A Single Regulator Mediates Strategic Switching between Attachment/Spread and Growth/Virulence in the Plant Pathogen Ralstonia solanacearum

ABSTRACT The PhcA virulence regulator in the vascular wilt pathogen Ralstonia solanacearum responds to cell density via quorum sensing. To understand the timing of traits that enable R. solanacearum to establish itself inside host plants, we created a ΔphcA mutant that is genetically locked in a low-cell-density condition. Comparing levels of gene expression of wild-type R. solanacearum and the ΔphcA mutant during tomato colonization revealed that the PhcA transcriptome includes an impressive 620 genes (>2-fold differentially expressed; false-discovery rate [FDR], ≤0.005). Many core metabolic pathways and nutrient transporters were upregulated in the ΔphcA mutant, which grew faster than the wild-type strain in tomato xylem sap and on dozens of specific metabolites, including 36 found in xylem. This suggests that PhcA helps R. solanacearum to survive in nutrient-poor environmental habitats and to grow rapidly during early pathogenesis. However, after R. solanacearum reaches high cell densities in planta, PhcA mediates a trade-off from maximizing growth to producing costly virulence factors. R. solanacearum infects through roots, and low-cell-density-mode-mimicking ΔphcA cells attached to tomato roots better than the wild-type cells, consistent with their increased expression of several adhesins. Inside xylem vessels, ΔphcA cells formed aberrantly dense mats. Possibly as a result, the mutant could not spread up or down tomato stems as well as the wild type. This suggests that aggregating improves R. solanacearum survival in soil and facilitates infection and that it reduces pathogenic fitness later in disease. Thus, PhcA mediates a second strategic switch between initial pathogen attachment and subsequent dispersal inside the host. PhcA helps R. solanacearum optimally invest resources and correctly sequence multiple steps in the bacterial wilt disease cycle.

M any microbes coordinate their behavior by means of a cell-to-cell signaling system called quorum sensing (QS) (1)(2)(3)(4). Quorum sensing is mediated by small diffusible signal molecules that accumulate as microbial populations increase or persist in confined spaces. If the signal reaches a critical concentration inside cells, it changes gene expression. The resulting density-dependent collective behavior ensures that quantitative traits such as production of bioluminescence, antibiotics, and virulence factors are expressed only when enough microbes are present to have a biological effect (2)(3)(4). Diverse animal and plant pathogens use QS to upregulate virulence factor expression at high cell densities (2,3), but less attention has been given to how bacteria use QS to modulate traits that are important for success at low cell densities. Identifying genes that are upregulated at low cell densities may reveal how microbes adapt and protect themselves under unfavorable conditions.
The soilborne plant pathogen Ralstonia solanacearum occupies diverse ecological niches. It can survive for years in soil or water as a saprophyte but can also live in plant vascular systems as an aggressive pathogen (5,6). The bacterium enters host plant through wounds, root elongation zones, or sites of lateral root emergence and colonizes the root cortex to invade its preferred in planta habitat, the water-transporting xylem vessels. There R. solanacearum rapidly multiplies to high population densities, spreading systemically and producing large amounts of extracellular polysaccharide (EPS), which results in wilting symptoms and, eventually, plant death (7).
R. solanacearum uses a sophisticated sensory and regulatory network to transition between its saprophytic and parasitic modes. At the center of this network is the Phc QS system, composed of the LysR-type transcriptional regulator PhcA and the products of the phcBSR operon, which control the levels of functional PhcA in response to cell density (8)(9)(10)(11)(12)(13).
As noted in the first description of R. solanacearum in 1896, this pathogen frequently becomes nonmucoid and avirulent in culture (14). In addition to lacking EPS, these spontaneous mutants have increased motility and aerotaxis (13,(15)(16)(17). In 1993, Denny and coworkers discovered that such spontaneous "phenotype converted" (PC) strains resulted from mutation of the phcA locus (13). Subsequent studies found that phcA mutants are hypermotile and overproduce the siderophore staphyloferrin B but that they make less EPS and plant cell wall-degrading endoglucanase and cellobiohydrolase than the wild type (12,18,19). The dysregulation of these multiple virulence factors explains why these mutants are nearly avirulent, although they can still colonize plant stems if introduced through a wound (12).
PhcA directly regulates expression of at least two genes, the endoglucanase structural gene egl and the EPS regulator gene xpsR, by binding to a conserved 14-bp PhcA box in their promoters (20). Many more traits are indirectly regulated by PhcA via other regulatory loci such as pehSR, vsrAD, vsrBC, flhDC, and solIR (5,21). The full extent of the PhcA regulon is unknown, but the many differences between the extracellular protein profiles of wild-type and phcA mutant strains of R. solanacearum indicate that numerous additional genes are directly or indirectly controlled by this system (22).
Most traits putatively regulated by PhcA have been experimentally verified only in vitro, with the important exception of EPS. Immunofluorescence microscopy of R. solanacearum-infected tomato plants using anti-EPS antibodies demonstrated that EPS production is dependent on cell density in planta (23,24). Accumulating evidence suggests that in vitro studies are not reliable for understanding R. solanacearum regulation during plant pathogenesis. For example, PhcA regulates two important bacterial wilt virulence factors differently in culture and in planta: swimming motility and the pathogen's type III secretion system (T3SS). Direct microscopic observation revealed that the percentage of motile cells was density dependent in culture, increasing rapidly as bacteria passed the QS threshold concentration of 10 7 CFU/ml (9,25). In contrast, bacteria sampled from xylem vessels were nearly all nonmotile, even when population sizes exceeded 10 9 CFU/ml xylem fluid (26). Similarly, PhcA repressed the T3SS 20-to 60-fold at high cell densities in culture (27). However, analyses performed with reporter genes, quantitative reverse transcriptase PCR (qRT-PCR), and microarray analyses all found that the T3SS is actually highly expressed during plant infections (28)(29)(30). These inconsistencies suggest that the roles of PhcA can be assessed only in the biologically relevant context of a plant host.
To define the full extent of the PhcA QS system and to identify novel virulence factors that are expressed only at low cell densities, we compared the transcriptome of wild-type R. solanacearum to that of a ⌬phcA mutant during colonization of tomato plants. About 12% of the pathogen's genes were differentially expressed in the ⌬phcA mutant, with an overall profile suggesting that at low cell densities R. solanacearum is optimized to obtain nutrients and attach to plant surfaces. Functional validation experiments confirmed that the ⌬phcA mutant grew better than the wild type on many specific nutrients and in infected tomato xylem sap. This low-cell-density-modemimicking mutant attached better to tomato roots but formed aberrantly dense mats inside xylem vessels and spread poorly in tomato stems.

RESULTS
Phenotypes of the low-cell-density-mimicking ⌬phcA mutant. We constructed an R. solanacearum strain GMI1000 mutant with a disrupted phcA coding sequence to study the behavior of R. solanacearum cells that are genetically locked in the low-celldensity mode. As expected, this mutant, here called the ⌬phcA mutant, formed small round nonmucoid colonies surrounded by brown pigment due to lack of EPS and overproduction of melanin (7) (Fig. 1A). Complementation of the ⌬phcA mutant by insertion of a copy of the wild-type phcA locus and its promoter into the selectively neutral att site restored the typical mucoid colony morphology and lack of pigment FIG 1 Phenotypes of the low-cell-density-mode-mimicking ⌬phcA mutant. (A) Colony morphology of R. solanacearum strains on tetrazolium chloride agar plates. Wild-type, phylotype I strain GMI1000; ⌬phcA, deletion mutant of GMI1000; ⌬phcAϩphcA, phcA deletion mutant complemented by inserting one copy of phcA into the chromosome. (B and C) Growth of wild-type R. solanacearum (blue) and of the ΔphcA mutant (red) was monitored over 48 h under aerobic conditions at 28°C in minimal broth supplemented with 0.2% glucose (three biological replicates and four technical replicates, n ϭ 12) (B) or in ex vivo xylem sap harvested from GMI1000-inoculated tomato plants showing early wilt symptoms (four biological replicates, each performed with five technical replicates, n ϭ 20) (C). Each replicate was grown in xylem sap pooled from eight plants. Growth was measured as A 600 levels by the use of a BioTek microplate reader. The ΔphcA mutant grew faster than the wild type both in minimal medium (Mann-Whitney test, P ϭ 0.0081) and in xylem sap (analysis of variance [ANOVA], P Ͻ 0.0001). (D) Population sizes of the wild-type strain and the ⌬phcA mutant in tomato stems following inoculation with 2,000 cells through a cut leaf petiole, as determined by serial dilution plating of ground stem slices. Data represent means of results from nine plants per treatment per time point; bars represent standard deviations.
R. solanacearum PhcA Modulates Attachment and Growth ® production (Fig. 1A). The ⌬phcA strain initially grew faster than the wild type in minimal medium (MM) with glucose (P ϭ 0.0081, Mann-Whitney test) (Fig. 1B). However, this high growth rate was not sustained and the wild-type strain ultimately showed a slightly higher maximum level of absorbance. The mutant also grew faster than the wild type in ex vivo xylem sap from infected tomato plants, where the ⌬phcA mutant reached and sustained a higher growth rate than the wild type (Fig. 1C). Together, these data suggest that R. solanacearum grows faster in low-cell-density mode than in high-cell-density mode. Finally, we compared the growth kinetics of the two strains in tomato stems following direct inoculation through a cut leaf petiole. Shortly after inoculation, the ⌬phcA mutant grew faster than the wild-type strain, but the two strains reached similar population levels 3 days postinoculation (dpi) and after day 4, ⌬phcA growth lagged (Fig. 1D).
The ⌬phcA mutant differentially expressed hundreds of genes during tomato colonization. To comprehensively identify traits controlled by the PhcA QS system, and to better understand how this regulator affects pathogen behavior in its natural host environment, we compared the transcriptomes of wild-type and ⌬phcA strains during colonization of tomato stems. Transcriptome sequencing (RNA-seq) of bacterial RNA harvested from colonized tomato stems 72 h after petiole inoculation, when the population sizes of the two strains were similar, revealed that the PhcA in planta regulon is very large. Overall, 620 genes were differentially expressed (DE) in the ⌬phcA mutant; representing 12.4% of the annotated coding sequences in the GMI1000 genome. Of these, 365 genes were downregulated, and 255 genes were upregulated (Fig. 2). Table S1A in the supplemental material lists all DE genes, defined as those that showed at least a 2-fold change in expression relative to the wild-type strain and that also met the criterion of a stringent false-discovery rate (FDR) of Յ0.005.
The RNA-seq data were robust and reliable. As expected with RNA extracted from ground plant tissue, most sequence reads mapped to the tomato genome while~1.2% of the ⌬phcA reads and~3.2% of the wild-type reads mapped to the R. solanacearum genome (see Fig. S1A in the supplemental material). Results of both multidimensional and scatter plot analyses indicated that the wild-type strain and ⌬phcA mutant gene Colored circles indicate DE genes encoding traits of interest: green, extracellular polysaccharide (EPS); yellow, plant cell wall-degrading enzymes (CWDE); blue, structural proteins and effectors of the type III secretion system (T3SS); pink, uptake or catabolism of carbon, nitrogen, sulfur, or iron sources (C/N/S/Fe); red, type VI secretion system (T6SS); orange, hemagglutinin-like proteins and adhesins. expression profiles differed significantly, while those of the biological replicates grouped together ( Fig. S1B and C). Hierarchical gene clustering clearly identified many genes that were regulated differently under the two biological conditions (Fig. S1D). Quantitative reverse transcriptase PCR (qRT-PCR) validated the relative expression levels of selected genes of interest; they were consistent with the levels determined by RNA-seq (Fig. S2B).
The in planta transcriptomic data generally agreed with the previously described phenotypes of phcA mutants. As expected, DE genes in the low-cell-density-mimicking mutant included those encoding synthesis of known virulence factors: extracellular polysaccharide (epsB and epsE [101-and 112-fold downregulated, respectively]), motility and chemotaxis, some type III secretion system structural proteins and several effectors, and plant cell wall-degrading enzymes (endoglucanase egl [90fold downregulated], cellobiohydroloase cbhA [228-fold downregulated], and exopoly-alpha-D-galacturonosidase pehB [4.7-fold downregulated]). Expression of the endo-polygalaturonase pehA gene, the exo-polygalacturonase pehC gene, and the pectin methylesterase pme gene, which are downregulated in ⌬phcA mutants growing in culture (31), trended down but did not meet our criteria for DE in planta (Table S1A; Fig. S2A). Along with these known virulence factors, PhcA upregulates the R. solanacearum type VI secretion system (T6SS), predicted to translocate factors into either prokaryotic or eukaryotic hosts in a contact-dependent manner (32,33). This is consistent with the discovery that the cryptic R. solanacearum T6SS contributes to virulence (34)(35)(36).
The distinct transcriptome profile of the ⌬phcA mutant suggests that R. solanacearum has a broader catabolic capacity and grows faster in low-cell-density mode. Collectively, the annotated functions of the 255 genes upregulated in the ⌬phcA strain suggested that the pathogen's core metabolism rates are strikingly different when the bacterium is at low versus high cell densities ( Fig. 3; Table S1A). We hypothesized that the low-cell-density-mimicking mutant could grow on sources of carbon, nitrogen, sulfur, and phosphorus that are not available to the pathogen after it reaches high cell density. Around 15% of the upregulated genes in the ⌬phcA mutant encode putative transporters (Table S1B). These 56 proteins are predicted to transport a wide array of nutrients, including diverse carbohydrates, amino acids, polyamines, iron, nitrate, phosphate, and sulfur. This suggests that R. solanacearum avidly scavenges environmental nutrients at low cell densities. Only nine transporter genes were downregulated in the low-cell-density-mimicking mutant.
To test the hypothesis that low-cell-density-mode R. solanacearum cells can access nutrients that are unavailable when the bacterium reaches high cell density, we measured the growth of the wild-type and ⌬phcA strains on Biolog phenotype arrays and in minimal media supplemented with individual carbon or nitrogen sources. The ⌬phcA mutant grew better than the wild-type strain on dozens of sugars, carboxylic acids, and amino acids; indeed, the wild-type strain did not grow at all on most of these nutrients ( Fig. 4A; Table S2).
The ΔphcA mutant upregulated dozens of genes involved in central carbon metabolism ( Fig. 3; Table S3). These encode catabolic enzymes such as glucosidase, glycogen synthase, galactaric acid dehydratase, and two glucarate dehydratases. The low-celldensity-mimicking strain also upregulated genes involved in fatty acid metabolism, such as the Swanson conversion of pyruvate to acetyl-coenzyme A (acetyl-CoA), acyl-CoA dehydrogenases and acetyltransferases, and the trifunctional enoyl-CoA hydratase/ delta3-cis-delta2-trans-enoyl-CoA isomerase/3-hydroxyacyl-CoA dehydrogenase that catalyzes the ␤-oxidation of fatty acids (Fig. 3). Conversely, expression of genes encoding degradation of some carbon sources available in xylem sap was repressed in the low-cell-density-mimicking mutant. These include genes encoding endoglucanase, exoglucanase A, trehalase, glycoside hydrolase, and a sucrose-specific phosphoenolpyruvatecarbohydrate phosphotransferase system.
Blast2Go node analysis identified nitrogen metabolism among the biological processes most affected by PhcA (Fig. S4A). Nitrate is relatively abundant in tomato xylem sap, and during pathogenesis R. solanacearum strongly expresses several pathways that metabolize inorganic nitrogen (29). Nitrogen assimilation and denitrifying respiration (which enables ATP production in the hypoxic xylem vessels) are both required for normal tomato colonization and bacterial wilt virulence (37,38). At high cell density, PhcA represses genes encoding nitrate reductase (Table S3) and putative transporters of nitrate, glutamate (gltJKL), and glutathione (Table S1B). In contrast, PhcA upregulates nitrite reductase, and direct measurements confirmed that the ⌬phcA mutant produces more nitrite than the wild type both in vitro and in planta ( Fig. S4B and C). Interestingly, the downstream denitrification steps function whether the quorum sensing system is active or not, because similar levels of nitric oxide were detected in the ΔphcA strain, the wild-type strain, and the phcA-complemented strain growing under anoxic conditions in vitro (Fig. S4D). Several genes encoding the amino acid catabolic pathways that generate tricarboxylic acid (TCA) cycle intermediates (proline, valine, glycine, glutamic acid, or serine) were also DE ( Fig. 3; Table S3).
The metabolic capacity of the ⌬phcA mutant was even broader with respect to sources of phosphorus and sulfur than for carbon and nitrogen sources. The ΔphcA mutant grew faster than the wild type on 83% of tested sulfur sources, including sulfate and many other sulfur-containing organic compounds present in soil and xylem sap (39, 40) ( Fig. S3B; Table S2). Four sulfur assimilation genes that lead to cysteine  Table S1B for specific gene expression levels. (B) Selected R. solanacearum metabolic pathways integrated with transcriptomic results. Metabolic pathways are based on KEGG analysis of the R. solanacearum GMI1000 genome. Compounds in blue are present in xylem sap (see references in Table S3). Purple arrows and boxes indicate reaction steps corresponding to enzymes whose expression was upregulated at low cell density (in the ⌬phcA mutant); green arrows and boxes indicate enzymes whose expression was downregulated at low cell density (in the ⌬phcA mutant). synthesis were upregulated in the ⌬phcA mutant ( Fig. 3; Table S3), as well as ABC transporters of sulfate and alkane sulfonate (Table S1B). The low-cell-density-mimicking mutant grew faster than the wild type on 81% of tested sources of phosphorus, which is key to many cellular processes, including signal transduction and energy storage and release. Organic phosphate sources more available to low-cell-density-mode R. solanacearum cells included glycolytic intermediates, sugar phosphates, nucleotides, and phospholipid precursors. Together, these results imply that R. solanacearum benefits from increased sulfur and phosphorus uptake at low cell densities.
Low-cell-density-mode-mimicking R. solanacearum grew better under low-iron conditions. Consistent with previous in vitro findings (19), we found that genes encoding ferric iron uptake were upregulated in the ⌬phcA mutant. These included predicted iron-specific transporters and half of the 20 genes related to siderophore synthesis or regulation ( Fig. S2A; Table S1A), suggesting that R. solanacearum is better able to scavenge iron at low cell densities than at high cell densities. We tested this hypothesis by comparing the growth rates of ΔphcA and wild-type bacteria in minimal medium supplemented with different concentrations of ferric chloride. The ΔphcA mutant grew better than the wild type under both low-iron conditions and high-iron conditions, but iron concentrations of Ͼ1 M provided no further growth advantage (Fig. 4B). These combined transcriptional and functional results indicate that R. solanacearum expends significant resources at low cell densities to obtain iron but that these systems are less active at higher cell densities, possibly because R. solanacearum can acquire sufficient iron by other means in the plant.
The ⌬phcA mutant is better at using nutrients present in xylem sap than the wild type. High-cell-density-mode R. solanacearum grew more slowly or not at all on many nutrients. Does PhcA repress catabolism of these nutrients because they are not present in xylem sap? To characterize the nutrients available to R. solanacearum during tomato pathogenesis, we compiled a list of compounds known to be present in tomato xylem sap, based on published analyses of sap from healthy plants and on our untargeted metabolomic analysis of sap from R. solanacearum-infected wilt-susceptible tomato cultivar "Bonny Best" (Table S3) (29,38,(41)(42)(43)(44). Transcriptomic data suggested that many of the metabolites present in tomato xylem sap (highlighted in blue in Fig. 3) are more accessible to the low-cell-density-mimicking ⌬phcA mutant. We functionally validated the RNA-seq results by comparing the growth rates of the wild-type and ⌬phcA strains on selected compounds from this list used either as the sole carbon source or, for N-containing compounds, as the sole nitrogen source (Table S3). The  (Table S3) as the sole carbon or nitrogen source. (B) R. solanacearum ⌬phcA grew better than the wild-type strain under low-iron conditions. R. solanacearum strains were grown aerobically in BMM broth supplemented with increasing concentrations of ferric chloride (FeCl 3 ). Bacterial growth was measured as absorbance at 600 nm at 24 h. The data show results from three biological replicates, each performed with three technical replicates (n ϭ 9). Asterisks indicate differences between the strains (ANOVA, P value Ͻ 0.001).
R. solanacearum PhcA Modulates Attachment and Growth ⌬phcA mutant was better able to use 36 of 76 tested xylem sap nutrients ( Fig. 4A; Table S3). Nine metabolites were used similarly by the two strains, and only galactose and trehalose supported growth of the wild-type strain alone. The remaining 29 tested xylem sap compounds did not support growth of either strain under our conditions. Thus, R. solanacearum can grow faster on many nutrients present in tomato xylem sap at low cell densities than it can after reaching high cell density. This may facilitate rapid pathogen growth at the earliest stages of host colonization.
However, the ⌬phcA mutant also grew better than the wild type on 18 carbon or nitrogen sources that are not known to be present in xylem sap, such as butyric acid, tartaric acid, and cysteine (Table S2). Upregulating the corresponding transporters and catabolic pathways at low cell densities could increase the fitness of R. solanacearum outside plants by enabling the bacterium to access diverse nutrients present in low concentrations in soil and water.
PhcA affects R. solanacearum attachment to abiotic surfaces. R. solanacearum forms biofilm-like aggregates inside xylem vessels in wilting plants, and R. solanacearum mutants that make abnormally thick biofilms in vitro are reduced in virulence (45,46). Interestingly, many genes potentially associated with bacterial attachment, including six that encode hemagglutinin-like adhesin proteins, were differentially expressed in the low-cell-density-mimicking ⌬phcA mutant. Three hemagglutinin genes were strongly upregulated, while three others were downregulated, suggesting that the PhcA QS system modifies the ability of R. solanacearum cells to aggregate and/or to adhere to surfaces in complex ways ( Fig. S2A; Table S1C). To better understand this trait, we measured bacterial attachment to abiotic surfaces in vitro. The ⌬phcA mutant formed much less biofilm than the wild type on polyvinyl chloride (PVC) plates stained with crystal violet (Fig. 5A). Confocal imaging analysis of 24 h biofilms on glass plates revealed that the ΔphcA mutant did not form structured three-dimensional biofilms like wild-type R. solanacearum, possibly because it was arrested in a low-cell-density mode optimized for initial substrate attachment. The ΔphcA mutant attached to a 3-foldlarger area than the wild-type strain or complemented mutant strain (P Ͻ 0.01, t test) ( Fig. 5B and C). The ΔphcA biofilms were 2-fold thinner and contained 7-fold less biomass than those of the wild type (P Ͻ 0.05, t test). After 48 h of static growth in rich medium, the wild-type cells formed a floating pellicle at the air-liquid interface, but the ⌬phcA mutant did not (Fig. 5D). Although this distinctive phenotype suggested that low-cell-density-mode R. solanacearum cells do not self-aggregate as well as they do at high cell density, its biological significance in a natural context remains to be determined.
⌬phcA cells colonizing tomato xylem are morphologically different from wildtype cells. We used scanning electron microscopy (SEM) to image colonized tomato stem tissue harvested at 72 h postinoculation, when the wild-type and ⌬phcA R. solanacearum cells had reached comparable population sizes. This is the same time point at which their transcriptomes were profiled. The wild-type bacteria had colonized most of the xylem vessels examined, where they formed loose aggregates of cells connected by thin, smooth fibers (Fig. 6B). In contrast, ΔphcA cells were present in fewer xylem vessels. In the vessels that were colonized, the ⌬phcA bacteria formed dense mats of cells embedded in fibers that were more numerous, more kinked, and rougher than the fibers in the wild-type aggregates (Fig. 6D, E, and F). These fibers were on average four times thicker than those in the wild-type cell aggregates. In planta, the ΔphcA cells were on average 25% shorter in their longest dimension than the wild-type cells (Fig. 6A and D). The ⌬phcA cells were oval, in contrast to the rod-shaped wild-type cells (Fig. 6A and D). Finally, the surfaces of the ΔphcA and wild-type cells had different textures in the SEM images; the wild-type cells had a uniform, rugose surface whereas the ⌬phcA cell surfaces appeared more irregular, possibly reflecting the absence of EPS ( Fig. 6C and F). Overall, the wild-type and ⌬phcA bacteria were qualitatively different during tomato plant colonization with respect to their distribution; aggregate matrix composition; and the size, shape, and surface texture of their cells.
The hyperattachment behavior of the ⌬phcA mutant restricts dispersal in the host. Because SEM analysis revealed that the ΔphcA mutant formed denser aggregates in xylem vessels than the wild type, we predicted that the ⌬phcA mutant would adhere more avidly to plant surfaces. We compared the levels of attachment of wild-type and ⌬phcA bacteria to tomato seedling roots. Nine times more ΔphcA cells than wild-type cells were attached to the roots after 2 h (Fig. 7A). This is consistent with the idea that at low cell density, R. solanacearum cells adhere better to plants, possibly as a result of higher expression of adhesins specific to biotic surfaces. Together, the increased root attachment and limited xylem colonization by the ΔphcA mutant suggested that low-cell-density-mimicking cells might not disperse well inside host plants. However, our metabolic analyses and the observation that ⌬phcA cells grew faster than wild-type cells in artificial media (Fig. S3A) and in ex vivo xylem sap (Fig. 1C) suggested that cells mimicking the low-cell-density condition would grow faster in planta and outcompete the wild-type strain.
To test these opposing hypotheses, we compared the abilities of the ΔphcA mutant to colonize tomato plants alone and in the presence of wild-type cells. When tomato plants were infected with a 1:1 suspension of wild-type and ΔphcA cells using a naturalistic soil-soak inoculation, the wild-type strain consistently outcompeted the ΔphcA mutant in tomato mid-stems (Fig. 7B). Even at 7 days after inoculation, few ΔphcA cells were detected in stems of coinoculated plants, although most plants contained more than 10 8 CFU of the wild-type strain per g stem at that time point (Fig. 7B). Interestingly, this defect was even more pronounced when the ΔphcA mutant was inoculated alone, suggesting that wild-type cells partially complement the ΔphcA colonization defect in trans, possibly by producing EPS or other secreted virulence factors (Fig. 7B). Similar results were obtained following direct coinoculation into the stem through a cut leaf petiole (Fig. 7C). Kinetic analysis showed that in the first 2 days after coinoculation, the ⌬phcA mutant grew better than the wild type at the site of inoculation, possibly due to its ability to better exploit the nutrients available in tomato xylem (Fig. S5). However, the wild-type cells quickly caught up, and by 5 dpi the wild-type bacteria outnumbered the ⌬phcA cells ( Fig. 7C and S5). Moreover, following coinoculation, ΔphcA cells were defective in moving up tomato stems, with a very poor competitive index (CI) of 8.4 ϫ 10 Ϫ4 at 4 cm above the site of inoculation (Fig. 7C).
To determine if the low-cell-density-mode-mimicking ΔphcA strain could spread as well as the wild-type strain in tomato stems, we measured the population sizes of each strain above and below the inoculated petiole 7 days after inoculation. The ΔphcA mutant was strikingly deficient in the ability to move both up and down tomato stems (Fig. 7D). At 4 cm above the inoculation point, the mean population size of the wild-type strain was 4 ϫ 10 8 CFU/g stem compared to only 2 ϫ 10 4 CFU/g stem for the ΔphcA mutant.
Overall, the behavior of the ⌬phcA mutant in planta suggests that it is adaptive for R. solanacearum to adhere to plant surfaces and to form aggregates there at early stages of infection. However, as the bacterium grows to high population densities in xylem vessels, this hyperattachment trait must be repressed so that the pathogen can disperse in the host plant.
The ⌬phcA mutant elicited lower tomato defense gene expression than the wild type. Infection with R. solanacearum causes tomato plants to increase expression of genes in both the ethylene (ET) and salicylic acid (SA) defense pathways (47). To determine if low-cell-density-mode-mimicking cells triggered plant defense gene expression as well as the wild type, we performed quantitative RT-PCR analysis of total RNA harvested from tomato stems 72 h postinoculation (hpi) with either ⌬phcA or wild-type bacteria. Plants infected with comparable populations of the ⌬phcA mutant expressed about 3-fold less of the SA-responsive defense gene PR-1a (P ϭ 0.007, t test) and about 4-fold less of the ET-responsive defense gene ACO5 (P ϭ 0.033, t test) than plants infected with the wild-type strain. Expression of the jasmonic acid (JA) pathway marker Pin2 was not altered (P ϭ 0.64, t test).

DISCUSSION
To understand how R. solanacearum makes the critical transition from living as scattered cells in water, soil, or on root surfaces to the very different conditions it experiences at high population densities in plant xylem vessels, we needed to profile the pathogen's gene expression at low cell density. Harvesting quality RNA from sparse, scattered cells is technically challenging, so we constructed a QS-blind ΔphcA mutant that mimics the behavior of the pathogen at low population densities. The transcrip- The PhcA quorum sensing system enables R. solanacearum to use different nutrient sources at low and high cell densities. Combining transcriptomic data with functional assays and xylem sap metabolomics identified specific mechanisms that expand the carbon sources available to R. solanacearum under low-nutrient conditions. Many metabolic pathways and transporters were expressed only at low cell densities, and functional assays confirmed that high-density R. solanacearum could not grow on the corresponding carbon sources. Thus, the pathogen can access some nutrients only at low cell densities and can also grow faster on many nutrients that are accessible to both modes. This suggests that the PhcA QS system regulates R. solanacearum core metabolism in ways that maximize growth early in colonization but that slow growth once bacteria reach high cell densities.
Among 76 potential nutrient sources detected in tomato xylem, wild-type R. solanacearum (high-cell-density mode) grew better than the ⌬phcA mutant only on trehalose and galactose, while the ⌬phcA mutant grew better or exclusively on 47% of the tested xylem metabolites. Several compounds that feed directly into central carbon metabolism (glycolysis and the TCA cycle) are more available to low-cell-density-modemimicking R. solanacearum cells. These include pyruvic acid, galactaric acid, glucaric acid, malate, citrate, and several others. Malate and citrate are important for the success of other plant-pathogenic bacteria such as Pseudomonas syringae and Xanthomonas campestris pv. vesicatoria (48,49).
R. solanacearum colonizes dead and lignified xylem tissue, which contains diverse carbohydrates such as glycosides, glycans, and glycoconjugates (50). Glucosidases convert these complex sugars to glucose, which is a preferred carbon source for R. solanacearum at both high and low cell densities. Glucosidase expression is upregulated 7-fold in ΔphcA cells, suggesting that host cell wall fragments are important carbon sources early in plant colonization. Conversely, trehalose and galactose, which are accessible only to high-cell-density-mode R. solanacearum, are likely key nutrients during full-blown wilt disease. Strategic use of a preferred carbon source such as glucose when the bacterium is at low cell densities not only provides the fitness advantage of rapid growth early in xylem colonization but also allows R. solanacearum to exploit less-preferred nutrient sources when it has already reached the critical density required for virulence. Further, rapid uptake of glucose, a widely preferred microbial carbon source, may help R. solanacearum outcompete other rhizosphere and endophytic bacteria.
Genes encoding nitrogen metabolism were upregulated in the low-cell-densitymimicking mutant, and the ΔphcA mutant grew better than the wild type on many nitrogen sources, including nitrate, nitrite, ammonia, and 16 amino acids. Plants infected with the ⌬phcA mutant contained higher levels of nitrite, which R. solanacearum can assimilate and convert into glutamine and glutamate via ammonia (51). These amino acids are present in tomato xylem sap, and the GltIJKL glutamate transporters and glutamate dehydrogenase GdhA were upregulated in the ΔphcA mutant, consistent with increased assimilation of these compounds at low cell densities. In addition to representing acquisition of nitrogen for growth, increased nitrite levels may improve R. solanacearum survival during initial infection in the hypoxic host environment (52). Nitrates protect Mycobacterium tuberculosis from hypoxic stress by replacing oxygen as the terminal electron acceptor in the respiratory chain (53). The M. tuberculosis nitrate importer NarK2 is highly induced during hypoxia to levels similar to those that R. solanacearum encounters in tomato xylem (38). The upregulation of NarK2 in the ΔphcA mutant, along with the finding that denitrifying respiration is critical for R. solanacearum success during pathogenesis (38), suggests that nitrate may play a similar protective role for R. solanacearum in planta.
A recent in silico model of R. solanacearum strain GMI1000 metabolism also suggested that a phcA mutant had a generally broader catabolic capacity than the wild type, with specific functional results mostly consistent with our data (54). However, in contrast to their report, but in agreement with our transcriptomic data, we found that the ΔphcA mutant did not grow on galactose either in the Biolog galactose or in minimal medium (MM) plus galactose. This difference may have been a consequence of the fact that Peyraud and coworkers starved R. solanacearum cells in water for 6 h before inoculating the Biolog plates, whereas we harvested metabolically active bacteria from mid-log phase. Minimal medium and Biolog growth assays gave inconsistent results for a few other key carbon sources, such as sucrose, trehalose, and glycerol.
Downregulation of genes encoding a trait did not always lead to loss of function. For example, despite the low expression levels of scrAB sucrose transport and catabolism genes in the ΔphcA mutant, the low-cell-density-mimicking mutant grew as well as the wild type in MM supplemented with 10 mM sucrose. Expression of the scrRAB locus is induced by sucrose; if the proprietary Biolog medium contains little sucrose, this could explain why the mutant did not grow in the Biolog assay (29,55). This carbon source is important in bacterial wilt disease development; a scrA mutant of R. solanacearum strain UW551 is reduced in virulence in tomato (29).
Taken together, the results of our transcriptomic and functional assays demonstrate that R. solanacearum uses many sources of carbon, nitrogen, sulfur, phosphorus, and iron more efficiently at low cell densities than when it is in high-cell-density mode. Indeed, the pathogen can use many of these substrates only at low cell density. How does this broad metabolic shift increase pathogen fitness? Perhaps the low-cell-densitymode bacteria are optimized to use nutrients available outside the plant in more chemically diverse and nutrient-poor rhizospheres and surface water habitats (56). This environmental nutrition hypothesis is consistent with the ability of the ΔphcA mutant to use compounds that have not been reported to be present in tomato xylem sap, including sulfur sources such as cysteine and glutathione. These may contribute to bacterial success in nonnutritional ways. For example, R. solanacearum at low cell densities could protect itself from oxidative damage inside the host plant by accumulating glutathione via upregulation of the gsiABCD glutathione ABC transporter and the ggtI and gstH glutathione S-transferases (57,58).
Surprisingly, quorum sensing does not optimize high-cell-density-mode R. solanacearum to use nutrients in tomato xylem sap, where the bacterium multiplies rapidly (59). Instead, PhcA represses many catabolic pathways so that, once it exceeds 10 7 CFU/ml, R. solanacearum uses only a few of the nutrients that it encounters during pathogenesis. We favor the hypothesis that low-cell-density-mode R. solanacearum more efficiently exploits compounds in xylem sap to fuel the pathogen's explosive growth early in disease development, before the QS system is activated. But the pathogen no longer uses all available xylem nutrients once it reaches high cell density and begins to express virulence factors. These findings are consistent with an in silico metabolic model suggesting that the PhcA QS system mediates a trade-off between optimal growth early in plant colonization and resource-demanding production of virulence factors later in disease development (54).
Similar trade-offs have been identified in many other bacteria (60). Environmental constraints and finite resources force bacteria to choose between two demanding traits at different developmental or disease stages (89). Their broader catabolic capacity and more rapid growth explain why spontaneous phcA mutants grow faster than the wild type in culture (7). Peyraud et al. proposed that spontaneous phcA mutants, which occur frequently (6), should outcompete wild-type cells in plants (54). We tested this hypothesis and found that it is not the case: wild-type cells outcompeted the ΔphcA mutant in planta by more than 10,000-fold. Moreover, phcA-type cells are rarely or never isolated from infected plants in the field (7). This suggests that, inside the plant, the growth advantage of phcA-type cells is outweighed by other disadvantageous traits. What are these traits?
The bacterium's ability to adhere to biotic and abiotic surfaces is dramatically modified by R. solanacearum cell density. While ΔphcA cells did not form biofilms on abiotic substrates, they attached to roots much better than wild-type cells. Avidly binding to roots of potential hosts would be adaptive for R. solanacearum at low cell densities in soil. However, SEM images revealed that once they were inside a plant, ⌬phcA cells formed abnormally thick, matted aggregates. All these phenotypes may be the result of altered expression of hemagglutinin-related nonfimbrial adhesins in the ⌬phcA mutant. Adhesins are surface proteins that help bacteria attach to eukaryotes, an essential step in host-microbe interactions (61). Nonfimbrial adhesins influence the attachment behavior of phytopathogenic bacteria such as Xylella fastidiosa and Xanthomonas fuscans, which are initially bound to a surface by nonfimbrial adhesins (62)(63)(64)(65). Once bacteria attach to a surface, fimbrial adhesins (pili, fimbriae) help bacteria stick to each other and form aggregates on the host surface. In general, adhesins contribute to virulence by playing specific roles at different disease stages (early infection, seed transmission, attachment to insect vectors, etc.). Nutrient availability can also influence attachment behavior; under low-nutrient conditions, Acinetobacter cells attached with increased stability and formed less-dispersive biofilm structures (66).
The R. solanacearum genome encodes about 20 putative hemagglutinin-related proteins, of which 6 were regulated by PhcA; 3 of the 6 showed higher expression at low cell densities, and the other 3 showed higher expression at high cell densities. A systematic analysis of mutants lacking these differentially expressed adhesins would identify their roles in attachment to various substrates throughout the disease cycle.
The low-cell-density-mimicking mutant upregulated genes encoding degradation of hydroxycinnamic acids (HCAs). These plant-produced compounds inhibit R. solanacearum, and we previously found that HCA degradation protects R. solanacearum from HCA toxicity in tomato roots, where these compounds are abundant (67). This is consistent with the hypothesis that, at low cell densities, the PhcA QS system upregulates traits that the pathogen needs early in disease development.
Inside the plant, the morphology of ΔphcA cells was strikingly different from that of wild-type cells: the ⌬phcA cells were smaller, rougher, and more oval. Similar morphological differentiation was seen in stationary-phase Escherichia coli and Arthrobacter cells undergoing a survival response (68,69). Under low-nutrient conditions, Acinetobacter sp. strain GJ12 cells switch from bacillar to coccoid morphology (66), which increases their surface-to-volume ratio and accommodates more nutrient transporter complexes (70). The increased roughness of the ΔphcA cell surfaces may be due to lack of EPS (71). Adding exogenous EPS to a spontaneous EPS-deficient rough mutant of R. solanacearum facilitated escape from agglutination (71), consistent with a role for EPS in colonization.
The ΔphcA cells were defective in plant colonization, occupying fewer xylem vessels and spreading up and down stems more slowly than the wild-type cells, possibly because they could not escape their abnormally matted aggregates. QS mutants of Pantoea stewartii subsp. stewartii also remained localized at the site of infection (72). Conversely, an rpfF QS mutant of Xylella fastidiosa had reduced adhesiveness, spread better in grapevine vessels than the wild type, and was hypervirulent after mechanical inoculation (73). Nonetheless, that mutant had very low fitness overall because it did not form biofilms in its insect vector and was poorly transmitted (74). R. solanacearum biofilms contain fibers made of extracellular DNA, and endonuclease-deficient mutants produced thick biofilms and spread poorly in tomato (46). A putative nucleaseencoding gene, RSp1603, was downregulated 35-fold in the ⌬phcA mutant. This may preserve the structural integrity of developing biofilms by protecting the DNA fibers. Further investigation is needed to explore the connection between extracellular DNA degradation and the structure of ⌬phcA aggregates in planta.
The hyperattachment trait of the ΔphcA mutant, which is consistent with its reduced competitive fitness and poor dispersal in plant stems, suggests that despite having the advantage of broad metabolic capacity, spontaneous phcA mutants are at a strong selective disadvantage in planta. Their altered morphology, root attachment, and aggregation suggest that these behaviors are advantageous at low cell densities in soil and early in plant colonization but must be downregulated for the pathogen to successfully colonize and spread as disease develops.
In an interesting parallel, it is well established that plants also execute an evolu-tionary trade-off between maximizing growth and maximizing defenses (75). Plants responded differently to R. solanacearum cells in the low-cell-density and high-celldensity modes. Wild-type R. solanacearum infection triggers rapid induction of salicylic acid and ethylene-responsive defense pathway genes, but both of these host signaling pathways were significantly less induced by the ⌬phcA mutant, even when comparable numbers of pathogen cells were present. Similarly, human bronchial epithelial cells produced much less of the defensive cytokine interleukin-8 in response to a QS mutant of Pseudomonas aeruginosa than in response to wild-type P. aeruginosa (76). The R. solanacearum QS mutant is compromised in expression of multiple virulence factors, some of which may trigger plant recognition and defense. It would clearly be advantageous for R. solanacearum to avoid activating host defenses early in plant colonization. Thus, reduced expression of these virulence factors not only saves energy but may also make R. solanacearum less visible to plant defenses. The lack of EPS in the ΔphcA mutant cannot explain this reduced defense gene expression because an EPSdeficient R. solanacearum mutant still induces wild-type defenses in wilt-susceptible plants such as the cultivar used here (47). Some R. solanacearum T3-secreted effectors suppress host defenses, but many T3SS-encoding genes were downregulated in the ⌬phcA mutant, including those encoding the system's major transcriptional activators, HrpB and HrpG. Nonetheless, the low-cell-density-mimicking ⌬phcA mutant ultimately failed to colonize plants. It is possible that the mutant's inability to spread normally meant that, overall, fewer plant cells perceived the pathogen, leading to a muted defense response. This could be tested using laser-capture microdissection to measure defenses in single plant cells that are in contact with either wild-type or ⌬phcA bacteria.
Along with identifying new QS-regulated phenotypes, this study also revealed some inconsistencies between the R. solanacearum gene expression levels measured in culture and the pathogen's transcriptional behavior in the natural environment of the host plant (summarized in Table S5). Most notably, the expression of the major T3SS regulator HrpB in culture is repressed by PhcA at high cell densities; this generated a model proposing that T3 secreted effectors were important only early in plant colonization (27). However, we found that many T3SS-related genes, including hrpB, were actually more highly expressed at high cell densities in planta. This is consistent with previous in planta experiments showing that T3SS genes are highly expressed during disease (28)(29)(30). It will be interesting to identify the intervening regulatory players that reverse this regulatory pathway in the plant.
Importantly, the gene expression levels reported here reflect averages across biologically heterogeneous bacterial populations that occupy distinct microenvironments. We extracted R. solanacearum RNA from infected tomato stem tissue, which included both the planktonic bacteria and those in aggregates or attached to plant surfaces. It is very likely that early in disease, cells in biofilms are in a different QS mode than individual cells in flowing xylem sap. Follow-up experiments could distinguish between gene expression levels in planktonic versus attached cells in xylem by using reporter gene fusions to compare levels of transcription of PhcA-responsive genes across subpopulations (77).
In conclusion, the in planta PhcA transcriptome identified novel traits that help R. solanacearum survive at low cell densities. The PhcA QS system controls two transitions essential for successful bacterial wilt pathogenesis. First, R. solanacearum expresses diverse transporters and catabolic pathways to maximize survival in soil or water and growth during early infection, but once infection has been established, these traits are repressed by PhcA in favor of virulence factor expression. Second, PhcA modulates the switch between a highly adhesive form that forms aggregates and attaches avidly to host surfaces and a more dispersive form that spreads rapidly within plants during host colonization. Because these transitions clearly play critical roles in the pathogen's complex disease cycle, they are attractive targets for disease control. Exogenously applied, microbiome-delivered, or host plant-produced QS inhibitors, quorum quenchers, or QS autoinducers (78, 79) could effectively reduce R. solanacearum success and, by extension, reduce crop losses due to bacterial wilt disease.
Data quality check and analysis. Illumina read raw data quality was verified with fastQC. The following genomes were used for mapping: the ITAG2.4 tomato genome (S_lycopersicum_chromosomes. 2.50.fa) and the R. solanacearum strain GMI1000 genome (http://www.ncbi.nlm.nih.gov/nuccore/NC _003295.1 and http://www.ncbi.nlm.nih.gov/nuccore/NC_003296.1). Raw sequence reads were mapped to genomes using the Subjunc aligner from Subread; most reads (Ͼ90% for all samples) aligned to the combined genomes, with about 1.2 to 4.5% total (84) reads per sample aligned to the bacterial genome. Bacterial alignment bam files were compared to the gene annotation GFF file, and raw counts for each gene were generated using the Subread featureCounts tool. Raw count data were normalized using the R Limma voom package (85) and then analyzed for differential expression. Blast2GO (86) was used for further data analysis and presentation.
Metabolomic analysis of xylem sap. Tomato xylem sap was harvested by root pressure. At 1 to 3 h after light onset, well-watered 26-to 28-day-old plants were detopped with a sharp blade. The sap that accumulated in the first 3 min was discarded, the stump was rinsed with water, and then sap was collected for 30 min. Five sap samples, each containing sap pooled from 4 plants, were analyzed. Samples were shipped on dry ice to the West Coast Metabolomics Center (Davis, CA) for gas chromatographymass spectrometry (GC-MS) analysis. Metabolites were detected on a Pegasus IV GC instrument (Leco Corp., St. Joseph, MI) coupled to a Pegasus IV time of flight-MS (TOF-MS) system as previously described (87).
Nitrite measurement. Nitrite was quantified both in vitro and in planta as previously described (38). The in vitro experiment included three biological replicates with three technical replicates each. For in planta measurements, homogenized stems of uninfected plants were used as controls. Nitrite concentrations were calculated on a per-cell basis. Fourteen plants were sampled for each treatment.
Scanning electron microscopy to visualize R. solanacearum in tomato xylem. Scanning electron microscopy was performed as previously described (46). The experiment was repeated twice with a minimum of three plants per treatment. Two sections from each plant were visualized using a LEO 1530 scanning electron microscope at the University of Wisconsin-Madison (UW-Madison) Material Sciences Center.
qRT-PCR. Expression of selected plant and bacterial genes was measured by quantitative reverse transcriptase PCR (qRT-PCR) as previously described (29). The threshold cycle (ΔΔC T ) method was used to calculate log fold expression changes; data reflect three biological replicates, with two technical replicates each. Primers used for qPCR are listed in Table S4.
Accession number(s). The complete experimental data set was deposited in the Gene Expression Omnibus (GEO) database under GenBank accession number GSE98074.