Novel genes and sex differences in COVID-19 severity

Abstract

Here, we describe the results of a genome-wide study conducted in 11 939 coronavirus disease 2019 (COVID-19) positive cases with an extensive clinical information that were recruited from 34 hospitals across Spain (SCOURGE consortium). In sex-disaggregated genome-wide association studies for COVID-19 hospitalization, genome-wide significance (P < 5 × 10-8) was crossed for variants in 3p21.31 and 21q22.11 loci only among males (P = 1.3 × 10-22 and P = 8.1 × 10-12, respectively), and for variants in 9q21.32 near TLE1 only among females (P = 4.4 × 10-8). In a second phase, results were combined with an independent Spanish cohort (1598 COVID-19 cases and 1068 population controls), revealing in the overall analysis two novel risk loci in 9p13.3 and 19q13.12, with fine-mapping prioritized variants functionally associated with AQP3 (P = 2.7 × 10-8) and ARHGAP33 (P = 1.3 × 10-8), respectively. The meta-analysis of both phases with four European studies stratified by sex from the Host Genetics Initiative (HGI) confirmed the association of the 3p21.31 and 21q22.11 loci predominantly in males and replicated a recently reported variant in 11p13 (ELF5, P = 4.1 × 10-8). Six of the COVID-19 HGI discovered loci were replicated and an HGI-based genetic risk score predicted the severity strata in SCOURGE. We also found more SNP-heritability and larger heritability differences by age (<60 or ≥60 years) among males than among females. Parallel genome-wide screening of inbreeding depression in SCOURGE also showed an effect of homozygosity in COVID-19 hospitalization and severity and this effect was stronger among older males. In summary, new candidate genes for COVID-19 severity and evidence supporting genetic disparities among sexes are provided.

Figure 1

Association results of SCOURGE for global and sex-disaggregated A1 hospitalization analysis. (A) Manhattan plot of results from global analysis. A quantile–quantile plot of the global analysis is also shown as an inset. (B) Miami plot of results from sex-disaggregated analyses (top: males and bottom: females).

Figure 2

Regional plot of a novel association at 9q21.32 found among females from the SCOURGE study. The x axis reflects the chromosomal position, and the y axis the −log(P-value). The sentinel variant is indicated by a diamond and all other variants are colour coded by their degree of LD with the sentinel variant in Europeans. Credible set for this signal is shown within a dashed square. The horizontal dotted blue line corresponds to the threshold for genome-wide significance (P = 5 × 10⁻⁸).

Figure 3

Lookup of previously found COVID-19 host risk factors in the SCOURGE study. Heatmap illustrating the results in all analyses performed in this study (rows) for the 13 leading variants in the COVID-19 HGI study (columns). Each box illustrates the top associated variant within the focal region. The color (gray to dark red) indicates the strength (significance level) of the association in SCOURGE. Note: In three cases (chr12: 112919388, chr19: 4719431 and chr21: 33242905), the imputed variants did not pass QC filters in SCOURGE and they were replaced by the nearest QC-ed imputed variant (chr12:112919404, chr19:4719822 and chr21:33241950, respectively).

Figure 4

Regional plots of novel association signals found from the meta-analysis between the SCOURGE and CNIO studies. Regional plots of novel association signals found in 9p13.3 (A–C), 17q21.31 (D–F) and 19q13.12 (G–I). The x axes reflect the chromosomal position, and the y axes the −log(P-value) of the SCOURGE-CNIO meta-analysis. On A, D and G the sentinel variant is indicated by a diamond and all other variants are color coded by their degree of LD with the sentinel variant in Europeans. Whenever a concise set of variants was prioritized, a credible set is shown within a dashed square. The horizontal dotted blue line corresponds to the threshold for genome-wide significance (P = 5 × 10⁻⁸). In the rest of panels, the x axes reflect the chromosomal position, and the y axes the −log(P-value) resulting from the eQTL analyses in whole blood (B, E and H) and in the lung (C, F and I) whenever a significant finding is available from GTEx v8.

Figure 5

Forest plot of the SNP-heritability estimates for the COVID-19 hospitalization risk analysis on the liability scale.

Figure 6

Effect of the ID on COVID-19 hospitalization in the SCOURGE cohort. Forest plots are shown for global analyses as well as for sex and age-disaggregated analyses.