Control of HIV-1 in Elite Suppressors despite Ongoing Replication and Evolution in Plasma Virus

Abstract

A subset of HIV-1-infected patients known as elite controllers or suppressors (ES) control the virus naturally. We have previously demonstrated sequence discordance between proviral and plasma gag clones in ES, much of which can be attributed to selective pressure from the host (J. R. Bailey, T. M. Williams, R. F. Siliciano, and J. N. Blankson, J. Exp. Med. 203:1357-1369, 2006). However, it is not clear whether ongoing viral replication continues in ES once the control of viremia has been established or whether selective pressure impacts this evolution. The cytotoxic T-lymphocyte (CTL) response in ES often targets Gag and frequently is superior to that of HIV-1 progressors, partially due to the HLA class I alleles B*57/5801 and B*27, which are overrepresented in ES. We therefore examined longitudinal plasma and proviral gag sequences from HLA-B*57/5801 and -B*27 ES. Despite the highly conserved nature of gag, we observed clear evidence of evolution in the plasma virus, largely due to synonymous substitutions. In contrast, evolution was rare in proviral clones, suggesting that ongoing replication in ES does not permit the significant reseeding of the latent reservoir. Interestingly, there was little continual evolution in CTL epitopes, and we detected de novo CTL responses to autologous viral mutants. Thus, some ES control viremia despite ongoing replication and evolution.

Chronic HIV-1 infection normally is characterized by ongoing viral replication and declining CD4⁺ T-cell counts. The low fidelity of HIV-1 reverse transcriptase results in high viral mutation rates, permitting the evasion of the host immune response through escape mutations (as reviewed in reference 17), and without highly active antiretroviral therapy (HAART), infection ultimately progresses to AIDS. Rare, infected individuals known as elite suppressors (ES), however, maintain control of HIV-1 without antiretroviral therapy (18, 23, 26, 35). The ES in our cohort have controlled the virus for several years, maintaining mostly normal CD4⁺ T-cell counts and plasma HIV-1 RNA levels below the limit of detection of clinical assays (50 copies/ml).

The level of viremia seen in ES is comparable to that of patients responding well to HAART. Patients on HAART have a very low level of residual viremia that does not show evidence of ongoing evolution (5, 34, and T. P. Brennan, unpublished data) and is not decreased by the intensification of therapy (28). It is most likely derived from the activation of latently infected resting CD4⁺ T cells and from other stable reservoirs rather than from ongoing cycles of replication. The source of viremia in ES, however, is unknown. ES may completely suppress viral replication in a manner comparable to HAART, or ongoing cycles of viral replication may continue in ES, which the patients nonetheless are able to control.

The cytotoxic T-lymphocyte (CTL) response has been associated with the control of HIV-1 in ES. Primary CD8⁺ T cells from many ES are able to suppress HIV-1 replication ex vivo (39, 40) by efficient granzyme B- and perforin-mediated killing of infected T cells (31), and the presence of a polyfunctional (1, 7) and proliferative (30) CD8⁺ T-cell response to HIV-1 antigens appears to be critical for the control of viremia. Additionally, the HLA class I alleles B*57, B*5801, and B*27 are overrepresented in ES (15, 19, 31, 32, 37, 41). Several peptides presented by these alleles are located in the Gag polyprotein, which includes highly conserved, structural components of the virion. The structural nature of Gag makes it difficult for escape mutations to develop without negatively affecting viral fitness, likely increasing the effectiveness of immune responses targeting this polyprotein (11, 25, 27).

We and others have shown previously that CTL exert selective pressure on HIV-1 replication in ES (3, 6, 29, 38). At a single time point, we observed that sequences of provirus in resting CD4⁺ T cells and virus in the plasma of ES segregate almost completely, with mutations in CTL epitopes found predominately in plasma clones (3, 6). No phylogenetic analysis was performed in this early study, however, and segregation is not necessarily suggestive of ongoing replication. CTL escape mutations develop rapidly in HLA-B*57/5801-restricted epitopes in the acute phase of infection (12, 13), even in patients who achieve impressive early control of viral replication (16, 36). Thus, the differences between proviral and plasma virus sequences may reflect evolution that occurs during acute infection, before the control of replication is established. Examining longitudinal proviral and plasma samples is necessary to determine whether viral evolution continues in ES despite the suppression of viremia to <50 copies/ml. Here, we examined proviral and plasma clones from six HLA-B*57/58 and/or -B*27 ES for a period spanning 5 years to determine whether evolution is continuing and, if so, whether the CTL response is influencing that evolution. We focused specifically on HLA-B*57/58- and -B*27-restricted epitopes in the Gag polyprotein. Additionally, we characterized the CTL response to autologous viral variants at these epitopes to understand the relative effectiveness of CTL against these mutants.

MATERIALS AND METHODS

Patients.

Clinical information on the six ES studied is documented in Table 1.

TABLE 1.

Patient characteristics^a

Patient	Yr of diagnosis	Last CD4+ T-cell count (cells/μl)	Mean viral load (copies/ml)	Allele(s)b
				HLA-A	HLA-B
ES1	1999	980	<50	24, 26	14, 27
ES3	1991	1,161	<1	25, 68	51, 5702
ES7	1994	1,104	51	30, 32	81, 5703
ES8	2003	626	26	02, 03	44, 5703
ES9	1999	1,027	<1	02, 30	2703, 5703
ES31	2008	758	<50	03	2705, 5801

^aViral loads from ES1 and ES31 are from a commercial ultrasensitive assay with a limit of detection of 50 copies/ml. Viral loads from ES3, ES7, ES8, and ES9 were obtained from an ultrasensitive assay with a limit of detection of 1 copy/ml. ES7 has routinely had viral loads of <50 copies in commercial ultrasensitive assays. The year of diagnosis is not necessarily indicative of the year of infection.

^bProtective alleles are shown in boldface.

Genomic DNA and RNA isolation.

Magnetic bead depletion was performed on peripheral blood mononuclear cells (PBMCs) to enrich for resting CD4⁺ T cells as described previously (6). Where both activated and resting CD4⁺ T cells were isolated, cells were negatively selected using the Miltenyi Biotech human CD4 isolation kit and then were stained with CD8 phycoerythrin (PE), CD16 PE, and HLA-DR fluorescein isothiocyanate (FITC). From the PE-negative cells, FITC-positive (HLA-DR⁺) and FITC-negative (HLA-DR⁻) cells were isolated via fluorescence-activated cell sorting (FACS). Genomic DNA and plasma RNA were purified as previously described (6).

PCR and RT-PCR amplification of gag.

To prevent resampling, gag genes were amplified from provirus in genomic DNA and from plasma-derived RNA by limiting-dilution digital nested PCR as previously described (6, 29). Briefly, gag was amplified via nested PCR utilizing a SuperScript III one-step reverse transcription-PCR (RT-PCR) system with Platinum Taq HiFi for the outer and inner reactions (Invitrogen). Proviral gag was amplified using nested PCR with Platinum Taq HiFi. PCR dilutions in which less than one in four reactions was positive were considered to be clonal.

Immune responses.

Enzyme-linked immunospot (ELISPOT) analysis was performed as previously described (4) with gamma interferon (IFN-γ) antibodies from Mabtech Inc. Triplicate reaction mixtures containing 100,000 PBMC/well were performed for each concentration of each peptide. The peptides were synthesized at Genemed Synthesis Inc. (San Antonio, TX). All plates were evaluated with an automated ELISPOT reader system (Carl Zeiss MicroImaging, Inc.) with KS4.8 software by an independent scientist in a blinded fashion (Zellnet Consulting). Negative controls consistently had fewer than 3 spot-forming cells (SFC) per well. Peptides that stimulated more than 5 SFC/well were considered epitopes. Intracellular staining was performed for ES1 using FITC-conjugated IFN-γ-specific monoclonal antibodies (MAbs) and allophycocyanin (APC)-conjugated pentamers consisting of HLA-B*27 and the immunodominant KK10 Gag epitope (Coulter). A total of 500,000 events were analyzed for each sample.

Phylogenetic analysis and statistics.

All clonal sequences obtained were included in the phylogenetic analysis, with the exception of APOBEC-mediated hypermutated sequences, which were removed. Sequences subsequently were handled as previously described (10).

Classical, maximum-likelihood, and Bayesian phylogenetic reconstruction for each patient was performed as previously described (10).

Nonsynonymous and synonymous p-distance calculations and the number of differences were based on the Nei-Gojobori method (33) and were calculated by comparing grouped sequences from the early time point(s) in each patient to sequences from the later time point(s) using MEGA 4.0 software (44). The definition of early and late sequences varied for each patient due to differences in the samples available for each patient, but samples from 0 to 2 years since the start of the study are considered early, and those from time points 3 to 5 years from the start of the study are considered late. For ES31, data were available only for 2 years, so data from the first year were considered early and those from the latter year were late.

Nucleotide sequence accession numbers.

All sequences have been submitted to GenBank and have been assigned accession numbers GU993311 to GU993518.

RESULTS

To address whether ongoing viral replication occurs in ES, we carried out a longitudinal study of proviral and plasma gag sequences from six HLA-B*57/5801 and/or -B*27 patients. We chose to focus on gag as it is a highly conserved gene, and because epitopes in Gag restricted by HLA-B*27 and -B*57/5801 are immunodominant and likely reflect ongoing CTL-mediated selective pressure. ES maintain extremely low levels of virus in their plasma (<50 copies/ml), but we obtained clonal sequences of the gag gene from free virus in the plasma at multiple time points using an extremely sensitive RT-PCR technique. Clonal proviral samples were derived from both resting (HLA-DR⁻) and activated (HLA-DR⁺) CD4⁺ T cells, when possible.

Evolution occurs in gag of the plasma virus of ES.

Phylogenetic analysis of clonal gag sequences revealed clear evolution in the plasma virus of ES over time. As shown in Fig. 1 and 2, the 23, 10, and 17 plasma clones from the early time points in ES7, ES9, and ES8, respectively, were clearly ancestral to those obtained from samples 4 to 5 years later (Fig. 1 and 2). A similar trend was observed with the smaller number of plasma clones obtained from ES1 and ES3 (Fig. 1). Notably, evolution was apparent between samples that were acquired 3 to 4 years apart, but sequences separated by only 1 or 2 years generally clustered together, as exemplified by ES9 and ES8. Indeed, ES31, for whom we could obtain samples only from time points 1 year apart, did not show evolution of plasma virus. Taken together, these results suggest that during the span of several years, ongoing evolutionary change is detectable in the plasma virus of ES. It is noteworthy that we saw evidence of evolution in ES9 in spite of the fact that this patient had a viral load of <1 copy/ml. A larger study will be needed to determine whether evolution occurs to a greater extent in ES who have higher levels of viremia.

FIG. 1.

Phylogenetic analysis of gag in elite suppressor patients. Phylogenies were estimated by using a classical approach, functioning under a maximum-likelihood (ML) optimality criterion. All sequences are clonal, and APOBEC-mediated hypermutated sequences were removed from analysis. Bootstrap values of 50 and higher are displayed. Protective HLA B alleles are noted beneath the patient number. Patients are not related in any way but are placed on the same tree for display purposes. Colors indicate time, with the scale below in years. Triangles represent clonal plasma sequences, squares represent proviral sequences from HLA-DR⁺ CD4⁺ T cells, and circles represent clonal proviral sequences from HLA-DR⁻ CD4⁺ T cells. To the right, escape mutations in HLA-B*57- and -B*27-restricted epitopes are denoted in black or gray and aligned with the appropriate symbol on the tree. Escape mutations are listed above the grid, with the epitopes in which the mutations occurred listed above the mutations.

FIG. 2.

Phylogenetic analysis of gag in ES8. Phylogenies were constructed as described for Fig. 1; ES8 is isolated due to the large number of sequences obtained for this patient. Colors indicate time, with the scale below in years. Triangles represent plasma sequences, squares represent proviral sequences from HLA-DR⁺ CD4⁺ T cells, and circles represent proviral sequences from HLA-DR⁻ CD4⁺ T cells. To the right, escape mutations in HLA B*57- and -B*27-restricted epitopes are denoted in black or gray and aligned with the appropriate symbol on the tree. Escape mutations are listed above the grid, with the epitopes in which the mutations occurred listed above the mutations.

In contrast to the evolution seen in plasma, no appreciable evolution was seen in proviral clones of this cohort. Proviral sequences from all time points clustered together, leaving no temporal footprint (Fig. 1 and 2). In addition, proviral clones from all six ES remained remarkably segregated from, and ancestral to, plasma clones. This is in stark contrast to the close phylogenetic relationship observed between proviral and plasma sequences in chronic progressors (43).

The amino acid sequences of plasma virus from each patient showed only rare nonsynonymous mutation and were strikingly similar in each patient during the course of several years. We used the Nei-Gojobori method to compare the p-distances of nonsynonymous and synonymous mutations in each patient between early and late plasma sequences. This parameter normalizes the frequency of substitutions to the number of possible synonymous or nonsynonymous sites, permitting comparisons between the two. As shown in Fig. 3A, synonymous mutations were a more significant factor in the evolution of plasma virus than were nonsynonymous mutations, possibly reflecting the highly conserved nature of the structural polyprotein that gag encodes. Figure 3B illustrates the relative incidence of synonymous and nonsynonymous mutation in the proviral clones of three ES for whom we had longitudinal proviral samples. Both nonsynonymous and synonymous mutations were less common in proviral samples than in plasma. This reflects the lack of evolution in the proviral compartment of ES. We additionally show the calculated number of differences between early and late plasma and proviral sequences from ES (Fig. 3C and D). These data again illustrate that there is significantly more synonymous mutation than nonsynonymous mutation, and that mutations are more frequent in the plasma virus than in the provirus.

FIG. 3.

Analysis of synonymous and nonsynonymous mutation in the plasma virus and proviral compartments. Shown are p-distance values for plasma (A) and proviral (B) sequences as determined by comparing early and late samples for each patient utilizing the Nei-Gojobori method. The numbers of differences also were calculated for plasma (C) and proviral (D) sequences using the Nei-Gojobori method.

Relationship between evolution of plasma virus in ES and CTL pressure.

Having established that evolution is occurring in the plasma virus, we hypothesized that CTL pressure would be an ongoing driving force for it. We identified putative CTL-targeted epitopes by examining documented HLA-restricted epitopes from the LANL database (22) and by identifying epitopes targeted by the patients via whole-proteome ELISPOT analysis (data not shown and reference 5).

Strikingly, there was little evidence of continuing evolution in HLA-restricted epitopes. Among the five patients for whom we had samples greater than 3 years apart, only 20 sites of nonsynonymous mutation were present in two or more plasma clones either from late or early time points but not from both time points. Of these 20, only 6 were in or adjacent to CTL-targeted epitopes. Additionally, plasma virus in one patient, ES7, showed convergent evolution in two HLA-B*57-restricted epitopes, QW10 (Gag 308-316) and IW9 (Gag 147-155).

Plasma clones in ES7 showed the most nonsynonymous evolution, and these clones were the only ones in which concerted change occurred in HLA-B*57-restricted epitopes. The E312D mutation in the QW10 epitope reverted over time and was supplanted by the I147L mutation in IW9. Interestingly, the E312D mutation was present in a fraction of proviral sequences from all time points, and it appears that the reversion of this mutation and the appearance of I147L occurred independently in both early and late plasma clones (Fig. 4). This is likely an example of convergent evolution, wherein a common E312D ancestor diverges along two distinct evolutionary paths, only to converge later to I147L. Additionally, late plasma clones from ES7 possessed the S173T mutation, which lies immediately adjacent to the KF11 epitope and has been seen previously in 5 out of 17 HLA-B*57 progressors (29). In contrast, all plasma clones from ES3 and ES9 had identical mutations in the HLA-B*57/HLA-B*58-restricted epitopes at all time points sampled, displaying no ongoing evolution in these epitopes over time (Fig. 4). ES1 had variation in the HLA-B*27 restricted KK10 epitope, but this variation was seen in both early and late time points and did not change over time. ES8 was unique in that he showed fluctuation in the TW10 epitope, but there was no clear resolution in the epitope during a period of 5 years.

FIG. 4.

Relevant sequence regions from clonal, near-full-length gag amplified from plasma (pl) or from either resting (HLA-DR⁻) or activated (HLA-DR⁺) CD4⁺ T cells. The date of sample acquisition and number of clones that are identical to the displayed sequences are noted. The HLA-B*27-restricted epitope KK10 (Gag 263-272) and HLA-B*57-restricted epitopes IW9 (Gag 147-155), KF11 (Gag 162-172), TW10 (Gag 240-249), and QW9 (Gag 308-316) all are denoted in shaded boxes. Sites of compensatory mutations for KK10 (S173) (42) and TW10 (H219, I223, M228) escape mutants also are shaded, and for ES7 the DL15 epitope is displayed. Sequences from 2004 and 2005 for ES3, ES7, ES8, and ES9 have been previously reported and are shown for comparative purposes only.

Aside from S173T in ES7, the five other nonsynonymous mutations that occurred within CTL-targeted epitopes were observed in ES8 and, again, in ES7. In an ELISPOT assay, ES7 targeted the Gag 329-343 epitope that later developed a K335R mutation. This mutation was present in all of the late plasma clones as well as in one proviral clone, but it was not present in early plasma clones (Fig. 4). ES8, in contrast, possessed four mutations in two different HLA-A*2 epitopes (Gag 275-283 and Gag 368-376), but these epitopes were not targeted by the patient in an ELISPOT assay (data not shown).

Mutations in HLA-restricted epitopes of ES provide limited escape.

We were interested in determining the impact of observed mutations in CTL epitopes on the patients' ability to respond to these epitopes. Because HLA-B*27- and -B*57/5801-restricted epitopes are immunodominant in the majority of patients, we examined the impact of mutations in these epitopes. We have shown previously that de novo or cross-reactive responses were made to many autologous TW10 and IW9 variants (5). For example, ES7 has comparable responses to the wild-type IW9 epitope and the I147L mutant. In contrast, he had no detectable response to both wild-type and mutant QW10 peptides. We thus limited our analysis here to previously undescribed escape mutations.

ES9 and ES1 both possess the HLA-B*27 allele and show discordance between plasma and proviral sequences in the immunodominant KK10 epitope (Gag 264-272). They do not, however, show subsequent evolution in this epitope. We have shown previously that ES9 N271H and L268M KK10 variants and to wild-type KK10 responds to in a similar fashion (6). We used intracellular staining to measure the IFN-γ response of ES1 to wild-type and autologous KK10 variants. ES1 showed a reduced response to the autologous variant, proving that the variant represented an escape mutation (Fig. 5A).

FIG. 5.

(A) Intracellular IFN-γ staining analysis of ES1 HLA-B27 tetramer-positive cells to autologous and wild-type KK10 (KRWIILGLNK; Gag 263-272) peptides. (B) IFN-γ ELISPOT analysis of ES8 CD8⁺ T cells to autologous TW10 variants and wild-type TW10 peptide (TSTLQEQIGW; Gag 240-249). (C) IFN-γ ELISPOT analysis of ES31 CD8⁺ T cells to autologous TW10 variant and wild-type TW10 peptides. (D) IFN-γ ELISPOT analysis of the ES7 CD8⁺ T cells to the wild-type and the autologous peptide containing the K335R mutation in DL15 (DCKTILKALGPAATL; Gag 329-343). (E) Magnitude of IFN-γ response by ES8, ES3, ES7, and ES9 to four epitopes in which evolution occurred in the plasma virus of either ES7 or ES8. Data reflect the IFN-γ response to the autologous variant of the epitope of each patient at the latest time point available. Open symbols indicate SFC < 50. Circles indicate the response by patients in whom no evolution was seen in this epitope, while diamonds indicate the response by the patient in whom the evolution of the plasma virus occurred in this epitope (ES8 for TW10 and ES7 for IW9, QW9, and KF11).

ES3, ES7, ES8, and ES31 all showed discordance between proviral and plasma sequence at the HLA-B*57/5801-restricted epitope TW10 (TSTLQEQIGW; Gag 239-249) (Fig. 4). Escape from the CTL response to this immunodominant epitope most frequently occurs via the T242N mutation. While this mutation was seen in plasma virus in ES3, ES7, and ES9, virus from ES8 and ES31 had alternative mutations in this epitope. ES8 acquired multiple, rare combinations of mutations (Fig. 4). Of 974 B clade Gag sequences in the LANL database (24), only three have mutations at Q244. Two of these three are Q244A, which is seen in ES8 samples from later time points. In contrast, 132 of the LANL database sequences exhibited mutation at T242, the majority of these being T242N. The unique combinations of mutations seen in ES8 did not confer escape from the CTL response, however, due to the appearance of de novo responses to the mutant epitopes (Fig. 5B). We previously have documented that while plasma variants with the T242N mutation conferred escape from TW10-specific CTL responses, ES8 developed a robust de novo IFN-γ response to the plasma Q244T, I247V, G248A TW10 triple mutant that was present in earlier time points (6). We looked for IFN-γ secretion to this and four more recent plasma variants 4 years later and found vigorous responses to all of the peptides tested, even at extremely low concentrations of peptide (Fig. 5B).

ES31 expressed a G248A variant in the TW10 epitope of all plasma samples that was not observed in the provirus. ELISPOT analysis revealed that, as with ES8, ES31 maintained a response to the wild-type epitope while mounting a strong de novo response to the autologous viral variant (Fig. 5C).

We also compared the response of CD8⁺ T cells from ES7 to that of the wild-type Gag 329-343 epitope and the autologous K335R mutant. Interestingly, the K335R mutation conferred almost complete escape from the CTL response (Fig. 5D).

We compared the baseline IFN-γ responses to those of the four HLA-B*57-restricted epitopes (6) in ES3, ES7, ES8, and ES9 to determine whether a stronger IFN-γ response was more likely to induce evolution. As shown in Fig. 5E, no consistent relationship was found between the magnitude of the immune response and viral evolution.

In sum, the mutations observed in the ES had various levels of impact on the CTL response. Interestingly, the mutation that had the most dramatic impact on the CTL response was the K335R mutation in ES7, which arose late and was not present in any plasma samples from early time points. In contrast, many mutations in the HLA-B*57/5801 and -B*27 epitopes had only a minor impact on the CTL response (Fig. 5 and reference 6). This is notable, considering that these immunodominant epitopes are thought to be important in the control of viremia. It is unclear what drove the selection of these mutants, considering that many had little impact on the CTL response. These mutations may be linked to beneficial mutations within other regions of the viral genome. More likely is that these mutations did aid the virus in evading the CTL response at the time at which they arose, but that a de novo response was elicited by the mutants, which abrogated the benefit. Presumably the mutations then were retained if they did not negatively affect viral fitness.

DISCUSSION

ES maintain incredibly low levels of viremia naturally, but the mechanism by which they do so is unclear. Central to understanding how ES control viremia is determining whether virus in ES actually engage in ongoing rounds of viral replication. Patients on effective HAART regimens maintain viral loads comparable to those of ES (14, 20, 31, 38), and residual viremia in HAART patients appears to be derived exclusively from the reactivation of the latent reservoir. It is conceivable that ES control mirrors HAART suppression: patients would naturally suppress viral replication after the initial phase of acute infection, preventing further ongoing replication events, and residual viremia therefore would be derived strictly from the latent reservoir. Were this the case, we would anticipate that no further evolution in the plasma virus of ES could be observed.

We examined longitudinal plasma and proviral gag sequences from six HLA-B*27 and -B*57/5801 ES and observed clear evolution in the plasma virus. Note that we amplified and analyzed a large number of clonal gag plasma sequences. This is extremely challenging, as many ES have less than 1 copy of HIV-1 RNA per ml of plasma (14, 20, 31, 38). We focused on gag because we wished to explore the relative import of CTL-targeted epitopes on ongoing evolution, and immunodominant epitopes presented by HLA alleles B*57/5801 and -B*27 are located in gag. It is thought that the CTL response to epitopes presented by these alleles is one cause of control in ES. Thus, we hypothesized that these epitopes would be under considerable selective pressure.

Interestingly, the majority of the observed evolution was derived from synonymous mutation and not from evolution in CTL-targeted epitopes. There are several possible explanations for these data. Free plasma virus in these patients may have reached a patient-specific fitness peak, such that purifying selection dominated the evolution in the population. The fixation of synonymous mutation in these patients may then be due to linkage between these mutations and mutations that have developed in other parts of the viral genome in response to host immune pressures, including NK cells and neutralizing antibody. Alternatively, they may be the result of stochastic processes. Given a sufficiently low effective population size and other contributing factors (45), plasma virus in these patients could be subjected to a significant amount of genetic drift. Despite the lack of evolution in CTL-targeted epitopes, it is of note that escape mutations present in the plasma virus did not revert. The persistence of the mutations implies that CTL continue to exert pressure on the virus.

While plasma virus evolved over time, the vast majority of proviral samples were ancestral to plasma clones and were almost completely segregated from the plasma sequences. This is notable, as we have cultured virus from resting CD4⁺ T cells in ES1, ES8, and ES9, proving that some of the virus in this compartment is replication competent (4, 8). This is in contrast to a prior study of an ES that attributed the lack of evolution of proviral sequence to excessive G-to-A hypermutation (46). Three ES, however, possessed rare proviral clones that showed a close phylogenetic relationship with plasma virus and that contained CTL escape mutations. ES9 was infected by an HLA-B*57 patient, and thus escape mutations in HLA-B*57-restricted epitopes in all proviral clones likely were transmitted (4). However, 1 of 32 proviral clones in this patient also had the L268M mutation in the HLA-B*27-restricted KK10 epitope and thus clustered phylogenetically with plasma clones (Fig. 1 and 4). Similarly, 2 out of 55 proviral clones in ES8 and 1 of 16 proviral clones in ES1 clustered with plasma clones (Fig. 1 and 2). The proviral clones from ES1 and ES9 were from resting CD4⁺ T cells, suggesting that limited reseeding of the latent reservoir is possible in ES. In ES8, 2 out of 25 proviral clones from activated CD4⁺ T cells contained CTL escape mutations. Thus, a small subset of activated CD4⁺ T cells potentially serves as one source for circulating plasma virus. Sequence analysis of a larger number of proviral clones from both resting and activated CD4⁺ T cells will help further define the relationship between plasma and provirus in ES.

Despite these exceptions, the lack of evolution in the proviral compartment suggests that ongoing replication is occurring at a sufficiently low rate so as to prevent significant reseeding of the latent reservoir and could suggest that plasma clones are derived from a viral reservoir other than peripheral blood CD4⁺ T cells. While elevated lipopolysaccharide (LPS) levels seen in some ES (2, 9, 21) suggest ongoing viral replication in the gastrointestinal tract, there have been no studies directly supporting this hypothesis. Studies analyzing viral clones obtained from lymphoid tissue in ES are needed to address this critical issue.

Our study for the first time documents the evolution of plasma HIV-1 in ES, demonstrating that ES control viremia in spite of ongoing viral replication. The majority of evolution in gag of this cohort was driven by synonymous mutation. While we observe the persistence of mutations in HLA-B*57/5801- and -B*27-restricted epitopes, there was minimal continuing evolution in these epitopes, suggesting that an optimal balance between immune evasion and viral fitness has been achieved. The data suggest that a therapeutic vaccine does not need to completely control viral replication in order to be effective.