The General Stress Response σS Is Regulated by a Partner Switch in the Gram-negative Bacterium Shewanella oneidensis*

Here, we show that a partner-switching system of the aquatic Proteobacterium Shewanella oneidensis regulates post-translationally σS (also called RpoS), the general stress response sigma factor. Genes SO2118 and SO2119 encode CrsA and CrsR, respectively. CrsR is a three-domain protein comprising a receiver, a phosphatase, and a kinase/anti-sigma domains, and CrsA is an anti-sigma antagonist. In vitro, CrsR sequesters σS and possesses kinase and phosphatase activities toward CrsA. In turn, dephosphorylated CrsA binds the anti-sigma domain of CrsR to allow the release of σS. This study reveals a novel pathway that post-translationally regulates the general stress response sigma factor differently than what was described for other proteobacteria like Escherichia coli. We argue that this pathway allows probably a rapid bacterial adaptation.

To adapt and survive to environmental changes or stresses, bacteria set up specific or general strategies. Among them, the general stress response (GSR) 3 has been deeply studied in Escherichia coli (1)(2)(3). When E. coli is submitted to stressful life conditions including nutrient starvation in stationary phase, osmolarity variations, or pH modifications, the alternative transcriptional sigma factor of the GSR, S (also called RpoS), allows a reorganization of gene expression by controlling the S regulon that contains more than 500 genes. The S regulon comprises genes involved in stress adaptation, biofilm formation, motility, and sometimes secretion of virulence factors (1, 4 -8). The availability of S in the cell is highly regulated at the transcriptional, translational, and post-translational levels (2,8,9). When bacteria encounter stressful events, the synthesis of S increases and S accumulates in the cell. The level of S decreases when regular conditions come back. At the post-translational level, the regulation of S is a sophisticated mechanism mediated by the ClpXP proteolytic machinery. To be degraded, S is addressed to ClpXP by the adaptor protein RssB (10 -14). RssB is an orphan response reg-ulator organized in two domains: a receiver domain and a phosphatase-like domain that is not functional in E. coli (15). RssB can be phosphorylated, but the role of this phosphorylation is not well understood yet (16). Moreover, another layer of regulation involves the anti-adaptor proteins IraD, IraM, and IraP that are synthesized in specific stress conditions when S is needed (17)(18)(19)(20). These proteins interact with RssB and prevent the degradation of S , leading thus to S accumulation in the cell. Surprisingly, except for enterobacteria, the post-translational regulation of S is unknown in other Gram-negative bacteria (21)(22)(23).
In the Gram-positive bacterium Bacillus subtilis, the factor controlling the GSR, B , is post-translationally regulated by a canonical partner-switching regulatory system that controls the release or the sequestration of the sigma factor according to the environmental conditions (24 -26). Partner-switching systems are one among other mechanisms by which bacteria connect cue transmission and gene expression (27)(28)(29). The partner-switching system that regulates B consists of an antisigma factor with a kinase activity (RsbW), two phosphatases (RsbU and RsbP), and an anti-sigma antagonist (RsbV) that contains a phosphorylatable serine residue in its sequence. In the absence of specific signal, the anti-sigma partner both phosphorylates the anti-sigma antagonist and sequesters the sigma factor (RsbW-B ), hampering its association with the RNA polymerase and, consequently, preventing expression of its regulon. When a stress or a specific signal arises, the antagonist protein is dephosphorylated by the phosphatases and interacts with the anti-sigma factor, leading then to the release of the sigma factor. In turn, the latter can thus associate with the RNA polymerase and can induce expression of its regulon. In this mechanism, the availability of the sigma factor depends on the phosphorylation state of the anti-sigma antagonist (30,31). Although partner-switching systems were initially unveiled in Gram-positive bacteria, they have also been found in some Gram-negative bacteria including Bordetella (BtrWVU), Vibrio fisheri (SypEA), and Pseudomonas aeruginosa (HsbRA). However, in these bacteria it has not been shown that these systems are involved in the direct regulation of sigma factors. BtrWVU may post-translationally control the type III secretion system, whereas SypEA promotes biofilm formation by an unknown mechanism, with HsbRA regulating swarming motility (32)(33)(34)(35)(36)(37).
Shewanella oneidensis, a Gram-negative Proteobacterium, colonizes aquatic and sedimentary environments that are char-acterized by their versatility in biotic and abiotic stresses. Thus, S. oneidensis is now an environmental model bacterium to study bacterial respiration and adaptation. Although it is described as presenting robust sensing and regulatory systems to respond and survive under harsh conditions, the GSR has not been studied yet in this organism (38,39).
In this study, we establish that contrarily to what was described in E. coli, S. oneidensis controls the availability of the GSR sigma factor, S , and, consequently, the expression of genes of the S regulon by a partner-switching mechanism close to that observed in Gram-positive bacteria. Our results revealed an unsuspected post-translational S regulatory system in Gram-negative bacteria.
The putative function of these proteins suggests that they could be part of a partner-switching regulatory system involved in the post-translational control of a sigma factor. Our working model is that the CrsR anti-sigma factor interacts with a dedicated sigma factor to block its transcriptional activity when it is not required, whereas under appropriate conditions, the CrsA anti-sigma antagonist binds CrsR, and the sigma factor is thus released. If true, this model raises essential questions including (i) Which sigma factor is targeted by the system? (ii) Are the anti-sigma domain of CrsR and the CrsA anti-sigma antagonist involved in the sequestration/release of this sigma factor? (iii) Is CrsR responsible for both the phosphorylation and the dephosphorylation of CrsA?
CrsR Interacts in Vitro with S -By analogy with B of B. subtilis, we hypothesized that S could be a target of the potential CrsR-CrsA partner switch. To test the role of the partnerswitching mechanism on the post-translational regulation of S , an in vitro approach was undertaken, and the successive steps leading to establish S regulation were detailed. With partner-switching regulation being based on sequestration and release of the controlled target, a potential binding between the anti-sigma domain of CrsR (CrsR D3 ) and S was investigated by coelution experiments with purified proteins. In these assays, His-tagged-CrsR or -CrsR D3 was incubated with Strep-tagged S , and CrsR or CrsR D3 was retained on cobalt beads. Coomassie Blue staining showed that S coeluted with CrsR and CrsR D3 , suggesting that the binding was mediated by the D3 anti-sigma domain of CrsR ( Fig. 2A).
We confirmed this result by a bacterial two-hybrid approach. In this experiment, S and CrsR D3 were fused to the T18 and T25 domains of the adenylate cyclase of Bordetella pertussis, respectively, and produced in an E. coli strain (43). If the two proteins interact, binding of the T18 and T25 domains of the adenylate cyclase occurs and leads to the expression of the lacZ reporter gene. Significant level of ␤-galactosidase activity was measured when T18-S and T25-CrsR D3 were produced together in E. coli, corroborating that S and CrsR D3 interact (Fig. 2B). Background levels of ␤-galactosidase activity were measured when the fusion proteins were produced alone.
Finally, we performed BioLayer Interferometry (BLI) experiments using S and His-tagged CrsR D3 and calculated a K D of 1.7 M with k d ϭ 0.00046 (Ϯ 6.9 ϫ 10 Ϫ6 ) s Ϫ1 and k a ϭ 260 Ϯ 9.7 M Ϫ1 s Ϫ1 (Fig. 2C). Altogether, these results clearly show that S is a target of the anti-sigma domain of CrsR.
CrsA Interacts with CrsR-To analyze the partner-switching mechanism, we tested whether CrsR D3 also interacts with CrsA. Indeed, CrsA is homologous to anti-sigma antagonists, and therefore it should interact with the anti-sigma domain CrsR D3 . We first looked for interaction by cross-linking exper-   (24,37,40,42). These conserved residues are indicated by D*. D, sequence alignment of the third domain of CrsR (CrsR D3 ) of S. oneidensis and the antifactor RsbW of B. subtilis, the anti-factor first domain of the response regulator SypE of V. fisheri, and the anti-factor third domain of the response regulator HsbR of P. aeruginosa. CrsR D3 GHKL ATPase domain contains the conserved N box residue Asn 460 , and the conserved G1 box residues Asp 517 and Gly 519 . The conserved ATP binding Bergerat-fold N, G1, and G2 boxes are framed, and the conserved residues required for serine kinase activity are indicated by N* (24,37,40,41). In A, C, and D, the sequence alignments were performed using Clustal Omega website. Black and gray boxes correspond to identical or similar residues, respectively, by comparison with Crs proteins.
iments. For that, CrsR D3 and CrsA were produced as Histagged recombinant proteins in E. coli and purified to homogeneity. The two proteins were incubated either together or alone and, in each case, in the presence or in the absence of EDC, a chemical cross-linker. After SDS-PAGE followed by Western blotting, the proteins were visualized using a His tag antibody (Fig. 3A). An additional band at ϳ30 kDa was observed when both CrsA and CrsR D3 were incubated together with the crosslinker, indicating that the two proteins interact (Fig. 3A, lane 6). Moreover, this experiment shows that CrsR D3 dimerizes because a supplementary band was observed when CrsR D3 alone was incubated with the cross-linker (Fig. 3A, lane 4). A bacterial two-hybrid approach was performed to confirm this interaction. To do that, CrsA and CrsR D3 were fused to the T18 and T25 domains of the adenylate cyclase of B. pertussis and produced in E. coli (43). High level of ␤-galactosidase activity was measured when T18-CrsA and T25-CrsR D3 were produced together in E. coli, corroborating that CrsA and CrsR D3 interact (Fig. 3B). In addition, this experiment also confirms that CrsR D3 dimerizes. Using BLI approach, we calculated a K D of 0.29 M with k d ϭ 0.00082 (Ϯ 7 ϫ 10 Ϫ6 ) s Ϫ1 and k a ϭ 2900 Ϯ 5.7 M Ϫ1 s Ϫ1 (Fig. 3C). Altogether, these experiments demon-strate that the D3 domain of CrsR interacts with CrsA, as observed in partner-switching systems.
Kinase and Phosphatase Activities of CrsR toward CrsA-As mentioned above, the canonical partner-switching mechanism involves transient phosphorylation of the anti-sigma antagonist protein by the anti-sigma protein. To investigate a potential phosphorylation of CrsA by CrsR D3 , purified proteins were incubated either alone or together in the presence of ATP. The samples were then analyzed by phosphate affinity SDS-PAGE (Phos-Tag TM ) (Fig. 4A). In this experiment, the migration of a phosphorylated protein is slowed down and results in an upper shift of the band. We observed that the band corresponding to CrsA was up-shifted in the presence of CrsR D3 and ATP, indicating that CrsA is phosphorylated by CrsR D3 in the presence of ATP. We also ran similar samples on native PAGE. Although CrsR D3 migrates as a ladder (Fig. 4B, lanes 1 and 2), we observed that in the presence of CrsR D3 and ATP, phosphorylated CrsA (CrsA-P) migrates as a down-shifted band compared with the non-phosphorylated protein (Fig. 4B, compare lane 14 with lanes 7, 8, and 13). It is noteworthy that in both Phos-Tag TM and native PAGE experiments, almost all CrsA present in the assay was phosphorylated by CrsR D3 as shown by the disap- pearance of the non-phosphorylated protein band. We then checked by native PAGE analysis if the full-length CrsR protein was also able to phosphorylate CrsA. As seen on Fig. 4C (lane 4), the band corresponding to CrsA was down-shifted when CrsA was incubated with full-length CrsR and ATP, confirming that CrsR can phosphorylate CrsA. As indicated above, the kinase activity of CrsR toward CrsA is very efficient because in this experiment there was an 8-fold excess of CrsA concentration compared with CrsR (Fig. 4C, lane 4).
According to sequence alignments with other anti-sigma antagonist proteins, phosphorylation of CrsA should occur on residue Ser 49 (Fig. 1A). To confirm this hypothesis, CrsA was extracted after electrophoresis both in its native and phosphor-ylated states and was submitted to trypsin digestion followed by MALDI-TOF mass spectrometry (Fig. 5, A and B). We observed that only the peptide containing residue Ser 49 shifted in a mass equivalent to a phosphate group (ϳ80 Da) in the phosphorylated protein compared with the native protein. This shift was reversed after alkaline phosphatase treatment (Fig. 5C). Finally, because two successive serine residues were present in this peptide, residue Ser 49 of CrsA was mutated either in alanine to prevent phosphorylation or in aspartate to mimic phosphorylation, and the mutant proteins were purified. As seen in Fig.  4B, CrsA-S49A alone migrates on a native PAGE like the wild type protein (compare lanes 3 and 7). However, after incubation with CrsR D3 and ATP, we did not observe a shift of the band corresponding to CrsA-S49A as seen with the wild type protein, indicating that the mutation actually prevents phosphorylation (Fig. 4B, lanes 10 and 14). In contrast, even in the absence of CrsR D3 and ATP, CrsA-S49D migrated between CrsA and CrsA-P as expected with this mutation (Fig. 4B,  compare lanes 5, 6, 11, and 12 with lanes 13 and 14). These  experiments confirm that CrsA is phosphorylated on residue Ser 49 .
The second domain of CrsR, CrsR D2 , has strong sequence homologies with proteins with phosphatase activity. Although we did not succeed in producing soluble CrsR D2 , we tested whether CrsR can also dephosphorylate phosphorylated CrsA (CrsA-P). Purified CrsA-P was incubated with CrsR in the presence and absence of MgCl 2 , which is necessary for serine phosphatase activity of PP2C type phosphatases (42), and the samples were subjected to native PAGE. We observed that in the presence of CrsR and MgCl 2 , CrsA migrates as the nonphosphorylated protein, clearly showing that CrsR exhibits a phosphatase activity toward CrsA-P (Fig. 4C, compare lane 7 with lanes 5 and 6).
CrsA Phosphorylation State Drives CrsR D3 -CrsA Binding-Together with the binding experiments, the kinase and phosphatase assays strengthen the hypothesis of a partner-switching system involving the anti-sigma domain CrsR D3 , the phosphatase CrsR D2 , and the anti-sigma antagonist CrsA. However, in a partner-switching mechanism, binding between the anti-sigma and the anti-sigma antagonist depends on the phosphorylation state of the anti-sigma antagonist (44). The following experiments clearly showed that CrsA-P cannot interact with CrsR D3 . First, when CrsR D3 and CrsA were incubated with a chemical cross-linker under conditions leading to CrsA phosphorylation, i.e. in the presence of ATP (Fig. 6, lane 1), we did not observe after Western blotting revelation, an additional band corresponding to the complex between CrsA and CrsR D3 , as seen in the absence of ATP (Fig. 6, lane 2). Second, a BLI experiment clearly established that the phosphorylated form of CrsA (CrsA-P) does not bind to CrsR D3 (Fig. 3C). These results indicate that, contrarily to the non-phosphorylated form of CrsA, CrsA-P cannot interact with CrsR.
Finally, we tested the interaction between CrsA and CrsR D3 in two-hybrid assay using the CrsA mutants S49A that abolishes phosphorylation and S49D that mimics phosphorylation. We found that although the interaction is conserved and even increased with CrsA-S49A, it is drastically decreased with CrsA-S49D, confirming that the phosphorylation hampered the interaction between the two partners (Fig. 3B). The conserved Asn residue that is present in the N box of CrsR D3 and homologs is involved in their kinase activity by chelating Mg 2ϩ , allowing then the ATP binding (Fig. 1D) (32). The residue was substituted to produce CrsR D3 N460A variant. Results of twohybrid experiments indicate that the binding of both CrsA and S on CrsR D3 N460A was drastically affected (Figs. 2B and 3B) as observed with other partner-switching systems (45,46). Based on the results provided by the in vitro approach, we postulate that CrsR and CrsA are part of a partner-switching regulatory system that regulates the S factor in S. oneidensis. To confirm this hypothesis, we carried out cross-linking experiments with CrsR D3 and S , in the absence or presence of CrsA, and the Western blot was revealed with antibody targeting the Strep-tag of S . As shown in Fig. 7, two additional bands were observed when both S and CrsR D3 were incubated together with the cross-linker. In the presence of CrsA, one of these bands disappeared, and the second one became fainter, strongly suggesting that CrsA prevents the interaction between S and CrsR D3 .
CrsR-CrsA Partner Switch Controls the Availability of S in Vivo-To test the validity of the results obtained from the in vitro approaches, we investigated the role of these proteins in vivo. In E. coli, it has been shown that the dps gene is under the transcriptional control of S (2,47). Because a dps ortholog (SO1158) is present in the S. oneidensis genome (48), we first checked whether its expression is regulated by S . For that, we constructed in S. oneidensis a fusion between the dps promoter and the lacZ gene in a wild type or a rpoS (the gene coding for S ) deletion strain. ␤-Galactosidase activity was measured during exponential and stationary phases. In the wild type context, ␤-galactosidase activity dramatically increased during the stationary phase compared with the level measured during exponential phase (Fig. 8). In contrast, in the rpoS deletion strain, the activity of the fusion was very low, and it did not change with growth phase (Fig. 8). Thus, as in E. coli, dps of S. oneidensis is under the control of S .
In our hypothesis, CrsA controls the release of the sigma factor by binding or not to CrsR. If true, in the absence of CrsA, the sigma factor could continuously bind to CrsR. We thus introduced the chromosomal fusion into a crsA deletion mutant, and its expression level was analyzed in both mutant and wild type crsA context under appropriate stress conditions. The absence of CrsA prevented the induction of the dps::lacZ fusion under the stationary phase, suggesting the activity of S was stopped in the mutant (Fig. 8). This experiment clearly appoints S as the target of the partner-switching system. Moreover, this result is also in good agreement with the hypothesis that, in the absence of the anti-sigma antagonist CrsA, the sigma factor is sequestered by the anti-sigma domain of CrsR, even under stress conditions. The chromosomal dps::lacZ fusion was then introduced into a crsR deletion mutant, and the ␤-galactosidase activity was measured during exponential and stationary phases. In the crsR mutant, the fusion was expressed during stationary phase at a similar level to that measured in the wild type (Fig. 8). During the exponential phase, the expression level of the fusion was lower in the crsR mutant than in the wild type strain. In agreement with the in vitro results, the absence of CrsR frees S , but our in vivo results suggest that S was not fully protected or active in such a genetic context. A possible explanation is that CrsR could protect S against degradation during the exponential phase.
To ascertain the role of each element in the cascade (i.e. CrsA, CrsR, and S ), we constructed double mutants and surveyed the expression of the dps::lacZ fusion in each context. The crsR-crsA double mutant behaved as the single crsR mutant, meaning that crsR is epistatic to crsA (Fig. 8). The crsR-rpoS double and the rpoS single mutants showed the same pattern of expression of the fusion, indicating that rpoS is epistatic to crsR (Fig. 8). In conclusion, these results strengthen the proposal of CrsR and CrsA regulating the S factor through a partner-switching mechanism and thus support our model (Fig. 9).

Discussion
This work establishes that S of S. oneidensis is post-translationally controlled by a partner-switching system composed of CrsR and CrsA. CrsR contains three domains: the first domain (D1) corresponds to a receiver domain of response regulators, and the second (D2) and the third (D3) domains have, respectively, phosphatase and kinase activities toward the anti-sigma antagonist CrsA. The phosphorylation state of CrsA depends on the modulation of CrsR activities probably in response to environmental cues and controls the release or the sequestration of S in S. oneidensis (Fig. 9). It is noteworthy that the affinity between CrsA and CrsR D3 is stronger than between S and CrsR D3 , as expected for a partner-switching mechanism.
The finding of a partner-switching mechanism in S. oneidensis was quite surprising because, although S from the ␥-proteobacteria S. oneidensis and E. coli are homologous (79% of identity), their post-translational regulation deeply differs. Interestingly, although S from S. oneidensis and B and EcfG are phylogenetically distinct, their post-translational regulation follows a similar mechanism based on a partner-switching system (29,49). Thus, according to our in vitro and in vivo results, the last steps of post-translational regulation of S in S. oneidensis are similar to those of B in B. subtilis. We propose that a mechanism of sequestration and release of an alternative sigma factor confers a selective advantage to bacteria living in highly versatile environments because in this case, response to stresses does not require de novo synthesis of a sigma factor to adapt, and it thus allows a rapid answer with a low energy cost. In E. coli, the binding of S to RssB leads to the degradation of S by the Clp machinery (2, 3, 10), a mechanism divergent from the CrsAR partner switch of S. oneidensis. The partner-switching mechanism is not only responsible for alternative factors regulation because recent studies have highlighted the involvement of such a mechanism for the post-translational control of the 70 and 66 housekeeping sigma factors of E. coli and Chlamydia trachomatis, respectively. 66 is controlled by a canonical RsbVWU-like system, whereas in the case of 70 , the partner-switching mechanism is conserved, but Rsd and HPr, which act as an anti-sigma and an anti-sigma antagonist, respectively, are not RsbVW homologs (50,51). It is thus clear that the partner-switching mechanism is a general pathway to efficiently regulate sigma factors. Studying the signal transduction pathway that governs the activation of the CrsAR partnerswitching system and in particular how the stress signal is transferred to CrsR will lead to a better understanding of this complex mechanism of regulation. Because the first domain of CrsR is similar to receiver domains of response regulator proteins, a still unknown histidine kinase is probably involved in this signal transduction pathway.

Experimental Procedures
Medium, Growth Conditions, Strains, and Plasmids-All strains used in this study are listed in Table 2 and were routinely grown at 28°C (S. oneidensis strains) or at 37°C (E. coli strains) in LB medium (52). When appropriated, antibiotics were used at the following concentrations: kanamycin (25 g/ml), strep- The proteins were first incubated either by two, by three, or alone at 30°C for 1 h. EDC was then added, and samples were further incubated at 30°C for 1 h. The samples were submitted to SDS-15% PAGE, and proteins were revealed by Western blotting using a Strep-Tag antibody. The stars indicate the dimer and multimers between CrsR D3 and S , and the lines indicate multimerization of S . tomycin (100 g/ml), and ampicillin (50 g/ml). All S. oneidensis strains are derivatives of the MR1-R strain referred as WT. Deletions of SO2118 (crsA), SO2119 (crsR), and SO3432 (rpoS) were done by cloning two 500-bp fragments flanking the gene to be deleted into the pKNG101 suicide vector (53) at the BamHI and SpeI (SO2118 and SO3432) and at SpeI and SalI (SO2119) restriction sites as described before (54). The resulting plasmids were transformed into E. coli CC118pir and then transferred to MR1-R wild type (single deletion) or crsR mutant (double deletion) by conjugation with the E. coli helper strain 1047/pRK2013 (55). The plasmid was integrated into the chromosome by a first recombination event. The plasmid was then removed from the chromosome by a second recombination event in the presence of 6% sucrose. Clones were tested by PCR to select the ones for which the recombination event led to gene deletion. The deletions include all the coding sequence (from start to stop codons) of the genes of interest. Deletion of the target genes was confirmed by sequencing the appropriate overlapping region. To construct the dps::lacZ fusion, two 500-bp fragments flanking SO1158 were cloned into the pKNG101 suicide vector at the BamHI and SpeI restriction sites as described for the deletion mutants, except that the restriction sites XbaI and XmaI were added between the two 500-bp fragments. In a second cloning step, the coding sequence of lacZ (ATG ϩ lacZ from codon 9) was added at the XbaI and XmaI restriction sites. The resulting plasmid was transformed into E. coli CC118pir strain and transferred to MR1-R wild type or mutant by conjugation with the E. coli helper strain 1047/pRK2013 (55). The plasmid was integrated into the chromosome by a first recombination event, leading to the addition of the E. coli lacZ coding sequence. Contrary to the deletion mutants, the integrated plasmid was not removed from the chromosome to keep a wild type copy of the dps gene. The fusions were confirmed by sequencing. Importantly, wild type S. oneidensis does not possess a gene homologous to lacZ.
Cells overproducing the proteins of interest were collected by centrifugation. For the soluble proteins CrsR D3 -His, CrsR-His, and Strep-CrsA, cells were resuspended in 40 mM Tris-HCl, pH 7.6, disrupted by French press, and centrifuged at 13,000 rpm, at 4°C, for 15 min. After an ultracentrifugation step at 45,000 rpm at 4°C for 45 min, the supernatant was loaded onto a HiTrap FF resin (GE Healthcare) for His-tagged proteins or a Strep-Tactin resin (IBA) for Strep-tagged proteins, and proteins were purified according to the manufacturer's protocol. Purified proteins were loaded onto a PD-10 desalting column (GE Healthcare) and recovered in 20 mM sodium phosphate, pH 7.4, 15% glycerol for the His-tagged proteins or 40 mM Tris-HCl, pH 7.6, 15% glycerol for the Streptagged proteins. For Strep-CrsA-P purification, equal volume of soluble supernatant overproducing CrsR D3 -His and supernatant overproducing Strep-CrsA were incubated in the presence of 1 mM ATP. The mixture was incubated on ice during 1 h, before being submitted to Strep-Tactin resin (IBA) and purified as described for Strep-CrsA.
For the insoluble proteins (CrsA-His and Strep-S ), cells were resuspended in wash buffer (20 mM Tris-HCl, pH 7.6, 100 mM NaCl, 1 mM EDTA, 1 mM DTT, and 5% glycerol). The cells were disrupted by French press and centrifuged at 10,000 rpm at 4°C for 15 min. The pellets containing the insoluble proteins were solubilized in wash buffer containing 6 M guanidine. After ultracentrifugation at 35,000 rpm at 4°C for 1 h, the supernatant containing the solubilized proteins was dialyzed against wash buffer. The proteins were then loaded onto HiTrap FF resin (GE Healthcare) or Strep-Tactin resin (IBA) and purified as described above. Protein concentration was determined by Bradford assay (Bio-Rad). All the proteins were stored at Ϫ80°C.
Bacterial Two-hybrid Assays-Plasmids used for two-hybrid experiments are listed in Table 3. Bacterial two-hybrid experiments were performed as described (43,57) with the following modifications. Briefly, the coding sequences of SO2118, SO2119-D3, and rpoS (SO3432) were PCR-amplified from S. oneidensis genomic DNA. The PCR fragments were cloned into pEB354 and pEB355 vectors, using the EcoRI and XhoI restriction sites, in frame at the 3Ј-extremity of the sequences coding for the T18 and T25 domains of the adenylate cyclase from B. pertussis. Point mutations in crsA and crsR were obtained using the QuikChange site-directed mutagenesis kit (Agilent) by amplification of the whole vector using mutated primers. For interaction experiments, combination of plasmids coding for the T18 and T25 fusions were transformed into BTH101 or BTH101 clpXP::cat rpoS::tet strains as indicated in figure legends, and the plates (LB agar, kanamycin (50 g/ml), and ampicillin (100 g/ml)) were incubated at 30°C for 2 days. Ten isolated colonies from each plates were used to inoculate 3 ml of LB containing kanamycin (50 g/ml), ampicillin (100 g/ml), and isopropyl ␤-D-thiogalactopyranoside (0.5 mM). After overnight growth at 30°C, ␤-galactosidase activity was measured as described before (52).
In Vitro Kinase and Phosphatase Assays-The Phos-Tag TM experiments were performed as follow. CrsA and CrsR D3 (10 M each) were incubated either alone or together (20 l final volume) in 50 mM Tris-HCl, pH 7.6, 100 mM NaCl, 20 mM MgCl 2 , 1 mM DTT and in the presence or absence of 1 mM ATP, for 3 h at 30°C. Samples were subjected to Phos-Tag TM -10% PAGE prepared as previously described (58,59). Proteins were revealed by Coomassie Blue staining (Instant Blue-Expedeon).
For kinase activity revealed by native PAGE, the proteins were incubated (final volume, 40 l) at the concentrations indicated in the figure legends for 45 min at 30°C in 20 mM phos- In Vitro Protein Interaction Assays and K D Determination-For cross-linking experiments, the proteins were incubated at 30°C for 1 h in 10 mM sodium cacodylate, pH 6 (final volume, 20 l). The chemical cross-linker 1-ethyl-3-(3-dimethylaminopropyl) carbodiimide hydrochloride (EDC, 10 mM) was added, and samples were further incubated for 1 h. The samples were analyzed by SDS-PAGE, and proteins were visualized after Western blot using His-probe HRP-conjugated antibody (Pierce) or Strep-Tactin probe HRP-conjugated antibody (IBA).
BLI assay experiments were performed using a BLItz system as described by the manufacturer (ForteBio) with some modifications. Briefly, CrsR D3 -His (4 l) was diluted at 6 M in a phosphate buffer containing 20 mM sodium phosphate, pH 7.4, 150 mM NaCl, 10 mM imidazole, 0.05% Tween 20 and then incubated with the biosensor for 120 s (loading step). Dissociation step in sodium phosphate buffer was performed during 180 s before the biosensor was incubated in a Tris buffer (30 mM Tris, pH 7.6, 150 mM NaCl, 10 mM imidazole, 0.05% Tween 20) for the Strep-S analyte or in the same phosphate buffer for Strep-CrsA analyte for 600 s. Then independent biosensors bound to either CrsR D3 or empty biosensors (negative controls) were incubated with 200 l of 0.1-4 M of S analyte diluted in Tris buffer or with 200 l of 0.05-0.5 M of CrsA or phosphorylated CrsA analyte diluted in phosphate buffer for 30 min. Dissociation step either in Tris buffer for S analyte or in phosphate buffer for CrsA analyte was done during 10 min. As a control, appropriate buffer also replaced analytes during the association step on CrsR D3 -bound biosensors. The data have been corrected using the BLItz Pro TM software to remove the signal of the buffer alone and the signal of nonspecific binding of analytes. K D calculation was done with the BLItz Pro TM software.
For pulldown assays (final volume, 20 l), Strep-S (2 M) was incubated with either the proteins CrsR D3 -His (2 M) or CrsR-His (0.75 M) or alone in 20 mM sodium phosphate, pH 7.4, 150 mM NaCl, 10 mM imidazole, 0.05% Tween 20 at room temperature for 60 min. HisPur cobalt resin (Pierce, 30 l) was added to each mixture that was further incubated for 30 min. After three washes, the samples were resuspended in 30 l of denaturing buffer and loaded onto SDS-PAGE.
Mass Spectrometry Experiments-For in-gel trypsin digestion of protein, pieces of electrophoresis gel were put into a 96-well microplate (Greiner) for sample digestion. A robotic work station (Freedom EVO 100, TECAN) was used to perform automated sample preparation, including multiple steps: washes, reduction and alkylation, digestion by trypsin (Promega, proteomics grade) with 0.0025% ProteasMax (Promega), and extraction and drying of mixed peptides.
For MALDI-TOF MS analysis, the dried digested peptides were dissolved in 0.1% trifluoroacetic acid in water, and desalted on the ZIP-TIP C18 column (Milipore). Then samples were eluted and spotted onto a MALDI target plate with a sat-urated solution of matrix ␣-cyano-4-hydroxycinnamic acid (70% acetonitrile in water, 0.1% TFA (v/v)) and were further analyzed by MALDI-TOF Microflex II (Bruker) in positive reflectron mode. External mass calibration was performed with calibrant mixture Pep_Cal from Bruker. A peak-picking list was performed on FlexAnalysis software with the PMF_PS_FAMS program. The peak list was manually checked and calibrated internally, and then contaminant peaks were removed (trypsin autolysis, keratin). Search was done on the Mascot web site against NCBI database. Searches were performed using a maximum peptide mass tolerance of 50 ppm, two missed cleavages allowed, a fixed modification of cysteines by iodoacetamide (carbamidomethyl), a variable modification of methionines (oxidation), and phosphorylation of serine. Proteins were considered as identified only when they had a probability p Ͻ 0.05.
For dephosphorylation experiments, control analysis was performed with 250 fmol of caseine monophosphopeptide (Sigma) [MϩH] ϩ ϭ 2061.8291 (Fig. 5, D and E). Phosphatase alkaline (Roche) experiments were performed as described (60), on the same spot directly onto the MALDI target plate with ␣-cyano-4-hydroxycinnamic acid matrix. The characterization of dephosphorylation of peptides was performed with BioTools software (Bruker) by observing the loss of 80 Da of mass of the phosphopeptides after alkaline phosphatase treatment.
In Vivo Assays-To follow the activity of the dps::lacZ fusion in exponential and stationary phases, the strain was grown at 28°C anaerobically in LB medium and because S. oneidensis is a non-fermentative bacterium, medium was supplemented with TMAO (10 mM) as final electron acceptor. ␤-Galactosidase activities were measured as previously described (52).
Author Contributions-All authors were involved in conduction of experiments, analysis of the results, and the writing of the paper.