DEFINING THE SEQUENCE SPECIFICITY OF THE SACCHAROMYCES-CEREVISIAE DNA-BINDING PROTEIN REB1P BY SELECTING BINDING-SITES FROM RANDOM-SEQUENCEOLIGONUCLEOTIDES

We have used a random selection protocol to define the consensus and r ange of binding sites for the Saccharomyces cerevisiae REB1 protein. T hirty-five elements were sequenced which bound specifically to a GST-R EB1p fusion protein coupled to glutathione-Sepharose under conditions in which more than 99.9% of the random sequences were not retained. Tw enty-two of the elements contained the core sequence CGGGTRR, with all but one of the remaining elements containing only one deviation from the core. Of the core sequence, the only residues that were absolutely conserved were the three consecutive G residues. Statistical analysis of a nucleotide-use matrix suggested that the REB1p binding site also extends into flanking sequences with the optimal sequence for REB1p b inding being GNGCCGGGGTAACNC. There was a positive correlation between the ability of the sites to bind in vitro and activate transcription in vivo; however, the presence of non-conformants suggests that the bi nding site may contribute more to transcriptional activation than simp ly allowing protein binding. Interestingly, one of the REB1p binding e lements had a DNAse 1 footprint appreciably longer than other elements with similar affinity. Analysis of its sequence indicated the potenti al for a second REB1p binding site on the opposite strand. This sugges ts that two closely positioned low-affinity sites can function togethe r as a highly active site. In addition, database searches with some of the randomly defined REB1p binding sites suggest that related element s are commonly found within 'TATA-less' promoters.