A Novel MHC-I Surface Targeted for Binding by the MCMV m06 Immunoevasin Revealed by Solution NMR*

Background: The m06/gp48 protein of MCMV binds to MHC-I proteins, diverting them to lysosomes. Results: Recombinant m06 binds weakly to H2-Ld MHC-I and tightly to mini-H2-Ld, which provides excellent NMR spectra for mapping the binding site. Conclusion: The binding site on MHC-I partially overlaps with the β2m interface. Significance: Thus, m06 may alter the conformation of β2m association with MHC-I heavy chain following m06 binding in a viral infection. As part of its strategy to evade detection by the host immune system, murine cytomegalovirus (MCMV) encodes three proteins that modulate cell surface expression of major histocompatibility complex class I (MHC-I) molecules: the MHC-I homolog m152/gp40 as well as the m02-m16 family members m04/gp34 and m06/gp48. Previous studies of the m04 protein revealed a divergent Ig-like fold that is unique to immunoevasins of the m02-m16 family. Here, we engineer and characterize recombinant m06 and investigate its interactions with full-length and truncated forms of the MHC-I molecule H2-Ld by several techniques. Furthermore, we employ solution NMR to map the interaction footprint of the m06 protein on MHC-I, taking advantage of a truncated H2-Ld, “mini-H2-Ld,” consisting of only the α1α2 platform domain. Mini-H2-Ld refolded in vitro with a high affinity peptide yields a molecule that shows outstanding NMR spectral features, permitting complete backbone assignments. These NMR-based studies reveal that m06 binds tightly to a discrete site located under the peptide-binding platform that partially overlaps with the β2-microglobulin interface on the MHC-I heavy chain, consistent with in vitro binding experiments showing significantly reduced complex formation between m06 and β2-microglobulin-associated MHC-I. Moreover, we carry out NMR relaxation experiments to characterize the picosecond-nanosecond dynamics of the free mini-H2-Ld MHC-I molecule, revealing that the site of interaction is highly ordered. This study provides insight into the mechanism of the interaction of m06 with MHC-I, suggesting a structural manipulation of the target MHC-I molecule at an early stage of the peptide-loading pathway.

The endogenous protein antigen processing and presentation pathway provides the cell with an important surveillance mechanism that protects against invading pathogens. This involves the cell surface display of intracellularly processed protein fragments within the peptide-binding groove of fully assembled MHC-I molecules (including the light chain ␤ 2 m) for recognition by CD8 ϩ cytotoxic T cells and NK cells. To counter this host defense, viruses that establish long term latent or persistent infections have evolved intricate strategies to evade the immune response (1). MCMV 5 encodes three proteins that interact with MHC-I to interfere with T cell and NK cell recognition of infected cells: m152/gp40, itself an MHC-I structural homolog, as well as m06/gp48 and m04/gp34, members of the MCMV m02-m16 family (2). The m152 protein is believed to associate transiently with MHC-I molecules and, by an unknown mechanism, arrest MHC-I maturation in the early secretory pathway (3,4). Although both m04 and m06 bind MHC-I, their association has varying effects; m06 reroutes MHC-I molecules to lysosomes using a dileucine sorting signal encoded in its cytoplasmic tail (5), whereas m04 partially counters the maturation arrest of m152, allowing a fraction of the cell's MHC-I molecules to reach the cell surface (6,7). Although there are several genetic and functional studies on the combined effects of m04 and m06 interference with MHC-I (7-10), the exact molecular mechanism of their interaction with MHC-I remains unknown, in part due to impediments to the co-crystallization of their complexes.
NMR spectroscopy provides an attractive alternative to crystallography for studying the structure and dynamics of macromolecular assemblies in solution. When isotope-labeled molecules can be recombinantly expressed and purified in milligram quantities, solution NMR is a powerful technique to map the interaction surfaces, to determine the conformations of bound ligands and complex structures, and to characterize any dynamic and conformational changes upon binding. Previous NMR (11) and x-ray (12) structural studies of the m04 luminal domain concurrently revealed an intricate variant of the Ig-fold that is conserved among members of the m02-m16 immmunoevasin family, including m06, suggesting a conserved mechanism for MHC-I binding. Using high resolution NMR, Varani et al. (13) exploited a truncated H2-L d molecule to map the binding site of the 2C T cell receptor on the peptide⅐MHC-I complex. First, they engineered mini-H2-L d , a minimal ␣1␣2 platform molecule derived from the murine H2-L d protein, which lacks the ␣3 domain and does not require the light chain ␤ 2 m for refolding in vitro (14). Subsequent structural characterization showed that this molecule preserves the binding epitopes for T cell recognition in their native orientation, as indicated by co-crystal structures of several peptide⅐MHC-I/TCR complexes (15). More recently, solution NMR studies of the 2C/mini-H2-L d system explored recognition dynamics at the MHC-I/TCR interface (16,17).
Here we examine the interaction of an additional member of the m02 family, m06, with the mini-H2-L d construct. We employ bacterially expressed recombinant m06 as well as fulllength H2-L d ⅐␤ 2 m and mini-H2-L d refolded with a high affinity peptide as probes to characterize these interactions. The ability to make complete backbone assignments of mini-H2-L d permitted us to characterize its interaction with m06 through a discrete site located under the peptide-binding platform that partially overlaps with the ␤ 2 m interface on the MHC-I heavy chain. The identification of this region is consistent with in vitro binding experiments showing much weaker complex formation between m06 and ␤ 2 m-associated MHC-I. NMR relaxation experiments characterize the picosecond-nanosecond dynamics of the free MHC-I molecule. This study defines an MHC-I binding site for a member of the m02-m16 family and suggests a novel strategy exploited by a viral immunoevasin to bind MHC-I molecules.

Experimental Procedures
Protein Production and Purification-The luminal domain of MCMV m06 was PCR-amplified from the plasmid gp48HA-PPM (originally amplified from Smith strain MCMV Bac DNA), the kind gift of Dr. A. Hudson (18), using primers 5Ј-TTTTTTCATATGGGAGAATCGCTAATA-3Ј and 5Ј-TTT-TTTGGATCCTTACTGGCGGCGGCGGCGTGATGG-3Ј and cloned into pET21b. The codon encoding the cysteine at m06 position 47 was mutated to alanine by PCR mutagenesis using primers 5Ј-ATCGTGGGGGCCAACGTTTCGCGTACC-GAG-3Ј and 5Ј-AACGTTGGCCCCCACGATAGGTTCCTC-3Ј and verified by DNA sequencing. The luminal domain of m06 was expressed in Rosetta 2 (DE3) Escherichia coli as inclusion bodies. Following washing in Tris/EDTA and solubilization in 6 M guanidine HCl, protein was refolded by dilution into refolding buffer (0.4 M arginine HCl, 0.1 M Tris, pH 8, 2 mM EDTA, 5 mM reduced glutathione, 0.5 mM oxidized glutathione) for 4 days at 4°C; dialyzed against 150 mM NaCl and 25 mM MES, pH 6.5; concentrated with an Amicon stirred cell concentrator using an Ultracel 10-kDa ultrafiltration-regenerated cellulose filter (Millipore); purified by gel filtration on Superdex HR 75 in the same buffer; and maintained at a temperature of 4°C. The luminal domain of H2-L d was expressed from a pET3a plasmid as inclusion bodies in BL21 E. coli and refolded with peptide YPNVNIHNF ("NIH peptide") (19) and murine or human ␤ 2 m, as described for H2-D d (20). The luminal domain of H2-D d was also refolded with peptide RGPGRAFVTI ("P18-I10 peptide") and murine ␤ 2 m as described (20). Mini-H2-L d in pET28a (14), the kind gift of Dr. D. Kranz, was expressed as inclusion bodies in BL21 Codon Plus (DE3) RIPL cells and refolded with the NIH peptide or QL9 peptide (sequence QLSPFPFDL (21)). Protein was purified by gel filtration and ion exchange chromatography on mono-Q, showing a single, monodisperse peak in the chromatogram. Isotope-labeled preparations of H2-L d were made with a similar protocol but instead using 13 C-, 15 N-, and 2 H-substituted M9 minimal media to prepare the inclusion bodies as described for m04 (11).
Surface Plasmon Resonance (SPR)-For the experiments shown, monoclonal antibody 30-5-7S (22), which binds the ␣2 domain of H2-L d (23), was covalently coupled to a BIAcore T100 CM5 surface by standard NHS-EDC chemistry as described previously (24). Then either full-length H2-L d refolded with the NIH peptide and either mouse or human ␤ 2 m or mini-H2-L d was captured on the antibody surface, followed by offering m06 to the captured H2-L d in graded concentrations. In competition experiments, mini-H2-L d was first captured and then exposed to different concentrations of m06 prior to being offered 1.2 mM murine ␤ 2 m. Apparent dissociation rates were calculated using BIAevaluation software. All SPR experiments were performed at 10°C.
Native Gel Shift Assays-Protein samples were incubated in native gel loading buffer (250 mM Tris, pH 8.8, 10% glycerol) for 30 min at 4°C. Samples were loaded on 8% polyacrylamide gels at 90 V (constant voltage) at 4°C for 3.5 h in 25 mM Tris and 190 mM glycine running buffer. Following electrophoresis, proteins were visualized with PageBlue protein staining solution (Thermo Scientific) or transferred to a nitrocellulose membrane for immunoblot analysis.
NMR Backbone and Side Chain Assignments, and Backbone Relaxation Rate Measurements-All experiments were recorded at a temperature of 25°C using 600-, 800-, and 900-MHz cryoprobe-equipped Bruker spectrometers. We used an array of TROSY-based triple-resonance assignment experiments A full set of R 1 , R 1 , and 15 N-{ 1 H} NOE relaxation spectra were recorded at 600 MHz and supplemented with R 1 measurements at 900 MHz from a 0.9 mM perdeuterated, amide 1 H sample using TROSY readout methods (30). R 2 rates were obtained from rotating frame R 1 rates (31) measured under a spin-lock field strength of 2 kHz, after correction for the 15 N off-resonance, tilted field. Uncertainties in the R 1 and R 1 measurements were estimated from the spectral noise levels using 21 Monte Carlo simulations, whereas uncertainties in the 15 N-{ 1 H} NOE ratios were propagated directly from the noise levels in the reference and attenuated spectra. To assess consistency between R 1 and R 2 rates and the presence of oligomer formation in the sample, we performed simulations of relaxation rates under the full spectral density function formalism using an inhouse perl script. To measure backbone amide residual dipolar couplings (RDCs), we prepared a liquid crystalline sample containing 15 mg/ml Pf1 phage in 50 mM NaCl (32). We used the ARTSY method for quantitative measurement of RDCs at 900 MHz (33). The agreement of the experimentally determined RDCs to the x-ray structure of mini-H2-Ld (PDB entry 3TF7) is quantified by the Q-factor, which reports the deviation of the back-calculated RDCs using the crystallographic coordinates relative to a range of RDCs estimated from a randomly distributed set of vectors assuming an alignment tensor of Known D a and R parameters, where D a and R represent the magnitude and rhombicity of the alignment tensor, and D calc and D obs are the calculated and observed RDCs, respectively. The five parameters of the alignment tensor were determined by best-fitting, using a singular value decomposition process (34), of the experimental RDCs to the x-ray coordinates of mini-H2-Ld in complex with the QL9 peptide (PDB entry 3TF7), yielding D a ϭ Ϫ15.4 Hz, and R ϭ 0.29. The fitting process was carried out using custom-built routines within the program Rosetta (35).

Results
Direct Interaction between m06 and H2-L d MHC-I-Initial preparations of the recombinant full-length luminal domain of m06, excluding the transmembrane and cytosolic domains, were prone to precipitation and interchain disulfide bond formation. From a sequence alignment with the structurally related m04 protein, we identified a putatively free cysteine in m06 at position 47 ( Fig. 1). Mutagenesis of Cys-47 to Ala resulted in the significantly more stable m06 protein used in this study. To evaluate the ability of m06 to bind MHC-I molecules, we first performed native gel shift assays, mixing recombinant m06 with different MHC-I constructs. Guided by previously reported co-immunoprecipitation results (5, 36), we probed for binding of the MHC-I allotype H2-L d to the recombinant m06 (Fig. 2a). The MHC-I constructs were refolded in vitro from bacterially produced inclusion bodies in the presence of murine ␤ 2 m and the high affinity NIH peptide (19,37). These first assays were consistently negative using different peptide⅐MHC-I complexes, including H2-D d and H2-L d refolded with human ␤ 2 m (Fig. 2, a and b). We then turned to the previously engineered mini-H2-L d protein that lacks the ␣3 domain and does not require ␤ 2 m for refolding or peptide binding (14). Preliminary screening of mini-H2-L d for m06 binding by a native gel shift assay revealed the appearance of a new band, migrating more slowly than the mini-H2-L d construct alone (Fig. 2c). To confirm that the new band represented a complex The positions of the two disulfide bonds shared by m04 core and the model of m06 are indicated with yellow brackets and connecting yellow lines, whereas the position of a free cysteine in m06 is indicated with a yellow bracket and an asterisk. Sequence alignment was generated using m04 sequence from PDB entry 2MIZ (11), and a homology-based model of m06 was generated with I-TASSER (38), using the first model in the NMR ensemble of m04 as a structural template (PDB entry 2MIZ) (11).
containing mini-H2-L d , we subsequently transferred proteins to a membrane and immunoblotted with the 64-3-7 antibody, which detects an epitope on the H2-L d ␣1 domain (39) (Fig. 2d). Considering that 64-3-7 recognizes a linear epitope present only on MHC-I molecules lacking peptide, we interpret this signal as detecting those H2-L d molecules that have lost peptide during the process of transfer to the nitrocellulose membrane, following native gel electrophoresis.
To validate further the interaction between m06 and H2-L d , we performed SPR binding studies, first capturing H2-L d on a surface coupled with 30-5-7S antibody to the ␣2 domain of H2-L d , followed immediately by exposure to up to 25.3 M m06 (Fig. 2, e-g). We found that m06 bound robustly to the 30-5-7S-captured mini-H2-L d protein. However, m06 interacts only weakly with full-length H2-L d ⅐m␤ 2 m with a more rapid apparent dissociation rate (Fig. 2f). H2-L d refolded with human ␤ 2 m showed detectable (within the limits of the method) but even weaker binding to m06 than the H2-L d ⅐m␤ 2 m complexes (Fig.  2g). The apparent dissociation rates, k d , of the binding of m06 to mini-H2-L d , H2-L d ⅐m␤ 2 m, and H2-L d ⅐h␤ 2 m, plotted in Fig. 2h, are consistent with the inability to detect complex formation of the full-length molecules in the gel shift assay. The continual dissociation of the captured H2-L d protein from the antibody surface complicates quantitative analysis of the binding of m06 in this system, but consistently we observed strong binding by mini-H2-L d and much attenuated binding in the case of H2-L d ⅐h␤ 2 m. The apparent ability of m06 to bind tightly to the mini-H2-L d suggests that m06 may recognize a surface that is partially occluded on folded full-length, ␤ 2 m-associated MHC-I molecules.
To explore further the potential for competition between m06 and ␤ 2 m, we performed SPR experiments, first capturing mini-H2-L d on a 30-5-7S surface, followed by exposure to m06, and then quickly shifting to a high, 1.2 mM concentration of murine ␤ 2 m (Fig. 2i). This concentration of ␤ 2 m resulted in clear association with the captured mini-H2-L d but no nonspecific association with the antibody surface without captured mini-H2-L d . When the captured mini-H2-L d was first exposed to m06, the increase in binding (resonance units) during the ␤ 2 m binding step proportionately decreased, suggesting that m06 occupancy on mini-H2-L d molecules prevents subsequent ␤ 2 m binding (Fig. 2, i and j). Finally, using NMR, we screened a sample containing 0.15 mM 13 C/ 15 N/ 2 H-labeled full-length H2-L d MHC-I heavy chain, refolded with unlabeled human ␤ 2 m and NIH peptide, for binding to unlabeled m06 at a 1:1 molar ratio. The absence of any chemical shift changes or line broadening confirms that the K d of the interaction between m06 and the full-length MHC is weaker than ϳ1 mM under the NMR conditions, consistent with a competition between ␤ 2 m and m06 for binding to a partially overlapping MHC-I epitope.
Peptide Optimization of Mini-H2-L d Construct and NMR Backbone Assignments-Whereas a single MHC-I allele can generally refold using any of a number of peptides that satisfy a set of sequence rules imposed by the structure of the MHC-Ibinding pockets (known as peptide "motifs" (40)), the stability of the resulting peptide⅐MHC-I complex is greatly influenced by the intrinsic affinity of the bound peptide because molecules that lose their peptides are prone to aggregation (41). NMR  allows the exploration of the conformational stability of MHC-I complexes with different bound peptides by monitoring the "fingerprint" features of two-dimensional amide 1 H-15 N correlation spectra over time. Toward this end, we prepared samples of the mini-H2-L d construct refolded in vitro with two different peptides. In particular, we used the QL9 peptide (21) (also employed previously to study binding to the 2C TCR (13,16,17)) and a high affinity NIH self-peptide. Although initially, both samples gave well dispersed two-dimensional TROSY spectra, each indicative of a properly conformed peptide⅐ MHC-I complex, the sample prepared with QL9 deteriorated within 1 week at 25°C. This is consistent with the previous NMR study of the QL9 peptide⅐mini-H2-L d complex, which reported transient self-association, leading to signal loss in the NMR experiments due to intermediate time scale conformational exchange with a high molecular weight form (13).
Notably, using the higher affinity NIH peptide, we were able to obtain stable samples of improved spectral quality and stability. We therefore prepared a 0.9 mM sample of 13 C/ 15 N/ 2 Hlabeled mini-H2-L d folded with the NIH peptide (Fig. 3). Using this sample, we recorded a three-dimensional 15 N-separated amide NOESY spectrum in addition to a full array of triple resonance experiments (HNCO, HN(CA)CO, HNCA, and HN(CA)CB). From these complementary data sets, we obtained complete backbone assignments by mapping residue connectivity through the amide hydrogens, in addition to the backbone CO, C␣, and C␤ atoms (42). The improved stability and lack of aggregation were evident even after 2 weeks of data collection at 25°C, where there was no loss of signal in twodimensional TROSY-HSQC spectra. The spectral quality of mini-H2-L d with NIH peptide enabled us to record RDCs under dilute alignment conditions using a Pf1 phage liquid crystalline sample (32) (as outlined in detail under "Experimental Procedures"). RDCs are sensitive probes of the local and long range structure that report on the relative orientation of backbone amide vectors along the protein sequence (43). Their quantitative comparison with the same parameters back-calculated from the x-ray coordinates of mini-H2-L d confirms that the solution structure of our NMR construct is very similar to the crystal structure, with a Q-factor of 31%. Such a Q-factor is typical for the levels of structural noise present in x-ray structures (44). With this information at hand, we proceeded to examine the backbone dynamics of mini-H2-L d in detail and to map the binding site for m06.
Backbone Dynamics of the Peptide⅐MHC-I Complex-The prolonged stability of the mini-H2-L d ⅐NIH peptide sample allowed us to perform a full suite of 15  Inspection of backbone relaxation rates of different amide sites along the MHC-I sequence (Fig. 4) shows several flexible regions of the molecule. In particular, loop residues 13-18 (␤1-␤2), 39 -43 (␤3-␤4), and 87-92 (␣1-␤5), all located on the ␣1 half of the molecule, show 15 N-{ 1 H} NOE and R 2 values that are much below average (ϳ0.8 and 20 s Ϫ1 , respectively, in Fig. 4, a  and b) and increased R 1 values (Fig. 4c), which are indicative of increased mobility on the picosecond-nanosecond time scale. The last four residues at the C terminus (not visible in the crystal structures) are also highly flexible. In addition, Asp-29 FIGURE 4. 15 N NMR relaxation data reporting on the backbone mobility of mini-H2-L d for each residue, using the crystal structure (PDB entry 3TF7) as a reference. a, 15 N-{ 1 H} NOE values recorded at 600 MHz using TROSYbased methods. b, R 2 values (600 MHz), as obtained from measured R 1 rates using a 2-kHz spin-lock field, after correction for off-resonance effects (31). c, longitudinal 15 N relaxation rates, R 1 , (red, 600 MHz data; yellow, 900 MHz). All data were recorded using a 0.9 mM 15 N/ 2 H-labeled sample of mini-H2-L d in the NMR buffer (20 mM PIPES, pH 6.4, 50 mM NaCl) and are consistent with a predominantly monomeric form (70%) with a rotational correlation time of 13.2 ns/rad. The secondary structure diagram based on the DSSP annotation of the crystal structure is shown at the top as a guide. located on the ␤ 2 -␤ 3 loop shows an increased R 2 rate, indicative of exchange between different conformations on the micro-to millisecond time scale. In contrast, the three shorter loops facing the ␣2 helix (residues 28 -30, 104 -108, and 126 -133) appear well ordered according to the NMR relaxation data and are stabilized by a network of hydrogen bonds in the crystal structure of the molecule. The comparative rigidity of the ␣2 domain as a whole is consistent with the stabilization resulting from the conserved disulfide bond bridging the ␣2 helix with the ␤-sheet of all MHC-I molecules.
Identification of a Discrete Binding Site in Slow Conformational Exchange-To identify the binding site of m06 on MHC-I, we mixed unlabeled m06 with 13 C/ 15 N/ 2 H-labeled mini-H2-L d at a ϳ0.8:1 molar ratio. As expected from the native gel shift results, m06 forms a tight complex with MHC-I under the NMR conditions, with nearly all MHC-I molecules in the bound state (Fig. 5a). The exchange between free and bound forms of MHC-I is slow on the NMR time scale, as indicated by the observation of a separate set of peaks for the complexed form (Fig. 5c). The rate of exchange between the free and bound forms was further investigated by NOE experiments and ZZexchange experiments with long mixing times relative to 1/R 1 (48). However, these did not show any observable cross-peaks between the free and complex peaks, consistent with an upper limit for the dissociation constant in the low micromolar range, assuming a diffusion-limited on-rate. Because the resonances in the complex are shifted significantly in the 15 N, amide 1 H, and 13 CO dimensions, resonance assignment for the complexed form in crowded regions of the spectra is not straightforward (Fig. 5b). We therefore recorded optimized HNCA and HN(CA)CB spectra, in addition to a three-dimensional 15 Nseparated amide NOESY spectrum. These spectra allowed us to unambiguously assign the peaks in the complex and identify the 15 N, amide 1 H, and 13 CO dimension chemical shift changes upon m06 binding to mini-H2-L d . The combined changes, scaled relative to 1 H (Fig. 6a), reveal several sites along the MHC-I surface that are affected by m06 binding. The chemical shift changes, although spread out along the primary sequence, cluster in a contiguous region on the MHC-I structure that is underneath the ␣1␣2 MHC-I platform, on the opposite face of the peptide-binding groove (Fig. 6b). The affected residues form a "wedge" on the surface of the MHC-I molecule, with the most strongly perturbed residues found in the ␤-strands at residues 95-99 and 115-125, forming a portion of the floor of the peptide-binding platform, as well as a region from residue 133 to 146, which forms part of the ␣2 helix. Notably, the association with m06 causes virtually no chemical shift changes in residues distant from the binding site, including all residues of the ␣1 and part of the ␣2 helix, which are responsible for retaining bound peptide (Fig. 6b). The absence of significant chemical shift changes in the ␣1 helix and most of the ␣2 helix indicates that binding of m06 does not lead to release of MHC-I-bound peptide. Formally, changes in chemical shift indicate perturbation of the magnetic environment of the affected atoms, which could result from a change in dynamics or from small changes in local electric fields caused by binding. Notably, the resonances of the shifted residues showed line widths that were very similar to sites not perturbed by binding, indicating the absence of substantial changes in backbone dynamics. The presence of a contiguous area of sites with substantially perturbed chemical shifts on the surface of the protein therefore can be safely interpreted as representing the m06 binding site (49).

Discussion
MHC-I interference by viral proteins is a multifaceted process involving the targeting of a number of distinct intermediates along the peptide loading and presentation pathway (1). For several proteins focused on the fully assembled MHC-I, the molecular mechanism of their interaction has been elucidated by x-ray crystallography of their MHC-I-bound complexes (50 -52). Nevertheless, several stages of the MHC-I pathway have been resistant to crystallographic study, perhaps due to the dynamic nature of the complexes involved. Our results show an example in which high resolution NMR on a previously engineered minimal MHC-I construct is used to probe the dynamics of the free MHC-I construct and to characterize the complex form, revealing a novel m06 binding site on MHC-I, which is used by the virus to thwart the antigen presentation pathway inside the cell. These results demonstrate that the luminal domain of the MCMV m06/gp48 immunoevasin interacts only weakly with fully assembled MHC-I peptide⅐heavy chain⅐␤ 2 m heterotrimers but binds with high affinity to mini-H2-L d , which renders it amenable to detailed analysis by NMR.
The minimal MHC-I construct used in this study is an ideal model system to map the binding of cognate T cell receptors as well as viral molecules that may recognize the peptide-binding platform domain of the molecule. Sample optimization through careful selection of the MHC-I-bound peptide has enabled us to prepare labeled samples of prolonged stability and in milligram quantities typically required for NMR studies. The previous NMR characterization of the same construct lacked assignments of multiple amino acids under the ␣1␣2 platform (13), and our assignments of those amino acids were instrumental in the precise mapping of the m06 binding site. The availability of a stable, fully assigned mini-H2-L d construct enables an approach toward rapid and direct characterization of the binding sites for other putative MHC-I-associated proteins in a systematic manner using NMR spectroscopy.
During an MCMV infection, the m06 protein binds to and redirects full-length MHC-I molecules to lysosomes (5). Previous studies characterizing the interaction between m06 and MHC-I by pulse-labeling immunoprecipitation have shown that the ␤ 2 m subunit is part of the complex that is bound by m06 (3,36). The ratio of recovered ␤ 2 m to heavy chain following m06 immunoprecipitation appears comparable with that observed in immunoprecipitations of properly folded, ␤ 2 m-associated heavy chain directly, suggesting that ␤ 2 m is an integral component of the MHC-I⅐m06 complex formed within the cell (3,36). Despite this, we have only observed tight association between m06 and the mini-H2-L d construct, whereas fulllength MHC-I molecules (lacking the transmembrane domain) in association with ␤ 2 m bind m06 weakly in vitro. This difference in in vitro binding efficiency can be partially explained by the overlap between the observed binding site on the mini-H2-L d protein and the binding site of ␤ 2 m on full-length MHC-I molecules, each on the underside of the MHC-I peptide binding platform (Fig. 7). Our ␤ 2 m competition experiments using SPR, in which binding of m06 to the mini-H2-L d construct prevents subsequent ␤ 2 m association, further validate the NMR binding observation. We note that a crystal structure of the HLA-Aw68 ␣1␣2 domain complexed with h␤ 2 m reveals that the interaction site of ␤ 2 m in the absence of the ␣3 domain is the same as that of the intact heavy chain (53). This further supports the identification of the m06 footprint on MHC-I that we have determined here.
The overlap between the binding sites of m06 and ␤ 2 m raises the question of how the simultaneous interaction between m06, MHC-I heavy chain, and ␤ 2 m occurs within the cell. Notably, m06 recognizes an extensive surface on MHC-I, including part of the ␣2 helix, thereby resulting in a much larger total footprint in the area under the ␣1␣2 platform than ␤ 2 m. This surface also partially overlaps with the putative MHC-I binding site of the chaperone protein tapasin, which has been mapped by mutagenesis (54 -56). m06 is expressed early during viral infection and first associates with MHC-I in the endoplasmic reticulum, the compartment where MHC-I folds and is assembled with ␤ 2 m and peptide (5). As a result, m06 might favorably compete with ␤ 2 m for the overlapping region on MHC-I, with the heavy chain using m06 as a scaffold for initial folding instead of or in addition to ␤ 2 m (Fig. 8). Our in vitro binding assays require first preparing m06 and MHC-I separately, thus not enabling m06 to co-assemble with MHC-I molecules to form the three-component complex. Our SPR data support this model. In comparison with the mini-H2-L d protein, properly conformed, full-length MHC-I shows reduced m06 binding in vitro. Furthermore, full-length H2-L d in complex with human ␤ 2 m, which is well known to have a tighter association with heavy chain than murine ␤ 2 m (57-59), further limits m06 binding in vitro.
The characterization of the binding site by NMR and by SPR binding competition suggests that in order for m06 to co-assemble with MHC-I, the MHC-I/␤ 2 m interface undergoes a local structural rearrangement. Knowledge of the dynamics of MHC-I heavy chain/␤ 2 m interaction has largely been restricted to analysis of crystal structures, in which flexible regions of either protein are constrained in a single conformation. Recent studies of ␤ 2 m using NMR, both free in solution and in complex with MHC-I heavy chain, reveal a significant degree of flexibility in certain regions of ␤ 2 m (60, 61). In particular, the ␤-strand/ loop composed of amino acids 53-63 of ␤ 2 m, which makes contact with the floor of the ␣1␣2 platform of the heavy chain, is flexible in free ␤ 2 m and retains much of that flexibility upon association with heavy chain (61,62). For example, the Lys-58 residue of human ␤ 2 m shifts by ϳ3.9 Å between its free and heavy chain-bound conformations and also shifts its conformation upon association with CD8␣␣ (62,63). The flexible 53-63 loop of ␤ 2 m interacts with the ␣1␣2 platform primarily in the same region bound by m06 (Fig. 7b). This observation suggests that the association between m06 and full-length, ␤ 2 m-bound heavy chain may involve a dislocation of this loop of the ␤ 2 m subunit, resulting in an alternate conformation of ␤ 2 m in complex with heavy chain and m06 in the context of an MCMV infection (Fig. 8). Alternatively, a contribution to the association by the transmembrane regions of the two molecules, not present in the constructs used here for the in vitro experiments, could explain the co-immunoprecipitation of m06 with MHC-I in the presence of ␤ 2 m.
The nucleotide sequences of many strains of MCMV have been determined, and their putative immunoevasion genes have been examined (64,65). In contrast to other members of the m02-16 family, which show high sequence divergence among MCMV strains, m06 sequences are extremely well conserved (65,66). On the other hand, sequences of the structurally related m04 protein, also known to associate with MHC-I, are significantly variable among strains (64). In functional studies, m06 displays a broad specificity for murine MHC-I molecules (10). These observations, in combination with the strong sequence conservation of both the ␤ 2 m-binding surface and tapasin-binding regions of the class I heavy chain, help to explain both the evolutionary preservation of the sequence of m06 and its target specificity. To illustrate this, we aligned the sequences of the ␣1␣2 domains of the mini-H2-L d used in this study with those of wild-type H2-L d as well as the more distantly related human HLA-A2 molecule (Fig. 7d). Due to constraints through their need to associate with ␤ 2 m, chaperone proteins like tapasin, and TCR and co-stimulatory molecules, the sequence variations among MHC-I alleles are primarily focused around the peptide-binding groove, whereas other surfaces are relatively conserved. The comparison of the sequence of the surface bound by the mini-H2-L d and HLA-A2 illustrates this well; with only a few exceptions, all residues on the m06binding surface of the mini-H2-L d are identical in HLA-A2, or the variations in HLA-A2 are on the floor of the peptide-binding ␤-sheet with the altered side chain oriented toward the bound peptide rather than the exterior of the molecule (Fig. 7). This surface, conserved across MHC-I molecules by necessity, is exploited for binding by m06, enabling a broad target specificity for m06 (10). This strategy is a recurring theme in the recognition of MHC-I by viral immunoevasins, as exemplified here, and also in the co-crystal structures of MHC-I with other viral proteins (50 -52).
Our identification of the m06 binding site on H2-L d is a first step toward understanding the effects of m06 in the context of the broader immune evasion network of MCMV that includes the m04 and m152 viral proteins that modulate MHC-I. The luminal part of the m152 protein is believed to transiently associate with MHC-I (3) and has been crystallized in association with one of its target proteins, Rae1␥, which is bound across the top of the ␣1/␣2 helices of the Rae1 MHC-I-like protein (67). Therefore, the interaction of m152 with MHC-I probably focuses on the opposite face of the peptide-binding platform relative to m06, consistent with the view that each protein binds a distinct site of MHC-I. Furthermore, despite m06 and m04 being closely related members of the same m02-m16 structural family, their behavior with regard to MHC-I binding may well be different. Whereas the luminal domain of m06 forms a tight complex with mini-H2-L d (K d in the low micromolar range by NMR, based on our observation of a tight complex with no detectable exchange), m04 shows only a weak affinity with H2-D d MHC-I (K d ϳ0.5-1 mM by SPR (11)). This result can be explained by the use of a full-length MHC-I molecule to probe m04 binding in our previous study and is in principle consistent with the binding site being partially occluded by ␤ 2 m in the fully assembled molecule. However, it has also been shown that the transmembrane domains of m04 and MHC-I contribute to their interaction (46), and the binding could thereby remain undetected in the recombinantly expressed luminal protein constructs used here and in previous m04 studies (11,12). Posttranslational modifications present in the naturally occurring molecules, not preserved in the bacterial molecules that we have studied, might also contribute to the interaction. The importance of the m04 transmembrane domain is further supported by the low sequence conservation of the m04 luminal domain among different viral isolates relative to m06 (64). The observation of a measurable interaction between m06 and MHC-I using just the luminal domains of the two molecules lacking the transmembrane domains suggests a distinct mode of engagement between m06 and m04. Taken together, these observations illustrate the diverse range of MHC-I binding sites and engagement strategies employed by MCMV to interfere with normal antigen processing of the cell, as exemplified by the m152, m06, and m04 paradigms.  . Model for the co-assembly of MHC-I heavy chain, ␤ 2 m, and m06 in the cell. MHC-I heavy chain is shown in blue, with the ␣1␣2 and ␣3 domains labeled. ␤ 2 m is shown in green, and m06 is shown in red. In the absence of m06, class I heavy chain associates normally with ␤ 2 m and peptide in the endoplasmic reticulum to form a properly conformed peptide⅐MHC-I complex for cell surface expression. m06 may exploit structural plasticity within the heavy chain/␤ 2 m interface or interactions via the transmembrane domains, allowing co-assembly with the MHC and resulting in an altered association between ␤ 2 m and heavy chain.