Synthetic growth phenotypes of Escherichia coli lacking ppGpp and transketolase A (tktA) are due to ppGpp-mediated transcriptional regulation of tktB


*E-mail; Tel. (+1) 301 5949841; Fax (+1) 301 4960243.


Many physiological adjustments to nutrient changes involve ppGpp. Recent attempts to deduce ppGpp regulatory effects using proteomics or gene profiling can rigorously identify proteins or transcripts, but the functional significance is often unclear. Using a random screen for synthetic lethals we found a ppGpp-dependent functional pathway that operates through transketolase B (TktB), and which is ‘buffered’ in wildtype strain by the presence of an isozyme, transketolase A (TktA). Transketolase activity is required in cells to make erythrose-4-phosphate, a precursor of aromatic amino acids and vitamins. By studying tktB-dependent nutritional requirements as well as measuring activities using PtalA-tktB′-lacZ transcriptional reporter fusion, we show positive transcriptional regulation of the talA-tktB operon by ppGpp. Our results show the existence of RpoS-dependent and RpoS-independent modes of positive regulation by ppGpp. Both routes of activation are magnified by elevating ppGpp levels with a spoT mutation (spoT-R39A) defective in hydrolase but not synthetase activity or with the stringent suppressor mutations rpoB-A532Δ or rpoB-T563P in the absence of ppGpp.


Bacteria have evolved with complex protective global responses to stress. In enteric bacteria, the accumulation of guanosine 5′-diphosphate, 3′-diphosphate and/or guanosine 5′-triphosphate, 3′-diphosphate, collectively referred to as the (p)ppGpp nucleotides (Cashel and Gallant, 1969), is a common response to different sources of nutritional stress. Here we shall use ppGpp as an abbreviation for pppGpp and ppGpp. There are many examples of regulation that involve ppGpp as a function of normal growth and during stress. Phenotypes associated with a complete deficiency of ppGpp (ppGpp0) include multiple amino acid requirements, filamentation, nucleoid-partitioning defects, agglutination changes, adhesion and motility defects arising from the absence of fimbriae and flagella, respectively, and decreased virulence in pathogenic bacteria (Xiao et al., 1991; Cashel et al., 1996; Magnusson et al., 2005; 2007; Braeken et al., 2006).

Random genetic approaches to define ppGpp function have not been frequently used. Instead, mutants of the relA and spoT genes were encountered in different strains of Escherichia coli; mapped, isogenic strains were constructed and phenotypes characterized. Selections based on these phenotypes identified mutations in relB and relC but not the genes involved in the wide range of functions just mentioned (Cashel et al., 1996). One way to identify cellular functions mediated by ppGpp is to isolate spontaneous extragenic suppressors of defects in ppGpp0 cells. Many such mutations have been isolated that reverse the multiple amino acid auxotrophic phenotype of ppGpp0 cells to allow growth on minimal glucose media. So far, such mutations occur exclusively in rpoB, rpoC and rpoD RNA polymerase subunit genes and have been termed ‘stringent’ mutations (Zhou and Jin, 1998; Murphy and Cashel, 2003).

We used a synthetic lethal approach to look for genes involved in ppGpp-dependent functions. Two mutations are synthetic lethal if either in isolation is viable but together cause inviability. Two separate non-lethal mutations that confer a growth defect more severe than either single mutation can be called synthetic growth inhibition (Phizicky and Fields, 1995; Ooi et al., 2006). The interpretation is that synthetic growth inhibition reflects an important genetic interaction, whereas synthetic lethality reflects an essential genetic interaction. Such interactions reveal genes that function in parallel pathways and ‘buffer’ each other biologically or function within the same pathway but independently contribute to the strength of the signal in the pathway. In general, synthetic lethal screens help uncover pathway(s) that are conditionally essential or significantly influence growth.

Here a genetic screen is used to search for pathways that show ppGpp-mediated regulation. We isolated an insertion in tktA that gives synthetic growth defects in a ppGpp0 strain. This led to the identification of tktB as a transcriptional target of ppGpp and evidence that activation of tktB transcription by ppGpp occurs both through the modulation of RpoS levels and independent of RpoS. The physiological relevance of the two modes of regulation is assessed.


A screen for synthetic lethal mutations in a ppGpp0 host identifies an insertion in tktA

The rationale behind the screen is that an unstable plasmid replicon carrying a gene required for growth would be retained through selection during growth conditions that favour plasmid loss through segregation (Phizicky and Fields, 1995). Low copy number plasmid pHR14 is a temperature-sensitive pSC101 replicon with functional spoT and lacIq genes. Replication of pHR14 is stable at 30°C significantly restricted at 38°C and completely abolished at 42°C. Growth at 38°C without selection causes plasmid loss and dilution of cellular LacI levels. In the ΔlacI (lacZ+) ΔrelA251ΔspoT207 strain CF11722 carrying plasmid pHR14, plasmid loss results in an increase in β-galactosidase expression and appearance of blue colonies in plates containing the chromogenic substrate Xgal. Mutations that limit plasmid loss will give rise to white or pale blue colonies. Among many other possibilities, mutations that render spoT gene functions essential for growth are expected to select against plasmid loss. A similar approach has been used to identify a synthetic lethal mutation in ftsEX mutant (Reddy, 2007).

CF11722 with pHR14 was subjected to Tn5 transposon mutagenesis and dilutions were plated on Luria–Bertani (LB) Xgal plates with trimethoprim to obtain about 200 well-separated single colonies on each plate. A pale blue colony was identified after screening 5000 blue colonies. The transposon insertion in this clone impaired growth when moved into the ppGpp0 strain CF10237 [by P1(vir) transduction] but not in the wild-type strain CF1648 (see Fig. 1B). Thus, growth inhibition is dependent on ppGpp deficiency and the growth phenotype from the transposon insertion is a severe growth impairment rather than lethality. Sequencing the transposon-chromosome junction localized the insertion to the distal half of the tktA open reading frame (ORF) and it was designated as tktA::Tp.

Figure 1.

Growth properties of transketolase mutants.
A. Strains on plate: tktA::Tp (CF13942), ppGpp0, i.e. relA256 spoT212 (CF10237), ppGpp0tktA::Tp (CF13926) and tktA tktB (CF13927).
B. LB agar after 18 h.
C. Minimal media with glucose and casamino acids after 36 h.
D. Minimal media with glucose casamino acids, tryptophan and pyridoxine after 36 h. All incubations were at 37°C.

In E. coli, tktA and tktB genes encode redundant transketolases that catalyse synthesis of a key metabolic intermediate, D-erythrose-4-phosphate. As shown in Fig. 2, lack of transketolase activity would result in the failure to synthesize erythrose-4-phosphate, a precursor required for biosynthesis of aromatic amino acids, aromatic vitamins like para-amino benzoic acid and pyridoxine (PN), a precursor of pyridoxal phosphate (Fraenkel, 1987; Zhao and Winkler, 1994; Pittard, 1996).

Figure 2.

Pathways for biosynthesis of the intermediary metabolite D-erythrose-4-phosphate and amino acids and vitamins derived from it. The enzymes involved at each step are indicated by gene names that encode them; broken arrows represent multiple steps in the pathway. Xly-5-P, D-xylulose-5-phosphate; Rib-5-P, D-ribose-5-phosphate; Gly-3-P, D-Glyceraldehyde-3-phosphate; Sedohep-7-P, Sedoheptulose-7-phosphate; Fruc-6-P, D-fructose-6-phosphate; Ery-4-P, D-erythrose-4-phosphate.

A ΔtktB::kan mutation confers synthetic growth defects in the tktA mutant similar to that observed from ppGpp deficiency

We examined LB growth in the presence of tktA and tktB mutations singly and in combination. We chose two tktA alleles, tktA::Tn10, an undefined insertion in tktA (Iida et al., 1993) and the ΔtktA::kan deletion-insertion allele from the Keio collection (National Bioresource project, Japan) as well as two ΔtktB::kan alleles (Iida et al., 1993 and Keio collection). The tktA mutants were slightly slower growing on LB while the tktB mutants showed no growth defect. However, when combined, the tktA tktB double mutant shows growth inhibition in LB comparable to that observed in the tktA ppGpp0 strains (Fig. 1B; Table 1A, last column).

Table 1.  Growth deficiency and requirements seen in tktA mutants in the absence of ppGpp arise from lowered transketolase B activity.
Relevant genotypesM9 Glucose minimal with (48 h)
20 AA + PN20 AA20 AA − phe20 AA − trp20 AA − tyrLB (24 h)
  1. Growth estimates are based on single colony size as described in the methods, with IPTG used at 0.25 mM where indicated. For plasmid bearing strains, ampicillin was used at final concentration of 100 μg ml−1 in LB and 50 μg ml−1 in minimal media. Strains in panel A rows 1–7 are: CF1648, CF13912, CF13969, CF13927, CF10237, CF13913 and CF13926; panel B CF13927 (rows 1–4) and CF13913 (rows 5–8). pHR30 is a plasmid for inducible expression of tktB.

 1 Wild type++++++++++++++++++
 2 tktA++++++++++++
 3 tktB++++++++++++++++++
 4 tktA tktB++±
 5 ppGpp0+++++++++++
 6 ppGpp0tktA::Tn10+++±
 7 ppGpp0tktA::Tp+++±
 1 tktA tktB/pHR30++++++++++++
 2 tktA tktB/pHR30 + IPTG++++++++++++++++++
 3 tktA tktB/vector++±
 4 tktA tktB/vector + IPTG++±
 5 ppGpp0tktA/pHR30++++±+
 6 ppGpp0tktA/pHR30 + IPTG+++++++++++
 7 ppGpp0tktA/vector+++±
 8 ppGpp0tktA/vector + IPTG+++±

Comparison of growth requirements between tktA ppGpp0 mutant and the tktA tktB double mutant

Transketolase mutants require aromatic amino acids and pyridoxine (Fraenkel, 1987; Iida et al., 1993; Zhao and Winkler, 1994). The growth requirements of the tktA tktB double mutants are due to the lack of erythrose-4-phosphate. This compound is generated from glyceraldehyde-3-phosphate and fructose-6-posphate by transketolase or from sedoheptulose-7-phosphate and glyceraldehyde-3-phosphate by transaldolase. However, the syntheses of the latter two substrates require transketolase. Therefore, erythrose-4-phosphate is not synthesized in a tktA tktB double mutant (Fig. 2).

Growth on minimal glucose casaminoacids plates with or without tryptophan and pyridoxine is shown in Fig. 1C and D. The growth requirements of tktA ppGpp0 strain differs from that seen in tktA tktB mutant; the former does not show an absolute requirement for pyridoxine; when each of the aromatic amino acids is additionally omitted, both strains fail to show growth (Table 1A). A requirement for pyridoxine as well as aromatic amino acids has been reported in a tktA tktB double mutant (Zhao and Winkler, 1994). As phenylalanine requirement is also observed in ppGpp0 strain (Xiao et al., 1991), it will not be considered further as a synthetic phenotype of the ppGpp0tktA mutant. The pyridoxine requirement seen in the tktA tktB double mutant is not observed in ppGpp0 strains with the tktA::Tn10 or the tktA::Tp alleles (Table 1A). This difference could be due to a trace of transketolase activity in the ppGpp0tktA strain as compared with the complete absence of activity in the tktA tktB mutant (see below). We assume that a small amount of pyridoxine is sufficient to support growth owing to the catalytic use of vitamins as opposed to stoichiometric consumption of amino acids. Unlike ppGpp0tktA mutant, the ppGpp0tktB mutant strain has growth phenotypes identical to the ppGpp0 parental strain (data not shown).

Supplementing LB medium with aromatic amino acids and/or pyridoxine did not improve growth while the addition of 0.2% glucose partially improved growth (data not shown). Colony sizes are equivalent on LB glucose and in minimal glucose with all 20 amino acids and PN when incubated for the same amount of time (data not shown). It is notable that the tktA tktB mutant does not appear to require para-amino benzoic acid and related vitamins under our growth and media conditions. We do not have a good explanation for this phenotype based on our current understanding of the metabolic pathways.

The ppGpp0tktA synthetic phenotypes arise from low transketolase activity

The tktA gene is located at the 66.3 min region of the genome with six ORFs (cmtB to yggC) downstream of tktA and oriented in the same direction (Fig. 3A). Therefore, tktA could be the first gene of an operon and the synthetic phenotypes of an insertion in tktA might be due to polar effects. In order to ensure that the loss of transketolase activity caused the observed phenotype, we looked for phenotypic rescue by ectopic expression of the transketolase B isozyme (74% amino acid identity with tktA) from an IPTG-inducible promoter in the ppGpp0tktA mutant. Table 1B shows that IPTG-induced expression of a minimal tktB gene from plasmid pHR30 completely reverses the synthetic growth defect of tktA-ppGpp0 strains while the plasmid vector had no influence on the synthetic growth phenotypes. This verifies that the phenotypes conferred by tktA insertions are a consequence of lowered transketolase catalytic activity. This result implies that ppGpp deficiency could lower transketolase B activity.

Figure 3.

Schematic representation of the genomic neighborhoods of tktA and tktB genes and the DNA segments in each lacZ operon fusion. A. tktA and proximal ORF′s. B. tktB and proximal ORF′s. Open-reading frames are represented by thick filled arrows; P1 and P2 refer to promoters characterized in the intergenic region (Lacour and Landini, 2004); Fusions A, B and C refer to the transcriptional fusions described in materials and methods. Horizontal bracketed lines refer to DNA segments present in each fusion; fusions A and C have identical start points upstream of the talA coding sequence and fusions B and C have identical end points within tktB; the contents of each fusion are described in materials and methods.

The growth phenotypes of the ppGpp0tktA mutant strain are alleviated by functional SpoT

There are two genes for ppGpp synthesis in E. coli, namely relA and spoT (Cashel et al., 1996). A relA tktA (spoT+) strain does not exhibit growth impairment on LB, but shows a partial tyrosine requirement (Table 2, rows 1–3). Tyrosine and tryptophan auxotrophy is observed when the entire spoT ORF is deleted (spoT212) in the relA tktA background (Table 2, row 4). We conclude that functional SpoT is sufficient to alleviate synthetic growth phenotypes, especially the growth defect on LB. The converse experiment of deleting spoT in a tktA relA+ strain could not be performed as such a construct is inviable because excess ppGpp inhibits growth (Xiao et al., 1991). However, as described below, RelA-mediated ppGpp synthesis also contributes to the synthetic growth phenotypes.

Table 2.  Growth phenotypes correlate with the loss of cellular ppGpp synthesis.
Relevant genotypeM9 glucose minimal (48 h)
20 AA20 AA − trp20 AA − tyrLB (24 h)
  1. Growth estimates are as in Table 1. Strains in rows 1–7 are: CF1648, CF15010, CF15036, CF15007, CF15025, CF15039 and CF15005. Mutant alleles: spoT-R39A is hydrolase defective and synthetase-proficient; spoT-E319Q is synthetase defective and hydrolase-proficient.

1 Wild type++++++++++++
2 ΔrelA++++++++++++
3 ΔrelA tktA+++++++
4 ΔrelA ΔspoT tktA+±
5 ΔrelA spoT-R39A tktA+++++++
6 spoT-E319Q tktA++++++++
7 ΔrelA spoT-E319Q tktA+±

The growth phenotypes in ppGpp0tktA mutant strain reflect the overall ppGpp biosynthetic capacity of the cell

There are at least two known functions for SpoT protein, namely, ppGpp synthesis and hydrolysis (Xiao et al., 1991). We wanted to understand the SpoT function required to alleviate synthetic growth phenotypes. It is even possible that this function is spoT-dependent but ppGpp-independent, because a number of proteins have been identified that interact with SpoT. We were unable to test this by providing a weak enough source of ppGpp to allow survival of a ΔrelAΔspoT strain (Table S2, row 3). Examples of proteins that interact with SpoT are acyl-carrier protein (Battesti and Bouveret, 2006): CgtA (Wout et al., 2004; Jiang et al., 2007) and numerous small and large subunit ribosomal proteins (Butland et al., 2005). The balance of SpoT hydrolase and synthetase activities respond to variety of environmental signals (Cashel et al., 1996; Murray and Bremer, 1996), but little is known of the mechanisms coupling SpoT responses to these signals except in the case of fatty acid synthesis (Battesti and Bouveret, 2006).

We constructed a pair of single amino acid substitution alleles of SpoT designed to eliminate either the hydrolase or the synthetase activity of SpoT but otherwise minimally altering the protein. To do this, we exploited predictions from mutants and structures solved for RelSeq, the SpoT homologue from Streptococcus equisimilis (Mechold et al. 2002; Hogg et al., 2004). The residues chosen to be altered in each of the two catalytic centres were SpoT-R39 to limit hydrolase activity (HS+) and SpoT-E319 to limit synthetase (H+S-) activity. The residues were selected because their homologues in RelSeq were deduced to display maximum movement during structural changes in their catalytic pocket when ligands bind the opposing catalytic centre (Hogg et al., 2004). Growth tests in a relA mutant to characterize spoT-R39A and spoT-E319Q alleles are described in supplementary information (Table S2). The hydrolase mutation (HS+) slows growth in LB and in minimal media (Table S2) consistent with higher basal levels of ppGpp during growth. The E319Q synthetase mutation in this host entirely eliminates ppGpp synthesis because the mutant fails to grow in minimal media when the relA256 in-frame ORF deletion is present (Xiao et al., 1991).

Substituting the synthetase mutant (H+S-) allele for a complete spoT deletion confers growth requirements in a tktA relA256 background. If this strain is made RelA+, growth is normal on LB and aromatic amino acid requirements are not seen; RelA becomes the source of ppGpp (Table 2, rows 6 and 7). Introduction of the spoT-R39A allele eliminates the growth requirements of the parental tktA ppGpp0 strain (Table 2, row 5) and is not viable in a relA+ background (data not shown). Apparently, transketolase B activity can be downregulated by a single residue change in the synthetase catalytic centre of the 702-residue SpoT protein. This suggests a key role for ppGpp synthetase function and eliminates other putative regulatory functions of SpoT protein that are unaltered in the E319Q allele. The simplest interpretation of the results is that ppGpp regulates transketolase B activity in the tktA relA256 mutant. The extent of tktB activation reflects the cellular capacity to synthesize ppGpp either from RelA or SpoT.

Independent and synergistic roles of ppGpp: DksA and RpoS modulate tktB-dependent growth requirements

Finding that ppGpp is required for tktB function leads to the need to assess the roles for DksA and RpoS, two proteins whose regulatory functions are coordinated with that of ppGpp in many instances. DksA, a multicopy suppressor of DnaK (Kang and Craig, 1990), functions at the level of transcription initiation in vitro as a cofactor of ppGpp to mediate both positive and negative regulatory effects on gene expression (Perederina et al., 2004; Paul et al., 2004a,b; 2005). Studying RpoS is relevant because during entry into stationary phase the accumulation of this stationary phase sigma factor is delayed in the absence of ppGpp or dksA; during exponential growth RpoS levels increase upon gratuitous induction of ppGpp (Gentry et al., 1993; Brown et al., 2002). A requirement for ppGpp exists not only at the level of accumulation of RpoS but also for RpoS-dependent gene expression (Kvint et al., 2000).

Deleting dksA confers several amino acid requirements, but these do not include tryptophan or tyrosine (Brown et al., 2002). The same dksA allele when combined with tktA reduces, but does not eliminate growth in the absence of tyrptophan or tyrosine (Table 3, row 2). Apparently, the absence of one cofactor (DksA) only partially mimics the absence of the other (ppGpp). The results could be interpreted as independent regulation of tktB expression by dksA or potentiation of ppGpp-mediated regulation by dksA. The latter is supported by the observation that absence of DksA and ppGpp gives phenotypes only as severe as those seen in the absence of ppGpp (Table 3, rows 3 and 4).

Table 3.  Independent and synergistic effects of ppGpp, DksA and RpoS in the modulation of tktB-dependent growth requirements.
Relevant genotypesM9 glucose minimal (48 h)
20 AA + PN20 AA20 AA − trp20 AA − tyr
  1. Growth estimates are as in Table 1. Strains in rows 1–10 are: CF14309, CF14955, CF14966, CF15240, CF14241, CF14965, CF15282, CF14965 with pHR30, CF14967, and CF14967 with pHR30. Wherever indicated, IPTG is present at 0.25 mM. For plasmid bearing strains ampicillin was used at 50 μg ml−1 final concentration.

1 dksA++++++++
2 dksA tktA++++++
3 ppGpp0tktA+++
4 dksA tktA ppGpp0+++
5 rpoS++++++++++++
6 rpoS tktA+++++±
7 rpoS tktA relA++
8 rpoS tktA/pHR30 + IPTG++++++++++++
9 rpoS tktA ppGpp0++
10 rpoS tktA ppGpp0/pHR30 + IPTG++++++++

In an otherwise wild-type host, deleting rpoS does not result in amino acid or vitamin requirements; upon further inactivation of tktA, growth impairment is slight with all amino acids and PN present, similar to a tktA mutant (Table 1, rows 1 and 2; Table 3, rows 5 and 6). However, the rpoS tktA double mutant, unlike each single mutant, shows partial requirements for tryptophan, tyrosine (Table 1, row 2; Table 3, rows 5 and 6) and phenylalanine (data not shown). Adding a relA deletion (rpoS tktA relA) gives strong growth requirements for PN and amino acids (Table 3, rows 7). A similar phenotype is observed in the rpoS tktA ppGpp0 strain, making these strains phenotypically identical to the tktA tktB mutant (Table 1 A, row 4; Table 3, rows 7 and 9). We confirmed that growth requirements in mutant strains arise from reduced levels of TktB by rescuing growth through ectopic tktB expression using plasmid pHR30 (Table 3, rows 6–10). The results indicate independent regulatory roles for ppGpp and rpoS in tktB transcription (see Discussion).

As mentioned previously, DksA overexpression using multicopy plasmid can suppress some ppGpp0 phenotypes. In Table 4, rows 3 and 4 show that DksA overexpression in the ppGpp0tktA mutant restores prototrophy for tryptophan and tyrosine and that the suppression requires RpoS. The pyridoxine requirement is overcome by DksA overexpression in a ppGpp0tktA rpoS mutant (Table 4, compare rows 2 and 4). Therefore, overexpression of DksA can suppress pyridoxine and amino acid requirement in the presence of RpoS and only the pyridoxine requirement in the absence of RpoS.

Table 4.  Suppression of growth requirements by multicopy DksA and the stringent rpoB mutations.
Relevant genotypeM9 glucose minimal with (48 h)
20 AA + PN20 AA20 AA − trp20 AA − tyr
  1. Growth estimates are as in Table 1. Strains in rows 1–10 are: CF15041, CF14967, CF14946, CF14953, CF14998, CF15000, CF14959, CF14999, CF15001 and CF14960. For plasmid bearing strains, ampicillin was used at 50 μg ml−1 final concentration.

1 ppGpp0tktA/pBR322+++
2 ppGpp0tktAΔrpoS++
3 ppGpp0tktA/pJK537++++++
4 ppGpp0tktAΔrpoS/pJK537+++
5 ppGpp0tktA rpoB-A532Δ++++++++
6 ppGpp0tktA rpoB-A532ΔΔrpoS++++±
7 ppGpp0tktA rpoB-A532ΔtktB+
8 ppGpp0tktA rpoB-T563P++++++++
9 ppGpp0tktA rpoBT-563PΔrpoS++++±
10 ppGpp0tktA rpoBT-563P tktB+

Suppression of auxotrophic requirements by RNA polymerase mutations

About 60 spontaneous mutant alleles have been isolated that restore the growth of ppGpp0 strain on minimal glucose and mapped to rpoB, rpoC and rpoD genes. Some of these have been studied extensively in vitro (Cashel et al., 1996; Zhou and Jin, 1998; Barker et al., 2001; Murphy and Cashel, 2003). We chose for this study two well-known rpoB alleles that confer rifampicin resistance, rpoB-T563P and rpoB-A532Δ(alias rpoB3370 and rpoB3449 respectively) which mimic ppGpp regulatory behaviour in vivo and in vitro (Zhou and Jin, 1998). We first asked if the alleles change RpoS expression pattern in the absence of ppGpp. Figure 4 is an immunoblot using anti-RpoS antibody in ppGpp0 strains with or without the rpoB3449 and rpoB3370 alleles. The presence of the suppressor mutations elevates RpoS protein levels 20-fold over the levels observed in ppGpp0 cells in the log phase of growth. The RpoS level in the rpoB mutant strains is about fivefold higher than in the wild-type strain in log phase (data not shown).

Figure 4.

Stringent rpoB suppressor mutations increase RpoS protein levels of exponentially growing cells. Extracts were made from LB grown cells taken at different stages of growth. Extracts from cells equivalent to 0.1 A600 were used for immunoblotting with anti-RpoS antibody. Lanes 1, 4 and 7 have extracts from ppGpp0 strains CF14276; lanes 2, 5 and 8 from the rpoBT563P derivative CF14278 and lanes 3, 6 and 9 from the rpoBA532Δ derivative CF14277. Log, early stationary phase and stationary phase correspond to A600 values of 0.6–0.8, 2–2.2 and 3.5–3.6 respectively.

Table 4 shows that both rpoB alleles completely suppress the growth requirements associated with low transketolase activity in the tktA ppGpp0 host (rows 5 and 8). This is consistent with their ability to induce RpoS accumulation. However, when rpoS is deleted, growth in the absence of PN and tryptophan or tyrosine persists although the suppression is considerably weakened (Table 4, rows 6 and 9). Therefore, the suppression activity of the rpoB alleles is not entirely RpoS-dependent. We thought it was also important to ask if suppression of growth requirements by the rpoB alleles is entirely through the activation of tktB or has an alternate explanation (say, the activation of a cryptic transketolase gene). In Table 4, rows 7 and 10 show that suppression requires tktB; these results indicate that the rpoB alleles can increase TktB activity independent of RpoS and ppGpp (see below).

RpoS and ppGpp activate tktB-lacZ transcriptional fusions

The nutritional requirements of the mutant strains indicate regulation of tktB expression by ppGpp, RpoS, DksA and the stringent rpoB mutations. To find out if transcription can account for regulation of growth requirements, reporter activity of tktB-lacZ operon fusions was measured during growth in LB using a tktA+ strain.

Previous studies have identified two closely spaced promoters upstream of talA (P1 and P2 in Fig. 3B), and one within the talA ORF just upstream of tktB (Lacour and Landini, 2004; Jung et al., 2005). We constructed three transcriptional fusions (Fig. 3B) to look at activity of the promoters. Fusion A measures transcription from P1 P2 promoters upstream of talA, fusion B from the promoter reported in the talA ORF with fusion joints identical to the one described in Jung et al. (2005). Fusion C detects transcription from the entire region upstream of tktB (Fig. 3B) and extends 238 nucleotides upstream of talA.

Table S3 is a survey of reporter activities using the three fusions in wild-type strains for exponential and stationary growth phases. Fusion B is marginally active, whereas fusions A and C have measurable activities during exponential growth, which are induced seven- and ninefold, respectively, in stationary phase. Under our conditions, talA and tktB genes seem to comprise an operon. We chose fusion C for the studies reported below.

The reporter activity from fusion C is shown in Fig. 5, measuring activities in log, early stationary and stationary phase for different strains. Transcription is lowered five- to sixfold during all phases of growth in the ppGpp0 strain (Fig. 5A and C), whereas the absence of RpoS lowers the tktB activity increasingly during growth from log to stationary phase (7- to 16-fold). In ppGpp0 strain, the absence of RpoS lowers activity at least threefold further in all growth phases (compared with rpoS mutant) and the activity is virtually absent in the log phase cells (∼0.6 Miller units). The results are consistent with an independent regulation of tktB expression by ppGpp and RpoS (see Discussion).

Figure 5.

Regulation of tktB transcription – the role of ppGpp, RpoS and stringent rpoB mutations. tktB transcription was monitored during growth in LB with the talA-tktB′::lacZ fusion C. Strains used are: CF14213 and CF14241 (columns A and B); CF14214 and CF14242 (columns C and D); CF15008 and CF15023 (columns E and F); CF14277 and CF14281 (columns G and H); CF14278 and CF14280 (columns I and J). For each culture activity was measured in log phase (A600 0.5–0.6), early stationary phase (A600 1.5–2) and stationary phase (A600 2.5–3.5). β-Galactosidase specific activities are the mean of three independent experiments expressed in Miller units. In the data table, but not the bar graph, values are rounded to the nearest whole number. SD, standard deviation.

Stringent rpoB alleles increase tktB-lacZ expression

The presence of the ‘stringent’rpoB suppressor mutations T563P and A532Δ (in the ppGpp0 background) results in a large increase in tktB expression (32-fold in rpoBA532Δ and 42-fold in rpoBT563P) during exponential growth. When these strains go into stationary phase, only a modest additional increase in expression occurs (Fig. 5G and I). For both rpoB mutants the increase in tktB expression is largely RpoS-mediated (Fig. 5H and J). However, in the absence of RpoS, they have a 20-fold higher activity in log phase compared with isogenic ppGpp0 strain (compare Fig. 5D with H and J).

Hydrolase-deficient spoT-R39A allele increases tktB-lacZ expression

The activation of tktB transcription by the hydrolase-deficient spoT-R39A allele is consistent with a positive regulatory role for ppGpp. The tktB expression pattern seen when spoT synthetase is not balanced by hydrolase, is strikingly similar to that observed in the rpoB mutant strains: a large increase in exponential phase and a moderate increase thereafter. An eightfold or 44-fold increase in expression is observed in exponential phase when compared with wild-type or ppGpp0 strains (Fig. 5, compare A, C and E). The RpoS-independent tktB expression in the spoT mutant is once again similar to that seen in the rpoB mutant strains (compare F, H and J in Fig. 5) underscoring a possibility of similar regulatory mechanisms (see Discussion).

Positive regulation of tktB-lacZ expression by DksA

A dksA deletion reduces tktB transcriptional activity roughly by half during all phases of growth, compared with wild-type strain; in a ppGpp0 strain the same deletion has no further effect (data not shown). DksA overexpression does not significantly alter tktB transcription in the presence of ppGpp, but restores expression close to wild-type levels in a ppGpp0 strain. This positive effect of DksA on tktB transcription in the ppGpp0 strain is primarily RpoS-dependent but the small RpoS-independent effect is also noted (Fig. 6B). These results are consistent with the growth phenotypes observed during DksA overexpression in the absence of ppGpp and in the presence or absence of RpoS (Table 4, rows 1–4).

Figure 6.

Effect of over-production of dksA on talA-tktB′::lacZ expression. β-Galactosidase specific activities are plotted against A600 during growth in LB in the presence of plasmid pJK537 or the vector control pBR322 in wild-type (CF14213) or ppGpp0 strains (CF14314).
A. The rpoS mutant derivatives of wild-type (CF14241) or ppGpp0 strains (CF14242).
B. The activities plotted are mean of three independent experiments expressed in Miller units.


This work shows that genetic screening for synthetic lethals can be applied to define ppGpp-dependent functions. We explain the synthetic growth defects arising from the inactivation of tktA in a ppGpp0 host strain as owing to inactivation of a tktB-dependent redundant pathway. We show that ppGpp regulates tktB transcription through the modulation of RpoS activity and by RpoS-independent positive stringent regulation as well. The contribution of each pathway may depend on growth conditions.

Transketolase activity and growth in LB

It is unclear how transketolase activity affects growth in LB agar or broth. Supplementing LB with glucose, aromatic amino acids and pryridoxine only marginally improves growth (data not shown). Unlike in LB, in minimal glucose media containing all the supplements, growth of the double mutant is only slightly slower than that of an isogenic wild-type strain (Table 1A). Presence of inhibitors in LB has been suggested previously (Zhao and Winkler, 1994). Alternatively, the unique need for transketolase activity could be specific for growth on carbohydrate-poor complex peptide digest medium like LB and related to its metabolic function at the intersection of gluconeogenesis and pentose phosphate shunt. The synthetic growth defect arising from tktA and ppGpp deficiency or the absence of both transketolase isozymes can be completely reversed by ectopic expression of TktB (Table 1B), indicating that growth defects are due to general transketolase deficiency rather than from a specific function of TktA. It is important to consider this possibility because a genetic selection for mutants with reduced chromosomal negative supercoiling has uncovered a role for TktA and DksA along with H-NS, Fis and SeqA/Pgm in the maintenance of chromosomal superhelicity (Hardy and Cozzarelli, 2005).

Positive regulation of tktB transcription by ppGpp: the RpoS-dependent and RpoS-independent modes of regulation

The assay for transketolase B function based on growth requirements indicates that the cellular transketolase B activity is almost entirely dependent on the presence of ppGpp and RpoS, because growth properties of tktA tktB mutant is identical to that of tktA rpoS ppGpp0 mutant. The tktB-lacZ fusion results show that regulation by ppGpp and RpoS is almost entirely at the level of transcription. The data also indicate that there are at least two routes for activation of tktB transcription; one requires ppGpp and RpoS while the other is transcriptional activation by ppGpp independent of RpoS.

The transcriptional fusion C in tktB shows a 16-fold drop in expression in stationary phase in the rpoS mutant, consistent with a microarray study that reported 14-fold RpoS-dependent increase in tktB transcripts in stationary phase (Weber et al., 2005). Previous studies have established that ppGpp0 strains phenocopy RpoS deficiency and gratuitous induction or increase in ppGpp basal levels leads to increased RpoS protein levels (Gentry et al., 1993; Brown et al., 2002). Also, ppGpp facilitates transcription by pre-existing RpoS (Kvint et al., 2000) and its absence diminishes the ability of RpoS to compete against RpoD for the core (Jishage et al., 2002). There is growing evidence from many other studies that lead to the proposal that ppGpp helps alternative sigma factors to compete for RNAP core (Magnusson et al., 2005). Therefore, downregulation of tktB expression observed in ppGpp0 strain can result from a combination of lowered RpoS protein levels and diminished RpoS function. It is possible that low RpoS levels could be a reason why tktB was not identified in a transcriptional profile of stringent response (Durfee et al., 2008), as the response was elicited in early exponential phase cultures using serine hydroxymate which inhibits protein synthesis. The induction by elevated ppGpp alone in the absence of RpoS was probably below detection levels.

The partial growth defects observed in rpoS tktA mutant indicate low-level TktB activity (Table 3). Additional inactivation of relA or the absence of ppGpp eliminates residual activity, because growth requirements of rpoS tktA relA, rpoS tktA ppGpp0 and tktA tktB mutants are identical (Table 3). Similarly, the low lacZ reporter activity in the rpoS mutant is lowered further to barely detectable levels in the absence of ppGpp (Fig. 5). These results are consistent with a role for ppGpp that is independent RpoS in the regulation of tktB expression, and suggest a mechanism for feedback regulation in the aromatic amino acid biosynthetic pathway (Fig. 2). Aromatic amino acid starvation activates RelA, leading to ppGpp synthesis, activation of tktB transcription and aromatic amino acid biosynthesis. This leads to disappearance of the signal.

Based on existing models, the transcriptional activation mediated by ppGpp we see could be due to either direct effect of ppGpp and DksA at the promoter (Paul et al., 2005) or due to an indirect effect resulting from increased availability of free RNAP as a consequence of inhibition of rRNA transcription (Zhou and Jin, 1998; Barker et al., 2001) that facilitates competition by alternative sigma factors for the RNAP core (Jishage et al., 2002; Magnusson et al., 2005; Szalewska-Palasz et al., 2007; Costanzo et al., 2008).

Two transcription start sites were localized upstream of talA and used to deduce promoters P1 and P2 based on a −10 consensus for RpoS-dependent promoters (Lacour and Landini, 2004). This microarray study found RpoS dependence for talA, but in contrast to our results and the other microarray study cited earlier (Weber et al., 2005) tktB transcripts were not found to be RpoS-dependent. Interestingly, the P1 promoter has a sequence tgctatgcttttt followed by the +1 transcription start site, that is, an extended −10 (underlined) together with AT rich discriminator. This could fuse RpoS-dependence with activation by ppGpp. It is also possible that the two promoters respond differentially, one to RpoS and another to ppGpp perhaps through sigma-70 or an alternative sigma factor.

RpoS-mediated positive and negative regulation of tktB and tktA, respectively, was reported; it was suggested that isozymes TktA and TktB may play their main roles in exponential and stationary phase respectively (Jung et al., 2005). However, using an identical construction (fusion B, Fig. 3) we are unable to observe activity (Table S2) or observe RpoS regulation (data not shown); the reason for the disparity is not apparent.

Overexpression of DksA and activation of tktB transcription

Suppression by overexpression of DksA is a feature found for many ppGpp0 phenotypes including the restoration of RpoS induction (Brown et al., 2002; Magnusson et al., 2007). Consistent with an effect mediated through RpoS, restoration of tktB expression by DksA overexpression requires functional RpoS (Fig. 6A and B). However, a small positive effect can be observed in the absence of RpoS and ppGpp; this is supported by the suppression of vitamin requirement observed in growth assays (Fig. 6B; Table 4). The ability of DksA to substitute for a complete absence of ppGpp argues that DksA and ppGpp are not cofactors (like cAMP and CRP).

Stringent rpoB mutations functionally mimic constitutive high levels of ppGpp

The spoT-R39A mutation is deduced to change the ppGpp hydrolase-synthetase balance in favour of synthesis and increase ppGpp basal levels (Table S2). The spoT-R39A mutant and the stringent rpoB mutant strains used in this study show essentially identical tktB transcriptional regulation and TktB activity in vivo as deduced from lacZ fusion data and growth requirements of these strains (Tables 2 and 4; Fig. 5). The phenotypic similarities between the strains persist in the absence of RpoS. The results suggest that an increase in intracellular ppGpp level and presumably polymerase conformation changes can have similar functional consequences for positive transcriptional regulation mediated by ppGpp. A previous study with the same polymerase mutants showed that they mimic the negative transcriptional regulation mediated by ppGpp (Zhou and Jin, 1998). The rpoB mutants increase RpoS protein levels (Fig. 4) and the same is observed for the spoT-R39A hydrolase mutant (data not shown). Together, the results provide further evidence for the passive regulatory model and extend it for ppGpp-mediated RpoS regulation. The spoT-R39A and stringent rpoB mutants soften the wild-type ‘stair step’ like increase in tktB expression with growth phase. If high ppGpp levels slow the metabolic turnover of RpoS during growth in LB like it does during phosphate starvation through iraP (Bougdour and Gottesman, 2007), similar effects might be expected.

This study shows that tktB is subject to RpoS-mediated regulation through ppGpp, excess DksA in the absence of ppGpp, and ‘stringent’rpoB mutations. In addition, transcription activation is also seen independent of RpoS (Fig. 7): (i) under conditions that increase cellular ppGpp levels, (ii) by ‘stringent’rpoB mutations in the absence of ppGpp and (iii) DksA overexpression in the absence of ppGpp. It is possible that in each case transcriptional activation is mediated through a similar conformational change in RNAP. In vitro transcription studies are warranted to find out if the effects are direct or involve additional factors.

Figure 7.

A model for the transcriptional regulation observed in this study for the talA-tktB operon. Two known promoters P1 and P2 upstream of talA are represented by a vertical line. Horizonal bars indicate the talA-tktB operon. Excess DksA refers to overexpression of the protein using plasmid pJK537. Continuous line arrows refer to activation signals; broken line arrows refer to activation signals observed only in the absence of ppGpp.

Experimental procedures

Media and growth conditions

Table S1 list the E. coli strain derivatives of MG1655 and plasmids used in this study. Cultures were grown in LB broth in rotary shaker flasks at 37°C. The media used are described by Miller (1972) but modified as follows: LB contains 0.5% NaCl and M9 glucose minimal contains 15 μM thiamine and may be supplemented with either 0.4% casiamino acids, all 20 amino acids (each at 40 μg ml−1), 19 amino acids lacking either phenylalanine, tryptophan or tyrosine and pyridoxine (10 μM). Final concentration of antibiotics was: ampicillin (100 μg ml−1 in LB and 50 μg ml−1 in minimal media), tetracycline (20 μg ml−1), kanamycin (30 μg ml−1) and trimethoprim (100 μg ml−1). Quantitative estimates of growth at 37°C are based on colony diameters: ± is 0.5 mm or less; + is 1–1.5 mm; ++ is 2–2.5 mm; +++ is 3 mm or more. In general, colony sizes are scored on LB plates and on minimal media after incubating for 24 h and after 48 h respectively.

Bacterial strains, plasmids and genetic procedures

Plasmid pMA2 is a temperature-sensitive, pSC101 replicon (obtained from the cloning vector collection at NIG, Mishima, Japan). pMA2 was digested with SphI and re-circularized to obtain plasmid pHR13, and was used to clone the spoT gene from plasmid pHX41 (Gentry et al., 1996) as a 2.1 kb SphI fragment yielding pHR14. Plasmid pJK537 has been described in Kang and Craig (1990). Plasmid pHR30 was constructed using pQE80L (Qiagen vector) for IPTG-inducible TktB expression. A 2041 bp fragment encompassing the entire tktB ORF was PCR-amplified from MG1655 genome using Pfu-polymerase and primers 5′-atgaattccagccacggagt-3′ (upstream) and 5′-ttggatccggcaatcacc atca-3′ (downstream). EcoRI and BamHI sites in the primers (underlined) were used to clone the fragment in the EcoRI/BamHI sites of pQE80L.

Phage P1(vir) transductions were performed by standard procedures (Miller, 1972). When constructing strains with the tktA rpoS, ppGpp0tktA and ppGpp0rpoS tktA genotypes, the last allele to be introduced is tktA and transductants were selected and maintained on LB plates containing 0.2% glucose which improved the growth of these strains. Plasmid DNA was extracted using Qiagen kits, transformations and recombinant DNA procedures were performed generally as described in Sambrook et al. (1989). Recombineering was performed as described in Yu et al. (2000), using the linear DNA transformation protocol to engineer the spoT– synthetase and hydrolase mutations.

Construction of transcriptional-lacZ fusions

The talA-tktB′::lacZ fusion (fusion C) was constructed using a 1331 bp fragment retrieved from the chromosome of strain MG1655. This fragment has 238 bp of non-coding region upstream of talA, the entire talA structural gene, the talA-tktB intergenic region and 124 nucleotides of tktB coding sequence; it was recombined into a linear pBR322-derived fragment using a protocol described in Court et al. (2002). The primers for generating the linear fragment for recombination with the chromosomal sequence (italics in the primer sequence) were 5′-cagggtaataatgtgcgccacgttgtgggcaggggaattcgcgt ttcggtgatgacggtg-3′ (bottom strand) and 5′-gggcatggctgatattgccgaagtgctgtggaacggatccga aagggcctcgtgatacgc-3′ (top strand). The chromosomal fragment was then subcloned into the operon fusion vector pRS415 using restriction enzyme EcoRI and BamHI (underlined in the primers) and recombined first into λRS45 and then into the λatt site on the chromosome as described in Simons et al. (1987). The talA′::lacZ fusion (fusion A) was constructed using a 319 bp fragment containing the same 5′ end point as fusion C and 81 bp of the talA coding sequence and obtained by PCR using primers 5′-atgaattccccctgccca caacg-3′ (top strand) and 5′-atggatccgatgataatggcgaatggactc-3′. The ‘talA-tktB′::lacZ fusion (fusion B) contains a 344 bp fragment with 200 bp of the 3′-coding sequence of talA, 19 bp of intergenic sequence and 124 bp of tktB coding sequence and obtained using the primers 5′-atgaattcctgcaggaaaaagtttcgcc-3′ (top strand) and 5′-atggatcctcgttccacagcac ttc-3′ (bottom strand). These fusions were also transferred to the λatt site as described for fusion C.

Transposon mutagenesis and sequencing of chromosomal transposon-insertion sites

Random mutagenesis was carried out using the EZ-Tn5TM <DHF R-1> transposon kit from Epicentre (Madison, WI) and the insertion sites were characterized by sequencing.

β-Galactosidase assays

Cultures were grown after a 1:500 dilution of an overnight culture and the lacZ fusions were assayed for β-galactosidase activity as described in Miller (1972) with activities reported in Miller units.

Gel electrophoresis and Western blotting for RpoS

Lysates were prepared after precipitation with cold trichloroacetic acid using final concentrations of 5% and 10% for exponential and stationary phase cells respectively. After centrifugation, pellets were washed once with 0.5 ml of ice-cold 80% acetone, air dried and resuspended in SDS-gel loading buffer. Equal quantities of proteins were separated on precast SDS-12% PAGE acrylamide gels (Invitrogen) and transferred to PVDF membranes. The membranes were incubated with anti-RpoS antibody (Neoclone, Madison, WI) at 1:1000 dilution and the blots developed using horseradish peroxide-conjugated goat anti-rabbit antibody by the enhanced chemiluminescence protocol (GE health sciences).


We would like to thank Manjula Reddy for discussions that inspired this study and for sharing plasmid pMA2 and Dick D’Ari making suggestions for the manuscript. This work is supported by the intramural research programme of NICHD/NIH.