{"paper_id":"bc1117db-2358-4575-ba4a-53a2815c33d0","body_text":"1 \n \nLowering of circulating sclerostin may increase risk of atherosclerosis and its \nrisk factors: evidence from a genome-wide association meta-analysis followed \nby Mendelian randomization \n \nJie Zheng1,2,3†, Eleanor Wheeler 4, Maik Pietzner 4,5, Till Andlauer 6, Michelle Yau 7, April E. Hartley 3, Ben \nMichael Brumpton8,9, Humaira Rasheed 3,9,10, John P Kemp 3,11,12, Monika Frysz 3,13, Jamie Robinson 3, Sjur \nReppe14,15,16, Vid Prijatel 17, Kaare M Gautvik 14, Louise Falk 3, Winfried Maerz 18,19,20, Ingrid Gergei 20,21, \nPatricia A Peyser 22, Maryam Kavousi 23, Paul S. de Vries 24, Clint L. Miller 25, Maxime Bos 23, Sander W. \nvan der Laan 26, Rajeev Malhotra 27, Markus Herrmann 18, Hubert Scharnagl 18, Marcus Kleber 19, George \nDedoussis28, Eleftheria Zeggini 29,30, Maria Nethander 31, Claes Ohlsson 31, Mattias Lorentzon 32, Nick \nWareham4, Claudia Langenberg 4,5, Michael V. Holmes 3,33,34,35, George Davey Smith 3†, Jonathan H. \nTobias3,13† \n \n1Department of Endocrine and Metabolic Diseases, Shanghai Institute of Endocrine and Metabolic Diseases, Ruijin Hospital, \nShanghai Jiao Tong University School of Medicine, Shanghai, China \n2Shanghai National Clinical Research Center for metabolic Diseases, Key Laboratory for Endocrine and Metabolic Diseases \nof the National Health Commission of the PR China, Shanghai Key Laboratory for Endocrine Tumor,State Key Laboratory \nof Medical Genomics,Ruijin Hospital, Shanghai Jiao Tong University School of Medicine, Shanghai, China \n3MRC Integrative Epidemiology Unit (IEU), Bristol Medical School, University of Bristol, Oakfield House, Oakfield Grove, \nBristol, BS8 2BN, United Kingdom \n4MRC Epidemiology Unit, Institute of Metabolic Science, University of Cambridge School of Clinical Medicine, Cambridge \nCB2 0QQ, UK. \n5Computational Medicine, Berlin Institute of Health at Charité – Universitätsmedizin Berlin, Germany \n6Department of Neurology, Klinikum rechts der Isar, School of Medicine, Technical University of Munich, Munich, Germany \n7 Marcus Institute for Aging Research, Hebrew SeniorLife, Harvard Medical School, Boston, MA, USA \n8K.G. Jebsen Center for Genetic Epidemiology, Department of Public Health and Nursing, NTNU, Norwegian University of \nScience and Technology, Trondheim, Norway \n9HUNT Research Centre, Department of Public Health and Nursing, NTNU, Norwegian University of Science \nand Technology, Levanger, 7600 Norway \n10Division of Medicine and Laboratory Sciences, Faculty of Medicine, University of Oslo, Norway \n11Institute for Molecular Bioscience, University of Queensland, Brisbane, Australia \n12The University of Queensland Diamantina Institute, The University of Queensland, Brisbane, Queensland, Australia \n13Musculoskeletal Research Unit, University of Bristol, Level 1 Learning and Research Building, Bristol, BS10 5NB, United \nKingdom \n14Unger-Vetlesen Institute, Lovisenberg Diaconal Hospital, Oslo, Norway \n15 Department of Plastic and Reconstructive Surgery, Oslo University Hospital, Oslo, Norway   \n16Department of Medical Biochemistry, Oslo University Hospital, Oslo, Norway \n17Department of Internal Medicine, Erasmus MC University Medical Center, Rotterdam, The Netherlands \n18Clinical Institute of Medical and Chemical Laboratory Diagnostics, Medical University of Graz, Austria \n19SYNLAB Academy, SYNLAB Holding Deutschland GmbH, Mannheim, Germany \n . CC-BY 4.0 International licenseIt is made available under a \n is the author/funder, who has granted medRxiv a license to display the preprint in perpetuity. (which was not certified by peer review)\nThe copyright holder for this preprint this version posted June 16, 2022. ; https://doi.org/10.1101/2022.06.13.22275915doi: medRxiv preprint \nNOTE: This preprint reports new research that has not been certified by peer review and should not be used to guide clinical practice.\n\n \n2 \n \n20Vth Department of Medicine (Nephrology, Hypertensiology, Rheumatology, Endocrinology, Diabetology), Medical \nFaculty Mannheim, University of Heidelberg, Mannheim, Germany \n21Therapeutic Area Cardiovascular Medicine, Boehringer Ingelheim International GmbH, Ingelheim, Germany \n22Department of Epidemiology, School of Public Health, University of Michigan, Ann Arbor, MI, 48109, United States \n23Department of Epidemiology, Erasmus MC, University Medical Center, Rotterdam, Wytemaweg 90, 3015 CN, Rotterdam, \nThe Netherlands \n24Human Genetics Center, Department of Epidemiology, Human Genetics, and Environmental Sciences, School of Public \nHealth, the University of Texas Health Science Center at Houston, Houston, TX, USA \n25Center for Public Health Genomics, Department of Public Health Sciences, University of Virginia, Charlottesville, \nVirginia, USA \n26Central Diagnostics Laboratory, Division Laboratories, Pharmacy, and Biomedical genetics, University Medical Center \nUtrecht, Utrecht University, Utrecht, the Netherlands \n27Cardiology Division, Dept of Medicine, Massachusetts General Hospital, Boston, MA \n28Department of Nutrition and Dietetics, School of Health Science and Education, Harokopio University, El. Venizelou 70, \n17671, Athens, Greece \n29Institute of Translational Genomics, Helmholtz Zentrum München, German Research Center for Environmental Health, \nNeuherberg, Germany \n30Technical University of Munich (TUM) and Klinikum Rechts der Isar, TUM School of Medicine, Munich, Germany \n31Sahlgrenska Osteoporosis Centre, Centre for Bone and Arthritis Research, Department of Internal Medicine and Clinical \nNutrition, Institute of Medicine, University of Gothenburg, Sweden \n32Institute of Medicine, Sahlgrenska Academy, Sahlgrenska University Hospital, Mölndal, 431 80 Mölndal, Sweden \n33Medical Research Council Population Health Research Unit, University of Oxford, Oxford, UK \n34Clinical Trial Service Unit & Epidemiological Studies Unit, Nuffield Department of Population Health, University of \nOxford, Oxford, UK \n35National Institute for Health Research, Oxford Biomedical Research Centre, Oxford University Hospital, Oxford, UK \n \n†Correspondence to:  \nJie Zheng, professor in Aetiological Epidemiology, Shanghai National Clinical Research Center for \nEndocrine and Metabolic Diseases, Key Laboratory for Endocrine and Metabolic Diseases of the National \nHealth Commission of the PR China, Shanghai Institute of Endocrine and Metabolic Diseases, Department \nof Endocrine and Metabolic Diseases, Ruijin Hospital, Shanghai Jiao Tong University School of Medicine, \nShanghai, China; and MRC Integrative Epidemiology Unit (IEU), Bristol Medical School, University of \nBristol, Oakfield House, Oakfield Grove, Bristol, BS8 2BN, United Kingdom. E-mail: \nJie.Zheng@bristol.ac.uk\n \n \nGeorge Davey Smith, professor in Clinical Epidemiology, MRC Integrative Epidemiology Unit (IEU), \nBristol Medical School, University of Bristol, Oakfield House, Oakfield Grove, Bristol, BS8 2BN, United \nKingdom. E-mail: kz.davey-smith@bristol.ac.uk \n \nJonathan H. Tobias, professor in Rheumatology, MRC Integrative Epidemiology Unit (IEU), Bristol \nMedical School, University of Bristol, Oakfield House, Oakfield Grove, Bristol, BS8 2BN, United \nKingdom; and Musculoskeletal Research Unit, University of Bristol, Level 1 Learning and Research \nBuilding, Bristol, BS10 5NB, United Kingdom. E-mail: jon.tobias@bristol.ac.uk\n  \n . CC-BY 4.0 International licenseIt is made available under a \n is the author/funder, who has granted medRxiv a license to display the preprint in perpetuity. (which was not certified by peer review)\nThe copyright holder for this preprint this version posted June 16, 2022. ; https://doi.org/10.1101/2022.06.13.22275915doi: medRxiv preprint \n\n \n3 \n \nABSTRACT \nSclerostin inhibition is a new therapeutic approach for increasing bone mineral density (BMD) \nbut its cardiovascular safety is unclear. We conducted a genome-wide association study \n(GWAS) meta-analysis of circulating sclerostin in 33,961 Europeans followed by Mendelian \nrandomization (MR) to estimate the causal effects of sclerostin on 15 atherosclerosis-related \ndiseases and risk factors. GWAS meta-analysis identified 18 variants independently \nassociated with sclerostin, which including a novel cis signal in the SOST  region and three \ntrans signals in B4GALNT3, RIN3 and SERPINA1 regions that were associated with \nopposite effects on circulating sclerostin and eBMD. MR combining these four SNPs \nsuggested lower sclerostin increased hypertension risk (odds ratio [OR]=1.09, 95%CI=1.04 \nto 1.15), whereas bi-directional analyses revealed little evidence for an effect of genetic \nliability to hypertension on sclerostin levels. MR restricted to cis (SOST) SNPs additionally \nsuggested sclerostin inhibition increased risk of type 2 diabetes (T2DM) (OR=1.26; \n95%CI=1.08 to 1.48) and myocardial infarction (MI) (OR=1.31, 95% CI=1.183 to 1.45). \nFurthermore, these analyses suggested sclerostin inhibition increased coronary artery \ncalcification (CAC) (β=0.74, 95%CI=0.33 to 1.15), levels of apoB ( β=0.07; 95%CI=0.04 to \n0.10; this result was driven by rs4793023) and triglycerides ( β=0.18; 95%CI=0.13 to 0.24), \nand reduced HDL-C (β=-0.14; 95%CI=-0.17 to -0.10). This study provides genetic evidence \nto support a causal effect of sclerostin inhibition on increased hypertension risk. Cis-only \nanalyses suggested that sclerostin inhibition additionally increases the risk of T2DM, MI, \nCAC, and an atherogenic lipid profile. Together, our findings reinforce the requirement for \nstrategies to mitigate against adverse effects of sclerostin inhibitors like romosozumab on \natherosclerosis and its related risk factors.   \n . CC-BY 4.0 International licenseIt is made available under a \n is the author/funder, who has granted medRxiv a license to display the preprint in perpetuity. (which was not certified by peer review)\nThe copyright holder for this preprint this version posted June 16, 2022. ; https://doi.org/10.1101/2022.06.13.22275915doi: medRxiv preprint \n\n \n4 \n \nInhibition of sclerostin is a therapeutic approach to increasing bone mineral density (BMD) \nand lowering fracture risk in patients with osteoporosis. However, two phase III trials of \nromosozumab, a first-in-class monoclonal antibody that inhibits sclerostin, reported higher \nnumbers of cardiovascular serious adverse events in the romosozumab treated group over its \ncomparator(1)(2). However, a similar imbalance of cardiovascular disease (CVD) was not \nseen in another study comparing romosozumab to placebo(3). Possibly, these different results \nreflect a beneficial effect of bisphosphonate treatment on risk of CVD. Another \nbisphosphonate, zoledronate, has been found to decrease all-cause mortality, to which \nreduced CVD mortality may contribute(4). However, a beneficial effect on mortality was not \nborne out in a meta-analysis of drug trials of zoledronate and other bisphosphonates (5). The \nrole of sclerostin in the vasculature is unknown, though some studies have shown that its \ninhibition may promote vascular calcification, which would increase the risk of CVD (6). \nGiven these concerns of CVD safety, marketing authorization for romosozumab indicates \nprevious myocardial infarction (MI) or stroke as contraindications, underlying the urgent \nneed to understand the causal role of sclerostin inhibition on CVD outcomes, so that further \nsteps can be taken to mitigate these potential adverse effects. \n \nContrary to trial evidence suggesting an increase in CVD risk following sclerostin inhibition, \nour recent observational study found that sclerostin levels are positively associated with CVD \nseverity and mortality, partly explained by a relationship between higher sclerostin levels and \nmajor CVD risk factors(7). Equivalent findings apply to BMD, with sclerostin levels found to \nbe positively related to BMD (8), despite trial evidence suggesting that sclerostin lowering \nincreases BMD(1)(2). Such discrepancies may reflect the influence of confounders or reverse \ncausality on findings from observational studies (9). Mendelian randomization (MR) uses \ngenetic variants as  proxies for an exposure to estimate the causal effect of a modifiable risk \nfactor on a disease(10)(11), in order to avoid bias from confounders or reverse causa lity. For \nexample, in a recent MR study using BMD-associated variants in the SOST region as a proxy \nfor sclerostin inhibition,  Bovijn et al. found genetic evidence consistent with a potential \nadverse effect of sclerostin inhibition on CVD-related events (12). However, as discussed in \nthe recent European Medicines Agency report on Romosozumab (13), this study has some \nweaknesses. For example, the SOST single nucleotide polymorphisms (SNPs) used in the \nanalysis by Bovijn et al. are >30kb downstream of the target gene. Another MR study using \nsclerostin gene expression in arterial and heart tissue as the exposure was interpreted as \nshowing no causal effect of sclerostin expression on risk of MI or stroke(14).  \n . CC-BY 4.0 International licenseIt is made available under a \n is the author/funder, who has granted medRxiv a license to display the preprint in perpetuity. (which was not certified by peer review)\nThe copyright holder for this preprint this version posted June 16, 2022. ; https://doi.org/10.1101/2022.06.13.22275915doi: medRxiv preprint \n\n \n5 \n \n \nAn alternative approach to instrument selection is to use SNPs identified from a well-\npowered genome-wide association study (GWAS) of circulating sclerostin. In an earlier \nGWAS of sclerostin levels, we identified three trans-acting genetic variants associated with \nsclerostin, including a top variant in the B4GALNT3 region. However, we only observed \nmarginal genetic associations in the cis-SOST region and had limited power to examine \ncausal relationships with extra-skeletal phenotypes (15). Therefore, a GWAS of circulating \nsclerostin including more participants is needed to identify more reliable genetic predictors, \nincluding in the cis -acting region. A further consideration is that a bidirectional causal \npathway appears to exist between sclerostin and BMD, whereby increased sclerostin levels \ncause a decrease in BMD, whereas higher BMD increases sclerostin levels, possibly \nreflecting a feedback pathway (15). Therefore, findings from a sclerostin GWAS are \npotentially subject to mis-specification of the primary phenotype (16)(17) , with genetic \nsignals being detected which are primarily related to BMD rather than sclerostin.   \n \nThe goal of the present study was to examine potential safety concerns of sclerostin \ninhibition on atherosclerosis and its risk factors using an MR approach, based on a set of \ninstruments derived from an updated GWAS meta-analysis of circulating sclerostin, and \nusing 15 atherosclerosis-related diseases and risk factors with available GWAS data, \nincluding outcomes such as coronary artery calcification (CAC), type 2 diabetes (T2DM) and \nlipid/lipoprotein risk factors, which is of interest but understudied. To enable sufficient power \nto examine causal effects on extra-skeletal phenotypes, we aimed to identify genetic \npredictors of sclerostin with good instrument strength, incorporating both cis- and trans -\nacting variants, having assembled a sample over three times the size of our previous \nstudy(15). Given cis-acting variants are more likely to be specific for the drug target under \ninvestigation(18), we also aimed to include sensitivity analyses in which MR analyses were \nrestricted to cis-acting variants.  \n  \n . CC-BY 4.0 International licenseIt is made available under a \n is the author/funder, who has granted medRxiv a license to display the preprint in perpetuity. (which was not certified by peer review)\nThe copyright holder for this preprint this version posted June 16, 2022. ; https://doi.org/10.1101/2022.06.13.22275915doi: medRxiv preprint \n\n \n6 \n \nRESULTS \nSummary of study design \nFigure 1 illustrates the design and participants of this study. We aimed to understand the \ngenetic architectural of sclerostin and the causal role of sclerostin inhibition on \natherosclerosis related diseases and risk factors. First, we conducted a GWAS meta-analysis \nand post-GWAS follow-up analyses of circulating sclerostin in 33,961 European individuals, \nwhich including nine cohorts: the Avon Longitudinal Study of Parents and Children \n(ALSPAC), Die Deutsche Diabetes Dialyse Study (4D), The Gothenburg Osteoporosis and \nObesity Determinants (GOOD), and the MANOLIS cohort, Fenland (19), INTERVAL(20), \nTrøndelag health study (HUNT) (21)(22)(23), the Osteoarthritis Initiative \n(OAI)(24)(25)(26)(27), the Ludwigshafen Risk and Cardiovascular Health (LURIC)(28)(29) . \nThe cohort details were included in Supplementary Note 1 . Second, we conducted MR \nanalyses of circulating sclerostin using genetic instruments from both cis and trans  regions \n(Supplementary Table 1 ) as well as from cis  region only ( Supplementary Table 2 ). The \noutcomes are 15 atherosclerosis related diseases and risk factors (Supplementary Table 3 ). \nThe bi-directional MR was further conducted for the 15 atherosclerosis related diseases and \nrisk factors (Supplementary Table 4) on sclerostin.  \n \nGenome-wide association signals of circulating sclerostin  \nGWAS results of circulating sclerostin were available in 33,961 European ancestry \nparticipants from a meta-analysis of nine cohorts (Table 1). Supplementary Figures 1 and 2 \nshow the Manhattan and QQ plots of association results from the fixed-effects meta-analysis \nof sclerostin, respectively. There was little evidence of inflation of the test statistics (genomic \ninflation factor \nλ =1.082; LD score regression intercept =1.023). Therefore, no genomic \ncontrol correction was applied to the meta-analysis results. Single trait LD score regression \nresults showed that common variants included in the GWAS meta-analysis explained 15.4% \nof the phenotypic variance of circulating sclerostin (SNP-based heritability h\n2=0.154, \nSE=0.021, P=3.01×10-13). \n \nIn total, 997 genetic variants were identified to be associated with circulating sclerostin at \ngenome-wide significance. After applying conditional analysis, 18 conditionally independent \nvariants within 15 genomic loci were associated with circulating sclerostin ( Table 2 ). The \nstrongest signal, rs215223, was close to the B4GALNT3 gene ( β=-0.136, SD change in \n . CC-BY 4.0 International licenseIt is made available under a \n is the author/funder, who has granted medRxiv a license to display the preprint in perpetuity. (which was not certified by peer review)\nThe copyright holder for this preprint this version posted June 16, 2022. ; https://doi.org/10.1101/2022.06.13.22275915doi: medRxiv preprint \n\n \n7 \n \ncirculating sclerostin per A allele, SE=0.008, P=2.44×10 -73, effect allele frequency=0.405, \nvariance explained=0.89%); the SNP is in perfect LD with the top signal reported in our \nprevious sclerostin GWAS, rs215226 (15) (Figure 2A ). One cis-acting variant in the SOST \nregion, rs66838809, showed a strong association with sclerostin ( β=-0.088, SD change in \ncirculating sclerostin per A allele, SE=0.015, P=1.45×10 -9, effect allele frequency=0.079, \nvariance explained=0.11%; Figure 2B). Another variant, rs28929474, in the SERPINA1 gene \nregion, was associated with circulating sclerostin ( β=0.173, SD change in circulating \nsclerostin per T allele, SE=0.027, P=1.1×10 -10, effect allele frequency=0.021, variance \nexplained=0.12%; Figure 2C). This missense rare variant constitutes the PiZ  allele, causing \nalpha-1 anti-trypsin (α1AT) deficiency in homozygous cases(30). The variant, rs7143806, in \nthe RIN3 gene region, was also associated with sclerostin (β=0.053, SD change in circulating \nsclerostin per A allele, SE=0.010, P=3.35×10 -8, effect allele frequency=0.181, variance \nexplained by this variant=0.08%; Figure 2D). The gene was reported to be associated with \nlower limb BMD (31). The other 12 genomic loci were FAF1  (rs61781020), PID1  \n(rs4973180), MAP3K1 (rs11960484), LVRN (rs34498262 and rs17138656), SUPT3H \n(rs75523462), LINC00326 (rs34366581), TNFRSF11B (rs11995824), TNFSF11  (rs9594738, \nrs34136735 and rs665632), TNFRSF11A (rs2957124), JAG1 (rs13042961) and rs6585816 in \nchromosome 10 (no nearby genes). These include SNPs related to TNFSF11 and TNFSF11A, \ntwo well established BMD loci that are assumed to increase sclerostin levels due to greater \nBMD, and to have no relevance when considering potential therapeutic effects of sclerostin \nlowering on BMD(32).  \n \nResults of the random effects meta-analysis were similar to those of the fixed-effect meta-\nanalysis (results not shown). The degree of heterogeneity was low across studies for most of \nthe identified genetic variants ( Table 2 ). Conditional analyses on the lead SNP in each \nassociation locus yielded one additional independent signal reaching genome-wide \nsignificance in the LVRN gene region and two additional independent signals in the TNFSF11 \ngene region (Supplementary Table 5).  \n \nGenetic colocalization analysis of sclerostin association signals with gene expression \nFor the 18 sclerostin associated variants, we identified four variants [rs215223 (in the \nB4GALNT3 region), rs28929474 (in the SERPINA1 region), rs66838809 (in the SOST region) \nand rs7143806 (in the RIN3 region)] where sclerostin-increasing alleles were associated with \n . CC-BY 4.0 International licenseIt is made available under a \n is the author/funder, who has granted medRxiv a license to display the preprint in perpetuity. (which was not certified by peer review)\nThe copyright holder for this preprint this version posted June 16, 2022. ; https://doi.org/10.1101/2022.06.13.22275915doi: medRxiv preprint \n\n \n8 \n \nlower eBMD (at P<0.001), whereas in the case of the remaining SNPs, variants were either \nnot associated with eBMD or the sclerostin-increasing alleles were associated with higher \neBMD, and were not considered further ( Supplementary Table 5 ). We conducted genetic \ncolocalization analysis for rs215223, rs28929474, rs66838809 and rs7143806 to confirm the \ncausal variants were shared between circulating sclerostin levels and gene expression levels \nof the related genes in tibial artery (data from GTEx v8). The expression of B4GALNT3 and \nSOST genes showed strong evidence of colocalization with circulating sclerostin levels \n(colocalization probability=99% and 98%, respectively; Supplementary Table 6A ). The \nSERPINA1 and RIN3 signal showed weaker evidence of colocalization with sclerostin \n(colocalization probability=7% and 30%; Supplementary Figure 3 and  Supplementary \nTable 6A). We also confirmed that the SOST SNP we identified was associated with altered \nSOST expression in iliac crest bone tissue (Supplementary Table 6B)(33).  \n \nBioinformatics functional follow-up \nWe investigated possible effects of rs215223 (in the B4GALNT3 region) and rs66838809 (in \nthe SOST region) on transcriptional activity. The chromatin accessibility analysis predicted \nby ChromHMM based on 5 chromatin marks for 127 epigenomes (34) identified that the \nB4GALNT3 variant (rs215223) overlapped with an enhancer region in osteoblast primary \ncells (OPCs), and the SOST variant (rs66838809) overlapped with an active transcription start \nsite (TSS) in OPCs (see Figure 2 ). The same analysis in other heart and vascular-related \ntissues showed that the SOST variant (rs66838809) also overlapped with an active TSS in \nthree heart tissues. The SERPINA1 and RIN3 variants overlapped with weak transcription \n(Supplementary Table 7 ). Regulatory elements analysis using RegulomeDB (35) graded \nB4GALNT3, SOST and RIN3  variants as rank 2B, 3A and 3A, respectively (lower rank \nimplies greater predicted functional impact; Supplementary Table 5 ). The eQTL lookup \nusing STARNET showed that our top hits rs215223 (for) and rs66838809 (for SOST) were \nassociated with gene expression levels of B4GALNT3 and SOST in free internal mammary \nartery (MAM) respectively (Supplementary Table 6C ). The gene-set enrichment analysis \nshowed that RIN3 and B4GALNT3 were enriched in the same enzyme linked receptor protein \nsignalling pathway (Gene Ontology ID: GO:0007167). SERPINA1 and SOST were enriched \nin two separate pathways that related to inflammatory response (GO:0072358). In addition, \nSOST was also enriched in a cardiovascular system development pathway (GO:0072358), \nwhere RIN3 was enriched in a pathway that positive regulate immune system process \n(GO:0002684) (Supplementary Table 6D).  \n . CC-BY 4.0 International licenseIt is made available under a \n is the author/funder, who has granted medRxiv a license to display the preprint in perpetuity. (which was not certified by peer review)\nThe copyright holder for this preprint this version posted June 16, 2022. ; https://doi.org/10.1101/2022.06.13.22275915doi: medRxiv preprint \n\n \n9 \n \n \nGenetic correlation between sclerostin levels and atherosclerosis-related traits \nAs expected, genetic correlation analysis between circulating sclerostin using genetic variants \nacross the whole genome revealed a relationship between lower sclerostin and higher eBMD \nand, to a lesser extent, lower fracture risk ( Supplementary Table S8 ). These analyses also \nshowed a genetic overlap of lower sclerostin with increased hypertension risk (r\ng=0.134, \nP=3.10×10-3), but not with any other atherosclerosis-related diseases or risk factors ( Table 3 \nand Supplementary Figure 4). \n \nSclerostin instruments and effects of lower sclerostin on risk of atherosclerosis-related \ndiseases and risk factors \nMR analyses of the effect of lower sclerostin levels on atherosclerosis risk combined the \nSOST cis variants with the B4GALNT3, SERPINA1 and RIN3 trans variants identified above. \nFor all four SNPs, the alleles associated with lower circulating sclerostin levels were \nassociated with increased eBMD and reduced fracture risk (Figure 3A).  Compared to the \nother three variants, the SOST  SNP showed a disproportionately strong association with \neBMD (Supplementary Table 1A ), relative to its association with circulating sclerostin \n(Figure 3B ). Together, these four SNPs explained 1.21% of the variance in circulating \nsclerostin and provided a strong genetic instrument (F-statistic 89.8, an F statistic of at least \n10 is indicative of evidence against weak instrument bias) (Supplementary Table 1). For the \nremaining 14 sclerostin variants, five variants were not associated with eBMD, where the \nother nine variants showed directionally similar effects on eBMD and sclerostin. These \nvariants did not fit with our selection criteria and were therefore excluded from the \ninstrument list (Supplementary Figure 5).  \n \nUsing these four conditionally independent SNPs to evaluate causal effects of lower \nsclerostin levels on atherosclerosis-related diseases and risk factors (Bonferroni-corrected \nthreshold=5.15×10\n-3), a predicted lower circulating sclerostin was found to be associated with \nan increased risk of hypertension (OR per SD decrease in sclerostin= 1.09, 95% CI=1.04 to \n1.15, P=7.93×10\n-4) (Figure 4A).  Sensitivity analyses including MR-Egger and a \nheterogeneity test suggested little evidence of horizontal pleiotropy (Egger regression \nintercept=-0.003, P=0.27) or heterogeneity (Cochran’s Q=2.85, P=0.42; Supplementary \nTable 9A ). In contrast, little evidence for a causal effect of lower sclerostin on any other \natherosclerosis-related disease or risk factor was identified. Given the low power to detect \n . CC-BY 4.0 International licenseIt is made available under a \n is the author/funder, who has granted medRxiv a license to display the preprint in perpetuity. (which was not certified by peer review)\nThe copyright holder for this preprint this version posted June 16, 2022. ; https://doi.org/10.1101/2022.06.13.22275915doi: medRxiv preprint \n\n \n10 \n \npleiotropy using the above methods due to the limited number of genetic instruments \navailable, we conducted a proteome-wide association scan of the four genetic variants to \nfurther examine potential pleiotropy. All variants, except rs28929474 within the SERPINA1  \ngene region, showed little evidence of association with any other proteins, where rs28929474 \nwas associated with additional 27 proteins (Supplementary Table 9B).  \n \nIn further analyses using cis -only instruments, we observed that the five correlated variants \n(rs66838809, rs1107747, rs4793023, rs80107551, rs76449013) together explained 0.4% of \nthe variance in circulating sclerostin and had acceptable instrument strength (F-statistic 24.8) \n(Supplementary Table 2 ). The cis -only analysis identified potential adverse effects of \nsclerostin inhibition on risk of hypertension (OR=1.08, 95% CI=1.01 to 1.15, P=0.03; Figure \n4A), T2DM (OR=1.26, 95% CI=1.08 to 1.48, P=0.004; Figure 4B ) and MI (OR=1.31, 95% \nCI=1.183 to 1.45, P=2.17×10\n-7; Figure 4C ). Genetically predicted lower sclerostin was \nassociated with higher levels of CAC ( β=0.74; 95% CI=0.33 to 1.15; P=4.27×10 -4; Figure \n4D). In addition, lower sclerostin in the cis-only analyses showed potential harmful effect on \nAAC and CAD, but with very wide confidence intervals ( Supplementary Table 10A ). In \naddition, sclerostin showed effects on four of the five lipids and/or lipoproteins in the cis-\nonly analysis but with little evidence of MR effects in the combined cis and trans  analysis \n(Figure 5A and Supplementary Table 10A). This could be partly caused by the LD between \nthe cis SOST variant (rs4793023) and the top-associated variant with mRNA CD300LG \nexpression (rs72836567; LD r 2=0.22 in the 1000 Genomes EUR population) (36), where \nCD300LG is known to be strongly associated with lipid measures (37). As a sensitivity \nanalysis, we excluded rs4793023 from the genetic predictor list and ran the cis-only MR \nusing the remaining four predictors. The results suggested that decreased sclerostin levels \nreduced HDL-C levels and increased triglycerides levels, whereas the MR effects on other \nlipids/lipoproteins were attenuated after this adjustment ( Supplementary Table 10B ). \nHeterogeneity analysis of MR estimates of each genetic instrument suggested little evidence \nof heterogeneity across the five genetic instruments (Cochran’s Q test P>0.05 for these four \nlipids and/or lipoproteins measures; Supplementary Table 10A and 10B). \n \nEffects of atherosclerosis-related diseases and risk factors on circulating sclerostin \nWe further conducted bidirectional MR (38) to evaluate the potential reverse causality of \natherosclerosis-related diseases and risk factors on circulating sclerostin. We used the 15 \n . CC-BY 4.0 International licenseIt is made available under a \n is the author/funder, who has granted medRxiv a license to display the preprint in perpetuity. (which was not certified by peer review)\nThe copyright holder for this preprint this version posted June 16, 2022. ; https://doi.org/10.1101/2022.06.13.22275915doi: medRxiv preprint \n\n \n11 \n \natherosclerosis-related diseases and risk factors as exposures, of which nine had valid \npredictors to conduct bidirectional MR ( Supplementary Table 4 ; small vessel disease had \nno valid genetic predictors and was therefore excluded from this analysis). Circulating \nsclerostin was treated as the outcome. IVW results using 226 T2DM-associated variants \nshowed a marginal positive relationship for liability of T2DM on sclerostin ( β=0.02, SD \nchange in sclerostin per unit increase of risk score of T2DM, 95%CI= 0.001 to 0.045, P=0.04; \nSupplementary Table 11A ). The IVW results showed that apoB was negatively associated \nwith sclerostin levels ( β=-0.03, 95%CI=-0.01 to -0.06, P=3.67×10 -3). However, the \nmultivariable MR including apoB, LDL-C and triglycerides in the same model suggested that \nincreased apoB levels increased sclerostin levels ( β=0.03, 95%CI=0.001 to 0.07, P=0.041; \nFigure 5B and Supplementary Table 11B). No other atherosclerosis-related disease or risk \nfactor showed a reverse effect on sclerostin (Supplementary Table 11A ). The MR-Egger \nintercept test did not suggest evidence of directional pleiotropy. The heterogeneity test \nshowed weak evidence for any heterogeneity of the causal estimates ( Supplementary Table \n11A). The Steiger filtering analysis further confirmed that the sclerostin instruments were \nlikely to first change the sclerostin level and then influence the atherosclerosis outcomes as a \ncausal consequence (Supplementary Table 11C).  \n  \n . CC-BY 4.0 International licenseIt is made available under a \n is the author/funder, who has granted medRxiv a license to display the preprint in perpetuity. (which was not certified by peer review)\nThe copyright holder for this preprint this version posted June 16, 2022. ; https://doi.org/10.1101/2022.06.13.22275915doi: medRxiv preprint \n\n \n12 \n \nDISCUSSION \nWe have presented findings from an updated GWAS meta-analysis of circulating sclerostin, \nwith three times the sample size of our previous study (15). We identified 18 sclerostin-\nassociated variants, of which four in the SOST , B4GALNT3, RIN3  and SERPINA1 genes \nprovided useful genetic instruments for determining the causal effects of lower sclerostin \nlevels on atherosclerosis-related diseases and risk factors based on inverse relationships \nbetween sclerostin levels and BMD. The B4GALNT3 result replicated the top hit, rs215226, \nfrom our previous study; RIN3 and SERPINA1 are novel sclerostin-associated trans signals, \nand the SOST finding represents a strong cis  signal which was only marginal in our previous \nstudy. MR analyses using these four SNPs as a combined cis and trans genetic instrument \nsuggested that lower sclerostin levels increase the risk of hypertension, however there was \nlittle evidence of a relationship with other atherosclerosis-related diseases or risk factors. On \nthe other hand, sensitivity analyses using a cis-only genetic instrument suggested that lower \nsclerostin levels increase the risk of hypertension, T2DM and MI and increase the extent of \nCAC. In addition, cis-only analyses suggested that lower sclerostin levels reduce HDL-C and \napoA-I, and increase apoB and TG; effects on HDL-C and TG persisted following exclusion \nof the SOST variant rs4793023 (in LD with rs72836567 in the adjacent CD300LG gene \npreviously found to be associated with lipid measures)(37).  \n \nLower sclerostin levels had similar predicted effects on hypertension risk using both cis-only \nand combined cis and trans instruments, and we observed evidence of genetic correlation \nbetween sclerostin inhibition and hypertension using variants across the whole genome, \nwithout evidence of reverse causality. Together, these findings provide reasonable evidence \nof a causal effect of sclerostin on risk of hypertension. cis-only analyses suggested additional \ncausal effects of lower sclerostin levels on atherosclerosis-related diseases and risk factors, \nand in particular that inhibition of sclerostin level increases risk of MI as a consequence of \ngreater CAC. That said, relationships between CAC and clinical events are potentially \ncomplex, with some evidence suggesting that CAC  reflects the presence of stable \natherosclerotic plaques with a reduced risk of coronary artery occlusion compared with \nuncalcified plaques(39)(40)(41). Bidirectional analyses broadly supported a causal effect of \nsclerostin inhibition on increased risk of atherosclerosis-related diseases and risk factors, as \nopposed to vice versa. That said, in addition to a causal effect of lower sclerostin levels on \nrisk of T2DM in cis-only analyses, there was marginal evidence for a causal effect of T2DM \n . CC-BY 4.0 International licenseIt is made available under a \n is the author/funder, who has granted medRxiv a license to display the preprint in perpetuity. (which was not certified by peer review)\nThe copyright holder for this preprint this version posted June 16, 2022. ; https://doi.org/10.1101/2022.06.13.22275915doi: medRxiv preprint \n\n \n13 \n \non sclerostin. Moreover, whereas the cis  instrument suggested a causal effect of sclerostin \ninhibition on CAC and MI, there was no equivalent effect on AAC.   \n \nCis instruments are more likely to directly link with biology, which aligns with our finding \nthat cis-only analyses identified more extra-skeletal effects of sclerostin. Trans instruments \nare, by their nature, more likely to be pleiotropic, and as a result, they may include additional \npathways responsible for this apparent divergence in effects on eBMD and extra-skeletal \npathways. Additionally, cis variants may be more predictive of tissue sclerostin levels \nresponsible for mediating biological effects. Based on eQTL data using bone tissue, the cis \nsignal is predicted to alter expression and hence local levels of sclerostin in bone. Osteocytes, \nembedded within bone, are the primary source of sclerostin, which then circulates locally \nthrough canaliculi to modulate the activity of other bone cells, including osteoblasts, leading \nto changes in bone mass and strength (42). Accordingly, the cis signal is expected to alter \ncirculating levels of sclerostin through exchange between bone tissue and the circulation. In \ncontrast, we previously hypothesised that the trans signal, B4GALNT3 , replicated in the \npresent study, primarily influences circulating sclerostin levels by affecting plasma clearance \ndue to altered protein glycosylation (15). Hence, any changes in tissue sclerostin levels \nresulting from the B4GALNT3 trans signal are likely secondary to altered circulating levels, \nrather than local production. Therefore, by its nature, the B4GALNT3 trans signal is \nexpected to produce smaller changes in tissue sclerostin levels compared to a cis SOST signal, \nleading to a weaker effect on eBMD.  \n \nThat the SOST cis signal is likely to produce greater increases in tissue sclerostin levels \ncompared to trans signals may also explain why the cis-only analyses predicted more extra-\nskeletal effects of sclerostin lowering compared to the cis+trans analyses. Sclerostin is also \nexpressed in vascular tissues including at sites of vascular calcification (43)(44), suggesting \nany effects of sclerostin on vascular tissues may also involve local sclerostin expression. \nSuch an effect is likely mediated by sclerostin’s well recognised action as a WNT \ninhibitor(45), given the contribution of WNT signalling to the development of \natherosclerosis(46). \n \nPharmacokinetic studies suggest that romosozumab is largely retained within the \ncirculation(13), in-keeping with the relatively large size of a monoclonal antibody. That said, \n . CC-BY 4.0 International licenseIt is made available under a \n is the author/funder, who has granted medRxiv a license to display the preprint in perpetuity. (which was not certified by peer review)\nThe copyright holder for this preprint this version posted June 16, 2022. ; https://doi.org/10.1101/2022.06.13.22275915doi: medRxiv preprint \n\n \n14 \n \nthe pharmacological action of romosozumab, involving neutralisation of sclerostin activity in \nbone tissue, depends on the antibody penetrating skeletal tissue after systemic administration, \nwhich is likely to involve convection or endocytosis/pinocytosis via endothelial cells (47). To \nthe extent that effects of romosozumab on CVD risk also involve local tissue penetration, a \ncis instrument reflecting tissue levels of sclerostin may be more likely to predict effects of \nromosozumab on CVD risk than a trans instrument more closely linked to systemic levels. \n \nThere have also been several previous observational studies examining associations between \ncirculating sclerostin and atherosclerosis related diseases and risk factors, of which the largest \nwas our recent study of over 5000 participants across two cohorts (7). We found that \ndecreased sclerostin levels increased risk of T2DM and triglyceride levels and reduced HDL-\nC levels; higher sclerostin was also associated with an increased severity of CAD as \nmeasured on angiogram, and an increased risk of death from cardiac disease during \nsubsequent follow up (7). These observed associations suggest causal effects in the opposite \ndirection to those predicted by our MR analyses,   particularly in analyses restricted to the cis \ninstrument. Interestingly, directionally opposite effects have also been observed in the case of \neBMD and atherosclerosis risk, with a protective effect found in an observational analysis but \na harmful effect predicted by MR analyses (48). The latter finding also raises the possibility \nthat any effect of sclerostin inhibition on atherosclerosis risk might be an indirect \nconsequence of increased BMD, as opposed to a specific effect of sclerostin. However, \narguing against this suggestion, there is little evidence that other therapeutic agents for \nosteoporosis acting to increase BMD affect atherosclerosis risk, apart from strontium ranelate \nfor which the European Medicines Agency issued a warning, restricting use in those with a \nhigh risk of CVD(49). \n \nTwo previous studies have used MR approaches to examine causal effects of sclerostin \ninhibition on atherosclerosis and related risk factors. Bovijn et al. reported that two \nconditionally independent SOST SNPs, selected on the basis of their association with eBMD \nin a previous UK Biobank GWAS (of which one SNP also showed evidence of colocalization \nwith lower SOST mRNA expression in tibial artery tissue), predicted higher risk of MI and/or \ncoronary revascularization, major cardiovascular events, hypertension, and T2DM (12). Our \nfindings, using the cis-only instrument for circulating sclerostin, are consistent with these \nobservations. In contrast, Holdsworth et al. found no association between gene expression \nlevel of SOST  in tibial artery/heart tissue and CVD risk, using three cis SOST eQTLs as \n . CC-BY 4.0 International licenseIt is made available under a \n is the author/funder, who has granted medRxiv a license to display the preprint in perpetuity. (which was not certified by peer review)\nThe copyright holder for this preprint this version posted June 16, 2022. ; https://doi.org/10.1101/2022.06.13.22275915doi: medRxiv preprint \n\n \n15 \n \ninstruments(14). Though one SOST SNP (rs9899889, in low LD with the SOST SNPs used \nhere (see Supplementary Table 12) was associated with lower triglyceride levels and higher \nHDL cholesterol and apoA levels, in contrast to our findings, this signal was entirely \nexplained by LD with the adjacent CD300LG gene known to be associated with lipid \nmeasures(37). Despite the distinct methods used to proxy sclerostin inhibition, our cis \ninstrument was in strong LD with those used in these other studies. Indeed, our cis instrument \nshared an identical SNP with the Holdsworth study (see Supplementary Table 12).  \n \nIn terms of other trans -acting pathways, we identified a further glycosylation enzyme, \nGALNT1, as being associated with circulating sclerostin levels in our earlier GWAS. \nHowever, this association did not replicate in the present expanded GWAS. On the other \nhand, we identified two new trans signals for sclerostin, RIN3 and SERPINA1.  Previous \nGWASs have identified RIN3 in association with lower limb and total BMD in children (31), \nand Paget’s disease of bone (50). Though altered sclerostin levels could conceivably underlie \nthis association with childhood BMD, this is less likely to apply to Paget’s disease, which is \nprimarily an osteoclast disorder. Homozygosity of SERPINA1 underlies deficiency of α1AT, \na glycoprotein mostly produced by the liver, which serves to protect lung tissue from tissue \ndamage caused by proteases released from neutrophils. The loss of function allele was \nassociated with higher sclerostin levels, and the mechanisms underlying this genetic \nassociation are unclear. α1AT deficiency causes early-onset chronic obstructive pulmonary \ndisease (COPD) (51), and whereas rs28929474 heterozygosity has been associated with \nincreased human height(51) , we are not aware of any previous findings relating α1AT to \nBMD or risk of osteoporosis. Given the lack of evidence of colocalization, it is also possible \nthat a different gene was responsible for the genetic signal identified at this locus.  \n \nIn terms of strengths, the present study had sufficient sample size to clearly detect a cis \n(SOST) signal, and our genetic instrument successfully accounted for bidirectional effects \nbetween sclerostin and BMD, by removing trans SNPs with the same direction of effect on \nsclerostin and eBMD. Our MR of sclerostin effects on atherosclerosis-related diseases and \nrisk factors used circulating protein level of sclerostin as the exposure, which may predict \nadverse effects from sclerostin antibody inhibition more accurately than previous studies \nusing BMD or SOST arterial expression as exposures. Finally, since genetic predictors in the \ncis- and/or trans -acting regions may yield different causal estimates on outcomes, we \n . CC-BY 4.0 International licenseIt is made available under a \n is the author/funder, who has granted medRxiv a license to display the preprint in perpetuity. (which was not certified by peer review)\nThe copyright holder for this preprint this version posted June 16, 2022. ; https://doi.org/10.1101/2022.06.13.22275915doi: medRxiv preprint \n\n \n16 \n \nconsidered these separately. In terms of weaknesses, though postmenopausal women are the \nmain target group for osteoporosis treatments such as romososumab, we were only able to \nexamine predicted effects of sclerostin inhibition in males and females combined, due to the \nlack of availability of sex-specific sclerostin GWAS dataset. In addition, the different cohorts \nused distinct methods to measure sclerostin, with the over half providing sclerostin measures \nthrough the SomaLogic platform, while the other half used a specific ELISA. However, \ndespite these methodological differences, there was little evidence of heterogeneity of genetic \nassociations between cohorts.  \n \nIn conclusion, our updated GWAS meta-analysis of circulating sclerostin now identified a \nrobust cis (SOST) signal, replicated our previous B4GALNT3 signal, and identified new trans \nsignals in the RIN3  and SERPINA1 genes. To predict adverse effects of sclerostin inhibition, \nthese signals were combined to provide a cis+trans instrument for an MR analysis of effects \nof lower sclerostin levels on atherosclerosis-related diseases and risk factors. Genetically \npredicted lower sclerostin levels were found to increase the risk of hypertension, a \nrelationship that was supported by the finding of an inverse genetic correlation between \nsclerostin and hypertension at the genome-wide level, and the lack of any evidence of reverse \ncausality. Analyses based on the cis (SOST) instrument alone found a similar causal effect of \nlower sclerostin levels on hypertension risk, and additionally suggested causal effects on risk \nof MI and T2DM, increased CAC, reduced HDL-C and increased apoB and TG levels. To the \nextent that genetically predicted lower lifelong exposure to sclerostin shares consequences \nwith pharmacological inhibition over 12 months, our results underscore the requirement for \nstrategies to mitigate potential adverse effects of romosozumab treatment on atherosclerosis \nand its related risk factors.  \n  \n . CC-BY 4.0 International licenseIt is made available under a \n is the author/funder, who has granted medRxiv a license to display the preprint in perpetuity. (which was not certified by peer review)\nThe copyright holder for this preprint this version posted June 16, 2022. ; https://doi.org/10.1101/2022.06.13.22275915doi: medRxiv preprint \n\n \n17 \n \nMaterials and Methods \nGWAS Meta-analysis of sclerostin \nSclerostin measures in the nine cohorts were standardized to SD units. Each cohort ran a \nGWAS across all imputed or sequenced variants. Age and sex and the first 10 principal \ncomponents (PCs) were included as covariates in all models (except INTERVANL and \nLURIC). For the INTERVAL study, age, sex, duration between blood draw and processing \n(binary, \n≤ 1 day / >1 day) and the first three PCs were included in the genetic association \nmodel. The Fenland, ALSPAC, 4D, and GOOD cohorts were imputed using the Haplotype \nReference Consortium (HRC) V1.0 reference panel (MANOLIS employed whole-genome \nsequencing). Linear mixed models BOLT-LMM and GEMMA were applied to the ALSPAC \nand MONOLIS cohorts, respectively, to adjust for cryptic population structure and \nrelatedness. Linear regression was performed using BGENIE (v1.3)(52) in the Fenland study. \nFor the INTERVAL study, a linear regression model was applied using genotype data \nimputed by a combined 1000 Genomes Phase 3-UK10K reference panel. The LURIC cohort \nwas imputed to the 1000 Genomes Phase 1 reference panel with linear regression performed \nin PLINK v1.90(53) using eight ancestry components as covariates to control for population \nsubstructure. In HUNT, the GWAS for protein was performed using rank transformed \nresiduals of protein adjusted for age, sex, batch effect, phase effect and principal components \nfrom 1-20. \n \nWe standardized the genomic coordinates to be reported on the NCBI build 37 (hg19), and \nalleles on the forward strand. Summary level quality control was conducted for Europeans \nonly in EasyQC(54). Meta-analysis (using a fixed-effect model implemented in METAL(55)) \nwas restricted to variants with a minimal sample size >10,000 individuals, MAF >1%, and \nhigh imputation quality score (R\n2 >0.8 for variants imputed in MaCH(56) and INFO >0.8 for \nvariants imputed in IMPUTE(57) (n=11,680,861 variants). Meta-analysed P value lower than \n5×10−8 was used as a threshold to define genome-wide significant associations. A random \neffects model meta-analysis was also conducted using GWAMA version 2.2.2 (58). \nHeterogeneity was assessed using the I2 statistic and Cochran's Q test.  \n \nConditional analysis and genetic fine mapping\n \nWe carried out an approximate conditional and joint genome-wide association analysis \n(GCTA-COJO) to detect multiple independent association signals at each of the sclerostin \nlocus(59). SNPs with high collinearity (Correlation r\n2 > 0.9) were ignored, and those situated \n . CC-BY 4.0 International licenseIt is made available under a \n is the author/funder, who has granted medRxiv a license to display the preprint in perpetuity. (which was not certified by peer review)\nThe copyright holder for this preprint this version posted June 16, 2022. ; https://doi.org/10.1101/2022.06.13.22275915doi: medRxiv preprint \n\n \n18 \n \nmore than 10 Mb away were assumed to be in complete linkage equilibrium (LD). A \nreference sample of 8,890 unrelated individuals of ALSPAC mothers was used to model \npatterns of LD between variants. The reference genotyping data set consisted of the same \n11.6 million variants assessed in our GWAS. Conditionally independent variants with \nP<5×10\n-8 were annotated to the physically closest gene with the hg19 gene range list \navailable in dbSNP (https://www.ncbi.nlm.nih.gov/SNP/).  \n \nFunctional mapping and annotation of sclerostin genetic association signals \nGenetic colocalization of gene expression quantitative trait loci (eQTLs) and the sclerostin \nsignals \nWe investigated whether the SNPs influencing serum sclerostin level were driven by cis-\nacting effects on transcription by evaluating the overlap between the sclerostin-associated \nSNPs and eQTLs within 500kb of the gene identified, using data derived from all tissue types \nfrom GTEx v8(36). Evidence of eQTL association was defined as P < 1×10\n-4 and evidence of \noverlap of signal was defined as high LD (r 2 ≥  0.8) between eQTLs and sclerostin-associated \nSNPs in the region. Where eQTLs overlapped with sclerostin-associated SNPs, we used \ngenetic colocalization analysis(60) to estimate the posterior probability (PP) of each genomic \nlocus containing a single variant affecting both circulating sclerostin and gene expression \nlevels in different tissues.  \n \nWe used Functional Mapping and Annotation of Genome-Wide Association Studies (FUMA), \nan integrative web-based platform ( http://fuma.ctglab.nl), containing information from 18 \nbiological data repositories and tools, to characterise the genetic association signals of \nsclerostin. According to: (i) functional consequences on gene functions, (ii) mapped genes \nand biological pathways, and (iii) associations with other phenotypes. The FUMA pipeline \nhas been described in detail elsewhere (61). First, we applied the basic plotting function of \nFUMA to create Manhattan and QQ plots of our sclerostin GWAS meta-analysis results as \nwell as regional plots for top loci. We then applied FUMA’s SNP2GENE function, which \nused the conditionally independent significant SNPs, and annotated the functional \nconsequences of these variants on gene functions (i.e., altering expression of a gene, affecting \na binding site or changing the protein structure). Functionally annotated SNPs were \nsubsequently mapped to genes based on functional consequences by (i) physical position on \nthe genome (positional mapping), (ii) eQTL associations (eQTL mapping), and (iii) 3D \nchromatin interactions (chromatin interaction mapping). Gene-based/gene-set analyses using \n . CC-BY 4.0 International licenseIt is made available under a \n is the author/funder, who has granted medRxiv a license to display the preprint in perpetuity. (which was not certified by peer review)\nThe copyright holder for this preprint this version posted June 16, 2022. ; https://doi.org/10.1101/2022.06.13.22275915doi: medRxiv preprint \n\n \n19 \n \nMAGMA were carried out to summarize SNP associations at the gene level and associate the \nset of genes to biological pathways. For each sclerostin-associated locus, we identified all \nSNPs in high LD with the top signal (LD r 2>0.8) and characterized their DNA features and \nregulatory elements in non-coding regions of the human genome using RegulomeDB \n(http://www.regulomedb.org/\n)(35), as implemented in FUMA. In addition, we estimated the \nchromatin accessibility of genomic regions (every 200bp) for the top loci ( SOST and \nB4GLANT3) using 15 categorical states predicted by ChromHMM (34) based on five \nchromatin marks for 127 epigenomes (lower state with higher accessibility).  \n \nFor the top genes been identified as associated with circulating sclerostin and eBMD, we \nsearched their eQTLs in 7 tissues related to cardio-metabolic phenotypes (which including \nblood, free internal mammary artery [MAM], atherosclerotic aortic root [AOR], \nsubcutaneous fat [SF], visceral abdominal fat [VAF], skeletal muscle [SKLM], and liver \n[LIV]) as well as gene-set enrichment using the STARNET web app(62).  \n \nLD score regression analyses \nEstimation of SNP heritability using LD score regression \n \nTo estimate the amount of genomic inflation in the data due to residual population \nstratification, cryptic relatedness, and other latent sources of bias, we used LD score \nregression(63). LD scores were calculated for all high-quality SNPs (i.e., INFO score [or R\n2] > \n0.9 and MAF > 0.1%) from the meta-analysis. We further quantified the overall SNP-based \nheritability with LD score regression using a subset of 1.2 million HapMap SNPs (SNPs in \nthe major histocompatibility complex [MHC] region were removed due to complex LD \nstructure).  \n \nEstimation of genetic correlations using LD Hub \n \nTo estimate the genetic correlation between reduced sclerostin level and 12 atherosclerosis-\nrelated diseases and risk factors and two bone phenotypes, we used a platform based on LD \nscore regression as implemented in the online web utility LD Hub (64). This method uses the \ncross-products of summary test statistics from two GWASs and regresses them against a \nmeasure of how much variation each SNP tags (i.e., its LD score). Variants with high LD \nscores are more likely to contain more true signals and thus provide a greater chance of \noverlap with genuine signals between GWASs. The LD score regression method uses \nsummary statistics from the GWAS meta-analysis of sclerostin and the atherosclerosis-\n . CC-BY 4.0 International licenseIt is made available under a \n is the author/funder, who has granted medRxiv a license to display the preprint in perpetuity. (which was not certified by peer review)\nThe copyright holder for this preprint this version posted June 16, 2022. ; https://doi.org/10.1101/2022.06.13.22275915doi: medRxiv preprint \n\n \n20 \n \nrelated diseases and risk factors/bone phenotypes, calculates the cross-product of test \nstatistics at each SNP, and then regresses the cross-product on the LD score. The small vessel \ndisease data had the heritability estimate out-of-bounds and was therefore removed.  \n \nMendelian randomization \nSelection of genetic predictors for sclerostin\n \nFrom the 18 conditionally independent sclerostin variants identified ( Supplementary Table \n1A), we selected valid genetic predictors of sclerostin for the MR using three further criteria: \n(i) the genetic variants showed predicted effects of sclerostin on BMD estimated using \nultrasound in heel (eBMD, data from UK Biobank; single SNP MR P value of sclerostin on \neBMD<0.001); (ii) the sclerostin reducing alleles of the genetic variants were associated with \nincreased BMD level (i.e., these variants showed a negative Wald ratio (65) for sclerostin on \nBMD). The final set of four genetic variants after applying these two additional criteria are \nlisted in Supplementary Table 1B. The analysis using these four variants is noted as the cis\n \nand trans analysis. \n \nDue to the greater relevance of the cis -acting variants, we conducted a sensitivity analysis \nusing genetic variants restricted to cis-acting variants (defined as ±500\n/i4 kb genomic region \nfrom the leading SOST SNP) (noted as the cis-only analysis). Of the 41 SNPs associated with \ncirculating sclerostin (at a regional-wide association threshold<1×10 -6) in the SOST region \n(±500/i4 kb genomic region from rs66838809), LD clumping identified five correlated SNPs \nwith LD r2<0.8 (Supplementary Table 2 ). Such an LD r 2 threshold was used here to avoid \nmulti-collinearity caused by SNPs in very high LD. These correlated instruments were used \nin a generalised MR model that considered LD among instruments (more details in later \nsection). Instrument strength was evaluated using F-statistics. \n \nOutcome selection\n \nFor the MR analysis estimating the potential adverse effects of sclerostin inhibition, we \nselected eight atherosclerosis-related diseases and seven atherosclerosis-related risk factors as \nprimary outcomes ( Supplementary Table 3 ). This list comprised two endpoints related to \nischaemic heart disease (coronary artery disease (CAD) and MI), four stroke endpoints \n(ischemic stroke, cardioembolic stroke, large vessel disease, small vessel disease), two \nmeasures of arterial calcification [CAC, abdominal aortic calcification (AAC)], hypertension, \nT2DM, and five lipids/lipoproteins risk factors [low density lipoprotein (LDL-C), high \n . CC-BY 4.0 International licenseIt is made available under a \n is the author/funder, who has granted medRxiv a license to display the preprint in perpetuity. (which was not certified by peer review)\nThe copyright holder for this preprint this version posted June 16, 2022. ; https://doi.org/10.1101/2022.06.13.22275915doi: medRxiv preprint \n\n \n21 \n \ndensity lipoprotein (HDL-C), triglycerides, apolipoprotein A-I (apoA-I), and apolipoprotein \nB (apoB)]. After applying PhenoSpD (66), which takes into account the correlation between \nthe 15 atherosclerosis-related diseases and risk factors, the number of independent tests was \n9.7 (Bonferroni corrected threshold=5.15×10 -3). We looked up GWAS results in datasets \nnon-overlapping to those used for the sclerostin GWAS, namely the T2DM GWAS from \nMahajan et al. (N cases=74,124, N controls=824,006) (67); four stroke GWASs from the \nMETASTROKE consortium (N cases=10,307, N controls=19,326) (68), the CAD GWAS \nfrom Van der Harst et al. (N cases=34,541, N controls=261,984) (69), the MI GWAS from \nHartiala et al. (N cases=61000, N controls=578000)(70), the CAC GWAS from Kavousi et al. \n(N=26,909 CHARGE participants with CAC score) (71); the hypertension, lipids and \nlipoproteins GWASs from UK Biobank IEU GWAS data release (N hypertension \ncases=119,731, N controls=343,202; N lipids/lipoproteins=441,016) (72) and the AAC \nGWAS from Malhotra et al. (sample size=9,417)(73).  \n \nMendelian randomization of sclerostin on atherosclerosis-related phenotypes\n \nFor the cis  and trans analysis, we applied a set of two-sample MR approaches [inverse \nvariance weighted (IVW), MR-Egger, weighted median, single mode estimator and weighted \nmode estimator](74) to estimate the effect of circulating sclerostin on the 15 atherosclerosis-\nrelated diseases and risk factors. Although we had a small number of relevant variants \navailable for this analysis, we still used the MR-Egger intercept term as an indicator of \npotential directional pleiotropy(75). Heterogeneity analysis of the instruments was conducted \nusing Cochran’s Q test.  \n \nFor the cis-only analysis, we applied a generalised IVW MR model followed by generalised \nEgger regression to account for LD structure between correlated SNPs in the SOST region \nand to boost statistical power (76). The generalised Egger regression intercept term was used \nas an indicator of potential directional pleiotropy. For the cis, trans, and cis -only analyses, \nthe above-mentioned Bonferroni corrected P-value threshold of 5.0×10\n-3 was used to control \nfor multiple testing.  \n \nBidirectional Mendelian randomization analysis of atherosclerosis-related phenotypes on \nsclerostin \nTo investigate the possibility of reverse causality between atherosclerosis-related diseases \nand risk factors and circulating sclerostin level, we used genetic variants associated with 15 \n . CC-BY 4.0 International licenseIt is made available under a \n is the author/funder, who has granted medRxiv a license to display the preprint in perpetuity. (which was not certified by peer review)\nThe copyright holder for this preprint this version posted June 16, 2022. ; https://doi.org/10.1101/2022.06.13.22275915doi: medRxiv preprint \n\n \n22 \n \natherosclerosis-related diseases and risk factors as genetic predictors (small vessel disease \ndata has no valid genetic predictors, therefore, we were not able to perform bidirectional MR \nfor this trait; for other genetic predictors, the genetic association data were extracted from \nrelevant GWAS listed in Supplementary Table 4A ). For this analysis, the circulating \nsclerostin level from our GWAS meta-analysis was used as the outcome. We applied the \nsame five two-sample MR approaches (IVW, MR-Egger, weighted median, single mode \nestimator and weighted mode estimator)(74)(17) . In addition, due to correlation between \nlipids and lipoproteins, we further applied a multivariable MR model (77) to estimate the \nindependent effect of each lipid and lipoprotein on sclerostin. For the genetic predictors of \nthese lipids and lipoproteins, see Supplementary Table 4B and 4C. We further estimated the \nstrength of the genetic predictors of the 15 atherosclerosis-related diseases and risk factors \nusing F-statistics. To further validate the directionality of the analysis, we conducted Steiger \nfiltering analysis (78) of the four selected sclerostin instruments on the 15 atherosclerosis-\nrelated diseases and risk factors.  \n \nAll MR analyses were conducted using the MendelianRandomization R package (79) and \nTwoSampleMR R package (github.com/MRCIEU/TwoSampleMR v0.5.6)(80). Results were \nplotted as forest plots using code derived from the ggplot2 R package. \n \n  \n . CC-BY 4.0 International licenseIt is made available under a \n is the author/funder, who has granted medRxiv a license to display the preprint in perpetuity. (which was not certified by peer review)\nThe copyright holder for this preprint this version posted June 16, 2022. ; https://doi.org/10.1101/2022.06.13.22275915doi: medRxiv preprint \n\n \n23 \n \n \n \nFIGURE LEGENDS \n \nFigure 1. Summary of the design and results of the current study. This study included \nfour major components: (1) meta-analysis of genome-wide association study of circulating \nsclerostin; (2) signal genetic trait analysis and functional annotation of the top sclersotin \nsignals; (3) Mendelian randomizaiton and genetic correlation analysis of sclerostin on 15 \natherosclerosis-related diseases and risk factors traits; (4) bidirectional Mendelian \nrandomization analysis of 15 atherosclerosis-related diseases and risk factors on sclerostin.  \n \nFigure 2. Regional plot for the B4GLANT3 (A), SOST (B), SERPINA1 (C) and RIN3 (D) \nregions. Description of the regulation elements listed in Supplementary Table 13. \n \nFigure 3. Genetic effect of sclerostin-associated SNPs on eBMD and fracture. (A) \nGenetic effects of four variants on sclerostin, fracture, and eBMD. The alleles presented in \nthe plot are the sclerostin-lowering alleles. Different colour refers to the three traits been \nplotted. (B) Genetic effect of three sclerostin variants on eBMD and fracture scaled to \nstandard deviation unit of sclerostin reduction.  \n \nFigure 4. Causal effects of circulating sclerostin inhibition on hypertension, type 2 \ndiabetes and coronary artery calcification using cis-only and cis+trans genetic \npredictors of sclerostin. (A) causal effect of sclerostin inhibition on hypertension risk; (B) \ncausal effect of sclerostin inhibition on type 2 diabetes risk; (C) causal effect of sclerostin \ninhibition on coronary artery calcification.  \n \nFigure 5. Bidirectional causal effects between circulating sclerostin and five lipids traits. \n(A) causal effect of sclerostin inhibition on five lipid traits using cis-only and cis+trans \ninstruments; (B) causal effect of five lipid traits on sclerostin using univariate (UVMR) and \nmultivariable Mendelian randomization (MVMR).  \n \n  \n . CC-BY 4.0 International licenseIt is made available under a \n is the author/funder, who has granted medRxiv a license to display the preprint in perpetuity. (which was not certified by peer review)\nThe copyright holder for this preprint this version posted June 16, 2022. ; https://doi.org/10.1101/2022.06.13.22275915doi: medRxiv preprint \n\n \n24 \n \nACKNOWLEDGEMENTS \nWe are extremely grateful to all the families who took part in the ALSPAC study, the \nmidwives for their help in recruiting them, and the whole ALSPAC team, which includes \ninterviewers, computer and laboratory technicians, clerical workers, research scientists, \nvolunteers, managers, receptionists and nurses. ALSPAC data collection was supported by \nthe Wellcome Trust (grants WT092830M; WT088806; WT102215/2/13/2), UK Medical \nResearch Council (G1001357), and University of Bristol. The UK Medical Research Council \nand the Wellcome Trust (ref: 102215/2/13/2) and the University of Bristol provide core \nsupport for ALSPAC. GDS works in the Medical Research Council Integrative Epidemiology \nUnit at the University of Bristol MC_UU_00011/1. The Osteoarthritis Initiative (OAI) is a \npublic-private partnership comprised of five contracts (N01-AR-2–2258; N01-AR-2–2259; \nN01-AR-2–2260; N01-AR-2–2261; N01-AR-2–2262) funded by the NIH. Sclerostin \nmeasurement in the OAI was funded by the Wellcome Trust (ref 20378/Z/16/Z) and analyses \nwere supported by NIH R01AR075356 and NIH P30DK072488. The Trøndelag Health \nStudy (HUNT) is a collaboration between HUNT Research Centre (Faculty of Medicine and \nHealth Sciences, NTNU, Norwegian University of Science and Technology), Trøndelag \nCounty Council, Central Norway Regional Health Authority, and the Norwegian Institute of \nPublic Health. The genotyping in HUNT was financed by the National Institutes of Health; \nUniversity of Michigan; the Research Council of Norway; the Liaison Committee for \nEducation, Research and Innovation in Central Norway; and the Joint Research Committee \nbetween St Olavs hospital and the Faculty of Medicine and Health Sciences, NTNU. The \ngenetic investigations of the HUNT Study is a collaboration between researchers from the \nK.G. Jebsen Center for Genetic Epidemiology, NTNU and the University of Michigan \nMedical School and the University of Michigan School of Public Health. The K.G. Jebsen \nCenter for Genetic Epidemiology is financed by Stiftelsen Kristian Gerhard Jebsen; Faculty \nof Medicine and Health Sciences, NTNU, Norway. S.W.v.d.L is funded through EU H2020 \nTO_AITION (grant number: 848146). We are thankful for the support of the Netherlands \nCardioVascular Research Initiative of the Netherlands Heart Foundation (CVON 2011/B019 \nand CVON 2017-20: Generating the best evidence-based pharmaceutical targets for \natherosclerosis [GENIUS I&II]), the ERA-CVD program ‘druggable-MI-targets’ (grant \nnumber: 01KL1802), and the Leducq Fondation ‘PlaqOmics’. Dr. Rajeev Malhotra was \nsupported by the National Heart, Lung, and Blood Institute (R01HL142809 and \nR01HL159514), the American Heart Association (18TPA34230025), and the Wild Family \nFoundation. J.P.K is funded by a National Health and Medical Research Council (Australia) \nInvestigator grant (GNT1177938). Paul S. de Vries and Patricia A. Peyser were supported by \nNational Heart, Lung and Blood Institute (NHLBI) grant number R01HL146860. \nInfrastructure for the CHARGE Consortium was supported in part by the NHLBI grant \nR01HL105756. \n \nDisclosures \nDr. Sander W. van der Laan has received Roche funding for unrelated work. \n  \n . CC-BY 4.0 International licenseIt is made available under a \n is the author/funder, who has granted medRxiv a license to display the preprint in perpetuity. (which was not certified by peer review)\nThe copyright holder for this preprint this version posted June 16, 2022. ; https://doi.org/10.1101/2022.06.13.22275915doi: medRxiv preprint \n\n \n25 \n \nREFERENCES \n1. E. M. Lewiecki, T. Blicharski, S. Goemaere, K. Lippuner, P. D. Meisner, P. D. Miller, A. \nMiyauchi, J. Maddox, L. Chen, S. Horlait, A Phase III Randomized Placebo-Controlled Trial \nto Evaluate Efficacy and Safety of Romosozumab in Men With Osteoporosis. J. Clin. \nEndocrinol. Metab. 103, 3183–3193 (2018). \n2. K. G. Saag, J. Petersen, A. Grauer, Romosozumab versus Alendronate and Fracture Risk in \nWomen with Osteoporosis. N. Engl. J. Med. 378, 195–196 (2018). \n3. F. Cosman, D. B. Crittenden, J. D. Adachi, N. Binkley, E. Czerwinski, S. Ferrari, L. C. \nHofbauer, E. Lau, E. M. Lewiecki, A. Miyauchi, C. A. F. Zerbini, C. E. Milmont, L. Chen, J. \nMaddox, P. D. Meisner, C. Libanati, A. Grauer, Romosozumab Treatment in Postmenopausal \nWomen with Osteoporosis. N. Engl. J. Med. 375, 1532–1543 (2016). \n4. I. R. Reid, A. M. Horne, B. Mihov, A. Stewart, E. Garratt, S. Bastin, G. D. Gamble, \nEffects of Zoledronate on Cancer, Cardiac Events, and Mortality in Osteopenic Older \nWomen. J. Bone Miner. Res. 35, 20–27 (2020). \n5. S. R. Cummings, L.-Y. Lui, R. Eastell, I. E. Allen, Association Between Drug Treatments \nfor Patients With Osteoporosis and Overall Mortality Rates: A Meta-analysis. JAMA Intern. \nMed. 179, 1491–1500 (2019). \n6. A. De Maré, B. Opdebeeck, E. Neven, P. C. D’Haese, A. Verhulst, Sclerostin Protects \nAgainst Vascular Calcification Development in Mice. J. Bone Miner. Res. (2022), \ndoi:10.1002/jbmr.4503. \n7. M. Frysz, I. Gergei, H. Scharnagl, G. D. Smith, J. Zheng, D. A. Lawlor, M. Herrmann, W. \nMaerz, J. H. Tobias, Circulating sclerostin levels are positively related to coronary artery \ndisease severity and related risk factors. J. Bone Miner. Res. (2021), doi:10.1002/jbmr.4467. \n8. P. Garnero, E. Sornay-Rendu, F. Munoz, O. Borel, R. D. Chapurlat, Association of serum \nsclerostin with bone mineral density, bone turnover, steroid and parathyroid hormones, and \nfracture risk in postmenopausal women: the OFELY study. Osteoporos. Int. 24, 489–494 \n(2013). \n9. G. D. Smith, S. Ebrahim, Data dredging, bias, or confounding. BMJ 325, 1437–1438 \n(2002). \n10. G. Davey Smith, S. Ebrahim, ‘Mendelian randomization’: can genetic epidemiology \ncontribute to understanding environmental determinants of disease? Int. J. Epidemiol. 32, 1–\n22 (2003). \n11. R. C. Richmond, G. Davey Smith, Mendelian Randomization: Concepts and Scope. Cold \nSpring Harb. Perspect. Med. 12 (2022), doi:10.1101/cshperspect.a040501. \n12. J. Bovijn, K. Krebs, C.-Y. Chen, R. Boxall, J. C. Censin, T. Ferreira, S. L. Pulit, C. A. \nGlastonbury, S. Laber, I. Y. Millwood, K. Lin, L. Li, Z. Chen, L. Milani, G. D. Smith, R. G. \nWalters, R. Mägi, B. M. Neale, C. M. Lindgren, M. V. Holmes, Evaluating the \ncardiovascular safety of sclerostin inhibition using evidence from meta-analysis of clinical \ntrials and human genetics. Sci. Transl. Med. 12 (2020), doi:10.1126/scitranslmed.aay6570. \n . CC-BY 4.0 International licenseIt is made available under a \n is the author/funder, who has granted medRxiv a license to display the preprint in perpetuity. (which was not certified by peer review)\nThe copyright holder for this preprint this version posted June 16, 2022. ; https://doi.org/10.1101/2022.06.13.22275915doi: medRxiv preprint \n\n \n26 \n \n13. Committee for Medicinal Products for Human Use (CHMP), Assessment report for \nromosozumab. European Medicines Agent (2019) (available at \nhttps://www.ema.europa.eu/en/documents/assessment-report/evenity-epar-public-assessment-\nreport_en.pdf). \n14. G. Holdsworth, J. R. Staley, P. Hall, I. van Koeverden, C. Vangjeli, R. Okoye, R. W. \nBoyce, J. R. Turk, M. Armstrong, A. Wolfreys, G. Pasterkamp, Sclerostin Downregulation \nGlobally by Naturally Occurring Genetic Variants, or Locally in Atherosclerotic Plaques, \nDoes Not Associate With Cardiovascular Events in Humans. J. Bone Miner. Res. (2021), \ndoi:10.1002/jbmr.4287. \n15. J. Zheng, W. Maerz, I. Gergei, M. Kleber, C. Drechsler, C. Wanner, V. Brandenburg, S. \nReppe, K. Gautvik, C. Medina-Gomez, E. Shevroja, A. Gilly, Y.-C. Park, G. Dedoussis, E. \nZeggini, M. Lorentzon, P. Henning, U. Lerne, K. Nilsson, S. Moverare-Skrtic, D. Baird, L. \nFalk, A. Groom, T. Capellini, E. Grundberg, M. Nethander, C. Ohlsson, G. D. Smith, J. \nTobias, Genome-wide mapping identifies beta-1,4-N-acetyl-galactosaminyl-transferase as a \nnovel determinant of sclerostin levels and bone mineral density. J. Bone Miner. Res. (2018). \n16. G. Davey Smith, G. Hemani, Mendelian randomization: genetic anchors for causal \ninference in epidemiological studies. Hum. Mol. Genet. 23, R89-98 (2014). \n17. E. Sanderson, M. M. Glymour, M. V. Holmes, H. Kang, J. Morrison, M. R. Munafò, T. \nPalmer, C. Mary Schooling, C. Wallace, Q. Zhao, G. D. Smith, Mendelian randomization. \nNature Reviews Methods Primers 2, 1–1 (2022). \n18. M. V. Holmes, T. G. Richardson, B. A. Ference, N. M. Davies, G. Davey Smith, \nIntegrating genomics with biomarkers and therapeutic targets to invigorate cardiovascular \ndrug development. Nat. Rev. Cardiol. 18, 435–453 (2021). \n19. M. Pietzner, E. Wheeler, J. Carrasco-Zanini, A. Cortes, M. Koprulu, M. A. Wörheide, E. \nOerton, J. Cook, I. D. Stewart, N. D. Kerrison, J. Luan, J. Raffler, M. Arnold, W. Arlt, S. \nO’Rahilly, G. Kastenmüller, E. R. Gamazon, A. D. Hingorani, R. A. Scott, N. J. Wareham, C. \nLangenberg, Mapping the proteo-genomic convergence of human diseases. Science , \neabj1541 (2021). \n20. E. Di Angelantonio, S. G. Thompson, S. Kaptoge, C. Moore, M. Walker, J. Armitage, W. \nH. Ouwehand, D. J. Roberts, J. Danesh, INTERVAL Trial Group, Efficiency and safety of \nvarying the frequency of whole blood donation (INTERVAL): a randomised trial of 45 000 \ndonors. Lancet 390, 2360–2371 (2017). \n21. S. Krokstad, A. Langhammer, K. Hveem, T. L. Holmen, K. Midthjell, T. R. Stene, G. \nBratberg, J. Heggland, J. Holmen, Cohort Profile: the HUNT Study, Norway. Int. J. \nEpidemiol. 42, 968–977 (2013). \n22. B. O. Åsvold, A. Langhammer, T. A. Rehn, G. Kjelvik, T. V. Grøntvedt, E. P. Sørgjerd, \nJ. S. Fenstad, O. Holmen, M. C. Stuifbergen, S. A. A. Vikjord, B. M. Brumpton, H. K. \nSkjellegrind, P. Thingstad, E. R. Sund, G. Selbæk, P. J. Mork, V. Rangul, K. Hveem, M. \nNæss, S. Krokstad, Cohort profile update: The HUNT study, Norway. bioRxiv (2021), \ndoi:10.1101/2021.10.12.21264858. \n . CC-BY 4.0 International licenseIt is made available under a \n is the author/funder, who has granted medRxiv a license to display the preprint in perpetuity. (which was not certified by peer review)\nThe copyright holder for this preprint this version posted June 16, 2022. ; https://doi.org/10.1101/2022.06.13.22275915doi: medRxiv preprint \n\n \n27 \n \n23. B. M. Brumpton, S. Graham, I. Surakka, A. H. Skogholt, M. Løset, L. G. Fritsche, B. \nWolford, W. Zhou, J. B. Nielsen, O. L. Holmen, M. E. Gabrielsen, L. Thomas, L. Bhatta, H. \nRasheed, H. Zhang, H. M. Kang, W. Hornsby, M. R. Moksnes, E. Coward, M. Melbye, G. F. \nGiskeødegård, J. Fenstad, S. Krokstad, M. Næss, A. Langhammer, M. Boehnke, G. R. \nAbecasis, B. O. Åsvold, K. Hveem, C. J. Willer, The HUNT Study: a population-based \ncohort for genetic research. bioRxiv (2021), doi:10.1101/2021.12.23.21268305. \n24. G. Lester, The Osteoarthritis Initiative: A NIH Public-Private Partnership. HSS J. 8, 62–\n63 (2012). \n25. L. K. Conlin, B. D. Thiel, C. G. Bonnemann, L. Medne, L. M. Ernst, E. H. Zackai, M. A. \nDeardorff, I. D. Krantz, H. Hakonarson, N. B. Spinner, Mechanisms of mosaicism, \nchimerism and uniparental disomy identified by single nucleotide polymorphism array \nanalysis. Hum. Mol. Genet. 19, 1263–1275 (2010). \n26. D. A. Peiffer, J. M. Le, F. J. Steemers, W. Chang, T. Jenniges, F. Garcia, K. Haden, J. Li, \nC. A. Shaw, J. Belmont, S. W. Cheung, R. M. Shen, D. L. Barker, K. L. Gunderson, High-\nresolution genomic profiling of chromosomal aberrations using Infinium whole-genome \ngenotyping. Genome Res. 16, 1136–1148 (2006). \n27. C. C. Laurie, K. F. Doheny, D. B. Mirel, E. W. Pugh, L. J. Bierut, T. Bhangale, F. \nBoehm, N. E. Caporaso, M. C. Cornelis, H. J. Edenberg, S. B. Gabriel, E. L. Harris, F. B. Hu, \nK. B. Jacobs, P. Kraft, M. T. Landi, T. Lumley, T. A. Manolio, C. McHugh, I. Painter, J. \nPaschall, J. P. Rice, K. M. Rice, X. Zheng, B. S. Weir, GENEVA Investigators, Quality \ncontrol and quality assurance in genotypic data for genome-wide association studies. Genet. \nEpidemiol. 34, 591–602 (2010). \n28. B. R. Winkelmann, W. März, B. O. Boehm, R. Zotz, J. Hager, P. Hellstern, J. Senges, \nLURIC Study Group (LUdwigshafen RIsk and Cardiovascular Health), Rationale and design \nof the LURIC study--a resource for functional genomics, pharmacogenomics and long-term \nprognosis of cardiovascular disease. Pharmacogenomics 2, S1-73 (2001). \n29. T. F. M. Andlauer, D. Buck, G. Antony, A. Bayas, L. Bechmann, A. Berthele, A. Chan, \nC. Gasperi, R. Gold, C. Graetz, J. Haas, M. Hecker, C. Infante-Duarte, M. Knop, T. Kümpfel, \nV. Limmroth, R. A. Linker, V. Loleit, F. Luessi, S. G. Meuth, M. Mühlau, S. Nischwitz, F. \nPaul, M. Pütz, T. Ruck, A. Salmen, M. Stangel, J.-P. Stellmann, K. H. Stürner, B. \nTackenberg, F. Then Bergh, H. Tumani, C. Warnke, F. Weber, H. Wiendl, B. Wildemann, U. \nK. Zettl, U. Ziemann, F. Zipp, J. Arloth, P. Weber, M. Radivojkov-Blagojevic, M. O. \nScheinhardt, T. Dankowski, T. Bettecken, P. Lichtner, D. Czamara, T. Carrillo-Roa, E. B. \nBinder, K. Berger, L. Bertram, A. Franke, C. Gieger, S. Herms, G. Homuth, M. Ising, K.-H. \nJöckel, T. Kacprowski, S. Kloiber, M. Laudes, W. Lieb, C. M. Lill, S. Lucae, T. Meitinger, S. \nMoebus, M. Müller-Nurasyid, M. M. Nöthen, A. Petersmann, R. Rawal, U. Schminke, K. \nStrauch, H. Völzke, M. Waldenberger, J. Wellmann, E. Porcu, A. Mulas, M. Pitzalis, C. \nSidore, I. Zara, F. Cucca, M. Zoledziewska, A. Ziegler, B. Hemmer, B. Müller-Myhsok, \nNovel multiple sclerosis susceptibility loci implicated in epigenetic regulation. Sci Adv 2, \ne1501678 (2016). \n30. F. Mihalache, A. Höblinger, F. Grünhage, M. Krawczyk, B. C. Gärtner, M. Acalovschi, \nT. Sauerbruch, F. Lammert, V. Zimmer, Heterozygosity for the alpha1-antitrypsin Z allele \nmay confer genetic risk of cholangiocarcinoma. Aliment. Pharmacol. Ther. 33, 389–394 \n(2011). \n . CC-BY 4.0 International licenseIt is made available under a \n is the author/funder, who has granted medRxiv a license to display the preprint in perpetuity. (which was not certified by peer review)\nThe copyright holder for this preprint this version posted June 16, 2022. ; https://doi.org/10.1101/2022.06.13.22275915doi: medRxiv preprint \n\n \n28 \n \n31. J. P. Kemp, C. Medina-Gomez, K. Estrada, B. St Pourcain, D. H. M. Heppe, N. M. \nWarrington, L. Oei, S. M. Ring, C. J. Kruithof, N. J. Timpson, L. E. Wolber, S. Reppe, K. \nGautvik, E. Grundberg, B. Ge, B. van der Eerden, J. van de Peppel, M. A. Hibbs, C. L. \nAckert-Bicknell, K. Choi, D. L. Koller, M. J. Econs, F. M. K. Williams, T. Foroud, M. C. \nZillikens, C. Ohlsson, A. Hofman, A. G. Uitterlinden, G. Davey Smith, V. W. V. Jaddoe, J. \nH. Tobias, F. Rivadeneira, D. M. Evans, Phenotypic dissection of bone mineral density \nreveals skeletal site specificity and facilitates the identification of novel loci in the genetic \nregulation of bone mass attainment. PLoS Genet. 10, e1004423 (2014). \n32. F. Rivadeneira, U. Styrkársdottir, K. Estrada, B. V. Halldórsson, Y.-H. Hsu, J. B. \nRichards, M. C. Zillikens, F. K. Kavvoura, N. Amin, Y. S. Aulchenko, L. A. Cupples, P. \nDeloukas, S. Demissie, E. Grundberg, A. Hofman, A. Kong, D. Karasik, J. B. van Meurs, B. \nOostra, T. Pastinen, H. A. P. Pols, G. Sigurdsson, N. Soranzo, G. Thorleifsson, U. \nThorsteinsdottir, F. M. K. Williams, S. G. Wilson, Y. Zhou, S. H. Ralston, C. M. van Duijn, \nT. Spector, D. P. Kiel, K. Stefansson, J. P. A. Ioannidis, A. G. Uitterlinden, Genetic Factors \nfor Osteoporosis (GEFOS) Consortium, Twenty bone-mineral-density loci identified by \nlarge-scale meta-analysis of genome-wide association studies. Nat. Genet. 41, 1199–1206 \n(2009). \n33. R. Jemtland, M. Holden, S. Reppe, O. K. Olstad, F. P. Reinholt, V. T. Gautvik, H. \nRefvem, A. Frigessi, B. Houston, K. M. Gautvik, Molecular disease map of bone \ncharacterizing the postmenopausal osteoporosis phenotype. J. Bone Miner. Res. 26, 1793–\n1801 (2011). \n34. J. Ernst, M. Kellis, ChromHMM: automating chromatin-state discovery and \ncharacterization. Nat. Methods 9, 215–216 (2012). \n35. A. P. Boyle, E. L. Hong, M. Hariharan, Y. Cheng, M. A. Schaub, M. Kasowski, K. J. \nKarczewski, J. Park, B. C. Hitz, S. Weng, J. M. Cherry, M. Snyder, Annotation of functional \nvariation in personal genomes using RegulomeDB. Genome Res. 22, 1790–1797 (2012). \n36. GTEx Consortium, The GTEx Consortium atlas of genetic regulatory effects across \nhuman tissues. Science 369, 1318–1330 (2020). \n37. J. Støy, U. Kampmann, A. Mengel, N. E. Magnusson, N. Jessen, N. Grarup, J. Rungby, \nH. Stødkilde-Jørgensen, I. Brandslund, C. Christensen, T. Hansen, O. Pedersen, N. Møller, \nReduced CD300LG mRNA tissue expression, increased intramyocellular lipid content and \nimpaired glucose metabolism in healthy male carriers of Arg82Cys in CD300LG: a novel \ngenometabolic cross-link between CD300LG and common metabolic phenotypes. BMJ Open \nDiabetes Research and Care 3, e000095 (2015). \n38. N. J. Timpson, B. G. Nordestgaard, R. M. Harbord, J. Zacho, T. M. Frayling, A. Tybjærg-\nHansen, G. D. Smith, C-reactive protein levels and body mass index: elucidating direction of \ncausation through reciprocal Mendelian randomization. Int. J. Obes. 35, 300–308 (2011). \n39. A. R. van Rosendael, I. J. van den Hoogen, U. Gianni, X. Ma, S. W. Tantawy, A. M. Bax, \nY. Lu, D. Andreini, M. H. Al-Mallah, M. J. Budoff, F. Cademartiri, K. Chinnaiyan, J. H. \nChoi, E. Conte, H. Marques, P. de Araújo Gonçalves, I. Gottlieb, M. Hadamitzky, J. A. \nLeipsic, E. Maffei, G. Pontone, S. Shin, Y.-J. Kim, B. K. Lee, E. J. Chun, J. M. Sung, S.-E. \nLee, R. Virmani, H. Samady, Y. Sato, P. H. Stone, D. S. Berman, J. Narula, R. Blankstein, J. \nK. Min, F. Y. Lin, L. J. Shaw, J. J. Bax, H.-J. Chang, Association of Statin Treatment With \n . CC-BY 4.0 International licenseIt is made available under a \n is the author/funder, who has granted medRxiv a license to display the preprint in perpetuity. (which was not certified by peer review)\nThe copyright holder for this preprint this version posted June 16, 2022. ; https://doi.org/10.1101/2022.06.13.22275915doi: medRxiv preprint \n\n \n29 \n \nProgression of Coronary Atherosclerotic Plaque Composition. JAMA Cardiol (2021), \ndoi:10.1001/jamacardio.2021.3055. \n40. G. S. Gulsin, A. J. Moss, Coronary artery calcium paradox and physical activity. Heart \n107, 1686–1687 (2021). \n41. K.-C. Sung, Y. S. Hong, J.-Y. Lee, S.-J. Lee, Y. Chang, S. Ryu, D. Zhao, J. Cho, E. \nGuallar, J. A. C. Lima, Physical activity and the progression of coronary artery calcification. \nHeart (2021), doi:10.1136/heartjnl-2021-319346. \n42. G. L. Galea, L. E. Lanyon, J. S. Price, Sclerostin’s role in bone’s adaptive response to \nmechanical loading. Bone 96, 38–44 (2017). \n43. R. Koos, V. Brandenburg, A. H. Mahnken, R. Schneider, G. Dohmen, R. Autschbach, N. \nMarx, R. Kramann, Sclerostin as a potential novel biomarker for aortic valve calcification: an \nin-vivo and ex-vivo study. J. Heart Valve Dis. 22, 317–325 (2013). \n44. D. Zhu, N. C. W. Mackenzie, J. L. Millán, C. Farquharson, V. E. MacRae, The \nappearance and modulation of osteocyte marker expression during calcification of vascular \nsmooth muscle cells. PLoS One 6, e19595 (2011). \n45. X. Li, Y. Zhang, H. Kang, W. Liu, P. Liu, J. Zhang, S. E. Harris, D. Wu, Sclerostin binds \nto LRP5/6 and antagonizes canonical Wnt signaling. J. Biol. Chem. 280, 19883–19887 \n(2005). \n46. A. Catalano, F. Bellone, N. Morabito, F. Corica, Sclerostin and Vascular \nPathophysiology. Int. J. Mol. Sci. 21 (2020), doi:10.3390/ijms21134779. \n47. S. Y. Lim, M. B. Bolster, Profile of romosozumab and its potential in the management of \nosteoporosis. Drug Des. Devel. Ther. 11, 1221–1231 (2017). \n48. W. Gan, R. J. Clarke, A. Mahajan, B. Kulohoma, H. Kitajima, N. R. Robertson, N. W. \nRayner, R. G. Walters, M. V. Holmes, Z. Chen, M. I. McCarthy, Bone mineral density and \nrisk of type 2 diabetes and coronary heart disease: A Mendelian randomization study. \nWellcome Open Res 2, 68 (2017). \n49. Committee for Medicinal Products for Human Use (CHMP), Recommendation to restrict \nthe use of Protelos/Osseor (strontium ranelate). European Medicines Agent (2013) (available \nat https://www.ema.europa.eu/en/documents/press-release/recommendation-restrict-use-\nprotelos/osseor-strontium-ranelate_en.pdf). \n50. O. M. E. Albagha, S. E. Wani, M. R. Visconti, N. Alonso, K. Goodman, M. L. Brandi, T. \nCundy, P. Y. J. Chung, R. Dargie, J.-P. Devogelaer, A. Falchetti, W. D. Fraser, L. Gennari, F. \nGianfrancesco, M. J. Hooper, W. Van Hul, G. Isaia, G. C. Nicholson, R. Nuti, S. Papapoulos, \nJ. del P. Montes, T. Ratajczak, S. L. Rea, D. Rendina, R. Gonzalez-Sarmiento, M. Di Stefano, \nL. C. Ward, J. P. Walsh, S. H. Ralston, Genetic Determinants of Paget’s Disease (GDPD) \nConsortium, Genome-wide association identifies three new susceptibility loci for Paget’s \ndisease of bone. Nat. Genet. 43, 685–689 (2011). \n51. I. Tachmazidou, D. Süveges, J. L. Min, G. R. S. Ritchie, J. Steinberg, K. Walter, V. \nIotchkova, J. Schwartzentruber, J. Huang, Y. Memari, S. McCarthy, A. A. Crawford, C. \nBombieri, M. Cocca, A.-E. Farmaki, T. R. Gaunt, P. Jousilahti, M. N. Kooijman, B. Lehne, \n . CC-BY 4.0 International licenseIt is made available under a \n is the author/funder, who has granted medRxiv a license to display the preprint in perpetuity. (which was not certified by peer review)\nThe copyright holder for this preprint this version posted June 16, 2022. ; https://doi.org/10.1101/2022.06.13.22275915doi: medRxiv preprint \n\n \n30 \n \nG. Malerba, S. Männistö, A. Matchan, C. Medina-Gomez, S. J. Metrustry, A. Nag, I. Ntalla, \nL. Paternoster, N. W. Rayner, C. Sala, W. R. Scott, H. A. Shihab, L. Southam, B. St \nPourcain, M. Traglia, K. Trajanoska, G. Zaza, W. Zhang, M. S. Artigas, N. Bansal, M. Benn, \nZ. Chen, P. Danecek, W.-Y. Lin, A. Locke, J. Luan, A. K. Manning, A. Mulas, C. Sidore, A. \nTybjaerg-Hansen, A. Varbo, M. Zoledziewska, C. Finan, K. Hatzikotoulas, A. E. Hendricks, \nJ. P. Kemp, A. Moayyeri, K. Panoutsopoulou, M. Szpak, S. G. Wilson, M. Boehnke, F. \nCucca, E. Di Angelantonio, C. Langenberg, C. Lindgren, M. I. McCarthy, A. P. Morris, B. G. \nNordestgaard, R. A. Scott, M. D. Tobin, N. J. Wareham, SpiroMeta Consortium, GoT2D \nConsortium, P. Burton, J. C. Chambers, G. D. Smith, G. Dedoussis, J. F. Felix, O. H. Franco, \nG. Gambaro, P. Gasparini, C. J. Hammond, A. Hofman, V. W. V. Jaddoe, M. Kleber, J. S. \nKooner, M. Perola, C. Relton, S. M. Ring, F. Rivadeneira, V. Salomaa, T. D. Spector, O. \nStegle, D. Toniolo, A. G. Uitterlinden, arcOGEN Consortium, Understanding Society \nScientific Group, UK10K Consortium, I. Barroso, C. M. T. Greenwood, J. R. B. Perry, B. R. \nWalker, A. S. Butterworth, Y. Xue, R. Durbin, K. S. Small, N. Soranzo, N. J. Timpson, E. \nZeggini, Whole-Genome Sequencing Coupled to Imputation Discovers Genetic Signals for \nAnthropometric Traits. Am. J. Hum. Genet. 100, 865–884 (2017). \n52. J. Mbatchou, L. Barnard, J. Backman, A. Marcketta, J. A. Kosmicki, A. Ziyatdinov, C. \nBenner, C. O’Dushlaine, M. Barber, B. Boutkov, L. Habegger, M. Ferreira, A. Baras, J. Reid, \nG. Abecasis, E. Maxwell, J. Marchini, Computationally efficient whole-genome regression \nfor quantitative and binary traits. Nat. Genet. 53, 1097–1103 (2021). \n53. C. C. Chang, C. C. Chow, L. C. Tellier, S. Vattikuti, S. M. Purcell, J. J. Lee, Second-\ngeneration PLINK: rising to the challenge of larger and richer datasets. Gigascience 4, 7 \n(2015). \n54. T. W. Winkler, F. R. Day, D. C. Croteau-Chonka, A. R. Wood, A. E. Locke, R. Mägi, T. \nFerreira, T. Fall, M. Graff, A. E. Justice, J. Luan, S. Gustafsson, J. C. Randall, S. Vedantam, \nT. Workalemahu, T. O. Kilpeläinen, A. Scherag, T. Esko, Z. Kutalik, I. M. Heid, R. J. F. \nLoos, Genetic Investigation of Anthropometric Traits (GIANT) Consortium, Quality control \nand conduct of genome-wide association meta-analyses. Nat. Protoc. 9, 1192–1212 (2014). \n55. C. J. Willer, Y. Li, G. R. Abecasis, METAL: fast and efficient meta-analysis of \ngenomewide association scans. Bioinformatics 26, 2190–2191 (2010). \n56. Y. Li, C. J. Willer, J. Ding, P. Scheet, G. R. Abecasis, MaCH: using sequence and \ngenotype data to estimate haplotypes and unobserved genotypes. Genet. Epidemiol. 34, 816–\n834 (2010). \n57. B. Howie, C. Fuchsberger, M. Stephens, J. Marchini, G. R. Abecasis, Fast and accurate \ngenotype imputation in genome-wide association studies through pre-phasing. Nat. Genet. \n44, 955–959 (2012). \n58. R. Mägi, A. P. Morris, GWAMA: software for genome-wide association meta-analysis. \nBMC Bioinformatics 11, 288 (2010). \n59. J. Yang, T. Ferreira, A. P. Morris, S. E. Medland, Genetic Investigation of \nANthropometric Traits (GIANT) Consortium, DIAbetes Genetics Replication And Meta-\nanalysis (DIAGRAM) Consortium, P. A. F. Madden, A. C. Heath, N. G. Martin, G. W. \nMontgomery, M. N. Weedon, R. J. Loos, T. M. Frayling, M. I. McCarthy, J. N. Hirschhorn, \nM. E. Goddard, P. M. Visscher, Conditional and joint multiple-SNP analysis of GWAS \n . CC-BY 4.0 International licenseIt is made available under a \n is the author/funder, who has granted medRxiv a license to display the preprint in perpetuity. (which was not certified by peer review)\nThe copyright holder for this preprint this version posted June 16, 2022. ; https://doi.org/10.1101/2022.06.13.22275915doi: medRxiv preprint \n\n \n31 \n \nsummary statistics identifies additional variants influencing complex traits. Nat. Genet. 44, \n369–75, S1-3 (2012). \n60. C. Giambartolomei, D. Vukcevic, E. E. Schadt, L. Franke, A. D. Hingorani, C. Wallace, \nV. Plagnol, Bayesian test for colocalisation between pairs of genetic association studies using \nsummary statistics. PLoS Genet. 10, e1004383 (2014). \n61. K. Watanabe, E. Taskesen, A. van Bochoven, D. Posthuma, Functional mapping and \nannotation of genetic associations with FUMA. Nat. Commun. 8, 1–11 (2017). \n62. S. Koplev, M. Seldin, K. Sukhavasi, R. Ermel, S. Pang, L. Zeng, S. Bankier, A. Di Narzo, \nH. Cheng, V. Meda, A. Ma, H. Talukdar, A. Cohain, L. Amadori, C. Argmann, S. M. Houten, \nO. Franzén, G. Mocci, O. A. Meelu, K. Ishikawa, C. Whatling, A. Jain, R. K. Jain, L.-M. \nGan, C. Giannarelli, P. Roussos, K. Hao, H. Schunkert, T. Michoel, A. Ruusalepp, E. E. \nSchadt, J. C. Kovacic, A. J. Lusis, J. L. M. Björkegren, A mechanistic framework for \ncardiometabolic and coronary artery diseases. Nature Cardiovascular Research 1, 85–100 \n(2022). \n63. B. K. Bulik-Sullivan, P.-R. Loh, H. K. Finucane, S. Ripke, J. Yang, Schizophrenia \nWorking Group of the Psychiatric Genomics Consortium, N. Patterson, M. J. Daly, A. L. \nPrice, B. M. Neale, LD Score regression distinguishes confounding from polygenicity in \ngenome-wide association studies. Nat. Genet. 47, 291–295 (2015). \n64. J. Zheng, A. M. Erzurumluoglu, B. L. Elsworth, J. P. Kemp, L. Howe, P. C. Haycock, G. \nHemani, K. Tansey, C. Laurin, Early Genetics and Lifecourse Epidemiology (EAGLE) \nEczema Consortium, B. S. Pourcain, N. M. Warrington, H. K. Finucane, A. L. Price, B. K. \nBulik-Sullivan, V. Anttila, L. Paternoster, T. R. Gaunt, D. M. Evans, B. M. Neale, LD Hub: a \ncentralized database and web interface to perform LD score regression that maximizes the \npotential of summary level GWAS data for SNP heritability and genetic correlation analysis. \nBioinformatics 33, 272–279 (2017). \n65. D. A. Lawlor, R. M. Harbord, J. A. C. Sterne, N. Timpson, G. Davey Smith, Mendelian \nrandomization: using genes as instruments for making causal inferences in epidemiology. \nStat. Med. 27, 1133–1163 (2008). \n66. J. Zheng, Richardson T.R., Millard L., Hemani G., Raistrick C., Vilhjalmsson B., \nHaycock P.C., Gaunt T.R., PhenoSpD: an integrated toolkit for phenotypic correlation \nestimation and multiple testing correction using GWAS summary statistics. GigaScience \n(2018) doi:10.1101/148627. \n67. A. Mahajan, D. Taliun, M. Thurner, N. R. Robertson, J. M. Torres, N. W. Rayner, A. J. \nPayne, V. Steinthorsdottir, R. A. Scott, N. Grarup, J. P. Cook, E. M. Schmidt, M. Wuttke, C. \nSarnowski, R. Mägi, J. Nano, C. Gieger, S. Trompet, C. Lecoeur, M. H. Preuss, B. P. Prins, \nX. Guo, L. F. Bielak, J. E. Below, D. W. Bowden, J. C. Chambers, Y. J. Kim, M. C. Y. Ng, \nL. E. Petty, X. Sim, W. Zhang, A. J. Bennett, J. Bork-Jensen, C. M. Brummett, M. Canouil, \nK.-U. Ec Kardt, K. Fischer, S. L. R. Kardia, F. Kronenberg, K. Läll, C.-T. Liu, A. E. Locke, \nJ. Luan, I. Ntalla, V. Nylander, S. Schönherr, C. Schurmann, L. Yengo, E. P. Bottinger, I. \nBrandslund, C. Christensen, G. Dedoussis, J. C. Florez, I. Ford, O. H. Franco, T. M. Frayling, \nV. Giedraitis, S. Hackinger, A. T. Hattersley, C. Herder, M. A. Ikram, M. Ingelsson, M. E. \nJørgensen, T. Jørgensen, J. Kriebel, J. Kuusisto, S. Ligthart, C. M. Lindgren, A. Linneberg, \nV. Lyssenko, V. Mamakou, T. Meitinger, K. L. Mohlke, A. D. Morris, G. Nadkarni, J. S. \n . CC-BY 4.0 International licenseIt is made available under a \n is the author/funder, who has granted medRxiv a license to display the preprint in perpetuity. (which was not certified by peer review)\nThe copyright holder for this preprint this version posted June 16, 2022. ; https://doi.org/10.1101/2022.06.13.22275915doi: medRxiv preprint \n\n \n32 \n \nPankow, A. Peters, N. Sattar, A. Stanč áková, K. Strauch, K. D. Taylor, B. Thorand, G. \nThorleifsson, U. Thorsteinsdottir, J. Tuomilehto, D. R. Witte, J. Dupuis, P. A. Peyser, E. \nZeggini, R. J. F. Loos, P. Froguel, E. Ingelsson, L. Lind, L. Groop, M. Laakso, F. S. Collins, \nJ. W. Jukema, C. N. A. Palmer, H. Grallert, A. Metspalu, A. Dehghan, A. Köttgen, G. R. \nAbecasis, J. B. Meigs, J. I. Rotter, J. Marchini, O. Pedersen, T. Hansen, C. Langenberg, N. J. \nWareham, K. Stefansson, A. L. Gloyn, A. P. Morris, M. Boehnke, M. I. McCarthy, Fine-\nmapping type 2 diabetes loci to single-variant resolution using high-density imputation and \nislet-specific epigenome maps. Nat. Genet. 50, 1505–1513 (2018). \n68. R. Malik, M. Traylor, S. L. Pulit, S. Bevan, J. C. Hopewell, E. G. Holliday, W. Zhao, P. \nAbrantes, P. Amouyel, J. R. Attia, T. W. K. Battey, K. Berger, G. B. Boncoraglio, G. \nChauhan, Y.-C. Cheng, W.-M. Chen, R. Clarke, I. Cotlarciuc, S. Debette, G. J. Falcone, J. M. \nFerro, D. M. Gamble, A. Ilinca, S. J. Kittner, C. E. Kourkoulis, R. Lemmens, C. R. Levi, P. \nLichtner, A. Lindgren, J. Liu, J. F. Meschia, B. D. Mitchell, S. A. Oliveira, J. Pera, A. P. \nReiner, P. M. Rothwell, P. Sharma, A. Slowik, C. L. M. Sudlow, T. Tatlisumak, V. Thijs, A. \nM. Vicente, D. Woo, S. Seshadri, D. Saleheen, J. Rosand, H. S. Markus, B. B. Worrall, M. \nDichgans, ISGC Analysis Group, METASTROKE collaboration, Wellcome Trust Case \nControl Consortium 2 (WTCCC2), NINDS Stroke Genetics Network (SiGN), Low-frequency \nand common genetic variation in ischemic stroke: The METASTROKE collaboration. \nNeurology 86, 1217–1226 (2016). \n69. P. van der Harst, N. Verweij, Identification of 64 Novel Genetic Loci Provides an \nExpanded View on the Genetic Architecture of Coronary Artery Disease. Circ. Res. 122, \n433–443 (2018). \n70. J. A. Hartiala, Y. Han, Q. Jia, J. R. Hilser, P. Huang, J. Gukasyan, W. S. Schwartzman, Z. \nCai, S. Biswas, D.-A. Trégouët, N. L. Smith, INVENT Consortium, CHARGE Consortium \nHemostasis Working Group, GENIUS-CHD Consortium, M. Seldin, C. Pan, M. Mehrabian, \nA. J. Lusis, P. Bazeley, Y. V. Sun, C. Liu, A. A. Quyyumi, M. Scholz, J. Thiery, G. E. \nDelgado, M. E. Kleber, W. März, L. J. Howe, F. W. Asselbergs, M. van Vugt, G. J. \nVlachojannis, R. S. Patel, L.-P. Lyytikäinen, M. Kähönen, T. Lehtimäki, T. V. M. Nieminen, \nP. Kuukasjärvi, J. O. Laurikka, X. Chang, C.-K. Heng, R. Jiang, W. E. Kraus, E. R. Hauser, \nJ. F. Ferguson, M. P. Reilly, K. Ito, S. Koyama, Y. Kamatani, I. Komuro, Biobank Japan, L. \nK. Stolze, C. E. Romanoski, M. D. Khan, A. W. Turner, C. L. Miller, R. Aherrahrou, M. \nCivelek, L. Ma, J. L. M. Björkegren, S. R. Kumar, W. H. W. Tang, S. L. Hazen, H. Allayee, \nGenome-wide analysis identifies novel susceptibility loci for myocardial infarction. Eur. \nHeart J. 42, 919–933 (2021). \n71. M. Kavousi, M. M. Bos, H. J. Barnes, C. L. Lino Cardenas, D. Wong, C. J. O’Donnell, L. \nF. Bielak, P. A. Peyser, R. Malhotra, S. W. van der Laan, C. L. Miller, CHARGE Subclinical \nand Clinical Atherosclerosis Working Group, Multi-ancestry genome-wide analysis identifies \neffector genes and druggable pathways for coronary artery calcification (2022), \ndoi:10.1101/2022.05.02.22273844. \n72. J. P. Kemp, J. A. Morris, C. Medina-Gomez, V. Forgetta, N. M. Warrington, S. E. \nYoulten, J. Zheng, C. L. Gregson, E. Grundberg, K. Trajanoska, J. G. Logan, A. S. Pollard, P. \nC. Sparkes, E. J. Ghirardello, R. Allen, V. D. Leitch, N. C. Butterfield, D. Komla-Ebri, A.-T. \nAdoum, K. F. Curry, J. K. White, F. Kussy, K. M. Greenlaw, C. Xu, N. C. Harvey, C. \nCooper, D. J. Adams, C. M. T. Greenwood, M. T. Maurano, S. Kaptoge, F. Rivadeneira, J. H. \nTobias, P. I. Croucher, C. L. Ackert-Bicknell, J. H. D. Bassett, G. R. Williams, J. B. \nRichards, D. M. Evans, Identification of 153 new loci associated with heel bone mineral \n . CC-BY 4.0 International licenseIt is made available under a \n is the author/funder, who has granted medRxiv a license to display the preprint in perpetuity. (which was not certified by peer review)\nThe copyright holder for this preprint this version posted June 16, 2022. ; https://doi.org/10.1101/2022.06.13.22275915doi: medRxiv preprint \n\n \n33 \n \ndensity and functional involvement of GPC6 in osteoporosis. Nat. Genet. 49, 1468–1475 \n(2017). \n73. R. Malhotra, A. C. Mauer, C. L. Lino Cardenas, X. Guo, J. Yao, X. Zhang, F. Wunderer, \nA. V. Smith, Q. Wong, S. Pechlivanis, S.-J. Hwang, J. Wang, L. Lu, C. J. Nicholson, G. \nShelton, M. D. Buswell, H. J. Barnes, H. H. Sigurslid, C. Slocum, C. O. Rourke, D. K. Rhee, \nA. Bagchi, S. U. Nigwekar, E. S. Buys, C. Y. Campbell, T. Harris, M. Budoff, M. H. Criqui, \nJ. I. Rotter, A. D. Johnson, C. Song, N. Franceschini, S. Debette, U. Hoffmann, H. Kälsch, \nM. M. Nöthen, S. Sigurdsson, B. I. Freedman, D. W. Bowden, K.-H. Jöckel, S. Moebus, R. \nErbel, M. F. Feitosa, V. Gudnason, G. Thanassoulis, W. M. Zapol, M. E. Lindsay, D. B. \nBloch, W. S. Post, C. J. O’Donnell, HDAC9 is implicated in atherosclerotic aortic \ncalcification and affects vascular smooth muscle cell phenotype. Nat. Genet. 51, 1580–1587 \n(2019). \n74. S. Burgess, A. Butterworth, S. G. Thompson, Mendelian randomization analysis with \nmultiple genetic variants using summarized data. Genet. Epidemiol. 37, 658–665 (2013). \n75. J. Bowden, F. Del Greco M, C. Minelli, G. Davey Smith, N. A. Sheehan, J. R. Thompson, \nAssessing the suitability of summary data for two-sample Mendelian randomization analyses \nusing MR-Egger regression: the role of the I 2 statistic. Int. J. Epidemiol. 45, 1961–1974 \n(2016). \n76. A. Gkatzionis, S. Burgess, P. J. Newcombe, Statistical Methods for cis-Mendelian \nRandomization. arXiv (2021) (available at http://arxiv.org/abs/2101.04081). \n77. S. Burgess, S. G. Thompson, Multivariable Mendelian randomization: the use of \npleiotropic genetic variants to estimate causal effects. Am. J. Epidemiol. 181, 251–260 \n(2015). \n78. G. Hemani, K. Tilling, G. Davey Smith, Orienting the causal relationship between \nimprecisely measured traits using GWAS summary data. PLoS Genet. 13, e1007081 (2017). \n79. O. O. Yavorska, S. Burgess, MendelianRandomization: an R package for performing \nMendelian randomization analyses using summarized data. Int. J. Epidemiol. (2017), \ndoi:10.1093/ije/dyx034. \n80. G. Hemani, J. Zheng, B. Elsworth, K. H. Wade, V. Haberland, D. Baird, C. Laurin, S. \nBurgess, J. Bowden, R. Langdon, V. Y. Tan, J. Yarmolinsky, H. A. Shihab, N. J. Timpson, \nD. M. Evans, C. Relton, R. M. Martin, G. Davey Smith, T. R. Gaunt, P. C. Haycock, The \nMR-Base platform supports systematic causal inference across the human phenome. Elife 7 \n(2018), doi:10.7554/eLife.34408. \n  \n . CC-BY 4.0 International licenseIt is made available under a \n is the author/funder, who has granted medRxiv a license to display the preprint in perpetuity. (which was not certified by peer review)\nThe copyright holder for this preprint this version posted June 16, 2022. ; https://doi.org/10.1101/2022.06.13.22275915doi: medRxiv preprint \n\n \n34 \n \nTable 1. Study information of the cohorts involved in the sclerostin GWAS meta-analysis. \n \nCohort/Study N sclerostin Age (SD) Ancestry Assay details \nFenland 10708 48.6 (7.5) European SOMALogic \nINTERVAL 3301 43.4 (14.1) European SOMALogic \nHUNT 3532 64.8 (10.1) European SOMALogic \nOAI 4484 61.2 (9.2) European Chemiluminescent assay \nLURIC 1884 62.9 (10.7) European Diasorin \nZheng et al 10584 34.9 (4.5) European ELISA/TECO/OLINK \n \n  \n . CC-BY 4.0 International licenseIt is made available under a \n is the author/funder, who has granted medRxiv a license to display the preprint in perpetuity. (which was not certified by peer review)\nThe copyright holder for this preprint this version posted June 16, 2022. ; https://doi.org/10.1101/2022.06.13.22275915doi: medRxiv preprint \n\n \n35 \n \nTable 2. Meta-analysis results for loci that reached genome-wide significance (P < 5 × 10−8). \nLocus SNP EA OA  EAF GENE Cis/trans BETA SE P Q Q P  I2 \nchr1 50566286 rs61781020 A G 0.049 FAF1 Trans 0.101 0.018 2.57×10-8 17.825 0.058 0.439 \nchr2 229236796 rs4973180 T C 0.821 PID1 Trans 0.059 0.010 1.04×10-9 6.691 0.754 0 \nchr5 56813327 rs11960484 A G 0.351 MAP3K1 Trans -0.049 0.008 1.40×10-10 14.823 0.139 0.325 \nchr5 115994797 rs34498262 A G 0.391 LVRN Trans 0.065 0.008 1.41×10-17 10.403 0.406 0.039 \nchr5 116013119 rs17138656 A G 0.124 LVRN Trans 0.096 0.011 3.16×10-17 14.273 0.161 0.299 \nchr6 45189983 rs75523462 T G 0.950 SUPT3H Trans 0.104 0.017 1.31×10-9 7.424 0.685 0 \nchr6 133044782 rs34366581 T G 0.326 LINC00326 Trans 0.047 0.008 2.80×10-9 13.204 0.212 0.243 \nchr8 119000461 rs11995824 C G 0.454 TNFRSF11B Trans 0.100 0.007 5.62×10-41 16.222 0.093 0.384 \nchr10 122342063 rs6585816 T G 0.209        / Trans 0.056 0.009 7.84×10-10 6.121 0.805 0 \nchr12 481093 rs215223 A G 0.405 B4GALNT3 Trans -0.136 0.008 2.44×10-73 83.297 1.13×10-13  0.880 \nchr13 42378009 rs9594738 T C 0.482 TNFSF11 Trans -0.056 0.007 6.48×10-14 9.217 0.512 0 \nchr13 42513606 rs34136735 T C 0.052 TNFSF11 Trans 0.171 0.017 5.69×10-23 17.750 0.059 0.437 \nchr13 42532378 rs665632 T C 0.813 TNFSF11 Trans 0.082 0.010 4.82×10-17 8.502 0.484 0 \nchr14 92637384 rs7143806 A G 0.181 RIN3 Trans 0.053 0.010 3.35×10-8 13.360 0.204 0.251 \nchr14 94378610 rs28929474 T C 0.021 SERPINA1 Trans 0.173 0.027 1.10×10-10 5.342 0.867 0 \nchr17 43721253 rs66838809 A G 0.079 SOST Cis  -0.088 0.015 1.45×10-9 13.369 0.147 0.327 \nchr18 62390996 rs2957124 A G 0.421 TNFRSF11A Trans -0.057 0.008 5.97×10-14 11.286 0.257 0.203 \nchr20 11231094 rs13042961 T C 0.955 JAG1 Trans 0.126 0.019 4.65×10-11 16.398 0.037 0.512 \n \nNote: Locus (chromosome and position of the SNP), EA (effect allele), OA (other allele), EAF (effect allele frequency), GENE (n earest gene to \nthe sclerostin associated SNP); Cis/trans (the associated SNP is close to the SOST region [noted as cis] or far away from this region [noted as \n . CC-BY 4.0 International licenseIt is made available under a \n is the author/funder, who has granted medRxiv a license to display the preprint in perpetuity. (which was not certified by peer review)\nThe copyright holder for this preprint this version posted June 16, 2022. ; https://doi.org/10.1101/2022.06.13.22275915doi: medRxiv preprint \n\n \n36 \n \ntrans]); BETA (SD change in serum sclerostin per effect allele), SE (standard error) and P (p-value)) .  Heterogeneity test (Q (Cochran's Q \nstatistics), Q_P (Cochran's Q P value), I2 (I2 statistics))  \n . CC-BY 4.0 International licenseIt is made available under a \n is the author/funder, who has granted medRxiv a license to display the preprint in perpetuity. (which was not certified by peer review)\nThe copyright holder for this preprint this version posted June 16, 2022. ; https://doi.org/10.1101/2022.06.13.22275915doi: medRxiv preprint \n\n \n37 \n \nTable 3. Mendelian randomization and genetic correlation analysis results of the effect of sclerostin inhibition on coronary ar tery calcification, \nhypertension and type 2 diabetes. \nExposure Outcome Model N SNPs Estimate  SE P OR LCI UCI \nSclerostin inhibition Coronary artery calcification Cis-only MR 5 0.740 0.210 4.27×10-4 ** / / / \nCis+trans MR 4 0.058 0.126 0.645 / / / \nSclerostin inhibition Hypertension Cis-only MR 5 0.073 0.034 0.030 * 1.080 1.010 1.150 \nCis+trans MR 4 0.086 0.026 7.93×10-4 ** 1.090 1.037 1.146 \nSclerostin inhibition Type 2 diabetes Cis-only MR 5 0.233 0.080 3.66×10-3 ** 1.262 1.079 1.477 \nCis+trans MR 3 0.034 0.138 0.804 1.035 0.789 1.358 \n \nTrait 1 Trait 2 Model N SNPs  rg SE_rg P_rg \nSclerostin inhibition \nSclerostin inhibition \nSclerostin inhibition \nCoronary artery calcification \nHypertension \nType 2 diabetes \nGenetic correlation All SNPs 0.007 0.083 0.933 \nGenetic correlation All SNPs 0.134 0.045 3.10×10-3 ** \nGenetic correlation All SNPs -0.041 0.073 0.573 \n \nNote: Model refers to different statistical methods/models been used. N_snps means the number of genetic variants been included  as predictors \nfor sclerostin. Estimate, SE and P (and r g, SE_r g, P_r g) are the association estimates, standard error and P value of the MR (or the genetic \ncorrelation analysis). OR, LCI and UCI are the odds ratio and 95% confidence interval of the MR estimates, which is not applica ble for the \ngenetic correlation analysis. Importantly, the Mendelian randomization and genetic correlation analyses have different assumptions therefore the \neffect estimate is not directly comparable. We listed them in the same table to compare the direction of effects and the P valu e estimates across \nthe two approaches. \n* Association reached marginal significance threshold of α =0.05 \n** Association reached Bonferroni-corrected significance threshold of α =4.17×10-3 \n \n . CC-BY 4.0 International licenseIt is made available under a \n is the author/funder, who has granted medRxiv a license to display the preprint in perpetuity. (which was not certified by peer review)\nThe copyright holder for this preprint this version posted June 16, 2022. ; https://doi.org/10.1101/2022.06.13.22275915doi: medRxiv preprint \n\n . CC-BY 4.0 International licenseIt is made available under a \n is the author/funder, who has granted medRxiv a license to display the preprint in perpetuity. (which was not certified by peer review)\nThe copyright holder for this preprint this version posted June 16, 2022. ; https://doi.org/10.1101/2022.06.13.22275915doi: medRxiv preprint \n\n . CC-BY 4.0 International licenseIt is made available under a \n is the author/funder, who has granted medRxiv a license to display the preprint in perpetuity. (which was not certified by peer review)\nThe copyright holder for this preprint this version posted June 16, 2022. ; https://doi.org/10.1101/2022.06.13.22275915doi: medRxiv preprint \n\n . CC-BY 4.0 International licenseIt is made available under a \n is the author/funder, who has granted medRxiv a license to display the preprint in perpetuity. (which was not certified by peer review)\nThe copyright holder for this preprint this version posted June 16, 2022. ; https://doi.org/10.1101/2022.06.13.22275915doi: medRxiv preprint \n\n . CC-BY 4.0 International licenseIt is made available under a \n is the author/funder, who has granted medRxiv a license to display the preprint in perpetuity. (which was not certified by peer review)\nThe copyright holder for this preprint this version posted June 16, 2022. ; https://doi.org/10.1101/2022.06.13.22275915doi: medRxiv preprint \n\n . CC-BY 4.0 International licenseIt is made available under a \n is the author/funder, who has granted medRxiv a license to display the preprint in perpetuity. (which was not certified by peer review)\nThe copyright holder for this preprint this version posted June 16, 2022. ; https://doi.org/10.1101/2022.06.13.22275915doi: medRxiv preprint","source_license":"CC-BY-4.0","license_restricted":false}