The role of internal packing interactions in determining the structure and stability of a protein.

Cassette mutagenesis has been used to investigate how internal packing interactions help to specify a protein's three-dimensional structure and stability. Three interacting residues in the hydrophobic core of the N-terminal domain of lambda repressor were randomized combinatorially. The randomization was restricted to the five amino acids Val, Leu, Ile, Met and Phe, thereby generating a sterically diverse set of core sequences composed solely of hydrophobic residues. We have isolated 78 of the 125 possible sequences generated by this randomization. Approximately 70% of the isolated sequences show some level of biological activity, and thus still carry sufficient information to encode the basic structure of lambda repressor. An assay based on the temperature dependence of activity in vivo has been used to estimate the relative activities and thermal stabilities of the set of mutants. In addition, nine mutants have been purified and their stabilities and DNA binding activities characterized in vitro. Of the 56 active sequences, only two, in addition to the wild-type, maintain the wild-type level of stability and activity. All three of these proteins satisfy stringent requirements for specifically shaped residues at each position. All of the remaining active sequences have reduced stabilities and/or reduced DNA binding affinities. These and previous results suggest that there are two levels of structural information encoded in core residues. At the first level, the basic structural information appears to reside largely in the hydrophobic character of these residues. The majority of sequences that simply maintain hydrophobicity at core positions are able to adopt the overall lambda repressor fold and maintain moderate stability. At the second, more detailed level, specific steric features of these residues and their packing interactions clearly act as important determinants of the protein's precise structure and stability. These results imply that many of the basic structural features of a protein could be predicted from relatively simple, degenerate sequence patterns.