RESEARCH Open Access
P omo e me hyla ion o he MGAT3
and BACH2 genes co ela es wi h he
composi ion o he immunoglobulin G
glycome in in lamma o y bowel disease
Ma ija Klasić
1†
, Do a Ma kulin
1†
, Aleksanda Voj a
1
, I ana Sama žija
1
, I an Bi uš
1
, Paula Dob inić
1
,
Nicholas T. Ven ham
2
, I ena T boje ić-Akmačić
3
, Mi na Šimu ina
3
, Je ko Š ambuk
3
, Genadij Razdo o
3
,
Nicholas A. Kennedy
2,5
, Jack Sa sangi
2,10
, Ana M. Dias
6
, Salome Pinho
6
, Vi o Annese
7
, Anna La iano
8
,
Rena a D’Inca
9
, IBD conso ium, Go dan Lauc
3,4
and Vla ka Zoldoš
1*
Abs ac
Backg ound: Many genome- and epigenome-wide associa ion s udies (GWAS and EWAS) and s udies o p omo e
me hyla ion o candida e genes o in lamma o y bowel disease (IBD) ha e demons a ed signi ican
associa ions be ween gene ic and epigene ic changes and IBD. Independen GWA s udies ha e iden i ied
gene ic a ian s in he BACH2,IL6ST,LAMB1,IKZF1,andMGAT3 loci o be associa ed wi h bo h IBD and
immunoglobulin G (IgG) glycosyla ion.
Me hods: Using bisul i e py osequencing, we analyzed CpG me hyla ion in p omo e egions o hese i e
genes om pe iphe al blood o se e al hund ed IBD pa ien s and heal hy con ols (HCs) om wo independen coho s,
espec i ely.
Resul s: We ound signi ican di e ences in he me hyla ion le els in he MGAT3 and BACH2 genes be ween bo h
C ohn’s disease and ulce a i e coli is when compa ed o HC. The same pa e n o me hyla ion changes was iden i ied
o bo h genes in CD19
+
B cells isola ed om he whole blood o a subse o he IBD pa ien s. A co ela ion analysis was
pe o med be ween he MGAT3 and BACH2 p omo e me hyla ion and indi idual IgG glycans, measu ed in he same
indi iduals o he wo la ge coho s. MGAT3 p omo e me hyla ion co ela ed signi ican ly wi h galac osyla ion, sialyla ion,
and bisec ing GlcNAc on IgG o he same pa ien s, sugges ing ha ac i i y o he GnT-III enzyme, encoded by his gene,
migh be al e ed in IBD. The co ela ions be ween he BACH2 p omo e me hyla ion and IgG glycans we e less ob ious,
since BACH2 is no a glycosyl ans e ase and he e o e may a ec IgG glycosyla ion only indi ec ly.
Conclusions: Ou esul s sugges ha epigene ic de egula ion o key glycosyla ion genes migh lead o an inc ease in
p o-in lamma o y p ope ies o IgG in IBD h ough a dec ease in galac osyla ion and sialyla ion and an inc ease o
bisec ing GlcNAc on digalac osyla ed glycan s uc u es. Finally, we showed ha CpG me hyla ion in he p omo e o he
MGAT3 gene is al e ed in CD3
+
T cells isola ed om in lamed mucosa o pa ien s wi h ulce a i e coli is om a hi d
smalle coho , o which biopsies we e a ailable, sugges ing a unc ional ole o his glyco-gene in IBD pa hogenesis.
* Co espondence: [email p o ec ed]
†
Ma ija Klasićand Do a Ma kulin con ibu ed equally o his wo k.
1
Depa men o Biology, Di ision o Molecula Biology, Facul y o Science,
Uni e si y o Zag eb, Ho a o ac 102a, 10000 Zag eb, C oa ia
Full lis o au ho in o ma ion is a ailable a he end o he a icle
© The Au ho (s). 2018 Open Access This a icle is dis ibu ed unde he e ms o he C ea i e Commons A ibu ion 4.0
In e na ional License (h p://c ea i ecommons.o g/licenses/by/4.0/), which pe mi s un es ic ed use, dis ibu ion, and
ep oduc ion in any medium, p o ided you gi e app op ia e c edi o he o iginal au ho (s) and he sou ce, p o ide a link o
he C ea i e Commons license, and indica e i changes we e made. The C ea i e Commons Public Domain Dedica ion wai e
(h p://c ea i ecommons.o g/publicdomain/ze o/1.0/) applies o he da a made a ailable in his a icle, unless o he wise s a ed.
Klasiće al. Clinical Epigene ics (2018) 10:75
h ps://doi.o g/10.1186/s13148-018-0507-y
Backg ound
In lamma o y bowel disease (IBD) is a ch onic in es inal
in lamma o y condi ion classi ied in wo majo o ms—
C ohn’s disease (CD) and ulce a i e coli is (UC)—which
exhibi e iologically and clinically dis inc ea u es. Now-
adays, IBD a ec s 2.5–3 million people in Eu ope and
causes conside able mo bidi y [1]. Despi e nume ous
clinical, gene ic, and o he expe imen al s udies, ou
unde s anding o IBD de elopmen and p og ession
emains incomple e.
I is gene ally accep ed ha IBD ep esen s an abe -
an immune esponse o gu mic obio a in gene ically
suscep ible indi iduals [2]. Genome-wide associa ion
s udies (GWAS) ha e iden i ied o e 200 gene ic suscep-
ibili y loci, he majo i y o which we e associa ed wi h
bo h o ms o IBD in genome-wide me a-analysis [3–7].
Howe e , common gene ic a ian s accoun only o 8.2
and 13.1% he i abili y o UC and CD, espec i ely [7].
In e ac iono anindi idual’s gu mic obiome, im-
mune sys em, gene ic backg ound, and en i onmen al
ac o s, such as smoking, die , d ugs, and physical
ac i i y [2,8–10], makes IBD a complex e iopa ho-
genic en i y. The challenge is he e o e o iden i y
addi ional ac o s in ol ed in he de elopmen and
p og ession o his disease, especially gi en i s apidly
inc easing incidence. I is p obable ha epigene ics
play a key ole in he in e ac ions be ween en i on-
men al, mic obial, and gene ic ac o s ha pa icipa e
in IBD de elopmen and p og ession. These include
DNA me hyla ion and his one modi ica ions, as well
as some o he epigene ic mechanisms [11–13]; o a
e iew, see [14,15].
DNA me hyla ion emains he mos s udied epigene ic
modi ica ion, eadily assayed in a la ge numbe o indi-
iduals/samples. Hype me hyla ion o gene p omo e s is
gene ally associa ed wi h gene silencing, while p omo e
hypome hyla ion is associa ed wi h gene ac i a ion [16].
En i onmen ally changed DNA me hyla ion pa e n may
con ibu e o he de elopmen o many complex diseases
by media ing he in e play be ween ex e nal and in e nal
ac o s and he gene exp ession [17–21]. The e a e also
da a o sugges ha he a o emen ioned en i onmen-
al modi ie s o IBD can also a ec DNA me hyla ion
[17–19,22]. Epigene ic componen o IBD has been
add essed in many s udies, mos ly by whole genome
me hyla ion analysis pe o med on pe iphe al blood
mononuclea cells (PBMCs) o mucosal issue, e eal-
ing egions di e en ially me hyla ed be ween he
diseaseandheal hys a e,aswellasbe weenCDand
UC [11–13,23–26].
The majo i y o euka yo ic p o eins a e modi ied by
addi ion o complex oligosaccha ides (glycans) h ough
he p ocess o glycosyla ion. The e o e, glycans a e an
in eg al pa o nea ly all memb ane and sec e ed
p o eins, including componen s o he immune sys em
[27]. Abe an p o ein glycosyla ion is implica ed in
i ually e e y human complex disease, including in lam-
ma ion [28–31]. P e ious s udies ha e sugges ed ha
N-glycosyla ion o sec e ed and memb ane p o eins
migh be egula ed epigene ically and ha abe an gly-
cosyla ion p o iles in disease can a ise h ough abe an
epigene ics [32–38]. A comp ehensi e e iew abou he
ole o p o ein glycosyla ion in IBD has been gi en
ecen ly [39]. N-glycosyla ion o se um-ci cula ing p o-
eins (such as he acu e phase p o eins; immunoglobulin
G, IgG; and immunoglobulin A, IgA) o whole plasma
N-glycome (i.e., N-glycans p esen on all plasma p o-
eins) has been he ocus o IBD bioma ke disco e y
[36,40–43]. In addi ion, ou pa ne s om IBD conso -
ium and o he s es ablished ha al e ed glycosyla ion o
IgG, which is a key e ec o o he humo al immune
sys em, has a ole in balancing in lamma ion a he sys-
emic le el [42–46].
GWA s udies indica ed associa ions o IBD wi h se -
e al loci in ol ed in p o ein glycosyla ion [47,48]. Mo e
ecen ly, he i s GWAS o IgG glycosyla ion iden i ied
16 loci speci ically associa ed wi h changes in IgG glyco-
syla ion [49]. In e es ingly, i e o hese loci showed plei-
o opy wi h IBD: MGAT3, a glyco-gene encoding o a
glycosyl ans e ase, GnT-III; LAMB1, a membe o
ansmemb ane glycop o ein amily o ex acellula
ma ix; he IL6ST, a signal ansduce sha ed by many
cy okines; IKZF1; and BACH2, ansc ip ion ac o s in-
ol ed in B cell di e en ia ion, ac i a ion, and ma u -
a ion. Only he MGAT3 is a classical glyco-gene wi h a
known unc ion in IgG glycosyla ion, while he exac
unc ional oles o o he ou GWAS hi s in IgG glyco-
syla ion o IBD emain unknown.
In his s udy, we in es iga ed p omo e me hyla ion
di e ences in hese i e genes, associa ed wi h bo h IBD
and IgG glycosyla ion, in pe iphe al whole blood o
se e al hund ed IBD pa ien s om wo independen
coho s. We also co ela ed p omo e me hyla ion da a
wi h IgG glycosyla ion da a analyzed p e iously o he
same IBD pa ien s by ou pa ne s om he IBD conso -
ium [43,46,50]. Pe iphe al blood was used o DNA
me hyla ion analysis and se um o plasma was used o
glycan analysis, since one o ou goals was he sea ch o
po en ial IBD bioma ke s. As pe iphe al whole blood is
a he e ogeneous cell mix u e wi h speci ic me hyla ion
pa e n o each o he cell ypes [51], we also analyzed
p omo e me hyla ion o ou candida e genes in CD19
+
B cells and CD3
+
T cells isola ed om pe iphe al blood
mononuclea cells (PBMCs). B cells we e o ou pa icu-
la in e es since hese cells p oduce IgG on hei mem-
b ane and a e p ecu so s o plasma cells which sec e e
IgG. We ha e u he explo ed i abe an p omo e
me hyla ion eco ded in pe iphe al whole blood o IBD
Klasiće al. Clinical Epigene ics (2018) 10:75 Page 2 o 14
pa ien s can be a p oxy o epigene ic e en s occu ing
in he in lamed mucosa. To add ess his ques ion, we
analyzed DNA me hyla ion om PBMCs, CD3
+
T cells
isola ed om PBMCs, and CD3
+
T cells isola ed om
in lamed colonic mucosa o UC pa ien s om he hi d
smalle coho , o which biopsies we e a ailable.
Me hods
Pa ien selec ion and e hics
Pa ien s we e ec ui ed p ospec i ely om Edinbu gh,
UK, and Flo ence, I aly, as a pa o he IBD-BIOM p o-
jec . The ec ui men o pa ien s om Edinbu gh has been
desc ibed elsewhe e [13,43]. B ie ly, we ec ui ed IBD
pa ien s p ospec i ely as close as possible o he da e o
diagnosis om gas oen e ology ou pa ien and endoscopy
appoin men s be ween 2012 and 2015. We ec ui ed
symp oma ic con ols om gas oen e ology clinics du -
ing he same pe iod. In hese indi iduals, we had excluded
IBD and o he o ganic bowel pa hology ollowing bio-
chemical and/o endoscopic in es iga ions. We ec ui ed
a u he heal hy olun ee coho wi h no gas oin es inal
symp oms. IBD pa ien s we e s a i ied by disease ype
(ulce a i e coli is, UC, and C ohn’s disease, CD). De ailed
gene ic, pheno ypic, and o he da a ega ding IBD cases
a e gi en in Addi ional ile 1: Tables S1 and S2. Flo ence
coho was collec ed h ough he ne wo k o he I alian
G oup o IBD (IG-IBD) since he beginning o 2001 and
i s desc ibed in 2005 [1] ollowing an in e nal alida ion
o pheno yping. Subsequen ly, longi udinal upda e has
been pe o med on a yea ly basis.
E hical app o als we e ob ained om Tayside Com-
mi ee on Medical E hics B, and all pa ien s and con ols
p o ided w i en, in o med consen (LREC 06/S1101/16,
LREC 2000/4/192).
Flo ence ec ui men de ails
IBD pa ien s we e p ospec i ely ec ui ed as close as
possible o he da e o diagnosis om gas oen e ology
ou pa ien and endoscopy appoin men s be ween yea s
2012 and 2015 in di e en e ia y e e al cen e s in
San Gio anni Ro ondo, Rome, Rozzano (Milan), Padua,
and Flo ence, I aly. Symp oma ic con ols we e ec ui ed
in he same cen e s (gas oen e ology clinics) du ing he
same pe iod. In hese indi iduals, IBD and o he o ganic
bowel pa hology we e excluded by biochemical and/o
endoscopic in es iga ions. IBD pa ien s we e s a i ied
by disease ype (ulce a i e coli is, UC, and C ohn’s
disease, CD). Samples we e ob ained wi h he same
me hodology (see u he ) and cen ally collec ed a San
Gio anni Ro ondo, I aly.
Sample collec ion
We collec ed whole blood a he ime o pa ien ec ui -
men in o 9-ml se um Z-clo ac i a o ubes (G eine ),
allowed hem o clo a 4 °C o 60 min, and hen cen i-
uged a 2500×g o 15 min. The se um was aliquo ed
o and s o ed a −80 °C un il u he analysis.
A subse o pa ien s and con ols ec ui ed in Edinbu gh
(Addi ional ile 1: Table S3) unde wen immunomagne ic
cell sepa a ion o ob ain CD19
+
B cells. The me hods ha e
p e iously been de ailed elsewhe e [13]. Venepunc u e
using 9-ml K3 EDTA acue e (G eine ) ubes was
pe o med o ob ain be ween o 18 and 36 ml o
EDTA-bu e ed blood. An ini ial Ficoll (Ficoll-Paque, GE
Heal hca e, Bucks, UK) densi y g adien cen i uga ion
was pe o med o ob ain pe iphe al blood mononuclea
cells. Cells labelled wi h an ibody-coa ed mic obeads
(human CD8
+
and CD19
+
mic obeads, 20 μlpe 1×10
7
cells) we e immunomagne ic sepa a ed using he au o-
MACs P o cell sepa a o (Mil enyi, Ge many). CD19
+
sepa a ions we e pe o med ollowing an ini ial CD8
+
deple ion s ep. Nucleic acids we e ex ac ed using AllP ep
(Qiagen, Hilden, Ge many) acco ding o he manu ac-
u e ’s guidance and s o ed a −80 °C.
Colonic biopsies om con ols and UC pa ien s wi h
inac i e and ac i e o m o disease we e mechanically
dissocia ed o p epa e single-cell suspensions using
Hanks’balanced sal solu ion modi ied medium, wi hou
calcium chlo ide and magnesium sul a e (HBSS) (Sigma),
wi h penicillin/s ep omycin and gen amicin. PBMCs
we e ob ained by densi y g adien cen i uga ion using
Lymphop ep. CD3
+
T cells ( om biopsies and blood)
we e magne ically so ed by using he EasySep™Human
T Cell En ichmen Ki (STEMCELL) ollowing he man-
u ac u e ’s ins uc ions. Following cell isola ion, DNA
ex ac ion was pe o med using he In iso b Spin Tissue
Mini Ki (S a ec Molecula ) ollowing he manu ac-
u e ’s ins uc ions.
DNA me hyla ion analysis
We analyzed p omo e me hyla ion o he candida e
genes in he DNA om whole blood, as well as om he
sepa a ed CD19
+
B cells. In addi ion, o he MGAT3,
which is a glycosyl ans e ase wi h di ec and known
unc ion in IgG glycosyla ion [52], we analyzed p omo e
me hyla ion—in DNA om PBMCs, CD3
+
T cells iso-
la ed om PBMCs, and CD3
+
T cells isola ed om he
colonic mucosa o heal hy con ols and UC pa ien s
(classi ied acco ding o ac i e and inac i e o m o he
disease) o he hi d independen smalle subcoho col-
lec ed by he Gas oen e ology Depa men o Cen o
Hospi ala do Po o-Hospi al de San o An ónio, Po ugal
(Addi ional ile 1: Table S4). All specimens we e sub-
jec ed o his ological examina ion and classi ica ion. All
pa icipan s ga e in o med consen abou all clinical
p ocedu es, and esea ch p o ocols we e app o ed by
he e hics commi ee o CHP/HSA, Po ugal (233/
12(179-DEFI/177-CES).
Klasiće al. Clinical Epigene ics (2018) 10:75 Page 3 o 14
Fo DNA me hyla ion analysis, 500 ng o DNA om
whole blood was bisul i e con e ed using EZ-96 DNA
Me hyla ion Gold ki (Zymo Resea ch, F eibu g,
Ge many), and 100 ng o DNA om CD19
+
B cells,
PBMCs, and T cells was con e ed using EZ DNA
Me hyla ion Gold ki (Zymo Resea ch, F eibu g,
Ge many) acco ding o he manu ac u e ’s p o ocol.
Two o six py osequencing assays we e de eloped o
p omo e egions o each o he i e candida e genes
(BACH2,MGAT3,IL6ST,IKZF1, and LAMB1). The se-
lec ion o analyzed CpG si es was andom o assays 2–5
o he MGAT3 gene. CpG si es wi hin he MGAT3 assay
1 we e selec ed based on he GEO (Gene Exp ession
Omnibus) da abase whe e me hyla ion da a we e
ob ained using Illumina HumanMe hyla ion450 Bead-
Chip 1.1 echnology. Fo he BACH2 gene, assays we e
selec ed based on loca ion o di e en ially me hyla ed
CpGs in di e en cell lines es ed by ENCODE p ojec ,
using Illumina HumanMe hyla ion450 BeadChip 1.1
echnology (a newe e sion, he In inium Me hyla ionE-
PIC 850K was no a ailable a he ime). We used
adi ional bisul i e-based p o ocols which canno
disc imina e be ween 5-me hylcy osine (5-mC) and
5-hyd oxyme hylcy osine (5-hmC) as oxida i e bisul i e
(oxBS-450K) me hod can [53]. Howe e , ecen s udies
ha e shown ha global DNA hyd oxyme hyla ion is e y
low in blood cells [54,55]. Fu he mo e, hyd oxyme hy-
la ion is signi ican ly deple ed om p omo o s and CpG
islands, while en iched in he gene bodies [53,56].
Based on he es ima ed s a is ical powe , we did ini ial
sc eening on 60 pa ien s o each py osequencing assay,
a e which we excluded hose genes (py osequencing
assays) ha did no show any s a is ically signi ican di -
e ences be ween IBD pa ien s and heal hy con ols.
Py osequencing assays o LAMB1,IL6ST, and IKZF1
a e shown in Addi ional ile 2: Figu e S1. We con inued
o analyze p omo e me hyla ion only in he BACH2 and
MGAT3 genes. Speci ic egions we e ampli ied using
Py oMa k PCR ki (Qiagen, Hilden, Ge many). The cyc-
ling condi ions o he BACH2 gene we e as ollows: ini-
ial polyme ase ac i a ion s ep o 15 min a 95 °C
ollowed by 50 cycles o 30 s dena u a ion a 95 °C, p i-
me annealing o 30 s a p ime -speci ic empe a u es
(Addi ional ile 1: Table S5), and 30 s a 72 °C, wi h inal
ex ension a 72 °C o 10 min. The cycling p o ocol used
o ampli ica ion o he MGAT3 gene agmen s was de-
sc ibed p e iously [35], wi h he annealing empe a u e
adjus ed o 55 °C o he agmen 1 pe o med on DNA
om CD19
+
B cells. Fo quan i a i e measu emen o
DNA me hyla ion le el a speci ic CpG si es,
PCR-ampli ied bisul i e-con e ed DNA was sequenced
using he Py oMa k Q24 Ad anced py osequencing sys-
em (Qiagen) acco ding o he manu ac u e ’s ecom-
menda ions. Sequences o PCR and py osequencing
p ime s o he BACH2 and he MGAT3 genes a e lis ed
in Addi ional ile 1: Table S5. EpiTec PCR Con ol DNA
Se (me hyla ed and unme hyla ed bisul i e-con e ed
human DNA, Qiagen) was used as a con ol o PCR and
py osequencing eac ions.
S a is ical analysis
The nonpa ame ic Mann-Whi ney U es was used o
compa e he me hyla ion s a us o CpG si es encom-
passed by he py osequencing assays in he MGAT3 and
BACH2 genes be ween he wo independen g oups: HC
compa ed o each o CD o UC. Signi icance h eshold
was se a p< 0.05 wi h addi ional Bon e oni co ec ion
o mul iple es ing. Gi en ha age was ou p ima y
conce n as a po en ial con ounde , we isualized he age
in he h ee g oups (CD, UC, and HC) o he samples
included in each analysis as iolin plo s (Addi ional ile 3:
Figu e S2) and assu ed he e was no signi ican di e -
ence be ween he age g oups (p> 0.05) using he
Mann-Whi ney U es . This was done o assu e he
alidi y and s eng hen he a ionale o he selec ion o
s a is ical me hods.
Fo heda ao heMGAT3 p omo e me hyla ion
om PBMCs, CD3
+
T cells isola ed om blood, and
CD3
+
T cells isola ed om in lamed colonic mucosa
( he Po o coho ), he Mann-Whi ney U es was
applied wi h Bon e oni co ec ion accoun ing o 15
CpG si es.
Glycan analysis
Glycans p esen on IgG we e analyzed om se um o
o e 1000 IBD (UC and CD) pa ien s and heal hy con-
ols in he Edinbu gh coho using ul a pe o mance
liquid ch oma og aphy (UPLC) [43,50]. In he Flo ence
coho , plasma samples o 3500 IBD pa ien s and heal hy
con ols was used o analysis o IgG glycopep ides by
liquid ch oma og aphy coupled o mass spec ome y
(LC-MS) [46]. The da a o IgG glycosyla ion analysis
we e used in his wo k o co ela ion analysis wi h p o-
mo e me hyla ion da a o MGAT3 and BACH2 genes,
wi h ma ching samples om he e y same pa ien s and
heal hy con ols.
Isola ion o IgG om blood plasma
IgG has been isola ed om blood plasma by a ini y
ch oma og aphy using CIM P o ein G 96-well pla e
(BIA Sepa a ions, Ajdo ščina, Slo enia) and acuum
mani old (Pall Co po a ion, Po Washing on, NY, USA)
as p e iously desc ibed [57,58]. In sho , plasma sam-
ples (50–90 μl) we e dilu ed wi h 1 × PBS, pH 7.4 in he
a io 1:7. All samples we e il e ed h ough 0.45 and
0.2-μm Ac oP ep GHP il e pla es (Pall Co po a ion)
using acuum mani old and immedia ely applied o p e-
condi ioned P o ein G pla e. A e washing o he
Klasiće al. Clinical Epigene ics (2018) 10:75 Page 4 o 14
P o ein G pla e, IgG was elu ed wi h 0.1 mol L
−1
o mic
acid and immedia ely neu alized wi h ammonium bica -
bona e o pH 7.0. P o ein G pla e was egene a ed and
s o ed a 4 °C.
IgG glycosyla ion analysis using ul a-pe o mance liquid
ch oma og aphy
N-glycans om isola ed IgG in he Edinbu gh coho
we e eleased wi h PNGase F a e d ying 300 μl o each
IgG elu ion ac ion, labeled wi h 2-aminobenzamide
and excess o egen s emo ed by clean-up using hyd o-
philic in e ac ion liquid ch oma og aphy solid phase
ex ac ion (HILIC-SPE). Fluo escen ly labeled and pu i-
ied N-glycans we e sepa a ed by HILIC-UPLC using
Acqui y UPLC ins umen (Wa e s, Mil o d, MA, USA)
as p e iously desc ibed [43]. Samples we e sepa a ed
in o 24 peaks [57], and he amoun o N-glycans in each
ch oma og aphic peak was exp essed as a pe cen age o
o al in eg a ed a ea (% a ea).
IgG glycosyla ion analysis using liquid ch oma og aphy
coupled o mass spec ome y
In he Flo ence coho , Fc-speci ic IgG glycopep ides
we e analyzed a e IgG pu i ica ion, o e nigh ypsin
diges ion a 37 °C, and e e se-phase pu i ica ion on
Ch omabond C18 beads using acuum mani old as
desc ibed [46,59]. Samples we e analyzed using nanoli-
quid ch oma og aphy coupled o mass spec ome y
(nanoLC-MS), on a nanoACQUITY UPLC sys em
(Wa e s, Mil o d Massachuse s, USA) coupled o
quad upole-TOF-MS (Compac ; B uke Dal onics,
B emen, Ge many) equipped wi h a shea h- low ESI
sp aye (capilla y elec opho esis ESI-MS sp aye ;
Agilen Technologies, San a Cla a, USA) as p e iously
desc ibed [46]. The nanoACQUITY UPLC sys em and
he B uke Compac Q-TOF-MS we e ope a ed unde
HyS a so wa e e sion 3.2.
Da a was p ocessed as desc ibed p e iously [46,60].
This esul ed in he ex ac ion o 16 IgG1, 16 IgG2/3,
and 11 IgG4 glyco o ms. The yp ic Fc-glycopep ides
o IgG2 and IgG3 subclasses ha e iden ical pep ide
moie ies in he Caucasian popula ion and a e he e o e
no dis inguishable wi h his me hodology. Anno a ion
o he spec a was done based on accu a e mass acco d-
ing o he ele an li e a u e [40,57].
Co ela ion analysis
Me hyla ion da a o he BACH2 (assay 2) and he
MGAT3 (assays 1 and 2) genes (ob ained o he wo
la ge coho s) we e il e ed acco ding o he peak quali y
by ejec ing peaks ma ked as “ ailed”by he py ose-
quencing so wa e. A e age me hyla ion ac oss all
assayed CpG si es was calcula ed o each py osequenc-
ing assay in each coho . Me hyla ion esul s o
indi idual pa ien s we e ma ched wi h hei co espond-
ing glycan p o iles. Sizes o da ase s and pa ien classes
ob ained a e including comple e eco ds (i.e., bo h
me hyla ion da a and glycan p o iles p esen ) a e shown
in Addi ional ile 1: Table S6. Indi idual glycan s uc-
u es we e ep esen ed as ela i e abundances and
ba ch-co ec ed. Pe cen age o s uc u es wi h bisec ing
N-ace ylglucosamine was calcula ed o each coho as a
de i ed ai a his poin . Glycan s uc u es iden i ied by
each me hod we e ansla ed o Ox o d no a ion, and
only he 13 s uc u es p esen in bo h he Edinbu gh
and Flo ence da ase s we e conside ed o co ela ion.
We used IgG1 da a om he Flo ence coho , as his
iso o m was he mos abundan . Th ee addi ional de-
i ed ai s we e calcula ed: a ios o FA2B o FA2,
FA2BG1 o FA2G1, and FA2BG2 o FA2G2.
Pea son co ela ion be ween CpG me hyla ion da a
and 17 glycan ea u es (13 s uc u es and 4 de i ed
ai s) was calcula ed. Signi icance h eshold was se a
p< 0.05 wi h addi ional Bon e oni co ec ion o
17- old mul iple es ing.
Me hyla ion o assayed CpG si es in p omo e s o he
BACH2 and MGAT3 genes was co ela ed wi h mea-
su ed glycan s uc u es. Pea son co ela ion coe icien
along wi h he associa ed p alue was calcula ed be ween
a e age CpG me hyla ion ( o all genes/assays) and each
measu ed IgG glycan s uc u e. Calcula ion was done on
pai wise comple e obse a ions. Only co ela ions wi h
he p alue below 0.01 we e conside ed u he . Nex ,
co ela ion coe icien s o all CpG assays we e calcu-
la ed, which was used o ank glycan s uc u es acco d-
ing o egula ion by he assayed egion. Glycan
s uc u es wi h he s onges co ela ion (ei he posi i e
o nega i e) o CpG me hyla ion we e hen used o
explain egula o y e ec s. All calcula ions and da a isu-
aliza ions we e done in R language and en i onmen o
s a is ical compu ing (R Founda ion o S a is ical Com-
pu ing, Vienna, Aus ia). Visualiza ion o co ela ions
was done using he R package “co plo .”
Resul s
P omo e me hyla ion o he candida e genes in whole
blood and B cells o IBD pa ien s
In o de o assess he le el o me hyla ion in CpG
islands o he i e candida e genes (BACH2,MGAT3,
IKZF1,LAMB1,andIL6ST), associa ed wi h bo h IBD
and IgG glycosyla ion by GWAS, we de eloped se -
e al py osequencing assays o each o he genes
(Fig. 1and Addi ional ile 2:Figu eS1).Wepe -
o med ini ial sc eening o he py osequencing assays
on 60 pa ien s. O e all cy osine me hyla ion le els
we e e y low o LAMB1 (a e age alue pe g oup <
8%), o IL6ST (< 3.5%), and o IKZF1 (< 4%) in he
assayed po ion o hei p omo e s; he e o e, we
Klasiće al. Clinical Epigene ics (2018) 10:75 Page 5 o 14

could no iden i y di e en ial me hyla ion. We hen
excluded hese genes om u he analysis.
MGAT3 and BACH2 p omo e me hyla ion was ana-
lyzed in se e al hund ed IBD pa ien s and heal hy con-
ols om wo independen coho s (Addi ional ile 1:
Tables S1 and S2). In hese genes, we analyzed me hyla-
ion le el a 47 CpGs co e ed by i e py osequencing as-
says in he BACH2 gene: 21 CpG si es we e in he
p omo e egion, 1 CpG si e was in i s exon, and 25
CpG si es we e loca ed in he i s in on o he gene. A
o al o 32 CpG si es, co e ed by i e py osequencing
assays, was analyzed o MGAT3: 18 CpG si es we e
loca ed in he p omo e egion and 14 CpG si es in he
i s in on (Fig. 1). Mos o hose CpG si es we e
loca ed wi hin CpG islands o he bo h genes. We ound
di e en ial CpG me hyla ion be ween IBD pa ien s and
HC wi hin he assay A2, loca ed a 213–368 bp up-
s eam ( ela i e o he gene o ien a ion) o he TSS in
he BACH2 p omo e and wi hin he assays A1 and A2,
loca ed in he CpG island 1 o he MGAT3 gene. The
same pa e n o di e en ial me hyla ion a hese CpG
si es was obse ed in whole blood o pa ien s and HC
om wo la ge independen coho s (Fig. 2). CpG
me hyla ion le el was gene ally low (up o 20%) in he
assayed po ion o he BACH2 p omo e ; howe e , sig-
ni ican di e ences be ween HC and CD me hyla ion
le el we e eco ded a CpG si es 4, 5, 6, and 8 (Fig. 2a).
Fo he assayed po ion o he MGAT3 p omo e , gene al
me hyla ion le el was high, wi h all CpG si es show-
ing a ep oducibly signi ican di e ence be ween HC
andbo hCDandUC.CpGsi es2,13,and15
showed signi ican di e ences only o CD bu no o
UC. Di ec ion o change was di e en o he wo
genes—di e en ially me hyla ed CpG si es wi hin he
BACH2 p omo e we e hypome hyla ed, while hose
o he MGAT3 gene we e hype me hyla ed in disease
compa ed o heal hy indi iduals.
These esul s we e con i med on CD19
+
Bcells
isola ed om pe iphe al whole blood o he independ-
en , smalle pa ien sample om he Edinbu gh
coho (67 samples). The CpG si es 1–5 and 12–13 in
he MGAT3 p omo e we e di e en ially me hyla ed
be ween CD and HC. Only CpG si e 5 wi hin he
assay A2 o he BACH2 gene showed change in he
me hyla ion le el be ween HC and CD in CD19
+
B
cells (Fig. 2b). The e we e no di e ences in he
me hyla ion le el o he same CpG si es wi hin
assayed agmen s o he BACH2 and MGAT3 genes
be ween UC and HC.
I is wo h no ing ha he same pa e n o CpG
me hyla ion di e ences was obse ed in PBMCs o he
IBD pa ien s and HC om bo h la ge independen
coho s, and mos o he CpG si es wi hin he assayed
po ion o he MGAT3 p omo e we e also di e en ially
me hyla ed in CD19
+
B cells om he subse o IBD
pa ien s om he Edinbu gh coho (Fig. 2a,b).
Fig. 1 Posi ions o he BACH2 and MGAT3 genes in he human genome and ela i e posi ions o he agmen s analyzed o me hyla ion le el
(py osequencing assays) wi hin hese genes. Fo each py osequencing assay (A1–A5), he egion ampli ied by PCR is shown. Posi ions o he genes on
he ch omosomes a e shown using ch omosome models ( ed e ical lines). Coo dina es a e ela i e o he hg19 human genome assembly. The genes
a e displayed in he di ec ion co esponding o hei eading ames. Anno a ions (CpG islands and py osequencing assays) a e o scale. TSS ansc ip ion
s a si e
Klasiće al. Clinical Epigene ics (2018) 10:75 Page 6 o 14
a
b
Fig. 2 (See legend on nex page.)
Klasiće al. Clinical Epigene ics (2018) 10:75 Page 7 o 14
P omo e me hyla ion o he MGAT3 gene in CD3
+
T cells
om PBMCs and in lamed colonic mucosa o UC pa ien s
We included in ou in es iga ion biopsy samples o UC
pa ien s om an independen coho om he Gas o-
en e ology Depa men o Cen o Hospi ala do
Po o-Hospi al de San o An ónio, Po ugal. Gi en he
echnical challenges in ob aining DNA and RNA om a
small numbe o pu i ied cells om in lamed colonic
mucosa, a subse o pa ien s wi h ac i e and inac i e
phase o UC was selec ed o me hyla ion analysis om
h ee sou ces: (1) PBMCs, (2) CD3
+
T cells isola ed om
PBMCs, and (3) CD3
+
T cells isola ed om colonic
mucosa (see also Addi ional ile 1: Table S4).
In e -indi idual a ia ion o MGAT3 me hyla ion le el
measu ed om PBMCs and om CD3
+
T cells isola ed
om PBMCs was qui e la ge—i a ied om 47 o 94%
and om 26 o 90%, espec i ely. The e o e, we could
no ind any di e ence in CpG me hyla ion le el
be ween UC pa ien s and HC in assayed agmen s o
he MGAT3 p omo e , nei he in PBMCs no in CD3
+
T
cells isola ed om PBMCs. Howe e , we eco ded a o al
o 7 (ou o 15) di e en ially me hyla ed CpG si es in
CD3
+
T cells isola ed om colonic mucosa o UC
pa ien s wi h ac i e disease compa ed wi h HC (Fig. 3).
O e all, he me hyla ion le el o CpGs wi hin assayed
agmen s o he MGAT3 p omo e was high in CD3
+
T
cells om heal hy colonic mucosa (be ween 77 and
98%). When compa ed o in lamed mucosa o UC
pa ien s wi h ac i e phase o he disease, he same CpG
si es we e hypome hyla ed, wi h he highes di e ence a
he CpG posi ion 10 (13.24%; p= 5.08 × 10
−5
; Fig. 3). In
inac i e UC, no signi ican di e ences could be ound
a e Bon e oni co ec ion o mul iple es ing.
I is wo h no ing ha he me hyla ion pa e n in
CD3
+
T cells isola ed om in lamed colonic mucosa di -
e ed om he me hyla ion pa e ns in PBMCs and o
CD3
+
T cells isola ed om PBMCs. The la e wo we e
e y simila and had much lowe me hyla ion le els han
ha measu ed o CD3
+
T cells om in lamed colonic
mucosa (Fig. 3). Also, MGAT3 me hyla ion le el was
inc eased in UC compa ed wi h HC when measu ed
om PBMCs o CD3
+
T cells isola ed om PBMCs
(hype me hyla ion), while i was dec eased (hypome hy-
la ion) when measu ed om CD3
+
T cells isola ed om
in lamed colonic mucosa in compa ison wi h heal hy
mucosa.
Co ela ion be ween he MGAT3 and BACH2 p omo e
me hyla ion and IgG glycosyla ion
The e we e s a is ically signi ican co ela ions ha epli-
ca ed ac oss assays and coho s be ween he MGAT3
p omo e me hyla ion and glycan s uc u es FA2,
FA2G2, FA2BG2, and FA2G2S1, as well as he de i ed
ai o he a io o FA2B o FA2 (Fig. 4a). All co ela-
ions excep wi h FA2 we e nega i e. No ep oducible
signi ican co ela ions could be ound be ween BACH2
p omo e me hyla ion and he glycan s uc u es
(Fig. 4a).
In o de o in e a mechanis ic pa hway o he
obse ed co ela ions, we mapped hem o he glycan
biosyn hesis pa hways (Fig. 4b). The a io o bisec ing
glycans o FA2 was aken as an indica o o MGAT3
(GnT-III) ac i i y. This in e p e a ion allowed us o in e
lowe GnT-III enzyma ic ac i i y when he p omo e o
he MGAT3 gene was me hyla ed. Inc ease in MGAT3
p omo e me hyla ion co ela ed wi h a dec ease in ce -
ain galac osyla ed and sialyla ed s uc u es (Fig. 4b). In
addi ion o he dec eased le els o bisec ing GlcNAc on
non-galac osyla ed glycans (B/FA2), he mos signi ican
e ec o he MGAT3 p omo e me hyla ion on IgG gly-
come composi ion was a dec ease o IgG galac osyla ion.
Discussion
Resul s om his s udy s ongly indica e ha he
MGAT3 and BACH2 genes play an impo an ole in
IBD pa hogenesis and sugges a possible disease pa hway
media ed by he p o-in lamma o y p ope ies o IgG
an ibodies acqui ed by al e a ions in Fc glycosyla ion.
Ou ecen s udy, pe o med on a la ge coho o o e
1000 IBD pa ien s, epo ed a signi ican di e ence in
IgG glycome composi ion in bo h UC and CD compa ed
o heal hy con ols [43,46]. We ound a dec ease in
quan i y o galacosyla ed glycans in bo h CD and UC, as
well as a dec ease in sialyla ed glycans and an inc ease
o bisec ing GlcNAc on digalac osyla ed glycan s uc-
u es on IgG in CD. Indeed, al e na i e N-glycosyla ion
o an IgG molecule in luences i s unc ion—p o-in lam-
ma o y and an i-in lamma o y ac i i y depends on he
(See igu e on p e ious page.)
Fig. 2 Box plo o CpG me hyla ion in pe iphe al whole blood o he BACH2 and MGAT3 genes in he Edinbu gh and Flo ence coho s and in B
cells om a subse o pa ien s om Edinbu gh coho . G oups we e compa ed using he Mann-Whi ney U es wi h signi icance h eshold o p= 0.05,
co ec ed o mul iple es ing using he Bon e oni me hod. aMe hyla ion le els we e gene ally low in he assayed po ion o he BACH2 gene p omo e ,
wi h signi ican di e ences be ween HC and CD me hyla ion a CpG si es 4, 5, 6, and 8 ( eplica ed in bo h coho s). Fo he MGAT3 gene, gene al
me hyla ion le el was high, wi h all CpG si es showing a ep oducibly signi ican di e ence be ween HC and bo h CD and UC, excep o CpG si es 2, 13,
and 15 o which ep oducible signi ican di e ences we e ound only be ween HC and CD. bIn B cells, isola ed om PBMCs o a subse o he pa ien s
om he Edinbu g coho , di e en ial me hyla ion was ound a he CpG posi ion 5 o he BACH2 gene (assay 2) be ween HC and CD, while o he MGAT3
gene, di e en ially me hyla ed we e CpG si es 1–5, 12, and 13 be ween HC and CD. CD C ohn’s disease, UC ulce a i e coli is, HC heal hy con ols
Klasiće al. Clinical Epigene ics (2018) 10:75 Page 8 o 14
glycans added on he Cy2 domain o i s Fc egion [29].
These glycans a e o a bian enna y complex ype wi h o
wi hou bisec ing GlcNAc, co e ucose, galac ose, and
sialic acid esidues [61]. Recen ly, his was con i med in
a la ge mul i-cen ic s udy o IgG glycome in IBD [46].
The e o e, glycan changes obse ed on IgG in pe iphe al
blood o UC and CD pa ien s a e ob iously associa ed
wi h inc eased in lamma o y po en ial o IgG, sugges ing
unc ional ele ance o IgG glycosyla ion o IBD.
He e, we p opose a possible mechanism unde lying
he abe an IgG glycosyla ion pa e n obse ed in IBD
[43,46]. Ou o i e candida e genes analyzed in his
wo k, he MGAT3, a glycosyl ans e ase which pa ici-
pa es in syn hesis o IgG glycans, and he BACH2,a
ansc ip ion ac o and a mas e egula o o a ne wo k
o genes ele an o B cell in eg i y [62,63], showed di -
e en ial me hyla ion in pe iphe al blood o bo h CD
and UC pa ien s when compa ed o heal hy indi iduals.
E en hough we iden i ied changes in me hyla ion le el
o bo h UC and CD compa ed o HC, he di e ences
we e mo e p onounced o CD. This is conco dan wi h
o he s udies ha explo ed ei he whole genome me hy-
la ion o p omo e me hyla ion o candida e genes in
IBD [11]. The ex en o he change in IgG glycome com-
posi ion was also consis en ly highe in CD han UC
compa ed o HC [43,46].
The p o ein encoded by he MGAT3 gene (N-ace yl-
glucosaminyl ans e ase III, GnT-III) is esponsible o
signi ican unc ional al e a ion o glycans on he Fc
egion o an IgG an ibody. The GnT-III adds N-ace yl-
glucosamine (GlcNAc) on β1,4-linked mannose in he
h ee-mannose co e o N-glycans, p oducing bisec ing
GlcNAc s uc u es. In he same CD pa ien s, who
showed changed MGAT3 p omo e me hyla ion le el in
pe iphe al blood cells, a signi ican inc ease in he pe -
cen age o bisec ing GlcNAc on glycans o ci cula ing
IgG an ibodies was eco ded, oo. The associa ion o he
MGAT3 wi h bo h IgG N-glycosyla ion [49] and C ohn’s
disease [4,5] sugges s ha N-glycans wi h bisec ing
GlcNAc could be in ol ed in CD pa hogenesis h ough
unc ional e ec on IgG an ibody.
Co ela ions be ween BACH2 and MGAT3 p omo e
me hyla ion and glycan s uc u es ha e gi en u he
insigh in o he changes o IgG glycosyla ion pa e n me-
dia ed by hose wo genes (Fig. 4b). The MGAT3
a
b
c
Fig. 3 Box plo o CpG me hyla ion le el in he MGAT3 gene p omo e
(assays A1 and A2) analyzed om PBMCs (a), CD3
+
T cells isola ed om
PBMCs (b), and CD3
+
T cells isola ed om in lamed colonic mucosa (c)
om he independen coho o Po o. Changes be ween UC pa ien s
wi h ac i e disease and HC we e s a is ically signi ican only in CD3
+
T
cells isola ed om in lamed colonic mucosa a CpG posi ions 3 and 7–12
(p< 0.05 a e Bon e oni co ec ion o 15 hypo heses). PBMC pe iphe al
blood mononuclea cells, UC ulce a i e coli is, HC heal hy con ols
Klasiće al. Clinical Epigene ics (2018) 10:75 Page 9 o 14