Ketone body 3-hydroxybutyrate as a biomarker of aggression

1
Vol.:(0123456789)
Scien i ic Repo s | (2021) 11:5813 | h ps://doi.o g/10.1038/s41598-021-84635-6
www.na u e.com/scien i ic epo s
Ke one body 3‑hyd oxybu y a e
as a bioma ke o agg ession
A. M. Whipp1*, E. Vuoksimaa1, T. Ko honen1, R. Pool2,3, A. Bu 4, L. Lig ha 2,
F. A. Hagenbeek2,3, M. Ba els2,3, L. H. Bogl1,5, L. Pulkkinen6, R. J. Rose7, D. I. Boomsma2,3 &
J. Kap io1,4
Human agg ession is a complex beha iou , he biological unde pinnings o which emain poo ly
known. To gain insigh s in o agg ession biology, we s udied ela ionships wi h agg ession o 11 low‑
molecula ‑weigh me aboli es (amino acids, ke one bodies), p ocessed using 1H nuclea magne ic
esonance spec oscopy. We used a disco e y sample o young adul s and an independen adul
eplica ion sample. We s udied 725 young adul s om a popula ion‑based Finnish win coho bo n
1983–1987, wi h agg ession le els a ed in adolescence (ages 12, 14, 17) by mul iple a e s and blood
plasma samples a age 22. Linea eg ession models speci ied me aboli es as he esponse a iable and
agg ession a ings as p edic o a iables, and included se e al po en ial con ounde s. All me aboli es
showed low co ela ions wi h agg ession, wi h only one—3‑hyd oxybu y a e, a ke one body
p oduced du ing as ing—showing signi ican (nega i e) associa ions wi h agg ession. E ec sizes
o di e en a e s we e gene ally simila in magni ude, while eache ‑ a ed (age 12) and sel ‑ a ed
(age 14) agg ession we e bo h signi ican p edic o s o 3‑hyd oxybu y a e in mul i‑ a e models. In an
independen eplica ion sample o 960 adul s om he Ne he lands Twin Regis e , highe agg ession
(sel ‑ a ed) was also ela ed o lowe le els o 3‑hyd oxybu y a e. These explo a o y epidemiologic
esul s wa an u he s udies on he ole o ke one me abolism in agg ession.
Agg ession is a signi ican beha iou al p oblem a ec ing child en and adul s, wi h consequences no only o
indi iduals, bu amilies, schools, and communi ies as well1. I is also a complex and he e ogenous beha iou ha
has been examined om mul iple pe spec i es, including epidemiologic2, gene ic3, epigene ic4,5, and bioma ke s6,
o name a ew. Wi h he he i abili y o agg ession in bo h child en and adul s ound o be abou 50%3, agg ession
has a s ong biological basis. Ye , much a ia ion in agg essi e beha iou emains o be explained. Some o he
mechanisms, pa hways, and molecules in ol ed in he biological-le el aspec s o he beha iou a e known; how-
e e , in bo h adul s and child en, he e emain biological pa hways o disco e and bioma ke s ye o iden i y6–8.
Wi h inc eased unde s anding o he biology o agg ession, he oppo uni y o imp o e ou in e en ions as
well as pha macological ea men s a ises. Thus, he e is a need o gene a e new hypo heses and app oaches o
enhance ou unde s anding o agg ession and i s associa ed p oblems.
One such sugges ed app oach is using p o on nuclea magne ic esonance (NMR) spec oscopy o ind new
bioma ke s, because i allows o a hypo hesis- ee way o examine pheno ypic associa ions9–11. NMR is no a new
echnology, bu applying he app oach o beha iou al/psychological p oblems, including agg ession, is gaining
in e es 6. The NMR me hodology is gene ally able o p oduce in o ma ion on he le els o lipids, amino acids,
o ganic acids, and o he small o ganic molecules. Associa ions be ween lipid concen a ions and agg ession ha e
p e iously been no ed6,12–15, while o he low-molecula -weigh molecules migh hold mo e p omise o no el
indings. Amino acids and ke one bodies (low-molecula -weigh molecules ca ego ized in NMR) ha e been asso-
cia ed wi h emo ional and beha iou al p oblems o he han agg ession16–20, and some s udies ha e shown amino
acids and simila molecules associa ed wi h ange and psychopa hy21, schizoph enia- ela ed iolence15, and
childhood agg ession22. Acco dingly, hese a eas o e ed a easonable place o ocus ou explo a o y in es iga ion.
Thus, in his s udy we combine an NMR app oach wi h he agg ession pheno ype in e o s o gene a e new
insigh s in o ou biological unde s anding o agg ession. We conduc ed explo a o y analyses on he associa ion
be ween agg ession and amino acid and ke one body bioma ke s, seeking o explo e associa ion(s) ound in mo e
OPEN
1Ins i u e o Molecula Medicine Finland (FIMM), Uni e si y o Helsinki, Helsinki, Finland. 2Depa men o
Biological Psychology, V ije Uni e si ei Ams e dam, Ams e dam, The Ne he lands. 3Ams e dam Public Heal h
(APH) Resea ch Ins i u e, Ams e dam, The Ne he lands. 4Clinicum, Depa men o Public Heal h, Uni e si y
o Helsinki, Helsinki, Finland. 5Depa men o Epidemiology, Cen e o Public Heal h, Medical Uni e si y o
Vienna, Vienna, Aus ia. 6Depa men o Psychology, Uni e si y o Jy askyla, Jy askyla, Finland. 7Depa men o
Psychological and B ain Sciences, Indiana Uni e si y, Blooming on, Indiana, USA. *email: alyce.[email p o ec ed]
2
Vol:.(1234567890)
Scien i ic Repo s | (2021) 11:5813 | h ps://doi.o g/10.1038/s41598-021-84635-6
www.na u e.com/scien i ic epo s/
dep h and wi h an a emp o eplica e hem in an independen sample. The disco e y da ase includes longi u-
dinal agg ession measu emen s in adolescence and bioma ke da a in young adul hood, while he eplica ion
da ase includes bioma ke and agg ession da a in young adul s.
Ma e ials and me hods
Disco e y da ase (FinnTwin12). The FinnTwin12 s udy is a longi udinal popula ion-based coho o
Finnish wins bo n in 1983–1987 ha was ini ia ed o ack beha iou al de elopmen and heal h habi s om
mid-childhood onwa d23,24. Twins and hei amilies we e iden i ied h ough he Finnish Cen al Popula ion
Regis y and ini ial en olmen o he wins occu ed be ween he ages o 11 and 12. Thebaseline esponse a e
was 87% (N = 5600 wins) and emained high (be ween 85 and 90%) o all da a collec ion wa es. Ques ionnai e
da a collec ion wa es occu ed a ound ages 11/12, 14, 17, and 22, and we e collec ed om pa en s (age 11/12),
eache s (age 11/12 and 14), and wins hemsel es (ages 14, 17 and 22).
Addi ionally, a subse o wins ( om 1035 amilies) was mo e in ensi ely s udied using addi ional assessmen s
(semi-s uc u ed psychia ic in e iew and ques ionnai es) in young adul hood (mean age = 22.4, SD = 0.70;
n = 1347, esponse a e 73.0%), as well as a plasma sample, an h opome ics and neu opsychological es s o
wins es ed in pe son in young adul hood (n = 779). P e ious in es iga ion has shown ha hese mo e in ensi ely
s udied wins do no di e om he es o he coho ega ding beha iou al o emo ional p oblems25.
A blood sample was collec ed a e an o e nigh as wi h he eques ha he pa icipan abs ain om
alcohol and obacco since he nigh be o e he sample. Plasma was ex ac ed immedia ely and s o ed a − 80°C.
Samples we e p ocessed in one ba ch in he au umn o 2010 using an au oma ed high- h oughpu se um NMR
me abolomics pla o m (i.e., he B ainshake/Nigh ingale pla o m)24,26,27. This pla o m has demons a ed good
quali y con ol ela ed o sample p epa a ion and expe imen al ac o s28,29. The me aboli es u ilized he e we e
he amino acids (alanine, glu amine, his idine, isoleucine, leucine, phenylalanine, y osine, aline) and ke one
bodies (ace a e, ace oace a e, 3-hyd oxybu y a e). All me aboli e da a we e a ailable wi h uni s mmol/l. P egnan
women (n = 53) and indi iduals on choles e ol medica ion (n = 1) we e excluded om analysis; hose excluded
we e no di e en in hei agg ession sco es compa ed o hose included. Thus, he inal sample o s udy was
725 win indi iduals.
E hical app o al o all wa es o da a collec ion was ob ained om he e hical commi ee o he Helsinki and
Uusimaa Uni e si y Hospi al Dis ic and Indiana Uni e si y’s Ins i u ional Re iew Boa d. All da a collec ion and
sampling was pe o med in compliance wi h he e hical guidelines. A ages 12 and 14, pa en s p o ided consen
o he wins, while wins hemsel es p o ided w i en consen o age 17 and 22 collec ion wa es.
Agg ession measu emen . Beha iou al and emo ional de elopmen in FinnTwin12 we e measu ed a ages 12,
14, and 17 ( om pa en s, eache s, wins hemsel es, and co- wins a ed hei win as well) using he modi-
ied Mul idimensional Pee Nomina ion In en o y (MPNI)30. The MPNI is a 37-i em ques ionnai e p oducing
mul iple subscales including agg ession (6 i ems; 4 di ec agg ession i ems, 2 indi ec agg ession i ems). Each
ques ion abou he child’s beha iou (e.g., “Does he child call people names when ang y wi h hem?”) has ou
esponse choices: ‘no obse ed in child’, ‘obse ed some imes, bu no consis en ly’, ‘de ini ely obse ed, bu is
no p ominen ’, and ‘clea ly obse ed’. Response choices a e sco ed 0–3, and agg ession sco es o indi idual
a e s (e.g., eache a ing a age 12) a e o med by aking he mean o all i ems in he agg ession subscale (no
missing alues we e allowed). A gene al agg ession sco e was c ea ed (using all 6 i ems). We use he ou di e -
en ypes o a e s a h ee di e en ages (se en agg ession a ings in o al) because he li e a u e indica es ha
di e en a e s obse e di e en aspec s o child en’s beha iou in he di e en se ings a ailable o hem31,32,
and by using mul iple a ings we can assess he obus ness o he bioma ke associa ions ac oss se ings, obse -
e s, and adolescence. Addi ionally, a composi e agg ession sco e (called ‘Agg ession Combined’) was used o
ini ial analyses in o de o minimize mul iple es ing and o combine in o ma ion om di e en in o man s,
who end o a e agg ession sligh ly di e en ly31,33. I was o med by calcula ing he mean o he agg ession sub-
scale sco es o he pa en a ing a age 12 and he eache a ings a ages 12 and 14 combined.
Co a ia es. Based on ca e ul li e a u e e iew, se e al co a ia es, measu ed a he ime o sample collec ion,
we e included in he eg ession model because hey we e ela ed o bo h bioma ke le els and agg ession, and
hey we e a ailable in bo h he disco e y and he eplica ion samples. Body mass index (BMI) was calcula ed
as kg/m2 (mean: 23.3, s anda d de ia ion (SD): 3.94). Leisu e- ime physical ac i i y was exp essed as me abolic
equi alen o ask (MET) hou s/day based on s uc u ed ques ions conside ing he in ensi y, equency, and
du a ion o ac i i ies34. Smoking s a us was ca ego ized as cu en , o me , ne e /expe imen e . Alcohol con-
sump ion equency was ca ego ized as none, once a mon h o less, 2–4 imes pe mon h, 2–3 imes pe week, 4
o mo e imes pe week. BMI, physical ac i i y, alcohol use and smoking ha e been shown o be ela ed o bo h
me aboli e le els26,35–37 and agg ession38–40. Finally, sel - a ed gene al heal h was dicho omized o ei he poo
heal h (poo + ai ) o good heal h (good + e y good). This measu e was included o oughly cap u e soma ic
heal h p oblems (e.g., diabe es) in hese young adul s ha could a ec hei me aboli e le els41, and which could
also be ela ed o agg ession42.
Replica ion da ase (NTR). The Ne he lands Twin Regis e (NTR), es ablished in 1987, collec s da a on
wins and hei amily membe s. Da a collec ion is na ionwide ia ques ionnai es mailed o pa en s o young
wins (ages 1–12) and o wins hemsel es a age 14 and beyond. Twins ec ui ed as adolescen s and adul s and
hei amilies a e in i ed o ake pa in da a collec ion by su ey e e y 2–4 yea s43. Mul iple su ey wa es ha e
included he Achenbach Sys em o Empi ically Based Assessmen (ASEBA) Adul Sel -Repo (ASR)44, which
includes ques ions ha allow o he calcula ion o an agg ession sco e. A subg oup o adul pa icipan s was
3
Vol.:(0123456789)
Scien i ic Repo s | (2021) 11:5813 | h ps://doi.o g/10.1038/s41598-021-84635-6
www.na u e.com/scien i ic epo s/
in i ed in o he NTR Biobank p ojec 45,46. As pa o his p ojec , hey we e isi ed a home whe e samples we e
collec ed be ween 7:00 and 10:00 a e o e nigh as ing. Pa icipan s we e asked no o smoke be o e sampling.
Fe ile women ook pa om day 2–4 o hei mens ual cycle, o in he pill- ee week. Du ing he isi , body
composi ion measu emen s we e aken and in o ma ion ega ding physical heal h and li es yle beha iou (e.g.,
smoking and d inking pa e ns, exe cise, medica ion use) was collec ed.
Me aboli e da a, p ocessed using he same B ainshake/Nigh ingale pla o m as FinnTwin1227, we e assessed in
5204 pa icipan s (age ange: 18–87yea s). As in FinnTwin12, we excluded women who we e p egnan (n = 19);
hose excluded we e no di e en in hei agg ession sco es compa ed o included indi iduals. Addi ionally, all
indi iduals o e he age o 30yea s a he ime o sampling (n = 3843) we e excluded o make he FinnTwin12
and NTR samples simila in age composi ion and because changes in body chemis y occu ha accompany,
o example, menopause and he ageing p ocess o bo h sexes47. The wo da a collec ion imepoin s (wa e 8
and 10) ha ob ained agg ession sco es (ASR ques ionnai e) and we e closes o he ime o blood sampling
we e analysed. ASR agg ession sco es a wa e 8 we e a ailable om 819 pa icipan s (mean age: 28.1yea s,
ange 18–38yea s), and ASR agg ession sco es om wa e 10 om 665 pa icipan s (mean age: 32.7yea s, ange
21–39yea s). The agg ession sco e used o he analyses was he sco e closes o he ime o blood sample col-
lec ion. O no e, agg ession sco es we e gene ally a ailable close o he ime o sampling o he NTR da ase
(mean 3.4yea s; ange 0–10yea s) han he FinnTwin12. The e we e 150 (non-p egnan , age 30 o younge a
ime o blood sampling) pa icipan s who had agg ession sco e alues collec ed be o e hei blood sample, and
hese sco es we e excluded o make he da ase empo ally uni o m (blood samples i s , agg ession sco es la e ;
his is he opposi e o he empo ali y o he FinnTwin12 sample). Thei agg ession sco es we e no di e en han
hose o included indi iduals. Thus, he inal sample o s udy was 960 indi iduals. Fo eplica ion, only he
me aboli e 3-hyd oxybu y a e (mmol/l) was examined.
NTR p ojec s we e app o ed by he Cen al E hics Commi ee on Resea ch In ol ing Human Subjec s o he
VU Uni e si y Medical Cen e, Ams e dam, an Ins i u ional Re iew Boa d ce i ied by he U.S. O ice o Human
Resea ch P o ec ions (IRB numbe IRB00002991 unde Fede al-wide Assu ance FWA00017598; IRB/ins i u e
code NTR 03-180). All da a collec ion and sampling we e pe o med in compliance wi h he e hical guidelines,
and all pa icipan s p o ided w i en in o med consen .
Agg ession measu emen . Agg ession in he Adul NTR was analysed om su eys 8 and 10, which bo h
included he ASR, pa o he ASEBA o assessing p oblems and s eng hs in child en and adul s, wi h mul iple
subscales including agg ession44. Each agg ession ques ion has h ee esponse choices which we e summed o e
i ems, i he numbe o missing i ems was no la ge han h ee. Fo pa icipan s wi h no mo e han h ee miss-
ing alues, missing i ems we e impu ed. I pa icipan s had comple ed he ASR a bo h su eys 8 and 10, we ook
hei agg ession sco e closes o he blood sampling (bu always a e blood sampling).
Co a ia es. We aimed o mi o a iables and ca ego ies as close o he FinnTwin12 da ase as possible. BMI
was calcula ed as kg/m2 ( om he biobank home isi ) (mean: 23.0, SD: 3.43). Leisu e- ime physical ac i i y
( om he same su ey wa e ha he agg ession sco e was aken) was calcula ed as MET minu es. Smoking
s a us was ca ego ized as cu en , o me , ne e ( om he biobank home isi ). Alcohol consump ion equency
( om he same su ey wa e as he agg ession sco e) was ca ego ized as none, once a mon h o less, 2–4 imes
pe mon h, 2–3 imes pe week, 4 o mo e imes pe week. Sel - a ed gene al heal h ( om he same su ey wa e
ha he agg ession sco e was aken) was dicho omized o ei he poo heal h (poo + ai ) o good heal h ( eason-
able + good + e y good). Since he NTR me aboli e da a we e p ocessed in di e en ba ches, he e was a ba ch
a iable a ailable o adjus men .
Analysis. Fo desc ip i e s a is ics (supplemen al), we calcula ed means and s anda d de ia ions (SD) and
anges o he agg ession sco es o he di e en a e s ha we e used in he analyses in FinnTwin12 (as well as
in he NTR eplica ion da ase ).
Explo a o y analyses in FinnTwin12 in ol ing he a ailable low-molecula -weigh me aboli es (n = 11; 8
amino acids and 3 ke one bodies) u ilised he composi e agg ession sco e (Agg ession Combined). Spea man co -
ela ions we e compu ed be ween he un ans o med me aboli es, and be ween he me aboli es and he Agg es-
sion Combined a iable. The ea e , due o hei non-no mal dis ibu ion, all me aboli es we e ank ans o med
(gene ally he bes solu ion o mos o he me aboli es). Then, linea eg ession was pe o med wi h indi idual
me aboli es as he esponse a iable and Agg ession Combined as he main p edic o a iable, adjus ed o
age, sex, and BMI. To ob ain obus s anda d e o s (i.e., co ec ed o amilial ela edness), he clus e op ion
in S a a was speci ied48. Be a coe icien s and 95% con idence in e als (CI) we e epo ed and hose eaching
signi icance (p < 0.05) we e in es iga ed u he . A e explo a o y analyses we e pe o med, we p oceeded o
in es iga e signi ican indings u he in an independen sample o young adul s.
To u he in es iga e 3-hyd oxybu y a e ( he only bioma ke ha eached a nominal p < 0.05 adjus ed o
mul iple es ing o 9 independen a iables ob ained om Ma ix Spec al Decomposi ion (ma SpD: h ps ://
gum p.qim .edu.a u/g ene al/daleN /ma Sp D/), which es ima ed he equi alen numbe o independen a iables in
ou co ela ion ma ix o 11 me aboli es), addi ional linea eg ession modelling was employed o es ablish con-
sis ency o he associa ion. In addi ion o he p ima y Agg ession Combined a iable, we examined he sepa a e
gene al agg ession a ings om di e en a e s as well (pa en s a age 12, eache (s) a age 12, eache (s) a age 14,
sel a ings a ages 14 and 17, and he win’s co- win a ages 14 and 17). Ini ial models designa ed 3-hyd oxybu-
y a e as he esponse a iable and he di e en gene al agg ession a ings as he main p edic o a iable, adjus ed
o age, sex, BMI, and amilial ela edness. Fu he models addi ionally adjus ed o leisu e- ime physical ac i i y
(METs), smoking s a us, alcohol consump ion equency, and sel - a ed gene al heal h. Sex in e ac ions we e also
4
Vol:.(1234567890)
Scien i ic Repo s | (2021) 11:5813 | h ps://doi.o g/10.1038/s41598-021-84635-6
www.na u e.com/scien i ic epo s/
in es iga ed. Supplemen al pos hoc analysis in es iga ed whe he including mul iple gene al agg ession a ings
in he same model could cla i y he esul s, since a p e ious in es iga ion o ou s indica ed ha mul iple a e s
in one model imp o ed agg ession in o ma ion (di e en a e s obse e in di e en se ings and can p o ide
unique in o ma ion31) and imp o ed p edic ion o a la e pe sonali y diso de 33.
To ob ain u he e idence o suppo o e u e he associa ion ound in ou disco e y sample, we ana-
lysed heNTR da ase o es eplica ion o ou esul s. The ini ial Spea man co ela ion o he un ans o med
3-hyd oxybu y a e and he agg ession a ing was compu ed. Then, he 3-hyd oxybu y a e a iable was ank ans-
o med, as in he FinnTwin12 da a. Linea eg ession models we e un wi h 3-hyd oxybu y a e as he esponse
a iable and he agg ession a ing as he p edic o a iable, adjus ed o age, sex, BMI, ba ch and amilial ela -
edness. Addi ionally, ully adjus ed models u he adjus ed o leisu e- ime physical ac i i y (METs), smoking
s a us, alcohol consump ion equency, and sel - a ed gene al heal h. Sex in e ac ions we e also in es iga ed o
s udy i he agg ession–3-hyd oxybu y a e ela ionship di e ed in males and emales49.
Finally, addi ional analyses we e un o e i y he maximum likelihood es ima es o he associa ion exam-
ined in he ully adjus ed equen is models abo e wi h a Bayesian app oach, a powe ul ool o inco po a ing
in o ma ion om p e ious s udies, o con olling con ounding, o in e p e a ion o s udy esul s and e alua ion
o hypo heses abou exposu e–disease ela ionships50. We i ed se e al Bayesian andom-in e cep (2-le el o
mixed) eg ession models sepa a ely on he FinnTwin12 and NTR da a wi h 3-hyd oxybu y a e as he esponse
a iable and he agg ession a ing as he p edic o a iable (we used ‘Agg ession Combined’ o FinnTwin12).
We adjus ed o age, sex, BMI, MET, smoking s a us, alcohol consump ion equency, and sel - a ed gene al
heal h (ba ch was also included o NTR), and an Bayesian models wi h and wi hou an in e ac ion o sex and
agg ession included, as well as s a i ied on sex. The clus e ing o he wins was accoun ed o by se ing each
win pai a andom in e cep .
The Bayesian model pa ame e s o es ima e included he o e all in e cep , he eg ession coe icien s and
he andom in e cep s along wi h hei a iance componen s. Fo he o e all in e cep and he eg ession coe -
icien s, we se a weakly in o ma i e (de aul ) p io (no mal (0,10,000)). Fo he a iance o he e o e m, we se
a non-in o ma i e (de aul ) p io (in e se-gamma p io wi h shape and scale pa ame e s equal o 0.1). Fo he
andom in e cep s, we se a hie a chical p io no mal (0,sigma), wi h he hype pa ame e sigma se a uni o m
p io wi h bounds a he 95% con idence in e al o he maximum likelihood es ima e o he a iance o he
andom in e cep s (i.e., 95% CI o he a iance om he equen is andom-in e cep model). In ac , he e
appea ed a con e gence p oblem when using a non-in o ma i e (de aul ) p io o he a iance o he andom
in e cep s. To o e come his p oblem, we inco po a ed in o ma ion on he p obable ange o he a iance ia he
p io dis ibu ion. Gibbs sampling was used o d aw samples om he pos e io dis ibu ion (20,000 i e a ions,
including a bu n-in o 7000 i e a ions). To con i m he con e gence o Ma ko chain Mon e Ca lo (MCMC)
chains, we pe o med a isual e alua ion ( ace, densi y and au oco ela ion plo s). F om hese models, we
epo ed he es ima ed eg ession coe icien o agg ession as he pos e io expec a ion (mean) and 95% c ed-
ible in e als assessed om he pos e io dis ibu ion. Las ly, o assess whe he he use o in o ma i e p io s
a ec s he esul s, we pe o med a sensi i i y analysis by inco po a ing in o ma ion on he associa ion ob ained
om he Bayesian sex-s a i ied analysis o he FinnTwin12 coho o he p io s o he agg ession coe icien s
in he NTR sex-s a i ied models. The coe icien s we e assigned he no mal dis ibu ion wi h he mean se o
he pos e io mean and he a iance se o he squa ed s anda d de ia ion ob ained om he co esponding
Bayesian sex-s a i ied analysis on he FinnTwin12 coho .
All analyses we e pe o med using S a a e sion 15.1 o 16.0 (S a a Co po a ion, College S a ion, TX, USA).
Resul s
Disco e y (FinnTwin12) esul s. Desc ip i e means (SD) and anges o he agg ession sco es om he
di e en a e s used in he analyses a e p o ided in Supplemen al Table1. Spea man co ela ions be ween
un ans o med me aboli es indica ed gene ally low o mode a e associa ions, excep he high co ela ions
be ween 3-hyd oxybu y a e and ace oace a e ( = 0.70) and he b anched-chain amino acids (leucine, isoleucine,
and aline) wi h each o he ( ange = 0.62–0.73) (Table1). Spea man co ela ions be ween me aboli es and he
Agg ession Combined a iable we e low, anging om − 0.13 o 0.14.
Nex , linea eg ession models wi h ank- ans o med me aboli es as he esponse a iable and Agg ession
Combined as he main p edic o a iable, also adjus ed o age, sex, BMI, and amilial ela edness, e ealed ha
only 3-hyd oxybu y a e was signi ican ly associa ed wi h agg ession, includinga e adjus men o mul iple
es ing (Fig.1; Supplemen al Table2). The associa ion was nega i e, wi h lowe 3-hyd oxybu y a e le els being
associa ed wi h highe agg ession le els.
In linea eg ession models u he explo ing 3-hyd oxybu y a e’s associa ion wi h agg ession, he basic
model (3-hyd oxybu y a e as he esponse a iable, and agg ession a ings om di e en a e s as main p edic o
a iables in sepa a e models, adjus ed o age, sex, BMI, and amilial ela edness) esul s indica ed ha nea ly all
agg ession a ings demons a ed he end o nega i e associa ion o 3-hyd oxybu y a e wi h agg ession, wi h
eache a ings a age 12 and sel a ings a age 14 showing he s onges e ec sizes (and also being s a is ically
signi ican ) (Fig.2; Supplemen al Table3). In ully adjus ed models (basic model plus MET, smoking s a us,
alcohol consump ion equency, and sel - a ed gene al heal h), he nega i e associa ions emained, wi h only
small a enua ion o he e ec sizes (Table2). No sex in e ac ions we e de ec ed (p- alue ange = 0.10–0.84;
Supplemen al Table4).
In supplemen al analyses, because he agg ession sco es om he di e en a e s we e only weakly o mod-
e a ely co ela ed ( ange: 0.14–0.47), mul iple agg ession a ings we e added in o he same model as p edic o
a iables, adjus ed o age, sex, BMI, and amilial ela edness. The es ed combina ions included he h ee agg es-
sion sco es ha made up he Agg ession Combined a iable (pa en age 12, eache age 12, eache age 14), bo h
5
Vol.:(0123456789)
Scien i ic Repo s | (2021) 11:5813 | h ps://doi.o g/10.1038/s41598-021-84635-6
www.na u e.com/scien i ic epo s/
Table 1. FinnTwin12 Spea man co ela ions o un ans o med me aboli es and agg ession (n=526).
Abb e ia ions: ace ace a e, acace ace oace a e, AggComb agg ession combined, bohbu 3-hyd oxybu y a e,
ala alanine, gln glu amine, his his idine, ile isoleucine, leu leucine, phe phenylalanine, y y osine, al aline.
a Agg ession is ’Agg ession Combined’ a iable (which is he mean o pa en -12, eache -12, and eache -14
agg ession a ings).
ace acace bohbu ala gln his ile leu phe y al
acace 0.11
bohbu 0.15 0.70
ala −0.12 −0.38 −0.36
gln 0.36 −0.03 −0.05 0.06
his 0.07 −0.01 −0.11 0.20 0.00
ile 0.11 0.22 −0.07 0.19 0.28 0.09
leu 0.12 0.13 −0.04 0.08 0.17 0.02 0.73
phe −0.06 −0.03 −0.05 0.16 −0.32 0.30 0.15 0.34
y 0.16 −0.12 −0.23 0.38 0.39 0.20 0.42 0.30 0.15
al 0.20 0.16 −0.02 0.07 0.37 0.05 0.68 0.62 0.03 0.55
AggComba−0.04 0.04 −0.13 0.04 0.09 0.07 0.14 0.10 −0.06 0.13 0.11
-150
-100
-50
0
50
100
150
aceacace bohbu alagln hisile leuphe y al
Be a Coe icien
Figu e1. FinnTwin12 eg ession coe icien s (and 95% con idence in e als) o Agg ession Combined
p edic o a iable in linea eg ession models wi h me aboli es as esponse a iables, adjus ed o age, sex, body
mass index, and amilial ela edness. ace ace a e, acace ace oace a e, bohbu 3-hyd oxybu y a e, ala alanine, gln
glu amine, his his idine, ile isoleucine, leu leucine, phe phenylalanine, y y osine, al aline.
-150
-100
-50
0
50
100
150
AggC P12T12 T14S14 TW14 S17TW17
Be a Coe icien
Figu e2. FinnTwin12 eg ession coe icien s (and 95% con idence in e als) o agg ession a ings
om di e en a e s as (sepa a e) p edic o a iables in linea eg ession models wi h ank- ans o med
3-hyd oxybu y a e as he esponse a iable, adjus ed o age, sex, body mass index, and amilial ela edness.
AggC agg ession combined, P12 pa en a ing age 12, T12 eache a ing age 12, T14 eache a ing age 14, S14
sel a ing a age 14, TW14 co- win a ing a age 14, S17 sel a ing a age 17, TW17 co- win a ing a age 17.

6
Vol:.(1234567890)
Scien i ic Repo s | (2021) 11:5813 | h ps://doi.o g/10.1038/s41598-021-84635-6
www.na u e.com/scien i ic epo s/
age 12 a ings, all h ee age 14 a ings, and bo h age 17 a ings, as well as he wo a ings ha we e consis en ly
signi ican ( eache age 12 and sel age 14). In e es ingly, in he model including eache age 12 and sel age 14
a ings, bo h we e s a is ically signi ican ly nega i ely associa ed wi h 3-hyd oxybu y a e (Supplemen al Table5).
Replica ion (NTR) esul s. In he NTR eplica ion da ase , he Spea man co ela ion be ween he agg es-
sion sco e and un ans o med 3-hyd oxybu y a e was − 0.07 (p = 0.045). The ini ial linea eg ession model wi h
ank- ans o med 3-hyd oxybu y a e as he esponse a iable and agg ession as he main p edic o a iable,
adjus ed o age, sex, BMI, ba ch and amilial ela edness, showed he same nega i e associa ion end (p = 0.138)
as he FinnTwin12 models (Table3). The e was no sex in e ac ion de ec ed in he ini ial model (p = 0.246), how-
e e , in he ully adjus ed model he e was (p = 0.015). Rega ding he ully adjus ed models, he nega i e associa-
ion was p esen o males only; i became sligh ly posi i e in di ec ion o emales, albei wi h wide con idence
in e als. Table3 displays esul s o he sexes pooled and males and emales modelled sepa a ely.
Table 2. FinnTwin12 mul i a iable linea eg ession models wi h 3-hyd oxybu y a e as he esponse a iable,
eg ession coe icien s o agg ession a ings om di e en a e s as p edic o a iables in sepa a e models,
adjus ed o age, sex, BMI, MET, smoking s a us, alcohol consump ion equency, sel - a ed gene al heal h, and
amilial ela edness. AggCombined agg ession combined, BMI body mass index, CI con idence in e al, MET
me abolic equi alen o ask, P12 pa en a ing age 12, S14 sel a ing a age 14, S17 sel a ing a age 17, T12
eache a ing age 12, T14 eache a ing age 14, TW14 co- win a ing a age 14, TW17 co- win a ing a age 17.
*This co- a ia e also signi ican (p < 0.05). a R-squa ed ep esen s he a ia ion explained om all a iables in
he model oge he .
Agg ession a ing Co- a ia es N Uns anda dised Be a Coe
(95% CI) S anda dised Be a Coe
(p- alue) R-squa eda
AggCombined Age, sex, BMI*, MET, smok-
ing, alcohol, heal h 524 − 84.1 (− 143.9, − 24.4) − 0.13 (0.006) 0.042
P12 Age, sex, BMI*, MET, smok-
ing, alcohol, heal h 677 − 4.5 (− 54.0, 45.0) − 0.08 (0.9) 0.024
T12 Age, sex, BMI*, MET, smok-
ing, alcohol, heal h 688 − 49.1 (− 78.4, − 19.9) − 0.14 (0.001) 0.039
T14 Age, sex, BMI*, MET, smok-
ing, alcohol, heal h 553 − 13.3 (− 60.3, 33.7) − 0.03 (0.6) 0.023
S14 Age, sex, BMI*, MET, smok-
ing, alcohol, heal h 684 − 56.2 (− 105.2, − 7.2) − 0.09 (0.02) 0.030
TW14 Age, sex, BMI*, MET, smok-
ing, alcohol, heal h 682 2.2 (− 38.2, 42.6) 0.004 (0.9) 0.022
S17 Age, sex, BMI*, MET, smok-
ing, alcohol, heal h 639 − 11.3 (− 53.5, 31.0) − 0.02 (0.6) 0.018
TW17 Age, sex, BMI, MET, smoking,
alcohol, heal h 632 − 15.8 (− 49.0, 17.4) − 0.04 (0.3) 0.018
Table 3. NTR mul i a iable linea eg ession models (ini ial and ully adjus eda) wi h 3-hyd oxybu y a e as
he esponse a iable, eg ession coe icien s o agg ession a ing as p edic o a iable, wi h sexes pooled and
sepa a ed. a Ini ial model: adjus ed o age, sex, BMI, ba ch, and amilial ela edness. Fully adjus ed model:
adjus ed o age, sex, BMI, ba ch, MET, smoking s a us, alcohol consump ion equency, sel - a ed gene al
heal h, and amilial ela edness. b R-squa ed ep esen s he a ia ion explained om all a iables in he model
oge he .
Model Co- a ia es N Uns anda dised Be a Coe
(95% CI) S anda dised Be a Coe
(p- alue) R-squa edb
Ini ial modela
Bo h sexes Age, sex*, BMI, ba ch* 811 − 22.6 (− 52.5, 7.3) − 0.05 (0.1) 0.151
Male Age, BMI, ba ch* 253 − 48.8 (− 118.7, 21.2) − 0.09 (0.2) 0.221
Female Age, BMI, ba ch* 558 − 14.1 (− 47.6, 19.5) − 0.03 (0.4) 0.108
Fully adjus ed modela
Bo h Sexes Age, sex*, BMI, ba ch*, MET,
smoking, alcohol*, heal h 567 − 11.9 (− 48.4, 24.7) − 0.03 (0.5) 0.153
Male Age, BMI, ba ch*, MET*, smok-
ing, alcohol, heal h 165 − 85.5 (− 171.8, 0.8) − 0.14 (0.052) 0.270
Female Age, BMI, ba ch*, MET, smok-
ing, alcohol*, heal h 402 11.8 (− 28.4, 52.1) 0.03 (0.6) 0.102
7
Vol.:(0123456789)
Scien i ic Repo s | (2021) 11:5813 | h ps://doi.o g/10.1038/s41598-021-84635-6
www.na u e.com/scien i ic epo s/
Bayesian modelling. Fo all he i ed Bayesian models, he con e gence checks e ealed no au oco ela-
ion and demons a ed he MCMC achie ed he a ge dis ibu ion. In FinnTwin12, he ini ial non-in o ma i e
p io s model wi hou he sex in e ac ion e m esul ed in an agg ession a ing a iable pos e io mean o − 84.0
(95% c edible in e al: − 91.0, − 76.9), while he model wi h he sex × agg ession in e ac ion e m indica ed a
pos e io mean o − 86.6 (95% c edible in e al: − 100.4, − 72.5) o he in e ac ion e m. Fo he males-only
model, he agg ession pos e io mean was − 37.9 (95% c edible in e al: − 43.8, − 32.0). Fo he emales-only
model, he agg ession pos e io mean was − 119.5 (95% c edible in e al: − 127.8, − 111.1). In NTR, he ini-
ial non-in o ma i e p io s model wi hou he sex in e ac ion e m esul ed in an agg ession a ing a iable
pos e io mean o 10.6 (95% c edible in e al: − 15.3, − 5.9), while he model wi h he sex × agg ession in e ac-
ion e m indica ed a pos e io mean o 83.5 (95% c edible in e al: 72.5, 94.4) o he in e ac ion e m. Fo
he males-only model, he agg ession pos e io mean was − 80.4 (95% c edible in e al: − 86.3, − 74.6). Fo he
emales-only model, he agg ession pos e io mean was 14.2 (95% c edible in e al: 9.9, 18.5). Finally, in he sen-
si i i y analysis o NTR sex-s a i ied models using (FinnTwin12-based) in o ma i e p io s o he sex-speci ic
eg ession coe icien s o agg ession, he esul s emained simila (da a no shown).
Discussion
In his s udy, we ini ially es ed 11 po en ial NMR-p ocessed low-molecula -weigh bioma ke s o associa ion
wi h a composi e agg ession sco e, and 3-hyd oxybu y a e, a ke one body, was he only nominally s a is ically
signi ican bioma ke (nega i ely associa ed wi h agg ession). Upon u he analysis, in bo h basic and ully
adjus ed models, nea ly all agg ession a ings ( om di e en a e s and ages) showed a consis en pa e n o
nega i e associa ion, wi h h ee being s a is ically signi ican . When including bo h eache (age 12) and sel (age
14) agg ession a ings in he same model, bo h we e signi ican ly associa ed wi h 3-hyd oxybu y a e, showing ha
an ex e nal a e ( eache ) and an adolescen ’s sel - a ing con ibu e uniquely o his associa ion. Impo an ly, we
we e able o assess his bioma ke –agg ession associa ion in a la ge and independen sample, he NTR. In bo h
basic and ully adjus ed NTR ( equen is ) models, he nega i e associa ion o 3-hyd oxybu y a e and agg ession
was ound, bu s a is ical signi icance was no eached un il he sexes we e sepa a ed in he ully adjus ed model
(as indica ed by he signi ican sex in e ac ion). Finally, Bayesian modelling indica ed a sex in e ac ion in bo h
FinnTwin12 and NTR coho s, and he nega i e associa ion—low 3-hyd oxybu y a e le els wi h high agg ession
le els— emained in sex-s a i ied models (wi h non-in o ma i e p io s) o bo h sexes in he FinnTwin12 coho
and o males only in he NTR coho .
The associa ion o 3-hyd oxybu y a e wi h agg ession is unlikely o be spu ious because i was seen in wo
independen samples, and was consis en ac oss a e s and ages. Because we we e no es ing any speci ic hypo h-
esis, and hus we e no guided solely by s a is ical signi icance (p- alues), ou ocus was on iden i ying epea ed
pa e ns o associa ions, which we saw he e. Addi ionally, al hough we we e no add essing causali y in hese
models, we saw s abili y in he end ega dless o empo ali y (bioma ke measu ed be o e o a e agg ession).
Sex di e ences we e e iden , bu hese will need o be cla i ied by u u e in es iga ions. Fu he mo e, he end
in FinnTwin12 was nea ly uni o m ac oss a e s and ages. This sugges s ha he associa ion is ela ed o gene -
alized agg ession (liabili y o beha e agg essi ely in di e en con ex s), which is a mo e se ious and impo an
ype o agg ession because i leads o he poo es ou comes51. Ou model including mul iple a e s o agg ession
in he same model also suppo s his no ion o gene alized agg ession, wi h a ings om bo h he eache s and
adolescen s hemsel es uniquely con ibu ing o he associa ion o agg ession wi h 3-hyd oxybu y a e. The een
him/he sel p o ides hei pe spec i e abou hei beha io ac oss all en i onmen s, while eache s ha e a unique
pe spec i e. Compa ed o pa en s o example, eache s egula ly obse e he child amongs o he child en o
he same age, making o mo e alid p edic ions o he s uden ’s liabili y o beha e agg essi ely33,51. Combining
bo h o hese ypes o obse e s p o ides unique in o ma ion, helping o cla i y he b ead h o his associa ion
o agg ession wi h 3-hyd oxybu y a e.
To ou knowledge, his is he i s ime ke one bodies, and 3-hyd oxybu y a e speci ically, ha e been associa ed
wi h agg ession. In a u ina y me aboli e s udy wi h agg ession in he NTR, no signi ican associa ions we e ound,
bu se e al nea -signi ican esul s poin ed o ene gy me abolism as a po en ial g oup o me aboli es wa an ing
u he in es iga ion22. Addi ionally, he e a e wo ecen s udies showing a link be ween 3-hyd oxybu y a e and
dep ession, one in mice ha was nega i ely co ela ed18 and one in humans ha was posi i ely co ela ed19. We
know om p e ious s udies ha dep ession and agg ession o en co-occu 25,52, and he concep o a ‘p ac o ’
ha po en ially unde lies all psychopa hology has been gaining ac ion53. Howe e , ou inding o a nega i e
agg ession–3-hyd oxybu y a e ela ionship is in con as wi h ea lie epo ed posi i e dep ession–3-hyd oxy-
bu y a e ela ionships and wa an s s udying hese ai s sepa a ely in ela ion o me aboli es.
In e alua ing he biological plausibili y o his associa ion, we can conside gene ally wha we know abou his
compound and how i migh ela e o agg ession. Du ing pe iods o exe cise o as ing (low a ailabili y o glucose
as an ene gy sou ce), ke one bodies become in ol ed in an al e na i e ene gy pa hway41. 3-Hyd oxybu y a e, in
pa icula , has been shown o be no only in ol ed in he al e na i e ene gy pa hway, bu i also has impo an
cell signalling and egula o y unc ions, including pa hways in ol ed in educing oxida i e s ess and in lamma-
ion (bo h linked o agg ession7,54) and epigene ic modi ica ions19,55. We also know ha while 3-hyd oxybu y a e
is gene ally p oduced in he li e , i can c oss he blood–b ain ba ie and ac in he b ain56,57, whe e we know
ha agg ession is also egula ed58,59. Finally, al hough he e appea o be complex biological mechanisms a play,
ke ogenic die s (which inc ease ke one body le els) ha e been popula ea men s in neu ologic diso de s, such
as epilepsy, o decades60. The ield o psychia y has ecen ly become mo e in e es ed in he po en ial bene i s
o his die , including in a en ion-de ici diso de s and au ism61, bo h o which can co-occu wi h agg ession,
howe e , agg ession and he ke ogenic die has ye o be conside ed. In iguingly, esea che s ha e shown ha
8
Vol:.(1234567890)
Scien i ic Repo s | (2021) 11:5813 | h ps://doi.o g/10.1038/s41598-021-84635-6
www.na u e.com/scien i ic epo s/
a ke ogenic die wi h be a-hyd oxybu y a e o male D osophila wi h agg ession a e auma ic b ain inju y
ma kedly educed he le els o agg ession62.
This s udy sugges ing a new bioma ke o agg ession has impo an s eng hs in i s la ge sample size, mul iple
a ings o agg ession ac oss s ages o de elopmen , and he di ec ion o associa ion eplica ed in an independen
sample. The e a e also limi a ions. Fi s , we did no ha e agg ession measu ed a he same ime as he biological
sample, in ei he da ase . Howe e , i is known ha agg ession is a ela i ely s able ai , especially in agg essi e
males51. Al hough agg ession le els in childhood and adolescence a e gene ally highe han in adul hood, agg es-
sion le els a e ela i ely s able in hei ank o de among simila aged pee s63,64. The e is longi udinal con inui y
in agg ession om adolescence o adul hood33,51,65, which is he de elopmen pe iod co e ed in he p esen s udy.
Addi ionally, a ai ly ecen s udy wi h simila empo al da a collec ion limi a ions, using childhood agg es-
sion measu es (as de elopmen al ajec o ies) and biological samples in young adul hood, was able o iden i y
cy okines (a measu e o in lamma ion) as a no el g oup o bioma ke s o agg ession7. Rega ding bioma ke
s abili y, while we do see high he i abili y in he ene gy me abolism bioma ke s (53.2% o 3-hyd oxybu y a e)66,
i would also be impo an in u u e s udies o ha e epea ed biological samples o con i m he s abili y o his
bioma ke o e ime, concu en wi h he s abili y o agg ession measu es. Howe e , epea ed blood samples
om child en may be challenging o ob ain, and hus his wo k migh need o be done in young adul s i s , o
using less in asi e samples such as u ine o sali a, which migh p oduce di e en esul s han plasma ( he ecen
u ine bioma ke s udy on NTR child en did no show any signi ican indings wi h 3-hyd oxybu y ic acid and
agg ession)22.
In conclusion, his explo a o y s udy using NMR-p ocessed me aboli es in plasma samples has e ealed a
po en ial new bioma ke o agg ession: 3-hyd oxybu y a e, a ke one body ela ed o an al e na i e ene gy pa h-
way. We obse ed a nega i e associa ion—high agg ession le els wi h low ke one bodies le els—in ou Finnish
da ase , which was sugges i ely eplica ed in ou independen Du ch da ase . The e is biological plausibili y o
associa ing his ke one body wi h agg ession, as well as he abili y o po en ially in luence he biological le els
o 3-hyd oxybu y a e ia die modi ica ion. Thus, i is well-wa an ed o u he in es iga e his epidemiological
associa ion a he biological le el, hope ully shedding mo e ligh on he complexi ies and he e ogenei y o he
agg ession pheno ype.
Da a a ailabili y
The da ase s gene a ed and/o analysed du ing he cu en s udy a e no publicly a ailable due o pa icipan
con iden iali y, bu a e a ailable om he co esponding au ho on easonable eques .
Recei ed: 19 May 2020; Accep ed: 9 Feb ua y 2021
Re e ences
1. Fe gusson, D. M., Ho wood, L. J. & Ridde , E. M. Show me he child a se en: The consequences o conduc p oblems in childhood
o psychosocial unc ioning in adul hood. J. Child Psychol. Psychia y 46, 837–849. h ps ://do i .o g/10.1111/j .1469-7610.2004.00387
.x (2005).
2. Fa ing on, D. P. & Loebe , R. Epidemiology o ju enile iolence. Child Adolesc. Psychia . Clin. N. Am. 9, 733–748 (2000).
3. Odin so a, V. V. e al. Genomics o human agg ession: Cu en s a e o genome-wide s udies and an au oma ed sys ema ic e iew
ool. Psychia . Gene . 29, 170–190. h ps ://doi.o g/10.1097/YPG.00000 00000 00023 9 (2019).
4. an Dongen, J. e al. Epigenome-wide associa ion s udy o agg essi e beha io . Twin Res. Hum. Gene . 18, 686–698. h ps ://doi.
o g/10.1017/ hg.2015.74 (2015).
5. Cecil, C. A. M. e al. DRD4 me hyla ion as a po en ial bioma ke o physical agg ession: An epigenome-wide, c oss- issue in es-
iga ion. Am. J. Med. Gene . B Neu opsychia . Gene . 177, 746–764. h ps ://doi.o g/10.1002/ajmg.b.32689 (2018).
6. Hagenbeek, F. e al. Disco e y o biochemical bioma ke s o agg ession: A ole o me abolomics in psychia y. Am. J. Med. Gene .
B Neu opsychia . Gene . 171, 719–732. h ps ://doi.o g/10.1002/ajmg.b.32435 (2016).
7. P o encal, N., Sude man, M. J., Vi a o, F., Szy , M. & T emblay, R. E. Childhood ch onic physical agg ession associa es wi h adul
cy okine le els in plasma. PLoS ONE 8, e69481. h ps ://doi.o g/10.1371/jou n al.pone.00694 81 (2013).
8. Kaplan, J. R. e al. Demons a ion o an associa ion among die a y choles e ol, cen al se o one gic ac i i y, and social beha io
in monkeys. Psychosom. Med. 56, 479–484. h ps ://doi.o g/10.1097/00006 842-19941 1000-00001 (1994).
9. Pa i, G. J., Yanes, O. & Siuzdak, G. Inno a ion: Me abolomics: The apogee o he omics ilogy. Na . Re . Mol. Cell Biol. 13, 263–269.
h ps ://doi.o g/10.1038/n m33 14 (2012).
10. Emwas, A. H. e al. NMR spec oscopy o me abolomics esea ch. Me aboli es h ps ://doi.o g/10.3390/me ab o9070 123 (2019).
11. Geb egiwo gis, T. & Powe s, R. Applica ion o NMR me abolomics o sea ch o human disease bioma ke s. Combin. Chem. High
Th oughpu Sc een. 15, 595–610. h ps ://doi.o g/10.2174/13862 07128 02650 522 (2012).
12. Repo-Tiihonen, E., Halonen, P., Tiihonen, J. & Vi kkunen, M. To al se um choles e ol le el, iolen c iminal o ences, suicidal
beha io , mo ali y and he appea ance o conduc diso de in Finnish male c iminal o ende s wi h an isocial pe sonali y diso de .
Eu . A ch. Psychia y Clin. Neu osci. 252, 8–11 (2002).
13. Hagenbeek, F. e al. Adul agg essi e beha io in humans and bioma ke s: A ocus on lipids and me hyla ion. J. Pedia . Neona .
Indi id. Med. 7, e070204 (2018).
14. Ramsay, H. e al. Cogni ion, psychosis isk and me abolic measu es in wo adolescen bi h coho s—CORRIGENDUM. Psychol.
Med. 48, 2628. h ps ://doi.o g/10.1017/S0033 29171 80021 92 (2018).
15. Chen, X. e al. Dys egula ion o amino acids and lipids me abolism in schizoph enia wi h iolence. BMC Psychia y 20, 97. h ps
://doi.o g/10.1186/s1288 8-020-02499 -y (2020).
16. Coppola, A. e al. B anched-chain amino acids al e neu obeha io al unc ion in a s. Am. J. Physiol. Endoc inol. Me ab. 304,
E405–E413. h ps ://doi.o g/10.1152/ajpen do.00373 .2012 (2013).
17. S -Jean, A. e al. B anched-chain and a oma ic amino acids in ela ion o beha io al p oblems among young Inui om Nuna ik,
Canada: A coho s udy. Pedia . Res. 82, 416–422. h ps ://doi.o g/10.1038/p .2017.115 (2017).
18. Chen, L., Miao, Z. & Xu, X. be a-hyd oxybu y a e alle ia es dep essi e beha io s in mice possibly by inc easing he his one3-
lysine9-be a-hyd oxybu y yla ion. Biochem. Biophys. Res. Commun. 490, 117–122. h ps ://doi.o g/10.1016/j.bb c.2017.05.184
(2017).
9
Vol.:(0123456789)
Scien i ic Repo s | (2021) 11:5813 | h ps://doi.o g/10.1038/s41598-021-84635-6
www.na u e.com/scien i ic epo s/
19. Se oyama, D. e al. Plasma me aboli es p edic se e i y o dep ession and suicidal idea ion in psychia ic pa ien s—A mul icen e
pilo analysis. PLoS ONE 11, e0165267. h ps ://doi.o g/10.1371/jou n al.pone.01652 67 (2016).
20. Yang, J. e al. Landscapes o bac e ial and me abolic signa u es and hei in e ac ion in majo dep essi e diso de s. Sci. Ad . h ps
://doi.o g/10.1126/sciad .aba85 55 (2020).
21. Gulsun, M. e al. Possible ela ionship be ween amino acids, agg ession and psychopa hy. In . J. Psychia y Clin. P ac . 20, 91–100.
h ps ://doi.o g/10.3109/13651 501.2016.11447 71 (2016).
22. Hagenbeek, F. e al. U ina y amine and o ganic acid me aboli es e alua ed as ma ke s o childhood agg ession: The ACTION
bioma ke s udy. F on . Psychia y 11, 165. h ps ://doi.o g/10.3389/ psy .2020.00165 (2020).
23. Kap io, J. The Finnish Twin Coho S udy: An upda e. Twin Res. Hum. Gene . 16, 157–162. h ps ://doi.o g/10.1017/ hg.2012.142
(2013).
24. Rose, R. J. e al. FinnTwin12 coho : An upda ed e iew. Twin Res. Hum. Gene . 22, 302–311. h ps ://doi.o g/10.1017/ hg.2019.83
(2019).
25. Ba els, M. e al. Childhood agg ession and he co-occu ence o beha iou al and emo ional p oblems: Resul s ac oss ages 3–16
yea s om mul iple a e s in six coho s in he EU-ACTION p ojec . Eu . Child Adolesc. Psychia y 27, 1105–1121. h ps ://doi.
o g/10.1007/s0078 7-018-1169-1 (2018).
26. Bogl, L. H. e al. Abdominal obesi y and ci cula ing me aboli es: A win s udy app oach. Me abolism 65, 111–121. h ps ://doi.
o g/10.1016/j.me ab ol.2015.10.027 (2016).
27. Soininen, P., Kangas, A. J., Wu z, P., Suna, T. & Ala-Ko pela, M. Quan i a i e se um nuclea magne ic esonance me abolomics in
ca dio ascula epidemiology and gene ics. Ci c. Ca dio asc. Gene . 8, 192–206. h ps ://do i .o g/10.1161/CIR CG ENETI CS.114.00021
6 (2015).
28. Holmes, M. V. e al. Lipids, lipop o eins, and me aboli es and isk o myoca dial in a c ion and s oke. J. Am. Coll. Ca diol. 71,
620–632. h ps ://doi.o g/10.1016/j.jacc.2017.12.006 (2018).
29. Ke unen, J. e al. Genome-wide s udy o ci cula ing me aboli es iden i ies 62 loci and e eals no el sys emic e ec s o LPA. Na .
Commun. 7, 11122. h ps ://doi.o g/10.1038/ncomm s1112 2 (2016).
30. Pulkkinen, L., Kap io, J. & Rose, R. J. Pee s, eache s and pa en s as assesso s o he beha iou al and emo ional p oblems o wins
and hei adjus men : The Mul idimensional Pee Nomina ion In en o y. Twin Res. 2, 274–285. h ps ://doi.o g/10.1375/13690
52993 20565 762 (1999).
31. Achenbach, T. M., McConaughy, S. H. & Howell, C. T. Child/adolescen beha io al and emo ional p oblems: Implica ions o
c oss-in o man co ela ions o si ua ional speci ici y. Psychol. Bull. 101, 213–232. h ps ://doi.o g/10.1037/0033-2909.101.2.213
(1987).
32. De Los Reyes, A. & Kazdin, A. E. In o man disc epancies in he assessmen o childhood psychopa hology: A c i ical e iew,
heo e ical amewo k, and ecommenda ions o u he s udy. Psychol. Bull. 131, 483–509. h ps ://doi.o g/10.1037/0033-
2909.131.4.483 (2005).
33. 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. h ps ://doi.o g/10.1007/s0078 7-018-1198-9 (2019).
34. Aal onen, S. e al. Mo o de elopmen and physical ac i i y: A longi udinal disco dan win-pai s udy. Med. Sci. Spo s Exe c. 47,
2111–2118. h ps ://doi.o g/10.1249/MSS.00000 00000 00065 0 (2015).
35. Kujala, U. M. e al. Associa ions o ae obic i ness and maximal muscula s eng h wi h me aboli es in young men. JAMA Ne w.
Open 2, e198265. h ps ://doi.o g/10.1001/jaman e wo kopen .2019.8265 (2019).
36. Wu z, P. e al. Me abolic p o iling o alcohol consump ion in 9778 young adul s. In . J. Epidemiol. 45, 1493–1506. h ps ://doi.
o g/10.1093/ije/dyw17 5 (2016).
37. Xu, T. e al. E ec s o smoking and smoking cessa ion on human se um me aboli e p o ile: Resul s om he KORA coho s udy.
BMC Med. 11, 60. h ps ://doi.o g/10.1186/1741-7015-11-60 (2013).
38. De ks, I. P. M. e al. Tes ing bidi ec ional associa ions be ween childhood agg ession and BMI: Resul s om h ee coho s. Obesi y
(Sil e Sp ing) 27, 822–829. h ps ://doi.o g/10.1002/oby.22419 (2019).
39. Pels, F. & Kleine , J. Does exe cise educe agg essi e eelings? An expe imen examining he in luence o mo emen ype and
social ask condi ions on es iness and ange educ ion. Pe cep . Mo . Skills 122, 971–987. h ps ://doi.o g/10.1177/00315 12516
64780 2 (2016).
40. G an , J. D. e al. The ole o conduc diso de in he ela ionship be ween alcohol, nico ine and cannabis use diso de s. Psychol.
Med. 45, 3505–3515. h ps ://doi.o g/10.1017/S0033 29171 50015 18 (2015).
41. La el, L. Ke one bodies: A e iew o physiology, pa hophysiology and applica ion o moni o ing o diabe es. Diabe es Me ab. Res.
Re . 15, 412–426 (1999).
42. Tu gay, A. Agg ession and dis up i e beha io diso de s in child en and adolescen s. Expe Re . Neu o he . 4, 623–632. h ps ://
doi.o g/10.1586/14737 175.4.4.623 (2004).
43. Lig ha , L. e al. The Ne he lands Twin Regis e : Longi udinal esea ch based on win and win- amily designs. Twin Res. Hum.
Gene . 22, 623–636. h ps ://doi.o g/10.1017/ hg.2019.93 (2019).
44. Achenbach, T. M., I ano a, M. Y. & Resco la, L. A. Empi ically based assessmen and axonomy o psychopa hology o ages 1(1/2)-
90+ yea s: De elopmen al, mul i-in o man , and mul icul u al indings. Comp . Psychia y 79, 4–18. h ps ://doi.o g/10.1016/j.
compp sych.2017.03.006 (2017).
45. Willemsen, G. e al. The Ne he lands Twin Regis e biobank: A esou ce o gene ic epidemiological s udies. Twin Res. Hum. Gene .
13, 231–245. h ps ://doi.o g/10.1375/ win.13.3.231 (2010).
46. Si o a, M. e al. E ec o genome and en i onmen on me abolic and in lamma o y p o iles. PLoS ONE 10, e0120898. h ps ://doi.
o g/10.1371/jou n al.pone.01208 98 (2015).
47. Au o, K. e al. A me abolic iew on menopause and ageing. Na . Commun. 5, 4708. h ps ://doi.o g/10.1038/ncomm s5708 (2014).
48. Williams, R. L. A no e on obus a iance es ima ion o clus e -co ela ed da a. Biome ics 56, 645–646 (2000).
49. Heida i, S., Babo , T. F., De Cas o, P., To , S. & Cu no, M. Sex and Gende Equi y in Resea ch: Ra ionale o he SAGER guidelines
and ecommended use. Res. In eg . Pee Re . 1, 2. h ps ://doi.o g/10.1186/s4107 3-016-0007-6 (2016).
50. Dunson, D. B. Commen a y: P ac ical ad an ages o Bayesian analysis o epidemiologic da a. Am. J. Epidemiol. 153, 1222–1226.
h ps ://doi.o g/10.1093/aje/153.12.1222 (2001).
51. Pulkkinen, L. In The Camb idge Handbook o Violen Beha io and Agg ession 2nd edn (eds Flanne y, D. J. e al.) 31–51 (Camb idge
Uni e si y P ess, Camb idge, 2018).
52. Achenbach, T. M., I ano a, M. Y., Resco la, L. A., Tu ne , L. V. & Al ho , R. R. In e nalizing/ex e nalizing p oblems: Re iew
and ecommenda ions o clinical and esea ch applica ions. J. Am. Acad. Child Adolesc. Psychia y 55, 647–656. h ps ://doi.
o g/10.1016/j.jaac.2016.05.012 (2016).
53. Caspi, A. e al. The p ac o : One gene al psychopa hology ac o in he s uc u e o psychia ic diso de s?. Clin. Psychol. Sci. 2,
119–137. h ps ://doi.o g/10.1177/21677 02613 49747 3 (2014).
54. Ca oll, J. E. e al. A u ina y ma ke o oxida i e s ess co a ies posi i ely wi h hos ili y among midli e communi y olun ee s.
Psychosom. Med. 72, 273–280. h ps ://doi.o g/10.1097/PSY.0b013 e3181 d0d72 b (2010).
55. Newman, J. C. & Ve din, E. be a-hyd oxybu y a e: Much mo e han a me aboli e. Diabe es Res. Clin. P ac . 106, 173–181. h ps ://
doi.o g/10.1016/j.diab es.2014.08.009 (2014).