Role of Sphingomonas sp. Strain Fr1 PhyR-NepR-σEcfG Cascade in General Stress Response and Identification of a Negative Regulator of PhyR

ABSTRACT The general stress response in Alphaproteobacteria was recently described to depend on the alternative sigma factor σEcfG, whose activity is regulated by its anti-sigma factor NepR. The response regulator PhyR, in turn, regulates NepR activity in a partner-switching mechanism according to which phosphorylation of PhyR triggers sequestration of NepR by the sigma factor-like effector domain of PhyR. Although genes encoding predicted histidine kinases can often be found associated with phyR, little is known about their role in modulation of PhyR phosphorylation status. We demonstrate here that the PhyR-NepR-σEcfG cascade is important for multiple stress resistance and competitiveness in the phyllosphere in a naturally abundant plant epiphyte, Sphingomonas sp. strain Fr1, and provide evidence that the partner switching mechanism is conserved. We furthermore identify a gene, designated phyP, encoding a predicted histidine kinase at the phyR locus as essential. Genetic epistasis experiments suggest that PhyP acts upstream of PhyR, keeping PhyR in an unphosphorylated, inactive state in nonstress conditions, strictly depending on the predicted phosphorylatable site of PhyP, His-341. In vitro experiments show that Escherichia coli inner membrane fractions containing PhyP disrupt the PhyR-P/NepR complex. Together with the fact that PhyP lacks an obvious ATPase domain, these results are in agreement with PhyP functioning as a phosphatase of PhyR, rather than a kinase.

The capacity to sense and respond to environmental fluctuations is crucial to bacterial survival. In addition to specific stress responses that reprogram cells to adapt to a particular stimulus, more complex responses exist, such as the general stress response (GSR), which is induced in response to diverse signals. The GSR has been described in many bacteria, and it allows them to adopt a preventive multiple stress resistance when facing adverse conditions. GSR regulatory cascades have been studied extensively in the Gram-negative Escherichia coli and the Gram-positive Bacillus subtilis. In both organisms, the central player of the response is a sigma factor, S in E. coli and B in B. subtilis; however, their regulation differs greatly. The work performed during the last 20 years has led to the discovery of exquisite mechanisms of regulation that generate fine-tuned responses via the integration of multiple signals for nutritional status, the environment and internal energy status (for recent reviews, see references 18, 19, and 36).
Despite several studies reporting the existence of a GSR in Alphaproteobacteria, no homologues for S or B have been identified within this class and, until recently, the regulation of this response remained unknown (12). The first evidence for a regulatory system came from the discovery of the response regulator PhyR in Methylobacterium extorquens and of the extracytoplasmic function (ECF) sigma factor RpoE2 and the putative anti-sigma factor NepR (SMc01505) in Sinorhizobium meliloti (13,14,39). Since the output domain of PhyR resembles an ECF sigma factor, the question was raised whether PhyR might function as a sigma factor regulated by phosphorylation. However, our recent work in the alphaproteobacterium M. extorquens AM1 indicated that PhyR acts instead as an anti anti-sigma factor, which led to the proposition of a novel partner switching mechanism (11). According to the proposed model, stress exposure induces phosphorylation of PhyR, which enables its sigma factor-like effector domain to interact with NepR and thus disrupt the complex between the ECF sigma factor EcfG and its anti-sigma factor NepR. Similar interactions reported in Bradyrhizobium japonicum, S. meliloti, and Caulobacter crescentus suggest that this mechanism is conserved within the alphaproteobacterial class (5,15,20,29).
Little is known about signals perception and transduction to PhyR. Interaction between PhyR and NepR in vitro requires PhyR phosphorylation, and a phyR allele in which the phosphorylatable aspartate has been replaced by an alanine cannot complement a phyR mutant in the Alphaproteobacteria analyzed thus far (5,11,15,20). In addition, the crystal structure of C. crescentus PhyR revealed that the receiver and sigma factor-like domains of unphosphorylated PhyR make tight contacts, and it was proposed that this closed conformation is disrupted upon PhyR phosphorylation (20). These data indicate that PhyR activity is controlled by phosphorylation, and thus one or several cognate PhyR histidine kinases should exist. In support of this hypothesis, histidine kinases are often found encoded at phyR loci in Alphaproteobacteria (47,48). Although these proteins possess diverse signaling domains, which probably reflect the species-specific signals activating the system, their DHp (dimerization and histidine phosphotrans-fer) domains remain conserved. Since this domain is the major determinant of specificity for partner recognition (7,27,46), these histidine kinases are prime candidates for functioning as cognate PhyR kinases. In agreement with this, it was recently shown that PhyK (CC_3474), the putative histidine kinase encoded at the phyR locus in C. crescentus, is required for phosphorylation of PhyR and activation of the cascade in response to stresses (29). However, no direct interaction between PhyK and PhyR was demonstrated in that study.
In the present study, we investigated the role and conservation of the PhyR-NepR-EcfG system in Sphingomonas sp. strain Fr1, a representative of the group of Sphingomonadaceae that comprises abundant epiphytes (9). In this organism, EcfG , NepR, PhyR, and a putative histidine kinase are encoded at the same locus. The results demonstrate that the PhyR-NepR-EcfG cascade is involved in multiple stress resistance in Sphingomonas sp. Fr1 and that its mode of action is conserved. In addition, our data suggest that the putative histidine kinase encoded at the phyR locus does not function as a histidine kinase activating PhyR but rather acts as a phosphatase of PhyR. According to its proposed function, this protein was termed PhyP, for PhyR phosphatase.
Plasmids and strain constructions. Plasmids used in the present study are listed in Table 1. The oligonucleotides used to construct plasmids are listed in Table SA1 in the supplemental material. All DNA manipulations were performed according to standard protocols (38). Phusion polymerase was used for all PCRs, and the restriction enzymes were from Fermentas. To construct a plasmid for the Sphingomonas sp. Fr1 phyR deletion, two overlapping PCR products were generated with the primer pairs oJVZ825/oJVZ558 and oJVZ824/ oJVZ556. The products were then mixed, annealed, and extended without the addition of primers using Phusion polymerase (PCR overlap extension). The resulting fragment was cloned into pK18mobsacB, a broad-host-range sacBbased vector for unmarked allelic exchange (40). To construct plasmids for the ecfG, phyP, and ecfG2 deletions, the regions flanking each open reading frame (ORF) were amplified by PCR with the following primer pairs: for ⌬ecfG::kan, oJVZ555/oJVZ604 and oJVZ557/oJVZ603; for ⌬phyP::kan, oJVZ822/oJVZ823 and oJVZ559/oJVZ560; for nepR::kan, oJVZ589/oJVZ588 and oJVZ590/ oJVZ591; and for ecfG2::kan, oJVZ593/oJVZ592 and oJVZ842/oJVZ595. Fragments were cloned sequentially into a suicide vector for antibiotic marker exchange, pCM184 (32), except for the ecfG2 fragment, which was cloned into a ampicillin-sensitive pCM184 derivative, pAK400, obtained by pCM184 digestion with ScaI/Eam1105I, Klenow fragment treatment and self-ligation. After antibiotic marker exchange in Sphingomonas, the floxed antibiotic resistance cassette was excised using the Cre recombinase of pCM157 (32) to generate unmarked ⌬ecfG::loxP, ⌬phyP::loxP ⌬phyR, and ⌬ecfG::loxP ⌬nepR::loxP strains. The nepR::loxP mutant (Km s ) was obtained 10 days after transformation with pCM184-nepR and selection on kanamycin and had spontaneously excised the kanamycin resistance cassette. Although it is unclear how the excision event happened, it can be explained by a strong polar effect of nepR replacement by the kanamycin resistance cassette on phyP expression and thus the selection pressure to eliminate this effect.
A derivative of the broad-host-range plasmid pBBR1 (3), pAK200, which is compatible with the pCM62 IncP origin of replication, was constructed as follows. A 3.0-kb fragment including the rep and mob genes was PCR amplified from pBBR1MCS-3 (26) using the primer pair oJVZ601/oJVZ602. This fragment was ligated to a 1.6-kb PCR product carrying the multiple cloning site (MCS), lacZЈ, and the kanamycin resistance cassette (nptII) of pK18mobsacB amplified with the primers oJVZ599 and oJVZ600. A derivative of pAK200, pAK206, was constructed by excision of the kanamycin resistance cassette of pAK200 using BglII/BcuI, followed by insertion of a BamHI/BcuI-digested PCR product generated with the primers oJVZ552 and oJVZ553, which carries the gentamicin resistance cassette (aacC1) of pUC18-mini-Tn7T-Gm-lux (8).
For transcriptional reporter experiments, the tetracycline resistance cassette and the lac promoter of pCM62 were replaced by an MCS and a kanamycin resistance cassette by cloning a PCR product generated with the primer pair oJVZ639/oJVZ640 and pCM66 as a template into the PscI/Eco47III sites of pCM62, yielding pLM01. sYFP2 was amplified from plasmid pDsYFP with the primers oJVZ845/oJVZ46 and cloned into the Acc65I/EcoRI sites of pLM01, generating pLM-sYFP2. The ecfG2 promoter was amplified with the primers oJVZ843 and oJVZ844 and cloned between the BcuI and NcoI sites of pLM-sYFP2, giving pLM-sYFP2-ecfG2p. For bacterial adenylate cyclase two-hybrid (BACTH) interaction studies, nepR, ecfG, and ecfG2 ORFs were amplified using the primer pairs oJVZ781/oJVZ782, oJVZ783/oJVZ784, and oJVZ785/ oJVZ786, respectively, and cloned via XbaI/Acc65I in pKT25 (nepR) or pUT18C (ecfG and ecfG2).
For protein production in E. coli, the phyR ORF was PCR amplified with the primers oJVZ820 and oJVZ821 and cloned in pET24b. The sequences encoding the PhyR sigma factor-like domain and the nepR ORF were amplified by PCR with the primer pairs oJVZ838/oJVZ839 and oJVZ836/oJVZ837, respectively, and cloned into pET26bII, a modified pET26 in which a thrombin cleavage site sequence has been inserted upstream of the hexahistidine tag sequence (kindly provided by W. Malaga). The ecfG ORF was PCR amplified with the primers oJVZ828 and oJVZ835, introducing an N-terminal thrombin cleavage site, and cloned into pENTR4, followed by recombination into pDEST544 (Addgene plasmid 11519) using the LR Clonase mix (Invitrogen), according to the manufacturer's instructions. phyP or phyP(H341A) was amplified with the primers oJVZ738 and oJVZ739 and cloned into pET26bII.
All plasmid inserts were verified by sequencing (Microsynth, Switzerland). Plasmids were introduced into Sphingomonas sp. Fr1 by electroporation or were transferred between E. coli S17-1 pir (44) and Sphingomonas sp. Fr1 via conjugative transfer, with E. coli counterselection by 10 g of colistin/ml on MM (for pCM184 derivatives) or by 50 g of carbenicillin/ml on LB (for pAK400 derivatives). For complementation studies with more than one plasmid, strains were transformed sequentially.
Selection for viable PhyP His-341 substitutions. To construct a library of phyP in which the codon encoding His-341 is randomized, a PCR overlap extension strategy with the primer pairs oJVZ748/oJVZ537 and oJVZ826/oJVZ749 was used, with oJVZ748 and oJVZ749 containing degenerate nucleotides (any of the four nucleotides at a similar frequency) at the positions coding for H341. The resulting PCR fragments were cloned into pAK206, approximately 2,500 individual E. coli clones containing an insert (as judged by blue/white screening) were pooled, and plasmids were extracted. This library was transformed into Sphingomonas sp. Fr1 ⌬phyP::loxP ⌬phyR, approximately 10,000 of the resulting colonies were pooled, and competent cells were prepared and transformed with pCM62-PhyR. Plates were incubated for 7 days to allow potential PhyP His-341 substitutions with reduced activity to catch up. A total of 20 colonies were visible on plates, and for 11 the sequence of the full-length phyP allele was determined.
5RACE. Total RNA was extracted using an RNeasy minikit (Qiagen) according to the manufacturer's instructions. Total RNA was treated with DNase I (Ambion) prior to cDNA synthesis. 5ЈRACE (5Ј rapid amplification of cDNA  ends) was performed as described previously (13). The following primers were used for cDNA synthesis and PCR, respectively: ecfG_R3 and ecfG_R4, phyR_R3 and phyR_R4, nepR_R2 and nepR_R3, and phyP_R5 and phyP_R7. Phenotypic assays. For all assays, bacteria were grown on NB supplemented with appropriate antibiotics. Precultures were inoculated with a small loop of bacteria from fresh NB plates and then grown for 8 to 10 h. The main cultures (20 ml in 100-ml baffled flasks) were inoculated from precultures at an optical density at 600 nm (OD 600 ) of 0.0005 to 0.001 and grown overnight at 28°C with orbital shaking at 220 rpm. Cultures at an OD 600 of 1 were used for phenotypic assays.
For heat shock assays, the main cultures were transferred to a water bath at 46°C and incubated with shaking. Aliquots (100 l) were removed after 20, 40, and 60 min of incubation, and then 10-fold serial dilutions were applied to NB plates. Aliquots taken before transfer to 46°C served as controls. For desiccation assays, 10-fold serial dilutions were spotted onto mixed cellulose ester filter membranes (HAWG-A0; Millipore), which were dried for 5 h under sterile airflow prior to transfer to NB plates. Control filter membranes were placed on NB plates immediately after spotting. For sucrose and sodium chloride stress assays, 10-fold serial dilutions of cultures were spotted onto NB plates containing 15% (wt/vol) sucrose or 300 mM NaCl, respectively, or NB plates (controls). Sensitivity to methylglyoxal was measured by disk-diffusion assays, as described previously (13), except that a 2% (vol/vol) methylglyoxal solution was used. Sensitivity to oxidizing agents was measured by disk diffusion assays using the following agents: 1 M hydrogen peroxide, 2.5 or 12.5% (wt/vol) paraquat (methyl viologen), 500 mM diamide, and 3% cumene peroxide. Sensitivity to ultraviolet (UV) light was tested by exposing plates with 10-fold serial dilutions to 254-nm UV light for 30 s as described previously (13). CFU were counted or images were taken after 3 days of incubation at 28°C. Three biological replicates were performed for all experiments.
Seed sterilization, plant growth conditions, and harvest. Seeds of Arabidopsis thaliana ecotype Col-0 were sterilized as described previously (41). Seeds were inoculated with 5-l suspensions of individual strains (wild type or ⌬ecfG::kan mutant) or 1:1 mixtures of both strains, as described previously (42), except that glucose was used instead of succinate for bacterial cultures. Plants were grown and bacteria were harvested as described previously (42). To distinguish wildtype (kanamycin-sensitive) and ⌬ecfG (kanamycin-resistant) bacteria, 10-fold serial dilutions of cell suspensions were applied to glucose-MM plates with or without kanamycin. Plates were incubated for 2 days at 28°C before CFU determination. In the few cases where the number of CFU was lower than the detection limit (250 CFU/plant), values just below the detection limit (225 CFU/plant) were included to calculate mean values.
Protein purification and analytical gel filtration. Histidine-tagged versions of PhyR, the N-terminal part of PhyR (PhyRN), and NepR proteins were produced in E. coli BL21(DE3) and purified in two steps, using Ni-NTA and gel filtration, as described previously (11). The EcfG protein was expressed as a fusion with a hexahistidine-tagged NusA protein cleavable by thrombin. The fusion protein was purified using Ni-NTA, an equimolar amount of NepR was added, and thrombin cleavage was performed overnight at 4°C in thrombin cleavage buffer (20 mM Tris-HCl [pH 8], 150 mM NaCl, and 2.5 mM CaCl 2 ). The resulting mixture was subjected to preparative exclusion chromatography (S75 HiLoad; GE Healthcare), and the purified EcfG -NepR complex was loaded onto an analytical gel filtration column (Superdex S75 10/300; GE Healthcare). For PhyR/NepR interactions, analytical gel filtration was performed as described previously (11), except that the presence of proteins was followed based on the absorbance at 220 nm, since the NepR primary sequence contains only one tyrosine residue and is thus difficult to detect at 280 nm.
Preparation of E. coli inner membrane fractions. Cultures (50 ml) of E. coli BL21(DE3) harboring pET26bII-PhyP or pET26bII-PhyPH341A were induced by addition of 1 mM IPTG (isopropyl-␤-D-thiogalactopyranoside) at an OD 600 of 0.6 and incubated for 2 h before harvesting. Inner membrane fractions were prepared by a modified protocol for enzymatic cell lysis (35). Spheroblasts were prepared by resuspending cells in 20 ml of 30 mM Tris-HCl (pH 8.0), 20% sucrose, 10 mM EDTA, 250 mg of lysozyme/liter, and 185 mg of chloramphenicol/liter. After 30 min of incubation at room temperature, spheroblasts were sedimented (13,000 ϫ g, 4°C, 30 min) and dispersed in 10 ml of 10 mM phosphate buffer (pH 6.6) containing 2 mM MgSO 4 and 20 mg of DNase I (Roche)/ml using a syringe with an 18-gauge needle. After incubation at 37°C for 30 min, the cells were sonicated (30 s at maximum frequency; Soniprep 150 [Sanyo]), and unlysed spheroblasts were removed by centrifugation (800 ϫ g, 4°C, 1 h). The supernatant was centrifuged for 1 h at 92,000 ϫ g, 4°C to collect membranes. Membrane pellets were resuspended in 500 l of buffer I (50 mM Tris [pH 8.0], 40 mM KCl, 5 mM MgCl 2 , 5% glycerol), and aliquots of 50 l were stored at Ϫ80°C. The total protein concentration was approximately 3 mg/ml, as determined by a Bio-Rad protein assay. Western blots with membrane preparations were performed with anti-His(C-term)-horseradish peroxidase antibodies (Invitrogen) to ensure PhyP and PhyP(H341A) are present in comparable concentrations.
PhyR-P/NepR complex formation and disruption. The PhyR-P/NepR complex was formed in 50 mM Tris-HCl (pH 8.0), 150 mM NaCl, 25 mM acetyl phosphate (Sigma catalog no. CAS 94249-01-1), and 10 mM MgCl 2 with 20 mM NepR and PhyR in a volume of 550 l for 30 min at room temperature and was subsequently purified by gel filtration using a Superdex S75 10/300 column (GE Healthcare) in buffer GF (50 mM Tris-HCl [pH 8.0], 150 mM NaCl). Peak elution fractions were pooled (1 ml). The PhyR-P/NepR complex was stable for at least 2 days at room temperature. For complex disruption experiments, 450 l of PhyR-P/NepR complex were incubated with 50 l of membrane preparation, 50 l of EDTA-free Complete protease inhibitor cocktail (Roche; 1 tablet dissolved in 500 l of double-distilled H 2 O), and 10 mM MgCl 2 (final concentration) overnight. The mixture was centrifuged to pellet membranes (92,000 ϫ g, 4°C, 1 h) and the supernatant was subjected to analytic gel filtration on a Superdex S75 10/300 column (GE Healthcare) in buffer GF. Elution fractions corresponding to the elution volumes of PhyR and PhyR-P/NepR complex were collected and subjected to SDS-PAGE (12.5%) and Western blotting with primary rabbit anti-PhyR and rabbit anti-NepR antibodies (BioGenes, Germany) and secondary goat anti-rabbit AP-conjugated antibodies (Bio-Rad). The experiments were repeated five times with three independent membrane preparations; the results were consistently reproducible, and the results of one representative experiment are shown in Fig. 5. Note that the Superdex column used in these experiments was different from the one used in the experiments described above for protein purification and analytical gel filtration, which explains the different elution volumes. Measurement of BACTH system interactions. The E. coli cya strain BTH101 was cotransformed with pKT25-and pUT18C-derived plasmids, and single colonies were used to inoculate 5 ml of LB supplemented with 0.5 mM IPTG. The ␤-galactosidase activity measurements were performed on overnight cultures grown at 28°C according to the method of Miller (34) with at least three biological replicates.

RESULTS
Description of the phyR locus. PhyR, NepR, and EcfG are conserved in essentially all free-living Alphaproteobacteria, and their genes are usually found at the same locus (13,48). A PhyR ortholog was identified through a BLASTp search (1) against all of the proteins predicted by the RAST server (4) for the partial genome sequence of Sphingomonas sp. Fr1 (draft genome composed of 229 contigs, 4.1 Mb, unpublished). The organization of the phyR locus of Sphingomonas sp. Fr1 is shown in Fig. 1A: phyR is located upstream of ecfG, a small ORF located upstream of ecfG encodes a putative membrane protein of unknown function conserved among several Sphingomonas species, and nepR and a putative histidine kinaseencoding gene, which we designated phyP, are transcribed divergently to phyR.
The proximity of nepR and phyP suggested that they might be organized as an operon, and this was confirmed by reverse transcription-PCR (data not shown). The transcriptional start sites for ecfG, phyR, nepR, and phyP were mapped using 5ЈRACE. All mapped promoters show homology to the proposed consensus of EcfG -type promoters, i.e., GAAC-N 17,18 -GTT ( Fig. 1) (2, 13, 39, 48). The mapped ecfG promoter overlaps the preceding predicted small ORF. Interestingly, as it has been proposed for other Alphaproteobacteria (48), phyR and the nepR-phyP operon are divergently transcribed from a bidirectional promoter in which the proposed Ϫ10 box on each strand appears to completely overlap the Ϫ35 box on the complementary strand.
PhyR and EcfG are involved in multiple stress resistance. The PhyR-NepR-EcfG cascade has been shown to control the general stress response in several Alphaproteobacteria (13,15,29,31,39). We thus examined whether this role is conserved in Sphingomonas sp. Fr1. Null mutants of phyR and ecfG, as well as their corresponding complemented strains, were constructed, and their sensitivity to different stresses was analyzed. The null mutants displayed the same growth rate as the wildtype strain under standard growth conditions, which indicated that neither protein is required for growth under optimal conditions (data not shown). Comparison of the ability to grow on high-osmolarity media (15% [wt/vol] sucrose or 300 mM NaCl) indicated that both mutants are more sensitive than the wildtype strain to osmotic stress ( Fig. 2A). When exposed to desiccation, the phyR and ecfG mutant strains exhibited a Ͼ1,000fold reduction in viability compared to the wild-type strain (Fig. 2B). Both strains were also more sensitive to a temperature shift from 28 to 46°C (Fig. 2C) and to exposure to the toxic electrophile methylglyoxal (Fig. 2D). In all experiments, in trans complementation of the phyR and ecfG mutant strains restored stress resistance to wild-type levels or better. The ecfG-complemented strain consistently showed higher resistance to stresses than the wild-type strain. This elevated resistance may have resulted from elevated EcfG levels generated by expression from a multicopy plasmid. Such an effect was not observed for the phyR-complemented strain, in line with earlier reports showing that the output of many response regulators is independent of their absolute concentrations, since their activity is determined by their phosphorylation status (37).
Under the experimental conditions tested, no differences in sensitivity to hydrogen peroxide, cumene peroxide, methyl viologen, UV light, ethanol, or the thiol-oxidizing agent diamide were observed between the wild-type and mutant strains (data not shown).
These findings indicate that PhyR and EcfG are important for resistance against various stresses that are distinct in nature, including osmotic stress caused by high salt and sucrose, methylglyoxal, which attacks the nucleophilic centers of macromolecules, and elevated temperatures, which primarily cause mis-or unfolding of proteins. Since PhyR and EcfG mediate resistance to multiple, unrelated stresses, these activities are referred to as a general stress response.
EcfG is important for plant colonization. PhyR is essential for plant colonization by M. extorquens AM1 (14), which shows the importance of the general stress response in this harsh environment. Since Sphingomonas sp. Fr1 was isolated from the phyllosphere, the role of EcfG in plant colonization was examined. Arabidopsis thaliana seeds were inoculated with wild-type and ecfG mutant strains, individually or in combination. In competition experiments, antibiotic resistance was used to distinguish wild-type from the kanamycin-resistant ecfG mutant strain. The average number of cells recovered from plants inoculated with the ecfG mutant was only slightly lower from plants inoculated with the wild-type (Fig. 2E), which indicates that this strain can colonize plants. However, when the ecfG mutant was tested for its capacity to compete with the wild type, only ca. 1% of recovered cells were ecfG mutants (Fig. 2E). Therefore, while not essential for growth in planta under laboratory conditions, EcfG confers a selective advantage in the phyllosphere.

VOL. 193, 2011
PhyR-NepR-EcfG CASCADE OF SPHINGOMONAS SP. Fr1 Protein interactions in the PhyR-NepR-EcfG cascade are conserved in Sphingomonas sp. Fr1. The observation that phyR and ecfG mutants exhibit the same phenotype is in agreement with their function in the same regulatory cascade. To analyze conservation of the PhyR-NepR-EcfG regulatory system further, we examined whether the protein interactions implicated in the proposed partner switching mechanism for Alphaproteobacteria (11) are conserved in Sphingomonas sp. Fr1, namely, whether NepR is the anti-sigma factor of EcfG and PhyR the anti-anti-sigma factor. In vivo, elevated levels of nepR rendered the wild-type strain more sensitive to osmotic stress (Fig. 2F), which is a finding consistent with the role of NepR as a negative regulator of the PhyR-NepR-EcfG cascade in other Alphaproteobacteria (11,39). To assess the role of PhyR phosphorylation in Sphingomonas sp. Fr1 in vivo, the predicted phosphorylatable aspartate (residue 194) of PhyR was changed to an alanine. The ability of this mutant allele to complement a phyR mutant was then tested in phenotypic assays. As shown in Fig. 2, expression of phyR(D194A) in a ⌬phyR background failed to restore the wild-type phenotype, which suggests that this mutant protein, indeed, represents an inactive form of PhyR that cannot interact with NepR. Note that PhyR levels were similar in strains complemented with the wild-type or D194A phyR alleles, as assessed by immunoblot experiments with antibodies against PhyR (data not shown).
The interactions between NepR and EcfG or PhyR in vitro were analyzed using size exclusion chromatography. PhyR and NepR were produced as histidine-tagged versions in E. coli, and EcfG as a fusion with histidine-tagged NusA, which was later removed by thrombin cleavage. Since EcfG tended to aggregate after cleavage, His-NusA-EcfG was incubated with NepR prior to cleavage and gel filtration experiments. EcfG and NepR coeluted in a peak corresponding to an apparent molecular mass of 33 kDa (Fig. 3A), which is consistent with the formation of a heterodimer (theoretical molecular mass of 34.6 kDa), whereas NepR alone eluted with an apparent molecular mass of 23 kDa, a size consistent with a dimeric or trimeric form of NepR (predicted molecular mass of 8 kDa; data not shown). These findings indicate that NepR and EcfG can form a complex, and both in vivo and in vitro data support the idea that NepR functions as an anti-sigma factor of EcfG in Sphingomonas sp. Fr1. colonization efficiencies of wild-type and ecfG mutant strains. Seeds were inoculated with wild-type or ⌬ecfG::kan strains or with a 1:1 mixture of wild-type and ⌬ecfG::kan strains. The number of CFU recovered from each plant is indicated as a dot (wild-type, black dot; ⌬ecfG::kan, gray dot), and the means are shown as a bar for each strain. The results of three independent experiments are shown. In panel F, wild-type (WT) and strains carrying nepR (pNepR) or phyP (pPhyP) on a multicopy plasmid were tested for their capacity to grow on high-osmolarity medium (NB with 300 mM NaCl). In panels A, B, and F, the results of one representative experiment of three biological replicates are shown. In each panel, spots of 10-fold serial dilutions are shown from left to right, starting with undiluted sample. In panel C, the percentage of viability corresponds to the ratio of treated to untreated cells and is the average of three biological replicates. In panel D, the diameter of inhibition indicated is the mean of three biological replicates. In panels C and D, error bars show the standard deviation of three biological replicates.
FIG. 2. Stress sensitivity assays and in planta experiments. Wildtype and mutant strains were assessed for their capacity to grow on high-osmolarity medium (NB containing 300 mM NaCl or 15% [wt/ vol] sucrose) (A) and for their sensitivity to desiccation (B), heat exposure (C), and methylglyoxal (D). Strains are indicated as follows: WT, wild-type/pCM62; ⌬ecfG, ⌬ecfG::kan/pCM62; pEcfG, ⌬ecfG::kan/pCM62-EcfG; ⌬phyR, ⌬phyR/pCM62; pPhyR, ⌬phyR/ pCM62-PhyR; and pD194A, ⌬phyR/pCM62-PhyRD194A. (E) Plant Interactions between PhyR and NepR were examined by analytical gel filtration with acetyl phosphate used as the phospho-donor for PhyR phosphorylation. In the absence of acetyl phosphate, the two proteins eluted separately with apparent molecular masses of 23 and 39.9 kDa (predicted to be 30 kDa) for NepR and PhyR, respectively (Fig. 3B). In the presence of acetyl phosphate, equimolar amounts of PhyR and NepR eluted within a peak corresponding to 46.8 kDa, indicating heterodimer formation (predicted molecular mass of 38 kDa; Fig. 3C). The elution volume of PhyR was the same regardless of the presence of acetyl phosphate (see Fig. SA1A in the supplemental material), indicating that phosphorylated PhyR does not dimerize, similar to what has been observed for M. extorquens PhyR (11). Finally, we tested whether the ECF sigma factor-like domain of PhyR was sufficient for interactions with NepR. Alone, the ECF sigma factor-like domain of PhyR eluted with an apparent molecular mass of 26.5 kDa (predicted molecular mass of 15 kDa; data not shown). When applied in equimolar amounts, NepR and the ECF sigma factor-like domain of PhyR eluted in a single peak with an apparent molecular mass of 32 kDa (the theoretical molecular mass of the complex was 23 kDa; see Fig. SA1B in the supplemental material). Thus, only the phosphorylated form of PhyR interacts with NepR in vitro, and this occurs via its ECF sigma factor-like domain, as previously described for M. extorquens proteins (11).
Together with the interaction between NepR and EcfG , these data demonstrate conservation of the mechanism underlying the PhyR-NepR-EcfG cascade in Sphingomonas sp. Fr1 and other Alphaproteobacteria.
Characterization of EcfG2 . A second gene encoding a EcfG -type sigma factor, ecfG2, is predicted in the Sphingomonas sp. Fr1 draft genome (Fig. 1B) based on homology to EcfG (43% identity, 61% similarity). To test whether this sigma factor might also be involved in the general stress response, the corresponding gene was deleted, and the resulting ecfG2 mutant subjected to phenotypic assays. No increased sensitivity to any of the stresses a phyR mutant is sensitive to was observed (data not shown). Bacterial two-hybrid assays based on Bordetella pertussis adenylate cyclase fragment complementation (24) further indicated that EcfG2 does not interact with NepR, whereas EcfG and NepR interacted in this system (Fig. 4A).
Finally, using transcriptional reporter fusions, it was demonstrated that the ecfG2 promoter is PhyR and EcfG dependent (Fig. 4B), a finding consistent with the presence of a EcfG -type promoter sequence in this region (Fig. 1B). These results suggest that EcfG2 is not involved in the PhyR-NepR-EcfG core cascade.
PhyP is a negative regulator of the PhyR-NepR-EcfG cascade. The importance of PhyR phosphorylation in the partner switching model implies the existence of modulators of PhyR phosphorylation. Putative histidine kinases are encoded at phyR loci in a number of Alphaproteobacteria (13,47,48), including Sphingomonas sp. Fr1. These proteins are prime candidates for the control of PhyR phosphorylation and thus PhyR activation. All efforts to obtain a null mutant of the gene encoding the putative kinase at the phyR locus in Sphingomonas sp. Fr1 failed, suggesting that phyP might be essential. Interestingly, work in S. meliloti has shown that overactivation of the cascade is lethal, whereby nepR is essential and ecfG cannot be overexpressed unless co-overexpressed with nepR (39). By analogy, if overactivation of the cascade was lethal in Sphingomonas sp. Fr1 and if phyP was a negative regulator of the cascade, then a phyP-null mutation would also be lethal.
The idea that PhyP functions as a negative regulator is based on the fact that (i) many histidine kinases are bifunctional, also acting on their cognate response regulators as phosphatases, and that (ii) deletion of these bifunctional kinases/phosphatases can lead to elevated levels of phosphorylated response regulator due to nonphysiological cross talk (27,45).
To pursue the hypothesis that PhyP functions as a negative regulator of the cascade, we attempted to generate a phyP deletion in the ⌬phyR or ⌬ecfG::loxP genetic backgrounds. Both phyR phyP and ecfG phyP double mutants were readily obtained, which suggests that the apparent lethality of a phyP mutant is dependent on an intact PhyR-NepR-EcfG cascade and that PhyP acts upstream of PhyR, since phyR is epistatic to phyP. In accordance with this, when phyR or ecfG was provided in trans on a multicopy plasmid in the phyR phyP or ecfG phyP double mutant, respectively, no viable colonies were obtained even after 10 days of incubation, and the viability of these strains could be rescued by simultaneously expressing phyP in trans. To test whether nepR was also essential, a similar experiment was performed. An ecfG nepR double mutant (ecfG::loxP nepR::loxP) was constructed and transformed with a plasmid expressing ecfG. This strain was viable, although it showed a delayed growth phenotype, suggesting that nepR is not essential. In fact, it was also possible to obtain a nonpolar single nepR mutant that had a similar delayed growth phenotype. These results might be explained by the presence of second, as-yet-unidentified nepR paralog in the Sphingomonas sp. Fr1 genome or another negative regulator of EcfG downstream of PhyR and do not preclude that overactivation of the cascade might, in fact, be lethal.
Predicted phosphorylation sites in PhyR and PhyP are essential for PhyP activity. If PhyP would keep PhyR in an inactive state, any condition mimicking this state should be viable in a phyP mutant background. Indeed, when the phyR phyP mutant was transformed with a plasmid containing the phyR(D194A) allele, encoding a phosphorylation-incompetent PhyR, the resulting strain was viable. Similarly, coexpression of the wild-type phyR allele with nepR from a multicopy plasmid in the phyR phyP mutant could rescue the lethal effect of the phyR allele, presumably because the elevated levels of NepR could titrate both active PhyR and EcfG .
His-341 of PhyP corresponds to the conserved phosphorylatable histidine residue essential for kinase function in characterized histidine kinases. A H341A substitution in PhyP was constructed to study the role of this conserved residue. The resulting phyP allele, phyP(H341A), could not complement a phyP phyR mutant coexpressing phyR, indicating that His-341 is important for PhyP function. Since phyP might encode a phos-phatase and some bifunctional histidine kinases/phosphatases retain phosphatase activity when the conserved residue is replaced by certain amino acids (21,23), the codon encoding His-341 of PhyP was randomized, and alleles were selected for that could restore viability in a phyR phyP mutant coexpressing phyR (see Materials and Methods). However, of 11 alleles recovered after selection, all encoded histidine at position 341 (7 by the CAC and 4 by the original CAT codon). These results suggest that His-341 is essential for PhyP function, although we cannot rule out the possibility that some substitutions might preserve residual phosphatase activity (if such an activity exists) that, however, would be insufficient in vivo.
Altogether, these findings indicate that PhyP acts catalytically on PhyR, thereby preventing complex formation between NepR and PhyR, rather than by a titration mechanism like NepR. This is also consistent with the fact that PhyP overexpression, unlike NepR overexpression, does not lead to increased sensitivity to osmotic stress (Fig. 2F). One obvious explanation is that PhyP acts as a phosphatase of PhyR.
PhyP disrupts the PhyR-P/NepR complex in vitro. In order to test whether PhyP could disrupt the complex between phosphorylated PhyR (PhyR-P) and NepR in vitro, the PhyR-P/NepR complex was purified by gel filtration, thus removing acetyl phosphate, and incubated with the purified cytoplasmic part of PhyP. No disruption of the complex was observed, as assessed by analytic size exclusion chromatography (data not shown). Since the truncated form of PhyP might be inactive, full-length PhyP was tested for its capacity to disrupt the complex. PhyP and PhyP(H341A), respectively, were expressed in E. coli, and inner membrane fractions were prepared, followed by incubation with purified PhyR-P/NepR complex. For PhyP(H341A), no complex disruption was observed as judged by a peak elution volume of 8.91 ml compared to 8.68 ml for the purified PhyR-P/NepR complex and 9.37 ml for PhyR alone (data not shown). In contrast, wild-type PhyP apparently disrupted the PhyR-P/ NepR complex, since the peak elution volume was 9.53 ml, close to the peak elution volume of 9.37 for PhyR alone (data not shown). To verify the presence of PhyR and/or NepR, elution fractions ranging from 7.8 to 10.6 ml were subjected to SDS-PAGE and Western blotting with antibodies against PhyR and NepR. As shown in Fig. 5, PhyR was present in both elution profiles, but NepR was present only in the one that had been incubated with the nonfunctional PhyP(H341A) protein, in which it coeluted with PhyR, thus confirming specific complex disruption by functional PhyP. Note that NepR was never observed in free form in these gel filtration experiments, which may be due to its hydrophobic character and in consequence its association with the membrane fraction (S. Campagne et al., unpublished data).
In summary, although we cannot exclude the unlikely possibility that a gene product conserved between distantly related E. coli and Sphingomonas sp. Fr1 mediates complex disruption of PhyR-P/NepR by PhyP, these results point toward a direct physical interaction of PhyP with PhyR, or the PhyR-P/NepR complex. Taken together, with the in vivo experiments, these results are in line with the hypothesis that PhyP acts as a phosphatase of PhyR.

DISCUSSION
The present study characterized the PhyR-NepR-EcfG cascade in the epiphytic bacterium Sphingomonas sp. Fr1. In this organism, PhyR, NepR and EcfG are encoded at the same locus, where a putative histidine kinase encoding gene, phyP, is also present. The phenotypic characterization of mutants determined that the PhyR-NepR-EcfG cascade is an important factor in multiple stress resistance. In vitro experiments demonstrated that NepR interacts with EcfG and with phosphorylated PhyR. These results are in agreement with the in vivo data and together, they suggest conservation of the partner switching mechanism already proposed to occur in other Alphaproteobacteria (11). Epistasis experiments and mutant alleles were used to identify PhyP as a negative regulator of PhyR phosphorylation. Together with in vitro experiments, these data suggest that PhyP maintains PhyR in an unphosphorylated state in the absence of stress.
Besides conservation of the role of the PhyR-NepR-EcfG cascade in multiple stress resistance, the system of Sphingomonas sp. Fr1 shares features observed in other Alphaproteobacteria. In Sphingomonas sp. Fr1, EcfG -type promoters drive the expression of phyR, ecfG, and the nepR-phyP operon. This aspect seems conserved in other genera, although experimental evidence exists for only a few species (2,15,39,48). However, other promoters probably exist, at least for phyR, since PhyR can still be detected by immunoblots in an ecfG mutant (data not shown). This finding is contrary to observations in B. japonicum and C. crescentus, where phyR expression appears solely dependent on EcfG (15,20). Another feature concerns the presence of a second EcfG , EcfG2 , which apparently is not directly controlled by NepR through protein-protein interac-tions but instead regulated at the transcriptional level by EcfG . This is reminiscent of the Caulobacter system, where two EcfG proteins (SigT and SigU) exist. While the former regulates the latter, SigU does not appear to be a direct player in the partner switching mechanism but regulates a rather small subset of the SigT regulon (2). Whether the situation is similar in Sphingomonas sp. Fr1 necessitates further elucidation.
Regulation of PhyR phosphorylation is essential for system function according to the proposed model (Fig. 1C). Although our results suggest that PhyP acts as a phosphatase of PhyR, it remains to be shown whether PhyP also displays additional histidine kinase activity toward PhyR, similar to other bifunctional histidine kinases/phosphatases. However, bioinformatics analyses suggest that this is not the case. (i) Pfam or SMART (28,43) do not predict any ATPase domain. (ii) No DUF or COG domains are predicted, which could be indicative of an as-yet-uncharacterized ATPase domain (10,30). (iii) Multiple sequence alignments of phyR-associated kinases reveal no conservation of key residues in N, G1, and G2 box catalytic motifs in PhyP (16, 25) (see Fig. SA2 in the supplemental material). (iv) Finally, a BLASTp search with the PhyP part C-terminal to the DHp domain gave no significant hits except PhyP orthologs of related Sphingomonas species (data not shown). Altogether, this suggests that PhyP is devoid of ATPase activity rather than that it contains a novel or weakly conserved catalytic and ATPase (CA) domain. That a CA domain is not essential for the phosphatase activity of several bifunctional histidine kinases/phosphatases (6,49) is in agreement with the proposed function of PhyP as a phosphatase of PhyR. In contrast, a CA domain is absolutely required for histidine kinase function, and this suggests that PhyP does not have kinase activity toward PhyR in addition.
How the PhyR response is triggered remains an open question. Our results indicate that lack of PhyP leads to activation of the PhyR-NepR-EcfG cascade. In consequence, it is plausible to assume that turning down the phosphatase activity of PhyP is in principle sufficient to trigger the cascade and this might be one mechanism to activate the system. Consistent with the idea of PhyP activity being tunable, PhyP contains a putative periplasmic sensor domain and a linker HAMP domain typical for membrane-associated histidine kinases. However, it remains unknown how PhyR is phosphorylated, and additional modulators of PhyR phosphorylation are expected to exist, such as cognate sensor histidine kinases or endogenous small phospho-donors. Defining the complement of factors controlling PhyR phosphorylation and how they interact will be the subject of future work. PhyR-NepR-EcfG CASCADE OF SPHINGOMONAS SP. Fr1 6637