Clinical significance of HIV-1 coreceptor usage

Despite an extensive literature on the subject, the process of in vivo transmission of HIV-1 remains largely unknown and most of the current models are essentially conjectural. In fact, there are virtually insurmountable difficulties in studying the earliest events of HIV-1 infection in the human species, while nonhuman primate models, albeit useful, provide only a partial and, in all likelihood, non-physiological picture. In fact, most of the observations on the acute phase of HIV-1 infection in humans are made several weeks or even months after the initial transmission event, typically in subjects who manifest clinical signs of acute retroviral syndrome. With this caveat in mind, the bulk of evidence indicates that R5 HIV-1 variants are largely prevalent during the acute phase of infection [67, 68]. Even if both R5 and CXCR4-using variants are present in the donor, most often only the R5 variants are detected in the recipient [69‐71]. Whether this early R5 predominance reflects a bona fide transmission bias or a superior in vivo fitness of R5 strains during the early phase of infection remains uncertain.

Several studies have attempted to correlate the predominance of R5 HIV-1 strains during the acute phase with a biological bottleneck inherent to the genital mucosa, variously related to trapping and inactivation of CXCR4-using virus by mucin and innate antiviral proteins, preferential transcytosis of R5 viruses in endothelial cells and/or preferential amplification of R5 viruses by resident macrophages, dendritic cells and/or Langerhans cells (reviewed in [72]. In reality, however, no conclusive evidence has been provided to indicate that CXCR4-using strains are less able or unable to sustain mucosal transmission. For example simian/human chimeric immunodeficiency viruses (SHIV) bearing an X4 HIV-1 envelope can be readily transmitted via the mucosal route in macaques, and have widely been used as a reference model [73].

Another important element that is rarely taken into consideration in the HIV-1 transmission equation is the transmitter bias, due to the fact that people with replicating CXCR4- using viruses may be in a more advanced stage of their disease and less prone to engage in risky sexual behavior [74]. If a majority of transmissions occur from asymptomatic individuals who at that time still only harbor R5 variants, then this would indeed contribute to the scarcity of transmission of CXCR4-using variants.

Another potential contributing factor in the marked R5-variant predominance during acute primary HIV-1 infection is a post-transmission amplification bias. In peripheral blood, the high proportion of CCR5⁺CD4⁺ T cells that are recruited during acute HIV- 1 infection may favor R5 variants to replicate and thereby outcompete putative co-transmitted CXCR4-using variants. Moreover, the high proportion of memory/activated CCR5⁺CD4⁺ T cells present in the mucosal-associated lymphoid tissue, which is considered a major site of HIV-1 replication during primary infection [75], may provide the optimal environment for preferential R5 HIV-1 amplification. Of importance, these cells also express high levels of integrin α4β7, an important facilitator of HIV-1 infection [76].

Although primary infection with CXCR4-using HIV-1 strains is believed to be a rare event, mixed R5/X4 primary infections have been clearly documented in a few patients studied longitudinally from an early stage post-infection [77, 78]; interestingly, however, the CXCR4- using component was selectively cleared from plasma with the transition to the chronic phase in which only R5 variants could be recovered, raising the possibility that mixed R5/X4 transmission may in fact be more frequent than it appears, albeit underestimated due to late sampling, subsequent to the disappearance of the CXCR4-using component from plasma. In a recent retrospective evaluation of a large number of patients (n = 390) enrolled in the PRIMO cohort in France between 1996 and 2007, a relatively high prevalence (15.9%) of predicted CXCR4-using viruses was documented during primary infection [79]. However, subsequent phenotypic analysis of the patients in this study infected with non-subtype B variants (n=131) showed that genotypic predictions overestimated the proportion of CXCR4-using variants in non-subtype B infected patients, resulting in a much lower prevalence (0.8%) of actual CXCR4-using variants [80]. In agreement, a low prevalence of CXCR4-using viruses (4-6%) was observed in three recent cohorts of seroconverters: one from the United States (n=150) during the period 1999-2003 [81], one from France (n=133) during the period 1995-2008 [82], and one from the Netherlands (n=46) during the period 2003-2008 (ABW et al., unpublished observations). Moreover, detailed investigations of individuals experiencing primary infection and sampled prior to seroconversion (Fiebig stage I-V) using single genome amplification have revealed a consistent pattern of CCR5 dependence at this stage [67, 68, 83‐88] indicating that any putative post-transmission amplification bias would have to occur within the first few days of exposure.

The most convincing argument in favor of the bona fide predominance of R5 variants in the initial transmission events is the high degree of protection from HIV-1 infection conferred by genetic deficiency of CCR5. Subjects that are homozygotes for the CCR5-Δ32 allele are highly protected from HIV-1 infection [89]. Indeed, this genotype is enriched among individuals who remain seronegative despite high-risk sexual behavior [90, 91] as well among hemophiliacs who had remained HIV negative despite exposure to batches of clotting factor that were known to be the cause of HIV infection in other hemophiliacs [92]. CCR5wt/Δ32 heterozygosity does not confer protection from HIV-1 infection, but this genotype has been associated with a significantly lower viral set point and delayed disease progression [93‐96]. A second crippling polymorphism, m303, that introduces a premature stop codon in the CCR5 gene has been identified in exposed-uninfected subjects [97].

Despite the strong protective effect conferred by congenital CCR5 deficiencies, a handful of infected CCR5-Δ32 homozygotes have been reported, all invariably harboring CXCR4- dependent HIV-1 strains [98]. Interestingly, two CCR5-Δ32 ^+/+ individuals were found to harbor an inherently (not mixed) dual-tropic virus (R5X4) that remained stable over time [99], suggesting that maintenance of the CCR5-using envelope conformation might provide a selective advantage in vivo despite the absence of a usable receptor. Although the viral load in infected CCR5-Δ32 ^+/+ subjects tends to be low, a rapid depletion of circulating CD4⁺ T cells has been noted [98]. However, the very limited number of individuals characterized thus far makes it difficult to evaluate whether the clinical and virological course of an X4 HIV-1 infection d'emblée differs from that of conventional R5 HIV-1 infection in individuals with wild-type CCR5 genes. In this respect, viral variants with X4/DM phenotype have been detected in individual case reports of acute HIV-1 infections with peculiarly severe clinical manifestations [100‐103], but again their number is too limited to draw any reliable conclusions and it is difficult to exclude other confounding cofactors.

Usage of the so-called “minor” coreceptors during primary HIV-1 infection has only marginally been investigated. Interestingly, a high frequency of CCR3 usage during primary infection has been detected by direct cloning and expression of viral sequences, while in vitro culture apparently selects against CCR3 usage leading to the expansion of variants with exclusive CCR5 usage [104]. By contrast, in a few cases of primary infection with subtype- C HIV-1 studied, all of R5 phenotype, there was a surprisingly high frequency of usage of GPR15, APJ and CXCR6, but not of CCR3 [105, 106]. Further studies are warranted to address the relevance of minor coreceptors in HIV-1 transmission.

During the asymptomatic phase of HIV-1 infection a homogeneous R5 virus population is commonly present that generally has the ability to replicate efficiently in both T cells and macrophages [70, 71]. In many patients, a typical pattern of viral evolution has been documented during the course of the infection with the emergence, usually in concomitance with the earliest signs of disease progression, of CXCR4-using variants. However, this pattern is not consistently observed in all patients progressing to AIDS (Figure2). For example, in patients infected with subtype-B HIV-1, variants that use CXCR4 can be isolated from approximately half of the patients who have developed AIDS [15, 24, 33, 106‐109]. Of note, such variants may first appear during the asymptomatic phase of infection, before AIDS is diagnosed [106], albeit after an initial decline of CD4⁺ T cells [110]. Patients progressing to AIDS often harbor viral populations that can use multiple coreceptors including CCR5, CXCR4 and one or more minor coreceptors [15]. Whether a promiscuous coreceptor usage provides a selective advantage for HIV remains uncertain since most of the minor coreceptors show a low and/or tissue-specific expression pattern.

Until recently, the general consensus was that CXCR4-using variants only emerge in a certain proportion of HIV-1 infected individuals, which varies according to the viral genetic subtype. However, recent evidence has challenged this view suggesting that the prevalence of CXCR4 usage is continuously evolving over time. Data from the Amsterdam Cohort Studies on HIV infection and AIDS (ACS) show a continuously ongoing X4 conversion rate at the population level, even after AIDS diagnosis (H.S. et al., unpublished results). Thus, it seems that the emergence of CXCR4 usage is a matter of time and that some HIV-1 infected individuals die before they develop CXCR4-using variants. Molecular cloning of the viral quasispecies has documented the frequent presence of DNA sequences predictive of CXCR4 usage in blood cells even in the absence of detectable replication of such variants, as well as transient appearances of CXCR4-using strains on a background of sustained R5 persistence [50, 111, 112]. These observations further challenge the simplistic concept of a single and irreversible coreceptor- switching event, depicting a complex dynamic state where variant predominance is a continuously evolving process governed by multiple interactive factors. After the appearance of CXCR4-using variants, R5 variants remain present in the vast majority of patients [108, 113]. Pure X4 virus populations are rather infrequently detected and often restricted to late-stage disease [114], although they have been documented earlier in rare cases of ab initio X4 transmission [70, 115, 116] often associated with the absence of CCR5 expression in the host [98].

Several lines of evidence suggest that CXCR4-using variants evolve in vivo from pre-existing R5 variants. Phylogenetic analysis of longitudinally obtained HIV-1 variants from individual patients has shown that although R5 and X4 variants from a single patient form separate monophyletic clusters in an unrooted tree, they do cluster together when an appropriate outgroup is used. Although this implies that R5 and CXCR4-using variants from a single individual are more related to each other than to HIV-1 variants from another person, it does not exclude that both variants may be initially transmitted together and then differentially expressed in vivo, with CXCR4-using variants maintained in a cryptic condition until immune deterioration occurs. However, the first detectable CXCR4-using variants are more closely related to the R5 variants obtained at the same time than co-existing R5 and CXCR4-using variants derived at later time points. Moreover, recently emerged CXCR4-using variants generally maintain the ability to use CCR5 and in phylogenetic trees these dual-tropic R5X4 populations cluster between pure R5 and later-stage CXCR4-using virus populations derived from the same individual. These observations are compatible with the hypothesis that the first CXCR4-using variants evolve from pre-existing R5 variants within the same individual, rather than remaining silent for many years after an initial independent transmission.

The emergence of CXCR4-using HIV-1 variants in a patient is almost invariably associated with a subsequent increase in the rate of decline of circulating CD4⁺ T cells, an accelerated disease progression, and a poor prognosis for survival [106] (Figure3). Indeed, studies in the era preceding the introduction of combination antiretroviral therapy (cART) demonstrated a significant acceleration in the rate of CD4⁺ T-cell decline after the first detection of CXCR4-using variants [106, 107]. However, the presence of CXCR4-using variants is not an obligatory prerequisite for disease progression and a significant proportion of individuals progress to AIDS and AIDS-related death while harboring exclusively R5 HIV-1 variants. Notably, the time from seroconversion to AIDS is not significantly different between individuals who harbor only R5 variants as compared to individuals with detectable CXCR4-using variants (Figure4).

The accelerated CD4⁺ T-cell decline after the first appearance of CXCR4-using HIV-1 variants seems to be determined by the greater proportion of CD4⁺ T cells that express CXCR4, which provides a larger target-cell pool for the virus. Indeed, up to 90% of CD4⁺ T cells express CXCR4 whereas CCR5 expression is restricted to a smaller subset (15-35%) that express a memory/activated phenotype [117‐119]. Specifically, CXCR4, but not CCR5, is expressed on naïve CD4⁺ T cells [118], a large population of non-antigen-primed T cells. Thus, naïve CD4⁺ T cells are exclusively a target for CXCR4-using variants [120]. The pathological consequences of infection of naïve CD4⁺ T cells may be dramatic, resulting in the depletion of the largest pool of CD4⁺ T cells in the body. Considering the potential beneficial effect of an expanded target-cell population, it is puzzling that CXCR4-using variants emerge in only a proportion of infected individuals [106] even though a limited number of amino acid substitutions in V3 is sufficient to confer CXCR4-using capability to R5 variants [38, 121].

It has been hypothesized that the absence of CXCR4-using variants early in HIV-1 infection may be due to their higher vulnerability to the host adaptive immune responses, in particular neutralizing antibodies. The fact that these variants appear more frequently after the initial decline of CD4⁺ T cells indeed suggests that they may represent a peculiar, congenic form of opportunistic infection. In a recent study, we found that the first appearing CXCR4-using variants are more sensitive to neutralizing antibodies directed against the CD4 binding site than their co-existing R5 variants [122]. A greater sensitivity to both monoclonal antibodies and sera from infected patients was also documented in sequential isolates derived from a cohort of perinatally-infected children (P.L. et al., unpublished data). Further evidence for a greater sensitivity of CXCR4-using variants to antibody-mediated neutralization has come from the characterization of a conserved neutralization epitope within the V3 domain of gp120: such epitope, designated D19, is invariably cryptic in R5 variants of different genetic subtypes, but it is consistently exposed in CXCR4-using variants, rendering such variants sensitive to neutralization by a specific antibody [123]. A role of cell-mediated immune responses was instead suggested by the re-emergence of CXCR4-using strains in dually-infected (R5+X4 SHIV) macaques after in vivo depletion of CD8⁺ T cells [124]. Additional data pointing to a lower in vivo fitness of CXCR4-using variants have come from the results of clinical trials with the CCR5- antagonist maraviroc (MVC). Resistance to MVC was shown to develop by two mechanisms: a reduced drug susceptibility associated with changes in the V3 loop that allow the R5 virus to use CCR5 in its MVC-bound conformation [125] or the emergence of variants that use CXCR4 [58]. In the latter scenario, the CXCR4-using virus seems to originate from an unrecognized pretreatment reservoir, indicating that screening assay sensitivity remains to be improved. Interestingly, however, the circulating virus was shown to revert to the R5 phenotype following cessation of MVC, indicating that the selective pressure acting against CXCR4 usage was preserved.

Altogether, the above observations suggest that effective host immune responses may exert a selective pressure that hinders the emergence of CXCR4-using variants. However, when the host immune competence begins to fade during the progression of HIV-1 disease, such pressure would start to wane, paving the way for the emergence of CXCR4-using variants. Although the host immune surveillance may explain, at least in part, the absence of CXCR4- using variants during the asymptomatic phase of HIV-1 infection, it cannot justify the apparent scarcity of CXCR4-using variants in the very early phase of infection when neutralizing antibodies and cytotoxic T lymphocytes are still absent. As discussed above, this may result from a lack of transmission or to the inability of recently transmitted CXCR4-using virus to compete with co-existing R5 HIV-1 during the earliest phase of infection.

As stated above, a significant fraction of patients progresses to full-blown AIDS without experiencing an overt switch to CXCR4 usage. However, accumulating evidence indicates that in spite of their ''monogamous'' CCR5 use late isolates from these patients are inherently more pathogenic [126] and RANTES-resistant than early isolates [127, 128]. In line with these observation is the ability of late-stage CCR5-restricted HIV-1 variants to use chimeric coreceptors in which parts of CCR5 have been replaced with segments of CXCR4 (R5 broad), whilst early CCR5-using HIV-1 variants are restricted to the use of wild-type CCR5 (R5 narrow) [127, 129, 130].

This in vivo evolution of CCR5-restricted HIV-1 in humans is similar to that observed in nonhuman primates infected with SIV, which never acquires CXCR4 usage even though its pathogenicity increases during the late disease stages [131]. Moreover, we have observed that the ability of R5 isolates to replicate in macrophages is progressively reduced during the course of infection, resulting in a predominantly T-cell tropic R5 HIV-1 quasispecies even before the progression to AIDS [108]. In about half of subtype-B HIV-1 infected individuals, this shift towards full T-cell tropism precedes the emergence of CXCR4 using HIV-1 variants.

Evidence suggests that the evolutionary changes in the V3 loop involved in the coreceptor-usage switch are gradual and accretive, and that dual coreceptor usage (R5X4) represents an intermediate transitional phase. R5X4 viruses can be more efficient in using either CCR5 (dual-R) or CXCR4 (dual-X), or can use both coreceptors with similar efficiency [5‐7]. As mentioned above, there is a continued evolution in viral coreceptor usage in vivo, resulting in a broad range of coreceptor affinities within the HIV-1 quasispecies. The limited number of amino acid substitutions required to confer CXCR4-using capability to R5 variants in vitro[38, 121], combined with the high mutation rate of HIV-1 and the larger population of target cells expressing CXCR4, would predict an early and nearly universal emergence of CXCR4-using variants during chronic infection. However, the fact that CXCR4-using variants seem to develop successfully only once during HIV-1 infection [112, 132], the scarcity of R5 virus variants with intermediate genotypes, and the fact that the newly emerged CXCR4-using variants differ from coexisting R5 variants by more than the minimally required number of amino acid mutations [133] altogether suggest that the virus evolves to the CXCR4-using phenotype through less-fit intermediate stages. The late emergence of CXCR4- using variants might be explained by an inability of these intermediate variants to compete with the well-established R5 virus population in spite of their broader cellular host range. Once established, however, R5X4 viruses may have developed the optimal fitness to predominate during the transition phase [134], although they may eventually be outcompeted by HIV-1 variants with a pure X4 phenotype.

The progressive divergence between co-existing R5 and CXCR4-using clones in phylogenetic trees reflects the continuous evolution of the variable loops of gp120 following the acquisition of CXCR4 usage. This implies that the structure of gp120 continues to evolve to optimize its interaction with the coreceptors. This evolution is accompanied by improved coreceptor-binding affinity, which in turn is reflected in a decreasing sensitivity of R5 variants to inhibition by CCR5-binding chemokines and small- molecule CCR5 antagonists [127, 128, 135, 136], as well as of CXCR4-using variants to SDF-1 and the synthetic antagonist AMD3100 [137].

Once established, CXCR4-using variants are likely to be more replication competent than R5 variants, given the broader target-cell range and the generally higher replication kinetics of CXCR4-using HIV-1 in vitro[132, 138, 139]. In light of these considerations, it seems paradoxical that R5 variants persist and may even expand in vivo after the emergence of CXCR4- using HIV-1 [113]. The most likely explanation for the observed coexistence of R5 and CXCR4- using variants is their distinct target cell range within an infected patient [120, 140, 141]. R5 variants seem to reside mainly in activated/memory CD45RO⁺CD4⁺ T cells, which express CCR5, whereas CXCR4-using variants reside in both CD45RA⁺ and CD45RO⁺CD4⁺ T cells [120]. Interestingly, evidence for frequent recombination events between co-existing R5 and CXCR4-using variants has been reported [142, 143]. While phylogenetic trees based on envelope sequences show a clear-cut separation of coexisting R5 and CXCR4-using virus variants, this separation is not observed when the phylogenetic tree is based on gag sequences. Apparently, R5 and CXCR4-using variants can co-infect cells that express both coreceptors, thus allowing for the recombination of their genetic material.

Interestingly, the distribution of R5 and CXCR4-using variants in different blood compartments may vary. Two recent studies reported a higher prevalence of predicted CXCR4-using envelopes in PBMC than in plasma [144, 145], although a third study could not confirm this discrepancy [146]. Differences in the sensitivity of the coreceptor usage predictors used may be responsible for the different study outcomes. Thus, the relevance of the choice of patient material for the determination of HIV-1 coreceptor usage remains to be established.