A principal component meta-analysis on multiple anthropometric traits identifies novel loci for body shape

ARTICLE Recei ed 19 Ma 2016 |Accep ed 21 Sep 2016 |Published 23 No 2016 A p incipal componen me a-analysis on mul iple an h opome ic ai s iden ifies no el loci o body shape Janina S. Ried and Janina M. Je e al.# La ge conso ia ha e e ealed hund eds o gene ic loci associa ed wi h an h opome ic ai s, one ai a a ime. We examined whe he gene ic a ian s a ec body shape as a composi e pheno ype ha is ep esen ed by a combina ion o an h opome ic ai s. We de eloped an app oach ha calcula es a e aged PCs (A PCs) ep esen ing body shape de i ed om six an h opome ic ai s (body mass index, heigh , weigh , wais and hip ci cum e ence, wais - o-hip a io). The fi s ou A PCs explain 499% o he a iabili y, a e he i able, and associa e wi h ca diome abolic ou comes. We pe o med genome-wide associa ion analyses o each body shape composi e pheno ype ac oss 65 s udies and me a-analysed summa y s a is ics. We iden i y six no el loci: LEMD2 and CD47 o A PC1, RPS6KA5/C14o 159 and GANAB o A PC3, and ARL15 and ANP32 o A PC4. Ou findings highligh he alue o using mul iple ai s o define complex pheno ypes o disco e y, which a e no cap u ed by single- ai analyses, and may shed ligh on o new pa hways. Co espondence and eques s o ma e ials should be add essed o M.M.-N. (email: [email p o ec ed]) o o R.L. (email: [email p o ec ed]). #A ull lis o au ho s and hei a filia ions appea s a he end o he pape . DOI: 10.1038/ncomms13357 OPEN NATURE COMMUNICATIONS | 7:13357 | DOI: 10.1038/ncomms13357 | www.na u e.com/na u ecommunica ions 1 La ge-scale me a-analyses o genome-wide associa ion s udies (GWAS) ha e iden ified nume ous loci o an h opome ic ai s, including mo e han 600 loci o heigh 1–3 and o e 160 loci o obesi y- ela ed ou comes, p edominan ly o commonly a ailable ai s such as body mass index (BMI)2and wais - o-hip a io (WHR)4,5, bu also o body a pe cen age6, childhood obesi y7and ex eme and ea ly onse obesi y7–9. While GWAS-me a-analyses ha e success ully e ealed new loci, so a , all hese s udies ha e ocused on one single an h opome ic ai a a ime and may no adequa ely cap u e di e ences in body shape be ween indi iduals who a e simila in one ai bu di e en in o he s. Fo example, wo indi iduals may ha e he same BMI, bu hei WHR and/o heigh can di e subs an ially, so ha each has a di e en body shape, which may ansla e in o di e ences in disease isk10,11. Se e al loci iden ified om p e ious single- ai GWAS on BMI, BMI-adjus ed WHR (WHRadjBMI) and heigh a e associa ed wi h mo e han one an h opome ic ai 1,2,4,12. Fo example, he loci nea MC4R and nea POMC/ADCY3 a e each associa ed wi h BMI and heigh . Howe e , he BMI-inc easing allele o he nea -MC4R locus is associa ed wi h inc eased heigh , whe eas he BMI-inc easing allele o he nea -POMC/ADCY3 locus is associa ed wi h educed heigh 1,2. Thus, hese loci a e likely each associa ed wi h a mo e comp ehensi e body shape pheno ype ha is no cap u ed by cu en GWAS ha only conside an h opome ic ai s indi idually. In ecen yea s, se e al app oaches ha e been de eloped o examine whe he single-nucleo ide polymo phisms (SNPs) influence mul iple co ela ed ai s associa ed wi h disease13,14. Howe e , mos app oaches es pheno ypes sepa a ely and a e hus subjec o mul iple es ing penal ies ha ul ima ely educe he s a is ical powe o de ec geno ype–pheno ype ela ionships among co ela ed ai s. One way o wa d is o apply a dimension educ ion me hod o he ai s o in e es , such as p incipal componen analysis (PCA) ha combines mul iple co ela ed ai s in o a se o unco ela ed ou comes p incipal componen s(p incipal componen s (PCs))15,16. This me hod is e y appealing o cap u e a composi e pheno ype, such as body shape. To da e, no la ge-scale GWAS me a-analyses ha e been epo ed ha aim o iden i y gene ic loci associa ed wi h body shape based on simul aneous analysis o mul iple an h opome ic ai s using PCA me hods. The e o e, he pu pose o ou s udy was wo old. Fi s , we aimed o cap u e body shape in i s mul i-dimensional s uc u e using PCs om se e al commonly a ailable an h opome ic ai s. To allow he me a-analysis o summa y s a is ics ac oss a la ge numbe o coho s, we de eloped an app oach ha calcula es a e aged PCs (A PCs) ha obus ly ep esen body shape ac oss a wide ange o s udies. Second, using his app oach, we aimed o iden i y gene ic loci associa ed wi h body shape based on he A PCs in 65 s udies o he GIANT Conso ium, including 4170,000 indi iduals. Resul s Defining composi e pheno ypes o body shape. As basis o ou analysis o body shape we used six an h opome ic ai s: BMI, WHR, heigh , weigh , hip and wais ci cum e ence. Fi s , we pe o med sepa a e PCA in a subse o 20 la ge popula ion-based s udies (up o 82,355 indi iduals, Supplemen a y Table 1) and compa ed he loadings o he an h opome ic ai s in each PC be ween s udies. Visual inspec ion o PCA loadings showed high conco dance ac oss s udies (Supplemen a y Fig. 1) and be ween men and women. Be ween-s udy a ia ion in a iance explained by he PCs was small (Supplemen a y Fig. 1, Supplemen a y Table 2). On a e age, he fi s ou PCs explained mo e han 99% o he a iance (Fig. 1, Supplemen a y Table 2), and we e he e o e pu sued as body shape ou comes o ou gene-disco e y e o . Gi en he ac oss-s udy s abili y o PCs, we de i ed a e age loadings ha we e calcula ed as weigh ed means o loadings om all 20 popula ion-based s udies ha we e analysed in his s ep. We used hese a e age loadings o calcula e a e age p incipal componen s (A PCs) as a ge s in each o he GWAS included in he fi s and second s age. In o he wo ds, he pheno ypes used o genome-wide associa ion we e cons uc ed in a consis en way ac oss s udies, such ha he summa y s a is ics could be me a- analysed. Each A PC ep esen s a specific composi ion o he six an h opome ic ai s and hus cap u es a specific aspec o body shape (Fig. 1). The fi s A PC, which explains on a e age 64.4% o he a ia ion in all ai s, shows high loadings o all ai s, excep o heigh . The loadings a e in he same di ec ion; meaning ha he A PC cap u es in e -indi idual a ia ion in ei he inc eased o dec eased BMI, weigh , WHR, hip and wais ci cum e ence. The e o e, a ia ion in his PC seems o p edominan ly cap u e o e all adiposi y. The second A PC, which explains 18.5% o he a ia ion, is cha ac e ized by pa icula ly high bu opposi e loadings on heigh and WHR. In o he wo ds, A PC2 cap u es a ia ion in a composi e pheno ype ha ep esen s all indi iduals wi h a small WHR o , ice e sa, sho indi iduals wi h a la ge WHR. The hi d A PC, explaining 13.8% o he a ia ion, also shows p edominan ly high loadings on heigh and WHR bu in he same di ec ion, wi h an opposi e loading o nea ly he same size on hip ci cum e ence. Gi en hese loadings, A PC3 disc imina es mainly be ween all indi iduals wi h a high WHR esul ing om a smalle hip ci cum e ence on one ex eme and sho indi iduals wi h low WHR, and a la ge Hip Hip Wais Wais Weigh WHR Weigh WHR –1 –0.5 00.5 BMI BMI 1–1 –0.5 0 0.5 1 a PC3 a PC4 a PC1 Loadings o A PCs Explained a iance a PC2 Heigh Heigh Hip Hip wais Wais Weigh 0 A e age explained a iance in % 20 40 60 80 100 PC1 PC2 PC3 PC4 PC5 PC6 WHR Weigh WHR –1 –0.5 0 0.5 BMI BMI 1 –1 –0.5 0 0.5 1 Heigh Heigh a b Figu e 1 | Loadings and explained a iance o A PCs o body shape. (a) Loadings o A PCs, and (b) explained a iance o A PCs o body shape. ARTICLE NATURE COMMUNICATIONS | DOI: 10.1038/ncomms13357 2NATURE COMMUNICATIONS | 7:13357 | DOI: 10.1038/ncomms13357 | www.na u e.com/na u ecommunica ions hip ci cum e ence on he o he ex eme. The ou h A PC explains on a e age 3% and is ha de o in e p e . I displays high loadings on BMI and body weigh , and opposi e loadings o a simila size on hip and wais ci cum e ence. These could be in e p e ed as a pheno ype anging be ween high BMI and weigh , wi h ela i ely small hip and wais ci cum e ence on he one hand and low BMI and weigh bu la ge wais and hip ci cum e ence on he o he hand. Consis en wi h he indi idual an h opome ic ai s, he ou A PCs ha desc ibe body shape a e also he i able. Using da a om ou isola e popula ions (n¼4,000), we es ima ed ha A PC2 has he highes he i abili y (75–80%), consis en wi h he ac ha heigh is he main con ibu ing ai o his A PC wi h a s ong gene ic componen 1. The he i abili y o A PC1 (35–50%), A PC3 (50–75%) and A PC4 (25–50%) we e mode a ely high and simila o he he i abili y o indi idual an h opome ic ai s17 (Supplemen a y Fig. 2). F om a clinical pe spec i e, each o he ou A PCs exhibi known co ela ions wi h ca dio- me abolic ai s (Supplemen a y Fig. 3), including dias olic blood p essu e, sys olic blood p essu e, o al choles e ol, low-densi y lipop o ein choles e ol, high-densi y lipop o ein choles e ol and o al iglyce ides le els. Genomic disco e y o body shape composi e pheno ypes.We pe o med a wo-s aged me a-analysis o iden i y gene ic loci ha a e associa ed wi h he ou A PCs (Supplemen a y Table 3, Supplemen a y Table 4). In he fi s s age, a me a-analysis o 43 s udies wi h impu ed genome-wide SNP da a including mo e han 133,000 indi iduals iden ified SNPs in 385 loci ac oss he ou A PCs (56 loci o A PC1, 205 o A PC2, 89 o A PC3 and 35 o A PC4) ha showed p omising associa ion (P alueo5106) o a leas one o he ou A PCs (Fig. 2, Supplemen a y Fig. 4). Lead SNPs (and p oxies; see ‘Me hods’ sec ion) o each locus we e aken o wa d o alida ion in a second s age, including da a om mo e han 39,900 indi iduals om 22 s udies o which 12 s udies had geno ypes om he Illumina Ca dioMe abochip and 10 s udies had impu ed genome- wide SNP da a. In he combined analyses, consis ing o he fi s and second s age s udies, he associa ion o 207 o he 385 loci eached genome-wide significance (P alue o5108) (31 o A PC1, 124 o A PC2, 45 o A PC3 and 7 o A PC4; Fig. 2, Fig. 3, Supplemen a y Fig. 4, Supplemen a y Table 6), o which 16 loci we e iden ified o wo A PCs and one showed significan associa ion wi h h ee A PCs (Supplemen a y Fig. 7, Supplemen a y Table 5) esul ing in a o al o 189 loci wi h –log10 (P alue)–log10 (P alue) –log10 (P alue) CD47 LEMD2 GANAB ARL15 ANP32B RPS6KA5 a PC1 a PC2 a PC3 Ch omosome 60 50 30 20 10 0 Obse e d –log10 (P alue) 60 50 30 20 10 0 Obse e d –log10 (P alue) 50 40 30 20 10 0 Obse e d –log10 (P alue) Obse e d –log10 (P alue) 30 20 10 10 0 0 024 Expec ed –log10 (P alue) 6 024 Expec ed –log10 (P alue) 6 024 Expec ed –log10 (P alue) 6 0246 50 40 30 30 20 20 10 10 0 0 a PC4 –log10 (P alue) 30 20 10 0 P e ious GIANT loci No el loci Addi ionally excluding no el bod y shape loci Excluding loci om p e ious GIANT analyses All SNPs Expec ed –log10 (P alue) Ch omosome Ch omosome Ch omosome 1 2 4 5 6 7 8 9 10 11 12 13 14 15 16 17 18 19 22 30 20 21 20 3 1 2 4 5 6 7 8 9 10 11 12 13 14 15 16 17 18 19 22 21 20 3 1 2 4 5 6 7 8 9 10 11 12 13 14 15 16 17 18 19 22 21 20 3 1 2 4 5 6 7 8 9 10 11 12 13 14 15 16 17 18 19 22 21 20 3 Figu e 2 | Manha an and QQ-plo s o associa ion esul s on A PCs o body shape. P alues o he fi s s age me a-analysis a e gi en in he Manha an and QQ-plo s. All genome-wide significan loci a e highligh ed. NATURE COMMUNICATIONS | DOI: 10.1038/ncomms13357 ARTICLE NATURE COMMUNICATIONS | 7:13357 | DOI: 10.1038/ncomms13357 | www.na u e.com/na u ecommunica ions 3 associa ion o a leas one A PC. To de e mine whe he he loci we iden ified we e independen o he loci p e iously ound o BMI, WHRadjBMI and heigh , we pe o med condi ional ana- lyses on SNPs epo ed in p e ious GIANT-GWAS publica ions on BMI, WHRadjBMI and heigh 1,2,4,5,18,19. A locus was conside ed independen o epo ed findings i he P alue in he analyses condi ioned on all p e iously iden ified loci emained sugges i e (P alue o5106). In o al, 183 loci had al eady been es ablished o BMI, WHRadjBMI o heigh (Fig. 3, Supplemen a y Fig. 7), whe eas six loci had no p e iously been iden ified o associa ion wi h con en ional an h opome ic ai s; wo o A PC1, wo o A PC3 and wo o A PC4 (Table 1, local associa ion plo s gi en in Supplemen a y Fig. 5). Fo hese six no el loci, he esul s o he lead SNPs we e checked in p e iously pe o med GWAS me a-analyses on an h opo- me ic and ca dio-me abolic ai s (Supplemen a y Table 7). Resul s o A PC1. Fo A PC1, we iden ified 31 genome-wide significan loci, o which wo we e no el (ups eam o LEMD2 and CD47). O he 29 p e iously es ablished loci, 24 ha e been associa ed wi h BMI only18, 3 wi h heigh only1,3, while wo loci ha e been epo ed o associa ions wi h bo h BMI and heigh 3,18 (Fig. 3a). While bo h no el loci showed some e idence o associa ion wi h BMI in he la es GIANT–GWAS (n4339,000; Po7.2 103; Table 1), hey did no each genome-wide significance. The lead SNP ( s943466) 7 kb ups eam o LEMD2 has been epo ed o be associa ed wi h exp ession o LEMD2 in li e (P¼1.66 109)20,21. Ano he a ian in LEMD2 ( s2296743 a 8 kb om ou lead SNP s943466; 2¼0.2, D0¼1.0) was p e iously epo ed o i s p omising associa ion (P alue ¼8106) wi h ene gy in ake a dinne in a small GWAS o 815 Hispanic child en22. The lead SNP ( s7640424) o he second no el locus was loca ed in an enhance egion 10 kb ups eam o CD47 ( e s 23,24), which encodes a memb ane p o ein ha migh be in ol ed in signal ansduc ion and memb ane anspo 25. No genome-wide significan associa ions ha e been epo ed o he lead SNP o o he SNPs in he CD47 gene be o e23–25. Howe e , a ecen s udy e ealed a link o die - induced obesi y in mice and sugges s CD47 as a po en ial d ug- a ge o comba obesi y and me abolic complica ions26,27. Resul s o A PC2. Fo A PC2, we iden ified no no el loci. Almos all (n¼122) o he 124 loci associa ed wi h A PC2 had p e iously been iden ified o heigh 1(Fig. 3b), which is consis en wi h A PC2’s high loadings on heigh and opposi e loadings on WHR. O hese 122 loci, 103 we e epo ed o associa ion o heigh only, whe eas o he 19 emaining loci, 4 we e p e iously associa ed wi h heigh , BMI and WHRadjBMI, 2 loci we e epo ed o heigh and BMI and 13 loci o e lapped wi h heigh and WHR. The wo A PC2 loci ha did no associa e wi h heigh we e p e iously iden ified o WHRadjBMI19. Resul s o A PC3. We iden ified 45 loci ha eached genome- wide significance o A PC3, o which 2 we e no el. Consis en wi h he loadings o A PC3, 43 o he associa ed loci had been epo ed be o e o heigh 1o WHR4,19 (Fig. 3c). The lead SNP o he fi s no el locus s7492628, ups eam o he genes RPS6KA5 (420 kb) and C14o 159 (430 kb), ailed o each genome-wide significance in p e ious WHRadjBMI GWAS (P alue ¼9.3 108) and was nominally associa ed wi h ex eme obesi y isk (P alue ¼7.26 105)28. The lead SNP o he o he no el locus, GANAB, s7949030, showed some e idence o associa ion wi h WHRadjBMI in he la es GIANT GWAS (P alue ¼3.3 106) and was epo ed o be an eQTL o se e al o he genes21: In monocy es, egula ion o MIR3654,EEF1G,EML3, BSCL2,HNRNPUL2-BSCL2,LRRN4CL was ound29–31.BSCL2 is o in e es , as i is a known candida e gene o he mos se e e lipodys ophy pheno ype32. In blood s7949030 was ound o be an eQTL o HNRNPUL2-BSCL2,AHNAK,LRRN4CL and INTS5 ( e s 33,34), while in skin and adipocy es i was ound as an eQTL o EML3 ( e s 30,31,35). Resul s o A PC4. Se en loci we e iden ified o A PC4, o which fi e had been p e iously epo ed; one o BMI and heigh , one o WHR and heigh , one o heigh only and wo o WHR only1,3,4,36 (Fig. 3). The lead SNPs o he wo no el loci iden ified wi h A PC4 we e bo h in onic, in ARL15 and ANP32. The allele associa ed wi h inc eased A PC4 o he lead SNP ( s4865796) in ARL15 was mode a ely associa ed wi h highe BMI (P alue ¼1.6 104), inc eased adiponec in le els (P alue ¼4.2 106ADIPOGEN37) and dec eased isk o diabe es (P alue ¼1.8 105, DIAGRAM38). This SNP was associa ed wi h as ing insulin ( s4865796, P¼2.1 10 8and 2.2 1012 a e adjus men o BMI39). O he nea by SNPs in high linkage disequilb ium (LD), ha e p e iously been epo ed o associa ions wi h BMI-adjus ed adiponec in le els ( s6450176/ s4311394, 2¼0.087, D0¼0.87 ( e s 37,40)), high densi y lipop o ein C (HDL-C) le els ( s6450176 ( e s 41,42)) and isk o ype 2 diabe es ( s702634, 2¼1.0, D0¼1.0 ( e . 38)). A duplica ion in ARL15, agged by s16992296) was p e iously ound o be associa ed wi h inc eased isk o childhood obesi y in Eu opean and A ican Ame icans43. Howe e , his duplica ion is independen o he associa ion we ound o s4865796-ARL15 and A PC4, which is in low LD ( 2EUR ¼0.065) wi h he duplica ion ( ep esen ed by s16992296), loca ed 168 kb ups eam. The lead SNP ( s7855432) o he second locus, ANP32B, was mode a ely associa ed wi h heigh (P alue ¼5.5 106)1. A SNP in high LD ( s4743150 2¼0.95, D0¼1.0) was epo ed o be p omisingly associa ed wi h co ona y hea disease isk (P alue ¼5106)44. a PC1 ab cd Heigh WHR Heigh WHR Heigh WHR Heigh WHR a PC2 a PC4a PC3 BMI BMI BMI BMI 24 2 2 2 0 00 0 0 0 0 0 0 0 2 2 21 1 1 2 103 13 4 4 8820 0 3 3 Figu e 3 | Numbe o loci associa ed wi h A PCs and known om p e ious GIANT analyses on BMI, WHR o heigh . (a–d) co esponds o each A gPC espec i ely. The Venn diag ams speci y o each A PC how many significan ly associa ed loci (p omising P alue in he fi s s age me a analysis (o5106) and genome wide significan in fi s and second s age combined analysis (o5108)) a e known om p e ious GIANT analysis on BMI, heigh o WHR. In he uppe igh co ne o each plo he numbe o loci is gi en ha a e no known om p e ious GIANT analyses. ARTICLE NATURE COMMUNICATIONS | DOI: 10.1038/ncomms13357 4NATURE COMMUNICATIONS | 7:13357 | DOI: 10.1038/ncomms13357 | www.na u e.com/na u ecommunica ions Discussion We de eloped a PCA-based app oach o cap u e a ia ion ac oss mul iple ai s simul aneously in a uni o m way ac oss mul iple s udies. Resul ing A PCs a e a obus c oss-pheno ype ep esen- a ion allowing hei use in la ge-scale me a-analyses. We assessed his app oach o cap u e body shape based on six indi idual an h opome ic ai s and iden ified six no el loci ha we e no iden ified be o e in much la ge GWAS-me a-analyses o BMI, WHRadjBMI and heigh 1,2,4. Ou findings sugges ha he body shape composi e pheno ype, assessed by A PCs, ep esen s in o ma ion ha is no ully cap u ed by indi idual (an h opome ic) ai s. Applica ion o his me hod o o he ela ed ai s, o example, in immune disease, di e en ypes o cance , ca diome abolic ai s, o o he co ela ed ai s migh compa ably e eal new loci, and po en ially new pa hways, ha ha e no been iden ified in single- ai GWAS. The A PCs a e combina ions o di e en an h opome ic ai s and he e o e cap u e mo e complex body shape pheno ypes han he single ai s. A PC1, ep esen ing o e all adiposi y, and A PC2, ep esen ing heigh wi h espec o WHR, a e he mos impo an con ibu o s o body shape, explaining on a e age mo e han 80% o he a ia ion. Mo e specific body shape ypes we e cap u ed by A PC3 and A PC4 and we e defined by impac o heigh and WHR (A PC3) o BMI, wais and hip (A PC4). Ou ini ial analyses demons a ed ha he loadings a e s able ac oss s udies, s udy designs and be ween men and women. Mo eo e , we ha e shown ha he A PCs a e he i able ai s and co ela ed wi h ca diome abolic ai s and isk ac o s. To u he demons a e he s eng h o his app oach, we compa ed o al a iance explained o single ai s and A PCs by SNPs p e iously iden ified in single- ai GWAS ( o BMI, WHRadjBMI, heigh 1,2,4). Fo example, he 97 loci ha ha e been epo ed o associa ion in he la es BMI single- ai GWAS (NB340,000) explain 8.7% o he a ia ion in A PC1, whe eas hey explained only 2.68% o he a ia ion in BMI2. These da a indica e ha ou PC-defined pheno ype o o e all body size (A PC1) cap u es a mo e composi e pheno ype compa ed wi h BMI as a single- ai . Explaining mo e o he a iance wi h he same gene ic a ian s as p e ious single- ai s udies in ou composi e pheno ype shows p omise o upda e and in o m exis ing me hods. So a , ypical GWAS ha e es ed o associa ion o gene ic a ian s wi h an h opome ic ai s, one ai a a ime. We define ‘body shape’ as a composi e o mul iple ai s defined by PCs. We fi s pe o med PC-analyses in ep esen a i e popula ion-based s udies and a e aged PC loadings ac oss hese s udies (A PCs). We subsequen ly use hese A PCs o calcula e PCs in all pa icipa ing s udies. This app oach ensu es ha PCs a e calcula ed in a uni o m manne ac oss all s udies, hus acili a ing subsequen me a-analyses. This app oach could be applied o cap u e gene ic a ia ion ac oss ela ed ai s ha is cu en ly no cap u ed by single- ai s GWAS ( o example, in he con ex o au oimmune disease, blood ai s, lipid le els, di e en cance s and so on.). Consis en wi h published an h opome ic ai s10,11,17, he de i ed A PCs a e he i able and co ela ed wi h clinically ele an ou comes. We iden ified addi ional loci, despi e a much smalle sample size compa ed wi h he la es single- ai GWAS analyses o BMI, heigh and WHRadjBMI1,2,4. This sugges s ha he A PC me hod cap u es pheno ype in o ma ion ha is no cap u ed by single- ai analyses and associa ed loci may highligh biological pa hways ha a e no e ealed wi h single- ai associa ed loci only. E en hough ou app oach has se e al ad an ages, i is no mean o eplace single ai GWAS analyses. A numbe o loci ha we e iden ified in he la es single- ai GWAS we e no iden ified in ou body shape GWAS; ha is, we iden ified 124 loci (o 14.2%) o he 837 loci ecen ly epo ed in he GIANT single- ai me a-analyses (Supplemen a y Fig. 6). This may be due o he ac ha hese ecen single- ai GWAS me a-analyses we e a leas wice as la ge as he cu en body shape GWAS. Howe e , e en when we compa e he numbe o iden ified loci in ea lie GWAS me a-analyses, which a e o simila size as he cu en body shape GWAS, we do no iden i y all p e iously epo ed loci o single ai s. Pe haps his is mos ob ious wi h heigh (la gely ep esen a i e o A PC2), whe e we only iden ified 91 (13.1%) o 697 loci iden ified o heigh . This is in pa due o he ac ha a conse a i e defini ion o linkage disequilib ium was applied ( 240.8), lack o powe due o sample size o SNPs o modes e ec s, o pe haps he A PCs in oduces noise o pu ely single ai s such as heigh . Consis en wi h his finding, we also obse e ha some single ai s also explain mo e o he a iance o body shape compa ed wi h A PCs. Ou compa ison o he a iance explained be ween p e ious single- ai s me a-GWAS and ou A PCs suppo his e idence o o e lapping associa ed a ian s. Since A PC2 ep esen s la gely a single ai , heigh , wi h la ge heigh loadings we we e unable o explain mo e o he a iance. In ac we explained less o he a iance, which is likely due o noise in oduced using his composi e A PCs pheno ype. This obse a ion is also e iden o a iance in body shape explained Table 1 | Associa ion esul s o no el loci wi h a PC o body shape. 1s s age up o 133,376 samples 2nd s age up o 39,904 samples 1s þ2nd s age combined up o 173,278 samples Condi ioned analysis on all GIANT ophi s P alue o SNPs in GIANT analysisw P alue o SNPs in GIANT analysisz ai SNP (lead SNP) Nex gene E ec / o he allele EAF*P alue P alue be a (sebe a) P alue Nbe a (sebe a) P alue BMI Heigh WHR BMI Heigh WHR a PC1 s7640424 CD47 C/T 69% 5.40E-07 0.0015 0.05 (0.008) 3.18E-09 171,544 0.05 (0.01) 5.80E-07 0.0072 0.74 0.25 2.28E-06 0.28 0.85 a PC1 s943466 ( s2281819) LEMD2 G/A 76% 6.39E-07 0.016 0.049 (0.009) 3.47E-08 172,174 0.049 (0.01) 7.28E-07 2.7E-04 0.045 0.54 9.34E-06 0.75 0.25 a PC3 s7949030 GANAB G/A 38% 2.74E-08 0.11 0.024 (0.004) 5.58E-09 139,195 0.025 (0.004) 6.36E-09 0.082 0.80 1.4E-04 0.54 0.041 3.3E-06 a PC3 s7492628 RPS6KA5 G/C 30% 8.75E-08 0.13 0.024 (0.004) 1.90E-08 139,874 0.024 (0.004) 7.93E-08 0.064 0.62 4.9E-05 0.0050 0.58 9.3E-08 a PC4 s4865796 ( s1664781) ARL15 G/A 32% 5.59E-07 0.011 0.008 (0.001) 2.25E-08 172,517 0.008 (0.002) 7.25E-07 5.1E-05 0.034 0.40 1.6E-04 0.020 0.84 a PC4 s7855432 ANP32B G/T 80% 1.40E-07 0.17 0.01 (0.002) 4.06E-08 140,805 0.01 (0.002) 1.78E-07 0.33 0.046 0.49 0.32 5.5E-06 0.91 The associa ion esul s o he fi s s age, second s age and fi s and second s age combined analysis is gi en o all six loci ha we e genome wide significan ly associa ed (p omising P alue in he fi s s age me a analysis (o510 6) and genome wide significan in fi s and second s age combined analysis e (o510 8)) wi h one o he a PCs and no el. Mo eo e , he P alues o he analysis condi ioned on all ophi s om he ecen GIANT publica ions on BMI, heigh and WHR. *EAF is mean o EAF o all s udies in he fi s s age me a analysis. wAll ophi s o he GIANT analysis published be o e 2014 ( e s 3,6). zAll ophi s o he GIANT analysis unpublished and/o published a e 2014 ( e s 1,2,4). NATURE COMMUNICATIONS | DOI: 10.1038/ncomms13357 ARTICLE NATURE COMMUNICATIONS | 7:13357 | DOI: 10.1038/ncomms13357 | www.na u e.com/na u ecommunica ions 5 by heigh compa ed wi h A PC3 and A PC4, bu is in con as o BMI, a complex ai comp ised o mul iple an h opome ic measu emen s, which explains less a iance in body shape compa ed wi h A PC3 and A PC4. I is impo an o emphasize ou app oach is mos in o ma i e o complex ai s such as BMI ha a e de i ed om a se ies o o he ai s. We belie e ha using PC space o define complex ai s is use ul o he de ec ion o loci in ol ed in mul iple pa hways ha migh go unde ec ed in a single ai se ing. We ha e de eloped a new s a egy ha applies a PCA app oach in a me a-analysis se ing o combine composi e pheno ypes in a ha monized way ac oss mul iple s udies. We success ully applied his app oach o an h opome ic ai s o cap u e body shape. The de i ed combined an h opome ic ai s (A PCs) we e shown o be he i able and co ela ed o ca dio-me abolic ai s. La ge-scale GWAS me a-analyses o he A PCs iden ified six new loci ha we e no iden ified by p e ious single- ai GWAS ha we e wice as la ge in samples size. This PCA app oach could maximize gene disco e y o o he co ela ed ai s, such as cance s, immune disease, hema ologic ai s and so on. and may iden i y genes ha poin owa ds sha ed physiological pa hways. Me hods S udy desc ip ion.In he fi s s age analyses, 43 s udies pa icipa ed (133,376 indi iduals) ha had HapMap 2 impu ed genome-wide da a a ailable. A subse o 20 s udies wi h un ela ed indi iduals was used o calcula ion o a e age loadings. Second s age analyses we e pe o med in 10 s udies (7,734 indi iduals) wi h genome-wide da a ha became a ailable a e he fi s s age and 12 s udies (32,170 indi iduals) wi h Ca dio-Me aboChip (by Illumina) da a (numbe o included s udies and indi iduals gi en in Supplemen a y Table 3). De ails on s udy pheno ypes, geno yping and impu a ion o each s udy a e gi en in he Supplemen a y Tables 8 and 9, espec i ely. E hics s a emen .All s udy pa icipan s ga e w i en in o med consen and e hic commi ees app o ed all s udies. The e hic s a emen o each s udy is gi en in he s udy specific acknowledgemen s. Calcula ion o a e age loadings.In 20 independen s udies (Supplemen a y Table 1) wi h un ela ed pa icipan s PCAs we e pe o med on six an h opome ic ai s (BMI, heigh , hip, wais , weigh and WHR). Each s udy pe o med a PCA on he s anda dized esiduals o he an h opome ic ai s adjus ed o age and gende . The same analyses we e done o men and women sepa a ely wi h esiduals adjus ed o age only. The esul o he PCA in each s udy is a se o six PCs ha a e o hogonal linea combina ions o he six an h opome ic ai s. In o he wo ds each PC is a weigh ed sum o he six ans o med an h opome ic ai s and independen o he o he PCs. The weigh s o each ai pe PC a e called loadings. Each s udy also calcula ed he explained a iance pe PC. The loadings and explained a iances we e compa able o all s udies (Supplemen a y Fig. 1 (1)). Wi h he in en ion o c ea e pheno ypes ha a e iden ically cons uc ed in all s udies, he esul s o single s udy PCAs we e used o deduce he a e age loadings. This app oach is easonable as he loadings o he s udy specific PCAs we e compa able. Wi h he use o he single s udy co ela ion ma ices a combined a e age co ela ion ma ix was de i ed (weigh ed sum di ided by numbe o indi iduals). This a e age co ela ion ma ix is hen used as basis o a PCA. The loadings ha esul om his PCA a e called a e age loadings (Fig. 1a) and Supplemen a y Table 2). This was pe o med o men, women and all indi iduals combined, howe e ul ima ely we used combined loadings o p ima y esul s epo ed in he manusc ip . Sex specific esul s a e epo ed in he Supplemen a y Ma e ial. The a e age loadings and explained a iance we e compa able o he s udy specific loadings and explained a iances (Supplemen a y Fig. 1). He i abili y analyses.He i abili y o he a PCs was calcula ed wi hin ou popula ion isola es, CROATIA-Vis (n¼909), CROATIA-Ko cula (n¼842), CROATIA-Spli (n¼499) and ORCADES (n¼866) using he ‘polygenic’ unc ion o he GenABEL package o R 45. A e age p incipal componen s as body shape pheno ype.The a e age loadings we e used in each s udy o calcula e he A PCs in a s anda dized way. The e o e, he a e age loadings we e dis ibu ed oge he wi h an R-sc ip (h p://www. - p ojec .o g/) ha calcula ed he A PCs as linea combina ion o esiduals o he s udy pheno ypes wi h he use o he a e age loadings. This was done o men and women sepa a ely and addi ionally o combined in s udies wi h ela edness s uc u e. As he fi s ou PCs explain on a e age mo e han 99% o he a iance (Fig. 1b) we decided o limi all analyses o hese ou PCs. S age 1 analyses.GWAS on he fi s ou A PCs we e calcula ed o men and women sepa a ely in s udies o un ela ed samples and combined o s udies wi h ela ed samples wi h an adjus men o s udy si e when necessa y. All s udies o he fi s s age analyses used HapMap 2 impu ed genome-wide da a. GWAS esul s unde wen ex ensi e quali y con ol and s udy-wise fil e ing (call a e 495%, P alue (HWE)4106, impu a ion quali y, mino allele coun (MAC) 43). The me a analyses o GWAS esul s o he fi s ou A PCs we combined sex-s a ified esul s o s udies wi h un ela ed indi iduals and uns a ified GWAS esul s o s udies wi h ela edness indi iduals. Me a analyses we e pe o med wi h METAL 46 using fixed e ec s in e se a iance-weigh ed me hod. Single s udy and he me a analysis P alues we e co ec ed by he genomic con ol infla ion ac o l(me a analysis lbe o e co ec ion: l(PC1) ¼1.29, l(PC2) ¼1.407, l(PC3) ¼1.236, l(PC4) ¼1.136). Resul s we e limi ed o SNPs ha a e in HapMap 2 and had esul s o mo e han 30,000 indi iduals. He e ogenei y analysis was pe o med wi h METAL. Each A PC all SNPs wi h a p omising P alue (P alueo510 6) we e iden ified in combined analyses. To iden i y p omising loci clus e ing (LD40.01, dis ance o1,000 kb) wi h PLINK47 based on HapMap 2 geno ypes was pe o med. All leading SNPs pe clump o A PCs we e aken o wa d o second s age analyses and named p omising SNPs in his manusc ip . Two SNPs ha we e p omising o he fi s p incipal componen had e y low he e ogenei y P alues ( s10847678 (P alue(he ) ¼8.8 10152), s13296358 (P alue(he ) ¼5.4 1067)). Fo bo h SNPs he e ec was d i en only by a single s udy and no o he SNP in high LD had a p omising P alue. The e o e, hese wo SNPs we e emo ed om u he analyses. S age 2 Analyses.As men ioned abo e o second s age analyses a mix u e o s udies wi h genome-wide SNP da a and Me aboChip geno ypes was a ailable. Some o he leading SNPs o he fi s s age analyses we e no geno yped on he Me aboChip. To inc ease he powe o all p omising SNPs o each A PC p oxies we e defined ha we e all SNPs close o p omising SNPs (dis ance o500 kb), in high LD (LD40.9) and a ailable in mo e han 70% o he indi iduals o he second s age. Resul s o he second s age analyses unde wen he same quali y con ol as fi s s age esul s. Combined analyses.The combined analyses o all fi s and second s age GWAS was pe o med wi h METAL35 wi h in e se a iance based me hod. Resul s o men and women we e combined as desc ibed o he fi s s age me a-analyses. All p omising loci o which a leas one p oxy had a genome-wide significan P alue in he combined analysis we e named genome-wide significan loci and he bes SNP o he combined analyses (la ges absolu e be a) was epo ed as opSNP o his locus. No el loci - condi ional analyses and look-ups in p e ious GIANT analyses. Two analyses we e pe o med o dis inguish be ween genome-wide significan body shape loci ha a e known om p e ious GWAS on BMI, heigh and WHR and no el body shape loci. Fi s , condi ional analyses we e pe o med. We used he 226 epo ed opSNPs (32 BMI, 180 heigh , 14 WHR) o published GIANT analyses on BMI, heigh and WHR1,2,4 o pe o m condi ional analyses o he fi s s age me a-analyses using GCTA15,48. The esul s o his analysis we e hen analysed condi ioned on 843 opSNPs (97 BMI, 697 heigh , 49 WHR) o he published GIANT analyses1,2,4. To iden i y he o e lap o he esul s o A PCs wi h he single an h opome ic ai s, he same condi ional analyses we e pe o med o BMI, heigh and WHR sepa a ely. Fo calcula ion o he LD-s uc u e geno ype da a om KORA F4 was used. Two opSNPs o he unpublished GIANT esul s had o be emo ed be o e analyses as hey we e in high co ela ion wi h wo o he opSNPs. I he body shape opSNPs we e independen loci iden ified by p e ious GIANT analyses, he P alue should s ay p omising (P alueo5106) in bo h condi ional analyses. Second, we checked by look-ups i hose genome-wide significan SNPs ha a e independen om he p e iously epo ed opSNPs we e no genome-wide significan (P alue4510 8)in GIANT analyses1,2,4. Genome-wide significan SNPs a e named no el SNPs i hey ulfil he ollowing condi ions: P alue o condi ioned analyses on opSNPs epo ed by p e ious GIANT analyses (on BMI, heigh , WHR) emained p omising (P alueo5106). P alue in p e ious GIANT analyses (on BMI, heigh , WHR) was no genome-wide significan (P alue45108). Pleio opic e ec s.Fo iden ifica ion o po en ial pleio opic e ec s se e al look-ups in a ious la ge-scale conso ia on di e en pheno ypes we e pe o med, including GIANT, DIAGRAM and MAGIC, all e e ences a e gi en in he esul s able o he look-ups (Supplemen a y Table 7). Fo compa ison o e ec di ec ions ARTICLE NATURE COMMUNICATIONS | DOI: 10.1038/ncomms13357 6NATURE COMMUNICATIONS | 7:13357 | DOI: 10.1038/ncomms13357 | www.na u e.com/na u ecommunica ions he loadings o each A PC ha e o be conside ed. Fo example A PC2 includes heigh wi h a posi i e loading and BMI wi h a nega i e loading. Tha means an inc easing e ec on A PC2 means an inc easing e ec on heigh bu a dec easing e ec on BMI. Fu he Analyses.PCA, u he analyses and plo s we e gene a ed wi h R (h p://www. -p ojec .o g/) i no s a ed o he wise. Apa om he GCTA analyses, which uses LD s uc u e o KORA F4, all LD analyses we e pe o med in PLINK based on HapMap 2 (CEU) geno ypes. Fo compa ison o findings be ween loci om di e en A PCs wo loci a e assumed o be iden ical i he opSNPs a e in high LD (LD40.8). Da a a ailabili y.Summa y s a is ics o all analyses can be downloaded om:h ps://www.b oadins i u e.o g/collabo a ion/gian / Re e ences 1. Wood, A. R. e al. Defining he ole o common a ia ion in he genomic and biological a chi ec u e o adul human heigh . Na . Gene . 46, 1173–1186 (2014). 2. Locke, A. E. e al. Gene ic s udies o body mass index yield new insigh s o obesi y biology. Na u e 518, 197–206 (2015). 3. Lango Allen, H. 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 467, 832–838 (2010). 4. Shungin, D. e al. New gene ic loci link adipose and insulin biology o body a dis ibu ion. Na u e 518, 187–196 (2015). 5. Liu, C. T. e al. Genome-wide associa ion o body a dis ibu ion in A ican ances y popula ions sugges s new loci. PLoS. Gene . 9, e1003681 (2013). 6. Kilpelainen, T. O. e al. Gene ic a ia ion nea IRS1 associa es wi h educed adiposi y and an impai ed me abolic p ofile. Na . Gene . 43, 753–760 (2011). 7. B adfield, J. P. e al. A genome-wide associa ion me a-analysis iden ifies new childhood obesi y loci. Na . Gene . 44, 526–531 (2012). 8. Wheele , E. e al. Genome-wide SNP and CNV analysis iden ifies common and low- equency a ian s associa ed wi h se e e ea ly-onse obesi y. Na . Gene . 45, 513–517 (2013). 9. Mey e, D. e al. Genome-wide associa ion s udy o ea ly-onse and mo bid adul obesi y iden ifies h ee new isk loci in Eu opean popula ions. Na . Gene . 41, 157–159 (2009). 10. Myin , P. K., Kwok, C. S., Luben, R. N., Wa eham, N. J. & Khaw, K. T. Body a pe cen age, body mass index and wais - o-hip a io as p edic o s o mo ali y and ca dio ascula disease. Hea 100, 1613–1619 (2014). 11. In e Ac , C. e al. Long- e m isk o inciden ype 2 diabe es and measu es o o e all and egional obesi y: he EPIC-In e Ac case-coho s udy. PLoS Med. 9, e1001230 (2012). 12. He, M. e al. Me a-analysis o genome-wide associa ion s udies o adul heigh in Eas Asians iden ifies 17 no el loci. Hum. Mol. Gene . 24, 1791–1800 (2015). 13. Pende g ass, S. A. e al. Phenome-wide associa ion s udy (PheWAS) o de ec ion o pleio opy wi hin he Popula ion A chi ec u e using Genomics and Epidemiology (PAGE) Ne wo k. PLoS Gene . 9, e1003087 (2013). 14. Denny, J. C. e al. PheWAS: demons a ing he easibili y o a phenome-wide scan o disco e gene-disease associa ions. Bioin o ma ics 26, 1205–1210 (2010). 15. Yang, J. e al. Condi ional and join mul iple-SNP analysis o GWAS summa y s a is ics iden ifies addi ional a ian s influencing complex ai s. Na . Gene . 44, S1–S3 (2012). 16. Ascha d, H. e al. Maximizing he powe o p incipal-componen analysis o co ela ed pheno ypes in genome-wide associa ion s udies. Am. J. Hum. Gene . 94, 662–676 (2014). 17. Polde man, T. J. e al. Me a-analysis o he he i abili y o human ai s based on fi y yea s o win s udies. Na . Gene . 47, 702–709 (2015). 18. Spelio es, E. K. e al. Associa ion analyses o 249,796 indi iduals e eal 18 new loci associa ed wi h body mass index. Na . Gene . 42, 937–948 (2010). 19. Heid, I. M. e al. Me a-analysis iden ifies 13 new loci associa ed wi h wais -hip a io and e eals sexual dimo phism in he gene ic basis o a dis ibu ion. Na . Gene . 42, 949–960 (2010). 20. Innocen i, F. e al. Iden ifica ion, eplica ion, and unc ional fine-mapping o exp ession quan i a i e ai loci in p ima y human li e issue. PLoS Gene . 7, e1002078 (2011). 21. A nold, M., Ra fle , J., P eu e , A., Suh e, K. & Kas enmulle , G. SNiPA: an in e ac i e, gene ic a ian -cen e ed anno a ion b owse . Bioin o ma ics 31, 1334–1336 (2015). 22. Comuzzie, A. G. e al. No el gene ic loci iden ified o he pa hophysiology o childhood obesi y in he Hispanic popula ion. PLoS ONE 7, e51954 (2012). 23. Wel e , D. e al. The NHGRI GWAS Ca alog, a cu a ed esou ce o SNP- ai associa ions. Nucleic Acids Res. 42, D1001–D1006 (2014). 24. Beck, T., Has ings, R. K., Gollapudi, S., F ee, R. C. & B ookes, A. J. GWAS Cen al: a comp ehensi e esou ce o he compa ison and in e oga ion o genome-wide associa ion s udies. Eu . J. Hum. Gene . 22, 949–952 (2014). 25. Li, M. J. e al. GWASdb: a da abase o human gene ic a ian s iden ified by genome-wide associa ion s udies. Nucleic Acids Res. 40, D1047–D1054 (2012). 26. So o-Pan oja, D. R., Kau , S. & Robe s, D. D. CD47 signaling pa hways con olling cellula di e en ia ion and esponses o s ess. C i . Re . Biochem. Mol. Biol. 1–19 (2015). 27. Maimai iyiming, H., No man, H., Zhou, Q. & Wang, S. CD47 deficiency p o ec s mice om die -induced obesi y and imp o es whole body glucose ole ance and insulin sensi i i y. Sci. Rep. 5, 8846 (2015). 28. Pa e nos e , L. e al. Genome-wide popula ion-based associa ion s udy o ex emely o e weigh young adul s-- he GOYA s udy. PLoS ONE 6, e24303 (2011). 29. Zelle , T. e al. Gene ics and beyond-- he ansc ip ome o human monocy es and disease suscep ibili y. PLoS ONE 5, e10693 (2010). 30. Wang, H. D. e al. DNA me hyla ion s udy o e us genome h ough a genome-wide analysis. BMC Med. Genomics 7, 18 (2014). 31. Tegha-Dunghu, J. e al. EML3 is a nuclea mic o ubule-binding p o ein equi ed o he co ec alignmen o ch omosomes in me aphase. J. Cell. Sci. 121, 1718–1726 (2008). 32. Wee, K., Yang, W., Sugii, S. & Han, W. Towa ds a mechanis ic unde s anding o lipodys ophy and seipin unc ions. Biosci. Rep. 34 (2014). 33. Wes a, H. J. e al. Sys ema ic iden ifica ion o ans eQTLs as pu a i e d i e s o known disease associa ions. Na . Gene . 45, 1238–1243 (2013). 34. Ramdas, M., Ha el, C., A moni, M. & Ka nieli, E. AHNAK KO Mice a e p o ec ed om die -induced obesi y bu a e glucose in ole an . Ho m. Me ab. Res. 47, 265–272 (2015). 35. G undbe g, E. e al. Mapping cis- and ans- egula o y e ec s ac oss mul iple issues in wins. Na . Gene . 44, 1084–1089 (2012). 36. Chambe s, J. C. e al. Gene ic loci influencing kidney unc ion and ch onic kidney disease. Na . Gene . 42, 373–375 (2010). 37. Das ani, Z. e al. No el loci o adiponec in le els and hei influence on ype 2 diabe es and me abolic ai s: a mul i-e hnic me a-analysis o 45,891 indi iduals. PLoS Gene . 8, e1002607 (2012). 38. Replica ion, D. I. G. e al. Genome-wide ans-ances y me a-analysis p o ides insigh in o he gene ic a chi ec u e o ype 2 diabe es suscep ibili y. Na . Gene . 46, 234–244 (2014). 39. Sco , R. A. e al. La ge-scale associa ion analyses iden i y new loci influencing glycemic ai s and p o ide insigh in o he unde lying biological pa hways. Na . Gene . 44, 991–1005 (2012). 40. Richa ds, J. B. e al. A genome-wide associa ion s udy e eals a ian s in ARL15 ha influence adiponec in le els. PLoS Gene . 5, e1000768 (2009). 41. Teslo ich, T. M. e al. Biological, clinical and popula ion ele ance o 95 loci o blood lipids. Na u e 466, 707–713 (2010). 42. Global Lipids Gene ics, C e al. Disco e y and efinemen o loci associa ed wi h lipid le els. Na . Gene . 45, 1274–1283 (2013). 43. Glessne , J. T. e al. A genome-wide s udy e eals copy numbe a ian s exclusi e o childhood obesi y cases. Am. J. Hum. Gene . 87, 661–666 (2010). 44. Co ona y A e y Disease Gene ics Conso ium. A genome-wide associa ion s udy in Eu opeans and Sou h Asians iden ifies fi e new loci o co ona y a e y disease. Na . Gene . 43, 339–344 (2011). 45. Aulchenko, Y. S., Ripke, S., Isaacs, A. & an Duijn, C. M. GenABEL: an R lib a y o genome-wide associa ion analysis. Bioin o ma ics 23, 1294–1296 (2007). 46. Wille , C. J., Li, Y. & Abecasis, G. R. METAL: as and e ficien me a-analysis o genomewide associa ion scans. Bioin o ma ics 26, 2190–2191 (2010). 47. Pu cell, S. e al. PLINK: a ool se o whole-genome associa ion and popula ion-based linkage analyses. Am. J. Hum. Gene . 81, 559–575 (2007). 48. Yang, J., Lee, S. H., Godda d, M. E. & Vissche , P. M. GCTA: a ool o genome-wide complex ai analysis. Am. J. Hum. Gene . 88, 76–82 (2011). Acknowledgemen s A comple e lis o acknowledgemen s is desc ibed in Supplemen a y No es. Au ho con ibu ions A de ailed lis o au ho con ibu ions is desc ibed in Supplemen a y No es. Addi ional in o ma ion Supplemen a y In o ma ion accompanies his pape a h p://www.na u e.com/ na u ecommunica ions Compe ing financial in e es s: Ka i S e ansson, Valge du S ein ho sdo i , Gudma Tho lei sson and Unnu Tho s einsdo i a e used by deCODE Gene ics/Amgen inc. Ge ´ a d Waebe and Pe e Vollenweide ecei ed an un es ic ed g an om GSK o build he CoLaus s udy. B uce M. Psa y se es on a DSMB o a clinical ial o a de ice unded by he manu ac u e (Zoll Li eCo ). NATURE COMMUNICATIONS | DOI: 10.1038/ncomms13357 ARTICLE NATURE COMMUNICATIONS | 7:13357 | DOI: 10.1038/ncomms13357 | www.na u e.com/na u ecommunica ions 7 Rep in s and pe mission in o ma ion is a ailable online a h p://npg.na u e.com/ ep in sandpe missions/ How o ci e his a icle: Ried, J. S. e al. A p incipal componen me a-analysis on mul iple an h opome ic ai s iden ifies no el loci o body shape. Na . Commun. 7, 13357 doi: 10.1038/ncomms13357 (2016). Publishe ’s no e: Sp inge Na u e emains neu al wi h ega d o ju isdic ional claims in published maps and ins i u ional a filia ions. This wo k is licensed unde a C ea i e Commons A ibu ion 4.0 In e na ional License. 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 he c edi line; i he ma e ial is no included unde he C ea i e Commons license, use s will need o ob ain pe mission om he license holde o ep oduce he ma e ial. To iew a copy o his license, isi h p://c ea i ecommons.o g/licenses/by/4.0/ The Au ho (s) 2016 Janina S. Ried1,*, Janina M. Je 2,*, Aud ey Y. Chu3, Jenni e L. B agg-G esham4, Jenny an Dongen5, Jenni e E. Hu man6, Ta un ee S. Ahluwalia7,8,9, Gemma Cadby10, Niina Eklund11, Joel E iksson12, To ˜nu Esko13,14,15,16, Ma y F. Fei osa17, Anuj Goel18,19, Ma hias Go ski20,21, Ca oline Haywa d6, Nancy L. Hea d-Cos a22,23, Anne U. Jackson24, Ee o Jokinen25, S a oula Kanoni26,27, Ka i K is iansson11,28, Zol a ´n Ku alik29,30,31, Ja i Lah i32,33, Jian’an Luan34, Reedik Ma ¨gi15,19, Anubha Mahajan19, Massimo Mangino35, Ca olina Medina-Gomez36,37, Ke i L. Monda38,39, Ilja M. Nol e40, Louis Pe ´ usse41,42, Inga P okopenko19,43,44, Lu Qi45,46, Lynda M. Rose3, E ika Sal i47,48, Megan T. Smi h49, Ha old Sniede 40, Alena S anc ˇa ´ko a ´50, Yun Ju Sung51, Ioanna Tachmazidou26, Alexande Teume 52,53, Gudma Tho lei sson54, Pim an de Ha s 55,56,57, Ryan W. Walke 2,58, Sophie R. Wang16,59,60,61, Sa ah H. Wild62, Sa a M. Willems63, And ew Wong64, Weihua Zhang65,66, E a Alb ech 1, Alexessande Cou o Al es67, S ephan J.L. Bakke 68, C is ina Ba lassina47,48, T aci M. Ba z49,69,70, John Beilby71,72, Clai e Bellis73,74, Richa d N. Be gman75, S en Be gmann29,30, John Blange o76, Ma hias Blu¨he 77,78, E ic Boe winkle79, Lo i L. Bonnycas le80, S e an R. Bo ns ein81, Ma cel B uinenbe g82, Ha y Campbell62, Yii-De Ida Chen83, Cha les on W.K. Chiang84, Pe e S. Chines80, F ancis S. Collins80, F acensco Cucca85, L. Ad ienne Cupples22, F ancesca D’A ila47,48, Eco J.C. de Geus5,86, Geo ge Dedoussis87, Ma ia Dimi iou26,87, Angela Do ¨ ing88,89, Johan G. E iksson32,90,91, Aliki-Eleni Fa maki87, Ma in Fa all18,19, Te esa Fe ei a19, K is a Fische 15, Ni a G. Fo ouhi34, Nele F ied ich92, Ane e P io Gjesing7, Nicola Glo ioso93, Ma iaelisa G a 38, Ha ald G alle 89,94,95, Niels G a up7,Ju¨ gen G a ¨le 96, Jag i G ewal65,66, Ande s Hams en97,98,99, Ma ie Nee gaa d Ha de 7, Ca ha ina A. Ha man100, Maija Hassinen101, Nicholas Has ie6, And ew Tym Ha e sley102, Aki S. Ha ulinna11, Ma kku Helio ¨ aa a11, Hans Hillege57, Albe Ho man36, Oddgei Holmen103, Geo g Homu h53, Jouke-Jan Ho enga5, Jennie Hui71,72,104, Lise Lo e Husemoen105, Pi o G. Hysi35, Aa on Isaacs63, Till I e mann52, Shapou Jalilzadeh18,19, Alan L. James106, To ben Jø gensen107,108,109, Pekka Jousilah i11, An i Jula11, Johanne Ma ie Jus esen7, Anne E. Jus ice38, Mika Ka ¨ho ¨nen110,111, Ma ia Ka ale he i112, Kay Tee Khaw113, Si kka M. Keinanen-Kiukaanniemi114,115, Leena Kinnunen116, Paul B. Knek 11, Heikki A. Kois inen116,117,118, I ana Kolcic119, Ishminde K. Koone 66, Seppo Koskinen11, Pe e Ko acs77, Theodosios Ky iakou18,19, Tomi Lai inen120,121, Claudia Langenbe g34,122, Alexand a M. Lewin67, Pe e Lich ne 123, Cecilia M. Lindg en13,19,124, Jaana Linds o ¨m11, Allan Linnebe g105,107,125, Robe o Lo bee 52, Ma ias Lo en zon12, Robe Luben126, Vale iya Lyssenko8,127, Sa u Ma ¨nnis o ¨11, Paolo Manun a128, I ene Ma eo Leach57, Wendy L. McA dle129, Ba ba a Mcknigh 49,70,130, Ka en L. Mohlke131, E elin Mihailo 15, Lili Milani15, Rebecca Mills66, May E. Mon asse 132, And ew P. Mo is19,133, Gab iele Mu¨lle 134, A hu W. Musk135, Na isu Na isu80, Ken K. Ong34,64,136, Ben A. Oos a63, Cli e Osmond137, Aa no Palo ie28,138, James S. Pankow139, La inia Pa e nos e 140, B enda W. Penninx141, I ene Pichle 142,143, Ma ia G. Pilia144, Oz en Polas ˇek62,119, Pe e P. P ams alle 142,143,145,146, Olli T. Rai aka i147,148, Tuomo Rankinen149, D.C. Rao51,150, Nigel W. Rayne 19,44,151, Rasmus Ribel-Madsen7, T e a K. Rice51,150, Ma cus Richa ds44,64, Paul M. Ridke 3,152, Fe nando Ri adenei a36,37, Ka hy A. Ryan132, Se ena Sanna144, Ma k A. Sa zynski149, Salome Schol ens40, Robe A. Sco 34, Syl ain Sebe 43,153,154, Lo aine Sou ham19,26, Thomas Hempel Spa sø7, Valge du S ein ho sdo i 54, Ka hleen S i ups26,27, Ronald P. S olk40, Kons an in S auch1,155, Hea he M. ARTICLE NATURE COMMUNICATIONS | DOI: 10.1038/ncomms13357 8NATURE COMMUNICATIONS | 7:13357 | DOI: 10.1038/ncomms13357 | www.na u e.com/na u ecommunica ions S ingham24, Mo is A. Swe z56, Amy J. Swi 80, Anke To ¨njes78, Emmanouil Tsa an akis156, Pe e J. an de Mos 40, Jana V. Van Vlie -Os ap chouk157, Liesbe h Vandenpu 12, E kki Va iainen11, C is ina Ven u ini35,158, Niek Ve weij57, Jo ma S. Viika i159,160, Ve onique Vi a 6, Ma ie-Claude Vohl42,161, Judi h M. Vonk40,Ge ´ a d Waebe 162, Elisabe h Wide ´n28, Gonneke Willemsen5, Tom Wilsgaa d163, Thomas W. Winkle 21, Alan F. W igh 6, Lau a M. Ye ges-A ms ong132, Jing Hua Zhao34, M. Ca ola Zillikens37, Do e I. Boomsma5, Claude Boucha d149, John C. Chambe s65,66,164, Daniel I. Chasman3,152, Daniele Cusi47,48, Ron T. Ganse oo 68, Ch is ian Giege 1,89,95, To ben Hansen7,165, And ew A. Hicks142,143, F ank Hu45,46, K is ian H eem103, Ma jo-Rii a Ja elin153,166,167,168, Ee o Kajan ie11,90,169,170, Jaspal S. Koone 66,164,171, Diana Kuh64, Johanna Kuusis o50, Ma kku Laakso50, Timo A. Lakka101,172, Te ho Leh ima ¨ki173,174, And es Me spalu15, Inge Njøls ad163,175, Claes Ohlsson12, Albe ine J. Oldehinkel100, Lyle J. Palme 176,177, Olu Pede sen7, Ma kus Pe ola11,15,28, Anne e Pe e s89,95,178, B uce M. Psa y70,179,180,181, Hannu Puolijoki182, Raine Rau amaa101,183, Igo Rudan62, Veikko Salomaa11, Pe e E.H. Schwa z81,184, Alan R. Shudine 132,185, Jan H. Smi 141, Tho kild I.A. Sø ensen7,186,187, Timo hy D. Spec o 35, Ka i S e ansson54,188, Michael S um oll77,78, Angelo T emblay41, Jaakko Tuomileh o189,190,191, And e ´G. Ui e linden36,37, Ma i Uusi upa192,193, Uwe Vo ¨lke 53,194, Pe e Vollenweide 162, Nicholas J. Wa eham34, Hugh Wa kins18,19, James F. Wilson6,62, Ele he ia Zeggini26, Goncalo R. Abecasis24, Michael Boehnke24, Ing id B. Bo ecki17, Panos Deloukas27,151,195, Co nelia M. an Duijn63,196, Ca oline Fox22,152, Lei C. G oop127,197, I is M. Heid21,198, Da id J. Hun e 13,45,46,199, Robe C. Kaplan200, Ma k I. McCa hy19,44,201, Ka i E. No h202, Je ey R. O’Connell132, Da id Schlessinge 203, Unnu Tho s einsdo i 54,188, Da id P. S achan204, Timo hy F ayling205, Joel N. Hi schho n13,14,16,59,60,61,206, Ma ina Mu¨lle -Nu asyid1,155,178,207,** & Ru h J.F. Loos2,34,58,208,209,** 1Ins i u e o Gene ic Epidemiology, Helmhol z Zen um Mu ¨nchen-Ge man Resea ch Cen e o En i onmen al Heal h, 85764 Neuhe be g, Ge many. 2The Cha les B on man Ins i u e o Pe sonalized Medicine, The Icahn School o Medicine a Moun Sinai, New Yo k, New Yo k 10029, USA. 3Di ision o P e en i e Medicine, B igham and Women’s Hospi al, Bos on, Massachuse s 02215, USA. 4Kidney Epidemiology and Cos Cen e , In e nal Medicine-Neph ology, Uni e si y o Michigan, Ann A bo , Michigan 48109, USA. 5Depa men o Biological Psychology, VU Uni e si y, 1081BT Ams e dam, The Ne he lands. 6MRC Human Gene ics Uni , Ins i u e o Gene ics and Molecula Medicine, Uni e si y o Edinbu gh, EH4 2XU Edinbu gh, Sco land. 7Facul y o Heal h and Medical Sciences, No o No disk Founda ion Cen e o Basic Me abolic Resea ch, Sec ion o Me abolic Gene ics, Uni e si y o Copenhagen, 2100 Copenhagen, Denma k. 8S eno Diabe es Cen e A/S, DK-2820 Gen o e, Denma k. 9COPSAC, Copenhagen P ospec i e S udies on As hma in Childhood, He le and Gen o e Hospi al, Uni e si y o Copenhagen, Led ebo g Alle ´34, DK-2820 Copenhagen, Denma k. 10 Cen e o Gene ic O igins o Heal h and Disease, Uni e si y o Wes e n Aus alia, C awley, Wes e n Aus alia 6009, Aus alia. 11 Depa men o Heal h, Na ional Ins i u e o Heal h and Wel a e (THL), FI-00271 Helsinki, Finland. 12 Depa men o In e nal Medicine and Clinical Nu i ion, Cen e o Bone and A h i is Resea ch, Ins i u e o Medicine, Sahlg enska Academy, Uni e si y o Go henbu g, 413 45 Go henbu g, Sweden. 13 B oad Ins i u e o he Massachuse s Ins i u e o Technology and Ha a d Uni e si y, Camb idge, Massachuse s 2142, USA. 14 Di isions o Endoc inology and Gene ics and Cen e o Basic and T ansla ional Obesi y Resea ch, Bos on Child en’s Hospi al, Bos on, Massachuse s 02115, USA. 15 Es onian Genome Cen e , Uni e si y o Ta u, Ta u 51010, Es onia. 16 Depa men o Gene ics, Ha a d Medical School, Bos on, Massachuse s 02115, USA. 17 Di ision o S a is ical Genomics, Depa men o Gene ics, Washing on Uni e si y School o Medicine, S . Louis, Missou i 63108, USA. 18 Di ision o Ca dio ascula Medicine, Radcli e Depa men o Medicine, Uni e si y o Ox o d, Ox o d OX3 9DU, UK. 19 Wellcome T us Cen e o Human Gene ics, Uni e si y o Ox o d, Ox o d OX3 7BN, UK. 20 Depa men o Neph ology, Uni e si y Hospi al Regensbu g, 93042 Regensbu g, Ge many. 21 Depa men o Gene ic Epidemiology, Ins i u e o Epidemiology and P e en i e Medicine, Uni e si y o Regensbu g, 93053 Regensbu g, Ge many. 22 Na ional Hea , Lung, and Blood Ins i u e, he F amingham Hea S udy, F amingham, Massachuse s 01702, USA. 23 Depa men o Neu ology, Bos on Uni e si y School o Medicine, Bos on, Massachuse s 02118, USA. 24 Depa men o Bios a is ics, Cen e o S a is ical Gene ics, Uni e si y o Michigan, Ann A bo , Michigan 48109, USA. 25 Hospi al o Child en and Adolescen s, Uni e si y o Helsinki, FI-00290 Helsinki, Finland. 26 Wellcome T us Sange Ins i u e, Human Gene ics, Hinx on, Camb idge CB10 1SA, UK. 27 William Ha ey Resea ch Ins i u e, Ba s and The London School o Medicine and Den is y, Queen Ma y Uni e si y o London, London EC1M 6BQ, UK. 28 Ins i u e o Molecula Medicine Finland, Uni e si y o Helsinki, FI-00290 Helsinki, Finland. 29 Swiss Ins i u e o Bioin o ma ics, 1015 Lausanne, Swi ze land. 30 Depa men o Medical Gene ics, Uni e si y o Lausanne, Lausanne, 1005, Swi ze land. 31 Ins i u e o Social and P e en i e Medicine, Uni e si y Hospi al Lausanne (CHUV), 1010 Lausanne, Swi ze land. 32 Folkha ¨lsan Resea ch Cen e, FI-00290 Helsinki, Finland. 33 Ins i u e o Beha iou al Sciences, Uni e si y o Helsinki, FI-00014 Helsinki, Finland. 34 MRC Epidemiology Uni , Uni e si y o Camb idge School o Clinical Medicine, Ins i u e o Me abolic Science, Uni e si y o Camb idge, Camb idge Biomedical Campus, Camb idge CB2 0QQ, UK. 35 Depa men o Twin Resea ch and Gene ic Epidemiology, King’s College London, London SE1 7EH, UK. 36 Depa men o Epidemiology, E asmus Medical Cen e , 3015GE Ro e dam, The Ne he lands. 37 Depa men o In e nal Medicine, E asmus Medical Cen e , 3015GE Ro e dam, The Ne he lands. 38 Depa men o Epidemiology, Uni e si y o No h Ca olina a Chapel Hill, Chapel Hill, No h Ca olina 27599, USA. 39 The Cen e o Obse a ional Resea ch, Amgen Inc., Thousand Oaks, Cali o nia 91320-1799, USA. 40 Depa men o Epidemiology, Uni e si y o G oningen, Uni e si y Medical Cen e G oningen, 9700 RB G oningen, The Ne he lands. 41 Depa men o Kinesiology, La al Uni e si y, Que ´bec, Que ´bec, Canada G1V 0A6. 42 Ins i u e o Nu i ion and Func ional Foods, La al Uni e si y, Que ´bec, Que ´bec, Canada G1V 0A6. 43 Depa men o Genomics o Common Disease, School o Public Heal h, Impe ial College London, London W12 0NN, UK. 44 Ox o d Cen e o Diabe es, Endoc inology and Me abolism, Uni e si y o Ox o d, Chu chill Hospi al, Ox o d OX3 7LJ, UK. 45 Depa men o Medicine, Channing Di ision o Ne wo k Medicine, B igham and Women’s Hospi al and Ha a d Medical School, Bos on, Massachuse s 02115, USA. 46 Depa men o Nu i ion, Ha a d School o Public Heal h, Bos on, Massachuse s 02115, USA. 47 Depa men o Heal h Sciences, Uni e si y o Milano a San Paolo Hospi al, 20139 Milano, I aly. 48 Fila e e Founda ion, NATURE COMMUNICATIONS | DOI: 10.1038/ncomms13357 ARTICLE NATURE COMMUNICATIONS | 7:13357 | DOI: 10.1038/ncomms13357 | www.na u e.com/na u ecommunica ions 9