Eur. J. Biochem. 259, 235–243 (1999) q FEBS 1999 Desulfovibrio gigas neelaredoxin A novel superoxide dismutase integrated in a putative oxygen sensory operon of an anaerobe Gabriela Silva 1,2 , Solange Oliveira 1,3 , Cla ´udio M. Gomes 2 , Isabel Pacheco 2 , Ming Y. Liu 4 , Anto ´nio V. Xavier 2 , Miguel Teixeira 2 , Jean LeGall 2,4 and Claudina Rodrigues-Pousada 1,2 1 Instituto Gulbenkian de Cie ˆncia, Oeiras, Portugal; 2 Instituto de Tecnologia Quı ´mica e Biolo ´gica, Universidade Nova de Lisboa, Oeiras, Portugal; 3 Universidade E ´ vora, Departamento de Biologia, E ´ vora, Portugal; 4 Department of Biochemistry and Molecular Biology, University of Georgia, Athens, GA, USA Neelaredoxin, a small non-heme blue iron protein from the sulfate-reducing bacterium Desulfovibrio gigas [Chen, L., Sharma, P., LeGall, J., Mariano, A.M., Teixeira M. and Xavier, A.V. (1994) Eur. J. Biochem. 226, 613–618] is shown to be encoded by a polycistronic unit which contains two additional open reading frames (ORF-1 and ORF-2) coding for chemotaxis-like proteins. ORF-1 has domains highly homologous with those structurally and functionally important in methyl-accepting chemotaxis proteins, including two putative transmembrane helices, potential methylation sites and the interaction domain with CheW proteins. Interestingly, ORF-2 encodes a protein having homologies with CheW proteins. Neelaredoxin is also shown to have significant superoxide dismutase activity (1200 U·mg –1 ), making it a novel type of iron superoxide dismutase. Analysis of genomic data shows that neelaredoxin-like putative polypeptides are present in strict anaerobic archaea, suggesting that this is a primordial superoxide dismutase. The three proteins encoded in this operon may be involved in the oxygen-sensing mechanisms of this anaerobic bacterium, indicating a possible transcriptional mechanism to sense and respond to potential stress agents. Keywords: anaerobes; oxygen-sensing mechanisms; sulfate-reducing bacteria; superoxide dismutase. The advent of an oxygen-rich atmosphere led to the appearance of more energetically efficient organisms, but at the same time introduced a stressful environment for both aerobic and anaerobic organisms. In fact, oxygen reduction leads to potentially hazardous species, such as the superoxide anion radical, the hydroxyl radical and hydrogen peroxide. To detoxify these products, living organisms developed an array of protective mechanisms, distributed to varying extents in the three life domains. In particular, catalase and superoxide dismutase (SOD) have been found in both aerobic and anaerobic prokaryotes . Sulfate-reducing bacteria were for a long time regarded as strict anaerobes , until it was shown that Desulfovibrio spp., although not capable of growing in the presence of oxygen, may not only survive in its presence but even use it to produce ATP [3,4]. In D. gigas a complete redox chain linking the oxidation of NADH to the reduction of dioxygen to water was characterized [5,6], in which a small mononuclear iron protein, rubredoxin , transfers electrons between an NADH oxidase  and an oxygen reductase [5,6]. Moreover, catalase and SOD have been shown to exist in some Desulfovibrio species . Taken together, these strategies show how organisms adapted and responded to the paradox resulting from the introduction of oxygen in their environment. Sulfate-reducing bacteria contain a large array of other mononuclear iron proteins, the function of which in most cases remains elusive. Recent reports have indirectly suggested that one such protein, desulfoferrodoxin (Dfx) , may play a role in defense mechanisms against oxygen radicals. In fact, overexpression of Desulfoarculus baarsii Dfx in an Escherichia coli mutant deficient in cytoplasmic SOD led to the unexpected result of restoring the wild-type phenotype. However, the protein was suggested to have little SOD activity [11,12]. Dfx is a small protein (14 kDa) present in several Desulfovi- brio species. It contains two mononuclear iron centers: center I, with a distorted tetrahedral cysteinyl sulfur co-ordination; and center II, having a square pyramidal geometry, with four histidine residues in the basal plane and a cysteine residue in the axial position . This protein, both in terms of amino acid sequence and tridimensional structure, seems to be built of two domains, each containing one of the iron centers. The first domain (residues 1–32) is very similar to desulforedoxin (Dx), a small iron protein so far only isolated from D. gigas , which has one iron atom bound in a distorted tetrahedral co-ordination to four cysteines . The second domain has homology with another mononuclear iron protein isolated from D. gigas which is blue and thus named neelaredoxin (Nlr). Nlr is a small protein (15 kDa) with two mononuclear iron sites. Its visible spectrum, consisting of a broad band at < 660 nm, is remarkably similar to that of center II of Dfx [16,17]. This observation, together with the N-terminal sequence available, showing high homology Correspondence to C. Rodrigues-Pousada, Instituto Gulbenkian de Cie ˆncia, 2781 Oeiras, Portugal. Tel. +351-1-4407900, Fax: +351-1-4407970, E-mail: [email protected]Abbreviations: SOD, superoxide dismutase; Dfx, desulfoferrodoxin; Dx, desulforedoxin; Nlr, neelaredoxin; MCP, methylaccepting chemotaxis protein. Note: The coding region of neelaredoxin has been deposited in GenBank with the accession number AF096317. (Received 20 July 1998; accepted 13 October 1998)
Desulfovibrio gigas neelaredoxinA novel superoxide dismutase integrated in a putative oxygen sensory operon of an anaerobe
Gabriela Silva1,2, Solange Oliveira1,3, ClaÂudio M. Gomes2, Isabel Pacheco2, Ming Y. Liu4, AntoÂnio V. Xavier2,
Miguel Teixeira2, Jean LeGall2,4 and Claudina Rodrigues-Pousada1,2
1Instituto Gulbenkian de CieÃncia, Oeiras, Portugal; 2Instituto de Tecnologia QuõÂmica e BioloÂgica, Universidade Nova de Lisboa, Oeiras, Portugal;3Universidade EÂ vora, Departamento de Biologia, EÂ vora, Portugal; 4Department of Biochemistry and Molecular Biology, University of Georgia,
Athens, GA, USA
Neelaredoxin, a small non-heme blue iron protein from the sulfate-reducing bacterium Desulfovibrio gigas [Chen,
L., Sharma, P., LeGall, J., Mariano, A.M., Teixeira M. and Xavier, A.V. (1994) Eur. J. Biochem. 226, 613±618] is
shown to be encoded by a polycistronic unit which contains two additional open reading frames (ORF-1 and ORF-2)
coding for chemotaxis-like proteins. ORF-1 has domains highly homologous with those structurally and functionally
important in methyl-accepting chemotaxis proteins, including two putative transmembrane helices, potential
methylation sites and the interaction domain with CheW proteins. Interestingly, ORF-2 encodes a protein having
homologies with CheW proteins.
Neelaredoxin is also shown to have significant superoxide dismutase activity (1200 U´mg±1), making it a novel
type of iron superoxide dismutase. Analysis of genomic data shows that neelaredoxin-like putative polypeptides are
present in strict anaerobic archaea, suggesting that this is a primordial superoxide dismutase.
The three proteins encoded in this operon may be involved in the oxygen-sensing mechanisms of this anaerobic
bacterium, indicating a possible transcriptional mechanism to sense and respond to potential stress agents.
The advent of an oxygen-rich atmosphere led to the appearanceof more energetically efficient organisms, but at the same timeintroduced a stressful environment for both aerobic andanaerobic organisms. In fact, oxygen reduction leads topotentially hazardous species, such as the superoxide anionradical, the hydroxyl radical and hydrogen peroxide. To detoxifythese products, living organisms developed an array ofprotective mechanisms, distributed to varying extents in thethree life domains. In particular, catalase and superoxidedismutase (SOD) have been found in both aerobic and anaerobicprokaryotes . Sulfate-reducing bacteria were for a long timeregarded as strict anaerobes , until it was shown thatDesulfovibrio spp., although not capable of growing in thepresence of oxygen, may not only survive in its presence buteven use it to produce ATP [3,4]. In D. gigas a complete redoxchain linking the oxidation of NADH to the reduction ofdioxygen to water was characterized [5,6], in which a smallmononuclear iron protein, rubredoxin , transfers electronsbetween an NADH oxidase  and an oxygen reductase [5,6].Moreover, catalase and SOD have been shown to exist in some
Desulfovibrio species . Taken together, these strategies showhow organisms adapted and responded to the paradox resultingfrom the introduction of oxygen in their environment.
Sulfate-reducing bacteria contain a large array of othermononuclear iron proteins, the function of which in most casesremains elusive. Recent reports have indirectly suggested thatone such protein, desulfoferrodoxin (Dfx) , may play a rolein defense mechanisms against oxygen radicals. In fact,overexpression of Desulfoarculus baarsii Dfx in an Escherichiacoli mutant deficient in cytoplasmic SOD led to the unexpectedresult of restoring the wild-type phenotype. However, theprotein was suggested to have little SOD activity [11,12].
Dfx is a small protein (14 kDa) present in several Desulfovi-brio species. It contains two mononuclear iron centers: center I,with a distorted tetrahedral cysteinyl sulfur co-ordination; andcenter II, having a square pyramidal geometry, with fourhistidine residues in the basal plane and a cysteine residue in theaxial position . This protein, both in terms of amino acidsequence and tridimensional structure, seems to be built of twodomains, each containing one of the iron centers. The firstdomain (residues 1±32) is very similar to desulforedoxin (Dx), asmall iron protein so far only isolated from D. gigas , whichhas one iron atom bound in a distorted tetrahedral co-ordinationto four cysteines . The second domain has homology withanother mononuclear iron protein isolated from D. gigas whichis blue and thus named neelaredoxin (Nlr). Nlr is a small protein(15 kDa) with two mononuclear iron sites. Its visible spectrum,consisting of a broad band at < 660 nm, is remarkably similarto that of center II of Dfx [16,17]. This observation, togetherwith the N-terminal sequence available, showing high homology
Correspondence to C. Rodrigues-Pousada, Instituto Gulbenkian de CieÃncia,
Note: The coding region of neelaredoxin has been deposited in GenBank
with the accession number AF096317.
(Received 20 July 1998; accepted 13 October 1998)
236 G. Silva et al. (Eur. J. Biochem. 259) q FEBS 1999
with that of the second domain of Dfx, led to the proposal thatNlr contained centers similar to that of Dfx center II [16,17].Although so far only isolated from D. gigas, genes coding forNlr-like proteins are present in the genomes of the archaeaArchaeoglobus fulgidus  and Methanoccocus jannaschii, albeit assigned as Dfx-like.
Besides the presence of detoxifying enzymes, micro-organ-isms also have chemosensory pathways to trigger physiologicalresponses towards the presence of oxygen. This is achievedeither by inducing motility or enzymic protective mechanisms,which enable the cells to adjust to their optimal oxygenenvironment. In bacteria such as Escherichia coli andRhodobacter sphaeroides, these mechanisms are quite wellunderstood. Signal transduction in these cases is often initiatedby transmembrane chemoreceptors, the methyl-accepting che-motaxis proteins (MCPs) . The chemosensory pathway inE. coli, namely the chemotaxis to oxygen and sugars, involvesseveral other proteins such as CheA, CheW and CheY,organized in a cascade of reactions that activates the flagelarmotor . The encoding genes for these proteins were found tobe organized as operons in E. coli , R. sphaeroides  andSinorhizobium meliloti .
In Desulfovibrio vulgaris Hildenborough, a multigene family,including dcrA-dcrL genes that encode putative MCP-likechemoreceptor proteins , was identified. The chemoreceptorDcrA, encoded by the dcrA gene, is a transmembrane proteinthat may function as a putative oxygen or redox sensor. The N-terminal sensor domain has a covalently bound c-type heme,which has been suggested to undergo a redox change in thepresence of oxygen, leading to conformational alterations thatmodulate the methylation process at the cytoplasmic signal-transduction domain . However, targeted gene-replacementmutagenesis of DcrA did not induce an aerotactic deficiency.
It was shown that expression of D. gigas Dx, equivalent to thefirst domain of Dfx, and of a truncated Desulfoarculus baarsiiDfx, lacking the second domain, cannot complement the SOD-deficient E. coli strain . This suggests that Dfx center II isresponsible for the protein SOD activity. Thus, the activity ofD. gigas Nlr, which contains a similar center, was investigated.The present work shows that Nlr is a novel type of Fe-SOD andis located in the same polycistronic unit as two predictedpolypeptides, homologous with chemosensory proteins.
Bacterial strains and plasmids
D. gigas (ATCC 19364) was grown as previously described. E. coli strains LE 392 and P2392 were used to screen thegenomic library and in the purification of the recombinantpositive phages. Competent cells of E. coli XL2-Blue Cells(Stratagene, La Jolla, Ca, USA) and E. coli TOP 10F 0
(Invitrogen, San Diego, Ca, USA), prepared according tostandard protocols , were used to transform the DNAfragments subcloned into the polylinker of the plasmidpZErOTM-1 (Invitrogen).
Preparation of genomic DNA
Genomic DNA from D. gigas was extracted as previouslydescribed . Plasmid DNA was prepared using the standardprotocols or the plasmid purification kit from Qiagen Inc.(Valencia, CA, USA).
Cloning and sequencing of the region containing the Nlr geneand two adjacent genes
A degenerate primer (5 0-ATGAARATGTAYGAYATGTTY-CARACNGCNGARTGGAAGGARAAR-3 0) corresponding tothe amino acid sequence of the N-terminus of Nlr was labeledwith [g-32P]dATP using the T4 phage polynucleotide kinase.This probe was used to screen a Sau3A1 genomic library fromD. gigas constructed in the vector l Dash II (Stratagene). Filterswere prehybridized and hybridized as described , using atemperature of 58±60 8C. Positive phages were purified andtheir DNA isolated as described previously . The DNA wasdigested with NotI, and an 11.5-kbp positive fragment wasidentified by Southern blotting. After appropriate restrictions,the NotI/NotI fragment was subcloned in pZErOTM-1 andsequenced using the Thermo Sequenase cycle sequencing kitaccording to the supplier's instructions (Amersham). Thisprocedure was used to solve the compressions, as the D. gigasgenome has a high G + C content . Sequential deletionswere made using the double-stranded nested deletion kit(Pharmacia). A search for protein sequence homology wasperformed using BLAST 2.0, and sequence alignments weremade using the GCG package and ClustalW . Sequenceretrieving from protein databases was achieved using NCBIEntrez protein sequence search.
Biochemical and spectroscopic procedures
Nlr was purified as previously described  and SDS/PAGEwas used to test the protein purity. The SOD assays wereperformed as described by McCord & Fridovich . Thissystem uses xanthine/xanthine oxidase as the superoxide-generating system, which reduces cytochrome c. One unit ofactivity is defined as the amount of enzyme required to obtain50% inhibition of the cytochrome c reduction. Sampleincubations with cyanide were performed overnight at 4 8C. Inthe assays with cyanide-incubated samples, cyanide wasincluded in the reaction medium (up to 10 mm) and appropriateblanks were included. Cyanide stock solutions were freshlymade and adequately buffered to avoid pH changes after
Fig. 1. Genomic organization of D. gigas Nlr transcriptional unit.
Partial restriction map of the right region of the subcloned 11.5-kb NotI/
NotI DNA fragment. The restriction sites for BamHI (B), EcoRI (E), PstI (P)
and NotI (N), and the coding units of Nlr and two other putative proteins
(ORF-1 and ORF-2) are shown. The vertical arrow indicates a putative rho-
independent terminator. The horizontal arrow indicates a potential promoter.
The grey arrows indicate the direction of transcription
q FEBS 1999 Novel superoxide dismutase in strict anaerobes (Eur. J. Biochem. 259) 237
Fig. 2. Nucleotide sequence of part of the subcloned 11.5-kb DNA fragment. The coding region of Nlr, as well as of ORF-1 and ORF-2, has been translated
to the single-letter amino acid code (deposited in GenBank with the accession number AF096317). Asterisks indicate translation stop codons. The ribosome-
binding sites (rbs) are shown. A putative rho-independent terminator (single underlining) and plausible promoters (± 35, ± 10) are indicated.
238 G. Silva et al. (Eur. J. Biochem. 259) q FEBS 1999
addition. Room temperature UV/visible spectra were recorded ina Beckman DU-70 spectrometer.
Location of Nlr gene in an oxygen-sensory operon
The sequence of part of the 11.5-kbp NotI/NotI DNA fragmentof D. gigas chromosomal DNA contains the Nlr-coding unit andtwo additional complete ORFs here designated as ORF-1 andORF-2 (Fig. 1). The two ORFs are transcribed in the samedirection as the Nlr gene. Nucleotide sequences, which possiblyact as ribosome-binding sites, were found in the upstream regionof each ORF (Fig. 2). The upstream region comprises a TATAsequence and other potential regulatory elements, which mightbe the promoter region of this operon. A predicted stem±loopstructure acting as a putative rho-independent transcriptiontermination signal is detected 23 bp downstream of the Nlr stopcodon (Fig. 2).
Analysis of Nlr sequence
D. gigas Nlr is 130 amino acid residues long and corresponds toa protein of molecular mass 14 634 Da. (While this work was inprogress, the Nlr sequence was deposited in the EMBL/GenBank/DDBJ databases (accession number AF034965) byN. V. Shenvi and D. M. Kurtz, Jr. This sequence, however,differs by two nucleotides in the intergenic region from thesequence reported here.) Nlr is homologous to the seconddomain of Dfxs (Fig. 3) with amino acid identities in this regionof 20±25% when compared with all Dfxs present in databanks(not shown). Most importantly, not only the ligands of theFeN4S center of D. desulfuricans Dfx  (His17, His45,His51, His118 and Cys116, in Nlr numbering) but also theresidues surrounding the binding sites are strictly conservedbetween Dfxs and Nlr (Fig. 3). Two putative archaeal proteins,namely ORF AF0344 from A. fulgidus  and M. jannaschiiORF MJ0741 , were considered Dfx-like. However, owingto their amino acid identities with Nlr (28% identity with ORFAF0344 and 51% with ORF MJ0741), the absence of the Dx-like domain and because the blue center putative binding sitesare conserved (Fig. 3), they should be regarded as Nlr-like.
A similar geometry for one of the Nlr iron sites and Dfxcenter II is very likely, as previously suggested on the basis of
the spectroscopic data. However, the Nlr sequence is notcompatible with the presence of two identical iron sites, as itdoes not contain two identical binding motifs. Nevertheless, thespectroscopic data on Nlr  suggest that it contains twoanalogous iron centers, both contributing to the 660 nm band,which in Dfx were assigned to a sulfur to iron charge transferband . Besides the residues needed for binding one of theiron atoms, three other histidines and four cysteines are present,some of which could be the ligands of the other iron site. Asexpected, the first domain of Dfx, containing the FeS4 centerhomologous with that of Dx, is not present in Nlr (Fig. 3).
MCP- and CheW-like sequences
The predicted polypeptide of 649 amino acids encoded by ORF-1 shows an overall amino acid identity to the MCPs of severalorganisms that ranges from 26% to 33% and appears to belimited to the C-terminal domain. It shows homology with theMCP transducer Tse (taxis to serine) from Enterobacteraerogenes (33% identity and 55% similarity), MCPII fromSalmonella typhimurium (33% identity and 53% similarity) andthe chemoreceptor proteins DcrH (27% identity and 52%similarity) and DcrA (26% identity and 50% similarity) fromD. vulgaris. It also shows similar homologies to E. coli MCPIII,Borrelia burgdorferi MCP-4 and Caulobacter crescentusMCPA. The alignment of the predicted polypeptide, namedMCP-like, with its counterparts reveals the conserved structuraland functional domains of the methyl-accepting chemotaxistransducers .
The hydropatic profile shows the presence of two hydro-phobic regions in D. gigas MCP-like protein, with rather lowsequence identity among the predicted polypeptides compared.These regions comprise the residues in positions 10±33 and195±218, which are equivalent to those predicted for the twotransmembrane domains of E. coli chemoreceptors . Both ofthese two hydrophobic regions (each 23 residues in length) arelong enough to span the lipid bilayer  and, as predicted bythe program predator , correspond mainly to a helices.Between these two regions there are about 160 amino acids, alength similar to that of the described periplasmic receptordomain . The low identity found in these three regions isconsistent with the fact that they include the periplasmicreceptor region for different chemoattractants or repellents, aswell as the two transmembrane-spanning domains.
Fig. 3. Sequence alignment of Nlr with homologous proteins. Dg Nlr, D. gigas Nlr; Af Nlr, A. fulgidus Nlr-like; Mj Nlr, M. jannaschii Nlr-like; Dd Dfx,
D. desulfuricans Dfx; Dg Dx, D. gigas Dx. Black boxes indicate the residues involved in the binding of Dfx center II. In the Nlrs shown, these residues are
q FEBS 1999 Novel superoxide dismutase in strict anaerobes (Eur. J. Biochem. 259) 239
240 G. Silva et al. (Eur. J. Biochem. 259) q FEBS 1999
As in most of the MCPs known , D. gigas MCP-likeprotein has an overall positive charge at its N-terminus. Twoputative methylation regions with several potential sites ofmethylation, based on the comparison with the E. colitransducer consensus methylation sequence A/S-X-X-E-E-X-A/S/T-A-S/T/A [38±40], are found in D. gigas MCP-likeprotein from residue 371 to 393 and from 545 to 570 (Fig. 4).The methylation sites could be glutamate or glutamine residues,as suggested for Bacillus subtilis MCP .
The C-terminal portion of D. gigas MCP-like protein alsoincludes the consensus sequence referred to as the highlyconserved domain, which is 46 residues long (residues 440±
486) (Fig. 4). In E. coli, it has been proposed on the basis ofgenetic evidence that this domain of MCPs interacts with thecytoplasmic CheW protein, to transduce intracellularly thesignal received by the receptor region . The importance ofthis domain is supported by its high conservation amongtransducers from several organisms [37,43].
The predicted polypeptide of 168 amino acids encoded byORF-2 shows homology with the chemotaxis protein CheWfrom several organisms. Amino acid similarities to and identitieswith CheWs from R. sphaeroides, Rhizobium meliloti, S. typhi-murium, E. coli and Ent. aerogenes ranges from 63 to 56% and33 to 28%, respectively. Alignment with the homologous
Fig. 6. Effect of cyanide on the visible band
of Nlr. Trace a, native Nlr; trace b, Nlr after
overnight incubation with 10 mm KCN, at 4 8C.
D. gigas Nlr was in 50 mm Tris/HCl buffer,
Fig. 5. Sequence alignment of CheW-like protein with homologous proteins. Dg CheW, D. gigas CheW-like putative protein; Rs CheWII and Rs CheW,
R. sphaeroides CheWII and CheW proteins; Rm CheW, Rhizobium meliloti CheW; St CheW, S. typhimurium purine-binding CheW; Ec CheW, E. coli purine-
binding chemotaxis protein CheW; Ea CheW, Ent. aerogenes purine-binding chemotaxis protein CheW. The amino acid residues involved in the interaction with
q FEBS 1999 Novel superoxide dismutase in strict anaerobes (Eur. J. Biochem. 259) 241
polypeptides (Fig. 5) revealed several conserved clusters which,in E. coli, were shown to represent the contact sites betweenCheW and the Tsr (taxis to serine and away from somerepellents) MCP transducer .
Nlr SOD activity
Purified Nlr has SOD activity, with a specific activity of1200 U´mg±1. This value is well within the range of thosedetermined for some Fe-SODs, isolated from bacterial [9,44,45]and even eukaryotic sources . This activity was shown to bepH independent between pH 5.5 and pH 9.0. Regarding theeffect of known SOD inhibitors and as observed for classical Fe-SODs , Nlr SOD activity is inhibited by azide (< 40%inhibition in the presence of 10 mm sodium azide). Cyanidedoes not inhibit Nlr SOD activity, but in fact increases it(fivefold to sixfold with 10 mm cyanide). By visible spectro-scopy it was found that cyanide binds to the iron centers, as a redshift of the visible band from 650 nm to 690 nm is observed onincubation with cyanide (Fig. 6). The extremely small amountof protein obtained from D. gigas cells  did not allow adetailed spectroscopic and kinetic analysis of cyanide binding.
Even though so far there is no structural information on Nlr,indirect information can be obtained by analysing the availableX-ray structure of Dfx . Comparing the geometry of Dfxcenter II with that of Fe-SODs (Fig. 7)  reveals someinteresting similarities: both iron sites are pentaco-ordinated, aremainly bound to histidine residues, and have a vacant co-ordination site, which can be used for substrate binding. Apartfrom this general geometric arrangement, no significant aminoacid sequence similarities are found among Nlr and the otherclassical Fe-SODs. A remarkable coincidence of reductionpotentials is observed between Nlr and Fe-SOD. This feature isparticularly relevant, as the most widely discussed reactionscheme for SOD catalysis involves alternate reduction andoxidation of the metal . Barrette et al.  have determinedthat at pH 7.0 the average value for reduction potentials of theFe-SOD from E. coli, Azotobacter vinelandii and Pseudomonasovalis is 260 mV, while a reduction potential of 190 mV atpH 7.5 was determined for both iron centers of Nlr .
As already mentioned, the archaea M. jannaschii andA. fulgidus contain genes encoding proteins homologous withD. gigas Nlr. On the other hand, analysis of the two genomes[18,19] reveals that there are no ORFs encoding classical SODs.However, SOD activities of 4±6 U´mg±1 and 50 U´mg±1 werefound in the soluble extracts of M. jannaschii and A. fulgidus,respectively (C. M. Gomes, H. Huber, K. O. Stetter and M.Teixeira, unpublished data). These figures are comparable with,or even higher than, those obtained in similar conditions forPropionibacterium shermanii (3.8 U´mg±1) , Streptococcus
mutans (0.8 U´mg±1)  and Methanobacterium thermoauto-trophicum (17.5 U´mg±1) . The SOD activity detected inM. jannaschii and A. fulgidus is probably due to the Nlr-likeprotein present in these micro-organisms.
With the exception of D. gigas rubredoxin for which a functionwas recently clarified , sulfate-reducing bacteria containseveral iron mononuclear proteins of unknown function. Thepresent data establish a function for another of these proteins,Nlr, as a novel type of Fe-SOD present in the bacteriumD. gigas and most probably in the archaea M. jannaschii andA. fulgidus. This suggests that Nlr is a primordial SOD. Ancientorganisms like these archaea may have developed mechanismsto deal with oxygen radicals produced in the primitiveatmosphere, either through the Urey effect  or as anadaptation to the emerging oxygen-rich atmosphere. Theseprotective mechanisms may have also been acquired by otheranaerobes, some of which (like D. gigas) retained these ancientforms of protective enzymes, or replaced them with alternativeproteins, like other Desulfovibrios, which apparently lack Nlrbut have other classical SODs . Moreover, it may beevolutionarily significant to find Nlrs both in the sulfatereducers D. gigas and A. fulgidus, and in the methanogenM. jannaschii: sulfate reducers and methanogens are known tooccur in symbiotic cultures, in which the carbon dioxide andhydrogen produced by the first are used by the second (reviewedin ). However, it should be stressed that the available proteinand genomic data for micro-organisms are too scarce to allowany sound evolutionary derivations.
It has been suggested that the gene coding for D. vulgaris Dfxarose from a gene fusion event involving the D. gigas Dx geneand the Nlr gene [16,55]. However, this hypothesis does notaccount for the presence of an additional iron center in Nlr or itsspecific function as reported in this paper, as Dfx does not showany significant SOD activity . Clearly, the evolutionaryrelationships among these proteins may indeed be morecomplex. In fact, besides having an ORF encoding a Dfx,A. fulgidus  also has an Nlr-like gene. However, there arenot enough cysteine or histidine residues available in itssequence for binding of a second iron center. In contrast, inM. jannaschii Nlr-like protein, classified as being Dfx-like ,these additional potential ligands are present (Fig. 3).
In spite of being an anaerobe, D. gigas has been shown tohave an oxygen-utilizing pathway [4,5]. The present resultsshowing that Nlr is a novel type of SOD, located in a putativeoxygen-sensory operon, contribute to our understanding of themechanisms by which anaerobes respond to oxygen. The codingunits for Nlr, MCP- and CheW-like proteins, are all grouped inthe same operon and are probably transcribed under the controlof the upstream promoter to express a polycistronic mRNA.Similarly, in Caulobacter crescentus [43,56], mcp and che genesare also located in a functionally related operon. In Rhizobiummeliloti  several che genes have been identified in the sameoperon, which further contains an ORF encoding a polypeptidereported to be MCP-like. However, it does not contain eithermethyl-accepting sites or hydrophobic transmembrane domains,and is therefore probably involved in a different signal-transduction pathway. It is noteworthy that the number offunctionally related genes within the same operon such as thoseinvolved in metabolic pathways is increasing, although thisfeature is not universal in either archaea or bacteria .
The cyanide effect on the catalytic activity was unexpected.The data do not allow us to establish if a change in co-ordination
Fig. 7. Comparison of the structure of the iron centers of Fe-SOD and
Dfx center II. (A) D. desulfuricans Dfx center II ; (B) Mycobacterium
242 G. Silva et al. (Eur. J. Biochem. 259) q FEBS 1999
geometry or a change in the number of ligands occurs. The factthat in Nlr the cyanide-bound form shows absorption in the near-infrared (< 700 nm) suggests that cysteine remains bound to theiron center(s) [16,17]. It is interesting to note that cyanide hasbeen proposed as one of the ligands bound to iron in the activesite of [NiFe] hydrogenases  and that D. gigas synthesizeslarge amounts of such an enzyme .
The results reported here suggest that in D. gigas, the proteinsof this putative operon are involved in an oxygen-sensorypathway. Oxygen or a toxic oxygen radical could activate thischemotaxis-like system and concomitantly induce the expres-sion of Nlr, a SOD. Similarly, in Helicobacter pylori, which is amicroaerophilic bacterium, a SOD gene is localized in anoxygen-sensing operon: analysis of its genome  shows thatthe ORFs encoding putative Fe-SOD, CheW, CheA and CheVproteins are located in adjacent positions and transcribed in thesame direction. This gene arrangement, similar to that observedin D. gigas, may reflect a transcriptional mechanism that isfinely tuned to respond rapidly to a potential stress agent.Whether or not this operon is involved in the anaerotacticresponse remains to be elucidaded. However, in the case ofD. vulgaris, it was shown  that oxygen is recognized as achemotactic repellent.
This work was supported by PRAXIS XXI (20 and 1075 to M.T.; 61/96 to
J.L.G. and 32/96 to C.R.P., BD9016/96 to G.S. and BD9793/96 to C.M.G.),
European Community Grant Bio4-CT96-0488 to M.T. and NIH grant
GM56001-01 to J.L.G. and M.Y.L. C.M.G. acknowledges the Programa
Gulbenkian de Doutoramento em Biologia e Medicina. We are grateful to A.
Coelho for providing D. desulfuricans Dfx co-ordinates.
1. Touati, D. (1997) Oxidative Stress and the Molecular Biology of
Antioxidant Defenses, Cold Spring Harbor Laboratory Press, Cold
Spring Harbor, NY.
2. LeGall, J. & Xavier, A.V. (1996) Anaerobes response to oxygen: The
sulfate-reducing bacteria. Anaerobe 2, 1±9.
3. Diling, W. & Cypionka, H. (1990) Aerobic respiration in sulfate-
reducing bacteria. FEMS Microbiol. Lett. 71, 123±128.