GarA was required for establishment of infection of M. tuberculosis in macrophages and in mice

We have previously constructed a conditional gene disruption mutant of M. tuberculosis and demonstrated that garA was essential in standard Middlebrook medium, which contains 3 mM glutamate, but dispensable when additional amino acid supplements were used (10 mM asparagine, glutamate or glutamine [9]). Using this knowledge, we have now constructed an unmarked garA deletion mutant ΔgarA_Mt (Fig 1). This strain grew poorly on standard Middlebrook medium and growth was restored by addition of asparagine or reintroduction of garA (Figs 1A and S1). Asparagine was chosen rather than glutamate or glutamine because of good solubility and low influence on pH or buffering capacity.

Download:

• PPT
  PowerPoint slide
• PNG
  larger image
• TIFF
  original image

Fig 1. garA is required for growth of M. tuberculosis in vitro, survival in macrophages, and virulence in mice.

(A) M. tuberculosis lacking garA was unable to grow on standard 7H10 medium unless supplemented with asparagine. Plasmid-borne garA restored the defect, but variants of garA with mutations at threonine 21 in the phosphorylation motif (ETTS) gave only partial complementation. Strains were grown in Middlebrook 7H9 plus 30 mM asparagine then washed and diluted in standard 7H9 and spotted onto standard 7H10 with or without 10 mM asparagine. Photographs are representative of at least 3 independent experiments. (B) M. tuberculosis lacking garA (red squares) had a defect in growth and survival in differentiated THP-1 cells compared to parental M. tuberculosis H37Rv (black circles). Re-introduction of GarA (black triangles) or variants of GarA lacking a single phosphorylation site (grey crosses and squares) restored growth but variant GarA lacking both phosphorylation sites (green triangles) did not. Data points show the mean and standard deviation from four replicates and are representative of two independent experiments. (C) M. tuberculosis lacking garA was avirulent in mice as it was eliminated from the lungs. BALB/C mice were infected intranasally with 10⁵ bacilli and bacterial burden was measured on days 1, 7, 21 and 28. Data points show the bacterial burden in individual animals. The bacterial burden of mice infected with ΔgarA_Mt (red squares) was significantly lower than those infected with M. tuberculosis H37Rv (black circles), or complemented ΔgarA_Mt (black triangles) at all time points from day 7 (p<0.005, t test). (D) M. tuberculosis lacking garA failed to disseminate to the spleen by day 28 (symbols match panel C). The minimum number of bacteria that could be detected was 45 CFU/organ, marked by a dashed black line in C and D.

https://doi.org/10.1371/journal.ppat.1006399.g001

ΔgarA_Mt was tested for its ability to infect and replicate in THP-1 macrophages (ATCC TIB-202) and was unable to replicate (Fig 1B). Plasmid-encoded GarA complemented the defect (Fig 1B) but addition of 20 mM asparagine to the cell culture medium did not restore replication (S2 Fig), suggesting that intracellular M. tuberculosis could be in an environment with non-permissive asparagine concentration (below 10 mM). These results reinforce our earlier findings with the conditional mutant strain [9], which suggest that garA is essential for M. tuberculosis in macrophages and that this essentiality could be due to amino acid deprivation inside the phagosome.

Since ΔgarA_Mt shows auxotrophy in axenic growth and attenuation in macrophages, we predicted it would be avirulent. To compare the virulence of ΔgarA_Mt with parental H37Rv, equivalent inocula, as measured by colony forming units (CFU), were used for intranasal infection of immune competent BALB/c mice. Virulent H37Rv and the complemented strain replicated in the lungs and disseminated to the spleen within 28 days, whereas ΔgarA_Mt failed to replicate within the lungs and no bacteria could be recovered from lungs or spleen after 21 days (Fig 1C & 1D). Notably, a clear confirmation of the attenuation was observed between the macrophage infection model and the mouse infection model, which is also compatible with the observed auxotrophy in axenic media. We conclude that garA was essential for virulence of M. tuberculosis in mice, which may be due to amino acid deprivation in vivo or because of the impact of primary metabolism on other aspects of virulence, such as stress tolerance [25]. According to our model, GarA and PknG act in the same pathway but with opposite effects on metabolism. Interestingly, loss of garA caused severe attenuation in our study (complete loss of replication and bacterial survival) whereas loss of pknG caused only partial attenuation in a previous study using two different mouse models [5].

The phosphorylation sites on GarA were needed for growth of M. tuberculosis

We next turned our attention to the role of phosphorylation of GarA and the putative responsible kinase PknG. The normal role of phosphorylation is to “switch off” the activity of GarA by preventing it from binding to its enzyme targets [8]. The phosphorylation sites are found in a conserved ETTS motif in an unstructured N-terminal extension distinct from the forkhead-associated domain [31]. Variants of GarA lacking phosphorylation sites are functional for enzyme binding [31] but cannot be “switched off” by kinase activity, and thus might have an opposite effect on bacterial metabolism from garA knockout. We used variants of garA with mutations in the phosphorylation motif to complement the knockout strain ΔgarA_Mt in order to examine the role of regulation of GarA by phosphorylation in M. tuberculosis. GarA variants lacking the first phosphorylation site, threonine 21 (EATS or EAAS at the motif) was less effective than normal GarA at restoring growth on Middlebrook medium (Fig 1A), despite confirmed expression (S3 Fig). In THP-1 macrophages GarA that lacks a single phosphorylation site (EATS or ETAS at the motif) restored growth of ΔgarA_Mt in THP-1 macrophages but GarA lacking both phosphorylation sites (EAAS) failed to restore growth (Fig 1B). This suggests that the normal function of GarA requires regulation by phosphorylation. The single mutants EATS and ETAS led to slightly different levels of complementation in macrophages compared to axenic growth, which may reflect the different nutrient sources utilized in the two conditions.

As PknG is the only kinase reported to phosphorylate GarA at T21 (the first threonine in the GarA ETTS motif), disruption of pknG might be expected to alter phosphorylation of GarA, mimicking the effects of mutations to the GarA phosphorylation motif. ΔpknG_Mt has previously been reported to have a defect in survival in bone marrow derived mouse macrophages [11]. Unlike ΔgarA_Mt, ΔpknG_Mt was able to replicate in THP-1 macrophages, albeit 10-fold less than parental M. tuberculosis (S4 Fig). Taken together, these results support the importance of phosphorylation in regulating the function of GarA in M. tuberculosis both in vitro and during infection. Phosphorylation at both sites (T21 and T22) may be important, meaning that PknG and at least one other kinase may be involved, as is thought to be the case for the homologous system in C. glutamicum [32].

Disruption of garA and pknG separately caused specific and opposing effects on nutrient requirements

Having established the importance of GarA for virulence, we probed the reasons for this essentiality and the extent of conservation between slow-growing M. tuberculosis and fast growing non-pathogenic model M. smegmatis. GarA binds to the same enzyme targets to promote the same effects on enzyme activity in both organisms [8, 31]. We predicted that pknG disruption (or disruption of GarA phosphorylation sites) would lead to an inability to catabolise glutamate (since excess unphosphorylated GarA would inhibit GDH and KDH), while garA disruption would lead to uncontrolled glutamate catabolism (since GltS would be less active and GDH and KDH would be uninhibited).

We have previously studied the phenotype caused by garA deletion in M. smegmatis (ΔgarA_Ms). This strain grew normally on standard mixed medium (Fig 2A) but relied on external glutamate or related amino acids for growth and also showed differences from wild type in the ability to utilize a range of carbon sources [9]. Here we examined the ability of truncated GarA, which lacks the phosphorylation motif, to complement the growth defect. Truncated GarA and EAAS GarA fully complemented the growth defect of ΔgarA_Ms on media lacking glutamate, indicating that these variant proteins are functional in stimulating glutamate production and preventing glutamate catabolism (Fig 2B). However, when glutamate was the only source of carbon (Fig 2C) or nitrogen (Fig 2D), strains lacking pknG or strains expressing non-phosphorylatable GarA formed clumps (Fig 2E) and grew poorly compared to the parent strain. The nutrient-specific growth phenotypes recorded in microplates (Fig 2) were also apparent when the same strains were cultured in flasks (S5 Fig).

Download:

• PPT
  PowerPoint slide
• PNG
  larger image
• TIFF
  original image

Fig 2. Disruption of pknG or removal of the phosphorylation motif of garA caused a nutrient-dependent growth defect in M. smegmatis.

(A) All strains grew at the same rate on standard Middlebrook 7H9 medium. (B) ΔgarA_Ms grew slower than wild type on minimal Sauton’s medium containing 20 mM propionate, 20 mM NH₄Cl plus 0.05% tyloxapol, and this growth defect could be fully complemented by GarA lacking phosphorylation sites (truncated “trunc.” garA). (C + D) ΔpknG_Ms grew slower than wild type and formed clumps (inset photo) on medium containing glutamate as sole carbon (C) or nitrogen source (D) (minimal Sauton’s with either 30 mM glutamate plus tyloxapol, or 1% glycerol, 10 mM glutamate plus tyloxapol). Data plotted are the mean and standard deviation of at least 3 independent experiments. (E) ΔpknG_Ms formed clumps when glutamate was the sole carbon or nitrogen source. The photograph shows a microplate from growth curve (D) imaged at 60 hours. Growth of ΔgarA_Ms complemented with phosphoablative GarA (EAAS) was equivalent to that of ΔpknG_Ms complemented with truncated GarA in all tested conditions so only the dataset for truncated GarA is shown for clarity.

https://doi.org/10.1371/journal.ppat.1006399.g002

Similarly, ΔpknG_Mt showed no growth defect compared to parental M. tuberculosis on minimum medium supplemented with glycerol (Fig 3A) or glucose or acetate (S6 Fig), but had a specific growth defect when asparagine or glutamate were used as the sole carbon source in liquid culture (Fig 3B & 3C). Reintroduction of pknG restored PknG expression (S7 Fig) and improved growth on asparagine and glutamate (Fig 3B & 3C), although full restoration of growth was not achieved, possibly due to deleterious effects of PknG over-expression.

Download:

• PPT
  PowerPoint slide
• PNG
  larger image
• TIFF
  original image

Fig 3. pknG disruption in M. tuberculosis caused a nutrient-dependent growth defect.

(A) All strains grew at the same rate on minimal medium supplemented with glycerol 0.2%. ΔpknG_Mt (blue diamonds) grew more slowly than wild type M. tuberculosis (black circles) when the sole carbon source was (B) glutamate 10 mM or (C) asparagine 10 mM. Plasmid-encoded PknG (empty circles) partially restored the growth defect. Graphs show the mean and standard deviation of three independent experiments.

https://doi.org/10.1371/journal.ppat.1006399.g003

The nutrient-specific phenotypes of M. tuberculosis and M. smegmatis gene knockouts were both specific to amino acid metabolism: garA-disrupted strains required glutamate or asparagine for growth, while pknG-disrupted strains had a defect in utilization of glutamate or asparagine. To investigate the likely conservation of function of the regulatory pathway between M. tuberculosis and M. smegmatis we used M. tuberculosis pknG and garA to complement the growth defects of M. smegmatis mutants (S8 Fig). Our results suggest conservation of function of PknG and GarA between a fast-growing saprophyte and a slow-growing pathogen (other differences between these organisms have been reviewed [33–35]).

In summary, the nutrient-specific phenotypes of pknG- and garA-disrupted M. smegmatis and M. tuberculosis support a role for these proteins in regulating amino acid metabolism.

PknG was needed for GarA phosphorylation in M. smegmatis and M. tuberculosis

To investigate kinase(s) that phosphorylate GarA in live mycobacteria and the stimuli that lead to kinase activity, we developed methods to distinguish phosphorylated GarA from the unphosphorylated form in cell extracts. Hexahistidine-tagged GarA shows a shift in mobility in SDS PAGE upon phosphorylation [8] but for untagged GarA this shift was too minor for reliable separation (for example S3 Fig shows a single band though Fig 4B demonstrates a mixture of phosphorylated and unphosphorylated GarA). Here we used three methods to determine whether GarA is phosphorylated in cells: (i) use of the Phos-tag reagent to retard the mobility of phosphorylated GarA in SDS PAGE of cell extracts from M. tuberculosis and M. smegmatis (Fig 4A), (ii) development of LC-MS/MS protocols to detect GarA in cell extracts of M. tuberculosis (Fig 4B), (iii) replacement of endogenous GarA with a hexahistidine-tagged version in M. smegmatis (Fig 4C).

Download:

• PPT
  PowerPoint slide
• PNG
  larger image
• TIFF
  original image

Fig 4. Investigation of GarA phosphorylation in M. smegmatis and M. tuberculosis.

(A) Addition of Phos-tag reagent to SDS-PAGE allowed separation of phosphorylated GarA (GarA-P) from GarA. Cell extracts were prepared from M. smegmatis and M. tuberculosis wild type and pknG deletion strains and complemented strains. (B) The same cell extracts of M. tuberculosis wild type and pknG deletion strain were analysed by LC-MS/MS to detect the tryptic peptide of GarA carrying the ETTS phosphorylation motif. The abundance of the peptide with no phosphorylation is shown in black and the T21-phosphorylated peptide is shown in red. A peptide from another part of the protein was also measured as a control (blue). (C) A reporter strain of ΔgarA_Ms carrying plasmids encoding hexahistidine-tagged garA, or variants of garA, confirmed that most phosphorylation occurred at the first threonine in the ETTS motif. GarA from cell lysates was visualised by Western blotting. Images shown are representative of three or more independent replicates.

https://doi.org/10.1371/journal.ppat.1006399.g004

M. smegmatis and M. tuberculosis growing in standard media contained two forms of GarA suggesting that cells contained a mixture of phosphorylated and unphosphorylated GarA (Fig 4A). In M. smegmatis lacking pknG the upper band was missing but could be restored by the introduction of plasmid-borne GarA, showing that PknG was the main kinase responsible for GarA phosphorylation. In M. tuberculosis lacking pknG the upper band was also missing, suggesting that PknG could also be responsible for phosphorylating M. tuberculosis GarA. Reintroduction of pknG to ΔpknG_Mt did not restore GarA phosphorylation, despite strong overexpression of pknG (S7 Fig). Non-physiological expression levels may influence GarA phosphorylation (see below for investigation into the conditions and stimuli that provoke phosphorylation).

To seek clarification about whether PknG may phosphorylate GarA in M. tuberculosis we decided to investigate the specific site of GarA phosphorylation. Several kinases have been reported to phosphorylate purified GarA at the second threonine (T22) while PknG is the only kinase shown to phosphorylate the first threonine (T21) [8, 36]. LC-MS/MS can distinguish between phosphorylation at T21 or T22 (S9 & S10 Figs). We have previously enriched phosphorylated GarA from cell extracts for detection by LC-MS/MS [8]. Here we developed a protocol avoiding enrichment to detect the various forms of GarA in cell extracts of M. tuberculosis. We used this protocol to determine the relative abundance of the three forms of GarA by comparison with peptide standards (unphosphorylated, T21-phosphorylated, and T22-phosphorylated). In wild type cells all three forms were detected (Figs 4B and S10) but the concentration of T22-phosphorylated form was always too low to quantitate reliably. However, in cell extracts of the M. tuberculosis pknG mutant strain there was no detectable T21-phosphorylated GarA, supporting the suggestion from Fig 4A that PknG may phosphorylate GarA in M. tuberculosis.

The equivalent peptides of M. smegmatis GarA were less amenable to mass spectrometry and so we created a reporter strain: ΔgarA_Ms + His₆garA, in which the garA deletion strain ΔgarA_Ms [9] is complemented by hexahistidine-tagged GarA (S11 Fig). We also generated variants with mutations in the phosphorylation motif of GarA ETTS. Western blotting showed that only those GarA variants lacking the phosphorylation site for PknG were predominantly unphosphorylated (Fig 4C). In summary, the data from Fig 4A–4C clearly demonstrate that PknG is the main kinase responsible for phosphorylation of GarA in M. smegmatis whereas phosphorylation by other kinase(s) may occur at lower levels, similar to the findings in C. glutamicum [32]. We also provide three independent lines of evidence to show that PknG may phosphorylate GarA in M. tuberculosis: site specificity (T21 phosphorylation in cells), loss of phosphorylation in ΔpknG_Mt, and conservation of function since M. tuberculosis pknG was able to complement the growth defect of M. smegmatis pknG knockout (S8 Fig).

Amino acids and optimal carbon sources triggered rapid phosphorylation of GarA

We next sought to identify the specific environmental signals that trigger phosphorylation or dephosphorylation of GarA. Since the active form of GarA is the unphosphorylated form, we predicted that this form would predominate in conditions where garA is essential, such as during amino acid deprivation. We used mass spectrometry to investigate GarA phosphorylation in M. tuberculosis and found only the unphosphorylated form in PBS-starved M. tuberculosis compared to a mixture in cells grown on standard media (S12 Fig). This trend of GarA phosphorylation in optimal medium but a lack of phosphorylation upon amino-acid starvation is similar to observations made on the homologous protein in C. glutamicum [37].

We then used the reporter strain of M. smegmatis to analyse a range of carbon and nitrogen sources separately for their effects on GarA phosphorylation (Fig 5A and 5B). Strikingly, the nutrients that led to the most phosphorylation (Fig 5A), are the amino acids that rescue the growth defect of ΔgarA_Mt and ΔgarA_Ms: glutamate, aspartate, glutamine and asparagine.

Download:

• PPT
  PowerPoint slide
• PNG
  larger image
• TIFF
  original image

Fig 5. GarA phosphorylation in M. smegmatis was regulated by nutrient availability.

The reporter strain of M. smegmatis was cultured in different media and cell lysates analysed by Western blot and densitometry. (A) Glutamate and related amino acids triggered phosphorylation of GarA: the nitrogen source is indicated and the carbon source was glucose. (B) The supplied carbon source affected phosphorylation of GarA: the carbon source is indicated and the nitrogen source was NH₄Cl. (C) Phosphorylation of GarA occurred rapidly when cells were cultured in poor medium and then given supplementary nutrients (initially 1% glucose with 10 mM NH₄Cl and 0.05% tyloxapol, then 1% v/v glycerol and 30 mM asparagine were added at time zero). (D) Dephosphorylation of GarA occurred slowly when cells switched from rich to poor medium (initially 1% glycerol with 30 mM asparagine and 0.05% Tween 80 then switched to 1% glucose with 10 mM NH₄Cl and 0.05% tyloxapol). (E) GarA was predominantly unphosphorylated when M. smegmatis were in stationary phase or starved in PBS. The reporter strain of M. smegmatis was grown in Sauton’s medium with shaking for 5 days. For the starvation experiment exponentially growing M. smegmatis were washed with PBS and incubated in PBS with 0.05% tyloxapol for 5 days. Values represent mean and standard deviation of at least three independent replicates.

https://doi.org/10.1371/journal.ppat.1006399.g005

When carbon sources were compared there was least phosphorylation during growth on acetate or glucose (Fig 5B). The correlation between the extent of GarA phosphorylation and the severity of growth phenotype of ΔgarA_Ms was weaker when comparing carbon sources, suggesting that there may be other sensory/regulatory input(s) that remain to be identified. Thus we conclude that nutrients are likely stimuli for PknG activity, and, at least in M. smegmatis, glutamate and related amino acids are the most important. Notably, the sensory mechanism remains to be identified and there are also likely to be additional stimuli influencing kinase activity.

In principle, reversible phosphorylation of GarA could allow cells to respond rapidly to changing nutrient availability. Since GarA interacts directly with enzymes of central carbon and nitrogen metabolism this would allow a more rapid response than alterations in gene expression level. To investigate the dynamics of adaption we grew the reporter strain on media promoting low or high phosphorylation of GarA and then exchanged the medium at mid-log phase, monitoring GarA phosphorylation until it had stabilized. Addition of glycerol/asparagine to a culture grown in medium containing glucose/ammonium chloride led to GarA phosphorylation within the shortest time period that could be sampled with accuracy (15 minutes) (Fig 5C). By contrast, when cells were transferred from standard Sauton’s medium to minimal Sauton’s medium, reductions in GarA phosphorylation were only seen after three hours, suggesting that dephosphorylation occurred slowly if at all (Figs 5D and S13). Similarly to M. tuberculosis, we observed only unphosphorylated GarA in starved or stationary phase M. smegmatis (Fig 5E). Our observations suggest that the PknG—GarA system may allow cells to adapt rapidly to an increase in amino acid availability: glutamate or related amino acids would stimulate PknG to phosphorylate GarA and hence enable glutamate catabolism. However, adaptation to nitrogen starvation may take longer and could involve new protein synthesis or protein dilution through cell division.

GarA was required for preservation of intracellular glutamate during extended stationary phase

Having established that GarA is predominantly in the active, unphosphorylated form during starvation and stationary phase in M. tuberculosis and M. smegmatis (S12 Fig and Fig 5E), we used a metabolomics approach with M. smegmatis to test the specific effects of garA knockout on intracellular metabolites. Since GarA stimulates glutamate synthase activity and inhibits enzymes involved in glutamate catabolism, we predicted that ΔgarA_Ms might have a lower concentration of intracellular glutamate compared to wild type M. smegmatis. We used a targeted mass spectrometry approach to monitor the intracellular concentration of glutamate and 39 additional metabolites of central carbon metabolism in stationary phase cultures. Of the 40 metabolites analysed, glutamate and glutamine showed the greatest difference between ΔgarA_Ms compared to wild type. In wild type cells the concentration of glutamate was maintained at a relatively steady level throughout 28 days while the concentrations of glutamine and many other metabolites declined during the first 7 days and were then steady (Figs 6A & S14). As predicted, ΔgarA_Ms had lower intracellular glutamate in stationary phase (from day 7 onwards). The intracellular concentration of glutamine was transiently elevated during entry of ΔgarA_Ms into stationary phase and then declined from day 7 onwards (Fig 6A). The decline in glutamate and glutamine in extended phase could be due to catabolism through uninhibited GDH and KDH. After 28 days the ΔgarA_Ms strain began to show loss of viability. Metabolite sampling was discontinued and viability was monitored for 10 further weeks, by which point cultures of ΔgarA_Ms contained 100-fold fewer CFU ml^-1 than wild type M. smegmatis (Fig 6B). The depletion of intracellular glutamate in stationary phase ΔgarA_Ms and the defect in long-term survival provides a functional demonstration of our predicted model of metabolic regulation and highlights the importance of glutamate homeostasis for bacterial viability.

Download:

• PPT
  PowerPoint slide
• PNG
  larger image
• TIFF
  original image

Fig 6. GarA was required during stationary phase for the maintenance of intracellular glutamate and for survival.

(A) M. smegmatis lacking garA failed to maintain intracellular glutamate and glutamine pools during extended stationary phase in 7H9 medium. Intracellular glutamate and glutamine were measured for wild type M. smegmatis (black circles), ΔgarA_Ms (red squares), and complemented ΔgarA_Ms (black triangles). Inset graphs show intracellular metabolites for the same experiment at day 28 (B) M. smegmatis lacking garA gradually lost viability during prolonged stationary phase. Cells were cultured in 7H9 medium over a time period of five months. Aliquots were withdrawn at regular intervals and surviving cells were plated on 7H10 to calculate CFU ml^-1. All experiments were repeated at least 3 times and data show the mean with standard deviation.

https://doi.org/10.1371/journal.ppat.1006399.g006

Metabolome analysis revealed amino acid metabolism as the main target of regulation by GarA and PknG in M. smegmatis

Disruption of pknG in M. tuberculosis has previously been shown to perturb intracellular glutamate and glutamine levels [5]. To examine the effect of garA disruption or pknG disruption on wider cell metabolism we grew M. smegmatis and variant strains for untargeted metabolome analysis of about four hundred annotated metabolites by mass spectrometry. An unbiased comparison of the metabolomes of ΔgarA_Ms with the parent and complemented strains identified a set of 15 metabolites with lower concentration in ΔgarA_Ms (log2(fold change)>0.5, and q-value<0.05) (Fig 7A and Table 1). Eight of the fifteen significantly changed metabolites were amino acids or intermediates in amino acid biosynthesis. Striking reductions were seen in the intracellular concentrations of glutamate and two direct products of glutamate: GABA and oxoproline/pyroglutamate (Fig 7A), and these changes were reversed by plasmid-borne garA (Fig 7B). Extracellular metabolites were also analysed but significant differences were not found (S15 Fig).

Download:

• PPT
  PowerPoint slide
• PNG
  larger image
• TIFF
  original image

Fig 7. Deletion of garA or disruption of GarA phosphorylation caused changes in intracellular glutamate and other changes in the metabolome of M. smegmatis.

Intracellular metabolites from wild type M. smegmatis and mutants were analysed by mass spectrometry using an untargeted metabolomics approach. M. smegmatis lacking garA has lower intracellular glutamate and metabolites related to glutamate than the parental strain (A) but plasmid encoded GarA reversed this change (B). M. smegmatis expressing truncated GarA that lacks phosphorylation sites had higher intracellular ornithine than wild type (C). Intracellular glutamate concentrations for the mutant strains are compared to wild type in panel (D), together with those metabolites that were significantly changed in >1 mutant strain but not in complemented strains. pSer is O-phosphoserine, Orn is ornithine, Cit is citrulline, MaltoP is maltopentaose. (A-C) Each point on the scatter plots represents a single metabolite. Metabolites with the greatest fold-change and statistical significance are highlighted (log2(fold change)>0.5 and q-value<0.05), thresholds marked as dashed lines on graphs): metabolites at lower concentration are blue and those at higher concentration are red. These data were taken from cells growing in Middlebrook 7H9 in early exponential phase and represent the mean from at least 3 independent experiments.

https://doi.org/10.1371/journal.ppat.1006399.g007

Download:

• PPT
  PowerPoint slide
• PNG
  larger image
• TIFF
  original image

Table 1. Summary comparing the intracellular metabolome of mutant strains with parental M. smegmatis.

Metabolites that are significantly changed in more than one strain are in bold type. Metabolites related to amino acid metabolism are highlighted with asterisks *. Metabolites are listed in order of fold-change compared to M. smegmatis, and those that were >2-fold changed are shaded.

https://doi.org/10.1371/journal.ppat.1006399.t001

To examine the impact of disrupting GarA phosphorylation, we next analysed the intracellular metabolites of ΔgarA_Ms carrying non-phosphorylatable GarA and ΔpknG_Ms (Tables 1 and S1, Figs 7C and S16). We predicted elevation in intracellular glutamate when GarA cannot be phosphorylated, a reversal of the glutamate deficiency when garA is deleted. Intracellular glutamate was indeed significantly higher (1.3-fold change, q-value<0.0001, Fig 7D, S1 Table) but below our chosen threshold for inclusion in Table 1 (log2(fold change)>0.5). The majority of the metabolites that were significantly changed were amino acids or involved in amino acid metabolism, notably intermediates of arginine biosynthesis (ornithine, citrulline), which were elevated in mutant strains. The wide reaching changes in amino acid metabolism could be consequences of perturbed glutamate metabolism, the central hub of NH₃ transfer. Beyond amino acid metabolism, we cannot differentiate whether changes in metabolite concentrations are indirect consequences of altered physiology and altered amino acid metabolism or indicative of other GarA/PknG targets.

Previously pknG deletion has been shown to cause raised intracellular glutamate and glutamine in M. tuberculosis [5] and raise glutamate production by C. glutamicum [37], whereas garA deletion abolished glutamate production by C. glutamicum. Our metabolome analysis of M. smegmatis reinforces the functional link between PknG and GarA and their role in regulating amino acid metabolism in Actinobacteria, which is further supported by perturbations in amino acid metabolism seen in the metabolome of pknG-disrupted M. bovis BCG (S7 Table). We chose a medium in which all strains grew at the same rate and in which the wild type strains of M. smegmatis and M. bovis BCG normally contain a mixture of phosphorylated and unphosphorylated GarA. We observed that perturbation of pknG caused changes in the concentrations of intracellular metabolites despite the absence of an obvious growth defect. Together with Figs 1–4, this highlights the relevance of phosphorylation for regulation of GarA function, and the importance of proper regulation of phosphorylation.

Bacterial strains, media, and culture

M. tuberculosis H37Rv and M. smegmatis mc²155 were routinely cultured on Middlebrook 7H10 agar (Oxoid) with 10% ADN (0.5% bovine serum albumin, 0.2% dextrose, 0.085% NaCl) and Middlebrook 7H9 medium (Oxoid) with 10% ADN and 0.05% Tween 80. A list of strains and plasmids used in this study is provided (S8 & S9 Tables). To analyse nutrient utilization a minimal version of Sauton’s medium was made (3.7 mM KH₂PO₄, 2 mM MgSO₄, 9.5 mM sodium citrate, 0.17 μM ferric ammonium citrate, pH 7.0 [51]), to which nutrients were added. Standard Sauton’s consisted of the recipe above with additional 1% glycerol, 30 mM asparagine and 0.05% Tween 80. To disperse the culture surfactants were added at 0.05% w/v: either Tween 80, which can be utilized as a carbon source, or tyloxapol, which cannot. When required antibiotics were used at the following concentrations: kanamycin (50 μg/mL), hygromycin (100 μg/ml).

Deletion of garA from M. tuberculosis H37Rv to generate ΔgarA_Mt

We have previously characterized an M. tuberculosis garA conditional knockdown mutant (cΔgarA_Mtb) in which the only functional copy of garA was inserted at the L5 att site under the transcriptional control of the repressible Pptr promoter using an L5-based integrative plasmid [9]. To obtain a garA null mutant, we decided to replace this plasmid with another one not containing garA. This is possible since introduction of an L5-integrative plasmid to a mycobacterial strain in which the L5 att site is already occupied by a similar plasmid leads to an efficient switching between the two plasmids [52]. To this end conditional mutant cΔgarA_Mtb was electroporated with pMV306, encoding hygromycin resistance, and switched mutants were selected on hygromycin (50 μg/ml) and asparagine (10 mM) to allow the growth of the garA null mutants. Hygromycin-resistant colonies were analysed to confirm the loss of kanamycin resistance (encoded in the garA containing integrative plasmid) and inability to grow in the absence of asparagine of the deletion strain ΔgarA_Mt.

Plasmid complementation of ΔgarA_Ms and ΔgarA_Mt with variants of GarA

M. smegmatis garA with its promoter region was cloned in the plasmid pRBexint [53] to create pRBexint-garA as described [9]. M. tuberculosis garA was cloned with a hexahistidine tag into pRBexint to create pRBexint-His₆garA and with a HA tag into pTTP1B [54] to create pTTP1B-garAHA. Variants of each gene were created by site directed mutagenesis to disrupt the phosphorylation motif ETTS. Also a truncated version of M. smegmatis garA was constructed, garA_39-143, with the first 38 residues missing.

Macrophage infection

THP-1 human cell line was grown at 37°C in a 5% CO₂ atmosphere and maintained in RPMI medium (Gibco) supplemented with 10% fetal bovine serum (Gibco). After expansion, THP-1 cells were differentiated into macrophages and infected with M. tuberculosis in 96-well plates with a multiplicity of infection of 1:20 CFU per macrophage as previously described [55]. After 90 minutes of incubation at 37°C, the medium was removed, and cells were washed twice with 100 μl of warm phosphate buffered saline to remove extracellular bacteria. Finally, 100 μl of warm RPMI (with added 20 mM asparagine for S2 Fig), was added to each well and the plate was incubated at 37°C. RPMI with or without additional asparagine was replaced every 48 hours. To enumerate intracellular bacteria the medium was removed from three wells, and 100 μl of 0.05% sodium dodecyl sulfate was used to lyse macrophages. The suspensions obtained were immediately diluted in 7H9 and plated to determine viable counts. About 95% of macrophages remained viable during the entire experiment, as determined by Trypan blue exclusion.

Ethics statement

All investigations involving animals were carried out according to the requirements of the Animals (Scientific Procedures) Act 1986 with the consent of the University of Leicester Animal Welfare & Ethics Board. The Home Office Licence number is 60/4327.

Assessment of virulence in mice

BALB/c mice (female, 6–8 weeks old) were purchased from Charles River, UK and acclimatised for 7 days prior to M. tuberculosis challenge. Frozen aliquots of bacterial strains were thawed and passed through a blunt needle 10 times to disperse clumps and adjusted to 2x 10⁶ CFU ml^-1 prior to infection. Mice were inoculated via the intranasal route by the drop-wise administration of 50 μl bacterial suspension onto the nostril of a lightly anaesthetised mouse (2.5% (v/v) flurothane over oxygen) held in a vertical position. Mice were monitored for full recovery from anaesthetic prior to return to their cages. Mice were housed in cages of 5 animals within a negative pressure rigid isolator (air change rate 25 changes/hr; pressure -100Pa). Mice had free access to water and diet (5LF2, LabDiet) and monitored daily for welfare and signs of disease over the 28-day experimental period.

Experimental groups (n = 20) were inoculated with either M. tuberculosis H37Rv wild type, ΔgarA_Mt deletion or complementation strain at 10⁵ CFU per animal. A cage of 5 mice from each group was euthanised by cervical dislocation at day 1, 7, 21 and 28 and the lungs and spleen were aseptically removed post-mortem (conformation of death by rigor mortis). Lung and spleen tissue (lung only at day 1) was homogenised using a FastPrep-24 (MP Biomedicals) in 15ml tubes containing 10 matrix S beads (MP Biomedicals) and 9 ml PBS. Three bursts of 20 seconds at 4 m/s with a five minute cool-down in-between was used to homogenate the organs for enumeration of bacteria on 7H10 agar; kanamycin (50 μg/ml) and/or hygromycin (100 μg/ml) was added as required.

Growth of M. tuberculosis garA deletion strain on Middlebrook 7H10 to examine the phenotype of garA gene deletion and complementation

M. tuberculosis H37Rv strains were grown in 7H9/ADN/Tween 80 with 30 mM asparagine until OD reached 0.6–0.9. The cultures were then diluted in Middlebrook 7H9 medium and plated on Middlebrook 7H10 ADN supplemented with and without 30 mM asparagine. Plates were incubated at 37°C and images taken after two weeks.

Plasmid complementation of ΔpknG_Mt and growth in minimum Sauton’s medium

The gene encoding PknG was amplified by Pfu Ultra Hf DNA polymerase (Agilent) using an upper primer (AC 145) designed to contain an NheI site immediately before the start codon, and a lower primer (AC146) designed to contain the HA-coding sequence in frame with the coding sequence of pknG, followed by a stop codon and a XbaI site (S9 Table). For expression in M. tuberculosis, fragments containing the Phsp60 promoter sequence from HindIII/NheI-digested pAL36 [56] and the PknG-HA encoding gene digested with NheI/XbaI were transferred to HindIII/XbaI-digested pMV306 vector [57] giving the mycobacterial plasmid pAL299.

The plasmid was electroporated into strain ΔpknG_Mt and kanamycin-resistant recombinants that had integrated the vector with the PknG insert at the attB site were selected on Middlebrook 7H11-OADC (BD) plates and subjected to PCR screening and Western blotting. This approach allowed one colony to be selected that showed the expected PCR amplification products as well as the expected band in PknG-specific Western blotting (S7 Fig) (anti-PknG serum was a generous gift from E Houben and J Pieters, VU Medical Center, Amsterdam and University of Basel). This clone, “ΔpknG_Mt + pknG” was used for further experiments to evaluate the growth under selected amino-acid deprivation conditions.

In these experiments, the growth of M. tuberculosis wild-type, deletion and complemented strains was measured by monitoring OD₆₀₀ of cultures grown in glass tubes at 37°C in standing conditions. Bacteria were grown until late-exponential phase (OD₆₀₀ 0.6–0.9) in 7H9 ADC medium, washed twice and diluted in the Sauton’s minimum medium supplemented by a specific carbon source to an initial OD₆₀₀ of 0.04–0.05. Sauton’s minimum medium was supplemented with ammonium chloride 10 mM plus one of the following carbon sources: L-Asparagine 10 mM, L-Glutamate 10 mM, Glycerol 0.2%, Glucose 1%, Acetate 0.2%. Data plotted represent the mean and standard deviation of at least three independent experiments.

Measurement of nutrient requirements of M. smegmatis

The growth of M. smegmatis was measured by monitoring OD of cultures grown in microplates at 37°C with shaking. The inoculum used was a late-exponential phase culture (OD₆₀₀ 0.6–0.9) in 7H9 ADN medium, which was dispersed by passing through a needle and then diluted in the required medium to an initial OD of 0.01. Growth curves used at least five wells per strain and were performed in triplicate. Figures show the mean and standard deviation for a representative experiment.

Growth of M. smegmatis to prepare cellular proteins for determination of GarA phosphorylation

In order to analyse the phosphorylation status of GarA in M. smegmatis mc²155, strains were grown in 7H9 medium to OD₆₀₀ 0.6–0.9 then diluted into 10 ml of 7H9 without ADN (or variant Sauton’s medium where specified) to OD₆₀₀ 0.01 in a 50 ml falcon tube. Cultures were grown with shaking at 37°C and cultures were harvested at OD₆₀₀ 0.6–0.9 by centrifugation at 4°C. Cells were either lysed by sonication either directly in SDS sample buffer or by sonication in cold Tris-buffered saline pH 8.0 containing PhosSTOP Phosphatase Inhibitors (Roche) and cOmplete Protease Inhibitors (Roche) followed by centrifugation and addition of SDS sample buffer. Protein concentration was measured by using BCA Protein Assay Reagent (Pierce).

Growth of M. tuberculosis and preparation of cellular proteins to determine GarA phosphorylation

M. tuberculosis H37Rv strains were grown in standard Sauton’s medium containing 0.05% Tween 80 to OD₆₀₀ 0.6–0.9 and then diluted into 30 ml of the specified Sauton’s medium to OD₆₀₀ 0.01. Cultures were grown with shaking at 37°C to OD₆₀₀ 0.6–1.0 and harvested by centrifugation. Cell pellets were resuspended in 1 ml SDS sample buffer (for mass spectrometry) or in 1 ml 1 M Tris HCl pH 8.0 containing PhosSTOP (Roche) and cOmplete (Roche) (for Phos-Tag analysis). Cells were lysed using a FastPrep (MP Biomedicals) with addition of glass beads 150–210 μm (Sigma) and insoluble material was removed by centrifugation. Cells were either killed by heating to 100°C for 30 minutes (for mass spectrometry) or were rendered non-infectious by filtration (for Phos-Tag). Protein concentration was measured by using BCA Protein Assay Reagent (Pierce).

Determination of GarA phosphorylation by Western blotting using Phos-tag

7.5 μg protein was separated by SDS PAGE (10% acrylamide) containing 50 mM acrylamide-pendant Phos-tag ligand (Wako Pure Chemical) and 100 mM of Zn(NO₃)₂. The running buffer (pH 7.8) contained 0.1 M Tris HCl, 0.1 M 3-(N-morpholino) propanesulfonic acid (MOPS), 0.1% w/v SDS and 5 mM sodium bisulfite. The gel was washed as described previously [58] before transfer onto nitrocellulose membrane (Hybond-C extra, Amersham Biotech). Rabbit anti-GarA serum was kindly provided by Dr I Rosenkrands (Statens Serum Institute) and GarA was detected using goat anti-rabbit alkaline phosphatase (Sigma) with SIGMAFAST BCIP/NBT (Sigma). Blots were photographed and images analysed using ImageJ 1.45s software (U.S. National Institutes of Health, Bethesda, MD, USA) [59]. For SDS PAGE and blotting without Phos-tag ligand, the same serum, antibodies and method of analysis were used, as previously [9], and representative images are supplied (S17 Fig).

Determination of GarA phosphorylation by mass spectrometry

Synthetic peptides corresponding to the tryptic peptides of GarA were purchased from CBio and analysed by LC-MS/MS using an RSLCnano HPLC system (Dionex) and 4000 Q-trap mass spectrometer (Applied Biosystems, Warrington, UK). Samples were loaded at high flow rate onto a reverse-phase trap column (0.3 mm i.d. x 1 mm), containing 5 μm C18 300 Å Acclaim PepMap media (Dionex) maintained at a temperature of 37°C. The loading buffer was 0.1% formic acid / 0.05% trifluoroacetic acid / 2% acetonitrile. After a 4 minute wash step, peptides were eluted from the trap column at the flow rate of 0.3 μl/min using an increasing proportion of mobile phase B (80% acetonitrile/0.1% formic acid); 4–45% B in 26 minutes, 45–90% B in 1 minute, held at 90% for 8 minutes, 90–4% in 1 minute, re-equilibration at 4% for 10 minutes. Eluted peptides were separated through a reverse-phase capillary column (75 μm i.d. x 250 mm) containing Symmetry C18 100 Å media (Waters) that was manufactured in-house using a high-pressure packing device (Proxeon Biosystems). The output from the column was sprayed directly into the nanospray ion source of the 4000 Q-Trap mass spectrometer.

The three forms of the peptide containing the phosphorylation sites (ETTS) and a control peptide (corresponding to a different tryptic peptide from the GarA protein) were found to separate by retention time and be distinguishable by their fragmentation spectra (Table 2). Multiple reaction monitoring (MRM) precursor/product ion transitions were chosen based on the fragmentation data and were used to produce standard curves for each peptide by injecting known amounts: 50, 100, 200, 400, 600 and 2000 fmol (S9 Fig).

Download:

• PPT
  PowerPoint slide
• PNG
  larger image
• TIFF
  original image

Table 2. Precursor/Product ion transition used for multiple reaction monitoring of GarA peptides.

https://doi.org/10.1371/journal.ppat.1006399.t002

Soluble protein extracts of M. tuberculosis were run on 1D-gels, the region of interest excised, and in-gel trypsin digestion carried out upon each. Gel slices was destained using 200 mM ammonium bicarbonate/20% acetonitrile, followed by reduction (10 mM dithiothreitol, Melford Laboratories Ltd., Suffolk, UK), alkylation (100 mM iodoacetamide, Sigma, Dorset, UK) and enzymatic digestion with trypsin (sequencing grade modified porcine trypsin, Promega, Southampton, UK) in 50 mM triethylammonium bicarbonate (Sigma) using an automated digest robot (Multiprobe II Plus EX, Perkin Elmer, UK). After overnight digestion, samples were acidified using formic acid (final concentration 0.1%) and analysed by LC-MS/MS using the gradient and MRM transitions outlined above. At the start and end of each analytical run, the synthetic peptides were analysed to allow comparison with the standard curves for the purpose of assessing technical reproducibility.

PBS starvation and extended stationary phase of M. smegmatis

For the PBS starvation experiment using M. smegmatis, strains were cultured and harvested as described above. Pellets were washed twice in PBS and resuspended in 30 ml PBS with 0.05% tyloxapol and incubated at 37°C without shaking for up to 5 days. Samples were taken at different time points for analysis of GarA phosphorylation. For analysis of GarA during stationary phase, cells were cultured as described above and incubated for 5 days at 37°C with shaking. For the extended stationary phase experiment strains were cultured from single colonies in 5 ml 7H9/ADN/Tween 80 in a closed 30 ml Universal tube for 3 months at 37°C with shaking. The first sample was taken after the culture reached late exponential phase (OD₆₀₀ 0.6–1.0) and bacterial viability was estimated by measuring CFU ml^-1 by plating aliquots of bacterial suspension on 7H10/ADN plates. Further samples were taken every 4 weeks and CFU determined.

Metabolomics

Samples for metabolic analysis were collected during early exponential growth phase (OD₆₀₀ 0.3–0.5) by fast filtration as described previously for Mycobacteria [60]. Briefly, a sample volume equivalent to a biomass of 4 ml at OD₆₀₀ of 1.0 was filtered (MF-Millipore Membrane, 0.45 μm), briefly washed with ammonium carbonate buffer (75 mM, pH 6.6) and transferred to 3 ml ethanol 60% (v/v) at 78°C for 2 min. Samples were dried at 30°C in a SpeedVac equipped with a cooling trap at -85°C. The dried extracts were dissolved in 100 μl water for metabolite analysis. Sample collection and processing of extracellular samples for metabolomics analysis was performed as previously described [61]. Quantification by targeted mass spectrometry was performed by ion pairing–reverse phase liquid chromatography tandem mass spectrometry on a Waters Acquity UHPLC coupled to a Thermo TSQ Quantum Ultra triple quadrupole instrument using fully U-¹³C-labled yeast extract as internal standard [62]. Non-targeted mass spectrometry was performed on an Agilent 6550 QTOF instrument [63]. Annotation was performed based on accurate mass determination of ions and the KEGG reference list (tolerance 0.001 Da). Removal of unknown ions and annotated ion adducts resulted in 397 putatively annotated ions with unique m/z. Two technical replicate measurements were performed for each sample and merged using their mean.

Differential analysis was performed applying an unequal t-test using MatLab (The Mathwork, Natick). For each metabolite, pairwise comparisons were made between test strain and wild type using an unequal t-test, resulting in fold-change and associated q-value (corrected for multiple hypotheses using the Benjamini Hochberg procedure). The results of pairwise comparisons are graphically represented as volcano plots. We chose criteria for significant changes of q<0.05 and absolute log2(fold change)>0.5. The borders are shown in the volcano plots and metabolites passing these thresholds are coloured red if higher than wild type or blue if lower. The metabolites scored as significant were compiled for each strain and these lists used for comparison between multiple strains. Metabolites that were significantly changed in >1 strain are highlighted by asterisks in the tables and fold-changes in concentration compared to wild type are summarized in Fig 7D. Some metabolites, including cAMP and mycobactin, were significantly altered in most strains, including complemented strains and other mutant strains unrelated to this project. It is possible that there were technical reasons making the extraction or quantification of these metabolites more variable, or alternatively these metabolite pools may be more susceptible to changes caused by cell stress. These variable metabolites were excluded from Table 1 and are listed in S1 Table.