Early and late cytotoxic T lymphocyte responses in HIV infection

Abstract

Breadth of HIV specific CTL responses CD8+ cytotoxic T lymphocytes (CTL) recognize virally infected cells by interacting with their T cell receptor and virus protein derived peptidic fragments, so-called epitopes presented by MHC class I molecules on the surface of infected cells or professional antigen presenting cells such as dendritic cells [1–3]. For many human pathologically relevant viruses, numerous epitopes have been identified which are presented by a large diversity of different HLA class I molecules. More than 180 optimal CTL epitopes have been defined for HIV by the end of 2002 and more are added every year [4]. These epitopes are spread over the entire HIV protein sequences with some regions more heavily targeted than others resulting in a distinct pattern of variable epitope density [5]. Such CTL epitope clustering may be explained by the variable HLA class I allele distribution in various ethnicities, which leads to the binding and presentation of specific epitopes which are restricted by alleles not present in other ethnicities. Alternatively, clustering may be due to the conserved character of heavily targeted sequences and antigen processing preferences and has been reported to differ in individuals of different ethnicity [5–7]. Indeed a recent report by Yusim et al. illustrates the importance of processing preferences for epitope clustering as proteasome cleavage sites were found to be predictable for the presence of CTL epitopes in a test sequence without previously described CTL epitopes [8]. In addition, this report also described a trend towards a loss of residues which seem to be required for high affinity epitope binding to the HLA class I molecules’ F pockets (C terminal end) as the virus evolves. The need for epitopes to possess such suitable C terminal ends has also been incorporated into the design of overlapping peptide sets used to detect HIV specific CTL responses [9]. Thus, there seem to be factors that limit the immunogenicity of certain HIV protein regions-suggesting that not every residue of the viral protein sequence will be included in CTL epitopes. Nevertheless, our own very recent data clearly shows that when the entire HIV sequence was synthesized as a 410 overlapping peptide set and used in Elispot assays, more than 80% of these peptides were recognized by at least some out of 120 individuals tested. This indicates that the majority of the HIV protein sequence is efficiently processed from the precursor protein and can indeed be targeted by HIV specific CTL [5]. Over the last few years, all viral proteins have been reported in numerous reports to be targeted by virus-specific CD8 T cell, although at varying degrees [4,5,10–15]. In a recent study by Chouquet et al., memory CTL responses were measured from circulating PBMC in a large cohort of 148 chronically-infected adult patients, enrolled before the institution of HAART. In that study, CTL activities were directed to all viral proteins and the breadth of HIV antigens recognized by memory CTL was shown to be a correlate of immune control of HIV-replication and disease progression [16]. However, anti-retroviral treatment, stage of disease, age at infection and time since infection appear to be important factors in the magnitude of total CTL responses as well as for the ability to detect single CTL specificities [13,17–23] and need to be kept in mind when interpreting CTL data. Besides time since infection and treatment status, CTL responses may also vary depending from what body location the samples were obtained. Work by Shacklett and colleagues found CTL activity in the gut-associated lymphoid tissue (GALT) and the mucosal cervicovaginal surface to differ in at least some cases in terms of specificity and magnitude [24,25]. Similarly, Altfeld et al. recently demonstrated that the CTL responses in lymph nodes and peripheral blood differed from each other [26]. Detailed analysis however revealed that the differences were mainly due to the magnitude of the responses detected in the two locations as all responses found in the peripheral blood were preceded by responses in the lymph nodes. In individuals where the repertoire and magnitude of responses was increased upon structured treatment interruptions, the responses previously limited to the lymph node became subsequently detectable in the peripheral blood as well, indicating that PBMC analyses may possibly underestimate the true range and strength of total virus specific CTL responses. In some of these studies [24], phenotypic analyses were performed which, together with some more recent studies, underscore the potential importance of maturation and homing markers for effective CTL responses [27]. This is also evidenced by a recent study by Chen et al., where the failure of HIV-specific tetramer binding CD8 T cells to efficiently traffick into the lymph nodes was explained by the lack of appropriate haming receptors [28]. Since a large proportion of HIV replication is believed to occur in the lymph nodes, this study identifies one potential explanation as to why CTL may not effectively control viral replication in vivo and may also offer an explanation to the conflicting data surrounding the association between the magnitude of CTL responses and viral load. Additional studies by a number of laboratories over the last year have added more evidence that phenotypic characterization of these responses may have predictive value for the identification of functionally competent CTL populations. Andersson et al recently demonstrated that in lymphoid tissue from patients with acute HIV-infection, granzyme A was induced in conjunction with a relative lack of perforin expression compared to lymphoid tissue from patients with acute-EBV-infection [29]. Similarly, studies by Oxenius et al showed particular CTL phenotypes, which were changed upon initiation of HAART, providing further evidence for a distinct development and activation pattern of HIV specific CTL compared to other viral infections [30]. Depending on the assay systems used, these differences may affect the ability to detect HIV specific CTL responses significantly and may explain different outcomes in different studies. For instance, while Demarest et al. found the absence of CTL in a rapid progressor to be the crucial factor for uncontrolled disease progression, Islam et al. found very strong CTL responses in an individual with similarly rapid disease progression [31,32]. Clearly, larger cohorts of rapid progressors need to be looked at to determine why persistent CTL responses can or can not be detected in these individuals and, if widely detectable, why they do not exert better control over viral replication. Recognizing these differences in specificities, breadth and magnitudes of responses may largely depend on the factors listed above and likely other (yet) undefined influences, appear crucial to the definition of effective immune correlates of HIV infection. Importantly, the rapid induction of these responses shortly after acute infection may carry significant importance and a number of reports have addressed these issues lately. However, whilst studies in SIV infected monkeys have revealed very early CTL responses (and CTL escape), these data largely await their confirmation in human HIV infection. Early CTL responses in SIV and HIV infection Through an elegant ‘reverse genetic’ approach, Allen et al. have identified early CTL responses in SIV infected macaques from which SIV rapidly escapes [33]. Nevertheless, these very early responss, detectable within a few weeks after injection were considered to (at least partly) control the peak of the initial viremia and possibly help limit the virus driven destruction of virus specific T-helper cells [33–35]. More recent studies from the same group have now demonstrated SIV specific CTL activity within even shorter period of times after acute infection; again directed against the initially described ‘early’ epitopes but also against a number of additional regions [36]. Interestingly, high binding affinity to their restricting MHC molecule appears necessary for these epitopes to induce early responses. These data suggest that CTL responses against SIV and likely HIV can be generated very rapidly and may indeed be crucial in limiting the initial hit on the immune system, especially by preventing ongoing infection of HIV-specific CD4 T cells [35]. However, studies in HIV infected humans are still limited mainly due to the difficulty to identify very early infection and the fact that the infecting viral isolate is generally unknown. Furthermore, the HLA diversity in the study subjects make similar studies in humans more difficult. Early and late responses in HIV infection Identification of individuals with acute HIV infection has allowed for a number of insightful studies focusing on the evolution of the virus specific T cell response from early on in the infection. The most striking observation in an initial study by Dalod et al. was that the magnitude of HIV-specific CD8+ T cell activity was much lower in the primary infection than in the chronic stage of infection [37]. In another study, looking at individuals with treated acute infection and subsequent treatment interruptions, Altfeld et al. found a narrow T cell response to only a few HIV derived CTL epitopes in early infection [19]. With repetitive, subsequent treatment interruptions, these responses broadened significantly and, in the presence of strong T-helper cell responses, were considered to be responsible for the observed relative control of viral replication during times off drugs. Interestingly, one of the generally immunodominant responses directed against the HIV Gag p17 derived, HLA-A*0201 restricted SL9 epitope (p17, aa 77–85, SLYNTVATL) was not among the responses detected early in infection [21]. By inference from the SIV model, it appeared possible that this response could have been induced and escaped from the CTL surveillance even before the patients presented for the first time [33,36]. However, sequence analyses of HIV Gag sequences demonstrated that this scenario was unlikely as no increase in potential CTL escape variants of this epitope were identified in HLA-A*0201 expressing individuals with acute HIV infection compared to HLA-A*0201 negative individuals [21]. This brings up the question as to why this SL9 specific CTL response has been found, with regular consistency, in about 70% of individuals with chronic HIV infection [38,39]. One possibility is that the early treatment of these individuals has prevented the induction of this response which may have a higher antigen threshold than other responses that were detectable early on. Alternatively, other responses detected in early stage of infectio could have mutated away, opening the field for a second wave of CTL specificities taking over in their place. Whether such responses could be effective in controlling viral replication is unclear and conflicting data have been published on the correlation between CTL activity and viral load [14,40,41]. Accordingly, CTL may not exert strong selection pressure, at least when looked cross-sectional and focusing on single epitopes [39]. This observation would be in line with an ineffective phenotype of these CTL and would support data that do not find a correlation between the magnitude of the CTL response and viral load. However, and as discussed below, many conflicting results might be due to different experimental approaches detecting different viral-specific CTL populations. Data by Betts et al., show a lack of indirect correlation between the magnitude of total HIV protein specific CTL and viral load [14]. Similarly, more recent data by Addo et al. confirm these data: when testing more than 50 individuals with a comprehensive set of overlapping peptides spanning the entire HIV protein sequence, neither the number of peptides targeted nor the total strength of these responses correlated with control of viral replication [42]. This suggest that control of viral replication is not a simple function of the total virus specific CTL response, rather, at least some of the ex-vivo detected responses may be irrelevant in terms of virus containment or, alternatively, the total CTL population may be functionally impaired. Indeed, dys-functional CTL populations against HIV have been described by many laboratories over the last few years (see above). An emerging consensus seems to be that strong T-helper cell responses are needed to maintain and maybe also to induce qualitatively meaningful CTL specificities [43]. An alternative explanation for these discrepancies might be of technical nature. In a study of 20 HLA-A*0201 positive HIV-infected children, the IFN-γ production in response to stimulation with two HLA A*0201-restricted immunodominant CD8 epitopes (SLYNTVATL (SL9) in Gag and ILKEPVHGV (IV9) in Pol) was measured using the Elispot assay. The results were compared to labeling with the corresponding tetramers. Comparison of results from tetramer and Elispot assays suggests that only a fraction of HIV-specific CD8+ T cells were functionally active as reported by others in HIV infection in adults [44]. Most importantly, the frequencies of IFN-γ producing CD8+ T cells were positively correlated with viral load, whereas frequencies of tetramer-binding CD8+ T cells were not [45]. Just how ineffective a person's anti HIV CTL response might be is suggested in a recent report by Altfeld et al. [46]. An individual who was identified with acute HIV infection and treated immediately after diagnosis developed strong and very broad CTL responses upon subsequent treatment interruptions [47]. During the 3rd treatment interruption, the individual who had mounted 25 different T cell responses became super-infected with a second virus. The super-infecting virus was functionally conserved in 8 of 16 sequenced epitopes and strong CTL responses were detected against these epitopes [46]. However, the individual was unable to control the second in-coming virus and presented with strongly elevated viral loads within weeks after super-infection. These findings are reasons for major concern for vaccine development and beg the question as to whether CTL responses against the known and well described epitopes are the correct immune correlates to be measured and taken as a reference when designing vaccine candidates. There may not be an easy answer to this question as the explanation likely is multifactorial and needs to take into account several aspects of antigen processing, presentation, TCR affinity as well as virus sequence conservation and replication fitness of variant virus. Epitopes restricted by HLA alleles associated with slow HIV disease progression Large surveys of HIV infected populations have identified certain HLA alleles that are over-represented among individuals with slow disease progression [48,49]. These include, among others, the HLA-B27 and HLA-B57 alleles. Interestingly, in HLA-B27 positive individuals, the immunodominant CTL response is directed against an epitope in HIV Gag p24 [50,51]. Upon mutations in this epitope, the dominant immune response is shifted to another, sub-dominant HLA-B27 restricted epitope. Loss of effective control of viral replication concurrent with changes in the viral sequence suggest that the initial response to the wildtype epitope represents an example of an effective CTL responses able to (at least partly), control viral replication [50,51]. However, once escaped, this epitope may not easily revert back to wild-type, even in the absence of the HLA-B27 allele and can thus be transmitted and reach fixation in the general population [51,52]. In the case of HLA-B57, a recent report characterized the evolution of CTL responses from early on in acute infection over time [13]. Interestingly, when testing for the total virus specific immune reactivity, the initial immune response appears to be heavily dominated by CTL restricted by HLA-B57. This again is evidence that at least some responses, especially those restricted by HLA alleles associated with slow disease progression, may indeed be able to control viral replication. This is further supported by the observation that HLA-B57 expressing individuals are under-represented in our acute infection cohort at Massachusetts General Hospital. It appears that these individuals control acute infection in a less symptomatic manner than individuals that do not express HLA-B57 and therefore rarely present at the HIV clinic. CTL responses against early expressed viral antigens and the impact of sequence variation It has been speculated that CTL responses against epitopes derived from viral proteins that are expressed early in the viral life cycle would give the immune system a crucial advantage over the virus by enabling CTL to recognize infected cells before they can assemble and release new virions and before other immune modulatory effects by other viral proteins can take effect on the antigen presentation pathway [33,53–55]. However, whether these responses are induced first in acute infection and whether these responses indeed can provide superior protection has not been demonstrated conclusively yet as the HLA diversity and differences in the processability of viral proteins complicate these analyses. In addition, if the induction of the most effective CTL responses depends on antigen cross-presentation by dendritic cells, structural antigens may be presented more efficiently compared to regulatory antigens which may rely on presentation on the infected cells themselves [3]. As described recently in large cohort of chronically-infected adult patients the most frequent targets of memory CTL were the structural proteins [16]. In this study, only 10% and 5% of the patients had at entry CTL activities directed to Rev and Tat respectively and, in contrast to previous reports, no preferential recognition of Rev and Tat was observed in non-progressors versus progressors [16]. Moreover, individuals with a certain HLA type may be able to present some regulatory protein derived epitopes with a higher binding affinity than other epitopes, making their response against early antigens potentially more potent, compared to individuals who do not have the HLA alleles that present early antigens very well. Therefore, although of high importance, the question as to whether early expressed epitopes would be beneficial, still remains to be fully clarified. Besides the genetic predisposition of the infected individual to mount an effective CTL response, early or late, the sequence variability of the infecting virus may play a major role in the development of these responses [52]. However, the CTL mediated selection pressure in controlling the course of viral evolution after infection remains debatable and at least two mechanisms might occur in vivo which may limit the development of effective CTL escape: (i) recognition of several epitopes variants by the same TCR and (ii) generation of several CTL populations specific for a single epitope recognizing different variant epitope sequences. A recent study by Buseyne et al. addressed these possibilities in a patient from whom two CTL populations specific for the same epitope were obtained at different times after infection. The CTL expressed different TCR, recognized different variants of the epitope and were tolerant to mutations in the antigenic epitope and even to amino acids variations on the HLA-B presenting molecule. This demonstrates that the flexibility of the TCR and the polyclonality of the CTL response may be crucial for control of a broad number of mutant viral sequences [56]. Besides the well documented sequence variation within the targeted epitope as a means to evade immune control, the deletion of the entire epitope or at least parts of it may serve as an additional mechanism for HIV-1 escape from CTL immune pressure. This has been reported recently by Singh et al. in a longitudinal follow-up of a patient over a ten years period [57]. Bulk unstimulated effector (eCTL) and stimulated memory CTL responses against Gag, Pol and Nef were studied and the patient was found to have a predominant, strong CTL response against Nef in unstimulated peripheral blood mononuclear cells with a peak during month 40 of the follow-up. PCR amplification and nucleotide sequencing of the plasma viral variants revealed a viral variant which had deleted the targeted epitope and which was detected early during the follow-up. This variant essentially replaced the wild type virus during the peak CTL response [57]. It thus appears that with a decline in the selection pressure, the wildtype virus took back over again, indicating the rapid evolution of viral quasispecies in vivo in dependence of more or less effective CTL pressure. Biased identification of CTL responses: role of viral variability and importance of the assay systems used As mentioned above, the sometimes conflicting or inconclusive data regarding the effectiveness of CTL responses in vivo, total as well as measured on a single epitope level, may have additional reasons besides the ones discussed. One major reason for inconsistencies might be a shift in the methodology used to assess these responses ex vivo (detection of activated effector CTL) or after in vitro stimulation and expansion (detection of memory CTL). During the past three decades most laboratories have studied HIV-specific CTL activities mainly using the traditional chromium release assay. This allows to measure the degree of lysis of target cells by CD8+ T effector cells thereby assessing their cytolytic potential. Alternative assays have been developed, including variations in the detection of released cellular components [58] or the detection of induced cell death in the target cells [59]. However, during the last few years, much more powerful and high throughput alternative techniques have been developed, including Elispot, intracellular cytokine staining (ICS) and tetramer analyses. While the first two are based on the detection of specific cytokines secreted by CD8+ T lymphocytes after antigen specific activation, the latter detects the presence of T cell receptors able to bind to the specific HLA/peptide complex without taking into account the functional activity of these cells. Thus, while these techniques measure different T cell properties and in combination can be very useful to identify ‘silent’ T cell responses-[44], the chromium release assay remains the only well established assay able to measure the actual functional cytolytic activity. Since this activity may be rather important in vivo in controlling viral replication, measuring the cytolytic activity, ideally directly ex-vivo, still can be an important factor in assessing the ‘usefulness’ of CTL responses. Another potential issue leading to inconsistent results may be the antigen reagents used in the various in vitro assay systems. Initial identification of HIV specific CTL responses were based on the use of recombinant vaccinia virus constructs or transfected cells expressing HIV proteins cloned from viral isolates generated early in the epidemic [60,61]. Since then, the virus has evolved dramatically and likely in response to population immune pressure [8,62,63] and current viral isolates can differ significantly from the initial viral isolates. Thus, testing individuals with acute infection today using reagents based on sequences from 20 years ago, may potentially underestimate CTL responses and lead to the biased identification of CTL epitopes located in more conserved regions than in variable [5,8]. Therefore, CTL responses identified over the last 15 years may have resulted in CTL epitopes that can not escape immune pressure or that revert rapidly back to wild type virus when no CTL pressure is present. The virus may have learned to life with this immune pressure by selecting for epitopes with reduced binding affinity or epitopes that are expressed late in the viral life cycle. As a consequence, when measuring total HIV specific CTL responses, ineffective responses that are unable to control viral replication may mask the responses that actually do the job. The detailed characterization of the evolution and the kinetics of CTL responses directed against multiple viral proteins from very early time points in infection will not only allow to appreciate the degree of acute CTL escape but will also help to identify the clinically most relevant relevant immune responses against HIV. These responses may be directed against early expressed proteins and the targeted epitopes may be presented on HLA alleles that are associated with slow HIV disease progression. In any case, it will be important to assess these responses using contemporary tools such as antigen sources based on autologous or currently circulating (consensus) viral sequences in order to prevent a bias towards regions with low sequence variability. The proper identification of these effective HIV specific CTL responses is a question that urgently requires clarification to allow for adequate vaccine design.