>

(f) cI Repressor, cro and l Operators

All six symmetrical operators of bacteriophage l are bound by both of the dimeric proteins repressor and cro (Ptashne et al., 1976, 1980; Johnson et al., 1981; Matthews et al., 1983). Maniatis et al. (1975) originally suggested that the sites are 17 basepairs wide, separated by A-T rich "spacers". Since then it has been thought that these regions are not part of the sites. However, a nonrandom sequence contains information. Chemical protection experiments that probed for guanine residues (Humayun et al., 1977a,b; Johnson et al., 1978; Pabo et al., 1982) did not address the issue since the region is almost completely devoid of G's and contacts in the region may not be directed to GC pairs. Adenine residues were unprotected either because the proteins do not cover that region or because the proteins bind to the opposite side of the DNA from the modifiable group. Two promoter mutations in these regions increase the A-T richness and do not affect repressor binding (Ptashne et al., 1976; prm116, Meyer et al., 1975; sex1, Kleid et al., 1976). One mutation, prm up-1, decreases the A-T richness. The effect of prm up-1 on repressor binding is said to be small (Johnson et al., 1979; Meyer et al., 1980). In contrast to this mutant, nuclease protection experiments show the sites to be 25 base pairs wide (Humayun et al., 1977a). Thus it is possible that a portion or all of the "spacers" are part of the binding sites. However, in keeping with the rules defined in Materials and Methods, we used a range 19 bases wide to avoid overlap between O $$\sum_{i=1}^{M}P_iu_i .$$3 and O $$\sum_{i=1}^{M}P_iu_i .$$2 (Fig. 7). (This also avoids the prm up-1 site.) Most information content of the "spacers" was lost by this procedure; R_sequence = 17.1, R_frequency = 19.3 bits per site. If overlaps are ignored, and the sites extended to the size protected from DNase (25 base pairs wide, -12 to +12), R_sequence becomes 19.0.