A Region of Conformational Variability Outside the Peptide-Binding Site of a Class I MHC Molecule

The elucidation of the static structures of several class I molecules, as well as two complexed with the TCR, have provided understanding of the molecular nature of immune recognition (1, 2, 3, 4). There remains the issue of how the conformation of the players in these interactions might undergo structural change during the formation of different complexes. While there have been numerous studies designed to look at the extent to which bound peptide or mutations in the α1 and α2 domains affect the overall conformation of the Ag-presentation domain of the complex, it is still unclear whether such changes may have structural consequences in the other regions of the molecule, or, conversely, the effect that conformational changes in α3 or β₂-microglobulin (β₂m)² have on the conformation of the Ag-presentation domain. A recent report describing the crystal structure of the complex between human CD8 and HLA-A2 finds that the position of the α3 domain is different from that in uncomplexed HLA-A2 (5). Taken together with the finding by Garcia et al. that the presence of CD8 enhances the formation of stable TCR/MHC complexes (6), these observations suggest that these shifts might occur to accommodate a higher-affinity interaction.

The present study describes features of the mouse class I molecule, K^b, which are important for the formation of the conformationally sensitive epitope of the mAb, AF6-88.5, produced by Loken and Stall (7). We find that even relatively subtle changes in any of the extracellular domains of the class I molecule, excluding the peptide-binding site, have an effect on the epitope recognized by AF6-88.5. This indicates that conformational perturbations that originate in one domain can radiate to affect the three dimensional structure of the other domains of the complex. Conformational plasticity provides a mechanism for transmitting signals along the class I molecule as a consequence of being bound by its ligands and receptors. While conventional views focus on the signals transmitted by the TCR and CD8 as a consequence of class I binding, we entertain the possibility that signals emanate from the class I molecule as a result of conformational changes induced by interactions with the TCR, CD8, or other ligands/receptors.

The B10.M/Sn, B10.S/SgMcdJ, BALB/cByJ, and B10.Q/SgJ mice, as well as the β₂m knockout and human β₂m (hβ₂m) transgenic mice used in this study were obtained from the colony of Chella David (Mayo Clinic, Rochester, MN); the C.B10-H2^b (BALB.B) were obtained from Peter Wettstein (Mayo Clinic, Rochester, MN); the C57BL/6J and C57BL/6J-H-2^bm3/Eg mice were obtained from The Jackson Laboratories (Bar Harbor, ME); the K^bm3/D^d transgenic mice as well as the K^b/L^d transgenic mice were produced in this laboratory in collaboration with the David Laboratory for unrelated studies; the TAP knockout mice (129/OLA TAP1) were obtained from Dr. Anton Berns (The Netherlands Cancer Institute, Amsterdam, Netherlands). The mice were maintained in the animal facility at the Mayo Clinic.

The R8 cells and its variants, which were a gift from Dr. Stanley Nathanson (Albert Einstein College of Medicine, Bronx, NY), have been described previously (8). The cell lines, which harbor the chimeric class I molecules, were generated by Ken Arakawa in this laboratory or described previously (9). Briefly, the mouse fibroblast L cells were transfected by CaPO₄ precipitation of the construct, along with a thymidine kinase gene, onto the cells. The cells were cultured in RPMI 1640 complete medium, and transfectants were selected for viability with hypoxanthine/aminopterin/thymidine.

The origin of the mAbs AF6-88.5, B8-24-3, K10.56, 28-13-3, and 64-3-7 have been listed previously (7, 10) and were used at binding-site-saturating concentrations. Detection of Abs was done using an FITC-conjugated goat anti-mouse IgG or IgM (Biosource International, Camarillo, CA). Flow cytometry was performed on a FACScan (Becton Dickinson, Mountain View, CA), and mean channel fluorescence values were obtained and converted to a linear scale.

AF6-88.5 reactivity of variant and chimeric class I molecules was first determined relative to an unaffected epitope, as indicated in the tables and figures, and that proportion was compared with the proportion found on the wild-type K^b molecule expressed in comparable cell types (i.e., freshly isolated spleen cells, transfected L cells, or the R8 lymphoma line). The values reported is the ratio of the percent of AF6-88.5 binding to the variant molecule relative to the percent of AF6-88.5 binding to the wild-type molecule. Each binding experiment was repeated at least three times, and all data reported were reproducible. Representative fluorescence values are shown in Table II.

AF6-88.5 is a mAb with high specificity for the class I allele K^b (7). The Ab was generated by immunizing a BALB/c mouse with splenocytes from a C57BL/6 mouse. During tests to confirm the reactivity of the Ab, we found that a mouse L cell line transfected with a chimeric class I molecule composed of the α1 and α2 domains of K^b, and the α3, transmembrane, and cytoplasmic domains of L^d failed to bind AF6-88.5 in a FACS assay. Reactivity against K^b was confirmed by showing that L cells transfected with genomic clones of K^b were positive, while transfectants of L^d were negative (Fig. 1 and Table I). Furthermore, splenocytes from a B6 mouse reacted positively with the Ab, while splenocytes from a B10.M transgenic mouse expressing the chimeric K^b/L^d molecule did not (data not shown).

Localizing the AF6-88.5 epitope to the α3 domain was unanticipated, as there is no allelic cross-reactivity reported for this Ab by the manufacturer or in the literature, and polymorphism in the α3 region is limited. Additionally, the epitope recognized by this Ab had been previously reported to reside in the α2 region of K^b (11), although no mapping data is presented. As there are nine amino acid residue differences between K^b and L^d in their α3 domains (Ref. 12, Fig. 3), we evaluated whether the epitope for AF6-88.5 might lie in the α3 domain of K^b. One prediction of this hypothesis is that a class I molecule with an identical α3 sequence would be able to bind AF6-88.5. K^q is the allele most closely related to K^b in the α3 domain by sequence, with differences at position 225 (Ile instead of Thr) and at position 268 (Glu instead of Lys) (12). Splenocytes from a B10.Q mouse failed to bind AF6-88.5 in a FACS assay. B10.S (H-2s) splenocytes expressing class I molecules that differ from the α3 sequence of K^b at additional positions than K^q also did not bind AF6-88.5 (data not shown). Examination of the three-dimensional structure of K^b shows that one of the nine amino acids different between K^b and L^d, position 264, lies in relatively close proximity to the α2 domain. We performed site-directed mutagenesis on the L^d α3 exon to change residue 264 to glutamic acid (the amino acid present at that position in K^b) to see if it participated in the formation of the AF6-88.5 epitope. This change in the chimera was found not to be sufficient to restore the AF6-88.5 reactivity (data not shown). To test the hypothesis that AF6-88.5 recognizes a structure determined by the α3 domain of K^b more directly, a chimeric L cell transfectant expressing a class I molecule bearing the α1 and α2 domains of L^d, and the α3, transmembrane, and cytoplasmic domains of K^b were evaluated and found not to bind the Ab (Table I). In an attempt to reconcile the previously published conclusion that AF6-88.5 recognizes an α2 epitope with our findings, a third chimeric class molecules consisting of the α1 domain of L^d and the α2, α3, transmembrane, and cytoplasmic domains of K^b was investigated. This molecule was found to be unstable, but could be monitored with L^d α1 and K^b α2-associated Abs (64-3-7 and K10.56, respectively) and thermal stabilization. Using the K^b α1-specific Ab, B8-24-3, as a negative control, it is shown that this chimera also failed to be bound by AF6-88.5 (Table I, Fig. 2).

The data thus far indicated that the AF6-88.5-defined epitope depended on sequences in the α1 and α3 regions. We next analyzed a panel of K^b mutants derived by in vitro selection of R8 cells with Abs specific for the α1 and α2 regions (8). This series of mutants contain single amino acid substitutions that modify surface-exposed residues on the peptide-binding domain. Each of the variants bound AF6-88.5 in a manner equivalent to the parent with the exception of the variants G162D, which has a slight disruptive effect, and N174K, which enhances binding (Table II). Also, while the five mutations that comprise the K^bm10 variant had no effect on AF6-88.5 binding (Table III and see below), two of the K^bm10 mutations, T163A and W167S, when analyzed individually each showed 150% AF6-88.5 reactivity, while the N174L mutation showed 80% reactivity (data not shown). As these positions, located on the α2 helix, are spatially removed from the α1 and α3 regions of the K^b molecule, it is evident that the AF6-88.5-defined epitope is sensitive to structural perturbations in all three of the extracellular heavy chain domains.

To determine the effects of previously described mutations that influence the structure of the peptide-binding site formed by the α1 and α2 domains of the K^b molecule, the ability of the AF6-88.5 Ab to bind splenocytes from the K^bm3, K^bm5, and K^bm8 mutant mice or L cells expressing transfected K^bm1 and K^bm10 genes were tested. Each of these mutant K^b molecules displays altered peptide-binding properties and TCR recognition. For most cases, that changes can be visualized with α1/α2-specific mAbs (9, 13, and our unpublished observations). None of the mutations had any significant influence on the expression of the AF6-88.5-defined epitope (Table III). Consistent with our previous finding that the α3 region influences AF6-88.5-recognition, cells transfected with a chimeric molecule that is comprised of the α1 and α2 domains of K^bm3 and the α3, transmembrane, and cytoplasmic tail of D^d were negative, as were splenocytes from a mouse expressing the chimeric class I molecule from a transgene (Table III).

To assess the role of peptides within the binding site, spleen cells from mice genetically deficient for the TAP complex were analyzed. Cells from these mice have reduced surface expression of K^b, as the class I molecules are not assembled with peptide within the endoplasmic reticulum (14). Reduced levels of K^b were detected relative to wild-type cells, and the proportion of AF6-88.5 reactivity relative to K10.56 reactivity was roughly half that proportion on wild-type cells (Table IV). Therefore, we concluded that the absence of high-affinity peptide had some effect on the AF6-88.5-defined epitope. A study has been previously reported that found no influence of 40 different OVA-derived peptide analogues on AF6-88.5 binding (11). This indicates that, while the presence of high-affinity peptide has an effect on the AF6-88.5 epitope, the nature of peptides that are bound did not.

We wished to assess whether β₂m participated in the formation of the AF6-88.5 epitope. We determined that K^b complexed to the “b” allele of β₂m had no effect on the AF6-88.5 epitope by analyzing BALB.B splenocytes (Table IV). To determine whether the AF6-88.5-defined epitope could be detected on β₂m deficient cells, spleen cells from a β₂m-deficient mouse expressing K^b were examined. These cells were found to be negative, which was not surprising as these cells do not express any known class I epitopes on their cell surface in appreciable quantities (15). Next, AF6-88.5 recognition was assessed on cells after restoring class I expression with a human β₂m transgene. Despite the presence of K^b on the surface of spleen cells from this mouse as detected by several mAbs, the AF6-88.5-defined epitope was not detected (Table IV). From this, we conclude that mAb AF6-88.5 recognizes an epitope that is dependent on all four extracellular domains of the K^b molecule.

In this report, we have characterized sequences that determine the structure of the binding site of the mAb, AF6-88.5, on the mouse class I molecule, K^b. The impetus for this came from our inability to use AF6-88.5 to visualize K^b on fixed tissue and by the subsequent discovery that the Ab would not bind a K^b chimera containing the α3 domain of L^d. We found it surprising that the nine amino acid differences in the α3 domain, most of which are far away from the Ag-binding domain in the three-dimensional structure (Fig. 3), would influence the binding of an Ab reported to recognize sequences in the α2 region. Furthermore, site-directed mutagenesis showed that the eight distal-most residues from the peptide-binding domain in α3 that differ between K^b and L^d were sufficient to disrupt AF6-88.5 binding. That K^q, as well as all the other chimeras between K^b and L^d, all failed to bind the Ab suggested to us that there must be some kind of interaction among the α1, α2, and α3 domains that determine the structure of the epitope. The epitope forms only when K^b sequences are present in each of the domains and is disrupted by amino acid substitutions in each of the external domains of the protein.

The R8 cell line mutants, as well as the “bm” mutant series, were originally isolated for changes in mAb and TCR-binding properties (8, 16). By definition, this set of mutants has structural divergence from K^b mapping in the α1 and α2 regions. With the exception of G162P and N174K, none had an impact on the AF6-88.5-defined determinant, indicating that the conformational changes they produce are not overlapping with the AF6-88.5-epitope. It should be pointed out that the substitution at position 162 results in a proline at a position that links two helices, and the substitution at 174 results in two adjacent lysine residues. The former might not be predicted to have an effect although it is destabilizing to AF6-88.5, and the latter might be predicted to have a disruptive effect on the structure although it results in stabilization of the structurally sensitive epitope.

Others have found that β₂m has a role in stabilizing the heavy chain, and there have been other mAb described that are dependent on β₂m (17). Indeed, some have found that xenogeneic β₂m has higher affinity for mouse heavy chains and provided for greater stability of the complex than mouse β₂m (18). Recent studies indicate that β₂m plays an essential role in the endoplasmic reticulum as it interacts with the heavy chain and facilitates its proper folding (19). It is also worth noting that numerous murine class I molecules, when complexed with human or bovine β₂m, acquire some affinity for the monomorphic anti-HLA class I mAb, W6/32. This binding appears to associate with the murine α2 domain (20, 21, 22), which would demonstrate an influence of β₂m on murine heavy chain structure. To assess what role β₂m might have in formation of the epitope, we used a β₂m-deficient mouse as well as a mouse β₂m-deficient cell line reconstituted with hβ₂m. We found that the AF6-88.5-defined epitope is dependent on both K^b and mouse β₂m (Table IV).

Thus, all of the major elements of the class I complex influence the AF6-88.5-defined determinant. A report from Chen et al. cites the use of two antigenic OVA peptides and their reciprocally substituted analogues bound to K^b and found that the AF6-88.5 epitope formed regardless of which peptide was used. We were able to recapitulate this finding using RMA-S cells and the dominant OVA peptide (data not shown).

Examination of the three-dimensional structure of K^b reveals a point where all three of the domains of the heavy chain as well as β₂m come together (Fig. 3). We have not been able to define the exact determinant for AF6-88.5 and cannot formally exclude that it is represented by this area. However, we find it unlikely that it represents the center of the Ab-binding site given the apparent inaccessibility of this point. Even if this were the area where this Ab was binding, the positions of the residues that influence binding still force us to conclude that structural perturbations can radiate throughout all the domains of the molecule.

The majority of the efforts made to assess changes in class I conformation have focused on using different peptides bound in the binding cleft or mutants of the heavy chain that directly alter the physical properties of the peptide-binding site (23, 24, 25, 26). These studies fit current paradigms well, as all the structural data reported over the past decade has shown that this is the region where immune recognition takes place. It has therefore been of the most potential value to study the peptide binding region of class I in an attempt to understand the molecular details surrounding class I-TCR interactions.

Here we have shown that the mAb, AF6-88.5, recognizes an epitope on the mouse class I molecule, K^b, which is sensitive to modifications of residues in all parts of the heavy chain as well as to the nature of the β₂m, but is insensitive to the nature of bound peptide. This implies the existence of an additional conformationally sensitive site(s) on class I molecules. The question of whether such changes have biological relevance remains to be answered.

There is increasing evidence that molecular interactions involving regions of the class I molecule outside of the peptide-presenting domain not only play important functions during T cell/APC interactions, but also have a role in the function of class I molecules on intracellular signaling pathways (27, 28, 29). An example of a large change in the positions of α3 and β₂m relative to α1α2 in the receptor-bound structure was observed and comes from the recent description of the high-resolution crystal structure of the entire complex of K^b bound by TCR (30). Another recent example of structural plasticity comes from the analysis of HLA-A2 complexed with CD8 (5). The observation of structural shifts in the α3 domain that accompany binding of CD8 to the class I molecule suggests a potential role for these structural changes in immune recognition. The finding that CD8 enhances the affinity of the class I TCR interaction (6) is supportive of this view. Another demonstration of large-scale structural changes that occur as a consequence of receptor-ligand interaction comes from Reich et al., where they described the oligomerization of TCR/peptide-MHC complexes while such oligomerization did not occur with either molecule by themselves or with a nonantigenic peptide (31).

While the static structures described above point to a functional role for regions of the class I molecule distinct from the TCR interaction domain, an example showing significance for such regions on molecules in a live membrane comes form the observation that Abs specific for the α3 domain have been shown to inhibit signals transduced through the TCR (32). That the cytoplasmic domain of the class I molecule is not required for this function implies that intermolecular interactions occurring between other membrane molecules and the extracellular domains of class I may be important (33). Whether activation of these negative signals by Abs is the result of conformational changes or cross-linking is not known.

There are now a large number of crystal structures of class I molecules that have been determined, and α3 and β₂m are the areas which show the greatest deviation from structure to structure. While this fact makes it difficult to conclude that such structural alterations are a consequence of receptor or ligand binding, it also indicates that these areas are likely to be capable of such fluctuations at the cell surface and that makes them attractive candidates for mediating other intermolecular interactions. Our finding that the AF6-88.5 mAb recognizes a conformationally sensitive epitope outside of the peptide-binding site of class I not only indicates that class I is capable of considerable structural fluidity, but that AF6 could serve as a valuable tool in exploring situations that may bring about a conformational shift in K^b.

We thank Michael Hansen and Kathleen Allen for their excellent technical assistance.