This is a two-sample mendelian randomization study. We obtained genetic instruments for nitric oxide and platelets from the largest publicly available published studies. Preferentially we selected genetic instruments on functional grounds, because functional variants known physiologically to correspond to a specific phenotype are most likely to capture it both comprehensively and exclusively. Otherwise, we selected genetic variants statistically (as reaching genome wide significance. We applied these genetic predictors of nitric oxide and platelets to the largest publicly available genome wide association study (GWAS) of COVID-19 to determine if people with genetically different levels of nitric oxide or platelets also differed in their vulnerability to very severe COVID-19, and obtained mendelian randomization estimates. We similarly considered any and hospitalized COVID-19.
Exposures
Nitric Oxide
We used three established functional genetic variants relevant to endothelial nitric oxide synthase (eNOS),13 i.e., rs2070744 (NOS3), rs1799983 (NOS3) and rs3918226 (NOS3).14 We did not include the 4b/4a VNTR variant (rs61722009 (NOS3)), because it is seldom included in GWAS, and is not in the 1000 Genomes reference panel.
Platelets
We used both platelet reactivity in response to thrombin and platelet count, because few genetic predictors of platelet reactivity are available, and platelet count may be phenotypically similar to platelet reactivity.10 One genetic variant (rs10886430 (GRK5)) functionally and statistically relevant to thrombin induced platelet reactivity is available.10 We extracted genetic variants for platelet count from a published GWAS giving independent genome-wide significant genetic predictors.15 We presented platelet reactivity estimates in terms of a square root transform, which was used to normalize the distribution,10 and platelet count in effect sizes.15
Outcome
We obtained associations with COVID-19 from the latest publicly available GWAS summary statistics. (https://www.covid19hg.org/results/) (accessed 2nd July 2020) comparing genetic make-up for different severities of COVID-19 with the population, i.e., very severe COVID-19 (cases=536, non-cases=329,391), hospitalized COVID-19 (n=3199, non-cases=897,488), and any COVID-19 (cases=6,696, non-cases=1,073,072). Case status for very severe COVID-19 was laboratory confirmed COVID-19 hospitalized with respiratory support or death. Case status for hospitalized COVID-19 was hospitalized with laboratory confirmed infection, hospitalization due to COVID-19-related symptoms or self-reported hospitalized COVID-19 positive. Case status for any COVID-19 was laboratory confirmed infection, doctor diagnosis or self-report. The COVID-19 GWAS is mainly based on people of European descent from existing cohort studies and was adjusted for study covariates, principal components, age, sex, age2 and sex*age, as appropriate. https://docs.google.com/document/d/1Pcq1jttF8W7ifEUXA6-a1WVMsUyEoAybS6IqvuP-Uv8/edit?ts=5e964dc2#.
Statistical Analysis
We aligned genetic variants for the exposures (eNOS or platelets) and genetic associations with COVID-19 on the same effect allele. There were no palindromic SNPs for eNOS or platelet reactivity. For the genetic variants predicting platelet count we did not drop palindromic SNPs because the GWAS for platelet count and COVID-19 both used the same strand direction. For genetic variants not in the COVID-19 GWAS, proxies (r2>0.8) were identified. We meta-analyzed genetic variant specific Wald estimates (genetic variant on COVID-19 divided by genetic variant on exposure) with the standard error obtained from the first term of Fieller’s theorem.16 We used inverse variance weighting, with multiplicative random effects for three or more genetic variants. We also used the weighted median and MR-Egger estimates as sensitivity analysis,17 where possible (i.e., when three or more independent genetic predictors of each exposure were available). Since variants for nitric oxide are correlated, we obtained their correlations using LdLink (https://ldlink.nci.nih.gov/), and included them in the analysis. We obtained mendelian randomization estimates using the MendenlianRandomization R package. All analyses were performed using R Version 3.6.1 (R Development Core Team, Vienna, Austria).
Ethics approval
This study only used publicly available summary statistics and hence no ethics approval is required. This study complies with the Declaration of Helsinki.