Maria Krook

Short-chain dehydrogenases/reductases (SDR) constitute a large protein family. Presently, at least 57 characterized, highly different enzymes belong to this family and typically exhibit residue identities only at the 15-30% level, indicating early duplicatory origins and extensive divergence. In addition, another family of 22 enzymes with extended protein chains exhibits part-chain SDR relationships and represents enzymes of no less than three EC classes. Furthermore, subforms and species variants are known of both families. In the combined SDR superfamily, only one residue is strictly conserved and ascribed a crucial enzymatic function (Tyr 151 in the numbering system of human NAD(+)-linked prostaglandin dehydrogenase). Such a function for this Tyr residue in SDR enzymes in general is supported also by chemical modifications, site-directed mutagenesis, and an active site position in those tertiary structures that have been characterized. A lysine residue four residues downstream is also largely conserved. A model for catalysis is available on the basis of these two residues. Binding of the coenzyme, NAD(H) or NADP(H), is in the N-terminal part of the molecules, where a common GlyXXXGlyXGly pattern occurs. Two SDR enzymes established by X-ray crystallography show a one-domain subunit with seven to eight beta-strands. Conformational patterns are highly similar, except for variations in the C-terminal parts. Additional structures occur in the family with extended chains. Some of the SDR molecules are known under more than one name, and one of the enzymes has been shown to be susceptible to native, chemical modification, producing reduced Schiff base adducts with pyruvate and other metabolic keto derivatives. Most SDR enzymes are dimers and tetramers. In those analyzed, the area of major subunit contacts involves two long alpha-helices (alpha E, alpha F) in similar and apparently strong subunit interactions. Future possibilities include verification of the proposed reaction mechanism and tracing of additional relationships, perhaps also with other protein families. Short-chain dehydrogenases illustrate the value of comparisons and diversified research in generating unexpected discoveries.

Different short-chain dehydrogenases are distantly related, constituting a protein family now known from at least 20 separate enzymes characterized, but with extensive differences, especially in the C-terminal third of their sequences. Many of the first known members were prokaryotic, but recent additions include mammalian enzymes from placenta, liver and other tissues, including 15-hydroxyprostaglandin, 17 beta-hydroxysteroid and 11 beta-hydroxysteroid dehydrogenases. In addition, species variants, isozyme-like multiplicities and mutants have been reported for several of the structures. Alignments of the different enzymes reveal large homologous parts, with clustered similarities indicating regions of special functional/structural importance. Several of these derive from relationships within a common type of coenzyme-binding domain, but central-chain patterns of similarity go beyond this domain. Total residue identities between enzyme pairs are typically around 25%, but single forms deviate more or less (14-58%). Only six of the 250-odd residues are strictly conserved and seven more are conserved in all but single cases. Over one third of the conserved residues are glycine, showing the importance of conformational and spatial restrictions. Secondary structure predictions, residue distributions and hydrophilicity profiles outline a common, N-terminal coenzyme-binding domain similar to that of other dehydrogenases, and a C-terminal domain with unique segments and presumably individual functions in each case. Strictly conserved residues of possible functional interest are limited, essentially only three polar residues. Asp64, Tyr152 and Lys156 (in the numbering of Drosophila alcohol dehydrogenase), but no histidine or cysteine residue like in the completely different, classical medium-chain alcohol dehydrogenase family. Asp64 is in the suggested coenzyme-binding domain, whereas Tyr152 and Lys156 are close to the center of the protein chain, at a putative inter-domain, active-site segment. Consequently, the overall comparisons suggest the possibility of related mechanisms and domain properties for different members of the short-chain family.

The enzyme NAD(P)H:flavin oxidoreductase (flavin reductase) catalyzes the reduction of soluble flavins by reduced pyridine nucleotides. In Escherichia coli it is part of a multienzyme system that reduces the Fe(III) center of ribonucleotide reductase to Fe(II) and thereby sets the stage for the generation by dioxygen of a free tyrosyl radical required for enzyme activity. Similar enzymes are known in other organisms and may more generally be involved in iron metabolism. We have now isolated the gene for the E. coli flavin reductase from a lambda gt11 library. After DNA sequencing we found an open reading frame coding for a polypeptide of 233 amino acids, with a molecular weight of 26,212 and with an N-terminal segment identical to that determined by direct Edman degradation. The coding sequence is preceded by a weak ribosome binding site centered 8 nucleotides from the start codon and by a promoterlike sequence centered at a distance of 83 nucleotides. In a Kohara library the gene hybridized to position 3680 on the physical map of E. coli. A bacterial strain that overproduced the enzyme approximately 100-fold was constructed. The translated amino acid sequence contained a potential pyridine nucleotide-binding site and showed 25% identity with the C-terminal part of one subunit (protein C) of methane monooxygenase from methanotropic bacteria that reduces the iron center of a second subunit (protein A) of the oxygenase by pyridine nucleotides.

During anaerobic growth of Escherichia coli an oxygen-sensitive ribonucleoside-triphosphate reductase, different from the aerobic ribonucleoside diphosphate-reductase (EC 1.17.4.1), produces the deoxyribonucleoside triphosphates required for DNA replication. The gene for the anaerobic enzyme has now been cloned and was found to contain a 2136-nucleotide coding region, corresponding to 712 amino acid residues, and an Fnr binding site 228 base pairs upstream of the initiator ATG. The deduced amino acid sequence shows 72% identity to a gene of coliphage T4, sunY, hitherto of unknown function, suggesting that the virus codes for its own anaerobic reductase. The location of an organic free radical formed during activation of the bacterial anaerobic reductase is proposed to be on Gly-681, since the pentapeptide RVCGY at positions 678-682 shows a striking similarity to the C-terminal sequence. RVSGY, of pyruvate formate-lyase. During activation of the anaerobically induced pyruvate formate-lyase, the glycine residue of the pentapeptide becomes an organic radical [Wagner, A. F. V., Frey, M., Neugebauer, F. A., Schäfer, W. & Knappe, J. (1992) Proc. Natl. Acad. Sci. USA 89, 996-1000]. The gene for the anaerobic reductase is located at a position around 96 min on the E. coli genomic map.

A specific ribonucleoside triphosphate reductase is induced in anaerobic Escherichia coli. This enzyme, as isolated, lacks activity in the test tube and can be activated anaerobically with S-adenosylmethionine, NADPH, and two previously uncharacterized E. coli fractions. The gene for one of these, previously named dA1, was cloned and sequenced. We found an open reading frame coding for a polypeptide of 248 amino acid residues, with a molecular weight of 27,645 and with an N-terminal segment identical to that determined by direct Edman degradation. In a Kohara library, the gene hybridized between positions 3590 and 3600 on the physical map of E. coli. The deduced amino acid sequence shows a high extent of sequence identity with that of various ferredoxin (flavodoxin) NADP+ reductases. We therefore conclude that dA1 is identical with E. coli ferredoxin (flavodoxin) NADP+ reductase. Biochemical evidence from a bacterial strain, now constructed and overproducing dA1 activity up to 100-fold, strongly supports this conclusion. The sequence of the gene shows an apparent overlap with the reported sequence of mvrA, previously suggested to be involved in the protection against superoxide (M. Morimyo, J. Bacteriol. 170:2136-2142, 1988). We suggest that a frameshift introduced during isolation or sequencing of mvrA caused an error in the determination of its sequence.