|Home | About | Journals | Submit | Contact Us | Français|
A distinctive feature of chronic human immunodeficiency virus type 1 (HIV-1) infection is the presence of multiple coexisting genetic variants, or subpopulations, that comprise the HIV-1 population detected in the peripheral blood. Analysis of HIV-1 RNA decay dynamics during the initiation of highly active antiretroviral therapy (HAART) has been a valuable tool for modeling the life span of infected cells that produce the bulk HIV-1 population. However, different HIV-1 target cells may have different turnover rates, and it is not clear whether the bulk HIV-1 RNA decay rate actually represents a composite of the decay rates of viral subpopulations compartmentalized in different cellular subsets with different life spans. Using heteroduplex tracking assays targeting the highly variable V3 or V4-V5 regions of the HIV-1 env gene in eight subjects, we found that all detectable coexisting HIV-1 variants in the peripheral blood generally decayed at similar rates during the initiation of HAART, suggesting that all of the variants were produced by cells with similar life spans. Furthermore, single genome amplification and coreceptor phenotyping revealed that in two subjects coexisting HIV-1 variants with distinct CXCR4 or CCR5 coreceptor phenotypes decayed with similar rates. Also, in nine additional subjects, recombination and a lack of genetic compartmentalization between X4 and R5 variants were observed, suggesting an overlap in host cell range. Our results suggest that the HIV-1 env subpopulations detectable in the peripheral blood are produced by cells with similar life spans and are not genetically isolated within particular cell types.
Infection with human immunodeficiency virus type 1 (HIV-1) is typically established by one or a few variants that give rise to an initially homogeneous viral population (12, 19, 38, 45). As infection progresses into the chronic phase, sequence diversification occurs throughout the viral genome, most dramatically in the envelope gene (env), as a result of selection of humoral and cytotoxic T-lymphocyte immune escape mutations (6, 20, 44, 74, 97). Sequence diversity in the env gene is clustered in variable regions, termed V1 through V5, that encode surface loops in the Env protein that are important targets of the host antibody response (20). Diversification of env results in the presence of multiple, coexisting env variants in the peripheral blood that continually evolve during the course of infection (31, 41).
Typically, the HIV-1 population early after infection uses CCR5 as the coreceptor (R5) (38, 93, 99, 101), and variants that can use CXCR4 (X4) arise later in the disease course in approximately one-half of individuals infected with subtype B virus (10, 42). A major determinant of coreceptor use is found in the V3 loop, and sequence evolution in this region is often linked to the virus's ability to use CXCR4 (17, 36, 54). After the emergence of X4 virus, env genes encoding both CCR5 and CXCR4 tropism can coexist for extended periods of time, maintaining a diverse V3 population (10, 54, 85). Variation in R5 Env proteins can also influence the ability of a virus to utilize various levels of CD4 and CCR5 found on different cell types, such as macrophages and T cells (23, 25, 64, 67, 89, 95).
Distinct biological characteristics encoded by different env variants, such as coreceptor use, cellular tropism, and sensitivity to immune pressure, may drive, or result from, HIV-1 compartmentalization. Genetic compartmentalization of HIV-1 variants has been well documented in anatomical compartments, such as the spleen (9, 18, 24), central nervous system (7, 29, 30, 43, 56, 73), and genital tract (13, 39, 60, 65, 70), and in different cellular compartments, such as monocyte/macrophage and CD4+ T cells (21, 47). Divergent coreceptor tropism in particular may lead to compartmentalization of virus in different cellular subsets. For example, both naïve and memory T lymphocytes express CXCR4 but differentially express CCR5 and have been shown to harbor unequal proportions of CXCR4- and CCR5-tropic viral variants that are concordant with their distinct coreceptor expression patterns (3, 91).
While genetic compartmentalization of HIV-1 between anatomically or cellularly segregated sequences has been extensively explored, evidence is limited as to the extent to which the coexisting env subpopulations circulating in the peripheral blood represent virus emanating from compartmentalized populations replicating independently of each other and in different cellular subsets. Although this question is challenging to address in the context of an infected individual, one approach stems from the examination of the rate of decay of HIV-1 RNA during the initiation of highly active antiretroviral therapy (HAART). Effective HAART prevents new rounds of HIV-1 infection, but cells already infected by HIV-1 continue to produce virus, and therefore the rate of viral RNA decay during effective HAART reflects the life span of the HIV-1-producing cells. This strategy was initially used by Ho et al. and Wei et al. (34, 96) to characterize the turnover rate of HIV-1-infected cells. These and subsequent studies (4, 49, 62, 63) suggested that the average life span of infected cells that produce 99% or more of the bulk HIV-1 population in the peripheral blood is short, with a half-life of 1 to 2 days, presumably reflecting the life span of activated T cells, whereas cells producing 1% or less of the population have a half-life measured in weeks, illustrated by a biphasic decay curve. However, the bulk HIV-1 RNA decay rate of the first phase may oversimplify the dynamics of the underlying viral genetic subpopulations because this rate may represent an average without accounting for possible moderate differences between decay rates of coexisting variants, as might be the case for variants infecting cellular subsets in different stages of maturation or activation (22, 59, 100). Based on this model, two or more coexisting viral populations with the same rates of decay during HAART could be presumed to be produced by the same cellular subset or, at minimum, two or more subsets with the same turnover rate. Conversely, the observation of different rates of decay would suggest that different cellular subsets are contributing toward a complex mixture of compartmentalized viral subpopulations.
In this study we characterized the decay rates of coexisting HIV-1 env variants in the peripheral blood of eight subjects initiating HAART. In two of these subjects, who possessed both CCR5- and CXCR4-tropic Env variants, we analyzed the decay rates of each phenotypically divergent env subpopulation. Furthermore, in an additional nine subjects, we characterized the degree of recombination between the R5 and X4 subpopulations and the degree to which each subpopulation was genetically compartmentalized. We found that all detectible HIV-1 env variants in the peripheral blood, regardless of coreceptor tropism phenotype, decayed with similar rates during the initiation of HAART. In addition, we provide evidence of recombination and a lack of compartmentalization between coexisting X4 and R5 subpopulations, consistent with the overlap in their target host-cell populations. Our findings suggest that the detectable HIV-1 env variants that coexist in the peripheral blood are produced by cells with similar life spans and, by this measure, are not compartmentalized in particular cell types.
Blood plasma samples used for the variant decay analysis were obtained from subjects initiating HAART who were recruited either through a previously described study carried out at the University of California, San Francisco (subjects 4015, 4021, 4022, 4030, and 5005) (87) or at the University of North Carolina, Chapel Hill (subjects 101, 106, 108, and 109), specifically for this study. Sampling of blood plasma was carried out on the day of treatment initiation and at 1- to 7-day intervals posttreatment initiation; the median time between samplings for all subjects was 3 days.
Additional blood plasma samples used for the analysis of recombination and compartmentalization in populations with mixed coreceptor use were excess tissue obtained from the baseline blood draw of subjects participating in the virology substudy of a ritonavir efficacy trial described previously (8) (subjects 1314, 1077, and 1551) or from baseline blood draws of subjects entering AIDS Clinical Trial Group 359 described elsewhere (subjects 432, 135, 139, 411, 413, and 310) (27). In all cases, written, informed consent was obtained and all protocols were subject to approval by the Institutional Review Board.
Viral RNA was extracted from virus particles pelleted from 1 ml of blood plasma (25,000 × g for 1.5 h) using a QiaAMP Viral RNA Kit (Qiagen, Chatsworth, CA). Two 5-μl aliquots of the 60-μl eluate were amplified in parallel using a OneStep RT-PCR kit (Qiagen, Chatsworth, CA). The PCR thermocycling procedures, the primers used to amplify the V3 and V4-V5 regions, and the heteroduplex tracking assays (HTAs) have been previously described (14, 31, 54). Briefly, PCR products were annealed to a radiolabeled probe to generate heteroduplexes, which were then separated in a polyacrylamide gel. Gel-separated heteroduplexes were visualized by autoradiography and quantitated on a Storm 840 phosphorimager using the ImageQuant software (Molecular Dynamics/GE, Pittsburgh, PA). The HTA was modified for this study by using a biotin-radiolabeled probe to facilitate isolation and sequencing of HTA bands, as described elsewhere (79). All samples were analyzed in duplicate to verify sampling reproducibility. Samples and time points that could not be reproducibly sampled were excluded from the analysis. Viral RNA loads (VLs) of individual variants were calculated as a product of the total viral RNA load and the fractional abundance of each variant. Variant half-lives were calculated using the time points between which VL was declining.
Viral RNA was extracted as described above. RNA was reverse transcribed using a Superscript III Reverse Transcriptase System (Invitrogen, CA) and oligo(dT). A region of the HIV-1 genome encompassing env through 3′ U3 was amplified using a limiting dilution, seminested PCR according to a method initially described by Simmonds et al. (82) and Edmonson and Mullins (16) and then modified by Palmer et al. (61) and Salazar-Gonzalez et al. (76). Primers and thermocycling procedures for amplification and sequencing were used as previously described in Keele et al. (38) but modified by replacing the downstream env amplification primer set with a primer that captures the U3, the sequence of which is 5′-AAGCACTCAAGGCAAGCTTTATTG-3′.
Coreceptor phenotype was predicted based on V3 sequences of extracted HTA bands and SGA amplicons using a position-specific scoring matrix (PSSM) generated using a training set of V3 sequences from envelopes with known coreceptor phenotypes on indicator cells expressing CD4 and either CXCR4 or CCR5 (36). PSSM was implemented through the web portal http://ubik.microbiol.washington.edu/computing/pssm/.
Phenotypic analysis was carried out as previously described by Kirchherr et al. (40) but with modifications. Briefly, a cytomegalovirus (CMV) promoter containing a 3′ tag matching the 5′ SGA primer binding site was linked to env SGA amplicons using overlapping PCR. For the CMV-env linking PCR, the 5′ primer sequence is specific for the start of the CMV promoter (5′-AGTAATCAATTACGGGGTCATTAGTTCAT-3′), and the downstream primer (5′-TGGGTGGCTCTGAAAAGAGCCTTTGGGCTGCTGGCTCAGCTCGTCTCATTCTTT C-3′) is specific for a sequence just 3′ of the end of env and contains a histone (H1e) mRNA 3′ stem-loop tag (underlined) for increased transcript stability. CMV-env amplicons were cotransfected with the pNL4-3.Luc.R−E− plasmid, obtained from the NIH AIDS Research and Reference Reagent Program (32), to generate a pseudotyped, single-cycle luciferase reporter virus. The coreceptor phenotype of pseudotyped virus was assessed on U87.CD4 indicator cell lines, expressing either CXCR4 or CCR5, obtained from the NIH AIDS Research and Reference Reagent Program (2).
All sequence alignments were generated using the program MAFFT (multiple alignment using fast Fourier transform) (37). Maximum-likelihood phylogenies were generated in PhyML using the HKY85 substitution rate model with the following parameters: use of four substitution rate categories and estimations of the transition/transversion rate ratio, proportion of invariant sites, and the gamma distribution parameter (26). A version of the Slatkin-Maddison (SM) test for gene flow was implemented using HyPhy (69, 83), and measures of KST* were obtained using the program DnaSP (75). KST* is calculated as described by Hudson, Boos, and Kaplan (35). Briefly, KST* = 1 − (KS*/KT), where KS* is the weighted average of the log of the pairwise differences within each of the two potentially compartmentalized subpopulations, and KT is the average number of pairwise differences between sequences, irrespective of their grouping. For both SM and KST* tests, comparison of the observed result to the distribution of 1,000 random permutations of the data was used to obtain a P value. Recombination between X4 and R5 variants was detected in sequence alignments using the bootscan/RECSCAN analysis implemented in the Recombination Detection Program 3 (50, 78). Bootscanning was carried out using a 200-bp window with a 20-bp step, and trees were constructed using the Jukes-Cantor model with 1,000 bootstrap replicates.
Sequences determined in this study have been deposited in the GenBank database under accession numbers FJ798320 to FJ798580.
We examined the decay rates of V4-V5 or V3 env variants in chronically infected subjects initiating HAART to characterize the life span of infected cells that produce coexisting genetic variants in the peripheral blood. Blood plasma samples from eight subjects were drawn every 1 to 7 days after the initiation of therapy, with a median interval of 3 days, for up to 3 weeks or over a 1- to 2-log10 drop in total VL. All subjects in this study achieved suppression of VL to below detectable levels. RNA was extracted from the plasma samples and then amplified by reverse transcription-PCR to generate amplicons of the variable regions V4-V5 or V3, which were then subjected to HTAs to resolve the coexisting sequence variants. The relative abundance of V4-V5 or V3 variants, potentially comprising as little as 1 to 3% of the total population, was measured by phosphorimaging analysis of HTA bands (71, 79). Duplicate reverse transcription-PCRs were analyzed to ensure sampling reproducibility; the ability to analyze duplicate samples to validate the quality of sampling is a key feature of the HTA strategy as the HTA pattern of two identical, complex populations will appear different if they are not adequately sampled (33); however, the HTA pattern represents only variants whose sequence differences cause a shift in migration rate. While much of the diversity in these highly variable regions of the genome is captured using this technique, in this work we are testing the hypothesis that these HTA variants are markers of potentially compartmentalized populations.
The decay curves of V4-V5 and V3 HTA variants from two representative subjects are depicted Fig. Fig.1.1. The log10 magnitude of the VL drop, the half-life (days) of the bulk VL, and the maximum differences in half-life for the env variants detected in each subject are shown in Table Table1.1. In all but one subject, virus decay appeared monophasic over the first 1- to 2-log10 decline in VL. In subject 106, the VL decay appeared more rapid between the first and second time points (days 0 and 1), with an apparent half-life of 0.28 days (based on a single time point), than the VL decline between the second and fifth time points, with an apparent half-life of 2.22 days (Fig. (Fig.1a).1a). The median half-life of the bulk VL decay rate for these subjects was 1.7 days. The decay rates of either V3 or V4-V5 variants within each subject did not vary significantly; the median value for the greatest differences between any two variants within a subject was 1.6-fold and did not exceed 2-fold within any subject, likely within the margin of error of this assay. The minor variation in decay rates of variants did not correlate with their relative abundance (Table (Table1).1). Thus, for variants comprising as little as 2% of the detected population and representing the first phase of decay, we conclude the following: (i) variants are not compartmentalized in anatomical locations or cellular compartments that are differentially targeted by antiviral activity; (ii) the life spans of potentially different, virus-producing cellular subsets do not differ to a significant degree for those cells producing the major variants of env; and (iii) variants are not otherwise compartmentalized in a manner that differentially affects their rates of decay.
The ability of a virus to use CXCR4 efficiently may allow infection of a different subset of target cells and hence may result in compartmentalization of X4 and R5 variants in cell types that differentially express these coreceptors, such as memory and naïve T-cell subsets (5). Many X4 variants retain the ability to enter cells using CCR5 (i.e., dual-tropic) although it is not clear this ability is utilized in vivo (48, 98). Furthermore, naïve and memory T-cell subsets have been shown to be preferentially infected by CXCR4 or CCR5 variants, respectively (3, 91).
We identified two subjects from the analysis described above who had coexisting CXCR4- and CCR5-tropic populations, which allowed us to determine the relative rates of decay of X4 and R5 variants specifically. Coreceptor usage was assessed by first using PSSM, followed by phenotypic analysis of the encoded Env protein in a pseudotyped virus entry assay, for a subset of sequences. One subject (109) had a viral population with both X4 and R5 variants identified by PSSM and confirmed in an entry assay. In the other subject (101), distinct V3 variants failed to meet the cutoff value for a CXCR4 tropism designation by PSSM but exhibited strong CXCR4-tropic activity in the entry assay. In both of these subjects, X4 variants retained some ability to use CCR5.
HTA analysis of the V3 and V4-V5 regions of env of virus from subject 109 revealed two and four HTA variants, respectively (Fig. 2A and B). Recovery and sequencing of HTA bands, along with SGA and phylogenetic and phenotypic analysis of full-length env genes from the first time point, allowed us to link env sequences and their coreceptor usage phenotypes to the specific V3 and V4-V5 HTA bands for this subject. For example, X4 and R5 V3 variants formed distinct phylogenetic lineages that were each represented by one of the two HTA variants (Fig. (Fig.3A),3A), and V4-V5 HTA variants were linked to X4 V3 sequences by the presence of a deletion in each variable loop (data not shown). The full-length env sequences also formed distinct phylogenetic lineages according to V3 genotype and coreceptor phenotype (Fig. (Fig.3B).3B). The decay of X4- and R5-linked V3 and V4-V5 HTA variants is depicted in Fig. 2A and B, respectively. The total VL decayed with a half-life of 2.3 days, and over the course of a 2-log10 drop in VL, there was no significant difference in the decay rates of HTA variants relative to each other in subject 109 (Fig. 2A and B; Table Table11).
HTA analysis of virus from subject 101 revealed two variants in the V4-V5 region that decayed with similar rates (Fig. (Fig.2C2C and Table Table1).1). However, we were not able to amplify the V3 region from this subject, presumably as a result of subsequently identified V3 primer binding site mismatches in this viral population. In an alternative approach taken for this subject, SGA was carried out on each time point in order to assess the change in the relative proportions of all SGA-amplified env variants as VL declined. env sequences that exhibited strong X4 usage in the entry assay clustered together in a phylogenetic tree, as did their V3 sequences, indicating linkage to coreceptor use (Fig. (Fig.4).4). While not meeting the threshold value for X4 usage according to PSSM, these V3 variants had distinctly higher PSSM values and were more positively charged relative to the rest of the population. V4-V5 genotypes were only weakly linked to tropism (data not shown). Weak X4 entry activity was detected in some envelopes whose sequences were intermingled with those of exclusively R5 envelopes (Fig. (Fig.4A)4A) although the biological significance of this low-level X4 activity is unclear.
The decay rate of the total VL for subject 101 was a half-life of 2.3 days, and over the course of a 2-log10 drop in VL, there was no significant difference in the decay rates of V4-V5 HTA variants relative to each other (Fig. (Fig.2C2C and Table Table1).1). In addition, we did not detect a significant change in the proportion of env sequence variants, sampled by SGA, that clustered with phenotypically identified X4 or R5 variants at the two time points following initiation of therapy. X4 variants constituted 26% of the amplicons in the first time point and 20% in the third time point (P = 0.7) (Fig. (Fig.4A).4A). That we did not detect differential decay of X4 versus R5 variants in subjects 101 or 109 leads us to conclude that either (i) the bulk of the tropism variants are not compartmentalized in different cellular subsets, or (ii) while tropism variants may be compartmentalized in different cellular subsets, these infected cells have similar life spans when productively infected. Furthermore, any differential effects that X4 or R5 infection may have on the life span of the infected cell or any differences in susceptibility of these tropic variants to antiviral inhibition are not apparent in these data.
We next examined the potential genetic compartmentalization of coexisting R5 and X4 variants using an alternative approach. For this analysis, we included the entry samples of the two subjects identified above (101 and 109) and an additional nine subjects, obtained from studies described in the Materials and Methods section, with coexisting X4 and R5 subpopulations identified by PSSM analysis of V3 sequences. Using SGA, which eliminates confounding recombination during PCR, we generated amplicons containing env, nef, and the U3 region of the viral genome and carried out PSSM and phenotypic analysis on a representative subset of env variants from each subject. Phenotypic analysis was consistent with the result of the PSSM analysis for every amplicon tested in all but subject 101, as described above. CXCR4-tropic variants from all but two subjects in this study (138 and 411) also exhibited CCR5-tropic activity in the reporter assay but will be referred to as X4 variants. In many subjects, phylogenetic analysis of env sequences showed a deep branch point between the R5 sequences and a monophyletic group (i.e., a group that has descended from a single ancestral virus) of X4 sequences, suggesting that the outgrowth of X4 variants derives from a clonal event. Representative examples of phylogenies of sequences encompassing V1 of env through the 3′ U3 from three subjects are depicted in Fig. Fig.55.
If these X4 and R5 variants exist in distinct compartments with no migration of viruses between compartments, then the independent evolution of these virus subpopulations will produce a number of evolutionary signatures in regions of the genome outside of V3. We looked for these signatures using two methods applied to sequences from three regions of the genome: the V4-V5 region of env, the gp41 region of env, and the U3 region of the long terminal repeat. First, we used the SM test (83) to investigate whether independent evolution of X4 and R5 variants in distinct compartments had produced similar phylogenetic patterns in all genome regions. Essentially, our use of the SM test examines the expectation that X4 viruses should form a monophyletic group, regardless of the genome region used to build the tree, if this group is compartmentalized. Second, we used a distance-based metric, KST* (see Materials and Methods) (34, 74, 89), to investigate whether independent evolution of X4 and R5 variants was occurring in distinct compartments and producing genetic differentiation between the compartments in regions outside of V3. The KST* statistic is a measure of whether the genetic distance (number of nucleotide differences) between the X4 and R5 subpopulations is significantly greater than the genetic distance among viruses within the X4 or R5 subpopulations. Gene flow between compartments would erode both of these signatures of independent evolution, and gene flow is expected to have the strongest impact on genomic regions that are most distant from V3, due to the increased likelihood of recombination.
Observed SM and KST values were compared to the distribution of 1,000 random permutations of branches or sequences to determine the level of significance of separation. Because X4 and R5 subgroups were essentially defined by their V3 genotypes, which in most subjects were monophyletic groups with highly similar genotypes, recombination events were inferred when the sequences of the regions analyzed did not cluster according to their linked V3 genotype, indicated by migration events in the SM analysis. However, a lack of compartmentalization can result in increased sequence homogeneity in regions distal to V3, and detection of recombination between X4 and R5 sequences in these regions is possible only in cases where there remain strong phylogenetic signals that result in a well-supported tree.
The analysis revealed statistical support, by both KST and SM measures, for compartmentalization of two genetic populations defined by coreceptor use for markers proximal to V3, such as V4-V5 (Fig. (Fig.6).6). This would be expected as mutations closer to the population-defining V3 sequence would likely persist in linkage disequilibrium for a longer period of time after the outgrowth of the X4 V3 mutations, especially if they are functionally linked to V3. However, in most subjects, SM migration events between the R5 and X4 populations were increased in regions more distal to V3, such as U3 (Fig. (Fig.6A),6A), indicating that these populations were replicating at least part of the time in a shared cell type, providing the opportunity for recombination. Representative examples of detected recombination between X4 and R5 sequences and the predicted location of breakpoints, in these cases outside of gp120, are illustrated in Fig. Fig.7.7. This result is consistent with previous observations of recombination between X4 and R5 viruses made using different methods (51, 77, 92). We also observed decreasing values of genetic differentiation between X4 and R5 groups, as measured by KST, for regions farther from V3 (Fig. (Fig.6b).6b). However, in several subjects (135, 411, 432, and 1314) a compartmentalization signal in the U3 region remained statistically significant, if decreased, by both measures (Fig. (Fig.6,6, bottom panels). Taken together, these results show that while X4 and R5 env variants are genetically distinct, in many of our subjects there is little evidence of genetic compartmentalization between X4 and R5 variants in regions outside of env, such as in the U3.
There are several reasons to hypothesize that variants that appear in the blood are potentially compartmentalized. Previous studies have demonstrated compartmentalization of HIV-1 variants between different anatomical compartments, such as the central nervous system, genital tract, and different lymphoid tissues (13, 15, 24, 28, 39, 43, 70, 73, 77, 81, 90), as well as tissue microenvironments (24). The gut-associated lymphoid tissue, in particular, represents a major source of active replication of potentially compartmentalized CCR5-tropic virus populations (1, 68, 90, 94). It is possible that virus spatially compartmentalized in these anatomical sites may be represented as distinct variants in the peripheral blood. Because the initiation of HAART abruptly blocks new rounds of HIV-1 infection, presumably without impacting viral RNA production from cells already infected, different decay rates of compartmentalized variants following suppression of viral replication should reflect different life spans of the cells from which they are emerging. Viral populations compartmentalized in either cell types or tissues that experience differential drug exposure may also decay at different rates if viral replication continues at some level in the presence of a suboptimal drug concentration. However, our observation that all detectable HIV-1 genetic variants declined at comparable rates suggests that the vast majority of the coexisting HIV-1 subpopulations in the peripheral blood are not compartmentalized either in cell types with different life spans or in cells or tissues with various degrees of antiretroviral drug bioavailability.
Our use of the HTA allowed for relatively sensitive detection of variants potentially comprising as little as 1 to 3% of the population (33, 79). However, one important limitation of this study is the sensitivity for detection of minority viral populations below this 1% threshold, which may be produced by cells with different life spans. Such viral populations almost certainly exist, based on the biphasic decay kinetics of the bulk HIV-1 RNA load during HAART (4, 49, 62). It has been hypothesized that the second phase of decay of the bulk HIV-1 RNA load represents virus produced by cells with a longer life span, presumably cells of the monocyte lineage. Compartmentalization of viral DNA populations between CD4+ T cells and monocytes has been reported (21, 47, 102), and assuming that monocytes have longer half-lives relative to activated CD4+ T cells, variants compartmentalized in these different cell types would be expected to exhibit different rates of decay during HAART (11, 34, 62, 96). However, the proportion of variants compartmentalized in productively infected monocytes may be too small to be detected in our assay (62, 63), and monocytes may not be productively infected and only produce virus upon entry into tissue and differentiation (72, 84). Furthermore, the decay characteristics of HIV-1 upon initiation of therapy that includes an integrase inhibitor suggest that much of the second phase of decay observed in conventional therapy represents cells that are slowly undergoing integration and that the proportion of productively infected, long-lived cells is smaller than previously thought (53, 80). Another limitation of this study is that it depends on the assumption that compartmentalized subpopulations can be distinguished by their env genotypes and, in particular, genotypes that can be resolved by HTA. However, in addition to its function in determining host cell tropism, the extreme genetic complexity of env within infected individuals makes it a highly sensitive target for the detection of coexisting viral subpopulations, and any other genomic region that may drive HIV-1 compartmentalization would likely be linked to distinct env variants as a result of founder effects, genetic isolation, or compartment-specific evolution. Furthermore, any compartmentalized variants would have likely diverged enough to be resolved by HTA analysis (30). Thus, we can conservatively conclude that the lack of genetic compartmentalization and the differential decay rates observed in this study apply to the bulk of the HIV-1 population in the peripheral blood that represents primarily the first phase of viral RNA decay during HAART.
Another potential opportunity for cellular compartmentalization is between naïve and memory CD4+ T cells. While naïve and memory T cells express similar levels of CXCR4, CCR5 is expressed only in memory cells (5, 46, 55). Previous studies have found a wide range of preferential infection by, and potential compartmentalization of, X4 and R5 variants in these cell types in a manner consistent with their coreceptor expression patterns (3, 57, 91). If X4 and R5 variants are compartmentalized in these two cell types to a significant degree, then similarity in decay rates would indicate that the life spans of infected naïve and memory cells are similar when they become activated and produce virus. It is thought that the bulk of viral replication occurs in activated CD4+ memory T cells (34, 49, 62, 86, 88, 100), in which case any potential compartmentalization of R5 and X4 variants observed in resting memory and naïve cells may represent only a small fraction of the total population. Also, activated and previously activated T cells express both CCR5 and CXCR4 (5, 55, 58), providing a potential source of mixing of R5 and X4 variants. There is evidence to suggest that the pool of cells supporting the bulk of virus replication is not homogeneous in its susceptibility to infection by X4 and R5 variants (22) and that X4 and R5 variants may be differentially affected by antiretroviral therapy (22, 66). However, we found no difference in the decay rates of X4 and R5 variants upon initiation of therapy, and the decay rates of these variants are within the range reported in other studies for the first phase of decay, presumably reflecting the life spans of the activated memory cells supporting ~99% of the virus population (34, 49, 62, 63, 96). This finding is consistent with a model where virus is emerging from a homogeneous pool of cells that is sufficiently susceptible to infection by both X4 and R5 variants to account for most of the production of these variants found in the periphery. This is also further supported by the finding of a lack of genetic compartmentalization between X4 and R5 populations in both subjects 101 and 109 for regions outside of env, indicating some overlap of target cell types. However, the lack of data indicating differential decay rates of variants does not allow a definitive conclusion to be drawn regarding the half-lives of infected cells in different cellular subsets until the degree of cellular compartmentalization of X4 and R5 variants can be more fully and directly accounted for in studies of this type.
The divergent X4 and R5 lineages indicate some degree of genetic compartmentalization between these variants. This could be due to physical isolation in different cell types or to genetic linkage selected across env for the ability to use different coreceptors. However, we detected an overall lack of compartmentalization and evidence of recombination between these populations in sequence regions increasingly distal of 3′ of V3, suggesting the potential for sequence mixing between R5 and X4 variants in a coinfected cell (Fig. (Fig.66 and and7).7). This observation is consistent with previous reports that have identified X4/R5 recombinants both within env and between env and other regions of the genome (51, 77, 92). However, our use of the single genome amplification approach avoided the possibility of recombination during PCR, which may have created artificial recombinants in some previous studies. These data support the conclusion that while X4 and R5 variants may preferentially replicate in distinct cellular compartments, they are not genetically isolated and must with some frequency infect the same cell types. Still, the deep branch points in the phylogenetic trees suggest that the initial outgrowth of X4 variants is from a monoclonal genotype.
This study found little evidence for differential decay and compartmentalization of env variants comprising the bulk of the virus in the peripheral blood, even in the case of divergent coreceptor phenotypes, indicating that HAART is equally effective on all the detectible variants making up the bulk virus in the peripheral blood. However, new technologies are becoming available that will allow sampling to below 1% (52), and the application of these technologies may yet reveal minor populations that exhibit differential rates of decay upon initiation of therapy.
This work was supported by NIH grant R37-AI44667 to R.S.; an AmFAR award to P.H.; NIH training grant support to W.I. (T32-GM07092), G.S. (T32-AI07001), M.P.C. (T32-AI07419), and P.H. (T32-CA09156); the UNC Center For AIDS Research (NIH award P30-AI50410); and the UNC General Clinical Research Center.
We thank Dale Kempf from Abbott and the AIDS Clinical Trial Group 359 study team for making samples available for this study.
Published ahead of print on 11 February 2009.