Structure and function of bacterial and mammalian dihydrofolate reductases.

Dihydrofolate reductase (EC 1.5.1.3) catalyses the NADPHdependent reduction of dihydrofolate to tetrahydrofolate. The product, together with other reduced folates, serves as a cofactor in a variety of 1-carbon transfers in the biosynthesis of purines, pyrimidines and amino acids (Blakley, 1969). The enzyme is considered to be the major cellular receptor for several important drugs including methotrexate (2.4-diamino-N'Omethylpteroylglutamate) (Hitchings & Burchall, 1965) and inhibition of the enzyme leads to inhibition of DNA synthesis and cell death. Certain inhibitors in particular trimethoprim 12,4-diamino-5-(3',4',5'-trimethoxybenzyl)pyrimidinel and pyrimethamine (2,4-diamino-5-p-chlorophenyl-6-ethylpyrimidine) are more potent inhibitors of bacterial dihydrofolate reductases than of the mammalian enzymes and, consequently, these compounds have been employed as antibacterial agents. In an attempt to understand the underlying molecular basis for the differential susceptibilities of dihydrofolate reductases to inhibition, a considerable effort has been devoted to the determination of the structures of these enzymes from a variety of bacterial and vertebrate sources. The complete amino acid sequences of the bacterial enzymes from Streptococcus faecium (Gleisner el al., 1974), Escherichia coli RT500 (Stone el al., 1977), E. coli MB1428 (Bennett el al., 1978). Lactobacillus casei (Bitar et al., 1977) and the enzyme specified in E. coli by the R-plasmid R67 (Stone & Smith, 1979) have been determined. In addition, the amino acid sequences of the enzymes from the following vertebrate sources have also been determined; mouse lymphoma L1210 (Stone et al., 1979), porcine liver (Smith et al., 1979), chicken liver (Kumar et al., 1980) and bovine liver (Lai et al., 1979). The primary sequences are compared in Fig. l., and in the following discussion all residue numbers refer to this Figure. With the exception of the R-67-plasmid-specified enzyme, which does not appear to be related to the other dihydrofolate reductases, the bacterial enzymes generally contain 159-167 amino acids, whereas the vertebrate enzymes all contain between 186 and 189. The vertebrate enzymes are closely homologous, possessing an overall identity of 72-89%, whereas the degree of identity observed between bacterial enzymes and between bacterial and vertebrate enzymes is of the order of 30%. The constant and variable residues are distributed throughout the molecules, suggesting that there is probably overall structural similarity, although it is Clem that certain restricted regions of the molecule show more extensive homology then is observed in other regions. In particular the regions between residues 16-29,49-76 11 1-125 and 135-152 are highly conserved, and it is interesting to consider these areas in terms of the X-ray-crystallographic studies of enzyme-ligand complexes that have been published to date. The structures of the E. coli-enzyme-methotrexate binary complex (Matthews et al., 1977) and the L. casei-enzymeNADPH-methotrexate ternary complex (Matthews et al., 1978) have both been solved at 2.5nm (25.4) resolution. Although the two proteins have substantially different amino acid sequences (29% identity), the protein backbone and inhibitor conformations in the two structures are very similar, and it seems likely, therefore, that all homologous dihydrofolate reductases will possess the same general structure. The overall folding of the polypeptide chain is dominated by an eight-stranded &sheet composed of seven parallel strands and a single antiparallel strand at the C-terminus. The sheet shows the usual right-handed twist. The molecule contains three helical regions. A cleft some 1.5nm (15 A) wide cuts across one face of the molecule and gives the structure a bi-lobed appearance. The pteridine ring of methotrexate, and, by analogy, the substrate dihydrofolate, occupies a hydrophobic area within this cleft and the nicotinamide ring of the cofactor is located in the other portion of the cleft. It is perhaps not surprising that all of the four highly conserved areas of the molecule described above contribute residues to the surfaces of this cleft, and thus, on the basis of the amino acid sequence homology, it seems likely that the active sites of mammalian and bacterial enzymes are structurally similar. This hypothesis is further strengthened by a consideration of the specific residues involved in binding methotrexate. According to the X-raycrystallographic studies, the pteridine ring is bound in the active-site cleft by hydrogen bonds formed between the 2-amino, 4-amino, N ( l ) and residues threonine136, isoleucine-7 and aspartic acid-30 respectively and by hydrophobic interactions with isoleucine-7, alanine-9, leucine-3 1, phenylalanine-34 and isoleucine1 14. These residues are all either conserved or conservatively replaced in the dihydrofolate reductase sequences that have been determined. Similarly residues leucine-3 1, isoleucine-60, leucine-67 and isoleucine1 14, all of which interact with the paminobenzoate portion of the methotrexate molecule are either conserved or conservatively replaced. The cofactor is bound in an extended conformation with its adenine portion on one side of the /?-sheet and the nicotinamide portion on the other side. The residues that interact with the cofactor, in particular the basic residue at position-54, which interacts with the 2'-phosphate, are well conserved, and this observation provides further support for the hypothesis that the enzymes possess very similar structures. The crystallographically observed structure of the ternary complex predicts that, assuming the pteridine ring of dihydrofolate is bound in the same orientation as that of methotrexate, the absolute configuration of C-6 of the product would be R according to the Cahn-Ingold-Prelog conversion (Cahn el al.,