Secondary anchor polymorphism in the HA-1 minor histocompatibility antigen critically affects MHC stability and TCR recognition

Detection of VLH-Specific T-Cell Responses in a Chronic Myeloid Leukemia Patient After DLI.

T-cell responses to the HA-1 mHag were detected in an HA-1^H-positive patient diagnosed with chronic myeloid leukemia (CML) who had undergone SCT from a matched, unrelated donor homozygous for the HA-1^R allele (Fig. 1). Following leukemic relapse, the patient received 3 escalating doses of DLI. After the third dose of DLI, VLH-specific CD8 T cells were detected within peripheral blood by binding to HLA-peptide tetramers, and their appearance correlated with elimination of the tumor. VLH-specific T cells were purified by magnetic selection and cloned by limiting dilution assay. The VLH-specific T-cell clone KP7 bound to VLH tetramer and was functionally specific for the VLH peptide presented by HLA-A2. It was also able to recognize naturally processed VLH epitope on VLH-positive LCL, in agreement with previously published data (7, 8) (see Fig. 1).

Proteasomal Generation and TAP Transport of VLH and VLR Variants Is Similar.

To understand the molecular basis of HA-1 mHag immunogenicity, we compared the relative propensities of VLH and VLR HA-1 peptides to undergo proteasomal generation and TAP transport. To determine proteasomal digestion directly, we synthesized extended (27 mer) peptides spanning the VLH/VLR epitope region and incubated these separately with 20S constitutive and immuno-proteasome preparations, before identification of the VLH/VLR species by mass spectroscopy (18). VLH and VLR nonamers, but not N-terminal-elongated precursors, were generated to a similar extent by the constitutive proteasome (within 2-fold), as indicated by appropriate cleavages before V(P1) and after A(P9) [supporting information (SI) Fig. S1A]. Interestingly however, neither the final epitopes nor any N-terminal-elongated precursors could be detected when either VLH or VLR 27 mers were incubated with immunoproteasome preparations, indicating a requirement for constitutive subunits for efficient epitope generation (see Fig. S1A). TAP transport was then assessed using a nonradioactive TAP binding assay based on fluorescence polarization (19). VLH and VLR each competed with the fluorescent test peptide similarly (Fig. S1B), indicating that proteasomal processing in the cytosol and transport into the endoplasmic reticulum by TAP is broadly equivalent in efficiency for the VLH and VLR peptides. These results therefore demonstrate that proteasomal cleavage and TAP transport are unlikely to substantially affect the relative immunogenicity of the two variants.

The VLR Polymorphic Variant Associates with HLA-A2 with Reduced Stability and Affinity.

To address whether the His>Arg polymorphism could affect association with HLA-A2, we carried out circular dichroism (CD) measurements to assess the relative stability of the 2 complexes. HLA-A2-VLH/R complexes refolded with comparable yields at 4 °C and were each purified to a high level at 25 °C. Consistent with this, HLA-A2-VLH/R complexes yielded qualitatively identical CD spectra at 25 °C, confirming each protein had a similar overall fold (Fig. 2A). However, thermal denaturation measurements indicated substantial differences in the stability of the 2 complexes. Whereas HLA-A2-VLH yielded a single denaturation transition (Tm 73.5 °C), an additional folding transition was apparent for HLA-A2-VLR with a Tm 43.1 °C, consistent with dissociation of the complex (Fig. 2B). To assess the effect of the His/Arg alteration on MHC affinity, we then measured the binding of VLH/R peptides to HLA-A2 in a radioactive competition binding assay (20). VLH bound with a mean IC₅₀ of 22 nM, broadly consistent with previous fluorescence-based affinity measurements (6). In contrast, multiple measurements indicated VLR bound extremely weakly to HLA-A2 (mean IC₅₀ 1,296 nM). These data indicate both variants can associate with HLA-A2. However, they also demonstrate that although the His>Arg change is not located at a primary anchor residue position, it has a substantial deleterious effect on peptide-MHC (pMHC) stability, which is manifest as a radical decrease in peptide affinity for HLA-A2.

Crystallographic Structure Determination of HLA-A2-VLH Complex.

To address the molecular basis of the effect of the His>Arg dimorphism on HA-1 affinity for HLA-A2, we determined the crystal structure of the HLA-A2-VLH complex to 1.3Å resolution (Fig. 3 A and B, Table S1). As expected, the VLH peptide adopts an extended structure in the antigen binding groove, with retention of stabilizing interactions from MHC residues to the peptide N and C terminus. The canonical Leu-2 anchor residue is oriented in a standard position in the B pocket, forming extensive contacts with surrounding hydrophobic residues. In contrast, the C-terminal anchor residue is alanine, which is poorly represented at this position in HLA-A2 peptides (21) and forms only extremely scant contacts with the F pocket (Fig. 3C). The paucity of contacts with Ala-9 suggested that interactions mediated by secondary anchor residues, such as the polymorphic His-3, might substantially affect the overall stability of the pMHC complex.

Incorporation of Arg at the P3 Secondary Anchor Actively Destabilizes pMHC Association.

Analysis of the HLA-A2-VLH structure indicated how His-3 is accommodated in the antigen binding groove and suggested possible mechanisms to explain how Arg-3 might decrease complex stability. The His-3 side chain fits relatively snugly in the D pocket (see Fig. 3 C and D), forming van der Waals and hydrophobic interactions with Tyr-99, Tyr-159, and Leu-156 lining the pocket. In addition, His-3 mediates indirect contacts to Gln-155 and the peptide main chain at P5 via a single bridging water molecule. These contacts, and the fact that they are mediated by the imidazole ring (see Fig. 3D), raised the question of whether the His-3 side chain substantially stabilized the pMHC interaction relative to VLR. To address this issue, we generated a variant peptide with a His>Ala substitution at P3 (VLA). Interestingly, the affinity of VLH (22 nM) and mutant VLA (20 nM) peptides for MHC was extremely similar. Furthermore, HLA-A2-VLH/A complexes exhibited similar thermal stabilities [Tm 73.5 °C (VLH) versus 66.1 °C (VLA)] (Fig. S2 A and B). This indicates that the interactions mediated by His-3 make minimal energetic contributions and do not substantially stabilize the pMHC complex.

These considerations suggested the presence of the Arg side chain at P3 actively destabilized the pMHC complex relative to VLH. Based on inspection of the HLA-A2-VLH structure, 2 possible mechanisms were envisaged. First, substitution of Arg-3 could electrostatically destabilize the peptide conformation by introducing a positively charged side chain in relatively close proximity to Asp-4 and Asp-5 side chains. Alternatively, steric or charge effects could hinder the Arg side chain from being accommodated optimally in the D pocket. To test the first hypothesis, we carried out thermal stability measurements on HLA-A2 complexes containing VLR variant peptides with either one or both Asp residues substituted to Ala (VLRD4A, VLRD5A, VLRD45A). In each case, an early unfolding transition was evident (Tm ≈42–49 °C), indicating a thermal stability similar to that of the parental VLR peptide (Fig. S2 C and D). These data indicated the destabilizing effect of Arg-3 was not dependent on the presence of Asp-4 or Asp-5 side chains. In contrast, inspection of the HLA-A2/VLH structure indicated a strong possibility of steric and electrostatic clashes with D pocket residues (e.g., Tyr-99, Gln-155, Leu-156, Leu-160, Tyr-159) when modeling Arg at P3 (Fig. S3 A–D). The sole Arg-3 rotamer that did not clash substantially with the D pocket did involve major clashes with the peptide main chain in the P4-P5 region (Fig. S3E). This suggested that binding of VLR was sterically and electrostatically unfavorable, and was likely to require conformational adjustments in either the central region of the peptide or the HLA-A2 heavy chain itself (most likely in the α2 helix). Because His-3 in VLH is oriented in a similar position to that of many HLA-A2 structures, such limitations were hypothesized to disfavor Arg at P3 in other HLA-A2 peptides. Consistent with this, bioinformatic analyses of HLA-A2-restricted T-cell epitopes confirmed a disproportionately low frequency of Arg at P3 (Fig. S3F), strongly suggesting the effects evident from our analyses of HA-1 VLH/R peptides also apply to HLA-A2 peptides in general.

Complete Discrimination of VLH and VLR Variants by a VLH-Specific TCR.

Despite a relatively low stability, the fact that the VLR peptide associated with HLA-A2 with an affinity similar to that of several low-affinity peptides eluted from cell-surface class I MHC molecules (16, 17, 22) suggested it may be presented at the cell surface at low levels in the steady state. This possibility raised the question of whether the VLH and VLR peptides could be discriminated by HA-1-specific T cells at the level of the TCR. To address this, we generated soluble HA-1-specific TCR heterodimer to characterize the molecular properties of HA-1 mHag recognition and assess how this was affected by His/Arg polymorphism. The TCR chains were amplified from the KP7 T-cell clone, which is functionally specific for the VLH peptide and was generated using standard methods, allowing analysis of KP7 TCR interaction with the HLA-A2-VLH/R complexes using surface plasmon resonance. The KP7 TCR bound HLA-A2-VLH with an affinity of 35 μM (Fig. 4, Table 1), somewhat lower than the typical values of 1–10 μM for human TCR/MHC interactions, most of which involve recognition of viral peptides (23). Consistent with this lower affinity, the KP7 TCR dissociated from HLA-A2-VLH with an extremely fast k_off of 3s⁻¹ (Fig. S4A) (23), suggesting TCR/MHC complexes are highly unstable at the cell surface. In addition, the thermodynamic profile of KP7/HLA-A2-VLH binding was similar to previously studied self-MHC-restricted TCR interactions (Fig. S4B, Table S2). These data suggested KP7 TCR recognition of HLA-A2-VLH was broadly similar to self-restricted TCR/MHC interactions, but involved a somewhat lower affinity and relatively faster dissociation kinetics compared to those previously measured.

To test the effect of the P3 His/Arg residue on TCR recognition of HA-1 nonamers, we directly compared binding of the KP7 TCR with HLA-A2-VLH and HLA-A2-VLR in parallel, at a temperature (25 °C) at which both complexes were stable (see Fig. 4A, Table 1). Whereas binding of TCR to HLA-A2-VLH complex was unambiguous, signals resulting from injection of KP7 over-flow cells containing immobilized HLA-A2-VLR were identical to those over control surfaces, indicating no interaction with the HLA-A2-VLR complex. To understand the energetic basis of this complete discrimination, we tested binding of KP7 to HLA-A2 presenting VLH peptide variants (see Table 1). Interestingly, substitution of Ala for His-3 resulted in an affinity similar to HLA-A2-VLH, suggesting the His-3 side chain itself is not essential for interaction with the KP7 TCR. In contrast, substitution of either Asp-4 or Asp-5 to Ala, both solvent-exposed, completely eliminated binding by the KP7 TCR. These results establish the importance of the central region of VLH for HA-1 recognition by the KP7 TCR and suggest that introduction of the lengthy and charged Arg at P3 is likely to substantially change the conformation and electrostatic nature of this energetically crucial interaction surface, thus providing a plausible mechanism for VLH/VLR discrimination by the TCR.

T-Cell Cloning and Functional Assays.

Patient NC27PM (HA-1^H+) received SCT for CML in 2001 from a matched-unrelated donor (HA-1^RR). In 2002, the patient relapsed and received 3 doses of DLI. PBMC were collected before and after DLI infusions and monitored for VLH-specific cytotoxic T lymphocytes using HLA-A2-VLH tetramers (31). After the third DLI dose, a VLH peptide T cell line was set up using standard procedures, coinciding with molecular disease remission. After enrichment using VLH tetramer, cells were cloned by limiting dilution assay using allogeneic PBMC and LCL. Tetramer-staining clones were tested for lysis of VLH- and VLR-peptide-loaded HA-1^R LCLs in standard chromium release assays (32).

Proteasome Digestion Analysis and TAP Transport Peptide Binding Assays.

Proteasomal generation of VLH and VLR nonamer species was tested by incubation of extended HA-1 peptide sequences (FAEGLEKLKECVLHDDLLEARRPRAHE and FAEGLEKLKECVLRDDLLEARRPRAHE) with purified 20S constitutive or immuno-proteasome preparations, as previously described (33). Peptide fragments were separated by capillary liquid chromatography and analyzed by mass spectrometry, with synthetic nonamer peptides VL(H/R)DDLLEA used as standards for relative quantification of epitope generation. Peptide binding to the human TAP, over-expressed in Sf9 insect cell microsomes as described (34), was measured using a nonradioactive fluorescence polarization assay (19), with relative TAP affinities of test peptides determined as the concentration required to inhibit 50% of specific TAP binding of the high affinity FITC-concjugated reporter peptide RRYNACTEL (IC₅₀). Each peptide was tested in at least 2 independent experiments. Unlabeled, unsubstituted reporter peptide RRYNASTEL was included in each assay for normalization of results. Results are expressed as the ratio of the IC₅₀ of the test peptide to the IC₅₀ of unlabeled R9L.

MHC Affinity and Stability Experiments.

The affinity of VLH and VLR peptides for HLA-A2 was determined using an equilibrium assay that measures the ability of test peptides to compete with the radio-labeled FLPSDYFPSV standard peptide for binding to HLA-A2 (20). Purified class I molecules were incubated with test peptides and after 48 h, class I-peptide complexes separated from free peptide by gel filtration. Class I-bound and -free radioactivity was measured and the doses of test peptides yielding 50% inhibition of labeled peptide binding (IC50, nM) were calculated. Each peptide was tested in 2 to 4 independent experiments. CD experiments were carried out using a JASCO J-810 Spectropolarimeter, in 20 mM Hepes pH 7.4, 10 mM NaCl.

Class I MHC Complex Production, Crystallization and Structure Determination.

HLA-A2-VLH and HLA-A2-VLR complexes were produced by dilution refolding (35), and subsequently purified by gel filtration. Crystals of HLA-A2-VLH were obtained in 20% PEG 10K, 0.1M Hepes pH 7.5 using a Mosquito crystallization robot (TTP Labtech). Before collection of 1.3Å data at ID14-EH4 (ESRF, France), the crystal was soaked in precipitant buffer containing 15% ethylene glycol and flash frozen to 100K. Data were integrated using MOSFLM (36) and scaled using the CCP4 suite (37). The structure was solved by molecular replacement in CNS (38) using an HLA–A2 complex as a search model. The model was subjected to alternate cycles of model building [TURBO-FRODO, (39)] and refinement [CNS and SHELX (40)]. The final structure comprises residues 1–274 (heavy chain), 2–99 (β₂M), 1–9 (peptide), and 336 water molecules, and has an R_fac of 16.9% and R_free of 20.4% (see Table S1 for statistics). Atomic coordinates and structure factors have been deposited in the RCSB Protein Data Bank (accession code 3D25).

Soluble TCR Generation and Surface Plasmon Resonance Studies.

KP7 TCR genes (TRAV41, TRAJ28; TRBV7–9, TRBJ2–1) were amplified by Rapid Amplification of cDNA ends (RACE) PCR using 3′ primers annealing to the TCR α and β constant regions. For binding studies, TCR-leucine zipper fusion protein was produced in Escherichia coli, as previously described (41). KP7 TCR preparations were purified by gel filtration immediately before use to eliminate protein aggregates. Surface plasmon resonance experiments were performed using a BIAcore 3000 in HBS-EP, as described (29). Biotinylated class I MHC complexes were immobilized by injection over streptavidin-coupled surfaces (42). For kinetic measurements, high-flow rates (50–100 μl/min) and low class I MHC immobilization levels (250–500RU) were used to eliminate mass transport effects (42). Data analysis was carried out in BIAevaluation 3.1 and Microcal Origin. The extinction coefficient of KP7 was determined by amino acid analysis to be 65860 M⁻¹·cm⁻¹. For more information, please see SI Text.