ARTICLE
Recei ed 9 Ma 2017 |Accep ed 8 May 2017 |Published 14 Jun 2017
Gene ic loci associa ed wi h hea a e a iabili y
and hei e ec s on ca diac disease isk
Ilja M. Nol e e al.#
Reduced ca diac agal con ol eflec ed in low hea a e a iabili y (HRV) is associa ed wi h
g ea e isks o ca diac mo bidi y and mo ali y. In wo-s age me a-analyses o genome-wide
associa ion s udies o h ee HRV ai s in up o 53,174 indi iduals o Eu opean ances y, we
de ec 17 genome-wide significan SNPs in eigh loci. HRV SNPs ag non-synonymous SNPs
(in NDUFA11 and KIAA1755), exp ession quan i a i e ai loci (eQTLs) (influencing GNG11,
RGS6 and NEO1), o a e loca ed in genes p e e en ially exp essed in he sinoa ial node
(GNG11,RGS6 and HCN4). Gene ic isk sco es accoun o 0.9 o 2.6% o he HRV a iance.
Significan gene ic co ela ion is ound o HRV wi h hea a e ( 0.74o
g
o0.55) and
blood p essu e ( 0.35o
g
o0.20). These findings p o ide clinically ele an biological
insigh in o he i able a ia ion in agal hea hy hm egula ion, wi h a key ole o gene ic
a ian s (GNG11,RGS6) ha influence G-p o ein he e o ime ac ion in GIRK-channel induced
pacemake memb ane hype pola iza ion.
Co espondence and eques s o ma e ials should be add essed o I.M.N. (email: [email p o ec ed]) o o H.S. (email: h.sniede @umcg.nl) o o E.J.C.d.G.
(email: [email p o ec ed]l).
#A ull lis o au ho s and hei a filia ions appea s a he end o he pape .
DOI: 10.1038/ncomms15805 OPEN
NATURE COMMUNICATIONS | 8:15805 | DOI: 10.1038/ncomms15805 | www.na u e.com/na u ecommunica ions 1
Hea a e a iabili y (HRV) is a physiological a ia ion in
ca diac cycle du a ion. When measu ed unde supine o
si ing condi ions, es ing HRV is mos p ominen ly
cen ed a ound he equency o espi a ion (B0.25 Hz) and he
in insic blood p essu e hy hm (B0.1 Hz). This eflec s
modula ion o onic ac i i y in he ca diac agal ne es
o igina ing in co ical and subco ical nuclei1by oscilla o y
inpu a he b ains em le el om ca dio espi a o y coupling, lung
s e ch- eflexes, and a e ial chemo- and ba o ecep o s1,2.
This agal ga ing gi es ise o oscilla o y agal e ec s on
he pacemake po en ials in he sinoa ial node ha scales wi h
he onic ac i i y in he agal ne es and p o ides a sou ce o
bea - o-bea a ia ion in hea a e. Due o i s good
ep oducibili y3and ease o measu emen , HRV is a widely
used non-in asi e esea ch and clinical ool o quan i y he
deg ee o agal con ol o hea a e4.
Loss o ca diac agal con ol as indexed by low HRV is
associa ed wi h mo ali y in pa ien s wi h ca dio ascula disease5.
Animal esea ch u he suppo s a ole o ca diac agal ac i i y
in p e en ing sudden dea h and en icula fib illa ion6.In
addi ion, hype ension7, end-s age enal disease8and diabe es9
a e all associa ed wi h low HRV. Al hough he abo e associa ions
may pa ly eflec impai ed ca diac agal con ol caused by hese
diseases, lowe ed HRV does no simply indica e disease se e i y
as i also p edic s all-cause mo ali y10 and ca diac mo bidi y and
mo ali y11,12 in appa en ly heal hy indi iduals.
La ge in e -indi idual di e ences in HRV exis in he basal
es ing s a e. Family and win s udies ha e uni o mly confi med a
subs an ial gene ic con ibu ion o es ing HRV wi h he i abili y
es ima es be ween 25 and 71% ( e . 13). Candida e gene s udies
based on cu en knowledge o pa asympa he ic ne ous sys em
biology ha e no yielded esul s ha hold up in eplica ion14.
To imp o e ou unde s anding o he gene ic basis o HRV,
we pe o med a wo-s age me a-analysis o genome-wide
associa ion s udies (GWAS) in up o 53,174 indi iduals
o Eu opean ances y on h ee HRV ai s ( he s.d. o he
no mal- o-no mal in e bea in e als (SDNN), he oo mean
squa e o he successi e di e ences o in e bea in e als
(RMSSD) and he peak- alley espi a o y sinus a hy hmia o
high equency powe (p RSA/HF)). These HRV ai s we e
measu ed du ing es ing, basal eco dings anging in leng h om
ul asho 10-s elec oca diog ams (ECGs) o up o 90 min o
si ing o om 2 o 12 h o day ime eco ding. Rele ance o he
iden ified loci o o he e hnici ies was examined in da a om
11,234 Hispanic/La ino and 6,899 A ican-Ame ican indi iduals.
In silico pos -GWAS analyses we e pe o med o es o
associa ion wi h ca diac disease isk ac o s and disease
ou comes and o p o ide insigh s in o he biological
mechanisms by which he iden ified loci influence ca diac agal
con ol and i s e ec on HRV.
We de ec 17 SNPs in eigh loci ha bou ing se e al genes
p e e en ially exp essed in he sinoa ial node and significan
nega i e gene ic co ela ions o HRV wi h hea a e and blood
p essu e. These findings p o ide clinically ele an biological
insigh in o he i able a ia ion in agal hea hy hm egula ion,
wi h a key ole o gene ic a ian s in p o eins (RGS6, GNG11)
known o influence G-p o ein he e o ime ac ion in
GIRK-channel induced pacemake memb ane hype pola iza ion.
Resul s
New loci associa ed wi h HRV. We me a-analysed esul s om
GWAS on h ee HRV ai s (see Me hods sec ion o de ails)
pe o med by 20 coho s o Eu opean ances y in up o 28,700
indi iduals (Fig. 1; Supplemen a y Figs 1–3; Supplemen a y
Tables 1–4). Using a significance h eshold o 1 106,
23 single-nucleo ide polymo phism (SNPs) in 14 loci ha we e
associa ed wi h one o mo e o hese HRV ai s we e aken
o wa d o we -lab geno yping o in silico eplica ion in 11
coho s including up o 24,474 addi ional indi iduals o Eu opean
ances y, ollowed by a second s age me a-analysis
(Supplemen a y Da a 1).
A e s age 2, we iden ified 17 lead SNPs (11 independen )
in eigh loci (Table 1) ha eached genome-wide significance
(Po5108). The loci on ch omosomes 14 and 15 con ained
h ee and wo independen signals, espec i ely, (Supplemen a y
Fig. 3). Condi ional analysis confi med he p esence o
independen ly associa ed a ian s in hese loci (Supplemen a y
Table 5). In o al, nine independen ly associa ed SNPs in se en
loci we e de ec ed o SDNN, nine independen ly associa ed SNPs
in eigh loci o RMSSD, and fi e independen ly associa ed SNPs
in fi e loci o p RSA/HF. Many o he SNPs we e associa ed wi h
a leas wo o he HRV ai s (Supplemen a y Da a 1). In ou
loci, he lead SNPs di e ed be ween ai s bu we e in linkage
disequilib ium (LD) wi h each o he (0.24o 2o0.90) (Table 1).
Fo es plo s show li le he e ogenei y in he gene ic associa ions
ac oss he en i e se o coho s o all SNPs (Supplemen a y
Fig. 4). Sex-s a ified analyses did no show di e ences in SNP
e ec s be ween men and women o he genome-wide associa ed
loci (Supplemen a y Table 6). Sepa a ely me a-analysing ac oss
coho s wi h sho labo a o y es eco dings e sus longe e m
ambula o y eco dings did no sugges sensi i i y o he esul s o
hese di e en eco ding me hods (Supplemen a y Table 7).
Resul s o VEGAS gene-based analyses co obo a ed hose o he
SNP-based analyses (Supplemen a y No e 1).
Va iance explained. Weigh ed gene ic isk sco es based on he
independen SNPs ha eached genome-wide significance a e
he second s age me a-analysis we e compu ed o he h ee HRV
ai s and used o p edic RMSSD, SDNN and p RSA/HF in
adul s om he Li elines (n¼12,101) and NESDA (n¼2,218)
coho s, adolescen s om he TRAILS-Pop coho (n¼1,191),
and child en om he ABCD coho (n¼1,094) (Table 2).
The mul i-SNP gene ic isk sco es we e all significan ly associa ed
wi h HRV and he pe cen ages o a iance explained o he
co esponding ai s we e 1.0–1.4% o SDNN, 1.1–2.4% o
RMSSD, and 0.9–2.6% o p RSA/HF. C oss- ai explained
a iances o gene ic isk sco es we e close o hose o he
co esponding ai .
To es he con ibu ion o SNPs ha did no each
genome-wide significance, we pe o med polygenic isk sco e
analyses using inc easingly mo e lenien significance h esholds
and de e mined he pe cen ages o explained HRV in he same
ou coho s (Supplemen a y Fig. 5; Table 3). Maximal a iance
explained by he polygenic isk sco e was 0.8–1.4% o SDNN,
0.9–2.3% o RMSSD and 0.9–2.3% o p RSA/HF. This was
eached a ela i ely small numbe s o SNPs ( 71) wi h
addi ional SNPs adding mo e noise han signal.
The o al a iance explained by common SNPs (SNP-based
he i abili y) es ima ed by Genomic Res ic ed Maximum
Likelihood o LD sco e eg ession analysis a ied be ween 10.8
and 13.2%, wi h only small di e ences in es ima es ac oss
me hods and HRV ai s (Supplemen a y No e 2).
Gene aliza ion o o he e hnici ies. In da a om up o 11,234
Hispanic/La ino indi iduals, fi e SNPs in fi e o he eigh loci
iden ified o RMSSD, se en SNPs in six o he se en loci o
SDNN and h ee SNPs in h ee o he fi e loci o p RSA/HF
showed a s a is ically significan associa ion ha was consis en in
di ec ion wi h he associa ion in indi iduals o Eu opean ances y
(Table 4). In da a om 6,899 A ican–Ame icans, ou SNPs om
ARTICLE NATURE COMMUNICATIONS | DOI: 10.1038/ncomms15805
2NATURE COMMUNICATIONS | 8:15805 | DOI: 10.1038/ncomms15805 | www.na u e.com/na u ecommunica ions
ou o he eigh loci we e associa ed wi h RMSSD, h ee SNPs in
h ee o he se en loci wi h SDNN and none wi h p RSA/HF. In
he combined me a-analysis in a maximum o 71,675 pa icipan s
om all e hnici ies, one SNP ( s6123471 on ch omosome 20) was
no longe significan (Table 4).
Co ec ing HRV o hea a e. The s ong in e se associa ion
be ween HRV and hea a e eflec s he well-es ablished
simul aneous biological e ec o ca diac agal ac i i y on hea
a e15 and HRV16, bu i also exp esses a ma hema ical
dependency o he a iance in in e bea in e al (IBI) on he
mean IBI ha is un ela ed o he unde lying biology. We
conduc ed h ee analyses o es whe he he associa ion o he
HRV SNPs was obus o co ec ion o he HRV ai s o hea
a e (Supplemen a y Table 8). Fi s , we used a ecen ly de eloped
analy ical echnique17 o ob ain he me a-analysis o he
coe ficien o a ia ion o SDNN and RMMSD om he
summa y s a is ics o he HRV and es ing hea a e
me a-analyses18. The coe ficien o a ia ion de ec s he amoun
o IBI a iabili y ela i e o he mean IBI o each subjec , and
deals wi h he p opo ionali y-based dependence o HRV on
hea a e19. Second, we es ablished he e ec o he 17 HRV
SNPs on he coe ficien s o a ia ion o SDNN and RMSSD in
he Li elines, NESDA and TRAILS-Pop coho s, and
me a-analysed he esul s. Thi d, we use a media ion analysis in
hese same coho s o see how much o he SNP e ec s on he
h ee HRV measu es was media ed by hea a e. In all h ee
analyses, we find some a enua ion o he HRV SNP associa ions.
The a e age media ion o he associa ion by hea a e was
B28%. Howe e , he co ec ion o hea a e le mos o he
HRV SNP associa ions in ac , pa icula ly in he fi s analysis ha
used he ull disco e y sample.
Associa ion o he HRV SNPs wi h es ing hea a e. Because
he HRV ai s eflec ca diac agal ac i i y, we expec ed he HRV
SNPs o ha e an e ec on es ing hea a e. We pe o med a
lookup o he 17 HRV SNPs in a GWAS me a-analysis on es ing
hea a e in 85,787 indi iduals18. Ou o he 17 HRV lead SNPs,
11 we e associa ed wi h hea a e a e co ec ing o mul iple
es ing (Supplemen a y Table 9, panel a). All e ec s we e in he
expec ed di ec ion such ha he HRV dec easing allele was
associa ed wi h highe hea a e (Supplemen a y Fig. 6). Six o
he HRV SNPs we e no significan ly associa ed wi h hea a e,
including ou op hi on ch omosome 19 ( s12974991 in
NDUFA11: p RMSSD ¼4.6 1046;phea a e ¼0.18).
Analysis o summa y s a is ics o he HRV and hea a e me a-
analyses as implemen ed in he g x R package showed ha mul i-
SNP gene ic isk sco es o HRV we e significan ly associa ed
wi h hea a e (Supplemen a y Table 9, panel b).
40
a
b
c
30
20
–log10 P alue
10
0
1
2
3
4
5
6
7
8
Ch omosome
PPIL1
PPIL1
PPIL1
GNG11
GNG11
GNG11
LINC00477
LINC00477
LINC00477
SYT10
SYT10
SYT10
RGS6
RGS6
NDUFA11
NDUFA11
NDUFA11
KIAA1755
HCN4
NEO1
9
10
11
12
13
14
15
16
17
18
19
20
21
22
1
2
3
4
5
6
7
8
Ch omosome
9
10
11
12
13
14
15
16
17
18
19
20
21
22
1
2
3
4
5
6
7
8
Ch omosome
9
10
11
12
13
14
15
16
17
18
19
20
21
22
40
30
20
–log10 P alue
10
0
40
30
20
–log10 P alue
10
0
Figu e 1 | Manha an plo s o he me a-analyses o s age 1 GWAS esul s. (a) SDNN, (b) RMSSD and (c) p RSA/HF in up o 28,700 indi iduals o
Eu opean ances y. Only SNPs wi h a mino allele equency 41% and ha we e p esen in a leas 1/3 o he sample a e plo ed. Significan loci a e shown
in blue, sugges i e ones in ed. The blue ho izon al line ep esen s he genome-wide significance h eshold. Genes closes o he lead SNPs a e indica ed o
he loci ha we e genome-wide significan ly associa ed wi h he ai a e he s age 1 þ2 combined me a-analysis.
NATURE COMMUNICATIONS | DOI: 10.1038/ncomms15805 ARTICLE
NATURE COMMUNICATIONS | 8:15805 | DOI: 10.1038/ncomms15805 | www.na u e.com/na u ecommunica ions 3
In addi ion, gene ic isk sco es based on he independen
genome-wide significan HRV SNPs om he combined s age 1
and 2 me a-analysis we e es ed o associa ion wi h hea a e in
he Li elines, NESDA, TRAILS-Pop and ABCD coho s
(Supplemen a y Table 9, panel b). The h ee mul i-SNP isk
sco es o he HRV ai s explained a small, bu mos ly significan
pe cen age o a iance in hea a e (0.09–1.13%). Polygenic
isk sco e analysis showed ha adding HRV SNPs below he
genome-wide significance h eshold did no u he inc ease he
a iance explained in hea a e (Supplemen a y Fig. 5;
Supplemen a y Table 9, panel c).
The e e se ques ion, whe he SNPs wi h e ec s on hea a e
a e associa ed wi h HRV, was also in es iga ed. The 21 hea a e
SNPs iden ified by he GWAS me a-analysis on hea a e18
explained be ween 0.2 and 0.9% o he a iance in he h ee HRV
ai s (Supplemen a y Table 10).
Associa ion wi h ca diome abolic ai s and diseases.
In addi ion o hea a e, we examined he associa ion o he 17
HRV-associa ed SNPs wi h o he confi med isk ac o s o
ca diac, me abolic and enal disease ai s and endpoin s using
da a om la ge-scale GWAS me a-analyses (Supplemen a y
Table 11). Mul i-SNP isk sco es we e compu ed based on ou
17 op SNPs and we es ed hei associa ion wi h he ou comes.
No e ec s o isk sco es using he 17 HRV SNPs we e obse ed
o sys olic o dias olic blood p essu e, body mass index,
enal unc ion, hea ailu e, sudden ca diac dea h, co ona y
a e y disease, a ial fib illa ion o ype 2 diabe es. Only o
a ial fib illa ion we obse ed indi idually significan SNPs. These
wo highly significan SNPs ( s10842383 nea LINC00477,
P¼3.45 107and s2680344 in HCN4,P¼4.34 107
(Supplemen a y Table 12) had la ge opposi e e ec s on a ial
fib illa ion, while bo h dec eased HRV. In addi ion o hese
lookups ha we e es ic ed o genome-wide significan SNPs, we
employed bi a ia e LD sco e eg ession20 ha uses he ull GWAS
summa y s a is ics o he HRV and ca diome abolic ai s and
diseases o compu e gene ic co ela ions. The gene ic co ela ions
sys ema ically poin ed o an o e lap in he gene ic a ian s
causing low HRV and inc eased isk o disease ( ha is, nega i e
Table 1 | S age 1 þ2 combined me a-analysis esul s o SDNN, RMSSD and p RSA/HF o loci ha we e genome-wide significan
(Po(5 108)/3) in he analysis o indi iduals o Eu opean ances y.
Locus Ch SNP Posi ion
(bp)
(build 36)
Closes gene Anno a ion T ai Allele S age 1 þ2
E/O NEAF b(s.e.) P alue
1 19 s12974991* 5845584 NDUFA11 IN RMSSD A/G 43,205 0.078 0.116(0.008) 4.57E-46
s12974440* 5845386 IN p RSA/HFwA/G 29,527 0.073 0.244(0.019) 1.91E-41
s12980262* 5844058 M SDNN A/G 46,046 0.076 0.060(0.006) 2.30E-23
2 12 s10842383 24663234 LINC00477 (C12o 67) IG, HRiSDNN C/T 47,808 0.863 0.049(0.004) 9.33E-31
RMSSD 43,223 0.862 0.065(0.006) 2.45E-29
p RSA/HFw31,085 0.865 0.124(0.013) 1.20E-25
3 6 s236349 36928543 PPIL1 IG SDNN G/A 51,379 0.651 0.033(0.003) 3.70E-25
RMSSD 46,795 0.655 0.035(0.004) 9.10E-17
p RSA/HFw33,654 0.645 0.069(0.009) 3.16E-15
4 12 s7980799z33468257 SYT10 IN, HRii RMSSD A/C 44,210 0.390 0.039(0.004) 3.19E-20
s1351682z33490042 IG, HRiii p RSA/HFwG/A 30,643 0.437 0.073(0.009) 5.70E-15
s1384598z33514166 IG, HRi SDNN T/A 47,358 0.432 0.023(0.003) 7.37E-13
5 7 s4262y93389364 GNG11 UTR5, Q, HR SDNN C/T 49,005 0.390 0.028(0.003) 4.26E-17
p RSA/HFw31,281 0.388 0.050(0.010) 1.84E-11
s180238y93388383 UP, Q, HR i RMSSD C/T 44,420 0.333 0.034(0.004) 7.99E-16
6 14b s4899412|| 71534015 RGS6 IN, Q SDNN T/C 48,252 0.253 0.026(0.004) 3.13E-13
s2052015|| 71556806 RMSSD T/C 45,492 0.165 0.036(0.006) 3.56E-10
14c s2529471 71883022 IN SDNN C/A 49,619 0.429 0.021(0.003) 1.88E-12
14a s36423 71422955 IG SDNN T/G 48,182 0.129 0.033(0.005) 6.25E-13
RMSSD 45,419 0.127 0.040(0.006) 5.36E-11
7 15a s2680344 71440538 HCN4 IN, HR ii SDNN A/G 51,370 0.777 0.024(0.004) 4.88E-11
15b s1812835 71294557 NEO1 IN, Q RMSSD A/C 44,421 0.418 0.025(0.004) 5.18E-10
8 20 s6123471 36273570 KIAA1755 UTR3, HR iii RMSSD T/C 46,789 0.534 0.024(0.004) 1.30E-08
Allele E/O, e ec allele/o he allele; bp, base pai posi ion based on build 36 (hg18); Ch , ch omosome; EAF, e ec allele equency; HR, HRV SNPs ha a e in pai wise LD (based on SNAP, HapMap
elease 22 CEU) wi h iden ified loci associa ed wi h hea a e (HR) om den Hoed e al.18; IG, in e genic a ian ; IN, in onic a ian ; N, sample size; M, missense a ian ; Q, associa ed wi h an eQTL;
s.e., s anda d e o o b; UTR3, a ian in he 30un ansla ed egion; UTR5, a ian in he 50un ansla ed egion; UP, ups eam a ian (wi hin 2kb); b, e ec size.
NOTE: Only SNPs ha we e independen ly associa ed ( ha is, lead SNPs) o he ai s a e shown. A some loci lead SNPs we e he same o he di e en ai s, a o he loci he e we e di e en
(dependen ) lead SNPs o he di e en ai s. SNPs a e so ed acco ding o P alue o he combined me a-analysis pe locus. Genome-wide significan associa ion ( wo-sided Po5108), co ec ed o
es ing h ee ai s ( ha is, Po510 8/3), is shown in bold. E ec alleles we e chosen o eflec an inc eased isk o low le els o HRV, hence b’s a e all nega i e.
i 2¼1 be ween s10842383 and s17287293[HR]; ii same SNP; iii 2¼0.782 be ween s1351682 and s7980799[HR]; i 2¼0.695 be ween s1384598 and s7980799[HR]; 2¼0.570 be ween
s4262and s180242[HR]; i 2¼0.893 be ween s180238 and s180242[HR]; ii 2¼0.505 be ween s2680344 and s4489968[HR]; iii 2¼1 be ween s6123471 and s6127471[HR].
* hese SNPs a e all in pe ec LD ( 2¼1).
wP alue, allele, EAF, N om P alue weigh ed me a-analysis o all coho s using METAL and b, s.e. om in e se- a iance me a-analysis o only HF coho s using GWAMA.
z 2¼0.782 be ween s7980799 and s1351682; 2¼0.695 be ween s7980799 and s1384598; 2¼0.903 be ween s1351682 and s1384598.
y 2¼0.600 be ween s4262 and s180238.
|| 2¼0.237 be ween s4899412 and s2052015.
ARTICLE NATURE COMMUNICATIONS | DOI: 10.1038/ncomms15805
4NATURE COMMUNICATIONS | 8:15805 | DOI: 10.1038/ncomms15805 | www.na u e.com/na u ecommunica ions
co ela ions wi h sys olic and dias olic blood p essu e, co ona y
a e y disease, hea ailu e, sudden ca diac dea h, BMI and ype 2
diabe es) compa ible wi h clinical ele ance o he HRV SNPs
iden ified, al hough significance was eached only o sys olic and
dias olic blood p essu e a e co ec ion o numbe o ou comes
es ed (Supplemen a y Table 11).
Po en ial unc ional impac o he HRV a ian s. To iden i y
unc ional a ian s agged by he 17 HRV SNPs, we pe o med
a ious pos -GWAS anno a ion (Supplemen a y Fig. 7). In silico
anno a ion (Supplemen a y Da a 2) showed ha he lead SNP o
SDNN on ch omosome 19 was a non-synonymous SNP
( s12980262 in NDUFA11) and ha he lead SNPs o RMSSD
( s12974991) and p RSA/HF ( s12974440) we e in pe ec LD
wi h his SNP (Table 1; Supplemen a y Da a 2). SNP s129080262
was cha ac e ized as dele e ious, wi h a so ing in ole an om
ole an (SIFT) sco e o 0.01 and a polymo phism pheno yping
(PolyPhen) sco e o 0.753 indica ing a possibly damaging e ec .
Func ional a ian analyses using RegulomeDB confi med ha
s12980262 and s12974440 in NDUFA11 on ch omosome 19
likely ha e unc ional consequences (Supplemen a y Table 13) by
binding o ansc ip ion ac o s o influencing he ch oma in
s a e. SNP s6123471 in he locus on ch omosome 20 was in
high LD wi h wo non-synonymous SNPs in he KIAA1755
gene ( s3746471 [ 2¼0.94] and s760998 [ 2¼0.55]) ha a e
p edic ed o yield ole a ed, benign amino acid changes
(Supplemen a y Da a 2).
We examined i he 17 HRV SNPs we e eQTLs in a la ge whole
blood da abase. Fou o he HRV SNPs we e significan ly ( alse
disco e y a e o5%) associa ed wi h gene exp ession in blood
(Supplemen a y Table 14): s1812835 wi h exp ession o NEO1,
s4899412 wi h exp ession o RGS6, and s180238 and s4262
wi h exp ession o GNG11. These ou SNPs we e all in s ong LD
wi h he op eQTL SNPs o hese genes ( 240.70) and los
significance a e condi ioning on he co esponding op eQTL.
The eQTLs o NEO1 and RGS6 we e eplica ed in a leas one
o he whole blood eQTL s udy (Supplemen a y Table 14). The
eQTL o GNG11 was eplica ed in he medulla (P¼2.8 104)
and he an e io ibialis a e y (P¼8.1 10 9). None o he 17
SNPs eached significance in a smalle hea eQTL da abase.
Nine o he 17 HRV SNPs we e in high LD ( 240.70) wi h
SNPs associa ed wi h me hyla ion le el o one o mul iple CpG
si es (me hyla ion quan i a i e ai loci [mQTLs]) in whole blood
(Supplemen a y Table 15). Two o he HRV SNPs ha we e
eQTLs also influenced me hyla ion o he same gene in whole
blood, s ongly sugges i e o a egula o y unc ion o hose SNPs.
eQTL s1812835 in NEO1 was associa ed wi h me hyla ion le el
o cg11357013, cg19281068, cg11552023 and cg17150474. eQTL
s4262 was associa ed wi h me hyla ion le el o cg08038054 and
cg06439941 in GNG11. The o he wo eQTL SNPs did no
achie e genome-wide significance le el o an associa ion wi h
me hyla ion, bu eQTL s4899412 in RGS6 was in high LD wi h a
p oxy SNP ( s2238280) ha was associa ed wi h me hyla ion le el
o cg19493789, which is loca ed in a CpG island shel nea RGS6.
Fi e o he HRV SNPs we e (in high LD wi h) mQTLs bu we e
no hemsel es eQTLs. Fo example, HRV SNPs s12974991,
s12974440 and s12980262 (ch omosome 19) we e associa ed
wi h me hyla ion le el o mul iple CpG si es (cg22854549,
cg03715305 and cg19211619) loca ed in o nea by NDUFA11,
bu we e no associa ed wi h exp ession le el o NDUFA11 in
whole blood. Such mQTLs may well exe a egula o y e ec on
NDUFA11 in o he issues. DEPICT issue en ichmen analysis
(Supplemen a y Da a 3; Supplemen a y Table 16, Supplemen a y
Fig. 8) showed NDUFA11 exp ession was weak in blood, bu
en iched in hea , senso y and endoc ine issues.
Discussion
This me a-analysis o GWAS o HRV yielded 17 lead SNPs
(11 independen ) in eigh loci ha we e genome-wide
significan ly associa ed, six o which gene alized o indi iduals
o A ican-Ame ican and Hispanic/La ino e hnici y. Va ious
ways ha co ec HRV o i s ma hema ical dependency on
es ing hea a e a enua ed he SNP e ec s, bu la gely le he
associa ions in ac . Toge he , he hi s in he eigh loci explained
0.9–2.6% o he a iance in es ing HRV in ou independen
coho s o Eu opean ances y. De ails o known biological
unc ions o he genes closes o hese loci a e gi en in he
Supplemen a y No e 6.
We no ed a s ong en ichmen o ou HRV loci in a
p e iously conduc ed me a-analysis o GWAS o es ing hea
a e18, a known isk ac o o ca diac mo bidi y and
mo ali y21,22. SNPs in fi e o he 21 es ing hea a e loci
( ha is, LINC00477 (C12o 67),SYT10,GNG11,HCN4 and
KIAA1755) we e associa ed wi h HRV a genome-wide
significance le el and six mo e a ained nominal significance,
wi h associa ions always in he expec ed di ec ion. Gene ic isk
Table 2 | Explained a iance in HRV ai s in he Li elines (n¼12,101), NESDA (n¼2,118), TRAILS-Pop (n¼1,191) and ABCD
(n¼1,094) coho s by he weigh ed mul i-SNP gene ic isk sco e based on he independen genome-wide significan SNPs in he
s age 1 þ2 me a-analysis.
Li elines NESDA TRAILS-Pop ABCDT ai Risk sco e
No. SNPs P alue DR2No. SNPs P alue DR2No. SNPs P alue DR2No. SNPs P alue DR2
SDNN SDNN 9 6.3E-33 1.00% 9 4.9E-10 1.39% 10 8.0E-05 1.28% 10 7.8E-04 1.03%
SDNN RMSSD 7 1.1E-30 0.93% 11 7.5E-10 1.35% 11 6.3E-06 1.69% 11 3.5E-05 1.56%
SDNN p RSA/HF 5 6.4E-25 0.75% 5 5.2E-07 0.89% 5 8.7E-07 1.99% 4 4.3E-03 0.75%
RMSSD SDNN 9 8.3E-37 1.13% 9 2.0E-11 1.54% 10 4.4E-06 1.73% 10 4.7E-04 1.11%
RMSSD RMSSD 7 8.8E-37 1.13% 11 6.1E-12 1.62% 11 5.3E-08 2.42% 11 2.5E-06 2.01%
RMSSD p RSA/HF 5 1.5E-30 0.93% 5 2.0E-11 1.54% 5 2.1E-09 2.92% 4 8.1E-04 1.02%
p RSA/HF SDNN 7 NA NA 9 1.5E-13 1.58% 10 4.3E-05 1.38% 10 2.0E-03 0.87%
p RSA/HF RMSSD 6 NA NA 11 7.1E-14 1.62% 11 1.3E-06 1.93% 11 1.5E-05 1.70%
p RSA/HF p RSA/HF 5 NA NA 5 7.7E-17 2.01% 5 1.4E-08 2.64% 4 1.8E-03 0.89%
NA, no a ailable.
NOTE:DR2is he di e ence in pe cen age o explained a iance by he mul i-SNP gene ic o polygene ic isk sco e be ween he models wi h and wi hou he isk sco e while adjus ing bo h o age, sex
and p incipal componen s.
Fo Li elines, NESDA and TRAILS-Pop he weigh s ( ha is, e ec s sizes) and numbe o genome-wide significan SNPs included in he isk sco e we e adjus ed by analy ically ex ac ing he coho ’s e ec
size and s.e. om he me a e ec size and s.e., espec i ely, and ecalcula ing he P alue based on hese adjus ed e ec sizes and s.e.’s, since hese coho s we e included in s age 1 and/o 2.
NATURE COMMUNICATIONS | DOI: 10.1038/ncomms15805 ARTICLE
NATURE COMMUNICATIONS | 8:15805 | DOI: 10.1038/ncomms15805 | www.na u e.com/na u ecommunica ions 5

sco es o HRV ai s we e also significan ly associa ed wi h
hea a e and LD sco e eg ession confi med ha he allelic
a ian s ha dec ease HRV in pa allel inc ease hea a e. This
sugges s ous ha pa o heHRVSNPsexe hei e ec on
hea a e h ough oscilla o y modula ion o pacemake ac i i y
by he agal ne es.
Supplemen a y Fig. 9 depic s he wo ou es by which
ace ylcholine eleased by he agal ne es in he sinoa ial node
is known o influence hea a e, bo h o which a e suppo ed
by ou esul s in GNG11,RGS6 and HCN4.Bybinding
hemusca inic ype2 ecep o (M
2
R) and dissocia ing he
G-p o ein he e o ime (Gabg)in oaGa
i/o
subuni and a Gbg
componen , ace ylcholine inhibi s he ongoing depola iza ion
o he pacemake cells by b1/b2-adenyla ecyclase ac i a ion o
unny (I
) channels and calcium channels23. In pa allel, i ac s
o ac i ely hype pola ize he pacemake cells by ac i a ion o
he GIRK1/4 channel. Each ou e accoun s o abou hal o he
onic dec ease in hea a e upon agal s imula ion23,bu
he esponse ime o M
2
R-GIRK e ec s on he sinus a e is
much sho e han o he M
2
R-HCN2/4 o he
b1/b2-adenyla ecyclase signalling pa hways. Only signalling
h ough he Gbg componen is as enough (B0.3 s) o apidly
ack changes in agal ou flow o he sinoa ial node, o
example, as hey occu wi hin he du a ion o a single
espi a ion (B4.5 s), whe eas signalling h ough he asubuni
is oo slow (43 s) o ack such phasic changes in ace ylcholine
elease1,24. GIRK signalling, he e o e, accoun s o mos o
HRV due o he phasic oscilla ion in agal ac i i y24,bu i
accoun s o only hal o he onic agal e ec s on hea a e.
The abo e leads o HRV only pa ially cap u ing he agal
e ec s on hea a e. Addi ional easons o he impe ec ela ion
be ween HRV and agal e ec s on hea a e1,2 a e indi idual
di e ences in: (i) es ing espi a ion a e and dep h; (ii) he
ampli ude o he in insic 0.1 Hz oscilla ions ela ed o bo h agal
Table 3 | Explained a iance in HRV ai s in he Li elines (n¼12,101), NESDA (n¼2,118), TRAILS-Pop (n¼1,191) and ABCD
(n¼1,094) coho s by he op imal polygenic isk sco es compu ed a he P alue h eshold ha explained he la ges pe cen age
o pheno ypic a iance.
T ai Risk sco e Coho Pcu o No. SNPs P alue DR2
SDNN SDNN Li elines o5E-7 13 6.8E-27 0.82%
NESDA o5E-8 6 2.6E-08 1.16%
TRAILS-Pop o5E-5 64 1.1E-04 1.23%
ABCD o5E-5 71 9.4E-05 1.39%
SDNN RMSSD Li elines o5E-8 8 2.4E-23 0.71%
NESDA o5E-6 23 1.2E-07 1.05%
TRAILS-Pop o5E-8 8 1.2E-04 1.23%
ABCD o5E-7 13 2.8E-06 2.00%
SDNN p RSA/HF Li elines o5E-8 7 3.1E-19 0.58%
NESDA o5E-8 4 3.5E-05 0.64%
TRAILS-Pop o5E-7 6 9.7E-06 1.61%
ABCD o5E-5 67 9.2E-04 1.01%
RMSSD SDNN Li elines o5E-7 13 8.9E-31 0.95%
NESDA o5E-8 6 1.6E-10 1.46%
TRAILS-Pop o5E-8 7 8.3E-06 1.63%
ABCD o5E-5 71 1.6E-04 1.30%
RMSSD RMSSD Li elines o5E-7 12 2.8E-30 0.94%
NESDA o5E-7 10 2.7E-10 1.43%
TRAILS-Pop o5E-7 11 3.4E-07 2.13%
ABCD o5E-7 13 3.8E-07 2.34%
RMSSD p RSA/HF Li elines o5E-8 7 1.4E-25 0.78%
NESDA o5E-8 4 3.6E-09 1.25%
TRAILS-Pop o5E-7 6 3.7E-08 2.47%
ABCD o5E-8 67 8.4E-04 1.02%
p RSA/HF SDNN NESDA o5E-8 6 1.1E-12 1.52%
TRAILS-Pop o5E-8 7 5.0E-05 1.36%
ABCD o5E-5 71 5.4E-04 1.09%
p RSA/HF RMSSD NESDA o5E-7 10 5.6E-14 1.69%
TRAILS-Pop o5E-7 11 3.3E-06 1.78%
ABCD o5E-7 13 1.9E-06 2.06%
p RSA/HF p RSA/HF NESDA o5E-8 4 4.4E-13 1.58%
TRAILS-Pop o5E-7 6 1.6E-07 2.25%
ABCD o5E-5 67 1.6E-03 0.90%
NA, no a ailable.
NOTE: Weigh ed polygenic isk sco e was de e mined based on independen SNPs in he s age 1 me a-analysis. Fo NESDA and TRAILS-Pop he weigh s ( ha is, e ec s sizes) and P alues we e adjus ed
by analy ically ex ac ing he coho ’s e ec size and s.e. om he me a e ec size and s.e., espec i ely, and ecalcula ing he P alue based on hese adjus ed e ec size and s.e., since hese coho s we e
included in s age 1.
ARTICLE NATURE COMMUNICATIONS | DOI: 10.1038/ncomms15805
6NATURE COMMUNICATIONS | 8:15805 | DOI: 10.1038/ncomms15805 | www.na u e.com/na u ecommunica ions
and sympa he ic blood p essu e egula ion h ough he ba o eflex
loops; (iii) mechano ansduc ion o in acellula pa hways
s imula ed by sinoa ial s e ch o (i ) he e ficiency o he
ac ual agal ga ing p ocess. These p ocesses can ha e a s ong
impac on HRV, bu less so on mean hea a e. We ound six
SNPs in ou loci, including ou op hi ( s12974991 in
NDUFA11), ha may ac on he indi idual di e ences in hese
p ocesses as hey had no disce nible e ec on hea a e, in spi e
o hei significan impac on HRV.
The genome-wide significan SNPs in GNG11,RGS6 and NEO1
we e eQTLs and in s ong LD wi h he op mQTLs and eQTLs o
he co esponding genes. Two o hese (GNG11,RGS6) eadily
p o ide a biological hypo hesis o accoun o he associa ions
de ec ed in he me a-analysis. The C alleles o s4262 and
s180238 o GNG11 coding o he g11 subuni o he
he e o ime ic G-p o ein complex Gabg cause dec eased
exp ession o his subuni and we e associa ed wi h lowe HRV.
The e ec s o he GNG11 eQTLs associa ed wi h lowe HRV a e
likely o lowe he a ailabili y o he g11 subuni , he eby
educing Gbg componen -induced GIRK ac i a ion. This
po en ially blun s he hea a e change in esponse o he
oscilla o y changes in ca diac agal ac i i y.
The egula o o he e o ime ic G-p o ein complex signalling,
ype 6 (RGS6) gene on ch omosome 14 was ound o be linked
o h ee independen signals o SDNN and RMSSD. RGS6 ac s as
a c i ical nega i e egula o o M
2
R signalling in he sinoa ial
node o he hea apidly e mina ing Gbg signalling and hus
cu ailing agal lowe ing o he hea a e25,26. The esul s o ou
me a-analysis a e consis en wi h a ole o RGS6 in dec easing
HRV p e iously hin ed a by animal expe imen a ion23,27 and a
human case epo 27,28. The T allele o ou eQTL RGS6 SNP
( s4899412) causes inc eased exp ession o RGS6. By inc easing
RGS6 exp ession, he T allele ac s as a gain-o - unc ion mu a ion
ha gi es ise o a dec ease in GIRK-channel signalling
and he obse ed dec ease in HRV. O no e, Rgs6/mice,
ha show he expec ed inc ease in HRV, a e cha ac e ized by
a s ong b adyca dia and an inc eased suscep ibili y o AV
block and a ial fib illa ion which is a ibu ed o an enhancemen
o GIRK-induced sinoa ial and a io en icula node
hype pola iza ion by emo ing he nega i e egula ion o Gbg
by RGS6 ( e s 23,26,28).
The associa ion o he s2680344 SNP in HCN4 is puzzling
because HCN signalling does no in ol e he as M
2
R-GIRK
channels and canno ansla e apid agal fluc ua ion in o
bea - o-bea a ia ion in IBI leng h, ha is, HRV. The e ec o
he HCN4 SNP on HRV may be seconda y o i s e ec s on he
a e age slope o he dias olic depola iza ion29. The HCN4 p o ein
is a key componen o he I
channel30–32 ha gene a es he
pacemake po en ial by a g adual depola iza ion o he sinoa ial
myocy e cell memb ane du ing dias ole. This ‘pacemake
depola iza ion’ phase is known o be slowed by loss-o - unc ion
mu a ions in he HCN4 ha lead o lowe hea a e31 and he
I
is he known si e o ac ion o i ab adine and o he he apeu ic
agen s used o slow hea a e in angina pa ien s32. O no e, bo h
i ab adine ea men 33 and loss-o - unc ion mu a ions inc ease
he isk o a ial fib illa ion34. In con as , gain-o - unc ion
mu a ions in he sensi i i y o HCN4 o cAMP lead o highe
hea a e30. This leads us o hypo hesize ha he A allele o
s2680344 in HCN4 ei he is i sel a gain-o - unc ion mu a ion o
ags such a mu a ion because i inc eases hea a e18.
High HRV is associa ed wi h lowe mo bidi y and mo ali y in
pa ien s wi h ca dio ascula disease5, hype ension7, end-s age
Table 4 | Me a-analysis esul s o he iden ified loci in o he e hnici ies and combined me a-analysis esul s wi h Eu opean
ances y.
Locus Ch SNP T ai Allele Hispanic/La ino A ican Ame ican EUR þHIS þA Am
E/O NEAF b(s.e.) P alue NEAF b(s.e.) P alue P alue
1 19 s12974991 RMSSD A/G 11,233 0.065 0.162 (0.018) 7.05E-20 6,673 0.455 0.077 (0.033) 1.10E-02 1.86E-63
s12974440 p RSA/HF* A/G 404 0.048 0.518 (0.174) 3.06E-03 1900 0.019 0.189 (0.158) 3.41E-01 4.53E-41
s12980262 SDNN A/G 11,233 0.048 0.161 (0.070) 1.04E-02 6675 0.093 0.046 (0.030) 6.48E-02 1.57E-24
2 12 s10842383 SDNN C/T 11,233 0.854 0.053 (0.012) 2.45E-06 6676 0.955 0.056 (0.026) 9.83E-01 7.61E-33
RMSSD 11,233 0.854 0.064 (0.012) 1.38E-07 6673 0.955 0.065 (0.030) 9.86E-01 4.23E-32
p RSA/HF* 404 0.830 0.140 (0.095) 1.40E-01 1901 0.959 0.068 (0.104) 6.79E-01 4.98E-25
3 6 s236349 SDNN G/A 11,234 0.684 0.034 (0.009) 6.15E-05 6676 0.724 0.017 (0.011) 6.57E-02 1.76E-28
RMSSD 11,234 0.684 0.034 (0.009) 1.67E-04 6673 0.724 0.021 (0.013) 4.79E-02 5.88E-20
p RSA/HF* 404 0.704 0.164 (0.080) 4.13E-02 1901 0.729 0.004 (0.043) 4.87E-01 4.64E-15
4 12 s7980799 RMSSD A/C 11,234 0.269 0.031 (0.010) 1.23E-03 6488 0.097 0.029 (0.021) 7.70E-02 1.57E-22
s1351682 p RSA/HF* G/A 404 0.348 0.166 (0.077) 3.19E-02 1901 0.142 0.082 (0.058) 6.91E-02 2.00E-14
s1384598 SDNN T/A 11,234 0.307 0.026 (0.009) 1.80E-03 6676 0.146 0.024 (0.015) 5.41E-02 2.88E-15
5 7 s4262 SDNN C/T 11,234 0.427 0.016 (0.008) 2.39E-02 6676 0.608 0.028 (0.011) 5.87E-03 5.36E-19
p RSA/HF* 404 0.410 0.014 (0.074) 8.46E-01 1901 0.618 0.055 (0.043) 1.15E-01 1.50E-11
s180238 RMSSD C/T 11,234 0.367 0.024 (0.009) 4.05E-03 6673 0.474 0.032 (0.011) 2.77E-03 8.07E-19
6 14b s4899412 SDNN T/C 11,234 0.329 0.012 (0.009) 8.27E-02 6676 0.419 0.009 (0.010) 1.86E-01 5.96E-13
s2052015 RMSSD T/C 11,234 0.173 0.015 (0.012) 9.94E-02 6673 0.098 0.001 (0.020) 4.83E-01 1.94E-09
14c s2529471 SDNN C/A 11,233 0.485 0.018 (0.008) 1.38E-02 6676 0.543 0.003 (0.010) 3.83E-01 2.08E-12
14a s36423 SDNN T/G 11,234 0.193 0.021 (0.011) 2.41E-02 6676 0.160 0.030 (0.015) 1.79E-02 1.60E-14
RMSSD 11,234 0.193 0.017 (0.011) 7.05E-02 6673 0.160 0.034 (0.016) 1.72E-02 1.02E-11
7 15a s2680344 SDNN A/G 11,234 0.681 0.005 (0.009) 2.97E-01 6676 0.450 0.024 (0.011) 1.32E-02 2.90E-11
15b s1812835 RMSSD A/C 11,234 0.426 0.012 (0.009) 8.83E-02 1388 0.140 0.009 (0.033) 3.98E-01 5.30E-10
8 20 s6123471 RMSSDwT/C 11,234 0.560 0.001 (0.009) 4.40E-01 6673 0.739 0.020 (0.013) 5.67E-02 5.14E-06
A Am, A ican Ame ican; Allele E/O, e ec allele/o he allele; Ch , ch omosome; bp, base pai posi ion based on build 36 (hg18); EAF, e ec allele equency; EUR, Eu opean; HIS, Hispanic/La ino;
N, sample size; s.e., s anda d e o o b;b, be a/e ec size.
NOTE: SNPs so ed as in Table 1 acco ding o he Eu opean ances y combined me a-analysis P alue pe locus. Significan Ps a e shown in bold (see ex o c i e ia). E ec alleles we e chosen o eflec
an inc eased isk o low le els o HRV, hence b’s a e all nega i e.
*P alue, allele, EAF, N om z-sco e weigh ed me a-analysis o all coho s using METAL and b, s.e. om in e se- a iance me a-analysis o only HF coho s using GWAMA.
wbo pa icipan s o Eu opean ances y di e s significan ly om ha o pa icipan s om A ican-Ame ican (di b¼0.044, P¼0.0012) o Hispanic/La ino ances y (di b¼0.023, P¼0.0195).
NATURE COMMUNICATIONS | DOI: 10.1038/ncomms15805 ARTICLE
NATURE COMMUNICATIONS | 8:15805 | DOI: 10.1038/ncomms15805 | www.na u e.com/na u ecommunica ions 7
enal disease8and diabe es9, bu also in appa en ly heal hy
indi iduals11,12. Using LD sco e eg ession on me a-GWAS
summa y s a is ics om a ious isk ac o s and endpoin s
we find some e idence o o e lap in he gene ic a ian s causing
low HRV and inc eased isk o disease, bu significance was
eached only o sys olic and dias olic blood p essu e a e
co ec ion o mul iple ou comes es ed. These gene ic
co ela ions a e compa ible wi h causal e ec s o ca diac agal
con ol in he ae iology o disease, bu hey could also be asc ibed
o e e sed causali y, whe e he disease p ocess leads o lowe
ca diac agal con ol. A s eng h o his s udy in his ega d is ha
analyses we e confined o indi iduals in good ca diac heal h, ha
is, coho s excluded pa ien s wi h exis ing ca dio ascula diseases
o medica ion po en ially impac ing HRV. Because we selec ed
indi iduals in good ca diac heal h e e se e ec s o disease on
HRV seem less likely, al hough some la en pa hology could ha e
been p esen . Howe e , an al e na i e explana ion ha is ha de
o ule ou is ha he gene ic co ela ion de i es om pleio opic
e ec s o gene ic a ian s common o bo h ou comes.
Fu he s eng hs o his s udy we e he consis ency o esul s
ac oss he di e en HRV ai s used o cap u e ca diac agal
con ol and he gene aliza ion o he HRV SNP e ec s o di e en
ances ies, in spi e o known e hnic di e ences in absolu e es ing
HRV35. Resul s also held in men and women sepa a ely and
ac oss a e y la ge ange o mean coho ages spanning om ea ly
childhood o he la e middle ages; in spi e o a s ong educ ion in
HRV alues wi h aging36. Al hough e ec s o age and sex on
HRV we e aken in o accoun in he analyses, many o he ac o s
we e no . The ideal design would ha e co ec ed o he known
e ec s o espi a ion dep h and a e on HRV, which a e
independen o agal ac i i y37. These could no be added as
co a ia es because hey we e no a ailable in mos coho s. We
we e libe al in excluding o he co a ia es like BMI, smoking and
exe cise in he GWAS analyses. These ai s a e subs an ially
he i able hemsel es and adjus ing o he i able co a ia es can
bias he genome-wide associa ion e ec s38 o e en induce
non-exis ing associa ions h ough collide bias39. Finally,
ins uc ions on p e-ECG eco ding beha iou s like physical
ac i i y, and ca eine, alcohol o nico ine use we e no
igo ously s anda dized ac oss coho s.
Di ec clinical ele ance o mos cu en GWAS findings is s ill
low and ou s udy is no excep ion. Po en ial u u e clinical use o
ou findings hinges on he abili y o ou gene ic a ian s o
cap u e (sub)co ical, b ains em and medulla y ansmission o
onic agal ac i i y o he sinoa ial node, no jus he impac o
ha ac i i y on hea a e. Subco ical gene a ion o onic agal
ac i i y is an impo an bioma ke o ca dio ascula heal h and
po en ially modifiable by in e en ions on psychosocial s ess40
and li es yle habi s41. I can e en be a ansdiagnos ic bioma ke
o psychopa hology and execu i e cogni i e unc ioning possibly
by eflec ing he in eg i y o p e on al co ex unc ioning42.
Gene ic ma ke s o HRV may p o e use ul as ins umen al
a iables in Mendelian andomiza ion43 o es causal hypo heses
on he e ec s o cen ally gene a ed agal ac i i y on beha iou al
and heal h ou comes.
In conclusion, his me a-analysis de ec s a c i ical ole o
gene ic a ia ion in Gbg and HCN signalling in explaining
indi idual di e ences in HRV. The HRV a ian s de ec ed can
help guide u he in es iga ions o he unc ional consequences
and po en ial he apeu ic implica ions o indi idual di e ences in
sinoa ial Gbg signalling.
Me hods
S udy coho s.App op ia e IRB app o al and in o med consen om pa icipan s
in all pa icipa ing coho s was ob ained. Full in o ma ion on consen p ocedu es
and de ails o he IRB boa ds a e p o ided in he Supplemen a y No e 8.
HRV measu emen .In his s udy, we in es iga ed h ee HRV ai s: SDNN, he
RMSSD and p RSA o HF. SDNN and RMSSD we e de i ed om he IBI ime
se ies ob ained om he R wa es in he ECG4. HF was calcula ed om Wa ele o
Fou ie decomposi ion wi h powe ob ained om a high equency band o ei he
0.15–0.40 Hz o 0.15–0.50 Hz. A ime domain measu e o RSA was de i ed by
p RSA using a espi a o y signal co- egis e ed wi h he ECG. Es ima es o p RSA
a e ob ained by sub ac ing he sho es IBI du ing hea a e accele a ion in he
inspi a ion phase om he longes IBI du ing hea a e decele a ion in he
expi a ion phase.
HRV ai s we e ex ac ed om he IBI ime se ies p e e ably based on
2–10 min pe iods o ECG in a s anda dized se ing, a es and in a si ing/supine
posi ion. I ambula o y da a we e a ailable, we ad ised coho s o ex ac a pe iod
o si ing s ill in he e ening, when his p o ed easible. Supplemen a y Table 2 lis s
he ac ual way HRV was assessed by he pa icipa ing coho s. Fo he coho s
analysed in s age 2, we ex ended ou HRV measu emen s o include coho s wi h
10 s and/o 20 s ECG eco dings, as RMSSD and SDNN based on hese ul a-sho
eco dings ha e shown a good ag eemen wi h 4–5 min eco dings3. Fu he mo e,
since IBI ime se ies equi e eliable de ec ion o he R-wa e only, a h ee-lead ECG
was conside ed su ficien while he use o mo e leads was encou aged. Fo p RSA,
an addi ional espi a ion signal o su ficien quali y o de ec beginning and end o
inspi a ion and expi a ion was needed.
SDNN and RMSSD ha e p e ailed in epidemiological s udies because hey a e
mo e easily assessed in la ge coho s and, as no ed abo e, can be ob ained e en
om sho ECG eco dings. HF and p RSA we e a ailable in ewe coho s, bu
hey be e eflec he ca dio espi a o y coupling ha d i es he oscilla o y
modula ion o agal e ec s in he sinoa ial node. In he ypical es ing espi a o y
equency ange, hese ime- and equency-domain measu es o RSA a e much less
con amina ed by oscilla ions in ca diac sympa he ic con ol han SDNN (and o he
measu es o HRV ha span a b oade equency ange). This is due o he empo al
dynamics o he sinoa ial node signalling pa hway ha ac s as a low pass fil e
allowing only oscilla ions in agal e ec s o ansla e in o HRV, whe eas o
sympa he ic e ec s o agal e ec s a p og essi ely highe espi a o y equencies
he node ac s as a leaky in eg a o causing mo e onic changes in hea a e1. Phasic
modula ion o agal e ec s is he e o e cap u ed mos pu ely by p RSA o HF.
Because p RSA and HF a e concep ually simila and highly co ela ed wi h each
o he ( 40.80) ac oss a wide ange o alues o espi a ion and hea a es44,
we g ouped he analyses on p RSA and HF unde he label p RSA/HF.
S udy popula ion.Coho s ha had da a on a leas one o he h ee HRV ai s
and genome-wide da a we e in i ed o pa icipa e in he fi s (disco e y) s age o
he Gene ic Va iance in Hea Ra e Va iabili y (V
g
HRV) conso ium. The s age 1
disco e y analysis was pe o med in up o 28,700 indi iduals o Eu opean ances y
om a maximum o 20 coho s. Independen coho s wi h ei he genome-wide o
gene-cen ed a ay da a o wi h he abili y o pe o m we -lab geno yping on he
single-nucleo ide polymo phism (SNPs) aken o wa d om he fi s s age we e
included in he second ( eplica ion) s age. This s age included addi ional da a om
up o 24,474 indi iduals om 11 coho s o Eu opean ances y (see Supplemen a y
Tables 1–4 o coho desc ip ions and de ails).
Associa ion analysis: s age 1 (disco e y).The ollowing exclusion c i e ia we e
applied a p io i: (1) indi iduals wi h hea disease ( o example, angina, pas
myoca dial in a c ion, le en icula ailu e) and (2) indi iduals known o use
an idep essan s (pa icula ly icyclic an idep essan s) and all an icholine gic
agen s ( o example, digoxin, a opine and ace ylcholines e ase inhibi o s) because
o he s ong e ec s ha hese d ugs ha e on HRV. Indi iduals epo ing o e he
coun e use o an icholine gic agen s we e no excluded.
Impu a ion o SNPs was done o ex end and c ea e simila SNP da abases
be ween coho s using di e en geno yping pla o ms. Mos o he coho s used he
HapMap Phase II elease 22 CEU panel as e e ence, bu la e eleases ( o example,
elease 24) o o he e e ence da a se s ( o example, 1000 Genomes) we e also used
(Supplemen a y Table 4).
Each coho pe o med linea eg ession analyses on all a ailable HRV ai s
using an addi i e SNP model adjus ing o age a he ime o ECG eco ding,
sex, p incipal componen s— o adjus o popula ion s a ifica ion—and o he
s udy-specific pa ame e s; all HRV ai s we e log- ans o med because o he
skewness o hei dis ibu ions. Only au osomal associa ions we e examined.
Analyses we e pe o med o all indi iduals as well as o men and women
sepa a ely.
S age 1 me a-analysis.Be o e me a-analysis, quali y con ol o all uploaded
coho files was pe o med using he QCGWAS package45. In case o issues he
coho s we e no ified and p oblems we e sol ed. Using he QCGWAS esul s,
specific impu a ion quali y and allele equency h esholds we e se o each coho .
An in e se- a iance, fixed-e ec s me a-analysis was pe o med o RMSSD and
SDNN o which SNPs o he di e en coho s we e me ged based on s-id. Fo
p RSA/HF, we pe o med a sample size weigh ed me a-analysis using z-sco es wi h
METAL46, since we combined esul s o wo HRV pheno ypes (p RSA and HF)
ha ha e di e en uni s and anges, and he e o e incompa able SNP e ec sizes.
To ge an idea o he size o he SNP e ec on p RSA/HF, we ob ained e ec sizes
ARTICLE NATURE COMMUNICATIONS | DOI: 10.1038/ncomms15805
8NATURE COMMUNICATIONS | 8:15805 | DOI: 10.1038/ncomms15805 | www.na u e.com/na u ecommunica ions
and s.e.’s om an addi ional fixed-e ec me a-analysis on he GWAS esul s o he
(majo i y o ) coho s ha measu ed HF. Resul s o he me a-analyses we e double
genomic con ol co ec ed47 o con ol o po en ial infla ion as a esul o
popula ion s a ifica ion wi hin and be ween coho s. The esul s included all SNPs
ha me he ollowing selec ion c i e ia: (a) a mino allele equency in he me a-
analysis o 41%, and (b) p esen in a leas one hi d o he coho s. This esul ed
in 2,555,913 SNPs being analysed o SDNN, 2,534,714 SNPs o RMSSD and
2,628,894 SNPs o p RSA/HF. Fo each ai sepa a ely, SNPs wi h a Po110 6
we e clumped o LD using pai wise LD checking in SNAP48 o asce ain
independen p ima y and seconda y signals ( 2o0.1). A o al o 23 lead SNPs in 14
loci we e selec ed o ollow-up in he second ( eplica ion) s age.
S age 2 me a-analysis.S age 2 coho s applied he same exclusion c i e ia and
pe o med he same associa ion analysis as in he disco e y s age, bu analyses we e
es ic ed o he 23 lead SNPs. I a SNP was no a ailable in a coho , he bes
a ailable p oxy was used ins ead based on s onges LD acco ding o he 1000
Genomes da abase. To e i y homogenei y o he esul s in he s age 2 coho s wi h
hose in he s age 1 coho s, he s age 1 me a-analysis e ec sizes o he 23 SNPs
we e co ela ed o he e ec sizes ob ained in each coho o each o he HRV
ai s. I a nega i e co ela ion ( o0) was ound, he coho / ai pai was excluded
om s age 2 analysis. Fo his eason esul s om one coho o SDNN we e
excluded. The eplica ion esul s we e hen me a-analysed pe ai using an
in e se- a iance fixed-e ec s me a-analysis o RMSSD and SDNN and a sample
size Pweigh ed me a-analysis using z-sco es in METAL46 o p RSA/HF. SNPs
we e ma ched based on s-id. Nex he associa ion esul s om bo h s ages we e
combined in he same way. A SNP was only conside ed o be significan ly
associa ed o HRV i i sa isfied he ollowing c i e ia: (1) i had Po1106in
s age 1, (2) i had a one-sided Po0.05 in he s age 2 me a-analysis cong uen wi h
he di ec ion o e ec in he s age 1 me a-analysis and (3) i had a genome-wide
significan Po5108/3 ( wo-sided) in he combined me a-analysis o s age 1
and 2 esul s, co ec ing o he es ing o h ee sepa a e ai s.
Condi ional analysis.In he disco e y s age independen SNPs we e selec ed
o ollow-up based on LD clumping ( 2o0.1). To confi m independence be ween
hese SNPs wi hin he loci on ch omosome 14 and 15, we applied he
condi ional-and-join analysis as implemen ed in he Genome-wide Complex T ai
Analysis so wa e package49 o he s age 1 summa y s a is ics o RMSSD and
SDNN wi h he geno ype da a o he NESDA coho 50 o 1,925 indi iduals as he
LD e e ence da a se . In addi ion, coho -le el indi idual da a on log- ans o med
RMSSD and SDNN o 12,101 indi iduals om he Du ch Li elines coho 51 we e
analysed using linea eg ession analysis wi h age and sex as co a ia es condi ioned
on he o he associa ed SNP(s) wi hin he locus.
Gene-based associa ion analysis (VEGAS).We pe o med gene-based es ing
wi h he ull se o B2.5 M HapMap SNPs om GWAS esul s o all h ee
pheno ypes, using VEGAS (Supplemen a y Table 17). This so wa e has he
ad an age o accoun ing o LD s uc u e and he possibili y o define a ange
beyond he gene bounds o include p omo e , 50UTR, in onic and 30UTR egions
in o he analysis. We defined a 50 kb ex a window beyond he genes, conside ed
e e y SNP in his window o he gene-based analysis, and an he analyses pe
ch omosome wi h up o 106pe mu a ions. A Po2.5 10 6(¼0.05/B20,000
genes) was conside ed as he h eshold o significance.
Va iance explained.The Li elines and NESDA coho s we e used o gene ic isk
sco e and polygenic isk sco e analyses o de e mine he pe cen age o a iance
explained by independen HRV SNPs ha we e genome-wide significan , and by
SNPs mee ing inc easingly lenien significance h esholds, espec i ely. Li elines
and NESDA ep esen examples o a popula ion-based coho and a coho
asce ained on case-con ol s a us ( o majo dep essi e diso de ). Bo h ec ui ed
adul pa icipan s. To es he s abili y o explained a iance ac oss he li e span, we
epea ed his analysis in wo o he Du ch coho s, he adolescen TRAILS-Pop
coho 52 (age 10–18) and he ABCD coho consis ing o young child en (age 5–7)
( e . 53).
Fo he gene ic isk sco e, s age 1 þ2 summa y s a is ics we e used o he
selec ion o HRV SNPs. No co ec ion was needed o ABCD as geno yping in his
coho had finished only a e comple ion o he me a-analyses. Howe e , he
NESDA coho had been included in bo h s age 1 and 2, TRAILS-Pop in s age 1,
and Li elines in s age 2, so he e ec sizes and s.e.’s o he HRV SNPs we e
co ec ed o sub ac he e ec s o hose coho s o ob ain independen alida ion
coho s54. Also, only SNPs we e used in he gene ic isk sco e i hey emained
genome-wide significan a e analy ically sub ac ing hese coho ’s e ec s om
he me a-analysis. Gene ic isk sco es o he emaining SNPs (Li elines: SDNN(9),
RMSSD(7), p RSA/HF(5); NESDA: SDNN(9), RMSSD(11), p RSA/HF(5);
TRAILS-Pop: SDNN(10), RMSSD(11), p RSA/HF(5)) weigh ed by he adjus ed
e ec size we e calcula ed o he pa icipan s o all ou coho s and eg essed on
he h ee HRV ai s (p RSA/HF was no a ailable in Li elines). Explained a iance
was compu ed as he change in R2 om a model wi h and wi hou he gene ic isk
sco e, while adjus ing bo h o age, sex and p incipal componen s.
To compu e he polygenic isk sco es, he impu ed geno ypes we e fi s
con e ed o bes -guess geno ypes. This was done ega dless o he impu a ion
quali y, since i was p e iously shown ha e en low-quali y SNPs migh con ibu e
o he a iance explained by SNPs ( e . 54). The SNP se was u he p uned o LD
using P io i yP une (h p://p io i yp une .sou ce o ge.ne /) o selec independen
SNPs, aking he significance o he SNP in he disco e y me a-analysis o each o
he HRV ai s in o accoun . This p o ided h ee LD-p uned SNP se s. Polygenic
isk sco es we e hen calcula ed in PLINK55 using significance h esholds o
5108,5107,5106,5105,5104, 0.005, 0.05, 0.5 and 1 and
associa ed wi h he h ee HRV ai s and es ing hea a e in he Li elines, NESDA,
TRAILS-Pop and ABCD coho s. Fo NESDA and TRAILS-Pop p uning and
polygenic isk sco e analysis was based on analy ically co ec ed esul s, since hese
coho s we e pa o s age 1 o ou s udy54.
He i abili ies and gene ic co ela ions.We applied genomic es ic ed
maximum likelihood analysis implemen ed in he Genomic Complex T ai
Analysis so wa e package56 in he Li elines coho (Supplemen a y Table 18) o
es ima e he pe cen ages o addi i e pheno ypic a iance ha can be explained by
common SNPs ( ha is, common SNP he i abili y deno ed as h2
SNP
). Fo his
analysis, SNPs om he HapMap Phase 3 p ojec we e selec ed o ob ain a se o
independen SNPs. We u he used LD sco e eg ession o es ima e he
he i abili ies o he h ee HRV ai s and he gene ic co ela ion among HRV ai s
and wi h hea a e20. The GWAS me a-analysis summa y s a is ics o RMSSD,
SDNN and p RSA/HF we e ob ained om s age 1 o he cu en s udy, and he
GWAS me a-analysis summa y s a is ics o hea a e om he disco e y s age o a
ecen GWAS me a-analysis o hea a e18. The LD sco es equi ed by he me hod
we e compu ed using 1000 Genomes da a o Eu opeans. The he i abili ies o he
h ee HRV measu emen s we e es ima ed using he uni a ia e model o his
me hod. C oss-pheno ype LD sco e eg ession analysis was pe o med using he
LDSC ool (LD SCo e) o es ima e gene ic co ela ions be ween pai s o
pheno ypes20.
In addi ion, we used he Oman Family S udy57 o pe o m uni a ia e and
bi a ia e analyses in fi e mul igene a ional highly inb ed pedig ees o es ima e he
he i abili ies o and he gene ic co ela ions be ween log- ans o med RMSSD,
SDNN, HF and hea a e using SOLAR ( 7.2.5) ( e . 58).
Gene aliza ion o o he e hnici ies.We u he examined he gene aliza ion o
loci iden ified a e me a-analysis o s age 1 and 2 esul s o o he e hnici ies using
da a om 11,234 indi iduals o wo Hispanic/La ino coho s, and 6,899 indi iduals
om fi e A ican-Ame ican coho s (Supplemen a y Tables 1–4). S age 3
me a-analyses we e pe o med in he same way as in s age 2 o his s udy o assess
he e ec o he HRV-associa ed SNPs in indi iduals o Hispanic/La ino and
A ican-Ame ican ances y, in he combined se o Eu opean and Hispanic/La ino
ances y, in he combined se o Eu opean and A ican-Ame ican ances y, and in
all h ee e hnici ies combined. He e, we applied he same c i e ia o significance as
in s age 2 desc ibed abo e, ha is, a SNP was only conside ed o be significan ly
associa ed o HRV i : (1) i had Po1106in s age 1 me a-analysis in Eu opean
indi iduals, (2) i had a one-sided Po0.05 in he new e hnici y specific
me a-analysis cong uen wi h he di ec ion o e ec in he s age 1 me a-analysis in
Eu opean indi iduals and (3) i had a genome-wide significan Po5108/3
( wo-sided) in he combined me a-analysis.
Co ec ing HRV o hea a e.The well-known in e se associa ion be ween
HRV and hea a e in pa eflec s a dependency o he a iance in IBI on he
mean IBI ha is un ela ed o ca diac agal ac i i y59. Tha is, he slowe he hea
a e, he longe he IBI, and he e o e, any p opo ionally mino bea - o-bea
di e ences in IBI a e mo e p onounced a slowe hea a es. This occu s on op o
he well-es ablished dual e ec o ca diac agal ac i i y ha lowe s hea a e and
inc eases HRV15,16. Al hough hese wo mechanisms (biological, mean- a iance
dependency) a e impossible o comple ely sepa a e, we conduc ed h ee analyses o
es whe he he HRV SNPs we e obus o co ec ion o he mean IBI.
Fi s , we co ec ed SDNN and RMSSD o hei dependency on mean IBI by
using he coe ficien o a ia ion, which is a mo e pa simonious solu ion19 han he
loga i hmic app oach sugges ed by Mon edi e al.29. We ob ained he summa y
s a is ics o he es ing hea a e GWAS me a-analysis18 om: h ps://
walke 05.u.hpc.mssm.edu/ and used he GWIS p ocedu e17 o in e a GWA
analysis o he coe ficien o a ia ion o he SDNN and he RMSSD. We
app oxima ed he coe ficien s o a ia ion by (SDNN/X) 100% and
(RMSSD/X) 100%, espec i ely, whe e X equals 60,000 pe hea a e.
T ans o ma ion om hea a e o IBI is equi ed as bo h e ms in he coe ficien o
a ia ion (HRV and IBI) a e in milliseconds, whe eas he hea a e GWAS
me a-analysis used hea a e in bea s pe minu e. As he coe ficien s o a ia ion
we e skewed, we used a log- ans o ma ion. As an example o he linea
app oxima ion by GWIS we assume ha he inc ease e ec o one allele o an
SDNN SNP is þ0.2 wi h he same SNP educing hea a e by 0.1. Gi en a
mean SDNN o 100 and mean hea a e o 60, we can hen app oxima e (omi ing
some nuances adequa ely explained in Nieuwboe e al.17) he e ec o he SNP on
NATURE COMMUNICATIONS | DOI: 10.1038/ncomms15805 ARTICLE
NATURE COMMUNICATIONS | 8:15805 | DOI: 10.1038/ncomms15805 | www.na u e.com/na u ecommunica ions 9
E a um: Gene ic loci associa ed wi h hea a e
a iabili y and hei e ec s on ca diac disease isk
Ilja M. Nol e e al.#
#A ull lis o au ho s and hei a filia ions appea s a he end o he pape .
Na u e Communica ions 8:15805 doi: 10.1038/ncomms15805 (2017); Published 14 Jun 2017; Upda ed 2 Aug 2017
In Supplemen a y Fig. 10 o his A icle, images o panels a and b we e inad e en ly omi ed. The co ec e sion o Supplemen a y
Fig. 10 is p o ided as Supplemen a y In o ma ion associa ed wi h his E a um. Fu he mo e, he a filia ion de ails o Azme aw T.
Ama e, Benedik on de Heyde, and Ma cel den Hoed a e inco ec in his A icle. The co ec a filia ion de ails o hese au ho s a e
gi en below.
Azme aw T. Ama e:
Depa men o Epidemiology, Uni e si y o G oningen, Uni e si y Medical Cen e G oningen, PO Box 30001, G oningen 9700 RB,
The Ne he lands.
School o Medicine, Uni e si y o Adelaide, Adelaide, Sou h Aus alia, SA 5005, Aus alia
College o Medicine and Heal h Sciences, Bahi Da Uni e si y, Bahi Da 6000, E hiopia.
Benedik on de Heyde:
Depa men o Immunology, Gene ics and Pa hology, Medical Gene ics and Genomics, Uppsala Uni e si y, Uppsala 75327, Sweden.
Science o Li e Labo a o y, Uppsala Uni e si y, Uppsala 75237, Sweden.
Ma cel den Hoed:
Depa men o Immunology, Gene ics and Pa hology, Medical Gene ics and Genomics, Uppsala Uni e si y, Uppsala 75327, Sweden.
Science o Li e Labo a o y, Uppsala Uni e si y, Uppsala 75237, Sweden.
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/
The Au ho (s) 2017
Ilja M. Nol e, M. Lo e o Munoz, Vinicius T agan e, Azme aw T. Ama e, Rick Jansen, Ahmad Vaez, Benedik on de Heyde,
Ch is y L. A e y, Joshua C. Bis, B am Die ckx, Jenny an Dongen, S ephanie M. Goga en, Philippe Goye e, Jussi He nesniemi,
Ville Huika i, Shih-Jen Hwang, Deepali Jaju, Ka hleen F. Ke , Alexande Klu ig, Bouwe P. K ij he, Ji ende Kuma ,
Sande W. an de Laan, Leo-Pekka Lyy ika
¨inen, Adam X. Maiho e , A pi Minassian, Pe e J. an de Mos ,
Ma ina Mu¨lle -Nu asyid, Michel Ni a d, E ika Sal i, James D. S ewa , Julian F. Thaye , Niek Ve weij, And ew Wong,
Delilah Zabaneh, Mohammad H. Za a mand, Abdel Abdellaoui, Sulayma Alba wani, Ch is ine Albe , Al a o Alonso, Fo am Asha ,
Juha Au inen, Tomas Axelsson, Dewleen G. Bake , Paul I.W. de Bakke , Ma eo Ba cella, Riad Bayoumi, Rob J. Bie inga,
Do e Boomsma, Gab ielle Bouche , Annie R. B i on, Ing id Ch is ophe sen, And ea Die ich, Geo ge B. Eh e , Pa ick T. Ellino ,
DOI: 10.1038/ncomms16140 OPEN
NATURE COMMUNICATIONS | 8:16140 | DOI: 10.1038/ncomms16140 | www.na u e.com/na u ecommunica ions 1

Ma kku Eskola, Janine F. Felix, John S. Flo as, Osca H. F anco, Pe e F ibe g, Maaike G.J. Gademan, Ma k A. Geye ,
Vilman as Gied ai is, Ca ha ina A. Ha man, Daiane Heme ich, Albe Ho man, Jouke-Jan Ho enga, Heikki Huiku i,
Nina Hu i-Ka
¨ho
¨nen, Xa ie Jou en, Juhani Jun ila, Ma kus Juonala, An i M. Ki iniemi, Jan A. Ko s, Meena Kuma i,
Ta iana Kuzne so a, Ca hy C. Lau ie, Joop D. Le and , Yong Li, Yun Li, Duanping Liao, Ma ian C. Limache , Hen y J. Lin,
Cecilia M. Lindg en, S e en A. Lubi z, Anubha Mahajan, Ba ba a McKnigh , Hen ie e Meye zu Schwabedissen, Yu i Milaneschi,
Nina Mononen, And ew P. Mo is, Mike A. Nalls, Ge jan Na is, Melanie Neij s, Kjell Nikus, Ka i E. No h, Daniel T. O’Conno ,
Johan O mel, Sieg ied Pe z, Anne e Pe e s, B uce M. Psa y, Olli T. Rai aka i, Vic o ia B. Risb ough, Mo i z F. Sinne ,
Da id Sisco ick, Johannes H. Smi , Nicholas L. Smi h, Elsayed Z. Soliman, Nona So oodehnia, Jan A. S aessen, Phyllis K. S ein,
Ad ienne M. S ilp, Ka a zyna S ola z-Sk zypek, Kons an in S auch, Johan Sunds o
¨m, Cees A. Swenne, Ann-Ch is ine Sy a
¨nen,
Jean-Claude Ta di , Ken D. Taylo , Alexande Teume , Timo hy A. Tho n on, Lesley E. Tinke , And e
´G. Ui e linden,
Jessica an Se en, And eas Voss, Melanie Waldenbe ge , Ki k C. Wilhelmsen, Gonneke Willemsen, Quenna Wong,
Zhu-Ming Zhang, Alan B. Zonde man, Daniele Cusi, Michele K. E ans, Halina K. G eise , Pim an de Ha s ,
Mohammad Hassan, E ik Ingelsson, Ma jo-Rii a Ja
¨ elin, S e an Ka
¨a
¨b, Mika Ka
¨ho
¨nen, Mika Ki imaki, Cha les Koope be g,
Diana Kuh, Te ho Leh ima
¨ki, La s Lind, Ca oline M. Nie e gel , Ch is J. O’Donnell, Albe ine J. Oldehinkel, B enda Penninx,
Alexande P. Reine , Ha ie
¨ e Riese, A ie M. an Roon, John D. Rioux, Je ome I. Ro e , Tama So e , B uno H. S icke ,
Henning Tiemeie , Tanja G.M. V ijko e, Folke W. Asselbe gs, Bianca J.J.M B undel, Susan R. Heckbe , E ic A. Whi sel,
Ma cel den Hoed, Ha old Sniede & Eco J.C. de Geus
2NATURE COMMUNICATIONS | 8:16140 | DOI: 10.1038/ncomms16140 | www.na u e.com/na u ecommunica ions
ERRATUM NATURE COMMUNICATIONS | DOI: 10.1038/ncomms16140