ABSTRACT
HIV's ability to persist during suppressive antiretroviral therapy is the main barrier to cure. Immune-privileged tissues, such as the testes, may constitute distinctive sites of HIV persistence, but this has been challenging to study in humans. We analyzed the proviral burden and genetics in the blood and testes of 10 individuals on suppressive therapy who underwent elective gender-affirming surgery. HIV DNA levels in matched blood and testes were quantified by quantitative PCR, and subgenomic proviral sequences (nef region) were characterized from single templates. HIV diversity, compartmentalization, and immune escape burden were assessed using genetic and phylogenetic approaches. Diverse proviruses were recovered from the blood (396 sequences; 354 nef-intact sequences) and testes (326 sequences; 309 nef-intact sequences) of all participants. Notably, the frequency of identical HIV sequences varied markedly between and within individuals. Nevertheless, proviral loads, within-host unique HIV sequence diversity, and the immune escape burden correlated positively between blood and testes. When all intact nef sequences were evaluated, 60% of participants exhibited significant blood-testis genetic compartmentalization, but none did so when the evaluation was restricted to unique sequences per site, suggesting that compartmentalization, when present, is attributable to the clonal expansion of HIV-infected cells. Our observations confirm the testes as a site of HIV persistence and suggest that individuals with larger and more diverse blood reservoirs will have larger and more diverse testis reservoirs. Furthermore, while the testis microenvironment may not be sufficiently unique to facilitate the seeding of unique viral populations therein, differential clonal expansion dynamics may be at play, which may complicate HIV eradication.
IMPORTANCE Two key questions in HIV reservoir biology are whether immune-privileged tissues, such as the testes, harbor distinctive proviral populations during suppressive therapy and, if so, by what mechanism. While our results indicated that blood-testis HIV genetic compartmentalization was reasonably common (60%), it was always attributable to differential frequencies of identical HIV sequences between sites. No blood-tissue data set retained evidence of compartmentalization when only unique HIV sequences per site were considered; moreover, HIV immune escape mutation burdens were highly concordant between sites. We conclude that the principal mechanism by which blood and testis reservoirs differ is not via seeding of divergent HIV sequences therein but, rather, via differential clonal expansion of latently infected cells. Thus, while viral diversity and escape-related barriers to HIV eradication are of a broadly similar magnitude across the blood and testes, clonal expansion represents a challenge. The results support individualized analysis of within-host reservoir diversity to inform curative approaches.
INTRODUCTION
The HIV reservoir, a small pool of primarily CD4+ T cells that harbor replication-competent virus despite long-term suppressive combination antiretroviral therapy (cART) (1), is the major barrier to cure. During untreated infection, HIV populations evolve within the host (2–5), and individual viral sequences are continually archived into the reservoir as integrated proviruses (6). There, they can persist long term within either the original infected cell or clonal descendants thereof (7–10). Much of our understanding of HIV persistence is derived from HIV sequences isolated from blood during long-term cART; these studies have revealed that the within-host HIV reservoir is genetically diverse (11–13), that it frequently contains immune escape mutations (14, 15), and that it commonly features clonally expanded cell populations harboring identical proviruses (10, 16–19).
The HIV reservoir in blood, however, may not reflect that within tissues. HIV reservoir composition in immune-privileged sites, defined as those that are protected by physical barriers that limit immune cell trafficking (20) and/or that are equipped with mechanisms to suppress immune responses locally (21, 22), may be distinct from that elsewhere, but this has been challenging to study in healthy humans. The testes represent such a site (21, 23). Nonhuman primate studies have demonstrated that the cytokine responses of testicular T cells to mitogen stimulation are lower than those of blood T cells (24, 25); moreover, human testes harbor higher frequencies of CD39+ T-regulatory cells (21), which are capable of suppressing HIV-specific CD8+ T-cell responses (26). Despite this, however, leukocyte populations, including central and effector memory CD4+ T cells that represent potential viral targets, are nevertheless present in the testes (24). Indeed, immunodeficiency viruses penetrate into the testes early during infection (22, 27) and subsequently persist (22, 28–30). As such, the testes have been hypothesized to represent a distinctive HIV sanctuary site (21, 25).
A key outstanding question is whether proviruses persisting in testes are genetically distinct from those in blood and, if so, how. Blood-tissue genetic compartmentalization could arise in a number of ways (31). It could happen if, during untreated HIV infection, tissue HIV populations replicated under reduced immune pressures and/or experienced limited subsequent mixing with blood; if this were the case, then provirus populations archived into blood and tissue HIV reservoirs would reflect their divergent evolution pre-cART. Genetically distinct HIV reservoirs could also arise as a result of restricted gene flow between compartments; for example, if only a small number of HIV strains initially penetrated into tissue, descendant viral populations in tissue could form nested subclades within the more diverse populations in blood. Differential proviral distributions could also arise via unequal clonal expansion and/or contraction of latently HIV-infected cells in blood and tissue either before or during cART. The latter would result in viral populations that are not different in terms of unique viral lineages but, rather, that are different in terms of their frequencies of particular identical proviruses. Investigating blood-tissue genetic compartmentalization will shed light on the extent to which tissue microenvironments shape latent within-host HIV populations and the extent to which the blood reservoir mirrors that elsewhere in the body, information that is relevant to the design of curative approaches guided by individualized analysis of within-host reservoir composition (32, 33).
While studies in nonhuman primates have compared the genetic structures of within-host simian immunodeficiency virus populations in male genital organs versus those of populations in blood (28, 34), no studies, to our knowledge, have investigated this in healthy, HIV-infected humans. We thus investigated the within-host proviral burden, genetic diversity, and compartmentalization in 10 HIV-infected, cART-suppressed individuals undergoing elective bilateral orchiectomy for gender affirmation surgery (21, 23).
RESULTS
HIV characterization in blood and testes.At the time of blood and tissue sampling, all participants had maintained plasma viremia suppression on cART for ≥6 months. At this time, the blood HIV DNA levels in the participants ranged from 11 to 4,003 (median, 512; interquartile range [IQR], 100 to 1,111) HIV DNA copies/106 CD4+ T cells. The presence of CD4+ T cells in the testes was confirmed by immunostaining (Fig. 1A). The testis HIV reservoir size, quantified in terms of the number of viral copies per total cell equivalent (CD4+ T cells were not isolated from tissue), ranged from trace quantities to 198 (median, 3; IQR, 0.4 to 10) copies/106 cells in the right testis and trace quantities to 39.2 (median, 1; IQR, 0.4 to 15) copies/106 cells in the left testis (Fig. 1B). The HIV DNA loads in the right and left testes correlated significantly between participants (Spearman's r = 0.73, P = 0.03; not shown). HIV DNA levels in blood and testes, calculated as the mean of the right and left sections, also correlated marginally between participants (Spearman’s r = 0.54, P = 0.1) (Fig. 1C).
Quantifying and characterizing the HIV reservoir in blood and testes. (A) Representative immunostaining of a frozen testis section (participant 5) showing cells expressing CD3 (red) or CD4 (green). Nuclei were counterstained with DAPI (blue). CD4 T cells (CD3+ CD4+) appear in yellow in the merged image. (B) HIV DNA loads, expressed in number of copies per million cells in whole-tissue lysate, in left and right testes. ND, not determined; trace, HIV DNA was detected but not quantifiable. The asterisk denotes the fact that HIV DNA loads in testes are measured as copies per million total cells (not per million CD4+ T-cells, as for blood). (C) Marginal positive correlation between HIV DNA loads in blood and testes (where the latter is expressed as the average measurements for the left and right testes). (D) Unique HIV sequence distribution in blood and testes, matched by participant. (E) Significant correlation between the HIV DNA load and unique HIV sequence frequency in blood. (F) Lack of a significant correlation between the HIV DNA load and unique HIV sequence frequency in testes.
In total, 722 subgenomic HIV sequences (nef) were isolated by single-genome amplification from blood (n = 396) and at least one tissue section (n = 326) for all 10 participants, even those for whom testis HIV DNA loads were below the limit of quantification (participants 7 and 10). Even though different tissue sections were used for HIV DNA quantification and sequence recovery, testis HIV DNA levels correlated with the total number of HIV sequences retrieved from ∼100 μg tissue (Spearman's r = 0.61; P = 0.07; not shown), suggesting the absence of major location-based sampling biases. After removal of 50 defective and/or hypermutated sequences and 9 putative within-host recombinants, 663 intact nef sequences remained: 354 (26 to 45 per participant; median, 34; IQR, 31 to 40) from blood and 309 (4 to 99 per participant; median, 16; IQR, 10 to 35) from testes (Table 1). For eight participants (participants 1 to 5, 7, 9, and 10), intact nef sequences were isolated from both the right and left testes; however, for participant 6, only the right testis yielded sequences, and for participant 8, only right testis sections were available for analysis. Due to the modest number of testis nef sequences recovered for some participants, our primary analysis combined all within-host testis sequences together regardless of sampling location, though we tested the validity of this assumption in participants for whom sufficient data were available (see below).
Participant HLA and HIV sequence characteristics
All participants harbored both unique HIV nef sequences (observed only once) and identical nef sequences (observed more than once), but at markedly different frequencies (Table 1). This could be visualized by plotting each participant's total number of unique sequences as a function of the total number of sequences collected (Fig. 2). While for some participants, notably, participant 6, unique sequences continued to be recovered during sampling, for many others (e.g., participant 1), ongoing sampling largely yielded sequences that had been recovered previously. Notably, there was no significant relationship between the frequency of unique sequences recovered from blood and testes within a given participant (Wilcoxon signed-rank test, P = 0.5; Fig. 1D). Participants 1 and 2 represented the most extreme examples of this: whereas 70% of HIV nef sequences isolated from participant 1's blood were unique, the sequences isolated from testes were almost all (92%) identical, whereas for participant 2 the opposite was true (Table 1). This indicates that one cannot generalize, based on location alone, where identical HIV sequences are more likely to be observed. In blood, we observed a highly significant positive correlation between HIV DNA levels and unique HIV nef sequence frequency (Spearman's r = 0.90, P = 0.0008; Fig. 1E). This supports the notion that larger blood reservoirs tend to be more genetically diverse (as opposed to more highly clonally expanded). In contrast, no significant correlation between HIV DNA levels and unique nef sequence frequency was observed in testes; in fact, the relationship trended in the opposite direction (Spearman’s r = −0.49, P = 0.2; Fig. 1F). This indicates that, in contrast to blood, the relationship between reservoir size and diversity in testes is less straightforward.
Relationship between the total number of unique HIV sequences collected per participant as a function of the total number of sequences collected. The dashed line denotes a hypothetical data set where every sampled sequence is unique.
A phylogeny inferred from an alignment comprising all intact HIV nef sequences plus select subtype reference strains confirmed that each participant's sequences formed monophyletic clades with high (≥99.7%) bootstrap support and revealed that participants 1 to 9 harbored HIV subtype B, while participant 10 harbored subtype C (Fig. 3). Though all within-host phylogenies featured both unique and identical HIV nef sequences, tree topology and diversity varied markedly between individuals (Fig. 4). Participant 2's phylogeny was remarkable, in that 25 of the 26 nef sequences isolated from blood were identical, but those isolated from testes were diverse (though the other lone unique blood nef sequence was also quite distant from the others) (Fig. 4A). Of note, the abundant blood nef sequence was also recovered from both the right and left testes; in addition, there was a second nef sequence that was independently recovered from both testes. Participant 6 harbored diverse HIV nef sequences within both blood and the right testis (no sequences were recovered from the left testis) and exhibited the highest within-host diversity overall (Fig. 4B). Nevertheless, identical nef sequences, including three examples of identical sequences isolated from both blood and the right testis, were identified. Participant 5 harbored multiple closely related monotypic clades, including one sequence that was abundant in both testes (and that was detected in independent samplings thereof) and that was also present in blood, alongside other diverse sequences in blood and tissue (Fig. 4C). Participants 1 and 9 were notable in that they both featured a large, genetically distinct monotypic clade detected solely in testes (in participant 1, this sequence was detected in both testes, while in participant 9, it was detected in the right testis only), yet both also featured other identical HIV nef sequences that were isolated from both blood and testes (Fig. 4D and E). Participant 8 featured six clusters of identical sequences, four of which were detected solely in blood and two of which were detected in both blood and the right testis (a left testis section was not available for this participant) (Fig. 4F). Participant 3 harbored four clusters of identical sequences: one large one that was found predominantly in blood but that was also recovered from the right testis and three others that were exclusively observed in either blood or tissue (including one sequence observed in both the right and left testes) (Fig. 4G). Of all individuals studied, participants 4, 7, and 10 exhibited the most limited within-host diversity (participant 10's proviruses were the least diverse of all, suggestive of early cART initiation; of note, this individual, the only one for whom some clinical information was available, had been diagnosed with infection only 3 years earlier) (Fig. 4H to J). Nevertheless, like all others studied, participants 4, 7, and 10 also featured at least one identical HIV nef sequence isolated from both blood and testes.
Between-host HIV phylogeny. The numbers on the branches indicate bootstrap values supporting within-host monophyletic clades. The scale is the estimated number of substitutions per nucleotide site. NL4-3 and MJ4 are subtype B and C reference sequences, respectively.
Within-host HIV reservoir diversity in blood and testes. Phylogenies, inferred from nucleotide sequence alignments (HIV nef), are midpoint rooted, with scales denoting the estimated number of substitutions per nucleotide site. Matched highlighter plots, made from amino acid sequence alignments, show substitutions relative to the master sequence (the top sequence in the phylogeny). Symbols denote sampling location: blood (filled circles), right testis (open squares), and left testis (open diamonds). The red A and black B markers for participants 1, 3, 4, 5, 7, 8, and 9 identify sequences collected from independent sections of the same testis (the HIV genetic compartmentalization results for these within-tissue comparisons are summarized in Table 3). Bootstrap values of between 70 and 90% are reported to the left of their respective nodes; those of >90% are marked with asterisks.
We next investigated the extent to which the HIV DNA diversity in the testes correlated with that in blood. Because identical sequences artificially lower average pairwise HIV diversity measures (and our data already indicated that identical sequence distributions varied markedly between compartments), we collapsed identical sequences down to a single copy per compartment and computed median within-host patristic (tip-to-tip phylogenetic) distances between all pairs of sequences sampled from each site. Median within-host patristic distances were 2.25 × 10−2 (IQR, 1.83 × 10−2 to 3.70 × 10−2) substitutions per nucleotide site in blood and 2.33 × 10−2 (IQR, 1.79 × 10−2 to 4.05 × 10−2) substitutions per nucleotide site in testes, values that correlated significantly with one another (Spearman's r = 0.94, P = 0.0002; Fig. 5). These results indicate that the diversity of unique HIV sequences in the testes generally reflects that in blood.
Significant positive correlation between blood and testis proviral diversity. Values represent average within-host patristic (tip-to-tip phylogenetic) distances between all pairs of sequences sampled from each site, after collapsing identical nef sequences to a single copy per compartment.
HIV genetic compartmentalization.Our major objective was to assess whether proviral sequences in the blood and testes exhibit significant population structure or genetic compartmentalization. As the presence of identical sequences influences compartmentalization detection (especially if their distribution varies markedly across sites), we analyzed the data in two ways: by including all sequences (overall) and by collapsing identical sequences down to a single copy per compartment (unique).
When assessing within-host data sets overall, both distance- and tree-based tests agreed that 6 of 10 participants (participants 1 to 5 and 9) displayed significant blood-testis HIV genetic compartmentalization, while participants 7 and 8 did not (Fig. 6A and Table 2). For participants 6 and 10, the distance-based test did not detect compartmentalization, but the tree-based test did; as per our predefined criteria, these participants' overall data sets were deemed not compartmentalized. Notably, all six participants whose overall data sets were deemed compartmentalized harbored at least one abundant identical HIV nef sequence that predominated or that was found exclusively in one compartment (Fig. 4).
Genetic compartmentalization results for each participant. “Distance” denotes the results of the Hudson, Boos, and Kaplan nonparametric test for population structure, “Tree” denotes the results of the Slatkin-Maddison test (57), and “Consensus” denotes the final result (compartmentalization was declared only if the results of both tests agreed). Blue denotes compartmentalization; gray denotes no compartmentalization. (A) Overall results when comparing all nef sequences from blood to all those from testes; (B) unique results with identical nef sequences collapsed to a single copy per compartment.
Blood-testis and right-left testis HIV genetic compartmentalization results
Importantly, however, when the analysis was restricted to unique nef sequences per site, no participant retained consistent evidence of compartmentalization between blood and testes (Fig. 6B and Table 2). These observations are notable for two reasons. First, the observation that 40% of the within-host data sets lacked any evidence of significant HIV population structure between blood and testes overall indicates that the HIV proviral composition in the testes is not always distinct from that in blood. Second, the observation that no participants exhibited evidence of blood-testis HIV compartmentalization when the analysis was restricted to unique nef sequences per compartment strongly suggests that HIV population structure, when present, is driven by the differential distribution of identical sequences across blood and tissue and not by the presence of distinct HIV lineages within these sites.
Our primary analyses combined all sequences from a given site together. However, given that only a limited amount of biological material was analyzed, sampling biases are a potential concern. For participants 5, 6, and 7, independent peripheral blood mononuclear cell (PBMC) aliquots were received and analyzed at different times, allowing us to investigate whether independent sampling of HIV nef sequences from blood yielded consistent distributions. It did in all cases: proviruses isolated from independent PBMC aliquots exhibited no evidence of compartmentalization by any method (Table 3), supporting their sampling from the same overall gene pool. Moreover, for seven participants (participants 1, 3 to 5, and 7 to 9), sufficient HIV sequences were isolated from different sections of the same testis (see the tip label annotations in Fig. 4), yielding eight opportunities to investigate whether independent samplings of the same site yielded consistent HIV sequence distributions (for participant 5, independent samplings of both the right and left testes yielded sufficient sequences for evaluation). Of these eight within-testis data sets, consistent evidence of overall compartmentalization was detected in only one: the right testis of participant 7 (Table 3). Note that consistent support for within-tissue compartmentalization in participant 7's right testis did not remain after the analysis was restricted to unique sequences (not shown), though limited statistical power is acknowledged. Phylogenetic segregation of the HIV sequences retrieved from these two independent samplings of the right testis is indeed apparent (Fig. 4J); however, further sampling of additional tissue sections yielded HIV sequences that intermixed with these, and it is important to underscore that there was no support for compartmentalization between blood and testes within this participant (Fig. 6). Together, these observations suggest that HIV sampling from blood and testes was not majorly biased by the specific aliquot or tissue section analyzed.
Genetic compartmentalization when sampling HIV sequences from the same site or tissue
For seven participants (participants 1 to 5, 7, and 9), we had sufficient HIV nef sequences to investigate compartmentalization between the right and left testes. When analyzing overall within-host data sets, only participant 9 consistently exhibited between-testis compartmentalization (Table 2). However, after restricting the analysis to only unique nef sequences per site, no consistent evidence of compartmentalization remained. Consistent with the results of our analyses of HIV sequences in blood and tissue, these results suggest that in the rare cases where population structure was observed between proviral populations in the right and left testes, this was driven by the differential distribution of identical sequences between sites.
HIV immune escape in blood and testes.Given that cellular immune responses are suppressed in the testes (21, 23), we wished to investigate whether blood and testis HIV sequences differed with respect to their immune escape mutation burden. We investigated this in two ways. First, the total escape burden was estimated by identifying all HIV codons associated with one or more host HLA alleles and classifying the HIV residue at that site as adapted (inferred escaped) or susceptible based on published lists of HLA-associated polymorphisms defined at the population level in HIV subtype B (35). As such, the analysis was restricted to the 9 subtype B-infected participants. For each sequence, we calculated the percentage of HLA-associated sites exhibiting an adapted (or possibly adapted) form and computed the mean for each within-host data set (e.g., participant 7’s data set is 43% adapted to the host HLA; Fig. 7A). Next, we estimated within-host escape complexity by quantifying the percentage of published or predicted optimally described HLA-restricted cytotoxic T lymphocyte (CTL) epitopes exhibiting within-host amino acid variation (e.g., 5/5 [100%] for participant 7). Both metrics correlated positively in blood and testes (Spearman's r = 0.98 and P = 0.0001 for escape burden [Fig. 7B] and Spearman's r = 0.93 and P = 0.001 for escape complexity [Fig. 7C]). In fact, these values were highly concordant (Lin’s concordance coefficient for the data presented in Fig. 7B was 0.97 [95% confidence interval {CI}, 0.90 to 0.99], while that for the data presented in Fig. 7C was 0.95 [95% CI, 0.80 to 0.99]). Examples of the generally concordant nature of within-host HIV variation in key HLA-restricted epitopes in blood and testes are shown in Fig. 7D. Consistent with previous reports (15), these examples also reveal that HLA-susceptible and -adapted forms of the same CTL epitope commonly coexist in the reservoir: 6 of the 9 subtype B-infected individuals harbored at least one epitope where this occurred.
Immune escape landscape in blood and testes. (A) Example of escape burden analysis using participant 7's HIV Nef amino acid sequence alignments in blood and testes. Red, orange, and blue represent HLA-adapted, possibly adapted, and susceptible forms, respectively. For each sequence, the sum of HLA-adapted and possibly adapted sequences is divided by the total number of HLA-associated sites to yield the overall percentage of adapted sites (values shown at the end of each sequence). These values are then averaged to yield site- and participant-specific averages. For viral sites associated with more than one HLA allele (e.g., Nef codon 105), the inferred escape profile is shown for the underlined allele. (B) Correlation between escape burden in blood and testes. (C) Correlation between epitope complexity, defined here as the percentage of known or predicted HLA-restricted CTL epitopes exhibiting amino acid variation, between blood and testes. (D) Examples of escaped and susceptible forms coexisting within known or predicted CTL epitopes within the participants' reservoirs, where the height of the residue represents its frequency.
DISCUSSION
The availability of matched blood and tissue from our unique cohort (21, 23) allowed the first genetic exploration of the hypothesis that the testes constitute a distinctive HIV sanctuary site in healthy HIV-infected humans on suppressive cART (21, 25). Our study confirmed that CD4+ T cells are present in testicular tissue and that HIV DNA could be detected in the testes of all participants, though in two cases these were below quantification limits. Overall, the testis proviral burdens measured in the present study were lower than those previously reported in an autopsy study of cART-treated individuals; however, in the latter study, some individuals had discontinued cART near the end of life, which could have increased proviral levels in tissue (30). The proviral burden in blood correlated with that in the testes (Fig. 1C), suggesting that individuals with larger blood reservoirs will tend to have larger tissue reservoirs (36). These observations confirm the testes as a site of HIV persistence during cART.
Our results also yielded insights into within-host proviral diversity. First, identical HIV nef sequences, which are suggestive of clonal expansion, were observed at markedly different frequencies within and between participants. While in some (e.g., participants 1 and 5) a particular nef variant dominated within the testes, in others (e.g., participant 2) a particular variant dominated in the blood, while in yet others (e.g., participant 6) identical nef sequences were relatively rare. These observations are consistent with the clonal expansion of latently infected cells as a driver of HIV persistence during cART in many persons (10, 16–18, 37, 38). However, the frequencies of identical nef sequences followed no consistent pattern between blood and testes, suggesting that one cannot generalize, based on location alone, where clonally expanded latently infected cell populations are more likely to reside. Disregarding clonal expansion, however, within-host blood and testis HIV diversity correlated significantly (Fig. 5), suggesting that individuals with diverse blood reservoirs also harbor diverse HIV sequences in the testes.
Our study also sheds some light on two outstanding questions: first, whether the within-host testicular microenvironment may be distinct enough to influence the proviral composition therein, and second, in the cases where blood-testis HIV genetic compartmentalization is detected, how this arises (31). Our observations reveal that, while blood-testis genetic compartmentalization was reasonably common if one analyzed within-host HIV nef sequences as a whole (two established tests consistently deemed 60% of overall within-host data sets to be compartmentalized), no participant retained consistent evidence of blood-testis compartmentalization when only unique HIV nef sequences per compartment were analyzed. Strictly speaking, therefore, the proviral populations in the testes of the majority of cART-suppressed individuals do differ from those in blood, but only because of their unbalanced distributions of identical sequences. This in turn suggests that the principal mechanism in which blood and testis HIV reservoirs differ from one another is not via seeding of divergent viral sequences therein pre-cART, nor is it due to restricted trafficking of viral lineages between blood and testes. Rather, our observations suggest that the extent and/or dynamics of proliferation of latently HIV-infected cells (and/or subsequent waning of descendant populations [10]) can differ substantially between blood and testes and that this in turn drives the differences in the HIV population structure between these sites. While it is tempting to speculate that the site harboring the sequence in abundance is where the clonal expansion occurred and the other is where clonal descendants subsequently migrated, we cannot rule out the possibility that cells harboring identical HIV proviruses preexisted in both sites as a result of a previous expansion and migration event and that these underwent another expansion, in parallel, more recently. Our observation that the HIV population structure, when present, is due to the differential distribution of identical sequences across anatomical sites is consistent with studies of viral diversity within the female genital tract (39, 40). More broadly, our observations are consistent with the notion that major blood-tissue compartmentalization during cART, at least in terms of unique HIV lineages, may not be the norm (41–43).
At first glance, our finding that the composition of unique proviral lineages in the testes does not differ significantly from that in the blood may appear to contradict the former's status as an HIV sanctuary site (21, 44). The recovery of proviruses from the testicular tissue of all participants, however, confirms the testes as a site of HIV persistence during cART. Rather, our results suggest that the testis microenvironment may not be sufficiently unique to persistently maintain distinct HIV lineages therein, at least not over the time course of cART suppression of the participants studied. Similarly, while reduced anti-HIV immune responses likely facilitate HIV persistence in the testes in general, our observation that the HIV immune escape burden correlated strongly between blood and testes (Fig. 5) suggests that the extent of local immunosuppression may not be sufficient to markedly influence the immune escape landscape therein. However, there is an alternative explanation for the lack of genetic compartmentalization of unique HIV lineages between blood and testes: we cannot exclude the possibility that low-level HIV replication may still occur in certain tissues, including the testes, during cART, and that this could lead to the ongoing reshaping of the HIV populations persisting in blood.
Other limitations of our study and data interpretation should also be acknowledged. Though proviral sequences were obtained via single-template amplification, only a subgenomic HIV region (nef) was analyzed due to the challenges of recovering HIV from the testes (including from two participants for whom the testis proviral loads were below quantification limits). As such, we must acknowledge the possibility that sequences that are intact within nef may harbor defects elsewhere in the genome, nor can we definitively classify identical nef sequences as being derived from clonally expanded cell populations. As HIV sequences were isolated from PBMCs or whole tissue, the specific immune cells harboring them are unknown. This may be particularly important for the testes, which are rich in macrophages, which may contribute to HIV persistence there (30). Limited sampling must be acknowledged, particularly from the testes, which could compromise our power to detect compartmentalization. However, our independent recovery of the same sequence from separate testis sections, our lack of observation of compartmentalization within and between testes in 21/22 comparisons performed (Tables 2 and 3), our recovery of at least one identical HIV sequence from both blood and testes in each participant, and the observation that, for many participants, increased proviral sampling largely yielded sequences that had been recovered previously suggest that our results are not simply attributable to biased and/or limited sampling. Furthermore, our qualitative observation that blood and testis sequences were reasonably well intermixed in all phylogenies (that is, that no data set exhibited overt segregation by site) further supports the notion that these two sites do not differ substantively in terms of a unique HIV lineage distribution. We must also acknowledge that limited within-host diversity can contribute to uncertainty in phylogenetic reconstruction (as evidenced by low bootstrap values for some nodes in the within-host trees), which could in turn influence the results of tree-based compartmentalization tests, but the relatively high concordance between the tree- and distance-based tests (Fig. 6; Tables 2 and 3) suggests that this is not a major concern. As nef was sequenced, the possibility of a differential distribution of antiretroviral resistance mutations across sites (as a result of the reduced penetration of some antiretrovirals into the testes [44]) could not be investigated. As clinical histories were unknown, we could not investigate correlates of blood-tissue HIV diversity and compartmentalization, nor can we rule out the possibility that evidence of genetic compartmentalization between blood and tissues may become apparent only after much more prolonged durations of cART suppression (e.g., as a result of latently infected cells having different half-lives in blood than in the testes as a result of reduced immune surveillance in the latter). Finally, proviral sequences could not be studied in the context of pre-cART viral populations, as these samples were not available.
In conclusion, and again noting the caveats associated with subgenomic HIV sequencing, our observation that a unique proviral sequence distribution and the immune escape mutation burden in testicular tissue are not unrepresentative of those within blood suggests that diversity and escape-related barriers to HIV eradication are likely to be of a broadly similar magnitude across these sites. However, the marked differences in the quantity and distribution of identical HIV sequences between hosts and between anatomical sites within each host underscore the potential challenges of differentially distributed clonally expanded latently infected cell populations to HIV eradication.
MATERIALS AND METHODS
Study participants and ethics statement.Blood and testicular tissues were donated by 10 HIV-infected but otherwise healthy adults aged 27 to 52 years who underwent elective bilateral orchiectomy for gender affirmation (the ORCHID cohort [21, 23]). At the time of surgery, all participants had maintained viremia suppression on cART for at least 6 months; however, full clinical histories (including infection date estimates) were unavailable. This study was approved by the Institutional Review Boards of the Research Institute of the McGill University Health Centre and Simon Fraser University. All participants provided written informed consent.
Blood and tissue collection.Peripheral blood mononuclear cells (PBMCs) were isolated from whole blood by Ficoll density gradient centrifugation and stored at −80°C until use. Testicular tissue was processed within 1 h of surgery (21, 23). For genetic analysis, ∼100-μg tissue sections were snap-frozen upon reception in liquid nitrogen; for histology, 10-mm3 sections were embedded in CryoMatrix material (Thermo Fisher) before snap-freezing.
Immunostaining.Frozen embedded tissues were cut into 5-μm-thick sections (Leica CM3050S cryostat), fixed in acetone-methanol for 5 min at 4°C, stained overnight at 4°C using anti-CD3 and anti-CD4 (Dako-Agilent), and exposed to conjugated Alexa Fluor secondary antibodies (Thermo Fisher) for 30 min at room temperature. The slides were counterstained using 4',6-diamidino-2-phenylindole (DAPI) before mounting in Fluoromount-G mounting medium (Southern Biotech). Images were acquired using a Nikon epifluorescence microscope with a Zyla sCMOS camera and analyzed using ImageJ (v1.47g) software (National Institutes of Health).
HIV quantification and sequencing.DNA was extracted from CD4+ T cells negatively isolated from PBMCs (StemCell Technologies) and directly from snap-frozen tissue (Qiagen). The HIV DNA burden was measured using a quantitative PCR assay capable of detecting a single viral genome per PCR, where primers are located in the HIV long terminal repeat-gag region and are optimized for detection of multiple HIV subtypes (21, 45). For each participant, a minimum of 5 million PBMCs plus at least four tissue sections, two each from the right and left testes, were allocated for HIV sequencing, with the exception of participant 8, for whom only two right testis sections were available. Genomic DNA was extracted from PBMCs directly (Invitrogen), whereas tissues were further sectioned into ∼50-μg fragments prior to genomic DNA extraction. HIV amplification was performed using single-genome amplification of a subgenomic fragment (HIV nef). The nef gene was selected based on its relatively high within-host diversity, its richness in phylogenetic signal, and its representativeness of within-host HIV evolution and diversity within the rest of the HIV genome (8). This gene also represents the one most likely to be intact within proviruses sampled during long-term cART (46). Single-genome amplification, performed using high-fidelity enzymes (Roche Expand HiFi) and oligonucleotide primers designed to amplify HIV sequences from all major subtypes, was achieved by endpoint dilution, such that ∼25 to 30% of the resulting nested PCRs were successful (8). Negative PCR controls, included in every amplification, always remained negative. Amplicons were sequenced on an automated DNA sequencer using BigDye (v3.1) chemistry (Applied Biosystems). Chromatograms were base called using the Sequencher (v5.0) program (GeneCodes). Sequences exhibiting nucleotide mixtures, defects, or hypermutations (identified using the Hypermut program [47]) were excluded, as were sequences exhibiting evidence of within-host recombination (identified using the RDP4 program [48]).
Sequences were codon aligned using the MAFFT method (49) and edited in the AliView (v1.18) alignment viewer and editor (50). For each within-host alignment, the best-fitting substitution models for both genetic distance and phylogenetic-based tests were determined using the jModelTest (v2.1.10) program (51). Phylogenies were inferred by approximate maximum likelihood using the FastTree2 (v2.1.10) program (52) under the best available nucleotide substitution model (for participants 7 and 10, the generalized time-reversible [GTR] model [53] with no rate heterogeneity among sites was the best fit; for all others, a GTR model with rate heterogeneity using a gamma distribution was the best fit). Node support values were derived from 1,000 bootstraps. Within-host phylogenies were midpoint rooted in FigTree (v1.4.4) software (http://tree.bio.ed.ac.uk/software/figtree/) and visualized with the ggtree program (54).
Genetic compartmentalization tests and other statistical analyses.As there are different ways to define genetic compartmentalization (e.g., as genetic heterogeneity, or the presence of distinct lineages, between two viral subpopulations), compartmentalization tests occasionally yield discordant results (55). As recommended, we applied more than one test and classified a data set as “compartmentalized” only when the results of both tests agreed (55). We employed one genetic distance-based test (56) and one tree-based test (57), both of which were implemented in the HyPhy (v2.22) program (58). For the distance-based test, we employed the Hudson, Boos, and Kaplan nonparametric test for population structure (56), a test that has been validated for within-host HIV data sets (59), including from the reservoir, where it is sometimes referred to as the “nonparametric test for panmixia” (59–61). This test compares the mean pairwise distances between sequences from different subpopulations versus the same subpopulation (compartmentalization is supported if the mean pairwise distances of sequences from the same subpopulation are smaller than those from different subpopulations) and computes the KST statistic, whose values range from 0 (denoting no compartmentalization) to 1 (denoting complete compartmentalization). Statistical significance, expressed as a P value, is assessed via a population structure randomization test. Using the jModelTest (v2.1.10) program (51), we identified the Tamura-Nei 1993 (TN93) nucleotide substitution model to be the best-fitting distance-based model available in the HyPhy (v2.22) program (58); use of this model is also consistent with previous applications of this test to within-host HIV data sets (55, 62). For the tree-based test, we used the Slatkin-Maddison test (57). This test determines the minimum number of between-compartment migrations to explain the distribution of compartments on the tree tips: the smaller the number, the stronger the support for compartmentalization. Statistical significance is assessed via a lineage permutation test. Importantly, because identical HIV sequences (especially those that are present in abundance and whose frequencies differ between sites) increase the likelihood of compartmentalization detection (39), we analyzed each data set in two ways: overall and with identical sequences collapsed to a single copy per compartment.
HLA class I sequence-based typing was performed to allele-level resolution (63). HLA-associated adapted (inferred HLA-escaped) and nonadapted (inferred HLA-susceptible) polymorphisms in HIV subtype B were defined using a published list derived from an analysis of linked HIV/HLA genotypes from an independent cohort of 1,888 subtype B-infected individuals (35). HLA-restricted, optimally described CTL epitopes in the participants' provirus sequences were defined using the Los Alamos HIV Molecular Immunology Database with current updates (64), where epitopes restricted by HLA alleles closely related to the participants' allele(s) were identified as putative epitopes (35). Correlations were assessed using Spearman's correlation. P values of <0.05 were considered statistically significant.
ACKNOWLEDGMENTS
The research reported in this publication was funded in part by Canadian HIV Cure Enterprise Team grants HIG-133050 (to J-P.R., Z.L.B., and A.F.Y.P.) and HB2-164064 (to J-P.R.) from the Canadian Institutes for Health Research (CIHR) in partnership with CANFAR and the International AIDS Society (IAS), by CIHR team grant HB1-164063 (to Z.L.B.), by CIHR project grants PJT-159625 (to J.B.J. and Z.L.B.) and PJT-155990 (to A.F.Y.P.), by the National Institutes of Health under award number NIHR21A127029 (to J.B.J., A.F.Y.P., Z.L.B.), by the National Institute of Allergy and Infectious Diseases of the National Institutes of Health with cofunding support from the National Institute on Drug Abuse, the National Institute of Mental Health, and the National Institute of Neurological Disorders and Stroke under award number UM1AI126617 (to Z.L.B.), and by a Simon Fraser University Next Big Question fund award (to Z.L.B.). This study was also supported by the Fonds de recherche du Québec-Santé (FRQ-S): Réseau SIDA/Maladies infectieuses and Thérapie cellulaire (to J-P.R.). N.N.K. was supported by a CIHR Frederick Banting and Charles Best M.Sc. award. F.H.O. was supported by the Canadian Queen Elizabeth II Diamond Jubilee Scholarships program, a joint initiative of the Rideau Hall Foundation, Community Foundations of Canada, and the Association of Universities and Colleges of Canada, and also received a fellowship from the Sub-Saharan African Network for TB/HIV Research Excellence (SANTHE), a DELTAS Africa Initiative (grant number DEL-15-006). The DELTAS Africa Initiative is an independent funding scheme of the African Academy of Sciences (AAS) Alliance for Accelerating Excellence in Science in Africa (AESA) and is supported by the New Partnership for Africa’s Development Planning and Coordinating Agency (NEPAD Agency) with funding from the Wellcome Trust (grant number 107752/Z/15/Z) and the UK government. M.-A.J. holds the CIHR Canada Research Chair tier 2 in immunovirology. A.F.Y.P. is supported by a CIHR new investigator award (FRN-130609). N.C. is supported by a research scholar career award of the Quebec Health Research Fund (FRQ-S; number 253292). Z.L.B. is supported by a Michael Smith Foundation for Health Research (MSFHR) scholar award. J.-P.R. is the holder of the Louis Lowenstein Chair in Hematology and Oncology at McGill University.
The content is solely the responsibility of the authors and does not necessarily represent the official views of the National Institutes of Health, AAS, the NEPAD Agency, the Wellcome Trust, the UK government, or the other funding agencies listed.
We thank Gabriel Stang and Jeff Knaggs for software programming and bioinformatics assistance and Mark Brockman for helpful discussions. We are grateful to the participants, who made this study possible by their donations of blood and tissue.
FOOTNOTES
- Received 6 May 2019.
- Accepted 8 June 2019.
- Accepted manuscript posted online 12 June 2019.
- Copyright © 2019 American Society for Microbiology.
