“ADP Sulfurylase” from Thiobacillus denitrificansIs an Adenylylsulfate:Phosphate Adenylyltransferase and Belongs to a New Family of Nucleotidyltransferases*

During AMP-dependent sulfite oxidation by some sulfur bacteria, the liberation of sulfate from adenosine-5′-phosphosulfate (APS) is catalyzed by APS:phosphate adenylyltransferase (APAT). Here we report the first biochemical and genetic characterization of APAT. We isolated this enzyme from the chemolithoautotroph Thiobacillus denitrificans and cloned the corresponding gene. The enzyme is homodimeric with 41,387-Da subunits and exhibits a specific activity of 2100 μmol min−1 mg−1. The K m values are K m (APS) = 300 μm andK m(Pi ) = 12 mm. Catalysis occurs by a ping-pong mechanism with a covalently bound AMP as reaction intermediate. The arsenolysis of APS , but not of ADP, CDP, GDP, UDP, or IDP, is also catalyzed, indicating a specific and unidirectional function. The former enzyme name ADP-sulfurylase implies that the reverse reaction is catalyzed; therefore, this name should not be used any longer. Histidine modification of APAT results in complete inactivation that can be suppressed by substrate addition. APAT is highly similar to galactose-1-phosphate uridylyltransferase and also related to Ap4A phosphorylase. Active site residues of galactose-1-phosphate uridylyltransferase are conserved in APAT and Ap4A phosphorylase, suggesting a histidine as the nucleotide-binding residue in all three enzymes, which together form a new family of nucleotidyltransferases.

Many bacteria are able to oxidize reduced sulfur compounds such as sulfide or thiosulfate to feed electrons into photosynthetic or respiratory electron transport (1,2). Two sulfite oxidation pathways may play a role in this dissimilatory oxidative sulfur metabolism as follows: (a) direct oxidation of sulfite to sulfate by sulfite:acceptor oxidoreductase (EC 1.8.2.1), and (b) indirect AMP-dependent oxidation of sulfite to sulfate via the intermediate APS. 1 In the latter pathway APS is formed from sulfite and AMP by the enzyme APS reductase (EC 1.8.99. 2) acting in reverse. Sulfate is released from APS in a second step either by ATP sulfurylase (EC 2.7.7.4) or by APS:phosphate adenylyltransferase (APAT) (3,4). The AMP moiety of APS is transferred either to pyrophosphate by ATP sulfurylase or to phosphate by APAT, resulting in the formation of ATP or ADP, respectively. Since ADP can be converted to ATP and AMP by adenylate kinase, both sulfate-liberating enzymes catalyze substrate phosphorylations that may be of energetic importance, especially in chemolithoautotrophic bacteria (5). Significant APAT activity has been reported from various chemotrophic and phototrophic sulfur bacteria (5)(6)(7)(8)(9)(10). However, in contrast to ATP-sulfurylase, APAT has never been purified from a bacterium, and a corresponding gene has never been cloned. This led some authors (11) to suggest that APAT does not exist as an enzymatic entity. Observed activities were tentatively explained by the occurrence of side reactions of other enzymes or by incorrect enzyme assays (11,12). To examine the existence and properties of APAT, we studied the activity from Thiobacillus denitrificans biochemically and genetically.

EXPERIMENTAL PROCEDURES
Bacterial Strains and Growth Conditions-Tb. denitrificans strain RT (DSMZ 807) was grown anaerobically on thiosulfate and nitrate as described elsewhere (13). Cells were harvested at late exponential growth phase and kept frozen at Ϫ70°C until use.
Synthesis of APS-APS was synthesized from AMP, sulfite, and ferricyanide using APS reductase activity of Tb. denitrificans crude extracts by a method essentially as described elsewhere (14) but modified to obtain salt-free APS after only one chromatographic step. APS synthesis was terminated by boiling for 10 min. Precipitated protein was removed by centrifugation (17,540 ϫ g, 4°C, 30 min). In 5-ml aliquots the supernatant was loaded onto a 95 ϫ 3 cm G-25 gel filtration column equilibrated with water (flow rate, 1 ml min Ϫ1 ). APS eluted after a yellow ferricyanide peak and before AMP. Fractions were tested by thin layer chromatography for purity (14). AMP-free fractions of APS were pooled. The APS concentration was determined photometrically (⑀ 259 nm ϭ 15.4 mM Ϫ1 cm Ϫ1 ) and enzymatically in APAT assays coupled with pyruvate kinase and lactate dehydrogenase (see below).
Activity Assays-APAT activity was measured in a coupled photometric assay system. The assay contained 50 mM Tris-HCl, pH 7.6, 50 mM potassium phosphate, pH 7.6, 400 M APS, 1 mM sodium phophoenolpyruvate, 200 M NADH, 2 mM MgCl 2 , 10 units of pyruvate kinase, 10 units of lactic acid dehydrogenase, and 10 l of the tested protein solution in a final volume of 0.5 ml. The activity was derived from the velocity of NADH oxidation (⑀ 366 nm ϭ 3.3 mM Ϫ1 cm Ϫ1 ). Since no interfering NADH oxidase activity was detectable, the reaction was started by addition of the protein tested. Start by addition of phosphate or APS gave identical velocities. For determination of the optimum pH, the mixture contained 100 mM Tris glycine buffer of the desired pH. K m values and V max for saturating second substrate concentrations were determined from initial velocities at varied substrate concentrations * This work was supported in part by the Deutsche Forschungsgemeinschaft Grant Da 351/1. 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) 148553.
Determination of Native and Subunit Molecular Weight-The native molecular weight was determined by gel filtration chromatography on a Superdex 200 column (calibration standards, Roche Molecular Biochemicals combithek and flavocytochrome c (16)). Purity and subunit molecular weight of the protein were determined by SDS-PAGE using the Mini Protean II-Cell (Bio-Rad) and the Laemmli method (17) with 12.5% T separating gels. Gels were Coomassie-stained. SDS-PAGE markers were obtained from Sigma.
Protein Modification-All incubations (100 l total volume) were carried out at room temperature. Bovine serum albumin (5 mg ml Ϫ1 ) was present to stabilize APAT. For APS protection assays 200 M APS was present. DEPC (2.5 mM) in 200 mM Bis-Tris, pH 6.0, was used to modify accessible histidines. The reaction was carried out at pH 6.0 for optimized His-specific modification (18). 4 mM 5,5Ј-dithiobis-(2-nitrobenzoic acid) (in 50 mM Na-EPPS, pH 8.0) and 10 mM iodoacetamide (in 50 mM Tris-HCl, pH 8.0) were used to modify accessible sulfhydryl groups (18,19). 10 mM phenylglyoxal (in 100 mM bicarbonate buffer pH 8.0) and 10 mM diacetyl (in 100 mM borate buffer pH 9.0) were used to modify accessible arginines (18). In all modification experiments 20 l of the solution were assayed after desired times for remaining APAT activity, using the photometric assay system.
Mass Spectrometry-1 ml of purified APAT was concentrated immediately after the gel filtration step to a final volume of 50 l (Amicon Centricon system, 10-kDa cut off). This concentration step stabilized enzymatic activity and allowed detection of the enzyme by matrixassisted laser desorption/ionization time of flight-mass spectroscopy (MALDI-TOF-MS). MALDI-TOF-MS was carried out using a Voyager RP Workstation (Perkin-Elmer). The sample (ϳ1-3 M) was 10-fold diluted with 0.1% trifluoroacetic acid, thereafter 1:1 mixed with 110 mM sinapinic acid (3,5-dimethoxy-4-hydroxycinnamic acid) in 0.1% trifluoroacetic acid, 67% acetonitrile, and air-dried. Measurements were carried out in a linear mode. Further parameters are given in the legend of the corresponding figure. For detection of the enzyme-AMPintermediate the sample was incubated with 17 M APS for 5 min at room temperature and treated as described above.
Genetic Methods-Standard methods were used for molecular techniques (20). Escherichia coli DH5␣ was used for all cloning steps. From the N-terminal amino acid sequence information obtained by automated Edman degradation (SDNQSAPNEIREIRINPIVP) degenerate primers were derived and used for polymerase chain reaction to amplify a 5Ј-region of the gene. The following primer combination gave a product of the expected size (53 base pairs): forward primer, 5Ј-ATG TC(GC) GA(CT) AA(CT) CA(GA) TC-3Ј; reverse primer, 5Ј-GG(GA) TT(GA) AT(GA) CG(GA) AT(TC) TC-3Ј. The product was cloned in pGEM-T (Promega) and sequenced using a silver sequencing kit (Promega). The sequence coded for the expected residues of the N terminus. The cloned fragment served as template for the polymerase chain reaction generation of a digoxygenin-labeled gene probe using again the degenerate oligonucleotides. A restriction map was constructed by Southern blot analysis (60°C hybridization and washing temperature), and a 2.97kilobase pair EcoRI-SalI fragment was isolated from a partial Tb. denitrificans genebank of 2.4 -3.2-kilobase pair EcoRI-SalI fragments in pBluescript SK II(ϩ). The correct clone (pTDA1) was identified by restriction analysis and verified by activity determination in the heterologous system. The complete fragment was sequenced on one strand, the gene for APAT (apt) was identified, and both strands of it were sequenced. Sequencing of pTDA1 was done commercially with Applied Biosystems sequencers and the dye terminator method, using either standard T7 or T3 promoter primers or sequence derived primers. The sequence of apt was deposited at GenBank TM (accession number AF148553).
Chemicals-All chemicals were obtained from Sigma, Aldrich, Fluka, or Roche Molecular Biochemicals and were of highest grade. For arsenolysis experiments, ADP was further purified by anion exchange, to separate it from AMP, which is present in commercially available ADP in significant amounts. Yeast galactose-1-phosphate uridylyltransferase was obtained from Sigma.

RESULTS
Purification Results and Some Molecular Properties-The purification data are summarized in Table I. The gel filtration purification step indicated a native molecular mass of 82 Ϯ 8 kDa. SDS-PAGE analysis of fractions after gel filtration showed a single band at 41 Ϯ 1 kDa which correlated in its intensity exactly with the elution peak absorption and APAT activity and therefore was identified as APAT ( Fig. 1). Native and subunit molecular weights strongly suggest a homodimeric structure of APAT. The purified enzyme exhibited a specific activity of 2100 units mg Ϫ1 under standard assay conditions. All further kinetic analysis and all experiments on the reactivity of the enzyme under various conditions were done using purified APAT which was stabilized with 5 mg ml Ϫ1 bovine serum albumin.
Kinetic Analysis-The optimum pH for catalysis was at pH 8.5-9.0. At higher pH some precipitation occurred. To ensure high activity of APAT and the coupled enzymes, all further kinetic analyses were carried out at pH 8.0. Variation of one substrate concentration at several fixed second substrate concentrations resulted in activities that gave a set of parallel lines in Lineweaver-Burk plots (Fig. 2). This finding strongly indicated a Ping Pong Bi Bi reaction mechanism that involves the formation of a stable enzyme-bound reaction intermediate from reaction with the first substrate (APS) before binding of the second substrate (P i ). The K m values are K m(APS) ϭ 300 M and K m(P i ) ϭ 12 mM at saturated second substrate concentrations. Due to the ping-pong mechanism significantly lower apparent K m values are observed at unsaturated second substrate concentrations. The theoretical V max of the reaction can be estimated to be at 3850 units mg Ϫ1 at optimum pH. Divalent cations in the assay (1 mM MgCl 2 or 1 mM MnCl 2 ) did not affect activity of the enzyme. A 1-h preincubation of the enzyme with trace metals had no or slightly negative effects on activity. The following metals (1 M) were tested (remaining activity): CoCl 2 (98%), CuCl 2 (100%), FeCl 2 (86%), MnCl 2 (96%), NiCl 2 (95%), or ZnCl 2 (89%).
Identification of the Reaction Intermediate-The purified enzyme was subjected to MALDI mass spectrometry. The enzyme molecular mass was determined with this method to be 41,376 Ϯ 20 Da (Fig. 3). After incubation of the enzyme with APS (see "Experimental Procedures"), the main mass peak decreased in parallel to an increase of a new peak of 330 Ϯ 2 Da higher mass (Fig. 3). Since a 329-Da mass difference is expected for covalently bound AMP, the new peak observed after APS addition can be attributed to a covalent enzyme-AMP complex. The bond must be formed between an enzyme nucleophile and the phosphorus atom of AMP, thereby releasing one water molecule, which has to be subtracted in the mass calculation.  . Data points in A and C correspond to average values of two measurements. All assays were carried out at 30°C and started by addition of 10 -100 ng of APAT after 3 min of preincubation of the assay mixture in the cuvette. The activity was calculated from 20-to 120-s time intervals. The ordinate of the secondary plots (B and D) is crossed at the reciprocal of V max at second substrate saturation, and the slope corresponds to K m /V max .
Reversibility of the Catalysis-In order to determine the reactivity of the enzyme with the reaction product ADP and with other NDPs, we carried out arsenolysis experiments. Na 2 HAsO 4 is a phosphate analogue that should be able to react with the putative enzyme-AMP intermediate. The resulting phosphate-arsenate anhydride is chemically instable and immediately hydrolyzes to AMP and arsenate. This AMP formation can be detected by thin layer chromatography. In a control experiment APS rapidly underwent arsenolysis, indicating the functionality of the assay system (Fig. 4). To our surprise, no reaction could be observed with ADP (Fig. 4). Other NDPs (GDP, UDP, CDP, and IDP) also did not react (data not shown). Even after 24 h of incubation arsenolysis was not detectable, although the enzyme still was active (tested by arsenolysis of APS). Obviously the enzyme-AMP complex cannot be formed from NDP substrates with significant rates. Although a back reaction should be possible in any enzyme-catalyzed reaction, its rate must be infinitely slow in the case of APAT so that the catalysis can be regarded as unidirectional. Therefore the former name "ADP-sulfurylase" is misleading and should not be used any longer.
Although an enzyme-AMP intermediate formation with ADP as a substrate was not detectable, ADP inhibited the reaction with APS. By thin layer chromatographic analysis, we found that half-maximum inhibition in assays containing 5 mM P i and 400 M APS was reached at 5-10 mM ADP. Traces of AMP were present in the ADP inhibition assays (commercially available ADP had to be used without further purification to reach the concentrations required for inhibition). AMP alone has a comparatively low effect on enzyme activity (21). Sulfate inhibits APS phosphorolysis at higher concentrations. Half-maximum inhibition in assays containing 5 mM P i and 400 M APS was observed at 100 mM sulfate.
The first half-reaction, the formation of an enzyme-AMP intermediate by reaction with APS, could in principle be reversible. Therefore we used the sulfate analogue molybdate instead of phosphate and examined molybdolysis with APS. Since the generated AMP-molybdate anhydride is unstable, the product rapidly hydrolyses into AMP and molybdate, and therefore the reaction product is taken out of the equilibrium continuously. No molybdolysis was observed. This can be taken as evidence against the possibility of a reaction of sulfate with the enzyme-AMP intermediate.
Enzyme Modification-To detect residues essential for catalysis, we carried out modification experiments with agents specific for cysteine, histidine, and arginine. The histidine-specific agent DEPC (2.5 mM) resulted in a rapid and complete loss of enzyme activity. Addition of APS slowed down the inactivation kinetics and protected the enzyme from complete inactivation (Fig. 5A). However, some decrease in activity was observed even when the enzyme was protected by bound substrate. The reason may be that DEPC slowly modifies additional unprotected histidines that are required for full activity. Another possibility is that the enzyme-AMP complex is not completely stable, and therefore DEPC slowly modifies even protected enzyme. This slow inactivation of the protected enzyme may be incomplete because DEPC is unstable in long term modifications. This could explain the nonlinear inactivation kinetics of APS-protected APAT as observed in the semilogarithmic plot (Fig. 5A). Whatever the reason for the slow decrease of activity, it is obvious that substrate binding protects the enzyme from rapid inactivation. Therefore it is most likely that at least one histidine essential for catalytic activity is present in the catalytic center. Incubation with the cysteine-specific agents 5,5Јdithiobis-(2-nitrobenzoic acid) (4 mM) or iodoacetamide (10 mM) had no significant effect on activity. We conclude that accessible cysteine residues with a catalytic function do not exist in APAT. Arginines were modified using phenylglyoxal (10 mM) or diacetyl (10 mM). Phenylglyoxal resulted in an incomplete inactivation to 55% residual activity which then remained completely stable. Diacetyl inactivated the enzyme slowly but completely (Fig. 5B). Since diacetyl is even more specific for arginine residues than phenylglyoxal, the inactivation indicates an essential role of arginine(s) for APAT. A slight decrease of activity was observed in borate buffer without modifying agent (Fig. 5B). APAT activity was not affected by any other buffer used. The inactivation by diacetyl was slowed down by substrate binding (200 M APS). The substrate binding effect was much weaker in the case of diacetyl modification compared with DEPC modification (Fig. 5). Since phenylglyoxal results in 55% residual activity, this comparatively bulky agent modifies less residues than diacetyl. It can be concluded that either arginines are involved in substrate binding or they are crucial for enzyme structure. If accessible and structurally essential arginines are located outside the catalytic site, substrate binding must induce conformational changes of the enzyme that cause slower modification kinetics.
Genetic Analysis-The determination of a sequence of 20 N-terminal amino acids allowed cloning and sequencing of the gene coding for APAT (apt) as described under "Experimental Procedures." A good matching ribosomal binding site (AAG-GAG) was identified in a six-nucleotide distance from the translational start codon. 281 C-terminal codons of an open reading frame were identified upstream of apt, which code for a protein with very high homology to the E. coli Ras-like protein (Era, BLAST P E value, 2 ϫ 10 Ϫ79 , release 2.0.6, see Ref. 22). Era is an essential G-protein that probably is involved in regulation of growth (23,24). The distance between era and apt (775 base pairs) and the existence of several AT-rich stretches in this region strongly suggest that apt and era transcription are not coupled. Various putative promoters are found upstream of apt. Therefore detailed genetic analyses are necessary to identify the functional promoter region. Downstream of apt a Rho-independent termination signal could not be identified in the remaining sequence of the cloned fragment (238 base pairs). It therefore cannot yet be excluded that downstream genes are transcriptionally coupled with apt. Further genetic analysis is in progress. Apt codes for a 41,387-Da protein of 370 residues (without the N-terminal methionine, which is cleaved off). This mass is in agreement with SDS-PAGE and MALDI results (see above). The calculated isoelectric point of the protein is at pH 6.0 (ProtParam, ExPASy home page).
Related Enzymes and Substrate Specificity-Available gene banks were screened for proteins with similarity to APAT (BLAST P, release 2.0.7). The sequence of APAT shows high homology to galactose-1-phosphate uridylyltransferases (GPUT, EC 2.7.7.12) from bacteria, archaea, fungi, plants, and animals (best E value: 3 ϫ 10 Ϫ27 for Thermotoga neapolitana, E value for E. coli-GPUT: 5 ϫ 10 Ϫ15 ). GPUT catalyzes the UMP transfer from UDP-glucose to galactose 1-phosphate. Lower homology exists to all known Ap 4 A phosphorylases (EC 2.7.7.-) that catalyze the NMP transfer from bis(5Ј-nucleosidyl)tetraphosphate dinucleotides to phosphate (see Ref. 25; best E value, 0.32 for Saccharomyces cerevisiae Ap 4 A phosphorylase 1). All three related enzymes have in common catalysis of NMP transfers to phosphate groups. Catalysis by Ap 4 A phosphorylase is not very specific; the yeast enzyme reversibly catalyzes phosphorolysis of various dinucleotides and in unphysiological reactions also a ␤-phosphate exchange of NDPs, the arsenolysis of NDPs, and very interestingly the arsenolysis and phosphorolysis of APS (25,26). For that reason we tested the ability of APAT to catalyze reactions with Ap 4 A. No catalysis was found with 10 mM Ap 4 A neither in a phosphorolysis nor in a hydrolysis reaction. Addition of divalent cations (1 mM MgCl 2 or CaCl 2 , 1 M of CoCl 2 , CuCl 2 , FeCl 2 , MnCl 2 , NiCl 2 , or ZnCl 2 ) did not result in any qualitatively detectable reactivity. We also tested the reactivity of GPUT from yeast (Sigma) and found that it neither reacted with Ap 4 A nor with APS. A summary of the activities of the three related enzymes is given in Table II. GPUT from E. coli is known to be a homodimeric 80-kDa metalloenzyme, containing one Zn 2ϩ and one iron atom (possibly Fe 3ϩ ) per subunit. Interestingly all four Zn 2ϩ ligands of GPUT (Cys-52, Cys-55, His-115, and His-164) are conserved in the sequence of APAT (Cys-53, Cys-56, His-120, and His-172), strongly suggesting the presence of Zn 2ϩ in APAT (Fig. 6A, see Refs. 27 and 28). Two of the iron ligands from E. coli GPUT (Glu-182, His-281) are also present in APAT (Glu-190 and His-310, see Fig. 6A). However, the other GPUT iron ligands (His-296 and His-298) cannot be clearly assigned (two of His-326, His-331, or His-333). Therefore iron might be not present in APAT. Interestingly, the iron ligands are not strictly conserved in some other GPUT sequences (e.g. from Homo sapiens, Caenorhabditis elegans, or Streptomyces lividans, CLUSTAL W analysis, Ref. 29), and the iron site is not near or part of the active center. For that reason, in these enzymes the iron site may not generally play an as important role as the Zn 2ϩ site which is positioned near the active site (see below, Ref. 28).
Identification of the Active Site-Only a short stretch from Val-155 to Ala-179 (numbering of APAT) was found to be conserved in all GPUT, Ap 4 A phosphorylase, and APAT sequences (Fig. 6B). This region has previously been identified in the E. coli GPUT to be that part of the active center that contains the UMP-binding histidine (30). A histidine is also present in the corresponding position of Tb. denitrificans APAT (His-174), rendering it very likely that during catalysis AMP is covalently bound to APAT via this residue (Fig. 6B). This idea is strongly supported by the results of histidine modification with DEPC (see above). Other residues that are strictly conserved in all sequences also may play important roles in nucleotide binding (in analogy to GPUT,Ref. 30). The carbonyl group of His-172 (E. coli His-164) probably interacts with the His-174 (E. coli His-166) imidazole during catalysis, and it can be postulated that Asn-161 and Gln-176 (E. coli Asn-153 and Gln-168) bind to the ␣-phosphate (Fig. 6B). The structure is probably stabilized

FIG. 5. Semilogarithmic plots of APAT inactivation by 2.5 mM DEPC (A) and 10 mM diacetyl (B) and protection by 200 M APS.
Diamonds, APAT ϩ modifying agent; squares, APAT ϩ modifying agent ϩ APS; circles, APAT without modifying agent (stability control). Activities were measured after indicated incubation times. and correctly oriented by the binding of Zn 2ϩ , because His-172 is one of the Zn 2ϩ ligands. As expected, most of the residues that are involved in glucose or galactose binding of GPUT are not found in APAT. An x-ray analysis of GPUT with bound UDP-glucose or UDP-galactose allows such assignments (31). Only Glu-317 finds its counterpart in APAT (Glu-347), whereas Lys-311, Phe-312, Val-314, Tyr-316, and Gln-323 of GPUT are lacking at the corresponding positions in APAT (31). Obviously APAT contains the residues important for covalent NMP binding (see Fig. 6B) but lacks residues that are needed to bind an additional hexose. This confirms the expected substrate specificity of APAT. Interesting is the finding that there is no cysteine within the conserved active site region in APAT. This is very much in agreement with the result that cysteine modification has no effect on activity. In contrast, in yeast Ap 4 A phosphorylase there are some cysteines whose modification may have negative influence on activity (APA 1: Cys-129 and Cys-176; APA 2: Cys-149 and Cys-179). This could explain the measured inactivation of this enzyme by the thiol-modifying agent p-chloromercuribenzoic acid (32). It is known from other enzymes that p-chloromercuribenzoic acid modification can lead to steric inhibition even when nonessential residues are modified (33). The Cys-160 of the active center from E. coli GPUT is not conserved among GPUT sequences and does not occur in the known sequences of Ap 4 A phosphorylases and APAT.
Beside the homologies mentioned above, a conspicious stretch of eight identical residues exists near the C terminus of APAT (Glu-360 to Arg-367) and E. coli GPUT (Glu-329 to Arg-336). These residues form an ␣-helix in GPUT which is neither close to the active center nor to the subunit interface (Rutgers University Protein Data Bank code 1HXQ, Ref. 30). The function of this conserved structure is not clear yet. DISCUSSION High APAT activity was present in crude extracts of Tb. denitrificans strain RT, and the responsible enzyme could be purified to homogeneity to allow further characterization and cloning of the corresponding gene. Results obtained from kinetic analysis, arsenolysis, molybdolysis, and MALDI experiments suggest the following overall reaction scheme.
where S i indicates sulfate. Phosphorolysis of the enzyme-bound intermediate is irreversible since the intermediate cannot be formed from ADP. APAT exhibits high sequence similarity to GPUT, which is also a homodimer of a very similar size (28). GPUT introduces galactose-phosphate into the sugar metabolism by transferring an UMP to its phosphate group. The product UDP-galactose can be epimerized to UDP-glucose which can be further metabolized. The reaction mechanism of both enzymes involves the formation of a covalent bond between a conserved active center histidine and a nucleotide. For GPUT this has been shown by x-ray analysis (30). We propose the same mechanism for APAT based on several lines of evidence. 1) The ping-pong kinetics of APAT suggests a tightly enzyme bound reaction intermediate (Fig. 2). 2) MALDI measurements are in agreement with the formation of a covalent enzyme-AMP intermediate (Fig. 3). 3) Histidine modification resulted in a complete loss of activity that was prevented by substrate addition (Fig. 5). 4) Sequence comparisons strongly suggest that the active site of APAT resembles that of GPUT, especially in a highly conserved region that contains the active histidine and some other intermediate stabilizing residues (Fig. 6).
In former modification studies with crude APAT preparations from other sources, only cysteine-modifying agents were tested (7). Complete inactivation was not achieved by those agents, and it has to be considered that N-ethylmaleimide, which resulted in highest inhibition (80%), also modifies histidines with a slower rate (34). However, it cannot be excluded that different types of APAT exist in organisms other than Tb. denitrificans and that cysteines may play more important roles in these enzymes.
Based on sequence similarities, the reaction mechanism of APAT and GPUT also has to be postulated for Ap 4 A phosphorylase. In agreement with this proposal, ping-pong kinetics have been observed for APS phosphorolysis of Ap 4 A phosphorylase (35). Ap 4 A phosphorylase from yeast is the only other purified enzyme with APAT activity (26) and catalyzes this reaction as an unphysiological side reaction which probably occurs because of the similarity of Ap 4 A phosphorylase to authentic APAT. In yeast there are two isoforms of Ap 4 A phosphorylase. It was shown by studies on mutants that Ap 4 A phosphorylase 1 is responsible for 85% of the APAT activity in this organism (36). Ap 4 A phosphorylase 2 is suggested to cause the remaining 15% activity (36). This is supported by the higher similarity of bacterial APAT to type 1 Ap 4 A phosphorylase (Blast P E value, 0.35) compared with Ap 4 A phosphorylase 2 (Blast P E value 8.1). Ap 4 A phosphorylase probably does not need to be more specific, because in yeast APS is formed only under sulfate assimilation conditions by a highly regulated ATP-sulfurylase (37). In addition, a substrate shuttle mechanism may transfer APS from ATP-sulfurylase to APS kinase, similar to the case of bifunctional sulfate-activating enzymes in higher eukaryotes (38). It is therefore unlikely that APS serves as a substrate for Ap 4 A phosphorylase in yeast. Interestingly, Ap 4 A phosphorylase catalyzes the arsenolysis of both ADP and APS (26). For that reason the reversibility of the APAT reaction by Ap 4 A phosphorylase has to be considered as possible, although energetically unfavorable. However, the enzyme-AMP complex of Ap 4 A phosphorylase might not react with sulfate, and in this case the APAT reaction of Ap 4 A phosphorylase may also be irreversible. Very interesting is the observation that Ap 4 A phosphorylase, similar to APAT, is sensitive against arginine-modifying agents (39). In both cases substrate binding has some effect on inactivation kinetics. However, it seems possible that the inactivation by arginine modification may not be due to a functional role of arginine in the catalytic center.
The three enzymes Ap 4 A phosphorylase, GPUT, and APAT belong to a new family of nucleotidyltransferases. For evaluation of former and future studies, it has to be considered that low specific activities of APAT in crude extracts (below ϳ100 milliunits mg Ϫ1 ) can be due to Ap 4 A phosphorylase or other enzymes of this family and therefore should not be overinterpreted. Higher activities (above ϳ100 milliunits mg Ϫ1 ), which have up to now only been detected in Thiocapsa roseopersicina (4), Thiobacillus thioparus (5), and Tb. denitrificans (this study), are likely to be due to a specific APAT.
Within the last years the arrangement of APAT into a cytoplasmic sulfite oxidation pathway was generally doubted (11,12). Tb. denitrificans is an organism for which it has been postulated that APAT functions in a pathway by which sulfite produced by a cytoplasmic reverse siroheme sulfite reductase is further oxidized to sulfate (40). This pathway involves APS reductase, which produces APS from sulfite and AMP (41). The formation of APS allows coupling of sulfate liberation to a substrate phosphorylation step. We are convinced that Thiobacillus APAT is involved in this sulfate liberation for various reasons as follows: 1) no other substrate than APS is known for Thiobacillus APAT; 2) the reverse reaction is not catalyzed; 3) the enzyme is present with high activities; 4) the enzyme is present in the same compartment as APS and therefore the reaction must take place if APS is not protected somehow (which is very unlikely at high turnover rates).
Interestingly, ATP-sulfurylase exists in the same organism during growth on reduced sulfur compounds. 2 All other known bacteria with high APAT activity also simultaneously contain high ATP-sulfurylase activity (Tc. roseopersicina see Ref. 10, Tb. thioparus see Ref. 5). We suggest the following model which may explain this finding at least for chemotrophs: ATP-sulfurylase activity is limited by pyrophosphate availability, since pyrophosphate is hydrolyzed by pyrophosphatase and used up by ATP-sulfurylase. APAT activity may allow a higher APS turnover and could thereby prevent the accumulation of toxic sulfite in the cytoplasm. Since ATP-sulfurylase is known to have a very high affinity for APS (K m(APS) ϭ 6.6 M, see Ref. 11) and pyrophosphate (K m(PP i ) ϭ 14 M, see Ref. 11), this enzyme is efficiently employed even in the presence of APAT as long as pyrophosphate is available. Note that for APAT the apparent K m values are higher (the K m(APS) is ϳ55 M with 2 mM P i and the K m(P i ) is ϳ1.7 mM with 80 M APS). APAT therefore could prevent accumulation of APS without interfering with ATPsulfurylase. By this mechanism a maximum energy conservation by both enzymes, ATP-sulfurylase and APAT, is guaranteed which could be of importance for chemolithoautotrophic growth of Tb. denitrificans.
Not only in Tb. denitrificans, but also in other sulfur compound oxidizing chemotrophic and phototrophic bacteria, sulfite may be formed in the cytoplasm. In those cases two scenarios for sulfite oxidation are possible (Fig. 7) as follows: sulfite can be either oxidized in the cytoplasm via the APS pathway or extruded into the periplasmic space where it is oxidized by sulfite:acceptor oxidoreductase (Fig. 7, A-C). Sulfite:acceptor oxidoreductases from various thiobacilli are known to be c-type cytochromes and therefore a periplasmic localization can be assumed (42). When the cytoplasmic APS pathway is employed, toxic sulfite accumulation in the cytoplasm may be prevented by APAT as rationalized above. We propose a mixture of both scenarios for those sulfur compound 2 T. Brü ser, unpublished results. oxidizing organisms that contain APS reductase and ATP sulfurylase but no APAT; in those cases accumulating sulfite could be extruded to the periplasm when ATP-sulfurylase is limited by low pyrophosphate availability (Fig. 7B).
After the herein reported biochemical and genetic analyses of Thiobacillus APAT, future studies will have to concentrate on the relation to other APATs and on the physiological role of APATs. The proposal of the various sulfite oxidation pathways will hopefully support such studies.