Starvation Induces Profound, Predominately (p)ppGpp-Dependent Metabolic Changes
To understand how cellular metabolism globally responds to environmental stressors, we extracted metabolites from exponentially growing and amino acid-starved B. subtilis
cells and quantified 131 metabolites with liquid chromatography-tandem mass spectrometry (LC-MS/MS) (Tu et al., 2007
) (). We identified 96 abundant species with high quality values (the remaining 35 were low-abundance species or could not be unambiguously quantified) (Table S1
). These 96 metabolites exhibited quantitative consistency among biological replicates yet displayed profound changes upon amino acid starvation, with half (48/96) altered significantly during starvation (1.5–15 fold change) (, Figure S1, and Table S2
Metabolic Profiling of Wild-type and (p)ppGpp0 Cells upon Amino Acid Starvation
We then examined the dependence of these metabolic changes on a single regulator, (p)ppGpp, which is rapidly induced to high concentration during amino acid starvation (Potrykus and Cashel, 2008
). We created (p)ppGpp-deficient cells (termed (p)ppGpp0
(Potrykus and Cashel, 2008
)) by deleting the three genes encoding (p)ppGpp synthetases: RelA (Wendrich and Marahiel, 1997
), YjbM, and YwaC (Nanamiya et al., 2008
; Srivatsan et al., 2008
). Compared to wild-type cells, we observed significantly attenuated or opposite metabolic responses in (p)ppGpp0
cells upon starvation (, Figure S1, and Table S2
). Among the 48 metabolites that changed significantly in wild-type cells, only 11 changed independently of (p)ppGpp, and most of these changes were mild (Table S2
As a complementary approach, we performed principal component analysis (PCA) to separate our samples by metabolic features (). We verified that profiles from the same strain and treatment are located near each other in the PCA plot, demonstrating that the experiments are reproducible. Untreated wild-type and (p)ppGpp0 samples are located in overlapping regions, indicating that they have similar metabolic profiles. In contrast, profiles of starved wild-type cells are located in a distinct cluster along the first principal component axis (PC 1), separate from those of starved (p)ppGpp0 cells and untreated cells. This accounts for the largest difference among metabolic profiles and thus supports that (p)ppGpp drives the metabolic response to starvation.
To identify metabolic pathways affected by (p)ppGpp during starvation, we performed pathway analysis using MetaboAnalyst (Xia and Wishart, 2011
). The highest-ranking pathway (p-value = 1.01 × 10−5
) differentially altered was purine biosynthesis (Table S3
). We observed differential changes between wild-type and (p)ppGpp0
cells in the pathways leading to production of ATP and GTP from IMP (). GMP, GDP, and GTP levels were reduced in starved wild-type cells but elevated in starved (p)ppGpp0
cells. In contrast, ADP and ATP were elevated in starved wild-type cells, but the increase was attenuated in (p)ppGpp0
(p)ppGpp Post-transcriptionally Blocks Two Steps in GTP Biosynthesis
Our metabolic profiling supports previous observations that starvation-induced reduction of GTP levels correlates with (p)ppGpp induction (Lopez et al., 1981
; Krasny and Gourse, 2004
). It also shows (p)ppGpp-dependent reduction of multiple intermediates in the GTP biosynthesis pathway. To identify key targets of (p)ppGpp, we quantitatively analyzed our results, focusing on adjacent intermediates in the GTP biosynthesis pathway () and reasoning that (p)ppGpp inhibition of a reaction would result in substrate accumulation and decreased product levels (). We observed such changes at the transitions from GMP to GDP (~10 fold decrease) and hypoxanthine to IMP (~40 fold decrease) upon starvation in wild-type cells but not in (p)ppGpp0
cells (). This suggests two enzymes—guanylate kinase (Gmk, which is proposed to convert GMP to GDP) and HprT (which converts hypoxanthine to IMP and guanine to GMP)—as in vivo
targets of (p)ppGpp.
(p)ppGpp Directly Inhibits Multiple GTP Biosynthesis Enzymes
To determine whether (p)ppGpp down-regulates Gmk and HprT via transcriptional versus post-transcriptional mechanisms, we examined transcript levels using microarrays. gmk and hprT mRNAs were not reduced following starvation (), indicating that (p)ppGpp-mediated regulation of Gmk and HprT occurs post-transcriptionally.
(p)ppGpp Directly and Potently Inhibits the Enzymatic Activities of Gmk and HprT
We next examined whether (p)ppGpp directly inhibits the enzymatic activities of Gmk and HprT. In B. subtilis,
Gmk is an essential enzyme whose putative function is to convert GMP to GDP (Kobayashi et al., 2003
). We purified B. subtilis
Gmk (Figure S2A
) and used a coupled enzymatic assay to verify that Gmk converts GMP to GDP ( and Figure S2C
). We found that pppGpp and ppGpp, but not GTP, potently inhibit Gmk activity, achieving 50% inhibition at ~20 µM (p)ppGpp ( and Table S4
). This inhibition is specific to Gmk and not attributable to inhibition of the coupling enzymes (Figure S2B
). We conclude that (p)ppGpp is a specific, direct, and potent inhibitor of Gmk enzymatic activity.
HprT, which converts both guanine to GMP and hypoxanthine to IMP, was previously suggested as a potential target of (p)ppGpp during amino acid starvation in B. subtilis
(Beaman et al., 1983
). To test whether (p)ppGpp inhibits HprT in vitro
, we purified B. subtilis
HprT (Figure S2A
) and performed kinetic assays (Figure S2D
). We found that pppGpp and ppGpp are potent inhibitors of HprT, achieving 50% inhibition at ~11 µM ( and Table S4
Inhibition of GuaB Plays a Minor Role in Starvation Response
GuaB (IMP dehydrogenase) converts IMP to XMP and is a proposed target of (p)ppGpp (Gallant et al., 1971
; Lopez et al., 1981
). To test whether (p)ppGpp inhibits B. subtilis
GuaB in vitro
, we purified GuaB (Figure S2A
) and performed kinetic assays (Figure S2E
). GMP moderately inhibits GuaB activity (), similar to results obtained in E. coli
(Gallant et al., 1971
). However, pppGpp and ppGpp only moderately inhibit GuaB activity, and 50% inhibition of GuaB activity requires relatively high levels of pppGpp and ppGpp (~0.3–0.5 mM, respectively) (Table S4
Thus, ~10–20 µM (p)ppGpp significantly inhibits both HprT and Gmk activity, while even at 2 mM (p)ppGpp, more than 30% of GuaB activity remains (). In relevant context, as the in vivo
concentration of (p)ppGpp increases up to 1–2 mM during amino acid starvation, it should be sufficient to strongly inhibit Gmk and HprT activity, thus lowering GTP pools in response to starvation; however, inhibition of GuaB activity by (p)ppGpp is likely a minor contributor. Correspondingly, our metabolomic data did not show a major block after IMP (), and overexpressing guaB
did not increase GTP levels during amino acid starvation (Figure S2F
(p)ppGpp is Indispensable for GTP Homeostasis, via Negative Feedback Control
While high levels of (p)ppGpp upon starvation strongly inhibit GTP synthesis, basal levels of (p)ppGpp in cells during normal growth (~10–20 µM) are comparable to the in vitro (p)ppGpp concentrations at which ~50% of activities of HprT and Gmk are inhibited. Therefore, we hypothesized that (p)ppGpp might regulate GTP levels even in the absence of starvation.
Interestingly, we found that pppGpp, produced from GTP, is in fact moderately induced by increased GTP levels in the absence of starvation. To increase GTP levels transiently, we added guanosine, which is converted to GTP via the salvage pathway (), and we measured levels of GTP and (p)ppGpp by thin layer chromatography (TLC). Following guanosine addition, pppGpp levels rise concomitantly with GTP levels (). Although the pppGpp level is much lower than that induced during amino acid starvation, it should be sufficient to inhibit GTP biosynthesis via HprT and Gmk inhibition, based on the in vitro potency of (p)ppGpp-dependent regulation ().
pppGpp Mediates Feedback Control of GTP Levels
Therefore, we proposed that pppGpp might be globally involved in GTP homeostasis via a negative feedback mechanism, as it is induced by increased GTP and subsequently inhibits GTP synthesis (). We thus examined GTP homeostasis in (p)ppGpp0 cells: Surprisingly, we observed complete dysregulation of cellular GTP levels. Correspondingly, two-dimensional TLC (to visualize nucleotides in 32P-labeled cell extracts sampled 20 minutes after guanosine addition) revealed no significant changes in GTP levels with respect to other nucleotides in wild-type cells (, left), indicative of tight feedback control of GTP levels. However, in the absence of (p)ppGpp, GTP levels strikingly rose to become the most dominant spot (, right).
We confirmed this result by quantifying label-free GTP levels with targeted LC-MS/MS. In wild-type cells, GTP levels initially rise following guanosine addition but quickly re-equilibrate within 5 minutes (), likely due to negative feedback regulation by increased levels of pppGpp. In (p)ppGpp0
cells, in contrast, GTP levels continuously increase up to ~20 fold (), while ATP levels are not dramatically altered (). Examination of the metabolites along the GTP salvage pathway in guanosine-treated cells verified that a major block occurs before GMP formation (Figure S3
), supporting our in vitro
results showing that (p)ppGpp inhibits HprT activity ().
In summary, we found that previously characterized homeostatic mechanisms are insufficient to protect GTP levels from perturbations in B. subtilis, and we demonstrate that pppGpp not only facilitates but is indispensable for maintaining GTP homeostasis via negative feedback control.
A Genetic Screen Reveals that (p)ppGpp Targets GTP Biosynthesis, not RNAP
We found that (p)ppGpp not only elicits changes in metabolites but also is required for B. subtilis cells to survive nutrient stress. Within 20 minutes of amino acid starvation, only ~3% of (p)ppGpp0 cells survive, compared to ~100% of wild-type cells (). In addition, similar to E. coli results, loss of (p)ppGpp renders B. subtilis cells unable to form colonies on minimal medium without amino acid supplementation ().
(p)ppGpp0 Phenotype and a Genetic Selection for Suppressor Mutations
To understand how (p)ppGpp exerts its protective role during nutrient stress, we performed an unbiased genetic selection for mutations that allow (p)ppGpp0
cells to form colonies on minimal medium. We inoculated (p)ppGpp0
cells in separate liquid cultures and plated them on minimal medium plates. We obtained a single colony from each plate () and examined these mutants by first sequencing the genes encoding the β and β' subunits of RNAP, rpoB
, and rpoC
, as a similar screen performed in E. coli
found mutations in rpoB
(Xiao et al., 1991
; Murphy and Cashel, 2003
). Interestingly, among the 105 suppressors we isolated, none contained mutations in rpoB
, suggesting that the physiologically critical targets of (p)ppGpp in B. subtilis
differ from those in E. coli
Combining Illumina whole-genome sequencing with gene-targeted DNA sequencing (), we identified mutations in 37 suppressors ( and Table S5
). Most mutations are located in genes along the de novo
GTP biosynthesis pathway—guaA
()—and several have mutations in the −10 and −35 canonical promoter sequences, presumably resulting in reduced transcription (Table S5
). This suggests that partial loss-of-function mutations in GTP biosynthesis genes rescue colony formation. Correspondingly, we placed the endogenous guaB
locus under the control of an IPTG-inducible promoter and observed that depletion of guaB,
upon removal of IPTG, also completely rescued colony formation (Figure S4A
In addition, a number of suppressor mutations map to codY
, which encodes a GTP-regulated transcription factor with numerous targets (Molle et al., 2003
). The majority of codY
mutants contained frame-shift mutations (Table S5
), and deletion of codY
partially rescued colony formation (Figure S4B
and ). Although CodY is not directly involved in GTP biosynthesis, it activates transcription of guaB
() (Molle et al., 2003
), and loss-of-function mutations in codY
could decrease GTP levels.
Decreased GTP Levels Allow Survival of Amino Acid Starvation
(p)ppGpp Regulation of GTP Biosynthesis is Essential for Survival During Starvation
Next, we tested whether these suppressors could survive sudden amino acid starvation. Interestingly, suppressor mutations that best rescue colony formation on minimal medium can prevent cell death completely upon starvation in liquid culture (). On the other hand, deletion of codY, which does not completely rescue colony formation on minimal medium, did not prevent cell death.
We confirmed that GTP levels in both untreated and starved suppressor mutants are decreased in comparison to (p)ppGpp0
cells (), suggesting that the mutants resist amino acid starvation due to lowered GTP levels. We also noticed that stronger suppressors had lower GTP levels and survived starvation in liquid culture, while weaker suppressors had higher GTP levels and did not survive. GTP levels upon amino acid starvation negatively correlate with the ability to survive starvation () and to form colonies on minimal medium (Figure S5A
). Although changes in ATP and GTP levels are inversely coupled during starvation in wild-type cells (), GTP but not ATP levels varied greatly from one suppressor allele to another (). There was also no significant correlation between ATP levels and resistance to starvation ( and Figure S5B
). Our results indicate that GTP levels or GTP/ATP ratios, but not ATP levels, correlate with the ability to withstand amino acid limitation.
To test the causal relationship between GTP levels and resistance to amino acid limitation, we treated cells with the GMP synthetase (GuaA) inhibitor decoyinine (Lopez et al., 1981
) to inhibit GTP biosynthesis and found that it increased the ability of (p)ppGpp0
cells to form colonies on minimal medium (). Conversely, increasing GTP levels by guanosine addition abolished the ability of the suppressors to form colonies on minimal medium (), demonstrating that lowering GTP levels enhances (and increasing GTP levels diminishes) resistance to amino acid limitation.
High Levels of GTP Kill (p)ppGpp0 Cells Independent of Starvation
Regulation of GTP Homeostasis Dictates Survival
Finally, we found that loss of (p)ppGpp-mediated GTP homeostasis drastically reduces cell viability even in the absence of starvation. Addition of guanosine to (p)ppGpp0 cells, thereby increasing GTP levels (), kills ~99% of cells within an hour (; confirmed using a Live/Dead test, ). (p)ppGpp0 cells also fail to form colonies on plates with guanosine even in the presence of all 20 amino acids ().
We found that the sensitivity of (p)ppGpp0 cells to guanosine is attributable to high levels of GTP (or potentially GDP), as suppressors affecting different steps of the GTP biosynthesis pathway show differential resistance to guanosine addition (). Upon guanosine addition, suppressors that inhibit only the de novo GTP biosynthesis pathway (guaB, codY) have high levels of GTP () and cannot form colonies even when all amino acids are present (). In contrast, a suppressor with a gmk mutation, which blocks both the de novo and salvage pathways prior to GDP formation, does not have high levels of GTP upon guanosine addition () and can form colonies (). Our results demonstrate that high GTP (or GDP) levels, but not their precursors, are toxic to cells and that this effect is independent of amino acid availability.