Structural Definition of Duck Major Histocompatibility Complex Class I Molecules That Might Explain Efficient Cytotoxic T Lymphocyte Immunity to Influenza A Virus

ABSTRACT A single dominantly expressed allele of major histocompatibility complex class I (MHC I) may be responsible for the duck's high tolerance to highly pathogenic influenza A virus (HP-IAV) compared to the chicken's lower tolerance. In this study, the crystal structures of duck MHC I (Anpl-UAA*01) and duck β2-microglobulin (β2m) with two peptides from the H5N1 strains were determined. Two remarkable features were found to distinguish the Anpl-UAA*01 complex from other known MHC I structures. A disulfide bond formed by Cys95 and Cys112 and connecting the β5 and β6 sheets at the bottom of peptide binding groove (PBG) in Anpl-UAA*01 complex, which can enhance IAV peptide binding, was identified. Moreover, the interface area between duck MHC I and β2m was found to be larger than in other species. In addition, the two IAV peptides that display distinctive conformations in the PBG, B, and F pockets act as the primary anchor sites. Thirty-one IAV peptides were used to verify the peptide binding motif of Anpl-UAA*01, and the results confirmed that the peptide binding motif is similar to that of HLA-A*0201. Based on this motif, approximately 600 peptides from the IAV strains were partially verified as the candidate epitope peptides for Anpl-UAA*01, which is a far greater number than those for chicken BF2*2101 and BF2*0401 molecules. Extensive IAV peptide binding should allow for ducks with this Anpl-UAA*01 haplotype to resist IAV infection. IMPORTANCE Ducks are natural reservoirs of influenza A virus (IAV) and are more resistant to the IAV than chickens. Both ducks and chickens express only one dominant MHC I locus providing resistance to the virus. To investigate how MHC I provides IAV resistance, crystal structures of the dominantly expressed duck MHC class I (pAnpl-UAA*01) with two IAV peptides were determined. A disulfide bond was identified in the peptide binding groove that can facilitate Anpl-UAA*01 binding to IAV peptides. Anpl-UAA*01 has a much wider recognition spectrum of IAV epitope peptides than do chickens. The IAV peptides bound by Anpl-UAA*01 display distinctive conformations that can help induce an extensive cytotoxic T lymphocyte (CTL) response. In addition, the interface area between the duck MHC I and β2m is larger than in other species. These results indicate that HP-IAV resistance in ducks is due to extensive CTL responses induced by MHC I.

, they typically do not show serious signs of the disease, even that caused by highly pathogenic IAVs (HP-IAVs), such as the H5N1 strain, which is lethal to chickens (8,9). Since 2002, many Asian lineage H5N1 HP-IAVs have been shown to produce similar symptoms and mortality, although clinical outcomes are also affected by the age and species of the ducks and the strains of IAVs (10,11). Moreover, ducks have more active and robust cellular immune responses than chickens (12). These findings suggest that at least some species of duck show efficient immune responses against IAV infection.
Studies have indicated that major histocompatibility complex class I (MHC I) plays a role in the anti-IAV response (13)(14)(15). MHC I can present viral epitope peptides to specific T cell receptors (TCRs), resulting in the proliferation of cytotoxic T lymphocytes (CTLs) and eventual clearance of the virus from the host (16)(17)(18). CTL responses mediated by MHC I play a significant role in primary IAV infection and provide cross-protection against different IAV strains in chickens, mice, and humans (19)(20)(21)(22)(23). Suppressive subtractive hybridization libraries have been constructed to enrich differentially expressed genes from the lungs of ducks infected with IAVs of high or low pathogenicity compared to mock-infected controls (13). These data showed that the MHC I gene was upregulated in ducks under both conditions, especially in the highly pathogenic group, in which the MHC I gene was increased by more than 1,000-fold (13). Another study showed that there was an increase in MHC I gene expression in two different duck species after injection with a commercial inactivated vaccine (24). These results suggest that duck MHC I may play a role similar to that of the human HLA against infection.
MHC I molecules are encoded by polymorphic alleles from several loci. These loci make up the peptide binding groove (PBG) and contain epitope peptides against  (hkl), where I i (hkl) is the observed intensity and ͗I(hkl)͘ is the average intensity from multiple measurements. c R ϭ ⌺ hkl || F obs | Ϫ k | Fcalc | |⌺ hkl | F obs |, where R free is calculated for a randomly chosen 5% of reflections and R work is calculated for the remaining 95% of reflections used for structure refinement. multiple pathogens. In humans and mice, there are three functional MHC I loci that provide polymorphic alleles for peptides (25,26); however, many birds predominantly express only one MHC class I gene, such as in chickens and ducks (27). Chickens express only one dominant MHC I locus, referred to as BF2, which is adjacent to the TAP gene (28). Although there are five different loci in the duck MHC I genome region (named the UAA to UEA loci), only UAA, which lies adjacent to the polymorphic TAP2 gene, is predominantly expressed for the CTL immune response (29,30). To date, more than 400 structures of the peptide-MHC I-␤2-microglobulin ␤2m) protein complexes (here called pMHC I) of different species have been determined, and most of them are from human and mouse. Structural studies revealed that the epitope peptides are fixed in the PBG of the MHC I heavy chain by six pockets (A to F) (31). Some pockets, typically the B and F pockets, play a critical role to bind the peptides and determine the peptide binding motif of a certain MHC I molecule. In recent years, pMHC I structures of other species have been solved, while in avian species, only chicken MHC I (BF2) was resolved (32)(33)(34). These structures have greatly facilitated the identification of MHC I-restricted CTL epitopes.
Several studies have indicated that "minimal MHC" is related to resistance and the susceptibility of chickens to viruses, such as the Rous sarcoma virus and Marek's disease . pAnpl-UAA*01 chains are shown with distinct colors: for pAnpl-UAA*01-RLI9, H in red and L in olive, and for pAnpl-UAA*01-MVM9, H in marine and L in green). Peptides are shown as spheres using the same colors with their light chains. (C to E) The additional disulfide bond of pAnpl-UAA*01 compared with chicken and human pMHC I complexes. Disulfide bonds are marked with spheres. The additional disulfide bond formed by Cys 95 -Cys 112 of pAnpl-UAA*01 is labeled in a box and amplified to show its position and electronic density map. Both chicken pBF2*2101 (green, Protein Data Bank code 3BEV) and human pHLA-A*0201 (lemon, Protein Data Bank code 3PWN) have just three disulfide bonds distributed in the ␣2 and ␣3 domains and ␤2m chains. virus (MDV) (32,35,36). Chickens expressing BF2*0401 (B4 haplotype) are susceptible to MDV, and chickens expressing BF2*2101 (B21 haplotype) are resistant to MDV. Crystal studies of BF2 structures have illustrated its resistance or susceptibility to MDV (32). The MHC I complex presents viral peptides to CTLs, depending on the six (A to F) pockets in its peptide binding groove (PBG). The MHC I polymorphisms determine the distinct three-dimensional (3D) structure of the MHC I PBG, and each classical MHC I molecule has a specific peptide binding motif. Crystal studies of B21 have shown that it can remodel its pockets to accommodate different peptides based on its wide binding groove and interplay of two charged residues (Arg 9 and Asp 24 ); B21 could bind multiple peptides from MDV and activate extensive T cell repertories to clear MDV (32). In contrast, B4 has a narrow peptide binding groove and strong polar pockets that restrict its peptide binding motif and the presented MDV peptides (33). Therefore, the CTL response to MDV induced by B4 is not strong enough to prevent infection and suggests that resistance is controlled by the type of MHC I and its 3D structure.
MHC I could also play a role in the susceptibility of viruses. For IAV, the higher resistances of some species of duck indicate that their MHC I alleles would present more IAV peptide epitopes than those of chickens (11,14). Our previous data, as well as a recently published study, confirmed the presence of two or more cysteines, which have never been found in humans or other species, in the PBGs of many duck MHC I alleles (30,37). To date, little is known about the 3D structure and peptide presentation of FIG 2 Structure-based amino acid sequence alignment of Anpl-UAA*01 and representatives of other crystallized MHC I molecules, with the secondary structure elements indicated. Black arrows above the alignment indicate ␤-strands; cylinders denote ␣-helices. Green numbers denote residues that form disulfide bonds. Cysteines at positions 95 and 112 are marked by green stars. Conserved residues are highlighted in red. Residues highlighted in green are species-specific amino acids that differ between duck and other animals. Solid blue dots indicate that the residues interact with Anpl-␤2m. The total amino acid (AA) identities between Anpl-UAA*01 and the listed MHC I molecules are given at end of each sequence. duck MHC I. In this study, we determined two crystal structures of the dominantly expressed duck MHC I (Anpl-UAA*01) molecules with two IAV peptides. The crystal structures showed that Anpl-UAA*01 exhibits characteristics consistent with the duck MHC I-peptide-␤2m complex (pMHC I) architecture; however, we also identified a disulfide bond in its PBG. Unlike the chicken MHC I, only B and F pockets of Anpl-UAA*01 are the primary anchor sites for peptide binding. The two pockets, whose compositions are similar to those of the B and F pockets of HLA-A2, could accommodate diverse residues and make Anpl-UAA*01 an extensive peptide binding motif. By screening the entire sequences of IAV strains, approximately 600 nonapeptides fit the identified peptide binding motif of Anpl-UAA*01. Our study demonstrates that Anpl-UAA*01 could present more IAV epitope peptides, suggesting that ducks with this MHC I haplotype may be protected from infection.

RESULTS
The Anpl-UAA*01 structure shows an unexposed disulfide bond in the PBG and a large interface between heavy and light (H and L) chains. Two trimer complexes formed by Anpl-UAA*01, duck ␤2m (Anpl-␤2m), and the MVM9 and RLI9 peptides were crystallized in the P1211 space group at resolutions of 1.71 Å and 2.06 Å, respectively ( Table 1). Both of the complexes displayed a canonical pMHC I structure, in which the peptide was located in the platform formed by the ␣1 and ␣2 domains of Anpl-UAA*01 above the ␣3 domain and light-chain Anpl-␤2m ( Fig. 1A and B). The most remarkable features of the pAnpl-UAA*01 complexes are the two disulfide bonds formed by 4 cysteines in the PBG (Fig. 1C). In other known pMHC I structures, there is only one disulfide bond formed by a pair of cysteines in their PBGs, such as in chicken pBF2*2101 (32) and human pHLA-A*0201 (38) (Fig. 1D and E). These two cysteines have been found only in ducks (Fig. 2). This novel disulfide formed by Cys 95 and Cys 112 of pAnpl-UAA*01 complexes connects the ␤5 and ␤6 sheets at the bottom of the PBG.
The no-peptide refolding product (only Anpl-UAA*01 and Anpl-␤2m) was used as a negative control to judge the refolding efficiencies of different peptides; however, a peak (approximately 300 milli-absorbance units [mAU]) eluted at the position representing the pAnpl-UAA*01. The gel filtration and SDS-PAGE results showed that the peak represented the complex of Anpl-UAA*01 and Anpl-␤2m (Fig. 3), which is different from results of studies on cattle, pig, chicken, and horse pMHC I refolding (33,(39)(40)(41). This result could be due to the Cys 95 -Cys 112 disulfide bond of pAnpl-UAA*01. In order to clarify the impact of this disulfide bond, Cys 95 and Cys 112 were mutated to alanine (A), alone and together. However, the results for refolding of the mutated Anpl-UAA*01s were similar to those for the wild type (Fig. 3), indicating that the additional disulfide bond is not directly related to the stability of Anpl-UAA*01 without peptide binding.
The distinct peptide conformations presented by the Anpl-UAA*01 molecule. The conformations of IAV-MVM9 and IAV-RLI9 presented by Anpl-UAA*01 are distinct, especially in the middle region from the residue at position 3 (P3) to that at P6 (Fig. 5). The IAV-RLI9 peptide is clear on the electronic density map and adopts an "M" overall conformation, which is common to other pMHC I structures. The B factors of the IAV-RLI9 middle part are relatively higher than at the N and C termini, indicating that P4 to P6 are more flexible than other residues (Fig. 5A). In IAV-MVM9, the electronic density map is missing at P3 and P4, which means that these two residues are quite flexible (Fig. 5B). The side chain orientations of P5 are obviously different between the IAV-MVM9 and IAV-RLI9 peptides. In IAV-MVM9, the side chain of P5 stretches to the ␣2 helix but in RLI9, its side chain stretches upward and most parts are exposed from the PBG (Fig. 5C).
The conformations of the MVM9 and RLI9 peptides were significantly distinct when the two pAnpl-UAA*01 structures were superimposed ( Fig. 6A and B); their deviation is focused mainly at the center, where the distance between the C␣ atoms of MVM9 and RLI9 P5 residues can reach 4.4 Å. The N and C termini of the two peptides matched well. Structural analysis showed that the N termini of the MVM9 and RLI9 peptides (P1 and P2 residues) form the same hydrogen bonds with the PBG (Fig. 6C). From the P3 to P7 residues, there are no hydrogen bonds between RLI9 and the PBG. Although the P3 and P4 residues are missing in MVM9, three hydrogen bonds were found between P5, P7, and the PBG (Fig. 6D). These distinct peptide conformations of this part should be related to the differences in hydrogen bonds found here. In the C termini, the hydrogen bonds between the main chain of peptides and the PBG are the same; the only difference is that the Lys 142 of pAnpl-UAA*01 forms 3 hydrogen bonds with the side chains of the P8 and P9 residues in the RLI9 peptide but does not form a bond with the MVM9 peptide (Fig. 6E).
These data suggest that the conformation of the middle section of the peptide presented by Anpl-UAA*01 is flexible and is determined by the peptide's specific amino acid sequence.
Anpl-UAA*01 selects peptides by relying on the B and F pockets. The compositions of the six pockets of Anpl-UAA*01 are shown in Fig. 7, and the interactions between peptides and these pockets are listed in Table 2. The RLI9 peptide was selected to illustrate the interactions with the six pockets in Fig. 6 because of its completeness.
The A pocket of Anpl-UAA*01, composed of Leu 5 , Tyr 7 , Tyr 58 , Glu 62 , Tyr 156 , Thr 160 , Cys 161 , Trp 164 , and Tyr 168 , fixes P1-Arg by hydrogen bonds and strong van der Waals forces (VDWs) (Fig. 7A; Table 2). As in the known pMHC I structures, P1-Arg forms the hydrogen bonds with the A pocket by its main-chain atoms, and its side chain stretches upward out of the A pocket (45,46). Therefore, although the binding between the A pocket and P1 residue is strong, the A pocket dose not play a restrictive role in the PBG of Anpl-UAA*01.
The B pocket is a primary anchor site and plays a restrictive role in peptide binding. The B pocket of Anpl-UAA*01 accommodates P2-Leu (Fig. 7B) The charged Glu 62 , which is on the top of the B pocket, can form a hydrogen bond with the main chain of P2-Leu. The side chain of the P2 residue inserts into the B pocket and is fixed by the VDWs provided by the surrounding residues ( Table 2).
The C, D, and E pockets usually connect with residues in the middle part of the binding peptides. The amino acid compositions of these three pockets of Anpl-UAA*01 are shown in Fig. 7C to E. The C and D pockets can interact with the side chains of the P3 and P6 residues, respectively, but the E pocket can only interact with the main chain of the P7 residue, as the orientation of the P7 side chain faces upward ( Table 2).
The additional disulfide bond formed by Cys 95 and Cys 112 is in the C pocket of Anpl-UAA*01. The impact of this disulfide bond on peptide binding was checked by the refolding of MVM9 and RLI9 peptides with the C95A, C112A, and C95A-C112A double mutant heavy chains. We found that that the mutant H chains can still form a complex with peptides, but the refolding is worse than that of the wild-type H chain, especially for the C112A mutant H chain ( Fig. 8A and B). The reduced refolding efficiencies of the mutants indicated that the disulfide bound in the C pocket helps with the peptide binding of Anpl-UAA*01; however, the distances between the peptide residues and disulfide bond are over 5.0 Å, which means that their direct interactions are negligible (Fig. 6C). Previous studies have suggested that Cys 95 and Cys 112 do not meet the preferred geometry to form the disulfide bond (30), and our pAnpl-UAA*01 structures show that the distance between Cys 95 and Cys 112 is less than that for the corresponding positions in HLA-A2 and B21 (Fig. 8C). To some extent, the additional disulfide bond of Anpl-UAA*01 alters the bottom of the C pocket and strengthens the stability. Improved peptide binding efficiency may strengthen the stability of the C pocket with the additional disulfide bound.
The F pocket is the most important anchor site at the C terminus of PBG. The F pocket of Anpl-UAA*01 is composed of Asn 76 , Thr 79 , Ala 80 , Arg 83 , Tyr 84 , Trp 93 , Phe 119 , and Ala 135 , and it has a strong binding affinity with P9-I (Fig. 7F). The main chain of P9-I can form 2 salt bridges with Arg 83 and Lys 142 and 2 hydrogen bonds with Asn 76 and Thr 139 in the F pocket. The side chain of P9-I inserts into the F pocket and is fixed by the strong hydrophobic forces (Table 2).
In order to determine the primary anchor residues of Anpl-UAA*01-presenting peptides and the vital restriction pockets for peptide binding, the RLI9 peptide was mutated by alanine scanning and circular dichroism (CD) spectral analysis was used to test the stabilities of Anpl-UAA*01 complexes with these mutant peptides (Fig. 9). The wild-type RLI9 peptide was used as a control, and its midpoint transition temperature (T m ) value was 43.3°C. Among all of the alanine mutant peptides, only the T m values of the P2-Ala and P9-Ala mutant peptides were significantly lower than that of the wild-type RLI9 peptide, indicating that the side chains of P2 and P9 play key roles in RLI9 peptide binding and that these two residues are the primary anchor residues. The pockets (B and F) at the two termini of the Anpl-UAA*01 PBG anchor the peptides and determine the peptide binding motif of Anpl-UAA*01.
The peptide binding motif of Anpl-UAA*01 is most similar to that of HLA-A2. The peptide binding manner of pAnpl-UAA*01 is similar to those of some mammalian pMHCs but not similar to that of chicken pMHC I; only the B and F pockets are the key anchor caves, similar to what is observed in human MHC I (also named human lymphocyte antigen [HLA]). By comparison with the solved structures in the Protein Data Bank (PDB), the peptide conformation and binding manner of pAnpl-UAA*01 were found to be most similar to those of HLA-A2 (Fig. 10). In the B pockets of Anpl-UAA*01 and HLA-A2, the side chains of P2-Leu residues have similar orientations and are fixed in a similar way by the residues at the same positions in the two pockets (Fig. 10A). A similar situation was also found in the two F pockets, where the hydrogen bonds fixing P9-I are almost the same (Fig. 10B). The peptide presentations and pocket compositions make the peptide conformations of pAnpl-UAA*01 and pHLA-A2 almost identical ( Fig.  10C and D). Previous studies have shown that the B pocket of HLA-A2 prefers hydrophobic anchor residues, such as I, L, V, A, and F, and can also accommodate the same polar residues, such as T. The most frequently occurring anchor residues in the F pocket of HLA-A2 are I, V, and L. The two peptides presented by Anpl-UAA*01, MVM9 and RLI9, perfectly fit the peptide binding motif of HLA-A2. Eighteen additional peptides from IAV and 11 mutant peptides were used to verify the peptide binding motif of Anpl-UAA*01 (Table 3). The results confirmed that the peptide binding motifs of these two MHC I molecules from different species are mostly overlapping ( Fig. 10E and F Many IAV peptides match the motif of pAnpl-UAA*01. Using the motif of Anpl-UAA*01 that we identified, the proteomes of several IAV strains were screened to identify the epitope peptides that could be presented by Anpl-UAA*01, including the H1N1, H3N2, H3N8, H4N6, H5N1, H7N9, and H9N2 subtypes (47-53). There were Structures of Duck MHC Class I Journal of Virology approximately 600 candidate nonapeptides from different IAV strains matching the peptide binding motif of Anpl-UAA*01 (Fig. 11). Most selected peptides also fit the motif of HLA-A2, and 11 nonapeptides were proven to activate HLA-A2-restricted CTL responses efficiently according to IEDB data (http://www.iedb.org/) (54). The in vitro refolding results for 31 peptides confirmed that the predicted IAV epitopes could be bound by Anpl-UAA*01 (Table 3). Based on the motifs of chicken B4 and B21, the longer peptides from IAV strains were also screened. There are fewer than 20 octapeptides for B4 and 24 to 27 decapeptides for B21 (Fig. 11), which are much less than the numbers of peptides found in Anpl-UAA*01.

DISCUSSION
The most distinguishable feature of Anpl-UAA*01 was the additional disulfide bond in the PBG. The mutant experiments showed that this disulfide bond located in the C pocket can increase the efficiency of peptide binding to pAnpl-UAA*01. Although all of the mutants could still form complexes with the RLI9/MVM9 peptide and ␤2m, their refolding efficiencies were significantly lower than that of the wild type ( Fig. 8A and B). Without the peptide, this disulfide bond did not increase the refolding production of A higher peak indicates a better efficiency of the peptide to help the MHC renature. (C) Results of further stabilization assays of the refolded products tested by anion exchange. A higher peak here also indicates better stability. With peptides P2-A and P9-A, the refolded complex proteins dissociated at the corresponding eluting NaCl concentration (9% to 11%), implying poor stabilities. the complex of Anpl-UAA*01 and ␤2m (Fig. 3). These data indicate that this additional disulfide bond can improve peptide binding to Anpl-UAA*01 and lead to the increased formation of a trimer complex; however, no significant contacts between the presented peptide and disulfide bond were identified. We also discovered that the additional disulfide bond makes the bottom ␤ sheet of the C pocket of Anpl-UAA*01 more compact than in other MHC I molecules. The refolding effect of the C112A mutant indicated that the free side chain of Cys 95 could disturb peptide binding in the C pocket region (Fig. 8A and B). Therefore, we hypothesize that this additional disulfide bond provides a more stable C pocket, which helps with peptide presentation of Anpl-UAA*01.
Another characteristic of the Anpl-UAA*01 structure is the intensive interaction between its H and L chains. In comparison with other known pMHC I structures, pAnpl-UAA*01 has the greatest interface area and many interchain bonds (Fig. 5). The interface area in ducks and chickens is significantly larger than that in mammals. Compared with BF2*2101 (32), pAnpl-UAA*01 had a 100-Å 2 -larger interface area and 4 more hydrogen bonds; this relatively stronger interaction could lead to the formation of complexes containing only Anpl-UAA*01 and ␤2m through in vitro refolding, which has never been identified in pMHC I structures of other species (33,(39)(40)(41).
Anpl-UAA*01 is similar to human HLA in peptide presentation (but not to chicken BF2), although the amino acid identities among Anpl-UAA*01 and BF2 molecules are much higher than those between Anpl-UAA*01 and HLA (Ͼ60% versus Ͻ50%, respectively). There are only two primary anchor pockets in pAnpl-UAA*01, the B and F pockets, which is common in HLA molecules (Fig. 10A and B). Using structural analysis and mutant peptide refolding tests, we found that the peptide binding motif of pAnpl-UAA*01 overlaps with HLA-A*0201 to a large degree. The B pocket of Anpl-UAA*01 can accommodate extensive uncharged residues, and the F pocket can accommodate multiple hydrophobic residues (Fig. 10E). However, BF2 exhibits peptidepresenting strategies that differ from those of HLA. B4 has a narrow and highly charged PBG, which limits the binding peptides that must fit its B, C, and F pockets together (33). B21 has a large central cavity and flexible Arg9, which make its binding motif promiscuous with three anchor residues (32). The moderate limits of the peptide binding motif suggest that ducks can accommodate more peptides from pathogens than chickens.
The role of MHC I mediating the CTL response in clearance of IAV has been confirmed in human and mouse. Although substantial data on anti-IAV CTL responses in duck and chicken are scarce, some evidence indicates that duck MHC genes respond RLIQNSITL ϩϩ a Base value for estimating the binding affinities of peptides with the NetMHCpan 3.0 server (http://www.cbs.dtu.dk/services/NetMHCpan/): rank threshold for strongly binding peptides, 0.100; rank threshold for weakly binding peptides, 1.000. b ϩϩ, peptide binds strongly and can tolerate anion-exchange chromatography; ϩ/Ϫ, peptide binds Anpl-MHC I but cannot tolerate anion-exchange chromatography. to infection with IAV (13,55). The studies showed that duck MHC locus genes (MHC I/II and TAP) are overexpressed after infection with IAV, especially the duck MHC I gene, and under the condition of HP-IAV infection its expression could be increased by about 1,000 times in the lung (13). Unlike mammals, both ducks and chickens have a "minimal MHC" region with a limited set of genes and express only one dominant MHC I allele.
The "minimal MHC" in chickens is critical to defense against a particular pathogen because it is completely dependent on whether or not it can load peptides from that pathogen. The best illustrated example is that chickens of the BF2*2101 genotype could defend against MDV, while chickens of the BF2*0401 genotype could not (32,33,56). The difference in duck versus chicken responses to IAV might also relate to the different peptide loading abilities of duck and chicken MHC I. Approximately 600 nonapeptides from different IAV strains matching the motif of pAnpl-UAA*01 were identified in this study, which is much higher than the numbers of BF2*2101 and BF2*0401 from chicken ( Fig. 11; see Data Set S1 in the supplemental material). Therefore, the duck Anpl-UAA*01 should exhibit stronger resistance to IAV strains than chicken BF2*2101 and BF2*0401 (57,58). In addition, the Anpl-UAA*01 structure showed that it can present IAV peptides in distinct conformations, which is believed to be useful for activating the T cell repertoire and inducing stronger CTL responses. Our study illustrates the structural basis of duck MHC I and provides a novel approach for explaining why ducks are more resistant to HP-IAV.

MATERIALS AND METHODS
Prediction and synthesis of IAV-derived peptides. All nonamer peptides that potentially bound to Anpl-UAA*01 were predicted using the NetMHCpan 3.0 server (http://www.cbs.dtu.dk/services/ NetMHCpan/) (59,60). The 31 peptides used in this study (Table 3) were synthesized and purified to 90% by reverse-phase high-performance liquid chromatography (HPLC) and mass spectrometry (SciLight Biotechnology). These peptides were stored in lyophilized aliquots at Ϫ80°C after synthesis and dissolved in dimethyl sulfoxide (DMSO) before use.
Assembly of the pAnpl-UAA*01 complex. To assemble the pAnpl-UAA*01 complex, the peptide, Anpl-UAA*01, and Anpl-␤2m inclusion bodies were refolded (in a 1:1:1 molar ratio) according to the gradual dilution method that we described previously (39,63). After a 24-h refolding step at 277 K, the remaining soluble portion of the complex was concentrated and purified using a Superdex 200 16/60 column (GE Healthcare), followed by Resource Q anion-exchange chromatography (GE Healthcare). Purified proteins were buffer exchanged with 10 mM Tris-HCl and 50 mM NaCl at a pH of 8.0.
Structure determination and refinement. The crystals of pAnpl-UAA*01-MVM9/RLI9 all belong to the P1211 space group, and their structures were solved by molecular replacement using Molrep and Phaser in the CCP4 package, with the chicken BF2*0401 structure (PDB code 4E0R) as the search model (67)(68)(69). Extensive model building was performed by hand with COOT (70), and restrained refinement was performed using REFMAC5. Additional rounds of refinement were performed using the Phenix refine program implemented in the PHENIX package (71) together with isotropic atomic displacement parameter (ADP) refinement and bulk solvent modeling. The stereochemical quality of the final model was assessed with the PROCHECK program (72). Detailed information about collection and refinement is shown in Table 1.
Structural analysis and generation of illustrations. Peptide-contacting residues were identified using the program CONTACT and were defined as residues containing an atom within 3.3 Å of the target partner (69). Structural illustrations and electron density-related figures were generated using the PyMOL molecular graphics system (http://www.pymol.org/). Solvent-accessible surface areas and the B factor were calculated with CCP4.
CD spectra and thermal unfolding. Circular dichroism (CD) experiments for the pAnpl-UAA*01 with parental RLI9 or mutant peptides were performed on a Jasco J-810 spectropolarimeter equipped with a water-circulating cell holder. The CD spectra were collected at a protein concentration of 8 m⌴ in pH 8.0 Tris buffer (20 mM Tris and 50 mM NaCl), using a 1-mm-optical-path-length cuvette with ellipticity at 218 nm. Thermal denaturation curves were determined as the temperature was raised from 25 to 80°C at a linear rate of 1°C/min. The temperature of the sample solution was directly measured with a thermistor. The fraction of unfolded protein was calculated from the mean residue ellipticity () using the standard method. The unfolded fraction (%) was expressed as ( Ϫ N)/(U Ϫ N), where N and U are the mean residue ellipticity values in the fully folded and fully unfolded states, respectively. The midpoint transition temperature (T m ) was determined by fitting data to the denaturation curves using the Origin 9.1 program (OriginLab) (73).

SUPPLEMENTAL MATERIAL
Supplemental material for this article may be found at https://doi.org/10.1128/JVI .02511-16. The funders had no role in study design, data collection and interpretation, or the decision to submit the work for publication.
We thank the staff at the Shanghai Synchrotron Radiation Facility of China (SSRF) for their technical assistance during data collection.
We declare no conflict of interest.