key: cord-0765263-0mvtnlbo
authors: Zhang, Kun; Dong, Shan-Shan; Guo, Yan; Tang, Shi-Hao; Wu, Hao; Yao, Shi; Wang, Peng-Fei; Zhang, Kun; Xue, Han-Zhong; Huang, Wei; Ding, Jian; Yang, Tie-Lin
title: Causal Associations Between Blood Lipids and COVID-19 Risk: A Two-Sample Mendelian Randomization Study
date: 2021-09-09
journal: Arterioscler Thromb Vasc Biol
DOI: 10.1161/atvbaha.121.316324
sha: 34899c09cb711c3f72ee70caea71ce5b67408433
doc_id: 765263
cord_uid: 0mvtnlbo

OBJECTIVE: Coronavirus disease 2019 (COVID-19) is a global pandemic caused by the severe acute respiratory syndrome coronavirus 2. It has been reported that dyslipidemia is correlated with COVID-19, and blood lipids levels, including total cholesterol, HDL-C (high-density lipoprotein cholesterol), and LDL-C (low-density lipoprotein cholesterol) levels, were significantly associated with disease severity. However, the causalities of blood lipids on COVID-19 are not clear. APPROACH AND RESULTS: We performed 2-sample Mendelian randomization (MR) analyses to explore the causal effects of blood lipids on COVID-19 susceptibility and severity. Using the outcome data from the UK Biobank (1221 cases and 4117 controls), we observed potential positive causal effects of dyslipidemia (odds ratio [OR], 1.27 [95% CI, 1.08–1.49], P=3.18×10(−3)), total cholesterol (OR, 1.19 [95% CI, 1.07–1.32], P=8.54×10(−4)), and ApoB (apolipoprotein B; OR, 1.18 [95% CI, 1.07–1.29], P=1.01×10(−3)) on COVID-19 susceptibility after Bonferroni correction. In addition, the effects of total cholesterol (OR, 1.01 [95% CI, 1.00–1.02], P=2.29×10(−2)) and ApoB (OR, 1.01 [95% CI, 1.00–1.02], P=2.22×10(−2)) on COVID-19 susceptibility were also identified using outcome data from the host genetics initiative (14 134 cases and 1 284 876 controls). CONCLUSIONS: In conclusion, we found that higher total cholesterol and ApoB levels might increase the risk of COVID-19 infection.

contrast, Peng et al 11 observed significantly increased level of LDL-C in patients with COVID-19 compared with age-and sex-matched controls where the levels of HDL-C and TC were inversely correlated with the severity of COVID-19. 11 Although above evidences demonstrated the associations between blood lipids and COVID-19, these findings were from observational studies which could be misguided by potential confounders, 12 whether these associations are causal is still unclear.

Mendelian randomization (MR) is an epidemiological method in which environmental exposure-related genetic variations are used as instrumental variables (IVs) to evaluate the association between exposures and outcomes. 13, 14 It can avoid the issues of confusion and has been demonstrated as an effective strategy to identify the causal effect. [14] [15] [16] Two-sample MR uses genetic associations with the exposure and outcome from the summary statistics of nonoverlapping genome-wide association studies (GWAS) and has facilitated the application of the MR methodology. 14, 17, 18 In this study, we conducted a 2-sample MR study to explore the possible causal effects of 7 blood lipids on COVID-19 susceptibility and severity using data from the UK Biobank (UKB) and the host genetics initiative (HGI).

The authors declare that all supporting data are available within the article and its Data Supplement. The data sets used in this study are listed in Major Resources Table in the Data Supplement. A step-by-step workflow in this study is presented in Figure 1 .

Details of the contributing GWAS summary data are list in Table 1 . The studies were selected for investigating blood lipids or COVID-19 having the largest sample sizes and consisting of similar populations (>70% White population/Europeans).

For each exposure, we filtered single-nucleotide polymorphisms (SNPs) using the following criteria:

1. Remove the SNPs located in the major histocompatibility complex region. 2. Remove the SNPs with minor allele frequency <0.01 in the 1000 genome European data. The estimated standardize the effect size (β) and SE for each GWAS data was obtained with the function of minor allele frequency and sample size as follows 24 :

where z=β/SE from the original summary data, p is the minor allele frequency, and n is the total sample size.

We selected independent and genome-wide significant GWAS SNPs using the clumping algorithm in PLINK (http://pngu. mgh.harvard.edu/purcell/plink/) 25 28 MR-Egger is based on the assumption which indicates instrument strength independent of the direct effects. 27 It can be evaluated by the regression dilution I 2 (GX) 29 according to the assumption that no measurement error in the SNP exposure effects. I 2 (GX) is an adaptative I 2 statistic which proposes to quantify the strength of no measurement error violation for the MR-Egger method. If I 2 (GX) 29 was sufficiently low (I 2 [GX]<0.9), the correction analysis was conducted to assess the causal effect by simulation extrapolation (SIMEX), which can substantially mitigate adverse effects by simulation extrapolation. 29 The intercept term of MR-Egger method can be used for evaluating the directional pleiotropic effect. 30 When the intercept is zero or its P value was not significant (P>0.05) were considered as nonpleiotropy. Moreover, we also used the Rucker Q′ statistic 31 to measure the heterogeneity for MR-Egger method. If the difference Q−Q′ is sufficiently extreme with respect to a χ 2 distribution with the 1 degree of freedom, we indicated that directional pleiotropy is an important factor, and MR-Egger model provides a better fit than the IVW method. 32 [33] [34] [35] and the persons with non-O types associated with greater risks of significant coronary artery disease, myocardial infarction, and SARS-CoV-2 infection. 23, [36] [37] [38] [39] Considering the associations between ABO blood group and blood lipids levels or COVID-19, we also rerun MR analysis after excluding SNPs in the ABO locus to avoid potential pleiotropy.

Leave-one-out sensitivity analysis was implemented to assess whether the significant results were driven by a specific SNP. 

We used the Bonferroni approach to address multiple comparisons issue. The significant threshold was set as P<7.14×10 −3 (0.05/7 exposures).

Details of the IVs after linkage disequilibrium clumping are represented in Table I in the Data Supplement. The number of remained IVs after harmonization and radial MR are shown in Table II in the Data Supplement.

Pleiotropy Assessment As shown in Table III in the Data Supplement, the evidence of pleiotropy was observed in TC to COVID-19 infection with the data from the HGI, thus we chose MR-Egger as the main MR method for this exposureoutcome pair and IVW was used in others. In addition, the assumption of no measurement error was not violated (I 2 [GX]>0.9), therefore, we did not perform SIMEX analysis.

For the outcome data from the UKB, the MR estimates ( We detected suggestive causal effects of LDL-C (OR, 1.13 [95% CI, 1.02-1.24], P=1.46×10 −2 ) in the preliminary results. However, the leave-one-out permutation identified 3 IVs with major effects. After excluding the main influential IVs, the marginal significant association of LDL-C was not observed (P>0.05, Table 2 ). For the other blood lipids (HDL-C, triglyceride, and ApoA1), we did not detect any association (P>0.05, Table 2 ).

Since TC is mostly LDL-C and the IVs overlap substantially, we further performed MR analysis after excluding the overlapped IVs between TC and LDL-C. The significant causal effect of TC was remained (OR, 1.25 [95% CI, 1.07-1.46], P=5.08×10 −3 ), but no association was detected for LDL-C (P>0.05, Table IV in the Data  Supplement) .

For the outcome data from the HGI (Table 1) , the MR analyses did not detect significant causal associations on COVID-19 susceptibility after Bonferroni correction (P<7.14×10 −3 ; Table 2 ). However, we successfully observed the possible causal effects of TC (OR, 1.01 [95% CI, 1.00-1.02], P=2.29×10 −2 ) and ApoB (OR, 1.01 [95% CI, 1.00-1.02], P=2.22×10 −2 ; P<0.05), but the association between dyslipidemia and COVID-19 infection was no longer significant.

If there were IVs located in the ABO locus, we reperformed MR analysis for TC and ApoB after excluding those SNPs. For ApoB, there were no IVs in the ABO locus. As shown in Table V We further performed MR analysis after removing IVs on chromosome 9. As shown in Table V 

We also measured the associations between blood lipids and severe COVID-19. According to the evidence of pleiotropy (Table III in Table 3 , we did not detect any significant association for all 7 traits (P>0.05). 

In this study, we implemented 2-sample MR analyses to investigate the possible causal associations of blood lipids profiles on COVID-19 infection or severity. We detected potential causal effects of dyslipidemia, TC, and ApoB on COVID-19 susceptibility. The clinical manifestation of dyslipidemia includes the maladjustment of TC level, LDL-C level, triglyceride level, HDL-C level, and other lipid and lipoprotein levels. 7 We acknowledge that it is hard to summarize the clinical risk factor for dyslipidemia. However, considering over 70% of the dyslipidemia instruments overlapped with the TC instruments in the same effect direction, and 83.22% of dyslipidemia in UKB subjects are hypercholesterolemia, it is likely that dyslipidemia is relatively homogeneous and the clinical causal risk factor of dyslipidemia might be high TC. The specific MR results of TC and ApoB could contribute to interpret the significant MR finding for dyslipidemia. We also assessed the phenotypic correlations between TC/ApoB and COVID-19 with baseline plasma lipid measures from the UKB. We selected independent white subjects and removed the individuals with confounders of cardiovascular disease, type 2 diabetes, and the treatment of statin. There were no significant correlations between TC/ApoB and COVID-19 (P>0.05). The reason might be that the UKB lipid data were acquired over a decade before COVID-19 pandemic, and the levels of TC/ApoB may be fluctuated with the influence of dietary habits, 41 physical activity, 42 and other variates in the past decade. Although we have excluded the interference of related diseases, medical treatment, and clinical characteristics as far as possible, the retrospective study may still be disturbed by potential confounders and produce bias in phenotypic associations. However, the MR estimation just relies on genetic determination and, therefore, can avoid these issues.

TC is mainly composed of LDL-C, HDL-C, and VLDL (very-low-density lipoprotein) cholesterol. The VLDL particles mainly carry triglycerides. ApoB is the major protein component of VLDL and LDL. 43 The lipoprotein particles are in the dynamic alteration of VLDL-ILDL-LDL (ILDL; intermediate-density lipoprotein) process involving in varying composition of cholesterol, ApoB, and triglyceride. 44 It has been found that the amount of cholesterol and ApoB within LDL particles are heterogeneous between persons, and the amount of LDL-C is lower in hypertriglyceridemia. [45] [46] [47] We did not detect significant effects of LDL-C, HDL-C, and triglyceride. It indicates that the potential causality between TC and COVID-19 susceptibility is not attributed to any single component but acted like a combined effect. Although we detected associations with dyslipidemia and COVID-19 more broadly, and with ApoB and TC more specifically, our findings did not clearly prioritize a specific lipoprotein fraction associated with COVID-19 susceptibility or severity.

The positive association between TC and COVID-19 susceptibility might be related to enhanced virus invasion process. Cholesterol is considered to be involved in fusion of the viral membrane to the host cell. 48 shown in the study by Guo et al 51 depleting plasma membrane cholesterol can disrupt the lipid rafts of cellular entry, resulting in suppressed infection of an avian coronavirus (infectious bronchitis virus). Consistently, membrane cholesterol has been found to similarly facilitate SARS-CoV-2 entry via lipid rafts. 48, 50 A further study 49 also showed that loading cells with cholesterol from serum would promote the endocytic entry of pseudotyped SARS-CoV-2, suggesting that high cholesterol levels in blood may contribute to SARS-CoV-2 infection in peripheral tissue. For the causal effect of ApoB on COVID-19 susceptibility, previous studies have identified that ApoB played a vital role in hepatitis C virus infection by facilitating the fusion of virus to host hepatocyte. 43, 52, 53 However, it is unclear whether ApoB could similarly regulate the fusion process of coronaviruses, and the linkage between ApoB and SARS-CoV-2 infection need to be further investigated.

Although a previous study 54 detected causal effect of LDL-C on COVID-19 susceptibility, a more recent MR study 55 covering more subjects only identified a suggestive association with LDL-C (P=0.04) that does not meet Bonferroni standards of significance. In our study, the MR result between LDL-C and COVID-19 was marginal significant (P=0.01) ignoring Bonferroni standards, but this causal link was driven by 3 influential SNPs which need to be cautious. The different MR estimates of LDL-C are likely due to the fact of different exposure and outcome data. In particular, for COVID-19 GWAS data, both of the 2 previous studies used the HGI data sets and individuals with unknown SARS-CoV-2 infection status was used as controls. In our study, we used the GWAS data from the UKB, and the controls were confirmed by polymerase chain reaction tests. For LDL-C GWAS data, we used data sets from the Million Veteran Program which had more subjects.

This study characterized the potential causality of blood lipids to the susceptibility and severity of COVID-19 using 2-sample MR design. Our findings broaden the understanding of COVID-19 risk and firstly address that higher TC and higher ApoB will increase the odds for COVID-19 susceptibility, which may be helpful to develop effective instructions and policies to control the spread of the disease to susceptible groups. However, the limitations of the current study should be addressed. First, due to the limitation of data resource, our findings are mainly based on the European/White cohort which cannot represent the universal conclusions for other ethnic groups. Second, to minimize the potential bias 56 of association analyses and strengthen our results, we have added another GWAS data from the HGI, which contains more subjects (14 134 cases and 1 284 876 controls), and the casual effects of TC and ApoB on COVID-19 susceptibility were also detected. Under current condition, we included all the available data sets and the consistent results support the causal effects of TC and ApoB on COVID-19. Third, although we have been able to evaluate the causal effects on COVID-19 based on the available data and multiple complementary methods, the findings should be verified by additional clinical resource and in-depth exploration on the potential mechanisms underlying these causalities is needed, as well.

In summary, we performed a 2-sample MR design for blood lipids and COVID-19 and found that higher TC and ApoB levels might increase the susceptibility of COVID-19. 

We thank the Million Veteran Program (MVP) staff, researchers, and volunteers, who have contributed to MVP, and especially participants who previously served their country in the military and now generously agreed to enroll in the study (see https:// www.research.va.gov/mvp/ for more details). The GWAS data set of blood lipids we used in this study is available from dbGaP (http://www.ncbi.nlm.nih.gov/gap) under accession number phs001672.v4.p1. We thank the GRASP, the coronavirus disease 2019 (COVID-19) Host Genetics Initiative, and all the genetics consortia for making summary statistics publicly accessible for this analysis. We also thank UK Biobank 

China Medical Treatment Expert Group for Covid-19. Clinical characteristics of coronavirus disease 2019 in China

Clinical features of patients infected with 2019 novel coronavirus in Wuhan

Metabolic syndrome: definitions and controversies

Behavioral counseling interventions to promote a healthy diet and physical activity for cardiovascular disease prevention in adults with cardiovascular risk factors: US preventive services task force recommendation statement

The emerging role of dyslipidemia in diabetic microvascular complications

Dyslipidemia in obesity: mechanisms and potential targets

Functional foods and dietary supplements for the management of dyslipidaemia

Hypolipidemia is associated with the severity of COVID-19

Declined serum high density lipoprotein cholesterol is associated with the severity of COVID-19 infection

Alteration of lipid profile and value of lipids in the prediction of the length of hospital stay in COVID-19 pneumonia patients

Cholesterol metabolism-impacts on SARS-CoV-2 infection prognosis. medRxiv

Association between low-density lipoprotein cholesterol levels and risk for sepsis among patients admitted to the hospital with infection

Mendelian randomization studies: a review of the approaches used and the quality of reporting

Mendelian randomization in the era of genomewide association studies

Avoiding bias from weak instruments in Mendelian randomization studies

Causal associations between risk factors and common diseases inferred from GWAS summary data

Smoller JW; Major Depressive Disorder Working Group of the Psychiatric Genomics Consortium. Assessment of bidirectional relationships between physical activity and depression among adults: a 2-sample Mendelian randomization study

Causal associations between serum bilirubin levels and decreased stroke risk: a two-sample Mendelian randomization study

MIGen) Consortium; Geisinger-Regeneron DiscovEHR Collaboration; VA Million Veteran Program. Genetics of blood lipids among ~300,000 multi-ethnic participants of the Million Veteran Program

Million Veteran Program: a mega-biobank to study genetic influences on health and disease

Genome-wide study for circulating metabolites identifies 62 loci and reveals novel systemic effects of LPA

The COVID-19 host genetics initiative, a global initiative to elucidate the role of host genetic factors in susceptibility and severity of the SARS-CoV-2 virus pandemic

Genomewide association study of severe COVID-19 with respiratory failure

Integration of summary data from GWAS and eQTL studies predicts complex trait gene targets

PLINK: a tool set for wholegenome association and population-based linkage analyses

Improving the visualization, interpretation and analysis of two-sample summary data Mendelian randomization via the radial plot and radial regression

Mendelian randomization with invalid instruments: effect estimation and bias detection through Egger regression

The combination of estimates from different experiments

Assessing the suitability of summary data for two-sample Mendelian randomization analyses using MR-Egger regression: the role of the I2 statistic

Evaluating the potential role of pleiotropy in Mendelian randomization studies

Treatmenteffect estimates adjusted for small-study effects via a limit meta-analysis

A framework for the investigation of pleiotropy in two-sample summary data Mendelian randomization

Correlation between ABO blood group, and conventional hematological and metabolic parameters in blood donors

Non-O blood groups can be a prognostic marker of in-hospital and long-term major adverse cardiovascular events in patients with ST elevation myocardial infarction undergoing primary percutaneous coronary intervention

ABO blood group in relation to plasma lipids and proprotein convertase subtilisin/kexin type 9

Relationship between ABO blood group distribution and clinical characteristics in patients with COVID-19

Testing the association between blood type and covid-19 infection, intubation, and death. medRxiv. Preprint posted online

ABO blood group is a cardiovascular risk factor in patients with familial hypercholesterolemia

Analysis of circulating cholesterol levels as a mediator of an association between ABO blood group and coronary heart disease

Mendelian randomization with a binary exposure variable: interpretation and presentation of causal estimates

American Heart Association Nutrition Committee of the Council on Lifestyle and Cardiometabolic Health

Thrombosis and Vascular Biology; Council on Cardiovascular and Stroke Nursing; Council on Clinical Cardiology; Council on Peripheral Vascular Disease; and Stroke Council. Dietary cholesterol and cardiovascular risk: a science advisory from the

Behavioral counseling to promote a healthy diet and physical activity for cardiovascular disease prevention in adults with cardiovascular risk factors: updated evidence report and systematic review for the US Preventive Services Task Force

APOB codon 4311 polymorphism is associated with hepatitis C virus infection through altered lipid metabolism

Metabolism and modification of Apolipoprotein B-containing lipoproteins involved in dyslipidemia and atherosclerosis

Clinical implications of discordance between low-density lipoprotein cholesterol and particle number

Should apolipoprotein B replace LDL cholesterol as therapeutic targets are lowered?

Composition and distribution of low density lipoprotein fractions in hyperapobetalipoproteinemia, normolipidemia, and familial hypercholesterolemia

Cholesterol, lipoproteins, and COVID-19: basic concepts and clinical applications

The role of high cholesterol in agerelated COVID19 lethality. bioRxiv. Preprint posted online

The role of lipid metabolism in COVID-19 virus infection and as a drug target

The important role of lipid raft-mediated attachment in the infection of cultured cells by coronavirus infectious bronchitis virus beaudette strain

Role of ApoB-516C/T promoter gene polymorphism in the risk of Hepatitis C virus infection in Egyptian patients and in gender susceptibility

Hepatitis C virus E1 envelope glycoprotein interacts with apolipoproteins in facilitating entry into hepatocytes

Causal Inference for genetic obesity, cardiometabolic profile and COVID-19 susceptibility: a Mendelian Randomization Study

Cardiometabolic risk factors for COVID-19 susceptibility and severity: a Mendelian randomization analysis

Collider bias undermines our understanding of COVID-19 disease risk and severity