The Same Major Histocompatibility Complex Polymorphism Involved in Control of HIV Influences Peptide Binding in the Mouse H-2Ld System*

Background: Major histocompatibility complex (MHC) class I polymorphisms at residue 97 have been associated with HIV control and progression. Results: Mouse MHC class I variant LdR97 bound a broader repertoire of peptides than LdW97. Conclusion: Aromatic-stacking interactions involving W97 may restrict peptide-binding plasticity. Significance: The homologous HLA-B allele (Arg97) likely binds a larger self-peptide repertoire than Trp97, resulting in reduced T cell diversity due to tolerance. Single-site polymorphisms in human class I major histocompatibility complex (MHC) products (HLA-B) have recently been shown to correlate with HIV disease progression or control. An identical single-site polymorphism (at residue 97) in the mouse class I product H-2Ld influences stability of the complex. To gain insight into the human polymorphisms, here we examined peptide binding, stability, and structures of the corresponding Ld polymorphisms, Trp97 and Arg97. Expression of LdW97 and LdR97 genes in a cell line that is antigen-processing competent showed that LdR97 was expressed at higher levels than LdW97, consistent with enhanced stability of self-peptide·LdR97 complexes. To further examine peptide-binding capacities of these two allelic variants, we used a high affinity pep-Ld specific probe to quantitatively examine a collection of self- and foreign peptides that bind to Ld. LdR97 bound more effectively than LdW97 to most peptides, although LdW97 bound more effectively to two peptides. The results support the view that many self-peptides in the Ld system (or the HLA-B system) would exhibit enhanced binding to Arg97 alleles compared with Trp97 alleles. Accordingly, the self-peptide·MHC-Arg97 complexes would influence T-cell selection behavior, impacting the T-cell repertoire of these individuals, and could also impact peripheral T cell activity through effects of self-peptide·Ld interacting with TCR and/or CD8. The structures of several peptide·LdR97 and peptide·LdW97 complexes provided a framework of how this single polymorphism could impact peptide binding.

Human leukocyte antigens (HLA) 2 encoded by the major histocompatibility complex (MHC) are among the most polymorphic genes in the population (1). Class I MHC gene products display short peptides of 8 -10 amino acids on the surface of all nucleated cells (2), and these complexes are recognized by the ␣␤ T cell receptors (TCR) on the surface of T cells. TCR binding of pep-MHC molecules above a minimal threshold results in T cell activation and effector function such as target cell lysis (3)(4)(5).
The MHC locus has been shown to be associated with protection from or susceptibility to various diseases, including infectious diseases of viral origin such as HIV, as well as autoimmune conditions (6 -12). Ultimately the MHC-associated control or progression of diseases is related to the presence of functional T cells that express TCRs that bind to antigenic (or self) peptide⅐MHC complexes. The presence of functional T cells, and the breadth of the TCR repertoire, is controlled both by thymic selection processes that occur during development of a T cell, and peripheral processes that regulate mature T cell activity (13). In the thymus, positive and negative selection operates on T cells through the ␣␤ TCR and the CD4 and CD8 co-receptors. T cells that have TCRs with a low affinity for self-peptides presented on self-MHC molecules are positively selected, yet T cells that express TCRs with higher affinity for self-peptide⅐MHC complexes are clonally deleted (14 -16).
It is widely hypothesized that the diversity of peripheral T cells of an individual correlates with the effectiveness of the adaptive immune system in eliminating viral infections, and that the breadth of the TCR repertoire correlates in turn with the diversity of self-peptides presented by MHC alleles of the individual during development (17). Thus, self-protein antigen processing pathways are critically linked to immune potential (16, 18 -20). Importantly, the peripheral regulatory processes are also influenced by interactions of T cells with self-pep⅐MHC complexes, which can control functional T cell activity by promoting the deletion or anergy of particular T cells (as reviewed in Refs. [21][22][23]. Recent studies have implicated several positions of the human MHC locus HLA-B in the control and progression of HIV infection. Positions 67, 70, and 97 of HLA-B located in the peptide-binding groove, showed the strongest associations with differences in the frequency of HIV controllers and progressors (9). It has been hypothesized that these amino acid positions in HLA-B could contribute to disease control or progression by affecting the ability of either foreign peptides or self-peptides to bind the different alleles; the former could operate by inducing favorable immunity, and the latter could operate by influencing thymic selection (as reviewed in Ref. 24). However, biochemical and structural explanations for these effects remain unknown, and thus a mechanistic explanation of the polymorphisms is yet to be developed.
The mouse MHC (known as H-2) represented the first system to reveal the properties of polymorphism and the process of T cell recognition that required MHC "restriction" of the antigen (25). The class I H-2 molecule L d has been well studied and is known to exhibit lower stability than many other class I products such as K b (26 -30). In addition, it appears that peptides that bind to L d may be influenced by more of their individual residues than peptides that bind to K b , which appear to have dominant anchor residues (31). Relevant to recent information about HIV and HLA-B polymorphisms are previous studies showing that two alleles called L q and L d contained the same polymorphism at residue 97 as in the HLA-B studies, and these alleles exhibited differences in stability between the ternary complexes (peptide, H-2L heavy chain, and ␤ 2 -microglobulin) (32). In an independent approach, the W97R substitution was identified in a "needle in a haystack" experiment in which more stable L d mutants were identified because they conferred higher surface levels of the protein in a yeast display system (33). Thus, completely different approaches have revealed the importance of position 97 in L d stability. The crystal structures of L d complexes with Trp (34,35) and Arg (36 -38) have also been solved, providing additional mechanistic insight.
To examine more fully the hypotheses regarding the impact of the position 97 polymorphisms on peptide binding, here we introduced two variants, L d W97 and L d R97, into a mouse cell line and examined various aspects of peptide binding and stability of the complexes. The binding analysis was facilitated by the use of a high affinity, soluble TCR that was available against the specific peptide complex called QL9-L d , allowing a quantitative assay of binding by a panel of peptides. The results showed that L d R97, compared with L d W97, conferred a greater average stability to peptide⅐L d complexes and resulted in a broader repertoire of peptide binding from the panel of peptides. Each peptide tested showed preferential binding to one of the L d variants, but most of them bound better to L d R97.
The structures of several different peptides bound to one or the other L d variants suggested that the Arg 97 residue acts not by having increased flexibility itself, but by tolerating more side chain flexibility in the vicinity of the peptide and adjacent MHC residues, compared with Trp 97 . Overall, these findings are consistent with the hypothesis that many self-peptides will be presented at physiologically higher levels by an allele with Arg 97 , compared with an allele with Trp 97 . Thus, our results support the notion that Arg 97 alleles exhibit a preference for a progressor phenotype in HIV patients because potentially effective T cells in these individuals have been deleted by negative selection in the thymus. It is also possible that these self-peptide⅐Arg 97 allele complexes may be capable of driving enhanced peripheral T cell anergy (as reviewed in Ref. 21).

EXPERIMENTAL PROCEDURES
Cell Lines and L d Transfectants-A derivative of BW5147, a mouse T cell thymoma (AKR mouse, H-2 k ), was used for the L d studies because it is readily transduced by the pMP71 retroviral vector, and it is capable of normal antigen processing of self and foreign peptides (e.g. it is TAP sufficient). The BW5147 subline called 58 Ϫ/Ϫ was used to generate L d -transfectants, as described below. BW5147 and the hybridoma 30-5-7 that secretes an anti-L d monoclonal antibody (against the ␣2 domain) were cultured in RPMI 1640 media containing 10% fetal calf serum at 37°C with 5% CO 2 .
The gene encoding the full-length murine MHC allele heavy chain L d W97, residues 1-338 with a 24-amino acid leader sequence, was cloned into retroviral vector pmp71 as a NotI-EcoRI fragment. The L d W97-pMP71 ligation product was obtained and a single-site mutation was made using the Stratagene QuikChange Lightning kit to generate a pMP71 construct containing the gene for L d R97. BW5147-L d W97 and BW5147-L d R97 cell lines were generated by retroviral transduction (38). Briefly, 20 g of L d -pMP71 DNA was incubated with 60 g of Lipofectamine 2000 (Invitrogen) in 3 ml of Opti-MEM media (Invitrogen) for 20 min at room temperature. DMEM culture supernatant was removed from 3 ϫ 10 6 adherent Plat-E retroviral packaging cells and 1.5 ml of the DNA/ Lipofectamine/Opti-MEM mixture was added. Plat-E cells were incubated with the DNA mixture for 3 h at 37°C, then the DNA mixture was removed. Plat-E cells were washed with 7 ml of RPMI 1640 media, then incubated with 6.5 ml of RPMI 1640 media for 48 h at 37°C. BW5147 cells (2 ϫ 10 6 ) were transduced by centrifuging the cells in a 24-well plate (Corning costar) at 652 ϫ g for 45 min with 3 ml of supernatant from transfected Plat-E cells. BW5147 cells were cultured overnight at 37°C before being transferred to a T25 flask with 5 ml of RPMI 1640 media and cultured at 37°C with 5% CO 2 . L d -positive populations of BW5147 cells, expressing L d W97 or L d R97, were enriched by staining with a saturating amount of 30-5-7 anti-L d monoclonal antibody (20 g/ml; purified from ascites fluid) followed by PE-conjugated goat F(abЈ) 2 anti-mouse IgG secondary antibody (1.25 g/ml; Southern Biotech) (39). After fluorescence activated cell sorting (FACS), isolated cells were cultured in RPMI 1640 at 37°C with 5% CO 2 .
Soluble High-affinity T Cell Receptor-The soluble TCR called 2C-m6 binds to the QL9⅐L d complex with high affinity (K D value 5 nM) (40). A single chain form of the 2C-m6 TCR (scTCR) was expressed in Escherichia coli, refolded from inclusion bodies, and purified as described previously (33). Purified 2C-m6 scTCR was biotinylated using the EZ-Link Micro Sulfo-NHS-Biotinylation Kit (Thermo Scientific) and stored at 4°C (41).
Measurement of Cell Surface L d Levels-Cell surface levels of L d variants were evaluated using transduced BW5147 cells and anti-L d antibody. BW5147 cells (2 ϫ 10 5 ) expressing L d W97 or L d R97 were incubated with RPMI alone or with peptides at various concentrations in RPMI for 3 h, then washed and incubated with 20 g/ml of 30-5-7 anti-L d mAb followed by PElabeled goat F(abЈ) 2 anti-mouse IgG secondary antibody. Samples were washed, fixed with 1% paraformaldehyde in PBS, and analyzed for PE fluorescence using an Accuri C6 flow cytometer. Data were analyzed using FCS Express.
Synthetic peptides were assayed to assess binding of soluble single chain 2C-m6 to each peptide⅐L d complex. BW5147-L d W97 or BW5147-L d R97 cells (2 ϫ 10 5 ) were incubated with 100 M peptide diluted in RPMI for 3 h at 37°C. Cells were washed and incubated with 10 g/ml of soluble biotinylated 2C-m6 scTCR for 40 min on ice. Cells were washed and incubated with 2.5 g/ml of phycoerythrin-conjugated streptavidin (Streptavidin-PE, BD Pharmingen) before fixing with 1% paraformaldehyde. Samples were analyzed by flow cytometry on an Accuri C6 cytometer, and data were analyzed using FCS Express.
L d -binding Assay-Because initial studies with a reference biotinylated peptide yielded very low signal to background, we developed a novel competition assay to evaluate the relative strength of each peptide in binding to L d W97 and L d R97. BW5147 cells (2 ϫ 10 5 ) expressing L d W97 or L d R97 were incubated with a mixture containing 50 nM QL9 and various concentrations of competitor peptides in RPMI (note that these competitor peptides were not detectably bound by the soluble TCR). After 2.5 h at 37°C, cells were washed and incubated with 10 g/ml of biotinylated 2C-m6 scTCR. After washing, cells were incubated with 2.5 g/ml of steptavidin-PE (BD Pharmingen). Samples were fixed with 1% paraformaldehyde in phosphate-buffered saline and analyzed for PE fluorescence using an Accuri C6 flow cytometer. Data were analyzed using FCS Express. Mean fluorescence units (MFU) of samples incubated with competitor peptide and QL9 (MFU 1 ) was divided by mean fluorescence units of samples incubated with QL9 alone (MFU 0 ). BD 50 values were obtained using non-linear regression analysis (GraphPad Prism).
Soluble L d R97 Stability Assay-The stabilized L d m31 form (21 kDa) of MHC L d was expressed in E. coli as inclusion bodies and refolded as described previously (33,37). Briefly, 3 M L d m31 inclusion bodies was refolded with 30, 6, or 1.2 M dimethyl sulfoxide-dissolved peptide (MCMV, FLS, or tumϪ) in 25 ml of buffer containing 100 mM Tris, pH 8.0, 400 mM L-arginine, 2 mM EDTA, 0.2 mM phenylmethylsulfonyl fluoride (PMSF), 5 mM reduced glutathione, and 0.5 mM oxidized gluta-thione for 48 h at 4°C, without stirring. Refolded mixtures were filtered through 0.22-m vacuum filter (Millipore) and concentrated to 50 l using a 10-kDa cut off filter (Amicon). Samples for SDS-PAGE were prepared using 8.3 l of each concentrated protein refold, and 4 -20% gradient polyacrylamide gels (Bio-Rad) were run for 45 min at 120 V, and stained with Coomassie Blue.
Scanned gel images were analyzed using the ImageJ public domain Java image processing program, first by converting the scanned image into 16-bit grayscale, and then by obtaining the area of each band and the average intensity of the band. The average intensity of each band was multiplied by the area of the band, to generate overall intensity. These values were considered proportionate to the yield of each peptide⅐L d complex. The overall intensity value of the band corresponding to L d m31 refolded in the absence of peptide (no peptide) was set to 1, and all protein yield values relative to this were calculated. Peptide

RESULTS
Complexes of Self-peptide⅐L d R97 Are Expressed at Higher Levels on Cells Than Complexes of Self-peptide⅐L d W97-A previous study showed that the two H-2 alleles L d and L q differed at six positions, including position 97 (Trp in L d and Arg in L q ) (32,42), the same single site polymorphism shown to be involved in HLA-B control of HIV (9). Experiments performed with single site mutants of L d (W97R) and L q (R97W) on the surface of DAP-3 cells using antibodies against assembled and empty L d or L q molecules suggested that Arg 97 may contribute to greater cell surface stability, stronger binding to endogenous peptides, and higher affinity for ␤ 2 -microglobulin compared with Trp 97 (32).
To determine whether this observation extended to a different antigen-processing competent cell line, we transduced the BW5147 (H-2 k ) cell line with wild-type L d (L d W97) and the single site mutant L d R97, using these genes inserted into the pMP71 retrovirus vector (43). To verify expression, cells were stained with the anti-L d mAb 30-5-7 and analyzed by flow cytometry (Fig. 1A). Surface levels of L d R97, as judged by the MFU with anti-L d staining were consistently twice as high as surface levels of L d W97 (Fig. 1A). As it is well known that the surface levels of MHC are controlled in part by the stability of the peptide-class I heavy chain⅐␤2m complex, this finding suggests that either there are many more endogenous self-peptides that can associate well with L d R97, or there is a smaller subset of abundant self-peptides that bind better to L d R97, or both.
In the L d system, "up-regulation" assays dependent on the ability of an exogenous synthetic peptide to stabilize the pep-L d ⅐␤2m complex and thus increase its total surface levels have often been performed to determine L d -binding abilities (44). To examine if L d -binding peptides could further increase the levels of L d W97 and L d R97 on BW5147, four peptides were examined in this assay. Cells were incubated with various concentrations of peptides QL9 (QLSPFPFDL), MCMV (YPHFMPTNL), FLS (FLSPFWFDI), and tumϪ (TQNHRALDL), and the surface levels of L d W97 and L d R97 were detected with the 30-5-7 anti-L d monoclonal antibody. The level of up-regulation in response to incubation with each peptide concentration was determined relative to L d surface levels in the absence of peptide, and the concentration of peptide resulting in 50% maximal up-regulation (SD 50 ) was determined (Fig. 1B). Peptide MCMV induced up-regulation of L d W97, with an SD 50 value of ϳ0.12 M. Peptide FLS induced modest up-regulation at only the highest concentration, 100 M, whereas neither QL9 nor tumϪ exhibited any detectable up-regulation. In contrast, peptides QL9, MCMV, and FLS induced significant up-regulation of L d R97 on the BW5147 cell surface, with SD 50 values in the order of MCMV Ͻ QL9 Ͻ FLS (Fig. 1B). Thus, the addition of exogenous peptides enabled the cell surface levels of L d R97 to be increased more effectively than the levels of L d W97.
The inability of the L d (especially L d W97) cell surface levels to be increased in the presence of some peptides is likely to be related to the expression of TAP and hence self-peptide complexes. We reasoned that perhaps L d levels could be enhanced at lower temperatures or in the presence of ␤2-microglobulin (␤2m). Previous studies using the RMAS cell line (45,46) showed that lower temperatures (26°C) allowed enhanced accumulation of empty MHC molecules at the cell surface, whereas addition of ␤2m during peptide loading also increased MHC surface levels on RMAS cells (47). However, neither incubation of the BW5147 cells at a lower temperature (26°C) nor inclusion of ␤2m during peptide loading increased surface levels of L d W97 or L d R97 (data not shown).
L d R97 Binds Some, but Not All, Peptides Better Than L d W97-As indicated, the higher levels of self-peptide L d R97 complexes here and in a previous study (32) could be due to a few abundant self-peptides that bind to L d R97 better, or to a large fraction of self-peptides that bind better. Because the L d up-regulation assay provides only a semi-quantitative assessment of L d -binding ability for some peptides, we developed a competition assay that took advantage of a unique soluble TCR reagent, called 2C-m6, which detects complexes of QL9⅐L d W97 and QL9⅐L d R97. Based on peptide titrations ( Fig. 2A), we estimated that the QL9 peptide binds ϳ2-3-fold better to L d R97 than to L d W97 (BD 50 values of 85 nM on L d W97 and 17 nM on L d R97).
To examine additional L d -binding peptides, we used a panel of peptides that had originally been isolated by one of several methods: previously characterized foreign peptides (tumϪ and MCMV), peptides derived from a yeast-display screen involving a single chain form of L d R97 (38), or peptides eluted from L d as part of an anti-L d affinity purification from BALB/c (L d W97 ϩ ) tissue. Thus, these peptides had been isolated based in part on their ability to associate with L d . These peptides, when bound to L d W97 or L d R97, are not detected by the soluble 2C-m6 TCR, although weak binding was detected with peptide FLSPFWFDI at very high concentrations. This allowed the competitive assay to be used to determine quantitatively the binding affinity of the collection of peptides for L d W97 and L d R97.
In the competition assay format, cells bearing either L d W97 or L d R97 were incubated with synthetic peptides that were titrated in the presence of 50 nM QL9, followed by staining with the 2C-m6 TCR. Decreased fluorescent signal indicated that the competitor peptide effectively competed for binding to the  (30-5-7) were evaluated by flow cytometry. Percent up-regulation was determined for each peptide and the concentration corresponding to 50% maximal up-regulation (SD 50 ) was determined. C, BW5147 cells expressing L d W97 and L d R97 were loaded with a mixture of 50 nM QL9 peptide and various concentrations of peptides MCMV, FLS, and tumϪ, and QL9-L d complexes were detected with soluble high affinity 2C-m6 scTCR. The concentration of each peptide that corresponded to 50% maximal competition (BD 50 ) with QL9 was determined. The data in C are repeated as part of a larger peptide panel in Table 1. D, L d m31 (Arg 97 with only the ␣1 and ␣2 domains) was refolded in the presence of 1.2 M peptides MCMV, FLS, and tumϪ and relative protein yields were determined by SDS-PAGE and quantification with ImageJ. The data in D are repeated as part of a larger panel of peptide concentrations shown in Fig. 4.
MHC, thus decreasing the 2C-m6 scTCR binding and fluorescent signal. In this assay, peptide QLSDVPMDL competed with the QL9 peptide on L d W97 and L d R97 with an average difference in BD 50 of 14.7 (Fig. 2B). In contrast, peptide tumϪ (TQNHRALDL) competed with the reference peptide QL9 on L d W97 and L d R97 with an average difference in BD 50 of 0.11fold (L d W97/L d R97) (Fig. 2C). Heptamer peptide MPKPLSL competed with QL9 peptide on L d W97 and L d R97 with an average difference in BD 50 of 0.54 (Fig. 2D).
In total, 8 peptides were evaluated for binding to L d W97 and L d R97 using the same peptide binding competition assay (Table  1). Peptide binding curves were used to calculate BD 50 values, and the relative peptide binding to L d R97 compared with L d W97 was determined by the ratio of BD 50 (L d W97)/BD 50 (L d R97). The results of this analysis indicated that six of the peptides bound better to L d R97 compared with L d W97, whereas the two others (tumϪ and MPKPLSL) bound better to L d W97 compared with L d R97 (Fig. 3). Peptide FLS (FLSP-FWFDI) showed the greatest disparity in binding L d R97 versus L d W97 (Fig. 3 and Table 1). Each of the peptides had a clear preference for binding to one or the other L d position 97 variant; none of the eight tested peptides bound equally well to both L d W97 and L d R97.
Peptide Stabilization of Soluble L d Correlates with the Cell Surface L d Binding Assay-To determine whether the results of the cell-based assays extended to a cell-free system, we used a soluble form of the L d R97 molecule called L d m31 (33). L d m31 contains only the ␣1 and ␣2 domains and the W97R mutation, thus eliminating effects due to ␤2m association. L d m31 was originally selected by yeast display of a library of L d mutants to overcome problems with the use of full-length L d , which has been difficult to express in soluble form in E. coli.
Thus, to assess peptide stability, inclusion bodies of the L d m31 (21 kDa) molecule (3 M) were refolded in the presence of three different concentrations of the peptides MCMV, FLS, and tumϪ. Samples from each refold were concentrated and analyzed by SDS-PAGE to determine amounts of refolded complexes as a measure of stability (Fig. 4A). As expected, L d m31 refolded in the absence of peptide resulted in a low yield of soluble 22-kDa protein. In contrast, L d m31 refolded in the presence of all concentrations of MCMV peptide had the largest yields of peptide⅐L d m31 complexes, followed by FLS⅐L d m31 complexes, and tum⅐L d m31 complexes.
The yields of peptide⅐L d m31 complexes, as determined by quantitative image analysis of gel bands (Fig. 4B), showed that MCMV at 1.2 M peptide was more stable than either FLS or tumϪ at 30 M, consistent with the results of the cell-based competition assay (Figs. 1C and 3, Table 1). A comparison of the yields at 1.2 M peptide in the refolding mixtures showed that the hierarchy of stability of MCMV, FLS, and tumϪ as soluble L d m31 complexes (Fig. 1D) correlated directly with the BD 50 values of the competition assay ( Fig. 1C and Table 1) and the SD 50 values of the up-regulation assay, each with cell surface L d R97 (Fig. 1B). Thus, we conclude that the cell-based competition assay is a valid surrogate for measuring actual stability of peptides with L d .
Arg 97 Tolerates Greater Side Chain Flexibility of Adjacent Residues than Trp 97 -The results above showed that all the peptides tested have a clear preference for binding to one particular L d variant, and more peptides have superior binding to L d R97 compared with L d W97. To explore the possible role of the position 97 residue on structural features of peptide binding, we were in the unique position of having high-resolution crystal structures of several of the peptides bound to either L d W97 (Fig. 5A) or L d R97 (Fig. 5B) (34 -36, 38). Three other adjacent residues, Tyr 156 , Phe 116 , and Glu 114 , are also known to be polymorphic in the human HLA-B gene, although these polymorphisms appeared to be less associated with HIV immunity than position 97 polymorphisms (8, 9, 48 -50). In the L d W97 complexes, Trp 97 formed aromatic stacking interactions with Phe 116 and Tyr 156 , and these three side chains were in similar spatial orientations regardless of the peptide sequence. Although these residues do not appear to interact directly with peptide, they could present a steric barrier that enforces bulging of the peptide backbone. This might be particularly the case when aromatic residues exist at positions 7 and 8 of the peptide (Fig. 5A).
In contrast to the L d W97 structures, in the L d R97 structures Arg 97 was in a position to mediate strong electrostatic interaction with Glu 114 (Fig. 5B). In addition, in the absence of Trp 97 , Tyr 156 and Phe 116 reside too far apart to formstacking interactions with each other. Perhaps as a consequence of the absence of such interactions, in each peptide-L d R97 structure, Tyr 156 was observed in a different spatial orientation, oriented toward the peptide, or away from the peptide, with various degrees of rotation in any plane.
Interestingly, in the peptide-L d R97 structures, the Arg 97 side chain does not apparently use additional flexibility to accommodate different peptides in the binding groove, as has been proposed in the HLA-B system (51). Rather, the absence of the tryptophan at position 97 allowed Tyr 156 to exhibit additional flexibility and thus to assume a variety of orientations. Accordingly, in L d R97 complexes, Tyr 156 is free to stack or interact with aromatic residues of the peptide, such as aromatic residues at position 5 as found in both QL9 and FLSPFWFDI. Two possible reasons for the rigidity of the Arg 97 side chain are evident from existing peptide-L d R97 structures. Electrostatic interactions between side chains of Arg 97 and Glu 114 , as well asstacking interactions between Arg 97 and Phe 116 may contribute to holding Arg 97 in the same position regardless of the peptide sequence. The corresponding plasticity of residue 156 may allow some peptides, including self-peptides, to interact more favorably with alleles that contain Arg 97 .

HLA-B Polymorphic Variants Have Similar Structural Features with L d W97 and L d R97-
The frequency of HLA-B alleles that contained Arg 97 was found to be 1.7-times greater in HIV progressors, compared with controllers (9). Conversely, the frequency of HLA-B alleles that contained Trp 97 was 2 times greater in HIV controllers, compared with progressors. Trp 97 and Arg 97 are positioned in the floor of the HLA-B peptide binding groove (Fig. 6A), in the same position as residue 97 in L d . Position 156 in HLA-B alleles with Trp 97 (40 total alleles) are overwhelmingly skewed toward leucine (97.5%), thus preventing aromatic interactions as seen in the L d system with Tyr 156 . In HLA-B alleles that contain Arg 97 (1360 total alleles), there is greater diversity at position 156 (e.g. 67.5% leucine, 14.3% aspartic acid, 14% tryptophan, etc). Based on our analysis of the L d system, this feature alone may allow additional plasticity in the binding of peptides by the Arg 97 alleles.   In HLA-B alleles with Trp 97 , the majority contained asparagine at position 114 and phenylalanine at position 116 (Fig. 7A). Thus, like L d , the Trp 97 HLA-B alleles likely promote aromatic interactions within the peptide-binding groove, reducing the plasticity associated with binding of some peptides. In HLA-B alleles with Arg 97 , the majority contained an aspartic acid residue at position 114 and a serine residue at position 116 (Fig. 7B). As in the L d system, Arg 97 and Asp 114 likely form electrostatic interactions that accommodate more flexibility in the pocket created at residues in positions 156 and 116 (48). Thus, collectively, the Arg 97 alleles are able to bind a larger collection of peptides, as observed in the L d R97 results here.

DISCUSSION
The results of the recent International HIV Controllers Study suggested that polymorphic position 97 of HLA-B was the most significant residue that correlated with HIV disease control or progression. Although much has been learned about aspects of the HLA-B alleles in terms of disease association and structure, little attention has been paid to a mouse MHC system (H2-L) that has structural and genetic similarities to the HLA-B system, and could inform studies of HIV immunity. Notably, these genetic similarities include two alleles, L d and L q , which express Trp 97 and Arg 97 polymorphisms, respectively (42). Furthermore, the Arg 97 substitution has been shown, by a completely independent approach, to stabilize the peptide⅐H-2L d complex (33).
Our conclusions here are that: 1) L d R97 allows a greater repertoire of self-peptides to be bound more stably, compared with L d W97; and 2) the structural correlate of these peptide-binding properties appears to involve increased plasticity in the class I pocket. The plasticity is enabled by arginine at position 97, but it is not due directly to flexibility of Arg 97 (51), but rather that Arg 97 allows adjacent residues the freedom to move in the pocket. We suggest that these same conclusions likely hold true for the HLA-B alleles, given the striking structural similarities between the two systems.
What is the structural basis for why a substantial number of the individual peptides bound better to L d R97 than to L d W97? Our analysis of the sequences of these peptides, and the structures of several of them in complex with L d R97 (Fig. 5B), does not reveal any obvious residue or motif shared among them (note that for L d , the anchor residues are thought to be a preferred proline at position 2, and a leucine or isoleucine at the C terminus). Thus, rather than pointing toward a specific structural feature of these peptides, the more general basis for their enhanced binding appears to be the nature of the pocket near residue 97.
Residue 97 is centrally located in the peptide-binding groove between the ␣1 and ␣2 domains of the L d heavy chain, with the side chain directed into the groove (Fig. 6B). A bulky tryptophan at that position forces the neighboring residues, including the side chains of residues 114, 116, and 156 into relatively invariant positions. This pocket cannot accommodate peptide variability to the same extent as an allele with Arg 97 at that position. This is because Arg 97 allows some of these same neighboring side chains (and residue 156 in particular) the flexibility to accommodate alternate peptides (Fig. 6B). For example, the structure of peptide QLSDVPMDL showed that the hydroxyl group of the Tyr 156 side chain was displaced by 9.1 Å and a 106°vertical angle compared with the complex with peptide FLSPFWFDI; among the 8 peptides, peptide QLSDVPMDL had significantly differential binding to Arg 97 versus Trp 97 , binding ϳ14-fold better to the Arg 97 variant (Fig. 3). In contrast, Tyr 156 showed no measurable vertical shifts within the peptide-L d W97 structures, but the side chain was observed within a modest 36°rotational angle range; the tyrosine side chains in both structures held both their direction from the backbone and position of the hydroxyl constant. These results also suggest that Arg 97 does not act directly in allowing flexibil-  ity of peptide binding, as has been suggested. Rather, the negatively charged residue at position 114 (Glu in L d and Asp in HLA-B) creates a stable electrostatic interaction that locks Arg 97 into the same position in different structures (Fig. 6B), moving it out of the way of other side chains in the pocket (49). A comparison of two HLA-B complexes (B*14:02 and B*35:01) with either Trp 97 or Arg 97 also shows the close proximities of this same group of residues (Fig. 6B), suggesting the critical importance of their interactions in peptide binding.
To further explore the possible structural parallels with our findings involving L d , and the HLA-B system, we examined in more detail the co-expressed polymorphisms in residues 97, 114, and 116 among the position 97 alleles that operated either as "HIV controllers" (Fig. 7A) or "HIV progressors" (Fig. 7B). We also examined a structural representative of each position 97 allele to gain further insight into the positions of these key residues, and the bound peptides. The analysis included four HLA-B alleles associ-  (Fig. 7).
The frequency analysis of all position 97 alleles (Val 97 , 82 alleles; Asn 97 , 118 alleles; Thr 97 , 491 alleles; and Ser 97 , 549 alleles), in addition to Trp 97 and Arg 97 alleles described above, revealed a clear predominance of commonly associated residue pairs at 114/116. For example, the majority of alleles with valine at position 97, which exhibited the highest allele frequency ratio (5.5) of controllers to progressors, contained an Asp at 114 and a Ser at 116. These exact two amino acids were also the predominant residues found in the Arg 97 alleles, which exhibited the HIV progressor phenotype. Thus, it is reasonable to predict that the combination of residues at these positions impact immune status, presumably by affecting the ability to associate with different numbers or types of self-peptides.
Inspection of the positioning and structures of the various peptides from complexes of the controller or progressor alleles did not provide evidence for a common motif. Although it is interesting that both progressor alleles (HLA-B*35:01 with Arg 97 and HLA-B*07:02 with Ser 97 ) have an unusual position 2 anchor (proline), it remains to be seen if this provides some biochemical feature that enables such peptides to bind more effectively to these alleles. In this regard, it has been suggested previously that there are fewer self-peptides in the human proteome with binding motifs of the controller alleles HLA-B*57:01 (Val 97 ) and HLA-B*27:06 (Asn 97 ) than there are selfpeptides with binding motifs of the progressor alleles HLA-B*35:01 (Arg 97 ) and HLA-B*07:02 (Ser 97 ) (17). It is clear that the size of the self-peptide repertoire for an allele might not only impact CD8 ϩ T cell capacity (e.g. through negative selection in the thymus), but could of course be related to the association of some alleles with autoimmunity.
Although observations showed an association between position 97 alleles of HLA-B and HIV immunity, the lack of a strict correlation implies that other factors also impact immunity. In this regard, a more recent study revealed disparate effector functions of T cells from controllers versus progressors that all expressed the "control-associated" allele HLA-B*27 (52). T cells from the controller population were more potent and crossreacted with variants of the HIV gag epitope. However, there was no evidence presented on the binding affinities and kinetics of the TCRs from the corresponding T cells. We speculate that the TCRs from the controllers exhibited higher affinities/longer off-rates, which would explain both their potency and cross-reactivity (40). If so, the question remains why the HLA-B*27-positive progressors did not have T cells with TCRs of similar functional capacities? We speculate that the answer could be that either these individuals had some polymorphic self-peptides that deleted these T cells, or these individuals had been exposed to foreign peptides (infectious or normal flora) that regulated the peripheral activity of these T cells. A mouse system, for example, L d /L q or HLA-B transgenic mice, could provide models with more limited variables (e.g. same repertoire of self-peptides) to examine some of these issues.
Our findings within the L d system do not support the view that Trp 97 alleles confer greater immunity because they result in a more robust response against some HIV peptide⅐HLA-B complexes. In this scenario, particular HIV antigenic peptides would bind better to Trp 97 allele products than Arg 97 allele products. Contrary to this notion, our data suggest that a substantial fraction of peptides actually bind better to L d R97, whereas fewer peptides bind preferentially to L d W97. Of course, we cannot definitively rule out that there is not some unique peptide from HIV, not yet mimicked in our H-2L study, that binds better to a Trp 97 allele and mediates an enhanced anti-HIV immune response.
Based on our results, we favor the possibility that a subset of self-peptides bind better to Arg 97 alleles, and that these complexes operate in a tolerance mechanism to eliminate or sup-press potential HIV-reactive T cells (8,49,50,53,54). These self-peptides may be predominantly those that have very weak binding to the Trp 97 alleles, and thus one might speculate that this increase in binding ability for Arg 97 sufficiently stabilizes the complexes such that they can operate in either a central or peripheral selection processes. In principle, the repertoire of self-peptides with enhanced binding to the Arg 97 alleles includes among them some that could have interacted with CD8 ϩ T cells during thymic development, resulting in deletion. Furthermore, this premise would require that some of these T cells would have been effective against HIV epitopes. It is possible that these self-peptide⅐R97 complexes function mechanistically not by inducing T cell deletion in the thymus but by stimulating T regulatory cells, which then suppress the HIV response. This mechanism cannot be ruled out. Alternatively, self-peptides can act as antagonists in T cell function (55), and can induce TCR-mediated signaling, which results in peripheral T cell tolerance (see Ref. 17 and reviewed in Ref. 56). Thus, the self-peptides associated with binding to Arg 97 alleles could act to tolerize a peripheral population of potential HIVreactive T cells. Whether there are dominant self-peptides, for example, with HIV peptide structural homologies, that operate in this capacity remains to be seen (57)(58)(59)(60)(61).