Biochemical studies on Francisella tularensis RelA in (p)ppGpp biosynthesis

Francisella tularensis RelA shows significant sequence differences from other members of the RelA family of enzymes. In the present study, we describe the functional similarities and differences between F. tularensis RelA and the model RelA from Escherichia coli.

In Escherichia coli, and by inference other βand γproteobacteria, the enzyme RelA is principally responsible for (p)ppGpp synthesis during amino acid starvation [9]. Since its discovery [10], the mechanism by which RelA activation leads to (p)ppGpp accumulation has been extensively studied [11][12][13][14]. A working hypothesis for RelA activation proposes that the enzyme is most active following its release from a stalled ribosomal complex [11,15] (Figure 1A). This hypothesis, termed the 'extended hopping model' [15], explains how the bacterium can respond to a reduced level of amino acids by sensing a hiatus in protein synthesis and is supported by both in vitro [11][12][13] and in vivo [15] experimental evidence.
The protein sequence for long RSH proteins [4] can be divided into an N-terminal region (containing catalytic sites) and a C-terminal region [containing regulatory threonyl-tRNA Figure 1 The extended hopping model for RelA activation, derived from the proposal of English et al. [15] Under conditions of amino acid starvation, unacylated-tRNA accumulates and binds to the acceptor (A) site of the ribosome, leading to stalling. This results in the release of RelA from the ribosome in an activated conformation, which catalyses the formation of (p)ppGpp. (B) Domain structure and corresponding amino acids for EcRelA and FtRelA. (C) Alignment of RelA enzymes synthetase domain active-site motif from pathogenic bacteria. . Figure 1B) [4]. Long RSH proteins can be divided into bifunctional (Rel or SpoT) or mono-functional (RelA) enzymes. Both classes are capable of synthesizing (p)ppGpp but only bifunctional enzymes are capable of (p)ppGpp hydrolysis [16]. Studies with RelA fragments indicate the C-terminal region functions to regulate the catalytic synthetase domain [17] and contains the recognition features that permit homodimerization [18]. F. tularensis RelA (FtRelA) contains a truncated C-terminal region, approximately 100 amino acids shorter than most RelAs and is one of only three RelA sequences known to not contain an ACT domain [4] ( Figure 1B). Proteins containing ACT domains can be found throughout nature, displaying a conserved βαββαβ fold of the domain and are involved in small molecule ligand recognition [19,20]. This domain originally described as a 'conserved, evolutionary mobile module' is proposed to have evolutionarily fused with proteins to facilitate the regulation of their catalytic activity by the allosteric binding of small molecules [21].
Besides its truncated C-terminal region FtRelA differs from the majority of other RelA enzymes in its synthetase active-site motif. The synthetase domain from RSH proteins are reported to contain either an RXKD or an EXDD motif, which are involved in the preferential binding of the pyrophosphate acceptor (i.e. GDP or GTP) [22]. In contrast, the synthetase domain in FtRelA contains the alternative EXSD motif ( Figure 1C), which interestingly can be found in a wide spectrum of Francisella species (Supplementary Figure S1). This alternative motif shows the replacement of the initial aspartate with a serine. The primary aspartate has been proposed to allow the co-ordination of a second magnesium ion [23]. The difference in this active-site motif in FtRelA raised the question of whether the (p)ppGpp synthetase activity in the presence of an EXSD motif is more closely aligned to the RSH enzymes featuring RXKD or EXDD motifs.
The functional analysis of FtRelA, described herein, offers insights into the importance of the distinctive sequence of the synthetase domain and the absence of the ACT domain, compared with other RelA enzymes.

MATERIALS AND METHODS
DTT, BSA and antibiotics were purchased from Melford Laboratories; polyacrylamide-bis polyacrylamide [

Cloning of F. tularensis RelA
The gene encoding FtRelA, FTT_1508c, was amplified from F. tularensis subspecies tularensis SCHU S4 genomic DNA using a forward primer (5 -ccgccatgggtcatcatcatcatcatcatcaagttattgactctaaacttctagatagt) paired with a reverse primer (5cgcctcgagttagctgacctcttcattatcatc). The PCR product was digested with NcoI and XhoI and the resultant fragment was ligated into a backbone derived from NcoI/XhoI restricted pET16b. The sequence of the resultant plasmid pET16b::relA was verified by sequencing.

Expression of FtRelA
The plasmid pET16b::relA was chemically transformed into E. coli BL21 (DE3) pLysS competent cells (Sigma-Aldrich). Single colonies were used to inoculate 2YT media [24] (10 ml, containing 100 μg/ml ampicillin and 30 μg/ml chloramphenicol) and cultured overnight at 37 • C. The overnight culture was used as a 1 % inoculum for flasks of 2YT (4 × 1.25 litre) which were induced with IPTG (final concentration of 0.4 mM) when the absorbance at 600 nm (A 600 ) reached 0.6 and then cultured overnight at 16 • C. The cell pellet was then collected by centrifugation (average yield of 5 g/ 1 litre of culture) and stored at -80 • C.

Characterization of FtRelA multimeric state by gel filtration
A Superdex 200 column (10 mm × 300 mm) was used to estimate the apparent molecular mass of purified FtRelA (∼1 mg/ml, ∼13.5 μM). Protein samples were applied (1 ml·min − 1 ) to a pre-equilibrated column in buffer D (50 mM Tris/HCl, pH 7.5, 100 mM KCl). Cytochrome C, carbonic anhydrase, BSA, alcohol dehydrogenase and β-amylase (Sigma-Aldrich) were used as protein standards for calibration of the column. Elution of the protein samples were monitored by absorbance at 280 nm.

Substrate specificity of FtRelA
Reaction mixtures (100 μl) were prepared with components at the following final concentrations: ribonucleotide di-or triphosphates (2 mM, either G, T or C), ATP (2 mM) and FtRelA (10 μM) in assay buffer (20 mM Tris, pH 8.0, 15 mM KCl, 15 mM MgCl 2 , 1 mM β-mercaptoethanol). Reactions were incubated at 30 • C for 60 min, prior to quenching by heating. All reactions quenched by heat were maintained at 80 • C for 2 min. The reaction mixtures were then kept on ice to ensure full protein precipitation (typically 10 min), before clearing by centrifugation (13 684 g, 4 • C, 5 min). An aliquot (40 μl) of the supernatant was then analysed by IP RP HPLC. Assays to determine substrate selectivity under modified conditions are described in 'Supplementary Methods' 1 and 2.

Steady state kinetics for FtRelA and substrates
Reaction mixtures for time course experiments (1 ml) contained FtRelA (10 μM) and were made up in assay buffer. When saturating, nts (GTP or ATP) were included at 2 mM. Reactions were incubated at 30 • C and at selected time points, aliquots (100 μl) were withdrawn and reaction quenched by heating. Precipitated protein was removed as described above and a sample (40 μl) was analysed by IP RP HPLC. The concentrations of nts were quantified by comparison with a standard calibration curve. The formation of product nts over time (never more than 15 % substrate turnover) was fitted to a linear function to determine initial rates (Supplementary Table S1). Initial rates, v, were fitted to an allosteric sigmoidal function (eqn 1) using Graphpad Prism software where V max is the rate of reaction at substrate saturation, K 1/2 is the concentration of substrate giving a rate of half of V max and h is the apparent

[ 31 P]-NMR analysis of FtRelA activity
FtRelA-catalysed reaction time courses were monitored with [ 31 P]-NMR. All data were recorded on a Bruker AVII400 FT-NMR Spectrometer using a 10-mm auto-tune and match broadband probe tuned to the sample prior to data collection (data acquisition size = 65536 points; sweep width = 395 ppm; 512 scans using 90 • pulse with a total acquisition time of 2.5 s per scan; chemical shifts referenced to H 3 PO 4 ). Reaction mixtures (3 ml) were prepared in assay buffer and contained GTP (2 mM), ATP (2 mM) and 10 % 2 H 2 O. The reaction was initiated through the addition of FtRelA (10 μM), thoroughly mixed and then data collected at 25 • C for 22 min followed by 22 min bins for the duration of the experiment (typically 110 min). Peak integrals were calibrated against a spectrum recorded prior to initiation. At each substrate concentration, the formation of product nts over time was fitted to a linear function to determine the rate of product formation. These rates were then fitted to an allosteric sigmoidal curve (eqn 1).

Product activation of FtRelA
The nt ppGpp (5 ,3 -dibisphosphate guanosine) was synthesized and purified as described by Carmona et al. [29] with minor modifications (Supplementary Method S3). Reaction mixtures (100 μl) were prepared in assay buffer with components at the following final concentrations: ribonucleotide triphosphates (2 mM, both A and G), FtRelA (5 μM) and purified ppGpp, AMP or KH 2 PO 4 (at indicated concentrations). Reactions were incubated at 30 • C for 60 min prior to quenching by heating. Protein was precipitated and removed from samples as detailed above, prior to analysis of the sample by IP RP HPLC. Assays were prepared alongside a negative control containing no additional small molecular component. The EC 50 value was calculated with a sigmoidal dose-response (variable slope) curve (eqn 2), using Graphpad Prism software where Y is the rate, Y max is the maximum rate, Y min is the basal rate and EC 50 is the concentration of ligand required to give 50 % of full activation and h is the apparent Hill slope.

Expression and purification of FtRelA
The RelA encoding gene, FTT_1508c, from F. tularensis subspecies tularensis SCHU S4 was cloned with primers designed to encode an N-terminal His 6 -tag to facilitate subsequent protein purification. Optimized expression of FtRelA in E. coli BL21 (DE3) pLysS was achieved by maintaining the cultures at 16 • C overnight after induction. The FtRelA purification required two chromatographic steps with the initial purification using Ni-IDA affinity chromatography (Figure 2A). SDS/PAGE analysis showed a distinct band corresponding to the expected molecular mass (74 kDa) of FtRelA alongside several minor impurities ( Figure 2B). FtRelA was further purified in a polishing step, using size exclusion chromatography (Superdex 200, Figures 2C  and 2D). The resultant highly purified FtRelA was isolated in a yield of 12 mg/ g cell pellet and could be concentrated up to 10 mg/ml (∼100 μM).

Multimeric state of purified FtRelA
The multimeric state of purified FtRelA was determined using size exclusion chromatography and comparison with protein standards. FtRelA was shown to form one distinct peak which had an apparent mass of 128 + − 1.57 kDa that approximately corresponds to a dimer state (calculated molecular mass of 148 kDa; Figure 3).

Substrate specificity of FtRelA
As other RSH enzymes have been shown to utilize GDP and GTP as pyrophosphate acceptors [30][31][32][33], the specificity of FtRelA for these nts was investigated. End-point FtRelA activity assays were prepared with FtRelA, ATP and either GDP, GTP or, as a negative control CTP (Supplementary Figure S2). Efficient formation of AMP and the 3 -pyrophosphorylated product were only observed in the presence of GTP as a co-substrate (Figure 4; Supplementary Figure S2); GDP and CTP were not accepted as substrates. Very low concentrations of AMP are still measurable for activity assays with GDP as a pyrophosphate acceptor (Figure 4), but the formation of ppGpp is not observed.  Table S1). The HPLC analysis provided measurements of the products concentrations: the AMP peak was clearly resolved, however slow degradation of pppGpp (5 -triphosphate-3 -diphosphate guanosine) to ppGpp resulted in two partially overlapping peaks, which were combined to give an overall (p)ppGpp concentration (Supplementary Figure S4). A better fit of the velocity curves (plotting rate against concentration of substrate) was achieved for a sigmoidal (R 2 = 0.947) rather than a hyperbolic function (R 2 = 0.899) for AMP production (Figures 5B and 5C; Table 1     observed to curve in a manner consistent with positive cooperativity [34] (Figure 5D). Fitting to eqn. (1) yielded the kinetic parameters in Table 1 including K 1/2 which denotes the concentration of substrate at which half the maximal activity of the enzyme is achieved. The

Kinetic characterization of FtRelA basal activity by [ 31 P]-NMR
As the sigmoidal saturation kinetic profile of FtRelA observed by HPLC analysis is somewhat unusual, we sought to verify it using [ 31 P]-NMR as an alternative measurement technique. Despite NMR not often being suitable for monitoring enzymatic reactions, due to its intrinsic low sensitivity and resultant long acquisition times, the relatively slow turnover of FtRelA in the absence of activators (as judged by HPLC analysis) suggested [ 31 P]-NMR might also provide a useful approach for this enzyme. Using nt standards dissolved in enzyme assay buffer, the signals for each phosphorus atom in the substrates and products were identified (Supplementary Table S2). In preliminary NMR experiments, some of these signals were observed to overlap including those for the 5 -α and 3 -α of ppGpp (∼6 ppm). The signal for the α-phosphate of AMP (3.32 ppm; Supplementary Figure  S5) however was well resolved. Despite the formation of an individual peak for the 5 -β phosphate of pppGpp (- 19.88 ppm; Supplementary Figure S5), the observed overlap with the neighbouring peak (5 -β phosphate of GTP/ATP) and slow hydrolytic conversion of pppGpp to ppGpp resulted in an increased error in measuring pppGpp formation. This resulted in the higher derived rates and apparent V max determined from calculated initial rates of pppGpp formation ( Figure 5).
The conversion of GTP to pppGpp was shown to occur in a highly specific manner, with only one minor by-product being observed, inorganic phosphate (1.94 ppm; Supplementary Figure  S5), which accumulates during the activity assays. This observation is consistent with either the instability of pppGpp under assay conditions or the presence of a very low level of contaminating phosphatase activity [35].
The intrinsic insensitivity of [ 31 P]-NMR results in a relatively weak signal, but 22-min spaced time points partly compensated for this. However, the resultant data come with a caveat that a significant proportion of the substrates (∼30 %) had been turned over by the third time point (66 min). Somewhat surprisingly, these [ 31 P]-NMR time courses (Supplementary Figure S6) showed approximate linearity in product formation to at least 1 h and reaction rates were calculated within this linear range (Supplementary Figure S7; Supplementary Table S3). The rates determined at a selected range of substrate concentrations were plotted to yield velocity curves ( Figures 5E and 5F). Once again, these fitted to a sigmoidal function comparable to the results observed with IP RP HPLC analysis of FtRelA activity assays, including the K 1/2 and V max for GTP and ATP (Supplementary  Table S4). Data obtained by [ 31 P]-NMR should however be considered as an estimate only, due to the higher than normal proportion of substrate turnover measured. Kinetic data sets measured by [ 31 P]-NMR and HPLC were fitted to eqn (1) with V max and K 1/2 as global (shared) constants to derive overall kinetic parameters ( Table 2; Supplementary Figure S8). For the reasons previously discussed, the initial rates calculated by [ 31 P]-NMR analysis of pppGpp formation were omitted from this global fit.

Activation of FtRelA by small molecules
Shyp et al. [14] have described the positive regulation of E. coli RelA (EcRelA) by the product of GDP pyrophosphorylation, pp-Gpp [14]. Kinetic analysis using both HPLC and [ 31 P]-NMR methods showed positive co-operativity of FtRelA activity and it was of interest to determine if small molecules such as ppGpp or other small ligands were regulating activity. The effect of adding putative activating factors, AMP, phosphate or ppGpp prior to initiating the activity assay at low (10 μM), medium (100 μM) or high (1000 μM) concentrations on the rate of product formation was measured using IP RP HPLC analysis. The background concentration of additional ppGpp or AMP was subtracted to give the concentration of each product newly formed during the experiment. No significant activation was observed in the presence of additional inorganic phosphate (KH 2 PO 4 ) or AMP ( Figure 6A); however, in the presence of medium to high concentrations of ppGpp, activation was observed. Repeating this activation measurement over a wider range of ppGpp concentrations (0-1000 μM) allowed determination of an EC 50 for ppGpp of 60 + − 1.9 μM ( Figure 6B) and a maximal 1.5-fold activation, similar to the approximate 2-fold activation observed for EcRelA [14].

Activation of FtRelA by stalled ribosomal complexes
Bio-safety considerations encouraged us to explore alternatives to F. tularensis as sources for ribosome isolation. Two alternative ribosome sources were investigated. Firstly, the well characterized E. coli MRE600 strain which lacks ribonuclease I [36] and has consequently been widely used for ribosome purification [26,37,38]. Secondly, another member of the Francisella genus, F. philomiragia, which is of low virulence [39] and also encodes within its genome a RelA enzyme lacking the ACT domain in its C-terminus [4]. Maguire et al. [26] developed an affinity chromatography method for ribosome purification using cysteine coupled SulfoLink resin (Pierce). Previous purification of ribosomes from clinical isolates of pathogenic bacteria using this method [26], suggested its potential use for the isolation of ribosomes from F. philomiragia. Ribosomes were isolated from bacterial cells collected from early-mid logarithmic phase cultures, as established by growth curves (Supplementary Figure  S9), to ensure the optimal recovery of ribosomes. The protein and RNA content of purified ribosomes were analysed primarily by SDS/PAGE and absorbance traces at 260 and 280 nm (

DISCUSSION
Virtually ubiquitous across bacterial species, the stringent response is co-ordinated by signalling molecules (p)ppGpp and is important in bacterial survival under nutrient deficient conditions. In βand γ -proteobacteria, the principle enzyme responsible for (p)ppGpp synthesis is RelA [4] and yet research on this enzyme has focused almost exclusively on E. coli RelA as a model enzyme [11][12][13]30]. In the present study, we discuss the relevance of the distinctive synthetase active-site motif and C-terminal truncation within FtRelA in the functional similarities and differences when compared with other defined RelA enzymes.
Purification of this enzyme by nickel affinity chromatography followed by size exclusion chromatography yielded highly puri- fied RelA, which could be concentrated to a 10-fold higher concentration than that reported for E. coli RelA [13]. The higher solubility of FtRelA may be of practical value in future crystallization studies.
Previous work with E. coli RelA has demonstrated the enzyme's ability to dimerize via its C-terminal region [17,18]. The ability of FtRelA to therefore dimerize was of interest given its truncated C-terminus. In the present study however, we demonstrate that FtRelA forms a stable dimer upon purification. We postulate that the presence of the identified key residues for EcRelA dimerization Cys 612 , Asp 637 and Cys 638 [17] in FtRelA facilitate the formation the dimer without the requirement of downstream residues. We note that this may not however be FtRelAs natural multimeric state and might relate in the present study to the high protein concentration. Further analysis of FtRelA dimeric state over a range of concentrations may elucidate the enzymes multimeric state at concentrations closer to those found in vivo.
Previous work on synthetase activity in long RSH enzymes from E. coli [30], Mycobacterium tuberculosis [44] and Bacillus subtilis [31] has demonstrated a preference for either GDP or GTP as a substrate but invariably both are accepted. Data presented in the present paper details the first example of a RelA enzyme which has an explicit specificity for one of these two main pyrophosphate acceptors. This specificity was also observed under activating conditions with stalled ribosomal complexes but interestingly not when FtRelA was activated with the primary alcohol methanol (Supplementary Figure S3). The effect of methanol on protein conformation has been previously demonstrated to strengthen hydrogen bonds and weaken hydrophobic interactions [45]. Structural alterations to the synthetase-active site by methanol could therefore account for the acceptance of GDP as a pyrophosphate acceptor under these conditions and suggests the EXSD motif contributes to substrate specificity along with other structural elements of the synthetase domain. The potential importance of the structural transition in a putative catalytic loop . . within the synthetase domain of bi-functional enzymes with an RXKD motif has previously been reported [22]. It is therefore tempting to speculate that the EXSD motif contributes to the observed substrate specificity, although mutational analysis would be required to provide evidence for this. With equal Mg 2 + concentrations across all activity assays, the acceptance of GDP in the presence of methanol leads us to suggest that the magnesium concentration is unlikely to be the dominant factor in determining this specificity. Further work will be required to elucidate the functional significance of this catalytic loop in RelA enzymes with the unusual EXSD motif.
Kinetic analysis of the EcRelA in the absence of full activation has shown typical Michaelis-Menten kinetics [22,30]. In the present study, we demonstrate the kinetic profile for the FtRelA in the absence of activating factors fits a sigmoidal curve and yields the derived kinetic parameters (termed V max and K 1/2 ). Measurement of both the nt products AMP and the (p)ppGpp by HPLC analysis gave comparable sigmoidal fits and calculated kinetic parameters. The observed sigmoidal curve was verified by a second technique, [ 31 P]-NMR spectroscopy, which yielded comparable kinetic parameters when calculated using AMP production for rate determination. The global fit of data gave a sigmoidal curve with an R 2 value of 0.93 and 0.88 for ATP and GTP respectively (with a worse fit resulting from a hyperbolic curve, R 2 = 0.85). A sigmoidal curve has been observed for the bifunctional long RSH enzyme, Rel Mtb , synthetase activity in the absence of activating factors [22]. This kinetic profile was linked to the RXKD motif found in the synthetase active site [22].
The substrate specificity and co-operative kinetic effects during (p)ppGpp synthesis by FtRelA more closely resembles that of bi-functional RSH enzymes rather than that of other RelA enzymes. Collectively, these data demonstrate a divergence from the current classification system used for RelA enzymes [4,22] and may indicate that FtRelA is instead an example of a new distinct sub-class of RelA enzymes within the protein superfamily.
Several mechanisms can account for a sigmoidal velocity curve, including the interesting possibility of allosteric regulation [46]. In 2012, Shyp et al. [14] suggested ppGpp was responsible for positive feedback regulation of E. coli RelA by allosteric activation. ACT domains have been shown to regulate enzyme catalytic activity by the downstream effects of binding small molecules, namely amino acids [19]. The regulatory small molecule involved in binding the ACT domain of RelA enzymes has yet to be identified. FtRelA is one of only three RelA enzymes to not contain this domain [4], therefore the observation of FtRelA activation by ppGpp ( Figure 6) is highly indicative that ppGpp is not the regulatory ligand for RelA enzymes ACT domains.
Experiments with FtRelA demonstrated that the (p)ppGpp synthetase activity could be stimulated in vitro by the presence of stalled ribosomal complexes formed with ribosomes from alternative species (Figure 8), as has been observed previously for RSH enzymes [47]. This weak activation of FtRelA however pales in comparison with that achieved by EcRelA, with maximal activities of 701.5 + − 30.5 picomoles AMP per picomoles RelA per minute and 2952 + − 99.14 picomoles AMP per picomoles RelA per minute respectively ( Figure 8). Conversely, strong activation of EcRelA (in the range of levels previously reported [11,13]), was observed in the presence of stalled ribosomal complexes formed with either F. philomiragia or E. coli MRE600 ribosomes ( Figure 8). This strongly suggests that the observed lower sensitivity of FtRelA to ribosomal activation is genuine and not related to the quality of purified ribosomes or the use of heterologous systems for ribosomal stalling. Key residues identified in EcRelA involved in ribosomal binding were amino acids 550-682 [18]. At only 647 amino acids in length, FtRelA is missing ∼22 of these identified residues involved in EcRelA ribosomal binding [18] and this could account for the weaker activation observed for FtRelA by stalled ribosomal complexes. Future studies may identify the mechanistic relationship between this truncation and the observed reduced activation of FtRelA.

CONCLUSIONS
FtRelA contains a variety of amino acid sequence differences when compared with a wide range of other RelA enzymes, including a truncated C-terminus and an alternative EXSD active site motif. The current model for all RelA enzymes has been based on that from E. coli. In the present study, we describe the similarities and differences of FtRelA compared with the accepted model. Observed differences include the specificity of FtRelA for the pyrophosphate acceptor GTP (except in the presence of methanol). Furthermore, the sigmoidal steady state kinetics observed for FtRelA are unlike those reported for EcRelA, but similar to that observed for the bi-functional RSH enzyme Rel Mtb . Conversely FtRelA behaves similarly to EcRelA in its apparent ability to dimerize and its activation by both stalled ribosomal complexes and nt ppGpp. Comparison of the degree of activation by stalled ribosomal complexes for FtRelA and EcRelAs suggest the Francisella enzyme is more weakly activated. A deeper understanding of the underlying reasons behind the observed lower activation and its value to Francisella species forms an interesting objective for future research.