Adaptive Immunity Is the Primary Force Driving Selection of Equine Infectious Anemia Virus Envelope SU Variants during Acute Infection

Abstract

Equine infectious anemia virus (EIAV) is a lentivirus that causes persistent infection in horses. The appearance of antigenically distinct viral variants during recurrent viremic episodes is thought to be due to adaptive immune selection pressure. To test this hypothesis, we evaluated envelope SU cloned sequences from five severe combined immunodeficient (SCID) foals infected with EIAV. Within the SU hypervariable V3 region, 8.5% of the clones had amino acid changes, and 6.4% had amino acid changes within the known cytotoxic T lymphocyte (CTL) epitope Env-RW12. Of all the SU clones, only 3.1% had amino acid changes affecting potential N-linked glycosylation sites. In contrast, a much higher degree of variation was evident in SU sequences obtained from four EIAV-infected immunocompetent foals. Within V3, 68.8% of the clones contained amino acid changes, and 50% of the clones had amino acid changes within the Env-RW12 CTL epitope. Notably, 31.9% of the clones had amino acid changes affecting one or more glycosylation sites. Marked amino acid variation occurred in cloned SU sequences from an immune-reconstituted EIAV-infected SCID foal. Of these clones, 100% had amino acid changes within V3, 100% had amino acid changes within Env-RW12, and 97.5% had amino acid changes affecting glycosylation sites. Analysis of synonymous and nonsynonymous nucleotide substitutions revealed statistically significant differences between SCID and immunocompetent foals and between SCID foals and the reconstituted SCID foal. Interestingly, amino acid selection at one site occurred independently of adaptive immune status. Not only do these data indicate that adaptive immunity primarily drives the selection of EIAV SU variants, but also they demonstrate that other selective forces exist during acute infection.

Equine infectious anemia virus (EIAV) is a macrophage-tropic lentivirus that causes persistent infections in horses (7, 31). Infected horses usually develop recurrent episodes of plasma viremia and associated acute clinical disease (fever, inappetance, lethargy, thrombocytopenia, and anemia) during the first few months of infection (7, 31, 45). In contrast to the progressive disease caused by other lentiviral infections, including human immunodeficiency virus type 1 (HIV-1), most horses control EIAV replication within a year and remain inapparent carriers (7, 31, 45). Results of EIAV infection in normal and in severe combined immunodeficient (SCID) Arabian foals and, recently, immune reconstitution in a SCID foal prior to EIAV challenge, indicate that this control of viral replication is mediated by virus-specific immune responses, including neutralizing antibody and cytotoxic T lymphocytes (CTL) (8, 29, 40). Despite robust virus-specific immune responses that control viral replication and clinical disease, EIAV is not eliminated. Therefore, EIAV infection in horses is a useful model system not only for the study of lentiviral immune control, but also for the study of lentivirus persistence in the face of effective immune responses.

Immune evasion is an important mechanism of persistence for lentiviruses, and adaptive immune responses, particularly neutralizing antibodies and CTL, could play a role in selecting viral variants that escape adaptive immune control (9, 14). In rhesus monkeys persistently infected with the simian immunodeficiency virus SIVmac239 molecular clone, amino acid changes accumulate within the variable regions of envelope gp120 (SU) in a marked nonrandom pattern, suggesting the presence of selection pressure (5). Neutralizing antibody could be a source of selection pressure, since sequential sera obtained from SIVmac239-infected monkeys neutralize the inoculum virus but not the variants that arise later during infection (4). During acute EIAV infection, each recurrent viremic episode is associated with the emergence of an antigenically distinct EIAV variant, as defined by neutralizing antibody, which neutralizes virus isolated during earlier disease episodes but not virus isolated during subsequent disease episodes (6, 13, 15, 18, 19, 38, 39, 43, 44). As in SIV, amino acid changes appear within variable envelope gp90 (SU) regions in a nonrandom pattern during persistent EIAV infection, suggesting the presence of selection pressure (22). Moreover, sequential EIAV gp90 variants arise during persistent infection that become increasingly resistant to neutralization by serum antibody (12), suggesting that neutralizing antibody is one driving selective force.

Numerous HIV-1 and SIV studies have provided evidence that CTL responses can result in selection of viral variants that escape CTL recognition. Convincingly, adoptive transfer of a Nef-specific CTL clone in an HIV-1-infected patient was followed by clinical disease progression associated with the appearance of an HIV-1 isolate lacking the Nef epitope (17). Additionally, HIV-1 quasispecies containing amino acid changes within CTL epitopes occur in infected patients, an observation consistent with adaptive evolution (48). Examination of nonsynonymous-to-synonymous amino acid substitutions within a Nef CTL epitope in proviral HIV-1 sequences obtained from an infected individual also suggests that CTL exert selection pressure on the viral population during acute HIV-1 infection (42). In SIV-infected rhesus monkeys, viral variants with amino acid substitutions within CTL epitopes occur only in monkeys with major histocompatibility complex (MHC) class I molecules capable of presenting the epitopes, providing additional evidence for adaptive viral selection by CTL (10). Furthermore, high-avidity CTL may exert more viral selection pressure than low-avidity CTL, since high-avidity CTL are more likely to rapidly select for escape variants following acute SIV infection in rhesus monkeys (36). Lastly, viral variants that escape high-avidity SU-specific CTL arise during EIAV infection (30), corroborating the observations made in other lentiviral systems.

Although it seems likely that the adaptive immune response is primarily responsible for the selection pressure exerted on lentiviral populations during infection, definitive proof of this is lacking. To specifically test this hypothesis, we evaluated plasma envelope SU cloned sequences obtained from five SCID Arabian foals infected with EIAV. Since SCID foals lack functional T and B lymphocytes (27), they provide a means of evaluating lentiviral quasispecies arising after natural infection in the absence of adaptive immune selection pressure. SCID in Arabian foals is caused by a frameshift mutation in the gene encoding the catalytic subunit of DNA-dependent protein kinase (47) and has an autosomal recessive mode of inheritance (41). The equine SCID defect is more severe than its murine counterpart, in that SCID foals are incapable of forming either coding or signal joints (47).

Plasma envelope SU sequences obtained from SCID foals 27 to 33 days after EIAV infection revealed a low frequency of amino acid changes within known SU hypervariable regions. In contrast, more frequent amino acid variation within these regions was found in plasma SU sequences obtained within the same timeframe from immunocompetent foals infected with the same EIAV strain. We then examined plasma SU sequences obtained from an EIAV-infected SCID foal that underwent adaptive immune system reconstitution by infusion of EIAV-specific T and B lymphocytes. These sequences contained an even higher frequency of amino acid variation than that observed in the immunocompetent foals. Moreover, the ratio of synonymous to nonsynonymous nucleotide substitutions per site (d_s/d_n) within the hypervariable SU regions was significantly lower in immunocompetent foals than in SCID foals, and significantly lower in the immune-reconstituted SCID foal compared to SCID foals. However, a lysine at SU position 319, present at low frequency in the stock virus, was present in a higher percentage of cloned sequences in all foals. Our results provide definitive evidence that adaptive immunity is the primary selection force on the viral population following acute EIAV infection, but that other selective forces are also present in vivo, independent of an adaptive immune response.

MATERIALS AND METHODS

Horses.

Archived frozen plasma from six SCID foals and four immunocompetent foals was used in the study. As part of previous studies, all SCID foals were inoculated intravenously with 10⁶ 50% tissue culture infectious dose of EIAV_WSU5, while all immunocompetent foals were inoculated intravenously with 10^6.6 to 10⁷ 50% tissue culture infectious dose of EIAV_WSU5. Compared with SCID foals, immunocompetent foals require higher doses of EIAV_WSU5 to produce clinical disease (8). The WSU5 strain of EIAV is a cell culture-adapted biological clone (37). Details pertaining to experimental EIAV_WSU5 inoculation and subsequent clinical disease course for SCID foals A2053, A2057, A2064, and A2073 and immunocompetent foals A2054 and A2072 are described elsewhere (8), as are those for SCID foals A2148 and A2149 (29) and immunocompetent foals A2140 and A2147 (29, 30). SCID foals A2148 and A2149 were inoculated at 1 month of age, while immunocompetent foals A2147 and A2140 were inoculated at 9 and 10 months of age, respectively. All other foals were inoculated at 2 months of age. SCID foal A2148 underwent successful EIAV-specific immune reconstitution at the time of EIAV inoculation, by adoptive transfer of EIAV-stimulated T and B lymphocytes derived from immunocompetent horse A2140 (29). A2140 had been infected with EIAV_WSU5 for 3 months at the time the donor lymphocytes were obtained, and these lymphocytes contained CTL recognizing an epitope in the envelope SU protein, designated Env-RW12 (30). Following the adoptive transfer, EIAV-specific CTL as well as neutralizing antibody were detected in SCID foal A2148 (29). Although similar immune reconstitution was attempted in SCID foal A2149, lymphocyte engraftment did not occur (29). All experiments involving horses and foals were approved by the Washington State University Institutional Animal Care and Use Committee.

Plasma viral RNA purification.

Plasma samples used in the present study were collected 24 to 39 days post-EIAV_WSU5 inoculation (DPI) (Table 1) and stored at −80°C. Viral RNA was isolated from 140 μl of frozen EDTA plasma by using a QIAamp viral RNA kit (QIAGEN Inc., Chatsworth, Calif.) (23), treated with DNase I on the spin column (DNase I set; QIAGEN Inc.), eluted in 60 μl of nuclease-free water, and frozen at −20°C. Only heparinized plasma was available for horse A2140; viral RNA was extracted as above, and then the eluate was treated with heparinase as described elsewhere (25).

TABLE 1.

EIAV_WSU5-infected SCID, immunocompetent, and immune-reconstituted SCID foal plasma used for viral RNA extraction

	Age at EIAVWSU5 inoculation (mos)	Plasma used (DPI)
SCID foals
A2053	2	29
A2057	2	33
A2064	2	32
A2073	2	29
A2149	1	29
Immunocompetent foals
A2054	2	39
A2072	2	31
A2140	10	24
A2147	9	28
Immune-reconstituted SCID foal
A2148	1	27

RT-PCR amplification of plasma EIAV env SU.

Two overlapping segments of the env gene encoding EIAV SU were amplified from plasma viral RNA by using a SuperScript one-step reverse transcription-PCR (RT-PCR) kit (Life Technologies). The 5′ (env1) and 3′ (env2) segments of the env gene encoding SU spanned nucleotides 5237 to 6250 and 6121 to 7029, respectively, of the EIAV_WSU5 genome (GenBank accession no. AF247394). The following primers were used to amplify env1: forward primer (5237F), 5′-AAGACTGAAGGCAATCCA-3′; nested forward primer (5249F), 5′-AATCCAACAAGGAAGACAACC-3′; reverse primer (6250R), 5′-GTATAATCCCCCCGTACTACTC-3′. The following primers were used to amplify env2: forward primer (6121F), 5′-GTGTACAGCCACCATTTTTTC-3′; reverse primer (7029R), 5′-CCATTTTGCTTACCCAGTCATC-3′; nested reverse primer (6710R), 5′-AGCAGCAATAGCAGTAGCGG-3′. For both env1 and env2, the following reaction conditions were used: initial RT-PCR, 40 min at 45°C and 2 min at 95°C followed by 35 cycles of 15 s at 95°C, 30 s at 53°C, 1 min at 72°C, and then 7 min at 72°C; nested PCR, 2 min at 95°C followed by 30 cycles of 10 s at 95°C, 20 s at 53°C, 1 min at 72°C, and then 7 min at 72°C. Initial RT-PCR was performed using 5 μl of template RNA eluate in a 50-μl total reaction volume. Nested reactions used 5 μl of the initial RT-PCR mixture in a 100-μl total reaction volume.

Cloning and sequencing of RT-PCR products.

Amplified env1 and env2 products from RT-PCRs were purified from agarose gels using a QIAquick gel purification kit (QIAGEN Inc.) and cloned into the pCR2.1-TOPO TA cloning vector (Invitrogen). A QIAprep Spin Miniprep kit (QIAGEN Inc.) was used to extract plasmid DNA from randomly selected clones with inserts, and the nucleotide sequences were determined by the Laboratory for Biotechnology and Bioanalysis (Washington State University, Pullman), using an ABI Prism 377 (Applied Biosystems) automated DNA sequencer. Primers 5249F and M13R (vector primer) were used to sequence env1, while primer 6121F was used to sequence env2. Deduced amino acid sequences from each of the env1 and env2 SU clones obtained were compared to the EIAV_WSU5 SU consensus sequence by using the ClustalW and Boxshade programs. The same methods were used to sequence the EIAV_WSU5 stock virus inoculum. The foals in this study were infected at different time points, and the stock virus used for inoculations was maintained at −80°C with periodic passage in cell culture. To account for possible SU variation in the virus stock, 20 env1 (Fig. 1a) and 19 env2 (Fig. 1c) clones derived from the EIAV_WSU5 stock virus used at the time to inoculate SCID foals A2053, A2057, A2064, and A2073 and immunocompetent foals A2054 and A2072 were sequenced. In addition, 20 env1 (Fig. 1b) and 20 env2 (Fig. 1d) clones derived from the EIAV_WSU5 stock virus used at the time to inoculate SCID foals A2148 and A2149 and immunocompetent foals A2140 and A2147 were sequenced. For each of the two virus stock aliquots, env1 was amplified in two independent RT-PCRs and env2 was amplified in two independent RT-PCRs. From each independent RT-PCR, 9 to 10 clones were sequenced.

FIG. 1.

(a) Deduced amino acid sequences of EIAV_WSU5 SU env1 clones (SU amino acids 1 to 297) obtained from the stock virus used at the time to inoculate SCID foals A2053, A2057, A2064, and A2073 and immunocompetent foals A2054 and A2072. Only the hypervariable regions are shown in clones with different sequences from the EIAV_WSU5 consensus sequence (GenBank accession no. AF247394). The number of clones with a given sequence over the total number of sequenced clones is indicated in parentheses to the right of the clone name. The known hypervariable regions (V1 to V5) are boxed, and the putative PND is indicated (2, 22). Numbers flanking the consensus sequence segments indicate SU amino acid positions. Potential N-linked glycosylation sites are underlined, and the Env-RW12 CTL epitope (30) is in bold. (b) Deduced amino acid sequences of EIAV_WSU5 SU env1 clones obtained from the stock virus used at the time to inoculate SCID foals A2148 and A2149 and immunocompetent foals A2140 and A2147. Annotations are the same as for panel a. (c) Deduced amino acid sequences of EIAV_WSU5 SU env2 clones (SU amino acids 285 to 446) obtained from the stock virus used at the time to inoculate SCID foals A2053, A2057, A2064, and A2073 and immunocompetent foals A2054 and A2072. The known hypervariable regions (V6 to V8) are boxed (22). Other annotations are the same as for panel a. (d) Deduced amino acid sequences of EIAV_WSU5 SU env2 clones obtained from the stock virus used at the time to inoculate SCID foals A2148 and A2149 and immunocompetent foals A2140 and A2147. The known hypervariable regions (V6 to V8) are boxed. Other annotations are the same as for panel a.

Analysis of synonymous and nonsynonymous nucleotide substitutions.

For each of the hypervariable regions in cloned SU sequences derived from SCID foals, immunocompetent foals, and the immune-reconstituted SCID foal, the frequencies of synonymous nucleotide substitutions per potential synonymous site (d_s) and nonsynonymous nucleotide substitutions per potential nonsynonymous site (d_n) and the ratio of synonymous to nonsynonymous nucleotide substitutions (d_s/d_n) were calculated using the Synonymous/Nonsynonymous Analysis Program (SNAP) available in the Los Alamos National Laboratory HIV Sequence Database (11, 20, 34). The consensus sequence to which the SU clones from the infected foals were compared was derived from the 40 env1 and 39 env2 sequences cloned from the EIAV_WSU5 virus stock (described above). The nonparametric Mann-Whitney test (32) was performed using GraphPad InStat version 3.06 (GraphPad Software, San Diego, Calif.) to make statistical comparisons between mean d_n values and between mean d_s/d_n values for SCID foals versus immunocompetent foals, SCID foals versus the immune-reconstituted SCID foal, and immunocompetent foals versus the immune-reconstituted SCID foal. It was hypothesized that d_n values were in the following order: immune-reconstituted SCID > immunocompetent foals > SCID foals. The d_s/d_n ratios were hypothesized to be in the following order: SCID foals > immunocompetent foals > immune-reconstituted SCID foal. Therefore, a one-tailed P value of <0.05 was considered significant for each comparison (33).

Nucleotide sequence accession numbers.

The cloned viral sequences have been submitted to GenBank under accession numbers AY522737 to AY522845, AY524299 to AY524414, AY569397 to AY569435, and AY582944 to AY582963.

RESULTS

SU amino acid sequences from EIAV-infected immunocompetent foals.

Thirty-seven env1 clones and 38 env2 clones were sequenced from the four immunocompetent foals. Five env1 clones and one env2 clone had full-length nucleotide sequences but contained premature stop codons. These clones likely represented defective virus and were not analyzed further. The env1 sequences from the immunocompetent foals were pooled for analysis and represented seven to nine clones from each of four independent RT-PCRs (one per foal). The env2 sequences were analyzed the same way, and represented 9 to 10 clones from each of four independent RT-PCRs (one per foal). Analysis of the deduced amino acid sequences revealed significant variation within some of the previously described hypervariable regions of SU (22). Notably, 11 of 32 (34.4%) of the env1 clones (Fig. 2a) contained amino acid changes (compared with the EIAV_WSU5 consensus sequence) not detected in the stock virus (Fig. 1a and b) within V1, and 22 of 32 (68.8%) contained amino acid changes not detected in the stock virus within V3. Potential N-linked glycosylation sites (NXT/S) were created in three clones with an A→T change at position 203 just downstream of V3, within the putative principal neutralizing domain (PND) of EIAV SU (2, 22). Potential N-linked glycosylation sites were abolished in nine clones containing an N→S change within V3 at position 186, in one clone with an S→P change at position 188, and in one clone with a T→I change at position 216. An N-linked glycosylation site was shifted within V4 by a T→N change at position 235 in three clones.

FIG. 2.

(a) Deduced amino acid sequences of EIAV_WSU5 SU env1 clones obtained from plasma from EIAV_WSU5-infected immunocompetent foals. Annotations are the same as for Fig. 1a. (b) Deduced amino acid sequences of EIAV_WSU5 SU env1 clones obtained from plasma from EIAV_WSU5-infected SCID foals. Amino acids different than the EIAV_WSU5 consensus but present in the cloned stock virus are bold and italicized. Annotations are the same as for Fig. 1a.

The ELA-A1 haplotype-restricted CTL epitope Env-RW12 (30) is recognized by A2140 CTL and is presented by the classical MHC class I molecule 7-6 (28). Env-RW12 has its amino terminus within V3 and is completely contained within the PND. Sixteen of 32 (50%) env1 clones had amino acid changes within Env-RW12, none of which were detected within the stock virus. Nine of these clones were derived from A2140 plasma (the horse with Env-RW12-specific CTL), while the other seven clones were derived from A2054 and A2072 plasma (CTL activity not assayed; MHC class I haplotypes are unknown).

Of the env2 clones, 5 of 37 (13.5%) had amino acid changes within V6 that were not detected in the stock virus (Fig. 3a). Although the N→K change at SU amino acid position 319 was present in 3 of 39 (7.7%) stock virus env2 clones (Fig. 1c and d), this change appeared to offer a selective advantage in vivo, as it was present in 26 of the 37 clones (70.2%) obtained from the immunocompetent foal plasma. Four of 37 (10.8%) env2 clones had amino acid changes within V7, 1 of which was a deletion, and 3 of 37 (8.1%) clones had amino acid changes within V8. Within V6, potential N-linked glycosylation sites were created in five clones by a D→N change at SU amino acid position 309; however, this change was present in 1 of 39 (2.6%) stock virus env2 clones (Fig. 1c and d). An N-linked glycosylation site was shifted in three clones by an N→K change at position 314 (also present in 1 of 39 stock virus env2 clones), an S→N change at position 316, and an L→T change at position 318. Also in V6, an N-linked glycosylation site was abolished in one clone by an N→S change at position 314; however, this change was also present in 1 of 39 stock virus clones. An N-linked glycosylation site was created just upstream of V7 by a P→S change at position 349 in six clones and a P→T change at the same position in one clone. Lastly, N-linked glycosylation sites were abolished within V8 in two clones by an S→G change at position 402 and in two clones with a T→A change at position 414, downstream of V8. Considering all 69 immunocompetent env1 and env2 clones together, 22 (31.9%) had amino acid changes not detected in the stock virus that affected one or more potential N-linked glycosylation sites.

FIG. 3.

(a) Deduced amino acid sequences of EIAV_WSU5 SU env2 clones obtained from plasma from EIAV_WSU5-infected immunocompetent foals. The known hypervariable regions (V6 to V8) are boxed. Amino acids different than the EIAV_WSU5 consensus but present in the cloned stock virus are bold and italicized. Other annotations are the same as for Fig. 1a. (b) Deduced amino acid sequences of EIAV_WSU5 SU env2 clones obtained from plasma from EIAV_WSU5-infected SCID foals. The known hypervariable regions (V6 to V8) are boxed. Amino acids different than the EIAV_WSU5 consensus but present in the cloned stock virus are bold and italicized. Other annotations are the same as for Fig. 1a.

SU amino acid sequences from EIAV-infected SCID foals.

Forty-eight env1 clones and 50 env2 clones were sequenced from the five SCID foals. One env1 clone and one env2 clone had full-length nucleotide sequences but contained premature stop codons and were not analyzed further. The env1 sequences from the SCID foals were pooled for analysis and represented 6 to 12 clones from each of five independent RT-PCRs (one per foal). The env2 sequences were analyzed the same way and represented 9 to 10 clones from each of five independent RT-PCRs (one per foal). Analysis of the deduced amino acid sequences of the SU clones derived from SCID foal plasma revealed less variation than in the clones derived from immunocompetent foals. Seven of 47 (14.9%) env1 clones had amino acid changes within V1 (Fig. 2b) that were not detected in the stock virus. Three clones had a Y→C change at position 27 and an N→K change at position 30. Four immunocompetent clones had these same changes (Fig. 2a), and it was possible that these clones were also present in the stock virus at a low enough frequency that they were not detected in the 40 stock virus env1 clones sequenced. Five of 47 (10.6%) clones had amino acid changes within V3, but one of these (R→C at position 195) was present in the stock virus. Four of 47 (8.5%) clones had amino acid changes within the Env-RW12 CTL epitope, but again, one of these (R→C at position 195) was present in the stock virus. Therefore, only three clones (6.4%) had unique amino acid changes within Env-RW12, and four clones (8.5%) had unique amino acid changes within V3. Lastly, 3 of 47 (6.4%) had amino acid changes within V5. None of the amino acid changes in SCID env1 clones affected potential N-linked glycosylation sites.

Of the 49 SCID env2 clones analyzed, 2 (4.1%) had amino acid changes within V6 that were not detected in the stock virus (Fig. 3b). Fifteen clones (30.6%) had only the N→K change at position 319, which was present in 7.7% of the stock virus env2 clones. As in the immunocompetent foals, this change appeared to offer some selective advantage in vivo. Eight of 49 (16.3%) clones had amino acid changes within V7; however, the Q→R change at position 360 (3 clones) and the N→K change at position 375 (1 clone) were present in the stock virus, leaving only 4 of 49 clones (8.2%) with amino acid changes within V7 not detected in the stock virus. Of the four clones with unique amino acid changes within V7, one had a deletion at position 370, resulting in a 1-position upstream shift of a potential N-linked glycosylation site. As in the immunocompetent SU clones, 2 of 49 (4.1%) clones had amino acid changes in V8 (S→G at position 402) that abolished an N-linked glycosylation site, but it was possible that this change was also present in the stock virus at too low a frequency to detect in the 39 stock virus env2 clones sequenced. Considering all 96 of the SCID env1 and env2 clones together, only 3 (3.1%) had amino acid changes affecting potential N-linked glycosylation sites.

SU amino acid sequences from an EIAV-infected SCID foal after immune reconstitution.

Twenty env1 and 20 env2 clones were sequenced from immune-reconstituted SCID foal A2148. The 20 env1 sequences were derived from two independent RT-PCRs (10 clones each), as were the 20 env2 sequences. All 20 env1 clones (100%) had amino acid changes within V3, V4, and V5, and all had amino acid deletions or substitutions within the CTL epitope Env-RW12 (Fig. 4a). Within V1, one clone (5%) had an N→S change at position 40, abolishing a potential N-linked glycosylation site. Within V3, seven clones (35%) had an N→S change at position 186, also abolishing an N-linked glycosylation site. All clones had an A→T change at position 203, creating an N-linked glycosylation site within the PND just downstream of V3. All clones had a T→N change at position 235, shifting an N-linked glycosylation site within V4, and an N-linked glycosylation site was created in all 20 clones within V5 by a D→N change at position 282. None of the above amino acid changes was detected in the stock virus.

FIG. 4.

(a) Deduced amino acid sequences of EIAV_WSU5 env1 clones obtained from plasma from immune-reconstituted, EIAV_WSU5-infected SCID foal A2148. Amino acids different than the EIAV_WSU5 consensus but present in the cloned stock virus are bold and italicized. Annotations are the same as for Fig. 1a. (b) Deduced amino acid sequences of EIAV_WSU5 SU env2 clones obtained from plasma from immune-reconstituted, EIAV_WSU5-infected SCID foal A2148. The known hypervariable regions (V6 to V8) are boxed. Amino acids different than the EIAV_WSU5 consensus but present in the cloned stock virus are bold and italicized. Other annotations are the same as for Fig. 1a.

Of the 20 env2 clones, all (100%) had amino acid changes within V6 at positions 309, 315, and 319 (Fig. 3b). The position 309 D→N, present in only 1 of 39 (2.6%) of the stock virus env2 clones, created a potential N-linked glycosylation site, while the position 315 V→E change occurred within an N-linked glycosylation site. As in the immunocompetent and SCID env2 clones, the position 319 N→K change (present in only 7.7% of the stock virus env2 clones) appeared to offer a selective advantage in this foal, as it was present in 100% of the clones. An N→K change at position 375 within V7 (11 of 20 clones; 55%) also appeared to offer a selective advantage in this foal, as it was present in only 2 of 39 (5.1%) virus stock env2 clones (Fig. 1c and d). One clone (5%) had an N→T change at position 314, abolishing an N-linked glycosylation site within V6. Three clones (15%) had a T→N change at position 338, creating an N-linked glycosylation site between V6 and V7. Twelve clones (60%) had a P→S change at position 349, creating an N-linked glycosylation site just upstream of V7. Within V7, 14 clones (70%) had amino acid changes, with 10 clones (50%) having an N insertion at position 370, affecting an N-linked glycosylation site, and 1 clone (5%) having an N→T change at position 372, creating an additional N-linked glycosylation site. Also in V7, one clone (5%) had an N→S change at position 375, creating yet-another N-linked glycosylation site. Considering all 40 of the reconstituted SCID env1 and env2 clones together, 39 (97.5%) had amino acid changes not detected in the stock virus that affected potential N-linked glycosylation sites.

Synonymous (d_s) and nonsynonymous (d_n) nucleotide substitutions.

The hypervariable regions of SU clones from SCID foals had significantly fewer nonsynonymous nucleotide substitutions per potential nonsynonymous site (d_n) than did the SU clones from immunocompetent foals (P = 0.0087) or from the immune-reconstituted SCID foal (P = 0.0048) (Table 2 and Fig. 5a and c). The d_s/d_n ratio for the SU clones from immunocompetent foals was significantly less than that for the SU clones from SCID foals (P = 0.0222) (Table 2 and Fig. 5b). Importantly, the d_s/d_n ratio for the SU clones from the immune-reconstituted SCID foal was also less than that for the SU clones from SCID foals, and the difference was highly significant (P = 0.0048) (Table 2 and Fig. 5d). Although the d_s/d_n ratio for the SU clones from the immune-reconstituted SCID foal was less than that for the SU clones from immunocompetent foals, the difference was not statistically significant (P = 0.0547) (Table 2 and Fig. 5f). However, d_n was significantly less for the SU clones from immunocompetent foals compared to that for the SU clones from the immune-reconstituted SCID foal (P = 0.0079) (Table 2 and Fig. 5e). These results were consistent with the hypothesis that EIAV SU variants arise during acute infection as a result of adaptive immune selection pressure.

TABLE 2.

Mean d_s, d_n, and d_s/d_n ratios for hypervariable regions of env1 and env2 in immunocompetent, SCID, and immune-reconstituted SCID foals

SU region	Immunocompetent foals			SCID foals			Immune-reconstituted SCID foal
	Mean ds	Mean dn	ds/dn	Mean ds	Mean dn	ds/dn	Mean ds	Mean dn	ds/dn
env1 V1	0.0168	0.0092	1.8261	0.0069	0.0040	1.7250	0.0000	NAa	NA
env1 V2	0.0000	0.0037	0.0000	0.0067	0.0012	5.5833	0.0000	NA	NA
env1 V3	0.0026	0.0113	0.2301	0.0026	0.0014	1.8571	0.0023	0.0282	0.0816
env1 V4	0.0063	0.0039	1.6154	0.0043	0.0018	2.3889	0.0000	0.0417	0.0000
env1 V5	0.0048	NA	NA	0.0049	0.0015	3.2667	0.0000	0.0259	0.0000
env2	0.0069	0.0115	0.6000	0.0018	0.0030	0.6000	0.0030	0.0199	0.1508
Overall mean	0.0062	0.0079	0.8543	0.0045	0.0022	2.5702	0.0009	0.0289	0.0581
Overall SD	0.0058	0.0039	0.8228	0.0021	0.0011	1.7147	0.0014	0.0092	0.0728

^aNA, mean d_n was not determined and d_s/d_n was not calculated because only one or no clones had nonsynonymous nucleotide substitutions.

FIG. 5.

Comparisons of mean nonsynonymous nucleotide substitutions per potential nonsynonymous site (d_n) and ratios of synonymous to nonsynonymous nucleotide substitutions (d_s/d_n) within SU hypervariable regions for immunocompetent, SCID, and immune-reconstituted SCID foals. Error bars are standard deviations. Nonparametric Mann-Whitney one-tailed P values for each comparison are indicated, and an asterisk denotes a P value of <0.05.

DISCUSSION

Substantial viral genetic diversity is generated during lentivirus infection due to the error-prone process of RT in conjunction with high levels of virus replication. Convincing evidence suggests that antiviral pressure exerted by neutralizing antibody and CTL allows variants within the viral population that lack the requisite epitopes to expand, resulting in disease progression and enhancing viral persistence. However, little is known regarding the lentivirus variants that arise during infection in a host lacking an adaptive immune system. Without this information, it would be difficult to conclude that adaptive immunity is the major selective force driving viral variation. The present study was designed to test the hypothesis that the adaptive immune system is primarily responsible for the selection of viral variants following acute lentiviral infection, and it exploited a unique and naturally occurring immunodeficiency model system in an outbred species, namely, EIAV infection of SCID foals. Since SCID foals lack functional T and B lymphocytes (27, 47), the hypothesis could be rigorously tested in a way unavailable in any other lentiviral model system. The foals in this study were infected with an EIAV_WSU5 stock virus that contained SU clones with some amino acid variation. Infection with an EIAV molecular clone would have avoided this possibility. However, the purpose of this study was not to determine the origin of SU variation, but rather to specifically determine if the presence of an adaptive immune response was primarily involved in the selection of amino acid variants in defined regions of SU. Therefore, the presence of sequence variants in the stock virus did not preclude the ability to address the stated hypothesis.

Cloned plasma viral SU sequences obtained 24 to 39 DPI from five SCID foals, four immunocompetent foals, and one immune-reconstituted SCID foal were evaluated and compared. Since EIAV-infected SCID foals only survive a maximum of 35 ± 3 DPI (8, 29, 40), it was not possible to evaluate SU sequences later in infection, although this would have been of interest since the differences between the immunocompetent and SCID SU sequences might have been even more profound. Regardless, differences were observed within the hypervariable regions of SU, with the most striking differences occurring within V3, which is contained within the putative PND of EIAV SU (2, 22). Notably, 68.8% of the immunocompetent SU clones contained amino acid changes within V3, compared with only 8.5% for the SCID SU clones. These SU sequences were obtained within a narrow timeframe from all the foals (24 to 39 DPI). Immunocompetent foals A2054 and A2072 were inoculated with EIAV at 2 months of age, similar to the SCID foals, while immunocompetent foals A2140 and A2147 were inoculated at 10 and 9 months of age, respectively. However, the degree of SU amino acid variation seen in the younger immunocompetent foals appeared similar to that in the older immunocompetent foals, making it unlikely that age differences between immunocompetent foals and SCID foals contributed to the difference in SU amino acid variation observed.

Previous studies examining SU sequences during early EIAV infection had somewhat disparate results, reporting both significant variation (1) as well as limited variation (24). Our results were more consistent with those of the former study. The latter study evaluated 12 clones from each of two ponies, one at 19 DPI and the other at 33 DPI. Despite reporting limited variation, all 12 clones from the pony evaluated at 33 DPI had amino acid changes within the PND region that were not present in the inoculum (24), consistent with our results in the immune-reconstituted and immunocompetent SCID foals. The present study and the latter study both evaluated plasma viral RNA following infection with a biologically cloned virus. Discrepancies could have been due to the greater number of horses and SU clones evaluated in the present study, as well as the longer duration of infection for horses in the present study. In the present study, the profound variation observed in the SU clones from the immune-reconstituted SCID foal resulted from the transfer of virus-specific lymphocytes from a donor that had been EIAV infected for 3 months. Thus, although the SU clones from the reconstituted SCID foal were obtained 27 DPI, the reconstituted immune system was more “mature” and exerted greater selection pressure on EIAV SU than did the immune systems of the immunocompetent foals.

The epitopes recognized by neutralizing antibodies in EIAV-infected horses are not known, but the PND/V3 region contains two epitopes recognized by neutralizing murine monoclonal antibodies (2). Additionally, the PND/V3 region contains the 7-6-restricted CTL epitope Env-RW12 (30). Over half of the immunocompetent SU clones had amino acid changes within these two neutralizing antibody epitopes or within Env-RW12. Immunocompetent horse A2140 is known to have Env-RW12-specific CTL (30), and every SU clone derived from this horse contained amino acid changes within the epitope. At the time these SU sequences were obtained from A2140 (24 DPI), plasma neutralizing antibody activity was considered absent, based on a virus reduction assay (<4% virus reduction), suggesting that CTL pressure was involved in selection of these variants (30). However, it is possible that the in vitro virus reduction assay used was not sensitive enough to detect neutralizing activity that was functionally relevant in vivo. Unfortunately, CTL assays were not performed in this horse before DPI 47, and so the contribution of CTL to the selection of DPI 24 SU variants could not be confirmed. Nevertheless, Env-RW12 escape variants arose later during infection in A2140 (30).

Envelope glycosylation is relevant to lentiviral immune evasion, since these surface glycans likely mask underlying B-lymphocyte epitopes as well as shield receptor binding sites. The N-linked glycosylation sites of HIV-1 gp120 are all utilized (21), and most occur on the exposed surface of the outer domain, likely rendering it less accessible to neutralizing antibody (49). Removal of a single N-linked glycosylation site at the base of the gp120 V3 loop can increase the neutralization sensitivity of HIV-1 primary isolates, indicating the importance of surface carbohydrates in protecting the virus from neutralizing antibodies (16). The overall structure of EIAV gp90, as well as its utilization of N-linked glycosylation sites, is probably similar to that of HIV-1 gp120. Importantly, EIAV variants with amino acid changes affecting N-linked glycosylation sites in the V3 and V4 regions of gp90 are neutralization resistant (12).

Given that envelope glycosylation is important in lentiviral immune evasion, it was expected in the present study that immunocompetent foals would have more SU sequences with amino acid changes affecting potential N-linked glycosylation sites than would SCID foals. Half of the immunocompetent SU clones with amino acid variation in the PND/V3 region had amino acid changes that affected potential N-linked glycosylation sites, while no changes in N-linked glycosylation sites were observed in this region in the SCID SU clones. Overall, 31.9% of all SU clones from immunocompetent foals contained amino acid changes affecting one or more glycosylation sites throughout SU compared with only 3.1% of all the SCID SU clones, suggesting the presence of adaptive immune selection, most likely by an early neutralizing antibody response.

Strikingly, all of the SU clones from immune-reconstituted SCID foal A2148 contained multiple amino acid changes within the PND/V3 region, V4, V5, V6, and V7. In addition, all of the A2148 SU clones had amino acid changes within the Env-RW12 CTL epitope, with 40% containing deletions of the first two residues (R and V). The V in position 2 of Env-RW12 is probably an anchor residue, and an amino acid change in this position prevents binding to MHC class I molecule 7-6, which presents Env-RW12 (28). Moreover, all of the A2148 SU clones had amino acid changes in multiple N-linked glycosylation sites. The higher number of nonsynonymous nucleotide substitutions per potential nonsynonymous site (d_n) and the lower d_s/d_n ratio in the immune-reconstituted SCID foal and immunocompetent foals compared to the SCID foals demonstrated that selection for amino acid changes within the hypervariable regions of SU correlated with the presence of an intact adaptive immune response.

It is highly unlikely that the source of the SU variants in A2148 was the infused lymphocytes derived from EIAV-infected lymphocyte donor A2140. First, A2140 was free of clinical disease and plasma virus at the time the donor peripheral blood mononuclear cells (PBMC) were obtained (29, 30). These PBMC were washed extensively prior to in vitro expansion in tissue culture and washed again prior to transfer to A2148 (29). Second, EIAV is detected only in tissue macrophages and not PBMC in acutely infected horses (46). In horses with subclinical infection, cell types other than monocytes/macrophages are not infected, and EIAV_WSU5 cannot be detected in PBMC (35). Therefore, even if monocytes had been infused, they probably would not have been infected with virus from the donor. Third, the donor PBMC were expanded in tissue culture for 2 weeks prior to infusion into A2148; thus, monocytes/macrophages were removed by adherence and not infused (29). Lastly, SCID foal A2149 also received an infusion of lymphocytes from an EIAV-infected immunocompetent horse (29). These lymphocytes did not engraft in A2149 (29), and minimal amino acid variation was observed in the SU clones derived from A2149.

Of interest was the observation that an amino acid present within V6 in only a few stock virus clones (N→K at position 319) was present in that position at much higher frequency in the SU clones from SCID, reconstituted SCID, and immunocompetent foals, regardless of adaptive immune status. The factors contributing to the selection of SU sequences independent of adaptive immunity are not known but could include the innate immune response (complement, cytokines, and natural killer cell activity), as this is intact in SCID foals (3, 26, 27), cell and tissue tropism, and replication rate.

In this study, analysis of SU sequences from EIAV-infected immunocompetent foals revealed a considerable degree of amino acid variation within the hypervariable regions and within regions affecting N-linked glycosylation sites. In contrast, SU sequences from EIAV-infected SCID foals displayed minimal amino acid variation in these regions. Importantly, a striking level of amino acid variation was observed in the SU sequences from a SCID foal that underwent adaptive immune reconstitution which included neutralizing antibody and CTL. Lastly, higher d_n values and lower d_s/d_n ratios in SU hypervariable regions correlated with host immune status. This study not only confirms that adaptive immune selection is the primary force driving selection of EIAV gp90 variants, but also demonstrates that selective forces independent of adaptive immunity are present in vivo during acute infection. These results further indicate that strategies to induce protective lentiviral immune responses must be directed towards viral regions with limited capacity for variation.