Main

The strongest CD8-positive T lymphocyte responses to HIV and SIV are observed in the first few weeks of infection, coincident with the initial decline in plasma viraemia. We reasoned that viral escape might occur from immune responses that exert selective pressure during this acute phase of infection. To test this hypothesis, we examined viral evolution during acute infection of 18 rhesus macaques with molecularly cloned SIV.

Every animal (10/10) expressing the rhesus major histocompatibility (MHC) class I molecule Mamu-A*01 made CD8-positive T-lymphocyte responses, which peaked between 3 and 4 weeks after infection, to a newly defined epitope in Tat28–35 STPESANL14 (SL8, Fig. 1). In two of these animals, up to 10% of their CD8/CD3-positive lymphocytes recognized this Tat epitope. However, the frequency of Tat-specific lymphocytes declined precipitously after the acute phase ( Fig. 1). We reasoned that this decline might be the result of viral escape from these Tat-specific responses. We investigated this possibility by sequencing the 5′ exon of Tat using virus derived from the ten Mamu-A*01-positive animals. By 8 weeks after infection, a high frequency of amino-acid substitution was observed in the SL8 epitope (Fig. 2a). Eighty-six per cent of clones contained variation in the CD8-positive T-lymphocyte epitope at this time point. In five out of the ten Mamu-A*01-positive animals, all sequenced clones contained mutations in the Mamu-A*01-restricted SL8 epitope. In contrast, little amino-acid variation was observed outside this SL8 epitope in Mamu-A*01-positive animals (Fig. 2b).

Figure 1: Quantitation of CD8-positive T-lymphocyte responses to various Mamu-A*01-bound peptides.
figure 1

The comparison of CD8-positive T-lymphocyte responses to 6 different epitopes in 10 Mamu-A*01-positive SIV-infected macaques during the first 12 weeks of infection shows that there is a strong CD8-positive T-lymphocyte response to Tat during the acute phase. The Mamu-A*01 Tat28-35 tetramer was initially constructed using an SIVMAC251-derived peptide (TTPESANL). This tetramer detected strong responses during the acute phase of SIVMAC239-infected macaques, even though the corresponding SIVMAC239 sequence was STPESANL. Subsequent staining with the Tat28–35 STPESANL tetramer yielded identical results (data not shown).

Full size image
Figure 2: Variation is present in the SL8 epitope of all 10 Mamu-A*01-positive animals after acute SIV infection using analysis of clones isolated from plasma virus.
figure 2

a, Variation in the SL8 epitope in Mamu-A*01-positive animals infected with molecularly cloned SIVMAC239 Nef stop or SIVMAC239 Nef open. Limited variation is detectable in the inocula. The predicted amino-acid translation of a minimum of 9 clones isolated 6–8 weeks after infection is shown. The frequency of the epitope variant is shown to the right of the sequence. b, Little variation outside the SL8 epitope in the 5′ exon of Tat in Mamu-A*01-positive animals. The entire 5′ exon of Tat in two Mamu-A*01-positive animals is shown. Amino-acid substitutions accumulate primarily in the STPESANL epitope during the first 8 weeks of infection. c, Viral populations in the Mamu-A*01-restricted SL8 epitope evolve rapidly during the acute phase. Predicted amino-acid sequences of virus derived 2, 4, 6 and 8 weeks after infection in animals 96114 and 96118. By 4 weeks after infection, sequence variation was detectable within the epitope. d, Temporal relationships between viral load, tetramer levels and the percentage of wild-type and escaped virus. Values were averaged for animals 96114 and 96118, illustrating that loss of wild-type virus is coincident with the peak of tetramer levels and with the decline in plasma virus concentrations.

Full size image

We then investigated whether these changes in the SL8 epitope resulted from a mixed population of variants in our inocula or whether they were selected for increased replicative fitness in Mamu-A*01-negative animals. As expected from a molecular clone, there was little variation in this epitope in either of the two inocula (Fig. 2a). In addition, only one out of eight Mamu-A*01-negative animals exhibited changes in the SL8 epitope (see Supplementary Information, Fig. 1 ). Thus, viral escape from the Mamu-A*01-restricted Tat-specific CD8-positive T-lymphocyte responses appeared to be the most consistent explanation for our findings.

We then performed a time-course analysis of viral evolution within the SL8 epitope and sequenced the entire virus after the acute phase in two of the Mamu-A*01-positive animals. At peak viraemia (2 weeks after infection), Tat-specific CTL responses were barely detectable and no changes in the Mamu-A*01-bound Tat epitope were present (Fig. 2c, d ). After resolution of peak viraemia (3 weeks after infection), Tat-specific CD8-positive T lymphocytes were at their highest level. One week later, extensive variation was apparent in the virus populations of both animals (Fig. 2c, d). Furthermore, direct sequencing of the open reading frames of the entire virus at 4 weeks after infection revealed only a single site of viral nucleotide diversity in the SL8 epitope in animal 96118. In animal 96114 there were three sites of viral nucleotide diversity, one of which was in the SL8 epitope, and the other two in Rev and Env (see Supplementary Information; Fig. 2). In animal 96118, the nucleotide substitution in RNA encoding the SL8 epitope caused a change in the overlapping reading frame of Vpr. In animal 96114, the change in Rev also caused a substitution in the overlapping open reading frame of Env. This Rev replacement is seen in most animals infected with this viral clone and appears to be selected for increased viral fitness. Analysis of the additional replacement in animal 96114 in Env by interferon-γ enzyme-linked immunospot (ELISPOT) assays of CD8 and CD4 lymphocytes, however, failed to show conclusively that this region contained any T-cell epitopes.

To determine whether the observed sequence changes in the SL8 epitope do represent viral escape variants, we characterized the functional consequences of the predominant variant epitopes on peptide binding to Mamu-A*01 and on CTL recognition. In vitro peptide-binding analyses showed that the new variants of the SL8 epitope did not bind as well as the wild-type peptide to Mamu-A*01 (Table 1). The substitutions of proline at P1 and leucine at P5 reduced peptide binding by more than 50% and 80%, respectively. The isoleucine substitution at P2 and the glutamine, arginine and proline substitutions at P8 abrogated binding (> 99% reduction). As P2 is a secondary anchor and P8 is the carboxy anchor15 for peptides bound by the Mamu-A*01 molecule, we would expect substitutions at anchor residues to have the most profound effect on peptide binding. Similarly, analyses of CTL lines generated from the peripheral blood mononuclear cells (PBMCs) of several Mamu-A*01-positive animals stimulated with the SL8 index peptide poorly recognized the new variant epitopes (Fig. 3 ). Notably, variant peptides with the P5 leucine substitution were the least efficient at sensitizing targets for CTL lysis, suggesting that this P5 mutation was probably interfering with T-cell receptor (TCR) recognition. Therefore, it seems probable that the new variants either reduced the amount of Tat-derived peptide/MHC class I complexes on the cell surface or reduced the ability of these complexes to be recognized by the T-cell receptor16.

Table 1 Peptide binding of mutant Tat epitopes to the rhesus MHC class I molecule Mamu-A*01
Full size table
Figure 3: CTL analyses of CD8-positive T-lymphocytes stimulated with the SL8 peptide.
figure 3

Cell lines were stimulated with the index peptide (SL8) and autologous B-LCL. After 2 weeks in culture, these T-cell lines were used in CTL analyses at an effector:target (E:T) ratio of 25:1 with the index peptide and the variants. Three different dilutions of peptides were tested.

Full size image

To test the hypothesis that viruses with amino-acid replacements within the SL8 epitope are favoured by natural selection, we compared the number of synonymous nucleotide substitutions per synonymous site (dS) and the number of non-synonymous nucleotide substitutions per non-synonymous site (dN) in the epitope and in the remainder of the sequence. In the SL8 epitope region of the virus from Mamu-A*01-positive animals, mean dN was significantly higher than mean dS both for comparisons between samples and the inoculum (dN = 5.7±0.4; dS = 0.4 ± 0.4, P < 0.001) and in comparisons within samples (dN = 7.3±1.0; dS = 0.7 ± 0.7, P < 0.001; see Supplementary Information , Table 1). Mean dN values in the SL8 epitope from Mamu-A*01-positive animals were almost 60-fold greater than the corresponding values for Mamu-A*01-negative animals (dN = 5.7 in Mamu-A*01-positive animals; dN = 0.1 in Mamu-A*01-negative animals). As a pattern of dN > dS is not expected under neutral evolution4,17,18, this result strongly implies that amino-acid replacements in the SL8 epitope are favoured by positive darwinian selection.

Notably, the 5′ exon of four of the eight Mamu-A*01-negative animals showed patterns of variation suggestive of escape from other Tat-specific cellular immune responses (see Supplementary Information, Figs 1 and 3). We therefore explored the possibility that animals with little evidence for selection in Tat should have higher plasma virus concentrations than animals with evidence for increased dN in Tat. As the 18 animals in our cohort were originally part of a vaccine study19, we excluded the 8 Mamu-A*01-positive animals that had been vaccinated from this analysis. The two naive Mamu-A*01-positive and four of the naive Mamu-A*01-negative animals exhibited evidence of increased dN peaks within the 5′ exon of Tat, whereas four Mamu-A*01-negative naive animals revealed little evidence of increased d Nin Tat (see Supplementary Information, Fig. 4). Averaging the plasma virus concentrations of these two groups of animals showed a significant difference of at least one log (P = 0.008) between the plasma virus concentrations of animals with high and low d N in Tat at 2 and 4 weeks after peak viraemia (see Supplementary Information, Fig. 5). Similarly, a significant inverse correlation was observed between peak dN and viral load 2 weeks (P = 0.007), 4 weeks (P = 0.008) and 8 weeks (P = 0.048) after peak viraemia ( Fig. 4; and data not shown). Of the four animals with low dN in Tat, two rapidly progressed to simian AIDS and had SIV plasma virus concentrations in excess of 100 × 106copies per ml within 6 months of infection. Therefore, animals with evidence of increased dN in Tat may have controlled wild-type virus better than those with less selective pressure on Tat.

Figure 4: Inverse correlation between plasma viral concentrations and peak dN in Tat.
figure 4

Peak dN in Tat was plotted against plasma viral concentrations at peak viraemia and at 2 weeks after peak viraemia, revealing a significant inverse correlation between peak dN and viral concentration at 2 weeks after peak viraemia.

Full size image

Vaccine-induced cellular immune responses against proteins expressed early in the viral life cycle may be better able to control HIV and SIV replication than responses directed against proteins that are expressed later in the viral life cycle. Viral escape from Tat-specific CD8-positive T lymphocytes occurred with kinetics similar to those seen during the emergence of drug-resistant mutants20. In five out of ten Mamu-A*01-positive animals, all clones isolated from plasma at 6–8 weeks after infection contained mutations in the Mamu-A*01-restricted SL8 epitope. This implies that Tat-specific CD8-positive T lymphocytes efficiently controlled replication of the original wild-type inoculum virus in these five animals. Responses directed against early proteins such as Tat may be particularly effective at controlling initial virus replication, as Tat and Rev are the only two viral proteins produced before Nef downregulates MHC class I molecules21. Tat-specific CTLs may, therefore, be potent inhibitors of early viral replication, whereas CTLs directed against peptides derived from other viral proteins may find few MHC class I/peptide complexes on the cell surface later in the course of the viral life cycle. The differences between the Gag and Tat-specific CTLs in their ability to exert selective pressure favouring viral escape are intriguing. Understanding the qualitative differences between these CTLs that account for these characteristics will be an important issue in the design of an effective HIV vaccine. Interestingly, vaccination of non-human primates with either Tat protein22,23 or recombinant viruses expressing Tat and Rev24 have reduced virus replication. In these studies it is possible that Tat-specific CD8-positive T-lymphocyte responses were involved in control of viral replication.

Methods

Tetramer analysis

Soluble tetrameric Mamu-A*01 MHC class I/SIV peptide complexes were constructed as described14,25. Background tetramer staining of fresh, unstimulated PBMCs from naive Mamu-A*01-positive animals was routinely less than 0.08%.

Amplification of viral RNA from plasma and sequence detection

We obtained SIV plasma virus sequence as described7. The primers used to amplify complementary DNA encoding the Mamu-A*01 Tat epitope included SIV 6511-F (5′-TGATCCTCGCTTGCTAACTG-3′) and 6900-R (5′-AGCAAGATGGCGATAAGCAG-3′). These primers were then used to isolate and sequence the cloned inserts. Seven overlapping PCR primer pairs (see Supplementary Information , Fig. 2b) were used to amplify cDNA spanning the entire SIV genome. The PCR products were directly sequenced from both cDNA strands. Overlapping sequence between the primers linked together sequences from the individual RT–PCR reactions. Sequence editing and finishing was performed using Auto Assembler v2.1 on a Macintosh. Nucleotide and predicted amino-acid sequences were aligned using MacVector 4.1 (Oxford Molecular).

Mamu-A*01 binding assay

We carried out quantitative assays for the binding of peptides to soluble Mamu-A*01 molecules on the basis of the inhibition of binding of a radiolabelled standard probe peptide to detergent-solubilized MHC molecules15. We used a position-1 C→A mutant of the SIV Gag 181–190 peptide (ATPYDINQML) as the radiolabelled probe. In the case of competitive assays, the concentration of peptide yielding 50% inhibition of the binding of the radiolabelled probe peptide was calculated. We initially tested peptides at one or two high doses. The half-maximal inhibitory concentration (IC50) of peptides yielding positive inhibition were then determined in subsequent experiments, in which 2–6 further dilutions were tested, as necessary. Because, under the conditions we used, where [label] < [MHC] and IC50 ≈ [MHC], the measured IC50 values are reasonable approximations of the true Kd values. Each competitor peptide was tested in 2–4 completely independent experiments. As a positive control, in each experiment we tested the unlabelled version of the radiolabelled probe and measured its IC50.

Generation of in vitro cultured CTL effector cells

We established CTL cultures from peripheral blood samples of SIV-infected rhesus macaques drawn in EDTA tubes, and cultured and assayed CTLs as described7.

Animals, viruses and infections

Rhesus macaques used in this study were identified as Mamu-A*01+ by PCR with sequence-specific primers (SSP) and direct sequencing as described26. All rhesus macaques used in this study were Mamu-A*01-positive, with the exception of animals 95003, 95112, 96020, 96081, 96093, 96072, 96104 and 96113. Rhesus macaques 95045, 96031, 95058, 96118, 96123, 95061, 96114 and 94004 were vaccinated with a DNA/Modified Vaccinia Ankara (MVA) regimen expressing the Gag181–189 peptide (CTPYDINQM)19. The Mamu-A*01-positive macaques 95114 and 95115 were not vaccinated before challenge. All rhesus macaques were infected intrarectally with a molecularly cloned virus; SIVMAC239 (ref. 27) either Nef stop (95045, 96031, 95058, 95114, 95115, 95003 and 95112) or Nef open (99118, 96123, 95061, 96114, 94004, 96020, 96081, 96093, 96072, 96104 and 96113)). Plasma viral concentrations were measured by branched DNA analysis (Chiron). The virus stock was amplified on rhesus PBMCs only. SIV-infected animals were cared for according to an experimental protocol approved by the University of Wisconsin Research Animal Resource Committee.

Statistical analysis of sequence and plasma virus concentration data

Numbers of synonymous nucleotide substitutions per synonymous site (dS) and of non-synonymous nucleotide substitutions per non-synonymous site (dN) were estimated as described28. For the sample of viral sequences taken from a given animal, the means of dS and dN were computed for (1) all pairwise comparisons between each viral sequence and viral sequences sampled from the inoculum; and (2) all pairwise comparisons among viral sequences within the sample. These quantities were computed for the Tat28–35 epitope and the remainder of the 98-codon portion of Tat that was sequenced. To evaluate the statistical significance of the difference in peak and post-peak viraemia between macaques with high and low dN in Tat, we compared the natural log (that is, ln) plasma virus concentrations among animals with high and low dN. Taking ln greatly improved the fit of the data to the assumptions of the statistical models used (that is, normality, homoscedasticity in a multivariate test). We also used multiple analysis of variance (MANOVA) which is dependent on fewer assumptions than the repeated measures ANOVA.