Sequences were aligned with MAFFT v7.164b with the L-INS-i strategy (Katoh and Standley, 2013), and alignments were visualized with Jalview (Waterhouse et al., 2009).

The strains and plasmids used in this study are listed in Supplementary Tables S1–S3. Oligonucleotides used in this study are provided in Supplementary Table S4. A detailed description of strain construction is provided in the Supplementary Material.

Escherichia coli BW25113 cells were transformed with expression constructs either based on a high-copy IPTG inducible vector pUC derivative pMG25 (pMG25:relA, Turnbull et al., 2019, pMG25:relA^ΔRRM, pMG25:spoR or pMG25:spoT^ΔRRM) or on a low-copy IPTG inducible vector, mini R1 plasmid pNDM220 which is present in one to two copies per chromosome (Molin et al., 1979) (pNDM220:relA, pNDM220:relA^ΔRRM, pNDM220:relA^{ΔRRM–H432E}, pNDM220:relA^{ΔRRM–R629E} or pNDM220:relA^{ΔRRM–C612A/C613A}). For solid medium growth assays, ten-fold serial dilutions of overnight LB cultures were spotted onto LB agar supplemented with 30 μg/mL ampicillin and 1 mM IPTG. For liquid medium growth assays, 1000-fold dilutions of the overnight LB cultures were made in liquid LB supplemented with 30 μg/mL ampicillin and 1 mM IPTG, seeded on a 100-well honeycomb plate (Oy Growth Curves AB Ltd, Helsinki, Finland), and plates incubated in a Bioscreen C (Labsystems, Helsinki, Finland) at 37°C with continuous medium shaking.

Bacillus subtilis strains were pre-grown on LB plates lacking the IPTG inducer overnight (10 h) at 30°C. Fresh individual colonies were used to inoculate filtered LB medium in the presence of indicated concentrations of IPTG and OD₆₀₀ adjusted to 0.01. The cultures were seeded on a 100-well honeycomb plate (Oy Growth Curves AB Ltd, Helsinki, Finland), and plates were incubated in a Bioscreen C (Labsystems, Helsinki, Finland) at 37°C with continuous medium shaking.

Growth rates (μ₂) were calculated as slopes of linear regression lines through log₂-transformed OD₆₀₀ data points.

The entire coding region of the B. subtilis rel gene was amplified by PCR using the synthetic oligonucleotide pQErelA_F and pQErelA_R containing a BamHI site and B. subtilis genomic DNA as a template. The resulting PCR fragment was cut with BamHI and then inserted into the BamHI sites of pQE60 (Qiagen), yielding plasmid pQErelA. pOErelA was transformed into E. coli M15 (pREP4) (Qiagen), fresh transformants were inoculated into LB liquid culture (1000 mL) with 100 μg/mL ampicillin and grown at 37°C with vigorous shaking. At OD₆₀₀ of 0.8 expression of Rel induced with 1 mM IPTG (final concentration). After 3 h of expression the cells were harvested by centrifugation, resuspended in buffer A (500 mM NaCl, 50 mM Tris-HCl pH 8.0) supplemented with 2 mM PMSF and lysed by sonication. Rel-His₆ inclusion bodies were collected by centrifugation, resuspended in buffer A supplemented with 8 M guanidine hydrochloride (= buffer B) and loaded onto an Ni-NTA agarose column (QIAGEN) pre-equilibrated in the same buffer. The column was washed with buffer B supplemented with 10 mM imidazole, and the protein was eluted with a 100–400 mM imidazole gradient in buffer B. The fractions containing Rel-His₆ was dialyzed against buffer A at 4°C overnight. Aggregated Rel-His₆ protein was collected by centrifugation, resuspended in buffer A supplemented with 6 M urea and used for rabbit immunization. Rabbit serum was used as a polyclonal anti-Rel antibody.

Bacillus subtilis strains were pre-grown on LB plates overnight at 30°C. Fresh individual colonies were used to inoculate 200 mL LB cultures that were grown at 37°C. At OD₆₀₀ of 0.2 amino acid starvation was induced by addition of isoleucyl tRNA synthetase inhibitor mupirocin (dissolved in DMSO, AppliChem) to final concentration of 700 nM for 20 min. As a mock control, a separate culture was treated with the same amount of DMSO. After 20 min the cells were collected by centrifugation (8,000 rpm, 5 min, JLA-16.25 Beckman Coulter rotor), dissolved in 0.5 mL of HEPES:Polymix buffer [5 mM Mg(OAc)₂] supplemented with 2 mM PMSF, lysed using FastPrep homogenizer (MP Biomedicals) by four 20 s pulses at speed 6.0 mp/s with chilling on ice for 1 min between the cycles), and clarified by ultracentrifugation (14,800 rpm for 20 min, Microfuge 22R centrifuge Beckman Coulter, F241.5P rotor). Clarified cell lysates were loaded onto 10-35% sucrose gradients in HEPES:Polymix buffer pH 7.5 (5 mM Mg²⁺ final concentration), subjected to centrifugation (36,000 rpm for 3 h at 4°C, SW-41Ti Beckman Coulter rotor) and analyzed using Biocomp Gradient Station (BioComp Instruments) with A₂₆₀ as a readout.

For Western blotting 0.5 mL fractions were supplemented with 1.5 mL of 99.5% ethanol, precipitated overnight at –20°C. After centrifugation at 14,800 rpm for 30 min at 4°C the supernatants were discarded and the samples were dried. The pellets were resuspended in 40 μL of 2xSDS loading buffer [100 mM Tris-HCl pH 6.8, 4% SDS (w/v) 0.02% Bromophenol blue, 20% glycerol (w/v) 4% β-mercaptoethanol], resolved on the 8% SDS PAGE and transferred to nitrocellulose membrane (Trans-Blot Turbo Midi Nitrocellulose Transfer Pack, Bio-Rad, 0.2 μm pore size) with the use of a Trans-Blot Turbo Transfer Starter System (Bio-Rad) (10 min, 2.5 A, 25 V). Membrane blocking was done for 1 h in PBS-T (1xPBS 0.05% Tween-20) with 5% w/v non-fat dry milk at room temperature. Rel was detected using anti-Rel primary combined with goat anti-rabbit IgG-HRP secondary antibodies. All antibodies were used at 1:10,000 dilution. ECL detection was performed using WesternBright^TM Quantum (K-12042-D10, Advansta) Western blotting substrate and an ImageQuant LAS 4000 (GE Healthcare) imaging system.

Wild type and H432E mutant variants of E. coli RelA were expressed and purified as described earlier (Turnbull et al., 2019).

Wild type and mutant variants of B. subtilis Rel were overexpressed in freshly transformed E. coli BL21 DE3 Rosetta (Novagen). Fresh transformants were inoculated to final OD₆₀₀ of 0.05 in the LB medium (800 mL) supplemented with 100 μg/mL kanamycin. The cultures were grown at 37°C until an OD₆₀₀ of 0.5, induced with 1 mM IPTG (final concentration) and grown for additional 1.5 h at 30°C. The cells were harvested by centrifugation and resuspended in buffer A (750 mM KCl, 5 mM MgCl₂, 40 μM MnCl₂, 40 μM Zn(OAc)₂, 1 mM mellitic acid (Tokyo Kasei Kogyo Co., Ltd.), 20 mM imidazole, 10% glycerol, 4 mM β-mercaptoethanol, 25 mM HEPES:KOH pH 8) supplemented with 0.1 mM PMSF and 1 U/mL of DNase I. Cells were lysed by one passage through a high-pressure cell disrupter (Stansted Fluid Power, 150 MPa), cell debris was removed by centrifugation (25,000 rpm for 40 min, JA-25.50 Beckman Coulter rotor) and clarified lysate was taken for protein purification.

To prevent possible substitution of Zn²⁺ ions in Rel’s Zn-finger domain for Ni²⁺ during purification on an Ni-NTA metal affinity chromatography column (Block et al., 2009), a 5 mL HisTrap HP column was stripped from Ni²⁺ in accordance with manufacturer’s recommendations, washed with 5 column volumes (CV) of 100 mM Zn(OAc)₂ pH 5.0 followed by 5 CV of deionized water. Clarified cell lysate was filtered through a 0.2 μm syringe filter and loaded onto the Zn²⁺-charged HisTrap 5 mL HP column pre-equilibrated in buffer A. The column was washed with 5 CV of buffer A, and the protein was eluted with a linear gradient (6 CV, 0-100% buffer B) of buffer B (750 mM KCl, 5 mM MgCl₂, 40 μM MnCl₂, 40 μM Zn(OAc)₂, 1 mM mellitic acid, 500 mM imidazole, 10% glycerol, 4 mM β-mercaptoethanol, 25 mM HEPES:KOH pH 8). Mellitic acid forms highly ordered molecular networks when dissolved in water (Inabe, 2005) and it was shown to promote the stability of Thermus thermophilus Rel (Van Nerom et al., 2019). Fractions most enriched in Rel (≈25–50% buffer B) were pooled, totaling approximately 5 mL. The sample was loaded on a HiLoad 16/600 Superdex 200 pg column pre-equilibrated with a high salt buffer (buffer C; 2 M NaCl, 5 mM MgCl₂, 10% glycerol, 4 mM β-mercaptoethanol, 25 mM HEPES:KOH pH 8). The fractions containing Rel were pooled and applied on HiPrep 10/26 desalting column (GE Healthcare) pre-equilibrated with storage buffer (buffer D; 720 mM KCl, 5 mM MgCl₂, 50 mM arginine, 50 mM glutamic acid, 10% glycerol, 4 mM β-mercaptoethanol, 25 mM HEPES:KOH pH 8). Arginine and glutamic acid were added to improve protein solubility and long-term stability (Golovanov et al., 2004). The fractions containing Rel were collected and concentrated in an Amicon Ultra (Millipore) centrifugal filter device (cut-off 50 kDa). To cleave off the His₁₀-SUMO tag, 35 μg of His₆-Ulp1 per 1 mg of Rel were added and the reaction mixture was incubated at room temperature for 15 min. After the His₁₀-SUMO tag was cleaved off, the protein was passed though 5 mL Zn²⁺-charged HisTrap HP pre-equilibrated with buffer D. Fractions containing Rel in the flow-through were collected and concentrated on Amicon Ultra (Millipore) centrifugal filter device with 50 kDa cut-off. The purity of protein preparations was assessed by SDS-PAGE and spectrophotometrically [OD₂₆₀/OD₂₈₀ ratio below 0.8 corresponding to less than 5% RNA contamination (Layne, 1957)]. Protein preparations were aliquoted, frozen in liquid nitrogen and stored at −80°C. Individual single-use aliquots were discarded after the experiment.

3.5 μL of 2 μM Rel protein was loaded onto a glow-discharged Cu₃₀₀ grid (TAAB Laboratories Equipment Ltd.) with manually layered 2.9 nm carbon. The sample was incubated on the grid for 1–3 min, blotted with Watman filter paper, than twice washed with water and blotted, stained with 1.5% uranyl acetate pH 4.2 for 30 s before the final blotting. Grids were dried on the bench and imaged by Talos L 120C (FEI) microscope with 92,000X magnification.

The 10X Polymix base was prepared as per (Antoun et al., 2004), with minor modifications. For preparation of the putrescine solution, 100 g of putrescine (1,4-diaminobutane) was dissolved in 600 mL of ddH₂O at 90°C, and the pH adjusted with acetic acid to 8.0 (approximately 100 mL of 100% acetic acid). After cooling to room temperature, the pH was adjusted further to 7.6 and the volume was adjusted to the final of 2 L by addition of 1.134 L of ddH₂O. One 100 mL cup of activated charcoal was added and the slurry was stirred under the hood for 30 min. The slurry was filtered through, first, Whatman paper and then through a 0.45 μm BA85 membrane. The final solution was stored at 4°C in a bottle wrapped in foil since putrescine is photosensitive. The preparation of 2 L of 10X Polymix buffer base used 141.66 g KCl, 5.35 g NH₄Cl, 21.44 g Mg(OAc)₂·4H₂O, 1.47 g CaCl₂·2H₂O, 5.092 g spermidine, and 160 ml of putrescine solution (described above). The salts were dissolved in ddH₂O (≈1,500 mL), then the putrescine solution was added and mixed well. Spermidine was dissolved in a small volume of ddH₂O and added to the mixture. The pH was adjusted to 7.5 with concentrated acetic acid or 5 M KOH, and after that the volume was adjusted by adding ddH₂O to 2 L. The buffer was filtered through 0.2 μm nitrocellulose filter (2–3 filters are needed). The resulting 10X Polymix buffer base was aliquoted and stored at −20°C. The final working HEPES:Polymix buffer was made using the 10X Polymix buffer base, 1M DDT and 1 M HEPES:KOH pH 7.5 and contains 20 mM HEPES:KOH pH 7.5, 2 mM DTT, 5 mM MgOAc₂, 95 mM KCl, 5 mM NH₄Cl, 0.5 mM CaCl₂, 8 mM putrescine, 1 mM spermidine.

Bacillus subtilis strain RIK2508 (trpC2 Δhpf) strain (Akanuma et al., 2016; Brodiazhenko et al., 2018) was pre-grown on LB plates overnight at 30°C. Fresh individual colonies were used to inoculate LB liquid cultures (25 × 400 mL) to OD₆₀₀ of 0.05 and grown at 37°C with vigorous shaking. At OD₆₀₀ 1.2 the cells were pelleted at 4°C (TLA10.500 (Beckman), 15 min at 5,000–8,000 rcf), resuspended with ice-cold PBS buffer, pelleted again in 50 mL Falcon tubes, frozen with liquid nitrogen and stored at –80°C. Approximately 20 g of frozen B. subtilis cells were resuspended in 50 mL of cell opening buffer (100 mM NH₄Cl, 15 mM Mg(OAc)₂, 0.5 mM EDTA, 3 mM β-mercaptoethanol, 20 mM Tris:HCl pH 7.5) supplemented with 1 mU Turbo DNase (Thermo Fisher Scientific), 0.1 mM PMSF and 35 μg/mL lysozyme, incubated on ice for 1 h, and opened by three passages on a high-pressure cell disrupter (Stansted Fluid Power) at 220 MPa. Lysed cells were clarified by centrifugation for 40 min at 40,000 rpm (Ti 45 rotor, Beckman), NH₄Cl concentration was adjusted to 400 mM, and the mixture was filtered through 0.45 μm syringe filters. The filtrated lysate was loaded onto a pre-equilibrated 80 mL CIMmultus QA-80 column (BIA Separations, quaternary amine advanced composite column) at a flow rate of 20 mL/min, and the column washed with 5 CV (CV = 80 mL) of low salt buffer [400 mM NH₄Cl, 15 mM Mg(OAc)₂, 3 mM β-mercaptoethanol, 20 mM Tris:HCl pH 7.5]. Ribosomes were then eluted in 45 mL fractions by a step gradient to 77% high salt buffer [900 mM NH₄Cl, 15 mM Mg(OAc)₂, 3 mM β-mercaptoethanol, 20 mM Tris:HCl pH 7.5] for 5 CV, followed by 100% high salt buffer for 1 CV. The fractions containing ribosomes were pooled, the concentration of NH₄Cl was adjusted to 100 mM, and the ribosomes were treated with puromycin added to a final concentration of 10 μM. The resultant crude 70S preparation was resolved on a 10–40% sucrose gradient in overlay buffer [60 mM NH₄Cl, 15 mM Mg(OAc)₂, 0.25 mM EDTA, 3 mM β-mercaptoethanol, 20 mM Tris:HCl pH 7.5] in a zonal rotor (Ti 15, Beckman, 17 h at 21,000 rpm). The peak containing pure 70S ribosomes was pelleted by centrifugation (20 h at 35,000 rpm), and the final ribosomal preparation was dissolved in HEPES:Polymix buffer [20 mM HEPES:KOH pH 7.5, 2 mM DTT, 5 mM Mg(OAc)₂, 95 mM KCl, 5 mM NH₄Cl, 0.5 mM CaCl₂, 8 mM putrescine, 1 mM spermidine (Antoun et al., 2004)]. 70S concentration was measured spectrophotometrically (1 OD₂₆₀ corresponds to 23 nM of 70S) and ribosomes were aliquoted (50–100 μL per aliquot), snap-frozen in liquid nitrogen and stored at –80°C.

Initiation complexes were prepared by as per (Murina et al., 2018a, b), with minor modifications. The reaction mix containing B. subtilis 70S ribosomes (final concentration of 6 μM) with E. coli IF1 (4 μM), IF2 (5 μM), IF3 (4 μM), ³H-fMet-tRNA_i^fMet (8 μM), mRNA MVFStop (8 μM, 5′-GGCAAGGAGGAGAUAAGAAUGGUUUUCUAAUA-3′; Shine-Dalgarno sequence is highlighted in bold, ORF is underlined), 1 mM GTP and 2 mM DTT in 1 × HEPES:Polymix buffer [20 mM HEPES:KOH pH 7.5, 95 mM KCl, 5 mM NH₄Cl, 5 mM Mg(OAc)₂, 0.5 mM CaCl₂, 8 mM putrescine, 1 mM spermidine, 1 mM DTT (Antoun et al., 2004)] was incubated at 37°C for 30 min. Then the ribosomes were pelleted through a sucrose cushion (1.1 M sucrose in HEPES:Polymix buffer with 15 mM Mg²⁺) at 50,000 rpm for 2 h (TLS-55, Beckman), the pellet was dissolved in HEPES:Polymix buffer [5 mM Mg(OAc)₂], aliquoted, frozen in liquid nitrogen and stored at −80°C.

3 μM E. faecalis RelQ (Beljantseva et al., 2017) was incubated in reaction buffer (18 mM MgCl₂, 20 mM DTT, 20 mM Tris-HCl pH 8.0) together with 8 mM ATP and 5 mM ³H-GTP (SA: 100 cpm/pmol) for 2 h at 37°C to produce ³H-pppGpp. The resultant mixture was loaded on strong anion-exchange column (MonoQ 5/50 GL; GE Healthcare), and nucleotides were resolved by a 0.5–1,000 mM LiCl gradient. Peak fractions containing ³H-pppGpp were pooled and precipitated by addition of lithium chloride to a final concentration of 1 M followed by addition of four volumes of ethanol. The suspension was incubated at –80°C overnight and centrifuged (14,800 rpm, 30 min, 4°C). The resulting pellets were washed with absolute ethanol, dried, dissolved in 20 mM HEPES-KOH buffer (pH 7.5) and stored at –80°C.

The reaction mixtures contained 140–250 nM Rel, 300 μM ³H-pppGpp, 1 mM MnCl₂, an essential cofactor for Rel’s hydrolysis activity (Avarbock et al., 2000; Mechold et al., 2002;Tamman et al., 2019), all in HEPES:Polymix buffer (5 mM Mg²⁺ final concentration). After preincubation at 37°C for 3 min, the reaction was started by the addition of prewarmed Rel and 5 μL aliquots were taken throughout the time course of the reaction and quenched with 4 μL 70% formic acid supplemented with a cold nucleotide standard (4 mM GTP) for UV-shadowing.

Assays with E. coli RelA were performed as described earlier (Kudrin et al., 2018). In the case of B. subtilis Rel, the reaction mixtures typically contained 500 nM B. subtilis 70S IC(MVF), 140 nM Rel, guanosine nucleoside substrate (300 μM ³H-GTP, PerkinElmer), 100 μM pppGpp, 2 μM E. coli tRNA^Val (ChemBlock), all in HEPES:Polymix buffer (5 mM Mg²⁺ final concentration). After preincubation at 37°C for 3 min, the reaction was started by the addition of prewarmed ATP to the final concentration of 1 mM, and 5 μL aliquots were taken throughout the time course of the reaction and quenched with 4 μL 70% formic acid supplemented with a cold nucleotide standard (4 mM GTP) for UV-shadowing. Individual quenched timepoints were spotted PEI-TLC plates (Macherey-Nagel) and nucleotides were resolved in 1.5 M KH₂PO₄ pH 3.5 buffer. The TLC plates were dried, cut into sections as guided by UV-shadowing, and ³H radioactivity was quantified by scintillation counting in EcoLite Liquid Scintillation Cocktail (MP Biomedicals).

As we have shown earlier, low-level ectopic expression of RelA^ΔRRM from a low copy number pNDM220 plasmid under the control of a P_A1/O4/O3 promoter has a more pronounced inhibitory effect on E. coli growth in comparison with expression of the full-length protein (Turnbull et al., 2019). We tested whether P_A1/O4/O3-driven high-level expression of RelA^ΔRRM from a high-copy pUC derivative pMG25 would cause a more pronounced growth defect (Figure 2A). For comparison, we tested the effects of expression of the second E. coli RSH enzyme – SpoT – using both the full-length and the ΔRRM variants. Even in the absence of the IPTG inducer, leaky expression of RelA^ΔRRM has a dramatic effect on E. coli growth, while the full-length protein does not have an effect. In the presence of 50 μM IPTG, both full-length and RelA^ΔRRM inhibit the growth, although the latter has a stronger effect; induction with 500 μM IPTG completely abrogates the growth in both cases. While expression of SpoT has a detectable inhibitory effect at 500 μM IPTG, the effect is the same for full-length and SpoT^ΔRRM. Since even leaky expression of RelA^ΔRRM inhibits growth, we concluded that this high-level expression system is ill-suited for follow-up microbiological investigations. Therefore, to test whether the toxicity of RelA^ΔRRM in E. coli is dependent on the interaction with starved complexes, we used the pNDM220-based low-level expression system used previously (Turnbull et al., 2019). Guided by the recent cryo-EM reconstructions of RelA (Arenz et al., 2016; Brown et al., 2016; Loveland et al., 2016), we designed a set of mutations that will specifically disrupt RelA’s interaction with starved ribosomal complexes.

To disrupt the interaction between RelA and the tRNA, we adopted the H432E mutation in the TGS domain that was earlier shown to specifically abrogate the recognition of the 3′ CCA end of the A/R tRNA (Winther et al., 2018). This conserved histidine residue stacks between the two cytosine bases and hydrogen-bonds the phosphodiester backbone (Figure 1B). Replacing it with glutamic acid introduces a charge repulsion effect as well as a steric clash. To disrupt the interaction with the ribosome we used mutations in the ZFD: R629E as well as a double substitution C602A C603A; both mutants are expected to compromise the recognition of the 23S rRNA ASF element that is crucial for RelA recruitment (Kudrin et al., 2018). The conserved double motif docks the ZFD α-helix into the major groove of the ASF; replacement by alanine is expected to abrogate this interaction (Figure 1C). The conserved arginine 629 residue is in close proximity to A886-A887 and A885-A886 phosphodiester bonds (3.5 and 5 Å, respectively), and, therefore, the R629E substitution is expected to cause electrostatic repulsion.

Low-level expression of full-length RelA has a minor, but detectable growth inhibitory effect both when tested on solid LB agar media (Figure 2B) and in liquid LB cultures [growth rate, μ₂, decreases from 0.69 (vector) to 0.47 h^–1) (Figure 2C]. Importantly, the spotting control on solid LB media lacking IPTG shows that the size of the inoculum is not affected by potential leaky expression in the overnight culture (Figure 2B). Deletion of the RRM renders RelA significantly more toxic (growth rate decreases to 0.29 h^–1), in good agreement with the accumulation of (p)ppGpp upon expression of the construct (Turnbull et al., 2019). The effect is countered by TGS H432E, ZFD R629E and even more so by the C612A C613A substitutions (Figures 2B,C). RelA^ΔRRM expression is equally toxic in the ΔrelA background and the effect is similarly countered by H432E, R629E and C612A C613A substitutions (Supplementary Figure S1), demonstrating that the growth inhibition is independent of the functionality of the endogenous RelA stringent factor. Finally, we directly confirmed the lack of activation by deacylated tRNA in the case of H432E E. coli RelA using biochemical assays (Figure 2D).

Taken together, our results suggest that RelA^ΔRRM toxicity in E. coli is dependent on the functionality of the interaction with starved ribosomes. To test the generality of this hypothesis, we next characterized B. subtilis Rel lacking the RRM domain.

We expressed ΔRRM Rel under the control of an IPTG-inducible P_hy–spank promotor (Britton et al., 2002) in ppGpp⁰ (Δrel ΔrelP ΔrelQ) (Nanamiya et al., 2008) or Δrel B. subtilis strains (Figure 3A). In the ppGpp⁰ background, inhibition of B. subtilis growth on LB plates is a marker of toxic (p)ppGpp overproduction. In the Δrel background, (p)ppGpp is overproduced by a SAS in the absence of Rel’s hydrolytic activity, causing a growth defect in B. subtilis (Nanamiya et al., 2008) and S. aureus (Geiger et al., 2014). Therefore, this experiment tests the complementation of hydrolase function of Rel that manifests in improved growth.

Unlike the full-length Rel, the Rel^ΔRRM truncation is toxic in the ppGpp⁰ and Δrel backgrounds, both when the growth is followed on plates (Figure 3A) and in liquid culture (Figures 3B,C and Supplementary Figures S2A,B). To probe the role of the interactions with starved ribosomes in Rel^ΔRRM toxicity, we used a set of substitutions in B. subtilis Rel corresponding to those used to study E. coli RelA (Figures 1, 2). The toxicity of the Rel^ΔRRM mutant is efficiently countered by the H420E substitution in the TGS as well as the R619E and C602A C603A substitutions in the ZFD (Figure 3D and Supplementary Figures S2D–F,H). This strongly suggests that the intact interaction with tRNA and starved ribosomes is essential for the toxicity of Rel^ΔRRM. For comparison, we tested, full-length Rel, Rel^ΔRRMΔZDF C-terminal truncation, as well as the NTD domain region alone. The Rel^ΔRRMΔZDF mutant is only slightly toxic in the ppGpp⁰ background and its expression promotes growth in the Δrel background (Figure 3A). NTD-alone construct displays no toxicity in the ppGpp⁰ background and does not promote the growth in the Δrel background, suggesting weak – or absent – synthetase activity.

To separate the effects of (p)ppGpp production from the effects of (p)ppGpp degradation, we tested synthesis-deficient SYNTH D264G mutants (Nanamiya et al., 2008) of C-terminally truncated Rel variants (Figure 3A and Supplementary Figure S2C). The toxicity of the ΔRRM variant is abolished by the D264G mutation demonstrating that it is, indeed, mediated by (p)ppGpp production and not, for example, through inhibition of protein synthesis via competitive binding to ribosomal A-site [the latter non-enzymatic mechanism of toxicity was shown for E. coli RelA^CTD (Turnbull et al., 2019)]. Both ΔRRM D264G and ΔZFDΔRRM D264G variants promote growth in the Δrel strain suggesting that neither deletion of ΔRRM alone – or both ΔZFD and ΔRRM – abrogates the hydrolysis activity of B. subtilis Rel (Figure 3A, bottom panel). At the same time expression of the synthesis-inactive D264G Rel^NTD has no effect, suggesting that the NTD does not efficiently hydrolyze (p)ppGpp. To test if further truncations of the NTD-only Rel (Rel^1–373) would induce hydrolytic activity, we tested several additional constructs of B. subtilis Rel – Rel^1–336, Rel^1–196 and Rel^1–155 – but neither of them could rescue the growth effect of Δrel B. subtilis (Supplementary Figure S3). Finally, the synthesis deficiency of D264G Rel and the lack of H420E Rel activation by deacylated tRNA was confirmed using biochemical assays (Supplementary Figure S4).

Taken together, our results demonstrate that (i) Rel^ΔRRM is toxic in ppGpp⁰ B. subtilis, (ii) this toxicity requires intact (p)ppGpp synthesis activity of the enzyme (iii) it is abrogated by mutations disrupting the interaction with tRNA and starved ribosomes and (iv) deletion of the RRM domain does not abrogate the hydrolysis activity of B. subtilis Rel.

To test the low-level, non-toxic, synthesis activity of Rel mutants and to validate the effects of point mutations disrupting the interaction of Rel with starved ribosomal complexes, we took advantage of the amino acid auxotrophy phenotype of the ppGpp⁰ B. subtilis (Δrel ΔrelP ΔrelQ) (Nanamiya et al., 2008).

When ppGpp⁰ B. subtilis is grown on Spizizen minimum medium (Spizizen, 1958) in the absence of casamino acids, neither of the D264G Rel mutants – either full-length or C-terminal truncations – promote growth, both whether or not expression is induced by 1 mM IPTG (Figure 3E, right panels). Full induction of NTD expression with 1 mM IPTG near-completely suppressed the auxotrophy phenotype (Figure 3E, top right panel), while the leaky expression in the absence of IPTG results in weak, but detectable suppression (Figure 3E, bottom right panel). This demonstrates that B. subtilis NTD has a weak net-synthesis activity. The Rel^ΔRRM is, as expected, highly toxic when expression is induced by IPTG; conversely, low-level leakage expression efficiently suppresses the amino acid auxotrophy phenotype. Removal of both RRM and ZFD domains renders the protein non-toxic. It is not trivial to reconcile this effect with the idea that removal of the RRM renders the protein toxic due to lack of auto-inhibition: one would expect that the additional removal of the ZFD domain would further compromise the CTD-mediated negative control in Rel^ΔRRM.

We next used the auxotrophy assay to test the effects of the H420E TGS and C602A C603A ZFD substitutions on the activity of Rel expressed from the native genomic locus under the control of the native promotor (Figure 3F). As a positive control we used a strain lacking the genomic copy of rplK (relC) encoding ribosomal protein L11. This ribosomal element is essential for E. coli RelA activation by starved ribosomal complexes (Parker et al., 1976; Wendrich et al., 2002; Shyp et al., 2012) as well as for cellular functionality of C. crescentus Rel (Boutte and Crosson, 2011). The ppGpp⁰ strain expressing H420E Rel fails to grow on the minimum media, reinforcing the crucial role of that this residue, while the C602A C603A can sustain the growth, suggesting that this substitution does not completely abrogate the activity.

Next, we probed the ribosomal association of ΔRRM and full-length Rel expressed in the ppGpp⁰ background using a centrifugation sucrose gradient followed by Western blotting using antiserum against native, untagged B. subtilis Rel (Figure 4A). Since deacylated tRNA promotes ribosomal recruitment of E. coli RelA (Agirrezabala et al., 2013; Kudrin et al., 2017), we probed the association of wild type and mutant Rel variants with the ribosome both under exponential growth and upon acute isoleucine starvation induced by the isoleucyl tRNA synthetase inhibitor antibiotic mupirocin (pseudomonic acid) (Thomas et al., 2010). In good agreement with the cryo-EM structures detailing multiple contacts between the RRM and the starved complex and therefore suggesting an importance of this element in ribosomal recruitment (Arenz et al., 2016; Brown et al., 2016; Loveland et al., 2016), we do not detect a stable association of ΔRRM Rel with the ribosome upon a mupirocin challenge. This suggests that the interaction with the ribosome is significantly destabilized in Rel^ΔRRM and the protein dissociates during centrifugation. It is noteworthy that, despite an unstable association of Rel^ΔRRM with starved ribosomes, the expression of this protein strongly induces the accumulation of 100S ribosomal dimers, which is indicative of (p)ppGpp overproduction (Tagami et al., 2012). Note that the 100S formation is abrogated when the culture is treated with mupirocin (Figure 4A and Supplementary Figure S5B), most likely due to complete inhibition of translation by the antibiotic hindering expression of the 100S-promoting Hibernation Promoting Factor (HPF) which, in turn, is induced by accumulation of (p)ppGpp (Tagami et al., 2012).

Our microbiological experiments demonstrate that both the C602A and C603A double substitution and the R619E point substitution render Rel^ΔRRM non-toxic (Figure 3D), which we attribute to further destabilization of Rel’s interaction with starved ribosomal complexes. To directly probe the effects of these mutations, we used centrifugation experiments with full-length Rel carrying the substitutions. As expected, both the C602A C603A double mutant and R619E full-length variants are compromised in recruitment to the ribosome upon a mupirocin challenge (Figure 4A).

Taken together, these results demonstrate that Rel^ΔRRM is significantly more toxic than the full-length protein and this toxicity is dependent on the interaction with starved ribosomes, which is, in turn, destabilized in this truncation. As a next step, we set out to test the effects of RRM deletion – either alone or in combination with mutations further compromising the interactions with starved ribosomes – on Rel’s enzymatic activity in a reconstituted B. subtilis biochemical system.

To purify untagged B. subtilis Rel we combined our protocols used for purification of E. coli RelA (Turnbull et al., 2019) and T. thermophilus Rel (Van Nerom et al., 2019) (Figure 5). Importantly, during all of the chromatography steps we followed both absorbance at 260 and 280 nm complemented with SDS PAGE analysis of fractions. This is essential in order to identify and specifically pool the fractions containing Rel free from RNA contamination. After the initial capture using immobilized metal affinity chromatography (IMAC) in high ionic strength conditions (750 mM KCl) using HisTrap HP column charged with Zn²⁺ in order to avoid possible replacement of the Zn²⁺ in the ZFD domain by Ni²⁺ ions (Block et al., 2009), His₁₀-SUMO-Rel was applied on size-exclusion chromatography (SEC) on HiLoad 16/600 Superdex 200 pg column (Figure 5D). Both in the case of the full length (Figure 5B) and ΔRRM Rel (Figure 5C), the RNA-free fractions constitute the minority of the protein that elute considerably later than the bulk of the RNA-contaminated Rel. While the SEC step is essential for generating RNA-free Rel preparations, the majority of the protein prep is lost at this stage. After the SEC step, the buffer was exchanged to storage buffer containing arginine and glutamic acid that improve protein solubility and long-term stability (Golovanov et al., 2004) (Figure 5D), His₁₀-SUMO tag was cleaved off and removed by passing the protein via second IMAC (Figure 5E). The quality of the final preparations was assessed by SDS-PAGE (Figure 5F and Supplementary Figure S6) as well as spectrophotometrically: OD₂₆₀/OD₂₈₀ ratio below 0.8 corresponding to less than 5% RNA contamination (Layne, 1957).

We have tested the effects of omission of the SEC step on the purification and activity of B. subtilis Rel preparations. Without the SEC step, the OD₂₆₀/OD₂₈₀ ratio was dramatically higher (1.9), suggesting that, counterintuitively, the ‘no SEC’ Rel preparation predominantly contains not protein but RNA. We have resolved the sample on 15% SDS-PAGE (Figure 6A) and denaturating 1.2% agarose (2% formaldehyde) (Figure 6B) gels, as well as subjected the samples to negative staining electron microscopy (Figure 6C). While the SDS-PAGE gel revealed multiple protein bands with Mw between 40 and 10 kDa, the agarose gel revealed that the RNA contaminant is dominated by three distinct populations of approximately 3000, 1500, and 100 nucleotides in length. Large (approximately 20 nm in diameter) particles are clearly visible on the to negative staining EM images. Collectively, this suggests that the RNA contamination is dominated by ribosomal particles, although it is unclear whether these are intact or partially degraded. Taking into account that 1 A₂₆₀ corresponds to 23 pmol 70S particles, we estimate that our ‘no SEC’ preparations contain 45 nM 70S ribosomes per 1 μM Rel, which corresponds to sub-stoichiometric contamination of 5% of Rel being in complex, and 95% free. We next tested the effects of SEC omission on the enzymatic activity of Rel. The effects are exceedingly mild. The synthetase activity is virtually unaffected; importantly, activation by deacylated tRNA remains strictly mRNA-dependent, with tRNA^Val inducing the enzymatic activity of ‘no SEC’ Rel only in the presence of 70S initiation complexes (70S IC) but not vacant 70S ribosomes (Figure 6D). Importantly, since ribosomes or starved complexes are added in our synthetase assays in excess over Rel (500 nM vs. 140 nM Rel), in the final reaction mixture purified ribosomes are in approximately 100x excess over the contaminant. Finally, the hydrolase activity of ‘no SEC’ Rel is approximately twofold lower (Figure 6E), which, however, could reflect variability in preparations.

Taken together, these results suggest that in the absence of a dedicated SEC step, Rel preparations are sub-stoichiometrically contaminated with ribosomes. While the effects of this contamination on the enzymatic activities of Rel are minor, it might interfere with other assays (see section Discussion).

We tested the ³H-pppGpp synthesis by full-length Rel as well as Rel^ΔRRM, either alone or activated by the ribosomes or starved complexes in a reconstituted system (Figure 4B). When the protein is tested by itself, the Rel^ΔRRM mutant is less active than the full-length, suggesting that deletion of the RRM domain does not lead to the loss of auto-inhibition. While Rel^ΔRRM remains less active than the full-length when activated by initiation complexes (about twofold), in the presence of starved ribosomal complexes the two proteins are equally active. This could be explained by tRNA stabilizing Rel on the ribosome and overriding the defect caused by the removal of the RRM domain.

The R619E mutation compromises activation of the full-length Rel by the initiation complexes (more than four-fold), and the effect is less pronounced in the presence of deacylated tRNA^Val (less than twofold decrease in activity). Just as in the case of the RRM deletion, a possible explanation is that the deacylated tRNA strongly stimulates the binding of Rel to the ribosome and offsets the effect of the mutation R619E. When the R619E substitution is introduced into ΔRRM Rel, the combination of the two mutations destabilizing Rel binding to the ribosome results in compromised activation both by the initiation (five-fold) and starved (seven-fold) ribosomal complex. Despite several attempts we failed to generate sufficiently pure and soluble C602A C603A Rel^ΔRRM, which precluded direct biochemical characterization of this mutant.

Taken together, our biochemical results demonstrate that while RRM is important in Rel recruitment to the ribosome, this domain is not absolutely essential for the activation of its (p)ppGpp synthesis activity by starved ribosomal complexes, which is consistent with the ribosome-dependent nature of the Rel^ΔRRM toxicity in live cells.

It was recently proposed that the RRM domain has a stimulatory effect on the hydrolysis activity of C. crescentus Rel, and the loss of this regulatory mechanism explains the toxicity of the ΔRRM mutant (Ronneau et al., 2019). This hypothesis does not explain the toxicity of the ΔRRM variant of the synthesis-only RSH RelA (Figure 2 and Turnbull et al., 2019) and our microbiological experiments showing that the synthesis-defective ΔRRM D264G variant of B. subtilis Rel remains active as a (p)ppGpp hydrolase (Figure 3A). Importantly, the variant characterized by Ronneau et al. (2019) (C. crescentus Rel^Δ668–719) does not completely lack the RRM, and it is possible that the remaining beta-strand alpha-helix turn structural element was interfering with the hydrolysis activity of the construct.

Our enzymatic assays following ³H-pppGpp degradation by full-length and Rel^ΔRRM show that the latter is approximately twice less active (Figure 4C). While the defect is detectable, it is quite minor. To test if the hydrolysis defect is more pronounced in the living cell, we re-tested the hydrolysis activity of synthesis-deficient SYNTH D264G full-length Rel as well as C-terminally truncated Rel variants in the Δrel background using the growth rate (μ₂) as a proxy (Figure 4D). We modulated the expression levels by titrating the inducer, IPTG, from 10 to 1000 μM. In good agreement with the biochemical results demonstrating a minor defect in hydrolysis caused by deletion of the RRM domain, ΔRRM D264G Rel mutant promotes the growth of Δrel B. subtilis only moderately less efficiently than the full-length D264G.

It was also recently reported that binding of branched-chain amino acids (BCAAs) to the ACT/RRM domain induces the hydrolysis activity of Rhodobacter capsulatus Rel (Fang and Bauer, 2018). The N651A substitution abrogates amino acid binding to R. capsulatus Rel^CTD and leads to (p)ppGpp accumulation in the cell, presumably due to lower hydrolysis activity of the mutant enzyme. It is, therefore, conceivable that B. subtilis Rel^ΔRRM is less hydrolytically active in the cell than the full-length protein due to the loss of BCAA-mediated activation. While the B. subtilis Rel^CTD fragment was shown to preferentially bind leucine with a K_D of 225 μM, no enzymatic assays were performed with this protein (Fang and Bauer, 2018). Notably, while the CTD region of E. coli RelA binds valine with high affinity (K_D of 2.85 μM) (Fang and Bauer, 2018), this interaction could not be regulating the hydrolysis activity of this synthesis-only RSH. It is, therefore, unclear whether B. subtilis Rel is, indeed, regulated by branched-chain amino acids similarly to R. capsulatus enzyme. Therefore, we tested the effect of 1 mM leucine on ³H-pppGpp degradation by B. subtilis Rel. We detect no stimulatory effect (Figure 7A). Furthermore, when we introduced the N685A substitution (equivalent to N651A in R. capsulatus) in the chromosomal rel gene, we detected no growth defect in comparison to wild-type 168 B. subtilis, either on solid or liquid LB media (Figures 7B–D). Taken together, these results suggest that amino acid binding to RRM should not automatically be equated with regulation of the hydrolysis activity.