LuxR Solos in the Plant Endophyte Kosakonia sp. Strain KO348

Cell-cell signaling in bacteria allows a synchronized and coordinated behavior of a microbial community. LuxR solos represent a subfamily of proteins in proteobacteria which most commonly detect and respond to signals produced exogenously by other microbes or eukaryotic hosts. Here, we report that a plant-beneficial bacterial endophyte belonging to the novel genus of Kosakonia possesses two LuxR solos; one is involved in the detection of exogenous N-acyl homoserine lactone quorum sensing signals and the other in detecting a compound(s) produced by the host plant. These two Kosakonia LuxR solos are therefore most likely involved in interspecies and interkingdom signaling.

that responds to plant compound(s) (Fig. 1B). Both LoxR and PsrR have two PFAM domains conserved among the LuxR-type family proteins, the autoinducer binding domain (PF03472) and the bacterial regulatory protein, LuxR-type DNA-binding HTH domain (PF00196).
LoxR is likely to respond to AHLs, whereas PsrR is likely to respond to a plant compound. In order to gain insights into the structure underlying autoinducer specificity of LoxR and PsrR, we have resorted to multiple-structure-based sequence alignment and to structure-based homology modeling. In particular, we focused on the pocket residues directly interacting with the ligand that are conserved and belong to three previously described clusters (37). Cluster 1 comprises the six conserved hydrophobic/aromatic residues previously reported (3,38), cluster 2 is composed of reasonably conserved residues directly involved in ligand binding (colored in green and cyan in Fig. 2 to 4), and cluster 3 consists of quite variable residues directly involved in ligand binding (variability patch colored in orange in Fig. 2 to 4).
This analysis revealed key differences between the ligand-binding pockets of LoxR and PsrR. We found in LoxR striking similarities to the structural determinants of the canonical QS LuxR family, while the ligand-binding site of the PsrR closely resembles the ones of PAB LuxR solo subfamily. In particular, not only has LoxR conserved the binding site core, including cluster 1 and 2 residues and delimiting the binding site floor and the distal wall, i.e., residues 70, 71, 72, 85, 113, and 129, according to The residues belonging to clusters 1 to 3 are highlighted in green, cyan, and orange, respectively. The 3D architecture of the boundaries of the ligand-binding site is schematized by r (roof), f (floor), p (proximal wall), and d (distal wall) and its tripartite topology by c (conserved core), s (specificity patch), and v (variable patch). TraR_At represents Agrobacterium tumefaciens TraR, and TraR_Sf represents Sinorhizobium fredii TraR.
Agrobacterium tumefaciens TraR (TraR_At) numbering (marked by "c" in Fig. 2), but also it has conserved all its residues of the specificity patch belonging to clusters 1 and 2 and mainly delimiting the binding site roof and the nearby proximal and distal walls, i.e., residues 57, 61, 73, 101, 105, and 110, according to TraR_At numbering (marked by "s" in Fig. 2). These differ from those in the subfamily of PAB LuxR solos and are conserved within the members of the canonical QS LuxR family, pinpointing that LoxR could have a common specificity toward AHLs. Residues belonging to both the conserved and the specificity patches in LoxR ( Fig. 3B and D, respectively) are identical to those of the corresponding regions in the canonical QS LuxR solo SdiA ( Fig. 3A and C, respectively). The crystal structure (PDB identifier [ID] 4LGW) of SdiA has been used as a template for this homology modeling since it displayed 68% and 83% overall primary sequence identity and homology, respectively. Moreover, some degree of conservation has been identified also in the respective variability patches ( Fig. 3E and F) that comprise cluster 3 residues (delimiting the proximal wall and the nearby roof and of the floor of the binding site, i.e., residues 49, 53, 58, and 62 according to TraR_At numbering), usually less conserved even within the members of canonical QS LuxR family. Indeed, besides two identical residues (i.e., Y63 and Q72), a conservative substitution (i.e., M68 in LoxR instead of V68 in SdiA) and a semiconservative substitution (i.e., V59 in LoxR instead of F59 in SdiA) have been found. Overall, a close specificity of LoxR for AHLs can therefore be inferred. Shown are results of mapping of the protein residues defining the three clusters (colored with the same color code as used in Fig. 2) that delineate the conserved core (A and B), the specificity patch (C and D), and the variability patch (E and F) on the structure of SdiA in complex with 3OC 6 -HSL (PDB ID 4Y15) (16) (A, C, and E) and on the 3D structure-based homology model of the LuxR solo LoxR (B, D, and F). The carbon, nitrogen, and oxygen atoms of the 3OC 6 -HSL ligand shown in the left column are represented by spheres and are colored in yellow, blue, and red, respectively. Structures were produced by PyMOL (73).
The molecular determinants of the PsrR binding site (Fig. 4, right) suggest a different specificity. All the residues belonging to the specificity patch are conserved with respect to the PAB LuxR solo subfamily (Fig. 4, left) and differ with respect to the canonical QS LuxR family. Among the residues of the roof of the binding site, PsrR and OryR share the residue W71, which belongs to cluster 1 and is highly conserved among all members of the PAB LuxR solo subfamily. Conversely, in the corresponding position, the canonical QS LuxR family is characterized by the presence of the highly conserved residue Y61 (according to TraR_At numbering). Similarly, the two roof residues PsrR and OryR V115 and a conserved hydrophobic/aliphatic residue (L/M) in position 120 are replaced by the quite conserved TraR_At F101 and A105 residues, respectively. Among the residues of the distal wall of the binding site, PsrR and OryR Q83, which belongs to cluster 2 and is highly conserved in the PAB LuxR solo subfamily, is replaced in the canonical QS LuxR proteins by a conserved hydrophobic/aliphatic residue (V/L), TraR_At V73. Nevertheless, important differences can be detected in cluster 3 residues that are less conserved: the only position in this cluster that PsrR and OryR share is C71. Indeed, PsrR is characterized by the presence of two charged residues (E and R, instead of hydrophobic/aliphatic residue M and of hydrophilic residue Q in positions 58 and 67, respectively) and V62, which replaces A62 in OryR. These key differences suggest a different specificity toward most probably unrelated plant compounds for PsrR and OryR. PipR is a recently reported PAB LuxR solo from Pseudomonas sp. strain GM79 and has been shown to respond to the plant compound N-(2-hydroxyethyl)-2-(2-hydroxyethylamino)acetamide (HEHEAA) (25). As PipR is the first PAB LuxR solo in which the inducing plant compound has been determined, it was of interest to compare its homology modeling to the one of PsrR. Figure 5 compares the ligand-binding sites of PsrR and PipR, highlighting the side chains of residues that are not conserved between PsrR and PipR. These residues define mainly the roof and part of the floor and of the proximal wall of the two binding pockets that are significantly different in terms of electrostatic potentials (E58 and R67 in PsrR instead of K96 and H105 in PipR) and hydrogen bond donor capabilities (V62 and M66 in PsrR instead of T100 and W104 in PipR) that may indicate different specificities of PsrR and PipR toward unrelated plant compounds.
LoxR is promiscuous and binds to a wide spectrum of AHLs. It was therefore of interest to determine if LoxR was able to bind AHLs due to its high homology to the SdiA three-dimensional (3D) structure and due to the conserved AHL binding docking residues (see above). His-tagged loxR was expressed in E. coli in the presence of different AHLs (C 6 -HSL, OHC 6 -HSL, OC 6 -HSL, OC 8 -HSL, and C 10 -HSL) by induction with 0.2 mM isopropyl-␤-D-thiogalactopyranoside (IPTG) overnight at 37°C. Figure S1 in the supplemental material compares the total extracts to the soluble fractions; LoxR was soluble when bound and insoluble when unbound to AHLs, as was also reported for many QS LuxR family proteins (39). This result indicated that LoxR was promiscuous toward many different AHLs, similar to what previously reported for its SdiA homolog.
To further verify the promiscuity of LoxR, the expression protocol was further modified because of the low stability of LoxR, which was due to proteolytic cleavage in the linker resulting in two protein bands with molecular weights (MW) of the two domains. The optimization as described by Studier (40) allowed the expression of loxR in the presence of two very different AHLs (C 6 -HSL and OC 12 -HSL); OC 12 -HSL was also able to solubilize LoxR (data not shown) like the other 6 AHLs shown in Fig. S1. LoxR was then purified by nickel-nitrilotriacetic acid (Ni-NTA) affinity chromatography followed by gel filtration (Fig. S2). After optimization of expression levels employing the autoinduction protocol (40), it was possible to obtain similar amounts of Lox complexes with a yield of 3 mg liter Ϫ1 . For confirmation of LoxR binding to AHLs, liquid chromatography-tandem mass spectrometry (LC-MS/MS) was performed for the protein eluted fractions corresponding to the putative LoxR-AHL complex ( Fig. S3 and S4). In these cases, selective ion monitoring was used to identify the presence of the coeluting unlabeled AHL. The identity of the binding protein was confirmed by MS-based protein identification (Tables S1 and S2). It was therefore concluded that AHLs solubilize and bind to LoxR.
LoxR and PsrR target gene studies. It was of interest to test whether some loci known to be regulated by LoxR or PsrR homologs in other bacteria were regulated by the two LuxR solos in Kosakonia. In the case of LoxR, we studied the transcription driven Shown is a superimposition of the 3D structure-based homology models of PsrR (A; colored in pale cyan) and PipR (B; colored in pale yellow), obtained from the orientation in Fig. 3 and 4 by 90°rotation around the y axis. Among the protein residues defining the three clusters (colored with the same color code as used in Fig. 2) that delineate the conserved core, the specificity patch, and the variability patch, the side chains of ones that are not conserved between PsrR and PipR are displayed. Structures were produced by PyMOL (73). by five gene promoters which have been reported to be regulated by SdiA (41). The loci regulated by these promoters were studied via transcriptional fusions with the lacZ reporter gene in a plasmid construct in Kosakonia KO348 and loxR mutant KO348 loxR. These gene promoters control the expression of the virulence-associated srgE gene and four type VI secretion system (T6SS)-related genes (one hypothetical gene and three vgrG genes). The T6SS has been shown to be regulated by SdiA in Enterobacter cloacae (41) and srgE in Salmonella enterica serovar Typhimurium. It was established that srgE, the hypothetical T6SS locus, and vgrG2 were not regulated by LoxR in the presence or absence of AHLs (Fig. 6). The promoters of vgrG1 and vgrG3, on the other hand, displayed differential expression in the presence and absence of AHLs in the KO348 loxR mutant in comparison to the wild type (WT) (Fig. 6). These differences were AHL independent and were only slight, albeit significant; hence, their meaning/implication remains to be determined. These little differences in gene promoter activities of vgrG1 and vgrG3 were restored to wild-type levels when the loxR mutant was complemented with a loxR gene in a plasmid (Fig. 6). Under the growth conditions used for the studies of gene promoter activities, no growth differences were observed between the KO348 wild type and all the derivatives (Fig. S5).
In the case of PsrR, since it belongs to the subfamily of PAB solos responding to plant low-molecular-weight molecules (15,25), it was of interest to study the expression of the adjacent pip gene, encoding a proline iminopeptidase. All members of the subfamily of PAB LuxR solos have a pip gene located adjacently which is regulated by the LuxR solo (15). We tested pip promoter activity in the presence of (i) rice root macerate, (ii) ethanolamine, and (iii) ethanolamine derivative HEHEAA. The last two compounds were recently reported to activate pip expression via the PAB LuxR solo PipR in a Pseudomonas endophyte (25). No PsrR-dependent pip promoter activity was detected and no induction of pip was observed in Kosakonia under any of the conditions tested ( Fig. S6A and B). Regardless of the fact that PsrR did not respond to plant macerate/extract, due to its very high homology to the PAB solos and proximity to the pip gene, it most likely responds to a plant-derived compound.
Phylogenetic clustering of PAB LuxR solos. In order to establish relatedness between PsrR and the PAB LuxR solos belonging to different plant-associated genera or species, a phylogenetic tree was created. LoxR is most closely related to other strains of Kosakonia and Enterobacter (data not shown). The tree of PAB LuxR solos belonging to many different plant-associated bacteria showed clustering of the proteins according to the genera that they belong to (Fig. 7). PsrR from Kosakonia and PipR from Pseudomonas GM79 were not closely related within this family, possibly suggesting that they might respond to related but different plant compounds, as supported in this study by the structural homology and pip gene promoter studies ( Role of LoxR and PsrR in in planta studies. Kosakonia strain KO348 is a very efficient rice root colonizer (35); thus, the role LoxR and PsrR LuxR solos play in colonization of the rhizoplane and root endosphere was studied. We generated loxR and psrR knockout mutants and determined rice rhizoplane and endosphere colonizing efficiencies. No statistically significant differences in colonization between the wild type and KO348 loxR were observed in the presence or absence of AHLs both in the rhizoplane and in the endosphere (Fig. S7). Similarly, no statistically significant difference in colonization of the rhizoplane between the wild type and KO348 psrR was observed. In the root endosphere, however, the wild type was a significantly better colonizer than the mutant KO348 psrR (Fig. 8), indicating the involvement of PsrR in the rice endosphere colonization of Kosakonia KO348.

DISCUSSION
LuxR solos are very widespread in proteobacteria (8,10,23); however, only a few have been thoroughly studied. Here, two LuxR solos in the recently described genus Kosakonia are reported; LoxR is an SdiA homolog and is able to bind AHLs promiscuously, whereas PsrR belongs to the PAB subfamily that responds to plant compounds.
Modeling analysis of LoxR showed high conservation of amino acid residues forming the binding pocket just like in E. coli SdiA (PDB ID 4LGW), indicating that it is likely to bind AHLs. LoxR has conserved amino acids Y63, W67, Y71, D80, and S134 (according to SdiA numbering) (Fig. 2), which are important for the docking of different types of AHLs (42). Experimental evidence for LoxR binding to AHLs include protein solubilization in the presence of AHLs, which is likely necessary for proper folding and homodimerization (39), and detection of LoxR-AHL complexes in the purified recombinant protein. It is therefore most likely that LoxR detects and responds to a wide range of exogenous AHLs produced by neighboring bacteria, as is the case of SdiA from Salmonella and E. coli (19,43). Promiscuity of LuxR receptors that respond broadly to AHL signals occurs also in QS LuxR proteins which possess cognate LuxI and hence can respond to non-self-signals (44). This property has implications in interspecies signaling, as bacteria are most commonly members of multispecies communities. SdiA is well conserved among members of the Enterobacteriaceae family; SdiA orthologs are present in Escherichia, Kosakonia, Salmonella, Enterobacter, Citrobacter, Cronobacter, Klebsiella, Pantoea, and Erwinia (45).
As SdiA is likely to be involved in interspecies signaling, identifying target loci will provide insights into its role in bacterial community lifestyle. In this study, some SdiA targets determined in other bacterial species were tested for LoxR regulation in Kosakonia. The srgE gene, encoding a type III effector protein, is regulated by SdiA in Salmonella (18,19). The srgE gene, however, is not regulated by LoxR in Kosakonia under the conditions tested in this study, indicating that it could have evolved to regulate a different set of target genes. In E. cloacae, SdiA regulates expression of a hypothetical protein from the T6SS cluster (41); interestingly, we have determined that in Kosakonia, two T6SS vgrG loci are negatively regulated by LoxR, indicating that the T6SS might be a conserved SdiA target. In Kosakonia KO348, T6SS plays a role in rhizoplane and rice root endosphere colonization (35); thus, LoxR could play a role in planta by regulating genetic loci in response to AHLs produced by neighboring bacteria in the rhizosphere and endosphere. The performed in planta experiments do not corroborate this possibility; however, in the wild, growth conditions are different from the experimental setup used in this study, since Kosakonia is likely to coexist with many more different microbes (see below). SdiA homologs and the T6SS are commonly found in endophytic bacterial genomes; however, they are more prevalent among the genomes of phytopathogens (46). More studies are needed to understand the role and regulation of T6SS by SdiA homologs in plant-beneficial bacteria.
In planta studies performed in this investigation showed no involvement of LoxR in rhizoplane or endosphere colonization. In a PGP E. cloacae strain, however, an SdiA mutant displayed 4-fold more efficient rhizoplane colonization with respect to that of the wild type (47). The in planta experiment described here has some limitations since it lacked soil containing a rich and native microbial community, which is likely to undergo cell-cell communication, including AHL signaling. AHLs were added in the experimental setup; however, it is not known if they were stable and remained at high concentrations throughout the days of the experiment. Therefore, the role of SdiA in root colonization under wild/field conditions may not be inferred from the results presented here.
PsrR is another member of the subfamily of PAB LuxR solos which responds to plant compounds (15). Homology modeling was employed to compare the structural determinants of PsrR with two previously described PAB LuxR solo proteins, OryR from the rice pathogen Xanthomonas oryzae pv. oryzae (22) and PipR from endophytic Pseudomonas strain GM79 (48). PipR is the only member of this subfamily which has been shown to respond to a specific plant compound (25). Importantly, some key differences between PipR and PsrR were found. PsrR displays the two important amino acids, M and W (at positions W57 and Y61 according to TraR numbering), that are part of the autoinducer binding site; these amino acids are conserved among many PAB LuxR solo subfamily members (49) and are also shared between PsrR and OryR. However, at the same positions, PipR shows a W and W (at positions W57 and Y61 according to TraR numbering) (23). This indicates a key difference in the binding site residue conformation between PsrR and PipR, probably resulting in a different response/specificity to plant compounds. PipR activates pip target gene expression in response to an ethanolamine derivative, N-(2-hydroxyethyl)-2-(2-hydroxyethylamino)acetamide, at particularly low concentrations (25); such a response did not occur in Kosakonia KO348. It cannot be excluded that this assay does not work under the conditions tested due to the lack of a positive response by PsrR in the regulation of the pip gene. The PsrR primary structure has a difference in the binding pocket conformation in comparison to PipR, which possibly suggests different inducer specificity, supporting its likely nonresponsiveness to the PipR signal. This difference in specificity could be the reason for the distance in clustering observed in the phylogenetic tree (Fig. 7). This analysis evidences clustering according to the genus and indicates possible evolution of binding to different, possibly related plant compounds.
The PAB LuxR solos are believed to regulate in response to a plant compound; the adjacently located pip gene encodes a proline iminopeptidase that is believed to be involved in the signal response (48). We tested whether PsrR could respond to plant extracts and regulated pip gene expression. Rice root extract could not activate pip expression in root endophyte Kosakonia KO348; several pip genes have, however, been shown to be activated by plant extract via PAB LuxR solos, as, for example OryR, PipR, and XccR (24,48,50), whereas some could not (49). The reason for this lack of response by plant extract by PsrR is currently unknown; it cannot be excluded that the signal molecule is present in very low concentrations or only in specific plant parts and/or growth stages. Alternatively, psrR could be itself regulated and under these experimental conditions the expression level is too low. In the promoter of the adjacent pip genes, a DNA-binding region resembling a lux box has been reported for several members of this LuxR solo subfamily (22). In this case, a DNA region resembling a lux box (or pip box) is present at the region from Ϫ19 to Ϫ39 upstream of the ATG start codon of the pip gene (CCCTGAAAGCCAGTTCACTT) and Ϫ176 to 196 of the psrR gene (CCCTGGTT ATCCGCGCCCGG). These findings could indicate regulation and autoregulation by PsrR, respectively; future research is needed to confirm the possible functionalities of these putative DNA-binding sites. It is also noted that the intergenic region between psrR and pip (Fig. 1B) is only 84 bp; hence, we cannot exclude an operonic structure between psrR and pip.
In planta studies indicated that PsrR is involved in root endosphere colonization, since the mutant displays significantly less colonization. This phenotype could not, however, be complemented when providing the wild-type gene in a plasmid; a possible explanation could be the high percentage of plasmid loss in the endosphere, as previously reported for Kosakonia (35). Other PAB LuxR solos have been implicated in plant colonization, for example, PsoR of Pseudomonas fluorescens, XocR of X. oryzae, PsaR2 of Pseudomonas syringae pv. actinidiae, and XagR of Xanthomonas axonopodis pv. glycines (21,23,49,51).
This study further highlights the presence of LuxR solos in proteobacteria, which are likely to play important roles in cell-cell signaling. These have most likely evolved from luxR genes which were part of a luxI/R system (45) and some have then evolved to respond to other endogenous or exogenous signals. A future challenge will be to identify the signals that they respond to.
Structure homology modeling. The structure-based sequence alignment of the ligand-binding domains of LoxR and PsrR was performed using the program Expresso (53) (Fig. 2). The following primary sequences were included in the multiple alignment: PipR from Pseudomonas sp. GM79 and OryR from Xanthomonas oryzae, prototypes of the PAB LuxR solo subfamily, and TraR from Agrobacterium tumefaciens TraR_At (PDB ID 1H0M) (54) and from Sinorhizobium fredii NGR234 TraR_Sf (PDB ID 2Q0O) (55) prototypes of the canonical QS LuxR proteins. The templates used in the structure-based homology modeling of LoxR and PsrR were SdiA from Escherichia coli (PDB ID 4LGW) (56) and QscR (PDB ID 3SZT) (57) and LasR (58) from P. aeruginosa. Structure-based homology modeling of full-length LoxR and PsrR was performed exploiting several approaches, and the resulting models were ranked according to protein model quality assessments based on LG score and MaxSub. In order to elucidate the molecular determinants of ligand discrimination in LoxR and PsrR, a cartographic analysis of the top-ranked models has been exploited as described in reference 37.
Several web-based servers were used to build the 3D structure-based homology models of full-length LoxR, PsrR, and PipR. The top-score models generated by the servers were then ranked and validated by the protein model quality predictor ProQ (59), employing PSIPRED (60) for the secondary-structure prediction. The IntFOLD server (61) produced the highest-quality 3D models for LoxR and PsrR from Kosakonia, according to the ranking obtained by ProQ, i.e., a predicted LG score and MaxSub value of 4.749 and 0.486, respectively, for LoxR and 4.125 and 0.706, respectively, for PsrR. The protein model of PipR was obtained by RaptorX (62), with a predicted LG score and MaxSub value of 4.039 and 1.010, respectively.
The templates used for LoxR modeling were SdiA from E. coli (PDB ID 4LGW [56]) and QscR from P. aeruginosa (PDB ID 3SZT [57]). The PsrR model was obtained using the same templates as for LoxR combined with LasR (PDB ID 3IX3 [58]) from P. aeruginosa. The template used for PipR modeling was QscR from P. aeruginosa (PDB ID 3SZT [57]).
Mass spectrometry. Aliquots of the peak fractions described above were denatured by heating to 95°C for 5 min and then subjected to reduction, alkylation, and digestion with trypsin. After digestion, they were dried down and resuspended in 20 l of 0.1% formic acid. LC-MS/MS of the digestion was performed using an Easy-nLC II coupled to an Amazon ETD mass spectrometer (Bruker Daltonics, Hamburg, Germany). The same digests were used to screen for the presence of AHLs. The elution times and fragmentation behavior of the AHLs were scouted using purified standards (C 6 -HSL and OC 12 -HSL; Sigma-Aldrich Inc., St. Louis, MO). Protein identification was performed using the X!tandem (www .thegpm.org) search engine and a Kosakonia KO348 database.
Construction of KO348 loxR and KO348 psrR. To generate KO348 loxR, an internal fragment (312 bp) of the loxR gene was amplified by PCR using the primers pKNloxR.Fw (GATGCAGCATTATCAG GCAGA) and pKNloxR.Rv (TGATCTTCCAGACGCGTTAA) and cloned as a BamHI-KpnI fragment into the corresponding sites of pKNOCK-Km (63), resulting in pKNOCKloxR. To generate KO348 psrR, an internal fragment (245 bp) of the psrR gene was amplified by PCR using the primers pKNpsrR.Fw (CTATCAACG CCCGGACAG) and pKNpsrR.Rv (ACAGCGGAAAGGCAGATT) and cloned as a BamHI-KpnI fragment into the corresponding sites of pKNOCK-Km, resulting in pKNOCKpsrR. Plasmid constructs were delivered to Kosakonia KO348 by electroporation, and transformants were selected after appropriate antibiotic selection. The loxR and psrR full-length genes (including their gene promoters) were amplified with the primers LoxR-comp_Fw (GGATCCCCGTTAACGTTGGCGTTAAA) and LoxR-comp_Rv (GAATTCTTTAAATCAT CCCTGTCGCC) for the loxR gene and PsrR-comp_Fw (GGATCCACGGTGCATCAGCATTCTCC) and PsrR-comp_Rv (GAATTCGGCGCTGAACACTAGCAAAA) for the psrR gene; the sequences were verified via DNA sequencing, and the resulting fragments were cloned into the pBBR1MCS-5 vector (64). The plasmids were electroporated in the mutant strains KO348 loxR and KO348 psrR, respectively, and selected for Km r and Gm r ; the resulting mutant complemented strains were named KO348 loxR(pBBRloxR) and KO348 psrR(pBBRpsrR). Mutants and complemented mutants were verified by colony PCR and DNA sequencing.
␤-Galactosidase activity was determined as previously described (52), with the modifications described in reference 66. Promoter activities were performed in biological triplicates at the onset of stationary phase. When necessary, two AHLs (C 6 -HSL and OC 12  Phylogenetic tree construction for PsrR. The amino acid sequence of Kosakonia KO48 PsrR solo was searched by BLAST against the genomes available in the U.S. Department of Energy IGM/M database, and the top homologs were identified. Representative homologs from the genera with previously described LuxR plant-associated solo subfamily sequences were chosen to be displayed phylogenetically. The sequences were aligned using Clustal Omega standard settings with FASTA-formatted output. The alignment was trimmed using TrimAI v1.3 using the "automated1" flag. The trimmed alignment was put into Fasttree 2.1.10 using standard settings. The tree was visualized on Interactive Tree of Life (67)(68)(69)(70)(71).
Rice root colonization assays. Rice root colonization experiments were performed as described previously (34), with the following modifications. Briefly, each strain was grown to an optical density at 600 nm (OD 600 ) of 0.8 and used to inoculate 9-day-old germinated rice plant roots (cultivar Baldo) by submerging them in the bacterial suspension for 1 h. Plants were then transferred to a 50-ml tube containing Hoagland's semisolid solution (72); when necessary, AHLs were added at a concentration of 1 M. All plants were followed for 14 days and then the Kosakonia strain was reisolated.
For rhizoplane colonization, rice roots were rinsed with sterile water for removing all adhered Hoagland's solution and then vortexed in 5 ml of phosphate-buffered saline (PBS) solution for 1 min. Serial dilutions of this PBS solution containing bacteria were then plated on the appropriate media with antibiotics for calculation of CFU per gram. For the determination of endosphere colonization, the roots were sterilized as described previously (34), macerated in PBS, and then plated after serial dilutions in Trypticase soy agar (TSA) containing the appropriate antibiotics; the plates were then incubated at 30°C for 48 h and CFU per gram were calculated. For comparing the rhizoplane and endosphere colonization ability, analysis of variance (ANOVA) was performed with Prism 7 (GraphPad Software, Inc.). Five biological replicates were used in each group.

SUPPLEMENTAL MATERIAL
Supplemental material is available online only. SUPPLEMENTAL FILE 1, PDF file, 5.1 MB.