DNA sequence recognition by Pax proteins: bipartite structure of the paired domain and its binding site

Previous DNA-binding studies indicated that an intact paired domain is required for interaction of the transcription factor BSAP (Pax-5) with DNA. We have now identified a subset of BSAP recognition sequences that also bind to a truncated BSAP peptide lacking 36 carboxy-terminal amino acids of the paired domain. Sequence comparison of this class of BSAP-binding sites made it possible to unequivocally align all known BSAP-binding sites and to deduce a consensus sequence consisting of two distinct half sites. We propose here a model for the paired domain--DNA interaction in which the paired domain is composed of two subdomains that bind to the two half-sites in adjacent major grooves on the same side of the DNA helix. The existence of these half sites and of the two paired domain subregions was directly demonstrated by methylation interference analysis and by in vitro mutagenesis of both the paired domain and its recognition sequence. Both half-sites contribute to the overall affinity of a given BSAP-binding site according to their match with the consensus sequence. However, none of the naturally occurring BSAP-binding sites completely conform to the consensus sequence. Instead, they contain compensatory base changes in their half-sites that explain the versatile and seemingly degenerate DNA sequence recognition of Pax proteins. Domain swap experiments between BSAP and Pax-1 demonstrated that the sequence specificity of the BSAP paired domain is determined by both its amino- and carboxy-terminal subdomains. Moreover, mutations affecting only one of the two subdomains restricted the sequence specificity of the paired domain. Such mutations have been shown previously to be the cause of mouse developmental mutants (undulated, Splotch, and Small eye) and human syndromes (Waardenburg's syndrome and aniridia) and may thus differentially affect the regulation of target genes by the mutated Pax protein.