Gene-centric functional dissection of human genetic variation uncovers regulators of hematopoiesis

Abstract

Genome-wide association studies (GWAS) have identified thousands of variants associated with human diseases and traits. However, the majority of GWAS-implicated variants are in noncoding regions of the genome and require in depth follow-up to identify target genes and decipher biological mechanisms. Here, rather than focusing on causal variants, we have undertaken a pooled loss-of-function screen in primary hematopoietic cells to interrogate 389 candidate genes contained in 75 loci associated with red blood cell traits. Using this approach, we identify 77 genes at 38 GWAS loci, with most loci harboring 1-2 candidate genes.Importantly, the hit set was strongly enriched for genes validated through orthogonal genetic approaches. Genes identified by this approach are enriched in specific and relevant biological pathways, allowing regulators of human erythropoiesis and modifiers of blood diseases to be defined. More generally, this functional screen provides a paradigm for gene-centric follow up of GWAS for a variety of human diseases and traits.We applied a gene-centric loss-of-function screening approach to GWAS of RBC traits. We focused on 75 loci associated with RBC traits that were identified by a GWAS performed in up to 135,000 individuals (van der Harst et al., 2012) spanning 6 RBC traits ( Figure S1A). Importantly, these 75 loci have been robustly replicated and show large effect sizes in more recently reported association studies performed on larger cohorts and thus represent ideal targets for perturbation studies (Astle et al., 2016;Lareau et al., 2018). We endeavored to select all genes that could potentially underlie these 75 GWAS signals. To do this, each of the 75 sentinel SNPs was first expanded to a linkage disequilibrium (LD) block including all SNPs in high LD (r 2 > 0.8, Figure 1A, Figure S1B), then further to the nearest genomic recombination hotspot. Based upon insights from previous expression quantitative trait locus (eQTL) studies (Montgomery and Dermitzakis, 2011; Rossin et al., 2011;Veyrieras et al., 2008), each gene annotated in the genome was expanded to include a wingspan encompassing 110 kb upstream and 40 kb downstream of the transcriptional start and end sites, respectively, to also capture potential functional regulatory elements. This resulted in selection of 389 genes overlapping or in the vicinity of the LD blocks to be tested in the pooled loss-of-function screen. These were distributed at a median of 4 genes per loci ( Figure S1C).Since the majority of common genetic variation underlying RBC traits appears to act in a cellintrinsic manner within the erythroid lineage, we decided to perturb the candidate genes during the process of human erythropoiesis Sankaran et al., 2012;Sankaran et al., 2008;Ulirsch et al., 2016). We chose a pooled short hairpin RNA (shRNA) based loss-of-

The genetics of human hematopoiesis and its disruption in disease

“…High‐throughput gene knockdown or knockout screens have also become a popular experimental approach to identify genes that are important for a phenotype of interest. These screens can be implemented with short hairpin RNA or CRISPR‐Cas9 genetic perturbation to systematically test which genes or regulatory elements are essential for a phenotype, such as hematopoietic lineage differentiation (Canver et al , ; Nandakumar et al , ).…”

An integrated platform to systematically identify causal variants and genes for polygenic human traits

“…With these exceptions, NG Capture-C performs with a high rate of success in identifying effector genes linked to potential causal variants. Three previous attempts with diverse methods to identify effector genes associated with RBC traits have been reported 5,6,11 . These were an annotation-based approach 11 , Promoter Capture-HiC 5 (PC-HiC), and a gene-centric shRNA screen 6 .…”

“…The ability to compare 3C data from multiple cell types allows tissue-specific and tissue-invariant interactions to be called by a wide range of statistical approaches 23,47,59–61 , increasing the throughput of accurate effector gene identification. These candidates can then be validated with functional follow-up, such as screening approaches 6 , or as shown here, in vivo modelling, to help to explain associated cellular phenotypes.…”

“…Although non-coding regions are increasingly well annotated, many variants do not correspond to known regulatory elements, and even when they do, it is rarely known which genes these elements control, and in which cell types. New technical approaches to link variants to the genes they control are rapidly improving but are often limited by their sensitivity and resolution 3–6 ; and because so few causal variants have been unequivocally linked to the genes they affect, the mechanisms by which non-coding variants alter gene expression remain unknown in all but a few cases; and, third, the complexity of gene regulation and cell/cell interactions means that knowing when in development, in which cell type, in which activation state, and within which pathway(s) a causal variant exerts its effect is usually impossible to predict. Although significant progress is being made, currently, none of these problems has been adequately solved.…”