Composition of the outgrowth medium modulates wake-up kinetics and ampicillin sensitivity of stringent and relaxed Escherichia coli

The transition of Escherichia coli from the exponential into the stationary phase of growth induces the stringent response, which is mediated by the rapid accumulation of the alarmone nucleotide (p)ppGpp produced by the enzyme RelA. The significance of RelA’s functionality during the transition in the opposite direction, i.e. from the stationary phase into new exponential growth, is less well understood. Here we show that the relaxed strain, i.e. lacking the relA gene, displays a relative delay in regrowth during the new exponential growth phase in comparison with the isogenic wild type strain. The severity of the effect is a function of both the carbon source and amino acid composition of the outgrowth media. As a result, the loss of RelA functionality increases E. coli tolerance to the bactericidal antibiotic ampicillin during growth resumption in fresh media in a medium-specific way. Taken together, our data underscore the crucial role of medium composition and growth conditions for studies of the role of individual genes and regulatory networks in bacterial phenotypic tolerance to antibiotics.

Scientific RepoRts | 6:22308 | DOI: 10.1038/srep22308 Rapid RelA-dependent accumulation of ppGpp is dubbed the stringent response 17 and leads to cessation of stable RNA synthesis, inhibition of translation and growth arrest 18 . Loss of function relA mutants display a so-called relaxed phenotype characterized by a waste of cellular resources on continuous production of stable RNA during amino acid starvation 18 , diminished antibiotic tolerance 19 , and reduced production of glycogen 20 .
The classical growth curve of bacteria in batch culture contains a lag phase, an exponential phase and a stationary phase 21 . RelA mediates rapid accumulation of (p)ppGpp during the exit from exponential phase to entry into the stationary phase 22 , preparing the bacteria for starvation and cessation of growth. Interest in the physiology of relaxed (ΔrelA) strains has been reignited in the last decade, since the functionality of ribosome-dependent RSH enzymes i.e. RelA in Beta-and Gammaproteobacteria and Rel in the rest of bacterial clades 4 has been linked to bacterial virulence 23 and antibiotic tolerance 19 . Given the multiple roles played by (p)ppGpp during bacterial stationary phase physiology (for review see Navarro Llorens and colleagues 24 ) we set out to systematically characterize how RelA functionality affects re-growth of E. coli from an overnight stationary culture in fresh media, a step involved in virtually all microbiological experiments, specifically focusing on the role of amino acids and carbon source composition of the outgrowth media.

Results
The growth resumption delay of a ΔrelA strain is dependent on the outgrowth medium and can be abolished by the addition of the complete set of 20 amino acids. We used two standard types of microbiology media: chemically defined minimal medium M9 25 and complex Lysogeny Broth (LB) medium 26 . LB is based on a mixture of nutrients originating from a pancreatic digest of casein from cow's milk and autodigest of Saccharomyces cerevisiae, and as different nutrients are sequentially consumed, E. coli cultures undergo a succession of diauxic shifts along the growth curve 27 . M9 in its simplest formulation consists of a buffering system, a mixture of essential inorganic salts and a carbon source -usually glucose, as is used here -and it satisfies minimal nutrient requirements for growth of E. coli, while supplements such as amino acids and vitamins can be added separately.
To test RelA's role in growth resumption, K-12 E. coli wild type strain BW25113 28 and isogenic relaxed ΔrelA were grown through exponential phase into stationary phase, kept in stationary phase (defined as less than 10% increase in OD 600 within 1 hour) for 15 hours and diluted into fresh medium. The OD 600 of cultures was followed throughout the time course. During the initial growth to stationary phase there is no substantial difference in the growth of the two strains, both in LB (designated with light beige shading) and M9 (designated with light blue shading) (Supplementary Figure 1), just as there is no difference in growth resumption of the wild type and the relaxed strain upon LB-to-LB transition (Fig. 1A, quantification of lag and doubling times is summarized in Table 1). At the same time, the ΔrelA strain showed a pronounced -around five hours -growth resumption delay during transition from LB to M9 medium supplemented with 0.4% glucose without additional supplements such as amino acids (Fig. 1D). As a simple numerical measure of the differences in growth resumption, we have plotted the ratio of OD 600 for ΔrelA to wild type strain ( Fig. 1A-F, red trace).
The growth resumption delay of the ΔrelA strain could, in principle, stem from lower effective inoculum size as measurements of colony forming units (CFU) do show slightly lower cell count of the ΔrelA strain compared to the wild type during the stationary phase (Supplementary Figure 2A). However, cross-inoculation experiments LB-to-M9 and M9-to-LB show that the appearance of the growth resumption delay in the ΔrelA strain is specific to the nature of the outgrowth medium, specifically it is present in M9 but not LB (Fig. 1C,D), suggesting that reduction of the inoculum size is not the cause of the phenomenon. Washing the cells with M9 during the LB-to-M9 transition in order to remove traces of LB has a dramatic effect on the relative growth delay of ΔrelA strain: when this step is omitted the effect is considerably less pronounced (compare Fig. 1D,F). However, the wash per se is not responsible for the delay, since addition of the wash step during LB-to-LB transition, if anything, promotes an earlier regrowth of the ΔrelA strain (compare Fig. 1E,A).
Eventual regrowth of the ΔrelA strain in M9 medium could, in principle, be mediated by a sub-population harboring compensatory mutations -a well-documented phenomenon for E. coli strains unable to produce (p) ppGpp due to a simultaneous disruption in both relA and spoT genes 29 . However, passage of the wild type and ΔrelA strain through a second regrowth phase faithfully replicated the growth delay effect ( Fig. 2A), supporting the idea of composition of the outgrowth medium being responsible for the effect. Since the growth resumption lag was not apparent in LB medium, which has a high concentration of easily metabolizable amino acids 27 , we have tested whether amino acid supplementation of M9 rescues delayed outgrowth of the ΔrelA culture. Indeed, the growth resumption delay is rescued by addition of a full set of 20 amino acids (each at 100 μg/ml) to the outgrowth minimal medium (Fig. 2B), suggesting that amino acid limitation in M9 is, indeed, responsible for the effect. Measurements of CFUs are in good agreement with the OD 600 trace (Supplementary Figure 2B).
Deprivation of methionine, valine and leucine in the outgrowth medium causes a relative delay in growth resumption of ΔrelA strain. To test whether any specific amino acid is the limiting factor responsible for the delay in the resumption of the ΔrelA strain we tested growth recovery in M9 minimal media supplemented with single amino acid drop out sets, M9 supplemented with 0.4% glucose and 19 individual amino acids added at final concentration of 100 μg/ml. Deprivation of methionine, lysine or any of the branched-chain amino acids (BCAA) -isoleucine, leucine and valine -resulted in a growth resumption delay in both strains, although to a somewhat different degree in each case (Fig. 3A,B). The effect, however, was substantially stronger in the case of the ΔrelA strain (compare Fig. 3A,B,C).
In order to separate amino acid dropout effects on bacterial growth per se from specific effects on growth resumption we have performed the same set of experiments using inoculum of E. coli cells from exponential, rather then stationary, phase -an approach that was used in the past to study auxotrophy of relA mutants [30][31][32] . When switched from minimal M9 medium lacking amino acid supplements into a 19 amino acid medium neither the wild type nor the ΔrelA strain were able to resume growth in isoleucine dropout media for 24 hours of observation: a well-known phenotype of the K-12 strains [33][34][35] (Fig. 4A,B). This is in stark contrast with the stationary phase cultures, which did start regrowth after 3.1 ± 0.1 (wt), 3.4 ± 0.3 (ΔrelA) hours (Fig. 3A,B, Table 1). values of a wild type BW25113 strain (filled circles) and an isogenic ΔrelA strain (empty circles) were followed in LB (A, C and E, light beige shading) or M9 medium supplemented with 0.4% glucose (hereafter M9, light blue shading) (B,D,F). The ratio of OD 600 for ΔrelA to OD 600 of wild type strain (red dotted line) serves as a numerical measure of the difference in growth resumption kinetics between the two. Prior to inoculation, the seeder culture was kept for 15 hours in stationary phase in either LB (light beige shading) (A,D,E,F) or M9 (B,C) media. Cross-inoculation experiments M9-to-LB (C) and LB-to-M9 (D) demonstrate that the growth defect of ΔrelA is specific to the outgrowth medium, i.e. present only in M9. During the LB-to-M9 transition (D), cells were washed with M9 (indicated by the red triangle on the x axis) to reduce carry-over of medium. The washing procedure itself had only mild effect on cells, and if anything, favored growth resumption of ΔrelA cells (E). Results are shown as mean values of biological replicates (n ≥ 3) and error bars (too small to be seen for some of the points) indicate standard error of the mean.
Additionally, the relaxed strain showed a specific growth delay when leucine is omitted. Tyrosine omission does not result in lower stationary phase OD 600 when we use exponential phase culture inoculum, but does with the use of stationary phase inoculum (compare Fig. 3A,B and Fig. 4A,B).
Addition of individual amino acids does not rescue the growth resumption delay of the ΔrelA strain. Next, we set out to determine if the addition of any specific amino acid rescues the relative delay in growth resumption of the ΔrelA strain by testing the effects of addition of individual amino acids at final concentration of 100 μg/ml. None of the amino acids reversed the defect; conversely, several amino acids exacerbated it for both strains (Fig. 3D-F). Addition of valine and cysteine strongly inhibited the regrowth of both wild type and relaxed strain; serine completely inhibited the regrowth of ΔrelA, but not wild type (Fig. 3D,E). Growth inhibition    by valine and cysteine is present when we use exponentially growing inoculum, suggesting that the effect is not specific for growth resumption but rather bacterial growth per se (Fig. 4). While the addition of histidine, serine, threonine, isoleucine and leucine caused a prolonged lag phase after stationary phase in both of the two strains (Fig. 3D,E), the severity of the effect was somewhat different, with serine causing a more pronounced growth resumption delay in the relaxed strain (Fig. 3E). The inhibitory effect of serine was absent in the case of exponentially growing cells (Fig. 4C,D).
Switching the carbon source of the outgrowth medium from glucose to glycerol abolishes the growth resumption delay of the relaxed strain. Rich LB and poor M9 minimal media dramatically differ in amino acid content: while in LB medium amino acids and peptides serve both as building blocks for protein as well as a source of carbon, ammonium and energy 27 , M9 usually lacks amino acids altogether and the most commonly used carbon source is glucose, as was used in the experiments described above (Figs 1-4). As we have shown, the addition of 20 amino acids set to M9 supplemented with 0.4% glucose abolishes the delay in growth resumption of the ΔrelA strain (Fig. 2). Importantly, addition of amino acids also decreases the doubling time of both the wild type and the relaxed strain almost twice (from 70 ± 3 to 44 ± 1 and from 82 ± 7 to 43 ± 0.3 minutes, respectively, Table 1). One could argue the relative growth delay of the relaxed strain in the absence of amino acids is merely a consequence of the necessity of relA functionality during slow growth per se, rather then a specific effect of the lack of amino acids.
To probe this conjecture, we have performed the regrowth experiments while reducing the growth rate in M9 lacking amino acids by substituting the glucose, a preferred carbon source for E. coli, for less optimal carbon source, glycerol. This further reduction of the growth rate can be counteracted by the addition of 20 amino acid set, which allows us to probe the connection amongst amino acid and carbon source composition, growth rate and growth resumption delay in the ΔrelA strain. While the doubling time increases to 109 ± 5 (ΔrelA) and Scientific RepoRts | 6:22308 | DOI: 10.1038/srep22308 123 ± 3 minutes (wt) in M9 supplemented with glycerol instead of glucose, the relaxed cells initiate the regrowth almost early as the wild type (Fig. 5A,B, Table 1). Addition of the 20 amino acids set to M9 medium supplemented with glycerol increases the growth rates to the levels similar to that in M9 supplemented with glucose. However, the growth resumption kinetics of the relaxed and wild type strains remain unchanged, i.e. ΔrelA and the wild type regrow similarly (Supplementary Figure 3, Table 1). Taken together, these results demonstrate that the relative growth delay of the relaxed strain is modulated by both carbon source and amino acid composition of the outgrowth media.
Relaxed strain is killed by ampicillin considerably slower then the wild type during growth resumption in M9 supplemented with either glucose or glycerol. The bacterial growth rate is a key factor affecting antibiotic susceptibility. In the case of the antibiotic ampicillin the killing efficiency is believed to be directly proportional to the rate of growth 36 . Therefore, the effects of relA's loss of functionality on growth resumption kinetics are expected to alter the antibiotic killing kinetics. To test this conjecture, we followed antibiotic killing by ampicillin after stationary phase cultures were diluted into M9 supplemented with either glucose (Fig. 5C) or glycerol (Fig. 5D). Surprisingly, the ΔrelA strain was killed considerably slower then the isogenic wild type under both conditions. In the case of the wild type strain there is a correlation between the regrowth and ampicillin killing kinetics, i.e. the earlier bacteria start regrowth, the more efficiently they are killed by ampicillin. At the same time the relaxed strain is killed by ampicillin considerably less efficient then the wild type even in if the growth kinetics are very similar in M9 supplemented with glycerol (compare Fig. 5B,D). As a result, the effect of relA disruption on ampicillin tolerance is heavily dependent on medium composition: while in the presence of glucose after 5 hours of incubation with ampicillin -time point that is often used for end-point persister , and the ratio of OD 600 for ΔrelA to OD 600 of wild type strain was plotted as a numerical measure of the differences in growth resumption between the two strains. To follow the ampicillin tolerance during E. coli regrowth in the presence of 0.4% glucose (C) or glycerol (D), the bacterial cultures were treated as described above but the regrowth medium was supplemented with ampicillin at 200 μg/ml and cell viability (colony forming units, CFU) was measured instead of OD 600 . Results are shown as mean values of biological replicates (n ≥ 3) and error bars indicate standard error of the mean.

Discussion
E. coli growth, nutrient availability and RelA functionality. We have systematically analyzed the effects of amino acid and carbon source availability, and RelA functionality in K-12 BW25113 E. coli strains during their transition from stationary phase to new exponential growth. The RelA-specific effects during this transition are confounded by two aspects that one has to consider. First are the defects in amino acid metabolism that are specific to K-12 E. coli strains, the workhorse of microbiology for almost a century 39 . Due to a frameshift mutation in one of the central isoenzymes of acetohydroxy acid synthase (AHAS) 34 , addition of valine to minimal medium leads to cessation of growth that can be rescued by the addition of isoleucine, although the exact mechanism behind it is still matter of debate 8,9,33 . We clearly see the valine effect in our experiments (Figs 3 and 4). Second is the role of RelA and (p)ppGpp in amino acid biosynthesis. (p)ppGpp is crucial for amino acid synthesis as evidenced by both ppGpp0 (i.e. completely lacking the alarmone) E. coli 12 and B. subtilis 40 being auxotrophic for several amino acids including methionine and branched chain amino acids leucine, isoleucine and valine. The knock out strain used in the current work, while lacking RelA does have an intact copy of the second enzyme synthesizing (p)ppGpp in E. coli -SpoT 12 . While not directly causing auxotrophy, disruption of relA does lead to perturbed regulation of amino acid biosynthesis. Simultaneous addition of "one-carbon" amino acids (serine, glycine and methionine, SMG) suppresses bacterial growth, but while the wild type can overcome it, the relaxed can not 30 ; and the effect is counteracted by addition of isoleucine 31,41 . The difference in the behaviors of wild type and relaxed strains is likely due the stringent response promoting biosynthesis of branched chain amino acids (BCAA), such as isoleucine 8,9 . We clearly see that omission of one of the BCAA results in RelA-specific retardation of growth resumption (Figs 3 and 4). Cysteine is known to cause transient amino acid starvation in the uropathogenic E. coli strain SP536 42 ; the mechanism behind this phenomenon is not understood. We see manifestations of cysteine-induced starvation in our background: while inhibition of wild-type growth is transient, growth inhibition is near-complete in the course of 24 hours of observation of the relaxed strain (Fig. 4).
While the effects of amino acid composition on regrowth of the ΔrelA strain were expected, the effects of substitution of the carbon source in M9 media from glucose to glycerol were surprising (Fig. 5). In the presence of glucose ΔrelA strain regrows with a delay in comparison to the wild type, and in the presence of glycerol the two strains regrow equally well. The cause of this is not obvious, connections between (p)ppGpp and carbon metabolism are known; for example expression of the receptor protein of the global catabolic modulator cAMP (CRP) is under direct negative control of (p)ppGpp 43 . There are parallels between the effects on re-growth observed in this study and previous observations of the differential requirements for RelA in glycogen accumulation during amino acid starvation in the presence of different carbon sources 44,45 . The relA gene is needed when glucose is the carbon source, while the high cellular levels of cyclic AMP relieve the requirement for relA when glycerol is the carbon source 20,45 . Moreover, branched-chain amino acid biosynthesis is promoted by cAMP 46 . Since (p)ppGpp and amino acid metabolism are interconnected with carbon metabolism via many other pathways, such as tricarboxylic acid cycle 47 the connections among carbon source, RelA functionality and re-growth are far from simple.
Bacterial regrowth kinetics is intimately connected with bacterial sensitivity to bactericidal antibiotics: the frequency of persisters is reflecting the awakening kinetics 48 . Increased cellular (p)ppGpp level was suggested to be the ultimate driver of persister formation 49 , and is implicated in antibiotic-specific tolerance mechanisms, i.e. protection from ampicillin acting via inhibition of cell wall biosynthesis 50 . Therefore, one could naively assume that the loss of RelA would result in, if anything, lower persister count, which is evidently not the case. Clearly, persistence is a multifaceted phenomenon, with media composition and growth conditions playing a major role via effects on metabolism 51 and growth rate 52 .
Media and growth conditions. Cells were grown with vigorous agitation (200-220 rpm) at 37 °C in LB (Becton, Dickinson and Company) and M9 minimal medium (48 mM Na 2 HPO 4 , 22 mM KH 2 PO 4 , 9 mM NaCl, 19 mM NH 4 Cl, 0.1 mM CaCl 2 and 2 mM MgSO 4 ) 25 or on LB agar plates (Becton, Dickinson and Company). M9 was supplemented with 0.4% (w/v) carbon source, which was glucose or glycerol. Amino acids were used at a concentration of 100 μg/ml, kanamycin at 25 μg/ml and ampicillin at 200 μg/ml. The data presented on Figs 3 and 4 were obtained using a 96-well plate reader Tecan Sunrise and the reset of the experiments were performed in flasks.
Growth recovery experiments. Bacterial cultures were started from single colonies on LB plate and grown until OD 600 of 0.8. Resulting seeder culture was used to inoculate the experimental culture to starting OD 600 of 0.1, which was grown aerobically into stationary phase (20 ml of medium in 125 ml flasks), kept in stationary phase for 15 h and directly diluted into fresh medium to OD 600 of 0.1 or, during shift from LB to M9, harvested by centrifugation and, washed with M9 before transfer into fresh medium. Experiments with inclusion of 1 or 19 amino acids were conducted as follows: after 15 h in stationary phase, cells were harvested by centrifugation (in carbon source experiments washed with carbon source depleted M9), resuspended to OD 600 of 0.1 and grown aerobically in fresh medium on 96-well plates in a volume of 80 μl per well. OD 600 readings of the 96-well plates (plate reader Tecan Sunrise) were converted to values for 1 cm path length (spectrophotometer Thermo Helios β ) (Supplementary Figure 5). The length of the lag phase was determined by an intercept between the initial inoculum density (OD 600 = 0.1) and the tangent of fastest exponential part of the growth curve that determines the doubling time. Lag and doubling times were calculated separately for individual growth curves (n ≥ 3). Data analysis was performed in R 53 and the code is provided in the Supplementary Information. Antibiotic killing. 15 h stationary phase cultures were prepared as described above for growth recovery experiments. The cells were then collected 10 min at 5000 g at room temperature, washed with M9 0.4% glucose or M9 0.4% glycerol, collected and resuspended again and diluted to OD 600 of 0.1 in 20 ml medium in 125 ml flasks. The following ampicillin killing assays were performed essentially as described in 54 . A 10 μl aliquot was used for a CFU count at the zero hour time point, and then ampicillin was added to the remaining culture at 200 μg/ml. During following time course of ampicillin killing, flasks were incubated at 37 °C 200 rpm. Colony forming units were determined by series of tenfold dilutions out of which 5 μl was spotted on an LB plate. After overnight incubation of the plates at 37 °C, colonies were counted and CFU/ml was calculated.