(p)ppGpp inhibits 70S ribosome formation in Staphylococcus aureus by impeding GTPase-ribosome interactions

During nutrient limitation, bacteria produce the alarmones (p)ppGpp as effectors of the stress signalling network termed the stringent response. Screening for (p)ppGpp-binding targets within Staphylococcus aureus identified four ribosome-associated GTPases (RA-GTPases), RsgA, RbgA, Era and HflX, each of which are cofactors in ribosome assembly, where they cycle between the ON (GTP-bound) and OFF (GDP-bound) states. Entry into the OFF-state from the ON-state occurs upon hydrolysis of GTP, with GTPase activity increasing substantially upon ribosome association. When bound to (p)ppGpp, GTPase activity is inhibited, reducing 70S ribosome assembly. Here, we sought to determine how (p)ppGpp impacts RA-GTPase-ribosome interactions by examining the affinity and kinetics of binding between RA-GTPases and ribosomes in various nucleotide-bound states. We show that RA-GTPases preferentially bind to 5′-diphosphate-containing nucleotides GDP and ppGpp over GTP, which is likely exploited as a regulatory mechanism within the cell. Binding to (p)ppGpp reduces stable association of RA-GTPases to ribosomal subunits compared to the GTP-bound state both in vitro and within bacterial cells by inducing the OFF-state conformation. We propose that in this conformation, the G2/switch I loop adopts a conformation incompatible with ribosome association. Altogether, we highlight (p)ppGpp-mediated inhibition of RA-GTPases as a major mechanism of stringent response-mediated growth control.


INTRODUCTION
The prokaryotic 70S ribosome is an essential and complex macromolecular assembly responsible for the translation of messenger RNA (mRNA) into functional proteins. It comprises a large 50S and a small 30S subunit, which consist of 33 ribosomal proteins (r-proteins: L1-L36) associated with two ribosomal RNAs (rRNA), and 21 r-proteins (S1-S21) with one rRNA, respectively. Due to the energetic cost of ribosome synthesis and the intricacy of assembly, cofactors play a vital role in ensuring the correct conformation of the complete 70S (1). One class of assembly cofactors are the ribosome-associated GTPases (RA-GTPases), a subset of P-loop GTPases within the Translation Factor Associated Due to the variation in accessory domains, each RA-GTPase associates with a distinct area of the ribosome to coordinate a maturation event. Cycling between the GTP-bound ON and GDP-bound OFF states enables these proteins to act as molecular checkpoints of ribosome assembly by monitoring the maturation state of individual subunits (7). Although it is unclear what the precise roles of RA-GTPases are in ribosomal maturation, they have been suggested to sterically prevent the premature association of other r-proteins (8). Unknown maturation events then act as activators of GTPase activity, enabling entry into the GDP-bound OFF state and subsequent dissociation from the ribosome (7). In addition to regulating the recruitment of r-proteins, RA-GTPases have been postulated to recruit RNA processing enzymes directly. For instance, the RA-GTPase Era can interact with several proteins involved in 16S rRNA maturation, including YbeY, an endonuclease involved in 16S processing in Escherichia coli (9), and CshA, a DEAD-box RNA helicase (10), pointing to a role for this group of enzymes as hub proteins that facilitate maturation events.
During periods of starvation, bacteria produce the alarmones guanosine penta-and tetraphosphate (collectively referred to as (p)ppGpp), which function as the mediators of a stress signalling system termed the stringent response (11). Amidst this response, the concentration of (p)ppGpp within the cell can reach between 1 -2 mM with a concurrent drop in GTP levels (12,13). This results in a plethora of downstream effects, including alterations to transcription, translation and DNA replication, as well as regulating late-stage growth phases such as sporulation or biofilm formation (14)(15)(16). Our previous work identified the four RA-GTPases (RsgA, RbgA, Era and HflX: Figure 1A) in the pathogenic bacterium Staphylococcus aureus as enzymes that can bind to and are inhibited by (p)ppGpp, resulting in a negative impact on 70S ribosome assembly and growth (17). RsgA is a nonessential, highly conserved late-stage 30S assembly cofactor (17,18), that has been implicated in the docking of helix 44 (h44) of the 16S rRNA into the correct conformation and therefore correct maturation of the decoding centre prior to subunit joining (4,19). Era is a highly conserved protein, known to interact with the 3' end of the pre-16S rRNA (3) where it monitors the ribonuclease processing by fluorescence automated sequencing by GATC. For protein expression and purification, all pET28b derived plasmids were transformed into E. coli strain BL21 (DE3). All S. aureus plasmids were first electroporated into RN4220 Δspa, before isolation and electroporation into LAC* Δera.

GTPase assays
GTPase activity assays were performed as previously described (10). Briefly, the ability of proteins to hydrolyse GTP was determined by incubating 100 nM recombinant protein with 100 nM S. aureus 70S ribosomes, 1 μM GTP and 2.78 nM α-32 P-GTP in 40 mM Tris pH 7.5, 100 mM NaCl (100 mM KCl for RbgA), 10 mM MgCl2 at 37°C for the indicated times. All reactions were also set up in the absence of enzymes to monitor spontaneous GTP hydrolysis. Reactions were heat inactivated at 95°C for 5 mins to precipitate proteins and release bound nucleotide. Precipitated proteins were pelleted by centrifugation at 17,000 x g for 10 min. Reaction products were visualized by thin layer chromatography (TLC) in PEI cellulose TLC plates (Macherey-Nagel) and separated using 0.75 M KH2PO4, pH 3.6 buffer.
The radioactive spots were exposed to a BAS-MS Imaging Plate (Fujifilm), visualised using an LA 7000 Typhoon PhosphorImager (GE Healthcare), and images quantified using ImageQuant (GE Healthcare).

Synthesis of 32 P-(p)ppGpp, differential radial capillary action of ligand assays (DRaCALA)
The synthesis of (p)ppGpp and DRaCALA binding and competition assays were performed as described previously (17).

Protein purifications
Proteins were purified from 1-2 L E. coli BL21 DE3 cultures. Cultures were grown at 37°C to an OD600 of 0.5-0.7, expression was induced with 1 mM isopropyl β-D-1-thiogalactopyranoside (IPTG) and incubated for 3 h at 30°C. Cell pellets were resuspended in 5 ml Buffer A (50 mM Tris pH 7.5, 150 mM NaCl, 5% glycerol, 10 mM imidazole) and lysed by sonication upon addition of 20 μg/ml lysozyme and 30 μg/ml RNase A. Protein purifications were performed by nickel affinity chromatography. The filtered cell lysate was loaded onto a 1 ml HisTrap HP Ni 2+ column (GE Healthcare) before elution using a gradient of Buffer B (50 mM Tris pH 7.5, 200 mM NaCl, 5% glycerol, 500 mM imidazole). Protein containing fractions were dialysed in 50 mM Tris-HCl pH 7.5, 200 mM NaCl, 5% glycerol before concentration using a 10 kDa centrifugal filter (Thermo Scientific) and storage at -80°C. Protein for use in crystallography was dialysed into 25 mM Tris-HCl pH 7.5, 200 mM NaCl and used immediately.

30S, 50S and 70S ribosome purification
70S ribosomes were purified as described (17), with the following exceptions: following purification of mature 70S ribosomes, the ribosome pellet was resuspended in dissociation buffer (20 mM Tris pH 7.5, 120 mM NH4Cl, 1.5 mM MgCl2 and 2 mM β-mercaptoethanol), and quantified using the absorbance at 260 nm as described (26). 50 A260 units of 70S ribosomes were applied to a 10-40% continuous sucrose gradient made up in dissociation buffer and separated at 111,000 x g for 16 hours. Gradients were fractionated by upwards displacement of 250 µl aliquots, which were analysed for RNA content at an absorbance of 260 nm. Fractions containing 30S and 50S ribosomal subunits were pooled separately, and purification was continued as described (26).

In vitro ribosome association assays
500 nM recombinant 6xHis-tagged RA-GTPase was incubated at room temperature for 5 mins with 200 nM S. aureus 70S ribosomes in dissociation buffer (20 mM Tris pH 7.5, 120 mM NH4Cl, 1.5 mM MgCl2 and 2 mM β-mercaptoethanol) in the apo form and in the presence of 40 µM GTP, GMPPNP, GDP, ppGpp or pppGpp. The resultant reaction (150 µl) was layered onto a 10-40% continuous sucrose density gradient in dissociation buffer. Subsequently, gradients were centrifuged for 16 h at 111,000 x g in order to separate the 30S and 50S subunits. Gradients were fractionated by upwards displacement of 250 µl aliquots, which were analysed for RNA content at an absorbance of 260 nm. Fractions containing 30S and 50S ribosomal subunits were pooled separately and the protein content was precipitated by the addition of 10% v/v trichloroacetic acid (TCA) and incubation for 3 h at 4°C. Samples were centrifuged at 17,000 x g for 5 mins and washed twice with ice-cold acetone prior to drying of the pellets at 37°C for 10 mins. Pellets were resuspended in 2x SDS-PAGE sample buffer (62.5 mM Tris-HCl pH 6.8, 2% SDS, 10% glycerol, 0.01% bromophenol blue, 10% v/v β-mercaptoethanol), proteins were separated using a 10% SDS-PAGE gel and transferred onto a PVDF Immobilon-P membrane (Merck Millipore). The membrane was blocked with 5% w/v milk in TBST (50 mM Tris-HCl pH 7.6, 150 mM NaCl, 0.1% Tween 20), probed using 1:500 monoclonal anti-His HRP-conjugated antibodies (Sigma) and imaged using a ChemiDoc MP (Bio-Rad). Band densitometry was performed using ImageJ.

Growth and in vivo ribosome association assays
S. aureus strains were grown overnight in TSB containing the appropriate antibiotics. Overnight cultures were diluted to a starting OD600 of 0.05 in the presence of 100 ng/ml Atet and appropriate antibiotics and grown at 37°C with aeration, with OD600 values determined at 2 h intervals. For ribosome association assays, a culture of LAC* Dera pCN55iTET-era-his was split at an OD600 of 0.6 and fractions were either left uninduced or were induced with either 0.05 or 60 µg/ml mupirocin at 37°C for 30 mins. After growth, all cultures were incubated with 100 μg/ml chloramphenicol at 37°C for 3 mins, then cooled to 4°C. Cells were centrifuged at 4,000 x g for 10 mins and pellets resuspended to an OD600 of 35 in dissociation buffer (20 mM Tris pH 7.5, 120 mM NH4Cl, 1.5 mM MgCl2 and 2 mM β-mercaptoethanol). Cells were lysed through the addition of 0.5 µg/ml lysostaphin and 75 ng/ml DNase for 60 mins at 37°C. Lysates were centrifuged at 17,000 x g for 10 min to remove cell debris and 250 µl of the lysate was layered onto a 10-40% continuous sucrose gradient in dissociation buffer. Subunit separation was continued as per the in vitro method and associated C-terminally histidine-tagged Era (Era-His) was quantified via western blotting and band densitometry (ImageJ). Crude lysates were loaded alongside pulled-down protein to verify Era-His expression level. Staining of the blotting membrane with Ponceau S in 5% acetic acid was used to ensure consistent lysate loading prior to membrane blocking. Membranes were incubated with staining solution for up to 5 minutes and washed with distilled water until the background was clear.
Following imaging, the Ponceau S was removed by repeated wash steps using PBS.

Ribosome profiles from S. aureus cell extracts
Crude isolations of ribosomes from S. aureus cell extracts were achieved as described by Loh et al. with some modifications (27). Briefly, 100 ml cultures of the different S. aureus strains were grown to an OD600 of 0.4 in TSB medium with 100 ng/ml anhydrotetracycline (Atet). 100 μg/ml chloramphenicol was added to each culture and incubated for 3 min before being cooled to 4°C to enhance the pool of 70S ribosomes. Pelleted cells were suspended in association buffer (20 mM Tris-HCl pH 7.5, 8 mM MgCl2, 30 mM NH4Cl and 2 mM β-mercaptoethanol) and normalized to an OD600 of 15. Cells were lysed by the addition of 0.2 μg/ml lysostaphin and 75 ng/ml DNase and incubated for 60 min at 37°C. Cell debris was removed by centrifugation at 17,000 x g for 10 min. Clarified lysates (250 μl) were layered onto 10-50% discontinuous sucrose density gradients made in association buffer. Gradients were centrifuged for 7 h at 192,100 x g. Gradients were fractionated by upwards displacement of 250 μl aliquots, which were analysed for RNA content by absorbance at 260 nm.

Crystallisation of RsgA
The purified recombinant protein consisted of 311 residues, comprising 291 residues of S. aureus RsgA with an N-terminal 20 residue tag MGSSHHHHHHSSGLVPRGSH. It was simultaneously buffer , with the exception of electron density maps which were generated using COOT (28,29).

RsgA-ppGpp
The concentrated RsgA solution was supplemented with 2 mM MgCl2 and 2 mM ppGpp. Successful crystallisation was observed when this sample was mixed 1:1 with well solution containing 0.2 M sodium citrate tribasic dihydrate, 0.1 M Bis-Tris propane pH 6.5 and 20% (w/v) PEG 3350, and incubated at 17°C. Rod shaped crystal clusters appeared after a few days. Crystals were transferred to a cryoprotectant solution consisting of mother liquor with 15% ethylene glycol added and flash cooled in liquid N2. X-ray diffraction data were collected from a single crystal on beamline i04 at the Diamond Light Source national synchrotron facility at a wavelength of 0.97949 Å. The ppGpp-bound crystals diffracted to a resolution of 1.94 Å (PDB: 6ZHL). Initial processing was completed using the Xia2 pipeline (30). The crystals belonged to the space group P212121 (Supplementary Table S3). The structure of RsgA-ppGpp was solved via molecular replacement, using the previously published Bacillus subtilis homologue YloQ (PDB: 1T9H) as a model. The structure contained one RsgA monomer in the asymmetric unit. Molecular replacement was carried out using Phaser from within the CCP4 suite (31,32). The structure was refined via rounds of manual model building and refinement using COOT (29) and REFMAC5 (33). The final model was validated using MOLPROBITY (34). Residues 181-200 were lacking electron density and as such were omitted from the final model.

Apo RsgA
Crystallisation of apo RsgA was achieved when the concentrated protein sample was mixed 1:1 with well solution containing 0.15 M ammonium sulphate, 0.1 M MES pH 6.0 and 15% (w/v) PEG 4000 and incubated at 17°C. A single rod shaped crystal formed after a few weeks and diffracted to 2.01 Å resolution (PDB: 6ZJO). Initial processing was completed using the Xia2 pipeline and the crystal belonged to the space group P1211 (Supplementary Table S3). The structure was solved via molecular replacement as above using the available RsgA-ppGpp structure as a model with ligands removed and contained two RsgA monomers in the asymmetric unit. Iterative rounds of modelling, refinement and validation were carried out as above. Residues 180-200 (Chain A) and 179-200 (Chain B) were lacking electron density and as such omitted from the model. and subsequent dialysis to remove imidazole. Labelling efficiency was calculated in accordance with the fluorescent dye manufacturer's guidelines.

Statistics
Statistical analyses were performed using Graphpad Prism 8.0 software. Statistical differences between samples were assessed using one-way analysis of variance (ANOVA), followed by Tukey's multiple comparisons test.

RA-GTPases preferentially bind 5´ diphosphate-containing nucleotides GDP and ppGpp
The RA-GTPases RsgA, Era, RbgA and HflX can bind to the guanosine nucleotides GTP, GDP, ppGpp and pppGpp. Our previous work observed higher binding affinities for ppGpp over GTP, pointing towards a difference in binding between 5´ di-or triphosphate nucleotides (17). To examine the nucleotide binding affinities of these RA-GTPases for GDP in comparison to ppGpp, pppGpp and GTP, we used a differential radial capillary action of ligand assay (DRaCALA) ( Figure Table S4). This supports a previous observation that ppGpp is a more potent inhibitor of GTPase activity than pppGpp (17).
Structural data places (p)ppGpp within the GTP-binding site of the RA-GTPase RbgA (6).
Based on our measured affinities (Supplementary Table S4), we speculate that both GDP and ppGpp will out-compete other nucleotides for occupancy of the binding site. To examine this, competition assays were performed in which the binding of a radiolabelled nucleotide was challenged with an excess of unlabelled nucleotides ( Figure  This suggests that GTP occupancy, and hence activity, of these RA-GTPases is strongly dependent on the cellular excess of GTP over GDP and ppGpp, which occurs during exponential growth when ribosomal biogenesis is at its peak (25). This ratio changes during stationary phase and upon induction of the stringent response, when cellular GTP levels decrease with a concurrent rise in (p)ppGpp (12,38), shifting binding to favour a ppGpp-bound state. The greater affinity of these RA-GTPases to diphosphate-containing nucleotides would hence aid a rapid transition between the GTP-bound and ppGpp-bound states under conditions of stress.

Interactions with (p)ppGpp reduce the affinity of RA-GTPases for the ribosome
It is well characterised that rRNA transcription decreases during the stringent response (39). In addition, the GTPase activity of ribosome assembly cofactors is inhibited by (p)ppGpp, both of which contribute to a reduction in mature ribosomes within the cell (17). To examine mechanistically how (p)ppGpp-GTPase interactions affect the ability of RA-GTPases to associate with ribosomal subunits, we examined the association of each GTPase to either the 30S or 50S ribosomal subunit in the presence of GDP, GTP, ppGpp, pppGpp, as well as GMPPNP, a non-hydrolysable analogue of GTP. His-tagged GTPases were preincubated with highly pure, salt washed 70S S. aureus ribosomes in a low magnesium buffer to encourage ribosomal subunit dissociation, and the amount of each GTPase associated with each of the subunits was quantified by western immunoblot using anti-His antibodies after sucrose gradient separation ( Figure 2). In all cases, we observed a marked decrease in association of each GTPase to the 30S or 50S subunits in the presence of GDP, ppGpp and pppGpp compared to the GMPPNP-bound state (Figure 2A-D). For Era and HflX, there was a similar level of subunit association when in the apo, GTP or GMPPNP-bound states, compared to a 2-fold reduction in ribosome binding when incubated with GDP, ppGpp or pppGpp ( Figure 2C, 2D), suggesting that these GTPases can associate to the ribosome in the unbound state. The ability of Era to bind the 30S in the absence of nucleotides has been reported previously, where it has been suggested that the apo form can bind to a secondary site (3,40). The patterns exhibited by RsgA and RbgA were slightly different, with strong binding in the GMPPNP-bound state, whereas 3-6 fold weaker binding was observed in the apo, GTP, GDP, ppGpp and pppGpp-bound states (Figure 2A, 2B). It is worth noting that previous studies have suggested that the association of RbgA with the 50S subunit is enhanced in the presence of pppGpp (24), a finding that is not replicated here. The apparent effect of ppGpp and pppGpp on ribosome association was comparable, which is not reflective of the differences in affinity  Table S4). From these data, we show that GTP binding favours association of RsgA, RbgA, Era and HflX to ribosomal subunits, and that this interaction is inhibited when in the GDP-, ppGpp-or pppGpp-bound states.

Binding kinetics of RA-GTPase-ribosome interactions
To gain further insight into the binding mechanism and how (p)ppGpp reduces the association of RA-GTPases to the ribosomal subunits, we used a stopped-flow technique with fluorescent derivatives of the RA-GTPases ( Figure 3A). Structural predictions of all four R A-GTPases were built by homology modelling using available structures to assess the availability of suitable residues for fluorescence labelling (Supplementary Figure S3A, S3B) (41). Both RbgA and HflX were amenable to covalent linkage to the fluorophore Atto-488 using maleimide chemistry with exposed cysteine residues. RbgA contains one wild-type cysteine residue (C277) that is surface exposed in the B. subtilis crystal structure (PDB: 1PUJ) and is located towards the C-terminus of the protein (Supplementary Figure S3A). Based on the E. coli structure (PDB: 5ADY), HflX contains two cysteines (Supplementary Figure S3B). C330 is predicted to be surface exposed and therefore amenable to labelling, while C45 is buried and is expected to show low accessibility for fluorescent labelling. Era, on the other hand, lacks any cysteine residues, while RsgA contains three conserved cysteine residues that coordinate the Zn 2+ ion within the For two-step reactions, the apparent rate under conditions tested, kapp1, is expected to increase linearly with increasing ligand concentration. On the other hand, kapp2 is expected to align to a hyperbolic relationship as ligand concentration increases. This was the case for HflX complexed with GTP ( Figure   3D, 3E). Thus, productive binding of the RA-GTPase appears to occur through two steps. When HflX was incubated with ppGpp, the kapp1 increased linearly ( Figure 3D), while kapp2 did not depend on ribosome concentration ( Figure 3E), indicating that ppGpp hampers the accommodation step of the binding mechanism. On the other hand, if HflX was complexed with pppGpp, neither kapp depended on 50S concentration, indicating that the alarmone drastically affects the mechanism of HflX binding. In this case, the reaction appears rate-limited by an isomerization step of the RA-GTPase at 5 s -1 ( Figure   3D). The linear increase in kapp1 was 2-fold greater for GTP than for ppGpp or pppGpp ( Figure 3D), suggesting a greater rate of the fast-phase reaction. The kapp2 of the GTP-bound form showed a hyperbolic relationship tending to 2 s -1 , while the linear relationship when bound to ppGpp was steady at 1.0 s -1 ( Figure 3E). This suggests that the second, slow-phase reaction is taking place while HflX is bound to GTP but is reduced 4-fold when bound to ppGpp. Additionally, this suggests that one or more of the microscopic constants which contribute to the kapp2 in the two-step association reaction remains incomplete while in the ppGpp-bound state.
Next, we used the sum and product of the kapp1 and kapp2 of each reaction ( Figure 3F, 3G) to estimate the microscopic constants defining the reaction for the GTP-and ppGpp-bound HflX (Supplementary Table S5). ppGpp reduced the value of the initial binding constant k1 by 2-3 fold, while drastically affecting k2, indicating that the alarmone hampers proper accommodation of HflX on the subunit ( Figure 3F, 3G, Supplementary Table S5). On the contrary, the dissociation rate constants k-1 and k-2 appeared less affected by ppGpp, remaining similar to those observed during the GTP-bound state (Supplementary Table S5). Altogether, our data indicates that (p)ppGpp induces a non-productive conformation of HflX, reducing the binding progression with the ribosomal subunit.
In the case of RbgA, all three tested nucleotides adhered to a two-step mechanism model, with kapp1 increasing linearly with 50S concentration, while kapp2 appeared hyperbolic (Supplementary Figure   S5A, S5B). Further analysis to estimate the microscopic constants indicated that (p)ppGpp increased the dissociation rate constant k-1 by 3-5-fold as compared to GTP, whereas its association velocity k1 appeared largely unaffected (Supplementary Table S5, Supplementary Figure S5C, S5D). Interestingly, ppGpp drastically reduced the accommodation rate constant k2 similarly to HflX, while pppGpp did not.
Altogether, our results indicate that (p)ppGpp can program RbgA to adopt different conformations that ultimately reduce their binding affinity for the ribosome (Supplementary Table S5).
For both RA-GTPases, the Kd of 50S binding is lower in the GTP-bound state compared to the (p)ppGpp-bound state (Supplementary Table S5). It appears that the main difference on a kinetic level, in agreement with our previous observations regarding the accommodation step, is that the binding of

Association of the RA-GTPase Era to the 30S subunit decreases upon induction of the stringent response
Upon induction of the stringent response, cellular levels of (p)ppGpp increase, while concentration of GTP drops (38). Having observed decreased association of RA-GTPases to ribosomal subunits in vitro, we wished to examine the interaction under more physiologically relevant conditions. To investigate RA-GTPases interactions with the ribosome in the bacterial cell, we used an era deletion mutant in the community-acquired methicillin-resistant S. aureus (CA-MRSA) strain LAC* that was available to us.
This strain has a growth defect ( Figure 4A) and has an abnormal cellular ribosomal profile when compared to the wild-type, with an accumulation of 50S subunits and a loss of 70S ribosomes ( Figure   4B, 4C) (10,27,42), suggesting that the absence of this GTPase is preventing mature ribosome formation and growth. In order to establish whether induction of the stringent response in bacterial cells leads to a decrease in the association of Era to the 30S subunit, the mutant was complemented with an anhydrotetracycline-inducible 6xHis-tagged version of era, yielding strain LAC* Dera iTET-era-His.
Having confirmed that the His-tagged version of the protein is expressed and restores the growth defect observed in era mutant strains ( Figure 4A), we grew cells to exponential phase and induced the stringent response with mupirocin, an antibiotic that inhibits isoleucyl tRNA synthetase and is known to activate the stringent response in S. aureus (43). Cells were lysed and applied to 10-40% sucrose gradients in ribosome dissociation buffer for subunit separation via isopycnic ultracentrifugation. The 30S pool was analysed for associated Era-His using a-His western immunoblotting ( Figure 4D). Crude lysates sampled prior to loading on the sucrose gradients were also analysed to ensure equal loading and equal expression of Era-His between samples (Supplementary Figure S6). In agreement with the in vitro data, the relative association of Era-His to the ribosome decreased at least 4-fold upon induction of the stringent response ( Figure 4D). Altogether this in vitro and in vivo data support a model in which the stringent response impairs 70S ribosome assembly by disrupting the association of RA-GTPases with the immature ribosomal subunits, thus preventing correct ribosome maturation.

Crystallisation of RsgA in the apo and ppGpp-bound states
GTPases act as molecular switches, cycling between OFF (GDP-bound) and ON (GTP-bound) states.
Structural studies of numerous GTPases have reported distinct conformations for both states, which are determined by the movement of the flexible switch I/G2 loop and the switch II/G3 loop (44). Often described as a loaded-spring mechanism, the conformational change occurs upon hydrolysis of GTP or the subsequent g-phosphate release. Both switch I/G2 and switch II/G3 are responsible for coordinating the Mg 2+ cofactor which interacts with the g-phosphate of GTP via a conserved threonine residue in G2 and a glycine in G3. Upon hydrolysis of the g-phosphate and Pi dissociation, the protein relaxes into the OFF conformation.
To look more at the mechanism of (p)ppGpp-mediated inhibition of RA-GTPases associating with ribosomal subunits, we solved the structures of RsgA in both the apo-( Figure 5A) and ppGppbound ( Figure 5B) states by X-ray crystallography (Supplementary Table S3 Figure 5A). Both the OB-fold and ZNF domains are involved in nucleotide recognition (45,46), and target RsgA to the 30S ribosomal subunit where they contact major helices of the 16S rRNA ( Figure 5C). The OB-fold is situated between h18 and h44, with the loop connecting β1 and β2 recognising the minor groove of h44 adjacent to the 30S acceptor site (4). The ZNF contacts the 30S head domain, making backbone contacts with h29 and h30, close to the interaction site of the P-site tRNA (4,47). In E. coli RsgA (YjeQ), the G-domain also contacts h44 by means of a clamp adjacent to the interaction site of h45 and h24. This clamping interaction is facilitated by the β6,7 hairpin and the switch I/G2 region (4), however this hairpin is lacking in S. aureus RsgA ( Figure 5A, 5B), so it is likely that the G-domain interacts with h44 singly through the switch I/G2 region.
The ppGpp ligand is bound in an elongated conformation, where the 3´ and 5´-phosphate moieties face away from each other (Supplementary Figure S7A).

ppGpp-bound RsgA mimics the GDP-bound OFF-state conformation
For RsgA, a catalytic histidine residue is located within the switch I/G2 loop, two residues upstream of the conserved G2 threonine (4). Therefore, correct docking of this region upon binding to either GTP or the 16S rRNA is thought to be instrumental for GTPase activity. It has also been previously proposed by Pausch et al. (6) that for RbgA, the 3´-diphosphate of (p)ppGpp prevents the movement of switch I/G2 into the ON conformation necessary for GTP hydrolysis and ribosome binding, explaining why the GTPase is incapable of hydrolysing (p)ppGpp in a similar manner to GTP (6). In order to determine whether a similar steric inhibition is occurring for RsgA, we compared our apo and ppGpp-bound structures with available structures of RsgA homologues, namely Aquifex aeolicus YjeQ bound to GDP (PDB: 2YV5) and E. coli YjeQ complexed with both the 30S subunit and GMPPNP (PDB: 5UZ4 (47)) ( Figure 6). Importantly, in both of these available structures, the switch I/G2 loops were partially resolved ( Figure 6A, 6B). Despite a similar overall fold of the G-domain, the switch I/G2 loop in the GDP-bound structure appears to extend distally from the main body of the protein, far from the associated ligand ( Figure 6A). Contrary to this, the GMPPNP-bound structure features a fully docked Switch I/G2 loop, positioned adjacent to the bound ligand and the binding site of the Mg 2+ ion, although the Mg 2+ ion itself is not resolved. Crucially, in this conformation, the docked switch I/G2 loop occupies the same space that the 3´-diphosphate moiety of ppGpp would ( Figure 6B, 6D). Additionally, the switch II/G3 loop conformation differs between the GDP-and GMPPNP-bound structures, being extended towards the γ-phosphate of GMPPNP in the latter. When compared to our apo ( Figure 6C) and ppGpp-bound ( Figure 6D) structures, the switch II/G3 region appears highly similar to that of the GDP-bound structure, leading us to hypothesise that the switch I/G2 loop will also adopt a similar conformation to the GDP-bound state due to steric inhibition by ppGpp. This lack of docking of switch I/G2 would inhibit GTPase activity by preventing proper docking of the catalytic histidine within switch I (4), coordination of the Mg 2+ cofactor by the G2 threonine (6), and subsequent interaction with the γ-phosphate of GTP.

Displacement of the G2 loop by (p)ppGpp inhibits RA-GTPase-ribosome interactions
The structure of RsgA in the GMPPNP-bound ON state has only ever been solved when associated with the 30S ribosomal subunit suggesting it is stabilised in this conformation (4,47). In order to assess the role of the switch I/G2 loop in ribosome association, we performed computational Ca alignments of both the available GDP-bound (PDB: 2YV5) and our ppGpp-bound structures with the GMPPNP-bound RsgA-30S ribosome complex (PDB: 5UZ4) ( Figure 7A-C). It has previously been shown that each of the 3 domains of RsgA interact with rRNA to provide a stable docking interaction ( Figure 5C) (4), and that for E. coli RsgA, the switch I/G2 loop and a b6, b7-hairpin clamp around h44, contacting the minor and major groove respectively ( Figure 7A). However, when the GDP-bound OFF-state structure from A. aeolicus is superimposed in place of the GMPPNP structure, it appears that the switch I/G2 loop is positioned in such a way that would cause steric clashing between the phosphate backbone of h44 ( Figure 7B). Likewise, the expected position of the switch I/G2 loop in the ppGpp-bound model would lead to similar steric clashing, with the 3′-diphosphate moiety of ppGpp preventing the switch I /G2 loop adopting the active conformation ( Figure 7C). While it is important to stress that this modelling is performed using protein models and 30S subunits from separate organisms, this leads us to hypothesise that the misalignment of the switch I/G2 loop and subsequent steric clashing between the RA-GTPase and h44 of the 16S rRNA could be responsible for (p)ppGpp-mediated inhibition of RA-GTPase association to the ribosome. We suggest that this region is not directly responsible for promoting subunit docking, however that the switch I region instead forms electrostatic interactions with conformationally mature h44 and h45 rRNA following ribosome association, enabling positioning of the switch I/G2 loop in a catalytically active conformation when the mature rRNA conformation is reached.
These interactions and the subsequent loop rearrangement may represent the slow stabilisation step (k2) observed in our stopped flow analysis (Figure 3).

DISCUSSION
The stringent response is a multi-faceted stress coping mechanism, ubiquitously used throughout the Bacteria to cope with nutrient starvation conditions. Recent transcriptomics data has highlighted the diversity and complexity of this response, with 757 genes being differentially regulated within 5 minutes of (p)ppGpp induction (15). For Gram-positive bacteria, the regulation of transcription by (p)ppGpp is intricately linked to purine nucleotide levels, which are impacted in a number of ways (48). Upon induction of the stringent response, GTP/GDP and ATP levels decrease as they are utilised by (p)ppGpp synthetase enzymes (12). Furthermore, once produced (p)ppGpp directly inhibits a number of enzymes involved in the guanylate and adenylate synthesis pathways, further reducing GTP/GDP levels (38,49). All of this results in a shift from high GTP/GDP and low (p)ppGpp levels in fast growing cells, to low GTP/GDP and high (p)ppGpp in nutritionally starved cells. For S. aureus, the impacts of this are wide-reaching, affecting transcription initiation (39), enzyme activities (50) and, as we show here, the regulation of the activity of RA-GTPases by tuning their capacity to interact with ribosomal subunits.
In the present work, we examine the nucleotide binding preferences of RA-GTPases, and the consequences of this binding on regulating the interactions of RA-GTPases with the ribosome. Cycling between the GTP-bound ON and GDP-bound OFF states is critically important for RA-GTPases, as it enables these proteins to act as molecular checkpoints of ribosome assembly. Here we show that RA-GTPases bind to guanosine nucleotides competitively and with differing affinities, with GDP and ppGpp binding with up to 6-times greater affinity than their 5´ trinucleotide-containing counterparts GTP and pppGpp (Supplementary Table S4). The consequence of differing nucleotide-bound states for interactions with ribosomal subunits are significant. We observe that GTP binding is required to promote RA-GTPase/ribosome interactions (Figure 2 & 3). Indeed, the binding of apo RbgA and HflX to the 50S subunit was almost undetectable by stopped-flow fluorescence ( Figure 3B, 3C), although Era and HflX did demonstrate background binding to the 30S and 50S subunits respectively by western immunoblotting. A cryo-electron micrograph (cryo-EM) structure of Era binding to the 30S subunit has previously been solved (40) and ObgE bind to ppGpp in a ring-like conformation (53)(54)(55), in which the 3´ and 5´ phosphate moieties point towards each other. While no structural reasoning for this difference in conformation is known, aside from to extend the breadth of responses controlled by (p)ppGpp, it has been suggested that proteins which bind (p)ppGpp in the ring-like conformation have 10-fold lower inhibitory constants and dissociation constants than those which bind in the elongated conformation (56,57). This could potentially influence the temporal or energetic threshold during the stringent response where a certain protein becomes inhibited, based on decreasing concentrations of GTP and increasing concentrations of (p)ppGpp (38).
Ribosomal rRNA production and biogenesis are not the only aspects of protein synthesis that (p)ppGpp regulates, given its ability to bind to the bacterial initiation factor 2 (IF2), elongation factor Tu (EF-Tu), elongation factor G (EFG), elongation factor Ts (EFTs) and release factor 3 (RF3) (58)(59)(60)(61)(62). In each case, competitive binding of (p)ppGpp to these GTPases results in an inhibition of activity and reduction of the elongation cycle. Unlike RA-GTPases involved in subunit maturation, both IF2 and EFG bind to GTP, GDP and (p)ppGpp with similar affinity (59,60,63), albeit with EFG demonstrating an overall lower affinity. Furthermore, IF2 binding to (p)ppGpp within the 30S pre-initiation complex alters the mRNA binding preference, enabling permissive translation of certain mRNAs such as tufA encoding EF-Tu (58), which may fine-tune the proteome on a translational level to better enable survival of nutrient deprivation.
With this work we have used complementary techniques to demonstrate that (p)ppGpp prevents stable association of RA-GTPases to the ribosome, both in vitro and within the bacterial cell. This is achieved by these enzymes having a stronger affinity for ppGpp over GTP, with ppGpp interactions holding these enzymes in an OFF-state conformation. Consequently, this imparts delays to 70S ribosome assembly, which in turn contributes to the growth defects that are observed upon induction of the stringent response. Altogether, we highlight RA-GTPases-(p)ppGpp interactions as important regulators of stringent response-mediated growth control.

DATA AVAILABILITY
The coordinates and electron density maps of RsgA-apo and RsgA-ppGpp have been deposited in the

CONFLICTS OF INTEREST
The authors declare no conflicts of interest.   Fluorescently labelled protein was also mixed with buffer lacking 50S subunits as a mixing control.
Fluorescence of the reaction was tracked using exponential sampling for 10 seconds and each curve is the mean average of at least 5 technical replicates. D) kapp1 dependence on 50S concentration for HflX complexed with GTP (green), ppGpp (pink), pppGpp (black). E) as (D) for the kapp2 dependence.  Figure   S4) were analysed by nonlinear regression using two exponential terms. The sum (F) and product (G) of apparent rates (kapp1 (D), kapp2 (E)) were plotted as a function of the total concentration of the 50S subunits and HflX protein to determine the microscopic constants k1, k-1, k2, and k-2 (Supplementary  Table S5)   Ribosomal subunits were separated and the amount of Era-His associated was detected using HRPconjugated α-His antibodies. Experiments were carried out in triplicate and one representative image is shown. Bottom: the mean signal intensities relative to the zero mupirocin sample of all repeats were plotted with error bars representing standard deviation. Statistical analysis was carried out using a oneway ANOVA followed by Tukey's multiple comparisons test (*** P < 0.001). was overlaid onto the model of YjeQ-GMPPNP (PDB: 5UZ4, chain Z (47)) using Cα alignment, relative to the 30S ribosomal subunit (PDB: 5UZ4, chain A (47)).