The specific DNA sequence recognized by the yeast transcription factor Gcn4 is sufficient to predict high-affinity binding on a genome-wide scale

Sequence-specific DNA-binding transcription factors are central to gene regulation. They are often associated with consensus binding sites that predict far more genomic sites than are bound in vivo. One explanation is that most sites are blocked by nucleosomes, such that only sites in nucleosome-depleted regulatory regions are bound. Alternatively, the consensus site may be poorly defined. We compared the binding of the yeast transcription factor Gcn4 in vivo using published ChIP-seq data (546 sites) and in vitro, using a modified SELEX method ("G-SELEX"), which utilizes short genomic DNA fragments to quantify binding at all sites. We confirm that Gcn4 binds strongly to an AP1-like sequence (TGACTCA) and weakly to half-sites. However, Gcn4 binds only some of the 1078 exact matches to this sequence, even in vitro. We show that the high-affinity site is RTGACTCAY (exact match), of which there are only 166 copies in the yeast genome, all occupied in vivo. Generally, RTGACTCAR/YTGACTCAY sites are bound much more weakly and YTGACTCAR sites are unbound, with biological implications for determining induction levels. We conclude that Gcn4 binding can be predicted using the genome sequence. Our study suggests that transcription factor consensus sequences should be defined more precisely using quantitative data. Overall design: Incubation of yeast genomic DNA fragments with recombinant Gcn4, followed by amplification of bound fragments (3 rounds) and paired-end sequencing