Processing the Ps in Times of Stress

(p)ppGpp and Poly-P in the Bacterial Stringent Response

by Daniel
 

A decade ago, Fred Neidhardt wrote a tour de force post for our blog on a case of convergence be­tween bacterial physiology and virulence that featured a review by Dalebroux et al. from Mi­chele Swanson's lab. The specific topic of that convergence was the role that the nucleotide guanosine tetraphosphate (ppGpp; 5'-diphosphate, 3'-diphosphate) plays in the intracellular differentiation of Legionella pneumophila as part of the molecular pathogenesis underlying Legionnaire's disease. Fred's post splits evenly between background on the physiology of ppGpp as the master regulator of bacterial growth during the 'stringent response', and a case study of the relevance of this 'alar­m­one' to L. pneumophila pathogenesis.
 

As an aside, the introduction of that post also bears rereading for its historical perspective. I would take it for granted that basic physiology and virulence must go hand-in-hand, but Fred opens his piece by highlighting that this wasn't always the perspective in bacteriology.
 

After a long absence, the topic of ppGpp reappeared here last year in a post by Elizabeth Mueller & Corey Westfall, a graduate student and post-doctoral fellow in the lab of my former PhD advisor, Petra Levin. In it they describe research demonstrating a role for ppGpp during 'normal' cellular growth homeostasis, apart from any stress response. While I was helping edit that post, I learned about a new side-note in the ppGpp story with research presented by Michael Gray at the Microbi­al Stress Response Gordon Research Conference and the Molecular Genetics of Bacteria and Bac­teriophage meeting. Gray's research, since published in the Journal of Bacteriology, presents re­sults that significantly update a long-standing model from Arthur Kornberg's group at Stanford on the connection between ppGpp and another, less-discussed, molecule involved in bacterial stress response: inorganic polyphosphate (poly-P).
 

Before introducing Kornberg's model and covering the results from Gray that revise it, I'll take a moment to describe poly-P and its roles. I'll also briefly remind readers about the basics of ppGpp, the stringent response, and the associated transcription factor DksA. For any readers not already familiar with these ppGpp-related topics, the above links to Small Things Considered posts and journal articles (including this more recent review) will be useful.
 

A Spot of Guanosine Magic for Stressed Cells

Nicknamed 'magic spots' upon their surprise discovery on thin-layer chromatography autoradio­grams just over fifty years ago by Michael Cashel in Jonathan Gallant's lab, guanosine tetra­phos­phate (ppGpp) and guanosine pentaphosphate (pppGpp) were first observed accumulating in cells during amino acid starvation. Accumulation of these GDP/GTP-derivatives leads to alterations of central cellular processes including transcription, translation, and replication. This so-called 'strin­gent response' serves to halt cell growth and regulate adaptive changes in gene expression, both positive and negative. This transcriptional reprogramming and its mechanism differs between bac­terial species, but typically includes genes involved in stress response, amino acid synthesis/trans­port, and rRNA synthesis. Originally defined specifically with regards to amino acid starvation, the term 'stringent response' has since come to encompass myriad cellular stresses that lead to accu­mu­lation of (p)ppGpp. As mentioned above, (p)ppGpp also plays a role in general growth rate ho­meostasis.
 

Figure 1. Identified sites of interaction bet­ween RNA polymerase, DksA, and ppGpp. A Graphical abstract from Ross et al. reporting two distinct ppGpp binding sites, 60 Å apart, on E. coli RNA poly­merase. In addition to the prev­iously iden­­tified Site 1 for ppGpp binding alone, tri­partite inter­actions between poly­merase, ppGpp, and DksA at the then newly iden­tified Site 2 reveal the struc­tural basis behind en­hance­ment of DksA’s reg­ulation of RNA Poly­merase by ppGpp. Strains with defects in Site 2 show im­paired growth follow­ing nut­ritional shifts due to mis­­managed trans­criptional reg­ulation as part of the strin­gent res­ponse. B Mod­ified graph­ical ab­stract from Molodtsov et al. reporting two modes of DksA binding to RNA Poly­merase ac­­cording to allo­steric reg­ulation by ppGpp to stab­ilize DksA in­hib­ition of trans­cription initia­­tion via destab­ilization of open complex for­m­a­tion.

In E. coli, while (p)ppGpp can directly interact with RNA polymerase to regulate transcription ini­tia­tion, its activity requires a protein co-factor named DksA, discovered in Richard Gourse's Lab. To­gether, DksA and (p)ppGpp as­sociate with RNA polymerase to either down- or up-regu­late tran­scription, and also reduce polymerase affini­ty for the housekeeping sigma factor while increasing its affini­ty for alternate sigma factors. You may wonder how the same partners can have comple­tely opposite effects on the same enzyme. The answer comes from a discrimi­na­tor sequence that lies in the DNA of a given operon be­tween the −10 promoter element and the transcriptional start site. Promoters for downregulated genes have GC-rich discrimina­tors that have relatively poor in­teraction with the RNA polymerase sigma factors. The presence of DksA/(p)ppGpp exacerbate the unstable open complex formation for transcription of these genes. In contrast, the opposite holds true for genes with AT-rich discrimi­nators.
 

A decade ago when Fred wrote his post on ppGpp and L. pneumophila, he noted that the structur­al details of ppGpp-binding to RNA polymerase along with DksA were still unclear. A few years back the Gourse Lab published structures showing that ppGpp can bind RNA polymerase both at a previously determined site (to exert effects on its own) and at a second site at the interface with DksA (Figure 1A). There, ppGpp is able to modulate DksA activities that DksA also has some ability to perform on its own. Later published crystallization data from the Murakami Lab further demon­strated that allosteric binding of ppGpp to DksA alters its binding of RNA polymerase during tran­scriptional inhibition (Figure 1B). However, the structural interactions during transcriptional acti­va­tion may differ.
 

As Fred noted in his post, "Nothing is simple in biological regulation (life is too interesting for sim­pli­­city and biologists have come to expect complexity)…" The tripartite interactions of (p)ppGpp, DksA, and RNA polymerase – plus the contribution of cis-regulatory elements of the DNA in ques­tion being regulated – makes the stringent response a fascinating molecular process that doubt­less has more magical secrets to reveal.
 

Indeed, compounding this topic even further: (p)ppGpp is not the only phosphate-containing mo­le­cule that accumulates as part of the bacterial stress response. Though less understood, cells al­so produce inorganic polyphosphate as an essential component of adaptation for growth and sur­vival.
 

Polyphosphate Production and its Relationship to (p)ppGpp

Both prokaryotic and eukaryotic cells regulate accumulation of inorganic polyphosphate, linear po­­­lymers of phosphates extending from dozens to hundreds of subunits, each linked by phospho­an­hydride bonds familiar in nucleotide di- and tri-phosphates. In the later years of his lab, Arthur Kornberg and his group devoted a large portion of their research to studying the biological roles of poly-P and arguing for its consideration both in textbooks and further experiment. This culmi­na­ted in an Annual Review of Biochemistry paper by Rao, Gómez-García, and Kornberg pub­lished in 2009, after Kornberg's death.
 

Long observed as electron-dense storage granules (volutin) and within vacuoles (such as the aci­­do­­­calcisome of trypanosomes), cellular deposits of poly-P are widespread in bacterial species and eukaryotic cytosol and organelles. Homologs of genes involved in poly-P me­tabolism have also been identified in archaea, but the archaeal representatives have been little studied. Abiotic (geo­logical) formation of inorganic polyphosphate also occurs within volcanoes and deep-sea vents. Given its apparent use across Domains, this has led to speculation of a pre­biotic role for poly-P in the development of life, perhaps linked to the chemistry of the NTPs found in an 'RNA world'.
 

Cells principally synthesize poly-P through the reversible activity of the enzyme Poly P Ki­nase, by taking γ-phosphates from ATP, and/or GTP, depending on the family member. Whether directly or directly, poly-P has been implicated in diverse phenotypes including Myxo­coc­cus xanthus preda­tion, virulence (including in Mycobacterium tuberculosis), sporulation (including of Acetonema longum), and DNA replication (including during lytic phage growth.) In the social amoeba Dictyo­stelium discoideum, poly-P also plays a role in cytoskeleton function and motor protein localiza­tion. Intriguingly, in this organism at least, a Poly P Kinase enzyme assem­bles into a cytoskeleton-like filament. Though Kornberg's group speculated poly-P may also play a role in prokaryotic cyto­skeletal structure, I don't know of any demonstration of this to date.
 

(click to enlarge)
Figure 2. Model for stress-induced poly-P accu­mulation proposed by the Kornberg lab. This mo­del consists of multiple reg­ulatory path­ways, but the path­way under nut­rient limit­ation pro­poses that ppGpp accum­ulation during the strin­gent res­ponse through RelA and SpoT acti­vity, leads to poly-P accum­ulation by PPX inhi­bi­tion and/or RpoS activ­ation. Based on known involve­ment of the RpoS sigma factor, they also pro­pose the exis­tence of an uniden­tified factor ("X"), which could lead to poly-P accum­ulation through any single path­way or com­bination of path­ways (arrows.) Source

Similar to mutants for genes involved in synthesis of (p)ppGpp, bacterial mutants (for example, of Pseudomo­nas aeruginosa) lacking Poly P kinase activity have a va­riety of pleiotropic growth related phenotypes, stemming from genome-wide changes in transcription. This sug­gests that poly-P levels somehow feed into transcription­al regulation. Furthermore, activation of the stringent response during amino acid starvation in E. coli leads to a >100-fold accumulation of poly-P, where it then activates LonA protease >20-fold to increase the amino acid pool.
 

Just as (p)ppGpp accumulation serves as part of general stress response, not just during amino acid starvation, so too does accumulation of poly-P. Together with the tem­poral stress-induced correlation between (p)ppGpp and poly-P synthesis to regulate transcription – and shared pleiotropic phenotypes – other observations have linked the two molecules. For one, the enzyme largely responsi­ble for degrading poly-P, Exopolyphosphatase (PPX), is inhibited by (p)ppGpp. Secondly, mutants lacking the genes responsible for (p)ppGpp synthesis (relA and spoT) are defective in poly-P synthesis. This led the Kornberg group to a model for poly-P accumulation that featured a dependency (at least in part) on synthesis of ppGpp (Figure 2). More specifically, this model proposed that certain stres­ses, such as amino acid starvation (nutrient limitation) required (p)ppGpp for poly-P accumulation, though other stresses (such as nitrogen or phosphate limitation, or salt stress) could act through other regulators.
 

This model held until recently, when research by Michael Gray further clarified these proposed relationships between (p)ppGpp with poly-P accumulation during the stringent response and re­vealed instead a more direct role for DksA and related transcription factors in poly-P synthesis.
 

The Dangers of Working with (p)ppGpp Mutants and Single Lab Strains

Gray revisited the Kornberg model of (p)ppGpp-dependent poly-P accumulation after getting re­sults seemingly contradictory to the model shown in Figure 2. Specifically, he had measured poly-P levels in E. coli following nutrient limitation across a number of strain backgrounds he had on hand in lab (Figure 3A). As expected, a Δppk strain did not produce poly-P. However, ΔrelA spoT:: cat⁺ (Cam^R) strains still accumulated poly-P in response to nutrient limitation, suggesting cells produced poly-P in the classical stringent response despite lacking (p)ppGpp. Given this contra­diction with the established model of (p)ppGpp regulating poly-P production under these con­di­tions, Gray validated his surprising results through several additional experiments. This included expressing ppGpp artificially from plasmid-borne relA⁺ under non-stress conditions. While this ppGpp slowed growth as would be expected, it did not result in poly-P increases.
 

Figure 3. (p)ppGpp is neither necessary nor suf­ficient for poly-P accumulation dur­ing nutri­ent limitation, unlike the Korn­berg model (Fi­gu­re 2.) A After nut­rient limit­ation an E. coli Δppk mut­ant made no poly-P, and a Δppx mut­ant ma­de slightly more than the wild type, as ex­­pec­ted. How­ever, un­like as reported prev­iously, ΔrelA, ΔrelA spoT::cat⁺, and Δppx ΔrelA spoT:: cat⁺ mutants accum­ulate at least as much poly­-P as the wild-type strain. B A strain descen­ded from the Korn­berg lab strain lacking ppGpp (CF1693(M⁺) relA::kan⁺ spoT:: cat⁺) does show def­ective poly-P accum­ulation during nut­rient limit­ation as prev­iously re­ported. How­ever, transfer of the relA and spoT mut­ations into a fresh strain back­ground does not recap­itulate the ppGpp dep­endence pheno­type, argu­ing that other un­charac­terized mut­ations were pre­sent in the ori­ginal Korn­berg lab strain to cause the observed defect in poly-P accum­ulation. Source

Why this seeming discrepancy? To further investigate matters, Gray obtained descendants from Kenn Gerdes of original strains used by the Kornberg lab for their studies and found that they did have significant defects in poly-P production when lacking genes for ppGpp syn­thesis (Figure 3B). Gray replaced the ΔrelA and spoT alleles of the Kornberg-derived alleles with those from his strains and found that they still had reduced poly-P synthesis upon nutrient limitation, unlike his strains. This suggested that something else in the Kornberg-derived strains was responsible for the differing results.
 

As mentioned above, strains lacking relA and spoT for ppGpp synthesis often demonstrate pleiotropic pheno­types, including ones stemming from complex amino acid requirements due to their inability to induce synthe­sis upon limitations. Freshly constructed relA spoT mu­tants, therefore, grow very poorly in minimal media. As a consequence these double mutants are under high se­lective pressure for accumulation of second-site sup­pres­sor mutations that permit growth. Well known examples of such suppressors include mutations in the β, β', and σ70 subunits of RNA Polymerase. These are believed to mimic the effects of ppGpp on RNA polymerase. Gray therefore sequenced RNA polymerase subunit genes and found the Kornberg lab-derived strains did indeed con­tain point mutations in both rpoB and rpoD. Though neither of the mutations he found had been previously identified as stringent mutants, replacement of the wild-type rpoB and rpoD alleles of his strains with these mu­tations now produced the loss of poly-P accumulation seen originally by the Kornberg lab.
 

Thus, Gray's results suggest that data linking poly-P synthesis to direct regulation by (p)ppGpp le­vels during the stringent response (giving rise to the Kornberg model) had likely been misinter­pret­ed due to unknown stringent mutations in RNA polymerase subunits selected for in the ori­gi­nal Kornberg lab strains. Loss of poly-P accumulation during nutrient limitation therefore is not a result of lost (p)ppGpp synthesis, but rather perhaps due to alterations in transcription due to changes in RNA polymerase structure, which perhaps affect transcription factor interactions.
 

Figure 4. DksA is required for poly-P ac­cumula­tion during nutrient limitation, re­gardless of the pres­ence of (p)ppGpp. Ad­ditionally, deletion of greA, en­coding an­other trans­cription fac­tor stru­c­turally sim­ilar to DksA, par­tially res­tores poly-P ac­cum­ulation in the ΔdksA back­ground, sug­gesting that inter­actions bet­­ween tran­s­crip­t­ion factors (not (p)ppGpp) and the secon­dary chan­nel of RNA poly­merase im­pacts reg­ulation of poly-P prod­uction (mod­ified.) Source

As discussed above, RNA polymerase works not only with (p)ppGpp to regulate transcription during the stringent response, but also with the transcription factor DksA, and dksA mutants typically display comparable phenotypes to relA spoT mutants lacking ppGpp. Gray therefore looked at the effects of DksA on poly-P production in the presence of wild-type RNA polymerase (no stringent mu­tations) (Figure 4). Unlike the relA spoT mutants, dksA mutants were deficient in poly-P accumulation during nutrient limitation, regardless of the presence of (p)pp­Gpp, consistent with DksA regulating poly-P synthesis. Many dksA defects can be compensated for through eli­mination of another, structurally similar transcription factor named GreA that interacts with the same regions of RNA polymerase (the secondary channel). As seen in Figure 4, deletion of greA restored poly-P accumulation in the dksA-null background, arguing further that interact­ions between transcription factors and the secondary channel of RNA polymerase impacts regulation of poly-P production.
 

Though (p)ppGpp may not be required for regulation of poly-P accumulation during the stringent response, Gray's results demonstrate that it still is clearly involved, highlighting the increased com­plexity behind the regulation of these molecules. Gray's future work sets to identify the currently unknown gene(s) whose transcription (regulated by DksA and GreA, but not ppGpp) promotes po­ly-P production. The mechanism by which this transmits to Ppk for increased poly-P production also remains unclear. And this is just E. coli.
 

Gray's work reminds us of the importance to always consider the assumptions we make when working with various lab strains. It is also a good reminder that sometimes contradictory data aren't wrong, but an indication of something more complex going on beneath the surface level of our models.
 

Finally, I am reminded again of Fred Neidhardt's comments on the complexities of biology in his ppGpp post a decade ago. We've gone huge strides further in our molecular understanding of the stringent response since then, but it appears there will be no shortage of questions still for the fu­ture.