Structure of the Cyanuric Acid Hydrolase TrzD Reveals Product Exit Channel

Cyanuric acid hydrolases are of industrial importance because of their use in aquatic recreational facilities to remove cyanuric acid, a stabilizer for the chlorine. Degradation of excess cyanuric acid is necessary to maintain chlorine disinfection in the waters. Cyanuric acid hydrolase opens the cyanuric acid ring hydrolytically and subsequent decarboxylation produces carbon dioxide and biuret. In the present study, we report the X-ray structure of TrzD, a cyanuric acid hydrolase from Acidovorax citrulli. The crystal structure at 2.19 Å resolution shows a large displacement of the catalytic lysine (Lys163) in domain 2 away from the active site core, whereas the two other active site lysines from the two other domains are not able to move. The lysine displacement is proposed here to open up a channel for product release. Consistent with that, the structure also showed two molecules of the co-product, carbon dioxide, one in the active site and another trapped in the proposed exit channel. Previous data indicated that the domain 2 lysine residue plays a role in activating an adjacent serine residue carrying out nucleophilic attack, opening the cyanuric acid ring, and the mobile lysine guides products through the exit channel.

Scientific RepoRts | 7:45277 | DOI: 10.1038/srep45277 neutral pH 6 . Given the chemical instability of carboxybiuret, it has not been possible to determine if there is some enzyme assistance in the decarboxylation reaction, and this is one of the issues addressed in the present study.
X-ray structures for cyanuric acid hydrolases are crucial for using the enzymes for treating disinfection waters 1 . The enzyme TrzD, a cyanuric acid hydrolase from Acidovorax citrulli 11 shows 58% and 50% sequence identity to the previously-studied enzymes from Pseudomonas sp. ADP 12 and Azorhizobium calindulans 9 , respectively, and TrzD is reported to have a k cat that is an order of magnitude greater than other known cyanuric acid hydrolases 11 . Given the importance of using a high k cat enzyme in commercial applications, and for obtaining insights into the outstanding mechanistic questions, the present study focused on determining the X-ray structure of TrzD. In the course of the study, we observed an unusual orientation of the second-domain lysine that interacts with the proposed catalytic second-domain serine. This observation simultaneously provided support that the second domain serine serves as the active site nucleophile and it gave evidence for a product exit channel. Moreover, the presence of trapped carbon dioxide in the enzyme is consistent with the view that the enzyme assists in the decarboxylation of carboxybiuret, and the cell does not rely on a spontaneous decarboxylation to produce the next intermediate, biuret, in the cyanuric acid biodegradation pathway (Fig. 1).

Methods
Protein Expression and Purification. The full-length trzD gene, cloned in pET28b( + )vector 6 , was expressed in Escherichia coli BL21 (DE3) that was grown in LB medium (10 g Bacto tryptone, 5 g yeast extract, 10 g NaCl per liter) supplemented with 50 ug/ml kanamycin. Cells were grown at 37 °C and induced with 0.5 mM of isopropyl β -D-1-thiogalactopyranoside (IPTG) for 30 hours at 20 °C. After collection by centrifugation (6400 × g, 10 min, 4 °C), the cells were resuspended in buffer A (50 mM Tris-HCl pH 7.2, 200 mM NaCl). Cells were lysed by three passes at 1000 p.s.i. through a French press, and the cell lysate was centrifuged at 48,000 × g for 45 min at 4 °C. After centrifugation, the cell lysate was applied onto a column packed with a Ni-NTA resin (Thermo Scientific) that had been equilibrated with the binding buffer. After washing, the bound protein was eluted from the column with a gradient using elution buffer B (50 mM Tris-HCl pH 7.2, 200 mM NaCl, 200 mM imidazole).
For further purification, gel filtration chromatography was performed using a HiLoad ™ Superdex-200 16/600 gel-filtration column (GE Healthcare) equilibrated with Buffer A. The enzyme was concentrated to 12 mg/ml and stored at 4 °C until used.
Crystallization. The protein was crystallized with its natural substrate cyanuric acid by hanging drop, vapor-diffusion experiment in which equal volume of protein (12 mg/ml concentration) and reservoir solution were mixed and allowed to equilibrate against the reservoir at 20 °C. The final concentration of cyanuric acid was 10 mM in the crystallization drop. The initial crystallization hit contained 0.05 M ammonium sulfate, 0.05 M BIS-TRIS pH 6.5, and 30% pentaerythritol ethoxylate (15/4 EO/OH). Needle-shaped crystals appeared in 48 hours. Two week old crystals were harvested and flash-cooled in liquid nitrogen without additional cryoprotectant, since 30% pentaerythritol ethoxylate (15/4 EO/OH) serves as good cryoprotectant. A single crystal was used for data collection.
Data Collection and Processing. The flash-cooled crystal was diffracted at beamline 23-ID-B in the Advanced Photon Source (APS) at 1.03322 Å wavelength with 0.5 degree oscillation per frame. A MAR CCD300 detector was used to collect 270 frames at a distance of 250 mm from the crystal. All data were processed and reduced using the HKL2000 package 13 . Crystals grown with cyanuric acid belonged to I222 space group with unit cell dimensions of 67.7, 90.8, and 120.3 Å and diffracted to 2.19 Å. Crystal showed an average mosaicity of 0.33 degree. The asymmetric unit contained one monomer with a Matthews coefficient of 2.32 with 47% solvent content.
Structure Determination and Refinement. Molecular replacement was performed using the software MOLREP 14 with the monomer of AtzD (PDB Code 4BVQ) 10 as the search model. Preliminary phases derived from the initial atomic models showed strong electron densities for ligands in the active site for crystals grown with cyanuric acid. Iterative cycles of model building and refinement were conducted using COOT 15 and PHENIX 16 alternatively against a dataset at 2.19 Å resolution. Refinement of the model against X-ray data was carried out until high quality electron density maps and satisfactory model statistics were achieved. MolProbity 17,18 and PROCHECK 19 were used for structure quality analysis. For the final structures, the Ramachandran plot showed that 98% of the residues are in the most favored region and the other 2% are in the allowed region. A summary of the data collection and model refinement statistics is shown in Table 1. a pseudo three-fold symmetry. They show significant divergence, most share ~40-60% sequence identity in pairwise alignment with other members of the cyanuric acid hydrolase/barbiturase protein family. Cyanuric acid hydrolase proteins differ from each other largely in their respective kinetic parameters and thermal stability 6,20 . The data were phased using coordinates from the X-ray structure of AtzD (cyanuric acid hydrolase from Pseudomonas sp. ADP) (4BVQ). The crystallographic asymmetric unit contained one TrzD molecule. The product-bound structure was refined to a resolution of 2.19 Å, with R work and R free values of 0.15 and 0.20, respectively, yielding high quality electron density maps. We could model the N-terminal His-tag. The crystallographic data collection and refinement statistics are listed in Table 1, and the coordinates and structure factors have been deposited as Protein Data Bank entry 5T13. Structural comparison of TrzD with available cyanuric acid hydrolase structures shows 58% sequence identity and 1.0 Å rmsd (on 364 carbon alpha) with AtzD and 50% sequence identity and 1.4 Å rmsd (on 351 carbon alpha) with ACAH (a cyanuric acid hydrolase from Azorhizobium calindulans). The next nearest structural homolog is chorismate mutase (4BPS) with 11% sequence identity and 4.0 Å rmsd (using the DALI server) 21 . An examination of the active site revealed, unexpectedly, the active site lysine residue in domain 2 (Lys163), that in other structures paired with the proposed nucleophilic serine (Fig. 1), exhibited a significant displacement out of the active site, and opening a clear channel from the buried active site to the protein surface. Moreover, the difference Fourier map (F o − F c ) clearly showed electron density corresponding to two CO 2 molecules and Lys163 within the TrzD structure (Figs S1 and S2). As expected, since CO 2 is a co-product, one is located in the active site, while the second one was observed at the interface of domain 2 and 3, proposed here as the exit channel for product release.
TrzD is a 370 amino acid, ~39.5 kDa protein that exhibits a complex α /β structure characteristic of the cyanuric acid hydrolase family enzyme. The active site is situated inside a core composed of 4 β -strands from each of the three-isostructural domains. Surface exposed elements are helices and loops. TrzD has a short β -hairpin and 9 significant helix-helix interactions. Domain 1 is composed of 107 (1 to 107), domain 2 is composed of 137 (113 to 249) and domain 3 is composed of 115 (255 to 369) residues. Each domain starts and ends with two antiparallel β -strands. Domain 2 has a 15 residue long extra helix, which is not present in the other two domains, making it slightly bigger. Residues important for catalysis, i.e., Ser, Lys and Arg residues from each domain, showed similar spatial disposition, when superimposed. The three structurally homologous domains form a pseudo-three-fold internal symmetry in the TrzD monomer ( Fig. 2A). The latter is likely selected by nature to accommodate the

Conserved Metal Binding Site. A conserved metal-binding loop sequence GGxEHQGPxGG is identified
throughout the cyanuric acid hydrolase/barbiturase protein family. Ser/Ala/Gly residues can substitute in the first X and the second X can be a Ala/Asp/Pro/Ser residue (Fig. S5). This conserved metal binding loop is situated ~13 Å away from the active site and has not been implicated in the catalytic mechanism. The present structure has an Mg +2 bound and makes 6-contacts with neighboring residues and one with a solvent molecule (2.8 Å). Mg +2 interacts with the backbone carbonyl oxygen of Ser352 (2.5 Å), Gln355 (2.7 Å), Pro357 (2.6 Å) and Gly360 (2.9 Å) and the side chain oxygens of Glu303 (2.8 Å) and Ser352 (2.9 Å). Glu303 is the only residue to interact with the metal ion that is not in the conserved loop (Fig. 2B). The metal binding site formed a uniform alcove in domain 3 and does not interact directly with domain 2.
Dynamic Nature of Second Domain Lysine163. An unusual displacement of the 2 nd domain catalytic lysine residue, away from the active site core and with strong electron density, was observed (Fig. 3). The movement does not affect the unique three -domain organization that coordinates and activates the three-fold symmetric substrate cyanuric acid in cyanuric acid hydrolases. The movement of Glu235, Leu238 and Glu242 near the domain 2-Lys163 provides more space for opening to the protein surface, but without extensive main-chain movement. By contrast, the active site lysine residues in domain 1 and 3 are strongly anchored and remain localized in the catalytic center (Fig. 4A,B). The average B-factor of the 2 nd domain's Lys, Ser and Arg are higher than those of domain 1 and 3. In domain 1, Lys39 is hydrogen bonded to Gly83, Gly84, & Val348 main chain carbonyl oxygen atoms, and with the Ser82 main-chain carbonyl oxygen and side-chain oxygen. Similarly, Lys301 of domain 3 is hydrogen bonded with the Ser349 main chain carbonyl oxygen and side chain oxygen and two consecutive glycine main chain oxygen atoms and the Thr232 side chain carbonyl oxygen. This is a unique conserved signature interaction of active site lysines observed in all other available CAH structures 9,10 . In contrast, Active Site CO 2 Binding Site. Density analysis clearly revealed two molecules of the co-product, carbon dioxide, with one in the active site and the other in the exit channel. The active site CO 2 was observed to be within a core made by tight clustering of three serines, three arginines and two lysines. The CO 2 molecule is stabilized by H-bonds with the side chain oxygen of Ser82, Ser233 and Ser349 (~2.9 Å), and the backbone amide nitrogen of Gly350 (3.3 Å), backbone carbonyl oxygen of Gly83 and also with a side chain nitrogen of Arg330 (Fig. 5A). A solvent molecule is also visible in the vicinity of carbon dioxide at a distance of 2.9 Å.
Domain Interface CO 2 Binding Site. Unexpectedly, an unambiguous density corresponding to a second bound CO 2 was observed in the domain interface, sandwiched between domain 1 and domain 2 near the surface. The oxygen atoms of the CO 2 molecule were stabilized by two paired H-bonds with the side chains of Arg51 and Glu237 from domain 1 and domain 2, respectively, with additional stabilization by the hydrophobic residue Ile236 backbone amide N and Glu237 backbone amide N (Fig. 5B). The CO 2 proximal oxygen is also H-bonded to a solvent molecule at 2.9 Å distance. On the basis of only structural observations, the domain interface-bound CO 2 would appear to be less stable than the active site CO 2 , suggesting it may be in a position penultimate to release from the proposed exit channel.
Entry and Exit Channels. The surface of the monomer has two cavities that could provide access to the active site for substrate entry and/or product release. One is shorter, around 10 Å from the surface, and is formed by Cys45, Tyr188, Met191, Arg195, and Thr326. Cys45, Tyr188, and Thr326 are proposed to be responsible for controlling substrate entry (Fig. 5C,D). Another clear channel was observed between the domain 1 and domain 2 interface that is longer (~15 Å) than the entry channel and formed by the displacement of active site Lys163. The terminal N-atom of this second domain Lys is hydrogen bonded to the adjacent Met81 and Ser233, in known cyanuric acid hydrolase structures 9,10 , and therefore does not open up into what we proposed here to be the exit channel. Arg51, parallel to Lys163 also helps to maintain the exit channel while Ile236, Glu237 and Gly52 form the outer surface of the exit channel. A CO 2 molecule is bound to this channel, supporting the exit channel concept.

Discussion
Previously reported cyanuric acid hydrolase structures have three pairs of Ser-Lys-Arg residues clustered in a closed active site pocket of around 155 Å 3 23 . Adjacent Ser-Lys residues have been proposed to serve as potential catalytic dyads 10 . The TrzD structure described here is the first study identifying the flexibility of the domain 2 lysine and observing the corresponding residue not in close proximity to its cognate serine. In the TrzD structure uniquely, Lys163, which is conserved in all cyanuric acid hydrolases, was re-oriented toward the surface, opening up an exposed channel between the domain 1 and 2 interface. Here we propose this to be the product exit channel. In all other cyanuric acid hydrolase structures, the terminal nitrogen atom of the second domain active site lysine was hydrogen bonded to the side-chain oxygen of its cognate serine, serving to close off the active site 9 . The displacement of Lys163 toward the surface, observed here, does not affect the conserved three-fold symmetry of the global structure. This movement of Lys163 is made possible by the relatively low contact density of this key active site residue. We note that this conformational change seems physically possible in other cyanuric acid hydrolases, for instance ACAH, where the corresponding lysine residue has room to move toward the surface, as is observed here in the structure of TrzD. It is unusual for an active site residue that participates in the reaction chemistry to have such freedom of movement 24,25 . As shown in Fig. S6, superimposition of TrzD with the ACAH structure shows that the Lys156 of ACAH could move similarly as Lys163 of TrzD. On the other hand, the other two active site lysines are anchored tightly and remain localized in the catalytic center forming four conserved hydrogen bonds each as shown in Fig. 4A,B. The Ser-Gly-Gly consensus sequence in each of the two domains, namely 1 and 3 has formed an identical geometry and interaction with their cognate lysine residues. The refined structure also revealed, unexpectedly and unambiguously, one CO 2 molecule in the active site. The ligand is bound at the center of the β -barrel in the catalytic site and has occupancies and B-factors fully compatible with the mobility of the surrounding atoms (for CO 2 and protein molecule are 42 and 40 Å 2 respectively, Fig. S4).
The new conformation of the Lys163 opens up a new channel connecting the active site with the protein surface. This proposed exit channel is overall hydrophobic in nature, and the electronic density maps reveal unam biguously a second bound CO 2 molecule that sits in the tunnel. In the view shown in Fig. S7, CO 2 is visible in the exit tunnel only 3 Å from the surface. Ile236 and Glu237 side-chain moved significantly in comparison to ACAH structure to form the channel for product exit, though their backbone movement is negligible. Phe49 also has some role to the exit channel opening in TrzD during product release, since the bulky side chain has moved.
A common motif in other protein-CO 2 binding sites is the presence of charged basic residues and a solvent molecule that is hydrogen bonded with CO 2 oxygen atoms 26,27 . In TrzD, a side chain NH 2 of Arg330 forms a hydrogen bond with the active site CO 2 and Arg51 is hydrogen bonded to the exit channel carbon dioxide. The Arg51 side chain NH 2 forms a water-mediated hydrogen bond with the other oxygen of the active site carbon dioxide (Fig. 5A,B). In the case of the active site CO 2 , ligating residues belong to the core beta-sheets, only Arg330 is an exception and comes from a 3 rd domain helix. In case of exit channel CO 2 , ligating residues belong to a loop arising at the end of the 2 nd domain similar to what has been observed in other CO 2 -binding proteins 26,27 . Curiously enough, only Arg51 derived from the 1 st domain helix. Comparison with available CO 2 bound protein structures revealed that the protein-based CO 2 bonding is guided by acid/base interaction rather than hydrophobic/philic interactions 26,28 . Preferences of secondary structural elements more amenable to CO 2 binding are loops and beta-sheets as opposed to helices.
The displacement of a catalytic lysine residue of domain 2, away from the active site core, suggests its dynamic nature during the reaction cycle. First, after cyanuric ring opening and hydrolysis of the enzyme-ester intermediate, carboxybiuret forms and decarboxylation is enzyme-assisted, producing biuret and carbon dioxide. The displacement of the domain 2 lysine then opens up a channel that allows product release. In silico, carbon dioxide remained bound, thus revealing the exit channel. By this model, domain 2 lysine's role in catalysis is not limited to activating catalytic serine to open the ring but also guides products towards the exit channel.