Future of Phylogeny in HIV Prevention

Abstract

The success of the HPTN 052 trial has led to revisions in HIV-1 treatment guidelines. Antiretroviral therapy (ART) may reduce the risk of HIV-1 transmissions at the population-level. The design of successful Treatment as Prevention interventions will be predicated on a comprehensive understanding of the spatial, temporal, and biological dynamics of heterosexual (HET), men having sex with men (MSM), and intravenous drug user (IDU) epidemics. Viral phylogenetics can capture the underlying structure of transmission networks based on the genetic interrelatedness of viral sequences and cluster networks that could not be otherwise identified. This article describes the phylogenetic expansion of the Montreal MSM epidemic over the last decade. High rates of co-clustering of primary infections are associated with one infection leading to 13 onward transmissions. Phylogeny substantiates the role of primary and recent stage infection in transmission dynamics, underlying the importance of timely diagnosis and immediate ART initiation to avert transmission cascades.

INTRODUCTION

Over the past thirty years, HIV-1/AIDS has evolved into an increasingly heterogeneous disease composed of multiple epidemics each influenced by a complex array of biological, behavioral, and cultural factors^1–3. Highly active antiretroviral therapy (HAART), introduced in Western World settings in 1995, has reduced morbility and mortality, stabilizing subtype B male having sex with male (MSM) and intravenous drug user (IDU) epidemics^1–4. Heterosexual (HET) epidemics in Africa have diversified to several major subtypes (A, C, D, F, and G) and circulating recombinant (e.g. CRF01_AE and CRF02_AG) forms.

Global initiatives to scale up ART over the last decade have led to 25–50% reductions in infections in Africa and Asia despite weak health care systems^5–9. The landmark HPTN052 trial showed that earlier ART initiation (550–350 vs. < 250 cells/μl) could result in a 96% reduction in the risk of transmission in HIV serodiscordant couples⁷. The success of HPTN 052, preexposure prophylaxis (PreP) and microbicide trials, as well as observational cohorts, has advanced the concept of “Treatment as Prevention” to avert new infections at a population level^7–18. Guidelines have been revised to reflect these goals, recommending universal annual testing and immediate ART initiation for all persons^{10, 19, 20}.

There remains a debate on the generalisability, feasibility, and sustainability of TasP intiatives^19–21. The resurgence of MSM epidemics and the rise in complex HET/IDU/MSM epidemics in Brazil, east Europe, China and Southeast Asia emphasize the need for tailoring ART with other prevention interventions.

One of the central disputes surrounds the issue as to whether transmissions in early stage infection, frequently undiagnosed, will compromise TasP strategies^22–26. Acute/early stage infection has been postulated to account for 5–70% transmissions depending on epidemiologic and mathematical modelling assumptions^27–30. Epidemiological analysis of MSM transmission dynamics is complicated by patterns of risk behaviour, frequent anonymity of sexual partnerships, low risk of infection per coital act, and long infectivity periods^27–33.

Phylogeny provides a unique framework to capture underlying structures of transmission networks that could not be otherwise identified^{23, 24, 34–40}. Phylogenetics can identify the genetic interrelatedness of viruses in HIV-infected persons^{23, 24, 34–40}. The “clustering” of sequences can infer transmission networks whereby dynamic HIV spread can be assessed on chronological and stage of infection time scales. Phylogenetic cluster analysis can be combined with epidemiological, demographic, and behavioural data to describe the underlying factors contributing to the growth of individual epidemics^{23, 24, 35, 36, 41, 42}.

This article will use phylogenetic findings based on the Montreal MSM cohort to illustrate the role of phylogeny in the design of prevention strategies. Transmission clustering is the driving force of 75% of the MSM wherein one infection can lead to 10 onward transmissions. These findings substantiate the necessity for targeted testing and immediate ART initiation to curb resurgent MSM epidemics^{23, 24, 34, 36, 37, 43–45}.

Phylogenetic analysis of MSM transmission dynamics

The Montreal MSM epidemic began in the early 1980's. By 2008, prevalence rates in sexually active MSM had risen to 15% despite low HIV incidence (0.62 per 100 person-years) with 75% of diagnosed persons receiving HAART⁴⁶. The provincial genotyping began in 2001 and has sequence datasets on half of the diagnosed HIV population. Transmissions dynamics have been assessed based on phylogenetic analysis of co-clustering patterns of newly diagnosed primary infections (subtype-B, male-only) over the last decade. Genotyping requisitions completed by prescribing physicians distinguish primary infection (PHI <6 months post-seroconversion) populations from chronic drug-naïve (PHI >6 months) and treatment-experienced populations^{23, 36}. Viral transmisson clustering has been based on robust criteria of high bootstrap values (>98%), short genetic distance (<1.5%) and similarity in signature mutational motifs.

In 2007, half of primary/early stage infections (PHI < 6 months) were observed to co-cluster with other primary infections although PHIs rarely co-clustered with drug-naïve and treated chronic populations (1% and 2.7% respectively)³⁶. High rates of co-clustering of primary stage cohorts are consistent with frequent re-transmissions among individuals who are recently infected and often unaware of their status^{23, 34, 36–38, 47–51}.

Three phylogenetic patterns of PHI clustering have been observed: unique “dead-end” primary infections, small cluster (2–4 PHI), and large cluster (5–60 PHI) networks (Fig. 1). The growth of the MSM epidemic can be attributed to the step-wise increase in large clustered transmissions, rising from 16 clusters in 2005 (n=140, 9 PHI/cluster) to 60 clusters in 2012 (n=750, 13 PHI/cluster). The cumulative contribution of large clusters to the epidemic has risen from 30% of the epidemic in 2005 to 54% of the epidemic in 2012 (Fig. 1). Unique transmissions have declined from 42% of infections in 2005 to 26% of infections in 2012. Small clusters (2–4 PHI) accounted for the remaining 28% and 20% infections in 2005 and 2012, respectively.

Fig. 1. Phylogenetic inferences on the expansion of the MSM epidemic in Montreal (2002–2012).

Phylogeny shows three patterns of clustering of primary infections. A. A short region in the Maximum-Likelihood tree shows three small clusters, one large cluster, and 13 transmissions. B. The cumulative increase of unique transmissions, small clusters and large clusters over the 2001–2005, 2001–2009, and 2001–2012 periods. C. The cumulative number of infections in the three cluster groups depict the rising role of large clustered transmissions representing 30%, 43% and 54% of transmissions in 2005, 2009, and 2012, respectively.

The temporal growth of individual small and large clusters highlight the role of primary (< 6 months) and early stage infection in onward transmission dynamics. Individual small clusters expanded over median 4.75 month periods with a 1 to 11 month interquartile range (Fig. 2). The temporal expansion of large clusters occurred over a median 11 month period with an 8–21 month interquartile range (Fig. 2). These results are similar to findings in the United Kingdom, Netherlands, and France^{23–25, 34, 36, 37, 52}. Taken together, 25–30% of transmissions in large clusters occur over a 6 month period and half of transmissions occur over a 14–17 month time period (Fig. 2).

Fig. 2. The temporal expansion of small and large clusters.

A frequency histogram plot shows the temporal spread of primary infections associated with small and large clusters, estimated by calculating the branch length (distance in months) of individual primary infections from the median dates of their respective clusters. The overall temporal interval was 4.75 months (1–11.5 month interquartile range) for small clusters and 11.0 months (3.5–25.5 month interquartile range) for large transmission chains.

Relationship to other studies

Comparisons of MSM transmission dynamics has been confounded by the use of different inclusion criteria and methodologies. Molecular phylogeny studies have been assessed using acute/PHI (<6 months) and recent infection (<12–18 months) MSM cohorts, as well as national genotyping programs that include chronic populations and different risk groups (MSM, HET and IDU).^{23, 34, 36–38}. The criteria for designation of transmission “clustering” have varied in bootstrap values (>95–98%) and genetic distance (<.015–0.045) criteria^{34–36, 51}. The rates of co-clustering of MSM early stage infections has varied from 17% to 70% in different regional settings, based on differences in prevalence rates, demographics and depth of sampling^{23, 34, 36–38, 47–51}.

Clustering in most MSM cohorts were related to early stage infection and high CD4 cell count^{36, 44, 51, 53}. The nationwide United Kingdom survey showed 15%, 21% and 15% of infections were inter-linked to 1, 2–10 and >10 infections with high bootstrap values (>95%) and genetic distance below 4.5%²⁴. The transmission interval occurred over median 17 month intervals with 20% of infections occurring over 6 month intervals, confirming the role of early infection in onward transmission^{24, 34}. The Brighton study showed 24% clustering of MSM infections with onward transmission related to recent infection, concomitant sexually transmitted disease, higher viral load, and unawareness of status; clustering was reduced by effective HAART⁵⁴. The Swiss HIV cohort, a mixture of HET, MSM and IDU epidemics, showed 42% overall clustering (bootstrap values >98%)⁵⁵. Inclusion in clusters was associated with MSM transmission (52% clustering) and recent infection (<1 year post-seroconversion, <0.5% ambiguity)⁵⁵.

Phylogenetic inferences re prevention strategies

Transmission clustering is clearly the driving force of MSM epidemics. The patterns of phylogenetic co-clustering implicate a complex interplay of biological, behavioral, and interventional factors in the rise of large cluster transmission cascades. Although 75% of persons may ultimately receive HAART, there remains the precarious ART-free period of early stage infection. The expansion of 60 clusters over 8–21 month intervals is inconsistent with a role of primary stage (<6 months), recent (1 year) and early stage (<24 months) in 25%, 50% and 75% of onward transmission. The duration of clusters indicate that onward transmission is not instantaneous but occurs over an extended period, involving the overlap of persons engaging in low and high risk behaviour. Unawareness of status and poor testing habits are fuelling onward transmission among treatment-naïve individuals.

Although it has been postulated that early stage infection will compromise TasP strategies, our findings argue that it is the delay in ART initiation that has contributed to the episodic development of new phylogenetic variants capable of overriding severe transmission bottlenecks. The failure to test, link to care, and initiation of early treatment is fuelling the epidemic. This has had dangerous implications in the spread of drug resistance and the introduction of non-B subtypes^56–60. Six large clusters in our cohorts (n= 60, n= 29, n=21, n=9, n=6, n=6) harbor G190A or K103N, conferring resistance to first generation non-nucleoside reverse transcriptase inhibitors⁵⁶. The crossover of non-B subtype HET and MSM epidemics has been rare, although three non-B subtype MSM clusters have arisen in Montreal, including CRF01_AE (n=6) and CRF02_AG and l CRF_AB (n=25) variants^{44, 59}.

Clearly, TasP interventions are needed to curb the development of drug-resistant sub-epidemics. High rates of TDR among drug-naïve MSM and IDU populations have been related to clustering^{56, 58, 59, 61, 62}. This is of concern in resource–poor settings, where stavudine, didanosine, and nevirapine-based regimens may facilitate development of K65R or NNRTI resistance^{63, 64}. Pooled drug resistance testing may be needed to identify emergent resistance in resource-poor settings⁶⁵.

The extended infectiousness of phylogenetic variants in large clusters may be related to multiple factors, including viral homogeneity, extended viremia, immature immune response and risk behaviour among those unaware of status⁶⁶. Fundamental research is needed to characterize the genotypic and phenotypic signatures of unique vs. cluster viral variants^67–72.

These findings argue that the success of TasP will be predicated on timely diagnosis^73–75. SPOT, a Montreal community-based initiative, was begun in 2008, both as an intervention and a research study, to understand structural and attitudinal barriers to frequent testing and linkage to care. The site provides anonymous, rapid testing and individualized motivational counselling. The SPOT findings point to the need to diversify services to reach priority populations who are less likely to use existing services. Half of the individuals seeking testing had not had an HIV test in the previous year. The overall rate of seropositivity was 2.1% (n=36/1718) as compared to the 0.14% seropositivity among MSM in the Montreal area (260,000 annual tests).

No persons at SPOT were identified with acute infections (n= 1682) using nucleic acid antigen testing, suggestive of a limited role of acute infection in transmission dynamics. Eight persons (25%) had primary infection (1–6 months since last test). Sequence-based assays, including nucleotide diversity, X4 env co-receptor usage, and next generation sequencing, as well as cluster association, were used to estimate recency of infection, since half of newly diagnosed persons had not had a test in the previous year^76–80. Overall, 80% of seropositive persons had early stage infection (<1 year) and were potentially infectious. Linkage to care and immediate ART is a viable option to curb the MSM epidemic.

Phylodynamics of HET epidemics

The global expansion of relatively few viral subtypes is indicative of clustering at a global level. Subtype C accounts for half of worldwide infections, distributed mostly in Ethiopia, central and southern Africa, Brazil, India and China^{1, 2, 4}. Subtypes A and CRF01_AE epidemics (17% of global infections), have spread from East Africa into Southeast Asia, China, and former Soviet Union nations through intravenous drug use (IDU), commercial sex work (CSW) and HET networks. Subtypes CRF02_AG and G (13% of global infections) have spread from West and North Africa into Europe^{1, 2, 4}. Subtype D remains mainly localized to Uganda. Subtype F, endemic in Angola, has spread to South America and Romania through MSM, IDU, and/or blood product infections. Newly emerging mosaic recombinant forms are emerging through the crossover of the HET, MSM, and IDU epidemics in different regional settings⁸¹.

The fastest growing epidemics worldwide are the IDU epidemics in Eastern Europe where subtypes A1 and CRF03_AB are most prevalent⁸². In heavily populated regions, including India, China and Southeast Asia, epidemics have rapidly shifted from predominant IDU epidemics to HET and MSM epidemics with selective expansion of subtype C, CRF07_BC, CRF08_BC and CRF01_AE subtypes^{81, 83, 84}.

There remains a paucity of phylogenetic studies on transmission dynamics of HET epidemics at the population level, although temporal cluster dynamics of the domestic subtype C epidemics in the United Kingdom appear to parallel those observed for the Montreal large cluster subset (Fig. 2)³⁸. Phylogenetic clusters are relatively small (2–4 infections) and represent ~20% of transmission events. It will become increasingly important to monitor increased clustering with the extended use of ART in resource-poor settings.

Phylogenetics remain an end-point metric in prevention trials of serodiscordant couples. The HPTN 052, Partners in Prevention, Zambia, and Uganda prevention trials showed that 21%, 26.5%, 13%, and 8%, respectively, of identified transmissions among enrolled couples were phylogenetically unlinked^{16, 17, 28}. Relationships outside partnerships may account for 10–65% of HIV transmissions in sub-Saharan Africa⁸⁵.

The HPTN 052 trial showed that the majority (83%) of linked transmission events involved the subtype C population in Africa, although this group represented only half of recruited participants^{7, 86}. The differential transmissibility of variants may affect the success of different clinical trials. A Botswana study showed that 34% of participants had extended viremia (>100,000 copies/ml) for median periods of 318 days (282–459), although no subtype differences were observed in The Partners to Prevention trial^{87, 88}.

Future directions for phylogeny in prevention

Testing, treatment and other prevention interventions require major public health commitments. Phylogenetics can delineate underlying trends in regional settings to establish evidence-informed decisions⁴⁰. The integration of phylogenetic, epidemiological, clinical, and demographic data will be important in delineating the role of linkage to care, behaviour, socioeconomic factors, and migration on transmission dynamics⁴⁰. Although early stage infection may dominate in regional settings with universal access to health care and ART coverage, significant contributions of chronic stage infections may be related to socioeconomic factors, including lower awareness of status and poor linkage to care and treatment^89–94. Phylogenetic inferences of local epidemics may assist in the design of targeted prevention policies for distinct demographic groups, such as young adults and racial/ethnic minorities^{92, 95, 96}.

The ultimate success of TasP will require improved strategies to target “Seek, Test, Link, Treat and Retain” most-at-risk populations^{97, 98}. Control interventions to limit HIV transmission are predicated on early diagnosis^{27, 29, 30, 33, 54, 99–101}. Rapid testing programs are needed to target most-at risk populations in a timely fashion. In Montreal, the SPOT site represents a MSM community-based initiative offering anonymous testing with peer-group motivational counselling. The newly instituted clinic-based initiative, Actuel-sur-la-Rue, now provides rapid testing for HIV-1 and sexually transmitted diseases testing with linkage to care. The success of both testing initiatives in recruitment will be assessed in real-time by phylogenetic analysis of cluster association and growth over time. Pylogeny will be used to assess the success of early treatment initiatives in reducing rates of clustering at a population level.

Sequence-based assays may be used to better monitor transmission dynamics and evaluate the impact of HIV prevention/intervention trials. The frequency of ambiguous calls in bulk sequencing can serve as a surrogate marker to distinguish recent infection (<0.44% ambiguity in the first year) from chronic infection (predictive value 98.7%)^{77, 79}. Single-genome amplification-direct sequencing, next generation sequencing and high resolution melting assays may be applied in dating the recency of infection and viral evolution in a highly accurate manner^{76, 102–104}.

The upcoming HPTN 071 (PopART) and Mochudi HIV-1 prevention project in Botswana will examine the benefit of early ART on population level HIV-1 incidence in Africa. Phylogenetic analyses may be of assistance in monitoring the success of intervention trials, vis-àvis i) assessment of viral linkage in partnerships; ii) clustering of transmission events and iii) determination of the proportion of new infections attributable to acute and chronic stage infection.

Future research will broaden our knowledge of underlying mechanisms leading to the preferential selection and expansion of transmitted ancestral strains. Phylodynamic inferences will be important in the design, implementation, and assessment of candidate public health, therapeutic and behavioral interventions for the ultimate prevention of new HIV infections.