Cytochrome c 551 from Starkeya novella CHARACTERIZATION , SPECTROSCOPIC PROPERTIES , AND PHYLOGENY OF A DIHEME PROTEIN OF THE SoxAX FAMILY *

Cytochromes from the SoxAX family have a major role in thiosulfate oxidation via the thiosulfate-oxidizing multi-enzyme system (TOMES). Previously characterized SoxAX proteins from Rhodovulum sulfidophilum and Paracoccus pantotrophus contain three heme c groups, two of which are located on the SoxA subunit. In contrast, the SoxAX protein purified from Starkeya novella was found to contain only two heme groups. Mass spectrometry showed that a disulfide bond replaced the second heme group found in the diheme SoxA subunits. Apparent molecular masses of 27,229 10.3 Da and 20,258.6 1 Da were determined for SoxA and SoxX with an overall mass of 49.7 kDa, indicating a heterodimeric structure. Optical redox potentiometry found that the two heme cofactors are reduced at similar potentials (versus NHE) that are as follows: 133 mV (pH 6.0); 104 mV (pH 7.0); 49 (pH 7.9) and 10 mV (pH 8.7). EPR spectroscopy revealed that both ferric heme groups are in the low spin state, and the spectra were consistent with one heme having a His/Cys axial ligation and the other having a His/Met axial ligation. The His/Cys ligated heme is present in different conformational states and gives rise to three distinct signals. Amino acid sequencing was used to unambiguously assign the protein to the encoding genes, soxAX, which are part of a complete sox gene cluster found in S. novella. Phylogenetic analysis of soxAand soxX-related gene sequences indicates a parallel development of SoxA and SoxX, with the diheme and monoheme SoxA sequences located on clearly separated branches of a phylogenetic tree.

Reduced sulfur compounds can be used as electron donors by a variety of photo-and chemolithotrophic microorganisms supporting autotrophic growth. Thiosulfate is one of the most abundant forms of reduced sulfur in nature, and hence the ability to oxidize this compound is found in many microorganisms belonging to a number of different phyla (1). Several pathways for thiosulfate oxidation have been proposed over the years, but in many cases investigations have been hampered by the reactivity of intermediates and the low growth yields common to chemolithotrophic microorganisms growing on sulfur. The most common (1) and also best studied pathway of thiosulfate oxidation involves the so-called thiosulfate-oxidizing multi-enzyme system (TOMES), 1 which can oxidize thiosulfate, sulfite, or sulfide to sulfate with cytochrome c as electron acceptor (2)(3)(4). The complex has been studied mainly in Paracoccus versutus (2), Paracoccus pantotrophus (4 -7), and Rhodovulum sulfidophilum (8) and is encoded by "sulfur oxidation" (sox) gene clusters. The TOMES from P. pantotrophus contains nonheme metalloproteins (SoxB, SoxC) as well as a number of cytochromes (SoxA, SoxX, SoxD, SoxE) and proteins without redox cofactors (SoxY, SoxZ) (2,3,5). Of these proteins, SoxAX, SoxB, and SoxYZ have been shown to be essential for complex function, whereas the role played by other components (SoxCD, SoxF, SoxE) has not yet been fully understood (3,5,9,10). The proposed reaction cycle of the complex involves a sequential oxidation of thiosulfate, which is bound covalently to the carrier protein SoxYZ. Sulfate and/or sulfite is thought to be liberated in the process (3) although no free reaction intermediates have been detected so far.
The SoxAX protein is the best characterized of the essential components of the TOMES. It is a heterodimeric protein composed of two heme c-bearing subunits with molecular masses of about 30 and 15 kDa, respectively. It has been studied mainly in R. sulfidophilum, in which the SoxA subunit contains two heme c groups with an unusual heme ligation (His/Cys and His/Cys-persulfide), whereas the SoxX subunit contains only one heme c group with a His/Met ligation (8,11). The crystal structure of SoxAX from R. sulfidophilum has recently been solved and showed a novel type of domain packing for the SoxA subunit, which might have arisen from a gene duplication event (12). Only one of the two SoxA heme groups is thought to participate in the reaction cycle, and this heme showed a modification of the ligating cysteine to a persulfide. The SoxAX protein from P. pantotrophus has been expressed in Escherichia coli and been shown to be functionally active in the recombinant form (13). As isolated, this form apparently did not contain a modified cysteine ligand, raising the question as to whether this modification is an artifact of an incomplete reaction cycle or is necessary for protein function. A SoxArelated protein that was reported to be a SoxA homodimer lacking a SoxX subunit has been purified from Chlorobium limicola (14).
Genes related to soxAX have been reported from a variety of * The costs of publication of this article were defrayed in part by the payment of page charges. This article must therefore be hereby marked "advertisement" in accordance with 18 U.S.C. Section 1734 solely to indicate this fact. microorganisms such as Aquifex aeolicus, Chlorobium tepidum, Chl. limicola, Ralstonia species, Bradyrhizobium japonicum, and others, usually as a result of genome sequencing projects (8,(15)(16)(17). They have uncovered considerable diversity within the SoxAX protein family. In particular, in addition to the well characterized triheme form of SoxAX found in R. sulfidophilum and P. pantotrophus, analysis of the gene sequences indicates that a number of diheme SoxAX proteins may exist in which the SoxA homologues possess only one heme group (8). Moreover, the SoxX homologues found in these proteins are usually much larger (25-75%) than their counterparts found in the triheme forms and exhibit low similarity values (about 36%) toward their characterized relatives. Here, we report for the first time the purification and biochemical characterization of a diheme member of this cytochrome family, the SoxAX protein from Starkeya novella. Structural aspects of the protein along with redox and spectroscopic properties of the heme groups present are reported herein. The diversity present within the SoxAX family is highlighted by analysis of the phylogenetic relationships between the different existing forms.

EXPERIMENTAL PROCEDURES
Materials and Reagents-All chemicals were purchased in analytical or corresponding grade unless otherwise stated. Polyvinylidene difluoride membranes were from Schleicher & Schuell.
Bacterial Strains, Media, and Plasmids-Bacterial strains and plasmids used in this study are specified in Table I. S. novella was cultivated on DSMZ medium 69 as described (18). E. coli strains were routinely cultivated on liquid or solidified Luria-Bertani medium; where necessary the medium was supplemented with antibiotics (19).
Molecular Biological and Genetic Techniques-Standard techniques were used throughout (19,20). DNA sequencing (21) was carried out using the Big Dye terminator chemistry (v. 2.0, Applied Biosystems). Probes for Southern blots were labeled nonradioactively by PCR incorporation using digoxigenen-dUTP (Roche Applied Science), and detection of signals was performed according to the manufacturer's instructions (Roche Applied Sciences). Plasmid purification kits were obtained from SIGMA and PCR purification and gel extraction kits from Qiagen.
Cloning of the S. novella sox Gene Region-The insert of the previously reported pTNSOX2 plasmid (22) was completely sequenced. Gene probes targeting the soxF (Soxdf5, GAA CGC CAG CTC GCG ACC; Soxdr1, GCA CAT GGC TGT CGC GAG) and soxY (Soxbf1, TCT CCT TCT CGC TCT CAT, Soxbr5, TCC GCC ATC GAC TTG TCG) genes were amplified by PCR from pTNSOX2 (22) and PCR-labeled using digoxigenen-dUTP. These probes were used in Southern blot experiments with genomic DNA of S. novella. A 2.5-kb PstI fragment and a 3.5-kb EcoRI fragment hybridized with the downstream (soxF) and upstream (soxY) gene probes for the S. novella sox gene cluster, respectively. Partial genomic libraries of PstI and EcoRI fragments in the appropriate size range were constructed in pBluescript and screened by Southern blotting. Two plasmids, designated pSNSOX3 and pSNSOX4, were identified as positives and their inserts sequenced on both strands (Table I, Fig. 1). A further round of screening using a gene probe targeting the soxX (M13rev primer, Soxyr3A TCA AGC GTT GAG GGA GGC; template, pSNSOX4) gene led to the identification of a 3.5-kb SstII fragment hybridizing with that gene probe. A plasmid, pSNSOX5, containing this insert was isolated from a partial clone library as described above and sequenced.
Purification of the Diheme SoxAX Cytochrome from S. novella-40 g of frozen S. novella cell paste were resuspended in 5 volumes of 10 mM Tris-HCl, pH 7.8, and broken by two passages through a French pressure cell (Aminco) at 1000 p.s.i. Preparation of a cell-free extract and ammonium sulfate precipitation were carried out as described (18)  Spectroscopic and Analytical Techniques-Optical spectra of the SoxAX protein were recorded using a Hitachi U-3000 double-beam spectrophotometer. Analytical ultracentrifugation was performed on samples with an A 280 of 0.4 at 13,000 rpm for 20 h at 20°C using a Beckman XL-1 Ultracentrifuge. Heme type and content were determined in alkaline pyridine solution (23). Denaturing and nondenaturing PAGE was performed according to Laemmli (24). Gels were routinely stained using Coomassie Brilliant Blue; staining of polyacrylamide gels for heme-dependent peroxidase activity was carried out according to Thomas et al. (25). N-terminal sequencing was performed by sequential Edman degradation of protein samples blotted onto a polyvinylidene difluoride membrane subsequent to denaturing PAGE. N-terminal protein sequencing was performed at the University of Queensland Protein Analysis Laboratory.
Mass Spectrometry-Protein samples for tryptic digestion were exchanged into 20 mM ammonium bicarbonate, pH 7.9. Porcine trypsin (sequencing grade) was from Promega. Tryptic digests were carried out for 20 h at 37°C with an enzyme to protein ratio of 1:20 according to the manufacturer's instructions. For mass analysis following SDS-PAGE, protein bands were cut out of Coomassie Brilliant Blue R250-stained gels and destained by repeated washing in 50% acetonitrile, 25 mM ammonium bicarbonate, pH 7.9 (10 min at 37°C with agitation per wash step) until the gel slice appeared clear. The destained gel slices were carefully dried under a gentle nitrogen flow. Tryptic digests were carried out in 10 l of 20 mM ammonium bicarbonate buffer, pH 7.9, using 200 ng of porcine trypsin/sample (37°C, 22 h) followed by enzyme inactivation at Ϫ20°C. Prior to mass analysis, 20 l of 50% acetonitrile, 1% trifluoroacetic acid was added, and the samples were sonicated in a water bath for 20 min. In some cases the destained gel slices were treated with dithiothreitol followed by alkylation of reduced SH groups with iodoacteamide as described previously (26).
Samples were analyzed on an Applied Biosystems Voyager DE STR 4316 MALDI-TOF mass spectrometer (Voyager TM 5.1 software with data explorer). Sinnapinic acid and ␣-cyano-4-hydroxycinnamic acid were used as matrices in sample preparations of whole proteins and peptides, respectively. Matrices were mixed with samples in a 1:1 ratio. Applied Biosystems Sequazyme TM calibration mixtures 2 and 3 were used in the analyses. Sequencing of tryptic peptides was carried out Amp r , pBluescript II derivate, 3.5-kb EcoRI fragment, partial soxXA soxYZ partial soxB This work pSNSOX5 Amp r , pBluescript II derivate, 3.5-kb SacII fragment, soxYZ soxAX soxW partial soxV This work using an Applied Biosystems Qstar/Pulsar I instrument equipped with a nanospray source (Protana), LINAC TM Pulsar collision cell, and a TOF mass detector. Applied Biosystems AnalystQS software was used in data acquisition and BioAnalyst (Applied Biosystems) for data analysis.
Thiosulfate:Cytochrome c Oxidoreductase Assays-Reaction mixtures contained 0.05 mM cytochrome c (horse heart, Sigma), 2 mM thiosulfate, and variable amounts of purified SoxAX in 20 mM Tris-HCl, pH 8.0. The assay volume was 1 ml, and all measurements were carried out at 25°C in a Hitachi UV-3000 double-beam spectrophotometer.
Redox Potentiometry-Heme redox potentials were determined by standard redox potentiometric methods (27) using a glove box and a N 2 atmosphere (O 2 concentration Ͻ2 ppm). The redox mediators used were tetramethyl-1,4-phenylenediamine (2 M), phenazinemethosulfate (2 M), 2-hydroxy-1,4-naphthoquinone (8 M), benzylviologen (2 M), and methylviologen (2 M); the reductant was Na 2 S 2 O 4 , and the oxidant was 6 ]. 20 mM phosphate buffer was used in titration experiments. The solution pH was adjusted with acetic acid before each titration, and the solution potential was measured upon equilibration with an ABB Kent Taylor combination Pt-Ag/AgCl electrode attached to a Hanna 8417 meter. Spectral differences between the oxidized and reduced hemes were monitored at 416 nm with a Shimadzu UV Mini 1240 spectrophotometer. The experimental data were fitted to the Nernst equation.
Electron Paramagnetic Resonance (EPR) Spectroscopy-X-band (ϳ9 -10 GHz) continuous wave EPR spectra were recorded on a Bruker Biospin Elexsys E500 EPR spectrometer fitted with a super-high Q cavity. Calibration of the magnetic field and microwave frequency was achieved with a Bruker ER 035 M gaussmeter and an EIP 548B microwave frequency counter, respectively. Liquid helium temperatures (1.5-50 K) were obtained with an Oxford ESR910 flow-through cryostat in conjunction with an Oxford Instruments ITC-4 temperature controller. Spectrometer tuning, signal averaging, and subsequent spectral comparisons and plotting were performed with Bruker's Xepr (version 2.1) software. SoxAX protein samples for EPR analysis contained 50 M protein.
Computer simulation of the EPR spectra was performed using version 1.1.3 of XSophe-Sophe-XeprView computer simulation software suite running on a personal computer with Mandrake Linux version 9.1 (28,29). The computational program Sophe employs a number of methods, including matrix diagonalization, SOPHE interpolation, and homotopy for the analysis of randomly oriented EPR spectra. In this research we employed matrix diagonalization in conjunction with mosaic misorientation to simulate the randomly oriented EPR spectra from the hemes. This method 2 significantly reduces the computational times. Comparisons of simulated and experimental spectra and data manipulation were performed with Xepr. Assignment of the g values to g x , g y , and g z was performed using the computer program ls_iron, which was developed using the theory described by Bohan (30).
Phylogenetic Analysis-Genes encoding proteins related to SoxAX were identified using the BLAST algorithm (31). Deduced amino acid sequences were aligned using ClustalX (32), and regions of low homology were removed from the alignment. This concerned mainly N-terminal regions and inserted amino acids that were found in only one of the sequences under consideration. The alignments constructed for SoxAand SoxX-related sequences spanned 274 and 190 residues, respectively. The data were then analyzed with the Phylip package (33). Different distance methods (neighbor-joining, fitch, kitsch) and a maximum parsimony analysis were applied to the data sets. Robustness of the resulting trees was verified by bootstrap and jackknife analysis with 1000 resampled data sets. 1000 randomly permutated data sets were generated to check the significance of the observed protein relationship.

Purification and Properties of the SoxAX Protein-Earlier
experiments had shown that a heterodimeric cytochrome c 551 with multiple heme groups is expressed to high levels in S. novella during chemolithoautotrophic growth (18). The soluble cytochrome c 551 was purified to homogeneity (Fig. 1C) as described under "Experimental Procedures" and designated SoxAX after its identity had been established by peptide sequencing (see below). Electronic absorption spectra of SoxAX confirmed that the protein is a cytochrome c 551 with peaks at 551 (␣), 522 (␤), and 416 nm (␥) for the reduced protein (oxidized: 410 (␥) and 362 nm (␦)) (Fig. 1B). The apparent molecular masses of the subunits were 24 This seems to indicate that the protein may be subject to some form of fragmentation in the laser beam and the possibility that other forms or modifications may exist. Additional mass peaks have also been observed for the SoxAX protein from P. pantotrophus (13).
The apparent molecular mass of the S. novella SoxAX holoprotein was determined to be 49.7 kDa by analytical ultracentrifugation, suggesting a heterodimeric composition (␣␤) of the protein. This value is close to the sum of the molecular masses determined by MALDI-TOF mass spectrometry (47,487.91 Da). Alkaline hemochrome spectra showed a single ␣ band at 550 nm, indicating the exclusive presence of c-type hemes. Both subunits could be stained for heme-linked peroxidase activity after SDS-PAGE (Fig. 1C), confirming the presence of covalently bound hemes in both polypeptides. However, an overall heme content of 0.7/subunit, determined according to Berry and Trumpower (23), indicated that, in contrast to the previously characterized SoxAX proteins from P. pantotrophus and R. sulfidophilum, SoxAX from S. novella is a diheme protein.
The purified hemoprotein could not be reduced by the addition of thiosulfate or sulfite to the samples. It also did not exhibit any thiosulfate:cytochrome c oxidoreductase activity (data not shown), which is consistent with the properties of purified SoxAX from R. sulfidophilum and P. pantotrophus. In the absence of other purified TOMES components such as SoxYZ, SoxB, and SoxCD, further activity analyses could not be undertaken.
soxAX Genes from S. novella-The cytochrome c 551 is encoded by the soxAX genes, which are part of the sox gene cluster found in S. novella. Sequencing of plasmids pTNSOX2 and pSNSOX3, -4, and -5 revealed the presence of a complete sox gene cluster in S. novella. It contains at least two transcriptional units, the soxAX genes and the soxYZBCD-orf1-soxF-orf2 genes, which are transcribed divergently (Fig. 1A) and encompass a putative promoter region of 227 bp. This arrangement of genes resembles that found in R. palustris. It differs from the most frequently found sequence of these genes, soxXYZA (8), which has been taken as being indicative of the intimately associated functions of the SoxAX and SoxYZ proteins; SoxAX is thought to have a role in attaching thiosulfate to the carrier protein SoxYZ (3).
The soxA gene from S. novella encodes a protein of 32.41 kDa (unprocessed, 289 residues, pI 8.15), with only the C-terminal heme binding site still intact and only one cysteine remaining of the N-terminally located heme-binding site found in P. pantotrophus and R. sulfidophilum. It appears likely that the heme found in SoxA from S. novella has a His/Cys ligation similar to the one described for the SoxA protein from R. sulfidophilum (11), as the Cys ligand residue in question (Cys-199, S. novellaprocessed SoxA numbering) is conserved. It has been suggested that the conserved cysteine residues at positions 73 (processed SoxA protein, Cys ligand to second heme) and 37 (former CXXCH motif) might form a disulfide bond, as observed in the related cytochrome c 551 from C. limicola (14,17), which, however, appears to be a SoxA homodimer. The S. novella soxA gene encodes a signal peptide with a predicted cleavage site between residues 39 and 40 (ARA2AE) that likely targets the protein for Sec-dependent export (40). However, the N-terminal sequence of the SoxA gene also contains a double arginine motif, which is located just downstream of an alternative initiation codon (position 15), but does not seem to contain any of the other conserved residues found in the twin arginine leader consensus sequence (41,42). Together with the strong hydrophobicity of the sequence segment following the double arginine, which is a typical feature of Sec export signals (40), these observations suggest that in this case, the occurrence of the motif is not indicative of the export mode. In general, twinarginine signal peptides that target the proteins to the Tat protein export system are not involved in directing cytochromes to the periplasm (43).
The SoxX protein encoded in the sox gene cluster has a predicted molecular mass of 22.3 kDa (208 residues, pI 6.74) before processing, which is about 5 kDa more than the molecular masses found in its counterparts from R. sulfidophilum and P. pantotrophus. Homology to these proteins is greatest around the heme-binding site and toward the C terminus. The S. novella SoxX protein shows only 36% similarity (26% identity) to the SoxX protein from R. sulfidophilum. Sequence anal-ysis suggests that the SoxX heme group has a His/Met coordination, as the Met residue that provides the ligand to the heme group of the R. sulfidophilum SoxX protein is conserved. The predicted molecular mass for the processed SoxX protein is 19.653 kDa (183 residues, pI 5.52), with the predicted cleavage site for a Sec transport signal peptide located between residues 25 and 26 (AHA2QE).
Structural Aspects of the SoxAX Protein-Because a number of microorganisms contain multiple copies of some sox genes, the identity of the purified SoxAX protein was confirmed by N-terminal sequencing and by sequencing of tryptic peptides using a Qstar/Pulsar I mass spectrometer. The sequencing of tryptic peptides corresponding to amino acids 19 -27 (m/z 1072.5, GEVLWSEPR) and 63-69 (m/z 890.4, VMDLEQR) in SoxA and amino acids 23-30 (m/z 973.4737, LPEGWESR) in SoxX resulted in the expected sequences, confirming that the purified cytochrome c 551 protein is indeed encoded by the S. novella soxAX genes. The N terminus of SoxX was not accessible to Edman degradation, for SoxA the sequence XQMIEDP was determined, which differs from the N-terminal sequence predicted from the gene sequence. However, the molecular mass of the SoxA protein assuming RQMIEDP (signal peptide cleavage site between amino acids 51-52) as the N terminus would be 26591 Da, which corresponds well to the value determined by mass spectrometry (27, native SoxAX and SoxA isolated after SDS-PAGE, we were able to show that the SoxA protein contains a disulfide bond formed by Cys-37 and Cys-73. These residues are located on tryptic fragments with predicted masses of 1518.7 and 1349.7 Da, respectively. Both of these fragments are absent from mass spectra generated from digested SoxAX protein; instead, a large peak of 2866.7 Da, corresponding to the combined masses of these fragments minus 2 Da, is present in all spectra ( Fig.  2A). In similar spectra generated from tryptic digests of SoxA isolated after SDS-PAGE, the 2866.7-Da peak is strongly diminished, and a peak at 1349.5 Da is observed first (Fig. 2B). In mass spectra obtained using reduced and iodoacetamide-treated tryptic digests, the 2866.7-Da peak was completely lost, and peaks corresponding to the single fragment masses (ϩ carbamidomethyl modification, 57 Da) could be observed (data not shown).
Analysis of SoxAX by EPR Spectroscopy-X-band EPR spectra of two independently prepared samples of the diheme S. novella SoxAX (Fig. 3, a and b) reveal the presence of three sets of low spin type II heme resonances, which were present in different proportions (compare Fig. 3, a and b). An adventitious, axially symmetric Cu(II) signal was also observed and could not be removed by treatment of the sample and EPR tubes with EDTA. Computer simulation (28,29) of the observed low spin Fe(III) species with the spin Hamiltonian parameters given in Table II and an orthorhombic electron Zeeman spin Hamiltonian produced the EPR spectra shown in Fig. 3, c-e.
The three low spin Fe(III) species have g matrices similar to those found by Cheesman et al. (11) (Table II) for the R. sulfidophilum triheme SoxAX and have been labeled LS1a, LS1b, and LS2 in accord with their nomenclature. LS1a and LS1b represent slight variations of one form of heme. The g matrices obtained from computer simulation studies were used to determine the assignment of the principal components of g (g x , g y , and g z ), the rhombicity parameter ͉R/͉ and the coefficients (D, E, and F) of the ground state doublet (Equation 1) (30). The molecular orbital coefficients for the low spin Fe(III) species observed in S. novella SoxAX (Table II) indicate that in all three species (LS1a, LS1b, and LS2), the unpaired electron is located predominantly in the ͉d xy Ͼ orbital. As observed for the R. sulfidophilum SoxAX protein, the contribution of LS1a, LS1b, and LS2 to the spectra varied between different enzyme preparations and also changed after treatment of the sample with ferricyanide. The proportions of the three species LS1a, LS1b, and LS2 in the EPR spectra of the two samples shown in Fig. 3, a and b, are as follows: LS1a, 87 and 7%; LS1b, 12 and 46%; LS2, 1 and 47% (Ϯ 5%), respectively. The g matrices for the low spin ferric heme LS1a,b species (Table II) are consistent with axial His/Cys ligation, and those for LS2 are consistent with the proposed axial His/modified-Cys ligation of a heme group. It has to be kept in mind, however, that the S. novella SoxAX protein contains only two heme groups, and only one of those, located on the SoxA subunit, appears to have a His/Cys axial ligation. Although EPR spectra of the as-prepared diheme SoxAX samples recorded over larger scan widths revealed no evidence (g ϭ 3.5 resonance (11)) for a type I heme with axial His/Met coordination, anaerobic oxidation of SoxAX with one equivalent of ferricyanide revealed a spectrum dominated by a g ϳ 2 resonance from ferricyanide and a weak asymmetric g ϭ 3.50 resonance from a type I heme (LS3 in the nomenclature of Ref. 11; Fig. 4), which is assigned to the heme group present on the SoxX subunit of the protein.
Analysis of the Redox Properties of Native SoxAX-The redox properties of the heme groups found in SoxAX were determined by optical redox potentiometry. A single heme-based, pH-dependent couple was identified. The Fe III/II potentials (versus NHE) are E o ϩ133 mV (pH 6.0), ϩ104 mV (pH 7.0), ϩ49 (pH 7.9), and ϩ10 mV (pH 8.7). The pH dependence indicates that both hemes undergo protonation concomitant with reduction to the ferrous state. Given the relatively narrow pH range that could be investigated, no pK a values could be determined for either the ferric or ferrous form of either heme, and so the sites of protonation are not known. In any case, the two hemes evidently exhibit very similar redox potentials despite the differences in their axial ligands. This fact was confirmed by measuring the EPR spectrum of a sample electrochemically poised at Ϫ100 mV (versus NHE), which exhibited no (ferric) heme-based signals (data not shown). The redox potentiometry data are available as supplemental material.
Phylogenetic Analysis of SoxA-and SoxX-related Proteins-Protein sequences related to SoxA and SoxX were retrieved from GenBank TM . While the SoxA-related sequences contained between 310 and 267 amino acids, the SoxX-related sequences exhibited much more variation with between 117 and 285 amino acids/sequence. This variation is also reflected in the observed similarities between the sequences; pairwise alignments of SoxA sequences against the S. novella sequence had a minimum length of 233 amino acids and a minimum similarity value of 39% over this length. The SoxX sequence alignments varied in length between 77 and 211 residues. Over a 77-amino acid alignment of the R. sulfidophilum and S. novella sequences, a similarity of 49% was obtained; however, an alignment of the same sequences using 174 residues only showed an overall similarity of 36%. Multiple sequence alignments were constructed after exclusion of low homology regions, and a phylogenetic analysis was undertaken. Distance methods (neighbor-joining, fitch, kitsch) yielded very similar phylogenetic tree topologies for both SoxA and SoxX (Fig. 5). Maximum Parsimony analysis of the data revealed that the placement of the Magnetococcus MC-1 soxA sequence relative to the central node could vary. Bootstrap (Fig. 5) and jackknife analysis of the data yielded virtually identical values. In both trees obtained, the sequences from S. novella, R. palustris, and the two Ralstonia species form a clade that appears to be closely related to the A. aeolicus sequences. The remaining sequences form another closely related group in which the Chlorobium and Magnetococcus sequences represent branches that are distinct from the Paracoccus/Rhodovulum group. The monoheme forms of SoxA are located in the Chlorobium and in the S. novella sequence group. The S. novella group also comprises the SoxX sequences with the most significant N-terminal extensions; all sequence in the S. novella group have gained residues (2-79 amino acids) relative to the A. aeolicus SoxX sequence and contain more than 200 amino acids. In contrast, those located in the other major group have less than 200 amino acids (mainly 140 -160 amino acids) and have lost between 8 and 89 residues. Whether the differing sizes of the SoxX subunits have implications for their function in the SoxAX dimer cannot be established at present because of a lack of information about the in vivo function of the proteins in the S. novella group.
The S. novella sox Gene Cluster-The genetic environment of the soxAX genes from S. novella was also analyzed in some detail. While some of the genes (soxVW, soxYZBC, and soxF) are closely related to genes for which the possible functions have been described elsewhere (5, 6, 8, 44 -46), there are a number of interesting new features present in this sox gene cluster, such as the existence of two separate transcriptional units (Fig. 1A) allowing for a differential expression of the TOMES components.
The soxD gene encodes a monoheme cytochrome of 22 kDa that is similar to the one found in R. sulfidophilum. By contrast, the SoxD protein from P. pantotrophus contains a second heme group in its C-terminal region. SoxD is a subunit of the SoxCD protein, which contains a molybdenum center at the active site of the SoxC subunit. The loss of a heme group from the protein changes the heme:Mo ratio from 2:1 to 1:1, and this in turn has implications for the turnover of the Mo center, which has to give up two electrons, taken up during catalysis, to the accessory heme group(s) before another reaction cycle can be initiated. It has been suggested that the loss of the C-terminal part of the SoxD protein with the concomitant loss of a heme group may render the SoxCD protein nonfunctional (8). This evidence was supported by experiments showing no detectable sulfite dehydrogenase activity in R. sulfidophilum cell extracts. However, it is also possible that the monoheme SoxCD is merely a variant form of the SoxCD protein, which in fact does not exhibit significant sulfite dehydrogenase activity in its purified form (10) and has recently been suggested to be a sulfur dehydrogenase (3). Also, a related sulfite dehydrogenase (SorAB) has a heme:Mo ration of 1:1 and is highly active toward sulfite (18,47).
The most interesting difference observed between the S. novella sequence and the sox clusters described so far is the replacement of the soxE gene, which encodes a cytochrome thought to be the heme subunit of the flavocytochrome encoded by soxF, by an open reading frame (orf1) encoding a transmembrane protein with an N-terminal CXXXC motif. The closest relative of orf1 found in GenBank TM is an unknown protein from A. aeolicus, Aq_1119, which, however, does not appear to be a transmembrane protein. The ORF1 protein is predicted to have a molecular mass of 18.7 kDa (173 amino acids) and a slightly acidic pI of 5.9. There does not appear to be an export signal, but predictions show that a central membrane-span-ning region from amino acids 70 to 93 is present. Two minor hydrophobic regions are found at positions 144 -168 and at the N terminus (ca. residues 10 -23). The topology of the protein is uncertain, as it depends largely on the number of transmembrane helices present. The CXXXC motif (positions [45][46][47][48][49] is related to the thioredoxin motif (CXXC) and has also been found in Sco-related proteins such as PrrC from Rhodobacter sphaeroides (48). Such a motif could be involved in the binding of sulfur compounds or metal ions; the latter function has been suggested for the CXXXC motif from PrrC (48). The ORF1 protein might be involved in anchoring the S. novella TOMES to the cell membranes, leading to a membrane-bound thiosulfate oxidizing activity as described for this microorganism (49,50). DISCUSSION In the present study we have reported the purification and characterization of a diheme form of a SoxAX cytochrome c 551 from S. novella, a protein shown to be an essential component of a thiosulfate-oxidizing multienzyme complex in a number of microorganisms. On the basis of the published crystal structure of the related triheme SoxAX protein from R. sulfidophilum and sequence comparison, we predict that the heme group located on the SoxA subunit has a His/Cys ligation, whereas the one located on the SoxX subunit should have a His/Met ligation. The second heme group found in the R. sulfidophilum SoxA has been replaced by a disulfide bond between a Cys (Cys-37) from the binding motif and the former heme ligand (Cys-73) (Fig. 2). The redox potentials of the two independent hemes could not be resolved, but both are coupled to a proton transfer reaction upon reduction. The sites of protonation remain uncertain, and it may be possible that the pH dependence of the potentials is in some way linked to changes in the predominant conformation of the SoxA heme group. The related triheme SoxAX from R. sulfidophilum has been crystallographically characterized in both its fully oxidized and dithionite-reduced forms. No significant conformational changes at either heme 2 or heme 3 (the two hemes conserved in S. novella SoxAX) were seen (12). However, it has been reported (11) on the basis of spectroscopic results that reduction of heme 2 in SoxAX from R. sulfidophilum is associated with either protonation or substitution of an axial ligand. The heme propionate residues are also commonly associated with proton-coupled electron transfer in c-type cytochromes (51). H-bonding and salt-bridge formation between the heme propionates and nearby amino acid side chains have been shown to have a significant influence on the pK a values of the propionates (52). Although we have no structural data for S. novella SoxAX, the corresponding heme propionates identified crystallographically in R. sulfidophilum SoxAX form salt bridges with three arginine residues (Arg-38 and Arg-260 in heme 2 (in SoxA) and Arg-70 in heme 3 (SoxX). The remaining heme propionate (on a The rhombicity parameter ͉R/͉ and molecular orbital coefficients (Eq. 1) were calculated with ls_iron as described in the text. b Unresolved resonance. c Value cannot be determined since the g matrix is not completely defined. d The assignment of the g values listed in Table 2 of Ref. 11 is incorrect. The correct assignments are given in this table, together with the rhombicity parameter ͉R/͉ and molecular orbital coefficients. heme 3) is H-bonded to Thr-90 and to water molecules. Although the residues in question are not conserved in the S. novella SoxAX amino acid sequences, they are conservatively replaced by amino acids with similar properties (exception, Arg-38). Hence, the observed pH dependence of the coincident heme potentials is in agreement with that seen in other c-type cytochromes exhibiting arginine salt bridges.
EPR spectra (Figs. 3 and 4) of the diheme S. novella SoxAX protein had the same complexity as those reported by Cheesman et al. (11) for the triheme SoxAX from R. sulfidophilum. The spectra revealed the presence of three low spin Fe(III) type II heme species (LS1a, LS1b, LS2) and an additional low spin Fe(III) type I heme species (LS3). The g ϭ 3.502 resonance for LS3 (Fig. 4) is consistent with axial His-Met ligation that has been proposed for the heme located on the SoxX subunit on the basis of sequence alignment with the R. sulfidophilum SoxX sequence. Although the electronic absorption spectrum of as isolated SoxAX (Fig. 1B) indicates that both hemes are fully oxidized, the absence of the g ϭ 3.5 resonance associated with LS3 suggests that electron transfer takes place between the two hemes and that, despite its optical appearance, the asprepared SoxAX protein is not fully oxidized. The EPR spectra (Fig. 3) of the type II hemes closely resemble the LS1a, LS1b, and LS2 spectra reported by Cheesman et al. (11) for the R. sulfidophilum (RS)SoxAX protein, with the g matrices for LS1 indicating a His-Cys axial ligation and those for LS2 being consistent with the proposed His/Cys-persulfide ligation (11,12). Similar to what was observed in the (RS)SoxAX protein, the proportions of the three type II heme components (LS1a, LS1b, LS2) were found to vary significantly between two preparations (LS1a, 87 and 7%; LS1b, 12 and 46%; LS2, 1 and 47% (Ϯ5%). However, whereas in the R. sulfidophilum enzyme the proportions of total LS1 and LS2 never fell below 100/100, the first sample of the diheme S. novella SoxAX contains almost exclusively LS1 forms and very little of the LS2 form. As previously noted, S. novella SoxAX is missing the heme I (type II) found in R. sulfidophilum SoxAX, and consequently the observed EPR spectra from the three low spin Fe(III) species, LS1a, LS1b, and LS2, must arise from the heme present in the SoxA subunit. Clearly, the different proportions of LS1(a,b) and LS2 indicate the presence of an equilibrium involving modification of the axial Cys ligand (LS1) to produce an axial Cyspersulfide ligand (LS2). The slightly different g matrices for LS1a and LS1b suggest different conformational states of the catalytic heme, which may facilitate catalysis. These observations are consistent with the proposal that in the (RS)SoxAX, the noncatalytic heme 1 is always in an LS1 (a,b) conformation, whereas the catalytic heme is present as either LS1 (a,b) or LS2 (11).
The phylogenetic analysis of the SoxAX protein family showed that considerable diversity exists within the SoxAX family, with the SoxA proteins more closely related to one another than the SoxX proteins. However, the strikingly similar topology of the phylogenetic trees obtained for the two subunits suggests that they may have evolved from a common ancestor. If that is the case, the loss of the N-terminal SoxA heme group must have occurred twice, as monoheme SoxA proteins are found in the S. novella group of proteins as well as in the Chlorobium branch of the P. pantotrophus group. Whether the N-terminal extensions of the SoxX subunit are indicative of a changed functionality is unclear at this stage.
Our results clearly show that the cytochrome c 551 from S. novella is a novel member of the SoxAX protein family. The presence of only two redox centers that correspond to those present at the catalytic site make it a useful tool in the study of the SoxAX catalytic mechanism and the changes in the ligand environment of the catalytic heme group present on the SoxA subunit.