A genome-wide association meta-analysis on lipoprotein(a) concentrations adjusted for apolipoprotein(a) isoforms

1834 Jou nal o Lipid Resea ch Volume 58, 2017 Copy igh © 2017 by he Ame ican Socie y o Biochemis y and Molecula Biology, Inc. This a icle is a ailable online a h p://www.jl .o g Abs ac High lipop o ein (a) [Lp(a)] concen a ions a e an independen isk ac o o ca dio ascula ou comes. Concen a ions a e s ongly in luenced by apo(a) k ingle IV epea iso o ms. We aimed o iden i y gene ic loci associa ed wi h Lp(a) concen a ions using da a om i e genome-wide associa ion s udies (n = 13,781). We iden i ied 48 indepen- den SNPs in he LPA and 1 SNP in he APOE gene egion o be signi ican ly associa ed wi h Lp(a) concen a ions. We also adjus ed o apo(a) iso o ms o iden i y loci a ec ing Lp(a) le els independen ly om hem, which esul ed in 31 SNPs (30 in he LPA, 1 in he APOE gene egion). Se en This wo k was suppo ed by Aus ian Science Fund G an P 26660-B13 o C.L. and he “Genomics o Lipid-associa ed Diso de s (GOLD)” o he “Aus ian Ge- nome Resea ch P og amme GEN-AU” o F.K. O he suppo was p o ided by he Biomedical Resea ch P og am a Weill Co nell Medicine-Qa a , unded by he Qa- a Founda ion o N.A.Y. The NHLBI Family Hea S udy was suppo ed by Na- ional Hea , Lung, and Blood Ins i u e G an s U01 HL67893, U01 HL67894, U01 HL67895, U01 HL67896, U01 HL67897, U01 HL67898, U01 HL67899, U01 HL67900, U01 HL67901, and U01 HL67902. The CoLaus s udy was and is suppo ed by esea ch g an s om GlaxoSmi hKline, he Facul y o Biology and Medicine o Lausanne, and Swiss Na ional Science Founda ion G an s 33CSCO- 122661, 33CS30-139468, and 33CS30-148401. The KORA s udy was ini ia ed and inanced by he Helmhol z Zen um München-Ge man Resea ch Cen e o En i onmen al Heal h, which is unded by he Ge man Fede al Minis y o Educa- ion and Resea ch (BMBF) and by he S a e o Ba a ia. KORA esea ch was also suppo ed wi hin he Munich Cen e o Heal h Sciences (MC-Heal h), Ludwig- Maximilians-Uni e si ä , as pa o LMUinno a i . The Young Finns S udy was inancially suppo ed by Academy o Finland G an s 286284, 134309, 126925, 121584, 124282, 129378, 117787, and 41071; he Social Insu ance Ins i u ion o Finland; Kuopio, Tampe e, and Tu ku Uni e si y Hospi al Medical Funds (G an X51001); Juho Vainio Founda ion; Paa o Nu mi Founda ion; Finnish Founda ion o Ca dio ascula Resea ch; Finnish Cul u al Founda ion; Tampe e Tube culosis Founda ion; Emil Aal onen Founda ion; and Y jö Jahnsson Founda- ion. The con en is solely he esponsibili y o he au ho s and does no necessa ily ep esen he o icial iews o he Na ional Ins i u es o Heal h. Au ho ’s Choice—Final e sion ee ia C ea i e Commons CC-BY license. Manusc ip ecei ed 10 Ma ch 2017 and in e ised o m 4 May 2017. Published, JLR Pape s in P ess, May 16, 2017 DOI h ps://doi.o g/10.1194/jl .M076232 A genome-wide associa ion me a-analysis on lipop o ein (a) concen a ions adjus ed o apolipop o ein (a) iso o ms Salome Mack,* S e an Coassin,* Rico Rueedi,†,§ Noha A. Yous i,‡,|| Ilkka Seppälä,# Ch is ian Giege ,$,**,†† Sebas ian Schönhe ,* Lukas Fo e ,* Ge aud E ha ,* Ped o Ma ques-Vidal,§§ Janina S. Ried,$ Ge a d Waebe ,§§ S en Be gmann,†,§ Do een Dähnha d ,* And ea S öckl,* Olli T. Rai aka i,‡‡,|||| Mika Kähönen,## Anne e Pe e s,**,$$,*** Thomas Mei inge ,†††,§§§,‡‡‡ Kons an in S auch,$,|||||| KORA-S udy G oup, Ludmilla Kedenko,### Be nha d Paulwebe ,### Te ho Leh imäki,# S e en C. Hun ,$$$,**** Pe e Vollenweide ,§§ Claudia Lamina,1,2,* and Flo ian K onenbe g1,2,* Di ision o Gene ic Epidemiology,* Depa men o Medical Gene ics, Molecula and Clinical Pha macology, Medical Uni e si y o Innsb uck, 6020 Innsb uck, Aus ia; Depa men o Compu a ional Biology,† Uni e si y o Lausanne, 1015 Lausanne, Swi ze land; Swiss Ins i u e o Bioin o ma ics,§ 1015 Lausanne, Swi ze land; Depa men o Physiology and Biophysics,‡ Weill Co nell Medical College-Qa a , Doha, Qa a ; Depa men o Compu e and Sys ems Enginee ing,|| Alexand ia Uni e si y, 21526 Alexand ia, Egyp ; Depa men o Clinical Chemis y,# Fimlab Labo a o ies and Uni e si y o Tampe e School o Medicine, 33520 Tampe e, Finland; Ins i u e o Gene ic Epidemiology,$ Ins i u e o Epidemiology II,** Resea ch Uni o Molecula Epidemiology,†† and Ins i u e o Human Gene ics,§§§ Helmhol z Zen um München-Ge man Resea ch Cen e o En i onmen al Heal h, 85764 Neuhe be g, Ge many; Depa men o Medicine,§§ In e nal Medicine, Lausanne Uni e si y Hospi al, 1015 Lausanne, Swi ze land; Depa men o Clinical Physiology,‡‡ Tu ku Uni e si y Hospi al, 20520 Tu ku, Finland; Resea ch Cen e o Applied and P e en i e Ca dio ascula Medicine,|||| Uni e si y o Tu ku, 20520 Tu ku, Finland; Depa men o Clinical Physiology,## Tampe e Uni e si y Hospi al and Uni e si y o Tampe e, 33521 Tampe e, Finland; Ge man Cen e o Ca dio ascula Resea ch (DZHK),$$ 80802 Munich, Ge many; Ge man Cen e o Diabe es Resea ch (DZD e.V.),*** 85764 Neuhe be g, Ge many; Ins i u e o Human Gene ics,††† Technische Uni e si ä München, 81675 München, Ge many; Munich Clus e o Sys ems Neu ology (SyNe gy),‡‡‡ 81377 Munich, Ge many; Ins i u e o Medical In o ma ics, Biome y, and Epidemiology,|||||| Ludwig-Maximilians-Uni e si ä , 81377 Munich, Ge many; Fi s Depa men o In e nal Medicine,### Pa acelsus P i a e Medical Uni e si y, 5020 Salzbu g, Aus ia; Ca dio ascula Gene ics Di ision,$$$ Uni e si y o U ah School o Medicine, Sal Lake Ci y, UT 84108; and Depa men o Gene ic Medicine,**** Weill Co nell Medicine, Doha, Qa a ORCID ID: 0000-0003-2229-1120 (F.K.) Abb e ia ions: CAD, co ona y a e y disease; CNV, copy-numbe - a ia ion; GWAS, genome-wide associa ion s udy; KIV, k ingle IV; LD, linkage disequilib ium; LDL-C, LDL choles e ol; Lp(a), lipop o ein (a); OR, odds a io; OxPL, oxidized phospholipid. 1 C. Lamina and F. K onenbe g con ibu ed equally o his wo k. 2 To whom co espondence should be add essed. e-mail: [email p o ec ed] (F.K.); [email p o ec ed] (C.L.) The online e sion o his a icle (a ailable a h p://www.jl .o g) con ains a supplemen .  Au ho ’s Choice a Uni e si y o Tampe e Lib a y, on Oc obe 13, 2017www.jl .o gDownloaded om .h ml h p://www.jl .o g/con en /suppl/2017/05/16/jl .M076232.DC1 Supplemen al Ma e ial can be ound a : Genome-wide me a-analysis on lipop o ein (a) 1835 SNPs showed a genome-wide signi ican associa ion wi h co ona y a e y disease (CAD) isk. A a e SNP ( s186696265; MAF 1%) showed he highes e ec on Lp(a) and was also associa ed wi h inc eased isk o CAD (odds a io = 1.73, P = 3.35 × 1030). Median Lp(a) alues inc eased om 2.1 o 91.1 mg/dl wi h inc easing numbe o Lp(a)-inc easing alleles. We ound he APOE2-de e mining allele o s7412 o be signi ican ly associa ed wi h Lp(a) concen a ions (P = 3.47 × 1010). Each APOE2 allele dec eased Lp(a) by 3.34 mg/dl co esponding o 15% o he popula ion’s mean alues. Pe o ming a gene-based es o associa ion, including suspec ed Lp(a) ecep o s and egula o s, esul ed in one signi ican associa ion o he TLR2 gene wi h Lp(a) (P = 3.4 × 104). In summa y, we iden i ied a la ge numbe o independen SNPs in he LPA gene egion, as well as he APOE2 allele, o be signi ican ly associa ed wi h Lp(a) con- cen a ions.—Mack, S., S. Coassin, R. Rueedi, N. A. Yous i, I. Seppälä, C. Giege , S. Schönhe , L. Fo e , G. E ha , P. Ma ques-Vidal, J. S. Ried, G. Waebe , S. Be gmann, D. Dähnha d , A. S öckl, O. T. Rai aka i, M. Kähönen, A. Pe e s, T. Mei inge , K. S auch, KORA-S udy G oup, L. Kedenko, B. Paulwebe , T. Leh imäki, S. C. Hun , P. Vollenweide , C. Lamina, and F. K onenbe g. A genome-wide associa ion me a-analysis on lipop o ein (a) concen a ions adjus ed o apolipop o ein (a) iso o ms. J. Lipid Res. 2017. 58: 1834–1844. Supplemen a y key wo ds  gene ics • epidemiology • co ona y a e y disease High lipop o ein (a) [Lp(a)] concen a ions ha e been shown o be an independen isk ac o o CVD, as well as ao ic al e calci ica ion and s enosis. Mendelian andom- iza ion s udies p o ided s ong e idence o causali y [ o e iew see (1)]. Lp(a) is a lipop o ein consis ing o a co e LDL-like pa icle and he glycop o ein, apo(a), ha a e co- alen ly linked o each o he . I is a quan i a i e gene ic ai unde p onounced gene ic con ol (1). Twin and am- ily s udies sugges ha 90–95% o a ia ion in Lp(a) is he i- able (2). The dis ibu ion o Lp(a) concen a ions in he popula ion is ex emely b oad, wi h a mo e han 1,000- old ange om below 0.1 mg/dl o mo e han 200 mg/dl (1, 3). The concen a ion o Lp(a) is mos ly in luenced by he size o he apo(a) iso o ms, which a e caused by a coding copy- numbe - a ia ion (CNV) in he LPA gene (2, 4, 5). This CNV con ains 1 o >40 epea ed plasminogen-like k ingle IV (KIV) domains, he so-called KIV epea s, leading o a high he e ogenei y o he apo(a) iso o m dis ibu ion in popula ions. Howe e , no all alleles a e exp essed (6), e- sul ing in di e en he e ozygosi y a es on he DNA le el (up o 95% he e ozygosi y) and p o ein le el (up o 70% he e ozygosi y) (1). These nonexp essed alleles a e no uni o mly dis ibu ed: sho e iso o ms a e mo e likely o be exp essed han longe iso o ms and a e associa ed wi h highe Lp(a) concen a ions (1). Lp(a) belongs o he s onges gene ically de e mined isk ac o s o CVD, conside ing ha 25–35% o he popula ion ca y sho iso o ms ha a e usually associa ed wi h high Lp(a) con- cen a ions and a doubling o he isk o CVD. Besides he numbe o KIV-2 domains, gene ic a ian s in he b oade LPA gene egion, as well as o he a ian s, ha e been ound o be associa ed wi h Lp(a) concen a ions (1). Fi y o nine y pe cen o he o e all gene ic a ia ion in Lp(a) is a ibu able o he LPA locus (1). So a , se e al genome-wide associa ion s udies (GWASs) on Lp(a) ha e been published (7–14). Howe e , hese s ud- ies we e limi ed by small sample size (7, 12, 14), es ic ed scope due o he use o a specialized ca dio ascula gene- chip (9, 10), o we e conduc ed in ce ain subg oups, such as pa ien s wi h ype 2 diabe es (11), o in popula ion iso- la es (8, 13). All o hese s udies p ima ily iden i ied SNPs in he LPA gene clus e on ch omosome 6q27 (SLC22A3- LPAL2-LPA-PLG). Howe e , he SNPs wi h he highes e - ec s we e no independen om he KIV epea (10). One ecen GWAS in A ican Ame icans (14) also ound one SNP in he APOE gene o be signi ican ly associa ed wi h Lp(a) concen a ions, which has been con i med by a ecen la ge s udy (15). Howe e , Lp(a) le els and he ypes and equencies o gene ic a ian s, especially in he LPA gene, di e conside ably be ween e hnic g oups (3). The e o e, his inding is no gene alizable o o he popula ions. Ou p ojec aimed o iden i y gene loci ha a e asso- cia ed wi h Lp(a) concen a ions on a hypo hesis- ee app oach using genome-wide SNP chips in s udies o Eu o- pean ances y. We included adjus men o apo(a) iso- o ms o iden i y loci ha a ec Lp(a) le els independen om hese apo(a) iso o ms. The hypo hesis- ee app oach, especially, is hough o p o ide new a enues o genes and he eby esea ch di ec ions o ind answe s o un esol ed ques ions ega ding physiology and pa hophysiology, as well as p oduc ion and ca abolism o Lp(a). Besides he hypo hesis- ee app oach, we pe o med a gene-based es o associa ion o a lis o possible candida e genes consis - ing o suspec ed Lp(a) ecep o s and egula o s discussed in he li e a u e. Al oge he , i e di e en p ima ily popu- la ion-based s udies wi h 13,781 indi iduals we e included and abou 10 million SNPs analyzed. MATERIALS AND METHODS S udy design and desc ip ion o coho s Genome-wide analyses we e pe o med in i e di e en s udies indi idually and we e hen me a-analyzed. The genome-wide sig- ni icance le el was se o 5 × 108. One addi ional s udy (SAPHIR s udy) was included o an in-dep h analysis o APOE geno ypes. A de ailed desc ip ion o he coho s is p o ided in he supplemen- al Ma e ials and Me hods and in supplemen al Table S1. Fo in o - ma ion on geno yping and impu a ion see supplemen al Table S2. Measu emen o Lp(a) concen a ions and apo(a) iso o ms Fo all pa icipa ing s udies, all Lp(a) measu emen s and apo(a) iso o ms we e pe o med by ELISA and immunoblo in he same labo a o y (Di ision o Gene ic Epidemiology, Inns- b uck, Aus ia). De ails on he measu emen echniques a e gi en in he supplemen al Ma e ials and Me hods. A s anda dized amoun o Lp(a) (150 ng) was loaded on he immunoblo . The p edominan ly exp essed apo(a) iso o m (in he e ozygous sam- ples), o he only band p esen (in homozygous samples) was used o adjus he s a is ical analysis o apo(a) iso o ms. In indi iduals a Uni e si y o Tampe e Lib a y, on Oc obe 13, 2017www.jl .o gDownloaded om .h ml h p://www.jl .o g/con en /suppl/2017/05/16/jl .M076232.DC1 Supplemen al Ma e ial can be ound a : 1836 Jou nal o Lipid Resea ch Volume 58, 2017 showing only one band in he Wes e n blo , he second allele was ei he no exp essed (16) o hey we e uly homozygous, al- hough he p opo ion o indi iduals being uly homozygous on DNA le el was expec ed o be lowe han 5%. S a is ical me hods GWAS analysis o single s udies and me a-analysis. Because he dis- ibu ion o he Lp(a) concen a ions we e highly skewed, an in- e se-no mal ans o ma ion was applied o he measu ed Lp(a) concen a ions. In each s udy, each SNP was es ed o associa ion wi h hese in e se-no mal ans o med Lp(a) concen a ions in an addi i e gene ic model using linea eg ession, adjus ing o age and sex (model 1). In addi ion, a second model, adjus ed o age, sex, and he apo(a) iso o m ha was p edominan ly exp essed in he immunoblo (model 2), was es ed. To ob ain in e p e able e - ec es ima es, linea eg ession was also pe o med on he o iginal scale o Lp(a) o bo h models. Genome-wide analysis in he FamHS s udy was done using a linea mixed model accoun ing o amilial dependencies desc ibed by a pedig ee-based kinship ma ix. Fo he me a-analysis o all GWASs, he so wa e, METASOFT (17), was used o all impu ed SNPs ha me impu a ion and quali y con ol c i e ia and we e p esen in a leas wo s udies (9.2 M SNPs). De ails on quali y con ol, il e ing c i e ia, and he me a- analysis app oach a e p o ided in he supplemen al Ma e ials and Me hods and supplemen al Fig. S1. Gende -s a i ied models we e also applied o bo h models, ollowed by a - es on e ec di e ences be ween men and women (18). Pai wise linkage dis- equilib ium (LD) was e alua ed using SNiPA wi h 1000 Genomes, phase1 3 da a (19). Condi ional analysis. To de ec independen ly associa ed SNPs, a condi ional s epwise analysis was pe o med using he p og am, GCTA [ e sion 1.24.7 (20)]. Fo each locus wi h a leas one P alue <5 × 108, he SNP wi h he lowes P alue was aken as he lead SNP. I was planned o include all SNPs wi hin a egion ±500 kb su ounding he lead SNP in he condi ional analysis. Because genome-wide signi ican SNPs we e also ound ou side o his ange o he LPA gene egion, he condi ional analysis was ex ended o a ange o 1.76 Mb. GCTA uses he summa y s a is ics o he me a-analysis plus one e e ence popula ion o LD calcula- ion. As e e ence popula ion, a combined geno ype da ase o KORA F3 and KORA F4 was used (n = 6,002). By de aul , he lead SNP was included in he model i s . Then, all SNPs in he in- cluded gene egion we e es ed o associa ion in addi ion o he al eady included SNPs in a s epwise manne . Using all indepen- den ly associa ed SNPs om model 1, an unweigh ed, as well as weigh ed, SNP-sco e was de i ed. The unweigh ed SNP-sco e co - esponded o he numbe o Lp(a)-inc easing alleles. Fo weigh - ing,  es ima es on in e se-no mal ans o med Lp(a) alues om he join model o all included SNPs we e aken. Gene-based and candida e gene analysis. In addi ion o he analy- sis o single SNP e ec s, a gene-based scan was pe o med using me a-analysis esul s om bo h models (wi h and wi hou adjus - ing o iso o ms) using he so wa e, KGG e sion 3.5 (21). Gene egions we e de ined as he gene ±20 kb acco ding o he Re - Gene da abase. Using his de ini ion, 66.5% o all a ailable SNPs we e included. Fo he gene-based analysis, he ex ended Simes es (GATES) was used as implemen ed in KGG (22). To adjus o mul iple es ing, he Bon e oni me hod was applied on he numbe o es ed genes (25,128 genes, which esul ed in a signi i- cance le el o 1.99 × 106). To calcula e LD be ween he SNPs, he 1000G phase1 3 Re e ence was used. In addi ion o he hypo he- sis- ee gene-based es , 21 candida e genes om li e a u e we e es ed o associa ion (supplemen al Ma e ials and Me hods) us- ing a Bon e oni signi icance P alue o 0.05/21 = 0.0024. Impac o Lp(a)-associa ed SNPs on co ona y a e y disease isk. To assess he ele ance o he iden i ied SNPs wi h ega d o isk on co ona y a e y disease (CAD), esul s om he CARDIoGRAM- plusC4D conso ium (23) we e downloaded. This GWAS me a- analysis comp ises s udies o mainly Eu opean, bu also Sou h Asian and Eas Asian descen , including 60,801 CAD cases and 123,504 con ols. Impu ed geno ypes we e based on 1000 Ge- nomes phase1 3. Log odds a ios (ORs) on CAD isk (assuming an addi i e model) and s anda d e o s o all independen ly wi h Lp(a)-associa ed SNPs we e e ie ed om he summa y-le el da a and ma ched o he mino allele. Va iance explained and he i abili y. The combined da ase o bo h KORA s udies (n = 6,002) was used o es ima e he genomic he i abili y, which is he p opo ion o pheno ypic a iance ex- plained by all es ed SNPs (24). In addi ion, he p opo ion o a iance explained by indi idual SNPs was calcula ed wi h da a om bo h KORA s udies using he so wa e GCTA ( .1.24.7) (20). In he FamHS s udy, he p opo ion o he addi i e (polygenic) a iance on he pheno ypic a iance, he na ow-sense he i abili y h2, was es ima ed using GenABEL’s polygenic unc ion, aking he kinship ma ix in o accoun . This na ow-sense he i abili y hus also includes he a iance explained by unmeasu ed SNP e ec s and o he ac o s (e.g., CNVs). E alua ing associa ion o APOE geno ypes wi h Lp(a) concen a- ions. We ollowed up he iden i ied associa ion o a SNP in he APOE gene ( s7412) in bo h KORA s udies and he SAPHIR s udy; he la e was no pa o he GWAS me a-analysis. Two SNPs ( s7412 and s429358) unambiguously de ined he apoE iso o ms (supplemen al Table S3). Because APOE E2/E2 and E4/E4 geno- ypes a e especially a he a e, a high impu a ion quali y o a leas 0.95 was equi ed o bo h SNPs indi idually o his analysis. The e o e, impu ed SNPs we e only aken om he KORA F3 s udy, whe eas de no o geno yping o hese wo SNPs was pe - o med o KORA F4 and SAPHIR. The associa ion analysis was pe o med o he APOE geno ypes, E2/E2, E2/E3, E2/E4, E3/ E4, and E4/E4, wi h he mos common geno ype, E3/E3, as he e e ence, adjus ed o age and sex and also age, sex, and apo(a) iso o ms. All h ee da ase s (KORA F3, KORA F4, and SAPHIR) we e combined in one da ase and mixed e ec models we e pe - o med bo h on in e se-no mal ans o med Lp(a) le els and on he o iginal scale o Lp(a). Bioin o ma ic analysis In o de o shed some ligh on po en ial unc ional elemen s unde lying he iden i ied SNPs, all o ou GWAS hi s, as well as all SNPs co ela ed wi h hem ( 2  0.8), we e in es iga ed o : 1) being a coding SNP o being loca ed a a canonical splice si e; 2) being loca ed in a ansc ip ion ac o binding si e epo ed by ENCODE (25); 3) being loca ed in a DNase hype sensi i e si e epo ed by ENCODE; o 4) being loca ed in a alida ed egula- o y elemen epo ed by ORegAnno (26). GTex (27) was sea ched o epo ed eQTL SNPs o LPA (i.e., SNPs being co - ela ed wi h LPA exp ession). De ailed me hods a e desc ibed in he supplemen al Ma e ials and Me hods. RESULTS Desc ip ion o coho s and quali y con ol Supplemen al Table S1 gi es he desc ip i e cha ac e is- ics o all con ibu ing s udies. The P-Z-plo s did no e eal any de ia ions o he epo ed P alues and he P alues a Uni e si y o Tampe e Lib a y, on Oc obe 13, 2017www.jl .o gDownloaded om .h ml h p://www.jl .o g/con en /suppl/2017/05/16/jl .M076232.DC1 Supplemen al Ma e ial can be ound a : Genome-wide me a-analysis on lipop o ein (a) 1837 calcula ed by he  coe icien and s anda d e o . The ge- nomic in la ion ac o  anged om 1.011 o 1.030 (sup- plemen al Table S2). Resul s o me a-analysis In bo h models (wi h and wi hou adjus ing o he p e- dominan ly exp essed apo(a) iso o m), SNPs in wo gene egions we e iden i ied: LPA and APOE. Manha an plo s o bo h me a-analyses a e shown in supplemen al Figs. S2 and S3 and co esponding QQ-plo s in supplemen al Figs. S4 and S5. On ch omosome 6, su ounding he LPA gene, 2,001 SNPs eached genome-wide signi icance in model 1, wi h he lowes P alue o SNP s55730499 (P = 3.6 × 10424, Fig. 1), and 1,961 SNPs eached genome-wide signi icance in model 2, wi h he lowes P alue o SNP s75692336 (P = 2.90 × 10216, supplemen al Fig. S6). Genome-wide signi ican SNPs o bo h models on ch omosome 6 we e sca e ed o e a b oad egion spanning 1.76 Mb (ch 6:159,991,850-161,753,083). Addi ionally, in he APOE gene on ch omosome 19, one genome-wide signi ican SNP ( s7412) was iden i ied in model 1 (P = 3.47 × 1010, Fig. 2), and h ee genome-wide signi ican SNPs in model 2 (lowes P alue o SNP s7412: 3.48 × 109; supplemen al Fig. S7). Fu he e alua ion o he LPA gene egion The condi ional analyses we e pe o med o bo h models, including all SNPs in he 1.76 Mb-spanning b oad LPA gene egion. Fo y-eigh SNPs we e independen ly associ- a ed wi h in e se-no mal ans o med Lp(a) in model 1 (see supplemen al Table S4 o cha ac e is ics o SNPs, supplemen al Table S5 o esul s o he single s udies, and supplemen al Table S6 o esul s o me a-analysis). These SNPs we e p ima ily loca ed wi hin he LPA gene, bu hey we e also sca e ed widely in a b oad egion up o 1 Mb away om he LPA gene (Fig. 1). The e ec sizes on he o igi- nal scale o Lp(a) anged be ween 0.05 and 64.74 mg/dl pe allele o he single SNPs and om 1.55 o 47.60 mg/dl pe allele in a join model including all 48 SNPs (median e ec size pe SNP: 5 mg/dl). Thi y SNPs emained s a is ically signi ican a e condi- ional s epwise analysis in model 2 (see supplemen al Table S4 o cha ac e is ics o SNPs, supplemen al Table S7 o esul s o he single s udies, and supplemen al Table S8 o esul s o me a-analysis). Again, hese SNPs we e dis ib- u ed o e he en i e b oad LPA gene egion, mos o hem in o close o he LPA gene. Nine een SNPs we e unique o his iso o m-adjus ed model and we e no included in he model whe e we did no adjus o he apo(a) iso o ms Fig. 1. Regional plo showing he genomic egion a ound he LPA gene (ch 6:159,991,850-161,753,083; LD e e s o s55730499, based on 1000G EUR); P alues a e de i ed om he me a-analysis on he i e coho s including 13,781 indi iduals on in e se-no mal ans o med Lp(a) concen a ions, adjus ed o age and sex. All 48 SNPs, which a e independen ly associa ed wi h Lp(a) in a join model, a e ci cled. a Uni e si y o Tampe e Lib a y, on Oc obe 13, 2017www.jl .o gDownloaded om .h ml h p://www.jl .o g/con en /suppl/2017/05/16/jl .M076232.DC1 Supplemen al Ma e ial can be ound a : 1838 Jou nal o Lipid Resea ch Volume 58, 2017 (supplemen al Table S4). Howe e , h ee o hem we e in LD ( 2 > 0.8) wi h SNPs om model 1 ( s140570886 in LD wi h s1510225 om model 1 and s3798220; s55730499 in LD wi h s118039278 om model 1 and s10466872; s59614420 in LD wi h s4252109 om model 1). The e- o e, 16 SNPs, which we e iden i ied in model 1, we e as- socia ed wi h Lp(a) independen ly om iso o ms and o he Lp(a)-associa ed SNPs. Bo h weigh ed and unweigh ed SNP-sco es we e de i ed om all 48 SNPs in he condi ional model 1. Figu e 3 shows ha median alues o Lp(a) inc eased om 2.1 mg/dl o indi iduals wi h he minimum numbe o Lp(a)-inc easing alleles o 91.1 mg/dl o indi iduals wi h he maximum numbe o Lp(a)-inc easing alleles. A linea inc ease was obse ed, bu in he medium ange o he isk sco e many ou lie s we e loca ed a he uppe ail o he Lp(a) dis ibu- ion. This migh be due o he skewed dis ibu ion o Lp(a) concen a ions and he e y wide ange o he singu- la SNP e ec s (see abo e). Ne e heless, he weigh ed SNP-sco e explained 36% o he pheno ypic a iance o Lp(a) concen a ions. E alua ion o po en ial unc ional o egula o y impac We i s in es iga ed o de e mine whe he he ma ke SNP o any o i s p oxies we e known coding a ian s (supplemen al Tables S4, S9). This esul ed in 12 SNPs be- ing loca ed in exons, 6 he eo being missense a ian s and 6 being synonymous a ian s. In e es ingly, all six missense a ian s we e loca ed in LPA, while only one o he six syn- onymous a ian s was loca ed in LPA. The s41272110 (LPA KIV-8), s41259144 (LPA KIV-4), and s41267807 (LPA, p o ease domain) we e conco dan ly epo ed as be- ing damaging by Polyphen (28) and SIFT (29), whe eas he well-known LPA SNP, s3798220, was classi ied as be- nign by SIFT, bu as “possibly damaging” by Polyphen. The wo emaining missense SNPs, s4252125 in PLG and s41267809 in LPA, we e conco dan ly p edic ed o be be- nign. The p e iously desc ibed SNP, s41272114, is loca ed on a splice si e and causes null alleles (30–32). O he known nonsense a ian s a e oo a e (33) o be de ec ed by GWASs o occu in di e en e hnici ies (34). Howe e , addi ional in silico e alua ion o hese SNPs was hampe ed by he ac ha GTEx, he la ges eQTL e- sou ce a ailable so a , does no ye epo eQTL da a o LPA in li e issue, while LPA exp ession is igh ly egula ed and happens nea ly exclusi ely in li e issue. Fu he e alua ion o he APOE gene egion Condi ional analyses, including SNPs om he APOE egion (lead SNP ±500 kb: ch 19: 44,912,079-45,912,079) Fig. 2. Regional plo showing he genomic egion de ined by he APOE lead SNP, s7412, ±500 kb (LD e e s o s7412, based on 1000G EUR); P alues a e de i ed om he me a-analysis on he i e coho s on in e se-no mal ans o med Lp(a) concen a ions, adjus ed o age and sex. a Uni e si y o Tampe e Lib a y, on Oc obe 13, 2017www.jl .o gDownloaded om .h ml h p://www.jl .o g/con en /suppl/2017/05/16/jl .M076232.DC1 Supplemen al Ma e ial can be ound a : Genome-wide me a-analysis on lipop o ein (a) 1839 yielded no addi ional genome-wide signi ican SNPs a e adjus ing o he op SNP ( s7412). This SNP de ined he APOE2 allele and explained 0.5% o he pheno ypic a iance o in e se-no mal ans o med Lp(a) le els. Each APOE2 copy dec eased Lp(a) concen a ions by 3.34 mg/dl co e- sponding o 15% o he popula ion’s mean alues (supple- men al Table S6). An ex ended e alua ion o APOE geno ypes was pe o med using da a om he KORA F3, KORA F4, and he SAPHIR s udies combined. The E2/E2 geno ype was associa ed wi h a dec ease o Lp(a) by 10.5 mg/dl com- pa ed wi h he E3/E3 geno ype (P = 7.77 × 105), while he e ec was abou hal as high (4.7 mg/dl) o E2/E3 (P = 1.05 × 108) and E2/E4 (P = 0.0191), indica ing a a he addi i e e ec o he E2 allele (Fig. 4). Con e sely, E4/E4 was associa ed wi h a nonsigni ican inc ease o Lp(a). Fu - he adjus ing o apo(a) iso o ms only a enua ed he e - ec es ima es o E4/E4, bu did no change he e ec es ima es and/o P alues o he o he APOE geno ypes. Gende -s a i ied analyses GWASs s a i ied o gende did no esul in any addi- ional genome-wide signi ican hi s ou side he wide LPA and APOE gene egion, nei he o men, no o women. The e was also no genome-wide signi ican SNP-gende in- e ac ion e ec o bo h adjus men models. Genome-wide a iance explained and he i abili y The genome-wide SNP-based explained a iance, including he en i e da ase o a ailable SNPs (genomic he i abili y), was es ima ed o be 49.3% in bo h KORA s udies [95% CI: (32.0%; 66.5%)]. A na ow-sense he i abili y h2 o Lp(a) was calcula ed o be 91.7% om he polygenic model in he amily-based FamHS s udy. This es ima e included no only he measu ed SNP e ec s, bu also unmeasu ed ac o s. Gene-based and candida e gene analyses The genome-wide gene-based associa ion scan esul ed in 23 signi ican genes o he in e se-no mal ans o med Lp(a) concen a ions in model 1 and in 20 signi ican genes in model 2. All o hem we e loca ed ei he in he b oad LPA o APOE gene egion (supplemen al Table S10). The gene-based es o associa ion o a lis o possible candida e genes consis ing o suspec ed Lp(a) ecep o s and egula o s discussed in he li e a u e esul ed in one signi ican associa ion o he TLR2 gene wi h Lp(a) o model 1 (P = 3.4 × 104; supplemen al Table S11). No sig- ni ican associa ion wi hin he candida e genes was ound o model 2 (supplemen al Table S11). Impac o Lp(a)-associa ed SNPs on CAD isk F om all 49 gene ic a ian s ha we e shown o be inde- penden ly associa ed wi h Lp(a) in model 1 (48 in he LPA gene egion plus s7412 in APOE), 40 we e p esen in sum- ma y-le el da a e ie ed om he CARDIoGRAMplusC4D conso ium. Nine low- equency a ian s (MAF <1%) we e missing. Figu e 5 shows how he e ec es ima es on Lp(a) ela e o he ORs o CAD isk. Se en SNPs we e e en sig- ni ican ly associa ed wi h CAD on a genome-wide scale (P < 5 × 108). The highes e ec was o s186696265, which showed an OR o 1.73 (P = 3.35 × 1030) wi h CAD isk o each copy o he mino allele (supplemen al Table S13). DISCUSSION The me a-analysis o i e di e en s udies e ealed SNPs in wo genomic egions o be signi ican ly associa ed wi h Lp(a) concen a ions. Wi hin a b oad egion su ounding he LPA gene, abou 2,000 genome-wide signi ican SNPs we e iden i ied, o which 48 we e s ill genome-wide signi i- can in a join model a e s epwise condi ional analysis. The apo(a) iso o ms de e mined ia Wes e n blo we e a ail- able o all 13,781 pa icipan s included in he me a-analysis. To ou knowledge, his is he la ges GWAS me a-analysis on Lp(a) concen a ions adjus ing o apo(a) iso o ms. Adjus ing o apo(a) iso o ms esul ed in 30 independen ly Fig. 3. Boxplo o Lp(a) concen a ion o g oups o a SNP-sco e [sum o Lp(a)-inc easing alleles] de i ed om he 48 independen SNPs in he b oad LPA gene egion. An unde lying ba plo shows he dis ibu ion o he sco e in KORA F3 and KORA F4. The blue line indica es he p edic ed alues o Lp(a) o mid-in e - al alues o he SNP-sco e, based on a linea eg es- sion om he SNP-sco e on Lp(a). a Uni e si y o Tampe e Lib a y, on Oc obe 13, 2017www.jl .o gDownloaded om .h ml h p://www.jl .o g/con en /suppl/2017/05/16/jl .M076232.DC1 Supplemen al Ma e ial can be ound a : 1840 Jou nal o Lipid Resea ch Volume 58, 2017 associa ed SNPs in he b oad LPA gene egion, mos o hem di e en om he SNPs ound in he p e ious model. In addi ion, we ound ha ca ie s o he APOE2 allele o he APOE locus had signi ican ly lowe Lp(a) le els com- pa ed wi h APOE E3/E3 geno ypes. The clinical impac o he SNPs ound o be associa ed wi h Lp(a) concen a ions was unde sco ed by he obse a ion ha se en SNPs we e e en signi ican ly associa ed wi h CAD on a genome-wide scale in he CARDIoGRAMplusC4D conso ium. Findings in he LPA gene egion The la ges con ibu o o Lp(a) concen a ion a iabili y was he numbe o he KIV epea s in he LPA gene. Fu - he mo e, se e al SNPs in he LPA gene egion we e ound o be associa ed wi h Lp(a) concen a ions. Fo many o he associa ions, i was no clea whe he hey causally egula e Lp(a) concen a ions, as discussed ecen ly (1). Howe e , he leading SNPs in GWASs on Lp(a) ha e no been such unc ional SNPs anyway, bu a he SNPs co ela ed wi h KIV epea s. Two SNPs ha e been epea edly epo ed and a e al eady well-es ablished: s10455872 and s3798220 (9–11, 13). Bo h SNPs ha e been ound o pa ially ag sho apo(a) iso o ms wi h 17–20 and 19–21 epea s, espec i ely. The e o e, ca ie s o he mino alleles o hese wo SNPs ha e highe Lp(a) le els and consequen ly highe isk o de eloping co ona y disease (10). Howe e , only abou hal o he sho high- isk apo(a) iso o ms a e agged by hese wo SNPs (35). Mo eo e , s3798220 has been epo ed o be associa ed wi h inc eased oxidized phospholipid (OxPL) ca iage on apoB-100 (36, 37). The 48 SNPs ha emained signi ican a e applying a s epwise condi ional model can be seen as a he indepen- den om each o he and explain 36% o he a iance in Lp(a) concen a ions. The lead SNP, which was selec ed in he condi ional analysis ( s118039278) was in pe ec LD ( 2 = 1) wi h he well-known KIV- agging SNP, s10455872 (P = 9.7 × 10418). The o he KIV- agging SNP, s3798220, was no included in he 48 SNP-sco e, bu was cap u ed by ano he SNP in pe ec LD ( s1510224). Th ee ou o he 48 SNPs ha e al eady been desc ibed in he li e a u e: s7770628 has been ound o be associa ed wi h Lp(a) con- cen a ions g ea e han 14 mg/dl (7); s41272114 causes a splicing de ec and esul s in a null-allele (38), and s186696265 has been iden i ied in wo ecen GWAS me a-analyses on lipid- ai s aiming o iden i y speci ically a e a ian s in al eady known lipid loci (39, 40). The SNP, s186696265, had a MAF o 1.1% (1000G phase1 3) and was shown o be associa ed wi h LDL choles e ol (LDL-C) and o al choles e ol, bu only in Eu opeans. In ou analy- sis, his SNP had he highes e ec size on Lp(a) concen- a ions; each copy o he mino allele o his SNP inc eased Lp(a) le els by 65 mg/dl, o by 48 mg/dl in a join model a e adjus men o all o he 47 o he independen SNPs. As Lp(a) pa icles con ain oughly 30% choles e ol (41) ha is included in each common choles e ol measu e- men me hod, he ecen ly ound associa ion o his SNP wi h LDL-C and o al choles e ol (39, 40) migh be ex- plained by he e ec o his SNP on Lp(a) concen a ions. An Lp(a)-inc easing e ec o 65 mg/dl would ansla e in o an inc ease o LDL-C o 20 mg/dl. O no e, his SNP Fig. 4. Resul s o a mixed model (using da a om KORA F3, KORA F4, and SAPHIR combined) e alua ing he e ec s om APOE geno- ypes, de ined as desc ibed in supplemen al Table S3, on un ans o med Lp(a) alues in millig ams pe decili e (A), as well as on in e se- no mal ans o med Lp(a) (B). Bo h panels show  es ima es and 95% CI o age- and sex-adjus ed models (in black), as well as age-, sex-, and iso o m-adjus ed models (in blue). P alues a e de i ed om he model using in e se-no mal ans o med Lp(a). a Uni e si y o Tampe e Lib a y, on Oc obe 13, 2017www.jl .o gDownloaded om .h ml h p://www.jl .o g/con en /suppl/2017/05/16/jl .M076232.DC1 Supplemen al Ma e ial can be ound a : Genome-wide me a-analysis on lipop o ein (a) 1841 is loca ed be ween wo in e genic enhance egions, which ha e been p oposed o egula e LPA exp ession (42) (sup- plemen al Fig. S8) and is in pa ial LD ( 2 = 0.69, D′ = 0.95) wi h a SNP wi hin he pu a i e egula o y enhance e- gion, DHIII ( s7758766) (43). This SNP also shows a pa ial LD ( 2 = 0.64) wi h s3798220, desc ibed by Cla ke e al. (10). To de ec SNPs whose e ec s did no me ely e lec LD wi h he KIV epea , we used models addi ionally adjus ing o apo(a) iso o ms. Fo me s udies looking a co ela ions o SNPs in he LPA gene egion wi h KIV epea s o iso o m size led o inconsis en esul s: one s udy ound no co - ela ion o SNPs wi h KIV epea s (44), while o he s conco dan ly ound a high co ela ion o one speci ic SNP ( s10455872) wi h KIV copy numbe and sho iso o m (10, 45). The la ges GWAS up o now ha adjus ed o apo(a) iso o ms was pe o med in 1,376 Old O de Amish (13). Howe e , only SNPs on ch omosome 6q25-26 we e adjus ed o iso o ms, which esul ed in a modes educ- ion o he s eng h o he SNPs in his egion wi h he Lp(a) choles e ol le els. As he published esul s a e he - e ogeneous, we concluded ha a sys ema ic adjus men o he apo(a) iso o ms on a la ge scale was needed o: 1) dis- en angle he in luence o pa ially co ela ed a ian s on Lp(a) le els wi h he LPA gene egion; and 2) inc ease powe o de ec ing o he genes ha ha e an in luence on Lp(a) le els. Howe e , adjus ing o apo(a) iso o ms is no s aigh o wa d. The analysis o iso o ms on he p o ein le el by Wes e n blo p o ides allele-speci ic iso o m sizes. Because no all apo(a) iso o ms a e exp essed, he de ec ed apo(a) iso o ms do no ully e lec he alleles on he DNA le el. Fu he mo e, e en i bo h iso o ms can be de ec ed, one o hem, usually he sho e one, migh be p edomi- na ing in plasma. F om se e al di e en s a is ical model- ing s a egies, he p edominan ly exp essed apo(a) iso o m explains he highes a iabili y o Lp(a) and is, in ou iew, he bes way o accoun o apo(a) iso o ms in s a is ical models. S ill, he con ibu ion o he second iso o m is igno ed and he e o e, adjus ing o he p edominan ly exp essed apo(a) iso o m can only pa ly emo e he asso- cia ion wi h SNPs ha is jus due o LD wi h KIV epea s. Consequen ly, known KIV- agging SNPs ( s10455872, s3798220) we e s ill highly signi ican ly associa ed wi h Lp(a) concen a ions a e adjus ing o apo(a) iso o ms, bu he associa ion was a enua ed massi ely. In his iso- o m-adjus ed model, a se o 30 SNPs independen ly associa ed wi h Lp(a) was iden i ied by using a s epwise condi ional model, 16 o hem unique and no co e- la ed wi h any SNP con ained in model 1. The op-hi ( s75692336) ep esen ed a clus e o co ela ed SNPs spanning o e he LPA gene, he in e genic egion be- ween LPA and PLG, and he PLG gene. I was highly co - ela ed wi h he missense a ian , s41272110, in KIV-8 [ 2 = 0.87; also known as T23P (30) o T12P (46)], which was p edic ed o be dele e ious o he p o ein unc ion and has been associa ed wi h educed Lp(a) le els be o e (30, 46). Mo eo e , we ecen ly ound s75692336 o ag a e- quen splice-si e a ian wi hin he KIV-2 epea , ha ex- plains 20% o he Lp(a) a iance in low molecula weigh iso o m ca ie s (47). This splice-si e a ian was disco - e ed by means o a nex gene a ion ul a-deep sequencing app oach, which aimed o explain Lp(a) alues ha we e disco dan o wha one would expec om he espec i e apo(a) iso o ms. In he same manne , ou iso o m-adjus men app oach seemed o ind no only iso o m-independen SNPs, bu pa icula ly SNPs leading o Lp(a) alues ha de ia ed om wha would be expec ed gi en he obse ed iso o m. As in he i s model, he highes e ec size on Lp(a) concen a ions was ound o SNP s186696265, loca ed nea he known in e genic enhance egions (supplemen al Fig. S8). Two ou o he 16 unique and independen ly associa ed SNPs we e al eady desc ibed in he li e a u e. The SNP, s3798221, was al eady ound o be associa ed wi h Lp(a) le els (10), MI (48), Lp(a) le els in women (49), and was ound o be associa ed wi h Lp(a) concen a ion and ma ginally associa ed wi h KIV copy numbe in h ee e hnici ies (Sou h Asians, Chinese, and Eu opean Caucasians) (45). The SNP, s56393506, was al- eady ound he be associa ed wi h Lp(a) le els (50). Be- cause hese 16 independen SNPs we e only iden i ied by adjus ing o he apo(a) iso o m, hey migh exe an e ec on Lp(a) and, as a esul , on CAD isk only in subg oups o Fig. 5. Sca e plo showing he e ec es ima es on in e se no mally ans o med Lp(a) le els (±95% CI) on he x axis and he ORs o CAD isk (±95% CI) on he y axis (de i ed om heCARDIoGRAMplusC4D conso ium) o all 40 SNPs, which we e iden i ied in model 1 (age- and sex-adjus ed) and which we e a ail- able in he CARDIoGRAMplusC4D esul s. All SNPs, which a e also genome-wide signi ican ly associa ed wi h CAD isk, a e ma ked in g een. a Uni e si y o Tampe e Lib a y, on Oc obe 13, 2017www.jl .o gDownloaded om .h ml h p://www.jl .o g/con en /suppl/2017/05/16/jl .M076232.DC1 Supplemen al Ma e ial can be ound a : 1842 Jou nal o Lipid Resea ch Volume 58, 2017 iso o ms (e.g., only in low o high molecula weigh iso- o ms), which should be subjec o u he esea ch. Associa ion wi h APOE P e ious s udies e alua ing he associa ion o a ian s in he APOE gene wi h Lp(a) we e mos ly non-GWASs and p oduced inconsis en esul s (supplemen al Table S12). While some epo ed no con ibu ion o APOE o Lp(a) a ia ion, o he s showed an Lp(a)-lowe ing e ec o APOE2 compa ed wi h APOE3, pa ially in a con ex -dependen manne , e.g., in women o apo(a) iso o m dependen . We clea ly show ha he Lp(a)-lowe ing e ec o APOE2 did no depend on ei he gende o apo(a) iso o m size. apoE is pa o chylomic ons, emnan s, and VLDL, IDL, and HDL pa icles and media es emo al o emnan s om ci cula ion ia LDLR, LRP1, and HSPG ecep o s (51). apoE has also been epo ed in a iglyce ide- ich subspe- cies o ci cula ing Lp(a) pa icles (52). Two missense SNPs ( s7412, s429358) de ine h ee iso o ms named E2, E3, and E4, he wo mu an iso o ms E2 and E4 p esen ing ma kedly di e en physiological unc ions (51). E2 p es- en s only 1% o he LDLR binding a ini y compa ed wi h E3 and E4 (51) and esul s in impai ed emo al o VLDL pa icles (53). Coincidence o homozygous APOE2 and condi ions causing high VLDL, such as obesi y, led o hype lipop o einemia ype III (54). The obse a ion ha APOE2 was associa ed wi h lowe Lp(a) le els con i ms p e ious epo s (15, 55), albei he mechanism has no been elucida ed ye . E ec s o he APOE geno ype on bo h Lp(a) ca abolism and syn hesis ha e been p oposed (15). Assuming a compe i ion o LDL-C, Lp(a), and apoE-ca ying iglyce ide- ich lipop o eins o common ecep o s, i has been sugges ed ecen ly ha Lp(a) concen a ions a e dec eased in APOE2 ca ie s be- cause he dec eased ecep o a ini y o apoE2 inc eases he numbe o ecep o s a ailable o LDL-C and Lp(a) (15). On he o he hand, an e ec on syn hesis a e is also concei able. APOE geno ypes exe se e al seconda y e ec s on lipop o ein me abolism by al e ing he hepa ic lipop o ein emnan me abolism (55). Gi en ha Lp(a) is syn hesized by he li e (3) and syn hesis a e ep esen s he main de e minan o Lp(a) le els (56), APOE2 may hus a ec he a ailabili y o subs a e o Lp(a) o ma ion (55). Gene-based candida e gene analyses To in es iga e mino e ec s no cap u ed by a single- SNP analysis, we pe o med a gene-based analysis o candi- da e genes ha we e epo ed o bind Lp(a) and/o egula e LPA o PCSK9. Finally, only TLR2 showed a signi i- can associa ion wi h Lp(a) le els. TLR2 has been shown o concu wi h CD36 and TLR6 in Lp(a)- anspo ed OxPL (57). Howe e , o ou knowledge, he e e se e ec o an in luence o TLR2 on Lp(a) le els has no been shown ye . Because nei he sca enge ecep o CD36, which in e nal- izes OxPL, no TLR6, which dime izes wi h TLR2 in OxPL ecogni ion (57), we e signi ican in ou analysis; he bio- logical basis o his associa ion migh ins ead eside in he o he mani old oles o TLR2 in inna e immuni y and in lam- ma ion (58). Acu e phase eac ions ha e been p oposed o a ec Lp(a) le els, albei he magni ude and di ec ion e- mains con o e sial (1). Fu he mo e, in mice, TLR2 ac i- a ion was also shown o esul in a 14- old inc ease in PCSK9 exp ession (59), ano he known egula o o Lp(a) in humans. Thus a ious po en ial ou es o ac ion o TLR2 on Lp(a) le els exis , al hough s ic eplica ion and u - he s udies will be needed o elucida e he mechanism o his inding. He i abili y and explained a iance o Lp(a) concen a ions The genome-wide-SNP-based explained a iance, in- cluding he en i e da ase o a ailable SNPs (genomic he i- abili y), was es ima ed o be 49.3% in bo h KORA s udies. In con as , he na ow-sense he i abili y h2 o Lp(a) de- i ed om he polygenic model in he amily-based FamHS s udy, was es ima ed o be 91.7%. The genomic he i abili y only includes he measu ed SNP e ec s. The na ow-sense he i abili y also includes unmeasu ed SNP e ec s, CNVs, and o he gene ic e ec s. A majo eason o he big di e - ence be ween hese wo he i abili y es ima es is mos p ob- ably due o he KIV epea s, which a e only pa ly co e ed by LD wi h he measu ed SNP e ec s. The weigh ed SNP-sco e ha was de i ed based on he 48 independen ly associa ed SNPs om model 1 explained 36% o he pheno ypic a iance o Lp(a) concen a ions. As wo SNPs o he SNP-sco e ( s118039278 and s1510224) we e in pe ec LD wi h he SNPs, s10455872 and s3798220, pa ially agging KIV, his es ima e o explained a iance migh also be in luenced by e ec s om he KIV epea s and no solely om he SNPs hemsel es. These wo SNPs, which a e usually aken as gene ic ins umen s o p edic Lp(a), join ly explain oughly 20% o he pheno ypic a i- ance o Lp(a) in he KORA s udies we used o ou me a- analyses. The e o e, we would expec an imp o emen in accu acy in analyses con aining a SNP sco e when aking he 48 SNPs ins ead o aking s10455872 and s3798220 only. Impac o Lp(a)-associa ed SNPs on CAD isk We obse ed a di ec p opo ional ela ionship o Lp(a)-associa ed a ian s wi h CAD isk inc ease o Lp(a)- inc easing a ian s and a dec ease in isk o Lp(a)-dec easing a ian s. This is in line wi h se e al p e ious s udies, which we e pe o med p ima ily on he highly ci ed a ian s, s10455872, s3798220, and s41272114 (10, 60, 61). In p e- ious s udies ha e alua ed he causal ela ionship o Lp(a) wi h CAD, s10455872 and s3798220, especially, ha e been used as ins umen al a iables o de e mine he “gene i- cally egula ed” p opo ion o Lp(a). Al hough bo h SNPs a e no included in ou iden i ied SNP se , hey a e e- placed by o he a ian s ( s118039278 and s3798220) in pe ec LD, which a e associa ed wi h an OR o CAD isk o 1.36 and 1.40, espec i ely. We could also show ha he SNP, s186696265, which p esen s he highes e ec size on Lp(a) in ou in es iga ion, also p esen s he highes OR o CAD o all in es iga ed SNPs (OR = 1.73) in he LPA gene egion. The e o e, wi h espec o CAD isk, s186696265 seems o be a leas equi alen o e en supe io o he a Uni e si y o Tampe e Lib a y, on Oc obe 13, 2017www.jl .o gDownloaded om .h ml h p://www.jl .o g/con en /suppl/2017/05/16/jl .M076232.DC1 Supplemen al Ma e ial can be ound a :