|Home | About | Journals | Submit | Contact Us | Français|
In the immune system, B cells, dendritic cells, NK cells, and T lymphocytes all respond to signals received via ligand binding to receptors and coreceptors. While the specificity of T cell recognition is determined by interaction of T cell receptors with MHC/peptide complexes, the development of T cells in the thymus and their sensitivity to antigen are also dependent on coreceptor molecules CD8 (for MHCI) and CD4 (for MHCII). The CD8αβ heterodimer is a potent coreceptor for T cell activation, but efforts to understand its function fully have been hampered by ignorance of structural details of its interactions with MHCI. Here we describe the structure of CD8αβ in complex with the murine MHCI molecule H-2Dd at 2.6 Å resolution. The focus of the CD8αβ interaction is the acidic loop (residues 222-228) of the α3 domain of H-2Dd. The β subunit occupies a T cell membrane-proximal position, defining the relative positions of the CD8α and CD8β subunits. Unlike the CD8αα homodimer, CD8αβ does not contact MHCIα2 or β2-microglobulin domains. Movements of the CD8α complementarity determining region-(CDR) 2 and CD8β CDR1 and CDR2 loops as well as flexibility of the H-2Dd CD loop facilitate the monovalent interaction. The structure resolves inconclusive data on the topology of the CD8αβ/MHCI interaction, indicates that CD8β is crucial in orienting the CD8αβ heterodimer, provides a framework for understanding the mechanistic role of CD8αβ in lymphoid cell signaling, and offers a tangible context for design of structurally altered coreceptors for tumor and viral immunotherapy.
The recognition of cell surface MHC-peptide complexes by T cells lies at the heart of the adaptive immune response. Specific identification of MHC-peptide complexes is mediated by the immunoglobulin-like, clonotypic α and β chains of the TCR and components of the multi-subunit CD3 complex that transduce signals to the T cell. However, TCR/MHC-peptide interactions alone do not efficiently trigger T cells, necessitating the engagement of the T cell coreceptors CD8 or CD4 by MHCI or MHCII respectively on the presenting cell, for potent T cell activation (1-3). Engagement of CD8 recruits the Src family kinase Lck to the TCR signalling complex, augmenting a stimulatory cascade (4) that involves conformational changes in the cytoplasmic tyrosine-based motifs of chains in the CD3 complex(5, 6). CD8/MHCI interactions play two overlapping roles: one related to the direct participation of CD8 as a component of the TCR/MHC signaling complex (coreceptor function) (7-11); and a second in which binding to neighboring MHC molecules contributes to stabilization of the T cell/antigen presenting cell (APC) interface (accessory function)(12-15). As a cell surface disulfide-linked dimeric glycoprotein, CD8 occurs in CD8αα and CD8αβ isoforms, which have distinct cellular distribution and function. CD8αα is broadly distributed and is found on intestinal intraepithelial lymphocytes, γδ T cells, subsets of dendritic cells, and NK cell subpopulations. The coreceptor function of CD8αα has been re-examined (16), and it has been proposed recently that CD8αα may negatively regulate cell activation in some lymphoid cell subsets (17). In contrast, CD8αβ is expressed by αβ TCR thymocytes and mature peripheral αβ T cells where it is indispensable for thymic selection of CD8 T cells (18-20) as well as for activation of peripheral CD8 T cells (21, 22). By linking thymic MHC recognition and TCR signaling, CD8αβ guides the developing TCR repertoire towards appropriate self-MHC recognition (23).
Despite the critical importance of CD8αβ for normal T cell development and function, the molecular basis for its biological differences from CD8αα is still clouded in controversy. A number of laboratories have evaluated biophysical parameters of binding of CD8αα and CD8αβ for MHCI, resulting in the general acceptance that, despite clear differences in function, both isoforms bind MHCI with essentially the same affinity. This binding is consistently observed to be independent of the nature of the particular MHC-bound peptide, although there are clear differences among MHCI alleles (7, 10, 24-26). Recently, measurement of the two-dimensional kinetics of the cell surface constrained CD8/MHCI interaction led to generally the same conclusions that CD8αα and CD8αβ interact with MHCI in an allele dependent but TCR and peptide independent manner (27). Experiments using a set of chimeric murine β chains expressed in CD8β–deficient mice suggest that the functional avidity advantage of CD8αβ derives from contributions of both its ectodomain and its cytoplasmic domain (2, 28, 29).
CD8 α and β chains consist of an Ig-like domain that is linked via a stalk to a transmembrane domain and a cytoplasmic tail. The tail serves CD8α as an Lck docking site. The superior coreceptor activity of CD8αβ has been attributed to the stalk region of the β chain and its glycosylation (30-33) and to a palmitoylation site in the β cytoplasmic tail (11), which targets CD8αβ and the associated TCR to lipid rafts. Recent studies indicate that a conserved peptide motif of the TCR α chain connecting peptide plays a crucial role in CD8β participation in signal transduction (34).
Extensive analyses of polymorphic variants and single site mutants of MHCI (35-40) and CD8 (41), initially guided by the MHCI structures, and subsequently by structures of CD8αα/MHCI complexes, have identified key residues mediating this interaction. However, the lack of a definitive structure of an MHCI/CD8αβ complex has hampered the interpretation of such data. Consideration of the surface electrostatic charge of the original HLA-A2/CD8αα structure led to speculation that CD8αβ would bind in the same general position and orientation, with the CD8β subunit replacing the CD8α2, in a T cell distal position (42). Early mutational studies suggested a similar orientation (43). However, consideration of the differences in the length of the stalk region of CD8α as compared to CD8β prompted others to favor the opposite orientation (26), with the CD8β chain in the T cell proximal CD8α1 location. Further mutational analyses of the HLA-A2/CD8 and H-2Kb/CD8 interactions provoked the unusual hypothesis that CD8αβ might bind MHCI in two distinct orientations (44, 45). To clarify the role of CDαβ in providing coreceptor signals, to illuminate structural aspects of the contribution of CDαβ to T cell development and activation, and to resolve ambiguities inherent in the interpretation of mutagenesis data, we have determined the structure of murine CD8αβ in complex with H-2Dd at 2.6 Å resolution. For comparison with an unliganded MHCI molecule, we also report a new high resolution (1.7 Å) structure of an H-2Dd/β2m/peptide complex. These structural data are further employed to interpret extensive mutational data in the literature.
Bacterial expression and purification of the H-2Dd/mβ2m/P18I10 complex have been described earlier (46). Expression and purification of H-2Kb/mβ2m/ISFK8 followed the same protocol. (The ISFK8 peptide, ISFKFDHL, has been described previously (47)). For mouse CD8αβ, E. coli codon-optimized DNA sequences encoding their extracellular domains were chemically synthesized (Genscript Corporation, Piscataway, NJ), separately subcloned into the pET21b bacterial expression vector (Novagen, EMD Chemicals, San Diego, CA), and transformed into E. coli Rosetta 2 (Novagen). The codon-optimized DNA sequences are included as Supplementary Figure 1, and have been deposited in GenBank (accession numbers GQ247790, GQ247791). The encoded protein extends from Gly -5 to Gly161 for CD8α and from Ser -3 to Gly147 for CD8β. This numbering scheme preserves that previously used for mouse CD8αα and CD8αβ structures (48, 49), which begin with Lys1 and Leu1 for CD8α and β respectively. CD8α and β were separately expressed with the Overnight Express Autoinduction System (Novagen). Inclusion bodies containing CD8α and CD8β protein obtained from 500 mL expression cultures were each denatured in 10 mL of 6M guanidine-HCl in TRIS-EDTA buffer pH8 containing 0.1mM dithiothreitol. Insoluble material was spun out, the supernatants mixed together, and added dropwise over 15 min to 1 L of chilled refolding buffer (0.4M arginine hydrochloride, 100mM TRIS pH8, 2mM EDTA, 3mM reduced glutathione, and 0.3mM oxidized glutathione). After incubation for four days at 4°C, the solution was dialyzed against 25mM MES pH5.5, concentrated and bound to a Hi-Trap SP cartridge (Pharmacia), and then eluted in MES buffer containing 1M NaCl. After overnight dialysis against 25mM HEPES pH7, 150mM NaCl, the protein was subjected to size exclusion chromatography on a Superdex 75 column in the same buffer. The major peak with the predicted retention time of the dimer was recovered and dialyzed against 25mM HEPES pH7, 50mM NaCl. The protein was then subjected to ion-exchange chromatography on a mono S column (Pharmacia) developed with a 0.05M-0.5M NaCl gradient in 25mM HEPES pH7. Two major peaks were resolved with the disulfide-linked CD8αα homodimer eluting earlier in the NaCl gradient than the disulfide-linked CD8αβ heterodimer. SDS-PAGE under reducing and nonreducing conditions revealed the disulfide linkage. Edman degradation sequencing of the amino terminal 15 residues confirmed the localization of the CD8αα and CD8αβ dimers to the earlier and later peaks, respectively, on mono S chromatography. The αα homodimer peak constituted 35 to 40% of the total protein in the two peaks. The CD8αβ peak was collected, the salt concentration adjusted to 50mM by dilution, and protein was concentrated to 10mg/ml for crystallization trials. In experiments requiring CD8αα the α chain alone was expressed, refolded, and purified as described above.
Surface plasmon resonance binding experiments were performed on a BIAcore™ 2000, CM-5 chip surfaces of which were covalently coupled with either CD8αα or CD8αβ. Data were analyzed with BIAeval 3.2. Coupling conditions and data analysis were as described previously for TCR (50).
The CD8αβ/H-2Dd complex was crystallized using the hanging-drop vapor diffusion method at room temperature. The H-2Dd/mβ2m/P18I10 complex and the CD8αβ heterodimer were mixed in a 1 to 2 molar ratio to a final protein concentration of 5 mg/ml, and crystals formed within one month in 12% PEG 3000, 50mM HEPES, pH7.5. A single crystal of 0.1 × 0.1 × 0.1 mm was frozen in liquid nitrogen after dipping in Paratone-N. X- ray diffraction data were collected under a nitrogen stream at 100 K at beamline 22ID-D at the Advanced Proton Source at Argonne National Laboratory, at a wavelength of 1.0 Å, using a MAR300 detector. The data were indexed, integrated, scaled and merged with HKL2000(51). The statistics of the crystallographic data collection are summarized in Table I.
For unliganded H-2Dd/mβ2m/P18I10, crystals were grown at room temperature in hanging drops over 16% PEG-3350 containing 0.2M magnesium formate, and cryopreserved in liquid nitrogen. Diffraction data to 1.7 Å were collected on a single crystal at beamline X29 at the National Synchrotron Light Source, Brookhaven, using a ADSC Quantum-315r CCD detector. Data were processed with HKL2000 and the structure was solved by molecular replacement. Data collection and refinement statistics are reported in Table I.
The structure of the CD8αβ/H-2Dd complex was determined by molecular replacement using the programs MOLREP (52) and Phaser (53) of the CCP4 suite (54) with the H-2Dd/P18I10 portion of the H-2Dd/P18I10/Ly49A complex (55) (PDB (56) code 1QO3) and mouse CD8αβ (49) (PDB code 2ATP) as search models, respectively. The crystal belonged to the space group P212121 with one complex (H-2Dd heavy chain, β2m, P18I10, CD8α, and CD8β) in the asymmetric unit. The structure of unliganded H-2Dd/mβ2m/P18I10 was solved by molecular replacement with AMORE (54) using our previously determined H-2Dd/P18I10 structure (1DDH)(46) as a search model.
For the CD8αβ/H-2Dd dataset, after an initial round of rigid body refinement, the model was fitted manually with Coot (57). The structure was refined with simulated annealing, energy minimization, B factor refinement and water addition using CNS (58). The final model with an Rwork of 24.8 and Rfree of 29.2 was obtained. The first three N-terminal residues (Lys-Pro-Gln) of CD8α were not visualized, and the first residue of the mature CD8β (Leu1) was in good density. Part of the C-terminal stalk region (Asp-Val-Leu-Pro) of CD8β was seen and built into the model. Uninterrupted electron density was observed for H-2Dd heavy chain residues 2 to 275, β2m light chain residues -1 to 99, and the decamer peptide, as well as for CD8α residues 4 to 122 and CD8β residues 1 to 120. However, although backbone density was observed for CD8α Leu69 to Phe75 and Leu89 to Lys91, side chain density was indistinct. No electron density was visible for the bulk of the stalk regions of either CD8 subunit.
For unliganded H-2Dd/mβ2m/P18I10 the refinement steps and water addition were carried out in PHENIX (59) followed by inspection of the maps in Coot. Anisotropic refinement of B factors was included in view of the relatively high resolution of this dataset.
Analysis of the resulting structures was accomplished with programs in CCP4 (54) and PDBsum(60). All molecular graphics figures were generated with PyMOL (http://www.pymol.org). Coordinates and structure factors have been deposited with the protein data bank (PDB) with accession codes 3DMM (CD8αβ/H-2Dd) and 3ECB (H-2Dd/β2m/P18I10), and can be accessed at http://www.rcsb.org (56).
Bacterial expression constructs encoding the extracellular portion of CD8α and β, including the first interchain disulfide, were expressed in E. coli and purified (see Materials and Methods and Supplemental Figure 1). Surface plasmon resonance (SPR) binding studies, using either the CD8αβ heterodimer or a similarly engineered CD8αα homodimer and recombinant H-2Dd and H-2Kb, revealed affinity constants (KD) for these carbohydrate-free, disulfide-linked CD8 proteins of 6.7 to 38.4 μM (Figure 1). The measured solution affinities are similar to those reported by some (7), but greater than those measured by others for mouse (61) and human (10, 62) molecules, and may reflect differences among MHC molecules (27). We observe little difference in the apparent affinity of CD8αβ as compared with CD8αα for MHCI, consistent with previous findings (7, 27, 63). In addition, comparisons of binding of H-2Dd complexes prepared with different peptides revealed no significant difference in binding to CD8αβ, consistent with the accepted view that the influence of peptide is minimal (7, 27, 64) (data not shown).
Crystallization conditions for the complex of H-2Dd with CD8αβ were determined, and synchrotron diffraction data to 2.6 Å were collected (see Materials and Methods and Table I). The structure, solved by molecular replacement, revealed continuous electron density for all five chains of the complex (CD8α, CD8β, H-2Dd, β2m, and the P18I10 peptide), but no density was visible for the bulk of the stalk regions of either CD8 chain. CD8α chain density extended from residue 4 to 121, and CD8β from 1 to 123. The overall complex, roughly 70 Å by 70 Å by 60 Å, reveals the canonical MHCI/β2m/peptide complex, bound to the two Ig-like domains of CD8α and CD8β (Figure 2A). The CD8αβ heterodimer focuses on the α3 domain of the H-2Dd heavy chain, consistent with early studies that mapped the binding site using mouse and human MHC variants (35, 36). CD8β is located in a position equivalent to that of the CD8α1 subunit of the three CD8αα/MHC complex structures (CD8αα/H-2Kb (48), CD8αα/TL (65), and CD8αα/HLA-A2 (42) (Figure 3). This region lies adjacent to, but not touching, the MHCI α2 domain platform, in an “upper”, T cell proximal, position. CD8α occupies the same relative position as the CD8α2 subunit of the CD8αα complexes, and is positioned closer to the carboxyl-terminus of the H-2Dd α3 domain, in a T cell distal location.
Although the recombinant CD8αβ protein contained residues of the stalk that join the Ig-like domains to the transmembrane region, the electron density map revealed very little of this region, presumably due to flexibility of this part of the molecule. However, the first several residues of the CD8β chain stalk, extending to CD8β Pro123 were visualized. These residues (from Pro123 on) seem to point towards the T cell. The disposition of the H-2Dd α3 domain relative to the peptide-binding α1α2 domain, and the relationship of the β2m subunit to the MHCI heavy chain are conserved in this complex structure.
The CD8αβ heterodimer only contacts H-2Dd residues located on the α3 domain (see Figure 2 and Table II). This contrasts with CD8αα which also makes contact with residues of the MHCI α2 and β2m domains. The exclusive focus of CD8αβ on the α3 domain of H-2Dd decreases the buried surface between CD8αβ and H-2Dd to 963 Å2, which differs from the buried surfaces of H-2Kb, TL, and HLA-A2 in complex with CD8αα of 1756, 1855, and 1302 Å2 respectively (Table III). The shape complementarity statistic (Sc)(66), an indicator of three-dimensional fit of a ligand for its receptor, calculated for the interface between the CD8 heterodimer and the MHC heavy chain is 0.59 for the CD8αβ/H-2Dd complex which is similar to that calculated for CD8αα/H-2Kb (0.61) and CD8αα/TL (0.65), but less than that calculated for CD8αα/HLA-A2 (0.72). The CD8β subunit of CD8αβ contributes almost equally (49%) to the buried surface of the interface. This contrasts to the three CD8αα/MHC complexes in which the CD8α1, “upper”, subunit contributes the bulk of the buried surface area (69%, 71%, and 74% of the interface for H-2Kb, TL, and HLA-A2 respectively). In the three CD8αα/MHC structures, residues of the three CDR loops of CD8α1 and of the N-terminus bind through hydrogen bonds and atomic contacts to both the MHC α3 domain and also to β2m. The footprint of CD8αβ on H-2Dd is compared graphically with that of CD8αα on H-2Kb in Figure 3E and F, emphasizing the more extensive interactions of CDα1 with residues of β2m as well as with H-2Kb residues of the α2 and α3 domains. In addition, the CD8α2 subunit (of CD8αα) interacts over a larger surface area and with more residues of the H-2Kb α3 domain as compared with the CD8α subunit's interactions with H-2Dd. In contrast, in the CD8αβ/H-2Dd complex only five residues of CD8β: one in the CDR1 loop (Lys27), three in the CDR3 loop (Gly100, Ser101, and Pro102), and one in β-strand F (Val99) contact H-2Dd (see Table II and Figure 2E). Ser101 and Pro102 participate via hydrogen bonds whose focus is on Thr225 and the highly conserved Gln226 of the H-2Dd α3 domain (Figure 2C and Figure 4A). Although CDR1 of CD8β makes contact through residue Lys27 to H-2Dd α3 domain residues Asp212 and Thr214, its CDR2 makes none at all (Table II). Of the residues of the H-2Dd α3 domain that interact with CD8αβ (Table II and Figure 4), Gln226 is the only one that interacts with both subunits (Figure 2C and Figure 3E and F). The footprint of the CD8α subunit on the α3 domain (Figure 2C and E, Figure 3E and F) is only slightly altered compared to that of its counterpart, the CD8α2 (lower) subunit in the CD8αα/H-2Kb complex (Figure 2A and B).
Superposition of the bound and free forms of CD8αβ reveals differences in the CDR1 and CDR2 loops of CD8β as well as the CDR2 loop of CD8α (Figure 5). These loops adjust and are stabilized by interaction with the H-2Dd α3 domain. The largest adjustments are in CD8β CDR1 where the Cα atom of residue Lys27 is displaced by 2.9 Å and its Nζ atom by 8.8 Å in the liganded structure (Figure 5D and E). Also, His60 of CD8α CDR2 approaches H-2Dd residue Glu227 in the bound structure. Notably, all CDR adjustments move the loops farther away from each other in the bound as compared to the unbound state. (The minor change in disposition of the DE loop of CD8α (Figure 5A) is not directly related to interaction with H-2Dd). These apparent adjustments of the loops of CD8β and of CD8α suggest that mobility of these loops permits a degree of “adaptive fit” to facilitate the interaction of the CD8 heterodimer with the MHC α3 domain. Changes in H-2Dd in the free and bound states are noted below.
Although the structure of unliganded H-2Dd has been determined previously (46, 67), for more precise comparison with the CD8αβ/H-2Dd complex we have determined the structure of the H-2Dd/mβ2m/P18I10 complex to 1.7 Å resolution (see Table I). Inspection of the 222–228 loop of the α3 domain of H-2Dd of the CD8-bound and free forms reveals mobility, particularly of the side chains of Gln226 and Glu227 (Figure 4C), which adapt to the pocket formed by the CD8αβ CDRs. Comparison of the hinge angle between the α1α2 domain platform and the α3 domain of H-2Dd in the CD8αβ bound state with the high resolution H-2Dd structure reveals slight movement of the α3 domain away from the platform domain by 6°, resulting in a larger angle for the bound versus free form of H-2Dd (76° and 70° respectively). This angle in the CD8αβ/H-2Dd complex is similar to that of H-2Dd bound to the Ly49A NK receptor (hinge angle of 77°), which binds in a similar region (55). As expected, interaction of CD8αβ with H-2Dd has essentially no effect on the conformation of the α1α2 domain and the bound peptide. (Superposition of this region has an rmsd of 0.44 Å for 189 superposed Cα atoms).
The crystal structure of the CD8αβ/H-2Dd complex now enables rationalization of extensive mutational analyses of the relative contributions of residues of CD8α and CD8β to binding and coreceptor functions (see Table IV). The structure emphasizes the critical role of the CD8β and CD8α CDR3 loops, both clamping down on the finger-like protrusion of H-2Dd residue Gln226 (Figure 2C, Figure 4, and Table II). Different laboratories have employed distinct assays of CD8/MHCI interactions: MHCI/peptide tetramer staining (65, 68), antigen-specific T cell hybridoma activation (32, 44, 49), and alloreactive and antigen specific T cell activation (13, 36, 37, 39, 69). Many of the effects of mutations observed in these assays can be explained either by previously reported CD8αα/MHCI structures or by the CD8αβ/MHCI structure reported here (Table IV). Mutations of CD8 and MHCI and polymorphisms of MHCI have been studied extensively. A number of CD8α mutants have been examined--those that decrease binding of transfectants by TL/β2m tetramers can be readily explained as mutants that affect contact residues to either CD8α1 or CD8α2 subunits (see Table IV). Although several mutations of CD8α that decrease binding by H-2Kb tetramers (Asn107Ala, Lys62Glu, Ser31Leu or Ala, Arg8Asp or Ala) can be rationalized because these are contact residues to either MHC or β2m, several others (Thr81Ala, Leu29Ala, and Lys12Glu) cannot be easily explained, although Thr81 resides at the edge of the H-2Kb interface with CD8α1. More interesting and also not easy to explain is the apparent augmentation of binding observed with the CD8α Lys73Ala mutant. This side chain is exposed to solvent on the backside C′ C″ loop, and thus cannot directly influence MHC interaction. Functional effects of Arg8Ala and Glu27Ala substitutions can be explained as these CD8α residues (in the CD8α1 subunit) contact β2m. (The residues of β2m involved are Lys58 and Asp59. Lys58 is conserved in the human β2m used for some tetramers, and Asp59 is substituted by Asn (in human β2m) which preserves size and hydrogen bonding ability).
A number of CD8β mutants, designed primarily because of their location in CDR loops of CD8β, have been examined both in tetramer binding (45) and in T cell hybridoma stimulation assays (44, 49). Mutants of CDR3β (Ser101Ala, Val99Arg, Pro102Ala) that diminish binding or functional activity are readily explained because the parental side chain interacts with MHC residues at or near the conserved MHCI Gln226 focus. The Oγ atom of Ser101 forms a hydrogen bond with the carbonyl oxygen of Thr225 of the H-2Dd α3 domain (Table II), an interaction that is eliminated by the Ser101Ala substitution. The CD8β Pro102Ala mutation, of a residue that contacts Gln226, also abrogated both CD8 coreceptor and binding activity (45, 49, 70). Mutation of the adjacent Lys103 to Ala reduced but did not abolish the activity of CD8αβ (45). Mutants of CDR3β Lys103, to either Asp or Ala, can be explained despite the lack of direct contact of the Lys sidechain with H-2Dd. The Lys103 sidechain is positioned to make a long range ion pair with CD8α Asp66 (Lys103/Nζ is 4.0 Å from Asp66/Oδ1). Indeed, in the structure of the unliganded mouse CD8αβ (49) this bridge is shorter, and involves both Asp66 carboxylate oxygens. Thus, we may speculate that CD8β Lys103 plays an important role in stabilization of the CD8αβ heterodimer. The CD8β Val99Arg mutation reduced staining in a tetramer-binding assay, a result that may be due to conformational effects on the CD8β CDR3 loop resulting from introduction of the long Arg side chain (45). CD8β CDR2 mutants have varied effects (44, 45, 49), perhaps because the CDR2 contacts to H-2Dd are more peripheral to the Gln226 focus. CD8β Lys55Asp and mutation of Gly56, Lys55, Ser54 and Ser53 to Ala all result in decreased tetramer binding or reduced T cell hybridoma stimulation (see Table IV). At first glance all of these are difficult to explain based on the structure of the complex, but on closer scrutiny we note that Lys55, which does not contact H-2Dd, clearly interacts with CD8α Ser108. Similar to the role of CD8β Lys103, Lys55 is involved in a heterodimer interdomain interaction, suggesting again that dimer stability is important in CD8αβ function. Point mutations of CD8β CDR1 have little or no effect, although the single CD8β CDR1 contact residue observed in the structure, Lys27, was not tested directly (45, 49). However, in the CD8αβ heterodimer interface this residue forms hydrogen bonds with Ser108 of CD8α. Thus, its mutation may destabilize the CD8αβ heterodimer. Some of the interactions of CD8α Ser108 with the CD8 β chain are illustrated in Supplemental Figure 2. Other CD8β mutants of residues not involved in MHC binding that have significant functional effects can also be explained by their role in disruption of heterodimer stability (Table IV).
Several mutants of CD8 are known to improve CD8αβ dependent T cell activation or binding to tetramers. In particular, CD8α Lys73Ala and CD8β Leu58Arg and Ser53Leu augmented binding of H-2Kb tetramers (45, 68). Perhaps elimination of the Lys73 to Asn90 hydrogen bond provided greater flexibility, allowing better accommodation of interaction. Leu58 is a contact residue to CD8α Ser108, and one might consider that the substitution by Arg would stabilize the heterodimer. CD8β Ser53Leu substitution may contribute to stabilization of the CD8β CDR2 loop. Recent studies of engineered TCR indicate that improvement of TCR αβ heterodimer stability by introduction of subunit bridging disulfide bonds contributes to the improvement of both expression and biological activity (71, 72). Our interpretation of changes of activity of CD8α and β mutants in the context of the structure of the CD8αβ/MHCI complex suggests that other mutations that might stabilize the CD8αβ heterodimer (such as introduction of interdomain salt bridges or disulfide bonds) might also lead to improved MHC binding and accessory function.
Examination of structure based amino acid sequence alignments of CD8α, CD8β, and the MHCI α3 domain offers additional insight into the conservation of the structure of the complex in other species (Figure 6). N-linked carbohydrate addition sites of both CD8α and β do not impinge on the interface with MHCI. Few contact residues of CD8β are particularly polymorphic, with the exception of Lys27, which is preserved in rodents. A major contact loop of CD8β, consisting of residues 99 to 102 (CDR3, FG loop), is highly conserved. Considerable effort has been expended to understand the molecular basis for the apparent higher affinity of the MHCI-like TL molecule for CD8αα as compared with CD8αβ (81). Examination of the CD8αβ/H-2Dd structure in comparison with CD8αα/TL and inspection of MHCI α3 domain sequences (Figure 6) suggest that the substitution of His in TL for Asp212 in H-2Dd, a residue that contacts Lys27 of CD8β, may play a significant role. Additional experiments will be needed to address this issue.
The structure reported here defines a single orientation of CD8αβ binding to MHCI, in which the β subunit occupies the “upper,” T cell proximal, CD8α1-equivalent position. Consideration of surface electrostatic interactions of CD8αα with HLA-A2 led to the suggestion that the CD8β subunit occupies the “lower”, T cell distal, CD8α2-equivalent, position (42) and mutagenesis data suggested a dual orientation model, in which the CD8β subunit can dynamically alternate between the “upper” and “lower” positions (44, 45). Although it is difficult to formally eliminate the dual conformation model, we emphasize that crystallographic capture of a single orientation CD8β in the T cell proximal position, that is consistent with most of the existing mutagenesis data argues strongly against such a model. Moreover, to our knowledge there is no precedent in the extensive literature on protein:protein interactions for a heterodimeric receptor that binds in dual, inverse orientations to the same ligand. Because of the sequence similarities of murine and human MHCI molecules as well as the similarities among species of CD8α and of CD8β (Figure 6), we expect that the domain relationships of the murine structure we report here are preserved in the CD8αβ/MHCI complexes of other species. Additional perspective gained from consideration of relative sizes of the subunits of a complete TCR/MHCI/CD8 complex may be gathered from the superposition of the structure of a TCR/H-2Dd complex (KN and DHM, unpublished) onto the CD8αβ/H-2Dd structure reported here (Figure 7). Although the role of CD8αβ as a T cell coreceptor dictates its “trans” interaction with the peptide-binding MHCI molecule on the APC, we note that the structure of the CD8αβ/MHCI complex does not eliminate the possibility of a “cis” interaction between CD8αβ and MHCI expressed on the T cell (73-76). Our structural analysis is consistent with the view that the shorter stalk of CD8β plays a crucial role in the orientation of the cytoplasmic domains of the CD8αβ heterodimer for their role in signal transduction (33). The structure of this murine CD8αβ/MHCI complex may serve as a guide to a mechanistic understanding of human CD8 mutants that result in immunodeficiency (77, 78) and provide a context for further examination of the role of the distinct domains of the CD8 molecule in binding and function.
We thank Zhongmin Jin of SER-CAT of the Advanced Photon Source at Argonne National Laboratory and Howard Robinson at beamline X29 at the National Synchrotron Light Source, Brookhaven, for data collection, and John E. Coligan and Sam Xiao for their comments on the manuscript.
1This research was supported by the Intramural Research Program of the NIAID, National Institutes of Health.
The authors have no financial conflict of interest.