The structure of HLA-DR52c: Comparison to other HLA-DRB3 alleles

Abstract

Class II major histocompatibility complex (MHCII) molecules present antigens to CD4⁺ T cells. In addition to the most commonly studied human MHCII isotype, HLA-DR, whose β chain is encoded by the HLA-DRB1 locus, several other isotypes that use the same α chain but have β chains encoded by other genes. These other DR molecules also are expressed in antigen-presenting cells and are known to participate in peptide presentation to T cells and to be recognized as alloantigens by other T cells. Like some of the HLA-DRB1 alleles, several of these alternate DR molecules have been associated with specific autoimmune diseases and T cell hypersensitivity. Here we present the structure of an HLA-DR molecule (DR52c) containing one of these alternate β chains (HLA-DRB3*0301) bound to a self-peptide derived from the Tu elongation factor. The molecule shares structurally conserved elements with other MHC class II molecules but has some unique features in the peptide-binding groove. Comparison of the three major HLA-DBR3 alleles (DR52a, b, and c) suggests that they were derived from one another by recombination events that scrambled the four major peptide-binding pockets at peptide positions 1, 4, 6, and 9 but left virtually no polymorphisms elsewhere in the molecules.

Both the mouse and the human have multiple isotypes of class II major histocompatibility complex (MHCII) molecules (1). The human MHC contains three major isotopes: DR, DP, and DQ. As in the mouse I-A molecule, both the α and β chains of DP and DQ are polymorphic. Because most of the polymorphisms involve amino acids lining the peptide-binding groove of the molecule, the different DP and DQ alleles bind peptides with different motifs in the major anchor amino acids of the peptide at positions 1, 4, 6, and 9. In contrast, DR molecules, like the mouse I-E molecule, all share a common, nearly invariant α chain. The similarity between the DR and I-E α chains is so extensive that often they can be exchanged with no loss of peptide-binding specificity (2). Thus, differences in the β chain account for the different peptide-binding motifs of the DR and I-E molecules. Whereas in the mouse these differences are confined to polymorphisms in a β chain encoded in a single gene, in the human, multiple genes can encode the β chain, thereby increasing the potential variations in the molecule within the human population. The most highly expressed β chain is encoded by the HLA-DRB1 gene, which is extremely polymorphic and often dominates antigen presentation for foreign peptide antigens.

A number of other genes linked to HLA-DRB1 encode the alternate β chains (3). Most of the alternate β-chain genes are pseudogenes; however, versions of three of them (DRB3[DR52], DRB4[DR53], and DRB5[DR51]) encode functional proteins. These “minor” DR subisotypes are less polymorphic and often less well expressed than HLA-DRB1, but nonetheless are fully functional in antigen presentation and contribute to the overall repertoire of MHC/peptide complexes that can result from the response to a single protein. Because these alternate β-chain genes are in linkage disequilibrium with the HLA-DRB1 gene, their function can be misassigned to the associated major HLA-DRB1 allele (4).

The DRB3 gene has 3 major alleles: DRB3*0101 (DR52a), DRB3*0202 (DR52b), and DRB3*0301 (DR52c) (5, 6), with allele frequencies 0.073, 0.136, and 0.120, respectively (7). Although the DRB3 gene-encoded molecules have not been as well studied as those encoded by DRB1 genes, these MHCII molecules have been linked to several autoimmune diseases; for example, DR52a has been associated with Crohn's sarcoidosis (8), Graves' disease (9, 10), and Löfgren's syndrome (11); DR52b has been associated with Graves' disease (12), multiple sclerosis (13), and essential hypertension caused by Chlamydia pneumoniae infection (14); and DR52c has been related to Crohn's disease (15) and cedar pollen allergies (16). Recently, we identified DR52c as the MHC-restriction element for a human nickel-reactive T cell isolated from a patient with contact sensitivity to nickel in jewelry (4). DR52c also was found to exclusively bind a minor histocompatibility antigen encoded by the Y chromosome, which evokes CD4⁺ T-helper cell responses, which can lead to the failure of stem cell grafts between HLA-identical siblings (17).

Here we present the structure of DR52c bound to a self-peptide derived from the Tu elongation factor (18). Although this structure is very similar to those of other MHCII molecules, features of the peptide-binding groove anchor pockets predict a unique peptide-binding motif for this molecule.

Results and Discussion

Overall Structure of DR52c Bound to a Self-Peptide Derived from the Tu Elongation Factor.

As described in Materials and Methods, we expressed, purified, and crystallized a soluble version of DR52c with a covalently attached self-peptide derived from the Tu elongation factor and subsequently solved its structure to a resolution of 1.8 Å. The electron density for the bound peptide was unambiguous and well defined (Fig. 1A). As expected, the overall structure of DR52c-Tu is very similar to that of the many other MHCII structures that have been reported previously (Fig. 1B). The Tu peptide is firmly bound in the peptide-binding groove between the helices of the α and β chains. The peptide backbone follows a familiar course through the binding groove, stabilized by conserved MHC interactions with the peptide backbone.

Fig. 1.

Crystal structure of the Tu peptide bound to DR52c. A wire frame representation of the Tu peptide is shown with coloring: carbon, white; oxygen, red; nitrogen, blue. The view is from the side looking through the β1 domain of DR52c. (A) A 2fo-fc electron density map of the residues P2-P10 [human elongation factor Tu (343–354)] of the Tu peptide contoured at 1 σ as it is bound to DR52c. (B) Superposition of the Tu peptide bound to DR52c and the human integrin peptide bound to DR52a (magenta). The structures were overlain through the backbones of the α1 and β1 domains using the O program.

Peptide-Binding Motif of DR52c.

The four amino acids at P1, P4, P6, and P9 are the major anchor residues for binding of the Tu peptide to DR52c. Their side chains occupy complementary pockets within the binding groove (Fig. 2A). The Ile side chain at P1 occupies a pocket lined with hydrophobic amino acids (Fig. 2B). Its size and shape predict that the pocket can optimally accept amino acids with aliphatic side chains, such as Ala, Val, Leu, and Ile. The DR52c molecule contains a hydrophilic pocket that is tailor-made for Asn at P4 (Fig. 2B). The nature of this pocket explains the strong selection for Asn (or Asp) at P4 among peptides known to bind to DR52c (18). The P4 pocket is stabilized by an intricate hydrogen-bonding network involving three water molecules and Gln α9, Glu β28, Lys β72, and Gln β74 of DR52c (Fig. 3A). These hydrogen bonds fix the orientation of the peptide P4 Asn side chain. The Pro at P6 sits in a pocket with both hydrophilic and hydrophobic properties that should accept virtually any aliphatic or short-polar amino acid side chain (Figs. 2B and 3B). The P6 pocket is connected to the P9 pocket through a tunnel at the bottom of the binding groove, which is blocked by two ordered water molecules (Fig. 2A). These water molecules foreshorten its depth; thus, in the absence of this water, amino acids with even larger side chains can be accepted (Fig. 3B). The similar tunnel also may be present in DR3, DR4, and other DR52 alleles, depending on the conformations of involved residues. Finally, the Ile at P9 sits at the top of a deep, mostly hydrophobic tunnel, at the bottom of which lies a Glu β9 (Fig. 2B). This pocket is large enough to accept virtually any hydrophobic amino acid, even Phe or Tyr. In a manner similar to the mouse I-E^k molecule, this pocket also can accept a Lys with its ε-amino group salt bridged to the carboxylate of Glu β9 (19).

Fig. 2.

DR52c peptide-binding pockets. The α1 domain is shown in cyan; the β1 domain, in magenta. The Tu peptide is colored as in Fig. 1. Water molecules deep in the pockets (shown as red balls) are labeled W0, W1, and W2. (A) Side view of the solvent-accessible surface (probe radius of 1.4 Å) of the peptide-binding groove. The peptide is shown as a wire frame. The β1 helix has been partially cut away for clarity. (B) Top view of the solvent-accessible surfaces of the four peptide amino acid-binding pockets (P1, P4, P6, and P9) of DR52c, showing only the side chains of the four bound residues.

Fig. 3.

The hydrogen-bonding network in the P4 and P6 pockets. The predicted hydrogen bonds among the atoms are shown as green dotted lines. The water molecules are shown as red balls. White, peptide carbon; cyan, α chain carbon; magenta, β chain carbon; blue, nitrogen; red, oxygen. (A) Details of the hydrogen bond interactions between P4 Asn and the DR52c. The side chains of Gln α9, Ser β13, Glu β28, Lys β72, and Gln β74 are shown as sticks. Three ordered water molecules are labeled W3, W4, and W5. (B) Details of the peptide-binding pocket P6. The side chains of P6 Pro and DR52c Asp α66, Asn α69, Glu β9, Tyr β30 and Leu β11 are shown as sticks. Two ordered water molecules are labeled W1 and W2. (C) A hypothetical network in the P6 pocket occupied by Arg.

Relatively little data are available on the nature of foreign and self-peptides that bind to DR52c. However, an examination of this short list of sequences readily identifies the DR52c-binding motif within each peptide (Fig. 4). The 10 peptides can be aligned with an aliphatic residue at P1, an Asn or Asp at P4, and acceptable amino acids at P6 and P9. P1 and P4 appear to be the major determinants of specific peptide binding. Some previous mutagenesis studies of DR52c-binding peptides support this view (16, 22); for example, in the CRj 1 peptide, the substitution of a P1 Leu to Ser completely abrogates binding to DR52c, and mutation of the P4 Asn to Ala or Gln also inhibits binding (16). Particularly revealing are studies of the Ras peptide. The natural Ras peptide with Gly at P4 fails to bind DR52c. Binding ability is gained by mutating the Gly to Asp but not to Arg, Val, Ser, or Ala (22). Thus in individuals with DR52c, a Gly-to-Asp mutation produces a neo-epitope for potential T cell recognition.

Fig. 4.

The DR52c peptide-binding motif. The sequences were aligned based on the predicted occupancy of the four binding pockets, with particular weight given to P1 and P4. Only the central regions of the peptides are shown, with the positions of the four anchor residues (P1, P4, P6, and P9) highlighted. References: elongation factor Tu (342–356), HLA DRα chain (110–127), and Ig λ chain (11–24) (18); tetanus toxin (830–843) (20); tetanus toxin (1273–1284) (21); P53 (243–253), G245S, P53 (278–291), G282W, Ras (3–20), and G12D (22); Crj 1 (335–346) (16); Y-chromosome encoded antigen (152–161) (17).

The P6 and P9 pockets are more tolerant of different residues; for example, changing the P6 Arg to Ser in the p53 (243–253) peptide does not inhibit binding to DR52c (22), and likewise, substitution of Gly to Ser or Ala in P6 or of Thr to Ala or Ser in P9 of the CRj 1 peptide does not impair DR52c's ability to activate T cells (16). Interestingly, from the structure, we can see how an Arg or Lys, such as found in the p53 peptide or one of the tetanus peptides shown in Fig. 4, also could occupy the P6 position. Within the P6 pocket are several negatively charged residues, similar to those seen in the P6 pocket of the mouse I-E^k MHCII molecule (23). Arg or Lys could be inserted into the bottom of P6 and replace the water molecule W1 (Fig. 3B), and Gln could rotate into the position of water molecule W2. Asp α66, Glu β9, and Asn α69 could form hydrogen bonds to Arg or Lys instead of to water molecules, thereby maintaining the hydrogen-bonding network (Fig. 3C). Intriguingly, Glu β9 could adopt different rotamers and form hydrogen bonds with positively charged residues in either the P6 or P9 pocket, but not in both.

Finally, stem cell transplantation between HLA-identical sex-mismatched siblings sometimes fails. Recently, Spiering et al. (17) discovered a human Y chromosome-encoded peptide, VIKVNDTVQI, presented by DR52c, that can activate CD4⁺ T-helper cells, leading to graft rejection. Our proposed binding motif for DR52c closely fits this peptide and predicts the surface-exposed amino acids at P-1, P-2, P-3, P-5, P-7, and P-8 that may make TCR contacts (Fig. 4).

Comparison of the Binding Pockets of DR52c with Other HLA-DR Molecules with DRB3 β Chains.

The sequences of the β1 domains of the three major DRB3 suballeles are very similar. All of the DNA sequences differences result in amino acid differences. The protein sequence identities between DR52c and DR52a or DR52b are 95.3% and 94.8%, respectively (Fig. 5). Remarkably, all of the amino acid differences except one (β51) reside in the peptide-binding pockets. These residues modify the DRB3 alleles to have different peptide preferences, which may explain the differences in their associations with particular diseases. The structure of DR52b is not available; however, the crystal structure of DR52a has been determined (24).

Fig. 5.

Sequence alignment of the β1 domains (aa1-aa90) of the DR52 alleles. The polymorphic differences are highlighted in red. The areas of DR52a and DR52b that may have recombined to generate the original DR52c gene are boxed.

The P1 pocket of DR52c is very hydrophobic and belongs to the group of MHCII molecules, including DR3 and DR15, that have a Val at β86. In contrast, both DR52a and DR52b have a Gly at β86. As pointed out previously (18), this Gly at position β86 makes the pocket larger so that it can accommodate virtually all hydrophobic residues, including large aromatic amino acids, whereas the Val partially blocks the pocket, selecting smaller aliphatic anchor residues.

DR52c and DR52a have very different P4 binding pockets. Arg β11 and Arg β74 are present in DR52a (24), making the P4 pocket highly selective for negatively charged residues. The P4 pocket of DR52c also is very selective. DR52c has a Gln at position β74, whereas most other DR molecules, including DR52a, have an Ala or Arg. Gln β74 together with other residues in the P4 pocket form a tightly knit hydrogen-bonding network that apparently limits the binding residues to Asn or Asp. Interestingly, DR52b has an identical amino acid composition in P4 as DR52c (Fig. 5) and also is selective for Asn (18).

The P6 pocket of DR52a is very shallow and usually occupied by residues with small side chains, giving rise to an unusual “1–4-9” peptide-binding motif (24). This is due mainly to the presence of Arg β11, which shortens the pocket considerably. In DR52c, this Arg β11 is replaced by a Leu, which gives the P6 pocket of DR52c more room to bind larger amino acids, even Arg or Lys. DR52b has a similar P6 pocket as DR52c (Fig. 5); however, the β30 of DR52b is His instead of Tyr, making the P6 pocket even larger.

The linings of the P9 pockets of DR52a and DR52c are very similar, with the only difference being the substitution of Leu β38 in DR52a for a Val in DR52c. Therefore, DR52a and DR52c should have a similar P9 pocket restriction. Within the P9 pocket, Asp β57 is a strongly conserved residue in both human and mouse MHCII molecules. It has been suggested that the MHCII molecules, with substitution of this Asp to other residues, may be more susceptible to disease (25). In the structures of both DR52a and DR52c, the substitution of Val for Asp at this position allows Arg α76 to move upward, because the conserved interaction between Asp β57 and Arg α76 is now missing. It is tempting to speculate that this change in the pocket structure may play a role in the disease susceptibilities associated with these alleles. The P9 pocket of DR52b is very different from that of DR52a and DR52c; it uniquely contains Tyr β37, Ala β38, Asp β57, and Tyr β60, making it more acceptable to smaller, polar, or charged residues.

The fact that the DR52 alleles are so closely related points strongly to their derivation from a common precursor gene through a gene conversion or other type of recombination event (5, 6). DR52a is nearly identical to DR52c from aa30 to aa73. DR52b is identical to DR52c for the first 29 aa and from aa74 to aa90 (Fig. 5). Therefore, one way of interpreting the sequence data are that DR52c was derived from a recombination between DR52a and DR52b (5). Because the structural data for DR52a and DR52c show that the differences among the DRB3 alleles were confined to the peptide-binding pockets, this recombination effectively scrambled the peptide-binding pockets. Thus, DR52c ended up with the P1 and P4 pockets from DR52b, the P9 pocket from DR52a, and a hybrid P6 pocket. This sharing of certain peptide-binding pockets may explain why some autoimmune diseases are associated with two of the three alleles; depending on the relative use of particular pockets, a given peptide might bind to more than one of these allelic forms of DR52.

Conclusion

The DRB3-encoded molecules share the same DRα chain and identical DRB3 amino acid sequences in regions important for TCR interaction. However, their few differences change the nature of the peptide-binding pockets, accounting for their differences in peptide selection and presentation.

Materials and Methods

Reagents.

Oligonucleotides were synthesized in the Molecular Resource Center of the National Jewish Medical and Research Center. Automated DNA sequencing also was performed in this facility.

Preparation of Soluble DR52c with a Bound Peptide.

A construct was prepared in a two-promoter Escherichia coli baculovirus transfer vector (26) as described in supporting information (SI) Fig. S1. In brief, the sequence encoding the signal peptide and extracellular region of the mature common DR α chain (truncated at the end of the α2 domain) was cloned behind the baculovirus p10 promoter. The sequence encoding the extracellular region of the mature DR52c β chain (truncated at the end of the β2 domain) was cloned behind the viral polyhedrin promoter fused to the sequence encoding the mouse I-A^b signal peptide. The sequence encoding a DR52c-binding aa13 peptide (QVIILNHPGQISA) derived from the Tu elongation factor followed by a flexible glycine-rich linker (GGGGSLVPRGSGGGG) was inserted into the frame between the signal peptide and the N terminus of the DR52c β chain, such that after expression and signal peptide cleavage, the peptide would occupy the peptide-binding groove of DR52c (26). The final construction was sequenced and then incorporated into baculovirus in SF9 insect cells using standard homologous recombination and Baculogold (PharMingen) as the recipient baculovirus DNA. The virus was cloned at limiting dilution, using a capture enzyme-linked immunosorbent assay [FK7.3.19.1 anti-DR52c capture (18), L243 anti-DRα detection (27)] to detect soluble DR52c in the culture supernatants of insect cells infected with recombinant clones. A single clone was selected and expanded to provide a high-titer stock for large-scale production of DR52c.

To prepare protein for crystallography, several liters of High Five insect cells (≈5 × 10⁵/ml) were infected at a multiplicity of infection of ≈5. After 5 days, the culture supernatant was collected by centrifugation and passed through a 0.2-μ filter. The soluble DR52c-Tu was isolated from the supernatant by immunoaffinity chromatography using the anti-DRα-specific monoclonal antibody LB3.1 (American Type Culture Collection). The eluate from the column (pH 11.4) was concentrated and further purified in a Superdex-200 size-exclusion chromatograph. The homogeneous peak corresponding to a molecular mass of ≈60 kDa was collected and concentrated to 7 mg/ml, and the buffer was exchanged into 5 mM NaN₃ and 10 mM 4-(2-hydroxyethyl)-1-piperazineethanesulfonic acid at pH 7.5. The overall yield was about ≈5 mg/l, and the protein appeared to be nearly homogeneous on sodium dodecyl sulfate polyacrylamide gel electrophoresis, isoelectric focusing, and native gel analysis. This stock was used for crystallography trials.

Crystallization and Data Collection.

DR52c-Tu was crystallized at room temperature by hanging drop vapor diffusion against 1 ml of mother liquor containing 0.2 M sodium chloride, 1.6 M ammonium sulfate, and 100 mM Tris·HCl (pH 7.5). In all of the crystallization setups, 0.5–1.0 μl of protein solution was mixed with an equal volume of reservoir solution. Crystals normally formed within 2 weeks. The x-ray diffraction data were collected under liquid-nitrogen cryo conditions at 100° K. The DR52c-Tu crystals were flashed-cooled in liquid nitrogen after a short soak in a cryo-protection solution consisting of the reservoir solution with 20% (wt/vol) glycerol added. The x-ray diffraction data were measured at SBC beamline ID-19 at the Advanced Photon Source, Argonne National Laboratory. The data were indexed, integrated, scaled, and merged using the HKL2000 program (28).

Structure Determination and Refinement.

The structure of DR52c-Tu was determined by molecular replacement using the AMoRe program (29) with human MHCII HLA-DR1 (pdb code: 1DLH) without a bound peptide as the search model. The initial fo−fc map revealed a clear positive density within the peptide-binding groove of DR52c. Tu peptide was modeled into this positive density. Models were manually adjusted using the O program (30).

Before refinement, an independent set of 5% reflections was set aside for R_free calculation (31) for all datasets. No sigma cutoffs were used in any of the refinements. Both conventional R-factor (R_working) and R_free were used to monitor the progress of refinement. The models were subjected to several rounds of alternating simulated annealing/positional refinement, followed by B factor refinement using the crystallography and NMR system software suite (32). Simulated annealing omit maps were routinely used to remove the model bias. The final model exhibited good stereochemistry, as determined by the PROCHECK program (33). Most of the residues were well defined; however, the loop at residues 106–110 of the DR52c β chain was disordered. Data collection and refinement statistics are summarized in Table 1. Figures were prepared using the PyMOL graphics and modeling package (34).

Table 1.

Data collection and refinement statistics

Data collection
Space group	P43222
Cell dimensions
a, b, c, Å	83.38, 83.38, 132.91
α, β, γ, °	90, 90, 90
Resolution (Å)	50–1.8 (1.86 −1.8)†
Rmerge‡	6.4 (80.3)
I/σ	14.9 (2.1)
Completeness, %	96.5 (87.7)
Redundancy	5.7 (4.6)
Refinement
Resolution, Å	41.7–1.8 (1.91–1.8)
No. of reflections	42,672 (6084)
Rworking/Rfree§	20.2/22.7 (27.9/28.7)
Number of atoms
Protein	3073
Carbohydrate	28
Water	381
B factors
Protein	26.8
Carbohydrate	73.9
Water	38.0
rmsds
Bond lengths, Å	0.005
Bond angles, °	1.3
Ramachandran plot
Most favored region, %	92.9%
Additional allowed region, %	7.1%
Generously allowed region, %	0.0%
Disallowed region, %	0.0%

^†All data (outer shell).

^‡R_merge = Σ(|I −<I>|)/Σ(I).

^§R_working and R_free = Σ||F_o| − |F_c||/Σ F_o|.