Abstract
Erythroid Kruppel-Like Factor (EKLF; KLF1) is the founding member of the Kruppel family of transcription factors, with 3 C2H2 zinc-fingers that bind a 9-base consensus sequence (NCNCNCCCN). The functions of EKLF, first identified as an activator of the beta-globin locus, include gene activation and chromatin remodeling. Our knowledge of genes regulated by EKLF is limited, as EKLF-deficient mice die by embryonic day 15 (E15), due to a severe anemia. Analysis of E13.5 wild type and EKLF-deficient fetal liver (FL) erythroid cells revealed that EKLF-deficient cells fail to complete terminal erythroid maturation (Pilon et al. submitted). Coupling chromatin immunoprecipitation and ultra high-throughput massively parallel sequencing (ChIP-seq) is increasingly being used for mapping protein-DNA interactions in vivo on a genome-wide scale. ChIP-seq allows a simultaneous analysis of transcription factor binding in every region of the genome, defining an “interactome”. To elucidate direct EKLF-dependent effects on erythropoiesis, we have combined ChIP-seq with expression array (“transcriptome”) analyses. We feel that integration of ChIP-seq and microarray data can provide us detailed knowledge of the role of EKLF in erythropoiesis. Chromatin was isolated from E13.5 FL cells of mice whose endogenous EKLF gene was replaced with a fully functional HA-tagged EKLF gene. ChIP was performed using a highly specific high affinity anti-HA antibody. A library of EKLF-bound FL chromatin enriched by anti-HA IP was created and subjected to fluorescent in situ sequencing on a Solexa 1G platform, providing 36-base signatures that were mapped to unique sites in the mouse genome, defining the EKLF “interactome.” The frequency with which a given signature appears provides a measurable peak of enrichment. We performed three biological/technical replicates and analyzed each data set individually as well as the combined data. To validate ChIP-seq results, we examined the locus of a known EKLF target gene, a-hemoglobin stabilizing protein (AHSP). Peaks corresponded to previously identified DNase hypersensitive sites, regions of histone hyperacetylation, and sites of promoter-occupancy determined by ChIP-PCR. A genome wide analysis, focusing on the regions with the highest EKLF occupancy revealed a set of 531 locations where high levels EKLF binding occurs. Of these sites, 119 (22%) are located 10 kb or more from the nearest gene and are classified as intergenic EKLF binding sites. Another 78 sites (14.6%) are within 10 kb of an annotated RefSeq gene. A plurality of the binding sites, 222 (42%), are within RefSeq coordinates and are classified as intragenic EKLF binding sites. Microarray profiling of mRNA from sorted, matched populations of dE13.5 WT and EKLF-deficient FL erythroid progenitor cells showed dysregulation of >3000 genes (p<0.05). Ingenuity Pathways Analysis (IPA) of the >3000 dysregulated mRNAs indicated significant alteration of a cell cycle-control network, centered about the transcription factor, E2f2. We confirmed significantly decreased E2f2 mRNA and protein levels by real-time PCR and Western blot, respectively; demonstrated that EKLF-deficient FL cells accumulate in G0/G1 by cell cycle analysis; and verified EKLF-binding to motifs within the E2f2 promoter by ChIP-PCR and analysis of the ChIP Seq data. We hypothesized that only a subset of the 3000 dysregulated genes would be direct EKLF targets. We limited the ChIP-seq library to display the top 5% most frequently represented fragments across the genome, and applied this criterion to the network of dysregulated mRNAs in the IPA cell cycle network. ChIP-seq identified peaks of EKLF association with 60% of the loci in this pathway. However, consistent with the role of EKLF as a transcriptional activator, 95% of the occupied genomic loci corresponded to mRNAs whose expression in EKLF-deficient FL cells was significantly decreased (p<0.05). The majority (59%) of these EKLF-bound sites were located at intragenic sites (i.e., introns), while a minority (15% and 26%) were found adjacent to the genes or in intergenic regions. We have shown that both the AHSP and E2f2 loci require EKLF to cause the locus to become activated and sensitive to DNase I digestion in erythroid cells. Based on the increased frequency of intragenic EKLF-binding sites, particularly in genes of the cell cycle network, we propose that the occupancy of intragenic sites by EKLF may facilitate chromatin modification.
Assessing Computational Steps for CLIP-Seq Data Analysis