REVIEWS

MHC class II proteins and disease:

a structural perspective
E. Yvonne Jones*, Lars Fugger‡§, Jack L. Strominger|| and Christian Siebold*
Abstract | MHC class II molecules on the surface of antigen-presenting cells display a range
of peptides for recognition by the T-cell receptors of CD4+ T helper cells. Therefore, MHC
class II molecules are central to effective adaptive immune responses, but conversely,
genetic and epidemiological data have implicated these molecules in the pathogenesis of
autoimmune diseases. Indeed, the strength of the associations between particular MHC
class II alleles and disease render them the main genetic risk factors for autoimmune
disorders such as type 1 diabetes. Here, we discuss the insights that the crystal structures
of MHC class II molecules provide into the molecular mechanisms by which sequence
polymorphisms might contribute to disease susceptibility.

Linkage disequilibrium
Two genetic factors are in
linkage disequilibrium when
the frequency with which they
occur together in a population
departs from random
expectation, as calculated by
the product of their individual
frequencies.
*Division of Structural
Biology, The Henry Wellcome
Building for Genomic
Medicine, The University
of Oxford, Roosevelt Drive,
Oxford, OX3 7BN, UK.
‡
Department of Clinical
Immunology, Aarhus
University Hospital, Skejby
Sygehus, DK-8200 N Aarhus,
Denmark.
§
MRC Human Immunology
Unit, Weatherall Institute
of Molecular Medicine,
John Radcliffe Hospital,
The University of Oxford,
Oxford, OX3 9DS, UK.
||
Department of Molecular
and Cellular Biology, Harvard
University, Cambridge,
Massachusetts 02138, USA.
Correspondence to E.Y.J.
and L.F.
e-mails:
yvonne@strubi.ox.ac.uk;
lars.fugger@molecular-
medicine.oxford.ac.uk
doi:10.1038/nri1805
A large number of chronic inflammatory diseases are
associated with genes in the MHC class II region. For
many of these diseases, the MHC class II association is
the main genetic association; indeed, for some diseases,
it is the only association that is confirmed. For most of
the MHC-class-II-associated diseases it has been diffi-
cult to unequivocally determine the primary disease-risk
gene(s), owing to the extended linkage disequilibrium in
the MHC region. However, recent genetic and functional
studies support the long-held assumption that common
MHC class II alleles themselves are responsible for the
disease associations (TABLE 1).
The molecular mechanisms by which MHC class II
sequence polymorphisms confer susceptibility to
chronic inflammatory diseases are still unclear. This
situation is further complicated by the fact that, for most
diseases, it is unknown which autoantigens (presented
by the disease-associated MHC class II molecules) are
primarily involved. Until recently, disease-associated
polymorphisms in the MHC alleles were analysed in
the one-dimensional context of linear sequences, even
though they function in the three-dimensional context
of the disease-associated protein. By integrating genetic,
epidemiological and structural analyses, it is now possible
to gain a deeper understanding of how disease-associated
MHC class II alleles might exert their pathogenic role.
This approach has the potential to uncover the relative
contribution of individual MHC class II polymorphisms,
their potential interplay and the significance of addi-
tive effects from individually subtle allelic variations.
An understanding of the MHC class II association at a
structural level therefore provides an effective tool with
which to begin to probe the molecular mechanisms that
underlie these disease pathways and prompt the devel-
opment of new therapies. In this Review, we assess the
power of this approach.
Analysis of a structural database
MHC class II molecules are expressed on the surface
of epithelial cells in the thymus and on professional
antigen-presenting cells in the periphery. These mol-
ecules consist of two non-covalently associated poly-
peptide chains: the α-chain and the β-chain. In addition
to their extracellular regions, both chains have a single
transmembrane sequence and a short cytoplasmic tail.
The amino (N)-terminal α1 and β1 regions of the chains
combine to form a membrane-distal peptide-binding
domain, whereas the remaining extracellular portions
of the two chains each form a membrane-proximal
immunoglobulin-like domain (α2 and β2 regions). The
peptide-binding domain consists of a groove, with a floor
provided by a β-sheet, and two walls, each formed from
an α-helix. This binding groove presents peptides to
CD4+ T cells, providing the mechanism by which MHC
class II molecules function in the maintenance of self-
tolerance and in the induction and regulation of adaptive
immune responses against invading pathogens. Since
the first MHC class II crystal structure was reported in
1993 by Brown et al.1, more than 20 MHC class II crystal
structures have been solved (these are freely available
from the RCSB Protein Data Bank). These structures
include examples of the human MHC class II isotypes
HLA-DR and HLA-DQ and the equivalent mouse iso-
types H2-IE and H2-IA. The peptide-binding grooves

NATURE REVIEWS | IMMUNOLOGY VOLUME 6 | APRIL 2006 | 271

© 2006 Nature Publishing Group
REVIEWS

Table 1 | Disease-associated MHC class II molecules class II molecules show marked deviations from the
canonical peptide-backbone conformation for positions
MHC class II molecules grouped MHC class II alleles Disease P6–P9 of the peptide (FIG. 1c). Other peptide side chains
by disease association association*
make contacts with the binding groove, most notably
Narcolepsy at position P7, which is directed sideways within the
HLA-DQ6.1 HLA-DQA1*0102/DQB1*06011 Negative groove and can therefore be described as binding to a
HLA-DQ6.2 HLA-DQA1*0102/DQB1*0602 Positive P7 pocket. However, for this Review, we focus on the
deeper P1, P4, P6 and P9 pockets that completely bury
Coeliac disease
the peptide side chains (FIG. 1d).
HLA-DQ2 HLA-DQA1*0501/DQB1*0201 Positive In this Review, we describe the properties of the pock-
HLA-DQ8 HLA-DQA1*0301/DQB1*0302 Positive ets for the different MHC class II isoforms and highlight
Type 1 diabetes the links between these properties and the disease asso-
ciations listed in TABLE 1. A systematic characterization
HLA-DQ2 HLA-DQA1*0501/DQB1*0201 Positive of pocket properties (for example volume, hydrophobic-
HLA-DR4.1 HLA-DRA1*0101/DRB1*0401 Positive ity and electrostatic charge) highlights the differences
HLA-DR4.3 HLA-DRA1*0101/DRB1*0403 Negative in their ability to accommodate specific amino-acid
HLA-DR4.5 HLA-DRA1*0101/DRB1*0405 Positive
side chains (that is, anchor specificity). The distinctive
properties of the four pockets define the population of
HLA-DQ6 HLA-DQA1*0102/DQB1*0602 Negative peptides that can be presented by each isotype. Here,
HLA-DQ8 HLA-DQA1*0301/DQB1*0302 Positive we show that by relating these properties to emergent
Rheumatoid arthritis patterns of autoimmune disease association and that by
pooling all of the available data, previously indiscernible
HLA-DR1 HLA-DRA1*0101/DRB1*0101 Positive
themes are becoming obvious.
HLA-DR4.1 HLA-DRA1*0101/DRB1*0401 Positive
HLA-DR4.2 HLA-DRA1*0101/DRB1*0402 Neutral or HLA-DQ and narcolepsy
negative Narcolepsy is a chronic neurological disorder with an
Multiple sclerosis unknown aetiology. It affects 0.02–0.05% of the popula-
HLA-DR2a HLA-DRA5*0101/DRB5*0101 Positive tion, and is mainly characterized by excessive daytime
sleepiness and disturbed nighttime sleep6. Narcolepsy is
HLA-DR2b HLA-DRA1*0101/DRB1*1501 Positive a complex disease7, and the only identified genetic risk
HLA-DQ6.2 HLA-DQA1*0102/DQB1*0602 Positive factor is the HLA-DQB1*0602 allele, which is carried by
*Positive association means that the associated MHC class II molecule increases susceptibility to 90–100% of patients with the most common form of the
the disease; negative association means that the associated MHC class II molecule protects disease8. Because the closely related HLA-DQB1*06011
against the disease; neutral association means that the associated MHC class II molecule has no
effect on susceptibility to the disease. allele protects against narcolepsy9, it is possible that
peptide-binding differences between these two allelic
of these MHC class II molecules are superimposable products determine whether they confer risk to or protect
(FIG. 1a), and as predicted from the first peptide–MHC against narcolepsy (TABLE 1).
class II structures2, such superposition (both within Given the strong HLA-DQB1*0602 association,
and between MHC class II isotypes and species) shows narcolepsy is thought to be autoimmune mediated,
that the conformation of the peptide backbone is highly but the target of the HLA-DQ6.2-restricted auto-
conserved (FIG. 1b). The peptide (for residues from posi- immune response is unknown. The neurotransmitter
tion P1 to P9) is bound in the groove in an extended hypocretin, which is secreted by neurons in the hypo-
conformation that is imposed by hydrogen bonding thalamus, represents a good candidate target for three
from conserved MHC class II residues (in the walls of reasons: hypocretin neurotransmission is important for
the groove) to the peptide backbone. Therefore, the sleep and wakefulness10, the amount of hypocretin is
conformation adopted by the bound peptide is inde- reduced in cerebrospinal fluid from patients with narco-
pendent of its sequence and bears no relationship to lepsy11, and this seems to be a result of the absence of
the conformation of the epitope sequence in the context hypocretin-secreting neurons12.
of the native antigenic protein3. This extended confor-
mation results in the side chains of peptide residues Structural comparison of HLA-DQ6.2 and HLA-DQ6.1.
at positions P1, P4, P6 and P9 being directed into the Siebold et al. 13 have recently reported the crystal
MHC class II peptide-binding groove. The residues at structure, at 1.8 Å resolution, of HLA-DQ6.2 in
these positions in the peptide are termed anchor residues complex with a peptide derived from hypocretin
because the interactions of their side chains with distinc- (amino-acid residues 1–13), which has been shown
Anchor tive pockets in the binding groove further stabilize the to bind HLA-DQ6.2 molecules with relatively high
Peptide ligands of most MHC peptide–MHC-class-II complex. The binding grooves of affinity. This molecule shares the same α-chain as,
molecules are anchored into MHC class II isotypes are therefore mainly character- and differs at only nine residues in the β-chain from,
the binding groove by specific ized by the properties of the P1, P4, P6 and P9 pockets, HLA-DQ6.1, which protects from narcolepsy (FIG. 2a).
binding to particular pockets
in the groove. Each MHC
which confer the specificity of the anchor residues. The Three of these residue differences could have effects
molecule has specificity only apparent exceptions to this rule are HLA-DR2a4 on T-cell receptor (TCR) recognition; however, in this
for two or three anchors. and HLA-DR2b5; the structures of these human MHC Review, we focus on the possible disease-related effects

272 | APRIL 2006 | VOLUME 6 www.nature.com/reviews/immunol

© 2006 Nature Publishing Group
 REVIEWS

a b
P8
P2 P5
P3 P7

P9
P1 P6
P4

c
P8
P5 P7
P2 P3

P9
P1 P4 P6
d
P6
P9

P1
P4

Figure 1 | Mining the structural database. a | A superimposition of selected peptide-binding grooves of MHC class II
molecules (green, HLA-DQ2; orange, HLA-DQ6.2; blue, HLA-DQ8; red, H2-IAb; yellow, H2-IAg7). The hypocretin peptide
from the hypocretin–HLA-DQ6.2 complex is depicted as sticks in atomic colouring (blue, nitrogen; red, oxygen; orange,
carbon). Superimpositions were calculated with the programme SHP73. b | A superimposition of representative HLA-DQ
and H2-IA peptides. The orientation is a 90° rotation around the y-axis relative to a.The peptides from the complexes of
hypocretin–HLA-DQ6.2 (orange), insulin B peptide–HLA-DQ8 (blue), α-gliadin–HLA-DQ2 (green), phosphoinositide-
3-kinase–H2-IAb (yellow), GAD–H2-IAg7 (red) are shown as sticks. c | A superimposition of representative HLA-DR peptides,
oriented is as in b. The peptides from the haemagglutanin–HLA-DR1 (yellow), MBP–HLA-DR2a (cyan), MBP–HLA-DR2b
(blue), EBV–HLA-DR2a (green), CLIP–HLA-DR3 (red) and collagen–HLA-DR4.1 (orange) complexes are shown as sticks
with only the carbon β atoms of the side chains depicted (as spheres). d | The solvent-accessible surface of HLA-DQ6.2
is shown, with the hypocretin peptide depicted as in a. The three-dimensional arrangement of each pocket is shown as a
coloured mesh. Electrostatic potentials were calculated using programme PYMOL.The volumes were calculated with the
programme VOLUMES (R. Esnouf, unpublished observations) and represents the size of a pocket available for a specific
side chain at this position. The volume calculation takes the solvent-accessible volume (with a probe of radius 1.4 Å)
around the MHC class II molecule without the peptide, selecting a part of the volume closest to the peptide side chain of
interest. Therefore, the pocket volume is delimited on the bottom and on either side by the solvent-accessible surface and
on the top by the main-chain atoms of the peptide, respectively. Orientation is as in a. EBV, Epstein–Barr virus; CLIP, MHC-
class II-associated invariant-chain peptide; GAD, glutamic-acid decarboxylase; MBP, myelin basic protein.

Deamidation
The modification of glutamine
to glutamic acid, or asparagine
to aspartic acid. The amine
group (NH2) of the side chain is
replaced by a hydroxyl group
(OH) resulting in a negatively
charged amino acid.
of residues in the peptide-binding groove, in this case
five residues. A comparison of the HLA-DQ6.2 crys-
tal structure with a model of HLA-DQ6.1 provides
an insight into potential disease-associated peptide-
binding characteristics13. Two pockets — P4 and P9
— are highlighted by the structural comparison. The
differences at the P9 pocket are mainly electrostatic: the
change at residue 37 in the β-chain (37β) from tyrosine
(that is, Tyr37β) in HLA-DQ6.2 to aspartic acid (that
is, Asp37β) in HLA-DQ6.1, introduces more negative
charge and could be expected to subtly modulate anchor
specificity. However, the most marked difference is in
the P4 pocket. Polymorphisms at residues 13β and 26β
(which change Gly13β and Leu26β in HLA-DQ6.1 to
Ala13β and Tyr26β in HLA-DQ6.2) effectively close
up the P4 pocket in HLA-DQ6.1 (FIG. 2b,c). This closure
places severe steric limitations on the peptide side chains
that can be accommodated in the HLA-DQ6.1 pocket;
for example, the hypocretin peptide is prevented from
binding because of the large size of the threonine resi-
due at position P4. Therefore, a structure-based analysis
indicates that differential peptide binding between two
closely related HLA-DQ6 molecules is central to their
positive and negative association with narcolepsy, an
observation that is consistent with narcolepsy being
an autoimmune disease13.
HLA-DQ and coeliac disease
Coeliac disease is an autoimmune-like disorder of the
small intestine that is caused by an immune response to
antigens (such as α-gliadin) in wheat gluten. It affects
0.5–1% of individuals in many populations and typically
leads to diarrhoea, malabsorption and failure to thrive14.
The vast majority of patients with coeliac disease are
either HLA-DQ2 positive (90%) or HLA-DQ8 posi-
tive (5%)15 (TABLE 1). Immunogenic gluten epitopes that
are recognized by HLA-DQ2- or HLA-DQ8-restricted
T cells have been identified, and these have sequences
in common that are rich in proline and glutamine
residues16,17. It has recently been shown that deamidation
of glutamine residues in these gluten peptides by the
enzyme transglutaminase 2 (the expression of which is
upregulated in inflamed intestines) generates epitopes
that bind efficiently to HLA-DQ2 and HLA-DQ8 and
are recognized by T cells isolated from lesions in the gut
of patients with coeliac disease18,19.

NATURE REVIEWS | IMMUNOLOGY VOLUME 6 | APRIL 2006 | 273

© 2006 Nature Publishing Group
REVIEWS
Structural comparison of HLA-DQ2 and HLA-DQ8.
The link between gluten epitopes and coeliac disease is
exemplified by the α-gliadin peptide (with an amino-
acid sequence PFPQPQLPY), which when selectively
deamidated by transglutaminase 2 and presented by
HLA-DQ2 as the amino-acid sequence PFPQPELPY,
induces potent T-cell responses 19. Kim et al. 20 have
reported the crystal structure, at 2.2 Å resolution, of
HLA-DQ2 in complex with this deamidated α-gliadin
peptide.
The crystal structure of HLA-DQ2 shows a distinc-
tive P6 pocket with a large volume and polar character
defined by the presence of Ser30β rather than Tyr30β,
which is typically found in other HLA-DQ molecules.
a the backbone nitrogen atoms of the proline residues
HLA-DQ6.2
9β
37β
13β 38β
66β
70β
26β
67β
b c Hypocretin peptide
P4-Thr
Hypocretin P4
peptide
Leu26β Tyr26β
Gly13β Ala13β
HLA-DQ6.2 HLA-DQ6.1
Figure 2 | Structural characteristics of HLA-DQ6.2 that are associated with
narcolepsy. a | A comparison of HLA-DQ6.2 and HLA-DQ6.1. The residues that differ
between the two molecules are highlighted as spheres (green, residues contributing to
peptide-binding pockets; blue, residues at the putative T-cell receptor (TCR)-
recognition surface). The position of residue 3β is not shown because it does not
contribute to peptide binding or TCR interaction. The orientation is as in FIG. 1a. b | The
P4 pocket. Selected residues that contribute to the P4 pocket of HLA-DQ6.2. The main
chain and selected residue side-chains are depicted as sticks (yellow, hypocretin
peptide; green, HLA-DQ6.2), c | A superposition with polymorphic residues modelled
for HLA-DQ6.1 coloured in magenta. The two arrows indicate positions at which the
distances between the atoms of the peptide P4 threonine side chain and the side chains
of Ala13β and Tyr26β are too small (that is, the model predicts severe steric clashes).
274 | APRIL 2006 | VOLUME 6
© 2006 Nature Publishing Group
This is a unique feature of HLA-DQ2, as is the presence
of a positively charged lysine residue at 71β; when com-
bined with the polar nature of the P4 and P9 pockets,
these characteristics make this MHC class II peptide-
binding groove the most suitable for accommodating
peptides with negatively charged anchor residues (FIG. 3a).
This is the key factor in allowing HLA-DQ2 to present
gluten-derived peptides that are rich in prolines and
glutamates (generated by deamidation of glutamines).
Both types of residue represent a specific challenge for
presentation by MHC class II molecules. The conserved
peptide conformation (FIG. 1b) in MHC class II molecules
is stabilized by hydrogen bonds between the peptide
backbone and the binding groove. A proline-rich pep-
tide would therefore be expected to bind more weakly
because of the loss of hydrogen-bonding potential for
(proline differs from other amino acids in that its side
chain is covalently bonded to both the nitrogen and car-
bon α atoms of the polypeptide main chain). However,
the peptide-backbone nitrogen atoms at residue posi-
tions P1, P3, P5, P7 and P8 are not involved in the
conserved hydrogen-bond network to the MHC class II
peptide-binding groove. These are therefore positions
at which proline can be incorporated without paying
a penalty for the loss of hydrogen-bonding potential
(FIG. 3b,c). Given the proline- and glutamate-rich nature
of gluten peptides, this constraint on where proline can
be bound makes it probable that glutamate(s) will have
to be accommodated at the anchor positions P4, P6 or
P9. Therefore, of all the MHC class II molecules, the
pockets of HLA-DQ2 (which are able to accommodate
glutamate at three anchor positions) provide the best fit
for α-gliadin (FIG. 3c). The crystal structure of a gluten-
derived peptide bound to HLA-DQ2 perfectly illustrates
this principle, showing an optimally bound glutamate
residue at position P6 and accommodating prolines at
positions P1, P3, P5 and P8 (REF. 20).
Although HLA-DQ2 is the main risk factor for
coeliac disease, HLA-DQ8 also confers susceptibility.
The crystal structure of HLA-DQ8 in complex with an
insulin B peptide4 can be used to exemplify the specific
features of the HLA-DQ8 peptide-binding groove. The
interactions between HLA-DQ8 and this peptide conform
to the canonical pattern for the hydrogen-bonding net-
work between MHC class II molecules and the peptide
backbone. Therefore, the backbone of a peptide bound to
HLA-DQ8 and the gluten peptide presented by HLA-DQ2
are superimposable. As argued above, the key criteria that
determine the successful presentation of gluten-derived
peptides are pockets that can accommodate glutamate
side chains. There are three main pockets in HLA-DQ2
(P4, P6 and P9) that show this capability. The comparison
with HLA-DQ8 shows that the glutamate-binding pro-
pensity of two of these pockets is reduced in this isotype.
Minor differences between HLA-DQ2 and HLA-DQ8
at residues 37β (Ile to Tyr) and 38β (Val to Ala) do not
change the charge characteristics of the P9 pocket. Indeed,
the insulin B peptide in the HLA-DQ8 crystal structure
has a glutamate residue in the P9 pocket4, confirming
that glutamate is also accommodated by the P9 pocket
www.nature.com/reviews/immunol
 REVIEWS

a peptide-binding groove to accommodate glutamate

anchor residues (and therefore gluten-derived pep-
tides) seems to be substantially reduced when com-
P6 P9 pared to the HLA-DQ2 peptide-binding groove. This
P1 is consistent with the weaker association of HLA-DQ8
with coeliac disease.

HLA-DQ and type 1 diabetes

P4 Type 1 diabetes affects 0.06–0.15% of the population
and, in most patients, it is the result of autoimmune-
mediated destruction of insulin-secreting pancreatic
HLA-DQ2 β-cells22. Recent evidence indicates that insulin is an
important source of antigen for MHC-class-II-restricted
b autoreactive T cells23. Type 1 diabetes is associated
Asn62α Asn69α
with alleles belonging to the HLA-DR3 and HLA-DR4
haplotypes, on which the strongest risk factors are the
HLA-DQB1*0302 (HLA-DQ8) and the HLA-DQ2
Tyr9α P6
P1 P9 alleles24,25. By contrast, individuals that carry the HLA-
DQB1*0602 allele are almost completely protected
against type 1 diabetes24,26 (TABLE 1). The HLA-DR3
allele (HLA-DRB1*0301) and certain HLA-DR4 alleles
(for example, HLA-DRB1*0405 and HLA-DRB1*0401)
P4 also predispose individuals to type 1 diabetes, whereas
Asn82β the HLA-DRB1*0403 allele seems to protect against this
disease27. Although the HLA-DR associations are not
c as strong as those of HLA-DQ, it is interesting to note
that insulin-reactive T cells derived from pancreatic
P1 P2 P3 P4 P5 P6 P7 P8 P9
draining lymph nodes of patients with type 1 diabetes
Pro X Pro Glu Pro Glu Pro Pro Glu were HLA-DR4.1 restricted rather than HLA-DQ8 or
Pro Phe Pro Gln Pro Gln Leu Pro Tyr HLA-DQ2 restricted23.
Much of our understanding of type 1 diabetes is
Pro Phe Pro Gln Pro Glu Leu Pro Tyr
based on studies in non-obese diabetic (NOD) mice, an
Figure 3 | Structural characteristics of HLA-DQ2 that are associated with animal model for type 1 diabetes, in which the mouse
coeliac disease. a | The solvent-accessible surface of HLA-DQ2 is coloured by the MHC class II molecule H2-IAg7 is crucial for disease
electrostatic potential (red, acidic; blue, basic) contoured using PYMOL. The α-gliadin development28. Mouse MHC class II molecules that
peptide is depicted in sticks in atomic colouring (blue, nitrogen; red, oxygen; yellow, protect against mouse diabetes have also been identified,
carbon). b | Peptide-binding properties of the α-gliadin–HLA-DQ2 complex. The and these include the H2-IAb molecule29.
peptide is shown and orientated as in a. Hydrogen bonds of the main-chain nitrogen
atoms of the α-gliadin peptide with HLA-DQ2 residues are shown as dashed orange
lines. The side chains of the peptide are represented by spheres marking the positions Structural comparison of molecules conferring risk to
of the carbon β atoms. HLA-DQ2 pockets P1, P4, P6 and P9 are shown as coloured type 1 diabetes: HLA-DQ2, HLA-DQ8 and H2-IAg7.
circles. c | Sequence motifs favoured by the distinctive characteristics of the peptide- Although genetic data show that the HLA-DR alleles
binding groove of HLA-DQ2. The top line shows the theoretical sequence motif, as HLA-DRB1*0401 and HLA-DRB1*0405 predispose to
discussed in the text (where X denotes any amino acid). The next two lines show the type 1 diabetes, whereas HLA-DRB1*0403 protects, so
actual motif of the immunogenic α-gliadin epitope in the natural (second line) and far the crystal structures of the encoded HLA-DR4.3
deaminated (third line) forms. and HLA-DR4.5 molecules have not been reported.
The crystal structure of HLA-DR4.1 is discussed later
of HLA-DQ8. There is only one polymorphism between in this Review in the context of rheumatoid arthritis.
HLA-DQ2 and HLA-DQ8 that affects residues contrib- Crystal structures have been determined for the
uting to the P4 pocket, and this is the substitution of type-1-diabetes-associated MHC class II molecules
Lys71β (in HLA-DQ2) with Thr71β (in HLA-DQ8), HLA-DQ2, HLA-DQ8 and H2-IAg7 (REFS 20,30–32).
but this reduction in positive charge does diminish the Some of the features of the two human molecules
propensity of this pocket to bind glutamate21. There have been discussed above. Two independent crystal
are important differences in the P6 pocket, namely structures have been determined for H2-IAg7: one in
polymorphisms at residues 9β (which is Phe9β in HLA- complex with a peptide derived from hen-egg lyso-
DQ2; Tyr9β in HLA-DQ8) and 30β (which is Ser30β in zyme31 and the other in complex with a peptide from
HLA-DQ2; Tyr30β in HLA-DQ8). The presence of ser- glutamic-acid decarboxylase32. Even before these crys-
ine at 30β at the entrance of the P6 pocket in HLA-DQ2 tal structures became available, it had been postulated
is crucial in allowing sufficient space for the glutamate that particular shared peptide-binding groove char-
side chain to take on an appropriate conformation. By acteristics distinguishing these three molecules from
contrast, tyrosine at this position, as in HLA-DQ8, steri- other MHC class II isotypes might provide insight into
cally hinders this option. The ability of the HLA-DQ8 disease pathways27.

NATURE REVIEWS | IMMUNOLOGY VOLUME 6 | APRIL 2006 | 275

© 2006 Nature Publishing Group
REVIEWS

a similar. The absence of Asp57β, which is a feature com-

mon to HLA-DQ2 and HLA-DQ8, has long been cited
MHC class II allele P4 pocket P6 pocket P9 pocket
to be associated with susceptibility to type 1 diabetes
13β 26β 9β 30β 66α 37β 38β 57β (based on amino-acid sequence comparisons of several
HLA-DQ6.2 Gly Leu Phe Tyr Ala Tyr Ala Asp MHC class II molecules that are either associated or not
HLA-DQ8 Gly Leu Tyr Tyr Leu Tyr Ala Ala associated with type 1 diabetes)24 (FIG. 4a). In most MHC
class II molecules, Asp57β forms a hydrogen bond to the
HLA-DQ2 Gly Leu Phe Ser Leu Ile Asp Ala
peptide backbone and a salt bridge, within the P9 pocket,
HLA-DQ6.1 Ala Tyr Leu Tyr Ala Ile Val Ala to Arg76α2,33 (FIG. 4b). The presence of Ala57β instead
H2-IA b
Gly Tyr Tyr Tyr Val Tyr Val Asp of Asp57β in HLA-DQ2 and HLA-DQ8 has three con-
H2-IAg7 Gly Leu His Tyr Glu Tyr Leu Ser sequences: first, the unpaired Arg76α imparts a strong
basic charge to the P9 pocket; second, the smaller Ala57β
b
side chain increases the accessible volume of the pocket
Ala-P9 Glu-P9
(FIG. 4b); and last, the electrostatically neutral alanine side
HLA-DQ6.2/H2-IAb HLA-DQ8/H2-IAg7
chain cannot form hydrogen bonds to the nitrogen atom
in the peptide backbone at position P10. The increased
volume of the pocket allows an acidic glutamate side
chain at the P9 position of the peptide to compensate for
the lost electrostatic interactions. Therefore, in the struc-
Asp57β Ala57β ture of HLA-DQ8 in complex with the insulin B peptide,
Tyr37β Tyr37β the P9 glutamate side chain forms hydrogen bonds to
Arg76α (FIG. 4b). Although in H2-IAg7, 57β is a slightly
Arg76α Arg76α larger serine residue, the P9 pocket still accommodates
a similar binding mode for glutamate31,32, consistent with
the reported preference for negatively charged P9 anchor
c
residues21. As noted by several authors, this distinctive
HLA-DQ6.2 HLA-DQ8 HLA-DQ2 H2-IAb H2-IAg7
feature of the P9 pocket in HLA-DQ2, HLA-DQ8 and
H2-IAg7 molecules might indicate similar pathophysi-
ological pathways for developing type 1 diabetes in
humans and NOD mice30,27.
So, given that there seem to be common peptide-
binding-groove characteristics that are associated
with susceptibility to type 1 diabetes between species,
107 Å3 72 Å3 176 Å3 81 Å3 69 Å3
are there also common trends in the peptide-binding
Figure 4 | Binding pockets that are implicated in type 1 diabetes. a | Disease- characteristics of MHC class II molecules that confer
associated residues are shown by the sequence alignment of selected HLA-DQ and protection to this disease?
H2-IA chains. b | Close-up view of selected residues that contribute to the P9 pocket of
MHC class II molecules that are negatively (HLA-DQ6.2, H2-IAb; left) and positively (HLA-
Structural comparison of molecules conferring risk
DQ8, H2-IAg7; right) associated with type 1 diabetes. HLA-DQ6.2 and HLA-DQ8 residues
are depicted as sticks in atomic colouring (blue, nitrogen; red, oxygen; cyan, HLA-DQ6.2 to or protection against type 1 diabetes. As discussed
carbon; yellow, HLA-DQ8 carbon; magenta, hypocretin–HLA-DQ6.2 and insulin-B- earlier, it has long been noted that for MHC alleles that
peptide–HLA-DQ8 carbon). A tightly bound water molecule is shown as a red sphere. are positively associated with type 1 diabetes, residue
The corresponding residues of H2-IAb are superimposed with HLA-DQ6.2, and H2-IAg7 57β of the P9 pocket is Ala57β (HLA-DQ2 and HLA-
residues are superimposed with HLA-DQ8; both are depicted in grey. Hydrogen bonds DQ8) or Ser57β (H2-IAg7 ), whereas for negatively
are represented as dashed orange lines. c | P6 pocket volumes of selected HLA-DQ and associated alleles (and in fact most other alleles) it is
H2-IA molecules. The volume of each pocket is shown as a coloured mesh. Asp57β24. There is, however, one example of an MHC
class II molecule that breaks this rule. This molecule
has the same α-chain as HLA-DQ8 but has a differ-
The P1 pocket is relatively variable and it is difficult ent β-chain, encoded by the HLA-DQB1*0401 allele,
to detect systematic trends in its properties for the HLA- which has aspartic acid at 57β and has been reported
DQ and H2-IA molecules, however, Suri et al. have to be positively associated with type 1 diabetes34,35.
noted that HLA-DQ8 and H2-IAg7 share a preference Lee et al. provide a plausible structural explanation for
for negatively charged anchor residues at position P1 this apparent contradiction, arguing that the presence
(REF. 21). Although there are some differences in detail, of Leu56β in the HLA-DQB1*0401-encoded β-chain
the overall polar nature of the P4 pocket is conserved indirectly causes conformational changes that cancel
between HLA-DQ2 and HLA-DQ8 (as discussed ear- out the effect of the negatively charged Asp57β side
lier). This conservation extends to H2-IAg7, but also to chain on the P9 pocket30. A comparison of the struc-
the protective HLA-DQ6.2 molecule, which indicates tures of the disease-associated versus protective MHC
that the P4 pocket does not have a role in susceptibility class II molecules hints at a second characteristic; the
to type 1 diabetes. In the previous section, we pointed P6 pocket shows a trend in volume size that seems to
out that there are differences between HLA-DQ8 and correlate from positive to negative disease association.
HLA-DQ2 at the P6 pocket but that their P9 pockets are Typically, P6 pockets have lower volumes in positively

276 | APRIL 2006 | VOLUME 6 www.nature.com/reviews/immunol

© 2006 Nature Publishing Group

associated molecules than in those molecules that are
negatively associated13 (FIG. 4c). For example, crystal
structures of the protective isotypes H2-IAb and HLA-
DQ6.2 (REFS 13,36) reveal deep P6 pockets with volumes
in excess of 80 Å3. The shape and charge characteristics
of these pockets indicate that they might bind a broad
range of residues, which is consistent with the reported
anchor-residue usage36,37. Conversely, the P6 pockets of
the type-1-diabetes-associated HLA-DQ8 and H2-IAg7
molecules are constricted by the presence of large side
chains at residues 66α (Leu66α in HLA-DQ8 and
HLA-DQ2; Ala66α in HLA-DQ6.2; Glu66α in H2-IAg7;
Val66α in H2-IAb) and 9β (Tyr9β in HLA-DQ8; Phe9β
in HLA-DQ2 and HLA-DQ6.2; His9β in H2-IAg7).
Perhaps the most striking example of this effect is the
formation of a strong electrostatic interaction between
Glu66α and His9β in H2-IAg7 (REFS 31,32). The result of
these polymorphisms in both HLA-DQ8 and H2-IAg7
is a P6 pocket with a volume of less than 70 Å3, a space
that is insufficient to accommodate anchor residues
with anything other than the smallest of side chains.
Superficially, HLA-DQ2, which has a calculated volume
of 176 Å3 for pocket P6, seems to be an exception to
this trend, but this apparently large volume results from
a very shallow pocket shape, which is unsuitable for
any large side chains other than the most conforma-
tionally flexible. Furthermore, when combined with
electrostatic potential, this shape imposes a restricted
anchor preference for glutamate38 (as discussed earlier
for the role of HLA-DQ2 in coeliac disease) or resi-
dues with relatively small side chains38,39. Although the
relevance of this trend in pocket volume remains to be
proven, it provides an encouraging indicator of how
subtle structural correlations with disease-associated
MHC class II molecules could become discernible as
structural databases grow.
HLA-DR and rheumatoid arthritis
Rheumatoid arthritis is a chronic inflammatory disease
that mainly affects the synovial joints40. One per cent of
individuals in Western populations suffer from rheuma-
toid arthritis and more than 90% of patients with severe
disease carry the HLA-DRB1*0401, HLA-DRB1*0404
and/or HLA-DRB1*0101 allele(s) 41,42 that encode
HLA-DR4.1, HLA-DR4.4 and HLA-DR1, respectively.
These proteins share a common stretch of amino acids
in their peptide-binding grooves at positions 67–74 of
the HLA-DR β-chain: the so-called shared epitope43.
The HLA-DRB1*0402-encoded molecule (HLA-DR4.2),
which is not associated with rheumatoid arthritis, differs
from this motif at positions 70 and 71, and has a different
peptide-binding specificity44. Based on the MHC associa-
tion, the presence of T cells in the synovial compartments
of patients with rheumatoid arthritis45, studies in human-
ized mouse models for rheumatoid arthritis expressing
HLA-DR4.1 or HLA-DR1 molecules46,47, and the clinical
benefits from treating active rheumatoid arthritis by
blocking T-cell co-stimulation and activation48, T cells are
thought to have an important role in the development of
rheumatoid arthritis. Although no definite autoantigen
has been identified, several candidates exist. One of these
NATURE REVIEWS | IMMUNOLOGY
© 2006 Nature Publishing Group
REVIEWS
is type II collagen, which has a single immunodominant
T-cell epitope (type II collagen residues 261–273) that is
bound by both HLA-DR1 and HLA-DR4.1 molecules47,49.
T cells that recognize this epitope are central to the devel-
opment of collagen-induced arthritis in humanized mice
that express HLA-DR1 and HLA-DR4.1 (REFS 50,51).
Structural comparison of HLA-DR1 and HLA-DR4.1.
The crystal structure of HLA-DR4.1 in complex with
a peptide derived from type II collagen (residues
1168–1180: QYMRADQAAGGLR), was determined by
Dessen et al.52 This peptide was selected for study on the
basis of its predicted ability to bind HLA-DR4.1 rather
than any known immunogenic activity, but served as a
good basis for a model of HLA-DR4.1 in complex with
the immunodominant type II collagen epitope (residues
261–273: AGFKGEQGPKGEP). The crystal structure
shows the HLA-DR4.1 peptide-binding groove to
be made up of electrostatically neutral P1, P6 and P9
pockets and a P4 pocket that can accommodate acidic
residues (for example, an aspartic acid in the type II
collagen epitope 1168–1180) in part through hydrogen
bonding with Lys71β. Lys71β lies within the shared
epitope, the presence of which predisposes individuals
to rheumatoid arthritis43. In total, four residues from
this portion of the β-chain contribute to the P4 pocket
(Leu67β, Gln70β, Lys71β and Ala74β). Prior to the crys-
tal structure determination, studies on peptide-binding
specificity highlighted the ability of the P4 pocket to
bind acidic residues as the key HLA-DR4.1 character-
istic associated with rheumatoid arthritis44. Analyses of
the HLA-DR4.1 structure reinforced this hypothesis52.
Such observations are also consistent with a recent
report53 indicating that the conversion of arginine to
citrulline in model peptides facilitates their binding to
HLA-DR4.1, as this modification changes the positively
charged arginine side chain to a polar form that can be
accommodated in the P4 pocket.
Although its importance is not as well established
as that of the P4 pocket, a second property of the HLA-
DR4.1 peptide-binding groove associated with the P6
and P9 pockets might also be noteworthy. Epitopes
that are derived from the triple-helical region of
type II collagen are rich in glycine (the sequence con-
tains a glycine as every third residue), for example the
type II collagen epitope 261–273, which is known to
be immunodominant in HLA-DRB1*0401-transgenic
mice46. An assessment of the pocket characteristics of
HLA-DR4.1 (FIG. 5a) indicates that these MHC class II
molecules might be particularly well suited to binding
such glycine-rich peptides because the P6 and P9 pock-
ets are shallow (FIG. 5a,b). This conclusion is supported
by the reported anchor preferences of the P6 and P9
pockets of HLA-DR4 molecules44,54, in particular for
the P9 pocket, which strongly favours glycine, alanine,
serine or cysteine (all of which are residues with small
side chains). The relatively weak contribution to bind-
ing by the shallow P6 and P9 pockets is compensated
by the deep, hydrophobic P1 pocket, which favours
anchor residues with aromatic side chains (such as
phenylalanine, tyrosine or tryptophan).
VOLUME 6 | APRIL 2006 | 277
REVIEWS

a HLA-DR4.1 c HLA-DR1
P6 (48 Å3)
P6 (64 Å3)

48

P9 (75 Å3)
P9 (92 Å3)

P1 (238 Å3) P1 (186 Å3)

P4 (87 Å3)
P4 (124 Å3)
b d
Ser53α Asn62α Asn69α Ser53α Tyr9α Asn62α Asn69α

Tyr9α
P8 P8
P3 P5 P3 P5

P6 P9
P7 P6
P1 P2 P9
P4 P1 P2 P7
P4
Asn82β Asn82β
Lys71β Arg70β

e
P1 P2 P3 P4 P5 P6 P7 P8 P9
Peptide residues favoured by HLA-DR4.1 Hydrophobic X Pro Polar Pro Gly (Pro) Pro Gly
Type II collagen peptide residues 263–271 Phe Lys Gly Glu Gln Gly Pro Lys Gly

Figure 5 | HLA-DR1 and HLA-DR4.1: structural characteristics associated with rheumatoid arthritis. a | Pocket
volumes of HLA-DR4.1 mapped onto the type II collagen peptide residues 1168–1180. The three-dimensional extent of
each pocket is depicted as a coloured mesh, and the pocket volumes are shown in parentheses. The peptide is shown as
sticks in atomic colouring (blue, nitrogen; red, oxygen; yellow, carbon) and orientated as in FIG. 1a. The volume of the P1
pocket differs slightly between HLA-DR1 and HLA-DR4.1 because of a difference in the main-chain conformation of
Gly86β, but this does not affect the pocket specificity. b | Hydrogen bonds between the main-chain nitrogen atoms of the
type II collagen peptide and HLA-DR4 residues are shown as dashed orange lines. Lys71β of HLA-DR4.1 (highlighted in a
red box) forms a hydrogen bond with the P4 aspartate side chain of the peptide. c | Pocket volumes of HLA-DR1 mapped
onto the influenza virus haemagglutinin peptide (residues 306–318). The three-dimensional extent of each pocket is
shown as a coloured mesh, and the pocket volumes are indicated in parentheses. The peptide is shown as sticks in atomic
colouring (blue, nitrogen; red, oxygen; cyan, carbon) and orientated as in FIG. 1a. d | Hydrogen bonds of the main-chain
nitrogen atoms of the haemagglutinin peptide with HLA-DR1 residues. Colour coding is as in c. The hydrogen bond
between the P4 glutamine side chain of the peptide and Arg70β of HLA-DR1 is highlighted in a red box. e | Sequence
motifs that are favoured by the distinctive characteristics of the peptide-binding groove of HLA-DR4.1. The top line of the
table shows the theoretical sequence motif discussed in the text (where X denotes any amino acid). Line two is the actual
motif of the immunodominant type II collagen peptide residues 261–273 (residues 263–271 are shown).

The crystal structure of HLA-DR1 in complex
with a peptide that is derived from the influenza-
virus protein haemagglutinin (residues 306–318;
PKYVKQNTLKLAT) 2 shows a peptide-binding
groove that has many distinctive features in common
with HLA-DR4.1. Indeed, there are several reported
examples of peptides that bind both HLA-DR4.1 and
HLA-DR1, including haemagglutinin 306–318, for
which a crystal structure in complex with HLA-DR4.1
shows that the epitope is bound in the same confor-
mation as it is in HLA-DR1 (that is, using the same
anchor residues)55. This is possible because the P1,
P4, P6 and P9 pockets of HLA-DR1 and HLA-DR4.1
have similar characteristics (compare FIG. 5a,b with
FIG. 5c,d). The sequence between β-chain residues 67
and 74 (the rheumatoid-arthritis-associated shared
epitope) in HLA-DR1 has arginine at 70β and 71β in
place of the glutamine and lysine residues that are seen
at these positions in HLA-DR4.1. Therefore, for both
molecules, the P4 pocket is able to bind acidic residues
(FIG. 5b,d). The shallow nature of the P6 and P9 pockets is
also retained between HLA-DR1 and HLA-DR4.1. Again,
this is reflected in the reported anchor preferences, with
glycine residues strongly favoured at both P6 and P9 pep-
tide positions for HLA-DR1 (REF. 56). Therefore, although
crystal structures for neither HLA-DR1 nor HLA-DR4.1
have been determined in complex with peptides con-
taining Gly-X-X repeats (where X denotes any amino
acid), the available structures do indicate that both of
these MHC class II molecules are well suited to binding
peptides that are derived from the glycine-rich collagen
proteins (FIG. 5e). Indeed, HLA-DR1 presents the type II
collagen epitope 261–273 to pathogenic T cells in a HLA-
DR1-expressing transgenic mouse model of rheumatoid
arthritis47. The Gly-X-X repeats in collagen proteins
often contain proline at one or both of the two X posi-
tions and, as discussed earlier for HLA-DQ2, the main-
chain properties of this residue can be accommodated at

278 | APRIL 2006 | VOLUME 6 www.nature.com/reviews/immunol

© 2006 Nature Publishing Group
 REVIEWS

a variety of non-anchor positions (P3, P5, P7 and P8; a

FIG. 5e). The distinctive ability of the HLA-DR1 and P6
HLA-DR4.1 peptide-binding grooves to bind peptides P1 P9
from collagen proteins is analogous to the ability of the
HLA-DQ2 peptide-binding groove to bind peptides
from α-gliadin. Particular MHC class II molecules have β
Trp61β
peptide-binding grooves that can accommodate atypical P4
MBP–HLA-DR2a
protein sequences: α-gliadin for HLA-DQ2, and type II
collagen for HLA-DR1 and HLA-DR4.1.

Structural comparison of HLA-DR4.1 and HLA-DR4.2. b

Although HLA-DR4.1 is associated with rheumatoid P6
P1 P9
arthritis, the closely related HLA-DR4.2 molecule is
not43 (TABLE 1). The important differences between the
sequences of these two MHC class II molecules, which
affect the properties of the peptide-binding groove, P4
Trp61β
are Asp70β instead of Gln70β and Glu71β in place of MBP–HLA-DR2b
Arg71β. Aspartic acid and glutamic acid are both acidic
residues and have been shown to swap the anchor prefer-
ence of the P4 pocket from one that accommodates nega-
tively charged anchor residues to one that has a strong c
preference for positively charged peptide side chains44,57. TCRβ
HLA-DR2b H2-IAu
The change of Arg71β to Glu71β also results in the loss
of a hydrogen bond to the peptide backbone carbonyl
group at position P5; this hydrogen bond is conserved
in all other HLA-DR structures (with the exception of
HLA-DR2b, which is discussed next). The combined TCRα
result of these changes is to markedly alter the peptide
anchor preferences of HLA-DR4.2; this is most obvious
at the P4 pocket, with the effect at other pockets being
indirect and more subtle. Despite HLA-DR4.1 and HLA-
DR4.2 having identical P6 and P9 pockets, there is no d EBV627–641
TCR
longer any preference for glycine at P6 (REF. 44), and this TCR MBP85–99
TCR TCR
is consistent with the need for tighter binding in the Lys
carboxyl-terminal half of the peptide to compensate for Val His Ile Val Pro
His Phe
the loss of the hydrogen bond to the peptide backbone
at position P5. The overall effect is a general reduction Pro Gln
in the propensity of this molecule to bind glycine-rich Gly Val
Lys Thr His
collagen peptides as well as the specific loss of binding Val
Val Phe Asn
Tyr
for the type II collagen epitope 261–273.
P9
P1 P4 P6
HLA-DR and multiple sclerosis
Multiple sclerosis is a chronic inflammatory and Figure 6 | Structural characteristics of HLA-DR2
degenerative disease of the central nervous system that molecules that are associated with multiple sclerosis.
affects 0.1% of individuals in Western populations58. a,b | The solvent-accessible surfaces of HLA-DR2a (a) and
HLA-DR2b (b) are coloured by the electrostatic potential
In high-risk (northern European) populations, mul-
(red, acidic; blue, basic). The myelin basic protein (MBP)
tiple sclerosis is associated with three different MHC peptides are depicted in sticks in atomic colouring (blue:
class II genes: HLA-DRB5*0101, HLA-DRB1*1501 nitrogen; red, oxygen; yellow, MBP–HLA-DR2a carbon;
and HLA-DQB1*0602 (REF. 59) (TABLE 1). Because these cyan, MBP–HLA–DR2b carbon). The position of Trp61β is
three alleles are in almost complete linkage disequilib- indicated. c | A schematic presentation of the HLA-DR2b–
rium, it has been difficult to determine which of them TCR (T-cell receptor) (left) and the H2-IAu–TCR (right)
confers the main susceptibility to multiple sclerosis. complexes. HLA-DR2b (red) and H2-IAu (green) are
However, a recent study in African Americans, who do depicted as ribbons. The MBP85–99–HLA-DR2b (magenta)
not have this strong linkage disequilibrium, indicates and MBPAc1–11–H2-IAu (cyan) peptides are shown as
that HLA-DRB1*1501, rather than HLA-DQB1*0602, beads. The interaction footprints of the corresponding
TCRs are drawn schematically and coloured according
constitutes the principal multiple sclerosis disease-risk
to their peptide chains (yellow, α-chain; orange, β-chain).
gene60. Because this study did not address the role of d | A superimposition of the Epstein–Barr virus (EBV)
the HLA-DRB5*0101 allele, it cannot be ruled out as an (orange) and MBP (blue) peptides based on MHC class II
additional susceptibility factor. HLA-DRB5*0101 and structures. The solvent-exposed residues that are crucial
HLA-DRB1*1501 encode the β-chains of HLA-DR2a for TCR recognition and the anchor residues that bind in
and HLA-DR2b, respectively. pockets P1, P4, P6 and P9 are highlighted with arrows.

NATURE REVIEWS | IMMUNOLOGY VOLUME 6 | APRIL 2006 | 279

© 2006 Nature Publishing Group
REVIEWS
Alanine-scanning analyses
Each residue of the peptide
is separately substituted by
alanine and the effect on T-cell
receptor stimulation and
binding by MHC class II
molecules is assayed.
280 | APRIL 2006 | VOLUME 6
There is strong evidence that autoimmunity to
myelin antigens has an important role in the develop-
ment of multiple sclerosis61. Several myelin-derived
autoantigenic targets have been described, and these
include myelin basic protein (MBP), proteolipid pro-
tein and myelin oligodendrocyte glycoprotein. There
has been a particular focus on MBP for at least two rea-
sons: MBP-specific TCR-expressing CD4+ T cells have
been found in brain tissue from patients with multiple
sclerosis62, and cells that present an immunodominant
MBP peptide (residues 85–99, denoted MBP85–99)
in complex with HLA-DR2b have been shown to be
present in multiple sclerosis lesions63. Also, human-
ized mice that express the HLA-DRB1*1501 gene and a
human TCR that recognizes the MBP85–99 peptide in
the context of HLA-DR2b either spontaneously or after
immunization with MBP85–99, develop experimental
autoimmune encephalomyelitis64, which has several
features in common with multiple sclerosis.
Structural comparison of HLA-DR2a and HLA-DR2b.
Given the emerging evidence indicating a secondary
role for HLA-DQ6.2 (which was discussed earlier, in the
context of narcolepsy) in multiple sclerosis, we focus here
on HLA-DR2a and HLA-DR2b. Crystal structures have
been determined for HLA-DR2a and HLA-DR2b4,5,13, and
these molecules show some similar characteristics in their
peptide-binding grooves. In the grooves of both mol-
ecules, the binding capacity of the pockets is large, which
favours the accommodation of large hydrophobic resi-
dues at positions P1 and P4. The P1 pocket in HLA-DR2a
is larger than in HLA-DR2b because a glycine replaces a
valine at residue 86β, resulting in a preference for larger
aromatic anchor residues in HLA-DR2a and aliphatic
residues in HLA-DR2b. The molecules differ at pockets
P6 and P9 because the peptide-binding groove of HLA-
DR2a contains a distinctive ridge formed by Trp61β4
(FIG. 6a,b). This residue is non-polymorphic but adopts
a radically different side-chain conformation in HLA-
DR2a compared with HLA-DR2b because of two poly-
morphic residues, 38β and 67β. In HLA-DR2a, Leu38β
causes a steric clash with the side chain of Trp61β in its
standard conformation; this large aromatic side chain is
therefore repositioned by a rotation of ~180° and stabi-
lized in the new environment by stacking against Phe67β.
The combination of Leu38β and Phe67β is almost unique
to HLA-DR2a (it also occurs in two relatively rare HLA-
DR12 alleles), and indirectly, through their interactions
with Trp61β, these polymorphisms impose a peptide-
backbone conformation that, from position P5, elevates
the bound peptide above the HLA-DR2a peptide-binding
groove (FIG. 1c), precluding optimal binding in the P6
and P9 pockets. This distinctive peptide conformation is
apparent in crystal structures of HLA-DR2a in complex
with MBP85–99 (REF. 4) and with a peptide of generally
unrelated sequence that is derived from Epstein–Barr
virus (EBV) (TGGVYHFVKKHVHES)65. Interestingly,
the structure of MBP85–99 in complex with HLA-DR2b
shows this same non-standard backbone conformation
(FIG. 1c), despite the anchor usage being shifted along
the peptide sequence by three residues5. In contrast to
© 2006 Nature Publishing Group
HLA-DR2a, this altered peptide conformation is not
imposed by the architecture of the HLA-DR2b peptide-
binding groove. Smith et al.5 propose that the MBP85–99
peptide adopts this conformation because the P6 and P9
residues are not optimal anchors: the asparagine at P6
is too large to insert fully into the P6 pocket5 and the
binding affinity of the MBP85–99 peptide is increased on
substitution of the P9 threonine by alanine66. It is unclear
whether this conformation is typical of other peptides
presented by HLA-DR2b, because no other structures are
available for peptide–HLA-DR2b complexes. It might,
however, be noteworthy that in HLA-DR2b, Ala71β
is unable to form the hydrogen bond to the peptide-
backbone carbonyl group at position P5 that is found
in most other MHC class II molecules (with the excep-
tion of HLA-DR4.2, as discussed earlier). Instead, this
polymorphism increases the volume of the P4 pocket.
Therefore, HLA-DR2b has particularly large hydrophobic
P1 and P4 pockets (FIG. 6b) that allow it to make strong
interactions to the N-terminal half of peptides with
suitable anchor residues and therefore to decrease the
relative importance of specific binding into the P6 and
P9 pockets.
The distinctive binding characteristics of HLA-DR2a
and HLA-DR2b each results in peptide residues from P6
to P9 having a raised position above the peptide-binding
grooves. This differs considerably from the canonical
mode of peptide presentation by other MHC class II mol-
ecules (FIG. 1b,c). Might this influence the possible binding
modes for TCR recognition of HLA-DR2a and HLA-
DR2b? Indeed, alanine-scanning analyses to identify the
residues in MBP85–99 that interact with TCRs indicate
that the N-terminal half of the peptide has the dominant
role in TCR binding to both HLA-DR2a and HLA-DR2b
complexes66,67. These observations are consistent with
the structures of two autoreactive (disease-associated)
TCRs: one in complex with MBP85–99–HLA-DR2b68
and the other with MBP85–99–HLA-DR2a69. For both
of these MHC-class-II–TCR complexes, the interaction
footprint of the TCR is markedly skewed towards the
N-terminal half of the peptide compared to that seen in
other complexes. If the nature of peptide presentation by
HLA-DR2a and HLA-DR2b favours TCRs that sample a
reduced portion of the peptide, the potential for cross-
reactivity with a series of peptides that, at least over a
limited portion of their surface, are structural mimics of
MBP85–99 will inevitably be increased (FIG. 6c). A com-
parison of the structures of EBV-peptide–HLA-DR2a and
MBP85–99–HLA-DR2b complexes has shown that two
peptide epitopes with relatively low sequence identity can
still appear structurally identical to a TCR if it only sam-
ples the peptide backbone conformation and the exposed
side chains at position P1, P2, P3 and P5 (REF. 65) (FIG. 6d).
Based on their resolution of the structure of a complex
between an autoreactive TCR and the mouse MHC class II
molecule H2-IAu presenting the MBPAc1–11 peptide,
Maynard et al.70 made a related point: to be crossreactive,
a TCR does not need to be ‘degenerate’ in its binding (that
is, binding many different peptide structures), but rather,
crossreactivity can result from the TCR sampling only a
few points on the peptide.
www.nature.com/reviews/immunol
 REVIEWS

Conclusions increasing power of this combined approach to inform

What general trends are revealed by surveying the
peptide-binding-groove characteristics of the MHC
class II molecules that are associated with autoimmune
disease? For each of the disease-associated MHC class II
molecules discussed in this Review, particular features
in the peptide-binding groove distinctively shape the
population of peptides that the molecule presents. In
at least two cases, the nature of the binding groove par-
ticularly favours peptides that are derived from proteins
with distinctive sequence patterns that are located at the
site of inflammation: α-gliadin in HLA-DQ2-associated
coeliac disease and collagen in HLA-DR1- and HLA-
DR4-associated rheumatoid arthritis. In HLA-DR2a and
HLA-DR2b, the actual mode of presentation of the pep-
tide seems also to be implicated in the disease pathogen-
esis, as it might actively select for a TCR population that
is intrinsically susceptible to crossreactivity. Already, the
combination of genetic, epidemiological and structural
analyses has highlighted previously unsuspected princi-
ples. As the databases that have been created by each of
the disciplines become even more richly populated, the
autoimmune research seems set to continue. Ultimately,
this should facilitate a detailed understanding of disease
pathways and the development of new and more effec-
tive drugs with fewer side effects. Such developments
could be facilitated in one of two ways. One possible
approach is that drug targets that are indirectly associ-
ated with the pathogenic MHC-class-II–TCR recogni-
tion events will be identified as the disease pathway is
elucidated. The second approach is to either directly
block the aberrant MHC-class-II–TCR recognition,
or to induce anti-inflammatory T-cell responses; both
these strategies have already been shown to have some
effect in the treatment of multiple sclerosis and its
mouse-model equivalent, experimental autoimmune
encephalomyelitis71,72. However, much further analysis
will be required if we are to generate the basic insights
necessary for these developments. A larger selection of
MHC-class-II–TCR complex structures must be deter-
mined and the functional role of molecular properties
revealed by these structures must be investigated in
mechanistic studies in vitro and in vivo.

1. Brown, J. H. et al. Three-dimensional structure of the
human class II histocompatibility antigen HLA-DR1.
Nature 364, 33–39 (1993).
The classic analysis of an MHC class II molecule
in complex with a single peptide. This structure
detailed the fundamental characteristics of the
peptide-binding groove.
2. Stern, L. J. et al. Crystal structure of the human class II
MHC protein HLA-DR1 complexed with an influenza
virus peptide. Nature 368, 215–221 (1994).
3. Schueler-Furman, O., Altuvia, Y. & Margalit, H.
Examination of possible structural constraints of
MHC-binding peptides by assessment of their native
structure within their source proteins. Proteins 45,
47–54 (2001).
4. Li, Y., Li, H., Martin, R. & Mariuzza, R. A. Structural
basis for the binding of an immunodominant peptide
from myelin basic protein in different registers by two
HLA-DR2 proteins. J. Mol. Biol. 304, 177–188
(2000).
This paper reports the crystal structure of
HLA-DR2a in complex with an immunodominant
epitope from MBP. Together with reference 5, the
structure illustrates how the same peptide can be
bound in different conformations by different MHC
class II molecules.
5. Smith, K. J., Pyrdol, J., Gauthier, L., Wiley, D. C. &
Wucherpfennig, K. W. Crystal structure of HLA-DR2
(DRA*0101, DRB1*1501) complexed with a peptide
from human myelin basic protein. J. Exp. Med.
188, 1511–1520 (1998).
This paper reports the crystal structure of
HLA-DR2b in complex with the multiple-sclerosis-
associated immunodominant epitope from MBP
described in reference 4. When presented in the
context of HLA-DR2b, this epitope is known to elicit
disease-causing T-cell responses.
6. Okun, M. L., Lin, L., Pelin, Z., Hong, S. & Mignot, E.
Clinical aspects of narcolepsy-cataplexy across ethnic
groups. Sleep 25, 27–35 (2002).
7. Mignot, E. Genetic and familial aspects of narcolepsy.
Neurology 50, S16–S22 (1998).
8. Matsuki, K. et al. DQ (rather than DR) gene marks
susceptibility to narcolepsy. Lancet 339, 1052 (1992).
9. Mignot, E. et al. Complex HLA-DR and -DQ interactions
confer risk of narcolepsy-cataplexy in three ethnic
groups. Am. J. Hum. Genet. 68, 686–699 (2001).
10. Taheri, S., Zeitzer, J. M. & Mignot, E. The role
of hypocretins (orexins) in sleep regulation and
narcolepsy. Annu. Rev. Neurosci. 25, 283–313
(2002).
11. Nishino, S., Ripley, B., Overeem, S., Lammers, G. J. &
Mignot, E. Hypocretin (orexin) deficiency in human
narcolepsy. Lancet 355, 39–40 (2000).
12. Peyron, C. et al. A mutation in a case of early onset
narcolepsy and a generalized absence of hypocretin
peptides in human narcoleptic brains. Nature Med.
6, 991–997 (2000).
13. Siebold, C. et al. Crystal structure of HLA-DQ0602
that protects against type 1 diabetes and confers
strong susceptibility to narcolepsy. Proc. Natl Acad.
Sci. USA 101, 1999–2004 (2004).
This structural analysis of HLA-DQ6.2 reveals
features of the peptide-binding groove that
correlate with susceptibility to narcolepsy and
protection against type 1 diabetes.
14. Farrell, R. J. & Kelly, C. P. Celiac sprue. N. Engl. J.
Med. 346, 180–188 (2002).
15. Sollid, L. M. et al. Evidence for a primary
association of celiac disease to a particular
HLA-DQ α/β heterodimer. J. Exp. Med. 169,
345–350 (1989).
16. Lundin, K. E., Scott, H., Fausa, O., Thorsby, E. &
Sollid, L. M. T cells from the small intestinal mucosa
of a DR4, DQ7/DR4, DQ8 celiac disease patient
preferentially recognize gliadin when presented by
DQ8. Hum. Immunol. 41, 285–291 (1994).
17. Lundin, K. E. et al. Gliadin-specific, HLA-DQ(α1*0501,
β1*0201) restricted T cells isolated from the small
intestinal mucosa of celiac disease patients.
J. Exp. Med. 178, 187–196 (1993).
18. van de Wal, Y. et al. Selective deamidation by tissue
transglutaminase strongly enhances gliadin-specific
T cell reactivity. J. Immunol. 161, 1585–1588
(1998).
19. Molberg, O. et al. Tissue transglutaminase selectively
modifies gliadin peptides that are recognized by
gut-derived T cells in celiac disease. Nature Med.
4, 713–717 (1998).
20. Kim, C. Y., Quarsten, H., Bergseng, E.,
Khosla, C. & Sollid, L. M. Structural basis for
HLA-DQ2-mediated presentation of gluten epitopes
in celiac disease. Proc. Natl Acad. Sci. USA 101,
4175–4179 (2004).
The structure of HLA-DQ2 in complex with a
gluten-derived peptide shows how MHC class II
peptide-binding-groove characteristics allow the
binding of peptides from atypical protein
sequences, in this case gliadin, leading to
susceptibility to coeliac disease.
21. Suri, A., Walters, J. J., Gross, M. L. & Unanue, E. R.
Natural peptides selected by diabetogenic DQ8
and murine I-Ag7 molecules show common
sequence specificity. J. Clin. Invest. 115, 2268–2276
(2005).
22. Atkinson, M. A. & Eisenbarth, G. S. Type 1 diabetes:
new perspectives on disease pathogenesis and
treatment. Lancet 358, 221–229 (2001).
23. Kent, S. C. et al. Expanded T cells from pancreatic
lymph nodes of type 1 diabetic subjects recognize
an insulin epitope. Nature 435, 224–228 (2005).
24. Todd, J. A., Bell, J. I. & McDevitt, H. O. HLA-DQ
β gene contributes to susceptibility and resistance
to insulin-dependent diabetes mellitus. Nature
329, 599–604 (1987).
The classic analysis demonstrating that MHC
class II polymorphisms can be linked to an
autoimmune disease.
25. Todd, J. A. & Wicker, L. S. Genetic protection from
the inflammatory disease type 1 diabetes in humans
and animal models. Immunity 15, 387–395 (2001).
26. Baisch, J. M. et al. Analysis of HLA-DQ genotypes and
susceptibility in insulin-dependent diabetes mellitus.
N. Engl. J. Med. 322, 1836–1841 (1990).
27. Cucca, F. et al. A correlation between the relative
predisposition of MHC class II alleles to type 1
diabetes and the structure of their proteins. Hum.
Mol. Genet. 10, 2025–2037 (2001).
28. Tisch, R. & McDevitt, H. Insulin-dependent diabetes
mellitus. Cell 85, 291–297 (1996).
29. Robinson, C. P. et al. A novel NOD-derived murine
model of primary Sjogren’s syndrome. Arthritis
Rheum. 41, 150–156 (1998).
30. Lee, K. H., Wucherpfennig, K. W. & Wiley, D. C.
Structure of a human insulin peptide-HLA-DQ8
complex and susceptibility to type 1 diabetes.
Nature Immunol. 2, 501–507 (2001).
This paper provided the first structural analysis
of an HLA-DQ molecule, HLA-DQ8. The structure
was of HLA-DQ8 in complex with a peptide derived
from human insulin (a candidate autoantigen in
type 1 diabetes).
31. Latek, R. R. et al. Structural basis of peptide binding
and presentation by the type I diabetes-associated
MHC class II molecule of NOD mice. Immunity
12, 699–710 (2000).
32. Corper, A. L. et al. A structural framework for
deciphering the link between I-Ag7 and autoimmune
diabetes. Science 288, 505–511 (2000).
33. Fremont, D. H., Hendrickson, W. A., Marrack, P. &
Kappler, J. Structures of an MHC class II molecule
with covalently bound single peptides. Science 272,
1001–1004 (1996).
34. Yamagata, K. et al. Aspartic acid at position 57 of DQ
β chain does not protect against type 1 (insulin-
dependent) diabetes mellitus in Japanese subjects.
Diabetologia 32, 762–764 (1989).
35. Awata, T., Kuzuya, T., Matsuda, A., Iwamoto, Y. &
Kanazawa, Y. Genetic analysis of HLA class II alleles
and susceptibility to type 1 (insulin-dependent)
diabetes mellitus in Japanese subjects. Diabetologia
35, 419–424 (1992).

NATURE REVIEWS | IMMUNOLOGY VOLUME 6 | APRIL 2006 | 281

© 2006 Nature Publishing Group
REVIEWS
36. Zhu, Y., Rudensky, A. Y., Corper, A. L., Teyton, L. &
Wilson, I. A. Crystal structure of MHC class II I-Ab
in complex with a human CLIP peptide: prediction
of an I-Ab peptide-binding motif. J. Mol. Biol. 326,
1157–1174 (2003).
37. Ettinger, R. A. & Kwok, W. W. A peptide binding motif
for HLA-DQA1*0102/DQB1*0602, the class II
MHC molecule associated with dominant protection
in insulin-dependent diabetes mellitus. J. Immunol.
160, 2365–2373 (1998).
38. Qiao, S. W. et al. Refining the rules of gliadin T cell
epitope binding to the disease-associated DQ2
molecule in celiac disease: importance of proline
spacing and glutamine deamidation. J. Immunol.
175, 254–261 (2005).
39. Arentz-Hansen, H. et al. The intestinal T cell response
to α-gliadin in adult celiac disease is focused on
a single deamidated glutamine targeted by tissue
transglutaminase. J. Exp. Med. 191, 603–612 (2000).
40. Lee, D. M. & Weinblatt, M. E. Rheumatoid arthritis.
Lancet 358, 903–911 (2001).
41. Stastny, P. Association of the B-cell alloantigen
DRw4 with rheumatoid arthritis. N. Engl. J. Med.
298, 869–871 (1978).
42. Wordsworth, B. P. et al. HLA-DR4 subtype frequencies
in rheumatoid arthritis indicate that DRB1 is the major
susceptibility locus within the HLA class II region. Proc.
Natl Acad. Sci. USA 86, 10049–10053 (1989).
43. Gregersen, P. K., Silver, J. & Winchester, R. J.
The shared epitope hypothesis. An approach to
understanding the molecular genetics of susceptibility
to rheumatoid arthritis. Arthritis Rheum. 30,
1205–1213 (1987).
44. Hammer, J. et al. Peptide binding specificity of HLA-
DR4 molecules: correlation with rheumatoid arthritis
association. J. Exp. Med. 181, 1847–1855 (1995).
45. Van Boxel, J. A. & Paget, S. A. Predominantly
T-cell infiltrate in rheumatoid synovial membranes.
N. Engl. J. Med. 293, 517–520 (1975).
46. Andersson, E. C. et al. Definition of MHC and
T cell receptor contacts in the HLA-DR4-restricted
immunodominant epitope in type II collagen and
characterization of collagen-induced arthritis in
HLA-DR4 and human CD4 transgenic mice.
Proc. Natl Acad. Sci. USA 95, 7574–7579 (1998).
47. Rosloniec, E. F. et al. An HLA-DR1 transgene confers
susceptibility to collagen-induced arthritis elicited
with human type II collagen. J. Exp. Med. 185,
1113–1122 (1997).
48. Kremer, J. M. et al. Treatment of rheumatoid arthritis
by selective inhibition of T-cell activation with fusion
protein CTLA4Ig. N. Engl. J. Med. 349, 1907–1915
(2003).
49. Fugger, L., Rothbard, J. B. & Sonderstrup-McDevitt, G.
Specificity of an HLA-DRB1*0401-restricted T cell
response to type II collagen. Eur. J. Immunol. 26,
928–933 (1996).
50. Svendsen, P. et al. Tracking of proinflammatory
collagen-specific T cells in early and late collagen-
induced arthritis in humanized mice. J. Immunol.
173, 7037–7045 (2004).
51. Latham, K. A., Whittington, K. B., Zhou, R., Qian, Z. &
Rosloniec, E. F. Ex vivo characterization of the
autoimmune T cell response in the HLA-DR1 mouse
model of collagen-induced arthritis reveals long-term
activation of type II collagen-specific cells and their
presence in arthritic joints. J. Immunol. 174,
3978–3985 (2005).
282 | APRIL 2006 | VOLUME 6
52. Dessen, A., Lawrence, C. M., Cupo, S., Zaller, D. M. &
Wiley, D. C. X-ray crystal structure of HLA-DR4
(DRA*0101, DRB1*0401) complexed with a peptide
from human collagen II. Immunity 7, 473–481
(1997).
The structural analysis of the rheumatoid-
arthritis-associated molecule HLA-DR4 revealed
a peptide-binding groove that is well suited to
binding the glycine-rich peptides derived from
type II collagen. The structure showed that a
sequence common to all the main rheumatoid-
arthritis-associated MHC class II molecules allows
negatively charged P4 residues to be
accommodated.
53. Hill, J. A. et al. Cutting edge: the conversion of
arginine to citrulline allows for a high-affinity peptide
interaction with the rheumatoid arthritis-associated
HLA-DRB1*0401 MHC class II molecule. J. Immunol.
171, 538–541 (2003).
54. Marshall, K. et al. Prediction of peptide affinity to
HLA-DRB1*0401. J. Immunol. 154, 5927–5933
(1995).
55. Hennecke, J. & Wiley, D. C. Structure of a complex of
the human α/β T cell receptor (TCR) HA1.7, influenza
hemagglutinin peptide, and major histocompatibility
complex class II molecule, HLA-DR4 (DRA*0101
and DRB1*0401): insight into TCR cross-restriction
and alloreactivity. J. Exp. Med. 195, 571–581
(2002).
56. Hammer, J., Takacs, B. & Sinigaglia, F. Identification
of a motif for HLA-DR1 binding peptides using M13
display libraries. J. Exp. Med. 176, 1007–1013
(1992).
57. Wucherpfennig, K. W. et al. Structural basis for major
histocompatibility complex (MHC)-linked susceptibility
to autoimmunity: charged residues of a single MHC
binding pocket confer selective presentation of self-
peptides in pemphigus vulgaris. Proc. Natl Acad. Sci.
USA 92, 11935–11939 (1995).
58. Keegan, B. M. & Noseworthy, J. H. Multiple sclerosis.
Annu. Rev. Med. 53, 285–302 (2002).
59. Fogdell, A., Hillert, J., Sachs, C. & Olerup, O.
The multiple sclerosis- and narcolepsy-associated
HLA class II haplotype includes the DRB5*0101 allele.
Tissue Antigens 46, 333–336 (1995).
60. Oksenberg, J. R. et al. Mapping multiple sclerosis
susceptibility to the HLA-DR locus in African
Americans. Am. J. Hum. Genet. 74, 160–167
(2004).
61. Sospedra, M. & Martin, R. Immunology of multiple
sclerosis. Annu. Rev. Immunol. 23, 683–747
(2005).
62. Oksenberg, J. R. et al. Selection for T-cell receptor
Vβ-Dβ-Jβ gene rearrangements with specificity for
a myelin basic protein peptide in brain lesions
of multiple sclerosis. Nature 362, 68–70 (1993).
63. Krogsgaard, M. et al. Visualization of myelin basic
protein (MBP) T cell epitopes in multiple sclerosis
lesions using a monoclonal antibody specific for
the human histocompatibility leukocyte antigen
(HLA)-DR2-MBP 85–99 complex. J. Exp. Med.
191, 1395–1412 (2000).
64. Madsen, L. S. et al. A humanized model for multiple
sclerosis using HLA-DR2 and a human T-cell receptor.
Nature Genet. 23, 343–347 (1999).
65. Lang, H. L. et al. A functional and structural basis
for TCR crossreactivity in multiple sclerosis. Nature
Immunol. 3, 940–943 (2002).
© 2006 Nature Publishing Group
The structure of HLA-DR2a, in complex with a
peptide derived from EBV, was compared with
the structure of HLA-DR2b in complex with the
immunodominant epitope from MBP, revealing
a clear example of molecular mimicry. These
complexes stimulate crossreactive T cells.
66. Wucherpfennig, K. W. et al. Structural requirements
for binding of an immunodominant myelin basic
protein peptide to DR2 isotypes and for its recognition
by human T cell clones. J. Exp. Med. 179, 279–290
(1994).
67. Vogt, A. B. et al. Ligand motifs of HLA-DRB5*0101
and DRB1*1501 molecules delineated from self-
peptides. J. Immunol. 153, 1665–1673 (1994).
68. Hahn, M., Nicholson, M. J., Pyrdol, J. &
Wucherpfennig, K. W. Unconventional topology of self
peptide-major histocompatibility complex binding by
a human autoimmune T cell receptor. Nature
Immunol. 6, 490–496 (2005).
69. Li, Y. et al. Structure of a human autoimmune TCR
bound to a myelin basic protein self-peptide and a
multiple sclerosis-associated MHC class II molecule.
EMBO J. 24, 2968–2979 (2005).
70. Maynard, J. et al. Structure of an autoimmune
T cell receptor complexed with class II peptide-MHC:
insights into MHC bias and antigen specificity.
Immunity 22, 81–92 (2005).
71. Bornstein, M. B. et al. A pilot trial of Cop 1
in exacerbating-remitting multiple sclerosis.
N. Engl. J. Med. 317, 408–414 (1987).
72. Stern, J. N. et al. Peptide 15-mers of defined
sequence that substitute for random amino acid
copolymers in amelioration of experimental
autoimmune encephalomyelitis. Proc. Natl Acad. Sci.
USA 102, 1620–1625 (2005).
73. Stuart, D. I., Levine, M., Muirhead, H. &
Stammers, D. K. Crystal structure of cat muscle
pyruvate kinase at a resolution of 2.6 Å. J. Mol. Biol.
134, 109–142 (1979).
Acknowledgements
We thank R. Esnouf for discussion. Work in the authors’ labo-
ratories is supported by the Danish and British Medical
Research Councils, Cancer Research UK, the Karen Elise
Jensen Foundation, the Lundbeck Foundation, the Danish
Multiple Sclerosis Society, the European Commission
Integrated Programme SPINE (Structural Proteomics in
Europe) and the European Commission Descartes Prize. E.Y.J.
is a Cancer Research UK Principal Research Fellow.
Competing interests statement
The authors declare no competing financial interests.
DATABASES
The following terms in this article are linked online to:
Entrez Gene: http://www.ncbi.nlm.nih.gov/entrez/query.
fcgi?db=gene
HLA-B | MBP
OMIM: http://www.ncbi.nlm.nih.gov/entrez/query.
fcgi?db=OMIM
Coeliac disease | multiple sclerosis | narcolepsy | rheumatoid
arthritis | type 1 diabetes
FURTHER INFORMATION
E. Yvonne Jones’s laboratory: http://www.stwbi.ox.ac.uk
RCSB Protein Data Bank: http://www.rcsb.org/pdb
Access to this links box is available online
www.nature.com/reviews/immunol