Transcriptional Control of Erythropoiesis Via a Kit-Response Cistrome

Cis-regulatory elements (cis-elements) interact with transcription factors to control when and how genes are expressed in response to specific extracellular stimuli. Chromatin signatures at cis-elements can be used to predict their activities and mechanisms of transcriptional control. However, accurate cell type-specific prediction of functional cis-elements remains a significant computational challenge due to the lack of ideal datasets focused on molecular features and locus-specific sequences. Moreover, static predictions of activity do not evaluate the potential for cis-element responses to extracellular growth factor stimulation. Despite these limitations, clinically-relevant cis-elements regulating hematopoietic processes have been identified using these approaches, leading to new treatments for sickle cell disease and screening for susceptibility to myelodysplastic syndromes.

To improve predictions of cis-element activity downstream of specific cell signaling pathways, we developed a multi-step approach to annotate, rank, and functionally-validate cis-elements. We chose to evaluate predictive features of cis-element activity at nearby or interacting regions with Kit-regulated genes (termed “Kit response element”, KRE) since this signal is essential for hematopoietic and erythroid progenitor cell (EPC) survival, proliferation, and lineage commitment. We hypothesize that Kit-response cis-elements (KREs) are required for transcription and EPC activity. RNA-seq data in acutely Kit-stimulated cells was filtered by Kit-Activated transcripts and annotated for potential nearby or interacting cis-elements. We then trained a machine learning model using random forest algorithm on multifactor prioritization criteria (transcription factor occupancy, histone modifications, and predicted activity by contact) compared to Kit-insensitive control regions. The overall accuracy of this model at predicting Kit-response loci was 89%. Among 750 possible KREs, 45 Kit-predictive features were identified and used to rank KREs based on confidence scores. The top Kit-predictive features included epigenetic regulators (ARID1B, SMARCA4, TBL1XR1, TCF12), Inflammatory Response (JUN, JUND), and blood cell maintenance (NCOR1, CBFA2T2, CBFA2T3). We generated ATAC-seq data post Kit-activation and performed motif footprint analysis. This revealed increased enrichment of Kit-activated inflammatory response transcription factors’ footprints, including Early Growth Response-1 (EGR1), known to regulate cell proliferation and differentiation and 37-fold upregulated in response to Kit pathway activation.

To test whether EGR1 transcriptional upregulation is required in a Kit-dependent genetic network, we used a CRISPRi (dCas9-kRAB) model to prevent Kit from upregulating EGR1 (8-fold, p<0.0001) in HUDEP-2 cells. This also resulted in a 2.5-fold decrease in the HUDEP-2 cell expansion rate compared to controls. We identified 550 EGR1-sensitive KREs using our previously implemented in silico approach. EGR1-sensitive predictive features included occupancy of transcription factors (GATA1, GATA2, LDB1) and epigenetic regulators (SMARCB1, KDM6A, CHAMP1, EHMT2). 47 EGR1-Sensitive KREs showed enrichment for key EPC transcription factors’ (KLFs, SPs, CTCF) footprints at EGR1-sensitive KREs in our ATAC-seq data. Given the EGR1 dependency for select Kit pathway transcriptional activity, we are testing locus-specific mechanisms in EGR1-dependent and -independent transcription for EPC functions (proliferation, survival, differentiation, progenitor activity).

Impaired Kit signaling resulting from Kit receptor mutations causes hematologic diseases, including leukemia. While many cellular outcomes of Kit signaling are established, chromatin targets of this pathway are poorly understood. With tumors developing resistance to inhibitors, it is critical to understand Kit signaling machinery and transcriptional targets at KREs that are required for its activity. Many of our predicted KRE target genes have poorly described or unknown roles in erythropoiesis. Annotating the KRE function will improve future predictive models and identify targets for drug development to treat Kit-related malignancies.