INTEGRATION OF BACTERIAL NUTRITIONAL RESPONSES
, Head, Section on Molecular Regulation
Rajendran Harinarayanan, PhD, Visiting Fellow
Christophe Penno, PhD, Visiting Fellow
Katarzyna Potrykus, PhD, Visiting Fellow
William Whalen, PhD, Research Associate
Helen Murphy, MS, Microbiologist
We wish to understand how nutrient availability coordinates bacterial genomic expression. Within this broad goal, we continue to focus on the roles of two regulatory analogues of GTP and GDP that are modified with ribose 3¢ pyrophosphates and collectively termed ppGpp. Remarkably, nutrient limitation elevates levels of the ppGpp nucleotides regardless of whether the organism is starved for sources of amino acids, phosphate, nitrogen, carbon, or iron. Only some of a bacterium’s sensing mechanisms are currently understood. Signaling roles are assigned to ppGpp because the elimination of ppGpp during starvation can abolish regulation and lead to difficulties in surviving starvation. Conversely, elevating ppGpp without starvation can mimic aspects of starvation. Recent work has revealed that transcription regulation by ppGpp requires the cofactor DksA, a structural homologue of the GreA and GreB elongation transcription factors. This year, we used genetic screens with multicopy gene fragment libraries to look for genes that complement ppGpp-deficiency (ppGpp0) phenotypes. Our approach revealed new candidates and possibly new mechanisms in the (p)ppGpp-DksA-GreA–regulatory network. Motility regulation studies unexpectedly disclosed involvement of GreA in this network. We are continuing to work on intermediary metabolism mutants to learn why they are (p)ppGpp0 synthetic lethals.
GreA, GreB, and DksA: similarities in structure can match similarities in phenotypes
The GreA, GreB, and DksA proteins show striking similarities that presumably result in their ability to bind to RNA polymerase (RNAP). Each protein has a small globular domain as well as a rigid coiled-coil finger, with a pair of acidic residues located at the finger tip. Structure and crosslinking studies show that the fingers are inserted in the secondary (substrate) channel of RNAP such that the finger tips nearly touch the catalytic center. The remaining residues bind to the rim of the secondary channel when the fingers are either missing or present. During elongation, the functions of GreA and GreB depend on an ability of the acidic residues to trigger an intrinsic RNase of RNAP. This activity reverses arrest of transcription by cutting backtracked nascent RNA to restore the proximity of the 3¢ hydroxyl end of RNA to the catalytic center, thereby allowing polymerization to resume. Despite similarities in binding to the secondary channel, DksA is not known to share other biochemical attributes of GreA/B activities at physiological concentrations. Conversely, GreA/B has no known cofactor-like interactions with ppGpp and fails to cleave ppGpp, which is a potential substrate with a diesterified 3¢ ribose hydroxyl. Nevertheless, other researchers have hypothesized that the two acidic residues in the DksA finger tip anchor ppGpp near the catalytic center by mutual coordination of Mg2+, thus stabilizing ppGpp binding. A single biological phenotype known for GreA led to GreA’s discovery. Multicopy greA suppresses the temperature sensitivity of a unique RNA polymerase mutant. GreB was found by sequence homology to GreA. DksA was originally discovered by its ability, in multicopy, to suppress the temperature sensitivity of a dnaK deletion—hence its name dnaK suppressor.
We found that GreA and GreB share the above-mentioned ability of DksA, in multicopy, to reverse the temperature sensitivity of a dnaK deletion mutant. Given that DksA does not display effects on transcription elongation, we sought to determine whether the RNAse-activating activities of GreA and GreB are required for their dks activity. To that end, we tested mutants of both acidic residues for GreA and GreB that are known from in vitro work to abolish activation of the intrinsic RNAP RNase. We call the mutants GreA* and GreB*. We used an analogous D* mutant as a test for the ppGpp-anchoring hypothesis. All three mutant proteins in multicopy suppress ΔdnaK hosts in much the same manner as their wild-type counterparts; that is, all three * mutants suppress the temperature sensitivity of the ΔdnaK strain as do their wild-type parent proteins. We concluded that suppression reflects a shared phenotype that does not appear to involve RNase activation for GreA and B and does not involve a ppGpp-anchoring mechanism.
Despite structural similarities, changing the balance among GreA, GreB, and DksA provokes different phenotypes
Last year, we found that adjusting the levels of the three transcription factors altered the inhibition of cell growth—the phenotype of ppGpp overproduction, which is the most sensitive phenotype of which we are aware. Growth sensitivity to ppGpp is modestly enhanced by deleting DksA and greatly enhanced by deleting both GreB and DksA. The outcome is the opposite of what would be expected if growth inhibition by ppGpp reflected negative regulation of stable RNA synthesis with DksA functioning as a cofactor. GreB levels weakly counterbalance ppGpp sensitivity but not nearly as much as DksA. No other single mutations or combination of double mutations alters growth sensitivity to ppGpp. Last year, we also provided evidence that the ratio of GreA to DksA—not the absolute amount of GreA or DksA—was the effector. We also reported that the multiple amino acid–requirement phenotype of ppGpp0 cells can be reversed by multicopy greA in a ppGpp0 ΔDksA strain. We found growth on minimal medium to be restored by the presence of multicopy greA, greA*, and greB* but not by the presence of multicopy greB. Restated, the acid residue mutants of both GreA and GreB are active suppressors, but the wild-type GreB is inactive.
This year, we extended our studies to find similar responses by looking at growth-inhibitory effects of GreA overexpression in ppGpp+ strains. A disk assay revealed that deleting dksA, but not greB, greatly sensitized growth to excess GreA. We found that the induced GreA* mutant inhibited growth even more (10,000-fold) when ppGpp was present. These features are reversed in a ppGpp0 ΔdksA strain; in this case, growth inhibition is not only less sensitive to multicopy greA and greA*, but growth is promoted in the absence of several amino acids. How the dksA deletion makes a ppGpp0 strain permissive for the toxicity of GreA or GreA* and how suppression of amino acid requirements occurs are two important unanswered questions.
Another physiological feature of ppGpp0 strains is a severe motility defect, seen as inhibition of bacterial growth on the surface of low-percentage agar plates. Given that the ppGpp0 motility defect can be completely reversed by overproducing DksA, ppGpp and DksA do not act as equal cofactors but rather as a pathway. Similar to the absence of ppGpp, a deletion of greA in a ppGpp+ strain also generates a motility defect. However, in this case, an excess of DksA cannot reverse the defect. The greA deletion therefore has a dominant negative effect on motility. Deleting GreB seems to have little effect on motility and does not compensate for motility defects of ppGpp0. These phenotypes strongly suggest that no single mechanism can explain the differential regulatory interactions among ppGpp, DksA, and GreA.
Multicopy library screens reveal new genes involved in the ppGpp-GreA-DksA–regulatory network
We screened a multicopy plasmid library of E. coli genes for the existence of genes that restore growth on minimal glucose media of a ppGpp0 ΔdksA host strain. The number of tested transformants represented a 10-fold genomic redundancy. The initial screening yielded 34 independent sequenced fragments, of which about two-thirds contained greA (an expected positive control). The 11 remaining isolates contained six different gene fragments. None of the genes on the six fragments had the coiled-coil–like finger structures that we thought might be present. We screened all six by Western blots to see if their presence led to elevated GreA. Only one gene fragment, represented twice, elevated the expression of GreA. In this case, suppression activity could have occurred through GreA expression given that multicopy greA itself suppresses. The GreA-inducing plasmid carries yhdW and yhdX genes, which both encode cryptic ABC transporter genes. A cloned minimal yhdW gene is sufficient to mimic the phenotype. Surprisingly, suppression activity by the yhdW yhdX plasmid was not abolished with a ΔgreA host. The required baroque explanation holds that the salient activity that suppresses amino acid requirements also induces GreA but that elevated GreA is not necessary for suppression.
A frequent isolate (six copies) is a plasmid carrying the glyQ region. Its role seems confirmed with a cloned minimal glyQ gene plasmid. The GlyQ protein is the alpha-subunit of glycyl tRNA synthetase, which, together with the beta subunit, represents the sole route to aminoacylate tRNAgly. Biochemical studies have demonstrated that the GlyQ protein alone is capable of forming the glycine-adenylate precursor to charging tRNA as well as a 5¢-5¢adenosine tetraphosphate (AppppA). Several tRNA synthetases share the latter synthetic activity. The literature describes a single glyQ mutant (Nishimura et al., Genes Cells 1997;2:401) whose physiological phenotype is a defect in coupling cell division to initiation of DNA synthesis. The biochemical phenotype is a 15-fold elevation in AppppA. According to the report, the authors also mapped a second mutation, which yields a phenocopy of the GlyQ mutant. The mutation is an insertion in the apaH gene, encoding the sole AppppA hydrolase. The apaH mutant elevated AppppA 100-fold. Multicopy wild-type glyQ, such as is present in our library isolates, suppressed the glyQ mutant. The authors found that induction of ApaH rendered AppppA undetectable and restored normal phenotypes for both the glyQ and ΔapaH mutants. A plausible inference is that our glyQ library isolate lowers AppppA levels and somehow contributes to the suppression of the multiple amino acid requirements. We are now constructing tools to explore this inference.
Characterization of ppGpp0 synthetic lethal mutants in intermediary metabolism genes
This year, we continued to characterize the ppGpp0 synthetic lethal insertion alleles: tktA, aceE, and proC. All three occur in genes involved in intermediary metabolism. The studies on tktA and aceE revealed why each yields extremely poor LB (Luria-Bertani medium) growth in ppGpp0 strains. An explanation for the behavior of proC remains a puzzle.
The tktA gene encodes the major transketolase and tktB the sole minor transketolase. A ΔtktA ΔtktB mutant yields a complete transketolase deficiency, resulting in an inability to form erythrose-4-phosphate, an important intermediate in the pentose phosphate shunt for biosynthesis of aromatic amino acids and pyridoxine-derived vitamins. A comparison of the tktA tktB double mutant with the tktA ppGpp0 strain revealed equivalent requirements for phe, trp, and tyr, whereas pyridoxine was not required in the tktA ppGpp0 strain. Apparently, the synthetic lethal situation is only marginally less severe than a complete transketolase deficiency. We made a series of tktB::lacZ transcriptional fusions to localize the tktB promoter. Measurements of lacZ reporter activities revealed that tktB expression was almost completely abolished in a ppGpp0 strain. The reason for the strong ppGpp dependence is that tktB expression requires expression of rpoS. RpoS is an alternative sigma factor responsible for expression of genes specifically in stationary phase. Several years ago, our laboratory discovered that ppGpp is required for rpoS expression as cells enter stationary phase. Furthermore, deletion of rpoS in a ppGpp+ background resulted in an almost complete loss of tktB::lacZ activity. Thus, the regulatory pathway relating ppGpp to tktB expression is ppGpp®rpoS®tktB. We confirmed this pathway by using RNAP mutants (rpoB A532Δ and rpoB T563P) that suppress multiple amino acid requirements of a ppGpp0 strain and many other ppGpp-dependent phenotypes. The mutants restore tktB::lacZ reporter activity in ppGpp0 ΔtktA strains and induce RpoS protein, as seen with Western blots. Last year, we described single-residue mis-sense mutants of the SpoT enzyme that are defective in either ppGpp synthetase or hydrolase activities and exploited them to reveal that the synthetase was the key activity for initiating the ppGpp-to-tktB regulatory cascade, with the regulatory signal thought to be ppGpp itself complexed with RNAP and DksA. Before we understood the tktB contribution, we asked if the synthetic lethality of the tktA mutant could be attributed to a general deficiency of pentose phosphate intermediates. We added gluconate and attempted to block it from feeding the glycolytic pathway by deleting glucose-6-phosphate dehydrogenase (zwf). The result was the surprising discovery that the zwf mutant itself is a synthetic lethal in a ppGpp0 background. We are still working on the mechanism that leads to that synthetic lethality.
The aceE gene encodes a subunit of pyruvate dehydrogenase. The mechanisms underlying ppGpp0 synthetic lethality of aceE mutants and tktA are similar. A redundant RpoS-dependent pathway requires ppGpp because the latter is needed to express rpoS. The metabolic deficiency in aceE ppGpp0 cells is acetyl-CoA. Supplementing LB medium with acetate restores viability. The nonviability of mutant aceE ppGpp0 cells is thought to result from a lack of acetyl-CoA, even though acetate is the direct metabolic product of pyruvate dehydrogenase and the immediate precursor of acetyl-CoA. The reasoning is that the addition of acetate restores the viability of aceE ppGpp0 cells while the growth-promoting effect of acetate depends on the presence of genes that encode enzymes required for the formation of acetyl-CoA from acetate. The poxB gene encodes the key enzyme in the redundant pathway for forming acetyl CoA when pyruvate dehydrogenase is inactive. Again, we used poxB::lacZ fusions and rpoS deletions to show that the ppGpp dependence of poxB expression is exerted through rpoS.
We do not yet understand why proC (or proA or proB) null alleles are ppGpp0 synthetic lethals. Growth tests reveal that multicopy dksA (but not greA or greB) plasmids allow growth, suggesting another cofactor role for DksA. The poor LB growth phenotype at 42°C is reversed with the addition of high levels of proline (200 µg/ml). The literature contains several descriptions of the chaperone effects of high proline concentrations on protein folding. Using standard carbonylation assays to detect protein oxidative changes resulting from misfolding, we have been unable to document quantitatively the misfolding of proteins in ppGpp0 strains with or without pro mutations. We thought we might gain evidence for genes involved in folding through the use of the multicopy library described above. In this case, the selection is for genes in multicopy that suppress the proC synthetic lethal phenotype. The library did yield the expected positive control plasmids expressing proC and dksA. We have isolated several candidates, but none is a chaperone gene while most encode putative or known membrane proteins. One of the latter is the apt gene, which catalyzes adenine uptake by a ribosylpyrophosphate transferase reaction. In many bacteria other than E. coli, apt is located adjacent to genes encoding ppGpp synthetases. Cloning has shown that a minimal apt clone suppresses just as well as the apt plasmid isolate. Given that suppression activity is abolished by an apt active site mis-sense mutant, we believe that apt is the relevant gene, thus suggesting that adenine transport is important. The addition of adenine or adenosine to LB also restores growth of the apt+ proC ppGpp0 strain. We found that adenine transport effects do not extend to other purines. No suppression occurs when either media are supplemented with guanine or gpt is present in multicopy. We are interested in this puzzle because of an intriguing report that the addition of AMP or adenosine to cells entering stationary phase with low ppGpp induces accumulation of an uncharged tRNA species that acts as a ribozyme with RNase activity toward a regulatory RNA (Wang et al., Microbiol 2006;152:3467).
Potrykus K, Vinella D, Murphy H, Szalewska-Palasz A, D’Ari R, Cashel M. Antagonistic regulation of Escherichia coli ribosomal RNA rrnB P1 promoter activity by GreA and DksA. J Biol Chem 2006;81:15238-48.
COLLABORATOR
Richard D’Ari, PhD, Institut Jacques Monod, Centre National de la Recherche Scientifique, Université Paris 7, Paris, France
Christophe Herman, PhD, Baylor College of Medicine, Houston, TX
Daniel Vinella, PhD, Institut Jacques Monod, Centre National de la Recherche Scientifique, Université Paris 7, Paris, France
For further information, contact mcashel@nih.gov.
