he first structure of a major histo- Tcompatibility complex (MHC) class Structural Immunology of MHC Class I Proteins, Ia protein, HLA-A2, had a significant impact on the immunological com- Homologs and Complexes munity and answered many of the fundamental questions of how are presented to the immune Roland K. Strong system (1). Many of the details of the recognition process have since been Div. of Basic Sciences, Fred Hutchinson Cancer Research Ctr., Seattle, WA 98109, USA elucidated, in part, by a plethora of subsequent structures of human and murine class I protein/peptide and class in the absence of an active infection. a small cytoplasmic domain. The a1 I protein/peptide/αβ receptor MHC class I/peptide complexes are and a2 domains together comprise the (TCR) complexes (2). In structural de- recognized by circulating cytotoxic T peptide- and TCR-binding “platform” terminations and sequence analyses, the (CTLs) through direct domain (Fig. 3). The distinctive MHC class I fold has been found interactions with -specific αβ topology of the platform consists of in proteins involved in a variety of TCRs and the co-receptor CD8, re- two long α-helices overlying an eight- biological processes, playing peripheral sulting in the elimination of infected stranded anti-parallel β-sheet. The two helices form the walls of the Roland K. Strong is Associate Member of the Div. of Basic peptide-binding groove. This groove Sciences, Fred Hutchinson Cancer Research Center. He worked as a Graduate student in the labs of Stephen C. Harrison, encompasses a series of pockets for the Harvard Univ. and of Gregory A. Petsko, MIT & Brandeis Univ. N- and C-termini of the peptide As Consultant in Payload Inc., Cambridge, he participated in and the side chains of the “anchor” efforts to assess the effect of microgravity on the crystallization residues, thus determining the peptide of biological macromolecules over the course of three missions specificity of a particular allele; exposed to the Russian MIR space station. He reached his PhD in the peptide side-chains and specific peptide lab of Pamela J. Bjorkman, Caltech and was Affiliate Assistant main-chain and peptide-induced MHC Professor of the Dept. of Immunology, Univ. of Washington. protein conformations are read out Research interests: Structural immunology - the analysis of proteins mediating by the TCR (Fig. 2 & 4). β -m 2 mucosal immune responses through molecular biology, solution biochemistry and x-ray and the heavy chain α3 domain fold crystallography. http://fhcr.org/labs/strong/ [email protected] into constant-type immunoglobulin roles in immune responses or even cells from the body (Fig. 2). The heavy domains. Association with β -m and 2 fulfilling functions uninvolved in im- chain of these proteins comprises three peptide is required for proper folding munology, interacting with a range extracellular domains (α1, α2 and α3), and cell-surface expression, although of different small-molecule and macro- a transmembrane-spanning domain and either or both of these requirements are molecular ligands (Table 1). Recently determined structures of the murine Table 1: Funcional roles of MHC class I proteins and homologs MHC class Ia protein H-2Dd in com- plex with its receptor Ly49A (3), the Molecule Ligand Receptor Function b murine MHC class Ib protein H-2T22 MHC Class Ia Endogenous αβ TCRs / CD8 Immune (4), the unusual MHC class I homolog (HLA-A, B & C) Surveillance zinc-α -glycoprotein (ZAG) (5), the 2 human natural killer (NK) cell and MHC Class Ib Class Ia Leader NKG2 / CD94 Check for MHC T cell target MIC-A (6) and the (HLA-E) Peptides Class Ia complex between HFE and transferrin Expression receptor (7) reveal the structural and MHC Class Ib N-formylated (on CD8+ T cells) Immune functional extremes that this fold family (H-2M3) Peptides Surveillance encompasses (Fig. 1). CD1a Bacterial Cell Wall αβ / γδ TCRs Immune • MHC class I proteins: fold and Components Surveillance function H-2T22b None? γδ TCRs Innate The human “classical” or class Ia MHC Immunity proteins (HLA-A, B and C) are cell surface, heterodimeric glycoproteins MIC-A None? NKG2D / ? Innate consisting of an integral-membrane Immunity heavy chain and a soluble light chain, FcRn IgG Not Applicable trans-Intestinal known as β -microglobulin (β -m) (8). IgG Transport 2 2 During infection, proteins from patho- HFE Transferrin Not Applicable Iron Metabolism gens are processed into peptides and Receptor packaged into MHC class Ia proteins for presentation on the cell surface; ZAG Hydrophobic and ? Fat Metabolism peptides from self proteins are presented Non-peptidic?

page Fig. 1: Representation of several cell- surface interactions mediated by MHC class I homologs. The crystal structures of the complex between dimeric transferrin receptor and HFE, the complex between two Ly49A dimers and H-2Dd, MIC-A and H-2T22b are shown to scale. The MHC class I proteins and Ly49A are shown as backbone ribbons with secondary structure elements indicated; transferrin receptor (red and orange) and the bound peptide (colored by atom type) in H-2Dd are shown in CPK representations. β2-microglobulin is colored blue. Presumed connections to transmembrane anchors are shown as dashed lines. The flexible loops in H-2T22b (red), and the disordered loop in MIC-A (red spheres) are shown. Figures were generated with SwissPDBViewer and rendered with POV-RAY3. dispensable in some members of the contacts with the peptide, and with the for crystallization, suggesting that carbo- class I family. Va domain overlying the α2 domain hydrate may contribute significantly to Together with the “non-clas- helix and the Vβ domain contacting Ly49A binding - even though Ly49A sical” or class Ib MHC proteins (HLA-E, the α1:H2 helix (Fig. 3). Fc binds to lacks several of the features thought F and G in humans) and their close, but FcRn in a pH dependent manner near to mark carbohydrate-binding CTLDs, distinct, structural relatives, the MHC the ends of the α1:H2 and the α2:H1 particularly two tightly-bound calcium class II proteins (DR, DP and DQ in helices, the edges of the b-strands of the ions. This first site represents yet humans), these proteins fill essential α2 domain, and residues of β -m near another binding surface on the MHC 2 functions in the by the N-terminus. class I fold, on the opposite side of presenting various types of peptides to the platform from the Fc binding site various immune system cells in various • MHC class I proteins and NK on FcRn. The second site, which compartments. In another variation cell receptors on this theme, the non-MHC encoded H-2Dd is a murine class Ia CD1 family of class I homologs present that is recognized by inhibitory not peptides, but mycobacterial lipids receptors on NK cells, a com- to T cells (9,10). While recognizably ponent of the innate immune class I-like, the structure of the binding system that destroys cells that pocket is altered to accommodate this fail to express class Ia proteins type of ligand (Fig. 4) (10). on their surfaces. Some patho- gens, such as cytomegalovirus • Functional diversity in MHC (CMV), have developed clever class I homologs strategies for down-regulating An early hint of things to come with the expression of class Ia pro- the description of an MHC class I-like teins, thus evading αβ T cell- (FcRn) which transports mediated destruction – only IgG across the intestinal epithelium of to be caught by NK cells. nursing neonates (11). The structure Dd is recognized through the of the FcRn/Fc complex showed a non- NK cell receptor Ly49A, a peptide binding MHC class I homolog member of the growing family using a different site to interact with of proteins that contain C-type ligand, a surface distinct from either lectin domains (CTLDs) (13). the TCR or CD8 binding sites (12): Two Ly49A homodimers bind αβ TCRs bind to the “top” of a class I to one Dd molecule in the molecule, covering much of the bound crystal (3). The first binding peptide; CD8 primarily binds the α3 site is not particularly extensive, domain but also contacts α2 and β -m but overlaps the position of 2 (Fig. 2). The TCR sits diagonally on the a conserved N-linked sugar, “top” of the platform making extensive absent from the protein used

Fig. 2: Footprints of protein ligands on several MHC class I molecules. CPK representations of the structures of HLA-A2, H-2Dd, FcRn, HFE, H-2T22b and MIC-A are shown. Bound peptides (red) and β2-microglobulin (dark gray) are shown when present. Ribbon representations of the fold of HLA-A2 are shown at the top of the figure to orient the views in each column; the fold of MIC-A is shown at the bottom of the figure in a ribbon representation to orient the view of MIC-A. Residues defining the ligand footprints (blue and purple) are labeled by molecule for HLA-A2, H-2Dd, FcRn and HFE (TfR: transferrin receptor); residues defining putative ligand binding sites in H-2T22b and MIC-A are similarly indicated. Fig. 3: Superpositions of the platform domains of HLA-A2 (red), MIC-A (blue) and H-2T22b (purple). A ribbon representation of HLA-A2 alone is shown to orient the views; secondary structure elements are indicated and the bound peptide is shown in ball-and- stick style and colored by atom type. This view looks down onto the top of the platform domain, the surface recognized by a TCR. In this orientation, the α1 domain is to the upper left and the α2 domain is at the lower right in each panel. The flexible loops in H-2T22b are colored yellow; the disordered loop in MIC-A is represented by spheres (red) connected by dotted lines to indicate the length of this segment.

complex is not yet ween binding and non-binding alleles available, mapping of of T22, and the closely related T10 allelic and species se- molecules, have been used to delineate overlaps the CD8 binding site, while quence conservations onto the MIC-A possible receptor binding sites, one of more extensive and intimate, may be a structure suggests that NKG2D may which lies in the most distorted part cis interaction between Dd and Ly49A bind to the underside of the platform of the T22 platform in the α2 domain on the surface of the NK cell, possibly domain (Fig 2.). Approximately half (Fig. 2). With only part of the structure playing a role in positioning the receptor of this surface corresponds to the β -m of a γδ TCR currently available (14), 2 for target binding. Ly49A homodimers binding site in class I proteins that the structure of a γδ/H-2T22b complex are able to crosslink Dd molecules associate with β -m. A receptor complex would be an important advance in 2 through the interactions seen in the structure will be necessary to sort out studying the function of these receptors, crystal, and do not occlude the TCR the details of this important interaction, since the radical changes seen in the binding site, both mechanisms possibly and determine if and how carbohydrate structure of the H-2T22b platform, contributing to signaling. Leading is involved. and differences in the complementarity questions raised by this structure include determining regions (CDRs) of the how Ly49A and other members of • MHC class I proteins and γδ T different classes of TCRs (15), almost the family distinguish between cell receptors guarantee that αβ and γδ TCRs recog- different murine MHC class Ia alleles The structure of the non-peptide bin- nize their ligands in distinct manners. and the details of the carbohydrate/ ding, murine MHC class Ib protein, Ly49A interaction. H-2T22b, which is recognized by a • MHC Class I Proteins and MIC-A is recognized by NK subset of γδ TCRs, reveals the most Metabolism cells and T cells through NKG2D, distorted example of an MHC class I Mutations in the HFE gene are re- which, unlike the Ly49A/Dd signal, platform fold yet seen, even though sponsible for the iron absorption stimulates, rather than inhibits, cytolysis. this molecule interacts with β -m in disorder hereditary hemochromatosis 2 MIC-A binds neither peptides, peptide the usual manner (4). surrogates (such as lipids) nor β -m. The distortions are partly 2 NKG2D is the most distantly related the result of deletions in member of the human NKG2 family of the sequence of H-2 CTLD-containing NK receptors, which T22b. Two loops, corre- have low sequence similarity to the sponding to the a1:H1 Ly49 family. The MIC-A platform and part of the α1:H2 differs dramatically from an MHC class helices, and the α2:H1 Ia platform, particularly in the loops at helix, are conformatio- the edges of the α1 domain and the nally flexible, dramati- apparent disordering of the α2:H2α cally demonstrated by helix into a flexible loop that is not comparisons of the four visible in the crystal structure (6). The molecules in the asym- platform and α3 domains are also metric unit of the crystal. flexibly linked. MIC-A and H-2Dd are The remainder of the structurally dissimilar in the regions of α1:H2 helix curves into the Ly49A binding sites, so it is very what would be the pep- unlikely that NKG2D binds MIC-A in tide-binding groove in a manner analogous to the Dd/Ly49A a previously unobserved interaction. Even though a receptor manner. Differences bet-

Fig. 4: Molecular surfaces of MHC class I proteins and homologs, colored by atom type (carbon: gray; oxygen: red; nitrogen: blue; sulfur: yellow). When present, bound peptides are shown in stick representa- tion colored by atom type. The outline of the windows into well defined binding pockets are shown in green when present. The orientation of the molecules is the same as in Fig. 3. (HH). HFE is a non-peptide binding interaction, suggesting that other mecha- fold can be suborned into a binding MHC class I homolog that binds β -m nisms, possibly interactions with other site for either a broad range of proteins, 2 and affects iron metabolism by binding molecules, may be affected in some smaller molecules, or both. Rather than transferrin receptor (TfR) in a pH forms of HH. recapitulating a single structural theme, dependent manner reminiscent of FcRn. ZAG is a soluble, secreted these recent structures demonstrate that Comparisons of the structure of TfR protein present in most body fluids and many of the key elements of the in the presence (7) or absence (16) has been found to accumulate in breast basic class I fold are mutable to yield of HFE reveal conformational changes cysts and some breast carcinomas. ZAG molecules with dramatically different that provide a possible explanation for has been associated with the catabolism functions. What surprises might be altering the affinity for transferrin of lipids and cachexia, though through revealed by structural studies of other versus HFE, which likely bind to an, as yet, uncharacterized mechanism. divergent MHC class I homologs, overlapping sites on TfR. A TfR ZAG binds neither β -m nor endo- such as the CMV protein UL18 (17), 2 homodimer binds to the top of two genously-expressed peptides. The crys- or functionally divergent interactions, HFE platforms, one on each chain of tal structure of ZAG revealed a more such as the interaction between class the TfR dimer, overlapping the αβ TCR- extensive interaction between the plat- I proteins and insulin receptor (18), binding surface in class Ia proteins (Fig. form and α3 domains, which has been remain to be seen. 2). However, the HFE-TfR interaction forwarded as the likeliest explanation surface consists of a three-helix bundle, accounting for the stable expression • REFERENCES two helices contributed from TfR and and of this molecule in the 1. Bjorkman PJ et al. Nature, 329, 506, 1987 one helix from the HFE α1 domain, absence of β -m association (5). ZAG 2. Maenaka K et al. Curr Opin Struct Biol, 9, 745, 1999 2 3. Tormo J et al. Nature, 402, 623, 1999 an interaction surface very unlike the also retains an open binding groove 4. Wingren C et al. Science, 287, 310, 2000 CDR loops in αβ TCRs. TfR is corresponding to the peptide binding 5. Sanchez LM et al. Science, 283, 1914, 1999 6. Li P et al. Immunity, 10, 577, 1999 also positioned over the “peak” formed groove of classical class I proteins, 7. Bennett MJ et al. Nature, 403, 46, 2000 by the bend between the α2:H1 and although with a significantly altered 8. Bjorkman PJ et al. Ann Rev Biochem, 90, 253 1990 α2:H2a helices, a feature that αβ TCRs shape. Electron density for a hydro- 9. Porcelli SA et al. Immunol Today, 19, 362, 1998 10. Zeng, Z.-H., et al., Science, 277, 339, 1997 avoid by their diagonal interaction with phobic, non-peptidic small-molecule 11. Simister NE et al. Eur J Immunol, 15, 733, 1985 class Ia proteins. The resulting packing ligand was visible in the ZAG groove. 12. Burmeister WP et al. Nature, 372, 379, 1994 of helices at the interface is unlike It remains to be seen if either further 13. Drickamer K, Curr Opin Struct Biol, 9, 585, 1999 14. Li H et al. Nature, 391, 502, 1998 coiled-coil interfaces seen other proteins. structural or biochemical analyses can 15. Davis MM et al. Immunol Today, 16, 316, 1995 The complex structure also suggests pin down the identity of this com- 16. Lawrence CM et al. Science, 286, 779, 1999 that pH-mediated changes in complex pound. 17. Fahnestock ML et al. Immunity, 3, 583, 1995 formation are the result of confor- 18. Verland S et al. J Immunol, 143, 945, 1989 mational changes in TfR sensed by • Future directions HFE. Interestingly, one HH mutation Structures of MHC class I homologs (His41) is not involved in the TfR illustrate that most of the surface of this