RESEARCH Open Access
Epigenome-wide DNA me hyla ion p o iling
o pe ip os a ic adipose issue in p os a e
cance pa ien s wi h excess adiposi y—a
pilo s udy
Yan Cheng
1,2
, Cá ia Mon ei o
3,4
, And eia Ma os
5,6
, Jiaying You
1
, A elino F aga
6,7
, Ca ina Pe ei a
3,8
,
Vic o ia Ca alán
9,10
, Amaia Rod íguez
9,10
, Ja ie Gómez-Amb osi
9,10
, Gema F ühbeck
9,10,11
,
Rica do Ribei o
3,5,6,12,13*†
and Pingzhao Hu
1*†
Abs ac
Backg ound: Pe ip os a ic adipose issue (PPAT) has been ecognized o associa e wi h p os a e cance (PCa)
agg essi eness and p og ession. He e, we sough o in es iga e whe he excess adiposi y modula es he
me hylome o PPAT in PCa pa ien s. DNA me hyla ion p o iling was pe o med in PPAT om obese/o e weigh
(OB/OW, BMI > 25 kg m
−2
) and no mal weigh (NW, BMI < 25 kg m
−2
) PCa pa ien s. Signi ican di e ences in
me hyla ed CpGs be ween OB/OW and NW g oups we e in e ed by s a is ical modeling.
Resul s: Fi e housand i e hund ed wen y-six di e en ially me hyla ed CpGs we e iden i ied be ween OB/OW and
NW PCa pa ien s wi h 90.2% hype me hyla ed. Fou hund ed eigh y- h ee o hese CpGs we e ound o be loca ed a
bo h p omo e s and CpG islands, whe eas he ep esen ing 412 genes we e ound o be in ol ed in plu ipo ency o
s em cells, a y acid me abolism, and many o he biological p ocesses; 14 o hese genes, pa icula ly FADS1,MOGAT1,
and PCYT2, wi h p omo e hype me hyla ion p esen ed wi h signi ican ly dec eased gene exp ession in ma ched
samples. Addi ionally, 38 genes we e co ela ed wi h an igen p ocessing and p esen a ion o endogenous an igen ia
MHC class I, which migh esul in a y acid accumula ion in PPAT and umo immune e asion.
Conclusions: Resul s showed ha he whole epigenome me hyla ion p o iles o PPAT we e signi ican ly di e en in
OB/OW compa ed o no mal weigh PCa pa ien s. The epigene ic a ia ion associa ed wi h excess adiposi y likely
esul ed in al e ed lipid me abolism and immune dys egula ion, con ibu ing owa ds un a o able PCa
mic oen i onmen , hus wa an ing u he alida ion s udies in la ge samples.
Keywo ds: DNA me hyla ion, Pe ip os a ic adipose issue, Obesi y, P os a e cance , Mic oen i onmen
Backg ound
P os a e cance (PCa) is one o he mos equen malig-
nancies in men and he second leading cause o cance -
ela ed dea h in he No h Ame ica and mos wes e n
Eu opean coun ies [1,2]. Epidemiological s udies suppo
obesi y o excess adiposi y as an impo an en i onmen al
isk ac o o PCa, being p ima ily associa ed wi h
ad anced disease and dea h [3]. Pe ip os a ic adipose issue
(PPAT), a whi e a depo su ounding he p os a e
capsula -like s uc u e, has been ecognized o ha e he
po en ial o exe p o- umo al endoc ine and pa ac ine
in luences on p os a e cance cell’s biological pheno ypes
[4]. The e is now e idence ha obesi y and o e weigh
esul in excess a deposi a PPAT [5], al e ed a y acid
p o ile [6], mig a ion o umo cells [7], sec e ion o a
a ie y o adipokines, such as in e leukin-1 be a (IL-1b),
os eopon in, lep in, umo nec osis ac o alpha (TNF-a),
* Co espondence: [email p o ec ed];[email p o ec ed]
†
Equal con ibu o s
3
Molecula Oncology G oup, Po uguese Ins i u e o Oncology, Po o,
Po ugal
1
Depa men o Biochemis y and Medical Gene ics & Depa men o
Elec ical and Compu e Enginee ing, Uni e si y o Mani oba, Winnipeg,
Canada
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.
Cheng e al. Clinical Epigene ics (2018) 10:54
h ps://doi.o g/10.1186/s13148-018-0490-3
and dec eased adiponec in, hus con ibu ing o a umo
mic oen i onmen ha ul ima ely acili a es PCa agg es-
si eness [7,8].
DNA me hyla ion is a well-known epigene ic mechan-
ism esul ing om he in e ac ion be ween en i onmen-
al ac o s and he genome [9]. DNA me hyla ion wi h
a ia ion o CpG si es is associa ed wi h issue-speci ic
gene modula ion and in ol ed in pheno ype ansmis-
sion and in he de elopmen o diseases [10]. Excess
adiposi y, as a consequence o en i onmen al ac o s
such as excessi e ood consump ion o inac i e li es yle,
has been iden i ied as a egula o o epigene ic modi ica-
ion in adipose issue. Recen indings om expe imen al
s udies sugges ed ha modi ica ion o DNA me hyla ion
pa e n in adipose issue and adipocy es was ela ed wi h
de elopmen o cance , ype 2 diabe es, and ca dio ascu-
la diseases h ough in luencing me abolism and in lam-
ma ion [11–13]. Addi ionally, se e al s udies epo ed
al e ed DNA me hyla ion in PCa cells as compa ed wi h
adjacen benign issue, and some signi ican ly me hyl-
a ed CpG si es and genes we e ound o be esponsible
o he occu ence and p og ession o PCa [14–16].
Ne e heless, he epigenome-wide DNA me hyla ion
p o ile o PPAT om excess adiposi y PCa pa ien s is
cu en ly unknown despi e i s po en ial mechanis ic
in ol emen in obesi y associa ion wi h PCa.
The aim o his s udy was o pe o m a epigene ic-
wide associa ion s udy (EWAS) in o de o e alua e
DNA me hyla ion p o ile o PPAT ob ained om obese/
o e weigh (OB/OW) in compa ison wi h no mal weigh
(NW) PCa pa ien s and iden i y di e en ially me hyla ed
si es. We also explo ed he consequen ial po en ial bio-
logical unc ions ha accoun o he e ec o PPAT
om OB/OW subjec s in PCa molecula mechanisms.
Me hods
S udy samples
This s udy included en p os a e cance pa ien s om
he Po uguese Ins i u e o Oncology, Po o Cen e. In-
clusion c i e ia and condi ions o his s udy ha e been
p e iously epo ed, including he p ocedu es o PPAT
collec ion, handling, and s o age [4]. B ie ly, PPAT was
collec ed and immedia ely p ocessed in he ope a ing
oom and anspo ed o he labo a o y wi hin 2 h in
app op ia e cul u e media and empe a u e condi ions,
in o de o minimize p e-analy ical e o s. Pa ien s’
signed in o med consen and esea ch p ocedu es we e
app o ed by he ins i u e’s e hics commi ee.
The clinical and pa hological cha ac e is ics o pa ici-
pan s a e p esen ed in Table 1. The en subjec s we e
selec ed om a la ge g oup o pa ien s unde going
p os a e su ge y (n= 51) [4,17] ha i ed he s ic
inclusion and exclusion c i e ia, in o de o con ol o
a iables ha migh in luence adipose issue gene
exp ession o me hyla ion (e.g., an i-diabe ic o an i-
dyslipidemia d ugs, s age o disease and PSA, concomi an
diseases such as diabe es, o he neoplasia o me abolic
synd ome). Subjec s we e ma ched o age a diagnosis,
PSA alue, Gleason g ade, and s age o disease, which
di e ed in body mass index (BMI). BMI was calcula ed by
di iding weigh in kilog ams by he squa ed heigh in
me e s and ca ego ized using he WHO (Wo ld Heal h
O ganiza ion) c i e ia: no mal weigh , BMI < 25 kg m
−2
,
o e weigh , 25 ≤BMI < 30 kg m
−2
, and obese, BMI ≥
30 kg m
−2
. Obese and o e weigh we e combined in o one
excess adiposi y g oup (n=5, BMI≥25 kg m
−2
) e sus
no mal weigh g oup (n= 5, BMI < 25 kg m
−2
). The e o e,
he wo g oups we e selec ed o di e only by BMI, in
o de o e lec ou objec i e o assessing whe he excess
adiposi y (BMI) in luences PPAT me hyla ion p o ile.
Epigenome-wide DNA me hyla ion analysis
DNA was isola ed om PPAT using Pu egene hisal
ex ac ion me hod (Qiagen/Gen a). B ie ly, he issue
was minced wi h scalpels in a s e ile pe i dish on ice
and hen ans e ed o Pu egene Cell Ki o o e nigh
P o einase K diges a 55 °C. A second P o einase K
diges was done he nex mo ning o 5 h. DNA om
he diges ed issue was pu i ied using Pu egene ex ac-
ion p o ocol (Qiagen/Gen a). Pu i ied DNA was
washed 2× wi h 70% e hanol and DNA pelle ai d ied
and ehyd a ed in TE (10 mM T is-Cl, 1 mM EDTA
pH 7.5). Epigenome-wide DNA me hyla ion was ana-
lyzed using he In inium Human Me hyla ion450
(HM450) BeadChip (Illumina, San Diego, CA, USA) in
he Cen e o Applied Genomics (To on o). This a ay
Table 1 Clinicopa hological cha ac e is ics o PCa pa ien s by
BMI ca ego y
Cha ac e NW (n= 5) OB/OW (n=5) P alue
Age (yea s) 65.2 ± 3.8 63.2 ± 2.5 0.67
a
BMI (kg/m
2
) 23.0 ± 0.3 29.0 ± 0.9 0.0003
a
Gleason sco e
< 7 2 (40%) 1 (20%)
≥7 3 (60%) 4 (80%) 1.00
b
S age
OCPCa 2 (40%) 2 (40%)
EPCa 3 (60%) 3 (60%) 1.00
b
Smoking s a us
Yes 1 5
No 4 0 0.05
b
PSA (ng/ml) 10.7 ± 2.7 12.1 ± 3.23 0.74
a
Da a a e p esen ed as mean ± SD o numbe (%). Signi ican di e ence
be ween OB/OW and NW was e alua ed using
a
es and
b
Fishe ’s exac es
OB/OW obese/o e weigh , NW no mal weigh , BMI body mass index; PSA,
p os a e speci ic an igen; PCa, p os a e cance ; OCPCa, o gan-con ined p os a e
cance ; EPCa, ex a-p os a ic PCa
Cheng e al. Clinical Epigene ics (2018) 10:54 Page 2 o 15
con ains 485,577 p obes, which co e 21,231 (99%)
Re Seq genes. B ie ly, DNA was bisul i e-con e ed using
he EZ DNA me hyla ion ki (Zymo Resea ch, O ange,
CA, USA) and hen used on he In inium Assay®
ollowed by he In inium HD Assay Me hyla ion P o ocol
(Illumina). The imaging da a on he BeadChips was
cap u ed by Illumina iScan sys em.
Da a il e ing and no maliza ion
Raw me hyla ion le el o each p obe was ep esen ed
by me hyla ion β alue, which was calcula ed based
on β= in ensi y o he me hyla ed allele/(in ensi y o
he unme hyla ed allele + in ensi y o he me hyla ed
allele + 100). M alues we e he logi ans o ma ion
o β alues based on M=log
2
(β/(1 −β)), which makes
he da a mo e homoscedas ic and app op ia e o
u he bioin o ma ic and s a is ical analysis.
Me hyla ion alues we e no malized using he unc-
ional no maliza ion algo i hm implemen ed in Min i R
package [18]. Quali y con ol was pe o med by exclud-
ing CpG p obes, which a e ound by Chen e al. o be
c oss- eac i e wi h a eas o he genome no a he si e o
in e es [19], as well as con ol p obes and p obes on sex
ch omosomes. We analyzed a o al o 438,458 CpG si es
om he PPAT o 5 OB/OW PCa pa ien s and 5 NW
PCa pa ien s.
Di e en ial me hyla ion analysis
A s a is ical linea modeling app oach was applied o he
de ec ed di e en ially me hyla ed CpG si es (DMCs) asso-
cia ed wi h obesi y in PPAT using he Bioconduc o
“limma”package [20]. Hype - o hypome hyla ion was
de e mined when me hyla ion le els o CpGs inc eased o
dec eased be ween he OB/OW PCa g oup and he NW
PCa g oup based on mean di e en β> 0 o < 0. False
disco e y a e (FDR)-co ec ed P alues we e de e mined
acco ding o he me hod o Benjamin and Hochbe g’s
(BH me hod) mul iple es ing p ocedu e [21].
Di e en ially me hyla ed egions (DMRs) we e iden i ied
using he “Bumphun e ”me hod implemen ed in he
“chAMP”Rpackage wi h he pa ame e s (B= 1000,
useWeigh s = TRUE, minP obes = 10, pickCu o = TRUE,
and o he se ings wi h de aul alues) [22].
The p opo ions o signi ican hype - o hypome hy-
la ed CpGs we e calcula ed and isualized acco ding o
hei ela ion o he nea es genes o o he CpG islands,
sepa a ely. Gene p omo e egion was de ined as 1500
base pai s (bp) and 200 bp ups eam o he ansc ip ion
s a si e (TSS) (TSS1500 and TSS200) [23]. Iden i ied
genes we e selec ed when mo e han wo signi ican ly
hype me hyla ed CpGs we e simul aneously loca ed in
he p omo e egion.
Func ions, pa hway, and ne wo k en ichmen analysis
Gene on ology (GO) and KEGG pa hway en ichmen ana-
lyses we e pe o med o explo e he biological unc ions
o signi ican ly me hyla ed genes using he online bioin o -
ma ic ool En ich [24]. P o ein-p o ein in e ac ion (PPI)
analysis o all DMC- ela ed genes was pe o med using
Ne wo kAnalys acco ding o STRING da abase [25].
Associa ion analysis be ween DNA me hyla ion and
gene exp ession
We ha e p e iously pe o med gene exp ession expe i-
men o he PPAT o he 5 OB/OW PCa pa ien s and
he 5 NW PCa pa ien s using he HG-U133 Plus 2.0
A yme ix GeneChip A ay (A yme ix, San a Cla a,
CA, USA) [4]. Di e en ial gene exp ession (DGE) ana-
lysis be ween he OB/OW PCa pa ien s and he NW
PCa pa ien s was e-pe o med using he Bioconduc o
“limma”package as p e iously desc ibed [4]. Spea man’s
ank co ela ion analysis was pe o med be ween he
me hyla ion p o iles o he hype me hyla ed CpGs and
he gene exp ession p o iles o he genes in PPAT.
Resul s
Clinical cha ac e is ics
Clinical cha ac e is ics o PCa pa ien s in his s udy we e
s a i ied acco ding o obesi y classi ica ion g oups and
a e p esen ed in Table 1. Mean age, PSA le el, Gleason
sum sco e, and cance s age in subjec s wi h PCa we e
simila (P alue > 0.05) be ween OB/OW and NW
g oups. As expec ed, he mean BMI o he OB/OW
g oup was signi ican ly highe han ha o he NW sub-
jec s (P alue < 0.01). All he pa ien s in he OB/OW
g oup a e ex-smoke s o ac i e smoke s, while only one
pa ien in he NW g oup is a smoke (P alue = 0.05).
Epigenome-wide DNA me hyla ion p o iling o PPAT
To s udy he impac o obesi y s a us on DNA me hyla-
ion p o iles and o iden i y di e en ially me hyla ed CpG
si es in PPAT om OB/OW and NW p os a e cance
pa ien s, we conduc ed epigenome-wide DNA me hyla-
ion analyses. A lowcha o he da a analysis is depic ed
in Addi ional ile 1: Figu e S1. A e quali y con ol and
il e ing, he In inium a ay gene a ed me hyla ion da a
o 438,458 CpG si es, om which 5526 we e di e en ially
me hyla ed a e FDR con ol in he PPAT o OB/OW
PCa pa ien s compa ed o NW (adjus ed P alue < 0.25;
Addi ional ile 2: Table S1 and Table 2). The unsupe ised
hie a chical clus e ing o DMCs showed di e en ial DNA
me hyla ion pa e ns in PPAT be ween OB/OW and NW
samples (Addi ional ile 3: Figu e S2). The majo i y o
DMCs we e hype me hyla ed (n= 4985, 90.2%), wi h 9.8%
hypome hyla ed CpG si es (n= 541) in OB/OW e sus
NW p os a e cance pa ien s (Fig. 1a, b, c).
Cheng e al. Clinical Epigene ics (2018) 10:54 Page 3 o 15
Ch omosomal dis ibu ion o he DMCs
To u he explo e he me hyla ion p o ile, we in es i-
ga ed he ch omosome dis ibu ion o DMCs. Resul s
showed ha hype me hyla ed CpG si es we e loca ed a
ch omosomes 1, 6, 11, and 17 (p opo ion > 6%, Fig. 1d)
and hypome hyla ed CpG si es we e loca ed a ch omo-
somes 1, 2, 6, 7, and 11 (p opo ion > 6%, Fig. 1e).
Me hyla ion a ia ions o hype me hyla ed DMCs and
hypome hyla ed DMCs we e ound mainly dis ibu ed
on ch omosomes 1, 6, and 11, sugges ing ha he DNA
me hyla ion al e a ions in hese ch omosomes we e
co ela ed wi h he body weigh changes in p os a e
pa ien s. Fu he mo e, we compa ed he dis ibu ion o
he DMCs (hype - and hypome hyla ed, sepa a ely) wi h
he dis ibu ion o all e alua ed CpG si es based on hei
ela ion o nea es gene egions (Fig. 2 , Addi ional ile 4:
Table S2) o hei ela ion o CpG islands (Fig. 2g,
Addi ional ile 5: Table S3) using χ
2
es . The esul s
showed ha hype me hyla ed CpGs a e mainly loca ed a
TSS1500 ( ansc ip ion s a si es 1500), IGR (in e genic
Table 2 Di e en ially me hyla ed CpG si es in PPAT be ween obese/o e weigh PCa pa ien s and no mal weigh con ols
P obe ID Ch omosome and coo dina e
(GRCh37)
Nea es gene Rela ion o
gene egion
Rela ion o
CpG island
DNAm β
di e ence (%)
P alue Adjus ed P alue
(< 0.25)
Hype me hyla ed CpG si es
cg09476130 ch 1:159870086 CCDC19 TSS200 Island 12.1 1.87E−03 0.213
cg21293934 ch 18:14748230 ANKRD30B TSS200 Island 11.2 1.83E−03 0.212
cg16925210 ch 2:216946718 PECR TSS200 Island 11.2 2.44E−03 0.226
cg11625005 ch 5:1295737 TERT TSS1500 Island 11.1 1.38E−03 0.196
cg07039560 ch 5:140683681 SLC25A2 TSS200 Island 10.5 2.24E−03 0.222
cg00329447 ch 8:145028170 PLEC1 TSS200 Island 10.1 3.41E−03 0.244
cg24463471 ch 1:25257978 RUNX3 TSS1500 Island 9.9 3.58E−04 0.155
cg26149485 ch 19:2428350 TIMM13 TSS1500 Island 9.7 3.36E−04 0.154
cg05156901 ch 22:51016646 CPT1B TSS200 Island 9.3 3.05E−03 0.238
cg18689454 ch 21:45705694 AIRE TSS200 Island 9.3 7.33E−04 0.174
cg01454592 ch 3:49236800 CCDC36 TSS200 Island 9.3 2.89E−03 0.236
cg24041556 ch 19:10736059 SLC44A2 TSS200 Island 9.1 2.15E−05 0.110
cg22257574 ch 9:135754383 C9o 98 TSS200 Island 9.0 1.97E−05 0.110
cg23005885 ch 15:90543450 ZNF710 TSS1500 Island 8.9 6.20E−04 0.169
cg05726756 ch 17:46608288 HOXB1 TSS200 Island 8.6 1.67E−03 0.206
cg12782180 ch 7:127880932 LEP TSS1500 Island 8.5 9.78E−04 0.184
cg04675542 ch 5:150284416 ZNF300 TSS200 Island 8.4 3.42E−03 0.244
cg10134527 ch 6:33283015 TAPBP TSS1500 Island 8.4 3.19E−03 0.241
cg23387569 ch 12:58120011 LOC100130776 TSS200 Island 8.4 3.39E−04 0.155
cg17205324 ch 14:23835595 EFS TSS1500 Island 8.3 1.46E−04 0.133
cg24402300 ch 19:55591437 EPS8L1 TSS1500 Island 8.2 2.02E−03 0.216
cg18081258 ch 14:21494161 NDRG2 TSS1500 Island 8.2 1.05E−03 0.187
cg00730561 ch 10:102279703 SEC31B TSS200 Island 8.1 4.82E−04 0.162
cg17791651 ch 1:38513489 POU3F1 TSS1500 Island 8.0 1.31E−04 0.133
Hypome hyla ed CpG si es
cg03462171 ch 16:1664488 CRAMP1L TSS200 Island −8.2 2.30E−03 0.223
cg11648730 ch 5:92907151 FLJ42709 TSS1500 Island −6.4 1.72E−03 0.207
cg04558166 ch 1:210001279 C1o 107 TSS200 Island −4.0 1.75E−03 0.209
cg25472897 ch 8:145560555 SCRT1 TSS1500 Island −3.1 1.53E−03 0.201
cg17612948 ch 5:110427863 WDR36 TSS200 Island −3.0 1.70E−03 0.207
cg21665057 ch 3:196295764 WDR53 TSS1500 Island −2.4 2.42E−03 0.225
cg12683173 ch 7:69063404 AUTS2 TSS1500 Island −1.4 3.30E−03 0.242
cg04872557 ch 1:76190008 ACADM TSS200 Island −1.0 1.96E−03 0.215
Cheng e al. Clinical Epigene ics (2018) 10:54 Page 4 o 15
egion), N-sho e, and S-sho e, and hypome hyla ed CpGs
a e mos ly loca ed a he gene body and open sea.
Func ional en ichmen analysis o signi ican ly
obesi y-associa ed DMCs
To in es iga e he po en ial biological ele ance o he sig-
ni ican DMCs, we u he il e ed 483 DMCs (dis ibu ed
wi hin 413 genes) om a o al o 5526 DMCs acco ding
o hei loca ions a bo h he gene p omo e and CpG
island (Addi ional ile 6: Table S4). Fou hund ed se en y-
i e o he 483 DMCs ( ep esen ing 404 genes) we e
hype me hyla ed. Func ional en ichmen analysis o he
hype me hyla ed genes showed ha hese genes we e
en iched o biological p ocesses, such as pa e n
speci ica ion p ocess, neu on di e en ia ion, neu on a e
speci ica ion, and nega i e egula ion o phospha e
me abolic p ocess (adjus ed P alue < 0.05, Addi ional ile 7:
Table S5), as well as molecula unc ions, such as
Fig. 1 Epigene ic p o iles o di e en ially me hyla ed CpGs o PPAT be ween OB/OW and NW g oups. Manha an plo s show epigene ic p o iles o all
inc eased me hyla ed CpGs (a) and all dec eased me hyla ed CpGs (b). The X-axis shows ch omosomes, and he Y-axisisa−log10 (P alue). The black line
ep esen s he h eshold o adjus ed P alue = 0.25. CpGs abo e he black line a e signi ican ly hype - o hypome hyla ed. The olcano plo o DNA
me hyla ion (c) shows a signi ican di e ence in PPAT be ween he OB/OW and NW g oups. Fou housand nine hund ed eigh y- i e hype me hyla ed CpGs
a e labeled in ed, and 541 hypome hyla ed CpGs a e labeled in g een (adjus ed P alue > 0.25). The p opo ions o hype - and hypome hyla ed CpGs on
each ch omosome a e shown in (d)and(e). The black line indica es i he p opo ions o hype -andhypome hyla edCpGs on a ch omosome a e highe
han 6%. The dis ibu ion o signi ican DMCs (hype - o hypome hyla ed CpGs) and globe DNA me hyla ion CpGs in loca ions ela ed o he nea es gene
egions and CpG islands a e shown in and g. Hype me hyla ed CpGs a e mainly loca ed a TSS1500 ( ansc ip ion s a si es 1500), IGR
(in e genic egion), N-sho e ( he 2 kb egions ups eam o he CpG islandbounda ies),andS-sho e( he2kb egions downs eam o he CpG island
bounda ies), and hypome hyla ed CpGs a e mos ly loca ed a he gene body and open sea. The di e ence o he p opo ion o CpGs among he h ee
CpG g oups was calcula ed based on he χ
2
es (*P< 0.05, **P< 0.01, ***P< 0.001, ns no signi ican ). CpG islands we e de ined as DNA sequences (500
base windows; excluding mos epe i i e Alu-elemen s) wi h a GC base composi ion g ea e han 50% and a CpG obse ed/expec ed a io o mo e han
0.6. The 2 kb egions immedia ely ups eam (N_Sho e) and downs eam (S_Sho e) o he CpG island bounda ies we e de ined as “CpG island sho es,”and
he 2 kb egions ups eam (N_Shel ) and downs eam (S_Shel ) o he CpG island sho es we e e e ed as “CpG island shel es.”Open seas we e he egions
mo e han 4 kb om CpG islands
Cheng e al. Clinical Epigene ics (2018) 10:54 Page 5 o 15

neu opep ide ecep o ac i i y and sequence-speci ic
DNA-binding RNA polyme ase II ansc ip ion ac o ac-
i i y (adjus ed P alue < 0.1, Addi ional ile 8: Table S6).
KEGG pa hway en ichmen analysis showed ha hype -
me hyla ed genes we e in ol ed in signaling pa hways
egula ing plu ipo ency o s em cells, a y acid me abol-
ism, basal cell ca cinoma, non-alcoholic a y li e disease
(NAFLD), and AMPK signaling pa hway (P alue < 0.05,
Addi ional ile 9: Table S7).
We mapped he 404 hype me hyla ed genes o he
STRING da abase and gene a ed a p o ein-p o ein in e -
ac ion (PPI) ne wo k by he Ne wo kAnalys . The la ges
subne wo k was iden i ied o include 247 nodes (genes)
and 403 edges (Fig. 2a). In he ne wo k, he size o he
nodes was based on hei deg ee alues and he colo o
nodes was based on hei P alues. This ne wo k con-
ained 118 seed genes om he DMCs, and he en ich-
men pa hway analysis showed ha he genes o he
Fig. 2 P o ein-p o ein in e ac ion analysis. aA subne wo k composing o 247 nodes and 403 edges was gene a ed using me hyla ed genes. Blue do s
ep esen he genes in ol ed in p os a e cance ; ed and pink do s ep esen he seeds (me hyla ed genes) acco ding o he di e en P alues; he
g ay do s ep esen he p o eins which we e closely in e ac ed wi h he seeds, and he ci cle size ep esen s he node deg ee. bThe pa hway
en ichmen analysis shows he subne wo k is mainly en iched in cance pa hways (P<0.0001)
Cheng e al. Clinical Epigene ics (2018) 10:54 Page 6 o 15
subne wo k we e mos ly in ol ed in he pa hways o p os-
a e cance and o he cance s (Fig. 2b,Addi ional ile10:
Table S8, adjus ed P alue < 0.05). Pa icula ly, he gene
UBC (ubiqui in C) was ound o be a hub connec ing wi h
many o he nodes in he ne wo k, sugges ing ha he
gene may play impo an biological oles in he PPAT o
obese PCa pa ien s.
Selec ed genes wi h mul iple me hyla ed CpG si es
In o de o explo e ep ession o genes by DNA me hy-
la ion modi ica ions, we selec ed genes which had mul-
iple hype me hyla ed CpG si es ( he numbe o
me hyla ed CpG si es ≥2, in a leas one o he si es wi h
ameandi e enceo β> 3% and an adjus ed P alue < 0.25)
(Addi ional ile 1: Figu e S1 and Addi ional ile 11:TableS9).
A o al o 38 genes wi h 100 di e en ially me hyla ed CpG
si es we e selec ed, which included TAPBP,RUNX3,CPT1B,
CPT1C,MOGAT3,WNT2,andAIRE (Addi ional ile 11:
Table S9). No ably, he p omo e egion o TAPBP (TAP-
binding p o ein) had eigh hype me hyla ed CpG si es in he
p omo e (Fig. 3a), which we e signi ican ly mo e me hyla ed
in he OB/OW han hose in he NW g oups (Fig. 3b), wi h
ameandi e enceo β alue g ea e han 5%
(Addi ional ile 10: Table S8). Spea man’s ank co ela ion
showed s ong associa ion (
2
=0.73–0.97) o he eigh
hype me hyla ed CpGs in he TAPBP p omo e wi h hei
me hyla ion le els (Fig. 3c). Pa hway analysis o hese genes
e ealed en ichmen o a y acid me abolism, PPAR
signaling pa hway, glucagon signaling pa hway, AMPK
signaling pa hway, glyce olipid me abolism, basal cell
ca cinoma, an igen p ocessing and p esen a ion, ECM
ecep o in e ac ion, and insulin esis ance (adjus ed
P alue < 0.25) (Addi ional ile 12: Table S10).
Di e en ial me hyla ed egions analysis
Ten DMRs we e iden i ied (P< 0.01) in obesi y PPAT
samples compa ed o no mal weigh con ols (Table 3).
The size o he DMRs a ied om 161 o 1287 bp. No e-
wo hy, ou ou o he en DMRs we e disco e ed on
ch omosome 6. Eigh egions we e loca ed in genes, and
wo we e in he in e genic egion. Fou egions we e in
he gene p omo e o FAM104A,C17o 80,HOXA4A,
and TAPBP.
Associa ion analysis be ween DNA me hyla ion and
mRNA exp ession
Inc eased DNA me hyla ion o p omo e in CpG islands
was ob iously linked o gene ansc ip ional silencing [26].
The e o e, we ela ed hype me hyla ed CpG si es in PPAT
wi h genes showed dec eased gene exp ession le el om
ou p e iously gene a ed mRNA exp ession da a [4]. DNA
me hyla ion o 16 CpG si es, co esponding o 14 genes,
was associa ed wi h signi ican ly dec eased ansc ip s in
OB/OW g oup (P alue < 0.05) (Table 4). The Spea man’s
ank co ela ion analysis showed ha eigh o he 14 genes
ha e signi ican ly nega i e associa ion (P alue < 0.05) be-
ween he me hyla ion p o iles and he gene exp ession
p o ileso hesegenes(Table4). The ep ession genes we e
mainly in ol ed in me abolic pa hways (Addi ional ile 13:
Table S11, adjus ed P alue < 0.25), such as MOGAT1
(glyce olipid me abolism), FADS1 ( a y acid me abolism
and biosyn hesis o unsa u a ed a y acids), and PCYT2
(glyce ophospholipid me abolism). The mRNA exp ession
le el o FADS1 was signi ican ly dec eased in he PPAT o
obese wi h p os a e cance s in ou p e ious s udy using
qRT-PCR [4]. Besides hese, GO en ichmen analysis
showed ha hese genes a e unc ionally ela ed o ecep o
binding (neu opep ide ecep o binding, dopamine ecep o
binding, and insulin ecep o binding) and enzyme ac i i y
(acid phospha ase ac i i y, me alloca boxypep idase
ac i i y, and acylglyce ol O-acyl ans e ase ac i i y)
(Addi ional ile 14: Table S12, adjus ed P alue < 0.25).
Discussion
This pilo s udy e ealed signi ican di e ences o DNA
me hyla ion p o iles be ween he PPATs om OB/OW
e sus NW PCa pa ien s. Va ia ions in global DNA
me hyla ion demons a ed ha excess adiposi y played
an impo an ole in DNA me hyla ion le el o PPAT
issues in p os a e cance pa ien s, which p o ide an
oppo uni y o explo e he e ec o obesi y on PPAT
epigene ic modi ica ion and subsequen ly on p os a e
cance . These indings epo ed o he i s ime in
PPAT depo a e in conco dance wi h p e ious wo ks
epo ing ha excess adiposi y and BMI ac i a e DNA
me hyla ion in adipose issue [27–29]. Thus, conside ing
he p esen unde s anding o he po en ial causal
ela ionship be ween excess adiposi y and cance [30],
diabe es [11], and ca dio ascula disease [31], ou esul s
p o ide me hyla ed candida e genes, which migh os e
esea ch on he po en ial biological mechanisms unde -
lying epigene ic egula ion o PPAT by excess adiposi y
and p os a e cance .
Gi en ha DNA me hyla ion o CpGs loca ed a p o-
mo e s and islands a e associa ed wi h gene ansc ip-
ion silencing, we pe o med a s ic il e ing o DMCs
and explo ed he biological unc ions o all p omo e
hype me hyla ed genes, aiming o ind he c i ical me h-
yla ed CpGs in he PPAT be ween he obese and no mal
weigh PCa pa ien s. Bioin o ma ic analysis showed ha
he en iched pa hways we e mos ly in ol ed in me a-
bolic diso de s, pa icula ly a y acid deg ada ion and
glyce olipid and choline me abolism. These pa hways a e
known o media e he p o- umo al e ec o whi e
adipose issue in umo s, hus con ibu ing o umo i-
genesis and me as asis [32,33], pa icula ly in p os a e
cance [5]. Findings om o he oncological models high-
ligh excess adiposi y-associa ed impac in me hyla ion
Cheng e al. Clinical Epigene ics (2018) 10:54 Page 7 o 15
Fig. 3 Visualiza ion and analysis o hype me hyla ed CpG si es in TAPBP p omo e .aThe ch omosome posi ions o hype me hyla ed CpG si es
show ha DMCs a e loca ed a ch 6 (p21.32), which is in he egion o TSS1500 (pink squa e) o TAPBP and a he loca ion o CpG island 31.
bMe hyla ion le els o eigh CpG si es in PPAT om OB/OW and NW PCa pa ien s ha e shown a signi ican di e ence. cCo ela ion analysis
shows s ong co ela ion (Spea man co ela ion coe icien
2
0.73~ 0.97) be ween he eigh hype me hyla ed CpG si es based on he mean
β alue di e ence o indi idual p obes
Cheng e al. Clinical Epigene ics (2018) 10:54 Page 8 o 15
ma ke s known o associa e wi h po en ial e ec in he
cance mic oen i onmen (e.g., a oma ase, p os aglandin
E
2
ecep o in b eas cance ) [34,35]. Obesi y has also
been shown o associa e wi h me hyla ion o cance -
ela ed genes (E-cadhe in, p16, and RAR-β(2)) di ec ly in
malignan b eas cells [36,37].
Pa hway en ichmen analysis showed a s ong associ-
a ion be ween p omo e hype me hyla ion o CPT1B,
CPT1C,ACADM, and FADS1, wi h a y acid me abol-
ism. CPT1B (ca ni ine palmi oyl ans e ase 1B) and
CPT1C (ca ni ine palmi oyl ans e ase 1C) genes encode
a e-limi ing enzymes in a y acid deg ada ion and play
c i ical oles in long-chain a y acid (LCFA) β-oxida ion
by con olling anspo a ion o long-chain a y acyl-
CoAs om he cy oplasm ac oss he ou e mi ochond ia
memb ane [38]. Maple e al. epo ed ha inc eased
me hyla ion o speci ic CpGs in he CPT1B p omo e
was co ela ed wi h dec eased CPT1B ansc ip s in he
skele al muscle a e lipid o e supply in se e e obesi y,
which esul ed in obese indi idual’s incapaci y o
inc ease a oxida ion, con ibu ing o me abolic in lex-
ibili y [39]. Al hough he biochemical unc ion o CPT1C
has been e i ied o be necessa y o he egula ion o
ene gy homeos asis in CPT1C knockou mouse b ain
[40], he s udy o CPT1C me hyla ion was absen .
CPT1B and CPT1C we e p e iously epo ed o be
highly exp essed in he muscle, b ain, and many o he
no mal issues including adipocy es [41]. Taken oge he ,
hese indings sugges ha me hyla ion o speci ic CpG si es
in he CPT1B and CPT1C p omo e s likely esul in gene
Table 3 Di e en ially me hyla ed egions (DMR) in PPAT o obese/o e weigh PCa pa ien s compa ed o no mal weigh con ols
DMR Ch S a –end (bp) Size (bp) P alue FDR Loca ed gene DMCs* Rela ion o CpG island
1 6 30,038,791–30,039,801 1010 5.11E−05 2.07E−02 RNF39 37(0) Island
2 6 29,648,161–29,649,084 923 1.54E−03 2.08E−01 ZFP57
#
22(0) Open sea
3 17 71,228,123–71,228,832 709 3.10E−03 2.86E−01 FAM104A 4(3) Island
4 17 71,228,123–71,228,832 709 3.10E−03 2.86E−01 C17o 80 10(8) Island
5 12 42,720,006–42,720,167 161 5.56E−03 2.86E−01 PPHLN1 4(0) Island
6 6 31,650,735–31,651,158 423 5.59E−03 2.86E−01 MIR4646
#
16(0) Island
7 7 27,169,674–27,170,961 1287 5.66E−03 2.86E−01 HOXA4 17(11) Island
8 6 33,282,736–33,283,145 409 5.87E−03 2.86E−01 TAPBP 18(18) Island
9 20 57,463,763–57,464,129 366 6.74E−03 2.86E−01 GNAS 15(0) Island
10 16 86,546,938–86,547,322 384 7.03E−03 2.86E−01 FOXF1 4(0) Sho e
#The DMR is loca ed a he in e genic egion
Ch ch omosome
*The numbe in he b acke is he quan i ies o DMCs loca ed a he p omo e (TSS200 and TSS1500) egions
Table 4 Genes hype me hyla ed in p omo e s wi h signi ican ly dec eased gene exp ession
Gene symbol DNA me hyla ion Gene exp ession Co ela ion analysis
P obe ID DNAm β
di . (%)
Adjus ed P alue
(< 0.25)
P obe ID FC P alue (< 0.05) Spea man’s ank
co ela ion coe icien
P alue (< 0.05)
UCN cg20442078 5.6 0.17 8051061 −1.12 3.61E−02 −8.42E−01 2.23E−03
CCHCR1 cg00160818 1.9 0.17 8124868 −1.14 1.73E−02 −7.45E−01 9.21E−03
CRB3 cg14782015 4.3 0.20 8025041 −1.13 1.84E−02 −7.21E−01 1.21E−02
AGBL4 cg21834207 3.2 0.13 7915971 −1.17 1.29E−02 −6.73E−01 1.97E−02
INSL3 cg10174482 4.2 0.13 8035345 −1.13 4.94E−02 −6.36E−01 2.72E−02
ANKRD30B cg21293934 11.2 0.21 8069499 −1.17 2.24E−02 −6.08E−01 3.11E−02
FADS1 cg16213375 3.6 0.16 7948612 −1.8 9.55E−04 −5.88E−01 4.01E−02
PAPL cg18481683 2.3 0.24 8028570 −1.19 1.45E−02 −5.52E−01 5.21E−02
MOGAT1 cg12678667 4 0.15 8048725 −1.28 3.87E−02 −4.67E−01 8.91E−02
PPP1R1B cg09762778 5 0.12 8006865 −1.27 4.74E−02 −4.67E−01 8.91E−02
PRUNE2 cg00390775 4.4 0.15 8161884 −1.31 1.91E−02 −3.82E−01 1.39E−01
CIDEA cg18309817 1.8 0.18 8020211 −1.32 3.21E−02 −2.97E−01 2.03E−01
PCYT2 cg19583655 6.2 0.21 8019280 −1.26 6.38E−04 −1.88E−01 3.04E−01
SCUBE1 cg07697597 1.7 0.23 8076586 −1.23 9.03E−03 −4.24E−02 4.59E−01
FC old change, DNAm βdi . DNAm βdi e ence
Cheng e al. Clinical Epigene ics (2018) 10:54 Page 9 o 15