|Home | About | Journals | Submit | Contact Us | Français|
Lumazine protein (LumP) is a fluorescent accessory protein having 6,7-dimethyl-8-(1′-d-ribityl) lumazine (DMRL) as its authentic chromophore. It modulates the emission of bacterial luciferase to shorter wavelengths with increasing luminous strength. To obtain structural information on the native structure as well as the interaction with bacterial luciferase, we have determined the crystal structures of LumP from Photobacterium kishitanii in complexes with DMRL and its analogues, riboflavin (RBF) and flavin mononucleotide (FMN), at resolutions of 2.00, 1.42, and 2.00 Å. LumP consists of two β barrels that have nearly identical folds, the N-terminal and C-terminal barrels. The structures of LumP in complex with all of the chromophores studied are all essentially identical, except around the chromophores. In all of the structures, the chromophore is tethered to the narrow cavity via many hydrogen bonds in the N-terminal domain. These are absent in the C-terminal domain. Hydrogen bonding in LumP-FMN is decreased in comparison with that in LumP-RBF because the phosphate moiety of FMN protrudes out of the narrow cavity. In LumP-DMRL, the side chain of Gln65 is close to the ring system, and a new water molecule that stabilizes the ligand is observed near Ser48. Therefore, DMRL packs more tightly in the ligand-binding site than RBF or FMN. A docking simulation of bacterial luciferase and LumP suggests that the chromophore is located close enough for direct energy transfer to occur. Moreover, the surface potentials around the ligand-binding sites of LumP and bacterial luciferase exhibit complementary charge distributions, which would have a significant effect on the interaction between LumP and luciferase.
Bioluminescent organisms are widely distributed in nature and comprise a remarkably diverse set of species (8, 11, 40). Among them, luminous bacteria have long been known to exist (15). Many studies of bacterial luminescence have been reported (34-36), and its biochemical mechanism has been the subject of many investigations (13, 14). The luminous bacteria illuminate using bacterial luciferases, which emit blue-green light through catalytic oxidation of reduced flavin mononucleotides (FMN) with long-chain aliphatic aldehydes. The maximal emission wavelengths (475 to 486 nm) of some Photobacterium strains are blue shifted with respect to those of purified luciferase (495 nm). This blue shift is induced by a fluorescent accessory protein called lumazine protein (LumP; 21 kDa) (22, 28, 29). LumP was first isolated from the luminescent marine bacterium Photobacterium phosphoreum (38) and later from Photobacterium leiognathi (28). Interestingly, LumP not only decreases the emission wavelength but also enhances the intensity of the light (29, 32). LumP possesses a noncovalently bound fluorophore, 6,7-dimethyl-8-ribityllumazine (DMRL), which is known to be the direct biosynthetic precursor of riboflavin (RBF) (10). Another accessory protein of bacterial luciferase, the yellow fluorescent protein (YFP; 23 kDa), was found in Vibrio fischeri Y-1. YFP contains RBF or FMN as a chromophore and modulates the emission wavelength of bacterial luciferase to yellow light (~540 nm) (3, 7, 25, 33). LumP and YFP share amino acid sequence homology (37%). These fluorescent accessory proteins are believed to interact with bacterial luciferases in the intermediate states during the luciferase reaction (29, 32) and to transform the chemical energy from the luciferase reaction into the excitation of their bound, fluorescent ligands. However, the detailed mechanisms of the interaction and the interactive process have not yet been understood.
LumP shares a high amino acid sequence homology with RBF synthase, which generates one RBF molecule from two DMRL molecules (10, 12). RBF synthase has a homo-trimer. RBF synthase consists of two segments with similar sequences and two closely related folds, the N-terminal and C-terminal domains (12, 24, 26, 39). In the crystal structure, each domain contained one DMRL derivative molecule. It is proposed that the catalytic site is located at the interface between the N-terminal domain of one subunit and the C-terminal domain of the adjacent subunit. Recently, Chatwell et al. reported the crystal structure of the lumazine protein of P. leiognathi (LumPPL) in complex with RBF at a resolution of 2.5 Å (5). The overall structure is similar to that of RBF synthase. RBF fits into a narrow cavity in the N-terminal domain, but no ligand was observed in the C-terminal domain. It should be noted that they determined the crystal structure of a mutant protein (L49N) whose ability to bind to DMRL and RBF is partially impaired (16). In addition, it is known that LumP modulates the color of light generated by bacterial luciferase only when it is complexed with DMRL. Therefore, structural information on the wild-type LumP-DMRL complex is required to reveal the details of the ligand-binding mode of LumP at the LumP-luciferase interaction site.
Here, we present the crystal structures of LumP from P. kishitanii (LumPPK) in complexes with the authentic chromophore, DMRL, and its analogues, RBF and FMN, at 2.00-, 1.42-, and 2.00-Å resolutions. These are the highest-resolution structures that have been determined for this family of proteins. In addition, a docking simulation of bacterial luciferase and LumPPK suggested that their chromophores are located closely enough for direct energy transfer to occur.
A luminous bacterium was collected and isolated from the cuttlefish Todarodes pacificus and identified as P. kishitanii from the sequence of the gyrB gene (36). The LumPPK gene was amplified using KOD Plus DNA polymerase (Toyobo, Osaka, Japan) and forward (5′-GGAATTCCATATGTTCAAAGGTATAGTTCA-3′ [the NdeI restriction site is underlined]) and reverse (5′-CGGAATTCTTAGTCCCATTCATTTGATA-3′ [the EcoRI restriction site is underlined]) primers. The PCR-amplified fragment was subcloned into the pT7 Blue T vector (Novagen, Madison, WI). The LumP expression plasmid was constructed by subcloning the fragment into the NdeI-EcoRI site of pET23a (Novagen).
The resulting plasmid was transformed into Escherichia coli strain BL-21(DE3). Transformed cells picked from a single colony were inoculated into 2 liters of LB medium containing ampicillin. After overnight culture at 37°C with shaking, the cells were harvested by centrifugation. The pellet was suspended in 50 ml of buffer A (50 mM Tris-HCl, pH 7.5), and the cells were disrupted by sonication. After recentrifugation, the pellet was suspended in 100 ml of buffer A containing 0.5 mM RBF. Urea was then added to the pellet suspension to give 1.5 M, and the mixture was left at 4°C for 4 h. After centrifuging again, the supernatant was dialyzed against buffer A. The dialyzed protein, to which 30% saturated ammonium sulfate had been added, was applied to a butyl-toyopearl column (Tosoh, Tokyo, Japan) and eluted using a 30 to 0% saturated ammonium sulfate gradient. After dialyzing again against buffer A, the protein was concentrated to 22 mg/ml using ultrafiltration.
RBF and FMN were purchased as commercial reagent grade chemicals (Wako Pure Chemical Industries). DMRL was prepared as described previously (17). The DNA fragment containing the rib operon genes (ribG, -B, -A, and -H) of Bacillus subtilis strain 168 was amplified from its genomic DNA by using KOD Plus DNA polymerase (Toyobo, Osaka, Japan) and forward (5′-CATGCCATGGAAGAGTATTATATGAAGCTGGCCT-3′ [the NcoI restriction site is underlined]) and reverse (5′-CGGATCCGTTATTCAAATGAGCGGTTTAAATTTGC-3′ [the BamHI restriction site is underlined]) primers. The PCR product was subcloned into the pT7 vector by TA cloning (pTRib). To prevent RBF synthase activity, the F2A mutation was introduced into the ribB gene, which codes for RBF synthase (19). An inactive mutant (F2A) was constructed using a QuikChange site-directed mutagenesis kit (Stratagene, La Jolla, CA), with the plasmid pTRib used as the template. This plasmid was transformed into E. coli, and the cells were grown overnight in a shaking flask containing 1 liter of M9 culture medium. The cells were collected by centrifugation, and the supernatant was used for the experiment as the DMRL solution.
Twenty milliliters of 0.5-mg/ml LumP-RBF in buffer A was dialyzed overnight against buffer A containing 4 mM dithiothreitol and 6 M urea. To prepare the LumP-FMN complex, this apo-protein solution was mixed with the FMN solution and then concentrated to 200 μl by ultrafiltration. To prepare the LumP-DMRL complex, the apo-protein solution was dialyzed overnight against the DMRL solution. Excess salts and nutrients were removed by dilution in buffer A and reconcentrated by ultrafiltration.
The crystals of LumP-RBF were grown using the hanging drop vapor diffusion method, a protein solution (5 mg/ml), and a reservoir solution (20% [wt/vol] polyethylene glycol 4000, 0.2 M MgCl2, 100 mM HEPES, pH 6.8). By use of the same method, crystals of the LumP-FMN and LumP-DMRL complexes were obtained using 20% (wt/vol) polyethylene glycol 4000, 0.2 M MgCl2, 100 mM Tris-HCl, pH 8.5, as the reservoir solution.
X-ray diffraction data for LumP-RBF were collected at 100 K using a charge-coupled-device detector system (Rigaku/MSC Jupiter210) on the BL26B1 beam line of SPring-8 (Hyogo, Japan). The LumP-FMN and LumP-DMRL data were collected at 95 K using an ADSC charge-coupled-device detector system (ADSC Quantum210r and Quantum 270) on the AR-NW12A and AR-NE3A beam lines at the Photon Factory (Tsukuba, Japan). The diffraction data were processed using the program HKL2000 (30). The initial phases of LumP-RBF were determined using a molecular replacement method, with the structure of LumPPL (Protein Data Bank [PDB] code 3DDY) used as a probe model, with the program Molrep (41). The structures of LumP-FMN and LumP-DMRL were also determined using the molecular replacement method, with the structure of LumP-RBF used as the probe. The models were corrected on the (2Fo-Fc) electron density map by using the program Coot (9), and the structures without solvent molecules were refined using maximum resolution data. The structural refinements were calculated using the program REFMAC5 (27) in the CCP4 suite (6). Solvent molecules were gradually introduced if the peaks above 4.0 σ in the (Fo-Fc) electron density map were in the hydrogen bond range. At this stage, each ligand was located at the ligand-binding site and added to the structure. To avoid overfitting of the diffraction data, a free R factor, in which 5% of the test set had been excluded from the refinements, was monitored. Finally, the structures were refined to Rwork/Rfree ratios of 0.192/0.234 (LumP-RBF), 0.196/0.255 (LumP-FMN), and 0.195/0.255 (LumP-DMRL). The data collection statistics and the refinement statistics are summarized in Table Table1.1. All of the structural figures in this paper were prepared using the program PyMOL (http://www.pymol.org/).
The atomic coordinates and structure factors for LumP-RBF, LumP-FMN, and LumP-DMRL have been deposited in the PDB under accession codes 3A35, 3A3B, and 3A3G, respectively.
The LumP gene from P. kishitanii was amplified and sequenced (DDBJ accession number AB508352). The sequence was identical to that of P. phosphoreum. This is not surprising, because all strains of luminous bacteria share very high nucleotide sequence homology (2, 20). LumP from Photobacterium kishitanii (LumPPK) has been expressed as inclusion bodies in E. coli, solubilized in a 1.5 M urea buffer containing RBF, refolded by removing the urea, and finally, purified by column chromatography. To confirm that the purified LumPPK still bound RBF, we measured the absorbance and fluorescence spectra of the purified protein. Both spectra exhibited absorbance and fluorescence emission peaks at 460 and 535 nm (excitation wavelength, 445 nm), respectively, which had been shifted to longer wavelengths relative to those of free RBF in the same buffer (data not shown). These spectra are very similar to those of LumPPL, indicating that RBF is properly bound to LumPPK. By use of size exclusion chromatography with a multiangle laser light scattering detector, the molecular mass of the LumPPK protein prepared was determined to be approximately 23 kDa. This is in good agreement with the deduced molecular mass of a monomer, 21 kDa (data not shown). These results coincide with the previous report well (18).
The crystal structures of LumPPK in complexes with RBF, FMN, and DMRL were determined at 1.42-, 2.00-, and 2.00-Å resolutions, respectively. The three LumPPK complex structures are isomorphous, and there are two protein molecules (Mol A and B) in each asymmetric unit. Each molecule exists as a monomer, as demonstrated by size exclusion chromatography with a multiangle laser light scattering detector. The LumP-RBF structure contains 184 residues (residues 1 to 184) of the 190 sequenced residues for both Mol A and B in the asymmetric unit. As the last 6 C-terminal residues constitute the loop region, the flexibility of the region seems to cause a disordered structure and a lack of electron density. Similarly, the last 6 C-terminal residues of LumP-FMN and LumP-DMRL were also not observed. In addition, no significant electron density was obtained for residues 88 to 94 in the LumP-FMN complex or for residues 87 to 93 in the LumP-DMRL complex. The residues correspond to flexible loop regions connecting the N-terminal and C-terminal domains. The electron densities for DMRL, RBF, and FMN were clearly visible. The final (2Fo-Fc) electron density maps of the crystal structures of the LumP complexes (contoured at 1 σ) showed that all of the amino acid residues, the ligand, and the solvent molecules were well fitted. The stereochemical quality of the models was analyzed using the program PROCHECK (21). In the three complexes, the structures of Mol A and Mol B are almost identical. The root mean square deviations (RMSD) (Å) for the Cα atoms between Mol A and Mol B are 0.534 in LumP-RBF, 0.438 in LumP-FMN, and 0.317 in LumP-DMRL. However, the ligand-binding site of Mol B is more exposed to the solvent than that of Mol A. Therefore, we describe the structure of Mol B hereafter.
The Cα RMSD values for each complex were calculated as 0.184 (LumP-RBF and LumP-DMRL), 0.214 (LumP-RBF and LumP-FMN), and 0.317 (LumP-FMN and LumP-DMRL) Å. The LumPPK monomer consists of two β barrels with nearly identical folds, i.e., the N-terminal and C-terminal barrel domains (Fig. (Fig.1).1). The two domains share 27% amino acid sequence identity. Both β-barrel domains are composed of six antiparallel β strands. The N-terminal barrel domain includes six β strands (β1 to β6) and two helixes (α1 and α2), whereas the C-terminal domain is formed by β strands (β7 to β12) and three helixes (α3, α4, and α5). The Cα RMSD between the N- and C-terminal domains (1 to 90 and 91 to 184) is calculated to be 0.902 Å in LumP-RBF. These structural features coincide well with those of RBF synthase (24) and LumPPL (5).
In the LumPPK-RBF complex, an RBF molecule is bound to the N-terminal domain, and no ligand is found at the C-terminal domain (Fig. (Fig.1).1). The RBF molecule is positioned in a shallow groove and is surrounded by residues 47 to 55 of strand β4, residues 58 to 63 of strand β5, and residues 64 to 69 of helix α2. In particular, Gln65 covers the ligand with its long side chain and is hydrogen bonded to Thr50 via two water molecules. The electron densities of RBF were clearly observed (Fig. (Fig.2).2). Almost all of the polar groups of RBF formed hydrogen bond contacts with LumPPK. We observed 29 direct and water-mediated intermolecular hydrogen bonds, involving Ser48, Thr50, Asp62, Asp64, Gln65, Ala66, Thr69, Thr99, Gly100, Asn101, and Ile102. In particular, Thr50, which is strongly conserved in LumP homologues (including RBF synthase), participates in two hydrogen bonds with the ligand: the Oγ atom is hydrogen bonded directly to N-5 of the ring system, and the Thr50-O atom is hydrogen bonded to O-4 of the RBF molecule via one water molecule. This water molecule was also observed in the crystal structures of E. coli and Schizosaccharomyces pombe RBF synthase and has been suggested to play an important role in stabilizing the ligand.
To determine why RBF is absent from the C-terminal domain, we compared the structures of the ligand-binding regions in the N- and C-terminal domains. In the C-terminal domain, the bulky side chain of His145 protrudes toward the binding site for the RBF ring system. Thus, steric hindrance by His145 prevents the binding of RBF in the C-terminal domain. The orientations of RBF in LumPPK are also completely matched with those of RBF synthase. In contrast, some distinct differences were observed in the orientation of the ribityl chain between the LumPPK and LumPPL complexes (Fig. (Fig.3).3). In the LumPPK complex, O-3′ of RBF is hydrogen bonded to Gln65, Ala66, Thr69, and Asn101 via three water molecules, and O-4′ of RBF is hydrogen bonded to Gly100 and Ile102. However, in the LumPPL complex, O-3′ of RBF is hydrogen bonded to Gly100 and Asn101 (corresponding to Gly100 and Asn101 of LumPPK), and O-4′ is hydrogen bonded to Gln65 (corresponding to Gln65 of LumPPK). These discrepancies may be due to the differences in the strains or preparations. However, considering the complete conservation of the amino acid residues in the ligand-binding site and the structural conservation of the ribityl side chain and ring system between the LumPPK complex and RBF synthase with RBF (26) and the inhibitor, 6-carboxyethyl-7-oxo-8-ribityllumazine (12), the LumPPL complex is likely to share the same orientation of its ribityl side chain as the LumPPK complex and RBF synthase. We should note the following possibilities. First, the resolution of the LumPPL complex may not be sufficient for evaluation of the structure of the water molecules in the ligand-binding site. It is obvious that the hydrogen bond networks including the hydration water molecules are inevitable for ligand stabilization. The lack of hydration water molecules might result in inaccurate orientation of the ribityl side chain. Second, the binding mode of RBF might be affected by the L49N mutation in the LumPPL complex. Leu49 is located at the ligand-binding site, and the mutation increased the dissociation constant greatly (16). Although the locations of residue 49 are not significantly different between the LumPPK and LumPPL proteins, the L49N mutation may affect the polarity of the ligand-binding site to induce a different orientation of the ribityl moiety in the mutant LumPPL complex.
The binding mode of FMN for binding to LumP (Fig. (Fig.4)4) is very similar to that of RBF (Fig. (Fig.2).2). The phosphate moiety of FMN protrudes out of the shallow cavity like a shrimp tail and forms new hydrogen bonds with Thr69 and Asn101. However, three hydration water molecules that are observed in the LumP-RBF complex (hydrogen bonded to O-3′ of RBF, Gln65, and Thr69) were not present. Moreover, O of Ile102 does not hydrogen bond to the ribitol moiety, because the ribitol terminal of FMN protrudes outside the binding site. Together, these results indicate that the hydrogen bond network is smaller in the LumP-FMN complex than in the LumP-RBF complex. This observation is consistent with the report that the dissociation constant for FMN is higher than that for RBF (31).
The structures of the ring system and the ribityl chain of DMRL are similar to those of RBF (Fig. (Fig.5a).5a). The binding arrangement of LumP-DMRL is well conserved with that of LumP-RBF, except around Gln65 and Ser48 (Fig. (Fig.5b).5b). The side chain of Gln65 is closer to the ring system than in LumP-RBF or LumP-FMN. In addition, Gln65 is hydrogen bonded to Thr50 via two water molecules. This hydrogen bond network is likely to contribute to the stability of DMRL in the binding site. Moreover, a new water molecule is observed near Ser48. This new water molecule fills the space opened up by the difference in the sizes of the rings of RBF and DMRL and is hydrogen bonded to Ser48 and a hydration water molecule that stabilizes the side of the methyl group of the lumazine ring and the hydrogen network around the ribityl moiety. Thus, DMRL is more tightly packed at the ligand-binding site than RBF or FMN. Consistent with these observations, Petushkov et al. studied the ligand-binding properties and concluded that the dissociation constant of LumPPL for DMRL is lower than that for RBF and FMN (31). On the other hand, when the ligand binds to the LumPPL and LumPPK proteins, a hypsochromic shift is observed for DMRL, but a bathochromic shift is observed for RBF (16). These fluorescence properties may be affected by the differences in the binding arrangements.
LumP has been reported to form a specific 1:1 complex with a catalytic intermediate species of bacterial luciferase (31). We performed virtual docking calculations of the LumPPK-DMRL complex with a bacterial luciferase from Vibrio harveyi (PDB code 3FGC) (4) to obtain structural information on their interaction. The amino acid sequence of V. harveyi luciferase has approximately 61% (luxA) and 48% (luxB) identity with that of P. kishitanii, and a LumP protein from P. phosphoreum that has an amino acid sequence identical to that of LumPPK was reported to interact with V. harveyi luciferase (23). The active site of bacterial luciferase is formed by residues in the α subunit with FMN (4).
We used the protein-protein docking program PatchDock (http://bioinfo3d.cs.tau.ac.il/PatchDock/patchdock.html) to build models of the LumPPK-luciferase complex and the program FireDock to refine the docked models (http://bioinfo3d.cs.tau.ac.il/FireDock/index.html) (1, 37). One of the models exhibiting a high score in FireDock showed that the ligand-binding site of LumP-DMRL contacted the reentrant surface of V. harveyi luciferase near the active site (Fig. (Fig.6a).6a). In this model, the ring system of DMRL is located very close to the isoalloxazine ring of the FMN of V. harveyi luciferase (approximately 10 Å). The close contact between the chromophores of LumPPK and the V. harveyi luciferase might support direct energy transfer between them. The electrostatic potential of the molecular surface shows that the DMRL binding site of LumPPK is covered by negatively charged residues (Asp19, Asp20, Asp62, and Asp64) (Fig. (Fig.6b),6b), while the active site of V. harveyi luciferase forms a concave shape, which is surrounded by positively charged residues (Arg107, Arg125, Lys259, Arg290, and Arg291). Thus, this countercharge distribution could be responsible for the interaction between LumPPK and bacterial luciferases. In the other calculated models, the C-terminal domain of the LumP protein that contained no DMRL molecule contacted with the active site of the α or β subunit of bacterial luciferase. These models are excluded because the energy transfer cannot occur. These models should be generated due to the structural homology between the N- and C-terminal domains of LumP.
In conclusion, we have determined the crystal structures of LumPPK in complexes with the authentic chromophore, DMRL, and its derivatives. This is the first report of the high-resolution structures of fluorescent accessory proteins for bacterial luciferase with its native chromophore. In addition, by using virtual docking calculations, we have proposed the structure of the bacterial luciferase complex and its fluorescent accessory protein for the first time. Although additional studies will be required, the present study gives insights into the LumP-bacterial luciferase reaction mechanism.
We are grateful to the beam line assistants at the Photon Factory (PF) for data collection at beam lines AR-NW12A and AR-NE3A. We also thank Yuki Nakamura, Go Ueno, Takaaki Hikima, and Masaki Yamamoto for automated data collection at SPring-8 by use of the mail-in system.
The work reported here is a part of the support program for improving graduate school education of the Human Resource Development Program for Scientific Powerhouse, which is financially supported by the Ministry of Education, Culture, Sports, Science and Technology, Japan, through Tokyo University of Agriculture & Technology.
Published ahead of print on 23 October 2009.