Cohort Profile: IAVI’s HIV epidemiology and early infection cohort studies in Africa to support vaccine discovery

Matt A Price,* William Kilembe, Eugene Ruzagira, Etienne Karita, Mubiana Inambao, Eduard J Sanders, Omu Anzala, Susan Allen, Vinodh A Edward, Pontiano Kaleebu, Patricia E Fast, Wasima Rida, Anatoli Kamali, Eric Hunter, Jianming Tang, Shabir Lakhi, Gaudensia Mutua, Linda Gail Bekker, Ggayi Abu-Baker, Amanda Tichacek, Paramesh Chetty, Mary H Latka, Pholo Maenetje, Heeran Makkan, Jonathan Hare, Freddie Kibengo, Fran Priddy, Elise Landais, Kundai Chinyenze and Jill Gilmour

Why were the cohorts set up?
In 2003, IAVI (formerly the International AIDS Vaccine Initiative) identified several gaps in HIV epidemiology and vaccine research. IAVI launched cohort studies to better understand at-risk populations, their suitability for clinical trial participation and their unmet needs for preventive services and products. Our goals included (i) improving our understanding of HIV incidence and volunteer retention among 'key populations' of at-risk persons suitable for participation in large-scale HIV prevention trials; (ii) identifying and addressing unmet needs for health care, counselling and prevention among these key populations; (iii) understanding host-virus interactions both shortly after virus acquisition and longer term; (iv) generating data and reagents from recently transmitted HIV to support new vaccine product discovery; and (v) understanding clinical outcomes of HIV disease in the African context to define clinical trial endpoints where antiretroviral therapy (ART) was not (yet) widely available, while building the clinical, laboratory and quality systems to support future trials. To this end, starting in 2004, IAVI established partnerships with nine experienced clinical research centres in east and southern Africa to enrol suitable volunteers ( Figure 1). This manuscript includes data from two broad protocols that followed persons at risk of HIV acquisition (1) 'A prospective, open cohort, observational feasibility study to determine HIV incidence in preparation for future preventive HIV vaccine clinical trials' (IAVI Protocol B) and (2) 'Heterosexual transmission of HIV in Africa' [Emory Heterosexual Transmission (HT) study]. Between 2005 and 2009, varying by research centre, through to December 2011, they served as the source populations for a third cohort study on the natural history of HIV infection, 'A prospective, observational, multi-center study to evaluate laboratory, clinical, immunologic and viral markers of disease progression in recently HIV-infected volunteers' (IAVI Protocol C). Table 1 summarizes the start and stop dates for each site and each cohort. As part of participating in each respective cohort study, clinical and laboratory teams were trained in good clinical practices and good clinical laboratory practices, laboratories were accredited, and all assays were standardized and conducted under an external quality control programme. 1 Study teams met annually, typically in Africa, to share experiences and results and for additional training.

Who is in the cohorts?
To prepare for HIV vaccine efficacy trials, each team began outreach to and recruitment of populations suitable for large-scale prevention trials. Enrolment criteria varied by research centre and cohort, but generally included screening adults (typically 18-49 years old) for risk behaviour associated with an increased risk of HIV acquisition. Protocol B maintained a diverse range of higher-risk study volunteers (see below) whereas the HT study focused primarily on heterosexual transmission risk in stable, HIVdiscordant couples (Table 1).
Starting in late 2004, the team in Masaka began recruiting volunteers in rural villages where previous research studies hadfoundahigherHIVprevalencethantheUgandannational average. 2 Annual HIV incidence was found to be low (1%), and the study team began to recruit the HIV-uninfected partner of HIV-discordant couples in 2006; this remained the sourcepopulationforstudyvolunteersatthisrecruitmentcentre for the duration of this study. 3 Discordant-couple recruit-mentexpandedtoEntebbein2006. 4 Alsoin2004,IAVIbegan the support and expansion of discordant couple cohorts in Rwanda and Zambia, recruited through voluntary counselling and testing (VCT) sessions for couples. [5][6][7] In Kilifi, recruitment began in 2005 with walk-in VCT attendees, then expanded to female sex workers and their clients. As the research centre became recognized as a site for non-judgmental care of sex workers, male sex workers began to attend outreach sessions and asked to be considered for enrolment; thus started the first prospective HIV epidemiology study among men who have sex with men (MSM) in Africa. 8,9 In Nairobi, enrolment began in 2005 and volunteers included female sex workers, their clients, and MSM. 8 In Cape Town, enrolment began in 2006, and because of high regional HIV prevalence and the generalized nature of the epidemic, recruitment included all persons reporting for VCT who reported any sexual activity; the Cape Town team was the only one to include adolescents, with an enrolment age range of 16-40 years old. 8 RustenburginitiatedProtocol B in 2009,also focusing onmen and women with more relaxed 'risk' criteria (any sexual activity). Many of these cohorts were active before recruitment for the Early HIV Infection Cohort began and continued after enrolment for that study ended. Additional details and a wider perspectiveonthisworkhavebeenpublishedelsewhere. 4 Enrolment for Protocol C, the Early HIV Infection Cohort, began in February 2006, but included some volunteers whose HIV infection was diagnosed in 2005 and who were enrolled once the study started. These latter volunteers, diagnosed prior to the start of the Early HIV Infection Cohort, were invited to remain in their respective HIV incidence study where we provided CD4 T cell counts, counselling and appropriate care. We did not collect peripheral blood mononuclear cells (PBMCs) from those volunteers during these visits, as this was not permitted in the HIV epidemiology studies at that time. We did allow (with permission from the volunteers via the consent process) data and samples (in this case, frozen blood plasma only) from the incidence studies to 'roll over' into the Early HIV Infection Cohort Study once it began. Over 90% of Protocol C volunteers were identified via Protocol B and the HT Study, with the remainder either identified at the time of screening for Protocol B by detectable p24 antigen in the absence of HIV antibodies (suggesting incident HIV infection) or through other sources (Table 1, see also 10 ). Suspected transmitting partners were also invited to enrol for a single study visit and 406 partners were enrolled, primarily from the HIV-discordant couple cohorts. In 2006, ART programmes in east and southern Africa were in their infancy, and ART was typically initiated when CD4 T cell counts were 200 cells/mm 3 . These programmes evolved considerably over the course of the Early HIV Infection Cohort, with 17 major guideline changes across five countries from 2005 to 2011. Volunteers' health was monitored closely and ART was initiated per guidelines that existed at the time.
How often have they been followed up?
Volunteers in Protocol B and the HT Study were typically followed quarterly, with a subset deemed at higher risk for HIV acquisition followed monthly for more prevention counselling and HIV testing to detect incident HIV soon after transmission (see 'What has been measured?' below). As volunteers were diagnosed with HIV infection they were invited to enrol in the Early HIV Infection Cohort and their follow-up schedule was determined based on their estimated date of HIV infection (EDI). The EDI was defined as the midpoint between the date of the last negative and first positive test in the case of detection by HIV antibody assay, 14 days prior to the test date in the case of detection by p24 assay only, or 10 days prior to the test date for those volunteers with a polymerase chain reaction (PCR)-positive result prior to antibody or p24 detection. If a volunteer could identify an obvious exposure event, the date of this event could be adopted as the EDI at the discretion of the research team. Once enrolled into the Early HIV Infection Cohort, volunteers were followed monthly for the first 3 months after EDI, quarterly through 2 years post-EDI, and every 6 months thereafter. Of the Early HIV Infection Cohort volunteers, 112 (18%) volunteers were identified very soon after acquiring HIV infection (typically prior to full HIV antibody seroconversion) and were invited into a different visit schedule for the first 3 months: weekly follow-up visits with PBMC collection in the first month following enrolment, then every 2 weeks for the second and third month. Subsequently, their follow-up schedule matched the other Early HIV Infection Cohort volunteers.

What has been measured?
Each HIV epidemiology cohort study measured basic demographics at enrolment, with risk behaviour assessed at   and p24-positive at screening for cohort study participation (n enrolment and at each quarterly visit thereafter. Medical history and physical examination were performed at baseline, genital examination was performed either at baseline or as indicated if sexually transmitted infection (STI) was suspected. HIV testing was done quarterly and followed the national guidelines; typically, they recommended two rapid tests followed by a tiebreaker if needed. A p24 antigen test was also performed to detect HIV infection prior to seroconversion. These data are summarized in Table 2. Plasma was stored from each visit. If incident HIV infection was detected, additional sampling and tests were done, including PCR testing of the preceding study visit sample to detect HIV infection prior to seroconversion ( Table 2). Newly transmitted HIV was subtyped by sequencing the pol region of the genome and transmitted antiretroviral mutations were assessed, as described. 11 Risk was evaluated at month 12 in Protocol B only, and volunteers who were no longer eligible in terms of behaviour risk were taken off study. Volunteers in the HT study continued follow-up while in a sexual relationship with an HIV-positive person. Once enrolled into the Early HIV Infection Cohort, larger volumes of blood were collected to allow processing and storage of plasma and PBMCs, HIV viral load testing, and CD4 and CD8 T cell counts. Human leukocyte antigen system (HLA) characterization was done for all Early HIV Infection Cohort volunteers, as described elsewhere. 12 Data on medication history, including antiretroviral drugs, was also collected. As these volunteers went on antiretroviral therapy, they were followed to assure they were medically stable and enrolled in a treatment programme, then taken off study.

What has been found? Key findings and publications
To date, data and samples from these cohorts have contributed to over 220 peer reviewed manuscripts. Highlights of this research include the following.
• HIV epidemiology: Even in the context of regular counselling and testing (and well before the era of widespread ART availability, pre-exposure prophylaxis and test and treat programmes) HIV incidence remained high. In these cohorts, we observed annual HIV incidences ranging from 1% in rural Ugandans, 4% in Ugandan discordant couples, 3% in Rwandan discordant couples, 8% in Zambian discordant couples, 6-7% in Kenyan MSM, 9-10% in South African women and 2-3% in Kenyan female sex workers. 2-4 ,8 HIV incidence tended to vary by sex, time on study, and sometimes by calendar year though there was considerable heterogeneity across cohorts. 4 Our Kenyan cohort of MSM was the first of its kind in Africa and highlighted many similar risk factors for HIV acquisition to non-African MSM, including unprotected sex, group sex, other STIs (e.g. gonorrhea) and receptive anal intercourse. 13 Working with these men, we have created training modules for health care workers to improve health care delivery and reduce prejudice. 14,15 Additional work included some of the first published reports on transmitted drug resistance mutations in Africa, 11 characterizing pregnancy outcomes in women with HIV infection, 16 describing a novel HLA type associated with favourable clinical outcomes, B*44 (12), and confirmation of other HLA types and other factors in disease progression. 17,18 Ground-breaking work with HIV discordant couples has reinforced the importance of couples' VCT in lowering HIV incidence, [19][20][21] leading the World Health Organization (WHO) to recommend couples' counselling wherever VCT is available. 22 • Clinical course of HIV infection: When the Early HIV Infection Cohort began, national ART programmes in the study countries were just beginning and criteria for when to start varied. Volunteers were followed closely, their CD4 T cell counts and viral loads monitored at every visit and they were referred for treatment according to national guidelines current at the time. The cohorts included regions with very different epidemic dynamics and viral diversity, allowing comparison of clinical outcomes across infecting HIV-1 subtype. We observed that HIV disease progression measured by three endpoints varied by subtype, with C-infected volunteers tending to progress to (i) AIDS, (ii) viral load 100 000 copies/mL and (iii) CD4 T cell count 350 cells/ml faster than those infected with subtype A. 23 Because we also enrolled volunteers with incident HIV infection, typically within 1-2 months of their EDI, we were able to see patterns in acute retroviral syndrome; those infected with subtype A appeared to have worse symptoms shortly after infection, with greater report of headache, lymphadenopathy, fever and other symptoms. 24 T cell decline in very early infection was much more pronounced than expected, with a majority of volunteers falling below 500 cells/mL (the WHO-recommended threshold to start treatment in 2013) within 6 months of acquiring HIV infection. 25 We also observed that 5% of volunteers appeared to control the virus, keeping viral load 2000 copies/mL. This too varied by infecting subtype; those with subtype A were more likely to control the virus compared with those with subtype C. 10 • HIV transmission: Enrolment of suspected transmitting partners allowed for in-depth analysis of events around the time of HIV transmission. We observed that 67-100% of suspected transmitting partners were truly the index case by comparing sequence between partners, and that this varied by study site. 4 Infection is typically established by a single genetic variant from the HIV swarm in the index case; we observed selection bias towards more fit viral variants establishing new infection, and that this bias was increased in men compared with women, suggesting a more permissive transmission environment in the female genital tract. 26 We also observed that these bottleneck events were not as pronounced when inflammation and/or STIs were present, which presumably compromised the mucosal barrier to viral entry allowing transmission of greater numbers and/or diversity of variants. 26,27 Pre-adapted HIV, that is, transmitted HIV that came from someone with a similar HLA profile to the recipient's, was also found to be associated with more negative clinical outcomes-this preadaptation likely grants recently transmitted HIV a level of invisibility to the new host's immune system to which it may, in part, have already been adapted through prior immune escape. 28,29 • HIV virology: Pol sequence was determined at an early timepoint for all volunteers, to estimate the HIV subtype. 12 Efforts to generate full-length sequence and infectious molecular clones from very early samples (typically within 2 months of the EDI) are underway to further define the extent of genetic recombination between regions and subtypes. Viral replicative capacity, as defined by cloning the virus' gag sequence into a replicationcompetent viral backbone (MJ4 and NL-43) and quantifying subsequent viral replication via an in vitro cell culture assay, 30 was found to correlate with immune decline, independent of viral load and HLA type; viruses with high replicative capacity were also found to more readily infect memory T cells, suggesting a more efficient seeding of the latent viral reservoir. 31 • Neutralizing antibodies to HIV infection: Understanding how the body produces neutralizing antibodies to HIV is currently an exciting topic for HIV vaccine design. Although individuals infected with HIV make neutralizing antibodies to the infecting virus, it rapidly escapes through mutations. 32,33 A goal, therefore, of the Early HIV Infection Cohort Study was to characterize the frequency and development of broad and potent neutralizing antibodies to HIV. In this cohort, we observed that 15% of volunteers developed broadly neutralizing antibodies to HIV, typically between 2 and 4 years postinfection. Higher viral loads, lower CD4 T cell counts, HLA A*03 allele and infection with subtype C HIV were all independently associated with the development of neutralizing antibodies. 34 In-depth characterization of the process over time by which these antibodies are developed provides guidance for immunogen design for an HIV vaccine to elicit neutralizing antibodies against a wide array of epitopes. [35][36][37] • Contributing to larger-scale analyses: Samples and data from these studies have contributed to larger work, including the development of assays to estimate HIV incidence from prevalent samples, 38,39 the creation of a repository of reagents of recently transmitted HIV to help researchers and assay developers, 40 and to answer questions about HIV and other infectious diseases in the context of European and global cohorts, including issues of host genetics and viral control, differences between European and African cohorts, and co-infections such as Hepatitis C. [41][42][43][44][45] The samples and data are being used by many African scientists and post-doctorate investigators to develop translational research capabilities, training opportunities and to strengthen north-south and southsouth partnerships within Africa.
What are the main strengths and weaknesses?
These cohorts were set up to both address questions of HIV epidemiology and volunteer retention, as well as recruit volunteers for Protocol C, our early infection cohort.
We enrolled a diverse set of cohorts across a diverse epidemic: HIV subtypes included A, C, D and many recombinant viruses. In part because of our strong relationships with each respective community, we had great success enrolling volunteers with incident HIV-90% of those diagnosed in Protocol B and the HT study enrolled into Protocol C (Table 1). Retention was high in the Early HIV Infection Cohort, with an annual attrition, on average, of 5%. Early timepoints and the collection of PBMCs has enabled us to answer questions related to the events early in infection, from viral dynamics to host immunology. However, because the Early HIV Infection Cohort was set up to answer questions relevant for HIV vaccine trials in an era prior to test and treat (e.g. clinical outcomes that might be amenable to therapeutic vaccines), we did not systematically follow volunteers once they started ART; questions about ART treatment programmes, ART effectiveness and their outcomes may not be well suited to this cohort. Additionally, as costs would have been prohibitive, we did not collect pre-infection PBMCs from all volunteers enrolled into the HIV incidence cohorts and thus have no details on immune status prior to HIV acquisition.
Can I get hold of the data? Where can I find out more?
We are committed to the tenets of Open Data and actively encourage investigators to reach out to us for more information, particularly African scientists from these regions. Data are currently available online and samples can be requested. These data are managed at the IAVI Dataspace, found at https://dataspace.iavi.org/. Data may also be found online with each respective publication that requires adherence to Open Data policies, and HIV sequence data have been submitted to GenBank. More information about IAVI's epidemiology program and how to obtain data and samples from these studies can be found online at https://www.iavi.org/our-work/iavi-dataspace.  • Volunteers were followed quarterly (a 'higher risk' subset had monthly follow-up). Risk behaviour, demographics and medical history were collected. HIV rapid tests and p24 antigen tests (to detect infection before seroconversion) were performed at each visit.

Collaboration and data access
Once HIV infection was detected, the follow-up schedule became more frequent to capture events in early infection; plasma and PBMCs were collected, and health and AIDS-related data were systematically collected.
• Information about IAVI's epidemiology program and how to obtain data and samples from these studies can be found online at https://www.iavi.org/ourwork/iavi-dataspace. A link to the IAVI DataSpace can also be found there (https://dataspace.iavi.org/).