ARTICLE
Placen al DNA me hyla ion signa u es o ma e nal
smoking du ing p egnancy and po en ial impac s
on e al g ow h
Todd M. E e son 1,38✉, Ma a Vi es-Usano2,3,4,38, Emie Sey e5,38, And es Ca denas6,7,
Ma ina Lacasaña 4,8,9, Je ey M. C aig10,11,12, Co ina Lesseu 13, Emily R. Bake 14,
No a Fe nandez-Jimenez 15,16,17, Ba ba a Heude 18, Pa ice Pe on19, Bea iz Gónzalez-Alzaga8,9,
Jane Halliday20,11, Maya A. Deyssen o h13, Ma ga e R. Ka agas21, Ca men Íñiguez4,22,23, Luigi Boucha d24,
Ped o Ca mona-Sáez 25,26, Yuk J. Loke10,11, Ke Hao 27, Thalia Belmon e 28, Ma ie A. Cha les18,
Jo di Ma o ell-Ma ugán 25,29, E elyne Muggli 11,20, Jia Chen13, Ma iana F. Fe nández 4,9,30,
Jo g Tos 31, An onio Gómez-Ma ín 32, S ephanie J. London 33, Jo di Sunye 3,4,34,35,
Ca men J. Ma si 1,36, Johanna Lepeule 5,38, Ma ie-F ance Hi e 6,37,38 & Ma iona Bus aman e 2,3,4,34,38✉
Ma e nal smoking du ing p egnancy (MSDP) con ibu es o poo bi h ou comes, in pa
h ough dis up ed placen al unc ions, which may be eflec ed in he placen al epigenome.
He e we p esen a me a-analysis o he associa ions be ween MSDP and placen al DNA
me hyla ion (DNAm) and be ween DNAm and bi h ou comes wi hin he P egnancy And
Childhood Epigene ics (PACE) conso ium (N=1700, 344 wi h MSDP). We iden i y 443
CpGs ha a e associa ed wi h MSDP, o which 142 associa ed wi h bi h ou comes, 40
associa ed wi h gene exp ession, and 13 CpGs a e associa ed wi h all h ee. Only wo CpGs
ha e consis en associa ions om a p io me a-analysis o co d blood DNAm, demons a ing
subs an ial issue-specific esponses o MSDP. The placen al MSDP-associa ed CpGs a e
en iched o en i onmen al esponse genes, g ow h- ac o signaling, and inflamma ion, which
play impo an oles in placen al unc ion. We demons a e links be ween placen al DNAm,
MSDP and poo bi h ou comes, which may be e in o m he mechanisms h ough which
MSDP impac s placen al unc ion and e al g ow h.
h ps://doi.o g/10.1038/s41467-021-24558-y OPEN
A ull lis o au ho a filia ions appea s a he end o he pape .
NATURE COMMUNICATIONS | (2021) 12:5095 | h ps://doi.o g/10.1038/s41467-021-24558-y | www.na u e.com/na u ecommunica ions 1
1234567890():,;
Almos one in en mo he s smoke du ing p egnancy, wi h
s a e-specific p e alence anging om as low as 1.8% o as
high as 27.1% in he USA1, while in Eu ope, he p e-
alence o ma e nal smoking du ing p egnancy (MSDP) anges
be ween 4.2 and 18.9% ( e . 2). Consequen ly, he nume ous
heal h e ec s o MSDP con ey a significan public heal h conce n.
The impac o his exposu e on e al de elopmen has been he
sou ce o significan in es iga ion, esul ing in MSDP being
ecognized as a cause o mul iple nega i e p egnancy and bi h
ou comes3.
The mechanisms ha unde lie his ep oduc i e and de elop-
men al oxici y a e pa ially unde s ood, and include molecula
and ana omical changes o he placen a4,5. In addi ion, expe i-
men al mouse models ha e ecen ly highligh ed he c i ical oles
o p ope placen al unc ion in ensu ing success ul p egnancy
ou comes6. Epigene ic esponses o p ena al exposu es ha e
eme ged as po en ial in e media e links be ween ea ly li e expo-
su es and de elopmen al heal h ou comes, and epidemiologic
s udies o DNAm, pa icula ly mul i-coho collabo a i e e o s,
a e powe ul app oaches o in es iga e hese ypes o esea ch
ques ions7. Mos s udies o MSDP and epigene ics ha e ocused
on DNA me hyla ion (DNAm) in co d blood, al hough some
s udies o placen a, pe iphe al blood, and lung issues ha e also
been pe o med8. The P egnancy and Childhood Epigene ics
(PACE) conso ium9published a la ge me a-analysis iden i ying
housands o MSDP-associa ed a ia ions in DNAm wi hin co d
blood and child pe iphe al blood10. Howe e , he placen al epi-
genome has no been as ho oughly s udied, al hough he pla-
cen a is likely a c i ical a ge o gan o MSDP-associa ed oxici y.
A hand ul o p io s udies ha e examined he ela ionships
be ween MSDP and (DNAm) in human placen a11–14, iden i ying
MSDP-associa ed CpGs some o which ha e been sugges ed
o pa ially media e he e ec s o MSDP on lowe bi h weigh
(BW)15.
These s udies ha e begun o cha ac e ize he impac ha
MSDP has on he human placen al epigenome, bu ha e been
limi ed by small sample sizes. We aimed o add ess his gap by
pe o ming a fixed e ec s me a-analysis examining he ela ion-
ships be ween MSDP and a ia ions in he placen al me hylome
ac oss se en independen s udies ha a e membe s o he PACE
conso ium. We also aimed o gain insigh s in o he po en ial
biological p ocesses ha migh be a ec ed and he ela ionships
wi h bi h ou comes by pe o ming addi ional analyses wi h
nea by mRNA exp ession, as well as unc ional, egula o y, and
pheno ypic en ichmen , and a seconda y me a-analysis o he
associa ions be ween placen al DNAm and bi h ou comes. He e,
we demons a e ha placen al DNAm is associa ed wi h MSDP
h oughou he genome, and hese di e ences in DNAm a e
associa ed wi h ges a ional age (GA) and bi h size me ics. We
also show ha many o hese CpGs a e co ela ed wi h nea by
gene exp ession and a e en iched o genes in ol ed in en i -
onmen al esponse, g ow h ac o signaling, and inflamma ion.
Fu he s udies a e needed o assess po en ial causali y and
media ion.
Resul s
S udy popula ion. Se en Ame ican, Aus alian, and Eu opean
s udies (N=1700) con ibu ed o he epigenome-wide associa-
ion s udy (EWAS) linking MSDP o placen al DNAm: including
Asking Ques ions abou Alcohol in p egnancy (AQUA)16, S udy
on he p ena al and ea ly pos na al de e minan s o child heal h
and de elopmen (EDEN)17, Gene ics o Glucose egula ion in
Ges a ion and G ow h (Gen3G)18, Gene ics, Ea ly Li e En i on-
men al Exposu es, and In an De elopmen in Andalusia
(GENEIDA), En i onmen and Childhood P ojec (INMA)19,
New Hampshi e Bi h Coho S udy (NHBCS)20, and Rhode
Island Child Heal h S udy (RICHS)21. Fo his me a-analysis, 344
(20.2%) mo he s epo ed any MSDP, defined as any ciga e e
smoking du ing any imes e o p egnancy. Any MSDP ended
o be less p e alen in he coho s om Canada and he USA
compa ed o hose om Aus alia and Eu ope (Table 1). Th ee
coho s (N=795, EDEN, GENEIDA, and INMA) con ibu ed o
he EWAS o sus ained MSDP, defined as ma e nal smoking
h oughou p egnancy, among which 163 (20.5%) mo he s
epo ed sus ained MSDP. Dis ibu ions o co a ia es by coho
a e p o ided in he Supplemen a y Ma e ials (Supplemen a y
Da a 1). The compa ison g oup o models o any o sus ained
MSDP included all mo he s ha did no epo smoking ciga -
e es du ing any imes e .
Genome-wide DNAm me a-analyses. We p oduced ou s a is-
ical models o each CpG si e, eg essing DNAm on bo h any
and sus ained MSDP, wi h and wi hou adjus men o pu a i e
cellula he e ogenei y, which was es ima ed wi h a e e ence- ee
decon olu ion algo i hm, Re F eeCellMix22. All models we e
adjus ed o ma e nal age, pa i y, and ma e nal educa ion.
Genomic infla ion ac o s om he coho -specific models ( an-
ging om λ=0.824 o 2.478) and me a-analyses ( anging om
λ=2.002 o 2.839; Supplemen a y Da a 2 and Supplemen a y
Fig. 1) e ealed po en ial esidual con ounding and infla ion
o es s a is ics. All EWAS esul s wi h Bon e oni-co ec ed
p alues < 0.05 a e included in he Supplemen a y Ma e ials
(Supplemen a y Da a 3–6). He e ogenei y in he associa ions
ac oss coho s was lowes o he models ha we e adjus ed o
Re F eeCellMix (Supplemen a y Da a 7), and hus we u ilized he
esul s om hese adjus men models o all downs eam ana-
lyses. To co ec o esidual bias and infla ion, we, hen, imple-
men ed BACON, which es ima es an empi ical null
dis ibu ion o he da a23. This subs an ially educed infla ion
Table 1 F equencies o any and sus ained MSDP wi hin pa icipa ing coho s.
Coho Coun y Any MSDP Sus ained MSDP
To al NN(%) nonsmoke s N(%) smoke s To al NN(%) nonsmoke s N(%) smoke s
AQUA Aus alia 99 75 (75.76%) 24 (24.24%) —— —
EDEN F ance 647 446 (68.93%) 201 (31.07%) 570 446 (68.93%) 124 (21.75%)
Gen3G Canada 151 138 (91.39%) 13 (8.61%) —— —
GENEIDA Spain 87 67 (77.01%) 20 (22.99%) 82 67 (77.01%) 15 (18.29%)
INMA Spain 166 119 (71.69%) 47 (28.31%) 143 119 (71.69%) 24 (16.78%)
NHBCS USA 310 290 (93.55%) 20 (6.45%) —— —
RICHS USA 240 221 (92.08%) 19 (7.92%) —— —
TOTAL 1700 1356 (79.76%) 344 (20.24%) 795 632 (79.50%) 163 (20.50%)
MSDP ma e nal smoking du ing p egnancy.
ARTICLE NATURE COMMUNICATIONS | h ps://doi.o g/10.1038/s41467-021-24558-y
2NATURE COMMUNICATIONS | (2021) 12:5095 | h ps://doi.o g/10.1038/s41467-021-24558-y | www.na u e.com/na u ecommunica ions
among coho -specific esul s ( anged 0.957–1.228), and me a-
analyses o hese BACON-co ec ed es ima es yielded 443 CpGs
ha we e associa ed wi h any o sus ained MSDP a e Bon e oni
co ec ion (Supplemen a y Da a 8 and Fig. 1). While 443 CpGs is
a la ge numbe o si es o iden i y in an EWAS, his is simila in
size o he 568 CpGs iden ified in a p e ious PACE me a-analysis
o co d blood DNAm associa ed wi h MSDP10. Sus ained MSDP
yielded simila , bu s onge e ec s o 93% o hese 443 CpGs
(Supplemen a y Fig. 2).
We did no adjus o GA because we hypo hesized ha i was
likely downs eam o he epigene ic esponse o MSDP o
possibly on he causal pa hway be ween MSDP and changes in
DNAm; in ei he case, i would likely esul in o e adjus men 24.
Howe e , we did explo e whe he addi ional adjus men o GA
al e ed he obse ed ela ionships in ou o ou coho s (EDEN,
INMA, NHBCS, and RICHS). We ound ha hese addi ional
adjus men s had almos no e ec on he es ima es o di e en ial
DNAm, wi h no appa en a enua ion o e ec s owa d he null
(Supplemen a y Fig. 3). Thus, MSDP-associa ed di e ences in
GA could no explain he obse ed ela ionships be ween DNAm
and MSDP.
The mos no able associa ion was obse ed a cg27402634,
loca ed ups eam o he LEKR1 gene and he noncoding RNA
LINC00886, which showed he la ges di e en ial DNAm and
smalles p alues in all me a-analyses. Placen as ha we e exposed
o any MSDP had 23.33% lowe DNAm (95% CI: 21.17–25.51%
lowe DNAm; in e se- a iance fixed-e ec me a-analysis p
alue =8.85E−99) and hose exposed o sus ained MSDP had
25.08% lowe DNAm (95% CI: 23.06–27.11% lowe DNAm;
in e se- a iance fixed-e ec me a-analysis p alue =2.20−E130),
when compa ed o mo he s ha did no smoke a all du ing
p egnancy. Al hough all coho s obse ed subs an ial hypo-
me hyla ion wi h MSDP a his CpG, he ac ual es ima es o he
associa ions we e highly a iable be ween coho s o models o
any MSDP (Coch an’sQ es p alue =1.74E−15), bu ela i ely
consis en o models o sus ained MSDP (Coch an’sQ es p
alue =1.61E−01; Fig. 2A). O e all, we obse ed consis ency in
he associa ions ac oss coho s o he as majo i y o he 443
CpGs: 93% and 96% o he 443 CpGs yielded he e ogenei y p
alues > 0.01, o any and sus ained MSDP, espec i ely. In
addi ion o cg27402634, we highligh hose ela ionships ha
yielded ha la ges magni udes o associa ion: |β
Any MSDP
| > 0.05
o cg26843110 (EDC3), cg20340720 (WBP1L), and cg17823829
(KDM5B; Fig. 2B–D). We iden ified nume ous o he no ewo hy
ela ionships bu due o he la ge numbe o genome-wide
significan associa ions, we highligh he ela ionship among he
20 mos s a is ically significan CpGs om he p ima y me a-
analysis o any MSDP going o wa d (Table 2), while esul s o
all BACON-adjus ed genome-wide significan CpGs, along wi h
de ailed anno a ions, a e included in he Supplemen a y Ma e ials
(Supplemen a y Da a 8).
Exp ession quan i a i e ai me hyla ion (eQTM) analyses.We
hen pe o med exp ession quan i a i e ai me hyla ion (eQTM)
analyses, es ing whe he he DNAm le els a MSDP-associa ed
CpGs we e associa ed wi h he exp ession o nea by mRNA
(wi hin 250 kb o he CpG) om 194 placen al samples in he
RICHS coho . We mapped each CpG o he gene ha was mos
s ongly associa ed wi h DNAm le els i he associa ion p oduced
ap alue < 0.05, hen used a Bon e oni-co ec ed h eshold o
iden i y hose eQTMs wi h he s onges e idence o a ela ion-
ship. Among he 421 CpGs ha we e wi hin 250 kb o a an-
sc ip ion s a si e (TSS), 258 CpGs we e mapped o eQTM genes
(p alues < 0.05), 40 o which we e significan a e Bon e oni
co ec ion (α=1.43E−05; Supplemen a y Da a 9). The majo i y
o mapped eQTMs exhibi ed in e se associa ions (65%) and
s a is ical significance was s onges o CpGs ha we e closes o
Fig. 1 Volcano and Manha an plo s o he in e se- a iance fixed me a-analysis esul s o any and sus ained MSDP. A Placen al DNAm associa ions
wi h any MSDP ( o al N=1700 independen samples om se en independen s udies; exposed =344). BPlacen al DNAm associa ions wi h sus ained
MSDP ( o al N=795 independen samples om h ee independen s udies; exposed =163). Fo bo h analyses, models we e adjus ed o ma e nal age,
pa i y, ma e nal educa ion, pu a i e cellula he e ogenei y, and esidual bias. In he olcano plo s, he x-axes show he es ima ed mean di e ence in DNAm
(e ec size), when compa ing mo he s ha smoked du ing p egnancy (MSDP) o hose ha did no , wi h a possible ange be ween 0 and 1, while he x-
axes in he Manha an plo s ep esen genomic loca ion; bo h plo s sha e he same y-axes wi h −log
10
(p alues). Bon e oni h esholds o s a is ical
significance a e shown as blue do s and a blue ho izon al line, o olcano and Manha an plo s, espec i ely. The y-axes we e unca ed o a minimum
p alue o 1 × 10−30 (o maximum −log
10
(p) o 30), o allow o be e isualiza ion o he majo i y o ou esul s.
NATURE COMMUNICATIONS | h ps://doi.o g/10.1038/s41467-021-24558-y ARTICLE
NATURE COMMUNICATIONS | (2021) 12:5095 | h ps://doi.o g/10.1038/s41467-021-24558-y | www.na u e.com/na u ecommunica ions 3
he TSS (Supplemen a y Fig. 4). Among he op 20 CpGs om
ou me a-analysis, 15 mapped o eQTM genes (Table 3) and 5 o
which we e significan a he Bon e oni-adjus ed h eshold:
SH3D21,TBC1D8,USP46,CRAT, and TGFB1.
Func ional and egula o y en ichmen analyses. En ichmen
analyses we e pe o med o gain insigh s in o he biological
p ocesses in placen a ha may be impac ed by MSDP h ough
al e ed epigene ic egula ion. We pe o med gene-se en ichmen
analyses using wo di e en gene se s (Supplemen a y Da a 10),
fi s using he genes anno a ed o he 443 MSDP-associa ed CpGs
acco ding o he Illumina anno a ion file (284 genes), and second
using he genes mapped o hese CpGs ia he eQTM analyses
(211 genes). The fi s gene lis le e ages as much o he da a as
possible ha was gene a ed om ou mos well-powe ed analysis.
Since DNAm is a ela i ely s able epigene ic ea u e, his could
ep esen a eco d o he genes whose me hyla ion le els we e
pe u bed by MSDP ac oss he p egnancy pe iod. On he o he
hand, gene exp ession is mo e dynamic han DNAm and is
a ec ed by mul iple s imuli. Thus, he second eQTM-mapped
gene se likely ep esen s he genes ha a e influenced by
DNAm a hese MSDP-associa ed CpGs a , o close o, he ime
o bi h.
We ound ha 46 and 9 biological pa hways we e significan ly
(q alue < 0.05) en iched among he Illumina anno a ed and
eQTM-mapped genes, espec i ely (Supplemen a y Da a 11
and 12). O e all, he eQTM genes we e in ol ed in inflamma o y
ac i i y (a yl hyd oca bon ecep o (Ah ) pa hway, Th17 cell
di e en ia ion, neu ophil deg anula ion, and pla ele deg anula-
ion), y osine kinase signaling (TYROBP causal ne wo k and
signaling by ecep o y osine kinases), ca cinogenesis (pa hways
in clea cell enal cell ca cinoma), adipogenesis, and pla ele s.
These pa hways included mul iple eQTM-mapped genes ha
we e wi hin he Bon e oni-co ec ed significance h eshold; mos
no ably, TGFB1, was in ol ed in almos all o hese pa hways, and
TNFRSF1B and ACLY, we e each in ol ed in wo o hese
pa hways. The Illumina anno a ed gene se , which included a
la ge numbe o genes anno a ed o mo e MSDP-associa ed
CpGs, was addi ionally en iched wi h nume ous pa hways
Fig. 2 Fo es plo s o coho specific and in e se- a iance fixed-e ec me a-analysis es ima es o associa ions be ween MSDP wi h placen al DNAm.
Es ima ed di e en ial DNAm (Mean Di .) and 95% confidence in e als (95% CI) a Acg27402634, Bcg26843110, Ccg20340720, and Dcg17823829
wi h any MSDP and sus ained MSDP. All models we e adjus ed o ma e nal age, pa i y, ma e nal educa ion, pu a i e cellula he e ogenei y, and o
esidual bias. The mean di e ence ep esen s he es ima ed di e ence in he p opo ion o DNAm a each CpG when compa ing mo he s ha smoked
du ing p egnancy (MSDP), o hose ha did no smoke du ing p egnancy ia linea eg ession.
ARTICLE NATURE COMMUNICATIONS | h ps://doi.o g/10.1038/s41467-021-24558-y
4NATURE COMMUNICATIONS | (2021) 12:5095 | h ps://doi.o g/10.1038/s41467-021-24558-y | www.na u e.com/na u ecommunica ions
in ol ing g ow h ac o signaling (FGFR, EGF-EGFR, and
PDGFR), ho mones (aldos e one, insulin, and TSH), immune
and inflamma o y signaling (IL2 and IL6), MAPK signaling,
myome ial and ascula smoo h muscle con ac ion, signal
ansduc ion, and cance pa hways. While only one pa hway was
en iched among bo h anno a ed and eQTM genes, signaling by
ecep o y osine kinases (anno a ed genes q alue =0.025 and
eQTM-mapped genes q alue =0.032, ia hype geome ic es ),
genes om bo h lis s we e in ol ed in nume ous inflamma o y
signaling pa hways.
Table 2 Top 20 me a-analysis esul s om models o any and sus ained MSDP.
Anno a ions Any MSDP Sus ained MSDP % Di .
CpG ID Loca ion Gene ( egion) β
1
pβ
1
p
cg25112191 ch 1:151804260 RORC (1s exon; 5′UTR) 0.021 3.56E−15 0.033 7.22E−21 55.1
cg17823829 ch 1:202765754 KDM5B (body) 0.062 6.29E−19 0.089 2.74E−23 44.8
cg26045080 ch 1:36807363 STK40 (body) 0.026 5.26E−17 0.034 8.48E−18 30.4
cg00534380 ch 2:101766586 TBC1D8 (body) −0.043 5.85E−19 −0.058 1.66E−20 35.7
cg19246018 ch 2:240031588 HDAC4 (body) 0.016 6.93E−18 0.017 1.86E−16 5.6
cg06202585 ch 2:38325802 2p22.2 −0.009 1.08E−16 −0.009 1.78E−06 0.1
cg23752985 ch 2:85803571 VAMP8 (TSS1500) −0.018 1.30E−14 −0.023 2.51E−11 28.7
cg00666842 ch 2:88366145 SMYD1 (TSS1500) 0.031 5.84E−17 0.049 3.35E−27 57.2
cg27402634 ch 3:156536860 3q25.31 −0.233 8.85E−99 −0.251 2.20E−130 7.5
cg09491670 ch 4:53529646 4q12 0.015 3.71E−15 0.022 1.06E−22 45.1
cg25585967 ch 5:14452105 TRIO (body) 0.040 3.50E−15 0.062 5.97E−19 54.8
cg14214914 ch 9:131870304 CRAT (body) 0.037 4.84E−18 0.052 1.21E−21 39.1
cg20340720 ch 10:104512523 WBP1L (body) −0.052 4.44E−30 −0.074 1.87E−41 41.3
cg26648103 ch 11:66791718 SYT12 (5′UTR) −0.033 3.29E−16 −0.042 2.06E−15 28.4
cg26115089 ch 11:93846406 HEPHL1 (3′UTR) 0.026 9.16E−25 0.033 1.53E−22 23.9
cg26843110 ch 15:74935742 EDC3 (body) −0.054 1.05E−22 −0.079 6.56E−26 46.3
cg26433445 ch 16:81764289 16q23.3 0.025 4.72E−17 0.035 8.84E−20 39.9
cg24177452 ch 17:27494295 MYO18A (5′UTR) 0.024 9.42E−15 0.032 2.55E−15 30.1
cg22018329 ch 19:3116716 GNA11 (body) 0.041 1.41E−14 0.060 6.48E−17 45.6
cg03313447 ch 19:41829042 CCDC97 (3′UTR) −0.019 3.28E−16 −0.027 2.70E−23 43.0
All models we e adjus ed o ma e nal age, pa i y, educa ion, pu a i e cellula he e ogenei y, and esidual bias; CpGs ha we e no anno a ed wi h a gene name in he Illumina 450 K anno a ion file ha e
been anno a ed wi h hei genomic egion (i.e., 2p22.2).
β
1
coe ficien o he associa ion be ween DNAm and MSDP, which ep esen he es ima ed di e ence in he p opo ion o DNAm a each CpG when compa ing mo he s ha smoked du ing p egnancy
(MSDP), o hose ha did no smoke du ing p egnancy; p alues om in e se- a iance fixed-e ec me a-analyses, % Di . he pe cen di e ence in e ec size when compa ing he β
1
o sus ained MSDP
o he β
1
o any MSDP (% Di .), MSDP ma e nal smoking du ing p egnancy.
Table 3 Resul s om eQTM models, DNAm e sus GA a bi h, and DNAm e sus BW.
CpG ID eQTM gene Ann. gene eQTM β
1
eQTM pGA β
1
GA pBW β
1
BW p
cg25112191 RORC RORC −7.28 5.17E−05 0.49 4.59E−01 −2.70 4.08E−05
cg17823829 CYB5R1 KDM5B 0.39 2.99E−02 −1.18 9.11E−06 −1.02 9.72E−05
cg26045080 SH3D21 STK40 −2.43 2.42E−07 −1.41 4.19E−03 −0.93 5.70E−02
cg00534380 TBC1D8 TBC1D8 −2.46 1.22E−08 3.53 9.26E−20 1.48 1.62E−04
cg19246018 NA HDAC4 1.53 4.05E−01 −1.39 1.74E−02 −1.03 7.67E−02
cg06202585 CYP1B1 2p22.2 −12.78 3.09E−03 2.23 5.94E−02 1.53 1.97E−01
cg23752985 CAPG VAMP8 4.23 3.24E−03 5.47 1.40E−15 −0.31 6.62E−01
cg00666842 NA SMYD1 −1.27 6.78E−02 0.45 2.94E−01 −0.44 3.07E−01
cg27402634 LEKR1 3q25.31 −2.34 3.41E−04 −0.28 1.41E−01 0.92 6.71E−07
cg09491670 USP46 4q12 −3.72 2.55E−06 −0.88 2.22E−01 −0.65 3.64E−01
cg25585967 NA TRIO 0.95 1.54E−01 −0.97 8.64E-03 −1.35 2.06E−04
cg14214914 CRAT CRAT −3.20 1.46E−10 −0.96 1.35E−02 −0.23 5.49E−01
cg20340720 WBP1L C10o 26 −0.69 2.81E−03 0.49 1.88E−01 1.86 2.42E−07
cg26648103 SYT12 SYT12 −1.80 7.31E−03 1.06 1.29E−02 1.66 7.56E−05
cg26115089 NA HEPHL1 −0.74 5.20E−01 −1.10 4.43E−02 −1.48 6.71E−03
cg26843110 UBL7-AS1 EDC3 −1.08 1.18E−02 2.30 5.09E−12 1.28 1.19E−04
cg26433445 NA 16q23.3 −0.51 4.40E−01 −0.38 5.45E−01 −2.99 9.06E−07
cg24177452 MYO18A MYO18A −0.66 2.36E−02 −0.76 1.24E−01 −0.74 1.29E−01
cg22018329 AES GNA11 −0.71 3.78E−03 −1.67 6.86E−07 −0.49 1.47E−01
cg03313447 TGFB1 CCDC97 −9.79 1.08E−14 −2.48 7.79E−04 2.29 1.58E−03
We p esen he eg ession coe ficien s and p alues om he models o eQTMs, ges a ional age a bi h, and bi h weigh , among he 20 CpGs p esen ed in Table 2. CpGs we e mapped o eQTM genes
whose exp ession le els we e associa ed wi h CpG DNAm le els (eQTM p alue < 0.05) and o he anno a ed genes om he Illumina 450 K anno a ion file; CpGs ha we e no anno a ed wi h a gene
name in he Illumina 450 K anno a ion file ha e been anno a ed wi h hei genomic egion (i.e., 2p22.2). Associa ions passing Bon e oni-co ec ion a e shown in bold.
β
1
eg ession coe ficien s om linea models, in which mRNA exp ession (eQTM), ges a ional age (GA; in e se no mal ans o med), and bi h weigh (BW; z-sco es) we e eg essed on DNAm a each
CpG, while adjus ing o con ounde s; p alues (p) o eQTM esul s a e om linea eg ession models, while p alues (p) o GA and BW esul s a e om in e se- a iance fixed-e ec me a-analysis,
eQTM exp ession quan i a i e ai me hyla ion.
NATURE COMMUNICATIONS | h ps://doi.o g/10.1038/s41467-021-24558-y ARTICLE
NATURE COMMUNICATIONS | (2021) 12:5095 | h ps://doi.o g/10.1038/s41467-021-24558-y | www.na u e.com/na u ecommunica ions 5

To u he unde s and he egula o y landscape o hese
di e en ially me hyla ed CpG si es, we used En ichR o es o
en ichmen o ansc ip ion ac o (TF) a ge s om ENCODE/
ChEA da abases25. The genes anno a ed o ou MSDP-associa ed
CpGs we e a ge s o GATA1 and GATA2, he and ogen
ecep o , TP63, SMAD4, RUNX1, and ZBTB7A (Supplemen a y
Da a 13). While he eQTM gene lis was no significan ly en iched
o TF binding, he op TF was RUNX1 (Supplemen a y Da a 14).
We hen examined whe he he MSDP-associa ed CpG si es we e
en iched o allele-specific ge mline di e en ially me hyla ed
egions (gDMR)26, egula o y ea u es om he placen a-specific
15-ch oma in s a e anno a ion om ROADMAP27, o placen a-
specific pa ially me hyla ed domains (PMD)28, which con ain
placen a-specific ep essed genes (anno a ed o he esul s files in
Supplemen a y Da a 8). Mos no ably, he MSDP-associa ed
CpGs we e subs an ially deple ed o PMDs (Supplemen a y
Fig. 5), and highly en iched in placen al enhance s (Supplemen-
a y Fig. 6); bo h o hese a e indica o s ha ou se o placen al
MSDP-associa ed CpGs occu wi hin highly ac i e egions o he
placen al me hylome. While we also explo ed whe he ou
findings we e en iched o allele-specific gDMRs, only one o
he 443 MSDP-associa ed CpGs was wi hin a candida e ma e nal
gDMR (cg05211790 anno a ed o RAI14), sugges ing ha gDMRs
a e no subs an ially a ec ed by MSDP.
Pheno ype en ichmen analyses. We es ed o en ichmen o
pheno ypes wi hin he da abase o Geno ypes and Pheno ypes
(dbGAP) using En ichR, o unde s and he ypes o heal h ou -
comes ha ha e been associa ed wi h hese genes. We again es ed
he gene lis s based on hose anno a ed o ou CpGs (n=284
genes) and hose mapped o eQTM genes (n=211 genes). The
Illumina anno a ed genes we e en iched o cell adhesion mole-
cules (CAM), as hma, body mass index (BMI), blood p essu e,
and an ipsycho ic agen s; while he gene lis based on eQTM
mapping did no yield significan ly en iched pheno ypes (Sup-
plemen a y Da a 15 and 16). The CAM designa ion in dbGAP,
which was he mos significan ly en iched pheno ype (Fishe ’s
exac p alue =6.51E−05), desc ibes a b oad a ay o molecula
unc ions ela ed o cellula mobili y and in eg a ion, wound
healing, and me as asis, which a e ela ed o many o he biolo-
gical unc ions iden ified in he abo e pa hway en ichmen es s.
While he o he pheno ypes a e ela ed o some o he heal h
ou comes ha ha e been associa ed wi h p ena al obacco smoke
exposu e: as hma, ca diome abolic e ec s (BMI and blood p es-
su e), and psychia ic e ec s (an ipsycho ic agen s).
P oximi y o gene ic a ian s linked o bi h ou comes.We
aimed o unde s and whe he he egions o he placen al genome
whe e DNAm is esponsi e o MSDP a e also impo an in
e al g ow h egula ion. Thus, we explo ed whe he gene ic a -
ian s ha ha e p e iously been associa ed wi h bi h ou comes ia
genome-wide associa ion s udies (GWAS) a e wi hin close
p oximi y o ou iden ified CpGs. We examined whe he MSDP-
associa ed CpGs we e wi hin ±0.5 Mb (1 Mb window) o single-
nucleo ide polymo phisms (SNPs) ha ha e been associa ed wi h
BW, bi h leng h (BL), head ci cum e ence (HC), and GA29–34
(Supplemen a y Da a 8). O he 330 bi h ou come SNPs in
au osomal ch omosomes, 61 SNPs we e wi hin 0.5 Mb o 51
CpGs (Supplemen a y Da a 17), including cg27402634 (LEKR1),
cg26843110 (EDC3), and cg20340720 (WBP1L), sugges ing ha
hese genomic egions ha a e esponsi e o MSDP appea o be
in ol ed in g ow h egula ion. We also explo ed whe he ou
MSDP-associa ed CpGs may be biased by me hyla ion quan i a-
i e ai loci (mQTLs), in which SNPs influence he me hyla ion
le els a nea by CpGs. Two s udies ha e examined his ques ion
in human placen a, iden i ying 866 ( e . 35) and 4342 ( e . 36)
placen al mQTLs. Ou findings did no appea o be biased by
gene ic a ia ion as only 5 o he 443 MSDP-associa ed CpGs
we e p e iously cha ac e ized placen al mQTLs.
DNAm associa ed wi h smoking- ela ed bi h ou comes.We
hen pe o med seconda y me a-analyses o examine he ela-
ionships be ween DNAm wi h GA a bi h, p e e m bi h, BW,
BL, and HC z-sco es. O he 443 CpGs es ed, 142 (32.1%) we e
ela ed o a leas one bi h ou come a e Bon e oni adjus men
(0.05/443). The majo i y o bi h ou come associa ions we e
ela ed o GA a bi h (121 CpGs) (Supplemen a y Da a 18).
P e e m deli e y, p oduced simila associa ions, al hough ewe
CpGs we e s a is ically significan (Supplemen a y Da a 19). We
also ound ha nume ous CpGs we e associa ed wi h bi h size
z-sco es, wi h he majo i y o hese being associa ed wi h BW (25
CpGs; Supplemen a y Da a 20), ollowed by BL (11 CpGs; Sup-
plemen a y Da a 21) and HC (2 CpGs; Supplemen a y Da a 22).
Some o he CpGs associa ed wi h GA we e also associa ed wi h
bi h size measu emen s, e en al hough BW, BL, and HC we e
s anda dized o GA, sugges ing independen associa ions wi h
bo h ges a ional du a ion and e al g ow h (Supplemen a y
Fig. 7). Fou CpGs (anno a ed o KDM5B, TTC7B, SFRS1, and
DUSP6) sha ed associa ions wi h bo h GA and BW, wo (anno-
a ed o TMEM51 and MYO7A) wi h bo h GA and BL, and one
wi h all h ee o GA, BW, and BL (anno a ed o KIAA1211).
Among he CpGs ha we e associa ed wi h a leas one o hese
bi h ou comes, hose ha ended o ha e posi i e associa ions
wi h BW, BL, HC, o GA we e ypically hypome hyla ed wi h
exposu e o MSDP, while CpGs ha exhibi ed in e se associa-
ions wi h bi h ou comes ended o be hype me hyla ed wi h
exposu e o MSDP.
Among ou op 20 CpGs ha we e associa ed wi h any MSDP,
5 we e associa ed wi h GA a bi h and 6 we e associa ed wi h
BW z-sco es (Table 3). DNAm a cg27402634 (LEKR1) and
cg20340720 (WBP1L), bo h loca ed close o BW-SNPs and o
which MSDP associa ed wi h lowe DNAm, we e associa ed
wi h la ge BW (p alue =6.71E−07 and p alue =2.42E−07,
espec i ely). On he o he hand, DNAm a cg26843110 (EDC3;
hypome hyla ed in esponse o MDSP and also close o BW-
SNPs) and a cg17823829 (KDM5B; hype me hyla ed) we e
associa ed wi h longe and sho e GAs a bi h, espec i ely
(p alue =5.09E−12 and p alue =9.11E−06, ia in e se-
a iance fixed-e ec me a-analysis). Fo es plo s o BW z-sco es
and GA o hese ou CpGs a e shown in Fig. 3.
Compa ison wi h CpGs associa ed wi h MSDP in co d blood.
We hen assessed whe he he DNAm signa u es o MSDP in he
placen a we e consis en wi h MSDP associa ions in co d blood
p e iously epo ed by he PACE conso ium10. Only ou CpGs
(anno a ed o CYP1A1,GNG12,RNF122, and ZBTB4) yielded
significan associa ions in bo h issues (Table 4). O no e, he
CpGs wi hin CYP1A1 and RNF122 showed opposi e di ec ions o
associa ion wi h MSDP in co d blood and placen a. The e was no
o e all genome-wide co ela ion ( 2< 0.1) o he eg ession
coe ficien s ac oss hese wo issues (Supplemen a y Fig. 8). We
also explo ed whe he he e was mo e consis ency i we used a
elaxed significance h eshold. O he 6073 CpGs ha we e wi hin
a 5% FDR om he co d blood analysis, 115 also yielded asso-
cia ions wi h FDR < 5% in placen a, 70 (61%) o which had
consis en di ec ions o e ec be ween he wo issues (Supple-
men a y Da a 23). Among hese, one CpG wi hin AHRR,
cg21161138, exhibi ed consis en hypome hyla ion wi h MSDP
ac oss bo h s udies; his was no able since CpGs wi hin he AHRR
ARTICLE NATURE COMMUNICATIONS | h ps://doi.o g/10.1038/s41467-021-24558-y
6NATURE COMMUNICATIONS | (2021) 12:5095 | h ps://doi.o g/10.1038/s41467-021-24558-y | www.na u e.com/na u ecommunica ions
gene ha e been mos consis en ly iden ified in s udies o MSDP
and co d blood DNAm.
Discussion
We iden ified 443 CpG si es wi h placen al me hyla ion le els ha
we e associa ed wi h any o sus ained MSDP. Di e en ial DNAm
was g ea e o he majo i y o hese CpGs when we s a ified o
sus ained MSDP, and a la ge p opo ion o he MSDP-associa ed
CpGs we e ela ed o bi h ou comes. Those CpGs ha we e
obse ed o ha e highe DNAm associa ed wi h MSDP, ended o
be in e sely associa ed wi h GA and bi h size, while CpGs
exhibi ing lowe DNAm wi h MSDP ended o be posi i ely
associa ed wi h GA and bi h size.
Fig. 3 Fo es plo s o he coho specific and in e se- a iance fixed-e ec me a-analysis es ima es o associa ion be ween highe le els o placen al
DNAm wi h ges a ional age a bi h and bi h weigh . Es ima ed di e ences in ges a ional age a bi h and bi h weigh z-sco es (slope) and 95%
confidence in e als (95% CI) associa ed wi h inc easing le els o DNAm a A,Bcg27402634, C,Dcg26843110, E,Fcg20340720, and G,Hcg17823829.
All models we e adjus ed o ma e nal age, pa i y, ma e nal educa ion, and pu a i e cellula he e ogenei y. The slopes and 95% confidence in e als (95%
CIs), ep esen he eg ession coe ficien s om linea models, in which ges a ional age (in e se no mal ans o med) and bi h weigh (z-sco es) we e
eg essed on DNAm a each CpG, while adjus ing o con ounde s.
NATURE COMMUNICATIONS | h ps://doi.o g/10.1038/s41467-021-24558-y ARTICLE
NATURE COMMUNICATIONS | (2021) 12:5095 | h ps://doi.o g/10.1038/s41467-021-24558-y | www.na u e.com/na u ecommunica ions 7
The mos s a is ically significan associa ion (cg27402634), also
iden ified in p io EWAS o MSDP in placen al issues15, yielded
d ama ically lowe DNAm le els in associa ion wi h MSDP
exposu e. This e ec size is much la ge in magni ude (~25%
di e ence o sus ained MSDP) compa ed o wha has gene ally
been obse ed in mos exposu e- ocused EWAS, al hough wi hin
he same ange as a CpG si e in AHRR (cg05575921; 18% di -
e ence be ween exposed and unexposed) om a p io EWAS o
cu en smoking and blood DNAm37. In addi ion, dec eased
placen al DNAm a cg27402634 co ela es wi h inc eased
exp ession o LEKR1, and associa es wi h smalle BW and BL.
Thus MSDP-associa ed hypome hyla ion a his CpG would be
consis en wi h he well-known e ec o ma e nal smoking,
esul ing in sho e ges a ion and smalle bi h size.
The unc ional ac i i ies o cg27402634, o co esponding
LEKR1 gene, in human placen al issues a e no known. Howe e ,
GWAS findings p o ide e idence ha gene ic a ian s wi hin his
egion (3q25.31) migh be in ol ed in e al g ow h and possibly
me abolic p og amming. Fo ins ance, he SNP s1482852 o i s
p oxies ( s900400; s13322435) ha e been associa ed wi h e al
g ow h38, adiposi y in newbo ns39,40, ma e nal adiponec in le els,
co d blood lep in40, and insulin elease a e an o al glucose
challenge41. These findings om gene ic s udies in combina ion
wi h ou cu en s udy, implica e ha his locus on ch omosome
3 (3q25.31) con ains ac i e de e minan s o e al g ow h egula-
ion and me abolic ac i i y, and ha placen al DNAm a
cg27402634 is highly esponsi e o ma e nal smoking. Fu u e
mechanis ic wo k is necessa y o in es iga e whe he he placen al
epigene ic egula ion a his locus specifically influences placen al
unc ions and/o o e all g ow h and me abolic unc ions o he
de eloping e us.
We iden ified nume ous o he no able MSDP-associa ed CpGs,
and highligh hose CpGs yielding he s onges magni udes o
e ec (cg20340720, cg26843110, and cg17823829). MSDP was
associa ed wi h lowe DNAm a cg20340720, loca ed wi hin
WBP1L (also anno a ed as C10o 26), while lowe DNAm a his
CpG co ela ed wi h lowe wi h BW and BL z-sco es. Gene ic
a ian s nea by o his CpG ha e been ela ed o BW34 and blood
p essu e42. We also obse ed lowe DNAm wi h MSDP a
cg26843110, which is wi hin he body o he EDC3 gene, and is
nea by o SNPs associa ed wi h BW ( s3784789, e . 34). Lowe
DNAm a cg26843110 associa ed wi h sho e GA a bi h, and
dec eased exp ession o CSK, which is in ol ed in ophoblas
di e en ia ion43, as well as blood p essu e and aldos e one
egula ion44. Finally, cg17823829 (anno a ed o KDM5B) was
hype me hyla ed wi h MSDP. Highe DNAm a his CpG
co ela ed wi h sho e GA a bi h and wi h lowe exp ession o
PPFIA4 gene, which can be induced in esponse o hypoxia45.
Ou en ichmen analyses iden ified nume ous in e ela ed
pa hways ha a e c i ical o placen al g ow h and de elopmen .
The Ah pa hway was significan ly en iched among he eQTM-
mapped genes, and is well ecognized o i s oles in esponding
o en i onmen al exposu es and influencing immune ac i i y,
pa icula ly among Th17 cells46. Mul iple p o-inflamma o y
immune cell pa hways, such as Th17 cell di e en ia ion, as well
as neu ophil and pla ele deg anula ion we e also en iched
among hese MSDP-associa ed eQTM genes. Th17 cells can
induce inflamma ion and oxida i e s ess in he placen a47, while
neu ophils a e in ol ed in he inflamma o y cascades ha a e
hough o con ibu e o p e e m bi h48. Pla ele s can ecognize
damaged issues and coo dina e T-cell-media ed inflamma o y
esponses, including Th17 cells49. Recep o y osine kinase
pa hways, which we e en iched among bo h he eQTM and
Illumina anno a ed gene se s, desc ibe a b oad a ay o cell su -
ace ecep o s ha can bind o cy okines, ho mones, and g ow h
ac o s, and include he epide mal g ow h ac o ecep o (EGFR).
The EGFR is highly exp essed in placen al issues, while EGF is
in ol ed in p o ec ing ophoblas s om hypoxia-induced
apop osis50, and pe u bed EGF-EGFR signaling has been asso-
cia ed wi h placen al pa hologies and g ow h es ic ion51.In
addi ion, when ac i a ed, he EGFR ini ia es mul iple signaling
pa hways in ol ing MAPK/JAK kinases o STAT TFs52, while he
MAPKAPK3,JAK1, and STAT5A genes we e among hose wi h
MSDP-associa ed CpGs and we e in ol ed in nume ous pa h-
ways iden ified in ou en ichmen analyses. Thus, hese unc ional
en ichmen analyses cha ac e ized nume ous in e ela ed pa h-
ways ha con ibu e o sensing and esponding o en i onmen al
s esso s, egula ing placen al inflamma o y ac i i y, and influ-
encing g ow h ac o signaling.
We aimed o be e unde s and he o e all egula o y land-
scape o hese CpGs by es ing o en ichmen o egula o y
ea u es (TSSs, enhance s, gene bodies, and un ansla ed egions),
PMDs, allele-specific gDMRs, and whe he hey we e en iched o
TF binding and pheno ypes ia En ichR. These CpGs we e
en iched in placen al enhance s while deple ed in PMDs28, sug-
ges ing ha hey a e loca ed wi hin ac i e egula o y egions. The
genes ha we e anno a ed o MSDP-associa ed CpGs we e en i-
ched o genes egula ed by specific TFs, mos no ably GATA1,
GATA2, and RUNX1. Toge he wi h PPARG and TP63, GATA
ac o s a e pa o he co e ansc ip ional egula o y ci cui ha
guides and main ains p ope ophoblas di e en ia ion53,54,
while angiogenic ac i i y is educed in placen as lacking
Table 4 Compa ing MSDP-associa ed CpGs om placen a o hose om co d blood.
CpG Gene Co d β
1
Co d pPlac.
any
β
1
Plac.
any
pPlac.
sus .
β
1
Plac.
sus
pDi .
cg06397161 SYNGR1 −0.008 3.44E−06 0.034 3.25E−14 0.046 6.66E−13 −/+
cg07565956 ZBTB4 −0.006 2.50E−10 −0.014 6.20E−12 −0.021 1.10E−12 −/−
cg08327744 RNF122 −0.010 1.37E−08 0.017 2.57E−08 0.022 2.26E−08 −/+
cg14351425 GRK5 0.007 1.09E−06 0.027 2.27E−08 0.037 1.21E−08 +/+
cg14541011 RALGDS −0.007 2.99E−07 0.019 2.17E−09 0.022 2.63E−07 −/+
cg16704246 RBM20 −0.008 6.21E−07 −0.017 4.03E−11 −0.023 6.17E−10 −/−
cg23160522 CYP1A1 0.006 2.33E−06 −0.029 2.60E−07 −0.042 2.48E−08 +/−
cg23680900 CYP1A1 0.003 8.73E−08 −0.005 1.52E−03 −0.016 6.88E−14 +/−
cg25189904 GNG12 −0.024 1.38E−32 −0.014 3.80E−08 −0.019 1.64E−09 −/−
We p esen he me a-analysis esul s o CpGs among ou 443 MSDP-associa ed CpGs ha we e also associa ed wi h MSDP in co d blood ( om a published PACE me a-analysis10) wi h p alues < 1E
−05; hose ha passed a Bon e oni-co ec ed h eshold in he co d blood s udy a e highligh ed in bold.
β
1
coe ficien o he associa ion be ween DNAm and MSDP, which ep esen he es ima ed di e ence in he p opo ion o DNAm a each CpG when compa ing mo he s ha smoked du ing p egnancy
(MSDP), o hose ha did no smoke du ing p egnancy; p alues a e om in e se- a iance fixed-e ec me a-analysis, MSDP ma e nal smoking du ing p egnancy, Di . di ec ion o e ec fi s o co d
blood, hen o placen a, Plac. esul s om placen al issue.
ARTICLE NATURE COMMUNICATIONS | h ps://doi.o g/10.1038/s41467-021-24558-y
8NATURE COMMUNICATIONS | (2021) 12:5095 | h ps://doi.o g/10.1038/s41467-021-24558-y | www.na u e.com/na u ecommunica ions
GATA255. RUNX1 on he o he hand, is a d i e o haema o-
poie ic de elopmen 56, and while no ho oughly s udied in
human placen a, he e is some e idence ha placen al inflam-
ma ion is associa ed wi h he up egula ion o placen al RUNX1
and hema opoiesis57. Genes ha we e anno a ed o MSDP-
associa ed CpGs ha e also been linked o human heal h and
disease ai s ia dbGAP, including a numbe o condi ions
ela ed o ca diome abolic heal h (BMI and blood p essu e),
which has p e iously been linked o MSDP58,59. This may indi-
ca e ha MSDP e ec s placen al genes ha a e in ol ed in ene gy
up ake and expendi u e, lipid and glucose me abolism, blood
p essu e egula ion, and inflamma ion, which a e some o he key
physiological p ocesses ha a e dis up ed in he pa hogenesis o
me abolic synd ome60. Genes we e also en iched o as hma and
psychia ic heal h, which a e known o be ela ed o MSDP61.
Thus, he epigene ic esponse o MSDP p edominan ly occu s a
CpGs ha a e in ac i e egions o he placen al genome, a genes
ha a e a ge ed by TFs in ol ed in angiogenesis and hema o-
poiesis, and wi hin genes ha ha e been linked o ca diome a-
bolic, espi a o y, and psychia ic heal h.
We also aimed o unde s and whe he he egions o he epi-
genome ha associa e wi h MSDP a e in ol ed in e al g ow h
egula ion. We ound ha many o hese CpGs, including some o
ou s onges hi s om he me a-analysis, we e in simila geno-
mic p oximi y (wi hin 0.5 Mb) o SNPs ha ha e p e iously been
associa ed wi h bi h size o GA a bi h29–34. In addi ion, ou
seconda y me a-analyses demons a ed ha DNAm le els a
almos hal o ou CpGs we e also associa ed wi h GA and/o BW
and BL. Thus, he CpG and gene lis s ha we ha e p oduced in
his s udy a e obus ly associa ed wi h MSDP and wi h bi h
ou comes ac oss mul iple independen popula ions, making hese
compelling candida es o u u e s udies aimed a pe o ming
causal media ion analyses. We did no pu sue media ion in his
s udy as i has been es ablished ha due o measu emen e o in
sel - epo ed smoking and he ac ha smoking- ela ed me hy-
la ion is an excellen bioma ke o exposu e, media ion can be
o e es ima ed62. Thus, media ion s udies a e mos sui able o
coho s ha collec objec i e smoking exposu e bioma ke s, such
as co inine o 4-(me hylni osamino)-1-(3-py idyl)-1-bu anol
(NNAL). Some s udies, howe e , ha e demons a ed ha pla-
cen al DNAm is an in e es ing candida e as a possible media o
be ween MSDP and lowe BW15,63, and we encou age addi ional
in es iga ion o his possibili y (while also conside ing o he bi h
ou comes), using he CpGs ha we iden ified as candida es and
using Mendelian Randomiza ion app oaches. I is also possible
ha MSDP, gene ic a ia ion, and placen al DNAm in hese
egions could yield addi i e o in e ac i e e ec s on bi h ou -
comes. P io s udies ha e add essed his ques ion in blood64,65,
bu u u e esea ch is needed o explo e he po en ial in e ac i e
e ec s be ween gene ic a ian s and MSDP in ela ion o DNAm
in he placen a o be ween gene ic a ain s and MSDP-associa ed
DNAm in ela ion o bi h ou comes. In addi ion, while ou
MSDP-associa ed CpGs a e in simila genomic egions (1 Mb
windows) wi h bi h ou come SNPs, e y ew o ou CpGs (4 o
443) a e known placen al mQTLs35,36, and hus we do no hink
ha ou findings we e subs an ially biased by nea by gene ic
a ia ion.
We compa ed ou findings o hose o a p e ious PACE me a-
analysis o MSDP and co d blood DNAm10. The mos s a is ically
significan associa ion wi h MSDP in placen a (cg27402634,
LEKR1) was no associa ed wi h MSDP in he co d blood me a-
analysis. Only wo CpG si es, anno a ed o GNG12 and ZBTB4,
we e di e en ially me hyla ed a genome-wide h eshold o s a-
is ical significance in bo h placen a and co d blood, wi h he
same di ec ion o associa ion in bo h issues. While wo CpG si es
wi hin CYP1A1 and RNF122 we e genome-wide significan in
bo h me a-analyses, bu wi h di e en di ec ions o associa ion in
co d blood e sus placen a. In e es ingly, we obse ed CYP1A1 o
be hypome hyla ed in placen a wi h exposu e o MSDP, which is
consis en wi h s udies o adipose, skin, and lung issues66, bu
his CpG was hype me hyla ed in co d blood10. E en when using
elaxed significance h esholds, such as FDR, only 70 CpGs
exhibi ed consis en e ec s be ween hese wo issues. One CpG
wi hin AHRR (cg21161138) did exhibi consis en hypome hyla-
ion wi h MSDP ac oss bo h issues a his elaxed FDR sig-
nificance h eshold. These obse a ions sugges ha he e a e
unique placen a-specific CpG me hyla ion esponses o his
exposu e. Howe e , as men ioned abo e, he Ah pa hway, was
significan ly en iched among ou eQTM genes, and he co d
blood s udy did iden i y some genes in ol ed in p o-
inflamma o y esponse and g ow h ac o signaling. Thus, while
he e was li le o e lap in specific MSDP-associa ed CpGs and
genes be ween hese wo issues, he e was some o e lap in he
o e all biological p ocesses.
The abo e findings should be in e p e ed wi hin he con ex o
his s udy’s limi a ions. MSDP was sel - epo ed and subjec o
misclassifica ion, al hough di e en ial misclassifica ion mos
likely would ha e biased ou findings owa d he null. We ound
ha sus ained MSDP p oduced la ge magni udes o associa ion,
as has been p e iously ound in s udies o blood DNAm67, bu we
did no explici ly compa e hose mo he s ha smoked h oughou
p egnancy o hose ha qui du ing ea ly p egnancy, no did we
assess dose– esponse pa e ns (i.e., numbe o ciga e es o bio-
ma ke concen a ions), bo h o which should be he ocus o
u u e in es iga ions. Ou s udy p edominan ly consis ed o
samples om mo he –in an pai s o Eu opean ances y, and hus
addi ional s udies in ol ing di e se ances ies and e hnic back-
g ounds a e needed, in o de o imp o e he gene alizabili y o
hese findings. The obse ed associa ions be ween DNAm le els
and ep oduc i e ou comes could be due o e e se causa ion,
which is one o he easons we did no pu sue o mal media ion
analyses. In addi ion, placen a is a he e ogeneous issue wi h
mul iple di e en cell ypes68 ha se e di e en unc ions and
hus ha e di e en epigene ic s a es69. To co ec o his, we
es ima ed and adjus ed o a iabili y using a da a-d i en
app oach, Re F eeCellMix22, since no e e ences o placen al
cell- ype me hylomes was a ailable a he ime. In addi ion, i is
possible ha coho -specific sampling p o ocols o o he coho -
specific di e ences in da a gene a ion could ha e esul ed in
placen al samples om some coho s ha ing g ea e he e o-
genei y han o he s. Residual con ounding om cellula he e o-
genei y o o he unmeasu ed con ounde s may ha e con ibu ed
o he obse ed infla ion in ou o iginal me a-analyses, hus we
implemen ed BACON23 and only p esen hose findings ha
we e s a is ically significan a e applying Bon e oni co ec ion
o mul iple es ing a e infla ion and bias co ec ion
wi h BACON.
Despi e hese limi a ions, ou s udy had nume ous s eng hs,
including a la ge sample size ac oss se en independen s udies,
ha monized defini ions o exposu e a iables and co a ia es, and
s anda dized p o ocols o quali y con ol, p e-p ocessing and
analyses o DNAm da a. We pe o med seconda y analyses
in ol ing mRNA exp ession, unc ional and pheno ype en ich-
men , o e lap wi h GWAS hi s o ep oduc i e ou comes, and
me a-analyses o DNAm a ia ion wi h bi h ou comes o p o ide
biological and heal h- ela ed in e p e a ions o ou findings.
O e all, we iden ified a DNAm signa u e o MSDP in he placen a
ha shows subs an ial di e ences om ha obse ed in co d
blood, mos no ably cg27402634 which is in e genic be ween
LINC00886 and LEKR1, whe e placen as ha we e exposed had
~25% lowe DNAm han hose ha we e no exposed. The
MSDP-associa ed CpGs a e wi hin ac i e egions o he placen al
NATURE COMMUNICATIONS | h ps://doi.o g/10.1038/s41467-021-24558-y ARTICLE
NATURE COMMUNICATIONS | (2021) 12:5095 | h ps://doi.o g/10.1038/s41467-021-24558-y | www.na u e.com/na u ecommunica ions 9