Doctor Leonardo Ferreria

International Immunology © The Japanese Society for Immunology. 2020. All rights reserved.
doi:10.1093/intimm/dxaa054 For permissions, please e-mail: [email protected]

Advance Access publication 18 August 2020
Immunodietica: interrogating the role of diet in
Downloaded from https://academic.oup.com/intimm/advance-article/doi/10.1093/intimm/dxaa054/5893934 by Serials Section, Dixson Library user on 08 September 2020
autoimmune disease
Iosif M. Gershteyn1–3, Andrey A. Burov1, Brenda Y. Miao4, Vasco H. Morais5 and
Leonardo M. R. Ferreira1,4,5,
1
Ajax Biomedical Foundation, Newton, 02461 MA, USA
2
ImmuVia LLC, Waltham, 02453 MA, USA
3
SoundMedicine LLC, Waltham, 02453 MA, USA
4
Diabetes Center and
5
Department of Surgery, University of California, San Francisco, San Francisco, 94143 CA, USA
Correspondence to: I. M. Gershteyn; E-mail: [email protected] or L. M. R. Ferreira; E-mail: [email protected]
Received 15 May 2020, editorial decision 8 August 2020; accepted 10 August 2020
Abstract
Diet is an environmental factor in autoimmune disorders, where the immune system erroneously
destroys one’s own tissues. Yet, interactions between diet and autoimmunity remain largely
unexplored, particularly the impact of immunogenetics, one’s human leukocyte antigen (HLA) allele
make-up, in this interplay. Here, we interrogated animals and plants for the presence of epitopes
implicated in human autoimmune diseases. We mapped autoimmune epitope distribution across
organisms and determined their tissue expression pattern. Interestingly, diet-derived epitopes
implicated in a disease were more likely to bind to HLA alleles associated with that disease than
to protective alleles, with visible differences between organisms with similar autoimmune epitope
content. We then analyzed an individual’s HLA haplotype, generating a personalized heatmap
of potential dietary autoimmune triggers. Our work uncovered differences in autoimmunogenic
potential across food sources and revealed differential binding of diet-derived epitopes to
autoimmune disease-associated HLA alleles, shedding light on the impact of diet on autoimmunity.
Keywords: alimentation, autoimmunity, epitope, HLA, immunogenetics
Introduction
Food has the potential to affect every aspect of one’s health. stiffness, swollen joints and other symptoms in rheumatoid
It provides not just energy and building blocks to our cells, arthritis (RA) patients (5). Food hypersensitivity was detected
but also many molecules with pharmacological properties, in 39% of infants tested in double-blind placebo-controlled
leading many to see food as medicine. Unsurprisingly, there food challenge (DBPCFC) studies (6). Several ways in which
is a push towards dissecting the composition of food at the diet impacts autoimmunity have been defined at the mo-
molecular level in greater detail (1). lecular level. High salt intake enhances the differentiation of
The role of diet in the prevalence and progression of pathogenic Th17 cells, a subset of CD4+ T helper cells of the
autoimmune disorders is poorly understood and presently adaptive immune system involved in autoimmunity, via the
understudied. There are almost 100 described autoimmune SGK1 kinase, and aggravates neural inflammation (7, 8). Red
disorders, affecting an estimated 50 million Americans. Still, meat contains a sugar absent in humans, Neu5Gc, which
the etiology of most cases remains unknown. Twin studies upon absorption in the intestine and incorporation in proteins
across different countries have revealed that genetics alone and lipids, is recognized by xenoautoantibodies, leading to
can only predict 22% of cases of common autoimmune dis- chronic inflammation (9). Yet, the connection between the
eases (2), leading to an increased recognition of the import- diet of autoimmune disease patients and their symptoms
ance of environmental factors (3). remains opaque.
Diet likely plays a key role among environmental factors Consumption of food-derived proteins containing auto-
affecting autoimmune disease incidence and severity. Milk immune epitopes may contribute to the commonly noticed
consumption positively correlates with multiple sclerosis links between certain foods and autoimmune disease flares.
(MS) incidence across 27 countries (4). Fasting, followed by Molecular mimicry is a phenomenon that occurs when a for-
1 year of vegetarian diet, significantly reduces pain, morning eign antigen shares an immunogenic sequence, an epitope,
Page 2 of 13 Interactions between diet and autoimmunity
with self-antigens (10). Epitopes can be either recognized derived from Bacteroides species in the intestine in mice (13).
directly by antibodies, either soluble or on the surface of B Increased levels of high affinity food-specific IgG antibodies
cells, or presented by specialized molecules, human leuko- have been detected in the intestine and serum of RA patients
cyte antigens (HLAs), to T cells. Within the HLA antigen (14). Fifteen years ago, pork abattoir workers processing
presentation system, HLA class I molecules (HLA-A, -B, -C) pig brains using compressed air developed an autoimmune
present to cytotoxic CD8+ T cells, whereas HLA class II mol- neurologic disorder. Studies concluded that aerosolized pig
ecules (HLA-DR, -DP, -DQ) present to CD4+ T helper cells neural antigens, of identity yet to be determined, induced
(11). Instances of molecular mimicry as the cause of auto- polyradiculoneuropathy, a normally slow-developing disease,
immune disease have been described both in mouse and in in 4 weeks (15).
human studies. Bacterial and viral infections were first impli- How are diet-derived epitopes recognized by the immune
cated in the pathogenesis of human autoimmune disease for system? Most food absorption takes place in the small intes-
Guillain-Barré syndrome, a neurological autoimmune disease tine. Intestinal barrier integrity is compromised by alcohol,
(12). Progression of myocarditis, an autoimmune disease processed food consumption, inflammation and aging (16–
where T cells recognizing myosin heavy chain destroy car- 18). This can lead to a ‘leaky gut’, allowing for the release
diac tissue, requires priming of these cells by peptides of bacterial products, such as flagellin and endotoxin, into
Fig. 1. Mapping interactions between diet and autoimmune disease. Autoimmune disease-implicated epitopes were aligned against the com-
plete proteomes of food organisms (top). Perfect matches were used to determine the autoimmunogenicity of food organisms with respect to
each of the 69 diseases analyzed (bottom center). These hits were then mapped back into human proteins to determine their organ and tissue
expression (bottom left), as well as tested for binding to autoimmune disease-associated HLA (human leukocyte antigen) molecules to con-
struct personalized sensitivity passports (bottom right).
Interactions between diet and autoimmunity Page 3 of 13
the blood stream and concomitantly partly undigested food a fully comprehensive database of autoimmune human-
antigens structurally similar to self-antigens in an inflamma- organism epitope overlap.
tory context (19, 20). Blood–brain barrier integrity is also
compromised with aging (21), so antibodies against diet- Mapping human autoimmune epitope tissue expression
derived epitopes mimicking neural epitopes can reach the
A list of 10 605 epitopes associated with human autoimmune
central nervous system and cause damage (22).
diseases was passed into a function converting them into
Here, we tackled the challenge of studying potential inter-
FASTA format and queried against the Swiss-Prot database
actions between diet and autoimmunity by creating a com-
(hosted by NIH) to acquire the UniProt name associated
prehensive database of overlap between human autoimmune
with the protein containing each epitope in the list. Queries
epitopes and food epitopes, mapping their tissue expression
were performed by directly parsing through the database.
pattern, and analyzing their presentation by HLA alleles asso-
In the instance where a direct lookup failed, BLASTp (ncbi-
ciated with autoimmune disorders (Fig. 1).
blast-2.10.0+) was run to find a match within the same data-
base. Because of the lengthy runtime and computational
Methods intensity associated with BLASTp queries, BLASTp was
only run if a linear scan of the database yielded zero results.
Human autoimmune epitope identification in food sources Threshold parameters were used to limit queries to only one
To identify epitopes implicated in human autoimmune result and only perfect matches (100% identity across all
disease prevalent in commonly consumed foods, we ag- amino acids in query and matching sequence). The gene
gregated 10 605 linear peptide epitopes implicated in 69 symbol for each epitope was obtained by querying db2db
autoimmune diseases, obtained from www.iedb.org (23) (re- API (bioDBnet) with the associated UniProt obtained in the
trieved September 10, 2019), and cross-referenced them previous query. This query resulted in a data table linking
with the proteomes of 24 organisms: alpaca (taxid:30538), the original list of epitopes with their associated gene sym-
bat (taxid:9397), bison (taxid:9901), salmon (taxid:8030), bols. Expression profiles for proteins in human tissues were
camel (taxid:419612), whale (taxid:9721), chicken obtained from the Human Protein Atlas (data from the Human
(taxid:9031), chimpanzee (taxid:9598), cow (taxid:9913), Protein Atlas version 19.3 and Ensembl version 92.38). For
duck (taxid:8839), human (Homo sapiens,taxid:9606), goat each protein in the tissue expression data, Ensembl name,
(taxid:9925), rice (taxid:39947), lettuce (taxid:4236), turkey gene symbol name, and tissue and cellular localization were
(taxid:9102), tilapia (taxid:8128), rabbit (taxid:9986), pig provided. In addition, the data detailed the level of expres-
(taxid:9823), potato (taxid:4113), quinoa (taxid:63459), rye sion (High, Medium, Low, Not detected) and the reliability
(taxid:4550), sheep (taxid:9940), soybean (taxid:3847) and of the described expression level and localization on the
wheat (taxid:4564). We then utilized Node.js, AWS EC2 and basis of existing literature (Approved, Enhanced, Supported,
PostgreSQL database technologies to automate data gath- Uncertain). To obtain the tissue-level localization of the auto-
ering and expedite analysis. Data gathering automation was immune disease-associated epitopes, the tissue expression
broken into two custom processes. Process #1 queried the data from the Human Protein Atlas and the list of epitopes
National Center for Biotechnology Information (NCBI) blastP were joined together using gene symbol as a matching iden-
system. The query request consisted of the unique epitope tity key. Data for proteins in which reliability was ‘Uncertain’
ID and the list of organisms. All the queries were checked or expression level was ‘Undetected’ were removed. A sep-
against the NCBI refseq_protein database. The API supports arate table labeling each tissue with its associated organ and
a list of organisms as a valid query criterion, significantly cut- tissue system was joined in order to provide tissue system
ting down the round trip of each query. Once a request was and organ-level localization of the epitopes. Each epitope in
made for a particular epitope in the database, Process #1 the resulting data table was labeled with its associated auto-
created a record with the Request ID where the uniqueness immune disease to create a database. Next, the same 10 605
of the record was the epitope ID and the organism. The re- epitopes were queried against the complete proteome of the
cord was marked as pending to indicate that it is ready to 24 organisms indicated above in Human autoimmune epi-
be retrieved for processing by Process #2. The process con- tope identification in food sources. Only matches with 100%
tinued until all the epitopes had been queried and marked as identity were retained. The same pipeline detailed above
pending. Process #2 retrieved the results of the queries via was followed and the final table of proteins and their organ-
Request ID from Process #1. As the query could take an un- level expression was matched against the complete set of
certain amount of time to finish, the process iterated through diet-derived epitopes generated in the previous step. The
the list of all the records that had a valid Request ID to obtain resulting database consisted of the amino acid sequence,
the results. In the query result data, the top hit for each or- gene symbol, associated organism and organ-level expres-
ganism was selected and then, if the Query Coverage and sion and localization for each epitope. The database was
Identity Percentage values both equaled 100, the record re- then filtered for cow (taxid:9913), pig (taxid:9823), chicken
sult was considered as a ‘hit’ (total match). In all other cases, (taxid:9031), salmon (taxid:8030) and rice (taxid:39947).
a ‘miss’ was recorded. Once hit or miss was recorded, the
process was marked as complete so the automated query
would ignore on the next pass of requests. The system con- Prediction of HLA binding and disease association
tinued polling NCBI servers until all results were gathered. The IEDB MHC-I and MHC-II Binding Prediction tools (tools.
Once both processes were fully executed, we had compiled immuneepitope.org) API were used to estimate binding affinity
of each autoimmune epitope to various HLA alleles in silico (24). for which previous OR data were available in at least three
For HLA class I alleles, 2306 epitopes were assessed, whereas predisposing and three protective alleles. Data were analyzed
for HLA class II, 4547 epitopes were analyzed. The length of using the SciPy statistics library (28). A Pearson coefficient
each queried peptide was set to 8–14 amino acids for each was calculated to determine the correlation between the ORs
HLA class I allele and 11–30 amino acids for each HLA class II and number of binding epitopes in each disease. Two-tailed
allele. As per previous protocols (24), epitope–HLA pairs with t-tests, with equal variances assumed, were used to identify
percentile rank scores <1% were considered strong binders to whether there was a statistically significant difference between
HLA class I and those with percentile rank scores <10% were the number of binding epitopes that were predisposing (OR >
considered strong binders to HLA class II, respectively, and 1) or protective (OR < 1) for each disease. P values of <0.05
used for downstream analysis. BLASTp was used to identify or- were considered significant.
ganisms commonly used as food sources that also had these Code created to carry out analyses for this study is avail-
strong-binding epitopes. The number of binding epitopes for able for researchers on the SourceForge associated with the
each associated organism and type of autoimmune disease Immunodietica project. Moreover, an interactive website for
were then recorded to create a map of predicted food sensi- researchers and the public to explore this database will be
tivity (Gershteyn-Ferreira sensitivity passport) for the set of HLA available at www.immunodietica.com.
alleles originally queried. The total number of binding epitopes
for each HLA was also compared to its corresponding odds
Results
ratio (OR), which were identified from previous genome-wide
association studies (25–27). Specifically, we looked at class II We built a database with all linear peptide epitopes impli-
alleles for RA (25), type 1 diabetes (T1D) (26) and MS (27), cated in human autoimmune diseases, obtained from www.
A D
Autoimmune disease epitopes vs. antigens acquired epidermolysis bullosa 800
Addison's disease
600 Behcet disease alopecia areata
ankylosing spondylitis
Multiple sclerosis antiphospholipid syndrome
autoimmune atherosclerosis
# antigens
400 autoimmune gastritis

autoimmune glomerulonephritis
Rheumatoid arthritis autoimmune haemolytic anaemia
autoimmune hepatitis
autoimmune optic neuritis
200 autoimmune pancreatitis
SLE autoimmune polyendocrine syndrome
autoimmune polyendocrine syndrome type 1
autoimmune thyroiditis
T1D Celiac disease autoimmune uveitis
0
0 500 1000 1500 2000 2500
autoimmune vasculitis
Behcet disease
600
# epitopes bullous pemphigoid
celiac disease
cicatricial pemphigoid
B Crohn's disease
cryoglobulinemia
% of total autoimmune disease epitopes cutaneous lupus erythematosus
demyelinating polyneuropathy
dermatomyositis
100 encephalomyelitis
Goodpasture's syndrome
Grave's disease
80 Guillain-Barre syndrome
hemolytic-uremic syndrome
60 immune thrombocytopenic purpura
type 1 diabetes
40
juvenile ankylosing spondylitis
juvenile rheumatoid arthritis
400
Lambert-Eaton myasthenic syndrome
mixed connective tissue disease
20 multiple sclerosis
myasthenia gravis
0 neuromyelitis optica
non-insulin-dependent diabetes mellitus
paraneoplastic polyneuropathy
ot on
d
im an
C n
A ia
us
nk an
Fu al
l
la e
B us
t
ea
an
ol
w
op a
r
pars planitis
m
Pr kt
Sl ozo
ng
um
yt Alg
te
r
no
m
er
Pl
Vi
ni
pemphigus
ac
e
H
pemphigus gestationis
U
peripheral neuropathy
Ph
prediabetes syndrome
primary biliary cirrhosis
C psoriasis
psoriatic arthritis
reactive arthritis 200
Autoimmune epitope content (%) relapsing polychondritis
rheumatic myocarditis
100 rheumatoid arthritis
sclerosing cholangitis
sensory neuropathy
80 Sjogrens syndrome
stiff-person syndrome
60 systemic lupus erythematosus
systemic scleroderma
thrombotic thrombocytopenic purpura
40 thyrotoxic exophthalmos
ulcerative colitis
vasculitis
20 vitiligo
Vogt-Koyanagi-Harada
0 Wegener's granulomatosis
0
quinoa
potato
turkey
bison
chicken
soybean
salmon
whale
duck
goat
camel
pig
rye
sheep
tilapia
alpaca
lettuce
rice
wheat
rabbit
cow
hi n
p
ha t
C le
Sh ow
p
lp t
a
a g
am t
C Bis el
c n
rk n
D ey
Sailap k
lm ia
So R on
yb ice
u n
Leota a
ttu to
ce
he e
at
W Ba
A Goa
C bbi
ac
T uc
P ino
W Ry
C ma
m
ee
R Pi
hi o
Tu ke
Q ea
u
H
Fig. 2. Mapping diet-derived epitopes implicated in human autoimmune disease. (A) Correlation between the number of autoimmune epitopes
and antigens across different diseases (R2 = 0.6620). (B) Attributed origin of epitopes implicated in human autoimmunity (www.iedb.org). (C)
Human autoimmune epitope content of 24 species. (D) Heat map of autoimmune epitopes present in organisms commonly consumed as food
(x-axis) per disease (y-axis).
iedb.org (23). The diseases with most identified epitopes Diet-derived antigens have been found to mimic a variety
were, in order of decreasing number of epitopes, MS, RA, of human tissue-specific antigens (30). Hence, we analyzed
celiac disease, systemic lupus erythematosus (SLE), T1D the expression pattern of all human autoimmune epitopes
(also known as insulin-dependent diabetes mellitus) and (Fig. 4A), as well as of those implicated in individual dis-
Behcet disease (Supplementary Figure S1). As expected, eases (Fig. 4B–F). Most epitopes found in MS patients are
these same diseases also had the highest number of iden- expressed in the brain (Fig. 4B), whereas most epitopes
tified protein antigens, with the exception of celiac disease implicated in T1D are present in the pancreas (Fig. 4E). Of
(Supplementary Figure S2; Fig. 2A). Interestingly, of all auto- note, epitopes for human autoimmune disease in general are
immune epitopes, <80% could be traced back to human, with most commonly expressed in the brain (Fig. 4A). Indeed, the
part of them being only found in cereals, bacteria or viruses disease with most identified epitopes is a neurological dis-
(Fig. 2B). To determine the autoimmunogenicity of different order, MS (Fig. 4B; Supplementary Figure S1) and brain is
foods, we mapped all 10 605 human autoimmune epitopes also the second most common tissue of expression for T1D
onto the proteomes of 24 organisms. Importantly, only exact epitopes (Fig. 4E). Strikingly, all analyzed organisms contain
matches were considered (Table 1), as the degree of peptide epitopes that map back to human proteins present in every
sequence similarity does not predict molecular mimicry or organ, from the cow and pig to rice (Fig. 5A–F).
cross-reactivity (29). Organisms commonly consumed as food Next, we sought to investigate the binding of the identified
fall in three categories according to their autoimmune epitope diet-derived autoimmune epitopes to HLA alleles. As per the
content: red meat, poultry and fish, and cereals and veget- International Immunogenetics Database, there are currently
ables. Curiously, the non-primate organism with the highest almost 20 000 different HLA class I and 10 000 HLA class II
autoimmune epitope content was the bat (Fig. 2C). Closing alleles identified in the world population, resulting in every
in at the individual disease level, we found that this pattern person having a unique HLA repertoire. Therefore, mapping
holds true for diseases with most known epitopes, such as diet-derived epitope–HLA binding has the potential to per-
MS, SLE and RA (Fig. 2D). Nonetheless, exceptions exist sonalize the assessment of the risk that each food poses in
in diseases with less-known epitopes. We divided diseases terms of autoimmunogenicity. In addition, many HLA alleles
in three categories according to the number of implicated have been either associated or found to be protective against
epitopes: ‘hot diseases’ with >500 epitopes, warm diseases multiple autoimmune disorders, allowing us to gain insight
with between 50 and 500 epitopes and cold diseases with into the relative affinity of a given disease-linked HLA allele to
<50 epitopes (Supplementary Table S1). Inspecting diseases diet-derived epitopes implicated in that same disease. HLA
in each one of these categories revealed, for instance, that pig class I-restricted epitopes are typically 8–11 amino acids
contains the most epitopes implicated in autoimmune gas- long (31). Yet, peptides as long as 15 amino acids have been
tritis, whereas the same is true for rabbit and demyelinating found to bind to some HLA class I alleles with affinity and
polyneuropathy, and rice for pemphigus gestationis (Fig. 3). stability comparable to those of canonical length (32, 33).
Table 1. Interspecies alignment of human autoimmune disease epitopes
Organism Insulin B9-23 (T1D) GDP-L-fucose synthase 96–105 (MS) Cardiac myosin S2 1052–1076 (myocarditis)
Wheat S+LVE+L------RG NV+HS----- ------DLKL------DLE+-----

Rye S+LVE+L------RG NV+HS----- ---------LTQ+++S+++L+N++Q+L
Lettuce ----EALY++++LVC+E-- NVLHSAFE-- ------LK+L++E++SI+M+LEN+KQ+L
Potato -HL++ALYL++GER- NV++++LHS+FEVG --KLE++L+++++KLTQ+SI+++E------
Quinoa -HL+EAL+LV----- N+LH++FEV- RK+++LE+++LKL+QE++MD+EN++QQ-
Soybean -HLV---YLVC++RG -VLHS+FE-- KLE+DLKL+++++++T+E+I++L++DK---
Rice -HL++ALY+VC+E+RG +VLH+AFE+G ---------L+Q++IM+DLEN+++KQ+L
Salmon SHLV+ALYLVCGE+G --L++AFEVG KRKLEGDLKLTQES+MDLENDKQQL
Tilapia SHLV+ALYLVCG+RG NVL++A+EVG KRKLEGDLKLTQES+MDLENDKQQL
Duck SHLVEALYLVCGERG NVLHSA+E-- KRKLEGDLKLTQESIMDLENDKQQL
Turkey SHLVEALYLVCGERG NVLHSA+E-- KRKLEGDLKLTQES+MDLENDKQQL
Chicken SHLVEALYLVCGERG NVLHSA+E-- KRKLEGDLKLTQES+MDLENDKQQL
Bison --LV+AL++VCG+RG NVLHSAFEVG KRKLEGDLKLTQESIMDLENDKQQL
Camel SHLVEALYLVCGERG NVLHSAFEVG KRKLEGDLKLTQESIMDLENDKQQL
Rabbit SHLVEALYLVCGERG NVLHSAFEV- KRKLEGDLKLTQES+MDLENDK+QL
Pig SHLVEALYLVCGERG NVLHSAFEVG KRKLEGDLKLTQESIMDLENDKQQL
Alpaca SHLVEALYLVCGERG NVLHSAFEVG KRKLEGDLKLTQESIMDLENDKQQL
Goat SHLVEALYLVCGERG NVLHSAFEVG KRKLEGDLKLTQESIMDLENDKQQL
Sheep SHLVEALYLVCGERG NVLHSAFEVG KRKLEGDLKLTQESIMDLENDKQQL
Cow SHLVEALYLVCGERG NVLHSAFEVG KRKLEGDLKLTQESIMDLENDKQQL
Whale SHLVEALYLVCGERG NVLHSAFEVG KRKLEGDLKLTQESIMDLENDKQQL
Bat SHLVEALYLVCGERG NVLHSAFEVG KRKLEGDLKLTQESIMDLENDKQQL
Chimpanzee SHLVEALYLVCGERG NVLHSAFEVG KRKLEGDLKLTQESIMDLENDKQQL
Human SHLVEALYLVCGERG NVLHSAFEVG KRKLEGDLKLTQESIMDLENDKQQL
Perfect matches are in bold, whereas non-matching sequences are in italics. For the latter, the sequence with highest homology to the queried
sequence is shown.
acquired epidermolysis bullosa 15

Addison's disease
alopecia areata
antiphospholipid syndrome
autoimmune gastritis
autoimmune glomerulonephritis
autoimmune haemolytic anaemia
autoimmune optic neuritis
autoimmune pancreatitis
autoimmune polyendocrine syndrome
autoimmune polyendocrine syndrome type 1
bullous pemphigoid
cicatricial pemphigoid
Crohn's disease 10
cryoglobulinemia
cutaneous lupus erythematosus
demyelinating polyneuropathy
dermatomyositis
encephalomyelitis
Guillain-Barre syndrome
hemolytic-uremic syndrome
juvenile ankylosing spondylitis
Lambert-Eaton myasthenic syndrome
non-insulin-dependent diabetes mellitus
paraneoplastic polyneuropathy
pars planitis
pemphigus gestationis 5
psoriatic arthritis
reactive arthritis
relapsing polychondritis
sclerosing cholangitis
sensory neuropathy
stiff-person syndrome
thrombotic thrombocytopenic purpura
thyrotoxic exophthalmos
ulcerative colitis
vasculitis
vitiligo
Vogt-Koyanagi-Harada
Wegener's granulomatosis
ankylosing spondylitis 200

whale
cow
sheep
goat
alpaca
pig
rabbit
camel
bison
chicken
turkey
duck
tilapia
salmon
rice
soybean
quinoa
potato
lettuce
rye
wheat
autoimmune atherosclerosis
autoimmune hepatitis
autoimmune thyroiditis
autoimmune uveitis
autoimmune vasculitis
celiac disease
Goodpasture's syndrome
Grave's disease
juvenile rheumatoid arthritis
mixed connective tissue disease
myasthenia gravis
neuromyelitis optica
pemphigus
peripheral neuropathy
prediabetes syndrome
primary biliary cirrhosis
psoriasis
rheumatic myocarditis
Sjogrens syndrome
systemic scleroderma
800
quinoa
potato
lettuce
rye
wheat
Behcet disease
whale
cow
sheep
goat
alpaca
pig
rabbit
camel
bison
chicken
turkey
duck
tilapia
salmon
rice
soybean
type 1 diabetes 600

multiple sclerosis 400
rheumatoid arthritis 200
systemic lupus erythematosus
whale
cow
sheep
goat
alpaca
pig
rabbit
camel
bison
chicken
turkey
duck
tilapia
salmon
rice
soybean
quinoa
potato
lettuce
rye
wheat
Fig. 3. Autoimmune epitope content in food sources by disease. Three groups of autoimmune diseases (on the y-axis) created based on the
number of known epitopes are displayed, with diet-derived disease-specific epitopes quantified per organism (on the x-axis). Organisms are
aligned in order of decreasing autoimmune epitope content. Red, ‘hot’ (>500 epitopes); orange, ‘warm’ (50–500 epitopes); blue, ‘cold’ (<50
epitopes).
Moreover, some seven amino acid long peptides can stably the 10 605 epitopes previously implicated in human auto-
bind to HLA class I and, curiously, peptides as short as four immune disorders, 2306 are potential HLA class I epitopes
amino acids can also occupy the HLA class I binding groove and 4547 are potential HLA class II epitopes, according to
in pairs, creating neoepitopes and eliciting CD8+ T-cell re- these thresholds. We focused our analysis on four common
sponses (34). HLA class II-restricted epitopes, on the other autoimmune diseases: MS, RA, T1D, and Behcet disease.
hand, usually span 13–20 amino acids in length (35, 36). These diseases have a large number of implicated epitopes
Nevertheless, unlike HLA class I, there is no known limit to (‘hot diseases’, >500 epitopes) and several HLA alleles pre-
the number of amino acids an HLA class II epitope can have, viously shown either to predispose individuals towards (OR
with peptides up to 40 amino acids long reported as natural > 1) or protect against (OR < 1) the development of these
ligands of HLA class II alleles (37). diseases (Fig. 6).
We used an in silico MHC binding prediction tool (24) Overall, organisms with a higher autoimmune epitope con-
(tools.immuneepitope.org) to interrogate 8–14 amino acid tent yielded more epitopes bound to the analyzed HLA al-
long and 11–30 amino acid long autoimmune epitopes leles. Strikingly, HLA alleles associated with a given disease
binding to HLA class I and class II alleles, respectively. Of have more epitopes predicted to bind to them than those
A B
All autoimmune diseases Multiple sclerosis
400
1500
Epitope Count
Epitope Count
# of Epitopes 300 # of Epitopes
400
1000 1500
1250 300
1000 200
200
750
500 100
500
100
0 0
p nd
ne lad ix
ar r
Br w
Es Br ain
la s
In ea r
te rt
Ki tine
Li y
m L er
n g
P Ov le
Pa anc ary
th as
id
ac nx
os ta
liv Re ate
y m
th S d
So mu in
tis e
S sue
om en
Te ch
To id
U Vag us
us
p nd
ne lad ix
ar r
Br w
Es Br ain
la s
H er
te rt
Ki tine
Li y
m L er
no g
P Ov e
Pa anc ary
th as
id
ac nx
os ta
liv Re ate
y m
th S d
So mu in
tis e
Sp sue
om en
Te ch
To id
U Vag us
us
M ode
M de
qu na
qu na
Th stis
U sil
Th stis
U sil
G h t
G ph ast
m de
H e
m de
lb u
lb u
e
op eas
ph un
an
ft scl
ph un
cl
an
ft scl
In ea
Bo B end
ro
Bo B end
ro
c
Ph yro
Ph ro
o
Pr en
ar ctu
Pr en
ar ctu
dd
dd
v
al ag
dn
ra re
Pl ary
n
r
ito
al ag
dn
ra re
Pl ary
n
r
ito
St le
Ap gla
St ple
Ap gla
bi i
bi i
us
us
yr
yr
t
te
te
s
o e
s
gl
gl
l
l
na
na
re
re
oo
oo
Ly
Ly
Ad
Ad
Sa
Sa
Sm
Sm
C D
300
Rheumatoid arthritis Systemic
y lupus erythematosus
100
Epitope Count
Epitope Count
# of Epitopes # of Epitopes
200
300 75 120
250 100
200
150 80
100
50
60
100 50
25
0 0
p nd
ne lad ix
ar r
Br w
Es Br ain
la s
H er
te rt
Ki tine
Li y
m L er
no g
P Ov e
Pa anc ary
th as
id
ac nx
os ta
liv Re ate
y m
th S d
So mu in
tis e
Sp sue
om en
Te ch
To id
U Vag us
us
M de
qu na
Th stis
U sil
p nd
om en
Te ch
To id
U Vag us
us
ne lad ix
ar r
Br w
Es Br ain
la s
H er
te rt
Ki tine
Li y
m L er
n g
Pa Ov e
Pa nc ary
th as
id
ac nx
os ta
liv Re ate
y m
th S d
So mu in
tis e
S sue
M ode
qu na
Th stis
U sil
G ph ast
m de
G ph ast
m de
lb u
ph un
cl
an
ft scl
In ea
lb u
ph un
cl
an
ft scl
Bo B end
ro
In ea
Ph ro
o
Pr en
ar ctu
Bo B end
ro
dd
v
Ph yro
o
Pr en
ar ctu
dd
al ag
dn
ra re
Pl ary
n
r
ito
Ap gla
St le
al ag
dn
ra re
Pl ary
n
r
ito
bi i
Ap gla
St ple
us
bi i
yr
t
te
us
yr
t
te
o e
gl
o e
gl
l
l
na
na
re
re
oo
oo
Ly
Ly
Ad
Sa
Ad
Sa
Sm
Sm
E F
Type 1 diabetes
T Behcet disease
300
150
Epitope Count
300 175
200
200
100 150
125
100
100
100 50
0 0
p nd
ne lad ix
ar r
Br w
Es Br ain
la s
H er
te rt
Ki tine
Li y
m L er
n g
Pa Ov e
Pa nc ary
th as
id
ac nx
os ta
liv Re ate
y m
th S d
So mu in
tis e
S sue
om en
Te ch
To id
U Vag us
us
p nd
ne lad ix
ar r
Br w
Es Br ain
la s
H er
te rt
Ki tine
Li y
m L er
no g
P Ov e
Pa anc ary
th as
id
ac nx
os ta
liv Re ate
y m
th S d
So mu in
tis e
Sp sue
om en
Te ch
To id
U Vag us
us
M ode
M de
qu na
qu na
Th stis
U sil
Th stis
U sil
G ph ast
G ph ast
m de
m de
lb u
lb u
e
ph un
cl
an
ft scl
ph un
cl
an
ft scl
In ea
In ea
Bo B end
ro
Bo B end
ro
Ph yro
Ph ro
o
Pr en
ar ctu
Pr en
ar ctu
dd
dd
v
dn
al ag
n
al ag
ra re
Pl ary
n
r
ito
dn
ra re
Pl ary
ito
Ap gla
St ple
Ap gla
St le
bi i
bi i
us
us
yr
yr
t
te
te
o e
o e
s
gl
gl
l
l
na
na
re
re
oo
oo
Ly
Ly
Ad
Ad
Sa
Sa
Sm
Sm
Fig. 4. Tissue expression pattern of human autoimmune epitopes. Expression by organ of epitopes implicated in (A) human autoimmune
disease, (B) MS, (C) RA, (D) SLE, (E) T1D and (F) Behcet disease.
A B
All organisms Cow
7500 600
Epitope Count
700
8000 600
5000 400
6000 500
400
4000 300
2500 200
0 0
id
p nd
ne lad ix
Ki tine
Li y
m L er
no g
P Ov le
Pa anc ary
th as
ac nx
os ta
liv Re ate
y m
th S d
So mu in
tis e
S sue
om en
Te ch
To id
U Vag us
us
ar r
Br w
Es Br ain
la s
In ea r
te rt
p nd
ne lad ix
ar r
Br w
Es Br ain
la s
H er
te rt
Ki ine
Li y
m L er
n g
P Ov e
Pa anc ary
th as
id
ac nx
os ta
liv Re ate
y m
th S d
So mu in
tis e
S sue
om en
Te ch
To id
U Vag us
us
M de
M ode
qu na
qu na
Th stis
U sil
Th stis
U sil
G h t
G h t
m de
H e
m de
lb u
lb u
e
op eas
ph un
an
ft scl
op eas
ph un
cl
an
ft scl
In ea
Bo B end
ro
Bo B end
ro
c
Ph yro
Ph yro
o
Pr en
ar ctu
Pr en
ar ctu
dd
dd
v
al ag
dn
ra re
Pl ary
n
r
ito
al ag
dn
ra re
Pl ary
n
r
ito
St ple
Ap gla
Ap gla
St ple
bi i
bi i
us
us
yr
st
yr
t
te
te
s
gl
gl
al
al
n
n
re
re
oo
oo
Ly
Ly
Ad
Ad
Sa
Sa
Sm
Sm
C D
Pig 300
Chicken
600
200
300
700
600 250
400
500 200
400 150
300
100 100
200
0 0
p nd
ne lad ix
ar r
Br w
Es Br ain
la s
H er
te rt
Ki tine
Li y
m L er
n g
P Ov e
Pa anc ary
th as
id
ac nx
os ta
liv Re ate
y m
th S d
So mu in
tis e
S sue
om en
Te ch
To id
U Vag us
us
p nd
ne lad ix
ar r
Br w
Es Br ain
la s
H er
te rt
Ki ine
Li y
m L er
n g
P Ov e
Pa anc ary
th as
id
ac nx
os ta
liv Re ate
y m
th S d
So mu in
tis e
S sue
om en
Te ch
To id
U Vag us
us
M ode
M ode
qu na
qu na
Th stis
Th stis
U sil
U sil
G ph ast
G h t
m de
m de
lb u
lb u
e
ph un
cl
an
ft scl
op eas
ph un
cl
an
ft scl
In ea
In ea
Bo B end
ro
Bo B end
ro
Ph yro
Ph yro
o
o
Pr en
ar ctu
Pr en
ar ctu
dd
dd
v
al ag
dn
ra re
Pl ary
al ag
dn
ra re
Pl ary
n
r
ito
a
n
r
ito
Ap gla
St ple
Ap gla
St ple
bi i
bi i
us
us
st
yr
yr
t
te
te
o e
gl
gl
l
l
na
na
re
re
oo
oo
Ly
Ly
Ad
Ad
Sa
Sa
Sm
Sm
E F
Salmon Rice
40
150
Epitope Count
# of Epitopes 30 # of Epitopes
40
100 150
30
20
100 20
50
10
0 0
p nd
ne lad ix
ar r
Br w
Es Br ain
la s
H er
te rt
Ki tine
Li y
m L er
n g
P Ov e
Pa anc ary
th as
id
ac nx
os ta
liv Re ate
y m
th S d
So mu in
tis e
S sue
om en
Te ch
To id
U Vag us
us
p nd
ne lad ix
ar r
Br w
Es Br ain
la s
H er
te rt
Ki tine
Li y
m Lu er
no g
P Ov e
Pa anc ary
th as
id
ac nx
os ta
liv Re ate
y m
th S d
So mu in
tis e
S sue
om en
Te ch
To id
U Vag us
us
M ode
M de
qu na
qu a
Th stis
U sil
Th stis
U sil
G ph ast
G ph st
m de
m de
lb u
lb u
e
ph un
cl
an
ft scl
ph n
cl
an
ft scl
In ea
In ea
bi in
Bo B end
ro
Bo B end
ro
Ph yro
Ph yro
o
Pr en
ar ctu
Pr en
ar ctu
dd
dd
v
al ag
dn
ra re
Pl ary
n
r
ito
o ea
al ag
dn
ra re
Pl ary
n
r
ito
Ap gla
St ple
Ap gla
St ple
bi i
us
us
yr
yr
t
te
te
o e
s
gl
gl
l
l
na
na
re
re
oo
oo
Ly
Ly
Ad
Ad
Sa
Sa
Sm
Sm
Fig. 5. Tissue expression pattern of food-derived human autoimmune epitopes. Expression by organ of epitopes present in (A) all organisms
commonly consumed as food analyzed, (B) cow, (C) pig, (D) chicken, (E) salmon and (F) rice.
found to be protective against that same disease (Fig. 6), HLA-DRB1*04:05 are also risk alleles for RA (25) (with OR
consistent with the notion that some HLA alleles are proof 2.80 and 3.27, respectively) but bind to visibly less diet-
tective because they do not present pathogenic epitopes as derived epitopes than HLA-DRB1*04:01 (Fig. 6B).
frequently as other HLA alleles, especially those associated Unlike most autoimmune diseases, Behcet disease has
with the disease. been mostly associated with HLA class I alleles, chiefly
Indeed, for MS, HLA-DRB1*15:01 [associated (27, HLA-B*51:01 (40, 41). Of note, Behcet disease is still
38), OR = 3.10] is predicted to present up to 20 epitopes target of debate with regards to whether it is an inflamma-
present in all red meat animals other than camel, whereas tory or an autoimmune disease. Nevertheless, several self-
HLA-DRB1*14:01 [protective (27, 39), OR = 0.54] presents antigens have been implicated in Behcet disease, such
almost none (Fig. 6A). For RA, HLA-DRB1*04:01, an allele as heat shock proteins, S-arrestin and alpha-tropomyosin.
associated (25) with the disease (OR = 3.31) presents the Infections are thought to trigger the onset of disease, at
most peptides, whereas the protective (25) alleles HLA- least in part due to molecular mimicry, given the existence
DRB1*07:01 (OR = 0.66) and HLA-DRB1*15:01 (OR = 0.87) of shared epitopes between bacterial and human heat
present the least (Fig. 6B). Of note, HLA-DRB1*04:04 and shock proteins (42). Curiously, there is a low activity variant
A B
Multiple sclerosis Rheumatoid arthritis
C D
Behcet disease Type 1 diabetes
E F
Type 1 diabetes Type 1 diabetes
Fig. 6. Differential binding of diet-derived autoimmune epitopes to disease-associated HLA alleles. Binding of epitopes implicated in four ‘hot’
diseases, i.e. MS (A), RA (B), Behcet disease (C) and T1D (D–F), to HLA alleles either associated with or protective against each disease was
predicted in silico using tools.immuneepitope.org. Each circle represents the autoimmune epitopes present in a species commonly consumed
as food associated with the disease (x-axis). The size of the circle represents the number of total epitopes contained in the organism that have
been associated with the disease (right hand ‘total epitopes’ legend), whereas the color represents the number of those epitopes predicted to
bind to each HLA allele along the y-axis (right hand ‘binding epitopes’ heatmap legend).
of endoplasmic reticulum aminopeptidase 1 (ERAP1), a different OR (45) (10.30 and 0.19, respectively), yet neither
key enzyme in protein antigen processing, associated with bind diet-derived T1D epitopes (Fig. 6D). As for class II
Behcet disease in epistasis with HLA-B*51:01. Studies alleles, epitope binding also does not always follow a pat-
dissecting the repertoire of peptides presented by HLA- tern consistent with the concept that associated HLA al-
B*51:01 in cells with the low activity ERAP1 mutant found leles bind disease epitopes, whereas protective alleles do
in Behcet disease are sources for implicated autoimmune not: while neither HLA-DRB1*15:01 nor HLA-DQA1*01:02/-
epitopes in www.iedb.org (43, 44). In agreement with this DQB1*06:02 bind to diet-derived T1D epitopes, agreeing
fact and human genetic association studies, we found with the OR of 0.03 for the HLA-DRB1*15:01–HLA-
that HLA-B*51:01(OR = 5.78) is predicted to bind a sig- DQA1*01:02–HLA-DQB1*06:02 haplotype (45), HLA-
nificant number of disease epitopes, whereas HLA-A*03:01 DRB1*14:01 and HLA-DQA1*01:01/-DQB1*05:03 bind to
(OR = 0.6) is not (Fig. 6C). many epitopes, yet the OR for the corresponding haplotype
Finally, we investigated diet-derived epitope binding to (45) is 0.02 (Fig. 6E and F). Nevertheless, for three haplo-
three categories of HLA alleles relevant in T1D biology: types associated with the disease (45), HLA-DRB1*04:05–
HLA-A and HLA-B, HLA-DRB, and HLA-DQA/HLA-DQB. HLA-DQA1*03:01–HLA-DQB1*03:02 (OR = 11.37),
With regards to class I alleles, the T1D-associated allele HLA-DRB1*04:01–HLA-DQA1*03:01–HLA-DQB1*03:02
(45) HLA-A*02:01 (OR = 1.35) bound the most epitopes, (OR = 8.39) and HLA-DRB1*03:01–HLA-DQA1*05:01–
with whale and rabbit containing dramatically more HLA-DQB1*02:01 (OR = 3.64), at least one of the HLA al-
binding autoimmune epitopes than other red meat sources. leles was found to bind a large number of T1D epitopes
Unexpectedly, HLA-B*39:06 and HLA-B*57:01 have vastly (Fig. 6E and F).
Interestingly, we found a weak, positive correlation be- of those epitopes have been implicated in. Focusing on the
tween the total number of disease-associated epitopes ‘hot’ diseases (>500 epitopes), we found that HLA-B*38:06
bound to each HLA and its corresponding OR for RA (Fig. 7A, and HLA-C*12:02 only bind robustly to Behcet disease
n = 9, P = 0.034). Yet, no statistically significant correlation epitopes. HLA-A*02:01 binds to MS and T1D epitopes, as ex-
was observed for T1D or MS in the same analysis (Fig. 7A; pected, yet mostly only to those found in whale and rabbit in
Supplementary Table S2), as well as no statistically signifi- the case of T1D. For HLA class II, all analyzed alleles bind
cant difference when comparing the number of binding to MS epitopes, with HLA-DRB1*04:01 also binding to RA
epitopes between HLA alleles that are predisposing towards and T1D epitopes (Fig. 8). Inspecting binding to epitopes
versus protective against each of the three diseases (Fig. 7B; found in ‘warm’ diseases (between 50 and 500 epitopes),
Supplementary Table S2). we found that HLA-DRB1*04:01 binds to epitopes from my-
These observations led us to develop a personalized asthenia gravis, Sjogren’s syndrome and systemic sclero-
scorecard to determine food autoimmunogenicity for indi- derma, whereas HLA-DRB1*03:01 and HLA-DRB1*15:01
vidual patients—the Gershteyn-Ferreira sensitivity passport. are only likely to present systemic scleroderma epitopes
As a proof-of-principle, we analyzed six (out of a theoretical (Supplementary Figure S3).
maximum of nine) pairs of HLA alleles taken from a pool of
de-identified individual’s haplotypes for binding to auto-
Discussion
immune epitopes found in different organisms commonly
consumed as food, and determined which diseases most Autoimmunity has been on the rise around the globe at a fast
pace, leading many to believe that environmental factors, not
A genetics, are mostly responsible for this trend. Processed
180 Type 1 diabetes foods play a large role in the American diet and have been
Rheumatoid arthritis shown to compromise intestinal barrier integrity (17). In the
Multiple sclerosis
160 USA, while processed meat consumption has remained
stable over the past two decades, total meat consumption
140 has decreased (46). Therefore, the proportion of processed
meat in American diets has increased.
Binding epitopes
120 Orally administered antigens interact with the immune system

in the mesenteric lymph nodes, typically leading to the gener-
100
ation of antigen-specific regulatory T cells (Tregs), a subset of T
lymphocytes dedicated to maintaining homeostasis by limiting
80
immune responses (47), and so called oral tolerance (48–50).
60
However, oral tolerance can be compromised. For instance,
mechanical injury to the skin promotes anaphylaxis, an acute
40 allergic reaction, to oral antigens: injury-activated keratinocytes
in the skin released IL-33 systemically, ultimately activating gut
mast cells and increasing small intestine permeability (51).
Formation of food antigen–immunoglobulin complexes, con-
taining either IgG or IgE, has been described in patients, with
B
Type 1 diabetes Rheumatoid arthritis Multiple sclerosis the deposition of these complexes in tissues causing local
100
damage and inflammation (52–54).
Celiac disease, an autoimmune disease where the lining of
90 the small intestine is destroyed, has a well-defined interaction
80
with diet and specific HLA alleles. Partly undigested gluten
Binding epitopes
peptides cross the intestinal barrier and are modified by

70 tissue transglutaminase 2 (TG2), which generates epitopes
60
that are then presented to autoreactive gluten-specific T cells
(20, 55). Over 80% of celiac disease patients carry HLA-DQ2
50 (HLA-DQA1*05:01/-DQB1*02:01, OR = 15.1) (56, 57). Gluten-
40
specific Tregs have been found to have a defective sup-
pressive capacity in celiac disease patients (58). Removing
30 gluten from the diet leads to remission of the disease. Can we
work towards determining the impact of diet on other auto-
Fig. 7. Association between number of bound autoimmune epitopes
immune diseases with the same degree of mechanistic de-
and disease OR among HLA molecules. (A) Correlation analysis be- tail? We predict that foods, containing epitopes implicated in
tween number of binding epitopes of OR for different HLA alleles a certain autoimmune disease, can be triggers to exacerbate
associated with an autoimmune disorder: T1D (blue), RA (orange) its symptoms upon frequent consumption.
or MS (green). The Pearson r value of each linear regression is dis- One limitation of the present study lies in the fact that only
played next to the respective curve. (B) Number of binding epitopes
for predisposing vs. protective alleles for T1D (blue), RA (orange) and unmodified linear peptide epitopes were considered. Hence,
MS (green). The box and whisker plot represents mean and standard post-translational modifications, such as citrullination (59), and
deviation, with outliers indicated by single dots outside each box. hybrid peptides, such as the hybrid insulin peptides (HIPs)
HLA-A*02:01 HLA-A*03:01
Type 1 diabetes
Rheumatoid arthritis
Behcet disease
Systemic lupus erythematosus
Multiple sclerosis
HLA-B*07:02 HLA-B*38:06
Type 1 diabetes
Behcet disease
Multiple sclerosis
HLA-C*05:01 HLA-C*12:02
Type 1 diabetes
Behcet disease
Multiple sclerosis
HLA-DRB1*03:01 HLA-DRB1*04:01
Type 1 diabetes
Behcet disease
Multiple sclerosis
HLA-DQA1*01:01-DQB1*05:01 HLA-DQA1*01:01-DQB1*06:01
Type 1 diabetes
Behcet disease
Multiple sclerosis
Fig. 8. Gershteyn-Ferreira sensitivity passport. Binding of total autoimmune epitopes present in food to 10 HLA alleles (6 HLA class I and 4
HLA class II alleles, with HLA-DQA and HALA-DQB being combined—alpha and beta chain of the HLA-DQ molecule) from one individual was
predicted in silico using tools.immuneepitope.org. Each circle represents the autoimmune epitopes present in a species commonly consumed
as food (x-axis) implicated in a human autoimmune disease (y-axis). The size of the circle represents the number of total epitopes contained in
the organism that have been associated with the disease (right hand ‘total epitopes’ legend), whereas the color represents the number of those
epitopes predicted to bind to the HLA allele (right hand ‘binding epitopes’ heatmap legend).
created by linkage of an insulin fragment to another peptide 2 Cooper, G. S., Miller, F. W. and Pandey, J. P. 1999. The role of
in beta cells (60), previously shown to be implicated in auto- genetic factors in autoimmune disease: implications for environ-
mental research. Environ. Health Perspect. 107(Suppl. 5):693.
immunity, were not analyzed. Curiously, MHC ligand assay- 3 Vojdani, A. 2014. A potential link between environmental triggers
derived implicated autoimmune epitopes were at most ~10% and autoimmunity. Autoimmune Dis. 2014:437231.
of total epitopes, except for one disorder: Behcet disease. In 4 Malosse, D. and Perron, H. 1993. Correlation analysis between
that instance, removing MHC ligand assay-derived epitopes bovine populations, other farm animals, house pets, and multiple
brought the number of Behcet disease epitopes down to 28. sclerosis prevalence. Neuroepidemiology 12:15.
5 Kjeldsen-Kragh, J., Haugen, M., Borchgrevink, C. F. et al. 1991.
Of these, only 10 were in the length range for binding to HLA Controlled trial of fasting and one-year vegetarian diet in rheuma-
class I and none of them are predicted to bind to any of three toid arthritis. Lancet 338:899.
Behcet disease-associated HLA alleles surveyed (Fig. 6C). 6 Bock, S. A. and Atkins, F. M. 1990. Patterns of food hypersen-
Future studies focused on improving our capacity to better sitivity during sixteen years of double-blind, placebo-controlled
food challenges. J. Pediatr. 117:561.
discern the relative importance of disease-implicated self- 7 Kleinewietfeld, M., Manzel, A., Titze, J. et al. 2013. Sodium
antigens found in large databases will be of value. Moreover, chloride drives autoimmune disease by the induction of patho-
food production and consumption practices can significantly genic TH17 cells. Nature 496:518.
affect the specific protein content of the final product. For in- 8 Wu, C., Yosef, N., Thalhamer, T. et al. 2013. Induction of patho-
stance, proteins found only in the breast muscle of fast-growth genic TH17 cells by inducible salt-sensing kinase SGK1. Nature
496:513.
chickens have been identified (61), whereas cooking food 9 Kawanishi, K., Dhar, C., Do, R., Varki, N., Gordts, P. L. S. M.
can alter their antigen composition and their impact on the and Varki, A. 2019. Human species-specific loss of CMP-N-
gut microbiome (62). In turn, an individual’s microbiome com- acetylneuraminic acid hydroxylase enhances atherosclerosis via
position can determine his or her propensity to intestinal in- intrinsic and extrinsic mechanisms. Proc. Natl Acad. Sci. USA
116:16036.
flammation and autoimmunity (63), adding an additional layer 10 Rojas, M., Restrepo-Jiménez, P., Monsalve, D. M. et al. 2018.
warranting future investigations. Finally, we predict that epi- Molecular mimicry and autoimmunity. J. Autoimmun. 95:100.
demiological studies will reveal differences in the incidence 11 Gough, S. C. and Simmonds, M. J. 2007. The HLA region and
and/or severity of certain autoimmune disorders in popula- autoimmune disease: associations and mechanisms of action.
tions with specific dietary habits, namely exclusion of one or Curr. Genomics 8:453.
12 Shahrizaila, N. and Yuki, N. 2011. Guillain-Barré syndrome animal
more foods, and that studies using humanized animals ex- model: the first proof of molecular mimicry in human autoimmune
pressing different HLA alleles subjected to dietary antigen ex- disorder. J. Biomed. Biotechnol. 2011:829129.
posure regimens will refine our mechanistic understanding of 13 Gil-Cruz, C., Perez-Shibayama, C., De Martin, A. et al. 2019.

the interaction between diet and autoimmunity. Microbiota-derived peptide mimics drive lethal inflammatory car-
diomyopathy. Science 366:881.
In summary, we systematically dissected the 14 Hvatum, M., Kanerud, L., Hällgren, R. and Brandtzaeg, P. 2006.
autoimmunogenicity of 24 organisms, 22 of which are com- The gut-joint axis: cross reactive food antibodies in rheumatoid
monly consumed as food (Fig. 1). We mapped not only their arthritis. Gut 55:1240.
content of epitopes previously implicated in human auto- 15 Lachance, D. H., Lennon, V. A., Pittock, S. J. et al. 2010. An out-
immune disorders (Fig. 2), but also assessed the capacity of break of neurological autoimmunity with polyradiculoneuropathy
in workers exposed to aerosolised porcine neural tissue: a de-
different HLA alleles to present these peptides. This research scriptive study. Lancet Neurol. 9:55.
revealed striking differences in binding, hence potential 16 Bischoff, S. C., Barbara, G., Buurman, W. et al. 2014. Intestinal
food autoimmunogenicity, across HLA alleles (Fig. 6), an im- permeability—a new target for disease prevention and therapy.
portant step towards personalizing this information for each BMC Gastroenterol. 14:189.
17 Lerner, A. and Matthias, T. 2015. Changes in intestinal tight junction
autoimmune disease patient (Gershteyn-Ferreira Sensitivity permeability associated with industrial food additives explain the
Passport, Fig. 8). Overall, our platform will help shed light on rising incidence of autoimmune disease. Autoimmun. Rev. 14:479.
the often observed yet seldom understood impact of diet on 18 Purohit, V., Bode, J. C., Bode, C. et al. 2008. Alcohol, intestinal
autoimmune disorders. bacterial growth, intestinal permeability to endotoxin, and med-
ical consequences: summary of a symposium. Alcohol 42:349.
19 Mu, Q., Kirby, J., Reilly, C. M. and Luo, X. M. 2017. Leaky gut as
Funding a danger signal for autoimmune diseases. Front. Immunol. 8:598.
This work was funded by Ajax Biomedical Foundation (Newton, MA). 20 Verdu, E. F. and Danska, J. S. 2018. Common ground: shared
risk factors for type 1 diabetes and celiac disease. Nat. Immunol.
19:685.
Acknowledgements 21 Erickson, M. A. and Banks, W. A. 2019. Age-associated changes
in the immune system and blood(-)brain barrier functions. Int.
We thank Dr Victor Goldmacher, Dr Andrey Vyshedskiy, Mikhail J. Mol. Sci. 20:1–28.
Gershteyn and Arkady Gershteyn for their support, as well as the an- 22 Jarius, S., Paul, F., Franciotta, D. et al. 2008. Mechanisms of
onymous reviewers for their comments and suggestions to improve disease: aquaporin-4 antibodies in neuromyelitis optica. Nat.
the manuscript. The graphical abstract and Fig. 1 of this manuscript Clin. Pract. Neurol. 4:202.
were created with BioRender.com. 23 Vita, R., Mahajan, S., Overton, J. A. et al. 2019. The Immune
Epitope Database (IEDB): 2018 update. Nucleic Acids Res.
Conflicts of interest statement: I.M.G. and L.M.R.F. received funding
47(D1):D339.
from Ajax Biomedical Foundation, a 501(c)(3) corporation. L.M.R.F. is
24 Fleri, W., Paul, S., Dhanda, S. K. et al. 2017. The Immune Epitope
the Jeffrey G. Klein Diabetes Fellow. The other authors declare no
Database and analysis resource in epitope discovery and syn-
conflicts of interest.
thetic vaccine design. Front. Immunol. 8:278.
25 Balandraud, N., Picard, C., Reviron, D. et al. 2013. HLA-DRB1
References genotypes and the risk of developing anti citrullinated protein anti-
body (ACPA) positive rheumatoid arthritis. PLoS ONE 8:e64108.
1 Barabási, A., Menichetti, G. and Loscalzo, J. 2020. The un- 26 Noble, J. A. and Erlich, H. A. 2012. Genetics of type 1 diabetes.
mapped chemical complexity of our diet. Nat. Food 1:33. Cold Spring Harb. Perspect. Med. 2:a007732.
27 Hollenbach, J. A. and Oksenberg, J. R. 2015. The immunogenetics 45 Noble, J. A. and Valdes, A. M. 2011. Genetics of the HLA region
of multiple sclerosis: a comprehensive review. J. Autoimmun. in the prediction of type 1 diabetes. Curr. Diab. Rep. 11:533.
64:13. 46 Zeng, L., Ruan, M., Liu, J. et al. 2019. Trends in processed meat,
28 Virtanen, P., Gommers, R., Oliphant, T. E. et al.; SciPy 1.0
unprocessed red meat, poultry, and fish consumption in the
Contributors. 2020. SciPy 1.0: fundamental algorithms for scien- United States, 1999–2016. J. Acad. Nutr. Diet. 119:1085.
tific computing in Python. Nat. Methods 17:261. 47 Ferreira, L. M. R., Muller, Y. D., Bluestone, J. A. and Tang, Q.
29 Rowntree, L. C., Nguyen, T. H. O., Halim, H. et al. 2018. Inability 2019. Next-generation regulatory T cell therapy. Nat. Rev. Drug
to detect cross-reactive memory T cells challenges the frequency Discov. 18:749.
of heterologous immunity among common viruses. J. Immunol. 48 Esterházy, D., Loschko, J., London, M., Jove, V., Oliveira, T. Y. and
200:3993. Mucida, D. 2016. Classical dendritic cells are required for dietary
30 Vojdani, A. 2019. Reaction of food-specific antibodies with dif- antigen-mediated induction of peripheral T(reg) cells and toler-
ferent tissue antigens. Int. J. Food Sci. Technol. 55:1800. ance. Nat. Immunol. 17:545.
31 Trolle, T., McMurtrey, C. P., Sidney, J. et al. 2016. The length distri- 49 Chen, Y., Kuchroo, V. K., Inobe, J., Hafler, D. A. and Weiner, H. L.
bution of class I-restricted T cell epitopes is determined by both 1994. Regulatory T cell clones induced by oral tolerance: sup-
peptide supply and MHC allele-specific binding preference. J. pression of autoimmune encephalomyelitis. Science 265:1237.
Immunol. 196:1480. 50 Pabst, O. and Mowat, A. M. 2012. Oral tolerance to food protein.
32 Hassan, C., Chabrol, E., Jahn, L. et al. 2015. Naturally processed Mucosal Immunol. 5:232.
non-canonical HLA-A*02:01 presented peptides. J. Biol. Chem. 51 Leyva-Castillo, J. M., Galand, C., Kam, C. et al. 2019. Mechanical
290:2593. skin injury promotes food anaphylaxis by driving intestinal mast
33 Remesh, S. G., Andreatta, M., Ying, G. et al. 2017. Unconventional cell expansion. Immunity 50:1262.
peptide presentation by major histocompatibility complex (MHC) 52 Mayadas, T. N., Tsokos, G. C. and Tsuboi, N. 2009. Mechanisms
class I allele HLA-A*02:01: BREAKING CONFINEMENT. J. Biol. of immune complex-mediated neutrophil recruitment and tissue
Chem. 292:5262. injury. Circulation 120:2012.
34 Xiao, Z., Ye, Z., Tadwal, V. S., Shen, M. and Ren, E. C. 2017. Dual 53 Sheffer, A. L., Lieberman, P. L., Aaronson, D. W. et al. 1988.

non-contiguous peptide occupancy of HLA class I evoke anti- Measurement of circulating IgG and IgE food-immune com-
viral human CD8 T cell response and form neo-epitopes with self- plexes. J. Allergy Clin. Immunol. 81:758.
antigens. Sci. Rep. 7:5072. 54 Severance, E. G., Gressitt, K. L., Halling, M. et al. 2012.

35 Stern, L. J., Brown, J. H., Jardetzky, T. S. et al. 1994. Crystal struc- Complement C1q formation of immune complexes with milk
ture of the human class II MHC protein HLA-DR1 complexed with caseins and wheat glutens in schizophrenia. Neurobiol. Dis.
an influenza virus peptide. Nature 368:215. 48:447.
36 O’Brien, C., Flower, D. R. and Feighery, C. 2008. Peptide length 55 Molberg, O., Mcadam, S. N., Körner, R. et al. 1998. Tissue

significantly influences in vitro affinity for MHC class II molecules. transglutaminase selectively modifies gliadin peptides that are rec-
Immunome Res. 4:6. ognized by gut-derived T cells in celiac disease. Nat. Med. 4:713.
37 van Lummel, M., van Veelen, P. A., de Ru, A. H. et al. 2016. 56 Mubarak, A., Spierings, E., Wolters, V., van Hoogstraten, I.,

Dendritic cells guide islet autoimmunity through a restricted and Kneepkens, C. M. and Houwen, R. 2013. Human leukocyte
uniquely processed peptidome presented by high-risk HLA-DR. antigen DQ2.2 and celiac disease. J. Pediatr. Gastroenterol. Nutr.
J. Immunol. 196:3253. 56:428.
38 Patsopoulos, N. A., Barcellos, L. F., Hintzen, R. Q. et al.; IMSGC; 57 Khosravi, A., Mansouri, M., Rostami-Nejad, M., Shahbazkhani, B.,
ANZgene. 2013. Fine-mapping the genetic association of the Ekhlasi, G. and Kalantari, E. 2016. The likelihood ratio and fre-
major histocompatibility complex in multiple sclerosis: HLA and quency of DQ2/DQ8 haplotypes in Iranian patients with celiac
non-HLA effects. PLoS Genet. 9:e1003926. disease. Gastroenterol. Hepatol. Bed Bench. 9:18.
39 Toro, J., Cuellar-Giraldo, D., Díaz-Cruz, C. et al. 2016. HLA-
58 Cook, L., Munier, C. M. L., Seddiki, N. et al. 2017. Circulating
DRB1*14 is a protective allele for multiple sclerosis in an admixed gluten-specific FOXP3+CD39+ regulatory T cells have impaired
Colombian population. Neurol. Neuroimmunol. Neuroinflamm. suppressive function in patients with celiac disease. J. Allergy
3:e192. Clin. Immunol. 140:1592.
40 Hughes, T., Coit, P., Adler, A. et al. 2013. Identification of mul- 59 Mathsson, L., Mullazehi, M., Wick, M. C. et al. 2008. Antibodies
tiple independent susceptibility loci in the HLA region in Behçet’s against citrullinated vimentin in rheumatoid arthritis: higher sen-
disease. Nat. Genet. 45:319. sitivity and extended prognostic value concerning future radio-
41 Ombrello, M. J., Kirino, Y., de Bakker, P. I., Gul, A., Kastner, D. L. graphic progression as compared with antibodies against cyclic
and Remmers, E. F. 2014. Behcet disease-associated MHC citrullinated peptides. Arthritis Rheum. 58:36.
class I residues implicate antigen binding and regulation of 60 Wiles, T. A., Powell, R., Michel, R. et al. 2019. Identification of hy-
cell-mediated cytotoxicity. Proc. Natl Acad. Sci. USA 111:8867. brid insulin peptides (HIPs) in mouse and human islets by mass
42 Tong, B., Liu, X., Xiao, J. and Su, G. 2019. Immunopathogenesis spectrometry. J. Proteome Res. 18:814.
of Behcet’s disease. Front. Immunol. 10:665. 61 Phongpa-Ngan, P., Grider, A., Mulligan, J. H., Aggrey, S. E. and
43 Guasp, P., Barnea, E., González-Escribano, M. F. et al. 2017. Wicker, L. 2011. Proteomic analysis and differential expression
The Behçet’s disease-associated variant of the aminopeptidase in protein extracted from chicken with a varying growth rate and
ERAP1 shapes a low-affinity HLA-B*51 peptidome by differential water-holding capacity. J. Agric. Food Chem. 59:13181.
subpeptidome processing. J. Biol. Chem. 292:9680. 62 Carmody, R. N., Bisanz, J. E., Bowen, B. P. et al. 2019. Cooking
44 Guasp, P., Alvarez-Navarro, C., Gomez-Molina, P. et al. 2016. shapes the structure and function of the gut microbiome. Nat.
The peptidome of Behcet’s disease-associated HLA-B*51:01 in- Microbiol. 4:2052.
cludes two subpeptidomes differentially shaped by endoplasmic 63 Zitvogel, L. and Kroemer, G. 2019. Immunostimulatory gut bac-
reticulum aminopeptidase 1. Arthritis Rheumatol. 68:505. teria. Science 366:1077.

Doctor Leonardo Ferreria

Uploaded by

Copyright:

Available Formats

Doctor Leonardo Ferreria

Uploaded by

Document Information

Copyright

Available Formats

Share this document

Share or Embed Document

Sharing Options

Did you find this document useful?

Is this content inappropriate?

Copyright:

Available Formats

Doctor Leonardo Ferreria

Uploaded by

Copyright:

Available Formats

International Immunology © The Japanese Society for Immunology. 2020. All rights reserved.

doi:10.1093/intimm/dxaa054 For permissions, please e-mail: [email protected]

Immunodietica: interrogating the role of diet in

Keywords: alimentation, autoimmunity, epitope, HLA, immunogenetics

400 autoimmune gastritis

Table 1. Interspecies alignment of human autoimmune disease epitopes

Wheat S+LVE+L------RG NV+HS----- ------DLKL------DLE+-----

acquired epidermolysis bullosa 15

ankylosing spondylitis 200

type 1 diabetes 600

120 Orally administered antigens interact with the immune system

peptides cross the intestinal barrier and are modified by

Systemic lupus erythematosus

Systemic lupus erythematosus

Systemic lupus erythematosus

Systemic lupus erythematosus

Systemic lupus erythematosus

You might also like