<i>Xanthomonas campestris</i> PqqD in the pyrroloquinoline quinone biosynthesis operon adopts a novel saddle‐like fold that possibly serves as a PQQ carrier

Abstract

Pyrroloquinoline quinone (4,5-dihydro-4,5-dioxo-1H-pyrrolo-[2,3-f]quinoline-2,7,9-tricarboxylic acid, PQQ) is a novel heat-stable, water-soluble redox cofactor present in a variety of bacterial alcohol and aldose dehydrogenases.1-5 It belongs to the third family of ortho-quinone cofactor following the well-known pyridine nucleotide- (such as NAD) and flavin-dependent (such as FAD) cofactors.6, 7 However, PQQ is unique in that it can act either as an antioxidant or as a pro-oxidant, depending on the conditions of different biological systems.8 Discovery of PQQ has also elicited considerable interests due to its role as a growth-promoting factor both in animals9 and plants.10 Although the structures and functions of many enzymes using PQQ as a cofactor has been elucidated,2-5, 11 relatively little is known about the pathway leading to the biosynthesis of this intriguing cofactor. However, some progresses have been made to date. For example, six genes belonging to the pqqABCDEF operon were found to be essential for the biosynthesis of PQQ in Klebsiella pneumoniae.12, 13 Of the six gene products, the first one (PqqA) is a 23-residue peptide with conserved glutamate and tyrosine residues that likely serves as the precursor for PQQ.14-16 The second one (PqqB) is not directly involved in the PQQ biosynthesis12 but may serve as a carrier for PQQ when PQQ is released from PqqC. This is because PQQ can react with oxygen to generate deleterious free radicals when set free without protection.17 Recently, the crystal structure of PqqB was released (PDB: 1XTO), which reveals a half-moon shaped molecule comprising a metallo-hydrolase/oxidoreductase fold and a putative PQQ binding site near a zinc atom. However, a comprehensive functional assignment for PqqB has not been achieved. The third one (PqqC) is possibly the most studied protein in the pqq operon and has been demonstrated to catalyze the last step in the PQQ biosynthesis pathway.18 Its tertiary structure has also been determined recently by X-ray crystallography (1OTV and 1OTW) and was found to adopt a homodimer, with each monomer folding into a compact seven helix bundle that provides a scaffold for the active site cavity.18, 19 Upon substrate binding, PqqC was found to undergo large conformational changes. PqqC is also unique because it is an oxidase yet contains no redox active metal or other cofactor. The next component PqqE was found to catalyze a radical reaction linking the conserved glutamate and tyrosine moieties of PqqA at atoms C9 and C9a of PQQ. Its tertiary structure is not determined, but was modeled as a molybdenum cofactor biosynthesis protein based on its sequence homology with the radical SAM protein family.20 An iron-sulfur cluster was proposed in the SAM motif and believed to be the center to initiate the free radical reaction necessary to link the C9 and C9a bond of PQQ. The model also features a tunnel through which the PqqA β-strand can pass and get to the active iron-sulfur cluster site. Similarly, PqqF was modeled as a metalloendopeptidase using homologous E. coli pitrilysin as the template.20 It was believed to be involved in cleaving the four bonds connecting the glutamate and the tyrosine side chains to PqqA. The modeled PqqF structure also features a triangular prism shaped chambers in the close form to trap the PqqA peptide inside the cavity. It is still unclear, however, how PqqF can cleave the four bonds using the same active site. Finally, PqqD is the smallest protein with a molecular weight of ∼10 kDa in the pqq operon. Although sequence alignment of PqqD with a number of different organisms shows some strictly conserved residues, no sequence homology with proteins outside the PqqD family could be detected. Modeling of this protein could only be carried out20 de novo using the robetta server,21 but the genuine structure and function of PqqD remain to be determined. Since there are many intriguing biochemical questions remaining in fully understanding the PQQ biosynthesis pathway,20 accurate structural determinations of PqqE, PqqF, and PqqD are essential to answer these questions. We now report in this manuscript the tertiary structure of PqqD from a plant pathogen Xanthomonas campestris (Xc) determined to a resolution of 1.66 Å by using X-ray crystallography. The crystal structure of XcPqqD was found to adopt a dimer of novel saddle-like fold that revealed considerable differences to the structure predicted de novo and carried out in the absence of suitable template. In addition, a phosphate ligand was found to be situated very well in the highly positively-charged central channel. Docking study of PqqD with PQQ located PQQ at the similar phosphate binding site as well, indicating that PqqD can possibly bind PQQ and serve as a PQQ carrier. The XcPqqD gene was PCR amplified directly from the plant pathogen Xanthomonas campestris pv. campestris str. 17 (Xcc) using a forward primer 5′-TACTTCCAATCCAATGCTGTGAGCACCATCAGCCGCGACAGC and a backward primer 5′-TTATCCACTTCCAATGTCATAACCGCAGCAAGCGCTTCTGGTG to form the fragment of required length. The PCR fragment exhibited a correct size in agarose-gel experiment, with its sequence confirmed by DNA sequencing. A ligation-independent cloning approach22 was used to obtain the desired constructs. The final construct codes for an N-terminal His6-tag, a glutathione S-transferase (GST) protein, a 17 amino acid linker, and an XcPqqD target protein under the control of a T7 promoter. Over expression of the fused target protein was induced by the addition of 0.5 mM IPTG at 293 K for 20 hours. The expressed protein was purified by immobilized metal affinity chromatography on a nickel column (Sigma). The His6-tag, GST and linker were cleaved from XcPqqD by Tobacco Etch Virus protease at 291 K for 12 hours. For crystallization, XcPqqD protein was further purified by passing through an anion exchanger column (AKTA, Pharmacia Inc.) and a Superdex 75 column (AKTA, Pharmacia Inc.). The final target protein exhibited a purity of greater than 99% and contained only an extra tripeptide (SNA) at the N-terminal end. Se-Met labeled XcPqqD was prepared in a similar way, and was produced using an E. coli strain BL21 (DE3) as the host in the absence of methionine, but in the presence of large amounts of Se-Met (100 mg/L). The M9 minimum medium consists of 1 g of NH4Cl, 3 g of KH2PO4, and 6 g of Na2HPO4, 0.3% (w/v) of MgSO4, and 10 mg of FeSO4 in one liter of double-distilled water supplemented with 20% (w/v) of glucose. The induction was conducted at 293 K for 20 hours by the addition of 0.5 mM IPTG. As XcPqqD contains only one Met at the first position, and was found to give anomalous difference signal insufficient for correct protein phase determination, we have constructed a L80M mutant by the QuikChange Site-Directed Mutagenesis Kit from Stratagene using appropriate primers containing the required mutations to improve the signal.23 The mutant protein was expressed and purified using a similar procedure as the native protein. For crystallization, the native protein was concentrated to 65 mg/mL in 20 mM Tris-HCl and 80 mM NaCl by using an Amicon Ultra-15 (Millipore). Screening for crystallization condition was performed using the sitting-drop vapor diffusion method in 96-well plate (Hampton Research) at room temperature by mixing 0.5 μL protein solution with 0.5 μL reagent solution. Initial screens included Hampton Clear Strategy Screen 1, Structure Screens 1 and 2, systematic PEG-pH screen, and PEG/Ion screen, and were performed using a Gilson C240 crystallization workstation. Hexagonal-like crystals appeared in one week from a reservoir solution comprising 0.1 M Tris (pH 7.0) and 1.6 M ammonium sulfate. Crystals suitable for diffraction experiments were grown by mixing 1.5 μL protein solution with 1.5 μL reagent solution at room temperature and reached dimensions of 0.1 × 0.05 × 0.1 mm after one week. Se-Met labeled XcPqqD was crystallized in the same way. Crystal was flash-cooled at 100 K under a stream of cold nitrogen. X-ray diffraction data was collected using the National Synchrotron Radiation Research Center (NSRRC) beamline 13B1 in Taiwan. A two-wavelength multiple anomalous dispersion data set diffracted up to 2.0 Å resolution was obtained for the L80M mutant. The data were indexed and integrated using HKL2000 processing software,24 giving data set that were ∼98% complete with an overall Rmerge of 3.9% on intensities. The refinement of selenium atom position, phase calculation, and density modification were performed using the SOLVE/RESOLVE program.25 After phase determination, the structure of XcPqqD was refined against the native data set to a resolution of 1.66 Å. The model was manually adjusted by using the XtalView/Xfit package.26 CNS27 was then used for refinement to a final Rcryst of 19.6% and Rfree of 22.8%, respectively. XcPqqD crystals belong to the C2221 space group, with one XcPqqD protein in each asymmetric unit. The data and refinement statistics were summarized in Table I. The docking study of PQQ against XcPqqD was performed using a CHARMm-based MD simulation algorithm of CDOCKER from Accelrys Discovery Studio. In this procedure, the protein was kept rigid, whereas the PQQ ligand was treated as fully flexible. The potential binding cavities were first searched, resulting with two mostly likely sites, situated at each side of the central tunnel. Ligand conformations generated from a random high-temperature MD were then translated into this binding cavity, followed by random rigid-body rotations. Simulated annealing using a grid potential was applied, with a final minimization step further carried out to refine the docked poses. Ten best poses were selected based on the final energy. The one with the lowest energy was shown in [Fig. 2(b)].28 The coordinates and structural factors of the XcPqqD monomer have been deposited in the Protein Data Bank with an ID of 3G2B. XcPqqD is a small protein with 92 amino acid residues and a theoretical PI of 4.82. Sequence alignment of XcPqqD with PqqD proteins from several different organisms revealed many strictly conserved residues throughout the sequence as shown in Figure 1(a). The structure of XcPqqD starts with a long coil (Thr3-Val16), followed by a β1-β2 hairpin, interrupted by a bulge comprising a Pro32-Glu33 dipeptide, and again followed by a β3 strand. After these coils and strands, three helices ensue, forming a three-helix bundle [Fig. 1(b)]. The three-helix bundle is stabilized extensively by intra-strand hydrophobic interactions, through the highly conserved residues such as Val45, Ile58, and Val77 hydrophobic triad, as well as other residues on top and bottom of this hydrophobic cluster such as the Leu43, Ala46, Tyr49, Leu55, Leu62, Phe66, Ile73, Leu80, and Leu84 residues and so forth. These hydrophobic residues are either strictly or highly conserved and were marked by gray dots below the sequences listed in Figure 1(a). (a) The sequence alignment of proteins homologous to XcPqqD. Bold red and bold blue characters indicate strictly conserved and well conserved residues, respectively. Characters marked by orange dots indicate polar residues, whereas those marked by gray dots indicate hydrophobic residues involved in tertiary interactions. Secondary structural elements of coil, β-strands, α-helices, and loops were shown as gray arrows, blue arrows, green columns, and gray lines, respectively. XcPqqD homologues were identified using the CLUSTAL W multiple sequence alignment program.34 Xc, Xanthomonas campestris (accession number: AAM42211); Rm, Rhizobium meliloti (Q9EXU9); Go, Gluconobacter oxydans (Q9L3B1); Kp, Klebsiella pneumoniae (P27506); Ra, Rahnella aquatilis (O33505); Pa, Pseudomonas aeruginosa (Q9I2C1); Ac, Acinetobacter calcoaceticus (P07781). (b) The XcPqqD dimer depicted in ribbon representation with each monomer colored green and red, respectively. Secondary structural elements of C1, β1-β3, and α1-α3 were labeled. A C2-fold axis parallel to the plane was also shown. (c) Close up stereo view of intersubunit interactions between the N-terminal coil residues and the L1 loop, α3 helix, and L3 loop residues from another subunit. H-bonds were drawn in red dots and water oxygen atoms depicted as magneta circles. All interacting residues were drawn in stick, with carbon atoms from N-terminal C1 coil colored gray, and labeled in blue, carbon atoms from loop L1 colored yellow and labeled in yellow, carbons atom from helix α3 colored orange and labeled in red, whereas carbon atoms from L3 colored green and labeled in green, respectively. Intriguingly, two XcPqqD molecules interconnect extensively to form a novel saddle-like structure, burying a substantial interface (30.9%, 2407 Å2) from solvent exposure as calculated from the Protein-Protein interface analysis server (http://www.bioinformatics.sussex.ac.uk/protorp/).29 Approximately half (48%) of the contacts comes from the highly conserved hydrophobic residues [Fig. 1(a)], and another half from the surface polar and charged residues. The interface comprises two major regions (circled in orange in [Fig. 1(b)], one consisting of a coil-coil interaction between the N-terminal coil C1 and coils C2′ and C4′ from another subunit, and another consisting of a mixed β1-β2-β3′ β-sheet formation. Figure 1(c) shows the stereo picture of interactions in the first region. Residues Thr3-Arg13 from coil C1 (annotated in blue letters) interacts extensively with residues Asp50′-Ala56′ from coil C2′ (annotated in yellow letters), residues Glu74′-Thr82′ from helix α3′ (annotated in red letters), and residues Arg88′-Leu92′ from coil C4′ (annotated in green letters), forming nine direct and four indirect H-bonds as shown in Figure 1(c). Some hydrophobic interactions, coming from residues Ile4, Leu90′ and Leu92′, are also detected. The N-terminal and C-terminal conformations of XcPqqD are thus well defined due to the presence of these extensive interactions. The interfacial interactions in the mixed β1-β2-β3′ β-sheet region are also extensive. These include the four regular H-bonds between the anti-parallel β2 and β3′ strands, the hydrophobic patch on top of the β2-β3′ strands from residues V28, L30, and V35′, and the hydrophobic cluster from residue L18 of β1 strand, residues Trp27 and Val29 of β2 strand, and residues V38′, L38′, L43′, and Ala46′ of β3′ strand. This hydrophobic patch was also found to extend into the Leu80/Leu84/Lys87/Leu89 cluster of the α1/α2/α3 three-helix bundle, forming an extensive hydrophobic interaction network for XcPqqD dimerization. Peculiarly, although extensive inter-subunit interactions were observed in the crystal, XcPqqD was found to form mainly as a monomer in solution, as demonstrated by the gel filtration and analytical ultracentrifugal experiments (data not shown). The reasons for this phenomenon are unclear. The function of PqqD is currently unknown, but has been speculated as: (1) it could play a role in releasing PQQ product from PqqC; (2) it could be involved in the binding of PqqB with PqqC; (3) it could function as a dioxygenase in the PQQ biosynthesis pathway.20 From the solved structure, we suggest that PqqD may function as a receiver or carrier for PQQ based on the following evidences. First, it adopts an electrostatic surface with rather uneven charge distribution as shown in Figure 2(a). The putative PQQ binding site in the central channel carries a significantly positive charge, whereas the two lobes at both sides carry significantly negative charges. As PQQ is a highly negative charged molecule containing three carboxylate groups, it could be situated well in the central channel. In fact, a phosphate anion was found to be located at this region, forming three H-bonds with the side chains of Arg13, Arg34, and Lys87 as shown in Figure 2a. A docking study of PqqD with PQQ also revealed that all PQQ molecules were found to be situated well at the similar phosphate binding site [Fig. 2(b)]. However, cocrystallization of XcPqqD with PQQ molecules has not been successful, possibly due to the reason that a PQQ may fit into the XcPqqD binding site in several different orientations. It is less likely that PqqD will function as a Fe(II)- and 2-oxoglutarate-dependent dioxygenase, since it lacks the HXD/E…H motif to bind with the iron-sulfur cluster required for the oxidation reaction.30, 31 No extra electron density of metal ion in the map was observed either. Possible PQQ binding site in XcPqqD. (a) Electrostatic surface plot of XcPqqD embedded with an XcPqqD monomer drawn in ribbon. The surface was colored according to the surface potential (positive, blue and negative, red). The phosphate ligand, forming three H-bonds with the side chains of R13, R34, and K87 (connected by dotted yellow lines), was colored orange. The phosphate binding site was circled in orange. (b) Stereo picture of modeled XcPqqD/PQQ complex in van der Waals representation from a docking study. The surface was colored according to the atom colors. The docked PQQ molecule was depicted in sticks, and the H-bonds with the surrounding residues were connected by dotted yellow lines. Two H-bonds, one between the PQQ O4 atom and the XcPqqD R13HN1 atom, and another between the PQQ O5 atom and the XcPqqD K87Nζ atom, were detected and are similar to those of the XcPqqD/phosphate complex shown in Figure 2(a). [Color figure can be viewed in the online issue, which is available at www.interscience.wiley.com.] The final structure of XcPqqD turns out to be very different from that predicted using a de novo method.20 DALI search against the Protein Data Bank returned many proteins with a Z-score higher than 5. However, most of these hits are due to the partial superimposition between the familiar helix-turn-helix motifs. When these hit proteins were superimposed with XcPqqD using several different programs such as PdbViewer,32 PyMOL (DeLano Scientific LLC at http://www.delanoscientific.com) or MUSTANG,33 they usually didn't overlap well. Overall speaking, XcPqqD can be considered to adopt a novel saddle-like fold. The authors appreciate the service of Structural Genomics Databases provided by the GMBD Bioinformatics Core (http://www.tbi.org.tw), NRPGM, Taiwan. We also thank the Core Facilities for Protein X-ray Crystallography in the Academia Sinica, Taiwan for help in crystal screening, the National Synchrotron Radiation Research Center (NSRRC) in Taiwan, and the SPring-8 Synchrotron facility in Japan for assistance of X-ray data collection. The National Synchrotron Radiation Research Center is a user facility supported by the National Science Council, Taiwan, ROC, and the Protein Crystallography Facility is supported by the National Research Program for Genomic Medicine, Taiwan, ROC.