Structural and Functional Constraints Limit Options for Cytotoxic T-Lymphocyte Escape in the Immunodominant HLA-B27-Restricted Epitope in Human Immunodeficiency Virus Type 1 Capsid

Abstract

Control of human immunodeficiency virus type 1 (HIV-1) by HLA-B27-positive subjects has been linked to an immunodominant CD8⁺ cytotoxic T-lymphocyte (CTL) response targeting the conserved KK10 epitope (KRWIILGLNK_263-272) in p24/Gag. Viral escape in KK10 typically occurs through development of an R₂₆₄K substitution in conjunction with the upstream compensatory mutation S₁₇₃A, and the difficulty of the virus to escape from the immune response against the KK10 epitope until late in infection has been associated with slower clinical progression. Rare alternative escape mutations at R₂₆₄ have been observed, but factors dictating the preferential selection of R₂₆₄K remain unclear. Here we illustrate that while all observed R₂₆₄ mutations (K, G, Q, and T) reduced peptide binding to HLA-B27 and impaired viral replication, the replicative defects of the alternative mutants were actually less pronounced than those for R₂₆₄K. Importantly, however, none of these mutants replicated as well as an R₂₆₄K variant containing the compensatory mutation S₁₇₃A. In assessing the combined effects of viral replication and CTL escape using an in vitro coculture assay, we further observed that the compensated R₂₆₄K mutant also displayed the highest replication capacity in the presence of KK10-specific CTLs. Comparisons of codon usage for the respective variants indicated that generation of the R₂₆₄K mutation may also be favored due to a G-to-A bias in nucleotide substitutions during HIV-1 replication. Together, these data suggest that the preference for R₂₆₄K is due primarily to the ability of the S₁₇₃A-compensated virus to replicate better than alternative variants in the presence of CTLs, suggesting that viral fitness is a key contributor for the selection of immune escape variants.

CD8⁺ cytotoxic T lymphocytes (CTL) play a critical role in the immune response against the human immunodeficiency virus type 1 (HIV-1) and against the simian immunodeficiency virus (SIV) in the macaque model. This assertion is supported by accumulating evidence that links effective CTL responses with viral containment (1, 6, 9, 36, 50) and ascribes a protective phenotype to particular major histocompatibility complex (MHC) class I alleles, such as human leukocyte antigen (HLA)-B57 and -B27 in HIV-1 infection (20, 25, 32, 41, 46, 57) or Mamu-A01 and -B17 in SIV infection (43, 61). Conversely, depletion of CD8⁺ cells results in impaired control of SIV infection (31, 40, 51), and in several reports, mutational escape in prominent CTL epitopes has been associated with disease progression for HIV-1 (1, 5, 11, 17, 26, 33, 35, 47, 52) and for SIV/SHIV (6).

HLA-B27-positive (HLA-B27⁺) individuals infected with HIV-1 clade B mount an immunodominant CTL response targeting the KK10 epitope (KRWIILGLNK_263-272) in p24 Gag (3, 26, 45, 56). Viral escape from this conserved epitope is reported to arise late in infection and to be associated with progression to AIDS (1, 17, 26, 33). We have recently demonstrated that the predominant CTL escape mutation R₂₆₄K in KK10 dramatically compromises in vitro viral replication capacity and impacts the interaction between the viral capsid and the host protein cyclophilin A (CypA) (52), highlighting the vulnerability of this region in p24 to host immune pressures. Replication of the mutated virus is restored to wild-type (WT) levels through incorporation of a rare upstream compensatory mutation observed in vivo, S₁₇₃A, which functions by restoring the normal CypA-dependent phenotype (52). These data suggest a significant impact of the respective mutations on the overall structure and/or function of HIV-1 capsid.

Besides R₂₆₄K, three alternative mutations at codon 264 are described to occur infrequently in HLA-B27⁺ individuals, namely, R₂₆₄T, R₂₆₄Q, and R₂₆₄G (4, 8, 17, 24, 33, 48, 52) (Table 1). As each of these substitutions also facilitates escape from the B27-KK10 response (4, 24, 26, 33), we sought to examine why the R₂₆₄K mutation is selected preferentially despite its dramatic impact on viral replication. Factors that might account for the predominant selection of R₂₆₄K include potential differences in the abilities of the escape mutations at codon 264 to impair HLA binding, their respective impacts on viral replication capacity in the absence and/or presence of CTL, and biased nucleotide substitution rates such that some escape variants are biologically favored during error-prone viral replication.

TABLE 1.

Reported CTL escape mutations in KK10 epitope

Viral variant or WT	Mutations for indicated positions and amino acid and nucleotide sequencesa																											Frequencyb	Reference(s)
	170							(…)				260										270
WT	P	M	F	S	A	L	S		P	V	G	E	I	Y	K	R	W	I	I	L	G	L	N	K	I	V	R
	CCC	ATG	TTT	TCA	GCA	TTA	TCA		CCA	GTA	GGA	GAA	ATC	TAT	AAA	AGA	TGG	ATA	ATC	CTG	GGA	TTA	AAT	AAA	ATA	GTA	AGA
SARKLM				A												K				M								22/707	1,c,d22,c,d24,c,d
				GCA												AAA				ATG									26,c,d33,c52
RTLM																T				M								0/707	17,c,d24,c,d
																ACA				ATG
RT																T												0/707	17,c,d52
																ACA
RQLM																Q				M								1/707	4,c48c,d
																CAA				ATG
EDRG												D				G												1/707	33c
												GAC				GGA
RG																G												0/707	8c
																GGA

^aThe numbering of the amino acids is according to NL4-3 Gag (GenBank accession no. AF324493). The amino acid and nucleotide sequences displayed are identical in the WT NL4-3 and B-clade consensus sequences (www.hiv.lanl.gov). Boldfaced amino acids indicate the KK10 epitope (amino acids 263 to 272).

^bFrequency, the number of sequences in which the mutation is present/total number of sequences. Values represent frequency in LANL HIV sequence database (www.hiv.lanl.gov) for sequences covering Gag amino acids 152 to 300, as of October 2007.

^cPublished sequences do not cover Gag codon 173.

^dNo accession numbers available.

Here we illustrate that despite the infrequent selection of these alternative mutations, substitutions R₂₆₄T, R₂₆₄G, and R₂₆₄Q were nonetheless better able to abrogate binding to HLA-B27 than R₂₆₄K. Similarly, while each of these mutations impaired viral replication, the R₂₆₄K mutation exhibited the most pronounced impact on viral replication. However, viral replication in both the absence and presence of KK10-specific CTL was compromised to a greater degree for the alternative escape mutations than for the R₂₆₄K virus containing the upstream S₁₇₃A compensatory mutation, suggesting that a minimal threshold of viral replication may be required to allow for selection of the dominant variant in the quasispecies. Notably, in analyzing codon usage of the respective variants, we also found that generation of the predominant R₂₆₄K mutation was favored over others due to a bias in the frequency of G-to-A nucleotide substitutions during the replicative cycle of HIV-1. Taken together, these data suggest that the preferential selection of R₂₆₄K in vivo is strongly influenced by structural and functional constraints on this region of p24/Gag and by the ability of the S₁₇₃A substitution to restore viral replication capacity. Despite the capability of the alternative escape variants R₂₆₄T, R₂₆₄G, and R₂₆₄Q to effectively impair HLA binding, their substantial impact on viral replication without sufficient compensation limits the possibility of these mutations to emerge.

MATERIALS AND METHODS

Peptide binding studies.

Quantitative assays to measure the binding of peptides to purified HLA B*2705 class I molecules are based on the inhibition of the binding of a radiolabeled standard peptide (human 60s rL28 38-46, sequence FRYNGLIHR; 50% inhibitory concentration [IC₅₀], 6.7 nM) and were performed as previously described (54, 55). Briefly, 0.1 to 1 nM of radiolabeled peptide was coincubated at room temperature with 1 nM to 1 μM of purified MHC in the presence of 1 to 3 μM human beta2-microglobulin (Scripps Laboratories, San Diego, CA) and a cocktail of protease inhibitors. Following a 2-day incubation, the percentage of MHC-bound radioactivity was determined by capturing MHC/peptide complexes on W6/32 antibody-coated Lumitrac 600 microplates (Greiner Bio-One) and measuring bound counts per minute using the TopCount microscintillation counter (Packard Instrument Co.). Peptides were tested at six different concentrations covering a 100,000-fold dose range and in three or more independent assays. Under the conditions utilized, where [label] < [MHC] and IC₅₀ ≥ [MHC], the measured IC₅₀ values are reasonable approximations of K_D (equilibrium dissociation constant) (14, 27).

Variant NL4-3 constructs.

HIV-1 strain NL4-3 was modified to express one or more mutations in p24 using the GeneTailor site-directed mutagenesis system (Invitrogen). A SacI-SbfI fragment (residues 491 to 2844) was isolated from pNL4-3 and ligated into pUC19. Mutations S₁₇₃A, R₂₆₄K, and L₂₆₈M were introduced as previously described (52). Mutagenesis of the other variants was performed using 5′ oligonucleotide primers R264T F (CCAGTAGGAGAAATCTATAAAACATGGATAATCCTG; nucleotide [nt] 1593), R264T on LM F (CCAGTAGGAGAAATCTATAAAACATGGATAATCATG; nt 1593), R264X R-1 (TTTTATAGATTTCTCCTACTGGGATAGGTGG; nt 1549), R264X R-2 (TTTATAGATTTCTCCTACTGGGATAGGTGG; nt 1549), R264Q F (CCAGTAGGAGAAATCTATAAACAATGGATAATCCTG; nt 1593), R264Q on LM F (CCAGTAGGAGAAATCTATAAACAATGGATAATCATG, nt 1593), E260D F (CCACCTATCCCAGTAGGAGACATCTATAAAA; nt 1578), E260D R (TCTCCTACTGGGATAGGTGGATTATGTGTC; nt 1539), R264G F (CCAGTAGGAGAAATCTATAAAGGATGGATAATCCTG; nt 1593), R264G on LM F (CCAGTAGGAGAAATCTATAAAGGATGGATAATCATG, nt 1593), S173T F (GAAGTAATACCCATGTTTACAGCATTATCAG; nt 1317), and S173T R (AAACATGGGTATTACTTCTGGGCTGAAAG; nt 1277). Mutated codons are underlined, and primer positions are numbered according to pNL4-3 (GenBank accession no. AF324493). Mutated SacI-SbfI fragments were then isolated from pUC19 and cloned back into pNL4-3. The complete HIV-1 coding region of the variant proviruses was sequenced using published primers (2) on an ABI 3730 XL DNA analyzer. Escherichia coli One Shot Stbl3 cells (Invitrogen) were used to propagate full-length proviral plasmids, and stocks were prepared using a QIAprep spin miniprep kit or HiSpeed plasmid midi kit (Qiagen).

Cells.

HEK293T cells were cultured at 37°C in a 5% CO₂ humidified incubator in Dulbecco's modified Eagle medium (DMEM; Gibco) supplemented with 10% fetal bovine serum (Atlantic Biologicals), penicillin (50 IU/ml), and streptomycin (50 μg/ml). CEM-GXR cells (12) were grown in R10+ medium (RPMI 1640 medium [Sigma] supplemented with 10% fetal bovine serum, 2 mM l-glutamine, penicillin [50 IU/ml], and streptomycin [50 μg/ml]) at 37°C and 5% CO₂. Peripheral blood mononuclear cells (PBMC) were separated from whole blood using Histopaque-1077 (Sigma) and then stimulated with a CD3:CD8-bispecific monoclonal antibody at a concentration of 0.5 μg/ml in R10-50 medium (R10+ medium containing 50 U/ml human interleukin-2 [Hoffmann-La Roche]) for 4 days prior to infection. A KK10-specfic CTL clone was generated using PBMC isolated from a B27⁺ HIV controller subject by peptide stimulation and limiting dilution using previously described methods (60). CTL were restimulated periodically using anti-CD3 antibody (12F6 clone) plus irradiated allogeneic feeder PBMC and maintained in R10-50 medium prior to use. Peptide-specific lytic activity was verified using chromium-51 release assays, and clonality was confirmed by sequencing of the T-cell receptor (TCR) alpha and beta chains (not shown) using standard methods (62).

Viral stocks.

Viral stocks were generated by transfection of HEK293T cells with 5 μg of plasmid DNA in antibiotic-free DMEM using Lipofectamine 2000 (Invitrogen). Pseudotyping with the vesicular stomatitis virus glycoprotein G (VSV-g) was performed by cotransfecting with 300 ng of pHEF-VSVG (13). Supernatants were harvested 48 h after transfection, and frozen aliquots stored at −80°C. The capsid concentration of the viral stocks was quantified by p24 enzyme-linked immunosorbent assay (ELISA) using the Alliance HIV-1 p24 ELISA kit (Perkin-Elmer). Initial titers were determined by flow cytometry using CEM-GXR cells.

Viral replication assays.

One million CEM-GXR cells were pelleted and resuspended with WT or variant virus at a multiplicity of infection (MOI) of 0.0015 in a total volume of 3 ml R10+ medium. Five hundred-microliter aliquots of the culture were harvested at the indicated times, and the volume was replaced with fresh R10+ medium. The proportion of green fluorescent protein (GFP)-expressing cells was determined by fluorescence-activated cell sorter (FACS). For the analyses of primary cells, one million stimulated PBMC were infected with inocula of WT, SARKLM, RTLM, RQLM, and EDRG viral variants, leading to similar fractions of CD4-positive cells with intracellular p24, as detected by FACS 72 h postinfection. Infection with RK and RKLM viruses was performed using 300 μl inocula. Day 3 infectious rates for WT, SARKLM, RTLM, RQLM, and EDRG viruses varied for the different PBMC samples between 2 and 10% of CD4-positive cells. Rates for RK and RKLM variants were consistently 1 log lower. PBMC were infected in a total volume of 1.5 ml R10-50 medium. Aliquots of 250 μl of the culture were harvested at the indicated times, and the volume was replaced with fresh R10-50 medium.

Staining and flow cytometry.

CEM-GXR cells were fixed in phosphate-buffered saline containing 2% paraformaldehyde, and GFP expression was determined using FACS. A signal 10-fold above the median fluorescence index of uninfected cells was considered positive, excluding greater than 99.95% of uninfected cells. PBMC were permeabilized using the Cytofix/Cytoperm kit (BD Biosciences) and stained for CD4 (CD4-allophycocyanin; BD Pharmingen) and intracellular p24/Gag (KC57-RD1; Beckman Coulter) to identify the proportion of infected CD4⁺ T cells in the cultures with the different viral variants. FACS analysis was done using a FACSCalibur flow cytometer (Becton Dickinson) and FlowJo software (TreeStar). A minimum of 25,000 cells was analyzed for each sample.

CTL coculture assays.

CEM-GXR cells were transduced with a murine stem cell virus-derived retroviral vector (Clontech) encoding HLA-B2705 and selected in R10 medium containing 5 μg/ml puromycin. Cells were inoculated with WT or variant viruses at an MOI of 0.0015, and the proportion of GFP-expressing cells was verified by FACS on day 2. Virus-infected cells were then washed and incubated in R10-50 medium in the absence or presence of a B27-KK10-restricted CTL clone mixed at an effector-to-target cell ratio of 1:10. To measure viral spread, the proportion of GFP-expressing CEM-GXR cells was assessed by FACS at days 3 through 7.

CsA experiments.

Single-cycle infectivity was determined by inoculation of CEM-GXR reporter cells with VSV-g pseudotyped virus in the presence or absence of cyclosporine A (CsA) (Sigma Aldrich). Cells were preincubated in R10+ medium containing 0.5 μM of drug for 30 min and then inoculated with WT or viral variants. Viral inocula were normalized to 50 ng p24 determined by ELISA. Infected cells were quantified using FACS to measure GFP expression at 48 h postinfection.

Nucleotide substitution rates.

Full-length HIV-1 sequences derived from seven subjects with acute HIV-1 infection and 6 months thereafter (37) were analyzed to determine the relative rates of each specific type of synonymous nucleotide substitution. All sequences were derived from proviral DNA, with hypermutated sequences excluded. Sequences correspond to accession numbers DQ676870 to DQ676875, DQ676877, DQ676878, DQ676880, DQ676881, DQ676883, DQ676884, DQ676886, and DQ676887.

Statistical analysis.

All analyses were performed using Prism 4.0 (GraphPad).

RESULTS

Alternative escape mutations in the B27-KK10 epitope impair viral replication to a lesser degree than R₂₆₄K.

Viral escape from the immunodominant HLA-B27-restricted KK10 epitope (KRWIILGLNK) in Gag has been described to typically consist of an early L₂₆₈M mutation arising at position 6 (P6) of the epitope, which is followed many years later by an R₂₆₄K mutation at the anchor residue P2 (26, 39). The L₂₆₈M mutation can partially impair TCR recognition (38), while R₂₆₄K imparts a substantial impact on binding to HLA-B27 and epitope presentation (4, 24, 26, 44). R₂₆₄T, R₂₆₄Q, and R₂₆₄G have been described as alternative escape mutations at P2 of KK10 (Table 1), although they occur at much lower frequencies than R₂₆₄K (4, 8, 17, 24, 33, 48). In a recent study of chronic clade B-infected subjects, we confirmed these results, with R₂₆₄K selected in 13 of 15 (87%) HLA-B27⁺ subjects developing mutations at P2 of KK10 (52) and with the R₂₆₄T mutation observed in the remaining two cases. The predominant R₂₆₄K mutation results in a significant defect in HIV-1 replication capacity, which can be restored through the generation of a compensatory substitution, S₁₇₃A (52).

We therefore hypothesized that differences in the replicative capacities of the respective escape variants may contribute to the overall frequencies of the mutations observed in vivo. To examine this, variant NL4-3 proviruses were constructed, (Table 2) containing either the R₂₆₄T (RT) or the R₂₆₄Q (RQ) mutation, as well as both in combination with the L₂₆₈M substitution observed in vivo (RTLM and RQLM) (4, 17, 24, 48). Since the R₂₆₄G escape mutation has been associated in some cases with an upstream E₂₆₀D mutation (ED) (33), this escape variant was engineered as single (RG), double (EDRG), and triple mutant (EDRGLM) forms. Replicative capacities of these variants were compared to WT NL4-3 and to viruses containing the predominant R₂₆₄K and L₂₆₈M escape mutations (RKLM). In addition, the compensated form of RKLM containing S₁₇₃A was included (SARKLM) (52). Viral particles were produced by transfection of HEK293T cells, and the kinetics of replication were evaluated by measuring the percentage of infected cells over 1 week using a CEM-based GFP reporter T-cell line, CEM-GXR (12).

TABLE 2.

Viral constructs and characteristics^e

Variant or WT	Mutation(s)	Observed in vivo	Replicative capacitya	HLA-B27 bindingb	% Inhibition by CTLc	Influence of CsA on infectivity	P2 substitution favored by mutational bias
WT		Yes	++	++	97.5	↓	NA
LM	L268M	Yes	++	++	ND	↓	NA
RK	R264K	No	−	+/−	1	↑	Yes
RKLM	R264K, L268M	Yesd	−	+/−	34.6	↑	Yes
SARKLM	S173A, R264K, L268M	Yes	+	+/−	37.8	↓	Yes
RT	R264T	Yes	+	−	39.8	↓	No
RTLM	R264T, L268M	Yes	+	−	20	↓	No
STRTLM	S173T, R264T, L268M	Yes	+	−	ND	ND	No
RQ	R264Q	No	+	−	73.1	↓	No
RQLM	R264Q, L268M	Yes	+	−	34.4	↓	No
RG	R264G	Yes	+	−	33.8	↓	No
ED	E260D	Yes	++	++	95.4	↓	NA
EDRG	E260D, R264G	Yes	+	−	23.7	↓	No
EDRGLM	E260D, R264G, L268M	No	−	ND	ND	ND	No
EDRK	E260D, R264K	No	−	+/−	ND	ND	Yes

^aMeasured as viral spread in CEM-GXR cell culture; ++, replicates at WT level or more than WT level; +, replicates at 50 to 100% of WT level; −, replicates at <10% of WT level.

^bMeasured by binding of peptide to HLA-B27 molecules; ++, binding capacity same as WT; +/−, binding capacity 1 to 2 log lower than WT; −, binding capacity >2 log lower than WT.

^cMeasured as percent reduction in GFP-positive cells compared to no CTL at day 7 after infection.

^dOnly in association with compensatory mutation.

^eND indicates “not determined”; NA indicates “not applicable”; ↓, decreased infectivity; ↑, increased infectivity.

The R₂₆₄T substitution appears to represent the second most frequent variant at P2 of KK10, according to an analysis of several HLA-B27⁺ subjects (52), although at the moment, it is not observed in sequences available in the LANL HIV sequence database (www.hiv.lanl.gov) (Table 1). Both RT and RTLM displayed a reduction in replicative capacity, having infected 10 times (RTLM) and 3 times (RT) fewer cells than WT NL4-3 after 7 days in culture (Fig. 1A). In contrast to RK (and RKLM; data not shown), which failed to replicate, RT and RTLM were capable of spreading in the culture. Notably, however, both RT and RTLM replicated slower than the SARKLM variant, which nearly achieved WT levels of replication. While not observed in vivo, we also combined the upstream S₁₇₃A mutation with the RT and RTLM variants, but no improvement in replication was observed for these viruses (data not shown). In previously published sequences, one instance with a combination of an S₁₇₃T mutation with R₂₆₄T was described (52), but in our assay S₁₇₃T did not compensate for the fitness defect of the RT, RTLM, or RK variant (data not shown). Therefore, the R₂₆₄T mutation impaired viral replication to a lesser degree than R₂₆₄K; however, the defect for R₂₆₄T was more severe than for the compensated R₂₆₄K escape variant SARKLM. Similar to RT and RTLM, both the RQ and RQLM variants displayed lower rates of viral replication than WT NL4-3 (Fig. 1B), and these variants replicated substantially better than viruses encoding R₂₆₄K, but less well than the compensated SARKLM variant.

FIG. 1.

Alternative escape mutations in KK10 impact viral replication in vitro to a lesser extent than R₂₆₄K. (A to C) One million CEM-GXR cells were incubated with WT or variant virus, and viral spread over 8 days was determined by flow cytometry. In all assays, the proportion of cells infected by RK declined over time, while SARKLM spread efficiently. RT and RTLM (A) as well as RQ and RQLM (B) replication was reduced by up to 1 log compared to the WT and by up to 0.5 log compared to SARKLM. Replication of RG (C) was reduced 1.5 logs compared to the WT and combining RG with E₂₆₀D (EDRG) compensated for this impairment. EDRGLM as well as EDRK failed to spread throughout the culture. Results shown are representative for three independent experiments. (D) Replication of KK10 variant viruses was examined using primary cells. PBMC from a healthy HIV-negative individual were isolated and CD4⁺ T cells were stimulated for 4 days. Cells were inoculated with WT NL4-3 or KK10 variants (MOI = 0.02), and viral replication was measured by detection of intracellular p24. The proportion of CD4⁺ p24⁺ cells at each time is displayed. As seen using CEM-GXR cells, the RK and RKLM variants were the most defective strains tested, while the SARKLM variant replicated better than the alternative escape mutants. Results are representative of three independent assays using PBMC from different donors.

Viral escape in KK10 through an R₂₆₄G mutation has also been described, although it is consistently in the absence of the early L₂₆₈M mutation (4, 8, 33). Furthermore, the neighboring mutation E₂₆₀D has been proposed to function as a compensatory mutation for R₂₆₄G (33). In examining these variants, we observed that replication of RG was more impaired than that of either RT or RQ, but that this variant still spread substantially better than RK (Fig. 1C). Incorporation of E₂₆₀D into RG (EDRG) improved its replicative capacity, although this variant still did not spread as efficiently as SARKLM or WT NL4-3. Notably, combining the E₂₆₀D mutation with the RK variant (EDRK) did not restore the replicative defect for RK. Therefore, E₂₆₀D functions to compensate only for the R₂₆₄G escape mutation, not for R₂₆₄K. Finally, a severe replication defect was observed when we combined L₂₆₈M with E₂₆₀D/R₂₆₄G (EDRGLM), which might explain the absence of this particular combination of mutations in vivo.

To ensure that these differences in replication capacities were not merely due to the reporter cell system used for these studies, we assessed the ability of each variant to spread in primary T-cell cultures. As shown in Fig. 1D, we observed a similar hierarchy of viral replication of the variants in primary T cells, verifying results from the reporter cell assays. Namely, the RTLM, RQLM, and EDRG variants spread more efficiently than RK and RKLM viruses, but each appeared less fit than the WT or the SARKLM variant.

Together, these results indicate that each of these CTL escape mutations in KK10 (R₂₆₄K, R₂₆₄T, R₂₆₄Q, and R₂₆₄G) imparts a defect in viral replicative capacity. Notably, each of the less frequently observed variants at this residue replicated better than the predominant escape mutation R₂₆₄K, but none of these alternative variants spread as well as the compensated SARKLM virus, which appeared similar to the WT in our assays. Overall, in vitro replication capacities corresponded well with the observed in vivo frequencies of the particular KK10 epitope variants, suggesting that viral fitness may be a key determinant during in vivo selection of CTL escape mutants.

Alternative CTL escape mutations in KK10 are dependent upon CypA for replication.

The severe fitness defect of the R₂₆₄K mutant is due in part to its altering of critical interactions between HIV-1 capsid and the host cellular protein CypA (52), where, unlike with wild-type HIV-1, the R₂₆₄K mutant is unable to replicate in the presence of CypA. Binding of CypA to the viral capsid is normally necessary for efficient infection by HIV-1 and is believed to regulate the uncoating process or to impair recognition of the capsid by host restriction factors (10, 19, 21, 58). The replication competence of the R₂₆₄K variant is partially recovered by treating target cells with CsA, a drug that binds to CypA and competes with HIV-1 for this interaction, or through replication in cell lines deficient in CypA (52). Notably, the normal CypA-dependent phenotype could be restored by introduction of the S₁₇₃A compensatory mutation into the R₂₆₄K variant virus (52). In order to determine whether a similar CypA-mediated impact on viral replication was induced in the RT, RQ, or RG escape variant, CEM-GXR cells were treated with 0.5 μM CsA and then infected with the respective viral variants. As expected, addition of 0.5 μM CsA increased the infectivity of the RK variant more than sevenfold (Table 2 and Fig. 2). In contrast, we observed declines in infectivity levels between two- and 2.5-fold for all of the alternative escape variants, which were similar to that of the WT virus, suggesting that only the R₂₆₄K mutation has a significant impact on CypA interactions (Table 2 and Fig. 2). Together with the replication data presented in Fig. 1, these results suggest that the less frequent selection of the alternative escape mutations at residue 264 is not due to a greater impact of these substitutions on p24 function than that of R₂₆₄K.

FIG. 2.

Infectivity of the alternative escape variants is reduced in the presence of CsA. CEM-GXR cells were infected with VSV-g-pseudotyped WT, RK, SARKLM, ED, RG, EDRG, RT, RTLM, RQ, and RQLM viruses (50 ng p24 equivalent each) in the presence (+) or absence (−) of 0.5 μM CsA. The percentage of GFP-expressing cells 48 h postinfection is shown. The percentage of infected cells for RK increased 7.5-fold, while the infectivity levels of all other viruses were reduced between 1.5- and 2.5-fold in the presence of the drug. Results displayed are mean values for duplicate parallel cultures representative of two independent experiments.

Rare alternative CTL escape mutations in KK10 efficiently impair binding to HLA-B27.

We next hypothesized that differences in the relative ability of KK10 mutations to escape CTL-mediated immune pressure might contribute to their selection frequencies in vivo. Since R₂₆₄ serves as the P2 binding site for the HLA-B27 molecule, we compared the HLA binding capacities of KK10 epitope variants to that of the WT peptide. As a control, we examined an HLA-B27-restricted influenza nucleoprotein epitope (SRYWAIRTR; positions 383 to 391; SR9) (29). HLA-B27 binding capacities were determined using a competition assay in which a radiolabeled standard peptide-HLA-B27 complex was incubated in the presence of increasing concentrations of unlabeled variant peptide, as described previously (54, 55). As expected, the L₂₆₈M mutation had little impact on HLA binding, while the R₂₆₄K mutation, alone or in combination with L₂₆₈M, impaired binding by nearly 2 logs (Fig. 3, Table 2, and Table 3) (26). Binding capacities for the less frequent escape mutants (R₂₆₄G, R₂₆₄T, and R₂₆₄Q) were reduced even more substantially, by nearly 3 logs or greater, and similar results were observed for R₂₆₄T and R₂₆₄Q in combination with L₂₆₈M, which have been associated in vivo (4, 24, 48). Given that an IC₅₀ of 500 nM is generally considered a physiological cutoff for peptide binding to MHC class I (53), these data indicate that all P2 mutations in KK10 substantially abrogated binding to HLA-B27. Notably, however, the less frequent mutations at P2 had an even greater impact on binding than R₂₆₄K in this assay. The results of this direct comparison of epitopes with all known mutations in KK10 agree with HLA binding capacities for the R₂₆₄T, R₂₆₄G, and R₂₆₄Q variant peptides which have previously been documented independently using differing assays (4, 24, 44). Therefore, an ability of the alternative escape mutations in KK10 to bind HLA did not appear to account for their decreased frequency.

FIG. 3.

Escape mutations at the P2 anchor residue abrogate binding to HLA-B27. A radiolabeled standard peptide (FRYNGLIHR) and variant KK10 peptides at different concentrations were incubated with purified HLA-B2705 molecules for 48 h, and binding capacity was determined by analyzing the percentage of MHC-bound radioactivity. The IC₅₀ in nM is displayed for the different peptides (influenza-SR9, SRYWAIRTR; WT-KK10, KRWIILGLNK; LM, KRWIIMGLNK; RK, KKWIILGLNK; RKLM, KKWIIMGLNK; RG, KGWIILGLNK; RT, KTWIILGLNK; RTLM, KTWIIMGLNK; RQ, KQWIILGLNK; and RQLM, KQWIIMGLNK [boldface indicates differences in residues compared to WT KK10]). Binding capacities for WT-KK10, influenza-SR9, and LM were not significantly different, while the binding of RK and RKLM was reduced by 2 logs. Peptides with alternative mutations without LM (RG, RT, and RQ) or with LM (RTLM and RQLM) reduced binding capacities by nearly 3 logs or greater. Raw binding capacity values (IC₅₀) are shown in Table 3.

TABLE 3.

HLA-binding capacities of variant HIV-1 B27-KK10 peptides and influenza SR9

Variant peptide, SR9, or WT	Peptide sequencea	IC50 (nM)
SR9	SRYWAIRTR	15
WT	KRWIILGLNK	8.3
LM	KRWIIMGLNK	13
RK	KKWIILGLNK	886
RKLM	KKWIIMGLNK	741
RG	KGWIILGLNK	6,816
RT	KTWIILGLNK	36,644
RTLM	KTWIIMGLNK	32,287
RQ	KQWIILGLNK	5,462
RQLM	KQWIIMGLNK	67,680

^aBoldfaced amino acids indicate differences in residues compared to WT KK10.

In vitro inhibition of R₂₆₄ variants by KK10-specific CTL.

Despite the low HLA-B27 binding affinities for KK10 peptides with mutations at the P2 anchor position, it is possible that differences in the levels of residual CTL recognition and/or processing efficiency of these epitopes could contribute to their propensity to be selected in vivo. To address these issues directly, a CEM-GXR cell line was generated that stably expressed the HLA-B2705 molecule. We infected these cells with variant viruses and measured viral replication capacity in the absence or presence of a highly active CTL clone recognizing the WT KK10 epitope. KK10-restricted CTL inhibited WT virus replication significantly during the 7-day coculture assay (Fig. 4A), while the compensated SARKLM variant was largely resistant to the effects of CTL (Fig. 4B). Similar assays were conducted for the alternative KK10 variants, and a wide range of viral replication capacities was observed in the absence and presence of CTL (Fig. 4C and Table 2). Notably, the compensated SARKLM virus appeared to be the most fit escape variant in the absence and presence of CTL, followed by the R₂₆₄T-containing variants. In addition, the remaining CTL-mediated inhibition among the KK10 variants in these coculture assays suggested that at least some CTL recognition still took place for the KK10 mutants tested. Together with the data represented in Fig. 1 and 3, our results indicate that viral replication capacity and the degree of residual CTL recognition are major driving forces during in vivo selection of KK10 escape variants.

FIG. 4.

The SARKLM variant displays the highest replication capacity in the presence of KK10-specfic CTL. CEM-GXR cells stably expressing HLA-B2705 were inoculated with WT NL4-3 or KK10 variants (MOI = 0.0015). KK10-specific CTL were added to one replicate well at an effector-to-target cell ratio of 1:10. Viral replication was measured using FACS at days 2 through 7, as described in the legend for Fig. 1, and the proportion of GFP⁺ CEM-GXR cells at each time is shown. Replication of WT virus was substantially inhibited in the presence of CTL (A), but the SARKLM variant was largely resistant to the effects of CTL (B). Replication capacity was highly variable for the different KK10 variants in the absence and presence of KK10-specific CTL, as demonstrated by the proportion of virus-infected cells present at day 7 using this in vitro coculture assay. Overall, however, the SARKLM virus exhibited the highest level of replication in either the absence or presence of CTL (C).

Mutational bias preferentially selects for the R₂₆₄K mutation.

Our results for viral replication and CTL recognition demonstrate that the compensated SARKLM is likely to be the most fit KK10 escape variant among those observed in vivo. However, the alternative escape variants also appeared to retain a relatively high replication capacity, particularly for the R₂₆₄T mutants, in the absence and presence of KK10-specific CTL (Fig. 4C). Nevertheless, the alternative escape mutants are observed less frequently in vivo than would be suggested from our functional studies. We therefore examined whether additional factors might directly influence the underlying evolutionary process at P2 of KK10 toward the R₂₆₄K mutation by analyzing differences in the rates of nucleotide substitutions that may affect codon 264.

It is known that the genomes of retroviruses vary considerably in their nucleotide composition, with a high content of adenosine (A) in the RNA of HIV-1 (7). This imbalance might result from the error proneness of HIV-1 reverse transcriptase, which has been described to preferentially incorporate G-to-T mismatches during minus-strand cDNA synthesis (59). Recently, it has also been suggested that HIV-1 sequence evolution may be influenced by the mutagenic potential of the APOBEC3 family of host proteins, which have been linked to the phenomenon of retroviral G-to-A hypermutation (28, 30). We hypothesized that the rates of the 12 possible nucleotide substitutions that occur during the replicative process of HIV-1 could additionally bias the mutational pathway during escape from CTL pressure against KK10 and hence contribute to the greater in vivo frequency of R₂₆₄K. To determine the rate of accumulation of each nucleotide change, we used previously published longitudinal full-length HIV-1 sequences from seven individuals collected during acute infection and 6 months thereafter (37). To avoid possible influences by particular selection pressures of the host, we disregarded all nonsynonymous mutations and calculated the number of synonymous substitutions that accumulated during this interval, relative to the total number of synonymous mutations possible. Using this approach, the most frequent changes observed during the first 6 months of infection were G-to-A mutations, which accumulated at approximately one mutation per 100 opportunities. The next most frequent synonymous nucleotide change was an A-to-G transition, which was observed at about half that rate (Table 4) (A. Schneidewind and Y. E. Wang, unpublished data).

TABLE 4.

Rates of nucleotide substitutions required for evolution of KK10 escape mutations at position P2

Viral variant or consensus	Nucleotide sequence at codon 264	Required nucleotide substitution(s)	Rate(s) of nucleotide substitution(s)a
Consensus R264	AGA
R264K	AAA	G to A	0.0096
R264G	GGA	A to G	0.0051
R264T	ACA	G to C	0.0000
R264Q	CAA	G to A and A to C	0.0096 (G to A) and 0.0006 (A to C)

^aIn vivo rates of synonymous substitutions as determined by analysis of pairs of longitudinal sequences derived from seven subjects over an interval of 6 months (37).

Of the four different amino acid mutations at P2 of KK10, only the predominant R₂₆₄K mutation derives from the relatively more common transition of G to A (Table 4). In contrast to R₂₆₄K, the R₂₆₄T mutation is generated by transversion from G to C, which happens very rarely according to our analysis. The R₂₆₄G mutation requires a transition from A to G that is only slightly less frequent than the G-to-A substitution, but this escape pathway may be hindered by the typically early generation of the L₂₆₈M substitution, which is incompatible with R₂₆₄G replication in our assays (Fig. 1C). Finally, the R₂₆₄Q substitution could arise through two pathways. The first would require the frequent G-to-A transition (AGA→AAA) followed by an additional A-to-C transversion event (AAA→CAA), which was observed to occur at a rate 16 times lower than that of G to A. Obviously, the occurrence of the AAA intermediate, which codes for lysine (K), might be limited due to the poor replication capacity of the R₂₆₄K variant. Alternatively, the R₂₆₄Q mutation could also arise through an initial A-to-C substitution (AGA→CGA) that would remain synonymous on the amino acid level, followed by a G-to-A substitution (CGA→CAA) generating R₂₆₄Q. However, here the CGA intermediate is rarely observed in vivo (0.11% of sequences in the LANL HIV sequence database; www.hiv.lanl.gov), which might further explain the limited frequency of this particular escape mutation. Therefore, mutational bias at the P2 position of KK10 during evolution of the HIV-1 quasispecies may preferentially select for the R₂₆₄K substitution (Table 2), which could also contribute to the higher observed frequency of this escape mutation.

DISCUSSION

Viral escape in the B27-KK10 epitope in Gag typically occurs through development of an R₂₆₄K mutation. Three alternative escape mutations at the same codon, R₂₆₄T, R₂₆₄Q, and R₂₆₄G, have been reported at lower frequencies (4, 8, 17, 24, 33, 48, 52), although the factors contributing to the preferential selection of R₂₆₄K remain unclear. Here our data illustrate that each of the alternative escape mutations at residue R₂₆₄ was capable of abrogating peptide binding to HLA-B27 to an even greater degree than R₂₆₄K, despite their infrequent selection in vivo. The alternative mutations also impaired viral replication capacity, but surprisingly, each had a less severe effect than the predominant R₂₆₄K mutation, and only the RK variant altered the critical interaction between the capsid and the host protein CypA. Notably, however, the S₁₇₃A-compensated R₂₆₄K variant demonstrated the highest level of replication capacity among known escape variants and restored normal CypA dependence. Moreover, while all of the escape variants successfully evaded inhibition by KK10-specific CTL in coculture, the S₁₇₃A-compensated R₂₆₄K variant replicated most efficiently. Therefore, in vitro replication capacities of KK10 variants in the presence of CTL correlated well to the relative frequencies of escape mutants in B27⁺ individuals in vivo, with SARKLM being selected most often.

During viral escape from the B27-KK10 response, the L₂₆₈M substitution develops significantly earlier than mutations at codon 264 (17, 22, 33, 52). Rather than functioning as a compensatory mutation, however, this L₂₆₈M mutation appears to represent an early partial escape mutation at a TCR contact residue (38, 52) and has little impact on the replicative capacity of HIV-1 (52). In the context of alternative KK10 mutations, R₂₆₄Q was always associated with L₂₆₈M (4, 48), and R₂₆₄T has been described to occur both in the presence and absence of L₂₆₈M (17, 26, 52). We have observed only minor differences in the replicative capacities for the RK, RQ, and RT variants in the presence of L₂₆₈M (Fig. 1A and 1B), further supporting the model that L₂₆₈M functions mainly as an early CTL escape mutation that is recognized by only a subset of KK10-specific TCR clonotypes (38, 52). Indeed, it is likely that a secondary de novo T-cell response against the LM variant drives the subsequent development of mutations at residue R₂₆₄.

The mutation E₂₆₀D is associated with R₂₆₄G and has been postulated to restore a replication defect (33), although in one report CTL escape in KK10 via development of R₂₆₄G was not associated with E₂₆₀D (8). We now demonstrate that E₂₆₀D does indeed compensate for a fitness defect of R₂₆₄G, although it had no similar effect on R₂₆₄K (Fig. 1C). Residue E₂₆₀D lies four amino acids N terminal of residue R₂₆₄, and simultaneous alanine substitutions at codons 260 and 264 have been shown to stabilize the viral core (18) without affecting the rate of capsid assembly (16). In contrast, single alanine substitutions at these charged amino acids led to faster and slower capsid assembly, respectively (16). These data suggest close interactions between residues at codons 260 and 264, whose side chains sit on the same planar face of the molecule, and imply that charge balances in this region might be critical for normal capsid function. Development of the R₂₆₄G mutation is also notable in that it occurs in the absence of L₂₆₈M (8, 33), and we observed that the combination of substitutions E₂₆₀D, R₂₆₄G, and L₂₆₈M (EDRGLM) led to a viral variant that was as severely compromised in its replicative capacity as RK (Fig. 1C). Therefore, the ability to generate the R₂₆₄G escape mutant might also be limited due to the early selection of the LM variant in the majority of HLA-B27⁺ individuals.

The alternative escape mutants efficiently abrogated HLA binding of the KK10 epitope at least as well as the RK variant (Fig. 3), and we observed that escape from CTL pressure was as complete for the alternative variants as for SARKLM in our coculture assays (Table 2). The in vitro replication kinetics of the various alternative escape mutants, including the compensated EDRG, however, were consistently slower than that of SARKLM in the absence or presence of KK10-specific CTL (Fig. 1 and 4). This suggests that a threshold of replicative capacity may exist that determines whether a mutant can dominate in the quasispecies. Indeed, despite its typically late development in vivo, the nearly WT-level replication capacity and profound CTL escape phenotype of SARKLM in our in vitro assays suggest that this variant would have a significant in vivo replicative advantage over the parental virus in the context of an immunodominant KK10-specific CTL response. In contrast, despite similar or even enhanced CTL evasion capabilities, the more substantially impaired replicative capacities observed for the RT/RTLM, RQ/RQLM, and RG/EDRG variants might still outweigh the selective advantage of a successful CTL escape. A potential model to explain the importance of viral replication capacity and evasion from CTL pressure relative to an in vivo threshold for HIV outgrowth is shown in Fig. 5. We propose that during chronic HIV-1 infection, only the SARKLM variant is suitably fit to escape KK10-specific CTL pressure and establish itself as the predominant viral strain in vivo.

FIG. 5.

Model summarizing forces that influence selection of the escape mutations in the B27-KK10 epitope. The relative influence of CTL escape mutations in the KK10 epitope on HLA-B27 binding, in vitro viral replication, and predicted in vivo viral replication capacity in the presence of KK10-specific CTL is illustrated. Viral sequences in the KK10 epitope are shown on the left, and mutations at residue 264 are indicated in red or blue text. The ability of the WT KK10 peptide (yellow) or KK10 P2 variants (red) to bind to HLA-B27 is classified as strong, weak, or very weak based on data presented in Fig. 3. In the panels on the right, in vitro and putative in vivo replication capacities of WT (black), R264K (red), R264T/G/Q (collectively in blue), and SARKLM (green) are displayed. The dashed line represents the hypothesized in vivo replication “threshold” that may need to be overcome for a given variant to dominate the quasi-species.

Several reports have addressed the question of how frequently each of the 12 possible nucleotide substitutions occurs during sequence evolution in HIV-1. The results of these analyses vary substantially, possibly due to differences in the data sets or in the methodologies applied. Goodenow et al. (23) compared short gag and env fragments of the viral quasispecies from seven subjects and noted a predominance of G-to-A substitutions, assuming that minor sequence populations were derived from the dominant ones. In their data set, 100% of all guanosine alterations in gag showed a G-to-A change. Moriyama et al. (42) estimated the relative frequency of nucleotide substitutions at the third codon positions in gag by analyzing published sequences of 15 independent HIV-1 isolates, proposing that 43.8% of all mutations were G-to-A changes, while only 16.4% were A-to-G changes. G-to-C changes and A-to-C changes were not observed in that study. In contrast, Plikat et al. (49) collected longitudinal nef sequences from one subject and reported relative frequencies for A-to-G and T-to-C transitions twofold higher than for G-to-A changes. Our analysis of synonymous nucleotide changes in seven longitudinal viral sequence pairs supports that the predominant nucleotide substitution introduced during the replicative cycle of HIV-1 is a G-to-A transition, which matches the change required to generate the R₂₆₄K mutation. Nucleotide substitutions necessary for evolution of the other KK10 escape variants, including R₂₆₄T (G to C), R₂₆₄G (A to G), and R₂₆₄Q (two mutations: A to C and G to A) arise significantly less often (Table 4). Our results correlate well with a recently published estimation of the substitution frequency during HIV evolution in vivo, which was derived from a considerably large data set of over 5,000 pol population sequence pairs (15).

The propensity for G-to-A substitutions during HIV-1 replication could therefore additionally hamper the generation of R₂₆₄T, R₂₆₄G, or R₂₆₄Q, while favoring the occurrence of R₂₆₄K. The significance of a mutational bias during viral replication to select for particular variants with even suboptimal replication capacity is supported by a study demonstrating that the initial selection of the less fit lamivudine drug-resistant mutant M₁₈₄I could be explained by its underlying G-to-A substitution (34). At later times, the M₁₈₄V variant displaying an improved replicative capacity becomes dominant, but this delay was attributed to its requirement for the infrequent A-to-G transition. In the context of CTL pressure, therefore, mutational bias may also alter the frequency of escape variants, despite seemingly significant differences in costs to replicative capacity. However, the notion that the nucleotide substitution bias that occurs during viral evolution also influences the preferential selection for R₂₆₄K is tempered by its escape pathway that requires two independent amino acid substitutions (i.e., R₂₆₄K and the S₁₇₃A compensatory change), while any of the alternative escape mutations at the P2 position of KK10 do not absolutely rely on additional mutations in vitro. S₁₇₃A itself is generated by a T-to-G substitution, which we observed at a rate of 0.0021. Since the S₁₇₃A mutation in the absence of R₂₆₄K has a relatively modest impact on replication capacity (52) and is present in 39 out of 707 (5.5%) sequences in the LANL HIV database (www.hiv.lanl.gov), it is likely that it occasionally arises as a minor species and therefore may not provide a substantial barrier to the generation of R₂₆₄K in the context of HLA-B27. Taking this into consideration, the existence of a particular fitness threshold, which must be surpassed by a viral mutant for it to become the dominant form in the quasispecies, seems even more probable. Viral escape in KK10 might generally be delayed until R₂₆₄K is generated on a viral backbone with a preexisting S₁₇₃A mutation, with the alternative escape variants failing to exceed the fitness level required for outgrowth. In this case of a viral backbone incorporating a preexisting S₁₇₃A mutation, the nucleotide substitution bias would then additionally favor the generation of an R₂₆₄K mutation over any of the alternative substitutions at the P2 position of the KK10 epitope. Furthermore, it is possible that each of the alternative escape variants also requires particular compensatory mutations in order to establish a variant quasispecies with sufficient replication competence to outcompete viruses with the WT or L₂₆₈M forms of the KK10 epitope. Analogous to the requirement for S₁₇₃A, the requirement for such compensatory mutations could then substantially reduce the chance for an alternative escape mutation in KK10 to arise. Unfortunately, due to the limited number of published sequences with R₂₆₄T and R₂₆₄Q mutations, we have been unable to identify potential compensatory changes (or in the case of R₂₆₄G, to identify others besides E₂₆₀D, which itself requires a rare A-to-C substitution) for the fitness defects associated with these mutations.

Despite the severe replicative defect of R₂₆₄K in vitro, this mutation is selected as the predominant substitution to escape in vivo immune pressure by the B27-KK10-specific CD8⁺ T-cell response. We have previously shown that the R₂₆₄K mutation occurs in nearly all cases concomitantly with the rare compensatory mutation S₁₇₃A. Therefore, in HLA-B27⁺ subjects, mutational escape at codon 264 is typically delayed until R₂₆₄K is generated together with S₁₇₃A. Presumably, HIV-1 remains under continuous, efficient CTL pressure for this extended length of time, which may contribute to long-term viral suppression (26, 56). We propose that the lower frequency of R₂₆₄G, R₂₆₄T, and R₂₆₄Q versus R₂₆₄K at P2 of the KK10 epitope can be explained by a combination of several viral constraints. We observed a replicative capacity for all of the alternative escape variants lower than that of the compensated SARKLM mutant in both the absence and the presence of KK10-specific CTL, indicating that a certain fitness threshold may need to be obtained before a variant can outgrow WT virus to become the dominant species. Additionally, we noted that R₂₆₄G displays a severe replication defect when combined with L₂₆₈M, which likely limits development of the RG variant in the context of this earlier escape mutation. Finally, we identified a bias in the nucleotide error rate during viral replication that may contribute to the preferential generation of R₂₆₄K. These observations underscore the notion that the KK10 epitope and especially the Gag codon 264 are located in a highly sensitive area of capsids with structural and functional constraints that limit viral escape from HLA-B27 CTL pressure. This region of Gag might therefore function as an attractive target for efficient immunomodulatory vaccine approaches.