B. A. Neilan, School of Biotechnology and Biomolecular Sciences, University of New South Wales, Sydney, NSW 2052, Australia Fax: +61 2 9385 1591 Tel: +61 2 9385 3235 E-mail: firstname.lastname@example.org
We report the first characterization of an l-arginine:glycine amidinotransferase from a prokaryote. The enzyme, CyrA, is involved in the pathway for biosynthesis of the polyketide-derived hepatotoxin cylindrospermopsin from Cylindrospermopsisraciborskii AWT205. CyrA is phylogenetically distinct from other amidinotransferases, and structural alignment shows differences between the active site residues of CyrA and the well-characterized human l-arginine:glycine amidinotransferase (AGAT). Overexpression of recombinant CyrA in Escherichia coli enabled biochemical characterization of the enzyme, and we confirmed the predicted function of CyrA as an l-arginine:glycine amidinotransferase by 1H NMR. As compared with AGAT, CyrA showed narrow substrate specificity when presented with substrate analogs, and deviated from regular Michaelis–Menten kinetics in the presence of the non-natural substrate hydroxylamine. Studies of initial reaction velocities and product inhibition, and identification of intermediate reaction products, were used to probe the kinetic mechanism of CyrA, which is best described as a hybrid of ping-pong and sequential mechanisms. Differences in the active site residues of CyrA and AGAT are discussed in relation to the different properties of both enzymes. The enzyme had maximum activity and maximum stability at pH 8.5 and 6.5, respectively, and an optimum temperature of 32 °C. Investigations into the stability of the enzyme revealed that an inactivated form of this enzyme retained an appreciable amount of secondary structure elements even on heating to 94 °C, but lost its tertiary structure at low temperature (Tmax of 44.5 °C), resulting in a state reminiscent of a molten globule. CyrA represents a novel group of prokaryotic amidinotransferases that utilize arginine and glycine as substrates with a complex kinetic mechanism and substrate specificity that differs from that of the eukaryotic l-arginine:glycine amidinotransferases.
Cyanobacterial toxins pose a serious health risk for humans and animals when they are present at hazardous levels in bodies of water used for drinking or recreational purposes. Under eutrophic conditions, cyanobacteria tend to form large blooms, which drastically promote elevated toxin concentrations. The problem is global, as most toxic cyanobacteria have a worldwide distribution [1–7]. The major toxin produced by the genus Cylindrospermopsis is cylindrospermopsin, which was first discovered after a poisoning incident on Palm Island (Queensland, Australia) in 1979, when 148 people, mainly children, were hospitalized with hepatoenteritis caused by contamination of a drinking water reservoir with Cylindrospermopsis raciborskii [8,9]. Cylindrospermopsin has hepatotoxic, nephrotoxic and general cytotoxic effects [10–12], and is a potential carcinogen . Besides C. raciborskii, five other cyanobacterial species have so far been shown to produce the toxin; they are Aphanizomenon ovalisporum, Umezakia natans, Rhaphdiopsis curvata, Aphanizomenon flos-aquae and Anabaena bergii [4,14–18].
Cylindrospermopsin is a polyketide-derived alkaloid with a central guanidino moiety and a hydroxymethyluracil attached to the tricyclic carbon skeleton  (Fig. 1). Putative cylindrospermopsin biosynthesis genes have been identified in A. ovalisporum  and C. raciborskii [18,21], and this led to the sequencing of the complete gene cluster (cyr) in an Australian isolate of C. raciborskii . The cyr gene cluster spans 43 kb and encodes 15 ORFs. On the basis of bioinformatic analysis of the gene cluster and isotope-labeled precursor feeding experiments , a putative biosynthetic pathway has been proposed . The first step in this proposed pathway is the formation of guanidinoacetate by the amidinotransferase CyrA. The nonribosomal peptide synthetase/polyketide synthase hybrid CyrB, followed by the polyketide synthases CyrC–F, then catalyze five successive extensions with acetate to form the carbon backbone of cylindrospermopsin. The biosynthesis is completed by formation of the uracil ring (CyrG–H), and tailoring reactions, such as sulfotransfer (CyrJ) and hydroxylation (CyrI).
Amidinotransferases catalyze the reversible transfer of an amidino group from a donor compound to the amino moiety of an acceptor . To date, l-arginine:glycine amidinotransferases from vertebrates and plants [25–28], an l-arginine:lysine amidinotransferase from Pseudomonas syringae [29,30], and the l-arginine:inosamine phosphate amidinotransferase (StrB) from Streptomyces species  have been described. More recently, another cyanobacterial amidinotransferase, SxtG, was discovered when the gene cluster for the biosynthesis of the neurotoxin saxitoxin in C. raciborskii T3 was sequenced . Amidinotransferases are a monophyletic group of enzymes with highly conserved sequences across distantly related organisms . They are key enzymes in the synthesis of guanidino compounds, which play an important role in vertebrate energy metabolism and in secondary metabolite production by higher plants and prokaryotes [24,27,30,34]. The best studied amidinotransferases are l-arginine:inosamine phosphate amidinotransferase (EC 220.127.116.11; StrB1) involved in the biosynthesis of the antibiotic streptomycin in the soil bacterium Streptomyces griseus , and l-arginine:glycine amidinotransferase (EC 18.104.22.168) involved in creatine biosynthesis in vertebrates . In cylindrospermopsin biosynthesis, the amidinotransferase CyrA is thought to catalyze the formation of guanidinoacetate, which suggests transamidination from arginine onto glycine in a manner similar to the vertebrate l-arginine:glycine amidinotransferase. Glycine and guanidinoacetate were confirmed as precursors in cylindrospermopsin biosynthesis by isotope-labeled precursor feeding experiments; however, incorporation of labeled arginine could not be confirmed, indicating an amidino group donor other than arginine . On the other hand, modeling of the active site of the CyrA homolog AoaA from A. ovalisporum, based on the crystal structure of AGAT, suggested the involvement of arginine as a possible substrate . Biochemical characterization of the enzyme is required to resolve this contradiction and identify the starting compounds for toxin production. Characterization of enzymes from the cylindrospermopsin pathway is also necessary to confirm the suggested mechanism for toxin production, as none of the cylindrospermopsin-producing organisms identified so far are amenable to genetic modification. In this article, we describe the cloning, purification and characterization of a novel amidinotransferase from C. raciborskii AWT205, in order to better understand the structure–function–stability relationship of this enzyme, which is responsible for the first step in the biosynthesis of a cyanotoxin.
CyrA is phylogenetically distinct from known amidinotransferases
To investigate the molecular phylogeny of CyrA within the amidinotransferase subfamily, an alignment of CyrA with 27 sequences spanning 376 residues was constructed. These sequences included representative proteins of the amidinotransferase subfamily, as well as uncharacterized genes annotated as ‘amidinotransferase’ from genome sequencing projects. A phylogenetic tree was constructed from the alignment (Fig. 2). The amidinotransferases fell into three major groups (groups 1–3) that were supported by high bootstrap values. Group 3 comprised StrB proteins from the prokaryote Streptomyces; these were only distantly related to other amidinotransferases. Group 2 encompassed two distinct subgroups. CyrA and the homolog AoaA from the cylindrospermopsin producer A. ovalisporum formed subgroup V. Subgroup IV in group 2 consisted of several experimentally uncharacterized (hypothetical) prokaryotic amidinotransferases that have been annotated as ‘glycine amidinotransferase’ (Fig. 2). CyrA is the first member of the phylogenetic group 2 amidinotransferases to be described experimentally.
Group 1 consisted of the eukaryotic l-arginine:glycine amidinotransferase in subgroup I and two prokaryotic enzymes in subgroup II. Subgroup III comprises the cyanobacterial amidinotransferases (SxtG) putatively involved in the biosynthesis of saxitoxin , together with one uncharacterized amidinotransferase from Beggiatoa.
Sequence analysis of CyrA reveals two active site substitutions
A structural alignment of CyrA and StrB1 with the well-characterized AGAT (Fig. S1) revealed that Asp254 and His303 (numbered according to the human protein), constituting part of the catalytic triad in the human and Streptomyces enzymes, are conserved in CyrA. The same applies to the active site Cys407, which was shown to form a covalent amidino–enzyme intermediate with the substrate’s amidino group. However, Met302, involved in arginine binding in AGAT, has been replaced by Ser247 in CyrA. A similar substitution has been reported in the ortholog AoaA from A. ovalisporum . Furthermore, Asn300, which contributes to the active site structure in AGAT, is replaced by Phe245 in CyrA.
Physicochemical properties of CyrA
The native cyrA gene is 1176 bp long and codes for a protein of 391 residues with a calculated molecular mass of 45.68 kDa and a theoretical pI of 5.1. Recombinant CyrA includes the N-terminal His6-fusion tag and 22 additional C-terminal vector-encoded amino acids, which increase the calculated molecular mass to 50.12 kDa and the pI to 5.6.
Yields of purified recombinant protein varied from 10.5 to 18.5 mg per liter of culture. After purification by immobilized metal ion affinity chromatography, recombinant CyrA was judged to be of >95% purity by SDS/PAGE (Fig. S2), and had the expected molecular mass of 50 kDa, as indicated by SDS/PAGE (Fig. S2), MALDI-TOF MS and LC-MS (Fig. S3). The presence of the His6-fusion tag and the identity of the purified protein as CyrA were confirmed by western blotting, MS intact mass analysis and peptide mass fingerprinting after enzymatic digestion (Table S1). The tryptic peptides covered 69% of the amino acid sequence of recombinant CyrA, including the N-terminal and C-terminal peptides, showing that the protein was expressed in its complete, nontruncated form.
Purified CyrA eluted from the size exclusion chromatography column in two peaks corresponding to molecular masses of 185 and 98 kDa (Fig. S2). SDS/PAGE analysis combined with activity assays confirmed that both peaks consist exclusively of CyrA. This indicated that CyrA is present in two forms, dimer and tetramer. Size exclusion chromatography was repeated four times with similar results, implying that the equilibrium between dimeric and tetrameric forms of CyrA is stable and reproducible under these conditions.
Amidinotransferase activity was found to be linear over a time period of 60 min in the presence of 20 mm l-arginine and 20 mm glycine, as well as a linear function of enzyme concentration. The plot of amidinotransferase activity at various pH values is bell-shaped (Fig. S4A). The highest activity of CyrA was detected at pH 8.5. At pH 7, only 25% of the original activity remained. For l-arginine:glycine amidinotransferases from pig, rat and soybean, pH optima of 7.5, 7.4 and 9.5, respectively, have been reported [25,35,36]. The optimum temperature (Topt) for CyrA was found to be 32 °C at pH 8. At 40 °C, 80% of the activity relative to Topt was lost (Fig. S4B). The Topt for soybean amidinotransferase was determined to be 37 °C .
Analysis of end-products confirmed CyrA as an l-arginine:glycine amidinotransferase
Isotope-labeled precursor feeding experiments confirmed glycine and guanidinoacetate as precursors for cylindrospermopsin biosynthesis, but could not confirm incorporation of ubiquitously labeled arginine into cylindrospermopsin . However, the transamidination of glycine from arginine by amidinotransferase, yielding guanidinoacetate, is common in vertebrates, and it was proposed that CyrA catalyzes the same reaction. All characterized amidinotransferases use arginine as the natural amidino group donor. In order to prove that the reaction catalyzed by CyrA converted l-arginine and glycine to ornithine and guanidinoacetate, 1H-NMR spectroscopy was used. Initially, several attempts were made to follow the reaction progress by NMR spectroscopy; however, the buffer component dithiothreitol and its oxidized form obscured key resonances. Therefore, the basic products (and reactants) were isolated by anion exchange chromatography prior to 1H-NMR. The presence of ornithine and guanidinoacetate was confirmed by the appearances of resonances for the α and δ protons of ornithine at 3.52 and 3.02 p.p.m., respectively, and the sharp single resonance for guanidinoacetate at 3.8 p.p.m. (Fig. 3). These assignments were confirmed by 1H–13C correlation spectroscopy and 1H–1H COSY spectra (data not shown).
CyrA has narrow substrate specificity
Apart from glycine and arginine, several structurally related compounds were tested for their ability to serve as substrates for CyrA. l-Homoarginine, agmatine, l-canavanine, guanidine hydrochloride, urea, γ-guanidinobutyric acid and β-guanidinoproprionic acid were tested as amidino group donors. l-Alanine, β-alanine, γ-aminobutyric acid, ethanolamine, taurine, l-lysine, α-amino-oxyacetic acid and l-norvaline were used as amidino group acceptors. The limit of detection for the assays was 0.5 mm hydroxyguanidine and 25 μm l-ornithine. Only incubation with hydroxylamine resulted in the detection of product. Therefore, it was concluded that CyrA only recognizes hydroxylamine as an amidino group acceptor. No other compound was an alternative substrate under these reaction conditions.
Kinetic analyses with natural substrates suggest a reaction mechanism different from that of other amidinotransferases
The formation of guanidinoacetate and ornithine from arginine and glycine obeyed regular Michaelis–Menten kinetics. Nonlinear regression analysis revealed kinetic constants as summarized in Table 1.
Table 1. Kinetic constants of CyrA.
a The values for human l-arginine:glycine amidinotransferase are given for comparison .
1.05 ± 0.05
kcat (min−1 per active site)
52.5 ± 2.5
K arginineM (mm)
3.5 ± 1.14
2.0 ± 0.5
K glycineM (mm)
6.9 ± 2.70
3.0 ± 1.0
In double-reciprocal plots with arginine as the varied substrate, the family of lines intersect to the left of the y-axis, below the x-axis (Fig. 4). This kinetic pattern is indicative of a random sequential mechanism, in which both substrates bind to the enzyme in a random order to form a compulsory ternary complex before the first product is released. The intercept below the origin suggests that binding of one ligand reduces the affinity for the other ligand .
Kinetic analyses with a non-natural acceptor reveal a complex kinetic mechanism
Initial reaction velocities for the formation of hydroxyguanidine and ornithine from hydroxylamine and arginine were measured over a wide range of hydroxylamine concentrations with a fixed concentration of arginine. The substrate versus velocity plot of these data revealed interesting features of the enzyme. First, the plot curves downwards (Fig. S5), suggesting substrate inhibition at high concentrations of hydroxylamine. Second, the plot is not a rectangular hyperbola but is sigmoidal, indicating allosteric behavior in the presence of hydroxylamine. The Hill constant (n) of 1.6 indicated positive cooperativity, with hydroxylamine binding to at least one peripheral site in addition to the active site. The theoretical maximum Hill constant for positive cooperativity is equal to the oligomeric state of the enzyme , i.e. either 2 or 4 for CyrA, which is an equilibrium of dimer and tetramer. Therefore, the Hill constant of 1.6 indicated a considerable to moderate cooperative effect of hydroxylamine.
Product inhibition also suggests a random sequential mechanism
A product inhibition study was conducted to further diagnose and confirm the kinetic mechanism of CyrA. Vertebrate l-arginine:glycine amidinotransferase display strong product inhibition by ornithine, with a Ki of 0.25 mm ; hence, it was speculated that CyrA might also be subject to inhibition by ornithine. Unfortunately, measurement of initial velocities in the presence of ornithine is not possible with the assay method of Van Pilsum et al. , which measures the formation of ornithine. Therefore, we measured initial reaction velocities at a saturating level of arginine and with varying noninhibitory concentrations of the non-natural acceptor hydroxylamine, in the presence of several fixed concentrations of ornithine, using the method of Walker . On a double-reciprocal plot of the data, the lines intercept in the upper right quadrant of the plot (Fig. 5). Such a kinetic pattern is characteristic of partial mixed inhibition . Ornithine therefore binds to the active site of CyrA at a binding site distinct from the hydroxylamine-binding site. This binding affects the rate of reaction by factor β, causing the noncompetitive component of the mixed inhibition. In addition, binding of ornithine to this distinct site also alters the affinity for hydroxylamine by factor α. This is most likely attributable to structural changes of CyrA induced by the binding of ornithine. The location of the common intercept in mixed-type inhibition systems depends on the actual and relative values of α and β. An intercept in the upper right-hand quadrant of the double-reciprocal plot, as is the case here (Fig. 5), indicates that β >> α .
The product inhibition study revealed another detail of this highly dynamic protein. The presence of ornithine not only has an inhibitory effect but also affects the affinity constant of hydroxylamine, modifying the allosteric behavior. The Hill constants for the individual series of velocity measurements in the presence of different ornithine concentrations ranged from 1.6 in the absence of ornithine to 2.1 and 2 in the presence of 3 and 6 mm ornithine, respectively (Fig. S6).
Analysis of reaction products with only the amidino group donor
In order to differentiate between a random sequential mechanism (both arginine and glycine must bind before ornithine is released) and a possible ping-pong mechanism (formation of an enzyme–amidino intermediate and release of ornithine in the absence of glycine), product formation by CyrA was investigated in the presence of arginine only.
CyrA incubated with arginine was subjected to MS and compared with CyrA that was not exposed to arginine in order to detect a possible enzyme intermediate by its difference in mass resulting from the bound amidino group (Fig. S3). CyrA samples were also digested with trypsin, endo-AspN or endo-LysC, and subjected to MALDI-TOF MS and LC-MS/MS (quadrupole time-of-flight) in order to identify the peptide fragment covalently linked to the amidino group (Table S1). However, an enzyme–amidino intermediate could not be detected.
GC-MS was employed to detect the reaction product ornithine in enzyme preparations that were incubated with arginine only. Ornithine was formed in the presence of only a single substrate, arginine, and its production therefore does not require the presence of the second substrate, glycine (Fig. S7). Incubation of 11 nmol of CyrA with 20 mm arginine produced only 81 nmol of ornithine in 1 h, which is equivalent to the slow rate of 0.0024 μmol·min−1·mg−1
CyrA is a thermolabile molten globule
Recombinant CyrA could be stored at – 80 °C in the presence of 20% glycerol for at least 2 months without significant loss of activity (>95% activity remaining). In contrast, total loss of activity occurred within 48 h when the enzyme was stored at 4 °C, despite the addition of reducing agents such as dithiothreitol, Tris(2-carboxyethyl) phosphine or β-mercaptoethanol. Similar observations were made for recombinant AGAT by Fritsche et al. . This prompted us to investigate the stability of CyrA in detail with the use of far-UV CD and fluorescence spectrophotometry to monitor the unfolding of secondary and tertiary structures, respectively. We compared fresh, active preparations of CyrA with samples that were inactive after storage at 4 °C for 2–4 days. We monitored the integrity of α-helical elements of active and inactive CyrA at 222 nm as a function of temperature, using far-UV CD (Fig. 6). For both active and inactive CyrA, an appreciable degree of secondary structure was still present after exposure to 94 °C, although active CyrA had greater preservation of secondary structure than inactive CyrA at all temperatures (Fig. 6A). There was a transition from higher to lower secondary structure for the active CyrA between 30 and 50 °C; however, the remaining structure was stable up to 94 °C. On the other hand, inactive CyrA did not show any transition, and seems to exist in a stable secondary structure conformation that is not affected at all by the increase in temperature. To confirm that the high ellipticity observed here represented α-helical elements that are stable at high temperatures, far-UV spectra of active and inactive CyrA were recorded in the presence and absence of urea as a denaturant. The addition of urea caused complete loss of ellipticity, confirming that the ellipticity was a result of secondary structure elements (Fig. 6B). Furthermore, the far-UV spectra for active and inactive CyrA were deconvoluted for the determination of relative amounts of α-helix and β-sheet. This revealed a shift of α-helical elements to β-strands upon formation of the inactive molten globule state, with a decrease in α-helix content from 19.9% to 11.4% and a concomitant increase in β-sheets from 27.5% to 34.2%.
As a significant degree of the secondary structure was retained at high temperatures, the unfolding of tertiary structure was investigated as the cause of the loss of observed enzyme activity. 8-Anilino-naphthalene-1-sulfonate (ANS) is a large hydrophobic molecule that is commonly used as a fluorescent probe of the hydrophobic surface exposed to solvent. The peak intensity of ANS fluorescence corresponds to the hydrophobic residues of a protein being maximally exposed, and the temperature at which this occurs is referred to as Tmax. The fluorescence melting curves of 0.1 mg·mL−1 active and inactive CyrA in the presence of 25 μm ANS as a function of temperature are shown in Fig. 7. ANS fluorescence in the presence of active CyrA showed a low intensity between 4 and 20 °C. This indicated a well-defined tertiary structure at low temperatures. The active CyrA also showed a sharp peak in intensity, with Tmax at 44.5 °C. Therefore, the tertiary structure loses integrity when the temperature is increased, leading to maximal exposure of the protein’s hydrophobic residues at ∼ 44 °C. In contrast, ANS fluorescence of inactive CyrA showed high intensity around 4 °C and Tmax at ∼ 10 °C, indicating exposure of hydrophobic residues to the solvent at these low temperatures. This demonstrates that inactive CyrA lacks a well-defined tertiary structure at any temperature.
From the experiments described above, it was clear that the loss of tertiary structure of CyrA stored at 4 °C is responsible for the loss of activity. CyrA rapidly loses its native tertiary structure when stored at 4 °C or when exposed to relatively mild temperatures (> 35 °C), with a concomitant retention of α-helical secondary structural elements. This state of the protein, when the tertiary structure has unfolded but the secondary structure remains intact, is reminiscent of a molten globule .
CyrA has optimum stability around neutral pH, in contrast to its alkaline pH activity optimum
In order to minimize loss of activity of CyrA during storage at 4 °C, we decided to investigate the stability of CyrA at different pH values and in the presence of NaCl, to identify conditions that would stabilize the enzyme. The stability of CyrA under these conditions was assessed by monitoring the unfolding of tertiary structure with ANS fluorescence. Fresh, active CyrA was exchanged into various buffers at 4 °C, and the Tmax was determined. ANS fluorescence of CyrA at pH 6.5, 7, 7.5 and 8.5 revealed defined peaks, with Tmax corresponding to 58, 54.5, 54 and 44.5 °C, respectively (Fig. 8). There was a clear trend towards increasing stability with a decrease in pH, with maximum stability around pH 6.5. At pH 6, a defined peak in fluorescence intensity was lacking, with maximum intensity around 4 °C indicating that the protein had already lost an appreciable amount of tertiary structure (data not shown). In contrast to stability, the activity of CyrA was found to be optimal at pH 8.5 (Fig. S4A). At the stability optimum (pH 6.5), CyrA retained only 10% of its activity as compared with pH 8.5. Therefore, the pH optimum for activity is not related to the stability optimum for this protein.
In the presence of 500 mm NaCl, Tmax at pH 7.5 decreased from 54 to 49 °C (Fig. S8), signifying a loss of stability at high ionic strength. This is an important consideration during purification procedures, as immobilized metal ion affinity chromatography buffers commonly have high ionic strength to minimize nonspecific interactions with the resin. We improved the purification and storage conditions of CyrA by reducing the NaCl concentration and lowering the pH of buffers to 7 when possible. Hence, knowledge of protein stability afforded optimization of protein purification and handling.
Cyanobacterial amidinotransferases play an important role in the biosynthesis of cyanotoxins such as cylindrospermopsin and saxitoxin; however, to date no amidinotransferase from cyanobacteria has been characterized. The data presented here indicate that the amidinotransferase from C. raciborskii AWT205 differs markedly from other known amidinotransferases with respect to its phylogeny, substrate specificity and kinetic mechanism. In addition, CyrA was found to be quite thermolabile, and existed in an intermediate state (molten globule) between its fully folded and unfolded states.
The phylogenetic analysis of CyrA showed that amidinotransferases fall into three different groups. Group 1 encompasses proteins from two different domains of life: eukaryotic enzymes involved in primary metabolism (subgroup I), and prokaryotic amidinotransferases involved in secondary metabolism (subgroups II and III). Surprisingly, the prokaryotic enzymes in groups 1 and 2 are more closely related to the eukaryotic l-arginine:glycine amidinotransferase (group 1) than to StrB from Streptomyces species (group 3). This close relationship between vertebrate l-arginine:glycine amidinotransferase and prokaryotic amidinotransferases is also illustrated by the fact that AGAT is regulated by end-product inhibition, a feature that is unusual in eukaryotic enzymes but common in prokaryotic enzymes [43,44].
Two uncharacterized proteins are present in group 1. The hypothetical protein from the enterobacterium Photorhabdus luminescens is closely related to l-arginine:lysine amidinotransferase (AmtA). Therefore, this enzyme might utilize an amidino group acceptor other than glycine, possibly lysine or a similar compound. An uncharacterized protein from Beggiatoa is annotated as AoaA in GenBank, but is more closely related to the SxtG amidinotransferase than to CyrA. Consequently, it seems unlikely that this enzyme represents a bacterial l-arginine:glycine amidinotransferase such as AoaA/CyrA.
CyrA clusters with group 2. The other amidinotransferases in this group are experimentally uncharacterized, but like CyrA and all other prokaryotic amidinotransferases discovered so far, these enzymes could also participate in secondary metabolite biosynthesis. Their participation in the primary metabolism (catabolic pathways) of arginine as a nitrogen, carbon or energy source seems unlikely, as the major enzymes utilized for arginine degradation in prokaryotes are arginase, arginine deiminase, arginine succinyltransferase and arginine oxidase [45,46]. The substrate specificity of CyrA could not be predicted from its phylogeny. The vertebrate l-arginine:glycine amidinotransferase (group 1) are not closely related to CyrA, despite their identical substrate specificity in vivo. This might reflect the difference in substrate use in vitro by CyrA and AGAT, with the stringent substrate specificity of CyrA being in stark contrast to the promiscuous behavior of AGAT. Furthermore, CyrA and SxtG are also phylogenetically distant, although both are involved in secondary metabolite biosynthesis in closely related or even the same species of cyanobacteria. Instead, SxtG is more closely related to AmtA. SxtG presumably utilizes an intermediate in saxitoxin biosynthesis as an amidino group acceptor. This compound (4-amino-3-oxo-guanidinoheptane) is structurally more similar to lysine, the substrate for AmtA.
As bioinformatic analysis yielded no relevant clues regarding the function of CyrA, we set out to biochemically characterize this enzyme. Arginase activity of overexpressed, purified CyrA was detected spectrophotometrically by following the formation of ornithine upon incubation with l-arginine and glycine. Although this indicated the utilization of l-arginine as a substrate by CyrA, as hypothesized, the question remained as to whether guanidinoacetate was a product of this reaction. Therefore, 1H-NMR analysis was carried out, and unambiguously identified the products of the reaction catalyzed by CyrA. This confirmed CyrA as the first prokaryotic l-arginine:glycine amidinotransferase to be described, and identified l-arginine and glycine as the starting units for cylindrospermopsin biosynthesis. Incorporation of the guanidino group of l-arginine could not be demonstrated in previous isotope-labeled precursor feeding experiments . This may be because not all cyanobacteria possess basic amino acid transporters [21,47].
CyrA shows allosteric behavior in the presence of hydroxylamine, resulting in positive cooperativity. Therefore, hydroxylamine might bind to a peripheral site on the enzyme, inducing a conformational change that causes activation by either increasing the affinity for the substrate or enhancing catalytic performance. Alternatively, the positive cooperativity could also be caused by the presence of multiple hydroxylamine molecules in the active site or by the oligomeric state of CyrA, e.g. because of differences in the Km values of the dimeric and tetrameric forms or cooperative binding of substrate to a neighboring active site. However, when the hydroxylamine concentration was increased, substrate inhibition was observed. This inhibition could either be kinetic (hydroxylamine binding to the wrong form of the enzyme) or allosteric (hydroxylamine binding to another peripheral site, which produces a conformational change that decreases activity). This allosteric site would have a lower affinity for hydroxylamine than the activating peripheral site, because it is occupied only at higher substrate concentrations.
Considering the high hydroxylamine concentrations tested, it is not surprising that allosteric and inhibitory effects were observed. Hydroxylamine was found to be a poor substrate for CyrA; activity was only detectable at concentrations above 20 mm. Substrate inhibition occurred at concentrations higher than 150 mm. At such high concentrations, a small polar molecule such as hydroxylamine would be expected to bind to additional sites on the protein.
It must be noted that substrate inhibition or allosteric effects were not observed with the natural substrates l-arginine and glycine at concentrations that exceed those in vivo (20 mm). Nevertheless, CyrA has characteristics of allosteric enzymes, such as a dynamic quaternary structure and ligand-induced conformational changes. Also, the flux of metabolites through biosynthetic pathways is often regulated at the first committed step in the pathway, which, for cylindrospermopsin biosynthesis, is catalyzed by CyrA. Ornithine inhibits CyrA and also affects its allosteric behavior in the presence of hydroxylamine. Therefore, it is possible that the activity of CyrA could be regulated in vivo by ornithine product inhibition.
The kinetic constants for CyrA were found to be similar to those of AGAT (Table 1). Similarly, the Km values for other mammalian and plant l-arginine;glycine amidinotransferase range from 1.8 to 9.21 mm for l-arginine and from 0.89 to 18 mm for glycine. Hence, the prokaryotic and eukaryotic l-arginine:glycine amidinotransferases have similar performances. However, the kinetic mechanism of CyrA in the presence of l-arginine and glycine as substrates differs from the well-established ping-pong mechanism of l-arginine:glycine amidinotransferase, as shown for the porcine  and human  enzymes. Initial velocity studies indicated a random sequential mechanism, and the noncompetitive inhibition of ornithine with respect to hydroxylamine confirmed this. Furthermore, the initial velocity study suggested that binding of one substrate reduces the affinity for the other. Similarly, the product inhibition study implied that binding of ornithine causes conformational changes that affect the binding of hydroxylamine, and therefore confirms the proposal that binding of one substrate/product affects binding of the other. Such ligand-induced, structural changes have been described for AGAT in the form of a ‘lid’ structure that opens and closes, regulating access to the active site. Binding of the large substrate/product (l-arginine/ornithine) to AGAT induces the open conformation of the lid, whereas binding of the small substrate/product (glycine/guanidinoacetate) induces the closed conformation .
In the classical random sequential mechanism, reaction products are not formed in the presence of only one substrate; however, here, the reaction product ornithine was formed in the presence of only one substrate, l-arginine, albeit in very low amounts and without the detection of an enzyme–amidino intermediate. Two explanations can reconcile these contradictory results. Water could act as the second substrate instead of glycine to accept the amidino group and produce ornithine. Although water is a weak nucleophile and CyrA has extremely stringent substrate specificity, this possibility cannot be excluded completely if one considers the high concentration of water (55 m), which will cause the reaction equilibrium to shift towards the formation of ornithine. Alternatively, an enzyme intermediate might have formed, as in a ping-pong mechanism, but be unstable, so that it decays to free enzyme and urea. For example, AGAT forms a covalent enzyme–amidino intermediate that is only stable at low pH [49,50]. Instability of the intermediate would make detection very difficult. The formation of product, ornithine, in the presence of only one substrate suggests that the reaction mechanism of CyrA is neither a classical sequential nor a ping-pong mechanism, but a hybrid of these two systems, in which an enzyme intermediate may be formed, but is not compulsory.
Many examples of enzymes that do not fall into the strict classification of sequential or ping-pong mechanisms, but lie somewhere in between these two systems, have been reported [51–57]. These studies show that it can be misleading to diagnose a kinetic mechanism on the basis of only initial velocity patterns, and recommend including additional experiments to confirm the mechanism. A hybrid ping-pong–random sequential mechanism fits all the data in this study, and helps to explain other features of CyrA, including its stringent substrate specificity. In such a mechanism, both substrates can bind to the enzyme simultaneously, but a partial reaction can still occur via formation of an enzyme intermediate. Therefore, the ternary complex of enzyme and both substrates is able to form but is not a requirement, as the reaction can also proceed as two partial reactions; hence the formation of product in the presence of only one substrate. Depending on conditions such as substrate concentration, the system will behave either like a rapid equilibrium random system or like a rapid equilibrium ping-pong system. Therefore, the system might appear as either a sequential or a ping-pong mechanism in initial velocity studies .
A hybrid ping-pong–random sequential mechanism also helps to explain the observed stringent substrate specificity of CyrA, because it postulates that there are distinct binding sites for each substrate. If the amidino group acceptor binds at the same site as the first substrate/product (l-arginine/ornithine), CyrA should be able to accept other compounds smaller than ornithine as amidino group acceptors. This is the case in l-arginine:glycine amidinotransferase which only have one substrate-binding site, giving rise to the classical ping-pong mechanism and acceptance of a wide range of substrates besides their physiological substrates [36,58]. However, CyrA does not accept l-alanine, β-alanine or ethanolamine as substrate. Even the smallest possible structures, such as hydroxylamine, are poor substrates for this enzyme. We therefore suggest that CyrA possesses a separate, specific binding site for the amidino group acceptor glycine. The most obvious difference between glycine, hydroxylamine and ethanolamine at physiologically relevant pH is the negative charge of glycine’s carboxyl group. It is likely that binding of glycine in CyrA’s active site is enhanced through ionic interaction with a charged residue of the enzyme.
To support this kinetic model, the level of conservation of residues involved in substrate binding and catalysis between AGAT and CyrA was investigated by structural alignment. The amino acids constituting the catalytic triad as identified in the crystal structure of AGAT (Asp254, His303 and Cys407) are conserved in CyrA (Asp197, His248 and Cys356). However, other amino acids located in the active site are substituted in CyrA, namely Asn300 and Met302 (AGAT numbering), which are replaced by Phe245 and Ser247 in CyrA. In particular, the replacement of a polar, hydrophilic amino acid (Asn300) with a large, nonpolar hydrophobic amino acid (Phe245) might explain the inability of CyrA to accept larger substrates. In StrB1, the Asn300 and Asn302 are replaced by the smaller amino acids alanine and threonine. This results in a much larger active site than in AGAT, and allows for binding of inosamine phosphate, the substrate of StrB1 .
CyrA was observed to be unstable at 4 °C. The low values for Tmax and the activity optimum are characteristic of cold-adapted, psychrophilic enzymes , in contrast to the host organism, which prefers growth at mesophilic temperatures of 25–28 °C . Investigations into the unfolding of the secondary and tertiary structures of the enzyme revealed a rapid loss of tertiary structure, whereas a substantial part of secondary structure persists in a native-like state upon storage at a temperature of 4 °C or higher. The inactive CyrA shows higher ANS fluorescence than the fully folded active CyrA. This is characteristic of the molten globule state, a protein folding intermediate that has less compact tertiary structure but substantial secondary structure [61,62]. Besides being an intermediate in protein folding, the molten globule state, with its increased structural flexibility, can also have biological functions. There are examples of molten globules that have an important role in the insertion of proteins into membranes and in binding of DNA or receptor molecules [63–65]. We cannot state whether CyrA exists as a molten globule state in vivo and has a particular function, or whether the molten globule is only observed as an in vitro folding intermediate. CyrA in the molten globule state is catalytically inactive, and CyrA is also not likely to bind to membranes or participate in molecular recognition.
In summary, CyrA represents a novel group of prokaryotic amidinotransferases that utilize arginine and glycine as native substrates, similarly to the vertebrate group of l-arginine:glycine amidinotransferases. The complex kinetic mechanism and substrate specificity of CyrA differ from those of the eukaryotic enzyme, and specific amino acid changes were identified that are likely to cause these differences. Further experiments to probe the mechanism and substrate binding of CyrA are underway, as are attempts to crystallize this enzyme for X-ray studies.
The amino acid sequence of native and recombinant CyrA (UniProt ID B0LI36) was submitted to the online tool protparam , accessed via the expasy proteomics server (http://au.expasy.org/tools/protparam.html), for calculation of the molecular mass, theoretical pI and molar extinction coefficient.
For phylogenetic analysis, amino acid sequences were aligned using the multiple sequence alignment tool clustalx 1.8 . The alignment was manually edited and submitted to the multiphyl webserver modelgenerator for substitution model selection [68,69]. Phylogenetic trees were constructed according to the maximum likelihood approach, using the LG+I+G+F substitution model  on the phyml webserver , with a bootstrap number of 100. The consensus tree was reproduced using the freeware figtree (http://tree.bio.ed.ac.uk/software/figtree/). Reference sequences were obtained from the Entrez Protein database, BRENDA and via psi-blast searches.
In addition, a structural alignment of CyrA and StrB1 (from Streptomyces griseus) against the crystal structure of AGAT was prepared using the online server fugue .
Cyanobacterial culture conditions
C. raciborskii AWT205  was grown in Jaworski medium  in static batch culture at 26 °C under continuous illumination (10 μmol·m−2·s−1).
Cloning, protein expression and purification
Cells from a 100 mL culture of C. raciborskii AWT205 were harvested by centrifugation (5000 g, 20 min), and genomic DNA was extracted from the cell pellet with the Ultra Clean Plant DNA Isolation Kit (MoBio, Carlsbad, CA, USA), according to the manufacturer’s instructions. PCR amplification of cyrA was performed using primers cyrA-F (5′-CATATGCAAACAGAATTGTAAATAGCT-3′) and cyrA-R (5′-CTCGAGAATAATGATGAAGCGAGAAAC-3′), which incorporated NdeI and XhoI restriction sites, respectively. The cyrA PCR product was cloned into the expression vector pET30a (Novagen, Madison, WI, USA) via pGEM-T Easy (Promega, Madison, WI, USA), verified by sequence analysis, and transformed into the Escherichia coli expression strain BL21(DE3) for expression as a C-terminal His6-tagged fusion protein. However, despite considerable efforts, the amidinotransferase could not be expressed in its soluble form. Therefore, cyrA was excised from pET30a and cloned into pET15b for expression with an N-terminal His6-tag. This resulted in the addition of 22 vector-encoded amino acids to the C-terminus of recombinant CyrA. For expression of recombinant His6-tagged CyrA, 1 L of MagicMedium (Invitrogen, Carlsbad, CA, USA) supplemented with 100 μg·mL−1 ampicillin was inoculated with a 2.5% overnight starter culture of E. coli BL21(DE3) harboring pET15-cyrA. Cultures were grown for 18–24 h at 30 °C before harvesting by centrifugation (8000 g, 15 min).
Cell pellets were subjected to one freeze–thaw cycle before suspension in one volume (w/v) of chilled lysis buffer (50 mm Hepes, pH 7, 500 mm NaCl, 5% glycerol, 0.5 mm dithiothreitol, 0.5 mm phenylmethanesulfonyl fluoride, 30 mm imidazole). Cells were lysed by sonication in the presence of lysozyme and DNase, and the lysate was cleared by centrifugation (16 000 g, 30 min, 4 °C). The lysate was then loaded onto a 1 mL HiTrap Chelating Column (GE Healthcare, Waukesha, WI, USA) charged with Ni2+ for purification of the recombinant protein by competitive elution with imidazole. The effluent collected from the column was subjected to SDS/PAGE  to identify fractions that contained pure recombinant protein. Such fractions were pooled, desalted with an Amicon Filtration Unit (10 kDa cut-off; Millipore, Billerica, MA, USA) and diluted in storage buffer (50 mm Hepes, pH 7, 5% glycerol, 10 mm dithiothreitol). Purified recombinant CyrA preparations were either used immediately or snap frozen in liquid nitrogen in the presence of 20% glycerol for storage at −80 °C. The protein concentration of purified CyrA preparations was determined with a spectrophotometer (NanoDrop Technologies, Wilmington, DE, USA), using the following relationship: A280 nm of 1.285 = 1 mg·mL−1.
Western blot analysis on total protein extract and on fractions of purified protein was performed with a nickel–nitrilotriacetic acid–alkaline phosphatase conjugate (Qiagen, Hilden, Germany), using 5-bromo-4-chloro-3-indolyl phosphate/Nitro Blue tetrazolium chloride (Sigma-Aldrich, St Louis, MO, USA). The protein band corresponding to 50 kDa on SDS/PAGE was excised, destained and extracted. The protein was analyzed by MALDI-TOF MS to determine the intact mass [75,76]. CyrA was then digested with trypsin and analyzed by MALDI-TOF MS [76–78]. The resulting data were analyzed with the mascot search engine (http://www.matrixscience.com/home.html).
Size exclusion chromatography for determination of native molecular mass
Size exclusion chromatography was performed on a Superdex 200 10/300 GL column (GE Healthcare) fitted to an ÄKTA Basic 900 series Fast Protein Liquid Chromatography System. The system was controlled by unicorn software, and the column was equilibrated in 50 mm Hepes (pH 7.5) and 112 mm NaCl at a flow rate of 0.4 mL·min−1, using a 0.2 mL sample loop. The column was calibrated with protein standards of known molecular mass, and approximately 100 μg of CyrA was applied. Eluted proteins were detected at 280 nm.
1H-NMR spectroscopy was employed to identify the products of CyrA upon incubation with arginine and glycine as substrates. 1H-NMR spectroscopy was performed at 600 MHz on a Bruker DMX 600 spectrometer (Bruker, Billerica, MA, USA) with a 5 mm QNP probe. Samples were dissolved in 5% D2O and run with the standard Bruker water suppression program.
To obtain sufficient amounts of product for analysis, a preparation of 10 mm l-arginine and 10 mm glycine in 100 mm Tris/HCl (pH 8.5), containing 1 mm dithiothreitol, was incubated overnight at 30 °C in the presence of 2 mg of CyrA in a total reaction volume of 10 mL. The reaction mixture was then applied to a Dowex 50 column (1 × 10 cm) in the H+ form. The column was washed with water (100 mL), and the basic products and reactants were eluted with 2 m NH3 (50 mL). After evaporation, the sample was dissolved in 5% D2O (500 μL) for NMR spectroscopy. Standards of l-arginine, glycine, ornithine and guanidinoacetate were run under identical conditions.
Amidinotransferase activity assays
The colorimetric assay for amidinotransferase activity was based on the specific reaction of ninhydrin with ornithine at low pH . Assays were carried out as previously described , with the following modifications: the total assay volume of 300 μL contained 50 mm Tris/HCl (pH 8.5), 20 mm l-arginine (donor), 20 mm glycine (acceptor) and 10–20 μg of enzyme. The reaction was carried out in 1.5 mL microcentrifuge tubes at 30 °C for 30 min. The absorbance at 505 nm was measured with a Cary 100 UV spectrophotometer (Varian, Palo Alto, CA, USA). Appropriate controls were included, consisting of reaction mixtures that were either stopped without incubation, contained one or no substrate, or contained boiled enzyme preparations. For accuracy, all assays were carried out in triplicate.
Various structurally related compounds were tested as potential amidino group acceptors instead of glycine, such as l-alanine, β-alanine, γ-aminobutyric acid, ethanolamine, taurine, l-lysine, α-amino-oxyacetic acid and l-norvaline, by measuring the amount of l-ornithine generated, as described above.
However, in order to test various amidino group donor analogs, a different colorimetric assay needed to be employed that relies on the ability of an amidinotransferase to amidinate the artificial substrate hydroxylamine to form hydroxyguanidine, which was measured after its reaction with pentacyanoaminoferrate . The donor analogs tested were l-homoarginine, agmatine, l-canavanine, guanidine hydrochloride, urea, γ-guanidinobutyric acid and β-guanidinoproprionic acid. The reaction mixture (45 μL) contained 100 mm Tris/HCl (pH 8.5), 90 mm test substrate, 200 mm hydroxylamine and 15 μg of purified CyrA. Assays were carried out for 45 min at 30 °C, as described previously . A standard curve for hydroxyguanidine–pentacyanoaminoferrate complex was constructed, and the absorbance at 490 nm was determined with a SpectraMax 340 microtiter plate reader (Molecular Devices, Sunnyvale, CA, USA). For this purpose, hydroxyguanidine was chemically synthesized as previously described . Purity of the synthesized compound was confirmed by spectral analysis from 400 to 560 nm.
The pH optimum of CyrA was determined by measuring the amount of ornithine formed at 30 °C at a pH range from 5.5 to 10 in the presence of 100 mm buffer (pH 5.5–6.5, Mes/NaOH; pH 7–7.4, K2HPO4/KH2PO4; pH 8–9, Tris/HCl; and pH 9.5–10, Ches/NaOH). The temperature optimum for CyrA was determined at pH 8 by measuring the amount of hydroxyguanidine formed at various temperatures, ranging from 15 to 61 °C.
In order to investigate the kinetic mechanism of CyrA, initial reaction rates (formation of ornithine) were measured over a range of l-arginine concentrations (2–24 mm) at fixed concentrations of glycine (3, 6, 9, 12, 16 and 20 mm). Kinetic constants were obtained by nonlinear regression analysis with the enzyme kinetics module 1.1 linked to sigmaplot 8.02. All kinetic analyses were repeated at least three times, with reproducible results.
Product inhibition study
Initial reaction rates were measured in the presence of different concentrations of ornithine (0, 1, 3, 6 and 15 mm) at a saturating level of l-arginine (50 mm), with hydroxylamine as the varied substrate (20–150 mm), using the method of Walker . Assays were performed in 45 μL with 25 μg of CyrA for 30 min at 30 °C in the presence of 100 mm Tris/HCl (pH 8).
MS analysis of the CyrA reaction sequence
CyrA (270 μg) was incubated in the presence of 18 mm l-arginine at 30 °C. After 2 h, the protein was separated from small molecular mass substrate and/or product with a Microcon YM-10 centrifugal filter device (Millipore). CyrA preparations treated in the same way but without the addition of l-arginine were used as the background control. The filtrate was subjected to electron capture negative ionization GC-MS for ornithine analysis, with an adaptation of a published method . A detailed description of this method can be found in Doc. S1.
The protein fraction was subjected to MALDI-TOF MS and ESI LC-MS for intact mass analysis. MALDI-TOF MS and LC-MS/MS were carried out after digestion with trypsin, endo-LysC or endo-AspN to obtain maximal sequence coverage. MS was performed according to previously described procedures [75–78], and further details can be found in Doc. S1.
Far-UV CD spectra in the range 190–260 nm and thermal unfolding curves at 222 nm in 100 mm Ches (pH 8.5) were generated with active CyrA and CyrA that showed no activity upon storage at 4 °C for ∼ 48 h. Spectra were recorded with a JASCO J-810CD Spectropolarimeter (Jasco, Easton, MD, USA) under constant nitrogen flow, connected to a Peltier temperature controller. The temperature was varied from 4 to 94 °C at a rate of 0.6 °C·min−1 to monitor unfolding at 222 nm. Far-UV spectra in the presence and absence of urea were recorded at 14 °C. The path length was 1 cm, and the protein concentration was 0.5 mg·mL−1. Spectra were corrected for buffer signal. Data are expressed as the mean residue ellipticity, [θ] (deg·cm2·dmol−1). The relative amounts of α-helix and β-sheet were deconvoluted from far-UV spectra for active and inactive CyrA with the online tool k2d2 .
Fluorescence spectra were collected using a Perkin Elmer LS50B Luminescence Spectrometer equipped with a Perkin Elmer PTP1 Peltier temperature programmer (Perkin Elmer, Waltham, MA, USA). In order to obtain thermal unfolding curves, the temperature was varied from 4 to 80 °C at a rate of 0.6 °C·min−1 (identical to the CD scan rate). The path length was 1 cm, and the protein concentration was kept at 0.1 mg·mL−1 in all experiments in which active and inactive CyrA were compared, to validate the comparison of fluorescence intensity between samples. The fluorescence of 25 μm ANS in the presence of active and inactive CyrA was monitored at 480 nm, with an excitation wavelength of 380 nm, in 50 mm Tris/HCl (pH 8.5). Additional thermal unfolding curves for active CyrA were obtained at pH 6 and 6.5 (50 mm Mes), pH 7 (50 mm Tris/HCl) and pH 7.5 (50 mm Hepes), as well as in the presence of 500 mm NaCl at pH 7.5. All spectra were corrected for buffer signal.
The authors would like to acknowledge O. Pilak for help with size exclusion chromatography and S. Murray for advice on phylogenetic analysis. This work was supported by The Australian Research Council. MS results were obtained at the Bioanalytical Mass Spectrometry Facility within the Analytical Centre of the University of New South Wales. This work was undertaken using infrastructure provided by NSW Government co-investment in the National Collaborative Research Infrastructure Scheme (NCRIS).