The R. solanacearum isolate UY031, phylotype IIB, sequevar 1, originally isolated from potato (Siri et al., 2011), carrying the LUX-operon under the psbA promoter (Monteiro et al., 2012) was used for all experiments. Bacteria were routinely grown in rich B medium as described (Monteiro et al., 2012).

S. commersonnii accessions F97 (susceptible to bacterial wilt) and F118 (moderately resistant) obtained from a segregating population were used in this work and propagated in vitro as described (Zuluaga et al., 2015).

As a control condition, bacteria were grown for 2 days on rich solid medium without tetrazolium chloride or antibiotics at the appropriate dilution to obtain separate colonies. Bacteria were recovered from plates and mixed with 5% of an ice-cold transcription stop solution [5% (vol/vol) water-saturated phenol in ethanol]. Cells were centrifuged at 4°C for 2 min at maximum speed and the bacterial pellet was immediately frozen in liquid nitrogen.

For plant RNA samples, S. commersonii F97 and F118 roots were inoculated as described in Zuluaga et al. (2015). Briefly, plant roots from 2-week old plants grown in soil were injured with a 1 ml pipette tip and inoculated by soil drenching with a bacterial solution at 10⁷ colony forming units (cfu)/ml. Control plants were mock-inoculated with water. After inoculation, plants were kept in a growth chamber at 28°C in long-day conditions. Luminescence quantification was used to select plants with comparable infection levels in the susceptible and the resistant accessions, corresponding to approximately 10⁵ colony forming units per g of tissue (Cruz et al., 2014).

Total RNA from bacterial cultures was extracted using the SV Total RNA Isolation System kit (Promega) following the manufacturer's instructions for Gram-negative Bacteria. Infected plant RNA extractions were carried out as described (Cruz et al., 2014). RNA concentration and quality was measured using the Agilent 2100 Bioanalyzer. For rRNA depletion, 2.5 μg of RNA were treated with the Ribo-zero^(™) magnetic kit for bacteria (Epicenter). Three biological replicates per condition were subjected to sequencing on an Illumina-Solexa Genome Analyzer II apparatus in the Shanghai PSC Genomics facility using multiplexing and kits specially adapted to obtain 100 bp paired-end reads in stranded libraries. Raw sequencing data is available in the Sequence Read Archive under the accession code SRP096020.

FASTQC was used to evaluate the quality of the RNA-seq raw data. R. solanacearum reads were identified from total infected root sequences using Bowtie2 (version 2.2.6; Langmead and Salzberg, 2012) as described in the results section. The completely sequenced genome of strain UY031 (Guarischi-Sousa et al., 2016) was used as reference. For identification of R. solanacearum reads, the Burrows-Wheeler Alignment (BWA) tool was initially used. However, a high number of reads from mock-inoculated control samples mapped to the bacterial genome (Table 1). Visual evaluation of these mapped reads using the Integrative Genomics Viewer (IGV) tool (Robinson et al., 2011; Thorvaldsdottir et al., 2013) showed that most contained mismatches to the R. solanacearum genome sequence, indicating that they likely belonged to contaminating bacteria. BWA was thus assayed with more stringent parameters (-B 20-O 30-E 5-U 85), to increase penalties for mismatches, gap openings, gap extension, and unpaired read pairs, resulting in a reduction of only half of the reads mapping to the genome. Finally, Bowtie2 was assayed, once more using stringent parameters to penalize mismatches and gaps (–mp 30–rdg 25,15–rfg 25,15). In this case, mapped reads levels in mock-inoculated plants could be considered background compared to the high read numbers from inoculated samples, thus, Bowtie2 was finally used in all samples analyzed, including RNA-seq reads coming from in vitro grown bacteria (Table 1). Alignments were summarized by genes on counting tables using HTSeq-count (version 0.6.1 p1; Anders et al., 2015) and NCBI's reference annotation (genome features were extracted from NCBI's RefSeq sequences NZ_CP012687.1 and NZ_CP012688.1); alignments with quality lower than 10 were discarded. Differential expression (DE) analysis was carried out with the DESeq2 (version 1.12.3; Love et al., 2014) package in R (version 3.3.2). Benjamini–Hochberg procedure was used for multiple testing corrections. Genes with log₂(fold-change) > 0.5 and q < 0.01 were considered as differentially expressed. We used these thresholds to select for relevant and robust differentially expressed genes. Final annotation of the genome was defined based on the NCBI gene locus and the gene name and description of the reference R. solanacearum GMI1000 genome annotation (Supplementary Table 1).

get_homologs (version 2.0; Contreras-Moreira and Vinuesa, 2013) was used for searching R. solanacearum UY031 homologous genes on R. solanacearum GMI1000, R. solanacearum IPO1609 and R. solanacearum UW551 strains as well as in Pseudomonas syringae pv. syringae B728a; NCBI RefSeq sequences GCF_001299555.1, GCF_000009125.1, GCF_001050995.1, GCF_000167955.1, and GCF_000012245.1, respectively. Default algorithm of bidirectional best-hits was used on homologous genes search.

R. solanacearum UY031's genes were functionally categorized using two different strategies. Firstly, functional categories from Pseudomonas syringae pv. syringae B728a as defined by Yu et al. (2013), were translated to R. solanacearum UY031 based on homology information between the two strains. Although the P. syringae-derived categories should be more specific and accurate for another bacterial plant pathogen, almost 70% of the R. solanacearum UY031 genes could not be classified using this method. Therefore, a second strategy based on Clusters of Orthologous Groups (COG) categories was applied. Genome features were extracted from NCBI's RefSeq annotation and cdd2cog.pl script (version 0.1; Leimbach, 2016) was used to assign COG IDs and functional categories to the differentially expressed genes (Supplementary Table 1).

cDNA libraries from rRNA-depleted RNAs isolated from S. commersonii roots inoculated with R. solanacearum were sequenced using Illumina technology as previously reported (Zuluaga et al., 2015). To generate the transcriptomic profile of the bacteria growing inside root tissues, R. solanacearum UY031 sequences were obtained following the pipeline detailed in Figure 1. First, reads from mock-inoculated plants were used as a control to determine the best alignment tool to map against the R. solanacearum UY031 reference genome (Guarischi-Sousa et al., 2016; see material and methods). The Bowtie2 alignment tool with stringent parameters was used, as it retained a number of R. solanacearum reads in mock-inoculated plants that could be considered background levels compared to the high read numbers from inoculated samples (Table 1). All samples were analyzed with Bowtie2, including RNA-seq reads coming from in vitro grown bacteria. We determined that around 1% of the total sequenced reads from plant tissues corresponded to R. solanacearum and these were retained for further analyses. S. commersonnii sequences accounted on average for 63.15% of the total reads sequenced and the remaining reads corresponded mostly to contamination by other bacterial endophytes. The retrieved bacterial sequences were quantified and differentially expressed (DE) genes comparing the different conditions were determined. Total RNAs from infected S. commersonii enabled transcript quantification for over 96% of R. solanacearum UY031 predicted genes (4,609 out of the 4,778; Guarischi-Sousa et al., 2016).

In order to compare the R. solanacearum gene expression patterns during infection of resistant and susceptible wild potato plants, we analyzed separately the bacterial reads obtained from infected S. commersonii accessions F118 and F97, respectively. Surprisingly, only two out of the 4,609 genes for which expression was detected showed differential expression between the two genotypes. The differentially-expressed (DE) genes, RSUY_RS08455, and RSUY_RS16950, were both up-regulated in bacteria grown inside the resistant accession (Table 2). The first gene corresponds to an uncharacterized member of the MarR transcriptional regulator family, while the second encodes a hypothetical protein.

Since R. solanacearum showed extremely similar (>99.9%) transcriptional behavior during interaction with both S. commersonii accessions, bacterial reads from both accessions were treated as biological replicates in the rest of this study.

The R. solanacearum in planta gene expression dataset was compared to a reference condition consisting of bacteria grown on solid rich B medium. Bacteria grown on solid medium were used as the reference condition instead of liquid cultures. R. solanacearum colonies grown on solid media better mimic the biofilms and microcolonies formed by R. solanacearum during early infection, when most bacteria occupy plant intercellular spaces (Mori et al., 2016). A total of 422 genes were differentially expressed during pre-symptomatic infection (231 up-regulated and 191 down-regulated), compared to growth on rich medium (Supplementary Table 2). These DE genes were classified into the functional categories previously used for gene expression studies in the plant pathogenic bacterium P. syringae (Yu et al., 2013; Supplementary Table 3). The number of successfully classified genes in each category was quantified in differentially induced or repressed groups and in the whole genome as a reference (Figure 2). This analysis revealed four categories highly over-represented in the up-regulated genes and under-represented in down-regulated genes: stress, secretion, chemosensing, and motility and phage and insertion sequences (IS). These categories represent together approximately 20% of the total induced genes in planta. The opposite trend (under-representation in up-regulated and over-representation in down-regulated genes) is observed in the categories including genes for transport and metabolism of amino acids and carbohydrates. In addition, the categories replication and DNA repair, transport, fatty acid metabolism and cofactor metabolism are strongly under-represented amongst the up-regulated genes in planta (Figure 2).

We used the P. syringae categories because they were created to describe the genes of a bacterial plant pathogen and are thus very informative for this study. However, the same analysis was carried out using the widely used but more general COG categories, and the results confirmed the previously-described tendencies (Supplementary Figure 1). Genes involved in carbohydrate, amino acid, lipid, cofactor, and secondary metabolism were over-represented among those down-regulated in planta. A clear enrichment of replication, cell motility and recombination and repair (where IS elements are included) was observed in the up-regulated genes. Interestingly, a clear asymmetry was seen for unclassified genes in this case, for they represent 40% of the up-regulated but only 7% of the down-regulated genes.

Closer scrutiny of the up-regulated genes in the plant revealed that the category secretion included 11 genes encoding the T3SS and its associated effectors and four chemosensing and motility genes, coding for pilus assembly and flagellum transcriptional activators (Table 3).

Taken together, these results show a major induction of stress-related activities and an inhibition of the central metabolism when the bacterium grows in planta compared to synthetic media.

Among the 422 genes DE during S. commersonii root colonization, 34% (80 induced and 65 repressed genes) had been identified in previous studies analyzing gene expression of R. solanacearum cells recovered from infected plant stems (see references below). Notably, 73 genes were also DE in microarray analyses of R. solanacearum UW551 -a phylotype IIB strain highly similar to UY031- isolated from tomato (Jacobs et al., 2012). Also, 42 genes have been shown to be induced in a temperature-dependent manner when bacteria grew in tomato xylem or rhizosphere (Bocsanczy et al., 2014; Meng et al., 2015). In addition, 31 DE genes (most of them induced in planta) are part of either the HrpB or HrpG regulons, which are known to trigger expression of the T3SS and other virulence genes in response to direct plant cell contact (Valls et al., 2006).

Amongst the R. solanacearum genes induced during plant colonization, 31 encode already reported virulence traits (Table 3). As expected, genes encoding the T3SS (hrpY, hrpX, hrpK, hrcT) and some of its related effectors (ripV2, popC, ripD, popF1, awr5_1, popB, and popA) were induced inside the plant (Boucher et al., 1987; Cunnac et al., 2004). Motility and adherence genes were also up-regulated, including type IV pili (pilG, pilH, pilN, pilM, pilY, pilW, and fimV), as well as the transcriptional activators of the flagellum genes flhC and flhD (Kang et al., 2002; Tans-Kersten et al., 2004). Other induced genes encoding described factors that are key for bacterial virulence included hdfA (Delaspre et al., 2007), efe (Valls et al., 2006), metE (Plener et al., 2012), and rpoN1 (Lundgren et al., 2015; Ray et al., 2015; Table 3). Peroxidases, catalases (katE, katG) and alkyl hydroperoxide reductases (ahpC1, ahpF), which have been described to combat the oxidative stress response during plant infection (Rocha and Smith, 1999; Flores-Cruz and Allen, 2009; Ailloud et al., 2016) were also induced. Similarly, the flavohemoprotein hmpX, involved in NO-detoxification (Dalsing and Allen, 2014), was also induced.

In contrast, only 10 reported virulence determinants were down-regulated, including the type III effectors ripQ, ripS2, and ripTPS, the quorum sensing regulator solI (Flavier et al., 1997) and the Type II secretion system genes gspE, gspJ (Table 3).

Thirty-six R. solanacearum genes previously described as related to plant colonization in gene expression studies in other plant species were also induced in potato. Few metabolic genes were induced in planta, being an exception nadB2, involved in the degradation of L-aspartate in the xylem (Brown and Allen, 2004) and the ptsN and narL nitrogen metabolism genes, known to be active during plant colonization (Dalsing and Allen, 2014; Dalsing et al., 2015; Table 3).

Amongst the down-regulated genes, 42 had also been described as specifically down-regulated during plant colonization (Jacobs et al., 2012). Most repressed genes encoded metabolic enzymes and transporters. Examples are the xylose transporters xylF, xylG, and xylH, glycine catabolism genes gcvP, gcvT, and gcvA, the adenilate cyclase coding gene RSUY_RS02845, four siderophore biosynthesis genes and 11 genes involved in amino acid metabolism (Table 3). Also, the stress response gene speE2 and five transcriptional and response regulators were repressed in planta.

Transcriptomic analysis of S. commersoni early root infection revealed highly induced R. solanacearum virulence factors still uncharacterized in this pathogen that may play a role at this stage of the interaction with the host. An example of this is suhB, a global virulence regulator controlling the type III and type VI secretion systems, flagellum biosynthesis, and biofilm formation in the human pathogens Burkholderia cenocepacia and Pseudomonas aeruginosa (Rosales-Reyes et al., 2012; Li et al., 2013). Similarly, a P. aeruginosa orthologue of the in planta induced type IV secretion gene Rhs has been described as a virulence determinant (Kung et al., 2012).

Metabolic traits that might be key at this point of plant infection are the assimilatory sulfate reduction pathway and phosphate mobilization, since cysD, cysN, and cysI (sulfate reduction) and pstB and pstS1 (phosphate mobilization) were induced during S. commersonii root infection. Also, carbonic anhydrase (RSUY_RS16085), which plays a role in disease establishment between potato and Phytophthora infestans (Restrepo et al., 2005), was also found to be up-regulated in the R. solanacearum interaction with wild potato.

The most important category amongst the R. solanacearum genes down-regulated in S. commersonii with so far no assigned functions in plant colonization or virulence was metabolite transporters. Almost half of these corresponded to the ABC-family, including five amino acid transporters. In contrast, the seven major facilitator superfamily (MFS) transporters found in this category are involved with carbohydrate transport. The rest of genes were classified as permeases or RND (Resistance-Nodulation-Division) efflux systems (Table 3). The major metabolic activities identified as repressed in planta for the first time were lipid mobilization and cofactor metabolism, such as the anaerobic cobalamin biosynthesis operon (cbiA, cbiG, and cbiL), and stress-response genes such as rtcA and rtcR, involved in RNA repair (Das and Shuman, 2013).

In sum, our work reflects important gene expression changes between parasitic life and growth in rich medium (see below). This was corroborated by the fact that seven genes annotated as response regulators were also DE, five of them induced (Table 3).

1/3 of the R. solanacearum genes DE during potato infection had been also found DE when the bacterium colonized other plant species and many of these correspond to virulence determinants. For instance, we found that genes encoding the type III secretion system and its associated effectors (popA, popB, popC, popF1, ripD, ripV2, and awr5_1) were induced in potato (Table 3). Except for awr5_1, all these effectors had already been described as up-regulated when the bacterium grew in tomato and in melon (Ailloud et al., 2016), likely indicating that they are part of the minimal gene set required for bacterial virulence. Similarly, the effector ripTPS was down-regulated both in potato (Table 3) and during the interaction with melon (Ailloud et al., 2016). Also sharing similar up-regulation in potato (Table 3) and tomato are the transcriptional activators flhC and flhD (Jacobs et al., 2012), which regulate flagellum-encoding genes (Tans-Kersten et al., 2004) and the nitrogen metabolism genes narL, ptsN, and hmpX (Dalsing and Allen, 2014; Dalsing et al., 2015), implying that they all play a key role during plant infection. Additional genes induced during potato colonization had been described as key for virulence on other plant hosts, including small molecule hdfA (Delaspre et al., 2007), the ethylene forming enzyme efe (Valls et al., 2006), the methionine metabolism gene metE (Plener et al., 2012) and the alternative sigma factor rpoN1 (Lundgren et al., 2015; Ray et al., 2015). These factors may be also considered essential for growth in planta, irrespective of the infected species.

Several transposable elements had been identified in an in vivo screening for genes expressed during R. solanacearum growth in tomato plants (Brown and Allen, 2004), and we found 16 transposases up-regulated in potato (Table 3). This may reflect common stressing conditions in various plant hosts, as stress is known to turn on transcription of transposable elements in various organisms (Capy et al., 2000). Oxidative stress seems also a condition generally encountered by R. solanacearum in plant tissues, as peroxidases, catalases, and peroxiredoxins, required for the bacterium to combat this stress in different plants (Rocha and Smith, 1999; Flores-Cruz and Allen, 2009; Ailloud et al., 2016), were also induced in potato.

Some of the R. solanacearum virulence genes DE in potato showed opposite trends in other host plants. ripQ and ripS2, two of the three type III secreted effectors inhibited in potato were, respectively, upregulated and not DE in melon, tomato and banana (Ailloud et al., 2016). Interestingly, these two downregulated effectors, together with the also repressed stress response gene speE2, are located in a genomic region that is deleted in the avirulent R. solanacearum strain UY043 (Siri et al., 2014), which suggests their involvement in bacterial virulence. Similarly, the effector awr5_1, which was described to trigger hypersensitive response (HR) in tobacco and to inhibit the TOR pathway (Sole et al., 2012; Popa et al., 2016b), showed opposite regulation in potato when compared to tomato and melon (Ailloud et al., 2016), suggesting that it may play host-specific roles. Similarly, genes pilG, pilH, pilN, pilM, pilY, and pilW, coding for structural components of the type IV pili involved in twitching motility and adherence (Liu et al., 2001; Kang et al., 2002) were induced in the current work but repressed in other plant species (Jacobs et al., 2012).

In addition, some virulence determinants well-described as induced during growth in planta were repressed or not DE in potato. Remarkably, the exopolysaccharide synthesis and regulation genes (eps) as well as most known cell wall degrading enzymes (pehA, pehB, pehC, egl, and cbhA), which are virulence determinants (Schell, 2000) induced during tomato infection (UW551 strain) infection (Jacobs et al., 2012) were absent from the potato DE dataset.

Differences in the host environment or in the tissue environment and disease stage are the two most plausible reasons for the discrepancies between virulence gene expression data in potato and in other plant hosts. We favor the latter explanation, as our samples were collected from bacteria growing in the root (including apoplastic and xylematic bacteria) at early times after inoculation while all previous transcriptomic work had been performed from bacteria extracted from xylem at later infection stages.

Three independent observations support the existence of stage-specific environmental cues that differentially affect gene expression in this work compared to previous studies. First, genes that are induced at high bacterial densities are absent from the potato DE genes. Examples are the mentioned exopolysaccharide synthesis genes or the quorum sensing regulator solI, repressed in our conditions but slightly induced in bacteria isolated from the tomato shoot xylem (Jacobs et al., 2012). In the low bacterial cell densities in the roots the phcA cell-density regulator was not induced, impeding solI or eps expression (Huang et al., 1995; Flavier et al., 1997). Second, three out of the six type III effectors that are induced in potato were described as secreted at early stages (Lonjon et al., 2016), two of them (popF1 and popA) also proposed to play an important role in the first steps of infection (Kanda et al., 2003). On the contrary, only two out of the 38 described as “late” effectors (ripD and popC) were induced in our root transcriptome. Third, the afore-mentioned transcriptional regulators flhC and flhD responsible for the activation of the flagellum genes were up-regulated in potato root samples (Table 3) and also in the tomato xylem (Jacobs et al., 2012), but only in the latter were the flagellum structural genes induced, suggesting that the potato transcriptome represents an earlier stage where complete activation of this regulon has not yet occurred. These observations imply that our transcriptome represents a snapshot of a precise stage of the genetic programs deployed consecutively during plant colonization.

Finally, we cannot rule out that changes in R. solanacearum DE genes in different studies are due to the use of different strains. Differing transcriptomes of two R. solanacearum strains in the same plant environment have already been reported (Ailloud et al., 2016). However, the fact that previous gene expression studies were performed with strain UW551, which is genetically extremely close to UY031 used here, render this explanation unlikely. Standardization of the plant inoculation and sampling procedures and a systematic analysis of plant-pathogen interactions dissecting gene expression over time in a defined strain-host pathosystem would clarify the nature of the observed discrepancies between transcriptomic studies.

From the transcriptomic information gathered in this work, we can infer for the first time the environmental conditions encountered by R. solanacearum in the root, the site where plant infection takes place.

A first observation is that the bacterium seems to start to run out of O₂. An indication of this is the highly induced Cbb3-cco, a high affinity cytochrome c oxidase known to contribute to the growth of R. solanacearum and other bacteria in microaerobic or anoxic environments (Colburn-Clifford and Allen, 2010; Hamada et al., 2014), such as the plant xylem (Pegg, 1985). Upregulation of the low O₂ affinity cytochrome ubiquinol oxidase genes cyoA1 and cyoB1 reinforces the notion of a microaerobic rather than an anoxic environment. In agreement with this, nrdB, which is required for growth in aerobiosis (Casado et al., 1991), was up-regulated, and nrdG and nrdD, required in strict anaerobiosis (Garriga et al., 1996; Ailloud et al., 2016) were not induced. Further, the cbiA, cbiL, and cbiG genes, which are involved in anaerobic cobalamin synthesis (Roessner and Scott, 2006), were repressed. Another indication of microaerobic conditions is the induction of genes driving nitrate and sulfate anaerobic respiration. Examples are the cys genes, involved in the assimilatory sulfate reduction pathway (Kredich, 1992), ptsN—a nitrogen-dependent regulatory protein, rpoN1, -the global nitrogen regulator- and narL -the nitrate/nitrite-responsive transcriptional regulator- were all induced in wild potato roots. All these findings suggest that during early root infection R. solanacearum is experiencing the transition from an aerobic environment to the anaerobic conditions established at the onset of disease during xylem colonization (Ailloud et al., 2016).

Another take home message from the root transcriptomes is that few central metabolic pathways seem to be active. It was previously described that a large proportion of the R. solanacearum genes involved in amino acid metabolism and transport was down-regulated during growth in the xylem (Ailloud et al., 2016) and we found that this was also the case during growth in the root tissues at early stages of infection. For instance, the glycine catabolism genes gcvP, gcvT, and gcvA as well as the dipeptide uptake gene dppD1 were repressed in both cases (Table 3; Ailloud et al., 2016). Other R. solanacearum metabolic genes previously known to be repressed in planta also down-regulated here included carbohydrate metabolism genes such as the xylose transporter operon xylFGH and Glucosamine 6-phosphate synthetase, the key enzyme controlling amino sugar biosynthesis (Milewski, 2002; Jacobs et al., 2012). Lipid metabolism was also strikingly repressed during root colonization. Out of the 21 DE genes involved in lipid mobilization, only 2 have been found in previous gene expression studies in R. solanacearum (Table 3; Jacobs et al., 2012). Thus, the downregulation of lipid metabolism could be specific to early infection stages or to wild potato colonization. In this sense, lipid metabolism has been reported to play an important role during plant-host interactions by modulating defense responses in plants and pathogen infection (Casadevall and Pirofski, 2001; Wenk, 2006). Cofactor metabolism was also repressed including the folate synthesis gene pabB (Table 3), already known to be down-regulated in planta (Shinohara et al., 2005), the cobalamin biosynthesis genes and adenilate cyclase. Repression of adenylate cyclase, which is a global metabolic regulator in bacteria (Ullmann and Danchin, 1980), illustrates the magnitude of the metabolic shutdown experienced by R. solanacearum in the roots of S. commersonii.

In contrast with the global metabolic shutdown, aspartate and tryptophan catabolism genes were up-regulated when R. solanacearum grew in the plant roots. The aspartate catabolism gene nadB2 had already been identified as an essential gene for in planta growth in an in vivo screening (Brown and Allen, 2004). Interestingly, aspartate is the second most abundant aminoacid in the tomato apoplast and less so in the xylem (Zuluaga et al., 2013), which is in agreement with the bacterium mostly thriving in the apoplastic root spaces at the early infection times analyzed. Also induced was the Tryptophan 2,3-dioxygenase. Concentrations of this aminoacid are high at lateral root emergence sites (Jaeger et al., 1999), and it was suggested that it is also present in the tomato apoplast (Yu et al., 2013). Induction of tryptophan catabolism would thus be indicative of early plant colonization.

These results likely indicate the existence of a trade-off between the expression of virulence and metabolic genes. This has already been described in a previous study where the quorum-sensing-dependent regulatory protein PhcA regulated a trade-off between production of R. solanacearum exopolysaccharides and bacterial proliferation (Peyraud et al., 2016).

RSUY_RS08455 and RSUY_RS16950 were found to be upregulated in a resistant S. commersonii accession compared to a susceptible one (Table 2), as well as during root colonization compared to rich medium (Table 3). Although these genes also appeared in the microarray transcriptome of bacteria extracted from infected tomato xylem vessels (Jacobs et al., 2012), they have not been characterized.

Similarly, the gene encoding an avrD-like protein was up-regulated in tomato xylem (Jacobs et al., 2012) and in wild potato (Table 3). AvrD is required in P. syringae for the synthesis of syringolide, small molecules that can elicit a hypersensitive response on resistant plants (Keen et al., 1990; Mucyn et al., 2014). In R. solanacearum the avrD-like protein encoding gene is activated by the master virulence regulator HrpG (Valls et al., 2006). Considering the persistence of these three genes among the up-regulated during plant colonization, we suggest that they encode for potential virulence factors, probably necessary independently of the host or the infection stage.

Three genes found up-regulated in S. commersonii (suhB, rhs and the carbonic anhydrase gene RSUY_RS16085, Table 3) have been involved in bacterial virulence on animals and constitute putative virulence genes in R. solanacearum. Although classified as a phosphate mobilization gene (Table 3), suhB is a super-regulator involved in the proper rRNA folding (Singh et al., 2016). It plays a role in virulence of animal bacterial pathogens, influencing T3SS, T6SS, flagellum and biofilm regulation and probably acts in opposite ways in different bacteria (Rosales-Reyes et al., 2012; Li et al., 2013). Interestingly, SuhB differential expression was also observed in two R. solanacearum strains (Meng et al., 2015). The function of Rhs (Rearrangement Hot Spot) proteins is ill-defined but they are considered to promote recombination (Lin et al., 1984). Interestingly, a member of the Rhs family was described to be induced during infection and associated with increased bacterial numbers and decreased survival in mice during pneumonia caused by P. aeruginosa (Kung et al., 2012). Finally, carbonic anhydrase catalyzes the inter-conversion between carbon dioxide and bicarbonate but is also required for growth of many animal pathogenic microorganisms (Capasso and Supuran, 2015). In addition, a role in disease establishment between potato and Phytophthora infestans was also reported (Restrepo et al., 2005), suggesting the possible implication of CAs during host colonization. These evidences suggest that suhB, rhs, and RSUY_RS16085 encode putative virulence factors shared between gram-negative bacterial pathogens that infect animals and plants.

The assimilatory sulfate reduction pathway (cysD, cysN, and cysI) and the phosphate mobilization (pstB and pstS1) were also induced during root colonization (Table 3). cysD and cysN, encode an ATP sulfurylase that produces APS, which can be in turn reduced to PAPS to ultimately synthesize cysteine by cysI. A study carried out in a closely related plant pathogenic bacterium, Xanthomonas oryzae pv. Oryzae, was demonstrated that mutation of either raxP or raxQ (homologs of cysD and cysN) impaired production of APS and PAPS and were required for the correct activity of the avirulence protein AvrXa21 (Shen et al., 2002). Further, several studies demonstrated that mutations on the pst system, responsible for phosphate uptake, affected virulence in diverse animal pathogenic bacteria (Rao et al., 2004; Lamarche et al., 2005, 2008). Altogether, these studies suggest that both systems might be regulators of bacterial pathogenicity, which could also be conserved in plant pathogens.

Finally, the rtcA and its regulator rtcR are down-regulated in planta (Table 3). The rtc system is involved in the regulation of the RNA repair system for ribosome homeostasis through the activation of rtcR by different agents and genetic lesions which in turn activates the rtcAB genes (Das and Shuman, 2013). The rtc system was also involved in the functioning of chemotaxis and motility in Escherichia coli (Engl et al., 2016), as mutations in either rtcA or rtcB increased motility. Since rtc acts a repressor of motility, its down-regulation in S. commersonii colonization could influence bacterial motility, a key virulence determinant.

The authors declare that the research was conducted in the absence of any commercial or financial relationships that could be construed as a potential conflict of interest.