Comparative Secretome Analysis of Ralstonia solanacearum Type 3 Secretion-Associated Mutants Reveals a Fine Control of Effector Delivery, Essential for Bacterial Pathogenicity*

Ralstonia solanacearum, the causal agent of bacterial wilt, exerts its pathogenicity through more than a hundred secreted proteins, many of them depending directly on the functionality of a type 3 secretion system. To date, only few type 3 effectors have been identified as required for bacterial pathogenicity, notably because of redundancy among the large R. solanacearum effector repertoire. In order to identify groups of effectors collectively promoting disease on susceptible hosts, we investigated the role of putative post-translational regulators in the control of type 3 secretion. A shotgun secretome analysis with label-free quantification using tandem mass spectrometry was performed on the R. solanacearum GMI1000 strain. There were 228 proteins identified, among which a large proportion of type 3 effectors, called Rip (Ralstonia injected proteins). Thanks to this proteomic approach, RipBJ was identified as a new effector specifically secreted through type 3 secretion system and translocated into plant cells. A focused Rip secretome analysis using hpa (hypersensitive response and pathogenicity associated) mutants revealed a fine secretion regulation and specific subsets of Rips with different secretion patterns. We showed that a set of Rips (RipF1, RipW, RipX, RipAB, and RipAM) are secreted in an Hpa-independent manner. We hypothesize that these Rips could be preferentially involved in the first stages of type 3 secretion. In addition, the secretion of about thirty other Rips is controlled by HpaB and HpaG. HpaB, a candidate chaperone was shown to positively control secretion of numerous Rips, whereas HpaG was shown to act as a negative regulator of secretion. To evaluate the impact of altered type 3 effectors secretion on plant pathogenesis, the hpa mutants were assayed on several host plants. HpaB was required for bacterial pathogenicity on multiple hosts whereas HpaG was found to be specifically required for full R. solanacearum pathogenicity on the legume plant Medicago truncatula.

livers virulence factors, the type 3 effectors (T3Es) into the host cells. Collectively, these effectors constitute one of the main weapons of the pathogenicity arsenal of R. solanacearum, as it is the case for many other Gram-negative phytopathogenic bacteria (i.e. Pseudomonas spp., Xanthomonas spp., and Erwinia spp.) (4). Many early substrates of the secretion apparatus are highly conserved among these bacteria, including the cytoplasmic, inner and outer membrane ring components, and the needle. The translocators (intermediate substrates) and the translocated effectors (late substrates) show more diversity, which may contribute to the host specificities of each pathogenic bacterium (4 -6). The core component proteins are encoded by the hrp (hypersensitive response and pathogenicity) and hrc (hrp-conserved) genes, located on the hrp gene cluster, whereas the T3Es are distributed throughout the genomes of these bacteria (7)(8)(9)(10).
R. solanacearum strains possess large T3E repertoires, with 72 Rips (Ralstonia injected proteins) identified in the model strain GMI1000 (11). Rip delivery through the T3SS is under a fine transcriptional control, orchestrated by the regulatory protein HrpB. This transcription factor activates both the expression of the T3SS encoding genes and of the T3E genes (12). The R. solanacearum T3SS transcriptional regulatory system is well described (13), whereas little is known about the post-translational control of Rip delivery. There were 94 Rip genes identified among strains of the R. solanacearum species complex sequenced to date (11). This large repertoire of virulence factors may explain the wide host range of this bacterium. However, because of Rip redundancy (14 -16), single Rip are often dispensable for bacterial pathogenicity on a given host, except for RipG7 (formerly named GALA7) on the legume plant Medicago truncatula (17). To date, many questions remain, we still do not know whether there could be a specific regulation of Rip delivery, depending on the host plant, the host tissues, or on the stage of bacterial colonization. Are all Rips delivered into the host at the same time, in equal quantities, or is there a hierarchical and quantitative control?
To better understand this complex mechanism, we studied the delivery of Rips via the T3SS through in vitro Rip secretion. We used optimized secretion conditions (18) in which the bacterium is in an HrpB-inducing environment, mimicking the in planta conditions and leading to Rip secretion. With this standardized and controlled system, we aimed at identifying sets of Rips with conserved and typical secretion patterns. Like it has been shown for other pathogenic bacteria (19,20), we can hypothesize that Rip delivery is post-translationally controlled by helper proteins. These helper proteins are described in both plant and animal pathogenic bacteria, but are not orthologs. They are known to regulate T3E transit through the T3SS at post-transcriptional and post-translational levels, stabilizing T3Es, preventing their degradation or ordering and mediating their recognition by the T3SS (19,21,22). Some secretion helper proteins have been well described, like the T3Cs (type 3 chaperones), or the T3S4 (type 3 secretion substrate specificity switch) proteins (6,22). Our previous work on R. solanacearum HpaP, a putative T3S4 protein, revealed that this regulatory protein could act as a modulator of the secretion level, targeting some T3Es, notably via direct protein interactions (23).
To advance our knowledge on the R. solanacearum T3S and on its post-translational control, we explored the complete secretome of R. solanacearum, taking advantage of the most recent tandem MS approach. In this work, we describe the most complete secretome (228 proteins including Rips) of a plant pathogenic bacterium using a MS-based shotgun approach with a label-free quantification (24). This allowed us to compare secretomes between hpa (hrp-associated) mutants and evaluate their respective involvement in T3S. In parallel, the phenotypic analysis of these hpa mutants on different host plants was carried out. This first in-depth proteomic approach on R. solanacearum enhances our knowledge on the T3S regulation and is a new mean to identify sets of Rips potentially coregulated, likely to be relevant for host specificity.

EXPERIMENTAL PROCEDURES
Bacterial Strains, Plasmids, and Growth Conditions-The bacterial strains used in this work are listed in Table I. Escherichia coli strains were grown at 37°C in Luria-Bertani medium. Ralstonia solanacearum strains were grown in complete B medium or minimal medium, as previously described (25). When needed, antibiotics were added at the following final concentrations (mg/L): kanamycin (50), spectinomycin (40), gentamycin (10), tetracycline (10) for R. solanacearum and kanamycin (25), spectinomycin (40), gentamycin (10), tetracycline (10), ampicillin (50), chloramphenicol (25) for E. coli. The plasmids used in this study, listed in Table I, were constructed by Gateway technology (Life Technologies, Carlsbad, CA) following the instructions of the manufacturer. Genes cloned in the pDONR207 plasmid were amplified in two steps. The first PCR was performed using the specific primers indicated in Supplemental Experimental Procedures. The second PCR was performed using 1 l of the first PCR as matrix and attB universal primers (oNP291 5Ј-GGGGACA-AGTTTGTACAAAAAAGCAGGCTTCGAAGGAGATAGAACCATG-3Ј and oNP292 5Ј-GGGGACCACTTTGTACAAGAAAGCTGGGTC-3Ј).
Generation of R. solanacearum Mutants-R. solanacearum mutants are listed in Table I. hrcV, hpaG, and hpaD mutants were previously generated (12). In this study we constructed hpaB and ripBJ (RSp0213) mutants. To generate hpaB mutation in the GMI1000 wild-type strain, a 1.6 kb DNA fragment encompassing hpaB was PCR-amplified using the following primers G-UPLNG 5Ј-AAGCTTC-CCGGACGGCTCAACGC-3Ј and HPAB-rev 5Ј-AAGCTTGTAGATGT-GCTGCTCGA-3Ј and then cloned into the pUC18 vector. The ⍀ -spectinomycin interposon (26) was cloned into the unique NotI restriction site within the hpaB open reading frame to generate plasmid pSG424. pSG424 was linearized by XbaI and used to transform R. solanacearum as previously described (27). Strain GRS474 was obtained by selecting a double recombination event using spectinomycin resistance and hpaB gene disruption was then checked by PCR. The same strategy was used to generate the strain GRS503 (RSp0213 mutant). p213fw 5Ј-TCTAGATCCAGGTGGCGCTGCAG-3Ј and p213rev 5Ј-GTTCGACGTGGATGCGGGC-3Ј primers were used to amplify and clone a 1.6 kb fragment containing RSp0213 into the pUC vector. The ⍀ -spectinomycin interposon (26) was cloned into the SmaI restriction site of RSp0213 (pSG493). Transformation into R.
solanacearum was performed and checked as previously described for generation of GRS474 strain.
Secretome Sample Preparation-Five R. solanacearum strains were used for MS secretome analysis: GMI1000 wild-type strain, GMI1694 hrcV mutant, GRS474 hpaB mutant, GRS266 hpaD mutant and GRS230 hpaG mutant (Table I). These R. solanacearum strains carrying the pAM5 plasmid (18), that leads to a higher transcriptional activity of T3SS-regulated genes, were first cultivated in B medium overnight, pelleted for 5 min at 4000 rpm and then resuspended in 15 ml of minimal medium (supplemented with 0.1 g/L congo red and 10 mM glucose and 10 mM glutamate, with 10 g/ml tetracycline) at OD 600 nm ϭ 0.2. Bacterial cultures were then harvested after 8 h of culture, set at the same concentration, according to the OD 600 nm and  1). Digestion was performed in 50 mM ammonium bicarbonate pH 8.0 and the quantity of modified trypsin (Promega, sequencing grade, Madison, WI) was 0.1 g per sample. Digestion was achieved for 6 h at 37°C. The supernatant was conserved. Peptides were extracted by 5% formic acid in water/acetonitrile (v/v). Supernatant and extracted tryptic peptides were dried and resuspended in 50 l of 0.1% (v/v) formic acid and 2% (v/v) acetonitrile.
Nano Liquid Chromatography Coupled to Mass Spectrometry in Tandem-A Q Exactive (Thermo Fisher Scientific) coupled to Eksigent 2Dnano LC (AB-Sciex, Framingham, MA) were used for the nano-LC-MS/MS analysis. Four microliters of sample were injected on the nanoLc-Ultra system (AB-Sciex) chain. Sample was loaded at 7.5 l/min on the precolumn cartridge (C18, 5 m, 120 Å, 20 mm Nanoseparations) and desalted with 0.1% formic acid. Then, peptides were separated with a gradient of acetonitrile on the reverse phase column C18 (stationary phase: C18 Biosphere, 3 m; column: 75 m i.d., 150 mm; Nanoseparations). Buffers were 0.1% formic acid in water (A) and 0.1% formic acid in acetonitrile (B). The peptide separation was achieved with a linear gradient from 5 to 30% B for 40 min at 300 nL/min Ϫ1 . Eluted peptides were analyzed on-line with a Q Exactive mass spectrometer (Thermo Ficher Scientific, San Jose, CA) using a nanoelectrospray interface. Ionization (1.8 kV ionization potential) was performed with stainless steel emitters (30 m i.d.; (Thermo Ficher Scientific). Peptide ions were analyzed using Xcalibur 2.1 with the following data-dependent acquisition steps: (1) full MS scan (massto-charge ratio (m/z) 400 to 1400) and (2) MS/MS. Step two was repeated for the eight major ions detected in step one. Dynamic exclusion was set to 40 s. Lock mass option was chosen "best," MS resolution 70,000 at m/z 400, auto gain control was 1e 6 , maximum injection time 250 ms. For MS2 the resolution was 17,500 at m/z 400, auto gain control was 2e 5 , maximum injection time 120 ms, isolation window m/z ϭ 3, normalized collision energy: 30, underfill ratio 0.5%, intensity threshold 8.3e 3 . Charge state: 2, 3, 4; dynamic exclusion 40 s.
Data Processing and Bioinformatics Analysis-The R. solanacearum strain GMI1000 database was downloaded from Database site (https://iant.toulouse.inra.fr/bacteria/annotation/cgi/ralso. cgi, October 2012, 5122 protein entries). This database was merged and in conjunction with reverse and contaminant databases, were searched by X!Tandem (Sledge Hammer version 2013.09.01.1, http:// www.thegpm.org/tandem/) using X!TandemPipeline (version 3.3.3) developed by PAPPSO platform (http://pappso.inra.fr/bioinfo/). Enzymatic cleavage was declared as a trypsin digestion with one possible miss-cleavage. Cys carboxyamidomethylation and Met oxidation were set to static and possible modifications, respectively. Precursor mass was 10 ppm and fragment mass tolerance was 0.02 Th. A refinement search was added with similar parameters except for semitrypsic peptides and possible N-ter proteins. For data of proteomic, only peptides with an E value smaller than 0.1 were reported. Identified proteins were filtered and grouped using X!TandemPipeline (http://pappso.inra.fr/bioinfo/xtandempipeline/) according to: (1) a minimum of two different peptides was required with an E value smaller than 0.05, (2) a protein log (E value) (calculated as the product of unique peptide E values) smaller than 10 Ϫ3 . These criteria led to a false discovery rate (FDR) of 0.08% for the wild-type strain and of 0.07% for the five strains (wild-type, hrcV and hpa mutants) for peptide and protein identification. To take redundancy into account, proteins with at least one peptide in common were grouped. This allowed to group proteins of similar function. Within each group, proteins with at least one specific peptide relatively to other members of the group were reported as subgroups. Only proteins detected on at least two samples on at least one strain were considered. Labelfree quantification of proteins was achieved in two steps. The first one is the spectral counting (SC) (28 -30), which is a strategy to determine a relative quantification of protein from their number of spectra obtained with tryptic peptides in MS. This quantification is based on the fact that the more of a particular protein is present in a sample, the more MS spectra are detected for peptides of that protein. The second step is performed using protein abundance index (PAI) for every protein and was obtained by dividing SC data of the observed protein by the number of theoretically observable tryptic peptides for that protein (31).
Experimental Design and Statistical Rationale-Concerning the secretome samples used for MS experiments, four independent biological replicates were generated for each of the five strains studied (wild-type strain, hrcV, hpaB, hpaD, and hpaG mutants). We found that four replicates were the best possible compromise in order to duly handle experimental variability and to do quantitative analysis. The hrcV mutant was used as a negative control of T3S, allowing discrimination of type 3 proteins. In order to be compared, samples were normalized dividing each PAI by the total sum of PAIs on the corresponding sample. Thus normalized PAI values were log-scaled to be able to adjust a linear model for every protein. With the aim of better handling multiple comparisons, general linear hypothesis tests (glht) (32) were subsequently run on the model to discriminate proteins differentially quantified between bacterial strains. All statistical procedures were carried out under [R] v3.1.1 environment (33) using base packages as well as "multcomp" (32), "limma" (34), and "venneuler" (35) packages. The R source code created for this study is available upon request.
Translocation assays were performed as previously described on Nicotiana tabacum (37). The cAMP production was measured using the cAMP Biotrak enzyme immunoassay (EIA) System kit by GE Healthcare (Buckinghamshire, UK). Four independent biological replicates were performed.
Plant Assays and Statistical Analysis-For Arabidopsis thaliana assay, 16 Col-0 ecotype plants were inoculated as previously described (38) using 10 8 bacteria/ml, without cutting the roots. For M. truncatula (ecotype A17), two types of inoculation were performed. For soil inoculations, 16 M. truncatula plants with cut roots were soaked with 800 ml of a bacterial solution at 10 6 bacteria/ml. In vitro inoculations were done as previously described (39) with 10 7 bacteria/ml. Disease development was monitored every day, and plants with at least 50% of wilting were considered as dead for the statistical survival analysis. To compare the disease development of two given strains, we used the Kaplan-Meier survival analysis with the Gehan-Breslow-Wilcoxon test. A p value smaller than 0.05 was considered significant, indicating that the Ho hypothesis of similarity of the survival experience of the tested strains can be rejected. Statistical analyses were done with Prism version 5.00 (GraphPad Software). Each experiment was repeated at least three times.
Internal Growth Curve-Internal Growth curve was determined mostly as previously described (39). Fourteen days post-inoculation, at least three pools of three plants were harvested, sterilized in 70% ethanol and rinsed three times in water. Plants were weighted, grinded and then resuspended in sterile water. Bacterial concentrations were determined by plating dilutions on B medium with antibiotic if necessary. Experiments were repeated three times.

Secretome of R. solanacearum GMI1000 Grown in Type 3
Secretion Inducing Conditions-Using LC-MS/MS, we analyzed the secretome of the R. solanacearum GMI1000 wildtype strain. Bacteria were grown in secretion inducing media, and four independent biological secretome samples were generated by a shotgun approach with label-free quantification. The experimental procedure scheme is shown in supplemental Fig. S2A. Identification of peptides was done using X!TandemPipeline developed by PAPPSO platform (http:// pappso.inra.fr/). Protein identification was done using the R. solanacearum genome database (https://iant.toulouse.inra.fr/ bacteria/annotation/cgi/ralso.cgi) and a contaminant database (trypsin, keratins, etc). A protein was considered as present in the supernatant when spectra were detected in at least two independent replicates. The number of spectra corresponding to each detected protein, the normalized protein abundance index (PAI), and all the peptide sequences are provided in supplemental Table S1. The analysis of strain GMI1000 secretome identified 228 proteins which were classified into functional groups according to the R. solanacearum proteome database (see supplemental Table S1; Fig. 1). A Bonferroni-corrected hypergeometric test (with a p value Ͻ 0.001) revealed a significant enrichment for three categories corresponding to macromolecule metabolism, structural elements, and type 3-associated proteins (Fig. 1). Among the 32 proteins belonging to the "macromolecule metabolism" class many previously characterized R. solanacearum secreted proteins were found, e.g. plant cell wall degrading enzymes (endoand exoglucanases, polygalacturonases) (supplemental Table  S1). The category of type 3-associated proteins showed the highest enrichment (p value ϭ 5.72.10 Ϫ34 ). The 38 detected type 3-associated proteins included three structural elements of the syringe, the HrpY pilin (the major component of the Hrp pilus) and the two translocators RipF1_1 and RipF1_2 (formerly named PopF1 and PopF2), which are thought to form a pore in the plant cell (supplemental Fig. S2B). The 35 other type 3 proteins correspond to Rip protein effectors.
Type 3 Secretome of R. solanacearum-In order to define the set of T3-secreted proteins, we compared the GMI1000 wild-type secretome with the T3SS-defective hrcV mutant. There were 224 proteins detected in the hrcV mutant culture supernatant (supplemental Fig. S3). A principal component analysis (PCA) on log-normalized PAI data was carried out to identify and visualize the sources of variability between all eight analyzed samples (four for each bacterial strain). The PCA on Fig. 2A clearly shows that the "strain" effect defining the first principal component, explains 66% of the overall variation. The positioning of the individual proteins on the PCA highlights that variance on this first component is quasi-exclusively because of T3-associated proteins (Fig. 2B). A similar analysis conducted excluding the annotated T3-associated proteins, shows that the strain effect is still present (supplemental Fig. S4A), but in this case the first principal component only explains 30% of the overall variance (supplemental Fig. S4B), with very few discriminating proteins being responsible for this component (supplemental Fig.   S4C). A linear model was fitted and the subsequent glht post-hoc tests were performed to discriminate the proteins differentially secreted between GMI1000 and the hrcV mutant, revealing 63 proteins differentially secreted (p value Ͻ 0.05) (supplemental Table S2). Twenty-nine out of these 63 proteins were found to be T3-associated proteins (HrpY and 28 Rips). These 29 proteins are among the most significantly different between both strains with the lowest p values (supplemental Table S2). The secretion pattern of all the T3-proteins detected in both strains is shown in Fig. 2C. A noticeable point is that none or, at best, very few spectra were detected for any of the Rip T3Es (except RipW) in the hrcV mutant (supplemental Table S3, Fig. 2C). The barely detected T3S SC data on the hrcV replicates suggest a minimal bacterial lysis in our secretome samples. Overall, these first data can be considered as  ). B, protein distribution on the same PCA shows that the most discriminating proteins are T3-associated proteins (in red), and reveals a new Rip candidate: RSp0213 (in purple). C, log-normalized PAI values of T3-associated proteins detected in the wild-type (red) and in the hrcV mutant (green) secretomes. The proteins were classified according to their level of secretion in the wild-type strain (from the strongest to the lowest log-normalized PAI values). Four biological replicates were analyzed for each strain. Each dot represents a replicate; dashes represent the mean values of all four replicates. Asterisks highlight proteins presenting significantly different secretion levels between both strains (glht post-hoc test, p value Ͻ 0.05). a good control for the whole process, from the protein extraction protocol in the supernatants to the statistical analysis of the MS detection.
RipBJ, a New R. solanacearum Type 3 Effector-The global comparison of the GMI1000 wild-type secretome with the hrcV mutant secretome by LC MS/MS allowed us to highlight proteins specifically secreted by the T3SS (Fig. 2B). We noticed that the RSp0213 hypothetical protein clustered within the T3-proteins in the PCA (Fig. 2B) and RSp0213 appeared to be differentially secreted between both strains, with a p value ϭ 1.50.10 Ϫ5 (Fig. 2C; supplemental Table S2). In order to test the hypothesis that RSp0213 encodes a previously overlooked Rip effector protein, we performed in vitro secretion assay using GMI1000 and hrcV mutant strains expressing an RSp0213-HA fusion construct. Immunoblotting using an anti-HA antibody revealed that RSp0213-HA fusion proteins were present in both strains, but the RSp0213-HA secretion was only detected from the wild-type strain (Fig. 3A). To definitely show that RSp0213 is a bona fide Rip T3E, we performed a translocation assay based on the adenylate cyclase reporter system (40). GMI1000 and hrcV mutant strains carrying RSp0213-CyaA' fusion constructs were generated, and infiltrated in tobacco leaves. Seven hours after infiltration, tobacco leaves were harvested and cAMP levels were measured using cAMP Biotrak competitive enzyme immunoassay (37). High amounts of cAMP were detected for the wild-type strain but not for the hrcV T3-defective mutant, indicating that RSp0213 was translocated into plant cells through the T3SS (Fig. 3B). These experiments showed that RSp0213 encodes a T3E, which was therefore renamed RipBJ following the proposed nomenclature for R. solanacearum T3Es (11). A ripBJ disruption mutant was generated and used to evaluate the contribution of this effector to pathogenicity on A. thaliana and M. truncatula, and to hypersensitive response (HR) elicitation after infiltration of tobacco leaves. No differences could be observed between the ripBJ mutant and the wild-type strain for none of these bioassays (supplemental Fig. S5).

Insights into Type 3 Secretion Regulation from the Secretome Analysis of hrp-Associated Mutants-
The comparative secretome analysis between the wild-type strain and a hrp deficient mutant showed that our analysis enables the fine characterization of T3-secretome, but also allows for the comparative quantification of detected proteins between strains. The sensitivity of the method incited us to explore and compare the secretomes of potential T3S regulators. Some of them were characterized in phytopathogenic bacteria for their participation to T3E delivery, by promoting or repressing secretion through protein-protein interactions (21,23). In R. solanacearum strain GMI1000, four potential Hpa proteins were identified: HpaB (RSp0853), HpaD (RSp0848), HpaG (RSp0842), and HpaP (RSp0862). Hpa genes are hrpB-regulated (41), physically associated to the hrp-cluster and not secreted. HpaB and HpaD have features of molecular chaperones (20, 42) with small sizes (16.8 kDa and 13.9 kDa, respectively) and acidic pI (4.4 and 4.2, respectively). HpaP was recently shown to modulate T3S (23) and HpaG (also known as LrpE) is a leucine-rich repeats (LRR) containing protein that negatively regulates the production of Hrp pili (43). We conducted secretome analyses of the hpaB, hpaD, and hpaG mutants using a similar experimental procedure as described above (supplemental Fig. S2C). We detected 260, 279, and 262 proteins in the secretomes of the hpaB, hpaD, and hpaG mutants, respectively (supplemental Table S3; supplemental Fig. S3). The first two components of the PCA conducted on the secreted proteins detected in the five strains (wild-type, hrcV and hpa mutants) clearly identified the strain effect, discriminating all strains and particularly singling out the hrcV mutant (Fig. 4A). The proteins responsible for this sample distribution are highlighted on Fig. 4B. Group 1 contains the RipF1 translocon proteins (44) and RipAB (formerly named PopB) (18). Group 2 consists of the HrpY pilin and the harpins RipX and RipW (formerly PopA and PopW) (45,46). Group 3 contains RipC1, RipD, RipG7, RipM, and RipAC, which are among the most abundant T3Es in the wild-type strain supernatant (see supplemental Table S1). Group 4 contains proteins with a lower secretion level in the wild-type strain (supplemental Table S1) but with a high discriminating value upon the glht post-hoc test (supplemental Table S2), whereas group 5 corresponds to the bulk of the remaining proteins. PCA performed on the five strain secretomes, excluding T3-associated proteins, shows no highly discriminating component (supplemental Fig. S6).
To more accurately evaluate the role of the hpa genes in T3S, we compared the detected T3-proteins in the wild-type strain versus each hpa mutant using a glht post-hoc test ( Fig.  5; supplemental Table S3). The hpaB mutant shows three classes of secretion patterns: (1) a class of proteins with a high level of secretion, but mainly not differentially secreted as compared with GMI1000, which corresponds to HrpY, RipF1, and RipAB proteins, along with the RipW and RipX harpins; (2) a class of 25 proteins with a reduced level of secretion in the hpaB mutant; and (3) a class of five effectors with a low level of detection, which are not detected in the hpaB mutant (Fig.  5A). The hpaB mutant appears therefore strongly altered in its secretion of effectors but not of the components of the T3SS machinery. The hpaD mutant shows a secretion pattern close to the wild-type strain, with only four Rip significantly differentially detected (Fig. 5B). In contrast, the hpaG mutant secretome reveals a global over-secretion pattern. Indeed, even if few Rips are differently secreted in the mutant compared with the wild-type (RipAB, RipF1_1, RipF1_2, and RipR), we can observe that numerous Rips are more secreted by the hpaG mutant than the wild-type (Fig. 5C). Six Rips were detected only in the hpaG mutant (RipV1, RipAK, RipAS, RipA4, RipAR, and RipAZ1).
Three Main Classes of Type 3 Effectors Based on Secretion Pattern Specificities-We used the log-normalized PAI data to perform a double hierarchical clustering in order to classify both the secreted Rips and the bacterial samples, according to the secretion of those Rip proteins. This analysis revealed two main groups of strains: hrcV and hpaB mutants on one hand, and hpaD, hpaG mutants, and GMI1000 on the other hand (Fig. 6). The separation of these two groups is caused by the differences in the quantification of 17 proteins from group "a" (including the newly identified RipBJ (RSp0213). These proteins are more abundant in the GMI1000/hpaD/hpaG secretomes than in the hpaB/hrcV ones. Likewise, the "b" group differentiates the hrcV from the other strains on the study, because of the absence of the Hrp pilus component and of the harpins in the hrcV mutant, thus revealing that regulation of RipAB and RipAM secretion is tightly linked at the post translational level. The "c" group corresponds to the six Rips only detected in the hpaG mutant secretome. The "d" subgroup clustered the 15 remaining Rips, which are also quantified in higher amount in the hpaG mutant.
In order to validate these specific secretion patterns obtained by MS, secretion assays were performed on several effectors showing (according to Fig. 2C) either high levels of secretion (RipAB, RipW, RipD, RipX, and RipG7), intermediate levels of secretion (RipH3, RipP1, RipAN, and RipP2), weak levels of secretion (RipAA, RipAV, and RipH2), or not detected by MS (RipH1). The abundance of these selected proteins was determined by immunoblotting protein extracts of cell pellets and of culture supernatants produced in the same conditions as the MS samples. In pellets, all the Rips were produced by the different strains in equal amount, except for hpaG mutant, as a slight overproduction can be observed for some Rips (Fig. 7). In supernatants, a different pattern of secretion can be observed: RipW, RipX, and RipAB were secreted in equal quantity by the three hpa mutants when compared with the wild-type strain (Fig. 7A). Combining MS data and Western blot experiments confirmed that the harpins RipW and RipX, as well as RipAB, share a similar secretion pattern. These Rips are thus secreted (in high amounts) independently of the three Hpa proteins. In Fig. 7B, we can observe that RipG7 is equally secreted in the hpaD and hpaG mutants compared with the wild-type, and less secreted in the hpaB mutant. This reflects the partial involvement of HpaB for strongly secreted Rips. Fig. 7C and 7D show the proteins (RipH3, RipP1, RipP2, RipAA, RipD, RipH2, RipAN, and RipAV) that are not secreted by hrcV nor hpaB mutants. This data confirms the secretion patterns observed by MS and highlights the specific requirement of hpaB for their secretion. Additionally, we observe an increased secretion for all these proteins in the hpaG mutant, which also validate the original over-secreting MS pattern of this mutant (Fig. 5C). We can notice that this over-secretion can be independent (Fig. 7C) or not (Fig. 7D) of an over-production of these proteins in the hpaG mutant. This reflects the involvement of HpaG in the FIG. 5. Log-normalized PAI values of T3-associated proteins detected in the secretomes of the wild-type (red) and of hpa mutants show different T3S patterns. A, hpaB mutant in cyan. B, hpaD mutant in black. C, hpaG mutant in blue. Four biological replicates were analyzed for each bacterial strain. Each dot represents a replicate; dashes represent the mean value for all four replicates. The proteins were classified according to their level of secretion in the wild-type strain (from the strongest to the lowest log-normalized PAI detection). Asterisks highlight proteins presenting significantly different secretion levels between strains (glht post-hoc test, p value Ͻ 0.05). control of T3E synthesis, but also potentially in transcriptional regulations or in protein stabilization. Finally, Fig. 7E shows that Western blot detection is more sensitive that MS detection as we can observe an hpaB-dependent pattern of secretion for RipH1, not detected in our MS analysis. However, the immunoblot analyses confirmed the specific patterns observed with MS analysis.
The Abundances of the Detected Effectors in the Bacterial Supernatants and That of Their Corresponding Gene Transcripts are Positively Correlated-In order to identify critical parameters determining the MS detection threshold, we investigated whether MS detection of the 44 secreted Rip effectors could be correlated either with structural features such as the isoelectric point (pI) or molecular weight (MW), or with transcription levels (as deduced from RNA sequencing approaches) (supplemental Fig. S7). The pI values of the detected Rips ranges from 4.2 (RipX) to 11.8 (RipAO), displaying a homogenous distribution between all the ranges of pI (supplemental Fig. S7A and S7B). The MW of the Rips ranges from 10 kDa (RipAG and RipAH) to 278 kDa (RipS4), with half of the Rips between 20 kDa to 80 kDa. The distributions observed in supplemental S7C and S7D could not clearly attribute a lower or higher detection rate according to the MW, since this detection ranged from RipAH (10 kDa) to RipR (186 kDa). However, the RipA and RipS protein families (formerly named AWR and SKWP families, respectively) having the highest molecular weight among R. solanacearum T3Es were less efficiently detected. Indeed, among the five paralogs of the RipA family (ranging from 113 kDa to 141 kDa), only RipA2 is detected and none of the six RipS paralogs (Ͼ250 kDa, except for RipS6) could be detected ( Fig. 2; see supplemental Table S4 for exhaustive pI and MW Rip data). We also looked for a possible correlation between RNA abundance (RNA-Seq data) and MS detection. This was done by taking advantage FIG. 6. Double hierarchical clustering of bacterial strains and T3-proteins. Bacterial strains thus produced two main clusters (arrows). Different protein subgroups according to their secretion patterns could be identified (a to d). White color represents a Z-score of 0 whereas red and blue shades represent higher and lower log-normalized PAI values, respectively. of RNA-Seq experiments performed in the same inducing medium (Guidot et al., in preparation). We observed a positive correlation between the Rip RNA abundance and the cognate protein detection in the supernatant (supplemental Fig. S7E and S7F). We fitted a linear model to explain the log-normalized PAI values by the squared rpkm values from the RNA-Seq dataset. We obtained the following formula log(normalized PAI) ϭ 0.1509 ϩ 0.0093*rpkm 2 where the p values for intercept (0.1509) and slope (0.0093) are 0.07 and 6.10 Ϫ4 , respectively (supplemental Fig. S7G). Therefore, except for the RipA and RipS family which correspond to very large proteins, no apparent differences could be attributed between detected and not-detected Rips in term of pI or MW, the threshold of Rip detection being very probably dependent on gene transcription, and thus to the amount of intracellular proteins produced.
Altered Secretion of Rip Proteins in the hpaB and hpaG Mutants is Associated with Impaired Virulence on M. truncatula-We then wanted to evaluate the impact of these modulations of Rip secretion on R. solanacearum pathogenicity on several host plants. We first validated that the hpaB, hpaD, and hpaG mutants were not affected in their growth in complete or minimal medium (data not shown). The three hpa mutant strains were leaf-infiltrated on N. tabacum, and rootinoculated on A. thaliana and M. truncatula. The hpaD mutant triggered an HR similar to the wild-type strain on tobacco plants, and virulence on A. thaliana and M. truncatula plants were undistinguishable from the wild-type strain (supplemental Fig. S8). The hpaB mutant was strongly affected in path-ogenicity (Fig. 8). Indeed, the hpaB mutant is no more able to trigger an HR on tobacco, a phenotype complemented by the wild-type hpaB allele (Fig. 8A). The hpaB mutant was also unable to induce disease on A. thaliana (Figs. 8B and 9C), or on M. truncatula (Fig. 8D). Finally, HpaG was found to be dispensable for HR induction on tobacco (supplemental Fig.  S8A) as well as for R. solanacearum pathogenicity on A. thaliana (Fig. 9A). However, disease development was delayed and significantly reduced on M. truncatula compared with the wild-type strain ( Fig. 9B and 9C). M. truncatula in planta bacterial growth showed that HpaG is also required for bacterial multiplication, with a decrease of more than 100-fold for the hpaG mutant compared with the wild-type strain. Bacterial multiplication was restored in the hpaG-complemented strain (Fig. 9C). DISCUSSION Phytopathogenic bacteria have evolved multiple strategies to infect plants. A major determinant of bacterial pathogenicity is the T3SS that delivers effector proteins directly into the host cells. Despite increasing knowledge on the control of the T3SS through transcriptional regulatory networks, only limited information on post-translational regulations is available. This study provides the first characterization of the T3-dependent secretome of R. solanacearum using the recent and advanced Q Exactive technology. To date, some T3-secretomes of animal pathogenic bacteria have been characterized (47)(48)(49), but only partial secretome and proteome analyses are available from plant pathogenic bacteria since these studies FIG. 7. Western blot validations of Rip secretion patterns. Secretion assays were performed in T3-inducing conditions, for the GMI1000 wild-type strain, and the hrcV, hpaB, hpaD, and hpaG mutants. Total proteins from bacterial pellets and supernatants were analyzed by Western-blot. Two to three biological replicates were done for each set of five strains. mainly used two-dimensional gel electrophoresis followed by MALDI-TOF-MS or nano-LC-MS/MS approaches (50 -55). With a total of 228 proteins detected in the wild-type strain supernatant (supplemental Table S1), this study therefore represents the most exhaustive secretome inventory to date among bacterial plant pathogens. In addition, the analysis of hpa mutants revealed specific T3-secretome patterns, showing both the sensitivity of the methodology and its efficiency to decrypt T3-secretomes and to describe the impact of T3post-translational regulators in the secretion process. Combining the data obtained for the four biological replicates of the five strains analyzed in this study, we could detect 320 secreted proteins. Among them, 66% were predicted to be secreted using SecretomeP 2.0 (http://www.cbs.dtu.dk/ services/SecretomeP/), SignalP 4.0 (http://www.cbs.dtu.dk/ services/SignalP-4.0/), and MaGe (Magnifying Genomes Microbial Genome Annotation System) (http://www.genoscope. cns.fr/agc/microscope/mage/index.php). We also showed with Western blot experiments that the non-detection of some Rips using MS is not because of a non-secretion in our experimental conditions, but is because of the MS detection threshold. Detection of 60% of the R. solanacearum GMI1000 known effectors (43 out of 72 (11)) is satisfactory compared with other T3-secretome studies (51)(52)(53)(54)(55) and considering that several Rip effectors may certainly be low-abundant proteins produced at the host cell contact. We observed a direct correlation between rip gene transcription and the quantitative detection of the corresponding proteins through MS analysis. Quantitative MS data were also supported by Western blot experiments. Absence or residual spectral counting for Rips in the hrcV mutant was confirmed by the absence of detection of Rips in bacterial supernatants after immunoblot experiments whereas these proteins were well produced in the hrcV bacterial pellets. Contrary to all the other Rips tested, RipW was detected as slightly secreted by the hrcV mutant as shown by MS (Fig. 2C) or Western blot experiments (Fig. 7A). The contrasting secretion patterns of the hpaB and hpaG mutants, with specific proteins either less or more secreted than the wild-type, were also validated by specific immunodetections. All these observations confirm the overall quality and sensitivity (reproducibility of detection of differentially secreted proteins) of the MS experimental process we used.
Our study identified a novel effector, RipBJ, which contains an N-terminal sequence homologous to RipAG and RipAH, but also to AvrRpm1 and AvrPto P. syringae effectors (supplemental Fig. S5D). This N-terminal sequence domain contains putative myristoylation/acylation signals which were shown to be required for the targeting of AvrRpm1 to the plasma membrane (56). However, no difference could be observed between the wild-type strain and the ripBJ mutant in pathogenicity assays, probably because of functional redundancy among Rip effectors (15)(16)(17). In a former study (23), we provided evidence of a post-translational control of T3E secretion in R. solanacearum by the HpaP protein. In this work, we combined proteomic and phenotypic analyses studying three additional hpa mutants, hpaB, hpaD, and hpaG. Our work showed that these hpa mutants have three distinct secretion patterns, with specific proteomic signatures highlighting Rip classes of T3S.
hpaB mutant secretome analysis showed the strongest Rip secretory phenotype. Indeed, secretion of almost all the Rips detected in the wild-type strain supernatant was partly or fully altered in the hpaB mutant. This global defect in Rip secretion is likely to be the cause of the nonpathogenicity phenotype of the hpaB mutant. This corroborates previous works on hpaB homologs in several phytopathogenic bacteria. In another strain of R. solanacearum, HpaB was shown to be required for the translocation of at least 66 T3Es (57,58). Studies in Xanthomonas campestris pv. vesicatoria (Xcv) showed that HpaB had a central role in Xcv efficient T3S, notably interacting with various control proteins and T3Es (6,22). HpaB homologs were shown to be required in bacterial pathogenicity but with different specificities according to the pathosys-tems studied (57,59,60). In our study, we show that the hpaB mutant is still capable of secreting HrpY, furthermore there is also efficient secretion of RipX and RipW harpins (45,46), altogether suggesting that the Hrp pilus is functional in this mutant. Interestingly, RipAB appears to be the most abundantly secreted Rip, and is still efficiently secreted in the hpaB mutant. Additionally, the hierarchical clustering obtained taking into account the five strains clustered the HrpY pilin, the RipF1_1 and RipF1_2 translocators, the RipX and RipW harpins with RipAB and RipAM. The RipF1 paralogs are required for efficient Rip translocation (11), but a direct interaction between RipF1_1 and HrpY could not be identified (44). Even if RipAM has a lower secretion level compared with the other proteins of this group, similar pattern is observed. All together, these data led us to hypothesize that RipX, RipW, RipAB, and RipAM could be involved, in an Hpa protein-independent manner, in the first stages of Rip translocation process. This hypothesis is in accordance with a polymutant approach in Pst DC3000 which highlighted a consortium of redundant translocators required to promote Pst T3E injection into plant cells (61,62). RipF1, RipW, RipX, RipAB, and RipAM all belong to the R. solanacearum core effectome (11). It will be of particular interest to study more precisely how they take part in the formation of the Hrp translocation apparatus as acces- The hpaG mutant secretome analysis showed a deregulation of the hrp-secretion with an unexpected MS-over-secretion pattern, numerous Rips being more detected compared with the wild-type strain. Although still able to induce an HR on tobacco plants and the full disease on A. thaliana, the hpaG mutant is specifically impaired in its pathogenicity on M. truncatula. In addition, RipG7, which is strictly required for R. solanacearum pathogenicity on M. truncatula (17), is still efficiently secreted by the hpaG mutant. A study with another strain of R. solanacearum on a HpaG homolog protein (LrpE) showed a slight virulence reduction of the lrpE mutant on eggplant (43). Thus, we hypothesize that these host phenotypes depend more on altered secretion of some effectors directly involved on certain hosts to confer disease. One or several Rips could be directly involved in this phenotype on the model legume plant, letting us hypothesize a potential HpaG function as a secretion suppressor of Rips involved in the induction of plant immunity. We also observed that oversecretion of Rips by the hpaG mutant can also be associated with specific over-production of Rips, suggesting a role for HpaG in transcriptional control or on protein stabilization for a subset of Rips.
This article constitutes another piece of the puzzle of T3S post-translational control, and even if many questions remain, this first R. solanacearum secretome study supports the existence of a T3S hierarchy with specific subsets of effectors having differential secretion patterns (Fig. 10). The study of three hpa mutants revealed previously unknown complex regulations of T3S, and Rips with similar regulation profiles. This work also showed that HpaB and HpaG proteins are posttranslational regulators of T3S, with opposite roles on T3S. Indeed, although there is no evidence for a role of HpaD in T3S post-translational control, HpaB clearly promotes secre-tion of numerous Rips and acts as a positive regulator of T3S, whereas HpaG seems to negatively regulate secretion of other Rips. A recent review on orchestration of T3S from animal and plant pathogenic bacteria classified the pilus as an "early" T3S substrate, translocon proteins as "intermediate" substrates and effector proteins as "late" substrates (6). Here, we identified a first set of Rips whose secretion is Hpaindependent which we propose to qualify as putative "early/ intermediate" substrates as they include secreted syringe components HrpY and RipF1 proteins. We highlighted a second set of Rips whose secretion is Hpa-dependent, which we hypothesize as putative "late" substrates, with either a strict, or a quantitative Hpa-secretion control. This also suggests a sequential secretion in R. solanacearum, between "early/intermediate" and "late" substrates, helping to define new subsets of Rips under secretion process (Fig. 10). This work illustrates how the Rip secretion is finely tuned and how this control can impact host specificity. We showed that we were able to monitor secretion of more than 40 Rips per assay. A polymutant approach could help to identify T3Es collectively involved on different host plants. This illustration proposes subsets of Rips for which T3S is coregulated, in an Hpa-independent or Hpa-dependent way. The model is based on the T3-specific secretion patterns of the GMI1000 wild-type strain and of the hpa mutants. Data obtained from the statistical analysis on MS assays, from the following hierarchical clustering, and from the Western blot validation experiments were combined. The two groups represent "early/intermediate" and "late" substrates of T3SS, respectively. Among the late substrates, the blue circle contains the Rips whose secretion is positively regulated by HpaB, whereas the red circle contains the Rips whose secretion is negatively regulated by HpaG.