into regional HIV transmission. AIDS 2012; 26(14):

86 Chapter 4 The role of viral introductions in sustaining community-based HIV epidemics in rural Uganda: evidence from spatial clustering, phylogenetics, and egocentric transmission models Mary K. Grabowski, Justin Lessler, Andrew D. Redd, Joseph Kagaayi, Oliver Laeyendecker, Anthony Ndyanabo, Martha I. Nelson, Derek A.T. Cummings, John Baptiste Bwanika, Amy C. Mueller, Steven J. Reynolds, Supriya Munshaw, Stuart C. Ray, Tom Lutalo, Jordyn Manucci, Aaron A.R. Tobian, Larry W. Chang, Chris Beyrer, Jacky M. Jennings, Fred Nalugoda 3, David Serwadda, Maria J. Wawer, Thomas C. Quinn, Ronald H. Gray, and the Rakai Health Sciences Program 4.1. Abstract Background It is often assumed that local sexual networks play a dominant role in HIV spread in sub- Saharan Africa. The aim of this study was to determine the extent to which continued HIV transmission in rural communities home to two-thirds of the African population is driven by intra-community sexual networks versus viral introductions from outside of communities. Methods and Findings We analyzed the spatial dynamics of HIV transmission in rural Rakai District, Uganda, using data from a cohort of 14,594 individuals within 46 communities. We applied spatial clustering statistics, viral phylogenetics, and probabilistic transmission models to 64

87 quantify the relative contribution of viral introductions into communities versus community- and household-based transmission to HIV incidence. Individuals living in households with HIV-incident (n = 189) or HIV-prevalent (n = 1,597) persons were 3.2 (95% CI: ) times more likely to be HIV infected themselves compared to the population in general, but spatial clustering outside of households was relatively weak and was confined to distances <500 m. Phylogenetic analyses of gag and env genes suggest that chains of transmission frequently cross community boundaries. A total of 95 phylogenetic clusters were identified, of which 44% (42/95) were two individuals sharing a household. Among the remaining clusters, 72% (38/53) crossed community boundaries. Using the locations of self-reported sexual partners, we estimate that 39% (95% CI: 32% 44%) of new viral transmissions occur within stable household partnerships, and that among those infected by extra-household sexual partners, 62% (95% CI: 55% 70%) are infected by sexual partners from outside their community. These results rely on the representativeness of the sample and the quality of self-reported partnership data and may not reflect HIV transmission patterns outside of Rakai. Conclusions Our findings suggest that HIV introductions into communities are common and account for a significant proportion of new HIV infections acquired outside of households in rural Uganda, though the extent to which this is true elsewhere in Africa remains unknown. Our results also suggest that HIV prevention efforts should be implemented at spatial scales broader than the community and should target key populations likely responsible for introductions into communities. 65

88 4.2. Introduction Effective prevention and control of the human immunodeficiency virus (HIV) builds upon an understanding of the dynamics that sustain viral transmission within sexual networks 1,2. These networks are comprised of sexual partnerships between individuals within households, between community members not sharing a household, and between individuals in different communities. While sufficiently large intracommunity sexual networks can potentially maintain local HIV epidemics, virus introduced from sources external to the community may also sustain incidence 3,4. The effectiveness of interventions designed to prevent HIV transmission within a given community or any other geographic unit depends in part upon the attributable fraction of new cases infected through partners residing within the targeted area and those infected from partners residing outside of that area 4 7. These proportions are particularly relevant to population-based antiretroviral therapy (ART) strategies for HIV prevention that aim to benefit individuals who do not themselves receive the treatment by reducing their risk of infection. In 2011, ART was established as a highly effective tool for HIV prevention in the landmark HPTN 052 clinical trial 8, which showed that ART almost universally prevents HIV transmission within HIV-discordant couples 8,9. The concept of ART for HIV prevention ( treatment as prevention ) is now widely accepted, and in 2012, it was adopted by the US President s Emergency Program for AIDS Relief as a key strategy for population-based HIV control 10. Despite the widely heralded success of HPTN 052, it is unknown whether ART can be scaled to levels necessary to interrupt community-level HIV transmission. Uncertainty remains, in part, because the treated population in HPTN 66

89 052 represented a unique subset of the total HIV-infected population: participants were in the chronic stages of HIV infection, receiving care for their disease, and in a stable sexual partnership 8. Transmission in the broader population occurs along a complex sexual network in which virus is transmitted by infected individuals in early and chronic stages of HIV infection and between individuals who may or may not be in stable sexual partnerships. These complexities have motivated large community-randomized controlled trials (CRCTs) of ART for HIV prevention in African populations, including the HPTN 071 study in Zambia and South Africa 11 and the Mochudi Prevention Project in Botswana 12. By virtue of their community-randomized design, these CRCTs presume that the preponderance of viral transmissions occur between partners residing within the same communities of randomization 13 ; however, it is unknown what fraction of HIV transmissions in Africa occur within communities versus across community boundaries. The empirical study of HIV transmission outside of stable couples is challenging, but new approaches to epidemiological inference and evolutionary biology provide unprecedented opportunities to understand the spatial scale of HIV transmission networks. Here we test the hypothesis that extra-household HIV transmission is predominately sustained through intra-community sexual networks using populationbased cohort data from 14,594 individuals, including 189 individuals with incident HIV residing within 46 communities in the Rakai District, Uganda. Rakai, bordered by Tanzania to the south and Lake Victoria to the east, is rural and represents one of the earliest epicenters of the HIV/AIDS epidemic in east Africa 14. Presently, HIV transmission in Rakai is endemic, with circulation of HIV-1 subtypes A, D, and C, and multiple recombinant viruses

90 Our study consists of three primary analyses, in all of which the primary geographic unit of interest was the community. In the first analysis we used the geographic coordinates of participant households and measured the tendency of HIVseropositive persons to spatially cluster within and outside of communities. If local transmission dynamics dominate, we expect infected persons to spatially cluster at geographic distances consistent with intra-community transmission. In the second analysis we examined the genetic relatedness of infecting viruses within communities. If transmission is sustained through local sexual networks, viruses within newly infected persons should be more similar to viruses of other HIV-infected persons within the community than to those of individuals outside the community. Finally, we used egocentric network information on the geographic locations of recent sexual partners to estimate the proportions of new transmissions occurring between household, community, and extra-community partners. In this third analysis we also estimated the proportion of household transmissions occurring within 1 y of an index household infection. Each of these three independent, yet complementary, analyses has its own strengths and weaknesses, and together they are a powerful set of inferential tools for understanding the spatial scale and structure of HIV transmission networks Materials and Methods Ethics Statement The study was independently reviewed and approved by Ugandan (Ugandan Virus Research Institute Security and Ethics Committee; Protocol GC/127/13/01/16) and US (Western Institutional Review Board; Protocol ) institutional review 68

91 boards. All study participants provided written informed consent at baseline and followup visits using institutional review board approved forms Study population and setting The Rakai Community Cohort Study (RCCS) is a well-characterized populationbased HIV surveillance cohort in the Rakai District, Uganda (Figure 1A). Methods for the RCCS have been described in detail elsewhere 6. Briefly, the RCCS enrolls all consenting persons aged y residing in 50 village communities. The RCCS defines households as a group of persons who sleep under one roof and eat out of a common pot, and a community as an administrative unit whose boundaries are determined by the Ugandan government (Local Council 1 and Local Council 2 units, the two smallest political units in Uganda). Eleven larger community groupings (2 8 communities each), referred to as geographic regions, were previously designated by the RCCS based upon geographic proximity and the frequency of cross-community contact (Figure 1B) 6. Study participants are administered a detailed questionnaire at visits occurring every mo and provide a serological sample at each visit. HIV serostatus is assessed by two enzyme immunoassays (Vironostika HIV-1, BioMerieux Inc., Charlotte, NC and Recombigen, Cambridge Biotech, Worcester, Massachusetts), with Western blot confirmation of discordant enzyme immunoassays and for all HIV seroconverters (HIV-1 WB, BioMerieux-Vitek, St. Louis, Missouri). RCCS participation rates are ~90% of persons present at time of survey, and follow-up rates between successive visits are ~75%. 69

92 In this study, we used data from RCCS survey round 13 (RCCS R13) for all data analyses (spatial clustering, viral phylogenetics, and egocentric transmission models). RCCS R13 was conducted between June 17, 2008 and December 7, 2009 within 46 of the 50 RCCS communities. It included surveys of 14,594 participants residing in 8,899 households, the collection of household GPS coordinates (8,156/8,899, or 91.6% of study households; resolution ~3 5 m), and viral sequencing for ART-naïve HIV-seropositive individuals. Participants who were HIV- seropositive upon entry into RCCS R13 were defined as HIV seroprevalent in all analyses. The average maximum distance between any two households within a community (i.e., the community size) was ~3 km (Figure S1). Though our three primary analyses use data drawn from the same study population (RCCS R13), each analysis was conducted independently of the others Spatial clustering analyses Using the geographic coordinates of participant households in RCCS R13 the spatial relatedness between HIV-seropositive individuals was characterized by τ(d 1,d 2 ), defined as the relative probability that a participant A residing within a distance range, d 1 to d 2, from an HIV-seropositive participant B was also HIV seropositive versus the probability that any RCCS participant was HIV seropositive, regardless of spatial location 16. It is estimated as: ˆ( τ d, d ) = 1 2 N z i 1 ( 1, 2) iz = j Ωi d d j N z i= 1 j Ωi ( d1, d2) i N i= 1 N j Ωi (0, ) i= 1 j Ωi ( 0, ) zz i z i j (1) 70

93 where Ω i (d 1,d 2 ) is the set of points in spatial range (d 1,d 2 ) of point i, and Z i indicates seropositivity. We also measured the spatial clustering of seroincident cases with other seroincident cases, and of seroincident cases with HIV-seroprevalent persons on and off ART. Values of τ(d 1,d 2 ) were calculated at 0 m (household) and for 250-m wide windows centered from 125 m to 30 km in 50-m increments. Where spatial clustering exists, τ(d 1,d 2 ) will be greater than 1. The significance of clustering was assessed by bootstrapping (1,000 iterations), where pairs of individuals were sampled with replacement. Instead of resampling individuals, samples were drawn from all possible pairs of individuals in the study to ensure no comparisons occurred between an individual and him/herself in bootstrapped samples Viral extractions and HIV-1 subtype assignment Viral RNA extractions were performed on sera of all ART-naïve HIVseropositive participants in RCCS R13 (n = 1,434) using the QiAmp Viral Mini Kit (Qiagen). Extracted RNA was amplified by reverse transcriptase polymerase chain reaction and an additional nested polymerase chain reaction in two separate assays for partial gag (HXB2 nucleotides 1249 to 1704) and env (HBX2 nucleotides 7858 to 8260) sequences, as previously described 15,17. RNA extractions and PCR assays were conducted in separate designated laboratory spaces for quality control. HIV amplicons were sequenced using direct Sanger methods on the Applied Biosystems 373xl DNA Analyzer. Results were examined immediately for contamination and batch effects. We also repeated testing for a subset of specimens (extraction through sequencing). Sequential samples from the same individual always clustered together when compared using phylogenetic methods (Figure S2). 71

94 HIV-1 subtype assignments were made using the US National Center for Biotechnology Information genotyping database and then confirmed phylogenetically with reference sequences from the Los Alamos National Laboratory HIV Sequence Database (HIVDB). Sequences were aligned with MUSCLE v3.7 and manually edited in Bioedit v Ambiguous regions in sequence alignments were removed using GBLOCKS v0.91b 19. Final alignments were ~564 bp in the gag gene and ~467 bp in the env gene. Sequences were scanned with all available methods in the Recombination Detection Program v Within-gene recombination events identified in one or more analyses were verified using jumping hidden Markov models 21. Intra-gene recombinant sequences were excluded from additional phylogenetic analyses (gag, n = 17; env, n = 8) Phylogenetic analysis Maximum likelihood (ML) methods under an HKY-85 model of nucleotide substitution were used to estimate genetic pairwise distances and reconstruct phylogenetic trees for gag and env genes and for HIV-1 A, D, and C subtypes separately (six datasets in total). African reference sequences (one per individual reference ID) were selected from the Los Alamos National Laboratory HIV Sequence Database for analyses. Using HKY-85 genetic pairwise distance, the three Los Alamos National Laboratory HIV Sequence Database reference sequences most similar to each participant s sequence were identified, and the unique subset of these sequences was defined as the reference set for RCCS R13 (Table S1). The reference set included viral sequences from all major geographic regions in sub-saharan Africa. 72

95 ML phylogenetic trees were reconstructed under two models of nucleotide substitution, the HKY-85 model and the general time reversible model with gamma distributed rate heterogeneity and a proportion of invariable sites (GTR+I+G) 22,23. In the GTR+I+G model all possible nucleotide substitution rates are estimated, whereas in the HKY-85 model only transition and transversion rates are estimated (six versus two substitution rate parameters). We defined a cluster of related HIV cases as two or more participants whose sequences were contained within a monophyletic group in ML trees in either one or both gene regions (gag or env) at a bootstrap threshold of 90% or greater (1,000 replications). Clusters also met intra-cluster median genetic distance thresholds, where thresholds were defined using RCCS genetic data from epidemiologically linked HIV-incident couples (i.e., where at least one of the partners was an incident case). Specifically, genetic distance thresholds for each gene region were defined as the 95% quantile of the distribution of ML branch length distances between epidemiologically linked sexual partners (i.e., known couples) where at least one of the partners was an incident case and the partner sequences were contained within a monophyletic cluster with moderately high clade support ( 70%; Figure S3). Distance thresholds estimated for gag and env genes were 1.3% and 2.6%, respectively. ML clusters were confirmed using Bayesian phylogenetics, where confirmation was established if the same sequences clustered together in the Bayesian tree with posterior probability equal to one. The ML tree topologies obtained using the more parameter-rich GTR+I+G model were similar to those obtained under the HKY-85 model, and so Bayesian confirmation of clusters was conducted using the HKY-85 model only. Bayesian analyses were conducted using MrBayes v3.2 24, where trees were 73

96 obtained through separate unconstrained phylogenetic analyses (i.e., no molecular clock) and each codon position was allowed to have its own site-specific rate. Four independent runs were performed for generations, and a burn-in of 25% was used for final analyses. Effective sample sizes for all parameters exceeded 200. We assessed the sensitivity of our cluster definition using alternate cluster definitions in the ML analysis: 70%, 80%, 90%, and 99% bootstrap thresholds with and without genetic distance thresholds for HKY-85 and GTR+I+G models of substitution. We present the ML radial and square phylogenetic trees estimated under the HKY-85 model as figures in this article and in the Supporting Information. Community and household labels used in the square trees were blinded (i.e., true RCCS identification numbers were not used), and the exact community locations were not labeled on geographic maps to ensure the privacy of our study participants. The ML phylogenetic trees constructed under the GTR+I+G model and the Bayesian phylogenetic trees are available from the authors upon request Egocentric transmission model Study participants in RCCS R13 were asked about their most recent sexual partners (up to four partners, restricted to last 12 mo). Stable partnerships were defined as either marriages or long-term consensual unions. All other partner types (boyfriend/girlfriend and casual) were defined as non-stable. Participants were asked whether each sexual partner s primary residence was within the same household, within the same community, or outside of that individual s community. As per protocol, RCCS participant identifiers could only be matched with a named partner for stable (usually household) partners.. If the stable partner was also an RCCS participant, we considered 74

97 those partners to be epidemiologically linked. In instances where the epidemiologically linked partner did not participate in RCCS R13 but did so in a prior RCCS survey and he/she was HIV seropositive at his/her last study visit, we considered that partner HIV seroprevalent. When discrepancies between the self-reported geographic locations of household partners and GPS data obtained through RCCS were identified (~2%, n = 256 self-reported partners), data were independently reviewed and adjudicated by study investigators (M. K. G., A. D. R.). We considered a household HIV-seropositive partner to be on ART if that person was on ART for 50% of the inter-survey interval in which their initially uninfected partner was at risk for HIV. The RCCS has identified no HIV seroconversions within serodiscordant couples where the HIV-infected partner is on ART since ART was introduced in Rakai in ; therefore, we assumed that HIV-seropositive household partners on ART posed no risk to their uninfected partners in this analysis. HIV sequence data for self-reported sexual partners was obtained only if those partners could be identified as being another RCCS participant, and this was possible only for stable partners. For phylogenetic methods to exclude any self-reported partner as a source of infection, sequences from all partners and the ability to detect co-infection are needed. As neither was available in this study, the egocentric transmission model and phylogenetics were conducted as independent, though complementary, analyses. We used egocentric sexual partner data from HIV-seronegative and -incident participants (excluding those HIV-seronegative participants who entered into the study for the first time in RCCS R13 or who had missed more than two previous study visits) to 75

98 model the probability of HIV infection from self-reported partners and unreported partners/sources as follows: n i wij z mij ij Pr ( Yi = 1) = 1 ( 1 ρi ) ( 1 α) ( 1 γ) ( 1 πij ) (2) j= 1 where Y i is equal to 1 if participant i is an incident case; n i is the number of partners of case i; w ij, z ij, and m ij are indicators of whether partner j of case i is ART-naïve seroprevalent, incident, or missing HIV status, respectively; α and γ are the probabilities of infection from ART-naïve seroprevalent and incident partners, respectively, between study rounds; π ij is the probability of case i being infected by a partner j with missing status given their respective locations; and ρ i is the probability of i being infected from an unnamed partner/source. The probability of infection from a self-reported partner of unknown HIV status was modeled as follows: ( ) 1 2C 3 logit π = π + π HH + π + π FC (3) ij o ij ij i ij where logit(π ij ) is the log odds that i was infected by partner j, HH ij is an indicator of whether participant i shares a household with partner j, C ij is an indicator of whether the partner is outside the community, and F i is an indicator of whether partner i is female. Parameters were estimated using Markov chain Monte Carlo methods. The numbers of infections attributable to specific partnership types were estimated by sampling parameters from the posterior distribution and then simulating sources of infections for each parameter set (250,000 iterations). In households where both partners had incident infection we initially randomly assigned one partner as having been infected 76

99 first (i.e., without an identifiable incident partner) and the other partner as having been infected second (i.e., with an identifiable incident partner). Assignments were updated in each Markov chain Monte Carlo iteration and accepted or rejected using the standard Metropolis-Hastings criteria. For each incident infection, the probability of infection by each type of partner was calculated based upon the current parameter set and then normalized so that they summed to one (i.e., calculated conditional on that individual having been infected). Which partner (or unknown source) infected each individual was then randomly selected based upon these probabilities. The sensitivity of the parameter estimates from our transmission model to unreported partnerships and misreported community status of partners was assessed by running 100 simulations where 10% of the reported partnerships in the original data were unreported and 100 simulations where the community status of 10% of extra-household partners was misreported (i.e., intra-community was changed to extra-community or vice versa) Results There were 14,594 individuals who participated in RCCS R13 ( ; Table 1), of whom 3,219 enrolled for the first time (7.8% were HIV seroprevalent at study entry, n = 252/3,219). More than 60% of the surveyed population was married (60.2%, n = 8,790/14,594), and slightly more than half of study participants were female (56.1%, n = 8,188/14,594). Study participants who were not in marital relationships included those who had never been married (27.3%, n = 3,982/14,594) or were previously but not currently married (12.3%, n = 1,795/14,594). Considering only married men, 15.3% (n = 560/3,664) were in polygamous unions. 77

100 We surveyed eligible adults aged y per community during RCCS R13, with 70% coverage of the censused target population (n = 14,594/21,275). There were 1,786 HIV-seropositive men and women who participated in RCCS R13, of whom 189 were incident cases. Among the HIV-seroprevalent individuals in this survey (n = 1,597), 1,345 had participated in a prior RCCS survey round, and 26.2% (n = 352/1,345) of these individuals were on ART. Among the HIV-seroprevalent men and women entering into the RCCS for the first time (n = 252), none were on ART. Overall, HIVseropositivity was 12.2% (n = 1,786/14,594), and incidence was 1.2 per 100 person-years (95% CI: ) (Table 1). Individuals who were lost to follow-up during the interval prior to RCCS R13 (30.9% attrition) were significantly more likely to be unmarried (Poisson unadjusted relative risk [RR] =1.59; 95% CI: ) and significantly more likely to be less than age 25 y (RR = 1.62; 95% CI: ) than those who remained in the study. Persons lost to follow-up were marginally more likely to be male (RR = 1.07; 95% CI: ) and HIV seropositive (RR = 1.09; 95% CI: ) Spatial clustering of HIV-seropositive individuals Spatial clustering of HIV-seropositive individuals within households. We observed strong spatial clustering of HIV-seropositive individuals within households (Figure 2A 2C). The probability that a participant living in the same household as an HIV-seropositive participant was also HIV seropositive was 3.2 (95% CI: ) times greater than the probability that any RCCS participant was HIV seropositive (shown in red, Figure 2A). Even stronger household spatial clustering was observed among HIVincident cases: the probability that a participant living with an HIV-incident case was also 78

101 HIV incident was 10.8 (95% CI: ) times the probability any participant was an HIV incident case (shown in blue, Figure 2C) Spatial clustering of HIV-seropositive individuals within communities. We explored whether there was spatial clustering of HIV-seropositive individuals outside of households at distances up to 30 km. We found statistically significant though weaker spatial clustering of HIV-seropositive persons outside of households. Compared to all study participants, persons living m from a HIVseropositive participant were 1.22 (95% CI: ) times as likely to be HIV seropositive themselves, and those living m away were 1.08 (95% CI: ) times as likely to be HIV seropositive (Figure 2A and 2D). We also examined whether incident cases spatially clustered with other HIVincident and -seroprevalent cases outside the household, since spatial clustering among all HIV-seropositive persons may reflect historic rather than recent patterns of HIV transmission. In contrast, we observed no statistically significant extra-household spatial clustering of HIV-incident cases with other incident or seroprevalent cases (Figure 2B and 2D), though incident cases appeared to weakly cluster with seroprevalent cases at geographic distances less than 500 m (shown in yellow, Figure 2B and 2D). There was no significant spatial clustering beyond 500 m in any spatial analyses and no significant intra-household or extra-household spatial clustering between HIV-incident and HIVseroprevalent persons on ART (Figure S4) HIV phylogenetics within and across communities Viral sequence data for the gag and env genes were obtained for 1,099/1,434 (76.6%) HIV-seropositive participants who were not on ART at the time of the RCCS 79

102 R13 survey (Table S2), including 164 of 189 (86.7%) incident cases (Table S3). On average, 15 (range 3 24) viral sequences were retrieved from HIV-incident cases, and 85 (range ) sequences were retrieved from HIV-prevalent cases per geographic region. Sequences were predominantly HIV-1 subtypes A1 or D, and both subtypes were found in all communities. Of those participants with sequence information in both gene regions (n = 842/1,099), 21.1% (n = 178/842) did not share the same HIV-1 subtype in gag and env genes. No statistically significant differences were found between HIVinfected individuals from whom viral sequences were obtained (in either or both genes) and those from whom no viral genetic data were obtained for duration of the participant s infection (prevalent or incident), gender, marital status, or geographic region of residence. However, there was a significant decrease in the number of sequences obtained with each increasing year of age (either gene: RR = 0.988; 95% CI: ; both genes: RR = 0.990; 95% CI: ) Genetic relatedness of HIV viruses within households. Our study population included 165 epidemiologically linked couples where both partners had participated in RCCS R13 and were HIV seropositive and not on ART at the time of the survey. Twenty-five percent (n = 42/165) of these couples included at least one incident case (both partners were HIV incident in 9/42 incident couples). Sequence information was available for at least one gene region (either gag or env) in 63.6% (n = 105/165) of epidemiologically linked couples, including 76.2% (n = 32/42) of those with one or more incident cases (n = 7/9, 77.7% of those where both cases were incident). Ninety-nine percent (n = 104/105) of epidemiologically linked pairs with sequence data shared a household, including all 32 incident couples. 80

103 The median genetic distance between epidemiologically linked couples with an incident case was 0.4% in gag (n = 24/32, interquartile range [IQR]: 0.3% 0.9%) and 0.9% in env (n = 27/32, IQR: 0.4% 1.3%; Figure 3A). All of these epidemiologically linked couples (n = 32/32) shared the same viral subtype in one or both genes, but only 71.9% (n = 23/32) shared a phylogenetic cluster in the ML trees in at least one gene region. In comparison, the median intra-subtype genetic distance between epidemiologically linked HIV-seroprevalent partners was 1.3% in gag (n = 47/73, IQR: 0.9% 2.2%) and 2.7% in env (n = 55/73, IQR: 2.0% 4.2%), and only 38.4% (n = 28/73) of these couples phylogenetically clustered in at least one gene region. There were 12 households where sequence data were available for two persons who were not epidemiologically linked, all of whom were HIV-seroprevalent pairs. Median intra-subtype genetic distance in these pairs was 6.4% in gag (n = 7/12, IQR: 3.0% 7.5%) and 9.4% in env (n = 10/12, IQR: 7.0% 10.7%), and only one pair phylogenetically clustered within the ML trees. A detailed summary of the HIV genetic data for all of the 105 epidemiologically linked couples with HIV sequence data is included in Table S Genetic relatedness of HIV viruses within and across communities. Shown in Figure 3B is the distribution of intra-subtype genetic distances in the gag gene for incident couples (i.e., one sequence obtained from an incident case) sharing the same community (median = 6.3%; IQR: 5.4% 7.3%). This distribution was nearly identical to that seen within geographic regions (median = 6.4%; IQR: 5.4% 7.4%) and across all communities (median = 6.4%; IQR: 5.5% 7.3%). Similar distributions were observed in the env gene (data not shown). 81

104 Limited geographic structure was observed in ML phylogenetic trees, regardless of the viral subtype or gene region examined (Figures 3C and S5 S9. More detailed phylogenetic trees, including information on both HIV status (i.e., incident or prevalent) and community of residence, showed that viral sequences from HIV-incident cases were distributed throughout the phylogenetic tree, with no apparent regard to community or geographic region of primary residence (Figures S5 S9). Two participants sharing a phylogenetic cluster suggests because of our strict cluster definition that they are separated by a relatively short and recent chain of transmission. Only 19.0% (209/1,099) of HIV-infected participants in RCCS R13 shared a phylogenetic cluster with at least one other RCCS study participant in either the gag or env genes. A total of 95 phylogenetic clusters were identified across all ML phylogenetic trees (n = 209 individuals; Tables 2 and S4). The majority of clusters included only two (86.3%, n = 82/95) or three HIV-infected persons (9.5%, n = 9/95). We also identified four additional phylogenetic clusters, of which two clusters contained four individuals each (2.1%, n = 2/95) and two clusters contained five individuals each (2.1%, n = 2/95). None of the identified phylogenetic clusters contained a reference sequence, and 40.0% (n = 38/95) contained at least one incident case, encompassing 50 incident cases in total (Table 2). Almost half of all phylogenetic clusters identified (44.2%, n = 42/95) were household pairs of two (63 prevalent cases; 21 incident cases). Of the 53 clusters that contained participants who spanned households (n = 53/95), 38 clusters crossed community boundaries (71.7%). These 38 cross-community clusters included 28 pairs (47 prevalent cases; nine incident cases); seven triplets (18 prevalent cases; three incident 82

105 cases), two clusters of size four (four prevalent cases; four incident cases), and one cluster of size five (one prevalent case; four incident cases). Nearly half of the crosscommunity clusters (47.4%, n = 18/38) also spanned geographic regions. Community clusters (n = 15/53) included 12 pairs (19 prevalent cases; five incident cases), two triplets (four prevalent cases; two incident cases), and one cluster of size five (three prevalent cases; two incident cases). When analyses were restricted to only those clusters containing at least one incident case (n = 38/95), similar geographic patterns were observed (Table 2). There were six phylogenetic clusters that contained only incident cases (6.3%, n = 6/95), of which five contained a single household pair (ten incident cases) and one contained two household pairs (four incident cases). Our definition of a phylogenetic cluster may have precluded the identification of some transmission chains; however, in sensitivity analyses the proportion of clusters with more than one household that crossed community boundaries was robust to the strictest (66.7%, n = 18/27 crossed community boundaries) and most relaxed (74.0%, n = 77/104 crossed community boundaries) phylogenetic cluster definitions assessed (Table S5). A detailed summary of each of the 95 phylogenetic clusters identified is included in Table S Probable infection from household, community, and extra-community sources A total of 11,992 recent sexual partners were self-reported by 5,368 women and 4,152 men who were HIV seronegative at a previous study visit (Table 3). Of these selfreported partners, 42.1% (n = 5,043) could be epidemiologically linked to another RCCS participant who participated in RCCS R13 or a previous survey round. Ninety-six percent (n = 5,159/5,368) of women reported only one sexual partner in the last 12 mo, compared 83

106 to 59.2% of men (n = 2,458/4,152) (Table S7. Of enumerated self-reported partners, 63.0% (n = 7,549/11,992) held primary residence within the participant s household, 19.5% (n = 2,342/11,992) were within the participant s community but outside of the household, and 17.5% (n = 2,101/11,992) had a primary residence outside of the participant s community (Table S8). Household partnerships were almost always stable partnerships (i.e., 99% were marital or long-term consensual unions), whereas partnerships outside of the household were usually not stable (95%; Table S8). The majority of extra-household sexual partners were reported by unmarried persons (n = 2,895/4,443, 65.2%) Attributable fractions of HIV infections from household-based transmission. Using the egocentric partner data, we estimated that 39.0% (95% CI: 32.3% 43.9%) of 189 incident cases were infected by a household sexual partner (Table 4). Those with an incident household partner (n = 9 household pairs) had an estimated 26.0% (95% CI: 13.4% 45.0%) probability of acquiring HIV from that partner (Table 5). In 20.6% of cases where infection was attributed to a household partner with known HIV status, that partner was him/herself an incident case. There were 38 incident events among 250 individuals in a stable sexual partnership with an ART-naïve HIVseroprevalent partner. After accounting for risk from other self-reported partners and unknown sources, we estimate that the probability of transmission from these seroprevalent household partners not on ART was 15.3% (95% CI: 10.9% 20.6%). Among at risk individuals who had an HIV-seroprevalent partner who was on ART for 50% or more of the risk interval (n=29), only one became HIV-infected; and there were 84

107 no infections among the 27 with partners who were on ART for 60% or more of the interval. The HIV status for the suspected index partner in 16.2% (95% CI: 11.6% 20.1%) of household transmissions was unknown Attributable fractions of HIV infections from community, extracommunity, and unknown sources. Infections from self-reported extra-household partners were estimated to account for 39.5% of new cases (95% CI: 33.9% 42.3%), of which the majority (62.1%, 95% CI: 54.9% 69.7%) were from self-reported partners outside the community (Table 4). Where the specific location of these self-reported extracommunity sexual partners was known (68%), 50% lived outside of the Rakai District and were geographically dispersed throughout Uganda (Figure 1A). While men were 1.8 times more likely to disclose an extra-community partner than women (1,061/4,152 versus 761/5,368; 95% CI: ), those women who reported an extra-community partner had higher odds of HIV acquisition from that self-reported partner than men who reported an extra-community partner (odds ratio = 5.0; 95% CI: ). Acquisition from unknown sources accounted for 21.4% of total infections (95% CI: 14.8% 29.6%), although the individual probability of such infections was low (0.3%; 95% CI: ) Sensitivity analyses. Sensitivity analyses were conducted to determine the robustness of the parameter estimates in Table 5 to underreporting and misreporting of self-reported sexual partnerships. In simulations where 10% of self-reported partnerships were considered unreported, the median bias in parameter estimates for the transmission model was less than 10% of the width of the 95% confidence interval in all cases except for the probability of infection from an unnamed source (ρ), which increased 85

108 as expected. Moreover, all 95% CIs included the original point estimate, with the exception of ρ, which differed as expected. In simulations where 10% of extra-household partnerships were considered to have a misreported geographic relationship with the study participant (i.e., extra-community partners were reported as community partners or vice versa), the median bias of each parameter estimate was less than 10% of the reported 95% CI width, and 97% or more of the 95% CIs from simulated estimates included the estimate from the original data Discussion Using spatial statistics, viral phylogenetics, and egocentric transmission models we find evidence that extra-community HIV introductions are frequent, and likely play a significant role in sustaining ongoing HIV incidence in rural Rakai, Uganda. We estimate that viral introductions combined with intra-household transmission account for the majority of incident infections in this HIV-endemic region, though our data also suggest that community-based sexual networks play a critical part in HIV spread. Our results underscore the complexities of HIV epidemic dynamics and sexual networks in rural Uganda and have important implications for the design and implementation of CRCTs and HIV prevention programs. Each of the analyses used illuminates a different aspect of HIV transmission networks, and together they provide a powerful framework for understanding the spatial scale and structure of HIV transmission networks (Figure 4). Spatial analyses reveal whether HIV incidence or prevalence is elevated in close proximity to HIV-infected persons, but cannot distinguish whether spatially related cases are part of the same sexual network. Viral phylogenetics provides insight into the relationship between spatial and 86

109 viral genetic similarity; however, high mutation rates and sparse sampling of networks make it impossible to definitively link cases to an infecting source. Egocentric transmission models relate the geographic distribution of personal sexual networks to individuals risk of HIV infection, but give minimal insight into global network structure. All three analyses suggest that frequent HIV introductions into communities play a critical role in ongoing HIV incidence in rural Rakai, Uganda (Figure 4). They show limited spatial clustering of HIV cases outside of households, multiple circulating HIV viruses within communities, and a significant proportion of incidence resulting from extra-community partnerships. Together, our data imply that there are frequent viral introductions into communities, followed by onward transmission within households (where we estimate over 1/3 of transmission occurs) and within small intra-community sexual networks. These findings do not rule out an important role for community-level sexual networks in the Rakai HIV epidemic, but do suggest that local HIV epidemics are not sustained through community-based viral transmission alone. Furthermore, they highlight the risks of applying the results of sexually transmitted infection studies in urban areas outside of Africa (e.g., studies showing strong spatial clustering of gonorrhea cases in Baltimore 26 ) to HIV control efforts within rural Africa. In this prospective population-based cohort, intra-household HIV transmission was common, accounting for approximately 39% of new incident cases. This fraction is within the range of that previously estimated in 18 sub-saharan African countries 27, but lower than the 55% 97% estimated in Zambia and Rwanda 28, both based on crosssectional Demographic and Health Surveys (DHS). Hence, targeting treatment to stable HIV-discordant couples could prevent substantial numbers of new infections, but the 87

110 effectiveness of this strategy is largely contingent on the rapid identification and treatment of HIV-infected index partners. Consistent with other studies 29,30, we found that the highest risk of HIV acquisition was within the first 18 mo of an index partner s infection. Chronically HIV-infected individuals also posed substantial, though lower, risk to their uninfected partners; however, ART appeared to eliminate this risk entirely. The strong protective effect of ART observed in this population-based study corroborates the findings from the HPTN 052 clinical trial and other observational studies of HIV transmission in Africa 25. Though no individuals in our study acquired HIV from an identifiable HIV-seroprevalent partner on ART, we cannot rule out the possibility that non-identifiable sexual partners of incident cases were taking ART at the time of transmission. While intra-household transmission was common, it is extra-household transmission that determines the geographic scale of HIV epidemics. Here we estimate that more than half of all household introductions were the result of extra-community partnerships, with a wide geographic range of sexual partner networks. Fifty percent of extra-community partners had a primary residence outside of Rakai, including major urban centers in Uganda (i.e., Kampala and Masaka). Within the Rakai District, but outside of the RCCS target area, there are fishing communities along Lake Victoria where HIV prevalence is extremely high (~40% in data from an unpublished pilot study of 2,106 individuals in fishing communities in the Rakai District). Preliminary data show that men from these high-risk fishing communities frequently travel to RCCS communities, which may in part explain the high rate of HIV infection we observed among unmarried women with extra-community partners. Mobility has long been 88

111 associated with HIV transmission in Africa 31,32, though how exactly it relates to local epidemic dynamics, including the persistence of viral transmission in African contexts, remains understudied. Studies of other infectious diseases and network simulations suggest that such long distance jumps even when infrequent can facilitate persistence of infection within broader contact networks We did not measure the impact of local treatment as prevention in this study; however, our results provide insight into the mechanisms and upper limits of its effectiveness when implemented given the relative fractions of community and crosscommunity HIV spread. Our results suggest that community-based ART programs could have a major impact on African epidemics, but also highlight the need to target extracommunity sources of HIV infection. Viral introductions could be reduced either through wider spread coverage of ART among HIV-infected persons or through prevention interventions that provide direct protection to uninfected individuals (e.g., male circumcision or pre-exposure prophylaxis). Targeting interventions that provide direct protection to those most likely to have extra-community partners may be an important addition to local HIV control strategies. Viral introductions pose significant challenges to epidemiological studies of HIV risk and prevention. Exposure misclassification may be common when using community viral load or other aggregated community-level measures of individual HIV risk 36,37. Similarly, in the case of CRCTs, indirect intervention effects may be obscured when cross-community transmissions are frequent 13. Incorporating phylogenetics and detailed information on individual partnerships into study design may facilitate interpretation of 89

112 results from community-based studies of treatment as prevention, including the upcoming HPTN 071 and Mochudi Prevention Project trials 11,12. Our study had several limitations. While RCCS demographics, including age distribution, marital status, and sexual behaviors, are largely representative of the broader Uganda population (Table S9) 38, our results may not be generalizable and suggest the need to study the spatial dynamics of HIV in other settings. In particular, uptake of HIV preventive services may be greater in RCCS communities, which could bias our estimates of per partnership risk if local HIV-infected partners were less likely to be infectious than partners outside of Rakai. A comparison of male circumcision prevalence in our study population versus in the general Ugandan population, as sampled in the DHS survey in 2011, revealed that the male circumcision rate was higher among RCCS participants than among DHS participants (39.4% versus 26.8%), though HIV prevalence and ART use among HIV-infected persons was similar between RCCS and DHS sampled populations (Table S9). We also considered newly enrolled HIV-seropositive persons to be HIV seroprevalent, potentially underestimating the effect of early HIV infections on transmission. Overrepresentation of particular types of partnerships in our sample could also have biased results. For example, oversampling of household partners could lead to overestimation of the importance of household transmission; however, the proportions of men and women who were married in RCCS were similar to those reported in the Ugandan DHS, and household partners were not selectively recruited over community partners 38. Another limitation is that we identified the geographic sources of HIV infection from self-reported sexual partner data that may be inaccurate. The presence of HIV-incident cases for which no possible infecting partner could be determined indicates 90

113 that some sexual partners were unreported. If these unreported sources of infection were evenly split between community and extra-community partners (as opposed to following the distribution in the data), our estimate of the percentage of extra-household transmission due to community partners would increase from 38% to 45%. Furthermore, sensitivity analyses show that randomly unreported partnerships or randomly misreported community status would not substantially bias the results. However, systematic biases in partnership reporting could bias our results. A notable of strength of our study was its prospective population-based study design, which captured a representative sample of the sexually active adult population in rural Rakai and yielded a sampling fraction of local sexual networks (~70% of the censused population) in the 46 surveyed communities. Individuals lost to follow-up during the interval of observation were more likely to be unmarried and younger than those who remained in the study. Such missing persons may be more mobile and at higher HIV risk. If so, our estimate of the frequency of cross-community transmission is likely an underestimate of the true value. Despite limited losses to follow-up and a high sampling fraction of the primary geographic unit of analysis (the community), we still observed minimal phylogenetic clustering between HIV sequences obtained from the same community, which limited our ability to identify HIV transmission chains using molecular epidemiological methods. Low levels of phylogenetic clustering are not uncommon in studies of HIV epidemics, particularly phylogenetic studies of heterosexual HIV transmission networks 39,40. Still, we were surprised to find so many singleton lineages within communities, given study participation rates. While it is true that we may have undersampled local sexual networks to some extent, high viral diversity within 91

114 communities, coupled with a lack of spatial clustering outside of households and a high probability of infection from extra-community partners, implies that the limited phylogenetic clustering is a reflection of frequent viral introductions, at least in part. Intra-host HIV evolutionary dynamics, including HIV co-infection and rapid HIV genetic drift, also may have obscured the identification of HIV transmission chains using our phylogenetic approaches. Taken together, our analyses reveal a complex picture of HIV dynamics in rural Uganda, and suggest that incidence is in part sustained through repeated introductions of HIV, with frequent intra-household transmission and some onward transmission through small intra-community networks. It remains unknown whether these patterns reflect broader source sink dynamics, in which localized key populations, such as fishing communities with high HIV prevalence, may have a major effect on regional HIV transmission dynamics. HIV introductions present a challenge to local HIV control programs and CRCTs, necessitating a commitment to widespread combination HIV prevention in sub-saharan Africa, and a deeper understanding of the extra-community partnerships that reintroduce infection into rural populations. 92

115 4.6. References 1. Anderson RM, May RM, Boily MC, Garnett GP, Rowley JT (1991) The spread of HIV-1 in Africa: sexual contact patterns and the predicted demographic impact of AIDS. Nature 352: Doherty IA, Padian NS, Marlow C, Aral SO (2005) Determinants and consequences of sexual networks as they affect the spread of sexually transmitted infections. J Infect Dis 191 (Suppl 1): S42-S Hagenaars TJ, Donnelly CA, Ferguson NM (2004) Spatial heterogeneity and the persistence of infectious diseases. J Theor Biol 229: Deredec A, Courchamp F (2003) Extinction thresholds in host-parasite dynamics. Ann Zool Fennici 40: Grosskurth H, Gray R, Hayes R, Mabey D, Wawer M (2000) Control of sexually transmitted diseases for HIV-1 prevention: understanding the implications of the Mwanza and Rakai trials. Lancet 355: Wawer MJ, Sewankambo NK, Serwadda D, Quinn TC, Paxton LA, et al. (1999) Control of sexually transmitted diseases for AIDS prevention in Uganda: a randomised community trial. Lancet 353: Garnett GP, Becker S, Bertozzi S (2012) Treatment as prevention: translating efficacy trial results to population effectiveness. Curr Opin HIV AIDS 7: Cohen MS, Mastro TD, Cates W Jr (2009) Universal voluntary HIV testing and immediate antiretroviral therapy. Lancet 373: Eshleman SH, Hudelson SE, Redd AD, Wang L, Debes R, et al. (2011) Analysis of genetic linkage of HIV from couples enrolled in the HIV prevention trials network 052 trial. J Infect Dis 204: US President s Emergency Program for AIDS Relief (2012) PEPFAR blue print: creating an AIDS-free generation. Washington (District of Columbia): Office of the Global AIDS Coordinator. 11. HIV Prevention Trials Network (2012) HPTN 071: population effects of antiretroviral therapy to reduce HIV transmission (PopART): a clusterrandomized trial of the impact of a combination prevention package on population-level HIV incidence in Zambia and South Africa. Available: Accessed 8 July

116 12. (2013) The Mochudi Prevention Project ART protocol. ClinicalTrials.gov. Available: Accessed 8 July Hayes RJ, Alexander ND, Bennett S, Cousens SN (2000) Design and analysis issues in cluster-randomized trials of interventions against infectious diseases. Stat Methods Med Res 9: Serwadda D, Mugerwa RD, Sewankambo NK, Lwegaba A, Carswell JW, et al. (1985) Slim disease: a new disease in Uganda and its association with HTLV- III infection. Lancet 2: Collinson-Streng AN, Redd AD, Sewankambo NK, Serwadda D, Rezapour M, et al. (2009) Geographic HIV type 1 subtype distribution in Rakai district, Uganda. AIDS Res Hum Retroviruses 25: Salje H, Lessler J, Endy TP, Curriero FC, Gibbons RV, et al. (2012) Revealing the microscale spatial signature of dengue transmission and immunity in an urban population. Proc Natl Acad Sci U S A 109: Conroy SA, Laeyendecker O, Redd AD, Collinson-Streng A, Kong X, et al. (2010) Changes in the distribution of HIV type 1 subtypes D and A in Rakai district, Uganda between 1994 and AIDS Res Hum Retroviruses 26: Edgar RC (2004) MUSCLE: multiple sequence alignment with high accuracy and high throughput. Nucleic Acids Res 32: Castresana J (2000) Selection of conserved blocks from multiple alignments for their use in phylogenetic analysis. Mol Biol Evol 17: Martin DP, Lemey P, Lott M, Moulton V, Posada D, et al. (2010) RDP3: a flexible and fast computer program for analyzing recombination. Bioinformatics 26: Schultz AK, Zhang M, Bulla I, Leitner T, Korber B, et al. (2009) jphmm: improving the reliability of recombination prediction in HIV-1. Nucleic Acids Res 37(Web Server issue): W647-W Miller MA, Pfeiffer W, Schwartz T (2010) Creating the CIPRES science gateway for inference of large phylogenetic trees [abstract] Gateway Computing Environments Workshop; 14 Nov 2010; New Orleans, Louisiana, US. 23. Zwickl DJ (2006) Genetic algorithm approaches for the phylogenetic analysis of large biological sequence datasets under the maximum likelihood criterion. Available: Accessed 8 July

117 24. Ronquist F, Teslenko M, van der Mark P, Ayres DL, Darling A, et al. (2012) MrBayes 3.2: efficient Bayesian phylogenetic inference and model choice across a large model space. Syst Biol 61: Reynolds SJ, Makumbi F, Nakigozi G, Kagaayi J, Gray RH, et al. (2011) HIV-1 transmission among HIV-1 discordant couples before and after the introduction of antiretroviral therapy. AIDS 25: Becker KM, Glass GE, Brathwaite W, Zenilman JM (1998) Geographic epidemiology of gonorrhea in Baltimore, Maryland, using a geographic information system. Am J Epidemiol 147: Bellan SE, Fiorella KJ, Melesse DY, Getz WM, Williams BG, et al. (2013) Extra-couple HIV transmission in sub-saharan Africa: a mathematical modelling study of survey data. Lancet 381: Dunkle KL, Stephenson R, Karita E, Chomba E, Kayitenkore K, et al. (2008) New heterosexually transmitted HIV infections in married or cohabiting couples in urban Zambia and Rwanda: an analysis of survey and clinical data. Lancet 371: Hollingsworth TD, Anderson RM, Fraser C (2008) HIV-1 transmission, by stage of infection. J Infect Dis 198: Wawer MJ, Gray RH, Sewankambo NK, Serwadda D, Li X, et al. (2005) Rates of HIV-1 transmission per coital act, by stage of HIV-1 infection, in Rakai, Uganda. J Infect Dis 191: Nunn AJ, Wagner HU, Kamali A, Kengeya-Kayondo JF, Mulder DW (1995) Migration and HIV-1 seroprevalence in a rural Ugandan population. AIDS 9: Pison G, Le Guenno B, Lagarde E, Enel C, Seck C (1993) Seasonal migration: a risk factor for HIV infection in rural Senegal. J Acquir Immune Defic Syndr 6: Read JM, Keeling MJ (2003) Disease evolution on networks: the role of contact structure. Proc Biol Sci 270: Messinger SM, Ostling A (2009) The consequences of spatial structure for the evolution of pathogen transmission rate and virulence. Am Nat 174: Keeling MJ, Woolhouse ME, Shaw DJ, Matthews L, Chase-Topping M, et al. (2001) Dynamics of the 2001 UK foot and mouth epidemic: stochastic dispersal in a heterogeneous landscape. Science 294:

118 36. Tanser F, Barnighausen T, Hund L, Garnett GP, McGrath N, et al. (2011) Effect of concurrent sexual partnerships on rate of new HIV infections in a highprevalence, rural South African population: a cohort study. Lancet 378: Das M, Chu PL, Santos GM, Scheer S, Vittinghoff E, et al. (2010) Decreases in community viral load are accompanied by reductions in new HIV infections in San Francisco. PLoS ONE 5: e Uganda Bureau of Statistics, ICF International (2012) Uganda demographic and Health survey 2011.Fairfax (Virginia): ICF International. 39. Hughes GJ, Fearnhill E, Dunn D, Lycett SJ, Rambaut A, et al. (2009) Molecular phylodynamics of the heterosexual HIV epidemic in the United Kingdom. PLoS Pathog 5: e Yirrell DL, Pickering H, Palmarini G, Hamilton L, Rutemberwa A, et al. (1998) Molecular epidemiological analysis of HIV in sexual networks in Uganda. AIDS 12:

119 Table 1. Summary statistics for the 46 Rakai communities (within 11 geographic regions) surveyed in RCCS R13. Region Community Participant s N Female n (Percent) Married a n (Percent) Households n HIV Seropositive n (Percent) HIV Seroinciden t n Person- Years n Incidence per 100 Person- Years b (95% CI) Overall 14,594 8,188 (56.1) 8,790 (60.2) 8,899 1,786 (12.2) , ( ) (64.6) 109 (55.1) (23.2) ( ) (61.7) 316 (58.2) (15.7) ( ) (55.2) 67 (43.5) (11.7) ( ) (58.5) 224 (57.7) (16.0) ( ) Total 1, (60.4) 716 (55.8) (16.4) 23 1, ( ) (58.5) 216 (54.0) (9.0) ( ) (54.8) 203 (49.0) (11.1) ( ) (53.0) 158 (58.5) (13.3) ( ) (53.3) 110 (61.1) (14.4) ( ) (54.1) 87 (42.0) (15.0) ( ) (59.8) 99 (47.4) (9.1) ( ) Total 1, (55.8) 873 (52.0) (11.5) 19 1, ( ) (53.8) 214 (67.7) (10.8) ( ) (53.9) 205 (54.7) (9.3) ( ) (50.6) 112 (63.6) (13.6) ( ) Total (53.2) 531 (61.2) (10.7) ( ) (53.3) 241 (63.6) (12.1) ( ) (56.9) 158 (60.3) (13.7) ( ) (56.8) 319 (53.4) (8.5) ( ) (55.9) 240 (60.2) ( 8.5) ( ) Total 1, (56.7) 958 (58.5) 1, (10.2) 16 1, ( ) (61.1) 160 (65.6) (20.5) ( ) (56.4) 97 (69.3) (20.0) ( ) (65.2) 82 (60.7) (18.5) ( ) (59.9) 130 (58.6) (22.1) ( ) Total (60.6) 469 (63.3) (20.5) ( ) (52.6) 254 (59.3) (11.7) ( ) (53.3) 205 (61.7) (14.5) ( ) (55.5) 448 (58.1) (10.2) ( ) (53.0) 317 (62.9) (11.1) ( ) Total 2,035 1,097 (53.9) 1,224 (60.1) 1, (11.4) 19 2, ( ) 97

120 (57.0) 204 (68.5) (11.4) ( ) (56.6) 284 (62.3) (7.9) ( ) (56.1) 364 (62.7) (8.1) ( ) (52.6) 176 (70.1) (11.2) ( ) (54.1) 130 (67.0) (14.4) ( ) Total 1, (55.7) 1,158 (65.1) 1, (9.7) 18 2, ( ) (53.4) 371 (57.8) (8.1) ( ) (51.4) 92 (49.7) (8.6) ( ) (56.2) 469 (65.1) (14.6) ( ) Total 1, (54.5) 932 (60.2) (11.2) 24 1, ( ) (54.4) 316 (62.3) (11.8) ( ) (57.1) 220 (64.7) (15.3) ( ) (52.4) 120 (62.8) (7.9) ( ) (57.3) 41 (54.7) (14.7) ( ) (56.8) 145 (58.0) (16.0) ( ) (61.8) 39 (51.3) (13.2) ( ) (57.5) 205 (64.1) (15.3) ( ) (59.4) 106 (58.9) (6.1) Total 1,939 1,093 (56.4) 1,192 (61.5) 1, (12.8) 27 2, ( ) (54.8) 426 (74.9) (7.9) ( ) (52.4) 115 (68.5) (17.9) ( ) (59.2) 86 (60.6) (12.7) ( ) Total (55.1) 627 (71.3) (10.6) 11 1, ( ) (64.0) 53 (53.0) (27.0) ( ) (62.3) 57 (53.8) (20.8) Total (63.1) 110 (53.4) (23.8) ( ) a Married refers to currently married individuals and includes married monogamous and married polygamous individuals. b Calculated using Poisson regression and assuming that HIV seroconversion occurred at the midpoint of the follow-up interval. 98

121 Table 2. Characteristics of 95 phylogenetic clusters identified in maximum likelihood phylogenetic analyses (HKY-85) of 915 gag sequences and 1,026 env sequences obtained from 1,099 HIV-infected participants in RCCS R13. Cluster Characteristic a Reference unit All Phylogenetic Clusters (n = 95) Phylogenetic Clusters with Incident Case(s) (n = 38) Cluster size distribution Number of participants in 2 (82), 3 (9), 4 (2), 5 (2) 2 (30), 3 (5), 4 (1), 5 (2) cluster (frequency) Clusters containing only incident Number of clusters (percent 6 (6.3) 6 (15.8) cases of total clusters) Household clusters b Number of clusters (percent 42 (44.2) 18 (47.4) of total clusters) Intra-community clusters c Number of clusters (percent 15 (15.8) 7 (18.4) of total clusters) Cross-community clusters d Number of clusters (percent 38 (40.0) 13 (34.2) of total clusters) Cross-regional clusters e Number of clusters (percent of total clusters) 18 (18.9) 6 (15.8) a Categories not mutually exclusive. b Refers to clusters of two individuals who shared the same household. c Refers to clusters of two or more individuals who spanned households but shared the same community. d Refers to clusters of two or more individuals who spanned households and communities. e Refers to clusters of two or more individuals who spanned households, communities, and geographic regions. 99

122 Table 3. Descriptive characteristics of HIV-seronegative and -incident participants in egocentric partner analysis (n = 9,520). Characteristic Women (n = 5,368) Men (n = 4,152) HIV- Seronegative n (Percent) HIV- Incident n (Percent) HIV- Seronegative n (Percent) HIV- Incident n (Percent) Total population 5,258 (98.0) 110 (2.0) 4,073 (98.1) 79 (1.9) Age in years y 290 (5.5) 6 (5.5) 320 (7.9) 2 (2.5) y 921 (17.5) 25 (22.7) 677 (16.6) 12 (15.2) y 1,287 (24.5) 25 (22.7) 820 (20.1) 20 (25.3) y 1,095 (20.8) 31 (28.2) 792 (19.4) 22 (27.8) y 710 (13.5) 13 (11.8) 681 (16.7) 14 (17.7) 40+ y 955 (18.2) 10 (9.1) 783 (19.2) 9 (11.4) Marital status Never married 590 (11.2) 19 (17.2) 906 (22.2) 11 (13.9) Unmarried, previously married 779 (14.8) 34 (30.9) 311 (7.3) 14 (17.7) Married, not polygamous 2,935 (55.8) 39 (35.5) 2,418 (59.4) 45 (57.0) Married, polygamous a 954 (18.1) 18 (16.4) 438 (10.8) 9 (11.4) Number of self-reported recent sexual partners (last 12 mo) b 1 5,060 (96.2) 99 (90.0) 2,425 (59.5) 33 (41.8) (3.6) 8 (7.2) 1,165 (28.6) 30 (40.0) (0.2) 3 (2.7) 483 (11.9) 16 (20.2) Locations of self-reported recent sexual partners Household only 3,907 (74.3) 59 (53.6) 1,918 (47.1) 27 (34.2) Community only 554 (10.5) 12 (10.9) 559 (13.7) 13 (16.5) Extra-community only 644 (12.2) 33 (30.0) 371 (9.1) 5 (6.3) Household and community 73 (1.4) 2 (1.8) 564 (13.8) 10 (12.7) Household and extra-community 44 (0.8) 1 (0.9) 411 (10.1) 14 (17.7) Community and extracommunity 33 (0.6) 3 (2.7) 159 (3.9) 5 (6.3) Household, community, and 3 (0.1) 0 (0.0) 91 (2.2) 5 (6.3) 100

123 extra-community One or more self-reported recent sexual partners with the following HIV serostatus c HIV seronegative 2,274 (43.2) 10 (9.1) 2,361 (58.0) 16 (20.2) ART naïve, HIV incident 16 (0.3) 9 (8.2) 8 (0.2) 9 (11.4) ART naïve, HIV seroprevalent 101 (1.9) 18 (16.4) 91 (2.2) 19 (24.1) Using ART d, HIV seroprevalent 14 (0.2) 0 (0.0) 14 (0.3) 1 (1.2) e Missing HIV serostatus 2,928 (55.7) 76 (69.0) 2,640 (63.3) 64 (81.0) a Married, polygamous for women refers to a woman in a marital relationship with a man who has multiple wives. b Self-reported sexual partners from the egocentric partnership block of the RCCS study questionnaire (records up to four partners in the last 12 mo). c Categories not mutually exclusive (i.e., participants may report multiple partners with different HIV serostatus). d Partners were considered to be on ART if they were using ART for 50% or more of the corresponding index participant s time at risk. e Partner on ART for 58% of the newly infected index participant s time at risk (previous to current survey interval). 101

124 Table 4: Attributable HIV transmissions by geographic location of sexual partner and gender of newly infected participant (estimated from egocentric transmission model). HIV Status of Partner Residential Location of Partner with Respect to Incident Case Men (n = 79) Women (n = 110) Overall (n = 189) Attributable Fraction 95% CI Attributable Fraction 95% CI Attributable Fraction 95% CI ART naïve, HIV seroprevalent ART naïve, HIV incident Household 21.0% 17.7% 22.8% 15.1% 14.6% 16.4% 18.1% 16.4% 19.1% Household 4.1% 1.3% 7.6% 5.2% 2.7% 7.3% 4.7% 4.2% 4.8% Missing HIV status Household 15.8% 10.1% 21.5% 16.6% 10.9% 20.9% 16.2% 11.6% 20.1% Missing HIV status Missing HIV status Unknown contacts/sources Extra-household, intra-community Extra-household, extra-community Unknown location 21.7% 15.2% 27.9% 10.2% 6.4% 13.6% 15.0% 11.1% 18.5% 16.1% 7.8% 22.3% 30.6% 27.3% 33.6% 24.6% 20.1% 28.0% 21.5% 12.7% 32.9% 21.4% 14.6% 30.0% 21.4% 14.8% 29.6% Household total a 39.0% 32.3% 43.9% Extra-household 39.5% 33.9% 42.3% total b a Estimate includes infections attributable to ART-naïve HIV-prevalent and -incident cases and household partners with missing HIV status. b Estimate includes extra-household, intra-community, and extra-community partners. 102

125 Table 5: Probability of HIV infection by partner type over 18-mo study interval. Partner Type Probability of HIV- Infection 95% CI ART-naïve, HIV-incident partner a 26.0% 13.4% 43.0% ART-naïve, HIV-seroprevalent partner a 15.3% 10.9% 20.6% Household partner with unknown HIV serostatus 1.1% 0.7% 1.7% Community partner with unknown HIV serostatus 1.3% 0.8% 2.0% Extra-community partner with unknown HIV serostatus, for women Extra-community partner with unknown HIV serostatus, for men 4.2% 3.0% 5.9% 0.9% 0.3% 1.8% Unknown contacts/undisclosed partners 0.3% 0.2% 0.5% a 99% of partnerships were intra-household. 103

126 Figure 1. Rakai District, Uganda. (A) Rakai (~2,200 km 2 ), a rural district in southwest Uganda, with population ~450,000 (~700 communities). RCCS R13 study participants (n = 1,085) reported 1,169 sexual partners with primary residence outside the Rakai District, but within Uganda (where disclosed, residential locations of sexual partners are indicated with red dots on the map). Only three sexual partners were reported to be living outside Uganda (two in Tanzania and one in the United Kingdom, not shown). (B) The Rakai district at a higher resolution, with the 11 geographic regions surveyed in RCCS R13 indicated in color. There are two primary highways (Masaka Road to Tanzania and the Trans-African National Highway to Rwanda and the Democratic Republic of the Congo [DR of Congo]) and numerous secondary roads that extend throughout the district. 104

127 Figure 2. Spatial clustering of HIV-seropositive persons within households (0 km) and in geographic windows of 250 m up to 10 km (the first window is m and windows are centered every 50 m starting at 125 m). Spatial clustering analyses show whether HIV prevalence or incidence is elevated within certain distances of other HIV-seropositive persons. We define the spatial clustering of HIV-seropositive individuals as τ(d 1,d 2 ), the relative probability that an HIV-seropositive person resides within a distance window, d 1 to d 2, from another HIV-seropositive person compared to the probability that any individual is HIV seropositive in the entire study population. Where spatial clustering exists, values of τ(d 1,d 2 ) exceed one. Shaded areas show the 95% bootstrapped confidence intervals for spatial clustering estimates. (A) The spatial clustering between HIV-seropositive persons (prevalent or incident cases with other prevalent or incident cases; red). (B) The spatial clustering of HIV-seroincident cases with ART-naïve HIV-seroprevalent persons (yellow). (C) The spatial clustering of HIV-seroincident cases with other HIV-seroincident cases (blue). (D) A blowup of the area where significant extra-household spatial clustering (<500 m) was identified among all HIV-seropositive persons (marked with black box in [A C]). Data are shown only up to 10 km (no significant spatial clustering was observed beyond this distance). 105

128 Figure 3. Maximum likelihood phylogenetic analyses of the HIV-1 gag gene. (A) Boxplots of the intra-subtype gag genetic pairwise distances for epidemiologically linked (Epi linked) incident couples (i.e., at least one member of the couple was an incident case) and for all epidemiologically unlinked incident pairs of individuals in RCCS R13. (B) Boxplots of intra-subtype gag genetic pairwise distances by the geographic distance between the incident pair. (C) A ML phylogenetic tree (radial) of HIV-1 subtype A gag sequences from HIV-seroprevalent (n = 245) and HIV-incident (n = 55) cases, where taxa are colored by the geographic region from which they were isolated. Reference strains (n = 87) are in black. Grey circles indicate nodes with bootstrap support of 70%; black circles indicate intra-household clusters; indicates an intra-household virus also sharing a cluster with at least one other household. Additional radial and rectangular phylogenetic trees for HIV-1 subtypes A, D, and C for gag and env genes are included in Figures S5 S

129 Figure 4. Summary of inferential methods and study results and conclusions. The dotted blue line represents the border of a hypothetical community. 107

130 Table S1. Accession numbers for HIVDB reference sequences used for maximum likelihood and Bayesian phylogenetic analyses. Accession Country Year HIV-1 Subtype Gene region AB UG A1 gag AB RW 1992 A1 gag AB UG 1992 A1 gag AB RW 1992 A1 gag AB UG A1 gag AB UG 1991 D gag AF KE 1994 A1 gag AF KE 1993 D gag AF ZA 1998 C gag AF BW 1996 C gag AF KE 2000 A1 gag AF KE 1999 A1 gag AF KE 2000 A1 gag AF KE 2000 A1 gag AF KE 1999 A1 gag AF KE 2000 A1 gag AF KE 2001 D gag AF UG 1999 A1 gag AF UG 1999 D gag AF UG 1999 D gag AF UG 1999 D gag AF UG 1999 D gag AF UG 1999 D gag AF UG 1999 A1 gag AF UG 1999 D gag AF UG 1999 D gag AF UG 1999 D gag AF UG 1999 D gag AF UG 1998 D gag AF UG 1998 A1 gag AF UG 1998 D gag AF UG 1998 D gag AF UG 1998 D gag AF UG 1999 D gag AF UG 1998 D gag AF UG 1999 D gag AY GA 1997 D gag AY TZ 2001 C gag 108

131 AY UG 1999 D gag AY KE 1986 A1 gag AY SN 1996 A gag AY SN 2001 A gag AY SN 1996 A1 gag AY KE 1997 D gag AY KE 1997 D gag AY KE 1998 A gag AY KE 1998 A gag AY KE 1999 A gag AY KE 1998 A gag AY UG 2001 A1 gag AY UG 2003 D gag AY UG 2003 D gag AY UG 2003 A1 gag AY UG 2003 D gag AY UG 2003 D gag AY UG 2003 D gag AY UG 2003 D gag AY UG 2003 A1 gag AY UG 2003 A1 gag AY UG 2003 A1 gag AY UG 2003 A1 gag AY UG 2004 D gag AY UG 2002 A1 gag AY UG 2002 A1 gag DQ KE 1999 A1 gag DQ KE 1999 A1 gag DQ KE 1999 A1 gag DQ KE 1999 A1 gag DQ KE 1999 A1 gag DQ KE 1999 A1 gag DQ KE 1999 D gag DQ KE 1999 A1 gag DQ KE 1999 A1 gag DQ KE 1999 A1 gag DQ KE 1999 A1 gag DQ KE 1999 A1 gag DQ ZM 2001 C gag EF UG 2005 D gag EF UG 2005 A1 gag 109

132 EF UG 2005 D gag EF UG 2005 D gag EF UG 2005 A1 gag EF UG 2005 D gag EF UG 2005 A1 gag EF UG 2005 C gag EF UG 2005 A1 gag EF UG 2005 A1 gag EF UG 2005 A1 gag EF UG 2005 D gag EF UG 2005 A1 gag EF UG 2005 D gag EF UG 2005 D gag EF UG 2005 A1 gag EF UG 2005 D gag EF UG 2005 A1 gag EF UG 2005 A1 gag EF UG 2005 A1 gag EU KE 2002 A1 gag FJ ZM 2005 C gag FJ ZM 1999 C gag FJ ZM 2005 C gag FJ ZM 2005 C gag FJ ZM 2005 C gag FJ ZM 2005 C gag FJ ZM 2000 C gag FJ ZM 1999 C gag FJ ZM 2006 C gag FJ ZM 2003 C gag FJ ZM 2003 C gag FJ ZM 2006 C gag FJ ZM 2005 C gag FJ ZM 2005 C gag FJ ZM 2005 C gag FJ ZM 2005 C gag FJ KE 2006 A1 gag FJ TZ 2003 C gag FJ TZ 2003 A1 gag FJ TZ 2003 A1 gag FJ TZ 2003 C gag FJ TZ 2003 A1 gag 110

133 FJ TZ 2003 A1 gag FJ TZ 2003 D gag GQ KE 1996 A1 gag GQ KE 1995 A1 gag GQ KE 2000 D gag GQ KE 1987 C gag GQ KE 1987 C gag GQ KE 1987 D gag GQ KE 1992 A1 gag GQ KE 1988 D gag GQ KE 2001 A1 gag GQ KE 1995 A1 gag GQ KE 1995 A1 gag GQ KE 1994 A1 gag GQ KE 1996 A1 gag GQ KE 1998 D gag GQ KE 1989 D gag GQ KE 2002 A1 gag GQ KE 1995 A1 gag GQ KE 1995 D gag GQ KE 1995 D gag GQ KE 1995 C gag GQ KE 1996 A1 gag GQ KE 1995 A1 gag GQ KE 1987 A1 gag GQ KE 1987 A1 gag GQ KE 1995 A1 gag GQ KE 1995 A1 gag GQ KE 1996 A1 gag GQ KE 1998 A1 gag HM ZA 2004 C gag HM ZA 2004 C gag HM ZA 2003 C gag HM ZA 2005 C gag HM ZA 2004 C gag HM ZA 2005 C gag HM ZA 2005 C gag HM ZA 2005 C gag HM ZA 2005 C gag HM ZA 2006 C gag HM ZA 2006 C gag 111

134 HQ UG 2008 A1 gag HQ UG 2008 A1 gag HQ UG 2008 D gag HQ UG 2008 D gag HQ UG 2008 D gag HQ UG 2008 D gag HQ UG 2008 D gag HQ UG 2008 A1 gag HQ UG 2008 D gag L11768 KE A gag L11770 KE A gag L11771 KE D gag L11773 KE A gag L11774 KE 1990 A gag L11775 KE A gag L11784 CD D gag L11787 CD C gag L11788 BI A gag L11801 UG 1991 D gag M62320 UG 1985 A1 gag U88824 UG 1994 D gag AB UG 1992 A1 env AB RW 1992 A1 env AB RW 1993 A1 env AB UG A1 env AB SN 1990 D env AF KE 1994 A1 env AF BW 1996 C env AF KE 1993 D env AF ZM 1996 C env AF TZ 1998 C env AF TZ 1998 C env AF TZ 1997 A1 env AF TZ 1997 C env AF ZA 1998 C env AF ZA 1998 C env AF KE 1995 A1 env AF KE 1995 A1 env AF KE 1994 A1 env AF BW 2000 C env AF BW 2000 C env 112

135 AF KE 2000 A1 env AF KE 2000 A1 env AF KE 1999 A1 env AF KE 2000 A1 env AF KE 2000 A1 env AF KE 2000 A1 env AF KE 1999 A1 env AF KE 2000 A1 env AF KE 2000 A1 env AF KE 2000 A1 env AF KE 2000 A1 env AF UG 1999 D env AF UG 1999 A1 env AF UG 1999 D env AF UG 1999 D env AF UG 1999 A1 env AF UG 1999 D env AF UG 1999 D env AF UG 1999 D env AF UG 1999 D env AF UG 1998 D env AF UG 1998 D env AF UG 1998 D env AF UG 1998 A1 env AF UG 1998 D env AF UG 1999 D env AF UG 1998 D env AM CM 1997 A env AM CM 1999 G env AY KE A env AY KE A env AY KE A env AY TZ 2001 A1 env AY TZ 2001 D env AY TZ 2001 A1 env AY KE 1993 A1 env AY KE 1986 A1 env AY ZA 2000 C env AY ZA 2001 C env AY UG 1994 D env AY ZA 1999 C env 113

136 AY ZA 1999 C env AY UG 1992 D env AY RW 1992 A1 env AY UG 1992 A1 env AY RW 1992 A1 env AY UG 1993 A1 env AY UG 1994 A1 env AY RW 1992 A1 env AY UG 1993 D env AY UG 1992 D env AY UG 1993 D env AY UG 1992 D env AY UG 1992 D env AY UG 1992 D env AY RW 1993 A1 env AY UG 1993 D env AY KE 2000 A env AY TZ 2000 A env AY ZA 2004 C env AY KE 1991 C env DQ ZA 2004 C env DQ ZA 2004 C env DQ ZA 2004 C env DQ KE A env DQ KE A env DQ KE A env DQ KE A env DQ KE A env DQ KE A env DQ ZA 2003 C env EF UG 1996 D env EF UG 1996 A1 env EF ZA 2005 C env EF UG D env EF UG D env EF UG D env EF UG A1 env EF UG A1 env EF UG A1 env EF UG A1 env EF UG A1 env 114

137 EF UG A1 env EF UG A1 env EF UG A1 env EF UG A1 env EF CM 1999 A1 env EF UG D env EF UG D env EF UG D env EF UG D env EF UG D env EF UG D env EF CM 1996 D env EF CM 1998 D env EU KE 2001 A1 env EU KE 2002 A1 env EU ZM 2003 C env EU ZM 2003 C env EU UG 1997 D env EU UG 1997 D env EU CM 2005 A1 env EU CM 2002 A1 env EU CM 2003 A1 env EU CM 2003 A1 env EU CM 2003 A1 env EU CM 2003 A1 env EU CM 2003 A1 env EU CM 2004 A1 env EU CM 2003 A1 env EU CM 2003 A1 env EU CM 2004 A1 env EU CM 2004 A1 env EU CM 2004 A1 env EU CM 2003 A1 env EU CM 2004 A1 env EU CM 2004 A1 env EU CM 2002 A1 env EU CM 2002 A1 env EU CM 2002 C env EU CM 2003 A1 env EU CM 2004 A1 env EU CM 2003 A1 env 115

138 EU CM 2004 A1 env EU CM 2002 A1 env EU CM 2004 A1 env EU CM 2001 A1 env EU CM 2001 A1 env EU CM 2001 A1 env EU CM 2001 A1 env EU CM 2001 A1 env EU CM 2001 A1 env EU CM 2001 A1 env EU UG 1997 D env EU UG 1997 D env EU UG 1998 A1 env EU UG 1997 D env EU UG 1997 D env EU UG 1997 D env EU UG 1997 D env EU UG 1997 D env EU UG 1999 A1 env EU UG 1999 A1 env EU UG 1997 D env EU UG 1997 D env EU UG 1997 A1 env EU UG 1997 A1 env EU UG 1998 D env EU UG 1998 D env EU UG 1998 D env EU UG 2001 D env EU UG 2001 D env EU UG 1997 D env EU UG 1997 D env EU ZA 2000 C env EU ZA 2000 C env FJ KE 2007 D env FJ KE 2007 A1 env FJ KE 2007 D env FJ KE 2007 A1 env FJ KE 2007 D env FJ KE 2007 D env FJ KE 2007 A1 env FJ KE 2007 A1 env 116

139 FJ KE 2007 A1 env FJ KE 2007 A1 env FJ KE 1998 A1 env FJ ZA 2007 C env FJ KE 2006 A1 env FJ KE 2006 A1 env FJ KE 2006 A1 env FJ KE 2006 A1 env FJ ZA 2001 A1 env FJ KE 1993 D env FJ KE 1997 D env GQ UG 1996 D env GU ZA C env GU ZM 2005 C env GU ZM 2005 C env HM TZ 2008 A env HM TZ 2008 C env HM UG 2007 D env HM UG 2007 D env HM UG 2006 D env HM UG 2007 A env HM KE 2007 A env HM TZ 2004 D env HM TZ 2002 A env HQ ZA 2005 C env HQ TZ 2008 C env L07082 RW A env L22943 KE 1990 A env L22948 UG 1990 C env L22950 UG 1990 D env M66533 RW A env U36867 UG D env U36884 UG D env U36886 UG D env U39239 BI 1991 C env U39241 BI 1991 C env U39245 BI 1991 C env U46016 ET 1986 C env 117

140 Table S2. Summary of HIV sequences from 1,434 HIV-1-seropositive participants in RCCS R13. Table includes the HIV-1 group M subtype assignment of isolated viruses in gag and env genes. gag Subtype (N=915) A C D R* NA** Total env Subtype (N=1026) A C D G R* NA** Total *** There were 1099 (77%) participants with sequence information in one or both genetic regions. Of those participants, 842 (59%) participants had viral sequence data for both gag and env genes and are highlighted in grey. *Recombinant viral sequence **Virus not amplifiable in gag/env gene region. ***Total number of ART naïve HIV-1 seropositive participants in study. 118

141 Table S3. Summary of HIV sequences obtained from 189 HIV-1- incident participants in RCCS R13. Table includes the HIV-1 group M subtype assignment of isolated viruses in gag and env genes. gag Subtype (N=139) A C D R* NA** Total env Subtype (N=153) A C D R* NA** Total *** There were 164 (87%) participants with sequence information in one or both genetic regions. Of those participants, 128 (68%) participants had viral sequence data for both gag and env genes and are highlighted in grey. *Recombinant viral sequence. **Virus not amplifiable in gag/env gene region. ***Total number of ART naïve HIV-1 incident cases in study. One participant was on ART at the time of study visit and was excluded from the viral sequence analysis. 119

142 Table S4. Summary phylogenetic data (HIV subtype, genetic pairwise distance, and clustering) for 105 epidemiologically-linked incident couples with phylogenetic data in either or both the gag and env gene regions. HIV-1 serostatus (Partner 1:Partner 2) HIV-1 gag subtype (Partner 1:Partner 2) gag genetic distance Phylogenetically clusters in gag HIV-1 env subtype (Partner 1:Partner 2) env genetic distance Phylogenetically clusters in env Incident:Incident D:D yes A:C no Incident:Incident A:A yes A:A no Incident:Incident A:A yes A:A yes Incident:Incident A:A yes A:A yes Incident:Incident D:D yes D:D yes Incident:Incident A:A yes A:A yes Incident:Incident A:A no Incident:Prevalent D:D yes Incident:Prevalent A:A no A:A yes Incident:Prevalent D:D yes A:A yes Incident:Prevalent A:A yes Incident:Prevalent D:D yes D:D yes Incident:Prevalent D:D yes D:D yes Incident:Prevalent A:A yes A:A yes Incident:Prevalent A:A yes Incident:Prevalent D:D yes C:C yes Incident:Prevalent A:A no Incident:Prevalent D:D no Incident:Prevalent D:D yes Incident:Prevalent A:A yes Prevalent:Incident A:A no A:A no Prevalent:Incident A:A no A:A no 120

143 Prevalent:Incident A:A no A:A yes Prevalent:Incident D:D yes D:D no Prevalent:Incident D:D yes A:A yes Prevalent:Incident D:D no A:D no Prevalent:Incident D:D yes A:A yes Prevalent:Incident D:D no D:D no Prevalent:Incident D:D no D:D yes Prevalent:Incident D:D no Prevalent:Incident D:D no Prevalent:Incident A:A yes Prevalent:Prevalent A:A yes Prevalent:Prevalent A:A no A:A no Prevalent:Prevalent D:D no D:D no Prevalent:Prevalent A:A yes A:A yes Prevalent:Prevalent A:A no A:A no Prevalent:Prevalent D:D no Prevalent:Prevalent A:C no A:C no Prevalent:Prevalent A:A yes A:A no Prevalent:Prevalent D:D no Prevalent:Prevalent D:D no D:D no Prevalent:Prevalent D:D no A:A no Prevalent:Prevalent D:D yes D:D no Prevalent:Prevalent D:D yes D:D yes Prevalent:Prevalent D:D no D:D yes Prevalent:Prevalent A:A no A:A no Prevalent:Prevalent A:A yes Prevalent:Prevalent D:D no D:D no 121

144 Prevalent:Prevalent D:D yes D:D no Prevalent:Prevalent A:D no A:A no Prevalent:Prevalent D:D no Prevalent:Prevalent D:D no D:D yes Prevalent:Prevalent D:D yes D:D no Prevalent:Prevalent D:D yes D:D yes Prevalent:Prevalent D:D no D:D no Prevalent:Prevalent A:A yes A:A no Prevalent:Prevalent D:D no Prevalent:Prevalent D:D no D:D no Prevalent:Prevalent C:C no D:D yes Prevalent:Prevalent A:D no A:D no Prevalent:Prevalent D:D yes D:D no Prevalent:Prevalent A:A yes A:A no Prevalent:Prevalent A:A yes A:A yes Prevalent:Prevalent D:D yes A:A yes Prevalent:Prevalent D:D yes D:D yes Prevalent:Prevalent A:D no A:A no Prevalent:Prevalent D:D no Prevalent:Prevalent A:A yes A:A yes Prevalent:Prevalent D:D yes D:D yes Prevalent:Prevalent A:D no A:D no Prevalent:Prevalent D:D yes D:D yes Prevalent:Prevalent A:C no A:C no Prevalent:Prevalent D:D no D:D no Prevalent:Prevalent D:D yes D:D no Prevalent:Prevalent A:A no A:A no 122

145 Prevalent:Prevalent D:D no D:D no Prevalent:Prevalent D:D no Prevalent:Prevalent D:D no Prevalent:Prevalent D:D no Prevalent:Prevalent D:D yes D:D no Prevalent:Prevalent A:A no A:A yes Prevalent:Prevalent D:D no D:D no Prevalent:Prevalent A:D no Prevalent:Prevalent A:D no A:A no Prevalent:Prevalent A:A no A:A no Prevalent:Prevalent A:A yes A:A yes Prevalent:Prevalent D:D no Prevalent:Prevalent D:D no Prevalent:Prevalent D:D yes Prevalent:Prevalent D:D no Prevalent:Prevalent D:D yes Prevalent:Prevalent A:A no Prevalent:Prevalent A:A no Prevalent:Prevalent A:A no Prevalent:Prevalent D:D yes Prevalent:Prevalent C:D no Prevalent:Prevalent A:D no Prevalent:Prevalent A:D no Prevalent:Prevalent D:D no Prevalent:Prevalent A:A no Prevalent:Prevalent D:D no Prevalent:Prevalent D:D no 123

146 Prevalent:Prevalent A:A no Prevalent:Prevalent A:A no 124

147 Table S5. Detailed summary data for each of the 95 phylogenetic clusters identified in maximum likelihood phylogenetic trees (HKY-85 model). Cluster ID HIV-1 gag subtype HIV-1 env subtype Total Participants Incident cases 125 Prevalent Cases Households Communities Geographic Regions 1 A A A A A A A A A A A A A A A A A A A A A A A A A A A D A A A A A A A A

148 24 A A A A A A A D D D D D D A D A D A D D D D D D D D D D D D D D D D A D A D A D D D D

149 51 D D D A D C D D D D D D D D D D D D D D D D C C C A A A A A A A A A A A A

150 78 - A D D D D D D D D D D D D D D D C C

151 Table S6. Sensitivity analyses of phylogenetic clustering results to choice of evolutionary model and bootstrap and genetic distance thresholds. Phylogenetic cluster analyses were conducted at 70%, 80%, 90%, and 99% bootstrap thresholds, with and without genetic distance cutoffs under the HKY-85 and GTR+I+G models of evolution. We present the cluster summary data shown in Table 2 under these different evolutionary models and genetic distance and bootstrap threshold criteria. Model of nucleotide evolution Bootstrap threshold (%) Total clusters Cluster size distribution Genetic distance threshold No. No. of Participants in cluster (frequency) Clusters containing only incident cases No. of Clusters (% of clusters) Household clusters a No. of Clusters (% of clusters) Intracommunity clusters b No. of Clusters (% of clusters) Cross community clusters c No. of Clusters (% of clusters) HKY yes 39 2 (34), 3 (4), 5 (1) 4 (10.2) 12 (30.1) 9 (23.0) 18 (46.1) 90 yes 95 2 (82), 3 (9), 4 (2), 5 (2) 6 (6.3) 42 (44.2) 15 (15.8) 38 (40.0) 80 yes (94), 3 (10), 4 (3), 5(2) 5 (4.6) 48 (44.0) 18 (16.5) 43 (39.4) 70 yes (97), 3 (10), 4 (3), 5 (2) 5 (4.4) 49 (43.8) 19 (17.0) 44 (39.2) HKY no 64 2 (54), 3 (7), 4 (1), 5 (1), 6 (1) 5 (7.8) 24 (37.5) 9 (14.0) 31 (48.4) 90 no (84), 3 (16), 4 (4), 5 (3), 6 (1) 6 (5.6) 42 (38.9) 17 (15.7) 49 (45.3) 80 no (107), 3 (22), 4 (4), 5 (3), 7 (1) 6 (4.8) 50 (36.5) 25 (18.2) 62 (45.2) 70 no (122), 3 (27), 4 (4), 5 (3), 7 (1), 9 (1) 6 (3.7) 54 (34.1) 27 (27.0) 77 (48.0) GTR+I+G 99 yes 39 2 (34), 3 (4), 5 (1) 4 (10.2) 13 (33.3) 8 (20.5) 18 (46.1) 90 yes 92 2 (80), 3 (8), 4 (2), 5 (1) 6 (6.5) 40 (44.6) 14 (15.2) 37 (40.2) 80 yes (91), 3 (9), 4 (3), 5 (2) 5 (4.7) 48 (45.8) 17 (16.2) 40 (38.1) 70 yes (95), 3 (9), 4 (3), 5 (2) 5 (4.6) 49 (45.0) 18 (16.5) 42 (38.5) GTR+I+G 99 no 40 2 (35), 3 (4), 5 (1) 4 (10.0) 13 (32.5) 8 (20.0) 19 (47.5) 90 no (82), 3 (14), 4 (4), 5 (2), 6 (1) 6 (5.8) 40 (38.8) 16 (15.5) 47 (45.6) 80 no (104), 3 (21), 4 (5), 5 (3), 7 (1) 5 (3.7) 51 (38.1) 23 (17.2) 60 (44.8) 70 no (113), 3 (27), 4 (5), 5 (4), 7 (1) 6 (4.0) 53 (35.3) 26 (17.3) 71 (47.3) a Refers to clusters of two individuals who share the same household b Refers to clusters of two more individuals who spanned households but shared the same community c Refers to clusters of two or more individuals who spanned households and communities 129

152 Table S7. Numbers of recent sexual partners self-reported by 9,520 HIV-seronegative and -incident participants in egocentric analysis by gender and marital status of the study participant. Females Unmarried-never married Unmarried-previously married Married - not polygamous Married - polygamous* Total N (%) N (%) N (%) N (%) N (%) 1 partner 573 (94.1) 755 (92.9) 2893 (97.3) 938 (96.5) 5159 (96.1) 2 partners 34 (5.6) 53 (6.5) 79 (2.7) 31 (3.2) 197 (3.7) 3 partners 2 (0.3) 5 (6.1) 2 (0.0) 3 (0.3) 12 (0.2) 4 partners 0 (0.0) 0 (0.0) 0 (0.0) 0 (0.0) 0 Total 0 (11.3) 813 (15.1) 2974 (55.4) 972 (18.1) 5368 Males Unmarried-never married Unmarried-previously married Married - not polygamous Married - polygamous Total N (%) N (%) N (%) N (%) N (%) 1 partner 655 (71.4) 211 (64.9) 1582 (64.2) 10 (2.2) 2458 (59.2) 2 partners 175 (19.1) 68 (20.9) 669 (27.2) 283 (63.3) 1195 (28.8) 3 partners 77 (8.4) 39 (12.0) 192 (7.8) 135 (29.8) 441 (10.6) 4 partners 10 (1.1) 7 (2.1) 20 (0.8) 21 (4.7) 58 (1.4) Total 917 (22.1) 325 (78.2) 2463 (59.3) 447 (10.8) 4152 * Married - polygamous for females refers to a female in a marital relationship with a man that has multiple wives. 130

153 Table S8. Summary of self-reported sexual partner data from 9,520 HIV-seronegative and -incident participants in egocentric analysis by gender of the study participant and geographic location of the sexual partner. Female Sexual partners of known HIV serostatus (N=5043)* Household Community Extra-community Total Household Community Extra-community Total N (%) N (%) N (%) N (%) N (%) N (%) N (%) N (%) Stable 2346 (99.9) 82 (97.6) 11(91.2) 2439 (99.8) Stable 2483 (99.9) 87 (85.2) 11 (91.2) 2581 (99.3) Unstable 1 (0.1) 2 (2.3) 1 (8.8) 4 (0.2) Unstable 3 (0.1) 15 (14.8) 1 (8.8) 19 (0.7) Total 2347 (96.1) 84 (3.4) 12 (0.5) 2443 Total 2486 (95.6) 102 (3.9) 12 (0.5) 2600 Female Sexual partners of unknown HIV serostatus (N=6949) Household Community Extra-community Total Household Community Extra-community Total N (%) N (%) N (%) N (%) N (%) N (%) N (%) N (%) Stable 1729 (99.2) 2 (0.3) 12 (1.6) 1743 (55.4) Stable 959 (98.7) 12 (0.8) 9 (0.7) 980 (25.8) Unstable 15 (0.8) 616 (99.7) 772 (98.4) 1403 (44.6) Unstable 13 (1.3) 1526 (99.2) 1284 (99.3) 2823 (74.2) Total 1744 (55.8) 618 (19.5) 784 (24.7) 3146 Total 972 (25.7) 1538 (40.4) 1293 (33.9) 3803 *83 incident cases occurred among individuals in which the HIV-status of at least one of their reported sexual partners was known. Male Male 131

154 Table S9. Comparison of demographics and sexual behaviors (% distribution) between the RCCS study population (RCCS R13, ) and the surveyed population in the 2011 Ugandan Demographic and Health Survey (DHS) RCCS R13, Ugandan DHS, 2011 Ugandan DHS, 2011 Central Region 1* All of Uganda Females Males Females Males Females Males N=8188 N=6406 N=767 N=188 n=8674 N=2295 % (95% CI) % (95% CI) Weighted % Weighted % Weighted % Weighted % Age distribution (15-49 years old) ( ) 21.8 ( ) 24.0 ( ) 21.7 ( ) 23.6 ( ) 25.5 ( ) ( ) 15.7 ( ) 17.8 ( ) 13.6 ( ) 18.9 ( ) 14.7 ( ) ( ) 16.8 ( ) 16.3 ( ) 24.0 ( ) 18.1 ( ) 16.6 ( ) ( ) 16.2 ( ) 13.5 ( ) 15.3 ( ) 12.5 ( ) 14.9 ( ) ( ) 13.7 ( ) 11.1 ( ) 9.5 ( ) 11.8 ( ) 12.3 ( ) ( ) 9.3 ( ) 7.9 ( ) 7.5( ) 8.4 ( ) 8.8 ( ) ( ) 6.4 ( ) 9.4 ( ) 8.4 ( ) 6.7 ( ) 7.2 ( ) Education No education 6.2 ( ) 3.4 ( ) 9.2 ( ) 5.6 ( ) 12.9 ( ) 4.5 ( ) Primary 61.5 ( ) 65.7 ( ) 54.4 ( ) 65.0 ( ) 59.4 ( ) 60.0 ( ) Secondary 26.7 ( ) 24.7 ( ) 30.5 ( ) 22.7 ( ) 22.5 ( ) 26.9 ( ) Higher than secondary 5.6 ( ) 6.2 ( ) 5.9 ( ) 6.7 ( ) 5.2 ( ) 8.4 ( ) Currently married or in cohabitating union 62.7 ( ) 57.3 ( ) 58.5 ( ) 60.1 ( ) 62.5 ( ) 58.3 ( ) Polygyny (married men) ( ) ( ) ( ) No. of sex partners in the last year None 18.0 ( ) 21.4 ( ) 26.1 ( ) 25.5 ( ) 27.3 ( ) 26.7 ( ) ( ) 44.3 ( ) 70.4 ( ) 46.0 ( ) 71.0 ( ) 53.9 ( ) ( ) 24.0 ( ) 3.1 ( ) 23.7 ( ) 1.6 ( ) 16.7 ( ) ( ) 10.3 ( ) 0.4 ( ) 4.9 ( ) 0.5 ( ) 2.7 ( ) Males circumcised ( ) ( ) ( ) HIV prevalence** 14.2 ( ) 9.7 ( ) 12.5 ( ) 8.4 ( ) 8.3 ( ) 6.1 ( ) *Central Region 1 is the geographic sub-region within the Ugandan DHS that contains the Rakai District. Central region 1 also includes Masaka District to the north of Rakai. **Information obtained from 2011 DHS AIDS Indicator Survey Report Approximately 19-22% of total HIV infected persons in DHS were taking ART (18% in Central Region 1) compared to 22.0% in Rakai. 132

155 Figure S1. The geographic scale of RCCS communities. Communities are color-coded according to their RCCS geographic region (see Figure 1 for color key). The means for the average and maximum geographic distances between households within a community (across all communities) are marked with dotted red lines. The size of the dot is proportional to the size of the surveyed population/community size. 133

156 Figure S2 Phylogenetic analyses of gag and envoi genes for specimens that underwent repeated viral RNA extraction and PCR testing. Repeated viral RNA extractions and PCR testing was performed for a sample of patient specimens for gag (n = 26) (A) and env (n = 46) (B) to assess the reliability of our laboratory methods. Sequences were compared using neighbor-joining trees (1,000 bootstrap replicates). Trees were constructed separately for each gene region using a Tamura-Nei model of nucleotide substitution. Results of the phylogenetic analyses showed that the laboratory methods yielded reliable sequence information: sequences obtained from the same individual always clustered together. A B 134

157 Figure S3. Genetic pairwise distances in gag and env genes for epidemiologically linked HIV-infected couples where at least one partner was an HIV-incident case. Figures show only those incident couples who shared a monophyletic clade in a ML tree with 70% or greater bootstrap support. These distributions were used to determine the genetic distance thresholds for phylogenetic cluster analyses. Figure S4. Spatial clustering of HIV-seroprevalent persons on ART with HIVincident cases within households (0 km) and in geographic windows of 250 m up to 10 km (every 50 m beginning at 125 m). Spatial clustering, τ(d 1,d 2 ), shown in black, is the relative probability that an HIV-seroprevalent person on ART resides within a distance range, d 1 to d 2, from an incident case compared to the probability that any individual participant is an incident case. The shaded area is the bootstrapped 95% confidence interval (1,000 iterations). 135

158 Figure S5. Maximum likelihood tree (radial) of gag HIV-1 subtype D sequences. Color corresponds to geographic region (see Figure 1 key). Reference sequences (n = 57) are in black. Grey circles indicate nodes with bootstrap support of 70%; black circles indicate intra-household clusters; indicates intra-household viruses also sharing a cluster with at least one other household. 136

159 Figure S6. Maximum likelihood tree (rectangular) of gag HIV-1 subtype C sequences. Taxa are labeled using participant gender/geographic region/community/household. Reference sequences (n = 37) are in black, and only bootstrap values 50% are shown. Color corresponds to the geographic region. 137

160 Figure S7. Maximum likelihood tree (radial) of env HIV-1 subtype A sequences. Color corresponds to geographic region (see Figure 1 key). Reference sequences (n = 107) are in black. Grey circles indicate nodes with bootstrap support of 70%; black circles indicate intra-household clusters; indicates intra-household viruses also sharing a cluster with at least one other household. 138

161 Figure S8. Maximum likelihood tree (radial) of env HIV-1 subtype D sequences. Color corresponds to geographic region (see Figure 1 key). Reference sequences (n = 70) are in black. Grey circles indicate nodes with bootstrap support of 70%; black circles indicate intra-household clusters; indicates intra-household viruses also sharing a cluster with at least one other household. 139

162 Figure S9. Maximum likelihood tree (rectangular) of env HIV-1 subtype C sequences. Taxa are labeled using participant gender/geographic region/community/household. Reference sequences (n = 37) are in black, and only bootstrap values 50% are shown. Color corresponds to the geographic regions. 140

163 Chapter 5 HIV-1 transmission dynamics in stable heterosexual couples: A retrospective study of initially concordant HIV-negative couples in Rakai, Uganda Mary K. Grabowski, Justin Lessler, Fred Nalugoda, Steve Bellan, Xiangrong Kong, Robert Ssekubugo, Godfrey Kigozi, Joseph Kagaayi, Steven J. Reynolds, Aaron A.R. Tobian, Thomas C. Quinn, David Serwadda, Maria J. Wawer, Ronald H. Gray, and the Rakai Health Sciences Program 5.1. Abstract Background: HIV transmission between stable heterosexual partners accounts for a substantial proportion of new HIV infections in sub-saharan Africa. However, the dynamics of HIV introduction into initially uninfected couples are poorly understood. Methods: We followed 4,570 concordant HIV-negative couples in stable marital or consensual partnerships for 21,535 couple-years in the Rakai Community Cohort Study between 1997and Multivariate Cox proportional hazards models were used to identify demographic and behavioral factors associated with HIV introduction into the relationship as well as to assess how the HIV introduction rate changed after antiretroviral therapy (ART) and male circumcision were (MMC) were introduced and scaled up in Rakai beginning in

164 Results: We identified 135 HIV introductions into 4,570 initially concordant HIVnegative couples. In 21.5% (n=29) of incident couples, both partners seroconverted during the same follow-up interval. Among the remaining 106 incident couples, 62.3% (n=66) of infections were introduced by males, and 37.7% (n=40) infections were introduced by females. The presence of at least one self-reported extra-couple partnerships in the past year was associated with a significantly greater hazard of an HIV introduction (male only: adjhr=2.1, 95%CI: ; female only: adjhr=6.6, 95%CI: ; both partners: adjhr=4.6, 95%CI: ); however, of newly HIV-infected partners in incident discordant couples, only 22.5% of female (9/40) and 69.7% (44/66) of male partners disclosed having external sexual partnerships during the risk interval prior to HIV seroconversion. Scale up of ART and MMC services after 2004 was associated with a significantly reduced risk of HIV introduction (post vs. pre scale-up period: adjhr=0.65; 95%CI: ). Conclusions: This is the first study of HIV introduction into stable initially uninfected couples in sub-saharan Africa before and after ART and MMC scale up. Both males and females introduced infection into relationships, highlighting the need for prevention strategies targeting both genders. Couples with a history of extra-marital relationships were at elevated risk; however, underreporting of external partnerships may hinder targeted prevention. Availability of ART and MMC was associated with a substantial reduction in risk of HIV introduction. 142

165 5.2. Introduction Between 50-70% of adults aged 15 to 49 years in East Africa are in marital or long term cohabitating (i.e. stable) heterosexual partnerships 1. Couples can be either concordant HIV negative (both partners uninfected), discordant HIV positive (one partner infected), or concordant HIV positive (both partners infected). Previous studies have shown that HIV-uninfected partners in stable discordant couples are at relatively high risk of acquiring HIV compared to individuals in concordant HIV-negative partnerships or to the general population, which also includes individuals not in stable unions 2 4. For example, a recent study of HIV transmission in Rakai, Uganda estimated that uninfected individuals in stable discordant partnerships were at more than fifteen times the risk of HIV acquisition than the general population 5. This same study also found that after HIV is introduced into a couple by either the male or female partner, HIV transmission to the remaining uninfected partner occurs within one year in approximately one fifth of incident couples, though other studies estimate that this proportion is higher 6 8. There have been four prior prospective studies of HIV introduction into initially concordant HIV-negative couples 3,4,9,10, including one former study in Rakai, Uganda conducted in This prior study in Rakai followed couples over a one year period and found that female partners accounted for half of HIV introductions into stable couples; however, the number of incident events identified in this study was only six. In contrast, a study in Masaka, Uganda found men were twice as likely to bring HIV infection into stable partnerships, though this study identified only twenty-five viral introductions (seventeen male vs. eight female introductions) 3. The two remaining studies were conducted in Tanzania 10. The first study, which was conducted among male factory 143

166 workers and their wives, identified fifteen incident cases (six in men, nine in women) 10. The second Tanzanian study found that males were more likely to introduce HIV infection, though this study also had few incident events (fifteen male vs. six female introductions) 4. None of these four prior studies examined factors associated with the formation of HIV discordance other than gender Here, we followed initially concordant HIV-negative couples between 1997 and 2011 for a first incident event (i.e. HIV introduction) and identified factors associated with HIV introduction. Importantly, our calendar period allowed us to assess the impact of population-based scale-up of antiretroviral therapy (ART) and medical male circumcision (MMC), which were introduced in 2004, on the rate of HIV introduction into couples. A detailed understanding of the mechanisms that drives formation of HIV discordance in stable couples may provide new insight into HIV epidemiology and novel opportunities for disease prevention among the largest demographic of persons at risk of HIV acquisition in East Africa Methods The Rakai Community Cohort Study The Rakai Community Cohort Study (RCCS) is a population-based cohort study of HIV incidence and risk behaviors in the Rakai District, Uganda that has been described in detail elsewhere 11. It is conducted by the Rakai Health Sciences Program (RHSP), which also provides HIV treatment and prevention services to individuals living in the Rakia District independent of cohort participation. Briefly, the RCCS conducts an approximately annual census of all households in 50 communities with no age truncation. 144

167 At the time of census, information on each household member is obtained, including name, age, gender and marital status (single, monogamous, polygamous). For those individuals who are in long term consensual or marital unions at census, additional relationship data are requested and used to identify marital or long-term stable partners who also participate in the RCCS. The RCCS survey, conducted after the census, includes all consenting residents aged years. Study participants are administered a detailed questionnaire by samesex interviewers in the local language, Luganda. Detailed information is obtained on demographics, sexual and health care seeking behaviors, and symptoms of sexually transmitted infections. Data are also obtained on recent sexual partners (i.e. egocentric network data). Individuals are asked only to disclose the names of sexual partners if the partnership is a marital or long-term consensual union. Biological specimens are collected for HIV and STI testing. Approximately 90% of eligible RCCS community residents present at time of survey agree to enroll and be interviewed, and of these over 95% provide biologic samples Identification of epidemiologically linked couples and primary study outcomes RCCS census and survey data were used to identify stable heterosexual couples that participated in the RCCS between June of 1997 and February of 2011, the period during which detailed egocentric network data were obtained at baseline and follow-up study visits. We defined two RCCS participants as in an epidemiologically linked couple if at a given visit individuals disclosed each other as marital or long-term consensual partners. Only those visits where individuals were in a stable couple together and both 145

168 partners participated in the RCCS survey were included in our analysis. We define the first visit at which individuals were first identified as a couple as the baseline couplevisit, though this was not necessarily the first visit that the male or female was observed within the RCCS. Sexual partners in epidemiologically linked couples are herein referred to as the index partners of one another. HIV serostatus was assessed by two enzyme immunoassays (Vironostika HIV-1, BioMerieux Inc., Charlotte, NC and Recombigen, Cambridge Biotech, Worcester, Massachusetts) with Western blot confirmation of discordant EIAs and all HIV incident cases (HIV-1 WB, BioMerieux-Vitek, St. Louis Missouri). Individuals were defined as HIV incident if they were HIV seropositive at a study visit and HIV seronegative at the prior RCCS survey, allowing for only one missed study visit. Couples who were concordant HIV-negative at their baseline couple-visit and who had at least one consecutive visit (i.e. two visits as a couple with no more than one missed visit in between visits) comprised our HIV incidence couples cohort. These couples were followed until one or both partners were lost to follow-up, the couple was administratively censored in 2011, or a first HIV incident event (i.e. an HIV introduction) occurred within the couple. We further denote a definitive male HIV introduction as a visit where the male partner was HIV incident and his female partner remained HIV seronegative and vice versa. HIV introductions in which both partners seroconverted during the same visit interval were considered non-definitively male or female (since the introducing partner could not be determined). Our primary study outcomes included HIV introduction into an epidemiologically linked couple by (1) the male or female index partner, (2) the male index partner only, and (3) the female index partner only. 146

169 Primary study exposures The predominate mode of HIV acquisition in East Africa is through sexual intercourse with an HIV infected partner, and so self-report of one or more extra-couple sexual partnerships by either or both index male and female partners at study visits was included as primary exposure variables in our analysis. At each study visit, participants were asked to disclose up to four stable or non-stable partners with whom they had sexual relations in the last twelve months, including their index stable partner (up to three extracouple partner total). In this analysis, extra-couple partnerships were considered as either stable or non-stable sexual partnerships, where an extra-couple stable partnership was defined as a cohabitating marital or long-term consensual union with an individual other than the primary index partner (i.e. polygamy) and a non-stable extra-couple partnership was defined as a non-cohabitating boyfriend/girlfriend or casual sexual partnership. Exposure variables analyzed included self-report of stable and non-stable extra-couple partnerships at the current/ first HIV positive study visit (T S ) and history of self-reported extra-couple partnerships at prior study visits (T 0, S-1 ). Primary exposures variables also included risk factors that directly reduce or augment the probability of HIV acquisition once individuals engage in sexual relations with HIV-infected persons. Specifically, we examined whether or not self-reported male circumcision status, female pregnancy and use of injectable contraceptives, and symptoms of genital ulcer in one or both partners were associated with risk of HIV introduction. In order to ensure that these exposures preceded the HIV introduction outcome, these variables were lagged by one visit (T S-1 ). 147

170 Risk of HIV introduction into couples also is determined by the probability that individuals engaging in extra-couple relations have infectious sexual encounters. Since HIV treatment and prevention services were substantially increased in the Rakai, District over our analysis period and these factors reduce the probability of having an infectious contact 12,13, we examined the association between HIV introduction and scale-up of antiretroviral therapies (ART) in HIV-infected persons and medical male circumcision (MMC) in non-muslin males using data from the entire RCCS population. ART was introduced into the Rakai District in 2004 through the Rakai Health Science Program, and indicated for HIV-infected individuals with CD4 counts 350 cell/μl or those who were otherwise clinically indicated. During this same period, MMC was also actively promoted and offered through the Rakai Health Sciences Program, at first through a clinical trial from and from then on to the general population. Data on individual ART status was linked to the full RCCS cohort through RHSP and was used to define two distinct calendar periods where couples were either unexposed or exposed to ART and MMC scale-up (i.e. pre vs. post ART and MMC scale-up period analysis). We also analyzed whether increasing proportions of HIV-infected individuals on ART and non-muslim males circumcised within a couple s local community affected risk of HIV introduction into couples Statistical Analyses Our primary unit of analysis was the couple, each consisting of one man and woman, and the primary outcome an HIV introduction into an initially concordant HIVnegative couple (by either partner or male or female partner). We did not exclude 148

171 polygamous males (17.1% of the incidence cohort); therefore, a male participant could have contributed multiple observations at the same study visit if he named multiple female RCCS participants as a stable partner at a given study visit. We also allowed monogamous individuals who had one stable partnership dissolve to re-enter into the study as part of another stable partnership at later visits. We first analyzed the probability of self-reporting non-stable extra-couple partnerships among men and women by couple demographics. Self-reported extra-couple data at both baseline and follow-up couple visits were used for gender-stratified analyses. Only non-stable sexual partnerships that were concurrent with the couple were considered extra-couple partnerships at baseline (i.e. partnerships occurring prior to couple formation were excluded), where self-reported start dates of sexual partnerships were used to determine partnership overlap and partnership duration at baseline visits. Start dates for epidemiologically linked couples were estimated as the median of all selfreported start dates by both partners at all couple-visits. Partnership duration at each visit was estimated as the time from the median partnership start date to the visit date. Poisson regression models with generalized estimating equations and robust variance estimators were used to quantify associations between time-varying couple demographics (i.e. male age, female age, partnership duration, type of marital union) and self-report of one or more non-stable extra-couple partnerships at a study visit using prevalence risk ratios (PRR). The probability of self-reporting a non-stable partnership was also estimated as a continuous function of age using generalized additive binomial models for males and females separately to detect any nonlinearity in the effect of age on the log-odds of self-reporting one or more extra-couple partnerships. 149

172 Next, we used Poisson regression models to estimate the incidence of (1) any HIV introduction by a male or female partner (2) HIV introduction by a male partner, and (3) HIV introduction by a female partner within four age groups (15-19, 20-29, 30-39, and 40+ years) for each gender. Gender-specific incidence rates of introduction were estimated using a weighted Poisson regression model, where non-definitive introduction events were weighted by the proportion of definitive introductions due to that gender and definitive introductions due to the opposite sex were censored. Cox proportional hazard models with robust variance and entry and exit time were used to estimate univariate and multivariate hazard ratios (HR) of primary exposures with HIV introduction outcomes. Weights were applied when outcomes were gender-specific introduction events as was done for Poisson models. Adjusted models included selfreport of extra-partnerships by either or both partners (T S ), circumcision status of male index partner (T S-1 ), self-report of genital ulcer by either or both partners (T S-1 ), male and female age, duration of partnership (T S ), and an indicator variable for whether or not the visit occurred before or after ART was introduced into the Rakai District (T S ). We found no association between female pregnancy and self-reported injectable contraceptive usage with HIV introduction in unadjusted or adjusted analyses, so these variables were not included in our final analysis. The proportional hazards assumption was verified using log-log plots of the scaled Schoenfeld residuals. Lastly, we examined whether specific characteristics of self-reported non-stable extra-couple partnerships were associated with HIV introduction using a nested casecontrol design, where partnerships of cases (incident males or females) were matched to partnerships of controls (HIV negative males or females) on age (+/- two years) and 150

173 survey visit at a 5:1 ratio. Analyses were conducted independently for each gender. Case partnerships selected included any non-stable extra-couple partnership reported at either the T S or T S-1 visits by an incident male or female in an epidemiologically linked couple. Conditional logistic regression was used to quantify factors associated with infectious partnerships. Conditional odds ratios are only shown for males since there were too few female partnerships disclosed to identify significant associations between partnership characteristics and incident infection Results We identified 10,597 epidemiologically linked couples consisting of 9,920 men and 10,454 women from 28,853 men and women who participated in the RCCS and selfreported being in marital or long term consensual unions. At baseline, 8,627 (81.4%) epidemiologically linked couples were concordant HIV-negative and 853 (8.1%) couples were concordant HIV-positive. The remaining 1,117 couples (10.5%) were discordant HIV-positive, where in 51.7% (n=577) of discordant couples the male was HIV positive and in 48.3% (n=540) of couples the female was HIV positive. Of the 8,627 concordant HIV-negative couples identified, 780 (9.0%) couples had their first couple-visit at the last RCCS visit included in the analysis period and were censored. Among the remaining concordant HIV-negative couples with opportunity for follow-up, 4,570 couples (57.1%) had two or more study visits (including baseline) and comprised our incidence couples cohort. These couples, including 4,328 men and 4,555 females, contributed an average of 2.7 follow-up study visits over a total of 21,535 couple-years. Concordant HIV-negative couples without the requisite follow-up visits 151

174 (n=3,307) had similar demographic and behavioral characteristics to couples in our incidence cohort, though those couples lost to follow-up after baseline had higher levels of self-reported non-stable partnerships (males: 29.4% vs. 27.1%; females: 4.2 vs. 2.8%) and genital ulcer disease (males: 12.5 vs. 10.6%; females: 12.7 vs. 11.1%) (Table S1). At baseline, the median ages of male and female partners in the incidence couples cohort was 29 (IQR: 25-36) and 23 (IQR: 20-29) years, respectively. The median age difference between male and female partners was 5.0 years (IQR: 3-8), and the median duration of partnerships was 4.1 years (IQR: ). There were 750 (17.3%) men who self-reported more than one stable sexual partner (i.e. polygamy) at baseline, of whom two or more partners were observed for 224 of these men. No women reported have more than one stable partner extra-couple sexual partner concurrently with their index partner Self-report of extra-couple sexual partnerships by age and gender Males more frequently self-reported having one or more non-stable extra-couple sexual partners in the last 12 months than their index female partners. Men self-reported a non-stable partnership at 27.1% of baseline (1238/4,570) and 29.2% (3596/12319) of follow-up visits, whereas women self-reported non-stable partnerships at only 2.8% (126/4570) of baseline and 2.1% (272/12319) of follow-up visits. At visits where a male partner self-reported non-stable extra-couple partnerships (4,834/16,889 or 28.6%), they reported one non-stable partner at 74.4% (3,647/4,834) of visits, two non-stable partners at 17.9% (864/4834) of visits, and three non-stable partners at 6.7% (323/4834) of visits; men reporting three non-stable partners could have had more partners since the RCCS 152

175 egocentric network is truncated at 4 sexual partners. Amongst women who self-reported a non-stable extra-couple sexual partnership, none reported more than one partner at a single study visit. The probability of males self-reporting a non-stable partner increased rapidly from 15 years of age, though after peaking in the mid-twenties, it declined precipitously (Figure 1). In contrast, the probability of a female self-reporting a non-stable partner peaked at earlier ages relative to their male counterparts with highest probability of disclosure among the youngest female participants aged years. At all ages the probability that a male self-reported having a non-stable extra-couple partner exceeded that for a female of the same age (Figure 1, Table 1). Other factors besides male and female age were associated with self-report of non-stable extra-couple partnerships at baseline and follow-up study visits (Table 1). Polygamous males were 33% less likely to self-report a non-stable partnership than men who were not in polygamous unions (adjprr=0.67, 95%CI: ). While males with younger index female partner were less likely to self-report non-stable partnerships (5-10 years younger: adjprr=0.90, 95%CI: ; 10 years younger: adjprr= 0.64, 95%CI: ), females with a younger index male partner were more likely to self-report non-stable partnerships (adjprr=1.77, 95%CI: ). Both male and female self-reports of non-stable extra-couple partnerships were more common if his or her index partner also self-reported a non-stable partnership at the same study visit (males: adjprr=1.18, 95%CI: ; females: adjprr=1.45, 95%CI: ). 153

176 Incidence of HIV introduction into initially concordant HIV-negative stable couples We identified 135 HIV introduction events in our incidence couples cohort. There were 106 HIV introductions for which the introducing partner could be determined (i.e. incident HIV discordant couples). In 66 (62%) of these incident discordant couples, the male introduced HIV into the partnership, wherein the other 40 (38%) incident discordant couples the female was the introducing partner. In the remaining 29 incident couples (21.2%), the male and female were HIV incident at the same follow-up visit, so the introducing partner could not be determined. Overall, the incidence of any HIV introduction by either a male or female index partner into the couple was 0.62/100 couple-years (95%CI: ), with significantly higher incidence of male introduction (0.39 per 100 couple-years; 95%CI: ) than female introduction (0.24 per 100 couple-years, 95%CI: ). Age-specific incidence of male and female HIV introductions was consistent with the gender and agespecific patterns of self-reported non-stable extra-couple partnerships (Fig 1A-D). After HIV was introduced into a couple, the non-introducing partner was at substantially higher risk for infection compared to visits prior to introduction. Assuming HIV acquisition occurs at the midpoint of the interval at risk (where the interval is again halved after introduction), subsequent transmission of HIV to the non-introducing partner within the same visit interval as his or her newly infected partner was 30.2 per 100 couple-years (29/95.74 couple-years; 95%CI: ). 154

177 Self-report of extra-couple sexual partnerships among introducing partners Self-report of having one or more non-stable extra-couple partnerships in the last twelve months at the seroconversion visit (T S ) was strongly associated with HIV introductions outcomes, particularly when index female partner self-reported extracouple partners (Table 2). Relative to couples in which neither partner reported a nonstable extra-couple partnership, the hazard of introduction was 2.29 times greater when only the male self-reported a non-stable partnership (95%CI: ), 6.91 times greater when only the female self-reported a non-stable partnership (95%CI: ), and 5.16 times greater when both partners self-reported reported non-stable partnerships (95%CI: ). Self-report of a non-stable extra-couple partnership by the female index partner was more strongly associated with a female HIV introduction than was the self-report of a non-stable partnership by a male associated with a male HIV introduction (females: adjhr=12.4; 95% CI: vs. males: adjhr=2.59; 95% CI: ). While males self-reporting additional stable sexual partners (i.e. polygamy) at the T S visit were at greater risk for introducing virus compared to men who did not report such partnerships (adjhr=1.73; 95%CI: ), the overall hazard of any HIV introduction, male or female, was not significantly greater in couples with polygamous males (adjhr=1.27 ; 95%CI: ). A prior history of self-reporting non-stable extra-couple partnership was also associated with significantly increased risk for HIV introduction (Table 2). Despite the strength of these associations, there was indication that both genders were underreporting extra-couple sexual partnerships. Among introducing women in incident discordant HIV-positive couples (n=40), only 22.5% of women (9/40) self- 155

178 reported an extra-couple partnership at the T S visit. Even when considering self-reporting at both the T S and T S-1 visits, rates of self-reporting extra-couple partnerships remained low: only 27.5% (n=11/40) of introducing women at either of these visits reported nonstable partnerships. Only 8/40 (20.0%) introducing females had any prior history of a self-reported non-stable partnership prior to their first positive visit of whom six also selfreported partnerships at the T S visit. Introducing males in incident discordant HIV-positive couples also underreported extra-couple sexual partnerships, though did so to a lesser extent than introducing females. At the T S visit only 69.7% (n=46/66) of introducing men in incident discordant couples self-reported stable or non-stable extra-couple sexual partnerships; 29/66 (43.9%) reported only non-stable partners, 11/66 (16.7%) reported only stable partners, and 6/66 (9.1%) reported non-stable and stable partners. The reporting rate was higher when considering both the T S and T S-1 visits: 83% (n=55/66) of introducing men self-reported a stable or non-stable either partner at either of these visits; 32/66 (48.5%) reported only non-stable partners, 8/66 (12.1%) reported only stable partners and 15/66 (22.7%) reported non-stable and stable partners. The characteristics of non-stable extra-couple partners self-reported at the T S or T S-1 visits by incident men and women were compared to the characteristics of non-stable partnerships self-reported by HIV-negative men and women (Table S2). No statistically significant differences were observed between the characteristics of female case and control extra-couple partners; however, consistent condom use with an extra-couple partner was significantly less common within male incident partnerships than control partnerships (OR matched =0.53; 95%CI: ). We found no significant differences in 156

179 the age differentials between index partnerships and extra-couple non-stable partnerships by gender or by HIV status (Figure S1) Genital ulcer disease and male circumcision as risk factors for HIV introduction Self-report of genital ulcer disease at the visit prior to HIV seroconversion was significantly associated with HIV introduction, but only when it was self-reported by the male and female index partners at the same visit (adjhr=4.64; 95%CI: ). This synergistic effect was observed regardless of whether virus was introduced by the male or female partner (Table 2). When the index male partner was circumcised, couples had a reduced risk of HIV introduction from men (adjhr=0.35; 95%CI: ); however, male circumcision did not significantly affect risk of introduction from female partners (adjhr=1.24; 95%CI: ) Scale-up of HIV treatment and prevention services and risk of HIV introduction ART was introduced into Rakai in June 2004 with a substantial increase in treatment uptake among HIV-infected persons participating in the RCCS through 2011 (Figure 2A). During the same time period uptake of medical male circumcision also was rapidly increasing among non-muslim male RCCS participants (Figure 2B; Pearson correlation =0.91). Despite increases in ART and MMC uptake, there were no significant changes in the proportion of individuals consistently using condoms with non-stable partners, either within the full RCCS population or our incidence cohort (Figure 2C). 157

180 Upward trends in ART and MMC uptake also were accompanied by modest increases in self-reported non-stable partnerships for both genders in the incidence cohort (Figure 2D- E). We also observed increased self-report of genital ulcer disease among males and females in the incidence and full RCCS cohorts (Figure 2F). We analyzed the incidence rate of HIV introduction in the incident couples cohort before and during scale-up of HIV treatment and prevention services in Rakai (i.e. pre vs. post scale-up periods). There were five RCCS visits in the period prior to ART and MMC scale-up (n=8,210 couple-years) and four visits in the period after scale-up had begun (n=13,325 couple years). Incidence of HIV introduction by a male or female was 0.80 per 100 couple-years (95%CI: ) prior to ART and MMC scale-up, and 0.52 per 100 couple-years (95%CI: ) during scale-up. Overall, we estimate a 35% reduction in the hazard of couple HIV introduction with ART and MMC scale-up (HR=0.71; 95%CI: ; adjhr=0.65; 95%CI: ); however, when gender-specific HIV introduction outcomes were examined, a significant reduction in risk was only observed for male introductions. Males had a 44% (HR=0.54; 95%: ) reduced hazard of introducing HIV in the post-scale up period (adjusted for circumcision status, genital ulcer disease, couple demographics and risk behaviors in both partners), whereas the hazard of female introduction was not significantly different between the two calendar periods (adjhr=0.89; 95%CI: ). Notably, the proportion of total HIV introductions definitively attributable to females increased from 19.7% (15/66) prior to scale-up to 39.1% (27/69) during the scale-up period (p=0.041, Figure 3C). We also examined whether the hazard of HIV introduction decreased with increasing uptake of ART and MMC within a couple s community of residence. Effects estimates for 158

181 community ART and MMC exposure were adjusted for factors included in primary adjusted model and community HIV prevalence. While we observed a greater reduction in the hazard of HIV introduction among higher tertiles and quartiles of community ART and MMC exposure, these results were not statistically significant (Table S3). To assess whether any one particular visit was driving the pre-post hazard ratio estimates, a model with a visit specific indicator was compared with our final model in which each visit was classified as either within the pre or post scale-up periods. Comparison of Akaike s or Bayesian Information Criteria (AIC and BIC, respectively) suggested that a pre vs. post model better explained the data (AIC pre/post = vs. AIC visit = ; BIC pre/post = vs. BIC visit =2219.2). We also conducted sensitivity analyses in which the pre-post period cutoff was changed by either moving the cutoff visit forward or backward one visit and sensitivity analyses excluding each RCCS visit from the unadjusted and adjusted models. In all cases we found a reduction in the hazard of HIV introduction during the ART and MMC scale-up period relative to the pre scaleup period (Table S4-5), though the effect size was reduced when the cutoff visit changed Discussion Concordant HIV-negative couples comprise the largest fraction of sexually active adults at risk for HIV infection in sub-saharan Africa; however, there have been few empirical studies of the levels and drivers of HIV incidence within this demographic using data from couples. Here, we sought to identify factors associated with HIV introduction, or first incident event in initially uninfected stable couples in Rakai, Uganda. While our findings suggest that incidence of HIV introduction into stable 159

182 couples is relatively uncommon when compared to HIV incidence within established discordant couples 15 18, it is the initial HIV introduction that ultimately gives rise to the higher risk discordant state. Discordant and concordant HIV-positive partnerships also dissolve at significantly higher rates than concordant HIV-negative partnerships 19, disrupting family structures and providing opportunity for HIV-positive persons to enter sexual partner pools and give rise to future transmissions 20,21. In 2013, the World Health Organization issued guidelines recommending that ART be provided to all HIV-positive persons in a discordant partnership regardless of stage of HIV infection 22 ; however, the timely identification of discordant HIV-positive couples is necessary for this prevention strategy to be successful. We found that the highest rates of self-report of extra-couple partnerships and formation of HIV discordance (i.e. HIV introduction) occur among couples in which male partners is under age 30 years, suggesting that ART-based interventions target younger couples 23. Our results also suggest that strategies that increase uptake of MMC and condom use with extra-couple partners both significantly associated with a decreased hazard of HIV introduction are likely to reduce formation of HIV discordance. While men were responsible for significantly more HIV introduction into stable couples than were females, we estimate female partners accounted for more than one third of HIV introductions into initially concordant HIV-negative couples. In contrast, earlier studies found that men were almost exclusively responsible for HIV introduction 4. These studies were conducted more than a decade earlier, and our results, obtained over a fourteen year period, suggest that the contribution of women to the formation of HIV discordance in stable couples may be changing with time. The substantial role that 160

183 females had in introducing virus was surprising considering the low levels of selfreported extra-couple partnerships among women compared to their male partners who self-reported partners at more than ten times the rate of women. Moreover, women who self-reported extra-couple partnerships were at considerably higher risk for HIV introduction than were males who self-reported extra-couple partners. Higher risk in females with extra-couple partners may be a result of females having older male partners who are more likely to be HIV-infected 24. It is also possible that females who choose to disclose extra-couple partnerships are inherently more risky themselves or that their extra-couple partners are more likely to be HIV-infected for reasons other than age. Prior studies have attempted to determine the validity of self-reported partnership data, but these analyses were based on assumptions about the underlying sexual network structure 25,26. For example, Ninko et al. assumed that for every female that has one extracouple partnership so does a male 26 ; however, this assumption is invalid if a small number of women (e.g., female sex workers) engage in sexual relations with many men. Our study presented the unique opportunity to examine the validity of self-report of extra-couple partnerships among newly HIV infected persons with HIV negative partners. We found evidence of underreporting of extra-couple sexual partnerships in both genders. Only 22% of introducing female partners self-reported reported an extracouple partner at the HIV seroconversion visit; however, it is unclear whether this high level of underreporting is similar among women in married couples who did not HIV seroconvert. We analyzed self-report of sexual partnerships among unmarried women who seroconverted in the RCCS over the same calendar period, and found that 99.1% (319/322) of these unmarried self-reported one or more sexual partners at their 161

184 seroconversion visit. These additional data suggest that underreporting of non-stable sexual partnerships among married is more common within this demographic than among single women which may reflect a social desirability bias due to fear of stigmatization or intimate partner abuse with disclosure. 26 Sensitive strategies that encourage women to disclose sexual behaviors to health care professionals may help target prevention interventions more effectively. We also found the risk HIV introduction into couples decreased with increasing ART and MMC uptake in the general Rakai population. Other temporal trends could potentially confound this association; however, we observed increasing self-report of HIV-related risk behaviors and symptoms of sexually transmitted infections during treatment and prevention service scale-up. That the availability of ART and MMC only reduced male risk of HIV introduction is consistent with the direct effectiveness of MMC for reducing male HIV acquisition 14. The lack of protective effect in women suggests that females who have extra-couple relationships are at especially high risk for HIV infection from partners who may not be engaged in treatment and prevention services. We previously found that Rakai women who self-reported extra-community partners with primary residence outside of RHSP coverage areas were at more than four times the risk for HIV acquisition when compared to women who selected partners locally 5. Our study has several limitations. While we sought to understand drivers of HIV discordance among couples, we only focused on one route through which HIV discordance forms (i.e. through viral introduction into initially concordant HIV-negative couples). A prior study of stable couples in the DHS found risk of HIV infection prior to couple formation is elevated among women 24. We also had substantial loss to follow-up 162

185 over the observation period (47%), and couples lost to follow up were more likely to report extra-couple partners and symptoms of genital ulcers. Our results could be biased if stable couples who acquired new infections dropped out before the introduction event was observed. Our study also excluded those couples where one or both partners were not present at the time of survey, so our incidence cohort likely represents a lower risk subset of concordant HIV-negative couples. Furthermore, with respect to concordant HIVincident couples, we assumed that when one partner introduced infection the other acquired HIV from that introducing partner. It is conceivable that some of our concordant incident pairs represent two independent HIV introductions, though a prior study in Rakai showed 100% of concordant incident couples were virologically linked 5. Our study also has several strengths. This is the largest prospective study of HIV incidence among concordant HIV-negative couples, and the only study to examine risk factors for HIV introduction using data on both partners. Couples in our incidence cohort were identified from population-based surveillance and were not self-selected through clinic based sampling. Data on extra-couple partnerships allowed assessment of predictors of extra-couple partnerships and reliability of self-reported extra-couple partnerships among newly infected individuals. The long duration of this study also permitted comparison of introduction risk before and after the scale-up of major HIV treatment and prevention interventions in Rakai. Individuals in concordant HIV-negative couples are at low risk for HIV infection, but because these individuals comprise the largest demographic of sexually active adults their ultimate contribution to HIV epidemics may be substantial 24. Strategies that promote couples voluntary counselling and testing, disclosure of high risk behaviors in females, 163

186 and uptake of HIV prevention services such MMC in younger males offer opportunities for HIV prevention in both partners. Increasing availability of ART and MMC within populations may reduce formation of HIV discordance and consequently, the rapid transmission following HIV introduction. 164

187 5.6. References 1. MESURE:DHS. Measure DHS: Demographic and Health Surveys. 2012; Available at: Accessed December/27, Chemaitelly H, Shelton JD, Hallett TB, Abu-Raddad LJ. Only a fraction of new HIV infections occur within identifiable stable discordant couples in sub-saharan Africa. AIDS 2013; 27(2): Carpenter LM, Kamali A, Ruberantwari A, Malamba SS, Whitworth JA. Rates of HIV-1 transmission within marriage in rural Uganda in relation to the HIV sero-status of the partners. AIDS 1999; 13(9): Hugonnet S, Mosha F, Todd J, et al. Incidence of HIV infection in stable sexual partnerships: a retrospective cohort study of 1802 couples in Mwanza Region, Tanzania. J Acquir Immune Defic Syndr 2002; 30(1): Grabowski M, Lessler J, Redd A, et al. Frequent introductions sustain local HIV epidemics in rural Africa. Conference on Opportunistic Infections and Retroviruses; March 2013; Atlanta, GA Wawer MJ, Gray RH, Sewankambo NK, et al. Rates of HIV-1 transmission per coital act, by stage of HIV-1 infection, in Rakai, Uganda. J Infect Dis 2005; 191(9): Hollingsworth TD, Anderson RM, Fraser C. HIV-1 transmission, by stage of infection. J Infect Dis 2008; 198(5): Cohen MS, Dye C, Fraser C, Miller WC, Powers KA, Williams BG. HIV treatment as prevention: debate and commentary--will early infection compromise treatment-asprevention strategies? PLoS Med 2012; 9(7): e Serwadda D, Gray RH, Wawer MJ, et al. The social dynamics of HIV transmission as reflected through discordant couples in rural Uganda. AIDS 1995; 9(7): Senkoro KP, Boerma JT, Klokke AH, Ng'weshemi JZ, Muro AS, Gabone R, Borgdorff MW. HIV incidence and HIV-associated mortality in a cohort of factory workers and their spouses in Tanzania, 1991 through J Acquir Immune Defic Syndr 2000; 23(2): Wawer MJ, Sewankambo NK, Serwadda D, et al. Control of sexually transmitted diseases for AIDS prevention in Uganda: a randomised community trial. Rakai Project Study Group. Lancet 1999; 353(9152):

188 12. Auvert B, Taljaard D, Rech D, et al. Association of the ANRS male circumcision project with HIV levels among men in a South African township: evaluation of effectiveness using cross-sectional surveys. PLoS Med 2013; 10(9): e Tanser F, Barnighausen T, Grapsa E, Zaidi J, Newell ML. High coverage of ART associated with decline in risk of HIV acquisition in rural KwaZulu-Natal, South Africa. Science 2013; 339(6122): Gray RH, Kigozi G, Serwadda D, et al. Male circumcision for HIV prevention in men in Rakai, Uganda: a randomised trial. Lancet 2007; 369(9562): Curran K, Baeten JM, Coates TJ, Kurth A, Mugo NR, Celum C. HIV-1 prevention for HIV-1 serodiscordant couples. Curr HIV/AIDS Rep 2012; 9(2): Guthrie BL, de Bruyn G, Farquhar C. HIV-1-discordant couples in sub-saharan Africa: explanations and implications for high rates of discordancy. Curr HIV Res 2007; 5(4): Wawer MJ, Gray RH, Sewankambo NK, et al. Rates of HIV-1 transmission per coital act, by stage of HIV-1 infection, in Rakai, Uganda. J Infect Dis 2005; 191(9): Cohen MS, Mastro TD, Cates W,Jr. Universal voluntary HIV testing and immediate antiretroviral therapy. Lancet 2009; 373(9669): 1077; author reply Porter L, Hao L, Bishai D, et al. HIV status and union dissolution in sub-saharan Africa: the case of Rakai, Uganda. Demography 2004; 41(3): Ankrah EM. The impact of HIV/AIDS on the family and other significant relationships: the African clan revisited. AIDS Care 1993; 5(1): Nabaitu J, Bachengana C, Seeley J. Marital instability in a rural population in south-west Uganda: implications for the spread of HIV-1 infection. Africa (Lond) 1994; 64(2): World Health Organization. Consolidated guidelines on the use of antiretroviral drugs for treating and preventing HIV infection. 2013;. 23. Mills EJ, Beyrer C, Birungi J, Dybul MR. Engaging men in prevention and care for HIV/AIDS in Africa. PLoS Med 2012; 9(2): e Bellan SE, Fiorella KJ, Melesse DY, Getz WM, Williams BG, Dushoff J. Extracouple HIV transmission in sub-saharan Africa: a mathematical modelling study of survey data. Lancet 2013;. 166

189 25. Clark S, Kabiru C, Zulu E. Do men and women report their sexual partnerships differently? Evidence from Kisumu, Kenya. Int Perspect Sex Reprod Health 2011; 37(4): Nnko S, Boerma JT, Urassa M, Mwaluko G, Zaba B. Secretive females or swaggering males? An assessment of the quality of sexual partnership reporting in rural Tanzania. Soc Sci Med 2004; 59(2):

190 Table 1. Univariate and multivariate associations between couple demographics and self-report of one or more extra-couple, unstable sexual partnerships at baseline (n=4,570) and follow-up study visits (n=12,319) by gender in 4,570 initially HIV-1 seroconcordant negative couples in the RCCS, Self-reports unstable extra-couple sexual partnership(s) Females Males No Yes PRR adjprr No Yes PRR adjprr Time-varying demographics Visits (%) Visits (%) (95% CI) (95% CI) Visits (%) Visits (%) (95% CI) (95% CI) Total Marital status Monogamous (82.7) 322 (80.9) 1.00 (Ref.) 1.00 (Ref.) 9791 (81.2) 4173 (86.3) 1.00 (Ref.) 1.00 (Ref.) Polygamous 2849 (17.3) 76 (19.1) 1.16 ( ) 1.24 ( ) 2264 (18.8) 661 (13.7) 0.59 ( ) 0.67 ( ) Duration of relationship <2.5 years 2273 (13.8) 79 (19.9) 1.33 ( ) 1.13 ( ) 1625 (13.5) 727 (15.0) 0.89 ( ) 0.89 ( ) 2.5 and <5 years 2847 (17.3) 66 (16.6) 1.00 (Ref.) 1.00 (Ref.) 1888 (15.7) 1025 (21.2) 1.00 (Ref.) 1.00 (Ref.) 5 and <10 years 4801 (29.1) 152 (38.2) 0.92 ( ) 1.04 ( ) 3266 (27.1) 1636 (33.8) 0.98 ( ) 1.03 ( ) 10 years 6570 (39.8) 79 (19.9) 1.06 ( ) 1.33 ( ) 5276 (43.8) 1446 (29.9) 0.74 ( ) 0.96 ( ) Male age years 89 (0.54) 159 (40.0) 1.02 ( ) 0.78 ( ) 59 (0.49) 33 (0.68) 0.86 ( ) 0.92 ( ) years 5845 (35.4) 139 (34.9) 1.00 (Ref.) 1.00 (Ref.) 3707 (30.8) 2297 (47.5) 1.00 (Ref.) 1.00 (Ref.) years 6643 (40.3) 97 (24.4) 0.84 ( ) 0.87 ( ) 4836 (40.1) 1946 (40.3) 0.82 ( ) 0.93 ( ) 40 years 3914 (23.7) 3 (0.75) 1.02 ( ) 1.01 ( ) 3453 (28.6) 558 (11.5) 0.49 ( ) 0.73 ( ) Female age (years) years 1670 (10.1) 66 (16.7) 1.71 ( ) 1.70 ( ) 1176 (9.8) 560 (11.6) 0.91 ( ) 0.94 ( ) years 8924 (54.1) 200 (50.3) 1.00 (Ref.) 1.00 (Ref.) 6049 (50.2) 3075 (63.6) 1.00 (Ref.) 1.00 (Ref.) years 4555 (27.6) 105 (26.4) 1.10 ( ) 0.95 ( ) 3630 (30.1) 1030 (21.3) 0.75 ( ) 0.88 ( ) 40 years 1342 (8.1) 27 (6.8) 0.95 ( ) 0.74 ( ) 1200 (10.0) 169 (3.5) 0.46 ( ) 0.63 ( ) Age difference Male younger 751 (4.6) 26 (6.5) 1.55 ( ) 1.77 ( ) 521 (4.3) 256 (5.3) 1.00 ( ) 1.07 ( ) Male 0 and <5 years older 6315 (38.3) 136 (34.2) 1.00 (Ref.) 1.00 (Ref.) 4303 (35.7) 2148 (44.4) 1.00 (Ref.) 1.00 (Ref.) Male 5 and <10 years older 6379 (38.7) 147 (36.9) 1.11 ( ) 1.13 ( ) 4264 (38.4) 1902 (39.4) 0.89 ( ) 0.90 ( ) Male 10 years older 3046 (18.5) 89 (22.4) 1.30 ( ) 1.46 ( ) 3607 (21.6) 256 (5.3) 0.54 ( ) 0.64 ( ) Partner self-reports extracouple sexual partnership 168

191 No (71.6) 247 (62.1) 1.00 (Ref.) 1.00 (Ref.) (98.0) 4683 (96.9) 1.00 (Ref.) 1.00 (Ref.) Yes 4583 (28.4) 151 (37.9) 1.39 ( ) 1.45 ( ) 247 (2.0) 151 (3.1) 1.18 ( ) 1.22 ( ) PRR=unadjusted prevalence risk ratio; adjprr=adjusted prevalence risk ratio; Adjusted model included the following covariates: marital status, duration of relationship, male age, female age, age difference in couple, partner self-reported extra-couple relationships, and study visit. 169

192 Figure 1. Probability of self-reporting one or more non-stable sexual partners and incidence of a first HIV introduction event (male or female, female only, and male only) among initially concordant HIVnegative couples by age and gender of the male and female index partners A) Probability of a female self-reporting one or more non-stable extra-couple partners by her age. B) Probability of a male self-reporting one or more non-stable extra-couple partners by his age. Note difference in y-axes between Figures 1A and 1B. C-D) Incidence of any HIV-1 introduction (black), female introduction only (red), and male introduction only (blue) by age of female (C) and age of male (D). 170

193 Table 2. Relative hazard of HIV introduction and self-report of extra-couple sexual partnerships, genital ulcer disease and male-circumcision status Couple HIV-1 introduction events Male or Female introduction Female introduction Male introduction M-/F- M+/F- M-/F+ M+/F+ Total Time* HR adjhr HR adjhr HR adjhr Visits (%) N (%) N (%) N (%) N (%) Years (95% CI) (95% CI) (95% CI) (95% CI) (95% CI) (95% CI) Overall Self-reports non-stable sexual partnership (T S ) Neither partner (69.9) 29 (43.9) 21 (52.5) 13 (44.8) 63 (46.8) (Ref.) 1.00 (Ref.) 1.00 (Ref.) 1.00 (Ref) 1.00 (Ref.) 1.00 (Ref) Male only 4607 (27.5) 34 (51.5) 10 (25.0) 16 (55.2) 60 (44.5) ( ) 2.13 ( ) 1.53 ( ) 1.42 ( ) 2.83 ( ) 2.59 ( ) Female only 281 (1.7) 2 (3.0) 6 (15.0) 0 (0.0) 8 (5.9) ( ) 6.64 ( ) 12.4 ( ) 12.6 ( ) 2.96 ( ) 2.70 ( ) Both partners 163 (0.97) 1 (1.5) 3 (7.5) 0 (0.0) 4 (3.0) ( ) 4.64 ( ) 9.65 ( ) 8.4 ( ) 2.15 ( ) 1.87 ( ) Prior history of self-reporting non-stable sexual partnerships (T 0, S-1 ) Neither partner 7202 (59.1) 19 (28.8) 19 (47.5) 14(48.3) 52 (38.5) (Ref.) (Ref.) (Ref.) - Male only 4564 (37.5) 40 (60.6) 13 (32.5) 14 (48.3) 67 (49.6) ( ) ( ) ( ) - Female only 204 (1.7) 3 (4.6) 4 (10.0) 0 (0.0) 7 (5.2) ( ) ( ) ( ) - Both partners 214 (1.8) 4 (6.1) 4 (10.0) 1 (3.5) 9 (6.7) ( ) ( ) ( ) - Male self-reports stable partnership (T S ) No (82.1) 49 (74.2) 34 (85.0) 26 (89.7) 109 (80.7) (Ref.) 1.00 (Ref.) 1.00 (Ref.) 1.00 (Ref.) 1.00 (Ref.) 1.00 (Ref.) Yes 2996 (17.9) 17 (25.8) 6 (15.0) 3 (10.3) 26 (19.3) ( ) 1.27 ( ) 0.74 ( ) 0.69 ( ) 1.35 ( ) 1.73 ( ) Self-reported symptoms of genital ulcer disease (T S-1 ) Neither partner 9984 (82.4) 46 (69.7) 28 (70.0) 23 (79.3) 97 (71.9) (Ref.) 1.00 (Ref) 1.00 (Ref.) 1.00 (Ref.) 1.00 (Ref.) 1.00 (Ref.) Male only 897 (7.4) 7 (10.6) 3 (7.5) 1(3.5) 11 (8.2) ( ) 1.12 ( ) 0.96 ( ) 0.94 ( ) 1.36 ( ) 1.23 ( ) Female only 996 (8.2) 8 (12.1) 6 (15.0) 2 (6.9) 16 (11.9) ( ) 1.64 ( ) 1.87 ( ) 1.79 ( ) 1.53 ( ) 1.58 ( ) Both partners 239 (2.0) 5 (7.6) 3 (7.5) 3 (10.3) 11 (8.2) ( ) 4.05 ( ) 4.59 ( ) 4.35 ( ) 4.69 ( ) 4.15 ( ) Male circumcised (T S-1 ) No 9337 (76.7) 59 (89.4) 26 (65.0) 28 (96.6) 113 (83.7) (Ref.) 1.00 (Ref.) 1.00 (Ref.) 1.00 (Ref.) 1.00 (Ref.) 1.00 (Ref.) Yes 2832 (23.3) 7 (10.6) 14 (35.0) 1 (3.4) 22 (16.3) ( ) 0.66 ( ) 1.24 ( ) 1.29 ( ) 0.35 ( ) 0.35 ( ) T S =seroconversion visit/current visit; T S-1 =visit prior to seroconversion/visit prior to current visit; T 0,S-1 =All visits prior to seroconversion visit/current visit (T S ); HR=hazard ratio; adjhr=adjusted hazard ratio; *Couple-time at risk for first introduction event; Adjusted model included male and female age (T S ), duration of partnership (T S ), male circumcision status (T S-1 ), self-reported genital ulcer disease in male and female partners (T S-1 ), and an indicator for whether or not the exposure period occurred before or during MMC and ART scale-up (T S ). 171

194 Figure 2. Time trends in ART and MMC coverage, consistent condom use with non-stable sexual partners, extra-couple partnership selection, and self-reported genital ulcer disease in the full RCCS cohort (solid lines) and in the incidence cohort of concordant HIV-negative couples (dotted lines). A) Percentages of all HIV infected males and females in the full RCCS cohort who were receiving anti-retroviral therapies B) Percentages of non-muslim men in full RCCS and incidence cohorts who were circumcised. C) Percentages of non-stable sexual partnerships in which there was consistent condom use as self-reported by males and females in the full RCCS and incidence cohorts. D) Percentage of males in the incidence cohort who reported one more non-stable sexual partnerships. E) Percentage of females in incidence cohort who reported one or more non-stable sexual partnerships. F) Percentages of HIV-negative males and females who self-reported symptoms of genital ulcer disease in the full RCCS and incidence cohorts. 172

195 Figure 3: Incidence of HIV introduction into stable HIV concordant negative couples before and after the scale-up of antiretroviral therapies (ART) and medical male circumcision (MMC) in Rakai. A) Incidence of HIV introduction per 100 couple-years and 95% CI pre and post ART and MMC scale-up. B) Relative hazard of any introduction (black), female introduction (red), and male (blue) and 95% CI in the post scale-up vs. pre-scale up eras. C) Proportion of virus introduced by males (M+/F-), females (M-/F+), or unknown partner (M+/F+) in pre and post scale-up eras. 173

196 Table S1: Comparison of selected baseline characteristics for concordant HIV-1 negative couples with and without follow-up in the RCCS, >=1 follow-up visits No follow-up visits N=4570 N=4057 N (%) N (%) Marital status Monogamous 3785 (82.8) 3218 (79.3) Polygamous 785 (17.2) 839 (20.7) Male age years 73 (1.6) 7.0 (1.7) years 2298 (50.3) 1955 (48.2) years 1389 (30.4) 1248 (30.8) 40 years 810 (17.7) 784 (19.3) Female age (years) years 1087 (23.8) 927 (22.9) years 2404 (52.6) 2097 (51.7) years 798 (17.5) 699 (17.2) 40 years 278 (6.1) 334 (8.2) Self-reported extra-couple, non-stable sexual partnership Neither male nor female 3207 (70.2) 2741 (67.6) Male only 1170 (25.6) 1130 (27.9) Female only 125 (2.7) 129 (3.2) Both male and female 67 (1.5) 57 (1.4) Self-reported symptoms of sexually transmitted infection* Neither male nor female 2971 (46.9) 2448 (60.7) Male only 521 (11.5) 517 (12.8) Female only 777 (17.2) 766 (18.9) Both male and female 250 (5.5) 304 (7.5) *Includes self-report of genital ulcer disease, dysuria, or genital warts 174

197 Table S2: Characteristics of non-stable extra-couple sexual partners of incident male and female HIV-1 cases (at T S and T S-1 visits) and partners of matched HIV-negative controls. Female partners Male partners Controls Cases Controls Cases MOR N (%) N (%) N (%) N (%) 95% CI N=51 N=13 N=741 N=147 Self-described relationship with partner Boyfriend/Girlfriend 45 (88.2) 11 (84.6) 613 (82.7) 127 (86.4) 1.00 (Ref.) Casual 5 (9.8) 1 (7.7) 125 (16.7) 19 (12.9) 0.73 ( ) Relative 1 (2.0) 1 (7.7) 3 (0.4) 1 (0.7) 1.72 ( ) Relationship with partner is ongoing at time of survey Yes 35 (68.6) 7 (53.9) 323 (53.5) 62 (54.4) 1.00 (Ref.) No 15 (29.4) 6 (46.2) 279 (46.2) 52 (45.6) 1.10 ( ) Don't know 1 (2.0) 0 (0.0) 2 (0.33) 0 (0.0) - Relative age of partner to him/herself Younger 4 (7.8) 2 (15.4) 534 (72.1) 99 (67.4) 1.00 (Ref.) Same age 9 (17.7) 1 (7.7) 36 (4.9) 12 (8.2) 1.35 ( ) Older 35 (68.6) 10 (76.0) 52 (7.0) 13 (8.8) 1.82 ( ) Don't know 3 (5.9) 0 (0.0) 119 (16.1) 23 (15.6) 1.06 ( ) Shares community with partner Yes 30 (58.8) 4 (30.8) 383 (51.7) 66 (44.9) 1.00 (Ref.) No 21 (41.2) 9 (69.2) 358 (48.3) 81 (55.1) 1.30 ( ) Partner s primary occupation Home agriculture 4 (7.8) 1 (7.7) 107 (14.4) 30 (20.4) 1.00 (Ref.) Selling agriculture products 12 (23.5) 4 (30.8) 15 (2.0) 3 (2.0) 0.69 ( ) Government/Clerical work 8 (15.7) 1 (7.7) 32 (4.3) 3 (2.0) 0.36 ( ) Trader/Shopkeeper 20 (39.2) 4 (30.8) 51 (6.9) 11 (7.5) 0.77 ( _ Bar worker 0 (0.0) 1 (7.7) 2 (0.27) 3 (2.0) 8.87 ( ) Trucking 1 (2.0) 0 (0.0) 0 (0.0) 0 (0.0) - Other 6 (11.8) 2 (15.4) 150 (20.2) 26 (17.7) 0.41 ( ) Don't know/remember 0 (0.0) 0 (0.0) 384 (51.8) 71 (48.3) 0.60 ( ) Condom use with partner Never 26 (51.0) 8 (61.5) 216 (29.2) 47 (32.0) 1.00 (Ref.) Sometimes 9 (17.7) 2 (15.4) 149 (20.1) 26 (31.3) 1.40 ( ) Always 12 (23.5) 1 (7.7) 295 (39.8) 25 (23.8) 0.53 ( ) Don't know/remember 4 (7.8) 2 (15.4) 81 (10.9) 19 (12.9) 1.13 ( ) Perceived HIV-risk from partner None/Unlikely 13 (13.7) 1 (6.6) 196 (26.7) 42 (29.2) 1.00 (Ref.) Don t know 3 (5.9) 2 (15.4) 139 (18.9) 33 ( ( ) Somewhat likely 27 (52.9) 8 (61.5) 347 (47.2) 63 (43.8) 0.84 ( ) Very likely 15 (27.5) 2 (15.4) 53 (7.2) 6 (4.2) 0.53 ( ) T S =seroconversion visit/current visit; T S-1 =visit prior to seroconversion/visit prior to current visit; MOR=Matched Odds Ratio; Partners of cases were matched to partners of HIV-negative controls on age of the index reporting male or female and the visit at which the case partner was reported 175

198 Figure S1. Age of male partner vs. age of female partner for stable sexual partnerships (index/incidence cohort partners) (A) and for extra-couple non-stable sexual partnerships self-reported by B) incidence cohort females and C) incidence cohort males. Colors indicate the HIV-1 status of RCCS participant. The black dotted is the diagonal. The black solid line is the best fit line of male index age vs. female index age using a linear generalized additive model. Similarly the pink line is the best fit line for female index vs. the age of her extra-couple partner and the cyan line the best fit line for index male vs. the age of his extra-couple partner. 176

199 Table S3. Risk of HIV introduction with higher community coverage of ART in HIV infected persons and MMC in non-muslim men Any HIV introduction Female HIV introduction Male HIV introduction HR (95% CI) adjhr (95%CI) HR (95% CI) adjhr (95%CI) HR (95% CI) adjhr (95%CI) ART Model Community ART coverage (T S ) No coverage 1.00 (Ref.) 1.00 (Ref) 1.00 (Ref) 1.00 (Ref.) 1.00 (Ref.) 1.00 (Ref.) 1st tertile ( %) 0.99 ( ) 0.89 ( ) 1.56 ( ) 1.39 ( ) 0.75 ( ) 0.68 ( ) 2nd tertile ( %) 0.94 ( ) 0.81 ( ) 1.19 ( ) 1.05 ( ) 0.86 ( ) 0.74 ( ) 3rd tertile ( %) 0.64 ( ) 0.61 ( ) 0.78 ( ) 0.69 ( ) 0.59 ( ) 0.62 ( ) Community HIV Prevalence (T S ) 1st quartile ( ) 1.00 (Ref.) 1.00 (Ref.) 1.00 (Ref.) 1.00 (Ref.) 1.00 (Ref.) 1.00 (Ref,) 2nd quartile ( ) 0.59 ( ) 0.67 ( ) 0.87 ( ) 1.06 ( ) 0.42 ( ) 0.46 ( ) 3rd quartile ( ) 2.11 ( ) 2.37 ( ) 2.04 ( ) 2.44 ( ) 2.14 ( ) 2.35 ( ) 4th quartile ( ) 2.04 ( ) 2.09 ( ) 1.87 ( ) 2.05 ( ) 2.13 ( ) 2.16 ( ) Circumcision model Community MMC prevalence (T S ) 1st quartile (0-5.56%) 1.00 (Ref.) 1.00 (Ref) 1.00 (Ref.) 1.00 (Ref.) 1.00 (Ref.) 1.00 (Ref.) 2nd quartile ( %) 1.24 ( ) 1.17 ( ) 1.43 ( ) 1.51 ( ) 1.17 ( ) 1.17 ( ) 3rd quartile ( %) 0.89 ( ) 0.81 ( ) 0.90 ( ) 0.80 ( ) 0.89 ( ) 0.72 ( ) 4th quartile ( %) 0.95 ( ) 0.83 ( ) 1.25 ( ) 1.15 ( ) 0.80 ( ) 0.79 ( ) Community HIV Prevalence (T S ) 1st quartile ( ) 1.00 (Ref.) 1.00 (Ref.) 1.00 (Ref.) 1.00 (Ref.) 1.00 (Ref.) 1.00 (Ref.) 2nd quartile ( ) 0.59 ( ) 1.27 ( ) 0.87 ( ) 1.02 ( ) 0.42 ( ) 0.50 ( ) 3rd quartile ( ) 2.11 ( ) 6.06 ( ) 2.04 ( ) 2.39 ( ) 2.14 ( ) 2.37 ( ) 4th quartile ( ) 2.04 ( ) 7.85 ( ) 1.87 ( ) 1.98 ( ) 2.13 ( ) 2.36 ( ) Proportion of HIV infected persons on ART within an individual's community of primary residence; Proportion of non-muslim males circumcised within an individual's community of primary residence 177

200 Table S4. Sensitivity of the unadjusted and adjusted pre-post ART and MMC hazard ratios to cutoff visit Any HIV introduction Female HIV introduction Male HIV introduction HR (95% CI) adjhr (95%CI) HR (95% CI) adjhr (95%CI) HR (95% CI) adjhr (95%CI) Shifted cut-off down one visit 0.99 ( ) 0.89 ( ) 1.19 ( ) ) 0.90 ( ) 0.84 ( ) Shifted cut-off up one visit 0.92 ( ) 0.84 ( ) 1.09 ( ) 0.95 ( ) 0.84 ( ) 0.80 ( ) Table S5. Sensitivity of the unadjusted and adjusted pre-post ART and MMC hazard ratios to exclusion of a single study visit Any HIV introduction Female HIV introduction Male HIV introduction Excluded visit HR (95% CI) adjhr (95%CI) HR (95% CI) adjhr (95%CI) HR (95% CI) adjhr (95%CI) Pre Scale Up ( ) 0.61 ( ) 0.85 ( ) 0.78 ( ) 0.58 ( ) 0.51 ( ) ( ) 0.62 ( ) 1.02 ( ) 0.94 ( ) 0.55 ( ) 0.45 ( ) ( ) 0.63 ( ) 0.98 ( ) 0.86 ( ) 0.57 ( ) 0.61 ( ) ( ) 0.71 ( ) 1.13 ( ) 1.01 ( ) 0.65 ( ) 0.60 ( ) ( ) 0.75 ( ) 1.11 ( ) 0.98 ( ) 0.72 ( ) 0.68 ( ) Post Scale Up ( ) 0.78 ( ) 1.21 ( ) 0.99 ( ) 0.74 ( ) 0.71 ( ) ( ) 0.70 ( ) 1.03 ( ) 0.93 ( ) 0.64 ( ) 0.62 ( ) ( ) 0.50 ( ) 0.80 ( ) 0.75 ( ) 0.45 ( ) 0.42 ( ) ( ) 0.59 ( ) 0.95 ( ) 0.83 ( ) 0.54 ( ) 0.50 ( ) 178

201 Chapter 6 High risk human papillomavirus viral load and persistence among heterosexual HIV-negative and HIV-positive men Mary K. Grabowski, Ronald H Gray, David Serwadda, Godfrey Kigozi, Patti E. Gravitt, Fred Nalugoda, Steven J. Reynolds, Maria J. Wawer, Stephen Watya, Thomas C. Quinn, Aaron A. R. Tobian 6.1. Abstract Objectives: High-risk human papillomavirus (HR-HPV) viral load is associated with HR-HPV transmission and HR-HPV persistence in women. It is unknown whether HR- HPV viral load is associated with persistence in HIV-negative or HIV-positive men. Methods: HR-HPV viral load and persistence were evaluated among 703 HIV-negative and 233 HIV-positive heterosexual men who participated in a male circumcision trial in Rakai, Uganda. Penile swabs were tested at baseline and 6, 12 and 24 months for HR- HPV using the Roche HPV Linear Array, which provides a semi-quantitative measure of HPV shedding by hybridization band intensity (graded:1-4). Prevalence risk ratios (PRR) were used to estimate the association between HR-HPV viral load and persistent detection of HR-HPV. 179

202 Results: HR-HPV genotypes with high viral load (grade:3-4) at baseline were more likely to persist than HR-HPV genotypes with low viral load (grade:1-2) among HIVnegative men (month 6: adjprr=1.83, 95%CI: ; month 12: adjprr=2.01, 95%CI: ), and HIV-positive men (month 6: adjprr=1.33, 95%CI: ; month 12: adjprr=1.73, 95%CI: ). Long-term persistence of HR-HPV was more frequent among HIV-positive men compared to HIV-negative men (month 24: adjprr=2.27, 95%CI: ). Persistence of newly detected HR-HPV at the 6 and 12 month visits with high viral load were also more likely to persist to 24 months than HR- HPV with low viral load among HIV-negative men (adjprr=1.67, 95%CI ). Conclusions: HR-HPV genotypes with high viral load are more likely to persist among HIV-negative and HIV-positive men, though persistence was more common among HIVpositive men overall. The results may explain the association between high HR-HPV viral load and HR-HPV transmission. 180

203 6.2. Introduction High risk human papillomavirus (HR-HPV) infection is a common sexually transmitted infection, and most sexually active individuals will acquire HR-HPV in their lifetime 1, 2. Persistent HR-HPV can cause oral, anal, penile, and cervical cancer, and greater than 85% of the disease burden is in developing countries 2-4. Cervical cancer is the leading cause of cancer mortality in women in Eastern Africa 4. Among women, numerous studies have demonstrated that HR-HPV DNA viral load is associated with both persistence 5, 6 and progression to cervical lesions, especially HR-HPV 16 5, 7. Data on the natural history of HR-HPV infection in men is limited. HR-HPV infection may be associated with HIV acquisition in men 8, 9. In addition, HR-HPV persistence among heterosexual men is associated with HIV infection, smoking, increased number of lifetime sexual partners, absence of male circumcision, and younger age Higher viral load among men is associated with detection of HR-HPV at multiple penile sites 14 and flat penile lesions 15. It has also been shown that HR-HPV transmission to female partners is associated with higher male viral loads 16. Since male HR-HPV infection is a key component of HPV transmission 2, 17, it is important to understand the natural history of HR-HPV in both sexes. However, neither risk factors for increased HR-HPV viral load nor the association between HR-HPV viral load and persistent infection among heterosexual men have been evaluated. In this study, HR- HPV viral load and persistence were assessed among HIV-negative and HIV-positive heterosexual men who participated in a male circumcision trial in Rakai, Uganda. 181

204 6.3. Materials and methods Study design and participants HR-HPV was assessed among men aged years during two trials of male circumcision for HIV and STI prevention in Rakai District, Uganda 18, 19. Men who had contraindications for surgery (e.g., anemia, active genital infection) were treated, and if their medical condition resolved, they were re-screened and enrolled into the trial. Those with anatomical abnormalities (e.g., hypospadias), other medical contraindications or indications for surgery (e.g., severe phimosis) were excluded. HIV-positive men with CD4 counts <250 cells/mm 3 or WHO stage 4 disease were excluded from the trials and referred for treatment. Participants provided written informed consent prior to screening and at baseline. Men were randomly assigned to receive immediate circumcision (intervention) or circumcision delayed for 24 months (control). Male circumcision was performed following the baseline visit for those men in the treatment arm of the trial. Infectious disease testing (HPV, HIV, HSV-2, and syphilis), physical examinations, and interviews to ascertain sociodemographic characteristics and sexual risk behaviors were conducted at and at 6, 12 and 24 months follow-up visits. Serum and swab samples were stored at -80 C. All subjects were offered free HIV counseling and testing, health education and condoms at each visits and those participants found to be HIV-positive were referred for free care to the Rakai Health Sciences Program. Of 6396 men enrolled in the two trials, 1790 men with a total of 5478 samples were tested for HR-HPV. Males were randomly selected from the trial population, except for married men, who were oversampled to permit a parallel study of HPV transmission to 182

205 their female partners 13. There were 949 swabs (17.3%) with no detectable cellular betaglobin or detectable HPV that were excluded in this analysis since the adequacy of the sample collection could not be ensured. HIV seroconverters (n=97) and men with an indeterminate HIV result at last visit (n=115) were excluded from this analysis since acute HIV infection is associated with a substantial increase of new HR-HPV infection 20. Of 1529 participants with 3961 observations, we restricted our analyses to only those men who provided swab samples at baseline and at least one follow-up visit (6, 12 and/or 24 months follow-up). A total 3084 study visits from 936 participants (703 HIV-negative and 233 HIV-positive men) were analyzed for this study. Our primary analysis focused on the association between HR-HPV DNA load and the persistence of HR-HPV infections detected among HIV-negative and HIV positive men at baseline to 6, 12, and 24 months. We also conducted a secondary analysis in which we analyzed the association between HR-HPV viral load and persistence among men with newly detected infections at either month 6 and 12 to month 24. The trials were approved by the Ugandan National Council for Science and Technology (UNCST, Kampala, Uganda), and by three institutional review boards: the Science and Ethics Committee of the Uganda Virus Research Institute (Entebbe, Uganda), the Johns Hopkins University Bloomberg School of Public Health IRB (Baltimore, MD, USA), and the Western Institutional Review Board (Olympia, WA, USA). The trials were overseen by independent Data Safety Monitoring Boards 21, and were registered with ClinicalTrials.Gov numbers NCT and NCT

206 HR-HPV Detection, Viral load quantification, and STI Testing Moistened Dacron swabs were rotated around the full circumference of the penis at the coronal sulcus and glans by trained male clinicians or nurses, and were stored in Digene specimen transport media at -80 o C until assay. HPV genotyping was performed using the Roche HPV Linear Array (Roche Diagnostics, Indianapolis, IN) 22. HPV genotypes 16, 18, 31, 33, 35, 39, 45, 51, 52, 56, 58, 59, 66, and 68 were considered high risk genotypes (HR-HPV). For each positive HR-HPV genotype, the band intensity was visually scored as 1-4, with intensity 4 representing the strongest LA hybridization. All laboratory technologists and evaluators of band intensity were blinded to trial arm and all demographic data. It has been previously demonstrated in both men and women that the linear array hybridization signal is linearly correlated with log-transformed HPV viral load (Spearman s r=073) 23, 24 ; a linear array band signal strength of 4 is approximately equivalent to a viral load of 2000 copies/5µl. Band signal strength of 3 is approximately copies/5µl. Thus, band signals of 3 and 4 represent >200 copies/5µl. The linear array results and band intensity were evaluated by two observers, independently, using a labelled acetate overlay that was provided by Roche, where the overlay indicated the position of the genotype probes on the test strip. Observer disagreement occurred rarely (1-4% of results) and was resolved by re-evaluation by the initial observers 23. Here, we estimated the proportion of linear array results with band intensities 3 and 4 (i.e. high viral load HR-HPV), relative to lower intensity bands 1-2 (i.e. low viral load HR- HPV), among men with detectable HR-HPV infections. 184

207 HSV-2 infection was determined by HSV-2 ELISA (Kalon Biological Ltd, Guilford, UK), as previously described 18. HIV status was determined using two separate ELISAs and confirmed by HIV-1 Western Blot, as previously described Statistical Analysis Baseline and follow-up characteristics of ART-naïve HIV-positive and HIV-negative men evaluated for HR-HPV were tabulated (n=936). Men were stratified by HIV and HR-HPV status, and differences were assessed using chi-squared tests with two-sided p- values. In all other analyses, the unit of observation was the HR-HPV genotype. Persistent detection of HR-HPV infection was defined as an initially HR-HPV infected man with continued detection of same genotype at the 6, 12, or 24 month follow-up visits. All analyses were stratified by HIV-status and were conducted independently for each follow-up visit. High viral load (band intensities 3-4) or low viral load (band intensities 1-2) was assessed for each HR-HPV genotype at baseline and follow-up visits, irrespective of the number of HR-HPV infections per individual. High viral load at baseline was the primary exposure variable and persistent HR-HPV detection at month 6, 12, or 24 months was the primary outcome. In a secondary analysis of newly detected HR-HPV at months 6 and 12, the exposure variable was viral load at time of initial detection and the outcome was the persistent detection of the same genotype at month 24. The sensitivity of the results to our exposure definitions was also assessed by comparing persistence with band intensity 4 vs. band intensities

208 We analyzed the association between HR-HPV viral load at baseline and the demographic and behavioral characteristics of HR-HPV-infected men using prevalence risk ratios (PRR) at each of the three follow-up visits. PRRs and 95% confidence intervals (CI) were visit-specific and were estimated using Poisson regression with generalized estimating equations (GEE) and robust variance to account for correlation between multiple HR-HPV within the same man. Since we were estimating the association between HR-HPV viral load and detection at each follow-up visit independently, no offset was incorporated into the regression models. All analyses were stratified by the HIV-status of the man from which the HR-HPV was isolated. Covariates in adjusted analyses included male age at baseline, which was highly correlated with marital status, treatment arm (i.e. circumcised or not), detection of betaglobin at baseline (for primary analyses) or follow-up (for secondary analyses), and sex with non-marital partners in the last year (time varying at 6, 12, and 24 months). Covariates were considered potential confounders if they were associated with the exposure and or causally associated with persistent detection, the latter of which was determined from the previously published literature 8-13, 23. We could not adjust for smoking status, since these data were not obtained in the primary clinical trial. All effect estimates presented in the main text are adjusted estimates unless otherwise noted. We also examined whether the association of viral load was modified by infection with multiple HR-HPV at baseline or follow-up, circumcision status, and HR-HPV genotype. Interaction terms were used to test for effect modification by multiple HR-HPV infections, circumcision, and HR-HPV genotype. 186

209 Statistical analyses were conducted in STATA Version 11.0 (STATA Corp LP, College Station, Texas) and R version Results Among 703 HIV-negative and 233 ART-naïve HIV-positive men evaluated for HR- HPV at baseline (Supplemental Table 1), at least one HR-HPV was identified in 264 HIV-negative men (37.6%, 403 genotypes total) and 164 HIV-positive men (70.4%, 399 genotypes total). Only one (55.8%) or two HR-HPV (20.8%) were detected in most men, however up to 9 unique HR-HPV genotypes were identified in a single study participant. The most frequently detected HR-HPV genotypes were HPV-16 (11.0%) and HPV-51 (11.5%) (Supplemental Table 2). Of the HR-HPV genotypes detected at baseline, 45.9% and 50.4% had a band intensity of 3-4 in HIV-positive and HIV-negative men, respectively (unadjusted PRR=0.91, 95%CI: ). Table 1 shows demographic and behavioral factors associated with high HR-HPV viral load at baseline. Among HIV-negative men, infections with high viral load were detected more frequently in individuals who were younger and unmarried. There were no significant associations between a high HR-HPV viral load and baseline demographic or behavioral factors among HIV-positive men. Persistent detection of HR-HPV at follow-up was significantly associated with a higher viral load at baseline (Table 2). Among HIV-negative men, HR-HPV infections with high viral loads were 1.83 times more likely to be detected at month 6 (95% CI: ) and 2.01 times more likely to be detected at month 12 (95% CI: ) 187

210 than infections with low viral load at baseline. By month 24, differences between high and low viral load infections among HIV-negative men were no longer statistically significant; though, the prevalence of HR-HPV with high viral load at baseline was still greater than HR-HPV with low viral load (15.2% vs. 8.8%, respectively, adjprr=1.58, 95%CI: ). Higher viral load at baseline was also significantly associated with persistent HR- HPV detection among HIV-positive men up to 12 months (Table 2). Prevalence of HR- HPVs with a high viral load at baseline were 1.33 times greater at 6 months (95% CI: ) and 1.73 times greater at 12 months (95% CI: ) than those HR-HPVs with low viral load. HR-HPV infections with low viral loads were detected more frequently at all follow-up visits in HIV-positive compared to HIV-negative men (6 months: adjprr=1.86, 95%CI: ; 12 months: adjprr=1.45, 95%CI: ; 24 months: adjprr=2.72, 95%CI: ) Persistent detection of baseline HR-HPV at 24 months was more frequent among HIV-positive than HIV-negative men (Figure 1). At two years, detection of HR-HPV was 2.27 times greater among HIV-positive after controlling for HR-HPV viral load at baseline, circumcision status, detection of beta-globin at baseline, non-marital sexual partners in the last 12 months, and age (95%CI: ). In our interaction analyses, the association between viral load and persistent HR-HPV detection during follow-up among HIV-negative or HIV-positive men did not differ by circumcision status, HR-HPV genotype, or infection with multiple HR-HPV at baseline or follow-up (data not shown), though we may have had limited power to detect differences among these subgroups. 188

211 Among HIV-negative men with HR-HPV at baseline and complete follow-up (n=139 men), HR-HPV with high viral load at baseline were 3.76 (95%CI: ) times more likely to be persistently detected at all follow-up visits (6, 12, and 24 months). Similarly, the persistent detection of HR-HPV with high viral load at baseline was 1.79 (95%CI: ) times greater than for HR-HPV with low viral load at baseline among HIVpositive men (n=86 men). Persistence of newly detected infections in month 6 and 12 was also analyzed and associated with viral load at initial detection (Table 3). There were 163 new HR-HPVs at month 6 and month 12 in 141 HIV-positive men, and 226 new infections in 87 HIVnegative men at the same visits. Associations between viral load and persistent detection in follow-up followed similar trends to that observed among baseline HR-HPV infections. Persistent detection of newly detected HR-HPV with high viral load more frequent than detection of new infections with low viral load among HIV-negative men; however, these results were not statistically significant (adjprr=1.67, 95% CI: ). High viral load viruses detected at 6 and 12 months were not more likely to persist than low viral load viruses among HIV-positive men (adjprr=1.10, 95% CI: ) Discussion The natural history of HR-HPV is poorly characterized among HIV-negative and HIV-positive men, especially in resource limited settings where the burden of HR-HPV associated diseases is greatest 4. The highest rates of penile cancers are found in sub- Saharan Africa, where both HR-HPV and HIV prevalence exceed that of all regions 189

212 globally 2, 4. Rates of cervical cancer in sub- Saharan Africa are also highest worldwide 4, and are almost universally associated with HR-HPV. The HR-HPV infections that ultimately result in cervical cancer are regularly acquired from HR-HPV infected men 2, 16. Here, we examined the association between high viral load and persistence of HR- HPV over a two year period among heterosexual men in Rakai District, Uganda. Among HR-HPV identified at baseline and newly detected HR-HPV at follow-up, we found that high viral load was associated with 6 and 12 but not 24 month HR-HPV persistence among HIV-negative and HIV-positive men. Persistent HR-HPV infections were more common among HIV-positive than HIV-negative men, and viruses with low viral load were detected at higher levels among HIV-positive men at all follow-up study visits. Most HR-HPV infections detected at the baseline visit were no longer detected after one year, however persistence of remaining infections was strongly and significantly associated with viral load up to 12 months. Though differences in prevalence at 24 months were not statistically significant, persistent detection of HR-HPV with high viral load at baseline was more frequent among HIV-positive and HIV-negative men after two years. Our findings are similar to reports among women in which both higher viral load at first detection and increasing viral load across study visits have been associated with long term persistence of HR-HPV and incidence of cervical lesions 6, 25, 26. High-HR HPV viral load in men may be associated with transmission of HR-HPV to their female partners 16, and may be due in part, to the longer duration of high viral load HR-HPV infections among males. 190

213 We also conducted a secondary analysis among men in whom HR-HPV was newly detected at 6 or 12 month visits. Such newly detected HR-HPV may be a combination of incident infection and reactivation of low-level persistent HR-HPV infection 27. While prevalent HR-HPV infections identified at baseline likely over represent persistent infections 5, 6, we observed very similar patterns in HR-HPV persistence between prevalent infections at baseline and new detections identified during follow-up suggesting that bias in the baseline results may be limited if most newly detected viruses in fact reflect incident infection. Some studies have found that HR-HPV 16 viral load is more strongly associated with persistent viral detection in women than viral load of other genotypes 5, 7, 28, but we observed no significant differences in persistence between genotypes in this study, though power was limited. Our results also suggested that younger and unmarried men were more likely to have high viral load among HIV-negative men, possibly because limited prior exposure is associated with a less robust immunologic response to HR-HPV acquired for the first time. HR-HPV persistence was more common among HIV-positive men and was associated with increased viral load. While viral load was significantly associated with persistent detection of HR-HPV initially detected at baseline in HIV-positive men, the effect estimates were lower than those observed among HIV-negative men. Moreover, we found no significant differences in HR-HPV persistence by viral load among newly detected HR-HPVs in HIV-positive men possibly because of reduced cellular immune responses needed for viral clearance and in the case of latent HPV infection, control of 191

214 viral replication 29. Clearance of HR-HPV is increased with circumcision among HIVnegative, but not among HIV-positive men 13. However, we found that the association between circumcision and HR-HPV persistence among HIV-negative men was independent of viral load. While a previous study showed that co-infection with multiple HPV modified the association between viral load and persistence among women 6, we did not observe any significant differences between men with single or multiple infections. There are inherent difficulties in the study interpretation of HPV epidemiology, particularly when the interval between sample collection is long. We may have misclassified persistent but low-level HR-HPV that were undetectable as cleared infection, and HR-HPV that were rapidly cleared and re-acquired during intervals as persistent infections. We also may have misclassified persistent infection as new events if we did not detect HR-HPV infection at the previous study visit. Newly detected virus may also reflect reactivation of latent HR-HPV infection after a loss of local immune response to viral infection, which is likely more common among those with increasing age and HIV co-infection 27, 29. Specifically, increased reactivation of latent HR-HPV in the presence of HIV co-infection could explain the lack of association between high viral load and persistent detection for newly detected virus in HIV-positive men observed here. There were other limitations to this study. We did not have samples for all men at each follow-up visit, and we have previously shown detection of beta-globin is significantly lower among uncircumcised men suggesting difficulty obtaining cellular material following circumcision 18. In women, HPV persistence and progression to 192

215 cervical lesions is associated with lower CD4 cell count 30. However, only men with a CD4 cell count >250 cells/mm 3 were included in the original trials, limiting any conclusions on the role of CD4 cell count and persistence in this study. We use a dichotomous measure of HR-HPV viral load, though we assessed the sensitivity of the findings to this exposure definition and found very similar results when only those viruses with the highest band intensity were classified as high viral load. Our results were also robust to adjustment for behavioral and clinical variables that may have confounded the association between viral load and persistent detection in follow-up; however, we were not able to adjust for smoking which may have confounded our results to some extent. In conclusion, we find that HR-HPV viral load is associated with persistent detection of HR-HPV infections in HIV-negative and HIV-positive men. Our finding highlight the role of viral load in the natural history of HR-HPV in men, and may explain in part increased levels of HR-HPV detection and persistence among individuals co-infected with HIV. 193

216 6.6. References 1. Tobian AA, Gray RH. Male foreskin and oncogenic human papillomavirus infection in men and their female partners. Future Microbiol 2011;6(7): Partridge JM, Koutsky LA. Genital human papillomavirus infection in men. Lancet Infect Dis 2006;6(1): Munoz N, Bosch FX, de Sanjose S, et al. Epidemiologic classification of human papillomavirus types associated with cervical cancer. The New England Journal of Medicine 2003;348(6): Ferlay J, Shin HR, Bray F, et al. Estimates of worldwide burden of cancer in 2008: GLOBOCAN Int J Cancer 2010;127: Gravitt PE, Kovacic MB, Herrero R, et al. High load for most high risk human papillomavirus genotypes is associated with prevalent cervical cancer precursors but only HPV16 load predicts the development of incident disease. Int J Cancer 2007;121(12): Xi LF, Hughes JP, Edelstein ZR, et al. Human Papillomavirus (HPV) type 16 and type 18 DNA Loads at Baseline and Persistence of Type-Specific Infection during a 2-year follow-up. J Infect Dis 2009;200(11): Dalstein V, Riethmuller D, Pretet JL, et al. Persistence and load of high-risk HPV are predictors for development of high-grade cervical lesions: a longitudinal French cohort study. Int J Cancer 2003;106(3): Tobian, AA, Graowski, MK, Kigozi, G, et al. Human papilloma virus clearance among males is associated with HIV acquisition and increased dendritic cell density in the foreskin. J Infect Dis 2013; 207(11): Smith JS, Moses S, Hudgens MG, et al. Increased risk of HIV acquisition among Kenyan men with human papillomavirus infection. J Infect Dis 2010;201(11): Backes DM, Bleeker MC, Meijer CJ, et al. Male circumcision is associated with a lower prevalence of human papillomavirus-associated penile lesions among Kenyan men. Int J Cancer 2012;130(8): Giuliano AR, Lu B, Nielson CM, et al. Age-specific prevalence, incidence, and duration of human papillomavirus infections in a cohort of 290 US men. J Infect Dis 2008;198(6): Giuliano AR, Lee JH, Fulp W, et al. Incidence and clearance of genital human papillomavirus infection in men (HIM): a cohort study. Lancet 2011;377(9769):

217 13. Tobian AA, Kigozi G, Gravitt PE, et al. Human papillomavirus incidence and clearance among HIV-positive and HIV-negative men in Rakai, Uganda. AIDS 2012;26(12): Flores R, Lu B, Nielson C, et al. Correlates of human papillomavirus viral load with infection site in asymptomatic men. Cancer Epidemiol Biomarkers Prev 2008;17(12): Bleeker MC, Hogewoning CJ, Voorhorst FJ, et al. HPV-associated flat penile lesions in men of a non-std hospital population: less frequent and smaller in size than in male sexual partners of women with CIN. Int J Cancer 2005;113(1): Bleeker MC, Berkhof J, Hogewoning CJ, et al. HPV type concordance in sexual couples determines the effect of condoms on regression of flat penile lesions. Br J Cancer 2005;92(8): Widdice L, Ma Y, Jonte J, et al. Concordance and Transmission of Human Papillomavirus Within Heterosexual Couples Observed Over Short Intervals. J Infect Dis 2013;. 18. Tobian AA, Serwadda D, Quinn TC, et al. Male circumcision for the prevention of HSV-2 and HPV infections and syphilis. N Engl J Med 2009;360(13): Gray RH, Kigozi G, Serwadda D, et al. Male circumcision for HIV prevention in men in Rakai, Uganda: a randomised trial. Lancet 2007;369(9562): Nowak RG, Gravitt PE, Morrison CS, et al. Increases in human papillomavirus detection during early HIV infection among women in Zimbabwe. J Infect Dis 2011;203(8): Gray RH, Kigozi G, Serwadda D, et al. Male circumcision for HIV prevention in men in Rakai, Uganda: a randomised trial. Lancet 2007;369(9562): Gravitt PE, Peyton CL, Apple RJ, et al. Genotyping of 27 human papillomavirus types by using L1 consensus PCR products by a single-hybridization, reverse line blot detection method. Journal of Clinical Microbiology 1998;36(10): Wilson L, Gravitt P, Tobian A, et al. Male circumcision reduces penile high risk human papillomavirus (HPV) viral load in a randomized clinical trial in Rakai, Uganda. Sex Transm Infect 2012;In press. 24. Wentzensen N, Gravitt PE, Long R, et al. Human papillomavirus load measured by linear array correlates with quantitative PCR in cervical cytology specimens. J Clin Microbiol 2012;50(5):

218 25. Marks M, Gravitt PE, Utaipat U, et al. Kinetics of DNA load predict HPV 16 viral clearance. J Clin Virol 2011;51(1): Monnier-Benoit S, Dalstein V, Riethmuller D, et al. Dynamics of HPV16 DNA load reflect the natural history of cervical HPV-associated lesions. J Clin Virol 2006;35(3): Gravitt PE, Rositch AF, Silver MI, et al. A cohort effect of the sexual revolution may be masking an increase in human papillomavirus detection at menopause in the United States. J Infect Dis 2013;207(2): Xi LF, Hughes JP, Castle PE, et al. Viral load in the natural history of human papillomavirus type 16 infection: a nested case-control study. J Infect Dis 2011;203(10): Stanley M. Immune responses to human papillomavirus. Vaccine 2006;24 Suppl 1:S Strickler HD, Burk RD, Fazzari M, et al. Natural history and possible reactivation of human papillomavirus in human immunodeficiency virus-positive women. J Natl Cancer Inst 2005;97(8):

219 Table 1: Associations between qualitative HPV viral load and demographic and clinical factors among HR-HPVs (n=802) detected at baseline HIV-negative (N=403) HIV-positive (N=399) All HR-HPVs (N=802) Qualitative viral load Unadjusted analysis Qualitative viral load Unadjusted analysis Unadjusted analysis (N=802) Low (1-2) High (3-4) PRR (95% CI) Low (1-2) High (3-4) PRR (95% CI) PRR (95% CI) N (%) N (%) N (%) N (%) All HR-HPVs 200 (48.1) 203 (52.6) 216 (51.9) 183 (47.4) Age (years) Marital status (28.0) 85 (41.9) 1.00 (Ref.) 34 (15.7 ) 28 (15.3) 1.00 (Ref.) 1.00 (Ref.) (36.5) 61 (30.1) 0.76 ( ) 46 (21.3) 52 (28.4) 1.12 ( ) 0.88 ( ) (21.5) 30 (14.9) 0.68 ( ) 69 (31.9) 56 (30.6) 0.98 ( ) 0.78 ( ) (14.0) 27 (13.3) 0.82 ( ) 67 (31.0) 47 (25.7) 0.91 ( ) 0.79 ( ) Not married 16 (8.0) 36 (17.7) 1.00 (Ref.) 11 (5.1) 12 (6.6) 1.00 (Ref) 1.00 (Ref.) Currently married 181 (90.5) 162 (79.8) 0.68 ( ) 164 (75.9) 144 (78.7) 0.89 ( ) 0.73 ( ) Previously married 3 (1.5) 5 (2.5) 0.90 ( ) 31 (19.0) 27 (14.8) 0.75 ( ) 0.66 ( ) Number of sexual partners during the past year Condom use (48.5) 101 (49.8) 1.00 (Ref.) 113 (52.3) 85 (46.6) 1.00 (Ref) 1.00 (Ref) 2 63 (31.5) 69 (34.0) 1.01 ( ) 53 (29.2) 63 (34.4) 1.16 ( ) 1.09 ( ) 3 23 (11.5) 24 (11.8) 1.00 ( ) 27 (12.5) 21 (11.5) 1.02 ( ) 1.01 ( ) (8.5) 9 (4.4) 0.66 ( ) 13 (6.0) 14 (7.7) 1.21 ( ) 0.91 ( ) Never 90 (45.2) 96 (48.0) 1.00 (Ref.) 107 (49.8) 81 (45.0) 1.00 (Ref.) 1.00 (Ref.) Sometimes 85 (42.7) 83 (41.5) 0.97 ( ) 95 (44.2) 92 (51.1) 1.15 ( ) 1.05 ( ) Always 24 (12.1) 21 (10.5) 0.87 ( ) 13 (6.1) 7 (3.9) 0.81 ( ) 0.89 ( ) Genital ulcer disease (selfreported) No 188 (94.0) 177 (87.2) 1.00 (ref) 143 (66.2) 133 (72.7) 1.00 (Ref.) 1.00 (Ref.) Yes 12 (6.0) 26 (12.8) 1.43 ( ) 73 (33.8) 50 (27.3) 0.84 ( ) 0.98 ( ) 197

220 Urethral discharge (self-reported) HSV-2 status No 192 (96.0) 190 (93.6) 1.00 (Ref.) 174 (80.6) 148 (80.9) 1.00 (Ref.) 1.00 (Ref.) Yes 8 (4.0) 13 (6.4) 1.25 ( ) 42 (19.4) 35 (19.1) 0.98 ( ) 1.02 ( ) Negative 106 (53.0) 108 (53.2) 1.00 (Ref.) 32 (14.8) 36 (19.7) 1.00 (Ref.) 1.00 (Ref.) Baseline indeterminate 20 (10.0) 29 (14.3) 1.15 ( ) 29 (13.4) 20 (10.9) 0.77 ( ) 0.97 ( ) Trial Arm Prevalent positive 74 (37.0) 66 (32.5) 0.94 ( ) 155 (71.8) 127 (69.4) 0.85 ( ) 0.90 ( ) No circumcision 107 (53.5) 109 (53.7) 1.00 (Ref.) 121 (56.0) 88 (48.1) 1.00 (Ref.) 1.00 (Ref.) Circumcision 93 (46.5) 94 (46.3) 0.99 ( ) 95 (43.4) 95 (51.9) 1.18 ( ) 1.08 ( ) Multiple HR-HPV infections* No 89 (44.5) 88 (43.4) 1.00 (Ref.) 35 (16.2) 27 (14.8) 1.00 (Ref.) 1.00 (Ref.) Yes 111 (55.5) 115 (56.6) 1.02 ( ) 181 (83.8) 156 (85.2) 1.06 ( ) 1.00 ( ) Detection of beta-globin at baseline Yes 168 (84.0) 179 (88.2) 1.00 (Ref.) 207 (95.8) 176 (96.2) 1.00 (Ref.) 1.00 (Ref.) No 32 (16.0) 24 (11.8) 0.77 ( ) 9 (4.2) 7 (3.83) 0.95 ( ) 0.87 ( ) *HR-HPVs in an individual with multiple HR-HPVs (between 2 and 9 co-infecting viruses). 198

221 Table 2: Persistence of high vs. low viral load baseline HR-HPVs at 6,12, and 24 months follow-up HIV-negative (N=403) Qualitative viral load at baseline Missing Low (1-2) High (3-4) Unadjusted analysis Adjusted analysis* Persisting Viruses/Total viruses Persisting Viruses/Total viruses N (%) (%) PRR (95% CI) P PRR (95% CI) P Month /168 (20.8) 74/179 (41.3) 1.79 ( ) ( ) <0.001 Month /144 (15.3) 54/160 (33.8) 1.94 ( ) ( ) <0.001 Month /148 (8.8) 25/165 (15.2) 1.61 ( ) ( ) HIV-positive (N=399) Qualitative viral load at baseline Low (1-2) High (3-4) Unadjusted analysis Adjusted analysis* Persisting Viruses/Total viruses Persisting Viruses/Total viruses (%) (%) PRR (95% CI) P PRR (95% CI) P Month /166 (36.8) 70/142 (49.3) 1.35 ( ) ( ) Month /163 (21.4) 51/141 (36.2) 1.72 ( ) ( ) Month /204 (25.0) 57/164 (34.8) 1.35 ( ) ( ) All HR-HPVs (N=802) Qualitative viral load at baseline Low (1-2) High (3-4) Unadjusted analysis Adjusted analysis* Persisting Viruses/Total viruses Persisting Viruses/Total viruses (%) (%) PRR (95% CI) P PRR (95% CI) P Month /334 (28.7) 144/321 (44.9) 1.51 ( ) < ( ) <0.001 Month /307 (18.6) 105/301 (35.2) 1.83 ( ) < ( ) <0.001 Month /352 (18.2) 82/329 (24.9) 1.34 ( ) ( ) *Adjusted for age, circumcision status, detection of beta-globin at baseline (present or absent), and sex with non-marital partners in the last year. 199

222 Figure 1: Viral load of HR-HPV detected at baseline at 6, 12, and 24 months follow-up visits by baseline HR-HPV viral load and HIV status. 200