Skip Navigation

Home > Section on Molecular Regulation

Global Regulation of Gene Expression by ppGpp

Michael Cashel, MD, PhD
  • Michael Cashel, MD, PhD, Head, Section on Molecular Regulation
  • Tamara James, PhD, Postdoctoral Intramural Research Training Award Fellow
  • Nathan Thomas, MS, Postbaccalaureate Training Award Fellow

There is a growing interest in unusual structural analogs of common purine ribonucleotides with a ribosyl 3′-pyrophosphate occurring together with GMP, GDP, GTP, AMP, ADP, or ATP. The best-studied family members are ppGpp and pppGpp, which have been investigated at the NIH during the past 45 years. The (p)ppGpp nucleotides signal nutritional, environmental, and metabolic stress in bacteria, archea, and plants by sensing subtle changes in central metabolism. The best-known among many mechanisms is synthesis activation when tRNA aminoacylation rates limit protein synthesis, which provides a sensitive sensor for metabolic availability of each amino acid addition to nascent proteins, whether limited by de novo synthesis, uptake, or catabolism. Hydrolysis of (p)ppGpp can also be inhibited when acylated carrier protein (ACP) limits acyl-donors for lipid synthesis, a circuit that provides a sensor of steady-state deficiencies in carbohydrate, energy, and membrane metabolism. We explored the possibility that pGpp, pApp, ppApp, and pppApp can provide novel types of regulation. Unique hydrolases with structures similar to bacterial (p)ppGpp hydrolases are found in flies, worms, and humans. We discovered that the substrate specificity of animal-kingdom hydrolases uniquely extends beyond (p)ppGpp to include ppApp and pppApp. Regulation is suggested because others have reported that deletion of the fly hydrolase alters development in addition to starvation responses. Our early research progress is documented below.

Regulation by (p)ppGpp is known be global, affecting about one third of E. coli gene transcripts and enzymatic reactions. In Gram-negative bacteria, (p)ppGpp interacts with the RNA polymerase (RNAP) itself to alter promoter recognition. The protein DksA can act with (p)ppGpp for positive or negative transcription regulation. Last year, we reported a site on RNA polymerase that binds to both pppGpp and ppGpp. Despite sharing a common binding site, ppGpp was a more potent regulator than pppGpp for five distinct regulatory phenomena.

The research is important because simple chemical signals can govern the fundamental ability of cells to adapt gene expression to stress. Thus, a therapeutic importance for (p)ppGpp has emerged. Most bacterial pathogens survive the stress of host defense mechanisms directly or indirectly by elaborating (p)ppGpp. It is also now appreciated that (p)ppGpp is a key factor for virulence as well as for pathogenic quiescence: (p)ppGpp is required for the generation of dormant antibiotic-resistant “persistence” of many bacterial pathogens, including M. tuberculosis. The promising antibiotic Relacin targets (p)ppGpp synthetase (Gad Glaser, Hebrew University of Jerusalem), which has already been useful as a tool for dissecting stages of pathogenesis. It is also noteworthy that (p)ppGpp–deficient attenuated bacterial vaccines have proven effective in models with mice and chickens.

To investigate regulation beyond (p)ppGpp and RNAP, this past year our research took three directions. The first explores determinants of substrate specificity for the animal kingdom ”Mesh” hydrolase, which includes (p)ppApp as well as to (p)ppGpp. Second, we want to devise a defined test system in E. coli to determine whether different regulators really do regulate differently, and we have initiated studies with Mesh and pGpp. Third, we have pursued our finding that the omega (ω) subunit is part of the (p)ppGpp RNAP binding site (1). Genetic associations now reveal that (p)ppGpp is surprisingly intertwined with chaperone functions, which, even more surprisingly, appear to be shared with RNAP accessory proteins.

Structural approaches to understand the promiscuous substrate specificity of Mesh hydrolase

We intend to ultimately devise a system that would enable one to test each putative regulator for different functions in E. coli strains. A valuable tool to learn of nucleotide-specific regulatory effects would be the availability of a hydrolase capable of degrading ppApp and pppApp but not ppGpp and pppGpp. We previously discovered hydrolases with the reverse specificity: (p)ppGpp but not (p)ppApp. A streptococcal enzyme (Relseq) is the gold standard for this property because we know its crystal structure. The animal-kingdom Mesh hydrolase degrades all four substrates. Given that Mesh-deletion mutants in flies give developmental phenotypes that are starvation-dependent, we wondered whether ppApp and pppApp function as newly discovered regulators. The biological persistence of this array of related analogs suggests there is evolutionary pressure for their continued existence, which implies they have different functions. The startling discovery of the ability of Mesh to degrade (p)ppApp was made by our collaborator Katarzyna Potrykus. Our collaboration with Potrykus continues for further studies. The crystal structures of both fly, human Mesh, and bacterial RelSeq hydrolases are available together with abundant bioinformatic sequence data that reveal evolutionary conservation. Tamara James used this information to model differences between the catalytic pocket of each enzymes to predict residues that could discriminate between the adenine or guanine purine rings of (p)pp(A,G)pp. Tamara James and Nathan Thomas constructed missense mutants, overexpressed and purified them, and measured their specific activities towards ppApp and ppGpp qualitatively by thin layer chromatography. The activity of Relseq towards (p)ppGpp is the standard for normalizing activity comparisons. Two rounds of analysis now have been completed (about six alleles in the first round followed nearly as many in a second round that incorporated what was learned from the first into double alleles). The results showed that two alleles define one region of the catalytic pocket as guanine-specific and two additional alleles define another region as adenine-specific. Future experiments will provide quantitative measurements of catalytic constants for both classes of substrates. Qualitatively, it is already evident that mutant Mesh is not yet completely (p)ppApp–specific, and additional rounds of Mesh mutagenesis are necessary to achieve this goal. Nevertheless, the work appears indicate that the catalytic pocket can be modified to achieve the very strong specificities.

Genetic approaches suggest that Mesh uniquely hydrolyzes an unknown substrate in E. coli essential for growth.

A Mesh gene cloned in an inducible plasmid allows us to determine whether it possesses a growth-altering activity that differs from those of bacterial hydrolases known to degrade only (p)ppGpp but not (p)ppApp. The approach is powerful because the comparison can be made without knowing the full range of possible substrates of both hydrolases. Cellular hosts for the Mesh plasmid are chosen from our existing collection of strains that differ with respect to presence, absence, or accumulation of (p)ppGpp resulting from alterations in synthetase, hydrolase, or other factors. We know that a complete deficiency of (p)ppGpp, designated (p)ppGpp0, is fully viable in rich media when a vector control is present. In contrast, the Mesh plasmid is lethal in ppGpp0 strains with or without induction. Mesh can be introduced into wild-type strains that maintain low basal levels of (p)ppGpp, but small colonies form that remain viable without induction. However, induction of Mesh protein in wild-type cells is lethal. Impaired growth frequently accompanies overexpression for many proteins and is usually thought to be attributable to nonspecific protein toxicity. Given that cloning Mesh in (p)ppGpp0 strains is lethal without induction, this explanation must be modified to include the possibility that either hypersensitivity to toxic effects occurs in the absence of (p)ppGpp or that small amounts of (p)ppGpp protect. To distinguish between the two explanations, we used a host strain with a weak constitutive synthetase but without bacterial hydrolase present and that gives an eight-fold elevated basal level of (p)ppGpp. The high basal level limits balanced growth and only tiny colonies appear on rich media after 24 hrs. The added presence of Mesh plasmid in these strains restores growth of the tiny colonies to normal, completely reversing the small-colony phenotype of wild-type cells. However, induction of Mesh is again lethal in the high basal-level strain. One explanation for large colony growth is that a slight elevation of ppGpp is degraded by Mesh, which titrates out a Mesh hydrolytic activity towards an unknown substrate, “X3′-pp,” presumed necessary for viability. This observation strongly argues that the Mesh protein toxicity is a consequence of its hydrolytic activity. Future experiments will aim to determine the identity of the presumably essential substrate. The substrate is predicted to be a purine ribonucleotide with pyrophosphate esterified to its ribosyl 3′-hydroxyl or a compound with this residue. An alternative explanation to test is that even basal levels of (p)ppGpp of wild-type strains are sufficient to limit expression of Mesh; we believe this is unlikely because this scenario has not been observed previously in our studies with expression of many other cloned proteins.

Studies of regulation by pGpp vs. ppGpp or pppGpp

We contributed to characterizing a new regulator (pGpp) found in Gram-positive bacteria in the lab of Jose Lemos. This opens the search for regulation specific to pGpp that is not shared by (p)ppGpp. So far, we found in E.coli that potent regulation by ppGpp is invariably shared to a lesser extent by pppGpp, with no examples of regulation unique to one or the other analog. RelQ is the enzyme that makes pGpp as well as (p)ppGpp in Enterococcus faecalis. The RelQ synthetase is unusual in that it lacks a hydrolase domain and its synthetase activity is redundant with RelEfa. RelEfa is a typical, large bacterial (p)ppGpp enzyme with both synthetase and hydrolase activities, like RelSeq. Deletion of RelEfa yields a cell completely lacking hydrolase while retaining the weak RelQ constitutive (p)(p)pGpp synthetase. Deletion of RelQ generates a cell with inducible (p)ppGpp but no pGpp. The Lemos lab showed that the absence of pGpp weakens the starvation response and pathogenicity, but the specific mechanism is so far elusive. A much stronger inhibition by pGpp (vs. ppGpp) was found assaying enzymes for GTP synthesis and 6-OH purine transport. We compared the ability of pGpp with those of ppGpp and pppGpp to inhibit initiation of E. coli ribosomal RNA transcription. Only a very weak inhibitory activity was found for pGpp, which ranks at the bottom of the potency hierarchy: ppGpp>pppGpp>>pGpp. So far, all regulatory activities found for pGpp are also shared to some extent with ppGpp or pppGpp. This is not unexpected, given that searches have explored only examples where (p)ppGpp is active.

(p)ppGpp regulation affects chaperone phenotypes, and chaperone activities are probably shared among the RNAP accessory proteins dksA greA, greB, and traR.

We applied genetics to resolve a dilemma raised by our finding that (p)ppGpp binds to a RNAP site that includes portions of the omega (ω) subunit (1). The dilemma arises because of our previous (1991) observation that cells deleted for ω do not have a phenotype. This is not consistent a regulatory function for (p)ppGpp, because many phenotypesl are found in (p)ppGpp0 strains when (p)ppGpp is absent. There are clues to the elusive ω function in the literature. Our strains deleted for ω have been intensively studied by others, who found that sizeable portions of immature RNAP are bound to the GroEL chaperone, known to be needed to fold large phage-head proteins. Moreover, ω-like proteins are biologically ubiquitous and judged, by weak evidence, to also be needed for multisubunit RNAP maturation. We began with a premise that is questionable, given that bacterial ω appears to have no function. It argues that the function of ω for RNAP maturation is so important that other chaperones can completely take over when ω is missing. The functional overlaps between ω, chaperones, RNAP accessory proteins, and (p)ppGpp are reinforced by a protein, called dksA, which often acts in concert with (p)ppGpp to regulate transcription and is named after its ability to reverse a phenotype due to a dnaK chaperone deletion (DNA K suppressor). The classical dnaK and dnaJ heat-shock chaperone phenotypes are partial temperature sensitivity. A matrix of all the possible combinations was studied (single, double, and triple) deletions of ω, relA (ppGpp), and each of four chaperones: dnaK, dnaJ, tig, or clpB. Chaperone choices were not random: dnaJ and dnaK were chosen because they are suppressed by dksA; trigger factor (tig) because a dnaK tig double mutant is a synthetic lethal suppressed by GroEL; ClpB because it has folding and proteolytic activity.

The work led to the discovery that the relA/dnaK (or relA/dnaJ) combination lowered survival temperatures from 37° to 24° ± ω and this property of dnaK/J was not shared with tig or clpB. This appears to be a striking new phenotype for relA, consistent perhaps with its enhanced ribosomal content saturating a limited chaperone capacity. This associates (p)ppGpp with dnaK/J and as just mentioned, dnaK/J are also linked to dksA by suppression. The next discovery required triple mutants: cells deleted for ω failed to grow at 32° whereas +ω grew at high temperatures, i.e., 42°. A complete triple mutant was required for temperature sensitivity; all double mutant combinations were insensitive. The dnaK/J deletions do not substitute. The finding links ω with (p)ppGpp, tig, clpB, and indirectly with GroE, as mentioned.

To finalize the logical association of dksA function with the triple-mutant phenotype, we then showed that overproduced DksA reverses the temperature sensitivity of the triple mutant. We reported in 2012 (2) that overproduction of DksA or GreA could reverse selected phenotypic features of (p)ppGpp0 mutants (auxotrophic requirements) and could themselves function as synthetic lethals, implying that GreA can display a function that is redundant with that of DksA. The question arises as to what functions are shared by these secondary channel RNAP–associated proteins. Several years ago in our lab, Rajendran Harinarayanan showed that it might be chaperone activity. He found that overproduced GreA and GreB reverse the temperature sensitivity of dnaKJ mutants in a similar manner to overproduced DksA. GreA and GreB functions are firmly established for reversing arrested transcription complexes. This well-known function is abolished in mutant variants of GreA and GreB, designated GreA* and GreB*. Analogous mutants exist for the (p)ppGpp help function of DksA, i.e., DksA*. Harinarayanan further demonstrated that GreA*, GreB*, and DksA* also reversed the temperature sensitivity of dnaKJ mutants. The association circle with tig/clpB, (p)ppGpp and ω was closed by showing that overproduced DksA* also reversed the temperature sensitivity of the triple mutant. The chaperone-like activities shared by these proteins, together with their involvement in regulation by (p)ppGpp suggest very strongly there is a commonality that is distinct from their better-known activities for reversal of arrested transcription complexes or DksA regulation of RNAP. The most apparent similarity is their structure, which uniformly consists of a long coiled-coil topped by a larger globular domain. The question now becomes how this type of a structure could have chaperone activity. There is literature evidence that GreA displays chaperone activity. The question will be studied more closely during the coming year as will the role of ω in (p)ppGpp regulation.


  1. Mechold U, Potrykus K, Murphy H, Murakami KS, Cashel M. Differential regulation by ppGpp versus pppGpp in Escherichia coli. Nucleic Acids Res 2013;41:6175-6189.
  2. Vinella D, Potrykus K, Murphy H, Cashel M. Effects on growth by changes of the balance between GreA, GreB and DksA suggest mutual competition and functional redundancy in Escherichia coli. J Bacteriol 2012;194:261-273.


  • James Kennison, PhD, Program in Genomics of Differentiation, NICHD, Bethesda, MD
  • Jose Lemos, PhD, University of Rochester Medical Center School of Medicine and Dentistry, Rochester, NY
  • Katsuhiko S. Murakami, PhD, Center for RNA Molecular Biology, Pennsylvania State University, University Park, PA
  • Katarzyna Potrykus, PhD, University of Gdańsk, Gdańsk, Poland
  • Agnieszka Szalewska-Palasz, PhD, University of Gdańsk, Gdańsk, Poland


For more information, email or visit

Top of Page