Cloning and nucleotide sequences of the genes for the subunits of NAD-reducing hydrogenase of Alcaligenes eutrophus H16

The genes hoxF, -U, -Y, and -H which encode the four subunit polypeptides alpha, gamma, delta, and beta of the NAD-reducing hydrogenase (HoxS) of Alcaligenes eutrophus H16, were cloned, expressed in Pseudomonas facilis, and sequenced. On the basis of the nucleotide sequence, the predicted amino acid sequences, and the N-terminal amino acid sequences, it was concluded that the structural genes are tightly linked and presumably organized as an operon, denoted hoxS. Two pairs of -24 and -12 consensus sequences resembling RpoN-activatable promoters lie upstream of hoxF, the first of the four genes. Primer extension experiments indicate that the second promoter is responsible for hoxS transcription. hoxF and hoxU code for the flavin-containing dimer (alpha and gamma subunits) of HoxS which exhibits NADH:oxidoreductase activity. A putative flavin-binding region is discussed. The 26.0-kilodalton (kDa) gamma subunit contains two cysteine clusters which may participate in the coordination of two [4F3-4S]centers. The genes hoxY and hoxH code for the small 22.9-kDa delta subunit and the nickel-containing 54.8-kDa beta subunit, respectively, of the hydrogenase dimer of HoxS. The latter dimer exhibits several conserved regions found in all nickel-containing hydrogenases. The roles of these regions in coordinating iron and nickel are discussed. Although the deduced amino acid sequences of the delta and beta subunits share some conserved regions with the corresponding polypeptides of other [NiFe] hydrogenases, the overall amino acid homology is marginal. Nevertheless, significant sequence homology (35%) to the corresponding polypeptides of the soluble methylviologen-reducing hydrogenase of Methanobacterium thermoautotrophicum was found. Unlike the small subunits of the membrane-bound and soluble periplasmic hydrogenases, the HoxS protein does not appear to be synthesized with an N-terminal leader peptide.

discussed. The 26.0-kilodalton (kDa) y subunit contains two cysteine clusters which may participate in the coordination of two [4Fe-4S] centers. The genes hoxY and hoxH code for the small 22.9-kDa 8 subunit and the nickel-containing 54.8-kDa ,B subunit, respectively, of the hydrogenase dimer of HoxS. The latter dimer exhibits several conserved regions found in all nickel-containing hydrogenases. The roles of these regions in coordinating iron and nickel are discussed. Although the deduced amino acid sequences of the 8 and ,1 subunits share some conserved regions with the corresponding polypeptides of other [NiFe] hydrogenases, the overall amino acid homology is marginal. Nevertheless, significant sequence homology (35%) to the corresponding polypeptides of the soluble methylviologen-reducing hydrogenase of Methanobacterium thermoautotrophicum was found. Unlike the small subunits of the membrane-bound and soluble periplasmic hydrogenases, the HoxS protein does not appear to be synthesized with an N-terminal leader peptide.
Alcaligenes eutrophus H16 is a gram-negative facultatively lithoautotrophic bacterium which can assimilate CO2 and utilize H2 as an energy source (reviewed in reference 5). Hydrogen oxidation is catalyzed by hydrogenases. These enzymes are found in phylogenetically diverse microorganisms, where they are responsible for both consumption and production of H2 (reviewed in reference 13). Two distinct hydrogenases have been purified and characterized from extracts of A. eutrophus (39,44). These two enzymes differ in cellular localization, cofactor content, polypeptide composition, and apparent molecular weight. The membranebound hydrogenase (HoxP) is a representative of the more common hydrogenase type. It consists of two heterologous polypeptides and is coupled to the respiratory chain via an unknown electron acceptor (43). The second hydrogenase (HoxS) of A. eutrophus resides in the cytoplasm (35). It is composed of four heterologous polypeptides, reduces NAD as the physiological electron acceptor, and contains flavin mononucleotide as the prosthetic group. Hydrogenases of this type have been found only in Alcaligenes species and in the gram-positive lithoautotrophic bacterium Nocardia opaca (reviewed in reference 5). A multimeric hydrogenase which utilizes deazaflavin as electron acceptor has been identified in methanogenic bacteria (23).
Three classes of hydrogenases can be distinguished on the basis of their metal content. The class of iron [Fe] hydrogenases is characterized by two ferredoxinlike [4Fe-4S] clusters and a third atypical iron-sulfur center supposedly involved in the reaction with H2. Both A. eutrophus hydrogenases belong to the nickel-iron [NiFe] type. These * Corresponding author. enzymes typically have two [4Fe-4S] centers and one [3Fe-3S] cluster in addition to nickel. A third class of hydrogenases contains nickel, iron, and selenium [NiFeSe]. An enzyme of this type has been described for Desulfovibrio and Methanococcus strains (reviewed in reference 13).
The genes coding for the two hydrogenases of A. eutrophus H16 lie in a cluster of genes on a 450-kilobase-pair (kb) conjugative megaplasmid (15,24). Molecular cloning of megaplasmid DNA in Escherichia coli and screening of the resultant hybrid plasmids for complementation of hydrogenase-deficient mutants led to the identification of hydrogenase (hox) genes. Subsequent studies revealed that the two structural gene loci hoxS and hoxP coding for the soluble NAD-linked hydrogenase and the membrane-bound hydrogenase, respectively, mark the left and right borders of the hox gene complex (12).
In this communication, we report the cloning of the hoxS locus, consisting of the four structural genes, hoxF, hoxU, hoxY, and hoxH, and the heterologous expression of this locus in Pseudomonas facilis. The complete nucleotide sequence of the hoxS region was determined, and the primary amino acid sequence was compared with that of other hydrogenases. Putative cofactor-binding sites are discussed.

MATERIALS AND METHODS
Organisms and plasmids. The bacterial strains, vectors, and recombinant plasmids used in this study are shown in Table 1.
Media and growth conditions. Strains of A. eutrophus and P. facilis were grown in mineral salts medium (40). Autotrophic cultures were grown under an atmosphere of hydrogen, oxygen, and carbon dioxide in a ratio of 8:1:1 (vol/vol/vol). Organic carbon sources were routinely added at a concentration of 0.4% (wt/vol). The concentration of the nitrogen source was 0.2% (wt/vol). Strains of E. coli were propagated in M9 medium (29) or in Luria broth (LB). LB medium with 0.2% (wt/vol) sodium chloride was used as the complex medium for the cultivation of hydrogen-oxidizing bacteria. Selective media contained antibiotics at the following concentrations: 15 jxg of tetracycline per ml and 350 ,ug of kanamycin per ml for A. eutrophus; 3 ,ug of tetracycline per ml for P. facilis; 15 ,ug of tetracycline per ml, 30 ,ug of kanamycin per ml, and 50 pug of ampicillin per ml for E. coli. Cloning of hoxS DNA. A cosmid library of megaplasmid pHG1 was constructed as described previously (12). pHG1 DNA was partially digested with the restriction endonuclease HindIlI and ligated with HindIlI-digested cosmid vector pVK102. The ligation products were packaged in vitro into bacteriophage lambda with a DNA packaging kit (Boehringer & Soehne GmbH, Mannheim, Federal Republic of Germany) according to the instructions of the manufacturer. After infection of E. coli HB101, tetracycline-resistant colonies were selected and screened for the presence of hydrogenase genes.
To identify hoxS clones, we transferred the cosmids to HoxSmutants of A. eutrophus (Table 1) via a triparental cross with pRK2013 as the mobilizing vector (11). Transconjugants were selected on mineral medium with 15 ,ug of tetracycline per ml and tested for autotrophic growth on hydrogen.
Standard DNA techniques. Standard DNA techniques were essentially as described by Maniatis et al. (29). Rapid, small-scale DNA isolation for clonal analysis was done by the method of Birnboim and Doly (4). For large-scale preparation of vector and recombinant plasmid DNA from E. coli, the DNA was further purified by ethidium bromidecesium chloride gradient centrifugation. Single-stranded template DNA was prepared after precipitation of the phage with polyethylene glycol. E. coli strains were transformed with the ligated DNA by the method of Mandel and Higa (28). DNA-DNA hybridization was conducted by the method of Southern (46). For labeling and detection of DNA, nick translation kits and Blue Gene kits (GIBCO-Bethesda Research Laboratories, Eggenstein, Federal Republic of Germany) were used as recommended by the manufacturer.
DNA sequence determination and analysis. Overlapping restriction fragments were cloned into phage M13mpl8 and M13mpl9 (32) or into plasmids pTZ18R and pTZ19R (33). Nested deletions were constructed by unidirectional exonuclease III-S1 nuclease digestion (Erase-a-base-system; Promega Corp., Madison, Wis.). The dideoxy-chain termination method was applied (37) with 35S-labeled a-dATP from Dupont, NEN Research Products (Dreieich, Federal Republic of Germany). Sequencing reactions were done with Klenow-polymerase, using 7-deaza-dGTP to avoid compression (pUC sequencing kit; Boehringer) or alternatively with T7 DNA polymerase (T7 sequencing kit; Pharmacia, Freiburg, Federal Republic of Germany) or Taq polymerase (Taquence DNA sequencing kit; United States Biochemical Corp., Cleveland, Ohio). The complete sequence was compiled from overlapping partial sequences determined throughout for both strands. Nucleotide and amino acid sequences (open reading frames, translation into amino acids, codon usage, sequence comparison, hydrophobicity) were analyzed with the programs of the Mac Molly sequence analysis package (Softgene GmbH, Berlin, Federal Republic of Germany).  I  I   I1   I  I  I  I  I  I   I   I  I  I  I   I  itoring hydrogen-dependent NAD reduction with detergenttreated cells (14). Ouchterlony immunodiffusion analyses were performed as described previously (15). Protein was determined with bovine serum albumin as the standard (27).
Chemicals. Restriction endonucleases, T4 DNA ligase, and the lambda DNA packaging kit were obtained from Boehringer & Soehne GmbH. All other chemicals were from E. Merck AG, (Darmstadt, Federal Republic of Germany).

RESULTS
Cloning of structural genes for NAD-reducing hydrogenase (HoxS) from A. eutrophus H16. Previous studies showed that genes coding for HoxS map in a 100-kb long region of megaplasmid pHG1 (24). An 11.6-kb EcoRI fragment from this region restored HoxS activity in some but not all mutants bearing structural gene mutations (12), indicating that only a part of the hoxS locus was present. To obtain a clone with the complete locus, we constructed a cosmid library of pHG1 by partial HindIII digestion. Resultant fragments were cloned into the broad-host-range vector pVK102 (11). Six recombinant plasmids were identified, which efficiently complemented all HoxSmutants. These recombinants shared a common 15-kb HindIII fragment, which was subcloned into pVK102 yielding the cosmid pGE15 ( Table 1). The 15-kb HindIII fragment and the 11.6-kb EcoRI fragment overlapped for a distance of 2.3 kb (Fig. 1).
Indirect evidence for the assumption that the 15-kb HindIll fragment encodes the entire set of structural genes for the NAD-reducing hydrogenase was obtained by heterologous expression of the hoxS DNA in P. facilis, which contains only a membrane-bound hydrogenase. Transfer of the recombinant cosmid pGE15 into two wild-type strains of P. facilis yielded transconjugants which grew faster on hydrogen than the parent strains. Thus, the aquisition of HoxS activity apparently enhanced the lithoautotrophic growth of P. facilis (Table 2). Enzymatic assays showed that the HoxS-containing transconjugants of P. facilis obtained NAD-reducing hydrogenase activity ( Table 2). The HoxS protein of the P. facilis transconjugants was immunologically identical to that of A. eutrophus ( Fig. 2A). In this context a prominent protein of A. eutrophus, designated B protein, is of interest. Its formation is coordinately regulated with HoxS activity, and it occurs only in those hydrogen bacteria which contain NAD-reducing hydrogenase. The physiological function of the B protein is still unknown (21). It was absent in wild-type strains of P. facilis but immunologically detectable in transconjugants containing the hoxS cosmid pGE15 (Fig. 2B). Thus, the gene of the B protein must also be located on the 15  0.10,a 0o35b a Cells were grown in mineral salts medium in an atmosphere of H2-C02-02 (8:1:1, vol/vol/vol). b Cells were grown in fructose-glycerol medium (14).
coli were unsuccessful. This may be attributable to the complex regulation of hox gene expression, to the inactivity of the hoxS promoter in E. coli, and/or to the requirement for specific processing of the HoxS gene products.
Nucleotide and derived amino acid sequences of NADreducing hydrogenase. The region presumed to encode the structural genes was sequenced. The nucleotide sequence revealed four adjacent open reading frames, each of which is preceded by a tentative ribosomal binding site (underlined in Fig. 3). Codon usage (data not shown) was in good agreement with that of previously sequenced genes of A. eutrophus (1,20). The open reading frames were designated as hoxF, -U, -Y, and -H. On the basis of comparison of the deduced amino acid sequences with the previously published N-terminal amino acid sequences (54), it was possible to assign these four genes to the subunits a, -y, 8, and of HoxS. Some discrepancies between the experimentally determined amino acid sequences and the deduced sequences were found. Our data predict Trp at position 24 of the a subunit and Val, Ala, and Val at positions 29, 31, 36, respectively, of the , subunit. Interestingly, the same amino acids were found at the corresponding positions in the NAD-reducing hydrogenase of N. opaca (54). Subunits ,B and y lack the initial methionine, which is apparently removed posttranslationally. The deduced molecular weights of the subunits were in good agreement with the biochemical values (Table 3).
Stop and start codons of the genes hoxF, -U, and -Y were found to overlap. The genes hoxY and hoxH are separated by 20 bases (Fig. 3) This arrangement predicts an organization as an operon. Two possible RpoN-specific promoter sequences with typical -24 and -12 consensus elements (17 lie upstream of the hoxF start codon, whereas typical procaryotic -35 and -10 consensus sequences were absent in this region. Primer extension experiments (data not shown) revealed a signal at nucleotide 654 ( Fig. 3) indicating that this is the transcription start point of the hoxS operon. This is compatible with the assumption that promoter 2 does in fact direct transcription of the hoxS operon. A large inverted repeat (indicated by arrows in Fig. 3) immediately following the hoxH stop codon could form a hairpinlike structure and may be involved in transcription termination. The free energy of this structure is 35.8 kcal (approximately 149.6 kJ)/mol according to the base-pairing rules of Tinoco et al. (47).
Amino acid sequence comparison of hydrogenase dimer. Sequence data are now available for six hydrogen-consuming hydrogenases (25,26,30,34,38,50,51). Similarities exist between [NiFe] and [NiFeSe] enzymes. They are predominantly heterodimers of comparable molecular weight. The respective structural genes are apparently organized as operons in which the genes for the small subunits precede those for the large ones (reviewed in reference 13). The genes for the NAD-dependent hydrogenase of A. eutrophus show a similar organization: the genes hox Y and hoxH encoding the hydrogenase dimer lie downstream of the genes hoxF and hoxU coding for the flavin-containing moiety of HoxS.
While there is little homology between the predicted primary amino acid sequences of the 8 and P subunits of HoxS and the corresponding membrane-bound and soluble periplasmic polypeptides, a considerable overall homology to the soluble hydrogenase subunits -y (34.0%) and 8 (36.6%) of Methanobacterium thermoautotrophicum AH (34) was found (Fig. 4).
The deduced amino acid sequences reveal that the 8 subunit of HoxS contains nine cysteine residues. A total of 9 to 13 cysteines are commonly found in the small hydrogenase subunits of various organisms (Fig. 5A). A number of these are situated at homologous positions within conserved sequence motifs (34). A conserved domain at the C terminus is not found in the small polypeptide of A. eutrophus; it seems to be truncated (Fig. 5A). On the other hand, an additional stretch of amino acids occurs at the N terminus which is missing in the corresponding polypeptides. Since this N-terminal moiety is present in the mature 8 subunit (54), a leader peptide, which has been predicted for the small subunits of membrane-bound and soluble periplasmic hydrogenases, can be excluded.
Our data indicate the presence of six cysteine residues in the ,1 subunit, four of which are highly conserved. Two of these are located near the N terminus, and the other two are at the C terminus. The primary amino acid sequence in their immediate neighborhood is also highly conserved. The cysteine clusters are located in hydrophobic regions, which is most explicit in the A. eutrophus enzyme (Fig. SB).
Comparison of flavin-containing HoxS dimer and other flavoproteins. Subunits a and -y of HoxS constitute an NADH:oxidoreductase and contain one molecule of enzyme-bound flavin mononucleotide (42,45). Comparison of the amino acid sequences of the a subunit and flavincontaining succinate dehydrogenase (53), fumarate reductase (9), and glycerol-3-phosphate dehydrogenase (10) from E. coli revealed no extensive homology (data not shown). In the N-terminal region, we found a pair of glycines (residues 42 and 44, Fig. 3 Schneider and Schlegel (44).

DISCUSSION
The four structural genes of the NAD-reducing hydrogenase of A. eutrophus H16 are tightly linked. Genes hoxF and hoxU and in turn hoxU and hoxY overlap at their respective stop and start codons. Genes hoxY and hoxH are separated by an intergenic region of 20 bases. This suggests that the four genes belong to an operon. Immunochemical analyses of HoxSinsertion mutants with antibodies raised against the individual subunits revealed a polar effect of mutations in hoxF and hoxU on the expression of the downstream genes hoxY and hoxH (U. Warnecke and B. Friedrich, unpublished data), indicating that all four genes are transcribed from a common promoter upstream of hoxF.
It has been shown previously that an rpoN-like gene of A. eutrophus, designated hno, controls the expression of diverse metabolic pathways including hydrogen oxidation (36). Various evidence suggests that this gene encodes an alter-native sigma factor of the RNA polymerase. Indeed, sequence analysis of the hno gene revealed extensive homology to the ntrA (rpoN) genes of enteric bacteria (J. Warrelmann and B. Friedrich, unpublished data). Not surprisingly, therefore, -24 and -12 consensus elements were found upstream of hoxF. These sequences are typical of RpoN-activatable promoters (3). The large inverted repeat immediately downstream of the hoxH stop codon may have a role in transcription termination. Preliminary DNA-RNA hybridization experiments indicate that the hoxS region directs the formation of an approximately 5-kb large transcript of hoxS (U. Oelmuller and C. G. Friedrich, personal communication). This length is in good agreement with the value postulated from the sequence data (4.8 kb).
A model for the spacial and functional organization of the NAD-reducing hydrogenase of A. eutrophus is presented in Fig. 6. This model is based on the results of sequence analyses, ultrastructural electron microscopic investigations (48; W. Johannssen and F. Mayer, personal communication), biochemical data, and electron spin resonance spectroscopy (45). The essential features of this model can be summarized as follows: The so-called hydrogenase dimer, consisting of subunits 1B and 8, contains the catalytic center involved in the generation of protons and reducing equivalents. Alignment of the amino acid sequences of the ,B subunit and the corresponding polypeptides of [NiFe] and [NiFeSe] hydrogenases suggests that highly conserved amino acids, in particular cysteine residues at the C-and N-terminal regions, play an important role in the catalytic function of the enzyme.
The nine cysteine residues of the small 8   [4Fe-4S] center and one [3Fe-3S] center. These clusters have been detected previously by electron spin resonance spectroscopy (45). The [3Fe-3S] center is presumably located in the small 8 subunit. The function of this subunit in catalysis is at present unclear. It may be involved in redox regulation of hydrogenase activity (8). Alternatively, it may represent an intermediate site in electron transport to the acceptor site. The acceptor site is tentatively assigned to the moiety of HoxS (45) composed of subunits a and -y (Fig. 6). The -y subunit contains sufficient cysteinyl residues to enable coordination of two [4Fe-4S] centers. Similar [4Fe-4S] centers have been identified in the NADH:oxidoreductase dimer of N. opaca (42). The predicted amino acid sequence reveals similarities to bacterial ferredoxins (7). Unlike the ferredoxins, however, the cysteine clusters are split into two parts, with the two parts separated by 47 and 46 amino acids (aa) respectively: (i) Cys-Leu-2aa-Cys-2aa-Cys-47aa-Cys-Pro and (ii) Cys-Ile-2aa-Cys-2aa-Cys-46aa-Cys-Pro.
The [2Fe-2S] center (Fig. 6) is spectroscopically distinguishable from the [4Fe-4S] cluster by its slow electron spin relaxation rate (45). Our data suggest that this cluster is coordinated by a series of cysteine residues of the a subunit. Three of these cysteines lie closely clustered: Cys-2aa-Cys-2aa-Cys at position 499. It is unclear which of the other cysteines belongs to this cluster. Sequence comparisons revealed that the a subunit contains some structural features of putative flavin-binding sites. This may be the site where reduction of the electron acceptor NAD+ takes place. Since NAD+ accepts two reducing equivalents at a time, flavin ,  (49). Therefore, reduction of flavin mononucleotide must involve two such clusters acting cooperatively. The pair of [4Fe-4S] clusters of the -y subunit may mediate this step. Sequence data indicate that archaebacterial and eubacterial hydrogenases have evolved from a common ancestor. Remarkably, the soluble methylviologen-reducing hydrogenase of M. thermoautotrophicum (34) and the hydrogenase dimer of HoxS of A. eutrophus share only slight homology with the corresponding membrane-bound and soluble periplasmic enzymes found in aerobic and anaerobic bacteria including HoxP of A. eutrophus (K. Horstmann, C. Kortluke, and B. Friedrich, unpublished data). In this context, it is worth noting that the plasmid-encoded hoxS genes of A. eutrophus are flanked by two copies of an insertion element (E. Schwartz and B. Friedrich, unpublished data). This transposonlike structure suggests that mobile elements play a role in the evolution of the lithoautotrophy gene cluster of megaplasmid pHG1.