Commentary Do basic region - leucine zipper proteins bend their DNA targets . . . does it matter ? Paul

Protein-induced bending of DNA has been described for numerous members of the large family of "basic regionleucine zipper" (bZIP) transcriptional modulators (1-4). However, for this class of proteins, there exists some uncertainty as to how much (or even whether) the bound target DNA is bent. This issue has been most sharply focused for the bZIP protein heterodimer, Fos-Jun, with two recent articles in the Proceedings (5, 6) presenting seemingly disparate views as to whether its recognition element (AP-1) is (6)-or is not (5)-bent. It has been known for some time that certain transcriptional regulators are capable of bending their DNA targets. For example, both solution (7) and crystallographic (8) studies of the Escherichia coli catabolite activator protein (CAP)-DNA complex suggest that CAP induces a -90° bend in its recognition sequence. A distortion of similar magnitude, albeit of very different local geometry, has been observed for a TATA box-binding polypeptide (9, 10). Distortions of this magnitude are highly improbable as thermal fluctuations; thus, one function of modulators such as CAP and TATA box-binding polypeptide may be to bring together otherwise distant DNA segments via protein-induced bends. In fact, tests of this general hypothesis have in some instances demonstrated that the specific function of protein-induced bends can be subserved either by heterologous protein-induced bends or by intrinsic (sequence-directed) elements ofDNA curvature (11). In 1991, Kerppola and Curran (12) reported that Fos-Jun heterodimers substantially distort the AP-1 site. Comparing the relative gel mobilities ofDNA molecules in which the AP-1 site was placed at various positions (permuted) within the molecules (circular permutation assay; refs. 13-15), those authors reported "flexure" angles of 940 for full-length FosJun, and 550 for a truncated form of the heterodimer (12). However, Kerppola and Curran (16) also performed a different form of gel experiment in which the AP-1 site was placed in varying torsional alignments with respect to a reference (sequence-directed) bend (phasing assay; ref. 17). This latter assay yielded 230 for the Fos-Jun-induced bend and 10-12° for the bend induced by the heterodimers. Furthermore, using the phasing assay, Kerppola and Curran (16) observed that the direction of the bend induced by Fos-Jun was nearly opposite to the direction created by the Jun-Jun homodimer, and they suggested that this directionality could contribute to the differential response of the glucocorticoid response element in the presence of Jun (activation) or Fos-Jun (repression). However, using a combination of gel and solution methods, Sitlani and Crothers (5) reached quite a different conclusion regarding the Fos-Jun-AP-1 interaction, namely, that Fos-Jun does not bend its target (i.e., by less than 50). Thus, there actually appear to be two sources of disagreement: one involves a large disparity in the apparent angles derived from permutation and phasing methods; a second involves inconsistent results from phasing and cyclization assays. Since both gel and solution approaches continue to enjoy widespread use, and since the question of Fos-Juninduced bending will necessarily turn on fine points of experimental method, it is worthwhile to consider each method. The details will turn out to be important, not only for Fos-Jun and other members of the bZIP family, but for the study of protein-induced bending of DNA in general.