Structures of the MHC-I molecule BF2*1501 disclose the preferred presentation of an H5N1 virus-derived epitope

Lethal infections by strains of the highly-pathogenic avian influenza virus (HPAIV) H5N1 pose serious threats to both the poultry industry and public health worldwide. A lack of confirmed HPAIV epitopes recognized by cytotoxic T lymphocytes (CTLs) has hindered the utilization of CD8+ T-cell–mediated immunity and has precluded the development of effectively diversified epitope-based vaccination approaches. In particular, an HPAIV H5N1 CTL-recognized epitope based on the peptide MHC-I–β2m (pMHC-I) complex has not yet been designed. Here, screening a collection of selected peptides of several HPAIV strains against a specific pathogen-free pMHC-I (pBF2*1501), we identified a highly-conserved HPAIV H5N1 CTL epitope, named HPAIV–PA123–130. We determined the structure of the BF2*1501–PA123–130 complex at 2.1 Å resolution to elucidate the molecular mechanisms of a preferential presentation of the highly-conserved PA123–130 epitope in the chicken B15 lineage. Conformational characteristics of the PA123–130 epitope with a protruding Tyr-7 residue indicated that this epitope has great potential to be recognized by specific TCRs. Moreover, significantly increased numbers of CD8+ T cells specific for the HPAIV–PA123–130 epitope in peptide-immunized chickens indicated that a repertoire of CD8+ T cells can specifically respond to this epitope. We anticipate that the identification and structural characterization of the PA123–130 epitope reported here could enable further studies of CTL immunity against HPAIV H5N1. Such studies may aid in the development of vaccine development strategies using well-conserved internal viral antigens in chickens.

Since the 1997 outbreak in Hong Kong, HPAIV 2 H5N1 viruses have posed a serious threat to both the poultry industry and public health worldwide. Deadly poultry outbreaks in Asian countries have been reported nearly every year since 2003, and human cases with greater than 50% mortality have been confirmed resulting from direct bird-to-human transmission during the past 15 years (1,2). Because of the high mutation rates and continuous circulation of influenza viruses in avian species, the emergence of new H5N1 virus strains adapted for human transmission is of great concern, as such viruses will eventually cause human pandemics (3). The current main strategy for the control of H5N1 viral infections is vaccination with inactivated viral vaccines (4), which can definitely induce humoral immunity against the highly-mutational hemagglutinin (HA) and neuraminidase (NA) proteins. However, this approach cannot provide long-lasting protection and often fails to prevent infection by new viral strains. Therefore, to cut off the transmission of poultry to humans, it is necessary to further elucidate the immune systems of poultry, such as chicken, and to develop safer and effective vaccines based on T/B cell peptide epitopes.
Recent studies have begun to address the importance of major histocompatibility complex class-I (MHC-I)-mediated CD8 ϩ T-cell immunity against influenza virus infection (5,6). For effective CD8 ϩ T-cell recognition, virus-derived CTL epitopes need to be presented by appropriate MHC-I molecules (7). MHC-I molecules are known to be highly polymorphic and are encoded by different loci, for example, three functional loci in humans and mice (8,9). MHC-I polymorphism provides diverse alleles that bind and present numerous peptides derived from the many pathogens encountered (10). To understand the mechanism underlying epitope presentation by MHC-I molecules, crystal structures of peptide-loading MHC-I and ␤ 2 -microglobulin (␤2m) complexes (pMHC-I) from different species have been determined over the years (11)(12)(13)(14)(15)(16)(17)(18)(19). These reports show that antigen-derived peptides are located in the peptidebinding groove (PBG) of pMHC-I molecules. PBG normally contains six pockets (A-F), among which the B and F pockets are anchor amino acid-binding sites and determine the peptide-binding motif of a given MHC-I molecule. Conformational features of the bound peptide have been demonstrated to influence the diversity of the responding T-cell receptor (TCR) repertoire (20), although the correlation between TCR diversity and the magnitude of the CD8 ϩ T-cell response apparently varies among different disease models (21)(22)(23). Moreover, it has been shown that a diverse array of MHC-I molecules can present well-conserved internal epitopes derived from influenza viruses, with different presentation and immunodominant features (24 -26). Therefore, understanding the unique structural features of distinct MHC-I haplotypes and identifying the peptide targets presented by each of these MHC-I molecules would provide useful information for the rational design of T-cell epitopes (5). A few studies have been carried out to identify influenza virus-derived T-cell epitopes in human and mouse models with defined MHC-I haplotypes (27,28). In this context, most of these epitopes were initially selected according to the binding motifs of the respective MHC-I molecules, and over the years, some of these epitopes have been determined to be immunodominant CTL epitopes involved in CD8 ϩ T-cell responses (22,29). In a well-established human study, CD8 ϩ T cells from individuals expressing the MHC-I haplotype HLA-A2 predominantly recognized the conserved influenza virus matrix 1 (M1) 58 -66 epitope (26,28,30). With over a billion people worldwide carrying the HLA-A2 haplotype, the M1 58 -66 epitope is considered a good candidate for vaccination (26). However, because of the restriction between the MHC-I and TCRs, these CTL epitopes cannot be used across species, only in humans and mice.
Chickens express only one dominant classical MHC-I locus (also known as BF2) (31)(32)(33). The "minimal essential MHC" of the chicken immune system has been demonstrated to have a close correlation with susceptibility and resistance to viruses, such as Marek's disease virus (MDV) and Rous sarcoma virus (16,34). Crystal structures of the BF2*2101 (B21 haplotype) and BF2*0401 (B4 haplotype) molecules have been studied to explain the mechanisms underlying the resistance and susceptibility of BF2 to MDV. The unique conformation of BF2*2101, with a wide PBG, allows the presentation of a wide range of peptides, which is believed to induce a wide range of CD8 ϩ T-cell responses to MDV (16). In contrast, the BF2*0401 structure exhibits a narrow PBG with highly and positively-charged pockets, resulting in rigidly-restricted peptide presentation (35). With a small repertoire of CD8 ϩ T-cell clones being activated, B4 chickens tend to be susceptible to viral infection, as observed for B19 chickens (36). In addition to crystal structures, the surface expression of different BF2 haplotypes has been studied to determine the correlation of the haplotype with peptide presentation and viral disease resistance or susceptibility (33,37,38). In contrast to research on other hosts, studies to identify HPAIV CTL epitopes in chickens are not only limited but also very preliminary. By in vivo depletion of T-cell subsets from immunized chickens, Seo et al. (39) demonstrated the protective immunity of CD8 ϩ T-cell subsets against lethal infection by the H5N1 virus. Moreover, studies on cross-protection, in which birds immunized with the H9N2 virus were protected from deadly H5N1 viral infection, also indicated the great potential of conserved internal viral antigens being presented and recognized by CD8 ϩ T cells (40). Based on previous work, several studies have been carried out to identify H5N1 virus-derived CTL epitopes with certain MHC-I haplotypes. A total of four epitopes were identified, including three nucleoprotein (NP)-derived epitopes (NP 67-74 , NP [89][90][91][92][93][94][95][96][97] ) and an HA-derived H5 246 -260 epitope (41)(42)(43). The latter was demonstrated to be presented by the MHC-I molecule of the B19 haplotype (43). Given the research progress described above, it is necessary to explore both the molecular features of different BF2 alleles and the immunogenic epitopes of certain highly-infectious pathogens, such as the HPAIV H5N1 strains. B15 chickens have been bred to obtain SPF animals, which can be used as laboratory animals for infection with various viruses, bacteria, and parasites. As a high-expression haplotype, B15 has been shown to be more stable on the cell surface than the low-expression haplotype B21 (44). In vitro analysis showed that the repertoire of peptides bound to the BF2*1501 molecule is actually larger than the previously-observed peptide repertoire on the cell surface of B15 chickens (34). TAP restriction is the main cause of the narrow peptide repertoire for the BF2*1501 molecule (44). Moreover, increasing BF2*1501 expression in B15/B21 heterozygous animals was observed in the study, indicating the possibility of more epitopes being received by the BF2*1501 molecule in heterozygotes than in homozygotes (44).
In this study, based on the BF2*1501 allele from the SPF strain, the CTL epitopes of HPAIV H5N1 strains were fully screened and identified. After that, a conserved CTL epitope (PA 123-130 ) derived from the polymerase acidic (PA) protein was elucidated. The PA 123-130 epitope-specific CD8 ϩ Tcell immunity was confirmed using the tetrameric peptide BF2*1501. Moreover, the crystal structure of the peptide BF2*1501 (pBF2*1501) was further determined to show a medium-wide PBG compared with the known BF2 complexes. The size of PBG in BF2*1501 is in agreement with the previously reported number of MDV peptides predicted to bind to the chicken B15 haplotype, which is less than the number bound to the B21 haplotype but more than the number bound to the B4 haplotype (16). Most importantly, the pBF2*1501 structure not only explains the previously-determined peptide-binding motif (XRXXXX(X,XX)Y) of the chicken B15 haplotype, but also clearly illustrates the preferential usage of basic amino acids Arg and Lys at P N of the bound peptides. Common PA 123-130 epitope-specific CD8 ϩ T-cell immunity was confirmed using the tetrameric peptide BF2*1501. With both structural elucidation and design and in vivo experiments, our results provide strong evidence for the newly-identified conserved epitope being a good target for anti-HPAIV CD8 ϩ T-cell immunity.

PA 123-130 peptide shows the highest conservation among HPAIV H5N1 strains
Because the peptide-binding motif of the MHC-I allele of the chicken B15 haplotype has been previously defined (34), an epitope-peptide map with the XRXXXX(X,XX)Y motif was proposed by screening the whole-protein sequences of the  Table 1, the seven selected peptides were derived from five segments of HPAIV: HA (one peptide); NP (one peptide); PA (two peptides); PB1 (polymerase basic 1, two peptides); and M2 (matrix 2, one peptide). To analyze the conservation of these peptides, alignment was performed for 203 H5N1 strains (Table S1) reported in China during 1996 -2016. As shown in Table 2, the residue Arg-2 of the selected peptides was present in Ͼ90% of the strains, except in peptides derived from the HA segment, and the residue Tyr-8/9 of the selected peptides was present in Ͼ98% of the strains. For each of the top seven selected T-cell epitope candidates, 72.5-100% of the aligned strains fit the XRXXXX(X,XX)Y motif, suggesting the highlyconserved properties of these peptides. Especially for the NPand PA-derived epitope candidates (NP 211-218 and PA 123-130 ), all aligned H5N1 strains had the anchor amino acids Arg-2 and Tyr-8 at their respective positions, indicating the potentially broad protection provided by these two epitope candidates once their immunogenicity was confirmed.

PA 123-130 stimulates epitope-specific CD8 T-cell responses in B15 SPF chickens
Considering the exceptionally high conservation of the NP 211-218 and PA 123-130 peptides, preliminary experiments designed to first activate chicken immune responses with inactivated H5N1 vaccine (attenuated H5N1 vaccines are unavail-able) and then provoke peptide-specific CD8 ϩ T-cell immunity with epitope candidates NP 211-218 and PA 123-130 were carried out to investigate whether these two peptides can stimulate epitope-specific CD8 ϩ T-cell responses in vivo (Fig. S1A). Tetrameric pBF2*1501 complexes were constructed for the NP 211-218 and PA 123-130 peptides to detect peptide-specific CD8 ϩ T cells in immunized animals. Data obtained from the initial experiment showed that the NP 211-218 peptide-specific CD8 ϩ T cells were not detectable in animals immunized with the NP 211-218 peptide, whereas small percentages of PA 123-130 peptide-specific CD8 ϩ T cells appeared in animals immunized with the PA 123-130 peptide (Fig. S1B). These results potentially indicate that the PA 123-130 peptide is more capable of stimulating antigen-specific CD8 ϩ T-cell responses than the NP 211-218 peptide.
To further investigate the ability of in vivo presentation and CD8 ϩ T-cell response to the PA 123-130 peptide, immunization was then conducted in B15 chickens with either the PA 123-130 peptide or inactivated H5N1 vaccine (Fig. 1A). Flow cytometric analysis showed that the percentage of CD8 ϩ T cells that were positive for the PA 123-130 peptide (i.e. CD8 ϩ Tet ϩ T cells) reached 4.5-5.8% for peptide-immunized animals and 0.96 -2.2% for inactivated vaccine-immunized animals (Fig. 1B). The appearance of CD8 ϩ T cells specific for the PA 123-130 peptide in vaccine-immunized animals was considered to be a result of the boost in immunization caused by the PA 123-130 peptide. However, the proportions of PA 123-130 peptide-specific CD8 ϩ T  Chicken BF2*1501 preferentially presents an HPAIV epitope cells were lower in vaccine-immunized animals than in peptide-immunized animals, indicating that cross-presentation of the PA 123-130 peptide could be lacking during the vaccine immunization, and therefore peptide-specific CD8 ϩ T cells in inactivated vaccine immunized animals cannot reach the same level as in peptide-immunized animals. Statistical analysis also confirmed the significant difference between PA 123-130 peptide immunization and inactivated vaccine immunization (Fig. 1C). The percentage of CD8 ϩ Tet ϩ T cells was shown to be significantly higher in the peptide-immu-nized group than in both the control group (p Ͻ 0.01) and vaccine-immunized group (p Ͻ 0.01). However, no significant difference was observed between the vaccine-immunized group and the control group. The reason for this result might be the variation observed in vaccine-immunized animals. Unlike the fairly-consistent proportion (4.5-5.8%) of CD8 ϩ Tet ϩ T cells detected in the PA 123-130 peptide-immunized animals, one animal in the vaccine-immunized group exhibited a much lower percentage (0.96%) of CD8 ϩ Tet ϩ T cells than the other two animals (2.15 and 2.23%). Nevertheless, these data indi-  [123][124][125][126][127][128][129][130] . B, PBMC were isolated from each of the nine chickens (three chickens in each group) and stained with both the PE-labeled pBF2*1501 Tet that incorporated the PA 123-130 peptide and FITC-labeled anti-CD8 mAb. Flow cytometric analysis was conducted to detect CD8 ϩ T cells that were specific for the PA 123-130 peptide (i.e. CD8 ϩ Tet ϩ T cells). Data are shown as pseudocolor plots. The percentages of CD8 ϩ Tet ϩ T cells (in rectangular gates) in CD8 ϩ T cell populations were determined to be 4.5-5.8% for the peptide-immunized animals and 0.96 -2.2% for inactivated vaccine-immunized animals. C, statistical analysis confirmed that the percentages of PA 123-130 epitope-specific CD8 ϩ T cells were significantly higher in the peptide-immunized group (red circles) than in the vaccine-immunized group (blue circles) (p Ͻ 0.01) and the control group (black circles) (p Ͻ 0.01). No significant difference was observed between the vaccine-immunized group and the control group. Data are shown as the mean Ϯ S.E. *, p Ͻ 0.05; **, p Ͻ 0.01 by unpaired Student's t test.

Chicken BF2*1501 preferentially presents an HPAIV epitope
cated that the PA 123-130 peptide can be presented by the BF2*1501 molecule in vivo and subsequently recognized by specific CD8 ϩ T cells.

BF2*1501 has a medium-wide groove to present the PA 123-130 peptide
In vivo detection of the BF2*1501 PA 123-130 peptide-specific CD8 ϩ T cells in immunized animals demonstrated that the PA 123-130 peptide is preferentially presented by chicken MHC-I molecule BF2*1501. To illustrate the structural basis for the preferential presentation and potential structural features for TCR recognition, chicken BF2*1501 in complex with the PA 123-130 peptide was crystallized in the P212121 space group with a high resolution of 2.1 Å (Table 3). Structural alignment revealed the root-mean-square deviations (RMSDs) of heavy chain (HC) between BF2*1501 and two other solved chicken BF2-peptide (pBF2) structures (BF2*0401, PDB code 4E0R; BF2*2101, PDB code 3BEW) to be 0.988 and 0.68 Å, respectively, suggesting agreement of the overall structure of pBF2*1501 with other resolved chicken MHC-I structures ( Fig. 2A). However, when the PBG shapes were compared, BF2*1501 showed a mediumwide groove, meaning that the pockets in PBG of BF2*1501 are wider than that of BF2*0401 but narrower than those of BF2*2101, particularly pockets C-E (Fig. 2

, B-F).
Structure-based sequence alignment illustrated that most of the amino acid residues composing the PBGs showed variation among the three chicken BF2 alleles (Fig. 3A). Previous stud- is the observed intensity, and ͗I(hkl)͘ is the average intensity from multiple measurements.
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.

Chicken BF2*1501 preferentially presents an HPAIV epitope
ies have shown that the different side chains of these variable amino acids determine the conformational space of PBGs (16,35). When the particular amino acid residues in pockets C-E of BF2*1501, BF2*0401, and BF2*2101 were compared, the residues from BF2*1501 showed partial similarity to the residues from either BF2*0401 or BF2*2101. Residues Arg-68 and Thr-69 from the ␣1 helix of BF2*1501 have large side chains such as Leu-68 and Asn-69 in BF2*0401, which are replaced by the much smaller residues Gly-68 and Ser-69 in BF2*2101; however, the small residues Gly-152 and Leu-153 from the ␣2 helix are conserved between BF2*1501 and BF2*2101, which are replaced by the much larger residues Arg-152 and Trp-153 in BF2*0401 (Fig. 3, B-D). These residues with half-similarity to BF2*0401 and BF2*2101 resulted in a medium-wide PBG in BF2*1501 to present the PA 123-130 peptide.

BF2*1501 has deep B and large F pockets for peptide anchoring
Previous studies have demonstrated that the B pockets in BF2*0401 and BF2*2101 are relatively shallow when compared with those in mammalian MHC-I molecules (16,35), and the peptide residue at position 2 (P-2) interacts with residues 62 and 69 from the ␣1 helix but does not extend under it (Fig. 4, B and C). In contrast, pocket B in BF2*1501 is rather deep, extending under the ␣1 helix, to enable the large side chain of peptide residue Arg-2 to form hydrogen bonds with residue 34 (Thr-34) of BF2*1501 (Fig. 4A). Comparisons of the particular residues composing pocket B of the three chicken BF2 molecules revealed that the deep pocket B in BF2*1501 was due to the small residues Thr-34 and Ala-43, which are replaced by the large residues Met-34 in BF2*2101 and Tyr-43 in BF2*0401(Figs. 3A and 4, B and C).
Pocket F of BF2*1501 is not only deep but wider than those of BF2*0401 and BF2*2101, so that the large peptide residue Tyr-8 with an aromatic ring can be accommodated and stabilized through interaction with residue 113 (Asp-113) of BF2*1501 (Fig. 4A). Although much smaller residues (Ser-113 and Ala-113) appeared at the same position, the F pockets of BF2*0401 and BF2*2101 cannot reach residue 113 due to blockage by the large residue Trp-95 (Fig. 4, B and C). In contrast, in BF2*1501, the large residue Trp-95 is replaced with a much smaller residue Leu-95 with its side chain shifting away from pocket F, resulting in a wider pocket than those in BF2*0401 and BF2*2101 (Fig. 4A).
As essential sites for peptide anchoring, the deep B pocket and the wide F pocket of BF2*1501 have a clear preference to accommodate peptide residues with long (for B pocket) and large (for F pocket) side chains. Moreover, because pocket B has an entirely negatively-charged surface, basic amino acids, such as Arg and Lys, would anchor to this site more easily. For pocket

Chicken BF2*1501 preferentially presents an HPAIV epitope
F, due to the spacious size and the presence of Asp-113 for the formation of hydrogen bonds, amino acids with aromatic rings (Tyr, Phe, and Trp) would be the most suitable choice to anchor in this pocket. These structural features of pockets B and F of BF2*1501 strongly agree with the early report of the peptidebinding motif of the chicken B15 haplotype, where Arg-2 and Tyr-8/9 gave the strongest signals during sequencing of the peptide pool for B15 (34).

Basic peptide residue at P N influences selection and stability of epitopes presented by BF2*1501
An interesting observation that has also been reported previously is the preferential usage of the basic residues Arg and Lys at P N of peptides bound to BF2*1501 (34,44). The crystallographic structure of pBF2*1501 in this study provided a detailed explanation for this phenomenon.
First, it is worth noting that only two or three fewer hydrogen bonds and salt bridges (seven in total) formed between the Arg-1 of PA 123-130 and the PBG than between the anchor residues and the PBG (10 for Arg-2 and 9 for Tyr-8), and more van der Waals interactions were observed for the Arg-1 residue than for the Arg-2 and Tyr-8 residues (Table 4), indicating an important role of Arg-1 in maintaining peptide stability in the PBG of BF2*1501.
Second, three salt bridges were formed between the peptide residue Arg-1 and the residue Glu-62 from the ␣1 helix of BF2*1501 (Fig. 5A). Structure-based sequence alignment showed that Glu-62 also exists in BF2*2101 (Fig. 3A). Moreover, in one of the reported pBF2*2101 structures (PDB code 3BEW), the presented peptide also had the basic amino acid Arg-1 at P N (16). However, Arg-1 and Glu-62 in pBF2*2101 did not form any salt-bridge interaction at P N . The main reason for this was the shifting of the Arg-1 side chain to the ␣2 helix in pBF2*2101, and the large side chain of Arg-61 in the ␣1 helix of BF2*2101 promoted this shifting (Fig. 5B). In contrast to pBF2*2101, the small residue Ser-61 was observed in pBF2*1501, providing room for the side chain of Arg-1 to interact with Glu-62 (Fig. 5A). Furthermore, an additional hydrogen bond was formed between the side chain of Arg-1 and the main chain of Tyr-58 in pBF2*1501. This hydrogen bond, together with the three salt bridges formed between Arg-1 and Glu-62, greatly enhanced the stability of the PA 123-130 peptide in the PBG of BF2*1501.
Finally, pocket A of BF2*1501 showed a negatively-charged surface, which preferentially contains basic amino acids and strongly repels acidic amino acids. This was confirmed by in vitro refolding of the selected H5N1-derived peptides with BF2*1501 HC and chicken ␤2m. The PA 123-130 peptide with the basic amino acid Arg-1 at P N showed high refolding efficiency for BF2*1501. In contrast, the peptide P7 (M2 43-51 , DRLFFKCIY) with an acidic amino acid Asp-1 at P N failed to
These structural features illustrated that the basic amino acid at P N also plays an important role in stabilizing the peptides bound to BF2*1501, which should be considered in the structure-based selection of antigenic epitopes.

Peptide epitope residue Tyr-7 is the potential hot spot for TCR interaction
In our preliminary peptide immunization experiment, peptide-specific CD8 ϩ T cells were detected for the PA 123-130 (RREVHTYY) peptide but were undetectable for the NP 211-218 (GRRTRIAY) peptide. Because both peptides bound to BF2*1501 with high efficiency (Fig. 5D), we speculate that the different CD8 ϩ T-cell responses might be induced by discrepant TCR recognition, where the PA 123-130 peptide is preferentially recognized by specific CD8 ϩ T cells.

Structures of pBF2*1501 help to prioritize other epitope candidates for immunogenicity assessment
Based solely on the peptide-binding motif of chicken B15 haplotype, a total of 14 epitope candidates, including octapeptides and nonapeptides, were identified from six protein segments of different avian influenza virus strains belonging to H1N1, H3N2, H5N1, and H7N9 subtypes ( Fig. 7 and Table S2). Of these epitope candidates, seven were found in all strains, although amino acid substitutions occurred among different subtypes. The peptides from HA and NA segments were lessconserved and identified only in certain stains. Comparing the sequences of the identified peptides showed clearly that the peptide derived from NA segment (NA 58 -66 ) has an acidic amino acid at P N , which makes it biochemically similar to the M2 43-51 peptide and would not be preferred by BF2*1501. The peptides derived from PB1 (PB1 209 -216 and PB1 428 -435 ), similar to the NP 211-218 peptide, have small residues Ser-7 and Thr-7 at P-7, which would be unfavorable for TCR recognition. Therefore, the peptides from the PA segment appeared to have the highest priority for immunogenicity assessment in B15 chickens. Among these peptides, the in vivo presentation and CD8 ϩ T-cell recognition of PA 123-130 were confirmed in this study. Future evaluation of immunogenicity and immunodominance status of PA 123-130 would aid greatly in epitope confirmation.

Discussion
The chicken is a very important economic animal. It is necessary to study various vaccines to prevent various viral diseases, especially zoonoses, such as HPAIV H5N1 avian influenza. However, the determination of immunological information such as defined T-cell epitopes and robust structural fea-  lines) were formed between Arg-1 (pink sticks) and PBG residues Tyr-␣7, Tyr-␣156, and Tyr-␣168 (pink lines) in pBF2*2101 (PDB code 3BEW, pink). No salt bridge was formed between Arg-1 and Glu-␣62 (pink sticks) in pBF2*2101. C, the A pocket of BF2*1501 shows a negatively-charged surface. The electrostatic surface potential was generated with PyMOL. Red is electronegative, and gray is neutral. The particular amino acid residues composing the A pocket are shown as lines. D, selected HPAIV H5N1 peptides showed different refolding efficiencies with BF2*1501 and ␤2m in vitro. Peaks 1-3 represent the aggregated H chain, the correctly refolded pBF2*1501 complex, and the extra ␤2m, respectively. The refolding efficiencies are represented by the height of peak 2 for each peptide. A high peak 2 indicates a high efficiency of the peptide in assisting MHC refolding. The height of peak 2 for P7 (M2 [43][44][45][46][47][48][49][50][51] , DRLFFKCIY) is similar to that for the negative control, indicating the failure of MHC-I binding for the peptide P7.

Chicken BF2*1501 preferentially presents an HPAIV epitope
tures of chicken CTL epitope presentation has lagged behind similar research in humans and experimental mice. Based on the SPF chicken B15 lineage, we have identified a highly-conserved HPAIV CTL epitope, PA 123-130 , which is preferentially presented by BF2*1501. Specifically, in vivo presentation and CD8 ϩ T-cell response to the conserved PA 123-130 epitope were demonstrated via epitope-specific CD8 ϩ T-cell detection, and the principles of preferential presentation and TCR recognition

Chicken BF2*1501 preferentially presents an HPAIV epitope
were illustrated through the structural determination of chicken pBF2*1501.
After performing a full screening of all protein segments of HPAI H5N1 viruses using a peptide map with the motif, seven peptides derived from five segments of H5N1, including both octapeptides and nonapeptides, were initially selected. Conservation analysis showed high conservation (100%) of these peptides among 203 avian influenza virus H5N1 strains reported in China from 1996 to 2016. All 203 strains fit the determined XRXXXX(X,XX)Y motif, especially for two peptides (NP 211-218 and PA 123-130 ) derived from the NP and PA proteins. The greatest advantage of the high degree of conservation of these peptides is that if their immunogenicity is confirmed, these peptides will be good candidates for peptide-based vaccine development. Experiments were then designed to illustrate the presence of specific CD8 ϩ T cells for the PA 123-130 peptide but not for the NP 211-218 peptide. Failure of CD8 ϩ T-cell responses to the NP 211-218 peptide has also been reported previously through an interferon ␥ (IFN-␥) detection assay (46). These results strongly suggest that although the two peptides can be presented by the same chicken MHC-I molecule, the PA 123-130 peptide can induce a stronger CD8 ϩ T-cell immune response. In addition, it is worth noting that inactivated H5N1 vaccine was used to prime immune responses to the virus. The main purpose of using inactivated rather than live or attenuated virus is to check whether cross-presentation of the PA 123-130 peptide occurs during the priming stage of immune responses, considering that inactivated vaccines are widely used in field. The observation of inefficient peptide-specific CD8 ϩ T-cell responses after inactivated vaccine immunization might indicate a lack of cross-presentation of the PA 123-130 peptide. Therefore, future studies involving the design of live or attenuated carrier viruses (47,48) to transport the PA 123-130 peptide for natural presentation and recognition would be necessary to ultimately determine immunogenicity and immunodominance status of the highly-conserved PA 123-130 peptide and would also provide a potential vaccination approach to immunize chickens against HPAIV infection.
The biostructures of pBF2*1501 determine both the molecular mechanisms of peptide presentation by the chicken B15 haplotype and the critical peptide conformation required for efficient TCR recognition. In contrast to the wide PBG in BF2*2101 and the narrow PBG in BF2*0401 (16,35), the PBG size of BF2*1501 was determined to be intermediate, with medium-wide pockets C-E to accommodate the epitope peptide (Fig. 2). This PBG was the direct result of particular amino acid residues in the ␣1 and ␣2 helices of BF2*1501, which were somewhat similar to the respective amino acid residues from BF2*0401 and BF2*2101 (Fig. 3). Functionally, the mediumwide PBG of BF2*1501 corresponds to the number of epitope peptides that were predicted to bind the MHC-I molecule of the chicken B15 haplotype. Previous studies of MDV genes predicted that BF2*1501 could predict a higher number of peptides than BF2*0401 and a smaller number than BF2*2101 (16). The same situation was observed in this study for HPAIV H5N1. The number of HPAIV peptides predicted to bind BF2*1501 (Table S2 and Fig. 7) was higher than the number of peptides predicted to bind BF2*0401 and lower than the number of pep-tides predicted to bind BF2*2101 (12). The second functional correlation between pBF2*1501 structures and epitope peptide presentation lies in the preferential binding of certain peptide residues at P N by BF2*1501. As in many reported structures of pMHC-Icomplexes,pBF2*1501showedconformationalrestrictions at pockets B and F, which determine the unique peptidebinding motif (XRXXXX(X,XX)Y) of BF2*1501. However, an additional restriction at the A pocket was observed for BF2*1501, with preferential accommodation of basic amino acids Lys and Arg at the N terminus of bound peptides (34). The structure of pBF2*1501 provided a clear picture of the molecular basis for this phenomenon (Fig. 5). The structural and biochemical features of residues Tyr-58, Ser-61, and Glu-62 from the ␣ helix of BF2*1501 directly resulted in the selection of the basic residue at peptide P N and the formation of additional interactions (one hydrogen bond and three salt bridges) in the A pocket. These interactions contributed greatly to maintaining the stability of the epitope peptides bound to BF2*1501, and substitution of the basic Arg-1 (in PA 123-130 ) by the acidic Asp-1 (in M2 [43][44][45][46][47][48][49][50][51] ) resulted in the failure of peptide refolding of BF2*1501. Whereas the structure of the PBG largely determines the functional presentation of epitope peptides, the conformations of epitope peptides have been demonstrated to greatly influence the outcome of CD8 ϩ T-cell recognition (20,49,50). Among the two conserved HPAIV peptides, PA 123-130 and NP 211-218 , peptide-specific CD8 ϩ T cells were predominantly detected for the PA 123-130 peptide. We speculate that the PA 123-130 peptide might have essential structural features for effective TCR recognition. After obtaining another pBF2*1501 structure incorporating a self-peptide CBP 22-29 , we aligned both structures and identified a protruding Arg-7 in the PA 123-130 peptide for potential TCR recognition (Fig. 6). In contrast, for the self-peptide CBP [22][23][24][25][26][27][28][29] , which cannot induce peptide-specific CD8 ϩ T-cell responses, the small residue Gly-7 was observed. Taken together, the structural features of pBF2*1501 and PA 123-130 might be the most favored epitope candidate that can be presented and recognized in vivo in chickens with the B15 haplotype. Direct evidence of the presentation of the PA 123-130 epitope by BF2*1501 in vivo was provided in this study via peptide immunization and pBF2*1501 tetramer staining. Incomplete Freund's adjuvant (IFA)-emulsified PA 123-130 peptide was used to induce epitope-specific CD8 ϩ T-cell immunity in SPF chickens expressing BF2*1501. Using IFA as an adjuvant during T-cell epitope vaccination has been demonstrated to allow the induction of protective CD8 ϩ T-cell responses (51)(52)(53). In this study, CD8 ϩ T cells specific for the PA 123-130 epitope were successfully detected in peptide-immunized SPF chickens; moreover, the percentages of epitope-specific CD8 ϩ T cells in peptide-immunized animals were significantly higher than those in inactivated vaccine-immunized animals (Fig. 1). These results clearly indicate that the PA 123-130 peptide can be presented by BF2*1501 in vivo. The detection of PA 123-130 -specific CD8 ϩ T cells after peptide immunization confirmed the immunogenicity of the PA 123-130 peptide to induce CD8 ϩ T-cell immunity in vivo.
In conclusion, through functional determination and structural illustration, this study identified an HPAIV H5N1 epitope candidate, PA 123-130 . The peptide immunization experiment Chicken BF2*1501 preferentially presents an HPAIV epitope confirmed the in vivo presentation of PA 123-130 by BF2*1501 and the presence of PA 123-130 -specific CD8 ϩ T cells in chickens with the B15 haplotype. The molecular mechanisms of preferred presentation of the PA 123-130 peptide were illuminated through biostructures of pBF2*1501. Moreover, the potential hot spot of the PA 123-130 peptide for TCR recognition was identified. Given the high conservation of the PA 123-130 peptide in HPAIV H5N1 strains, our study provides necessary information for structural design and CD8 ϩ T-cell immunity both for the further investigation of chicken CD8 ϩ T-cell immunity against HPAIV and for the development of epitope-based vaccination approaches.

Cloning and expression of the B15 class I molecule
The BF2*1501 HC gene (GenBank TM accession number L28958.2, residues 22-292) was cloned as reported previously (56). Briefly, total RNA was isolated from the spleens of SPF chickens using TRIzol reagent and reverse-transcribed to cDNA using oligo(dT) and Moloney murine leukemia virus reverse transcriptase. The ␣1-␣3 domains of the BF2*1501 HC gene were then amplified using the primers 5Ј-CCGGAATTC-GAACTGCATACCCTGCGTTACA-3Ј and 5Ј-GCGCAAGC-TTTTACCATGAGTAGAGGCCGGGCTGGGG-3Ј with the following PCR conditions: 98°C for 5 min; 30 cycles of 94°C for 1 min, 65°C for 1 min, and 72°C for 2 min; and extension at 72°C for 10 min. The amplified PCR products were analyzed by agarose gel electrophoresis. The expected PCR product was purified and sequenced by Life Technologies, Inc. (Beijing, China). The cloned genes were then ligated to the pET-21a(ϩ) vector (Novagen, Darmstadt, Germany) and expressed in the Escherichia coli strain BL21(DE3). The previously constructed pET-21a(ϩ)-Ch␤2m vector was also transformed into the E. coli strain BL21(DE3) for expression of the chicken-␤2m (Ch␤2m) gene. The recombinant proteins of BF2*1501 HC and Ch␤2m were both expressed as inclusion bodies and separately dissolved in 6 M guanidinium chloride buffer to a final concentration of 30 mg/ml.

Tetramer preparation for the detection of peptide-specific CTL immunity
Tetrameric pBF2*1501 complexes that incorporated the HPAIV PA 123-130 peptide were constructed using a previously described method (15). Briefly, a sequence containing a BirA enzymatic biotinylation site was added to the C terminus of the BF2*1501 HC with the forward primer 5Ј-CCGGAATTCGA-ACTGCATACCCTGCGTTACA-3Ј and the reverse primer 5Ј-GCGCAAGCTTTTAACGATGATTCCACACCATTTTCT-GTGCATCCAGAATATGATGCAGGCTGCCCCATGAGT-AGAGGCCGGGCTG-3Ј. The resulting construct was cloned into the pET-21a(ϩ) plasmid (Novagen, Darmstadt, Germany) and transfected into competent E. coli BL21(DE3) cells for protein expression. The purified recombinant BF2*1501 HC containing the BirA site and Ch␤2m was refolded with the PA 123-130 peptide as described above. The pBF2*1501 complexes were then purified and biotinylated using the BirA enzyme (Avidity, Aurora, CO). Subsequently, pBF2*1501 tetramers were prepared by mixing the biotinylated pBF2*1501 complexes and PE-labeled streptavidin (Sigma) at a molar ratio of 4:1, after which, the samples were separated using 100-kDa Millipore tubes. SDS-PAGE was used to determine the efficiency of tetramerization.

Identification of the PA 123-130 peptide as a CTL epitope of HPAIV H5N1
A total of nine 3-week-old SPF chickens expressing the B15 MHC-I molecule were provided by Merial (now Boehringer-Ingelheim) in Beijing and divided into three groups: the inactivated vaccine-immunized group; the peptide-immunized group; and the control group (three chickens in each group). Animals in the vaccine-immunized group were initially immunized with inactivated H5N1 vaccine according to the manufacturer's instructions (Harbin Pharmaceutical Group Biovaccine Co., Ltd.; H5N1 Re-4ϩRe-5). After 7 days, the immunized animals were subcutaneously injected with the IFA-emulsified (1:3 emulsification) PA 123-130 peptide (0.5 mg/kg of body weight). For the peptide-immunized group, animals were immunized (subcutaneous injection) twice with the IFA-emulsified PA 123-130 peptide (0.5 mg/kg of bodyweight) with a 7-day interval. Animals in the control group were not treated during the experimental period. PBMC of immunized and control animals were isolated using Ficoll-Paque according to the manufacturer's instructions (Haoyang Biological Manufacture Co., Tianjin, China) and incubated for 30 min at 4°C in FACS buffer (PBS with 0.1% BSA and 0.1% sodium azide) containing the PE-labeled tetrameric pBF2*1501 complex and an FITC-labeled anti-CD8 mAb (SouthernBiotech). The PBMC were then washed twice with FACS buffer and detected via flow cytometry. More than 10 6 cell events were acquired for each sample. Cells positive for PE-labeled pBF2*1501 tetramers and FITClabeled anti-CD8 mAb were counted as epitope-responsive CD8 ϩ T cells.

Assembly of the peptide-loading BF2*1501 complex
pBF2*1501 complexes were assembled using the gradual dilution method as described previously (35). The BF2*1501 HC and Ch␤2m inclusion bodies and the synthetically-pre-Chicken BF2*1501 preferentially presents an HPAIV epitope pared HPAIV H5N1-derived peptides were refolded in a 1:1:1 molar ratio. Additionally, the BF2*1501 HC and Ch␤2m inclusion bodies were refolded without peptide as a negative control and refolded with the cCBP-RRA peptide as a positive control. After incubation for 48 h at 4°C, the remaining soluble portion of the complex was concentrated and then purified by chromatography on a Superdex 200 16/60 HiLoad size-exclusion column 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.

Crystallization of pBF2*1501 and data collection
The purified complexes (45 kDa) of pBF2*1501 with HPAIVderived peptides and the positive-control peptide were dialyzed against crystallization buffer (20 mM Tris-HCl (pH 8.0) and 50 mM NaCl) and concentrated to 12 mg/ml. Crystallization trials were set up with Index (Hampton Research, Riverside, CA) at 18°C using the hanging drop method. Two drops containing equal volumes (2 l each) of each protein solution (at 6 and 12 mg/ml) and reservoir crystallization buffer were placed over a well containing 400 l of reservoir solution using a VDX plate (HR3-142; Hampton Research). Crystals of the BF2*1501-PA 123-130 complex were obtained under optimized conditions (0.1 M BisTris (pH 6.5), 0.2 M sodium chloride, and 1.5 M ammonium sulfate at 25°C). The crystals were soaked for 20 -30 s in reservoir solution supplemented with 20% glycerol as a cryoprotectant and then flash-cooled directly in liquid nitrogen. Complete data sets were collected to a resolution of 2.1 Å for the BF2*1501-PA 123-130 complex at the Shanghai Synchrotron Radiation Facility (SSRF) using beamline BL17U at a wavelength of 1.5418 Å (Shanghai, China). Data were indexed, integrated, corrected for absorption, scaled, and merged using the HKL2000 package (57).

Structure determination and refinement of pBF2*1501
The structure of pBF2*1501 with the HPAIV PA 123-130 peptide was solved via molecular replacement using the MOLREP program with BF2*0401 (PDB code 4E0R) as the search model (58,59). Extensive model building was performed by hand with COOT, and restrained refinement was performed with REF-MAC5 (60,61). Additional rounds of refinement were conducted using the phenix.refine program implemented in the PHENIX package (62) with isotropic atomic displacement parameter refinement and bulk solvent modeling. The stereochemical quality of the final model was assessed with the PRO-CHECK program (63). Data collection and refinement statistics are listed in Table 3.

Structural analysis and generation of figures
Residues that were in contact with peptides were identified using the program CONTACT and defined as residues containing an atom that was within 3.3 Å of the target partner (64). Illustrations of the structures and electrostatic potential surfaces were generated using the PyMOL molecular graphics system (DeLano Scientific, RRID:SCR_000305). Structurebased sequence alignment was performed using ESPript 3.0 (RRID:SCR_006587).

Statistical analysis
The FACS data are presented as the mean Ϯ S.E. for the three animals in each group. Statistical analysis was performed using GraphPad Prism 7 (RRID:SCR_002798) for Windows. Significant differences (p Ͻ 0.01) between means were assessed by a two-tailed Student's t test.
The animal trials in this study were performed according to the Chinese Regulations for Laboratory Animals-The Guidelines for the Care of Laboratory Animals (Ministry of Science and Technology of People's Republic of China) and Laboratory Animal Requirements for Environment and Housing Facilities (GB14925-2010, National Laboratory Animal Standardization Technical Committee). The license number associated with this research protocol is CAU20140305-2, which was approved by The Laboratory Animal Ethics Committee of China Agricultural University. The protocol adhered to the recommendations in the Institute for Laboratory Animals Research's "Guide for the Care and Use of Laboratory Animals."

Data availability statement
The coordinates and structural characteristics of pBF2*1501 have been deposited in the Protein Data Bank under accession numbers 6IRL and 6KX9.