Proteolysis of histidine kinase VgrS inhibits its autophosphorylation and promotes osmostress resistance in Xanthomonas campestris

In bacterial cells, histidine kinases (HKs) are receptors that monitor environmental and intracellular stimuli. HKs and their cognate response regulators constitute two-component signalling systems (TCSs) that modulate cellular homeostasis through reversible protein phosphorylation. Here the authors show that the plant pathogen Xanthomonas campestris pv. campestris responds to osmostress conditions by regulating the activity of a HK (VgrS) via irreversible, proteolytic modification. This regulation is mediated by a periplasmic, PDZ-domain-containing protease (Prc) that cleaves the N-terminal sensor region of VgrS. Cleavage of VgrS inhibits its autokinase activity and regulates the ability of the cognate response regulator (VgrR) to bind promoters of downstream genes, thus promoting bacterial adaptation to osmostress.

I n both eukaryotic and prokaryotic cells, signalling pathways can be generally divided into two major modes according to the persistence of protein participants: reversible and irreversible. For example, proteolysis is an irreversible, post-translational signalling cascade that modulates cell physiology 1,2 , whereas protein phosphorylation catalysed by receptor kinases is reversible 3 . Although, the regulatory mechanisms of these two kinds of cellular signalling pathways have been extensively investigated, the cross-regulation between them is not fully understood. Recent studies in eukaryotes revealed that the proteolysis of receptor kinases is critical for regulating development, apoptosis and tumour genesis [4][5][6] . However, in prokaryotic cells, how proteolysis modifies receptor kinases and controls their regulatory functions are incompletely understood.
In gram-negative bacteria, regulated proteolysis catalysed by PDZ-domain-containing proteases, such as the HtrA-family proteases (high temperature requirement A) and tail-specific proteases (Tsp), modulates multiple physiological pathways, including virulence, stress response, quorum-sensing, protein quality control and antibiotic resistance 7,8 . Thus, these proteases are potential molecular targets for the development of novel antibacterial agents 9 . These proteases contain multiple domains, including one or more PDZ domains, and they are mostly located in the periplasm of gram-negative bacteria 10 . PDZ domaincontaining proteases regulate physiological processes by binding or cleaving their protein substrates in cells. For example, under stress conditions, the DegS protease of Escherichia coli cleaves anti-σ factor RseA to release σ E from the membrane and then activates the transcription of various stress-response genes 11 . In Vibrio cholerae, a Tsp protease cleaves TcpP to initiate the subsequent proteolysis and inactivation of the membrane-bound transcription factor (TF) 12 . In addition to anti-σ factors and TFs, studies revealed that PDZ domain-containing proteases also process misfolded proteins 13 , outer-membrane proteins, phosphatases 14 , haemagglutinin 15 and autotransporters 16 . However, identification of the physiological substrates of proteases remains technically challenging, and thus, the regulatory functions of most proteases are unclear 17 .
The majority of bacterial species belonging to the genus Xanthomonas are plant pathogens. Among them, Xanthomonas oryzae pv. oryzae is the causative agent of rice bacterial blight disease, and Xanthomonas campestris pv. campestris causes the black rot disease of crucifers. Both bacterial species encode at least six PDZ domain-containing proteases 18 . The inactivation of one orthologous gene, prc (also named tsp), results in the attenuation of virulence and hypersensitivities to multiple environmental stresses 18,19 . Prc of X. oryzae pv. oryzae is a periplasmic protein, and its cleavage substrate remains to be discovered. When the bacterium is grown under osmolarity stress (osmostress), Prc binds to a virulence-associated dipeptidyl peptidase (DppP). Rather than cleaving DppP, Prc stabilizes DppP, suggesting it has a chaperone activity 18 . Therefore, DppP is not a proteolytic substrate of Prc. More direct and effective approaches are necessary to identify the physiological substrates of this protease, which will allow the regulatory role of Prc to be determined.
In the present study, we aim at identify the physiological substrates of Prc and investigate the role of Prc-catalysed proteolysis in controlling the environmental adaptation of X. campestris pv. campestris, which is more genetically amenable than X. oryzae pv. oryzae. Under osmostress, Prc directly binds and cleaves the N-terminal sensor region of VgrS, a canonical histidine kinase (HK) that controls bacterial virulence and stress responses. Cleavage of the VgrS sensor inhibits its autokinase activity, which then modulates the TF activity of the cognate VgrR in binding the promoters of downstream genes. This process significantly promotes bacterial resistance to osmostress.
Our study reveal that the proteolytic processing of the sensor region of HK by protease is a molecular mechanism to control the TCS signalling of gram-negative bacteria.

Results
Prc regulates virulence and bacterial stress resistance. In X. campestris pv. campestris 8004, XC_0714 is the orthologue of Prc (PXO_04290) in X. oryzae pv. oryzae PXO99 A (BlastP search, e-value = 0, identities = 94%). Both proteins encode three domains: an N-terminal PDZ domain, a central peptidase domain and a C-terminal DUF3340 domain with an unknown function (Fig. 1a). The Prc protein of X. campestris pv. campestris contains a predicted 26-aa signal peptide with a cleavage site between the 26th and 27th positions. According to a MEROPS peptidase database search result 20 , Prc belongs to the S41A serine endopeptidase subfamily (e-value = 2.40 × 10 −99 ) that contains two conserved catalytic residues (Ser 475 and Lys 500 ). This is different from HtrA family proteases (S1C subfamily) that usually contain a catalytic triad (His-Asp-Ser). Western blotting revealed that the Prc of X. campestris pv. campestris localized to both the periplasm and cytosol (Fig. 1b). In addition, an RT-PCR analysis revealed that prc is located in a four-gene operon that also contains XC_0711, XC_0712, and XC_0713 because positive RT-PCR products were amplified from a cDNA template derived from the intergenic transcripts of these genes ( Supplementary Fig. 1).
We constructed various prc mutants and characterized their phenotypic alterations, including biofilm development, virulence, the production of extracellular polysaccharides and enzymes, and resistance to multiple antibiotics and environmental stresses. Inframe deletions of the prc coding sequence or its PDZ (ΔPDZ), peptidase (ΔPEP) or DUF (ΔDUF) domains′ coding sequences independently caused significant decreases in bacterial virulence and in planta growth against susceptible cabbage (Brassica oleraceae cv. Jingfeng No. 1; Fig. 1c, d, Supplementary Fig. 2a-2c) and tolerance to the antibiotics erythromycin and kanamycin (Fig. 1e, f), as well as bacterial resistance to a high Fe 2+ concentration (2.5 mM; Fig. 1g) and osmostress (1.0 M sorbitol; Fig. 1h). Genetic complementation was performed by providing a full-length prc (Δprc-prc) or the heterogeneous prc of X. oryzae pv. oryzae PXO99 A (Δprc-prc Xoo ) and fully restored phenotypic deficiencies to levels similar to those of the wild-type (WT) strain ( Fig. 1c-h, Supplementary Fig. 2). Thus, prc may regulate bacterial virulence and resistance to multiple environmental stresses, and the biological functions of the prc of X. campestris pv. campestris and X. oryzae pv. oryzae appear to be highly conserved. Here, we focused on the role of Prc in regulating bacterial responses to osmostress.
Prc is a bona fide serine endopeptidase. Our previous study failed to detect the in vitro peptidase activity of Prc in X. oryzae pv. oryzae 18 . In this work, a mature form of Prc from X. campestris pv. campestris was successfully obtained by expressing the full-length prc in the Escherichia coli BL21(DE3) strain. N-terminal amino acid sequencing of the purified, soluble Prc showed that the 26-aa signal peptide was removed, indicating that E. coli cells could recognize and process the cleavage site of the signal peptide. As shown in Fig. 2a and Supplementary  Fig. 3a, the mature Prc of X. campestris pv. campestris not only degraded the universal substrate β-casein (Fig. 2a, Lanes 4 and 5) but also had endopeptidase activity that cleaved azocasein, a general chromogenic substrate for endopeptidases (Fig. 2b). When the core catalytic residues Ser 475 and Lys 500 were substituted, the degradative activities of recombinant Prc S475A and Prc K500A , respectively, were completely lost or significantly decreased (Fig. 2a, Lanes 8 and 9, and Lanes 12 and 13, respectively; Fig. 2b). Using Michaelis-Menten kinetic curve and Lineweaver-Burk plot analyses, Prc protease azocaseinhydrolysing activities were measured at 37°C. As shown in Fig. 2c, d, the maximum protease activity and the Michaelis constant (K m ) value were 0.0023 ± 0.001 μmol/min/μg and 2.48 ± 0.32 μM, respectively. This kinetic value is similar to that of the Tsp from E. coli (K m = 4.4 ± 0.6 μM) and represents a relatively strong peptidase activity relative to other Tsp and HtrA proteases, which usually have K m values ranging from micromolar to nanomolar levels [21][22][23] . Prc protease activity was very stable under various conditions. The optimized temperature range for Prc activity was~28-37°C, and changing the pH value from 4.0 to 10.0 did not remarkably affect Prc activity (Supplementary Fig. 3b  and 3c). In addition, when the dithiothreitol concentration in the reaction buffer was changed from 0 to 200 mM, the Prc activity level was only slightly affected ( Supplementary Fig. 3d). These genetic and biochemical results confirmed the theoretical prediction that Prc is a serine endopeptidase.
Monomeric Prc directly binds the VgrS sensor. An affinity proteomic approach using tandem affinity purification (TAP) together with a nanoscale liquid chromatography-mass spectrometry (nanoLC-MS/MS) analysis 24 was employed to screen for Prc-binding proteins. Thus, a recombinant bacterial strain (Δprcprc S475A -HA-FLAG) was constructed for the TAP analysis. In this strain, HA and FLAG epitope tags were fused in tandem to the C-terminus of an inactive Prc in which the Ser 475 residue has been substituted with Ala. This substitution inactivated Prc protease activity to avoid the degradation of binding proteins during the analysis. The C-terminal HA and FLAG tags did not affect bacterial resistance to osmostress because the corresponding strain (Δprc-prc-HA-FLAG), which encoded a functional Prc, showed a growth rate similar to that of the WT strain on NYG-sorbitol plates ( Supplementary Fig. 2f). The bacterial cells of the Δprc-prc S475A -HA-FLAG strain were treated with osmostress consisting of 1.0 M sorbitol before being lysed by freeze grinding, which prevented the dissociation of Prc-protein   interactions as much as possible. With or without osmostress treatment, the affinity proteomic screening, through two independent rounds of immunoprecipitation, obtained 20 and 36 proteins, respectively, with 12 being present under both conditions (Fig. 3a). These 44 proteins were classified into 15 functional categories (Fig. 3b, Supplementary Table 3) and represent a subset of proteins potentially bound to Prc. Although further functional verification is needed, the diversity of these candidate proteins indicated that Prc may bind various proteins and modulate bacterial osmostress responses. The TAP analysis identified a membrane-bound HK VgrS (XC_1050) that is important in regulating bacterial virulence and multiple-stress responses (Supplementary Table 3) 25 . VgrS is a metal receptor that contains a periplasmic, ferric iron-binding sensor, a transmembrane (TM) helix, a HAMP linker and a typical C-terminal transmitter domain (Fig. 3c, upper panel) 25,26 . Because both mature Prc and the sensor region of VgrS are located in the periplasm, we hypothesized that Prc bound to the VgrS sensor. As shown in Fig. 3c (lower left panel), a gel filtration analysis, together with analytical ultracentrifugation, revealed that Prc was present as a trimer (molecular weight = 272.43 ± 24.21 kDa) and monomer (molecular weight = 75.83 ± 10.17 kDa) in vitro. We purified a VgrS sensor peptide (Fig. 3c, lower right panel) and then used surface plasmon resonance (SPR) to measure the possible binding events between the VgrS sensor and different forms of Prc. The VgrS sensor physically bound the Prc monomer with a dissociation constant (K d ) of 33.9 μM (Fig. 3d), suggesting an intermediate level of protein-protein interaction. However, the VgrS sensor did not bind to the Prc trimer (Fig. 3e). A microscale thermophoresis assay (MST) also confirmed direct binding between the Prc monomer and VgrS sensor in solution (K d = 0.65 ± 0.090 μM; Supplementary Fig. 4a). Although Prc bound the cytoplasmic region of VgrS that contained the HAMP and transmitter regions, the binding affinity was too weak to have any physiological significance (K d = 191 μM; Supplementary  Fig. 4b and 4c). The difference in K d values detected by SPR and MST may be caused by the nature of these two approaches: SPR detects Prc-VgrS sensor interaction on a solid, CM5 sensor chip in which Prc is embedded, whereas the MST-associated binding event was directly measured in a solution, which is closer to the cellular environment.
To determine the region in Prc that binds the VgrS sensor, three recombinant Prc S475A proteins, each with a deletion of one of the domains (PDZ, PEP or DUF3340), were obtained and purified ( Supplementary Fig. 4d). However, SPR analyses showed that none of these proteins bound the VgrS sensor (data not shown). A subsequent analysis revealed that the VgrS sensor bound a recombinant Prc S475A containing peptidase and DUF regions with a K d value of 22.4 μM (Fig. 3f). An MST analysis also confirmed that the VgrS sensor bound either the recombinant Prc S475A containing peptidase and DUF regions (K d = 0.27 ± 0.028 μM; Supplementary Fig. 4e) or another recombinant Prc S475A in which the PDZ domain was deleted (K d = 1.95 ± 0.113 μM; Supplementary Fig. 4f). Thus, the VgrS sensor bound the peptidase-DUF region of the Prc monomer rather than the PDZ domain that usually acts as a docking site in protein-protein interactions 27 .
Prc cleaves the VgrS sensor to inhibit autokinase activity. VgrS is an HK with autokinase activity that is controlled by its periplasmic sensor region 25 . The Prc binding to the VgrS sensor region suggests that Prc regulates the latter's autophosphorylation. Thus, an in vitro phosphorylation assay was To further investigate whether the VgrS sensor is a proteolytic substrate of Prc in vitro, the mature Prc protein was co-incubated with a purified, recombinant VgrS sensor. As shown in Fig. 4d, within 60 min, Prc cleaved the VgrS sensor into two parts and then completely degraded the sensor in vitro. However, when the inactive form of Prc (Prc S475A ) was added to the reaction mixture, the VgrS sensor remained intact during the co-incubation period ( Fig. 4e). Furthermore, to determine whether this proteolytic process took place in vivo, recombinant strains in the ΔvgrS mutant and the ΔprcΔvgrS double-mutant genetic backgrounds were constructed (ΔvgrS-vgrS HA and ΔprcΔvgrS-vgrS HA , respectively; Supplementary Table 1). In the two strains, an HA-epitope tag was fused to the N-terminus of VgrS (between the 6th and 7th aa). Chloramphenicol was added to the bacterial cultures of the two strains to eliminate interference by de novo protein synthesis. Then, the bacterial cells were separately challenged with osmostress (1.0 M sorbitol). As revealed by western blotting analysis (Fig. 4f), after 5 min of osmostress stimulation, the amount of the N-terminal HA-tag, which was detected by the HA monoclonal antibody, gradually decreased as osmostress treatment progressed (Fig. 4f, upper panel, Lanes 6-10). Additionally, western blotting using a polyclonal antibody against VgrS revealed that VgrS was cleaved, and additional bands were detected (Fig. 4f, middle panel, Lanes 6-10). Thus, the N-terminal HA tag was removed from VgrS by Prc cleavage. In the double mutant, in which prc was deleted, the amount of HA-tag remained stable regardless of the presence or absence of osmostress stimulation (Fig. 4f,     To determine the exact, primary cleavage sites in the VgrS sensor processed by Prc, active Prc and VgrS sensor proteins were co-incubated, and then a matrix-assisted laser desorption/ionization time of flight mass spectrometry (MALDI-TOF-MS/MS) analysis was used to identify the VgrS sensor fragments. A peptide with a mass-to-charge ratio of 2771.5 was obtained (Fig. 5a, upper panel), and subsequent MS/MS analysis revealed that Prc cleaved the VgrS sensor at the site between Ala 9 and Gln 10 after 10 min of co-incubation (Fig. 5b, c, Supplementary Fig. 5). In the negative control experiment, the site remained intact when the inactive form of Prc S475A was added to the reaction mixture (Fig. 5a, middle panel). To verify this result in vitro, a recombinant VgrS sensor with the Ala 9 and Gln 10 sites substituted (Sensor A9G-Q10A ) was obtained and purified. MALDI-TOF-MS/MS analysis revealed that the Sensor A9G-Q10A protein was not cleaved by active Prc because no corresponding fragment was identified (Fig. 5a, lower panel). To verify this cleavage site in vivo, recombinant bacterial strains that encoded VgrS, in which the two sites were substituted in the ΔvgrS-vgrS HA and ΔprcΔvgrS-vgrS HA genetic backgrounds were constructed (ΔvgrS-vgrS HA-A9G-Q10A and ΔprcΔvgrS-vgrS HA-A9G-Q10A , respectively; Supplementary Table 1). As shown in Fig. 5d, in the two recombinant strains, the N-terminal HA-tag of VgrS remained stable regardless of osmostress treatment (Fig. 5d, Lanes 1-5 and Lanes 6-10), indicating that recombinant VgrS A9G-Q10A resisted Prc cleavage in vivo. As previously shown in Fig. 4a, b, after Prc cleavage, the difference in the theoretical molecular weights between the fulllength and the truncated VgrS was~1.0 kDa, and the two forms were hard to discriminate by SDS-PAGE analysis. Thus, we employed high-resolution, QTRAP LC-MS/MS analysis to directly detect the N-terminal proteolytic peptide generated by Prc cleavage. An 8-aa peptide (NRNIDFFA, corresponding to 2nd to 9th aa of VgrS) was chemically synthesized and used as a standard. As Fig. 5e and Supplementary Fig. 6 show, after a 5-min treatment of the full-length VgrS IMV by Prc, QTRAP LC-MS/ MS analysis detected the NRNIDFFA peptide in the reaction mixture. The amounts of this peptide gradually increased along with the treatment time ( Fig. 5e and Supplementary Fig. 6). These biochemical results demonstrated that Prc primarily cleaved the VgrS sensor between the Ala 9 ↓Gln 10 site in vivo and in vitro.
Furthermore, to measure the effect of Ala 9 ↓Gln 10 cleavage on the autokinase activity of VgrS, we obtained and purified a truncated, recombinant VgrS protein (VgrS Δ9 , embedded in IMV) containing a deletion from the 2nd to the 9th aa at the Nterminus. An in vitro phosphorylation assay then revealed that the autophosphorylation level of the VgrS Δ9 IMV was significantly decreased relative to that of the full-length VgrS IMV (Fig. 5f, Lane 2). As controls, two truncated VgrS IMV proteins (VgrS Δ72 and VgrS Δ84 , which were truncated in the sensor region) retained phosphorylation levels that were similar or even + + + + + + + + + VgrS H186A (IMV)  higher than that of the WT VgrS protein (Fig. 5f, Lanes 4 and 5), while the autokinase activity of VgrS Δ58 was also substantially decreased ( Fig. 5f, Lane 3). In addition, the recombinant VgrS A9G-Q10A protein had autokinase activity similar to that of the WT VgrS (Fig. 5g). Together with the data shown in Fig. 4a, the results of the phosphorylation analyses demonstrate that Ala 9 ↓Gln 10 of VgrS was the primary cleavage site of the Prc protease. Prc proteolysis resulted in a substantial decrease in the VgrS autophosphorylation level.
Cleavage of VgrS promotes bacterial osmostress resistance. The cleavage of a bacterial HK by a protease has not been previously reported, and the adaptive significance of the signalling process remains unclear. To genetically investigate the consequence of VgrS sensor cleavage on bacterial resistance to osmostress, an inframe deletion mutant of vgrS (vgrS Δ9 ), in which the coding sequence from the 2nd to the 9th aa of the chromosomal vgrS was deleted (Fig. 5b), was constructed to mimic the truncated protein product cleaved by Prc. Without osmostress, the mutant vgrS Δ9 exhibited a growth rate similar to that of the WT strain (Fig. 6a). However, when this bacterial mutant was challenged by osmostress (1.0 M sorbitol), it grew faster than the WT strain (Fig. 6b). Under osmostress, genetic complementation, in which a plasmid-borne, full-length vgrS was provided to this vgrS Δ9 mutant, decreased bacterial growth to the WT level (Fig. 6b). In addition, the growth of a bacterial strain encoding a recombinant VgrS A9G-Q10A , which is resistant to Prc cleavage (vgrS A9G-Q10A ), was significantly reduced under stress conditions (Fig. 6d). Thus, cleavage of the VgrS sensor by Prc may be beneficial for the growth of X. campestris pv. campestris under osmostress. Because Prc-catalysed VgrS cleavage resulted in a substantial decrease in its autokinase activity, we hypothesized that dephosphorylation of the VgrS-VgrR system promotes bacterial resistance to osmostress. As predicted, vgrS deletion mutants and point mutants of vgrS and vgrR, with substitutions in their conserved phosphorylating aa (vgrS H186A and vgrR D51A ,  respectively), also exhibited remarkably increased capabilities to adapt to osmostress (Fig. 6c-f). However, similar to the prc mutant, the vgrR deletion mutant was highly susceptible to sorbitol treatment (Fig. 6e, f), suggesting that intact, dephosphorylated VgrR is indispensable for regulating bacterial responses to stress. To determine the epistatic relationship between prc and vgrS, a set of double mutants was constructed in the Δprc genetic background by mutating vgrS. As shown in Fig. 6g, h, the inactivation of prc almost completely arrested bacterial growth under osmostress, but the growth rates of the double mutants Δprc-vgrS Δ9 , ΔprcΔvgrS and ΔprcΔvgrS-vgrS H186A were significantly restored towards the WT level. These epistatic analyses demonstrate that mutations causing the dephosphorylation of VgrS efficiently suppressed the growth deficiency of the prc mutation in resisting osmostress. Thus, vgrS was located downstream of prc regulation during the bacterial osmostress response. In addition to osmostress resistance, other phenotypic changes in the vgrS Δ9 mutant were measured. The virulence of this mutant against a susceptible host cabbage (B. oleraceae cv. Jingfeng No. 1) decreased slightly, while the resistance of the vgrS Δ9 mutant to iron stress and erythromycin was significantly decreased relative to the WT strain ( Supplementary Fig. 7). Thus, under osmostress, the phosphorylation of VgrS was detrimental to bacterial growth. Prc-catalysed cleavage of the VgrS sensor to dephosphorylate this TCS system was specifically beneficial to bacterial osmostress adaptation.
Prc controls the VgrR promoter-binding landscape. VgrS autophosphorylates and then transfers the phosphoryl group to the cognate VgrR to control its TF activity in binding promoters 25 . Because Prc cleaved the VgrS sensor and decreased its autokinase activity, we hypothesized that Prc will modulate VgrR-DNA-binding affinity.
Chromatin immunoprecipitation sequencing (ChIP-seq) was employed to compare the VgrR-binding promoters in the genome of X. campestris pv. campestris 8004. As shown in Fig. 7a and Supplementary Table 4, after being challenged by 1.0 M sorbitol, ChIP-seq analysis revealed that the VgrR of the WT strain was potentially bound to the promoters of 87 genes (GenBank GEO: GSE120292). In the Δprc mutant, VgrR bound the promoters of 105 genes. Only 41 VgrR-regulated genes were shared by the two strains under osmostress (Fig. 7a). prc may regulate VgrRpromoter binding, and, therefore, a prc deletion may cause changes in genome-scale binding events. The predicted VgrRbinding DNA motif is highly similar to that previously described (Fig. 7b) 25 . All of the identified VgrR-regulated genes were functionally classified into 16 categories (Fig. 7c, Supplementary  Table 4). Among them, the osmostress-related genes, such as those involved in lipopolysaccharide synthesis (XC_3814), lipoic acid synthesis (XC_0713) and transportation (XC_1619, XC_0969 and XC_4223), were absent in the Δprc mutant's dataset. Notably, The prc promoter was identified by ChIP-seq analysis (Supplementary Table 4). The promoter region of prc also contains a VgrR-binding motif that was then verified by an in vitro electrophoretic mobility shift assay (EMSA) (Fig. 7d). In the WT strain, the prc mRNA level was significantly induced after sorbitol stimulation (Fig. 7e), and in vivo ChIP, together with a quantitative PCR (ChIP-qPCR) assay, revealed that more VgrR was bound to the promoter regions of prc under stress conditions (Fig. 7f). However, in both vgrR and prc mutants containing 669-2,139-bp deletions of prc, prc (or remaining prc sequence, 497-665 bp) mRNA levels were significantly decreased, regardless of the osmostress stimulation (Fig. 7e). In the prc mutant, the level of VgrR occupancy on the prc promoter was significantly decreased, while genetic complementation significantly increased the level of VgrR occupancy (Fig. 7f). Thus, the transcription of prc was directly regulated by VgrR, which may form a feedback regulatory loop between vgrS-vgrR and prc.
We selected six genes from the ChIP-seq data, XC_0690 (encoding a sugar kinase), XC_0943 (conserved hypothetical protein), XC_2164 (YciE orthologue of E. coli that is involved in osmotic stress responses) 28 , XC_3300 (outer-membrane protein) XC_3301 (oxidoreductase), and XC_3576 (outer-membrane protein), for further functional investigation. As shown in Fig. 8a, independent deletions of XC_0943, XC_2164, XC_3300 and XC_3301 resulted in decreases in bacterial growth on NYG plus 1.0 M sorbitol plates, while the inactivation of XC_0690 and XC_3576 did not have any recognizable impacts, suggesting that XC_0943, XC_2164, XC_3300 and XC_3301 are involved in the   osmostress response. The EMSA revealed that VgrR physically bound to the promoter regions of these four genes (Fig. 8b, c, Supplementary Fig. 8). Competition upon the addition of various concentrations of unlabelled DNA probes gradually decreased the isotopic signals representing the VgrR-DNA complexes, demonstrating that VgrR directly binds the promoters of these genes. To investigate the role of Prc in controlling the transcription of these VgrR-regulated genes, a qRT-PCR analysis revealed that in the WT strain, sorbitol treatment significantly increased the expression levels of the four genes to 158% (XC_0943), 547% (yciE), 412% (XC_3300) and 174% (XC_3301) of the untreated sample (Fig. 8d, e, Supplementary Fig. 8). However, in the prc or vgrR mutants, stress-induced increases in the mRNA levels of these genes were not seen, whereas genetic complementation by overexpressing prc or vgrR fully restored deficiencies in the upregulation of the four genes (Fig. 8d, e, Supplementary Fig. 8). This result demonstrated that the inactivation of prc caused decreases in the mRNA levels of the four genes relative to the WT strain.
Prc control of the transcription levels of VgrR-regulated genes led to the hypothesis that in bacterial cells, proteolysis of VgrS by Prc decreased the phosphorylation level of VgrR, which modulated the TF activity of VgrR. MST analyses then revealed that unphosphorylated VgrR bound to the promoter regions of Pprc, PXC_0943, PXC_2164, PXC_3300 and PXC_3301 with K d values from 0.46 to 7.62 μM (Supplementary Fig. 9). However, when VgrR was phosphorylated by VgrS in the presence of ATP, VgrR-P did not bind to the promoter regions of Pprc, PXC_0943, PXC_2164 or PXC_3301, and the binding affinity between VgrR-P and PXC_3300 was decreased (K d = 30.7 μM). When inactive VgrS was added to the reaction, the binding affinities between VgrR and DNA were similar to those of unphosphorylated VgrR (Supplementary Fig. 9). Thus, unphosphorylated VgrR bound the promoter regions of downstream genes in vitro with higher affinity levels than those of phosphorylated VgrR.
To determine the changes in VgrR-DNA-binding affinities in vivo, ChIP-qPCR was used to estimate the levels of VgrR occupancy in the promoter regions of the selected genes, XC_0943 and yciE. As shown in Fig. 8f, g, after stimulation by osmostress, the amounts of promoter DNA bound by VgrR were significantly increased in the WT strain to 1110% (PXC_0943) and 362% (PyciE) relative to the unstimulated samples. However, in the prc mutant, the amounts of promoter DNA bound to VgrR after stress stimulation were significantly decreased relative to the levels in the WT strain. In the genetic complementation strain of the prc mutation, in which prc was overexpressed, levels of VgrR-DNA binding were restored to, or even higher than, WT levels ( Fig. 8f, g). Therefore, these genetic and biochemical analyses revealed that Prc positively regulates the binding between VgrR and the promoters of these osmostress response genes.

Discussion
Bacterial cells, except those of Mycoplasma spp., usually encode several to over a hundred HKs to detect various environmental stimuli 29,30 . Because HKs react to stimuli mainly through autophosphorylation and catalyse phosphoryl transfer to their cognate RRs, which control multiple cellular responses, regulation of the tempo and mode of HK phosphorylation is important for bacterial adaptation 31 . Chemical ligands, environmental cues, phosphatases (including cognate RRs), heterogeneous HKs and auxiliary proteins directly adjust the phosphorylation levels of HKs 32,33 . All of these regulatory modes are reversible in controlling HK phosphorylation. However, irreversible proteolytic modifications of HKs are less well-studied. In the present work, we demonstrated that under osmostress, a Prc protease of X. campestris pv. campestris specifically cleaved the periplasmic sensor region of an HK, VgrS, and inhibited its autokinase activity. Quenching VgrS phosphorylation then modulated the TF activity of VgrR to bind the promoters of downstream genes. In addition, the transcription of prc itself was subject to regulation by VgrR, which then formed a positive feedback loop between the protease and the RR. This signalling process significantly and specifically promoted bacterial resistance to osmostress (Fig. 9). Thus, proteolysis of the sensor region of HK appears to be a molecular mechanism for regulating bacterial TCS phosphorylation and has an unambiguous physiological impact on bacterial adaptation. Although studies investigating the relationship between proteases and HKs are lacking in bacteria, the proteolysis of Thr/Ser/ Tyr receptor kinases to modulate their activities has been extensively studied in eukaryotic cells 5,34,35 . In Proteobacteria, Prc/Tsp-family proteases are widely distributed. These proteases not only recognize and process the C-termini of protein substrates but also degrade the substrates into small peptides 36,37 . For example, the Prc of E. coli controls cell wall synthesis and antibiotic resistance by degrading peptidoglycan hydrolases or outer-membrane lipoproteins 38,39 . The results of our study indicated that the cleavage of HK sensors by proteases localized in periplasmic space, including Prc/Tsp-family (MEROPS S41Afamily) and HtrA-family proteases (MEROPS S1-family) 40 , is important for regulating bacterial adaptation because, in gramnegative bacteria, a majority of the HKs contain periplasmic sensors to detect environmental signals 33 . Compared with reversible dephosphorylation, proteolysis of the HK sensor holds advantages in regulating TCS signalling: cleavage takes place in the periplasm, rather than the cytoplasm, which is more efficient because it shortens the distance travelled by TM signals. In particular, irreversible proteolysis avoids the continuing phosphorylation of existing HKs, which is more effective if reversible phosphorylation is extremely detrimental to bacterial survival in certain circumstances 1,17 . However, more experimental studies are needed for the following reasons: (1) there is no consensus recognition or processing sequence for the Prc/Tsp-family and HtrA-family proteases, which makes it difficult to predict the physiological substrates of these proteases. (2) as our results revealed, proteolytic processing of the VgrS sensor by the Prc protease is specific and adaptive to the osmostress response (Fig. 6), suggesting that protease-HK interactions are highly specific to environmental signals. VgrS-VgrR is a pleiotropic TCS of X. campestris pv. campestris. It controls virulence, multiple-stress responses, bacterial growth and the capability of eliciting hypersensitivity reactions in nonhost plants 25,26 . The VgrS sensor contains an EPQE motif to directly bind ferric iron. In iron-depleted environments, VgrS maintains a high level of autophosphorylation and then phosphorylates VgrR to regulate the expression of virulence factors 25 . Substitutions of the conserved phosphorylation sites in VgrS (His 186 ) or VgrR (Asp 51 ) caused decreases in virulence against host cabbage. However, here, in a prc mutant that exhibited a growth deficiency under osmostress, various mutations of vgrS, including deletion of the N-terminal sequence that mimics cleavage by Prc, effectively suppressed the growth deficiency caused by the null mutation of prc (Fig. 6h). In addition, substitution of the conserved His 186 of VgrS or the Asp 51 of VgrR (VgrR D51A ) increased the bacterial growth rate under stress conditions (Fig. 6f). Thus, a low level of VgrS phosphorylation may promote bacterial resistance to osmostress, and the inhibition of VgrS autokinase activity by Prc proteolysis is an effective mechanism. In bacteria, a number of RRs are degraded by proteases in the cytosol, such as CtrA of Caulobacter crescentus and DegU of Bacillus subtilis 41,42 . Compared with the proteolysis of RRs, the cleavage of HKs by proteases may be more flexible in TCS regulation because of the following: (1) HK is a receptor, rather than an effector, in cell signalling 43 . Thus, degradation of the HK only impacts the phosphorylation level of the cognate RR (effector). This does not cause complete inactivation of the TCS because the dephosphorylated RR can continue to regulate physiological processes. (2) Bacteria usually encode paralogous TCSs that are generated by operon duplication during evolution 44 . These TCSs form regulatory networks, and the proteolysis of an HK reinforces the signalling of the existing compensatory pathways of these paralogous TCS. This is probably important for the VgrS-VgrR system because there are two functionally unknown, paralogous TCSs in X. campestris pv. campestris (XC_3126-XC_3125 and XC_3452-XC_3451) 25 . (3) The phosphatase activity of HK is essential for maintaining high levels of specificity between cognate HKs and RRs 45,46 . The proteolysis of an HK releases the inhibition on the illegitimate phosphorylation of the RR and then provides an opportunity for cross-talk among different TCSs. Future experiments are needed to verify these possibilities.
Our proteolytic experiment showed that Prc cleaved and degraded the purified VgrS sensor in a low-velocity in vitro experiment (~60 min, 28°C, Fig. 4d), which seemed slower than the cleavage of full-length VgrS in vivo (~5-10 min, 28°C, Fig. 4f). This difference indicated that the purified VgrS sensor may undergo a conformational change that interferes with Prc recognition in vitro. Alternatively, other adaptors or cellular factors are involved in Prc proteolysis within bacterial cells. Adaptors are proteins that bind proteases to help with the recognition of substrates or proteolysis. In E. coli, NlpI is an adaptor of Prc that helps facilitate the protease's degradation of its physiological substrate, MepS, an outer-membrane protein that is associated with peptidoglycan biosynthesis 47 . The presence of NlpI enhances Prc protease activity by 50-fold 48 . Our TAP analysis provided information on Prc-binding proteins, but it is difficult to predict which one is the Prc adaptor used during processing of the VgrS sensor because adaptors are diverse in structure and include lipoproteins, RRs and arginine kinases 49 . In addition, how the Prc protease is activated to cleave the VgrS sensor remains unclear. Our preliminary experiment showed that a high concentration of sorbitol (200 mM to 1.0 M) did not affect the enzymatic activity of Prc, indicating that Prc itself is not a sensor that directly detects osmostress. As Fig. 3c revealed 48 , Prc exists as a monomer or trimer in vitro, and only monomeric Prc bound and degraded the VgrS sensor (Fig. 3c, e). This implied that alteration of the quaternary structure of Prc might act as a molecular mechanism to activate its protease activity in vivo.
In addition, the regulatory functions of the proteolytic products of VgrS are unknown. In eukaryotes, fragments of receptor kinases generated by proteases are usually stably maintained in cells 50,51 . These fragments serve as dominant negative factors, decoy receptors or even cellular signals 52,53 . As Fig. 5 and Supplementary Fig. 6 show, Prc proteolysis produced a short Nterminal peptide and a truncated VgrS. The two fragments were detected by western blotting and MS analysis, and they were relatively stable. In TCS regulation, the majority of HKs constitutively form homodimers that are autophosphorylated in a trans manner 33 . The presence of a truncated, "decoy" VgrS may disturb this process. Furthermore, if VgrS and its paralogous HKs (XC_3125 and XC_3451) constitute regulatory heterodimers, VgrS proteolysis could affect the signalling of other TCSs. The biological roles and fates of the two fragments remain unclear. Future investigations are needed to study the activating mechanism of Prc under osmostress and the regulatory roles of HK fragments. The results will provide insights into the cross-talk between TCS signalling and regulation through proteolysis.

Methods
Bacterial strains and growth conditions. Strains used in this study are listed in Supplementary Table 1. E. coli strain BL21 (DE3) (Novagen) was used to express recombinant proteins, and E. coli strain DH5α (Lab collection) was used for molecular cloning. The E. coli strain was cultured at 37°C in Luria-Bertani medium supplemented with the appropriate antibiotics. Xanthomonas campestris pv. campestris 8004 was cultured at 28°C in NYG medium with the appropriate antibiotics. Xanthomonas oryzea pv. oryzea PXO99 was cultured at 28°C in PS medium with the appropriate antibiotics. The following concentrations of antibiotics were used: 100 μg/ml ampicillin, 50 μg/ml kanamycin, 100 μg/ml spectinomycin, 34 μg/ml chloromycetin and 25 μg/ml rifampicin. The transformation of bacterial competent cells was performed according to previous studies 18,25 . Bacterial electro-competent cells were prepared by extensively washing fresh bacterial cells for three times with ice-cold glycerol (10%). Transformation condition of bacterial competent cells were set as 1.6 kV cm −1 to 2.0 kV cm −1 and conducted in a Bio-Rad Pulser XCell (Bio-Rad, USA).

Construction of bacterial recombinant strains.
In-frame deletion mutants and recombinant strains of X. campestris pv. campestris were constructed using the homologous, double-cross-over method with the suicide vector pK18mobsacB as described in a previous study 25 . Briefly, the upstream and downstream genomic sequences of the region to be deleted were amplified using primers listed in Supplementary Table 2. After restriction enzyme digesting and ligating into pK18mobsacB, the recombinant plasmid was electroporated into competent cells to generate single-cross-over mutants by selection on NYG plates containing kanamycin. The putative single-cross-over mutants were confirmed by PCR, cultured in NYG medium for second-round homologous cross-over, and grown on NYG plates containing 10% sucrose to select in-frame deletion mutants, which were further verified by PCR and subsequent sequencing. Genetic complementation was achieved in trans by providing full-length genes using the broad host vector pHM1.
Phenotypic characterization. Plant inoculations and virulence assays were performed as described previously 26 . Briefly, strains of X. campestris pv. campestris were inoculated into susceptible 4-week-old cabbage (Brassica oleraceae cv. Jing-Feng 1) hosts by clipping the leaves with pre-sterilized scissors dipped into the bacterial culture. The lesion level was measured at 10 d after inoculation with at least 12 repetitions. Virulence level was scored by a semi-quantitative standard as previously reported 19,54 : 0-level, no visible effect; 1-level, limited chlorosis around the cut site; 2-level, chlorosis extending from the cut site; 3-level, blackened leaf veins, death and drying of tissue within the chlorotic area; 4-level, extensive vein blackening, death and drying of tissue. For in planta growth assay, bacterial strains were infiltrated into plant leaves by a 1-ml syringe and leave disks (diameter = 0.8 cm) were periodically collected, ground in 1 ml NYG medium, and serially diluted before plating and colony counting. For each assay, bacterial population in six leave disks was determined. Bacterial stress resistance assays were performed on NYG plates supplemented with 2.5 mM FeSO 4 (plus 5 mM vitamin C) and 1.0 M sorbitol. Strains were cultured to 10 6 CFU/ml (OD 600 = 0.4) in NYG medium with the appropriate antibiotics, and then 1 μl of 10-fold serial-dilution cultures were inoculated onto the plates and grown for 2-3 d at 28°C.
To determine the bacterial sensitivity to antibiotics (erythromycin and kanamycin), strains were cultured to 10 6 CFU/ml (OD 600 = 0.4) in NYG medium. Then, 1 ml of culture was mixed with 30 ml of melted NYG agar plate, just before the agar solidified. After the plate was fully solidified, filter paper disks (Φ = 0.6 mm) containing 5 μl erythromycin (20 μg/ml) or kanamycin (50 μg/ml) were placed on the plate to inhibit bacterial growth. Minimal inhibition concentration determinations were performed using the standard serial-dilution method 55 .
Protein expression and purification. The pET System (Novagen, Madison, WI, USA) was used for the cloning and expression of recombinant proteins in E. coli. Recombinant vectors used to express proteins are listed in Supplementary Table 1. Induction, expression, and purification were performed using affinity chromatography with Ni-NTA agarose beads (Novagen, USA) according to the manufacturer's manual. Purified His 6 -tagged proteins were concentrated with Amicon Ultra-4 and -15 Centrifugal Filter Units (Merck Millipore, Germany) and dissolved in storage buffer (50 mM Tris-HCl, pH 8.0, 0.5 mM EDTA, 50 mM NaCl, and 5% glycerol) before use.
Gel filtration chromatography. The purified proteins were concentrated with Amicon Ultra-4 and -15 Centrifugal Filter Units and subsequently subjected to gel filtration with Fast Protein Liquid Chromatography AKTA Purifier 10 with Frac-900 (GE Healthcare, USA). The ATKA system was pre-equilibrated with 20 mM Tris-HCl, pH 8.0, at a flow rate of 0.5 ml/min and then applied to a Superdex 75 10/300 GL column to separate dimers and oligomers from the monomer. The elution profiles were collected at A 280 and confirmed by SDS-PAGE gels and western blots.
Analytical ultracentrifugation. Analytical ultracentrifugation was performed at 20°C in a ProteomeLab XL-1 analytical ultracentrifuge (Beckman, USA) equipped with absorbance optics and an An60-Ti rotor. The sedimentation velocity analysis of the affinity chromatography-purified Prc S475A was conducted at 240,000 g using double sector cells, and data were collected at the concentration of A 280 = 0.7 in sodium phosphate buffer (pH 8.0) plus 150 mM NaCl for each peak sample. These values were normalized to standard conditions by correcting for buffer density and viscosity. Interference sedimentation coefficient distributions were calculated from the sedimentation velocity data using the SEDFIT software program (www. analyticalultracentrifugation.com).

Subcellular cell fraction.
To prepare the total cell protein fraction, the bacterial cells were collected and resuspended in 1X SDS-PAGE loading buffer (50 mM Tris-HCl, pH 6.8, 2% SDS, 0.1% bromophenol blue, 10% glycerol and 100 mM DTT) and immediately boiled for 3 min to denature the protein.
To prepare the periplasmic fraction, the bacterial cell pellets were thoroughly resuspended in Buffer A (30 mM Tris-HCl, pH 8.0, 20% sucrose and 1 mM EDTA). After rotating at room temperature for 10 min, the cells were centrifuged and the cell pellet was resuspended in ice-cold Buffer B (5 mM MgSO 4 ). After rotating at 4°C slowly for 10 min on ice, the periplasmic proteins were released into the buffer. Finally, the periplasmic fraction was concentrated by ultrafiltration. To collect the soluble cytoplasmic fraction, the bacterial cells were collected and resuspended in protein lysis buffer [50 mM Tris-HCl, pH 7.5, 0.5 M NaCl, 10% glycerol, 0.1% Triton X-100 and 1 mM phenylmethylsulfonylfluoride (PMSF)]. After sonicating, the supernatant represented the cytoplasmic fraction.
To isolate the inverted membrane vesicles, the bacterial cells were collected, resuspended and sonicated in a low-salt buffer (100 mM sodium phosphate, pH 7.0, 10% glycerol, 5 mM EDTA, 10 mM DTT, 1 mM PMSF). The inverted membrane vesicles were collected by ultracentrifugation (150,000 × g for 60 min at 4°C) before removing the unbroken cells and debris through two rounds of centrifugation (8000 × g for 10 min at 4°C). After washing with high salt buffer (20 mM sodium phosphate, pH 7.0, 2 M KCl, 10% glycerol, 5 mM EDTA, 5 mM DTT and 1 mM PMSF), the inverted membrane vesicles were resuspended in storage buffer (20 mM Tris-HCl, pH 7.5 and 10% glycerol).
Enzymatic and endo-protease activity assays. The qualitative enzymatic activity assay was performed in the enzymatic reaction buffer (50 mM Tris-HCl, pH 7.5, 2 mM DTT, 25 mM NaCl, 25 mM KCl and 5 mM MgCl 2 ) with the proper amounts of protease and substrate co-incubated at 28°C for the indicated time, or at a different temperature, DTT concentration, or pH value for 5 min. The reactions were stopped by adding SDS-PAGE loading buffer before loading the samples into a 12% PAGE gel and performing the electrophoresis. The gel was stained with Coomassie brilliant blue to determine the remaining substrate.
The quantitative endo-protease activity assay was carried out using azocasein (Megazyme) as the substrate. The pre-equilibrated protease (5 μM) was coincubated with pre-equilibrated azocasein (400 μM) in Buffer A (100 mM sodium phosphate). The solution was fully mixed and incubated for 10 min at 37°C. The enzymatic reaction was terminated by the addition of 2.5 times the volume of 5% trichloroacetic acid to precipitate the non-hydrolyzed azocasein. The reaction was vigorously stirred for 5 s and stood for 5 min at room temperature. Then, it was centrifuged at 3000 × g for 10 min, and the absorbance of the supernatant solutions at 440 nm was determined. The samples, minus the protease, which was substituted with Buffer A, served as blanks.
To measure the enzyme kinetics of the protease, the same amount of protease was co-incubated with different amounts of azocasein in Buffer A (100 mM sodium phosphate). The subsequent procedure was performed in the same manner as the qualitative endo-protease activity assay. The K M , which is the Michaelis-Menten constant, and V max , which is the maximum reaction velocity, values were determined using the Lineweaver-Burk plot method.
TAP and MS identification. TAP was carried out following the protocol of the FLAG HA Tandem Affinity Purification Kit (Sigma-Aldrich). Briefly, the bacterial cells were collected and ground before being resuspended in lysis buffer [50 mM Tris-HCl, pH 8.0, 0.15 M NaCl, 1 mM EDTA, pH 8.0, 1 mM PMSF and 1 tablet/20 ml protease inhibitor cocktail tablets (Roche)]. They were centrifuged at 13,000 × g for 20 min at 4°C, and the supernatant represented the whole cell lysate. Prewashed ANTI-FLAG M2 resin was added to the sample lysate and incubated from 2 h to overnight at 4°C with slow shaking. For each wash, the resin was gently agitated in the lysis buffer, centrifuged at 3000 × g for 1 min, and then the remaining final wash volume was decanted without losing any resin. The resin-protein complex was washed with lysis buffer using three rounds of lowspeed centrifugation. The remaining resin was transferred into a spin column, and 2.5 volume of 3X FLAG peptide (150 ng/µl) was added and co-incubated for at least 10 min at 4°C. The column was spun before removing the tip to a clean microcentrifuge tube and keeping the eluate, which contained the eluted FLAG-protein complex. The elution process was repeated twice. Then, 40 μl pre-washed ANTI-HA resin in lysis buffer was added and incubated 30 min to 2 h at 4°C with slowly shaking. The supernatant was removed after the incubation, and the ANTI-HA resin-protein complex was washed with lysis buffer through three rounds of lowspeed centrifugation. Then, 50 μl 8 M urea was added and co-incubated for a minimum of 10 min at room temperature. The sample was centrifuged at 3000 × g for 1 min, and the eluate was loaded and separated on a 12% SDS-PAGE gel. After silver staining, the differential gel bands between sample and control were manually excised, and each band then underwent enzymatic digestion.
The protein bands in the gel were further cut into small plugs, washed in 100 µl freshly made 30 mM K 3 Fe(CN) 6 and 100 mM Na 2 S 2 O 3 (1:1, v-v) buffer for seconds, rinsed with 25 mM NH 4 HCO 3 , dehydrated in 100% acetonitrile for 10 min, and dried in a SpeedVac (Labconco) for 15 min. The reduction procedure was performed in solution A (10 mM DTT and 25 mM NH 4 HCO 3 ) for 45 min at 56°C, and then, the alkylation procedure was performed in solution B (40 mM iodoacetamide and 25 mM NH 4 HCO 3 ) for 45 min at room temperature in the dark. The gel plugs were washed with 50% acetonitrile in 25 mM ammonium bicarbonate twice. The gel plugs were then dried in the SpeedVac for 15 min and digested with sequence-grade modified trypsin (50 ng per band) in 25 mM NH 4 HCO 3 overnight at 37°C. Formic acid was added to a 1% final concentration to stop the enzymatic reaction before transferring the solution to a sample vial for subsequent LC-MS/MS analysis.
The nanoLC-MS/MS identification of proteins was performed on a Thermo Finnigan LTQ linear ion trap mass spectrometer in line with a Thermo Finnigan Surveyor MS Pump Plus HPLC system. Tryptic peptides were loaded onto a trap column (300 SB-C18, 5 × 0.3 mm, 5-μm particle; Agilent Technologies, Santa Clara, CA, USA). The peptides were eluted over gradient solution C (80% acetonitrile and 0.1% formic acid) at a flow rate of 500 nl/min and introduced into the online linear ion trap mass spectrometer (Thermo Fisher Corporation, San Jose, CA, USA) using nano electrospray ionization. The five most abundant ions (one microscan per spectra) were selected for fragmentation from a full-scan mass spectrum by collision-induced dissociation for data-dependent scanning.
MS data were analysed with SEQUEST against NCBI's X. campestris pv. campestris 8004 protein database and displayed with Bioworks 3.2. Peptides with +1, +2, or +3 charge states and with cross correlations of ≥1.90, >2.5, and >3.0, respectively, were accepted. Carbamidomethylation on cysteine and oxidation on methionine were selected as residue modifications. SEQUEST was searched with a peptide tolerance of 3.0 Amu and a fragment of 1.0 Amu.
Microscale thermophoresis (MST) measurement. The MST measurement was used for detecting protein-protein interactions. Briefly, 10 μM purified VgrS sensor was labelled with a Monolith NT Protein Labeling Kit RED-NHS (Nano Temper Technologies GMBH, München, Germany) using red fluorescent dye NT-647 N-hydroxysuccinimide (amine-reactive) according to the manufacturer's instructions. The excess labelling reagents were removed by buffer-exchange column chromatography, the labelled VgrS sensor was eluted in NTA buffer (300 mM NaCl and 50 mM sodium phosphate buffer, pH 7.0). The binding assays were performed on a Monolith NT.115 Microscale Thermophoresis instrument (Nano Temper Technologies GMBH) using standardly treated capillaries. Equal amounts of labelled protein were titrated by the purified Prc S475A -monomer, PrcΔPDZ S475A , and PEP + DUFΔ20 S475A , which were exchanged into the 1X NTA buffer with 0.05% Tween, using a 1:1-series dilution method. The curves were fitted by Nano Temper Analysis Version 1.5.41 from three replicates, and the value of dissociation constant (K d ) was calculated.
MST was also used to quantify the VgrR-DNA binding affinities. 5′-FAMlabelled oligonucleotide primers for downstream genes were synthesized and used in PCR. The labelled double-stranded DNA product was added to serially diluted protein reaction volumes (at an initial concentration of 40 µM) containing 50 mM Tris-HCl (pH 7.4), 150 mM NaCl, 10 mM MgCl 2 and 0.05% (v/v) Tween-20. The KD Fit function of NanoTemper Analysis software version 1.5.41 was used for curve fitting and calculation of the value of the dissociation constant (K d ).
SPR. The binding kinetics of protein-protein interactions were estimated by the SPR assay, which was performed using a Biacore 3000 (GE Healthcare) at 25°C. The running buffer was PBST (0.27 g/L KH 2 PO 4 , 1.42 g/L Na 2 HPO 4 , 8 g/L NaCl, 0.2 g/L KCl, pH 7.4 and 0.005% Tween-20). As the manufacturer described, one flow cell of a CM5 sensor chip was activated with a mixture of 0.2 M N-(3dimethylaminopropyl)-N'-ethylcarbodiimide hydrochloride and 0.05 M N-hydroxysuccinimide in ddH 2 O. The ligand protein, in 10 mM sodium acetate (pH 4.5), was injected over the flow cell for 10 min at a flow rate of 10 μl/min. The remaining binding sites were blocked by 1 M ethanolamine (pH 8.5). In total, 3000 response units of the ligand protein were immobilized. Protein at different concentrations (double dilution) was injected into the immobilized protein and blank flow cells for 5 min at a flow rate of 30 μl/min. The bound protein was removed with 0.01% SDS for 30 s after the 5-min dissociation phase. The curves were fitted to the 1:1 Langmuir binding model or steady state affinity model (BIA evaluation 4.1 software) to obtain the equilibrium and kinetic constants.
RT-PCR, qRT-PCR and western blotting. The total RNA was extracted using TRIzol reagent (Invitrogen, Carlsbad, CA, USA), according to the manufacturer's instructions. RNA concentrations were determined using a NanoDrop 1000 spectrophotometer (Thermo Scientific, Wilmington, DE, USA). Then, 1 μg total bacterial RNA was treated with RNase-free DNAse (Ambion, Austin, TX, USA.) to remove contaminating DNA, and the first strand cDNA was synthesized with random primers (Promega, Madison, WI, USA) using Superscript III reverse transcriptase (Invitrogen).
To analyse the Prc operon's structure, with cDNAs as templates, primers were used to amplify the possible transcripts from the intergenic regions between genes. Real-time quantitative reverse transcription PCR (qRT-PCR) was used to quantify the mRNA levels. qRT-PCR was performed using Maxima SYBR Green (Fermentas, Glen Burnie, MD, USA) with a CFX96 real-time PCR detection system (Bio-Rad, Hercules, CA, USA). The transcript level of 16 s rRNA was amplified as the internal control. The qRT-PCR conditions were as follows: 95°C for 10 min; 44 cycles of 95°C for 15 s, 60°C for 30 s, and 7°C for 30 s; and 95°C for 15 s. The melt curve was performed at 65-95°C, in increments of 0.5°C for 5 s.
Western blotting was conducted by transferring the proteins onto PVDF membrane (Millipore, USA). Polyclonal antiserum of VgrS or monoclonal antibodies of HA (M20003-L) and His 6 tags (M20001L, Abmart, China) were used to detect the proteins. All of the antibodies were diluted 5000-10,000-folds before use.
Identification and quantification of proteolytic peptide by QTRAP LC-MS/ MS. For detecting the N-terminal proteolytic peptide of VgrS, Prc-treated VgrS inverted membrane vesicles were collected at 0, 5, 10, 30, 60, 120 and 180 min and immediately subjected to liquid nitrogen for fixing the proteolytic status. After centrifuging the mixer with 3K Ultrafiltration devices (Pall Corporation, USA) at 11,000 × g for 10 min, the peptides below 3 kDa were concentrated for further analyses. The chemically synthesized VgrS N-terminal peptide (2nd to 9th aa) was used as a standard. The samples were analyzed on an Agilent 1200 pump LC system (Agilent Technologies, Santa Clara, CA, USA) equipped with a Synergy 4 μM Hydro-RP 80 Å LC column (150 × 2 mm, Phenomenex), which is coupled to a QTRAP 4500 Mass Spectrometer instrument with a TurboIonSpray ionization source (Applied Biosystems, AB SCIEX, CA, USA). For all the separations, the flow rate of the mobile phase was 400 μL/min., and eluent A and B of the mobile phase was water and 100% methanol, respectively. The injection volume was 3 μL. The gradient separation and elution was as follow: time = 0 min: 98% A, 2% B; 3 min: 98% A, 2% B; 24 min: 0% A, 100% B; 28 min: 0% A, 100% B; 29 min: 98% A, 2% B and 32 min: 98% A, 2% B. The ESI source was operated in positive ionization mode with dwell time of 150 ms for MRM analysis (461/120). The MS acquisition with the ESI interface was performed under the following condition: ion spray voltage, 4500 V; curtain gas 20 (arbitrary units); GS1 (Nebulizer Gas) and GS2 (Heater Gas), 60 psi, respectively; probe temperature, 500°C. The selected reaction monitoring (SRM) measurements were performed using declustering potential (DP) of 80 V, entrance potential (EP) of 10 V, collision cell exit potential (CXP) of 13 V, and collision energy (CE) values of 18 optimized by the software. Evaluation of the data was performed with Analyst software (v.1.6.1) and the relative and quantitative amount of each sample was determined with Analyte Peak Area (counts).
In vitro phosphorylation assay. In the in vitro autokinase activity assay, the inverted membrane vesicles containing full-length wide-type or mutated VgrS were incubated with or without Prc or Prc S475A for 10 min at 28°C, then incubated with 1 μl ATP containing 10 μCi [γ-32 P]-ATP (PerkinElmer) in 20 μl autophosphorylation buffer (50 mM Tris-HCl, pH 7.8, 2 mM DTT, 25 mM NaCl, 25 mM KCl and 5 mM MgCl 2 ) for 2 min. The phosphorylation reaction was stopped using 5X SDS-PAGE loading buffer. Ten micrometers of VgrR was added for phosphotransfer analysis if needed. The phosphorylated proteins were separated by 12% SDS-PAGE and exposed to a phosphor screen for 2-3 h. After being scanned using a PhosphorImage system (GE Healthcare, Chicago, IL USA), the gels were stained with Coomassie brilliant blue to determine the location and amount of the autokinase protein.
ChIP-seq and ChIP-qPCR. ChIP-seq was performed as previously reported 25,56 . Briefly, the bacterial strains grown in NYG medium until OD 600 = 0.8 ± 0.05 were treated with or without 800 mM sorbitol for 5 min. After collection by