Structural Investigations of N-carbamoylputrescine Amidohydrolase from Medicago truncatula: Insights into the Ultimate Step of Putrescine Biosynthesis in Plants

Putrescine, 1,4-diaminobutane, is an intermediate in the biosynthesis of more complexed polyamines, spermidine and spermine. Unlike other eukaryotes, plants have evolved a multistep pathway for putrescine biosynthesis that utilizes arginine. In the final reaction, N-carbamoylputrescine is hydrolyzed to putrescine by N-carbamoylputrescine amidohydrolase (CPA, EC 3.5.1.53). During the hydrolysis, consecutive nucleophilic attacks on the substrate by Cys158 and water lead to formation of putrescine and two by-products, ammonia and carbon dioxide. CPA from the model legume plant, Medicago truncatula (MtCPA), was investigated in this work. Four crystal structures were determined: the wild-type MtCPA in complex with the reaction intermediate, N-(dihydroxymethyl)putrescine as well as with cadaverine, which is a longer analog of putrescine; and also structures of MtCPA-C158S mutant unliganded and with putrescine. MtCPA assembles into octamers, which resemble an incomplete left-handed helical twist. The active site of MtCPA is funnel-like shaped, and its entrance is walled with a contribution of the neighboring protein subunits. Deep inside the catalytic cavity, Glu48, Lys121, and Cys158 form the catalytic triad. In this studies, we have highlighted the key residues, highly conserved among the plant kingdom, responsible for the activity and selectivity of MtCPA toward N-carbamoylputrescine. Moreover, since, according to previous reports, a close MtCPA relative from Arabidopsis thaliana, along with several other nitrilase-like proteins, are subjected to allosteric regulation by substrates, we have used the structural information to indicate a putative secondary binding site. Based on the docking experiment, we postulate that this site is adjacent to the entrance to the catalytic pocket.


INTRODUCTION
Polyamines (PAs) are small, aliphatic, polycationic compounds that bear at least two amino groups. Although PAs are found in all domains of life, the most abundant PA in nature is 1,4-diaminobutane, commonly named putrescine (PUT), which, together with 1,5-diaminopentane, cadaverine (CAD), owes its name to the noxious odor of putrefying cadavers. More complexed PAs include, but are not restricted to, triamine spermidine (SPD), tetraamine spermine (SPM) and thermospermine. It has been shown that PAs are some of the key players during plant growth and development, as well as response to both abiotic and biotic stresses (Do et al., 2014;Jimenez-Bremont et al., 2014;Minocha et al., 2014;Tiburcio et al., 2014;Liu et al., 2015). To list only a few examples, exogenous application of PUT can enhance thermotolerance of wheat (Kumar et al., 2014) and alleviate salt stress in apple callus (Liu et al., 2006). Also, supplementation with SPD mitigated salt stress symptoms in tomato and sorghum (Zhang et al., 2015;Yin et al., 2016). Application of exogenous PAs was shown to be not the only way to improve stress tolerance by increasing PA levels. Overexpression of genes involved in PA biosynthesis results in accumulation of PAs and, therefore, greater stress tolerance (Roy and Wu, 2001;Capell et al., 2004;Kasukabe et al., 2004;Alcazar et al., 2010;Wang et al., 2011). PAs are also related to nodulation, i.e., the symbiotic interaction between a plant hosts from legume (Fabaceae) family and nitrogen-fixing bacteria. Medicago truncatula (Mt), which is the source organism of the enzyme studied in this work, is a model legume plant. Legumes form special organs, root nodules, which are dwelled by nitrogenfixing bacteria from Rhizobium genus. This extraordinary plantmicrobe symbiosis allows legume plants to overcome limited nitrogen availability because rhizobia utilize atmospheric N 2 and convert it into ammonia, which plants are able to assimilate. It has been demonstrated that the concentration of PAs affects both number and biomass of root nodules (Vassileva and Ignatov, 1999;Terakado et al., 2006). Moreover, PA levels in the nodule tissue are elevated by a factor of 5-10, when compared to the rest of the plant organs (Fujihara et al., 1994;Efrose et al., 2008).
In plant cells, the homeostasis of PAs is attained by regulation of their biosynthesis, catabolism, conjugation and transport . The elementary PA that plants biosynthesize is PUT. Synthesis of SPD and SPM involves enzymatic conjunction of PUT with aminopropyl groups from decarboxylated S-adenosylmethionine. There are two routes leading to PUT synthesis (Figure 1) (Martin-Tanguy, 2001). One originates from ornithine, which, in a single reaction, catalyzed by ornithine decarboxylase (ODC, EC 4.1.1.17), is converted into PUT. In the other pathway, arginine is the starting molecule, and some plants, such as Arabidopsis thanliana (At), rely solely on this method (Hanfrey et al., 2001). The arginine route involves three enzymatic steps: (i) arginine decarboxylation to agmatine by arginine decarboxylase (ADC, EC 4.1.1.19); (ii) deimination of agmatine by agmatine iminohydrolase (AIH, EC 3.5.3.12) and (iii) hydrolysis of N-carbamoylputrescine (NCP) to putrescine by N-carbamoylputrescine amidohydrolase (CPA, EC 3.5.1.53), with the release of carbon dioxide and ammonia. The latter enzyme, CPA from Medicago truncatula (MtCPA), is the subject of the studies presented here.
Plants are the only eukaryotes able to produce PUT from arginine, which makes enzymes involved in PUT biosynthesis good targets for herbicide design. It has been shown that the plant pathway originates from bacteria (Illingworth et al., 2003;Fuell et al., 2010). Even though all three enzymes, ADC, AIH and CPA are examples of endosymbiotic horizontal gene transfer, they are encoded by nuclear DNA. It is, however, not surprising, as for instance a big part of cyanobacterial genes has been incorporated into the A. thaliana genome (Martin et al., 2002). Only ADC contains chloroplast-targeting signal peptide, whereas AIH and CPA are present in cytoplasm (Illingworth et al., 2003). Consistently, only the ADC sequence is present in many cyanobacteria, meanwhile most cyanobacteria use single agmatine ureohydrolase (EC 3.5.3.11) instead of AIH and CPA (Fuell et al., 2010).
N-carbamoylputrescine amidohydrolase enzymes belong to nitrilases, and more precisely, to C-N hydrolases breaking non-peptide bonds using Cys-Glu-Lys catalytic triad (Bork and Koonin, 1994;Pace and Brenner, 2001;Chen et al., 2003). Enzymatic activity of A. thaliana CPA (AtCPA) has been examined by Piotrowski et al. (2003). Using protein overexpressed in Escherichia coli cells, authors reported that AtCPA does not follow Michaelis-Menten kinetics, and its enzymatic activity can be described by Hill equation. The reported Hill coefficient, equal 2.2, indicated positive cooperative binding. The maximal velocity, V max was calculated as 86 nanokatals per mg of protein, and half maximal velocity was achieved at 135 µM of NCP. Based on native gel electrophoresis, authors also show that AtCPA in solution forms homooctamers (Piotrowski et al., 2003).
Even though the overall mechanism catalyzed by nitrilase enzymes has been established, particular subclasses must have evolved several modifications that allow for binding chemically different substrates. So far, there have been no structural information on any CPA enzyme from whichever domain of life. Therefore, it has been unknown how CPAs catalyze the enzymatic reaction, and which amino acid residues interact with the substrate and/or product. It also has been elusive what is the mechanism of the positive cooperative binging. In this studies, we address those fundamental questions, and using the first crystal structures of a CPA protein, expand the knowledge about these housekeeping enzymes.

Cloning, Overexpression and Purification of MtCPA
MtCPA protein was produced using a protocol adapted from the reported previously for obtaining δ 1 -pyrroline-5-carboxylate reductase from the same organism, . M. truncatula leaves were used for isolation of the total RNA with RNeasy Plant Mini Kit (Qiagen). cDNA was obtained with the use of SuperScript II reverse transcriptase (Life Technologies) and oligo dT (15 and 18) primers. PCR with the primers (Forward: TACTTCCAATCCAATGCCATGGCGGAAGACAAGGGTAGA AAAGTA and Reverse: TTATCCACTTCCAATGTTATCACAGA ACAGGATTTTTGCCGTCCAAA) and cDNA as a template was performed to obtain the coding sequence of MtCPA open reading frame. The PCR product was incorporated into pMCSG68 vector (Midwest Center for Structural Genomics) by ligase-independent cloning (Kim et al., 2011). The N-terminus of the expressed protein contains His 6 -tag, followed by the TEV protease cleavage site, and non-cleavable Ser-Asn-Ala linker that precedes the genuine sequence of MtCPA.
BL21 Gold E. coli competent cells (Agilent Technologies), transformed with the vector, were used for overexpression of the protein. Culture was carried out to the A 600 of 1.0 at 37 • C in LB media with the addition of ampicillin (150 µg/ml). The culture was cooled to 18 • C before induction. Then, IPTG (final concentration of 0.5 mM) was added to the culture. The culture was grown for 20 h and then cooled to 4 • C. Cells were pelleted by centrifugation at 3500 × g for 20 min. After the supernatant removal, cells were resuspended in 35 ml of the binding buffer (50 mM Tris-HCl pH 8.0; 500 mM NaCl; 20 mM imidazole; 1 mM TCEP) and frozen at −80 • C. Thawed culture was subjected to sonication to disrupt the cells. Bursts of total 4 min duration, with appropriate intervals for cooling, were applied. Disrupted cells were pelleted by centrifugation at 25 000 × g for 30 min at 4 • C. The supernatant was used to retrieve the His 6 -tagged MtCPA protein on the column packed with 5 ml of HisTrap HP resin (GE Healthcare), which was connected to the Vacuum Manifold (Promega). After application of the supernatant, column was washed five times with 40 ml of the binding buffer. The elution of MtCPA was performed with 20 ml of elution buffer (50 mM Tris-HCl pH 8.0; 500 mM NaCl; 300 mM imidazole; 1 mM TCEP). His 6 -tagged TEV protease was added to the sample to the final concentration of 0.1 mg/ml to cleave the His 6 -tag. The sample was dialyzed overnight at 4 • C against the buffer: 50 mM Tris-HCl pH 8.0; 500 mM NaCl; 1 mM TCEP. HisTrap HP resin was used to remove the cleaved His 6 -tag and the His 6tagged TEV protease. Final purification was performed by the size exclusion chromatography on HiLoad Superdex 200 16/60 column (GE Healthcare), equilibrated with the buffer: 50 mM Tris-HCl pH 8.0, 200 mM NaCl and 1 mM TCEP. The column was connected to the AKTA FPLC system (Amersham Biosciences).
MtCPA-C158S mutant was obtained by introduction of the C158S mutation to the vector used for overexpression of wild-type MtCPA, according to the Polymerase Incomplete Primer Extension method (Klock and Lesley, 2009). Primers used for PCR reaction were designed as follows -Forward: GCTATTTCCTGGGATCAGTGGTTTCCGGA and Reverse: AT CCCAGGAAATAGCGACTCCAATTTTCGCATA. The same protocol of overexpression and purification as for MtCPA was used in case of MtCPA-C158S. Sequences of MtCPA and MtCPA-C158S mutant were confirmed by DNA sequencing.

Crystallization and Data Collection
Both, MtCPA and MtCPA-C158S mutant proteins, were concentrated with Amicon concentrators (Millipore) to 20 mg/ml, as determined by the absorbance measurement at 280 nm, with the extinction coefficient of 38640. Initial crystallization trials were performed by sitting drop vapor diffusion method using the crystallization robot (Mosquito). Conditions were established for MtCPA as the 64 th reagent of PEG/Ion Screen HT (Hampton research). Both proteins were then crystallized manually by hanging drop method in the conditions containing 20% PEG3350 and 8% Tacsimate at pH 7.0 (Hampton Research) or 20% PEG3350, 8% Tacsimate at pH 7.0 and 10% glycerol (GOL). Crystals with the best morphology were obtained by streak seeding.
Complexes of MtCPA with N-(dihydroxymethyl)putrescine (DHMP) and MtCPA-C158S with PUT were obtained by incubation of the unliganded proteins with PUT (10 mM final concentration) for 15 min. Samples were then centrifuged at 17 000 × g before crystallization setup. The complex of MtCPA with CAD was formed using the same protocol, with the final 17 mM concentration of CAD. Crystals were cryoprotected by transfer to the solution containing 20% PEG3350, 8% Tacsimate at pH 7.0 and 25% GOL, supplemented with the appropriate ligand.
The diffraction data were collected at Southeast Regional Collaborative Access Team (SER-CAT) 22-ID beamline at the Advanced Photon Source, Argonne National Laboratory, USA. The diffraction images were processed with XDS (Kabsch, 2010); for details see Table 1. Initial data of MtCPA were processed to the resolution of 2.65 Å. The online version of BALBES, a molecular-replacement pipeline (Long et al., 2008) was used for structure determination. Twentyone structures with at least 15% identity to the given MtCPA sequence were recruited by BALBES. Subsequently, six structures with the best similarity of 39.6% along a 96-residue fragment were used for the structure determination. The following models from the PDB (Berman et al., 2000) were used: 3IVZ, 2W1V, 1F89, 1ERZ, 2VHH and 3AA0. All possible space groups in the orthorhombic system were tested. The best solution was obtained using the model 1ERZ with 574 residues in two monomers. Eight dimers of 1ERZ were found in the asymmetric unit of the MtCPA crystal lattice. The given model and the quality of electron density maps were inspected in Coot (Emsley et al., 2010). PHENIX AutoBuild (Terwilliger et al., 2008) was used for the model building, starting from the solution obtained from BALBES. After rebuilding, the model was taken for the successive steps of manual and automatic refinement with Coot and REFMAC5 (Murshudov et al., 2011). This model was used for the determination of all isomorphous MtCPA and MtCPA-C158S complexes by rigid body refinement. The refinement procedures were following: ligands and solvent molecules were deleted; the atomic displacement parameters were reset to 20 Å 2 for all atoms and rigid body refinement was used in the first refinement cycle. At the later stages of the structure refinement, TLS parameters (Winn et al., 2001(Winn et al., , 2003 were introduced. Restraints for DHMP were generated using PHENIX eLBOW (Moriarty et al., 2009) based on the chemical formula, while all the other ligands were refined against standard CCP4 libraries (Winn et al., 2011). The quality of refined structures was controlled by R work , R free factors (Brunger, 1992) and geometric parameters. PROCHECK (Laskowski et al., 1993) and MolProbity (Chen et al., 2010) were used for evaluation of the final models. Summary of data collection and refinement statistics are given in Table 1.

Docking
Docking calculations were carried out using Autodock version 4.2.6 ( Morris et al., 2009). Non-polar hydrogens were removed from the ligand, and their partial atomic charges were united with the bonded carbon atoms. The docking runs were performed using the Lamarckian genetic algorithm with grid sizes of 60 × 60 × 60 (grid spacing 0.375 Å), yielding 10 docked conformations. During the docking computation, free rotation was allowed around the two C-N aliphatic bonds. Three binding energy terms were taken into account in the docking step: the van der Waals interaction represented as a Lennard-Jones dispersion-repulsion term, the hydrogen bonding term, and the Coulombic electrostatic potential. The resulting 20,000 poses for NCP were analyzed by grouping them into clusters with similar ligand orientations.

MtCPA Forms Helical Octamers
The polypeptide chain of MtCPA contains 301 amino acids and its calculated molecular weight is 33.8 kDa. MtCPA crystallizes in P2 1 2 1 2 space group, with two octamers (chains A-H and I-P) as the asymmetric unit. In each of the structures presented herein, the octamer-flanking monomers (subunits A and H in the first octamer, and I and P in the second) present higher degree of flexibility than the inner subunits of the octamers. This is most probably due to lack of intermolecular contacts, which allows for higher conformational fluctuations. As a result, the electron density maps for the flanking protein subunits (monomers A, H, I, P) were not clear enough to model ligand molecules, and the ligands were placed for 12 out of 16 protein chains (B-G and J-O). CPAs show high degree of similarity, especially among plant species (Figure 2). Therefore, many conclusions, inferred from the presented structures, may be extrapolated to other CPA enzymes. The octameric quaternary structure in the crystal lattice of MtCPA is in agreement with the assembly reported, based on native gel electrophoresis, for the very close ortholog from A. thaliana (85% identity, 92% similarity) (Piotrowski et al., 2003). Based on the crystal structure, MtCPA octamers resemble an incomplete left-handed helical twist with an outer diameter of about 110 Å (Figure 3). The helical octamer surrounds a central void of 20 Å diameter. MtCPA octamer can be divided into four symmetric dimers that are shifted by about 30 Å along the long axis of adjacent dimer(s). The gap between the first and the last dimer within the octamer is approximately 10 Å. The dimers present αββααββα topology, and are formed by two monomers of αββα sandwich fold, which is very characteristic for the members of nitrilase superfamily (Nakai et al., 2000;Raczynska et al., 2011). To form a dimer, MtCPA monomers mutually swap their C-terminal fragments from His275 to Leu301 (Figure 4) on the inner side of the octamer. The C-terminal swap anchors the monomers to each other in an arm-to-arm manner, sturdily enhancing the monomer-monomer interactions. Furthermore, residues Asp295:Leu301 do not interact with a dimer mate, but instead increase the dimer-dimer interfaces in all but first (A/B and I/J) dimers in each octamer.
The oligomerization state varies among the nitrilases, although they all present a common architecture of αββααββα dimers, with swapped C-terminal fragments analogical to those observed in MtCPA. The length of the swapped fragments varies among proteins with available structures from nitrilase family, from short swap of as little as 10-residue fragment (Kumaran et al., 2003) to longer of about 70 residues (Agarkar et al., 2006). The quaternary assemblies of nitrilases are universally composed of even number of protein subunits. More precisely, the reported functional oligomers vary from dimers (Barglow et al., 2008;Liu et al., 2013), tetramers (Wang et al., 2001), hexamers (Agarkar et al., 2006), octamers, like MtCPA or β-alanine synthase from Drosophila Melanogaster (DmβAS) (Lundgren et al., 2008), through higher tube-like spiral multimers Sewell et al., 2003;Thuku et al., 2007).
In the PDB, one protein of the nitrilase family, DmβAS (Lundgren et al., 2008) (PDB ID: 2VHI), presents helically shaped octamers that resemble those formed by MtCPA. Even though the structures of MtCPA and DmβAS proteins present a similar assembly of the dimers at the first glance, the twist of the MtCPA octamer is more coiled. In contrast, the "uncoiled" DmβAS octamer has a diameter of approx. 120 Å, void of ≈ 30 Å and 65-70 Å pitch (rise along the axis per turn), whereas the corresponding values for MtCPA are: 110, 20, and 45 Å, with the latter value being only theoretical because the pitch of an incomplete helix is virtual. Authors report that DmβAS can freely form higher helical oligomers by attachment of further dimers to the octamer. In contrast, the helix pitch of MtCPA is too small to accommodate another dimer (≈ 75 Å along the long axis), and the addition of further dimers would either create severe steric clashes or force significant structural rearrangements (uncoiling) of the octamer. Consistently, the flanking dimers of the MtCPA octamer are separated by only a 10 Å gap, much narrower than in the DmβAS octamer. In DmβAS, the gap between the first and last dimer is almost 40 Å, which is sufficient to accommodate additional dimer, necessary for the formation of the theoretically infinite helical assembly. The structures of wild-type MtCPA, crystallized in the presence of PUT and CAD in the solution, were determined at 1.97 and 2.19 Å resolution, respectively. Inspection of the electron density maps, representing the structure of MtCPA cocrystallized with PUT, brought unexpected, but very informative results. Instead of linear PUT, the reaction intermediate, DHMP, was observed in the active site ( Figure 5A). More interestingly, DHMP was covalently bound to the catalytic Cys158. There are two conceivable scenarios of the DHMP adduct formation during crystallization. It is possible that PUT reacts with carbon dioxide (or carbonic acid) in solution, outside the catalytic site of the enzyme, for it is known that primary amines react with CO 2 in aqueous solution (Penny and Ritter, 1983;McCann et al., 2009). The resulting N-carbamicputrescine (with carboxylic COO − group attached to one of the PUT amines) may enter the active site and, similarly to the NCP, most likely is susceptible to the nucleophilic attack of the Cys158. The DHMP covalent adduct is stabilized by placement of one of its oxygen atoms within the oxyanion hole, created by Nζ of Lys121 and the backbone N of Trp159 (see Active Site and the Reaction Mechanism). The other scenario of events, which might lead to the covalent enzyme-DHMP adduct in the crystallization drop, involves an initial reaction of carbon dioxide (or carbonic acid) with the active Cys158. To our judgment, however, this situation is much less likely for two reasons. Firstly, the covalent C-N bond in that case would need to result from a nucleophilic attack of N atom of PUT on the S-bound carboxyl C, which is very questionable. Secondly, if the second scenario was true, one would expect a similar state in the structure of MtCPA cocrystallized with CAD, because nothing should prevent the Cys158 carboxylation in the first place. In that instance, there would be not enough room to bind entire CAD molecule. This is not the case, and we do indeed observe CAD in the active site ( Figure 5B). It is, of course, possible that CAD molecules in solution also form CO 2 adducts, but due to steric clashes (see Substrate Specificity), they cannot be recruited to the catalytic site, meanwhile non-modified CAD molecules can. CAD in MtCPA complex reaches deeper than PUT into the catalytic pocket, however, due to presence of sulfhydryl group of Cys158 it bends toward Glu132.
The crystal structure of wild-type MtCPA without addition of any ligands was determined at 2.55 Å resolution (data not shown). Unfortunately, the electron density maps for the content of the catalytic site were unclear, and represented most probably disordered ingredients of the crystallization solution, GOL, and partially oxidized Cys158. Combined with the rather poor data resolution, we were not able to unambiguously interpret the content of the active site of "unliganded" wildtype MtCPA, and we do not report this structure herein. Instead, we decided to construct the MtCPA-C158S mutant. Substitution of sulfhydryl with hydroxyl group preserved binding properties of residue 158 but strongly diminished its reactivity. The structure of unliganded MtCPA-C158S, determined at 2.39 Å resolution, presented significantly lower noise in the electron density maps around the active site when compared to the unliganded wild-type MtCPA. This improvement allowed to identify the ligand bound inside the active site as GOL ( Figure 5C). However, GOL does not penetrate deep inside the pocket and presents no direct interactions with either of the residues from catalytic triad. Instead, it is hydrogen-bonded with the side chain oxygen of Glu187.
The crystal structure of MtCPA-C158S mutant cocrystallized with PUT clearly shows the ligand without any modifications ( Figure 5D). In this case, the absence of nucleophilic sulfur prevents from the formation of covalent bond between DHMP and residue 158. In fact, PUT in MtCPA-C158S complex does not interact directly with the residue 158, but creates water-mediated hydrogen bond with the hydroxyl group of Ser158 instead.

Active Site and the Reaction Mechanism
Each MtCPA monomer contains a cavity, which is accessible from the outer surface of MtCPA octamer and, in its depth, comprises the active site. In order to present the active sites in a clear manner, we divide the MtCPA cavities to four parts, each buried deeper in the protein core: (i) the entrance preceding the genuine active site; (ii) the tail amine-binding region, which binds the non-reacting amine group of PA; (iii) the non-polar section, interacting with the ligand aliphatic fragment; and (iv) the catalytic triad, which performs the reaction. The cavities are in a very close proximity to the neighboring protein subunits, with the exception of the outer monomers A, H, I, and P that only have a single neighbor. The entrances to the funnel-like catalytic cavities are actually shaped not only by the amino acids from the protein subunit to which a particular site belongs, but are also fenced by protein chains of its immediate neighbor(s) (Figure 6). Moreover, each dimer-mate subunit surrounds a larger fragment of the active site entrance than the protein chain from the neighboring dimer.
Glu187, Asp190, together with several carbonyl oxygens of Trp277 * (asterisk indicates a residue that belongs to dimer-mate subunit), Cys276 * , Asp126 and Pro128 are responsible for the negative potential localized around the active site ( Figure 6A). Most probably, they attract the positively charged NCP, when it penetrates its way through the cavity. Walls of the middle section of the active site are composed of the side chains of: Tyr130, Tyr54, Pro125, Trp159, Trp162, Ile184, Ala183, and by Trp277 * from the swapped C-terminal fragment ( Figure 6B). In this manner, MtCPA provides the hydrophobic milieu of the central section of the active pocket. This fragment is, therefore, perfectly adapted to bind the aliphatic fragment of the substrate, NCP by hydrophobic interactions.
As in other nitrilases, the substrate hydrolysis is carried out by the catalytic triad, composed in MtCPA of Glu48 (proton acceptor), Lys121 (proton donor) and Cys158 (nucleophile). These three residues are placed deep inside the narrow part of the active site. The nucleophilic Cys, as in any other nitrilaselike protein, is positioned at the apex of the nucleophile elbow (Kumaran et al., 2003), formed in MtCPA by β7-turn-η6 motif, and its tightness enhances the nucleophilic character of Cys158 by enforcing energetically unfavorable conformation.
The nucleophilic character of the active Cys158 is additionally fortified by two glutamates, Glu48 and Glu132, which attract the sulfhydryl proton. Both glutamates, i.e., Glu48, as the member of the catalytic triad, but also Glu132 are highly conserved among nitrilase superfamily. Most probably, this pair attracts the NH 2 group of the NCP, therefore, properly orients the carbamoyl carbon of the substrate for the nucleophilic attack by Cys158. Moreover, Glu48 and Glu132 affect the closest environment of Lys121 and preserve its positive charge, crucial for the hydrolysis.
The hydrolysis of NCP by MtCPA most likely follows a mechanism similar at first stages to the proposed earlier for Helicobacter pylori formamidase (HpAmiF) (Hung et al., 2007), but after creation of the oxyanion geminal diol intermediate (see below), it carries on in its own way. Briefly, activated Sγ of Cys158 performs nucleophilic attack on the carbamoyl C atom of the substrate, creating the covalent, tetrahedral (geometry concerns the initially planar C atom, which ultimately leaves as HCO 3 ) diaminothioalcoholate intermediate (Figure 7). The generated oxyanion is stabilized by the oxyanion hole, formed by Nζ of Lys121 and backbone N of Trp159, similarly to the case reported for Pyrococcus abyssi nitrilase (PaNit) (Raczynska et al., 2011). Subsequently, the amino group of the intermediate accepts a proton, donated by Glu48, and leaves as the first by-product, ammonia, resulting in the planar S-acyl intermediate-enzyme complex, which is hydrated to yield the tetrahedral oxyanionic geminal diol (present in the crystal structure of MtCPA/DHMP complex). Then, most probably, the water molecule tethered to Oε of Gln161 and Nε of Trp162 is added to the intermediate, and, in an apparently concerted manner, C-S and C-N bonds are cleaved, which results in the release of PUT and the second by-product, HCO 3 − .

Substrate Specificity
Enzymatic assays on CPA enzymes from At (Piotrowski et al., 2003) and Pseudomonas aeruginosa (PaerCPA, Nakada and Itoh, 2003) show that CPAs are highly active only against NCP. Plant AtCPA showed no activity toward substrates of other nitrilaselike enzymes, e.g., amino acids, N-Carbamoyl-β-alanine, agmatine. Also, the bacterial PaerCPA very poorly hydrolyzed both longer and shorter substrates in comparison to NCP: N-carbamoylcadaverine and N-carbamoyldiaminopropane, respectively. We addressed the question of structural features that are responsible for such narrow substrate specificity of CPA enzymes. The comparison of the product (PUT), its longer analog (CAD) and the intermediate (DHMP) bound within the active site of MtCPA highlights key elements of the active site architecture, which determine the enzyme preference for NCP (Figure 8). All bound amines, CAD, PUT and DHMP, present the same hydrogen bond between terminal amine group and the side chain oxygen of Glu187. Additionally, these amino groups create two water-mediated hydrogen bonds with carbonyl oxygens of Asp126, Trp277 * , and mediated by two waters hydrogen bonds with Pro128 and Cys276 * . Glu187, conserved in CPA enzymes, prevents from substrate mismatch, and stabilizes  substrate position within the active site. Moreover, the position of negatively charged Glu187 at the active site entrance determines both the charge and length of the bound substrate. Longer or shorter molecules than NCP, would not be able to create a hydrogen bond with the oxygen of Glu187, and/or would collide with other parts of the active site.
The pocket presents very narrow and highly hydrophobic middle section, which excludes bulky substrates form binding. It also limits the spectrum of possible substrates to those which have an unbranched aliphatic fragment of four carbon atoms. From the MtCPA/CAD complex, we can assume that longer substrates than NCP, such as N-carbamoylcadaverine, would not obtain stable conformation inside the active site. Although CAD is able to interact via hydrogen bond with Glu187 at the entrance of the pocket, similarly to PUT and DHMP, deep inside the cleft CAD passes by the Sγ of Cys158 and bends toward the pair of glutamates. Thus, this would not leave enough space for the carbamoyl moiety of such longer substrate. In other words, the proper orientation of N-carbamoylcadaverine, with the preservation of the hydrogen bond with the Glu187 side chain is impossible due to steric clashes either with the catalytic triad or with the residues in the middle section of the pocket.
The commercial availability of NCP is extremely limited, therefore, during our studies, we attempted to obtain complexes with other ligands, such as urea, sarcosine, L-lysine, L-arginine, which would mimic the substrate. Unfortunately, their complexes with neither wild-type MtCPA nor its C158S mutant allowed to unambiguously identify any other ligand of interest within the active site, except for those presented in this paper.

Proposed Secondary Binding Site
As it was shown by Piotrowski et al. (2003), the hydrolysis of NCP by CPA from A. thaliana does not follow the Michaelis-Menten kinetics. Instead, in the excess of substrate, an accelerating mechanism is triggered. Since PAs are linked with the plants response to stress factors, this phenomenon may actually be significant in vivo. The positive allosteric regulation was found also in other nitrilase-like proteins, β-alanine synthases (Tamaki et al., 1987;Matthews et al., 1992;Sakamoto et al., 2001;Lundgren et al., 2008). In these enzymes, the mechanism was attributed to the change of oligomerization caused by increasing substrate concentration, rather than to the existence of an allosteric regulatory site (Matthews et al., 1992). However, in MtCPA, assuming that it also follows the Hill kinetics, due to disabled formation of higher oligomers, a different mechanism may control the enzyme behavior.
The architecture of close neighborhood of the catalytic site shows a solvent/cytoplasm accessible niche near the substrate binding pocket. In every structure presented herein, this niche is occupied by a GOL molecule. Considering the fact that in the MtCPA-C158S mutant structure, GOL enters the active site, and mimics the substrate or product, the niche-bound GOL quite likely might represent the secondary binding site. In that case, in higher concentration of NCP, this site might be occupied by the substrate. Docking simulation seems to support this hypothesis. In fact, the carbamoyl moiety of the docked substrate, NCP, significantly overlaps with the GOLbinding site from the presented here complexes (Figure 9). The substrate is situated in the niche with the carbamoyl moiety aiming at the entrance to the active site. In the predicted alternative binding site, NCP, would form three hydrogen bonds with Asp194, His198 and Glu248 * . The calculated binding energy of the secondary binding site is −3.6 kcal mol −1 , whereas the corresponding value for the genuine catalytic cavity equals −5.8 kcal mol −1 . It is, therefore, suggestive that the substrate affinity is significantly lower in the alternative binding site, nonetheless, still within a reasonable range.
At this point, however, the mechanism of allosteric regulation remains unrevealed, for it appears that binding of the second NCP molecule, should neither directly facilitate binding of the substrate within the active site, nor it should impact the catalytic triad. It is possible that the alternative NCP binding location represents a standby spot for substrate molecule. Even though it would not be an allosteric regulation of the enzyme activity per se, it would mitigate the delay needed for the substrate diffusion into the active site. By far, it is merely a speculation, and many additional experiments are required to validate this hypothesis, however, it is worth to mention that a standby substrate position was reported, e.g., for tyrosine phosphatase from Selenomonas ruminantium (Puhl et al., 2007), unrelated to CPA proteins.

Conclusions, Physiological Significance and Future Outlook
This study was focused on the structural characterization of CPA from the model legume plant, M. truncatula. The octameric MtCPA enzyme is assembled into an incomplete helical twist, made by four tight dimers with swapped C-termini. This dimeric αββααββα organization is highly conserved among nitrilaselike enzymes, however, the analogical octameric assembly was observed only in DmβAS (Lundgren et al., 2008).
The octameric structure of plant CPAs, reported previously for A. thaliana (Piotrowski et al., 2003) and herein for M. truncatula brings up a fundamental question of physiological benefit from forming high oligomers. The role of PUT in response to abiotic stresses (mentioned in the Introduction), such as high salinity, drought and/or thermal extrema, could provide a conceivable answer. Assembling into higher oligomers undoubtedly increases enzyme's tolerance to adverse conditions, for a smaller portion of the protein surface is exposed toward the cytoplasm and, therefore, is less susceptible to changing environment. A similar rationale was anticipated to explain oligomerization of other abiotic-stress-related plant proteins, which form ring-like decamers, δ 1 -pyrroline-5-carboxylate reductases .
N-carbamoylputrescine amidohydrolases are enzymes highly specific towards NCP (Piotrowski et al., 2003), and the architecture of the MtCPA catalytic cavity, formed as a deep pocket with broadened entrance, explains this phenomenon. The entrance and its middle section is built with a contribution of the residues from neighboring monomers Cys276 * and Trp277 * . Two characteristic features influence the substrate specificity of MtCPA: the narrow and hydrophobic channel in the middle section of the catalytic cavity (Tyr130, Tyr54, Pro125, Trp159, Trp162, Ile184, Ala183 and Trp277 * from dimer mate) and Glu187, placed at the entrance, the key residue responsible for the proper substrate orientation inside the active site. The catalytic triad at the bottom of the cavity is composed of Glu48, Lys121 and Cys158. Glu48, the proton acceptor, in the cooperation with Glu132 enhances the nucleophilic character of Cys158 and maintains the proper charge of Lys121, the proton donor. Lys121, together with backbone N of Trp159 forms the oxyanion hole, which stabilizes the intermediate state during NCP hydrolysis. Cys158, placed at the apex of the nucleophile elbow, formed in MtCPA by β7-turn-η6 motif, is responsible for the nucleophilic attack on the substrate, and creation of the covalent intermediate-enzyme complex. The catalytic water molecule, which performs subsequent nucleophilic attack on the intermediate, is stored within the active site in a small niche, near the catalytic triad. This water is hydrogen-bonded to the side chains of Gln161 and Trp162. The reaction leads to the breakage of two covalent bonds of the substrate, which ends with the release of PUT, ammonia and carbon dioxide. Presented here indication of the key residues, responsible for specificity and activity of MtCPA, sheds new light on CPA enzymes overall. Any future in-depth physiological study of CPAs can be more rational, as the research can benefit, or even directly arise from the structural knowledge presented herein.
The crystallographic analysis of the complexes of MtCPA and its C158S mutant confirmed high affinity of their active sites for ligands, which were either absorbed from crystallization solution (GOL), or intentionally added to obtain MtCPA complexes (PUT and CAD). Interestingly, in the complex of the wild-type MtCPA, obtained by the cocrystallization with PUT, the reaction intermediate, DHMP, was covalently bound to the active Cys158, inside the active site instead of PUT. DHMP adduct was most probably formed as the result of the reaction of the amine group of PUT with carbon dioxide, absorbed by the crystallization drop, followed by nucleophilic attack of the active Cys158. This showed that the oxyanion hole, next to the MtCPA catalytic triad, is able to stabilize the otherwise labile geminal diol. It would be very interesting to see whether a similar scenario could be relevant in planta, that is, if CO 2 to a significant extent reacts with PUT in the cytoplasm and (as carbamic acid) enters the catalytic cavity of MtCPA. If so, the enzyme activity would be modulated by CO 2 concentration, which would give another link between PAs and plant stress response.
Several members of nitrilase superfamily have been previously reported to exert enzymatic mechanism suggesting cooperative binding. Due to lack of a candidate secondary substrate-binding position, for most of them this behavior was associated with changes of the oligomerization state. MtCPA, however, is not able to change the oligomerization state, as its octamers are formed in a very coiled manner. We examined MtCPA structure to indicate a secondary substrate-binding site. The analysis led to designation of the conserved (in MtCPA structures) GOL binding location as the potential alternative substrate binding site. This hypothesis was supported by docking experiment, which indicated that the substrate may indeed bind also in a niche, with carbamoyl moiety of the docked NCP partially overlapping the GOL binding site. This venue is very close to the entrance of the active site, but separated from the catalytic triad. However, with present data we are not able to unambiguously determine the activity-accelerating mechanism linked to the alternative binding site. We may suppose that MtCPA might not be allosterically regulated per se, but rather own a site, where the substrate can standby, prior to entering the catalytic site. In that case, the turnover rate of the enzyme would be increased, as the delay needed for the substrate to diffuse in the cavity would be alleviated if it was bound right next to the entrance. This hypothesis, however, needs to be verified using parallel functional and structural approaches, nonetheless, it sketches a perspective for future studies of CPA enzymes. From the physiological perspective, accelerating mechanism is very beneficial, as it triggers increased enzyme activity when it is most needed, bypassing regulatory mechanisms of transcription and translation. During abiotic and biotic stresses as well as in the course of nodulation, when the demand for PAs is aggravated, plant can produce sufficient amount of PUT without engaging the protein synthesis machinery to deliver more molecules of CPA, but instead simply providing more NCP.

AUTHOR CONTRIBUTIONS
BS produced the protein and solved the crystal structures. BS, MR, and ZD analyzed the data. MM performed docking calculations. MR and BS wrote the manuscript.