The maintenance of H3K9 and DNA methylation at imprinting control regions (ICRs) during early embryogenesis is key to the regulation of imprinted genes. Here, we reveal that ZFP57, its cofactor KAP1, and associated effectors bind selectively to the H3K9me3-bearing, DNA-methylated allele of ICRs in ES cells. KAP1 deletion induces a loss of heterochromatin marks at ICRs, whereas deleting ZFP57 or DNMTs leads to ICR DNA demethylation. Accordingly, we find that ZFP57 and KAP1 associated with DNMTs and hemimethylated DNA-binding NP95. Finally, we identify the methylated TGCCGC hexanucleotide as the motif that is recognized by ZFP57 in all ICRs and in several tens of additional loci, several of which are at least ZFP57-dependently methylated in ES cells. These results significantly advance our understanding of imprinting and suggest a general mechanism for the protection of specific loci against the wave of DNA demethylation that affects the mammalian genome during early embryogenesis.

The DNA-binding protein CTCF (CCCTC binding factor) mediates enhancer blocking insulation at sites throughout the genome and plays an important role in regulating allele-specific expression at the Igf2/H19 locus and at other imprinted loci. Evidence is also accumulating that CTCF is involved in large scale organization of genomic chromatin. Although CTCF has 11 zinc fingers, we show here that only 4 of these are essential to strong binding and that they recognize a core 12-bp DNA sequence common to most CTCF sites. By deleting individual fingers and mutating individual sites, we determined the orientation of binding. Furthermore, we were able to identify the specific finger and its point of DNA interaction that are responsible for the loss of CTCF binding when CpG residues are methylated in the imprinted Igf2/H19 locus. This single interaction appears to be critical for allele-specific binding and insulation by CTCF.

Specific DNA binding to the core consensus site GAGAGAG has been shown with an 82-residue peptide (residues 310-391) taken from the Drosophila transcription factor GAGA. Using a series of deletion mutants, it was demonstrated that the minimal domain required for specific binding (residues 310-372) includes a single zinc finger of the Cys2-His2 family and a stretch of basic amino acids located on the N-terminal end of the zinc finger. In gel retardation assays, the specific binding seen with either the peptide or the whole protein is zinc dependent and corresponds to a dissociation constant of approximately 5 x 10(-9) M for the purified peptide. It has previously been thought that a single zinc finger of the Cys2-His2 family is incapable of specific, high-affinity binding to DNA. The combination of an N-terminal basic region with a single Cys2-His2 zinc finger in the GAGA protein can thus be viewed as a novel DNA binding domain. This raises the possibility that other proteins carrying only one Cys2-His2 finger are also capable of high-affinity specific binding to DNA.

The GATA family of vertebrate DNA binding regulatory proteins are expressed in diverse tissues and at different times of development. However, the DNA binding regions of these proteins possess considerable homology and recognize a rather similar range of DNA sequence motifs. DNA binding is mediated through two domains, each containing a zinc finger. Previous results have led to the conclusion that although in some cases the N‐terminal finger can contribute to specificity and strength of binding, it does not bind independently, whereas the C‐terminal finger is both necessary and sufficient for binding. Here we show that although this is true for the N‐terminal finger of GATA‐1, those of GATA‐2 and GATA‐3 are capable of strong independent binding with a preference for the motif GATC. Binding requires the presence of two basic regions located on either side of the N‐terminal finger. The absence of one of these near the GATA‐1 N‐terminal finger probably accounts for its inability to bind. The combination of a single finger and two basic regions is a new variant of a motif that has been previously found in the binding domains of other finger proteins. Our results suggest that the DNA binding properties of the N‐terminal finger may help distinguish GATA–2 and GATA‐3 from GATA‐1 and the other GATA family members in their selective regulatory roles in vivo .