Associations of functional alanine-glyoxylate aminotransferase 2 gene variants with atrial fibrillation and ischemic stroke

1 Scien i ic RepoR s | 6:23207 | DOI: 10.1038/s ep23207 www.na u e.com/scien i ic epo s Associa ions o unc ional alanine- glyoxyla e amino ans e ase 2 gene a ian s wi h a ial ib illa ion and ischemic s oke Ilkka Seppälä1, Ma cus E. Klebe 2, S e e Be an3, Leo-Pekka Lyy ikäinen1, Niku Oksala1,4, Jussi A. He nesniemi1,5, Ka i-Ma i Mäkelä1, Pe e M. Ro hwell6, Ca hie Sudlow7, Ma in Dichgans8, Nina Mononen1, E hymia Vlachopoulou9, Juha Sinisalo10, G aciela E. Delgado2, Reijo Laaksonen1, Tuomas Koskinen11,12, Hube Scha nagl13, Mika Kähönen14, Hugh S. Ma kus15, Win ied Mä z2,13,16 & Te ho Leh imäki1 Asymme ic and symme ic dime hyla ginines (ADMA and SDMA) impai ni ic oxide bioa ailabili y and ha e been implica ed in he pa hogenesis o a ial ib illa ion (AF). Alanine–glyoxyla e amino ans e ase 2 (AGXT2) is he only enzyme capable o me abolizing bo h o he dime hyla ginines. We hypo hesized ha wo unc ional AGXT2 missense a ian s ( s37369, V140I; s16899974, V498L) a e associa ed wi h AF and i s ca dioembolic complica ions. Associa ion analyses we e conduc ed using 1,834 indi idulas wi h AF and 7,159 una ec ed indi iduals om wo co ona y angiog aphy coho s and a coho comp ising pa ien s unde going clinical exe cise es ing. In co ona y angiog aphy pa ien s wi hou s uc u al hea disease, he mino A allele o s16899974 was associa ed wi h any AF (OR = 2.07, 95% CI 1.59-2.68), and wi h pa oxysmal AF (OR = 1.98, 95% CI 1.44–2.74) and ch onic AF (OR = 2.03, 95% CI 1.35–3.06) sepa a ely. We could no eplica e he associa ion wi h AF in he o he wo coho s. Howe e , he A allele o s16899974 was nominally associa ed wi h ischemic s oke isk in he me a-analysis o WTCCC2 ischemic s oke coho s (3,548 cases, 5,972 con ols) and wi h ea lie onse o i s -e e ischemic s oke (360 cases) in he coho o clinical exe cise es pa ien s. In conclusion, AGXT2 a ia ions may be in ol ed in he pa hogenesis o AF and i s age- ela ed h omboembolic complica ions. Ci cula ing asymme ic and symme ic dime hyla ginine (ADMA and SDMA) p edic mo ali y in ca dio ascu- la diseases1,2. Mo eo e , accumula ing epidemiological and expe imen al e idence ha e implica ed ADMA in he pa hogenesis o a ial ib illa ion (AF)3–7, whe eas li le a en ion has been paid o he ole o SDMA2. In pa ien s 1Depa men o Clinical Chemis y, Fimlab Labo a o ies and School o Medicine, Uni e si y o Tampe e, Tampe e, Finland. 2V h Depa men o Medicine (Neph ology, Hype ensiology, Endoc inology, Diabe ology, Rheuma ology), Medical Facul y Mannheim, Uni e si y o Heidelbe g, Heidelbe g, Ge many. 3School o Li e Science, Uni e si y o Lincoln, Lincoln, UK. 4Di ision o Vascula Su ge y, Depa men o Su ge y, Tampe e Uni e si y Hospi al, Tampe e, Finland. 5Hea Hospi al, Tampe e Uni e si y Hospi al, Tampe e, Finland. 6S oke P e en ion Resea ch Uni , Nu ield Depa men o Clinical Neu oscience, Uni e si y o Ox o d, Ox o d, UK. 7Di ision o Clinical Neu osciences and Insi i u e o Gene ics and Molecula Medicine, Uni e si y o Edinbu gh, UK. 8Ins i u e o S oke and Demen ia Resea ch, Klinikum de Uni e si ä München, Ludwig-Maximilians-Uni e si ä , Munich, Ge many & Munich Clus e o Sys ems Neu ology (SyNe gy), Munich, Ge many. 9T ansplan a ion Labo a o y, Haa man Ins i u e, Uni e si y o Helsinki, Helsinki, Finland. 10Hea and Lung Cen e , Helsinki Uni e si y Hospi al and Helsinki Uni e si y, Helsinki, Finland. 11Resea ch Cen e o Applied and P e en i e Ca dio ascula Medicine, Uni e si y o Tu ku, Tu ku, Finland. 12Sa akun a Cen al Hospi al, Depa men o Su ge y, Po i, Finland. 13Clinical Ins i u e o Medical and Chemical Labo a o y Diagnos ics, Medical Uni e si y o G az, G az, Aus ia. 14Depa men o Clinical Physiology, Tampe e Uni e si y Hospi al and Uni e si y o Tampe e, Tampe e, Finland. 15Clinical Neu osciences, Uni e si y o Camb idge, Camb idge, UK. 16Synlab Academy, Synlab Se ices GmbH, Mannheim, Ge many. Co espondence and eques s o ma e ials should be add essed o I.S. (email: [email p o ec ed]) Recei ed: 27 No embe 2015 Accep ed: 02 Ma ch 2016 Published: 17 Ma ch 2016 OPEN www.na u e.com/scien i ic epo s/ 2 Scien i ic RepoR s | 6:23207 | DOI: 10.1038/s ep23207 wi h a his o y o AF, he induc ion o AF du ing ca he e abla ion esul ed in signi ican ly ele a ed ADMA le els in he a ia and sys emic ci cula ion7, whe eas in ano he s udy elec ical ca dio e sion o AF dec eased ADMA le els o no mal alues wi hin 24 hou s a e AF was e mina ed6. Simila ly, highe ADMA le els we e associa ed wi h he ecu ence o AF a e elec ical ca dio e sion o ca he e abla ion in pa ien s wi h pe sis en AF3,5. These da a sugges s ha ele a ed ADMA le els a e no only a consequence o AF bu may also be in ol ed in he de elopmen o AF by modi ying he a ial subs a es o AF. Indeed, al hough only ADMA di ec ly inhibi s ni ic oxide syn hase (NOS), bo h dime hyla ginines can educe he bioa ailabili y o ni ic oxide (NO) by si ing he NOS enzyme om he p oduc ion o NO o supe oxide anion8–10, and hus p omo e oxida i e s ess, endo helial dys unc ion, and in lamma ion, all o which a e associa ed wi h a ial emodelling and inc eased ulne abili y o AF11–13. The p ima y pa hway o ADMA deac i a ion is ca alysed by he dime hyla ginine dime hylaminohyd olase (DDAH) enzymes, whe eas he enal clea ance plays a majo ole in he elimina ion o SDMA om he sys- emic ci cula ion. In con as o DDAH, alanine–glyoxyla e amino ans e ase 2 (AGXT2), a nuclea -encoded mi ochond ial p o ein, can me abolise no only ADMA bu also SDMA in humans14,15. In addi ion o a s ong exp ession o AGXT2 in he kidney and li e , AGXT2 mRNA exp ession was ecen ly de ec ed in se e al o he human o gans including he hea 16. As he ac i i y o DDAH is educed and he gene a ion o ADMA inc eased ia up- egula ion o p o ein a ginine me hyl ans e ase 1 (PRMT1) in animal ib illa ing a ia4 also explaining he inc eased ADMA le els in human AF, i can be hypo hesized ha he al e na i e AGXT2-media ed elimina ion pa hway o dime hyla ginines could play an impo an ole in he pa hogenesis o AF. We ecen ly ine-mapped he AGXT2 gene egion on ch omosome 5p13 o associa ions be ween single-nucleo ide polymo phisms (SNP) and se um SDMA le els and iden i ied wo missense mu a ions in AGXT2 ( s37369, p.V140I; s16899974, p.V498L) o be s ongly and independen ly associa ed wi h SDMA le els in a me a-analysis o wo la ge coho s o Eu opean descen 15. Expe imen al da a indica es ha s37369 modi ies he enzyme ac i i y16,17 whe eas an in silico analysis showed ha he s16899974 a ian may modi y he enzyme s abili y17. The e o e, we conside ed ha he wo unc ional AGXT2 a ian s may se e as na u ally occu ing gene ic models o s udy he ole o AGXT2 in human AF and i s h omboembolic complica ions, i.e. s oke and i s sub ypes. Mo e speci ically, we hypo hesized ha hese unc ional a ian s a e (1) associa ed wi h pa oxysmal and/o ch onic AF in pa ien s e e ed o co ona y angiog aphy, (2) associa ed mo e s ongly wi h ci cula ing le els o ADMA and SDMA in pa ien s wi h AF compa ed o pa ien s in sinus hy hm, and (3) associa ed wi h ischemic s oke and i s sub ypes. As a gene ic con ol, we epea ed some o he analyses wi h a known a ian a 4q25 p e iously associa ed wi h AF. Resul s Associa ions o AGXT2 and 4q25 con ol a ian s wi h a ial ib illa ion and i s sub ypes in pa ien s e e ed o co ona y angiog aphy (LURIC and Co ogene). The clinical cha ac e is ics o he s udy pa icipan s o all s udies a e shown in Supplemen a y Table 1. In LURIC, 381 (13.0%) pa ien s had p e alen AF a baseline o which 175 (50.3%) and 173 (49.7%) we e u he classi ied as pa oxysmal o ch onic AF, espec i ely. Those wi h AF we e olde (mean age 66.4 e sus 62.2 yea s, P < 0.0001) and had less likely co - ona y a e y disease (56.8% e sus 71.6%, P < 0.0001) han pa ien s in sinus hy hm. The associa ions o AGXT2 and 4q25 a ian s wi h AF and i s sub ypes a e shown in Fig.1. In he whole s udy popula ion, s16899974 showed a signi ican associa ion wi h any AF (P = 6.6 × 10−4) in a ully adjus ed model. Mo eo e , when ana- lysing pa oxysmal and ch onic AF cases sepa a ely, he associa ion was especially ma ked wi h he pa oxysmal sub ype (P = 1.8 × 10−4), whe eas he e was no associa ion wi h ch onic AF (P = 0.53). We ound s ong e idence ha a his o y o al ula hea disease a enua ed he e ec o s16899974 and s37369 on AF (p o in e ac- ion o 4.3 × 10−4 and 0.015, espec i ely, Table1). Nominally signi ican in e ac ion was also obse ed be ween s16899974 and p e alen ca diomyopa hy on AF (P = 0.043). When excluding he pa ien s wi h s uc u al hea Figu e 1. Associa ions o he AGXT2 and 4q25 a ian s wi h p e alen AF and i s sub ypes in LURIC. ORs pe one mino allele inc ease om logis ic eg ession models assuming addi i e gene ic e ec a e shown. Model 1: adjus ed o age, sex and body mass index. Model 2: Model 1 u he adjus ed o a e ial hype ension, diabe es, co ona y a e y disease (> 50% s enosis), se um NT-p oBNP and eGFR. S uc u al hea disease e e s o ca diomyopa hy o al ula hea disease. www.na u e.com/scien i ic epo s/ 3 Scien i ic RepoR s | 6:23207 | DOI: 10.1038/s ep23207 disease om he analysis, a highly signi ican associa ion o AGXT2 s16899974 was obse ed wi h any AF (P = 3.9 × 10−8), and wi h pa oxysmal AF (P = 3.0 × 10−5) and ch onic AF (P = 7.0 × 10−4) sepa a ely. In con- as o he AGXT2 SNPs, he 4q25 a ian showed a s onge associa ion wi h ch onic AF han pa oxysmal AF in bo h he whole s udy popula ion and when excluding he pa ien s wi h s uc u al hea disease (Fig.1). When conduc ing a genome-wide scan using he ully adjus ed model in pa ien s wi hou s uc u al hea disease, no addi ional genome-wide signi ican associa ions we e iden i ied apa om he signal a he AGXT2 locus (da a no shown). We sough o eplica e he associa ions o he AGXT2 a ian s wi h AF and i s sub ypes in an independen coho . In con as o LURIC, all subjec s in he Co ogene s udy wi h geno ype da a a ailable had an acu e co o- na y synd ome (Supplemen a y Table 1). O he 2,208 pa ien s included in he analysis, 265 (12.0%) had p e a- len AF o which 141 (6.4%) and 107 (4.8%) had pa oxysmal and ch onic AF, espec i ely. No associa ions we e obse ed be ween AGXT2 a ian s and AF o i s sub ypes in Co ogene (all P > 0.05, Supplemen a y Table 2). Howe e , in acco dance wi h he associa ions seen in LURIC, he 4q25 a ian showed signi ican associa ions wi h any AF (OR = 1.50, 95% CI 1.18–1.91, P = 0.001) and ch onic AF (OR = 1.84, 95% CI 1.29–2.61, P = 0.001) bu no wi h pa oxysmal AF (OR = 1.25, 95% CI 0.89–1.74, P = 0.20). Associa ions wi h inciden clinical AF and age a AF onse (FINCAVAS). O he 3,862 FINCAVAS pa ien s unde going clinical exe cise es be ween 2001 and 2008, 1,188 (30.8%) had hei i s AF e en diag- nosed be ween 1987 and 2015. The mean (SD) age a he diagnosis was 60.5 (13.3) and 64.6 (13.7) yea s o men and women, espec i ely (Supplemen a y Figu e 1). The AF- ee su i al cu es s a i ied by he s16899974 gen- o ypes o all pa ien s and AF cases only a e displayed in Fig.2A,B, and o he AGXT2 s37369 and 4q25 a ian s in Supplemen a y Figu es 2 and 3. No associa ions we e seen be ween he s16899974 geno ypes and inciden AF ei he in he whole s udy popula ion o in he case-only-analysis (bo h P > 0.05). As seen in Fig.2A,B, he su i al cu es o he h ee geno ype g oups s a o sepa a e only a e a ound age 75, sugges ing ha o he ac o s a e esponsible o clinical AF e en s in younge pa ien s. In a mul i a iable analysis, s16899974 was associa ed wi h inciden AF and age a AF diagnosis in pa ien s aged ≥ 75 yea s a he end o he ollow-up, bu no in pa ien s aged < 75 yea s a censo ing o da a (Table2, Supplemen a y Table 3). The 4q25 a ian showed a s ong associa ion in he whole s udy popula ion; howe e , s16899974 appea s o be mo e s ongly associa ed wi h inciden AF in pa ien s aged ≥ 75 yea s. Associa ions wi h his o y o ischemic s oke and i s sub ypes (WTCCC2). We examined whe he he wo unc ional AGXT2 a ian s a e associa ed wi h ischemic s oke and i s sub ypes in he me a-analysis o WTCCC2 ischemic s oke coho s. As shown in Supplemen a y Table 4, he e was no signi ican associa ion o s37369 wi h all-cause ischemic s oke (IS) whe eas s16899974 showed s a is ically signi ican associa ions wi h all-cause IS and la ge-a e y a he oscle osis (LAA) s oke sub ype wi h pe A allele ORs o 1.04 (95% CI 1.00– 1.08, P = 0.032) and 1.10 (95% CI 1.01–1.19, P = 0.021), espec i ely. Mo eo e , s16899974 showed a bo de line signi ican associa ion in he same di ec ion o ca dioembolic (CE) s oke wi h an OR o 1.08 (95% CI 1.00–1.16, P = 0.057). The SNPs we e no associa ed wi h small- essel disease (SVD) s oke sub ype. AF isk ac o Risk ac o cases/con ols AF cases/con ols AGXT2 s37369 (T) AGXT2 s16899974 (A) 4q25 s6817105 (C) ORin (95% CI) pORin (95% CI) pORin (95% CI) p A e ial hype ension 1,721/1,202 381/2,542 0.99 (0.56–1.74) 0.98 0.88 (0.60–1.28) 0.49 0.68 (0.46–1.02) 0.065 Diabe es 1,181/1,742 381/2,542 0.81 (0.46–1.40) 0.45 0.82 (0.56–1.19) 0.30 0.91 (0.61–1.36) 0.65 Val ula hea disease 518/2,405 381/2,542 0.45 (0.23–0.86) 0.015 0.46 (0.30–0.71) 0.00043 1.00 (0.63–1.59) 0.99 Ao ic s enosis 146/2,405 272/2,279 0.45 (0.09–2.19) 0.32 0.46 (0.18–1.15) 0.096 1.46 (0.57–3.73) 0.43 Ao ic insu iciency 80/2,403 269/2,214 0.18 (0.02–1.58) 0.12 0.38 (0.14–1.05) 0.063 1.37 (0.47–4.03) 0.57 Mi al s enosis 16/2,405 260/2,161 0.10 (0.01–1.18) 0.068 0.19 (0.02–1.59) 0.13 0.44 (0.08–2.35) 0.34 Mi al insu iciency 225/2,394 315/2,304 0.36 (0.15–0.88) 0.025 0.48 (0.27–0.85) 0.012 0.88 (0.47–1.63) 0.68 O he 34/2,403 262/2,175 2.66 (0.42–16.9) 0.30 0.83 (0.25–2.76) 0.76 0.39 (0.07–2.01) 0.26 Ca diomyopa hy 307/2,616 381/2,542 0.79 (0.37–1.68) 0.54 0.61 (0.38–0.99) 0.043 1.31 (0.75–2.27) 0.34 Ischemic 134/2,612 326/2,420 0.62 (0.21–1.81) 0.38 0.58 (0.29–1.16) 0.12 1.10 (0.52–2.34) 0.79 Dila ed 147/2,614 335/2,426 0.76 (0.25–2.26) 0.62 0.46 (0.24–0.90) 0.024 1.34 (0.62–2.91) 0.46 Res ic ed 24/2,616 296/2,344 1.76 (0.10–32.5) 0.70 2.52 (0.46–13.8) 0.29 0.67 (0.08–5.73) 0.71 Myoca dial in a c ion 1,217/1,706 381/2,542 1.43 (0.81–2.51) 0.21 1.22 (0.82–1.80) 0.33 0.85 (0.55–1.30) 0.46 CVD e en (s oke/TIA) 264/2,659 381/2,542 1.27 (0.58–2.78) 0.55 1.53 (0.86–2.74) 0.15 1.18 (0.64–2.18) 0.59 Table 1. E ec modi ica ion o AGXT2 and 4q25 a ian s on a ial ib illa ion in LURIC. S a is ics: Resul s a e om logis ic eg ession analyses adjus ed o age, sex and BMI. In e ac ion e ec s on a mul iplica i e scale (ORin ) pe one mino allele inc ease assuming addi i e gene ic model a e shown. ORin = 1 means no in e ac ion on a mul iplica i e scale. No es: In he subg oup analyses o al e disease and ca diomyopa hy, con ols did no ha e any al ula disease o ca diomyopa hy, espec i ely. Abb e ia ions: CVD, ce eb o ascula disease; TIA, ansien ischemic a ack. www.na u e.com/scien i ic epo s/ 4 Scien i ic RepoR s | 6:23207 | DOI: 10.1038/s ep23207 Associa ions wi h inciden ischemic s oke and age a i s s oke diagnosis (FINCAVAS). O he 3,862 FINCAVAS pa ien s unde going clinical exe cise es be ween 2001 and 2008, 360 (9.3%) had hei i s ischemic s oke diagnosed be ween 1987 and 2015. The mean (SD) age a he diagnosis was 65.9 (10.7) and Figu e 2. Kaplan-Meie e en - ee su i al o pa ien s pa icipa ing he FINCAVAS s udy. Su i al cu es a e shown as a unc ion o he AGXT2 s16899974 geno ype g oups o inciden AF (A) and ischemic s oke (C) and o cases o inciden AF (B) and ischemic s oke (D) only (age o onse analysis). Locus SNP cases/N HR (95% CI) P All AGXT2 s37369 972/3,122 1.02 (0.87, 1.19) 0.83 AGXT2 s16899974 1188/3,862 1.05 (0.96, 1.16) 0.28 4q25 s6817105 972/3,122 1.51 (1.36, 1.69) 9.7 × 10−14 Age < 75 yea s AGXT2 s37369 802/2,562 0.95 (0.80, 1.12) 0.54 AGXT2 s16899974 991/3,210 0.95 (0.85, 1.05) 0.32 4q25 s6817105 808/2,587 1.55 (1.37, 1.74) 6.1 × 10−13 Age ≥ 75 yea s AGXT2 s37369 170/560 1.07 (0.71, 1.61) 0.75 AGXT2 s16899974 197/662 1.38 (1.10, 1.74) 0.0063 4q25 s6817105 170/560 1.26 (0.95, 1.67) 0.11 Table 2. Associa ions o AGXT2 and 4q25 a ian s wi h inciden clinical AF in FINCAVAS. S a is ics: HRs pe one mino allele inc ease a e om Cox eg ession models, assuming an addi i e gene ic e ec . The baseline haza d unc ion o Cox models a e s a i ied by sex. No es: Age is used as he ime scale in Cox models. Analyses a e s a i ied based on he age a he end o he ollow-up. www.na u e.com/scien i ic epo s/ 5 Scien i ic RepoR s | 6:23207 | DOI: 10.1038/s ep23207 68.5 (12.4) yea s o men and women, espec i ely. O he 360 inciden ischemic s oke cases, 83 (23%) had a p e ious AF diagnosis. The ischemic s oke- ee su i al cu es s a i ied by he s16899974 geno ypes o all pa ien s and ischemic s oke cases only a e displayed in Fig.2C,D, and o he AGXT2 s37369 and 4q25 a ian s in Supplemen a y Figu es 2 and 3. None o he h ee SNPs showed a s a is ically signi ican associa ion wi h inciden ischemic s oke. Howe e , s16899974 was associa ed wi h age a onse o a i s ischemic s oke bo h in he uni a ia e analysis (Fig.2D) and a e con olling o sex and p e iously diagnosed clinical AF (Table3). In e es ingly, his inding was seen only in pa ien s wi h c yp ogenic ischemic s oke (n = 233) wi h a pe mino allele HR o 1.57 (95% CI 1.21–2.04, P = 0.0007) and app oxima ely 3.8 (95% CI 1.5–6.0, P = 0.0009) yea s ea - lie age a he diagnosis wi h each addi ional copy o he mino allele bu no in hose wi h ano he diagnosis o ischemic s oke (n = 127, bo h P > 0.05). Associa ions wi h ci cula ing dime hyla ginines in pa ien s wi h sinus hy hm and AF (LURIC). As displayed in Fig.3A, he e we e s a is ically signi ican s epwise inc eases o ADMA and SDMA le els in pa ien s wi h pa oxysmal AF and ch onic AF compa ed o pa ien s in sinus hy hm. The end ac oss he AF s a- us g oups o ADMA emained highly signi ican a e adjus ing o age, sex and es ima ed glome ula il a ion a e (eGFR) (p o linea end = 1.9 × 10−9) whe eas he associa ion wi h SDMA was a enua ed, bu s ill signi - ican (p o linea end = 0.0023). The unadjus ed associa ions o he AGXT2 a ian s wi h ADMA, SDMA and eGFR by AF s a us a e shown in Fig.3B. As we expec ed, bo h SNPs showed s ong associa ions wi h ci cula ing SDMA le els in pa ien s wi h sinus hy hm. Nei he o he AGXT2 a ian s was associa ed wi h ADMA o eGFR in any o he AF s a us g oups; howe e , s16899974 showed a bo de line signi ican associa ion wi h ADMA in pa ien s wi h pa oxysmal AF (P = 0.057). Mo eo e , i seems ha he e ec es ima es o SDMA in he pa oxys- mal AF g oup a e la ge han in hose wi h no AF o ch onic AF (Fig.3B). To es his hypo hesis he in e ac ion be ween he AF s a us g oups and he AGXT2 geno ypes on SDMA le els we e es ed using a wo-way ANOVA model wi h in e ac ion. S a is ically signi ican in e ac ion was obse ed o s16899974 in unadjus ed model (F = 3.8, P = 0.0044), and when age, sex and eGFR we e used as co a ia es (F = 2.6, P = 0.036). No such s a is i- cally signi ican in e ac ions we e obse ed o s37369 o when using ADMA as a dependen a iable. Discussion In he p esen s udy, we analysed he p esence o associa ions o wo unc ional missense AGXT2 a ian s wi h AF and i s h omboembolic complica ions in ou independen s udy coho s. The p.V498L a ian ( s16899974) was associa ed wi h an inc eased isk o bo h pa oxysmal and ch onic AF in pa ien s e e ed o co ona y angi- og aphy bu wi h no s uc u al hea disease. Mo eo e , he same A allele o s16899974 was nominally associa ed wi h an inc eased isk o any ischemic s oke in he me a-analysis o WTCCC2 ischemic s oke coho s. Finally, s16899974 was associa ed wi h an ea lie age a he i s ischemic s oke diagnosis in pa ien s unde going exe - cise s ess es ing and wi h inciden clinical AF in pa ien s aged ≥ 75 yea s a he ime o AF diagnosis in he same coho . The wo s udied AGXT2 SNPs ha e no been iden i ied o be associa ed a a genome-wide signi ican le el wi h he isk o p e alen o inciden AF18 o ischemic s oke sub ypes19 in p e ious la ge scale me a-analyses o genome-wide associa ion s udies. This is in acco dance wi h he da a om he FINCAVAS s udy showing lack o associa ion wi h inciden AF and ischemic s oke in he whole s udy popula ion. Fu he mo e, no associa- ions we e obse ed wi h p e alen AF o i s sub ypes in he Co ogene s udy including co ona y angiog aphy pa ien s wi h acu e co ona y synd ome. In s iking con as wi h hese indings we ound a genome-wide signi i- can associa ion o s16899974 wi h any AF in pa ien s wi hou s uc u al hea disease in he LURIC s udy. This Locus SNP E ec allele EAF cases/N HR (95% CI) P a) inciden i s e e ischemic s oke AGXT2 s37369 T 0.094 295/3,122 0.97 (0.73, 1.30) 0.86 AGXT2 s16899974 A 0.240 360/3,862 1.05 (0.88, 1.24) 0.61 4q25 s6817105 C 0.156 295/3,122 1.22 (0.90, 1.40) 0.96 b) age a he i s ischemic s oke diagnosis Linea eg ession Locus SNP E ec allele EAF N β (95% CI) P AGXT2 s37369 T 0.086 295 − 2.60 (− 5.75, 0.55) 0.11 AGXT2 s16899974 A 0.232 360 − 3.24 (− 5.24, − 1.25) 0.0015 4q25 s6817105 C 0.151 295 − 2.60 (− 5.14, − 0.074) 0.044 Cox model Locus SNP E ec allele EAF N HR (95% CI) P AGXT2 s37369 T 0.086 295 1.34 (0.99, 1,81) 0.062 AGXT2 s16899974 A 0.232 360 1.44 (1.19, 1.75) 0.00018 4q25 s6817105 C 0.151 295 1.18 (0.93, 1.51) 0.18 Table 3. Associa ions o he AGXT2 and 4q25 a ian s wi h (a) inciden ischemic s oke and (b) age a ischemic s oke onse in FINCAVAS. S a is ics: HRs and β s (in yea s) pe one mino allele inc ease a e om Cox and linea eg ession models, espec i ely, assuming an addi i e gene ic e ec . Models a e con olled o sex and a his o y o clinical a ial ib illa ion. www.na u e.com/scien i ic epo s/ 6 Scien i ic RepoR s | 6:23207 | DOI: 10.1038/s ep23207 associa ion is suppo ed wi h he obse a ion in FINCAVAS showing ha s16899974 is associa ed wi h hospi al discha ge da a based inciden clinical AF in pa ien s aged 75 and o e a he ime o AF diagnosis. Al hough, he e is a sepa a e me a-analysis o lone AF20, de ined as AF wi h an onse be o e 66 yea s o age in indi iduals wi hou o e hea disease, no la ge-scale s udies ha e sough o iden i y common gene ic a ian s unde lying AF isk in he elde ly o o pa oxysmal and mo e pe sis en o ms o AF sepa a ely. In e es ingly, s6817105 a 4q25, p e iously associa ed wi h AF in genome-wide associa ion s udies, was mainly associa ed wi h p e alen ch onic AF in bo h LURIC and Co ogene. These associa ions sugges ha he isk allele ca ie s ha e a ela i ely high a e o p og ession om pa oxysmal AF o ch onic AF, a leas in pa ien s e e ed o co ona y angiog aphy. In con as , he associa ions o s16899974 wi h bo h pa oxysmal and ch onic AF in LURIC, wi h la e onse clinical AF and wi h ea lie age o onse o c yp ogenic ischemic s oke sugges a ole o AGXT2 in mo e subclinical o ms o AF cha ac e ized by a ela i ely low a e o p og ession o ch onic AF; howe e , his hypo hesis needs o be alida ed in a p ospec i e se ing. In line wi h he AF associa ions in LURIC and FINCAVAS, he AF isk allele o s16899974 was associa ed wi h an ea lie onse o c yp ogenic ischemic s oke al hough no associa ion was obse ed wi h he ischemic s oke isk in he whole FINCAVAS s udy popula ion. The associa ion wi h age a ischemic s oke diagnosis was s eng hened a e adjus ing he analysis o gende and p io AF diagnosis. Mo eo e , he majo i y o ischemic s oke cases did no ha e a p io AF diagnosis and we e no ha ing an o al an icoagula ion he apy agains ca di- oembolic s oke. The associa ion wi h he age a ischemic s oke diagnosis could be explained wi h he obse ed associa ions wi h pa oxysmal AF in LURIC and AF in pa ien s aged 75 and o e in FINCAVAS. Pa oxysmal AF is mo e likely o be subclinical and asymp oma ic han mo e pe sis en o ms o AF and he e o e acu e ischemic s oke may be he i s clinical mani es a ion o he unde lying subclinical a ial ib illa ion. The possible bias due o he unde diagnosed pa oxysmal AF cases in FINCAVAS may explain he obse a ion ha he associa ion wi h inciden AF is seen only in he elde ly wi h p obably mo e symp oms and mo e equen sc eening o AF han younge pa ien s. In he LURIC s udy, ci cula ing ADMA and SDMA le els showed s ep-wise inc eases as he AF s a us was changed om sinus hy hm h ough pa oxysmal and ch onic AF independen o age, sex and enal unc ion. In line wi h hese esul s, a simila s ep-wise inc ease in ADMA le els we e seen in pa ien s wi h pa oxysmal AF and non-pa oxysmal AF compa ed o pa ien s wi h no AF in a p e ious s udy o co ona y angiog aphy pa ien s whe eas he associa ion o SDMA wi h AF was no in es iga ed21. Mo e ecen ly, ADMA, bu no SDMA, was independen ly associa ed wi h he de elopmen o symp oma ic AF in pa ien s wi h acu e myoca dial in a c- ion22. In con as , SDMA, bu no ADMA, was signi ican ly associa ed wi h a ial ib illa ion in pa ien s wi h acu e ischemic s oke2. The lack o associa ion be ween SDMA and AF in mos o he p e ious s udies may e lec he ela i ely small numbe o AF cases s udied. Mo eo e , he me hod o ADMA and SDMA quan i ica ion is no s anda dized complica ing compa isons be ween di e en s udies. We obse ed a s a is ically signi ican in e ac ion be ween s16899974 and AF s a us g oups on ci cula ing SDMA le els independen o age, sex and enal unc ion. Toge he wi h he obse a ion ha AGXT2 is exp essed in he human hea 16 his in e ac ion sugges s a possible ole o AGXT2 in he local dime hyla ginine me abolism Figu e 3. Box plo s o un ans o med concen a ions o ADMA, and SDMA (μmol/l) acco ding o he a ial ib illa ion (AF) s a us in LURIC (A). Da a a e shown as he 25 h, 50 h, and 75 h pe cen iles ( ep esen ed by g ay boxes), ange (shown as whiske s; ou lie s ha e been emo ed), and he median (black ho izon al line). *P < 0.05, **P < 0.01, ***P < 0.001. Unadjus ed β s o ADMA, SDMA and eGFR by AGXT2 SNPs and AF s a us assuming addi i e gene ic e ec on he log scale (B). Poin s ep esen poin es ima e o he β ; e ical ba s; he 95% CI. www.na u e.com/scien i ic epo s/ 7 Scien i ic RepoR s | 6:23207 | DOI: 10.1038/s ep23207 in ca diac issue; howe e , due o he obse a ional na u e o ou s udy, his in e es ing hypo hesis clea ly needs u he in es iga ion. Fu he mo e, he lack o associa ion o s16899974 wi h he sys emic ADMA le els does no exclude he possibili y ha AGXT2 could con ibu e o he ADMA me abolism locally in ca diac issues. This can be explained by he ac ha , unlike SDMA, ADMA is p ima ily me abolised by he DDAH enzymes which may be able o compensa e impai ed me abolism by AGXT2 bo h locally and a he sys emic le el23. To u he suppo he ole o dime hyla ginines behind he associa ions wi h AF, in a ecen la ge popula ion-based s udy24, SDMA was co ela ed wi h ma ke s o a ial emodelling such as le a ial size and a ial conduc ance ime al hough no independen associa ion was seen be ween SDMA and AF. These asso- cia ions a e in line wi h he concep ha dime hyla ginines could p omo e he elec ical and s uc u al a ial emodeling du ing AF seconda y o oxida i e s ess and NO syn hase uncoupling. I is also possible ha he gene ic a ian s o AGXT2 wi hin ca diac issues al e he NO- ela ed signalling pa hways ha egula e almos all ca diac ionic channels and modula e he suscep ibili y o bo h a ial and en icula a hy hmias25. In e es ingly, a ecen epo showed ha ADMA was independen ly associa ed wi h le a ial appendage h ombus in pa ien s wi h non- al ula a ial ib illa ion26, u he suppo ing he concep ha ADMA-induced endoca dial dys unc- ion could con ibu e o he inc eased isk o ca dioembolic s oke in AF. Finally, because he AGXT2 enzyme ha e se e al subs a es o he han dime hyla ginines27 and he sys emic le els o dime hyla ginines can me ely e lec he sys emic and/o local ac i i y o AGXT2, ou esul s suppo u he mechanis ic s udies on he ole o dime hyla ginines behind he obse ed associa ions. Whe he u u e genome-wide associa ion s udies on age a onse o any ischemic o c yp ogenic s oke could iden i y no el gene ic a ian s associa ed addi ionally wi h pa oxysmal o subclinical o ms o AF is wo h in es- iga ing because diagnosing o silen AF is challenging. Mo eo e , as he isk o AF inc eases subs an ially wi h age, an age-a -onse in o med28 o age s a i ica ion-based29 app oaches could de ec no el AF loci whose magni- ude o gene ic e ec s di e wi h age. Whe he he AGXT2 a ian s could p edic he p esence o a silen o pa ox- ysmal AF in pa ien s wi h acu e c yp ogenic ischemic s oke wa an s u he in es iga ion. Finally, expe imen al da a o he unc ional ole o s16899974 is cu en ly lacking. Some deg ee o misclassi ica ion o AF cases is expec ed, al hough his would mos likely a enua e he ue gene ic e ec a he han c ea e alse posi i e indings. Because a ial ib illa ion is o en in e mi en and asymp- oma ic, i is no possible o exclude he possibili y ha some o hose how a e classi ied o ha e no AF ha e unde- ec ed silen AF. Howe e , his limi a ion is likely o be p esen in all clinical da a se s in which AF is de ined based on hospi al discha ge diagnos ic codes a he han p olonged ambula o y ECG moni o ing. In con as , acu e ischemic s oke wi h neu ologic de ici s is less likely o be le unde ec ed and wi hou a diagnosis. Mo eo e , he associa ions wi h AF and age a ischemic s oke onse canno be di ec ly gene alized o he gene al popula ion, o he pa ien g oups o indi iduals o o he ances al backg ounds. Finally, s16899974 may be in linkage disequi- lib ium wi h he causal a ian (s) unde lying he associa ions wi h AF and ischemic s oke pheno ypes. Conclusions We ound s ong e idence ha he AGXT2 p.V498L polymo phism is associa ed wi h bo h pa oxysmal and ch onic o ms o AF in co ona y angiog aphic pa ien s wi hou s uc u al hea disease in ul asound, and ea - lie age a onse o ischemic s oke in pa ien s unde going exe cise s ess es ing. Hence, ou s udy sugges s ha AGXT2 a ia ions a e in ol ed in he genesis o AF and i s age- ela ed h omboembolic complica ions. Fu u e mechanis ic s udies should in es iga e whe he hese associa ions a e media ed h ough local me abolism o dime hyla ginines by AGXT2 in ca diac issues. Me hods S udy popula ions. The LURIC s udy consis s o 3,316 Caucasian pa ien s hospi alized o co ona y angi- og aphy be ween 1997 and 2000 a a e ia y ca e cen e in Sou hwes e n Ge many. Clinical indica ions o angi- og aphy we e ches pain o a posi i e non-in asi e s ess es sugges i e o myoca dial ischemia. To limi clinical he e ogenei y, indi iduals su e ing om acu e illnesses o he han acu e co ona y synd ome (ACS), ch onic non-ca diac diseases and a his o y o malignancy wi hin he i e pas yea s we e excluded. The Co ogene s udy included 5295 consecu i e Finnish pa ien s assigned o co ona y angiog am in 4 hospi- als se icing 1.5 million people in he Hospi al Dis ic o Helsinki and Uusimaa. O he Co ogene s udy, 2500 pa ien s wi h acu e co ona y synd ome (ICD-10: I20–I25) we e included in a genome-wide associa ion s udy. The FINCAVAS s udy included all consecu i e pa ien s e e ed o a clinically indica ed exe cise es using a bicycle e gome e a Tampe e Uni e si y Hospi al be ween Oc obe 2001 and he end o 2008 and willing o pa icipa e. A o al o 4,068 pa icipan s had a echnically success ul exe cise es . The main indica ions o he exe cise es we e suspicion o co ona y hea disease (CHD, equency 46%), e alua ion o wo k capaci y (26%), es ing ulne abili y o a hy hmia du ing exe cise (25%), and adequacy o he CHD ea men (13%); some pa ien s had mo e han one indica ion. Disco e y s oke coho s in WTCCC2 ischemic s oke GWAS included samples om he UK and Ge many, wi h a o al o 3,548 cases and 5,972 con ols. See u he desc ip ion o he s udy coho s and labo a o y analyses in Supplemen a y Me hods 1–2 and Supplemen a y Tables 1, 5 and 6. This s udy was conduc ed acco ding o he Decla a ion o Helsinki p inciples and w i en in o med consen was ob ained om all pa icipan s. The s udy was pe o med in acco dance wi h app o ed guidelines and eg- ula ions. The LURIC s udy was app o ed by he e hics commi ee a he Ä z ekamme Rheinland-P alz. The Co ogene s udy was app o ed by app op ia e E hics Commi ees o he Helsinki and Uusimaa Hospi al egion. The FINCAVAS s udy was app o ed by he E hical Commi ee o he Hospi al Dis ic o Pi kanmaa, Finland. Fo he WTCCC2 ischemic s oke coho s, he ec ui men o pa ien s was app o ed by he ele an local e hics commi ees om all he pa icipa ing cen e s. www.na u e.com/scien i ic epo s/ 8 Scien i ic RepoR s | 6:23207 | DOI: 10.1038/s ep23207 Geno yping and quali y con ol. In LURIC, geno yping was done by using he A yme ix Human SNP A ay 6.0 a he LURIC S udy acili y. Geno ype impu a ion was pe o med using he IMPUTE2 so wa e and he 1000 Genomes Ma ch 2012 haplo ypes as a e e ence. Geno yped da a was used o s37369 and impu ed da a o s16899974 wi h an excellen impu a ion quali y (in o~0.935). The Co ogene coho was geno yped wi h Illumina 660 K BeadChip a ay a he Sange Ins i u e (Hix on, Camb idge, UK) and impu ed using he 1000 Genomes Ap il 2012 e e ence panel as a e e ence. In FINCAVAS, we geno yped s16899974 success ully o 3,889 pa icipan s using Taqman@SNP Geno yping Assay C__25742181_10 and ABI P ism 7900HT Sequence De ec ion Sys em (Applied Biosys ems, Fos e Ci y, CA, USA). No e idence o de ia ion om he Ha dy-Weinbe g equilib ium was obse ed ( χ1 2 = 0.21, p = 0.65). The da a o s37369 and s6817105 we e ob ained o 3,195 indi iduals by geno yping using he Illumina HumanCa dio-Me abo BeadChip o HumanCo eExome chip a ays and impu a ion using he IMPUTE2 so - wa e and 1000 Genomes Ma ch 2012 haplo ypes as a e e ence. Fo he WTCCC2 samples, Illumina BeadChips we e used o genome-wide geno yping and geno ype impu- a ion was pe o med using MACH based on HapMap Phase 2 Eu opean (CEU) e e ence da a. Classi ica ion o a ial ib illa ion cases. In LURIC, 2,923 pa ien s had bo h geno ype and a ial ib il- la ion (AF) s a us da a a ailable. O he 2,923 pa icipan s, 360 had a his o y o AF a baseline. In addi ion, 161 indi iduals we e in AF hy hm du ing he index co ona y angiog aphy, o which 21 we e no included in he 360 cases diagnosed p e iously. The e o e, a o al o 381 indi iduals we e classi ied as any AF. O he 381 AF cases, 348 we e u he classi ied as ha ing ei he pa oxysmal AF (n = 175) o ch onic AF (n= 173). Fo he Co ogene s udy, po en ial p e alen AF cases we e sc eened om he baseline da abase. Fo hese pa ien s, we asce ained he AF s a us and i s sub ype om hei medical eco ds. O he 2,208 pa ien s included in his s udy, 265, 141 and 107 had any AF, pa oxysmal AF and ch onic AF, espec i ely. Fo he FINCAVAS pa icipan s, clinical AF was asce ained om he cen al uni e si y hospi al discha ge diagnos ic codes (ICD-9-CM 427.3, 427.31, o 427.32; o ICD-10 I48) om 1987 o 2015 (1179 inciden AF cases), and s udy popula ion a ea-wide elec ical ECG eco dings om 2005 onwa d (16 addi ional inciden AF cases). In addi ion, o u he asce ain he AF s a us a he index exe cise s ess es , we u ilized he da a om he baseline examina ion. Acco ding o he FINCAVAS baseline da abase, 13 pa ien s had a his o y o AF/ lu e o de eloped one du ing he exe cise s ess es bu did no ha e any p e ious discha ge diagnosis o AF wi h he da e o diagnosis and we e he e o e excluded om all analyses. Bo h he geno ype and pheno ype da a we e a ailable o 1,188 inciden AF cases and 2,674 censo ed con ols. Classi ica ion o ischemic s oke cases. Fo he WTCCC2 ischemic s oke coho s, T ial o O g 10172 in Acu e S oke T ea men (TOAST) classi ica ion30 was pe o med by an in-house neu ologis and all s oke cases we e classi ied in o mu ually exclusi e ae iologic sub ypes: la ge-a e y a he oscle osis (LAA), small- essel disease (SVD), ca dioembolic s oke (CE), o he ae iology, o unknown ae iology. Fo he FINCAVAS coho , age a he i s ischemic s oke was asce ained om he cen al uni e si y hos- pi al discha ge diagnos ic codes (ICD-9 433.x1, 434 (excluding 434.x0), o 436; o ICD-10 I63.0–I63.9). O he 3,862 indi iduals wi h bo h geno ype and pheno ype da a a ailable, he e we e 360 inciden ischemic s oke cases diagnosed be ween 1987 and 2015. In he majo i y o he cases 233 (65%), he e iology (LAA, SVD o CE) was unce ain a he ime o he diagnosis and we e he e o e diagnosed as a c yp ogenic ischemic s oke (I63.9). S a is ical me hods o c oss-sec ional associa ion and in e ac ion analyses. S a is ical analyses we e pe o med unde he R s a is ical en i onmen . The associa ions o he AGXT2 and 4q25 a ian s wi h AF and i s sub ypes we e es ed using mul i a iable logis ic eg ession analyses assuming an addi i e gene ic model. In addi ion, we es ed he in e ac ions o he s udied polymo phisms wi h es ablished AF isk ac o s on any AF in LURIC. Fo he WTCCC2 coho s, associa ions o he AGXT2 a ian s we e es ed wi h unad- jus ed logis ic eg ession using PLINK31 unde an addi i e gene ic model and esul s we e me a-analyzed wi h in e se- a iance-weigh ed me hod implemen ed in he METAL so wa e32. S a is ical me hods o su i al and age o onse analyses. Su i al analyses we e used o assess associa ions wi h inciden AF/s oke and di e ences in age o AF/s oke onse as a unc ion o he s udied a - ian s. In addi ion, o he age o onse analyses, we applied linea eg ession conside ing he age o onse as a quan i a i e ai . Su i al cu es o ime o AF/s oke onse we e es ima ed using he Kaplan–Meie me hod. We used a Cox eg ession analysis o examine he e ec o he SNPs oge he wi h co a ia es on he su i al unc ions o inciden AF and ischemic s oke. We used ch onological age as he undamen al ime scale in all analyses. All pa ien s we e ollowed o a ixed da e (Ap il 2015). Those wi h no inciden AF o ischemic s oke du ing he obse a ion pe iod we e igh censo ed a hei las isi a he cen al hospi al o a hei las elec ical ECG eco ding in he communi y, whiche e came la e . Signi icance was accep ed a P < 0.05 in all analyses. Re e ences 1. Meini ze , A. e al. Symme ical and Asymme ical Dime hyla ginine as P edic o s o Mo ali y in Pa ien s Re e ed o Co ona y Angiog aphy: The Ludwigsha en Risk and Ca dio ascula Heal h S udy. Clinical Chemis y 57, 112–121 (2011). 2. Schulze, F. e al. Symme ic dime hyla ginine p edic s all-cause mo ali y ollowing ischemic s oke. A he oscle osis 208, 518–523 (2010). 3. Xia, W. e al. Asymme ic dime hyla ginine concen a ion and ea ly ecu ence o a ial ib illa ion a e elec ical ca dio e sion. Pacing Clin. Elec ophysiol. 31, 1036–1040 (2008). 4. Liu, H. e al. Va iance o DDAH/PRMT/ADMA pa hway in a ial ib illa ion dogs. Biochem. Biophys. Res. Commun. 377, 884–888 (2008). 5. Yang, L. e al. Asymme ic dime hyla ginine concen a ion and ecu ence o a ial achya y hmias a e ca he e abla ion in pa ien s wi h pe sis en a ial ib illa ion. J. In e . Ca d. Elec ophysiol. 32, 147–154 (2011). www.na u e.com/scien i ic epo s/ 9 Scien i ic RepoR s | 6:23207 | DOI: 10.1038/s ep23207 6. Goe e, A. e al. The impac o apid a ial pacing on ADMA and endo helial NOS. In . J. Ca diol. 154, 141–146 (2012). 7. Lim, H. S. e al. E ec o a ial ib illa ion on a ial h ombogenesis in humans: impac o a e and hy hm. J. Am. Coll. Ca diol. 61, 852–860 (2013). 8. Sydow, K. & Munzel, T. ADMA and oxida i e s ess. A he oscle . Suppl. 4, 41–51 (2003). 9. Bode-Böge , S. M. e al. Symme ical Dime hyla ginine: A New Combined Pa ame e o Renal Func ion and Ex en o Co ona y A e y Disease. Jou nal o he Ame ican Socie y o Neph ology 17, 1128–1134 (2006). 10. Felie s, D., Lee, D. Y., Go in, Y. & Kasina h, B. S. Symme ic dime hyla ginine al e s endo helial ni ic oxide ac i i y in glome ula endo helial cells. Cell. Signal. 27, 1–5 (2015). 11. Cai, H. e al. Down egula ion o endoca dial ni ic oxide syn hase exp ession and ni ic oxide p oduc ion in a ial ib illa ion: po en ial mechanisms o a ial h ombosis and s oke. Ci cula ion 106, 2854–2858 (2002). 12. Kim, Y. M. e al. A myoca dial Nox2 con aining NAD(P)H oxidase con ibu es o oxida i e s ess in human a ial ib illa ion. Ci c. Res. 97, 629–636 (2005). 13. Reilly, S. N. e al. A ial sou ces o eac i e oxygen species a y wi h he du a ion and subs a e o a ial ib illa ion: implica ions o he an ia hy hmic e ec o s a ins. Ci cula ion 124, 1107–1117 (2011). 14. Caplin, B. e al. Alanine-glyoxyla e amino ans e ase-2 me abolizes endogenous me hyla ginines, egula es NO, and con ols blood p essu e. A e ioscle . Th omb. Vasc. Biol. 32, 2892–2900 (2012). 15. Seppälä, I. e al. Genome-wide associa ion s udy on dime hyla ginines e eals no el AGXT2 a ian s associa ed wi h hea a e a iabili y bu no wi h o e all mo ali y. Eu opean Hea Jou nal 35, 524–531 (2014). 16. Lünebu g, N. e al. Genome-Wide Associa ion S udy o l-A ginine and Dime hyla ginines Re eals No el Me abolic Pa hway o Symme ic Dime hyla ginine. Ci cula ion: Ca dio ascula Gene ics 7, 864–872 (2014). 17. Ki el, A. e al. Alanine-glyoxyla e amino ans e ase 2 (AGXT2) polymo phisms ha e conside able impac on me hyla ginine and be a-aminoisobu y a e me abolism in heal hy olun ee s. PLoS One 9, e88544 (2014). 18. Ellino , P. T. e al. Me a-analysis iden i ies six new suscep ibili y loci o a ial ib illa ion. Na . Gene . 44, 670–675 (2012). 19. T aylo , M. e al. Gene ic isk ac o s o ischaemic s oke and i s sub ypes ( he METASTROKE collabo a ion): a me a-analysis o genome-wide associa ion s udies. Lance Neu ol. 11, 951–962 (2012). 20. Ellino , P. T. e al. Common a ian s in KCNN3 a e associa ed wi h lone a ial ib illa ion. Na . Gene . 42, 240–244 (2010). 21. Chao, T. F. e al. Plasma asymme ic dime hyla ginine and ad e se e en s in pa ien s wi h a ial ib illa ion e e ed o co ona y angiog am. PLoS One 8, e71675 (2013). 22. S amboul, K. e al. A ial Fib illa ion Is Associa ed wi h a Ma ke o Endo helial Func ion and Oxida i e S ess in Pa ien s wi h Acu e Myoca dial In a c ion. PLoS One 10, e0131439 (2015). 23. Rodiono , R. N. e al. Role o alanine:glyoxyla e amino ans e ase 2 in me abolism o asymme ic dime hyla ginine in he se ings o asymme ic dime hyla ginine o e load and bila e al neph ec omy. Neph ol. Dial. T ansplan . 29, 2035–2042 (2014). 24. Ramuschka , M. e al. ADMA, subclinical changes and a ial ib illa ion in he gene al popula ion. In . J. Ca diol. 203, 640–646 (2016). 25. Tama go, J., Caballe o, R., Gomez, R. & Delpon, E. Ca diac elec ophysiological e ec s o ni ic oxide. Ca dio asc. Res. 87, 593–600 (2010). 26. Xia, W. e al. Asymme ic dime hyla ginine p edic s le a ial appendage h ombus in pa ien s wi h non- al ula a ial ib illa ion. Th omb. Res. 136, 1156–1159 (2015). 27. Rodiono , R. N., Ja zebska, N., Weiss, N. & Len z, S. R. AGXT2: a p omiscuous amino ans e ase. T ends Pha macol. Sci. 35, 575–582 (2014). 28. T aylo , M. e al. A no el MMP12 locus is associa ed wi h la ge a e y a he oscle o ic s oke using a genome-wide age-a -onse in o med app oach. PLoS Gene . 10, e1004469 (2014). 29. Simino, J. e al. Gene-age in e ac ions in blood p essu e egula ion: a la ge-scale in es iga ion wi h he CHARGE, Global BPgen, and ICBP Conso ia. Am. J. Hum. Gene . 95, 24–38 (2014). 30. Adams, H. P.,J e al. Classi ica ion o sub ype o acu e ischemic s oke. De ini ions o use in a mul icen e clinical ial. TOAST. T ial o O g 10172 in Acu e S oke T ea men . S oke 24, 35–41 (1993). 31. 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). 32. Wille , C. J., Li, Y. & Abecasis, G. R. METAL: as and e icien me a-analysis o genomewide associa ion scans. Bioin o ma ics 26, 2190–2191 (2010). Acknowledgemen s We hank he LURIC eam in ol ed in pa ien ec ui men , sample and da a handling, and he labo a o y s a s a he Ludwigsha en Gene al Hospi al, he Uni e si ies o F eibu g, Ulm, and G az. We also hank he Wellcome T us Case Con ol Conso ium 2. LURIC has ecei ed unding om he 6 h F amewo k P og am (in eg a ed p ojec Bloodomics, g an LSHM-CT-2004–503485) and 7 h o F amewo k P og am (in eg a ed p ojec A he oRemo, g an ag eemen numbe 201668 and RiskyCAD, g an ag eemen numbe 305739) o he Eu opean Union and by he INTERREG IV Obe hein P og am (P ojec A28, Gene ic mechanisms o ca dio ascula diseases) wi h suppo om he Eu opean Regional De elopmen Fund (ERDF) and he Wissenscha so ensi e TMO. FINCAVAS has been inancially suppo ed by he Academy o Finland: G an 286284, Social Insu ance Ins i u ion o Finland, Kuopio, Tampe e and Tu ku Uni e si y Hospi al Medical Funds, Finnish Founda ion o Ca dio ascula Resea ch, Tampe e Tube culosis Founda ion and Emil Aal onen Founda ion, and Y jö Jahnsson Founda ion. I.S is suppo ed by Labo a o iolääke ie een edis ämissää iö and he Finnish Medical Founda ion. The unding o he WTCCC2 s oke s udy was p o ided by he Wellcome T us , as pa o he Wellcome T us Case Con ol Conso ium 2 p ojec (085475/B/08/Z and 085475/Z/08/Z and WT084724MA). Au ho Con ibu ions I.S. w o e he main manusc ip ex . I.S., S.B. and E.V. analysed he da a. M.E.K., S.B., L.-P.L., N.O., J.A.H., K.-M.M., G.E.D., T.K. and H.S. p o ided expe ise o da a analysis and in e p e a ion, commen ed he pape . P.M.R., C.S., M.D., J.S., R.L., M.K., H.S.M., W.M. and T.L. con ibu ed o da a collec ion. N.M. pe o med he geno yping in FINCAVAS. All au ho s con ibu ed o and ha e app o ed he inal manusc ip . Addi ional In o ma ion Supplemen a y in o ma ion accompanies his pape a h p://www.na u e.com/s ep Compe ing inancial in e es s: The au ho s decla e no compe ing inancial in e es s.