Binding of Flavin Adenine Dinucleotide to Molybdenum-containing Carbon Monoxide Dehydrogenase from Oligotropha carboxidovorans

The carbon monoxide (CO) dehydrogenase ofOligotropha carboxidovorans is composed of anS-selanylcysteine-containing 88.7-kDa molybdoprotein (L), a 17.8-kDa iron-sulfur protein (S), and a 30.2-kDa flavoprotein (M) in a (LMS)2 subunit structure. The flavoprotein could be removed from CO dehydrogenase by dissociation with sodium dodecylsulfate. The resulting M(LS)2- or (LS)2-structured CO dehydrogenase species could be reconstituted with the recombinant apoflavoprotein produced inEscherichia coli. The formation of the heterotrimeric complex composed of the apoflavoprotein, the molybdoprotein, and the iron-sulfur protein involves structural changes that translate into the conversion of the apoflavoprotein from non-FAD binding to FAD binding. Binding of FAD to the reconstituted deflavo (LMS)2 species occurred with second-order kinetics (k +1 = 1350m −1 s−1) and high affinity (K d = 1.0 × 10−9 m). The structure of the resulting flavo (LMS)2 species at a 2.8-Å resolution established the same fold and binding of the flavoprotein as in wild-type CO dehydrogenase, whereas theS-selanylcysteine 388 in the active-site loop on the molybdoprotein was disordered. In addition, the structural changes related to heterotrimeric complex formation or FAD binding were transmitted to the iron-sulfur protein and could be monitored by EPR. The type II 2Fe:2S center was identified in the N-terminal domain and the type I center in the C-terminal domain of the iron-sulfur protein.

O 2 , thereby producing hydrogen peroxide (H 2 O 2 ) and superoxide (O 2 . ). With Streptomyces thermoautotrophicus, the O 2 . produced by CO dehydrogenase serves as the electron donor in a superoxide-dependent nitrogenase reaction (4). CO dehydrogenase is a prototype of the molybdenum hydroxylase sequence family, which includes the aldehyde oxidoreductase (Mop) from Desulfovibrio gigas as well as enzymes, oxidizing aromatic N-heterocyclic compounds such as xanthine dehydrogenases/oxidases from different procaryotic and eucaryotic sources, quinoline 2-oxidoreductase from Pseudomonas putida, and nicotine dehydrogenase from Arthrobacter nicotinovorans (5,6). The crystal structure of CO dehydrogenase from O. carboxidovorans has been solved at a 2.2-Å resolution (7) and shows a dimer of heterotrimers (see Fig. 1). Each heterotrimer is composed of a 88.7-kDa molybdoprotein (L), a 30.2-kDa flavoprotein (M), and a 17.8-kDa iron-sulfur protein (S).
The molybdoprotein of CO dehydrogenase carries the molybdopterin cytosine dinucleotide (MCD) 1 -type of molybdenum cofactor and the unique active-site loop Gly 383 -Val-Ala-Tyr-Arg-Cys-Ser-Phe-Arg 391 , which positions the catalytically essential S-selanylcysteine 388 in a distance of 3.7 Å to the molybdenum ion. The iron-sulfur protein carries the two 2Fe:2S centers of CO dehydrogenase. One center is located in the N-terminal domain, and the other is located in the C-terminal domain of the iron-sulfur protein (see Fig. 1). The N-terminal 2Fe:2S center is positioned at the interface between the iron-sulfur protein, and the flavoprotein and is exposed to the solvent. It lies 8.7 Å apart from the FAD in the flavoprotein (Fig. 1). The flavoprotein binds the FAD as a prosthetic group in a fold formed by its N-terminal and middle domains. Binding of FAD involves strong interactions of two vanillyl alcohol oxidase-like double glycine motifs on each of the two domains, mainly with the pyrophosphate moiety of the FAD. On the N-terminal domain a loop between the sheet ␤2 and the helix ␣2 of a ␤␣␤ unit (P-loop) containing the motif Ala 32 -Gly 33 -Gly 34 -His 35 -Ser 36 is involved, and on the middle domain a small ␣-helix (␣7) containing the major part of the motif Thr 111 -Ile 112 -Gly 113 -Gly 114 -Asn 115 is involved. The C-terminal domain of the flavoprotein carries a Tyr 193 residue, which shields the central part of the isoalloxazine ring from the solvent (7). This paper presents a structural and functional analysis of FAD binding in CO dehydrogenase employing a CO dehydro-genase species in which the native flavoprotein has been replaced by a recombinant one. The FAD binding site on the free flavoprotein is not able to bind FAD. However, in the heterotrimeric complex of the flavoprotein with the iron-sulfur protein and the molybdoprotein, the flavin binding site is capable to integrate FAD.
A 0.95-kilobase BamHI-HincII fragment of pCDH1 containing the CO dehydrogenase structural gene coxM (5) was cloned into the vector pBluescript I KSϩ, yielding the plasmid pSK7. This plasmid was used as a template for polymerase chain reaction to amplify the first 244 nucleotides of coxM, using the following primers. The 34-mer 5Ј-TGTA-GAGCTCCATATGATCCCTGGTTCATTTGAT-3Ј was used to generate a NdeI site by replacing the start codon GTG by ATG and to alter the second codon for isoleucin from ATA to ATC, resulting in a suitable codon usage for expression in E. coli (14). The 25-mer 5Ј-ATTATCG-CATGCTGAGTGGTCATCG-3Ј contained a SphI site to cleave the amplified polymerase chain reaction product for cloning. The fragment amplified by polymerase chain reaction was cloned into the vector pET16b together with the SphI-XhoI fragment from pSK7, which contained the remaining part of coxM, yielding pETSK1. The NdeI-BamHI fragment from pETSK1 was cloned into pET11a, yielding pETSK2, which was then transformed into E. coli BL21(DE3)/pREP4 (11) yielding E. coli BL21(DE3)/pREP4/pETSK2. Overexpression of CoxM was induced with 2 mM isopropyl-␤-D-thiogalactopyranoside in LB medium at 20°C and at an A 600 of 0.7.
Assay of CO Dehydrogenase-The oxidation of CO or H 2 by CO dehydrogenase was coupled to the reduction of 50 M methylene blue (2), 100 M 1-phenyl-2-(4-iodophenyl)-3-(4-nitrophenyl)-2H-tetrazolium chloride (INT), 20 M 1-methoxyphenazine methosulfate (MPMS) (15), or 100 M 2,6-dichlorophenolindophenol in 50 mM NaH 2 PO 4 -NaOH (pH 7.2) and followed spectrophotometrically at 30°C employing ⑀ 615 of 37.11 mM Ϫ1 cm Ϫ1 (methylene blue), ⑀ 496 of 17.98 mM Ϫ1 cm Ϫ1 (INT/ MPMS), and ⑀ 600 of 16.10 mM Ϫ1 cm Ϫ1 (2,6-dichlorophenolindophenol). Assays of 1-ml total volume were flushed with pure CO or H 2 in screw-capped cuvettes provided with a rubber septum. Reactions were FIG. 1. Three-dimensional structure of native CO dehydrogenase (A), the reconstituted flavo (LMS) 2 species (B), its disordered active-site loop (C), and the postulated pathway of electron flow in native CO dehydrogenase (D). A, structure of native CO dehydrogenase adopted from Dobbek et al. (7) showing a dimer of two heterotrimers. The two molybdoproteins (L subunits) are presented in blue and light blue. They build up the dimer interface by head to head arrangement. Each molybdoprotein carries an MCD-type of molybdenum cofactor. The two iron-sulfur proteins (S subunits) are shown in same colors. The C-terminal domain, shown in orange, carries the type I 2Fe:2S center proximal to the molybdenum cofactor. The N-terminal domain, shown in red, carries the type II 2Fe:2S center distal to the molybdenum cofactor. The two flavoproteins (M subunits) are shown in the same colors. FAD is bound in a cleft formed by the N-terminal domain (shown in yellow) and the middle domain (shown in light green); the C-terminal domain is in dark green. B, structure of the flavo (LMS) 2 species reconstituted from the (LS) 2 species obtained from CO dehydrogenase, the recombinant M subunit produced in E. coli, and FAD. The figure shows one LMS monomer of the dimeric enzyme. The molybdoprotein, iron-sulfur protein, and flavoprotein are shown in blue, red, and yellow, respectively. Disordered amino acids are presented in silver. C, active site of native CO dehydrogenase showing the amino acids that are disordered in the reconstituted flavo (LMS) 2 species of B. The amino acids in the disordered sequence Gly 383 -Val-Ala-Tyr-Arg-Cys-Ser 389 are given as atom-colored stick models. D, postulated linear pathway of electron flow in native CO dehydrogenase, involving S-selanylcysteine 388, the molybdenum cofactor, the type I and II 2Fe:2S centers, and FAD. initiated by injection of 20 l of enzyme.
Preparation and Purification of Proteins-All purification steps were carried out below 4°C. For purification of CO dehydrogenase cell paste of CO-grown O. carboxidovorans (about 100 g wet mass) was suspended in 200 ml of 50 mM Hepes/NaOH (pH 7.2) (buffer A) containing 1 mM Na 2 EDTA, 0.2 mM phenylmethylsulfonylfluoride, and 5 mg of DNase I, disrupted in a high pressure homogenizer (Rannie AS), and subjected to low spin centrifugation, yielding crude extracts. Cytoplasmic fractions were prepared from crude extracts by ultracentrifugation for 2 h at 100,000 ϫ g. Anion exchange chromatography was on Source 30 Q (Amersham Pharmacia Biotech) equilibrated with buffer A. Cytoplasmic fractions (210 ml) were applied to the column (dimensions 15 cm by 5 cm) and eluted with 640 ml of buffer A followed by 1440 ml of a linear gradient of 0 to 1 M NaCl in buffer A. Fractions with CO dehydrogenase activity were pooled, supplemented with 1.3 M ammonium sulfate and 1 mM Na 2 EDTA, and gently stirred for 60 min. Precipitated protein was removed by low spin centrifugation, and the supernatant (220 ml) was loaded onto a hydrophobic interaction chromatography column (butyl-Sepharose 4 fast flow (Amersham Pharmacia Biotech) column (dimensions 6.5 cm by 5 cm) that has been equilibrated with 0.85 M ammonium sulfate in buffer A containing 1 mM Na 2 EDTA. Proteins were desorbed with 360 ml of equilibration buffer followed by 840 ml of a linear gradient combining decreasing ammonium sulfate concentrations (0.85 to 0 M) with increasing 2-propanol concentrations (0 to 20%, vol/vol) in buffer A. Fractions with CO dehydrogenase activity were pooled, desalted by gel filtration in buffer A, and concentrated by ultrafiltration. The purified enzyme was frozen in liquid nitrogen and kept at Ϫ80°C until use.
For the preparation of enzyme species devoid of the flavoprotein, typically 200 mg of CO dehydrogenase in 75 ml of buffer A containing 1% (mass/vol) SDS were incubated for 45 min at 30°C followed by gel filtration on Sephadex G-75 (Amersham Pharmacia Biotech, column dimensions 35 cm by 5 cm). Excluded CO dehydrogenase species were pooled, loaded onto Source 30 Q (column dimensions 12.5 cm by 2.6 cm), and eluted with 540 ml of a linear gradient of 0 to 1 M NaCl in buffer A. Purified CO dehydrogenase species with the subunit compositions (LMS) 2 , M(LS) 2 , and (LS) 2 proteins were pooled separately, concentrated, desalted, and stored as described above.
Recombinant apoflavoprotein was purified from E. coli BL21(DE3)/ pREP4/pETSK2 as follows. 12 g of cell paste suspended in 40 ml of buffer B (50 mM Tris/HCl, pH 8.7) containing 2 mg of DNase I were ultrasonicated (model Labsonic U, probe tip 40 T, Braun Biotech), subjected to ultracentrifugation (2 h at 100,000 ϫ g), loaded onto a Source 30 Q column (dimensions 12.5 cm by 2.6 cm), eluted with 180 ml of buffer B followed by 750 ml of a linear gradient of 0 to 0.5 M NaCl in buffer B, and gel filtered on Sephadex G-75 in buffer B containing 150 mM NaCl. Fractions were analyzed for the recombinant apoflavoprotein by SDS-PAGE and Western blotting, pooled, and stored on ice for less than 2 days until use.
Reconstitution of CO Dehydrogenase Species-For a typical reconstitution, 4 ml of the CO dehydrogenase species (LS) 2 (10 mg ml Ϫ1 ) or 6.8 ml of M(LS) 2 (10 mg ml Ϫ1 ) were mixed with 42 ml of the recombinant apoflavoprotein (0.4 mg ml Ϫ1 ) in buffer B supplemented with 150 mM NaCl and kept on ice for 60 min. Excess apoflavoprotein was removed by chromatography on Source 30 Q (for conditions, refer to purification of recombinant apoflavoprotein). Reconstituted CO dehydrogenase species were pooled, concentrated, desalted in buffer A and stored at Ϫ80°C. For reconstitution with FAD, typically 2 ml of reconstituted deflavo (LMS) 2 (3 mg ml Ϫ1 ) in buffer A were mixed with 0.22 ml of 5 mM FAD in the same buffer and kept on ice for 60 min. Excess FAD was removed by gel filtration on Sephadex G-25.
Determination of Association Constants and Association Rate Constants-Association constants for FAD with deflavo (LMS) 2 were determined from fluorescence quenching data at equilibrium binding (16,17). Association rate constants were determined assuming the simple model, where A is a deflavo LMS monomer (apoenzyme), F the FAD, and H is the flavo LMS monomer (holoenzyme). The equation describes the decrease of fluorescence (Fl). Since the concentrations of F and A exceeded the assumed K d by 3 orders of magnitude, the term of complex dissociation is negligible. For a reaction with second-order kinetics and initial concentrations of Association experiments were carried out at concentrations of 6.47, 3.24, and 1.61 M of F and A in buffer B containing 150 mM NaCl at 8°C, respectively.

Analysis of Pterins, Cofactor-bound Mononucleotides, Flavins, and Related
Compounds-Oxidation of MCD to form A with I 2 /KI followed published procedures (18). Mononucleotides were released from MCD or FAD by treatment of CO dehydrogenase species with sulfuric acid followed by HPLC analysis (19). Flavins were analyzed spectrophotometrically in supernatants of trichloroacetic acid precipitates (2). For estimation of flavin or nucleotide binding to reconstituted deflavo (LMS) 2 , CO dehydrogenase was incubated for 300 min in buffer A containing a 50-fold molar excess of the indicated compounds, followed by gel filtration. Bound nucleotides were released by boiling the proteins for 90 s with SDS (2%, mass/vol) in buffer A, followed by ultrafiltration (Microcon-3, Millipore) and analysis by isocratic anion exchange HPLC (ET 250/8/4 Nucleosil 100 -10 SB, Macherey and Nagel) employing 0.5 M NaH 2 PO 4 (pH 4.2) as the mobile phase at a flow rate of 1 ml min Ϫ1 .
Analytical PAGE was carried out according to Laemmli (22). For SDS-PAGE, 7.5% (mass/vol) acrylamide-stacking gels and 12% (mass/ vol) acrylamide-running gels were used. For nondenaturing PAGE, 7.5% (mass/vol) acrylamide-running gels were used. Protein staining was performed with Coomassie Brilliant Blue G-250. Protein transfer from acrylamide gels to polyvinylidene difluoride membranes and Nterminal amino acid sequencing were carried out as described (19). Immunoblot analysis was done with rabbit polyclonal IgG antibodies directed against the M subunit of O. carboxidovorans CO dehydrogenase. Primary IgG antibody binding was detected using goat anti-rabbit IgG alkaline phosphatase conjugate (Sigma). Isoelectric focussing was performed on IsoGel agarose plates (FMC, Bio-products) according to the instructions of the manufacturer. Selenium contents were estimated by neutron activation analysis (native CO dehydrogenase) or by inductively coupled plasma mass spectroscopy (CO dehydrogenase species M(LS) 2 and (LS) 2 ). Iron, molybdenum, and acid-labile sulfur were estimated spectrophotometrically as described (23)(24)(25).
CD spectra were recorded on a spectropolarimeter (J-600, Jasco). Contents of secondary structural elements were calculated as described (26).
X-band EPR spectra were recorded on a Bruker EMX spectrometer operated with a helium cryostat. The experimental conditions and determinations of g values were essentially as described before (27)(28)(29).
Reconstituted flavo (LMS) 2 species were crystallized under the same conditions effective for native CO dehydrogenase (7). Molecular replacement methods with data sets obtained from native CO dehydrogenase were used for structure determination.
Chemicals-Gases were purchased from Linde. All other chemicals employed were obtained from the usual commercial sources.

RESULTS AND DISCUSSION
The Recombinant M Subunit of CO Dehydrogenase Produced in E. coli Is Deflavo and Does Not Bind FAD-The medium (M) subunit of CO dehydrogenase was heterologously overexpressed. After induction of E. coli BL21(DE3)/pREP4/pETSK2 with isopropyl-␤-D-thiogalactopyranoside for at least 6 h, the recombinant protein comprised 30 to 40% of the total cell protein and comigrated with the authentic M subunit (Fig. 2). Induction at 20°C and a pH of 8.7 of the sonication buffer were essential to obtain maximum yields of the soluble recombinant M subunit (25% soluble and 75% aggregated).
The recombinant M subunit was purified 8-fold by ultracentrifugation, anion exchange chromatography, and gel filtration with an electrophoretic purity of 86% and a yield of 25% (Fig.  2B, lane 4). The protein turned out to be fragile and remained soluble only at concentrations below 1 mg ml Ϫ1 and temperatures below 8°C. The identity of the recombinant M subunit is apparent from comigration with the authentic M subunit upon SDS-PAGE (Fig. 2B, lanes 4 and 7) and an N terminus of MIPGSFDYHRPKSIADAVALLTKL, which is identical to that of the wild-type flavoprotein (5). The apparent molecular masses of the recombinant M subunit determined from SDS-PAGE (30 kDa, Fig. 2B, lane 4) or by gel filtration (41 kDa) indicate a monomeric subunit structure. The recombinant M subunit revealed a far-UV CD spectrum with a maximum at 190 nm and a trough extending from 210 to 225 nm, which is typical of ␣-helix/␤-sheet proteins and indicates that it was folded. The recombinant M subunit revealed the following percentages of secondary structural elements: ␣-helices (24 Ϯ 1), antiparallel ␤-sheets (20 Ϯ 1), parallel ␤-sheets (4 Ϯ 1), ␤-turns (22 Ϯ 1), and other elements (30 Ϯ 1). Its isoelectric point was 7.4. The recombinant M subunit was deflavo on the basis of its UV-visible absorption spectrum (protein concentration of 0.8 mg/ml, ⑀ 277 ϭ 19.6 mM Ϫ1 cm Ϫ1 ) or the spectrum of trichloroacetic acid supernatants. When recombinant M protein was incubated for 18 h at 4°C with a 50-fold molar excess of FAD or FMN and subjected to gel filtration, and the protein fraction was analyzed spectrophotometrically, no flavins were identified. These data indicate that the recombinant M subunit is devoid of flavins and has no affinity for FAD or FMN.
Removal of the M Subunit from Native CO Dehydrogenase and Characterization of the Resulting Protein Species-The anionic detergent SDS was found to remove specifically one or both M subunits from CO dehydrogenase, resulting in the formation of proteins of different mobility, designated band I to IV (Fig. 3A). The formation of the band I to IV proteins was paralleled by a time-dependent, gradual loss of CO dehydrogenase activity (Fig. 3B). The 277-kDa band (Fig. 3, band I) represented catalytically competent CO dehydrogenase. It was converted into an inactive 217-kDa protein (Fig. 3, band III) with the intermediate formation of a partially active 247-kDa protein (band II). The data indicate a sequential release of the two 30-kDa M subunits (Fig. 3, band IV) from native (LMS) 2structured CO dehydrogenase, resulting in the formation of an intermediate M(LS) 2 species (band II) and a final (LS) 2 species (band III). Band IV (Fig. 3) was identified as the M subunit on the basis of its correct N terminus of MIPGSFDY (5) and migration as a 30-kDa polypeptide upon SDS-PAGE (data not shown). The low mobility of the M subunit (Fig. 3, band IV) indicates that it became partially or fully unfolded under the experimental conditions applied.
The purified band I protein (Fig. 2C, lane 1) had a (LMS) 2 subunit structure on the basis of 89-, 30-, and 18-kDa polypeptides appearing at a 1:0.8:0.8 molar ratio upon SDS-PAGE (Fig.  2B, lane 1) and represents native CO dehydrogenase. The pu-rified band II protein (Fig. 2C, lane 2) had an M(LS) 2 subunit structure on the basis of 89-, 30-, and 18-kDa polypeptides appearing at a 1:0.5:0.9 molar ratio (Fig. 2B, lane 2). The purified band III protein (Fig. 2C, lane 3) had a (LS) 2 subunit structure on the basis of 89-and 18-kDa polypeptides appear- ing at a 1:0.9 molar ratio (Fig. 2B, lane 3). From now on we are referring to the homogeneous proteins of band II and III as M(LS) 2 and (LS) 2 , respectively. Native CO dehydrogenase and the M(LS) 2 and (LS) 2 species eluted from the anion exchanger (Source 30 Q) at 0.29, 0.33, or 0.35 M NaCl, respectively, referring to an increase in the net overall negative charges.
Analyses of native CO dehydrogenase, M(LS) 2 , and (LS) 2 by the trichloroacetic acid precipitation revealed a 1:1 molar correlation between the M subunits and FAD (Table I). The contents of molybdenum, 5Ј-CMP, and form A indicate a 1:1 mononuclear complex of molybdenum and MCD in each of the three proteins (Table I). The contents of iron and acid-labile sulfur agree with the presence of both 2Fe:2S centers in each of the three proteins (Table I).
The spectrum of the (LS) 2 species is characterized by a complete loss of the FAD absorption maxima at 387 and 449 nm (Fig. 4A, spectrum c) and is very similar to that of the Mop protein from D. gigas (30), which resembles (LS) 2 in containing a 1:1 mononuclear complex of molybdenum and MCD, the 2Fe:2S centers, and no FAD (31,32). The shoulder at 387 nm in the spectrum of native CO dehydrogenase originates from the short wavelength absorption maximum of FAD (Fig. 4A, spectrum a). The extended iron-sulfur absorption that centers around 550 nm was the same in the spectra of native CO dehydrogenase, M(LS) 2 , and (LS) 2 (Fig. 4A, spectra a to c). The spectrum of the flavoprotein of CO dehydrogenase, i.e. the M subunit with the bound FAD, can be obtained from the difference spectra of native CO dehydrogenase minus (LS) 2 (Fig. 4A,  spectrum d) or native CO dehydrogenase minus M(LS) 2 (Fig.  4A, spectrum e). The shoulder at 478 nm in the flavoprotein spectrum (Fig. 4A, spectra d and e) is not apparent with free FAD in solution and, thus, originates from cofactor-protein interactions.
We have shown that controlled dissociation of CO dehydrogenase with SDS produces a rather rigid deflavo (LS) 2 substructure from which both M subunits have been removed (Figs. 2 and 4). The noncovalent interactions of the individual L and S subunits and of two LS monomers are apparently rather tight and are contrasted by only weak binding of the M subunits to (LS) 2 . The existence of a discrete M(LS) 2 species containing FAD at a 1:1 molar ratio (Figs. 2 and 4) and exhibiting about 50% full enzymic activity (Table I) indicates that each LMS heterotrimer in CO dehydrogenase represents an independent catalytic unit as concluded before from a molybdenum to molybdenum distance of 53 Å (7). The hierarchy of subunit interactions resolved here for CO dehydrogenase is also apparent with Mop from D. gigas. Mop is composed of two covalently linked protein domains that are equivalent to the L and the S subunits of CO dehydrogenase. The flavodoxin, employed by Mop as electron acceptor, is not a permanent constituent of the enzyme (33). (LS) 2 derived from CO dehydrogenase and Mop show high sequence similarities in their small subunit or domain (70%, S of CO dehydrogenase and amino acids 1 to 166 of Mop) as well as their large subunit or domain (58%, L of CO dehydrogenase and amino acids 167 to 809 of Mop). Circular dichroism measurements on the (LS) 2 species of CO dehydrogenase revealed a content of 30% ␣-helices and 25% ␤-sheets, which compares to the content of secondary structural elements of Mop (28% ␣-helices, 21% ␤-sheets (32)) and agrees with a very similar fold of Mop and LS in (LMS) 2structured CO dehydrogenase (7).
Reconstitutions of M(LS) 2 (Fig. 2C, lane 2) with the recombi-   2 were produced by controlled dissociation of native CO dehydrogenase with SDS; the flavo (LMS) 2 species was obtained by reconstitution of (LS) 2 with the recombinant M subunit and FAD. The contents of form A are given relative to native CO dehydrogenase, set to 100; all other contents are in mol/mol of protein; values with S.D. are the mean from three independent determinations. DCPIP, 2,6-dichlorophenolindophenol. nant M subunit (Fig. 2C, lane 4) resulted in the formation of a (LMS) 2 species (Fig. 2, B and C, lanes 5), which carried 1 mol of FAD/mol of protein originating from the M(LS) 2 species. In mixtures of equimolar amounts of the deflavo (LMS) 2 species and FAD, the fluorescence emission of free FAD (Fig.  5A, trace a) was quenched (Fig. 5A, trace b). The extent of fluorescence quenching upon the reconstitution of the deflavo (LMS) 2 species with FAD was proportional to the molar ratios of (LMS) 2 and FAD. The quenching of FAD fluorescence reflects the binding of FAD to deflavo (LMS) 2 since the fluorescence emission characteristic of free FAD (Fig. 5A, trace a) is absent in native CO dehydrogenase (Fig. 5A, trace c). Reconstitution of the deflavo (LMS) 2 species with FAD led to the integration of 2 mol of FAD/mol of protein (Fig. 4B, spectra g  and h). The data indicate that the deflavo (LMS) 2 species can be reconstituted with free FAD, yielding a flavo (LMS) 2 species. In this respect the deflavo (LMS) 2 species is similar to the deflavo species of milk xanthine oxidase or chicken liver xanthine dehydrogenase produced by treatment with 2 M CaCl 2 or 3 M KI (34,35).
Since the recombinant M subunit was not able to bind FAD, it is apparent that the flavin binding site on M is only functional in the fully assembled CO dehydrogenase exhibiting the complete heterotrimeric subunit structure. As will be shown later, the assembly of the M subunit with (LS) 2 (Fig. 5A, inset) and passed through the origin, which indicates that FAD binding follows true second-order kinetics. The association rate constant k ϩ1 of 1350 Ϯ 150 M Ϫ1 s Ϫ1 is 7-to 740-fold lower than that of other flavoenzymes (36). The (LS) 2 species showed no detectable affinity for FAD and quenched the FAD fluorescence only slightly due to its absorbance at 450 nm and the resulting inner filter effect. The addition of the recombinant M subunit to an FAD-containing solution led to a rapid increase of FAD fluorescence (Fig. 5B, traces a and b), suggesting interactions between the two compounds that can be explained by destabilization of the intramolecular complex between the adenine and the isoalloxazine moieties of FAD in solution (refer to Mü ller (37) for a discussion of FAD fluorescence). As the M subunit and FAD were completely separated upon gel filtration, they do not form a stable complex. If it is true that the FAD interacts with the flavin binding site and not with other parts of the recombinant M subunit, then the flavin binding site must be in an open conformation since a closed conformation would not allow the observed interactions. The addition of the (LS) 2 species to a mixture of the recombinant M subunit and FAD resulted in rapid quenching of FAD fluorescence (Fig.  5B, traces b and c), indicating again that both the M subunit as well as the (LS) 2 species are required for FAD binding.
The association constant of FAD with deflavo (LMS) 2 of 1.0 ϫ 10 9 M Ϫ1 (K d ϭ 1.0 ϫ 10 Ϫ9 M), determined from the quenching of flavin fluorescence in equilibrium binding experiments, groups CO dehydrogenase among the most efficient noncovalently flavin-binding proteins (36).
Binding of Compounds other than FAD to the Deflavo (LMS) 2 Species-To check the ability of the deflavo (LMS) 2 species to integrate compounds representing structural elements of FAD, the protein was incubated for 300 min with a 50-fold molar excess of these compounds and analyzed for bound compounds. Under these conditions the following relative ranking of binding was observed (mol of compound/mol of (LMS) 2 ): FAD (2.10 Ϯ 0.06) Ͼ ADP (1.94 Ϯ 0.10) Ͼ FMN (1.52 Ϯ 0.02) Ͼ ATP (1.26 Ϯ 0.05). Riboflavin, AMP, CDP, and GDP did not bind, since less than 0.05 mol of these compounds were resolved/mol of (LMS) 2 . That the strongest interactions of FAD with the protein involve the pyrophosphate linkage is indicated by the ability of deflavo (LMS) 2 to bind stoichiometric amounts of ADP and substoichiometric amounts of ATP and FMN and the inability to bind riboflavin and AMP. The isoalloxazine ring, the ribitol, and the mononucleotide provide only weak or no interactions at all. The adenine moiety apparently determines the specificity for FAD since deflavo (LMS) 2 has no affinity for CDP or GDP. The binding of 0.99 Ϯ 0.08 mol of 8-azido-ATP/mol of (LMS) 2 offers the possibility to establish photochemical crosslinks. Native CO dehydrogenase exhibited no binding affinity for ADP or ATP, and both compounds were not hydrolyzed by native CO dehydrogenase or the deflavo (LMS) 2 species.
EPR of the Type II Iron Sulfur Center Reports the Interaction of the Subunits M and S in CO Dehydrogenase-The 2Fe:2S center in the N-terminal domain of the S subunit is located at the interface that interacts with the M subunit and is exposed to the solvent (Fig. 1A). The C-terminal 2Fe:2S center is located at the interface of the subunits L and S and is buried in the protein structure 11 Å below the surface of the protein (Fig.  1A). It is, therefore, immediately obvious that the environment of the N-terminal iron-sulfur center is likely to be altered by the presence or absence of the M subunit, whereas the Cterminal center is unlikely to be affected. EPR has been proven a most suitable technique for the analysis of the two iron-sulfur centers in CO dehydrogenase (27)(28)(29). Fig. 6A shows the EPRspectra of native CO dehydrogenase. At 49 K almost only the signals originating from the type I 2Fe:2S center were developed (g 1 ϭ 2.024, g 2 ϭ 1.947, g 3 ϭ 1.901, g av ϭ 1.957), whereas at 16 K, a combined EPR spectrum originating from the type I (g 1 ϭ 2.025, g 2 ϭ 1.947, g 3 ϭ 1.900, g av ϭ 1.957) and type II (g 1 ϭ 2.159) 2Fe:2S centers was obtained. Fig. 6B shows the ironsulfur EPR spectra of the (LS) 2 species, which differ from those of native CO dehydrogenase depicted in Fig. 6A. Fig. 6B shows at 49 K a spectrum that originates from the type I and type II 2Fe:2S centers. The signals of the (LS) 2 protein at g 1 ϭ 2.024, g 2 ϭ 1.948, g 3 ϭ 1.903, g av ϭ 1.958 are the same as in native CO dehydrogenase (Fig. 6A, 49 K) and must, therefore, be ascribed to the type I 2Fe:2S center, whereas the signal at g ϭ 2.048 is not visible in native CO dehydrogenase (Fig. 6A, 49 K). It is apparent that the removal of the M subunits has no effect on the EPR properties of the type I center in (LS) 2 . Fig. 6C shows the EPR difference spectrum obtained by subtracting various amounts of the pure type I spectrum of native CO dehydrogenase at 49 K (Fig. 6A) from the combined EPR spectrum of the (LS) 2 species at 49 K (Fig. 6B). The resulting EPR difference spectrum of Fig. 6C is rhombic and originates from the type II 2Fe:2S center of (LS) 2 . The g values (g 1 ϭ 2.048, g 2 ϭ 1.949, g 3 ϭ 1.915, g av ϭ 1.971) of the difference spectrum are similar to those reported for the type II 2Fe:2S center in Mop (g 1 ϭ 2.057, g 2 ϭ 1.970, g 3 ϭ 1.900, g av ϭ 1.976 (30,38)), and the temperature dependence of the difference spectrum of the (LS) 2 species (Fig. 6C) and Mop is the same. The type II signal of (LS) 2 was fully sharpened at 43 K and scarcely detectable at 60 K. In Mop, the type II signal was fully sharpened at 45 K and scarcely detectable at 65 K to 70 K (30,38). In contrast, the type II signal of native CO dehydrogenase was fully sharpened at 16 K and scarcely detectable at 26 K. The data can be explained by assuming that the removal of the M subunits from native CO dehydrogenase leads to a Mop-like environment of the type II center in (LS) 2 . Fig. 6E shows the EPR spectra of the reconstituted flavo (LMS) 2 species at 49 K (g 1 ϭ 2.024, g 2 ϭ 1.951, g 3 ϭ 1.900, g av ϭ 1.958) and 16 K (type I: g 1 ϭ 2.024, g 2 ϭ 1.950, g 3 ϭ 1.900, g av ϭ 1.958, type II: g 1 ϭ 2.165). The spectra are almost indistinguishable from those of native CO dehydrogenase under the same conditions (Fig. 6A), indicating that the Mop-like type II 2Fe:2S center in (LS) 2 has been rearranged to that of native CO dehydrogenase.
The type II 2Fe:2S centers are also weak reporters of FAD binding to deflavo (LMS) 2 . This can be seen by comparing the EPR spectra shown in Fig. 6, D and E, which show at 16 K a shift of the type II signal at g 1 ϭ 2.190 in deflavo (LMS) 2 (Fig.  6D) to g 1 ϭ 2.165 in flavo (LMS) 2 (Fig. 6E). The shift can be explained by an induced fit-like mechanism of flavin binding to the M subunit, where cleft closure effects conformational changes in the environment of the type II center. The FAD binding cleft is formed by the N-terminal and the middle domain of the M subunit (Fig. 1A), where the strongest interactions originate from the FAD pyrophosphate with the doubleglycine motifs Ala 32 -Gly-Gly-His-Ser 36 in the N-terminal and Thr 111 -Ile-Gly-Gly-Asn 115 in the middle domain (7). Since both domains also contribute to the interface interacting with the S subunit, it is likely that they report cleft closure to the Nterminal 2Fe:2S center.
The data discussed so far show that interactions of the proteins (LS) 2 and M are reported only by the type II 2Fe:2S center and have no effect on the type I center. It is apparent from the three-dimensional structure of CO dehydrogenase that the M subunit binds to the N-terminal domain of the S subunit ( Fig.   FIG. 6. Binding of the M subunit and/or FAD to (LS) 2 ; analysis by iron-sulfur and molybdenum EPR. The indicated CO dehydrogenase species (36 M in 50 mM Hepes/NaOH, pH 7.2) were reduced with 4 mM dithionite (A to E) or incubated for 3.25 h with pure CO (F to H) under anoxic conditions and frozen in liquid nitrogen. In C, the difference spectrum of (B minus 0.8 ϫ A) at 49 K is given. EPR spectra were recorded with a microwave frequency of 9.47 GHz and a modulation amplitude of 1 millitesla. EPR spectra indicative of Mo(V) and FADH q (g ϭ 2.005) were run at 120 K and a microwave power of 10 mW, and iron-sulfur EPR were at 49 K and a microwave power of 10 mW or at 16 K and a microwave power of 200 mW.