Combining evidence from Mendelian randomization and colocalization: Review and comparison of approaches

The interplay between inflammatory cytokines and cardiometabolic disease: bi-directional mendelian randomisation study

Objective

To leverage large scale genetic association data to investigate the interplay between circulating cytokines and cardiometabolic traits, and thus identifying potential therapeutic targets.

Design

Bi-directional Mendelian randomisation study.

Setting

Genome-wide association studies from three Finnish cohorts (Northern Finland Birth Cohort 1966, Young Finns Study, or FINRISK study), and genetic association summary statistics pooled from observational studies for expression quantitative trait loci and cardiometabolic traits.

Participants

Data for 47 circulating cytokines in 13 365 individuals from genome-wide association studies, summary statistic data for up to 21 735 individuals on circulating cytokines, summary statistic gene expression data across 49 tissues in 838 individuals, and summary statistic data for up to 1 320 016 individuals on cardiometabolic traits.

Interventions

Relations between circulating cytokines and cardiovascular, anthropometric, lipid, or glycaemic traits (coronary artery disease, stroke, type 2 diabetes mellitus, body mass index, waist circumference, waist to hip ratio, systolic blood pressure, glycated haemoglobin, high density lipoprotein cholesterol, low density lipoprotein cholesterol, total cholesterol, triglycerides, C reactive protein, glucose, fasting insulin, and lifetime smoking).

Main outcome methods

Genetic instrumental variables that are biologically plausible for the circulating cytokines were generated. The effects of cardiometabolic risk factors on concentrations of circulating cytokines, circulating cytokines on other circulating cytokines, and circulating cytokines on cardiometabolic outcomes were investigated.

Results

Genetic evidence (mendelian randomisation P<0.0011) suggests that higher body mass index, waist circumference, smoking, higher concentrations of lipids, and systolic blood pressure increase circulating concentrations of several inflammatory cytokines and C reactive protein. Evidence for causal relations (mendelian randomisation P<0.0011) were noted between circulating cytokines, including a key role of vascular endothelial growth factor on influencing the concentrations of 10 other cytokines. Both mendelian randomisation (P<0.05) and colocalisation (posterior probability >0.5) suggested that coronary artery disease risk is increased by higher concentrations of circulating tumour necrosis factor related apoptosis-inducing ligand (TRAIL), interleukin-1 receptor antagonist (IL1RA), and macrophage colony-stimulating factor (MCSF).

Conclusion

This study offers insight into inflammatory mediators of cardiometabolic risk factors, cytokine signalling cascades, and effects of circulating cytokines on different cardiometabolic outcomes.

Mendelian randomization as a tool to inform drug development using human genetics

Drug development is essential to the advancement of human health, however, the process is slow, costly, and at high risk of failure at all stages. A promising strategy for expediting and improving the probability of success in the drug development process is the use of naturally randomized human genetic variation for drug target identification and validation. These data can be harnessed using the Mendelian randomization (MR) analytic paradigm to proxy the lifelong consequences of genetic perturbations of drug targets. In this review, we discuss the myriad applications of the MR paradigm for human drug target identification and validation. We review the methodology and applications of MR, key limitations of MR, and potential future opportunities for research. Throughout the review, we refer to illustrative examples of MR analyses investigating the consequences of genetic inhibition of interleukin 6 signaling which, in some cases, have anticipated results from randomized controlled trials. As human genetic data become more widely available, we predict that MR will serve as a key pillar of support for drug development efforts.

Using genetic association data to guide drug discovery and development: Review of methods and applications

Evidence on the validity of drug targets from randomized trials is reliable but typically expensive and slow to obtain. In contrast, evidence from conventional observational epidemiological studies is less reliable because of the potential for bias from confounding and reverse causation. Mendelian randomization is a quasi-experimental approach analogous to a randomized trial that exploits naturally occurring randomization in the transmission of genetic variants. In Mendelian randomization, genetic variants that can be regarded as proxies for an intervention on the proposed drug target are leveraged as instrumental variables to investigate potential effects on biomarkers and disease outcomes in large-scale observational datasets. This approach can be implemented rapidly for a range of drug targets to provide evidence on their effects and thus inform on their priority for further investigation. In this review, we present statistical methods and their applications to showcase the diverse opportunities for applying Mendelian randomization in guiding clinical development efforts, thus enabling interventions to target the right mechanism in the right population group at the right time. These methods can inform investigators on the mechanisms underlying drug effects, their related biomarkers, implications for the timing of interventions, and the population subgroups that stand to gain the most benefit. Most methods can be implemented with publicly available data on summarized genetic associations with traits and diseases, meaning that the only major limitations to their usage are the availability of appropriately powered studies for the exposure and outcome and the existence of a suitable genetic proxy for the proposed intervention.

Statistical methods for cis-Mendelian randomization with two-sample summary-level data

Mendelian randomization (MR) is the use of genetic variants to assess the existence of a causal relationship between a risk factor and an outcome of interest. Here, we focus on two-sample summary-data MR analyses with many correlated variants from a single gene region, particularly on cis-MR studies which use protein expression as a risk factor. Such studies must rely on a small, curated set of variants from the studied region; using all variants in the region requires inverting an ill-conditioned genetic correlation matrix and results in numerically unstable causal effect estimates. We review methods for variable selection and estimation in cis-MR with summary-level data, ranging from stepwise pruning and conditional analysis to principal components analysis, factor analysis, and Bayesian variable selection. In a simulation study, we show that the various methods have comparable performance in analyses with large sample sizes and strong genetic instruments. However, when weak instrument bias is suspected, factor analysis and Bayesian variable selection produce more reliable inferences than simple pruning approaches, which are often used in practice. We conclude by examining two case studies, assessing the effects of low-density lipoprotein-cholesterol and serum testosterone on coronary heart disease risk using variants in the HMGCR and SHBG gene regions, respectively.

Exploring genes for immunoglobulin A nephropathy: a summary data-based mendelian randomization and FUMA analysis

BACKGROUND

Immunoglobulin A nephropathy (IgAN) is a complex autoimmune disease, and the exact pathogenesis remains to be elucidated. This study aimed to explore genes underlying the pathogenesis of IgAN.

METHODS

We conducted the summary data-based Mendelian randomization (SMR) analysis and performed functional mapping and annotation using FUMA to explore genetic loci that are potentially involved in the pathogenies of IgAN. Both analyses used summarized data of a recent genome-wide association study (GWAS) on IgANs, which included 477,784 Europeans (15,587 cases and 462,197 controls) and 175,359 East Asians (71 cases and 175,288 controls). We performed SMR analysis using Consortium for the Architecture of Gene Expression (CAGE) expression quantitative trait loci (eQTL) data and replicated the analysis using Genotype-Tissue Expression (GTEx) eQTL data.

RESULTS

Using the CAGE eQTL data, our SMR analysis identified 32 probes tagging 25 unique genes whose expression were pleiotropically associated with IgAN, with the top three probes being ILMN_2150787 (tagging HLA-C, P_SMR= 2.10 × 10^-18), ILMN_1682717 (tagging IER3, P_SMR= 1.07 × 10^-16) and ILMN_1661439 (tagging FLOT1, P_SMR=1.16 × 10^-14). Using GTEx eQTL data, our SMR analysis identified 24 probes tagging 24 unique genes whose expressions were pleiotropically associated with IgAN, with the top three probes being ENSG00000271581.1 (tagging XXbac-BPG248L24.12, P_SMR= 1.44 × 10^-10), ENSG00000186470.9 (tagging BTN3A2, P_SMR= 2.28 × 10^-10), and ENSG00000224389.4 (tagging C4B, P_SMR= 1.23 × 10 ^-9). FUMA analysis identified 3 independent, significant and lead SNPs, 2 genomic risk loci and 39 genes that are potentially involved in the pathogenesis of IgAN.

CONCLUSION

We identified many genetic variants/loci that are potentially involved in the pathogenesis of IgAN. More studies are needed to elucidate the exact mechanisms of the identified genetic variants/loci in the etiology of IgAN.

Novel Causal Plasma Proteins for Hypothyroidism: A Large-scale Plasma Proteome Mendelian Randomization Analysis

CONTEXT

Although several risk proteins for hypothyroidism have been reported in recent years, many more plasma proteins have not been tested.

OBJECTIVE

To determine potential mechanisms and novel causal plasma proteins for hypothyroidism using Mendelian randomization (MR).

METHODS

A large-scale plasma proteome MR analysis was conducted using protein quantitative trait loci (pQTLs) for 2297 plasma proteins. We classified pQTLs into 4 different groups. MR analyses were conducted within the 4 groups simultaneously. Significant proteins were discovered and validated in 2 different cohorts. Colocalization analysis and enrichment analysis were conducted using proteins found with MR.

RESULTS

Thirty-one proteins were identified in the discovery cohort. Among them, 13 were validated in the validation cohort. Nine of the 13 proteins are risk factors (ISG15, Fc receptor-like protein 2, tumor necrosis factor ligand superfamily member 14, Rab-2A, FcRL3, thrombomodulin, interferon [IFN]-lambda-1, platelet glycoprotein Ib alpha chain, IL-7RA) for hypothyroidism, whereas others are protective proteins (protein O-glucosyltransferase 1 [POGLUT1], tumor necrosis factor ligand superfamily, 3-hydroxyisobutyryl-CoA hydrolase, transferrin receptor protein 1). Among the significant proteins, POGLUT1 strongly colocalized with expression quantitative trait loci from whole blood (posterior probability of colocalization [PP4] = 0.978) and the thyroid (PP4 = 0.978). Two different trans-pQTLs (rs2111485 PP4 = 0.998; rs35103715 PP4 = 0.998) for IFN-lambda-1 strongly colocalized with hypothyroidism in different chromosomes.

CONCLUSION

Thirteen various proteins were identified and validated to be associated with hypothyroidism using univariable MR. We reinforced and expanded the effect of IFN on hypothyroidism. Several proteins identified in this study could explain part of the association between the coagulation system and hypothyroidism. Our study broadens the causal proteins for hypothyroidism and provides the relationships between plasma proteins and hypothyroidism. The proteins identified in this study can be used as early screening biomarkers for hypothyroidism.

SUMMER: a Mendelian randomization interactive server to systematically evaluate the causal effects of risk factors and circulating biomarkers on pan-cancer survival

Genome-wide association studies (GWASs) underlying case-control design have uncovered hundreds of genetic loci involved in tumorigenesis and provided rich resources for identifying risk factors and biomarkers associated with cancer susceptibility. However, the application of GWAS in determining the genetic architecture of cancer survival remains unestablished. Here, we systematically evaluated genetic effects at the genome-wide level on cancer survival that included overall survival (OS) and cancer-specific survival (CSS), leveraging data deposited in the UK Biobank cohort of a total of 19 628 incident patients across 17 cancer types. Furthermore, we assessed the causal effects of risk factors and circulating biomarkers on cancer prognosis via a Mendelian randomization (MR) analytic framework, which integrated cancer survival GWAS dataset, along with phenome-wide association study (PheWAS) and blood genome-wide gene expression/DNA methylation quantitative trait loci (eQTL/meQTL) datasets. On average, more than 10 traits, 700 genes, and 4,500 CpG sites were prone to cancer prognosis. Finally, we developed a user-friendly online database, SUrvival related cancer Multi-omics database via MEndelian Randomization (SUMMER; http://njmu-edu.cn:3838/SUMMER/), to help users query, browse, and download cancer survival results. In conclusion, SUMMER provides an important resource to assist the research community in understanding the genetic mechanisms of cancer survival.

Mendelian randomization eradicates the causal relationship between educational attainment, household income, and oropharyngeal cancer

Background

It was reported that educational attainment and household income are associated with oropharyngeal cancer. However, whether such an association is causal is still unknown.

Methods

The Mendelian randomization (MR) design was performed to disentangle their causal relationship. Initially, genetic variants proxied for educational attainment and household income were extracted from the largest genome-wide association studies (GWAS), and two oropharyngeal GWAS datasets were used in the discovery and validation stages separately. A reverse MR analysis was carried out to judge whether oropharyngeal cancer affects educational attainment and household income. The results from the two stages were combined using meta-analysis. The heterogeneity and horizontal pleiotropy were appraised using several methods.

Results

All selected genetic variants were valid. In the discovery stage, genetically elevated years of education might decrease the risk of oropharyngeal cancer (IVW OR = 0.148 [0.025, 0.872], p-value = 0.035), while such a result became insignificant in the validation stage (IVW p-value >0.05). Household income cannot change the risk of oropharyngeal cancer at both stages. The reverse MR suggested that oropharyngeal cancer should slightly alter household income (IVW OR = 1.001 [1.000, 1.003], p-value = 0.036) in the discovery set, but the result cannot be replicated in the validation stage. The meta-analysis did not find any significant results either. The results were also assessed by sensitivity analyses, and there was no heterogeneity or horizontal pleiotropy in the analyses. The statistical powers were all above 80% at the discovery stage.

Conclusions

There should be no causal association between educational attainment, household income, and oropharyngeal cancer.

Association between genetically proxied PCSK9 inhibition and prostate cancer risk: A Mendelian randomisation study

BACKGROUND

Prostate cancer (PrCa) is the second most prevalent malignancy in men worldwide. Observational studies have linked the use of low-density lipoprotein cholesterol (LDL-c) lowering therapies with reduced risk of PrCa, which may potentially be attributable to confounding factors. In this study, we performed a drug target Mendelian randomisation (MR) analysis to evaluate the association of genetically proxied inhibition of LDL-c-lowering drug targets on risk of PrCa.

METHODS AND FINDINGS

Single-nucleotide polymorphisms (SNPs) associated with LDL-c (P < 5 × 10-8) from the Global Lipids Genetics Consortium genome-wide association study (GWAS) (N = 1,320,016) and located in and around the HMGCR, NPC1L1, and PCSK9 genes were used to proxy the therapeutic inhibition of these targets. Summary-level data regarding the risk of total, advanced, and early-onset PrCa were obtained from the PRACTICAL consortium. Validation analyses were performed using genetic instruments from an LDL-c GWAS conducted on male UK Biobank participants of European ancestry (N = 201,678), as well as instruments selected based on liver-derived gene expression and circulation plasma levels of targets. We also investigated whether putative mediators may play a role in findings for traits previously implicated in PrCa risk (i.e., lipoprotein a (Lp(a)), body mass index (BMI), and testosterone). Applying two-sample MR using the inverse-variance weighted approach provided strong evidence supporting an effect of genetically proxied inhibition of PCSK9 (equivalent to a standard deviation (SD) reduction in LDL-c) on lower risk of total PrCa (odds ratio (OR) = 0.85, 95% confidence interval (CI) = 0.76 to 0.96, P = 9.15 × 10-3) and early-onset PrCa (OR = 0.70, 95% CI = 0.52 to 0.95, P = 0.023). Genetically proxied HMGCR inhibition provided a similar central effect estimate on PrCa risk, although with a wider 95% CI (OR = 0.83, 95% CI = 0.62 to 1.13, P = 0.244), whereas genetically proxied NPC1L1 inhibition had an effect on higher PrCa risk with a 95% CI that likewise included the null (OR = 1.34, 95% CI = 0.87 to 2.04, P = 0.180). Analyses using male-stratified instruments provided consistent results. Secondary MR analyses supported a genetically proxied effect of liver-specific PCSK9 expression (OR = 0.90 per SD reduction in PCSK9 expression, 95% CI = 0.86 to 0.95, P = 5.50 × 10-5) and circulating plasma levels of PCSK9 (OR = 0.93 per SD reduction in PCSK9 protein levels, 95% CI = 0.87 to 0.997, P = 0.04) on PrCa risk. Colocalization analyses identified strong evidence (posterior probability (PPA) = 81.3%) of a shared genetic variant (rs553741) between liver-derived PCSK9 expression and PrCa risk, whereas weak evidence was found for HMGCR (PPA = 0.33%) and NPC1L1 expression (PPA = 0.38%). Moreover, genetically proxied PCSK9 inhibition was strongly associated with Lp(a) levels (Beta = -0.08, 95% CI = -0.12 to -0.05, P = 1.00 × 10-5), but not BMI or testosterone, indicating a possible role for Lp(a) in the biological mechanism underlying the association between PCSK9 and PrCa. Notably, we emphasise that our estimates are based on a lifelong exposure that makes direct comparisons with trial results challenging.

CONCLUSIONS

Our study supports a strong association between genetically proxied inhibition of PCSK9 and a lower risk of total and early-onset PrCa, potentially through an alternative mechanism other than the on-target effect on LDL-c. Further evidence from clinical studies is needed to confirm this finding as well as the putative mediatory role of Lp(a).

A large-scale plasma proteome Mendelian randomization study identifies novel causal plasma proteins related to primary biliary cholangitis

Background and aims

Primary biliary cholangitis (PBC) is a progressive chronic autoimmune cholestatic liver disease characterized by the destruction of small intrahepatic bile ducts leading to biliary cirrhosis. Liver biopsy is required in the diagnosis of Antimitochondrial antibody-negative patients. Therefore, novel biomarkers are needed for the non-invasive diagnosis of PBC. To identify novel biomarkers for PBC, we conducted large-scale plasma proteome Mendelian randomization (MR).

Methods

A total of 21,593 protein quantitative trait loci (pQTLs) for 2297 circulating proteins were used and classified into four different groups. MR analyses were conducted in the four groups separately. Furthermore, the results were discovered and replicated in two different cohorts of PBC. Colocalization analysis and enrichment analysis were also conducted.

Results

Three plasma proteins (ficolin-1, CD40 and protein FAM177A1) were identified and replicated as being associated with PBC. All of them showed significant protective effects against PBC. An increase in ficolin-1 (OR=0.890 [0.843-0.941], p=3.50×10^-5), CD40 (OR=0.814 [0.741-0.895], p=1.96×10^-5) and protein FAM177A1 (OR=0.822 [0.754-0.897], p=9.75×10^-6) reduced the incidence of PBC. Ficolin-1 (PP4 = 0.994) and protein FAM177A1 (PP4 = 0.995) colocalized with the expression of the genes FCN1 and FAM177A1 in whole blood, respectively. Furthermore, CD40 (PP4 = 0.977) and protein FAM177A1 (PP4 = 0.897) strongly colocalized with PBC.

Conclusions

We expand the current biomarkers for PBC. In total, three (ficolin-1, CD40, and protein FAM177A1) plasma proteins were identified and replicated as being associated with PBC in MR analysis. All of them showed significant protective effects against PBC. These proteins can be potential biomarkers or drug targets for PBC.

Investigating the prediction of CpG methylation levels from SNP genotype data to help elucidate relationships between methylation, gene expression and complex traits

As popularised by PrediXcan (and related methods), transcriptome-wide association studies (TWAS), in which gene expression is imputed from single-nucleotide polymorphism (SNP) genotypes and tested for association with a phenotype, are a popular approach for investigating the role of gene expression in complex traits. Like gene expression, DNA methylation is an important biological process and, being under genetic regulation, may be imputable from SNP genotypes. Here, we investigate prediction of CpG methylation levels from SNP genotype data to help elucidate relationships between methylation, gene expression and complex traits. We start by examining how well CpG methylation can be predicted from SNP genotypes, comparing three penalised regression approaches and examining whether changing the window size improves prediction accuracy. Although methylation at most CpG sites cannot be accurately predicted from SNP genotypes, for a subset it can be predicted well. We next apply our methylation prediction models (trained using the optimal method and window size) to carry out a methylome-wide association study (MWAS) of primary biliary cholangitis. We intersect the regions identified via MWAS with those identified via TWAS, providing insight into the interplay between CpG methylation, gene expression and disease status. We conclude that MWAS has the potential to improve understanding of biological mechanisms in complex traits.

Bayesian Genetic Colocalization Test of Two Traits Using coloc

Genetic colocalization is an approach for determining whether a genetic variant at a particular locus is shared across multiple phenotypes. Genome-wide association studies (GWAS) have successfully mapped genetic variants associated with thousands of complex traits and diseases. However, a large proportion of GWAS signals fall in non-coding regions of the genome, making functional interpretation a challenge. Colocalization relies on a Bayesian framework that can integrate summary statistics, for example those derived from GWAS and expression quantitative trait loci (eQTL) mapping, to assess whether two or more independent association signals at a region of interest are consistent with a shared causal variant. The results from a colocalization analysis may be used to evaluate putative causal relationships between omics-based molecular measurements and a complex disease, and can generate hypotheses that may be followed up by tailored experiments. In this article, we present an easy and straightforward protocol for conducting a Bayesian test for colocalization of two traits using the 'coloc' package in R with summary-level results derived from GWAS and eQTL studies. We also provide general guidelines that can assist in the interpretation of findings generated from colocalization analyses. © 2022 Wiley Periodicals LLC. Basic Protocol: Performing a genetic colocalization analysis using the 'coloc' package in R and summary-level data Support Protocol: Installing the 'coloc' R package.

Calcium, Its Regulatory Hormones, and Their Causal Role on Blood Pressure: A Two-Sample Mendelian Randomization Study

CONTEXT

Vitamin D (Vit-D), parathyroid hormone (PTH), and fibroblast growth factor 23 (FGF23) are the major calciotropic hormones involved in the regulation of blood calcium levels from the intestine, kidney, and bone through a tight endocrine feedback loop system. Altered levels of calcium itself or through the effect of its regulatory hormones could affect blood pressure (BP), but the exact mechanisms remain unclear.

OBJECTIVE

To evaluate whether a causal relationship exists between serum calcium level and/or the regulatory hormones involved in its homeostasis with BP, we performed a two-sample Mendelian randomization (MR) study.

METHODS

From 4 large genome-wide association studies (GWAS) we obtained independent (r2 < 0.001) single nucleotide polymorphisms (SNPs) associated with serum calcium (119 SNPs), Vit-D (78 SNPs), PTH (5 SNPs), and FGF23 (5 SNPs), to investigate through MR their association with systolic BP (SBP) and diastolic BP (DBP) in a Swedish urban-based study, the Malmö Diet and Cancer study (n = 29 298). Causality was evaluated by the inverse variance weighted method (IVW) and weighted median, while MR Egger and MR-PRESSO were used as sensitivity analyses.

RESULTS

Genetically predicted serum calcium level was found to be associated with DBP (IVW: beta = 0.10, SE = 0.04, P = 0.007) and SBP (IVW: beta = 0.07, SE = 0.04, P = 0.04). Genetically predicted Vit-D and PTH showed no association with the traits, while FGF23 was inversely associated with SBP (IVW: beta = -0.11, SE = 0.04, P = 0.01), although this association lost statistical significance in sensitivity analysis.

CONCLUSION

Our study shows a direct association between genetically predicted calcium level and DBP, and a weaker association with SBP. No such clear association was found for genetically predicted calciotropic hormone levels. It is of interest to detect which target genes involved in calcium homeostasis mediate the effect of calcium on BP, particularly for improving personalized intervention strategies.

Dissecting causal relationships between nonalcoholic fatty liver disease proxied by chronically elevated alanine transaminase levels and 34 extrahepatic diseases

BACKGROUND

Nonalcoholic fatty liver disease (NAFLD) is prevalent worldwide and is associated with the risk of many extrahepatic diseases. However, whether NAFLD is a risk marker or a common cause of extrahepatic diseases is unclear.

METHODS

We searched PubMed to identify NAFLD-related extrahepatic diseases. Genetic instrumental variables (IVs) for NAFLD surrogated by chronically elevated alanine transaminase levels and eligible extrahepatic diseases were retrieved from the corresponding genome-wide association analysis. We proposed a procedure for Mendelian randomization (MR) analysis and performed validation analyses to dissect the association between NAFLD and extrahepatic diseases. The Bonferroni method was used to correct the bias of multiple testing.

RESULTS

In total, 34 extrahepatic diseases were included and 54 SNPs were used as IVs for NAFLD. The MR analysis gave a robust and significant (or suggestive) estimate for the association between NAFLD and 9 extrahepatic diseases: type 2 diabetes (odds ratio [OR] = 1.182, 95 % confidence interval [CI] 1.125-1.243, P = 5.40 × 10^-11), cholelithiasis (OR = 1.171, 95%CI 1.083-1.266, P = 7.47 × 10^-5), diabetic hypoglycemia (OR = 1.170, 95%CI 1.071-1.279, P = 5.14 × 10^-4), myocardial infarction (OR = 1.122, 95%CI 1.057-1.190, P = 1.46 × 10^-4), hypertension (OR = 1.060, 95%CI 1.029-1.093, P = 1.18 × 10^-4), coronary artery disease (OR = 1.052, 95%CI 1.010-1.097, P = 1.58 × 10^-2), heart failure (OR = 1.047, 95%CI 1.006-1.090, P = 2.44 × 10^-2), dementia (OR = 0.881, 95%CI 0.806-0.962, P = 5.01 × 10^-3), and pancreatic cancer (OR = 0.802, 95%CI 0.654-0.983, P = 3.32 × 10^-2). Validation analyses using IVs from biopsy-confirmed and imaging-determined NAFLD reported similar results to the main analysis. For the remaining 25 outcomes, no significant or definitive association was yielded in MR analysis.

CONCLUSIONS

Genetic evidence suggests putative causal relationships between NAFLD and a set of extrahepatic diseases, indicating that NAFLD deserves high priority in clinical practice.

Evaluating the impact of glucokinase activation on risk of cardiovascular disease: a Mendelian randomisation analysis

BACKGROUND

Glucokinase activators (GKAs) are an emerging class of glucose lowering drugs that activate the glucose-sensing enzyme glucokinase (GK). Pending formal cardiovascular outcome trials, we applied two-sample Mendelian randomisation (MR) to investigate the impact of GK activation on risk of cardiovascular diseases.

METHODS

We used independent genetic variants in or around the glucokinase gene meanwhile associated with HbA_1c at genome-wide significance (P < 5 × 10^-8) in the Meta-Analyses of Glucose and Insulin-related traits Consortium study (N = 146,806; European ancestry) as instrumental variables (IVs) to mimic the effects of GK activation. We assessed the association between genetically proxied GK activation and the risk of coronary artery disease (CAD; 122,733 cases and 424,528 controls), peripheral arterial disease (PAD; 7098 cases and 206,541 controls), stroke (40,585 cases and 406,111 controls) and heart failure (HF; 47,309 cases and 930,014 controls), using genome-wide association study summary statistics of these outcomes in Europeans. We compared the effect estimates of genetically proxied GK activation with estimates of genetically proxied lower HbA_1c on the same outcomes. We repeated our MR analyses in East Asians as validation.

RESULTS

Genetically proxied GK activation was associated with reduced risk of CAD (OR 0.38 per 1% lower HbA_1c, 95% CI 0.29-0.51, P = 8.77 × 10^-11) and HF (OR 0.54 per 1% lower HbA_1c, 95% CI 0.41-0.73, P = 3.55 × 10^-5). The genetically proxied protective effects of GKA on CAD and HF exceeded those due to non-targeted HbA_1c lowering. There was no causal relationship between genetically proxied GK activation and risk of PAD or stroke. The estimates in sensitivity analyses and in East Asians were generally consistent.

CONCLUSIONS

GKAs may protect against CAD and HF which needs confirmation by long-term clinical trials.

TwoStepCisMR: A Novel Method and R Package for Attenuating Bias in cis-Mendelian Randomization Analyses

Mendelian randomisation (MR) is an increasingly popular method for strengthening causal inference in epidemiological studies. cis-MR in particular uses genetic variants in the gene region of a drug target protein as an instrumental variable to provide quasi-experimental evidence for on-target drug effects. A limitation of this framework is when the genetic variant is correlated to another variant that also effects the outcome of interest (confounding through linkage disequilibrium). Methods for correcting this bias, such as multivariable MR, struggle in a cis setting because of the high correlation among genetic variants. Here, through simulation experiments and an applied example considering the effect of interleukin 6 receptor signaling on coronary artery disease risk, we present an alternative method for attenuating bias that does not suffer from this problem. As our method uses both MR and the product and difference method for mediation analysis, our proposal inherits all assumptions of these methods. We have additionally developed an R package, TwoStepCisMR, to facilitate the implementation of the method.

ELF5 is a potential respiratory epithelial cell-specific risk gene for severe COVID-19

Despite two years of intense global research activity, host genetic factors that predispose to a poorer prognosis of COVID-19 infection remain poorly understood. Here, we prioritise eight robust (e.g., ELF5) or suggestive but unreported (e.g., RAB2A) candidate protein mediators of COVID-19 outcomes by integrating results from the COVID-19 Host Genetics Initiative with population-based plasma proteomics using statistical colocalisation. The transcription factor ELF5 (ELF5) shows robust and directionally consistent associations across different outcome definitions, including a >4-fold higher risk (odds ratio: 4.88; 95%-CI: 2.47-9.63; p-value < 5.0 × 10-6) for severe COVID-19 per 1 s.d. higher genetically predicted plasma ELF5. We show that ELF5 is specifically expressed in epithelial cells of the respiratory system, such as secretory and alveolar type 2 cells, using single-cell RNA sequencing and immunohistochemistry. These cells are also likely targets of SARS-CoV-2 by colocalisation with key host factors, including ACE2 and TMPRSS2. In summary, large-scale human genetic studies together with gene expression at single-cell resolution highlight ELF5 as a risk gene for severe COVID-19, supporting a role of epithelial cells of the respiratory system in the adverse host response to SARS-CoV-2.

Multi-ancestry Mendelian randomization of omics traits revealing drug targets of COVID-19 severity

BACKGROUND

Recent omic studies prioritised several drug targets associated with coronavirus disease 2019 (COVID-19) severity. However, little evidence was provided to systematically estimate the effect of drug targets on COVID-19 severity in multiple ancestries.

METHODS

In this study, we applied Mendelian randomization (MR) and colocalization approaches to understand the putative causal effects of 16,059 transcripts and 1608 proteins on COVID-19 severity in European and effects of 610 proteins on COVID-19 severity in African ancestry. We further integrated genetics, clinical and literature evidence to prioritise drug targets. Additional sensitivity analyses including multi-trait colocalization and phenome-wide MR were conducted to test for MR assumptions.

FINDINGS

MR and colocalization prioritized four protein targets, FCRL3, ICAM5, ENTPD5 and OAS1 that showed effect on COVID-19 severity in European ancestry. One protein target, SERPINA1 showed a stronger effect in African ancestry but much weaker effect in European ancestry (odds ratio [OR] in Africans=0.369, 95%CI=0.203 to 0.668, P = 9.96 × 10-4; OR in Europeans=1.021, 95%CI=0.901 to 1.157, P = 0.745), which suggested that increased level of SERPINA1 will reduce COVID-19 risk in African ancestry. One protein, ICAM1 showed suggestive effect on COVID-19 severity in both ancestries (OR in Europeans=1.152, 95%CI=1.063 to 1.249, P = 5.94 × 10-4; OR in Africans=1.481, 95%CI=1.008 to 2.176; P = 0.045). The OAS1, SERPINA1 and ICAM1 effects were replicated using updated COVID-19 severity data in the two ancestries respectively, where alternative splicing events in OAS1 and ICAM1 also showed marginal effects on COVID-19 severity in Europeans. The phenome-wide MR of the prioritised targets on 622 complex traits provided information on potential beneficial effects on other diseases and suggested little evidence of adverse effects on major complications.

INTERPRETATION

Our study identified six proteins as showing putative causal effects on COVID-19 severity. OAS1 and SERPINA1 were targets of existing drugs in trials as potential COVID-19 treatments. ICAM1, ICAM5 and FCRL3 are related to the immune system. Across the six targets, OAS1 has no reliable instrument in African ancestry; SERPINA1, FCRL3, ICAM5 and ENTPD5 showed a different level of putative causal evidence in European and African ancestries, which highlights the importance of more powerful ancestry-specific GWAS and value of multi-ancestry MR in informing the effects of drug targets on COVID-19 across different populations. This study provides a first step towards clinical investigation of beneficial and adverse effects of COVID-19 drug targets.

FUNDING

No.