INTRODUCTION

Top Abstract Introduction Materials and Methods Results Discussion References

Human immunodeficiency virus type 1 (HIV-1) infection is characterized by an asymptomatic period of highly variable length between acute infection and AIDS (32, 38, 42, 53-55, 72). Among the viral factors that may affect the rate of HIV disease progression, HIV strains with a syncytium-inducing (SI) phenotype on MT-2 cells (now referred to as X4 or R5/X4 for viruses with dualtropism that includes CXCR4 [3]) have long been associated with faster disease progression, although the mechanism of this effect is unclear (7, 13, 22, 36, 64, 70, 78). Also unclear is the role of viral evolutionary change in affecting the duration of the asymptomatic period. While several studies have found an inverse relationship between the rate of viral diversification and disease progression (15, 16, 23, 45, 46, 73, 82, 83), others have not (26, 49, 52); one study found instances of both inverse and direct relationships (51).

To understand more clearly the relationship of viral evolutionary changes, the emergence of X4 viral phenotypes, and the rate of progression over the course of HIV-1 disease, we studied the evolution of the C2-V5 region of the HIV-1 env gene in nine homosexual men enrolled in the Multicenter AIDS Cohort Study (MACS) (34). These nine men constituted a subset of individuals previously selected for a study of changes in plasma viremia, T-cell subsets, and cytotoxic memory T-cell responses in MACS participants who had a downward inflection in CD3⁺ T-cell numbers (65). A high prevalence (75%) of such inflections has been found in MACS participants who have developed AIDS, occurring a median of 1.7 years prior to the onset of AIDS (24). The C2-V5 region of env was chosen for the analysis of viral changes because it encodes an important target for immune responses, determines coreceptor specificity, and exhibits a high degree of phylogenetically informative variability (41, 42). An average of 12 time points per person were studied, covering 6 to 12 years of infection.

As the present study is one of the most comprehensive longitudinal studies of HIV-1 evolution in vivo to date, it provides unique insights into the patterns of viral change over time, as well as the emergence and representation of X4 viruses and changes in T-cell subsets associated with disease progression. The observed patterns and associations may explain inconsistencies among previous reports relating HIV-1 sequence evolution to disease progression and enhance our understanding of the factors that lead to progression.

	MATERIALS AND METHODS

Top Abstract Introduction Materials and Methods Results Discussion References

Study participants and clinical specimens. The MACS is an ongoing prospective cohort study, with over 5,000 homosexual men participating, whose study design has been described in detail (34). At semiannual visits, participants provide responses to structured, interviewer-administered questionnaires (including detailed information about therapies taken since the last visit) and provide blood specimens that are stored at 70°C (serum and plasma) or 135°C (peripheral blood mononuclear cells [PBMCs]) at a central repository. T-cell subsets were determined by flow cytometry by previously described methods (25, 68), with antibodies obtained from Becton Dickinson Immunocytometry Systems (San Jose, Calif.) and an ELITE flow cytometer (Coulter Electronics, Hialeah, Fla.).

PCR amplification, cloning, and sequencing. Ficoll-Hypaque-purified PBMCs were lysed by the addition of quick lysis buffer (10 mM Tris [pH 8.3], 50 mM KCl, 2.5 mM MgCl2, 0.45% Nonidet P-40, 0.45% Tween 20) and digested with 100 µg of proteinase K per ml at 56°C for 1 h. Lysates from PBMCs were treated with GeneReleaser (BioVentures, Murfreesboro, Tenn.) according to the manufacturer's protocol and used for PCR. Nested PCR with hot start (19) was used on serial dilutions of the lysate, with amplification of the entire gp160 in the first round followed by amplification of the C2-V5 region in the second round. Template quantitation was done by measuring the frequency of positive and negative reactions at different dilutions (66). Primer sequences for ED3, ED12, ED31, DR7, and DR8 have been described previously (14, 17, 45); primer Nef3 is 9028-5'-TAAGTCATTGGTCTTAAAGGTACC, where the number corresponds to the position in the NL4-3 genome (GenBank accession no. M19921) of the 5' nucleotide. First-round primers, ED3 and Nef3, were used with the following cycling conditions: 94°C for 2 min; 3 cycles of 94°C for 30 s, 55°C for 30 s, and 72°C for 3 min; 36 cycles of 94°C for 30 s, 58°C for 30 s, and 72°C for 3 min; and a final incubation at 72°C for 10 min. The second-round PCR included a 1:20 dilution of the first-round products in fresh reaction buffer plus primers DR7 and DR8 with the following cycling conditions: 94°C for 2 min; 36 cycles at 94°C for 30 s, 55°C for 30 s, and 72°C for 1 min; and a final incubation at 72°C for 10 min. Plasma samples were processed for RNA extraction and cDNA synthesis as previously described (45). cDNA was amplified first with primers ED12 and ED31, followed by a second round as for the PBMC samples.

Sequence analysis. Sequences bearing open reading frames (~90% of all sequences determined) were first aligned with the Pileup program in the GCG suite (Genetics Computer Group, Madison, Wis.) and then manually edited. Pairwise evolutionary nucleotide distances (excluding gaps in the pairwise alignment) were estimated with the Kimura two-parameter model of evolution (35) with a transition/transversion ratio of 2 as implemented in MEGA (40) or PHYLIP, version 3.5 (20). DNA distances were also estimated with a general time-reversible model with site-to-site variation in substitution rates (discrete approximation of a  distribution with a shape parameter,  = 0.5 with four bins [77]). Identical trends were seen with either method, although as expected, the Kimura two-parameter method tended to underestimate DNA distances at higher divergence levels. This comparison is available online (71a).

Analysis of viral divergence and diversity. Both viral divergence from the founder strain and viral diversity were estimated at each time point. To estimate viral diversity at a given time point, we determined the mean and standard deviation for pairwise nucleic acid distances between all sequences obtained at that time point. To estimate viral divergence at a given time point, a founder sequence was approximated as the sequences found at the initial virus-positive time point. Nearly identical values for viral divergence were observed when these distances were estimated with a single consensus sequence from the first time point. The mean and standard deviation for all pairwise comparisons between sequences from the first positive time point and subsequent time points were then calculated. The rate of substitution was measured by estimating the expected number of substitutions that accumulated along the branch from an ancestral sequence (77) as described below.

(1)

(2)

Coreceptor usage and SI phenotype prediction and determination. The occurrence of viruses utilizing the CCR5 and CXCR4 coreceptors was determined by one or more of three different assays. (i) Virus isolates derived from fresh or frozen viable PBMCs were tested for syncytium induction on MT-2 cells (31). (ii) The phenotypic use of coreceptors was determined by the growth of viral isolates cultured with cells expressing CD4 and coreceptors (cells graciously provided by D. Littman, Skirball Institute of Biomolecular Medicine, New York, N.Y.) (65). (iii) The X4 phenotype was predicted from viral sequence data, based on deduced basic amino acid substitutions at positions 306, 319, and 320 of the mature gp120 protein (3, 4, 7, 13, 22, 36, 64, 70, 78). None of these methods distinguished between X4 viruses and dualtropic R5/X4 viruses. For simplicity, viruses predicted (via sequence analysis) or observed (via phenotypic culture) to use the CXCR4 coreceptor are referred to as X4.

Nucleotide sequence accession numbers. All sequences described were deposited in the GenBank database under accession no. AF137629 to AF138163, AF138166 to AF138263, and AF138305 to AF138703.

	RESULTS

Top Abstract Introduction Materials and Methods Results Discussion References

Consistent trends in viral evolution. A total of 1,300 sequences bearing an open reading frame were obtained from the nine men at a total of 106 time points, for an average of 12.3 sequences from each time point. Paired PBMC and plasma samples were examined at 25 time points. Figure 1 shows sequence comparisons for two participants, illustrating three features of env gene evolution that were observed among all participants (for complete data sets, see reference 71a). First, the changes in viral divergence (from an approximated founder strain; see Materials and Methods) were essentially the same in both cell-associated viral DNA (Fig. 1a) and plasma RNA (Fig. 1b). The patterns of increase in viral diversity (the breadth of population at a given time point) were also similar in the viral DNA and RNA populations (Fig. 1c and d), although at later times they appeared to diverge in participant 2, perhaps due to the unavailability of RNA sequences from the two populations at all time points. Thus, divergence and diversity measurements were not affected by the different half-lives of cells versus virions, in agreement with another recent report (2). For simplicity, therefore, we combined DNA and RNA data for the analyses shown in subsequent figures. Second, divergence increased linearly for several years after seroconversion but then appeared to slow or stabilize late in infection, again at approximately the same times in DNA and RNA (Fig. 1a, b, and f). Third, the breadth of viral population diversity at a given time point increased in parallel with divergence for a few years after seroconversion before reaching a peak and then leveling off or decreasing prior to the point of divergence stabilization. Figures 1a to d, f, and g also illustrate the variability in pairwise comparisons of divergence and diversity around mean values. The data were distributed approximately symmetrically around the means with some evidence of a multimodal distribution in diversity measures at later time points. The evolutionary relationships between all of these sequences showed a significantly time-ordered, progressive evolutionary trend away from the founder strain (P < 0.001) (Fig. 1e and h).

View larger version (64K): [in this window] [in a new window]   FIG. 1.   Viral population analyses for participants 2 and 9. Analyses of sequences from participant 2 (a to e) and participant 9 (f to h) are shown. Genetic distances between all possible pairs of nucleotide sequences for viral DNA in PBMC (a, c, f, and g) and viral RNA in plasma populations (b and d) are shown (means at each time point indicated by symbols connected by thick lines) as a function of time since seroconversion. Parallel estimates with sequences sampled from plasma are shown with a thin blue line (a and c) and those from PBMC as a thin red line (b and d). Divergence from the founder strain (i.e., the first virus-positive time point; see Materials and Methods) is shown (a, b, and f). Viral population diversity (i.e., distances between all pairs of sequences within each time point; see Materials and Methods) is shown (c, d, and g). Dotted vertical lines correspond to a peak or a slowdown in the rate of change of each measure. Neighbor-joining phylograms (e and h) are shown for participants 2 and 9, respectively, derived from maximum-likelihood distances between all the sequences in each patient with PHYLIP (20). Sequences are represented by a square for PBMC sequences or a triangle for plasma sequences, with an arbitrary color gradient corresponding to the time of sampling following seroconversion, as indicated in the key. An asterisk (*) indicates sequences with a basic amino acid substitution in the V3 loop specifying the X4 genotype.

View larger version (28K): [in this window] [in a new window]   FIG. 2.   Summary of sequences, viral phenotypes, and T-cell measures. Participant identifiers above each panel are from a previous study (65); participants 4 and 10 were excluded from the current study due to a lack of available specimens. The horizontal axis of each panel indicates the time relative to seroconversion. (a) Genetic distances of the combined population of viral DNA and RNA sequences relative to the founder strain (divergence, thin lines) and within individual time points (diversity, thick lines). The left (dotted) vertical line for each participant (and the single dotted line for participants 5 and 11) indicates the time of peak viral diversity, at which a significant stabilization or decrease in the slope of population diversity growth was observed as defined in the text. The final time point from participant 7 had an increase in overall viral diversity due to the appearance of unique divergent virus populations in the plasma versus the those in the PBMCs, although these populations taken individually had much lower levels of diversity (see appendix at reference 71a). The right (dashed) vertical line for each participant (except participants 5 and 11) indicates the time at which the divergence from the founder strain began to slow down or stabilize (divergence stabilization). (b) Viral genotype analysis. Each panel shows the proportions of deduced amino acid sequences containing mutations predictive of the X4 phenotype, based on encoding a basic amino acid (lysine or arginine, indicated by K/R in the figure) at residues 306, 319, or 320 within the envelope glycoprotein. For participant 7, mutations at both position 319 and position 320 (open symbol) were found at one time point. (c) Viral phenotype and coreceptor usage analysis. Data taken from Rinaldo et al. (65). The squares indicate CXCR4 coreceptor usage, and the diamonds indicate CCR5 coreceptor usage by virus isolates derived from the indicated time points. Open symbols indicate no growth, and filled symbols indicate growth on cells expressing the specified coreceptor plus CD4 (see Materials and Methods). The filled circles indicate syncytium formation (SI phenotype), and the open circles indicate a lack of syncytium formation (NSI phenotype) when the virus isolates were added to MT-2 cells. (d) Clinical progression. Data was taken from Rinaldo et al. (65). CD4+ T-cell levels are shown with a dotted line. Patient 2 had exceptionally high CD4+ T-cell levels early in infection which are plotted with a different scale as indicated in the panel. CD3+ T-cell numbers are shown with the thick line, whereas plasma RNA levels are shown with the filled circles connected by a thin line. Antiretroviral treatment, AIDS diagnosis by development of opportunistic infections (AIDS), and survival time are also indicated. Participants' visits, at which time the antiretroviral drugs were prescribed, are indicated (ZDV, zidovudine; d4T, stavudine; 3TC, lamivudine; ddI, didanosine; SQV, saquinavir). Five participants died after the period of analysis shown, and this time (in years) is indicated in parentheses below the dagger ().

View larger version (22K): [in this window] [in a new window]   FIG. 3.   Trends of viral diversification across individuals. Data from the nine participants shown in Fig. 2 are replotted to show the similarities of viral population intra-time-point diversity (a and b) and divergence from founder strains (c and d) across individuals. Participant 1 is shown with a black line, 2 with red, 3 with orange, 5 with gray, 6 with green, 7 with cyan, 8 with blue, 9 with magenta, and 11 with purple. (a and b) Viral population intra-time-point diversity plotted relative to the time of seroconversion (a) or the time of peak or stabilization of diversity (vertical dotted lines) (b). (c) For each participant, all of the sequences from the first virus-positive visit (the founder strain) were compared to all sequences sampled at each subsequent time point. Similar slopes were also generated when mean values at each time point were used to calculate a combined slope (data not shown). (d) Viral population divergence from the founder strain as shown in panel c but plotted relative to the estimated time of divergence stabilization (vertical dotted line). Participants 5 and 11, who had no evident stabilization, are omitted. (e) Relationship between the times from seroconversion to peak viral diversity and to divergence stabilization.

Association of viral evolutionary change and X4 viruses. Mutations predictive of the X4 viral phenotype (basic amino acid substitutions at positions 306, 319, and 320 of the mature gp120 protein) were detected at one or more time points in all nine individuals (Fig. 2b). Individual participant data, plotted in Fig. 2b, demonstrated a pattern in which X4 viruses were initially absent, rose to constitute >80% of the viral population in six cases, and then declined. To determine if these patterns of X4 virus prevalence were related to times of seroconversion, peak diversity, and divergence stabilization, data from all individuals were plotted relative to those events. Relative to seroconversion (Fig. 4a), the times of X4 emergence were spread out over the course of infection. However, X4 emergence appeared to be closely associated with, and usually just prior to, the time of peak diversity (Fig. 4b) (mean = 0.32 years; 95% CI, 1.26 to 0.63). Moreover, the peak prevalence of X4 viruses appeared to be closely associated with and just prior to the time of divergence stabilization (Fig. 4c) (mean = 0.35 years; 95% CI, 1.26 to 0.56). The strength of these associations is quantified by the correlations between the time of emergence of X4 viruses and the time to peak diversity (Fig. 4g) (r² = 0.67) and the time of peak X4 with the time to divergence stabilization (Fig. 4h) (r² = 0.79).

View larger version (38K): [in this window] [in a new window]   FIG. 4.   Trends of X4 genotype representation and CD4+ T-cell numbers across individuals. Data from all nine participants shown in Fig. 2 are replotted with the proportion of the virus population with an X4 genotype (a to c) or levels of circulating CD4+ T cells (d to f), relative to the time of seroconversion (a and d), time of peak viral population diversity (b and e), and time of divergence stabilization (c and f). Participant color-coding is the same as described in the legend to Fig. 3. The dotted vertical lines indicate the times of peak diversity (b and e) or divergence stabilization (c and f). The horizontal line across the lower panels indicates a CD4+ lymphocyte level of 200 cells/µl. (g to i) Relationships between the times from seroconversion to the times of the indicated viral population events. Slope = 0.94 ± 0.25 (g), 0.96 ± 0.22 (h), and 0.91 ± 0.19 (i).

Relationship of disease progression with viral evolutionary changes and X4 viruses. Across all participants, consistent relationships were noted between times of the viral events described earlier (i.e., peak diversity, divergence stabilization, and emergence and peak representation of X4 viruses) and changes in T-cell populations. The participants experienced different rates of decline of CD4⁺ T cells over time (Fig. 4d). However, seven of nine subjects (all except participants 9 and 11) transitioned to 200 CD4⁺ T cells/µl within 2 years following the point of divergence stabilization, with these two events highly associated in time (r² = 0.83; mean interval = 0.97 years; 95% CI, 0.18 to 1.76) (Fig. 4i). The CD4⁺ T cells in subject 9 dropped to 202 cells/µl 1 year following the point of divergence stabilization (included in Fig. 4i), at which time he began highly active antiretroviral therapy (HAART). Subject 11 experienced the most protracted decline in CD4⁺ T cells and did not reach a level of 200 CD4⁺ T cells or a point of divergence stabilization as of the last time point studied.

View larger version (20K): [in this window] [in a new window]   FIG. 5.   Relationships between times from seroconversion to CD3+ T-cell inflection and times to viral evolutionary changes and CD4+ T-cell numbers. The time of the CD3+ cell inflection point is plotted relative to the time of peak viral diversity, slope = 0.71 ± 0.22 (a); divergence stabilization, slope = 0.70 ± 0.17 (b); peak representation of X4 viral genotypes, slope = 0.51 ± 0.28 (C); and transition of the CD4+ T-cell levels to 200/µl, slope = 0.72 ± 0.14 (d).

Evolutionary patterns in late progressors. Two participants initially selected as nonprogressors experienced evidence of disease progression with further follow-up. One of those (participant 9) was studied in more detail after this progression occurred. After a period of 8 years of stable CD3⁺ and CD4⁺ T-cell counts and a 4-year period of increasing viral load, he experienced a rapid decline in CD4⁺ T cells to 202 cells/µl at the last time point available before receiving HAART. Preceding the decrease in CD4⁺ T cells, a peak in viral diversity and an outgrowth of X4 viruses were observed. Participant 11 had stable T-cell counts and low viral loads for 7 years and remained asymptomatic through 1998 without drug therapy. However, attenuation in viral diversity and a minor outgrowth of X4 viruses were noted at 6 years after infection. These were followed by a decline in CD4⁺ and CD3⁺ T cells, indicating that he was experiencing disease progression.

	DISCUSSION

Top Abstract Introduction Materials and Methods Results Discussion References

Our results suggest that in the individuals studied, the time between primary HIV-1 infection and AIDS could be subdivided into three phases based on patterns of viral evolution over the C2-V5 region of env. The phases were defined by changes in viral divergence from the founder strain and intra-time point diversity; these definitions in turn were reinforced by observed changes in the prevalence of X4 viruses (probably including dualtropic X4/R5 viruses). In the first phase, both viral population divergence and diversity increased linearly, with diversification essentially keeping pace with divergence at a rate of about 1% per year. The beginning of the second phase was signaled by a leveling off, or decrease, in diversity. In this phase, the viral population continued to diverge from the founder strain at the same rate, while diversity continued to plateau or constrict. Finally, the increase in divergence also slowed or stabilized, marking the beginning of the third phase, the latter being characterized by a decline in diversity.

The peaking of diversity was strongly supported by statistical analysis. Although similar support for the point of divergence stabilization was not found, the validity of this concept, as well as the general framework put forward here, is strongly supported by the consistency of the patterns described among participants and by the close linkages observed between the three defined phases and the patterns of X4 virus prevalence and T-cell subset landmarks. Thus, the second phase was strongly associated with X4 viruses, which in most participants first appeared very close to the beginning of this phase and peaked in prevalence very close to the end of it. The third phase in turn was closely associated in time with the decline of CD4⁺ T cells to <200/µl and with the failure of T-cell homeostasis and a loss of phenotypically naive (CD45RA⁺RO) CD4⁺ T cells (48a). The proposed framework is also generally consistent with previous sporadic reports of constriction of HIV diversity (15, 51, 58) and similar rates of divergence from the founder strain (39, 43, 45, 46, 82-84). Here, we have extended these previous reports by systematically documenting the consistency of these patterns and rates.

Other arguments support the idea that the proposed framework, though derived from a small number of participants, may have broad applicability. (i) Participants were studied in substantial depth from seroconversion to advanced disease in those developing disease progression. (ii) The findings were quite consistent. (iii) Participants were representative of most HIV-1-infected people in terms of the rate of disease progression (median survival time = 9.1 years after seroconversion compared to 10.2 years among the MACS participants as a whole [56]) and occurrence of T-cell homeostasis failure before AIDS (seen in ~75% of MACS participants who develop AIDS [24, 34, 48]). (iv) The patterns were seen even in two participants who were initially selected for lack of progression of HIV disease but exhibited evidence of progression as follow-up advanced.

Taken together, a summary sequence of events can be inferred (Fig. 6) that describes the virological and T-cell changes that occur in the asymptomatic period of HIV infection. It should be stressed that these phases constitute general trends found in the six individuals we studied who developed AIDS during the period of analysis, although all nine experienced the initial events we described. The sequence begins with the detection of X4 viruses and the attainment of peak viral diversity at a variable time after seroconversion (4 to 9 years in these participants). If the time of X4 virus appearance is taken as time zero, then peak diversity occurs an average of 0.3 years later, followed by X4 peak representation at 1.5 years, divergence stabilization at 2.2 years, T-cell homeostasis failure at 2.5 years, and CD4⁺ cell transition through 200 cells/µl at 2.9 years. Based on a study of 212 seroconverters and 1,129 seroprevalant infections within the MACS cohort, the occurrence of clinical AIDS-defining conditions follows T-cell homeostasis failure by 1.7 years (24, 48), or approximately 4.2 years from the emergence of X4 viruses.

View larger version (59K): [in this window] [in a new window]   FIG. 6.   Schematic illustration of proposed consistent patterns in development of HIV disease in moderate progressors. (a) Clinical phases of HIV infection as well as typical patterns of CD4+ and CD3+ T cells and plasma viral RNA loads. The initial transient increase in viral RNA is estimated from the literature, whereas the later RNA increase and the CD3+ and CD4+ T-cell declines are based on the data from this study and that of Rinaldo et al. (65). (b) Patterns of viral sequence evolution within the asymptomatic period of infection identified in this study. Circle diameters represent the mean viral population diversities at increasing intervals following seroconversion. Vertical displacement of the circles represents the extent of viral population divergence from the founder strain. Shading represents the proportion of the viral population comprised of viruses with an X4 genotype. Dotted vertical lines represent (from left) the end of acute infection, peak viral diversity, stabilization of divergence from the founder strain, and the development of AIDS. (c) Characteristic changes in viral evolution in the three periods of the asymptomatic phase identified in this study (, increasing; , decreasing; , stable).

Several caveats should be noted. It is not clear whether the patterns and sequence of events observed can be generalized to people with other rates (i.e., very rapid or very slow) of disease progression or over shorter observation periods (23, 45, 46, 67, 76). Seven of the participants we studied were typical progressors, whereas two (participants 9 and 11), were initially chosen as examples of nonprogressors but subsequently did progress. Second, the length of the intervals between the noted events is variable within the individuals we studied; the durations estimated here are not meant to set limits and are mentioned only as a frame of reference for the evaluation of additional HIV-infected individuals. Third, we drew our inferences about viral evolutionary patterns on only 650 bp of env; other genes and gene regions evolve more slowly and may not exhibit the same patterns.

Despite these concerns, recognition of the aforementioned patterns of HIV-1 evolution suggests a way to reconcile conflicting reports about the relationship between the rate of HIV-1 disease progression and the degree and rate of increase in viral diversity (15, 16, 23, 26, 27, 44-46, 49, 51, 52, 60, 73, 82, 83). Because only extensive sampling can detect the complex pattern of viral evolution we describe, viral diversification could appear spuriously slow if viral populations are assessed infrequently, e.g., only early and late in infection (thereby missing the point of peak diversity) or in people with highly atypical disease pathogenesis (e.g., the slowest and the most rapid progressors). A previous study of env sequences by Shankarappa et al. (71) suggested the existence of a diversity peak with continuing divergence from the founder strain, although sampling was less frequent and potentially compromised by resampling (44).

In this study, X4 viruses were detected in the blood of all nine participants and grew to predominance in six. This represents a significantly higher frequency of detection of X4 viruses than in approximately 50% of the individuals observed in previous studies of people progressing to AIDS (36, 70, 78). Previous studies have used phenotypic assays to detect X4 viruses, as opposed to the genotypic analysis done here. Both methodologies can miss detection of X4 viruses on occasion. However, we also noted that the representation of X4 viruses peaked and then diminished over time, which could lead to falsely low estimates of the proportion of people who have circulating X4 viruses if they were sampled subsequent to such a decline. Our higher detection rate by sequence analysis could also reflect the fact that PBMC DNA harbors defective and latent proviruses for extended periods (12, 61, 63, 69) and thus may provide a record of past transient replication of X4 viruses. Evidence for the transience of X4 viruses has also been seen in previous studies (30, 51). We have also detected transient representation of X4 viruses as well as a similar pattern of viral evolutionary dynamics in the two other patients we have studied throughout infection: a gay man with rapidly progressing disease (43, 45) and a perinatally infected child (40a). Overall, our results support the view that X4 viruses are likely to occur in significantly more than 50% of people who develop AIDS. It should be noted that X4 viruses have often been linked to AIDS in HIV-1 subtype A, B, D, and E infections (18, 37, 59), have been found with significant frequency in early HIV-1 subtype C infections (79), but are seen at low frequency in cases of late-stage subtype C infections (59). Thus, it has not been established that our findings apply to infections with all HIV-1 subtypes.

Interestingly, the timing of evolution and rapidity of outgrowth of X4 viruses relative to the point of peak viral diversity appeared to differ by mutation. X4 viruses with a mutation at position 306 (participants 5 and 8) increased rapidly, reaching a peak level shortly after initial detection. In contrast, viruses with a mutation at position 320 (participants 1, 2, and 3) reached peak levels more gradually, and the appearance of the position 319 mutation was most variable and tended to be found at low levels (participants 3, 6, 7, and 8 but not 9) and in individuals with other X4-specific mutations. Thus, although these data are preliminary, the specific X4 mutation may be important to viral growth properties and in disease progression.

These observations suggest an important role for X4 viruses in T-cell destruction late in the progression of HIV-1 infection in the participants studied. This is plausible in view of the preferential expression of CXCR4 on naive T cells (9) and on thymocytes (50), which are needed to replenish memory T cells (6). Among other recent findings suggesting a differential impact of R5 and X4 viruses on T-cell populations (5, 8, 29, 33, 75, 81) is the mediation of apoptosis of CD8⁺ cells by X4 Env (29), a mechanism that may apply to CXCR4-expressing naive CD4⁺ cells as well.

Our data also suggest that the degree and rapidity of immune reconstitution (1) that occurs under HAART may differ according to the phase at which HAART is initiated, with greater reconstitution possible prior to some of the milestones we described. For example, it may be that people treated with HAART before X4 viruses are detected will be able to restore lost immunological functioning more effectively than those whose treatment is begun later. In addition, given the serious long-term side effects of HAART (10, 28), the ability to predict AIDS onset ~4 years in advance based on detection of X4 viruses (57) may help refine therapeutic strategies.