Pseudomonas aeruginosa Exopolyphosphatase Is Also a Polyphosphate: ADP Phosphotransferase

Pseudomonas aeruginosa exopolyphosphatase (paPpx; EC 3.6.1.11) catalyzes the hydrolysis of polyphosphates (polyP), producing polyPn−1 plus inorganic phosphate (Pi). In a recent work we have shown that paPpx is involved in the pathogenesis of P. aeruginosa. The present study was aimed at performing the biochemical characterization of this enzyme. We found some properties that were already described for E. coli Ppx (ecPpx) but we also discovered new and original characteristics of paPpx: (i) the peptide that connects subdomains II and III is essential for enzyme activity; (ii) NH4 + is an activator of the enzyme and may function at concentrations lower than those of K+; (iii) Zn2+ is also an activator of paPpx and may substitute Mg2+ in the catalytic site; and (iv) paPpx also has phosphotransferase activity, dependent on Mg2+ and capable of producing ATP regardless of the presence or absence of K+ or NH4 + ions. In addition, we detected that the active site responsible for the phosphatase activity is also responsible for the phosphotransferase activity. Through the combination of molecular modeling and docking techniques, we propose a model of the paPpx N-terminal domain in complex with a polyP chain of 7 residues long and a molecule of ADP to explain the phosphotransferase activity.


Introduction
polyP are linear polymers containing few to several hundred residues of orthophosphate linked by energy-rich phosphoanhydride bonds. The presence of this polymer has been detected in all kinds of living organisms, including higher organisms. Main enzymes involved in the synthesis of polyP are the polyphosphate kinases (Ppks; EC 2.7.4.1), which catalyze the formation of polyP from ATP (Ppk1) and GTP (Ppk2). Endo-and exopolyphosphatases are the most important enzymes involved in the utilization of polyP. In bacteria only exopolyphosphatases have been described [1].
The implication of Ppk and polyP in the virulence of P. aeruginosa has been clearly demonstrated, since a ppk1 knockout mutant was defective not only in various forms of motility [2,3] but also in biofilm development, quorumsensing, synthesis of virulence factors such as elastase and rhamnolipid, virulence of the burned-mouse pathogenesis model [4], and general stress and stringent responses [5]. However, little is known about the relation between Ppx and its possible role in the pathogenesis of this bacterium [3,6,7]. We have recently confirmed that Ppx is relevant for pathogenesis in P. aeruginosa [8], due to the fact that the ppx null mutant was defective in the production of factors associated to both acute infection (e.g., motilitypromoting factors, blue/green pigments production, and quorum-sensing C6-C12 homoserine lactones) and chronic infection (e.g., rhamnolipids and biofilm formation). Thus, there is enough evidence that both Ppk and Ppx as well as polyP balance contribute to the pathogenesis of P. aeruginosa.
Concentrations of polyP in P. aeruginosa are severalfold greater than in E. coli [9,10]. P. aeruginosa Ppx (paPpx) activity was described for the first time by Miyake and collaborators [11], and its gene was later cloned and expressed in E. coli [12]. The crystal structure and analysis of E. coli Ppx (ecPpx) have been reported [13,14] and the active site of 2 Enzyme Research the enzyme was suggested as a cleft where the polyP chain would bind. Several authors have determined that subdomains I and II of the N-terminal region of ecPpx represent the catalytic portion, whereas the C-terminal region, formed by subdomains III and IV, was proposed to be involved in substrate binding [13][14][15].
The role of polyP as phosphate donor in phosphoryl transfer reactions has also been described [16]. Polyphosphate AMP phosphotransferase transfers a P i from polyP to AMP producing ADP; afterwards, ATP is regenerated by the sequential action of the adenylate kinase. In addition, the glucose phosphotransferase catalyzes the phosphoryl transfer from polyP to glucose or glucosamine producing glucose 6-phosphate and glucosamine 6-phosphate, respectively. Finally, it is well known that Ppk is capable of catalyzing the inverse reaction to produce ATP from ADP + polyP (known as polyP:ADP phosphotransferase activity) (see [16] and references cited therein). To our knowledge, a similar activity of Ppx in prokaryotes has never been reported. Considering that other enzymes of the same family, such as the Ppx/GPPA of Aquifex aeolicus (aaPpx; EC 3.6.1.40), act on a nucleotide (pppGpp) [17], we hypothesized that in the active site of paPpx there could be enough space to bind a nucleotide molecule such as ADP and that the reaction of ATP production using polyP as P i donor could be possible. In the present work, we present results of the cloning and purification of paPpx, in full-length, N-terminal, and Cterminal domains variants and we describe the preference of the full-length enzyme for long polyP chains, pointing the Cterminal domain as responsible for this behavior. We show that paPpx is also a polyphosphate:ADP phosphotransferase and that the active site is the same as that one involved in the hydrolase activity. Finally, we present a structural model of full-length paPpx, in closed conformation, based on the atomic coordinates of ecPpx [13] and a model of the N-terminal domain of paPpx in an open state based on atomic coordinates of aaPpx [17]. We also propose a model of the paPpx N-terminal domain in complex with a polyP chain of 7 residues long and a molecule of ADP to explain the phosphotransferase activity through docking techniques.   ), 314 (N-paPpx  ), 303 (N-paPpx  ), and 192 (C-paPpx (315-506) ) amino acids of Ppx were amplified from P. aeruginosa PAO1 wild-type chromosomal DNA through PCR with the oligonucleotides listed in Table 1. PCR-amplified DNA fragments were cloned into the pCR2.1-TOPO vector to generate pCR-ppx, pCR-Nppx, pCR-N303ppx, and pCR-Cppx, respectively. These plasmids were later transformed into E. coli XL10-Gold, followed by selection of ampicillin-resistant transformants. For gene expression, the different amplified ppx fragments were restricted by EcoRI-NdeI enzymes and subcloned into pET-15b (Novagen) as N-terminal fusions to a 6xHis-tag, generating pET-ppx, pET-Nppx, pET-N303ppx, and pET-Cppx, respectively. These plasmids were then transformed into E. coli BL21-CodonPlus. The resulting transformants were grown and induced as previously described by [18]. Affinity purification on Ni-agarose columns was used to perform protein purification, following the manufacturer's protocol (the QIA expressionist, Qiagen). Pure recombinant proteins were dialyzed against 10 mM Tris-HCl, pH 8.0, 150 mM NaCl, 150 mM imidazole, and 30% glycerol. Consecutive dialysis steps were performed to reduce the imidazole concentration to approximately 10 mM. The 6xHis-tag was subsequently removed using a thrombin cleavage capture kit (Novagen), and protein dialysis was repeated.

Enzyme Activities and Protein
Assay. Ppx activity was measured after incubation at 37 ∘ C for 30 min in 200 L of 50 mM Tris-HCl buffer pH 8.0, 80 mM KCl, and 5 mM MgCl 2 . The substrates used were polyP 25 , polyP 45 , polyP 65 , and polyP 75 . The P i released after incubation was measured by the Katewa and Katyare method [19] with modifications. Briefly, 100 L of the reaction mixture was added to 400 L of a solution with 2.5% (NH 4 ) 6 Mo 7 O 24 ⋅(H 2 O) 4 in 3 N H 2 SO 4 and 400 L of 2% ascorbic acid/2% hydrazine in 0.1 N H 2 SO 4 , and the solution was brought to a final volume of 1200 L with triple glass-distilled water. Quantification of free P i was performed after 30 min of incubation at 37 ∘ C through measurement of the absorbance at 820 nm. One unit of exopolyphosphatase was defined as the amount of enzyme that releases 1 nmol of P i per minute at 37 ∘ C.
The phosphotransferase activity was measured after incubation at 37 ∘ C for 30  Probes). In sum, 50 L of the mixture described above was added to 450 L of 25 mM tricine pH 7.8, 5 mM MgSO 4 , 0.08 mM EDTA, 0.08 mM Na-azide, 10 mM DTT, 0.5 mM D-luciferin, and 10 g of firefly luciferase. The mix was incubated for 10 min and subsequently the emission spectra were measured between 500 and 650 nm, without excitement, with increments of 1 nm, an integration time of one second, and an emission slit of 10. The fluorescence measurements were performed in a Spex Fluoromax 3 spectrofluorometer (Jovyn-Ivon HORIBA). One unit of phosphotransferase was defined as the amount of enzyme that produces 1 nmol ATP per minute at 37 ∘ C. and max values were estimated by nonlinear fitting of initial rate data according to the ion concentrations, respectively. Protein concentration was determined by spectrophotometric measurement at 280 nm using the correspondent theoretical molar extinction coefficient calculated with the "ProtParam" tool [20] for physicochemical parameter prediction, which is available at the Expasy server (http://www.expasy.com/).

Molecular Modeling and Molecular
Dynamics. The search of paPpx homologues through the use of the BLASTp algorithm resulted in the identification of an ortholog protein in E. coli (ecPpx) with a 41% identity and 58% similarity. Two solved structures of this protein are available: PDB: 2FLO, 2.2Å resolution [14] and PDB: 1U6Z, 1.90Å resolution [13]. According to this, we used 1U6Z as template because the resolutions of the template had a large impact on the quality of the resulting model. Enzyme Research A homology model of paPpx in a closed conformation was constructed by comparative modeling using the ICM program [21]. For loop modeling, we performed a conformational sampling of these regions by means of the Sampling Loop module (Monte Carlo) implemented in ICM program.
The final model was validated using ProSA [22], ANOLEA [23], and PROCHECK [24]. Tautomeric states of histidine residues in the model were assigned according to the local environment using the Check Sidechains Plugin from VMD software [25]. The center of mass of the active site (residues Glu 126 , Asp 149 , Gly 151 , Ser 154 , and Glu 156 ) was measured with VMD software, and in its place an atom of Mg 2+ , which is an essential cofactor, was added prior to the molecular dynamics.
The model of the N-paPpx  in an open state was constructed using the atomic coordinates of PPX/GPPA phosphatase from A. aeolicus (aaPpx) in complex with the alarmone ppGpp [26] (PDB: 2J4R). Considering not only the low percentage of identity between paPpx and aaPpx [27] but also the homology in secondary structure, the model was constructed by threading using the "one-to-one threading" option of Phyre Server [27] (http://www.sbg.bio .ic.ac.uk/phyre2/). The model was obtained with 100% of confidence and 292 residues of a total of 314 were aligned. Tautomeric states of histidine residues in the model were assigned according to the local environment and Mg 2+ ion was located in the active site similarly as we described for the full-length model. The obtained models were subjected to MD calculations to reach a minimum energy state. For this, the models were embedded in a 15Å water box, and KCl 80 mM was used not only to mimic the optimal conditions for enzyme activity but also to neutralize the total charge of the system. The initial configuration of both systems was optimized using energy minimization followed by an equilibration through a molecular dynamics (MD) simulation in the NPT ensemble at 310 ∘ K for 1 ns, using a backbone restriction of 0.5 kcal/molÅ. NAMD program was used to perform all molecular dynamic simulations [28]. The electrostatic interactions were computed with no truncations using the particle mesh Ewald algorithm [29] under periodic boundary conditions.
The DYNDOM server was used to assess the opening degree model in open conformation with respect to the closed conformation [30]. By APBS software the potential electrostatic calculations were performed [31]. The charge and vdW radius assignment were determined with the software PDB2PQR [32] and the CHARMM force-field. pKa values were calculated using propKa [33].

Docking Assays.
Docking studies were carried out as previously described [34], defining the N-paPpx  model in open conformation as receptor. ICM [21] version 3.4 was used and the icmPocketFinder function was employed to detect possible binding sites with a tolerance of 4.6 by default. Tolerance is related to flexibility for sites prediction. The lower the tolerance value the higher the number of pockets predicted and vice versa. The value used was the one recommended by software developers.
Nucleotide ligand inputs were extracted from the Pub-Chem database (http://pubchem.ncbi.nlm.nih.gov/) and the polyP ligand input was constructed by using ICM molecule editor. All ligand charges were assigned by ICM software. All of the structures were protonated and optimized using standard ICM protocols. The thoroughness parameter, which represents the length of the docking simulation, was set at 2.0, as recommended by software developers when metals are present in the docking binding site. For the docking of ADP, the gridbox size was 16.95Å × 19.23Å × 16.05Å, and the center was located at the points (−5.602 × 7.844 × −4.172)Å. The binding site was determined by structural alignment with aaPpx and the region at which ppGpp is located was used. Thus, the docking pocket was composed of the residues Asn 25 , His 28 , Gly 152 , Gly 224 , Asp 272 , and Arg 274 . For the docking of polyP 7 , the N-paPpx-ADP complex was used as receptor. The binding site was one of those detected by the icmPocketFinder tool and was consistent with the S-shaped canyon described by [14].

Biochemical Characterization of paPpx
3.1.1. Cloning and Overproduction of paPPX Variants. The exopolyphosphatase gene of P. aeruginosa (ppx, PA5241) was cloned and overproduced as the full-length recombinant protein and also the two peptides representing the N-terminal and C-terminal domains. All recombinant proteins were purified and used to study some of the biochemical properties of the enzyme. The identification of N-and C-terminal domains was performed considering the crystallographic structure reported by [13,14]. The full-length protein comprised 506 aminoacyl residues (paPpx (1-506) ), while the N-terminal domain contained the first 314 aminoacyl residues and the remaining 192 aminoacyl residues corresponded to the C-terminal domain (N-paPpx  and C-paPpx (315-506) , resp.). The theoretical MW for paPpx  , N-paPpx  , and C-paPpx while C-paPpx (315-506) lacked enzymatic activity. The specific activities of paPpx  and N-paPpx (1-314) were 1.49 and 0.38 mol of P i min −1 mol −1 protein, respectively. In order to determine the fragment responsible for the enzymatic activity of paPpx, we took into consideration the N-terminal construct obtained by [15], and we produced an N-terminal variant formed by the first 303 aminoacyl residues, named N-paPpx  , which were found to be inactive.

Effect of the polyP Chain
Length on paPpx Activity. The kinetic behavior evaluated and apparent catalytic constants obtained for paPpx  and N-paPpx  are displayed in Figure 1 and summarized in Table 2. With saturating Mg 2+ and K + concentrations, paPpx  increased its affinity for polyP in the following order: polyP 25 < polyP 45 < polyP 65 < polyP 75 . The decreasing (app) values were accompanied by a sharp rise in the catalytic efficiencies, with this behavior being more noticeable with the increment in the length of the substrate chain (Figure 1(a)). This trend did not occur with N-paPpx since both (app) and catalytic efficiency were similar independently of the length of the polyP chain (Figure 1(b), Table 2). Indeed, the analysis of the results obtained with polyP 75 demonstrated that the full-length enzyme presented  of complementation, where a mixture of N-paPpx (1-314) /C-paPpx (315-506) , in a 1 : 10 ratio, produced an active enzyme with higher affinity for the substrate polyP 75 . The (app) value measured in the mix was 11 ± 2 M, which represented an increase of ≈2.3-fold.

Ion Dependence of Ppx Activity
Divalent Ions Dependence. We studied the behavior of the full-length paPpx  and N-paPpx  against different concentration of divalent ions such as Mg 2+ , Zn 2+ , Ca 2+ , and Mn 2+ as effectors, in presence of a saturating concentration (8 M) for the substrate polyP 65 . The activation of both enzyme variants by Mg 2+ was similar and showed no inhibition at high concentrations of this ion (Figure 1(c)). This result differed from the one obtained by [36,37] for the ecPpx activity that found a sharp decrease in the activity with Mg 2+ concentrations of 1 mM and higher. The values of 0.5(app)Mg 2+ in paPpx  and N-paPpx  were 0.30 ± 0.02 mM and 0.28 ± 0.02 mM, respectively. Zn 2+ was able to activate the enzyme only 20% compared to Mg 2+ , whereas the activation by Ca 2+ and Mn 2+ was negligible (3% and 2%, resp.). These data, added to the fact that the interaction between paPpx  and Mg 2+ occurs in the N-terminal domain, showed a clear preference of the enzyme for Mg 2+ without inhibition by ion concentration up to 10 mM.
As expected, K + was a nonessential activator of paPpx. This result is in good agreement with other studies performed on Ppxs [11,15]. To assess the net effect of Mg 2+ , we performed a saturation curve of the divalent ion with the full-length enzyme, with and without K + . The presence of K + did not affect the affinity of the enzyme for Mg 2+ ; ( 0.5(app)Mg 2+ values were similar: 0.30 ± 0.02 mM (K + ) and 0.33 ± 0.01 mM (no K + )).

Monovalent Ions Dependence.
Considering the activation produced by K + in the activity of paPpx  , we decided to test the effect of other monovalent ions. We observed that Li + and Na + presented no effects on enzyme activity while NH 4 + , K + , Rb + , and Cs + were activators of paPpx  .
Taking into account the physiological relevance of NH 4 + and K + , saturation curves of paPpx with these ions were performed in the presence of 8 M of polyP 65 and 5 mM of Mg 2+ . The curves obtained with NH 4 + and K + were sigmoid and reached their maximum activity at concentrations of 30 mM and 80 mM, respectively (Figure 1(d)). 0.5(app)NH 4 + was 10 ± 0.4 mM and 0.5(app)K + was 42 ± 0.5 mM for K + .
In view of the activation of paPpx  by NH 4 + , we tested alkylammonium ions with different degrees of methylation as activators. We found that activation decreased as the number of methyl substituents increased. Considering the NH 4 + activation as 100%, the percentages of activity with methylamine, dimethylamine, and trimethylamine were of 28, 9, and 5%, respectively. Tetramethylammonium was not an activator of paPpx  .

Phosphotransferase Activity.
Since the Ppx/GPPA from A. aeolicus acts on a nucleotide (pppGpp) and the structure of this enzyme in presence of the product (ppGpp) (PDB: 2J4R) is available [17,26], we hypothesized that in the active site of paPpx there could be enough space to bind a nucleotide molecule such as ADP; thus we decided to test the paPpx as a polyphosphate:ADP phosphotransferase. The production of ATP from polyP and ADP was measured in paPpx  and N-paPpx  with polyP lengths of 25 and 65 residues as substrates. Both variants presented phosphotransferase activity and, similarly to results obtained for the phosphatase activity, the N-paPpx (1-314) had a lower catalytic efficiency (Figure 2). The catalytic parameters of the phosphotransferase activity compared to those of the phosphatase activity are listed in Table 2. In both variants, Enzyme Research 7 Table 3: Effect of mono-and divalent cations on phosphotransferase activity of the full-length paPpx and of the N-paPpx variant.

Mono-or divalent cations
paPpx  N-paPpx ( Values are means ± SD of at least three experiments performed independently. the phosphotransferase activity was independent of the polyP chain length since the (app) and the cat were in the same order of magnitude for the polyP 25 or the polyP 65 .
The phosphotransferase activity was dependent on Mg 2+ ; however, the 0.5(app)Mg 2+ value was approximately half that observed for phosphatase activity. The fact that the phosphotransferase activity was insensitive to the addition of K + or NH 4 + (Table 3) constituted an interesting finding. On the other hand, the affinity for ADP was independent of the length of the polyP chain used as phosphate donor. The (app) value was in the order of 90 M for both substrates and variants tested (data not shown).

Molecular Modeling of paPpx: Open and Closed Conformation.
Initially, we modeled a full-length paPpx by comparative modeling, using the atomic coordinates of the crystal structure of ecPpx [13] (PDB: 1U6Z). The template structure shared 41% of identity and 58% of similarity with paPpx  and was reported in what the authors named "closed conformation." In second place, we modeled the Nterminal domain of paPpx in the "open conformation" based on the atomic coordinates of aaPpx [26] (PDB: 2J4R).
Based on the homology in secondary structure and in spite of the low identity level between paPpx and aaPpx (27%), the model was constructed by threading using the "one-to-one threading" option of Phyre Server [27]. The model was obtained with 100% of confidence and 292 residues of 314 were aligned. The resulting structures had an architecture that is characteristic within the actin-like ATPase domain superfamily, composed of two subdomains in the N-terminal domain and other two subdomains in the Cterminal (in the case of full-length paPpx). In the proteins of this family, movements of up to 30 ∘ were described to be related to the catalytic function of the enzymes. In aaPpx a rotational movement of 22.5 ∘ between both domains around a single hinge region was described, indicating the access to the active site, located at the interface between domains. Kristensen and collaborators [26] described the access to the active site through a "butterfly-like" cleft opening. Figure 3(a) shows the superimposition of N-terminal domains of the models of paPpx in the open and closed conformations. A rotation of 24 ∘ between subdomains I and II was detected by DYNDOM server between both structures [30], with G 123 :R 124 and R 304 :E 305 as bending residues.

Active Site.
By homology to what has been described in other exopolyphosphatases [13,14,17,26], the active site of paPpx is formed by residues E 126 , D 149 , G 151 , S 154 , and E 156 . We confirmed the role of these amino acids in the active site of paPpx  by performing nonconserved site-directed mutations, replacing individually each of these residues by alanine. The two activities, phosphatase and phosphotransferase, were measured and the release of both P i and ATP production was severely affected (Table 4).

Docking with polyP and ADP.
After finding that paPpx can also act as a phosphotransferase, we were interested in proposing a three-dimensional model with polyP as P i donor and ADP as the acceptor. Therefore, we performed docking assays that complemented our biochemical findings.
Sequential with 7 residues long polyP (polyP 7 ) chain in presence of Mg 2+ . Among the pockets detected by the IcmPocketFinder tool there was one that covered all the cleft, which was postulated to be opened or closed with N-terminal domain movements. Considering that this region was consistent with the location of the active site and the putative polyP binding sites described by [13,14], we used this pocket for the docking assays. After the selection of the best conformer, the N-paPpx-polyP 7 complex was used as the receptor molecule to perform docking with ADP. The possible conformations adopted by the nucleotide seemed to be led by the P i moiety which is always located in the same position, and the guanosine group showed some degree of rotation resulting in slightly different positions inside the same binding site.
The location of the terminal P i nearby the Mg 2+ atom was consistent with the proposed catalytic mechanism. It was also possible to see the rest of the polyP chain running along the cleft. The final complex is shown in Figure 3( Values are means ± SD of at least three experiments performed independently. Both activities were measured with 8 M polyP 65 , 5 mM of Mg 2+ , and 80 mM of K + . 1 mol of P i min −1 mol protein −1 . 2 nmol of ATP min −1 mol protein −1 .
3 Mutated variants are all full-length paPpx  .
collaborators [13], the model of N-paPpx  in the open conformation with both ligands coexisting in the cleft was aligned with the N-terminal domain of ecPpx, which had SO 4 2− ions in the cleft. It is assumed that these ions describe a path for the polyP chain. The criterion for alignment was based on the active site residues, which remain in the same position regardless of the state of the conformation (opened or closed). This superimposition is shown in Figure 3(b). We also structurally aligned the complex obtained by docking with aaPpx and it is noted that phosphates residues of 3 are superimposable with the polyP residue while the 5 phosphates residues are superimposable with 5 phosphates of ADP (Figure 3(b)).

Electrostatic Potential Calculations.
A key aspect that remains to be characterized in the paPpx is the binding site of the polyP chain. Considering that the substrate of Ppx is a polyanion, it is possible that coulombic interactions govern the union between polyP and Ppx. Taking into account the relationship that may exist between the electrostatic potential of Ppx and the area where the polyP would bind [14], we performed the calculations of the electrostatic potential in paPpx  model and ecPpx. These results are shown in Figures 3(c) and 3(d). It is notorious that in paPpx the electropotential was much less positive than in ecPpx, especially around the cleft.

Discussion
We have recently demonstrated the involvement of Ppx in the pathogenesis of P. aeruginosa [8]. This finding led us to investigate in greater detail the enzyme from a biochemical approach. Previous studies performed by several authors have established that the ecPpx is composed of two independently folded domains: N-and C-terminal domains [13][14][15]. Based on this knowledge, we constructed several recombinant variants of paPpx: paPpx  , N-paPpx (1-314) , N-paPpx (1-303) , and C-paPpx (315-506) .
Our results indicate that the catalytic moiety of paPpx was localized in the N-terminal portion formed by the first 314 amino acid residues. Consistently with this, Alvarado and collaborators [13] also found that the N-terminal region of ecPpx, formed by 320 amino acid residues, was responsible for the catalytic activity. However, Bolesch and Keasling [15] did not find ecPpx activity in the N-terminal portion after limited proteolysis with Staphylococcus aureus V8 protease (Glu-C). A possible explanation for the discrepancy between these results may lie in the construction of the N-terminal variants. The peptide 304 EMEGRFRHQDVRSRTAS 320 located in the carboxyl end of the N-terminal variant constructed by Alvarado and collaborators [13] was absent in the Nterminal variant produced by Bolesch and Keasling [15]. This peptide constitutes a part of the connecting segment between domains II and III. When we compared the Cterminal end of the N-ecPpx 295-321 and N-paPpx 301-326 , we obtained a high degree of identity between these two enzymes (Figure 4(a)). The analysis of residue conservation through multiple sequence alignment showed that the residues forming the last -helix of the N-terminal domain (297-304 in ecPpx and 303-310 in paPpx) are conserved (Figure 4(b)). Structurally this helix is stepped between subdomains I and II of the N-terminal domain near the active site and constitutes a sort of separation between both subdomains (Figure 4(c)). We propose that this -helix is involved in ligand interaction and/or folding of paPpx, since no enzymatic activity was detected in the variant N-paPpx  . Likely, the lack of this helix may prevent the proper folding of the domain or it may cause a disruption affecting the active site, resulting in loss of activity.
Although N-paPpx  presented enzymatic activity, it had no preference for long polyP chains, and its (app) and catalytic efficiency presented roughly the same values for all tested substrates. These results strongly suggest that the C-terminal domain is important for the recognition and/or interaction with long polyP chains. These findings are in concordance with those reported by [15] that stated that Cterminal domain of ecPpx is involved in the recognition and processivity of long polyP chains.
The activity of paPpx was dependent on Mg 2+ , as it was demonstrated for other Ppxs [1]. Studies performed with the ecPpx showed that the maximum activity was achieved in presence of Mg 2+ 1 mM [35] while higher concentrations produced a sharp inhibition. It was reported for the Ppx of the archaea bacterium Sulfolobus solfataricus that Mn 2+ is needed as an activator in concentrations ≈1 mM and that Mg 2+ is needed to a lesser extent [38,39]. Furthermore, for the mitochondrial Ppx of Saccharomyces cerevisiae [40] the requirements determined for divalent metals were similar: about 1 mM of Mg 2+ , Mn 2+ , and Co 2+ , including Zn 2+ to a lesser extent. Results of these reports suggest similar catalytic mechanisms. Our results related to the activation of paPpx by divalent cations are in good agreement with the work of Dudev and Lim [41], who demonstrated that the Mg 2+ binding sites are not often specific for Mg 2+ and that Zn 2+ may replace it in that location. We found that K + and NH 4 + are nonessential activators of paPpx. Our results considerably  N  301  302  303  304  305  306  307  308  309  310  311  312  313  314  315  316  317  318  319  320  321  322  323  324  325  differ from those reported for the ecPpx [36,37], where more than twice the amount of K + was necessary to achieve the maximal activity (175 mM versus 80 mM) and where ammonium sulfate was not considered as an activator of the E. coli enzyme. We believe that the lack of activation by ammonium sulfate may be due to the presence of sulfate ions which can bind to the enzyme and mimic the phosphate residues of the substrate, thus blocking the activity. The crystal structure reported by [13] also supports our inference, since numerous sulfate ions were bound to the enzyme, which led the authors to postulate that the presence of these ions represented the polyP chain path.
The whole data presented confirm that paPpx activity is dependent on Mg 2+ and is stimulated by K + in a similar manner to ecPpx. Our present study also adds a new perspective to previous results, with the finding that NH 4 + can act as a direct activator of paPpx at lower concentrations than K + . It was demonstrated that bacteria may accumulate transiently polyP during the stationary phase and especially under conditions of nitrogen and P i limitation and during osmotic stress [9]. It is also known that K + is the most prevalent cation in the cytoplasm and it becomes more concentrated as the osmolarity increases [42]. Therefore, in a hyperosmolar condition, the increase of K + is useful to activate directly the paPpx. As the NH 4 + concentration is limited or in the presence of a nonpreferential nitrogen source (amino acids, choline, etc.), the activation of paPpx is at transcriptional level since the expression of the ppx gene is under the control of the global regulator NtrC [8]. On the other hand, with a good availability of nitrogen in the environment, NH 4 + directly stimulates the enzymatic activity and the bacteria obtain energy to start their growth.
Taking all these reports together, it was evident that polyP levels are involved in the global energetic state of the cell. Thus, we wondered if paPpx could have a second function, such as the synthesis of ATP by transferring P i from polyP to a molecule of ADP. Confirming what we previously hypothesized, we found that paPpx is also a polyP:ADP phosphotransferase. As the phosphatase activity, the phosphotransferase activity is located in the N-terminal domain and depends on Mg 2+ ; however, it is insensitive to the addition of K + or NH 4 + , suggesting that the catalytic mechanism is somehow different, probably acting nonprocessively.
Additionally, we performed in silico assays to complement our experimental data. We modeled the full-length paPpx in a closed conformation and the N-paPpx in an open conformation.
In E. coli, the active site was proposed to lay in a cleft between subdomains I and II [13] and a glycine-rich phosphate-binding loop more likely stabilizes the transition state during catalysis [14]. paPpx has full conservation of the proposed active site residues from ecPpx that would be formed by Glu 156 , Asp 149 , Gly 151 , Ser 154 , and Glu 126 and the glycine-rich loop consisting of residues Gly 141 -Ser 154 .

11
The conserved Glu 126 is proposed as the residue which activates a water molecule for the nucleophilic attack to the phosphodiester bond. The solved structures of ecPpx lack the Mg 2+ bound to the enzyme. The putative Mg 2+ binding site was suggested by comparison with the binding of aaPpx to a Ca 2+ atom [17]. By analogy to the description of ecPpx, the acidic residues Asp 149 and Glu 156 are predicted to contribute to the coordination sphere of Mg 2+ in paPpx [14]. We confirmed the participation of Glu 126 , Asp 149 , Ser 154 , Gly 151 , and Glu 156 residues in the active site by nonconservative site-directed mutagenesis in the full-length paPpx. The phosphotransferase activity of paPpx was also measured in the mutated variants of paPpx and no ATP production was detected. This result indicates that the active site of paPpx is the same for both activities: the phosphotransferase and the hydrolase.
It has been reported that ecPpx and aaPpx also presented pppGpp activity [17,26,43,44], suggesting that the active site would have enough space to bind a nucleotide. Other authors [13,14,26] have debated whether there are two different active sites or a common one. Considering the work presented by Kristensen and collaborators [26] who have reported the structure of aaPpx (in presence and absence of the alarmone ppGpp), it is clear that there is a single active site for hydrolysis of both polyP and pppGpp. As it was suggested by [14], ecPpx must undergo a conformational change from the closed to the open state to allow the entry of the nucleotide pppGpp. A phosphotransferase activity was never described for exopolyphosphatases or PPX/GPPA phosphatases; this work is the first reporting ATP production in a Ppx. Our results suggest that in paPpx the same active site is responsible for this activity, and we think that the transition from the closed to the open conformation occurs also in paPpx to allow the entry of ADP. The pppGpp molecule in the active site is occupying the space that would occupy both polyP and ADP (Figure 3(b)). Considering that we propose that the active site for phosphotransferase and phosphatase activities is the same, we would expect that other Ppx/GPPA phosphatases also have the ability to synthesize ATP. A major aspect that remains to be characterized in the paPpx is the binding site of the polyP chain. This site is not well defined because so far there is no crystal structure of a Ppx bound to the substrate polyP. The polyP binding site has been suggested in E. coli by two different approaches. Alvarado and collaborators [13] have described a possible binding area based on several sulfate ions present in the crystal, assuming that sulfates may mimic phosphates. On the other hand, Rangarajan and collaborators [14] suggested the presence of a channel of highly positive electrostatic potential for polyP binding at the dimerization interface. Both approaches are consistent in the suggested areas for polyP binding to ecPpx. These areas include the residues Arg 165 , Arg 166 , Arg 189 , and Lys 197 corresponding to a monomer and His 378 , His 384 , Arg 383 , Arg 413 , Lys 414 , and Lys 488 corresponding to the other monomer. Within these, it was found that Arg 166 , Lys 197 , His 378 , and Arg 413 triads are linked to sulfate ions in a crystal. It is very striking that in paPpx only half of these residues are conserved. Thus, the channel formed after dimerization had a less positive electrostatic potential (Figure 3). Residues in this region of paPpx are Leu 171 , Gln 172 , Gln 195 , Glu 203 , His 383 , His 389 , Lys 388 , Arg 418 , Arg 419 , and Gln 491 . This leads us to believe that while there is a channel, the potential is much less positive and, therefore, there must be differences in the binding of both enzymes with polyP. One of the possible roles of K + in the activity of paPpx may be the stabilization of the negative charges of polyP. This ion would also be involved in the constant attachment and detachment of the polymer during processive catalysis. If so, the interaction between the polyP and the enzyme in ecPpx would be stronger than in paPpx and, thus, greater amounts of K + would be required to detach the polyP in each catalysis cycle. We believe that this is one of the reasons why ecPpx needs 175 mM of K + for maximal activity whereas paPpx needs only 80 mM.

Concluding Remarks
In the present survey we show that, similarly to ecPpx, the catalytic activity of paPpx is found in the N-terminal region formed by subdomains I and II. This N-terminal domain is unable to distinguish the long polyP chain of beyond 15 residues in length. As occurs with ecPpx, the activity of paPpx depends on Mg 2+ and is activated by K + . In addition, we found and described new and original properties of paPpx, including that the polypeptide connecting the -helices from subdomains II and III is necessary for the catalytic activity and NH 4 + is an activator of the enzyme and may work at lower concentrations than K + . Finally, we demonstrate that paPpx has also a phosphotransferase activity capable of producing ATP. Surprisingly, this activity is dependent on Mg 2+ but is not activated by NH 4 + or K + , suggesting that, in spite of the fact that the active site is the same, the catalytic mechanism is slightly different.
The regulation of the degradation of polyP is complex and dual, since it involves regulation at transcriptional and biochemical levels. Thus, bacteria have mechanisms which ensure that paPpx is active in various physiological situations such as (i) under nitrogen limitation, where ppx is activated by NtrC, mediated by a 54 -dependent promoter [8]; (ii) under P i deficiency mediated by PhoB by a 70 -dependent promoter; (iii) in the presence of preferential nitrogen source (NH 4 + ), where the transcription of ppx is inhibited but the existing enzyme can be directly activated; and (iv) under hyperosmolarity, where paPpx is rapidly activated by K + , the most prevalent cation in the cytoplasm. This enzyme with its two functions can release P i for a direct synthesis of ATP or meet the nutritional needs where P i is necessary either to generate more energy or to initiate metabolic processes.
Together, the foregoing data and observations point out that paPpx is an enzyme of remarkable relevance due to its implication in polyP metabolism and, consequently, in virulence and pathogenesis of P. aeruginosa. Despite the fact that several specific aspects related to the paPpx enzyme, such as the detection of the specific binding site of the polyP chain, remain to be characterized, our present work contributes to the understanding of activity of the enzyme and of