Background: Finding the cells where risk variants act

Explaining the genetic risk of inflammatory bowel disease (IBD) requires identifying not only where a variant lies, but also which gene’s expression it is associated with and in which cells. The IBDverse study, published in Nature, combined single-cell expression data from intestinal tissue and blood with genotypes and nominated candidate effector genes at 180 of 321 known IBD risk loci. Regulation of Notch signaling in immune cells and Wnt-related regenerative processes in epithelial cells emerged as leading candidate mechanisms.

Most association signals for complex diseases identified by genome-wide association studies (GWAS) lie in regions that do not directly encode proteins. The nearest gene cannot necessarily be assumed to be the actual effector gene. Analyses that combine multiple cell types may also obscure expression-regulatory effects found only in specific cells. This study examined these issues by comparing the same data at different cellular resolutions.

Key findings: Cell-type analysis reveals regulatory signals

The researchers recruited 421 participants, including 296 without IBD and 125 with Crohn’s disease, and generated single-cell data from intestinal biopsies and blood. The final analysis matched to genotype data included about 2.2 million cells from 396 participants. The team aggregated expression values for each participant at the levels of 86 cell types, nine major cell groups, and all cells in each sample, then examined cis-expression quantitative trait loci (cis-eQTLs), associations between nearby genetic variants and gene expression.

Of 88,548 identified eQTLs, 25,980 (29.3%) were detected only at the cell-type level. The median distance between these variants and the transcription start sites of their associated genes was 200 kb, farther than the 114 kb at the major-cell-group level and 52 kb at the all-cell level. They were also relatively more concentrated in regions annotated as enhancers, which regulate gene expression. Among eQTLs detected only at the cell-type level, the nearest gene was the expression-associated target in just 19.8% of cases. This illustrates why distance alone is insufficient for assigning effector genes.

According to the paper’s abstract, cell-type-level eQTLs were more than 3.5 times as likely as tissue-level eQTLs to colocalize with IBD GWAS signals. Colocalization is a statistical analysis that assesses the likelihood that expression-association and disease-association signals share the same causal variant. Applying a threshold requiring a posterior probability above 0.75 for a shared causal variant, the researchers proposed 419 candidate genes at 180 loci that colocalized with eQTLs or interaction eQTLs. At 74 of these loci, the Open Targets comparison data had not identified a candidate gene through colocalization.

In dendritic cells, reduced expression of the Notch-regulatory genes MAML2 and ZMIZ1 was linked to disease-risk signals. In intestinal stem cells, increased MYC expression colocalized with Crohn’s disease risk signals. In intestinal enterocytes under inflammatory conditions, Crohn’s disease risk alleles were also associated with reduced RNF14 expression. These findings illustrate why Wnt-related results should be interpreted according to the gene, cell, and inflammatory state rather than collectively described as decreased signaling.

Meaning and outlook: A map for target validation, with remaining limitations

These findings provide a research map for narrowing down candidate effector genes and cellular contexts for IBD risk variants. The Notch- and Wnt-related results are mechanistic hypotheses that require follow-up experiments; colocalization alone does not establish causality. The next step is to perturb candidate genes in relevant cells and assess the resulting functional changes.

Statistical power was limited for rare cell types. For 57% of the analyzed cell types, there were fewer than 100 cells per participant, and the colocalization analysis assumed a single causal variant. The recruited disease cohort consisted of people with Crohn’s disease, and participants of non-European ancestry were excluded during genotype quality control. Replication in tissue from people with ulcerative colitis and in other ancestry groups also needs to be assessed. The study cannot be interpreted as having validated patient-specific prognosis prediction or the efficacy and safety of targeted treatments.

The study population, analytical results, methods, and limitations are described in the Nature article.