Phylogenetic Analysis of Human Immunodeficiency Virus from People Who Inject Drugs in Indonesia, Ukraine, and Vietnam: HPTN 074

Abstract

Background

HIV Prevention Trials Network (HPTN) 074 evaluated human immunodeficiency virus (HIV) prevention interventions for people who inject drugs (PWID) in Indonesia, Ukraine, and Vietnam. Study interventions included support for HIV infection and substance use treatment. The study enrolled index participants living with HIV and injection partners who were not living with HIV. Seven partners acquired HIV infection during the study (seroconverters). We analyzed the phylogenetic relatedness between HIV strains in the cohort and the multiplicity of infection in seroconverters.

Methods

Pol region consensus sequences were used for phylogenetic analysis. Data from next-generation sequencing (NGS, env region) were used to evaluate genetic linkage of HIV from the 7 seroconverters and the corresponding index participants (index-partner pairs), to analyze HIV from index participants in pol sequence clusters, and to analyze multiplicity of HIV infection.

Results

Phylogenetic analysis of pol sequences from 445 index participants and 7 seroconverters identified 18 sequence clusters (2 index-partner pairs, 1 partner-partner pair, and 15 index-only groups with 2–7 indexes/cluster). Analysis of NGS data confirmed linkage for the 2 index-partner pairs, the partner-partner pair, and 11 of the 15 index-index clusters. The remaining 5 seroconverters had infections that were not linked to the corresponding enrolled index participant. Three (42.9%) of the 7 seroconverters were infected with more than 1 HIV strain (3–8 strains per person).

Conclusions

We identified complex patterns of HIV clustering and linkage among PWID in 3 communities. This should be considered when designing strategies for HIV prevention for PWID.

Clinical Trials Registration

NCT02935296.

In this cohort of people who inject drugs, we observed clusters of genetically related human immunodeficiency virus (HIV) infections. Most seroconverters acquired HIV infection from a source other than their enrolled injection partner living with HIV. Multiple HIV strains were detected in some seroconverters.

People who inject drugs (PWID) have a high risk of human immunodeficiency virus (HIV) infection. A UNAIDS report in 2014 estimated that 13% of PWID were living with HIV worldwide and that infections in PWID accounted for 30% of new HIV infections outside of sub-Saharan Africa [1]. Drivers of the HIV epidemic among PWID include sharing of injection equipment and lack of access to programs for needle exchange and medically assisted treatment for substance use. Criminalization of drug use and possession may lead to increased risk behaviors among drug users [2].

HIV prevalence among PWID has been reported at rates as high as 53.4% in Eastern Europe and 44.5% in Asia [3]. Between 2010 and 2016, HIV incidence in the general population increased 60% in Eastern Europe and central Asia and decreased 13% in other regions of Asia and the Pacific [4]. In this study, we analyzed HIV from PWID enrolled in a randomized clinical trial conducted in Indonesia, Ukraine, and Vietnam (HIV Prevention Trials Network [HPTN] 074). Indonesia and Vietnam have rapidly expanding HIV epidemics that are primarily concentrated among PWID, female sex workers, men who have sex with men, and transgender women [3]. In 2016, HIV prevalence among PWID was 24.8% in Indonesia and 11% in Vietnam [5]. Ukraine has one of the highest rates of HIV infection in Europe [6]. The HIV epidemic in this country was initially driven by PWID, with drug injection serving as the primary route of HIV transmission [7–9]. In 2016, HIV prevalence among PWID was 21.9% in Ukraine [5]; more recently, heterosexual transmission surpassed injection drug use as the primary mode of HIV transmission [9, 10].

The HPTN 074 trial enrolled PWID living with HIV (index participants) and their injection partners who were not living with HIV. Index-partner groups were randomized to receive either the standard of care (control arm) or supported referral for antiretroviral therapy (ART) and substance use treatment (intervention arm). At 52 weeks of follow-up, the following outcomes in index participants were significantly associated with the study intervention: increased self-reported ART, increased participation in a substance use treatment program, increased viral suppression, and decreased mortality [11]. Seven partners acquired HIV infection during follow-up (seroconverters), all in the control study arm [11]. These results show promise for improving health and reducing HIV incidence in PWID populations.

Phylogenetic analysis has been widely used to identify HIV transmission clusters [12, 13] and genetically linked infections [14, 15]. Phylogenetic analysis of HIV infections among PWID can provide information about the source of new infections in this high-risk group. This information may be helpful for optimizing HIV prevention and treatment interventions for PWID. In this study, we examined the phylogenetic relationships of HIV infections among PWID enrolled in HPTN 074. This included analysis of the genetic linkage of infections in index-partner pairs and the multiplicity of HIV infection in seroconverters.

METHODS

Study Cohort

HPTN 074 was conducted in Jakarta, Indonesia; Kiev, Ukraine; and Thai Nguyen, Vietnam. Index participants had HIV viral loads ≥1000 copies/mL at study screening. Each index participant recruited up to 5 concurrent injection partners not living with HIV. The study aimed to recruit a cohort where at least half of the index participants at each site were ART-naive at study entry.

Laboratory Testing

HIV diagnostic testing, viral load testing, and CD4 cell count testing were performed at study sites [11]. Additional testing was performed retrospectively at the HPTN Laboratory Center (Johns Hopkins University, Baltimore, MD). This included HIV diagnostic testing (repeated for all participants at all visits using a single testing algorithm [11]), HIV viral load testing (repeated for quality control using a single testing platform), HIV sequencing, and phylogenetic analysis.

HIV pol Sequencing

The HIV pol region was sequenced using the ViroSeq HIV-1 Genotyping System, v2.0 (Abbott Molecular, Des Plaines, IL). The system generates 1302 base-pair sequences that encode HIV protease and amino acids 1–335 of HIV reverse transcriptase. Testing was performed using samples collected at study enrollment for index participants and samples collected at the first HIV-positive visit for seroconverters. For seroconversion cases, sequencing was also performed using index samples collected at the visit after enrollment and partner samples collected at the second HIV-positive visit. HIV drug resistance was assessed using ViroSeq system software v3.0.

HIV Subtyping

HIV subtypes were determined by analyzing HIV pol sequences with the REGA v3.0 [16], COMET [17], and RIP [18] subtyping tools and by phylogenetic analysis that included HIV subtype reference sequences from the Los Alamos National Laboratory HIV database [19]. Neighbor-joining phylogenetic trees that included study sequences and HIV subtype reference sequences were reconstructed using PHYLIP v3.695. A subtype was assigned if the same result was obtained using at least 3 of the 4 subtyping approaches noted above.

Phylogenetic Analysis of HIV pol Sequences

A BLAST search of sequences in GenBank was performed to identify the 10 background pol sequences that were most closely related to each pol sequence generated in the study. Duplicated background sequences and study sequences with nucleotide ambiguity >5% were excluded from analysis. Sequences were aligned using MAFFT v6.864 [20]. Recombination analysis was conducted using the Recombination Detection Program 4 [21]. Separate phylogenetic trees were constructed for each study site using RAxML v8.2.10 [22]. Median pairwise Tamura-Nei 93 (TN93) genetic distance was calculated using the R “ape” package [23]. Transmission clusters were identified using phylogenetic methods. Two or more sequences were considered to form a sequence cluster if the branch support was ≥90% and the genetic distance between each sequence and at least 1 other sequence in the cluster was ≤1.5% [24, 25]. The pol sequence clusters were also identified using HIV-TRACE [26], using a pairwise genetic distance threshold of 1.5%. Phylogenetic trees were visualized using interactive Tree of Life v3.0 [27].

Analysis of Genetic Linkage Using HIV env Sequences

Genetic linkage was assessed by analyzing env sequences generated by next-generation sequencing (NGS) using the MiSeq system (Illumina, Inc, San Diego, CA; see Supplementary Material 1). Phylogenetic trees were generated for seroconversion cases (index-partner pairs) and for other indexes in pol sequence clusters. Multiple nucleotide alignments and phylogenetic trees were generated as described above. Participants were considered to have genetically linked HIV infections if their env sequences formed a distinct monophyletic cluster with bootstrap support ≥90%. Demographic, clinical, and behavioral variables were evaluated for participants with linked HIV infections.

Analysis of the Multiplicity of HIV Infection

The multiplicity of HIV infection in seroconverters was evaluated by visual inspection of Highlighter plots constructed using the Highlighter tool [28]. Neighbor-joining phylogenetic trees were constructed using PAUP [28], in combination with mathematical modeling [28] incorporated into Poisson-Fitter v2 tool [29]. Intra-host consensus sequences were calculated using Consensus Maker tool [30]. Sequences were evaluated for APOBEC3G-induced hypermutation using Hypermut v2.0 [31]; hypermutated sequences were excluded from further analysis. Within-individual pairwise genetic distances were calculated as noted above.

Sequence Data

The pol sequences were submitted to GenBank (accession numbers: MK228159-MK228624).

Ethics Statement

Written informed consent was signed by each participant in the HPTN 074 trial. The study was approved by institutional review boards and ethics committees of the participating institutions in the United States and at each study site.

RESULTS

Study Cohort

HPTN 074 enrolled 502 index participants living with HIV (121 in Indonesia, 187 in Ukraine, 194 in Vietnam) and 806 injection partners not living with HIV (mean, 1.6 partners/index). Seven partners acquired HIV infection during the study (3 in Indonesia, 1 in Ukraine, 3 in Vietnam) [11]. HIV pol sequencing was performed for 473 index samples and the 7 seroconverters (Figure 1). Genotyping results were obtained for 467 (98.7%) of the 473 index participants and all 7 seroconverters; additional sequences were obtained for seroconverters and the corresponding index participants (2 sequences per person). Phylogenetic analysis was performed using sequences from 445 index participants (100 from Indonesia, 166 from Ukraine, 179 from Vietnam) and all 7 seroconverters; the major HIV subtypes for these samples were CRF01_AE (Indonesia [n = 103, 83.1%], Vietnam [n = 182, 92.4%]) and A1 (Ukraine [n = 167, 88.8%]). Sequences from 22 index participants were excluded from analysis (3 subtype B, 11 unique recombinant forms, 8 with high sequence ambiguity). Demographic, clinical, and behavioral characteristics of participants included in this study are provided in Supplementary Material 2.

Figure 1.

The HIV Prevention Trials Network 074 study cohort. The flow chart provides an overview of the study cohort, indicating the number and source of HIV sequences used in the analysis. One index participant did not have a sample from the enrollment visit available for testing (indicated with an asterisk); in that case, a sample collected at the screening visit was used for analysis. Abbreviations: HIV, human immunodeficiency virus; VL, viral load.

Identification of pol Sequence Clusters

First, we evaluated the phylogenetic relationships of HIV pol sequences from 445 index participants and 7 seroconverters. This analysis was performed separately for each study site (Figure 2). The median pairwise genetic distance among study sequences was 2.9% for Indonesia, 2.9% for Ukraine, and 3.2% for Vietnam (detailed analysis of regional genetic diversity is described in Supplementary Material 3). Eighteen sequence clusters were identified (Table 1). Two clusters included index-partner pairs; the median genetic distance between the index and partner sequences in these 2 cases was 0.4% and 0.2%. The remaining 5 seroconverters had sequences that did not cluster with the corresponding enrolled index participants; the median genetic distance for the index and partner sequences in those cases ranged from 1.51% to 3.3%. One cluster included 2 seroconverters who were enrolled at the same study site with different index participants (partner-partner pair). The remaining 15 clusters included pairs or groups of index participants (index-index clusters). This included 13 index-index pairs, 1 cluster of 3 index participants, and 1 cluster of 7 index participants; the clusters of 3 and 7 index participants were from Vietnam. In all 18 cases, sequences in each cluster were from participants at the same study site. The same clusters were identified using HIV-TRACE with a genetic distance threshold of 1.5% (data not shown).

Figure 2.

Phylogenetic trees of study and background pol sequences from each study site. Phylogenetic trees were generated using pol gene sequences generated by population sequencing. Trees were generated using RAxML for each study site: (A) Indonesia, 109 study sequences; (B) Ukraine, 169 study sequences; and (C) Vietnam, 188 study sequences. Gray dots corresponding to tree branches represent background (control) sequences obtained from a sequence database (see the Methods section). Dark blue dots represent study sequences from index participants. Light blue dots represent study sequences from injection partners who acquired human immunodeficiency virus (HIV) infection during the study (seroconverters). The letter “C” indicates a cluster of 2 or more index participants (1 sequence per person). The letter “I” indicates sequences from the 7 index participants whose partners acquired HIV infection during the study (I1–I7; 2 sequences per person). The letter “P” indicates sequences from the 7 seroconverters (P1–P7; 2 sequence per person). Numbers indicate that the index and partner were enrolled as part of an injection group (eg, index I1 was enrolled with partner P1). Brackets indicate paired sequences from the same individual obtained from samples collected at different study visits. Branches for sequences in clusters are highlighted with shading. Purple shading indicates index sequences that are part of an index-partner cluster (participants I1–I7); light blue shading indicates seroconverter sequences (participants P1–P7); gray shading indicates clusters of sequences from 2 or more index participants (clusters C1–C15). Note that the cluster designated C15 includes 7 index participants and 2 nonstudy background sequences (C). Abbreviations: I, index; P, partner; C, cluster; RAxML, Randomized Axelerated Maximum Likelihood.

Table 1.

pol Gene Phylogenetic Clusters and Linkage Status Based on Next-Generation Sequencing Analysis of Human Immunodeficiency Virus env

Cluster Type	Cluster Identifier	Study Site	Cluster Size	Median pol Pairwise Genetic Distance, %	Linkage Status (Next-Generation Sequencing, env)
Index-Partner	IP1	Indonesia	2	0.31	Linked
	IP4	Ukraine	2	0.4	Linked
Partner-Partner	PPa	Indonesia	2	0.08	Linked
Index-Index	C1	Indonesia	2	0.0	Linked
	C2	Indonesia	2	0.0	Linked
	C3	Ukraine	2	1.23	Data not available
	C4	Ukraine	2	0.54	Unlinked
	C5	Ukraine	2	1.18	Linked
	C6	Ukraine	2	1.33	Linked
	C7	Ukraine	2	0.0	Linked
	C8	Ukraine	2	0.93	Linked
	C9	Ukraine	2	1.26	Unlinked
	C10	Ukraine	2	0.38	Linked
	C11	Vietnam	2	1.41	Linked
	C12	Vietnam	3	0.33	Linked
	C13	Vietnam	2	1.19	Linked
	C14	Vietnam	2	1.02	Unlinked
	C15b	Vietnam	7	0.16	Linked

The table shows the cluster type, study site, number of participants included in the cluster (cluster size), median pol pairwise genetic distance, and linkage status for 18 pol sequence clusters (see Figure 1). Linkage status was determined by analysis of env sequences generated by next-generation sequencing; linkage could not be determined in 1 case (C3). The 14 groups of linked infections included 2 index-partner pairs (IP1 and IP4), 1 partner-partner pair (PP), 9 index-index pairs (C1, C2, C5-C8, C10, C11, C13), 1 group of 3 index participants (C12), and 1 group of 7 index participants (C15).

Abbreviations: C, cluster; IP, index-partner; PP, partner-partner.

^aThe PP cluster includes 2 partners enrolled with different index participants (P2 and P3).

^bCluster C15 includes sequences from 7 index participants and 2 background sequences.

Analysis of Genetic Linkage Between Participants in Index-Partner Pairs Using env NGS Data

Genetic linkage in the 7 seroconversion cases was analyzed using HIV env sequences generated by NGS (Figure 3). Two of the 7 seroconversion cases were classified as linked (Figure 3A and 3D). The remaining 5 index-partner pairs were characterized as unlinked (Figure 3B, 3C, and 3E–3G).

Figure 3.

Phylogenetic trees of env sequences from index-partner (IP) pairs and a partner-partner (PP) pair with genetically linked human immunodeficiency virus (HIV) infections. Phylogenetic trees were generated using env sequences from next-generation sequencing for 7 seroconversion cases (IP pairs, A–G) and 1 PP cluster (H) are shown. Colors indicate the source of study sequences, as noted in the lower right of the figure. Each tree includes nonstudy HIV env sequences from each participating country and background sequences (black branches). Bootstrap values were obtained for 1000 replicate trees; values ≥90% are shown. Case numbers (IP1–IP7, PP) refer to cases described in Table 1. In 1 case (IP1), the partner sequences clustered with low-level viral variants present in 1 index sample; these variants represented approximately 5% of the env read counts in the index sample.

Analysis of Genetic Linkage Between Participants in pol Sequence Clusters Using env NGS Data

Genetic linkage of HIV from participants in pol sequence clusters was also analyzed using env sequences generated by NGS. This analysis included 2 partners enrolled with different indexes whose pol sequences clustered together (P2 and P3, Figure 3H) and index participants whose pol sequences clustered together (14 of the 15 index-index clusters; 2–7 participants/cluster; there was insufficient plasma for analysis of 1 cluster). Genetic linkage was confirmed for the partner-partner pair (Figure 3H) and for 11 (78.6%) of the 14 index-index clusters. Representative phylogenetic trees are shown for 1 unlinked index-index case and 1 linked index-index case (Figure 4A and 4B). Phylogenetic trees are also shown for the clusters that included 3 and 7 index participants (C12 and C15; Figure 4C and 4D); the infections in these 2 cases were classified as linked. Overall, analysis using NGS env sequences confirmed genetic linkage in 2 index-partner pairs, 1 partner-partner pair, and 11 groups of index participants (Table 1).

Figure 4.

Phylogenetic trees of env sequences from a subset of the index-index clusters. Phylogenetic trees generated using env gene sequences from next-generation sequencing are shown for 4 index-index clusters (see Table 1): a representative unlinked case (cluster C14; A), a representative linked case (cluster C7; B), the cluster involving 3 index participants (cluster C12; C), and the cluster involving 7 index participants (cluster C15; D). Colors indicate the source of study sequences, as noted in the upper left of each panel. Each tree includes nonstudy human immunodeficiency virus env sequences from each participating country and background sequences (black branches). Bootstrap values were obtained for 1000 replicate trees; values ≥90% are shown. Case numbers (C14, C7, C12, and C15) refer to cases described in Table 1.

Characteristics of Participants With Linked HIV Infections

Thirty-four participants had linked HIV infections based on NGS analysis (Table 1; 12 in linked pairs and 10 in larger linkage groups). Participants in pairs or groups of linked infections varied in sex, age, marital status, number of sexual and injection partners, and education level (Table 2).

Table 2.

Characteristics of Participants With Genetically Linked Human Immunodeficiency Virus Infections

Cluster Type	Cluster Identifier	Study Site	Sex	Age, y	Marital Statusa	Number of Injection Partnersb	Number of Sexual Partnersc	Educationd
Index-Partner	IP1	Indonesia	F/M	22*/28	Married/Married	1*	1	Secondary/Higher*
	IP4	Ukraine	F/F	31/33*	Married/Married	2–4*	0*	Secondary/Higher*
Partner-Partner	PPe	Indonesia	M/M	29/36	Single/Married	–	–	Higher/Higher
Index-Index	C1	Indonesia	M/M	27/41	Single/Single	≥5, 2–4	0, 1	Secondary/Higher
	C2	Indonesia	M/M	34/35	Single/Single	2–4, 2–4	0, 1	Higher/Higher
	C5	Ukraine	M/F	31/33	Married/Married	2–4, 1	1, 1	Secondary/Higher
	C6	Ukraine	M/F	36/40	Married/Married	2–4, ≥5	1, 1	Secondary/Secondary
	C7	Ukraine	F/M	30/40	Single/Single	2–4, ≥5	1, 1	Higher/Higher
	C8	Ukraine	M/M	35/39	Single/Married	1, 2–4	0, 1	Higher/Secondary
	C10	Ukraine	F/M	34/39	Married/Married	2–4, 2–4	1, 1	Secondary/Higher
	C11	Vietnam	M/M	41/45	Single/Married	2–4, 2–4	0, 0	Secondary/Primary
	C12	Vietnam	All M	23–43	All single	1 or 2–4	0 or 1	All secondary
	C13	Vietnam	M/M	29/36	Married/Single	≥5, 1	1, 0	Secondary/Primary
	C15f	Vietnam	All M	24–42	Single or married	1 or 2–4	0, 1 or ≥2	All secondary

The table shows demographic and behavioral characteristics of participants who had genetically linked infections. The 14 groups of linked infections included 2 IP pairs (IP1 and IP4), 1 PP pair (PP), 9 index-index pairs (C1, C2, C5-C8, C10, C11, C13), 1 group of 3 index participants (C12), and 1 group of 7 index participants (C15). Characteristics of study participants are shown. For each group, characteristics are listed in the order of participant age (younger to older); for the 2 IP clusters, index characteristics are noted with an asterisk if the index and partner characteristics were different. Age range is shown for participants in the groups of 3 and 7 individuals.

Abbreviations: F, female; M, male.

^aMarried marital status indicates that the participant reported that he/she was married or had a partner.

^bThe number of injection partners is shown for index participants only; these data were reported for the month prior to enrollment. A dash indicates that these data were not available for injection partners. Aggregate data on the number of injection partners for the 7 seroconverters is provided in Supplementary Material 2.

^cThe number of sexual partners is shown for index participants only; these data were reported for the 3 months prior to enrollment. A dash indicates that these data were not available for injection partners. Aggregate data on the number of injection partners for the 7 seroconverters is provided in Supplementary Material 2.

^dPrimary education indicates that the participant reported that he/she had either no education or a primary education only.

^eThe PP cluster includes 2 partners enrolled with different index participants (P2 and P3).

^fCluster C15 includes sequences from 7 index participants and 2 background sequences.

Analysis of the Multiplicity of HIV Infection in Seroconverters

Next, we analyzed the multiplicity of infection in the 7 seroconversion cases. The median duration of infection for these cases was 50 days (range, 7–117; Table 3). Maximum within-individual env genetic distances in these cases ranged from 0.57% to 5.3% (median, 0.56%–2.3%). In 4 cases (P1, P3, P5, and P7), inspection of Highlighter plots and neighbor-joining phylogenetic trees identified a single distinct low-diversity group of env sequences that corresponded to a single monophyletic lineage. These individuals were most likely infected by a single viral variant or multiple closely related viral variants. In the remaining 3 cases, phylogenetic analysis indicated infection with multiple viral variants; maximum within-individual env diversity in these 3 cases ranged from 2.4% to 5.3% (median, 0.85%–2.3%). The median number of distinguishable viral variants (lineages) in these individuals was 7 (range, 3–8). For comparison, the maximum within-individual env diversity among 41 index participants with available NGS data ranged from 1.2% to 10.7%. Figure 5 shows representative neighbor-joining phylogenetic trees, Highlighter plots, and the distribution of within-individual env pairwise genetic distances for a participant infected with a single HIV variant and a participant infected with multiple HIV variants. Table 3 shows demographic and clinical characteristics, the duration of HIV infection, within-individual env diversity, and the number of viral variants detected for the 7 seroconverters. All 7 seroconverters (including those infected with multiple HIV variants) were infected with a single HIV subtype.

Table 3.

Individual Characteristics and Within-Individual env Diversity Analysis in Seroconverters

								Number of Viral Variants
Case	Study Site	Estimated Duration of Human Immunodeficiency Virus Infection, da	Age at Enrollment, y	Sex	CD4 Cell Count,b cells/mm3	Viral Load,c Log10 copies/mL	Maximum (Median) env Genetic Distanced	Model Basede	Phylogeny Basedf
P1	Indonesia	7	22	F	527	3.7	1.1 (0.56)	1	1
P2	Indonesia	81	36	M	726	5.2	2.6 (0.5)	>1	3
P3	Indonesia	50	29	M	338	6.7	0.57 (0.57)	1	1
P4	Ukraine	61	31	F	486	3.7	2.4 (0.8)	>1	8
P5	Vietnam	19	22	M	522	5.2	1.1 (0.57)	1	1
P6	Vietnam	117	27	M	223	5.6	5.3 (2.3)	>1	7
P7	Vietnam	7	32	M	501	5.1	1.7 (0.57)	1	1

The table shows demographic characteristics, clinical characteristics, behavioral characteristics, within-individual env genetic diversity, and the number of viral variants detected for 7 partners who acquired human immunodeficiency virus (HIV) infection during the HIV Prevention Trials Network 074 study (seroconverters).

Abbreviation: F, female; M, male; P, partner.

^aSamples used for phylogenetic analysis were collected at the first and second HIV-positive visits. The duration of HIV infection was calculated as mid-time point in days between last HIV-negative visit and the first HIV-positive visit, except in 2 cases. Cases P1 and P7 had acute infection at the first HIV-positive visit; in those cases, the estimated duration of infection was classified as 7 days prior to the acute infection visit.

^bCD4 cell count testing was performed a median of 5 days after the first HIV-positive visit (range, 2–67 days).

^cViral load testing was performed at the first HIV-positive visit, except in 2 cases: case P1 had viral load testing performed 7 days after the first HIV-positive visit; case P7 had viral load testing performed 63 days after the first HIV-positive visit.

^dThe env pairwise genetic distances were calculated using the dist.dna function in R “ape” package.

^eModel by Keele et al 2008 [28].

^fThe number of phylogenetically distinct HIV strains (viral lineages) was determined by visual inspection of Highlighter plots and neighbor-joining phylogenetic trees.

Figure 5.

Analysis of within-individual env diversity. The figure shows representative neighbor-joining phylogenetic trees (top panels), Highlighter plots (middle panels), and histograms of the distribution of pairwise genetic distances (lower panels). Data are shown for a representative participant infected with a single human immunodeficiency virus (HIV) variant (P5; A) and a representative participant infected with multiple HIV variants (P6; B). Highlighter plots show the number of sequence mismatches as a function of nucleotide position (env alignment position) compared with an intrahost consensus sequence (master). Nucleotide mismatches are shown using the following colors: thymine, red; adenine, green; cytosine, blue; and guanine, orange. The histograms show the distribution of within-individual pairwise genetic distance; the red vertical line shows the median value. The same results were obtained when the analysis was performed using maximum-likelihood (RAxML) phylogenetic trees. Abbreviation: m, master.

DISCUSSION

Analysis of HIV strains from PWID in Indonesia, Ukraine, and Vietnam revealed complex HIV dynamics. In 5 of 7 seroconversion cases, linkage analysis indicated that the partner acquired HIV infection from a source other than the enrolled index participant. In addition, 2 seroconverting partners had infections that were genetically linked to each other, and 11 index participants had infections that were linked to infections in other index participants (9 index-index pairs and 2 larger index-index clusters). All 7 seroconverters in HPTN 074 were in the control (standard-of-care) study arm; this suggests that the study intervention may have averted some HIV transmission events [11]. These results must be interpreted cautiously, however, since the study was not designed to have sufficient power to show an effect of the intervention on transmission and since the control group was larger than the intervention group. In the 5 unlinked seroconversion cases, the partner’s infection would not have been directly averted by interventions targeted to their enrolled index partner. However, study interventions may have indirectly impacted HIV risk among enrolled partners (eg, by reducing HIV transmission to individuals who were not enrolled in the study but were also members of a partner’s extended injection network).

In this study, we used env sequences from NGS to establish genetic linkage of infections. A similar approach was used in the HPTN 052 study to identify linked transmission events among HIV discordant heterosexual couples [14]. This analysis complemented the analysis of pol sequences derived from population sequencing. The env sequences provide more specificity for determining genetic linkage because of the high level of genetic diversity in this region. Analysis of sequences from NGS is also more sensitive for detecting linked infections, since this approach can identify cases where 2 individuals share a viral strain that is, at present, at a low level in 1 or both individuals. This approach may be especially important in studies of PWID, since these individuals may be infected with multiple HIV strains.

We also used env NGS sequences to analyze the multiplicity of infection in seroconverters; we detected multiple HIV variants in 3 (42.9%) of 7 cases. More diverse infections may be seen in PWID compared with those with sexually acquired HIV infection, due to the absence of a mucosal barrier for transmission. Higher multiplicity of HIV infection may have clinical implications. In prior studies, HIV infection with multiple viral variants was associated with higher HIV viral loads [32] and faster disease progression [33].

This study had some limitations. First, the number of participants from each study community was limited. A larger sampling would provide more information about transmission dynamics in each region, since some individuals in the injection networks were likely unsampled and since sexual partners of study participants were not included in the study cohort. In this study, 5 (71.4%) of the 7 seroconverters had HIV that was not linked to any index participant in the cohort. Unlinked infections were also observed in the HPTN 052 study; in that study, 36% of seroconverters had infections that were not linked to their enrolled sexual partners [34]. In HPTN 074, a larger sampling size might also have revealed differences in transmission dynamics at the 3 study sites, reflecting structural, cultural, or other differences. Differences in recruitment strategies at the 3 sites may also have impacted enrollment of participants with linked infections. Recruitment in Ukraine was based primarily at a nongovernment organization, recruitment in Vietnam involved considerable community outreach, and recruitment in Indonesia occurred at HIV clinics and within the local community [11]. Differences were also noted in types of injected drugs used at the study sites [35]; use of different drugs and/or injection practices could have impacted the probability of HIV transmission/acquisition. Demographic characteristics were similar across the 3 study sites; notable exceptions were the higher rate of women enrolled in Ukraine and the higher education level of participants in Ukraine and Indonesia compared with Vietnam [35].

This study identified complex patterns of HIV transmission and infection among PWID in Indonesia, Vietnam, and Ukraine, with multiple sources of partner infections. The study also demonstrates that participants with epidemiologically linked infections (enrolled as injection partners) may not represent direct transmission events and that phylogenetic analysis is needed for accurate identification of transmission chains. In this study, individuals outside of the enrolled cohort were the source of infection in 5 seroconversion cases. Further research is needed to evaluate the relative contributions of HIV transmission from injection drug use and sexual transmission among PWID. Larger phylogenetic studies that include intensive sampling of PWID, sexual partners, and others may be useful for understanding the spread of HIV infection among PWID. Heterogeneity of sexual and injection risk behaviors among PWID may also complicate efforts to reduce HIV transmission in this high-risk population. A range of interventions, including ART for prevention, preexposure prophylaxis, medication-assisted substance use treatment, and enhanced education and counseling, may be needed to decrease HIV transmission among PWID.

Supplementary Data

Supplementary materials are available at Clinical Infectious Diseases online. Consisting of data provided by the authors to benefit the reader, the posted materials are not copyedited and are the sole responsibility of the authors, so questions or comments should be addressed to the corresponding author.

Notes

Author contributions. M. V. S. performed human immunodeficiency virus (HIV) sequence analysis, analyzed data, and drafted the manuscript. M. K. G. provided input on HIV sequence analysis and drafted the manuscript. Y. Z. performed laboratory testing and HIV sequence analysis, analyzed data, and drafted the manuscript. P. J. P. performed HIV genotyping, analyzed data, and drafted the manuscript. X. G. was an HIV Prevention Trials Network (HPTN) 074 data analyst. E. P.-M. was an HPTN 074 laboratory center quality assurance-quality control representative. E. H. was a HPTN 074 study coordinator. T. V. H., S. A., D. I., V. G., M. L., and S. D. were HPTN 074 site investigators. I. H. was the HPTN 074 protocol co-chair. W. M. was the HPTN 074 protocol chair. S. H. E. was the HPTN 074 virologist, who was responsible for study design, analyzed data, and drafted the manuscript. All authors contributed to manuscript preparation and reviewed the manuscript before publication.

Acknowledgments. The authors thank the HPTN 074 study team and participants for providing the samples and data used in this study. The authors also thank the laboratory staff at the study sites and HPTN Laboratory Center who helped with sample management and testing.

Financial support. This work was supported by the HPTN sponsored by the National Institute of Allergy and Infectious Diseases, National Institute on Drug Abuse, and Office of AIDS Research of the National Institutes of Health (grant UM1-AI068613 to S. H. E. and grants UM1-AI068617 and UM1-AI068619).

Potential conflicts of interest. S. H. E. has collaborated on research studies with investigators from Abbott Laboratories (distributor of the ViroSeq HIV-1 Genotyping System). Abbott Laboratories has provided reagents for collaborative research studies. All other authors report no potential conflicts. All authors have submitted the ICMJE Form for Disclosure of Potential Conflicts of Interest. Conflicts that the editors consider relevant to the content of the manuscript have been disclosed.

Presented in part: Conference on Retroviruses and Opportunistic Infections. Seattle, WA, 7–10 March 2019. Abstract 1670.