Discovery of core genes for systemic lupus erythematosus via genome-wide aggregated trans-effects analysis

Abstract

The "omnigenic" hypothesis postulates that the polygenic effects of common variants on a typical complex trait coalesce on relatively few core genes through trans-effects on their expression. Our aim was to identify core genes for systemic lupus erythematosus (SLE) by testing for association with genome-wide aggregated trans-effects (GATE) scores for gene expression in a large genetic dataset (5267/4909 SLE cases/controls). SLE was strongly associated with upregulation of expression of eight interferon-stimulated genes driven by shared trans-effects. We estimate that trans-effects on interferon signaling account for 9% of the total genetic effect on SLE risk. Outside this pathway, GATE analysis detected twenty putative core genes for SLE. Direct protein measurements for these genes were strongly associated with SLE in UK Biobank. Two putative core genes (TNFRSF17, TNFRSF13B) encode receptors (BCMA, TACI) expressed on B cells; their ligands (BAFF, APRIL) are targeted by drugs licensed or in development for SLE. Four genes (PDCD1, LAG3, TNFRSF9, CD27) encode receptors that have been characterized as immune checkpoints, and three (CD5L, SIGLEC1, CXCL13) are biomarkers of SLE disease activity. These results provide genetic support for existing drug targets in SLE (interferon signaling, BAFF/APRIL signaling) and identify other possible therapeutic targets including immune checkpoint receptors.

Fig. 1. Study overview.

A Flowchart of Genome-wide Aggregated Trans-Effects (GATE) analysis. B Summary of follow-up analyses and external evidence for validating putative core genes. C Overview of key findings from application to systemic lupus erythematosus (SLE). Parts of this image were created in BioRender. Iakovliev, A. (2024), https://BioRender.com/d04z288.

Fig. 2. Heatmap of correlations among GATE scores for genes detected by GATE analysis ordered by hierarchical clustering.

Correlations were computed using SLE controls in the discovery dataset. The three-letter prefix, eqtl or pqtl denotes if the corresponding gene was detected via aggregated trans-effects on expression or protein level, respectively. Blue labels are used to highlight the cluster of interferon stimulated genes (ISGs) with highly correlated GATE scores.

Fig. 3. Manhattan plot for the association of genetic variants with SLE using summary statistics from Bentham et al. [15] annotated with known SLE GWAS hits from the GWAS Catalog (top) and with eQTLs for MX1 gene expression (bottom).

Blue lines point to trans-eQTLs for MX1. Magenta line points to the cis-eQTL.

Fig. 4. Effects of individual eQTLs for MX1, on MX1 expression level (x-axis) and on SLE risk (y-axis).

Blue points show effects of trans-eQTLs, labeled by nearby genes. Red point shows effects of the cis-eQTL. The size of the marker is inversely proportional to the standard error of the Wald ratio estimate for the effect of MX1 expression level on SLE based on each individual eQTL. The slope of the line is the causal effect estimate of MX1 expression (exposure) on SLE risk (outcome) using all trans-eQTLs as instrumental variables. The red shaded area contains all possible lines passing through the origin whose slope is within one standard error of the estimate. Estimation is performed by marginalizing over the posterior distribution of (unobserved) pleiotropic effects of the trans-eQTLs. IFIH1, IKZF1 and WDFY4 are known GWAS hits for SLE. Rare variants in IFIH1, IKZF1 and RIGI (also known as DDX58) are known to cause monogenic forms of lupus.

Fig. 5. Forest plot for the associations of SLE with measured transcript levels (in PRECISESADS) for genes detected by eGATE analysis, and with measured protein levels (in UKB-PPP) for genes detected by pGATE analysis.

For each gene the log(odds ratio) and 95% confidence interval corresponding to one standard deviation change in the measured transcript/protein level is shown. Gene symbols in brackets represent a cluster of interferon stimulated genes (ISGs) with highly correlated GATE scores. LGALS2, MX1, ZG16, and ALKBH2 were not measured in UKB-PPP and could not be evaluated in this analysis.