Acute phase viral escape from CTL responses is a universal feature of AIDS virus infection, both in SIV-infected rhesus macaques and HIV-infected humans 
. Descriptions of viral escape from cryptic epitope-specific CTL responses are rare and little is known about whether CTL target cryptic epitopes in the first several weeks of infection and whether these responses drive viral evolution. We initially showed that SIV escapes CTL responses that target a cryptic epitope restricted by Mamu-B*017:01 in Indian Rhesus macaques, the expression of which is correlated with extraordinary control of SIV replication 
. Subsequent reports demonstrated that HIV likewise evolves to evade cryptic epitope-specific CTL responses 
. However, few studies have shown that CTL responses targeting cryptic epitopes within the critical first few weeks of AIDS virus infection can exert sufficient selective pressure on AIDS viruses to drive viral evolution. This report fills that gap and underscores the importance of these responses to the total AIDS virus-specific CTL response in vivo.
ARF-1 is putatively a 69 amino acid polypeptide starting with the first AUG codon and ending with a stop codon and translated from the +1 reading frame relative to the Env-encoding ORF. We previously published two epitopes in ARF-1, one of which was targeted early in infection 
. ARF-10 is putatively a 73 amino acid polypeptide translated from the Env-encoding mRNA, via translation initiation at the Rev AUG start codon and continued translation into the intron 
. For the present and previous study 
, we tested for CTL responses against these polypeptides using overlapping peptides (15-mers overlapping by 11) that encompass the predicted translated portions of the ARFs. For ARF-10, these peptides included the last several amino acids of the first exon of Rev to ensure identification of responses that target epitopes overlapping the exon/intron boundary. We used these peptides in IFN–γ ELISPOT assays to identify T cell responses targeting this polypeptide in Rhesus macaques infected with SIVmac239 (or experimental mutants of SIVmac239) as part of previous studies. Identified responses were mapped to the minimal optimal epitope using serial dilutions of peptides and either PBMC from the responding animal or antigen-specific T cell lines grown against the targeted 15-mer (from r04028, explained in detail in the Results and Discussion and Materials and Methods sections).
In this study, we sought not only to identify epitopes targeted in ARFs -1 and -10 during the acute phase of infection but also to measure CTL-mediated selection pressure on these epitopes. Sequence evolution was identified using deep pyrosequencing of a small, highly targeted PCR amplicon that spanned ARFs -1 and -10, including the first exon of Rev and the first portion of the Env protein, which overlaps both ARFs. The genomic locations of ARFs -1 and -10 and the forward and reverse primers used to amplify the region from SIV are depicted in . The primers used, including Roche adaptor, multiplex identifier (MID) and sequence-specific primer are depicted in . Finally, we conducted dN/dS analyses and compared the kinetics of the accumulation of non-synonymous intra- vs. extra-epitopic changes in ARF-1 and ARF-10 in order to measure CTL-induced selection on the epitopes.
Animals r04028 and rh2261 were recently infected with SIVmac239 experimentally mutated to harbor “pre-escaped” epitopes typically presented by Mamu-B*008:01 as part of an unrelated previous study 
. Three weeks after infection, we used IFN—γ ELISPOT to enumerate SIV-specific T cell responses targeting ARFs -1 and -10. Both animals responded to overlapping peptides in ARF-10 at magnitudes near 1,000 spot forming cells (SFC) per million PBMC (). We next used a panel of peptides ranging in length from 8 to 11 amino acids spanning the targeted 15-mers to map the responses to the minimal optimal epitopes. We used serial dilutions of these peptides and cryopreserved PBMC from animal rh2261 in IFN–γ ELISPOT assays to map the targeted epitope to the QW9 epitope (). PBMC from animal r04028 were not available from an acute phase time point and later time points showed no or very little response (data not shown). Therefore, to map the targeted epitope, we expanded antigen-specific CTL from this animal by exposing PBMC from 12 weeks post infection to irradiated autologous BLCL pulsed with the 15-mer peptide against which the response was originally detected (WESAAYRHLAFKCLW). After weekly stimulation for four weeks, we used this cell line in ELISPOT assays, similar to rh2261, and identified the minimal optimal epitope in this animal as the AF8 epitope (). In the absence of PBMC from an acute time point, the use of a CTL line expanded against the parent 15-mer peptide for a minimum number of in vitro stimulations was the least biased approach to mapping this epitope. The locations of the mapped epitopes within ARF-10 are show in .
Acute phase CTL responses against ARF-10 in two SIV-infected Rhesus macaques.
We next wished to determine if the CTL targeting ARF-10 were sufficiently potent to select for viral escape mutations in vivo. We used primers from to reverse transcribe and amplify replicating SIV isolated from cell-free plasma from four, eight and 12 weeks post-infection in animal r04028 and weeks four, ten and 16 in animal rh2261. We used 454 Life Sciences (Roche) deep pyrosequencing of these amplicons and found clusters of non-synonymous mutations within the targeted epitopes by eight weeks post-infection in virus isolated from both animals (). Viral escape from CTL responses is generally characterized by small numbers of amino acid substitutions (often just one) within a restricted epitope that either reduces binding of the virus-derived peptide to the MHC-I molecule or reduces the affinity of the T cell receptor (TCR) on the responding CTL for the MHC-I: peptide complex. Alternately, mutations can arise outside the epitope that reduce the efficiency with which the epitope itself is liberated from the parent polypeptide by cellular proteasomes and immunoproteasomes. Our pyrosequencing data was restricted to acute time points soon after the initial CTL responses arose and do not demonstrate which mutations resulted in stable escape. Therefore, we next used Sanger sequencing to sequence this portion of the circulating virus from plasma samples taken at the time of euthanasia (81 weeks post infection for r04028 and 106 weeks post infection in rh2261) to determine if viral evolution had coalesced onto a single dominant mutation. In both animals a single amino acid change was identified within the epitopes, likely representing the stable escape mutation.
Viral sequence evolution in overlapping reading frames encoding Env and ARFs -1 and -10.
The clear clustering of non-synonymous mutations within the targeted epitopes, particularly the AF8 epitope targeted by r04028, strongly suggests archetypal viral escape from a CTL response. dN/dS analysis supported this suggestion (p<0.05). Indeed, in both animals, the dramatic drop in the portion of sequence reads matching the inoculum within the targeted epitopes during the acute phase (, solid shapes) compared to the relative constancy of the extra-epitopic portion of ARF-10 is strongly suggestive of viral escape from these CTL responses in vivo.
To test if these mutations detected late in infection reduced or eliminated CTL recognition, we thawed cryopreserved PBMC from animal rh2261 from four weeks post infection and used IFN-γ ELISPOT to measure recognition of serial dilutions of the wild type QW9 peptide versus QW9 harboring a position 8, S to P mutation. Recognition of the mutated peptide was reduced significantly as compared to wild type. However, it was not eliminated (). This data provides further evidence that the mutation was the result of selection by antiviral CTL. It is curious that selection did not favor a mutant in an MHC anchor residue, assuming the virus could do so without fitness consequences. It is possible that this intra-epitope mutation impairs processing of the MHC-bound 9-mer, which would not be evidenced by recognition of peptides but would require introduction of the mutation into the virus and measurement of recognition of infected cells. This situation was described for a cryptic epitope in HIV-1 
. As mentioned above, acute phase PBMC were no longer available from animal r04028 that could be used to compare recognition of the wild type AF8 peptide with the mutated peptide harboring a position 8 F to S mutation (the intra-epitope mutation detected at time of euthanasia). To remedy this, we used an antigen specific T cell line expanded against the 15-mer harboring the AF8 epitope (WESAAYRHLAFKCLW). We found this cell line to have high background for IFN-γ so we instead used intracellular cytokine staining and measured its ability to produce TNF-α in response to serial dilutions of the wild type and the mutant peptide. This cell line responded well to several dilutions of the wild type peptide but very poorly to the mutant peptide (). This is not surprising, given the mutation is in a presumed anchor residue (position 9 of 9).
We note two important observations in our evolutionary analysis of CTL recognition of ARF-10. First, in virus circulating in animal r04028, nearly all mutations within the targeted AF8 epitope that exceeded our threshold of 1% were synonymous in the overlapping Env-encoding ORF, presumably indicating that this short stretch of Env was under pressure to maintain its amino acid sequence. Finally, we note that viral evolution in the QW9 epitope targeted by rh2261 was entirely clustered in the carboxyl end of the epitope. We hypothesize that selection might have favored variants that escaped the CTL response while maintaining the amino acid sequence of the first exon of Rev and the RNA sequence surrounding the Rev splice donor site, the disruption of which would likely incur a substantial cost to viral fitness.
In our previous study 
, we showed that several animals made high-frequency CTL responses against the RP9 epitope in ARF-1. In this report, we show a comprehensive examination of the CTL responses against this epitope in two animals and the kinetics of the subsequent viral escape from this response. Animal r97035 was infected with SIVmac239 engineered to harbor escape mutations in epitopes in Gag, Tat and Nef 
. This animal was a long-term non-progressor and survived with low but detectable viremia for more than four years after initial SIV infection. Animal r97111 was infected with wild type SIVmac239 as part of an unpublished previous study. Each animal responded to overlapping 15-mer peptides in ARF-1 (). The responses were fine-mapped to the same RP9 epitope using serial 10-fold dilutions of peptides representing truncations of the targeted 15-mer in IFN–γ ELISPOT assays (). The position of the RP9 epitope within the ARF-1 polypeptide is shown in . In both animals, nearly identical functional avidities are shown between one of the targeted 15-mers and the RP9 peptide. Presumably, this is due to highly efficient liberation of the RP9 9-mer from this peptide for unknown reasons. PBMC were not available from the same acute time points in which the responses were first detected and, hence, the magnitudes of the responses in both animals are lower in the epitope mapping ELISPOTS than they are in the initial ELISPOTS.
Acute phase CTL responses against ARF-1 in two SIV-infected Rhesus macaques.
We next tested whether selection favored variants of RP9 that evaded the RP9-specific CTL response. To accomplish this, we used the same focused pyrosequencing approach we used with ARF-10. Specifically, we sequenced virus extracted at acute phase time points to determine whether viral escape occurs in this epitope and how long after infection escape becomes detectable. SIV from both animals harbored clusters of mutations in this epitope as early as 3 and 3.7 weeks post infection. By 16 weeks post-infection in r97111, there was a clear pattern favoring the position two Q to R mutation. Indeed, when we used Sanger sequencing at time points much later in infection (110 weeks p.i. in r97111 and more than 4 years p.i. in r97035), we detected the same single mutation in both animals at position two of the epitope (). Position two of an MHC-I restricted epitope is nearly always important for anchoring peptides to the MHC-I molecules. Hence, this escape pattern would indicate that the virus evolves in convergent fashion in these animals to escape the RP9-specific CTL response. Patterns of acute viral escape in these animals was similar to that in the AF8 epitope in r04028 in that a cluster of mutations appears early in infection that is almost entirely synonymous in the overlapping Env ORF. We interpret these data to mean that there is pressure to maintain the Env amino acid sequence in this small region. It should also be noted that the patterns substitutions could suggest immune targeting of the region of ARF-10 overlapping the RP9 epitope within ARF-1. However, we did not detect T cell activity against ARF-10 in these animals (data not shown).
Viral sequence evolution in overlapping reading frames encoding Env and ARFs -1 and -10.
Finally, we assessed CTL-induced selection in the RP9 epitope in SIV from each animal. As with the ARF-10 restricted epitopes, dN/dS analysis was significant (p<0.05) and we also observed dramatic accumulations of non-synonymous mutations within the epitope, represented by loss of inoculum sequence (, closed symbols). In contrast, the extra-epitopic portions of ARF-1 were largely maintained as wild type through acute phase sampling (, open symbols), and even at much later time points as measured by Sanger sequencing. We next thawed acute phase (week 4 post infection) PBMC from animal r97111 to compare recognition of the wild type RP9 epitope to the mutant peptide. We detected strong recognition of the wild type peptide over several orders of magnitude of peptide concentrations but no recognition of the mutant peptide (). Animal r97035 was infected several years prior to this study and acute phase CTL were no longer available to complete a comparable experiment. However, since the responses mapped to the same minimal epitope and viral evolution was highly convergent in each animal, these data provided strong evidence that the virus evolved as a direct result of the CTL targeting the RP9 epitope in both animals.
The challenge viruses used to inoculate several of the animals used in this study contained mutations in established T cell epitopes in order to test hypotheses distinct from those tested here. It is unknown whether this affected the kinetics of the ARF-1 and ARF-10 specific CTL responses and subsequent viral escape. In the case of ARF-1, this appears not to be the case. Animal r97035 was infected with a version of SIVmac239 containing escape mutations in three epitopes 
, while animal r97111 was infected with wild type SIVmac239, and the kinetics of the ARF-1 RP9 targeted CTL response and viral escape were nearly identical in both animals. Hence, cryptic epitope-specific CTL responses are not inherently subdominant. The RP9-specific CTL response appears to be an immunodominant acute phase CTL response, irrespective of epitope sequences in the challenge virus outside this epitope.
Taken together, our data substantiate an important role for CTL against cryptic epitopes during AIDS virus infection. Viral escape is an important measure of in vivo CTL efficacy. Hence, CTL against cryptic epitopes could be important components of the total AIDS virus-specific responses and their role in vaccine modalities should be investigated. Our data also suggest that ARFs -1 and -10 appear to be translated as a normal function of SIV replication, despite being favorite targets of host T cell responses. It may prove informative to investigate the biochemical properties of these novel polypeptides in greater detail.