In Vivo GATA-1 Binding to an LCR Core Element Requires Flanking Krupple-Like Factor Biniding Sites Which Direct the Formation of a Chromatin Mini-Domain.

The GATA-1 transcription factor is involved in the expression of most, if not all erythroid genes. In vitro, GATA-1 binds the WGATAR consensus and related sequences. If GATA-1 binding to DNA in nuclear chromatin were only dependent on these recognition sequences, there would be roughly 3 million sites per genome – or approximately 1 site per 1000 bp. The alternative is that mechanisms beyond the consensus sequence are involved in targeting GATA-1 (and other transcription factors) to sites in chromatin. To investigate which model is correct, we performed GATA-1 ChIP assays in MEL cells on 9 WGATAR sites in the murine β-globin gene locus that were not associated with known regulatory elements. GATA-1 bound to none of these sites, while it did bind to positive control sites in LCR HS2 and the GATA-1 promoter. This implied that factors other than DNA sequence are involved in GATA-1 targeting. We hypothesized that factor binding sites flanking the WGATAR motifs are necessary for GATA-1 binding. To test this idea, we chose the human LCR HS4 core as a model. Here, a cluster of 6 binding sites, contained within ~100 bp, is found in the following arrangement: 5′-K-N-K-GG-K-3′ (K=Krupple-like factor, N=NF-E2, G=GATA). To determine which sites were required for GATA-1 binding, we used site-directed mutagenesis to insert combinations of the sites into a non-expressed fragment of DNA in a pUC-based plasmid. Constructs were then stably transfected into MEL cells and pools of 25 clones were analyzed for GATA-1 binding to the integrated test sites. Each ChIP assay was performed in triplicate and analyzed by quantitative RT-PCR, also in triplicate. Internal positive and negative control sites for GATA-1 binding were included for each ChIP assay. GATA-1 did not bind the following constructs in vivo: no sites, N alone, GG alone, N-GG, N-K-GG, N-K-GG-K, and K-N-K-GG. Only when all sites were present, K-N-K-GG-K, did GATA-1 bind. These results showed that both the 5′ and 3′ flanking KLF sites were necessary for GATA-1 binding. To determine whether they were also sufficient for GATA-1 binding, we next evaluated K-K and K-GG-K constructs, where the K sites were the 5′ and 3′ flanking sites in their normal locations. While both the K-K and K-GG-K constructs formed DNAse I HSs and directed H3 hyperacetylation in the region (the control without binding sites showed neither characteristic), indicating that the site was accessible for factor binding, the K-GG-K construct still did not bind GATA-1. We next treated cells containing the GG construct with trichostatin A to verify that histone hyperacetylation alone was insufficient to allow GATA-1 binding. Despite a large increase in total H3 acetylation (Western blot) and a doubling of local H3 acetylation (ChIP), GATA-1 still did not bind the GG construct. Based on these results we propose a mechanism for GATA-1 binding to the LCR HS4 core in which flanking KLF sites initially bind factors that form a localized “mini-domain” characterized by histone hyperacetylation, HS formation and trans-factor accessibility, but that factors binding to the NF-E2 and/or the internal KLF site are required for stabilizing GATA-1 binding within this open chromatin domain. These results are relevant not only to GATA-1 binding, but to understanding how transcription factors, in general, are targeted to specific sites in nuclear chromatin.