Systematic assessment of COVID-19 host genetics using whole genome sequencing data

Abstract

Courses of SARS-CoV-2 infections are highly variable, ranging from asymptomatic to lethal COVID-19. Though research has shown that host genetic factors contribute to this variability, cohort-based joint analyses of variants from the entire allelic spectrum in individuals with confirmed SARS-CoV-2 infections are still lacking. Here, we present the results of whole genome sequencing in 1,220 mainly vaccine-naïve individuals with confirmed SARS-CoV-2 infection, including 827 hospitalized COVID-19 cases. We observed the presence of autosomal-recessive or likely compound heterozygous monogenic disorders in six individuals, all of which were hospitalized and significantly younger than the rest of the cohort. We did not observe any suggestive causal variants in or around the established risk gene TLR7. Burden testing in the largest population subgroup (i.e., Europeans) suggested nominal enrichments of rare variants in coding and non-coding regions of interferon immune response genes in the overall analysis and male subgroup. Case-control analyses of more common variants confirmed associations with previously reported risk loci, with the key locus at 3p21 reaching genome-wide significance. Polygenic scores accurately captured risk in an age-dependent manner. By enabling joint analyses of different types of variation across the entire frequency spectrum, this data will continue to contribute to the elucidation of COVID-19 etiology.

Fig 1. The DeCOI and the DeCOI_EUR cohort.

(A) Individuals in the DeCOI cohort are classified into three phenotypes based on WHO definition. In addition, the cohort was subsetted to an unrelated cohort of the European population (DeCOI_EUR) for association analyses. Based on the phenotypes, case-control definitions were established within DeCOI_EUR. (B) Composition of the DeCOI cohort according to sex (inner circle), phenotype (color coded, middle circle), and population (outer circle). Shaded intervals in the outer circle represent non-European individuals. (C) Age distribution of individuals from the DeCOI cohort (n = 1,220) and the European subcohort (DeCOI_EUR; n = 1,017), as stratified according to severity (color coded). In both subcohorts, the average age increases with disease course severity. Numbers indicate individuals in the respective group. (D) Phenotype distribution of individuals harboring ClinVar-annotated variants, as grouped according to disorder class. Autosomal recessive patterns of inheritance (AR/likely compound-heterozygous (CH), n = 6 diseases in six individuals) are displayed in the upper panel, and autosomal dominant inheritance patterns (AD, n = 79 diseases in 77 individuals) are displayed in the lower panel.

Fig 2. Effect sizes of nominally significant gene-set based tests in the DeCOI_EUR cohort.

Gene-sets and the corresponding functional masks (S4 Table) that were tested are given on the y-axis. On the x-axis, effect size estimates (betas) are shown as markers with error bars indicating the standard errors of betas. Note that phenotypes are color-coded, and the markers outlined in black indicate analyses that only included males. Nominally significant findings were only obtained in the overall analyses and male sub-stratification. None was observed in female-only or age-stratified analyses. A list of genes that were included in each gene-set can be found in S3 Table.

Fig 3. Analysis of common variants within the DeCOI_EUR cohort.

(A) and (B): Manhattan plots of association analyses of single variants (MAF>0.5%) in DeCOI_EUR (n = 1,017 individuals), for phenotype Ex (272 severely affected individuals vs. 362 mild controls) and B1 (655 hospitalized individuals vs. 362 non-hospitalized controls), respectively. Genomic inflation factors were 1.04 (Ex) and 1.00 (B1). Among the strongest associations is the well-established risk locus at 3p21.31. Panels (C) and (D) show the distribution of individual polygenic risk scores (PRS) among cases (orange or yellow) and controls (gray) of Ex (C) or B1 (D) overall (density plots in the left parts) or when stratified according to age below or above 60 years (box plots in the right parts). The elements of the box plots correspond to the following values: thick line: median, box: 25th and 75th percentile, whiskers: largest / smallest value not further away from the box than 1.5 times the interquartile range, points: values outside of the range of the whiskers. *: p<0.05, ***: p<0.001; Wald test followed by Bonferroni correction. MAF: Minor Allele Frequency.