ArticlePDF AvailableLiterature Review

Genetics of ankylosing spondylitis

Authors:
Molecular
Immunology
57 (2014) 2–
11
Contents
lists
available
at
ScienceDirect
Molecular
Immunology
j
ourna
l
ho
me
p
age
:
www.elsevier.com/locate/molimm
Review
Genetics
of
ankylosing
spondylitis
Philip
C.
Robinson,
Matthew
A.
Brown
University
of
Queensland
Diamantina
Institute,
Translational
Research
Institute,
37
Kent
Road,
Princess
Alexandra
Hospital,
Brisbane,
Australia
a
r
t
i
c
l
e
i
n
f
o
Article
history:
Received
12
June
2013
Received
in
revised
form
18
June
2013
Accepted
19
June
2013
Available online 31 July 2013
Keywords:
Ankylosing
spondylitis
Spondyloarthritis
Single
nucleotide
polymorphism
association
a
b
s
t
r
a
c
t
Ankylosing
spondylitis
(AS)
is
a
chronic
inflammatory
arthritis
that
affects
the
spine
and
sacroiliac
joints.
It
causes
significant
disability
and
is
associated
with
a
number
of
other
features
including
peripheral
arthritis,
anterior
uveitis,
psoriasis
and
inflammatory
bowel
disease
(IBD).
Significant
progress
has
been
made
in
the
genetics
of
AS
have
in
the
last
five
years,
leading
to
new
treatments
in
trial,
and
major
leaps
in
understanding
of
the
aetiopathogenesis
of
the
disease.
© 2013 Elsevier Ltd. All rights reserved.
1.
Introduction
Ankylosing
spondylitis
(AS)
is
a
chronic
inflammatory
arthritis
that
affects
the
spine
and
sacroiliac
joints.
It
causes
significant
dis-
ability
and
is
associated
with
a
number
of
other
features
including
peripheral
arthritis,
anterior
uveitis,
psoriasis
and
inflammatory
bowel
disease
(IBD).
Significant
progress
has
been
made
in
the
genetics
of
AS
have
in
the
last
five
years,
leading
to
new
treatments
in
trial,
and
major
leaps
in
understanding
of
the
aetiopathogenesis
of
the
disease.
2.
Major
Histocompatibility
Complex
Genetics
of
AS
It
has
long
been
known
that
AS
runs
strongly
in
families,
with
the
risk
of
disease
in
first-degree
relatives
of
AS
cases
being
>52
times
that
of
unrelated
subjects
(Brown
et
al.,
2000).
Whether
this
cofamiliality
was
due
to
shared
environmental
or
genetic
factors
was
unclear
until
the
demonstration
in
the
early
1970s
of
asso-
ciation
of
the
HLA-B27
allele
with
the
disease
(Brewerton
et
al.,
1973b;
Schlosstein
et
al.,
1973).
Heterozygote
HLA-B27
carriage
has
an
odds
ratio
of
50
for
AS,
and
homozygosity
with
an
odds
ratio
of
100
(International
Genetics
of
Ankylosing
Spondylitis
et
al.,
2013;
Jaakkola
et
al.,
2006;
Khan
et
al.,
1978).
The
recurrence
risk
for
AS
in
monozygotic
twins
is
63%,
first
degree
relatives
8.2%
and
second
This
article
belongs
to
Special
Issue
on
The
Pathogenetic
Role
of
HLA-B27
and
other
Genes
in
Ankylosing
Spondylitis.
Corresponding
author
at:
University
of
Queensland
Diamantina
Institute,
Level
7,
37
Kent
Road,
Princess
Alexandra
Hospital,
Woolloongabba,
Brisbane,
Queensland
4102,
Australia.
Tel.:
+61
7
3442
7017.
E-mail
address:
matt.brown@uq.edu.au
(M.A.
Brown).
degree
relatives
1.0%.
The
parent–child
recurrence
risk
is
7.9%
and
the
sibling–sibling
recurrence
risk
is
8.2%
(Brown
et
al.,
2000).
The
HLA-B27
association
in
AS
remains
amongst
the
strongest
genetic
association
with
any
common
human
disease,
but
the
molecular
mechanism
underlying
this
association
remains
unclear.
In
nearly
all
populations
studied
worldwide,
HLA-B27
is
strongly
associated
with
AS.
One
hundred
and
thirty
subtypes
of
HLA-
B27
have
now
been
reported
(European
Bioinformatics
Database
Immuno
Polymorphism
Database,
2013),
and
AS
has
been
reported
to
occur
with
the
following
subtypes:
B*2702
(MacLean
et
al.,
1993),
*2703
(Reveille
et
al.,
2000),
*2704
(Lopez-Larrea
et
al.,
1995),
*2705
(MacLean
et
al.,
1993),
*2706
(Gonzalez-Roces
et
al.,
1997),
*2707
(Armas
et
al.,
1999),
*2708
(Armas
et
al.,
1999),
*2710
(Garcia
et
al.,
1998),
*2714
(Garcia-Fernandez
et
al.,
2001),
*2715
(Garcia-
Fernandez
et
al.,
2001),
and
*2719
(Djouadi
et
al.,
2001).
The
vast
majority
of
HLA-B27
subtypes
occur
in
too
few
individuals
to
defini-
tively
establish
their
association
with
disease.
Of
those
studied
in
sufficient
numbers
of
carriers,
HLA-B*2702-5,
*2707,
*2708
and
*2710
clearly
significantly
increase
AS
risk.
There
is
some
evidence
suggesting
that
HLA-B*2704
may
carry
higher
risk
than
the
ances-
tral
HLA-B*2705
allele,
and
that
the
risk
associated
with
B*2703
may
be
lower.
Two
subtypes,
B*2706
and
B*2709,
are
not
associ-
ated
with
disease,
but
AS
has
been
reported
in
carriers
of
each
allele,
indicating
that
they
are
not
protective
for
AS.
It
is
beyond
the
scope
of
this
article
to
discuss
the
potential
mechanisms
by
which
HLA-
B27
induces
AS,
but
as
we
and
others
have
proposed
previously,
any
hypothesis
as
to
explain
this
must
also
explain
how
these
two
subtypes
are
not
AS-associated
(Brown,
2009;
McLean
et
al.,
1985).
It
has
also
long
been
suspected
that
other,
non-HLA-B27,
MHC
alleles
are
involved
in
AS-pathogenesis.
Whilst
many
studies
have
attempted
to
study
this
further,
the
difficulty
of
distinguishing
link-
age
disequilibrium
with
HLA-B27
from
direct
associations
across
0161-5890/$
see
front
matter ©
2013 Elsevier Ltd. All rights reserved.
http://dx.doi.org/10.1016/j.molimm.2013.06.013
P.C.
Robinson,
M.A.
Brown
/
Molecular
Immunology
57 (2014) 2–
11 3
the
MHC
has
until
recent
large
scale
studies
precluded
convincing
demonstration
of
any
replicated
non-HLA-B
association
with
AS.
An
exception
to
this
has
been
the
association
of
HLA-B60
with
AS,
first
reported
by
Robinson
et
al.
in
1986
(Robinson
et
al.,
1989)
and
con-
firmed
subsequently
in
populations
of
both
white
European
(Brown
et
al.,
1996)
and
east
Asian
ancestry
(Wei
et
al.,
2004).
The
International
Genetics
of
AS
Consortium
has
recently
extended
this
observation,
using
the
findings
of
10,619
AS
cases
and
15,145
controls
densely
genotyped
across
the
MHC
to
genetically
dissect
the
region
further
(International
Genetics
of
Ankylosing
Spondylitis
et
al.,
2013).
A
major
finding
from
this
study
was
the
identification
of
a
single
tagSNP,
rs116488202,
that
is
highly
sensi-
tive
and
specific
for
HLA-B27
(>98.5%,
within
the
limit
of
accuracy
of
direct
genotyping
of
HLA-B27),
and
which
can
be
genotyped
for
a
small
fraction
of
the
cost
of
typing
HLA-B27
itself.
This
should
have
a
major
impact
on
genetic
screening
for
AS
either
in
high
risk
or
even
in
the
general
population.
An
early
application
is
that
it
has
enabled
imputation
of
HLA-B27
in
large
cohorts,
and
thus
the
ability
to
study
non-HLA-B27
MHC
associations
of
AS
in
sufficiently
powered
studies.
These
studies
have
shown
genomewide
signif-
icant
association
of
HLA-A*0201
with
AS
with
odds
ratios
of
1.21
and
1.36
in
HLA-B27
positive
and
negative
cases
respectively.
It
is
likely
that
the
further
associations
of
MHC
genes
will
be
identified
in
future
studies,
particularly
as
better
imputation
methods
for
HLA
loci
become
available.
3.
Non-MHC
genetic
associations
Strong
epidemiological
evidence
of
the
existence
of
significant
non-MHC
genetic
associations
of
AS
was
presented
well
before
any
such
genes
were
convincingly
identified.
HLA-B27
positive
first-
degree
relatives
of
AS
cases
are
5.6–16
times
more
likely
to
develop
disease
themselves
than
HLA-B27
positive
carriers
in
the
general
community
(Calin
et
al.,
1983;
van
der
Linden
et
al.,
1983).
Identi-
cal
twins
are
much
more
likely
to
be
concordant
for
AS
(60–75%)
than
HLA-B27-positive
dizygotic
twins
(24%)
(Brown
et
al.,
1997;
Pedersen
et
al.,
2008).
The
GWAS
era
has
enabled
rapid
progress
in
identifying
non-
MHC
associations
of
AS
to
be
made.
These
findings
have
highlighted
a
number
of
important
pathways
in
AS
pathogenesis
including
the
IL-23
pathway,
aminopeptidases
and
peptide
presentation,
innate
immune
stimulation
and
the
interaction
and
homeostasis
of
resi-
dent
microbial
communities.
Included
among
them
are
pathways
for
which
we
currently
have
therapeutics
available,
particularly
in
the
IL-23
pathway,
and
it
will
be
important
to
test
these
therapeu-
tics
in
AS
patients.
3.1.
Ubiquitination,
aminopeptidases
and
MHC
class
I
presentation
Ubiquitination
is
the
process
of
adding
ubiquitin
groups
onto
proteins
that
directs
them
to
a
specific
sub-cellular
compartment
or
for
degradation
via
the
multi-catalytic
complex
called
the
pro-
teasome
(Pickart,
2001).
The
proteasome
degrades
the
protein
and
either
recycles
the
resultant
products
or
the
peptides
can
be
pre-
sented
on
MHC
class
I
molecules
on
the
cell
surface.
Ubiquitination
therefore
plays
an
important
role
in
determining
what
antigens
are
presented
to
the
immune
system.
UBE2E3
and
UBE2L3
have
recently
been
associated
with
AS
and
these
genes
encode
the
enzymes
UbcH9
and
UbcH7
respectively
(International
Genetics
of
Ankylosing
Spondylitis
et
al.,
2013).
However,
UbcH9
is
unable
to
form
bonds
with
ubiquitin
but
can
with
a
similar
protein
SUMO
(Desterro
et
al.,
1997).
When
a
protein
undergoes
sumoylation
it
is
not
degraded,
it
may
enhance
the
proteins
stability,
change
its
location
or
direct
involvement
in
other
cellular
processes
like
signal
transduction
(Muller
et
al.,
2001).
Variants
in
UBE2L3
are
involved
with
NF-␬␤
regulation
and
UBE2L3
has
been
associated
with
a
num-
ber
of
other
inflammatory
diseases
implicating
this
intracellular
signalling
pathway
as
a
shared
pathogenic
pathway
(Wang
et
al.,
2012).
Two
loci
(chromosomes
5p15
and
17q21)
containing
genes
encoding
four
aminopeptidases
have
now
been
implicated
in
AS
aetiology
(International
Genetics
of
Ankylosing
Spondylitis
et
al.,
2013).
The
chromosome
5p15
locus
contains
genes
that
encode
the
aminopeptidases
endoplasmic
reticulum
aminopeptidase
(ERAP)-
1,
ERAP2,
and
insulin
regulated
aminopeptidase
(IRAP
or
LNPEP).
The
chromosome
17q21
locus
contains
the
NPEPPS
gene
that
encodes
puromycin-sensitive
aminopeptidase.
The
same
primary
ERAP1
haplotype
associated
with
AS
is
also
associated
with
psoria-
sis
(Strange
et
al.,
2010),
and
the
AS-associated
ERAP2
haplotype
is
also
associated
with
both
psoriasis
and
inflammatory
bowel
disease
(Chapman
et
al.,
2010).
Suggestive
evidence
has
been
presented
that
ERAP1
variants
may
be
associated
with
type
1
diabetes
and
cervical
cancer
but
these
findings
have
not
been
universal
and
in
no
study
have
definitive
associations
been
reported
(Dendrou
et
al.,
2009;
Mehta
et
al.,
2007).
ERAP1
and
2
are
resident
in
the
endoplasmic
reticulum
(ER)
and
are
an
integral
part
of
the
MHC
class
I
presentation
pathway.
Once
peptides
have
been
processed
through
the
proteasome,
the
Trans-
porter
associated
with
Antigen
Processing
(TAP)
takes
the
resultant
peptide
from
the
cytoplasm
into
the
ER.
ERAP1
and
ERAP2
then
trim
any
N-terminally
extended
peptides
longer
than
9
amino-
acids
down
to
that
length
(Chang
et
al.,
2005;
York
et
al.,
2002),
which
is
the
favoured
length
for
subsequent
loading
onto
MHC
class
I
molecules
such
as
HLA-B27.
ERAP1
was
the
first
aminopeptidase
associated
with
AS
(Burton
et
al.,
2007).
The
association
has
been
widely
replicated
with
similar
allelic
and
haplotypic
associations
reported
in
both
populations
of
white
European
ancestry
and
in
east
Asians,
implying
that
common
variants
are
involved
rather
than
multiple
rare
variants
(Choi
et
al.,
2010;
Davidson
et
al.,
2009,
2011;
Harvey
et
al.,
2011;
Li
et
al.,
2010;
Lin
et
al.,
2011;
Maksymowych
et
al.,
2009;
Pimentel-Santos
et
al.,
2009).
Recently
interaction
between
HLA-B27
and
ERAP1
variants
has
been
described
in
AS,
such
that
variants
of
ERAP1
were
shown
only
to
influence
AS
disease
risk
if
HLA-B27
is
present
(Evans
et
al.,
2011).
This
suggests
that
ERAP1
associated
variants
operate
by
effects
on
peptide
processing
prior
to
HLA
Class
I
presentation.
It
had
previ-
ously
been
reported
that
ERAP1
is
involved
in
cleavage
of
cytokine
receptors
from
cell
membranes
(Cui
et
al.,
2002,
2003a,b;
Goto
et
al.,
2011),
but
studies
comparing
serum
levels
of
receptor
levels
in
wild
type
and
ERAP
knockout
mouse
found
no
differences
(Evans
et
al.,
2011).
Further,
in
humans
there
is
no
association
between
ERAP1
polymorphisms
and
serum
cytokine
receptor
levels
(Haroon
et
al.,
2010).
Fine-mapping
studies
indicate
that
the
SNP
rs30187
(Lys528Arg)
is
directly
disease-associated,
and
that
the
variation
rs10050860
(Asp575Asn)
marks
an
AS-associated
haplotype
that
also
carries
rs17482078
(Arg725Gln)
(Evans
et
al.,
2011).
The
link-
age
disequilibrium
between
rs10050860
and
rs17482078
is
very
tight,
and
it
has
not
been
possible
thus
far
to
determine
which
of
these
two
polymorphisms
is
the
key
AS-associated
variant,
or
if
both
are
actually
disease
associated.
Certainly
they
have
a
profound
effect,
with
a
reduction
in
AS
risk
in
HLA-B27
positive
homozygous
carriers
of
ERAP1
protective
variants
of
3–4
fold
(Evans
et
al.,
2011).
In
vitro
studies
of
peptidase
activity
of
ERAP1
and
its
variants
show
that
the
protective
variants
of
rs30187
and
rs17482078
are
associated
with
a
40%
reduction
in
peptidase
function,
whereas
rs10050860
has
no
effect
(Evans
et
al.,
2011).
This
suggests
that
rs10050860
is
not
the
true
disease-associated
variant
but
rather
rs17482078
is.
Using
recombinant
ERAP1
and
comparing
wild
type
protein
with
rs30187
and
rs27044
(Q730E)
variants,
Evnouchidou
4P.C.
Robinson,
M.A.
Brown
/
Molecular
Immunology
57 (2014) 2–
11
et
al.
demonstrated
that
the
functional
properties
of
these
variants
depend
on
the
substrate
concentration
and
sequence
(Evnouchidou
et
al.,
2008).
Crystallization
of
the
ERAP1
enzyme
and
resolution
of
its
struc-
ture
has
shown
that
AS-associated
variants
are
variously
located
in
the
transition
zone,
C-terminal
cavity
and
in
the
active
site
(Kochan
et
al.,
2011;
Nguyen
et
al.,
2011).
ERAP1
efficiently
pro-
cesses
peptides
9–16
residues
long
and
prefers
substrates
with
a
positively
charged
side
chains
and
a
large
hydrophobic
C
ter-
minal
residue
(Chang
et
al.,
2005;
Evnouchidou
et
al.,
2008).
But
variants
shown
functionally
to
negatively
affect
trimming
veloc-
ity
(rs30187,
rs17482078)
influence
the
transition
of
the
enzyme
from
the
open
to
the
closed
state
(Evans
et
al.,
2011;
Kochan
et
al.,
2011).
Therefore
it
is
more
likely
the
other
mapped
variants
are
not
the
functionally
important
variations,
but
are
in
LD
with
the
transition
state
variants.
ERAP
knockout
(KO)
mice
(mice
do
not
have
ERAP2)
are
fertile
with
no
obvious
phenotype
except
they
are
more
susceptible
to
intra-cellular
infection
with
toxoplasmo-
sis
(Blanchard
et
al.,
2008).
ERAP
KO
mice
have
20–80%
fewer
cell
surface
class
I
molecules
and
the
resultant
complexes
are
more
unstable,
and
don’t
illicit
normal
CD8T
cells
responses
(Hammer
et
al.,
2006).
HeLa
cells
treated
with
90%
effective
siRNA
show
sig-
nificantly
increased
MHC
class
I
expression
and
a
greater
supply
of
peptide
for
presentation
(York
et
al.,
2002).
This
suggests
that
ERAP1
degrades
some
antigenic
peptides
and
prevents
their
pre-
sentation.
Therefore
AS-associated
variants
in
ERAP1
change
the
amount
and
length
of
peptides
presented
as
well
as
the
stability
of
the
peptide-MHC
molecule
(Garcia-Medel
et
al.,
2012).
Therefore
ERAP1
could
influence
AS
disease
risk
through
either
changes
in
the
type
or
volume
of
peptides
presented.
ERAP2
was
first
reported
to
be
associated
with
AS
along
with
the
initial
description
of
association
of
ERAP1
with
the
disease
(Burton
et
al.,
2007).
Harvey
et
al.
demonstrated
haplotypic
association
across
the
ERAP1
locus
which
was
not
fully
explained
by
association
at
any
individual
SNP
(Harvey
et
al.,
2011),
and
Tsui
et
al.
reported
a
SNP
haplotype
involving
ERAP1
and
ERAP2
to
be
associated
with
AS
(Tsui
et
al.,
2010).
This
was
better
defined
by
a
conditional
anal-
ysis
demonstrating
an
associated
haplotype
spanning
ERAP2
and
LNPEP
influencing
AS
risk
independently
of
ERAP1
(International
Genetics
of
Ankylosing
Spondylitis
et
al.,
2013).
Because
of
the
extreme
linkage
disequilibrium
between
ERAP1
and
HLA-B27
and
between
ERAP1
and
ERAP2,
it
is
not
possible
to
determine
if
ERAP2
is
associated
with
AS
in
HLA-B27
positive
cases,
but
it
is
clearly
AS-associated
in
HLA-B27
negative
cases.
ERAP2
is
also
associated
with
psoriasis,
inflammatory
bowel
disease
and
juvenile
idiopathic
arthritis
(Chapman
et
al.,
2010;
Hinks
et
al.,
2013;
Tsoi
et
al.,
2012).
ERAP2
is
another
ER
resident
aminopeptidase
that
trims
N-
extended
peptides
prior
to
loading
onto
MHC
class
I
molecules;
it
has
49%
sequence
homology
to
ERAP1
(Birtley
et
al.,
2012).
ERAP1
and
ERAP2
have
been
identified
to
form
heterodimers
to
a
minor
degree
(10–30%)
and
in
such
a
way
synergistically
cleave
some
peptides,
but
the
significance
of
this
finding
is
not
known
(Saveanu
et
al.,
2005).
The
crystal
structure
of
ERAP2
has
been
determined
and
its
internal
cavity
has
more
positively
charged
side
chains
compared
to
ERAP1,
potentially
explaining
the
differ-
ing
peptide
trimming
preferences
(Birtley
et
al.,
2012).
Further,
ERAP2
does
not
have
the
additional
pocket
for
C-terminal
peptide
recognition
that
enables
ERAP1
to
trim
in
a
length
dependent
way
(Birtley
et
al.,
2012).
There
are
two
functionally
important
vari-
ants
in
ERAP2
that
are
associated
with
AS
(International
Genetics
of
Ankylosing
Spondylitis
et
al.,
2013).
The
most
important
is
a
vari-
ant
(rs2248374)
that
changes
the
strength
of
the
exon
10
splice
site
resulting
in
an
extended
exon
10
transcript
that
contains
two
stop
codons
(Andres
et
al.,
2010).
A
truncated
mRNA
is
transcribed
and
nonsense-mediated
decay
then
degrades
this
truncated
mRNA
and
the
protein
is
never
produced.
This
variant
causing
loss
of
ERAP2
protein
is
associated
with
reduced
MHC
class
I
surface
expression
in
cell
lines
(Andres
et
al.,
2010).
This
loss-of-enzyme
variant
is
protective
of
AS,
in
an
analogous
way
to
the
loss-of-function
vari-
ants
in
ERAP1
being
protective
of
AS
(International
Genetics
of
Ankylosing
Spondylitis
et
al.,
2013).
Another
variant
(rs2549782)
has
been
shown
to
alter
both
the
rate
and
specificity
of
ERAP2
trimming,
and
is
also
associated
with
AS
(International
Genetics
of
Ankylosing
Spondylitis
et
al.,
2013;
Evnouchidou
et
al.,
2012).
This
variant
is
in
strong
LD
with
the
loss-of-enzyme
variant
(from
1000
Genomes
D=
1.00,
r2=
0.90),
and
therefore
is
almost
never
transcribed
into
functioning
protein.
It
therefore
implicates
the
loss-of-enzyme
variant
(rs2248374)
as
the
functional
AS
associated
variant.
The
conditional
association
of
the
chromosome
5p15
locus
that
implicates
ERAP2
also
includes
the
LNPEP
gene
(International
Genetics
of
Ankylosing
Spondylitis
et
al.,
2013).
This
gene
encodes
the
aminopeptidase
IRAP.
IRAP
is
involved
in
trimming
of
peptides
during
the
process
of
cross-presentation
(Saveanu
et
al.,
2009).
Cross
presentation
is
primarily
described
in
dendritic
cells
(Joffre
et
al.,
2012),
whereby
exogenous
antigens
are
internalized
and
loaded
onto
class
I
molecules
either
through
the
traditional
MHC
class
I
pathway
in
the
ER
or
through
phagosomes
in
the
cytosol,
and
then
presented
on
the
cell
surface.
There
are
no
known
functional
variants
identified
in
LNPEP
associated
with
AS
and
the
relevance
to
AS
of
cross-presentation
is
unknown,
so
further
research
in
this
area
is
required.
At
the
chromosome
17
locus
harbouring
the
aminopeptidase
NPEPPS
there
still
remains
to
be
fine
mapping
and
functional
work
is
required
to
localize
the
association.
This
locus
has
two
indepen-
dent
signals,
one
involving
NPEPPS,
and
the
other
involving
the
neighbouring
genes
TBX21
and
TBKBP1.
NPEPPS
exists
in
the
cyto-
plasm
and
is
the
only
cytosolic
aminopeptidase
known
to
cleave
polyQ
sequences
(Menzies
et
al.,
2010).
Research
in
murine
mod-
els
of
neurodegenerative
diseases
shows
over-expression
reduces
protein
accumulation
and
enhances
macro-autophagy
(Kudo
et
al.,
2011).
The
genetic
association
of
IBD
with
autophagy
genes
such
as
ATG16L1
identified
for
the
first
time
a
role
for
autophagy
in
IBD
pathogenesis
(Rioux
et
al.,
2007).
Whilst
AS
is
not
associated
with
ATG16L1
(Danoy
et
al.,
2010),
evidence
of
autophagy
has
been
demonstrated
in
ileal
biopsies
in
AS,
and
has
been
shown
to
be
a
driver
of
IL-23
expression
in
the
gut
(Ciccia
et
al.,
2013).
Whether
NPEPPS
operates
in
AS
through
effects
on
autophagy
or
through
alternate
effects
on
intracellular
peptide
handling
requires
further
study.
Potentially
other
aminopeptidases
may
be
implicated
in
AS
aeti-
ology
in
the
future.
Peptides
are
trimmed
either
in
the
cytosol
or
ER
and
the
relative
importance
of
each
of
these
compartments
is
dependent
on
their
individual
N-terminal
sequence
(Hearn
et
al.,
2010;
Weimershaus
et
al.,
2013).
There
are
many
other
aminopep-
tidases
present
in
the
cytosol
that
may
play
a
role
in
trimming
of
peptides
that
could
potentially
impact
on
the
risk
of
AS,
and
new
aminopeptidases
continue
to
be
described
(Akkad
et
al.,
2012;
Dutoit
et
al.,
2012;
Saveanu
et
al.,
2002).
3.2.
IL-23
pathway
Multiple
constituents
of
the
IL-23/IL-17
pathway
have
now
been
associated
with
AS.
The
IL-23
cytokine
is
made
up
of
two
subunits:
IL-12p40
and
IL23p19.
IL12p40
is
encoded
by
the
IL12B.
Variants
near
IL12B
have
established
association
with
AS
(Danoy
et
al.,
2010;
Evans
et
al.,
2011).
IL-23
signals
through
the
IL23R,
and
protective
variants
that
reduce
sensitivity
to
IL-23
stimulation
are
strongly
associated
with
AS
(Di
Meglio
et
al.,
2011;
Sarin
et
al.,
2011).
The
finding
was
initially
described
in
white
Europeans
(Burton
et
al.,
2007),
and
the
corre-
sponding
alleles
were
not
polymorphic
in
Han
Chinese
(Davidson
P.C.
Robinson,
M.A.
Brown
/
Molecular
Immunology
57 (2014) 2–
11 5
et
al.,
2009),
but
a
sequencing
study
has
now
found
different
IL23R
variants
in
Han
Chinese
which
are
associated
with
AS
(Davidson
et
al.,
2013),
with
the
protective
alleles
of
these
variants
being
loss-of-function
variants.
Once
stimulated,
IL-23R
signals
through
the
JAK-STAT
pathway,
involving
the
kinases
TYK2
and
JAK2,
and
STAT3,
to
the
nucleus.
The
genes
encoding
each
of
these
proteins
have
been
associated
with
AS
susceptibility
(International
Genetics
of
Ankylosing
Spondylitis
et
al.,
2013;
Danoy
et
al.,
2010),
with
TYK2
variants
achieving
genomewide
significance.
Indeed
a
rare
variant
of
TYK2
which
is
predicted
to
affect
TYK2
splicing
has
an
odds
ratio
of
7.7
for
the
dis-
ease,
making
it
the
strongest
non-MHC
genetic
effect
as
measured
by
odds
ratio
(International
Genetics
of
Ankylosing
Spondylitis
et
al.,
2013).
These
findings
were
the
first
evidence
that
the
IL-23
pathway
was
involved
in
AS
(Brown,
2008),
and
have
stimulated
trials
of
antagonists
to
the
pathway
in
AS.
There
is
already
good
evidence
that
anti-IL-12p40
antibody
treatment
(ustekinumab)
and
anti-IL-
17
antibody
treatment
(secukimumab)
are
effective
in
AS,
and
trials
of
tofacitinib
(a
JAK
inhibitor)
and
fostamatinib
(a
TYK2
inhibitor)
are
now
underway
in
the
disease.
Research
stimulated
by
the
genetic
findings
is
now
demon-
strating
that
the
IL-23
pathway
is
a
critical
pathway
in
the
aetiopathogenesis
of
AS,
and
that
IL-23
over-expression
alone
is
sufficient
to
cause
spondyloarthritis
in
mouse
models
of
disease
(Adamopoulos
et
al.,
2011).
It
is
known
that
IL-23
is
over-expressed
in
the
bowel
of
AS
patients
(Ciccia
et
al.,
2009).
Over-expression
of
IL-23
with
a
mini-circle
vector
has
previously
been
shown
to
cause
arthritis
(Adamopoulos
et
al.,
2011).
Sherlock
and
col-
leagues
used
a
mini-circle
vector
to
endogenously
over-express
IL-23
in
mice,
resulting
in
the
development
of
axial
and
periph-
eral
spondyloarthritis
and
aortitis.
They
identified
a
population
of
IL-23R
positive
cells
resident
in
the
enthesis,
a
key
site
where
dis-
ease
occurs
in
AS
(Sherlock
et
al.,
2012).
These
entheseal
resident
cells
secreted
TNF,
IL-17
and
IL-22
in
response
to
IL-23;
inhibiting
either
IL-17
or
IL-22
could
block
disease,
but
IL-17
over-expression
alone
was
not
sufficient
to
cause
disease.
This
suggests
that
IL-23
may
be
the
key
cytokine
controlling
many
of
the
disease
manifesta-
tions
in
AS
and
other
spondyloarthropathies.
Consistent
with
these
data,
all
major
mouse
and
rat
models
of
spondyloarthritis
show
involvement
of
the
IL-23
pathway,
including
the
proteoglycan-
induced
model,
TNF-overexpression
model,
skg
model
and
even
B27-transgenic
rats
(Boldizsar
et
al.,
2009;
DeLay
et
al.,
2009;
Kontoyiannis
et
al.,
2002;
Ruutu
et
al.,
2012).
Thus
it
appears
that
almost
irrespective
of
the
cause
of
IL-23
over-expression,
the
func-
tional
consequence
is
a
spondyloarthritis-like
phenotype.
3.3.
Intergenic
regions
Two
replicated
associations
with
AS
have
been
to
intergenic
regions
at
chromosomes
2p15
and
21q22,
where
no
translated
gene
product
is
encoded
(Reveille
et
al.,
2010).
At
chromosome
21
PSMG1
is
the
nearest
gene,
but
is
82
kb
away,
and
there
is
no
difference
in
expression
of
this
gene
in
AS
cases
compared
with
con-
trols
or
in
relation
to
the
AS-associated
SNPs
at
this
locus
(Reveille
et
al.,
2010).
At
both
these
intergenic
regions
RNA-seq
data
identi-
fied
novel
long
non-coding
RNA
transcripts,
raising
the
possibility
that
the
molecular
mechanism
underlying
these
associations
is
through
effects
of
germline
sequence
variation
on
non-coding
RNA
sequence
or
transcription
(Reveille
et
al.,
2010).
3.4.
Lymphocyte
development
and
activation
Multiple
transcription
factors
involved
in
lymphocyte
develop-
ment
and
activation
have
been
associated
with
AS.
These
include
EOMES,
RUNX3,
TBX21,
ZMIZ1,
IL7,
and
IL7R
(Danoy
et
al.,
2010;
International
Genetics
of
Ankylosing
Spondylitis
et
al.,
2013;
Reveille
et
al.,
2010).
RUNX3
influences
thymic
T
cell
develop-
ment
(Park
et
al.,
2010)
and
AS
associated
RUNX3
variants,
as
well
as
the
AS-associated
variants
in
IL7R
and
ZMIZ1,
are
associ-
ated
with
reduced
CD8T
cell
counts
in
healthy
individuals
(Evans
et
al.,
2011).
RUNX3
knockout
(KO)
mice
spontaneously
develop
inflammatory
bowel
disease
at
four
weeks
of
age,
demonstrating
its
important
role
in
AS-organ
specific
immune
function
(Brenner
et
al.,
2004).
IL7R
is
another
gene
involved
in
T
cell
development
as
IL7R
chain
KO
mice
are
devoid
of
␥␦
T
cells
and
very
deficient
in
␣␤
T
cells
and
B
cells
the
defect
causes
a
lack
of
T
cell
recep-
tor
locus
rearrangement
(Candeias
et
al.,
1997).
EOMES
encodes
eomesodermin,
a
transcription
factor
involved
in
CD8
differentia-
tion,
the
expression
of
which
is
induced
by
RUNX3
(Cruz-Guilloty
et
al.,
2009;
Pearce
et
al.,
2003;
Yagi
et
al.,
2010).
Eomesodermin
deficiency
promotes
IL-17
expression
by
CD8
T-cells
(Intlekofer
et
al.,
2008).
IL-7
acts
through
IL-7R
to
induce
RUNX3
expres-
sion
in
developing
T-cells,
in
turn
favouring
differentiation
towards
the
CD8-lineage
(Franke
et
al.,
2010).
ZMIZ1
is
a
transcriptional
co-activator
of
the
Protein
Inhibitor
of
Activated
STAT
(PIAS)-
like
family;
thus
it
may
affect
STAT-mediated
cytokine
signalling.
ZMIZ1
interacts
with
activating
NOTCH1
mutations
to
induce
T-cell
acute
lymphoblastic
leukaemia,
suggesting
a
role
in
T-cell
differen-
tiation
(Rakowski
et
al.,
2013).
TBX21
encodes
T-bet,
a
transcription
factor
which
is
important
in
regulation
of
T-
and
NK-cell
function
and
development.
T-bet
expression
is
controlled
by
IL-23
(Klose
et
al.,
2013),
and
T-bet
itself
controls
IL-22
secretion
by
innate
lymphoid
cells
in
the
gut
(Sciume
et
al.,
2012).
These
associations
implicate
the
differentiation
of
immune
cells
as
being
important
to
AS
pathogenesis,
but
it
is
not
clear
which
of
these
cells
are
primar-
ily
affected
by
the
genetic
variants
identified.
The
recent
discovery
of
the
expansion
of
innate
lymphoid
cell
populations,
such
as
␥␦T
(Kenna
and
Brown,
2013)
and
NKp46
IL-22
expressing
cells
(Ciccia
et
al.,
2012),
in
AS
cases
is
particularly
relevant
as
these
genes
are
known
to
be
major
determinants
of
differentiation
and
activation
of
these
cell
types,
suggesting
that
they
play
a
causative
role
in
AS
pathogenesis
(reviewed
in
Kenna
and
Brown,
2013).
3.5.
G-protein
coupled
receptors
Three
G-protein
coupled
receptors
(GPR)
(GPR35,
GPR37
and
GPR65)
have
now
been
associated
with
AS
(International
Genetics
of
Ankylosing
Spondylitis
et
al.,
2013).
Generally
these
receptors
are
not
well
characterized,
but
are
known
to
mediate
a
wide
range
of
cellular
signalling
(Rosenbaum
et
al.,
2009).
GPR35
is
known
to
be
expressed
predominately
in
immune
and
gastrointestinal
tis-
sues,
both
sites
important
in
AS
(Wang
et
al.,
2006).
It
is
thought
indoleamine
2,3-dioxygenase
(IDO),
an
important
immune
system
mediator
signals
through
GPR35
(Ren
et
al.,
2009).
Amongst
other
actions,
IDO
influences
tolerance
in
dendritic
cells
(Pallotta
et
al.,
2011)
and
therefore
alterations
in
its
ability
to
signal
could
well
influence
inflammatory
disease.
GPR37
was
noted
to
be
down
regulated
in
children
with
severe
influenza,
but
other
immune
functions
are
unknown
(Heltzer
et
al.,
2009).
It
is
also
called
Parkin-Associated
Endothelin
Receptor-
Like
Receptor,
over-expression
of
which
causes
it
to
misfold,
accumulate
and
induce
macro-autophagy
(Marazziti
et
al.,
2009).
This
association
with
macro-autophagy,
along
with
the
role
of
NPEPPS
in
clearing
large
accumulated
proteins
may
suggest
this
pathway
is
important
in
AS,
as
it
is
in
IBD
(Khor
et
al.,
2011).
GPR65
is
also
known
as
T
cells
death
associated
gene
8
(TDAG8)
has
been
shown
to
be
involved
in
thymocyte
apoptosis,
implicating
this
important
process
of
T
cell
development,
already
highlighted
by
the
genetic
associations
described
above
(Tosa
et
al.,
2003).
6P.C.
Robinson,
M.A.
Brown
/
Molecular
Immunology
57 (2014) 2–
11
3.6.
IL-1
genes
Prior
to
the
GWAS
era,
suggestive
evidence
of
the
involvement
of
IL-1
genes
in
AS-pathogenesis
was
reported
(Djouadi
et
al.,
2001;
Sims
et
al.,
2008;
van
der
Paardt
et
al.,
2002).
These
asso-
ciations
have
not
been
subsequently
replicated
in
GWAS
studies,
likely
because
this
region
is
now
known
to
be
subject
to
substan-
tial
population
stratification
effects
that
earlier
studies
were
not
able
to
control
for.
Nonetheless,
GWAS
studies
have
now
conclu-
sively
demonstrated
the
association
of
the
IL1R1-IL1R2
locus
with
AS
(International
Genetics
of
Ankylosing
Spondylitis
et
al.,
2013;
Reveille
et
al.,
2010).
Two
independent
association
signals
are
found
at
this
locus,
implicating
both
genes
in
AS
pathogenesis.
Innate
immune
stimulation
through
pattern
recognition
receptors
stimu-
lates
the
production
of
pro-IL-1
that
is
cleaved
to
IL-1
by
the
inflammasome
and
released.
IL1R1
mediates
many
of
the
responses
of
IL-1
whereas
IL1R2
encodes
a
decoy
receptor
for
IL-1,
compet-
itively
binding
IL-1
and
preventing
its
binding
to
IL-1R1.
A
role
for
IL-1
in
AS
inflammation
is
supported
by
the
efficacy
of
anakinra
in
AS,
although
the
level
of
benefit
from
this
agent
was
much
lower
than
that
observed
with
TNF-inhibitors
and
therefore
it
has
not
entered
clinical
practice
for
AS
management
(Haibel
et
al.,
2005;
Tan
et
al.,
2004).
Historically
B
cells
were
thought
not
to
play
a
role
in
AS
patho-
genesis,
but
recently
the
efficacy
of
rituximab
has
been
reported,
especially
in
those
naïve
to
TNF
blockers
(Song
et
al.,
2013).
A
high
prevalence
of
antibodies
to
CD74,
that
target
the
HLA
class
II
invari-
ant
chain
peptide,
has
also
been
reported,
particularly
in
early
AS
(Baerlecken
et
al.,
2013;
Baraliakos
et
al.,
2013).
Intriguingly,
sug-
gestive
association
of
CD74
with
AS
has
previously
been
reported
(Burton
et
al.,
2007),
suggesting
a
possible
role
in
disease
causation.
These
data
suggest
that
contrary
to
the
long
held
belief,
B
cells
may
play
some
role
in
AS
pathogenesis.
This
hypothesis
is
supported
by
the
recent
association
of
AS
with
the
B-cell
specific
transcription
factor
BACH2.
This
gene
transcribes
the
BACH2
protein
that
has
high
expression
in
B
cells
prior
to
the
plasma
cell
stage
of
B
cell
develop-
ment
(Muto
et
al.,
2004).
BACH2
is
a
critical
player
in
class
switch
recombination
and
somatic
hypermutation
(Muto
et
al.,
2004).
Lit-
tle
more
is
known
about
the
mechanisms
by
which
this
association
operates.
The
receptor
through
which
IL-6
signals,
IL6R,
has
been
asso-
ciated
with
AS
(International
Genetics
of
Ankylosing
Spondylitis
et
al.,
2013).
The
AS
associated
SNP
has
been
shown
to
be
associ-
ated
with
asthma,
CRP
levels
and
pulmonary
function
in
previous
studies
(Dehghan
et
al.,
2011;
Evans
et
al.,
2011;
Wilk
et
al.,
2007).
IL6
is
a
proinflammatory
cytokine
with
pleiotropic
effects,
and
is
instrumental
in
directing
pathogenic
TH17
responses
(Burton
et
al.,
2007).
The
IL6R
SNP
correlates
with
CRP
levels,
which
are
elevated
in
AS.
Two
therapeutics
directed
against
IL-6R
have
been
trialled
and
found
not
to
be
sufficiently
effective
to
progress
through
full
development
and
marketing.
4.
Non-axial
AS
manifestations
4.1.
Age
of
symptom
onset
The
age
at
which
AS
patients
develop
their
first
symptoms
is
significantly
influenced
by
genetics,
with
family
studies
suggesting
the
heritability
of
age
of
onset
is
33–50%
(Brown
et
al.,
2003;
Sims
et
al.,
2007).
Many
studies
have
found
HLA-B27
positive
patients
develop
symptoms
3–9
years
earlier
than
their
B27-negative
coun-
terparts
(Feldtkeller
et
al.,
2003;
Jaakkola
et
al.,
2006;
Queiro
et
al.,
2008).
A
large
Finnish
family
study
also
found
association
between
HLA-DRB1
alleles
and
age
at
symptom
onset
(Jaakkola
et
al.,
2006).
There
does
not
seem
to
be
an
association
between
age
of
onset
and
severity,
suggesting
susceptibility
and
severity
are
not
linked
(Brophy
and
Calin,
2001).
4.2.
Anterior
uveitis
Autoimmune
anterior
uveitis
(AU)
is
a
T
cell
mediated
condition
that
causes
inflammation
of
the
iris
and
ciliary
body
of
the
eye.
It
occurs
in
30–40%
of
AS
patients
episodically
and
is
also
strongly
associated
with
HLA-B27
(Brewerton
et
al.,
1973a).
It
also
occurs
in
other
autoimmune
conditions
such
as
Bechet’s
disease,
sarcoidosis
and
Vogt–Koyanagi–Harada
disease.
One
of
the
issues
with
determining
uveitis
specific
genetic
asso-
ciations
apart
from
HLA-B27
is
the
challenge
of
separating
genes
that
are
AS-specific,
AU-specific
or
predispose
to
both
conditions.
Few
genes
have
been
robustly
replicated
as
AU
specific
risk
genes.
Non-B27
MHC
alleles
that
have
been
associated