Transcription Factor Networks Disproportionately Enrich for Heritability of Blood Cell Phenotypes

The production of blood cells varies between individuals, with a significant underlying genetic component. We and others have uncovered much of the heritable variation underlying these traits through genome-wide association studies (GWAS), particularly for phenotypes impacting the erythroid lineage. However, our understanding of the mechanistic basis for this variation remains poorly understood. While some efforts have focused on variant-centric approaches, given that this variation is enriched in transcriptional regulatory elements, we sought to tackle this challenge by developing systematic approaches that would be applicable to all variation in the genome. Specifically, we developed a workflow to directly perturb individual master regulator transcription factors (TFs) and then sought to read out their effects on chromatin accessibility and gene expression in individual cells to identify TF perturbation-sensitive gene regulatory networks. We then sought to use the resultant TF-driven gene regulatory networks to decipher whether there was specific enrichment for variation impacting hematopoietic phenotypes.

We initially designed a lentiviral library of guide RNAs targeting 18 hematopoietic master regulator TFs, which was introduced into adult CD34+ purified human stem and progenitor cells (HSPCs) at a low MOI so each cell carries out a single perturbation, along with Cas9 protein. These cells were then subject to erythroid differentiation. Across 4 timepoints that encompass the full spectrum of hematopoietic and erythroid differentiation, 137,604 perturbed cells were analyzed for joint single-cell ATAC and RNA sequencing to identify perturbation-sensitive regulatory elements and genes, as well as to establish correlations between these. As expected, we uncovered that most of the TF perturbation-sensitive genes were found in the vicinity of TF perturbation-sensitive enhancers. For example, GFI1B perturbation-sensitive genes were highly correlated with genes neighboring GFI1B perturbation-sensitive enhancers, enabling a reliable connection between regulatory elements and genes controlled by a given TF. By integrating our data with existing ChIP-seq data for GATA1 and NFE2, we could validate these findings and demonstrate enrichments in relevant regions. However, we would note that not all TF bound regions were equivalently sensitive to perturbation of specific TFs, illuminating the value of this perturbation-based dataset.

Given the regulatory networks we had identified, we then sought to identify how phenotype-associated variation can alter these regulatory networks at specific stages of hematopoiesis and erythroid differentiation. We initially examined specific regulatory networks and could identify potential targets of phenotype-associated variation, such as at the GATA1 perturbation-sensitive CPEB4 locus. Remarkably, we found that these TF-perturbation sensitive gene regulatory networks were >100-fold enriched for heritable variation underlying relevant hematopoietic traits, demonstrating how these TF-sensitive gene regulatory networks uniquely identify regulatory regions subject to interindividual heritable variation.

Our work paves the way to systematically map hematopoietic gene regulatory networks and to better characterize how human genetic variation influences the process of hematopoiesis. Ultimately, these insights could enable many further opportunities to use genetic insights to modulate human hematopoiesis for therapeutic purposes.