Actively Seeking Activating Sequences

In the spring of 1986, I was recovering from yet another disappointment. Mark Ptashne, my Ph.D. supervisor, had been actively promoting the looping model for how regulatory proteins act at a distance on DNA. Inspired by the ;~ repressor work of Ann Hochschild (1986), another student in the lab, I wanted to determine whether the yeast activator Gal4 had to sit on the same side of the DNA helix as a nearby TATA box to activate transcription, as predicted by the looping model. But my experiments showed that Gal4 worked equally well regardless of whether its binding site and the TATA box were on the same or opposite side of the DNA helix. Reflecting on this result, I thought it might be due to the modular-perhaps highly flexible--nature of this large protein of 881 aa. Recent experiments by Roger Brent (1985), a postdoctoral fellow in the lab, and two students, Liam Keegan and Grace Gill (1986), had just shown that while the N-terminal 74 aa of Gal4 were responsible for bringing the protein to DNA, the remaining 90% of the protein provided an activation function. I began to wonder if I could generate smaller Gal4 derivatives to determine whether they might exhibit the anticipated helical periodicity. After a few pilot tests, I soon initiated a systematic deletion analysis of Gal4, progressively removing protein sequences both internally and from the C terminus. The urgency of this analysis was also fueled by a rumor that Keith Yamamoto's lab was able to generate large truncations of the glucocorticoid receptor without destroying its activation function. My experiments (1987a) identified two short segments of Gal4 (49 and 114 aa), each sufficient to activate transcription when linked to its DNA binding domain. Derivatives of Gal4 lacking as much as 80% of their protein sequence can activate transcription. Later, Doug Ruden (1988), another student in the lab, and I looked for helical periodicity using small derivatives of Gal4, but we never found any. Before the completion of the Gal4 deletion analysis, I read a report by lan Hope and Kevin Struhl (1986), who had delineated a 19 aa activating region for another yeast activator Gcn4. Interestingly, the two Gal4 activating sequences and that of Gcn4 shared no obvious homology among themselves except being highly acidic. As noted by Hope and Struhl, the Gcn4 activating region also failed to show sequence homology to other yeast activators available at the time. An idea began to emerge that many different sequences--perhaps sharing some loosely defined common features such as an excess of acidic residues--could activate transcription in yeast,