Lude Franke

Human genetic studies have identified key regulators of fetal haemoglobin (HbF) expression, including BCL11A, resulting in therapeutic advances ^1–8 . Yet the mechanisms by which HbF expression is activated remain incompletely understood ⁹ . Here we conduct a large multi-ancestry genome-wide association study of HbF levels in 28,279 individuals that identifies 91 conditionally independent associations across 12 genomic regions. In one previously uncharacterized associated region, the high-HbF-linked causal variant rs1010474-C reduces BACH2 expression and elevates HbF levels. Direct perturbation or inhibition of BACH2 likewise increases HbF expression. Mechanistically, BACH2 restrains activation of the HbF-encoding γ-globin genes, while loss of BACH2 enhances NRF2 chromatin occupancy and promotes the formation of activation foci at the γ-globin genes. Although BACH2 and NRF2 binding motifs in the γ-globin promoters overlap, they can be selectively edited to activate or repress γ-globin, respectively, and do so independently of BCL11A. These findings illustrate how human genetic variation continues to advance our understanding of therapeutically relevant regulatory mechanisms underlying HbF expression.

Genome-wide association studies have identified thousands of variants associated with immune-related diseases, yet most lie in non-coding regions, complicating mechanistic interpretation. Regulatory quantitative trait loci (QTLs), such as expression QTLs (eQTLs) and chromatin accessibility QTLs (caQTLs), offer a powerful framework for prioritization and interpretation of these disease-associated genetic variants. When analyzed together, they offer deeper insights into the regulatory architecture underlying disease. We generated same-cell, single-cell multi-omics data, integrating transcriptomic and chromatin accessibility information, from 563,100 matched peripheral blood mononuclear cells collected from 264 individuals, either unstimulated or stimulated for 24h with C. albicans (CA). Across six major immune cell types, we mapped both cis -eQTLs and -caQTLs, identifying 1,571 eGenes and 28,862 caPeaks, with 41% and 11% showing a stimulation-dependent effect. Finally, to dissect the regulatory mechanisms underlying these QTL effects, we applied two complementary strategies: 1. overlapping caQTLs with eQTLs; 2. applying SCENIC+ to identify regulatory triplets containing a transcription factor, the chromatin region it may bind to and the candidate target genes it thereby may regulate. With the first approach, we identified 1,861 dual-acting QTLs. These dual-QTLs showed 1.9-fold stronger enrichment for immune-related disease associations than single-modality QTLs, highlighting their relevance for disease interpretation. With the second approach, we found 62,932 regulatory triplets, of which 1.7% were under genetic control. By then leveraging the SCENIC+-derived TF activity measurements we could study how genetic variants can rewire TF control of gene expression, ultimately shaping inter-individual variation in disease risk. Together, our network-based approach offers new insights into the cellular contexts and gene programs perturbed in disease, providing a foundation for prioritizing therapeutic targets and informing strategies for disease prevention.

Many non-coding variants influence complex traits and diseases through gene regulation, yet the mechanisms linking these variants to downstream biology remain poorly understood. Here, we present eQTLGen Phase 2 , a comprehensive genome-wide analysis of gene expression quantitative trait loci (eQTLs) in 43,301 blood samples from 52 datasets. Beyond local cis -effects, this sample size enabled the first systematic mapping of trans -eQTLs at scale. We identify cis -eQTLs for nearly all expressed genes (94.7%) and trans -eQTLs for over half (56.2%). Second, by colocalizing cis -eQTLs with trans -eQTLs, we infer a directed gene regulatory network comprising 47,554 directed gene regulatory relationships. These networks reveal how genetic perturbations in upstream regulators produce dose-dependent downstream effects, supported by Perturb-seq and ChIP-seq data. Third, integrating this network with 87 genome-wide association studies allows us to systematically prioritize trait-relevant pathways and candidate genes. Variants exerting both cis – and trans -effects are markedly more likely to colocalize with trait associations than cis -only variants, delineating a subset of functionally active cis -eQTLs from a large group with limited downstream impact. This distinction provides a conceptual framework for identifying regulatory variants that truly mediate complex trait biology. Together, these results provide a publicly available resource of cis – and trans -eQTLs and an in vivo scaffold for human gene-regulatory networks, elucidating how propagation of cis -effects modulates complex disease.

Promoters are the core regulatory elements of all genes. Their activity ensures the correct transcription level of each individual gene, which is essential for cellular homeostasis and responses to a wide range of signals. One of the major challenges in genomics is to build computational models that accurately predict genome-wide gene expression from the sequences of regulatory elements¹. Here we present promoter activity regulatory model (PARM), a cell-type-specific deep-learning model trained on specially designed massively parallel reporter assays (MPRAs) that query human promoter sequences. PARM is experimentally and computationally lightweight so that cell-type-specific and condition-specific models can be generated that reliably predict autonomous promoter activity across the genome from the DNA sequence alone. PARM can also design purely synthetic strong promoters. We leveraged PARM to systematically identify binding sites of transcription factors that probably contribute to the activity of each natural human promoter and to detect the rewiring of these regulatory interactions after various stimuli to the cells. We also uncovered and experimentally confirmed substantial positional preferences of transcription factors that differ between activating and repressive regulatory functions and a complex grammar of motif–motif interactions. Our approach provides a highly economic strategy towards a deeper understanding of the dynamic regulation of human promoters by transcription factors.

Genetic effects on gene expression are often cell type-specific and obscured in bulk analyses. To resolve this context-dependent regulation, we performed a federated cis- eQTL meta-analysis across 12 PBMC datasets (2,032 individuals, 2.5 million cells). Across six immune cell types, we identified cis- eQTLs for 6,592 genes and fine-mapped 14,985 independent loci. Notably, the 42% of eQTLs that were undetected in a bulk eQTL study on 43,301 whole blood samples also showed stronger enrichment for disease GWAS loci. We further identified three genome-wide significant and 65 suggestive loci affecting the abundance of (rare) immune cell types and validated these using previously reported hematological GWAS and bulk-derived trans- eQTLs. Integrating single-cell cis -eQTLs with bulk trans- eQTLs enabled us to anchor 6,382 trans- eGenes (37.2% novel) to upstream regulators and reconstruct directed gene regulatory relationships. For example, a hemorrhoidal disease-associated variant showed a CD4+ T cell-specific cis- eQTL on BACH1 that colocalized with 45 immune and metabolic trans -eGenes. These results demonstrate the power of single-cell QTL meta-analysis in interpreting complex trait genetics.

Identifying causal mechanisms from genome-wide association studies (GWAS) requires an understanding of how disease-associated genetic variants influence gene expression in specific cell types. Here, we present scMetaBrain, a large-scale single-nucleus RNA-sequencing (snRNA-seq) resource derived using 1,260 samples from 785 individuals spanning 10 brain datasets. By analyzing 3.9 million transcriptomes, we identified 19,371 unique expression quantitative trait locus (eQTL) genes (eGenes) at a major cell type level, with the largest number of eQTLs observed in excitatory neurons. Notably, 31% of the eQTLs detected were highly cell-type-specific, with most restricted to excitatory neurons (69%). We compared the eQTLs with bulk RNA-seq datasets across different tissues and with a newly generated single nucleus dataset of 123 donors from peripheral blood mononuclear cells. We observed that differences in eQTL effect sizes between brain cell types are often as large as comparing eQTLs between brain tissue and non-brain tissue from bulk RNA-seq studies. Furthermore, we observe that eQTL effect size agreement was highest for cell types with similar function, even when comparing brain to blood cells. This suggests that that bulk analyses substantially overestimate eQTL agreement, likely due to tissue-level averaging of cellular regulatory effects. Through colocalization, we prioritized 662 genes for 11 brain-related traits and prioritized a single cell type in 68% of genes. Our findings demonstrate that eQTL effects are far more cell-type-specific than previously recognized, underscoring the need to expand single-cell eQTL studies across diverse tissues and cell types to fully capture the regulatory architecture of genetic variants.

Aims/hypothesis: Despite playing critical roles in the pathophysiology of type 2 diabetes and other metabolic disorders, the molecular mechanisms underlying circulating adipokine levels remain poorly understood. By identifying genomic regions involved in the regulation of adipokine levels and adipokine-mediated disease risk, we can improve our understanding of type 2 diabetes pathogenesis and inter-individual differences in metabolic risk. Methods: We conducted an epigenome-wide meta-analysis of associations between serum adiponectin ( n =2791) and leptin ( n =3661) and leukocyte DNA methylation at over 400,000 CpG sites across five European cohorts. The resulting methylation signatures were followed up using functional genomics, integrative analyses and causal inference methods. Results: Our findings revealed robust associations with adiponectin at 73 CpGs and leptin at 211 CpGs. Many of the identified sites were also associated with risk factors for the metabolic syndrome and located in enhancers close to relevant transcription factor binding sites. Integrative analyses additionally linked 35 of the adiponectin-associated CpGs to the expression of 46 genes, and 100 of the leptin-associated CpGs to the expression of 151 genes, with implicated genes enriched for lipid transport (e.g. ABCG1 ), metabolism (e.g. CPT1A ) and biosynthesis (e.g. DHCR24 ). Bidirectional two-sample Mendelian randomisation further identified two specific CpG sites as plausible drivers of both adiponectin levels and metabolic health: one annotated to ADIPOQ , the gene encoding adiponectin; and another linked to the expression of SREBF1 , an established modifier of type 2 diabetes risk known to exert its effects via adiponectin. Conclusions/interpretation: Taken together, these large-scale and integrative analyses uncovered links between adipokines and widespread, yet functionally specific, differences in regulation of genes with a central role in type 2 diabetes and its risk factors. Graphical

Although genome-wide association studies have provided valuable insights into the genetic basis of complex traits and diseases, translating these findings to causal genes and their downstream mechanisms remains challenging. We performed trans expression quantitative trait locus (trans-eQTL) meta-analysis in 3734 lymphoblastoid cell line samples, identifying four robust loci that replicated in an independent multi-ethnic dataset of 682 individuals. The trans-eQTL signal at the ubiquitin specific peptidase 18 (USP18) locus colocalised with a GWAS signal for systemic lupus erythematosus (SLE). USP18 is a known negative regulator of interferon signalling and the SLE risk allele increased the expression of 50 interferon-inducible genes, suggesting that the risk allele impairs USP18’s ability to effectively limit the interferon response. Intriguingly, the USP18 trans-eQTL signal would not have been discovered in a meta-analysis of up to 43,301 whole blood samples, reaffirming the importance of capturing context-specific genetic effects for GWAS interpretation.

Wellbeing is associated with both behavioral phenotypes as well as several key life outcomes, such as health, employment, and coping with stressful events. These phenotypes associated with wellbeing could be potential indicators of differential epigenetic patterns between individuals that differ in their levels of wellbeing. We performed the largest epigenome-wide (EWAS) meta-analysis of wellbeing to date by combining whole blood DNA methylation data (Illumina 450k array) from 13 cohorts from Europe, Australia, and the USA ( N = 10,757 participants). After correcting for smoking and BMI, no epigenome-wide significant methylation sites were identified. We tested whether a weighted methylation score (MS) based on leave-one-cohort-out EWAS meta-analysis summary statistics predicted wellbeing in an independent cohort, and whether prediction was significant over and above the polygenic score (PGS) for wellbeing. The MS was associated with wellbeing (variance explained=0.22%, p =0.03) and was no longer significant after adding the polygenic score (PGS; variance explained=0.43%, p =0.0046, MS; variance explained=0.07%, p =0.2842). We further compared DNA methylation levels in 16 pairs of monozygotic twins discordant for wellbeing. These analyses revealed no significant within-pair DNA methylation differences at the top-sites from the meta-analysis or in MS. Our results suggest that larger EWAS meta-analyses with uniform phenotype assessment are required to identify methylation sites associated with wellbeing.

In coeliac disease (CeD), the epithelial lining (EL) of the small intestine is severely damaged by a complex auto-inflammatory response, leading intraepithelial lymphocytes to attack epithelial cells. To understand the intestinal changes and genetic regulation in CeD, we investigated the heterogeneity in the transcriptomic profile of the duodenal EL using RNA-seq and eQTL analysis on predicted cell types. The study included duodenal biopsies from 82 patients, grouped into controls, gluten-free diet treated CeD and untreated CeD. We identified 1 862 differential expressed genes, which clustered into four sets. Two sets, one upregulated for cell cycle function (n = 366) and one downregulated for digestion, transmembrane transport, and laminin pathways (n = 543), defined three sample groups based on inflammation status: non-inflamed, mild inflammation or severe inflammation. The remaining two sets of genes were enriched for immune (n = 458) and extracellular matrix and barrier functions (n = 495) and were sufficient to classify samples into their disease conditions. Finally, deconvoluting eQTL effects from epithelial and immune cells identified 6 and 15 cell-type-mediated eQTL genes, respectively. In sum, we identified genes expressed in the duodenal EL whose expression reflect heterogeneity in CeD and that may be used as biomarkers to assess CeD condition and its mucosal and immune status.