Tungstate Uptake by a Highly Specific ABC Transporter inEubacterium acidaminophilum *

The Gram-positive anaerobeEubacterium acidaminophilum contains at least two tungsten-dependent enzymes: viologen-dependent formate dehydrogenase and aldehyde dehydrogenase.185W-Labeled tungstate was taken up by this organism with a maximum rate of 0.53 pmol min− 1mg− 1 of protein at 36 °C. The uptake was not affected by equimolar amounts of molybdate. The genestupABC coding for an ABC transporter specific for tungstate were cloned in the downstream region of genes encoding a tungsten-containing formate dehydrogenase. The substrate-binding protein, TupA, of this putative transporter was overexpressed inEscherichia coli, and its binding properties toward oxyanions were determined by a native polyacrylamide gel retardation assay. Only tungstate induced a shift of TupA mobility, suggesting that only this anion was specifically bound by TupA. If molybdate and sulfate were added in high molar excess (>1000-fold), they were also slightly bound by TupA. The K d value for tungstate was determined to be 0.5 μm. The genes encoding the tungstate-specific ABC transporter exhibited highest similarities to putative transporters from Methanobacterium thermoautotrophicum, Haloferax volcanii, Vibrio cholerae, and Campylobacter jejuni. These five transporters represent a separate phylogenetic group of oxyanion ABC transporters as evident from analysis of the deduced amino acid sequences of the binding proteins. Downstream of the tupABCgenes, the genes moeA, moeA-1,moaA, and a truncated moaC have been identified by sequence comparison of the deduced amino acid sequences. They should participate in the biosynthesis of the pterin cofactor that is present in molybdenum- and tungsten-containing enzymes except nitrogenase.

The function of tungsten as an essential trace element for some archaea and bacteria has now been fully recognized (1)(2)(3). It is incorporated into a pterin cofactor that is required for assembly and function of enzymes such as acetylene hydratase, aldehyde ferredoxin oxidoreductase, carboxylic acid reductase, formaldehyde dehydrogenase, formate dehydrogenase, formylmethanofuran dehydrogenase, and glyceraldehyde-3-phosphate ferredoxin oxidoreductase (3)(4)(5)(6)(7)(8)(9)(10). Tungsten is very similar in size and chemical behavior to molybdenum, and therefore, tungsten can act as a specific antagonist of molybdenum-containing enzymes. Tungsten-containing enzymes seem to catalyze similar or even the same reactions as related molybdoenzymes (1)(2)(3). In contrast to tungstate (11,12), the uptake of molybdate into cells has been well studied and is mediated by an ABC transporter system (13)(14)(15). The molybdate-binding protein ModA has been identified to bind specifically molybdate and tungstate (16), and the structures of the ModA proteins from Escherichia coli and Azotobacter vinelandii have been analyzed at 1.75 and 1.25 Å resolution, respectively (17)(18)(19). The specificity of ModA to bind molybdate and tungstate, but not sulfate or other anions, is determined mainly by the size of the binding pocket for molybdate/tungstate (17). Because tungstate and molybdate have nearly identical sizes, ModA cannot discriminate between these anions (16 -19). At present, only a little is known about a specific uptake mechanisms for tungstate into microorganisms. Labeling studies were done for certain enzymes, and the tungstate requirement for optimal growth was analyzed for some organisms, most of them being (hyper)thermophilic (1)(2)(3).
The Gram-positive anaerobe Eubacterium acidaminophilum grows at mesophilic temperatures, degrades amino acids by Stickland reactions, and requires the addition of selenite but not molybdate or tungstate for growth (20). Two enzymes are involved in its metabolism: formate dehydrogenase and aldehyde dehydrogenase, which are now described to be specifically dependent on tungsten availability to exhibit enzymatic activity. This correlates with a specific incorporation into both enzymes (21,22). This report focuses on uptake experiments using resting cells of E. acidaminophilum and radioactive labeled tungstate as well as on cloning and sequencing of an ABC transporter. The extracytoplasmic binding protein of this transporter, TupA, was overexpressed in E. coli, and its anion binding characteristics show a specificity for tungstate. Thus, this is the first report of a transport system specific for tungstate.

EXPERIMENTAL PROCEDURES
Materials-All chemicals were obtained from commercial sources unless otherwise specified.
Bacterial Strains and Plasmids-E. acidaminophilum (DSM 5388 T ) was grown anaerobically as described (20). Glycine (50 mM) or the substrate combination serine (10 mM), formate (40 mM), betaine (60 mM) was used as growth substrate. The organism was also grown without the addition of tungstate and molybdate as described in this paper. E. coli XL1-Blue MRFЈ and E. coli SOLR were obtained from Stratagene (Heidelberg, Germany) and used for cloning purposes and expression of proteins. They were grown in Luria broth medium (23) or on agar plates containing 1.5% (w/v) agar. 100 g ml Ϫ1 ampicillin was added, and 40 g ml Ϫ1 5-bromo-4-chloro-3-indolyl ␤-D-galactopyranoside and 48 g ml Ϫ1 isopropyl ␤-D-thiogalactopyranoside if needed. The plasmid vector pBluescipt KS ϩ was obtained from Stratagene, the protein expression vector pASK-IBA2 was from Institut fü r Bioanalytik (Göttingen, Germany).
Formate Dehydrogenase and Aldehyde Dehydrogenase Assay-Formate dehydrogenase and aldehyde dehydrogenase were measured by a * This work was supported by grants from the Land Sachsen-Anhalt and the Fonds der Chemischen Industrie. 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.
The nucleotide sequence(s) reported in this paper has been submitted to the GenBank TM /EBI Data Bank with accession number(s) AJ291988. standard procedure (12) at 34°C under anaerobic conditions by monitoring the substrate-dependent reduction of benzyl viologen at 578 nm (⑀ ϭ 8.3 mM Ϫ1 cm Ϫ1 ) or methyl viologen at 600 nm (⑀ ϭ 13.1 mM Ϫ1 cm Ϫ1 ). To assay formate dehydrogenase, the reaction mixture contained in a final volume of 1 ml, 50 mM Tris buffer (pH 8.0), 5 mM benzyl or methyl viologen, 20 mM sodium formate, and 0.1-10 l of extract. 1-5 l of 50 mM sodium dithionite was added to the reaction mixture to obtain a light reduction of the viologen as an indication for strict anaerobic conditions. Aldehyde dehydrogenase was measured under similar conditions (6) except that the reaction mixture contained in a final volume of 1 ml, 50 mM Tris buffer (pH 8.5), 25 mM benzyl viologen, and 0.5 mM acetaldehyde. Blanks were run with the same reaction mixture excluding the substrate. The reactions were always started by the anaerobic addition of substrate. 1 unit of enzyme activity was defined as the amount of enzyme catalyzing the reduction of 2 mol of viologen min Ϫ1 .
Uptake For uptake experiments, cells (10 ml) were grown overnight with serine/formate/betaine as substrates, harvested anaerobically at 2,000 ϫ g, and washed twice with 10 ml of ice-cold 50 mM Na 2 HCO 3 buffer (pH 7.5). Cells were suspended in the same buffer, and 10 mM serine, 40 mM formate, 60 mM betaine, and 30 g ml Ϫ1 chloramphenicol were added to give a final volume of 9.9 ml. Subsequently, cells were incubated at 30°C, and 100 l of a 10 M [ 185 W]tungstate solution was added. At different time intervals, 1-ml samples were taken and collected on nitrocellulose filter discs (pore size 0.45 m, Schleicher & Schü ll, Dassel, Germany). The filter discs had been preincubated in 50 mM Na 2 HCO 3 buffer (pH 7.5) containing 1 mM tungstate and were washed after cell collection with 20 ml of this buffer. Filter discs were dried for 30 min at 60°C, incubated in scintillation fluid (Beckmann, Mü nchen Germany), and the radioactivity was determined in a Beckmann LS6500 scintillation counter with an efficiency of nearly 90%.
To label proteins of E. acidaminophilum with [ 185 W]tungstate, cells were grown overnight with glycine as substrate in 3 liters of medium supplemented with [ 185 W]tungstate (4⅐10 5 Bq) and different concentrations of unlabeled tungstate to give the final concentration. Cells were harvested by centrifugation at 10,000 ϫ g and 15 min and were washed two times with potassium phosphate buffer (50 mM, pH 7.5) and suspended in this buffer. After rupture of the cells by two passages through a French pressure cell, cell debris were removed by centrifugation at 20,000 ϫ g for 20 min. To separate the labeled proteins, the supernatant was desalted using a PD10 column (Amersham Pharmacia Biotech) and applied to a Superdex-200 gel filtration column (Amersham Pharmacia Biotech). The column had been preequilibrated with 50 mM potassium phosphate buffer (pH 7.5) containing 150 mM KCl.
DNA Manipulations and Sequence Determination-Routine DNA techniques were performed as described by Sambrook et al. (23). Genomic DNA from E. acidaminophilum was isolated according to Saito and Miura (24). Enzymes were used according to the recommendations of the manufacturer. Plasmid preparations were done using the Qiagen (Hilden, Germany) kits as outlined in the manufacturer's manual. Nucleotide sequences were determined by the dideoxy chain termination method (25) using the Rhodamine Terminator Cycle Sequencing Ready Reaction kit (PerkinElmer Life Sciences) and analyzed using an Applied Biosystems PRISM 377 DNA sequencer. Sequence information of large inserts were obtained using the GPS TM genome priming system from New England Biolabs (Frankfurt, Germany) as recommended by the manufacturer. Alternatively, the primer-walking method was used. The oligonucleotides were synthesized by Metabion (Martinsried, Germany). DNA fragments were labeled, and hybridizing bands were detected using the DIG DNA labeling and detection kit from Roche Diagnostic according to the manufacturer's manual. A ZAP II library was constructed using the ZAP II-predigested EcoRI/CIAPtreated vector Kit (Stratagene). Genomic DNA of E. acidaminophilum was digested with the restriction endonuclease EcoRI and separated by centrifugation for 24 h at 200,000 ϫ g using a sucrose density gradient (10 -40% (w/v) sucrose). DNA fragments of 5-10 kb 1 were ligated into the EcoRI site of the ZAP II vector and subjected to in vitro packaging according to the manufacturer. The resulting phage particles represented a library of E. acidaminophilum DNA and were screened by plaque hybridization after infection of E. coli XL1-Blue MRFЈ. M13 ExAssist helper phage and E. coli SOLR were used for in vivo excision of the pBluescript phagemid according to the instructions provided by Stratagene.
RNA Techniques-E. acidaminophilum was grown in medium containing glycine or serine/formate/betaine as substrates to an absorbance of 0.9 and harvested by centrifugation at 4,000 ϫ g and 4°C for 10 min. Total RNA was isolated using the RNeasy Mini Kit (Qiagen) with modifications. 10 -25-ml cultures were used, and the cells were lysed in 20 mg of lysozyme ml Ϫ1 for 10 min. Northern hybridizations were performed as described (26).
For RT-PCR experiments, purified RNA was treated with DNase RQ1 (Promega, Mannheim, Germany) following the instructions of the supplier to obtain DNA-free RNA. For reverse transcription of mRNA, 1 l of purified RNA and 1 l of hexanucleotide mixture (500 g/ml) were added to 10 l of H 2 O, denatured for 10 min at 70°C, and cooled for 2 min on ice. Subsequently, 4 l of first strand buffer, 2 l of 0.1 M dithiothreitol, 1 l of 20 mM dNTP, and 0.7 l of reverse transcriptase (Superscript II, Life Technologies, Inc.) were added, and the mixture was incubated for 10 min at room temperature and 1 h at 42°C. The reaction was terminated by incubation at 94°C for 5 min. This reaction mixture contained the synthesized cDNA, and 1 l of it was used as template in the following PCR. 100 pmol of forward and 100 of pmol reverse primer, 1 l of 20 mM dNTP, 1 l of Taq DNA polymerase (Quantum Appligene, Heidelberg), 5 l of Taq buffer, and 42 l of H 2 O were mixed, and PCR was performed with 30 cycles: 30 s at 94°C for denaturing, 30 s of annealing at 42-72°C (depending on the primers used), and 1-3-min extension (depending on the distance between both primers) at 72°C.
Cloning of tupA into the pASK-IBA2 Vector-To clone the tupA coding sequence into the pASK-IBA2 expression vector (IBA, Göttingen), the gene was amplified by PCR using Pwo DNA polymerase (Roche Diagnostics), primers PA3 and PA3r (Table I), and chromosomal DNA from E. acidaminophilum as template. The leader peptide of TupA including the conserved cysteine residue was omitted and replaced by the OmpA leader peptide provided by the vector pASK-IBA2. Both primers contain a BsaI restriction site and were cloned into the respective site of pASK-IBA2 resulting in the TupA fusion protein with an N-terminal OmpA signal peptide and a C-terminal Strep-tagII. This plasmid, pASK-tupA, was transformed into E. coli XL 1-Blue MRFЈ. The insert of pASK-tupA was sequenced for confirmation.
Expression and Purification of TupA-An E. coli XL1-Blue clone harboring the plasmid pASK-tupA was inoculated into 3 ml of LB medium containing 100 g ml Ϫ1 ampicillin and was grown overnight at 30°C. 2 ml of this culture was transferred to 100 ml of LB medium containing 100 g ml Ϫ1 ampicillin. After the culture had reached an absorbance of 0.5 at 550 nm, 10 l of an anhydrotetracyclin solution (2 mg ml Ϫ1 in dimethyl sulfoxide) was added to induce expression of the binding protein. After 3 h at 30°C the cells were harvested by centrifugation for 12 min at 4,000 ϫ g and 4°C and used directly for the preparation of crude extract or stored at Ϫ20°C.
E. coli cells overexpresing TupA were suspended in buffer A (100 mM Tris (pH 8.0), 0.5 g of cells ml Ϫ1 buffer) and were lysed by sonication using a GM 60 HD sonicator (Uni Eqip Laborgerä tebau, Martinsried, Germany) at intervals of 5 s until most cells were broken. After each interval the samples were cooled on ice for 10 s. Cell debris were removed by centrifugation at 25,000 ϫ g and 4°C for 5 min. The resulting supernatant was used directly or stored at Ϫ20°C. The supernatant containing TupA was applied to a 1-ml Strep-Tactin column (IBA) that had been equilibrated previously with buffer A. The column was washed with 5 ml of buffer A and subsequently, the TupA-binding protein was eluted in 3 ml of buffer A containing 2.5 mM desthiobiotin.
TupA Protein Gel Shift Assay-The ligand-dependent gel shift assay developed by Rech et al. (16) was used to analyze anion binding by TupA. Aliquots of purified TupA were incubated with the indicated concentration of anion in binding buffer (50 mM potassium acetate (pH 5.0), 100 mM Tris (pH 8.0)) for 30 min on ice. Samples were mixed with 0.25 volume of a sucrose solution (30% w/v) containing bromphenol blue and applied to a native 12% polyacrylamide gel. Protein was separated in a mini-Gel system (Biometra, Göttingen). The gel was buffered with 50 mM Tris (pH 8.5), and the running buffer contained 0.1 M Tris and 0.1 M glycine. Electrophoresis was done at 150 mV and 4°C until bromphenol blue left the gel.
Analytical Methods-The concentration of proteins was determined by the method of Bradford (27) with bovine serum albumin as a standard. Whole cells were incubated in 0.1 M NaOH for 10 min at 95°C and neutralized with 0.5 M HCl prior to protein determination. The extinction coefficient for TupA was determined to 24,460 mol Ϫ1 cm Ϫ1 according to Gill and von Hippel (28) and used to calculate protein concentration of the purified protein. SDS-polyacrylamide gel electrophoresis was done with the Laemmli buffer system as described (29). Proteins were blotted from an SDS gel onto a polyvinylidene difluoride membrane according to Towbin et al. (30) in 50 mM NaBO 3 buffer (pH 9.0) containing 20% (v/v) methanol at 1.2 mA cm Ϫ2 membrane for 1-2 h. For Edman degradation, blotted proteins were cut from the polyvinylidene difluoride membrane and analyzed in a 476A amino acid sequencer (Applied Biosystems, Weiterstadt, Germany). Molybdenum and tungsten were determined by ICP-MS with an Agilent 7500 Series ISP-MS (Agilent Technologies, Waldbronn, Germany).

Uptake of [ 185 W]Tungstate by Growing Cells of E. acidaminophilum and
Labeling of Proteins-Omission of tungstate and molybdate from the added trace element solution did not affect growth of E. acidaminophilum over hundreds of generations in a defined mineral medium on different substrates. After growth of E. acidaminophilum in the presence of 10 Ϫ9 M [ 185 W]tungstate, labeled proteins were analyzed by gel filtration. This revealed a strong radioactivity peak at 74 kDa and one shoulder at about 160 kDa. If 10 Ϫ8 M radioactive tungstate was present during growth, the specific radioactivity eluting at about 160 kDa was nearly unchanged, whereas the value of the radioactivity peak at 74 kDa increased 2-fold, and a new radioactivity peak appeared at about 40 kDa which became prominently labeled if higher concentrations of tungstate were provided (data not shown). The simultaneous addition of equimolar amounts of molybdate had a negligible effect on the observed labeling pattern (data not shown), pointing to a high specificity for tungstate.
However, a coelution of radioactivity and enzyme activity could already be obtained for the 160-kDa protein with a viologendependent formate dehydrogenase (31,32) and for the 74-kDa protein with a viologen-dependent acetaldehyde dehydrogenase (32). Both enzymes were now independently purified to homogeneity and contain tungsten but no molybdenum. 2,3 Tungstate and molybdate had quite different effects on both enzyme activities. In tungstate/molybdate-depleted cells of E. acidaminophilum, viologen-dependent acetaldehyde dehydrogenase activity was no longer measurable, whereas viologendependent formate dehydrogenase activity was still present at 18% of its maximum value (100% ϭ 1.1 unit mg Ϫ1 ) which was obtained after growth in the presence of 10 Ϫ7 M tungstate added to the medium. The latter concentration was also optimal for aldehyde dehydrogenase activity (0.56 unit mg Ϫ1 ). Formate dehydrogenase activity was unaffected by higher tungstate concentrations added to the medium, whereas the aldehyde dehydrogenase activity declined by about 30% if 10 Ϫ5 M tungstate was present during growth. In the absence of tungstate, the addition of molybdate at physiological concentrations (10 Ϫ9 to 10 Ϫ7 M) did not give rise to an aldehyde dehydrogenase activity. However, if a high concentration of molybdate (10 Ϫ5 M) was added to the medium, the specific activities of acetaldehyde dehydrogenase and formate dehydrogenase increased to values of 14 and 62% of their maximum activity, respectively. These activity levels corresponded to those determined for a suboptimal tungstate supplementation by 10 Ϫ9 M. Because molybdenum salts are contaminated by tungsten (33), the induction of acetaldehyde dehydrogenase and formate dehydrogenase activity by high molybdate concentrations added to the medium might be caused by a contami-nation by tungstate. However, it might also be possible that both tungsten-containing enzymes incorporate molybdenum and form an active enzyme when this metal is added in high concentrations to the medium. At least, these physiological results suggested a strong preference for tungstate compared with molybdate and a specific uptake of tungsten even at quite low concentrations by growing cells.
Uptake of [ 185 W]Tungstate into Resting Cells of E. acidaminophilum-Cells of E. acidaminophilum were grown with serine/formate/betaine as carbon and energy source. Uptake experiments using [ 185 W]tungstate were carried out with 1 M labeled tungstate and resting cells in carbonate buffer and serine/formate/betaine as energy source. The highest uptake rate of 0.53 pmol min Ϫ1 mg of protein Ϫ1 was obtained at 36°C (Fig. 1). Significant lower values were obtained at room temperature (22°C) and at 50°C. This is in good agreement with the optimal growth temperature of E. acidaminophilum, which is between 30 and 35°C (20). Uptake experiments at 0°C, performed as control, gave a strongly reduced uptake rate of 0.1 pmol min Ϫ1 mg of protein Ϫ1 (Fig. 1). Tungstate uptake was not influenced by the presence of equimolar amounts of molybdate in the uptake buffer (data not shown). No impairment of tungstate uptake was observed if the ATPase inhibitor arsenate (5 mM) or the uncoupler 2,4-dinitrophenol (2 mM) was present during the uptake experiments by resting cells (data not shown).
Cloning of the tupABC Operon-The genes encoding a tungstate uptake system were identified in the downstream region of the genes encoding the tungsten-containing formate dehydrogenase I and some enzymes involved in pterin-cofactor biosynthesis of E. acidaminophilum (21) (Fig. 2A, depicted as 7.3-kb EcoRI fragment). Sequence comparison and later biochemical analysis (see below) revealed that the three genes in the immediate downstream region of the mentioned 7.3-kb fragment encode an ABC uptake system that seems to be specific for tungstate. According to the results presented below, these genes have been named tup for tungstate uptake (Fig. 2). To clone this EcoRI fragment downstream of formate dehydrogenase I genes, a Southern blot analysis was performed using probe A, which was deduced from the 3Ј-region of the 7.3-kb EcoRI fragment (Fig. 2 and Table I primers P probeA ). First, a 3-kb SacI fragment was identified which covers the 3Ј-part of the 7.3-kb EcoRI fragment and the 5Ј-part of the following 6.9-kb EcoRI fragment (Fig. 2). SacI-digested chromosomal DNA from E. acidaminophilum was separated in a sucrose density gradient and fractionated. Using probe A, a positive reacting fraction was identified which contains fragments of 2-4 kb. Subsequently, these fragments were ligated into pBSK Bluescript and transformed into E. coli XL1-Blue MRFЈ. All transformants were washed from the agar plates and used to inoculate 100 ml of LB medium. After growth overnight, the plasmid mixture was isolated from this culture and was used directly as a template in a PCR that was done with primers FDH long and universal primer (Table I). A 3.1-kb fragment was amplified and partly sequenced to obtain sufficient sequence information from the so far unknown DNA region downstream of the EcoRI site to generate probe B (Fig. 2, Table I primers P probeB ). With that probe a ZAP-EcoRI library of chromosomal DNA from E. acidaminophilum was screened. A positive phage was isolated, and after in vivo excision plasmid pKME5 containing a 6.9-kb EcoRI insert was obtained (Fig. 2). On this 6.9-kb fragment seven open reading frames including a truncated one were identified ( Fig. 2A). The deduced amino acid sequences of the genes tupABC located at the 5Ј-end of this DNA fragment exhibited similarities to transporters of the ABC-type uptake permeases (34, 35). TupA is similar to substrate-binding lipoproteins (36), TupB to the permease part, and TupC contains amino acid sequences similar to ATPases. The deduced amino acid sequence of the other open reading frames exhibited similarities to proteins involved in the biosynthesis of the molybdopterin cofactor and were thus named: moeA, moeA-1, moaA, and moaCЈ (Fig. 2). From the last, a sequence of only 87 amino acids was encoded on the plasmid. The genes tupA and tupB were identified by Southern hybridization to be present as a single copy in the genome of E. acidaminophilum (data not shown).
The N-terminal amino acid sequence of TupA was similar to leader peptides of bacterial lipoproteins (36). A conserved cysteine residue at position 21 preceded by a glycine or alanine residue is characteristic for such leader peptides, which are split at the N-terminal side of the cysteine after transport across the cytoplasmic membrane. Subsequently, the cysteine is modified in this class of proteins to contain a fatty acid at the amino group and a diacyl glycerol thioether at the thiol group. This lipid anchor is used to attach the protein to the outside of the cytoplasmic membrane in Gram-positive bacteria or mycoplasmas (37).
Expression and Purification of TupA-If the TupA protein was expressed in E. coli containing its native leader peptide, 90% of TupA was found in the cytoplasmic membrane of E. coli (data not shown), and only a very small amount of soluble but inactive TupA was obtained. Thus, TupA was expressed as a C-terminal Strep-tag fusion protein containing the OmpA leader peptide as outlined under "Experimental Procedures" resulting in a higher yield (5 mg of TupA from 100 ml of E. coli culture) of soluble and active protein. Five bands were present in the purified TupA preparation as revealed by analysis using native polyacrylamide gel electrophoresis (Fig. 3A), and no further purification/separation could be obtained (data not shown). The N-terminal amino acid sequences of these protein bands revealed that all bands represented only TupA, which differed slightly by its N-terminal start (Fig. 3B). Although these TupA protein species differed only by a few amino acids in length, their size seems to vary between 34 and 38 kDa as judged by denaturing SDS-polyacrylamide gel electrophoresis (data not shown). Most likely these forms of TupA were processed differently by the signal peptidase of E. coli. The native size of TupA as determined by gel filtration eluted in one peak with a maximum at 38 kDa, indicating that the expressed TupA was present as a monomer (data not shown).
Specificity of the TupA Protein for Binding Divalent Oxyanions-To examine the binding specificity of TupA, the native polyacrylamide gel mobility assay as developed by Rech et al. (16) was used. The anions sulfate, molybdate, tungstate, phosphate, chlorate, chromate, vanadate, and selenate were tested and supplied in a 90-fold molar excess of TupA (Fig. 3A). None of these anions except tungstate caused a mobility shift of TupA, indicating that TupA specifically binds tungstate. Binding specificity was not altered even testing pH values of 5 and 8 in the binding buffer (data not shown). If molybdate and sulfate anions were added in a higher than 1,000-fold molar excess to TupA, a mobility shift was observed, but it was weaker than the shift induced by tungstate (data not shown). 11 M TupA was titrated with 0 -30 M tungstate to estimate the dissociation constant for tungstate (Fig. 3C). An apparent K d value of ϳ0.5 M was determined for tungstate binding using Equation 1 where P 0 is the protein concentration, and A 0 is the anion concentration with 50% of the protein bound to the anion. TupA (286 amino acids, 30.9 kDa) exhibited no significant similarities to the molybdate-binding proteins ModG, ModE, and Mop and only weak similarities to ModA proteins and to other anion-binding proteins (Fig. 4 and data not shown) (14,38). Highest similarities were obtained to hypothetical proteins from Methanobacterium thermoautotrophicum (accession no.  2. Organization of the sequenced tup genes and the downstream region in E. acidaminophilum. A, the restriction sites (EcoRI, SacI) and the probes (probe A, probe B) relevant for cloning of the 6.87-kb EcoRI fragment are shown. A putative termination signal is indicated. B, putative transcripts were identified using RT-PCR and are indicated. G69162), Haloferax volcanii (CAB42540), Campylobacter jejuni (B81301), and Vibrio cholerae (A82188) (Fig. 4). Therefore, these proteins are termed TupA-homologous proteins in this publication. They are all included in operons encoding putative ABC transporters and represent the substrate-binding component (Fig. 5). The TupA-homologous protein from H. volcanii was identified during mutational analysis to be essential for nitrate respiration (39). The Tup-homologous ABC transporter from M. thermoautotrophicum has been suggested in the database entry to be an ABC transporter specific for sulfate. A phylogenetic tree was derived for TupA and the mentioned homologous proteins as well as for binding proteins specific for other oxyanions such as molybdate, sulfate, and phosphate. Additionally, other hypothetical binding proteins obtained during similarity searches are included. TupA and its homologs clustered together in a distinct phylogenetic group compared with the other oxyanion-binding proteins. Because the crystal structure of two ModA proteins is available, the amino acids that bind molybdate have been identified to be Ser-12, Ser-39, Ala-125, Val-152, and Tyr-170 in E. coli ModA and Thr-9, Asn-10, Ser-37, Tyr-118, and Val-147 in A. vinelandii ModA (Fig. 4) (17-19). TupA and its homologs show a conserved TTTS motif close to the corresponding molybdate-binding amino acids Ser-12 (E. coli) and Thr-9/Asn-10 (A. vinelandii) and a conserved threonine flanked by two glycine residues corresponding to Ser-37 (E. coli) and Ser-39 (A. vinelandii) (Fig. 4). Ala-125 of E. coli ModA and Tyr-118 of A. vinelandii ModA are replaced in TupA and its homologs by the positively charged Arg-135 present in the motif SRGDXSGT (Fig. 4). Arg-135 and Asp-137 are also conserved in the phosphate-binding protein of E. coli, and Arg-135 is involved in phosphate binding and linked to Asp-137 by salt bridges (40). Additionally, Val-147/ 152 in ModA is replaced by glycine in the TupA group (Fig. 4).
TupB (228 amino acids, 24.5 kDa) exhibited highest similarities (36% identity) to the permease protein MTH478 of the ABC transporter from M. thermoautotrophicum, whose binding protein was similar to TupA (Fig. 5). Nearly identical similarity values were obtained to the permease components of the putative ABC transporters from C. jejuni (AL139078, 34% identity), V. cholerae (AE004231, 33% identity), and H. volcanii (AJ238877, 32% identity), which are located in the operons encoding the TupA-homologous proteins (Fig. 5). Additionally, lower similarity values were determined for permease components of other ABC transporters including ModB from the molybdate-specific ABC transporter (Ͻ20% similarity). Five transmembrane helices were postulated to be present in TupB using the dense alignment surface method (41), which indicates a location of this protein in the cytoplasmic membrane (data not shown). The conserved C-terminal region EAAX 2 GX 9 IXLP, which is generally present in permease components of ABC transporters was identified to be RIGX 2 LGX 8 LXIR in TupB and is, thus, only slightly conserved. This sequence was suggested to constitute a recognition site for the ATP-binding protein component (42). TupC (214 amino acids, 23.6 kDa) was similar to a wide variety of ATP-binding proteins of ABC transporters with values of up to 39% identical amino acids calculated on the basis of the length of TupC. Other ATP-binding subunits of ABC transporters are up to 120 amino acid residues longer than TupC. The ATP-binding components of the ABC transporter of C. jejuni and of V. cholerae, whose putative binding proteins and permease proteins are similar to TupA and TupB, were also homologous to TupC with values of 34 and 32% identity, respectively. The respective proteins from both archaea H. volcanii and M. thermoautotrophicum showed identity values below 28%. In TupC amino acid sequences were present, which are similar to the Walker A (Gly-31 through Thr-38) motif and the Walker B motif (Leu-149 through Asp-153) (43), and should be responsible for nucleotide binding (data not shown). Near to the Walker B motif a conserved linker peptide ( 129 LSG-GETQRV 137 ) typical for the ATP-binding subunit of ABC transporters (42) was present in TupC. A conserved histidine residue (His-187) was identified in TupC which might also be involved in nucleotide binding (44).
Sequence Analysis of MoeA, MoeA-1, MoaA, and MoaC-The proteins encoded by moeA and moeA-1 exhibited just 53% identity to each other and differed in size. They exhibit similarities to many MoeA proteins from different sources present in the data bases. However, the highest similarity values were obtained to MoeA proteins from hyperthermophilic archaea like Pyrococcus abyssi and Archeoglobus fulgidus and lower values to MoeA proteins from bacterial sources. MoeA is involved in the biosynthesis of the molybdopterin cofactor. Hasona et al. (45) suggest that MoeA participates in the synthesis of a molybdenum-sulfur complex leading to formation of a thiomolybdate, which is apparently the molybdenum donor in the production of the molybdopterin cofactor. Additionally, evidence was provided that MoeA might regulate the expression of formate hydrogen lyase and respiratory nitrate reductase (46).
Sequence Analysis of MoaA and MoaCЈ-MoaA and MoaCЈ are similar to proteins involved in the formation of precursor Z, which is an intermediate during molybdopterin cofactor biosynthesis (47,48). MoaA contains an N-terminal and a C-terminal cysteine-rich region. The N-terminal motif (CXXXCYXC) has been also identified in NifB and in PqqE, a protein involved in the biosynthesis of PQQ (49). A mutation of these cysteine to serine residues leads to an inactivation of MoaA (50). It is speculated that this motif is necessary for the formation of iron-sulfur clusters (49). The function of the C-terminal cysteine motif in MoaA (CXXCX 14 C) is not known.
Transcriptional Analysis of the tupABC Gene Region-Transcription of the genes tupABC, moeA, moeA-1, moaA, and moaC was analyzed using RT-PCR because no signals were detected by Northern blot analysis, probably due to a low transcription rate of these particular genes. RNA was isolated from cells grown in a medium containing serine/formate/betaine as substrates. tupA and tupB were transcribed together on one transcript (transcript I) (Fig. 2B). Stronger signals were obtained when RT-PCR experiments were carried out with primers deduced from the coding region of tupA compared with a situation where one primer was deduced from tupA and one primer from tupB (data not shown). A loop structure (111 kJ mol Ϫ1 ) was identified in the sequence between tupA and tupB, which might represent a rho-independent transcriptional termination signal (Fig. 2). Probably transcription started at a promoter structure upstream of tupA, and one part of this transcript is terminated at the loop structure between tupA and tupB, the other part runs through this structure and covers tupB. Thus, an elevated transcription of tupA would be the result. By using RT-PCR no common transcript of tupAB and tupC has been identified. tupC was transcribed together with moeA on one extra transcript (transcript II) and also moeA-1, moaA, and moaCЈ were transcribed together (transcript III). Probably the last transcript extended into the unsequenced downstream region covering the complete moaC gene. No difference in these mRNA pattern was evident from RT-PCR data if RNA was isolated from cells grown with (1 M) or without the addition of the anions tungstate and molybdate to the medium (data not shown).   (40).

DISCUSSION
All data presented here indicate that tungstate is taken up in E. acidaminophilum by a high affinity and highly specific transport system. For the first time such a specificity is described. The deduced amino acid sequences exhibit similarities to ABC-type uptake permeases from different sources, particular to the sulfate, phosphate, and molybdate uptake transporters (SulT, PhoT, and MolT family, TC 3.A.1.6-8) according to Saier (35).
Two different key metabolic enzymes formate dehydrogenase I and II (21,31,32) and aldehyde dehydrogenase (22,32) of E. acidaminophilum contain tungsten. Thus, an uptake system specific for tungstate is essential for this organism (20). It has to be very efficient because the organism grew well and contained a reduced, but still active, tungsten-dependent formate dehydrogenase activity if no tungstate was added. Because E. acidaminophilum was cultivated over hundreds of generations without tungsten supplementation, and no growth impairment was observed (20, 32, and data not shown), it seems unlikely that tungstate was stored.
So far, no molybdenum-dependent enzyme has been detected in E. acidaminophilum. Therefore, a specific uptake system for tungstate, discriminating molybdate, would be reasonable. Otherwise, if a classical molybdoenzyme, e.g. xanthine dehydrogenase, would be present as in purinolytic anaerobes (17), E. acidaminophilum should contain a separate molybdate uptake system, and additional principles must be exerted by the cell to incorporate the proper metal ion into the respective enzyme. So far, no molybdate specific ABC transporter has been identified in E. acidaminophilum using heterologous probes deduced from the respective transporter genes from E. coli and Arthrobacter nicotinovorans (data not shown). However, the presence of two genes similar to moeA (moeA-1, moeA) might be indicative of proteins responsible for a selective incorporation of tungsten and molybdenum, respectively.
The ABC transporter described in this paper exhibits the typical structure of ABC transporter systems (for a review, see Refs. 34 and 35). It includes a substrate-specific binding protein such as TupA, which is localized in the periplasm of Gramnegative bacteria or attached as a lipoprotein to the outer side of the cytoplasmic membrane in Gram-positive bacteria; a permease (TupB), which is an integral membrane protein that facilitates the transport across the membrane; and a protein (TupC) attached to the cytoplasmic side of the membrane that hydrolyzes ATP to provide energy for the transport.
TupA had a K d value of 0.5 M for tungstate, indicating a high specificity. Molybdate and sulfate were bound weakly by TupA when they were added in a more than 1,000-fold molar excess, indicating that the K d values for these two anions are orders of magnitude higher than for tungstate or that both substances contained tungstate as contaminant. Other anions like vanadate and phosphate were not bound to TupA. Thus, TupA is able to discriminate between tungstate and all other tested anions including molybdate. Molybdate and tungstate are chemically very similar and have the same ionic radii (cited in Ref. 1). In contrast to TupA, the molybdate-binding protein ModA of E. coli cannot discriminate between molybdate and tungstate, and thus both anions induce a mobility shift of ModA in native polyacrylamide gels (16). The K d values were determined to be 3 M for molybdate and 7 M for tungstate in ModA (16); however, other authors reported a K d value of 0.02 M for molybdate, but giving no value for tungstate (51). The K d value for molybdate binding to the regulatory protein ModE of E. coli was determined to be 0.8 M (52).
In contrast to the ModA proteins investigated, the molybdate repressor ModE (14,51), and the small molybdate/tungstatebinding protein Mop of Sporomusa ovata (53), TupA was able to discriminate between the anions molybdate and tungstate. It is an interesting questions how the protein accomplishes this specificity at the molecular level. A discrimination by size seems unlikely because both anions are nearly identical in size (1,17). The pK a value of tungstate is 4.7, the pK a of molybdate is 3.8 (1), and thus, TupA might be able to take advantage of this difference. Hydrogen bond energy increases when the pK a values of donor and acceptor become matched at the transition state (54). The ligands binding tungstate should reflect this behavior exhibiting stronger hydrogen bonds to the more basic tungstate and thus bind this anion with higher affinity than molybdate and other oxyanions. At least, the amino acids known to be involved in molybdate chelation by ModA (17,19) exhibited noticeable changes in the TupA of E. acidaminophilum and related proteins (Fig. 4). To identify the molecular basis of the ability of TupA to discriminate tungstate and other anions, the structure of TupA has to be analyzed. Experiments to crystalize TupA are under way in our laboratory.
The ABC transporters exhibiting highest similarities to the tupABC-encoded proteins were identified in archaea (H. volcanii, M. thermoautotrophicum) and Gram-negative proteobacteria (V. cholerae, C. jejuni). TupA from E. acidaminophilum and the four TupA-homologous proteins form a phylogenetic separate group compared with other anion-binding proteins (Fig. 5). These data suggest that all of these organisms might have a specific tungstate uptake system and consequently should have tungsten-containing cell components. Tungsten-dependent enzyme activities are known for M. thermoautotrophicum (1,55), and it is likely that they should also occur in H. volcanii because predominantly archaea are known to contain tungstoenzymes (1,3). To our knowledge, there is no report that V. cholerae and C. jejuni contain tungsten-dependent enzymes. However, both organisms can grow anaerobically and might express tungstoproteins under anaerobic conditions.