Insights into the reaction mechanism of Escherichia coli agmatinase by site-directed mutagenesis and molecular modelling A critical role for aspartate 153 Mo ´ nica Salas 1 , Rolando Rodrı ´ guez 2 , Nelia Lo ´ pez 2 , Elena Uribe 1 , Vasthi Lo ´ pez 1 and Nelson Carvajal 1 1 Departamento de Biologı ´ a Molecular, Facultad de Ciencias Biolo ´ gicas, Universidad de Concepcio ´ n, Casilla 160-C, Concepcio ´ n, Chile; 2 Center for Genetic Engineering and Biotechnology, Habana, Cuba Upon mutation of Asp153 by asparagine, the catalytic activity of agmatinase (agmatine ureohydrolase, EC 3.5.3.11) from Escherichia coli was reduced to about 5% of wild-type activity. Tryptophan emission fluorescence (k max ¼ 340 nm), and CD spectra were nearly identical for wild- type and D153N agmatinases. The K m value for agmatine (1.6 ± 0.1 m M ),aswellastheK i for putrescine inhibition (12 ± 2 m M ) and the interaction of the enzyme with the required metal ion, were also not altered by mutation. Three- dimensional models, generated by homology modelling techniques, indicated that the side chains of Asp153 and Asn153 can perfectly fit in essentially the same position in the active site of E. coli agmatinase. Asp153 is suggested to be involved, by hydrogen bond formation, in the stabilization and orientation of a metal-bound hydroxide for optimal attack on the guanidinium carbon of agmatine. Thus, the disruption of this hydrogen bond is the likely cause of the greately decreased catalytic efficiency of the D153N variant. Keywords: agmatinase; Asp153; site-directed mutagenesis; homology-modelling; E. coli. Agmatinase (agmatine ureohydrolase, EC 3.5.3.11) cata- lyses the hydrolysis of agmatine to putrescine and urea [1]. Agmatine, which results from decarboxylation of arginine by arginine decarboxylase [2], is a metabolic intermediate in the biosynthesis of putrescine and higher polyamines [1] and may have important regulatory roles in mammals [3–5]. Agmatinases from Escherichia coli and human tissues, and putative agmatinases from Synechocystic sp. Schizo- saccharomyces pombe and Bacillus subtilis, have been cloned and the deduced amino acid sequences indicate their homology to all sequenced arginases [4–7]; all these enzymes catalyse an hydrolytic reaction with production of urea. The question arises therefore as to whether a similar or identical mechanism is involved in catalysis by these enzymes, which apparently evolved from a single primordial protein [6,7]. In this context, both enzymes exhibit an absolute requirement for Mn 2+ for catalytic activity [8,9]; the well established requirement of a binuclear metal cluster for full catalytic activity of arginase [8] is probably also valid for agmatinase [9]. This is reinforced by the fact that residues known to be metal ligands in arginase are strictly conserved in the sequence of agmatinase [7]. Moreover, a critical role for one conserved histidine residue (His163 in the sequence of E. coli agmatinase) has been shown by chemical modifica- tion and site-directed mutagenesis of human and rat liver arginases [10,11] and E. coli agmatinase [12]; similar infor- mation was deduced from X-ray crystallographic data for arginase from Bacillus caldovelox [13]. Based on the crystal structure of rat liver arginase, it was suggested that arginine hydrolysis involves the participation of a metal-bound hydroxide, which is stabilized for optimal nucleophilic attack at the substrate, by donating an hydrogen bond to Asp128 [8,14,15]. In this connection, the D128G variant of human liver arginase was described as inactive [16,17], although the possible influence of structural changes accompanying the mutation were not examined. Since this aspartate is conserved among all sequenced arginases and agmatinases [4–7], a critical role for the equivalent residue in agmatinase (Asp153), may be reason- ably expected. This expectation is supported by our present findings of a markedly decreased activity of a D153N variant of E. coli agmatinase. From the enzymic properties of D153N agmatinase and a modelled structure, we conclude that the lower activity of the mutant may be ascribed to the loss of an acceptor hydrogen bond to a metal-bound hydroxide, as a consequence of replacement of a carboxylate oxygen with an amide group. MATERIALS AND METHODS Materials All reagents were of the highest quality commercially available (most from Sigma Chemical Co.) and were used without further purification. Restriction enzymes, as well as enzymes and reagents for PCR were obtained from Promega. The plasmid pKA5, bearing the speB gene of E. coli agmatinase, was kindly supplied by S. Boyle (Vir- ginina Polytechnic Institute and State University). The pQE60 E. coli expression vector and the Ni-nitrilotriacetic acid resin were obtained from Qiagen, and synthetic Correspondence to N. Carvajal, Departamento de Biologı ´ a Molecular, Facultad de Ciencias Biolo ´ gicas, Universidad de Concepcio ´ n, Casilla 160-C, Concepcio ´ n, Chile. Fax: + 56 41 239687; E-mail: ncarvaja@udec.cl (Received 5 June 2002, revised 9 September 2002, accepted 12 September 2002) Eur. J. Biochem. 269, 5522–5526 (2002) Ó FEBS 2002 doi:10.1046/j.1432-1033.2002.03255.x nucleotide primers from the Centro de Estudios Avanzados (Universidad de Chile). Enzyme preparations and enzyme assays Bacteria were grown with shaking at 37 °C in Luria broth in the presence of ampicillin (100 lgÆmL )1 ). The wild-type and D153N agmatinase cDNAs were directionally cloned into the histidine tagged pQE60 E. coli expression vector and the histidine-tagged enzyme was expressed in E. coli strain JM109, following induction with 1 m M isopropyl thio-b- D -galactoside. The histidine-tagged enzymes were purified to homogeneity by metal chelate chromatography over Ni-nitrilotriacetic acid resin, according to the instructions of the manufacturer. A single protein band was detected by SDS/PAGE of purified enzymes. Enzyme activities were determined by measuring the formation of urea from agmatine in 50 m M glycine/NaOH (pH 9.0). Urea was determined by a colorimetric method with a-isonitrosopropiophenone [18] and protein concen- trations were estimated by the method of Bradford [19], with bovine serum albumin as standard. Kinetic data were analyzed by double reciprocal plots, and the K i value for putrescine inhibition was determined from a replot of slopes vs. inhibitor concentration. All lines were computer-fitted to the appropriate equations. Site-directed mutagenesis The D153N mutant form of E. coli agmatinase was obtained by a two-step PCR [20], using the plasmid pKA5 containing the speB gene of E. coli agmatinase as a template. A first PCR product was obtained using the 5¢ sense primer 5¢-AGTCCATCCATGGGCACCTTAG-3¢ and a 3¢ complementary primer corresponding to nucleo- tides 448–468 of agmatinase with a CfiT substitution at nucleotide 457 (sequence: 5¢-CGCATAGGTAT TG GTGTGGGC-3¢). Similarly, the second PCR product was obtained using the 5¢ sense primer corresponding to nucleotides 448–468 of agmatinase with a GfiA substitu- tion at nucleotide 457 (sequence: 5¢-GCCCACACC AAT ACCTATGCG-3¢) and the 3¢ complementary primer 5¢-AT TAATGGCATGCTTTACCCGT-3¢.UsingthePCR products of agmatinase with the GfiAandCfiT substi- tutions in the coding and noncoding strands, respectively, and using the 5¢ and 3¢ primers mentioned above, the full length agmatinase cDNA coding for the D153N mutant was generated by a second round of PCR. The expected mutation was confirmed by DNA sequence analysis. That no unwanted mutations had been introduced during the mutagenesis process was verified by automated sequencing. The H163F E. coli agmatinase was obtained as described previously [10]. Fluorescence spectra Fluorescence measurements were made at 25 °Cona Shimadzu RF-5301 spectrofluorimeter. The protein concen- tration was 30 lgÆmL )1 and emission spectra were measured with the excitation wavelength at 295 nm. The slit width for both excitation and emission was 1.5 nm, and spectra were corrected by subtracting the spectrum of the buffer solution (5 m M Tris/HCl, pH 7.5) in the absence of protein. Circular dichroism CD experiments were performed on a Jasco J-810 spectro- polarimeter thermostated at 22 °C. CD spectra of wild-type and D153N mutant enzymes (5.5 l M ) were measured in the range 200–250 nm, with a bandwidth of 1 nm and a scan speed of 50 nmÆmin )1 . The buffer solution contained 5 m M Tris/HCl (pH 7.5) and 2 m M Mn 2+ . The reported spectra represents the averages of five repeat scans. Spectra were smoothed and analysed for protein secondary structures by using the software package provided with the instrument. Molecular modelling The agmatinase structural model was obtained by homology methods, using the structure of B. caldovelox arginase (PDB id: 1CEV) as a template and the modelling package WHAT IF [21]. An amino-acid sequence identity of 29% was calculated for E. coli agmatinase and B. caldovelox arginase. The sequence alignment used in the modelling experiment was derived from the structural superposition of two arginase structures (rat liver and B. caldovelox arginases, PDB id 1RLA and 1CEV, respectively). The agmatinase SPEB_ECOLI sequence, obtained from the Swissprot database (accession number P16936), was separately aligned with the sequences of 1CEV and 1RLA and pasted into the structural alignment. The alignment was then corrected by hand. Since the sequences of the two arginases and agmatinase greately differ in the N terminal region, the agmatinase and rat liver arginases were stripped of the first 32 and five amino acids, respectively. The location of the gaps was optimized by repetitive modelling, shuffling those aminoacids that were not conserved in the structural alignment, to obtain the shortest possible Ca-Ca distance. The resulting sequence alignment, along with the secondary structure elements, is shown in Fig. 1. All loops that were different due to insertions or deletions in the agmatinase sequence were modelled using the DGLOOP set of options in WHATIF ; the whole loops and the two connecting amino acids at the beginning and the end, were mutated to glycines and, after remodelling back to their side chains in agma- tinase, the whole hydrogen bond network was optimized. The position of the active site manganese ions was calculated using the averaged distance template of 1CEV and 1RLA, and then a full hydrogen bond network optimization was performed again. Agmatine was added to agmatinase using the same superposition matrix calculated for manganese ions, in order to conserve the ligand-manganese distances. RESULTS AND DISCUSSION By site-directed mutagenesis, a D153N mutant form of E. coli agmatinase was obtained. The mutant enzyme retained about 5% of wild-type activity and it was equally active when assayed in the presence or absence of Mn 2+ . However, as shown in Fig. 2, it was half inactivated by dialysis against 10 m M EDTA in 5 m M Tris/HCl (pH 7.5) for 4 h at 4 °C and full recovery of enzyme activity was produced by incubation of the EDTA-treated species with 2m M Mn 2+ for 20 min at 37 °C; as expected, the effect of the metal ion was again reversed by EDTA. These results indicate the presence of tightly and weakly bound mangan- ese ions in fully active species of the mutant enzyme. Ó FEBS 2002 A critical role for Asp153 in E. coli agmatinase (Eur. J. Biochem. 269) 5523 Identical behaviour was previously described, and con- firmed here (Fig. 2), for wild-type agmatinase and this was interpreted as supporting the presence of a binuclear metal center in the active site of fully activated agmatinase [9]. The K m for agmatine (1.6 ± 0.1 m M ) and the K i for competitive inhibition by putrescine (12 ± 2 m M ), were also essentially equal for wild-type and D153N agmatinases. It is clear therefore that altered interactions with the substrate or a significantly altered affinity for the activating metal ion, are not the explanations for the greatly decreased catalytic activity of the D153N variant. To evaluate possible structural changes that may result from mutation, wild-type and D153N enzymes were com- pared by using fluorescence and CD spectrometry. The tryptophan emission fluorescence spectra were not altered by mutation (k max ¼ 342 nm), indicating that the environ- ment of tryptophan residues is essentially conserved in the mutant enzyme. On the other hand, the absence of major differences in the CD spectra of the wild-type and D153N agmatinases indicates that mutation had no effect on their respective secondary structures (Fig. 3). As an example, the percentage values calculated for the a-helix were 22.3 and 21.5% for wild-type and D153N enzymes, respectively. Therefore, based on the criteria used in this study, we may discard a gross structural change as the explanation for the lower catalytic activity of D153N agmatinase. Since an experimentally derived structure is not yet available for any agmatinase, to further evaluate the consequences of the Asp153fiAsn substitution, we used a modelled structure of the E. coli enzyme, constructed by using homology-modelling techniques and the 3D structure of the binuclear form of B. caldovelox arginase as a template. Principal attention was given to the modeling of the active site. The modelled structure was very similar to the template, with respect to the number and arrangement of structural elements (a-helix and b-sheets), and one of the major differences concerned the surface loops. However, as showninFig.4,B. caldovelox and rat liver arginases also Fig. 2. Effect of added Mn 2+ and EDTA on the catalytic activity of wild-type and D153N agmatinase. The enzymes were assayed before (Control) and after dialysis against 10 m M EDTA in 5 m M Tris/HCl (pH 7.5) for 4 h at 4 °C. Enzyme activities were measured with and without a previous incubation with 2 m M Mn 2+ for 20 min at 37 °C. Activities are expressed as percentage of the corresponding control not preincubated with Mn 2+ andassayedintheabsenceofaddedmetal ion. Fig. 1. Structural sequence alignment of B. caldovelox arginase (1cev), E. coli agma- tinase (AUH) and rat liver arginase (1rla). H, S, T and 3 stands for a-helix, strand, turn and 3 10 -helix. Conserved residues are marked by an asterisk. 5524 M. Salas et al. (Eur. J. Biochem. 269) Ó FEBS 2002 differed in these areas. A more specific difference concerned a loop located at the entrance of the active site cleft, and defined by residues 124–141 in the sequence of the bacterial arginase. Whereas the backbone and some of the side chains are very precisely conserved in both arginases, the agma- tinase loop was shorter when compared with the same region of the arginases (Fig. 4). Since the arginase loop contains residues that interacts with the a-carboxylate group of the substrate arginine [13], and this is the part of the molecule that makes arginine different from agmatine, it seems reasonable to assume that differences in this loop area are key factors in determining the difference in substrate specificity between arginase and agmatinase. This aspect is presently under investigation in our laboratory. In any case, despite the structural differences between agmatinase and arginase, agmatine and arginine were fixed in essentially the same position in the corresponding active site. The same position for the scissile guanidine carbon of the substrates, with respect to the metal ions and conserved, catalytically important residues, agree with a similar, if not identical, mechanism for both enzymes. Upon replacement of Asp153 with asparagine, the whole topology of the active site was found to be conserved in agmatinase. In the modelled active site structure, the side chains of Asp153 and Asn153 can be accommodated at essentially the same position, with the whole distance network remaining almost intact (Fig. 5). Moreover, one of the carboxylate oxygens of Asp153 and the carboxamide oxygen of Asn153 are positioned in such a way as to allow metal coordination interaction with one of the manganese ions. This, together with the fact that the positions of other Fig. 4. A superimposition of the structures of B. caldovelox arginase (red), rat liver arginase (yellow) and the modelled structure of E. coli agmatinase (blue). Note the shorter extension, in the case of E. coli agmatinase, of the loop indicated by the letter a. Fig. 5. Scheme of the binuclear manganese cluster and the localization of the side-chains of Asp153 and Asn153 in the modelled structures of wild-type and D153N mutant forms of E. coli agmatinase. Average distances (in A ˚ ) are indicated by the numbers. For simplicity, other active site residues, including other metal ligands, are not indicated. Fig. 3. CD spectra of wild-type (solid line) and D153N mutant (dotted line) E. coli agmatinases. The far-UV CD spectra were recorded at 22 °C. Ó FEBS 2002 A critical role for Asp153 in E. coli agmatinase (Eur. J. Biochem. 269) 5525 potential metal ligands are not substantially affected by the mutation, as revealed by the conservation of the whole topology of the active site, would explain the essentially unaltered interaction of the D153N variant with the metallic cofactor. In the modelled structure, the noncoordinated carboxy- late oxygen atom of Asp153 is within hydrogen-bonding distance of the metal-bound water molecule. This is interesting if one considers a catalytic mechanism involving a nucleophilic attack of a metal-bound hydroxide on the guanidinium carbon of agmatine [8,9]. By donating an hydrogen bond to Asp153, the nucleophile would be stabilized and oriented for optimal catalysis. On this basis, the disruption of this stabilizing hydrogen bond, due to the replacement of the carboxylic oxygen by the amide nitrogen, would be expected to result in a less efficient catalysis by the metal-bound hydroxide in the D153N mutant enzyme. The low, but significant catalytic activity of the mutant indicates that, even in the absence of the hydrogen bond to the noncoordinating carboxyl oxygen, the metal-bound hydroxide could still serve as a catalytic nucleophile, although considerably less efficiently. Further studies are, evidently, required to clarify this aspect. The proposed role for the critical Asp153 in E. coli agmatinase reinforces the relationships between this enzyme and the evolutionary related arginase. It would be, thus, of interest to examine the effects of the corresponding aspartate to asparagine mutation in this enzyme. ACKNOWLEDGEMENTS This research was supported by Grant 2990049 from FONDECYT and P.I. 98.031.076-1.0 from the Direccio ´ ndeInvestigacio ´ n, Universidad de Concepcio ´ n. We are greateful to Dr Enrique Pe ´ rez Paya (Universidad de Valencia, Espan ˜ a) for assistance with the CD spectra. REFERENCES 1. Satishchandran, C. & Boyle, S.M. (1986) Purification and prop- erties of agmatine ureohydrolyase, a putrescine biosynthetic enzyme in Escherichia coli. J. Bacteriol. 165, 843–848. 2. Buch, J.K. & Boyle, S.M. (1985) Biosynthetic arginine decarboxylase in Escherichia coli is synthesized as a precursor and locatedinthecellenvelope.J. Bacteriol. 163, 522–527. 3. Gilad, G.M., Wollam, Y., Iaina, A., Rabey, J.M., Chernihovsky, T. & Gilad, V.H. (1996) Metabolism of agmatine into urea but not into nitric oxide in rat brain. Neuroreport 7, 1730–1732. 4. Iyer,R.K.,Kim,H.K.,Tsoa,R.W.,Grody,W.W.&Cederbaum, S.D., (2002) Molecular cloning and characterization of human agmatinase. Genet. Metabol. 75, 209–218. 5. Mistry, S.K., Burwell, T.J., Chambers, R.M., Rudolph-Owen, L., Spaltmann, F., Cook, W.J. & Morris, S.M. (2002) Cloning of human agmatinase. An alternate path for polyamine synthesis induced in liver by hepatitis B virus. Am. J. Physiol. Gastrointest Liver Physiol. 282, G375–G381. 6. Ouzounis, C.A. & Kyrpides, N.C. (1994) On the evolution of arginases and related enzymes. J. Mol. Evol. 39, 101–104. 7. Perozich, J., Hempel, J. & Morris, S.M. (1997) Roles of conserved residues in the arginase family. Biochim. Biophys. Acta 1328, 23–37. 8. Kanyo, Z.F., Scolnick, L.R., Ash, D.E. & Christianson, D.W. (1996) Structure of a unique binuclear manganese cluster in argi- nase. Nature 382, 554–557. 9. Carvajal, N., Lo ´ pez,V.,Salas,M.,Uribe,E.,Herrera,P.&Cerpa, J. (1999) Manganese is essential for catalytic activity of Escherichia coli agmatinase. Biochem. Biophys. Res. Commun. 258, 808–811. 10. Carvajal, N., Olate, J., Salas, M., Uribe, E., Lo ´ pez, V., Herrera, P. & Cerpa, J. (1999) Chemical modification and site-directed mutagenesis of human liver arginase: evidence that the imidazole group of histidine-141 is not involved in substrate binding. Arch. Biochem. Biophys. 371, 202–206. 11. Cavalli, R.C., Burke, C., J., Kawamoto, S., Soprano, D.R. & Ash, D.E. (1994) Mutagenesis of rat liver arginase expressed in Escherichia coli: role of conserved histidines. Biochemistry 33, 10652–10657. 12. Carvajal, N., Olate, J., Salas, M., Lo ´ pez, V., Cerpa, J., Herrera, P. & Uribe, E. (1999) Evidence that histidine-163 is critical for cata- lytic activity, but not for substrate binding to Escherichia coli agmatinase. Biochem. Biophys. Res. Commun. 264, 196–200. 13. Bewley, M.C., Jeffrey, P.D., Patchett, M.L., Kanyo, Z.F. & Baker, E.N. (1999) Crystal structures of Bacillus caldovelox arginase in complex with substrate and inhibitors reveal new insights into activation, inhibition and catalysis in the arginase superfamily. Structure 7, 435–448. 14.Ash,D.E.,Cox,J.D.&Christianson,D.W.(2000)Arginase: a binuclear manganese metalloenzyme. Met Ions Biol. Syst. 37, 407–428. 15. Christianson, D.E. & Cox, J.D. (1999) Catalysis by metal-acti- vated hydroxide in zinc and manganese metalloenzymes. Annu. Rev. Biochem. 68, 33–57. 16. Vockley, J.G., Tabor, D.E., Kern, R.M., Goodman, B.K., Wiss- man,P.B.,Kang,D.S.,Grody,W.W.&Cederbaum,S.D.(1994) Identification of mutations (D128G, H141L) in the liver arginase gene of patients with hyperargininemia. Hum. Mutat. 4, 150–154. 17. Ash, D.E., Scolnick, L.R., Kanyo, Z.F., Vockley, J.G., Ceder- baum, S.D. & Christianson, D.W. (1998) Molecular basis of hyperargininemia: structure–function consequences of mutations in human liver arginase. Mol. Genet. Metabol. 64, 243–249. 18. Archibald, R.M. (1945) Colorimetric determination of urea. J. Biol. Chem. 157, 507–518. 19. Bradford, M.M. (1976) A rapid and sensitive method for the quantitation of microgram quantities of protein utilizing the principle of protein-dye binding. Anal. Biochem. 72, 248–254. 20. Ho, S.N., Hunt, H.D., Horton, R.M., Pullen, J.K. & Pease, L.R. (1989) Site-directed mutagenesis by overlap extension using the polymerase chain reaction. Gene 77, 51–59. 21. Vriend, G. (1990) WHAT IF: a molecular modelling and drug design program. J. Mol. Graph. 8, 52–56. 5526 M. Salas et al. (Eur. J. Biochem. 269) Ó FEBS 2002 . Insights into the reaction mechanism of Escherichia coli agmatinase by site-directed mutagenesis and molecular modelling A critical. of E. coli agmatinase (blue). Note the shorter extension, in the case of E. coli agmatinase, of the loop indicated by the letter a. Fig. 5. Scheme of the