This is a massive team effort! 💪 Huge thanks to the donors and staff @finngen.bsky.social and Finnish Blood Service for sample collection, the KCO team @broadinstitute.org for multiome sequencing, @thexavierlab.bsky.social @dalygene.bsky.social for supervision, and the entire team for their help!

Summary:
1️⃣ Population-scale multiome maps immune regulation
2️⃣ Genetic regulation operates through hierarchical, cell-type specific mechanisms
3️⃣ GWAS coloc & base editing reveal causal disease drivers
4️⃣ Regulatory buffering reconciles the GWAS-eQTL paradox

At the IL4R/IL21R locus, we identified two distinct causal variants within the same Finnish-enriched haplotype. Using our multiome and BBJ GWAS, we pinpointed the exact non-coding driver for IL21R expression in NK cells, distinct from the adjacent IL4R missense variant.

We identified this TNRC18 variant overlaps chromatin accessibility peak and decreases gene expression in CD8+ T cells.

Functional validation via STAG-seq validated its down-regulation and provided deeper mechanistic insight.
See Liu et al for details: www.biorxiv.org/content/10.1...

The Finnish-enriched variants help us identify disease drivers 🇫🇮

e.g., TNRC18 intron variant (114x enriched) is a risk factor for IBD but protective against autoimmune hypothyroidism.

Evolution tolerates chromatin variation but maintains essential gene expression levels via attenuated enhancer–gene links. This buffering makes caQTLs more detectable than eQTLs at disease loci targeting highly constrained genes 🧬🛡️

We found 1) constrained genes actually colocalize MORE with eQTLs but with smaller effects, and 2) "multi-layered regulatory buffering" where caQTLs occur with normal effect sizes, but transmission to gene expression is attenuated via systematically weaker enhancer–gene links

Motivated by higher GWAS colocalization for caQTLs (52%) than eQTLs (27%), we tackled a major paradox: Why do disease variants target constrained (essential) genes, yet eQTLs are often depleted? 🤔

c.f. Mostafavi et al (2023)
www.nature.com/articles/s41...

Crucially, these regulatory mechanism predicts disease relevance 🧬🚨

Full cascade variants (caQTL + eQTL + Link) show 2x higher GWAS colocalization rates compared to those affecting chromatin alone, establishing a clear hierarchy for prioritizing causal variants.

How do variants translate to function? We classified ~120K fine-mapped molQTLs into mechanistic categories, identifying ~3% to show a "Full Cascade" (Variant → Chromatin Δ → Expression Δ). The majority affect chromatin without detectable expression changes in baseline PBMCs.

To characterize cell-type specificity and regulatory mechanisms of molQTLs, we developed a new framework called CASCADE. We identified that while ~73% of eGenes are shared across lineages, individual regulatory variants often act in specific cell types.

Leveraging this population-scale multiome, we identified 51,083 cis-eQTLs for 20,829 genes, 338,100 cis-caQTLs for 210,584 peaks, 119,094 fine-mapped variants, and 496,488 enhancer–gene links. This allows us to trace regulatory mechanisms across immune cell types 📊🩸