Protein-induced DNA bending is an important element in the structure of many protein-DNA complexes, including those involved in replication, transcription, and recombination. To understand these structures, the path followed by the DNA in each complex must be established. We have generated an empirical relation between the degree of bending and the altered electrophoretic mobility in polyacrylamide gels that allows estimation of protein-induced bends. This technique has been used to analyze 17 different protein-DNA complexes formed by six proteins including the four proteins involved in lambda site-specific recombination. The simplicity of this technique should make it useful in estimating angles for the construction of models of protein-DNA complexes and readily applicable to many systems where questions of higher-order structure are important for understanding function.

Alignments of 105 site-specific recombinases belonging to the Int family of proteins identified extended areas of similarity and three types of structural differences. In addition to the previously recognized conservation of the tetrad R-H-R-Y, located in boxes I and II, several newly identified sequence patches include charged amino acids that are highly conserved and a specific pattern of buried residues contributing to the overall protein fold. With some notable exceptions, unconserved regions correspond to loops in the crystal structures of the catalytic domains of lambda Int (Int c170) and HP1 Int (HPC) and of the recombinases XerD and Cre. Two structured regions also harbor some pronounced differences. The first comprises beta-sheets 4 and 5, alpha-helix D and the adjacent loop connecting it to alpha-helix E: two Ints of phages infecting thermophilic bacteria are missing this region altogether; the crystal structures of HPC, XerD and Cre reveal a lack of beta-sheets 4 and 5; Cre displays two additional beta-sheets following alpha-helix D; five recombinases carry large insertions. The second involves the catalytic tyrosine and is seen in a comparison of the four crystal structures. The yeast recombinases can theoretically be fitted to the Int fold, but the overall differences, involving changes in spacing as well as in motif structure, are more substantial than seen in most other proteins. The phenotypes of mutations compiled from several proteins are correlated with the available structural information and structure-function relationships are discussed. In addition, a few prokaryotic and eukaryotic enzymes with partial homology with the Int family of recombinases may be distantly related, either through divergent or convergent evolution. These include a restriction enzyme and a subgroup of eukaryotic RNA helicases (D-E-A-D proteins).

A combination of two methods for detecting distant relationships in protein primary sequences was used to compare the site‐specific recombination proteins encoded by bacteriophage lambda, phi 80, P22, P2, 186, P4 and P1. This group of proteins exhibits an unexpectedly large diversity of sequences. Despite this diversity, all of the recombinases can be aligned in their C‐terminal halves. A 40‐residue region near the C terminus is particularly well conserved in all the proteins and is homologous to a region near the C terminus of the yeast 2 mu plasmid Flp protein. This family of recombinases does not appear to be related to any other site‐specific recombinases. Three positions are perfectly conserved within this family: histidine, arginine and tyrosine are found at respective alignment positions 396, 399 and 433 within the well‐conserved C‐terminal region. We speculate that these residues contribute to the active site of this family of recombinases, and suggest that tyrosine‐433 forms a transient covalent linkage to DNA during strand cleavage and rejoining.

No abstract