HIV-1 diversity and prevalence differ between urban and rural areas in the Mbeya region of Tanzania

Objective: To characterize HIV-1 strains in a potential vaccine trial cohort (CODE) in the Mbeya region of southwest Tanzania.

Design: Study volunteers (n = 3096) were recruited from urban areas in Mbeya Town, using two different recruitment strategies, and in a nearby rural village.

Methods: Cryopreserved plasma from 507 HIV-1 prevalent cases was the source of viral RNA for HIV-1 genotyping by the Multi-region Hybridization Assay, the MHA_acd, and selected strains were confirmed by complete genome sequencing.

Results: The overall HIV-1 prevalence was 16.6% [95% confidence interval (CI), 15.3-17.9] within the cohort. HIV-1 prevalence was higher among women, and in urban areas. Recruitment through advertisement targeted a high-risk urban male population for HIV-1 infection [adjusted odds ratio (adj. OR), 1.68; 95% CI, 1.13-2.51] when compared with men recruited door-to-door. The complexity of the HIV-1 epidemic was also higher in urban areas evidenced by the high-risk of HIV-1 infection with a recombinant strain (adj. OR, 2.69; 95% CI, 1.08-6.69) and HIV-1 dual infection (adj. OR, 5.16; 95% CI, 1.07-24.9), mainly driven by urban men recruited through advertisement.

Conclusions: Overall the urban epidemic was more genetically complex, with higher prevalence and more recombinants and dual infections. Vaccine trials in Mbeya region can assess a complex HIV-1 population dynamic and determine vaccine efficacy in relationship to the genetic diversity of HIV-1 strains that challenge vaccines.

Introduction

Inter-subtype recombinant HIV-1, and the co-infections that generate them, are a key feature of the HIV-1 pandemic. Recombination following dual infection with genetically distinct subtypes or strains of HIV-1 is now recognized with increasing frequency in the global epidemic [^1-4; Piyasirisilp S, et al. 2005, unpubl. results] HIV-1 risk behaviors and community prevalence often correlate with the frequency of HIV-1 dual infection [Herbinger KH, et al. 2005, unpubl. results]. Population-based studies are essential to assess the impact of HIV dual infection in vaccine trials and its role in the genesis of recombinant strains. The high degree of genetic variation, maintained and expanded through mutation and recombination, poses a challenge for AIDS vaccine development ^[5].

In Tanzania HIV-1 subtypes, A, C, and D, have been documented ^[6,7]. Full-length sequencing has provided new insight into HIV-1 diversity in Tanzania, providing an accurate estimate of the proportion of recombinants in the region ^[7-11]. The level of complexity imposed by recombinant strains, and the number of individuals to be genotyped, is rising far beyond the capacity of full-genome sequencing alone. The development of a multi-region hybridization assay (MHA), able to determine HIV-1 subtypes A, C and D in multiple genome regions and efficiently discriminate pure from recombinant strains, offers a high throughput alternative for the surveillance of HIV-1 in East Africa ^[12]. This assay is ideal for HIV vaccine cohort development, where large population-based samples must be addressed.

The Mbeya region of Tanzania is the site of an ongoing AIDS control program and several important studies in preparation for vaccine trials. Within the framework of the Mbeya Medical Research Programme (MMRP), the Cohort Development (CODE) study was designed to ascertain the incidence and prevalence of HIV-1 infection among urban and rural populations in the Mbeya region, and to compare the success of enrollment and retention using two different recruitment strategies. Our sub-hypothesis was that urban and rural areas might reflect somewhat distinct HIV-1 epidemic dynamics, resulting in differences in HIV-1 prevalence, subtype distribution and frequency of dual infections. In this study we present the HIV-1 prevalence and molecular epidemiological results from seroprevalent cases, using MHA_acd on 507 subjects and phylogenetic analysis of 14 complete genome sequences.

Materials and methods

Study population

From September 2002 to April 2003, 3096 volunteers were recruited in the south-western region of Mbeya, Tanzania. Two different recruitment strategies were used. Volunteers from Ghana ward, an urban area in the heart of Mbeya Town, and from a small rural village called Itende, were recruited in a door-to-door campaign. The third subgroup of volunteers was recruited through public advertisement (Advert) in all wards of Mbeya Town. Sera were tested with dual enzyme-linked immunosorbent assay strategy (HIV-Determine; Abbot Laboratories, Abbot Park, Illinois, USA and Enzygnost HIV-1+2 plus; Behring, Liederbach, Germany) and confirmed by HIV-1 Western blot (HIVblot 2.2 Genelabs/Abbott, Wiesbaden, Germany). Samples were identified only by a specimen alphanumeric identification code, CO0001-CO2999 from Ghana, CO3000-CO5999 from advertisement in Mbeya Town, and CO6000 and above from Itende. Peripheral blood mononuclear cells (PBMC) and plasma samples were used for full-length sequencing and genotyping by MHA_acd, respectively. All volunteers completed informed consent, and the study was reviewed and approved by the human subject ethics and safety committees in compliance with all relevant federal guidelines and institutional policies.

MHA_acd version 2

Viral RNA was extracted from 200 μl of plasma using the MagNA pure total nucleic acid robotic extraction procedure (Roche Diagnostics Corp., Indianapolis, Indiana, USA). Negative controls, phosphate buffer saline (PBS) and HIV-1 seronegative samples were included in each extraction procedure and in the analysis. For the RNA-MHA_acd, amplification from the plasma RNA was performed as described by Arroyo et al. ^[7]. Each amplicon was then distributed to three second-round polymerase chain reactions (PCRs), each with a different fluorescent subtype-specific probe, in a TaqMan real-time PCR format. A fourth real-time PCR reaction containing 2.5 μl of reverse transcriptase (RT)-PCR product, 600 nmol/l primers and TaqMan Syber Green Mix (Applied Biosystems, Foster City, California, USA), was carried out to determine whether sample amplification had occurred. Second round real-time PCR amplification was performed on a 384-well spectrofluorometric ABI 7900HT sequence detection system (Applied Biosystems) as described by Arroyo et al. ^[7]. Fluorescence intensity was monitored during the reaction and was analyzed using the SDS v2.1 software (Applied Biosystems). Following amplification in the presence of Syber Green, dissociation curve analysis was performed to confirm the identity of PCR amplicons. The dissociation curve was carried out using the following conditions: 95°C for 15 s, 30°C for 15 s and 95°C for 15 s. The identity of amplification products was determined by the melting temperature (T_m) deduced from the dissociation curve plot. PCR primers and probes described by Hoelscher et al. ^[12] were used for the MHA_acd with the following modifications. For the pol region, a new probe PolA.es1 (position 2121-2145 according to the HXB2 numbering) 5′-CTGYTCTGA GGAAAATTYCCTGGC-3′, was used. For the gp41 region, first round reverse primer, Jl88 (9038-9062) 5′-TAAGTCATTGGTCTTAAAGGTACCTG-3′, second round forward primers, mhgp3.esA/D (8559-8580) 5′-RCG AGGAYTGTGGAA CTTCTG-3′, and mhgp3.esC (8679-8706) 5′-CAATAGCAGTAGCTGAAGGAACAGATAG-3′, second round reverse primers, mhgp4.esA/D (8756-8778) 5′-RGGTAYGTKARAA AWASCTCTA-3′ and mhgp4.esC (8927-8954) 5′-TGTAAGYGCYCCATRTTTITC TAAGTC-3′, and probes, Gp'A'es3 (8684-8708) 5′-CCCTATCTGTCCAGCCAGC TACTGC-3′, Gp'C'es2 (8768-8794) 5′-TTATWGCAAAGCTGC TTCAAAGCCCTG-3′, and Gp'D'es4 (8632-8664) 5′-AAGCTAATAGCACTATTCTT YAGTTCCTGATC-3′ were used.

Sequencing and phylogenetic analysis

Fresh PBMC were isolated by Ficoll-Histopaque separation and cryopreserved. DNA was extracted from 14 randomly selected samples, which according to MHA_acd results represented the most prevalent strains, either subtype C or AC recombinant. Complete genomes were amplified by nested PCR using primers and reaction conditions as described ^[13,14]. Amplicons were sequenced with big dye terminators using an ABI 3100 capillary sequencer (Applied Biosystems), as described ^[14]. Sequencher 3.0 program (Gene Codes Corp., Ann Arbor, Michigan, USA) was used to analyze, edit and assemble sequences. Phylogenetic analysis was performed using the SEQBOOT, DNADIST, NEIGHBOR, CONSENSE modules of the Phylip software package (V3.2c) and TREETOOL ^[15,16].

Statistical analysis

HIV-1 prevalence rates were calculated for men and women stratified by recruitment strategy/location. HIV-1 prevalence was defined as the number of HIV-1 infected individuals per 100 persons tested for HIV-1 infection and is expressed as a percentage. Odds ratios (OR), with 95% confidence intervals (CI), for HIV-1 seropositivity were determined using logistic regression (JMP 5.0 software, SAS Institute, Cary, North Carolina, USA). In multivariate regression, adjusted odds ratios (adj. OR) were calculated adjusting for age.

GeneBank accession numbers

HIV-1 complete genome sequences were identified by year of sampling, country code (TZ), and a specimen alphanumeric identification, and can be found under accession numbers: AY734550-AY734563, inclusive.

Results

Population demographics and HIV-1 prevalence

HIV-1 infection was diagnosed in 514 volunteers at the first visit, establishing an overall prevalence of 16.6% (95% CI, 15.3-17.9). A description of different components of CODE and their HIV-1 prevalences are shown in Table 1. Volunteers were drawn in nearly equal numbers from urban Ghana, urban Advert, and rural Itende. The female/male recruitment ratio in all sites was 1.3, and the average age was 27.7 years with only slight variation among sites (data not shown). The age distribution among females and males was essentially the same, with the majority of volunteers between 17 and 24 years.

Table 1

The HIV prevalence varied almost two-fold in different CODE components, from a low of 11.5% in rural men to over 21.2% in urban women. This broad range permitted further evaluation of prevalence by gender, location, urbanity and recruitment strategy (Table 2). The risk for HIV-1 infection was higher among urban women (adj. OR, 1.87; 95% CI, 1.38-2.55). Recruitment by advertisement in the urban area targeted a high-risk male population (adj. OR, 1.52; 95% CI, 1.03-2.24), when compared with the door-to-door recruitment strategy. No statistically significant difference in risk was observed among urban women between the two recruitment strategies.

Table 2

HIV-1 subtype distribution analysis by MHA_acd

In the MHA_acd, subtype assignment was given when two or more regions showed probe hybridization. Previous results by Arroyo et al., demonstrated that this criteria provided an accurate measure of the subtype distribution when compared with the full-length sequencing genotyping ^[7]. Ninety-five percent of the 507 samples analyzed were successfully genotyped, ranging from two to five regions positive. The overall number of probe hybridization sites per sample genotyped in this study was 3.6 (± 0.95). Similar results were obtained for each of the different groups analyzed, Ghana; 3.6 (± 0.96), Advert; 3.6 (± 0.97), Itende 3.6 (± 0.91). The frequency of the number of probe hybridizations per sample genotyped was 14.1% for two regions; 31.6% for three regions; 35.5% for four regions; and 18.9% for five regions. Sample distribution between these categories was similar among the study groups. These results demonstrated that the performance of the MHA_acd was similar and not biased by the performance of the assay or sample distribution.

The subtype distribution in the 487 samples that could be genotyped was 18% A; 43% C; 3% D; 20% AC; 4% AD; 4% CD; and 8% ACD. Subtype C was the predominant strain, accounting for the majority of pure subtype infections and also contributing substantially to the recombinant component. Among recombinants, more than 85% contained subtype C sequences in their genome and more than 50% were AC recombinants. Figure 1 shows surprising differences in the proportions of strains in different components of CODE. For example, rural men harbored subtype C at a rate of 60%, whereas in their urban counterparts this rate was only 38%. Recombinant strains were more abundant in urban men recruited by advertisement and lowest in rural men, 48 versus 19%, respectively. These disproportions prompted a complete statistical analysis by gender and recruitment area (Table 2).

Fig. 1

The risk for infection with a pure subtype C strain was higher in the rural village of Itende, when compared with both urban groups (adj. OR, 0.40; 95% CI, 0.18-0.89). This difference was only observed among rural men, and was not influence by any of the recruitment strategies. Correspondingly, the risk of infection by recombinant strains was higher among urban men (adj. OR, 2.69; 95% CI, 1.08-6.69), and mainly driven by men recruited through advertisement (adj. OR, 3.84; 95% CI, 1.62-9.12).

MHA_acd genotyping of this cohort was highly successful, with more than 95% of individuals genotyped, and the results add another dimension to the epidemic picture. The lower HIV-1 prevalence rural men was paralleled by less viral complexity, with pure subtype C elevated to 60% of strains and a corresponding decrease in recombinant forms. In contrast, the rural women had an HIV-1 subtype distribution that was almost as complex as that of their urban counterparts.

Estimating the frequency of HIV-1 dual infection

HIV-1 dual infections are suspected when two or more different probes hybridize in any genome region by MHA_acd. The overall frequency of putative HIV-1 dual infections and the differences among the CODE populations are shown in Fig. 2. Putative dual infections were concentrated in the urban sites, especially the Advert component. The overall frequency of putative dual infection among HIV-1 positive individuals was higher in the urban Ghana (13.0%) and, especially, Advert (20.1%) populations, in comparison with the rural (7.6%). Only 12.9% of these cases showed dual probe reactivity in two or more regions of the genome analyzed.

Fig. 2

The urban population also showed a higher risk for HIV-1 dual infection (adj. OR, 5.16; 95% CI, 1.07-24.9) when compared with the rural population. This difference was mainly driven by urban men recruited through advertisement (adj. OR, 8.69; 95% CI, 1.92-39.3), and to a lower but significant degree, by urban men recruited door-to-door and urban women recruited through advertisement. Thus, the low HIV-1 prevalence and low subtype complexity in rural men is reinforced by another aspect, the relative absence of putative dual infections by the MHA_acd assay. The frequency of dual infection was elevated in non-rural areas, especially among men.

Complete molecular characterization of selected HIV-1 strains from CODE

The phylogenetic analysis of complete genomes was in full agreement with the MHA_acd typing results identifying subtype C and AC recombinant strains (Fig. 3). The availability of these sequences permits a detailed evaluation of the accuracy of individual MHA probes (Fig. 3). We calculated the MHA_acd probe specificity by dividing the number of probe-positive samples that hybridized only to the corresponding sequence subtype over the total number of probe hybridizations. Based on this small evaluation, the MHA_acd probe specificity was 98%. The only discrepant result in the envelope region of sample 03TZCO3720, caused by the presence of an atypical probe target sequence not predicted from the overall sequence. This analysis fortifies the MHA_acd results on the overall sample set and provides confidence that the assay provides a highly accurate representation of the proportion of subtypes and recombinant strains in Mbeya region.

Fig. 3

Discussion

Understanding the molecular dynamics of the HIV-1 epidemic is crucial in preparation for vaccine trials in Mbeya, Tanzania. The underlying structure of the HIV-1 epidemic in Mbeya is starkly revealed by the results from CODE. HIV-1 prevalence is higher in urban compared with the rural areas, and the epidemic in women exceeds that in men in the urban centers. The disparity between genders is modest in rural areas and reached its highest level in urban settings. Other studies in the Mbeya region have provided insight on the prevalence and molecular epidemiology of HIV-1 among different risk groups [^7,17; Piyasirisilp S, et al. 2005, unpubl. results; Herbinger KH, et al. 2005, unpubl. results] and provide support for the results presented here. These studies confirm that subtype C is the most prevalent strain and have further demonstrated that pure subtype C strains outnumber recombinant forms in low-risk groups but are closer to parity in diverse high-risk groups. The HIV-1 Super Infection Study (HISIS), conducted among high risk women in Mbeya, Tanzania, has raised awareness of dual infection in the population, and strengthened the link between dual infection and the proportion of recombinant strains [Piyasirisilp S, et al. 2005, unpubl. results]. Dual infection can be a catalyst for an increasing genetic complexity of HIV-1 in populations.

To our knowledge, CODE is the first large cohort study in which HIV-1 prevalence, combined with the frequency of recombination and dual-infection, has been assessed in different geographical sites among community risk individuals. In agreement with previous studies, our results show that the prevalence of HIV-1 infection in Mbeya differs considerably from one geographic area to another, despite their close proximity ^[17]. This is consistent with other sites throughout Africa, where patterns of sexual networks and behaviors may be contributing to the differences observed ^[18]. These differences allowed us to suggest that the sexual networks related to HIV-1 transmission are at least partially, and may be substantially, distinct in urban and rural populations, a contention that is well supported by differences in prevalence, subtype distribution, and frequency of putative dual infections.

The development of the MHA_acd has provided a new avenue to estimate not only subtypes and recombinants, but also the frequency of HIV-1 dual infection in East African countries where subtypes A, C and D are circulating ^[12]. The specificity of detection of dual infection by MHA_acd is at present known to be reasonably high, but cannot be estimated with precision until more samples reach the confirmatory stage. We refer to dual infection detected by MHA_acd as 'putative' to reflect this, and it is our expectation that the majority of these cases will be confirmed. Our study is the first to address the frequency of HIV-1 dual infection among low-risk individuals. In this study the observed frequency of HIV-1 putative dual infection was 14%, 18% in men and 12% in women.

These results can be an overestimation of the real frequency of HIV-1 dual infections due to probe cross-reaction on atypical strains, or may be an underestimation due to highly unequal proportion of strains in any given sample. The confirmation of these HIV-1 putative dual infection cases was beyond the scope of this study, but analysis of selected samples is underway. Putative HIV-1 dual infection results detected by MHA_acd in the higher risk cohort in Mbeya (HISIS) and confirmed by sequence analysis will be reported elsewhere [^19,20: Piyasirisilp S, et al. 2005, unpubl. results].

Are the proportions of subtypes and recombinants determined by MHA_acd sufficiently accurate to support the differences in different genders and population sites that are observed? When comparing the MHA_acd and full-length sequencing genotyping results we observed a 100% agreement in the subtype assigned by the two methodologies. Only sample 03TZCO3720 showed an atypical sequence in the envelope hybridization site, but, because other regions of the strain accurately typed, there was no discrepancy with the subtype assigned by sequencing and phylogenetic analysis. Based on this small subset, the probe specificity was 98%, which is similar to the initial assessment performed in a panel of 45 east African strains ^[12]. We conclude that MHA_acd is a powerful tool to dissect the details of the East African epidemic and to understand variables that may affect vaccine development and evaluation.

In summary, we have found that the HIV-1 epidemic in Mbeya urban areas was more complex than in a nearby rural area. A relationship between HIV-1 prevalence and genetic complexity was also observed in this cohort. Elements of this complexity included more recombinants, and more dual infection cases, in the urban setting. In some of the study parameters, the Advert urban area stood out with higher HIV-1 prevalence and complexity, possibly due to the different method of recruitment used. It is possible that recruitment by advertisement will yield a different population of potential vaccine trial participants.

The complexity observed in Mbeya Town may be explained in part by the accessibility of the local community to the Trans-African highway, which connects Tanzania with neighboring countries and provides an epidemiological route for HIV-1. Communities along the highway, which are known to engage in social interactions and casual sexual contact between travelers and local inhabitants, might also be contributing to the complex epidemiology observed. Geographical features such as mountainous terrain and a broad river, and the more limited availability of motorized transportation among Itende villagers, are some of the factors that might be responsible for the lack of interaction between rural and urban populations, despite their close proximity. The results of this study will be taken into consideration for the development of cohorts in preparation for vaccine trials. The high number of recombinants and dual infection cases in Mbeya, Tanzania will make vaccine trials both complex and informative, centered in a complex social milieu.

Acknowledgements

The authors will like to thank Gudrun Schoen, Eric Sanders-Buell, Sucheep Piyasirisilip and Sodsai Tovanabutra for their contributions. This work was supported by a cooperative agreement between the Henry M. Jackson Foundation for the Advancement of Military Medicine and the United States Department of Defense. The US Military HIV Research Program is jointly planned and funded by the DOD and NIAID/NIH by an interagency agreement. This manuscript was approved for publication by the Walter Reed Army Institute of Research. The views and opinions expressed herein do not necessarily reflect those of the US Army or the Department of Defense.