Mendelian randomisation identifies alternative splicing of the FAS death receptor as a mediator of severe COVID-19

Abstract

Severe COVID-19 is characterised by immunopathology and epithelial injury. Proteomic studies have identified circulating proteins that are biomarkers of severe COVID-19, but cannot distinguish correlation from causation. To address this, we performed Mendelian randomisation (MR) to identify proteins that mediate severe COVID-19. Using protein quantitative trait loci (pQTL) data from the SCALLOP consortium, involving meta-analysis of up to 26,494 individuals, and COVID-19 genome-wide association data from the Host Genetics Initiative, we performed MR for 157 COVID-19 severity protein biomarkers. We identified significant MR results for five proteins: FAS, TNFRSF10A, CCL2, EPHB4 and LGALS9. Further evaluation of these candidates using sensitivity analyses and colocalization testing provided strong evidence to implicate the apoptosis-associated cytokine receptor FAS as a causal mediator of severe COVID-19. This effect was specific to severe disease. Using RNA-seq data from 4,778 individuals, we demonstrate that the pQTL at the FAS locus results from genetically influenced alternate splicing causing skipping of exon 6. We show that the risk allele for very severe COVID-19 increases the proportion of transcripts lacking exon 6, and thereby increases soluble FAS. Soluble FAS acts as a decoy receptor for FAS-ligand, inhibiting apoptosis induced through membrane-bound FAS. In summary, we demonstrate a novel genetic mechanism that contributes to risk of severe of COVID-19, highlighting a pathway that may be a promising therapeutic target.

Figure 1:

Mendelian randomisation study design and data sources. Severity-associated protein biomarkers were identified from the studies by Filbin et al and Gisby et al.

Figure 2:. Mendelian Randomisation (MR) of soluble FAS protein levels and COVID-19 outcomes.

a) Regional association plot (hg19 genome build) showing the cis-pQTL for soluble FAS (plasma) and the associations with COVID-19 outcomes. Posterior probability of a shared causal variant (PP H4) between FAS protein levels and very severe COVID-19 = 0.95. b) MR estimates of the causal effect of soluble FAS protein on different COVID-19 outcomes: susceptibility to infection, hospitalisation and very severe disease. c) Soluble FAS protein levels in COVID-19 patients, stratified by clinical severity, and non-infected controls (data from Gisby et al.). Boxplots showing distribution of plasma protein levels according to COVID-19 status at the time of blood draw. Boxplots indicate median and interquartile range. n=256 samples from 55 COVID-19 patients and 51 samples from non-infected patients. ‘COVID-19 status’ indicates clinical severity score of the patient at the time the sample was taken. Mild n=135 samples; moderate n=77 samples; severe n=29 samples; critical n= 15 samples.

Figure 3:

a) Regional association plot showing (from top to bottom): transcript isoforms, the soluble FAS cis-pQTL, and the associations with FAS exon 6 splicing. FASΔEx6 indicates the transcript isoform lacking exon 6 (red circle). The posterior probability of a shared causal variant (PP H4) between FAS protein levels and exon 6 splicing = 0.99. b) Boxplot showing exon 6 splice quantitative trait locus (sQTL). Number of individuals by genotype at rs982764: 443 (CC), 1992 (CT), 2329 (TT). In 14 individuals genotype could not be reliably ascertained. Y-axis represents % of transcripts with exon 6 skipping. P for association with genotype 4×10⁻²² (linear model). c) Proposed model by which genetic variation in FAS increases risk of severe COVID-19. A non-coding variant acts as a splice QTL. The risk allele for very severe COVID-19 (rs982764:T) is associated with an increased proportion of transcripts lacking exon 6 resulting in higher levels of soluble FAS (sFAS). sFAS acts as a decoy receptor for FAS-ligand (FASL), blocking FASL binding to membrane-bound FAS (mFAS) on the cell surface and thus reducing apoptosis.