Uncovering the complex genetics of human temperament

Molecula Psychia y
h ps://doi.o g/10.1038/s41380-018-0264-5
ARTICLE
Unco e ing he complex gene ics o human empe amen
Igo Zwi 1,2 ●Ja ie A nedo 1,2 ●Co al Del-Val2●Lau a Pulkki-Råback3●Be ina Kon e 4●Sa ah S. Yang5●
Rocio Rome o-Zaliz2●Mi ka Hin sanen6●Ke in M. Cloninge 7●Danilo Ga cia 8,9 ●D agan M. S akic1●
Sando Rozsa1●Ma ibel Ma inez1●Leo-Pekka Lyy ikäinen 10 ●Ina Giegling4,11 ●Mika Kähönen12 ●
Helena He nandez-Cue o13 ●Ilkka Seppälä 10 ●Emma Rai oha ju10 ●Gab iel A. de E ausquin14 ●Olli Rai aka i15 ●
Dan Rujescu4●Teodo T. Pos olache16,17 ●Joohon Sung5●Liisa Kel ikangas-Jä inen3●Te ho Leh imäki10 ●
C. Robe Cloninge 1,18
Recei ed: 12 Feb ua y 2018 / Re ised: 21 July 2018 / Accep ed: 15 Augus 2018
© The Au ho (s) 2018. This a icle is published wi h open access
Abs ac
Expe imen al s udies o lea ning sugges ha human empe amen may depend on he molecula mechanisms o associa i e
condi ioning, which a e highly conse ed in animals. The main gene ic pa hways o associa i e condi ioning a e known in
expe imen al animals, bu ha e no been iden ified in p io genome-wide associa ion s udies (GWAS) o human
empe amen . We used a da a-d i en machine lea ning me hod o GWAS o unco e he complex geno ypic–pheno ypic
ne wo ks and en i onmen al in e ac ions ela ed o human empe amen . In a disco e y sample o 2149 heal hy Finns, we
iden ified se s o single-nucleo ide polymo phisms (SNPs) ha clus e wi hin pa icula indi iduals (i.e., SNP se s) ega dless
o pheno ype. Second, we iden ified 3 clus e s o people wi h dis inc empe amen p ofiles measu ed by he Tempe amen
and Cha ac e In en o y ega dless o geno ype. Thi d, we ound 51 SNP se s ha iden ified 736 gene loci and we e
significan ly associa ed wi h empe amen . The iden ified genes we e en iched in pa hways ac i a ed by associa i e
condi ioning in animals, including he ERK, PI3K, and PKC pa hways. 74% o he iden ified genes we e unique o a specific
empe amen p ofile. En i onmen al influences measu ed in childhood and adul hood had small bu significan e ec s. We
confi med he eplicabili y o he 51 Finnish SNP se s in heal hy Ko ean (90%) and Ge man samples (89%), as well as hei
associa ions wi h empe amen . The iden ified SNPs explained nea ly all he he i abili y expec ed in each sample (37–53%)
despi e a iable cul u es and en i onmen s. We conclude ha human empe amen is s ongly influenced by mo e han 700
genes ha modula e associa i e condi ioning by molecula p ocesses o synap ic plas ici y and long- e m memo y.
In oduc ion
Tempe amen is classically defined as hose aspec s o
pe sonali y ha exp ess basic emo ions like ea , ange , and
disgus , and ha a e de elopmen ally s able and he i able,
a he han lea ned [1]. Howe e , his classical defini ion is
inadequa e because human beings ha e h ee majo sys ems
o lea ning and memo y wi h dis inc i e gene ic and bio-
logical bases ha e ol ed in succession o e he long
phylogene ic lineage leading om p imi i e animals o
mode n human beings [2–4]. P ocedu al lea ning o habi s
is p esen in all animals h ough highly conse ed molecula
mechanisms o associa i e condi ioning, including classical
and ope an condi ioning [5–9]. In con as , e idence o
in en ional cogni i e p ocesses, such as pu pose ul goal-
seeking, social econcilia ion, and abs ac symboliza ion o
ac s, a e p esen in he p ima e lineage o human beings,
bu no in ep iles [2–4,10]. E idence o au onoe ic
o au obiog aphical lea ning appea s o be p esen only
wi h he ad en o a and science in mode n Homo sapiens
[2,11–15].
Ea ly esea ch assessing empe amen ocused on
de elopmen ally s able ea u es o ac i i y and a ec , bu
some ecen wo k has ex ended assessmen s o empe amen
o include aspec s o a en ion and sel - egula o y p ocesses
ha eme ged la e in e olu ion and ha de elop in esponse
o bo h indi idual expe ience and social no ms [1,2,16]. In
con as , Cloninge ook an e olu iona y pe spec i e o
*C. Robe Cloninge
[email p o ec ed]
Ex ended au ho in o ma ion a ailable on he las page o he a icle.
Elec onic supplemen a y ma e ial The online e sion o his a icle
(h ps://doi.o g/10.1038/s41380-018-0264-5) con ains supplemen a y
ma e ial, which is a ailable o au ho ized use s.
1234567890();,:
1234567890();,:
lea ning in de eloping he Tempe amen and Cha ac e
In en o y (TCI), defining empe amen as ha aspec o
pe sonali y based on associa i e condi ioning [17–19]. The
TCI measu es ou empe amen dimensions ha ha e been
empi ically confi med by unc ional b ain imaging o
quan i y indi idual di e ences in associa i e condi ioning
and ela ed human b ain ci cui y: Ha m A oidance (i.e.,
ea ul, pessimis ic s. isk- aking, op imis ic) [20–22],
No el y Seeking (i.e., impulsi e, exci able s. delibe a e,
ese ed) [23,24], Rewa d Dependence (i.e., iendly,
sen imen al s. de ached, objec i e) [21,24], and Pe sis-
ence (i.e., de e mined, ambi ious s. easily discou aged,
unde achie ing) [25,26]. Ha m A oidance is an indica o
o nega i e alence ha measu es passi e a oidance
lea ning and inc eased sensi i i y o ea ul s imuli media ed
by ac i a ion o he amygdala, subgenual cingula e co ex,
and he insula salience ne wo k [22,27,28]. No el y
Seeking is an indica o o posi i e alence ha measu es
app oach o no el s imuli [29,30], e en i hey do no
p edic ewa ds [24], whe eas Rewa d Dependence is p e-
dic i e o social a filia ion and app oach o ewa ds based
on a di e en pa e n o ac i a ion o dopamine gic neu ons
in he nucleus accumbens and subs an ia nig a [24] and on
oxy ocine gic neu ons in he hypo halamus [31]. Pe sis-
ence quan ifies di e ences in a es o ex inc ion o in e -
mi en ly ewa ded beha io s in esponse o us a i e non-
ewa d by ac i a ion o a ci cui connec ing he nucleus
accumbens, an e io cingula e, and en ola e al on al
co ex [25,26].
S udies o gene exp ession in esponse o associa i e
condi ioning in expe imen al animals ha e consis en ly
documen ed he ac i a ion o specific molecula pa hways
ha igge synap ic plas ici y, which is a undamen al basis
o long- e m memo y [7,32–34]. The Ras-MEK-ERK
cascade (also known as he Mi ogen-ac i a ed P o ein
Kinase (MAPK) pa hway) and he PI3K-AKT-mTOR cas-
cade a e majo cellula mechanisms o esponding o
ex acellula s imuli, and hei ac i a ion igge s in acel-
lula p ocesses ha p omo e synap ic plas ici y and asso-
cia i e condi ioning, including long- e m po en ia ion (LTP)
and long- e m dep ession (LDP) [7,32,33,35]. The cell-
su ace ecep o s o hese pa hways can be ac i a ed by a
wide a ie y o soma ic, psychological, and social s esso s
ha a y in posi i e and nega i e alence and in con-
sequences o su i al and ep oduc ion [6,33,36]. Chan-
ges in hese pa hways in esponse o associa i e
condi ioning occu in a coo dina ed manne wi h ela ed
p ocesses including s ess eac i i y [37], neu onal and glial
g ow h [38], and neu o ansmission [39]. The e o e, we
hypo hesized ha genes in he same molecula pa hways
iden ified in non-human animals o associa i e condi ion-
ing and ela ed p ocesses would be associa ed wi h human
empe amen p ofiles. This hypo hesis was al eady
suppo ed indi ec ly by ou finding ha genes in hese
pa hways we e associa ed wi h he dependen and apa he ic
cha ac e p ofiles in which sel - egula o y pe sonali y ai s
we e inadequa e o egula e empe amen in a heal hy
manne , esul ing in s ess eac i i y and ill-heal h [40].
Un o una ely p io genome-wide associa ion s udies
(GWAS) o empe amen ha conside ed only he a e age
e ec s o genes ha e iden ified ew genes associa ed wi h
pe sonali y and ha e specifically ailed o unco e he genes
associa ed wi h long- e m memo y whe he he TCI o o he
pe sonali y in en o ies we e used [41,42]. Such ailu e is
an example o he “missing”[43]o “hidden”[44] he i -
abili y p oblem in s udies o complex pheno ypes. Tem-
pe amen as measu ed by he TCI and o he in en o ies is
known o be s ongly influenced by gene–gene [45–48] and
gene–en i onmen in e ac ions [49–51]. Such complexi y is
expec ed om he ex ensi e eedback in e ac ions among
he molecula pa hways ha a e ac i a ed in non-human
animals in esponse o associa i e condi ioning [52].
As in ou accompanying GWAS o human cha ac e
[40], we ha e chosen o use s ic ly da a-d i en me hods o
deep clus e analysis in GWAS o unco e he complex
geno ypic and pheno ypic a chi ec u e o empe amen [53–
55]. We pos ula e ha he genes in molecula pa hways
ela ed o empe amen a e no missing bu a e dis ibu ed in
di e en ne wo ks o in e ac ing genes and en i onmen s
ha influence di e en people [54–57]. Mo e specifically,
we hypo hesize ha he genes associa ed wi h empe amen
will be en iched in he molecula pa hways expe imen ally
ac i a ed by associa i e condi ioning in non-human
animals.
Subjec s and me hods
Subjec s and me hods we e he same as de ailed in an
accompanying pape [40], so essen ials a e b iefly sum-
ma ized he e.
Desc ip ion o he samples
Ou disco e y sample was he Young Finns S udy, an
epidemiological s udy o 2149 heal hy Finnish child en
ollowed egula ly om 1980 (ages 3–18 yea s) o 2012
(ages 35–50 yea s) [58]. All Finnish subjec s (56% women)
had ho ough s anda dized geno ypic, en i onmen al, and
pheno ypic assessmen s, including adminis a ion o he
TCI [16,58].
We eplica ed he esul s in wo independen samples o
heal hy adul s om Ge many [59,60] and Ko ea [61,62]in
which compa able geno ypic and pheno ypic ea u es we e
a ailable (see Supplemen ). The Ko ean s udy in ol ed
1052 un ela ed indi iduals ex ac ed om a na ional egis e
I. Zwi e al.
(aged 28–81 yea s, 57% women). The Ge man s udy
in ol ed 902 subjec s (aged 20–74 yea s, 49% women)
andomly selec ed om he Munich ci y egis e and
sc eened o exclude anyone wi h a his o y o psychia ic
illness in hemsel es o hei fi s -deg ee ela i es.
Pe sonali y assessmen
All subjec s comple ed he TCI o assess se en he i able
dimensions o pe sonali y [18,63]. The TCI measu es ou
well- alida ed dimensions o empe amen (No el y Seek-
ing, Ha m A oidance, Rewa d Dependence, Pe sis ence)
and h ee dimensions o cha ac e , as desc ibed in he
“In oduc ion”and in mo e de ail in Supplemen a y Sec-
ion 1 and Table S1 [18,63]. The 12 empe amen subscales
om he TCI we e used as he p ima y pheno ypic da a in
all h ee samples (Supplemen a y Sec ion 2 and Table S1).
Pe sonali y heal h indices
People a isk o unheal hy pe sonali y we e iden ified as he
bo om decile o he sum o TCI Sel -di ec edness and
Coope a i eness [64], a p e iously alida ed indica o o ill-
being [65,66]. In con as , people wi h heal hy pe sonali ies
we e iden ified as he op decile o he p oduc o all h ee
TCI cha ac e ai s, a p e iously alida ed indica o o
well-being [64,67,68]. Ou ill-being and well-being indices
we e used o measu e he heal h s a us o subjec s con-
sis en ly in all h ee samples.
We also iden ified an empi ical index o empe amen
(Supplemen a y Sec ion 3 and Table S2) as a single com-
p ehensi e measu e o empe amen ha could be used in
SNP-se Ke nel Associa ion Tes (SKAT) [56,57] and
he i abili y analyses.
Geno yping
The Finnish sample was geno yped by using Illumina
Human670-Quad Cus om, (i.e., Illumina 670k cus om)
a ays [69]. The Ko ean sample used A yme ix Genome-
Wide Human SNP A ay 6.0 and Illumina HumanCo e [61].
The Ge man sample used A yme ix Genome-Wide
Human SNP A ay 6.0, Illumina OMNI Exp ess and he
300 A ay, p e-phased and impu ed wi h SHAPEIT2 and
IMPUTE2. Some Ge man indi iduals had also been geno-
yped on Illumina Omni1-Quad. Quali y con ol was pe -
o med o all samples as in p io wo k [55] (Supplemen a y
Sec ion 3).
A e quali y con ol, he PLINK so wa e sui e [70] was
used o educe he la ge sea ch space by p e-selec ing a
subse o SNPs using a gene ously inclusi e h eshold (p-
alue < 0.01 wi hou Bon e oni co ec ion) o possible
associa ion wi h empe amen , aking gende and e hnici y
in o accoun as co a ia es o he indi idual SNPs, as
de ailed in an accompanying pape [40]. We accoun ed o
e hnici y in each sample by using he fi s h ee p incipal
componen s o ances al s a ifica ion o SNP geno ypes
(Supplemen a y Sec ion 3) [71].
Compu a ional p ocedu es
The clus e analyses used he Gene alized Fac o iza ion
Me hod [72–75] including Non-nega i e Ma ix Fac o iza-
ion (NMF), which op imizes pa e n ecogni ion and na u-
ally occu ing associa ions be ween pa e ns ac oss di e en
ypes o da a. The clus e ing was en i ely da a-d i en wi h-
ou es ic i e assump ions abou he numbe o con en o
he clus e s [54], as de ailed elsewhe e [53–55,72,76]. The
s eps o his analy ic p ocedu e a e summa ized and sche-
ma ically ela ed o unsupe ised Deep NMF Lea ning in
Supplemen a y Figu e S1. The ad an ages o his clus e ing
app oach o e al e na i e analyses o single o mul iple
ma ke s a e desc ibed in Supplemen a y Sec ion 4.
Ou web se e applica ion o Pheno ype–Geno ype
Many- o-many Rela ions Analysis (PGMRA) in GWAS is
published [54] and a ailable online a h p://phop.ug .es/
enogeno. The PGMRA me hod and algo i hm a e also
summa ized in Supplemen a y Sec ions 5 and 6, which
include a semi-supe ised classifie o pheno ypes om
geno ypes. PGMRA p ope ly accoun s o Linkage Dis-
equilib ium (LD) e ficien ly (i.e., wi hou loss o in o ma ion
abou complex geno ypic–pheno ypic ela ions) (Supple-
men a y Sec ion 4). S a is ical analysis co ec ing o mul iple
compa isons, as well as gende and e hnici y as co a ia es o
he SNP se s, was pe o med by SKAT [56,57], also
accessible ia PGMRA. He i abili y was es ima ed om a
immed eg ession o SNPs on he empi ical index o em-
pe amen con olling o ou lie s and en i onmen al a iables
[77,78] (see also Supplemen a y Sec ion 7).
Replicabili y o esul s was e alua ed in he h ee inde-
penden samples o SNP se s, pheno ypic se s, and geno-
ypic–pheno ypic ela ions using mul i-objec i e
op imiza ion echniques [55], as de ailed in Supplemen a y
Sec ion 8. The PGMRA classifie was used o p edic
empe amen pheno ypes om he geno ypic se s (Supple-
men a y Sec ion 9). Fu he de ails a e a ailable in Sup-
plemen a y In o ma ion and elsewhe e [72–75].
Resul s
Iden i ying SNP se s as candida es o causal
a iabili y
902 Non-iden ical bu possibly o e lapping SNP se s we e
exhaus i ely iden ified by PGMRA in he Finnish sample
Unco e ing he complex gene ics o human empe amen
Table 1 Desc ip ion o 51 SNP se s associa ed wi h Tempe amen se s (p<1E−05)
Finnish sample P obabili y o heal h # Gs
SNP se s SNP se names Cha ac e % Coding SKAT p- alue A e age SNPs Bes SNP Wo s SNP # Subjec s # SNPs Well- being Ill-being
G_13_3 ERK-condi ioned impulsi i y 71 7.38E−14 1.50E−01 1.46E−04 1.00E+00 95 158 0.03 0.32 21
G_8_8 Global inosi ol/chemokine pa hways O ganized 60 1.06E−14 4.39E−01 5.04E−05 1.00E+00 224 611 0.08 0.07 286
G_13_10 Choline gic neu omodula ion 71 3.13E−08 9.38E−02 8.81E−06 5.99E−01 148 57 0.09 0.08 17
G_21_18 Cogni i e flexibili y 73 2.75E−05 3.53E−01 1.10E−03 1.00E+00 116 47 0.13 0.09 15
G_30_10 TNF-based esilience 50 1.21E−05 2.33E−01 1.74E−04 1.00E+00 47 45 0.09 0.06 6
G_7_3 Neu ogenesis O ganized 66 3.32E−06 4.07E−01 4.02E−05 1.00E+00 133 364 0.17 0.36 128
G_38_23 Senso y sensi i i y 63 7.60E−07 1.23E−02 3.90E−04 1.12E−01 39 37 0.05 0.49 16
G_25_3 Ace ylcholine biosyn hesis 50 2.06E−06 3.62E−03 1.66E−04 1.54E−02 16 31 0.13 0.50 2
G_31_8 Neu o ophin O ganized 57 4.34E−16 2.55E−01 4.02E−05 1.00E+00 54 183 0.09 0.54 60
G_28_15 Es ogen neu oplas ici y Dependen 52 3.10E−06 3.66E−01 2.05E−04 1.00E+00 101 123 0.08 0.38 29
G_11_7 HPA s ess eac i i y 64 9.17E−10 1.49E−01 1.66E−04 1.00E+00 26 92 0.08 0.38 11
G_26_14 Glucose anspo Apa he ic 60 1.12E−07 2.78E−01 1.31E−04 1.00E+00 46 75 0.09 0.24 25
G_41_33 GPCR neu oplas ici y Dependen 47 1.47E−06 2.70E−01 7.41E−04 1.00E+00 56 76 0.11 0.21 15
G_21_16 Ace ylcholine biosyn hesis 100 8.24E−06 2.41E−03 1.66E−04 1.54E−02 37 26 0.14 0.22 1
G_38_38 Ion pe meabili y 67 2.67E−12 1.60E−01 1.08E−04 1.00E+00 38 79 0.00 0.16 18
G_7_2 GPCR dys egula ion Dependen 61 2.12E−18 3.39E−01 2.95E−04 1.00E+00 211 303 0.09 0.23 147
G_12_11 Ras-Ak in e ac ion 75 8.00E−08 5.89E−02 2.77E−04 5.95E−01 105 44 0.03 0.21 4
G_16_1 PI3K-based memo y 64 1.72E−05 2.61E−01 3.35E−04 1.00E+00 108 53 0.16 0.27 11
G_37_14 Neu oexci abili y 58 1.25E−05 2.50E−02 1.74E−04 8.12E−01 21 42 0.05 0.14 12
G_13_12 Ace ylcholine biosyn hesis 100 6.26E−06 8.60E−02 1.66E−04 1.00E+00 78 47 0.23 0.31 1
G_12_1 Episodic lea ning C ea i e 61 2.92E−13 2.44E−01 1.08E−04 1.00E+00 146 189 0.20 0.06 64
G_28_10 WD/CDK neu oplas ici y 50 7.40E−06 3.11E−02 3.22E−04 2.76E−01 46 30 0.15 0.15 8
G_5_3 Regula ion pa hways 0 1.40E−05 4.24E−03 1.74E−04 3.74E−02 172 50 0.28 0.05 2
G_38_13 Glucu onidase habi ex inc ion 57 6.25E−06 6.73E−02 1.74E−04 7.31E−01 60 63 0.13 0.20 7
G_35_22 PI3K-based memo y 80 2.90E−07 1.04E−01 1.31E−04 5.71E−01 43 36 0.07 0.21 5
G_33_4 ERK-PKA in e ac ion 50 4.53E−05 1.80E−01 3.35E−04 1.00E+00 24 51 0.00 0.33 6
G_12_8 Neu op o ec ion Resou ce ul 63 3.30E−22 2.68E−01 6.89E−05 1.00E+00 173 285 0.09 0.03 111
G_39_21 RGS nega i e emo ionali y 60 8.48E−06 1.03E−01 3.22E−04 7.53E−01 56 37 0.11 0.07 5
G_7_7 Ol ac ion Dependen 52 9.84E−08 3.41E−01 1.66E−04 1.00E+00 145 193 0.03 0.10 58
G_21_3 Cellula senescence Apa he ic 64 8.23E−07 3.00E−01 1.52E−04 1.00E+00 60 117 0.10 0.23 39
G_39_26 mTOR myelina ion 62 1.08E−09 2.89E−01 5.50E−05 1.00E+00 20 118 0.20 0.30 26
G_42_39 App oach-a oidance conflic 45 2.42E−06 2.28E−01 1.31E−04 1.00E+00 19 52 0.16 0.11 11
G_35_7 PI3K-based memo y 67 2.64E−05 2.15E−01 6.19E−04 9.15E−01 32 35 0.09 0.13 12
I. Zwi e al.
Table 1 (con inued)
Finnish sample P obabili y o heal h # Gs
SNP se s SNP se names Cha ac e % Coding SKAT p- alue A e age SNPs Bes SNP Wo s SNP # Subjec s # SNPs Well- being Ill-being
G_19_3 Glucu onidase habi ex inc ion 80 1.05E−05 6.97E−02 3.35E−04 1.00E+00 48 61 0.08 0.17 5
G_22_6 Blood-b ain ba ie Dependen 60 1.09E−07 2.67E−01 1.66E−04 1.00E+00 37 93 0.08 0.16 30
G_20_2 Enhanced memo y C ea i e 78 2.39E−07 3.15E−01 1.08E−04 1.00E+00 25 80 0.24 0.12 18
G_21_17 TGFβ esis ance o aging 65 2.38E−05 3.38E−01 4.23E−04 1.00E+00 67 105 0.18 0.09 26
G_36_18 B ain-RNA-biogenesis 0 1.50E−06 2.06E−03 2.05E−04 8.20E−03 19 25 0.05 0.26 4
G_36_29 Elec on anspo O ganized 57 4.26E−08 3.57E−01 6.25E−05 1.00E+00 25 185 0.08 0.48 49
G_14_12 Ras-based s ess memo y 55 1.84E−07 2.58E−01 8.81E−06 1.00E+00 83 74 0.12 0.07 22
G_25_20 Fa y acid oxida ion 67 9.84E−07 1.03E−01 1.66E−04 7.16E−01 33 62 0.03 0.12 3
G_33_33 TGFβmemo y enhancemen 62 1.36E−07 1.65E−01 1.31E−04 1.00E+00 49 47 0.10 0.08 13
G_9_2 Se o onin–cy okine in e ac ion 73 1.33E−05 3.16E−02 3.35E−04 1.00E+00 140 56 0.04 0.24 11
G_30_9 E k-IP3-PKC In e ac ion 75 2.80E−16 2.00E−01 1.80E−05 1.00E+00 69 138 0.17 0.14 52*
G_38_17 MAPK memo y enhancemen 46 8.72E−06 2.20E−01 5.60E−04 1.00E+00 14 42 0.00 0.43 13*
G_40_5 Mannosidase habi ex inc ion 67 3.40E−05 5.49E−02 3.22E−04 2.76E−01 16 30 0.06 0.38 3*
G_30_28 Hippocampal synap ic plas ici y 30 2.64E−06 2.76E−01 1.31E−04 1.00E+00 34 53 0.06 0.12 10*
G_16_5 E k-IP3-PKC in e ac ion-based s ess memo y 71 6.39E−10 3.87E−01 4.62E−04 1.00E+00 87 324 0.09 0.15 1*
G_16_15 IL-2 neu oimmune esponse 43 3.E−04 9.20E−01 4.94E−01 1.00E+00 94 14 0.16 0.15 7**
G_37_6 Me hyla ion-based gene silencing 41 5.00E−08 1.13E−01 7.93E−04 1.61E−01 26 34 0.35 0.25 23#
G_41_37 PI3K-MAPK cogni i e unc ion 51 2.08E−05 4.95E−02 8.22E−04 2.96E−01 41 38 0.10 0.25 11#
The SNP se s a e named based on molecula pa hways and neu onal unc ions o he genes ha dis inguish he se s om one ano he (see Supplemen a y Table S4). % coding indica es he
pe cen age o p o ein-coding genes. S eng hs o associa ion a e compa ed o he SNP se , he bes SNP, and a e age SNP based on SKAT p- alues. The numbe o subjec s and SNPs comp ising
each SNP se is specified. The p obabili ies o he well-being and ill-being a e gi en o subjec s in each SNP se (see also Supplemen a y Table S2). Cha ac e indica es he associa ion o he se
wi h he Cha ac e pheno ype (published elsewhe e [40]). # Gs indica es he numbe o genes mapped by he SNP se s (Figu e S6), whe e genes can be mapped by mo e han one SNP se
*indica es SNP-se s di ec ly associa ed only wi h empe amen se s
Unco e ing he complex gene ics o human empe amen

wi hou knowledge o he pheno ype, as in ou analysis o
cha ac e [40]. Among hese, he SNP se s ela ed o
empe amen had di e en numbe s o subjec s and/o
SNPs and associa ed heal h isks (Table 1, Supplemen a y
Table S2). The SNPs mapped o di e se classes o
gene ic a ian s dispe sed ac oss all he ch omosomes
(Figs. 1aand2a; Supplemen a y Figu e S2, Supplemen-
a y Table S3).
I. Zwi e al.
Iden i ying clus e s o subjec s wi h dis inc
empe amen p ofiles
118 Tempe amen se s we e exhaus i ely iden ified by
PGMRA in he Finnish sample using he 12 empe amen
subscales wi hou knowledge o he geno ype. These fine-
g ained se s we e iden ified in clus e ing solu ions wi h he
possible numbe o se s anging om 2 o 15. Hie a chically
clus e ing hese 118 fine-g ained se s wi h PGMRA, we
iden ified 3 empe amen supe -se s ha minimized he
Cophene ic Co ela ion Coe ficien (Table 2). In o he
wo ds, 3 g oups o people had highly dis inc empe amen
p ofiles.
The h ee empe amen p ofiles we e named Reliable,
An isocial, and Sensi i e based on adi ional labels o
hei p ominen ea u es [17]. People in he Reliable
p ofile we e high in Rewa d Dependence (i.e., sen i-
men al, iendly, app o al-seeking), high in Pe sis ence
(i.e., de e mined), low in No el y Seeking (i.e., delibe a e,
h i y, o de ly), and low in Ha m A oidance (i.e., op i-
mis ic, confiden , ou going, igo ous). This p ofile e-
quen ly is associa ed wi h heal hy and us wo hy
beha io (Table 2). In con as , people in he An isocial
p ofile we e low in Rewa d Dependence (i.e., cold,
de ached, independen ), low in Pe sis ence (i.e., easily
discou aged), and high in No el y Seeking (i.e., ex a-
agan , ule-b eaking, bu no inquisi i e), which is e-
quen ly associa ed wi h unheal hy an isocial conduc
(Table 2). People wi h he Sensi i e p ofile we e high in
Ha m A oidance (i.e., pessimis ic, ea ul, shy, and
a igable), high in No el y Seeking (i.e., impulsi e,
ex a agan ), and high in Rewa d Dependence (i.e., sen-
imen al, iendly), which is equen ly associa ed wi h
app oach-a oidance conflic s and emo ional sensi i i y
(Table 2).
P edic ion o empe amen p ofiles by SNP se s
We compu ed he associa ion o SNP se s wi h empe a-
men in Finnish subjec s. SKAT showed ha he asso-
cia ion o he empi ical index o empe amen wi h
pa icula SNP se s was s onge han wi h he a e age
e ec s o hei cons i uen SNPs (Table 1). We ound 51
SNP se s had significan associa ions wi h empe amen
(p<4E−04). SNP se s we e labeled by a geno ypic
iden ifica ion “G”, ollowed by 2 numbe s indica ing he
maximum numbe o clus e s and he o de o hei
selec ion by he algo i hm. Fo example, he SNP se
G_13_3 has a p- alue o 7.38E−14, whe eas he bes and
a e age SNPs wi hin his se ha e 1.46E−04 and 1.50E
−01 p- alues, espec i ely (Table 1). SKAT [56]and
PLINK [70] me hods es ima ed simila p- alues o he
indi idual SNPs (R2=0.95, Fs a is ics, p<1E−41),
which showed ha SKAT did no infla e esul s.
The 51 SNP se s associa ed wi h empe amen a e
desc ibed in Table 1. We assigned names o he SNP se s
basedonp ominen molecula p ocessesandpa hways
ha dis inguished hem (Supplemen a y Table S4). The
empe amen - ela ed SNP se s we e comp ised o ne -
wo ks o SNPs ha mapped o 736 genes, nea ly all o
whicha eknown oinfluence indi idual di e ences in
b ain unc ions. In pa icula , hese SNP se s we e
in ol ed in he egula ion o synap ic plas ici y, long-
e m memo y based on associa i e condi ioning (long-
e m po en ia ion and dep ession, ea condi ioning,
ewa d ein o cemen , habi ex inc ion), and ela ed
Fig. 1 aTwo examples o SNP se s a e ep esen ed as Hea Map
subma ices o biclus e s. SNP se s we e iden ified by dis inc pa e ns
o molecula ea u es o SNPs in subg oups o subjec s. Allele alues
a e indica ed as BB (da k blue), AB (in e media e blue), AA (ligh
blue), and missing (black). SNP se s we e labeled o specifici y by a
pai o numbe s ep esen ing he maximum numbe o clus e s om
which he biclus e was selec ed (e.g., 16 clus e s may p oduce mo e
specific han 5) and he o de in which hey we e selec ed by he
me hod (e.g., 3 d biclus e o ac o selec ed by FNMF when he
maximum numbe o clus e s was 5) and usually ha e a p efix G o
geno ype o P o pheno ype. Only a subse o op imal and cohesi e
se s a e selec ed ac oss all numbe o clus e s (see Supplemen a y
Me hods). The SNPs wi hin each SNP se can map o di e en ch o-
mosomes (e.g., 6 and 20) and exhibi dis inc molecula consequences
(see Supplemen a y Table S3). The pie cha shows he pe cen age o
SNPs wi hin a SNP se ha belong o each ype o consequence.
bDissec ion o a GWAS in a Finnish popula ion o iden i y he
geno ypic and pheno ypic a chi ec u e o pe sonali y measu ed by he
TCI. The geno ypic ne wo k is depic ed as nodes (SNP se s) linked by
sha ed SNPs (blue lines) and/o subjec s ( ed lines). Each SNP se
maps o one o mo e genes (see Supplemen a y Table S6 o a ull lis
o genes associa ed wi h each SNP se ). SNP se s associa ed wi h each
o he h ee gene al empe amen p ofiles a e dis inguished by colo -
coding as shown in he legend (see Table 3). c,dCompa ison o le el
o ill-being (cwhe e high alues indica e ill-being) and o le el o
well-being (dwhe e high alues indica e well-being) in g oups o
subjec s wi h each o he h ee empe amen p ofiles specified by bo h
pheno ypic and geno ypic in o ma ion (e alua ed by ANOVA).
(Compa e wi h ei he gene ic o pheno ypic assessmen alone in
Supplemen a y Figu e S5.) eVa ia ion in heal h s a us o SNP se s:
well (blue, see (d)), ill (o ange, see (c)), in e media e (g ay). 19
geno ypic–pheno ypic pipelines connec di e en se s o genes o he
same empe amen dimension (see also Supplemen a y Tables S9–
S11). Red lines indica e di ec connec ions, whe eas blue lines and
“&”indica e composi e connec ions. gSu ace showing he pa e n o
heal h s a us o he subjec s in his s udy based on SNP se in o ma ion
only (i.e., in e pola ion om Table 1). The p obabili y o well-being in
he z-axis a ies om high ( ed o high well-being) o low (g een).
The o de o he SNP se s is based on sha ed subjec s (x-axis) and on
sha ed SNPs (y-axis) measu ed by hype geome ic s a is ics, so SNP
se s sha ing mo e SNPs and/o subjec s a e nea by. (See ill heal h
su ace in Supplemen a y Figu e S3.) hSu ace showing he pa e n o
heal h s a us o subjec s based on bo h geno ypic in o ma ion (SNP
se s) and pheno ypic in o ma ion ( empe amen se s) (as in Table 3).
The p obabili y o well-being in he z-axis a ies om high ( ed, high
well-being) o low (g een). The sha ing o subjec s is shown o bo h
SNP se s (x-axis) and empe amen se s (y-axis). (See ill heal h su ace
in Supplemen a y Figu e S4.)
Unco e ing he complex gene ics o human empe amen
p ocesses in ol ing s ess eac i i y, neu o ansmission
(choline gic, monoamine gic, GABAe gic, glu amine -
gic), esis ance o aging, neu onal and glial g ow h,
myelina ion, and ene gy p oduc ion (Table 1, Supple-
men a y Tables S4–S6).
Complex geno ypic–pheno ypic ela ionships in
empe amen p ofiles
We ound 44 o he 118 empe amen se s we e significan ly
associa ed wi h pa icula SNP se s (Hype geome ic
I. Zwi e al.
s a is ics, 1E−11 < p<1E−03, Table 3). The geno ypic–
pheno ypic ela ions we e complex, demons a ing pleio-
opy and he e ogenei y. Fo example, G_13_3 (ERK-con-
di ioned impulsi i y) is comp ised o mul iple genes ha
egula e beha io al disinhibi ion in associa i e lea ning
asks, such as DAB1 and CDH13 (Table 1, Supplemen a y
Table S4); i was equen ly associa ed wi h sensi i e
empe amen se s, bu some imes wi h an isocial o
eliable p ofiles (Table 3). The 44 empe amen se s we e
associa ed wi h he 51 SNP se s in 158 ela ionships ha
we e significan by a pe mu a ion es (Table 3, empi ical
p< 4.6E−03).
Clus e s o indi iduals sha ing SNPs and/o subjec s
(Fig. 1b) o en had simila empe amen p ofiles associa ed
wi h pa icula molecula p ocesses (Table 3, Supplemen-
a y Tables S4, S7). As p edic ed, each o he empe amen
p ofiles was s ongly associa ed wi h egula ion o synap ic
plas ici y and associa i e condi ioning by genes egula ing
he Ras-MEK-ERK and PI3K-AKT-mTOR cascades in
in e ac ion wi h one ano he , P o ein Kinases A, B (also
known as AKT), and C, and a ious physiological and
psychosocial s esso s (Fig. 2c, Table 1, Supplemen a y
Table S4).
Specific componen s o hese complex molecula cas-
cades dis inguished each empe amen p ofile (Supplemen-
a y Tables S4, S7). Fo example, SNP se s in ol ing
neu oexci abili y (G_35_7, G_37_14), dopamine gic ac i-
a ion (G_16_1, G_35_22, G_39_26), and ol ac ion
(G_7_7) we e associa ed wi h he an isocial p ofile
(Table 1, Fig. 1b, Supplemen a y Table S4). SNP se s
in ol ing esis ance o aging and s ess (G_12_8, G_20_2,
G_21_17, G_30_10, G_33_33), cogni i e flexibili y
(G_21_8, G_38_17), and choline gic neu omodula ion
(G_13_10) we e associa ed wi h he Reliable p ofile. SNP
se s in ol ing senso y sensi i i y (G_38_21), suscep ibili y
o ea condi ioning (G_30_9, G_39_21), s ess eac i i y
(G_7_2, G_11_7, G_26_14), and se o onin–cy okine
in e ac ions in esponse o s ess (G_9_2) we e associa ed
wi h he Sensi i e p ofile.
Rela ions among SNP se s wi h one ano he and
molecula p ocesses
We ound 17 single and disjoin nodes, and a leas 3 sub-
ne wo ks composed o highly connec ed nodes, shown in
Fig. 1b (see Supplemen a y In o ma ion, 9. Iden ifica ion o
sub-ne wo ks). SNP se s G_8_8 (Inosi ol-Chemokine sig-
naling), G_9_2 (Se o onin–Chemokine in e ac ion), and
G_7_3 (Neu ogenesis) each ep esen he hub o sub-
ne wo ks by hei di ec connec ions o 6 o 7 o he SNP
se s. These ne wo ks we e ela i ely disjoin (i.e., sha ing
ew SNPs and subjec s; see Supplemen a y Sec ion 6 (i )),
sugges ing ha hese a e dis inc an eceden s o pe sonali y.
He e ogenic pa hways influence he same
empe amen ai
The genes associa ed wi h each o he h ee empe amen
p ofiles we e la gely unique o ha p ofile. 73.6% o he 736
genes associa ed wi h empe amen we e unique o a single
empe amen p ofile: 266 wi h eliable, 236 wi h sensi i e,
and 40 wi h an isocial (Supplemen a y Table S8). Conse-
quen ly, he e we e mul iple clus e s o genes ha lead o
each indi idual empe amen ai , as depic ed in Fig. 1 .
Fo example, high No el y Seeking is a composi e o
indi iduals wi h he an isocial o sensi i e empe amen
p ofiles because bo h a e associa ed wi h ea u es o high
No el y Seeking. Likewise, high Rewa d Dependence is a
composi e o indi iduals wi h Sensi i e o Reliable p ofiles.
Mo e gene ally, we e e o he mul iple geno ypic–
pheno ypic ne wo ks ha con ibu e o indi idual ai s as a
pipeline, as depic ed in Fig. 1 . The specific genes and
molecula p ocesses in he pipelines o each o he ou
empe amen ai s a e desc ibed in Supplemen a y
Tables S9–S11.
Complex geno ypic–pheno ypic ela ionships
influence heal h s a us
Combining geno ypic and pheno ypic in o ma ion p o ided
mo e in o ma ion han ei he alone o bo h well-being
(Fig. 1g s. 1h) and ill-being (Supplemen a y Figu es S3
s. S4). When heal h s a us was based on he join ela-
ionship o SNP se s and empe amen se s, all h ee em-
pe amen p ofiles we e well dis inguished in e ms o he
p obabili ies o bo h ill-being (p< 1.58E−42, ANOVA
s a is ics, Fig. 1c) and well-being (p< 1.05E−23, ANOVA,
Fig. 1d). In con as , when heal h s a us was based on
empe amen sco es only, he p obabili ies o ill-being (p-
alue < 1.27E−06, ANOVA s a is ics, Supplemen a y
Fig. 2 a,bTypes o gene ic a ian s mapped by SNP se s associa ed
wi h empe amen . aSpecific molecula consequences (Supplemen a y
Table S5) and b hei sub ypes. Genes ela ed only o empe amen se s
( ed) we e less o en p o ein coding and mo e o en RNA genes han
hose also associa ed wi h empe amen se s (blue colo ). cCell dis-
playing he molecula pa hways con aining genes associa ed wi h he
Sensi i e and An isocial p ofiles. The unco e ed genes influence he
Ras-MEK-ERK (MAPK), PI3K-AKT-mTOR, and P o ein Kinase A,
B, C pa hways ha egula e associa i e condi ioning (see also Sup-
plemen a y Tables S4, S7). dMul iple SNPs wi hin a SNP se can
a ec a single o mul iple genes in many ways (Supplemen a y
Table S3). The PIP4K2A, he ARMC3 di e gen egula o y egion,
and he ARMC3 coding egion a e illus a ed. SNPs in he SNP se
G_41_37 may a ec egula o y egions ( he eby inhibi ing ansc ip-
ion), whe eas SNPs om SNP se 39_26 a e mos ly loca ed in in onic
egions ( he eby blocking o dec easing p o ein p oduc ion). The SNP
se s a e associa ed wi h p ofiles exhibi ing dis inc empe amen ea-
u es (sensi i e s. an isocial)
Unco e ing he complex gene ics o human empe amen
sample, and 37% in he Ko ean sample (Supplemen a y
Table S21). In addi ion, 87% o he SNP se s we e s ongly
associa ed wi h he empi ical empe amen index (5E−08 >
p- alue > 5E−73). In o he wo ds, he SNPs ha comp ise
he di e en SNP se s s ongly dis inguished he empe a-
men ea u es o he subjec s in each se , indica ing ha each
indi idual SNP se con ibu ed significan ly o explain he
o al dis ibu ed he i abili y (Supplemen a y Sec ion 9).
Consequen ly, when he geno ypic se s we e used o clas-
si y he well- and ill-being o he subjec s using he
PGMRA classifie , he p edic ed alues we e highly accu-
a e (a e age A eas Unde Cu e o he classifica ions we e
0.940 and 0.922, espec i ely) (Supplemen a y Figu e S8).
We also conside ed en i onmen al influences in he
Finnish sample. The e we e di ec associa ions o se s o
en i onmen al influences in childhood and adul hood
(Supplemen a y Table S22A) wi h empe amen se s (Sup-
plemen a y Table S22B) and wi h SNP se s (Supplemen a y
Table S22C). The impac o hese co ela ions was small, so
he he i abili y es ima e was s ill 46–52% in he Finnish
sample when adjus ed o gene–en i onmen co ela ion
(Supplemen a y Table S21).
Fu he mo e, 12 no el associa ions be ween SNP se s
and empe amen se s we e unco e ed when en i onmen al
influences we e used as media o s (Supplemen a y
Table S22D). Se en SNP se s associa ed wi h he an isocial
p ofile depended on exposu e o low pa en al income du ing
childhood, s ess ul li e e en s in adul hood, and u al
esidence in childhood o adul hood (p< 3.4E−03 o 6.3E
−04). Two SNP se s associa ed wi h sensi i e p ofiles
depended on he expe ience o ole ance and low income in
childhood (p< 9.7E−04 o 4.7E−05). One SNP se asso-
cia ed wi h eliable p ofiles depended on high pa en al
income h oughou childhood (p< 1.5E−04).
Discussion
SNPs ha map o 736 genes explained 48% o he a ia-
bili y in empe amen in he Finnish sample, he eby
accoun ing o nea ly all he he i abili y o human em-
pe amen expec ed om win s udies. Mo e specifically,
mos o he genes ha we iden ified in a s ic ly da a-d i en
manne a e known o egula e synap ic plas ici y, associa-
i e condi ioning, and ela ed p ocesses o s ess eac i i y
and neu o ansmission. These findings confi m ou
hypo hesis ha he highly conse ed molecula p ocesses
ha egula e associa i e condi ioning in expe imen al ani-
mals accoun subs an ially o he he i abili y o human
empe amen . Ou findings a e suppo ed in independen
eplica ions by GWAS and by independen s udies o
gene exp ession du ing habi lea ning in expe imen al ani-
mals [7,32,33].
Molecula pa hways o empe amen and
associa i e condi ioning
Mos o he SNP se s associa ed wi h empe amen
we e in ol ed in he egula ion o habi lea ning and
synap ic plas ici y in esponse o ex acellula s imuli
media ed mainly by he Ras-MEK-ERK and he PI3K-
AKT-mTOR cascades (Table 1,Fig.2c). As p edic ed,
hese main pa hways o as adap i e esponse ope a ed
in conjunc ion wi h ela ed p ocesses o s ess
eac i i y, neu o ansmission, ch oma in plas ici y, neu-
onal and glial g ow h, myelina ion, neu op o ec ion, and
ene gy p oduc ion (Table 1, Supplemen a y Tables S4–
S6). The iden ified pa hways o associa i e condi ioning
a e known o in e sec o egula e each o he and o
co- egula e downs eam unc ions [52], as illus a ed
specificallyinFig.2c. The mechanisms o in eg a ion
o he ERK and PI3K cascades include mechanisms
o c oss-ac i a ion, c oss-inhibi ion, nega i e eedback,
and posi i e and nega i e influences ha con e ge on
he same complex (e.g., mTOR in Fig. 2c). In addi ion,
p o ein kinases A, B (also known as AKT), and C
ha egula e hese pa hways a e a he non-selec i e
[52]. Such in e ac ions a e expec ed o p oduce
complex geno ypic–pheno ypic ela ionships, as we
obse ed.
These findings abou specific molecula pa hways o
human empe amen ha e impo an implica ions. Fi s ,
hey confi m ou hypo hesis ha he human empe amen
is based on he highly conse ed mechanisms o habi
lea ning. This suppo s a p ecise defini ion o empe a-
men in e ms o associa i e condi ioning [17,18]. Sec-
ond, he independen expe imen al suppo o specific
molecula pa hways o associa i e condi ioning p o ides
suppo o he alidi y o he s ic ly da a-d i en me hod
we used o analyze and in e p e genome-wide associa ion
da a.
These esul s should encou age widesp ead use o
PGMRA o analysis o complex pheno ypes in a a ie y
o se ings, including GWAS [54,55] and neu oimaging
[53]. Fo example, PGMRA p o ides an e ec i e way o
allow o epis asis and gene–en i onmen in e ac ions ha
a e p ominen in complex pheno ypes, he eby o e -
coming he hidden he i abili y p oblem ( ha is, he con-
sis en inabili y o accoun o mos o he he i abili y o
complex ai s when only he a e age e ec s o genes a e
conside ed). The gene alized clus e ing me hod imple-
men ed in PGMRA can be in e p e ed as a deep unsu-
pe ised NMF lea ning p ocess ha can iden i y clus e s
o indi iduals wi h dis inc ea u es om a ious ypes o
in o ma ion, such as he geno ypes, pheno ypes, and
en i onmen s (Supplemen a y Figu e S1). Such clus e s,
SNP se s, and empe amen se s can be used as au o-
I. Zwi e al.

encode s used by ecommende sys ems in p ecision
medicine [55].
S eng hs and limi a ions
The majo s eng h o hese findings is he s ong eplic-
abili y o he findings in h ee independen samples om
di e en cul u es and in independen s udies o gene
exp ession du ing beha io al condi ioning o expe imen al
animals. While i is ue ha clus e analysis is a hypo hesis-
gene a ing me hod in which he e is no unique solu ion o
he numbe o clus e s, which ea u es a e ele an o a
clus e , o he deg ee o homogenei y o be demanded o
each clus e , PGMRA included a p ac ical and obus
solu ion o each o hese p oblems [53,54].
Conclusions and ecommenda ions o u u e
esea ch
We we e able o desc ibe and eplica e he complex geno-
ypic–pheno ypic isk a chi ec u e o empe amen in h ee
independen samples o people. Ou unbiased da a-d i en
findings confi m he hypo hesis ha empe amen is based
on associa i e condi ioning and ela ed p ocesses, pa icu-
la ly s ess eac i i y in esponse o ex acellula s imuli.
We ha e ound ha di e en molecula and cogni i e p o-
cesses a e associa ed wi h cha ac e [40], bu heal h s a us
depends on geno ypic–pheno ypic ela ions ha influence
bo h empe amen and cha ac e . The e o e, we ecommend
u he wo k o examine he o e lap and in e ac ions
be ween empe amen and cha ac e .
Acknowledgemen s The Young Finns S udy has been financially
suppo ed by he Academy o Finland: g an s 286284, 134309 (Eye),
126925, 121584, 124282, 129378 (Sal e), 117787 (Gendi), and 41071
(Skidi); he Social Insu ance Ins i u ion o Finland; Compe i i e S a e
Resea ch Financing o he Expe Responsibili y a ea o Kuopio,
Tampe e and Tu ku Uni e si y Hospi als (g an X51001); Juho Vainio
Founda ion; Paa o Nu mi Founda ion; Finnish Founda ion o Ca -
dio ascula Resea ch; Finnish Cul u al Founda ion; Tampe e Tube -
culosis Founda ion; Emil Aal onen Founda ion; Y jö Jahnsson
Founda ion; Signe and Ane Gyllenbe g Founda ion; Diabe es
Resea ch Founda ion o Finnish Diabe es Associa ion: and EU Ho -
izon 2020 (g an 755320 o TAXINOMISIS). The Ame ican Foun-
da ion o Suicide P e en ion suppo ed he s udy o heal hy Ge mans.
The na ional Heal hy Twin Family Regis e o Ko ea suppo ed he
s udy o heal hy Ko eans. The An h opedia Founda ion and he
Spanish Minis y o Science and Technology TIN2012-38805 and
DPI2015-69585-R including FEDER unds suppo ed his collabo a-
ion. We hank bios a is icians Cha les Gu and Miguel Angel Rubio
Escude o o ad ice on s a is ical analysis, and fi e anonymous
e iewe s o hei help ul commen s on he manusc ip .
Compliance wi h e hical s anda ds
Conflic o in e es The au ho s decla e ha hey ha e no conflic o
in e es .
Open Access This a icle is licensed unde a C ea i e Commons
A ibu ion 4.0 In e na ional License, which pe mi s use, sha ing,
adap a ion, dis ibu ion and ep oduc ion in any medium o o ma , as
long as you gi e app op ia e c edi o he o iginal au ho (s) and he
sou ce, p o ide a link o he C ea i e Commons license, and indica e i
changes we e made. The images o o he hi d pa y ma e ial in his
a icle a e included in he a icle’s C ea i e Commons license, unless
indica ed o he wise in a c edi line o he ma e ial. I ma e ial is no
included in he a icle’s C ea i e Commons license and you in ended
use is no pe mi ed by s a u o y egula ion o exceeds he pe mi ed
use, you will need o ob ain pe mission di ec ly om he copy igh
holde . To iew a copy o his license, isi h p://c ea i ecommons.
o g/licenses/by/4.0/.
Re e ences
1. Shine RL, Buss KA, McClow y SG, Pu nam SP, Saudino KJ,
Zen ne M. Wha is Tempe amen Now? Assessing p og ess in
empe amen esea ch on he wen y-fi h anni e sa y o Gold-
smi h e al. (1987). Child De Pe spec . 2012;6:436–44.
2. Cloninge CR. The e olu ion o human b ain unc ions: he
unc ional s uc u e o human consciousness. Aus N Z J Psy-
chia y. 2009;43:994–1006.
3. Cloninge CR, Kedia S. The phylogenesis o human pe sonali y:
iden i ying he p ecu so s o coope a ion, al uism, & well-
being. In: Sussman RW, Cloninge CR, edi o s. The o igins o
coope a ion and al uism. 1s ed. New Yo k: Sp inge ; 2011.
p. 63–110.
4. MacLean PD. E olu iona y psychia y and he iune b ain. Psy-
chol Med. 1985;15:219–21.
5. Cloninge CR, Gilligan SB. Neu ogene ic mechanisms o lea n-
ing: a phylogene ic pe spec i e. J Psychia Res. 1987;21:457–72.
6. A kins CM, Selche JC, Pe ai is JJ, T zaskos JM, Swea JD. The
MAPK cascade is equi ed o mammalian associa i e lea ning.
Na Neu osci. 1998;1:602–9.
7. Lusche C, Malenka RC. NMDA ecep o -dependen long- e m
po en ia ion and long- e m dep ession (LTP/LTD). Cold Sp ing
Ha b Pe spec Biol. 2012;4:1–16.
8. Kandel ER. The molecula biology o memo y s o age: a dialogue
be ween genes and synapses. Science. 2001;294:1030–8.
9. Colling idge GL, Peineau S, Howland JG, Wang YT. Long- e m
dep ession in he CNS. Na Re Neu osci. 2010;11:459–73.
10. de Waal FB. P ima es—a na u al he i age o conflic esolu ion.
Science. 2000;289:586–90.
11. Tul ing E. Episodic memo y and common sense: how a apa ?
Philos T ans R Soc Lond B Biol Sci. 2001;356:1505–15.
12. Tul ing E. Episodic memo y: om mind o b ain. Annu Re
Psychol. 2002;53:1–25.
13. Tul ing E. Mul iple memo y sys ems and consciousness. Hum
Neu obiol. 1987;6:67–80.
14. Le ine B. Au obiog aphical memo y and he sel in ime: b ain
lesion e ec s, unc ional neu oana omy, and li espan de elop-
men . B ain Cogn. 2004;55:54–68.
15. Po inelli DJ. Folk physics o apes: he chimpanzee’s heo y o
how he wo ld wo ks. New Yo k: Ox o d Uni e si y P ess; 2000.
16. Jose sson K, Jokela M, Cloninge CR, Hin sanen M, Salo J,
Hin sa T, e al. Ma u i y and change in pe sonali y: de elopmen al
ends o empe amen and cha ac e in adul hood. De Psycho-
pa hol. 2013;25:713–27.
17. Cloninge CR. A sys ema ic me hod o clinical desc ip ion and
classifica ion o pe sonali y a ian s. A p oposal. A ch Gen Psy-
chia y. 1987;44:573–88.
18. Cloninge CR, S akic DM, P zybeck TR. A psychobiological
model o empe amen and cha ac e . A ch Gen Psychia y.
1993;50:975–90.
Unco e ing he complex gene ics o human empe amen
19. Cloninge CR. A unified biosocial heo y o pe sonali y and i s
ole in he de elopmen o anxie y s a es. Psychia De .
1986;4:167–226.
20. Co PJ, Kuma i V, Wilson GD, Checkley S, G ay JA. Ha m
a oidance and a ec i e modula ion o he s a le eflex: a epli-
ca ion. Pe s Indi id Di . 1997;22:591–3.
21. Co PJ, Picke ing AD, G ay JA. Pe sonali y and ein o cemen in
associa i e and ins umen al lea ning. Pe s Indi id Di . 1995;
19:47–71.
22. Pezawas L, Meye -Lindenbe g A, D aban EM, Ve chinski BA,
Munoz KE, Kolachana BS, e al. 5-HTTLPR polymo phism
impac s human cingula e–amygdala in e ac ions: a gene ic sus-
cep ibili y mechanism o dep ession. Na Neu osci. 2005;
8:828–34.
23. Houillon A, Lo enz RC, Boehme W, Rapp MA, Heinz A, Gal-
lina J, e al. The e ec o no el y on ein o cemen lea ning. P og
B ain Res. 2013;202:415–39.
24. K ebs RM, Scho BH, Duzel E. Pe sonali y ai s a e di e en-
ially associa ed wi h pa e ns o ewa d and no el y p ocessing in
he human subs an ia nig a/ en al egmen al a ea. Biol Psy-
chia y. 2009;65:103–10.
25. Gusna d DA, Ollinge JM, Shulman GL, Cloninge CR, P ice JL,
Van Essen DC, e al. Pe sis ence and b ain ci cui y. P oc Na l
Acad Sci USA. 2003;100:3479–84.
26. Cloninge CR, Zoha AH, Hi schmann S, Dahan D. The psy-
chological cos s and benefi s o being highly pe sis en : pe son-
ali y p ofiles dis inguish mood diso de s om anxie y diso de s. J
A ec Diso d. 2012;136:758–66.
27. Yang TT, Simmons AN, Ma hews SC, Tape SF, F ank GK,
Bischo -G e he A, e al. Adolescen subgenual an e io cingula e
ac i i y is ela ed o ha m a oidance. Neu o epo . 2009;20:
19–23.
28. Paulus MP, Rogalsky C, Simmons A, Feins ein JS, S ein MB.
Inc eased ac i a ion in he igh insula du ing isk- aking decision
making is ela ed o ha m a oidance and neu o icism. Neu o-
image. 2003;19:1439–48.
29. Nagha i HR, Lind J, Nilsson LG, Adol sson R, Nybe g L. Pe -
sonali y ai s p edic esponse o no el and amilia s imuli in he
hippocampal egion. Psychia y Res. 2009;173:94–9.
30. Lei X, Chen C, Xue F, He Q, Liu Q, Moyzis RK, e al. Fibe
connec i i y be ween he s ia um and co ical and subco ical
egions is associa ed wi h empe amen s in Chinese males. Neu-
oimage. 2014;89:226–34.
31. Tos H, Kolachana B, Hakimi S, Lemai e H, Ve chinski BA,
Ma ay VS, e al. A common allele in he oxy ocin ecep o gene
(OXTR) impac s p osocial empe amen and human
hypo halamic-limbic s uc u e and unc ion. P oc Na l Acad Sci
USA. 2010;107:13936–41.
32. Peng S, Zhang Y, Zhang J, Wang H, Ren B. ERK in lea ning
and memo y: a e iew o ecen esea ch. In J Mol Sci.
2010;11:222–32.
33. T onson NC, Co co an KA, Jo ase ic V, Radulo ic J. Fea
condi ioning and ex inc ion: emo ional s a es encoded by dis inc
signaling pa hways. T ends Neu osci. 2012;35:145–55.
34. Rapanelli M, F ick LR, Zanu o BS. Di e en ial gene exp ession
in he a hippocampus du ing lea ning o an ope an condi ioning
ask. Neu oscience. 2009;163:1031–8.
35. B ems B. Spon aneous decisions and ope an condi ioning in ui
flies. Beha P ocess. 2011;87:157–64.
36. Sakai N, Ohno H, Tomioka M, Iino Y. The in es inal
TORC2 signaling pa hway con ibu es o associa i e lea ning in
Caeno habdi is elegans. PLoS ONE. 2017;12:e0177900.
37. Diamond DM, Pa k CR, Campbell AM, Woodson JC. Compe i-
i e in e ac ions be ween endogenous LTD and LTP in he hip-
pocampus unde lie he s o age o emo ional memo ies and s ess-
induced amnesia. Hippocampus. 2005;15:1006–25.
38. Rapanelli M, F ick LR, Zanu o BS. Lea ning an ope an con-
di ioning ask di e en ially induces gliogenesis in he medial
p e on al co ex and neu ogenesis in he hippocampus. PLoS
ONE. 2011;6:e14713.
39. Rapanelli M, F ick LR, Be na dez-Vidal M, Zanu o BS. Di e en
MK-801 adminis a ion schedules induce mild o se e e lea ning
impai men s in an ope an condi ioning ask: ole o buspi one and
ispe idone in amelio a ing hese cogni i e defici s. Beha B ain
Res. 2013;257:156–65.
40. Zwi I, A nedo J, Del-Val C, Pulkki-Råback L, Kon e B, Yang SS,
e al. Unco e ing he complex gene ics o human cha ac e . Mol
Psychia y. 2018. h ps://doi.o g/10.1038/s41380-018-0263-6
41. Se ice SK, Ve weij KJ, Lah i J, Congdon E, Ekelund J, Hin -
sanen M, e al. A genome-wide me a-analysis o associa ion s u-
dies o Cloninge ’s Tempe amen Scales. T ansl Psychia y. 2012;
2:e116.
42. Ve weij KJH, Zie sch BP, Medland SE, Go don SD, Benyamin B,
Nyhol DR, e al. A genome-wide associa ion s udy o Cloninge ’s
empe amen scales: implica ions o he e olu iona y gene ics o
pe sonali y. Biol Psychol. 2010;85:306–17.
43. Eichle EE, Flin J, Gibson G, Kong A, Leal SM, Moo e JH, e al.
Missing he i abili y and s a egies o finding he unde lying
causes o complex disease. Na Re Gene . 2010;11:446–50.
44. Williams SM, Haines JL. Co ec ing away he hidden he i abili y.
Ann Hum Gene . 2011;75:348–50.
45. Kelle MC, Co en y WL, Hea h AC, Ma in NG. Widesp ead
e idence o non-addi i e gene ic a ia ion in Cloninge ’s and
Eysenck’s Pe sonali y Dimensions using a Twin Plus Sibling
Design. Beha Gene . 2005;35:707–21.
46. Ve weij KJ, Yang J, Lah i J, Veijola J, Hin sanen M, Pulkki-
Raback L, e al. Main enance o gene ic a ia ion in human pe -
sonali y: es ing e olu iona y models by es ima ing he i abili y
due o common causal a ian s and in es iga ing he e ec o
dis an inb eeding. E olu ion. 2012;66:3238–51.
47. Benjamin J, Oshe Y, Ko le M, G i senko I, Nemano L, Bel-
make RH, e al. Associa ion be ween idimensional pe sonali y
ques ionnai e (TPQ) ai s and h ee unc ional polymo phisms:
dopamine ecep o D4 (DRD4), se o onin anspo e p omo e
egion (5-HTTLPR) and ca echol O-me hyl ans e ase (COMT).
Mol Psychia y. 2000;5:96–100.
48. S obel A, Lesch KP, Ja zke S, Pae zold F, B ocke B. Fu he
e idence o a modula ion o No el y Seeking by DRD4 exon III,
5-HTTLPR, and COMT al/me a ian s. Mol Psychia y. 2003;
8:371–2.
49. Kel ikangas-Ja inen L, Pu onen S, Ki imaki M, Ron u R, Leh-
imaki T. Cloninge ’s empe amen dimensions and epide mal
g ow h ac o A61G polymo phism in Finnish adul s. Genes B ain
Beha . 2006;5:11–8.
50. Kel ikangas-Ja inen L, Salo J. Dopamine and se o onin sys ems
modi y en i onmen al e ec s on human beha io : a e iew. Scand
J Psychol. 2009;50:574–82.
51. Kel ikangas-Ja inen L, Pulkki-Raback L, Elo ainio M, Rai aka i
OT, Viika i J, Leh imaki T. DRD2 C32806T modifies he e ec o
child- ea ing en i onmen on adul hood no el y seeking. Am J
Med Gene B Neu opsychia Gene . 2009;150B:389–94.
52. Mendoza MC, E EE, Blenis J. The Ras-ERK and PI3K-mTOR
pa hways: c oss- alk and compensa ion. T ends Biochem Sci.
2011;36:320–8.
53. A nedo J, Mamah D, Ba ange DA, Ha ms MP, Ba ch DM,
S akic DM, e al. Decomposi ion o b ain di usion imaging da a
unco e s la en schizoph enias wi h dis inc pa e ns o whi e
ma e aniso opy. Neu oimage. 2015;120:43–54.
54. A nedo J, del Val C, de E ausquin GA, Rome o-Zaliz R, S akic
D, Cloninge CR, e al. PGMRA: a web se e o (pheno ype x
geno ype) many- o-many ela ion analysis in GWAS. Nucleic
Acids Res. 2013;41:W142–9.
I. Zwi e al.
55. A nedo J, S akic DM, Del Val C, Rome o-Zaliz R, He nandez-
Cue o H, Fanous AH, e al. Unco e ing he hidden isk a chi-
ec u e o he schizoph enias: confi ma ion in h ee independen
genome-wide associa ion s udies. Am J Psychia y. 2015;
172:139–53.
56. Wu MC, K a P, Eps ein MP, Taylo DM, Chanock SJ, Hun e
DJ, e al. Powe ul SNP-se analysis o case-con ol genome-wide
associa ion s udies. Am J Hum Gene . 2010;86:929–42.
57. Wu MC, Lee S, Cai T, Li Y, Boehnke M, Lin X. Ra e- a ian
associa ion es ing o sequencing da a wi h he sequence ke nel
associa ion es . Am J Hum Gene . 2011;89:82–93.
58. Rai aka i OT, Juonala M, Ronnemaa T, Kel ikangas-Ja inen L,
Rasanen L, Pie ikainen M, e al. Coho p ofile: he
ca dio ascula isk in Young Finns S udy. In J Epidemiol. 2008;
37:1220–6.
59. Cook TB, B enne LA, Cloninge CR, Langenbe g P, Igbide A,
Giegling I, e al. La en In ec ion wi h Toxoplasma gondii: asso-
cia ion wi h ai agg ession and impulsi i y in heal hy adul s. J
Psychia Res. 2015;60:87–94.
60. Ree es GM, Mazahe i S, Sni ke S, Langenbe g P, Giegling I,
Ha mann AM, e al. A posi i e associa ion be ween T. gondii
se oposi i i y and obesi y. F on Public Heal h. 2013;1:73.
61. Gomboja B, Song YM, Lee K, Yang S, Kho M, Hwang YC,
e al. The Heal hy Twin S udy, Ko ea upda es: esou ces o omics
and genome epidemiology s udies. Twin Res Hum Gene .
2013;16:241–5.
62. Sung J, Cho SI, Lee K, Ha M, Choi EY, Choi JS, e al. Heal hy
Twin: a win- amily s udy o Ko ea—p o ocols and cu en s a us.
Twin Res Hum Gene . 2006;9:844–8.
63. Cloninge CR, P zybeck TR, S akic DM, We zel RD. The
Tempe amen and Cha ac e In en o y (TCI): a guide o i s
de elopmen and use. S . Louis: Washing on Uni e si y Cen e
o Psychobiology o Pe sonali y; 1994.
64. Cloninge CR, Zoha AH. Pe sonali y and he pe cep ion o heal h
and happiness. J A ec Diso d. 2011;128:24–32.
65. S akic DM, Whi ehead C, P zybeck TR, Cloninge CR. Di e -
en ial diagnosis o pe sonali y diso de s by he se en- ac o model
o empe amen and cha ac e . A ch Gen Psychia y. 1993;
50:991–9.
66. Cloninge CR. A p ac ical way o diagnosis pe sonali y diso de : a
p oposal. J Pe s Diso d. 2000;14:99–108.
67. Zoha AH, Cloninge CR, McC a y R. Pe sonali y and hea a e
a iabili y: explo ing pa hways om pe sonali y o ca diac
cohe ence and heal h. J Social Sci. 2013;1:32–9.
68. Cloninge CR. Wha makes people heal hy, happy, and ulfilled in
he ace o cu en wo ld challenges? Mens Sana Monog .
2013;11:16–24.
69. Smi h EN, Chen W, Kahonen M, Ke unen J, Leh imaki T, Pel-
onen L, e al. Longi udinal genome-wide associa ion o ca dio-
ascula disease isk ac o s in he Bogalusa hea s udy. PLoS
Gene . 2010;6:e1001094.
70. Pu cell S, Neale B, Todd-B own K, Thomas L, Fe ei a MA,
Bende D, e al. PLINK: a ool se o whole-genome associa ion
and popula ion-based linkage analyses. Am J Hum Gene .
2007;81:559–75.
71. P ice AL, Pa e son NJ, Plenge RM, Weinbla ME, Shadick NA,
Reich D. P incipal componen s analysis co ec s o s a ifica ion
in genome-wide associa ion s udies. Na Gene . 2006;38:904–9.
72. Zwi I, Shin D, Ka o A, Nishino K, La ifiT, Solomon F, e al.
Dissec ing he PhoP egula o y ne wo k o Esche ichia coli and
Salmonella en e ica. P oc Na l Acad Sci USA. 2005;102:2862–7.
73. Zwi I, Huang H, G oisman EA. Analysis o di e en ially-
egula ed genes wi hin a egula o y ne wo k by GPS genome
na iga ion. Bioin o ma ics. 2005;21:4073–83.
74. Rome o-Zaliz R, Del Val C, Cobb JP, Zwi I. On o-CC: a web
se e o iden i ying Gene On ology concep ual clus e s. Nucleic
Acids Res. 2008;36:W352–7.
75. Ha a i O, Pa k SY, Huang H, G oisman EA, Zwi I. Defining he
plas ici y o ansc ip ion ac o binding si es by decons uc ing
DNA consensus sequences: he PhoP-binding si es among
gamma/en e obac e ia. PLoS Compu Biol. 2010;6:e1000862.
76. Ma quand AF, Wol e s T, Mennes M, Bui elaa J, Beckmann CF.
Beyond lumping and spli ing: a e iew o compu a ional
app oaches o s a i ying psychia ic diso de s. Biol Psychia y
Cogn Neu osci Neu oimaging. 2016;1:433–47.
77. In e na ional Schizoph enia Conso ium, Pu cell SM, W ay NR,
S one JL, Vissche PM, O’Dono an MC. e al. Common poly-
genic a ia ion con ibu es o isk o schizoph enia and bipola
diso de . Na u e. 2009;460:748–52.
78. Lango Allen H, Es ada K, Le e G, Be nd SI, Weedon MN, Ri -
adenei a F, e al. Hund eds o a ian s clus e ed in genomic loci and
biological pa hways a ec human heigh . Na u e. 2010;467:832–8.
A filia ions
Igo Zwi 1,2 ●Ja ie A nedo 1,2 ●Co al Del-Val2●Lau a Pulkki-Råback3●Be ina Kon e 4●Sa ah S. Yang5●
Rocio Rome o-Zaliz2●Mi ka Hin sanen6●Ke in M. Cloninge 7●Danilo Ga cia 8,9 ●D agan M. S akic1●
Sando Rozsa1●Ma ibel Ma inez1●Leo-Pekka Lyy ikäinen 10 ●Ina Giegling4,11 ●Mika Kähönen12 ●
Helena He nandez-Cue o13 ●Ilkka Seppälä 10 ●Emma Rai oha ju10 ●Gab iel A. de E ausquin14 ●Olli Rai aka i15 ●
Dan Rujescu4●Teodo T. Pos olache16,17 ●Joohon Sung5●Liisa Kel ikangas-Jä inen3●Te ho Leh imäki10 ●
C. Robe Cloninge 1,18
1Depa men o Psychia y, Washing on Uni e si y School o
Medicine, S . Louis, MO, USA
2Depa men o Compu e Science, Uni e si y o G anada,
G anada, Spain
3Depa men o Psychology and Logopedics, Uni e si y o
Helsinki, Helsinki, Finland
4Depa men o Psychia y, Ma in-Lu he -Uni e si y Halle-
Wi enbe g, Halle, Ge many
5Depa men o Epidemiology, School o Public Heal h, Ins i u e o
Heal h and En i onmen , Seoul Na ional Uni e si y, Seoul, Ko ea
6Uni o Psychology, Facul y o Educa ion, Uni e si y o Oulu,
Oulu, Finland
7An h opedia Founda ion, S . Louis, MO, USA
8Depa men o Psychology, Uni e si y o Go henbu g,
Go henbu g, Sweden
Unco e ing he complex gene ics o human empe amen
9Blekinge Cen e o Compe ence, Blekinge Coun y Council,
Ka lsk ona, Sweden
10 Fimlab Labo a o ies, Depa men o Clinical Chemis y, Facul y o
Medicine and Li e Sciences, Finnish Ca dio ascula Resea ch
Cen e -Tampe e, Uni e si y o Tampe e, Tampe e, Finland
11 Uni e si y Clinic, Ludwig-Maximilian Uni e si y,
Munich, Ge many
12 Depa men o Clinical Physiology, Facul y o Medicine and Li e
Sciences, Tampe e Uni e si y Hospi al, Uni e si y o Tampe e,
Tampe e, Finland
13 Depa men o Psychia y and Neu osu ge y, Uni e si y o Sou h
Flo ida, Tampa, FL, USA
14 Depa men o Psychia y and Neu ology, Ins i u e o
Neu osciences, Uni e si y o Texas Rio-G ande Valley School o
Medicine, Ha lingen, TX, USA
15 Depa men o Clinical Physiology and Nuclea Medicine, Tu ku
Uni e si y Hospi al, Tu ku, Finland
16 Depa men o Psychia y, Uni e si y o Ma yland School o
Medicine, Bal imo e, MD, USA
17 Rocky Moun ain Men al Illness, Resea ch, Educa ion and Clinical
Cen e o Ve e an Suicide P e en ion, Den e , CO, USA
18 Depa men o Psychological and B ain Sciences, School o A s
and Sciences, and Depa men o Gene ics, School o Medicine,
Washing on Uni e si y School o Medicine, S . Louis, MO, USA
I. Zwi e al.