Structure of Ddi2, a highly inducible detoxifying metalloenzyme from Saccharomyces cerevisiae

Cyanamide (H2N-CN) is used to break bud dormancy in woody plants and to deter alcohol use in humans. The biological effects of cyanamide in both these cases require the enzyme catalase. We previously demonstrated that Saccharomyces cerevisiae exposed to cyanamide resulted in strong induction of DDI2 gene expression. Ddi2 enzymatically hydrates cyanamide to urea and belongs to the family of HD-domain metalloenzymes (named after conserved active-site metal-binding His and Asp residues). Here, we report the X-ray structure of yeast Ddi2 to 2.6 Å resolution, revealing that Ddi2 is a dimeric zinc metalloenzyme. We also confirm that Ddi2 shares structural similarity with other known HD-domain proteins. HD residues His-55, His-88, and Asp-89 coordinate the active-site zinc, and the fourth zinc ligand is a water/hydroxide molecule. Other HD domain enzymes have a second aspartate metal ligand, but in Ddi2 this residue (Thr-157) does not interact with the zinc ion. Several Ddi2 active-site point mutations exhibited reduced catalytic activity. We kinetically and structurally characterized H137N and T157V mutants of Ddi2. A cyanamide soak of the Ddi2-T157V enzyme revealed cyanamide bound directly to the Zn2+ ion, having displaced the zinc-bound water molecule. The mode of cyanamide binding to Ddi2 resembles cyanamide binding to the active-site zinc of carbonic anhydrase, a known cyanamide hydratase. Finally, we observed that the sensitivity of ddi2Δ ddi3Δ to cyanamide was not rescued by plasmids harboring ddi2-H137N or ddi2-TI57V variants, demonstrating that yeast cells require a functioning cyanamide hydratase to overcome cyanamide-induced growth defects.

Cyanamide (H 2 N-CN) can be regarded as the amide of cyanic acid (HO-CN) or equivalently as the nitrile of carbamic acid (H 2 N-COOH). Cyanamide is bifunctional, being capable of reacting as a nucleophile at the amino group or as an electro-phile at the cyano group. The hydration of cyanamide (H 2 N-CN) to form urea (H 2 N-CO-NH 2 ) is similar to reactions carried out by nitrile hydratase (NHase) enzymes, which catalyze the hydration of nitriles (R-CN) to the corresponding amides (R-CO-NH 2 ), although it is worth noting that nitrile hydratases do not convert cyanamide to urea (1). However, the soil fungus Myrothecium verrucaria and the budding yeast Saccharomyces cerevisiae have both been shown to contain a cyanamide hydratase (CAH, EC 4.2.1.69) 3 enzyme that converts cyanamide to urea (2,3), and CAH belongs to the diverse HD-domain family of metalloproteins (3,4).
Cyanamide is used as both a fertilizer and herbicide in agriculture and is toxic to plant cells because it causes oxidative stress via inhibition of catalase (5). Furthermore breakdown of cyanamide into nitroxyl (HNO) by catalase inhibits aldehyde dehydrogenases (5). Plants apparently do not contain a homologous CAH gene, at least according to BLAST searches using the yeast Ddi2 protein sequence (3,6). Treatment of perennial woody plants with cyanamide is known to break bud dormancy and has revealed the plant metabolic pathways that are activated in response to cyanamide treatment (7)(8)(9). These pathways involve activation of the jasmonate and cytokinin pathways via cyanide (cyanamide is first broken down into cyanide (HCN) and HNO by the action of catalase), plus activation of several cyanide detoxification and oxidative stress response enzymes. Interestingly, cyanamide is not only a substrate of catalase, it also functions as a competitive inhibitor (10 -12). Oxidative stress in woody plants exposed to cyanamide presumably is initiated by cyanamide-inhibiting catalase and subsequent hydrogen peroxide accumulation (5,(7)(8)(9)(10).
Cyanamide has been used as an alcohol deterrent, because in humans it is similarly metabolized to nitroxyl and HCN by catalase; nitroxyl then inhibits aldehyde dehydrogenase and thus interferes with the oxidation of acetaldehyde to acetate (10 -12). Cyanamide is also known to be both a poor substrate and an inhibitor of carbonic anhydrase II (13,14). Carbonic anhydrases are zinc metalloenzymes that hydrate carbon dioxide to carbonate and thereby help maintain physiological pH via carbonate/bicarbonate buffering. The linear arrangement of N-CN in cyanamide approximately resembles that in CO 2 . Crystallographic studies have revealed, however, that cyana-mide and CO 2 have different binding modes at the carbonic anhydrase active site (13).
Biodegradation of cyanamide in microorganisms was first credited to the soil fungus M. verrucaria that specifically catalyzed the formation of urea from cyanamide but did not catalyze the conversion of other nitriles to their respective amides (2). M. verrucaria cyanamide hydratase was found to contain zinc, and the functional enzyme was assessed to be a hexamer of identical subunits. The deduced K m for cyanamide was reported to be 27 mM for M. verrucaria CAH (2). We recently demonstrated that S. cerevisiae Ddi2, a homolog of M. verrucaria CAH (37% pairwise sequence identity), is also a bona fide CAH, enzymatically converting cyanamide to urea with kinetic parameters comparable with those of the M. verrucaria enzyme (3). Closely related orthologs of cyanamide hydratase identified by BLAST are all from fungi and appear to be limited to the Ascomycota and Basidiomycota.
Because cyanamide is synthesized by limited species of plants presumably as a chemical defense against fungi and weeds (15)(16)(17), it is not surprising that some microorganisms, especially those living in soil or associated with plants, have developed a cyanamide biodegradation pathway to utilize the nitrogen in cyanamide and to protect against oxidative stress related to cyanamide reaction with catalase.
The yeast Ddi2 CAH enzyme was originally brought to our attention because microarray studies demonstrate that S. cerevisiae DDI2 expression is highly induced by the DNA alkylating agent methyl methanesulfonate (MMS) (18,19). Recently, we demonstrated that Ddi2 is also induced upon adding dimethyl sulfate or cyanamide to the growth medium (3).
The S. cerevisiae Ddi2 protein is encoded by two identical duplicated genes DDI2 and DDI3 (two copies of DDI2 orthologs were also found in Aspergillus nidulans and Aspergillus oryzae) and belong to the HD-domain superfamily of enzymes, reflecting the presence of key conserved histidine and aspartate residues in the primary sequence that bind metal ions involved in enzyme catalysis (3,4). Henceforth, we will refer to Ddi2/Ddi3 as Ddi2. Site-specific mutations of the conserved HD signature residues in Ddi2 resulted in the loss of CAH enzymatic activity, indicating that the core Ddi2 HD residues are required for enzyme stability and/or catalysis (3). HD domain proteins have a mostly helical structure and are either divalent metal-dependent phosphohydrolases (20 -25) or oxidases (26,27).
Hence Ddi2 and its fungal CAH orthologs represent a novel and distinct subgroup of HD domain proteins that display novel catalytic activity (2)(3)(4). To better understand the effects of cyanamide on S. cerevisiae physiology and the mechanism of cyanamide hydration, we here describe the crystallographic determination of the S. cerevisiae Ddi2 structure, identify conserved amino acid residues essential for catalysis, and deduce the mode of cyanamide substrate binding to the active-site Zn 2ϩ atom in a Ddi2-T157V mutant enzyme. Furthermore, we demonstrate that only enzymatically active Ddi2 protein is capable of rescuing yeast growth defects in the presence of cyanamide.

Ddi2 is a dimeric zinc metalloenzyme
The Ddi2 structure solution and refinement are described under "Experimental procedures." The structure was refined using diffraction data to 2.6 Å resolution, and the model consists of 9 identical subunits of 225 residues (chains A-I), with each subunit being present in a symmetrical dimer (Fig. 1, Table  1, and Fig. S1). Chains A-G had significantly lower B factors than chains H and I, which exhibited overall weak electron density and high B-factors. The refined Ddi2 model showing zinc coordination and the deduced active site is depicted in Fig. 1. Secondary structure assignments for the Ddi2 polypeptide, sequence conservation in fungal homologs, and the location of active-site residues are summarized in Fig. 2. In each Ddi2 monomer, the zinc ion is coordinated by HD residues His-55, His-88 and Asp-89, and a water molecule or hydroxide ion, forming approximate tetrahedral geometry and having typical zinc-ligand distances of 2.0 -2.2 Å (28). Structural similarities between Ddi2 and the HD-domain phosphohydrolase YpgQ (25), including core conserved elements in the HD fold, are The zinc ion is shown as a gray sphere. Rainbow colors are shown for the secondary structure: the N terminus is blue, and the C terminus is red. Side chains for residues His-55, His-88, Asp-89, His-137, Gln-138, and Thr-157 near the zinc-binding site are depicted as stick models with nitrogen in blue and oxygen in red. B, drawn as for A but showing the observed dimer. The panels were drawn with PyMOL (42).

Structure of Ddi2: a HD-domain cyanamide hydratase
summarized in Fig. 3 and Table S2 (see also below). Conserved Ddi2 residues Phe-104, His-137, Gln-138, Thr-157, Asn-161, and Trp-186 form a pocket that is likely to be the active site (Figs. [1][2][3][4]. Four water molecules (Wat-1 to Wat-4) were observed near the zinc site for chains A-G, and this was confirmed by noncrystallographic symmetry averaging of ͉F o ͉ Ϫ ͉F c ͉ omit electron density maps for the seven well-defined subunits in the structure (Fig. 4). The Wat-1 molecule is bound directly to the zinc ion in the absence of substrate; the remaining water molecules form hydrogen bonds with the side chains of Asp-89 and His-137 (Wat-2), Gln-138 and Asn-161 (Wat-3), and Thr-157 (Wat-4) (Fig. 4). Wat-4 also hydrogen-bonds to the backbone carbonyl groups of residues 153 and 154 in helix ␣8, one helical turn prior to residue Thr-157.

Ddi2 shares overall structural similarity with HD-domain family proteins
Structural overlays of the Ddi2 polypeptide onto other known HD domain proteins ( Fig. 3 and Fig. S2 and Table S2) identified residues 51-161 of Ddi2 as the core HD motif, including five core helices (here labeled ␣A-␣E) of the HD domain region and in some cases a sixth helix at the N terminus preceding helix ␣A (depicted as ␣N) ( Fig. 3 and Table S2). Briefly the conserved HD-domain helices in Ddi2 are as follows: ␣N corresponds to ␣2, ␣A corresponds to ␣3, ␣B corresponds to ␣4, ␣C corresponds to ␣6, ␣D corresponds to ␣7, and ␣E corresponds to ␣8 ( Fig. 3 and Table S2). Most structurally characterized HD domain superfamily members overlaid reasonably well with the core HD-domain fold of Ddi2 (20 -27). RMS differences on backbone atoms ranged between 1.6 and 2.4 Å for 79 equivalent residues, including conserved helix ␣N (Table  S2). Structural similarity searches with FATCAT (29) and BLAST (6) searches of the PDB confirmed the similarity of Ddi2 to known HD domains.
A few HD domain proteins stood out as having the most informative structural similarity to Ddi2. They are YpgQ (PDB code 5DQW), a bacterial nucleotide phosphohydrolase (25), PA4781 (PDB code 4R8Z), a cyclic di-GMP phosphodiesterase (24), and phosphodiesterase 2A (Pde2A) (PDB code 4D08) (30) (Fig. 3 and Fig. S2). For instance, there is a conserved salt-bridge interaction between helices ␣A (Arg-58 in Ddi2) and ␣E (Asp-160 in Ddi2) that is observed in PA4781 (PDB code 4R8Z) and YpgQ (PDB code 5DQW) (24, 25) ( Fig. 3 and Fig. S2). The side-chain carboxylate of Asp-160 of Ddi2 and the equivalent residue in the above-mentioned HD-domain proteins form a hydrogen bond with the ND1 side-chain nitrogen of metalcoordinating residue His-55 and likely helps orient the side chain of His-55 to coordinate its metal ion.
Ddi2 is somewhat unusual because it contains three C-terminal helices (␣10, 4, and ␣11) that form a partial lid over the active-site entrance; these helices are not found in most other HD-domain family members (Fig. 3). However, two of the HD domain proteins noted above (YpgQ and Pde2A) (25,30) do have two C-terminal helices arranged in a hairpin fashion that resemble helices ␣10 and ␣11 of Ddi2. In particular, the C-terminal region of Pde2A aligns reasonably well with the Ddi2 polypeptide in BLAST searches.
Of particular note, a second aspartate ligand found in most HD-domain proteins on helix ␣E is missing in Ddi2, being substituted by threonine (Thr-157), and Thr-157 does not coordinate the zinc metal ion, although it is nearby. The placement of a threonine residue rather than an aspartate at this position in the Ddi2 sequence is very likely important for the Zn 2ϩ ion specificity of the enzyme and permits the binding of a water/ hydroxide molecule at the fourth ligand site (Fig. 4).

Kinetic and substrate-binding studies of Ddi2 active-site mutants
We determined the pH optimum for the WT Ddi2 reaction to be 7.5 using our standard enzyme assay operating under sat- Rotamer outliers (%) 1.9 a The details of data collection and processing are provided under "Experimental procedures." b The values in parentheses correspond to the highest resolution shell.
͉F calc (hkl)͉ are the observed and calculated amplitudes, respectively, for the structure factor F(hkl). e R free is the equivalent of R work for 5% of the reflections (randomly selected) that were not used in structure refinement.

Structure of Ddi2: a HD-domain cyanamide hydratase
urating substrate conditions (see "Experimental procedures") and then redetermined the kinetic parameters for WT recombinant Ddi2 (using the cleaved GST-fusion protein) (see Table  3). Based on the locations of conserved residues near the zincbinding site, we also qualitatively characterized a series of active-site point mutations (Fig. 5) using a cyanamide depletion assay (31) with mutant Ddi2-His 6 proteins purified to homogeneity (Table 3 and Fig. S3). Incidentally we also determined the kinetic parameters for Ddi2-His 6 , and they were indistinguishable from Ddi2 purified from the cleaved GST-fusion protein (not shown). All His-tagged mutants assayed (H137N, Q138E, Q138A, T157V, and N161A) displayed a significantly reduced ability to convert cyanamide to urea (Fig. 5). We then kinetically characterized and crystallized the H137N and T157V mutant proteins (Tables 2 and 3 and Fig. 5). Because the C-terminally His-tagged Ddi2 proteins did not crystallize, and we wanted to also structurally characterize these two mutants in the hope of trapping bound substrate, we purified recombinant Ddi2 mutants from the GST-fusion system. We chose to mutate residue His-137 because it is highly conserved in Ddi2 (Fig. 2), and modeling suggested the His-137 side chain may hydrogen-bond to the cyanamide substrate. In addition, we chose mutation of Thr-157 because it is unusual in HD-domain proteins, normally being an aspartate, but also because it may perturb the zinc site and catalytic rate because of its spatial proximity to the zinc atom. The H137N mutant protein exhibited very low turnover   Fig. 2. B, superposition of the HD motif of Ddi2 (blue) onto the HD motif of YpgQ (orange, PDB code 5DQV), highlighting the conserved five conserved ␣-helices: ␣A-␣E (in this case a sixth N-terminal helix ␣-N (magenta) is also conserved). The zinc atom of Ddi2 is shown as a blue sphere, and the bound water is shown as a red sphere. C, as in B, but looking from the top at the Ddi2 active site. Conserved active-site residues (including Arg-58 and Asp-160) are shown as stick models colored by atom type for each model. Residue equivalences and root-mean-square deviation values for the superposition are given in Table S2 in the supplementary information. The panels were drawn with PyMOL (42).

Structure of Ddi2: a HD-domain cyanamide hydratase
levels of substrate, and we resorted to using 100 g of mutant enzyme to measure its kinetic parameters (typically 1 g of Ddi2 is used in the assay). Ddi2-H137N exhibited a near WT K m value (33 Ϯ 60 mM) but a significantly reduced value of k cat (0.0040 Ϯ 0.0029 s Ϫ1 ). Ddi2-T157V exhibited a K m value of 25 Ϯ 6 mM, comparable with the WT enzyme but a more modestly compromised turnover number, k cat ϭ 1.3 Ϯ 0.1 s Ϫ1 (Table 3). We conclude from these measurements that the H137N mutant is severely compromised in catalysis, whereas the T157V mutant binds substrate as efficiently as the WT enzyme and is moderately catalytically impaired.

Binding of cyanamide at the Ddi2 active site
To observe substrate binding in Ddi2 crystals, we utilized the T157V and H137N mutants, because extensive soaking and cocrystallization experiments of the WT Ddi2 protein with cyan-amide or cyanide yielded negative results. Crystals of Ddi2-H137N and Ddi2-T157V were briefly (30 s) soaked in cryobuffer containing 0.2-0.3 M cyanamide before flash-freezing in liquid nitrogen for X-ray diffraction experiments. Diffraction data to 3.0 Å resolution were collected for the Ddi2-H137N mutant crystals; the Ddi2-T157V mutant crystals diffracted to 2.90 Å resolution ( Table 2). The structures were solved by rigidbody and maximum-likelihood refinement (see "Experimental procedures"), because the unit cell dimensions and crystal morphology were isomorphous to the WT Ddi2 crystals ( Table 2). Analysis of the H137N mutant structure revealed that the Asn-137 side chain rotated away from the position of His-137 in the WT enzyme and made hydrogen bonds to the side chain of nearby residue Thr-157 (Fig. S4). In addition, there was no electron density observed for cyanamide at the active site.
In contrast, the cyanamide soak of the T157V mutant revealed electron density consistent with cyanamide binding directly to the zinc for seven of the nine independent active sites in the crystal structure. To objectively evaluate cyanamide binding, we averaged the ͉F o ͉ Ϫ ͉F c ͉ difference electron density for subunits A, B, C, E, F, and G onto subunit D (Fig. 4) for the cyanamide soak data. Real-space refinement of a cyanamide molecule into the 7-fold averaged difference electron density demonstrated that cyanamide bound directly to the zinc atom and hydrogen-bonded to the side chains of Asp-89 and His-137, thereby displacing the bound water molecules Wat-1 and Wat-2 observed in the apo-enzyme structure (Fig. 4). This observation seems to rule out a catalytic role for the zinc-bound water molecule observed in the apo-enzyme structure, because it otherwise would have been a candidate for the attacking nucleophile in the hydration reaction. Binding of cyanamide to the active site of the T157V mutant is consistent with our kinetic measurements, because the T157V mutant exhibited a near WT K m value for substrate but a 10-fold lower k cat value, thereby permitting the trapping of bound substrate (Fig. 5). In addition, we observed a change in side chain orientation at residue 157 upon substituting threonine with valine. In the WT enzyme structure, the side chain of Thr-157 adopts a 1 torsion angle value of ϩ60, and the OG1 atom of Thr-157 hydrogenbonds to Wat-3 and the backbone carbonyl oxygen atoms of residues 153 and 154 (Fig. 4). In the T157V mutant, these H-bonds are no longer possible, and the valine side chain adopts a presumably lower energy conformation ( 1 ϭ 180) in which the CG1 methyl group now points almost directly at the zinc atom (Figs. 4 and 6). This small structural change results in a small displacement of the zinc ion away from the CG1 atom of Val-157, and a more dramatic displacement of the zinc-bound water molecule at the active site (Wat-1), presumably as a result of repulsive van der Waals forces ( Fig. 6 and Fig. S4). This suggests the position of the zinc-bound cyanamide would be similarly displaced in the T157V mutant by steric clashes with the Val-157 side chain, possibly affecting the catalytic rate of hydration. For the WT enzyme, Thr-157 hydrogenbonds directly to Wat-3 that in turn hydrogen-bonds to Wat-4. Wat-4 would sit above the nitrile group of cyanamide in a Michaelis complex, almost within van der Waals distance of the substrate nitrile carbon. Hence not only is Wat-4 a candidate for the attacking water nucleophile in the hydra- The assays were carried out in duplicate, and standard errors for initial velocity measurements are indicated with black bars. For the kinetic assays, typically 1 g of purified recombinant protein is used, but in the case of the H137N mutant, 100 g was used to obtain measurable initial velocities.

Structure of Ddi2: a HD-domain cyanamide hydratase
tion reaction, but its occupancy and hence the hydration rate may be dependent on the presence of the Thr-157 side chain through the bridging hydrogen bond to Wat-3. Cryo-bufferonly soaks containing no cyanamide were also collected for the T157V mutant, processed, and refined (not shown) so that we could be certain we were observing the bound substrate in difference electron density maps and not two nearby water molecules (Fig. 4, compare A and B).
The complex of Ddi2-T157V and cyanamide revealed useful information regarding the geometry of cyanamide binding. Cyanamide appears bound directly to zinc via either the N2 nitrile atom or the N1 amino nitrogen (Fig. 4B). At the current limited resolution of 2.9 Å for the T157V substrate soak, we were unable to experimentally discern between these two possible binding modes for cyanamide. We also observed that the cyanamide molecule adopts a slightly different orientation than the two bound water molecules (Wat-1 and Wat-2) in the WT enzyme structure (Figs. 4 and 6). In analogy with high-resolution structures of carbonic anhydrase bound to cyanamide (carbonic anhydrase is a zinc metalloenzyme that also contains a bound, activated water molecule), we fitted a cyanamide molecule to the observed difference electron density with the nitrile nitrogen bound to the zinc (9). The zinc-cyanamide N2 distance is ϳ1.9 Å, similar to other typical zinc-ligand distances but slightly closer than the optimal distance for our fitted water molecules in the apo-enzyme structure and closer than the zinc-N2 cyanamide distance of 2.1 Å in carbonic anhydrase (Fig. 6) (13). In agreement with the structure of carbonic anhydrase bound to cyanamide, the zinc-N2 bond is at an angle with the N2-C1-N1 cyanamide molecular axis. In addition, the N2 atom of cyanamide forms a hydrogen bond with the side chain ND2 atom of Asn-161 (d ϭ 3.0 Å), possibly explaining why an N161A mutation exhibited low levels of catalytic activity (Fig.  5). This leaves the presumed N1 amino group of cyanamide forming hydrogen bonds with the NE2 atom of His-137 and the OD1 atom of Gln-138 at roughly equal lengths of 2.7-2.8 Å. The N1 amino group of cyanamide also forms a weak hydrogen bond or salt bridge with the OD2 atom of zinc ligand Asp-89 at a distance of 3.3 Å. Given that the Ddi2-H137N mutant has a significantly lower k cat than the native enzyme or the T157V mutant, it would appear that the hydrogen bond observed between cyanamide and the side chain NE2 of residue His-137 may be important for optimal substrate positioning prior to catalysis.

Ddi2 homologs in bacteria
A BLAST (6) search for Ddi2 homologs in the NCBI protein database indicated that a large number of Gram-positive bacteria contain Ddi2 homologs (Fig. S5). Homologs of Ddi2 are also found in the Actinobacteria. No significant hits were observed for sequences belonging to plants, archaea, or animals.
Conserved residues in Ddi2 and its presumed bacterial homologs include Phe-104 and His-137 (Ddi2 numbering) that are part of the Ddi2 substrate-binding pocket at the metal ion site, suggesting that they are important to support the activesite architecture, and His-137 may play a similar role in binding substrates or catalysis. However, Gln-138, a residue important for CAH activity as assessed in this study, is variable as either leucine or threonine in bacterial homologs, suggesting that bacterial Ddi2 homologs may not bind or hydrate cyanamide (Fig.  S5). Another active-site residue not conserved in bacterial ho- The values in parentheses correspond to the highest resolution shell. d R work ϭ ⌺ hkl ͉͉F obs (hkl)͉ Ϫ ͉F calc (hkl)͉͉/⌺ hkl ͉F obs (hkl)͉, where ͉F obs (hkl)͉ and ͉F calc (hkl)͉ are the observed and calculated amplitudes, respectively, for the structure factor F(hkl). e R free is the equivalent of R work for 5% of the reflections (randomly selected) that were not used in structure refinement.

Structure of Ddi2: a HD-domain cyanamide hydratase
mologs of Ddi2 is Asn-161, which hydrogen-bonds to cyanamide and a nearby water molecule. This residue is valine in bacterial homologs. Because valine has a hydrophobic side chain, it cannot form a hydrogen bond with solvent or cyanamide as Asn-161 does in Ddi2. However, the bacterial homologs of Ddi2 will likely bind Zn 2ϩ , because although Thr-157 is mutated to alanine or valine in bacterial homologs, it is not an aspartate in any of the sequences, as would be expected based on known HD motifs in other HD-domain proteins.

ddi2-H137N and ddi2-T157V are unable to rescue yeast ddi2⌬ ddi3⌬ mutant phenotype
To assess the functional implications of Ddi2 enzyme activity in vivo, we cloned the DDI2 gene including promoter and terminator sequences into a yeast multicopy plasmid (YEPlac195) and transformed it into yeast ddi2⌬ ddi3⌬ (ddi2/3⌬) double mutant cells. As expected, the ddi2/3⌬ double mutant displayed an increased growth defect in the presence of cyanamide in a gradient plate assay (Fig. 7). Yeast transformants harboring the YEpU-DDI2 plasmid carrying the WT DDI2 gene rescued the ddi2/3⌬ double mutant sensitivity to growth on cyanamide, exhibiting growth comparable with WT levels. In sharp contrast, the same plasmid containing the ddi2-H137N or ddi2-T157V mutant did not rescue growth defects of the ddi2/3⌬ strain grown in the presence of cyanamide (Fig. 7), strongly suggesting that the efficient hydrolysis of cyanamide by Ddi2 plays important roles in its biological function in vivo.

Implications for the presence of cyanamide hydratases in yeasts
Given that Ddi2 orthologs are widespread in yeasts and fungi, it is interesting to speculate on their likely role in cellular physiology and metabolism. The cellular response to cyanamide treatment in woody shrubs, for the breaking of bud dormancy, may provide valuable clues (7)(8)(9). Both transcriptome studies and physiological studies indicate that cyanamide exposure induces cyanide degradation pathways and that cellular H 2 O 2 levels rise because of the inhibition of catalase by cyanamide (7)(8)(9). Increased H 2 O 2 levels resulting from catalase inhibition appear crucial for bud break in woody perennials and also induce expression of several oxidative stress response proteins (7)(8)(9). Catalases also break down cyanamide to cyanide in the presence of H 2 O 2 (10). In fact, cyanide levels initially increase in plant tissue after application of cyanamide (8). Hence, if cyanamide has similar effects on yeast catalases, it would result in catalase inhibition and concomitant H 2 O 2 buildup, as well as production of significant amounts of toxic cyanide via residual catalase activity. The rapid induction of yeast DDI2 upon exposure to cyanamide would presumably remove the cyanamide before significant amounts of cellular catalase could undergo cyanamide inhibition. Hence the presence of Ddi2 would protect yeast cells from the unwanted buildup of H 2 O 2 resulting from cyanamide inhibition of catalase.
Given that nitroxyl generated from catalase-mediated breakdown of cyanamide is also known to inhibit aldehyde dehydrogenase (5,11,12), it is reasonable to assume that cyanamide might interfere with alcohol metabolism in yeast and hence be of considerable metabolic consequence. Hence, cyanamide hydratase would be expected to be beneficial under these circumstances. Finally, cyanamide can react via its electrophilic cyano group with ornithine or lysine to form arginine or homoarginine, respectively. Again, such chemical modifications may have important repercussions for the function of many cellular proteins in yeast.

Structure of Ddi2: a HD-domain cyanamide hydratase Other possible enzymatic activities for Ddi2
Not only can carbonic anhydrase function as a cyanamide hydratase, it is also known to function as an esterase using its zinc-bound water molecule to hydrolyze esters (14). Given that Ddi2 also contains a zinc-bound water/hydroxide molecule, perhaps Ddi2 also has an esterase function. Ddi2 is known to be strongly induced by exposure to MMS; therefore, it seems reasonable to suppose that MMS is hydrolyzed and therefore metabolically inactivated by the zinc-bound water molecule in Ddi2. Modeling suggests that MMS could easily fit into the Ddi2 active site near the zinc-bound water molecule (not shown). It is also relevant that most HD-domain family members have esterase functions, being phosphodiestereases. Further studies on the enzymology of Ddi2 and the effects of cyanamide on yeast physiology and metabolism will shed light on the roles of CAH enzymes in these organisms.

Ddi2 has unique metal coordination geometry as compared with other HD-domain proteins
The structure of yeast Ddi2 adds to the diversity of HD-domain family metalloenzymes and their respective chemistries because Ddi2 binds zinc and carries out a simple hydration reaction of cyanamide, yielding urea as the product. The structure of the enzyme is somewhat atypical for a HD family member because it has a significantly restricted active-site entrance compared with most HD-domain phosphohydrolase enzymes, which normally exhibit an "open" active-site architecture to accommodate relatively large nucleotide or dinucleotide substrates. Ddi2 also lacks the second aspartate metal ligand found in other HD-domain proteins.
Our structural studies on Ddi2 indicate that the zinc coordinated water molecule of the apo-enzyme is displaced by the substrate cyanamide where the nitrile N2 atom of the substrate binds directly to the zinc, as occurs in cyanamide bound carbonic anhydrase (13). However, we cannot rule out that it is the amino nitrogen of cyanamide that coordinates the zinc, given the resolution limits of our structural analysis. Our results are also consistent with the zinc directly binding to and activating the cyanamide for hydration by a nearby water molecule (Wat-4 in our structure). In the WT Ddi2 structure, Wat-3 is hydrogen-bonded to and likely stabilizes the position and occu-pancy of Wat-4 at the active site. The absence of Wat-3 in the T157V mutant enzyme may result in a lower occupancy of Wat-4 in Ddi2-T157V, and if Wat-4 is the attacking nucleophile, a lower catalytic rate (as observed) would be expected.
Another seemingly less likely possibility for catalysis is that the zinc-bound water in the apo-enzyme is actually the attacking nucleophile, and the cyanamide molecule observed bound directly to the zinc in the T157V mutant structure represents an abortive complex and not a true intermediate. However, this scenario would not explain the kinetic evidence of near WT substrate binding in the T157V mutant. In addition, this idea would be difficult to reconcile with the requirement for His-137 in substrate binding and catalysis, because there really is no other obvious place for cyanamide to bind at the active site than directly to the zinc, as we observe.
Together, these observations provide compelling evidence that the pocket containing the zinc site is the actual active site of Ddi2. Given that the pH optimum for Ddi2 is 7.5, it is reasonable to assume that His-137 does not carry a positive charge at pH 7.5, but the cyanamide N1 amino group may carry a positive charge via protonation. Hence His-137 more than likely functions to correctly position the substrate for catalysis via hydrogen bonding.
The current structural model of Ddi2-T157V with bound cyanamide has a relatively low resolution of 2.90 Å for definitive mechanistic studies, and hence we are unable to discriminate the exact chemical nature of cyanamide binding to the zinc ion (N1 or N2) at this resolution. Therefore, the possibility of cyanamide binding to the zinc ion via the N1 amino group cannot currently be ruled out. More concrete mechanistic inferences on the Ddi2 family of enzymes will have to await higher resolution structures of trapped substrates or inhibitors.
In summary, we have solved and refined the structure of S. cerevisiae Ddi2. The enzyme is a zinc-containing member of the HD domain superfamily and forms tightly associated dimers. Using crystallography, we also captured the bound substrate cyanamide in a catalytically compromised T157V mutant. Overall, the mode of cyanamide binding in Ddi2 is reminiscent of the binding of cyanamide to carbonic anhydrase. We also identified residues His-137, Gln-138, and Asn-161 as being important for substrate binding and/or catalysis because of their proximity to the zinc site or via interactions with the substrate. Finally, yeast lacking a functional cyanamide hydratase exhibit growth defects in the presence of cyanamide, strongly suggesting that oxidative and chemical stress brought on by cyanamide are harmful to yeast, and Ddi2 provides protection from these stresses.

Cloning of the DDI2/3 gene into a bacterial overexpression system
The cloning of DDI2 into a PGEX-6P1 GST-fusion vector has been described previously (3). The cloning process adds an 8-residue tag to the N terminus of the cleaved recombinant Ddi2 protein (GPLGPSEF). To clone DDI2 into a C-terminally His 6 -tagged expression vector, the DDI2 ORF was similarly amplified from yeast genomic DNA using primers 5Ј-GGCCC- Figure 7. Rescue of cyanamide sensitivity of a ddi2 ddi3⌬ mutant by WT Ddi2 or its active-site mutant derivatives. WT strain BY4741 and its ddi2⌬ ddi3⌬ mutant derivative WXY3149 were transformed with plasmids as indicated and used for a gradient plate assay as described under "Experimental procedures." The plates were incubated for 48 h before taking the photograph. Only a single representative clone for each strain is shown on the plates. The arrow points in the direction of increasing cyanamide concentration.

Structure of Ddi2: a HD-domain cyanamide hydratase
ATGGGCATGTCACAGTACGGA-3Ј and 5Ј-GGCGCGGCC-GCGTTATACCCAAATGTATT-3Ј. PCR products were digested with NcoI and NotI and inserted into plasmid pET28b (EMD Millipore) at the corresponding multiple cloning site. The resultant pET-DDI2-His 6 plasmid contains three additional alanine residues prior to the His 6 sequence. This plasmid was then used for creating active-site point mutants and subsequently used in cyanamide depletion kinetic studies (see Urease-based enzymatic assay of cyanamide hydratase activity).

Site-specific mutagenesis
A total of five mutant pET-DDI2 plasmids (H137N, Q138E, Q138A, T157V, and N161A) and two mutant PGEX-DDI2 plasmids (T157V and H137N) were constructed by using the QuikChange site-specific mutagenesis kit (Agilent). Site-specific mutations were introduced by carrying out PCR using primers that contain the mutated codons (Table S1). First, plasmid pET-DDI2-His 6 (for colorimetric cyanamide depletion assays) or pGEX-DDI2 (for steady-state kinetics and crystallography) were used as the template for PCR. After PCR, the DNA was digested by DpnI. The PCR product was directly transformed into Escherichia coli cells. The resulting plasmids containing the desired point mutations were recovered and verified by DNA sequencing. Recombinant mutant Ddi2 proteins were expressed and purified as described below. The cyanamide hydratase activity of Ddi2 mutants was measured using our enzymatic assay as described below (3).
To construct a yeast DDI2 expression plasmid, the DDI2 coding sequence plus its 888-bp upstream and 325-bp downstream sequences was amplified by PCR of genomic DNA using primers 5Ј-AGCTACGGTACCTTCAAAGGTTAAAC-TCGC-3Ј and 5Ј-TTGATGGAATTCTCACTGCAATGATT-TTCC-3Ј. The PCR product was digested with KpnI and EcoRI (recognition sequence underlined in the primers) and cloned into plasmid YEplac195 (32) to form YEpU-DDI2. H137N and T157V point mutations were introduced into YEpU-DDI2 by site-specific mutagenesis and confirmed by DNA sequencing.

Yeast cell survival assay
S. cerevisiae WT BY4741 (MATa his3⌬1 leu2⌬0 met15⌬0 ura3⌬0) and its isogenic ddi2⌬ ddi3⌬ mutant derivative WXY3149 (Chr14⌬::NAT ddi3⌬::HIS3) were transformed with YEpU-DDI2 derivatives, and several individual transformants for each strain were examined. The gradient plate assay was performed as previously described (33) to measure the relative sensitivity of S. cerevisiae cells to the presence of cyanamide. Briefly, the gradient was made by pouring 30 ml of molten YPD agar containing the predetermined cyanamide concentration into a tilted square Petri dish. After agar solidification, the dish was returned flat, and 30 ml of the same molten agar without cyanamide was poured to form the top layer. Overnight cell cultures grown in selective media were printed across the gradient using a microscope slide, and the plate was incubated at 30°C for 48 h before photographic recording.

Recombinant protein expression and purification
Recombinant plasmids pGEX-DDI2 or pET-DDI2-His 6 were transformed into E. coli strain BL21 (DE3), and protein expres-sion/purification was carried out as previously described (3). For GST-fused Ddi2, a further ion-exchange purification step was carried out after GST tag removal and prior to kinetic assays or protein crystallization. Partially purified Ddi2 was dialyzed into 20 mM Tris-HCl, pH 8.0, containing 20 mM NaCl before loading onto a Source Q anion-exchange column (GE Healthcare) and was eluted using a salt gradient from 0.02-2 M NaCl in 20 mM Tris-HCl, pH 8.0. Eluted protein fractions were analyzed by SDS-PAGE. Fractions containing highly purified Ddi2 were combined and concentrated to 10 mg/ml by using UltraSpin (EMD Millipore) microconcentrators with a molecular mass cutoff of 10 kDa. Concentrated Ddi2 was dialyzed into storage buffer containing 20 mM bis-Tris propane, pH 7.0, 150 mM NaCl, and 0.01 mM ZnCl 2 .
Similarly, C-terminally His-tagged Ddi2 was purified from E. coli BL21 (DE3) cells transformed with pET-DDI2-His 6 using nickel-nitrilotriacetic acid affinity chromatography. The cell lysate was equilibrated into buffer containing 50 mM Tris-HCl, pH 8.0, 500 mM NaCl, 20 mM imidazole and was then loaded onto a nickel-Sepharose FF column (GE Healthcare) and washed extensively prior to elution in the same buffer containing an additional 500 mM imidazole.

Urease-based enzymatic assay of cyanamide hydratase activity
Steady-state kinetic studies on purified recombinant WT Ddi2 and active-site point mutants thereof (T157V and H137N) were carried out as previously described (3). In the case of measuring the pH dependence of the cyanamide hydratase reaction, the urease and glutamate dehydrogenase reactions were carried out separately as previously described (3). To monitor the depletion of cyanamide in solutions containing purified Ddi2-His 6 or point mutants thereof, cyanamide was added to a final concentration of 2 mM to a solution of 3.8 M Ddi2 enzyme, and the solution was incubated at room temperature for 2 h. The cyanamide concentration was monitored using a colorimetric assay as described (31). Briefly, 100 l of reaction mixture was added to 500 l of PBS, followed by adding 400 l of 0.1 M sodium carbonate-bicarbonate buffer, pH 10.4, and 200 l of a 4% sodium pentacyanoammine-ferroate (II) (TCI Chemicals) solution as the color reagent. After reaction in the dark for 10 min, A 530 nm was measured to determine the remaining cyanamide concentration. The decrease of A 530 nm after incubation indicates the consumption of cyanamide in solution, reflecting the activity of cyanamide hydratase. In addition, initial velocity measurements were carried out for each mutant as described in Ref. 3, and no urea was detected after 20 min. Kinetic parameters and standard errors were calculated using the nonlinear least squares fit to the Michaelis-Menten equation as implemented in Prism/GraphPad.

Protein crystallization and cryoprotection
Crystals of Ddi2 or mutants thereof were grown by the hanging-drop vapor diffusion method at 20°C using 0.8 -2.0 l of protein solution (8 -11 mg/ml) mixed with an equal volume of precipitant solution. Initial crystallization hits were found using various commercial sparse-matrix crystallization kits, including Wizard I and Wizard II (Rigaku), and the ammonium sul-

Structure of Ddi2: a HD-domain cyanamide hydratase
fate suite (Qiagen). Single crystals of Ddi2 grew in solutions containing 1.1-1.3 M ammonium sulfate, 0.2 M arginine, 0.1 M N-morpholino ethane sulfonic acid, pH 5.2-6.0. Crystals took 2-3 weeks to grow to a size that allowed X-ray diffraction data to be recorded. Cryoprotection was attained by the sequential addition of mother liquor solution supplemented with 24 -26% (v/v) glycerol to drops containing crystals, followed by subsequent flash-freezing in liquid nitrogen or a stream of liquid nitrogen boil-off vapor. Substrate soaking was performed by addition of 0.20 -0.35 M cyanamide to the cryo-buffer, followed by subsequent flash-freezing in liquid nitrogen.

X-ray data collection and structure solution/refinement
Crystals of Ddi2 belong to the space group P321, a ϭ b ϭ 264.4 Å, c ϭ 119.2 Å. This space group is relatively rare for proteins (0.5% of PDB entries) and lacks a translational symmetry element along the principal 3-fold axis, but the presence of a 3-fold screw axis (space groups P3 1 21 P3 2 21, P3 1 , and P3 2 ) was clearly ruled out by the diffraction data based on the presence of numerous strong (0, 0, l) reflections for l 3n from l ϭ 3 up to l ϭ 43. In addition, indexing in lower symmetry Laue groups (e.g. 3) did not improve the merging statistics and indexing in 31m dramatically worsened the merging statistics. Furthermore, analysis of intensity statistics effectively ruled out the presence of merohedral twinning. We confirmed that P321 was the correct space group using Pointless (34,35), which yielded a probability of 0.998 that the space group was indeed P321 with no likely alternate choices. The structure of Ddi2 was hence solved to 3.2 Å resolution in space group P321 using the singlewavelength anomalous dispersion method (36). Three isomorphous zinc K-edge peak wavelength ( ϭ 1.28146 Å) data sets were collected under identical conditions (each data set was from a different crystal). Each data set consisted of 500 frames of 0.2°oscillations of 4 s at a crystal to film distance of 320 mm. All data frames from the three crystals were simultaneously scaled and merged within Denzo/HKL2000 and Scalepack (37) ( Table 1). Anomalous pairs were scaled independently in this process because both I ϩ and I Ϫ were at least 8-fold redundant overall. Using Solve (38), 10 candidate zinc sites were located and refined using anomalous scattering differences to a resolution of 8.0 Å resolution with an overall figure of merit of 0.28. Subsequent heavy atom substructure refinement and phasing in PHENIX-Autosol yielded nine stable sites and these were used to calculate initial phases and electron density maps to 3.2 Å resolution (39). The figure of merit after heavy atom refinement was 0.13 to 3.2 Å resolution. Solvent flattening and automated model building using PHENIX-Autosol improved phases significantly and generated an interpretable electron density map and partial model with the correct zinc ligands placed in most of the nine subunits (39). Structure building and refinement were then performed using COOT (40) and PHE-NIX (39). The final model was refined against native data to 2.6 Å resolution, and the unit cell contained 77% solvent. X-ray diffraction data on all Ddi2 crystals or point mutations thereof were collected using the bending magnet Beamline 08-B1 at the Canadian Light Source. The data collection strategy was identical to that used for the zinc single-wavelength anomalous dispersion peak data sets, and all indexing, scaling, and merging of diffraction intensities were carried out as described above. Further manipulations of the resultant diffraction intensities were performed using the CCP4 suite of programs (41). Structural figures were prepared using PyMOL (42).