Wild-type (WT) Pseudomonas putida KT2440 and the ΔrelA and ppGpp⁰ derivative strains used in this study were kindly provided by Fernando Govantes (Centro Andaluz de Biología del Desarrollo, Universidad Pablo de Olavide) (Diaz-Salazar et al., 2017) (see Supplementary Table 1). All strains were stored at −80°C in LB medium containing 15% (vol/vol) glycerol. For all experiments, the strains were first streaked onto LB agar plates (with 50 μg/ml kanamycin for the ΔrelA and ppGpp⁰ mutants) and incubated overnight at 30°C. A 3 mL sample of LB medium (containing 50 μg/ml kanamycin when required) was then inoculated from a single isolated colony and incubated at 30°C and 200 rpm in an orbital shaker (Inova 4230, New Brunswick Scientific, New Brunswick, NJ, United States). After 8 h of incubation, the cells were first diluted (1/1000) in 25 ml of M9 medium containing 3 g⋅L^–1 glucose, 17.4 g⋅L^–1 Na₂HPO₄ ⋅ 12H₂O, 3.0 g⋅L^–1 KH₂PO₄, 2.0 g⋅L^–1 NH₄Cl, 0.5 g⋅L^–1 NaCl, 0.5 g⋅L^–1 MgSO₄, 3.3 mg⋅L^–1 CaCl₂, and 1 ml of a trace element solution containing 15 mg⋅L^–1 Na₂EDTA ⋅ 2H₂O, 4.5 mg⋅L^–1 ZnSO₄ ⋅ 7H₂O, 0.3 mg⋅L^–1 CoCl₂ ⋅ 6H₂O, 1 mg⋅L^–1 MnCl₂ ⋅ 4H₂O, 1 mg ⋅L^–1 H₃BO₃, 0.4 mg⋅L^–1 Na₂MoO ⋅ 2H₂O, 3 mg⋅L^–1 FeSO₄ ⋅ 7H₂O, and 0.3 mg⋅L^–1 CuSO₄ ⋅ 5H₂O. The glucose, MgSO₄ and trace element solutions were filtered through a 0.2 μm filter (Minisart 0.2 μm syringe filter; Sartorius, Göttingen, Germany) and the other solutions were autoclaved. Since a kanamycin resistance gene was inserted into the genome of the mutants, kanamycin was then not added to the M9 precultures and cultures to avoid any unforeseen effects of the antibiotic. After overnight incubation at 30°C and at 200 rpm, exponentially growing cells were harvested by centrifugation (sigma 3–18K centrifuge, Sigma-Aldrich, Seelze, Germany) at 5,000 g for 5 min at room temperature, washed twice in fresh medium without glucose, and used to inoculate 50 ml of M9 medium with 3 g⋅L^–1 glucose at an optical density at 600 nm (OD₆₀₀) of 0.15 and incubated at 30°C at 200 rpm. The cells were grown aerobically in baffled shake-flasks in three separate biological replicates and the cultures were monitored at OD₆₀₀ with a Genesys 6 spectrophotmeter (Thermo, Carlsbad, CA, United States). Stringent response was triggered by adding 0.2 mM serine hydroxamate (SHX) to the cell cultures in mid-exponential phase at biomass concentrations of between 0.75 and 1.0 g⋅L^–1.

Samples and intracellular metabolites were extracted as described previously (Patacq et al., 2020) with some modifications. Briefly, 50 μl of culture was withdrawn 0, 1, 2, 5, 10, 20, 30, 45, 60, 90, 120, 150, and 180 min after SHX induction and added to 1 ml of a precooled methanol-acetonitrile-H₂O (4:4:2) solution at −20°C to quickly quench metabolic activity (Millard et al., 2014). ¹³C-labeled metabolites cell extract (25 μl, see below) was added as an internal standard (Wu et al., 2005). The samples were then vigorously vortexed, incubated at −20°C for at least 2 h, evaporated overnight to dryness in a SpeedVac (SC110A SpeedVac Plus, ThermoSavant, Waltham, MA, United States) and stored at −80°C until analysis.

¹³C-labeled metabolites were extracted from P. putida KT2440 cells grown in a 500 ml bioreactor (my-Control, Applikon Biotechnology INC, Sunnyvale, CA, United States) containing 200 ml of M9 medium supplemented with 7 g⋅L^–1 U-¹³C labeled glucose (Euriso-Top, Saint Aubin, France) and containing 1.0 g⋅L^–1 NH₄Cl, 0.5 g⋅L^–1 KH₂PO₄, 0.25 g⋅L^–1 NaCl, 0.25 g⋅L^–1 MgSO₄, 3.3 mg⋅L^–1 CaCl₂, and 1 ml of the trace element solution described above. The pH was maintained at 7 by automatically adding 14% (g/g) ammonia or 10% (g/g) phosphoric acid and the temperature was set to 30°C. Adequate aeration of the culture was achieved by automatically controlling the stirrer speed and aeration to maintain > 30% oxygen saturation. The stringent response was triggered by adding 0.8 mM of SHX to the culture when the OD₆₀₀ reached 3.5. One hour after SHX exposure, 50 ml of culture was withdrawn from the bioreactor and directly quenched in 200 ml of a precooled (4:4:2) methanol-acetonitrile-H₂O solution at −20°C. The solution was then vigorously vortexed, incubated at −20°C for 2 h, and evaporated overnight to dryness in a SpeedVac (SC110A SpeedVac Plus, ThermoSavant, Waltham, MA, United States). The ¹³C-metabolites were then resuspended in 50 ml deionized water and stored at −80°C until use.

Extracellular metabolites of WT P. putida KT2440, and of the ΔrelA and ppGpp⁰ mutants were identified and quantified by nuclear magnetic resonance (NMR). After SHX induction, 1 ml of broth was collected every 30 min, centrifuged at 14,500 g for 3 min, and the supernatants were stored at −20°C until analysis. The supernatants were mixed with 20 μl of an internal standard consisting of D₂O and 2.35 g/l deuterated trimethylsilylpropanoic acid (TSP-d4). Proton NMR spectra were recorded on an Avance III 500-MHz spectrometer equipped with a 5-mm z-gradient BBI probe (Brucker, Rheinstatten, Germany). Quantitative ¹H NMR analysis was performed at 286 K, using a single pulse and a relaxation of 10 s. Thirty-two scans were accumulated (32 k data points with a spectral width of 10 ppm) after four dummy scans. The spectra were processed and metabolites were quantified using Topspin 3.1 (Bruker, Rheinstatten, Germany). Three separate biological replicates were analyzed in this way for each strain.

After resuspension of the cell extract samples in 50 μl of water, cell debris were removed by centrifugation at 14,500 g for 10 min at 4°C. The samples were then analyzed using an ion chromatograph (IC; Thermo Scientific Dionex ICS- 50001 system; Dionex, Sunnyvale, CA, United States) coupled to an LTQ Orbitrap mass spectrometer (Thermo Fisher Scientific, Waltham, MA, United States) equipped with a heated electrospray ionization (ESI) probe. The ion chromatography method was adapted from Patacq et al. (2018). The KOH gradient was modified as follows: 0.5 mM for 1 min; 0.5 to 4.1 mM from 9.5 to 14.6 min; 9.65 mM at 24 min; 60 mM at 36 min; 60 to 90 mM from 36.1 to 43 min; 90 to 100 mM from 43.1 to 48 min; and 100 to 0.5 mM from 48.1 to 55 min. The total analysis time was 55 min and the injected volume was 15 μL. Fourier transform mass spectra were recorded in full-scan negative ion mode at a resolution of 60,000 (at m/z 400), with the following source parameters: a capillary temperature of 350°C, a source heater temperature of 350°C, a sheath gas flow rate of 50 AU (arbitrary units), an auxiliary gas flow rate of 5 AU, an S-lens RF level of 60%, and an ion spray voltage of 2.7 kV. Metabolites were identified based on their m/z values with a tolerance of 5 ppm. The data were acquired using the Xcalibur software (Thermo Fisher Scientific, Waltham, MA, United States) and analyzed using TraceFinder 4.0 (Thermo Fisher Scientific, Waltham, MA, United States). Three samples from three independent biological replicates were analyzed.

Growth rates and extracellular metabolite fluxes including glucose, gluconate, ketogluconate, acetate, pyruvate and valine exchange (production and/or consumption) were all calculated from the NMR data. All calculations were performed using Physiofit 1.0.2 (Peiro et al., 2019), a program designed to estimate growth rates and exchanges fluxes from time-course cultures of microorganisms grown under metabolic (pseudo) steady-state. Extracellular metabolite fluxes were determined from the rates of disappearance (or appearance) of substrates and products, in culture supernatants, as measured by NMR.

Metabolite concentration changes (expressed as mean log₂ ratios with respect to the WT strain) were clustered based on the Pearson correlation between metabolites using MEV v4.0.9.

The macroscopic effects of SHX addition were characterized in Pseudomonas putida KT2440 cells growing exponentially under aerobic conditions in minimal medium with 3 g⋅L^–1 glucose (16.6 mM) as sole carbon source. Results are reported as the mean ± standard deviation of three separate biological replicates. For clarity, the data of one representative biological experiment out of three are shown in Figures 1–5 and those obtained from the other biological replicates are given in supplementary material.

Before SHX induction, the growth was exponential (μ = 0.57 ± 0.03 h^–1) and accompanied by gluconate and 2-ketogluconate accumulation as previously reported (Nikel et al., 2015; Kohlstedt and Wittmann, 2019). As expected, the addition of 0.2 mM SHX in mid-exponential phase strongly inhibited growth, but not immediately, as the biomass concentration kept increasing before reaching a plateau at t = 60 min (Figure 1A and Supplementary Figure 1A). During that first hour, glucose continued to be consumed at half of the estimated rate during the exponential growth phase (3.97 ± 0.93 mmol g^–1 h^–1 versus 8.32 ± 1.01 mmol g^–1 h^–1, respectively) (Figure 2A and Supplementary Figure 2A), gluconate and 2-ketogluconate were either consumed or excreted (Figures 2B,C and Supplementary Figures 2B,C), and the concentrations of acetate, pyruvate and valine increased sharply (at 0.16 ± 0.04 mmol g^–1 h^–1, 0.89 ± 0.19 mmol g^–1 h^–1 and 0.06 ± 0.09 mmol g^–1 h^–1, respectively). The latter were not detected during the exponential phase (Figures 2D–F and Supplementary Figures 2D–F) or when no SHX was added (Supplementary Figure 3). These results reveal that the cells remained metabolically active after SHX addition, despite severe growth inhibition.

Growth then resumed until t ≈ 150 min, at which stage glucose and the other main carbon sources—including gluconate, acetate and pyruvate but not valine or 2-ketogluconate—were fully depleted in the medium. The final biomass concentration was 1.35 ± 0.14 g L^–1 (Figure 2 and Supplementary Figure 2). Note that the final biomass concentration for the WT strain grown in similar conditions using 3 g L^–1 glucose was 1.73 ± 0.02 g L^–1 when no SHX was added (Supplementary Figure 3). The difference in biomass yields (0.45 ± 0.05 g_CDW g_Glucose^–1 versus 0.58 ± 0.01 g_CDW g_Glucose^–1) can be explained by non-growth-associated glucose consumption consequent to SHX exposure. During the second growth phase, the growth rate was roughly half the value measured in the first exponential growth phase, 0.29 ± 0.05 h^–1), glucose, gluconate and to a lesser extent 2-ketogluconate are co-consumed, and pyruvate and acetate continued to accumulate before peaking at t = 90 min and t = 120 min, respectively. In contrast, the concentration of valine increased gradually right until the end of the experiments (Figure 2 and Supplementary Figure 2).

The intracellular concentrations of ppGpp and pppGpp, measured by LC-MS (Patacq et al., 2018; Figures 1B,C), varied strongly throughout the experiments. Starting from 0.28 ± 0.02 μmol⋅g_CDW^–1 just before SHX addition, ppGpp began to accumulate about 10 min after SHX addition, with a first peak (2.28 ± 0.42 μmol⋅g_CDW^–1) roughly 30 min after SHX addition. The ppGpp concentration then decreased continuously for 60 min (1.05 ± 0.38 μmol⋅g_CDW^–1), before increasing once again up to 3.29 ± 1.06 μmol⋅g_CDW^–1 at t = 150 min. Interestingly, the pppGpp concentration followed the same trend, albeit with a more moderate initial increase, peaking at 1.78 ± 0.34 μmol g_CDW^–1 after 30 min and then at 4.25 ± 1.1 μmol g_CDW^–1 after 120 min (Figure 1C). The two major phases of alarmone accumulation appear even more clearly when the ppGpp and pppGpp concentrations are pooled (Figure 1D).

In E. coli, SHX is well-known to promote (p)ppGpp accumulation and provoke growth arrest (Tosa and Pizer, 1971; Ross et al., 2016) and we showed in a previous study (Patacq et al., 2020) that growth is inhibited by SHX alone, while the stringent response induced by (p)ppGpp accumulation is crucial to overcome the growth disruption caused by SHX. To confirm that this is also the case in P. putida, the experiments were repeated with a ΔrelA mutant and a ppGpp⁰ strain missing the (p)ppGpp synthetase-encoding genes relA and SpoT and therefore unable to synthesize (p)ppGpp. The latter strain was used to validate results obtained with the ΔrelA mutant and exclude any effect related to SpoT.

The ΔrelA mutant grew slightly faster than the WT strain (0.61 ± 0.02 h^–1 versus 0.58 ± 0.02 h^–1) (Supplementary Figures 4A–C). As observed for the WT strain, SHX interrupted the growth of the ΔrelA mutant (Figure 3). (p)ppGpp was not detected by LC-MS either before or after SHX addition, confirming that SHX inhibited growth by itself, without (p)ppGpp. For the ΔrelA mutant furthermore, contrary to the WT strain, no second growth phase was observed (Figure 3 and Supplementary Figure 4C); the growth rate estimated from the slight increase in biomass concentration only reached a tenth of the value estimated in the initial exponential phase. This means that the ability of P. putida to overcome the growth disruption caused by SHX addition almost completely disappears without the stringent response. However, glucose continued to be consumed at a similar rate as observed during growth arrest for the WT strain (q_glc = 4.24 ± 0.73 mmol g^–1 h^–1) showing that the cells were metabolically active. They also consumed gluconate and produced 2-ketogluconate. While acetate and pyruvate did not accumulate after SHX addition as observed for the WT strain, valine excretion was much more pronounced in the ΔrelA mutant than in the WT strain (Supplementary Figure 5, left panel). Together, these results show that while the ΔrelA cells stopped growing after SHX addition, they remained metabolically active.

As reported previously (Sze et al., 2002), unlike the equivalent E. coli strain, the ppGpp⁰ strain of P. putida was able to grow in minimal medium without adding any amino acids. The ppGpp⁰ strain’s estimated growth rate was 0.45 ± 0.03 h^–1 (about 70% of the WT’s estimated growth rate) (Supplementary Figures 4A,D). The ppGpp⁰ strain accumulated gluconate but not 2-ketogluconate during the exponential phase either because of its reduced growth rate or because of its impaired stringent response. As in the ΔrelA mutant, adding SHX to ppGpp⁰ cultures led to growth arrest and no transient (p)ppGpp accumulation as expected (Figure 3 and Supplementary Figure 4D). Growth did not resume but the cells continued to consume glucose at 2.86 ± 0.85 mmol g^–1 h^–1, indicating that they remained metabolically active. After SHX addition, the gluconate concentration plateaued and the concentration of valine increased in the medium (Supplementary Figure 5, right panel). These results confirm that growth inhibition and the metabolic activity of P. putida are independent of (p)ppGpp synthesis, whereas the stringent response is crucial for growth resumption.

In summary, these comparative dynamics analyses show that: (i) the first peak in (p)ppGpp accumulation in the WT strain is mediated by RelA-dependent (p)ppGpp synthesis in response to SHX addition, (ii) the cells’ ability to resume growth is conditioned on the stringent response, and (iii) the second peak in (p)ppGpp accumulation, concurrent with carbon source exhaustion, is very likely due to the cells entering the stationary phase (Cavanagh et al., 2010).

Having established the macroscopic effects of SHX addition, we investigated the dynamic metabolic response by quantifying metabolites by LC-MS. Quantitative data over the entire course of the experiments were obtained for 67 intracellular metabolites, representing a substantial fraction of P. putida’s central carbon metabolism and nucleotides (the other metabolites were below the quantification limit or too unstable to yield reliable information). Metabolite concentrations were determined relative to their U-¹³C labeled forms by isotope dilution mass spectrometry (IDMS) (Mashego et al., 2004). A higher sampling frequency was used in the first hour of the experiments to catch the expected metabolic regulatory effects of (p)ppGpp. The data, expressed as log₂ ratios of the relative metabolite concentration changes, are given for the three biological replicates in Supplementary Material.

We used hierarchical clustering [Pearson correlation; (Eisen et al., 1998)] to visualize relative changes over time and compare metabolite profiles between strains (Figure 4). Similar clusters were obtained from the independent analyses of biological replicates (see Supplementary Figure 6). Several general conclusions can be drawn from these data. First, in the WT strain, the variations in metabolite concentrations follow the (p)ppGpp accumulation profiles remarkably closely. The metabolite concentrations varied little in the first 5 min of the experiments, when (p)ppGpp concentrations were low. During the first peak in (p)ppGpp accumulation, metabolites in the purine pathway, which formed a remarkably consistent cluster, decreased drastically in concentration while the concentration of certain nucleotides and of metabolites belonging to the upper parts of the Embden-Meyerhof-Parnas (EMP), Entner-Doudoroff pathway (ED) and pentose phosphate (PP) pathways (termed EDEMPP) increased. When (p)ppGpp levels decreased (after t = 90 min), purine intermediates returned to the levels observed before SHX addition, whereas almost all other metabolites accumulated in the cells. During the second peak in (p)ppGpp accumulation, the concentrations of purine intermediates dropped further but so did the concentrations of the EDEMPP pathway metabolites, contrary to what was observed during the first (p)ppGpp peak. Since the upper parts of the EMP and PP pathways are fed by phosphorylated glucose (G6P), this is probably a consequence of glucose depletion in the culture medium. Second, the metabolite profiles of the ΔrelA and ppGpp⁰ strains differ substantially from those of the WT strain. Almost all metabolites increased in concentration after SHX addition, notably the purine intermediates, whose levels decreased at this stage in the WT strain. Third, the variations in metabolite concentrations are less pronounced in the WT than in the ΔrelA and ppGpp⁰ strains, with, respectively, 28.4, 67.2, and 80.6% of metabolites increasing or decreasing by a factor of 4 or more. This indicates that the stringent response moderates fluctuations in intracellular metabolites levels. In the same vein, the proportion of metabolites whose levels varied by less than a factor of 2 before glucose exhaustion, was much higher in the WT strain (34.3%) than in the ΔrelA mutant (13.4%) or the ppGpp⁰ strain (4.5%). This set of homeostatic compounds included nucleotides such as XMP, AMP, ADP, ATP, GTP and GDP. A significant part of the metabolome thus remains homeostatic despite the sudden and severe growth disruption.

In summary, the intracellular levels of several metabolites involved in purine synthesis decrease in the first minutes after SHX addition in WT P. putida, but not in ΔrelA or ppGpp⁰ strains. The fact that these changes occur extremely soon after SHX addition suggests that (p)ppGpp exerts a regulatory effect on purine nucleotide biosynthesis. These results are in line with previous studies in E. coli (Hochstadt-Ozer and Cashel, 1972; Wang et al., 2019, 2020) and B. subtilis (Kriel et al., 2012) and highlight the conserved nature of this form of metabolic regulation in bacteria.

Variations in intracellular metabolite concentrations can be analyzed to identify enzymes whose activity is regulated by (p)ppGpp in vivo. The approach consists in examining the response of upstream and downstream metabolic intermediates to a modulation of the enzyme’s activity, provided its abundance does not change dramatically. While SHX addition can affect transcript levels (Durfee et al., 2008), ppGpp has been shown to have a negative effect on ribosome assembly, and thus limits mRNA translation (Li et al., 2018). Provided (p)ppGpp concentrations remain high therefore, enzyme levels should not vary dramatically on short timescales. Accordingly, we interpreted fluctuations in metabolite levels during the first phase of (p)ppGpp accumulation (0 to 60 min) as reflecting the metabolic regulation of the target enzymes by (p)ppGpp, while the later changes were assumed to reflect a more complex interaction of both metabolic and hierarchical regulation.

We investigated the regulatory effect of (p)ppGpp on purine biosynthesis in more detail by examining the concentration profiles of the precursors and intermediates of this pathway alongside those of (p)ppGpp during the first hour (Figure 5).

Some of the metabolites in the purine pathway (pRpp, GAR, FGAR, AICAR and IMP) varied sharply in concentration in the first minutes after SHX addition. Strikingly, the corresponding increase in (p)ppGpp concentration was weaker than observed later on (a ∼2-fold increase versus a ∼20-fold increase relative to the basal level). At the end of this period (t ≈ 10 min), the intracellular levels of most metabolites tended toward the basal value measured before SHX addition. Then, as (p)ppGpp concentration increased in the WT strain, the levels of GAR, FGAR, AICAR and AS, and to a lesser extent that of IMP, decreased sharply, down to 11, 21, 27, 7, and 50% of the levels measured before (p)ppGpp accumulation and the minima coincided with the (p)ppGpp concentration maximum (t = 30 min). In the ΔrelA and ppGpp⁰ strains in contrast, the concentrations of these metabolites are constant (AICAR, AS and IMP) or even increased dramatically (GAR and FGAR) (Supplementary Figure 7). In the meantime, the levels of pRpp, AMP, XMP and GMP varied little in the WT despite strong (p)ppGpp accumulation.

The dramatic decrease in GAR, FGAR and AICAR combined with the invariance of the pRpp concentration can be explained by a reduction in PurF (pRpp synthase) activity, the decrease in downstream metabolites likely reflecting the inhibition of PurF by (p)ppGpp. This result is in line with previous studies in other bacteria where (p)ppGpp has been shown to interact and inhibit PurF in vitro (Ahmad et al., 2019; Wang et al., 2019). Our data match remarkably well the changes reported in these studies in pRpp, GAR, FGAR and AICAR levels upon inhibition of purine biosynthesis. Similarly, AS profile is also consistent with the one recently associated with in vitro (p)ppGpp inhibition of PurA (AS synthetase) in Firmicutes (Yang et al., 2020) and much earlier in E. coli (Gallant and Cashel, 1967; Stayton and Fromm, 1979). This effect may explain the strong decrease in AS upon (p)ppGpp accumulation that was not observed in the ΔrelA mutant (Supplementary Figure 7).

This analysis identified two enzymes in the purine pathway (PurF and PurA) as potentially controlled in vivo by (p)ppGpp in P. putida. PurF catalyzes the first step of the de novo purine biosynthesis pathway and PurA catalyzes the committed step of AMP biosynthesis (and of de novo ATP synthesis) from IMP. (p)ppGpp may therefore exert dual control on purine biosynthesis, acting simultaneously on the committed step and on a metabolic branch of the pathway.