A Newly Recognized Pairing Mechanism of the α- and β-Chains of the Chicken Peptide–MHC Class II Complex

This article is featured in In This Issue, p.1419

In jawed vertebrates, the genetic region of the MHC class II (MHC-II) has evolved to encode proteins that are critical for the adaptive arm of the immune system (1). The physiological function of classical MHC-II is to bind exogenous peptides and present them on the surfaces of APCs for recognition by CD4⁺ T cells, inducing cytokine secretion for Ab generation and immune system regulation (2, 3).

The structures of classical MHC-II molecules in humans and mice are well characterized (4, 5). In general, heterodimeric MHC-II complexes are formed by one α-chain and one β-chain. Each chain (α and β) consists of two domains, α1 and α2 and β1 and β2 (6). Α1 and β1 assemble to form a peptide-binding groove (PBG), whereas the two membrane-proximal IgC1-like domains, α2 and β2, support the floor of the α1/β1 U (7). Unlike in MHC-I, the PBG in MHC-II is open ended, with increased tolerance for binding peptides ranging from 13 to 25 residues in length (8, 9). Peptides bound to MHC-II adopt an extended polyproline type II conformation (10). The canonical conformation of peptides is maintained by a hydrogen-bonding network that involves binding of the peptide backbone to residues of MHC-II molecules (7). Studies of peptide–MHC-II (pMHC-II) structures have confirmed the role of four specific peptide-binding pockets, P1, P4, P6, and P9, with additional contributions from the P3, P7, and P10 pockets, which control the selection of specific antigenic peptides (11, 12). Furthermore, polymorphic MHC-II residues located in areas forming pockets in pMHC-II structures are key determinants of T epitope immunogenicity (13). Overall, structural determination of pMHC-II complexes is crucial for elucidating the molecular mechanism of the process of peptide loading (14).

The chicken immune system provides an invaluable model for studies on basic immunology in nonmammalian vertebrates (15). The chicken MHC region includes the chicken MHC region B (MHC-B) locus and the chicken MHC region Y (MHC-Y) locus (16). The MHC-B locus, consisting of highly polymorphic B-F, B-L, and B-G regions, is more compact and simple in chickens than in mammals and is also arranged in a different manner (17–19). The highly polymorphic classical MHC-II β-chain is encoded by two closely related BLB genes (BLB1 and BLB2) in the B-L region. A monomorphic α gene (BLA) is located outside of the MHC-B locus and is not linked to nonclassical class II β genes in the MHC-Y (18, 20). The two classical BLB genes exhibit tissue-specific expression rather than dominant expression, and BLB2 is more highly expressed at the RNA level in the spleen than is BLB1 (21). The most striking feature of the chicken MHC system is that the MHC-B is associated with resistance and susceptibility to various infectious diseases, including viruses, bacteria, and parasites (22–24). However, knowledge regarding the pMHC-II complex in chickens and other nonmammalian vertebrates is lacking to date, yet studies in chicken may provide insight into the evolution of the MHC-II system. It is worth noting that there are three classical MHC-II isotypes (DR, DP, and DQ) in humans and two isotypes (I-A and I-E) in mice (25). Both the α- and β-chains of DP, DQ, and I-A are polymorphic. In contrast, the β-chains in I-E and DR molecules are polymorphic, but the α-chains are monomorphic (26). In chickens, there are two closely related and highly polymorphic BLB chains and one monomorphic BLA chain (20). Clearly, chicken BLA is a nonmammalian homolog of human HLA-DRα. However, the reason why only one α-chain is maintained in chicken and the mechanism by which one monomorphic α-chain combines with polymorphic β-chains to form a functional heterodimer remain unknown.

To elucidate the pairing mechanism of the monomorphic α-chain and polymorphic β-chains and the peptide-binding mode of chicken MHC-II, in this study, the crystal structure of chicken pMHC-II (pBL2*019:01) in complex with a 17-residue peptide was the first solved. Moreover, the pairing mechanism of one α-chain with several β-chains was elucidated. Additionally, comparison of the pMHC-II structures among chicken, human, and mouse provides evolutionary information regarding the ancestral MHC-II system in jawed vertebrates.

A gene encoding the extracellular residues 1Leu-184Glu of the chicken MHC-II α-chain, namely, BLA (GenBank accession no. AY357253), was synthesized and inserted into the expression vector pET-21a (Novagen), followed by transformation into Escherichia coli strain BL21 (DE3) (TransGen Biotech, Beijing, China) (20). When the OD₆₀₀ of the bacterial culture reached 0.6 at 37°C, 0.5 mM isopropyl β-d-thiogalactoside was added to induce expression of BLA for 6 h. The bacteria were harvested by centrifugation at 6000 × g for 10 min and resuspended in cold PBS. After sonication, the sample was centrifuged at 16,000 × g, and the pellet containing inclusion bodies was washed three times with a solution consisting of 20 mM Tris-HCl (pH 8.0), 100 mM NaCl, 1 mM EDTA, 1 mM DTT, and 0.5% Triton X-100. The inclusion bodies were then dissolved to 30 mg/ml in guanidinium chloride (Gua-HCl) buffer (27). A construct for the chicken MHC-II β-chain, namely, BLB2 (GenBank accession no. DQ008584), which encodes a fusion protein comprising ribosomal protein RPL30 (PGDSDIIRSMPEQTSEK), connected by a 16-mer peptide poly-Gly/Ser linker (SGGGSLVPRGSGGGGS) to residues 1Thr-190Pro of the BLB2 chain, was cloned into the expression vector pET-21a and then transformed into E. coli strain BL21 (DE3) (28–30). The endogenous peptide RPL30 originating from the ribosomal protein was identified following peptide elution from BL2*019:01 (29, 30). RPL30–BLB2 was expressed in inclusion bodies and purified as described above for BLA.

For in vitro refolding, purified BLA and RPL30–BLB2 inclusion bodies were diluted to a final concentration of 40 mg/l each in a refolding solution (containing 50 mM Tris-HCl, 20% [w/v] glycerol, 0.5 mM EDTA, 3 mM reduced glutathione, and 0.9 mM oxidized glutathione [pH 8.0]) (28). After 8 d at 4°C, the folding solution was concentrated and purified by chromatography using a Superdex 200 16/60 column (GE Healthcare), followed by RESOURCE Q anion-exchange chromatography (GE Healthcare) as previously described (27).

Purified pBL2*019:01 was concentrated to 8 mg/ml in a buffer containing 20 mM Tris-HCl (pH 8.0) and 50 mM NaCl for crystallization. After mixing with the reservoir buffer at a 1:1 ratio, the concentrated pBL2*019:01 complex was crystallized according to the sitting-drop vapor diffusion method at 18°C. After 7 d, pBL2*019:01 crystals were obtained with solution no. 17 from the Crystal Screen 1 Kit (100 mM Tris base/hydrochloric acid [pH 8.5], and 200 mM lithium sulfate) (Hampton Research). The 1.9-Å diffraction data for the pBL2*019:01 crystal were collected at 100 K; the data collection was performed at the Shanghai Synchrotron Radiation Facility using beamline BL17U at a wavelength of 1.5418 Å (Shanghai, China) (31). The collected intensities were indexed, integrated, corrected for absorption, scaled, and merged using the HKL2000 package (32).

The crystal of pBL2*019:01 belongs to the C2 space group, and the structure was solved by molecular replacement using MolRep and Phaser in the CCP4 package, with the human HLA-DR1 structure (Protein Data Bank [PDB] code: 1AQD) as the search model (33–35). Extensive model building was performed manually with Coot (36), and restrained refinement was conducted using REFMAC5 (37). Additional rounds of refinement were carried out using the Phenix refine program implemented in the Phenix package together with isotropic atomic displacement parameter refinement and bulk solvent modeling (38). The stereochemical quality of the final model was assessed with the PROCHECK program (39). Detailed information about collection and refinement is shown in Table I.

SignalP 5.0 server was used to predict the presence and location of signal peptide cleavage sites (40). Structural illustrations and electron density–related figures were generated using PyMOL (http://www.pymol.org/) and UCSF Chimera (http://www.cgl.ucsf.edu/chimera/). The isotropic B factor was calculated using the equation B = 8π²μ². Solvent-accessible surface areas were calculated with the Protein Data Bank in Europe Proteins, Interfaces, Structures and Assemblies webpage (http://www.ebi.ac.uk/pdbe/pisa/picite.html), and comparison of amino acid sequences from different proteins was performed using Clustal Omega (http://www.ebi.ac.uk/Tools/msa/clustalo/). The crystal structure has been deposited in PDB (http://www.pdb.org/pdb/home/home.do) under accession number 6KVM.

The crystal structure of pBL2*019:01 was determined at 1.9-Å resolution by molecular replacement. The final refinement of the structure generated R/R_free factors of 18.7/21.6% (Table I). In this structure, only one heterodimer forms by the BLA and BLB2 (BLB2*019:01) chains in the asymmetric unit. The BLA chain is split into α1 and α2 domains, and the BLB2 chain is split into β1 and β2 domains (Fig. 1A). The α1 and β1 domains contribute approximately equal halves to the PBG, and the membrane-proximal α2 and β2 domains consist of two Ig-C1 domains. Superposition of the pBL2*019:01 structure with the solved human and mouse pMHC-II structures results in a root mean-square difference in α-carbon positions of <1.7 Å (Supplemental Fig. 1A, Supplemental Table I). The pBL2*019:01 structure is more similar to the human HLA-DR2 structure than other resolved pMHC-II structures of human and mouse (Supplemental Table I).

A total of 38 hydrogen bonds form between the BLA and BLB2 chains. Among them, 16 hydrogen bonds between the main chains at the central interface occur between the α1 and β1 strands of the PBG (Table II). The remaining 22 hydrogen bonds are distributed between the main chain and side chains or between the side chains of the BLA and BLB2 chains. All residues of the BLB2 chain involved in the formation of the 22 hydrogen bonds are conserved in the chicken BLB sequences (Fig. 2, Table II). However, those amino acids differ from those of human isotypes and other mammalian MHC-II sequences (Supplemental Table II).

Interactions among the four domains (α1, α2, β1, and β2) of pBL2*019:01 were analyzed. Only one hydrogen bond occurs between Asp^α145 of the α1 domain and Arg^β34 of the β2 domain, located in the loop rings (Fig. 1B). pBL2*019:01 shows eight hydrogen bonds between the α2 and β2 domains, which are dispersed (Fig. 1C). The number of hydrogen bonds between the α2 and β2 domains of the human and mouse structures vary from two to six, and those hydrogen bonds are concentrated in the intermediate region (Table II). In short, all residues in the BLB chains, excluding those involved in hydrogen bonds between the main chains at the central interface, interacting with the BLA chain are conserved, regardless of whether the chain is BLB1 or BLB2.

The α2 and β2 domains are the major regions of interaction with CD4 in mammals (41). The key residues that interact with CD4 are identical in pBL2*019:01; Val^β142-Thr^β145, Leu^β158, Glu^β162, Glu^α91, and Arg^α179 are conserved (Fig. 1D). A significant change is Arg^β114, which may not affect natural binding to CD4 because of the mutation of CD4 Tyr⁴⁰, and other positions display changes in residues of the same nature. The chicken CD4 structure was modeled based on the HLA-DR1–CD4 complex structure (PDB code: 3S4S). Because the chicken CD4 molecule has a five-residue deletion at the corresponding position (positions 55–59) of the human CD4 molecule, the distance between this region of the chicken CD4 molecule and pBL2*019:01 is changed, suggesting that interaction between pBL2*019:01 and chicken CD4 is different from the corresponding interaction in known pMHC-II complexes.

The bottom of the PBG in pBL2*019:01 is composed of eight antiparallel strands, with the α1 domain and β1 domain each contributing four strands (Fig. 3A). The four antiparallel strands of the α1 domain and β1 domain were termed α1–α4 strands and β1–β4 strands, respectively. Based on superposition of pBL2*019:01 with resolved pMHC-II structures, it was found that the α1 strand, α2 strand, and β1 strand in the PBG are the longest (Fig. 3A, Supplemental Fig. 1). One reason for this finding is that, compared with other α-chains, 4 aa (GPDK) are inserted in the α1 domain of BL2*019:01 at positions 16–19 in the bottom of the PBG, which lengthens the α1 strand (Fig. 3B). The lengthened strands are accompanied by an increase in the number of hydrogen bonds in the main chains. In pBL2*019:01, there are 16 hydrogen bonds between the main chains of the α1 strand and the β1 strand (Fig. 3C). Only 10–13 hydrogen bonds at the central interface are present in the resolved human and mouse MHC-II structures, regardless of the number of hydrogen bonds formed between the α- and β-chains (Fig. 3C, 3D, Table II).

In addition, the lengthening of the α1 strand results in interaction between the bottom of the PBG and the sidewall of the α1 helix, because BLA is more dispersed (Fig. 4A–D). The hydrogen bond between Glu^α16 and Arg^β82 in BL2*019:01 is located at the C terminus of the PBG and does not exist in the human and mouse structures. Additionally, there are more interactions between the α1 helix on the side of the PBG and the strands at the bottom of the α1 domain, reaching six hydrogen bonds, with only one to four hydrogen bonds in the human structure (Fig. 4A–D). This result might indicate that the stability of invariant chicken BLA itself is enhanced. Importantly, the four-residue insertion at positions 16–19 of the α1 domain is found in avian and reptiles, suggesting that the increase in main-chain hydrogen bonding caused by the insertion may be conserved in these animal groups (Fig. 3B).

Analysis of the interactions between the α1 helix of the α1 domain and the α2 helix of the β1 domain reveals no direct interaction between the two helices in pBL2*019:01. Instead, the two helices connect indirectly through a molecule of water and P10-Ser (Fig. 5A). Similarly, Asn^α79 and Leu^β53 are connected by a water molecule. Comparison of the structures shows a difference in the direction and distance of the key residues (Fig. 5B). Thus, the condition for forming the salt bridge is absent. The main reason for this change is that the key residue Arg^α79, which is completely conserved in the human and mouse pMHC-II structures, is replaced by Asn^α79 in pBL2*019:01 (Fig. 5D). The Arg^α79 residues in humans and mice always form hydrogen bonds with conserved Leu^β53 (Fig. 5C). Because of the residue difference at position β57 in the human and mouse pMHC-II structures, the conserved Arg^α79 forms hydrogen bonds and/or salt bridges with the β57 residue or connects via a water molecule. Gln is present at position β57 in pBL2*019:01 (Fig. 5E).

According to the clear electron density, the peptide adopts a type II polyproline helical conformation and consists of 17 residues, even though nine residues are generally observed in the binding site of the PBG (Fig. 6A). Peptide side chains that orient away from the PBG have increased B factors, and the side chains are flexible. The peptide forms a hydrogen-bonding network that extends from the main chain of the peptide to both the main-chain and side-chain residues in the PBG of pBL2*019:01 (Fig. 6B). There are 18 hydrogen bonds between the PBG and the main-chain atoms of the peptide, 7 derive from the BLA chain and 11 from the BLB2 chain (Fig. 6B). Twelve of the eighteen hydrogen bonds that involve the PBG of pBL2*019:01 are conserved in most human and mouse pMHC-II structures. Asn^α79 and Gln^β57 form four hydrogen bonds, and because the peptide to which pBL2*019:01 binds has 17 aa, it is those four hydrogen bonds that cause the peptide C terminus to dip. Thus, differences in some hydrogen bonds lead to changes in the direction of the peptide backbone.

Four pockets accommodate the side chains of the peptide at the P1, P4, P6, and P9 positions, but the PBG of pBL2*019:01 also forms the P7 and P10 clefts (Fig. 7). The P3 and P8 pockets consist of only BLA-chain amino acids, whereas the P2 pocket consists of only BLB2-chain amino acids. The P1, P4, P6, P7, and P9 pockets comprise both BLA and BLB2 chain amino acids. The P1 pocket is composed of the BLB2-chain amino acids His^β81, Asn^β82, Gly^β85, and Val^β86 (Fig. 7B). Gly^β85 and Val^β86 affect the shape and size of the P1 pocket. β85 Is located on the sidewall of the P1 pocket; β85 is Gly in BL2*019:01, but other β85 residues in chicken are Ile/Val. Therefore, the P1 pocket sidewall of BL2*019:01 is the lowest and more open. β86 Is located at the bottom of the P1 pocket, and the change in β86 of the BLB chain results in a different depth of the P1 pocket and restricts the size of peptide side chains that might be accommodated at this position. The P6 pockets are large but filled with four water molecules rather than proline (Fig. 7D). The water molecules occupy the depth of the pocket; thus, in the absence of those water molecules, amino acids with even larger side chains can be accommodated. The P7 pocket has a negative charge and is stabilized by a hydrogen-bonding network involving two water molecules and Arg^β71, Tyr^α30, and Asn^β73 (Fig. 7E). As the side chain of Asn^α79 is short and cannot participate in the formation of the P9 pocket, the boundary of the P9 pocket in pBL2*019:01 consists of Val^α75 and Asn^β57; in other structures, it is composed of Val^α75, β57, and Arg^α79 (Fig. 7F). Therefore, the P9 pocket of pBL2*019:01 is more open than that in other structures. The surface-exposed amino acids at P-5, P-1, P3, P5, P8, and P11 may dock TCRs.

The nonclassical MHC-II protein DM plays a critical role in the endosomal peptide selection process. Based on the HLA-DR1–DM structure (42), the DM molecule mainly contacts the α1 domain of pMHC-II close to the P1 pocket (Fig. 8A). In the DM–DR1 structure, two key residues (Lys^α41 and Glu^α43) of DR are at the interface with DM; these charged residues form an extended hydrogen-bonding network that includes Asp^α183, Arg^α98, and His^α180 of DM. The importance of this network is highlighted by previous mutagenesis data: the mutation Glu^α43 in DR results in unresponsiveness to DM. Residues located at Ala^α41, Ala^α42, and Gln^α43 in the BLA chain are converted to noncharged amino acids, which may hinder the formation of hydrogen bond networks with chicken DM molecule. In addition, polymorphic residues of the BLB2 chain are almost located in areas forming pockets in the pBL2*019:01 structure (Figs. 2, 8B).

The pBL2*019:01 structure links nonmammals and mammals to provide an understanding of the presence of Ag peptides in the PBG as well as activation of T cells to further promote the production of specific humoral and cellular immunities in birds. It is important that the three dimensional structure of pBL2*019:01, the first nonmammalian pMHC-II crystal structure, is elucidated.

Although the architecture of pBL2*019:01 is similar to that of human pMHC-II (Fig. 1A), many unique structural features were found in pBL2*019:01, which may explain the novel α- and β-chain–pairing mechanism. Chickens have only one classical monomorphic BLA gene and two classical highly polymorphic BLB genes (20). The α1 strand of the BLA chain contains four more amino acids than in humans and mice, which results in a longer α1–α2 region and increased main-chain hydrogen bonding at the central strand interface, stabilizing the bottom of the PBG (Fig. 3). In addition, the residues interacting with the chicken BLA chain to form these hydrogen bonds are highly conserved in known polymorphic chicken BLB chains (Fig. 2). At the C terminus of the α1 and α2 helices, Asn^α79 in the chicken structure replaces the conserved basic residue Arg^α79 in the human and mouse structures. Furthermore, β57 is a nonpolar Asn residue in the chicken structure, which prevents formation of the salt bridges present in the human and mouse structures; instead, the two helices employ a novel method to connect via water molecules or peptide residues (Fig. 5). Based on the premise of the above structure, the chicken BLA chain rationally matches all chicken β-chains from BLB loci. In classical mammalian MHC-II molecules, the α-chain and β-chain are encoded by A and B genes, which are generally located adjacent to each other (43), and this suggests that these genes can coevolve relatively easily to generate functional dimers. The BLA locus is located outside of BLB genes in chickens, and this genetic separation may have allowed the β-chain to be highly polymorphic and forced the α-chain to become monomorphic toward an average fit. Thus, only one α-chain remains to best function with all β-chains. In addition, the BLA chain might be the α-chain partner with the nonclassical class II β-chains encoded by the MHC-Y (22). HLA-DRα and I-Eα are also monomorphic in humans and mice. Evolutionary analysis has shown that the chicken α-chain is highly similar to the HLA-DRα–chain and I-Eα–chain in humans, mice, and other mammals (20). The short α1–α2 region found in most MHC-II α1 domains is probably not the original MHC form, as the region tends to be longer in the MHC-I α1 domain (Supplemental Fig. 1). Therefore, the α1–α2 region of MHC-II may have lost length to reduce sequence-independent interstrand interactions and to rely more on specific binding, facilitating evolution of A and B pairs that do not share each other’s A or B chain. The β strand extensions cause an increase in main-chain hydrogen bonding, which makes it more difficult to select A-to-B binding based on lineage-specific interactions. This suggests that in species that show the α1 domain region extension, namely birds and reptiles, there is reduced evolutionary pressure to maintain multiple classical MHC-II lineages, with A and B molecules that do not cross-react. Indeed, the suborder Galloanserae (Anseriformes [e.g., duck] and Galliformes [e.g., chicken]) and probably also Serpentes (snakes) appear to express only classical α-chains, with which the multiple classical β-chains found in the same species likely interact (20, 44). Conversely, in mammals, the lineages DP, DQ, and DR stably coexist and have specific interaction motifs that are conserved for each lineage, which may explain why the A and B lineages do not show productive cross-lineage interactions.

Activation of T cells by the PBG-presenting Ag peptide is the core function of pMHC-II (45). The PBG is more open at both ends in the 3D structure of pBL2*019:01 than in other structures (Fig. 7). Because the chicken CD4 molecule carries a five-residue deletion, interaction between BL2*019:01 and CD4 may be different (Fig. 1D). Key residues of the BLA chain involved in HLA-DM–DR complex interactions have been substituted, suggesting that the DM–MHC-II binding mode in chickens may be different from that in humans (Fig. 8A). Because the α-chain is invariant, the diversity of the presented Ag peptide depends on the β-chain. In the heterodimer, the function of the α-chain may be to bind with the β-chain to stabilize the conformation, whereas the β-chain may tend to select and bind peptides. Understandably, the difference in binding peptides leads the α-chain to adjust accordingly. Finally, it must be emphasized that the chicken pMHC-II α- and β-chain–pairing mechanism found in this study can be used as a reference for understanding the pairing mechanism of human and mouse MHC-II. The chicken pMHC-II complex establishes a structural principle for designing epitope-related vaccines for the prevention of diseases in chicken.

We acknowledge the assistance of the staff at the Shanghai Synchrotron Radiation Facility of China. We thank Prof. Jianxun Qi (Institute of Microbiology, Chinese Academy of Sciences) for help in structure refinement.

The authors have no financial conflicts of interest.