The Mechanism of Assembly and Cofactor Insertion into Rhodobacter capsulatus Xanthine Dehydrogenase*

Rhodobacter capsulatus xanthine dehydrogenase (XDH) is a molybdo-flavoprotein that is highly homologous to the homodimeric mammalian xanthine oxidoreductase. However, the bacterial enzyme has an (αβ)2 heterotetrameric structure, and the cofactors were identified to be located on two different polypeptides. We have analyzed the mechanism of cofactor insertion and subunit assembly of R. capsulatus XDH, using engineered subunits with appropriate substitutions in the interfaces. In an (αβ) heterodimeric XDH containing the XdhA and XdhB subunits, the molybdenum cofactor (Moco) was shown to be absent, indicating that dimerization of the (αβ) subunits has to precede Moco insertion. In an (αβ)2 XDH heterotetramer variant, including only one active Moco-center, the active (αβ) site of the chimeric enzyme was shown to be fully active, revealing that the two subunits act independent without cooperativity. Amino acid substitutions at two cysteine residues coordinating FeSI of the two [2Fe-2S] clusters of the enzyme demonstrate that an incomplete assembly of FeSI impairs the formation of the XDH (αβ)2 heterotetramer and, thus, insertion of Moco into the enzyme. The results reveal that the insertion of the different redox centers into R. capsulatus XDH takes place sequentially. Dimerization of two (αβ) dimers is necessary for insertion of sulfurated Moco into apo-XDH, the last step of XDH maturation.

Mammalian XORs catalyze the hydroxylation of hypoxanthine and xanthine, the last two steps in the formation of urate, and exist originally as the dehydrogenase form (XDH, EC 1.17.1.4) but can be converted to the oxidase form (XO, EC 1.1.3.22) either reversibly by oxidation of sulfhydryl residues of the protein molecule or irreversibly by proteolysis (4). XDH shows a preference for NAD ϩ reduction at the FAD reaction site, whereas XO exclusively uses dioxygen as a terminal electron acceptor, leading to the formation of superoxide and hydrogen peroxide (5). The enzyme has been implicated in diseases characterized by oxygen radical-induced tissue damage, such as postischemic reperfusion injury (6). The oxidation of xanthine takes place at the molybdenum center, and the electrons thus introduced are rapidly distributed to the other centers according to their relative redox potentials (1). The re-oxidation of the reduced enzyme by the oxidant substrate, either NAD ϩ or molecular oxygen, occurs through FAD (7). The two [2Fe2S] clusters (FeSI and FeSII) are indistinguishable in terms of their absorption spectra, but the midpoint redox potential of FeSII is generally more positive than that of the FeSI center (8,9). The FeS centers from enzymes of the XO family have been characterized earlier by EPR (10 -12). The FeSI center of eukaryotic XOR exhibits a rhombic EPR signal, well observable at temperatures up to 60 K, slightly different from those found in the regular plant-type [2Fe2S] ferredoxins (9), whereas FeSII exhibits an unusual broad EPR signal that is characteristic of some molybdenum-containing hydroxylases and can be observed only at 20 K or lower temperature.
The crystal structures of the mammalian XDH/XO from bovine milk (3) and the structure of the highly homologous bacterial XDH from Rhodobacter capsulatus (13) have been solved. The bacterial XDH can be expressed in high quantities in a heterologous Escherichia coli system in a highly active form (14). The amino acid sequence of R. capsulatus XDH has a high degree of similarity to eukaryotic XORs (up to 39% identity to bovine milk XOR), however, in contrast to the homodimeric (␣) 2 structure of eukaryotic XORs, R. capsulatus XDH has an (␣␤) 2 heterotetrameric structure, and the cofactors were identified to be located on two different polypeptides: the iron-sulfur clusters and the FAD are bound by the XdhA subunit, and the Moco is bound by the XdhB subunit (15). Despite differences in subunit composition, the folds of bovine XDH and R. capsulatus XDH are highly similar but differ in important details. The NAD ϩ binding pocket of the bacterial XDH resem-bles that of the dehydrogenase (XDH) form of the bovine enzyme rather than that of the oxidase (XO) form, and it was shown that R. capsulatus XDH is a true dehydrogenase that is not converted to an oxidase (14). The Moco was found to be deeply buried in the XdhB subunit at the end of a funnel-shaped passage giving access only to substrate molecules like pterins, purins, and aldehydes (13). The FeS clusters of R. capsulatus XDH showed notable differences in comparison to bovine XO (16), with FeSI having the highest redox potential and showing an EPR spectrum with axial symmetry.
Mammalian XORs exist as dimers, and it was reported that the two monomers act as independent catalytic subunits (1). A non-cooperative mechanism was also reported for R. capsulatus XDH (14). Because, with the exception of the monomeric DMSO reductase (17), most molybdoenzymes have two independent acting catalytic active sites (18), this raised the oldaged question of why some proteins are dimers. However, a more recent report by Tai and Hwang (19) showed that the two bovine XOR subunits are strongly cooperative in both binding and catalysis.
To investigate whether the two (␣␤) subunits of R. capsulatus XDH act independently or whether an intramolecular electron transfer mechanism occurs between the two ␤-subunits, we have generated an (␣) 2 (␤ 1 wt/␤ 2 E730A) chimeric XDH variant containing one active XdhB half and one inactive XdhB half. To further characterize the mechanism of assembly and intramolecular electron transfer of R. capsulatus XDH, sitedirected mutagenesis was performed modifying the cysteines ligating FeSI and FeSII, and further, the role of Gln A 102 for electron transfer from Moco to FeSI was investigated. In addition, amino acid exchanges at the dimer interface of the XdhB subunit were introduced to create an (␣␤) dimeric variant of XDH. The influences of the amino acid exchanges on the assembly of the two XDH subunits, cofactor insertion, the kinetic constants, and the EPR properties were investigated. Our results show that the assembly of XDH is a complex process, which occurs in an ordered manner. A model for the assembly and cofactor insertion into R. capsulatus XDH is presented.

EXPERIMENTAL PROCEDURES
Bacterial Strains, Plasmids, Media, and Growth Conditions-E. coli TP1000(⌬mobAB) cells (20) were used for the expression of XDH wild type and variants from plasmid pSL207 (16). E. coli RK4353(DE3)mobAB Ϫ cells (16,21) were used for heterologous expression of the chimeric R. capsulatus XDH after cotransformation of plasmids pSL239/pAK22 and pSL239/pSS2 (see below). E. coli cultures were generally grown in LB medium under aerobic conditions at 30°C. When required, 1 mM sodium molybdate, 150 g/ml ampicillin, or 50 g/ml chloramphenicol were added to the medium.
Construction of Expression Vectors-By using PCR mutagenesis the corresponding amino acid exchanges Q A 102A, Q A 102G, E B 220R/D B 517R, C A 44A/C A 47A, and C A 134A/ C A 136A were introduced by base pair exchanges into plasmid pSL207 for the expression of the xdhABC genes of R. capsulatus XDH (16).
For the expression of a chimeric (␣) 2 (␤ 1 wt/␤ 2 E B 730A) XDH, the amino acid exchange E B 730A was introduced into plasmid pAK22 (22), containing the xdhB gene as a NdeI/KpnI fragment cloned into vector pTYB2 (New England Biolabs), allowing the expression of XdhB as a C-terminal fusion to an intein tag containing a chitin binding domain. The resulting plasmid was designated pSS2. For coexpression of wild-type xdhB in conjunction with xdhA and xdhC, the plasmid pACYC-duett-1 (Novagen) was used. Primers were designed that allowed cloning of the xdhAB genes into the NcoI and HindIII sites of MCS1 of pACYC-duett-1. In addition, the 3Ј-HindIII primer contained the coding sequence for six histidines, resulting in an XdhB fusion protein containing a C-terminal His 6 tag. Subsequently, an NdeI-KpnI fragment containing the coding sequence of XdhC was cloned into the MCS2 of this vector. The resulting plasmid was designated pSL239.
Expression and Purification of Different XDH Variants-Recombinant R. capsulatus wild-type XDH was purified using the procedure described by Leimkühler et al. (14), with affinity chromatography on Sepharose 4B/folate gel as the final step. The generated XDH variants were expressed under the same conditions as the wild-type enzyme and purified by nickel-nitrilotriacetic (NTA), Q-Sepharose, and size-exclusion chromatography. The purified enzymes were concentrated by ultrafiltration, gel filtered using a PD10 gel filtration column (GE Healthcare), equilibrated with 50 mM Tris, 1 mM EDTA, 2.5 mM dithiothreitol, pH 7.5, and stored at Ϫ70°C until used. For expression of chimeric (␣) 2 (␤ 1 wt/␤ 2 E B 730A) XDH E. coli RK4353(DE3)mobAB Ϫ cells cotransformed with pSL239 and pSS2 were grown in LB medium supplemented with 150 g/ml ampicillin, 50 g/ml chloramphenicol, 1 mM molybdate, and 0.02 mM isopropyl-␤-D-thiogalactopyranoside until the A 600 ϭ 1. These precultures were used 1:500 to start the main culture and subsequently grown at 30°C until A 600 ϭ 0.3 was reached. The cultures were induced with 100 M isopropyl-␤-D-thiogalactopyranoside, and the growth was continued for 24 h at a temperature of 16°C. Cells were harvested by centrifugation at 8000 ϫ g, the pellet was resuspended in 3 volumes of 50 mM sodium phosphate, 300 mM NaCl, pH 8.0, and cell lysis was performed by several passages through a French Pressure cell. The supernatant was incubated with 1.7 ml of Ni 2ϩ -NTA resin (Qiagen) per 2 liters of cell growth. The slurry was transferred to a column and washed with 10 column volumes of 50 mM sodium phosphate, 300 mM NaCl, pH 8.0, containing 10 mM imidazole, followed by a wash with 10 column volumes of the same buffer containing 20 mM imidazole. His 6 -tagged XDH was eluted with 50 mM sodium phosphate, 300 mM NaCl, pH 8.0, containing 250 mM imidazole. The eluted fractions were analyzed by SDS-PAGE, and the ones containing XDH were combined and dialyzed against 20 mM Tris, 500 mM NaCl, 1 mM EDTA, pH 8.5. To the dialyzed samples 7 ml of Chitin matrix (New England Biolabs) was added, and the mixture was incubated for 2 h at 4°C. The slurry was transferred to a column and washed with 20 mM Tris, 500 mM NaCl, 1 mM EDTA, pH 8.5. To induce the intein-catalyzed self-cleavage reaction, a buffer with 250 mM Tris, 500 mM NaCl, 1 mM EDTA, 50 mM dithiothreitol, pH 8.5, was added and left on the column for 24 h at 4°C. Proteins were eluted with the same buffer and analyzed by 10% SDS-PAGE. XDH-containing fractions were combined and dialyzed against 50 mM Tris, 1 mM EDTA, pH 7.5, and stored at Ϫ70°C until used.
Absorption Spectra during Anaerobic Reduction with Xanthine and NADH-XDH in 0.5 ml of 50 mM Tris, 1 mM EDTA, pH 7.5, was incubated in an anaerobic chamber (Coy Lab Systems) for 2 h at 4°C before either xanthine or NADH was added to a final concentration of 100 M. Complete reduction was achieved by the addition of 20 mM sodium dithionite. Spectra were recorded in 0.15-ml cuvettes using a Shimadzu UV-2401 PC spectrophotometer.
Enzyme Assays-Routine enzyme assays were carried out at 25°C in either 50 mM Tris, 1 mM EDTA, pH 7.5, or 50 mM Tris, 0.2 mM EDTA, pH 7.8, monitoring the absorbance change at 340 nm due to reduction of NAD ϩ to NADH. The enzyme concentration was determined from the absorbance at 465 nm using an extinction coefficient of 31.6 mM Ϫ1 cm Ϫ1 (14). XDH assays were performed using a Shimadzu UV-2401PC spectrophotometer. 5 nM XDH was incubated with 20 -100 M xanthine as substrate, and 20 -100 M NAD ϩ or 20 -100 M dichlorphenolindophenol (DCPIP) as electron acceptor.
Metal and Moco/MPT Analysis-The molybdenum and iron contents of the purified proteins were quantified by inductively coupled plasma-optical emission spectroscopy analysis with a PerkinElmer Optima 2100 DV. The samples were wet-ashed at a concentration of 10 M in a volume of 500 l by the addition of 500 M 65% nitric acid and incubated overnight at 100°C. The samples were further diluted by the addition of 4 ml of water. As reference, the multi-element standard solution XVI (Merck) was used.
Moco was quantified by conversion to Form A as described earlier (23). Moco was extracted from 0.1 M XDH by the addition of 50 l of acidic iodine. Following incubation at room temperature for 14 h, excess iodine was removed by the addition of 55 M 1% ascorbic acid, and the sample was adjusted with 1 M Tris to pH 8.3. The phosphate monoester of Form A was cleaved by the addition of 16 l of 1 mM MgCl 2 and 1 unit of calf intestine alkaline phosphatase. After the addition of 10 l of acetic acid, Form A was identified and quantified by high-performance liquid chromatography analysis with a C18 reversed phase high-performance liquid chromatography column (4.6ϫ 250-mm ODS Hypersil, particle size 5 m) with 5 mM ammonium acetate, 15% methanol at an isocratic flow rate of 1 ml/min. In-line fluorescence was monitored by an Agilent 1100 series detector with an excitation at 383 nm and emission at 450 nm.
CD Spectroscopy-UV-visible CD spectra of 1.7 mg/ml enzyme samples were recorded in 50 mM Tris, 1 mM EDTA, pH 7.5, using a Jasco J-715 CD spectrophotometer.
EPR Spectroscopy-9.5-GHz X-Band EPR spectra were recorded on a Bruker ESP300E spectrometer equipped with a rectangular microwave cavity in the TE 102 mode. For temperature control between 6 K und 100 K the sample was kept in an Oxford ESR 900 helium flow cryostat with an Oxford ITC4 temperature controller. The microwave frequency was detected with an EIP frequency counter (Microwave Inc.). The magnetic field was calibrated using an LiLiF standard with a known g-value of 2.002293 Ϯ 0.000002 (24). Samples were prepared as frozen solutions (typically 0.1 mM enzyme) in quartz tubes with 4-mm outer diameter. Chemical reduction, to generate the reduced Fe(II)Fe(III) clusters, has been performed with a 20-fold excess of sodium dithionite. Baseline corrections, if required, were performed by subtracting a background spectrum, obtained under the same experimental conditions from a sample containing only a buffer solution. Simulations of the experimental EPR spectra have been carried out with the program EasySpin (25). Second integrals from the simulated FeSI and FeSII spectra were used to estimate the relative amount of both clusters in the respective samples.

Characterization of an R. capsulatus XDH Mutant Affecting the Binding Site of FeSI-
The two non-identical [2Fe2S] clusters located in the XdhA domain of R. capsulatus XDH are coordinated by the amino acid motifs Cys 103 -X 2 -Cys 106 -X n -Cys 134 -Arg-Cys 136 for FeSI and Cys 39 -X 4 -Cys 44 -X 2 -Cys 47 -X n -Cys 63 for FeSII ( Fig. 1). FeSI is the cluster in proximity to the pterin ring of Moco, whereas FeSII is in proximity to the FAD (13,16). It has been reported before that a mutation of Arg A 135 in R. capsulatus XDH, being located in-between the two conserved cysteine residues Cys A 134 and Cys A 136 of FeSI resulted in the production of two different forms of XDH: an (␣␤) 2 heterotetrameric form indistinguishable from the wild-type protein and an (␣␤) heterodimeric form lacking FeSI (16). Thus alterations in FeSI seem to influence the correct assembly of the protein. To characterize alterations in the FeSI binding site on the (␣␤) 2 heterotetramer formation via the XdhB subunit further, an XDH variant was created exchanging two FeSI-coordinating ligands Cys A 134 and Cys A 136 to alanine. The XDH-C A 134A/C A 136A variant was purified after heterologous expression in E. coli cells and characterized ( Fig. 2A). As shown in Table 1, the protein contained an iron complement of 62%, and the molybdenum content of the protein was reduced to 8%. Further analyses revealed that the low molybdenum content of the protein corresponded well with a reduced MPT content, showing that the protein was produced in a form almost free of Moco (Table 1). Size exclusion chromatography identified that the XDH-C A 134A/C A 136A variant was mainly expressed as an (␣␤) heterodimer (Fig. 2B). However, the small shoulder at 11 min might indicate that a small amount of the protein existed as a heterotetramer. The UV-visible absorption spectra of the protein variant in comparison to native XDH showed that the heterodimer displayed diminished absorbance in the visible  region, with a shift in the absorbance peak from 465 to 450 nm ( Fig. 2A). Also, the absorbance of the FeS clusters and Moco at 320 nm was strongly influenced. However, because the iron content was shown to be 62%, this finding suggests that the protein variant lacks a significant amount of FeSI and contains a modified version of the FeSI cluster in a small amount of the protein. Steady-state kinetics of the XDH-C A 134A/C A 136A variant showed no xanthine-oxidizing activity with NAD ϩ or DCPIP as electron acceptor. In contrast to amino acid exchanges in the ligating cysteines for FeSI, it was not possible to purify a double variant of the ligands Cys A 44 and Cys A 47 for FeSII substituted by alanine, due to the instability of the protein (data not shown).
Generation of an (␣␤) Dimeric XDH Variant by Amino Acid Exchanges in the Dimer Interface-Because the absence or a modified FeSI prevented the formation of the XDH (␣␤) 2 heterotetramer and a modified FeSII influenced the stability of the protein, cofactor insertion into XDH seems to be a coordinated process. To investigate whether Moco insertion into XDH pre-cedes dimerization of the XdhB subunit to form the (␣␤) 2 heterotetramer, amino acid exchanges in the dimer interface region of XdhB were introduced. The crystal structure of R. capsulatus XDH suggested that introducing positively charged amino acids in the dimer interface region might interrupt dimerization (13). Thus a variant of XDH was generated exchanging Asp B 220 and Glu B 517 to arginine (Fig. 2C). Size exclusion chromatography identified that the XDH-D B 220R/ E B 517R variant was mainly purified as an (␣␤) heterodimer after heterologous expression in E. coli (Fig. 2D). The small peak at 11 min showed that a very small portion of the protein existed as a heterotetramer (Fig. 2D). Compared with native XDH, the UV-visible absorption spectra of the protein variant showed that the heterodimer also displayed a diminished absorbance in the visible region, with a shift in the absorption peak from 465 to 450 nm (Fig. 2C). Also, the absorbance of the FeS clusters and Moco at 320 nm was strongly influenced; however, the absorbance changes differed from the XDH-C A 134A/ C A 136A variant ( Fig. 2A). The iron content was shown to be 81%, suggesting an almost full complement of FeS clusters ( Table 1). Analysis of the molybdenum and MPT content of the variant indicated that the protein was purified in a form containing only residual amounts of Moco (Table 1), showing similarities to the XDH-

Investigations of the FeS Clusters of (␣␤) Heterodimeric and (␣␤) 2 Heterotetrameric XDH Variants
Using EPR Spectroscopy-Because of the differences in UV-visible spectra of R. capsulatus XDH and the XDH-C A 134A/C A 136A and XDH-D B 220R/E B 517R variants, the EPR properties of the FeS clusters were investigated. Fig. 3A shows the EPR spectra of the FeS clusters of dithionite-reduced R. capsulatus wild-type XDH (trace A), together with the corresponding simulations (traces B-E). The spectra show signals from the reduced FAD cofactor   Table 2). FeSII cluster has been simulated with g-values obtained from the Moco-deficient form (16). The EPR spectrum from the de-Moco XDH is a superposition of four paramagnetic centers. Fig. 3A shows a decomposition of all contributions obtained from simulations (except for Mo V ).
The g-values of the wild-type FeSI signals are similar to the ones of the de-Moco form, only the line widths are somewhat larger. The double integrated simulations for the single iron-sulfur clusters (traces C and D) display a ratio of 1:1 indicating the presence of both clusters in the same amount in the protein. The flavin semiquinone (FAD) has been simulated by using an isotropic g-value of 2.0 and a line width of 3.9 milliteslas (mT). This line width is larger as compared with that from usual flavins (ϳ1.9 mT) (11) and points to a magnetic interaction observed at low temperatures with the metal centers. At 80 K the line width of FAD is only 2.1 mT, in accordance to that from other FAD cofactors (11). The Moco (Mo V ) has been neglected in the simulations.
The EPR spectra from the FeSI and FeSII clusters of variants C A 134A/C A 136A and XDH-D B 220R/E B 517R differ from those of the wild type, although broad similarities remain (Fig. 3B). The spectrum of variant C A 134A/C A 136A (trace B, complete simulation in trace C) shows predominantly FeSII, which exhibits g-values, shifted closer together, and lines, showing broadening due to g-strain. The FeSI signal of the C A 134A/ C A 136A variant is very weak. The relative EPR signal intensity (second integral) of FeSI versus FeSII (Յ0.1) is much smaller as  .008 for FeSII. g y from FeSII has a larger error due to overlap with the Mo(V) signal. c The g-strain was included in the simulation with 0.025, 0.025, and 0.025 for g x , g y , and g z . d The strain was included in the simulation with 0.0, 0.012, and 0.05 for g x , g y , and g z . for the wild type (1.0) (Fig. 3A, trace A) indicating a loss of a significant fraction of FeSI in this variant. This is also reflected in the reduced iron content of 62%, determined for this variant (Table 1).
For the D B 220R/E B 517R variant (trace D) a spectrum more similar to that from the wild type was observed, with strong signals from FeSI. Both signals, from FeSI and from FeSII, are similar to that of the wild type, except for a smaller value of g x and slightly shifted values of g y and g z for FeSII. Furthermore, the line width is increased due to g-strain. The integrated signal intensities from FeSII versus FeSI, estimated from second integrals (trace E shows the complete simulation) show a relative ratio close to 1:1 indicating the presence of almost equal amounts of both clusters in this variant, consistent with the iron content of 81% (Table 1).
Iron, which may be bound in an unspecific way to the protein, can be sometimes observed in enzymes with active iron centers, in particular in variants, when the protein structure may be modified or in the worst case damaged. In its oxidized form this unspecific iron, Fe(III), usually gives a strong EPR signal with g ϭ 4. We recorded therefore EPR spectra over the full field range from 50 to 380 mT from all samples in the oxidized form, prior to reduction with dithionite. In all cases we found only marginal amounts of g ϭ 4 signal, corresponding at maximum (D B 220R/E B 517R variant) to Ͻ2% as compared with the EPR signals from the reduced FeSI and FeSII centers. This shows that practically all of the iron was present in the form of FeSI and/or FeSII clusters.
CD Spectroscopy-To obtain more information on the FeS centers in wild-type XDH and the D B 220R/D B 517R and C B 134A/C B 136A variants, CD spectra were measured in the visible region in both the reduced and oxidized forms (Fig. 4). The spectrum of the oxidized wild-type enzyme exhibited strong negative dichroic bands at ϳ350 -400 nm and 520 -580 nm, and intensive positive bands between 400 and 500 nm (Fig.  4A). From the various maxima and infections, transitions can be identified at 374 (Ϫ), 434 (ϩ), 470 (ϩ), and 552 (Ϫ) nm. On the reduction with dithionite, the spectrum changes markedly with less intense transitions at 371 (Ϫ), 409 (ϩ), 461 (Ϫ), and 573 (Ϫ) nm. The visible CD spectra of reduced and oxidized R. capsulatus wild-type XDH are very similar in shape and intensity to those of Comamonas acidovorans XDH (26). The spectra of the two variants were measured and compared with wildtype XDH. As seen in Fig. 4, the CD spectra derive largely from the FeS centers, thus, they provide further information about the content of correctly assembled FeSI and FeSII in the enzymes. CD ⑀434 and ⑀470 values for wild-type XDH, the D B 220R/D B 517R variant, and the C B 134A/C B 136A variant were determined from the spectra and used to calculate the amount of FeS centers in the variants, assuming the FeS center are the sole contributor to the CD spectrum at 434 nm and 470 nm. As obvious in Fig. 4 (A and B), the spectra of variant D B 220R/D B 517R in comparison to wild-type XDH have essentially the same form, but show different intensities. The intensity for the D B 220R/D B 517R variant is ϳ80% of the wild-type XDH spectrum, consistent with the reduced iron content of the enzyme. However, apparently both FeSI and FeSII are present at lower concentrations. The spectrum of the D B 220R/D B 517R could be reduced to the same extent as wild-type XDH. In contrast, the CD spectrum of the C B 134A/C B 136A variant was different in comparison to the two other spectra. As shown in Fig.  4C, the FesI signal at 470 nm of the C B 134A/C B 136A variant is almost completely absent. This is consistent with the EPR spectra and the reduced iron content of the C B 134A/C B 136A vari- ant to 62% in comparison to wild-type XDH. FeSII is present in a similar extent as in wild-type XDH (Fig. 4, A and C), thus the iron content of 62% is likely derived from unspecific bound FeIII and to a small extent from an inactive form FeSI.
Amino Acid Substitutions in the Gln A 102 Ligand of the Molybdenum Cofactor Bound to the XdhB Subunit-Especially mutations in the ligands for FeSI seemed to strongly influence the assembly of XDH and the subsequent insertion of Moco into the (␣␤) 2 heterotetrameric protein. The amino acid Gln A 102 , which lies in proximity to FeSI (Fig. 1), was identified to be the only residue of the XdhA subunit with direct contact to the pterin ring of Moco (13). To analyze whether Gln A 102 is involved in the electron transfer reaction or influences the assembly of the XdhA and XdhB subunits, we changed this residue to alanine and glycine by site-directed mutagenesis. Size exclusion chromatography identified that the XDH-Q A 102G and XDH-Q A 102A variants were expressed as an (␣␤) 2 heterotetramer. The UV-visible absorption spectra of the protein variant in comparison to native XDH showed little difference in the region around 300 -350 nm (Fig. 5). Analysis of the molybdenum and iron content revealed that the protein contained a full complement of iron, whereas the molybdenum content corresponded well with the MPT content determined to be ϳ60% for both variants. The kinetic parameters at pH 7.8 of the XDH-Q A 102G and XDH-Q A 102A variant with either NAD ϩ or DCPIP as electron acceptors are given in Table 1. The k cat xanthine and K m xanthine for NAD ϩ as electron acceptor for the protein variants were found to be very similar, whereas the val-ues with DCPIP as electron acceptor varied between the two variants and the wild-type XDH. Because no major differences in comparison to the wild-type enzyme were determined, it is rather unlikely that the electron transfer pathway from Moco to FeSI occurs via this amino acid residue.
Generation of an XDH Variant Containing Only One Active Site per (␣␤) 2 Tetramer-So far it was believed that both of the two identical (␣␤) heterodimers of R. capsulatus XDH act independently during catalysis (14). To analyze whether cooperativity exists in R. capsulatus XDH of an intramolecular electron transfer occurs between two XdhB subunits, we created an XDH variant containing two different active sites in the (␣␤) 2 heterotetramer: while one site contained the wild-type structure, the second active site contained the amino acid exchange E730A, which was previously shown to result in a fully inactive enzyme (14). The chimeric protein was created by fusion of two different affinity tags to the C terminus of the XdhB subunit (see "Experimental Procedures"). The two different versions of XdhB, containing a His 6 tag and an intein tag, were co-expressed together with the XdhA and XdhC subunits, and purified first by Ni-NTA chromatography and then by chitin affinity chromatography. The two affinity purification steps should separate all assembled proteins containing either two active wild-type XdhB subunits or two inactive XdhB-E730A subunits from the chimeric XdhB-wt/E730A variant. Fig. 6 shows a Coomassie-stained SDS-PAGE gel of the different purifications steps of the (␣) 2 (␤ 1 wt/␤ 2 E730A) chimeric XDH. For better comparison, the wild-type XDH was purified according to the same procedure (Fig. 6). Analysis of the Moco content showed that both proteins were purified in a form containing similar amounts of Moco with 76% for wild-type XDH and 70% for the (␣) 2 (␤ 1 wt/␤ 2 E730A) chimeric XDH (Table 3). For further char-  acterization of the activity of both proteins, reduction spectra with xanthine and NADH were recorded. Immediately after mixing the reduction level with xanthine as substrate resulted in ϳ46 and 26% of the initial absorbance at 465 nm for wild-type XDH and the (␣) 2 (␤ 1 wt/␤ 2 E730A) chimeric XDH, respectively. Reduction with NADH showed a level of 66% for wild-type XDH and 74% for the (␣) 2 (␤ 1 wt/␤ 2 E730A) chimeric XDH (Table 3). Because the reduction with NADH results in a reduced XDH to the four-electron state, both proteins are fully functional at the FAD site and the 2x[2Fe2S] clusters. However, the reduction with xanthine indicated that only 46% of wildtype XDH contained the terminal MoϭS ligand required for activity. In comparison, the (␣) 2 (␤ 1 wt/␤ 2 E730A) chimeric XDH was only reduced to a level of 26%, which is about half of the activity determined for wild-type XDH expressed simultaneously under the same conditions.
Steady-state Kinetics of the (␣) 2 (Table 4). While K m xanthine and the K m NAD are comparable for both enzymes, the k cat xanthine is only half of that of the chimeric enzyme in comparison to wild-type XDH. This result shows that with a reduced k cat of 50% the chimeric (␣) 2 (␤ 1 wt/␤ 2 E730A) XDH is able to reduce the substrate xanthine independent of the inac-tive half of the enzyme, showing that both subunits of R. capsulatus carry out catalysis without any cooperativity.

DISCUSSION
Biosynthesis of the functional form of XOR is clearly a multistep process. In the case of the homodimeric mammalian XOR, this requires the incorporation of FAD, 2x[2Fe2S] centers, and Moco into each subunit (1). To date it has not been investigated whether this occurs during the synthesis of the protein or post-translationally before or after dimer formation. Biosynthesis of R. capsulatus XDH is even more complicated, requiring the assembly of an (␣␤) 2 heterotetramer, which is a dimer of dimers. Here, the FAD and [2Fe2S] centers are inserted into the XdhA subunit, whereas the Moco is bound to the XdhB subunit (15). Our studies suggest that insertion of the three different redox centers into R. capsulatus XDH is a complex process that occurs in an ordered manner (Fig. 6). The data both from EPR and CD show that in the XDH-C A 134A/ C A 136A variant a significant amount of FeSI is missing. Only Յ10% of FeSI remains in this variant. Because the XDH-C A 134A/C A 136A variant was found to have mainly an (␣␤) dimeric structure with a little portion remaining as (␣␤) 2 heterotetramer ( Fig. 2A), we conclude that the absence of an active FeSI cluster results in a structure unable to form the XDH (␣␤) 2 heterotetramer. Also Moco was missing in this protein variant. Previous studies showed that a stable Moco-free (␣␤) 2 heterotetrameric form of XDH can be purified from an E. coli moaA-deficient strain (22). Thus, we conclude that Moco insertion is the last step of XDH assembly and occurs after the formation of the (␣␤) 2 heterotetramer.
To analyze whether intramolecular electron transfer occurs from Moco via the residue Gln A 102 , which is in close proximity to the ligand Cys A 103 of FeSI and the only residue of the XdhA subunit involved in coordination of Moco, Gln A 102 was substituted by alanine or glycine. Because no changes of the steadystate kinetics were found, any impact of Gln A 102 on the intramolecular electron transfer rates can be ruled out. Modification of the cysteines Cys A 44 and Cys A 47 coordinating FeSII result in an unstable protein. Surprisingly, modification of two amino acid residues at the dimer interface of the XdhB subunit to positively charged residues results in a stable (␣␤) dimeric XDH form. Because the dimeric protein contained both FeSI and FeSII and FAD but no Moco, this is additional evidence that Moco insertion might occur after the formation of the (␣␤) 2 heterotetramer. However, we cannot rule out the possibility that the structure of the (␣␤) heterodimer was altered by the two amino acid exchanges in such a manner that both dimerization via the XdhB subunit and Moco insertion were influenced, although FAD and FeS insertion are not affected.
An altered structure of the (␣␤) dimer can be inferred from the slightly modified EPR spectrum of FeSII from the XDH-E B 220R/D B 517R variant and the larger EPR line width (g-strain). However, the CD and EPR data show that both FeSI and FeSII are present, and possibly the structure of the (␣␤) dimer is stabilized and thus slightly changed after formation of the (␣␤) 2 heterotetramer.
The failure to express the XdhA subunit separately as reported before (16) implies that XdhA has to dimerize with the  XdhB subunit to form a stable complex prior to insertion of FAD and the two [2FeS] clusters. Also, the separately purified XdhB subunit, which was stabilized by specific purification conditions, has been shown to be monomeric and free of Moco (22).
We conclude that the presented data support the view that the assembly of XDH is a highly ordered process, which involves the synthesis of the XdhA and XdhB subunits, the dimerization of both subunits, the insertion of FeSI, FeSII, and FAD into the XdhA subunit, dimerization of two (␣␤) dimers via the XdhB subunit, and finally, insertion of sulfurated Moco into XdhB, resulting in an active enzyme (Fig. 7). The biosynthesis of Moco is additionally a complex process involving more than a dozen different proteins (27), with the insertion of sulfurated Moco by the XdhC protein being the last step of XDH maturation (28). This step is strictly regulated in R. capsulatus, because in vivo di-oxo Moco is not inserted into R. capsulatus XDH (29). Thus XdhC performs two reactions: (i) to ensure that Moco is sulfurated by the interaction with the L-cysteine desulfurase NifS4 (30) before insertion into XDH and (ii) to insert the sulfurated Moco in the formed (␣␤) 2 heteroteramer of XDH. Because Moco is deeply buried in the protein, it is also believed that XdhC acts as a chaperone being involved in proper folding of XDH after Moco insertion (29).
In contrast to a previous report on bovine XO (19), we have shown that both (␣␤) dimers of R. capsulatus XDH act independently without cooperativity or intramolecular electron transfer between the XdhB subunits. We were able to purify a chimeric (␣) 2 (␤ 1 wt/␤ 2 E730A) XDH with one active XdhB subunit and one inactive XdhB subunit (containing the amino acid exchange E730A at the active site). Although the K m values for the substrates xanthine and NAD remained unaltered, k cat for xanthine was reduced to 50% compared with the wild-type protein, ruling out any cooperativity or intramolecular electron transfer between the active subunits. This result is in agreement with early reports for bovine XO, suggesting that both subunits of bovine XO carry out catalysis independently (1). However, these findings are in contrast to a recent report by Tai and Hwang (19) who demonstrated that binding of slow substrates like 6-formylpterin at one active site affects the binding affinity and catalysis rate at the other active site. This finding may be taken as an indication for a difference in subunit interaction between R. capsulatus XDH and bovine XO. It would be, therefore, instructive, to apply the same approach used in this work also to bovine XO to unambiguously confirm or rule out a cooperative mechanism for the mammalian enzyme.
The present results for R. capsulatus XDH demonstrate that dimerization of the (␣␤) subunits is required to stabilize a structure of the protein that makes the protein suitable for Moco insertion. In contrast to our results, a sulfite oxidase variant was identified in a human patient containing the amino acid exchange G473D in impaired dimerization of the protein but still contains a full complement of Moco (31). For this protein variant it was shown that misfolding prevented dimerization of sulfite oxidase, thereby preventing an efficient electron transfer between the Moco domain and the heme domain of sulfite oxidase (31). So far, no specific chaperone/scaffold protein for Moco insertion like the XdhC protein for XDH has been reported for sulfite oxidase, thus a "quality control" is missing for sulfite oxidase ensuring that the enzyme adopts the correct conformation suitable for Moco insertion. This observation together with our data supports the idea that dimerization of molybdoenzymes is important for stabilizing a structure for the intramolecular electron transfer between various cofactors in most molybdoenzymes.