- Email: [email protected]
Antigen Presentation: Kissing cousins exchange CLIP
FRANCES SANDERSON AND JOHN TROWSDALE
FRANCES SANDERSON AND JOHN TROWSDALE
ANTIGEN PRESENTATION
ANTIGEN PRESENTATION
Kissing cousins exchange CLIP HLA-DM, a recent addition to the immunoglobulin family, isstructurally most like class IImolecules of the MHC. DM facilitates exchange of CLIP, a temporary filler of the groove of class IImolecules, for antigenic peptides. A central event in the development of a specific immune response is recognition by the T-cell receptor of an antigenic peptide bound within the groove of a class I or class II molecule of the major histocompatibility complex (MHC) [1]. Class I molecules handle peptides from proteins synthesized within the cell, such as viral proteins, which they present to lymphocytes bearing the cell-surface marker CD8. Class I-binding peptides are generated by a multi-component protease, the proteasome, and are loaded into the groove of the class I molecule after translocation into the endoplasmic reticulum (ER). Class II molecules, on the other hand, are responsible for presenting peptides derived from extracellular pathogens to the CD4-bearing subset of T cells.
shown that the CLIP region is both necessary and sufficient for the blocking and stabilizing effects of intact Ii on MHC class II molecules. Two observations suggest that the binding of CLIP to HLA-DR is different from the binding of antigenic peptides. First, certain antibodies (such as the monoclonal 16.23) that normally recognize cell-surface class 1I molecules are unable to bind the 1-CLIP complex. Second, although mature class II-peptide complexes are characteristically stable in the detergent SDS, most
The human class II molecules - HLA-DP, -DQ and -DR, of which HLA-DR is the prototype (Fig. 1) - are heterodimers of two immunoglobulin (Ig) superfamily protein chains, ax and 13, the genes for which reside together in the MHC. Upon synthesis, class II a13 heterodimers associate with the invariant chain, Ii, a type II (inverted) membrane protein with which they are exported from the ER in a nonameric complex. Unlike MHC class I molecules, newly synthesized class II molecules do not follow the constitutive secretory pathway. Instead, to allow them to capture peptides from external sources, they are guided by targeting signals contained in the cytoplasmic tail of Ii onto a pathway that intersects that of endocytic traffic (Fig. 2). This journey to the cell surface takes several hours, during which time Ii is degraded and removed from the class II molecule, which is in turn loaded with peptide in an acidic compartment. Ii has other important functions: it helps assembly of class II a and chains and prevents their aggregation; it also blocks the groove of class II a13 heterodimers, so preventing binding of self peptides in the lumen of the ER. A strong indication as to how Ii fulfils its blocking and stabilizing roles is provided by mutant cell lines in which the processing of antigens to be presented by class II molecules is faulty. Most of the class II molecules on the surfaces of these cells are bound not by the normal spectrum of antigenic peptides but by a nested set of fragments that span residues 81-105 of Ii and are known as CLIP (class II-associated invariant chain peptides). The normal fate of class II-bound Ii is to be attacked by proteases which remove it in sections, systematically. CLIP is the smallest fragment of Ii to be retained, and ua3-CLIP complexes form a natural intermediate in class II maturation in vivo [2]. Genetic truncation and deletion experiments have 1372
Fig. 1. HLA-DM is a distant cousin of class I and class II
molecules of the MHC. Sequence comparisons show that DM may have split from the primordial class I and class IIsequences early on in evolution; the protein structures are shown, highly schematically, below the tree. DM is similar in arrangement to other class II structures, except for the extra cysteine (C)residues in the ul and 31 domains, the amino-acid sequences of which are only weakly related to those of other HLA molecules. DM has no recognizable CD4-binding site.
© Current Biology 1995, Vol 5 No 12
DISPATCH
Fig. 2. Behaviour of DR in DM+ and DM- cells. Peptides in the lumen of the ER are prevented from binding to class II by i. Ii and the a and chains of the class II molecule form a nonameric complex, (1i) 3, for transport, held together by a 'trimerization' sequence in the carboxy-terminal (luminal) region of Ii (see also Fig. 4). Destruction of the trimerization domain of i, at a later stage, as the nonamer moves to the endocytic vesicles under the guidance of amino-terminal Ii signals, may be a key step in the process. DM has its own targeting signal that may enable the molecule to be retrieved from the cell surface [181. DM may also interact with i in the ER. Degradation of endocytosed proteins takes place so that peptides are available to newly synthesized class II molecules as i is released. DM accumulates in the MHC class II compartment (MIIC), where it is thought to catalyze peptide loading onto class II molecules; class II molecules loaded with peptide move to the cell surface. There is evidence for cell-type-specific variations in this pathway, especially in phagocytic cells, as well as for recycling of some class II molecules from the cell surface. The intravesicular pH through the class II trafficking pathway decreases as it does inendocytosis, to lower than pH 5 inthe MIIC. CLIP-bound class II molecules dissociate in SDS into their constituent a and 3 chains. Both findings point to a structure for xot3-CLIP that is different from that adopted by other class II-peptide complexes. In addition, attracting and repelling forces between peptide side chains and the polymorphic residues lining the class II peptide-binding groove result in the specificity of most peptides for a limited range of class II allotypes; in contrast, CLIP binds to most, if not all, class II allotypes. Although it seemed possible that CLIP inhibits peptide binding indirectly, the wide range of binding affinities exhibited by CLIP for different class II molecules, and the allele-dependent manner in which mutations introduced into synthetic CLIP influence binding, has strongly favoured a model in which CLIP occupies the groove [3,4]. Structure determination now confirms this [5]. Ghosh et al [5] have crystallized HLA-DR3 in an ax3-CLIP complex and solved the structure to 2.75 A resolution (Fig. 3). Remarkably, the o4-CLIP complex has a structure almost indistinguishable from that of class II with bound antigenic peptide (specifically, HLA-DR1 with a peptide from influenza virus haemagglutinin, HA 306-318). The CLIP is bound to HLA-DR3 in a conformation and with a network of hydrogen bonds similar to those in the HLA-DR1-HA complex, and with similar
side-chain-pocket interactions. The only conformational difference between the DR3-CLIP and DR1-HA complexes is in the second a-helical region of the DR3 3 chain (residues 65-74), which in the DR3-CLIP complex has moved 1.5 A towards the peptide. This helical section corresponds to the predicted epitope of the conformation-specific monoclonal antibody 16.23 [6], and its displacement could provide an elegant explanation for the recognition by the antibody of mature DR3-peptide but not DR3-CLIP complexes [7]. The reason for the SDS-instability of class II-CLIP complexes (including DR3-CLIP) is less clear. Certainly it can no longer be attributed to any major difference in the io3 portion of the structure. Nor is CLIP more superficially bound: the conservation of hydrogen bonds between the two complexes and the comparability of solvent-accessible surfaces argue that, as predicted from affinity measurements in vitro, DR3-CLIP and DR1-HA complexes will have similar stabilities. In the context of intact Ii, the residues spanned by the CLIP fragment need not necessarily bind within the ot groove nor with the conformation adopted by the isolated peptide. But if they do, the deep interactions demonstrated in the oaL-CLIP crystal structure could be responsible for Ii-mediated stabilization of xa13in much
1373
1374
Current Biology 1995, Vol 5 No 12 of CLIP from DR, as well as facilitating loading with antigenic peptides. Incorporation of DR molecules
from wild-type cells into DM-containing membranes accelerates peptide binding in vitro [15]. Membrane reconstitution is not essential, however: immunoprecipitated DM, purified soluble recombinant DM, and even detergent-solubilized DM-containing membranes have similar effects on DR molecules in solution [13,15]. All three studies have demonstrated that, at acidic pH in vitro, DM promotes the dissociation of CLIP from a-CLIP complexes.
Fig. 3. Side view of CLIP within the HLA-DR3 peptide-binding site [5]. (Graphic courtesy Partho Ghosh).
the same way as stability is conferred by the binding of antigenic peptides. Analysis of the same antigen-processing mutants that led to the identification of CLIP as the universal class II-binding peptide has also identified a function for two novel class II genes, HLA-DMA and HLA-DMB [8]: their products are required for successful CLIP removal from, and peptide loading onto, class II molecules [9,10]. Inspection of the DM protein sequences, as shown schematically in Figure 1, reveals that the DM genes could have split from the primordial class I/II sequences several hundred million years ago, at around the time that class I and class II diverged from each other. The DM u-2 and 32 domains are homologous to the membrane-proximal domains of class I and II molecules; the al and 31 regions, which in DR form the peptide-binding groove, contain little primary sequence homology but can be modelled into a structure similar to the DR archetype (C. Thorpe and P. Travers, personal communication). Unlike the majority of class II heterodimers, little DM is found at the cell surface. Instead, it accumulates together with class II molecules in the specialized late endosomal/lysosomal MHC class II compartments (MIICs) [11]. These membrane-bound vesicles co-migrate on Percoll density gradients with compartments in which immunostimulatory class II molecules first appear [12], and they are the putative site of class II peptide loading. Thus, DM is well placed for its apparent role in expediting the association between antigenic peptide and class II molecules. So how does DM effect peptide exchange? Initially the most attractive theories, given the molecule's class II-like structure, proposed that DM might act either by binding CLIP and removing it from the class II molecule to make way for other peptides -
the 'CLIP sink' model -
or
by binding a range of peptides in endocytic vesicles and bringing them to class II in a peptide-loading compartment. Evidence from recent studies favours a different scheme [13-15]. Rather than participating in a peptideshuttle type of mechanism, it seems that DM acts as a catalyst which at low pH increases the dissociation rate
A requirement in vivo for DM in order to release Iiderived peptides from DR would account not only for the persistence of DR-CLIP complexes in DM-negative mutant cells, but also for the apparent DM-independence of antigen presentation by the mouse allele I-Ak, which stands out among class II molecules as having a remarkably low affinity for CLIP [4]. Much of the effect of HLA-DM on peptide binding appears to be attributable to DM-catalyzed CLIP release, but not all. Denzin and Cresswell [14] compared the efficiencies with which either DM or the detergent octylglucoside produces peptide-loaded DR molecules from DR-CLIP complexes in vitro. Although CLIP is released faster in the presence of detergent, comparable peptide loading requires 100-fold higher concentration of peptide. Thus, DM appears to have a direct effect on peptide loading. To date, there is no evidence that DM binds peptide; in fact, under conditions in which it is clearly able to remove CLIP from anx-CLIP complexes, no DM with bound CLIP was detected [14]. To explain these data, it has been proposed that DM interacts fleetingly with ota-CLIP complexes, promoting CLIP release and then stabilizing a peptide-receptive conformation adopted by the empty oa3 class II molecule [14]. Support for the notion that a physical interaction between the two molecules is required for function comes from work on a DR molecule with a proline-serine substitution at position 96 of the ax chain (DR3Ser 96 ), which introduces an extra N-linked glycosylation site [13]. The phenotype of the DR3Ser 96 mutant in vivo is similar to that of DR in DM-negative cells. In vitro, the altered DR molecule is refractory to the effects of DM on CLIP removal. A similar effect was produced by the pre-incubation of a1-CLIP with an anti-CLIP antibody. It is reasonable to propose that both the bulky carbohydrate group on the mutant DR molecule and the anti-CLIP antibody prevent interaction with DM. So far, direct evidence for association between DR and DM has been difficult to obtain. However, data from our laboratory show a non-covalent DM-DR interaction, under conditions that approximate those of the appropriate endosomal compartment (ES., C. Thomas, J. Neefjes and J.T., unpublished observations). It is odd that DM, the catalyst for DR peptide-loading, is a molecule that shares a structural relationship with its substrate yet does not appear to bind peptide. A precedent can be found in the rat neonatal Fc receptor,
DISPATCH
Fig. 4. Speculative model of the role of DM in catalyzing peptide loading of DR. For details see text.
which has the primary structure of a typical class I molecule, but its binding groove proved to be distorted and empty [16]. A closer inspection of the sequence of HLA-DM suggests that here too the putative groove may be malformed, with bulky hydrophobic residues obstructing it at one end. A speculative model for the function of HLA-DM is therefore as follows. The class II-Ii complex leaves the ER as a nonamer, and it is significant that this complex is not a substrate for exchange mediated by DM [14]. As the trimerization domain is in the section of the Ii chain that is removed first, it seems likely that the dissolution of the (Ii) 3 nonamer takes place upon entering, or just before entering, the MIIC. Which fragment of Ii is removed by DM in vivo is not entirely clear. DM is certainly able to catalyze the dissociation of fragments larger than CLIP in vitro [14], and it is possible (as indicated in Fig. 4) that class II molecules bound to these larger fragments serve as one, or possibly even the principal, substrate for DM in vivo. In this model, DM does not have to bind peptide itself in order to effect peptide exchange on DR. Rather, as mentioned above, the evidence points to a direct interaction between the two MHC-encoded molecules. It is necessary only to postulate a fleeting 'kiss' between DM and DR (Fig. 4); in vitro, in the absence of peptide, unoccupied class II o3 heterodimers become refractory to peptide binding [17] and these may be stabilized by association with DM. This interaction, by a mechanism analogous to that by which an enzyme stabilizes a transition state, may account for both the release of CLIP and the facilitation of peptide binding in vitro. Enzyme-substrate binding is often of low affinity, and certainly the association between DM and DR is difficult to demonstrate. Although the structures of DR1-HA and DR3-CLIP complexes are similar, the structure of the empty ad intermediate is unknown. It is possible that a change in conformation effected by peptide binding is the trigger for the dissociation of DM, which is then free to catalyze another round of peptide exchange.
One consequence of this model, and also one of the most provocative features of DM-catalyzed peptide dissociation in vitro, is that CLIP is not the only peptide to be removed from DR [13]. The half-life of a DR1-myelin basic protein complex was also shortened by DM, and to a comparable extent to that of ot-CLIP complexes. It has been suggested that the DM-catalyzed dissociation of non-CLIP peptides in vivo serves an important function [13]. By favouring the formation of a subset of oa3-peptide complexes with long half-lives, DM may be responsible for the observed selection of a small number of immunodominant epitopes from the array of available class II-binding peptides. Clearly, the next step is to determine whether it is the affinity of the peptide, a specific structural feature of the complex, or some other characteristic of the class II-peptide interaction, that confers susceptibility to DM-mediated exchange. There are many fascinating examples of immunoglobulin superfamily members put to work in a novel ways. But few, like DM, are altruistic: DM helps out its immediate class II family with perhaps no chance of binding a peptide of its own. Acknowledgements: We would like to thank P. Travers, N. Koch and G. Malcherek for comments, and S. Giles for secretarial help. References 1. Germain RN: MHC-dependent antigen processing and peptide presentation: providing ligands for T lymphocyte activation. Cell 1994, 76:287-299. 2. Avva RR, Cresswell P: In vivo and in vitro formation and dissociation of HLA-DR complexes with invariant chain-derived peptides. Immunity 1994, 1:763-774. 3. Malcherek GC,Gnan V, lung G, Rammensee H-G, Melms A: Supermotifs enable natural invariant chain-derived peptides to interact with many major histocompatibility complex-class II molecules. J Exp Med 1995, 181:527-536. 4. Sette A, Southwood S, Miller , Appella E: Binding of major histocompatibility complex class II to the invariant chain-derived peptide, CLIP, is regulated by allelic polymorphisms in class II. J Exp Med 1995, 181:677-683. 5. Ghosh P, Amaya E, Mellins E, Wiley DC: The structure of an
1375
1376
Current Biology 1995, Vol 5 No 12
6. 7.
8. 9. 10.
11.
12. 13.
intermediate in class II MHC maturation. HLA-DR3 complexed with the invariant chain fragment CLIP. Nature 1995, 378:457-462. Marsh SGE, Bodmer JG: HLA-DR and -DQ epitopes and monoclonal antibody specificity. Immunol Today 1989, 10:305-312. Mellins E,Cameron P, Amaya M, Goodman S, Pious D, Smith L, Arp B: A mutant human histocompatibility leukocyte antigen DR molecule associated with invariant chain peptides. I Exp Med 1994, 179:541-549. Kelly AP, Monaco IJ,Cho S, Trowsdale J: A new human HLA class Il-related locus, DM. Nature 1991, 353:571-573. Fling SP, Arp B, Pious D: HLA-DMA and -DMB genes are both required for MHC class 11/peptide complex formation in antigenpresenting cells. Nature 1994, 368:554-558. Morris P, Shaman J, Attaya M, Amaya M, Goodman S, Bergman C, Monaco JJ, Mellins E: An essential role for HLA-DM in antigen presentation by class II major histocompatibility molecules. Nature 1994, 368:551-554. Sanderson F, Kleijmeer MJ, Kelly AP, Verwoerd D, Tulp A, Neefjes JJ, Geuze HI, Trowsdale J:Accumulation of HLA-DM, a regulator of antigen presentation, in MHC class II compartments. Science 1994, 266:1566-1569. Qiu Y, Xu X, Wandinger-Ness A, Dalke DP, Pierce SK: Separation of subcellular compartments containing distinct functional forms of MHC class II. J Cell Biol 1994, 125:595-605. Sloan VS, Cameron P, Porter G, Gammon M, Amaya M, Mellins E,
14. 15. 16. 17.
18.
Zaller DM: Mediation by HLA-DM of dissociation of peptides from HLA-DR. Nature 1995, 375:802 806. Denzin LK, Cresswell P: HLA-DM induces CLIP dissociation from MHC class II alpha beta dimers and facilitates peptide loading. Cell 1995, 82:155-165. Sherman MA, Weber DA, Jensen PE: DM enhances peptide binding to class II MHC by release of invariant chain-derived peptide. Immunity 1995, 3:197-205. Burmeister WP, Gastinel LN, Sinister NE, Blum ML, Bjorkman PJ: Crystal structure at 2.2 A resolution of the MHC-related neonatal Fc receptor. Nature 1994, 372:336-343. Mason K, Denney Dl, McConnell HM: Myelin basic protein peptide complexes with the class II MHC molecules I-A(U) and I-A(K) form and dissociate rapidly at neutral pH. J Immunol 1995, 154: 521 6-5227. Marks MS, Roche PA, van Donselaar E, Woodruff L, Peters PJ,Bonifacino JS: A lysosomal targeting signal in the cytoplasmic tail of the 3 chain directs HLA-DM to MIIC class II compartments. J Cell Biol 1995, 131:351-369.
Frances Sanderson and John Trowsdale, Human Immunogenetics Laboratory, Imperial Cancer Research Fund, P.O. Box 123, Lincoln's Inn Fields, London WC2A 3PX, UK.
THE FEBRUARY 1996 ISSUE (VOL. 8, NO. 1) OF CURRENT OPINION IN IMMUNOLOGY will include the following reviews, edited by Kirsten Fischer-Lindahl and Hans-Georg Rammensee, on Antigen recognition: Developing and shedding inhibitions: how MHC class II molecules reach maturity by E. Mellins T-cell receptor affinity and structure by D. Fremont and H. Kozono Human minor histocompatibility antigens by E. Goulmy Nonclassical MHC ligands including CD1 by M. Brenner Natural killer cell receptors specific for MHC class I molecules by M. Colonna T cell epitope determination by P. Walden Process and delivery of peptides presented by MHC class I by P. Cresswell Affinity maturation and class switching by M. Wabl and C. Steinberg The same issue will also include the following reviews on Innate immunity, edited by Alan Ezekowitz and Jules Hoffman: Signal transduction in plant immunity by K. Shirasu, R.A. Dixon and C. Lamb Innate immunity in higher insects by J.A. Hoffman, J.M. Reichhart and C. Hetru Immunity to eukaryotic parasites in vector insects by EC. Kafatos and A. Richman Innate immunity in lower vertebrates by M. Zasloff Scavenger receptors in innate immunity by A.M. Pearson and M. Kreiger The collectins in innate immunity by J. Epstein, Q. Eichbaum, S. Sheriff and A. Ezekowitz Mechanisms of phagocytosis by L-A.H. Allen and A. Aderem The role of haemolymph coagulation in innate immunity by T. Muta and S. Iwanaga
FRANCES SANDERSON AND JOHN TROWSDALE
ANTIGEN PRESENTATION
ANTIGEN PRESENTATION
Kissing cousins exchange CLIP HLA-DM, a recent addition to the immunoglobulin family, isstructurally most like class IImolecules of the MHC. DM facilitates exchange of CLIP, a temporary filler of the groove of class IImolecules, for antigenic peptides. A central event in the development of a specific immune response is recognition by the T-cell receptor of an antigenic peptide bound within the groove of a class I or class II molecule of the major histocompatibility complex (MHC) [1]. Class I molecules handle peptides from proteins synthesized within the cell, such as viral proteins, which they present to lymphocytes bearing the cell-surface marker CD8. Class I-binding peptides are generated by a multi-component protease, the proteasome, and are loaded into the groove of the class I molecule after translocation into the endoplasmic reticulum (ER). Class II molecules, on the other hand, are responsible for presenting peptides derived from extracellular pathogens to the CD4-bearing subset of T cells.
shown that the CLIP region is both necessary and sufficient for the blocking and stabilizing effects of intact Ii on MHC class II molecules. Two observations suggest that the binding of CLIP to HLA-DR is different from the binding of antigenic peptides. First, certain antibodies (such as the monoclonal 16.23) that normally recognize cell-surface class 1I molecules are unable to bind the 1-CLIP complex. Second, although mature class II-peptide complexes are characteristically stable in the detergent SDS, most
The human class II molecules - HLA-DP, -DQ and -DR, of which HLA-DR is the prototype (Fig. 1) - are heterodimers of two immunoglobulin (Ig) superfamily protein chains, ax and 13, the genes for which reside together in the MHC. Upon synthesis, class II a13 heterodimers associate with the invariant chain, Ii, a type II (inverted) membrane protein with which they are exported from the ER in a nonameric complex. Unlike MHC class I molecules, newly synthesized class II molecules do not follow the constitutive secretory pathway. Instead, to allow them to capture peptides from external sources, they are guided by targeting signals contained in the cytoplasmic tail of Ii onto a pathway that intersects that of endocytic traffic (Fig. 2). This journey to the cell surface takes several hours, during which time Ii is degraded and removed from the class II molecule, which is in turn loaded with peptide in an acidic compartment. Ii has other important functions: it helps assembly of class II a and chains and prevents their aggregation; it also blocks the groove of class II a13 heterodimers, so preventing binding of self peptides in the lumen of the ER. A strong indication as to how Ii fulfils its blocking and stabilizing roles is provided by mutant cell lines in which the processing of antigens to be presented by class II molecules is faulty. Most of the class II molecules on the surfaces of these cells are bound not by the normal spectrum of antigenic peptides but by a nested set of fragments that span residues 81-105 of Ii and are known as CLIP (class II-associated invariant chain peptides). The normal fate of class II-bound Ii is to be attacked by proteases which remove it in sections, systematically. CLIP is the smallest fragment of Ii to be retained, and ua3-CLIP complexes form a natural intermediate in class II maturation in vivo [2]. Genetic truncation and deletion experiments have 1372
Fig. 1. HLA-DM is a distant cousin of class I and class II
molecules of the MHC. Sequence comparisons show that DM may have split from the primordial class I and class IIsequences early on in evolution; the protein structures are shown, highly schematically, below the tree. DM is similar in arrangement to other class II structures, except for the extra cysteine (C)residues in the ul and 31 domains, the amino-acid sequences of which are only weakly related to those of other HLA molecules. DM has no recognizable CD4-binding site.
© Current Biology 1995, Vol 5 No 12
DISPATCH
Fig. 2. Behaviour of DR in DM+ and DM- cells. Peptides in the lumen of the ER are prevented from binding to class II by i. Ii and the a and chains of the class II molecule form a nonameric complex, (1i) 3, for transport, held together by a 'trimerization' sequence in the carboxy-terminal (luminal) region of Ii (see also Fig. 4). Destruction of the trimerization domain of i, at a later stage, as the nonamer moves to the endocytic vesicles under the guidance of amino-terminal Ii signals, may be a key step in the process. DM has its own targeting signal that may enable the molecule to be retrieved from the cell surface [181. DM may also interact with i in the ER. Degradation of endocytosed proteins takes place so that peptides are available to newly synthesized class II molecules as i is released. DM accumulates in the MHC class II compartment (MIIC), where it is thought to catalyze peptide loading onto class II molecules; class II molecules loaded with peptide move to the cell surface. There is evidence for cell-type-specific variations in this pathway, especially in phagocytic cells, as well as for recycling of some class II molecules from the cell surface. The intravesicular pH through the class II trafficking pathway decreases as it does inendocytosis, to lower than pH 5 inthe MIIC. CLIP-bound class II molecules dissociate in SDS into their constituent a and 3 chains. Both findings point to a structure for xot3-CLIP that is different from that adopted by other class II-peptide complexes. In addition, attracting and repelling forces between peptide side chains and the polymorphic residues lining the class II peptide-binding groove result in the specificity of most peptides for a limited range of class II allotypes; in contrast, CLIP binds to most, if not all, class II allotypes. Although it seemed possible that CLIP inhibits peptide binding indirectly, the wide range of binding affinities exhibited by CLIP for different class II molecules, and the allele-dependent manner in which mutations introduced into synthetic CLIP influence binding, has strongly favoured a model in which CLIP occupies the groove [3,4]. Structure determination now confirms this [5]. Ghosh et al [5] have crystallized HLA-DR3 in an ax3-CLIP complex and solved the structure to 2.75 A resolution (Fig. 3). Remarkably, the o4-CLIP complex has a structure almost indistinguishable from that of class II with bound antigenic peptide (specifically, HLA-DR1 with a peptide from influenza virus haemagglutinin, HA 306-318). The CLIP is bound to HLA-DR3 in a conformation and with a network of hydrogen bonds similar to those in the HLA-DR1-HA complex, and with similar
side-chain-pocket interactions. The only conformational difference between the DR3-CLIP and DR1-HA complexes is in the second a-helical region of the DR3 3 chain (residues 65-74), which in the DR3-CLIP complex has moved 1.5 A towards the peptide. This helical section corresponds to the predicted epitope of the conformation-specific monoclonal antibody 16.23 [6], and its displacement could provide an elegant explanation for the recognition by the antibody of mature DR3-peptide but not DR3-CLIP complexes [7]. The reason for the SDS-instability of class II-CLIP complexes (including DR3-CLIP) is less clear. Certainly it can no longer be attributed to any major difference in the io3 portion of the structure. Nor is CLIP more superficially bound: the conservation of hydrogen bonds between the two complexes and the comparability of solvent-accessible surfaces argue that, as predicted from affinity measurements in vitro, DR3-CLIP and DR1-HA complexes will have similar stabilities. In the context of intact Ii, the residues spanned by the CLIP fragment need not necessarily bind within the ot groove nor with the conformation adopted by the isolated peptide. But if they do, the deep interactions demonstrated in the oaL-CLIP crystal structure could be responsible for Ii-mediated stabilization of xa13in much
1373
1374
Current Biology 1995, Vol 5 No 12 of CLIP from DR, as well as facilitating loading with antigenic peptides. Incorporation of DR molecules
from wild-type cells into DM-containing membranes accelerates peptide binding in vitro [15]. Membrane reconstitution is not essential, however: immunoprecipitated DM, purified soluble recombinant DM, and even detergent-solubilized DM-containing membranes have similar effects on DR molecules in solution [13,15]. All three studies have demonstrated that, at acidic pH in vitro, DM promotes the dissociation of CLIP from a-CLIP complexes.
Fig. 3. Side view of CLIP within the HLA-DR3 peptide-binding site [5]. (Graphic courtesy Partho Ghosh).
the same way as stability is conferred by the binding of antigenic peptides. Analysis of the same antigen-processing mutants that led to the identification of CLIP as the universal class II-binding peptide has also identified a function for two novel class II genes, HLA-DMA and HLA-DMB [8]: their products are required for successful CLIP removal from, and peptide loading onto, class II molecules [9,10]. Inspection of the DM protein sequences, as shown schematically in Figure 1, reveals that the DM genes could have split from the primordial class I/II sequences several hundred million years ago, at around the time that class I and class II diverged from each other. The DM u-2 and 32 domains are homologous to the membrane-proximal domains of class I and II molecules; the al and 31 regions, which in DR form the peptide-binding groove, contain little primary sequence homology but can be modelled into a structure similar to the DR archetype (C. Thorpe and P. Travers, personal communication). Unlike the majority of class II heterodimers, little DM is found at the cell surface. Instead, it accumulates together with class II molecules in the specialized late endosomal/lysosomal MHC class II compartments (MIICs) [11]. These membrane-bound vesicles co-migrate on Percoll density gradients with compartments in which immunostimulatory class II molecules first appear [12], and they are the putative site of class II peptide loading. Thus, DM is well placed for its apparent role in expediting the association between antigenic peptide and class II molecules. So how does DM effect peptide exchange? Initially the most attractive theories, given the molecule's class II-like structure, proposed that DM might act either by binding CLIP and removing it from the class II molecule to make way for other peptides -
the 'CLIP sink' model -
or
by binding a range of peptides in endocytic vesicles and bringing them to class II in a peptide-loading compartment. Evidence from recent studies favours a different scheme [13-15]. Rather than participating in a peptideshuttle type of mechanism, it seems that DM acts as a catalyst which at low pH increases the dissociation rate
A requirement in vivo for DM in order to release Iiderived peptides from DR would account not only for the persistence of DR-CLIP complexes in DM-negative mutant cells, but also for the apparent DM-independence of antigen presentation by the mouse allele I-Ak, which stands out among class II molecules as having a remarkably low affinity for CLIP [4]. Much of the effect of HLA-DM on peptide binding appears to be attributable to DM-catalyzed CLIP release, but not all. Denzin and Cresswell [14] compared the efficiencies with which either DM or the detergent octylglucoside produces peptide-loaded DR molecules from DR-CLIP complexes in vitro. Although CLIP is released faster in the presence of detergent, comparable peptide loading requires 100-fold higher concentration of peptide. Thus, DM appears to have a direct effect on peptide loading. To date, there is no evidence that DM binds peptide; in fact, under conditions in which it is clearly able to remove CLIP from anx-CLIP complexes, no DM with bound CLIP was detected [14]. To explain these data, it has been proposed that DM interacts fleetingly with ota-CLIP complexes, promoting CLIP release and then stabilizing a peptide-receptive conformation adopted by the empty oa3 class II molecule [14]. Support for the notion that a physical interaction between the two molecules is required for function comes from work on a DR molecule with a proline-serine substitution at position 96 of the ax chain (DR3Ser 96 ), which introduces an extra N-linked glycosylation site [13]. The phenotype of the DR3Ser 96 mutant in vivo is similar to that of DR in DM-negative cells. In vitro, the altered DR molecule is refractory to the effects of DM on CLIP removal. A similar effect was produced by the pre-incubation of a1-CLIP with an anti-CLIP antibody. It is reasonable to propose that both the bulky carbohydrate group on the mutant DR molecule and the anti-CLIP antibody prevent interaction with DM. So far, direct evidence for association between DR and DM has been difficult to obtain. However, data from our laboratory show a non-covalent DM-DR interaction, under conditions that approximate those of the appropriate endosomal compartment (ES., C. Thomas, J. Neefjes and J.T., unpublished observations). It is odd that DM, the catalyst for DR peptide-loading, is a molecule that shares a structural relationship with its substrate yet does not appear to bind peptide. A precedent can be found in the rat neonatal Fc receptor,
DISPATCH
Fig. 4. Speculative model of the role of DM in catalyzing peptide loading of DR. For details see text.
which has the primary structure of a typical class I molecule, but its binding groove proved to be distorted and empty [16]. A closer inspection of the sequence of HLA-DM suggests that here too the putative groove may be malformed, with bulky hydrophobic residues obstructing it at one end. A speculative model for the function of HLA-DM is therefore as follows. The class II-Ii complex leaves the ER as a nonamer, and it is significant that this complex is not a substrate for exchange mediated by DM [14]. As the trimerization domain is in the section of the Ii chain that is removed first, it seems likely that the dissolution of the (Ii) 3 nonamer takes place upon entering, or just before entering, the MIIC. Which fragment of Ii is removed by DM in vivo is not entirely clear. DM is certainly able to catalyze the dissociation of fragments larger than CLIP in vitro [14], and it is possible (as indicated in Fig. 4) that class II molecules bound to these larger fragments serve as one, or possibly even the principal, substrate for DM in vivo. In this model, DM does not have to bind peptide itself in order to effect peptide exchange on DR. Rather, as mentioned above, the evidence points to a direct interaction between the two MHC-encoded molecules. It is necessary only to postulate a fleeting 'kiss' between DM and DR (Fig. 4); in vitro, in the absence of peptide, unoccupied class II o3 heterodimers become refractory to peptide binding [17] and these may be stabilized by association with DM. This interaction, by a mechanism analogous to that by which an enzyme stabilizes a transition state, may account for both the release of CLIP and the facilitation of peptide binding in vitro. Enzyme-substrate binding is often of low affinity, and certainly the association between DM and DR is difficult to demonstrate. Although the structures of DR1-HA and DR3-CLIP complexes are similar, the structure of the empty ad intermediate is unknown. It is possible that a change in conformation effected by peptide binding is the trigger for the dissociation of DM, which is then free to catalyze another round of peptide exchange.
One consequence of this model, and also one of the most provocative features of DM-catalyzed peptide dissociation in vitro, is that CLIP is not the only peptide to be removed from DR [13]. The half-life of a DR1-myelin basic protein complex was also shortened by DM, and to a comparable extent to that of ot-CLIP complexes. It has been suggested that the DM-catalyzed dissociation of non-CLIP peptides in vivo serves an important function [13]. By favouring the formation of a subset of oa3-peptide complexes with long half-lives, DM may be responsible for the observed selection of a small number of immunodominant epitopes from the array of available class II-binding peptides. Clearly, the next step is to determine whether it is the affinity of the peptide, a specific structural feature of the complex, or some other characteristic of the class II-peptide interaction, that confers susceptibility to DM-mediated exchange. There are many fascinating examples of immunoglobulin superfamily members put to work in a novel ways. But few, like DM, are altruistic: DM helps out its immediate class II family with perhaps no chance of binding a peptide of its own. Acknowledgements: We would like to thank P. Travers, N. Koch and G. Malcherek for comments, and S. Giles for secretarial help. References 1. Germain RN: MHC-dependent antigen processing and peptide presentation: providing ligands for T lymphocyte activation. Cell 1994, 76:287-299. 2. Avva RR, Cresswell P: In vivo and in vitro formation and dissociation of HLA-DR complexes with invariant chain-derived peptides. Immunity 1994, 1:763-774. 3. Malcherek GC,Gnan V, lung G, Rammensee H-G, Melms A: Supermotifs enable natural invariant chain-derived peptides to interact with many major histocompatibility complex-class II molecules. J Exp Med 1995, 181:527-536. 4. Sette A, Southwood S, Miller , Appella E: Binding of major histocompatibility complex class II to the invariant chain-derived peptide, CLIP, is regulated by allelic polymorphisms in class II. J Exp Med 1995, 181:677-683. 5. Ghosh P, Amaya E, Mellins E, Wiley DC: The structure of an
1375
1376
Current Biology 1995, Vol 5 No 12
6. 7.
8. 9. 10.
11.
12. 13.
intermediate in class II MHC maturation. HLA-DR3 complexed with the invariant chain fragment CLIP. Nature 1995, 378:457-462. Marsh SGE, Bodmer JG: HLA-DR and -DQ epitopes and monoclonal antibody specificity. Immunol Today 1989, 10:305-312. Mellins E,Cameron P, Amaya M, Goodman S, Pious D, Smith L, Arp B: A mutant human histocompatibility leukocyte antigen DR molecule associated with invariant chain peptides. I Exp Med 1994, 179:541-549. Kelly AP, Monaco IJ,Cho S, Trowsdale J: A new human HLA class Il-related locus, DM. Nature 1991, 353:571-573. Fling SP, Arp B, Pious D: HLA-DMA and -DMB genes are both required for MHC class 11/peptide complex formation in antigenpresenting cells. Nature 1994, 368:554-558. Morris P, Shaman J, Attaya M, Amaya M, Goodman S, Bergman C, Monaco JJ, Mellins E: An essential role for HLA-DM in antigen presentation by class II major histocompatibility molecules. Nature 1994, 368:551-554. Sanderson F, Kleijmeer MJ, Kelly AP, Verwoerd D, Tulp A, Neefjes JJ, Geuze HI, Trowsdale J:Accumulation of HLA-DM, a regulator of antigen presentation, in MHC class II compartments. Science 1994, 266:1566-1569. Qiu Y, Xu X, Wandinger-Ness A, Dalke DP, Pierce SK: Separation of subcellular compartments containing distinct functional forms of MHC class II. J Cell Biol 1994, 125:595-605. Sloan VS, Cameron P, Porter G, Gammon M, Amaya M, Mellins E,
14. 15. 16. 17.
18.
Zaller DM: Mediation by HLA-DM of dissociation of peptides from HLA-DR. Nature 1995, 375:802 806. Denzin LK, Cresswell P: HLA-DM induces CLIP dissociation from MHC class II alpha beta dimers and facilitates peptide loading. Cell 1995, 82:155-165. Sherman MA, Weber DA, Jensen PE: DM enhances peptide binding to class II MHC by release of invariant chain-derived peptide. Immunity 1995, 3:197-205. Burmeister WP, Gastinel LN, Sinister NE, Blum ML, Bjorkman PJ: Crystal structure at 2.2 A resolution of the MHC-related neonatal Fc receptor. Nature 1994, 372:336-343. Mason K, Denney Dl, McConnell HM: Myelin basic protein peptide complexes with the class II MHC molecules I-A(U) and I-A(K) form and dissociate rapidly at neutral pH. J Immunol 1995, 154: 521 6-5227. Marks MS, Roche PA, van Donselaar E, Woodruff L, Peters PJ,Bonifacino JS: A lysosomal targeting signal in the cytoplasmic tail of the 3 chain directs HLA-DM to MIIC class II compartments. J Cell Biol 1995, 131:351-369.
Frances Sanderson and John Trowsdale, Human Immunogenetics Laboratory, Imperial Cancer Research Fund, P.O. Box 123, Lincoln's Inn Fields, London WC2A 3PX, UK.
THE FEBRUARY 1996 ISSUE (VOL. 8, NO. 1) OF CURRENT OPINION IN IMMUNOLOGY will include the following reviews, edited by Kirsten Fischer-Lindahl and Hans-Georg Rammensee, on Antigen recognition: Developing and shedding inhibitions: how MHC class II molecules reach maturity by E. Mellins T-cell receptor affinity and structure by D. Fremont and H. Kozono Human minor histocompatibility antigens by E. Goulmy Nonclassical MHC ligands including CD1 by M. Brenner Natural killer cell receptors specific for MHC class I molecules by M. Colonna T cell epitope determination by P. Walden Process and delivery of peptides presented by MHC class I by P. Cresswell Affinity maturation and class switching by M. Wabl and C. Steinberg The same issue will also include the following reviews on Innate immunity, edited by Alan Ezekowitz and Jules Hoffman: Signal transduction in plant immunity by K. Shirasu, R.A. Dixon and C. Lamb Innate immunity in higher insects by J.A. Hoffman, J.M. Reichhart and C. Hetru Immunity to eukaryotic parasites in vector insects by EC. Kafatos and A. Richman Innate immunity in lower vertebrates by M. Zasloff Scavenger receptors in innate immunity by A.M. Pearson and M. Kreiger The collectins in innate immunity by J. Epstein, Q. Eichbaum, S. Sheriff and A. Ezekowitz Mechanisms of phagocytosis by L-A.H. Allen and A. Aderem The role of haemolymph coagulation in innate immunity by T. Muta and S. Iwanaga