HIV-1-Specific CD8⁺ T Cell Responses and Viral Evolution in Women and Infants¹

The CD8⁺ T cell response plays an important role in controlling HIV-1 or SIV replication and subsequent disease progression. In primary macaque SIV or adult human HIV-1 infection, the detection of HIV-1- and SIV-specific CD8⁺ T cell responses correlates temporally with the early decline in plasma RNA following peak viremia (1, 2, 3, 4, 5, 6) and the generation of variant viral sequences suggestive of potent in vivo selective pressures (7, 8). Furthermore, in the macaque model of SIV infection, CD8⁺ T cell depletion results in impaired control of viral replication (4, 5).

HIV-1-infected individuals harbor many genetically related variants referred to as quasispecies. Quasispecies have been defined as a “dynamic distribution of nonidentical but closely related mutant and recombinant genomes subjected to a continuous process of genetic variation, competition and selection, and which act as a unit of selection” (9). These dynamic distributions are due mainly to viral and host factors such as an error-prone reverse transcriptase, high viral replication rates, and diverse host selective pressures (including CD8⁺ T cell responses) (10, 11). The generation of these variant viral sequences may also result in diminished CD8⁺ T cell recognition. CD8⁺ T cell escape has thus been proposed as a mechanism for CD8⁺ T cell failure in acute and chronic phases of HIV-1 and SIV infections (7, 12, 13, 14).

We have previously reported mother-to-child transmission (MTCT)³ of antiretroviral therapy (ART)-resistant viruses (15). The detection of ART resistance mutations in infants was associated with only transient responses to therapy. MTCT of CD8⁺ T cell escape variants may also have important consequences. Because infants share at least three HLA class I alleles with their mothers, MTCT of CD8⁺ T cell escape variants may compromise the infant’s ability to mount CD8⁺ T cell responses restricted by shared HLA alleles, resulting in reduced control of viral replication. MTCT of CD8⁺ T cell escape variants have been documented (16), but it is unclear as to how commonly this occurs (17). Although we and others have reported previously the detection of virus-specific CD8⁺ T cells in the peripheral blood of young infants, the extent to which early infant CD8⁺ T cell responses are functional in vivo is unknown.

Therefore, we undertook this study to better understand the potential selective pressures exerted by CD8⁺ T cells on viral sequences in HIV-1-infected women and their infants. Maternal plasma sequences were analyzed at delivery along with early infant viral sequences to detect the presence of CD8⁺ T cell variants in maternal plasma and transmission of these variants to their infants. Serial plasma samples obtained from infants over the first year of life were also studied to determine the fate of transmitted variants and the potential selective pressures exerted by infant CD8⁺ T cell responses in early infection.

Five HIV-1 infected mother-infant pairs were studied. Diagnostic studies (DNA PCR, viral isolation) in all but one infant (P-1115) were consistent with HIV-1 infection during the intrapartum period (18). P-1115 was considered infected in utero following viral isolation from cord blood and from peripheral blood at 1 day of age. RNA was extracted from maternal plasma at delivery and from infant plasma obtained at three different time points (1–3, 3–6, and 11–15 mo of age). This study examined plasma samples obtained before the routine use of ART to interrupt mother-to child HIV-1 transmission, and none of the mothers was on ART at delivery. Characteristics of the mother-infant pairs studied are shown in Table I. The Human Studies Committee at the University of Massachusetts Medical School approved these studies and informed consent was obtained for participation.

Viral RNA was extracted from 200 μl of plasma using High Pure Viral RNA kit (Roche Diagnostics). We focused on gag and nef genes because these are highly immunogenic regions of HIV-1 (19). cDNA was generated using random hexamers or a reverse primer specific for the 5′ end of pol RTRD: 5′-CTGTCCACCATGCTTCCC-3′, complementary to positions 3742–3759 according to HXB2 numbering (〈http://hiv-web.lanl.gov〉) to allow for complete amplification of gag and nef genes (GeneAmp RNA PCR kit; Applied Biosystems). After cDNA synthesis, nested PCR were used to amplify both genes (HotStarTaq DNA polymerase; Qiagen). For the first amplification, gag primers were as follows: FGF60, 5′-CAGACCCTTTTAGTCAGTGTGGAAAATC-3′, positions 600–627; and RTRA, 5′-GTTGACTCAGATTGGTTGCA-3′ complementary to positions 2519–2538). Primers for the nef gene were as follows: 587, 5′-AATCTCCTACAGTATTGGAGTCAG-3′, positions 8616–8639; and NEF OA, 5′-GCCCAGGCCACGCCTCCCTGG-3′, complementary to positions 372–392. For the second amplification, gag primers were as follows: GAG000; 5′-GACTAGCGGAGGCTAGAAG-3′, positions 764–782; and GAG G10, 5′-TACTGTATCATCTGCTCCTGTATC-3′, complementary to positions 2325–2348. Second round primers for nef gene were as follows: NEF IS, 5′-CCACATATCTAGAAGAATAAGACAGGG-3′, complementary positions 8747–8773; and NEF IA, 5′-AGTCCCCCGCGGAAAGTCCCTTGTAGC-3′, complementary to positions 343–369. Following verification of specific amplification by gel electrophoresis, gag and nef amplicons were ligated into the PCR4 TOPO vector and transformed into TOP 10 cells (TOPO TA Cloning kit; Invitrogen Life Technologies). Clones were selected after overnight incubation on kanamycin-infused Luria-Bertani plates. Plasmid DNA were amplified overnight in kanamycin-infused slant cultures and isolated using the QIAprep Spin Miniprep kit (Qiagen). Positive clones were determined by EcoRI restriction digest analysis. Ten clones per gene were sequenced at each time point for each patient using universal primers T3 and T7 (Davis Sequencing).

The nucleotide sequences from the subtype B reference strain HXB2 and from different subtypes were downloaded from the Los Alamos database (〈http://hiv-web.lanl.gov〉). Sequences were aligned with the Clustal X program (20) and later edited by hand. All positions with alignments gaps in at least one sequence were excluded from further analysis. The number of different bases in pairwise comparisons and phylogenetic analysis were determined by the Molecular Evolutionary Genetic Analysis (MEGA) program (21). DNA distance matrices were calculated by the Kimura two-parameter method. Phylogenetic trees were constructed with distance matrices, using the neighbor-joining method as implemented in MEGA. The robustness of each tree was evaluated by bootstrap analysis of 1000 replicas. The dN and dS values were calculated according to the Suzuki and Gojobori method (22) as the average of values obtained at each amino acid position.

Shannon entropy values were calculated as another measure of variability. The normalized Shannon entropy was calculated as -Σ_i ((p_i × lnp_i)/lnN)) in which p_i is the frequency of each sequence in the quasispecies and N is the total number of sequences compared. A Shannon entropy (23) value of 1 means that all genomes in the quasispecies distribution differ in at least one amino acid (highly heterogeneous population), whereas a value of 0 means that all genomes are identical (highly homogeneous population).

PBMC were isolated from whole blood using the Ficoll-Paque (Amersham Biosciences) density centrifugation method and were viably cryopreserved in RPMI 1640 containing 10% DMSO. PBMC used for ELISPOT assays were thawed and washed twice in RPMI 1640 supplemented with 10% FBS, 25 mM HEPES, and 10 mg/L gentamicin (R10 medium) before counting. Autologous B-LCL were generated for each infant by transformation with EBV and maintained in R10 medium.

Genomic DNA was extracted from 4 to 5 million PBMC or B-LCL using the QIAamp DNA minikit (Qiagen). HLA class I A, B, and C typing was performed using the Biotest ABC SSPtray (Biotest Diagnostics).

Peptides for use in the study were synthesized by Genemed Synthesis. All peptides were HPLC purified with a minimum purity of 95% and analyzed by mass spectrometry.

HIV-1-specific CD8⁺ T cell responses were quantified by ELISPOT assay as described previously. Before the addition of cells, a 96-well flat-bottom plate (MAIPN1450; Millipore) was coated with 1 mg/ml anti-IFN-γ mAb (D1K; Mabtech) and allowed to sit overnight at 4°C. After washing the plate with cold PBS, each plate was blocked for nonspecific Ab binding by the addition of 200 μl/well R10 medium for 2 h at 37°C. PBMC (10⁵) from each time point were added to each well in triplicate per study subject. In infant plates, 10⁵ autologous B-LCL previously incubated with peptides (10 μg/ml) were added to appropriate wells to bring the total volume in each well to 100 μl. This constituted an E:T ratio (PBMC:B-LCL) of 1:1. To ensure assay consistency, all study time points per study subject were assayed on the same ELISPOT plate. Each plate was incubated for 16–20 h overnight at 37°C in 5% CO₂. After overnight incubation, the plate was washed with cold PBS, and 0.5 mg/ml of a secondary, biotinylated anti-IFN-γ mAb (7-B6-1; Mabtech) was added for 3 h at room temperature. The plate was washed again with cold PBS, and 0.5 mg/ml streptavidin-alkaline phosphatase conjugate (Mabtech) was added for 2 h at room temperature. Spot-forming cells (SFC) were visualized after alkaline phosphatase color development (Bio-Rad).

SFC were counted using an automatic plate reader (ImmunoSpot Software Version 3; Cellular Technology) and were expressed as SFC/10⁶ PBMC. Two criteria were used to define positive responses. First, each well (from triplicate wells) must have had a minimum of 6 SFC (corresponding to 60 SFC/10⁶ PBMC). Second, the experimental SFC frequency must have exceeded the mean of all triplicate negative control wells (PBMC or PBMC + B-LCL without peptide) by 2 SDs. PHA stimulation of PMBC samples from all study time points served as positive controls for IFN-γ release.

The wild type HIV-1 proviral DNA used was plasmid pNL43. A 1.9-kb fragment was amplified with primers FGF60 and RTRA. This amplicon was cut with BssHII-ApaI restriction enzymes, and a 1.3-kb fragment was obtained. This fragment was subcloned into the PCR4 TOPO vector as a target plasmid for mutagenesis. The mutation found in a B*4002-restricted epitope (E17K (p17)) was performed using mutation primers 5′-GGGGAGAATTAGATAAATGGAAAAAAATTCGGTTAAGGCCAGG-3′ and 5′-CCTGGCCTTAACCGAATTTTTTTCCATTTATCTAATTCTCCCC-3′ and QuikChange site-directed mutagenesis kit (Stratagene) according to the manufacturer’s instructions. The presence of the desired mutation was confirmed by sequencing the whole fragment inserted. To generate the mutated HIV-1 proviral DNA clone, the BssHII-ApaI fragment, carrying the E17K mutation, was reinserted into the parental pNL43 plasmid. The mutated pNL43 plasmid has been designated as pNL43p17E17K.

Hela and MT-4 cells were maintained in DMEM with 10% FBS and R10 medium, respectively. Hela cells (3 × 10⁶) were seeded into 75-cm² tissue culture flasks 24 h before transfection. HeLa cultures were transfected by the calcium phosphate precipitation technique with 20 μg of each proviral plasmid DNA (pNL43 and pNL43p17E17K). After 48 h posttransfection, supernatants were collected, passed through 0.45-μm pore size filters, and quantified viral production by p24 Ag detection assay. Equivalent p24 amounts (10 ng/ml) of pNL43 and pNL43p17E17K were used to infect 4 × 10⁶ fresh MT-4 cells. Viral replication was monitored by p24 Ag detection.

The nucleotide sequences that we obtained from each patient have been submitted to GenBank with accession numbers AY786790 to AY786979 for the gag clones and AY786591 to AY786789 for the nef clones.

Several investigators have reported limited diversity of HIV-1 envelope (env) third variable region gene sequences of viruses obtained in early pediatric HIV-1 infection (24, 25, 26, 27, 28, 29) and have proposed the transmission or posttransmission selection of single viral strains. However, other investigators have reported heterogeneity of other regions of HIV-1 env or gag p17, suggesting the transmission of multiple variants (30, 31, 32). Therefore, we thought it important to evaluate the heterogeneity of maternal and early infant HIV-1 gag and nef sequences.

A total of 20 samples from five mother-infant pairs were studied (Table I). The nucleotide sequences from 190 clones were obtained (10 clones/sample; the gag gene in infant P-1026 could not be amplified at the second time point). Phylogenetic trees constructed using mother-infant sequences and subtype reference sequences (〈http://hiv-web.lanl.gov〉) showed that all mother-infant sequences clustered in the same clade as subtype B reference sequences. All sequences in both genes from each mother-infant pair formed distinct clusters with bootstrap values of 100% (data not shown).

Intrapatient genetic distances were calculated using the MEGA program. Mean maternal intrapatient genetic distances ranged from 0.2 to 1.6% of nucleotides for gag and 0.2 to 1.7% for nef (Table II). At the first time point (1–3 mo of age), the mean intrapatient genetic distances for infant sequences ranged from 0.6 to 1.5% for gag and 0.2 to 1.4% for nef. Between the first and second time point (median elapsed time = 2 mo), mean intrapatient genetic distances increased by 1–1.5% for gag and 0.4–1.2% for nef. Additional diversification of infant gag and nef sequences were observed between the second and third time points (0.8–1.5% for gag and 0.9–1.7% for nef; median elapsed time = 7 mo).

To further examine the diversity of maternal and early infant viral variants, amino acid sequences from each mother at delivery and each infant at the first time point (1–3 mo of age) were aligned, and the alignments were used to calculate individual Shannon entropy values (Table II). The Shannon entropy values for all maternal gag sequences at delivery were 1.0, suggesting a high degree of heterogeneity of maternal gag sequences. By contrast, the Shannon entropy values for maternal nef sequences at delivery were more variable and ranged from 0.53 to 1. Shannon entropy values for first infant gag sequences (1–3 mo of age) were also uniformly at or close to 1, indicating heterogeneity of the sequences. First infant nef sequence Shannon entropy values ranged from 0.85 to 1, suggesting a high degree of heterogeneity of these early viral sequences.

Altogether, the analyses of the intrapatient genetic distances and the Shannon entropy values document a high degree of heterogeneity of early infant viral sequences. These data suggest that multiple viral strains are commonly transmitted from mothers to their infants. Alternatively, early viral replication may be subject to sufficient selective pressures to drive rapid evolution of the viral quasispecies (33). Analyses of additional early samples obtained at more closely spaced intervals may help to distinguish between these two possibilities.

We were interested next in determining whether CD8⁺ T cell epitope variants could be detected in maternal plasma viral sequences at delivery. To identify potential CD8⁺ T cell escape variants, pairwise comparisons of HXB2 and maternal plasma amino acid sequences were performed. We used the HXB2 reference sequence because the majority of optimal CD8⁺ T cell epitopes described in the HIV-1 Immunology Database (〈www.hiv.lanl.gov〉) are based on this sequence. Several amino acid differences between maternal and HXB2 sequences were detected (Table III). To select amino acid changes possibly due to CD8⁺ T cell selective pressures, we established conservative criteria by focusing on variants within epitopes restricted by maternal HLA alleles that were present in >30% of the maternal quasispecies and that were represented in <10% in the alignments from all subtypes and recombinant forms available at the HIV-1 Sequence Database (〈www.hiv.lanl.gov〉). These criteria minimize the likelihood that the changes occurred through chance alone but may underestimate CD8⁺ T cell escape variants that have become fixed in the population (11, 34)

A total of 17possible (9 in gag and 8 in nef) CD8⁺ T cell escape variants was identified (Table III). As expected, several variants were found within CD8⁺ T cell epitopes restricted by maternal HLA alleles (e.g., Figs. 1, top panel, and 2). Of interest was that potential CD8⁺ T cell variants were also found within CD8⁺ T cell epitopes restricted by HLA alleles commonly represented in our clinical populations (HLA A*2, A*24, and B*7) but not shared by the mother (Fig. 1, lower panels). For example, a variant of a HLA A*2-restricted epitope in nef (aa 83–91; AAVDMSHFL) was detected in the plasma of a HLA A*2⁻ woman (M-1002), and a variant of a HLA A*24-restricted epitope in nef (aa 186–194) was detected in an HLA A*24-negative woman (M-1001).

Using ELISPOT assays, we next investigated whether maternal and infant CD8⁺ T cells recognized the transmitted epitope variants. Because of the limited numbers of PBMC available, nine epitope variants restricted by five different HLA alleles were tested: A*2-AL9 (AAVDMSHFL, nef 83–91), A*2-GL9 (GALDLSHFL, nef 83–91), A*24-DM9 (DSRLAFQHM, nef 186–194), A*24-DK9 (DSTLAFQHK, nef 186–194), A*24-QW9 (QYKLKHIVW, gag 28–36), A*30-RY10a (RLRPGGKKQY, gag 20–29), A*30-RY10b (RLRPGGKKRY, gag 20–29), B*7-RW9 (RPMTHQAAW, nef 77–85), and B*40-GI9 (GELDRWKKI, gag 11–19); throughout text, underlined letters in sequences denote amino acid substitutions (mutations) in the peptide (see also footnote a in Table IV). Five of the nine epitope variants (AL9, DM9, DK9, QW9, and GI9) were recognized at lower responder cell frequencies compared with wild type (Fig. 1). An A*24⁺ mother (M-1003) recognized a HLA A*24-restricted gag epitope but did not recognize the epitope variant. Her HLA A*24⁺ infant (P-1189) did not recognize either the wild-type sequence or the variant. Variants of a HLA A*24-restricted nef epitope (aa 186–194; AVDLSHFLK) were detected in the plasma of two mothers (M-1003, HLA A*24⁺ and M-1001, HLA A*24⁻). CD8⁺ T cell responses to the wild-type and variant sequences were not detected in PBMC from these women. Responses to the wild-type and variant sequences were detected in infant P-1189 (HLA A*24⁺) at 2 mo of age, but the frequency of responses to the variant sequence was ∼3-fold less than the response to the wild-type sequence. Infant P-1024 responded to both the wild-type and the variant sequences at 4 mo of age; the responder cell frequency to the variant sequence was 65% of the responder cell frequency to the wild-type sequence. At 15 mo of age, P-1024 did not recognize either the wild-type or the variant epitope, and reversion of sequences to wild type was noted. Additional PBMC were not available from these infants to do additional assays (including peptide titrations) to confirm reduced recognition of the variant sequences. To confirm that DK9 and DM9 epitope variants were escape mutants, a peptide titration using PBMC from another HLA A*24⁺ individual was performed. Markedly reduced responder cell frequencies to both variants were detected compared with wild type (Fig. 3). Of the four remaining variants, two epitopes (RY10a and GL9) were not recognized on ELISPOT assays using maternal PBMC, and two variants (RW9 and RY10b) were recognized as well as wild-type epitopes by the B*7 and A*30⁺ infants (data not shown).

To determine whether CD8⁺ T cell epitope variants found in maternal sequences were transmitted to the infants, infant plasma viral sequences from the first time point (1–3 mo of age) were compared with their maternal viral sequences at delivery (Table IV). Subsequent comparisons with infant viral sequences from the second (3–6 mo of age) and the third time points (11–15 mo of age) were done to analyze fixation over time (Table III).

In all but one case, variants similar to those found in maternal plasma samples were detected in the first infant plasma specimen. When mothers and infants shared the restricting HLA alleles, these variant amino acid sequences were usually found at subsequent time points (e.g., QW9 and DK9; Table IV). Even when mothers and infants did not share the restricting HLA allele, most variant sequences were maintained over time (e.g., the Q28K substitution in an HLA A*30-restricted gag p17 epitope in P-1189 or the P14N in an HLA B*8-restricted nef epitope in P-1024; Table III). Because many of the persistent variants represented single nucleotide substitutions, their persistence likely indicates that these mutations do not significantly limit viral fitness. Alternatively, the lack of reversion may have been due to CD8⁺ T cell-mediated selective pressure restricted by an infant HLA allele not shared with the mother; however, we did not find evidence to support this.

Amino acid substitutions were also detected within optimal epitopes restricted by HLA alleles commonly represented in our clinical populations (HLA A*2, A*24, B*7) but not shared by the mother. These variant sequences were also passed to the infants; when the infant had the restricting HLA allele, these variants were often detected in the infants’ subsequent plasma samples. For example, a variant of a HLA A*2-restricted epitope in nef (aa 83–91; AAVDMSHFL) was detected in a HLA A*2- mother (M-1002; Fig. 1); the same amino acid substitution was found in plasma samples from her HLA A*2⁺ infant (P-1031) at 2, 4, and 11 mo of age. Reduced recognition of the variant sequences was noted using PBMC obtained from the infant at 4 mo of age (50% of wild type) but variant sequences were not recognized at 11 mo of age.

Only one documented maternal escape variant was not represented in the infant’s first sequences. This variant was a rare amino acid substitution E17K in a B*40-restricted gag p17 (aa 11–19; GI9) epitope that was present in all 10 maternal viral clones at delivery (Fig. 2,A). ELISPOT assays performed using maternal PBMC revealed recognition of the wild-type but not the variant sequence (Fig. 2,B). Of interest, is that this amino acid substitution was not detected in any plasma samples obtained from her infant (at 2, 4, and 11 mo of age). Lack of detection of the virus in the infant’s plasma despite common detection of this mutation in the maternal plasma suggests that either a minor wild-type variant was transmitted or that a variant was transmitted but rapidly reverted in the absence of continued CD8⁺ T cell-selective pressure. This epitope was noted to lie in the α helical region-1 of p17, which is known to be important for viral replication in culture (35). Thus, site-directed mutagenesis was performed to introduce the E17K substitution into a pNL4-3 background. HeLa cultures were then transfected with each proviral plasmid DNA (pNL4-3 and pNL43P17E17K), viruses were collected, and then used to infect MT-4 cells. After infection with equivalent amounts of p24 Ag (10 ng/ml), the mutant virus showed markedly lower levels of replication compared with the wild-type virus, confirming that the E17K substitution adversely affected viral replication (Fig. 2 C).

Although we have previously reported the detection of HIV-1-specific CD8⁺ T cell responses in young infants (36, 37, 38), an important question is whether these early virus-specific CD8⁺ T cell responses are active in vivo. We and others have previously demonstrated that the analysis of viral sequences over time can be used to detect and measure the effects of selective pressures exerted by antiretroviral agents or immune responses in vivo. As indicated by the phylogenetic analyses above, infant viral sequences appeared to diversify over time. To evaluate the potential contribution of CD8⁺ T cell-selective pressures in vivo, we calculated the ratio of synonymous to nonsynonymous amino acid substitutions (dN/dS) in the gag and nef genes over time. Pairwise comparison between first time point consensus sequences and sequences from the second time point revealed a dN:dS ratio > 1 in the nef gene sequences from four infants (Table V). Nonsynonymous mutations were distributed throughout nef (Fig. 4).

Similar pairwise comparisons of gag p55 gene sequences revealed dN/dS > 1 in only one of four infants. Generally, dN/dS tends to be <1 (purifying selection) when a large gene is analyzed globally because of the requirement to maintain protein function. Silent (synonymous) nucleotide changes usually predominate among the surviving variant sequences. Given the possibility that analysis of the whole gene may have obscured selective pressures on individual proteins within the gag gene, we analyzed the dN/dS ratio in gag p17, p24, and p15 (p2p7p6p1). Analysis of the number of the nonsynonymous mutations compared with length of each protein over gag confirmed that the mutations clustered mainly in p15 (p = 0.001, χ² test; Fig. 4). A significant relation was observed between accumulation of dN/dS values > 1 (positive selection) in the nef gene compared with dN/dS values > 1 in the gag p55 gene (p = 1.3 × 10 ⁻⁶; χ² test). A significant relation was also observed between accumulations of dN/dS values > 1 in p15 compared with p17 and p24 together (p = 0.0032; χ² test).

Analysis of sequences within potential epitopes restricted by infant HLA alleles revealed evidence of CD8⁺ T cell selective pressures as early as 2–3 mo of age. In a HLA A*24⁺ infant (P-1026), a Y135F amino acid substitution (resulting from a single nucleotide substitution TAT to TTT) was detected in an A*24-restricted nef epitope (aa 134–143; RYPLTFGWCF). Epitope variant RFPLTFGWCF (mutation Y135F) was present as early as 3 mo of age and at 13 mo of age was the predominant variant (Fig. 5). This mutation was not detected in the HLA A*24-negative mother’s plasma at delivery. This variant has been described as a common escape mutant in HLA A*24⁺ hemophiliacs (39). Of interest, is that recognition of A*24⁺CD8⁺ T cells pulsed with variant RFPLTFGWCF peptide was equivalent to recognition of the unaltered sequence. However, this mutation resulted in decreased recognition of cells expressing the Nef variant by CD8⁺ T cell clones established from two A24⁺ patients. The latter was associated with reduced expression of native Nef proteins by Western blot analysis, likely due to a disruption of the normal processing of native Nef proteins. Similarly, we detected diminished recognition of A*24⁺CD8⁺ T cells pulsed with variant RFPLTFGWCF peptide only at a very high molar peptide concentration 10⁻⁴ (data not shown), confirming that the mechanism for escape was not likely due to impaired HLA binding but was more likely due to abnormal processing of Nef, as previously reported.

In another HLA A*24⁺ infant (P-1024), Q192H substitution was detected within an A*24-restricted nef epitope (aa 186–194; DSRLAFQHM) in 6 of 10 clones by 2 mo and 7 of 10 clones at 4 mo coincident with the detection by ELISPOT of enhanced CD8⁺ T cell responses to the wild-type sequence compared with the variant sequence (Fig. 1); this mutation represented the minority (3 of 10 clones) of maternal sequences at delivery. Of interest, is that reversion of this and all but one other CD8⁺ T cell escape mutations were noted in this infants’ plasma at 15 mo of age coincident with a decline in CD4⁺ T cell count and the lack of detectable epitope-specific responses by ELISPOT. Thus, our data suggest that CD8⁺ T cell selective pressures were responsible for the generation of the variant sequences and that reversion occurred when CD8⁺ T cell responses were no longer detectable.

These studies were undertaken to better understand the potential selective pressures exerted by CD8⁺ T cells on viral sequences in HIV-1-infected women and their infants. Heterogeneity of maternal viral sequences was demonstrated, and CD8⁺ T cell escape variants were commonly detected in maternal plasma. Evaluation of early infant plasma viruses documented heterogeneity of gag and nef gene sequences and common MTCT of CD8⁺ T cell escape variants. Infant HLA haplotype and viral fitness appeared to determine the stability of the CD8⁺ T cell escape mutants in the infant over time. Finally, changes within CD8⁺ T cell epitopes were detected in infants’ plasma over the first year of life, suggesting that early infant CD8⁺ T cell responses are active in vivo and exert selective pressures on the viral quasispecies.

Evaluation of the evolution of the quasispecies over time can provide important clues to selective pressures operant in vivo. However, the degree to which various factors may exert selective pressure on the virus is highly dependent not only on the specificity and strength of the selective force but also on the functional importance of the region of the targeted viral gene. Our data are compatible with population-based studies (11), as well as studies of infected individuals (43), which suggest that HIV-1 epitope-specific CD8⁺ T cell responses exert significant selective pressures on the viral quasispecies; SIV epitope-specific CD8⁺ T cell responses also appear to exert selective pressures in vivo (reviewed in Ref.43).

Variability of founder viral sequences may present a challenge to immune control of viral replication. Selective transmission of a few viral variants has been proposed in horizontal as well as vertical transmission due to the detection of highly homogeneous env sequences in recently infected individuals and greater env gene diversity found in individuals chronically infected (24, 25, 26, 27, 28, 29). However, heterogeneous populations in HIV-1 env have also been described. To our knowledge, our data are among the first to document heterogeneity of early infant gag and nef gene sequences (26, 30, 44). Although it is possible that some or all of the heterogeneity detected in the first infant sequences resulted from early host selective pressures, all but one of the infants studied acquired infection during birth, and the first samples represent time points very close to the acquisition of infection.

CD8⁺ T cell escape variants were commonly found in maternal plasma specimens. Because we focused on defined optimal epitopes, our studies likely underestimate the frequency of CD8⁺ T cell-mediated pressure in our study population. Most occurred in epitopes restricted by maternal HLA alleles, implying that they arose as a result of escape from maternal CD8⁺ T cell responses. However, several CD8⁺ T cell escape variants were detected that occurred within epitopes not known to be restricted by any of the maternal HLA alleles; these variants occurred within epitopes restricted by common HLA alleles (e.g., HLA A*2) and likely represent imprints of viral passage through prior hosts. In the same way that viruses resistant to antiretroviral therapies can be transmitted (15), it appears that CD8⁺ T cell escape variants may become fixed and may be propagated through populations (11). Vaccines that incorporate these epitopes may thus provide little protection against infection.

Studying the fate of CD8⁺ T cell escape variants over time following transmission between individuals may be a powerful tool to determine the fitness cost of the mutations in vivo. The detection of CD8⁺ T cell escape variants restricted by nonmaternal HLA alleles in the mothers studied and the persistence of maternal CD8⁺ T cell escape variants following their transmission to infants who do not share the restricting alleles suggest that the selective pressures exerted by CD8⁺ T cells commonly target areas of the virus in which mutations do not significantly affect the fitness of the virus to replicate in vivo. In contrast, escape variants with a high fitness cost to the virus are going to be no longer present in the recipient’s viral population when they are transmitted to a non-HLA-matched individual. As an example, we found a rare mutation E17K in gag p17 (α helix 1) in a B*4002-restricted epitope GELDRWEKI (GI9) in sequences from a B*40⁺ mother (M-1002). The amino acid substitution E17K (GELDRWKKI) was not detected in her B*40⁻ child’s viral population. Of note was that the amino acid substitution was located in a very conserved region (gag p17, α helix 1) previously described to be important for viral replication in Jurkat cells (35). We also demonstrated reduced recognition of this epitope variant GELDRWKKI in ELISPOT assays and that NL4-3 viruses containing this mutation E17K did not replicate in MT-4 cells. Of note is that this mother must have had replicating virus to transmit HIV to her infant. Therefore, we cannot exclude the existence of compensatory secondary mutations that partially restored viral fitness in vivo. Another possible explanation for detecting only wild-type sequences in the B*40⁻ child at 2 mo of age is that a minor population of wild-type viruses was transmitted and predominated in the absence of the selective pressure. CD8⁺ T cell epitopes in which variants have impaired fitness may represent good targets for inclusion into HIV-1 vaccines. Further studies should be done to identify these.

We have previously reported MTCT of ART-resistant viruses; the detection of ART resistance mutations in infants before therapy was associated with only transient response to therapy (15). The transmission of CD8⁺ T cell escape variants may similarly compromise the ability of infants to control HIV-1 replication. When CD8⁺ T cell escape variants were transmitted to infants with the restricting HLA allele, the variants were commonly detected in sequential infant plasma samples throughout the first 11–15 mo of life. This, together with documentation of reduced recognition of the variants by infant CD8⁺ T cells, suggests that the transmission of maternal CD8⁺ T cell variants may compromise the infants’ ability to generate effective HIV-1-specific CD8⁺ T cell responses. We and others have previously described the detection of virus-specific CD8⁺ T cells in fetuses (45), cord blood (36), or in peripheral blood samples taken as early as 1 day of age (38, 46), implying that the ability to generate virus-specific T cells begins in utero. However, direct evidence for the in vivo activity of these early virus-specific CD8⁺ T cell responses has been lacking. We observed de novo generation of an amino acid substitution (Y135F) within an A24-restricted epitope in infant P-1026. This mutation has been described previously as a common escape mutant in HLA A24⁺ hemophiliacs. Our data demonstrating the detection of CD8⁺ T cell escape variants as early as 2–3 mo of age is thus the first to document that young infants can generate CD8⁺ T cell responses that exert selective pressures in vivo and contribute to viral evolution over the first year of life. These data also provide support for the use of vaccines to prevent or modify the course of HIV-1 infection in young infants.

We thank the infants and their families for their participation in these studies. We also thank Margaret McManus for data management, Linda Lambrecht and Joyce Pepe for technical assistance, Mei Gong for assistance with viral cloning, and Wanda DePasquale for preparation of the manuscript.

The authors have no financial conflict of interest.