The Mechanism of Nucleotide-assisted Molybdenum Insertion into Molybdopterin

The molybdenum cofactor (Moco) is synthesized by an ancient and conserved biosynthetic pathway. In plants, the two-domain protein Cnx1 catalyzes the insertion of molybdenum into molybdopterin (MPT), a metal-free phosphorylated pyranopterin carrying an ene-dithiolate. Recently, we identified a novel biosynthetic intermediate, adenylated molybdopterin (MPT-AMP), which is synthesized by the C-terminal G domain of Cnx1. Here, we show that MPT-AMP and molybdate bind in an equimolar and cooperative way to the other N-terminal E domain (Cnx1E). Tungstate and sulfate compete for molybdate, which demonstrates the presence of an anion-binding site for molybdate. Cnx1E catalyzes the Zn2+-/Mg2+-dependent hydrolysis of MPT-AMP but only when molybdate is bound as co-substrate. MPT-AMP hydrolysis resulted in stoichiometric release of Moco that was quantitatively incorporated into plant apo-sulfite oxidase. Upon Moco formation AMP is release as second product of the reaction. When comparing MPT-AMP hydrolysis with the formation of Moco and AMP a 1.5-fold difference in reaction rates were observed. Together with the strict dependence of the reaction on molybdate the formation of adenylated molybdate as reaction intermediate in the nucleotide-assisted metal transfer reaction to molybdopterin is proposed.

Molybdenum forms the active center in all molybdenum enzymes (1), catalyzing key metabolic reactions in the global sulfur, nitrogen, and carbon cycles in organisms ranging from bacteria to human. With the exception of nitrogenase, in all other molybdenum enzymes molybdenum is activated and chelated to the molybdenum cofactor (Moco) 3 (2) consisting of molybdenum covalently bound to the dithiolate moiety of a tricyclic pterin (molybdopterin; MPT), whose structure is conserved in eukaryotes, eubacteria, and archaea (3). Molybdenum enzymes are essential for diverse metabolic processes such as nitrate assimilation in autotrophs and phytohormone synthesis in plants (4) or sulfur detoxification and purine catabolism in mammals (5). Human Moco defi-ciency is a severe hereditary metabolic disorder for which recently a first substitution treatment has been described (6).
Moco is synthesized by a conserved biosynthetic pathway that can be divided into four main steps (2,7), according to the biosynthetic intermediates cyclic pyranopterin monophosphate, MPT, and adenylated MPT. At least six gene products catalyzing Moco biosynthesis have been identified in humans (8), plants (7), and bacteria (2). In the final and most diverse steps of Moco biosynthesis, a single molybdenum atom is ligated to one or two MPT dithiolates (3). Furthermore, in bacteria nucleotides are attached via a pyrophosphate bond to MPT forming the so-called MPT dinucleotide cofactors that are found either in mono-MPT (9) or bis-MPT containing enzymes (10). After completion of biosynthesis mature Moco has to be inserted into molybdenum enzymes. In prokaryotes a complex of proteins synthesizing the last step(s) of Moco biosynthesis is proposed to donate the mature cofactor to the appropriate apo-enzymes (11) assisted by enzyme-specific chaperones (12). In eukaryotes, a Moco carrier protein has been been described in Chlamydomonas rheinhardtii (13).
Eukaryotic molybdenum insertion is catalyzed by Cnx1 in plants (14) and gephyrin in mammals (15). Both proteins exhibit two conserved functional domains (E and G) homologous to the Escherichia coli proteins MoeA (16) and MogA (17), respectively. Cnx1 G domain (Cnx1G) binds MPT with high affinity (18) and its active site was mapped structurally to a large surface depression with a clear discrimination between substrate binding and catalysis (19) (20,21). Structures with bound MPT and adenlyated MPT (MPT-AMP), a recently identified intermediate of Moco biosynthesis, have uncovered the function of Cnx1G (22). Cnx1G catalyzes the adenylation of MPT in a Mg 2ϩ -and ATP-dependent manner (23) (Fig. 1A). ATP hydrolysis and pyrophosphate release were dependent on bound MPT. Catalytically inactive Cnx1G variants were found to be defective in MPT adenylation and therefore we named Cnx1G as MPT adenylyl transferase (23). An unexpected observation in the MPT-bound Cnx1G structures was the identification of copper bound to the MPT dithiolate sulfurs. Copper might play a role in sulfur transfer to cyclic pyranopterin monophosphate, in protecting the MPT dithiolate from oxidation and/or presenting a suitable leaving group for molybdenum insertion (22).
As we found a Mg 2ϩ and Cnx1E dependence of in vitro Moco synthesis, a function of Cnx1E in MPT-AMP hydrolysis and molybdenum insertion has been proposed (22). Here, we show that MPT-AMP is hydrolyzed by Cnx1E. Both, molybdate and MPT-AMP bind in a cooperative manner to Cnx1E. Molybdate is essential for the Zn 2ϩ /Mg 2ϩdependent MPT-AMP hydrolysis, which is competitively inhibited by tungstate and sulfate. Upon MPT-AMP hydrolysis Moco is formed and quantitatively transferred to apo-plant sulfite oxidase (apo-PSO). AMP and copper are released as additional products of the Cnx1E reaction.
Our data demonstrated a nucleotide-dependent molybdenum insertion reaction using equimolar amounts of substrates. The formation of adenylated molybdate as reaction intermediate is proposed.

EXPERIMENTAL PROCEDURES
Expression and Purification of Cnx1G, Cnx1E, and Apo-SO-Cnx1G and variants D515H and S583A were expressed from pQE60Cnx1g628 in strain SE1581 (moeA Ϫ ) for 36 h at 22°C in LB medium containing 10 M isopropyl ␤-thiogalactoside. Cnx1G-MPT-AMP complex and C-terminally His-tagged Cnx1E were purified as described (18,22). Apo-PSO was expressed in BL21 using plasmid pQE80At-sox (24). Expression was induced with 0.5 mM isopropyl ␤-thiogalactoside at A 600 of 0.5 and continued for 18 h at 25°C. Harvested cells were extracted and purified using 5 ml nickel nitrilotriacetic acid superflow matrix per 1-liter culture under denaturating conditions as recommended (Qiagen). After washing, column-bound apo-PSO was refolded by a stepwise gradient against refolding buffer (100 mM Tris/HCl, 300 mM NaCl, 10 mM dithiothreitol, pH 8.0) and subsequently eluted with refolding buffer containing 500 mM imidazol, pH 8.0. Eluted pure protein was desalted against PSO buffer (20 mM Tris acetate, pH 8.0), concentrated to 2 mg/ml, and frozen in 25-l aliquots in liquid nitrogen.
Size Exclusion Chromatography-Size exclusion chromatography was performed using a HR10/30 Superdex 200 column (Amersham Biosciences) equilibrated in 20 mM Tris/HCl, 250 mM NaCl, pH 8.0. Proteins (15-30 M) were loaded in a total volume of 500 l and separated with 0.2 ml/min flow rate. Fractions of 500 l were collected, pooled, concentrated, and MPT as well as MPT-AMP content was determined by HPLC FormA or FormA-AMP analysis, respectively.
MPT-AMP Hydrolysis-Cleavage of MPT-AMP by Cnx1E was performed aerobically in two different ways, either in a mixture of Cnx1Gbound MPT-AMP and Cnx1E or with purified Cnx1E-MPT-AMP complex. In the first case 20 M Cnx1G-bound MPT-AMP, 10 M Cnx1E, and if indicated 20 M sodium molybdate were incubated in a final volume of 200 l of 100 mM Tris/HCl, pH 7.2 with 1 mM MgCl 2 . At different times aliquots of 20 l were taken and MPT-AMP as well as MPT contents were determined by HPLC FormA-AMP or FormA analysis, respectively. In the other case, purified Cnx1E-MPT-AMP complex was prepared by transferring MPT-AMP from Cnx1G to Cnx1E. In 500 l total volume 30 M Cnx1G-bound MPT-AMP, 15 M Cnx1E, and 30 M sodium molybdate were co-incubated in 100 mM Tris/HCl, pH 7.2 for 1 h at room temperature under anaerobic conditions before purification by size exclusion chromatography as described above. MPT-AMP hydrolysis was performed in 20 l total volume containing 10 M Cnx1E-MPT-AMP complex (200 pmol of MPT-AMP). The reaction was started by adding the indicated divalent cations and either stopped by the addition of 390 l 100 mM Tris/HCl, pH 7.2, and 50 l of 1% I 2 , 2% KI, 1 M HCl, or 1 mM EDTA.
Determination of MPT and AMP-MPT-MPT was detected by HPLC FormA analysis as described (18). To separate MPT and MPT-AMP oxidation was performed for 2 h at room temperature. By this procedure MPT-AMP was converted into FormA-AMP that was separated from FormA by anion exchange chromatography and subsequently treated with pyrophosphatase and alkaline phosphatase (Sigma) thus yielding FormA dephospho, which was quantified as described (18).
Molybdate Binding Assay-Quantitative molybdate binding to Cnx1E was performed in a final volume of 200 l of 100 mM Tris/HCl, pH 7.2, containing 12 M sodium molybdate. Protein-bound and unbound molybdate was separated using NICK desalting columns (Amersham Biosciences) equilibrated in 100 mM Tris/HCl, pH 7.2.
Molybdate was determined in the salt fraction (1 ml) by a modified colorimetric method (25). Samples were incubated with 100 l of concentrated sulfuric acid for 10 min before adding the toluene 3,4-dithiolate reagent (Sigma) (25). The formed non-soluble dithiolate complex was separated by centrifugation for 10 min at 15,000 ϫ g, and the resulting green pellet was re-suspended in 200 l of isoamyl acetate. The amount of molybdate was determined spectrophotmetrically according to a standard curve by the absorption difference between 550 and 680 nM wavelength.
Inductively Plasma-coupled Mass Spectrometry (ICP-MS)-For analysis of metal concentrations, samples were prepared by HNO 3 closedvessel microwave digestion and diluted in water to a final concentration of 6.5% HNO 3 for analysis by ICP-MS. ICP-MS was performed by using a HP4500 Series 300 ShieldTorch system instrument (Agilent, Waldbronn, Germany) in peak-hopping mode with spacing at 0.05 atomic mass unit, three points/peak, three scans/replicate, two replicate/sample, and an integration time of 500 ms per point as described (26). The rate of plasma flow was 15.8 liter/min with an auxiliary flow of 1.0 liter/min and a blend gas flow rate of 0.1 liter/min. The RF power was 1.21 killowatts. The sample was introduced by using a cross-flow nebulizer at a flow rate of 1.02 liter/min. The apparatus was calibrated by using a 6.5% HNO 3 solution containing copper and molybdenum at 1, 5, 10, 25, 50, 100, and 200 parts per billion with Rh-103 as internal standard for all isotopes of copper and molybdenum.
Apo-PSO Recontistution Assay-Transfer of Moco from Cnx1E to apo-PSO was performed in 75 l reaction volume containing up to 0 -20 M Cnx1E-MPT-AMP complex, 10 M apo-SO, and either 500 M ZnCl 2 or 100 M MgCl 2 and incubated the indicated time. After reconstitution 75 l of PSO buffer containing 100 M ferricyanide was added. The assay was started with 150 l of PSO buffer containing 25 M sulfite, and PSO activity was determined by monitoring the reduction of ferricyanide at 420 nm using a 96-well plate reader (Versa Max, Molecular Devices). For each reaction a control without sulfite was performed and resulting rates were subtracted from the sulfite-dependent activity. 1 unit of PSO activity corresponds to a change in A 420 of 1 milliabsorbance unit/min. AMP Determination-AMP was detected by an HPLC-based method (27) with minor modifications. AMP-containing samples were loaded onto a C18 reversed phased column (Hypersil ODS, 5 M, 250 ϫ 4.6 mM, Techlab) equilibrated in 6 mM tetrabutyamoniumhydroxide (Fluka), 10% methanol at 1 ml/min flow rate. After injection, AMP was eluted with a 13-ml gradient of methanol between 35 and 85% and detected at A 260 . To monitor the release of AMP upon MPT-AMP hydrolysis 10 M Cnx1E-MPT-AMP complex (200 l volume) was incubated for different times, stopped by the addition of 1 mM EDTA, and injected onto the HPLC.

Isolation of MPT-AMP by Purifying the Cnx1G-Product Complex-
To obtain sufficient amounts of MPT-AMP for subsequent mechanistic studies we performed MPT-AMP co-purifications with different Cnx1G variants. We expressed and purified wild-type Cnx1G domain as well as S583A and D515H variants under conditions of MPT-AMP accumulation (22). Upon expression in the E. coli moeA mutant SE1581 (28) preferentially MPT-AMP was co-purified with all three variants ( Fig. 1) demonstrating a metabolic block in further processing of MPT-AMP in the moeA mutant (Fig. 1A), which is not seen in mogA mutants (22). Previously we have also shown that D515H is unable to synthesize MPT-AMP (22,23). Therefore one can conclude that co-purified MPT-AMP was synthesized by the Cnx1G-homologous E. coli MogA protein ( Fig. 1A). The product-substrate ratio (MPT-AMP:MPT) was best in the S583A variant (50% MPT-AMP saturation as determined by FormA analysis) due to its high MPT adenylation rate (23). Therefore we used S583A for subsequent biochemical studies and named it Cnx1G-bound MPT-AMP. Cnx1G concentrations were adjusted to 50% MPT-AMP saturation.
Hydrolysis of MPT-AMP by Cnx1E-Cnx1E is important for Mg 2ϩdependent in vitro Moco synthesis (22). Therefore, we investigated the ability of purified Cnx1E to hydrolyze Cnx1G-bound MPT-AMP ( Fig.  2A). In the presence of 1 mM MgCl 2 a time-dependent cleavage of MPT-AMP (detected as FormA-AMP) was observed (Fig. 2B). Simultaneously, the appearance of non-adenylated pterin (detected as FormAdephospho) was found (Fig. 2C). If not stated otherwise, we will use the term MPT for non-adenylated pterin throughout the manuscript. Interestingly, the increase in MPT was only linear up to 100-min reaction time; longer incubation resulted in a progressive decrease. A 6-fold increase in reaction rate was observed when only 20 M molybdate (2-fold molar excess) was added (Fig. 2B), and MPT-AMP hydrolysis was completed after 90-min reaction time. As MPT formation also reached a maximum at 90 min ( Fig. 2C), the strong decay observed with longer reaction times suggests a degradation of the non-andenylated pterin product generated in this reaction. These findings show that Cnx1E cleaves MPT-AMP in a Mg 2ϩ -dependent and molybdate-enhanced manner for which Cnx1E has to bind either to MPT-AMPloaded Cnx1G or MPT-AMP has to be transferred from Cnx1G to Cnx1E.
MPT-AMP and Molybdate Binding to Cnx1E-We have co-incubated equimolar amounts of MPT-AMP-loaded Cnx1G with Cnx1E (7.5 nmol each) in the presence or absence of 15 nmol molybdate and separated the reaction mixture by size exclusion chromatography (Fig.  3, A and B). In the absence of molybdate both proteins eluted at the same volume (data not shown) as the individually loaded proteins (Fig. 3A). MPT-AMP was only detected in Cnx1G-containing fractions (Fig. 3C).
In the presence of molybdate no change in the elution profile of Cnx1E and Cnx1G was observed (Fig. 3B), but MPT-AMP was shifted into the Cnx1E-containing fraction (Fig. 3D, 6.1 nmol) resulting in about 80% MPT-AMP-saturated Cnx1E. These data demonstrate that Cnx1E is able to bind MPT-AMP with high affinity but only in the presence of molybdate. Furthermore, no physical interaction between both proteins could be detected.
Molybdate-dependent MPT-AMP binding to Cnx1E indicated that also molybdate binds to the protein. Therefore, molybdate binding was measured by incubating Cnx1E with 2-fold molar excess of molybdate (12 M), desalting, and determining unbound molybdate in the salt fraction using a colorimetric molybdenum assay (25) (Fig. 4A and B). Low concentrations of molybdate were used in order to measure accu-  Moco biosynthesis is shown as described before (22). Cnx1G and presumably the homologous MogA convert MPT in MPT-AMP by Mg 2ϩ -dependent hydrolysis of ATP. As Cnx1G variants were expressed in a Cnx1E-homologous moeA strain no further processing of MPT-AMP was assumed. B, co-purification of MPT and MPT-AMP with wild-type, S583A, and D515H variants of Cnx1G. Proteins were expressed in E. coli SE1581 (moeA), purified as described under "Experimental Procedures," and 650 pmol of purified proteins were analyzed. Error bars were derived from triplicate measurements.

Nucleotide-assisted Molybdenum Activation
rate changes in the remaining unbound substrate as the determination of protein-bound molybdate was hampered by protein precipitations (data not shown). In the absence of MPT-AMP no molybdate binding was observed to Cnx1E (Fig. 4B). When adding 6 M MPT-AMP to 6 M Cnx1E, a time-dependent binding of molybdate reaching saturation at 6 M was observed (Fig. 4A) indicating equimolar binding. Next we have titrated either MPT-AMP or Cnx1E against constant amounts of Cnx1E or MPT-AMP (loaded to Cnx1G), respectively (Fig. 4B). In both cases molybdate binding reached saturation at 6 M according to the respective concentration of the constantly kept MPT-AMP or Cnx1E, respectively. Therefore we conclude that MPT-AMP and molybdate bind cooperatively in an equimolar manner to Cnx1E. As quantitative binding was observed upon gel filtration the affinity for both substrate can be considered in the submicromolar range. The cooperative binding of MPT-AMP and molybdate to Cnx1E raised the question how molybdenum is attached to the Cnx1E-MPT-AMP complex. Due to our previous finding that Cnx1G-bound MPT and MPT-AMP contain an ene-dithiolate-coordinated copper atom (22), we examined the molybdenum and copper content of Cnx1G-and Cnx1E-MPT-AMP complexes by ICP-MS (Fig. 4C). First, we confirmed the previously found binding of copper to MPT-AMP (22) by detecting 200 pmol of copper on 1 nmol of Cnx1G. The submolar stoichiometry of copper to MPT-AMP (500 pmol) might indicate a dissociation of copper during purification from Cnx1G-MPT-AMP complex (Fig. 4C). Molybdenum was undetectable on Cnx1G or Cnx1E. After transfer of MPT-AMP (500 pmol) from Cnx1G to Cnx1E in the presence of 1 nmol of molybdate an equimolar transfer of molybdenum to the protein fraction was observed (Fig. 4C, "p"). Remaining unbound molybdate was regained in the salt fraction (Fig. 4C, "s"). Most importantly, the MPT-AMP-coordinated copper remained in the protein-bound fraction demonstrating that molybdenum has not replaced the copper upon binding to Cnx1E. Also Cnx1E-MPT-AMP complex purified by gel filtration showed almost similar ratios between molybdenum, copper, and MPT-AMP (Fig. 4C, "g "). The slight decrease in copper content also points again to a loss of copper by dissociation.
Tungsten and Sulfate Compete for the Molybdate-binding Site on Cnx1E-After showing that molybdate binds to Cnx1E we wanted to know if other anions might compete for the molybdate-binding site on Cnx1E. Tungstate and sulfate are known competitors of molybdate in molybdate-binding proteins (29). Molybdate binding experiments were performed as described above and increasing amounts (0 -100 mM) of tungstate and sulfate (Fig. 4D) were added. Already 20-fold excess of tungstate (250 M) showed detectable competition for the molybdatebinding site and reached a maximum at 5 mM, which is consistent with the known antagonistic effect between tungstate and molybdate (29). Also sulfate showed inhibition of molybdate binding but to a much lower extent.
Next we wanted to know if both anions could also promote MPT-AMP binding to Cnx1E. Therefore, Cnx1E was co-incubated with a 2-fold excess of MPT-AMP (Cnx1G-bound, compare Fig. 3) in the presence or absence of molybdate (20 M), tungstate (5 mM), or sulfate (100 mM). Cnx1E and Cnx1G were separated by size exclusion chromatography and MPT-AMP as well as MPT contents were determined (Fig.  4E). In the presence of any of the three anions a nearly complete transfer of MPT-AMP was observed indicating a cooperative binding of MPT-AMP to Cnx1E in the presence of tetrahedral anions. Furthermore, these data demonstrate that molybdate binds to a distinct anion-binding site in Cnx1E.
Metal-and Anion-specific MPT-AMP Hydrolysis by Cnx1E-After quantitative transfer of MPT-AMP and molybdate to Cnx1E we performed MPT-AMP hydrolysis experiments with different divalent cations (1 mM each) to determine the appropriate co-substrate (Fig. 5A). The data show that not only Mg 2ϩ but also Ni 2ϩ and Mn 2ϩ are able to promote MPT-AMP hydrolysis. However the most efficient cation was Zn 2ϩ , which is known to act as a cofactor in similar reactions catalyzed by pyrophosphatases (30).
Due to the fact that not only molybdate but also other anions are able to bind cooperatively to Cnx1E we investigated MPT-AMP hydrolysis as a function of different anions bound to the proposed molybdatebinding site in Cnx1E. This experiment clearly shows that only molybdate-loaded Cnx1E is able to hydrolyze MPT-AMP, while tungstate and sulfate show no activity under the experimental conditions used (Fig.  5B). With much longer incubation times (1-2 h) low hydrolysis rates were observed (data not shown) according to the molybdate-independent cleavage shown in Fig. 2. These data unequivocally demonstrate that Cnx1E-catalyzed MPT-AMP hydrolysis is strictly dependent on bound molybdate.
Reaction Rates of MPT-AMP Hydrolysis-We also determined the catalytic parameters of MPT-AMP hydrolysis (Fig. 5, C-E). As only one molecule per Cnx1E reaction could be cleaved the experimental set up resembled a "pre-steady-state"-type of kinetic analysis. MPT-AMP hydrolysis rates with Zn 2ϩ (Fig. 5C) were much higher than with Mg 2ϩ (Fig. 5D) as shown in the double-reciprocal plot (Fig. 5E) with a k cat for Zn 2ϩ of 2.5 s Ϫ1 , whereas the Mg 2ϩ -dependent reaction rate was 10 times lower (k cat ϭ 0.25 s Ϫ1 ). The K m for both metals was in the same range (133-255 M, Fig. 5F). It is important to note that the formation of non-adenylated pterin was significantely slower than the simultaneous hydrolysis of MPT-AMP (data not shown). In Vitro Synthesis of Mature Moco by Cnx1E-We have shown that Cnx1E binds molybdate and hydrolyzes bound MPT-AMP thus yielding non-adenylated pterin. Now we want to demonstrate that during molybdate-dependent MPT-AMP hydrolysis molybdenum is transferred to the MPT ene-dithiolate and mature Moco is released. So far there is no direct measure for Moco, and it is commonly established that free Moco is rapidly degraded (31), which fits well to our previous observation (Fig. 2C). To prove Moco formation we used the simplest molybdenum enzyme, PSO, in its apo-state. Large amounts of Moco-free PSO were obtained by purification under denaturating conditions and subsequent on-column refolding (see "Experimental Procedures").
Apo-PSO was incubated with increasing amounts of Cnx1E-MPT-AMP complex in the presence of 500 M Zn 2ϩ , which resulted in rapid MPT-AMP hydrolysis (compare Fig. 5C). Maximal activation of 10 M apo-PSO was reached with 10 M Cnx1E-MPT-AMP complex (Fig. 6A) pointing to an equimolar and quantitative transfer of Moco formed upon MPT-AMP hydrolysis. The reaction was completed within 10 -15 min (Fig. 6B). It was very important that apo-PSO was present before starting the reaction, as increasing preincubation times of Cnx1E-MPT-AMP complex with Zn 2ϩ before adding apo-PSO resulted in rapid degradation of Moco with an estimated half-life of 6 min (Fig. 6C). As all experiments were performed with 10 M Cnx1E-MPT-AMP complex and 10 M bound molybdate these data clearly show an equimolar conversion of MPT-AMP and molybdate into Moco, which reflects the physiological situation of metal-cofactor assembly.
Rates of Substrate Decomposition and Formation of Products-To gain more details into the complex reaction of Moco formation by Cnx1E we compared the rates of apo-PSO activation (Fig. 7A) with rates of MPT-AMP hydrolysis (Fig. 7B). To slow down MPT-AMP hydrolysis only 100 M MgCl 2 was used and the reaction was stopped by 1 mM EDTA at indicated times. PSO reconstitution showed a linear dependence with reaction time and reached maximum after 20 min with a rate of 0.272 unit/min (Fig. 7A). According to the results shown in Fig. 6 we conclude that 6 units of SO activity correspond to fully reconstituted PSO (10 M) on which basis a Moco synthesis rate of 9 pmol/min was calculated. MPT-AMP hydrolysis was also completed after 15-20 min, but only the first 10 min of the reaction can be considered as linear with a rate of 14 pmol MPT-AMP hydrolyzed per min (Fig. 7B). After stopping the reaction by EDTA an extra 15 min were given to complete PSO reconstitution to avoid interfering effects of Moco transfer. When comparing both reaction rates a 1.5-fold difference is seen between Moco synthesis and MPT-AMP hydrolysis that might point toward a second half reaction during Moco formation. We have also performed these experiments in the absence of apo-PSO showing the same rate of MPT-AMP hydrolysis as with apo-PSO (Fig. 7B).
Upon MPT-AMP hydrolysis AMP should be released as second substrate, which was determined by HPLC analysis (Fig. 7C). Under similar concentrations as shown in Fig. 7, A and B, we detected a time-dependent release of AMP (Fig. 7D). The rate of AMP release (9 pmol/min) was comparable with the rate of Moco formation (Fig. 7A) and therefore again significantly lower than the rate of MPT-AMP hydrolysis, which confirms the above mentioned rate difference in substrate decomposition and formation of products.
Finally, we also monitored the traffic of Cnx1E-bound metals during Moco synthesis. by measuring molybdenum and copper contents in  Cnx1E-MPT-AMP complexes before and after Moco formation (Fig.  7E). Upon the addition of Zn 2ϩ both metals were found in the salt fraction after gel filtation showing that Moco and also copper were ultimately released after MPT-AMP hydrolysis. When including apo-PSO into the reaction, only copper was found in the salt fraction because molybdenum was then bound as Moco to the reconstituted PSO (Fig.  7E). Despite the submolar saturation with copper this experiment clearly showed that during Moco synthesis all the bound copper is replaced by molybdenum and copper is immediately released to the solvent.

DISCUSSION
One of the first biochemical observations of Moco mutants were molybdate-repairable phenotypes pointing to a defect in molybdate uptake or molybdenum insertion into the cofactor (32). Later in E. coli the mod locus encoding a high affinity molybdate uptake system (33) and the mogA gene essential for Moco biosynthesis (34) have been identified. In contrast to the single-locus defect in bacterial MogA plant, fungal and mammalian molybdate-repairable mutants showed a mutation in one of the two conserved domains of proteins like Cnx1 (14), CnxE (35), or gephyrin (15), respectively. The two-domain nature of Cnx1 and its homologues pointed to a functional "cooperation" between both domains such as product-substrate-channeling. For Cnx1G the synthesis of MPT-AMP (22,23) was recently presented. Here, we show that it is Cnx1E, which catalyzes the hydrolysis of MPT-AMP in a molybdate-dependent manner. We have unequivocally demonstrated that MPT-AMP and molybdate bind with high affinity in a cooperative and equimolar manner to Cnx1E. This finding explains the previously observed in vivo molybdate binding of Cnx1 that was dependent on the ability of the cell to synthesize MPT (36). Our competition studies and metal analyses have shown that both substrates bind to different sites on Cnx1E. As other tetrahedral anions are able to compete for the anionbinding site of molybdate we proposed a conformational change induced by the anion that enables Cnx1E for MPT-AMP binding.
Once transferred from Cnx1G to Cnx1E, MPT-AMP is rapidly hydrolyzed in the presence of Mg 2ϩ or Zn 2ϩ with rates that are several orders of magnitude higher than MPT-AMP synthesis (23). Therefore MPT-AMP synthesis seems to be the rate-limiting step in Cnx1 reaction. Based on the reaction catalyzed, Cnx1E can be considered as a member of the large family of Nudix hydrolases, enzymes that catalyze the hydrolysis of nucleoside diphosphates linked to other moieties (37). Within this family of diverse enzymes ADP-ribose pyrophosphatase might be the closest functional homologue to Cnx1E because it catalyzes the hydrolysis of ADP-ribose by nucleophilic substitution of water at the adenosyl phosphorus to yield AMP and ribose 5-phosphate, the latter could correspond to MPT (Moco) in the Cnx1E reaction. In general, pyrophosphatases can use different divalent cations as co-substrates, but in most cases Mg 2ϩ is preferentially used and the activity observed with Zn 2ϩ , Mn 2ϩ , or Co 2ϩ is considered as non-physiological (30). Therefore we also believe that despite of the higher rates with Zn 2ϩ Cnx1E is mainly using Mg 2ϩ as co-substrate.
Molybdate was essential for both MPT-AMP binding and hydrolysis indicating that Cnx1E catalyzes the metal insertion reaction thus forming Moco. To demonstrate Moco synthesis we used PSO (24) as acceptor enzyme thereby generating a fully defined in vitro system for Moco synthesis. We showed that Cnx1E-mediated MPT-AMP hydrolysis in the presence of bound molybdate resulted in the formation of equimolar amounts of Moco that is quantitatively transferred to PSO. Reconstitution was completed within 15 min, which is in the range of human sulfite oxidase (20 -30 min) (38). As Moco is sensitive to oxidation it was very important that PSO was present before starting MPT-AMP hydrolysis to protect newly synthesized and released Moco.
A comparison between the rate of MPT-AMP hydrolysis, Moco formation and AMP release revealed an 1.5-fold difference between substrate hydrolysis and product release. Therefore, we propose a second half-reaction after MPT-AMP hydrolysis that is needed to complete the synthesis of active Moco. The model presented in Fig. 8 depicts a proposed mechanism for the Cnx1E-catalyzed molybdenum insertion into MPT. First, MPT-AMP and molybdate cooperatively bind to Cnx1E (Fig. 8, step 1). Next, MPT-AMP is hydrolyzed by the addition of Zn 2ϩ or Mg 2ϩ (step 2). According to the action of pyrophosphatases the metal stabilizes the negative charge of the pyrophosphate thereby making the phosphorous more electrophile. Whereas in ADP-ribose pyrophosphatase a water is mediating the hydrolysis (37), we propose for Cnx1E that a molybdate oxygen promotes pyrophosphate cleavage. Consequently, this attack would result in the formation of adenylated molybdate as hypothetical and probably very transient reaction intermediate. Such activations by adenylyl transfers are known from other processes like aminoacyl-tRNA synthesis (39), thiamin biosynthesis (40), sulfate assimilation (41), or even MPT synthesis (42).
Supporting evidence for the hypothesis presented above comes from studies of the sulfate assimilation pathway. There, molybdate is known to act as substrate analogue of ATP sulfurylase (molybdolysis (43)), the first enzyme in the sulfate assimilation catalyzing the adenylation of sulfate (41). In the presence of molybdate ATP sulfurase hydrolyzes ATP. However, so far the formation of adenylated molybdate was not demonstrated probably due to its transient nature (44). Another line of evidence for the existence of an adenylated molybdate comes from in vivo studies with E. coli moeA mutants that have a defect in the Cnx1Ehomologous MoeA protein (28). These mutants are suppressed under condition of sulfur starvation, which results in high expression of ATP sulfurylase that is know to act on molybdate (45). Therefore the formation of adenylated molybdate either by Cnx1E and MoeA or non-physiologically by ATP sulfurylase is proposed.
Returning to the model, adenylated molybdate would be able to react with the MPT dithiolate thereby replacing bound copper (Fig. 8, step 3), which was clearly demonstrated by shifting copper from the proteinbound to the salt fraction. Our current data also indicate that copper is not involved in molybdenum insertion as MPT-AMP species not loaded with copper were also converted into Moco, which favors the hypothesis that copper plays a protecting function during Moco synthesis. We believe that submolar saturation of MPT-AMP with copper reflects its dissociation during purification and that in vivo the MPT/MPT-AMP dithiolate is saturated with copper. Finally, the observed inhibition by external copper and the suppression by the addition of Cnx1E (22) clearly points to a block in the Cnx1E catalyzed molybdenum insertion when excess copper is provided.
The coordination of molybdenum in the reaction product of Cnx1E (Fig. 8, stage 4) is shown with three oxygens as derived from the molybdate adenylate. Later, during Moco insertion into molybdenum enzymes, one of the oxygens might be replaced either by a cysteine that coordinates the molybdenum center in enzymes of the sulfite oxidase family (46) or by a terminal sulfur ligand forming the center of enzymes of the xanthine oxidoreductase family (47). Upon formation, Moco is released from Cnx1E (Fig. 8, step 4), which is supported by fast reconstitution of PSO as well as its sensitivity to oxygen in the absence of the protecting environment of a molybdenum enzyme.
In eukaryotes a defect in any of the two conserved domains of Cnx1 or homologous proteins causes a molybdate-repairable phenotype (14,35,48). This finding suggests that in the presence of excess molybdate the nucleotide-asssisted metal insertion step might be bypassed by direct chelation of molybdenum to the MPT dithiolate. However, under physiological conditions molybdenum concentrations in the environment are low, and therefore a high affinity uptake (still unknown in eukaryotes) and processing system (Cnx1) are needed. In E. coli, mogA mutants are molybdate-repairable but moeA mutants are not. Together with the inability of Cnx1E to reconstitute E. coli MoeA function (14) one can argue that the metal insertion step is different between eukaryotes and bacteria, which might be due to the synthesis of a bis-MPT type cofactor in prokaryotes. Recently, it was shown that recombinant MoeA is also able to transfer molybdenum to MPT (49) but using a much higher concentration of both MoeA and molybdate (49) that does not reflect the physiological situation.
In summary we have demonstrated that Cnx1E converts MPT-AMP, the product of Cnx1G, in a very efficient way into active Moco by inserting molybdenum from bound molybdate. We have shown a novel nucleotide-assisted mechanism and propose the formation of adenylated molybdate as intermediate that reacts with the MPT dithiolate thus forming mature Moco.