General stress response in α-proteobacteria: PhyR and beyond


E-mail; Tel. (+49) 89 2180 74622; Fax (+49) 89 2180 74626.


In addition to stress-specific responses, most bacteria can mount a general stress response (GSR), which protects the cells against a wide range of unspecific stress conditions. The best-understood examples of GSR are the σB-cascade of Bacillus subtilis and the RpoS response in Escherichia coli. While the latter is conserved in many other proteobacteria of the β-, γ- and δ-clades, RpoS homologues are absent in α-proteobacteria and their GSR has long been a mystery. Recent publications finally unraveled the core of the GSR in this proteobacterial class, which is mediated by EcfG-like σ-factors. EcfG activity is controlled by NepR-like anti-σ factors and PhyR-like proteins that act as anti-anti-σ factors. These unusual hybrid proteins contain an N-terminal EcfG-like domain that acts as a docking interface for NepR, and a C-terminal receiver domain typical for bacterial response regulators. Upon phosphorylation, PhyR titrates NepR away from EcfG, thereby releasing the σ-factor to recruit RNA polymerase and initiate transcription of its target genes. In this issue of Molecular Microbiology, Herrou et al. describe the function and three-dimensional structure of PhyR from Caulobacter crescentus. This structure is key to understanding the mechanism of the reversible, phosphorylation-dependent partner switching module that orchestrates the GSR in α-proteobacteria.


Bacterial life in its natural context stands in stark contrast to the defined laboratory situation, where pure bacterial cultures are grown under fixed and usually optimal conditions. Out in the wild, physicochemical parameters vary greatly, as does the nutrient supply. Moreover, bacteria have to struggle with countless competitors from various species for the available resources and ecological niches. The ability to adapt to the changing conditions present in their natural habitats is a central aspect but also one of the biggest challenges of bacterial life. Adaptation requires the accurate monitoring of critical parameters and a precise and specific information flow in order to mount an adequate response.

Bacteria have evolved a number of basic mechanisms to facilitate signal transduction. One-component systems, such as TetR or LacI, combine input and output functions in a single protein and predominantly respond to intracellular signals (Ulrich et al., 2005). In contrast, two major classes of phylogenetically unrelated mechanisms allow bacteria to respond to extracellular cues. In two-component systems, consisting of a histidine kinase that acts as a sensor and a corresponding response regulator that mediates the output (usually by acting as a transcriptional regulator), signal transduction is based on reversible phosphoryl group transfer (Gao and Stock, 2009). In contrast, the activity of so-called extracytoplasmic function (ECF) σ factors is controlled by their cognate anti-σ factors through direct protein-protein interactions: in the presence of inducing conditions, the anti-σ releases its grip, thereby allowing the σ-factor to bind alternative target promoters and induce expression of the downstream target genes (Helmann, 2002; Butcher et al., 2008).

Most bacteria, especially those living in complex environments, harbour many (often more than 100) such signal transducing systems to adapt to a plethora of different conditions (Ulrich et al., 2005). Typically, most of these systems respond to a specific stimulus by an adequately specific cellular output. In addition to mounting such specific responses, bacteria also have to adjust their physiology to an overall decline in environmental conditions, caused by hunger, overpopulation or changes in crucial physical parameters that affect many cellular processes simultaneously.

Under laboratory conditions, such global adjustments can, for example, be observed at the onset of stationary phase, when the overall living conditions deteriorate: with cell density reaching critical levels, nutrients start running low, and toxic metabolic byproducts begin to accumulate. In the environment, where logarithmic growth rarely occurs, these periods of hardship might be the rule rather than the exception. In bacteria living in complex, quickly changing habitats such as the soil or the phyllosphere, these general protection responses often also facilitate adaption to sudden stress conditions such as heat or oxidative stress that affect the cellular physiology at many different levels and therefore necessitate an overall adjustment of the bacterial physiology. Therefore, these responses are important both for long-term gradual adjustments, but also for quick and transient adaptations to protect the cells. Because of this broad spectrum of inducing conditions and the complexity of the resulting output, perhaps affecting the expression of hundreds of target genes, such adjustments have been termed ‘general stress responses’ (GSR).

From a signalling point of view, GSR requires the computation and integration of numerous different input signals in order to mediate a graded response that adequately protects and stabilizes the cell. Not surprisingly, the underlying regulatory cascades mediating GSR are often very complex, and may involve the activity of dozens of regulatory proteins that perceive stimuli, integrate signals and transduce information to mediate the desired output. Here, we will first give a brief overview of the two best-understood GSR networks, to highlight the variability but also to extract some common themes involved in bacterial GSR.

Paradigms of bacterial GSR: Escherichia coli RpoS and Bacillus subtilisσB

One important common theme that emerges from these two well-investigated examples is that the ultimate function of both signalling networks is to control the activity of a single alternative σ-factor. In this sense, ‘activity’ basically means how much of this specific GSR σ-factor is freely available to compete with the remaining cellular pool of other σ-factors for RNA polymerase core enzyme in order to induce gene expression of its target promoters. But the way that this activation is achieved differs dramatically between these two paradigms.

RpoS of E. coli

This alternative σ-factor is closely related to the primary (or vegetative) σ-factor. It is unique in the sense that its activity is not regulated by a typical anti-σ factor, in contrast to most other stress-inducible alternative σ-factors. Instead, its activity is tightly regulated at the level of transcription, translation and especially protein turnover (Hengge, 2010). RpoS is gradually induced upon entry into stationary phase to ensure long-term survival. This fine-tuned mechanism of activation is achieved by a combination of regulation at all the three levels mentioned above. Moreover, RpoS also quickly responds to sudden stresses, such as heat shock, hyperosmotic or acid stress, UV irradiation or sudden starvation. These fast responses are mainly mediated by a rapid inhibition of RpoS proteolysis, leading to a fast increase in the cellular pool of this σ-factor. The central player for this stress-inducible activation is the adaptor protein RssB, an unusual response regulator, which delivers RpoS to the ClpXP protease under non-inducing condition. This protein therefore represents a functional analogue to a classical anti-σ factor. Its activity is negatively affected by at least three anti-adaptors, which titrate RssB away from RpoS in response to various sudden stresses, such as phosphate starvation (IraP), magnesium starvation (IraM), as well as oxidative stress and DNA damage (IraD). Hence, these proteins represent analogues to the anti-anti-σ factors, which are present in the GSR described below.

RpoS controls the expression of more than 500 genes that are involved in stress-protective functions, alternative metabolic pathways, cell envelope biosynthesis and cell shape, biofilm formation and pathogenicity. RpoS also integrates input signals by cooperating with various transcription factors, thereby participating as a central player in a complex regulatory network. Moreover, RpoS-dependent gene regulation is also strongly affected by second messengers such as cAMP, ppGpp or c-diGMP. Further details on various aspects of this σ-factor can be found in a number of recent reviews (Hengge, 2008; 2009; 2010).

σB of B. subtilis

In contrast, activity of the σB GSR regulator from B. subtilis (and many other Firmicutes bacteria) is ultimately regulated by a controlled and reversible release from its anti-σ factor RsbW. The major signalling features of this GSR are phosphorylation-dependent partner-switching modules mediated by opposing kinase/phosphatase pairs that integrate input signals from two distinct pathways in response to environmental or energy stress (Fig. 1, right side) (Hecker et al., 2007; Price, 2010). Signal integration occurs at the level of the anti-anti-σ factor RsbV, which gets dephosphorylated by the two phosphatases RsbP (energy pathway) and RsbU (environmental pathway) in the presence of stress conditions (Fig. 1). Active RsbV than titrates RsbW (anti-σ factor) away from σB, thereby ultimately releasing the σ-factor from its inhibitory grip to redirect transcription initiation to the GSR regulon. In the absence of stress, the anti-σ factor RsbW actively phosphorylates (and thereby inactivates) RsbV. This mechanism is also used to reset the system to the inactive pre-stimulus state.

Figure 1.

Regulatory networks orchestrating the general stress response in α-proteobacteria (left) and Bacillus subtilis (right). Colour code: kinases yellow, phosphatases red, phospho-acceptors blue, σ-factors green. Inhibition is indicated by T-shaped lines, phosphoryl groups by grey circles, labelled ‘P’. HK = histidine kinase. Partner switching modules are underlain in grey, with the switching proteins represented as tilted squares. See text for detail.

The σB response is induced by a plethora of signals, such as entry into stationary phase, addition of uncouplers of the proton motive force (such as CCCP and nitric oxide), starvation for carbon, phosphate or oxygen (energy pathway). Moreover, environmental stresses, such as acid, ethanol, heat or osmotic stress and blue light irradiation also trigger the GSR (environmental pathway). Deletion of sigB leads to increased sensitivity to some, but not all inducing conditions. σB controls about 150–200 genes, only few of which encode obvious stress-related functions. In similarity to the situation described for RpoS above, many σB-dependent genes are co-regulated by other transcription factors (Hecker et al., 2007; Price, 2010).

GSR in α-proteobacteria: distribution and conservation of the PhyR–NepR–EcfG cascade

Comparative genomic analyses have revealed that homologues to core components of these two different types of GSR networks can be found in many bacterial genomes. On the other hand, some phylogenetic groups such as the α-proteobacteria completely lack any such regulators, and how the GSR was mediated in this bacterial class was unclear for a long time. Recent reports have begun to reveal the nature of the α-proteobacterial GSR. Ground-breaking studies from the group of Julia Vorholt identified the central players of the GSR response in the methylotrophic bacterium Methylobacterium extorquens (Gourion et al., 2006; 2008; Francez-Charlot et al., 2009). Their data were confirmed by results obtained in other α-proteobacteria, such as Bradyrhizobium japonicum, Caulobacter crescentus, Rhizobium etli and Sinorhizobium meliloti (Alvarez-Martinez et al., 2007; Sauviac et al., 2007; Gourion et al., 2009; Martinez-Salazar et al., 2009; Bastiat et al., 2010).

The picture that emerged from these studies is that the regulatory cascade orchestrating the α-proteobacterial GSR shows some similarities with the σB-dependent response, but is based on a set of unique proteins not found in any other bacterial stress signalling pathway to date (Fig. 1, left side). Instead, it reshuffles and combines domains from two phylogenetically unrelated signalling archetypes, two-component systems and ECF σ-factors. The input section of this GSR is made up of a specialized TCS consisting of as yet unidentified stress-responsive sensor histidine kinases and an unusual response regulator, PhyR, which will be described in more detail below. The output section is represented by a unique anti-σ/σ-factor pair of NepR- and EcfG-like proteins respectively. In the absence of stress signals, NepR binds to EcfG, thereby keeping this ECF σ-factor inactive. EcfG-dependent gene expression is induced by stress conditions very reminiscent of those described for other GSR, including heat and osmotic stress, carbon and nitrogen starvation, and desiccation (Francez-Charlot et al., 2010). Under such conditions, the postulated sensor kinase(s) phosphorylates their cognate PhyR-like response regulator, which then binds NepR, thereby titrating it away from EcfG. Phosphorylation PhyR therefore acts as an anti-anti-σ factor. This phosphorylation-dependent partner switch – which is remiscent of the σB-cascade in B. subtilis (Fig. 1) – ultimately releases EcfG, which then recruits RNA polymerase core enzyme to initiate transcription of its target genes. Once the stress is relieved, dephosphorylation of PhyR again reverses the switch, hence shutting off the GSR (Fig. 1).

Comparative genomics analyses revealed that all three proteins are conserved within, but also restricted to, the class of α-proteobacteria (Starońet al., 2009; Francez-Charlot et al., 2010). In this bacterial clade, they are only absent in some degenerated obligate symbionts highly adapted to very constant host conditions, including the genera Anaplasma, Rickettsia or Wolbachia. The composition and genomic organization of the known GSR components from a number of prominent α-proteobacteria is summarized in Table 1.

Table 1.  Distribution of EcfG, PhyR and NepR in selected α-proteobacterial genomes.a
OrganismPhyR locus organizationbPhyRNepRcEcfGHWEdHisKA2d
  • a. 

    A complete table containing a detailed list of proteins from all α-proteobacteria is available on request from the authors.

  • b. 

    E, EcfG; HK, Histidine kinase harbouring Pfam:HisKA_2 domain; HW, histidine kinase harbouring Pfam:HWE-HK domain; N, NepR; P, PhyR. The arrows indicate the operon organization and orientation.

  • c. 

    nepR is not always annotated.

  • d. 

    Histidine kinases harbouring either HWE or HisKA_2 domains. Values in parentheses: number of (cytoplasmic-/periplasmic-sensing) histidine kinases.

Acetobacter pasteurianus<HK-N-E-E< >P>1121 (0/1)
Agrobacterium tumefaciens<E-N< >P-Hw> <Hw<1135 (4/1)1 (1/0)
Bartonella spp.<HK-E-N< >P-Hw>1111 (1/0)1 (0/1)
Bradyrhizobium japonicum<P< >N-E>1116 (3/3)3 (3/0)
Bradyrhizobium sp. BTAi1<P< >N-E> <HW<11112 (6/6)3 (2/1)
Brucella spp.<HW-X-P< >N-E-HK>1112 (2/0)1 (0/1)
Caulobacter crescentus<HK-E-N< >P>1127 (5/2)3 (1/2)
Erythrobacter litoralis>P><HW< >N-E>1114 (3/1)3 (3/0)
Gluconobacter oxydans<P< >E-E-HK>1022 (2/0)2 (1/1)
Hyphomonas neptunium<HK-E-N-X< >X-P>// >P> <HW<2124 (3/1)1 (0/1)
Methylobacterium extorquens<N< >P>11611 (9/2)5 (4/1)
Methylobacterium sp. 4–46<N< >P>111011 (9/2)6 (4/2)
Nitrobacter winogradskyi<P< >N-E> <HW<1112 (0/2)0
Paracoccus denitrificans<HK-E-N< >P>1121 (0/1)1 (0/1)
Rhizobium etli<HW-P< >N-E>1123 (1/2)3 (3/0)
Rhodobacter sphaeroides<HK-E-N< >P>1124 (4/0)5 (1/4)
Rhodopseudomonas palustris<P< >N-E> <HW<1118 (7/1)3 (2/1)
Rhodospirillum centenumP1003 (3/0)5 (3/2)
Roseobacter denitrificans<HK-E< >P>// <HK-E-N< >P>21202 (0/2)
Sinorhizobium medicae<HW-E-N< >P>// <P< >N>2225 (4/1)1 (1/0)
Sinorhizobium meliloti<HW-E-N< >P>// P2127 (6/1)0
Sphingopyxis alaskensis<E< >X-E> <P< >HW>1022 (1/1)0

The ‘typical’ locus consists of a phyR gene, which is divergently transcribed from a shared intergenic region with the nepR–ecfG operon. Due to its small size, homologues of nepR– while normally present – are not always annotated. In addition, genes encoding sensor kinases are usually in the vicinity and often co-transcribed with either of the two cistrons. With few exceptions, most genomes encode a single copy of PhyR and NepR. In contrast, the number of EcfG-like σ-factors can vary significantly (Table 1), indicative for regulatory diversification at the level of gene expression (Fig. 1). This represents a remarkable difference compared with all the other GSR described before, which invariantly co-ordinate the activity of a single σ-factor. The physiological significance of this observation remains to be identified. But one could imagine that more than one output σ-factor either allows for more graded responses or the induction of distinct sub-regulons.

PhyR: an unusual hybrid protein acting as a phosphorylation-dependent anti-anti-σ factor

As mentioned above, PhyR represents the central regulatory switchboard of the α-proteobacterial GSR. It presumably integrates the information of different sensor kinases (see below) and indirectly regulates the activity of EcfG-like σ-factors that orchestrate the GSR.

PhyR is an unusual protein in a number of respects. It is a response regulator that carries its receiver domain (the interaction interface with the cognate sensor kinases) at its C-terminus. This organization, while not unprecedented, is only found in a small minority of other response regulators, including CheW-like proteins, all of which do not function as transcriptional regulators (Galperin, 2010). The N-terminal output domain of PhyR shows high sequence similarity to ECF σ-factors, at least at first glance. But a closer inspection revealed that this domain lacks a number of residues crucial for DNA-binding, and indeed no such activity has been described for PhyR. Instead, this domain shows specific homology to EcfG-like proteins (Starońet al., 2009; Francez-Charlot et al., 2010). Biochemical studies revealed that – upon PhyR phosphorylation – this domain serves as a docking module for NepR (Francez-Charlot et al., 2009). Hence, NepR switches its partner under stress conditions, thereby releasing the EcfG to initiate expression of the GSR regulon.

The high-resolution crystal structure of PhyR from Caulobacter crescentus, published in this issue of Molecular Microbiology, offers fascinating insight into how its functional role is achieved by the three-dimensional architecture of this protein (Herrou et al., 2010). Initially, the authors genetically infer that C. crescentus PhyR indeed functions as a phosphorylation-dependent anti-anti-σ factor, in line with previous reports on PhyR proteins (Francez-Charlot et al., 2009; Gourion et al., 2009; Bastiat et al., 2010). Subsequently, they solve the crystal structure of this protein at 1.25Å. This structure yields a number of important insights.

In accordance with previous results in solution, the crystal structure is consistent with a model in which PhyR functions as a monomer. In contrast to canonical response regulators, which dimerize upon phosphorylation, PhyR activity is mediated by phosphorylation-dependent changes in intramolecular interactions.

PhyR can be divided into three structural domains that – by themselves – do not offer a lot of surprises. The N-terminal EcfG-like domain contains two sub-domains with folds typical of regions σ2 and σ4, which are responsible for binding the −10 and −35 promoter regions, respectively, in all σ70 proteins. Hence the σ-like domain of PhyR is structurally homologous to other ECF σ factors, such as RpoE from E. coli and R. sphaeroides (Campbell et al., 2008). In this unphosphorylated structure of PhyR, the σ-like domain exists in a closed conformation and there is evidence that the loop region linking regions σ2 and σ4 is highly flexible. As expected, the C-terminal receiver domain of PhyR adopts a structure typical of classical response regulators (Bourret, 2010).

What is most remarkable about the PhyR structure is the identification of extensive interaction interfaces between the C-terminal receiver domain and both regions of the EcfG-like domain. Considering that the latter alone is sufficient to perform the anti-anti-σ activity, the authors suggest that the receiver domain therefore can be viewed as an anti-anti-anti-σ domain that inhibits the EcfG-like anti-anti-σ domain through direct interactions in the absence of phosphorylation. The inhibitory interaction interface differs from the one necessary for the transient interactions between sensor kinase and response regulator mediating phosphoryl-group transfer. Hence, the former does not get in the way of the latter. The structure offers evidence that, in the absence of inducing signals, the unphosphorylated receiver domain of PhyR tethers the EcfG-like domain in a closed conformation. Phosphorylation of the receiver domains usually induces intramolecular conformational changes that ultimately activate the output domain of a response regulator (Bourret, 2010). In the case of PhyR, it is attractive to hypothesize that phosphorylation changes its conformation to release the two regions of the EcfG-like domain from its inhibitory grip, thereby allowing this anti-anti-σ domain to interact with NepR-like anti-σ factors (Herrou et al., 2010). But how this interaction is achieved remains to be analysed.

Outlook: NepR interactions, sensor kinases and beyond

The PhyR structure presented in this issue of Molecular Microbiology is an important step towards understanding the mechanism underlying the PhyR–NepR–EcfG-mediated signal transduction of the GSR in α-proteobacteria. But it will require further structural studies to truly understand how the partner switching module works.

Specifically, how is the NepR anti-σ factor unlocked from EcfG and titrated to the EcfG-like domain once PhyR gets phosphorylated under stress conditions? For this mechanism to be efficient, one has to postulate that the EcfG-like domain of PhyR – in its phosphorylated state – has a higher affinity for NepR than the corresponding σ-factor itself. NepR does not share any sequence similarity with other anti-σ factors. While structural studies on the RpoE-RseA pair of E. coli and RpoE–ChrR of R. sphaeroides have revealed the mechanism underlying the inhibition of ECF σ-factors by ‘classical’ anti-σ factors (i.e. those containing a so-called anti-sigma domain) (Campbell et al., 2008), it is unclear whether this mode of interaction will also hold true for NepR.

An additional question regards the nature of the histidine kinases that sense the stress conditions and – upon autophosphorylation – serve as phospho-donors for PhyR-like response regulators. Here, the first indications come from comparative genomics. Numerous such analyses have demonstrated that, at least in bacteria, genomic context conservation is usually a reliable indicator of functional links between proteins encoded by genes adjacent to each other on the chromosome. And indeed, we do find genes encoding sensor kinases in the vicinity of quite a number of phyR loci (Table 1). But at first glance, these sensor kinases – if present at all – look rather diverse, including both cytoplasmic- and periplasmic-sensing proteins of varying domain architecture (Fig. 2).

Figure 2.

Domain architecture of histidine kinases (HK) associated with PhyR. Proteins are symbolized as grey lines and drawn to scale. Putative transmembrane regions are represented by blue vertical bars. All other domains according to the graphical output of the SMART database (, with modifications. See text for details.

But bearing in mind that the specificity of a histidine kinase for its partner response regulator is defined by the dimerization and histidine phosphotransfer (DHp) domain (Gao and Stock, 2009), a closer look revealed one important similarity between all kinases genetically linked to PhyR. They all do not contain ‘typical’ DHp domains (Pfam:HisKA), but instead harbour either a Pfam:HWE-HK or a Pfam:HisKA_2 DHp domain. Remarkably, the HWE and HisKA_2 domains are very similar to each other, but differ significantly from the vast majority of classical sensor kinases (Karniol and Vierstra, 2004). What is even more striking is that both types of sensor kinases are either predominantly (HisKA_2; 55%), or in the case of the HWE kinases almost exclusively (95%) found in genomes of α-proteobacteria. These observations strongly suggest that the only likely phospho-donors for PhyR belong to either of the two kinase groups. This hypothesis, if true, narrows down the range of potential proteins significantly, which should allow their identification even in cases where no candidate gene is found next to phyR on the chromosome (Table 1). Lastly, the observation that such kinases can be either cytoplasmic- or periplasmic-sensing offers the possibility to integrate both environmental and intracellular signals, again analogous to the situation described for the σB-dependent cascade of B. subtilis. But experimental evidence will be required to verify these hypotheses on the nature of the sensor kinases able to phosphorylate PhyR-like proteins.

Although the study of GSR in the α-proteobacteria is just beginning, the data already available indicate that more surprises are likely in store. And structural studies, like the one of PhyR reported in this issue of Molecular Microbiology, will be pivotal to ultimately unravel to mechanism behind this novel GSR cascade.


We would like to thank Sean Crosson and John Helmann for critical reading of the manuscript, and Regine Hengge, Chet Price and Julia Vorholt for sharing manuscripts prior to publication. Work in the authors' lab was supported by grants from the Deutsche Forschungsgemeinschaft (MA2837/2-1) and the Fonds der Chemischen Industrie.