Patterns of Immunodominance in HIV-1–specific Cytotoxic T Lymphocyte Responses in Two Human Histocompatibility Leukocyte Antigens (HLA)-identical Siblings with HLA-A*0201 Are Influenced by Epitope Mutation

The 24 HLA-A*0201–positive adult donors were selected from asymptomatic HIV-infected individuals attending genito–urinary clinics in the Oxford Region, St. Mary's Hospital (London), or the Oxford Haemophilia Centre. No donors had received antiretroviral therapy.

Donors 003 and 023 are HLA-identical hemophiliac brothers, aged 27 and 25, respectively. Both 003 and 023, and also donor 008, were infected between August and December, 1983, after infusion of a particular batch of HIV-contaminated Factor VIII concentrate. These donors were exposed only to this single batch of Factor VIII between August and December, 1983, inclusive, during which donor 003 received 25,000 U and donor 023 received 27,000 U (the mean usage for a hemophiliac being ∼30,000 U per yr; (Spooner, R., personal communication). Donor 023 first became seropositive for HIV in December, 1983 (aged 10 yr), while both 008 and 003 (aged 12 yr) first became seropositive in February, 1984, having been seronegative previously. The brothers 003 and 023, each having a natural factor VIII level of <1%, received very similar doses of Factor VIII. They have lived in the same house throughout their lives; Factor VIII has been exclusively home administered since 1976; all supplies of Factor VIII have been shared; and there has been no batch of Factor VIII of which one brother has received significantly more than the other. The original batches of contaminated Factor VIII unfortunately were unavailable for this study.

Donor 46M first presented in 1994 when her child was diagnosed as HIV infected. The route and duration of infection in 46M are unknown.

HLA tissue typing was performed by amplification refractory mutation system (ARMS)-PCR using sequence-specific primers (29).

Peptides were synthesised by Research Genetics, Inc. (Huntsville, AL). The purity of the peptides was determined by HPLC and the identity of the peptide confirmed by mass spectrometry measurement. Peptides were dissolved in DMSO and diluted in RPMI-1640 (Sigma Chem. Co., St. Louis, MO). Overlapping 15–20-mer peptides were supplied by the Medical Research Council AIDS Directed Programme (MRC-ADP, Hertz, UK).

Vaccinia virus recombinants expressing the HIV-1pol and gp120 genes were provided by the MRC-ADP, the vaccinia expressing the gag gene was as described previously (30), that expressing the nef gene was provided by Transgene (Strasbourg, France), and the control influenza PB2 vaccinia was kindly provided by Dr. B. Moss (National Institutes of Health, Bethesda, MD).

Bulk-cultured CTLs were generated as previously described (30). In brief, one-fifth of PBMC separated from whole blood were added to the other four-fifths after 24 h stimulation with PHA to activate expression of autologous HIV. These cells were cultured in RPMI-1640 supplemented with antibiotics and 10% heat-inactivated FCS (Globepharm, Esher, UK) (R10) for 7 d, after which 10% Lymphocult T (Biotest) was also added to the medium. Peptide-specific lines were generated by pulsing PBMC with 100 μM peptide for 1 h, followed by 3-d culture in R10 with added IL-7 (40 ng/ml). Subsequent media exchanges were performed using R10 containing 10% Lymphocult T. Assays were performed on days 14–24 of culture.

Dominant responses were established by assaying bulk CTL against autologous EBVtransformed B-lymphoblastoid cells lines (BCL)¹ infected with a panel of vaccinia recombinants expressing HIV-1 gp120, gag, pol, and nef genes. Overlapping peptides (MRC-ADP), 15–20 amino acids in length, were then used to define the region(s) containing recognized epitopes. Additionally, previously defined epitopes presented by matching HLA class I molecules were tested for recognition by bulk CTL.

Vaccinia-infected targets were infected with 3 PFU/cell for 90 min, washed, and incubated in R10 overnight. In peptide-based assays, targets were pulsed with peptide for 1 h, washed, and aliquots of 5 × 10³/well in 100 μl were used in duplicates. Targets were labeled with ⁵¹Cr and washed a total of three times before use in the assay. Effectors were added to each well in 100 μl and incubated with targets for 4–6 h. Percent lysis was calculated from the formula: 100 × (E − M)/(T − M), where E was the chromium release from 20 μl supernatant of wells containing targets and effectors, M the release from wells containing targets and medium only, and T the release from wells containing targets and 5% Triton X-100. Where the chromium release from duplicate wells differed by >15% from the mean, the data were discarded. Spontaneous release (M/T) was always <30%. Percent of specific lysis was calculated by subtracting percent lysis of targets pulsed with no peptide or infected with the control flu–vaccinia from percent lysis of targets pulsed with peptide or infected with HIV-1 protein–vaccinia, respectively.

Full-length p17 gag was amplified from proviral DNA isolated from PHA-activated PBMC using nested PCR. The primers used were: 5′-GTTCTAGGTGATATGG-3′, 5′-ACTAGCGGAGGCTAG-3′ for the primary reaction; and 5′-AATTCCATGGGTGCGAGAGCGT-3′ and 5′-CAGTTCTAGATCTAGTAATTTTGGGCTGACC-3′ for the secondary reaction. PCR products were cloned into the T vector system (Amersham) and epitopes sequenced with insert-specific primers using the Sequenase Version 2.0 protocol (USB, Cleveland, OH).

The HLA class I and class II tissue type of the HLA-identical siblings, donors 003 and 023, was the following: HLA-A*0201, -A*03 -B*07, -B*51 -Bw*4, -Bw6 -Cw*7, -Cw14 -DR*1, -DR*8 -DQ*4, -DQ*5. The respective CD4 counts and virus load are shown in Fig. 1,a. The CTL response profile was quite different in the two brothers (Fig. 1,b, Table 1). CTL from donor 003 recognized the HLA-A*0201–restricted peptide SLYNTVATL as the immunodominant response, but two HLA-A3–restricted epitopes within p17 Gag, one in Nef, and one in Env were also seen by CTL from this donor. The dominant responses identified in donor 023 were to two HLA-B7–restricted epitopes not previously described (Fig. 1,b), one in p24 Gag, and one in Nef, neither of which was recognized by CTL from donor 003. A relatively weak response to the HLA-A*0201 epitope in Pol ILKEPVHGV was identified in bulk-cultured lymphocytes from donor 023. This peptide was not seen in the bulk response by 003, although it could be seen as a weak response when PBMC were stimulated directly by the peptide (Fig. 1,c). Conversely, no responses were observed in donor 023 to the HLA-A*0201– and HLA-A3–restricted p17 Gag epitopes, the strongest responses observed in bulk CTL from donor 003 (Fig. 1 b). This pattern of Gag-specific CTL responses observed in donors 003 and 023 has been stable since 1986 (31).

One potential explanation for the relative failure of donors to generate a response to a particular epitope is that autologous antigenpresenting cells may fail to process it. The possibility of processing polymorphism being associated with a particular response was not the explanation in donors 003 and 023, who are HLA class I and class II identical. Using Gag–vaccinia-infected BCL from 023 and from an HLA-A*0201– matched donor (5M) as targets, the Gag peptide SLYNTVATL was processed and presented as efficiently to 003 bulk CTL as by autologous BCL (Fig. 2,a). Likewise, BCL from Gag responders were able to process and present the Pol peptide very efficiently (Fig. 2 b). Gag–vaccinia-infected BCL from other Gag nonresponders (see below) similarly were able to present SLYNTVATL efficiently to 003 CTL (data not shown).

Provirus DNA from donor 003 does not differ significantly from the consensus B clade (wild-type) Gag sequence, and in particular, no deviation from the wildtype sequence encoding the two immunodominant epitopes SLYNTVATL (HLA-A*0201) and RLRPGGKKK (HLAA3; our unpublished data) was seen in more than 40 clones (Table 2). In contrast, proviral DNA from donor 023 differs strikingly from the wild-type sequence. The proviral sequences from 023 encoding the two epitopes best recognized by CTL from 003 were significantly altered, encoding the peptides SLHNAVAVL and RLRPGGKKC in 100% of sequences cloned from the most recent timepoint. No response to these peptides or the wild-type peptides was seen in bulk CTL from donor 023 (see Fig. 1,b; data not shown) and no response was made to the variant peptides by bulk CTL from donor 003 (Fig. 3,b; data not shown). The distinctive epitope changes (His for Tyr at position 3 in the SLYNTVATL epitope) observed in proviral DNA from donor 023 have been stably present at least since the earliest timepoint sequenced, a period of almost a decade, and similarly the wild-type sequence has been stable within donor 003 for this duration (Table 2). Thus, the two immunodominant responses in 003 may be absent in 023 because both epitopes have mutated to nonrecognized forms.

To determine which of these responses was typical of HIV-positive donors with HLA-A*0201, we determined the HLA-A*0201–restricted CTL responses in 22 further donors, using autologous virus-infected activated T cells to stimulate the responses. One hemophiliac donor, 008, was infected by the same batch of contaminated Factor VIII as 003 and 023. One donor made no detectable CTL response (SC7), and in two donors there was an equal response to the Gag and Pol epitopes (069, responses to both peptides; and 606, responses to neither). The response in donor 48M to the Pol peptide was not tested. Bulk CTL from 15 of the other 20 donors preferentially responded to the Gag epitope, while 5 (including 023 and 008) preferentially responded to the Pol epitope.

Analysis of the p17 gag sequences within Gag responders shows that there is little divergence away from the wild-type sequence (Table 2). In contrast, there was greater variation in this region in sequences from Gag nonresponders. In donors 008 and 606 (Gag nonresponders), the encoded Gag peptides were not well recognized by peptide-specific CTL, whereas variation transiently observed in the Gag responder 868 did not affect CTL recognition (Fig. 3 a). Although mutations within the epitope and within amino acids flanking the epitope were observed more frequently in Gag nonresponders, some nonresponders (065 and 241) carried proviral sequences encoding changes within the epitope that did not alter CTL recognition, and some Gag responders carry sequences encoding changes flanking the epitope (1M, 868).

When bulk cultured CTL from donor 46M were first assayed in 1994, a clear response to the SLYNTVATL Gag epitope was observed. Subsequently, this response has disappeared (Fig. 3,c), while other responses remain readily detectable. In 1994, 50% of autologous provirus sequences encoded the index peptide; 2 yr later, 100% of sequences encode the variant SLFNTVATL (Table 2). This Tyr→ Phe change at position 3 in the epitope results in poor CTL recognition (Fig. 3,a). In all Gag nonresponders tested so far (023, 241, 065, 606, and 46M; timepoint 12/96) stimulation of PBMC with the SLYNTVATL peptide allowed detection of SLYNTVATL-specific CTL (specific lysis at E/T ratio of 5:1 ranging from 9 to 40%; Fig. 3,c; data not shown), but stimulation of 46M PBMC with the SLFNTVATL variant did not result in the generation of any peptide-specific CTL (Fig. 3 d).

CTLs play a central role in the immune response to virus infections, as well as to other intracellular pathogens and to tumor cells. The importance of CTL at all stages of HIV infection is increasingly apparent. There is also growing evidence that there are qualitative differences in the responses generated by infected individuals, and that these may influence the long-term control of the infection and, ultimately, the prognosis. The determinants of immunodominance of MHC class I–restricted CTL responses are of considerable relevance to disease progression.

We have studied the HIV-specific CTL responses directed through the commonest HLA class I molecule found in caucasoid and many other populations, HLA-A*0201 (32). The majority of donors recognise the Gag epitope SLYNTVATL predominantly, and a minority, the Pol epitope, ILKEPVHGV. Because the response is not universally directed towards one epitope rather than the other, the explanation cannot lie solely in differences of affinity of the peptides to HLA-A*0201. In fact, our binding studies show that the Pol peptide binds significantly better than the Gag peptide (data not shown), and other studies have similarly shown very strong binding of the Pol peptide to HLA-A2 (33). This is consistent with what is known of the peptidebinding motif for HLA-A*0201, where the preferred anchor residues are leucine at position 2 and valine at position 9 (COOH-terminal anchor, PC). Of 16 HLA-A*0201– restricted CTL epitopes listed (34), 12 had valine at PC, 2 had leucine. A previous study comparing the relative binding affinities of polyalanine substituted amino acids to HLAA*0201 showed that the substitution of valine for leucine at PC resulted in a 10-fold increase in relative binding (35).

The predominance of CTL responses to the Gag epitope is in better agreement with data estimating the abundance of each peptide on the cell surface, where >30 times as many SLYNTVATL molecules were present on the surface of stably infected Jurkat-A2 cells as ILKEPVHGV molecules (27). We investigated the possibility that HIV-infected cells from Gag responders might be presenting the SLYNTVATL peptide more efficiently than the Pol peptide, but experiments using Gag–vaccinia-infected BCL did not support this (Fig. 2,a; data not shown). Similarly, BCL from Gag responders, when infected with Pol–vaccinia, presented the Pol peptide at least as well as Pol–vaccinia-infected BCL from a Pol responder (Fig. 2 b).

The observation has often been made that HIV-specific CTL epitopes tend to cluster in immunodominant regions of proteins (17). Because of the report of HLA-B8 and HLA-B*2702 MHC class I molecules competing for the presentation of the overlapping influenza-specific epitopes ELRSRYWAI and LRSRYWAI (36), the possibility was raised that this phenomenon might influence immunodominance of HIV-specific CTL responses similarly. Whereas donor 065 generated no response to the Gag epitope SLYNTVATL, a strong response was made to the HLA-B8–restricted overlapping epitope ELRSLYNTV (Goulder, P.J.R., S.W. Reid, D.A. Price, A.J. McMichael, R.E. Phillips, and E.Y. Jones, manuscript submitted for publication), and a weaker response to the HLA-A1–restricted epitope, GSEELRSLY (our unpublished data). Although Gag–vaccinia-infected HLA-A*0201/B8 targets appeared to express the SLYNTVATL epitope adequately (data not shown), we cannot exclude that competition between MHC molecules for overlapping epitope peptides could influence the development of immunodominance of the response, favoring a predominant response to the Pol epitope in this donor.

Donors 003 and 023 are two HLA-identical brothers who were infected by a virus at the same time, probably the same quasispecies of virus, at a very similar dose. Neither donor has progressed to disease over 12 yr of infection. Donor 003 has a virus load below the level of detection (<500 RNA copies/ml), and that of donor 023 is low (8,500 RNA copies/ml plasma); neither has received antiretroviral therapy. The CTL responses are quite different, which demonstrates that, in this instance, the pattern of response is not determined solely by the HLA genotype, or by TAP polymorphism or by proteasome-related processing effects.

It is difficult to demonstrate unequivocally that viral isolates from the two individuals have derived from a single original virus, because selection acting early on in the course of infection may have resulted in dramatically divergent sequences. Unfortunately, we do not have access to the contaminated batch of Factor VIII believed to have infected these two brothers, and the earliest sequence available for analysis is 3 yr after infection. Comparison of p17 gag sequences would show major dissimilarities between 003 and 023 sequences, because (as argued below) selection pressure exerted by CTL early on in infection has resulted in epitope variation and switching of immunodominant responses to different epitopes in donor 023. Neighboring sequences outside epitopes might be expected to show compensatory changes, increasing the interindividual sequence divergence further. Comparison of other regions (such as env) of the genome might be equally misleading because of different selective pressures acting on the viruses in the two brothers.

The clear epidemiological evidence that these brothers were infected with the same original virus is the detailed documentation that throughout life they have shared identical supplies of Factor VIII, they were exposed at the same time exclusively to one particular batch of Factor VIII, and shortly afterwards both brothers seronconverted. Donors 003 and 023 are now 27 and 25 yr of age, respectively, and first became seropositive within 10 wk of one another (December, 1983 and February, 1984) when aged 12 and 10 yr, having been exposed by home-administered treatment exclusively to the same batch of non–heat-treated Factor VIII between August and December, 1983.

The study by Tsomides et al. (27) implies that an excess of the Gag SLYNTVATL epitope compared with the Pol ILKEPVHGV epitope is presented by infected CD4⁺ cells in both donor 003 and donor 023 (27). The remarkable difference between these two donors is in the provirus sequence encoding the Gag epitope. The peptide encoded by 100% of provirus from donor 023, SLHNAVAVL, is not recognized by autologous bulk CTL or Gag-specific CTL lines (Fig. 3,a). Equally striking is the sequence change observed in provirus from donor 023 encoding the HLA-A3–restricted epitope, RLRPGGKKK, the second of two dominant responses in donor 003 (Fig. 1,b). In this case, the PC anchor lysine residue is switched to a cysteine, a change resulting in abrogation of recognition by CTL from donor 003 (Fig. 3 b). (No epitope peptide has been described with a cysteine at the COOH terminus [34], so this peptide may not even be processed or transported.) This combination of mutations is unique for each epitope, when compared with 32 B clade Gag amino acid sequences in the database (38). It is also noteworthy that the unusual Tyr→ His sequence change has been stable for at least 9 yr.

Are the unique sequence changes observed in the provirus of donor 023 and the unusual preference in the CTL response of this donor for the Pol epitope causally connected? In another hemophiliac donor, 008, also probably infected by the same viral quasispecies as donors 003 and 023, provirus also encodes an epitope that is not recognized by Gag-specific CTL, and this donor, too, is unusual in being a Pol responder. Together with the PC anchor sequence change observed within the HLA-A3–restricted epitope RLRPGGKKK in donor 023 provirus, these data are suggestive of CTL escape occurring early on in infection and strongly influencing the subsequent stable pattern of CTL response observed. In donor 023, two mutations together might permit escape from both CTL epitopes, whereas in 003 this does not occur, perhaps because of the presence of other CTL responses.

The disappearance of a bulk response to SLYNTVATL in donor 46M in association with the increased frequency to fixation of the F3 variant is suggestive that this pattern of events may have occurred in other Gag nonresponders. In support of this is the finding that SLYNTVATL-specific CTL could be detected using the sensitive method of peptide stimulation of PBMC in all Gag nonresponders tested, implying the presence of memory CTL. Use of SLYNTVATL–HLA-A*0201 complexes (39) has shown that some Gag nonresponders nonetheless have a relatively high frequency of SLYNTVATL-specific effectors within PBMC, which implies that these Gag nonresponders have previously generated SLYNTVATL-specific responses that are no longer active.

A previous study described two donors with HLA-A11 and three other donors with HLA-B18, who made CTL responses to Nef-specific epitopes, provided that autologous virus sequences did not show mutations that abrogated CTL recognition (40). The data we describe show that such epitope variation is not necessary for a failure of a response towards the epitope to be observed; but often this is the case, and may be the result of multiple early escape mutations, occurring within 3 yr of seroconversion. Studies of CD8⁺ TCR Vβ usage after seroconversion have illustrated the spectacular Vβ-specific expansions and deletions that occur rapidly at this time (41). The swift emergence of CTL escape mutations at seroconversion has also been well described (42–43). The precise relationship of these dramatic fluctuations to escape variation within the cognate epitope has not been determined.

It is clear from these data that early escape mutation within the normally immunodominant epitope SLYNTVATL is not the only factor determining the pattern of an individual CTL response. It is possible, as in the Pol responder 065 (whose sequence carries no mutations within the epitope), that competition between class I molecules for overlapping epitopes may have an influence. In this study, we have not investigated the role of flanking mutations, which occurred universally in the Gag nonresponder group (including 065) and relatively infrequently in the Gag responder group (Table 2). The clustering of nonsynonymous mutations at sites flanking HIV-specific CTL epitopes has been noted previously (40, 44), and flanking changes have been demonstrated to affect presentation in some systems (19, 45).

In summary, the immunodominance of CTL responses in two HLA-identical siblings, probably infected by the same quasispecies of virus and at the same time, is not determined by HLA genotype or processing differences alone. The CTL responses in HLA-A*0201 seropositive donors to two welldefined HLA-A*0201–restricted epitopes is likely to be determined primarily by the high abundance on the cell surface of the Gag epitope relative to the Pol epitope. In the minority of individuals who fail to respond to the Gag epitope, mutation within the Gag epitope occurring early in infection that abrogates CTL recognition may influence the pattern of the CTL response away from the Gag epitope. In addition, variation within flanking regions, reducing the presentation of the Gag epitope, or competition by other HLA class I molecules for epitopes overlapping with the Gag epitope, may also determine the pattern of the CTL response.

We thank M. Brooks, M. Fletcher, and A. Tippett, the staff at the Oxford Haemophilia Centre, the Radcliffe Infirmary, and Wycombe Hospital Trust GU Department for the collection of blood specimens; C. Rizza, A. Gallimore, L. Tussey, S. McAdam, and P. Klenerman for help and discussion; and T. Rostron and R. Tan for viral load measurements.

This work was supported by the Medical Research Council (P.J.R. Goulder, A.J. McMichael, R.E. Phillips) and the Wellcome Trust (A.K. Sewell, D.G. Lalloo, D.A. Price, J. Whelan, R.E. Phillips).