Alkyl Hydroperoxide Reductase from Salmonella typhimurium SEQUENCE AND HOMOLOGY TO THIOREDOXIN REDUCTASE AND OTHER FLAVOPROTEIN DISULFIDE OXIDOREDUCTASES*

The DNA sequence of the Salmonella typhimurium ahp locus was determined. The locus was found to contain two genes that encode the two proteins (C22 and F52a) that comprise the S. typhimurium alkyl hydroperoxide reductase activity. The predicted sequence of the F52a protein component of the alkyl hydroperoxide reductase was found to be highly homologous to the Escherichia coli thioredoxin reductase protein (34% identity with many conservative substitutions). The homology was found to be particularly striking in the region containing the redox-active cysteines of the thioredoxin reductase molecule, and among the identities were the redox-active cysteines themselves. Aside from the strong similarity to thioredoxin reductase, overall homology between the F52a protein and other flavoprotein disulfide oxidoreductases such as glutathione reductase, dihydrolipoamide dehydrogenase, and mercuric reductase was found to be rather limited, and the conserved active site segment common to the three proteins was not observed within the F52a protein. However, three short segments that have been implicated in FAD and NAD binding were found to be conserved between the F52a protein and the other disulfide reductases. These results suggest that the alkyl hydroperoxide reductase is the second known member of a class of disulfide oxidoreductases which was represented previously by thioredoxin reductase alone; they also allow the putative assignment of several functional domains.

The DNA sequence of the Salmonella typhimurium ahp locus was determined.
The locus was found to contain two genes that encode the two proteins (C22 and F52a) that comprise the S. typhimurium alkyl hydroperoxide reductase activity.
The predicted sequence of the F52a protein component of the alkyl hydroperoxide reductase was found to be highly homologous to the Escherichia coli thioredoxin reductase protein (34% identity with many conservative substitutions).
The homology was found to be particularly striking in the region containing the redox-active cysteines of the thioredoxin reductase molecule, and among the identities were the redox-active cysteines themselves.
Aside from the strong similarity to thioredoxin reductase, overall homology between the F52a protein and other flavoprotein disulfide oxidoreductases such as glutathione reduetase, dihydrolipoamide dehydrogenase, and mercuric reductase was found to be rather limited, and the conserved active site segment common to the three proteins was not observed within the F52a protein.
However, three short segments that have been implicated in FAD and NAD binding were found to be conserved between the F52a protein and the other disulfide reductases. These results suggest that the alkyl hydroperoxide reductase is the second known member of a class of disulfide oxidoreductases which was represented previously by thioredoxin reductase alone; they also allow the putative assignment of several functional domains.
The alkyl hydroperoxide reductase activity of Salmonella typhimurium and Escherichia coli was first recognized through the isolation of mutant cells that were resistant to mutagenesis by alkyl hydroperoxides (1). By using as an assay the conversion of a model alkyl hydroperoxide to its corresponding alcohol, the activity was purified to homogeneity from a regulatory mutant strain that produced it at high levels (1). The purified activity was found to be composed of two proteins that had been identified previously by two-dimensional gel electrophoresis as hydrogen peroxide-inducible proteins F52a and C22, both of which were known to be positively regulated by the S. typhimurium oxyR locus (2). The F52a protein was found to be composed of two identical chains, each containing a bound FAD cofactor and having a subunit molecular mass of -57 kDa. It can use either NADH or NADPH as electron donors for the direct reduction of redox dyes or of alkyl hydroperoxides when combined with the C22 protein. Mutant S. typhimurium strains that lack the alkyl hydroperoxide reductase are extremely sensitive to killing by organic hydroperoxides, and therefore a likely in uiuo function for this activity would be the detoxification of lipid and other hydroperoxides that are produced during an oxidative stress (3).
Four other well studied flavoprotein oxidoreductases are glutathione reductase, mercuric reductase, dihydrolipoamide dehydrogenase, and thioredoxin reductase (4). Although each of these four enzymes is specific for a different substrate, each contains a pair of redox-active cysteines involved in the transfer of reducing equivalents from the FAD cofactor to the substrate. Although the mechanism and active site segment of the alkyl hydroperoxide reductase have not yet been determined, the activity has been shown to be inhibited by thiolreactive reagents, suggesting that its mechanism may also involve catalytic cysteine residues (1).
The amino acid sequences of glutathione reductase, mercuric reductase, and dihydrolipoamide dehydrogenase have each been determined from two or more organisms (5-12). These sequences show substantial sequence homology to one another, particularly around the 2 redox-active cysteine residues that are separated by 4 conserved amino acids. However, the sequence of E. coli thioredoxin reductase shows quite poor overall sequence homology to these other disulfide reductases, even around the active site, and contains 2 redox-active cysteines separated by only 2 amino acids (13,14).
We report here the DNA sequence of both genes encoding the alkyl hydroperoxide reductase activity located near 13 min on the S. typhimurium chromosome. The predicted amino acid sequence of the F52a protein is compared with the known sequences of other flavoprotein disulfide oxidoreductases and is shown to be highly homologous to the thioredoxin reductase molecule although exhibiting similarities to other reductases that are limited to short segments probably involved in FAD and NAD binding. Reactions were primed with either the "universal primer" or oligonucleotides complementary to insert sequences as described in Fig. 1.

Organization and Nucleotide
Sequence of the ahp Locus-In a previous study we reported the isolation of a plasmid (pAQ9) from a S. typhimurium library that contained the ahp locus on an 8.6-kilobase fragment (3) showed that a smaller DNA fragment, a 3.0-kilobase BglII-EelI fragment from pAQ9 (see Fig. l), contained all sequences sufficient for both complementation of the extreme sensitivity to organic hydroperoxides characteristic of ahp deletion mutants and restoration of expression of the two proteins that comprise the alkyl hydroperoxide reductase activity (data not shown). The 3.0-kilobase BglII-Bali fragment was sequenced using the strategy shown in Fig. 1. The DNA and predicted amino acid sequences of the two long open reading frames contained on this clone are shown in Fig. 2 identified the probable position of the translation start point. We are designating this gene as nhpC. The C22 protein is predicted to be 183 amino acids long with a molecular mass of 20,699 daltons (minus the initiating methionine). This predicted mass is in reasonable agreement with a subunit molecular weight of 22,000 extrapolated from its mobility in denaturing polyacrylamide gels (1). Shortly upstream of the ahpC coding region is a sequence that closely matches the bacterial Shine-Dalgarno consensus sequence. The start point of the ahp transcript (indicated on Fig. 2) and the regulation of its promoter are described in a separate report (16). A comparison of the predicted amino acid sequence of the C22 protein with two protein data bases (NBRF/Dayhoff and translated GenBank) showed no significant sequence similarities to other known proteins.  (1) confirmed that this gene encodes the F52a protein component of the alkyl hydroperoxide reductase. We are designating this gene as ahpF. The N-terminal protein sequence was also used to identify the likely translation start point on the ahpF transcript, and inspection of this region revealed a Shine-Dalgarno consensus sequence just upstream of the initiating methionine codon. The F52a protein is predicted to be 521 amino acids long with a molecular mass of 55,959 daltons (without the FAD), in good agreement with a molecular weight of 57,000 extrapolated from its mobility on denaturing polyacrylamide gels (1). The NBRF/Dayhoff and translated GenBank data bases were searched for proteins with sequences similar to that of the F52a protein using the FASTP and TFASTN computer programs (21). The protein giving both the highest initial and optimized scores observed in any search was E. coli thioredoxin reductase (see Table I). Computer alignment of the entire E. coli thioredoxin reductase protein with residues 207-521 of the F52a protein revealed 34% identity and many conservative changes (Fig. 3). A particularly striking align-  (16) are also indicated. ment requiring no deletions or insertions occurs in the 65 showed that the overall sequence simiiarity to these enzymes residues surrounding the redox-active cysteines of the thio-was rather poor and limited to three short regions of strong redoxin reductase molecule in which there is 46% identity local sequence similarity. The redox-active segment, which is and many conservative changes. Among the identities are the highly conserved among glutathione reductase, mercuric reredox-active cysteines themselves. ductase, and dihydrolipoamide dehydrogenase, is not observed Comparison of the F52a Protein Sequence with That of Other in the F52a protein. The three regions of local sequence Disulfide Oxidoreductases-Computer searches of the NBRF/ similarity which are shared by the F52a protein and other Dayhoff and GenBank data bases using FASTP and TFASTN FAD-containing disulfide oxidoreductases are shown in Table  showed that the functionally related proteins glutathione II. A similar alignment has been used by Russel and Model reductase, mercuric reductase, and dihydrolipoamide dehydro-to show homologous segments between thioredoxin reductase genase also showed sequence similarity scores significantly and several other disulfide reductases (13). The first region above the mean score (Table I). Further analysis, however, of sequence similarity (the most N-terminal of the three  The corresponding region in human erythrocyte glutathione reductase has been implicated in the binding of FAD based on its position in the crystal structure (22). The second region of sequence similarity includes residues 357-371 of the F52a protein and shows strong homology to the known NADP-binding domain of glutathione reductase (9). The third region (residues 470-488 in the F52a protein), which is near the carboxyl end of each of the proteins, aligns well with the central domain of glutathione reductase, which has also been shown to make contacts with the FAD molecule (22). More recently, the analogous region in a crystal of Azotobacter vinilandii dihydrolipoamide dehydrogenase has also been shown to make contactswith FAD (23). \ DISCUSSION We have sequenced the S. typhimurium ahp locus and have shown that it contains the two genes that encode the two proteins comprising the alkyl hydroperoxide reductase activity. We have designated the more upstream of the two genes as ahpC (encoding the C22 protein) and the more downstream gene as ahpF (encoding the F52a protein). Earlier work has shown that the synthesis of both proteins is regulated by the oxyR locus (2), and an oxyR-regulated promoter has been identified just upstream of the ahpC gene (16). Since single Z'nlO insertions within the ahp locus eliminate both the basal and oxyR-stimulated levels of both the C22 and F52a proteins, the ahpC and ahpF genes are likely to comprise or be part of an operon with a promoter upstream of the ahpC gene regulating the levels of both gene products.
The purified alkyl hydroperoxide reductase has been well characterized with respect to its substrate specificity and cofactor requirements (1). However, little is known about the mechanism by which reducing equivalents are transferred from the NADH/NADPH cofactor to the organic peroxide substrates. Based on its functional relationship and size sim- Amino acids that are identical or similar between the alkyl hydroperoxide reductase and human glutathione reductase (as determined by FASTP) are marked with two or one asterisks (*), respectively. The locations of amino acid residues that are identical (A) or chemically similar (.) between the alkyl hydroperoxide reductase and at least seven of the nine other aligned proteins are indicated on the bottom line. *** ** l *. * l * l ** l * l t *a* * l * *t l *,******* l ** I) **i*** *** * l **** *tt********* ilarity to the well characterized disulfide oxidoreductases glutathione reductase, dihydrolipoamide dehydrogenase, and mercuric reductase, it was proposed that the catalytic cycle of the F52a protein involves the transfer of electrons from the NADH/NADPH cofactor to the FAD and then to a redoxactive disulfide (1). All three of the glutathione reductase family of proteins have been shown to contain a characteristic redox-active disulfide within a highly conserved active site region. However, comparison of the amino acid sequence of the F52a protein with the known amino acid sequences of glutathione reductase, dihydrolipoamide dehydrogenase, and mercuric reductase from several species shows no evidence for this conserved active site peptide within the F52a protein.
Also, the F52a protein shows rather poor overall homology to these proteins unlike the 26-30% identity seen among the three proteins.
There are, however, three short segments within the F52a protein which show significant homologies to these other reductases. Two of these segments (the more Nand carboxyl-terminal segments) align well with two segments in human glutathione reductase which have been shown to make important contacts with FAD. The middle segment within the F52a protein corresponds to the NADP-binding domain of glutathione reductase and is therefore likely to bind either NAD or NADP in the case of the F52a protein. This same arrangement of adenine dinucleotide-binding domains is found in other flavin reductases.
The protein thioredoxin reductase is another example of a well characterized disulfide oxidoreductase. However, several differences between it and the glutathione reductase family of disulfide reductases, particularly the lack of sequence conservation around the redox-active cysteines, have led to the proposal that thioredoxin reductase has arisen through convergent evolution toward a mechanism similar to that of these other reductases (25). The recently deduced sequence of the entire thioredoxin reductase molecule has been reported to show poor overall homology with glutathione reductase, di-hydrolipoamide dehydrogenase, and mercuric reductase (13), thus supporting that hypothesis. Comparison of the amino acid sequence of the F52a protein with that of thioredoxin reductase shows a high degree of similarity throughout the length of the thioredoxin reductase molecule. Among the identities between the two proteins are the known redoxactive cysteines of the thioredoxin reductase molecule, which are found in identical locations within the sequences of the two proteins relative to the NAD/NADP-binding sites (no deletions or insertions).
The strong homology between the two proteins makes it highly probable that the alkyl hydroperoxide reductase activity is a disulfide oxidoreductase and suggests the location of its redox-active cysteines within the primary structure of the molecule. The number of identities between the two proteins and the large number of substitutions that are frequently observed during protein evolution make it likely that these two proteins have diverged from a common ancestor. The F52a protein is therefore likely to be the second member of a class of disulfide oxidoreductases which is distinct from the class of reductases represented by glutathione reductase, dihydrolipoamide dehydrogenase, and mercuric reductase. Mutagenesis of the likely active site region of the alkyl hydroperoxide reductase should complement the ongoing mutagenesis and x-ray crystallography studies of the thioredoxin reductase molecule (27,28) and help elucidate the mechanistic details of this type of disulfide reductase.
The role of the C22 subunit of the alkyl hydroperoxide reductase activity is still not well understood.
No significant homologies were found between the C22 protein and other proteins in the available data bases, including the protein thioredoxin that, together with thioredoxin reductase, comprises a general protein disulfide-reducing system. The F52a subunit of the alkyl hydroperoxide reductase has been shown to be capable of reducing a number of substrates even in the absence of the C22 protein (1). However, the C22 protein subunit is required for the reduction of organic hydroperox-ides (1). It is therefore likely that the C22 protein plays a role in substrate binding and targeting the activity for hydroperoxide substrates. The role of the N-terminal domain of the F52a protein is also unclear. Although the F52a protein and thioredoxin reductase align well throughout the length of the thioredoxin reductase molecule, the smaller size of thioredoxin reductase (322 amino acids as opposed to 521 for the F52a protein) results in a segment of approximately 200 amino acids at the N terminus of the F52a protein which has no counterpart in thioredoxin reductase. This N-terminal region showed no significant similarities to other proteins in the NBRF/Dayhoff or translated GenBank data bases. Aside from the common subunit organization and sequence homology between the F52a protein and thioredoxin reductase, there is another striking similarity between these two proteins. Both interact with a smaller protein chain to fulfill their physiologically important function. It will therefore he interesting to see if other reductases that show homology to the F52a protein and thioredoxin reductase will be found whose function also requires an additional protein factor.