Unique Features of Alarmone Metabolism in Clostridioides difficile

ABSTRACT

The “magic spot” alarmones (pp)pGpp, previously implicated in Clostridioides difficile antibiotic survival, are synthesized by the RelA-SpoT homolog (RSH) of C. difficile (RSH_Cd) and RelQ_Cd. These enzymes are transcriptionally activated by diverse environmental stresses. RSH_Cd has previously been reported to synthesize ppGpp, but in this study, we found that both clostridial enzymes exclusively synthesize pGpp. While direct synthesis of pGpp from a GMP substrate, and (p)ppGpp hydrolysis into pGpp by NUDIX hydrolases, have previously been reported, there is no precedent for a bacterium synthesizing pGpp exclusively. Hydrolysis of the 5′ phosphate or pyrophosphate from GDP or GTP substrates is necessary for activity by the clostridial enzymes, neither of which can utilize GMP as a substrate. Both enzymes are remarkably insensitive to the size of their metal ion cofactor, tolerating a broad array of metals that do not allow activity in (pp)pGpp synthetases from other organisms. It is clear that while C. difficile utilizes alarmone signaling, its mechanisms of alarmone synthesis are not directly homologous to those in more completely characterized organisms.

IMPORTANCE Despite the role of the stringent response in antibiotic survival and recurrent infections, it has been a challenging target for antibacterial therapies because it is so ubiquitous. This is an especially relevant consideration for the treatment of Clostridioides difficile infection (CDI), as exposure to broad-spectrum antibiotics that harm commensal microbes is a major risk factor for CDI. Here, we report that both of the alarmone synthetase enzymes that mediate the stringent response in this organism employ a unique mechanism that requires the hydrolysis of two phosphate bonds and synthesize the triphosphate alarmone pGpp exclusively. Inhibitors targeted against these noncanonical synthetases have the potential to be highly specific and minimize detrimental effects to stringent response pathways in commensal microbes.

INTRODUCTION

Clostridioides (formerly Clostridium) difficile is a Gram-positive, spore-forming anaerobic gastrointestinal pathogen that causes Clostridioides difficile infection (CDI), which is multidrug resistant and has a high rate of recurrence (1). C. difficile is transmitted to susceptible hosts via fecal-oral transmission (2). Once spores are ingested, they survive the acidic pH of the stomach and travel through the intestine, where they interact with primary bile salts and amino acids that facilitate their germination into vegetative cells that secrete protein toxins which cause symptomatic disease (3). The most significant risk factor for CDI development is alteration of the normal gut microflora after exposure to broad-spectrum antibiotics (1, 4).

C. difficile must navigate a dynamic and hostile environment to establish symptomatic infection. The proliferation of vegetative cells in the gut environment is impacted by several factors, including the host immune response, oxygen levels, nutrient limitation, secondary bile acids, short-chain fatty acids, competing gut microflora, and various pHs throughout the gastrointestinal tract (5–8). The immune response to gastrointestinal infection can include fever of 1 to 4°C, antimicrobial host defense peptides, sequestration of metals to limit their availability, and oxidative stress from reactive oxygen species (ROS) and reactive nitrogen species (RNS) (9–13).

The stringent response (SR) is a stress survival mechanism that enables bacterial cells to quickly adapt to extracellular stress by temporarily halting growth and cell division and inducing transcription of stress survival genes (14, 15). The SR contributes to regulation of bacterial growth, virulence, antimicrobial peptide survival, oxidative stress resistance, antibiotic tolerance, and persistence in diverse bacterial pathogens (16). This process is canonically driven by the accumulation of two intracellular alarmone nucleotides, guanosine pentaphosphate (pppGpp) and guanosine tetraphosphate (ppGpp), collectively known as (p)ppGpp and sometimes referred to as “magic spots” (17). These signaling molecules are metabolized by multidomain bifunctional synthetase/hydrolase enzymes in the long RelA-SpoT homolog (RSH) family and by monofunctional small alarmone synthetase (SAS) and small alarmone hydrolase (SAH) domains (18). Extracellular stresses, including amino acid starvation, alkaline shock, cell wall stress from cell wall-active antibiotics, heat stress, oxidative stress, and acid stress can trigger the SR by modulating the transcription of either RSH or SAS genes (14, 19, 20). Posttranslational regulation of RSH enzymes typically involves protein-protein interactions or binding of uncharged tRNAs, while posttranslational regulation of SAS enzyme seems to depend on allosteric binding of (p)pGpp (21–24). The SAS enzymes also appear to play a stress-independent role in maintaining basal alarmone levels and regulating guanosine homeostasis in Bacillus subtilis and Enterococcus faecalis (25–27).

Guanosine substrate utilization varies among characterized alarmone synthetases, hinting at different biological roles. We have previously reported that the RSH of C. difficile (RSH_Cd) binds both GDP and GTP in vitro but exclusively utilizes GDP as a phosphoacceptor (28). This is unusual among RSH family enzymes, which utilize both GDP and GTP with various affinities to synthesize ppGpp and pppGpp and typically have higher affinity for GTP in Gram-positive organisms (18, 29–33). In addition, RSH homologs from B. subtilis and Methylorubrum (formerly Methylobacterium) exotorquens have recently been reported to synthesize (p)ppApp when incubated with adenosine phosphoacceptors (34, 35). The RelP and RelQ enzymes of B. subtilis, E. faecalis, and Staphylococcus aureus all exhibit greater affinity for GDP than GTP (25, 36, 37). RelZ from Mycobacterium smegmatis, which contains an RNase H domain in addition to the SAS, utilizes GMP preferentially (38). RelS from Corynebacterium glutamicum, part of a subfamily found only in Acinetobacter, has a higher affinity for GTP (39). C. glutamicum also contains a RelP homolog, RelP*, which lacks conserved amino acids at its C terminus and has no (pp)pGpp synthetic capacity (39).

Firmicutes species typically encode a single RSH and one or two SASs (18). The highly conserved synthetase domains of both families catalyze the transfer of a pyrophosphate moiety from ATP to 3′-OH of GDP and GTP to produce ppGpp and pppGpp, respectively (17). SAS enzymes from several Gram-positive species were recently discovered to synthesize a third alarmone, pGpp, utilizing GMP as a substrate (25, 39, 40). In addition, NUDIX hydrolases remove 5′ phosphates or pyrophosphates from ppGpp or pppGpp to produce pGpp (25, 38, 40–42). The physiological role of pGpp as a third alarmone is still under investigation. It has been speculated to allow a prolonged SR after the depletion of cytoplasmic GTP and GDP (25, 39). We henceforth refer to the three alarmones collectively as (pp)pGpp where appropriate.

Here, we report that expression of clostridial (pp)pGpp synthetase genes is induced by antibiotic stress and several stresses that mimic those produced by the immune system, including acid and oxidative stress and limited metal ion availability. We further report that the RelQ_Cd enzyme utilizes both GDP and GTP as substrates but that the only form of alarmone produced by either the RSH_Cd or RelQ_Cd enzyme is pGpp. Activity of both enzymes depends upon 5′ β phosphate bond hydrolysis of the GXP substrate. We found that RelQ_Cd, like RSH_Cd, tolerates a remarkable diversity of divalent metal ion cofactors. The seemingly exclusive production of pGpp is unique to C. difficile among previously studied bacteria.

(The thin-layer chromatography [TLC] image in Fig. 4B was previously reported in Appendix D of Astha Pokhrel’s doctoral dissertation, “Evaluating the Role of the Stringent Response Mechanism in Clostridioides difficile Survival and Pathogenesis” [43].)

RESULTS

C. difficile encodes one conserved SAS and one divergent SAS.

We have previously reported that C. difficile encodes an RSH family enzyme and the SAS RelQ_Cd (CDR20291_0350) and confirmed the enzymatic activity of RSH_Cd (CDR20291_2633) (28). In addition to RelQ, C. difficile encodes another putative SAS enzyme, CDR20291_1607, which we have named RelC. RelC contains a putative SAS domain and a C-terminal 250-residue region that contains no conserved domain, and its only homologs are in Firmicutes species (18). RelQ_Cd exhibits high sequence conservation in the ATP and GDP binding motifs identified in crystal structures of RelQ of B. subtilis (RelQ_Bs) and RelP of S. aureus (RelP_Sa) (Fig. 1) (31, 44). RelC has a highly conserved ATP binding motif, but the putative GDP binding motif lacks several conserved residues, including a tyrosine (Y116 in RelQ_Bs, Y151 in RelP_Sa, and Q137 in RelC) that stacks with the guanosine base in the GDP substrate and is invariant in SAS and RSH synthetase domains (31, 44). In addition, RelC contains a 10-residue insert within the GDP recognition motif that has no homolog in any characterized SAS (Fig. 1). Initial attempts to express RelC in Escherichia coli for purification were unsuccessful, but further experimentation will be necessary to determine whether RelC binds any form of guanosine substrate or serves as a pyrophosphotransferase (Fig. 1).

FIG 1.

Substrate binding sites of SAS enzymes. The sequences of characterized SAS enzymes from Bacillus subtilis, E. faecalis, S. aureus, S. mutans, C. difficile, C. glutamicum, and M. smegmatis are aligned with the crystal structure of Bacillus subtilis RelQ (PDB code 5DED). The clostridial sequences RelQ and RelC are highlighted in yellow. β-Strands are shown as arrows, and α-helices are shown as corkscrews. Conserved residues are shown in red, and highly conserved residues are shown in white on a red background. ATP and GDP binding motifs identified in RelQ_Bs and RelP_Sa are indicated at the bottom. The preferred guanosine nucleotide substrate of enzymes that have been kinetically characterized are shown on the left. “ND” indicates that the substrate preference of an enzyme has not been determined; “n/a” indicates that the enzyme is not active as a (pp)pGpp synthetase. Alignment was generated with ESPript 3 (https://espript.ibcp.fr/ESPript/ESPript/).

C. difficile induces rsh and relQ transcription in response to clinically relevant antibiotic stress.

In our earlier work, we utilized an anaerobic fluorescent transcriptional reporter incorporating the oxygen-independent fluorescent flavoprotein phiLOV2.1 to monitor transcription of the clostridial rsh and relQ genes (28). We observed the increased transcription of these genes upon exposure to sublethal concentrations of clindamycin and metronidazole (28). Antibiotic-induced synthetase gene expression differed between the epidemic strain C. difficile R20291 and the laboratory strain C. difficile 630Δerm despite the complete conservation of the relQ promoter between strains and the nearly complete (except for a single purine substitution) conservation of the rsh promoter. The protein sequences of both synthetases and of RelC are completely conserved between the strains (28). In this study, we extended these findings using two other antimicrobials, vancomycin and fidaxomicin, the currently recommended antibiotic treatments for severe or recurring CDI (45, 46). Exponentially growing cells were exposed to drug concentrations 0.5× those sufficient to inhibit growth (Fig. 2A; see also Fig. S1 in the supplemental material). In the C. difficile R20291 and C. difficile 630Δerm strains, promoter activity in the tetracycline-inducible control strain was unaffected by either antibiotic. In C. difficile R20291, vancomycin had no effect on transcription from the rsh or relQ promoter, while fidaxomicin stimulated a 96% increase in P_rsh activity and a 3.1-fold increase in P_relQ activity (Fig. 2B). In C. difficile 630Δerm, the P_rsh reporter had no response to either drug, while activity from the P_relQ reporter increased 3.6-fold upon exposure to vancomycin (Fig. 2C).

FIG 2.

Transcriptional response of (pp)pGpp synthetase genes to antibiotic stress. (A) Overnight growth of R20291 at the indicated concentrations of vancomycin (VAN) and fidaxomicin (FIX). Shown are the means and standard deviations of nine biologically independent samples. (B and C) PhiLOV fluorescence normalized to cell density after 2 h of exposure to 1.7 μg mL⁻¹ of vancomycin or 0.067 μg mL⁻¹ of fidaxomicin. Shown are the means and standard deviations of at least six biologically independent samples in the R20291 (B) and 630Δerm (C) backgrounds. Conditions with antibiotics were compared to those without (NT) by two-way analysis of variance (ANOVA). ****, P < 0.0001; *, P < 0.05. ns, not significant.

rsh and relQ transcription during oxidative, heat, and pH stress.

To assess the potential role of (pp)pGpp signaling and the stringent response in C. difficile survival of environmental stresses within the host, we monitored rsh and relQ promoter activity in C. difficile R20291 under conditions mimicking potential stressors in vivo. Activity from both promoters was indistinguishable after incubation at either 37°C or 41°C, suggesting that even high fever within a mammalian host would not activate transcription of (pp)pGpp synthetase genes (Fig. 3A). Oxidative stress was mimicked through exposure to copper, a transition metal previously shown to inhibit C. difficile growth, and to diamide, an oxidant that mimics oxidative stress in anaerobes by instigating disulfide bonds (47, 48). We have previously shown that 4 μM CuSO₄ prevents C. difficile growth (47). In this study, we have determined that 1 mM diamide is similarly inhibitory (Fig. S1). Three hours of exposure to copper did not stimulate increased activity from the rsh or relQ promoter; indeed, rsh promoter activity was somewhat decreased after copper treatment (Fig. 3B). Treatment with diamide stimulated a 3.8-fold increase in rsh promoter activity but had no effect on the relQ promoter (Fig. 3C). While the baseline fluorescence of the P_tet reporter strain increased under alkaline conditions, transfer to media buffered at neutral or alkaline pH had no effect on the activity of either clostridial promoter relative to that of the of the P_tet reporter (Fig. 3D). However, transfer from unbuffered brain heart infusion medium supplemented with 5% yeast extract (BHIS; initial pH measured at 7.4) to BHIS buffered at pH 6 did stimulate a 3.2-fold increase in P_rsh activity (Fig. 3D). Notably, when we measured the pH of spent BHIS after 24 h of C. difficile growth, we found that R20291 acidified its growth medium to pH 6. Medium acidification by 630Δerm was less pronounced (Table 1). We have previously shown that rsh transcription in C. difficile 630Δerm is induced by stationary-phase onset. It is possible that medium acidification, in addition to nutrient depletion, triggers this response.

FIG 3.

Transcriptional response of (pp)pGpp synthetase genes to environmental stress in R20291. (A) PhiLOV fluorescence normalized to cell density after 30 min of incubation at the indicated temperature. Shown are the means and standard deviations of six biologically independent samples. (B) PhiLOV fluorescence normalized to cell density after 2 h of exposure to the indicated concentration of copper sulfate. Shown are the means and standard deviations of six biologically independent samples. (C) PhiLOV fluorescence normalized to cell density after 3 h of exposure to the indicated concentration of diamide. Shown are the means and standard deviations of three biologically independent samples. (D) PhiLOV fluorescence normalized to cell density 3 h after transfer from unbuffered BHIS to BHIS buffered at the indicated pH with potassium phosphate buffer. Activity from the P_rsh and P_relQ promoters was compared to that from the P_tet promoter under the same condition by two-way ANOVA. *, P < 0.05; **, P < 0.01.

TABLE 1.

pHs of fresh and spent BHIS medium^a

C. difficile strain	pH
	Fresh BHIS	24-h spent BHIS
R20291	7.44	6.06
630Δerm	7.44	6.94

^apH of growth medium was measured before and 24 h after inoculation with single colonies.

RelQ_Cd is an active (pp)pGpp synthetase.

We have previously reported the in vitro synthetase activity of RSH_Cd, which transfers pyrophosphate to GDP phosphoacceptors exclusively (28). In this study, we confirmed that the putative synthetase RelQ from this organism is also active in vitro, capable of transferring a pyrophosphate moiety from [γ-³²P]ATP to a guanosine nucleotide phosphoacceptor (Fig. S2A). Unlike RSH_Cd, RelQ_Cd activity is concentration dependent and exhibits little to no pyrophosphotransfer activity at concentrations below 0.4 μM (Fig. S2A). This may explain why overexpression of RelQ_Cd in E. coli does not arrest cell division as we previously observed with RSH_Cd overexpression (Fig. S2B) (28). SAS from B. subtilis, S. aureus, and E. faecalis function as tetramers, raising the possibility that RelQ_Cd is not expressed in E. coli at levels sufficient to allow oligomerization and alarmone synthesis in vivo (31, 44, 49).

RelQ_Cd utilizes GDP and GTP but synthesizes pGpp exclusively.

In order to assess the substrate specificity of RelQ_Cd, we performed synthesis assays utilizing the monophosphate, diphosphate, and triphosphate forms of both adenosine and guanosine nucleotides as phosphoacceptors. We found that RelQ_Cd is incapable of transferring pyrophosphate from ATP to any adenosine nucleotides or to GMP (Fig. 4A). RelQ_Cd utilizes both GDP and GTP as phosphoacceptors but synthesizes the same-size product rather than the expected ppGpp and pppGpp, which should migrate differently on a chromatogram due to the additional mass and charge of the pentaphosphate alarmone (Fig. 4A). We performed synthesis assays using the same range of adenosine and guanosine phosphoacceptors with the RelQ enzyme from Bacillus subtilis, which is known to utilize GMP, GDP, and GTP in vitro in order to synthesize pGpp, ppGpp, and pppGpp, respectively (27, 34). As predicted, RelQ_Bs utilized the three guanosine nucleotide substrates to synthesize three differently sized products. Unexpectedly, the single product synthesized by RelQ_Cd had the same apparent size and charge as pGpp synthesized by RelQ_Bs from ATP and GMP, rather than ppGpp (Fig. 4A). Upon further exploration of guanosine utilization by these enzymes, it was found that RSH_Cd utilized GDP to synthesize an apparent pGpp product rather than ppGpp as we had previously reported (28). In this assay, RSH_Cd exhibited some trace synthesis of an apparent pGpp product using GTP as a substrate, although the pGpp spot was very faint and could have resulted from GDP contamination in the GTP sample (Fig. 4B). RelQ_Bs again utilized mono-, di-, and triphosphate guanosine substrates to synthesize tri-, tetra-, and pentaphosphate products, respectively. RelQ_Cd utilized both GDP and GTP as substrates but only synthesized the apparent triphosphate pGpp product. Neither clostridial enzyme utilized GMP as a substrate, and neither synthesized any product other than the apparent triphosphate magic spot pGpp.

FIG 4.

RelQ_Cd utilizes GDP and GTP but exclusively synthesizes pGpp. (A) Pyrophosphotransfer after 60 min from [γ-³²P]ATP to a 500 μM concentration of the indicated purine nucleotide phosphoacceptors by 2.0 μM RelQ_Cd or 3.0 μM RelQ_Bs. (B) Pyrophosphotransfer after 60 min from [γ-³²P]ATP to a 500 μM concentration of the indicated guanosine nucleotide phosphoacceptors by 3.0 μM RSH_Cd, 3.0 μM RelQ_Bs, or 2.0 μM RelQ_Cd. (C) After RelQ_Bs was incubated with GDP or GTP for 60 min, RelQ_Cd was added to the reaction for the indicated time and reaction products were analyzed by TLC. Each of the images shown is representative of at least three experiments. (D) The proposed reaction of pGpp synthesis via pyrophosphate transfer from ATP to the 3′ hydroxyl group of GDP with concurrent phosphate bond hydrolysis and inorganic phosphate release from the 5′ diphosphate of GDP.

To rule out the possibility that the clostridial enzymes synthesize ppGpp or pppGpp and then convert it to pGpp through an intrinsic hydrolysis capability, we incubated RelQ_Cd with ppGpp and pppGpp produced by RelQ_Bs. After an hour of incubation, the exogenously produced magic spots had not been hydrolyzed and no pGpp spot had emerged, suggesting that under the experimental conditions tested, RelQ_Cd synthesizes pGpp directly rather than converting ppGpp or pppGpp intermediates (Fig. 4C). This suggests that the clostridial synthetases could hydrolyze 5′ phosphate bonds on their guanosine substrates even as they ligate pyrophosphates to the 3′ hydroxyl groups (Fig. 4D).

Under the conditions studied, RelQ_Cd appeared to consume all of the [γ-³²P]ATP provided in the presence of GTP but leave more residual ATP in the presence of GDP, suggesting a higher affinity for the triphosphate substrate (Fig. 4A). Kinetic analysis of enzyme activity at multiple concentrations of guanosine substrate yielded calculated Michaelis constants of 45.4 μM for GTP and 53.6 μM for GDP, consistent with the visual observations (Table 2; Fig. S3). In order to further quantitate the affinity of RelQ_Cd for its guanosine substrate, we performed isothermal titration calorimetry. We found that the affinity constant of RelQ_Cd for GDP is roughly half that for GTP (Table 2). While the two methods did not produce a consensus for the absolute binding affinity of RelQ for its guanosine substrates, they were consistent in indicating that the enzyme has higher affinity for the triphosphate substrate under the experimental conditions. As RSH_Cd has a higher affinity for GDP than GTP in vitro, these results suggest that the two synthetases may be differentially active under discrete cellular conditions when either GTP or GDP is more prevalent.

TABLE 2.

Substrate affinity of RelQ_Cd at 37°C^a

Type of analysis and substrate	Km (μM−1)	K (M−1)	ΔH (cal mol−1)	ΔS (cal mol−1°C−1)
Michaelis-Menten
GDP	45.4
GTP	53.6
Isothermal titration calorimetry
GDP		2.04E3 ± 2.25E3	−8.18E7 ± 8.21E7	−2.64E5
GTP		4.87E4 ± 3.64E3	−1.639E6 ± 7.412E4	−5.25E3

^aThe binding of RelQ_Cd at 0.006 mM to GDP/GTP at 0.060 mM was measured by ITC. The data were fitted to a single-binding-site model. Shown are the values for K_m, K (the equilibrium binding constant), ΔH (the enthalpy change associated with protein binding to the ligand), and ΔS (the entropy change associated with binding). Each value is the average of two repeat experiments, and the standard deviations are shown.

The clostridial magic spot is pGpp.

To confirm the identity of the clostridial alarmone product, we investigated the ³¹P chemical shifts of the clostridial synthetase reactions. When ATP is mixed with GDP or GTP, the resonant peaks of each phosphorus are clearly visible and are consistent with the peaks of the individual nucleotides (Fig. 5C; Fig. S4) (50, 51). The chemical shifts of each of the phosphorus groups are listed in Table S1. The putative triphosphate alarmone pGpp contains a 5′ α phosphate with resonant peak near those in AMP and GMP (Fig. 5A to C). The alarmone also contains a 3′ diphosphate with α and β phosphates whose resonant peaks are very near those of the 5′ α and β phosphates of GDP or the α and γ phosphates of GTP (Fig. 5A, Fig. S4, and Table S1). When ATP and GDP were incubated with RelQ_Cd, the 5′ α peak of the ATP triphosphate at −10.71 ppm was reduced and a 5′ α monophosphate peak identical to that of the AMP peak at 3.49 ppm appeared at 3.52 ppm. Incubation with RelQ also reduced the ATP β and γ phosphate peaks at −19.16 ppm and −5.58 ppm, consistent with ATP hydrolysis (Fig. 5B, D, and E). The peaks at −10.10 ppm and −6.18 ppm, ascribed to the 5′ α and β phosphates of GDP, were broadened, presumably by overlap with the peaks of the chemically similar 3′ α and β phosphates of pGpp (Fig. 5D and E). A new peak appeared at 4.02, close to but distinct from that of the 5′ α monophosphate peak of GMP at 3.28 ppm, which we ascribe to a 5′ α monophosphate (Fig. 5C and E). In addition, an enormous peak appeared at 2.2 ppm, identical to published values for inorganic phosphate (Fig. 5E, F and H). These results are all consistent with transfer of a pyrophosphate from ATP to the 3′ OH of GDP, accompanied by hydrolysis of the 5′ β phosphate bond of GDP and release of inorganic phosphate.

FIG 5.

³¹P NMR evidence that the clostridial magic spot is pGpp. (A) Structures of nucleotide alarmone substrates and products. Nuclei are labeled according to the peaks assigned to them in the ³¹P spectra. (B) AMP standard. (C) GMP standard. (D) GDP and ATP standards. (E) GDP and ATP incubated with 2.0 μM RelQ_Cd for 40 min. (F) GDP and ATP incubated with 3.0 μM RSH_Cd for 60 min. (G) GTP and ATP standards. (H) GTP and ATP incubated with 2.0 μM RelQ_Cd for 40 min.

Similar results were obtained when ATP and GDP were incubated with RSH_Cd, confirming that our previous identification of the RSH_Cd product as ppGpp was erroneous (Fig. 5D and F). The signal from the α, β, and γ 5′ phosphates of ATP diminished, while a 5′ α monophosphate AMP peak and a 5′ α pGpp peak emerged (Fig. 5F). The α and β phosphate peaks of GDP differentiated into doublet peaks, presumably from interaction with the similar pGpp 3′ α and β phosphates. We attribute the greater loss of ATP peaks and greater differentiation of the guanosine α and β phosphate peaks to greater enzymatic activity by RSH, for which GDP is the preferred substrate, and subsequently less residual ATP and GDP in the reaction mixture.

When ATP and GTP were mixed, the α and γ phosphates of the two triphosphate substrates presented as doublet peaks, while the β phosphates presented as one broad peak (Fig. 5G). When ATP and GTP were incubated with RelQ_Cd, the 5′ α phosphate peak of pGpp appeared at 4.03 ppm and peaks appeared at −5.83 ppm and −10.39 ppm, consistent with the diphosphate of GDP or pGpp and not present in GTP appear (Fig. 5G and H).

When the enzymes were incubated with guanosine substrates in the absence of ATP, the inorganic phosphate peaks were very small relative to the peaks from the guanosine phosphates, suggesting that either ATP hydrolysis or pyrophosphotransfer stimulates the phosphorolysis of the guanosine substrate (Fig. S5). No peaks attributable to pGpp emerged, confirming that hydrolysis depends on ATP (Fig. S5).

Clostridial synthetases require phosphorolysis of the guanosine substrate.

To confirm that the guanosine substrate is hydrolyzed during clostridial magic spot synthesis, we employed 5′-β-thio-diphosphate (GDPβS), which has a nonhydrolyzable thiol bond between the α and β phosphates, as a phosphoacceptor. Both RelQ_Cd and RSH_Cd synthesized pGpp when GDP was supplied as a substrate, but neither clostridial enzyme generated any form of magic spot when GDPβS was the substrate (Fig. 6). RelQ_Cd did appear to deplete the radioactive ATP substrate in the presence of GDPβS but could not complete phosphotransfer, while RSH_Cd displayed no activity at all in the absence of a chemically labile β phosphate bond on the guanosine substrate (Fig. 6). RelQ_Bs was able to utilize GDPβS as a phosphoacceptor. The product formed by RelQ_Bs using GDPβS was larger than ppGpp formed from GDP since the nonhydrolyzable analog contains three lithium atoms that make the product heavier than ppGpp and closer in apparent size to pppGpp formed from GTP (Fig. 6).

FIG 6.

Clostridial synthetases hydrolyze the beta phosphate bond on GXP to synthesize pGpp. Pyrophosphotransfer from [γ-³²P]ATP by 2.0 μM RelQ_Cd, 3.0 μM RSH_Cd, or 3.0 μM RelQ_Bs to 500 μM GDP, nonhydrolyzable GDPβS, or GTP. The image shown is representative of three experiments.

RelQ_Cd utilizes diverse metal cofactors and may evade nutritional immunity.

In vitro analyses of RSH or SAS hydrolysis are typically carried out in the presence of micromolar levels of magnesium (17, 25, 30, 31, 39). RSH from Methylorubrum extorquens has previously been reported to utilize cobalt more efficiently than magnesium, although its activity is severely curtailed by manganese, calcium, or nickel cofactors, indicating that the enzyme’s metal utilization is highly specific (35). We have previously reported that RSH_Cd is capable of utilizing multiple structurally diverse divalent cations as cofactors with little to no loss of efficacy compared to magnesium (28). In this study, we found that RelQ_Cd also accepts multiple metal ion cofactors with a broad range of atomic radii, both smaller and larger than magnesium (Fig. 7; Table 3). Pyrophosphotransfer to a GDP substrate was robust under almost every condition studied. GDP utilization was modestly reduced by ferrous iron, the smallest metal ion tested (Fig. 7A; Table 3). RelQ_Cd utilization of GTP was more sensitive to metal cofactor identity. GTP utilization was decreased by iron, nickel, and copper, despite the last two being very close in size to magnesium (Fig. 7A; Table 3). Even calcium, which is substantially larger than magnesium, effectively substituted as a cofactor. In contrast, RelQ_Bs was much more selective for metal ion cofactors. This enzyme exhibited strong pyrophosphotransferase activity to both GDP and GTP substrates with cobalt, nickel, manganese, or magnesium cofactors but very little activity in the presence of iron, zinc, copper, or calcium (Fig. 7B). RelQ_Bs does appear to discriminate on the basis of ionic size, as the acceptable metal substitutions are those closest in size to the conventionally used magnesium (Fig. 7B; Table 3). Nickel ions appear to disfavor the use of GTP relative to GDP by both enzymes despite the fact that nickel is very close in size to magnesium, indicating that size is not the only factor that influences metal ion cofactor specificity.

FIG 7.

RelQ_Cd accommodates a wide range of metal ion cofactors and metal starvation induces rsh. (A and B) Pyrophosphotransfer after 60 min from [γ-³²P]ATP to a 500 μM concentration of the indicated GXP phosphoacceptor in the presence of a 5.0 mM concentration of the indicated metal ion cofactor by 2.0 μM RelQ_Cd (A) or 3.0 μM RelQ_Bs (B). Metal ions are arranged in order of increasing ionic size. Shown are the means and standard deviations of three independent samples. (C) PhiLOV fluorescence normalized to cell density after 1 h of exposure to 2 mM EDTA and/or 1 mM FeCl₂ and 1 mM ZnCl₂. Activity from the P_rsh and P_relQ promoters was compared to that from the P_tet promoter under the same condition by two-way ANOVA. ****, P < 0.0001.

TABLE 3.

Ionic radii of divalent metal cofactors^a

Cation	Effective ionic radius (Å)
Fe2+	0.75
Co2+	0.79
Mn2+	0.81
Ni2+	0.83
Mg2+	0.86
Cu2+	0.87
Zn2+	0.88
Ca2+	1.1

^aCrystal radii of metal ions in the +2 oxidation state with octahedral coordination. Where two electron configurations are possible, the low spin radius is listed (62, 63).

Because magnesium chloride was present in the protein purification buffer, we considered the possibility that residual magnesium rather than the added divalent cations was responsible for the observed activity. To address this, we performed ³¹P nuclear magnetic resonance (NMR) in the presence of 5 mM MgCl₂ and in the absence of any added metal cations. When magnesium chloride was added to the reactions, incubation of RelQ_Cd with ATP and GDP resulted in the loss of the ATP 5′ β phosphate signal and the concurrent emergence of 5′ α phosphate peaks from pGpp and AMP (Fig. 8A and C). Signal intensity from the ATP 5′ α and 5′ β phosphate peaks was also reduced (Fig. 8A and C). In the absence of supplemental magnesium chloride, incubation of RelQ_Cd with ATP and GDP still resulted in the loss of the ATP 5′ β phosphate signal but no gain of the 5′ α phosphate signal, suggesting that ATP was hydrolyzed but pyrophosphotransfer did not occur under this condition (Fig. 8B). Incubation of RSH_Cd with ATP and GDP without magnesium chloride abolished the ATP 5′ β phosphate signal and appeared to result in a small AMP 5′ α phosphate peak, suggesting that residual magnesium was sufficient for RSH_Cd ATP hydrolysis to ADP and possibly to AMP (Fig. 8D). Without added metal, RSH_Cd did not generate any 5′ α phosphate signal from pGpp, which was robust when RSH_Cd was incubated with magnesium chloride (Fig. 8D and E). RelQ_Cd incubated with ATP and GTP was similarly capable of eradicating the ATP 5′ β phosphate peak in the absence of added metal but did not generate a pGpp 5′ α phosphate peak unless magnesium chloride was present (Fig. 8F to H). RelQ_Cd supplemented with magnesium also resulted in separation of the previously overlapping ATP and GTP 5′ α and 5′ γ phosphate peaks, presumably as ATP was hydrolyzed to ADP (Fig. 8F and H). It appears that residual magnesium contamination was sufficient to allow ATP hydrolysis but not pyrophosphotransfer by both of the clostridial synthetases but that the alarmone synthesis observed in the presence of diverse metal cations was in fact dependent upon those metal ion cofactors.

FIG 8.

³¹P NMR evidence that alarmone synthesis depends on added divalent cations. (A) ATP and GDP incubated with MgCl₂. (B) RelQ_Cd (2.0 μM) incubated with ATP, GDP, and no residual metal for 40 min. (C) RelQ_Cd (2.0 μM) incubated with ATP, GDP, and MgCl₂ for 40 min. (D) RSH_Cd (3.0 μM) incubated with ATP, GDP, and no residual metal for 60 min. (E) RelQ_Cd (2.0 μM) incubated with ATP, GDP, and MgCl₂ for 40 min. (F) ATP and GTP incubated with MgCl₂. (G) RelQ_Cd (2.0 μM) incubated with ATP, GTP, and no residual metal for 40 min. (H) RelQ_Cd (2.0 μM) incubated with ATP, GTP, and MgCl₂ for 40 min.

We speculate that tolerance for the clostridial enzymes for a broad range of metal ion cofactor sizes could be an adaptive response to “nutritional immunity,” which is the ability of mammalian immune systems to chelate and sequester metal ions that are limiting factors for the growth of foreign pathogens (52). Metal starvation can trigger the stringent response in some bacteria (53, 54). We monitored phiLOV expression under the control of the rsh and relQ promoters after 1 h of exposure to the metal ion chelator EDTA. Metal depletion by EDTA had no significant effect on relQ promoter activity but did increase the activity of P_rsh (Fig. 7C). When an equimolar amount of the divalent cations zinc and iron, the best-documented targets of nutritional immunity, was added with the EDTA, the EDTA-induced increase in P_rsh activity was abolished. Supplemental zinc and iron did not affect P_rsh activity alone (Fig. 7C).

DISCUSSION

The outputs regulated by the SR are highly conserved, as accumulation of the magic spot alarmones causes similar phenotypes in diverse bacteria. However, different mechanisms lead to these convergent results. In this study, we explored the biological role and biochemistry of the clostridial SAS RelQ_Cd. Despite the high sequence conservation between RelQ_Cd and SAS from other organisms, our results were not entirely as expected.

We have previously demonstrated that stationary-phase onset and exposure to the antibiotics clindamycin and metronidazole stimulate rsh transcription in C. difficile and that genetic or chemical inhibition of RSH_Cd increases antibiotic susceptibility (28). In this study, we confirmed that antibiotic stress stimulates transcription of clostridial stringent response genes in a strain-specific manner; relQ responds to fidaxomicin but not vancomycin in R20291 and to vancomycin but not fidaxomicin in 630Δerm. It is tempting to speculate that strain-specific induction of synthetase genes may contribute to observed differences in antibiotic tolerance among C. difficile strains, although this will need to be verified by testing in more clinical strains before a conclusion can be drawn. We have previously found that transcription of rsh but not relQ is induced by stationary-phase onset (28). In this study, we investigated the response of the clinically relevant ribotype 027 strain C. difficile R20291 to extracellular stresses representative of those likely to be encountered in vivo and found that rsh but not relQ is upregulated by oxidative and acid stress in this strain. It is possible that RSH is the primary mediator of a stress-induced stringent response in this organism and that the primary role of RelQ is regulation of metabolic homeostasis as has been reported for B. subtilis and E. faecalis (26, 27). This would be consistent with the greater affinity of RelQ_Cd for GTP, which suggests a biological role for RelQ_Cd when cytoplasmic GTP is not depleted. Our finding that medium acidification caused by stationary-phase onset in R20291 is sufficient to induce rsh transcription in nutrient-rich fresh growth medium suggests that medium acidification due to metabolic activity is one of the signals that regulate the expression of stationary-phase genes in this organism.

Intriguingly, we report that the two clostridial magic spot synthetase enzymes share characteristics that appear to be unique to C. difficile. While some SAS enzymes can directly synthesize pGpp using GMP as a substrate, there is little precedent for a synthetase to produce the triphosphate signal using GDP or GTP as substrates (25, 34, 40–42, 55). There are no known examples of an organism that produces pGpp exclusively. Direct synthesis of pGpp via hydrolysis of GDP or GTP 5′ β phosphate bonds has only been reported as a minor in vitro activity of E. coli RSH, dependent on the presence of an EKDD motif in the synthetase active site (29). To date, every organism that synthesizes pGpp directly or indirectly also synthesizes ppGpp and pppGpp (25, 34, 40, 55, 56). Utilization of GDP and GTP to exclusively synthesize pGpp by both an RSH and an SAS enzyme has not previously been reported. While it is still possible that RelC synthesizes the longer alarmones despite its nonconserved guanosine binding motif, the dependence of RSH_Cd and RelQ_Cd on guanosine β phosphate bond hydrolysis indicates that pGpp is a very important, and possibly the only, clostridial alarmone. It is also unusual that RelQ_Cd exhibits a higher affinity for GTP than GDP. The only obvious divergence between RelQ_Cd and characterized SAS with substrate preferences for GDP is the presence of a glycine in the first position of the GXP binding motif, which is a polar charged residue in all other SAS homologs except C. glutamicum RelS, which has a valine at that position and preferentially utilizes GTP (Fig. 1). Previous work in RSH enzymes has posited that an acidic EXDD sequence at the beginning of the guanosine binding motif results in a preference for GDP, while a basic RXKD sequence correlates to a preference for GTP (29). This association does not appear to hold in SAS enzymes, several of which have a basic motif and prefer GDP (Fig. 1). However, it is likely that this position does contribute to the enzyme’s ability to utilize GDP or GTP. Future mutational analysis will likely be necessary to determine what residues play a role in guanosine β phosphate bond hydrolysis.

In addition to the unexpected triphosphate alarmone product, the characterized clostridial synthetases share another trait that does not appear to be widespread among RSH or SAS family enzymes: a remarkable ability to utilize structurally diverse divalent cationic cofactors. While few other synthetases have been tested against an expansive panel of divalent cations, cobalt, zinc, copper, nickel, iron, and calcium do not allow synthetase activity by Mycobacterium tuberculosis Rel and manganese, nickel, and calcium reduce or abolishes synthetase activity by M. extorquens Rel (33, 35). Zinc, but not iron or nickel, stimulates enzymatic activity of crystallized RelP from S. aureus (44). There does not appear to be any precedent for the extremely broad range of metals successfully utilized as cofactors by both clostridial synthetases (28). As it appears that metal ion limitation is another stress that induces expression of clostridial rsh, it is possible that promiscuous utilization of divalent cations in the enzymes that enact the clostridial stringent response could prevent dependence on any specific cation and provide a survival advantage in metal-limited environments. It is interesting that residual magnesium present after protein purification is sufficient for ATP hydrolysis but not pyrophosphotransfer and that RSH appears to remove a pyrophosphate from ATP while RelQ removes a phosphate. Future structural analyses will be necessary to explore possible mechanistic differences in metal ion usage by the two clostridial enzymes.

Given the role of the SR in antibiotic tolerance and virulence, there is interest in inhibiting it as an antimicrobial strategy (57, 58). However, the ubiquity of this signaling pathway among bacteria means that synthetase inhibitors are likely to exhibit a broad spectrum of action that could diminish the protective effect of the gut microbiota. The discovery that the SR in C. difficile is mediated by synthetases whose alarmone product and metal utilization are conserved with each other but different from that of their homologs in other bacterial species raises the possibility that the clostridial enzymes share structural features that distinguish them from synthetases in other organisms. Further characterization of the clostridial synthetase enzymes may provide a foundation for specific, targeted inhibition of the clostridial stringent response, which could be applied as an adjuvant to antibiotic therapy to diminish C. difficile antibiotic survival.

MATERIALS AND METHODS

Cell culture.

The bacterial strains and the plasmids used in this study are listed in Table S2. E. coli strains were grown in Luria-Bertani (LB) medium at 37°C. C. difficile strains were grown anaerobically in brain heart infusion medium supplemented with 5% yeast extract (BHIS). Anaerobic bacterial culture was maintained in a Coy anaerobic chamber (Coy Laboratory Products, Grass lake, MI) with an atmosphere of 85% N₂, 10% CO₂, and 5% H₂ at 37°C. Unless otherwise stated, all the experiments were conducted in C. difficile R20291. The pH of the C. difficile filtered growth medium was monitored before and after 24 h of incubation using a benchtop pH meter (Mettler-Toledo). To avoid adding spent liquid medium to the fresh samples, cultures used for medium pH experiments were inoculated with single colonies picked from BHIS agar plates. All plastic consumables were equilibrated in the anaerobic chamber for at least 72 h prior to use. Bacterial strains carrying plasmids were maintained using the corresponding antibiotics at the indicated concentrations: 10 μg mL⁻¹ of thiamphenicol (Tm10; Alfa Aesar) and 50 μg mL⁻¹ of kanamycin (Kan; Bio Basic).

Promoter activity analysis.

The in vivo promoter activity of the reporter strains was studied as previously described (28). Briefly, saturated starter cultures of C. difficile strains containing the phiLOV2.1 reporter plasmid were grown anaerobically at 37°C in BHIS-Tm10 for 12 to 16 h and inoculated 1:50 into fresh BHIS-Tm10 containing indicated treatments (1.70 μg mL⁻¹ of vancomycin [VWR], 0.67 μg mL⁻¹ of fidaxomicin [Cayman Chemical], 1.0 mM diamide [MP Biomedicals], 4.0 μM copper sulfate [Fisher Scientific]). For metal starvation, log-phase cells (optical density at 600 nm [OD₆₀₀], 0.50 to 0.70) were exposed to 1 mM FeCl₂ (Acros Organics) and 1 mM ZnCl₂ (Acros Organics) or 2 mM freshly suspended EDTA (Fisher Scientific) with 1 mM FeCl₂ and 1 mM ZnCl₂, and grown for 1 h at 37°C. To monitor temperature-induced promoter activity, log-phase cells (OD₆₀₀, 0.50 to 0.70) of each of the strains were divided into halves and incubated in parallel at 37°C and 41°C for 30 min. pH-induced promoter activity was measured by inoculating exponentially growing cells 1:50 into BHIS-Tm10 containing 0.1 M potassium phosphate buffer at pH 6.0, 7.0, 8.0, or 9.0. OD₆₀₀ was recorded for each sample upon collection. To minimize discrepancies in fluorescent signal from cellular autofluorescence, cell numbers in each sample were equalized on collection. Each sample was collected at a volume that would give a cell count equivalent to 3 mL of the culture with the lowest OD₆₀₀. After collection, cells were pelleted anaerobically in a microcentrifuge and suspended in 400 μL of anaerobic 1× phosphate-buffered saline (PBS). Duplicate 200-μL samples were aliquoted into a clear-bottomed black 96-welled microplate (Thermo Fisher Scientific) and were removed from the anaerobic chamber to measure sample fluorescence intensity. Sample fluorescence using 440-/30-nm excitation and 508-/20/nm emission filters and sample OD₆₃₀ were measured on a BrandTek plate reader. The instrumental parameters for all fluorescence measurements included a sensitivity limit of 65. Measurements were blanked against 1× PBS and were reported as E₅₀₈/A₆₃₀. All measurements were performed on at least six biologically independent samples. Statistical analysis was performed using Prism (GraphPad).

Cloning of RelQ_Cd.

The relQ gene (CDR20291_0350) was amplified from C. difficile R20291 genomic DNA using gene specific primers (5′relQ_NdeI and 3′relQ_XhoI). The relQ amplicon was ligated into pET24a expression vector at the NdeI-XhoI restriction sites and upstream of the vector-derived 6×His tag to generate pET24a::relQ plasmid. The plasmid was transformed into E. coli BL21 (New England BioLabs [NEB]) and confirmed by PCR using relQ gene-specific primers (relQ_a_F and relQ_a_R) (Table S2).

In vivo overexpression of RelQ.

Growth curve assays following the induction of C. difficile relQ in E. coli BL21 cells carrying pMMBneo::relQ were performed in 96-well microtiter plates (Brand Plates) using a BioTek Synergy plate reader. Exponential-phase cultures of E. coli were inoculated in LB medium in the presence or absence of 0.5 mM isopropyl-β-d-thiogalactopyranoside (IPTG) inducer. The plate was incubated at 37°C with shaking for a total of 16 h while monitoring cell growth every 30 min.

RelQ_Cd expression and purification.

Saturated starter cultures of E. coli BL21 cells containing pET24a-RelQ_Cd-His plasmid were grown overnight in LB medium with 50 μg mL⁻¹ of kanamycin in a shaking incubator at 250 rpm and 37°C. Cultures were diluted 1:20 into LB medium with 50 μg mL⁻¹ of kanamycin and grown at 37°C with shaking until the optical density at 600 nm reached 0.5 to 0.6. Once cells reached the desired OD₆₀₀, they were allowed to stand at room temperature for 10 min and induced with 0.5 mM IPTG for 3 h at 30°C for protein expression. Cells were centrifuged at a relative centrifugal force (rcf) of 4,030 using a Beckman coulter JA12 rotor for 30 min at 4°C, and pellets were frozen at −20°C. Cells were lysed using lysis buffer (50 mM Tris-HCl [pH 8.0], 500 mM NaCl, 5 mM MgCl₂, 10 mM imidazole, 3 mM β-mercaptoethanol (BME), and 100 mM phenylmethylsulfonyl fluoride [PMSF]) via sonication. The cells were burst for 10 s at 40% amplitude with 30-s pause for 8 cycles, followed by centrifugation at 13,680 × g for 15 min at 4°C in a benchtop Scilogex centrifuge. The soluble protein fraction was collected as supernatant and subjected for purification using HisPur nickel-nitrilotriacetic acid (Ni-NTA) resin affinity chromatography (Thermo Fisher Scientific). Briefly, the supernatant was mixed with equilibrating buffer (same as lysis buffer) in 1:1 ratio and subjected to a 1.0-mL Ni-NTA resin column equilibrated with the equilibrating buffer. Initially the column was washed with five column volumes of the lysis buffer and then further washed and eluted using a gradient of imidazole: 30 mM, 50 mM, and 75 mM concentrations of imidazole were used for column washing, while the protein was eluted using 150 mM and 300 mM imidazole in two fractions. Purified protein was subsequently dialyzed overnight at 4°C in a buffer consisting of 30 mM Tris (pH 8.0), 300 mM NaCl, 5 mM MgCl₂, 5 mM BME, and 20% glycerol. Protein concentration was determined spectroscopically, measuring A₂₈₀ using a BioDrop instrument. A molar extinction coefficient of 36,455 M⁻¹cm⁻¹ as determined by ExPASy (59) for RelQ_Cd was used to estimate the concentration.

Expression and purification of the control synthetase.

E. coli BL21 cells carrying pET28 bearing the Bacillus subtilis SAS1/RelQ gene were generously provided by Mingxu Fang, University of California, San Diego, and Carl E. Bauer, Indiana University, Bloomington. Bacillus subtilis RelQ was expressed and purified using a protocol developed by Mingxu Fang as previously published (43). Briefly, E. coli BL21 cells bearing pET28a::RelQ_Bs were grown in LB media containing 50 μg mL⁻¹ of kanamycin at 37°C with shaking until the OD₆₀₀ reached 0.4. Protein expression was induced overnight at 16°C by the addition of 200 μM IPTG. Cell pellet was collected by centrifugation at 4°C and resuspended in StrepTactin buffer containing 50 mM Tris (pH 8.9), 1 M NaCl, and 20% glycerol along with 2 mM PMSF. The cells were sonicated with a 10-s burst and 30-s pause cycle at 40% amplitude for 8 cycles, followed by centrifugation to collect the clarified lysate. The protein was purified by using StrepTactin resin column (IBA Life sciences) following the manufacturer’s instructions.

In vitro measurement of synthetase activity.

(pp)pGpp synthesis was carried in a reaction volume of 10.0 μL as described previously (28, 60) with a fixed concentration of RelQ_Cd or RelQ_Bs. The reaction mixture consisted of a buffer containing 10 mM Tris-HCl (pH 7.5), 5 mM ammonium acetate, 2 mM KCl, 0.2 mM dithiothreitol (DTT), 0.15 mM ATP, 5 mM MgCl₂, 1.0 μCi of [γ-³²P]ATP (PerkinElmer), and desired concentrations of AMP (Acros Organics)/ADP/ATP (Bio Basic)/GMP (BioPlus Chemicals)/GDP/GTP (Bio Basic) phosphoacceptor. Where indicated, the pH of the reaction was adjusted with HCl or NaOH. Where indicated, MgCl₂ was replaced with 5 mM metal salts in the +2 oxidation state (MnCl₂, CoCl₂, CuCl₂, ZnCl₂, NiBr₂, CaCl₂, and FeSO₄). The reactions occurred at 37°C and were stopped by spotting 2.0 μL of each sample onto polyethyleneimine (PEI)-cellulose thin-layer chromatography (TLC) plates, allowing the spots to dry. Unless otherwise indicated, reactions were run for 60 min before quenching. The plates were subsequently developed in 1.5 M KH₂PO₄ (pH 3.64 ± 0.03) and autoradiographed using a Storm 860 phosphorimager (GE Healthcare Life Sciences). pGpp signal was quantitated using ImageJ software to measure signal in each spot on the TLC plate and correct for background signal as previously described (60, 61). RelQ activity was expressed as the percentage of [γ-³²P]ATP substrate converted into product at each time point as described previously (60). Michaelis-Menten assays were performed by running the reactions for 5 min at the indicated concentration of GDP or GTP. Data shown are the means and standard deviations from two experiments. The line of best fit for the Michaelis-Menten equation was calculated using nonlinear regression with least-squares fit and the equation y = (V_max × X)/(K_M + X), where X represents the GXP substrate concentration and y represents the initial reaction velocity, and graphed with Prism (GraphPad).

Hydrolase contaminant check.

RelQ_Bs synthesis reaction mixtures containing either GDP or GTP as substrates were conducted for 60 min at 37°C as described above. RelQ_Cd was added to each reaction at a final concentration of 2.0 μM and incubated for an additional 60 min. At 0, 30, and 60 min after RelQ_Cd addition, 2.0 μL of each reaction mixture was spotted into the PEI-TLC plates for development as described above.

GXP phosphatase activity.

Guanosine nucleotide hydrolysis was assessed through synthetase assays performed as described above using guanosine 5′-β-thio-diphosphate trilithium salt (GDPβS) as the only nucleotide phosphoacceptor (Millipore Sigma). The reactions were caried out for 60 min at 37°C, and the spots were autoradiographed as reported above.

³¹P NMR spectroscopy.

The structural properties of the newly synthesized product were also evaluated by phosphorus NMR. Synthesis reactions were carried out with clostridial synthetases as mentioned above for 40 min at 37°C, and the reaction mixtures were subjected to acquire phosphorus NMR spectra. For “no metal” experiments, reactions were performed in 10 mM Tris-HCl (pH 7.5) containing 5 mM ammonium acetate, 2 mM KCl, 0.2 mM DTT, and no additional divalent cations. In addition, phosphorus NMR spectra of the standard nucleotides (ATP, GTP, GDP, GMP, and AMP) solutions in the reaction buffer containing 5 mM MgCl₂ were also acquired for signal comparison and assignments. D₂O at 15% was added into each of the samples to enable locking of the magnetic field in the NMR spectrometer. NMR spectra were indirectly referenced to 85% external phosphoric acid. ³¹P NMR recordings with proton decoupling were performed using a 400-MHz Bruker Avance III HD NMR spectrometer at 162 MHz (³¹P) equipped with a 5-mm PABBO BB/19F-1H/D Z-GRD Z108618/0798 probe. A total of 14,000 scans for each sample at 303 K were performed to acquire NMR spectra. All the measurements were carried out using pulse sequences supplied by the spectrometer manufacturer (Bruker TopSpin 3.2).

Isothermal calorimetry (ITC).

ITC200 microcalorimeter (Malvern) was employed for the analysis of the binding affinity of RelQ_Cd toward GXP as described previously (28). Briefly, a total volume of 300 μL of reaction buffer containing 0.006 mM protein was added into the cell component and 40 μL of 0.06 mM GXP was filled into the syringe, from which a total of 38 injections were made for the analysis at 37°C. Both the protein and ligands were prepared in the same buffer containing 10 mM Tris-Cl, 5 mM ammonium acetate (AmAce), 5 mM MgCl₂, 0.2 mM DTT, 0.12 mM ATP, and 2 mM KCl. The data obtained were processed in Origin software using a single-binding-site model. The heats of dilution acquired through the titration of ligand into the reference solution were subtracted from the binding curves during peak integration and the calculation of thermodynamic parameters. Two independent samples were measured.

Data availability.

The data that support the findings of this study are available from the corresponding author upon reasonable request.