Pro-inflammatory fatty acid profile and colorectal cancer risk: A Mendelian randomisation analysis

O iginal Resea ch P o-in lamma o y a y acid p o ile and colo ec al cance isk: A Mendelian andomisa ion analysis Sebas ian May-Wilson a,1 , Ami Sud a,1 , Philip J. Law a,1 , Kimmo Palin b,c , Sa i Tuupanen b,c , Alexand a Gyl e b,c , Ul ika A. Ha ¨nninen b,c , Ta iana Cajuso b,c , Tomas Tanskanen b,c , Johanna Kondelin b,c , Ee i Kaasinen b,c , An i-Pekka Sa in d , Johan G. E iksson g,h,i , Ha i Rissanen g , Paul Knek g , Ee o Pukkala j,k , Pekka Jousilah i g , Veikko Salomaa g , Samuli Ripa i d,e, , Aa no Palo ie d,l,m,n , Lau a Renkonen-Sinisalo o , Anna Lepis o ¨ o , Jan Bo ¨hm p , Jukka-Pekka Mecklin q , Nada A. Al-Tassan , Clai e Palles s , Susan M. Fa ing on , Ma ia N. Timo ee a , B ian F. Meye , Salma M. Wakil , Ha y Campbell u , Ch is ophe G. Smi h , Shelley Idziaszczyk , Timo hy S. Maughan w , Da id Fishe x , Rachel Ke y , Da id Ke z , Michael N. Passa elli aa , Jane C. Figuei edo ab,ac , Daniel D. Buchanan ad,ae , Aung K. Win ae , John L. Hoppe ae , Ma k A. Jenkins ae , No alane M. Lindo a , Polly A. Newcomb ag , S e en Gallinge ah , Da id Con i ai , F ed Schumache ai , G aham Casey aj , Lau i A. Aal onen b,c , Je emy P. Cheadle , Ian P. Tomlinson s , Malcolm G. Dunlop , Richa d S. Houls on a, * a Di ision o Gene ics and Epidemiology, The Ins i u e o Cance Resea ch, London, SW7 3RP, UK b Genome-Scale Biology Resea ch P og am, Resea ch P og ams Uni , Uni e si y o Helsinki, Helsinki, 00014, Finland c Depa men o Medical and Clinical Gene ics, Medicum, Uni e si y o Helsinki, Helsinki, 00014, Finland d Ins i u e o Molecula Medicine Finland (FIMM), Uni e si y o Helsinki, Helsinki, 00014, Finland e Wellcome T us Sange Ins i u e, Wellcome T us Genome Campus, Hinx on, Camb idge, CB10 1SA, UK Depa men o Public Heal h, Uni e si y o Helsinki, Helsinki, 00014, Finland g Na ional Ins i u e o Heal h and Wel a e, Helsinki, 00271, Finland h Folkha ¨lsan Resea ch Cen e, Helsinki, 00250, Finland i Uni o Gene al P ac ice and P ima y Heal h Ca e, Uni e si y o Helsinki and Helsinki Uni e si y Hospi al, Helsinki, 00014, Finland *Co esponding au ho : Di ision o Gene ics and Epidemiology, The Ins i u e o Cance Resea ch, 15 Co swold Road, London, SM2 5NG, UK. Fax: þ44 (0) 722 4365. E-mail add ess: [email p o ec ed] (R.S. Houls on). 1 These au ho s con ibu ed equally o his wo k. h p://dx.doi.o g/10.1016/j.ejca.2017.07.034 0959-8049/ª2017 The Au ho s. Published by Else ie L d. This is an open access a icle unde he CC BY license (h p://c ea i ecommons.o g/ licenses/by/4.0/). A ailable online a www.sciencedi ec .com ScienceDi ec jou nal homepage: www.ejcance .com Eu opean Jou nal o Cance 84 (2017) 228e238 j Finnish Cance Regis y, Ins i u e o S a is ical and Epidemiological Cance Resea ch, Helsinki, 00130, Finland k School o Heal h Sciences, Uni e si y o Tampe e, Tampe e, 33014, Finland l Analy ic and T ansla ional Gene ics Uni , Depa men o Medicine, Massachuse s Gene al Hospi al, Bos on, MA, 02114, USA m P og am in Medical and Popula ion Gene ics, The B oad Ins i u e o MIT and Ha a d, Camb idge, MA, 02142, USA n Depa men o Neu ology, Massachuse s Gene al Hospi al, Bos on, MA, 02114, USA o Abdominal Cen e , Depa men o Su ge y, Helsinki Uni e si y Hospi al, Helsinki, 00029, Finland p Depa men o Pa hology, Cen al Finland Cen al Hospi al, Jy a ¨skyla ¨, 40620, Finland q Depa men o Su ge y, Jy a ¨skyla ¨Cen al Hospi al, Uni e si y o Eas e n Finland, Jy a ¨skyla ¨, 40620, Finland Depa men o Gene ics, King Faisal Specialis Hospi al and Resea ch Cen e , Riyadh, 12713, Saudi A abia s Molecula & Popula ion Gene ics Labo a o y, Wellcome T us Cen e o Human Gene ics, Uni e si y o Ox o d, Ox o d, OX3 7BN, UK Colon Cance Gene ics G oup, Uni e si y o Edinbu gh and MRC Human Gene ics Uni , Wes e n Gene al Hospi al, Edinbu gh, EH4 2XU, UK u Cen e o Popula ion Heal h Sciences, Uni e si y o Edinbu gh, Edinbu gh, EH8 9AG, UK Di ision o Cance and Gene ics, School o Medicine, Ca di Uni e si y, Ca di , CF14 4XN, UK w CRUK/MRC Ox o d Ins i u e o Radia ion Oncology, Uni e si y o Ox o d, Ox o d, OX3 7DQ, UK x MRC Clinical T ials Uni , A ia ion House, London, WC2B 6NH, UK y Ox o d Cance Cen e, Depa men o Oncology, Uni e si y o Ox o d, Chu chill Hospi al, Ox o d, OX3 7LE, UK z Nu ield Depa men o Clinical Labo a o y Sciences, Uni e si y o Ox o d, John Radcli e Hospi al, Ox o d, OX3 9DU, UK aa Depa men o Epidemiology, Geisel School o Medicine a Da mou h, Da mou h-Hi chcock Medical Cen e , Lebanon, NH, 03756, USA ab Samuel Oschin Comp ehensi e Cance Cen e , Ceda s-Sinai Medical Cen e , Los Angeles, CA, USA ac Keck School o Medicine, Uni e si y o Sou he n Cali o nia, Los Angeles, CA, USA ad Colo ec al Oncogenomics G oup, Gene ic Epidemiology Labo a o y, Depa men o Pa hology, The Uni e si y o Melbou ne, Vic o ia, 3010, Aus alia ae Cen e o Epidemiology and Bios a is ics, The Uni e si y o Melbou ne, Vic o ia, 3010, Aus alia a Depa men o Heal h Sciences Resea ch, Mayo Clinic, Sco sdale, AZ, 85259, USA ag Cance P e en ion P og am, F ed Hu chinson Cance Resea ch Cen e , Sea le, WA, 98109, USA ah Lunen eld-Tanenbaum Resea ch Ins i u e, Moun Sinai Hospi al, To on o, ON, M5G 1X5, Canada ai Depa men o P e en i e Medicine, Uni e si y o Sou he n Cali o nia, Los Angeles, CA, 90033, USA aj Cen e o Public Heal h Genomics, Uni e si y o Vi ginia, Cha lo es ille, VA, 22908, USA Recei ed 16 Decembe 2016; ecei ed in e ised o m 20 July 2017; accep ed 22 July 2017 A ailable online 19 Augus 2017 KEYWORDS Mendelian andomisa ion; Colo ec al cance ; Risk; Plasma a y acids; Fa y acids Abs ac Backg ound: While die a y a has been es ablished as a isk ac o o colo ec al cance (CRC), associa ions be ween a y acids (FAs) and CRC ha e been inconsis en . Using Mendelian andomisa ion (MR), we sough o e alua e associa ions be ween polyunsa u a ed (PUFA), monounsa u a ed (MUFA) and sa u a ed FAs (SFAs) and CRC isk. Me hods: We analysed geno ype da a on 9254 CRC cases and 18,386 con ols o Eu opean ances y. Ex e nally weigh ed polygenic isk sco es we e gene a ed and used o e alua e asso- cia ions wi h CRC pe one s anda d de ia ion inc ease in gene ically de ined plasma FA le els. Resul s: Risk educ ion was obse ed o oleic and palmi oleic MUFAs (OR OA Z0.77, 95% CI: 0.65e0.92, PZ3.9 10 3 ;OR POA Z0.36, 95% CI: 0.15e0.84, PZ0.018). PUFAs li- noleic and a achidonic acid had nega i e and posi i e associa ions wi h CRC espec i ely (OR LA Z0.95, 95% CI: 0.93e0.98, PZ3.7 10 4 ;OR AA Z1.05, 95% CI: 1.02e1.07, PZ1.7 10 4 ).The SFA s ea ic acid was associa ed wi h inc eased CRC isk (OR SA Z1.17, 95% CI: 1.01e1.35, PZ0.041). Conclusion: Resul s om ou analysis a e b oadly consis en wi h a p o-in lamma o y FA p o ile ha ing a de imen al e ec in e ms o CRC isk. ª2017 The Au ho s. Published by Else ie L d. This is an open access a icle unde he CC BY license (h p://c ea i ecommons.o g/licenses/by/4.0/). S. May-Wilson e al. / Eu opean Jou nal o Cance 84 (2017) 228e238 229 1. In oduc ion Colo ec al cance (CRC) is one o he mos common cance s and a majo cause o cance - ela ed mo ali y in economically de eloped coun ies [1]. Geog aphical di e ences in CRC incidence be ween coun ies and mig a ion s udies ha e es ablished he impo ance o li es yle and die as majo de e minan s o CRC isk [2]. Wo ldwide CRC is cu en ly diagnosed in o e one million indi iduals annually; howe e , i s incidence is se o inc ease wi h adop ion o wes e n li es yles in de el- oping coun ies [3]. Gi en he impo ance o die as a isk ac o o CRC, i s modi ica ion o e s he p ospec o impac ing signi ican ly on disease incidence h ough public heal h ini ia i es. Die a y a has been widely implica ed as a isk ac o o cance , and me a-analyses o epidemiological s udies ha e ended o associa e CRC isk wi h a highe con- sump ion o ed and p ocessed mea [4]. The associa ion be ween a in ake on cance isk howe e , is likely o depend no only on he quan i y, bu also on he speci ic ype o a y acid (FA). Animal models and ecological s udies ha e ended o implica e animal a [5], sa u a ed a y acid (SFA) and ce ain omega-6 polyunsa u a ed a y acids (u-6 PUFAs) wi h an inc eased isk, and u-3 PUFA in ake wi h a educed isk [6e8]. E idence o a causal ela ionship wi h in ake o speci ic ypes o a om epidemiological s udies has howe e la gely been inconclusi e. Reasons o inconsis encies in obse a- ional s udies include he inhe en p oblem o elici ing accu a e measu emen s o long- e m die , con ounding and e e se causa ion [9]. Mendelian andomisa ion (MR) analysis ep esen s an adjunc o he con en ional epidemiological obse - a ional s udy o examining associa ions be ween an exposu e wi h a disease. The MR s a egy makes use o allelic a ian s ha a e andomly assigned du ing meiosis and a e obus ly associa ed wi h ai s o in e - es , as ins umen al a iables (IVs). Using gene ically de ined IVs as p oxies o modi iable exposu e a oids con ounding by en i onmen al ac o s, is no subjec o e e se causali y and can in o m on li e-long exposu e [10,11]. Since s udies ha e shown ha FA in ake in- luences plasma le els o FAs in heo y MR makes an a ac i e s a egy o link die a y FA o CRC isk [12,13]. We ha e he e o e sough o iden i y associa ions be ween gene ically p edic ed plasma PUFA, MUFAs and SFA le els and CRC isk. Speci ically: (1) he u-6 PUFAs, linoleic acid (LA), a achidonic acid (AA) and dihomo-g-linolenic acid (DGLA); (2) he u-3 PUFAs, eicosapen aenoic acid (EPA), docosapen aenoic acid (DPA) and docosahexaenoic acid (DHA); (3) he MUFAs, oleic acid (OA) and palmi oleic acid (POA); and (4) he SFAs, palmi ic acid (PA), a achidic acid and s ea ic acid (SA). 2. Me hods 2.1. Colo ec al cance da ase s We in es iga ed he ela ionship be ween gene ic isk sco es o le els o MUFAs, PUFAs, and SFAs and CRC isk adop ing a wo-sample MR s a egy using da a om se en epo ed genome-wide associa ion s udies (GWAS) o CRC (Table 1). B ie ly, hese GWAS we e based on indi iduals wi h Eu opean ances y: CCFR1, CCFR2, COIN, FINLAND, UK1, Sco land1 and VQ58 [14]. Each s udy was app o ed by espec i e ins i u ional e hics e iew boa d and pe o med/con- duc ed in acco dance wi h he Decla a ion o Helsinki. 2.2. Geno yping da a Comp ehensi e de ails o he geno yping and quali y con ol o he se en GWAS ha e been p e iously e- po ed [14]. B ie ly, we excluded single nucleo ide poly- mo phisms (SNPs) wi h a mino allele equency o <1%, low call a e <95%, hose SNPs iola ing Ha - dyeWeinbe g equilib ium, and indi iduals wi h non- Eu opean ances y as assessed using da a om Hap- Map 2 [15]. IMPUTE 2 so wa e [16] was used o eco e un yped SNP geno ypes using a me ged e e - ence panel consis ing o Sequencing Ini ia i e Suomi ( o he FINLAND da a) o UK10K ( o he emaining da a) and 1000 Genomes P ojec da a [17,18]. Poo ly impu ed SNPs, de ined by an INFO sco e o <0.9, we e excluded. Summa y s a is ics om he se en GWAS we e used o calcula e he odds a ios (ORs) o FA- ela ed SNPs. 2.3. Gene a ian s used o cons uc gene ic isk sco es Gene ic isk sco es o IVs o each plasma FA we e de eloped om SNPs p e iously iden i ied by The Co- ho s o Hea and Aging Resea ch in Genomic Epidemiology (CHARGE) Conso ium. We conside ed SNPs associa ed a genome-wide signi icance (i.e. P5.0 10 8 ) in indi iduals wi h Eu opean Ances y. To a oid co-linea i y be ween SNPs o each FA we imposed a h eshold 2 alue o 0.01 o linkage disequilib ium (LD) including only he SNPs wi h he s onges e ec on he ai in gene ic isk sco es (Table 2,[19e22]). Fo each iden i ied SNP, we eco e ed he ch omosome posi ions, he isk alleles, associa ion es i- ma es and s anda d e o s. Fo each SNP, he allele ha was associa ed wi h inc eased FA le el was conside ed he e ec allele. 2.4. S a is ical analysis The associa ion be ween he plasma le el o each FA and CRC was examined using MR on summa y S. May-Wilson e al. / Eu opean Jou nal o Cance 84 (2017) 228e238230 Table 1 Summa y o he se en colo ec al cance genome-wide associa ion s udies. Se ies S udy se ing S udy cen e Geno yping pla o m No. cases No. con ols CCFR1 Colon Cance Family Regis y Uni e si y o Sou he n Cali o nia Illumina 1M, 1M Duo 1290 1055 CCFR2 Colon Cance Family Regis y Uni e si y o Sou he n Cali o nia Illumina 1M, Omni exp ess 796 2236 COIN COIN ial: Mul icen e s udy o ce uximab and o he he apies in me as a ic CRC. Con ols we e unselec ed blood dono s Ca di Uni e si y A yme ix Axiom 2244 2162 FINLAND Finnish Colo ec al Cance P edisposi ion S udy Helsinki Uni e si y Illumina 610K/Illumina HumanOmni2.5M 1172 8266 UK1 CORGI (colo ec al Tumou Gene Iden i ica ion Conso ium) Ox o d Uni e si y Illumina Hap550 940 965 Sco land1 COGS (Colo ec al Cance Suscep ibili y S udy) Edinbu gh Uni e si y Illumina Hap300/240S 1012 1012 VQ58 Cases: VICTOR, pos - ea men s ages o a phase III, andomised ial o o ecoxib (VIOXX) in pa ien s a e po en ially cu a i e he apy. QUASAR2, mul i-cen e s udy o capeci abine be acizumab as adju an ea men . 1958 Bi h coho con ols Ox o d Uni e si y Illumina Hap300/370, Illumina 1M 1800 2690 Table 2 E ec sizes o plasma a y acid con en (pe s anda d de ia ion inc ease in le els) o genome-wide signi ican (P<510 8 ) ins umen al a iables epo ed by CHARGE conso ium. FA sub ype Fa y acid SNP ID Ch Posi ion (bp) a Allele bS dE P- alue Va iance explained b SFA A achidic acid (20:0) s680379 20 12917400 A/G 0.098 0.01 5.81 10 13 e Palmi ic acid (PA) (16:0) s2391388 1 95485825 C/A 0.18 0.03 2.72 10 11 0.21e0.98% S ea ic acid (SA) (18:0) s6675668 1 95515637 G/T 0.17 0.02 2.16 10 18 0.37e1.39% s11119805 1 211918244 T/A 0.17 0.03 2.8 10 09 <0.01e0.72 s102275 11 61557803 T/C 0.18 0.02 1.33 10 20 0.33e1.34% u-3 PUFA Docosahexaenoic acid (DHA) (22:6n-3) s2236212 6 10995015 G/C 0.11 0.01 1.26 10 15 0.7% Docosapen aenoic acid (DPA) (22:5n-3) s780094 2 27741237 T/C 0.02 0.003 9.04 10 09 e s3734398 6 10982973 C/T 0.04 0.003 9.71 10 43 8.6% s174547 11 61570783 T/C 0.07 0.003 3.79 10 154 2.8% Eicosapen aenoic acid (EPA) (20:5n-3) s3798713 6 11008622 C/G 0.035 0.005 1.93 10 12 0.4% u-6 PUFA A achidonic acid (AA) (20:4n-6) s174547 11 61570783 T/C 1.69 0.03 3.30 10 971 3.7e37.6% s16966952 16 15135943 G/A 0.2 0.03 2.43 10 10 0.1e0.6% Dihomo-g-linolenic acid (DGLA) (20:3n-6) s174547 11 61570783 C/T 0.36 0.01 2.63 10 151 8.7e11.1% s16966952 16 15135943 G/A 0.22 0.02 7.55 10 65 2.0e4.5% Linoleic acid (LA) (18:2n-6) s10740118 10 65101207 G/C 0.25 0.04 8.08 10 09 0.2e0.7% s174547 11 61570783 C/T 1.47 0.04 4.98 10 274 7.6e18.1% s16966952 16 15135943 A/G 0.35 0.04 1.23 10 15 0.5e2.5% u-7 MUFA Palmi oleic acid (POA) (16:1n-7) s780093 2 27742603 T/C 0.02 0.003 9.80 10 10 0.23e0.93% s6722456 2 134529091 G/A 0.05 0.009 4.12 10 08 <0.01e0.57 s603424 10 102075479 G/A 0.03 0.004 5.69 10 15 0.28e1.57% s11190604 10 102302457 G/A 0.02 0.004 5.69 10 09 0.02e0.71% s102275 11 61557803 C/T 0.02 0.003 6.60 10 13 0.15e1.03% u-9 MUFA Oleic acid (OA) (18:1n-9) s102275 11 61557803 C/T 0.23 0.02 2.19 10 32 0.32e2.14% FA, a y acid; SNP, single nucleo ide polymo phism; bp, base pai ; SFA, sa u a ed a y acid; MUFA, monounsa u a ed a y acid; PUFA, polyunsa u a ed a y acid; S dE , s anda d e o . E ec allele in luencing each FA ai is ma ked in bold. a hg19 NCBI build. b Taken om CHARGE conso ium, as a pe cen age o o al se um a y acids, calcula ed by (b 2 *2*MAF*(1-MAF))/Va (Y) whe e bis he eg ession coe icien , MAF is he mino allele equency and Va (Y) is he a iance in le els o he a y acid. IVs ob ained om Re s. [19,20,22]. S. May-Wilson e al. / Eu opean Jou nal o Cance 84 (2017) 228e238 231 s a is ics as pe Bu gess (2015) [23]. The a io es ima e ðbbÞo all SNPs associa ed wi h each a y acid, com- bined, on CRC was calcula ed as ollows: bbZPkXkYks2 Yk PkX2 ks2 Yk : whe e X k co esponds o he associa ion o SNP k(as log o he OR pe isk allele) wi h he a y acid ai Y, Y k is he associa ion be ween SNP kand CRC isk (as log o he OR) wi h s anda d e o sYk. The es ima e o ðbbÞ ep esen s he causal inc ease in he log odds o he CRC, pe uni change in a y acids. The s anda d e o o he combined a io es ima e is gi en by: sebbZffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffi 1 PkX2 ks2 Yk s: A me a-analysis o s a is ics o each speci ic FA gene a ed o each CRC coho was combined unde ixed-e ec s models o de i e he summa y ORs and con idence in e als (CIs). To assess he impac o be- ween s udy he e ogenei y, we also de i ed ORs unde a andom-e ec s model. A cen al ene in MR is he absence o pleio opy (i.e. a gene in luencing mul iple ai s) be ween he SNPs in luencing CRC isk and FA le els. This would be e ealed as de ia ion om a linea ela ionship be ween SNPs and hei e ec size o any FA and CRC isk. To examine o iola ion o he s anda d IV assump ions in ou analysis, we pe o med in e se a ian weigh ed (IVW) and MR-Egge eg ession es s [24]. We conside ed a signi icance le el o P0.05 as being sa is ac o y o de i e a conclusion. While o di- na ily i would be app op ia e o impose a Bon e oni- co ec ed h eshold, his assumes an independence o IVs ac oss all FA ai s, which is no he case in he p esen analysis. All s a is ical analyses we e unde aken using R e sion 3.1 so wa e [25]. 2.5. Exp ession quan i a i e ai locus analysis To examine he ela ionship be ween SNP geno ype and exp ession o FA me abolism genes, we pe o med exp ession quan i a i e ai locus (eQTL) analysis using da a om The Cance Genome A las (TCGA) and he geno ype issue exp ession (GTEx)p ojec [26,27]. 3. Resul s The FA-associa ed gene ic a ian s and hei GWAS- epo ed cha ac e is ics ha we e used o de i e IVs o FAs a e de ailed in Table 2. A educed isk o CRC was obse ed o gene ic a ian s associa ed wi h inc eases in he MUFAs s udied (Table 3). In all bu one o he se en coho s inc eased le els o OA we e associa ed wi h educed CRC isk (Fig. 1). In he me a-analysis o hese se en coho s he OR OA was 0.77 (95% CI: 0.65e0.92, PZ3.9 10 3 ) wi h li le e idence o be ween-s udy he e ogenei y (P he Z0.23, I 2 Z26%). Simila ly, inc eased le els o POA we e associa ed wi h educed CRC isk wi h an OR POA o 0.36 (95% CI: 0.15e0.84, PZ0.018, P he Z0.08, I 2 Z47%; Fig. 1). The u-6 PUFAs LA and AA bo h showed associa- ion wi h CRC isk, bu in di e en di ec ions. Speci - ically, LA was associa ed wi h educed isk (OR LA Z0.95, 95% CI: 0.93e0.98, PZ3.7 10 4 , P he Z0.03, I 2 Z57%; Fig. 1) and AA wi h an inc eased isk (OR AA Z1.05, 95% CI: 1.02e1.07, PZ1.7 10 4 , P he Z0.03, I 2 Z56%). The associa ion be ween one s anda d de ia ion inc ease in each o he o he PUFAs de ined by hei espec i e IVs and CRC isk we e null (Supplemen a y Fig. 1). O he h ee SFAs s udied, inc eased SA was nomi- nally associa ed wi h CRC isk (OR SA Z1.17, 95% CI: 1.01e1.35, PZ0.041, P he Z0.04, I 2 Z55%). To o mally assess he impac o he e ogenei y on s udy indings we de i ed ORs unde a andom-e ec s Table 3 Odds a ios (ORs) and 95% con idence in e als (CI) o one s anda d de ia ion inc ease in gene ically p edic ed plasma a y acid le els and colo ec al cance isk. Fa y acid Signi ican associa ions OR ( ixed e ec s) 95% CI ( ixed e ec s) P- alue ( ixed e ec s) OR ( andom e ec s) 95% CI ( andom e ec s) P- alue ( andom e ec s) I 2 P he A achidic acid 0.92 0.61e1.39 0.7 0.93 0.61e1.40 0.71 3% 0.41 Palmi ic acid (PA) 0.97 0.78e1.21 0.82 0.97 0.78e1.21 0.82 0% 0.47 S ea ic acid (SA) 1.16 1.01e1.35 0.04 1.2 0.95e1.49 0.12 55% 0.04 Docosahexaenoic acid (DHA) 1.32 0.94e1.87 0.11 1.32 0.94e1.87 0.11 0% 0.65 Docosapen aenoic acid (DPA) 1.58 0.99e2.52 0.06 1.63 0.97e2.73 0.06 17% 0.3 Eicosapen aenoic acid (EPA) 0.39 0.13e1.21 0.1 0.39 0.13e1.21 0.1 0% 0.57 A achidonic acid (AA) 1.05 1.02e1.07 1.7 10 4 1.05 1.02e1.09 4.9 10 3 56% 0.03 Dihomo-g-linolenic acid (DGLA) 0.91 0.83e1.00 0.06 0.95 0.80e1.01 0.07 23% 0.26 Linoleic acid (LA) 0.95 0.93e0.98 3.7 10 4 0.95 0.91e0.99 8.9 10 3 57% 0.03 Oleic acid (OA) 0.77 0.65e0.92 3.9 10 3 0.76 0.62e0.94 9.7 10 3 26% 0.23 Palmi oleic acid (POA) 0.36 0.15e0.84 0.018 0.32 0.10e1.07 0.06 47% 0.08 P he ,P- alue o he e ogenei y; I 2 , p opo ion o he o al a ia ion due o he e ogenei y; SFA, sa u a ed a y acid; PUFA, polyunsa u a ed a y acid; MUFA, monounsa u a ed a y acid. S. May-Wilson e al. / Eu opean Jou nal o Cance 84 (2017) 228e238232 model. Associa ions be ween AA, LA and OA and CRC isk emained signi ican (Table 3). We assessed he impac o possible classical pleio- opism on MR es ima es using bo h IVW and MR- Egge eg ession es s. The e was no e idence o iola ion o he s anda d IV assump ions used o MR analysis, such as a dependence on con ounde s (Table 4). In he p esen analysis, we used he SNP s102275 in combina ion wi h o he SNPs o gene a e a polygenic isk sco e o SA, OA and POA, whe eas s174547, which is in LD wi h s102275 ( 2 Z1.0 and D0Z1.0), was used o DPA, AA, DGLA and LA. Bo h SNPs anno a e he FADS2 gene. FADS2 is a a e-limi ing enzyme in he desa u a ion o LA o AA, and a-lino- lenic acid in o DHA and EPA (Fig. 2). These FAs a e p ecu so s o p os aglandins and leuko ienes, which a e key media o s o he in lamma o y esponse. In an eQTL analysis s174547 and s102275 geno ype we e shown o be s ongly co ela ed wi h FADS2 exp es- sion ac oss a ange o di e en issue ypes, including blood (PZ3.98 10 29 ), no mal colon (PZ1.65 10 10 )andCRC(PZ2.07 10 5 ) (Supplemen a y Table 1). Fig. 1. Me a-analysis odds a ios (OR) o colo ec al cance pe uni inc ease in gene ic isk sco e (s anda d de ia ion o ai ) o sig- ni ican a y acid associa ions. (a) Oleic acid; (b) a achidonic acid; (c) s ea ic acid; (d) linoleic acid; (e) palmi oleic acid; I 2 : p opo ion o he o al a ia ion due o he e ogenei y. Boxes: OR poin es ima e; i s a ea is p opo ional o he weigh o he s udy. Diamond: o e all summa y es ima e, wi h con idence in e als gi en by i s wid h. Ve ical line: null alue (OR Z1.0). S. May-Wilson e al. / Eu opean Jou nal o Cance 84 (2017) 228e238 233 4. Discussion While die a y a in ake has been associa ed wi h he CRC isk, easing ou speci ic FA associa ions and hei mechanis ic basis has p o en o be challenging. A numbe o obse a ional s udies ha e epo ed associa- ions be ween se um le els o speci ic FAs wi h CRC [28,29], suppo ing ou indings. A majo s eng h o he MR s a egy o iden i y causal associa ions is ha i is no in luenced by ecall bias and con ounding ha can a ec adi ional obse - a ional s udies. Ne e heless, a key assump ion in MR is ha he a ian s used o gene a e gene ic sco es a e associa ed wi h he exposu e being que ied. He ein, we only made use o SNPs associa ed wi h each FA a genome-wide signi icance om hypo hesis- ee GWAS. Fu he mo e, we only used da a om indi iduals o Eu opean descen so as o limi bias om popula ion s a i ica ion. Ano he cen al assump ion in MR is ha a ian s a e associa ed wi h CRC only h ough he exposu e and a e no con ounded by pleio opy, which would be e ealed by a posi i e co ela ion be ween inc easing e ec sizes in he IVs and CRC isk. While we did no obse e such ela ionship, we acknowledge ha IVs o a numbe o he FAs we e solely based on only one o wo SNPs, p e en ing assessmen by IVW and MR-Egge analysis. One s a egy o o e come his and ully in es iga e any pleio opy would be o measu e FA se um le els in co ela ion wi h CRC isk. In his analysis, he same SNP ( s102275, o co e- la ed SNP s174547) was used o make causal de- duc ions be ween mul iple FAs and CRC isk. The e o e, SNPs ha e been used each ime assuming ha he exposu e indi idually accoun s o he disease associa ion. The gene ic a ian associa ion wi h CRC isk is consequen ly double-coun ed, in ha he e ec is a ibu ed o di e en FA exposu es [30]. Wi h such e ical pleio opism, single locus MR analyses canno obus ly deciphe which FA is p ima ily d i ing he ela ionship wi h CRC isk. Such conside a ions ha e no been add essed in p e ious s udies o he ela ion- ship be ween PUFAs and p os a e cance [31] o be- ween b anched-chain amino acids and diabe es [32]. While we did no demons a e a causal associa ion be ween o he FAs including se e al PUFAs, SFAs and CRC isk, we acknowledge ha ou powe o demon- s a e a ela ionship was limi ed. Fo example, wi h espec o EPA: assuming he a iance explained by he alleles is 0.04%, based on epidemiological obse a ional s udy da a, and a ela i e isk o 1.04 we had <10% powe o demons a e a ela ionship [33]. Accep ing hese ca ea s we ha e p o ided suppo o di e ing e ec s o OA, and u-6 PUFAs LA and AA on CRC isk. Ou indings b oadly acco d wi h he indings om many o he published ecological and epidemio- logical obse a ional s udies. No ably, inc eased le els o AA con ibu e as a isk ac o o CRC de elopmen [34,35], while inc eased in ake o oli e oil, which is high in OA, is associa ed wi h dec eased isk [36]. A numbe o epidemiological s udies ha e p o ided e idence ha a Medi e anean die , wi h a highe oli e oil in ake, is associa ed wi h educed CRC isk [36e38]. In he eQTL analysis, bo h s102275 and s174547 show e idence o cis- egula o y e ec s on FADS2 exp ession. In iguingly, s174547 has p e iously been epo ed o ha e opposing e ec s on FADS2 and FADS1 exp ession in CRC [39]. Collec i ely, hese da a p o ide o ela ionship be ween die , geno ype, FA me abolism and CRC isk h ough modula ion o an in lamma o y esponse. E en so, a biological basis o associa ions be ween speci ic FAs and CRC isk emain o be es ablished. I is howe e , p edic ed a p io i ha wi hin any FA class, di e en membe s ha e di e en ac ions and e ec s. Wi h Table 4 IVW and MR-Egge es esul s o combined a y acid ins umen al a iables. Fa y acid sub ype Fa y acid IVW MR-Egge Slope Es ima e (95% CI) P- alue Es ima e (95% CI) P- alue SFA S ea ic acid (SA) 0.1 (0.33 o 0.64) 0.30 In e cep 0.68 (4.79 o 3.43) 0.28 Slope 4.10 (19.86 o 28.06) 0.27 u-3 PUFA Docosapen aenoic acid (DPA) 0.46 (2.32 o 3.23) 0.55 In e cep 0.09 (0.56 o 0.39) 0.26 Slope 2.01 (7.9 o 11.61) 0.23 Eicosapen aenoic acid (EPA) 0.59 (7.99 o 9.16) 0.54 In e cep 0.11 (N/A) e Slope 2.2 (N/A) e u-6 PUFA A achidonic acid (AA) 0.04 (0.2 o 0.33) 0.29 In e cep 0.04 (N/A) e Slope 0.02 (N/A) e Dihomo-g-linolenic acid (DGLA) 0.09 (2.48 o 2.29) 0.70 In e cep 0.25 (N/A) e Slope 0.90 (N/A) e Linoleic acid (LA) 0.05 (0.17 o 0.07) 0.22 In e cep 0.02 (0.64 o 0.67) 0.77 Slope 0.07 (0.81 o 0.68) 0.46 MUFA Palmi oleic acid (POA) 1.03 (2.64 o 0.58) 0.15 In e cep 0.11 (0.27 o 0.05) 0.12 Slope 3.13 (3.16 o 9.41) 0.21 CI, con idence in e al; MUFA, monounsa u a ed a y acid; PUFA, polyunsa u a ed a y acid; SFA, sa u a ed a y acid; IVW, in e se a ian weigh ed. *FA ai s wi h wo IVs, p e en ing calcula ion o CIs and P- alue. S. May-Wilson e al. / Eu opean Jou nal o Cance 84 (2017) 228e238234 espec o u-6, e idence suppo s he in lamma o y e ec s o AA h ough COX-2 p oduc ion o in lamma o y me- dia o s [40] including p os aglandin E2, which a ec CRC ca cinogenesis [41e43]. This implies ha die s high in AA, such as mea o eggs, may lead o mo e in lamma o y compounds, which in u n may inc ease CRC isk. While inc easing die a y LA, an essen ial FA, migh po en ially en ich issues wi h AA due o hei me abolic link [44],a geneeen i onmen in e ac ion may exis o in luence colon FA con en [45]. The e is howe e , con adic o y e idence om s udies ha ha e associa ed LA wi h bo h an inc eased [46] and dec eased isk o CRC, possibly by al e ing u-6 o u-3 FA a ios [47] o al e na i ely p o- duc ion o eac i e oxygen species [48]. The abili y o aspi in o i e e sibly inhibi COX-1 and COX-2 and he e o e lowe p o-in lamma o y signals independen o geno ype and die , has hus p o ed an a ac i e op ion o CRC chemop e en ion [49]. Fig. 2. Pa hway o a y acids. Shown a e he a ious a y acids analysed, and he enzymes in ol ed in hei me abolism. COX: cyclo- oxygenase, LOX: 5-lipoxygenase. S. May-Wilson e al. / Eu opean Jou nal o Cance 84 (2017) 228e238 235 In conclusion, i espec i e o he biological basis o associa ions be ween FAs and CRC isk ou indings a e consis en wi h he obse a ion ha he die a y composi ion o MUFAs in Medi e anean die s a e isk educing, and ha a p o-in lamma o y die a e isk inc easing [50]. While we may no be a a s age whe e we can jus i iably ad ise indi iduals o al e hei in ake o speci ic FAs o dec ease he isk o de eloping CRC, i seems he cu en guidelines o mode a e o al a and SFA consump ion and inc ease unsa u a ed FA in ake is likely o be bene icial. Con lic o in e es s a emen None decla ed. Acknowledgemen s A he Ins i u e o Cance Resea ch, his wo k was suppo ed by Cance Resea ch UK (C1298/A8362 - Bobby Moo e Fund o Cance Resea ch UK). Addi- ional suppo was p o ided by he Na ional Cance Resea ch Ne wo k. S. M-W was in eceip o a PhD s uden ship om The Ins i u e o Cance Resea ch. A.S. is suppo ed by a clinical ellowship om Cance Resea ch UK. In Edinbu gh he wo k was suppo ed by P og amme G an unding om Cance Resea ch UK (C348/A12076). In Ox o d addi ional unding was p o- ided by he Ox o d Comp ehensi e Biomedical Resea ch Cen e and he EU FP7 CHIBCHA g an . Co e in as uc u e suppo o he Wellcome T us Cen e o Human Gene ics, Ox o d was p o ided by g an (090532/Z/09/Z). We a e g a e ul o many col- leagues wi hin UK Clinical Gene ics Depa men s ( o CORGI) and o many collabo a o s who pa icipa ed in he VICTOR and QUASAR2 ials. We also hank colleagues om he UK Na ional Cance Resea ch Ne wo k ( o NSCCG). Suppo om he Eu opean Union (FP7/207-2013, g an 258236) and FP7 collabo- a i e p ojec SYSCOL and COST Ac ion in he UK is also acknowledged (BM1206). The COIN and COIN-B ials we e unded by Cance Resea ch UK and he Medical Resea ch Council and we e conduc ed wi h he suppo o he Na ional Ins i u e o Heal h Resea ch Cance Resea ch Ne wo k. COIN and COIN-B ans- la ional s udies we e suppo ed by he Bobby Moo e Fund om Cance Resea ch UK, Teno us, he Kidani T us , Cance Resea ch Wales and he Na ional Ins i- u e o Social Ca e and Heal h Resea ch Cance Ge- ne ics Biomedical Resea ch Uni (2011e2014). N.A.A., B.F.M. and S.M.W. we e unded and suppo ed by KFSHRC. In Finland, his wo k was suppo ed by g an s om he Academy o Finland (Finnish Cen e o Excellence P og am 2012e2017, 250345), he Jane and Aa os E kko Founda ion, he Finnish Cance Socie y (K.P.), he Eu opean Resea ch Council [ERC; 268648], he Sig id Juselius Founda ion, SYSCOL, he No dic In o ma ion o Ac ion eScience Cen e (NIASC), he No dic Cen e o Excellence inanced by No dFo sk (p ojec 62721, K.P.) and S a e Resea ch Funding o Kuopio Uni e si y Hospi al (B1401). We acknowledge he compu a ional esou ces p o ided by he ELIXIR node, hos ed a he CSCeIT Cen e o Science, Finland, and unded by he Academy o Finland (g an s 271642 and 263164), he Minis y o Educa ion and Cul u e, Finland. V.S. was suppo ed by he Finnish Academy (g an numbe 139635). Sample collec ion and geno yping in he Finnish Twin Coho has been sup- po ed by he Wellcome T us Sange Ins i u e, ENGAGE eEu opean Ne wo k o Gene ic and Genomic Epidemiology, FP7-HEALTH-F4-2007 (201413), he Na ional Ins i u e o Alcohol Abuse and Alcoholism [g an s AA-12502 and AA-00145 o Richa d J Rose and K02AA018755 o Danielle M Dick] and he Academy o Finland (100499, 205585, 265240 and 263278 o J.K.). The wo k o he Colon Cance Family Regis y (CCFR) was suppo ed by om he Na ional Cance Ins i u e (UM1 CA167551), Na ional Ins i u es o Heal h and h ough coope a i e ag eemen s wi h he ollowing CCFR cen e s: Aus alasian Colo ec al Can- ce Family Regis y (U01 CA074778, U01/U24 CA097735), USC Conso ium Colo ec al Cance Fam- ily Regis y (U01/U24 CA074799), Mayo Clinic Coop- e a i e Familial Regis y o Colon Cance S udies (U01/U24 CA074800), On a io Familial Colo ec al Cance Regis y (U01/U24 CA074783), Sea le Colo- ec al Cance Family Regis y (U01/U24 CA074794), and Uni e si y o Hawaii Colo ec al Cance Family Regis y (U01/U24 CA074806). The CCFR Illumina GWAS was suppo ed by unding om he Na ional Cance Ins i u e, Na ional Ins i u es o Heal h (U01 CA122839 and R01 CA143237 o G.C.). Addi ional suppo was p o ided om he Su eillance, Epidemi- ology and End Resul s (SEER) P og am o he Na ional Cance Ins i u e o F ed Hu chinson Cance Resea ch Cen e (Con ol Nos. N01-CN-67009, N01-PC-35142 and Con ac No. HHSN2612013000121), he Hawai’i Depa men o Heal h (Con ol Nos. N01-PC-67001, N01-PC-35137, and Con ac No. HHSN26120100037C), and he Cali o nia Depa men o Public Heal h (con ac HHSN261201000035C) awa ded o he Uni e si y o Sou he n Cali o nia. The con en o his manusc ip does no necessa ily e lec he iews o policies o he Na ional Cance Ins i u e, any SEER p og am o any o he collabo a ing cen es in he CCFR, no does men ion o ade names, com- me cial p oduc s, o o ganisa ions imply endo semen by he US Go e nmen , SEER o he CCFR. We a e g a e ul o all indi iduals who pa icipa ed in he a ious s udies. This s udy made use o geno yping da a om he 1958 Bi h Coho , kindly made a ailable by he Wellcome T us Case Con ol Conso ium 2. A ull lis o he in es iga o s who con ibu ed o he S. May-Wilson e al. / Eu opean Jou nal o Cance 84 (2017) 228e238236