Comparative genomics, based on an ever-increasing number of complete genome sequences, can be used to reveal numerous insights into host–pathogen interactions. However, hypotheses obtained using these strategies should be experimentally demonstrated by functional approaches. The application of STM to a range of microbial pathogens has resulted in the identification of novel fitness-related genes and virulence determinants in each screen performed to date (Autret & Charbit, 2005). To our knowledge, the only STM study involving a bacterial phytopathogen identified 19 Erwinia amylovora genes required during the infection process of shoots of apple trees (Malus domestica), most of which were metabolic-related genes (Wang & Beer, 2006). However, our application of this technique to the interaction of P. savastanoi pv. savastanoi with olive plants represents the first STM analysis of a bacterial strain belonging to the P. syringae complex. Screening for virulence factors based on STM relies on the premise that virulence of a given mutant strain within a complex mixed infection reflects its virulence in an individual infection (Hensel et al., 1995), a phenomenon dependent on the dose of inoculation, including plant infection by P. syringae (Macho et al., 2007). In relation to P. savastanoi pv. savastanoi infection of olive plants, optimization of the inoculum dose for STM analysis has been previously reported (Rodríguez-Moreno et al., 2008). Complete coverage of the NCPPB 3335 genome was not achieved in this study, as the number of open reading frames (ORFs) predicted in the genome sequence of this strain (5232 ORFs; Rodríguez-Palenzuela et al., 2010) is slightly higher than the total number of STM mutants screened (4741 strains). Considering the frequency of multiple Tn5 insertions into the genome of NCPPB 3335 (Pérez-Martínez et al., 2007), as well as the number of siblings and independent insertions in the same gene identified in this study (Table S3), the total number of STM mutants tested in olive plants would represent a genome coverage of c. 67%. Nevertheless, STM libraries composed of lower numbers of mutants have been successfully applied for the identification of virulence factors in several bacterial human pathogens (Saenz & Dehio, 2005; Wang & Beer, 2006).
Metabolic pathways required by P. savastanoi pv. savastanoi in olive plants
Plant-pathogenic P. syringae are nutritionally specialized to use a limited set of nutrients that are abundant in the plant apoplast, in leaf exudates, and on the leaf surfaces (Rico & Preston, 2008). Metabolic-related genes interrupted by the transposon in FAM strains (Fig. 2, Table 2) are, in general, not associated with virulence per se but provide information about the nutritional limitations of P. savastanoi inside the host and the metabolic pathways that are relevant for infection of olive plants by this pathogen. As expected for the screening method used in this study, over 50% of the FAM strains selected were metabolically deficient mutants; together, they had disruptions in the biosynthetic pathways of nine amino acids and three vitamins (Fig. 2, Table 2), indicating that the concentrations of these compounds are limiting for bacterial growth in olive plants but are not restrictive for growth in LB medium. Although it cannot be ruled out that mutants with disruptions in the remaining amino acid biosynthetic pathways either were not represented in our library or were discarded, these results could also indicate that some of these amino acids could be nonlimiting compounds in the apoplast. This would mean that mutants for their biosynthetic pathways would not be selected through our screening. Although no data related to the amino acid content of the olive apoplast are currently available, a recent study has revealed that P. syringae pv. tomato DC3000 uses amino acids that are abundant in the plant apoplast (Rico & Preston, 2008). Interestingly, and with the exception of glutamate, a key molecule in cellular metabolism, the amino acid biosynthetic pathways identified in this study were those corresponding to amino acids less abundant in the tomato (Solanum lycopersicum) apoplast. In addition, our screening identified FAM strains with a disrupted citrate (citN) or glutamate transporter (gltP) (Fig. 2, Table 2), indicating that these mutants are defective in the acquisition of these compounds, which are abundant in the tomato apoplast (Rico & Preston, 2008).
Bioinformatics analysis of the genome of P. savastanoi pv. savastanoi NCPPB 3335 revealed the existence of various genes encoding candidate enzymes involved in the complete degradation of aromatic compounds to intermediates of the Krebs cycle (Rodríguez-Palenzuela et al., 2010). Related to this battery of genes, the catJ mutant (Fig. S2) showed a drastic reduction in both fitness (Fig. 1) and virulence in planta (Figs 3, 5). Together, these results suggest that P. savastanoi pv. savastanoi could be adapted to use or detoxify aromatic compounds present at high concentrations in the tissues of woody plants. In fact, the production of phenolic compounds is greatly increased in olive tree knots upon P. savastanoi pv. savastanoi attack (Cayuela et al., 2006), which strongly suggests that bacterial resistance to phenols could be of paramount importance in the pathogenicity of this bacterium. However, all these hypotheses remain to be tested.
Fitness and virulence genes not previously identified in P. savastanoi pv. savastanoi
Our STM screening identified mutations affecting previously reported virulence and pathogenicity factors in P. savastanoi, such as the biosynthesis of IAA (iaaH gene, strain FAM-110) and the T3SS (hrpR gene, strain FAM-117), which validates the use of young micropropagated olive plants for the identification of virulence genes in this pathogen. In relation to the T3SS, the virulence phenotype, the cellular localization in olive tissue and the structure of the tumors observed for the hrpR mutant (Figs 3, 5, 6) were in agreement with our previous results (Pérez-Martínez et al., 2010).
Fitness and virulence genes of P. savastanoi pv. savastanoi identified in this study included those related to the Sec pathway, which is essential in other bacteria for peptide translocation across or into the bacterial membrane and for the functional assembly of the T3SS and T4SS (Chen et al., 2000; Kimbrough & Miller, 2002). Peptides translocated through the transmembrane SecYEG complex are then correctly folded by the periplasmic chaperone PpiD (Antonoaea et al., 2008). Transposon insertions into either secG or ppiD resulted in a significant fitness reduction in planta (Fig. 1), which in the case of the secG mutant was clearly reflected by a reduction of the GFP fluorescence observed in tumors induced by this strain (Fig. 5). In addition, the spatial distribution of pathogen cells and the internal structure of the tumors induced by this strain were similar to those visualized for the hrpR mutant (Fig. 6), indicating that the P. savastanoi pv. savastanoi Sec pathway is also essential for host cell lysis and for the spatial distribution of the pathogen inside olive tissues.
Type IV secretion systems are multiprotein complexes that mediate translocation of macromolecules (proteins, DNA or DNA–protein complexes) across the bacterial cell envelope into recipient cells (Álvarez-Martínez & Christie, 2009). T4SSs have been subgrouped into type IVA (vir genes) and type IVB (tra genes) (Christie et al., 2005). Although the role of the T4SSs in virulence is still not well understood in Pseudomonas, in other animal pathogenic bacteria, such as Bartonella (Schroder et al., 2011), Helicobacter (Backert & Clyne, 2011) and Brucella (de Jong & Tsolis, 2011), this system has an important role in pathogenesis. Our screening identified transposon insertions in both type IVA (virB4 gene) and type IVB (traY gene). These mutants showed a significant virulence reduction in adult olive plants, indicating that these systems are also relevant for P. savastanoi pathogenicity.
One of the most rapid plant defense reactions to pathogen attack is the so-called oxidative burst, which constitutes the production of ROS at the site of attempted invasion (Apel & Hirt, 2004). Bacteria have developed mechanisms to survive in these hostile conditions; however, the roles of these mechanisms in the virulence of the P. syringae complex have not been studied in detail. Our screening identified three P. savastanoi pv. savastanoi mutants with disruptions in rubB, pqiB, and msrA. Rapid metabolic reduction of ROS is activated in the electron transport chain, where the protein RubB acts (Hagelueken et al., 2007). However, genes induced by high levels of radicals include pqiB, whose precise function in detoxification is still unknown. Finally, MsrA contributes to the virulence of Dickeya dadantii by repairing oxidized proteins via reduction of methionine sulfoxide to methionine (García-Olmedo et al., 2001; Ezraty et al., 2005). Tumors induced by rubB or pqiB mutants also exhibited a compact tissue structure (Fig. 6), indicating that host cell lysis mediated by P. savastanoi pv. savastanoi is impaired under oxidative stress conditions. Also related to oxidative stress, P. savastanoi pv. savastanoi mutants disrupted in homologs of arsB and cioA, encoding a pump membrane protein involved in arsenite resistance (Parvatiyar et al., 2005; Patel et al., 2007) and subunit A of a cyanide-insensitive terminal oxidase involved in adaptation of Pseudomonas aeruginosa to copper limitation (Frangipani et al., 2008), respectively, were selected in our screening, although virulence reduction in olive plants was more pronounced for the cioA mutant (Fig. 3).
The cell wall or exoskeleton is a critically important structural entity in bacteria. STM mutants disrupted in peptidoglycan (PG)-related genes (pomA, mltB, and ampG mutants) and the cls gene, encoding cardiolipin synthetase, were hypovirulent in woody olive plants (Fig. 3). Interestingly, and although a GFP-tagged ampG mutant induced tumors whose sizes were similar to those induced by the wild-type strain, GFP emission by these strains did not cover the entire knot surface, suggesting that intact AmpG is required for bacterial propagation inside olive tissue.
The second messenger 3′,5′-cyclic diguanylic acid (c-di-GMP), which is synthesized by diguanylate cyclase (DGC) and degraded by phosphodiesterase A, is a central regulator of the prokaryote biofilm lifestyle. Recent evidence also links this molecule to virulence (Cotter & Stibitz, 2007). Over 30 different genes encoding GGDEF-domain DGCs were recently identified in the genome of P. savastanoi pv. savastanoi NCPPB 3335 (Rodríguez-Palenzuela et al., 2010), among which AER-0004088 was interrupted by the transposon in strain FAM-030. Only a subtle reduction of virulence was observed for FAM-030 in woody olive plants (Fig. 3), perhaps as a result of functional redundancy of DGC proteins in P. savastanoi pv. savastanoi NCPPB 3335.
Although many bacterial virulence determinants are thought to be host-specific, several universal virulence mechanisms are shared by bacterial pathogens of plants, insects, and mammals (Rahme et al., 2000). Further supporting this notion, several of the P. savastanoi pv. savastanoi virulence genes identified in this study (Fig. 3, Table 3) were previously selected in other STM screenings involving human and animal bacterial pathogens (Table S5). Although some of the specific virulence-associated genes identified here have orthologs in other pathogenic bacteria, for example, arsB, cioA and spmAB, the exact roles of these genes in pathogenesis remain unclear.
In summary, this novel application of STM to a bacterial phytopathogen belonging to the P. syringae complex allowed us to identify metabolic pathways required for full fitness of P. savastanoi pv. savastanoi in olive plants and, additionally, revealed novel mechanisms involved in the virulence of this pathogen, such as the type IV secretion system, a battery of genes involved in tolerance and detoxification of ROS, a set of genes required for the biosynthesis of the cell wall, and a gene that regulates c-di-GMP levels. Other features identified in this study but not analyzed in detail include a chemotaxis-related protein, several DNA-binding proteins and seven hypothetical proteins. Further functional studies of the genes identified here may promote a better understanding of the pathogenic processes of bacterial phytopathogens and, more specifically, of the interaction of P. savastanoi pv. savastanoi with olive plants.