Defining target sequences of DNA-binding proteins by random selection and PCR: determination of the GCN4 binding sequence repertoire.

We developed a simple and accurate method to define the sequence recognition properties of DNA-binding proteins. The method employs polymerase chain reaction (PCR) amplification of sequences selected from a mixture of random oligonucleotides by the gel mobility-shift assay. We used this method to define the sequence requirement of the binding domain of the yeast transcriptional activator GCN4. Using a total of 200 ng of purified protein and four cycles of binding and subsequent amplification, we identified the TGA-(C/G)TCA sequence as the binding consensus of GCN4, which is consistent with the previously reported recognition sequence. In addition, our data indicate that GCN4 can bind with lower affinity to sequences that differ from the optimal sequence in one or even two positions. The most common variation was the C to A at position +2. The majority of the substitutions that still allowed binding were 3' to the central C residue indicating that the two sides of the palindromic recognition sequence are not equivalent.