Crystal Structures of Nonoxidative Zinc-dependent 2,6-Dihydroxybenzoate (γ-Resorcylate) Decarboxylase from Rhizobium sp. Strain MTP-10005*

Reversible 2,6-dihydroxybenzoate decarboxylase from Rhizobium sp. strain MTP-10005 belongs to a nonoxidative decarboxylase family. We have determined the structures of the following three forms of the enzyme: the native form, the complex with the true substrate (2,6-dihydroxybenzoate), and the complex with 2,3-dihydroxybenzaldehyde at 1.7-, 1.9-, and 1.7-Å resolution, respectively. The enzyme exists as a tetramer, and the subunit consists of one (αβ)8 triose-phosphate isomerase-barrel domain with three functional linkers and one C-terminal tail. The native enzyme possesses one Zn2+ ion liganded by Glu8, His10, His164, Asp287, and a water molecule at the active site center, although the enzyme has been reported to require no cofactor for its catalysis. The substrate carboxylate takes the place of the water molecule and is coordinated to the Zn2+ ion. The 2-hydroxy group of the substrate is hydrogen-bonded to Asp287, which forms a triad together with His218 and Glu221 and is assumed to be the catalytic base. On the basis of the geometrical consideration, substrate specificity is uncovered, and the catalytic mechanism is proposed for the novel Zn2+-dependent decarboxylation.

Recently, Ishii et al. (18) reported the purification and characterization of GRDC from Rhizobium radiobacter WU-0108, Agrobacterium tumefaciens IAM12048 (19), and Pandoraea sp. 12B-2 (20). They reported that these enzymes also catalyze the reversible decarboxylation of 2,6-dihydroxybenzoate without cofactors and have a similar substrate specificity. Orotidine 5Ј-monophosphate decarboxylase catalyzes the cofactor-independent decarboxylation. On the basis of the x-ray structure of the enzyme, it is proposed that the decarboxylation of orotidine 5Ј-monophosphate proceeds by an electrophilic substitution mechanism in which decarboxylation and carbon-carbon bond protonation by Lys 62 occur in a concerted way (21,22).
To elucidate the overall and active-site structure, the substrate recognition, and the reaction mechanism, we have determined the crystal structures of GRDC from the Rhizobium sp. strain MTP-10005 in the native form, GRDC complexed with the substrate 2,6-dihydroxybenzoate, and GRDC complexed with substrate analogue 2,3-dihydroxybezaldehyde at 1.7-, 1.9-, and 1.7-Å resolutions, respectively. Interestingly, these structures revealed that Zn 2ϩ ion is bound to the active center to catalyze the reaction. We now report the first structure of GRDC as a novel Zn 2ϩ -dependent reversible decarboxylase.

EXPERIMENTAL PROCEDURES
Crystallization and Data Collection-The purification of the enzyme has been reported elsewhere (13). The initial screening for the crystallization conditions was performed using the * This work was supported by the National Project on Protein Structural and Functional Analyses and in part by research grants from the Japan Foundation of Applied Enzymology and the Kansai University High Technology Research Center. The costs of publication of this article were defrayed in part by the payment of page charges. This article must therefore be hereby marked "advertisement" in accordance with 18 U.S.C. Section 1734 solely to indicate this fact.  (23). Crystallization of the native GRDC was carried out at 293 K by the hanging drop vapor diffusion method, using 3.6 mg/ml protein solution and a reservoir solution containing 15% (w/v) PEG8000, 5% (v/v) ethylene glycol, 10% (v/v) glycerol, 100 mM Tris-HCl, pH 7.5. The native crystal was obtained from a drop composed of a 1:1 volume ratio of the reservoir solution and the protein solution. Within 2 weeks, crystals had grown to dimensions of ϳ0.05 ϫ 0.15 ϫ 0.40 mm 3 . Crystals of GRDC⅐2,6-dihydroxybenzoate and GRDC⅐2,3-dihydroxybenzaldehyde were obtained at 293 K by soaking the native crystals in a reservoir solution containing 40 mM 2,6-dihydroxybenzoate for 10 min and 40 mM 2,3-dihydroxybenzaldehyde for 40 min, respectively, before data collection. The x-ray diffraction data set for the native crystal was collected to 1.7 Å resolution at 100 K using a wavelength of 1.00 Å on the NW12 station equipped with an ADSC Quantum 4R CCD detector system at the photon factory, KEK (Tsukuba, Japan). The x-ray data sets for the GRDC⅐2,6-dihydroxybenzoate crystal and the GRDC⅐2,3-dihydroxybenzaldehyde crystal were collected to 1.9 and 1.7 Å resolution at 100 K using a wavelength of 1.0 Å on the BL41XU station and a ADSC Quantum 315 CCD detector at the SPring-8 (Hyogo, Japan). These three crystals are isomorphous with a space group of P2 1 2 1 2 1 and have average cell dimensions of a ϭ 109.2, b ϭ 113.66, and c ϭ 119.2 Å. There are four subunits in the asymmetric unit, and ϳ44% of the crystal volume is occupied by solvent (24). The heavy atom derivatives were obtained by soaking unliganded crystals in solutions containing heavy atom reagents. The data sets for heavy atom derivatives were collected to 2.15 and 2.45 Å resolution at 100 K on the NW12 station at the photon factory (KEK). All the data were processed and scaled using the program HKL2000 (25) ( Table 1).
Structure Determination and Refinement-The structure of the native GRDC was solved by the multiple isomorphous replacement with anomalous scattering method, using two isomorphous data sets. The scaling of all the data and the map calculations were performed with the CCP4 program suite (26). The positions of one mercury and five platinum sites were determined with the program SOLVE (27). The resulting isomorphous replacement with anomalous scattering map has a mean figure-of-merit of 0.335 at a resolution of 20-3.0 Å. The map was significantly improved by the process of solvent flattening and local symmetry averaging with the program RESOLVE (28). The mean figure-of-merit reached 0.712 with the same resolution range. The map was of good quality, and the model of the subunit was gradually built into the 3.0 Å map through several cycles of model building using the program O (29). The model was refined by simulated annealing and energy minimization with the program CNS (30). A random sample with 10% of the reflections was set apart for the calculation of the R free . Solvent molecules were picked up on the basis of the peak heights (3.0 ) and distance criteria (4.0 Å from protein or solvent) from the sigmaA-weighted F o Ϫ F c map. The procedure converged to R factor and R free values of 0.180 and 0.200, respectively, calculated for 163,233 reflections observed in a 40.3-1.70 Å resolution range ( Table 1).
The same refinement procedure was applied to GRDC⅐2,6dihydroxybenzoate and GRDC⅐2,3-dihydroxybenzaldehyde but using the coordinates of the unliganded enzyme as the initial model. When the R factor value became less than 30%, the simulated annealing 2F o Ϫ F c map showed residual electron densities corresponding to the bound 2,6-dihydroxybenzoate and 2,3-dihydroxybenzaldehyde. Solvent molecules were picked up from the sigmaA-weighted F o Ϫ F c map. Further model building and refinement cycles gave R factor values of 0.176 and 0.184 and R free values of 0.206 and 0.208 for GRDC⅐2,6-dihydroxybenzoate and GRDC⅐2,3-dihydroxybenzaldehyde, respectively ( Table 1).
Quality The final models were validated with the PROCHECK (31) and CNS (30) programs. 99.4% of the amino acid residues in the final model were located in the most favorable and additionally allowed regions of the Ramachandran plot (Table 1). We note that one residue in the active site, Asp 27 , is located in the disallowed region in subunits A and B of the native GRDC. It is confirmed that the conformation of Asp 27 is correct on the basis of the electron density map. We use subunit A and subunit B, if necessary, as the reference standards for the native and complex forms of GRDC. Structure diagrams were drawn using the programs MOLSCRIPT (32), RASTER3D (33), and POV-SCRIPTϩ (34).

RESULTS AND DISCUSSION
Overall and Subunit Structure-The overall structure of GRDC⅐2,6-dihydroxybenzoate is shown in Fig. 1A. The molecular mass of the enzyme is about 151 kDa by gel filtration and that of the subunit is 37.5 kDa by SDS-PAGE, suggesting that the GRDC exists as a homotetramer (13). The polypeptide chain is folded into a tetrameric form with a noncrystallographic 222 axis, having approximate dimensions of 95 ϫ 85 ϫ 50 Å. One subunit in the tetramer interacts with the other three subunits, and the surface areas of the subunit interfaces are 3431 Å 2 for subunits A and B, 546 Å 2 for subunits A and C, and 206 Å 2 for subunits A and D. The largest of these areas is found between the subunits A and B (or C and D) indicating that the tetramer may be considered to be an assembly of two dimer units (a dimer of dimers) around a 2-fold axis.
The subunit structure of GRDC⅐2,6-dihydroxybenzoate with secondary structure assignments by the program DSSP (35) is shown in Fig. 1, B and C. The subunit C␣ carbon atoms except for Ala 21 -Asp 28 in the native GRDC can be superimposed onto the corresponding ones in GRDC⅐2,6dihydroxybenzoate and GRDC⅐2,3-dihydroxybenzaldehyde within root mean square deviations of 0.08 and 0.05 Å with the maximum displacement of 0.25 and 0.23 Å, respectively. Thus, the overall structure of the native GRDC is quite similar to that of the complex with the substrate or analogue when Ala 21 -Asp 28 is neglected. However, the region of Ala 21 -Asp 28 is mobile and plays an important role in substrate recognition (see "Induced Fit of Linker 1").
The subunit is divided into a (␣␤) 8 barrel, three linkers (linker 1, Glu 9 -Asp 38 ; linker 2, Arg 166 -Thr 86 ; and linker 3, His 232 -Glu 255 ), and a C-terminal tail (Ser 304 -C-terminal). The barrel has a typical TIM-barrel structure (␤-strand S1, ␣-helix H3, S3, H4, S4, H5, S5, H6, S6, H7, S7, H8, S8, H10, S9, and H11) (Fig. 1, B and C). The linker 1 is inserted between S1 and H3 forming S2, H1, and H2. The random coiled linker 2 connects S6 and H7. The linker 3, composed of a random coil followed by H9, is between H8 and S8. The linkers play an important role in the formation of the dimeric unit because the residues from the linkers contribute to 46.5% of accessible surface area in the interface between subunits A and B. The C-terminal tail located at the dimer-dimer interface forms 40.5 and 46.6% of the interfaces between subunits A and C and subunits A and F, respectively, serving to form the tetramer.
The amino acid sequence of GRDC displays no significant homology to those of the known decarboxylases requiring a cofactor such as pyridoxal 5Ј-phosphate, thiamine pyro-  (39)), 15.9 for renal dipeptidase (13%, 1ITQ (40)), 14.9 for phosphotriesterase (15%, 1PSC (41)), 14.0 for the catalytic domain of cytosine deaminase (11%, 1K6W (42)), 13.6 for the catalytic domain of n-acetylglucosamine-6-phosphate deacetylase (9%, 1UN7 (43)), 13.2 for the ␣-sub-unit of urease (13%, 1UBP (44)), 13.1 for D-aminoacylase (9%, 1M7J (45)), and 13.1 for adenosine deaminase (12%, 1A4M (46)). Because these enzymes are members of the recently identified TIM-barrel metallo-dependent hydrolase superfamily, GRDC is classified as a member of this superfamily. The superfamily is mainly divided into two subgroups depending on a bi-metal or mono-metal enzyme (47,48), and GRDC thus belongs to the mono-metal-dependent hydrolase subgroup together with Fe 3ϩ -dependent cytosine deaminase (42) and Zn 2ϩ -dependent adenosine deaminase (46). Possibly, the active site structure of GRDC is more similar to that of adenosine deaminase than cytosine deaminase in terms of the mono-Zn 2ϩ binding. Induced Fit of Linker 1-One of the interesting features of GRDC is the insertion loop's conformational change to close the active site upon binding of the substrate (Fig. 1C). The C␣ atoms of 314 residues (1-19 and 31-325) of 325 residues in the native form can be superimposed onto the corresponding ones in GRDC⅐2,6-dihydroxybenzoate within 0.5 Å with a root mean square deviation of 0.08 Å. However, the root mean square deviation for 11 residues excluded from the calculation is 1.36 Å with a maximum displacement of 2.48 Å. This indicates that, upon binding of the substrate (2,6dihydroxybenzoate), a part (Ser 20 -Ala 30 ) of the linker 1 changes its main-chain conformation to close the active site from the solvent region. The bound substrate is almost isolated from the solvent region because accessible surface area of the substrate is reduced from 298 to 1 Å 2 (Fig. 1A). This is also the case for the GRDC⅐2,3-dihydroxybenzaldehyde with the reduced accessible surface area of 3 Å 2 . The linker 1 acts as an active site lid and plays important roles in the substrate recognition and catalytic actions of the enzyme. It has been reported that isoaspartyl dipeptidase belonging to the hydrolase superfamily displays a conformational change similar to that of GRDC (39).
Zinc Coordination and Active Site Structure-When the R factor value became less than 30%, the simulated annealing 2F o Ϫ F c map showed a marked electron density peak located within 2.0 -2.3 Å from Glu 8 , His 10 , His 218 , and Asp 287 . The peak was identified as that of Zn 2ϩ ion by an anomalous scattering experiment, although Zn 2ϩ ion was not contained in the buffer solution for purification or crystallization, and GRDC has been reported to require no cofactor. Zn 2ϩ absorption edge was determined to be 1.2805 Å by x-ray absorption experiment. An electron density was observed at the position on Bijvoet difference map based on the data at ϭ 1.0000 Å, whereas there was no significant electron density on that based on the data at ϭ 1.2820 Å. We also determined the Zn 2ϩ concentration by atomic absorption spectroscopy. Zinc atomic absorbencies followed at a wavelength of 213.9 nm, using a Hitachi Z-8200 model. The calculated number of Zn 2ϩ ion per each tetramer was 4.01.
Active site components and bound Zn 2ϩ ion in the native GRDC are shown in Figs  three equatorial ligands. The side chains of Glu 8 , His 10 , His 164 , and Asp 287 are arranged in geometrical positions favorable to act as ligands for Zn 2ϩ by interactions with the neighboring residues of the active site. Glu 8 , His 10 , and His 164 interact with Tyr 162 , Ser 58 , and Asn 128 , respectively. The carboxylate oxygen of Asp 287 is hydrogen-bonded to the N-1 atom of His 218 , and the N-3 atom of His 218 is further hydrogen-bonded to the carboxylate of Glu 221 , implying that the Asp:His:Glu triad forms the charge relay system. The Asp 287 and His 218 dyad is conserved in adenosine and cytosine deaminases, but it has been observed in the disallowed region of the Ramachandran plot (42,46). The Zinc coordination scheme is very similar to those observed in adenosine deaminase and cytosine deaminase, although Glu 8 in GRDC is replaced by a His residue in both deaminases.
The active site cavity, filled with eight water molecules, resides at the C-terminal side groove of the (␣␤) 8 barrel. The bottom of the cavity is composed of Zn 2ϩ ion and the surrounding residues, and the active site wall is encircled by the C-terminal residues of the barrel (Ala 61 , Pro 165 , Pro 185 , Phe 189 , and Phe 290 ), the linker 1 residues (Ser 20 and Phe 23 ), and the linker 3 residues (Asn 234 * (* indicates a residue from another subunit of the dimer unit) and Arg 229 *).
Binding Mode of the Substrate and Substrate Analogue-2,6-Dihydroxybenzoate is a true substrate for GRDC, and the GRDC⅐2,6-dihydroxybenzoate crystal was obtained by soaking the native crystal in a solution supplemented with 2,6-dihydroxybenzoate for 10 min. The stereo structure of GRDC⅐2,6dihydroxybenzoate and the hydrogen-bonding scheme are shown in Figs. 2B and 3B, respectively. Upon binding of the substrate, three water molecules are expelled from the cavity. Concurrently, the linker 1 changes not only its main-chain but also its side-chain conformation to enclose the substrate in the cavity. The linker 1 residues, Ser 20 and Phe 23 , behave like a lid for the active site cavity. Phe 23 changes its side-chain direction toward the active site, and its side-chain phenyl group makes a van der Waals contact with the bound substrate with the shortest contact of 3.5 Å. The active site residues except for the linker 1 do not change their main-chain and side-chain conformations and retain the interactions among them upon binding of the substrate.
The substrate carboxylate is coplanar with the substrate phenyl ring. One side of the substrate phenyl ring plane is perpendicular to the phenyl ring of Phe 189 with the shortest contact of 3.8 Å. The other side is parallel to Phe 290 withinteraction within 3.5 Å. One of the carboxylate oxygen atoms of the substrate is directly coordinated to Zn 2ϩ ion with a distance of 2.0 Å. The other oxygen atom forms a hydrogen bond with a water molecule. The substrate 2-hydroxy group is hydrogen-bonded to the carboxylate oxygen of Asp 287 , which is a ligand for Zn 2ϩ ion and to two water molecules (W1 and W2), which are further hydrogen-bonded to Arg 229 *. Asp 287 forms a triad together with His 218 and Glu 221 . The 6-hydroxy group is hydrogenbonded to a water molecule, which interacts with Ser 20 and Ala 61 .
The activity of GRDC is inhibited by 2,3-dihydroxybenzaldehyde, which is an analogue of the substrate 2,3-dihydroxybenzoate. The stereo structure of GRDC⅐2,3-dihydroxybenzalde-hyde and the hydrogen-bonding scheme are shown in Figs. 2C and 3C, respectively. The GRDC⅐2,3-dihydroxybenzaldehyde crystal was generated by the soaking method. The side-chain arrangements of the active site residues are quite similar to those observed in GRDC⅐2,6-dihydroxybenzoate. The location of 2,3-dihydroxybenzaldehyde is roughly the same as that of 2,6-dihydroxybenzoate, although the phenyl ring of 2,3-dihydroxybenzaldehyde rotates by about 30°around the axis perpendicular to the ring at the C-1 atom (Fig. 3C). The aldehyde carbonyl group in cis-position with respect to the C-1-C6 bond is coordinated to a Zn 2ϩ ion. This might cause the phenyl ring rotation of 2,3-dihydroxybenzaldehyde compared with 2,6-dihydroxybenzoate. Asn 234 * is hydrogen-bonded to the 2and 3-hydroxy groups. Arg 229 * and Asp 287 interact with the 2-hydroxy group in the medium of two water molecules, W1 and W2.
GRDC catalyzes the decarboxylation of both 2,6-dihydroxybenzoate and 2,3-dihydroxybenzoate but does not act on 2,4-dihydroxybenzoate, 2,5-dihydroxybenzoate, 3,4-dihydroxybenzoate, 3,5-dihydroxybenzoate, 2-hydroxybenzoate, or 3-hydroxybenzoate. This suggests that to catalyze the decarboxylation, it is necessary that the carboxylate and two hydroxy groups occupy three consecutive positions of the benzene ring (13). The substrate specificity shown by GRDC might be explained by the binding mode of 2,6-dihydroxybenzoate and 2,3-dihydroxybenzaldehyde to the active site (Figs. 2 and 3). The substrate carboxylate is coordinated to bind Zn 2ϩ ion with the 3-position of the benzene ring directed toward the entrance of the cavity. Pro 185 , Phe 189 , and Phe 290 limit the substrate to having a planar shape. Asp 287 , Arg 229* , and Asn 234 * form a hydrophilic cluster together with water molecules to interact with the 2-hydroxy group or the 2and 3-hydroxy groups. Ser 20 acts as the recognition site for the 6-hydroxy group with the aid of a water molecule. These results show that the active site can recognize the hydroxy groups at positions 2, 3, and 6 of a benzene ring. When 3,4-or 3,5-dihydroxybenzoate is superimposed onto 2,6-dihydroxybenzoate in GRDC⅐2,6-dihydroxybenzoate, the hydrophilic 4-or 5-hydroxy group makes short contacts of 2-3 Å with the hydrophobic phenyl ring of Phe 23 as a lid for the active site cavity. The lack of activity for decarboxylation of 2,4-, 2,5-, or 3,5-dihydroxybenzoate is attributable at least in part to the unfavorable access between the 4-or 5-hydroxy group and Phe 23 .
Mechanism-The crystallographic study revealed that GRDC is Zn 2ϩdependent decarboxylase, suggesting that the reaction mechanism is different from that of decarboxylase dependent on thiamine diphosphate (49,50), pyridoxal 5Ј-phospahte (51), NAD (52,53), NADP (54), FMN (55), or a pyruvoyl group (56,57). The proposed mechanism of GRDC based on the geometrical consideration is shown in Fig. 4. Upon access of the substrate 2,6-dihydroxybenzoate to the active site cavity, GRDC changes the linker 1 conformation to enclose the substrate in the cavity. One of the substrate carboxyl oxygen atoms interacts with Zn 2ϩ ion, and the substrate 2-hydoxyl group is hydrogen-bonded to the carboxylate of Asp 287 . Asp 287 is liganded to Zn 2ϩ ion and is involved in a triad together with His 218 and Glu 221 . This indicates that Asp 287 plays an important role in the catalysis. The water molecule (W1) interacts with positively charged His 218 and Arg 229 * and the substrate 2-hydroxy group. The water molecule (W2), which interacts with Arg 229 *, bridges Asp 287 and the substrate 2-hydroxy group by hydrogen bonds. Zn 2ϩ ion, the triad (Asp 287 , His 218 , and Glu 221 ), W1, and W2 form the specific recognition site for the carboxylate and the adjacent 2-hydroxy group of the substrate.
The negative charge on the substrate carboxylate group migrates via Zn 2ϩ ion to the carboxylate of Asp 287 , which is polarized by the His 218 and Glu 221 pair (Fig. 4A). This increases the basicity of the carboxylate of Asp 287 , which then acts as a catalytic base to abstract a proton from the substrate 2-hydroxy group. The negative charge generated on the substrate O-2 flows into the aromatic ring and accumulates on the C-1 atom to form the substrate carbanion (Fig. 4B). The next step is the addition of a proton to the C-1 atom. The water molecule W1 interacts with positively charged Agr 229 * and His 218 , thereby  NOVEMBER 10, 2006 • VOLUME 281 • NUMBER 45 reducing its pK a value. The pK a value would be further reduced after the 2-hydoxy proton abstraction by Asp 287 ; the protonation at Asp 287 weakens the hydrogen bond between Asp 287 and His 218 and decreases the electron density at N-1 of His 218 , thereby reducing the interaction between W1 and His 218 . W1, which is at a distance of 3.76 Å from the C-1 atom in Fig. 4A, approaches the C-1 atom within the distance of 3.4 Å on the formation of the carbanion followed by the adoption of sp 3 hybridization by the C-1 atom. W1 is thus reasonably assumed to be the conjugate acid to shuttle its proton to the C-1 atom (Fig. 4C). Throughout this process, the repulsive interaction between the substrate carboxylate and the C-1 carbanion is suppressed by the coordination of the carboxylate to Zn 2ϩ ion. The substrate C-2 carbonyl group then abstracts a proton from the carboxylate of Asp 287 in concert with the decarboxylation of the substrate to yield carbon dioxide and resorcinol (Fig. 4, C  and D).

X-ray Structure of Zinc-dependent GRDC
The enzyme also catalyzes the decarboxylation of 2,3-dihydroxybenzoate. 2,3-Dihydroxybenzoate is modeled into the active site of GRDC⅐2,6-dihydroxybenzoate on the assumption that 2-hydroxybenzoate moieties of both substrates occupy the same position. The 3-hydroxy group of 2,3-dihydroxybenzoate is located at a position favorable for forming a hydrogen bond with Asn 234 * (Fig. 2C). Therefore, 2,3-dihydroxybenzoate is a good substrate for GRDC. After submission of this paper, we learned that the native structure of mono-metal-dependent decarboxylase (␣-amino-␤-carboxymuconate-⑀-semialdehyde decarboxylase) belonging to the TIM-barrel metallo-dependent hydrolase superfamily has been published (58).
In conclusion, we have determined the three-dimentional structures of the novel decarboxylase, GRDC, in the native form, complexed with substrate 2,6-dihydroxybenzoate and complexed with 2,3-dihydroxybenzaldehyde. This is the first structure of a Zn 2ϩ -dependent decarboxylase complex with the substrate or substrate analog belonging to the hydrolase superfamily. The arrangements of the active site residues, the substrate, the substrate analogue, and the interactions among them have been elucidated to provide a putative mechanism for catalysis of the enzyme. Asp 287 and Zn 2ϩ ion are identified as the catalyst.