Gag Cytotoxic T Lymphocyte Escape Mutations Can Increase Sensitivity of HIV-1 to Human TRIM5α, Linking Intrinsic and Acquired Immunity^,

Abstract

Although laboratory-adapted HIV-1 strains are largely resistant to the human restriction factor TRIM5α (hTRIM5α), we have recently shown that some viruses carrying capsid (CA) sequences from clinical isolates can be more sensitive to this restriction factor. In this study we evaluated the contribution to this phenotype of CA mutations known to be associated with escape from cytotoxic T lymphocyte (CTL) responses. Recombinant viruses carrying HIV-1 CA sequences from NL4-3 and three different clinical isolates were prepared, along with variants in which mutations associated with CTL resistance were modified by site-directed mutagenesis, and the infectivities of these viruses in target cells expressing hTRIM5α and cells in which TRIM5α activity had been inhibited by overexpression of TRIM5γ were compared. For both hTRIM5α-sensitive viruses studied, CTL-associated mutations were found to be responsible for this phenotype. Both CTL resistance mutations occurring within HLA-restricted CA epitopes and compensatory mutations occurring outside CTL epitopes influenced hTRIM5α sensitivity, and mutations associated with CTL resistance selected in prior hosts can contribute to this effect. The impact of CTL resistance mutations on hTRIM5α sensitivity was context dependent, because mutations shown to be responsible for the TRIM5α-sensitive phenotype in viruses from one patient could have little or no impact on this parameter when introduced into another virus. No fixed relationship between changes in hTRIM5α sensitivity and infectivity was discernible in our studies. Taken together, these findings suggest that CTL mutations may influence HIV-1 replication by modifying both viral infectivity and sensitivity to TRIM5α.

INTRODUCTION

Cytotoxic T lymphocytes (CTLs), which target virus-infected cells through recognition of viral peptides presented by class I HLA molecules, play a critical role in controlling HIV-1 replication in the infected host (26, 72). Numerous studies have demonstrated that the HLA alleles expressed by infected patients have an important bearing on viral load at set point, the rate of CD4⁺ T cell decline, and progression to AIDS (12, 26, 34). Certain HLA alleles (e.g., B*57, B*27, B*14, and B*52) are more frequent in patients that spontaneously control HIV-1 replication to low levels, reflecting differences in the nature of the HLA-viral peptide interactions permitted by these “protective” alleles, including which viral sequences (epitopes) are presented and how the epitope binds to the HLA molecule (38, 54).

In turn, HIV-1 attempts to escape from this immune pressure by introducing mutations in or near the targeted epitopes that either disrupt the processing of the epitopes, impair their binding to the restricting HLA molecule, or weaken interactions with the T cell receptor (3, 18, 33, 57). Predictable patterns of escape mutations in epitopes recognized by a given HLA allele have been identified in population studies (9, 50). Importantly, resistance mutations in some epitopes can impair viral replicative capacity, and this is observed more frequently for mutations occurring in CA than for those occurring in other viral proteins (42, 44, 49, 71). Thus, resistance mutations in CA epitopes are often accompanied by compensatory mutations that serve to correct the impairment in viral replication produced by the resistance mutations (8, 9, 33, 61, 62). The observations that the presence of Gag CTL resistance mutations correlates with lower viral loads in infected patients and that such mutations revert following transmission of the virus to HLA-discordant recipients indicate that the restoration of viral replicative capacity by compensatory mutations is often incomplete (3, 7, 16, 23, 40, 69).

The reason that CTL escape mutations in CA are more likely to be deleterious than those in other viral proteins is not fully understood. The CA is a highly conserved structure, and it is known that many mutations can disrupt optimal CA stability (6, 19, 20). In addition, unlike other viral proteins, the CA is targeted by the intrinsic restriction factor TRIM5α (43, 52, 66, 70).

TRIM5α interacts with the mature capsid lattice after its entry into target cells; this interaction can directly promote rapid disassembly of the capsid structure, thereby preventing the completion of reverse transcription (67). In addition, TRIM5α possesses an E3 ubiquitin ligase activity that is amplified following interaction of TRIM5α with the capsid, thereby stimulating a cascade that both promotes innate immune signaling and contributes directly to viral restriction by TRIM5α (35, 55). Restriction exerted by TRIM5α on retroviral replication varies according to the virus and to the host species, reflecting pressure exerted on the TRIM5 gene over the course of mammalian evolution by retroviral infections (31). HIV-1 is strongly sensitive to restriction by rhesus macaque TRIM5α (66). In contrast, HIV-1 was until recently considered largely resistant to human TRIM5α (hTRIM5α); results from several groups have shown that laboratory-adapted HIV-1 strains such as NL4-3 show only very modest (approximately 2-fold) sensitivity to hTRIM5α expressed at physiological levels (27, 32, 65, 68). We have recently shown, however, that HIV-1 carrying CA sequences derived from some clinical isolates are more sensitive to hTRIM5α than NL4-3 and that increasing the expression of hTRIM5α in target cells, including CD4⁺ T cells, by treatment with alpha interferon (IFN-α) further increases the activity of hTRIM5α against these viruses (5). Thus, it is possible that the introduction of mutations in CA in response to CTL pressure, in addition to directly modifying the functional properties of CA, could alter its sensitivity to recognition by TRIM5α, a possibility that has been suggested by other authors (29, 49).

To study this question, we have evaluated the role played by mutations associated with CTL resistance in producing the increased TRIM5α sensitivity observed for the viruses identified in our previous study that carried CA sequences derived from clinical isolates. We demonstrate that both CTL resistance mutations and compensatory mutations can result in increased viral sensitivity to hTRIM5α, a phenomenon that involves not only mutations likely to have developed de novo in the current host but also mutations selected in prior hosts. The impact of CTL resistance and compensatory mutations on hTRIM5α sensitivity, however, is strongly dependent on the overall context of the CA sequence in which they occur. These findings suggest that modifications in hTRIM5α sensitivity resulting from CTL mutations are likely to influence HIV-1 replicative capacity in vivo but that the impacts of this effect are likely to be different for different patients.

MATERIALS AND METHODS

Cell culture.

U373-X4 cells in which hTRIM5α activity had been inhibited by stable overexpression of untagged TRIM5γ and the corresponding control cell line that overexpresses LacZ were established by transduction with pLenti6/V5-D-TOPO-based vectors using previously described techniques (5). Two independently derived pairs of these transduced cell lines were studied in parallel in all studies reported here. The infectivity of N-tropic and B-tropic murine leukemia viruses (N-MLV and B-MLV) in IFN-α-pretreated U373-X4-TRIM5γ cells and U373-X4-LacZ cells was evaluated as previously described (5) before and after completion of these studies. In both cases, U373-X4-LacZ cells continued to strongly inhibit the infectivity of N-MLV, whereas similar titers of N-MLV and B-MLV were observed in U373-X4-TRIM5γ cells (data not shown).

Viruses.

The production of the vesicular stomatitis virus (VSV)-pseudotyped pNL4-3-based recombinant viruses that contain a deletion in env, that express Renilla luciferase in place of nef, and whose Gag-PR sequences were derived from clinical isolates (NRC1, NRC2, and NRC10; accession numbers JN408075 to JN408077) or from NL4-3 has previously been described (5, 45). To create variants of these viruses, the BssHII-ClaI fragment from the proviral plasmid was subcloned into a modified pBluescript vector in which one of the two BssHII sites in the polylinker had been removed and one or more nucleotide substitutions were introduced by the procedures described in Tables S1 and S2 in the supplemental material. The modified BssHII-ClaI fragment was then reinserted into the original plasmid. All constructions were verified by sequencing the entire Gag-PR region of the proviral plasmid. VSV-pseudotyped viral stocks were produced as previously described and stored at −80°C (5, 45).

Measurement of viral sensitivity to TRIM5α and infectivity.

The procedures used to measure sensitivity to TRIM5α and viral infectivity are summarized in Fig. 1. HIV-1 infectivity was measured by determining luciferase activity in target cells after infection as previously described (5, 45). Briefly, U373-X4 cells overexpressing LacZ or TRIM5γ were incubated in the presence of 1,000 U/ml IFN-α for 24 h to increase the expression of TRIM5α and infected in triplicate with serial 2-fold dilutions of virus (nominally 20, 10, and 5 μg p24/ml) in the presence of 2 μg/ml DEAE-dextran, and luciferase activity was measured 40 h after infection (Fig. 1A). The actual p24 contents of the virus preparations used to infect the cells were measured in each experiment, and these values were used to calculate slopes. The mutations introduced in CA were shown not to influence quantification of p24 (data not shown). The slope (RLU versus ng of p24) was calculated for each virus in each cell line. For all viral variants, RLU values were linearly related to the amount of virus used to infect the cells. To assess sensitivity to TRIM5α (Fig. 1B), the results were expressed as a ratio and normalized to that obtained for NL4-3 according to the formula (slope for U373-X4-TRIM5γ/slope for U373-X4-LacZ)/(slope for NL4-3 in U373-X4-TRIM5γ/slope for NL4-3 in U373-X4-LacZ). To evaluate viral replicative capacity, as measured in the target cells in which hTRIM5α activity had been inhibited by overexpression of TRIM5γ (Fig. 1C), the slope of RLU versus amount of p24 obtained for U373-X4-TRIM5γ cells was normalized to that obtained for NL4-3 in the same experiment.

Fig. 1.

Measurement of sensitivity to hTRIM5α and viral infectivity. (A) Recombinant proviruses containing a deletion in env, expressing Renilla luciferase in place of nef, and carrying the Gag-PR sequence from NL4-3 or the clinical isolate NRC10 were prepared, and VSV-pseudotyped viral stocks were obtained by transfecting 293T cells with the proviral construct and a plasmid expressing the VSV G protein. U373-X4 cells transduced with a control vector causing them to overexpress LacZ (U373-X4-LacZ) or with a vector causing them to overexpress TRIM5γ (U373-X4-TRIM5γ) were infected with three dilutions of virus. Luciferase activity (relative light units [RLU]) was measured 40 h after infection and plotted as a function of amount of virus, and the slopes of the curves (RLU/ng p24) were determined. (B) To measure viral sensitivity to hTRIM5α, the ratio of the slope obtained for U373-X4-TRIM5γ cells to that obtained for U373-X4-LacZ cells was determined for each viral variant. The hTRIM5α sensitivity of NRC10 relative to that of NL4-3 was determined by calculating the ratio of these two values. (C) To assess infectivity, measurements of viral infectivity in the target cells in which hTRIM5α activity had been inhibited by overexpression of TRIM5γ (U373-X4-TRIM5γ cells) were used; values were expressed as percentages of results obtained for NL4-3 in the same experiment.

Statistical analysis.

All results are expressed as means ± standard errors of the means (SEM) unless otherwise indicated. Groups within a single series of experiments were compared by analysis of variance (ANOVA) using the R statistical package (version 2.11.1). Posttest comparisons were performed using t tests with nonpooled standard deviations (SD) and adjusted for multiple comparisons by the method of Holm (1). Correlations were evaluated using the Pearson test.

RESULTS

Evaluation of hTRIM5α sensitivity.

To evaluate the effect of changes in the HIV-1 CA on the sensitivity to hTRIM5α, we used VSV-pseudotyped NL4-3-based recombinant viruses that contained a deletion in env, that expressed Renilla luciferase in the place of nef, and that had Gag-protease (Gag-PR) sequences derived from NL4-3 or from clinical isolates NRC1, NRC2, and NRC10. As described below, the CA sequences of these viruses were modified by site-directed mutagenesis, and these viruses were used to infect U373-X4 cells that had been transduced with vectors resulting in the stable expression of either TRIM5γ or, as a control, β-galactosidase. The truncated TRIM5γ isoform forms heterodimers with TRIM5α, blocking its activity, and the cells overexpressing TRIM5γ did not restrict TRIM5α-sensitive virus, as demonstrated by the similar infectivities of N-MLV and B-MLV viruses in this cell line (5). In the experiments presented here, sensitivity to TRIM5α was expressed as the ratio of luciferase activity measured in U373-X4-TRIM5γ cells to that measured in U373-X4-LacZ cells after normalization with results obtained in the same experiment for viruses carrying the NL4-3 CA (Fig. 1). Consistent with prior studies performed using other cell types (27, 32, 65, 68), the infectivities of viruses carrying the NL4-3 CA were 2-fold higher in U373-X4-TRIM5γ cells than in U373-X4-LacZ cells. Thus, the infectivity of a virus that is 10 times more sensitive to hTRIM5α than NL4-3 was approximately 20-fold higher in the U373-X4-TRIM5γ cells than in the control cells.

We have previously found that the sensitivities to TRIM5α of recombinant viruses NL4-3, NRC1, NRC2, and NRC10 were similar when VSV-pseudotyped viruses and those expressing the HIV-1 envelope were compared (5). Thus, the entry pathway used during infection does not influence this parameter.

CTL resistance mutations induce susceptibility to hTRIM5α in viruses from patient NRC10.

We have previously shown that a recombinant virus carrying Gag-PR sequences from plasma viruses obtained from patient NRC10 (whenever a designation such as NRC10 refers to a patient rather than a virus, “patient” appears before the designation) was more sensitive to hTRIM5α than a virus expressing Gag-PR sequences from NL4-3 (5). Three different Gag-PR sequences obtained from this patient showed increased sensitivity to hTRIM5α, indicating that this property was shared by the swarm of viruses from this patient and was not due to a rare deleterious point mutation (data not shown). Replacement of the CA coding sequence of this isolate with the corresponding sequence from NL4-3 completely eliminated hTRIM5α sensitivity, indicating that this phenotype mapped uniquely to CA (data not shown).

Patient NRC10 is HLA-B27 positive (Table 1), and the CA sequences of viruses from this patient contained the classical R264K and L268M resistance mutations in the HLA-B27 restricted epitope KK10, as well as the I267V substitution, which has not been implicated in CTL resistance (Fig. 2 and Table 2). Viruses expressing KK10 resistance mutations often also express the S173A compensatory mutation, which alleviates the decrease in replicative capacity resulting from R264K (61). Instead of S173A, viruses from patient NRC10 carried the S173T mutation, which has also been observed in association with KK10 resistance mutations (61). To evaluate the impact of the S173T, R264K, I267V, and L268M mutations on sensitivity to hTRIM5α, one or more of these mutations were removed from the NRC10 sequence by site-directed mutagenesis and the resulting viruses were tested for susceptibility to hTRIM5α (Fig. 3A). As previously reported (5), NRC10 was 7.2-fold more sensitive to hTRIM5α than NL4-3. Removing the compensatory S173T mutation alone (NRC10-T173S) increased sensitivity to 16.1 times that of NL4-3. In contrast, removing S173T together with resistance mutation R264K alone (NRC10-T173S-K264R) or together with both R264K and L268M (NRC10-T173S-K264R-M268L) reduced hTRIM5α sensitivity to 3.5 and 3.7 times that of NL4-3, respectively. Replacing S173T by S173A (NRC10-T173A) also reduced hTRIM5α sensitivity to 3.6 times that of NL4-3. Reverting the I267V mutation in the NRC10-T173S-K264R-M268L variant (producing NRC10-T173S-K264R-V267I-M268L) had no effect on sensitivity to hTRIM5α. Thus, removing all the mutations implicated in resistance to CTLs targeting KK10 did not completely eliminate the sensitivity of this virus to hTRIM5α. Taken together, these results indicate that (i) the R264K mutation increases the sensitivity of NRC10 to hTRIM5α and (ii) the impact of R264K can be partially (S173T) or completely (S173A) eliminated by compensatory mutations but that (iii) other mutations independent of those implicated in resistance to CTLs targeting KK10 also contribute to the sensitivity of NRC10 to hTRIM5α.

Table 1.

Class I HLA genotypes of patients evaluated in this study

Patient	Class I HLA types
	HLA-A	HLA-B	HLA-C
NRC10	A*01, A*11	B*27, B*55	C*03, C*06
NRC2	A*23, A*24	B*40, B*57	NDa
NRC1	A2, A30	B14, B44	ND

^aND, not determined.

Fig. 2.

Annotated sequence of the amino-terminal domain of HIV-1 capsid. Shown is the amino acid sequence of the NL4-3 CA (amino acids 1 to 151). The seven helical regions, as described in reference 22, are boxed and numbered (numbers in circles). The CTL epitopes mentioned in the text are indicated, with the exception of QW9, which is located in the C-terminal CA domain. Amino acids that are commonly mutated to induce resistance to CTLs recognizing these epitopes (red) or to compensate for impaired replication resulting from CTL resistance mutations (green) and which were evaluated in this study are indicated. The amino acids are numbered both according to the amino acids in CA (numbers in parentheses) and Gag.

Table 2.

Mutations in capsid common to all viral clones from patient NRC10^a

Mutation	Epitope	HLA type(s)	Effect	Reference(s), source
V159I	SV9	B*0702, B*8101	CTL resistance	23, 73
S165N	KF11	B*57/5801, B63	CTL resistance	17, 21, 74
S173T	KK10	B*27	Compensatory	60, this article
	KF11	B*57	CTL resistance	48, 53
E203D	KA9	B*4002	NDb
V218A	TW10	B*57/5801, B63	Compensatory	11, 14, 29, 36
M228I	TW10	B*57/5801, B63	Compensatory	4, 8, 11, 14, 29, 36, 51
G248A	TW10	B*57/5801, B63	CTL resistance	8, 20, 21, 39, 49, 63
R264K	KK10	B*27	CTL resistance	25, 33, 60, 61
I267V	KK10	B*27	ND
L268M	KK10	B*27	CTL resistance, compensatory	25, 33, 60, 61
R286K	ND
S310T	QW9	B*5301, B*57/B*5801	CTL resistance	2, 48, 51
G340A	ND

^aFor mutations implicated in resistance to CTLs, the CTL epitope involved, its major histocompatibility complex (MHC) restriction, and the effect of the mutation are also shown.

^bND, not determined.

Fig. 3.

Mutations associated with escape from CTL increase sensitivity of viruses from patient NRC10 to hTRIM5α. Recombinant viruses carrying the Gag-PR sequence from the clinical isolate (NRC10) and variants in which one or more mutations in the gag coding sequence were introduced by mutagenesis were prepared. The mutations resulted in the indicated amino acid substitutions in the KK10 epitope and/or Gag amino acid 173 (a) or in the indicated substitutions in the TW10 epitope and/or in or near the CypA binding loop, accompanied or not by reversion of three mutations (S173T, R264K, and L268M) associated with resistance to CTL targeting the KK10 epitope (b). hTRIM5α sensitivity was determined as shown in Fig. 1. All experiments were performed 4 times using in parallel two independently derived pairs of transduced cell lines (n = 8); two different viral stocks were used in the course of the experiments for each series. The infectivity of NL4-3 was 1.8 ± 0.2 fold (a) and 2.3 ± 0.4 fold (b) greater in U373-X4-TRIM5γ cells than in U373-X4-LacZ cells. *, P < 0.05; **, P < 0.01 (compared to NRC10). ††, P < 0.01 compared to NL4-3. ‡, P < 0.05; ‡‡, P < 0.01; ‡‡‡, P < 0.001 (compared to NRC10-T173S-K264R-M268L).

In addition to the four mutations studied above, 9 other amino acids in CA were common to all viral clones obtained from patient NRC10 but differed from those present in NL4-3 (Table 2). Interestingly, 6 of these mutations have been implicated in resistance to CTLs targeting epitopes presented by HLA-B57. Because patient NRC10 is HLA-B57 negative, these mutations are likely to represent “footprints” selected during prior passage of the virus in an HLA-B57-positive individual (46, 47). Consistent with this idea, the virus from NRC10 did not contain either A163G or T242N, mutations which frequently occur in the KF11 and TW10 epitopes targeted by CTL in HLA-B57-positive individuals but which impair viral replicative capacity and often revert when virus is transmitted to HLA-B57-negative individuals (16, 17, 69).

Two of the HLA-B57 footprint mutations found in NRC10 (S165N and S310T in the KF11 and QW9 epitopes) had no effect on hTRIM5α susceptibility (data not shown). In contrast, reversion of the three mutations associated with escape from CTLs targeting the HLA-B57 restricted TW10 epitope (G248A in the TW10 epitope and V218A plus M228I in the cyclophilin A [CypA]-binding loop) led to a substantial decrease in hTRIM5α sensitivity (Fig. 3b, NRC10-A218V-I228M-A248G). This reduction in hTRIM5α sensitivity was significantly greater than that observed following reversion of the mutations associated with resistance to the HLA-B27 restricted epitope KK10 (NRC10-T173S-K264R-M268L). Furthermore, reverting the mutations associated with escape from CTLs targeting both the TW10 epitope and the KK10 epitope did not reduce hTRIM5α sensitivity to a greater extent than reverting only the mutations associated with escape from CTLs targeting the TW10 epitope (Fig. 3b; compare NRC10-A218V-I228M-A248G and NRC10-T173S-K264R-M268L-A218V-I228M-A248G).

Among the three mutations associated with escape from CTLs targeting the TW10 epitope, the V218A compensatory mutation had the greatest impact on hTRIM5α sensitivity. Using the variant in which mutations associated with escape from CTLs targeting the KK10 epitope had been reverted as a starting point (NRC10-T173S-K264R-M268L), reverting V218A alone (NRC10-T173S-K264R-M268L-A218V), reverting both V218A plus M228I (NRC10-T173S-K264R-M268L-A218V-I228M), or reverting all three mutations (NRC10-T173S-K264R-M268L-A218V-I228M-A248G) led to similar reductions in sensitivity to hTRIM5α, whereas reverting only the G248A resistance mutation within the TW10 epitope (NRC10-T173S-K264R-M268L-A248G) did not significantly change hTRIM5α sensitivity.

Taken together, these findings indicate that the relatively high sensitivity of viruses from patient NRC10 to hTRIM5α required the coexistence of both the V218A compensatory mutation associated with escape from the HLA-B57 restricted epitope TW10 and the R264K resistance mutation associated with escape from the HLA-B27 restricted epitope KK10. In the absence of the V218A mutation, the sensitivity to hTRIM5α of NRC10 was not influenced by the presence of R264K. Conversely, in the absence of R264K, the impact of the V218A mutation on hTRIM5α sensitivity was relatively small. When both mutations were present, however, sensitivity to hTRIM5α was substantially higher and could be further increased by removing the S173T compensatory mutation. Consistent with this interpretation, we found that adding the R264K mutation to NL4-3, which does not carry the V218A mutation, had no significant effect on hTRIM5α sensitivity (data not shown).

CTL resistance mutations induce susceptibility to hTRIM5α in viruses from patient NRC2.

Recombinant viruses carrying Gag-PR sequences of viruses from patient NRC2 are also more sensitive than NL4-3 to hTRIM5α (5). Patient NRC2 is HLA-B57 positive (Table 1), and viruses from this patient contained the classical T242N and G248A escape mutations in the HLA-B57 restricted epitope TW10 (Table 3). Viruses expressing TW10 resistance mutations often express one or more compensatory mutations in or near the CypA-binding loop, and viruses from patient NRC2 expressed the I223V mutation. To evaluate the impact of these mutations on sensitivity to hTRIM5α, one or more of these mutations were removed from the NRC2 sequence and the sensitivities of viruses carrying these sequences were evaluated (Fig. 4A). Removing the I223V compensatory mutation (NRC2-V223I) did not influence sensitivity to hTRIM5α. Removing all three mutations associated with resistance to CTLs targeting the HLA-B57 restricted epitope TW10 (NRC2-V223I-N242T-A248G) or reversion of I223V plus T242N (NRC2-V223I-N242T) significantly reduced the hTRIM5α sensitivities of these viruses compared to that of NRC2, whereas reversion of I223V plus G248A (NRC2-V223I-A248G) did not significantly reduce hTRIM5α sensitivity. Adding two additional TW10-associated compensatory mutations to NRC2 (NRC2 plus H219Q and M228L) had no significant effect on hTRIM5α sensitivity relative to that of NRC2. These findings indicate that the resistance mutations in the HLA-B57 restricted epitope TW10, especially T242N, explain most of the increased sensitivity of NRC2 to hTRIM5α.

Table 3.

Mutations in capsid common to all viral clones from patient NRC2^a

Mutation	Epitope(s)	HLA type(s)	Effect	Reference(s)
A146P	IW9, VL10, VV9	B*57/5801, B*15, B*13	CTL resistance	2, 18, 28, 39, 51
A147L	IW9, VV9	B*57/5801, B*13	CTL resistance	2, 28, 39, 51
A163G	KF11	B*57/5801, B63		15, 17, 21
S165N	KF11	B*57/5801, B63	CTL resistance	17, 21, 74
I223V	TW10	B*57/5801, B63	compensatory	4, 8, 11, 14, 29, 36, 51
T242N	TW10	B*57/5801, B63	CTL resistance	2, 8, 21, 39, 49, 63
G248A	TW10	B*57/5801, B63	CTL resistance	2, 8, 20, 21, 39, 49, 63
E260D	KK10	B*27	NDb	33, 60
T280V	RL9	A2	ND	37
E312D	QW9	B*57	CTL resistance?	48, 51, 53
T342S	ND	B*57, C*0804		41, 56

^aFor mutations implicated in resistance to CTLs, the CTL epitope involved, its MHC restriction, and the effect of the mutation are also shown.

^bND, not determined.

Fig. 4.

Effect of mutations associated with escape from CTL targeting the HLA-B57 associated epitope TW10 on hTRIM5α sensitivity of viruses from patients NRC2 and NRC1. Recombinant viruses carrying the Gag-PR sequence from the clinical isolate NRC2 (a) or NRC1 (b) and variants in which one or more mutations in the gag coding sequence were introduced by mutagenesis, resulting in the indicated amino acid substitutions in the TW10 epitope and/or substitutions in or near the CypA binding loop were prepared. Sensitivity to hTRIM5α was evaluated as described for Fig. 1. The experiments were performed 4 times (a) or 3 times (b) using in parallel two independently derived pairs of transduced cell lines (n = 8 and n = 6, respectively); two different viral stocks were used in the course of the experiments. The infectivity of NL4-3 was 1.6 ± 0.1 fold (a) and 1.8 ± 0.1 fold (b) greater in U373-X4-TRIM5γ cells than in U373-X4-LacZ cells. ††, P < 0.01; †††, P < 0.001 (compared to NRC2 [a] or NRC1[b]).

In addition to the three mutations studied above, 6 of the 8 other amino acids in the CA from NRC2 that differed from those present in NL4-3 are associated with resistance to CTLs targeting epitopes presented by HLA-B57 (Table 3). Removing the A146P and A147L mutations, involved in resistance to CTLs targeting the IW9 epitope, did not influence the sensitivity of NRC2 to hTRIM5α (data not shown). The impact of mutations involved in resistance to CTLs targeting the KF11 and QW9 epitopes was not studied in this isolate; the S165N mutation in the KF11 epitope, however, did not influence the sensitivity of NRC10 to hTRIM5α (data not shown).

The effect of resistance mutations in the TW10 epitope on hTRIM5α sensitivity is context dependent.

Although the T242N and G248A mutations in the TW10 epitope caused a 3-fold increase in the sensitivity of NRC2 viruses to hTRIM5α, the viruses from a third patient, patient NRC1, which express these mutations along with 3 compensatory mutations in the CypA-binding loop, showed little sensitivity to hTRIM5α (Fig. 4B). As was observed for NRC2, progressive removal from the NRC1 virus of the compensatory mutations in or near the CypA-binding loop did not influence hTRIM5α sensitivity (NRC1-A218V, NRC1-A218V-Q219H, NRC1-A218V-Q219H-V223I, and NRC1-A218V-Q219H-V223I-L228M). Removing, in addition, the T242N resistance mutation (NRC1-A218V-Q219H-V223I-L228M-N242T) or both T242N and G248A mutations (NRC1-A218V-Q219H-V223I-L228M-N242T-A248G) reduced hTRIM5α sensitivity to levels that were not significantly different from that for NL4-3. Nevertheless, the presence of the T242N and G248A mutations in NRC1 had substantially less impact on sensitivity to hTRIM5α sensitivity than was observed for NRC2. Thus, as described earlier for the resistance mutations in the HLA-B27 restricted epitope KK10, the impact of resistance mutations in the HLA-B57 restricted epitope TW10 on hTRIM5α sensitivity is dependent on variation in the CA sequence outside this epitope.

hTRIM5α sensitivity and viral infectivity.

For the mutations studied above, we also examined the relationship between their effects on sensitivity to hTRIM5α and viral infectivity, as measured in the target cells in which hTRIM5α activity had been inhibited by overexpression of TRIM5γ (Fig. 1). No significant correlations between hTRIM5α sensitivity and infectivity were observed, either when variants of a given CA sequence were analyzed separately (Fig. 5) (Pearson r: −0.10 to +0.52; P > 0.29 for all comparisons) or when all viral variants were analyzed together (Pearson r = −0.35; P > 0.09).

Fig. 5.

Relationship between sensitivity to hTRIM5α and viral infectivity. Sensitivity to hTRIM5α and viral infectivity, as measured in the target cells in which hTRIM5α activity had been inhibited by overexpression of TRIM5γ, were determined as described for Fig. 1 for recombinant viruses carrying the Gag-PR sequence from the clinical isolates (solid symbols) or viral variants produced by mutagenesis (open symbols) whose sensitivity to hTRIM5α is shown in Fig. 3 and 4.

In some cases, changes in hTRIM5α sensitivity and infectivity were inversely related. For viruses from patient NRC10, in the absence of compensatory mutations at amino acid 173, the R264K and L268M mutations in the HLA-B27 restricted epitope KK10 increased sensitivity to hTRIM5α (Fig. 3A) and significantly impaired infectivity (Table 4); the presence of either the S173T or S173A compensatory mutation decreased sensitivity to hTRIM5α and increased infectivity (Table 4) (P < 0.02 for both comparisons). In other cases, however, this inverse relationship was not observed. Adding the R264K and L268M mutations to NL4-3 had no significant effect on sensitivity to hTRIM5α but, as previously described (61), resulted in a 4-fold reduction in infectivity (data not shown). Removing the T242N and G248A resistance mutations in the TW10 epitope along with compensatory mutations in the CypA-binding loop reduced both hTRIM5α sensitivity (Fig. 4) and infectivity of NRC2 and NRC1 (Table 4). Thus, no fixed pattern relating changes in hTRIM5α sensitivity and infectivity was discernible in our studies.

Table 4.

Viral infectivity, as measured in the target cells in which hTRIM5α activity had been inhibited by overexpression of TRIM5γ

Variant	Infectivity (% NL4-3)		n	Pa
	Mean	SD
NRC10	21.4	10.8	16	††
NRC10-T173S	5.5	2.3	8	*
NRC10-T173S-K264R	26.4	17.5	8
NRC10-T173S-K264R-M268L	46.8	28.9	16	††
NRC10-T173A	36.1	13.1	8	†
NRC10-T173S-K264R-V267I-M268L	42.9	17.1	10	††
NRC10-A218V-I228M-A248G	13.6	4.4	8
NRC10-T173S-K264R-M268L-A218V	20.0	12.2	8
NRC10-T173S-K264R-M268L-A218V-I228M	36.7	16.8	8
NRC10-T173S-K264R-M268L-A218V-I228M-A248G	12.5	6.4	8
NRC10-T173S-K264R-M268L-A248V	16.9	6.8	8
NRC2	36.2	2.6	4
NRC2-V223I	27.7	10.2	8
NRC2-V223I-A248G	11.6	3.2	8	‡
NRC2-V223I-N242T	33.7	7.9	8
NRC2-V223I-N242T-A248G	18.0	2.5	8	‡
NRC2-H219Q-M228L	40.9	16.0	8
NRC1	89.3	4.3	6
NRC1-A218V	101.8	10.7	6
NRC1-A218V-Q219H	66.2	13.5	6
NRC1-A218V-Q219H-V223I	50.2	9.2	6	§
NRC1-A218V-Q219H-V223I-L228M	78.3	37.2	6
NRC1-A218V-Q219H-V223I-L228M-N242T	62.9	15.1	6
NRC1-A218V-Q219H-V223I-L228M-N242T-A248G	37.5	14.2	6	§

^a*, P < 0.01 compared to NRC10. †, P < 0.02; ††, P < 0.01 (compared to NRC10-T173S). ‡, P < 0.01 compared to NRC2; §, P < 0.01 compared to NRC1.

DISCUSSION

We recently found that HIV-1 expressing CA sequences from some clinical isolates can be more sensitive to hTRIM5α than are laboratory-adapted strains (5). This study extends these findings by demonstrating that (i) the increased hTRIM5α sensitivities of these viruses can be attributed to mutations associated with escape from CTL responses against epitopes in CA, (ii) both resistance mutations within CTL epitopes and associated compensatory mutations can contribute to this process, (iii) the impact of these mutations on hTRIM5α sensitivity is strongly context dependent, in that the effects of a given mutation can be strikingly different depending on the CA sequence in which it occurs, and (iv) CTL resistance mutations and/or associated compensatory mutations previously selected by prior hosts expressing distinct HLA haplotypes appear to influence the impact on hTRIM5α sensitivity of CTL resistance mutations arising in the recipient. Taken together, these findings suggest that modifications in hTRIM5α sensitivity resulting from CTL mutations may influence HIV-1 replication in vivo, but, as discussed below, numerous factors are likely to modulate the magnitude of this effect.

Several of the resistance mutations studied here are thought to impair viral replication in vivo because the mutations revert when viruses carrying these mutations are transmitted to individuals not expressing HLA molecules capable of presenting the involved epitope. When evaluated in single-cycle infectivity assays, effects of these mutations on infectivity have ranged from a <2-fold reduction (T242N mutation in the B57 restricted TW10 epitope) to a 20-fold reduction (R264K mutation in the B27 restricted KK10 epitope) (8, 44, 61). These effects can be different in different target cells (8, 44), and the possible contribution of hTRIM5α expression by the target cells was not evaluated. Because we evaluated the effect of hTRIM5α sensitivity using different target cells and treated these target cells with IFN-α to increase hTRIM5α expression, it is difficult to compare these prior findings with our results. In some cases studied by us, resistance mutations were found to both impair infectivity and increase sensitivity to hTRIM5α, and the addition of compensatory mutations corrected, at least partially, both of these defects. In these cases, the selective pressures to improve infectivity and reduce hTRIM5α sensitivity would presumably act in consort to favor selection of the compensatory mutations. Further studies are required to evaluate the extent that hTRIM5α sensitivity influences viral replication in vivo and to assess the relative importance of the effect of CTL mutations on hTRIM5α sensitivity and viral infectivity.

It is likely, however, that the pressure exerted by hTRIM5α on viral replication by a given host may vary for several reasons. First, we found that the extent that a given CTL resistance mutation influences hTRIM5α sensitivity is dependent on the CA sequences of viruses infecting that individual. In addition, host TRIM5α activity can be variable depending on the level of expression of hTRIM5α (64), possibly reflecting polymorphisms in transcription factor-binding sites (10), on the extent of exposure of cells to IFN-α (5, 13, 58), and/or on the presence of hTRIM5α polymorphisms that can modify its ability to inhibit replication of a given viral isolate (5, 24, 30, 59). Furthermore, several of the mutations studied by us, including both the CTL resistance mutations and mutations in the CypA-binding loop, have been previously shown to modulate the effects of inhibiting CA-CypA interactions on viral infectivity (8). The inhibition of CypA-CA interactions often directly reduces the stability of HIV-1 capsid and impairs viral infectivity, although mutations in or near the CypA binding loop can render viruses “CypA independent” (8, 14, 29, 70). The inhibition of CypA-CA interactions has been reported to have no effect on the sensitivity of HIV-1 to hTRIM5α (32, 65, 68), but we have recently shown that inhibiting CypA-CA interactions can also increase or decrease hTRIM5α sensitivity in an isolate-specific fashion (5). These findings suggest that differences in cellular CypA levels may also influence the effect of CTL mutations on both viral replicative capacity and hTRIM5α sensitivity. Finally, it is conceivable that, for a given virus, escape mutations for certain epitopes targeted by CTL may increase hTRIM5α sensitivity to such an extent that their emergence is precluded, leaving the virus more sensitive to CTL pressure.