Crystal structure of HLA-A*2402 complexed with a telomerase peptide

Overall structure of A24 complexed to VYG (A24 VYG)

The structure of A24 complexed to VYG (A24 VYG) was solved at 2.8 Å and refined to an Rcryst and Rfree of 18.3% and 25.6%, respectively (Table 1, 2). The complex crystallized in space group C2 with unit cell parameters a = 125.65 Å, b = 132.04 Å, c = 91.45 Å and β = 128.9º. The overall model consists of two molecules of A24 VYG in the asymmetric unit (AU). The A24 VYG molecules form a dimer linked by a sulfate-mediated salt-bridge. A pseudo twofold axis, centered on the sulfur atom, is formed between residues R79 and R83 of one A24 VYG molecule and their equivalents of the other copy in the AU (Fig. 1). Extensive non-specific crystal contacts are also made, stabilizing the crystal lattice, between the α3 domain of one molecule and the α1 and β2 m domains of symmetry equivalents. The root mean square deviation (RMSD) between the two molecules in the AU is 0.57 Å over all Cα atoms (Fig. 2). The A24 VYG molecule comprises a heavy α chain and a light β2 m chain, similar to other HLA I structures. The α chain is made up of three distinct domains: The α1 and α2 domains make up the peptide-binding interface, formed by an antiparallel β sheet and two long α helices, which mediate contacts with the TCR; the α3 domain, consisting primarily of antiparallel β sheets, forms the proximal membrane domain and makes extensive contacts with β2 m, which is another antiparallel β sheet domain. Superposition with the model of the A2 TAX: A6 structure (1A07) 20 used for molecular replacement (RMSD 0.90Å and 0.96Å with complex 1 and complex 2, respectively) showed that the overall conformations remain very similar. However, the TAX and VYG peptides have markedly different conformations in A2 and A24, respectively, since their sequences are quite different. This indicates a low likelihood of model bias in the present model.

Peptide binding interface

The electron density around the VYG peptide in the two copies of the AU clearly revealed an extended conformation in the HLA binding groove (Fig. 3). The surface area buried by peptide binding was calculated as 736 Ų and 737 Ų for complexes 1 and 2, respectively. The peptide binding groove has been classified elsewhere into six pockets, A to F 21. The same nomenclature was adopted here and was used in the analysis below. The general features of the binding are similar to those observed in other peptide-HLA I (pHLA I) complexes. The peptide termini provide binding deep into the HLA I binding groove with many specific contacts between peptide main/side chains and HLA I groove residues (Fig. 4). Specifically, VP1 forms a close hydrogen bond with Y159 and a more tentative bond with E63 in the A pocket of the A24 molecule. This feature would suggest some importance for this residue in anchoring the peptide, although peptide screening of A24-compatible molecules suggests that V is not conserved at this position. Of more importance is YP2, which is conserved in virtually all observed A24-compatible peptides (Table 3), indicating that the hydrogen bond formed between the bulky aromatic side chain and H70 in the B pocket of the HLA I is vital for peptide stabilization at the N terminus of the peptide 22 (Fig. 5). GP3 makes a hydrogen bond with K66 in pocket D, but this residue is not conserved and is probably not vital for peptide anchoring. Similarly, strong hydrogen bonds are made between VP5 with Q156 in pocket C, and AP7 and T73 in pocket E. These ‘middle’ contacts are presumably less important for peptide binding in general as they represent the more variable residues in terms of A24 peptide presentation. LP9 is the other important anchor residue and forms hydrogen bonds with Y84 and T143 in pocket F (Fig. 5), allowing the non-anchor residues to bulge out of the HLA α1:α2 binding groove (Fig. 6). These contacts are vital for pHLA I stability and presentation to T cells via the TCR.

The peptide conformation is unusual in having a middle anchor point at VP5, a feature similar to that observed only in B8 HIV-Gag and A1 HIV-1 peptide complexes 23, 24. The peptide adopts an ‘M’-shaped conformation that sees the backbone of positions 3 to 7 twice rising out of and dipping into the groove at alternate residues. Anchor residues at GP3 and AP7 lock the peptide into place on either side of the prominently exposed FP4 and RP6 residues, which form the two tops of the ‘M'. The anchor residue at VP5 secures the peptide in the groove between the two exposed residues, forming the dip in the ‘M’ (Fig. 3). In other examples, there is a tendency for two adjacent residues to bulge out from the HLA I surface, e.g. P4 and P5 or P5 and P6, on a more exposed central bulge for TCR docking 21, 25, 26.

Projected TCR:pHLA docking

Like other pHLA I complexes, the bulge is also observed around the central residues of the peptide (Fig. 6). The two most prominent and probably most important residues for TCR interaction are FP4 and RP6, which make little contact with the HLA I binding groove, leaving them free to present their side chains conspicuously for TCR docking. FP4 is not as prominent as RP6, illustrated by the relatively higher thermal parameter (45–50 Ų compared with 55–70 Ų for the latter) indicating greater mobility. When the two copies of the peptide in the AU are superposed, the peptides show almost identical topology, with an RMSD of 0.26 Å for the Cα domains. However, slight reorientation of the RP6 side chain was apparent (Fig. 6), exploiting the free volume around it. Such flexibility could be involved in allowing a greater range of TCR to bind. In other A24-compatible peptides, the FP4 and RP6 positions are not conserved, in agreement with the notion that these residues are pivotal to the specificity of TCR recognition. CP8, although not as prominently exposed as FP4 and RP6, also makes little contact with the HLA I surface and might be involved in TCR docking. Furthermore, VP1, which is not conserved in other A24-binding peptides, could potentially contact the TCR as the N-terminal side chain points away from the HLA I surface (Fig. 3); however, it is equally possible that the side chain could simply be pointing out into the solvent region, away from the TCR interface.

Generation of A24 and β2m expression plasmids

The A24 and β2 m expression plasmids were generated by QuickChange PCR mutagenesis (Stratagene) from pEX517, which contains A24 α chain residues 1–284, and JMB034B, which contains β2 m residues 1–100 under the control of the T7 promoter. A24 and β2 m sequences were confirmed by automated DNA sequencing (Lark Technologies).

Protein expression, refolding and purification

Of plasmid stock containing amino acids 1–284 of the A24 α chain, or amino acids 1–100 of β2 m, transformed into 20 µL of Rossetta DE3 competent E. coli cells, 1 µL was used to inoculate 1 L warm TYP medium (16 g/L tryptone, 16 g/L yeast extract, 5 g/L NaCl, 2.5 g/L K₂HPO₄) containing 100 µM ampicillin. The transformed cells were induced using 0.5 mM isopropyl β-D-thiogalactopyranoside, harvested by centrifugation, and the subsequent inclusion bodies were harvested by lysis in 20 mL lysis buffer (10 mM Tris, 10 mM MgCl₂, 150 mM NaCl, 10% glycerol), sonication and treatment with triton wash buffer (0.5% Triton X-100, 50 mM Tris pH 8, 100 mM NaCl, 10 mM EDTA). The protein, in the form of inclusion bodies, was washed and then dissolved in guanidine buffer (6 M guanidine, 50 mM Tris, 2 mM EDTA, 100 mM NaCl). Prior to refolding, protein concentration was estimated using a Bradford reagent kit (Sigma-Aldrich) using BSA as standard.

Refolding of A24 VYG was carried out as described 35. Briefly, 30 mg/L α chain was mixed with 30 mg/L β2 m at 37°C and 4 mg/L synthetic peptide for 15 min. This was then added to cold refold buffer (50 mM Tris pH 8, 2 mM EDTA pH 8, 400 mM L-arginine, 0.74 g/L mercaptoethylamine, 0.83 g/L cysteine). The refold was allowed to mix at 4°C for at least 1 h before dialysis. Dialysis was carried out at 4°C for 2 days, changing the dialysis buffer once. The refold was then filtered and anion-exchanged using a POROS 50HQ column (Applied Biosystems) in 10 mM Tris pH 8, and then gel-filtrated into crystallization buffer (10 mM Tris, 10 mM NaCl) using a Superdex 200HR column (Amersham). The purity of each protein was measured using an SDS-PAGE 10% Bis-Tris gel.

Crystallization

A24 VYG was concentrated to 20 mg/mL in 10 mM Tris, 10 mM NaCl buffer. Crystal screens were initiated using Crystal Screen (CS)1 (Hampton research), CS2 and CS Cryo 1 buffers 1–48 with 0.5-µL drops of protein and 0.5-µL drops of crystallization buffer, using the hanging drop method. Plates were incubated at 291 K and analyzed after 24 h, 48 h and 1 wk. Crystals were observed in CS2 buffers 23 and 32 after 24 h. Crystals were flash-cooled in liquid nitrogen in Cryo loops and the working data set was collected from crystals grown in CS2 buffer 23 (1.6 M AmSO₄, 0.1 M NaCl, 0.1 M HEPES, 10% dioxane).

Diffraction data collection and model refinement

Data were collected with the rotation method at Station 14.1 of the Synchrotron Radiation Source (SRS; Daresbury, UK) using an ADSC Quantum 4 CCD detector system. (Rigaku/MSC). The wavelength (λ) was set to 1.488 Å. The data set was collected from two crystals, where a total of 320 frames were recorded, each covering 1° of rotation. Reflection intensities were estimated with the MOSFLM package (http://www.ccp4.ac.uk/download.php) 36, and the data scaled, reduced and analyzed with the CCP4 package (http://www.ccp4.ac.uk/download.php) 37. Crystal data and relevant statistics are given in Table 1.

The refined model of the A6TCR:HLA-A2 TAX complex (Protein Data Base code 1AO7) was used as a search probe in AMoRe. The final fitted coordinates from AMoRe were checked for close contacts using the graphics programs ‘O’ (http://www.bio.net) 38. The model sequence was corrected prior to refinement with REFMAC5 (http://www.ccp4.ac.vk/download.php) 39. Manual adjustment of the model by visual inspection of the electron density maps was performed at various stages. Remaining positive density was filled, as appropriate, with solvent (ARP/wARP procedure in CCP4) or ligand molecules. The thermal parameters were allowed to refine, but were restrained to be similar to those of neighboring atoms (Table 1). The two pHLA I complex copies found in the AU were refined with medium non-crystallographic restraints applied. The latter graphics sessions for model adjustment were performed with COOT (http:/www.ysbl.york.ac.vk/∼emsley/coot/) 40. The final refinement statistics are listed in Table 2. The final coordinates were submitted to the Protein Data Base (code 2BCK).