Dynamic CO-Localization of GATA1, NFE2, and EKLF and Changes in Gene Expression During Hematopoiesis

Abstract 741

Regulation of lineage choice during the development and differentiation of erythroid cells in hematopoiesis is a complex process. GATA1, NFE2, and EKLF are transcription factors critical for erythropoiesis. Focused studies, including detailed analyses of the human beta globin gene locus and a select group of erythrocyte membrane protein genes, have revealed that these three transcription factors may co-localize at common regulatory sites in erythroid-expressed genes. To address the hypothesis that GATA1, NFE2, and EKLF frequently co-localize on critical regulatory elements responsible for cell-type specific gene expression during erythropoiesis, chromatin immunoprecipitation coupled with ultrahigh throughput sequencing (ChIP-seq) was used to identify sites of GATA1, NFE2, and EKLF occupancy in human primary hematopoietic stem and progenitor cells (HSPCs) and human primary erythroid cells. ChIP was done using CD34+ HSPCs prepared by immunomagnetic bead selection and cultured CD71+/GPA+ erythroid cells (R3/R4 population) using antibodies against GATA1, NF-E2, and EKLF. The MACS algorithm (Zhang et al. Genome Biol, 2008) was used to identify regions of DNA-protein interaction, with a p-value ≤10e-5. Sites identified by MACS were ordered by p-value, and the 7000 sites with the most stringent p-values were selected for further analysis. Sites which occurred within 200bp of each other were treated as a single site. Unexpectedly, sites of GATA1, NFE2, and EKLF occupancy were common in HSPCs, with 6643 GATA1, 6657 NFE2, and 6579 EKLF sites identified, respectively. Sites identified in HSPCs were primarily in enhancers (>1kb from a RefSeq gene; 44% of GATA1, 49% of NFE2, and 51% of EKLF sites) and in introns (32% of GATA1, 34% of NFE2, and 34% of EKLF sites), with only a few sites at proximal promoters (within 1kb of a TSS; 7% of GATA1, 6% of NFE2, and 7% EKLF sites.) In erythroid cells, 6895 GATA1, 6907 NF-E2, and 6874 EKLF sites were identified. For all 3 factors, binding site occupancy varied greatly from that observed in HSPCs. Proximal promoter binding was much more common in erythroid cells than in HSPCs, with 19% of GATA1, 28% of NFE2 and 38% of EKLF sites found at promoters. Binding was frequently found at enhancers (41% of GATA, 38% NFE2, and 32% EKLF sites) and in introns (29% of GATA1, 26% of NFE2, and 21% of EKLF). To gain insight into three factor co-occupancy on a genome-wide scale, GATA1, EKLF, and NFE2 binding sites were compared using the Active Region Comparer (http://dart.gersteinlab. org/). Surprisingly, co-localization of all three factors was common in HSPCs, occurring at 2666 sites (40%, 40% and 45% of GATA1, NFE2, and EKLF sites). Sites of GATA1-NFE2-EKLF co-localization in HSPCs were located primarily at enhancers (51% of sites), in introns (32% of sites), and rarely at proximal promoters (6% of sites). In erythroid cells, co-localization of all three transcription factors was also common, occurring at 2445 sites (35%, 35%, and 36% of GATA1, NFE2, and EKLF sites, respectively). In contrast to HSPCs, sites of GATA1-NFE2-EKLF co-localization in erythroid cells were located primarily at proximal promoters (35% of sites) and enhancers (34% of sites), with co-localization in introns accounting for 20% of sites. A limited subset of sites, 1429 GATA1, 921 NFE2, and 1038 EKLF sites, were present in both HSPC and erythroid cells. Throughout the genome, there were only 233 sites of three factor co-localization in common in both HSPC and erythroid cells. Gene expression in HSPC and erythroid cell was analyzed via RNA hybridization to Illumina HumanHT-12 v3 Expression BeadChip arrays. In erythroid cells, genes with GATA1-NFE2-EKLF co-localization from 5kb upstream to 2kb downstream had significantly higher levels of mRNA expression than genes without GATA1-NFE2-EKLF co-localization (p<2.2e-16). The reverse was observed in HSPCs, where genes with GATA1-NFE2-EKLF co-localization had significantly lower levels of mRNA expression than genes without GATA1-NFE2-EKLF co-localization (p<7.3e-05). These data support the hypothesis that co-localization of GATA1, NFE2, and EKLF is a common finding in hematopoietic cells. Significant differences in factor co-localization and gene expression in HSPC and erythroid cells suggest that this coordinated binding orchestrates different patterns of gene expression during hematopoiesis.