|Home | About | Journals | Submit | Contact Us | Français|
The network controlling the general stress response in Bacillus subtilis requires both the RsbP phosphatase and the RsbQ α/β hydrolase to convey signals of energy stress. RsbP contains three domains: an N-terminal PAS, a central coiled-coil, and a C-terminal PP2C phosphatase. We report here a genetic analysis that established the functional interactions of the domains and their relationship to RsbQ. Random mutagenesis of rsbP yielded 17 independent bypass suppressors that had activity in an rsbQ null strain background. The altered residues clustered in three regions of RsbP: the coiled-coil and two predicted helices of the phosphatase domain. One helix (α0) is unique to a subfamily of bacterial PP2C phosphatases that possess N-terminal sensing domains. The other (α1) is distinct from the active site in all solved PP2C structures. The phenotypes of the suppressors and directed deletions support a model in which the coiled-coil negatively controls phosphatase activity, perhaps via the α0-α1 helices, with RsbQ hydrolase activity and the PAS domain jointly comprising a positive sensing module that counters the coiled-coil. We propose that the α0 helix characterizes an extended PP2C domain in many bacterial signaling proteins, and suggest it provides a means to communicate information from diverse input domains.
Understanding how distant sites interact to control protein function is of interest for diverse areas of biology. The regulation of cellular processes provides numerous examples of interdomain communication to control protein activity, and in many cases the means of this communication is not well understood. Here we report a genetic analysis that outlines the signaling pathway within the RsbP serine phosphatase, a multi-domain protein which activates the general stress response of the bacterium Bacillus subtilis (Vijay et al., 2000).
The general stress response brings about a special physiological state that enhances bacterial survival in the natural environment and in certain pathogenic interactions. In B. subtilis and related organisms, this response is controlled by the σB transcription factor, whose activity is regulated by a signaling network that employs the “partner switching” mechanism (reviewed in Hecker et al., 2007). As shown in Fig. 1, the mechanism uses serine phosphorylation to control interactions among a characteristic set of regulatory proteins. In this network, the primary regulator is the RsbW anti-σ factor/kinase, and its association with σB is governed by the phosphorylation state of the RsbV anti-anti-σ factor (Dufour and Haldenwang, 1994; Alper et al., 1996). This state reflects the balance between the serine kinase activity of RsbW and activity of the input phosphatases, RsbU and RsbP, which are differentially required for response to environmental or energy stress signals (Voelker et al., 1996; Yang et al., 1996; Vijay et al., 2000). In unstressed cells the RsbV anti-anti-σ is phosphorylated and cannot interact with RsbW, which then holds σB in an inactive complex. When either of the input phosphatases overbalances RsbW kinase activity, dephosphorylated RsbV binds RsbW to force the release of σB, initiating transcription of the general stress regulon.
Here our focus is the RsbP phosphatase and the RsbQ α/β hydrolase, both of which are necessary for energy stress activation of σB (Vijay et al., 2000; Brody et al., 2001). RsbP belongs to the PPM serine phosphatase family (Protein Phosphatase, Mg+2- or Mn+2-dependent), members of which manifest a range of signaling roles in all three kingdoms of life (Bork et al., 1996; Zhang and Shi, 2004). They are often called PP2C phosphatases after the defining member of the family, human PP2Cα (Das et al., 1996). Phylogenetic analysis suggests that bacterial PP2C phosphatases originated in eukaryotes (Ponting et al., 1999), with subsequent horizontal gene transfer and diversification into two distinct subfamilies (Zhang and Shi, 2004). Subfamily I retains more sequence similarity to eukaryotic PP2C phosphatases and its members are thought to antagonize the activity of standard Hanks serine-threonine kinases. Most have only a PP2C catalytic domain, but some have a C-terminal membrane anchor that can extend outside the cell (Wehenkel et al., 2007). The four bacterial PP2C phosphatases whose structures have been solved all belong to this group (Pullen et al., 2004; Bellinzoni et al., 2007; Rantanen et al., 2007; Schlicker et al., 2008). By contrast, Subfamily II has diverged further from the eukaryotic archetype (Zhang and Shi, 2004), and its experimentally characterized members antagonize a class of serine-threonine kinases evolved from bacterial histidine protein kinases (Duncan and Losick, 1993; Min et al., 1993). This subfamily is larger than the first, and its representatives usually possess a multidomain organization that embraces a variety of N-terminal sensory domains. Subfamily II includes B. subtilis RsbP, which contains an N-terminal PAS (Per-Arnt-Sim) domain and a C-terminal PP2C domain (Vijay et al., 2000). PAS domains typically bind small molecules that alter PAS communication with intra- or intermolecular signaling partners (Taylor and Zhulin, 1999). In some cases the residues lining the interior of the PAS fold appear capable of directly interacting with a small metabolite to trigger PAS signaling (Amezcua et al., 2002), and in others a cofactor such as FAD, FMN, or heme is needed to transduce fluctuations in redox, light, or gas concentration into a biologically useful signal (Harper et al., 2004; Kurokawa et al., 2004; Key et al., 2007).
Although the presence of the PAS domain implies that RsbP alone could sense and transmit signals of energy stress, the RsbQ protein is also necessary for the response (Brody et al., 2001). Both are encoded in the bicistronic rsbQP operon, and the proteins interact in the yeast two-hybrid system. However, this interaction is insufficient for co-immune precipitation from B. subtilis cells or affinity pull-downs of the co-expressed proteins in E. coli cells (M. S. Brody, unpublished), suggesting a transient rather than long-lived association. From sequence analysis RsbQ is a hydrolase of the α/β fold family, members of which mediate a wide variety of reactions (Nardini and Dijkstra, 1999). Notably, residues comprising the proposed catalytic triad within RsbQ are essential for the energy stress response (Brody et al., 2001), leading us to hypothesize that RsbQ activity modifies either (i) a small molecule signal of energy stress or (ii) RsbP itself, possibly through maturation of an associated cofactor. Kaneko et al. (2005) have solved the crystal structure of RsbQ, confirming our earlier assignment of the catalytic triad. Based on their structural analysis, Kaneko and his colleagues suggested that RsbQ acts upon a small hydrophobic molecule, the identity of which remains unknown.
In the absence of a candidate substrate for the RsbQ hydrolase or firm knowledge of the signal that activates the RsbP phosphatase, we used a genetic approach to establish the relationship between the two proteins, and to map the dependencies of the signaling pathway onto regions of RsbP. This approach relied upon isolation of suppressor alterations within RsbP that bypassed the requirement for RsbQ function. Because close homologues of the RsbQ hydrolase and the RsbP PAS domain are jointly associated with different output domains in other bacteria (E. V. Nadezhdin, M. S. Brody and C. W. Price, in preparation), our results shed light on the operation of a conserved and widely distributed sensing module. Moreover, our results implicate a predicted α helix immediately N-terminal to the recognized PP2C domain in controlling phosphatase activity. In other representatives of the multidomain subfamily that includes RsbP, this helix is conserved in fusions with diverse input domains, suggesting a common mechanism linking signal sensing with control of PP2C activity.
A PAS sensory domain and a PP2C effector domain within RsbP were predicted previously (Vijay et al., 2000). Before beginning the suppression study we defined the boundaries of these domains by sequence analysis, finding a third region connecting them in the order PAS-coiled-coil-PP2C (Fig. 2A)
The requirement of RsbQ hydrolase activity for stress activation suggests a model in which RsbQ acts on a signal or sensor of energy flux. This model predicts that RsbP alterations which bypass the need for RsbQ could identify elements important for conveying the signal or controlling phosphatase activity. Because RsbQ has a positive role in the signaling pathway, we anticipated that most of the bypass suppressors would affect inhibitory regions.
We developed a plate screen that relied on the observation that overexpression of wild type RsbP artificially invokes the energy stress response, measured using a σB-dependent lacZ reporter fusion, even in energy-sufficient, exponentially-growing cells (Brody et al., 2001). By contrast, overexpression of wild type RsbP in an rsbQ null strain background does not induce the response. We began with a medium-copy plasmid that places the entire rsbP coding region under control of the inducible Pspac promoter, as previously described (Brody et al., 2001). This plasmid was randomly mutagenized by passage through an E. coli mutDST strain to create three independent pools, which were then transformed into a B. subtilis strain bearing an rsbQ null allele. When plated on selective medium containing XGal and inducer, colonies containing presumptive RsbQ-independent alterations appeared blue against a background of white.
Plasmids were recovered from 17 blue candidates, and their rsbP coding regions were examined by DNA sequence analysis. All 17 plasmids contained single missense changes within rsbP, and these mapped in three discrete regions (Fig. 2B). Four mapped to the coiled-coil region; two mapped within the α0 helix; and eleven mapped to the α1 helix or its near environs in β2 and α2 (with P242S and R246C each recovered twice, from different pools). Because four of the 17 independently isolated alterations were recovered more than once, we infer that the screen was beginning to approach saturation. To ensure that the RsbQ-independent phenotype depended upon the observed alterations in rsbP, the coding region of each candidate was recloned into the unmutagenized vector before further characterization.
An overexpression assay in buffered Luria Broth (BLB) quantified the increase in reporter expression elicted by each recloned alteration. Figs. 4A and B show the results for representative suppressors from each of the three discrete regions identified: A132T, N181S and R246C. Both sets of assays were done using cells lacking RsbQ, and growing exponentially so they were not subjected to energy stress. These assay conditions ensured that activity of the RsbW kinase was held constant (Alper et al., 1996; Delumeau et al., 2002), and therefore reflected only differences in activity of the overexpressed RsbP phosphatases.
The assays shown in Fig. 4A were done in the same genetic background as the original screen, with an in-frame deletion of the chromosomal copy of rsbQ that leaves the downstream rsbP intact. When an additional, plasmid-borne copy of wild-type rsbP was overexpressed in this background, the absence of rsbQ prevented induction of the reporter fusion (Fig. 4A). By contrast, when an RsbQ-independent form of rsbP was overexpressed, fusion activity increased 20-100 fold, depending on the alteration. This increase was less than that of the positive control, in which the plasmid-borne copy of wild-type rsbP was overexpressed in a background containing an intact chromosomal copy of rsbQ.
We next asked whether the phenotypes of the RsbQ-independent rsbP alleles on the plasmid were dependent on a wild type copy of rsbP on the chromosome. These assays were conducted in a genetic background that deleted both rsbQ and rsbP from the chromosome, with the expectation that the data would now solely reflect the activity of the plasmid-borne form. As shown in Fig. 4B, the results for overexpression of the wild type and RsbQ-independent forms of rsbP were much the same as in Fig. 4A. Thus the RsbQ-independent alleles manifested the same phenotype in the presence or absence of wild type rsbP. We chose to conduct all remaining assays with wild type rsbP on the chromosome, as was done for the experiments shown in Fig. 4A.
Fig. 4C summarizes the reporter gene activity elicited by overexpression of the RsbQ-independent versions of rsbP in the absence of rsbQ, revealing a diversity of suppressor phenotypes. We tested some of the weaker suppressors (N124D, A129T, A136T, N181S and R246H) and found they no longer manifested an obvious plate phenotype after transfer from the primary mutagenized plasmid, even though they recorded increases in reporter expression by liquid assay. These likely came through the screen in part due to the presence of a second mutation in the original vector that increased rsbP expression beyond that of the parent plasmid (data not shown). By contrast, some of the stronger suppressors (A132T, G230R, R246C) retained their plate phenotype after transfer. The clustering of these 15 varied RsbQ-independent alterations implicates the three regions they identify in controlling RsbP activity.
If the RsbQ hydrolase acts on a signal or sensor of energy stress, then the PAS domain is the likely element to recognize the signal, or to interact with the sensor. Therefore, an intriguing outcome of our genetic screen was the absence of RsbQ-independent suppressor alterations mapping within the PAS domain itself. Such suppressor alterations also failed to emerge from a subsequent screen that specifically targeted the PAS domain with error-prone mutagenesis (E. V. Nadezhdin and C. W. Price, unpublished). One explanation for this result is the PAS domain, like RsbQ, might have a positive role in the signaling pathway. Gain-of-function alterations to activate a positive element would likely be rare.
To directly test the regulatory role of the PAS domain, we made two different null mutations within rsbP and exchanged each for the wild-type allele on the B. subtilis chromosome. As shown in Fig. 5A, a large in-frame deletion within the PAS domain abolished energy stress response, with a phenotype closely resembling that of an rsbQ null allele. This deletion removed residues 8-108 of the domain. An N25A missense alteration, expected to critically disrupt the cap of the first PAS helix (Pellequer et al., 1998), had an equally negative effect. These loss-of-function phenotypes were not due to a large decrease in RsbP protein levels in the altered strains. In a Western blot, the strain bearing the large in-frame deletion within the PAS domain had a stronger RsbP signal than wild type, whereas that bearing N25A had a slightly weaker signal than wild type (Fig. 5A, insert).
These results are consistent with a model in which both RsbQ and the PAS domain are required for signal recognition, with the PAS domain communicating this signal to other regions of RsbP. Such a model predicts that suppressor alleles isolated on the basis of their RsbQ-independent phenotype would also bypass the need for PAS function. To test this prediction, we combined the in-frame deletion of the PAS domain with representative RsbQ-independent alterations affecting the coiled-coil, α0 and α1 regions of RsbP – A132T, N181S and R246C. As shown in Fig. 5B, the mutant form of RsbP bearing the PAS deletion manifested the same loss-of-function phenotype in the overexpression assay as it did in single copy on the chromosome (Fig. 5A). Significantly, each of the RsbQ-independent alterations tested restored activity to rsbP constructions bearing the PAS deletion. The activity of these suppressed forms of RsbP was in some cases even greater than the wild type control, a feature we attribute to the apparently higher protein levels in derivatives carrying the PAS deletion (see Fig. 5A insert for effect on RsbP levels in single copy and Fig. 6A insert for effect in multicopy). The experiment shown in Fig. 5B was done in an rsbQ+ genetic background. To eliminate the possibility that suppression occurred via an rsbQ-dependent pathway, we repeated the experiment in an rsbQ null background and observed similar results (Fig. 5C). We conclude that suppressors isolated on the basis of their ability to bypass loss of RsbQ function also bypassed loss of PAS function. From the phenotypes of the loss-of-function and suppressor alleles, we infer that RsbQ and the PAS domain are positive elements jointly required for signal perception, with the coiled-coil, α0 and α1 regions exerting their influence downstream.
We next used the overexpression assay to determine the effects of a large, in-frame deletion that removed the PAS domain and the coiled-coil, producing a construct that lacks residues 8-168. As a result, this form of RsbP consists of the α0 helix together with the recognized core of the PP2C domain. As shown in Fig. 6A, deletion of the coiled-coil region reversed the loss-of-function phenotype caused by deletion of the PAS domain. We conclude that the coiled-coil acts as a negative element, and that loss of coiled-coil function overrides loss of PAS function.
The high activity of the form of RsbP missing the PAS and coiled-coil regions allowed a further test of the system. The previous finding that three RsbQ-independent alterations also suppress the loss of PAS function (Figs. 5B and 5C) strongly suggests that RsbQ and the PAS domain functionally interact. We therefore asked whether a reciprocal suppression supports this relationship. The results in Fig. 6A indicate that loss of coiled-coil function overrides loss of PAS function, implying that loss of the coiled-coil would also override loss of RsbQ. This is indeed the case. As shown in Fig. 6B, removing the PAS and the coiled-coil regions from RsbP rendered the remaining portion RsbQ-independent. Together our genetic results support a model in which RsbQ and the PAS domain form a sensory input module that positively controls phosphatase activity by counteracting the negative effect of the coiled-coil.
We then investigated the effects of an in-frame deletion that removed the PAS domain, the coiled-coil and the α0 helix, producing a construct that lacks residues 8-188. This form of RsbP represents the PP2C core alone. In contrast to the high activity of the form lacking the PAS and coiled-coil regions, the additional removal of the α0 helix greatly decreased the response in the overexpression assay (Fig. 6A). However, this truncated form of RsbP accumulated at significantly reduced levels in cell extracts (Fig. 6A, insert). Thus, these results alone do not allow us to distinguish whether the α0 helix has a role in controlling activity of the adjacent PP2C core or is primarily important for protein stability.
In an attempt to address this issue, we added the RsbQ-independent R246C alteration to the α1 helix-coding region of the construct that deleted PAS, coiled-coil and α0. Any suppression of the loss-of-function phenotype would provide additional evidence that α0 plays a role in controlling phosphatase activity. However, the activity of this construct was indistinguishable from that of the original PAS-coiled-coil-α0 deletion (data not shown). It appears that protein levels were so reduced by the absence of the α0 helix that any contribution from the R246C alteration was undetectable. Nonetheless, the existence of the RsbQ-independent E173K and N181S alterations within the α0 helix (Fig. 2B), coupled with a broad phylogenetic distribution of the helix (Fig. 3), support the inference that it is important for the proper functioning of the adjacent core of the PP2C domain.
In order for RsbP to respond to changes in energy levels, it must sense and communicate a signal to the phosphatase domain. Our genetic results can be summarized in the following scheme:
We propose that a molecule transformed by the RsbQ hydrolase interacts with the PAS domain, which is a positive element that controls phosphatase activity via the coiled-coil. This proposal is consistent with the strong inference that RsbQ activity is required for signaling (Brody et al., 2001), and with the structural analysis that suggests the RsbQ substrate is a small hydrophobic molecule (Kaneko et al., 2005). It is also consistent with a transient rather than long-lived interaction between RsbQ and RsbP (Brody et al., 2001; M. S. Brody, unpublished). Because RsbQ and the PAS domain are both positive elements (Fig. 5), our present genetic experiments do not explicitly establish their order of action. However, since PAS domains often bind small molecules and signal to adjacent regions of the same protein, the order of RsbQ → PAS is the simplest interpretation of the data.
Deletion analysis has shown that the coiled-coil region is a negative element in the absence of the PAS domain (Fig. 6), and the sum of our results supports the hypothesis that PAS acts by countering the negative effect of the coiled-coil. Loss of the positive input from either RsbQ or the PAS domain can be ameliorated via bypass alterations within the coiled-coil, or within the α0 or α1 regions of the PP2C domain (Figs. (Figs.44 and and5).5). The phenotype of the deletion that removes both the PAS and coiled-coil further indicates that these elements are not required for phosphatase activity but instead serve a regulatory role. Thus a plausible signaling model involves action at a distance: a binding event within PAS causes a conformational change that is communicated to the adjacent coiled-coil, relieving its negative action on the phosphatase domain. In this model, the function of the coiled-coil is to hold the phosphatase domain inactive until a signal is received. A similar role has been proposed for the S-helix, a predicted coiled-coil that connects a variety of input and output domains in bacterial, fungal and animal signaling proteins (Anantharaman et al., 2006). Here we have experimentally linked this role to a coiled-coil unrelated to the S-helix, suggesting a wider application of the design.
The isolation of bypass alterations within the α0 and α1 helices indicates that these elements influence phosphatase activity in vivo. To estimate the positions of the target residues within the PP2C domain, we constructed a computation model using the i-TASSER server, which relies on threading alignment, ab initio modeling and iterative structural refinement (Zhang, 2008). The model shown in Fig. 7 depicts the core domain (residues N189-V402) threaded against the structures of the two bacterial PP2C phosphatases (Rantanen et al., 2007; Schlicker et al., 2008) that were chosen by the server as the best available templates. Such a model is not a substitute for an experimentally determined structure, but it can be useful in assessing key features. In this regard, the α helices and β strands are likely well represented, as indicated by the favorable confidence score of −0.07, which suggests that, on average, the alpha carbons lie within about 5Å of their true positions (Zhang, 2008). Both threading templates belong to Subgroup I of the bacterial PP2C phosphatases and therefore lack an α0 helix, which is not included in the model.
In Fig. 7 the invariant aspartate residues that coordinate divalent metals in the active site are shown as red spheres, and the positions of the five strongest bypass suppressors are shown as magenta sticks within β2 and α1. These bypass residues are distinct from the active site, and all are concentrated in one region of the PP2C domain. Even if the register of the α1 helix differs by as much as half a turn in the true structure, alterations at each of the five residues potentially affect the same surface. At a minimum, these suppressors define sites of allostery revealed by genetic alteration (Goodey and Benkovic, 2008). Parsimony therefore suggests that this region of the PP2C domain is involved in the signaling process. Moreover, in the model shown in Fig. 7, the α0 helix joins the domain at a site distant from the active site and cannot extend to occlude it. Therefore, the bypass suppressors within α0 also define sites of allostery. We hypothesize that the α0 and α1 helices are the recipients of signals from the PAS and coiled-coil regions, serving as conduits that ultimately control phosphatase activity.
The isolation of bypass suppressors within the α0 helix indicates its importance to RsbP function. Based on the widespread occurrence of similar helices preceding the PP2C domains of many prokaryotic phosphatases belonging to Subfamily II, we suggest that the α0 helix plays a general role in controlling phosphatase activity. A BLAST search using the α0 helix and PP2C domain of RsbP revealed more than 500 similar sequences that possess such a helix (see Fig. 3 for an alignment of representative examples). These predicted multidomain phosphatases are found in bacteria from varied lineages as well as in some archaea, and they possess diverse input domains capable of sensing internal or external signals.
Figs. 8A and B show schematic depictions of three well-characterized examples from partner switching networks in Bacilli. In each example, an α0-PP2C domain is joined to one of the three most commonly observed input domains. The first is our focus here, the RsbP energy-signaling phosphatase of B. subtilis, in which an N-terminal PAS domain is linked via a predicted coiled-coil to the α0 helix and PP2C core (Fig. 8A). A similar domain configuration is found in a few other Bacilli, and for organisms classified as Nocardia, Mycobacterium and Streptomyces species. The second example is the RsbU environmental-signaling phosphatase of B. subtilis, which, like RsbP, dephosphorylates RsbV-P to activate the general stress response (Fig. 1). Here the N-terminal RsbU_N domain provides an essential binding determinant for the RsbT regulator (Fig. 8B), but the mechanism by which binding activates the C-terminal PP2C domain remains unexplained (Delumeau et al., 2004). This configuration is restricted to Bacillus species and close relatives, including organisms that lack RsbT orthologs, such as members of the genus Staphylococcus (Fig. 3). The third example is the RsbY phosphatase from B. cereus and its relatives; RsbY also activates a general stress response (van Schaik et al., 2005). Here the N-terminal region contains a receiver domain, presumably modified by a histidine protein kinase (Fig. 8B). In our search, receiver domains were frequently linked to the PP2C output, with examples from Bacteroides, Bradyrhizobium, Geobacillus, Leptospira, Nostoc, Pseudomonas, Rhodopseudomonas, Vibrio and Yersinia species, among others (not shown).
Based on the combined genetic and bioinformatic analysis, we suggest that the α0 helix provides communication between different sensory domains and the PP2C catalytic domain of Subfamily II enzymes. Study of the mechanism coupling input and output within RsbP may therefore be relevant to understanding the control of PP2C phosphatase activity in diverse signaling pathways.
Our data indicate that both RsbQ and RsbP-PAS are positive elements in the energy-signaling pathway, acting upstream from the coiled-coil and PP2C domains. These genetic results lead us to hypothesize that RsbQ and RsbP-PAS form a sensory input module, in which the RsbQ hydrolase provides a small molecule required for PAS signaling. This hypothesis gains additional support from a bioinformatic analysis that finds RsbQ and RsbP-PAS orthologs encoded by adjacent genes in a wide variety of bacteria, with the PAS domain fused to diverse output domains. We present the details of this analysis elsewhere (E. V. Nadezhdin, M. S. Brody and C. W. Price, in preparation). Here in Figs. 8A and C we show three examples in which the RsbQ-PAS module is joined to one of the three output domains with which it is commonly associated. The first is the RsbQ-RsbP pair under consideration here, with the input module joined to a PP2C output domain (Fig. 8A). The second is from Pseudomonas putida of the γ proteobacteria, with the module linked to a histidine protein kinase domain (Fig. 8C). And the third is from Kineococcus radiotolerans of the Actinobacteridae, with the module linked to diguanylate cyclase and phosphodiesterase domains (Fig. 8C).
We propose that RsbQ and RsbP-PAS jointly sense a common input signal in phylogenetically diverse bacteria. This input is then coupled with different output domains to regulate cellular processes that – outside of Bacillus subtilis – remain largely uncharacterized. What is the nature of this input signal that appears to be significant to the physiology of many different bacteria? In Bacillus subtilis the phenotypes of rsbQ and rsbP null alleles suggest that their products detect a change in energy flux, but this need not be the case. Of the two input phosphatases in the general stress signaling network, RsbP is the more active against the RsbV-P anti-anti-σ factor in exponentially growing cells, and is therefore responsible for setting the basal level of σB activity under these conditions (Eymann and Hecker, 2001). It is not yet established whether RsbP activity is rapidly regulated by a change in energy flux per se, or whether RsbP provides a more slowly modulated point against which the activity of the RsbW kinase rises or falls as cells experience energy sufficiency or depletion (Vijay et al., 2000). In the event of slow modulation, the RsbW kinase may provide the primary signal of energy stress (Alper et al., 1996; Delumeau et al., 2002). If this is the case, the RsbQ-PAS input module could adjust the activity of RsbP in response to another signal that serves to set the basal activity of the σB transcription factor. Identification of the small molecule that binds RsbP-PAS should help resolve this issue, and reveal the signal that the module senses in diverse bacteria.
B. subtilis strains are shown in Table 1; the plasmids used for these constructions are in Table 2. Standard recombinant DNA methods and natural transformation of B. subtilis were as previously described (Brody et al., 2001).
Random mutagenesis of rsbP was performed by transforming the multicopy expression plasmid pKV2 into the DNA repair-deficient E. coli strain XL1-Red (Stratagene, La Jolla, CA), according to the manufacturer's protocol. pKV2 contains a 1.2 kb SalI-SphI fragment bearing the rsbP coding region and its ribosomal binding site, under control of the IPTG (isopropyl-β-D-thiogalactoside) inducible Pspac promoter of pDG148 (Brody et al., 2001). Transformants were grown overnight at 37°C in Luria broth, diluted 1:100 in fresh medium, and grown again for a total of five overnight cycles. Plasmid DNA was extracted and used to transform B. subtilis strain PB605 (rsbQΔ2 amyE::ctc-lacZ trpC2). To screen for RsbQ-independent suppressor alterations, transformants were plated on tryptose blood agar base plates (Becton Dickson, Sparks, MD) containing 5 μg/ml neomycin, 160 μg/ml X-Gal (5-bromo-4-chloro-3-indolyl-β-D-galactopyranoside), and 1 mM IPTG. Blue colonies were isolated, with the plasmid recovered and transformed again into PB605 to confirm that the blue color was plasmid-linked. DNA sequencing located changes that led to amino acid substitutions within the rsbP coding region. The 1.2 kb SalI-SphI fragments encoding rsbP in candidate plasmids were re-cloned into the unmutagenized pDG148 vector before further characterization.
Two different mutations were constructed within the PAS coding region of the chromosomal rsbP gene. To make an in-frame deletion, we used the four-primer method of site-directed mutagenesis (Ho et al., 1989) to remove 303 bp within rsbP, deleting triplets 8 to 108, as confirmed by DNA sequencing. This allele (rsbPΔ3) was substituted for wild type on the B. subtilis chromosome by I-SceI mediated allelic exchange (Janes and Stibitz, 2006), using plasmids pMB5947 and pSS4332; substitution was confirmed by PCR. Introduction of a σB-dependent ctc-lacZ reporter fusion into this mutant background yielded strain PB1069. The strain bearing the rsbPN25A allele was likewise constructed using the four-primer method and two-step allele replacement (K. Vijay, PhD thesis, UC Davis 2001). It was kindly provided by Kamni Vijay, and designated PB623. DNA sequence analysis confirmed that the desired N25A allele had been introduced into the rsbP region.
Three different in-frame deletions were constructed within the rsbP coding region on the pDG148 multicopy expression plasmid. Four-primer mutagenesis was used to make deletions with the same 5′ end, but with different 3′ ends that removed an increasing number of domains and regions. The outer PCR primer pair in all three clones was the same as used previously to amplify wild-type rsbP along with its predicted ribosomal binding site (Brody et al., 2001). The in-frame deletion of the PAS domain was identical to the rsbPΔ3 allele described above, i.e., we removed 303 bp from within the 1209 bp coding region, deleting triplets 8 to 108. To make the larger in-frame deletion of the PAS and coiled-coil regions, we removed 483 bp, deleting triplets 8 to 168. Finally, to make the largest in-frame deletion of the PAS, coiled-coil and α0 regions, we removed 543 bp, deleting triplets 8 to 188. Each final PCR product was cloned into pDG148 to yield pMB5918, pMB5919, and pMB5920, respectively, and the sequences were verified. We also introduced the A132T, N181S, or R246C suppressor mutations into pMB5918, using the QuikChange Lightning site-directed mutagenesis kit (Stratagene, La Jolla, CA) according to the manufacturer's instructions. The 0.9 kb SalI-SphI fragments containing the rsbPΔ3 allele together with each of the point mutations were re-cloned into unmutagenized pDG148 vectors before further use. These overexpression clones were transformed into strains of B. subtilis that bore the ctc-lacZ reporter fusion (Table 1).
Shake cultures were grown at 37°C to mid-exponential phase in buffered Luria broth (BLB) lacking salt (Boylan et al., 1993), then diluted 1:25 into fresh BLB medium. This second culture was treated in two different ways, depending on the experiment. First, to artificially induce expression of wild or mutant rsbP under Pspac control in the pDG148 vector, 1 mM IPTG was added during the early exponential phase of growth. In all such overexpression experiments, neomycin was included (2.5 μg/ml final concentration) throughout to promote plasmid retention. Second, to elicit the conditions that normally trigger the energy stress response, cells were allowed to grow until they had entered stationary phase. For both kinds of assays, samples were collected at the times indicated and treated according to Miller (1972), as previously described (Brody et al., 2001). Protein levels were determined using the Bio-Rad Protein Assay reagent (Bio-Rad Laboratories, Hercules, CA). Activity was defined as ΔA420 × 1,000 per minute per mg of protein.
Rabbit anti-RsbP antibody was made by the UC Davis Animal Resources Service, using purified recombinant protein we supplied (Vijay et al., 2000). Specificity was tested by Western blot analysis of wild and mutant cell extracts. To compare steady-state levels of RsbP and its mutant forms, cells were harvested either one generation (20 min) before the onset of stationary phase or one generation after IPTG addition, depending on the experiment. Cells were broken by sonication; total cell protein was separated by SDS-PAGE, transferred to polyvinylidene difluoride membranes (Bio-Rad Laboratories) and treated as previously described (Brody et al., 2001). Briefly, after exposure to primary antibody, membranes were washed and incubated with IgG peroxide-conjugated anti-rabbit secondary antibody (Santa Cruz Biotechnology Inc., Santa Cruz, CA). Bound antibody was detected using the ECL Plus Western Blotting kit (GE Healthcare Bio-Sciences, Piscataway, NJ) according to the manufacturer's instructions
We thank Scott Stibitz for his kind gift of plasmid pSS4332, William Burkholder for providing pUS19, Tanya Gaidenko for sharing pTG5916, and Kamni Vijay for constructing strains PB622 and PB623. This research was supported by Public Health Service grant GM42077 from the National Institute of General Medical Sciences.