Structural basis for the binding of an immunodominant peptide from myelin basic protein in different...

doi:10.1006/jmbi.2000.4198 available online at http://www.idealibrary.com on J. Mol. Biol. (2000) 304, 177±188

Structural Basis for the Binding of anImmunodominant Peptide from Myelin Basic Proteinin Different Registers by Two HLA-DR2 Proteins

Yili Li1, Hongmin Li1, Roland Martin2 and Roy A. Mariuzza1*

1Center for Advanced Researchin Biotechnology, University ofMaryland BiotechnologyInstitute, 9600 Gudelsky DriveMD 20850, USA2Neuroimmunology BranchNational Institute ofNeurological Disorders andStroke, National Institutes ofHealth, Bethesda, MD 20892-1400, USA

Y.L. and H.L. contributed equallAbbreviations used: MS, multipl

central nervous system; MBP, myeEAE, experimental allergic encephaphosphate-buffered saline; CA, conTCR, T cell receptor.

E-mail address of the [email protected]

0022-2836/00/020177±12 $35.00/0

Susceptibility to multiple sclerosis (MS) is associated with certain MHCclass II haplotypes, in particular HLA-DR2. Two DR b chains, DRB1*1501and DRB5*0101, are co-expressed in the HLA-DR2 haplotype, resulting inthe formation of two functional cell surface heterodimers, HLA-DR2a(DRA*0101, DRB5*0101) and HLA-DR2b (DRA*0101, DRB1*1501). Bothisotypes can present an immunodominant peptide of myelin basic protein(MBP 84-102) to MBP-speci®c T cells from MS patients. We have deter-mined the crystal structure of HLA-DR2a complexed with MBP 86-105 to1.9 AÊ resolution. A comparison of this structure with that of HLA-DR2bcomplexed with MBP 85-99, reported previously, reveals that the peptideregister is shifted by three residues, such that the MBP peptide is boundin strikingly different conformations by the two MHC molecules. Thisshift in binding register is attributable to a large P1 pocket in DR2a,which accommodates Phe92, in conjunction with a relatively shallow P4pocket, which is occupied by Ile95. In DR2b, by contrast, the small P1pocket accommodates Val89, while the deep P4 pocket is ®lled by Phe92.In both complexes, however, the C-terminal half of the peptide is posi-tioned higher in the binding groove than in other MHC class II/peptidestructures. As a result of the register shift, different side-chains of theMBP peptide are displayed for interaction with T cell receptors in theDR2a and DR2b complexes. These results demonstrate that MHC mol-ecules can impose different alignments and conformations on the samebound peptide as a consequence of topological differences in theirpeptide-binding sites, thereby creating distinct T cell epitopes.

# 2000 Academic Press

Keywords: multiple sclerosis; myelin basic protein; MHC class II;HLA-DR2; X-ray crystallography
*Corresponding author
Introduction

Multiple sclerosis (MS) is an in¯ammatorydemyelinating disease of the central nervous sys-tem (CNS) believed to be mediated by autoreac-tive T cells (Steinman, 1996). The activation ofresting T cells reactive with CNS antigens, suchas myelin basic protein (MBP), is thought toconstitute the primary autoimmune event in MS.

y to this work.e sclerosis; CNS,lin basic protein;lomyelitis; PBS,albumin peptide;

ing author:

Similar to other polygenic chronic in¯ammatorydiseases, such as rheumatoid arthritis andinsulin-dependent diabetes, susceptibility to MSis associated with certain MHC class II haplo-types, in particular HLA-DR2 (Hillert et al., 1994;Oksenberg et al., 1996; Ebers et al., 1996). TwoHLA-DR b chains, DRB1*1501 and DR5*0101,are co-expressed in the HLA-DR2 haplotype.Both pair with the non-polymorphic HLA-DR achain (DRA*0101), resulting in the formation oftwo functional cell surface heterodimers, HLA-DR2b (DRA*0101, DRB1*1501) and HLA-DR2a(DRA*0101, DRB5*0101) (Sone et al., 1985). Thesegene products are believed to confer disease sus-ceptibility by de®ning the speci®city of animmune attack against the white matter of thebrain by autoreactive T cells (Steinman, 1996). Anumber of autoantigens, including MBP, myelin

# 2000 Academic Press

http://www.idealibrary.com

mailto:[email protected]

178 Structure of Human MHC Class II/Peptide Complex

oligodendrocyte glycoprotein and proteolipidprotein, have been proposed as candidate anti-gens in MS based on their encephalitogenicity inexperimental allergic encephalomyelitis (EAE), ananimal model of MS (Steinman, 1996).

In humans, an immunodominant T cell epitopehas been identi®ed in the middle portion of MBP,corresponding to residues 84 to 102 (Ota et al.,1990; Martin et al., 1990; Pette et al., 1990). Theencephalitogenic potential of MBP 84-102 hasrecently been shown in a trial with an altered pep-tide ligand of this peptide (Bielekova et al., 2000).The relevance of DR2b in conjunction with thesame peptide is also strongly supported by datafrom a humanized transgenic mouse expressingDR2b and a T cell receptor (TCR) speci®c for MBP84-102 (Madsen et al., 1999) and by pathologicalstudies demonstrating the presentation of MBP 85-99 in the context of DR2b in MS brain tissue(Krogsgaard et al., 2000). While the above evidencestrongly supports the relevance of MBP 84-102 inMS, it is not yet clear which of the two DR2 allelesis more important in presenting autoantigenic pep-tides, since MBP 84-102 can be presented by bothDR2a and DR2b to T cell clones from MS patients(Martin et al., 1990; Pette et al., 1990; Jaraquemadaet al., 1990; Vogt et al., 1994; Wucherpfennig et al.,1994; Vergelli et al., 1997). It has also been shownto bind a number of other DR haplotypes, includ-ing DR1 and DR4 (Valli et al., 1993). Binding toHLA-DR is particularly important for this epitopebecause it protects the peptide from degradationby cathepsin D, an abundant protease in proces-sing compartments (Vergelli et al., 1997). The abil-ity of certain peptides to bind multiple MHC classII molecules is a frequently observed phenomenon(O'Sullivan et al., 1991; Hammer et al., 1993) and isnot restricted to MBP 84-102. For example, a num-ber of peptides derived from the circumsporozoiteprotein of malaria have been found to be recog-nized by T cells in association with most MHCclass II alleles, both mouse and human (Singagliaet al., 1988; Doolan et al., 2000).

In principle, peptide cross-presentation by differ-ent MHC class II molecules may be explained bytwo general mechanisms: (i) the peptide binds par-ticular class II molecules simply because theirpeptide-binding grooves are structurally very simi-lar, such that the overall alignment and confor-mation of the bound peptide are maintained in thedifferent MHC/peptide complexes; or (ii) the pep-tide is able to bind selected class II molecules withtopologically distinct grooves but, as a conse-quence, the alignment and/or conformation of thebound peptide are altered. The interaction of MBP84-102 with HLA-DR2a and HLA-DR2b appears tofall into the latter category. Therefore, Val89 andPhe92 of the MBP peptide were identi®ed as theP1 and P4 anchor residues for HLA-DR2b bindingbased on binding assays using peptide analogsand on alignments of naturally processed foreignand self-peptides eluted from puri®ed DR2 mol-ecules (Vogt et al., 1994; Wucherpfennig et al.,

1994). However, for the binding to HLA-DR2a,Phe92 was identi®ed as the P1 anchor residue,implying that DR2a and DR2b present MBP 84-102to T cells in different registers.

Recently, the crystal structure of HLA-DR2bcomplexed with MBP 85-99 has been reported(Smith et al., 1998). A distinguishing feature of theDR2b peptide-binding site is a large, predomi-nantly hydrophobic P4 pocket created by an ala-nine residue at the polymorphic 71b position. Aspredicted (Vogt et al., 1994; Wucherpfennig et al.,1994), this pocket is occupied by Phe92 of the MBPpeptide, while the smaller P1 pocket accommo-dates Val89. Here, we describe the crystal structureof HLA-DR2a complexed with MBP 86-105 to1.9 AÊ resolution. The structure reveals that theMBP peptide is bound in a strikingly differentfashion by HLA-DR2a compared to HLA-DR2b,such that the peptide register is shifted by threeresidues. This shift in register is attributable to alarge P1 pocket, which accommodates the bulkyPhe92 side-chain, in combination with a relativelyshallow P4 pocket, which is occupied by Ile95. Asa consequence of the shift in peptide register,different side-chains of MBP are exposed for inter-action with TCR in the DR2a and DR2b complexes.Thus, a given peptide residue does not have aunique function of always interacting with MHCor TCR, but may switch from one role to the otheras it is presented by different MHC alleles. Thisstudy represents the ®rst demonstration by X-raycrystallography that MHC molecules can imposemarkedly different conformations on the samebound peptide as a result of topological differencesin their peptide-binding grooves.

Results and Discussion

Overall structure of the HLA DR2a/MBP86-105 complex

The structure of HLA-DR2a complexed with theimmunodominant MBP 86-105 self-peptide wasdetermined to 1.9 AÊ resolution, the highest yetreported for a human MHC class II/peptide com-plex (Stern et al., 1994; Ghosh et al., 1995; Murthy& Stern, 1997; Dessen et al., 1997; Smith et al.,1998). We attribute this improvement, at least inpart, to the homogeneity of the bacterially-expressed protein, compared to recombinant classII from insect cells. The complex was assembled byrefolding the DR2a a and b chains, produced inEscherichia coli as inclusion bodies, in the presenceof the MBP 86-105 peptide. Stockel et al. (1994)have reported the reconstitution of biologicallyactive DR2a/MBP 86-105 complex from thedenatured a and b chains using the mild detergentdodecy-b-D-maltoside to reduce aggregation andincrease refolding ef®ciency. We instead used highconcentrations (25 %) of glycerol for this purpose,as described for HLA-DR1 (Frayser et al., 1999).The electron density maps for the complex are ofhigh quality, as shown in Figure 1(a) for the bound

Figure 1. (a) Electron density for the bound peptide in the HLA-DR2a/MBP 86-105 complex. The Fo ÿ Fc omit mapat 1.9 AÊ resolution is contoured at 2s. (b) Superposition of the MBP 86-105 peptide (red) bound by HLA-DR2a ontothe MBP 85-99 peptide (green) bound by HLA-DR2b (Smith et al., 1998). (c) Alignment of MBP 86-105 and MBP 85-99 according to their registers in the peptide-binding grooves of HLA-DR2a and HLA-DR2b, respectively.

Structure of Human MHC Class II/Peptide Complex 179

peptide. In the ®nal model of the HLA-DR2a/MBP86-105 complex, the two molecules in the asym-metric unit (hereafter referred to as molecule I andII) displayed electron density for all MHC a and bchain residues, with only a loop composed of bresidues 109 to 111 of molecule II and N-terminal aresidues 1 to 2 of molecule I and b residues 1 to 3of molecule II lacking electron density. The pep-tide-binding a1 and b1 domains of HLA-DR2aappear to be the most ordered, as re¯ected byaverage temperature B-factors that are 9 AÊ 2 lowerthan those for the a2 and b2 domains (Table 1).Continuous and unambiguous density was alsoseen for the complete MBP 86 to 105 peptide (20residues) in molecule I (Figure 1(a)). For molecule

II, 18 MBP residues (87 to 104) were clearlyde®ned. Molecules I and II in the asymmetric unitare virtually identical. Superposition of the a1/b1domains gives a r.m.s.d. in a-carbon positions ofonly 0.07 AÊ ; for the complete HLA-DR2a mol-ecules, the r.m.s.d. is 0.38 AÊ . The structures of thetwo non-crystallographically related peptides arealso very similar (r.m.s.d. of 0.05 AÊ ).

The overall structure of HLA-DR2a is, asexpected, very similar to that of other MHC class IImolecules, both mouse and human (Stern et al.,1994; Ghosh et al., 1995; Fremont et al., 1996;Murthy & Stern, 1997; Dessen et al., 1997; Scottet al., 1998; Fremont et al., 1998; Smith et al., 1998)(Figure 2(a)). Superposition of the a1 and b1

Table 1. Crystal data and structure re®nement statistics

A. Data collectionSpace group P21

Unit cell a � 63.19 AÊ ,b � 114.89 AÊ ,c � 63.18 AÊ ,

a � 90.0 �, b � 90.9 �,g � 90.0 �

No. mol./asymm. unit 2Resolution (AÊ ) 1.9Observations 181,399Unique reflections 63,997Completeness (%) 90.8 (75.6)a

Mean I/s(I) 12.8 (2.0)a

Rsym (%)b 8.1 (34.9)a

B. RefinementResolution range 100-1.9 AÊ

Rwork (%)b 23.3Rfree (%)b 26.8Non-hydrogen protein atoms 6897Solvent molecules

Water molecules 557SO4 3Glycerol 2

Average atom B-factors (AÊ 2)MHC 33.8Peptide 38.6Solvent 37.2

r.m.s.d. from idealityBonds (AÊ ) 0.006Angles (deg.) 1.5

a Values in parentheses correspond to the highest resolutionshell (2.0-1.9 AÊ ).

b Rsym � �j(Ihkl ÿ Ihhklij/�(Ihhkli), where Ihhkli is the meanintensity of all re¯ections equivalent to re¯ection hkl by symme-try; Rwork(Rfree) � �jjFoj ÿ jFcjj/�jFoj; 5 % of data were used forRfree.


domains of DR2a onto those of DR1 (Stern et al.,1994), DR2b (Smith et al., 1998), DR3 (Ghosh et al.,1995), DR4 (Dessen et al., 1997), I-Ek (Fremont et al.,1996), I-Ad (Scott et al., 1998) and I-Ak (Fremontet al., 1998) gives average r.m.s.d. in a-carbon pos-itions of 0.36, 0.44, 0.49, 0.59, 0.75, 1.26 and 0.91 AÊ ,respectively. Therefore, the three-dimensionalstructure of HLA-DR2a is closest to that of HLA-DR1; the sequence of the DR2a b chain is also clo-sest to that of DR1 among all DRB1 alleles. In thecrystal, HLA-DR2a exists as a dimer of ab hetero-dimers. Dimerization of MHC class II heterodimershas also been observed in crystals of DR1 (Brownet al., 1993; Stern et al., 1994; Murthy & Stern,1997), DR2b (Smith et al., 1998) and DR3 (Ghoshet al., 1995), in which two DR molecules are paral-lel to each other with their peptide-binding sitesfacing the same direction. By contrast, the DR2a/MBP complex forms an anti-parallel dimer, inwhich the two peptide-binding sites face oppositedirections.

Conformation of the bound MBP peptide

The MBP 86-105 eicosamer (NPVVHFFKNIVTPRTPPPSQ) binds in an extended conformationwith a pronounced twist (f/c angles near those of

polyproline) (Figure 1(b)) and with the N and Ctermini protruding from the ends of the bindinggroove (Figure 2(a)). The conformation of the MBPpeptide was directly compared with those of pep-tides of other MHC class II/peptide complexes(Stern et al., 1994; Ghosh et al., 1995; Fremont et al.,1996; Murthy & Stern, 1997; Dessen et al., 1997;Scott et al., 1998; Fremont et al., 1998; Smith et al.,1998) by superposition of the a1/b1 domains ofthe class II molecules (Figure 3(a)). The backboneatoms of MBP residues 90 to 95 superpose wellonto the N-terminal segments of these peptides(P-2 to P4), where P1 is the ®rst peptide anchorresidue at an arbitrary distance from the peptideN terminus (MBP residues 86 to 105 are therebydesignated P-6 to P14). However, the peptides dis-play considerable divergence between residues P6and P10. Compared to other peptides bound toclass II molecules, the C-terminal half of MBP86-105 is positioned higher in the HLA-DR2a bind-ing groove, like the C terminus of MBP 85-99 inthe HLA-DR2b complex (Smith et al., 1998)(Figure 3(a)), as discussed below.

Residues P-1 Phe and P10 Pro of the MBPpeptide partially protrude from the DR2a bind-ing site. Amino-terminal residues P-6 to P-2 andC-terminal residues P11 to P14, though unam-biguously de®ned in the electron density(Figure 1(a)), extend completely beyond thegroove (Figure 2(a)). Residues P-6 to P-2 are ina fully extended conformation (Figure 1(b)),which may be partly stabilized by crystal con-tacts between Val P-3 and His P-2 in one com-plex molecule and DR residues Gln107b andThr103b, respectively, in a symmetry-mate. Inagreement with the crystal structure, truncationof the C terminus of the MBP peptide to P9 Thrgreatly diminished binding, whereas truncationto P10 Pro had little effect (Vogt et al., 1994;Wucherpfennig et al., 1994). It is surprising thathowever, truncation of the N terminus to P-1Phe resulted in a marked reduction in af®nity,even though P-2 His stands outside the bindingsite. This may be explained by the loss of ahydrogen bond between the main-chain carbonylof P-2 His and the main-chain amide of DR2aresidue Ser53a.

Peptide interactions in the HLA-DR2abinding cleft

In the peptide-binding site, 42 HLA-DR2a resi-dues account for a total of 942 AÊ 2 of buried surfacearea, while 1176 AÊ 2 of peptide surface areabecomes inaccessible to solvent upon binding. TheMBP 86-105 eicosamer forms a total of 12 hydro-gen bonds with seven different residues in theHLA-DR2a binding cleft. Most of the conservedhydrogen bonds between main-chain peptideatoms and side-chain atoms of MHC class II resi-dues present in other class II/peptide structures(Stern et al., 1994; Ghosh et al., 1995; Fremont et al.,1996; Murthy & Stern, 1997; Dessen et al., 1997;

Figure 2. Comparison of theHLA-DR2a/MBP and HLA-DR2b/MBP complexes. (a) Top view ofthe HLA-DR2a/MBP 86-105 com-plex. (b) View of the large P1pocket of HLA-DR2a that is occu-pied by P1 Phe92. (c) View of thesmall P4 pocket of HLA-DR2a thatis occupied by P4 Ile95. (d) Puta-tive TCR-contacting residues of theMBP peptide in the HLA-DR2a/MBP complex. (e) Top view of theHLA-DR2b/MBP 85-99 complex(Smith et al., 1998). (f) View of thesmall P1 pocket of HLA-DR2b thatis occupied by P1 Val89. (g) Viewof the large P4 pocket of HLA-DR2b that is occupied by P4 Phe92.(h) Putative TCR-contactingresidues of the MBP peptidein the HLA-DR2b/MBP complex.Figures were drawn using GRASP(Nicholls et al., 1991). Blue denotesregions of positive potential andorange regions of negative poten-tial.


Scott et al., 1998; Fremont et al., 1998; Smith et al.,1998) are also observed in the DR2a/MBP com-plex. However, two conserved hydrogen bondsbetween Arg76a and Asp57b and peptide residueP10 are absent in the DR2a/MBP complex becauseof the raised peptide C terminus (Figure 3(a)).These two residues retain their salt bridge beneaththe peptide.

The general peptide-binding motifs proposed forMHC class II molecules de®ne amino acid prefer-ences at position P1, P4, P6, and P9 (Madden,1995). By sequencing naturally processed peptideseluted from HLA-DR2 molecules puri®ed from celllysates, Vogt et al. (1994) found that HLA-DR2arequires a bulky hydrophobic residue (phenyl-alanine or tyrosine) at P1 as a primary anchor and

an aliphatic residue (valine, isoleucine, methionineor glutamine) at P4 as a secondary anchor. Thestructure of the DR2a/MBP complex reveals thebasis for these preferences. In the complex, fourout of 20 peptide residues are completely, oralmost completely, buried in the DR2a bindingpocket: the surface areas of P1 Phe, P4 Ile, P6 Thrand P9 Thr are 98, 92, 96 and 89 %, respectively,inaccessible to solvent, consistent with results frombinding assays using peptide analogs (Vogt et al.,1994; Wucherpfennig et al., 1994). Therefore, asingle phenylalanine-to-alanine substitution at P1almost completely abolishes the binding of MBP85-99 to HLA-DR2a. In the complex, P1 Phe bindsin a deep hydrophobic pocket formed by DR2aresidues Phe24a, Ile31a, Phe32a, Trp43a, Ala52a,

Figure 3. (a) Superposition of peptides bound tohuman MHC class II molecules. The MBP 86-105 pep-tide bound to HLA-DR2a is red and the MBP 85-99 pep-tide bound to HLA-DR2b (Smith et al., 1998) is green. Inyellow are peptides bound to other HLA-DR molecules:DR1 (Stern et al., 1994), DR3 (Ghosh et al., 1995) andDR4 (Dessen et al., 1997). HLA-DR2a is shown in gray.The peptide backbones superpose well from P-2 to P4,but display considerable divergence beginning at P6. Inparticular, the C-terminal portions of the MBP peptidespresented by DR2a and DR2b are positioned higher inthe binding groove than the C-terminal portions of pep-tides presented by DR1, DR3, or DR4. (b) Confor-mations of Trp61b in HLA-DR2a and HLA-DR3. In redare the MBP peptide bound to HLA-DR2a and the side-chains of residues Trp61b and Leu38b of this DR. In yel-low are CLIP peptide bound to HLA-DR3 (Ghosh et al.,1995) and the side-chains of Trp61b and Val38b. HLA-DR2a and HLA-DR3 are gray and blue, respectively.Superpositions were done using the a1 and b1 domainsof the DR molecules.


Phe54a, Asn82b, Val85b, Gly86b and Phe89b(Figure 2(b)), all of which except Gly86b are con-served in DR molecules. This pocket, which canaccommodate large aromatic and non-polar side-chains, is identical to that reported for DR1 (Sternet al., 1994) and DR4 (Dessen et al., 1997). In con-trast, DR2b (Smith et al., 1998) and DR3 (Ghoshet al., 1995) have a relatively shallow P1 pocket,resulting in the selection of smaller aliphaticanchor residues (Figure 2(f)). This change in bind-ing preference is due to the presence of valine,

instead of glycine, at position 86b at the base of theP1 pocket. The P4 pocket of HLA-DR2a, which isoccupied by P4 Ile, is formed by Gln9a, Asn62a,Tyr13b, Tyr26b, Arg71b, Ala74b and Tyr78b(Figure 2(c)). It is similar in size to the P4 pocket ofHLA-DR1, but is less deep than the P4 pocket ofHLA-DR2b, in which residue 71b is alanine(Figure 2(g)).

Residue P6 Thr of the MBP peptide is seques-tered from solvent in a cleft lined by Glu11a,Asn62a, Val65a, Asp66a, Asp11b and Tyr13b. Asnoted in other MHC class II structures (Sternet al., 1994; Ghosh et al., 1995; Fremont et al.,1996; Murthy & Stern, 1997; Dessen et al., 1997;Scott et al., 1998; Fremont et al., 1998), this pock-et is negatively charged due to the presence ofconserved residues Glu11a and Asp66a, whichare thought to determine the pH dependence ofpeptide binding or exchange (Fremont et al.,1996). In DR2a, the P6 pocket is considerablymore acidic than in other DR molecules becauseof polymorphic residue Asp11b, which isencoded by all DRB5 alleles, as well as byDRB1*09012. The net negative charge of the P6pocket implies that negatively-charged peptideresidues at this site are incompatible with highaf®nity binding, in agreement with the sequencesof self-peptides isolated from HLA-DR2 (Vogtet al., 1994). The acidic character of this pocketextends to pockets P7 and P9, creating a cleftthat is negatively charged throughout much ofits length, a unique feature of the DR2a peptide-binding site (compare Figure 2(a) and (e)). Poly-morphic residues Asp30b and Asp37b accountfor this difference. Asp30b is only encoded byDRB5 alleles, whereas other DR alleles encodetyrosine, cysteine, leucine, glycine or histidine atthis position. In addition, although Asp37b isalso present in DRB1*0806, DRB1*1115 andDRB1*1124, the combination of Asp30b andAsp37b, which cluster with Asp11b, is only seenin DRB5 alleles. This may explain the preferencefor lysine or arginine at position P9 amongendogenous peptides bound to HLA-DR2a (Vogtet al., 1994), although P9 is threonine in MBP86-105.

The side-chain of DR2a Asp30b is rotated byabout 100 � relative to that of the correspondingresidue in other DR molecules (Cys in DR1 andTyr in DR2b, DR3 and DR4). As a consequence,the Asp30b side-chain is directed toward the P9,rather than P6, pocket. There, Asp30b forms twohydrogen bonds with the Ne2 group of Gln9b, aswell as hydrogen bonds with two ordered watermolecules that occupy a cavity beneath the MBPpeptide. Except for DRB3*0212, which also encodesa glutamine at position b9, Gln9b is unique toDRB5 alleles. Due to the particular orientation ofAsp30b side-chain, and that of Trp61b (see below),the P7 pocket of the DR2a peptide-binding site isrelatively deep. It is only partly ®lled by P7 Probecause of the elevated position of the C-terminalhalf of the MBP peptide, from P5 to P14. Three


water molecules ®ll the space usually occupied bythe P7 side-chain in other class II structures. Thesewater molecules form a hydrogen-bonding net-work linking the carbonyl oxygen atom of P7 Prowith the Ne2 and NH1 atoms of Arg71b and thehydroxy group of Tyr13b. Water molecules arealso observed in the peptide-binding grooves ofDR1, DR2b and DR4 (Dessen et al., 1997; Murthy &Stern, 1997; Smith et al., 1998), though fewer thanin DR2a; this difference may be an intrinsic featureof the DR2a/MBP complex, or could be attribu-table to the higher resolution of the latter structure.As in other MHC/peptide complexes, these watermolecules serve to increase shape and chemicalcomplementarity by ®lling cavities between theinteracting surfaces and by contributing to thehydrogen bonding network linking peptide andMHC.

Residue P7 Pro packs against Trp61b, Phe67band Arg71b. Although Phe67b and Arg71b arepolymorphic, they (or similar residues) occur fre-quently in DR alleles. In DR2a, Arg71b displays aside-chain conformation similar to those observedfor Arg71b in DR1 (Stern et al., 1994; Murthy &Stern, 1997) and Lys71b in DR3 (Ghosh et al., 1995)and DR4 (Dessen et al., 1997). Residue Trp61b isstrictly conserved in all HLA-DR, -DQ and -DOalleles. However, the side-chain conformation ofTrp61b in DR2a is signi®cantly different fromthose in the known MHC class II structures (Sternet al., 1994; Ghosh et al., 1995; Murthy & Stern,1997; Dessen et al., 1997; Smith et al., 1998), wherethe indole ring points into the peptide-bindingcleft. In contrast, the Trp61b side-chain in theDR2a/MBP complex is rotated about 150 � andprojects upward, creating a large hydrophobicridge between binding pockets P7 and P9(Figure 3(b)). The result is that the MBP 86-105peptide cannot sit as low in the DR2a groove aspeptides bound to DR1, DR3 or DR4; otherwise,the peptide would make sterically unfavorablecontacts as close as 1.6 AÊ . In addition, due to itsunusual side-chain conformation, the hydrogenbond between Trp61b and peptide P8 O observedin other MHC class II/peptide complexes (Sternet al., 1994; Ghosh et al., 1995; Murthy & Stern,1997; Dessen et al., 1997; Smith et al., 1998) isabsent in the DR2a/MBP complex. Because of theraised position of the C-terminal half of the MBPmain-chain, P9 Thr does not fully occupy the avail-able space in the P9 pocket. In addition to the twowater molecules discussed above that form hydro-gen bonds with Asp30b, one additional water ishydrogen bonded to Asp37b Od1 at the base of thispocket, helping to ®ll the cavity beneath thepeptide.

The preference for positively charged residuesat MBP positions P2 and P8 for binding toDR2a is unclear (Vogt et al., 1994;Wucherpfennig et al., 1994). Residues P2 Lys andP8 Arg are not bound in DR2a pockets andtheir side-chain amino and guanidinium groups,respectively, are fully exposed to solvent, such

that they could be recognized by TCR. Inaddition, the P2 and P8 side-chains do not makeany direct hydrogen bonds or salt bridges toDR2a residues. However, the Nz atom of P2 Lysforms a hydrogen bond with a water moleculewhich, in turn, hydrogen bonds to the main-chain carbonyl oxygen of DR2a residue Thr77b.An arginine residue at P2 has been shown to berequired for tight binding of peptides to HLA-DR1 (Hammer et al., 1994). In the crystal struc-ture of the DR1/A2 complex (Murthy & Stern,1997), the guanidinium group of Arg P2 forms adirect hydrogen bond with Thr77b O, whichmay play an important role in stabilizing theDR1 b1 helix. Similar hydrogen bonds betweenpeptide P2 residues and the main-chain carbonyloxygen atom of Thr77b have been observed inthe crystal structures of DR2b (Smith et al.,1998), DR3 (Ghosh et al., 1995), DR4 (Dessenet al., 1997) and I-Ek (Fremont et al., 1996) com-plexed with speci®c peptides. Although the ArgP8 side-chain does not interact directly withDR2a, the side-chain guanidinium group formsseveral solvent-mediated hydrogen bonds toAsn69a Od1 and P9 O, which may contribute tostabilizing the DR2a/MBP complex.

Comparison with the DR2b/MBP complexreveals a shift in peptide binding register

The MBP 86-105 peptide can be presented byboth DRB1*1501 and DRB5*0101 alleles (Vogt et al.,1994; Wucherpfennig et al., 1994). Figure 1(b)shows an alignment of MBP peptides in the DR2a/MBP 86-105 and DR2b/MBP 85-99 complexes(Smith et al., 1998), following least-squares super-position of the a1/b1 domains of the class II mol-ecules. It is apparent that the binding of the MBPpeptide is strikingly different in DR2a compared toDR2b, such that the peptide register is shifted bythree residues (Figure 1(c)). Therefore, the MBPpeptide binds DR2a using Phe92 and Ile95 as pri-mary anchor residues in pockets P1 and P4,respectively. In contrast, the MBP peptide bindsDR2b with anchor residues Val89 in the P1 pocketand Phe92 in the P4 pocket (Smith et al., 1998). TheHLA-DR2b molecule has a relatively shallow P1pocket (Figure 2(f)), which can accommodate ali-phatic amino acids such as leucine and valine, anda large hydrophobic P4 pocket (Figure 2(g)), whichselects for bulky aromatic anchor residues, particu-larly tyrosine and phenylanine (Vogt et al., 1994);MBP residues Val89 and Phe92 ful®ll these require-ments. Conversely, HLA-DR2a displays a large P1pocket (Figure 2(b)) and a relatively shallow P4pocket (Figure 2(c)). As a consequence, the MBPpeptide must undergo a three-residue shift inorder to place a bulky aromatic residue (Phe92) inpocket P1 and a less bulky aliphatic residue (Ile95)in pocket P4. The change in volume of the P1 pock-et is attributable to a Gly/Val dimorphism at pos-ition 86 of the DR b chain, as discussed above,while the volume difference between P4 pockets in


DR2a and DR2b is due to the polymorphism atposition 71b of DR molecules. In HLA-DR2b, analanine residue is encoded at position 71b, whichcontributes to the formation of a large, primarilyhydrophobic P4 pocket lined by additional DRresidues Arg13b, Phe26b, Gln70b, Ala74b, Tyr78b,and Asn62a. In contrast, DR2a encodes an arginineresidue at 71b, whose side-chain is directedtowards the P4 pocket, greatly reducing its size. Itis also worth noting that the guanidinium group ofpolymorphic residue Arg13b in DRB1*1501 formstwo hydrogen bonds to the Od1 atom of MBP resi-due P6 Asn (Smith et al., 1998), which are absent inthe DR2a/MBP complex. These two additionalinteractions may contribute to the approximatelytenfold higher af®nity of MBP 84-102 for DR2bthan for DR2a (Wucherpfennig et al., 1994), sincecoupling energies up to 4 kcal/mol have beenmeasured for hydrogen bonds involving charged-neutral pairs (Fersht & Serrano, 1993).

As shown in Figure 1(b), the main-chain con-formations of the shifted MBP peptides in theDR2a and DR2b structures are very similarbetween residues P1 and P5. The differences inpositions of the a-carbon atoms of these residuesare <0.50 AÊ , with a minimum displacement of0.08 AÊ in the position of the P4 a-carbon. Thepeptides begin to arch up at position P6 in bothcomplexes (Figure 3(a)). In the DR2b/MBP com-plex, peptide residue P6 Asn binds to pocket P6,where Arg13b is thought to be at least partiallyresponsible for the raised position of the MBPC terminus within the binding groove (Smithet al., 1998). In the DR2a/MBP complex, P6 Throccupies the P6 pocket. The position of theTyr13b side-chain in DR2a is similar to that ofArg13b in DR2b. The side-chain of P6 Thr of theMBP peptide in the DR2a complex is alsooriented similarly to that of P6 Asn in the DR2bcomplex. Because the threonine side-chain is onecarbon atom shorter than that of asparagine, thea-carbon atom of P6 Thr lies 1.2 AÊ deeper in theDR2a binding groove than the a-carbon of P6Asn in the DR2b groove. Conversely, the a-car-bon position of P8 Arg in DR2a is 1.1 AÊ higherthan that of P8 Thr in DR2b due to the uniqueside-chain conformation of Trp61b, which createsa prominent ridge between the P7 and P9 pock-ets of DR2a.

It should be noted that register shifting as ameans by which certain peptides can be presentedby multiple MHC molecules is only possible forMHC class II, since the MHC class I bindinggroove is closed off at both ends, with the N andC-terminal residues of the peptide serving asanchors (Madden, 1995). Therefore, peptides cross-presented by different class I alleles (e.g. P18-I10 ofHIV gp120 that can be presented by H-2Dd, H-2Dp

and H-2Du (Shirai et al., 1997)) must bind in thesame register, though not necessarily in the sameconformation.

Binding of MBP 86-105 to other HLA-DR alleles

In addition to DR5*0101 (DR2a) and DR1*1501(DR2b), MBP 86-105 has been shown to bind toDRB1*0101 (DR1), DRB1*0401 (DR4), DRB1*0404,DRB1*0701, and DRB1*1101 (Vogt et al., 1994).However, only DRB1*1501 and DRB5*0101 occurat increased frequency in MS patients, where theyare believed to confer disease susceptibility byin¯uencing T cell repertoire selection, or by pre-senting self-peptides against which an autoimmuneresponse is directed (Steinman, 1996). An alanineat polymorphic position 71b is responsible for theincreased size of the P4 pocket in DR1*1501 (Smithet al., 1998) (Figure 2(g)), compared to other DRalleles able to present MBP 86-105, where aminoacid 71b is arginine (DRB1*0101, DRB1*0404,DRB1*0701, DRB1*1101, DRB5*0101) or lysine(DRB1*0401). The bulky arginine or lysine side-chain should prevent the MBP peptide frombinding to these alleles in the same register as toDR2b. Most likely, therefore, DRB1*0101,DRB1*0401, DRB1*0404, DRB1*0701 andDRB1*1101 bind MBP 86-105 in the same align-ment as DRB5*0101, since all are predicted (orhave been shown) to display a large hydrophobicP1 pocket and a relatively shallow P4 pocket.Compared to peptides presented by DR1 (Sternet al., 1994; Murthy & Stern, 1997), DR3 (Ghoshet al., 1995) and DR4 (Dessen et al., 1997), the MBPpeptides presented by DR2a and DR2b displayelevated C-terminal portions, starting at P5(Figure 3(a)). As discussed above, the raisedposition of the peptide C terminus in the DR2a/MBP complex is the result of a rotation of the side-chain of Trp61b, a strictly conserved residue in allDR alleles. In the DR2b/MBP complex, on theother hand, the raised peptide C terminus is due tothe side-chain of polymorphic residue Arg13b (Tyrin DR2a), which prevents P6 Asn from binding asdeeply in the P6 pocket as the correspondingresidue in other DR/peptide complexes (Smithet al., 1998). The rotated Trp61b side-chain in DR2anot only lifts the MBP peptide higher in thebinding groove, it displaces it laterally, closer tothe MHC a chain. A comparison of known DRstructures reveals that the change in conformationof the Trp61b side-chain is attributable to asubstitution at position 38b, which is completelyburied in the binding site. Whereas HLA-DR2aencodes a leucine at 38b, all other DR allelesknown to bind MBP 86-105, including HLA-DR2b,encode a valine at this position. A reorientation ofthe Trp61b side-chain in DR2a is necessary inorder to avoid a steric clash with Leu38b(Figure 3(b)). Therefore, we predict that MBP86-105 will bind DRB1*0101, DRB1*0401,DRB1*0701, and DRB1*1101 with its C terminuspositioned deeper in the groove than in theDR2a/MBP and DR2b/MBP complexes, asobserved for peptides bound to DR1 (Stern et al.,1994; Murthy & Stern, 1997), DR3 (Ghosh et al.,1995) and DR4 (Dessen et al., 1997). The association


of MS development with DRB1*1501 andDRB5*0101 may involve TCR recognition of theunique C-terminal conformation of self-peptidesfrom myelin/oligodendrocyte antigens, such asMBP 86-105, when presented by these two HLA-DR alleles.

Interactions with TCR

Although no three-dimensional structure of aTCR/MBP/DR2 ternary complex has beenreported, the structure of a TCR complexed with aconalbumin peptide (CA) bound to the mouseMHC class II molecule I-Ak has been determined(Reinherz et al., 1999). Assuming that the dockingmode observed in the TCR/CA/I-Ek complex isconserved in other TCR/peptide/MHC class IIcomplexes, a model of a TCR/MBP/DR2a complex(not shown) may be constructed by least-squaressuperposition of the a1 and b1 domains of HLA-DR2a onto those of I-Ak in the TCR/CA/I-Ak com-plex; a TCR/MBP/DR2b complex may be modeledsimilarly. Due to the change in peptide alignmentbetween DR2a and DR2b, the predicted TCR-con-tacting residues are very different in the two com-plexes. In the DR2a complex, MBP residues Lys93,Asn94, Val96, Thr97, Pro98 and Arg99 are recog-nized by the TCR (Figure 2(d)), whereas His90,Phe91, Lys93, Asn94, Ile95 and Val96 contact TCRin the DR2b complex (Figure 2(h)). Therefore, onlythree putative TCR-contacting residues (Lys93,Asn94 and Val96) are common to both complexes.This is consistent with the ®nding that T cell linesspeci®c for MBP 84-102 presented by HLA-DR2afail to recognize this peptide in the context ofHLA-DR2b, and vice versa (Martin et al., 1990; Petteet al., 1990; Vogt et al., 1994; Wucherpfennig et al.,1994; Vergelli et al., 1997). Therefore, distinct Tcell epitopes may arise from the imposition ofdifferent alignments on the same bound peptideby MHC molecules with topologically distinctpeptide-binding grooves.

Materials and Methods

Construction of truncated genes encoding theHLA-DR2a aaa and bbb subunits

The HLA-DR2a b chain (DRB5*0101) gene was modi-®ed for T7 polymerase-directed expression of its extra-cellular domain in E. coli as described for the HLA-DR1b chain (Frayser et al., 1999). Oligonucleotide primers forPCR were designed to incorporate an EcoRI restrictionenzyme site, ribosome binding site (AGGAGG), spacersequence (AATTTAAA), and initiation codon at the 50end of the gene; a stop codon (TAA) and HindIII restric-tion site were incorporated at the 30 end. The resultingPCR product, encoding DR2a b chain residues Gly1 toSer192, was cloned immediately after the T7 promoter ofthe expression vector pLM1 (MacFerrin et al., 1990). Asimilar pLM1-based construct encoding extracellular resi-dues Ile1 to Ala181 of the non-polymorphic HLA-DR1 achain (DRA*0101) was kindly provided by L.J. Stern

(Massachesetts Institute of Technology) (Frayser et al.,1999).

Expression and purification of aaa and bbbchain polypeptides

For protein expression, plasmids encoding the HLA-DR2a a and b polypeptides were separately transformedinto E. coli strain BL21(DE3)pLysS. Individual colonieswere grown in Luria-Bertani (LB) medium containing60 mg/ml ampicillin, 35 mg/ml chloramphenicol, and2 mg/ml glucose. The bacteria were grown at 37 �C toan absorbance of 0.6-0.8 at 600 nm and isopropyl-b-D-thiogalactoside was added to a ®nal concentration of1 mM. After ®ve hours of further incubation, the cellswere harvested by centrifugation, lysed by French press,and pelleted at 10,000 g for 20 minutes. The inclusionbody pellets were washed three times by suspendingwith Dounce homogenizer in 50 mM Tris-HCl, 0.5 %(v/v) Triton X-100, 1 mM EDTA (pH 8.0), and once inthe same solution without Triton X-100. The washedinclusion bodies were stored at ÿ80 �C, or dissolvedimmediately in 50 mM Tris-HCl, 8 M urea, 10 mM DTT,0.5 mM EDTA (pH 8.0), and centrifuged at 30,000 g for45 minutes at 4 �C. The a and b polypeptides were puri-®ed by anion exchange (Frayser et al., 1999) on a PorosHQ20 column (Perspective Biosystems) in 50 mM Tris-HCl, 8 M urea, 1 mM DTT (pH 8.0) (DRa) or (pH 9.0)(DRb), with a linear gradient over ten column volumesof 0 to 400 mM NaCl in the same buffer. Fractions con-taining DRa or DRb by SDS-PAGE were pooled andconcentrated to approximately 6 mg/ml; 0.5 mM EDTAwas added and the subunits stored at ÿ80 �C.

In vitro folding and purification of the HLA-DR2a/MBP complex

Puri®ed subunits were thawed and diluted to a ®nalconcentration of 30 mg/ml each in a folding solution con-taining 50 mM Tris-HCl, 25 % (w/v) glycerol, 0.5 mMEDTA, 3 mM reduced glutathione, and 0.9 mM oxidizedglutathione (pH 8.0). Peptide MBP 86-105 (Multiple Pep-tide Systems) was added to a ®nal concentration of5 mM, and the folding mixture was kept for two weeksat 4 �C. These folding conditions are similar to thosedescribed for HLA-DR1, but were optimized forHLA-DR2a. The amount of folded DR2a in the mixturewas monitored with a capture ELISA using monoclonalantibody L243, speci®c for the folded conformation,as previously described (Stern & Wiley, 1992; Frayseret al., 1999).

The ®nal folding solution was concentrated tenfoldusing an Amicon RA2000 concentrator and applied to anL243 af®nity column equilibrated with phosphate-buffered saline (PBS). The column was washed with 50volumes of PBS and eluted with 50 mM CAPS-HCl,pH 11 (Stern & Wiley, 1992; Frayser et al., 1999). Frac-tions were immediately neutralized with 1 M Tris-HCl(pH 7.0), and those containing DR2a were pooled andbuffer-exchanged with 50 mM Tris-HCl (pH 8.0). Furtherpuri®cation was carried out on a Mono Q anionexchange FPLC column (Pharmacia) equilibrated withthe same buffer and developed with a linear NaCl gradi-ent. Recombinant HLA-DR2a/MBP 86-105 eluted as twowell-resolved peaks at about 0.3 M NaCl, both of whichbound L243 and crystallized; material from the secondpeak was used for further analysis.


Crystallization, data collection andstructure determination

Initial crystallization conditions for the HLA-DR2a/MBP 86-105 complex were established using HamptonCrystal Screen Kit I (Hampton Research). Thinplate-shaped crystals, with dimensions up to0.3 mm � 0.5 mm � 0.01 mm, were obtained at roomtemperature in hanging drops by mixing equalvolumes of the complex solution at 5 mg/ml and awell solution containing a 2:3 dilution of Formula 15of the screen (30 % (w/v) PEG 8000, 0.1 M sodiumcacodylate, pH 6.5, 0.2 M ammonium sulfate). Fordata collection, crystals were washed three times withmother liquor and then transferred to a cryoprotectantsolution (mother liquor containing 20 % glycerol), priorto ¯ash-cooling in a nitrogen stream. X-ray diffractiondata were measured from one complex crystal at100 K on beamline X12B (l � 0.978 AÊ ) of the Brookha-ven National Synchrotron Laboratory using a Quan-tum-4 four-module CCD-based detector (Area DetectorSystems). The DR2a/MBP 86-105 complex crystallizesin space group P21 with cell dimensions a � 63.2 AÊ ,b � 114.9 AÊ , c � 63.2 AÊ , b � 90.9 � and two complexmolecules per asymmetric unit. Diffraction data wereprocessed and scaled using DENZO/SCALEPACK(Otwinowski & Minor, 1997), followed by datareduction using programs from the CCP4 suite (Com-putational Collaborative Project No. 4, 1994). The dataset is 90.8 % complete to 1.9 AÊ resolution with Rsym�8.1 %. Data collection statistics are shown in Table 1.

The structure was solved by the molecular replace-ment method with the program AMoRe (Navaza, 1994),using the HLA-DR2b MHC class II molecule (Smith et al.,1998), stripped of the peptide, as a search model. Afterrigid body re®nement, iterative cycles of re®nementincluding simulated annealing, positional re®nement,and temperature factor (B) re®nement with a maximumlikelihood target were carried out using CNS0.4(Brunger et al., 1998). A randomly sampled Rfree set con-sisting of 5 % of the collected data was set aside forcross-validation and sA weighting. A bulk solventcorrection was calculated to allow the inclusion of lowresolution diffraction data in the re®nement procedure.Non-crystallographic symmetry restraints were usedthroughout the re®nement and released only for the ®nalcycles. Gradually, the HLA-DR2a polymorphic residuesand the MBP peptide were manually built into themodel according to sA-weighted Fo ÿ Fc and 2Fo ÿ Fc

electron density maps using TURBO-FRODO (Roussel& Cambillau, 1989). The side-chains of 30 DR2a residueswith no electron density were truncated to ®t thedensity; none of these residues is located in the peptide-binding groove. The ®nal model includes 6897 non-hydrogen protein atoms, 454 water molecules, twoglycerol molecules, and three sulfates, resulting in a ®nalR-value of 0.233 with a Rfree of 0.267 at 1.9 AÊ resolution.Model geometry was examined using PROCHECK(Laskowski et al., 1993). The program AREAIMOL in theCCP4 suite was used to analyze solvent accessibility.The re®nement statistics are summarized in Table 1.

Acknowledgements

This work was supported by grants from the NationalMultiple Sclerosis Society (RG2747) and the National

Institutes of Health (AI36900). We thank L.J. Stern (Mas-sachusetts Institute of Technology) for discussions andfor the gift of HLA-DR1 expression plasmids. We arealso grateful to E.O. Long (National Institutes of Health)for providing HLA-DR2a cDNAs. Atomic coordinateshave been deposited in the Protein Data Bank as entry1fv.

References

Bielekova, B., Goodwin, B., Richert, N., Cortese, I.,Kondo, T. & Afshar, G., et al. (2000). Encephalito-genic potential of myelin basic protein peptide (83-99) in multiple sclerosis: results of a phase II clinicaltrial with an altered peptide ligand. Nature Med. 6,1167-1175.

Brown, J. H., Jardetzky, T. S., Gorga, J. C., Stern, L. J.,Urban, R. G., Strominger, J. L. & Wiley, D. C.(1993). Three-dimensional structure of the humanclass II histocompatibility antigen HLA-DR1.Nature, 364, 33-39.

Brunger, A. T., Adams, P. D., Clore, G. M., DeLano,W. L., Gros, P. & Grosse-Kunstleve, R. W., et al.(1998). Crystallography & NMR system: a new soft-ware suite for macromolecular structure determi-nation. Acta Crystallog. sect. D, 54, 905-921.

Collaborative Computational Project No. 4 (1994). TheCCP4 suite: programs for protein crystallography.Acta Crystallog. sect. D, 50, 760-763.

Dessen, A., Lawrence, C. M., Cupo, S., Zaller, D. M. &Wiley, D. C. (1997). X-ray crystal structure of HLA-DR4 (DRA*0101, DRB*0401) complexed with a pep-tide from human collagen II. Immunity, 7, 473-481.

Doolan, D. L., Southwood, S., Chestnut, R., Appella, E.,Gomez, E. & Richatds, A., et al. (2000). HLA-DR-promiscuous T cell epitopes from Plasmodium falci-parum pre-erythrocytic-stage antigens restricted bymultiple HLA class II alleles. J. Immunol. 165, 1123-1137.

Ebers, G. C., Kukay, K., Bulman, D. E., Sadovnick, A. D.,Rice, G. & Anderson, C., et al. (1996). A full genomesearch in multiple sclerosis. Nature Genet. 13, 472-476.

Fersht, A. R. & Serrano, L. (1993). Principles of proteinstability derived from protein engineering exper-iments. Curr. Opin. Struct. Biol. 3, 75-83.

Frayser, M., Sato, A. K., Xu, L. & Stern, L. J. (1999).Empty and peptide-loaded class II MHC proteinsproduced by expression in E. coli and refoldingin vitro. Protein Expr. Purif. 15, 105-114.

Fremont, D. H., Hendrickson, W. A., Marrack, P. &Kappler, J. (1996). Structures of an MHC class IImolecule with covalently bound single peptides.Science, 272, 1001-1004.

Fremont, D. H., Monnaie, D., Nelson, C. A.,Hendrickson, W. A. & Unanue, E. R. (1998). Crystalstructure of I-Ak in complex with a dominantepitope of lysozyme. Immunity, 8, 305-317.

Ghosh, P., Amaya, M., Mellins, E. & Wiley, D. C. (1995).The structure of an intermediate in class II MHCmaturation: CLIP bound to HLA-DR3. Nature, 378,457-462.

Hammer, J., Valsasnini, P., Tolba, K., Bolin, D., Higelin,J., Takacs, B. & Sinigaglia, F. (1993). Promiscuousand allele-speci®c anchors in HLA-DR-binding pep-tides. Cell, 74, 197-203.

Hammer, J., Belunis, C., Bolin, D., Papadopoulos, J.,Walsky, R. & Higelin, J., et al. (1994). High-


af®nity binding of short peptides to major histo-compatibility complex class II molecules byanchor combinations. Proc. Natl Acad. Sci. USA,91, 4456-4460.

Hillert, J., Kall, T., Vrethem, M., Fredrikson, S., Ohlson,M. & Olerup, O. (1994). The HLA-Dw2 haplotypesegregates closely with multiple sclerosis in multi-plex families. J. Neuroimmunol. 50, 95-100.

Jaraquemada, D., Martin, R., Rosen-Bronson, S.,Flerlage, M., McFarland, H. F. & Long, E. O. (1990).HLA-DR2a is the dominant restriction molecule forthe cytotoxic T cell response to myelin basic proteinin DR2Dw2 individuals. J. Immunol. 145, 2880-2885.

Krogsgaard, M., Wucherpfennig, K. W., Canella, B.,Hansen, B. E., Svejgaard, A. & Pyrdol, J., et al.(2000). Visualization of myelin basic protein (MBP)T cell epitopes in multiple sclerosis lesions using amonoclonal antibody speci®c for the human histo-compatibility leukocyte antigen (HLA)-DR2-MBP85-99 complex. J. Exp. Med. 191, 1395-1412.

Laskowski, R. A., MacArthur, M. W., Moss, D. S. &Thornton, J. M. (1993). PROCHECK: a program tocheck the stereochemical quality of protein struc-tures. J. Appl. Crystallog. 26, 283-291.

MacFerrin, K. D., Terranova, M. P., Schreiber, S. L. &Verdine, G. L. (1990). Overproduction and dissec-tion of proteins by the expression-cassette polymer-ase chain reaction. Proc. Natl Acad. Sci. USA, 87,1937-1941.

Madden, D. (1995). The three-dimensional structure ofpeptide-MHC complexes. Annu. Rev. Immunol. 13,587-622.

Madsen, L. S., Andersson, E. C., Jansson, L.,Krogsgaard, M., Andersen, C. B. & Engberg, J., et al.(1999). A humanized model for multiple sclerosisutilizing HLA-DR2 and a human T cell receptor.Nature Genet. 23, 343-347.

Martin, R., Jaraquemada, D., Flerlage, M., Richert, J.,Whitaker, J., Long, E. O., McFarlin, D. E. &McFarland, H. F. (1990). Fine speci®city and HLArestriction of myelin basic protein-speci®c cytotoxicT cell lines from multiple sclerosis patients andhealthy individuals. J. Immunol. 145, 540-548.

Murthy, V. L. & Stern, L. J. (1997). The class II MHCprotein HLA-DR1 in complex with an endogenouspeptide: implications for the structural basis of thespeci®city of peptide binding. Structure, 5, 1385-1396.

Navaza, J. (1994). AMoRe: an automated package formolecular replacement. Acta Crystallog. sect. A, 50,157-163.

Nicholls, A., Sharp, K. A. & Honig, B. (1991). Proteinfolding and association: insights from the interfacialand thermodynamic properties of hydrocarbons.Proteins: Struct. Funct. Genet. 11, 281-296.

Oksenberg, J. R., Seboun, E. & Hauser, S. L. (1996).Genetics of demyelinating diseases. Brain Pathol. 6,289-302.

O'Sullivan, D., Arrhenius, T., Sidney, J., Guercio, M.,Albertson, M. & Wall, M., et al. (1991). On the inter-action of promiscuous antigenic peptides withdifferent DR alleles. J. Immunol. 147, 2663-2669.

Ota, K., Matsui, M., Milford, E. L., Mackin, G. A.,Weiner, H. L. & Ha¯er, D. A. (1990). T-cell recog-nition of an immunodominant myelin basic proteinepitope in multiple sclerosis. Nature, 346, 183-187.

Otwinowski, Z. & Minor, W. (1997). Processing X-raydiffraction data collected in oscillation mode.Methods Enzymol. 276, 307-326.

Pette, M., Fujita, K., Wilkinson, D., Altmann, D. M.,Trowsdale, J. & Giegerich, G., et al. (1990). Myelinautoreactivity in multiple sclerosis: recognition ofmyelin basis protein in the context of HLA-DR2products by T lymphocytes of multiple sclerosispatients and healthy donors. Proc. Natl Acad. Sci.USA, 87, 7968-7972.

Reinherz, E. L., Tan, K., Tang, L., Kern, P., Liu, J. &Xiong, Y., et al. (1999). The crystal structure of a Tcell receptor in complex with peptide and MHCclass II. Science, 286, 1913-1921.

Roussel, A. & Cambillau, C. (1989). TURBO-FRODO. InSilicon Graphics Geometry Partners Directory, pp. 77-78, Silicon Graphics, Mountain View, CA.

Scott, C. A., Peterson, P. A., Teyton, L. & Wilson, I. A.(1998). Crystal structures of two I-Ad-peptide com-plexes reveal that high af®nity can be achievedwithout large anchor residues. Immunity, 8, 319-329.

Shirai, M., Kozlowski, S., Margulies, D. H. & Berzofsky,J. A. (1997). Degenerate MHC restriction reveals thecontribution of class I MHC molecules in determin-ing the speci®city of CTL recognition of an immu-nodominant determinant of HIV-1 gp160 V3 loop.J. Immunol. 158, 3181-3188.

Sinigaglia, F., Guttinger, M., Kilgus, J., Doran, D. M.& Matile, H., et al. (1988). A malarial T-cell epi-tope recognized in association with most mouseand human MHC class II molecules. Nature, 336,778-780.

Smith, K. J., Pyrdol, J., Gauthier, L., Wiley, D. C. &Wucherpfennig., K. W. (1998). Crystal structure ofHLA-DR2 (DRA*0101, DRB1*1501) complexed witha peptide from human myelin basic protein. J. Exp.Med. 188, 1511-1520.

Sone, T., Tsukamoto, K., Hirayama, K., Nishimura, Y.,Takenouchi, T., Aizawa, M. & Sasazuki, T. (1985).Two distinct class II molecules encoded by thegenes within HLA-DR subregion of HLA-Dw2 andDw12 can act as stimulating and restriction mol-ecules. J. Immunol. 135, 1288-1298.

Steinman, L. (1996). Multiple sclerosis: a coordinatedimmunological attack against myelin in the centralnervous system. Cell, 85, 299-302.

Stern, L. J. & Wiley, D. C. (1992). The human class IIMHC protein HLA-DR1 assembles as empty abheterodimers in the absence of antigenic peptide.Cell, 68, 465-477.

Stern, L. J., Brown, J. H., Jardetzky, T. S., Urban, R.,Strominger, J. L. & Wiley, D. C. (1994). Crystalstructure of the human class II MHC protein HLA-DR1 complexed with an in¯uenza virus peptide.Nature, 368, 215-221.

Stockel, J., Meinl, E., Hahnel, C., Malotka, J., Seitz, R.,Drexler, K., Wekerle, H. & Dornmair, K. (1994).Refolding of human class II major histocompatibil-ity complex molecules isolated from Escherichia coli.J. Biol. Chem. 269, 29571-29578.

Valli, A., Sette, A., Kappos, L., Oseroff, C., Sidney, J. &Miescher, G., et al. (1993). Binding of myelin basicprotein peptides to human histocompatibility anti-gen class II molecules and their recognition by Tcells from multiple sclerosis patients. J. Clin. Invest.91, 616-628.

Vergelli, M., Kalbus, M., Rojo, S. C., Hemmer, B.,Kalbacher, H. & Tranquill, L., et al. (1997). T cellresponse to myelin basic protein in the context of themultiple sclerosis-associated HLA-DR15 haplotype:


peptide binding, immunodominance and effectorfunctions of T cells. J. Neuroimunol. 77, 195-203.

Vogt, A. B., Kropshofer, H., Kalbacher, H., Kalbus, M.,Rammensee, H.-G., Coligan, J. E. & Martin, R.(1994). Ligand motifs of HLA-DRB5*0101 andDRB1*1501 molecules delineated from self-peptides.J. Immunol. 153, 1665-1673.

Wucherpfennig, K. W., Sette, A., Southwood, S.,Oseroff, C., Matsui, M., Strominger, J. L. &Ha¯er, D. A. (1994). Structural requirements forbinding of an immunodominant myelin basic pro-tein peptide to DR2 isotypes and for its recog-nition by human T cell clones. J. Exp. Med. 179,279-290.

Edited by I. Wilson

(Received 28 July 2000; received in revised form 27 September 2000; accepted 28 September 2000)

Structural basis for the binding of an immunodominant peptide from myelin basic protein in different...

Documents

Transcript of Structural basis for the binding of an immunodominant peptide from myelin basic protein in different...