REVIEWS

HLA variation and disease

Calliope A. Dendrou1, Jan Petersen2,3, Jamie Rossjohn2–4 and Lars Fugger5,6
Abstract | Fifty years since the first description of an association between HLA and human
disease, HLA molecules have proven to be central to physiology, protective immunity and
deleterious, disease-causing autoimmune reactivity. Technological advances have enabled
pivotal progress in the determination of the molecular mechanisms that underpin the association
between HLA genetics and functional outcome. Here, we review our current understanding of
HLA molecules as the fundamental platform for immune surveillance and responsiveness in
health and disease. We evaluate the scope for personalized antigen-specific disease prevention,
whereby harnessing HLA–ligand interactions for clinical benefit is becoming a realistic prospect.

Linkage disequilibrium
The nonrandom association of
alleles at different loci, for
example, owing to close
physical proximity within a
genomic region.
Gene conversion
The process by which one
allele is converted to another
by mismatch repair
mechanisms.
Heterozygote advantage
The increased relative fitness of
an organism conferred by
having two different forms of a
genetic variant, as opposed to
having two identical copies of
either of the two forms.
Frequency-dependent
selection
An evolutionary process
whereby fitness of a given
phenotype depends on its
frequency relative to other
phenotypes in a study
population. Positive selection
will occur if the fitness of the
phenotype increases as its
frequency increases, whereas
negative selection will occur if
the fitness decreases as the
frequency of the phenotype
increases.
Correspondence to L.F. and J.R. 
lars.fugger@imm.ox.ac.uk;
jamie.rossjohn@monash.edu
doi:10.1038/nri.2017.143
Published online 2 Jan 2018
HLA gene variation was first linked to human disease
with the discovery of the association between HLA‑B
and Hodgkin lymphoma1. Since then, the MHC has
been established as the region of the genome that
is associated with the greatest number of human
diseases 2. The majority of these diseases have an
immuno­logical component, which is consistent with
the enrichment for key immune genes within the MHC
region. The MHC is divided into three subclasses: the
class I region, which includes the classical, highly poly­
morphic HLA‑A, HLA‑B and HLA‑C genes, as well as
the nonclassical HLA‑E, HLA‑F and HLA‑G genes,
which exhibit limited polymorphism; the class  II
region, which includes the HLA‑DPA1, HLA‑DPB1,
HLA‑DQA1, HLA‑DQA2, HLA‑DQB1, HLA‑DQB2,
HLA-DRA, HLA‑DRB1, HLA‑DRB2, HLA‑DRB3,
HLA‑DRB4 and HLA‑DRB5 genes as well as less var­
iable genes involved in antigen processing and pres­
entation; and the class III region, which contains genes
implicated in inflammatory responses, leukocyte matu­
ration and the complement cascade. Nonpolymorphic
antigen-presenting molecules that are encoded by genes
outside of the MHC locus (for example, CD1a–d and
MHC class-I‑related protein) have been reviewed else­
where3 and are not the main focus of this Review. The
MHC locus has a dense clustering of immune-­relevant
genes that can show extreme polymorphism and can
be in strong linkage disequilibrium; this complexity has
complicated the determination of the exact genes and
alleles that are responsible for signals of disease associ­
ation in the region2. However, the classical HLA class I
and II genes are often thought to drive associations
with disease. This is in keeping with the crucial role
that HLA class I and II molecules have in presenting a
vast array of antigenic peptides to T cells, thus enabling
the immune system to discriminate between self and
non-self.
More than 15,000 different classical HLA class I and
II alleles have been identified4. The diversity of HLA
molecules expressed in humans at the population level
is likely to maximize the probability that at least some
individuals within the general population can mount an
immune attack against an emerging infection and survive.
Genetically, this diversity may arise partly due to point
mutations but also more frequently due to mecha­nisms
such as gene conversion2. Much of the poly­morphism in
HLA genes leads to nonsynonymous amino acid changes
in the peptide-binding groove of HLA molecules, thus
indicating a strong selection pressure on the peptide-­
binding groove and the importance of the HLA–peptide
interaction. Two main mechanisms of balancing selection
have been postulated that favour the maintenance of HLA
diversity in a population over time: heterozygote advan-
tage and frequency-dependent selection5. The potential of
pathogens to drive HLA selection is suggested through
investigations of emerging infectious diseases, such as
HIV; in HIV infection, HLA class I allele heterozygosity,
as opposed to homozygosity, correlates with delayed dis­
ease progression6. However, resistance to infection may
not be the only driver of HLA polymorphism, as HLA
molecules are implicated in reproductive fitness via the
regulation of maternal and fetal survival7.
In an individual who is heterozygous at each of the
six classical class I or II HLA loci, it has been estimated
that an antigen-presenting cell (APC) could theo­retically
present over 1012 different peptides8. The specific recog­
nition of HLA–peptide combinations is mediated by αβ T
cell receptors (TCRs) on CD8+ T cells, which bind class I
molecules, and on CD4+ T cells, which bind class II mole­
cules. The TCR displays substantial sequence hetero­
geneity that arises as different variable (V), diversity (D)
and joining (J) gene segments come together through
somatic, convergent recombination; additional variation
is introduced by the semi-random insertion or deletion of

NATURE REVIEWS | IMMUNOLOGY ADVANCE ONLINE PUBLICATION | 1

©
2
0
1
7
M
a
c
m
i
l
l
a
n
P
u
b
l
i
s
h
e
r
s
L
i
m
i
t
e
d
,
p
a
r
t
o
f
S
p
r
i
n
g
e
r
N
a
t
u
r
e
.
A
l
l
r
i
g
h
t
s
r
e
s
e
r
v
e
d
.
REVIEWS

Author addresses
1
Nuffield Department of Medicine, The Wellcome Centre for Human Genetics,
University of Oxford, Roosevelt Drive, Oxford OX3 7BN, UK.
2
Australian Research Council Centre of Excellence for Advanced Molecular Imaging,
Monash University, Wellington Road, Clayton, Victoria 3800, Australia.
3
Infection and Immunity Programme and Department of Biochemistry and Molecular
Biology, Biomedicine Discovery Institute, Monash University, Wellington Road, Clayton,
Victoria 3800, Australia.
4
Division of Infection and Immunity, Cardiff University School of Medicine, Heath Park,
Cardiff CF14 4XN, UK.
5
Danish National Research Foundation Centre PERSIMUNE, Rigshospitalet,
University of Copenhagen, Copenhagen DK-2100, Denmark.
6
Oxford Centre for Neuroinflammation, Nuffield Department of Clinical Neurosciences,
Division of Clinical Neurology and Medical Research Council Human Immunology Unit,
Weatherall Institute of Molecular Medicine, John Radcliffe Hospital,
University of Oxford, Headley Way, Oxford OX3 9DS, UK.
nucleotides at segment junctions9. As a result, each TCR
chain bears three highly variable complementarity deter­
mining region (CDR) loops: CDR1 and CDR2 diversity
is germline-encoded, whereas the CDR3 loop is hyper­
variable, as it is encoded by the sequences generated by
nucleotide insertion and deletion. In addition to binding
TCRs (the focus of this Review), HLA class I molecules
can also bind killer cell immunoglobulin-like receptors
(KIRs) and C‑type lectin-like CD94/NKG2 family of
receptors found on natural killer (NK) cells10 as well as
leukocyte immunoglobulin-like receptors (LILRs) found
on APCs and several other cell types11. These interactions
are thought to be less peptide-specific than HLA–TCR
binding, although there is evidence that they can be
tuned in a peptide-dependent fashion11–13.
The specificity of HLA–peptide–TCR tripartite inter­
actions is fundamental in enabling the adaptive immune
system to mount an efficient and appropriate response
to counteract infection and malignancy while maintain­
ing self tolerance and preventing autoimmune disease.
Understanding the molecular principles that govern
these interactions may give mechanistic insight into the
role of HLA in driving and protecting against immuno­
pathology; this presents an ongoing biomedical research
challenge but also holds much therapeutic promise.
In this Review, we consider our current understand­
ing of HLA function in immune-mediated diseases
with a particular emphasis on autoimmune conditions.
Following this, we discuss how technological progress
is facilitating the study of HLA biology, and we explore
the prospects for developing and using antigen-specific
strategies for disease prevention and therapy.
HLA interactions in disease
The structural determination of disease-relevant
peptide–HLA and HLA–peptide–TCR complexes is
crucial for the elucidation of the molecular mechanisms
responsible for the development of specific protective
HLA restriction immunity against pathogens, as well as for the deleter­
A property of T cells whereby a ious T cell reactivity that promotes disease. One model
given T cell receptor will proposes that the HLA–peptide–TCR interactions that
recognize and respond to an
antigen only when it is
underpin self and non-self discrimination are guided by
presented by a particular HLA a set of rules that are perturbed in immunopathology.
molecule. The study of more than 50 distinct HLA–peptide–TCR
Figure 1 | HLA-dependent molecular mechanisms of ▶
peptide and T cell receptor binding. a | Conventional
docking. Central docking geometry with the T cell receptor
(TCR) α-chains and β-chains positioned diagonally with
respect to the peptide, thus maximizing the coverage of the
peptide–HLA surface and allowing the TCR to access
exposed peptide side chains. b | Alternate docking.
Unconventional amino-terminal docking geometry with
α-chains and β-chains positioned perpendicularly with
respect to the peptide axis, thus limiting coverage of the
peptide–HLA surface and preventing interactions between
the TCR and the carboxy‑terminal portion of the peptide.
Top row of parts a and b depicts the interface viewed along
the TCR–peptide–HLA axis; bottom row of parts a and b,
and parts c–i depict a cross section of the interface along
the peptide backbone. c | TCR stabilization of weak
peptide–HLA complexes. Left: unstable peptide–HLA
complex with suboptimal peptide anchor residues.
Right: TCR binding to the peptide–HLA stabilizes the weak
complex interactions. d | Altered register. Polymorphic
differences between HLA‑DR1 and HLA‑DR15 result in
different anchor residues being used for α3135–145 peptide
presentation, resulting in vastly different antigenic surfaces.
e | Hotspot binding and molecular mimicry. TCRs that form
focal interactions with only a limited portion of the cognate
peptide–MHC complex (left) tend to recognize mimic
peptides (right) that differ in their primary sequence.
f | Post-translational modification (PTM)-altered HLA
binding. PTMs can substantially change peptide binding to
the HLA and expose new features to the TCR. Left: a native
peptide is unable to bind the HLA. Right: the peptide
carrying PTMs is presented by the HLA because of its
changed chemistry and exposes a modified residue for TCR
recognition. g | PTM-altered antigen processing. Left: an
antigenic peptide is cleaved (red dashes) into fragments too
short for HLA presentation. Right: PTM of the same peptide
obscures a protease site, resulting in productive antigen
processing for HLA presentation. h | HLA presentation of
hybrid peptides. Top: two peptides with suboptimal anchor
residues are poorly bound by the HLA. Bottom: the hybrid
peptide has enhanced anchor residues for HLA binding and
provides novel features for TCR recognition. i | Stability and
peptide repertoire. The comparatively wide and shallow
peptide-binding groove of HLA‑C*05 enables presentation
of peptides with large, aromatic anchor residues, whereas
the deep and narrow peptide-binding groove of HLA‑C*07
selects for smaller anchor residues. This difference
contributes to HLA‑C*05 forming more stable complexes,
which results in higher expression levels on the cell surface
of HLA‑C*05 compared with HLA‑C*07.
complex structures has suggested that there are certain
principles that govern HLA restriction and TCR docking
geometry, but that they are malleable; instead of a conven­
tional, central TCR docking geometry (FIG. 1a), a variety
of molecular mechanisms that affect HLA–peptide–TCR
interactions are implicated in disease14 (FIG. 1b–i). A still
open question is whether any of these mechanisms are
shared between conditions with common aetiological
features. For example, HLA–peptide–TCR binding has
been of particular interest in the context of different
autoimmune diseases. A common feature of these dis­
eases is tissue damage facilitated by autoreactive T cells
that escape both negative selection in the thymus and
peripheral tolerance mechanisms. However, it remains

2 | ADVANCE ONLINE PUBLICATION www.nature.com/nri

©
2
0
1
7
M
a
c
m
i
l
l
a
n
P
u
b
l
i
s
h
e
r
s
L
i
m
i
t
e
d
,
p
a
r
t
o
f
S
p
r
i
n
g
e
r
N
a
t
u
r
e
.
A
l
l
r
i
g
h
t
s
r
e
s
e
r
v
e
d
.
 REVIEWS

a Conventional docking b Alternate docking c TCR stabilization of weak peptide–HLA complexes

TCRβ Peptide
TCRβ

C C
N N
HLA
TCRα TCRα
Peptide

HLA

d Altered register e Hotspot binding and molecular mimicry

Focal interaction
αβTCR

α3–HLA-DR1 α3–HLA-DR15 Cognate peptide–MHC Mimic peptide

f PTM-altered HLA binding g PTM-altered antigen processing

Site of peptide
cleavage

h HLA presentation of hybrid peptides i Stability and peptide repertoire

HLA-C*05 HLA-C*07
Hybrid peptide

Native
Mimic peptide
peptide PTM residues
residues
residues

to be determined whether autoreactive T cells are directly
generated and activated by the mechanisms that are
described below, such as atypical HLA–peptide–TCR
binding orientation, low-affinity peptide binding that
facilitates thymic escape, TCR-mediated stabilization of
weak peptide–HLA interaction and presentation of pep­
tides in a different binding register. Other mechanisms
that may generate and activate autoreactive T cells are
likely to be driven by epitope variation, including molec­
ular mimicry, post-translational epitope modification
and the generation of hybrid peptides, and processes that
regulate HLA expression and stability are also emerging
as being increasingly relevant (FIG. 1b–i).
Alternate docking. The Ob.1A12 TCR, which was orig­
inally isolated from CD4+ Nature
T cellsReviews
from a |patient with
Immunology
relapsing–remitting multiple sclerosis, recognizes resi­
dues 85–99 of the immunodominant peptide of myelin
basic protein (MBP), which is presented by the dis­
ease-associated HLA‑DR15. The pathogenic potential of
the Ob.1A12 TCR in promoting central nervous system
(CNS) disease has been demonstrated in transgenic mice
expressing Ob.1A12 TCR and HLA‑DR15 and immu­
nized with the MBP peptide15. At the molecular level, the
Ob.1A12 TCR displays an off-centre binding topology
(FIG. 1b), whereby it docks over the amino‑terminal side
of the peptide–HLA platform16. This alternate docking

NATURE REVIEWS | IMMUNOLOGY ADVANCE ONLINE PUBLICATION | 3

©
2
0
1
7
M
a
c
m
i
l
l
a
n
P
u
b
l
i
s
h
e
r
s
L
i
m
i
t
e
d
,
p
a
r
t
o
f
S
p
r
i
n
g
e
r
N
a
t
u
r
e
.
A
l
l
r
i
g
h
t
s
r
e
s
e
r
v
e
d
.
REVIEWS

topology was associated with reduced affinity of the peptide-specific TCRs22 and probably exerts its com­
TCR for the peptide–HLA complex and has an uncon­ pensatory effect through the maximal binding of MBP.
ventional geometry that may have allowed autoreactive As the HLA‑DR4‑MBP peptide interaction is highly
T cell evasion of thymic negative selection, although unstable, a low autoantigen density in the thymus could
still maintaining the capacity of the TCR to stimulate an have enabled MS2‑3C8‑TCR-bearing T cells to receive
inflammatory response in the periphery (TABLE 1). a weak signal for positive selection that would not reach
the required threshold for negative selection. Once in
Low-affinity-mediated thymic escape. The 1E6 TCR, the periphery, MS2‑3C8 TCR+ cells could then be acti­
which was isolated from CD8+ T cells from a newly vated in response to higher antigen densities, potentially
diagnosed patient with type 1 diabetes (T1D), medi­ triggering an autoimmune response against the CNS.
ates pancreatic islet β-cell killing by recognizing amino
acids 15–24 of the preproinsulin (PPI) signal peptide Altered register. The peptide binding register refers to
presented by disease-associated HLA‑A*02:01. The 1E6 the ~9‑mer window of a peptide that sits directly within
TCR binds this peptide–HLA complex in a conventional the peptide-binding groove at a given time. Alterations
fashion, which is a characteristic of the majority of HLA in this register, whereby the same peptide binds a
class-I‑restricted TCRs studied to date17. However, the peptide-­binding groove utilizing a different 9‑mer win­
affinity of the 1E6 TCR for the peptide–HLA complex dow, can potentially have a differential impact on the
is very low and may be below the threshold for thymic generation of autoreactive T cells. Diabetogenic T cells
nega­tive selection induction, but it is nevertheless suf­ from nonobese diabetic (NOD) mice recognize an insu­
ficient to trigger β-cell killing (TABLE 1). Given the low-­
lin B‑chain peptide, which is bound to the MHC class II
affinity mode of binding of the TCR, the level of the I-Ag7 molecule in a low-affinity register 23, thus demon­
PPI peptide presented in the thymus may be related strating the potential importance of the peptide bind­
to the escape from negative selection. PPI expression in ing register. Similar results were seen in patients with
the thymus is under the influence of a variable number T1D, and there is also some evidence in these patients of
of tandem repeats (VNTRs) in the 5ʹ regulatory region of potential priming of diabetogenic T cells through molec­
the INS gene. The presence of few VNTRs is a strong risk ular mimicry of the self peptide (discussed below) with
factor for T1D and is associated with lower levels of PPI- microbial antigens24.
encoding mRNA transcripts in the thymus18. Notably, In addition to the potential role of the peptide
the patient from whom the 1E6 TCR was derived was binding register in increasing the risk of auto­immune
homozygous for VNTR alleles with few repeats in the disease, another fundamental question is its rele­
INS gene. In the periphery, the escapee CD8+ 1E6‑TCR- vance to disease resistance. This was the focus of a
bearing cells may be primed by microbial-derived recent study investigating the mechanisms by which
antigens, particularly as this TCR focuses on only two HLA‑DR15 and HLA‑DR1 confer respective risk of
residues within the PPI peptide, suggesting a higher or dominant protection against the renal autoimmune
chance of cross recognition. This model is supported by condition Goodpasture disease25. Autoreactive T cells
the finding that the 1E6 TCR can bind with a high affin­ specific for the α3 chain of type IV collagen (α3135–145)
ity to a peptide from the human pathogen Clostridium were expanded in patients with Goodpasture disease,
asparagi­forme19. In the pancreas, the pro-­inflammatory and immunization of HLA‑DR15 transgenic mice
environment that accompanies T1D development with the α3135–145 self peptide led to the infiltration
induces islet β-cells to upregulate HLA class I expres­ of peptide-­specific T cells into the kidney, resulting
sion, thus enabling the autoreactive T cells to exert theirin disease. Compared with HLA‑DR15, HLA‑DR1
cytotoxic function17. has a distinct peptide repertoire and binding prefer­
ences and presents the α3135–145 epitope in a different
T cell receptor stabilization of weak peptide–HLA binding register (FIG. 1d; TABLE 1). This altered regis­
complexes. Intriguingly, there is also evidence that ter promotes the generation of peptide-specific CD4+
auto­reactive TCRs can bind self peptide–HLA com­ T cells with a predominantly forkhead box protein P3
plexes with a conventional binding topology and (FOXP3)+ regulatory phenotype in both HLA‑DR1+
a high affinity 20. The MS2‑3C8 TCR was isolated and HLA‑DR15+HLA‑DR1+ transgenic mice, which
from CD4 + T  cells from a patient with relapsing–­ were resistant to disease. Consistent with this, healthy
remitting multiple sclerosis and binds an MBP pep­ human donors who were HLA‑DR15+ or HLA‑DR1+
tide presented by HLA‑DR4. The pathogenic potential displayed altered α3135–145-specific TCR usage and had
of this TCR has been demonstrated through the use of peptide-specific T cells that primarily showed either an
HLA‑DR4‑transgenic mice that developed CNS disease effector or a regulatory pheno­type, respectively 25. Thus,
via adoptive transfer of MS2‑3C8 TCR+CD4+ T cells in one mechanistic basis for dominant protective effects
the absence of antigen administration21. Structural ana­ of HLA alleles in autoimmunity, via the effect on the
lyses have shown that although the MBP peptide binds generation of CD4+CD25+ regulatory T (Treg) cells, has
HLA‑DR4 only weakly and is unstably accommodated been revealed.
in the peptide-binding cleft, the MS2‑3C8 TCR binds
the complex tightly and compensates for the weak MBP ‘Hotspot’ molecular mimicry. In the ‘hotspot’ molecular
peptide–HLA interaction20 (FIG. 1c; TABLE 1). The affinity mimicry model, autoreactive TCRs with a highly focused
of this TCR is in the range of that observed for pathogen footprint, which predominantly binds only a small area

4 | ADVANCE ONLINE PUBLICATION www.nature.com/nri

©
2
0
1
7
M
a
c
m
i
l
l
a
n
P
u
b
l
i
s
h
e
r
s
L
i
m
i
t
e
d
,
p
a
r
t
o
f
S
p
r
i
n
g
e
r
N
a
t
u
r
e
.
A
l
l
r
i
g
h
t
s
r
e
s
e
r
v
e
d
.
 REVIEWS

Table 1 | Autoimmune disease–HLA associations for which molecular mechanisms of action have been identified
HLA allomorph (effect Autoimmune Molecular mechanism of action Refs
on disease) disease
Alternate docking
HLA‑DR15 (risk) Multiple sclerosis • Alternate docking of a TCR may allow MBP peptide-specific T cells to escape thymic 15,16
selection
• In the periphery, such cells may be cross-reactive with microbial peptides
Low-affinity-mediated thymic escape
HLA‑A*02:01 (risk) T1D • Low affinity of a TCR for preproinsulin signal peptide–HLA may allow autoreactive 17–19
T cells to escape thymic selection
• This mechanism may depend on the thymic expression level of the autoantigen
• In the periphery, these cells may bind microbial peptides with a high affinity
TCR stabilization of weak peptide–HLA complexes
HLA‑DR4 (risk) Multiple sclerosis • Weak MBP peptide–HLA‑DR4 interaction is stabilized by binding of a 20–22
patient-derived TCR
• Low autoantigen density in the thymus may enable escape of autoreactive T cells
from negative selection
• Higher autoantigen density in the periphery could trigger an autoreactive response
Altered register
HLA‑DQ2 and HLA‑DQ8 T1D • Insulin B‑chain peptide binds HLA‑DQ2 and HLA-DQ8 with a low-affinity peptide 23,24
(risk) register
• Low affinity may allow autoreactive T cells to escape thymic selection
• In the periphery, autoreactive T cells may be cross-reactive with microbial peptides
via molecular mimicry
HLA‑DR15 (risk) Goodpasture disease • HLA‑DR15 and HLA‑DR1 both bind an α3‑chain of type IV collagen peptide 25
HLA‑DR1 (dominant (α3135–145) but with a different binding register
protection) • When HLA‑DR1 is present, this promotes the generation of autoantigen-specific
regulatory T cells as opposed to effectors
‘Hotspot’ molecular mimicry
HLA‑DR15 (risk) Multiple sclerosis • The probability of aberrant, off-target TCR reactivity induced by pathogen-derived 16,
peptides is increased for autoreactive TCRs with a highly focused footprint that 26–28,
predominantly binds only a small area of peptide 32
• Patient TCRs have been identified that cross-react with HLA‑DR15‑restricted MBP
and Escherichia coli or EBV-derived peptides
Post-translational modification
HLA‑DQ2.5 and HLA‑DQ8 Coeliac disease • HLA‑DQ2.5 and HLA-DQ8 present deamidated gliadins with a high kinetic stability, 38–41
(risk) leading to sustained antigen presentation that may promote a pathogenic T cell
response
HLA‑DR4 (risk) Rheumatoid arthritis • Citrullination of autoantigens (for example, vimentin) facilitates binding to HLA‑DR4 34,35
• Citrullination can also modify peptide cleavage, enabling retention of autoantigens
HLA‑DR4 (risk) T1D • Vicinal disulfide bond creation in an insulin A‑chain peptide presented by HLA‑DR4 37
is needed for recognition by a patient-derived autoreactive TCR
Hybrid peptides
HLA‑DQ8 (risk) T1D • Hybrid proinsulin peptides present in pancreatic β-cells may drive a breakdown in 51
immune tolerance
Regulation of HLA expression
MHC risk variants in distal Systemic lupus • Risk variants alter the binding of the IRF4 and CTCF factors that regulate the 52
intergenic XL9 regulatory erythematosus transcription of HLA‑DRB1, HLA‑DQA1 and HLA‑DQB1, resulting in a 2.5‑fold
element increase in the levels of the HLA‑DR and HLA‑DQ proteins implicated in lupus
Highly expressed HLA‑C Crohn’s disease • HLA‑C allotypes with a high expression correlate with Crohn’s disease risk 53
allotypes (risk) • This expression is partly regulated by the microRNA miR‑148a, which is itself
subject to regulation by genetic variation
HLA stability
HLA‑DQ2 and HLA‑DQ8 T1D, coeliac disease • HLA‑DQ2 and HLA-DQ8 instability could confer risk by allowing escape of 71–76
(risk) autoreactive T cells from thymic negative selection but also by the formation of
T1D, autoimmune
weak peptide–HLA complexes in the periphery with a range of autoantigens
HLA‑DQ6 (protection) polyglandular
• By a converse mechanism, the stability of HLA‑DQ6 may confer protection
syndrome, IgA
against autoimmune disease; despite such a mechanism, HLA‑DQ6 confers risk of
deficiency
narcolepsy
CTCF, transcriptional regulator CTCF; EBV, Epstein–Barr virus; IgA, immunoglobulin A; IRF4, interferon regulatory factor 4; MBP, myelin basic protein;
T1D, type 1 diabetes; TCR, T cell receptor.

NATURE REVIEWS | IMMUNOLOGY ADVANCE ONLINE PUBLICATION | 5

©
2
0
1
7
M
a
c
m
i
l
l
a
n
P
u
b
l
i
s
h
e
r
s
L
i
m
i
t
e
d
,
p
a
r
t
o
f
S
p
r
i
n
g
e
r
N
a
t
u
r
e
.
A
l
l
r
i
g
h
t
s
r
e
s
e
r
v
e
d
.
REVIEWS

Neoantigens
Non-germline-encoded (and
thus non-inherited) antigens
that may arise because of
somatic mutation (as in cancer)
or other processes such as
post-translational modification
and splicing together of
peptides.
of peptide, increase the probability of aberrant, off-­target
TCR reactivity that is induced by pathogen-­derived
peptides (FIG. 1e). For example, the alternate docking
strategy used by the Ob.1A12 TCR when binding MBP
peptide16 was also observed when this TCR recognizes
a cross-reactive peptide from Escherichia coli 26 (TABLE 1).
Importantly, this finding provided the first evidence that
atypical binding orientations are not only a feature of
aberrant reactivity but are also implicated in responses
to pathogen and that hotspot molecular mimicry may be
a trigger for multiple sclerosis26. Similarly, the Hy.1B11
TCR, which was derived from a patient with multiple
sclerosis, binds the MBP85–99 peptide, but only one of the
CDR3 loops makes contact with the peptide, thereby
limiting the specificity of the TCR to only a few resi­
dues within the peptide27. As the Hy.1B11 TCR relies on
only a few key residues, it can be activated not only by
MBP but also by peptides from several different viruses,
although these viral peptides show a low similarity only
to the MBP85–99 peptide28. The potential importance of
molecular mimicry between self and microbial peptides
in driving autoimmune disease is also supported by epi­
demiological investigation and shared genetic factors.
For example, for patients with multiple sclerosis, there
is an increased comorbidity with Epstein–Barr virus
(EBV)-dependent infectious mononucleosis29, and hav­
ing HLA‑DRB1*15:01 increases the risk of developing
both conditions30,31; these findings present the possibil­
ity that autoreactive T cells could in some instances be
activated by EBV antigens. Although numerous differ­
ent mechanisms may act to trigger multiple sclerosis, a
patient-derived TCR has been identified that recognizes
both HLA‑DRB1*15:01‑restricted MBP and an EBV
peptide presented by HLA‑DRB5*01:01 (TABLE 1), which
is encoded by an allele in high linkage disequilibrium
with the HLA‑DRB1*15:01 allele32. This supports the
model of molecular mimicry acting to increase the risk
of autoimmune disease.
Post-translational modification. The HLA-mediated
presentation of neoantigens has been proposed as another
mechanism by which T cells could mount an attack against
self tissue (FIG. 1f). Post-translational modification (PTM)
can generate such neoantigens through deamidation and
citrullination, processes that have been implicated in sev­
eral autoimmune diseases33–35. In addition, neoantigens
can be generated through glycosylation and phosphory­
lation36. Interestingly, post-translational vicinal disulfide
bond creation between adjacent cysteine residues was
found to be required for the TCR-mediated recognition
of an insulin A‑chain peptide in T1D, although this modi­
fication does not alter the binding of the peptide to HLA37.
Correlating processes of PTM to disease-­associated anti­
genic diversity is still a nascent research area. In the con­
text of autoimmune disease, an assumption is made by
researchers that PTM occurs in the periphery but not
in the thymus during central tolerance establishment;
however, this concept requires formal testing.
The impact of deamidation has been revealed through
studies of T cell recognition in coeliac disease, where the
disease-associated HLA‑DQ2.5 and HLA‑DQ8 molecules
present gluten-derived peptides (gliadins) that have been
deamidated by trans­glutaminase 2 (REF. 38). Gliadins are
bound to HLA‑DQ2.5 with a high kinetic stability that
leads to their sustained presentation, which is a require­
ment for the generation of the pathogenic T cell response.
By contrast, HLA‑DQ2.2 confers a low risk of coeliac dis­
ease, and its geometry is such that gliadins cannot be sta­
bly accommodated in the peptide-­binding cleft 39,40. The
disease-associated HLA‑DQ8 lacks an aspartic acid resi­
due at position 57β, which creates a positively charged P9
pocket with a preference for nega­tively charged peptides41.
The autoimmune disease rele­vance of the HLA‑DQ8
polymorphic residue 57β was first identified for T1D42.
Furthermore, autoantigenic glutamic acid decarboxy­
lase 65 (GAD65) peptides bind with high affinity to the
NOD mouse MHC class II I-Ag7 molecule, which also
lacks this aspartic acid43. In the context of coeliac dis­
ease, the absence of this amino acid may better accom­
modate deamidated gliadins and may also promote the
recruitment of cross-reactive TCRs that carry a negative
signature charge in CDR3β, which can respond to both
the modified and unmodified gluten peptides41 (TABLE 1).
A shared feature of different coeliac disease-relevant
peptide–HLA‑DQ complexes is their bias towards inter­
acting with certain TCRs; an arginine residue within the
CDR3 loop of the TCRs studied is thought to anchor
the interaction with HLA‑DQ2.5–deamidated gliadin
and HLA‑DQ8–deamidated gliadin complexes44,45. These
HLA–deamidated gliadin–TCR interactions may be at
the centre of a pathological T cell–B cell cooperation34.
The transglutaminase 2 enzyme that mediates gliadin
deamidation also serves as a target for auto­antibodies
in coeliac disease46. Thus, B cells with the capacity to
generate such autoantibodies would initially bind trans­
glutaminase 2–gliadin complexes via their B cell recep­
tors (BCRs), thereby mediating BCR-dependent gliadin
internalization and presentation by HLA‑DQ2.5. CD4+
T cells that become activated by the interaction of their
TCR with HLA‑DQ2.5–gliadin would then provide
help to the B cells, leading to the secretion of the auto­
antibodies that can bind transglutaminase 2, affinity
matu­ration, further gliadin uptake and presentation, and
an ensuing boost of the T cell response34.
A similar mechanism may also occur in rheumatoid
arthritis. CD4+ T cells that respond to citrullinated pep­
tides presented by disease-associated HLA‑DR mole­
cules that are present on citrullinated-protein-specific
B cells can then provide help to these B cells; the B cells
can then undergo class switching to secrete antibodies
specific for anti-citrullinated protein47 and may in turn
further stimulate the T cells. Notably, citrullination facil­
itates the binding of autoantigens to the rheumatoid-
arthritis-­associated HLA‑DR4 protein. In addition, when
the rheumatoid arthritis-associated protein vimentin
becomes citrullinated, it also modifies protease-mediated
cleavage so that self peptides derived from vimentin tend
to be retained rather than degraded35 (FIG. 1 g; TABLE 1).
Hybrid peptides. Another mechanism for neo­antigen
generation is through the creation of hybrid self
peptides in peripheral tissues (FIG. 1 h). Proteasomal

6 | ADVANCE ONLINE PUBLICATION www.nature.com/nri

©
2
0
1
7
M
a
c
m
i
l
l
a
n
P
u
b
l
i
s
h
e
r
s
L
i
m
i
t
e
d
,
p
a
r
t
o
f
S
p
r
i
n
g
e
r
N
a
t
u
r
e
.
A
l
l
r
i
g
h
t
s
r
e
s
e
r
v
e
d
.
 REVIEWS

Pancreatic islets
Clusters of different cell types
found throughout the pancreas
that include the
insulin-producing β-cells, which
are a main target of the
autoimmune response
occurring in patients with
type 1 diabetes.
degradation of intracellular proteins to produce the
peptides typically presented by HLA class I mole­
cules can promote the generation of spliced epitopes.
Although this was once thought to be a rare occurrence,
it has now been demonstrated that up to one-third of
the HLA class I peptidome comprises spliced peptides,
thus indicating a greater diversity of presented self anti­
gens or pathogen-derived antigens than previously
appreciated48. Consequently, it is not surprising that
viral immune evasion mechanisms include resistance
to proteasomal degradation49. In the context of auto­
immunity, N-terminal additions to a chromogranin A
peptide, an autoantigen in the NOD mouse model pre­
sented on class II MHC molecules, could substantially
improve T cell activation. This finding raises the pos­
sibility that hybrid peptides presented by MHC class II
molecules could drive autoreactivity 50. A more recent
study in T1D has now shown that HLA class II proteins
can indeed present spliced antigens that are generated
by covalent crosslinking of peptides51. Pathogenic CD4+
T cell clones isolated from NOD mice were found to
recognize proinsulin hybrid peptides present in the
secretory granules of pancreatic β-cells. Furthermore,
CD4+ T cells with an analogous specificity were also
isolated from the residual pancreatic islets of two organ
donors with T1D. Therefore, T cell targeting of hybrid
insulin peptides specifically present in the pancreas
could provide an explanation for the breakdown of
immune tolerance in T1D51 (TABLE 1).
Regulation of HLA expression. In addition to the spe­
cific presentation of antigens that trigger the stimulation
of autoreactive T cells, the level of expression of HLA
molecules at the cell surface, and thus the density of
relevant antigens, has been found to be associated with
autoimmune diseases and other conditions (TABLE 1). For
example, the distal intergenic XL9 element in the MHC
locus has been found to alter the binding of interferon
regulatory factor 4 (IRF4) and transcriptional regulator
CTCF that modulates the transcription of HLA‑DRB1,
HLA‑DQA1 and HLA‑DQB1, resulting in a 2.5‑fold
increase in the expression of HLA‑DR and HLA‑DQ
proteins that predispose to systemic lupus erythema­
tosus 52. Interestingly, genetically determined high
expression levels of HLA‑C increase the risk of Crohn’s
disease but also promote the control of infections such
as HIV. Consequently, the opposing effects of differential
HLA expression may exert a selection pressure in favour
of retaining allelic lineages with an HLA expression pro­
file that is associated with autoimmune disease but that
protects against infection53. In a more extreme scenario,
genetic defects that lead to partial or complete HLA
class I or class II deficiency can cause bare lympho­cyte
syndrome, which is typically characterized by extreme
susceptibility to bacterial, protozoan and viral infec­
tions54. Moreover, many pathogen immune evasion strat­
egies include the downregulation of pathogen-­derived
peptide presentation by inhibiting HLA expression.
For example, EBV can inhibit HLA class I synthesis
through host shut-off 55 and encodes a repressor of MHC
class II transactivator (CIITA), thereby blocking HLA
class II transcription56. However, the interplay between
EBV immune evasion mechanisms and viral latency and
the link between EBV and multiple sclerosis remain to
be determined.
The relevance of HLA expression levels in transplan­
tation has also been considered fairly recently. HLA‑C
expression levels can influence permissivity to donor
mismatch and hence mortality after transplantation
and graft-versus-host disease (GVHD)57. Furthermore,
HLA‑DPB1‑mismatched transplants from donors
with genetically determined low HLA‑DPB1 expres­
sion can lead to a high risk of GVHD if the recipients
have high expression levels of HLA‑DPB1 (REF. 58).
Notably, genetic variants that reduce the expression of
HLA‑G, a nonclassical HLA molecule found on fetal
extravillous trophoblast cells that invade the maternal
decidua and generate the maternal–fetal interface, are
associated with pregnancy-related disorders such as
pre-eclampsia59,60.
HLA stability. Given the involvement of HLA expres­
sion levels in disease, post-translational regulation of
HLA stability has also been implicated in aberrant T cell
reactivity. The interdependence of HLA class I levels
and the peptide loading and binding process have long
been recognized: the folding and assembly of class I
molecules in the endoplasmic reticulum (ER) lumen is
dependent on the multiprotein peptide-­loading com­
plex, and suboptimal assembly can promote ER reten­
tion or instability at the cell surface61. Tapasin, which
forms part of the peptide-loading complex, stabilizes the
peptide-free conformation of class I molecules, thereby
increasing the chance of occupancy by a high-affinity
peptide62,63. Notably, HLA molecules that differ by only
one amino acid residue in the peptide-­binding groove,
such as the HLA‑B*44:02 and HLA‑B*44:05 mole­
cules, can differ strongly in their tapasin dependence,
thus altering their trafficking to the cell surface63–65.
As HLA‑B*44:05 has a tyrosine rather than an aspar­
tic acid at residue 116, it can constitutively acquire
peptides in a manner that is not dependent on the pep­
tide-loading complex and is thus less susceptible to viral
interference66.
Compared with HLA‑A and HLA‑B, HLA‑C proteins
are present at the cell surface at a tenfold lower level67,
but these too can be differentially expressed depending
on their peptide-binding capacity. HLA‑C*05 is found
more abundantly at the cell surface than HLA‑C*07, as
its peptide-binding cleft has a flatter structure, which
is more permissive to the binding of a larger range of
peptides that consequently help to stabilize the HLA
molecule68 (FIG. 1i). HLA-derived peptides themselves can
have an impact on HLA molecule stability. For example,
polymorphic HLA‑B leader peptide sequences can vary
in their capacity to bind HLA‑E and stabilize its expres­
sion levels at the cell surface, which in turn can skew NK
cell education towards that mediated by the binding of
HLA‑E to CD94/NKG2 receptors69.
Differential stability of HLA class II proteins has also
been reported and may have implications in disease
development. Antibody-dependent quantification of

NATURE REVIEWS | IMMUNOLOGY ADVANCE ONLINE PUBLICATION | 7

©
2
0
1
7
M
a
c
m
i
l
l
a
n
P
u
b
l
i
s
h
e
r
s
L
i
m
i
t
e
d
,
p
a
r
t
o
f
S
p
r
i
n
g
e
r
N
a
t
u
r
e
.
A
l
l
r
i
g
h
t
s
r
e
s
e
r
v
e
d
.
REVIEWS

Allomorphs
Different MHC protein forms
encoded by different alleles.
HLA‑DQ allomorphs on human primary B cells indicated
relatively little variation in the expression level of dif­
ferent allomorphs70. However, a study using a cell-line-
based MHC expression system suggests that HLA‑DQ
allomorphs have a differential intrinsic stability. Some
variation in the cell-surface density of HLA‑DQ mole­
cules may correlate with risk or resistance to auto­
immune disorders 71. For example, T1D‑associated
HLA‑DQ proteins have been found to be unstable,
whereas T1D‑protective allomorphs are stable. Amino
acid variation at HLA‑DQ residue 47α, which is located
on the outside of the peptide-binding groove, has been
pinpointed as a key regulator of stability, potentially
through a differential dependency on the MHC II pep­
tide-loading regulator HLA-DM72. In addition, evolu­
tionary analysis suggests that this resi­due was the target
of positive diversifying selection71. HLA‑DQ instability
may confer risk of autoimmune diseases by allowing
the escape of autoreactive T cells from thymic negative
selection but also by facilitating the formation of weak
peptide–HLA complexes in the periphery with a wide
range of self peptides. The rela­tive instability of certain
HLA class II proteins could explain why the same HLA
alleles can be associated with either risk or resistance
to several different autoimmune diseases but cannot
explain all cross-disease association patterns. For exam­
ple, although the HLA‑DRB1*15:01−HLA‑DQB1*06:02
haplotype, which encodes both the relatively less sta­
ble HLA‑DRB1*15:01 protein and the intrinsically
stable HLA‑DQB1*06:02 protein, is protective against
T1D, autoimmune polyglandular syndrome and
immuno­globulin A (IgA) deficiency 73–75, this haplotype
confers risk of multiple sclerosis and narcolepsy 30,76. The
increased risk of multiple sclerosis can be associated
with the less stable gene product of HLA‑DRB1*15:01;
however, this is not the case for narcolepsy, in which
HLA‑DQ6‑mediated autoimmunity arises despite the
intrinsic high stability of the allomorph (TABLE 1).
Collectively, these different mechanisms of HLA–
peptide–TCR binding may promote autoimmunity
by altering the capacity for autoreactive effector T cell
activation and escape from the regulatory processes
that establish and maintain tolerance (FIG. 2). The sto­
chasticity of TCR generation probably contributes to the
chance of an individual carrying autoreactive TCRs with
an unconventional binding mode. However, auto­reactive
T cells exist not only in patients with auto­immune dis­
ease but also in healthy individuals with HLA risk allo­
morphs; thus, the presence of autoreactive TCRs alone
is unlikely to be sufficient for disease development.
Instead, this may further require numerous other genetic
and environmental influences that modulate the avail­
ability, processing and presentation of relevant antigens
as well as T cell activation thresholds. A fundamental
challenge for our understanding of how HLA–peptide–
TCR interactions promote disease is not only to deter­
mine how these different factors come together to break
self tolerance but also to elucidate the relative breadth
and importance of perturbed molecular mechanisms of
HLA–peptide–TCR binding in individual patients and
across patient and control cohorts.
Technological advances and challenges
The identification of the numerous molecular mecha­
nisms by which HLA molecules are implicated in dis­
ease development and resistance represents the outcome
of several key technological advances in silico, in vitro
(BOX 1) and in vivo (BOX 2) that are enabling key questions
regarding HLA biology to be addressed.
Cross-comparing HLA genetic associations. Beyond
identifying HLA genetic associations with individual
diseases, the capacity to compare association patterns
across autoimmune diseases can provide hints regard­
ing shared and distinct pathophysiological pathways
that may have implications for understanding disease
aetiology and therapy. In the future, such compari­
sons will be increasingly enabled by the availability
of large resources, such as the UK Biobank (http://
www.uk­­biobank.ac.uk/), which allow for the simulta­
neous analysis of matched genetic and clinical infor­
mation from different types of patients and controls77.
Intriguing patterns include the observation that the
HLA class I risk alleles HLA‑B*27 (associated with
ankylosing spondylitis), HLA‑C*06:02 (associated
with psoriasis), HLA‑B*51 (associated with Behçet
Figure 2 | HLA–peptide–T cell receptor interactions can ▶
promote autoimmunity. a | HLA class I allomorphs
associated with autoimmune disease can promote the
activation of autoreactive effector CD8+ T cells, the T cell
receptors (TCRs) of which are generated stochastically in
the thymus. Autoreactive HLA–peptide–TCR interactions
may be affected by HLA gene expression levels, altered
peptide processing and the antigen affinity of
autoreactive TCRs, ultimately leading to increased
TCR-mediated activation of the autoreactive T cells. The
threshold for T cell activation may be further modulated
by non-HLA risk genes and environmental influences that
collectively help lead to the development of a pathogenic
autoimmune response. b | HLA class II allomorphs
associated with autoimmune disease can promote the
activation of autoreactive effector CD4+ T cells, the TCRs
of which are generated stochastically in the thymus.
Autoreactive HLA–peptide–TCR interactions may be
affected by HLA gene expression levels and differential
HLA protein stability, neoantigen creation (through
protein post-translational modification and hybrid
peptide generation), ‘hotspot’ molecular mimicry,
alternate TCR docking, TCR-mediated stabilization of
weak peptide–HLA complexes and an altered TCR binding
register. The downstream consequence of these different
mechanisms is the increased TCR-mediated activation of
autoreactive effector T cells (and reduced generation or
function of regulatory T cells in some instances). The
threshold for T cell activation may be further modulated
by non-HLA risk genes and environmental influences that
collectively help lead to the development of a pathogenic
autoimmune response. α, MHC class II alpha chain;
β, MHC class II beta chain; β2m, β2-microglobulin;
APC, antigen-presenting cell; CLIP, class-II‑associated
invariant chain peptide; ER, endoplasmic reticulum; ERAD,
ER-associated protein degradation; ERAP, ER
aminopeptidase; MIIC, MHC-class-II‑containing
compartment; miRNA, microRNA; TAP, transporter
associated with antigen processing.

8 | ADVANCE ONLINE PUBLICATION www.nature.com/nri

©
2
0
1
7
M
a
c
m
i
l
l
a
n
P
u
b
l
i
s
h
e
r
s
L
i
m
i
t
e
d
,
p
a
r
t
o
f
S
p
r
i
n
g
e
r
N
a
t
u
r
e
.
A
l
l
r
i
g
h
t
s
r
e
s
e
r
v
e
d
.
 REVIEWS

Unfolded protein response
A cellular stress response
triggered because of the
accumulation of unfolded or
misfolded proteins within the
endoplasmic reticulum.
disease) and HLA‑A*29 (associated with birdshot
chorio­retinopathy) all display epistatic interactions
with ER aminopeptidase (ERAP) gene polymorphisms,
suggesting a key role for altered peptide processing 78.
However, it remains to be elucidated whether these
interactions confer disease risk by leading to aber­
rant peptide presentation, changes in HLA class I
homodimerization and expression levels at the cell
surface or an unfolded protein response79 (FIG. 2).
For several of the conditions in which HLA class II
associations are the strongest determinants of disease
risk, interactions between different HLA class II alleles
have been identified, as is the case for multiple sclero­
sis30, rheumatoid arthritis, T1D and coeliac disease80.

a b
Nucleus
Transcription Altered HLA
expression

ER
Altered HLA gene
expression → impact Translation
of miRNAs +
HLA-DM
β α

Golgi

TAP Proteasome • Self proteins

• Viral proteins CLIP

HLA
class I
ERAP Early
ERAD MIIC endosome
Altered peptide
ER processing:
Golgi • Aberrant peptide Hybrid
presentation? peptides Protease
• Altered HLA
homodimerization? ‘Hotspot’
• Unfolded protein molecular Differential
response? mimicry HLA stability
HLA
APC class II APC
β2m
HLA class I

HLA–peptide–TCR Antigen
interaction TCR HLA–peptide– Post-translationally
TCR interaction modified protein
Low-affinity- CD3
mediated Microbial proteins
thymic escape containing cross-
reactive antigens
Low affinity for
LCK autoantigen LCK
enables thymic Alternate docking
escape
Higher affinity for
cross-reactive ↑ Cross-reactivity TCR stabilization ↑ Escape from
microbial peptide with microbial of weak peptide– thymic
peptide HLA complexes selection

Altered register
Differential generation
of autoreactive effector
versus regulatory T cells

Modulation by Increased TCR-mediated

non-HLA risk genes activation of autoreactive Modulation by
and environmental effector cells Increased TCR-mediated non-HLA genes
influences activation of autoreactive and environmental
effector cells influences

Autoimmunity Autoimmunity
CD8+ T cell CD4+ T cell

Nature Reviews | Immunology

NATURE REVIEWS | IMMUNOLOGY ADVANCE ONLINE PUBLICATION | 9
©
2
0
1
7
M
a
c
m
i
l
l
a
n
P
u
b
l
i
s
h
e
r
s
L
i
m
i
t
e
d
,
p
a
r
t
o
f
S
p
r
i
n
g
e
r
N
a
t
u
r
e
.
A
l
l
r
i
g
h
t
s
r
e
s
e
r
v
e
d
.
REVIEWS

Box 1 | Studying HLA biology: the immunologist’s in silico and in vitro toolbox
Genetics and genomics
Mapping the relationship of HLA alleles and haplotypes with single-nucleotide
polymorphisms (SNPs) in the MHC123 has revealed that, despite the high level of HLA
heterogeneity, the extensive linkage disequilibrium in the region allows for genetic
variation to be captured through the use of tag SNPs, such that HLA alleles and
haplotypes can be imputed124. This has facilitated the interrogation of large case–control
cohorts, leading to the accelerated discovery of HLA associations with even modest
effect sizes across many different diseases. However, attributing associations to specific
HLA alleles or haplotypes may not be as biologically informative as the capacity to
pinpoint the exact polymorphisms implicated in disease risk. Identifying the precise
coding or non-coding variants involved — by using improved analytical methods125 or
MHC deep sequencing126 — may provide further insight into peptide-binding groove
amino acid substitutions that could alter HLA–antigen interactions and into the
differential regulation of HLA gene expression.
Tetramers
Peptide–HLA tetramers bind T cell receptors (TCRs) with a superior avidity compared
with peptide–HLA monomers (which exhibit a high rate of dissociation) and thus can
sensitively detect specific T cells present at a frequency as low as 0.02%. Moreover, even
lower frequency T cells can be detected when tetramer binding is coupled to magnetic
enrichment techniques. The study of low-affinity peptide and HLA–peptide–TCR
interactions, such as those relevant in certain autoimmune diseases, still presents technical
problems127. This is particularly the case for HLA class-II‑restricted T cells, as the CD4
co-receptor does not facilitate the HLA–peptide–TCR interaction, leading to the
underestimation of specific self-reactive T cell populations128. However, methods for
boosting HLA multimer staining are enabling the detection of T cells for which the TCR
affinity for peptide–HLA complexes has a dissociation constant greater than 1 mM (REF. 94).
For coeliac disease, this may be due to HLA‑DQA1*05:01
and HLA‑DQB1*02:01 genotypes that are carried on dif­
ferent haplotypes enabling the formation of HLA‑DQ2.5
heterodimers in trans81, whereas for multiple sclerosis,
the protective effect of the HLA‑DQA1*01:01 haplotype
in carriers of the disease-associated HLA‑DRB1*15:01
genotype30 may be related to TCR cross reactivity, as pro­
posed for another multiple-sclerosis-related interaction
— HLA‑DRB1*15:01–HLA‑DRB5 (REF. 82). Intriguingly,
this latter HLA-DRB5 haplotype is also associated with
Parkinson disease. Although Parkinson disease is con­
sidered a primary neurodegenerative disorder, a recent
study has demonstrated that both HLA‑DRB1*15:01
and HLA‑DRB5*01:01 can present peptides from
α‑synuclein, a protein aggregated in the brain of patients,
and the antigenic epitopes displayed can drive T cell
responses83. This suggests that the condition could have
an autoimmune component despite the absence of a
prominent immune cell infiltrate.
In the context of infectious disease, the coordi­
nated study of host and viral genetic variation in dif­
ferent ethnic groups is revealing the capacity of HLA
micropolymorphisms to diversify the repertoire of
pathogen-specific cytotoxic lymphocytes, as exem­
Epitope spreading plified by infection with hepatitis B virus84. A recent
Describes how a self-directed
genome-wide association study of host and hepatitis
immune response induced by a
single peptide (or epitope) C virus genomes indicated that different HLA alleles
could spread to include other also exert selective pressures on the hepatitis C virus
peptides (or epitopes), not only genome, presumably via the impact of the adaptive
on the same autoantigen immune system, thereby driving in vivo viral evo­
(intramolecular spreading) but
also on other self molecules in
lution85. Thus, evaluating the relationship between
close vicinity to the target cell HLA, T cell and virus genetic variation is important
(intermolecular spreading). for designing effective vaccination and treatment
strategies86,87. In cancer, the study of HLA genetics has
not only focused on disease risk but has also found that
certain tumours display recurrent somatic mutations
in HLA genes88, including mutations that promote a
loss of HLA allele heterozygosity 89; consequently, such
tumours are more likely to undergo immune eva­
sion as there is less-efficient presentation of tumour
neoantigens. This has important thera­p eutic impli­
cations — particularly as a recent analysis has demon­
strated that tumour HLA class  I genotype can be
used to predict which oncogenic mutations are more
likely to arise on the basis of the peptide sequences that
are less effectively presented by the patient HLA class I
molecules90.
Interrogating the antigen-specific T cell repertoire.
Although HLA allele associations may suggest how
specific HLA allomorphs contribute to disease suscep­
tibility and protection, more definitive evidence of their
molecular mode of action requires the identification
of the rele­vant peptides and receptors that bind them.
Exogenous and endogenous antigens presented by HLA
molecules can vary according to tissue site, cellular route
of uptake and presentation, temporally owing to muta­
tion and modi­fication, and as an organism’s environ­
mental exposures change, which all present technical
challenges. However, advances in proteomic and other
mass-spectrometry-based approaches are set to facili­
tate the capturing of antigenic diversity, as evidenced by
the identification of novel hybrid peptides48. Crucially,
these methods should be coupled to the investigation
of the diseased tissue itself rather than peripheral blood
alone, which is easily accessible but may be less inform­
ative. There is a fundamental need to examine peptide
and TCR repertoires from the site of immunopathology,
where the disease-relevant epitopes and cells are local­
ized. Moreover, even within a single individual, epitope
spreading can either contribute to ongoing pathology or
even have a protective effect91. In chronic conditions (for
example, coeliac disease), the range of epitopes driving
pathological responses may change by epitope focus­
ing, which is when T cell clones that are directed against
immunodominant epitopes outcompete other clones92.
Thus, antigenic diversity in the context of polymorphic
HLA molecules can promote the evolution of the T cell
repertoire, from the early events that occur in the thymus
during the establishment of central tolerance through to
the events that occur over the course of a lifetime in the
periphery.
The observations above indicate the value of moni­
toring and tracking specific T cells during the course
of disease or upon vaccination or immunotherapy. The
development of HLA tetramers (BOX 1) and higher-­order
multimers has been transformative for the study of T cell
responses, as it has facilitated the ex vivo identification
and quantification of antigen-specific T cells as well,
as the mapping of their epitopes93,94. Ultimately, the
combination of fluorescently labelled tetramer staining
with the interrogation of markers of cellular phenotype
and function has yielded important insights into the
dynamics of immune responsiveness in a wide range of

10 | ADVANCE ONLINE PUBLICATION www.nature.com/nri

©
2
0
1
7
M
a
c
m
i
l
l
a
n
P
u
b
l
i
s
h
e
r
s
L
i
m
i
t
e
d
,
p
a
r
t
o
f
S
p
r
i
n
g
e
r
N
a
t
u
r
e
.
A
l
l
r
i
g
h
t
s
r
e
s
e
r
v
e
d
.
 REVIEWS

pathophysiological settings95, and the experimental use
of peptide–HLA tetramers is increasingly being cou­
pled to newer mass cytometry 96, TCR sequencing and
computational technologies97.
Exploiting the rules of HLA–peptide–T cell receptor
engagement. Advances in deep-sequencing technolo­
gies, particularly at the single-cell level98, are enabling the
detailed characterization of TCR sequence hetero­geneity,
and this has spurred a concomitant effort to ascribe a
meaningful order to this variation. A recent study inte­
grating TCR sequences, TCR antigenic specificity (deter­
mined by peptide–HLA tetramer panel screening) and
structural data has shown that it is possible to group
TCRs with shared specificities, both within and between
individuals, on the basis of conserved motifs and CDR3
sequence similarity 97. The number and size of these
TCR groups may reflect immune response complexity,
and clustering across individuals may suggest sharing of
specif­icities for important antigens. This indicates that
TCR sequence analysis may be of value clinically for
monitoring responses to infection or vaccination and
that it might facilitate antigen discovery.
To investigate autoimmunity-prone T cell repertoires,
a study has explored the relative structural properties
of TCRs and found that the presence of hydro­phobic
residues at positions 6 and 7 of the β‑chain CDR3 loop
promotes reactivity to self antigens99. An investi­gation
of the specificity of peptide–MHC recognition by
TCRs in infection, through the use of TCR selection of
peptide–MHC libraries in conjunction with deep
sequencing, confirmed that for any given TCR inter­
action, there is a greater tolerance for variation of pep­
tide residues that are distal to the TCR interface. This
was true to the extent that specific TCR ligands can be
identified computationally 100.
Not all interactions between TCRs and peptide–HLA
molecules are confined to broad docking geometry con­
straints. A notable example of a large-scale change in
peptide–HLA binding geometry that is not easily pre­
dictable is that of the HLA‑A*02:01‑restricted melanoma
DMF4 TCR; this TCR engages two homologous antigens
Box 2 | Studying HLA biology: the immunologist’s in vivo toolbox
Humanized mouse models
The creation of humanized mouse models of disease, in which human genes or cells
are transferred to mice to enable the functional dissection of specific HLA molecules,
has been instrumental to studying the biological consequences of disease-relevant
antigen-specific responses in vivo129. The use of such models has facilitated the
pinpointing of disease risk alleles that cannot be distinguished as pathophysiological
drivers by statistical means alone owing to high linkage disequilibrium. Interactions
between a multiple sclerosis HLA risk allele and disease-modifying alleles have been
investigated in humanized mice and indicate the potential for antigenic-specific
immunomodulation82, as does the investigation of a Goodpasture-disease-protective
HLA allele25. Studies in nonobese diabetic HLA‑A*02:01‑expressing mice have
identified CD8+ T cell populations that can also be found in patients with type 1
diabetes carrying HLA‑A*02:01, further suggesting that mechanistic insights relevant
to human disease can be gained from the careful interrogation of humanized mice130.
Developments in genetic engineering such as the use of CRISPR–Cas9 coupled to
improved microscopy techniques such as intravital imaging may further facilitate the
study of antigen-specific immune responses in vivo.
via distinct geometries that differ by a 15° rotation of
the whole TCR101. Other atypical MHC–peptide–TCR
interactions include reports of cross-class recognition
(for example, when MHC class II molecules bind CD8+
TCRs)102 and the discovery of a reversed polarity of TCR
recognition whereby TCR binding occurs with a 180°
rotation relative to the conventional binding mode103,104.
These examples challenge conventional views on the
rules of TCR engagement; nevertheless, our growing
understanding of the typical and atypical molecular
mechanisms of TCR recognition, in conjunction with
the application of next-generation sequencing methods,
may come to expedite the in silico identification of HLA–
peptide–TCR binding partners and their correlation with
cellular and clinical phenotypes through time.
Harnessing HLA function clinically
The progress in characterizing HLA diversity, HLA asso­
ciations with human disease and HLA–peptide–TCR
interactions and their mechanistic implications has
galvanized one of the most fundamental goals of this
area of research: to harness the improved understanding
of HLA function for clinical benefit (TABLE 2). With the
current aim of achieving precision or even personalized
medicine, the capacity to stratify patients on the basis of
their HLA genotypes and to administer antigen-specific,
patient-tailored disease prevention or immunotherapy
presents an especially attractive target.
Noncellular disease prevention strategies and thera­
pies: vaccination. Perhaps the most successful use of
antigen-­specific approaches for promoting human
health is through vaccination to prevent the spread of
infectious diseases. Despite substantial achievements,
combating infections such as HIV, tuberculosis, hepa­
titis C and malaria is an ongoing challenge. However, for
malaria, some progress has been reported based on the
findings of a phase III trial of the RTS,S/AS01 candidate
malaria vaccine105. Emerging infections, such as Ebola,
Zika and Dengue viruses, highlight the need for rapid
vaccine generation and trialling strategies that rely on
the efficient identification of appropriate viral epitopes
to safely but effectively induce immunogenicity 106,107.
Even currently licensed vaccines have limitations
that need to be overcome, such as the failure of recip­
ients to develop or maintain protective immunity and
the occurrence of adverse events. For example, the per­
tussis resurgence observed in children and adolescents
who had previously received the acellular vaccine is at
least partially related to waning immunity 108. For sev­
eral viral infections, including measles, mumps, rubella,
influenza and hepatitis B virus, the induction and sus­
tainment of a humoral immune response following
vaccination correlates with specific HLA class I and II
alleles that are determinants of nonresponsiveness or
hyper-­responsiveness; this partly depends on the type
of T helper cells generated by specific HLA alleles109.
Antigenic changes in influenza virus require the
yearly reformulation and administration of vac­
cines. The 2009–2010 influenza A(H1N1)pdm09
(Pandemrix; GlaxoSmithKline) vaccination campaign

NATURE REVIEWS | IMMUNOLOGY ADVANCE ONLINE PUBLICATION | 11

©
2
0
1
7
M
a
c
m
i
l
l
a
n
P
u
b
l
i
s
h
e
r
s
L
i
m
i
t
e
d
,
p
a
r
t
o
f
S
p
r
i
n
g
e
r
N
a
t
u
r
e
.
A
l
l
r
i
g
h
t
s
r
e
s
e
r
v
e
d
.
REVIEWS

Table 2 | Antigen-specific disease prevention and therapeutic strategies, desired mode of action and challenges
Preventative or therapeutic strategy
Noncellular
Vaccination against infection and cancer
Peptide immunotherapy for allergy and
autoimmune disease
Peptide–HLA-coated nanoparticles for
autoimmune disease
Peptide–HLA-bearing exosomes for
cancer
Engineered TCRs; for example, bispecific
ImmTACs with enhanced affinity and
T cell response redirection capacity for
cancer
Cellular
DC adoptive transfer for cancer
immunotherapy
Regulatory T cell adoptive transfer for
autoimmune disease and transplantation
T cell and/or total mononuclear
cell adoptive transfer for cancer
immunotherapy
DC, dendritic cell; ImmTACs, immune mobilizing monoclonal TCRs against cancer molecules; TCR, T cell receptor.
Mode of action
Boosting effector immune
response
Promoting peripheral
tolerance and/or pathogenic
immune response dampening
Promoting peripheral
tolerance
Boosting effector immune
response
Promoting effector immune
response
Boosting effector immune
response
Promoting peripheral
tolerance
Promoting effector immune
response
Development and implementation challenges
• Failure of development or maintenance of immune response
• Emerging infections needing rapid antigen identification and trialling
• Unexpected side effects and differential efficacy
• Peptide and adjuvant selection
• Peptide selection
• Generation of desired type and magnitude of immune response
• Time of treatment relative to disease course
• Peptide selection
• Generation of desired type and magnitude of immune response
• Time of treatment relative to disease course
• Trialling in humans required for safety and efficacy
• Peptide selection
• Generation of desired type and magnitude of immune response
• Target selection and validation
• Temporal changes in target expression
• Unwanted effects due to target expression on nonmalignant cells
• Off-target effects
• Heterogeneity of DC molecular composition
• Immunosuppression of DCs after transfer
• Poor storage capacity and handling difficulty
• Maintenance of functional stability
• Purity and storage
• Potency after transfer in pathophysiological milieu
• Tumour antigen loss
• Tumour immune evasion
• Peptide and adjuvant selection
• Immunosuppression of cells after transfer

saw an unanticipated rise in the incidence of narcolepsy,
particularly in individuals carrying the predisposing
HLA‑DQB1*06:02 allele110. Intriguingly, there is also
some evidence that Pandemrix administration inter­
fered with the potential therapeutic benefit of the pep­
tide immunotherapy GAD-alum (GAD formulated with
aluminium hydroxide; discussed below) in a phase III
trial in individuals with new-onset T1D111. Collectively,
these findings emphasize the importance of understand­
ing the factors that regulate vaccine immunogenicity,
including the recipient’s HLA genetic profile and prior
immunogenic exposures.
Noncellular disease prevention strategies and thera-
pies: peptide immunotherapy. Despite the elucidation
of many molecular mechanisms of HLA–peptide–TCR
binding in the context of autoimmune disease, anti­
gen-specific immunotherapy for these conditions has
yet to become standard clinical practice. A key challenge
is determining whether peptide immunotherapy should
aim either to delete effector T cells (to skew their patho­
genic phenotype towards a more anti-­inflammatory pro­
file) or to promote Treg cell function (in order to prevent
disease progression due to continued organ damage).
This in turn depends on the nature of the HLA–peptide–
TCR interactions that drive disease in any given indi­
vidual and whether they are the outcome of insufficient
thymic negative selection, the availability of peripheral
neoantigens or inadequate Treg cell function.
Nevertheless, peptide immunotherapy with dis­
ease-relevant autoantigenic epitopes has been con­
sidered as a potential option to treat autoimmune
disease112. A recent small placebo-controlled trial in
patients with newly diagnosed T1D has demonstrated
that pro­insulin peptide immunotherapy is safe, does
not expedite β-cell dysfunction and is correlated with
markers of immunomodulation 113. The efficacy of
antigen-­specific immunotherapy depends on adequate
HLA-mediated antigen presentation to T cells. Boosting
of HLA–peptide–TCR interactions may be mediated
by the use of adjuvants and small molecules that act as
cell-surface HLA-loading enhancers, which have also
been tested in the context of cancer immunotherapy 114.
Notably, for GAD-alum administration to patients with
T1D, there is some initial evidence that antigen presenta­
tion may be further facilitated by direct intralymphatic
injection of the therapeutic115.
Noncellular disease prevention strategies and therapies:
peptide–HLA delivery. A possible therapeutic strategy
that bypasses the need for antigen uptake or cell-surface
loading that is associated with peptide immuno­therapies
is the use of peptide–HLA-coated nanoparticles. These
can be tailored with respect to the peptide–HLA

12 | ADVANCE ONLINE PUBLICATION www.nature.com/nri

©
2
0
1
7
M
a
c
m
i
l
l
a
n
P
u
b
l
i
s
h
e
r
s
L
i
m
i
t
e
d
,
p
a
r
t
o
f
S
p
r
i
n
g
e
r
N
a
t
u
r
e
.
A
l
l
r
i
g
h
t
s
r
e
s
e
r
v
e
d
.
 REVIEWS

molecules used and have been shown to drive the are not maintained; thus, the repertoire of neoantigen-­
expansion of antigen-specific regulatory responses in specific T cells must change through time in parallel
humanized mice116, although there are as yet no reports with newly occurring neoantigens121. This suggests the
of testing and efficacy in humans. need for strategies that broaden neoantigen-­specific
A similar approach is being trialled in malignancies T cell responses, especially as autologous patient T cells
including metastatic melanoma, advanced non-small- recognize only a fraction of predicted melanoma neo­
cell lung cancer and advanced colorectal cancer using antigens. To address this problem, a recent study
dendritic cell (DC)-derived exosomes, which are nano­ describes a system for identifying a larger number of
metre-sized membrane vesicles that express functional neoantigen-specific TCRs with the use of the naive
peptide–HLA complexes at their surface and which can T cell repertoire of healthy blood donors as a source of
induce T and NK cell-mediated immune responses in diverse TCRs restricted to the same HLA‑A molecule
patients117. Exosome-based immunotherapy is antigen-­ as the patient. The TCRs were then introduced into
specific and has distinct advantages over DC‑based can­ re­directed T cells by genetic engineering, and these cells
cer therapies, which are complicated by the difficulties ofwere capable of recognizing and killing patient-derived
culturing and storing DCs and high costs. melanoma cells expressing the relevant neoantigens122.
This approach therefore demonstrates the poten­
Noncellular disease prevention strategies and therapies: tial power of immunotherapeutic strategies in which
engineered T cell receptors. Other noncellular, antigen-­ HLA, antigen and TCR specificity are concomitantly
specific cancer immunotherapies include engineered considered and exploited for clinical benefit.
TCR-dependent strategies, such as the use of immune
mobilizing monoclonal TCRs against cancer (ImmTAC) Concluding remarks
molecules118. Native cancer-specific TCRs often have a After decades of studying HLA biology and its role in
weak affinity as a result of thymic selection, given that human disease, progress in genetic and molecular profil­
cancer-specific antigens are typically derived from self ing technologies, computational methods and structural
proteins; thus, boosting the activity of existing can­ analyses is broadening the scope of how the interactions
cer-specific T cells may be of limited benefit. However, between HLA molecules, the antigens they present and
ImmTACs are based on affinity-enhanced monoclonal the receptors that bind them can be utilized in a clini­
TCRs that bind their cognate peptide–HLA complexes cal setting. The many mechanisms identified to date by
with high, picomolar affinity. In addition to the specific which HLA–peptide–TCR interactions can promote
TCR-mediated binding of ImmTACs to tumour cells, disease represent individual case studies. Thus, it is still
these molecules are also fused to an anti‑CD3 anti­ unknown whether certain mechanisms are more com­
body fragment that then potentiates a redirected T cell monly utilized in particular diseases or in patient sub­
response and subsequent tumour cell killing 118. groups carrying the same HLA risk alleles or whether
multiple mechanisms may be implicated in individual
Cell-based therapies. DC‑based strategies to promote patients. Despite progress in identifying the specific
antigen-specific anticancer immune responses or toler­ HLA genes and alleles implicated in disease risk and
ogenesis are fraught with practical and economic chal­ resistance through large case–control cohort analyses, a
lenges, whereas T cell-based therapeutic approaches more high-throughput determination of the molecular
have shown more substantial promise. In the context of and cellular basis of these associations with respect to
autoimmune disease and transplantation, pilot studies the exact antigens and TCRs involved is a major goal
and early trials of the adoptive transfer of autologous, for the future. However, studying and monitoring HLA–
polyclonal Treg cells have provided support for the safety peptide–TCR interactions and their downstream impact
and possible capacity of these cells to induce tolerance in patient cohorts are becoming increasingly feasible
in vivo119,120. and will be required to meet the added challenges pre­
However, the greatest progress in the use of T cell- sented by the changing antigenic landscape within and
based therapies is in the cancer arena. An important across individuals. At the same time, ongoing research
finding has been the observation that successful treat­ efforts are already culminating in the use of antigen-­
ment of patients with melanoma with adoptive T cell specific disease prevention and therapeutic strategies,
therapy is driven by T cell reactivity against muta­ the design and implementation of which will probably
tion-derived neoantigens. As the neoantigens are lost be further aided by advances in genetic engineering and
owing to the impact of the T cells, the T cells themselves nanotechnology.

1. Amiel, J. in Study of the Leukocyte Phenotypes in
Hodgkin’s Disease in Histocompatibility Testing (ed.
Teraski, P. I.) 79–81 (Munksgaard, 1967).
2. Trowsdale, J. & Knight, J. C. Major
histocompatibility complex genomics and human
disease. Ann. Rev. Genom. Hum. Genet. 14,
301–323 (2013).
3. Godfrey, D. I., Uldrich, A. P., McCluskey, J.,
Rossjohn, J. & Moody, D. B. The burgeoning family of
unconventional T cells. Nat. Immunol. 16, 1114–1123
(2015).
4. Robinson, J., Soormally, A. R., Hayhurst, J. D. &
Marsh, S. G. E. The IPD-IMGT/HLA database — new
developments in reporting HLA variation.
Hum. Immunol. 77, 233–237 (2016).
5. Parham, P. & Ohta, T. Population biology of antigen
presentation by MHC class I molecules. Science 272,
67–74 (1996).
6. Carrington, M. et al. HLA and HIV‑1:
heterozygote advantage and B*35‑Cw*04
disadvantage. Science 283, 1748–1752
(1999).
7. Parham, P. & Moffett, A. Variable NK cell
receptors and their MHC class I ligands in
immunity, reproduction and human evolution.
Nat. Rev. Immunol. 13, 133–144 (2013).
8. Sewell, A. K. Why must T cells be cross-reactive?
Nat. Rev. Immunol. 12, 668–677 (2012).
9. Quigley, M. F. et al. Convergent recombination shapes
the clonotypic landscape of the naive T‑cell repertoire.
Proc. Natl Acad. Sci. USA 107, 19414–19419
(2010).
10. Saunders, P. M. et al. A bird’s eye view of NK cell
receptor interactions with their MHC class I
ligands. Immunol. Rev. 267, 148–166
(2015).
11. Hudson, L. E. & Allen, R. L. Leukocyte Ig‑Like
Receptors — a model for MHC class I disease
associations. Front. Immunol. 7, 281 (2016).

NATURE REVIEWS | IMMUNOLOGY ADVANCE ONLINE PUBLICATION | 13

©
2
0
1
7
M
a
c
m
i
l
l
a
n
P
u
b
l
i
s
h
e
r
s
L
i
m
i
t
e
d
,
p
a
r
t
o
f
S
p
r
i
n
g
e
r
N
a
t
u
r
e
.
A
l
l
r
i
g
h
t
s
r
e
s
e
r
v
e
d
.
REVIEWS

12. Fadda, L. et al. Peptide antagonism as a mechanism
for NK cell activation. Proc. Natl Acad. Sci. USA 107,
10160–10165 (2010).
13. Sim, M. J. et al. Canonical and cross-reactive binding
of NK cell inhibitory receptors to HLA‑C allotypes is
dictated by peptides bound to HLA‑C. Front. Immunol.
8, 193 (2017).
14. Miles, J. J., McCluskey, J., Rossjohn, J. & Gras, S.
Understanding the complexity and malleability of T‑cell
recognition. Immunol. Cell Biol. 93, 433–441 (2015).
15. Madsen, L. S. et al. A humanized model for multiple
sclerosis using HLA‑DR2 and a human T‑cell receptor.
Nat. Genet. 23, 343–347 (1999).
16. 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. Nat. Immunol. 6,
490–496 (2005).
17. Bulek, A. M. et al. Structural basis for the killing of
human beta cells by CD8+ T cells in type 1 diabetes.
Nat. Immunol. 13, 283–289 (2012).
18. Pugliese, A. et al. The insulin gene is transcribed in the
human thymus and transcription levels correlate with
allelic variation at the INS VNTR‑IDDM2 susceptibility
locus for type 1 diabetes. Nat. Genet. 15, 293–297
(1997).
19. Cole, D. K. et al. Hotspot autoimmune T cell receptor
binding underlies pathogen and insulin peptide cross-
reactivity. J. Clin. Invest. 126, 2191–2204 (2016).
20. Yin, Y., Li, Y., Kerzic, M. C., Martin, R. &
Mariuzza, R. A. Structure of a TCR with high affinity for
self-antigen reveals basis for escape from negative
selection. EMBO J. 30, 1137–1148 (2011).
21. Quandt, J. A. et al. Unique clinical and pathological
features in HLA‑DRB1*0401‑restricted MBP
111‑129‑specific humanized TCR transgenic mice.
J. Exp. Med. 200, 223–234 (2004).
22. Armstrong, K. M., Piepenbrink, K. H. & Baker, B. M.
Conformational changes and flexibility in T‑cell
receptor recognition of peptide-MHC complexes.
Biochem. J. 415, 183–196 (2008).
23. Stadinski, B. D. et al. Diabetogenic T cells recognize
insulin bound to IAg7 in an unexpected, weakly binding
register. Proc. Natl Acad. Sci. USA 107, 10978–
10983 (2010).
24. Yang, J. et al. Autoreactive T cells specific for insulin
B:11‑23 recognize a low-affinity peptide register in
human subjects with autoimmune diabetes. Proc. Natl
Acad. Sci. USA 111, 14840–14845 (2014).
25. Ooi, J. D. et al. Dominant protection from HLA-linked
autoimmunity by antigen-specific regulatory T cells.
Nature 545, 243–247 (2017).
This study reveals a mechanistic basis for the
dominantly protective effect of HLA in autoimmune
disease via an effect on self-epitope-specific
Treg cells.
26. Harkiolaki, M. et al. T cell-mediated autoimmune
disease due to low-affinity crossreactivity to common
microbial peptides. Immunity 30, 348–357 (2009).
27. Sethi, D. K. et al. A highly tilted binding mode by a
self-reactive T cell receptor results in altered
engagement of peptide and MHC. J. Exp. Med. 208,
91–102 (2011).
28. Wucherpfennig, K. W. & Strominger, J. L. Molecular
mimicry in T cell-mediated autoimmunity: viral
peptides activate human T cell clones specific for
myelin basic protein. Cell 80, 695–705 (1995).
29. Belbasis, L., Bellou, V., Evangelou, E.,
Ioannidis, J. P. A. & Tzoulaki, L. Environmental risk
factors and multiple sclerosis: an umbrella review of
systematic reviews and meta-analyses. Lancet Neurol.
14, 263–273 (2015).
30. Moutsianas, L. et al. Class II HLA interactions
modulate genetic risk for multiple sclerosis. Nat.
Genet. 47, 1107–1113 (2015).
31. Nielsen, T. R. et al. Effects of infectious mononucleosis
and HLA‑DRB1*15 in multiple sclerosis. Mult. Scler. J.
15, 431–436 (2009).
32. Lang, H. L. E. et al. A functional and structural basis
for TCR cross-reactivity in multiple sclerosis. Nat.
Immunol. 3, 940–943 (2002).
33. Babon, J. A. et al. Analysis of self-antigen specificity of
islet-infiltrating T cells from human donors with type 1
diabetes. Nat. Med. 22, 1482–1487 (2016).
34. Koning, F., Thomas, R., Rossjohn, J. & Toes, R. E.
Coeliac disease and rheumatoid arthritis: similar
mechanisms, different antigens. Nat. Rev. Rheumatol.
11, 450–461 (2015).
35. Scally, S. W. et al. A molecular basis for the association
of the HLA‑DRB1 locus, citrullination, and rheumatoid
arthritis. J. Exp. Med. 210, 2569–2582 (2013).
36. Mohammed, F. et al. Phosphorylation-dependent
interaction between antigenic peptides and MHC
class I: a molecular basis for the presentation of
transformed self. Nat. Immunol. 9, 1236–1243
(2008).
37. Mannering, S. I. et al. The insulin A‑chain epitope
recognized by human T cells is posttranslationally
modified. J. Exp. Med. 202, 1191–1197 (2005).
38. Molberg, O. et al. Tissue transglutaminase selectively
modifies gliadin peptides that are recognized by gut-
derived T cells in celiac disease. Nat. Med. 4,
713–717 (1998).
39. Bodd, M., Kim, C. Y., Lundin, K. E. & Sollid, L. M. T‑Cell
response to gluten in patients with HLA‑DQ2.2 reveals
requirement of peptide-MHC stability in celiac disease.
Gastroenterology 142, 552–561 (2012).
40. Fallang, L. E. et al. Differences in the risk of celiac
disease associated with HLA‑DQ2.5 or HLA‑DQ2.2
are related to sustained gluten antigen presentation.
Nat. Immunol. 10, 1096–1101 (2009).
41. Hovhannisyan, Z. et al. The role of HLA‑DQ8 β57
polymorphism in the anti-gluten T‑cell response in
coeliac disease. Nature 456, 534–538 (2008).
42. 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).
43. Corper, A. L. et al. A structural framework for
deciphering the link between I-Ag7 and autoimmune
diabetes. Science 288, 505–511 (2000).
44. Broughton, S. E. et al. Biased T cell receptor usage
directed against human leukocyte antigen
DQ8‑restricted gliadin peptides is associated with
celiac disease. Immunity 37, 611–621 (2012).
45. Petersen, J. et al. T‑Cell receptor recognition of
HLA‑DQ2‑gliadin complexes associated with celiac
disease. Nat. Struct. Mol. Biol. 21, 480–488
(2014).
46. Dieterich, W. et al. Identification of tissue
transglutaminase as the autoantigen of celiac disease.
Nat. Med. 3, 797–801 (1997).
47. Verpoort, K. N. et al. Isotype distribution of anti-cyclic
citrullinated peptide antibodies in undifferentiated
arthritis and rheumatoid arthritis reflects an ongoing
immune response. Arthritis Rheum. 54, 3799–3808
(2006).
48. Liepe, J. et al. A large fraction of HLA class I ligands
are proteasome-generated spliced peptides.
Science 354, 354–358 (2016).
This study demonstrates that the proteasome-­
generated hybrid peptide pool accounts for as
much as one-third of the HLA class I
immunopeptidome, with implications for disease
mechanisms and antigen-specific therapies.
49. Hansen, T. H. & Bouvier, M. MHC class I antigen
presentation: learning from viral evasion strategies.
Nat. Rev. Immunol. 9, 503–513 (2009).
50. Jin, N. et al. N‑Terminal additions to the WE14
peptide of chromogranin A create strong autoantigen
agonists in type 1 diabetes. Proc. Natl Acad. Sci. USA
112, 13318–13323 (2015).
51. Delong, T. et al. Pathogenic CD4 T cells in type 1
diabetes recognize epitopes formed by peptide fusion.
Science 351, 711–714 (2016).
This study demonstrates that hybrid insulin
peptides can be found in pancreatic islet β-cells in
T1D and are antigenic for CD4+ T cells.
52. Raj, P. et al. Regulatory polymorphisms modulate the
expression of HLA class II molecules and promote
autoimmunity. eLife 5, e12089 (2016).
53. O’Huigin, C. et al. The molecular origin and
consequences of escape from miRNA regulation by
HLA‑C alleles. Am. J. Hum. Genet. 89, 424–431
(2011).
54. Shrestha, D., Szollosi, J. & Jenei, A. Bare lymphocyte
syndrome: an opportunity to discover our immune
system. Immunol. Lett. 141, 147–157 (2012).
55. Rowe, M. et al. Host shutoff during productive
Epstein–Barr virus infection is mediated by BGLF5
and may contribute to immune evasion. Proc. Natl
Acad. Sci. USA 104, 3366–3371 (2007).
56. Balan, N., Osborn, K. & Sinclair, A. J. Repression of
CIITA by the Epstein–Barr virus transcription factor
Zta is independent of its dimerization and DNA
binding. J. Gen. Virol. 97, 725–732 (2016).
57. Petersdorf, E. W. et al. HLA‑C expression levels define
permissible mismatches in hematopoietic cell
transplantation. Blood 124, 3996–4003 (2014).
58. Petersdorf, E. W. et al. High HLA‑DP expression and
graft-versus-host disease. N. Engl. J. Med. 373,
599–609 (2015).
59. Ferreira, L. M. R., Meissner, T. B., Tilburgs, T. &
Strominger, J. L. HLA‑G: at the interface of maternal-
fetal tolerance. Trends Immunol. 38, 272–286
(2017).
60. Quach, K., Grover, S. A., Kenigsberg, S. &
Librach, C. L. A combination of single nucleotide
polymorphisms in the 3’ untranslated region of HLA‑G
is associated with preeclampsia. Hum. Immunol. 75,
1163–1170 (2014).
61. Ljunggren, H. G. et al. Empty MHC class‑I molecules
come out in the cold. Nature 346, 476–480
(1990).
62. Wearsch, P. A. & Cresswell, P. Selective loading of
high-affinity peptides onto major histocompatibility
complex class I molecules by the tapasin‑ERp57
heterodimer. Nat. Immunol. 8, 873–881 (2007).
63. Williams, A. P., Peh, C. A., Purcell, A. W., McCluskey, J.
& Elliott, T. Optimization of the MHC class I peptide
cargo is dependent on tapasin. Immunity 16, 509–
520 (2002).
64. Rizvi, S. M. et al. Distinct assembly profiles of HLA‑B
molecules. J. Immunol. 192, 4967–4976 (2014).
65. Thammavongsa, V., Raghuraman, G., Filzen, T. M.,
Collins, K. L. & Raghavan, M. HLA‑B44
polymorphisms at position 116 of the heavy chain
influence TAP complex minding via an effect on
peptide occupancy. J. Immunol. 177, 3150–3161
(2006).
66. Zernich, D. et al. Natural HLA class I polymorphism
controls the pathway of antigen presentation and
susceptibility to viral evasion. J. Exp. Med. 200,
13–24 (2004).
67. Apps, R. et al. Relative expression levels of the HLA
class‑I proteins in normal and HIV-infected cells.
J. Immunol. 194, 3594–3600 (2015).
68. Kaur, G. et al. Structural and regulatory diversity
shape HLA‑C protein expression levels. Nat. Commun.
8, 15924 (2017).
69. Horowitz, A. et al. Class I HLA haplotypes form two
schools that educate NK cells in different ways. Sci.
Immunol. 1, eaag1672 (2016).
70. Duggal, P. et al. Genome-wide association study of
spontaneous resolution of hepatitis C virus infection:
data from multiple cohorts. Ann. Intern. Med. 158,
235–245 (2013).
71. Miyadera, H., Ohashi, J., Lernmark, Å., Kitamura, T. &
Tokunaga, K. Cell-surface MHC density profiling
reveals instability of autoimmunity-associated HLA.
J. Clin. Invest. 125, 275–291 (2015).
72. Zhou, Z. & Jensen, P. E. Structural characteristics of
HLA‑DQ that may impact DM editing and
susceptibility to type‑1 diabetes. Front. Immunol. 4,
262 (2013).
73. Ferreira, R. C. et al. High-density SNP mapping of the
HLA region identifies multiple independent
susceptibility loci associated with selective IgA
deficiency. PLOS Genet. 8, e1002476 (2012).
74. Weinstock, C. et al. Autoimmune polyglandular
syndrome type 2 shows the same HLA class II pattern
as type 1 diabetes. Tissue Antigens 77, 317–324
(2011).
75. Hu, X. L. et al. Additive and interaction effects at three
amino acid positions in HLA‑DQ and HLA‑DR
molecules drive type 1 diabetes risk. Nat. Genet. 47,
898–905 (2015).
76. 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).
77. Cortes, A. et al. Bayesian analysis of genetic
association across tree-structured routine healthcare
data in the UK Biobank. Nat. Genet. 49, 1311–1318
(2017).
78. McGonagle, D., Aydin, S. Z., Gul, A., Mahr, A. &
Direskeneli, H. ‘MHC‑I‑opathy’-unified concept for
spondyloarthritis and Behcet disease. Nat. Rev.
Rheumatol. 11, 731–740 (2015).
79. Brown, M. A., Kenna, T. & Wordsworth, B. P. Genetics
of ankylosing spondylitis — insights into pathogenesis.
Nat. Rev. Rheumatol. 12, 81–91 (2016).
80. Lenz, T. L. et al. Widespread non-additive and
interaction effects within HLA loci modulate the risk of
autoimmune diseases. Nat. Genet. 47, 1085–1090
(2015).
81. Vader, W. et al. The HLA‑DQ2 gene dose effect in
celiac disease is directly related to the magnitude and
breadth of gluten-specific T cell responses. Proc. Natl
Acad. Sci. USA 100, 12390–12395 (2003).
82. Gregersen, J. W. et al. Functional epistasis on a
common MHC haplotype associated with multiple
sclerosis. Nature 443, 574–577 (2006).

14 | ADVANCE ONLINE PUBLICATION www.nature.com/nri

©
2
0
1
7
M
a
c
m
i
l
l
a
n
P
u
b
l
i
s
h
e
r
s
L
i
m
i
t
e
d
,
p
a
r
t
o
f
S
p
r
i
n
g
e
r
N
a
t
u
r
e
.
A
l
l
r
i
g
h
t
s
r
e
s
e
r
v
e
d
.
 REVIEWS

83. Sulzer, D. et al. T cells from patients with Parkinson’s
disease recognize α-synuclein peptides. Nature 546,
656–661 (2017).
This study demonstrates that the HLA associations
with Parkinson disease may be related to the
presence of α‑synuclein-specific T cells, suggesting
a relevance of the immune system in this
neurodegenerative condition.
84. Tan, A. T. et al. Host ethnicity and virus genotype
shape the hepatitis B virus-specific T‑cell repertoire.
J. Virol. 82, 10986–10997 (2008).
85. Ansari, M. A. et al. Genome-to‑genome analysis
highlights the effect of the human innate and adaptive
immune systems on the hepatitis C virus. Nat. Genet,
49, 666–673 (2017).
86. Swadling, L. et al. A human vaccine strategy based on
chimpanzee adenoviral and MVA vectors that primes,
boosts, and sustains functional HCV-specific T cell
memory. Sci. Transl. Med. 6, 261ra153 (2014).
87. Park, S. H. et al. Subinfectious hepatitis C virus
exposures suppress T cell responses against
subsequent acute infection. Nat. Med. 19,
1638–1642 (2013).
88. Shukla, S. A. et al. Comprehensive analysis of cancer-
associated somatic mutations in class I HLA genes.
Nat. Biotechnol. 33, 1152–1158 (2015).
89. McGranahan, N. et al. Allele-specific HLA loss and
immune escape in lung cancer evolution. Cell 171,
1259–1271.e11 (2017).
90. Marty, R. et al. MHC‑1 genotype restricts the
oncogenic mutational landscape. Cell 171,
1272–1283.e15 (2017).
91. Vanderlugt, C. L. & Miller, S. D. Epitope spreading in
immune-mediated diseases: implications for
immunotherapy. Nat. Rev. Immunol. 2, 85–95 (2002).
92. Vader, W. et al. The gluten response in children with
celiac disease is directed toward multiple gliadin and
glutenin peptides. Gastroenterology 122,
1729–1737 (2002).
93. Altman, J. D. et al. Phenotypic analysis of antigen-
specific T lymphocytes. Science 274, 94–96 (1996).
94. Dolton, G. et al. Comparison of peptide-major
histocompatibility complex tetramers and dextramers
for the identification of antigen-specific T cells. Clin.
Exp. Immunol. 177, 47–63 (2014).
95. Nepom, G. T. MHC class II tetramers. J. Immunol.
188, 2477–2482 (2012).
96. Newell, E. W. et al. Combinatorial tetramer staining
and mass cytometry analysis facilitate T‑cell epitope
mapping and characterization. Nat. Biotechnol. 31,
623–629 (2013).
97. Glanville, J. et al. Identifying specificity groups in the
T cell receptor repertoire. Nature 547, 94–98 (2017).
This study describes the development of an
algorithm for analysing large numbers of TCR
sequences and defining TCR specificity groups
across TCRs and individuals, which may facilitate
analyses of T cell responses and ligand
identification.
98. Stubbington, M. J. et al. T cell fate and clonality
inference from single-cell transcriptomes.
Nat. Methods 13, 329–332 (2016).
99. Stadinski, B. D. et al. Hydrophobic CDR3 residues
promote the development of self-reactive T cells.
Nat. Immunol. 17, 946–955 (2016).
100. Birnbaum, M. E. et al. Deconstructing the peptide-
MHC specificity of T cell recognition. Cell 157,
1073–1087 (2014).
101. Borbulevych, O. Y., Santhanagopolan, S. M.,
Hossain, M. & Baker, B. M. TCRs used in cancer gene
therapy cross-react with MART‑1/Melan‑A tumor
antigens via distinct mechanisms. J. Immunol. 187,
2453–2463 (2011).
102. Hansen, S. G. et al. Cytomegalovirus vectors violate
CD8+ T cell epitope recognition paradigms.
Science 340, 1237874 (2013).
103. Beringer, D. X. et al. T cell receptor reversed
polarity recognition of a self-antigen major
histocompatibility complex. Nat. Immunol. 16,
1153–1161 (2015).
104. Gras, S. et al. Reversed T cell receptor docking on a
major histocompatibility class I complex limits
involvement in the immune response. Immunity 45,
749–760 (2016).
105. RTS,S Clinical Trials Partnership. Efficacy and safety of
RTS,S/AS01 malaria vaccine with or without a booster
dose in infants and children in Africa: final results of a
phase 3, individually randomised, controlled trial.
Lancet 386, 31–45 (2015).
106. Henao-Restrepo, A. M. et al. Efficacy and effectiveness
of an rVSV-vectored vaccine expressing Ebola surface
glycoprotein: interim results from the Guinea ring
vaccination cluster-randomised trial. Lancet 386,
857–866 (2015).
107. Hadinegoro, S. R. et al. Efficacy and long-term safety
of a Dengue vaccine in regions of endemic disease.
N. Engl. J. Med. 373, 1195–1206 (2015).
108. Plotkin, S. A. The pertussis problem. Clin. Infect. Dis.
58, 830–833 (2014).
109. Ovsyannikova, I. G., Dhiman, N., Jacobson, R. M. &
Poland, G. A. Human leukocyte antigen
polymorphisms: variable humoral immune responses
to viral vaccines. Expert Rev. Vaccines 5, 33–43
(2006).
110. Bomfim, I. L. et al. The immunogenetics of narcolepsy
associated with A(H1N1)pdm09 vaccination
(Pandemrix) supports a potent gene-environment
interaction. Genes Immun. 18, 75–81 (2017).
111. Tavira, B. et al. Effect of simultaneous vaccination with
H1N1 and GAD-alum on GAD65‑induced immune
response. Diabetologia 60, 1276–1283 (2017).
112. Larche, M. & Wraith, D. C. Peptide-based therapeutic
vaccines for allergic and autoimmune diseases.
Nat. Med. 11, S69–S76 (2005).
113. Alhadj Ali, M. et al. Metabolic and immune effects of
immunotherapy with proinsulin peptide in human new-
onset type 1 diabetes. Sci. Transl. Med. 9, eaaf7779
(2017).
114. Afridi, S., Hoessli, D. C. & Hameed, M. W.
Mechanistic understanding and significance of small
peptides interaction with MHC class II molecules for
therapeutic applications. Immunol. Rev. 272,
151–168 (2016).
115. Ludvigsson, J., Wahlberg, J. & Casas, R.
Intralymphatic injection of autoantigen in type 1
diabetes. N. Engl. J. Med. 376, 697–699 (2017).
116. Clemente-Casares, X. et al. Expanding antigen-specific
regulatory networks to treat autoimmunity. Nature
530, 434–440 (2016).
This study demonstrates that systemic delivery of
nanoparticles coated with autoimmune-disease-­
relevant peptides bound to MHC class II molecules
can trigger the generation and expansion of antige
n-specific regulatory CD4+ T cells, with potential
therapeutic implications.
117. Pitt, J. M. et al. Dendritic cell-derived exosomes for
cancer therapy. J. Clin. Invest. 126, 1224–1232
(2016).
118. Liddy, N. et al. Monoclonal TCR-redirected tumor cell
killing. Nat. Med. 18, 980–987 (2012).
119. Bluestone, J. A. et al. Type 1 diabetes immunotherapy
using polyclonal regulatory T cells. Sci. Transl. Med. 7,
315ra189 (2015).
120. Chandran, S. et al. Polyclonal regulatory T cell therapy
for control of inflammation in kidney transplants. Am.
J. Transplant. 17, 2945–2954 (2017).
121. Verdegaal, E. M. E. et al. Neoantigen landscape
dynamics during human melanoma‑T cell interactions.
Nature 536, 91–95 (2016).
122. Stronen, E. et al. Targeting of cancer neoantigens with
donor-derived T cell receptor repertoires. Science
352, 1337–1341 (2016).
This study demonstrates that a higher frequency of
tumour mutations is more immunogenic than initially
estimated and that healthy donors can be a valuable
source of T cells that are reactive to these tumour
neoantigens, thereby indicating a new, personalized
approach for cancer immunotherapy.
123. de Bakker, P. I. W. et al. A high-resolution HLA and
SNP haplotype map for disease association studies in
the extended human MHC. Nat. Genet. 38,
1166–1172 (2006).
124. Dilthey, A., Cox, C., Iqbal, Z., Nelson, M. R. &
McVean, G. Improved genome inference in the MHC
using a population reference graph. Nat. Genet. 47,
682–688 (2015).
125. Raychaudhuri, S. et al. Five amino acids in three HLA
proteins explain most of the association between MHC
and seropositive rheumatoid arthritis. Nat. Genet. 44,
291–296 (2012).
126. Zhou, F. S. et al. Deep sequencing of the MHC region
in the Chinese population contributes to studies of
complex disease. Nat. Genet. 48, 740–746 (2016).
127. Dolton, G. et al. More tricks with tetramers: a practical
guide to staining T cells with peptide-MHC multimers.
Immunology 146, 11–22 (2015).
128. Sabatino, J. J. Jr, Huang, J., Zhu, C. & Evavold, B. D.
High prevalence of low affinity peptide-MHC II
tetramer-negative effectors during polyclonal CD4+
T cell responses. J. Exp. Med. 208, 81–90 (2011).
129. Wekerle, H., Flugel, A., Fugger, L., Schett, G. &
Serreze, D. Autoimmunity’s next top models.
Nat. Med. 18, 66–70 (2012).
130. Yu, D. et al. Roquin represses autoimmunity by
limiting inducible T‑cell co‑stimulator messenger RNA.
Nature 450, 299–303 (2007).
Acknowledgements
C.A.D. is supported by the Wellcome Centre and the Royal
Society. J.R. is supported by grants from the National Health
and Medical Research Council (Australia), the Cancer Council
of Victoria, the Australian Research Council (ARC) and
Worldwide Cancer Research and is an ARC Laureate Fellow.
L.F. is supported by the Wellcome Centre, the Medical
Research Council, the Danish National Research Foundation,
Takeda and the Oak Foundation.
Author contributions
C.A.D. and J.R. researched data for the article; L.F., C.A.D.,
J.R. and L.F. had a substantial input into the discussion of
content; C.A.D., J.R and L.F. wrote the article; and L.F. and
C.A.D. carried out the review and editing.
Competing interests statement
The authors declare no competing interests.
Publisher’s note
Springer Nature remains neutral with regard to jurisdictional
claims in published maps and institutional affiliations.
Reviewer information
Nature Reviews Immunology thanks L. Klareskog, and two
anonymous reviewers, for their contribution to the peer
review of this work.
FURTHER INFORMATION
UK Biobank: http://www.ukbiobank.ac.uk/
ALL LINKS ARE ACTIVE IN THE ONLINE PDF

NATURE REVIEWS | IMMUNOLOGY ADVANCE ONLINE PUBLICATION | 15

©
2
0
1
7
M
a
c
m
i
l
l
a
n
P
u
b
l
i
s
h
e
r
s
L
i
m
i
t
e
d
,
p
a
r
t
o
f
S
p
r
i
n
g
e
r
N
a
t
u
r
e
.
A
l
l
r
i
g
h
t
s
r
e
s
e
r
v
e
d
.