In this study, we provide genome-wide chromatin occupancy analysis of the erythroid MTF GATA-1. Comparison to a new comprehensive dataset of GATA-1 induced gene expression changes allowed us to characterize features of GATA-1 in vivo occupancy that correlate with its site selectivity and gene context-dependent transcriptional activity.
We combined transcription factor metabolic biotinylation with streptavidin-based ChIP-seq, and were able to confirm 97% and 91% of our peak calls by independent standard ChIP assays in MEL or primary erythroid cells, respectively. Comparison of our dataset with 63 validated enrichment peaks identified in a recent GATA-1 ChIP-chip study of 66 Mb of mouse chromosome 7 in induced G1-ER4 cells (Cheng et al., 2008
) shows an overlap of 21 peaks (33%) (Tables S6
). Of 59 sites identified in the ChIP-chip study that did not validate, none were called as peaks in our dataset. If we relax our threshold call from 14 to 8, then the ChIP-seq dataset picks up 30 of the 63 ChIP-chip validated peaks (48%), and only one of the 59 non-validated sites. Thus, the SA-biotin ChIP-seq technique as applied here has relatively high specificity, but perhaps limited sensitivity, at least based on comparison to this one prior study.
Although our dataset may fail to identify all of the bona fide GATA-1 occupancy sites, ascertainment of a large number of high-confidence sites allowed us to apply statistical methods to further understand GATA-1's transcriptional activity. We found that the binding motif that best predicts global GATA-1 in vivo is slightly more extended and has more sequence preference than the canonical motif defined in previous DNAse I footprinting and in vitro studies, and that palindromic GATA binding motifs are significantly enriched at in vivo occupancy sites. Moreover, a higher overall number of GATA motifs is predictive of in vivo occupancy. While our data may be biased toward high affinity sites, these findings may help explain GATA-1's in vivo site selectivity.
Comparison of our ChIP-seq dataset with GATA-1 induced gene expression changes enabled the identification of global cell processes that are under direct GATA-1 control. As expected, we found marked enrichment for genes involved in hemoglobin synthesis, providing further validation of our dataset. In addition, we found that many GATA-1 direct target genes are involved in cell cycle control and Ras/GTPase signaling.
GATA-1's role in regulating cell proliferation has been previously studied. Rylski et al.
showed that GATA-1 represses expression of the cyclin-dependent kinase 6 (Cdk6) and cyclin D2, and activates expression of the Cdk inhibitors p18INK4C
(Rylski et al., 2003
). This occurs, in part, by direct transcriptional repression of the oncogene c-myc. Munogalavadla et al.
found that GATA-1 directly regulates c-myb leading to altered cell cycle (Munugalavadla et al., 2005
). In addition, GATA-1 deficiency causes marked hyperproliferation of murine megakaryocytes (Vyas et al., 1999a
), and exclusive production of a short GATA-1 isoform (GATA-1s) leads to transient myeloproliferative disorder in Down syndrome neonates (Muntean et al., 2006
). Our findings add to the list of cell cycle related genes, including E2F4, Cdc6,
that are under direct GATA-1 transcriptional control.
Ras signaling has previously been shown to affect erythroid differentiation. K-ras−/−
mice die during embryonic development from severe anemia (Johnson et al., 1997
), and expression of oncogenic N-ras and H-ras perturbs erythroid differentiation (Darley et al., 1997
; Zhang et al., 2003
). Since Ras/GTPase signaling is also involved in cell survival, our findings may partly explain GATA-1's anti-apoptotic activity during erythroid development (Weiss and Orkin, 1995
Motif search analysis of GATA-1 bound regions of direct target genes revealed enrichment for Zbtb7a binding sequences, particularly at activated genes. Zbtb7a occupancy was confirmed at a significant number of GATA-1 enrichment sites in stage-sorted primary fetal liver erythroid cells, although it was found at both activated and repressed genes. Zbtb7a is highly expressed in CD71+
primary erythroid cells and is transcriptionally activated by EKLF (Hodge et al., 2006
; Maeda et al., 2007
mice die at around e16.5 from severe anemia (Maeda et al., 2007
). Yet, the mechanisms underlying the anemia have not been reported. Our data suggest that combinatorial transcriptional activity with GATA-1 may be involved in this phenotype. Zbtb7b (Th-Pok), a close family member of Zbtb7a, is a key regulator of CD4/CD8 lineage choice during T-cell development, a process that involves the GATA family member GATA-3 (He et al., 2005
; Sun et al., 2005
). It is possible that similar combinatorial processes occur between Zbtb7b and GATA-3 during lymphoid development.
Compared to gene activation, less is known about how GATA-1 functions as a transcriptional repressor. Polycomb repressive complexes play critical roles in epigenetic gene silencing during development. Unlike in Drosophila, in which polycomb protein complexes are recruited to Polycomb Response Elements (PREs), the recruitment of polycomb in mammalian cells is poorly understood. In FACS-sorted primary fetal liver erythroblasts, we found significant levels of H3K27me3 at a number of GATA-1 bound and repressed genes, but not activated genes or non-GATA-1 bound repressed genes. Although we cannot conclude that GATA-1 directly recruits PRC2 to these sites, several pieces of data support PRC2 involvement in late stages of GATA-1 mediated gene silencing, at least for a subset of genes. First, GATA-1 physically associates with Suz12 and EZH2 in erythroid cells (). Second, activation of GATA-1 in G1-ER4 cells results in increased Suz12 chromatin occupancy and H3K27me3 levels at some GATA-1 direct target genes within 48 hours. Third, erythroid-specific deletion of the core PRC2 component EED results in reduced H3K27me3 at direct GATA-1 target genes. Fourth, erythroid specific deletion of EED results in impaired erythroid maturation.
We feel that it is unlikely that PRC2 is involved in the initial steps of GATA-1 mediated gene repression. Rather, we favor the view that it participates in stabilizing epigenetic silencing once the initial decision to turn off the gene is made. Gfi1b is a transcriptional repressor that is required for normal erythroid development (Saleque et al., 2007
). It interacts with LSD1, which has specific H3K4 demethylase activity. Interestingly, Gfi-1b co-purifies with Suz12 and GATA-1 in MEL cells (), and Gfi-1b occupies GATA-1 bound regions of several repressed genes (). The actions of Gfi-1b, via LSD1, may be the initial step in reversing gene activation by removing H3K4 methylation at genes that are initially on during early erythroid development, such as GATA-2
. The H3K27 methyltransferase activity of PRC2 may then act to stabilize the silencing at a subset of genes (). GATA-1 may coordinate these activities, acting as a platform for both Gfi-1b and PRC2. It is also possible that the absence of SCL/TAL1 complexes might enable recruitment of PRC2. Future studies will be needed to address these possibilities.
Model of transcription factor and cofactor occupancy at in vivo GATA-1 binding sites correlating with gene activation versus repression.
In summary, our data provide a genome-wide analysis of GATA-1 chromatin occupancy, facilitating examination of its transcriptional mechanisms. The findings implicate Zbtb7a as a factor involved in GATA-1 mediated gene regulation, and the PRC2 complex as being involved in late stages of silencing of some GATA-1 repressed genes. This dataset should provide a valuable resource to other investigators studying the transcriptional regulation of terminal cell maturation in mammalian systems.