Na u e Men al Heal h | Volume 1 | Oc obe 2023 | 751–760 751
na u e men al heal h
h ps://doi.o g/10.1038/s44220-023-00124-xA icle
The e ec o en i onmen on dep essi e
symp oms in la e adolescence and ea ly
adul hood: an exposome-wide associa ion
s udy and win modeling
Zhiyang Wang 1, S ephanie Zelle s1, Alyce M. Whipp1,2, Ma ja Heinonen-Guzeje 2,
Ma ia Fo as e 3,4,5,6, Jo di Júl ez4,7, I ene an Kamp8 & Jaakko Kap io 1,2
The exposome ep esen s he o ali y o en i onmen al e ec s, bu
sys ema ic e alua ion be ween i and dep essi e symp oms is scan . He e
we sough o comp ehensi ely iden i y he associa ion o he exposome
wi h dep essi e symp oms in la e adolescence and ea ly adul hood and
de e mine gene ic and en i onmen al co a iances be ween hem. Based
on he FinnTwin12 coho (3,025 pa icipan s in young adul hood and 4,127
a age 17), he exposome-wide associa ion s udy (ExWAS) design was used
o iden i y signi ican exposu es om 12 domains. Bi a ia e Cholesky win
models we e i ed o an exposome sco e and dep essi e symp oms. In
ExWASes, 29 and 46 exposu es we e signi ican ly associa ed wi h dep essi e
symp oms in young adul hood and a age 17, espec i ely, and amilial
exposu es we e he mos in luen ial. Twin models indica ed conside able
gene ic and en i onmen al co a iances be ween he exposome sco e and
dep essi e symp oms wi h sex di e ences. The indings unde sco e he
sys ema ic app oach o he exposome and he conside a ion o ele an
gene ic e ec s.
Dep essi e symp oms a e a ype o ch onic men al heal h condi ion
wi h complex e iology, and majo dep essi e diso de (MDD) is he
clinical diso de diagnosed when dep essi e symp oms each a h esh-
old o se e i y and du a ion. Dep essi e symp oms and MDD lead o a
se ious public heal h bu den. The upda ed Global Bu den o Diseases
s udy showed ha he age-s anda dized p e alence o MDD was 4%
(3,951 pe 100,000 people) in Wes e n Eu ope, highe han he global
le el, and unde lined he hea y bu den on people aged be ween 15 and
24 ( e . 1). Among adolescen s, a 2021 sys ema ic e iew indica ed ha
he pooled p e alence o sel - epo ed dep essi e symp oms was 34%
and o MDD was 5% om he s udies be ween 2001 o 2020, and he
p e alence is inc easing
2
. The COVID-19 pandemic exace ba ed he
al eady g owing end o ha dship. Gi en a g owing body o e idence
on he en i onmen al e ec on dep essi e symp oms and MDD3,4, mo e
sys ema ic in es iga ion is u gen ly needed, especially among you h.
The concep o he ‘exposome’, which depic s he dynamic o ali y
o he en i onmen ha an indi idual expe iences, was aised in 2005
5
.
The exposome is di ided in o h ee pa s—speci ic ex e nal, gene al
Recei ed: 6 Ap il 2023
Accep ed: 9 Augus 2023
Published online: 25 Sep embe 2023
Check o upda es
1Ins i u e o Molecula Medicine Finland, Helsinki Ins i u e o Li e Science, Uni e si y o Helsinki, Helsinki, Finland. 2Clinicum, Depa men o Public
Heal h, Uni e si y o Helsinki, Helsinki, Finland. 3PHAGEX Resea ch G oup, Blanque na School o Heal h Science, Uni e si a Ramon Llull (URL),
Ba celona, Spain. 4ISGlobal, Ba celona, Spain. 5Uni e si a Pompeu Fab a (UPF), Ba celona, Spain. 6CIBER Epidemiología y Salud Pública (CIBEREsp),
Mad id, Spain. 7Clinical and Epidemiological Neu oscience (Neu oÈpia), Ins i u d’In es igació Sani à ia Pe e Vi gili (IISPV), Reus, Spain. 8Cen e o
Sus ainabili y, En i onmen and Heal h, Na ional Ins i u e o Public Heal h and he En i onmen (RIVM, Ne he lands), Bil ho en, he Ne he lands.
e-mail: jaakko[email p o ec ed]
Na u e Men al Heal h | Volume 1 | Oc obe 2023 | 751–760 752
A icle h ps://doi.o g/10.1038/s44220-023-00124-x
amily membe s is an e icien way o demons a e he p esence (o lack
o ) gene ic e ec s. Thus, he combina ion o exposome and win s ud-
ies could ad ance ou knowledge o he complexi ies be ween genes
and en i onmen s, imp o e ou unde s anding o exis ing de icien-
cies in exposome measu es, and p oduce u he esea ch ques ions.
A na u al ex ension is hen o include measu ed geno ypes, ei he
a ge ing speci ic genes such as hose in ol ed in he me abolism o
ex e nal compounds o mo e b oad-based genome-wide app oaches
o de i e polygenic sco es o gene ic suscep ibili y.
In his s udy, based on he FinnTwin12 coho , we aim o (1) comp e-
hensi ely and sys ema ically de e mine exposu es ha a e signi ican ly
associa ed wi h dep essi e symp oms and MDD in la e adolescence
and ea ly adul hood h ough h ee ExWASes and (2) es ima e o wha
ex en he exposome and dep essi e symp oms sha e he same gene ic
and en i onmen al isk ac o s.
Resul s
Cha ac e is ics o he s udy, pa icipan s, and exposu es
Figu e 1 shows he lowcha o he analysis pipeline, which consis ed
o h ee ExWASes and he ollowing bi a ia e win modeling. Pe
he FinnTwin12 coho , he e we e 3,025, 1,236, and 4,127 indi idual
wins included in h ee sepa a e ExWASes wi h he ou comes o gen-
e al beha io in en o y (GBI) sco e in young adul hood (p ima y),
he incidence o MDD in young adul hood, and GBI sco e a age 17,
espec i ely. The cha ac e is ics o each ExWAS a e shown in Table 1.
Fo indi idual wins included in ExWASes o all ou comes
(Table 2), he majo i y we e emale and om dizygo ic (DZ) pai s, and
hei pa en al educa ion le els we e limi ed (less han high school). A
age 17, 25.4% o indi idual wins epo ed being cu en smoke s, and
82.6% we e ull- ime s uden s and no wo king. In young adul hood,
25.3% o indi idual wins epo ed ha hey we e cu en ly smoking,
and 51.4% had a ull- ime job. The mean GBI sco es a age 17 and in young
adul hood we e 5.0 (s.d.: 4.9) and 4.4 (s.d.: 4.7), espec i ely, and he
ex e nal, and in e nal exposomes—and he ex e nal exposome could be
u he subdi ided in o he amilial, social, buil exposome, and so on.
Ins ead o s udying a single o small g oup o exposu es, an exposome
s udy aims o in es iga e he o e all e ec o he en i onmen while,
una oidably, complexi ies such as in e ac ion o ubiqui y inc ease he
di icul y
6
. An exposome- (en i onmen al, exposu e) wide associa-
ion s udy (ExWAS), like o he ‘WAS’ s udies, deno es an agnos ic and
sys ema ic me hod o hypo hesis gene a ing, which is compa a i ely
app op ia e o he exposome’s spa io empo al a iabili ies and mul i-
le el s uc u e
7
. Se e al ExWAS s udies ha e a ge ed men al heal h
8–10
,
and Choi e al11. used clinical inciden dep ession as he ou come and
iden i ied mul iple modi iable ac o s. As i is he ea ly wa ning sign
o MDD, ocusing on dep essi e symp oms in adolescence o young
adul hood could be easie o guide ansla ional in e en ion as ea ly
as possible, which would be mo e cos e ec i e.
Despi e he bene i s o he exposome app oach, he e a e some
o he hind ances. Fi s , unde he cu en echnique, we canno meas-
u e e e y possible exposu e ( a om eaching ‘1-genome’), and he
exposome keeps upda ing, expanding, and en iching. Mo eo e , some
s udies ha e emphasized exposu es’ non-gene ic p ope ies, which
igno es how he en i onmen in e ac s wi h gene ics h ough mul iple
mechanisms among many ai s, including dep ession
12,13
. Medda and
colleagues, on he basis o he I alian Twin Regis y, demons a ed he
subs an ial gene ic ole in exogenous me allomics, whe e he es ima-
ions o s anda dized gene ic a iance, as a p opo ion o o al a iance
o he measu ed exposu es, anged om 0.15 (a senic) o 0.79 (zinc)14.
As a na u al expe imen , win and amily s udies p o ide a me hod o
e alua e gene ic and en i onmen al ela ionships be ween ai s and
exposu es. This design decomposes he a iance o ai s in o addi i e
gene ic (A), dominan gene ic (D), common en i onmen al (C), and
unique en i onmen al (E) componen s, which con ain he dis inc
ea u es o he exposome as he o e all en i onmen al e ec . Such
indi ec e idence o gene ic e ec s based on gene ic ela ionships o
ExWAS o GBI in young adul hood
(indi idual win n = 3,025)
Bi a ia e win
modeling in male
Only include he ull win pai
and d op he opposi e-sex DZ pai
385 en i onmen al
exposu es we e included
ExWAS o GBI a age 17
(indi idual win n = 4,127)
Only included he ull win pai
and d op he opposi e-sex DZ pai
Gene a e exposome sco e by CFA
based on signi ican exposu es
Gene a e exposome sco e by CFA
based on signi ican exposu es
Bi a ia e win
modeling in emale
Impu a ion Impu a ion
Pos hoc MMRM
(indi idual win n = 3,025)
ExWAS o incidence o MDD in young
adul hood (indi idual win n = 1,236)
Impu a ion
Excluded 99 exposu es
due o empo ali y
Bi a ia e win
modeling in male
Bi a ia e win
modeling in emale
Fig. 1 | Flowcha o he analysis pipeline. Flowcha o he analysis pipeline demons a ing he pa h om he choice o exposu es o ExWAS analysis and ending wi h
bi a ia e win modeling. The ull pa h was used o dep essi e symp oms (GBI) a wo ages. Only ExWAS was comple ed o MDD.
Na u e Men al Heal h | Volume 1 | Oc obe 2023 | 751–760 753
A icle h ps://doi.o g/10.1038/s44220-023-00124-x
wo measu es co ela ed wi h 0.49. The incidence o li e ime MDD in
young adul hood was 12.3%.
Exposu es’ code names, desc ip ion, and s a is ics based on wins
included in he ExWAS o GBI in young adul hood (be o e impu a ion)
a e p esen ed in Supplemen a y Table 1. The e a e 12 domains o expo
-
su es, colo ed in he ollowing plo s: ai pollu ion, building, blue and
g een spaces, popula ion densi y, geocoo dina es, p ena al exposu es,
passi e smoking, amily and pa en s, iend and oman ic ela ionships,
school and eache s, s ess ul li e e en s, and social indica o s. In p in-
cipal componen analysis (PCA), he i s p incipal componen (PC1)
a ibu ed only 10.93% and 10.66% o he o al a iabili y o all included
exposu es in young adul hood and a age 17, espec i ely (Ex ended
Da a Fig. 1). F om he sca e plo s o PC1 and PC2, we iden i ied some
po en ial clus e s o exposu es om domains o building, blue and
g een spaces, and social indica o s ia isual assessmen .
ExWAS o dep essi e symp oms and MDD in young adul hood
The adjus ed coe icien and –log10(P alue) o all exposu es included
o bo h adul ou comes a e p esen ed in Supplemen a y Table 2. The e
we e 40 signi ican P alues in 29 exposu es, which we e associa ed
wi h log- ans o med GBI sco e in young adul hood, iden i ied om
385 exposu es (Fig. 2a). The e we e 24, 2, and 3 exposu es belonging o
he domains o amily and pa en s, iend and oman ic ela ionships,
and school and eache s, espec i ely. Fo he mos p o ec i e expo-
su e, compa ed wi h wins who el hei home en i onmen was com-
ple ely un ai , qui e un ai , o somewha un ai a age 17 (un ai _A17),
wins who el i was no a all un ai a age 17 we e associa ed wi h a
0.40 lowe log- ans o med GBI sco e (95% con idence in e al (CI):
−0.50, −0.31) (Fig. 2b). Fo he mos ha m ul exposu e, compa ed wi h
wins who we e comple ely sa is ied wi h hei ela ionships wi h
iends a age 14 (sa _ iend_A14), wins who el somewha sa is ied,
mainly no sa is ied, o no a all sa is ied a age 14 we e associa ed wi h
a 0.42 highe log- ans o med GBI sco e (95% CI: 0.29, 0.55) (Fig. 2b).
By con as , none o he exposu es showed a signi ican associa ion
wi h MDD (Ex ended Da a Fig. 2).
ExWAS o dep essi e symp oms a age 17
The adjus ed coe icien and –log10(P alue) o he age 17 ou come
a e p esen ed in Supplemen a y Table 2. The e we e 71 signi ican
P alues in 46 exposu es, which we e signi ican ly associa ed wi h log-
ans o med GBI sco e, iden i ied om 286 exposu es (Ex ended Da a
Fig. 3a). The e we e 32, 6, 4, and 4 exposu es belonging o he domains
o amily and pa en , iend and oman ic ela ionship, school and
eache s, and s ess ul li e e en s, espec i ely. Fo he mos ha m ul
exposu es, compa ed wi h wins who we e comple ely sa is ied wi h
hei success a wo k o s udies a age 17 (sa _s udywo k_A17), wins
who el mainly no sa is ied o no a all sa is ied a age 17 we e associ-
a ed wi h a 0.65 highe log- ans o med GBI sco e (95% CI: 0.55, 0.74)
(Ex ended Da a Fig. 3b). Fo he mos p o ec i e exposu e, he same
as he esul in young adul hood, compa ed wi h wins who el hei
home en i onmen was comple ely un ai , qui e un ai , o somewha
un ai a age 17 (un ai _A17), wins who el i was no a all un ai a
age 17 we e associa ed wi h a 0.50 lowe log- ans o med GBI sco e
(95% CI: −0.57, −0.43) (Ex ended Da a Fig. 3b). The e a e 27 exposu es
ha a e signi ican ly associa ed wi h log- ans o med GBI sco es
bo h in young adul hood and a age 17, and 22 exposu es belong o
he domain o amily and pa en s.
Twin modeling o dep essi e symp oms wi h exposome sco es
Be o e he bi a ia e modeling, he bes - i uni a ia e AE model (had he
lowes Akaike in o ma ion c i e ion compa ed wi h ADE and E models)
indica ed E explained 61% o he a iance o dep essi e symp oms in
males and 45% in emales a age 17, and he numbe s sligh ly educed
o 59% and 42%, espec i ely, in young adul hood (Supplemen a y
Table 3). The exposome sco e was c ea ed by con i ma o y ac o analy-
sis (CFA) based on he signi ican exposu es om ExWASes. The s and-
a dized oo mean squa e esidual o models in young adul hood and a
age 17 we e 0.100 and 0.078, espec i ely, indica ing accep able model
i . MDD was no included in he CFA o ollowing win modeling due o
he smalle sample size and no signi ican exposu e being iden i ied.
Then we used he exposome sco e o conduc bi a ia e win modeling
be ween he exposome sco e and dep essi e symp oms. Gi en he sex
di e ences in he p e alence o dep essi e symp oms, he di e ences
in he i abili y, and he ac ha sex-limi ed bi a ia e models also indi-
ca ed signi ican sex di e ences (Supplemen a y Table 4) a bo h age
poin s, we an he bi a ia e models sepa a ely o males and emales.
Figu e 3 and Supplemen a y Table 5 show he pa h coe icien s
o he model o exposome sco e and log- ans o med GBI sco e in
young adul hood (mean age: 23.9). Unique en i onmen al ac o s
accoun ed o 23% and 13% o he co a iances in males and emales,
espec i ely, while addi i e gene ic ac o s accoun ed o 77% in males
and 87% in emales. In males, s anda dized a iances o E
exposome
and
EGBI we e 0.32 (95% CI: 0.26, 0.39) and 0.51 (95% CI: 0.42, 0.62); he
numbe s educed o 0.25 (95% CI: 0.21, 0.30) and 0.50 (95% CI: 0.42,
0.58) in emales. The emaining sha e o a iance was accoun ed o
by addi i e gene ic e ec s.
Ex ended Da a Fig. 4 and Supplemen a y Table 5 show he pa h
coe icien s o he model o exposome sco e and log- ans o med
GBI sco e a age 17. Unique en i onmen al ac o s accoun ed o 31%
and 13% o he co a iances in males and emales, espec i ely. Addi-
i e gene ic ac o s accoun ed o 69% in males and 87% in emales.
The s anda dized a iances o Eexposome a age 17 a e simila o Eexposome
in young adul hood ega dless o sex. The s anda dized a iance o
Eexposome is 0.26 (95% CI: 0.22, 0.30) and 0.22 (95% CI: 0.19, 0.25) and o
EGBI is 0.64 (95% CI: 0.55, 0.73) and 0.44 (95% CI: 0.38, 0.50) in males and
emales, espec i ely. The emaining sha e o a iance was accoun ed
o by addi i e gene ic e ec s.
Pos hoc mixed model epea ed measu es. On he basis o he longi-
udinal design and 27 signi ican exposu es selec ed by bo h ExWASes
o log- ans o med GBI sco e, a e adjus ing o co a ia es and base-
line e ec , all he exposu es we e s ill signi ican ly associa ed wi h log-
ans o med GBI sco e in young adul hood. The esul s a e p esen ed
in Supplemen a y Table 6.
Discussion
Using da a on dep essi e symp oms and diagnosed MDD om he
FinnTwin12 s udy and a wide ange o exposu es om mul iple sou ces,
we applied a wo-s age analysis o i s sc een he exposome and hen
es ima e he en i onmen al sou ces o co ela ion be ween he expo-
some and dep essi e symp oms ia win modeling. Fi s , mul iple expo-
su es by sel - epo ha e been iden i ied ac oss domains o amily and
pa en s, iend and oman ic ela ionships, school and eache s, and
s ess ul li e e en s, which we e signi ican ly associa ed wi h dep essi e
symp oms in young adul hood and a age 17. By con as , none o he
exposu es co ela ed wi h he incidence o MDD in young adul hood.
Second, a e gene a ing an exposome sco e based on signi ican ly
associa ed exposu es, he bes - i ing bi a ia e AE models indica ed
Table 1 | Cha ac e is ics o ExWASes
Ou come Numbe o
indi idual
wins
Numbe o
exposu es Numbe o
P alues Signi ican
h eshold
(–log10(P alue))
GBI in young
adul hood 3,025 385 501 3.51
Incidence o
MDD in young
adul hood
1,236 385 501 3.47
GBI a age 17 4,127 286 394 3.44
Na u e Men al Heal h | Volume 1 | Oc obe 2023 | 751–760 754
A icle h ps://doi.o g/10.1038/s44220-023-00124-x
ha unique en i onmen al e ec s accoun ed o a ma ked ac ion o
he co a iance be ween he exposome sco e and dep essi e symp oms.
This en i onmen al ac ion was highe in males han in emales, sug-
ges ing a no able sex di e ence. Ou esul implies ha en i onmen al
e ec s a e mo e impac ul compa ed wi h gene ic e ec s in males
han in emales.
In luence om he amilial componen o he social exposome,
especially om he amilial a mosphe e, was demons a ed by ou e i-
dence as ha ing he mos subs an ial impac on dep essi e symp oms
in la e adolescence and ea ly adul hood and hei ajec o y. A la ge
Chinese su ey also ound ha amilial ac o s such as cohesion,
con lic , and con ol co ela ed wi h he occu ence o dep essi e
symp oms among uni e si y s uden s
15
. O he s udies ha e e ealed
he connec ion o amily iangula ion (pa en –child coali ion and
alliance) and sa is ac ion wi h dep essi e symp oms om childhood
o la e adolescence ac oss coun ies16,17. Fai ness (la ges p o ec i e
e ec size o GBI a bo h age poin s), as a dimension o pa en i ica ion,
was demons a ed as a unique p edic o o men al heal h symp oms
18
.
These exis ing con en ional in es iga ions we e consis en wi h ou s,
while ou ExWAS mo e sys ema ically e alua ed a wide ange o expo-
su es and educed he chance o ype I e o wi hou any p e-iden i ied
hypo hesis. Mo eo e , ins ead o adi ional scales o assessing amil-
ial and in e pe sonal ela ionships, we ea ed each scale componen as
an ‘independen ’ exposu e in models, which helped us o iden i y new
co ela ions, de ec he ela i e impo ance, and p epa e o u he
analysis o mo e in ica e ela ionships be ween di e en componen s
and dep essi e symp oms.
Resul s om bi a ia e win modeling e eal a complex ela ionship
among genes, en i onmen s, and dep essi e symp oms. Al hough he
unique en i onmen al ac o explains a no able amoun o co a iance
be ween exposome sco e and dep essi e symp oms, he addi i e
gene ic ac o explained ela i ely mo e. Many signi ican exposu es
we e chosen unde he guidance o he exposome pa adigm, bu i
does no necessa ily imply a pu e en i onmen al e ec . Many amilial
in luences a e conside ed ‘inhe i able ac o s’ be ween gene a ions o a
ce ain ex en , acco ding o he in e gene a ional ansmission heo y.
Such e ec s can be ansmi ed om pa en s o child en h ough
sha ed genes bu also by sha ed en i onmen s. Ea ly s udies ha e ound
ha li e sa is ac ion o amily iolence om pa en s and o igin amilies
led o an impo an impac on he de elopmen o subsequen simila
amilial en i onmen s among o sp ing19,20. Mo eo e , we should con-
side he exis ence o he gene–en i onmen in e ac ion (G×E), which
sugges s he di e en e ec s o a geno ype on disease isk in pe sons
wi h di e en en i onmen al exposu es21. Choi e al11. s a i ied he
ExWAS by polygenic isk sco es o majo dep ession and ound ha
some signi ican ac o s in he ull sample became null in he gene i-
cally a - isk sample. Ano he s udy sugges ed he mul iple modula ion
pa hways by exposu e o DNA me hyla ion, h ough nume ous es ing,
ega ded as he G×E-WAS22. In addi ion, p e ious win s udies ound
geog aphic con ounding in he assessmen o A, C, and E a iances,
possibly a ibu able o di e ences in gene ic ances y. Resul s om
he Ne he lands Twin Regis e ound 1.8% o he a iance in child en’s
heigh was cap u ed by egional clus e ing
23
. In he Ne he lands, he e
we e s ong gene ic di e en ia ions be ween he no h and sou h,
be ween he eas and wes , and be ween he middle band and he es o
he coun y by PCA on genome-wide da a
24
. In he Finnish popula ion,
also a subs an ial popula ion s uc u e di e ence is obse ed be ween
he eas and wes pa s o coun y25. In b ie , he hidden he i able and
gene ic ac o s c i ically in luence he associa ion be ween he expo-
some and dep essi e pheno ype h ough a ious mechanisms, which
po en ially lead o a p opensi y o weak associa ions in ou indings.
Fu he mo e, exposu es om he mo e ex e nal domains, pa icu-
la ly in he physical exposome, also showed, a mos , weak connec ions
wi h dep essi e symp oms. While i may be he case ha he ela i e
impo ance o he physical exposome is much less han ha o he
social and amilial exposome wi h espec o dep essi e symp oms,
he e a e possibly o he explana ions. Fi s , a mo e complex s uc u e
o he exposome, such as he in e ac ion o co ela ion be ween indi-
idual exposu es and ex e nal exposome, may exis . Some p e ious
exposome analyses ha e indica ed his
26,27
, bu he ExWAS design can-
no cha ac e ize i . Fo example, he social exposome is an explaining
pa o he physical exposome, which could no be comple ely sepa-
a ed. We aim o in es iga e he complica ed e ec o he dep essi e
pheno ype in he plu alis ic pla o m like machine lea ning on he
Table 2 | Cha ac e is ics o included wins acco ding o he
ExWAS
Cha ac e is ics N (%) / mean (s.d.)
Pa icipan s included in he ExWAS o
GBI (indi idual
win n = 3,025) Incidence o
MDD (indi idual
win n = 1,236)
GBI (indi idual
win n = 4,127)
In young adul hood A age 17
GBI sco e 4.4 (4.7) —5.0 (4.9)
MDD incidence
Yes —152 (12.3) —
No —1,084 (87.7) —
Sex
Male 1,318 (43.6) 564 (45.6) 1,988 (48.2)
Female 1,707 (56.4) 672 (54.4) 2,139 (51.8)
Zygosi y
MZ 1,050 (34.7) 513 (41.5) 1,362 (33.0)
DZ 1,833 (60.6) 721 (58.3) 2,577 (62.4)
Unknown 142 (4.7) 2 (0.2) 188 (4.6)
Pa en al educa ion
Limi ed 1,743 (57.6) 672 (54.4) 2,392 (58.0)
In e media e 666 (22.0) 305 (24.7) 950 (23.0)
High 616 (20.4) 259 (21.0) 785 (19.0)
Smoking
Ne e 1,617 (53.5) 614 (49.7) 1,218 (29.5)
Fo me 339 (11.2) 115 (9.3) 1,418 (34.4)
Occasional 304 (10.1) 132 (10.7) 445 (10.8)
Cu en 765 (25.3) 375 (30.3) 1,046 (25.4)
Wo k (young adul hood)
Full- ime wo k 1,556 (51.4) 497 (40.2) —
Pa - ime wo k 388 (12.8) 236 (19.1) —
I egula wo k 368 (12.2) 338 (27.4) —
No wo king 713 (23.6) 165 (13.4) —
Seconda y-le el school (young adul hood)
Voca ional 1,025 (33.9) 377 (30.5) —
Senio high school 1,826 (60.4) 778 (62.9) —
None 174 (5.8) 81 (6.6) —
Age (young
adul hood) 24.2 (1.7) 22.4 (0.7) —
S udy and wo k s a us (age 17)
Nei he s udy
no wo k — — 150 (3.6)
Only s udy — — 3,406 (82.5)
Bo h s udy and wo k — — 571 (13.8)
Na u e Men al Heal h | Volume 1 | Oc obe 2023 | 751–760 755
A icle h ps://doi.o g/10.1038/s44220-023-00124-x
basis o ou indings in he u u e. Second, Finland has been anked
e y high in he bene icial en i onmen al e ec on he child by UNICEF
(Uni ed Na ions Child en’s Fund), p o iding en i onmen s wi h low ai
pollu ion, high g eenness, sa e wa e , and o he cons uc i e aspec s
ela i ely equally o mos esiden s in childhood and adolescence
28
.
I could explain null esul s wi h ex e nal li ing en i onmen s due o
a lack o indi idual a ia ion in exposu es. Ano he ma e con ibu -
ing o la ge amilial e ec s is he o e lap be ween in e pe sonal ela-
ionships and dep essi e symp oms. In a Swedish win s udy among
emales, in e pe sonal ela ionships con ibu ed be ween 18% and 31%
o he a iance o dep essi e symp oms
29
. Some pe sonali y diso de s
a e igh ly connec ed wi h in e pe sonal ela ionships, o example,
bo de line, a oidan , and pa anoid pe sonali y diso de s’ liabili y
ac o s o e lapped subs an ially wi h MDD’s, in pa icula , clus e s
among No wegian young adul s30. This o e lapping may ha e led o
an o e es ima ion o he impo ance o in e pe sonal ela ionships.
Fo social indica o s, besides he c i ical pe iod, a ious isk mod-
els such as accumula ion o ajec o y may exis , which may also explain
he null esul s. Mo issey and Kinde man con i med he hypo hesis
ha accumula ion o ad e se inancial ha dship nega i ely a ec s
men al heal h, bu no he hypo hesis o c i ical pe iods
31
, while ou
isk model is he ‘c i ical pe iod’. Ano he s udy demons a ed he
complica ed e ec among changes in acial composi ion, neighbo -
hood socioeconomic s a us, and dep essi e symp oms
32
. The social
indica o s de i ed om S a is ics Finland’s (s a . i/ ilas o ie o) egis-
e s a e a he pos al code o municipali y le el, which leads o some
conce n abou he inaccu a e measu emen o an indi idual’s exposu e
(in o ma ion bias).
Se e al p e ious ExWAS s udies linking he exposome o men al
heal h had some simila o he e ogeneous esul s o ou s. an de Weije
e al10. iden i ied se e al social indica o s such as sa e y and income
being linked o men al well-being, bu he links we e weak in ou analy-
sis. This may be due o using di e en ou comes, he olde age in hei
samples, and di e en s a is ical me hods be ween he wo coun ies’
au ho i ies
10
. Al hough he ExWAS o Choi e al. was on he gene al pop-
ula ion in he Uni ed Kingdom, hey also ound ha a highe equency
o isi s wi h amily/ iends educed he odds o dep ession incidence,
and Mendelian andomiza ion ein o ced he causali y o his associa-
ion11. Howe e , we do no ha e many common a iables wi h Choi e
al
11
. in which hey included many li es yle ac o s (speci ic ex e nal
exposome), while we ha e mo e gene al ex e nal exposome a iables.
Ano he ExWAS on psycho ic expe iences iden i ied many s ess ul
li e-e en ac o s, a esul ha was simila o ou s udy
8
. Despi e he
di e gen indings, he accumula ion o ExWAS indings om di e en
coun ies, popula ions, and age g oups enhances ou unde s anding
o g owing concep s o he exposome on dep ession, as well as b oad
men al heal h. The inclusion o a la ge numbe o exposu es abou
in e pe sonal and pe son–socie al ela ionships is also an impo an
addi ion o he exis ing e idence. No ably, some o he in o ma ion
was p o ided by he pa en s, no only he wins. Fu he mo e, some
scien is s ha e aised he concep o an ‘eco-exposome’ o ho oughly
assess he in e nal exposome, including molecules a ec ed by exog-
enous exposu es33, which could be assimila ed in o u he esea ch.
The sex di e ence is no able. Ou p e ious s udy ound ha male
wins end o s ay oge he longe , implying mo e exposu e o any
amilial impac
34
. In a Swedish s udy, amily s uc u e, con lic , and
16
Domain
a b
Ai pollu ion
Signi ican exposu e Adjus be a
(95% CI)
Sa _s udywo k_A17$4
Sa _ iend_A14$3
Open_A17$4
T us _A17$3
Sa _dad_A17$3
Suppo _A17$3
Sa _mon_A17$3
Wa m_A17$3
T us _A14$3
Talk_A14$3
Suppo _A14$3
Sa _dad_A14$3
Wa m_A14$3
Money_A14$3
Open_A14$3
Sa _school_A14$3
F eq_ ip_A12$3
Sa _dad_A17$2
Sa _ eache _A14$3
Popula _1_A14
Open_A17$3
F eq_spo _A14$3
Sa _ iend_A14$2
Suppo _A17$2
Sa _dad_A14$2
T us _A17$2
Sa _mon_A17$2
Daily_A14$2
Con lic _A17$2
Un ai _A14$3
Con lic _A14$2
Un ai _A17$2
Indi e en _A14$3
Con lic _A14$3
Con lic _A17$3
Indi e en _A17$3
Un ai _A17$3
–0.4 00.4
Sa _mon_A14$3
Wa m_A12$3
0.42 (0.29, 0.55) Sa is ac ion wi h
ela ionship wi h
iends a age 14
Un ai home
a mosphe e a age 17
0.38 (0.25, 0.51)
0.34 (0.22, 0.47)
0.33 (0.24, 0.41)
0.32 (0.24, 0.40)
0.31 (0.23, 0.40)
0.31 (0.21, 0.40)
0.30 (0.16, 0.44)
0.29 (0.17, 0.40)
0.28 (0.14, 0.41)
0.27 (0.18, 0.36)
0.27 (0.18, 0.37)
0.26 (0.17, 0.35)
0.26 (0.17, 0.35)
0.26 (0.15, 0.36)
0.22 (0.12, 0.31)
0.22 (0.13, 0.30)
0.20 (0.11, 0.29)
0.19 (0.10, 0.28)
0.18 (0.11, 0.25)
0.18 (0.09, 0.27)
0.17 (0.10, 0.24)
0.17 (0.08, 0.26)
0.17 (0.08, 0.25)
0.16 (0.10, 0.22)
0.15 (0.08, 0.22)
0.15 (0.08, 0.21)
0.14 (0.07, 0.21)
0.13 (0.07, 0.20)
0.12 (0.06, 0.19)
0.12 (0.06, 0.18)
–0.17 (–0.24, –0.10)
–0.18 (–0.26, –0.09)
–0.21 (–0.28, –0.13)
–0.23 (–0.33, –0.14)
–0.24 (–0.37, –0.11)
–0.28 (–0.37, –0.20)
–0.33 (–0.41, –0.24)
–0.37 (–0.50, –0.24)
–0.40 (–0.50, –0.31)
Building
Blue and g een spaces
Popula ion densi y
Geocoo dina es
P ena al exposu es
Passi e smoking
Family and pa en s
F iend and oman ic ela ionships
School and eache s
S ess ul li e e en s
Social indica o s
Signi ican line: –log10(P alue) = 3.51
14
12
10
–log10(P alue)
8
6
4
2
0
Sa _mon_A14$2
Fig. 2 | Associa ion esul s be ween exposu e and log- ans o med GBI sco e
in young adul hood, adjus ed o co a ia es (indi idual win n = 3,025),
using gene alized linea eg essiona. a, Manha an associa ion plo o
exposu es in ela ion o log- ans o med GBI sco e in young adul hood. The y
axis is showing s a is ical signi icance as –log10(P alue) o he adjus men o
mul iple es ing. b, Fo es plo o he adjus ed be a o signi ican exposu es in
descending o de om op o bo om ( om ha m ul o p o ec i e). The cen e
do and ba p esen he e ec size (coe icien o linea eg ession) and 95%
CI, and he sizes o he do s p esen he e ec size ela i ely. The colo legend
applies o bo h a (Manha an associa ion plo ) and b ( o es plo ). The adjus ed
co a ia es we e sex, zygosi y, pa en al educa ion, smoking in young adul hood,
wo k s a us in young adul hood, seconda y-le el school in young adul hood, and
age when wins p o ided he GBI assessmen in young adul hood.
Na u e Men al Heal h | Volume 1 | Oc obe 2023 | 751–760 756
A icle h ps://doi.o g/10.1038/s44220-023-00124-x
child disclosu e o in o ma ion o pa en s we e associa ed wi h o end
-
ing beha io in boys, while only one ac o was salien in gi ls35. Ano he
B i ish s udy ound ha boys in de imen al amilial en i onmen s
we e inc easingly disad an aged in school achie emen compa ed
wi h gi ls
36
. The e idence hin s ha males a e mo e easily a ec ed
by he amily en i onmen , which could explain he highe con ibu-
ion o E on he co a iance be ween he exposome and dep essi e
symp oms in males. This in e ence is no ce ain, and he e is con-
a y e idence
37
. Mo eo e , sex di e ences exis in many biological
mechanisms ega ding how he body neu ophysiologically e lec s
he ex e nal en i onmen . Se e al sex-di e en ially exp essed neu-
o ansmi e s o ho mones, such as p oges e one in emales, a e
in ol ed in sys emic dys egula ion, inducing dep ession
38
. Fu he -
mo e, en i onmen al endoc ine-dis up ing chemicals a e able o al e
neu ode elopmen wi h sex-speci ic e ec s a e y ea ly de elopmen-
al s ages
39
. In he u u e, in eg a ing wi h he in e nal exposome such
as me aboli es and o he omics will help us ad ance he s udy o sex-
di e ence mechanisms on he ela ionship be ween he exposome and
dep essi e pheno ype.
As a pa o he Eu opean Human Exposome Ne wo k, ou o e -
a ching goal is o e alua e he impac o he exposome on human heal h
ac oss a ious age g oups and wi h espec o mul iple ou comes. The
p esen analysis ep esen s one indi idual analysis, and by pooling ou
collec i e e o s, impo an implica ions o clinical p ac ice can be
d awn in he u u e. Ou indings sugges ha s udies on he amilial
componen o social exposome should be no iced and in es iga ed in
he imp o emen o cu en he apy. I does no mean ha we should
igno e he physical exposu e g oup, due o ubiqui y, e en hough hei
ele ance is no salien
40
. In addi ion, i is impe a i e o inco po a e he
conside a ion o amilial e ec s and gene ic liabili y a he same ime
o a mo e ho ough unde s anding in u u e s udies.
The e a e some o he limi a ions in ou s udy. Fi s , compa ed
wi h o he ExWASes, ou sample size is ela i ely small. Al hough
Chung e al. indica ed ha a sample size be ween 1,795 and 3,625
pa icipan s is adequa e when using he Bon e oni co ec ion
41
,
we did no s a i y he ExWAS by sex due o he sample size being
educed by hal . Second, we did no u he assess he causali y.
Causal in e ences a e c i ical o u he policymaking and in e en-
ion. Mendelian andomiza ion in la ge samples is a u u e di ec ion.
Thi d, he ExWAS, CFA, and win modeling we e all pe o med on he
basis o he FinnTwin12 coho , which aises conce ns abou model
o e i ing and leakage. Di e en models wi h di e en pu poses,
hypo heses, and me hodologies in wo s ages educe he isk o
o e i ing and leakage. ExWAS was used o iden i y salien exposu e,
while CFA and win modeling we e used o explo e. The obse a ional
uni was each win pai in win modeling, while in ExWAS and CFA, i is
each indi idual win. Replica ion on o he win coho s and in amily
da ase s is wa an ed.
Conclusion
This s udy applied a wo-s age analysis. Fi s , in ExWAS, we iden i ied
ha exposu es om amily and pa en s, iend and oman ic ela ion-
ships, school and eache s, and s ess ul li e e en s we e signi ican ly
associa ed wi h dep essi e symp oms in la e adolescence and young
adul hood. The amily and pa en exposu es we e he mos in luen-
ial. Second, win modeling be ween he exposome and dep essi e
Exposome sco e
(scaled)
Log- ans o med
GBI sco e in young
adul hood
Unique en i onmen al ac o explained 23.06% o he co a iance
Male (188 MZ and 162 DZ pai s)
Exposome sco e
(scaled)
Log- ans o med
GBI sco e in young
adul hood
Unique en i onmen al ac os explained 12.85% o he co a iance
Female (278 MZ and 218 DZ pai s)
Aexposome
= 0.68
Eexposome
= 0.32
Aexposome
= 0.75
a11 = 0.80 a12 = –0.11 a22 = –0.59 a11 = –0.85 a12 = –0.20 a22 = 0.58
e11 = –0.55 e12 = 0.05 e22 = –0.61 e11 = 0.49 e12 = –0.06 e22 = –0.60
AGBI
= 0.49
AGBI
= 0.50
Eexposome
= 0.25
EGBI
= 0.51
EGBI
= 0.50
Fig. 3 | Bi a ia e Cholesky AE model o he exposome sco e and log-
ans o med GBI sco e in young adul hood ( win pai n = 846).
A, s anda dized a iance o addi i e gene ic e ec ; E, s anda dized a iance
o unique en i onmen al e ec . The a and e s and o pa hway coe icien s om
A and E, espec i ely, o bo h he exposome sco e and log- ans o med GBI sco e.
The 95% CIs o s anda dized a iances and pa hway coe icien s a e p esen ed in
Supplemen a y Table 4.
Na u e Men al Heal h | Volume 1 | Oc obe 2023 | 751–760 757
A icle h ps://doi.o g/10.1038/s44220-023-00124-x
symp oms unco e ed a complex ela ionship among genes, en i on-
men s, and dep essi e symp oms wi h sex di e ences. The indings
unde line he impo ance o sys ema ic e alua ion o he en i onmen-
al e ec s on dep essi e symp oms and ecommend he conside a ion
o gene ic e ec s in u u e s udies.
Me hods
S udy pa icipan s
The pa icipan s came om he FinnTwin12 coho , which is a na ion-
wide p ospec i e coho among all Finnish wins bo n be ween 1983
and 1987. Fi s , he o e all epidemiological s udy consis ed o all 5,184
wins who esponded (age 11–12) a wa e 1, and he e a e h ee gene al
ollowing wa es a ages 14, 17, and in young adul hood (mean age: 21.9).
Mo eo e , 1,035 amilies wi h 2,070 wins we e in i ed o ake pa in an
in ensi e s udy wi h psychia ic in e iews, some biological samples,
and addi ional ques ionnai es
42
. A age 14 (wa e 2), 1,854 wins pa -
icipa ed. They we e hen in i ed o pa icipa e again as young adul s
(wa e 4) o he s udy. Psychia ic in e iews in young adul hood we e
comple ed o 1,347 wins in he in ensi e s udy, including assessmen
o MDD using he Semi-S uc u ed Assessmen o Gene ics o Alcohol
based on Diagnos ic and S a is ical Manual o Men al Diso de s IV
c i e ia
43,44
. The wins also comple ed ques ionnai es on heal h, heal h
beha io s, wo k, and mul iple psychological scales. The lowcha o
gene al FinnTwin12 coho is p esen ed in Ex ended Da a Fig. 5. An
upda ed e iew has been published45. .
The e hics commi ee o he Depa men o Public Heal h o he
Uni e si y o Helsinki and he Ins i u ional Re iew Boa d o Indiana
Uni e si y app o ed he FinnTwin12 s udy p o ocol om he s a
o he coho . The e hical app o al o he e hics commi ee o he
Helsinki Uni e si y Cen al Hospi al Dis ic (HUS) is he mos ecen
and co e s he mos ecen da a collec ion (wa e 4) (HUS/2226/2021).
The HUS e iews he s udy annually, and 2023’s s a emen is numbe
4/2023, da ed 1 Feb ua y 2023. All pa icipan s and hei pa en s/legal
gua dians ga e in o med w i en consen o pa icipa e in he s udy.
The au ho s asse ha all p ocedu es con ibu ing o his wo k comply
wi h he e hical s anda ds o he ele an na ional and ins i u ional
commi ees on human expe imen a ion and wi h he Helsinki Decla a-
ion o 1975, as e ised in 2008.
Measu es
The p ima y ou come is he sho - e sion GBI sco es in young adul -
hood. I is a sel - epo ed in en o y o e alua e he occu ence o
dep essi e symp oms, which is composed o en Like -scale ques-
ions46. The o al sco e anges om 0 o 30, and a highe sco e implies
mo e dep essi e symp oms occu ed. The e a e wo seconda y ou -
comes: GBI sco es a age 17 and incidence o MDD in young adul hood.
In o al, we cu a ed 385 en i onmen al exposu es unde he con-
cep o he Equal-Li e p ojec
47
om mul iple sou ces and g ouped
hem in o 12 domains. Ai pollu ion exposu es came om he annual
a e age ai quali y o each obse a ion s a ion om he Finnish Me e
-
o ological Ins i u e. Domains o building, blue and g een spaces,
popula ion densi y, and a pa o geocoo dina es we e om Equal-
Li e en ichmen . Thei desc ip ion can be ound in a p e ious s udy48
and is p esen ed in Supplemen a y No e 1. Exposu es om p ena al
exposu es, passi e smoking, amily and pa en s, iend and oman ic
ela ionships, school and eache s, and s ess ul li e e en s domains
we e om FinnTwin12 ques ionnai es by sel - epo o pa en epo
and a e desc ibed in a published e iew45. Social indica o s we e om
S a is ics Finland and a e desc ibed in Supplemen a y No e 1. Excep o
FinnTwin12 ques ionnai es, exposu es om o he sou ces we e linked
o indi idual wins ia EUREF-FIN geocoo dina es. The ull esiden ial
his o y o he wins om bi h onwa d un il 2020 was ob ained as geo
-
coo dina es and da es o mo ing in and ou o speci ic add esses om
he Digi al and Popula ion Da a Se ices Agency in Finland34. The ypes
o exposu es a e con inuous, bina y, and ca ego ical. Conside ing he
empo ali y, we included epea ed exposu es o he c i ical-pe iod
isk model, and Ex ended Da a Fig. 6 p esen s he imeline o he s udy.
The e a e h ee exposu e inclusion c i e ia: (1) wins ha e a ailable
esiden ial his o y, (2) wins and hei amily comple ed a leas one
ques ionnai e a any wa e, and (3) he pe cen age o missing alues is
less han 20% in ExWAS. The code names o each exposu e we e de el-
oped om he desc ip ion as closely as possible, and hei domains,
esou ces, and da es a e p esen ed in Supplemen a y Table 1. The
missing pa e ns o each exposu e in each ExWASes a e p esen ed in
Supplemen a y Table 7
Fo analysis o ou comes in young adul hood, we a p io i iden i-
ied se en co a ia es: sex (male, emale), zygosi y (monozygo ic (MZ),
DZ, unknown), pa en al educa ion (limi ed, in e media e, high)
49
,
smoking (ne e , o me , occasional, cu en ), wo k s a us ( ull- ime,
pa - ime, i egula , no wo king), seconda y-le el school ( oca ional,
senio high school, none), and age. The la e ou a iables we e
epo ed by wins as young adul s (wa e 4). Fo analysis o ou come
a age 17, sex, zygosi y, pa en al educa ion, smoking ( epo ed a
age 17) emained. S udy and wo king s a us (nei he s udy no wo k,
only s udy, only wo k) we e included when mos pa icipan s we e
in school a age 17 (wa e 3). The inclusion o co a ia es, besides sex,
zygosi y, and age, was based on he p e ious li e a u e, which shows
co ela ions wi h he en i onmen and dep essi e symp oms50–52.
Pa en al educa ion was adjus ed o o ep esen he amily esou ces
and esilience 49.
Da a p e-p ocessing and desc ip i e s a is ics
Pa icipan s missing in o ma ion on ou come o co a ia es we e
excluded om he co esponding age’s analyses. Due o he skewness
o he GBI sco e, we added one o he GBI sco e and log- ans o med i .
App op ia e eg ouping was conduc ed o ca ego ical exposu es, and
hen we used simple impu a ion by he median o eplace he missing
alues o exposu es. As a dimension educ ion echnique, PCA was
u ilized o measu e he p opo ion o o al a iabili y o all included
exposu es a ibu ed o each PC and isually assess he po en ial clus-
e s o exposu es (co ela ed) on he basis o he wo-dimensional
coo dina e wi h he i s and second componen s. I was conduc ed
only o ou comes o GBI a age 17 and in young adul hood, no o he
incidence o MDD.
Exposome-wide associa ion s udy
To conduc he ExWAS, a gene alized linea eg ession model wi h
Gaussian dis ibu ion (essen ially linea eg ession) o he ou comes
o log- ans o med GBI sco e was epea edly pe o med o each expo-
su e. We used Bon e oni co ec ion by he numbe o e ec i e es s
(calcula ed by PCA) o adjus o mul iple es ing and accoun o
co ela ion be ween exposu es
53
. Co a ia es we e adjus ed and he
clus e e ec o sampling based on amilies o win pai s was con olled
o by he obus s anda d e o . Fo he ou come o he incidence o
MDD, he dis ibu ion was swi ched o be binomial. The numbe o
included exposu es o seconda y ou comes was smalle due o he
hi d exposu e inclusion c i e ia, and he sample size a ied; hus, he
P- alue h esholds a ied. Due o ca ego ical exposu es, he numbe o
P alues was highe han he numbe o exposu es. We used he expo-
some package in he R en i onmen ( e sion 4.2.3) 54.
We u he calcula ed powe using he R package WebPowe (R
en i onmen , e sion 4.2.3) o ExWASes o he log- ans o med GBI
sco e a bo h age poin s. These calcula ions we e based on he smalles
absolu e e ec size among signi ican esul s (0.12 in young adul hood
and 0.10 a age 17), sample size (3,025 in young adul hood and 4,127
a age 17), numbe o p edic o a iables in a single model (8 in young
adul hood and 6 a age 17), and signi ican h esholds (3.09 × 10−4
in young adul hood and 3.63 × 10−4 a age 17). The powe s we e 1 o
ExWASes bo h in young adul hood and a age 17, indica ing adequa e
sample sizes in his s udy.
Na u e Men al Heal h | Volume 1 | Oc obe 2023 | 751–760 758
A icle h ps://doi.o g/10.1038/s44220-023-00124-x
Gene a ing exposome sco e
Based on he signi ican exposu es selec ed om he ExWAS, CFA was
used o es ima e an exposome sco e, p epa ing o he ollowing win
modeling. Acco ding o he concep o he en i onmen ’s o ali y, we
indica ed a one- ac o s uc u e o he exposome. The CFA assumes
he co ela ion be ween exposu es due o he exposome sco e and
e i ies i based on s uc u al equa ion modeling as heo y d i en.
We used maximum likelihood o es ima e he sco e and s anda dized
oo mean squa e esidual o e alua e he model i 55. The clus e e ec
was con olled like be o e. Due o mul iple subg oups in ca ego ical
exposu es, we included he whole exposu e a iable when he e was a
leas one subg oup ha was signi ican compa ed wi h he e e ence
in ExWAS. The coe icien s o signi ican exposu es we e p esen ed in
Supplemen a y Tables 8 and 9 o ou comes o GBI in young adul hood
and a age 17, espec i ely. In addi ion, we conduc ed explo a o y ac o
analysis (EFA) es ima ed by maximum likelihood wi h 100 op imiza-
ions, whe eas a la ge numbe o e ained ac o s indica ed po en ial
o e i ing o EFA. The CFA and EFA we e pe o med using S a a 18.0
(S a aCo p), and package sem was used.
Twin modeling
In win modeling, he gene ic e ec is usually di ided in o addi i e
and dominan gene ic e ec s
56
. Since MZ wins a e oughly gene ically
iden ical and DZ wins sha e oughly hal o hei seg ega ing genes, he
co ela ion o A is se o 1.0 and 0.5 and o D is se o 1.0 and 0.25 wi hin
MZ and DZ win pai s, espec i ely. The epis a ic e ec is a pa o A.
The en i onmen al e ec is also di ided in o wo componen s: com-
mon en i onmen , whose co ela ion is assumed o be 1.0 ega dless
o zygosi y, and unique en i onmen (no co ela ion), which includes
unmeasu ed e o s. The use o he win model assumes he absence
o asso a i e ma ing o he ai unde s udy among he pa en s and
equal e ec s o he en i onmen by zygosi y.
The in apai co ela ions o GBI in DZ (ρ = 0.22 in young adul -
hood and 0.16 a age 17) and MZ (ρ = 0.52 in young adul hood and 0.51
a age 17) indica ed o use an ADE model ini ially, ins ead o he ACE
model (ρ
MZ
> 2ρ
DZ
). Due o using only he win pai design, ins ead o he
ex ended amily design, we could no use an ACDE model. The sa u a ed
win model was pe o med o es he assump ions o equal means and
a iances o win o de and o zygosi y, ia cons ain means and
a iances, and o de ec he sex di e ence ia sex limi a ion. In he
sa u a ed model (Supplemen a y Table 10), he Akaike in o ma ion
c i e ion and likelihood a io es be ween models sugges ed ha he
assump ions we e basically me . Resul s o he sex-limi a ion sa u a ed
model (Supplemen a y Table 10) indica ed a no able sex di e ence.
Finally, o assess how he cu en exposome sco e explains he
a iance o dep essi e symp oms, we employed he bi a ia e Cholesky
AE model o i he exposome sco e and log- ans o med GBI sco e
(Ex ended Da a Fig. 7) a bo h age poin s, which e icien ly decomposes
he pheno ypic co ela ion and o e s he a ibu ion (%) o gene ic
and en i onmen al ac o s57. Two la en ac o s (Aexposome and Eexposome)
in luence bo h he exposome sco e (a
11
and e
11
) and log- ans o med
GBI sco e (a
21
and e
21
), and ano he wo la en ac o s (A
GBI
and E
GBI
)
in luence only he log- ans o med GBI sco e (a22 and e22). The o e all
co ela ion be ween he exposome sco e and GBI could be calcula ed
as
a11 ×a12 +e11 ×e12
. Va iances o A
exposome
, E
exposome,
A
GBI
, and E
GBI
we e
calcula ed as
a2
11 +a2
12
,
e2
11 +e2
12
,
a2
22
, and
e2
22
, espec i ely. We also e-
assess he sex di e ence ia an addi ional sex-limi ed sa u a ed bi a i-
a e win model.
Only ull MZ and DZ win pai s we e included in he win modeling.
We d opped he opposi e-sex DZ pai s and s a i ied he uni a ia e
and bi a ia e win models by sex. The cha ac e is ics o included and
excluded indi idual wins in he win modeling a e p esen ed in Supple
-
men a y Table 11, and we did no obse e a la ge di e ence, sugges ing
low selec ion bias isk due o sex, zygosi y, and win pai . Age, epo ed
in he young adul hood su ey, was adjus ed in uni a ia e and bi a ia e
models o he ou come in young adul hood. We used he OpenMx
package in he R en i onmen ( e sion 4.2.3) 58.
Pos hoc mixed models o epea ed measu es. On he basis o he
exposu es signi ican ly associa ed wi h GBI a bo h ime poin s, we
pe o med he mixed models o epea ed measu es (MMRM) as a
pos hoc analysis o u he explo e he e ec s on he ajec o y o
dep essi e symp oms. This me hod analyzes he in luence on he
log- ans o med GBI in young adul hood by bo h exposu es o in e -
es ( ixed e ec ) and ‘baseline’ log- ans o med GBI a age 17 ( andom
e ec )59. The sample size and co a ia es o he MMRM we e he same
as in he ExWAS o log- ans o med GBI sco e in young adul hood. The
clus e e ec was con olled by he obus s anda d e o . The mul iple
es ing was con olled by he alse disco e y a e (Q alue < 0.05 was
conside ed s a is ically signi ican ). These pos hoc analyses we e
pe o med using S a a 18.0 (S a aCo p).
Repo ing summa y
Fu he in o ma ion on esea ch design is a ailable in he Na u e Po -
olio Repo ing Summa y linked o his a icle.
Da a a ailabili y
The FinnTwin12 da a a e no publicly a ailable due o he es ic ions o
in o med consen . Howe e , he FinnTwin12 da a a e a ailable h ough
he Ins i u e o Molecula Medicine Finland (FIMM) Da a Access Com-
mi ee (DAC) ([email p o ec ed]i) o au ho ized esea che s who
ha e IRB/e hics app o al and an ins i u ionally app o ed s udy plan.
To ensu e he p o ec ion o p i acy and compliance wi h na ional da a
p o ec ion legisla ion, a da a use/ ans e ag eemen is needed, he
con en and speci ic clauses o which will depend on he na u e o he
eques ed da a. Reques s will be add essed in a easonable ime ame
(gene ally wo o h ee weeks), and he p ima y mode o da a access is
by ei he pe sonal isi o emo e access o a secu e se e .
Code a ailabili y
No new so wa e, package, o algo i hm was de eloped. All code o
da a cleaning and analysis associa ed wi h he cu en submission is
a ailable upon easonable eques o he co esponding au ho o Z.W.
(zhiyang.[email p o ec ed]).
Re e ences
1. GBD 2019 Men al Diso de s Collabo a o s. Global, egional,
and na ional bu den o 12 men al diso de s in 204 coun ies
and e i o ies, 1990–2019: a sys ema ic analysis o he
Global Bu den o Disease S udy 2019. Lance Psychia y 9,
137–150 (2022).
2. Sho ey, S., Ng, E. D. & Wong, C. H. J. Global p e alence o
dep ession and ele a ed dep essi e symp oms among
adolescen s: a sys ema ic e iew and me a-analysis. B . J. Clin.
Psychol. 61, 287–305 (2022).
3. Köhle , C. A. e al. Mapping isk ac o s o dep ession ac oss
he li espan: an umb ella e iew o e idence om me a-analyses
and Mendelian andomiza ion s udies. J. Psychia . Res. 103,
189–207 (2018).
4. Mai , C., Roux, A. V. D. & Galea, S. A e neighbou hood
cha ac e is ics associa ed wi h dep essi e symp oms?
A e iew o e idence. J. Epidemiol. Communi y Heal h 62, 940
LP–940946 (2008).
5. Wild, C. P. Complemen ing he genome wi h an ‘exposome’: he
ou s anding challenge o en i onmen al exposu e measu emen
in molecula epidemiology. Cance Epidemiol. Bioma ke s P e .
14, 1847–1850 (2005).
6. Guloksuz, S., an Os, J. & Ru en, B. P. F. The exposome pa adigm
and he complexi ies o en i onmen al esea ch in psychia y.
JAMA Psychia y 75, 985–986 (2018).
Na u e Men al Heal h | Volume 1 | Oc obe 2023 | 751–760 759
A icle h ps://doi.o g/10.1038/s44220-023-00124-x
7. Zheng, Y. e al. Design and me hodology challenges o
en i onmen -wide associa ion s udies: a sys ema ic e iew.
En i on. Res. 183, 109275 (2020).
8. Lin, B. D. e al. Nongene ic ac o s associa ed wi h psycho ic
expe iences among UK biobank pa icipan s: exposome-wide
analysis and Mendelian andomiza ion analysis. JAMA Psychia y
79, 857–868 (2022).
9. Ni, M. Y. e al. De e minan s o physical, men al and social well-
being: a longi udinal en i onmen -wide associa ion s udy. In . J.
Epidemiol. 49, 380–389 (2020).
10. an de Weije , M. P. e al. Expanding he en i onmen al scope:
an en i onmen -wide associa ion s udy o men al well-being.
J. Expo. Sci. En i on. Epidemiol. 32, 195–204 (2022).
11. Choi, K. W. e al. An exposu e-wide and Mendelian andomiza ion
app oach o iden i ying modi iable ac o s o he p e en ion o
dep ession. Am. J. Psychia y 177, 944–954 (2020).
12. Le e s, H. C. B., Lange, T., Collins, C., Ul -Mølle , C. J. & Jacobsen,
S. The s udy o in e ac ions be ween genome and exposome in
he de elopmen o sys emic lupus e y hema osus. Au oimmun.
Re . 18, 382–392 (2019).
13. Wiche s, M. e al. Mechanisms o gene–en i onmen in e ac ions
in dep ession: e idence ha genes po en ia e mul iple sou ces o
ad e si y. Psychol. Med. 39, 1077–1086 (2009).
14. Medda, E. e al. The esponse o oxida i e s ess and me allomics
analysis in a win s udy: he ole o he en i onmen . F ee Radic.
Biol. Med. 97, 236–243 (2016).
15. Yu, Y. e al. The ole o amily en i onmen in dep essi e
symp oms among uni e si y s uden s: a la ge sample su ey in
China. PLoS ONE 10, e0143612 (2015).
16. S a opoulos, V., Laza a ou, H., Ma ini, E. & Dikeos, D. Low amily
sa is ac ion and dep ession in adolescence: he ole o sel -
es eem. J. Educ. De . Psychol. 5, 109–118 (2015).
17. Wang, L. & C ane, D. R. The ela ionship be ween ma i al
sa is ac ion, ma i al s abili y, nuclea amily iangula ion, and
childhood dep ession. Am. J. Fam. The . 29, 337–347 (2001).
18. Hoope , L. M. & Wallace, S. A. E alua ing he pa en i ica ion
ques ionnai e: psychome ic p ope ies and psychopa hology
co ela es. Con emp. Fam. The . 32, 52–68 (2010).
19. Cappell, C. & Heine , R. B. The in e gene a ional ansmission o
amily agg ession. J. Fam. Violence 5, 135–152 (1990).
20. Powd ha ee, N. & Vignoles, A. Men al heal h o pa en s
and li e sa is ac ion o child en: a wi hin- amily analysis o
in e gene a ional ansmission o well-being. Soc. Indic. Res. 88,
397–422 (2008).
21. O man, R. Gene–en i onmen in e ac ion: de ini ions and s udy
design. P e . Med. 25, 764–770 (1996).
22. Ca e as-Gallo, N. e al. The ea ly-li e exposome modula es he
e ec o polymo phic in e sions on DNA me hyla ion. Commun.
Biol. 5, 455 (2022).
23. Tamimy, Z. e al. Mul ile el win models: geog aphical egion as a
hi d le el a iable. Beha . Gene . 51, 319–330 (2021).
24. Abdellaoui, A. e al. Popula ion s uc u e, mig a ion, and
di e si ying selec ion in he Ne he lands. Eu . J. Hum. Gene . 21,
1277–1285 (2013).
25. Ke minen, S. e al. Fine-scale gene ic s uc u e in Finland. G3 7,
3459–3468 (2017).
26. Ohanyan, H. e al. Machine lea ning app oaches o
cha ac e ize he obesogenic u ban exposome. En i on. In . 158,
107015 (2022).
27. Ohanyan, H. e al. Associa ions be ween he u ban exposome
and ype 2 diabe es: esul s om penalised eg ession by leas
absolu e sh inkage and selec ion ope a o and andom o es
models. En i on. In . 170, 107592 (2022).
28. Places and Spaces: En i onmen s and Child en’s Well-Being
(UNICEF IRC, 2022).
29. Spo s, E. L. e al. Accoun ing o dep essi e symp oms in women:
a win s udy o associa ions wi h in e pe sonal ela ionships.
J. A ec . Diso d. 82, 101–111 (2004).
30. Reichbo n-Kjenne ud, T. e al. Majo dep ession and dimensional
ep esen a ions o DSM-IV pe sonali y diso de s: a popula ion-
based win s udy. Psychol. Med. 40, 1475–1484 (2010).
31. Mo issey, K. & Kinde man, P. The impac o childhood
socioeconomic s a us on dep ession and anxie y in adul li e:
es ing he accumula ion, c i ical pe iod and social mobili y
hypo heses. SSM Popul. Heal h 11, 100576 (2020).
32. Lee, H. & Es ada-Ma ínez, L. M. T ajec o ies o dep essi e
symp oms and neighbo hood changes om adolescence o
adul hood: la en class g ow h analysis and mul ile el g ow h
cu e models. In . J. En i on. Res. Public Heal h 17, 1829 (2020).
33. Scholz, S. e al. The eco-exposome concep : suppo ing an
in eg a ed assessmen o mix u es o en i onmen al chemicals.
En i on. Toxicol. Chem. 41, 30–45 (2022).
34. Wang, Z., Whipp, A., Heinonen-Guzeje , M. & Kap io, J. Age a
sepa a ion o win pai s in he FinnTwin12 s udy. Twin Res. Hum.
Gene . h ps://doi.o g/10.1017/ hg.2022.17 (2022).
35. Nilsson, E.-L. Analyzing gende di e ences in he ela ionship
be ween amily in luences and adolescen o ending among boys
and gi ls. Child Indic. Res. 10, 1079–1094 (2017).
36. Mensah, F. K. & Kie nan, K. E. Gende di e ences in educa ional
a ainmen : in luences o he amily en i onmen . B . Educ. Res. J.
36, 239–260 (2010).
37. Blaauboe , M. & Mulde , C. H. Gende di e ences in he impac
o amily backg ound on lea ing he pa en al home. J. Hous. Buil
En i on. 25, 53–71 (2010).
38. Sikes-Keilp, C. & Rubinow, D. R. In sea ch o sex- ela ed media o s
o a ec i e illness. Biol. Sex Di e . 12, 55 (2021).
39. Rebuli, M. E. & Pa isaul, H. B. Assessmen o sex speci ic
endoc ine dis up ing e ec s in he p ena al and p e-
pube al oden b ain. J. S e oid Biochem. Mol. Biol. 160,
148–159 (2016).
40. Ma ia, F. e al. Associa ion o long- e m exposu e o a ic- ela ed
ai pollu ion wi h blood p essu e and hype ension in an adul
popula ion–based coho in Spain ( he REGICOR s udy). En i on.
Heal h Pe spec . 122, 404–411 (2014).
41. Chung, M. K., Buck Louis, G. M., Kannan, K. & Pa el, C. J.
Exposome-wide associa ion s udy o semen quali y:
sys ema ic disco e y o endoc ine dis up ing chemical
bioma ke s in e ili y equi e la ge sample sizes. En i on. In .
125, 505–514 (2019).
42. Whipp, A. M. e al. Ea ly adolescen agg ession p edic s an isocial
pe sonali y diso de in young adul s: a popula ion-based s udy.
Eu . Child Adolesc. Psychia y 28, 341–350 (2019).
43. Bucholz, K. K. e al. A new, semi-s uc u ed psychia ic in e iew
o use in gene ic linkage s udies: a epo on he eliabili y o he
SSAGA. J. S ud. Alcohol 55, 149–158 (1994).
44. Bell, C. C. DSM-IV: Diagnos ic and S a is ical Manual o Men al
Diso de s. JAMA 272, 828–829 (1994).
45. Rose, R. J. e al. FinnTwin12 coho : an upda ed e iew. Twin Res.
Hum. Gene . 22, 302–311 (2019).
46. Kokko, K. & Pulkkinen, L. Unemploymen and psychological
dis ess: media o e ec s. J. Adul De . 5, 205–217 (1998).
47. an Kamp, I. e al. Ea ly en i onmen al quali y and li e-cou se
men al heal h e ec s: he Equal-Li e p ojec . En i on. Epidemiol.
6, e183 (2022).
48. Jul ez, J. e al. Ea ly li e mul iple exposu es and child cogni i e
unc ion: a mul i-cen ic bi h coho s udy in six Eu opean
coun ies. En i on. Pollu . 284, 117404 (2021).
49. Huppe z, C. e al. The e ec s o pa en al educa ion on exe cise
beha io in childhood and you h: a s udy in Du ch and Finnish
wins. Scand. J. Med. Sci. Spo s 27, 1143–1156 (2017).
Na u e Men al Heal h
A icle h ps://doi.o g/10.1038/s44220-023-00124-x
Ex ended Da a Fig. 6 | Calenda imeline o included exposu es and ou comes. The in o ma ion s a ed o be eco ded in 1983 du ing he p egnancy o wins’
mo he s and un il 2015.
Na u e Men al Heal h
A icle h ps://doi.o g/10.1038/s44220-023-00124-x
Ex ended Da a Fig. 7 | Diag am o bi a ia e Cholesky AE decomposi ion model ( : co ela ion). No e: ci cles deno e la en ac o s o addi i e gene ic (A) and
unique en i onmen al (E) componen s, while ec angles deno e measu ed/calcula ed a iables.
