The Radical S-Adenosyl-l-methionine Enzyme QhpD Catalyzes Sequential Formation of Intra-protein Sulfur-to-Methylene Carbon Thioether Bonds*

Background: The small subunit of quinohemoprotein amine dehydrogenase contains three Cys-to-Asp/Glu thioether bonds. Results: The radical S-adenosyl-l-methionine (SAM) enzyme QhpD catalyzes the single-turnover reaction of thioether bond formation in the protein substrate. Conclusion: The thioether bond formation by QhpD proceeds sequentially in an N- to C-terminal direction of the polypeptide. Significance: Our findings uncover another challenging reaction of radical SAM superfamily of enzymes. The bacterial enzyme designated QhpD belongs to the radical S-adenosyl-l-methionine (SAM) superfamily of enzymes and participates in the post-translational processing of quinohemoprotein amine dehydrogenase. QhpD is essential for the formation of intra-protein thioether bonds within the small subunit (maturated QhpC) of quinohemoprotein amine dehydrogenase. We overproduced QhpD from Paracoccus denitrificans as a stable complex with its substrate QhpC, carrying the 28-residue leader peptide that is essential for the complex formation. Absorption and electron paramagnetic resonance spectra together with the analyses of iron and sulfur contents suggested the presence of multiple (likely three) [4Fe-4S] clusters in the purified and reconstituted QhpD. In the presence of a reducing agent (sodium dithionite), QhpD catalyzed the multiple-turnover reaction of reductive cleavage of SAM into methionine and 5′-deoxyadenosine and also the single-turnover reaction of intra-protein sulfur-to-methylene carbon thioether bond formation in QhpC bound to QhpD, producing a multiknotted structure of the polypeptide chain. Homology modeling and mutagenic analysis revealed several conserved residues indispensable for both in vivo and in vitro activities of QhpD. Our findings uncover another challenging reaction catalyzed by a radical SAM enzyme acting on a ribosomally translated protein substrate.

The bacterial enzyme designated QhpD belongs to the radical S-adenosyl-L-methionine (SAM) superfamily of enzymes and participates in the post-translational processing of quinohemoprotein amine dehydrogenase. QhpD is essential for the formation of intra-protein thioether bonds within the small subunit (maturated QhpC) of quinohemoprotein amine dehydrogenase. We overproduced QhpD from Paracoccus denitrificans as a stable complex with its substrate QhpC, carrying the 28-residue leader peptide that is essential for the complex formation. Absorption and electron paramagnetic resonance spectra together with the analyses of iron and sulfur contents suggested the presence of multiple (likely three) [4Fe-4S] clusters in the purified and reconstituted QhpD. In the presence of a reducing agent (sodium dithionite), QhpD catalyzed the multiple-turnover reaction of reductive cleavage of SAM into methionine and 5-deoxyadenosine and also the single-turnover reaction of intra-protein sulfur-to-methylene carbon thioether bond formation in QhpC bound to QhpD, producing a multiknotted structure of the polypeptide chain. Homology modeling and mutagenic analysis revealed several conserved residues indispensable for both in vivo and in vitro activities of QhpD. Our findings uncover another challenging reaction catalyzed by a radical SAM enzyme acting on a ribosomally translated protein substrate.
Quinohemoprotein amine dehydrogenase (QHNDH) 2 is a bacterial enzyme that catalyzes oxidative deamination of various aliphatic primary amines for use as energy, carbon, and nitrogen sources (1)(2)(3). Although the enzyme was originally reported to occur only in a limited number of Gram-negative bacteria (2,3), our recent survey with the updated genome sequence database revealed a wide distribution of the enzyme not only in more than 200 Gram-negative species but also in some Gram-positive bacteria with the associated genes constituting an operon designated "qhp" (4). QHNDH consists of three nonidentical subunits, with each subunit exhibiting distinct structural features ( Fig. 1A) (5,6). The large subunit of 60 kDa, termed ␣ and encoded by the qhpA gene, has four domains with two hemes that mediate electron transfer from the substrate to an external electron acceptor, such as cytochrome c 550 and azurin (2,3), contained in the N-terminal di-heme cytochrome c peroxidase-like domain. The mid-sized subunit of 36 kDa, termed ␤ and encoded by the qhpB gene, has a ␤-propeller structure that is highly conserved in quinoprotein dehydrogenases. The smallest subunit of ϳ9 kDa, termed ␥ and encoded by qhpC, has an unprecedented protein structure with four internal thioether cross-links (Fig. 1B). One of the cross-links (Cys-37-Trp-43) is formed between the sulfur atom of a Cys residue and the indole group of a Trp residue that is specifically modified to contain ortho-quinone groups, building a peptidyl quinone cofactor, cysteine tryptophylquinone (CTQ). Three other thioether cross-links (Cys-7-Glu-16, Cys-27-Asp-33, and Cys-  are formed between the sulfur atom of a * This work was supported by Japan Society for the Promotion of Science gap-CX 2 CX 5 CX 3 C-gap-C) in the C-terminal half (15), which are likely involved in binding of auxiliary [4Fe-4S] clusters (16,17). Moreover, multiple sequence alignment of QhpD and its homologs with other SPASM proteins (Fig. 2) reveals strict conservation of several residues throughout the subgroup and also some residues only among QhpD homologs that constitute the QHNDH maturation protein family (9), in addition to the N-terminal signature CX 3 CX 2 C motif that ligates the SAMbinding [4Fe-4S] cluster (10 -12). These SPASM proteins are primarily involved in cofactor and peptide maturation processes, acting on ribosomally translated proteins (peptides) encoded in the vicinity of their genes (14).
In view of the reactions catalyzed, QhpD resembles AlbA and some other related enzymes (SkfB, ThnB, and TrnC/D) that catalyze the regiospecific formation of the sulfur-to-␣-carbon thioether bonds in the biosynthesis of sactipeptides (18 -20). However, QhpD differs from these proteins in that it acts on an enzyme subunit to form a multiknotted structure of the polypeptide chain with sulfur-to-methylene carbon thioether crosslinks (7). In this study, we describe efficient expression of recombinant QhpD from Paracoccus denitrificans in complex with its substrate QhpC and biochemical characterization as a radical SAM enzyme, and we demonstrate that QhpD does catalyze the formation of intra-protein sulfur-to-methylene carbon thioether bonds in QhpC. Based on the modeled structure for the QhpC⅐QhpD complex, we propose a possible mechanism of the sequential formation of multiple thioether bonds, FIGURE 1. Crystal structure of quinohemoprotein amine dehydrogenase. A, ribbon model of QHNDH from P. denitrificans (PDB code 1JJU) with ␣, ␤, and ␥ subunits colored green, yellow, and magenta, respectively. Hemes and CTQ are drawn as ball-and-stick models. B, enlarged view of the ␥ subunit in an orientation different from A. CTQ and thioether cross-links are shown as ball-and-stick models. This figure was prepared using MOLSCRIPT (49) and Raster3D (50). C, schematic structure of ␥ subunit is shown with the stereochemistry of the thioether bonds and the amino acid residues in a single-letter code.
which uncovers another challenging reaction catalyzed by a radical SAM enzyme belonging to the SPASM subgroup.

EXPERIMENTAL PROCEDURES
Expression and Purification of the QhpC⅐QhpD Complex-Expression plasmids for QhpC and QhpD from P. denitrificans Pd1222 and their derivatives were constructed using either an Escherichia coli expression vector pET-15b or a broad host range vector pBBR1, mostly according to standard molecular genetic protocols (Fig. 3). In addition, a hexa-His (H 6 ) tag or a Twin-Strep (St 2 ) tag was appended on the N terminus of QhpD or the C terminus of QhpC to aid purification of the QhpC⅐QhpD complex and analysis of their interaction. PCR primers used in construction of these plasmids as well as those for site-specific mutants of QhpC and QhpD are summarized in Table 1. Site-specific mutants were obtained by the PCR-based site-directed mutagenesis method. Codon replacements and the absence of PCR-derived errors, if any, were confirmed by sequencing the entire coding regions. E. coli C41 (DE3) cells carrying expression plasmids for QhpD and/or QhpC were grown at 37°C by reciprocal shaking at 180 rpm in LB medium (1% (w/v) bactotryptone, 0.5% (w/v) yeast extract, and 0.5% (w/v) NaCl) supplemented with the appropriate antibiotics (50 g/ml ampicillin, 30 g/ml streptomycin, and 10 g/ml tetracycline). When the cell density reached an OD 600 of ϳ0.5, 0.17 mg/ml ammonium ferric citrate and isopropyl ␤-D-thiogalactopyranoside to a final concentration of 0.1 mM were added, and the cells were further grown at 30°C for 6 h. The cells were harvested by centrifugation at 5,000 ϫ g for 10 min and stored at Ϫ20°C until use.
For purification of the QhpC⅐QhpD complex, such as the complex between a short version of QhpC (QhpC (Ϫ28)Ϫ23 ; hereafter designated sQhpC) and QhpD, frozen cells were suspended in 50 mM potassium phosphate buffer, pH 7.8, contain-        (17) are shown above the alignment; the region including Gln-345-Leu-360 of Pd_QhpD was re-aligned manually so that Cys-353 of QhpD corresponds to Cys-261 of anSMEcpe to coincide with the QhpD structure model, described later (Fig. 15). Identical and highly conserved residues are shown by white letters on red background and red letters boxed in blue, respectively. Mutated residues of Pd_QhpD are indicated with green triangles above the alignment. Numbers in orange circles shown below the alignment represent Cys residues ligating the auxiliary clusters (Aux) in anSMEcpe (Aux I: 1, 2, 3, and 7; Aux II: 4, 5, 6, and 8) (17). The figure was produced using ESPRIPT (52).

Bs_AlbA 1 M F I E Q M F P F I N E S V R V H Q L P E G G V L E I D Y L R D N V S I S D F E Y L D L N K T A Y E L C M R M D G Q K T A E Q I L
ing 300 mM NaCl, 1 mM sodium dithionite (DT), and cOmplete Protease Inhibitor (EDTA-free; Roche Applied Science) (1 tablet/50 ml of cell suspension), and disrupted by ultrasonic oscillation. The cell extract was obtained by centrifugation at 20,000 ϫ g for 1 h. All subsequent procedures were carried out in a vacuum-type glovebox (AS ONE Corp.) filled with 99.999% (v/v) argon gas; oxygen concentration was routinely maintained below 0.01% (v/v) throughout the purification by monitoring with an oxygen sensor. The stock solution of 100 mM DT in 20 mM Tris-HCl, pH 8, was also prepared freshly in the glovebox. The cell extract was applied to a nickel-chelate column ( To prevent thioether bond formation in sQhpC occurring in E. coli cells during expression of the sQhpC⅐QhpD complex (see below), Cys-7 (the sulfur donor for the Cys-7-Glu-16 thioether bond) in sQhpC was replaced by Ser (sQhpC C7S ) (Fig.  3), which did not affect expression and purification of the complex. In addition, for size discrimination from the wild-type QhpC, sQhpC C7S was expressed occasionally as a C-terminally fusion protein with GST (sQhpC C7S -GST).
For purification of the full-length QhpC (QhpC (Ϫ28)Ϫ82 ; designated fQhpC) without thioether bonds that are partially formed in E. coli cells when expressed as the complex with QhpD, fQhpC was first co-expressed with an inactive mutant of QhpD (QhpD C116S ) and then the fQhpC⅐QhpD C116S complex was purified by nickel-chelate affinity chromatography as above. Subsequently, the eluted solution was incubated at 70°C for 20 min to denature the QhpD C116S protein; fQhpC is heatstable. After standing on ice for 20 min, the precipitated QhpD C116S was removed by centrifugation. The fQhpC remaining in the supernatant was further purified using a Strep-trap column as described above under atmospheric conditions. Site-specific mutants of fQhpC (fQhpC E16Q , fQhpC D33N/D49N , and fQhpC D49N ) that do not contain preformed thioether bonds were also obtained by the same method. The fQhpC⅐QhpD complex, containing no thioether bonds in fQhpC, was prepared by anaerobically incubating the 5Ј-GCTCTAGATTTTGGAGGATGGTCGCCACCACCAAACG-3Ј XbaI a Plus and minus signs in parentheses denote sense and antisense strands, respectively. b Restriction sites are underlined. Mismatching nucleotides at the mutated sites are shown in bold. APRIL 24, 2015 • VOLUME 290 • NUMBER 17 sQhpC C7S -GST⅐QhpD complex (60 M), which was expressed and purified as described above, with 20 M fQhpC containing no thioether bonds prepared as above for more than 6 h (to exchange sQhpC C7S -GST with fQhpC). The mutant fQhpC⅐QhpD complex was also obtained by the same method.

Enzymatic Formation of Intra-protein Thioether Bonds
In these experiments, the concentration of fQhpC added was kept lower than that of the complex so that most the fQhpC was bound to QhpD.
Reconstitution-QhpD in the as-purified complex was reconstituted essentially according to the published procedure (18). Briefly, the complex (protein concentration, 0.1-0.2 mM) eluted from the Strep-trap column was incubated with 100 eq of DTT (final concentration, 10 -20 mM) for 1 h on ice in the glovebox. Then 10 eq of ammonium ferric citrate (final concentration, 1-2 mM) was slowly added under careful mixing, and the mixture was incubated for 5 min on ice. Finally, 10 eq of lithium sulfide (final concentration, 1-2 mM) was added, and the mixture was incubated overnight at 4°C in a sealed bag with gentle rotary shaking. The resulting solution was filtrated through a 0.22-m membrane and then desalted with a PD-10 column (GE Healthcare) equilibrated with buffer B containing 1 mM DTT. The reconstituted QhpC⅐QhpD complex thus obtained was stored at 4°C in a sealed bag containing an oxygen absorber for a few days without appreciable loss of SAM cleaving and thioether bond forming activities.
Determination of Protein Concentration and Iron and Sulfur Contents-Protein concentrations of the as-purified QhpC⅐ QhpD complex were determined by the Bradford assay (Nacalai Tesque) using BSA as the standard. For determination of the Bradford correction factor (21), the sQhpC⅐QhpD complex (4.06 Ϯ 0.21 mg/ml) was first precipitated with 10% (w/v) TCA at Ϫ20°C for 30 min and washed twice with cold acetone to remove the bound [Fe-S] clusters. The precipitates dried in air were then dissolved in a small volume of 8 M urea (3.93 Ϯ 0.02 mg/ml; recovery, 97%). As the urea solution exhibited a spectrum of simple protein without absorption above 350 nm (data not shown), the protein concentration was calculated to be 3.23 mg/ml from the absorbance at 280 nm using the molar extinction coefficient (52,370 M Ϫ1 cm Ϫ1 , assuming all Cys residues are reduced) obtained by the web-based tool, ProtParam, in ExPASy (22). The Bradford correction factor (3.23/3.93 ϭ 0.822) thus obtained was used for all the subsequent enzyme characterizations. The iron and sulfur contents were determined according to published procedures (23,24).
Spectral Measurements-Solutions of the as-purified and reconstituted QhpC⅐QhpD complexes were prepared anaerobically in a 1-ml quartz cuvette with a gas-tight screw cap within the glovebox. UV-visible absorption spectra were recorded at 25°C with an Agilent 8453 photodiode array spectrophotometer.
For measurement of EPR spectra, the reconstituted QhpC⅐ QhpD complex (80 M in buffer B) was reduced with 2 mM DT for 5 min and transferred to EPR tubes, which were sealed with screw caps, in the glovebox. The samples were immediately flash-frozen and kept in liquid nitrogen until EPR analysis. X-band (9.23 GHz) microwave frequency EPR spectra were recorded at 15 K on a Varian E-109 EPR spectrometer with 100-kHz field modulation. The microwave frequency was cali-brated with a microwave frequency counter (Takeda Riken Co., Ltd., model TR5212). An Oxford flow cryostat (ESR-900) was used to achieve temperatures down to 15 K.
Assay for the SAM Cleavage Reaction-The reductive SAM cleavage activity of QhpD was measured for the sQhpC⅐QhpD and sQhpC C7S ⅐QhpD complexes (63 M) under strictly anaerobic conditions maintained in the glovebox. The reconstituted QhpC⅐QhpD complex in buffer B containing 1 mM DTT was first reduced with a 10-fold molar excess of DT for 10 min and then reacted with 1 mM SAM (Sigma) at an ambient temperature. The reaction was quenched by the addition of 10 mM K 3 [Fe(CN) 6 ] at 10 -300 min after the addition of SAM, and the precipitated protein was removed by centrifugation. For quantification of the 5Ј-deoxyadenosine (5Ј-dA) produced, a 10-l aliquot of the supernatant was injected into a Cosmosil 5C 18 -PAQ reversed-phase column (Nacalai Tesque) attached to a Hitachi L-7000 HPLC system with an L-7420 UV detector. Elution was performed with 0.1% (w/v) KH 2 PO 4 for 5 min followed by a linear gradient to 0.1% (w/v) KH 2 PO 4 containing 40% (v/v) methanol for 15 min and then with the latter solution for 5 min at a flow rate of 0.5 ml/min, with the absorbance monitored at 260 nm. Under these conditions, SAM and 5Ј-dA are eluted at 6.8 and 24.5 min, respectively. The concentration of 5Ј-dA was determined by comparison with the integrated peak areas of 5Ј-dA standards at known concentrations. The eluate collected at the 5Ј-dA peak was also analyzed by TOF-MS (Accu TOF, JMS-T100LC, JEOL) equipped with a direct-analysis-in-realtime ion source (MS-54141 direct-analysis-in-real-time, JEOL).
For determination of the methionine produced, the supernatant of the reaction mixture (20 l) was diluted 10-fold with 100 mM sodium borate, pH 9.3, containing 1 mM EDTA and incubated with 50 mM 4-(N,N-dimethylaminosulfonyl)-7-fluoro-2,1,3-benzoxadiazole (DBD-F) (Dojindo) at 50°C for 30 min to convert the methionine into a fluorescent derivative (DBD-methionine). A 20-l aliquot of the reaction mixture was injected into a Cosmosil 5C 18 -PAQ reversed-phase column attached to the L-7000 HPLC system with a Hitachi F-1050 fluorescence detector. Elution was done with 0.1% (v/v) TFA for 5 min, and then a linear gradient to 0.08% (v/v) TFA containing 80% (v/v) acetonitrile for 15 min at a flow rate of 0.5 ml/min, monitored by excitation at 450 nm and emission at 590 nm. Under these conditions, the DBD-methionine was eluted at ϳ18 min. The concentration of methionine was determined by comparison with the integrated peak areas of methionine standards at known concentrations that had been treated with DBD-F as described above.
Modification of free cysteine residues with N-(9-acridinyl)maleimide (NAM) (Tokyo Chemical Industry) was carried out as follows. The dried residue obtained by precipitation of the reaction mixture (50 l) with 10% (w/v) TCA, as described above, was dissolved in 50 l of 1% (w/v) SDS in 50 mM potassium phosphate, pH 7.5, containing 1 mM tris(2-carboxyethyl)phosphine and incubated at 95°C for 5 min. To 10 l of the solution was added 1 l of 5 mM NAM in acetone, and the mixture was left to stand at room temperature for 1 h. After the addition of 3 l of 50% (w/v) glycerol, the sample was analyzed by SDS-PAGE. A UV transilluminator (ChemiDoc XRS system, Bio-Rad) was used to detect fluorescent bands by excitation at 355 nm and emission at 465 nm.
Structure Modeling-The structure model of QhpD was built by applying the P. denitrificans QhpD amino acid sequence (gi: 119384441) to the website SWISS-MODEL (25), and the resultant initial model was modified manually by Coot (26). The model of the QhpC leader peptide without structural data was also built by Coot (26) initially as an ideal ␣-helix on the basis of secondary structure predictions by Jpred (27), GOR (28), and PSIPRED (29), which was then slightly bent around Gly(Ϫ8) to fit to the shape of a groove of the QhpD model. The modeled leader peptide was then connected at the N terminus of the core peptide of QhpC, the structure of which was extracted from the QHNDH crystal structure (PDB code 1JJU). Finally, according to the progress of the processive action of QhpD (as described below), the structure model of the QhpC core peptide with or without the thioether bonds (three Cys-to-Asp/Glu cross-links and one in the CTQ cofactor) was also built by Coot (26), using the coordinates of the ␥ subunit in the QHNDH crystal structure as an initial structure.

RESULTS
Expression and Purification of the QhpC⅐QhpD Complex-To achieve overproduction of recombinant QhpD from P. denitrificans Pd1222, we constructed various plasmids carrying the genes encoding QhpD and its substrate QhpC (Fig. 3). For QhpC, we also prepared a series of full-length and truncated forms with or without the N-terminal 28-residue leader peptide that is essential for the formation of thioether bonds in QhpC (7) but must be removed subsequently by a subtilisin-like protease QhpE (8). Initial attempts to express QhpD alone or by co-expression with a plasmid carrying E. coli iscRSUA-hscBAfdx-iscX genes (30) or sufABCDSE genes (31), which are expected to assist the assembly of iron-sulfur clusters in radical SAM proteins (21), were all unsuccessful for efficient production of QhpD in the soluble fraction of the E. coli C41 (DE3) cell lysate. However, when QhpD was co-expressed with substrate QhpC, its expression was dramatically improved. Analysis of the cell-free extracts using a His-trap spin column revealed that expressed QhpD was tightly associated with QhpC containing the leader peptide plus the mature protein region with a sufficient length (at least residues involved in the first thioether bond between Cys-7 and Glu-16) (Fig. 4A). Co-expression of shorter QhpC with or without the leader peptide (QhpC (Ϫ28)Ϫ7 , QhpC 1-23 ) was essentially ineffective for QhpD expression, suggesting that overproduction of soluble QhpD was a result of the formation of a stable complex with QhpC containing both the leader peptide and the mature protein region of more than 23 residues. Densitometric comparison of the stained protein bands of the purified QhpC⅐QhpD complex (Fig. 4B) strongly suggested that the protein composition was a 1:1 complex of QhpC and QhpD. It is also noteworthy that co-expression of the full-length QhpC without the leader peptide (QhpC 1-82 ) did not enhance the QhpD production. Thus, the leader peptide of QhpC is essential for the interaction with QhpD, leading to the efficient expression of the QhpC⅐QhpD complex.
For purification of the QhpC⅐QhpD complex, we routinely employed co-expression of QhpD and sQhpC, carrying an N-terminal H 6 tag and a C-terminal St 2 tag, respectively, encoded in separate plasmids. Transformed E. coli cells were grown in the medium supplemented with excess ferric ions for loading iron into QhpD. By using the first His trap and the following Strep trap affinity columns in an anaerobic chamber, both proteins were efficiently co-purified from the cell-free extracts to near homogeneity (purity, Ͼ90% on SDS-PAGE) as the sQhpC⅐QhpD complex (Fig. 4B). The as-purified complex was rather unstable, forming protein precipitates even in an overnight storage under anaerobic conditions. Therefore, it was reconstituted immediately after the purification, unless otherwise stated. The reconstituted complex thus obtained was stable for a few days when stored at 4°C in a buffer containing 10% (w/v) glycerol under rigorously anaerobic conditions; addi- tion of glycerol was essential to prevent protein precipitation. All the subsequent experiments were done within 1-3 days after the reconstitution.
Spectral Properties of the Purified QhpC⅐QhpD Complex-After purification, the sQhpC⅐QhpD complex was colored dark brown with absorption bands at ϳ320 and 410 nm, thereby supporting the presence of [4Fe-4S] clusters (32,33) in QhpD (Fig. 5). Chemical reconstitution resulted in a further increase of the absorption bands not only at 320 and 410 nm but also at 280 nm. Using a molar extinction coefficient at 410 nm reported for the [4Fe-4S] cluster (15,000 M Ϫ1 cm Ϫ1 ) (34,35) and protein concentrations determined by the Bradford method and corrected with the Bradford correction factor (21), the contents of the [4Fe-4S] cluster were calculated to be about 1.6 and 2.7 mol/mol of as-purified and reconstituted sQhpC⅐QhpD complexes, respectively. Colorimetric iron/sulfur analysis also indicated the presence of multiple (likely three) iron-sulfur clusters (Fig. 5). These results are consistent with the presence of three iron-sulfur cluster-binding motifs conserved in the QhpD homologs (Fig. 2). The absorption band at ϳ410 nm disappeared after incubation with DT (Fig. 5), suggesting reduction of the [4Fe-4S] 2ϩ cluster.
The reconstituted sQhpC⅐QhpD complex was EPR-silent ( Fig. 6), indicating that the [4Fe-4S] clusters in QhpD are in the diamagnetic oxidized form ([4Fe-4S] 2ϩ ) (36). The EPR spectrum characteristic for the reduced form of the cluster ([4Fe-4S] ϩ ) (37) was observed at 15 K after the addition of excess DT (Fig. 6), but not at 77 K (data not shown). The signals derived from the reduced QhpD were almost identical in the complex with either sQhpC (wild type) or sQhpC C7S (a mutant that does not undergo ether bond formation by QhpD, as described later), except for slight broadening of the g ʈ signal at g ϭ 2.041 (Fig. 6).
SAM Cleavage Activity-The common activity for most, if not all, radical SAM enzymes reported is the reductive homolytic cleavage of the S-5ЈC bond of SAM, producing methionine, and the 5Ј-deoxyadenosyl radical (5Ј-dA ⅐ ) that abstracts a hydrogen atom from the substrate in their respective coupled reactions, or from protein or solvent in the uncoupled reaction, to form the final stable product 5Ј-deoxyadenosine (5Ј-dA) (10 -12). The reductive SAM cleavage activity of the sQhpC⅐ QhpD complex was determined by measuring both methionine and 5Ј-dA by liquid chromatography of the reaction mixture. Control experiments indicated that the reaction occurred only when the DT-reduced QhpC⅐QhpD complex was anaerobically incubated with SAM, yielding a product identifiable as 5Ј-dA (calculated [M ϩ H] ϩ ϭ 252.24) by a TOF-MS analysis of the HPLC eluate (Figs. 7 and 8A). The reaction by the sQhpC C7S ⅐QhpD complex continued linearly over 60 min for production of both methionine and 5Ј-dA with a multiple-turn-over rate constant (k cat ) of about 0.193 min Ϫ1 at 1 mM SAM, when measured immediately after reconstitution (Fig. 8B). This rate is within the range of the reported values for SAM cleavage activities of other radical SAM enzymes, including the SPASM subgroup, such as PqqE (0.011 min Ϫ1 ) (38), anSME from Clostridium perfringens (termed anSMEcpe) (1.09 nmol min Ϫ1 mg Ϫ1 (ϭ 0.047 min Ϫ1 )) (15), AtsB (0.32 min Ϫ1 ) (16), and lipoate synthase (0.175 min Ϫ1 ) (39). By measuring the initial rates of 5Ј-dA production at different SAM concentrations, the apparent K m value for SAM was determined to be 45.3 Ϯ 5.4 M. The observation that the rate of 5Ј-dA production with the wild-type sQhpC⅐QhpD complex was slightly higher than that with sQhpC C7S ⅐QhpD (Fig. 8B) suggests that the SAM cleavage reaction is further stimulated, albeit only by about 30%, when coupled with the following reaction (thioether bond formation). Collectively, these results show that QhpD catalyzes the multiple-turnover reaction of reductive SAM cleavage irrespective of the subsequent coupled reaction, yielding 5Ј-dA and methionine considerably exceeding the amount of QhpC⅐ QhpD complex used.
Thioether Bond Formation in the sQhpC⅐QhpD Complex-We expected that the formation of thioether bonds within the bound QhpC in the QhpC⅐QhpD complex would occur in vitro, if the 5Ј-dA ⅐ produced by QhpD in the reductive cleavage of SAM is used efficiently for the subsequent coupled reaction. Because the mass difference before and after the thioether bond formation is only 2 mass units and protein chemical determination of intra-peptidyl thioether bonds is not straightforward (see also Fig. 9A) (40), the reaction was assayed by monitoring the loss of a free sulfhydryl group from a Cys residue by chemical modification with IAA or NAM (Fig. 10, A and D). A similar assay using IAA has been employed previously for the AlbAcatalyzed formation of sulfur-to-␣-carbon thioether bonds in subtilosin A (18). Thus, modification of a Cys residue with IAA and NAM gives an overall difference of ϳ59 and 276 mass units, respectively, for the molecular ion peaks observed before and after the thioether bond formation (Fig. 10, B and E). Disappearance of the fluorescent band on SDS-PAGE is also a convenient way to assay semi-quantitatively the reaction using NAM that forms a fluorescent conjugate upon reaction with a free Cys residue (Fig. 10D). However, the assay methods using IAA and NAM are both inapplicable to the sQhpC C7S mutant without a free SH group. Nevertheless, exact matching of the mass peaks before and after the reaction (Fig. 10, C and F) and protein chemical analysis (Fig. 9B) revealed that the sQhpC C7S mutant does not undergo ether bond formation between Ser-7 and Glu-16 by QhpD.
The thioether bond formation was first examined with the purified and reconstituted sQhpC⅐QhpD complex, where only a single Cys residue (Cys-7) forming a thioether bond is contained in sQhpC. As shown in Fig. 11, A and B, the thioether bond formation (i.e. disappearance of a free SH group) occurred only in the presence of both DT and SAM. The reaction followed pseudo-first order kinetics and proceeded with the disappearance of Ͼ80% modifiable Cys residue within 60 min (Fig.  11C), indicating that the thioether bond formation is a singleturnover reaction occurring in the sQhpC⅐QhpD complex. The pseudo-first order rate constant for the initial 20-min reaction FIGURE 6. EPR spectra of wild-type and mutant QhpD complexes with sQhpC or sQhpC C7S . The X-band EPR spectra of the reconstituted sQhpC⅐ QhpD, sQhpC C7S ⅐QhpD, sQhpC⅐QhpD C116S , sQhpC⅐QhpD C353S , sQhpC⅐ QhpD G161A , and sQhpC⅐QhpD R373A complexes (80 M) are shown with important g values including those for g ʈ and g Ќ signals. X-band (9.23 GHz) microwave frequency spectra before (gray) and after (black) the addition of 2 mM DT were recorded at 15 K with modulation frequency, 100 kHz; modulation amplitude, 1 millitesla (mT); and microwave power, 10 milliwatts. was estimated to be ϳ0.05 min Ϫ1 (3 h Ϫ1 ) (Fig. 11D), which corresponds to about 26% of the SAM cleavage activity. Although this activity was considerably lower than those of the multiple-turnover reactions catalyzed by other radical SAM enzymes that range from 0.175 min Ϫ1 (lipoate synthase) (39) to ϳ27 s Ϫ1 (lysine 2,3-aminomutase) (41), the low single-turnover activity of QhpD is probably sufficient for the post-translational modification (maturation) of QhpC (␥ subunit of QHNDH) coordinated with bacterial cell growth.
The thioether bond formation in the sQhpC⅐QhpD complex appeared to occur also in the recombinant E. coli cells, although slightly, as judged from the presence of a small amount of sQhpC that could not be modified with IAA even before the reaction (with a relative peak height corresponding to about 12% of the main peak) (Figs. 11A and 12A). It is assumed that intracellular SAM and a physiological electron donor serving for reduction of the SAM-binding [4Fe-4S] cluster promoted the formation of the thioether bond by QhpD in the recombinant E. coli cells. Supporting this assumption, co-expression of SAM synthetase (MetK) resulted in a significant increase in the amount of sQhpC already containing the thioether bond in the purified QhpD⅐sQhpC complex (with a relative peak height corresponding to about 26% of the main peak) (Fig. 12B).
Thioether Bond Formation in the fQhpC⅐QhpD Complex-We next studied thioether bond formation in fQhpC, which contains three Cys residues (Cys-7, Cys-27, and Cys-41) that are involved in thioether cross-links to the methylene carbon atom of Glu or Asp and one Cys residue (Cys-37), forming the CTQ cofactor (Fig. 1B). When fQhpC was co-expressed with QhpD, the purified fQhpC in the fQhpC⅐QhpD complex was found to be a mixture of species that contains 2-4 Cys residues reactive with IAA (Fig. 13A, green trace), indicating again that thioether bond formation in fQhpC partially occurs in E. coli cells. The fQhpC without thioether bonds, which contains the full four Cys residues reactive with IAA (Fig. 13A, blue trace), could only be obtained by expression as a single polypeptide or by co-expression with an inactive QhpD mutant (QhpD C116S , as described later) as the QhpC⅐QhpD complex followed by removal of heat-denatured QhpD C116S (fQhpC is heat-stable). The mixed fQhpC containing 2-4 Cys residues could finally be converted to the form that contains only a single free Cys (Cys-37, which is not involved in the sulfur-to-methylene carbon FIGURE 7. Analysis of 5-dA produced in the SAM cleavage reaction by QhpD. After TCA quenching and removal of the precipitated proteins, the reaction mixtures were analyzed by reversed-phase column chromatography using Cosmosil 5C 18 -PAQ column (Nacalai Tesque) equipped on an HPLC system: A, DT; B, sQhpC⅐QhpD; C, SAM; D, sQhpC⅐QhpD ϩ DT ϩ SAM; E, 5Ј-dA standard. The elution was monitored at absorbance at 260 nm. F, P2 peak of sQhpC⅐QhpD ϩ DT ϩ SAM (D) was collected and analyzed by TOF MS. The peaks with masses larger than that of 5Ј-dA are assumed to be derived from impurities contaminated during the procedure. thioether bonds) by anaerobic incubation with DT and SAM for 2 h (Fig. 13A, red trace). The time course of the reaction measured by MS (Fig. 13B) shows that the reaction in fQhpC practically completes within 1 h. Analysis by SDS-PAGE also indicates the time-dependent formation of fQhpC containing only a single Cys (ϭ3 thioether bonds formed), which is well separated from the initial unmodified polypeptide (Fig. 13B, inset). These results unequivocally show that formation of the sulfurto-methylene carbon thioether bond is catalyzed by QhpD in the complex with its substrate fQhpC.
Furthermore, to study whether the thioether bond formation occurs from the N-to C-terminal direction (i.e. in the order of Cys-7-Glu-16, Cys-27-Asp-33 and Cys-41-Asp-49) or initiates at any position in the fQhpC sequence, we prepared mutants of substrate fQhpC and reacted these mutants with QhpD in the presence of DT and SAM. The Glu-to-Gln or Asp-to-Asn mutants of fQhpC are expected not to serve as a sulfur acceptor at the methylene carbon atoms, presumably due to the absence of the side-chain carboxyl group that is assumed to interact with the conserved Arg-373 in QhpD, as discussed later. The reaction products after 20-min of incubation with DT and SAM were treated with IAA and analyzed by TOF-MS (Fig.  13C). To confirm the products after a longer reaction time (1-4 h), the products were also analyzed by SDS-PAGE as the position of the stained bands differs depending on the number of thioether bonds present; although the bands without and with one thioether bond cannot be separated (Fig. 13C). The wildtype fQhpC was converted to the major product containing three full thioether bonds (i.e. containing only 1 IAA-reactive SH group), whereas the single fQhpC E16Q and double fQhpC D33N/D49N mutants were unmodified (i.e. containing four IAA-reactive SH groups) or were converted to the product containing only a single thioether bond (i.e. containing three IAA-reactive SH groups) after reaction with QhpD. In contrast, the single fQhpC D49N mutant was converted to the product that contained two thioether bonds (i.e. containing two IAAreactive SH groups) (Fig. 13C). These results clearly show that thioether bond formation proceeds sequentially in an N-to C-terminal direction in the bound fQhpC but not randomly at any position, as shown schematically in Fig. 13C.
Dissociation of the QhpC⅐QhpD Complex-An important issue to be addressed is whether the substrate fQhpC dissociates from QhpD or stays bound in the complex during or after formation of the three thioether bonds. This was studied by using a His trap affinity column (Fig. 14A), and most of fQhpC was found to bind in the complex with QhpD even after the three thioether bonds were formed; only a small amount of fQhpC with three thioether bonds was found in the flowthrough and the first washing fractions. This result indicates that formation of the three thioether bonds in fQhpC is catalyzed progressively by a single QhpD molecule bound in the same complex (i.e. a single-turnover reaction in terms of fQhpC) and thus suggests the sliding of the fQhpC polypeptide chain along the active site of QhpD during the sequential formation of the three thioether bonds, as discussed below. Quantification of the stained bands of fQhpC (Fig. 14B) also supports that the fully cross-linked fQhpC is only slightly more dissociable from QhpD than the unmodified one. Thus, it is concluded that the substrate fQhpC most probably stays bound in the complex during and after formation of the three thioether bonds, which is certified by the tight binding of the leader peptide of fQhpC to QhpD.
Homology-based Structure Modeling of QhpD-Presumably due to the instability after purification and/or heterogeneity of the bound QhpC, we have not yet succeeded in crystallization of the QhpC⅐QhpD complex for x-ray diffraction analysis.  ) are also shown near each peak. In the case of the wild-type sQhpC shown in A, the peak height of fragment iii was significantly decreased after the cross-linking reaction, probably due to the inability of Glu-C to cleave the peptide bond next to the cross-linked Glu-16. In contrast in B, such effect was not observed after the reaction, indicating that the C7S mutant does not undergo the ether bond formation between Ser-7 and Glu-16 by QhpD.
However, a highly plausible structure model of QhpD could be obtained by homology alignment-based structure modeling (Swiss-Model) (25). Although the crystal structure of anSMEcpe (PDB code 4K36) (17) was auto-selected as the template based on sequence homology, the loop region from Gln-345 to Leu-360, adjacent to one of the predicted auxiliary clusters, had to be manually modified to create sufficient space to accommodate the substrate QhpC peptide segment in the active site. Consequently, Cys-353 of QhpD that corresponded to Cys-255 of anSMEcpe in the initial model was re-aligned to correspond to Cys-261 of anSMEcpe ( Fig. 2A) to build the final model, containing residues Ile-99 -Asn-463 (Fig. 15); the N-terminal 98 residues that are absent in the anSMEcpe sequence were not included in the model. This model is consistent with the prediction that the majority of radical SAM superfamily enzymes would have a common core fold comprising a partial (␣/␤) 6 triose-phosphate isomerase barrel (42,43). The lateral opening (an opening on the side) of the triose-phosphate isomerase barrel in the QhpD model is as large as that of anSMEcpe and hence appears to be able to accommodate proteinaceous substrates. Moreover, in addition to the consensus CX 3 CX 2 C motif (Cys-112, Cys-116, and Cys-119 in QhpD) that ligates the rad- ical SAM [4Fe-4S] cluster (designated RS cluster) (10 -12), seven Cys residues conserved in the C-terminal SPASM domain (Fig. 2B) are ideally positioned to bind two auxiliary clusters (Aux I and Aux II ) at sites very similar to those in the anSMEcpe structure. The Aux I cluster in QhpD, however, is ligated by three Cys residues (Cys-353, Cys-371, and Cys-422) unlike in the anSMEcpe structure, in which the auxiliary cluster I is ligated by four Cys residues (17), whereas the Aux II cluster is fully ligated by four Cys residues (Cys-409, Cys-412, Cys-418, and Cys-441) (Fig. 15). Thus, the noncoordinated iron site in the Aux I cluster may bind the substrate as proposed for the formylglycine-generating enzyme AtsB (16), the subtilosin A-synthesizing enzyme AlbA (18), and the molybdenum cofactor biosynthetic enzyme MoaA (44,45). The RS cluster to Aux I cluster distance is estimated to be 16.3 Å, which is comparable with that (16.9 Å) in the anSMEcpe structure (17).
Site-specific Mutagenesis of Conserved Residues-To study the roles of residues strongly conserved in the SPASM subgroup ( Fig. 2A), a number of site-specific mutants were prepared for QhpD. The first mutant prepared involved mutating the middle Cys residue (Cys-116) in the N-terminal signature motif (CX 3 CX 2 C) into Ser (QhpD C116S ). The second mutant prepared replaced the Gly residue (Gly-161) in the GGE motif, predicted to be involved in binding of SAM, with Ala (QhpD G161A ). Two Cys residues (Cys-353 and Cys-412) in the 7-Cys motif of the SPASM domain (15) were replaced by Ser (QhpD C353S and QhpD C412S ). Finally, Arg-373 that is invariant in QhpD homologs (Fig. 2B) was replaced by Ala (QhpD R373A ). We previously reported that the n-butylamine-dependent growth of the ⌬qhpD mutant strain of P. denitrificans was not rescued by complementation with a plasmid encoding QhpD C116S , QhpD G161A , or QhpD C412S mutants (7). Using the same method, QhpD C353S and QhpD R373A mutants were also found to be unable to recover the growth of the ⌬qhpD strain and to produce active QHNDH (Fig. 16), demonstrating the importance of these conserved residues for the in vivo activity of QhpD.
Interaction of these QhpD mutants with the co-expressed sQhpC was examined using a His-trap spin column (Fig. 17). Three mutants (QhpD C116S , QhpD G161A , and QhpD R373A ) were produced and bound to sQhpC as efficiently as the wildtype QhpD. However, expressions of the QhpD C353S and QhpD C412S mutants were significantly lower and negligible, respectively, suggesting that mutation of Cys-353 and Cys-412 resulted in a significant structural change of QhpD that impaired the interaction with QhpC. Because of negligible protein expression, the QhpD C412S mutant was not analyzed further in subsequent studies.
Absorption spectra of the QhpD C116S , QhpD G161A , QhpD C353S , and QhpD R373A mutants co-purified with sQhpC are shown in Fig. 5, together with their iron/sulfur contents. Increases in visible absorption bands as well as iron/sulfur contents observed after reconstitution of the purified complex were similar among the wild-type QhpD and QhpD G161A and QhpD R373A mutants but were considerably greater in the QhpD C116S and QhpD C353S mutants over the wild-type enzyme. These results suggest that in these Cys mutants, additional iron and sulfur atoms are adventitiously bound to surface residues on the protein (21) and/or to the remaining Cys residues in the canonical and auxiliary [4Fe-4S] cluster-binding sites possibly in different chemical structures (12).
The reduced QhpD C116S mutant in the complex with sQhpC exhibited an EPR spectrum considerably different from that of the wild-type QhpD complex, particularly in the g Ќ region showing a broader trough of the derivative type signal with an increased rhombic character (Fig. 6). The shoulder signal at g ϭ 2.075 is still observed in the QhpD C116S spectrum and therefore appears to be derived from Aux I and Aux II . The spectra of the reduced QhpD G161A and QhpD R373A mutants in the complex with sQhpC were very similar to the spectrum of the wild-type QhpD complex, except that the signal at g ϭ 2.075 was almost absent in the QhpD G161A mutant, and the trough at g ϭ 1.900 was considerably shifted to a higher magnetic field, suggesting that the mutation of the SAM-and/or substrate-binding sites  B, the as-purified sQhpC-H 6 ⅐QhpD (blue) and that co-expressed with MetK (red) were used without subjecting to the in vitro reaction. In other panels, the reconstituted complexes before (blue) and after (red) the reaction with 1 mM DT and 1 mM SAM for 1 h were used for analysis. Note that in the sQhpC-St 2 ⅐QhpD G161A complex (D), the thioether bond formation in E. coli cells (before the in vitro reaction) was almost unobserved as compared with the wild-type sQhpC-St 2 ⅐QhpD in A. The method of sample preparation was the same as described in the legend to Fig. 10 also affected the structures of the [4Fe-4S] cluster-binding regions. From the spectrum of the reduced QhpD C353S mutant in the complex, it is likely that the signal around g ϭ 1.92 is derived from Aux II and/or RS clusters that are left in this mutant.
Regarding the SAM cleavage activity, the QhpD G161A mutant had significantly reduced activity (about 20% of the wild-type enzyme), whereas the QhpD C116S mutant showed no activity (Fig. 18), as expected because these residues are involved in binding SAM. It is also interesting to note that even the QhpD C353S and QhpD R373A mutants, both of which are assumed not to directly participate in SAM binding, exhibited very low SAM cleavage activities. Presumably, these residues play allosteric roles in stabilizing the protein architecture necessary for SAM binding and/or cleavage. Alternatively, there is a possibility that electrons are supplied from Aux I and/or Aux II to the RS cluster during the reductive SAM cleavage reaction, as proposed for the role of auxiliary clusters in anSMEcpe (15,17).
The in vitro thioether bond formation was also examined with various QhpD mutants in the complex with sQhpC (Fig.  12, C-F). Unexpectedly, the QhpD G161A mutant exhibited significant activity that was comparable with the wild-type QhpD at 1 mM SAM (Fig. 12D). This observation is apparently inconsistent with the inability of the QhpD G161A mutant to rescue the growth of the ⌬qhpD strain of P. denitrificans (7). It is likely that the intracellular concentration of SAM in P. denitrificans is lower than 1 mM and is therefore insufficient for the QhpD G161A mutant to function in vivo; the mutant is probably unable to form thioether bonds in QhpC during the QHNDH maturation in the ⌬qhpD strain. Conversely, the QhpD C116S , QhpD C353S , and QhpD R373A mutants had no thioether bond forming activity (Fig. 12, C, E, and F), as predicted from their no or negligible SAM cleavage activities (Fig. 18) that should provide 5Ј-dA ⅐ needed for the subsequent thioether bond formation. It is noteworthy that the QhpD R373A mutant was unable to form thioether bonds not only in sQhpC but also in fQhpC (Fig.  12G).

DISCUSSION
In this study, we have demonstrated that the radical SAM enzyme QhpD catalyzes the sequential formation of sulfur-tomethylene carbon thioether bonds of its protein substrate QhpC in the presence of an artificial reducing agent (DT) and SAM. The intra-protein thioether bond formation is a chemically challenging reaction, accompanying the removal of a hydrogen atom from an inert methylene carbon atom of Asp and Glu residues with high bond dissociation energies for homolytic cleavage (394.8 and 418.6 kJ/mol, respectively) (46). These values are even higher than those of ␣-hydrogen atoms of a polypeptide chain, which are abstracted in the sulfur-to-␣carbon thioether bond formation by analogous radical SAM enzymes AlbA and SkfB (18,19), requiring 345.1, 348.9, and 338.9 kJ/mol for ␣-H abstraction from Thr, Phe, and Met, respectively (46).
Based on the spectral properties of the purified and reconstituted QhpC⅐QhpD complex and the results of site-specific mutagenesis of conserved Cys residues, the presence of multiple (most likely three) [4Fe-4S] clusters (RS, Aux I , and Aux II ) in QhpD is strongly suggested, as also predicted from homology modeling of the QhpD structure (Fig. 15). However, except for the canonical role of the RS cluster that binds SAM and reductively cleaves it into methionine and 5Ј-dA ⅐ , the roles of the Aux I and Aux II clusters remain inconclusive. Decreased binding of the QhpD C353S mutant with substrate QhpC (Fig. 17) and loss of thioether bond forming activity (Fig. 12E) both support the role of the Aux I cluster in binding of protein substrate. In addition, Aux I may also act as an electron acceptor from substrate and transfer the accepted electron to the RS cluster during the thioether bond formation, as proposed below. In the QhpD model (Fig. 15), the Aux II cluster is located rather close to the molecular surface and behind the predicted binding region for the leader peptide of QhpC (see below); mutation of one of the Aux II cluster-ligating Cys residues (Cys-412) completely abolishes the interaction with QhpC (Fig. 17), suggesting a structural role to support the stable QhpC⅐QhpD complex. Arg-373 that is completely conserved in the QhpD homologs (Fig. 2B) is located in the vicinity of the Aux I cluster and assumed to be involved in electrostatic binding of acidic residues (Asp or Glu) of substrate QhpC; the anchoring of the FIGURE 14. Analysis of the interaction between QhpD and fQhpC before and after the cross-linking reaction. A, reconstituted fQhpC⅐QhpD (60 M) (before reaction) was reacted with 1 mM DT and 1 mM SAM for 8 h (after reaction). The reaction mixture (2 ml) was applied to a His trap column (2 ml) (QhpD contained N-terminal H 6 -tag), and the flow-through (FT, 2 ml), first and second wash (W1 and W2, each 2 ml), and eluted (E, 2 ml) fractions were collected and analyzed by SDS-PAGE (10 l/lane). The protein bands were stained with colloidal Coomassie G-250 (54). B, enhanced image of the boxed area in A. Band intensities were quantified using the ImageJ program (National Institutes of Health) and are shown below each lane as a percentage relative to the band intensity of uncross-linked fQhpC before the reaction. carboxyl group of Asp/Glu is likely to be essential for thioether bond formation by QhpD (Figs. 12, F and G, and 13C). We also noted the presence of a large groove consisting of highly conserved and positively charged residues (including above mentioned Arg-373) in the QhpD model, which has sufficient space to accommodate the core QhpC polypeptide containing several negatively charged residues, as depicted in Fig. 19, A and B (see also supplemental Movie). The RS and Aux I clusters are harbored in the bottom of this groove, thus constituting the active site of QhpD.
Precursors of most ribosomally translated natural products such as antibiotics contain an N-terminal leader extension in addition to the C-terminal core peptide that is processed to the final mature compound by post-translational tailoring reactions (47). Likewise, QhpC contains the 28-residue leader peptide that is essential for stable binding with QhpD (Fig. 4A). The leader peptides in general show high sequence conservation as compared with the core peptides with greater sequence variability. However, the leader peptide of QhpC is less conserved than the mature protein region (Fig. 20A) that undergoes the multiple thioether bond formation by QhpD, indicating that QhpD has rigorous sequence specificity for protein substrate but is rather tolerant for the leader peptide sequence. Although the details as to how leader peptides of natural products are recognized by the processing enzymes are largely unknown, it has been noted that many leader peptides have a propensity to form amphipathic ␣-helices, either in solution or when bound to the enzymes (47). A major part of the 28-residue leader peptide of QhpC is indeed assumed to form ␣-helices by secondary structure prediction (Fig. 20B). Thus, it appears very likely that formation of the stable complex between QhpC and QhpD is facilitated through binding of the putative ␣-helix of the QhpC leader peptide to a small groove located above the Aux II cluster and connected to the large groove with some specified ionic interactions between the conserved residues of QhpC and QhpD (Fig. 19, A-C). There is an additional possibility that about 100 N-terminal residues of QhpD that are not included in the model may also interact with the leader peptide, particularly at the conserved N-terminal portion including the invariant Lys(Ϫ19) (Fig. 20A).
There are important structural features common to the three Cys-to-Asp/Glu thioether bonds formed in QhpC (Fig. 1C), which should be taken into consideration in modeling the structure of the QhpC⅐QhpD complex and also in deciphering the reaction mechanism of QhpD as follows. (a) The sulfur-donating Cys residue always precedes the partner Asp/Glu residue in the sequence of QhpC, i.e. Cys-7-Glu-16, Cys-27-Asp-33, and Cys-41-Asp-49, suggesting that the segments of QhpC to be cross-linked bind to the active site of QhpD always in the same orientation. This could be achieved by binding of QhpC at two distinct sites, one with the Cys sulfur at the noncoordinated iron site in the Aux I cluster of QhpD and another with the carboxyl group at the conserved Arg-373. (b) The length of the FIGURE 19. Structure modeling of the QhpC⅐QhpD complex. A, sequence conservation is mapped onto the molecular surfaces of the models of QhpD and QhpC, colored in gradation from white to green, corresponding to the score of 0.5 (nonconserved) and 1.0 (fully conserved), respectively. The conservation score was calculated using Scorecons (55) with all QhpD and QhpC homologs (ϳ270 sequences) identified by a BLAST search. B, electrostatic surface potential is mapped onto the modeled QhpD surface, colored in gradation from red (Ϫ5 kT) to blue (ϩ5 kT), where k is Boltzmann's constant and T is the absolute temperature, based on the calculation by PyMOL (Schrödinger, LLC) and APBS (56). C, predicted interactions between the leader peptide of QhpC and surface residues of QhpD. Residue numbers of QhpC and QhpD are indicated in red and black, respectively. Sequence conservation is also shown as in A. D, modeled structures of the fQhpC⅐QhpD complexes during the processive reaction along the fQhpC polypeptide are shown in ribbon representation with molecular surfaces of QhpD (green ribbon). The leader and the core peptides of QhpC are colored orange and magenta, respectively. SAM, iron-sulfur clusters, thioether bonds, and their precursor residues are shown by stick models. Modeled structures of the cross-linked segments, Cys-7-Glu-16 in step II, Cys-7-Glu-16 and Cys-27-Asp-33 in step III, Cys-7-Asp-33 and Cys-41-Asp-49 in step IV, and the fully cross-linked core peptide in step V of fQhpC were adopted from the coordinate of ␥ subunit (maturated QhpC) in the QHNDH crystal structure with other regions built and refined using the regularize zone module in Coot (26) to improve model stereochemistry. All molecular drawings were generated using PyMOL.
polypeptide chain between the Cys and Asp/Glu is not too short and not too long, i.e. 5-8 amino acids, containing one Pro residue (Pro-13, Pro-29, and Pro-44) that is often observed in a loop region of polypeptide chains. The limited chain length between the cross-linked partner residues excludes creation of the ladder-like topology found in sactibiotics (18). Thus, upon coordination of the substrate Cys residue to the Aux I cluster, the polypeptide segment should assume a loop structure with an appropriate chain length so that the site of hydrogen abstraction (the methylene carbon atom of downstream Asp/Glu) is positioned close to the Cys residue. (c) The carbon atoms involved in the thioether bonds are all in the S-configuration (5,6). This consistent stereochemistry suggests that thioether bond formation proceeds through stereospecific abstraction of either one of the two prochiral hydrogen atoms, in contrast to the reaction by AlbA that proceeds nonstereospecifically with net retention of the configuration at Phe-22 and inversion at Thr-28 and Phe-31 in subtilosin A (18,48).
Altogether, based on the above considerations and referring to the reaction mechanism suggested for AlbA (18), we propose a plausible radical mechanism of QhpD for formation of the thioether bond in its protein substrate QhpC, as shown in Fig.  21. The QhpC core polypeptide following the N-terminal leader peptide bound to the molecular surface of QhpD would be arrested within the large groove consisting of positively charged residues of QhpD ( Fig. 19B; see also supplemental Movie). The starting model for the reaction includes coordination of the QhpC-Cys-7 sulfur atom to the Aux I cluster of QhpD and anchoring of QhpC-Glu-16 through a salt bridge of the carboxyl group to the QhpD-Arg-373 guanidino group, together with SAM binding to the unique iron of the reduced RS cluster (Fig. 21, step i). The orientation of anchored Glu-16 directs the C␥ position toward the 5Ј-C of SAM bound to the RS cluster. We postulate that the initial reductive cleavage of SAM and concomitant oxidation of the RS cluster will proceed in the same way as other radical SAM enzymes (10 -12). The 5Ј-dA ⅐ generated by the reductive cleavage of SAM then abstracts the hydrogen atom of the Glu-16 C␥ methylene carbon (step ii). Subsequently, the Cys-7-Glu-16 loop would change the conformation to position the Cys-7 sulfur and Glu-16 methylene carbon atoms in a juxtaposition (step iii), followed by formation of the thioether bond between the carbon-centered radical of Glu-16 and the sulfur atom of Cys-7 ligated to the Aux I cluster, accompanied by simultaneous electron transfer to the Aux I cluster (step iv). Finally, due to the restricted conformation following the formation of the thioether bond (step v), the crosslinked Cys-7-Glu-16 loop is released from the active site to evacuate for the second cross-linked segment (Cys-27-Asp-33) (step vi). The reduced Aux I cluster could then transfer intramolecularly the electron to the RS cluster in the [4Fe-4S] 2ϩ form FIGURE 21. Possible reaction mechanism for the formation of thioether bond catalyzed by QhpD. The active site of QhpD is depicted by a stick model extracted from the modeled QhpD structure (Fig. 15) with the ribbon model of QhpC built by Coot (26) and shown in purple, except for the cross-linked loop structure containing Cys-7-Glu-16 in step vi, which was adopted from the coordinates of the ␥ subunit (maturated QhpC) in the QHNDH crystal structure. Distances between the two atoms connected by cyan dotted lines are indicated in angstroms (Å). Red and purple curves with an arrowhead show presumed routes of electron transfer. All molecular drawings were generated using PyMOL.