The plant defense signal galactinol is specifically used as a nutrient by the bacterial pathogen Agrobacterium fabrum

The bacterial plant pathogen Agrobacterium fabrum uses periplasmic-binding proteins (PBPs) along with ABC transporters to import a wide variety of plant molecules as nutrients. Nonetheless, how A. fabrum acquires plant metabolites is incompletely understood. Using genetic approaches and affinity measurements, we identified here the PBP MelB and its transporter as being responsible for the uptake of the raffinose family of oligosaccharides (RFO), which are the most widespread d-galactose–containing oligosaccharides in higher plants. We also found that the RFO precursor galactinol, recently described as a plant defense molecule, is imported into Agrobacterium via MelB with nanomolar range affinity. Structural analyses and binding mode comparisons of the X-ray structures of MelB in complex with raffinose, stachyose, galactinol, galactose, and melibiose (a raffinose degradation product) revealed how MelB recognizes the nonreducing end galactose common to all these ligands and that MelB has a strong preference for a two-unit sugar ligand. Of note, MelB conferred a competitive advantage to A. fabrum in colonizing the rhizosphere of tomato plants. Our integrative work highlights the structural and functional characteristics of melibiose and galactinol assimilation by A. fabrum, leading to a competitive advantage for these bacteria in the rhizosphere. We propose that the PBP MelB, which is highly conserved among both symbionts and pathogens from Rhizobiace family, is a major trait in these bacteria required for early steps of plant colonization.

The plant-rhizospheric microbial population interaction is dynamic and largely influenced by root exudates, with either beneficial or harmful consequences for plant growth develop-ment and health (1). The germinating seeds in contact with the surrounding soil and microorganisms have strong influences on the rhizosphere composition and favor fast-growing microorganisms able to exploit carbon released, to resist to antimicrobial compounds, and to outcompete other surrounding bacteria (2)(3)(4). Raffinose and stachyose from the raffinose family of oligosaccharides (RFO) 5 accumulate in plant seeds as energystorage metabolites, and are released during plant germination (2,5). The precursor of RFO synthesis, namely galactinol, a D-galactose bound to an inositol, is produced by the plant enzyme galactinol synthase (GolS) (5, 6) ( Fig. 1). Galactinol plays an important role in plant health, being involved in plant resistance against abiotic (drought and temperature) (7,8) and biotic stresses (9 -11). Indeed, this molecule, which accumulates in plants in response to bacterial inoculation, is involved in the induced systemic resistance to phytopathogens (9). Raffinose and stachyose are synthesized from sucrose by the subsequent addition of activated galactose moieties donated by galactinol using plant raffinose and stachyose synthases, respectively (Fig. 1). Therefore, RFOs are ␣-(1,6)-galactosyl extensions of sucrose.
Periplasmic-binding proteins (PBPs) associated with their ATP-binding cassette (ABC) transporter are essential for transport (12). A PBP-mediated transport system is responsible for RFO uptake from seed exudates into bacterial cells as previously shown in Ensifer meliloti 1021 (13,14). RFOs, which are degraded by ␣-galactosidases in this latter strain, are used as nutrients, and their assimilation may be involved in bacterial survival in plant rhizosphere (13,14). In more detail, raffinose and stachyose can be degraded into melibiose and fructose, and raffinose and galactose, respectively (15,16). Bacterial assimilation of RFOs and melibiose was associated with trophic advantage in plant-bacteria interaction (14,17), whereas nothing was known for galactinol. Agrobacteria are telluric and rhizosphere bacteria, commonly isolated from roots of numerous plants as commensal bacteria. They can also be pathogenic with the presence of the tumor-inducing plasmid (18,19). They are then able to create their own ecological niche after plant cell transformation that leads to tumor formation in a wide range of plants (18). An in silico analysis of ␣-galactosidases distribution in bacteria indicated that Agrobacterium fabrum C58 strain contains an operon putatively involved in RFO transport and degradation (20 -22). This operon that we named mel is similar to the agp operon of E. meliloti (13,14), and encodes the PBP MelB (Atu4661) which shares 73% sequence identity with the PBP AgpA, its associated ABC transporter (Atu4662-Atu4664), and two ␣-galactosidases (Atu4660 and Atu4665). All these latter proteins display between 74 and 86% sequence identity with their E. meliloti corresponding homologues (Fig. 2a).
We hypothesized that this mel operon was responsible for the transport and assimilation of ␣-galactosides in A. fabrum. Here, we focused on its transport function, and investigated the genetic and molecular role of the PBP MelB through an integrative approach using a defective mutant in cellulo and in planta, crystallography, and affinity measurements. We showed that MelB was the PBP responsible for RFO, melibiose, and galactinol import into A. fabrum C58, displaying the highest affinity for galactinol in nanomolar range and preferring to bind a 2-unit ligand. We structurally characterized the binding mode of MelB for its different ligands. Overall, our work highlights how the capacity of agrobacteria to assimilate plant ␣-galactosides confers on them an advantage in colonizing efficiently the plant tomato rhizosphere, explaining why the PBP MelB is highly conserved among symbionts and pathogen rhizobiales.

The PBP MelB is responsible for galactinol, melibiose, and RFOs (raffinose and stachyose) uptake
The growth profiles of A. fabrum C58 wildtype (WT) and C58⌬melB-defective mutant for MelB were compared in rich (YPG) and minimal medium containing RFOs, their derivatives or succinate (as control) as the sole source of carbon. The C58⌬melB mutant has the same growth rate as the WT strain in minimum medium with succinate and in rich medium. However, the C58⌬melB mutant did not grow on galactinol, melibiose, raffinose, and stachyose, in contrast to WT (Fig. 2b). Therefore, MelB associated to its ABC transporter is the transport system responsible for the uptake and is necessary for the assimilation of these four molecules in pure culture.

MelB exhibits a high affinity for galactinol
Binding of galactinol, melibiose, raffinose, and stachyose to the purified recombinant mature protein MelB was explored using tryptophan fluorescence spectroscopy (MelB possesses 16 tryptophans) and isothermal titration microcalorimetry (ITC). Intrinsic protein fluorescence titration experiments yielded apparent dissociation constant K D values of 10 Ϯ 1 nM and 72 Ϯ 4 nM with galactinol and melibiose, respectively, showing that MelB is very efficient for galactinol binding. Reactions for synthesis (a) are represented with black arrows whereas those for degradation (b) are with gray arrows. The first step of RFOs biosynthesis starts with the formation of galactinol from UDP-galactose and inositol catalyzed by the plant enzyme galactinol synthase. Raffinose, stachyose, and verbascose are synthesized from sucrose by the subsequent addition of activated galactose moieties donated by galactinol using plant raffinose and stachyose synthases and galactan:galactan galactosyltransferase (GGT), respectively. Cleavage, shown by scissors, of raffinose by an ␣-galactosidase leads to the formation of either melibiose and fructose or sucrose and galactose. For RFOs of higher degree of polymerization (DP), an ␣-galactosidase activity results in RFOs of lower DP and galactose.

Galactinol and melibiose bound to MelB
Reducing the ligand size to a monosaccharide (galactose) or increasing it resulted in a substantial affinity reduction compared with galactinol: K D values of 35-fold higher for the raffinose, and over 1000-fold higher for both galactose and stachyose, respectively (Table 1 and Fig. S1). The K D values were slightly higher using ITC than those determined by autofluorescence, but this increased K D was consistent (Table 1 and Fig. S1). Because MelB was not stable at high concentration during the time course of ITC experiment, we were not able to measure an interpretable signal for the stachyose binding. The ITC data confirmed the 1:1 binding stoichiometry for all ligands and revealed a high enthalpy of binding for galactinol and melibiose meaning that both ligands use the same binding mechanism mainly involving polar interactions. In contrast, the binding mode of raffinose was characterized by an unfavorable enthalpy contribution and a strong entropy term indicating that hydrophobic interactions may play a predominant role and/or a displacement of water molecules occurs upon ligand binding. The galactose interaction adopts an intermediate behavior with a high entropy term, accompanied by a weakly favorable enthalpy of binding, suggesting that polar bonds are less important for galactose alone compared with a 2-unit ligand, resulting in a lower enthalpy, thus a weaker affinity. MelB is specific for ␣-(1,6)-galactosides (RFOs). Indeed, no interaction could be measured with glucose, sucrose, cellobiose, lactose, and ␣-(1,6)-glucosides.
Thermal denaturation experiments revealed a contribution of more than 3°C for two ligands for protein stability (Fig.  S2). Indeed, adding galactinol or melibiose led to a melting temperature (T m ) of over 46°C compared with the 43°C for the unliganded protein in agreement with the measured K D values. Galactose or stachyose binding did not stabilize MelB whereas the raffinose binding produced a slight effect with a T m of 45°C.

MelB is a PBP from cluster C
The mature MelB expression plasmid was a synthetic gene lacking the first 18 signal sequence residues that serve for localization to bacterial periplasm. The numbering used for the description of residues corresponds to the mature protein of 677 amino acids. Because MelB is the biggest PBP so far and shares low sequence identity (around 20%) compared with PBPs with known three-dimensional structures, we first solved the structure of seleniated MelB in complex with raffinose at 2 Å resolution by single wavelength anomalous dispersion method. The asymmetric unit is composed of four similar MelB-raffinose complexes (average root mean square deviation (RMSD) value of 0.4 Å). By the molecular replacement method, we then solved the structure of MelB in complex with galactinol, melibiose, galactose, and stachyose at 2.2, 1.8, 2.5, and 2.1 Å resolution, respectively ( Table 2). The asymmetric unit of the galactinol and melibiose complexes also contains four very similar molecules with RMSD values between monomers ranging from 0.2 to 0.4 Å whereas that of the galactose and stachyose complexes possesses two identical molecules. Moreover, the five ligand-bound structures are very similar with an average RMSD value of 0.4 Å. They all adopt a closed conformation. MelB fold is monomeric, composed of two lobes, each formed by a central ␤-sheet flanked by ␣-helices (Fig. 3a). The biggest lobe (lobe 1) consists of residues 8 -354 and 619 -678 and the smallest (lobe 2) comprises the residues 364 -610. Two short segments (Fig. 3a) define the hinge region connecting the two lobes. MelB possesses a typical fold of cluster C within the PBP structural classification (12) as SSM-EBI (http://www.ebi. ac.uk/msd-srv/ssm) 6 reports: RMSD between MelB and similar PBP structures binding oligopeptide are over 2.6 Å for 450 C␣ Figure 2. The mel operon structure and MelB involvement in galactinol, melibiose, and RFO consumption. a, atu4660-atu4665 genes belong to the same transcription unit (operon prediction by Westover et al. (22) that we called the mel operon). The mel promoter P mel indicates the gene transcription direction. Both atu4660 and atu4665 genes are annotated as two ␣-galactosidases, atu4661 as the PBP MelB and atu4662-atu4664 genes as the associated ABC transporter. Comparison is between the mel operon genes and their homologues in E. meliloti 1021 (13,14). the percentages of sequence identity between each homologous protein are indicated; for example, the PBPs AgpA and MelB share 73% sequence identity. b, 2.5 days growth (A at 600 nm) of A. fabrum C58 WT strain (in white) and the C58⌬melB mutant (in black) in AT minimal medium supplemented with different carbon sources. Standard deviations were calculated from five technical and two biological replicates. Asterisks indicated significant differences (Mann-Whitney p value ϭ 0.05).
atoms. Nonetheless, a detailed structural comparison is irrelevant because MelB presents a distinct ligand-binding site.

Ligand-binding site of MelB
All ligands are bound between the two closed lobes of MelB. The size and the volume of the ligand-binding site constrain the conformation of bound RFOs. Indeed, raffinose and stachyose bind in a very compact form (Fig. 3b), and the addition of a galactosyl moiety at the nonreducing end of stachyose corresponding to verbascose will abolish its binding, because the pocket is not large enough to accommodate a pentasaccharide. All ligands are well defined in their electron density maps except the fructose moiety of the stachyose likely responsible for its low affinity (Fig. 3, c-h). They share a buried nonreducing end galactosyl unit wedged between two aromatic residues (Trp 317 and Trp 639 ) which superimposes very well with the bound galactose alone (Fig. 3, c-h). These galactoses at position 1 make 10 similar protein contacts involving the main chain amino group of Gly 111 and the side chains of Arg 320 , Asn 333 , Glu 335 , and Glu 641 from lobe 1 and both side chains of Tyr 487 and Arg 533 from lobe 2. The O6 atom of this pyranose interacts with a conserved water molecule observed in each complexed structure (except in molecule B of MelB-galactose complex), which in turn makes hydrogen bonds with the amino group of Gly 112 and the side chain of Asp 114 . In both structures of MelB in complex with galactose and stachyose, the galactose at position 1 does not interact with Asn 333 but because of two hydrogen bonds with the side chain of Glu 335 , it conserves 10 interactions with MelB. Modeling a glucosyl unit at the nonreducing end (position 1) creates steric hindrance between the equatorial C4-OH and the Trp 639 indol explaining the specificity of MelB for a galactosyl unit at position 1 as shown by the affinity measurement.
In contrast to position 1, at position 2, the glucosyl moieties of melibiose and raffinose, the inositol moiety of galactinol and the galactose moiety of stachyose do not superimpose and can shift up to 3.5 Å to allow, for example, the fructose or the sucrose accommodation of raffinose and stachyose in the ligand-binding site, respectively (Fig. 3, b and c). A conformational change less than 1 Å for Trp 110 and Tyr 487 side chains is observed to accommodate a glucosyl unit at position 2 ( Fig. 3c). More arrangements from amino acids of lobe 2 can occur to accommodate the glucosyl units at positions 3 and 4 ( Fig. 3c). For example, Trp 557 is pushed away by more than 1 Å compared with the other liganded structures to find a room for the fructose at position 4. All units at position 2 have in common the stacking onto the aromatic indol of Trp 110 , with the optimum one for the inositol because of the shorter link (one carbon shorter) between the two subunits of galactinol compared with melibiose and RFOs. The melibiose's glucose and the inositol  Table 2 Crystallographic data and refinement parameters Values for the highest resolution shell are in parentheses.

Galactinol and melibiose bound to MelB
share only one protein contact: an oxygen interacts with the NH of the Trp 639 side chain (Fig. 3, e and f). The inositol makes three additional hydrogen bonds with Asn 484 , Tyr 487 , and Ser 515 from lobe 2 whereas the rest of the oxygen of the melibiose's glucose interacts with the protein side chains via water molecules only. In contrast, the glucose moiety of raffinose and the galactose moiety of stachyose at position 2 make an additional H-bond compared with inositol involving the side chains of Asp 519 (Fig.  3, g and h). The fructose at position 3 in raffinose interacts with the main chain of Arg 533 from lobe 2 and the NH of the Trp 639 indol from lobe 1 (Fig. 3g). In contrast, Arg 532 side chain makes two H-bonds with the glucose moiety at position 3 in stachyose whereas only one is present between the fructose moiety at position 4 and the Trp 639 side chain (Fig. 3h).

MelB is highly conserved among rhizobiales
Searching for MelB conservation in the bacterial kingdom (protein database at NCBI), and subsequent phylogenetic anal-ysis revealed Ͼ310 PBPs above 65% sequence identity (Fig. 4). Galactinol-, melibiose-, and raffinose-binding signatures share 10 amino acids Trp 110 -Gly 111 -Trp 317 -Arg 320 -Asn 333 -Glu 335 -Tyr 487 -Arg 533 -Trp 639 -Glu 641 . Stachyose binding shares only nine of these latter because Asn 333 does not interact with stachyose. Two additional residues (Asn 484 and Ser 515 ) are involved in galactinol, raffinose, and stachyose binding compared with melibiose binding. One additional residue Asp 519 belongs to the raffinose and stachyose signatures. A last additional residue Arg 532 defines the stachyose-binding signature composed of 13 residues in total. Members of the MelB subgroup (67 PBPs) display more than 90% identity sequence with the conserved binding signature. They all belong to Rhizobium and Agrobacterium genera. Outside the MelB cluster, the signature slightly degenerates for galactinol and raffinose binding. Nonetheless, modeling indicates that their binding would not be affected. Remark-

Galactinol and melibiose bound to MelB
ably, all these PBPs belong to soil-and plant-interacting genera.

Galactinol and melibiose are inducers of mel operon genes
We constructed the C58 pOT1e-P mel reporter fusion strain to study in cellulo gene expression of mel operon in the presence of commercial compounds. Compared with succinate, slight but significant inductions were observed with raffinose, galactose, and stachyose (2-, 1.6-, and 1.25-fold change values, respectively) (Fig. 5a). In contrast, galactinol and melibiose are efficient inducers with 4.8-and 6.5-fold change values, respectively.

mel operon genes are expressed in early plant colonization
At two early stages of plant colonization, in planta expression of mel operon has been studied in the WT strain harboring the pOT1eM-P mel plasmid reporter fusion (m-cherry constitutive expression and egfp inducible expression). 48 h after seed imbibition and inoculation, most bacteria in contact with the radicle cells expressed mel operon, as shown by the yellow cells in Fig. 5, b and c. 14 days after seed imbibition and inoculation, among bacterial cells present on plant roots (red and yellow cells), some of them were still expressing RFO uptake and degradation genes (yellow cells in Fig. 5d). Thus, mel operon was more expressed at the beginning of plant colonization.

The PBP MelB confers a competitive advantage in colonizing tomato rhizosphere
The colonization of plant rhizosphere by A. fabrum C58 WT and C58⌬melB mutant was evaluated at 2 days post inoculation (dpi) and 14 dpi (Fig. 6). When tomato seeds were inoculated with each strain individually, the bacterial colonization level did not significantly differ at 2 dpi (Mann-Whitney p value of 0.09) whereas at 14 dpi, this was slightly higher for the WT strain (Mann-Whitney p value Ͻ2.2e-16e) (Fig. 6a). When A. fabrum C58 WT and C58⌬melB mutant were co-inoculated, a slight and a drastic reduced fitness was observed for the C58⌬melB mutant at 2 and 14 dpi, respectively (Fig. 6b), revealing a selective advantage conferred by galactinol/melibiose/RFO exploitation under a competitive challenge.

Discussion
This work reveals the molecular and ecological roles, and structural basis of the PBP MelB, encoded by the linear chromosome of A. fabrum C58.
At the beginning of the study, we made the straightforward assumption that MelB was behaving like its homologous PBP AgpA from E. meliloti described as an ␣-galactoside transporter (13). Our gene expression analyses and growth assays experiments showed that similarly to what has been reported for expression of agp operon genes in E. meliloti, ␣-galactosides

Galactinol and melibiose bound to MelB
induce expression of mel operon genes in A. fabrum, and are used as carbon sources after being imported by MelB. Indeed, in contrast to the WT strain, a MelB defective mutant is unable to grow on ␣-galactosides. Moreover, using two different biophysical methods, we demonstrated that MelB can bind melibiose, raffinose, and stachyose. Because of the small volume cavity of the ligand binding, stachyose displays a weak affinity for MelB (micromolar range) compared with melibiose and raffinose (nanomolar range). The accommodation of stachyose requires drastic conformational constraints on the ligand when bound to the protein. Indeed, only small local rearrangements of few protein side chains (Trp 110 and Trp 557 ) forming the binding site can occur. With an affinity in the micromolar range, MelB also binds galactose, which is the sugar common to all ␣-galactosides present at the nonreducing end. Nonetheless, the low affinity of MelB for galactose prevents galactose from competing with melibiose and raffinose. From our results, MelB can appear as an alternative galactose transporter suspected by Kemner et al. (23), which allowed a chvE-gguABC defective mutant to grow on galactose (23,24). Conversely, the presence of the ChvE-GguABC sugar transporter can explain why melB defective mutant was still able to grow on galactose.
An unexpected outcome of our study was that galactinol, which is the precursor of ␣-galactosides production in plants, was uptaken into agrobacteria via the MelB-mediated transport system. Moreover, MelB displays a preference for galactinol with high affinity (nanomolar range) indicating that this molecule must be efficiently imported into A. fabrum, in line with the gene expression results. Remarkably, MelB recognizes similarly the nonreducing end galactose common to all tested ␣-galactosides and galactinol. Overall, using genetic, structural, and affinity data, this work demonstrates that the MelB-mediated transport system contributes to the import of ␣-galactosides with a strong preference for a 2-unit ligand (melibiose) and mainly contributes to that of galactinol. To our knowledge, this is the first description of a bacterial PBP allowing galactinol import.
The imported sugars are used to sustain bacterial growth. Galactinol-and ␣-galactoside-rich environments would facilitate the settlement of bacteria capable to assimilate these plant compounds efficiently. From in planta competition assays on tomato between the A. fabrum WT and melB defective mutant, we showed that MelB confers a marked selective advantage in colonizing tomato rhizosphere, and since early time. This observation correlates with the presence of melibiose and raffinose in the plant rhizosphere, as they are highly abundant in the seeds (25) and known to be released during seed germination (14). Hence, the competitive advantage of the WT in the tomato rhizosphere could be because of a trophic advantage of the strain during seed germination. A germinating seed can indeed be considered as a new environment to be colonized. The community composition of the mature plant is influenced by historical contingency (timing and order of arrival) of the seed community members and their ability to efficiently settle in that environment (26), as the first establishing species are known to affect the ability of potential immigrants to establish. This is called the priority effect (27). Thus, the ability of bacteria to compete and settle in germinating seed environment by growing on galactinol or ␣-galactosides released at this time could have long effect on their ability to persist and colonize plant rhizosphere. This is consistent with our findings that at 14 dpi, the competitive advantage of the WT strain is even higher than the one measured at 2 dpi.
Besides the trophic advantage linked to the mel operon, this operon could be associated to another aspect of bacterial plant interactions, linked to plant defense signaling and protection against pathogens. Indeed galactinol is a plant compound involved in plant defense (9 -11). For example, Pseudomonas chlororaphis O6-mediated induced systemic resistance was shown associated with an elevation of galactinol content within plants, which conferred disease resistance against pathogen attack (9). The disease resistance was associated with induction of the expression of a set of pathogen-responsive genes (9,10). Moreover, in Arabidopsis, deletion of enzymes that decrease the galactinol and/or raffinose content has been shown to increase plant resistance against the phytopathogenic nema-

Galactinol and melibiose bound to MelB
tode Heterodera schachtii (11). Similarly, Agrobacterium melmediated activity could modify the level of plant galactinol and/or raffinose, which could either drive bacterial recognition by the plant or reduce plant defense signaling through the fall of galactinol content. Agrobacterium is known to be able to bypass and overcome plant defenses (28). It would thus be of interest to study the involvement of the mel operon in that situation.
In this study, we defined the galactinol-/melibiose-/raffinose-binding signature and found out that this is strictly conserved in MelB homologues in Agrobacterium and Rhizobium, which share a high sequence identity over 90% with MelB. Moreover, phylogenetic and structural data showed that this wide occurrence is extended among Rhizobiaceae, all plantinteracting genera (Mesorhizobium, Allorhizobium, Ensifer, Martelella, Pleomorphomonas, Kaistia, and Devosia). Therefore, whatever advantage it gives, it is tempting to speculate that galactinol/melibiose (and to a lesser extent raffinose) may be associated with a selective pressure toward the acquisition of binding, transport, and degradation functions in microorganisms, making the PBP MelB a major trait in the first step of tomato colonization and likely of other plant species.

Bacterial culture conditions
Bacteria and plasmids used in this study are shown in Table  S1. Escherichia coli strains were grown at 37°C in LB medium supplemented when it was necessary with appropriate antibiotics (tetracycline 10 g/ml, gentamicin 15 g/ml, ampicillin 100 g/ml). A. fabrum C58 strain and its derivatives were cultivated at 28°C in YPG (yeast extract, 5 g per liter, peptone, 5 g per liter, glucose, 10 g per liter, and pH adjusted to 7.2) rich medium supplemented when required with neomycin (25 g/ml), kanamycin (25 g/ml), and/or gentamycin (20 g/ml). In growth assays, AT minimal medium supplemented with 10 mM ammonium sulfate and 10 mM carbon sources was used. 200 l were inoculated in Bioscreen honeycomb 100-well sterile plates and incubated in a Bioscreen C Reader (Labsystems, Helsinki, Finland) at 28°C during 5 days. Cell growth was measured every 20 min. Analyses were performed in five technical replicates and in three biological replicates.

Construction of melB defective mutant in A. fabrum C58 and transcriptional fusion
The A. fabrum C58⌬melB defective mutant was constructed as described previously (29) without marker exchange. Briefly, the recombinant region containing the upstream and downstream region flanking the melB gene (amplified by PCR using primers listed in Table S2) was inserted into pJQ200sk vector (30) leading to a nonpolar mutant. The resulting plasmid was introduced into A. fabrum C58 by electroporation. Bacteria were spread on YPG medium plates containing gentamicin (20 g/ml) for the first selection and gentamicin-resistant colonies were spread on YPG plates supplemented with 5% sucrose for the second selection. The deletion of melB was verified by sequencing (GenoScreen, Lille, France).
The C58 pOT1e-P mel transcriptional reporter fusion strain was obtained as follows: the promoter region of atu4660-atu4665 genes named P mel was PCR amplified (using primer listed in Table S2) and the PCR fragment obtained was ligated into ClaI-SalI digested vector pOT1e. pOT1eM-P mel plasmid was obtained by cloning P mel into SpeI digested pOT1eM vector as described previously (31). Transcriptional reporter constructions were introduced into A. fabrum by electroporation.

Cloning, expression, and purification of mature MelB
The mature MelB expression plasmid was chemically synthesized using codon optimization for the expression in E. coli and inserted into pET-9a plasmid using NdeI and BamHI restriction enzyme (GenScript, Piscataway, NJ). E. coli BL21 competent cells transformed with pET9a-MelB were grown in LB media at 37°C until A 600 of 0.6. 0.5 mM of isopropyl ␤-D-thiogalactopyranoside (IPTG) was added to the culture for overnight expression at 20°C. The cells were pelleted by centrifugation at 4000 ϫ g for 15 min at 4°C, resuspended in 50 mM Tris-HCl, pH 8, 300 mM NaCl, and 20 mM imidazole, and disrupted by sonication. After centrifugation at 25,000 ϫ g for 30 min, the filtered supernatant was injected on a nickel affinity column (HiTrap 5 ml, GE Healthcare). After a washing step of 6% 50 mM Tris-HCl, pH 8, 300 mM NaCl, and 300 mM imidazole (Buffer B), the protein was eluted with Buffer B and injected on a gel filtration Superdex 200 26/60 (GE Healthcare) using 50 mM Tris-HCl, pH 8, and 150 mM NaCl. The protein fractions were pooled, concentrated at 10.7 mg/ml, and stored at Ϫ80°C.

Expression and purification of mature seleniated MelB
The E. coli BL21 cells transformed with the plasmid pET9a-MelB were grown overnight at 28°C in M9 media supplemented with 0.4% glucose; 2 mM MgSO 4 ; 1 M CaCl 2 ; 100 mg/liter of lysine, threonine, and phenylalanine; and 50 mg/liter of leucine, valine, isoleucine, and methionine. The pelleted cells were resuspended in fresh M9 media (same as above) with 100 mg/liter of selenomethionine instead of methionine for 1 h at 37°C before inducing the expression with 0.5 mM isopropyl ␤-D-thiogalactopyranoside overnight at 20°C. The cells were centrifuged at 4000 ϫ g for 15 min at 4°C. The purification protocol was the same as described above.

Crystallization and data collection of MelB
Crystallization conditions for seleniated MelB in the presence of 2 mM raffinose were screened using Qiagen kits (Valencia, CA) with a Cartesian NanoDrop robot (Genomic Solutions). The crystals were manually reproduced in hanging drops experiments by mixing equal volumes of protein solution and the precipitant solution 25% PEG 4000, 0.2 M NaCl, 0.1 M Mes, pH 6.5, and 0.2 M CaCl 2 . For the four other complexes, a similar condition without CaCl 2 and 0.6 M NaCl was used. Crystals were transferred to a cryoprotectant solution (mother liquor supplemented with 25% PEG 400) and flash-frozen in liquid nitrogen. X-ray diffraction data sets were collected at 100 K on the Proxima 1 or 2 beamlines (SOLEIL synchrotron, Saint-Aubin, France). Data processing was performed using the XDS package (32) ( Table 2).

Structure determination and refinement of MelB
The crystal structure of the MelB-raffinose complex was determined by SAD method from selenomethionine-labeled Galactinol and melibiose bound to MelB protein and refined at 2 Å resolution. Solvent content analysis using CCP4 (Collaborative Computational Project, Number 4) indicated the presence of four monomers in the asymmetric unit. The positions of 12 over 15 selenium atoms per monomer were found using SHELX suite program (33) The phases were calculated using PHASER (34) and density modification was performed by PARROT (CCP4 suite). An iterative process of manual building in COOT combined with phase calculation where a partial model was used as input, allowed the modeling of the complete polypeptide chain. The structures of all other liganded MelB were solved using the SeMet-MelB monomer as a search model. Refinement of each structure was performed with BUSTER-2.10 (34), NCS restraints, and TLS group. Because of the strong anisotropy of the crystals of MelBstachyose, the DEBYE and STARANISO programs developed by Global Phasing Ltd. were applied to the data scaled with AIMLESS using the STARANISO server (http://staraniso. globalphasing.org). 6 These programs perform an anisotropic cut-off of merge intensity data on the basis of an analysis of local I/s(I); compute Bayesian estimates of structure amplitudes, taking into account their anisotropic fall-off; and apply an anisotropic correction to the data. The corrected anisotropic amplitudes were used for further refinement of the MelBstachyose structure with BUSTER-2.10. Inspection of the density maps and manual rebuilding were performed using COOT (35). The three-dimensional models of stachyose and galactinol were generated with the ProDRG webserver (36), whereas those of melibiose and raffinose were found in the Protein Data Bank. Refinement details of each structure are shown in Table 2. Molecular graphics images were generated using PyMOL.

Fluorescence titration measurements of MelB
Each ligand bound to MelB was monitored by autofluorescence by exciting the protein at a wavelength of 295 nm and monitoring the quenching of fluorescence emission of tryptophans at 335 nm. All experiments were performed at 22°C in 96-well plates (1/2 Area Plate-96F, PerkinElmer Life Sciences) using Tecan Infinite M1000 (Tecan, Männedorf, Switzerland) in 25 mM Tris-HCl, pH 8.0, and 150 mM NaCl with a fixed amount of proteins (1 M) and increasing concentrations of ligand. Each ligand has no emission signal at 335 nm. The data were analyzed using Origin® 7 software and fitted to the following equation.

Isothermal titration microcalorimetry measurements of MelB
Isothermal titration microcalorimetry experiments were performed with an ITC200 isothermal titration calorimeter from MicroCal (GE Healthcare). The experiments were carried out at 20°C. Protein concentration in the microcalorimeter cell (0.2 ml) varied from 10 to 300 M. Nineteen injections of 2 l of ligand solution (raffinose, stachyose, melibiose, galactose, and galactinol) concentration from 0.1 to 2.8 mM were performed at intervals of 180 s while stirring at 500 rpm. The experimental data were fitted to theoretical titration curves with software supplied by MicroCal (ORIGIN®). This software uses the rela-tionship between the heat generated by each injection and ⌬H (enthalpy change in kcal mol Ϫ1 ), K a (the association binding constant in mol Ϫ1 ), n (the number of binding sites), total protein concentration, and free and total ligand concentrations (37).

Differential scanning calorimetry
Thermal stability of the WT and liganded MelB (13 M and 50 M for protein and ligand, respectively) was studied by differential scanning calorimetry (DSC) on a MicroCal model VP-DSC in a standard buffer. Each measurement was preceded by a baseline scan with the standard buffer. All solutions were degassed just before loading into the calorimeter. Scans were performed at 1 K⅐min Ϫ1 between 20 and 90°C. The heat capacity of the buffer was subtracted from that of the protein sample before analysis. Thermodynamic parameters were determined by fitting the data to the following equation, where K d is the equilibrium constant for a two-state process, ⌬H vH is the enthalpy calculated on the basis of a two-state process, and ⌬H cal is the measured enthalpy.

Phylogenetic analysis
Sequences were analyzed using BlastP from NCBI (https:// blast.ncbi.nlm.nih.gov/) and MicrosScope (https://www. genoscope.cns.fr/). 6 Alignments of MelB and related sequences were conducted using ClustalW software. Relationship tree was build using Mega software, version 7. The bootstrap consensus tree inferred from 1000 replicates was taken to represent the evolutionary history of the taxa analyzed. The evolutionary distances were computed using the Poisson correction method and are in units of the number of amino acid substitutions per site.

Measurement of mel operon gene expression
Expression of mel operon genes was measured in the C58 pOT1e-P mel strain. Quantification of fluorescence was carried out in a microplate filled with 200 l of AT medium, supplemented with different carbon sources at a final concentration of 10 mM. Microplate wells were inoculated with overnight cultures to obtain an A 600 of 0.2. A TECAN apparatus (Tecan Spark TM 15 M, Männedorf, Switzerland) was used to read microplates after 24 h of incubation at 28°C. The following parameters were used: absorbance at 600 nm, fluorescence excitation at 488 nm, and emission at 510 nm. Results were normalized by the A 600 and -fold change values were obtained by dividing the fluorescence by the corresponding value obtained from the empty pOT1e vector. The fluorescence level comparison was carried out using the Tukey's test (p value ϭ 0.05) and computed with the "vegan" package in the R v3.1.3 statistical software environment (R Core Team, 2014).

Plant inoculation
For bacterial colonization, competition assays, and confocal observation studies, tomato seeds (Solanum lycopersicum Galactinol and melibiose bound to MelB "Marmande") were sterilized as described (38). Seeds were plated on 0.8% agar plant cell culture supplemented with 1.5 g/liter of the Plant-Prod 15-15-30 High K nutrient solution (Master Plant-Prod Inc., Brampton, Ontario, Canada). They were inoculated with 10 l of overnight culture (10 6 cfu/ml) of a single strain (A. fabrum C58 pTiatu6148:K m derivative of the WT strain (39) or C58⌬melB mutant), or with a mixture of both at 1:1 ratio (competition). Petri dishes were placed 2 days in the dark and then in a climatic chamber at 24°C with 18/8 h for light/dark and 65% of humidity. To determine bacterial colonization level, roots were ground at 2 and 14 dpi. Serial dilutions of crushed roots were plated on the YPG medium and colonies were counted after 2 days of incubation at 28°C. Significant difference in the population level resulting from five plants per strain, with enumeration of three Petri dishes for each plant, was evaluated with Mann-Whitney test (p value ϭ 0.05) performed with the R v3.1.3 statistical software environment.
In competition experiments, the colonized bacteria were nonselectively recovered at 2 and 14 dpi. To that end, crushed roots were first plated with a spiral plater (EasySpiral®, Interscience, Saint-Nom-la-Bretèche, France) on YPG medium without antibiotics to enable a biologically unbiased recovery of both C58 pTiatu6148:K m WT strain and C58⌬melB mutant (kanamycin sensitive). Two hundred individual colonies were then plated in YPG medium with kanamycin/neomycin to determine the relative proportions of C58 pTiatu6148:K m and C58⌬melB mutant strains (output ratio). The determination of the initial strains ratio of the inoculum was realized using the same protocol. The experimental assays were performed with 10 independent assays and repeated four times. The proportions of WT strains between the initial strain ratio and in planta output ratios were compared with the test of equal or given proportions (p value ϭ 0.05).

Confocal microscopy analyses
Visualization of reporter bacterial cells harboring pOT1eM-P mel on tomato radicles and roots was performed using a confocal laser scanning microscope (LSM 800 Meta Confocal Microscope, Zeiss, Oberkochen, Germany). In the reporter strain, M-Cherry (red color) is constitutively expressed and enhanced GFP (eGFP) (green color) is expressed under P mel control. The red color indicates the bacteria presence (red cells) whereas eGFP reports the induction of P mel shown as yellowgreen cells. At 2 and 14 dpi, tomato radicles and roots were mounted between a slide and a coverslip in a commercial mounting fluid (Aqua Poly/Mount, Polysciences, Inc., Warrington, PA). The eGFP and the M-cherry were excited with argon laser at 488 nm and 584 nm, respectively, and fluorescence was captured at 528 nm and 607 nm. Analyses of images (five plants per condition) were performed thanks to LSM 800 software (Zeiss, Oberkochen, Germany).