The Two PPX-GppA Homologues from Mycobacterium tuberculosis Have Distinct Biochemical Activities

Inorganic polyphosphate (poly-P), guanosine pentaphosphate (pppGpp) and guanosine tetraphosphate (ppGpp) are ubiquitous in bacteria. These molecules play a variety of important physiological roles associated with stress resistance, persistence, and virulence. In the bacterial pathogen Mycobacterium tuberculosis, the identities of the proteins responsible for the metabolism of polyphosphate and (p)ppGpp remain to be fully established. M. tuberculosis encodes two PPX-GppA homologues, Rv0496 (MTB-PPX1) and Rv1026, which share significant sequence similarity with bacterial exopolyphosphatase (PPX) and guanosine pentaphosphate 5′-phosphohydrolase (GPP) proteins. Here we delineate the respective biochemical activities of the Rv0496 and Rv1026 proteins and benchmark these against the activities of the PPX and GPP proteins from Escherichia coli. We demonstrate that Rv0496 functions as an exopolyphosphatase, showing a distinct preference for relatively short-chain poly-P substrates. In contrast, Rv1026 has no detectable exopolyphosphatase activities. Analogous to the E. coli PPX and GPP enzymes, the exopolyphosphatase activities of Rv0496 are inhibited by pppGpp and, to a lesser extent, by ppGpp alarmones, which are produced during the bacterial stringent response. However, neither Rv0496 nor Rv1026 have the ability to hydrolyze pppGpp to ppGpp; a reaction catalyzed by E. coli PPX and GPP. Both the Rv0496 and Rv1026 proteins have modest ATPase and to a lesser extent ADPase activities. pppGpp alarmones inhibit the ATPase activities of Rv1026 and, to a lesser extent, the ATPase activities of Rv0496. We conclude that PPX-GppA family proteins may not possess all the catalytic activities implied by their name and may play distinct biochemical roles involved in polyphosphate and (p)ppGpp metabolic pathways.


Introduction
All living cells appear to have the physiological ability to synthesize and degrade inorganic polyphosphate (poly-P) molecules. These linear biopolymers comprise chains of phosphate (Pi) residues linked via 'high-energy' phosphoanhydride bonds, and range from a few to several hundred phosphate residues in length. In bacterial systems, poly-P is involved in a diverse range of biochemical, physicochemical and biological processes; e.g. the modulation of membrane structure and permeability; cell morphogenesis; DNA replication; as well as RNA and protein degradation. Poly-P also acts as an intracellular phosphate store and a biochemical phosphorylation agent (reviewed in refs. [1][2][3][4]). Within pathogenic species of bacteria, polyphosphate has been associated with enhanced levels of virulence, motility, stationary phase survival, persistence, resistance to complement-mediated cell lysis and increased biofilm formation [1][2][3][4][5][6][7][8][9][10]. Bacterial polyphosphate metabolism is also of notable environmental importance, playing a key role in the biological removal of phosphate from wastewater [11]. Consequently, the modulation of intracellular poly-P concentrations is of pivotal importance to numerous physiological processes involved in bacterial growth, viability, adaptability and infection.
Members of the Actinobacteria, (e.g. mycobacteria, corynebacteria, actinomycetes) lack identifiable PPX or GPP proteins. Instead, they generally encode two 'PPX-GppA' family proteins (pfam02541) ca. 300-350 aa in length, which share 20-35% aa identity with each other, and share homology with both PPX and GPP proteins [10,[24][25][26][27][28] (see Figure S1). There is a dearth of biochemical data for 'PPX-GppA' enzymes, which are generally assumed to have the ability to hydrolyze both polyphosphate and pppGpp substrates. This prompted us to comprehensively investigate the activities of the two PPX-GppA homologues encoded by the bacterial pathogen Mycobacterium tuberculosis: Rv0496 and Rv1026, which share ca. 27% amino acid identity. Polyphosphate and (p)ppGpp metabolism have been relatively well-studied within this organism at both the biochemical and biological levels [8][9][10]21,[29][30][31][32][33][34][35][36][37][38][39]. Evidence indicates that polyphosphate and the stringent response both appear to play pivotal roles in the ability of this organism to form virulent or persistent infections. However, many fundamental questions relating to the putative interplay between these two processes remain to be established.
Here we show that the Rv0496 (MTB-PPX1) protein functions as a short-chain exopolyphosphatase, whose activities are inhibited by (p)ppGpp alarmones. Most notably, neither MTB-PPX1 nor Rv1026 have the ability to hydrolyse pppGpp to ppGpp. Our results indicate that these two PPX-GppA protein homologues possess notably different biochemical activities.

Rv0496 (MTB-PPX1) has exopolyphosphatase activities
The rv0496 and rv1026 genes encoded by the H37Rv strain of M. tuberculosis were cloned into pMAL-c2 expression vectors, and over-expressed in E. coli. Milligram quantities of the respective Nterminal Maltose Binding Protein (MBP) fusion proteins were subsequently isolated in ca. 90-95% purity after one-step affinitypurification on amylose resin. The MBP-fusions (ca. 42 kDa) were cleaved using the Factor Xa protease, and the untagged recombinant Rv0496 and Rv1026 proteins were purified to .95% homogeneity using gel-filtration chromatography (Figure 1, Panels A and B). As the catalytic activities of the MBP-tagged and untagged forms of Rv0496 and Rv0126 were indistinguishable from one another; we used the tagged forms for the determination of their enzymatic kinetic parameters (see below), as they had slightly better in vitro stabilities (data not shown). We also cloned, expressed and purified the previously characterized E. coli PPX (EC-PPX) [12] and GPP (EC-GPP, GppA) [16] proteins, so that they could be included as side-by-side controls in the relevant assays (Supplementary Figure 2, Panels A and B). The multimeric states of the four recombinant proteins were determined by two independent approaches: light scattering and gel-filtration chromatography. As may be seen in Supplementary Figure 2D, consistent results were obtained for all four proteins using both approaches; namely they all adopted stable dimeric arrangements in solution. The dimeric arrangements of the E. coli PPX and GPP proteins established here is in agreement with previous reports [12,16], as well as the X-ray crystal structures previously obtained for the E. coli PPX protein [14,15].
A spectrophotometric phosphate release assay was used to determine the exopolyphosphatase activities of the Rv0496 and Rv1026 proteins, using EC-PPX and EC-GPP as positive controls. This assay was analogously used to determine the activities of the Cg0488 (CG-PPX1) and Cg1115 (CG-PPX2) proteins from Corynebacterium glutamicum [24]. A range of different buffers, salt concentrations and divalent metal ions were surveyed for their ability to promote the catalytic activities of the Rv0496 and Rv1026. It was subsequently found that the Rv0496 protein catalyzed the hydrolysis of polyphosphate in the presence of Mn 2+ ions; and to a lesser extent, Mg 2+ and Zn 2+ ions; but no activity was detected in the presence of Ca 2+ , Co 2+ , Cu 2+ or Fe 2+ ions. Exopolyphosphatase activities were highest in HEPES buffer (50 mM, pH 6.8) containing 1 mM MnCl 2 and 25 mM KCl (data not shown). No significant exopolyphosphatase activities could be detected for the recombinant Rv1026 protein under any condition tested. The respective abilities of the Rv1026 and Rv0496 proteins to catalyze the hydrolysis of a representative polyphosphate substrate (poly-P 130 ) in the presence and absence of 1 mM MnCl 2 , may be seen in the poly-P PAGE gel shown in Figure 1C. It may be seen that he Rv0496 protein hydrolyzed the poly-P 130 to shorter chain length polyphosphate (and phosphate) only in the presence of Mn 2+ ions (compare lanes 2 and 3). We confirmed that there was no detectable polyphosphate hydrolysis activities in the absence of protein (lane 6), or in the presence of the purified maltose binding protein (lane 1). Under the same conditions, the E. coli GPP (EC-GPP) protein hydrolyzed the poly-P 130 to undetectable levels. Consistent with the results from the spectrophotometric assays, the Rv1026 protein had no detectable activities (lane 5). We therefore renamed the Rv0496 protein 'MTB-PPX1', analogous to the C. glutamicum nomenclature [24]. However, we thought it inappropriate to refer to Rv1026 as 'MTB-PPX2', as this activity was not demonstrated.
To confirm the results from these assays within a more biologically-relevant environment, we performed a set of complementation experiments using a crude cell lysate prepared from the CF6032 (DgppA Dppkx) mutant strain of E. coli, analogous to those previously performed by Kuroda et al. [22]. This triple mutant strain is defective for GPP, PPK and PPX protein expression, and hence lacks the ability to hydrolyze polyphosphate or pppGpp molecules to any significant extent. Recombinant MTB-PPX1 (Rv0496), Rv1026, E. coli PPX (EC-PPX) or MBP (negative control) proteins were added to a buffered reaction mixture containing cell lysate and poly-P 130 . After incubation at 37uC for 2 hours, polyphosphate content was analyzed on polacrylamide gels (see Figure 2, Panel A). It may be seen that the CF6032 lysate supplemented with MTB-PPX1 protein (lane 8) effectively mediated the hydrolysis of poly-P 130 to shorter chain products (and phosphate); analogous to the results previously obtained using the fully-defined in vitro conditions (Figure 1 Panel C, lane 3). Under similar conditions, the EC-PPX protein (lane 10) hydrolyzed poly-P 130 to undetectable levels, with no apparent formation of intermediate chain length products. In contrast, neither the MBP (lane 7) nor Rv1026 (lane 9) proteins were able to complement the CF6032 lysate for polyphosphate hydrolysis activity. A lysate that was analogously prepared from the wild type MG1655 strain of E. coli digested a significant proportion of the poly-P 130 under equivalent conditions (lane 5) in the absence of added GPP or PPX protein, unlike the mutant strain (lane 6).
These assays were repeated using a crude cell lysate analogously prepared from cultured Mycobacterium smegmatis mc 2 155 cells ( Figure 2, Panel B). As may be seen in lane 5, the M. smegmatis lysate did not possess detectable polyphosphate hydrolase activities. Similar to what was observed for the assays containing E. coli DgppA Dppkx lysate, the addition of the MTB-PPX1 protein resulted in a significant proportion of the poly-P 130 being hydrolyzed to shorter chain length polyphosphate (lane 7). The consistent observation of significant amounts of (partially-digested) polyphosphate molecules of intermediate chain length (ca. 60-100 phosphate residues) in these assays indicates that Rv0496 does not function as a highly processive exopolyphosphatase, unlike EC-PPX and EC-GPP [12,16]. Consistent with previous results, the addition of Rv1026 to the M. smegmatis lysate led to no detectable alterations in polyphosphate hydrolysis levels (lane 9). This further indicated that there were no components present in the M. smegmatis lysate that were able to stimulate or otherwise interact with the Rv1026 protein, to promote the hydrolysis of polyphosphate molecules.

Rv0496 (MTB-PPX1) prefers short-chain length polyphosphates as substrates
The substrate specificities and kinetic parameters of the MTB-PPX1 (Rv0496) protein were probed in detail, using synthetic polyphosphate molecules of various chain lengths (poly-P 14 , poly-P 60 , poly-P 130 ). The K m and k cat values determined for each of the three poly-P chain lengths are shown in Table 1. The K m (poly-P 14 ) for MTB-PPX1 was 5.960.3 mM, which was ca. two-fold lower than the K m for poly-P 60 , and ca. 3 times lower than the K m for poly-P 130 . The rate of hydrolysis for the three polyphosphate chain lengths followed a similar trend, with poly-P 14 being hydrolyzed most rapidly (k cat = 10.260.1 s 21 ). This clearly indicated that MTB-PPX1 preferentially hydrolyzed short-chain poly-P molecules. Contrastingly, results from analogous experiments revealed that both the EC-GPP and EC-PPX proteins bound long-chain polyphosphate substrates most effectively. The K m (poly-P 130 ) values for the EC-GPP and EC-PPX proteins were 3.261.1 mM and 1.160.3 mM, respectively, which is reasonably consistent with previous findings [12,16]. However, we found that the E. coli PPX and GPP proteins hydrolyzed the shorter-chain poly-P 14 and poly-P 60 substrates significantly more rapidly than poly-P 130 , even though these substrates had higher K m values. The turnover numbers and catalytic efficiencies of EC-GPP and EC-PPX were notably higher than those of MTB-PPX1 (see Table 1), clearly indicating that the two E. coli enzymes had substantially higher exopolyphosphatase activities under the conditions tested.

Rv0496 and Rv1026 do not hydrolyze guanosine pentaphosphate (pppGpp)
To the best of our knowledge, there is no commercial supplier of pppGpp. Previous investigations have used a promiscuous purine nucleotide pyrophosphotransferase from Streptomyces morookaensis [40] or a purified E. coli ribosomal fraction that contains the RelA protein [40] to convert ATP + GTP into pppGpp. We have found that the RelQ protein homologue from Enterococcus faecalis (EF-RelQ, EFFG_01794) has excellent (p)ppGpp synthesis activities,  [41], was used to enzymatically synthesize pppGpp and ppGpp from ATP + GTP, or from ATP + GDP, respectively. Anion exchange chromatography was used to separate the synthesized (p)ppGpp alarmones from the starting materials and AMP biproduct (see Figure S3) . The purified pppGpp and ppGpp nucleotides were then desalted on Sephadex G10 resin [42] and characterized as described by Hardiman et al. [43] to confirm their identity and purify.
As previous reports have failed to investigate the putative guanosine pentaphosphate 59-phosphohydrolase activities of the PPX-Gppa homologues from A. aeolicus and C. glutamicum [24,26,27], we felt it was highly-important to determine whether the MTB-PPX1 (Rv0496) or Rv1026 proteins had this catalytic ability. As the E. coli GPP and PPX appear to be the only proteins clearly demonstrated to have this activity, they were both included as positive controls [16,22]. Levels of pppGpp and ppGpp present after prolonged enzymatic incubation with MTB-PPX1 or Rv1026 were analyzed using anion exchange chromatography (see Figure 3). Under the conditions used, ppGpp and pppGpp had distinct retention times (eluting at 11.38 ml and 11.72 ml, respectively). Whilst both the EC-GPP and EC-PPX proteins effectively cleaved the 59-terminal phosphate group, the MTB-PPX1 and Rv1026 proteins completely lacked this activity under all conditions employed (see Figure 3).
To further confirm the pppGpp hydrolase abilities of the MTB-PPX1 (Rv0496) and Rv1026 proteins, we performed sets of complementation assays using crude cell lysates prepared from E. coli MG1655 (wild type), CF6032 (DgppA Dppkx) and M. smegmatis mc 2 155, analogous to those used to analyze their exopolyphosphatase activities. In these assays, 2 mg of purified protein and cell lysate (2 mg total protein) were incubated with pppGpp (0.1 mM) in buffered reaction mixtures. The pppGpp and ppGpp levels were then quantified using anion exchange chromatography. The chromatograms obtained are shown in Figures S4, S5 and S6, respectively. As may be seen in the chromatogram shown in Figure  S4 Panel A, more than 80% of the pppGpp was hydrolyzed to ppGpp by the wild type E. coli MG1655 lysate. In contrast, there was no significant hydrolysis of pppGpp mediated by the E. coli These data clearly indicated that the Rv1026 was a more active and efficient ATPase than MTB-PPX1. Both proteins could also catalyze the hydrolysis of ADP to AMP (Figure 4, Panels E and F), with Rv1026 being slightly more active than MTB-PPX1 in this regard. As may be seen in Panels A and B, the Rv1026 protein mediated the sequential hydrolysis of ATP to ADP to AMP under the conditions employed; clearly indicating it was more efficient ATPase and ADPase than MTB-PPX1. The MTB-PPX1 and Rv1026 enzymes also had the ability to synthesize small amounts of ATP from ADP (see Figure 4, Panels E and F). From this observation, we speculate that there may a phosphory transfer to an active site residue. If this phosphorylated enzyme intermediate is sufficiently long-lived; after the release of the initial AMP product, there could be subsequent binding and phosphorylation of a second ADP molecule, resulting in the synthesis of ATP.
The MTB-PPX1 and Rv1026 proteins possessed extremely low GTPase activities, which were only marginally above background (non-enzymatic) levels under any of the conditions tested ( Figure 5). Notably, the addition of polyphosphate or phosphate ions (p)ppGpp alamones inhibit the exopolyphosphatase activities of MTB-PPX1 As the MTB-PPX1 and Rv1026 proteins lacked the ability to hydrolyze pppGpp to ppGpp, we next investigated whether their exopolyphosphatase were inhibited by either of these alarmone molecules. The E. coli PPX and GPP proteins were included as positive controls, as Kuroda et al. have previously shown that pppGpp, and to a lesser extent ppGpp, inhibit their poly-P hydrolysis activities [22]. The K i values reported for E. coli PPX were 10 mM for pppGpp, and 200 mM for ppGpp. We consequently used 1 mM concentrations of either pppGpp or ppGpp in our inhibition experiments in order to obtain definitive results. In comparative experiments, we quantified the maximal rate of poly-P 130 hydrolysis (V max ) by the EC-GPP, EC-PPX and MTB-PPX-1 enzymes in the presence of pppGpp or ppGpp, versus alarmone-free controls. Results are summarized in Table 2A. Consistent with the findings of Kuroda et al. [22], pppGpp inhibited the exopolyphosphatase activities of EC-PPX ca. 10-fold more effectively than ppGpp. Analogous results were obtained for the EC-GPP protein, with pppGpp inhibiting poly-P hydrolysis 5-fold more effectively than ppGpp. Similarly, we found that the maximal rate of poly-P 130 hydrolysis by the MTB-PPX1 protein was reduced by ca. 38% in the presence of 1 mM pppGpp. Under the same conditions, 1 mM ppGpp lowered the rate of poly-P hydrolysis by only ca 6% (approximately 6-fold less effectively). In contrast, the exopolyphosphatase activities of MTB-PPX1 were not significantly affected by 59-mono-, di-and triphosphates of adenosine or guanosine, even at 5 mM concentrations (data not shown). Taken together, these experiments revealed that the pppGpp and ppGpp alarmones had analogous inhibitory effects on the exopolyphosphatase activities of the E. coli GPP, PPX and MTB-PPX1 proteins; with pppGpp being a significantly more potent inhibitor.

(p)ppGpp alamones inhibit the ATPase activities of Rv1026 and MTB-PPX1
We also investigated whether the pppGpp or ppGpp alarmones could inhibit the ATPase activities of the MTB-PPX1 or Rv1026 proteins. 1 mM pppGpp inhibited the ATPase activities of Rv1026 (V max reduced by ca. 32%), but ppGpp had no significant effects (see Table 2A). pppGpp molecules also inhibited the ATPase activities of the MTB-PPX1 protein under these conditions, but to a lesser extent (V max reduced by ca. 8%). ppGpp alarmones had no detectable inhibitory effects. Data obtained from further inhibition assays performed over a range of ATP concentrations, were consistent with pppGpp inhibiting the ATPase activities of Rv1026 in a competitive manner (see Table  S2).

Discussion
Here we show that the MTB-PPX1 (Rv0496) protein functions as an exopolyphosphatase, exhibiting a distinct preference for relatively short-chain polyphosphate substrates. Its activities appear to be similar to those of the CG-PPX2 (Cg1115) protein from C. glutamicum [24]. Both enzymes prefer short-chain polyphosphate substrates, require Mg 2+ or Mn 2+ ion cofactors, and have optimal activities at slightly acidic pH values (pH 6.8). The sequences of the MTB-PPX1, Rv1026 and CG-PPX2 proteins align with the N-terminal regions of the E. coli PPX and GPP proteins ( Figure S1). The available structural and bioinformatic data indicates that polyphosphate and pppGpp substrates are hydrolyzed at a single active site located within this region [14,15,24,26,27]. Both the MTB-PPX1 and Rv1026 proteins contain conserved amino acid residues implicated in the catalytic mechanism. Asp135 and Glu142 (MTB-PPX1) and Asp146 and Glu153 (Rv1026) are predicted to be responsible for binding the essentially-required Mn 2+ or Mg 2+ ions. Arg84 and Glu112 (MTB-PPX1) and Arg90 and Glu123 (Rv1026) are implicated in the hydrolysis of the phosphoanhydride bond linking the terminal and penultimate phosphate units in the poly-P or ATP (and ADP) substrates. X-ray crystal structures of the E. coli PPX protein suggest that the polyphosphate chain is bound primarily by residues located within the C-terminal region [14,15]. This is supported by biochemical and biophysical data obtained from truncated forms of the EC-PPX protein lacking the C-terminal domain [13]. As these ca. 150-200 C-terminal residues are absent in MTB-PPX1 (and CG-PPX2), it most likely adopts an alternative mode of polyphosphate binding. This may be responsible for the differences in polyphosphate chain length preferences observed for MTB-PPX1, compared to the E. coli PPX and GPP proteins; which have higher affinities for short and longchain polyphosphate substrates, respectively (Table 1). Bolesch and Keasling previously defined the N-terminal ca. 300 residues of EC-PPX as functioning as a 'quasi-processive' exopolyphosphatase [13]. This is what the MTB-PPX1 protein appears to function as. The activities of additional diverse PPX-GppA homologues will need to be characterized to determine whether this putative structure-function relationship generally holds true.
Our finding that the E. coli GPP protein has notable GTPase activities is not consistent with the previous report by Hara and Sy [17]. We tentatively speculate that these authors may have inadvertently purified and characterized the E. coli PPX protein; as their report predates the identification and characterization of EC-PPX [12]; and the discovery that this protein also possesses guanosine pentaphosphate 59-phosphohydrolase activities [22]. Future investigations may substantiate or repudiate this speculation.
Transposon mutagenesis has previously shown that the rv0496 gene was non-essential in the H37Rv strain, whilst rv1026 was required for optimal growth [44]. Rv0496 (MTB-PPX1) was found to be secreted from M. tuberculosis cells, and was identified as a T-cell antigen with potential for vaccine development [45,46]. Thayil et al. [10] have recently reported that the 344 aa MT0516 protein from M. tuberculosis CDC1551, which has an identical sequence to Rv0496 (MTB-PPX1) from the H37Rv strain, could hydrolyze a 65-mer of polyphosphate; noting that its polyphosphate hydrolase activities were not inhibited in the presence of 1 mM of ppGpp. The authors also reported that the intracellular poly-P levels were elevated in a MT0516-deficient strain of M. tuberculosis during the mid-logarithmic and late stationary growth phases. Using a guinea pig lung model, they further demonstrated that the activities of MT0516 played a key role in enabling M. tuberculosis cells to grow and persist within necrotic granulomas. The authors did not report the activities of the 319aa MT1054 protein, which has an identical amino acid sequence to the Rv1026 protein characterized here. Our results are broadly consistent with the preliminary biochemical data reported by Thayil et al., and greatly extend upon it.
The PPK1 protein appears to be the major source of polyphosphate in M. tuberculosis, synthesizing a relatively broad spectrum of poly-P chains lengths, ranging from ca. 200 to 800mers [3,8,30]. It should be noted that MTB-PPX1 is one of a number of proteins that possess the catalytic ability to catabolize polyphosphate molecules via hydrolysis or other biochemical processes. The PPK2 (Rv3232c) protein functions primarily as a polyphosphate-dependent ATP (and possibly GTP) regenerating enzyme, rather than a source of poly-P [4,8,30,38,39,47]. Other net-consumers of poly-P are polyphosphate glucokinase (Rv2702) [34] and PpnK/NADK (Rv1695) [35]. The physiological significance of the MTB-PPX1 protein's preference for shortchain poly-P remains obscure, as the distributions of polyphosphate chain lengths in M. tuberculosis cells are not accurately known.
The biological role of the Rv1026 protein has not yet been established. Its expression was previously shown to be up-regulated in cells treated with translation-inhibitors [48], and it was also induced during intracellular growth within macrophages [49]. Our results reveal that Rv1026 lacks the ability to hydrolyze poly-P or pppGpp substrates. It possibly plays a role in adenine nucleotide equilibration; promoting the formation of ADP and AMP, at the expense of ATP (Figure 3). However, the physiological significance of these biochemical activities and the relevance of the pppGpp-mediated inhibition of its ATPase activities remain obscure. Shi et al. recently reported that the over-expression of Rv1026 in M. smegmatis led to a small, temporary decrease in the relative abundance of polyphosphate [25]. Our results suggest that the over-expression of Rv1026 may lower polyphosphate levels in M. smegmatis cells in an indirect manner. It would putatively reduce the polyphosphate-synthesizing activities of the M. smegmatis PPK1 protein, by lowering its substrate (ATP) levels and increasing it product (ADP)-mediated inhibition; amongst other possible pleiotropic effects [8,30,39,50]. However, we cannot exclude the possibility that Rv1026 functions as an exopolyphosphatase when expressed within M. tuberculosis or other mycobacterial cells.
As mediators of the stringent response, pppGpp and ppGpp alarmones affect global changes in stable RNA levels, DNA synthesis, amino acid biosynthesis and protein degradation [18,19,20,21]. The stringent response plays a pivotal role in the ability of M. tuberculosis cells to exist in a prolonged state of dormancy within macrophage phagosomes, and to form persistent and virulent infections [29,31]. Our results clearly demonstrate that the both the MTB-PPX1 and Rv1026 proteins lack the ability to hydrolyze pppGpp to ppGpp (Figure 3). It remains to be seen whether M. tuberculosis encodes an alternative protein with GPP functionality, or does not require this alarmone-converting activity. The bifunctional Rel MTB (Rv2583c) protein is the only source of pppGpp and ppGpp molecules in M. tuberculosis [32,33]. Polyphosphate molecules modulate the transcription of rel MTB via a two-component MprAB/SigE pathway, thereby regulating (p)ppGpp production [8]. Via this mechanism, increased polyphosphate levels lead to increased (p)ppGpp levels. In E. coli, there is positive feedback via the (p)ppGpp-mediated inhibition of PPX activities [22]; thereby prolonging the intracellular lifetime of polyphosphate. As pppGpp, and to a lesser extent ppGpp, inhibit the exopolyphosphatase activities of MTB-PPX1 (Table 2A), our results suggest that this regulatory feedback is also present in M. tuberculosis.
To briefly conclude, our results demonstrate that the Rv0496 (MTB-PPX1) protein functions as a short-chain exopolyphosphatase, whose activities are inhibited by (p)ppGpp alarmones produced during the bacterial stringent response. Neither MTB-PPX1 nor Rv1026 have the ability to hydrolyze pppGpp, a property that makes them notably different to the GPP and PPX proteins from E. coli. The data presented here reveals that members of the PPX-GppA protein family possess notable differences in their catalytic activities, indicating that overall sequence homology may not be a reliable indicator of biochemical or biological functionality.

Gene cloning procedures
Rv0496 (MTB-PPX1) and Rv1026. The rv0496 and rv1026 genes were PCR amplified from M. tuberculosis H37Rv genomic DNA using the Rv0496for2 and Rv0496rev2, and Rv1026for and Rv1026rev primer pairs, respectively, with the use of LongAmp Taq DNA polymerase from New England Biolabs (NEB). After TOPO cloning (pCR2.1 TOPO-TA cloning kit from Invitrogen, Life Technologies), amplified genes were subcloned (XbaI/ HindIII for Rv0496, BamHI/HindIII for Rv1026), into similarly digested pMAL-c2 expression vectors (NEB), to encode Nterminal maltose binding protein (MBP; MalE) fusions. The MBP protein was expressed from unmodified plasmid pMAL-c2, for use as a negative control. For a list of the primers used in this study, see Table S1.

Gel filtration chromatography
The molecular masses of the recombinant Rv0496 (MTB-PPX1), Rv1026, E. coli GPP, E. coli PPX and E. faecalis RelQ proteins were determined by size exclusion chromatography on a Superdex 200 gel filtration column (GE Healthcare) using an AKTA purifier system (GE Healthcare). Calibration curves were constructed using protein standards (HMW and LMW gel filtration calibration kits, GE Healthcare).

Exopolyphosphatase assays
The exopolyphosphatase activities of Rv0496 (MTB-PPX1), EC-PPX and EC-GPP were determined using continuous spectrophotometric assays, quantifying phosphate (Pi) release using the EnzChek phosphate assay kit (Invitrogen) according to the manufacturer's protocol; analogous to the method reported by Lindner et al. [24]. Unless otherwise stated, assays were performed in 96-well plates at 37uC, in 200 ml of 50 mM HEPES pH 6.8, 1 mM MnCl 2 , 25 mM KCl, containing EC-PPX, EC-PPX or Rv0496 (MTB-PPX1) protein (2 mg, 0.1-0.2 mM), polyphosphate (0-150 mM) of the stated chain length (poly-P 14 , poly-P 60 , poly-P 130 ); quantifying changes in the absorbance at 360 nm using a Spectra Max 340 plate reader (Molecular devices). Inorganic polyphosphate samples with average chain lengths of P14, P60 and P130 were provided by Dr. T. Shiba (RegeneTiss Inc., Japan). V max and K m values were determined by fitting data to the Michaelis-Menten equation using Origin 6.0 (OriginLab). Experiments were performed in quadruplicate, and the mean values 6 standard deviation are reported.

Enzymatic preparation of ppGpp and pppGpp alarmones
The ppGpp and pppGpp nucleotides were synthesized enzymatically using recombinant EF-RelQ protein. Reaction mixtures (100 ml) contained EF-RelQ (5 mg) in 50 mM Tris-HCl pH 8.6, 100 mM NaCl, 10 mM MgCl 2 , 1 mM DTT, 1 mM ATP and 1 mM of GDP (to make ppGpp) or GTP (to make pppGpp), and were incubated for 1 h at 30uC. Synthesized pppGpp or ppGpp were purified by anion exchange chromatography (1 ml Resource Q column, GE Healthcare) on an AKTA purifier system (GE Healthcare). The eluent was monitored at 254 nm to detect and quantify nucleotide-containing fractions (see supplementary Figure  3). Identical runs using known concentrations (0-1 mM) of ATP, ADP, AMP, GTP, GDP, GMP, ppGpp and pppGpp were performed to enable unambiguous nucleotide identification and quantification. Fractions containing pure pppGpp or ppGpp were desalted by gel filtration chromatography on Sephadex G-10 (GE Healthcare), analogous to the method of Krohn and Wagner [42] and were characterized as described by Hardiman et al. [43]. ATP hydrolysis assays ATP hydrolysis activities were determined by quantifying Pi release using the EnzChek phosphate assay kit, analogous to the method described above with minor modifications. Reactions (200 ml) contained Rv0496 (MTB-PPX1) or Rv1026 protein (1 mg), ATP (0.5-10 mM), 1 mM MnCl 2 , 4 mM (NH 4 ) 2 SO 4 , in 50 mM Tris-HCl pH 7.4; were incubated at 37uC for 1 hour. Various concentrations of ATP (0.5, 1, 2, 5, 10 mM) were used to establish the steady-state kinetic constants for Rv0496 and Rv1026 proteins. Data were taken every minute, and V max and K m values were determined by fitting data to the Michaelis-Menten equation using Origin 6.0 (OriginLab). Experiments were performed in quadruplicate, reporting the mean values 6 standard deviation.
EnzChek phosphate release assays analogous to those described above were performed at 37uC in 50 mM Tris-HCl pH 7.4, containing: Rv0496 or Rv1026 protein (1 mg, ca. 0.1 mM), 1 mM ATP, 1 mM MnCl 2 , 4 mM (NH 4 ) 2 SO 4 ; and the indicated amount of purified and desalted pppGpp or ppGpp (0-2 mM). For both sets of assays, V max values were determined and compared with corresponding values obtained in the absence of (p)ppGpp. The % decrease in V max values were reported as the mean 6 standard deviation, obtained from three independent experiments, each performed in duplicate.
Reaction mixtures were analyzed on TBE 12% polyacrylamide gels as described above.