The opgC gene is required for OPGs succinylation and is osmoregulated through RcsCDB and EnvZ/OmpR in the phytopathogen Dickeya dadantii

Osmoregulated periplasmic glucans (OPGs) are a family of periplasmic oligosaccharides found in the envelope of most Proteobacteria. They are required for virulence of zoo- and phyto-pathogens. The glucose backbone of OPGs is substituted by various kinds of molecules depending on the species, O-succinyl residues being the most widely distributed. In our model, Dickeya dadantii, a phytopathogenic bacteria causing soft rot disease in a wide range of plant species, the backbone of OPGs is substituted by O-succinyl residues in media of high osmolarity and by O-acetyl residues whatever the osmolarity. The opgC gene encoding a transmembrane protein required for the succinylation of the OPGs in D. dadantii was found after an in silico search of a gene encoding a protein with the main characteristics recovered in the two previously characterized OpgC of E. coli and R. sphaeroides, i.e. 10 transmembrane segments and one acyl-transferase domain. Characterization of the opgC gene revealed that high osmolarity expression of the succinyl transferase is controlled by both the EnvZ-OmpR and RcsCDB phosphorelay systems. The loss of O-succinyl residue did not affect the virulence of D. dadantii, suggesting that only the glucose backbone of OPGs is required for virulence.

In D. dadantii, opgG and opgH mutant strains are completely devoid of OPGs and exhibit pleiotropic phenotypes including a total loss of virulence, loss of motility, increased synthesis of exopolysaccharides and induction of a general stress response indicating impairment in the perception of the environment 12,13 . Most of these phenotypic characteristics are the result of the requirement of OPGs to achieve a low level of activation of the RcsCDB two-component signalling pathways [14][15][16] .
Two-component systems (also called phosphorelays) are the key to gene expression plasticity in response to environmental variations. The paradigm of the two-component systems is the EnvZ/OmpR system of Escherichia coli. Under increasing osmolarity, a sensor histidine kinase, EnvZ, autophosphorylates and transfers its phosphate group to its cognate regulator, OmpR, which in turn regulates the expression of different target genes involved in the osmotic stress responses 17 . Actually, the only known role of EnvZ/OmpR in D. dadantii is the regulation of two oligogalacturonate outer membrane channels 18 .
The RcsCDB phosphorelay is restricted to Enterobacteria. Under various stimuli, one of them being increasing osmolarity, RcsC, the transmembrane kinase, phosphorylates the RcsB cytoplasmic regulator via the transmembrane protein RcsD. In several Enterobacteria, such as D. dadantii, it was shown that activation of the RcsCDB system enhanced exopolysaccharide synthesis, cell division and decreased virulence and motility 15,[19][20][21] .
The following study was conducted to identify the gene responsible for the succinylation of the OPGs in D. dadantii and the regulation of its expression. We show that the regulation of this gene by osmolarity variation occurs through both the RcsCDB and the EnvZ/OmpR phosphorelay systems and that only the glucose backbone of OPGs is involved in virulence.

ABF-0020261 (opgC) encodes the O-succinyl-transferase of the Osmoregulated Periplasmic Glucans (OPGs).
OpgC, annotated as a membrane acyl-transferase, was first identified in the non-pathogen bacteria E. coli K12 22 and then in Rhodobacter sphaeroides 23 . In our attempt to identify the opgC gene in Dickeya dadantii (opgC DD ), we first looked unsuccessfully for an homologous gene of opgC from E. coli (opgC EC ) or R. sphaeroides (opgC RS ) in the D. dadantii genome. We then looked for a gene encoding a protein with the two main characteristics found in the OpgC of E. coli and R. sphaeroides, i.e. 10 transmembrane segments and an acyl-transferase domain. We found 72 genes with an acyl-transferase domain, including 4 genes with 9 or 10 transmembrane domains (Table S1). These 4 genes were inactivated and OPGs of each of them were extracted. The presence of succinate on OPGs was analyzed by direct quantification of succinate after desubstitution by alkaline treatment of OPGs, then confirmed by mass spectrometry on native OPGs (see Experimental procedures). As expected for the wild-type strain (EC3937), no succinic acid could be detected in low osmolarity medium, while a total of 21.58 ± 3 μ g of succinic acid per mg of glucose was found on OPGs at high osmolarity. Three mutants, ABF-0020770 (NFB7306), ABF-0016719 (NFB7307), and ABF-0018319 (NFB7435), displayed a similar profile to the wild-type, respectively 23.38 ± 7.63, 28.5 ± 5.1 and 24.43 ± 3.1 μ g of succinic acid per mg of glucose at high osmolarity. In the last mutant, ABF-0020261 (NFB7305), no succinic acid could be detected regardless of the osmolarity. To confirm that the inactivation of the ABF-0020261 gene provokes the lost of succinyl substitution of OPGs, a single ectopic copy of the ABF-0020261 wild-type gene was introduced by Tn5 transposon into the mutant (NFB7305). The complemented strain (NFB7326) restored succinyl substitution of OPGs by succinyl residues only at high-osmolarity: 24.6 ± 2.54 μ g of succinic acid per mg of glucose.
The extracted OPGs from the wild-type (EC3937), ABF-0020261 mutant (NFB7305) and the complemented strains (NFB7326) were analyzed by mass spectrometry at low and high osmolarity (Fig. 1). In the mass spectra of extracted OPGs from the wild-type, OPGs with 3-12 residues of glucose were observed. This was clearly seen in mass spectra with the ten [M+ Na]+ molecular ions at m/z 527, 689, 851, 1013, 1175, 1337, 1499, 1661, 1823, 1985 (black arrow, Fig. 1A,B). For several molecular ions and regardless of the osmolarity, an increment of 42 Da was observed (blue arrow, Fig. 1A,B) and corresponds to the substitution of OPGs by O-acetyl residues. Only for the high osmolarity, an increment of 100 Da was observed (red arrow, Fig. 1B); this corresponds to the substitution of OPGs by O-succinyl residues. In the mass spectra of the OPG extracted from the ABF-0020261 mutant, OPGs with only O-acetyl residues were observed, regardless of the osmolarity (Fig. 1C,D), confirming the loss of succinylation of OPGs. The complemented strain displayed mass spectra similar to the wild-type strain, i.e. OPGs with O-acetyl residues at low and high osmolarity and O-succinyl residues only at high osmolarity (Fig. 1E,F).
The ABF-0020261 gene encodes a membrane protein with an acyl-transferase required for the succinylation of the OPGs and was named opgC. opgC is regulated by the variation of the osmolarity. Dependence of succinyl substitution toward osmolarity led us to study the relationship between osmolarity, succinyl residues on OPGs and expression of this gene.
We started by measuring the expression of the opgC-uidA fusion gene and the amount of O-succinyl residues on the OPGs (Fig. 2). The wild-type strain was grown from 70 to 770 mosM until mid-log phase and the opgC level expression was measured ( Fig. 2A). Until medium osmolarity, (from 70 to 570 mosM), the level of expression remained similar and low. Since no O-succinyl residue was detected on OPGs from 70 to 570 mosM, this low level of expression of opgC was insufficient for the transfer of succinyl residues on OPGs (Fig. 2B). The expression of the opgC gene increased abruptly 3-fold at high osmolarity (both at 670 and 770 mosM) ( Fig. 2A). This increased level of expression was correlated with a substitution of OPGs by O-succinyl residues, the amounts being similar at both osmolarities: 22.4 ± 2.4 and 25.10 ± 3.5 μ g of succinic acid per mg of glucose of OPGs. The absence or presence of O-succinyl residues on the OPGs was confirmed by mass spectrometry for extracted OPGs from each osmolarity (data not shown). Thus, the succinylation of the OPGs responds in a simple manner by an On/Off system, i.e. presence or absence of succinyl residues.
The EnvZ-OmpR and RcsCDB phosphorelays system control the opgC gene. In D. dadantii, perception of osmolarity is mainly sensed by the EnvZ-OmpR and RcsCDB phosphorelay systems 24 . We speculated that the O-succinylation of the OPGs is regulated through one or both of these phosphorelay systems.
To test this hypothesis, the opgC expression level was measured at low and high osmolarity in single rcsC, rcsB, rcsF, envZ, ompR mutant strains and in the double ompR rcsB mutant strain (Fig. 3). Expression of opgC in the rcsF mutant strain is similar to the expression observed for the wild-type strain for each osmolarity, and succinyl residues were measured at 23.4 ± 4 μ g of succinic acid per mg of glucose and seen in mass spectra from OPGs extracted at high osmolarity from the rcsF mutant strain (Fig. 3A,B and F). In the rcsC and rcsB single mutant strains, opgC expression is reduced two-fold regardless of the osmolarity, indicating that regulation by osmolarity still occurs. Induction of the opgC gene in both mutant strains remains too low to have on effect on the OPGs substitution since no O-succinyl residue could be observed by mass spectra or measured on OPGs extracted from these mutant strains at any osmolarity tested (Fig. 3B,D and E). In the envZ and ompR single mutant strains, the opgC expression was not induced at high osmolarity and remains similar to the level observed at low osmolarity. As expected, no succinyl residue could be observed by mass spectra nor measured from OPGs extracted from these mutant strains at any osmolarity (Fig. 3A,B,G and H). In the rcsB ompR double mutant strain, there was no opgC expression and no succinyl residue could be observed by mass spectra nor measured from OPGs extracted from these mutant strains at any osmolarity (Fig. 3A,B,G and H).
These data indicate that the EnvZ-OmpR and RcsCDB phosphorelay systems are both required in the regulation of the O-succinylation of the OPGs. The EnvZ-OmpR phosphorelay system is involved in the induction of the system. The RcsCDB phosphorelay system is most likely involved in the enhancing of opgC expression in an RcsF-independent manner. OmpR and RcsB response regulator bind on the opgC promoter. To know whether OmpR and RcsB regulation is achieved by a direct binding of both the transcriptional regulators on the opgC promoter, we performed an electrophoretic mobility shift assay (Fig. 4). After purification of the OmpR and the RcsB regulator proteins, we found that both proteins were able to bind separately on the opgC promoter as shown by the gel shift of the DNA probe at a minimal concentration of 500 ng and 300 ng for OmpR and RcsB respectively. Both proteins were able to bind together to this promoter, which significantly reduced both minimal concentrations to 100 ng and 200 ng for OmpR and RcsB, respectively. When the middle of the opgC coding sequence was used as a negative DNA control, neither OmpR nor RcsB, separately or together, bound on this DNA fragment of the opgC gene. When BSA was used as a negative protein control, we were not able to observe any gel shift with any part of the opgC gene. These results showed that OmpR and RcsB directly regulate opgC gene expression by binding on its promoter.
The opgC genes of E. coli and of D. dadantii seem to have two different origins. We were wondering why we were not able to find the opgC gene of D. dadantii (opgC DD ) by direct BLAST of the opgC gene of E. coli (opgC EC ). We started by analyzing the identities and the similarities between the amino acid sequences by using EMBOSS Needle 25 . Only 19.5% identities and 32.5% similarities between OpgC DD and OpgC EC were found. We extended our study by a phylogenetic analysis. Three phylogenic trees of OpgG, OpgH (the two proteins required for backbone synthesis) and OpgC, obtained by the maximum-parsimony method, were constructed with OpgC, identified by homology with OpgC EC or OpgC DD (Fig. 5). Three cellulose synthases from cyanobacteria were used as an outgroup for OpgG and OpgH tree. Three transmembrane acyl-transferases from cyanobacteria were used as an outgroup for OpgC. Both OpgH and OpgG phylogenetic trees displayed a similar organization, which closely follows the organization of the tree based on the 16S RNA. In contrast, the OpgC phylogenetic tree revealed two major phylogenetic groups of OpgC: group I and group II (Fig. 5). Group I is composed of E. coli K12, four zoopathogenic bacteria (Shigella flexneri, Salmonella enterica pv. Typhimurium, Klebsiella pneumoniae, Yersinia enterocolitica) and three phytopathogenic bacteria (Erwinia billingiae, Erwinia tasmaniensis, Erwinia amylovora). Group II is composed of D. dadantii, five phytopathogenic bacteria (Dickeya chrysanthemi, Dickeya zeae, Dickeya paradisiaca, Pectobacterium carotovorum, Pectobacterium atrosepticum). Seven bacteria (Pseudomonas aeruginosa, Bradyrhizobium japonicum, Rhodopseudomonas palustris, Rhodobacter sphaeroides, Brucella melitensis biovar Abortus and Sinorhizobium meliloti) didn't belong in either the group I or the group II. We analyzed the synteny of the opgC gene in the bacteria used for the phylogenetic tree. Two genetic organizations were identified (Fig. 6). In group I, the opgC gene was located downstream of the opgGH operon. Upstream of the opgGH operon, the synteny was conserved on 18 genes. Group II shared the same macrosyntheny from the opgGH operon and the upstream area with group I. The opgC gene is not closely related to the opgGH operon, as observed in group I. In group II, the areas upstream and downstream of the opgC gene displayed a conserved macrosynteny on 13 genes. The seven bacteria not included in group I or group II did not share any common genetic organization. The phylograms and the synteny showed two major clusters, strongly suggesting two different origins for the O-succinyl transferase gene among the strains.
A functional complementation occurs between opgC of E. coli and opgC of D. dadantii. As opgC EC and opgC DD seem to have two different origins, we assayed a functional complementation between opgC EC and opgC DD . We constructed various strains in E. coli and D. dadantii with plasmids containing the opgC EC gene (pNF418) or the opgC DD gene (pNFW412). All the strains were grown at low and high osmolarity. The OPGs were extracted and analyzed by mass spectrometry. In the D. dadantii strain with opgC DD , the OPGs were substituted by O-succinyl residues only at high osmolarity as seen in the wild-type strain. In the D. dadantii strain with opgC EC , the OPGs were substituted by O-succinyl residues regardless of the osmolarity, as seen in the E. coli wild-type strain. In E. coli strains, whatever the combination of the gene or the osmolarity, the OPGs were always  13,14 . Previously, we demonstrated that the OPGs concentration directly affects the activation level of the RcsCDB phosphorelay system. Furthermore, OPGs are synthesized at a high level to repress the RcsCDB phosphorelay system, which in turn avoids the repression of the pectinase genes set through the Rsm system, resulting in full virulence throughout the entire infectious process 26 . To figure out the role of the O-succinylation of the OPGs in virulence, we assayed the impact on virulence of the opgC strain. Chicory leaves were inoculated with the wild-type strain and opgG and opgC single mutant strains. Virulence severity was observed 2 days post-inoculation (Fig. 7). As expected, the wild-type showed the visible symptoms of the soft rot disease at 2 days post-inoculation. The opgG mutant strain showed no symptoms. The opgC strain showed the same phenotype as the wild-type.
This result was not surprising since we previously showed that D. dadantii synthesizes OPGs without O-succinyl residues throughout the infectious cycle 26 . On the contrary, the constant presence of O-succinyl residues on the OPGs could affect virulence. To test this hypothesis, we used the opgC DD mutant strain of D. dadantii  with the opgC EC gene, which constitutively succinylated OPGs in D. dadantii (see preceding paragraph). Bacteria were inoculated into chicory leaves and the development of the disease was observed 2 days post-inoculation (Fig. 7). The leaves showed a similar development of the soft rot disease. OPGs extracted during infection into chicory leaves from D. dadantii with the opgC EC gene synthetized a similar amount of OPGs succinated as the wild-type strain at high osmolarity (data not shown). Taken together, these results strongly suggest that only the backbone of the OPGs has a role in virulence.

Discussion
In a previous paper, our lab showed that succinylation of OPGs in D. dadantii was osmodependent 11 . Here, we characterize the opgC gene responsible for the succinylation of OPGs in D. dadantii and decipher the mechanism by which the regulation of opgC occurs.
In this study, we identified by in silico analysis the opgC gene (ABF-0020261). A direct search based on the opgC from E. coli was unsuccessful. Indeed we showed that the opgC gene from E. coli and opgC from D. dadantii seem to have two different origins. However, major structural features of the protein were conserved (Fig. 5). This idea of two different origins was strengthened by the analysis of their respective genomic organizations. D. dadantii and E. coli do not share the same organization (Fig. 6). One additional difference is that opgC of E. coli is constitutively expressed 22 . Interestingly, in D. dadanti, opgC is included in the narXL operon. This operon is known to be involved in the regulation of metabolism 27 . However, the inactivation of the narXL genes had no effect on OPGs synthesis or on its substitutions 26 . opgC gene expression is increased when the external osmolarity increases and occurrs through the EnvZ-OmpR and the RcsCDB phosphorelay systems (Figs 2, 3, 8). No expression of opgC was observed when these two phosphorelays were simultaneously inactivated, while a basal expression was observed when only one of the two phosphorelay was inactivated, regardless of the osmolarity (Fig. 3). Both regulators OmpR and RcsB bind separately the opgC regulatory DNA region (Fig. 4). However, the required amount of both proteins for binding severely decreased when added together. These two systems are necessary and sufficient for full expression and regulation of opgC. Both regulators are able to activate the expression of the opgC gene to a lower level in medium or low osmolarity although without succinylation of OPGs. Thus, succinylation of OPGs in D. dadantii appears to be an all-or-nothing phenomenon: succinylation on OPGs occurs only when OpgC reaches the adequate level (i.e. high osmolarity). Lowering this level, at least to 60% (see rcsB and rcsC mutant strains in Fig. 3), led the synthesis of OPGs completely devoid of succinyl residue (Fig. 3). Expression results show that the EnvZ-OmpR system is directly involved in the induction of the opgC gene by high osmolarity while RcsCDB is more likely involved in the enhancing of opgC gene expression. Interestingly, we showed previously that RcsCDB system activation is decreased when the amount of OPGs increased (i.e. in medium of low osmolarity) 14,15 . At high osmolarity, the concentration of the OPGs is low and in this way the RcsCDB system is activated 14 , which in turn enhances the expression level of the opgC gene (Figs 3 and 8A). The role of this feedback loop between the substitution of the OPGs and the activation of the RcsCDB system remains unclear but strengthens the relationship between them.
Osmoregulated periplasmic glucans are known to be an important virulence factor for many Gram-negative bacteria 8 . It was previously shown that OPGs are not substituted by succinyl residue during the infection cycle and this absence may be of importance for full virulence 14 . Thus, the normal virulence of strains synthesizing non succinylated OPGs was logical. The normal virulence of strains synthesizing OPGs constitutively succinylated indicates that the absence of O-succinyl substitutions on OPGs is not a prerequisite for virulence. The origin of the succinyl is probably the same in D. dadantii and in E. coli (succinyl CoA) since proteins show strong common structural features. Increased diversion of this succinyl CoA from the TCA cycle may be an additional signal to help perception of high osmolarity in D. dadantii relieving the RcsCDB and/or EnvZ-OmpR signals in an unknown mechanism. This additional signal cannot be used by E. coli even though it encounters the same kind of environmental variation.
The biological role of succinylated glucans under high osmolarity conditions remains unclear. At least, two hypothesis could be proposed. The succinylation of the OPGs gives an acidic character on OPGs synthesized. The presence of positive charge could bind free anions from the outside directly in the periplasm and avoid that too many anions arrived in the cytoplasm. In this case, the succinylated OPGs will play a role of osmoprotective molecules. We can also hypothesized that if the substrate is the succinyl CoA, used succinyl CoA could affect directly the TCA cycle by increasing or decreasing the production of ATP.
The nature of substitutions of OPGs depends on the species considered 9 . In addition, this work highlights the differences in regulation of the same substituent depending on the species. Thus, substituents of OPGs may be considered as an intrinsic property playing a biological role, eventually defined by the lifestyle of bacteria. In Brucella abortus 44 , the succinylated OPGs are required under hypo-osmotic condition but not during the virulence. However, no role had been assigned to any substituent in other bacterial species.

Methods
Bacterial strains, media and growth conditions. Bacterial strains are described in Table 1. Bacteria were grown at 30 °C for D. dadantii or at 37 °C for E. coli in lysogeny broth (LB) 28 , or in minimal medium LOS (4 g of casein hydrolysate, 0.5 mg of FeSO 4 , 18 mg of MgCl 2 , 200 mg of (NH 4 ) 2 SO 4 and 175 mg of K 2 HPO 4 per liter, pH 7.2) supplemented with a carbon source at a concentration of 2 g.l −1 29 . Solid media were obtained by adding agar at 15 g.l −1 .
Osmolality of growth media (mosM) was measured with vapor pressure osmometer (model 3320, Advanced Instruments, Inc., MA, USA). Construction of the opgC, rcsF, envZ and ompR mutations. In order to inactivate opgC, the opgC gene was amplified by PCR (opgCml-F: GCAGTCTGGGTGGTGCAAG and opgCml-R: ATGTCAGATAAAACAGTGGGTG) and cloned into the pCR2.1 plasmid (topo TA cloning kit, Life Technologies). A chloramphenicol DNA cassette, obtained by digestion of pNFCml 15 by EcoRV was introduced into the resulting plasmid, digested by HincII (pNFW414).
In order to inactivate rcsF, a DNA fragment containing this gene was generated by P C R ( rc s F S m a -F : A AC C C G G G AT C C T G C T G G T G G T T C T G G T T TAT C a n d rc s F X b a -R : AATCTAGACTGCGCCAGCACCTGGCTGATC) and cloned into the pCR2.1 plasmid (topo TA cloning kit, Life Technologies). A chloramphenicol DNA cassette, obtained by digestion of pNFCml 15 by EcoRV was introduced into the resulting plasmid, digested by AgeI and blunt-ended by the Klenow fragment of the DNA polymerase (Biolabs) (pNFW184).
In order to inactivate ompR, a DNA fragment containing this gene was amplified by PCR (ompRKpn-F: AGGTACCGCAAACGCAGCAGGCGCTAC and ompRKpn-R: AGGTACCCGCGTTGAGGTCGGCCACTT) and cloned into the pCR2.1 plasmid (topo TA cloning kit, Life Technologies). A gentamycin DNA cassette, generated by PCR and digested by HpaI 14 , was introduced into the resulting plasmid, digested by AfeI (pNFW457). In order to inactivate envZ, a DNA fragment containing this gene was amplified by PCR (envZ-F: GCGAATCCTTCCATCTGATC and envZ-R: GCTGCTGGAAAACGTGACC) and cloned into the pJET1.2 (cloneJET PCR cloning kit, Life Technologies) A gentamycin DNA cassette 14 , was introduced into the resulting plasmid, digested by SfoI (pNFW515). After electroporation of the four resulting plasmids, the mutations were integrated into the D. dadantii chromosome by marker exchange recombination in the presence of the appropriate antibiotic after successive cultures in low phosphate medium 30 . Cloning of the opgC from D. dadantii and E. coli. To clone the opgC gene from D. dadantii, the opgC DNA fragment was amplified by PCR (opgC-F: GCTCTAGAGCACCGCCACGCCGATGTC and opgC-R: GGGGTACCGTTCAGACGATGCATCGATG), digested by XbaI and KpnI and introduced into pUC18Not, digested by the same enzymes (pNFW413).
To obtain the pUTmini-Tn5-Sp 31 containing opgC DD , pNFW413 was digested by NotI and the released opgC DD DNA fragment was cloned into the pUTmini-Tn5-Sp plasmid digested by the same enzyme. This plasmid was introduced into D. dadantii by conjugation.
To obtain the pUTmini-Tn5-Sp containing opgC EC , pNF418 22 was digested by PstI and inserted into pUC18 Not digested by the same enzyme. This last plasmid was digested by NotI and the released opgC EC DNA fragment was cloned into the pUTmini-Tn5-Sp plasmid digested by the same enzyme. This plasmid was introduced into D. dadantii by conjugation.
Construction of the opgC::uidA gene fusion. To construct the opgC::uidA gene fusion, the coding DNA sequence of the uidA gene was amplified by PCR (GusSph-F: CACAGCATGCATTCCGGACCAGTATTATTATC and GusHind-R: CACAAAGCTTATCCGCTCACAATTCCAC) from pCRS548 32 , digested by SphI and HindIII and cloned into the pUC18not plasmid digested by the same enzymes. The opgC regulatory DNA region was amplified by PCR (promXba-F: CACATCTAGACGATGGTTATCTGCTCAAGG and promSph: CACAGCATGCTCATCCAAGCAGAAACCGGAG), digested by XbaI and SphI and cloned into the pUC18 Not-uidA plasmid digested by the same enzymes.
To obtain the pUTmini-Tn5-Sp containing opgC::uidA, this last plasmid was digested by NotI and the released opgC::uidA DNA fragment was cloned into the pUTmini-Tn5-Sp plasmid digested by the same enzyme. This plasmid was introduced into D. dadantii by conjugation. To construct the recombinant His-tagged OmpR protein, the ompR DNA fragment was amplified by PCR (ompR-F: CACCATGCAAGAGAATTATAAAATTCTGG and ompR-R: CCATCGAATCATGCTTTACTGCCG) and cloned into the His-tag expression vector pET100/D-Topo (Invitrogen, Life Technologies).

Cloning, expression and purification of His-tagged
Both recombinant proteins were expressed in E. coli BL21(DE3) and purified as described previously 14 . Briefly, after growing the cell was lysed by sonication and the crude extract was transferred to a Ni-NTa column (Macherey Nagel) to purify the protein.
Transduction, conjugation and transformation. Transformation of E. coli cells was carried out by the rubidium chloride technique 29 . Construction of strains was performed by transferring genes from one strain of D. dadantii to another by generalized transduction with phage Φ EC2 as described previously 33 . Plasmids were introduced in D. dadantii by conjugation or electroporation.
Transposon mutagenesis. To allow integration of a single ectopic copy of the mini-Tn5-Sp-opgC::uidA gene fusion, mini-Tn5-Sp-opgC DD and mini-Tn5-Sp-opgC EC , transposon mutagenesis was performed as described previously 15 . Briefly, after conjugation between an E. coli strain harboring the pUTmini-Tn5-Sp plasmid carrying the appropriate construction and a D. dadantii strain, Sp r mutants were selected on M63 plates containing sucrose as a unique carbon source and spectinomycin.
Determination of enzyme activities. β -glucuronidase assays were performed on crude extracts obtained from bacteria disrupted by sonication 2 × 20 s (Sonifier cell disruptor B-30, Branson, 70% duty cycle, 7 microtip limit, Hold time, continuous, appropriate probe) after growth in vitro (LB medium) or in planta and extracted from chicory leaves as described elsewhere 14 . β -glucuronidase activity was determined by spectrometric monitoring of the hydrolysis of PNPU (4-nitrophenyl-β -D-glucuronide) at 405 nm.
The protein concentration was determined by the Bradford assay with bovine serum albumin as a standard 34 .
Pathogenicity test. Chicory leaves were inoculated as previously described 13 . Briefly, bacteria from an overnight culture in LB medium were recovered by centrifugation and diluted in physiological water. After wounding, plants were inoculated with 10 7 bacteria and incubated in a dew chamber at 28 °C for 48 h.

Extraction of OPGs.
Bacteria were grown overnight in LOS with the indicated NaCl concentration. Bacteria were collected by centrifugation at 4 °C for 15 min at 8,000 g. Cell pellets were resuspended in distilled water and lysed with 5% trichloroacetic acid (TCA). After centrifugation at 4 °C for 15 min at 8,000 g, the supernatant was neutralized with ammonium hydroxide 10% and concentrated by rotary evaporation. The resulting material (2 ml) was then fractionated by gel filtration on a Bio-Gel P-4 column (lengh: 55 cm, diameter; 1.6 cm, Bio-Rad) equilibrated with 0.5% acetic acid. The column was eluted in the same buffer at a flow rate of 15 ml h −1 , and fractions of 1.5 ml were collected. Presence of oligosaccharides in each fraction was determined by the colorimetrically anthrone procedure. Fractions containing OPGs were pooled and total content was determined by the same procedure 35 . For mass spectrometry analysis and succinate quantification, OPGs were subsequently desalted in water on a Bio-Gel P-2 column (lengh: 90 cm, diameters; 1.6 cm, Bio-Rad,) and fractions containing OPGs were lyophilized.

Determination of succinate content of OPGs.
After purification of the OPGs as described above, one milligram of OPGs was dissolved in 0.2 ml of 0.5 M NaOH and incubated at 100 °C for 30 min to liberate the succinyl from OPGs. Glucosidic backbones were removed by absorption on 50 mg of charcoal suspended in 0.3 ml of water, and the charcoal was then washed three times with 0.5 ml of water. The four supernatants were pooled (2 ml) and neutralized with Dowex AG 50W-X8 (Bio-Rad) on H+ form. Succinic acid content was determined with a succinic acid kit (R-Biopharm) following the manufacturer's instruction. The limit of detection is of 2.1 mg/mg of OPGs.
Mass spectrometry. All mass spectra were acquired on a Voyager Elite (DE-STR) reflectron time-of-flight (TOF) mass spectrometer (Perseptive Biosystems, Framingham, MA), equipped with a pulsed nitrogen laser (337 nm) and a gridless delayed extraction ion source. Oligosaccharide samples were analyzed in delayed extraction mode using an accelerating voltage of 20 kV, a pulse delay time of 200 ns and a grid voltage of 66%. Detector bias gating was used to reduce the ion current for masses below 500 Da. Between 100 and 200 scans were averaged for each mass spectrum. Oligosaccharide alditols were co-crystallized with 2,5-dihydroxybenzoic acid (DHB) as matrix [10 mg ml −1 of DHB in methanol/water (50/50) containing 0.1% Trifluoro Acetic acid (TFA)]. For all measurements, the "dried droplet" preparation technique was used. Typically, 1 μ l of the matrix was mixed on-target with 1 μ l of water-dissolved oligosaccharides and allowed to dry under an air stream. They were analyzed in positive ion mode.
Gel shift assays. DNA fragments of about 600bp containing either the opgC regulatory region or middle of the opgC gene coding region (the control fragment) were amplified by PCR (same primer pair used for opg-C::uidA gene fusion (see before) or opgCmid-F:AACTGACCAGCATGGGATTC) and opgCmid-R:GGGGTTG-CGTTGTAGCAGG respectively). PCR products were purified using the NucleoSpin Gel and PCR Clean-up kit (Macherey-Nagel Phylogeny tree. The protein sequences of the 21 strains were aligned with CLUSTAL OMEGA 37 with default parameters. The trees were generated using the protein maximum likelihood (Proml) method in the Phylip package (version 3.69). The node reliability of the phylogenetic trees was evaluated with 100 bootstrap replicates (seqboot, consense and retree methods of the Phylip package). The protein sequence of the strain Gloeobacter violaceus was used as the outgroup. Phylogenetic trees were displayed with Dendroscope (version 3.2.10).
Statistical Analysis. For statistical analyses, Graph-prism6 software was used. Data were analyzed by paired t-test; a value of p < 0.05 was considered significant.