We used the SctC, SctN, SctT, and SctV sequences of seven well characterized T3SS representatives (Abby and Rocha, 2012) to identify the T3SS clusters of Burkholderia s.l. species through homology screening using BLASTp (Supplementary Table 1). The amino acid sequences of SctC, SctN, SctT, and SctV from the representatives of each T3SS type were individually screened for in a database representing the diversity of Burkholderia s.l. with one representative strain for each species (mostly the type strain and alternatively the strain with the genome of highest quality; Supplementary Table 2). The database additionally holds 10 new strains that were isolated from rice roots and belong to either Burkholderia or Paraburkholderia (Bioproject PRJEB31911 and PRJEB42536). Strains were collected, and genomes sequenced as described in Wallner et al. (2019). Genome annotation was performed by the MicroScope platform (Vallenet et al., 2020). Duplicate hits from different BLAST runs were curated and the hit of highest homology conserved.

To infer the phylogenetic evolution of Burkholderia s.l. strains nine coding sequences (dnaG, ftsZ, glnA, gyrB, pykA, recA, rho, rpoB, and secA) of 146 Burkholderia s. l. strains and two outgroup strains (i.e., Ralstonia solanacearum K60 and Cupriavidus metallidurans CH34) were used. The genes were aligned separately using the MUSCLE software and subsequently concatenated resulting in a 16,630 bp alignment. An independent alignment was generated using the same loci for the 80 T3SS-bearing Burkholderia s.l. strains.

To specifically study the evolution dynamics of the T3SS, four protein sequences (SctC, SctN, SctT and SctV) were aligned separately using MUSCLE and subsequently concatenated to form a 1,652 aa alignment for 136 T3SS found in Burkholderia s.l. and 15 reference strains (Supplementary Table 2). Additional sequences of the protein homologs in the flagella coding cluster (FliI, FliR, FlhA) of E. coli were added (No SctC homolog exists) to form the outgroup.

All phylogenetic trees were constructed using the BEAST v1.10.2 software with GTR + G + I and LG + G + I substitution models for nucleotide and amino acid sequences, respectively (as determined to be the most suitable using the MEGA-X built-in tool for model selection), and assuming constant lineage birth rate for each branch with a Yule tree prior (best suited for describing the relationships between individuals from different species). The analysis was run for 5,000,000 generations, sampling every 1,000 generations, and a burn-in of 500,000 was applied. Posterior probabilities above 95% are reported in the form of black dots at the nodes.

The PIP-box promoter region was screened for using the prosite format sequence: T-T-C-G-[CGT]-N(15)-T-T-C-G-[CGT]-N(30,32)-[CT]-A-N(3)-T and the EMBOSS v6.5.7.0 fuzznuc software to infer abundance in target genomes. To specifically detect PIP-box regions that are concurrently found in promoting regions of genes, the EDGAR2.3 platform was used (Blom et al., 2016). Screening for homologs of experimentally validated effectors was conducted using BLASTp with a threshold of 30% alignment identity and 1^–10 E-value, on a database of 515 T3SS effectors involved in pathogenesis or symbiosis (Hu et al., 2017). The 1,538 hits were analyzed with the EffectiveT3 prediction tool (Eichinger et al., 2016) resulting in 464 putative effectors with a prediction score > 0.1. The list was further refined by removing hits in strains lacking any kind of T3SS type and hits resulting from homology with effectors that yielded homologs with low identity in > 90% of screened Burkholderia s.l. strains.

Burkholderia vietnamiensis LMG10929 (Bv) and Paraburkholderia kururiensis M130 (Pk) were cultured using Luria’s low salt LB (LBm; Sigma-Aldrich). Fragments of Bv’s hrcV and Pk’s hrcC genes were inserted into a pGEM-T Easy vector (Promega) using specific primers (hrcV-fwd: 5′-GTTCATCGTGTCGGTGGTC-3′, hrcV-rev: 5′-GAACGACCATTTCGGGAAG-3′, hrcC-fwd: 5′-GTTGAAGGCAAAGATCTCAAGG-3′, hrcC-rev: 5′-CGCCACCATTACCACCTC-3′) and cloned into heat-shock competent E. coli JM109 cells. Positive cells were multiplied, and their plasmid extracted using the Wizard Minipreps DNA purification system (Promega). The gene fragments were liberated from the plasmid by an EcoRI digestion and separated on a 1% agarose gel followed by a purification using a NucleoSPin Gel and PCR Clean-up kit (Macherey-Nagel). The target vector pSHAFT2 (Shastri et al., 2017) carrying a chloramphenicol (Cm) resistance gene was opened by an EcoRI digestion and treated with an alkaline phosphatase. Insert and plasmid were mixed in a 5:1 ratio regarding their respective molecular size and ligated using a T4 DNA ligase (Promega). The resulting vectors, pSHAFT2-hrcC and pSHAFT2-hrcV, were cloned into heat-shock competent E. coli CC118 (λpir) which were to be used as donor cells during the subsequent triparental mating process (Supplementary Table 3). Spontaneous streptomycin (St) resistant Bv and Pk mutants were used as acceptor cells. Donor, acceptor (i.e., Pk-St^R and Bv-St^R) and E. coli DH5α carrying a pRK2013 helper plasmid were grown overnight (in liquid LB at 37°C for E. coli species with 50 μg.mL^–1 kanamycin for helper strain and 50 μg.mL^–1 chloramphenicol for donor strain; in liquid LBm at 28°C with 100 μmol.L^–1 streptomycin for Bv and Pk). The cultures were spun down and the cells were concentrated to OD₆₀₀≈50. Acceptor, donor and helper cells were mixed in a 10:1:1 ratio and 50 μL spots were spread out on square LBm plates. The plates were incubated at 28°C for 12 and 48 h for Bv and Pk, respectively. The bacterial spots were then resuspended in liquid LBm before being plated on LBm plates supplemented with 100 μg.mL^–1 streptomycin and chloramphenicol to counter select for the presence of E. coli cells and acceptor cells having not integrated the plasmid. To validate the insertion of the plasmid through homologous recombination at the desired T3SS sites, we checked for the positive amplification of the cat2 Cm-resistance gene and the negative amplification of a set of primers binding upstream and downstream of the targeted area.

Oryza sativa subsp. japonica cv. Nipponbare plants were cultured as described in King et al. (2019). Briefly, seeds were sterilized using successive 70% ethanol and 3.6% sodium hypochlorite treatments and germinated seedlings were transferred onto sterile perlite in an airtight hydroponic system. Plants were grown for 5 days before being inoculated with Pk-St^R, Bv-St^R, Pk-ΔhrcC, and Bv-ΔhrcV. Each individual hydroponic system contained 5 plants. Each plant was inoculated at the stem basis with 1 mL of water containing 1 × 10⁸ bacterial cells. Control plants were inoculated with 1 mL of water. Two systems (2 × 5 plants) were prepared for each condition and time point. At 7 and 14 dpi, the five fittest looking plants were harvested for each condition. The entire root system was sampled and transferred to a screw cap tube containing 1 mL of sterile water and a sterile ceramic bead. Roots were weighted and then pulverized using a Fastprep-24 (MPbio) at 6 m.s^–1 for 40 s. A serial dilution of the resulting solution was spotted out on square LBm plates supplemented with 100 μg.mL^–1 streptomycin and incubated at 28°C for 48 h before counting CFU.

The screening for injectisome coding regions within 146 Burkholderia s.l. strains, representing 139 different species yielded a total of 134 T3SS clusters distributed in 84 strains. Our selection of four core proteins, i.e., SctC, SctN, SctN, SctV, concatenated into a 1652 amino acid alignment and used for the phylogenetic reconstruction resulted in the clustering of Burkholderia s.l. T3SS into seven different clades (Figure 1 and Supplementary Table 4). Five of those clades cluster together with reference T3SS which were included to infer homology to validly described T3SS, i.e., Hrp-1, Hrp-2, SPI-1, SPI-2, and Ysc (see Experimental procedures). No Burkholderia s.l. T3SS is phylogenetically close to the Rhizobiales- or Chlamydiales-like T3SS. However, two additional categories distinctly appear which do not cluster with any reference sequence (named BCI and PSI, described in detail below). One unique T3SS, belonging to Paraburkholderia madseniana RP11 is phylogenetically isolated (Figure 1) and could potentially represent an eighth T3SS category.

We analyzed conservation levels of T3SS-related operon synteny within and between the different phylogenetic types found amongst Burkholderia s.l. strains (Figure 2). The thirteen core components are consistently found in every T3SS type. The overall chromosomal organization of the T3SS coding operons is strongly varying from type to type. Within T3SS types however, there is little variation even between phylogenetically distant strains sharing the same type of injectisomes, e.g., Ralstonia solanacearum CFBP2957 and Paraburkholderia kururiensis M130 (Supplementary Figure 1). The new clusters identified in the previous section (BCI and PSI) display a distinct chromosomal organization from the remaining T3SS types (Figure 2).

Based on phylogenetic analysis and operonic organization, we observed two novel injectisomes. The first one, which we refer to as Paraburkholderia Short Injectisome (PSI), is dominantly present in Paraburkholderia with 11 representative strains. Additionally, two Caballeronia species harbor this new type as well as Burkholderia hayleyella (Figures 1, 3). Its existence was already reported in studies on P. phytofirmans which insisted on its atypical chromosomic organization (Mitter et al., 2013; Angus et al., 2014). Compared to other injectisome families present in Burkholderia s.l. the regulon coding for this peculiar T3SS is remarkably short with only 12–14 kb and containing 14–16 genes. Although most structural sct components are present, we could not identify sctF, the canonical gene coding for the needle-forming component. Still, several genes with unidentified functions within the operon could fulfill this function as they are in the size range of previously reported sctF genes. The needle component sctF is extremely variable (Lombardi et al., 2019) and is usually undetectable by simple homology approaches. To confirm its presence would require a thorough and dedicated investigation, including protein structure analyses and functional validation (Lombardi et al., 2019).

The second uncharacterized T3SS family, which we refer to as Burkholderia cepacia complex Injectisome (BCI), is almost ubiquitously present among species of the Burkholderia cepacia complex (BCC). It is further found in four Trinickia and seven Paraburkholderia species (Figures 1, 3). The operonic structures coding for this injectisome contains the thirteen core genes. A confident identification of the needle component sctF proved here again to be unsuccessful. Still, both BCI and PSI types have been transferred independently to multiple species and been conserved throughout speciation (Figure 1). This strongly suggests that the PSI and BCI remain functional and that selection favored their conservation.

The particularity of the BCI is the presence of a large > 10 kb element separating the sctR and sctQ genes from the rest of the operon (Figure 2). The protein encoded by this large element is predicted to possess a peptidoglycan binding LysM domain. An adjacent gene is encoding a putative lytic transglycolase, allowing peptidoglycan degradation. Another interesting feature is the upstream luxR-type regulator which could hint toward a quorum sensing dependent activation of the operon.

By a homology search using the concatenated sequences also used in the T3SS phylogeny, we verified if the PSI and BCI types are present outside Burkholderia s.l. species. For the PSI type, the closest hits had a similarity of less than 50% sequence homology and a divergent chromosomal organization. For the BCI, however, there were several hits with > 55% sequence homology and a concordant operon structure. These BCI are mainly found in Variovorax species but also in other families of the Burkholderiales order which it seems restricted to Supplementary Figure 2.

The last new T3SS presents no taxonomical closeness to any model used in our phylogenetical reconstruction where it is placed at the interface between the BCI and the PSI (Figure 1). It is found in P. madseniana RP11 and also has a chromosomal organization that diverges from the remaining types found in Burkholderia s.l. (Figure 2). The available genome for this strain is relatively fragmented (393 contigs) and the T3SS operon is split. Nine out of thirteen core genes could nonetheless be detected and do not show signs of degeneration suggesting that these divergent genes remain active. However, as we do not have any diversity supporting the description of this non-canonical T3SS type we will not further insist on the characterization of this putative injectisome.

Among the Burkholderia s.l. representatives tested here, 58% possess a T3SS. However, there are great disparities between genera. Indeed, only six T3SS-positive strains were detected amongst 31 Caballeronia species (Figure 3). Several species belonging to this genus were described in endosymbiotic associations with insects or plant leaves (Miller, 1990; Kikuchi et al., 2005) but none are T3SS-positive.

It is not uncommon for Burkholderia s.l. species to bear two or more T3SS clusters (22% of strains). For instance, out of the six validly described Trinickia species, four have three injectisome-coding clusters of different types.

We attempted to infer correlations between the T3SS type carried by a strain and its ecologic isolation source. The presence of a SPI-1 type is strongly correlated with a strain’s isolation from a clinical setting as it is present essentially in the Burkholderia mallei complex (Figure 3 and Supplementary Table 2). While species of the BCC are also frequently described as potential human opportunistic pathogens, they are equally present in association with plants, in the rhizosphere or as environmental strains. The BCI which is the most frequent T3SS among BCC strains, is dominantly found in environmental or plant-associated bacteria (Figure 3 and Supplementary Table 2). The Hrp-2, which has a demonstrated role in plant-interaction, is also occasionally found in clinical species but dominantly in environmental and plant-isolates. Finally, the PSI type T3SS was never detected in clinical isolates among the surveyed Burkholderia s.l. strains. Furthermore, 10 out of 16 legume nodulating Paraburkholderia and Trinickia species lack a T3SS (Figure 3).

The Ysc type injectisome is rare in Burkholderia s.l. species (Figure 3). Beside Robbsia andropogonis it is restricted to three species which share amoeba as a common host (Haselkorn et al., 2019), Paraburkholderia agricolaris and the genetically distant Burkholderia bonniea and B. hayleyella (Figure 3). Unlike the remaining T3SS types they share, the Ysc appears to be well conserved among these 3 species (Figure 1).

There are great disparities in the distribution of the different T3SS types between different Burkholderia s.l. species (Figure 3 and Supplementary Figure 3). The Hrp2 distribution along every genus, and especially in the early diverging genera Mycetohabitans and Robbsia, suggests that it could have been present in the common ancestor of the Burkholderia s.l. genera, subsequently lost in different species and regularly reacquired, supporting its central role in several species. The second most abundant injectisome is the newly described BCI type which is present in almost every species of the BCC and Trinickia spp. as well as several Paraburkholderia species. The BCI sequences recovered from Burkholderia s.s. species form a single clade (Figure 1), suggesting not only a single emergence event in this group, but also its maintenance all along the diversification of this genus, reflecting its evolutionary importance. Both SPI-1 and PSI injectisomes are more occasional and are almost completely restricted to the Burkholderia mallei complex (BMC) and a Paraburkholderia sub-clade, respectively. The three remaining T3SS types, SPI-2, Hrp-1, and Ysc are only sporadically present in Burkholderia s.l. species.

Many Burkholderia s.l. species harbor two or more different injectisomes which always belong to different types except for Hrp-2. Eight of the ten species bearing two Hrp-2 have one copy closely related to the Hrp-2 of R. solanacearum suggesting the existence of Hrp-2 subfamilies with potential diverging functions.

We further verified if the complex gain and loss dynamics of T3SS are also displayed at the intraspecific level. Many Burkholderia s.l. strains were only recently described and are consequently represented by a single whole-genome sequence. Still, several species with either clinical importance or peculiar ecologies and functions (such as nodulation capacity) have received increased attention and had several genomes sequenced.

Using all 87 publicly available genomes, we validated that B. vietnamiensis uniformly possesses a single T3SS of the BCI type (screened strains and T3SS types for the following species are listed in Supplementary Table 5). This was likewise verified for all 11 P. tropica representatives which uniformly display an Hrp-2 type. Trinickia symbiotica is the only representative of its genus lacking a T3SS (Figure 3) and we confirmed that this was consistently valid in all six available strains (Supplementary Table 5). Similarly, the species-wide absence of T3SS was validated for eight P. caballeronis, and five P. silvatlantica strains. P. fungorum is the only representative of PSI-bearing species with more than four representative and sequenced strains. All P. fungorum uniformly possess a single T3SS of the PSI type. For all these species, there is therefore a uniformity of their genomic content of secretion systems.

While B. lata strain 383 was identified to lack any kind of T3SS, screening of the remaining 23 available genomes of this species yielded a more diverse landscape with eight strains being positive for one T3SS (Supplementary Table 5). Still, the absence of injectisome is the main trend in this species. While not a ubiquitous feature, two B. servocepacia strains were detected to have a Hrp-2 injectisome in addition to their BCI. The recently described B. servocepacia species englobes strains that were affiliated to B. cenocepacia but show significant genomic divergence (Wallner et al., 2019). The Hrp-2 was never observed in B. cenocepacia strains despite a wider diversity of available genomes and could thus be a sign of adaptation toward improved interaction with plants that is displayed by B. servocepacia.

Our prediction for B. contaminans using strain LMG23361, however, seems to be the exception among the diversity of strains for this species. Indeed, the Hrp-2/BCI/SPI-1 trinity of injectisomes was not detected in any other of the 35 genomes available for B. contaminans. Only five strains display both an Hrp-2 and a BCI type while the remaining strains present either one of those injectisome types (Supplementary Table 5).

In Burkholderia s.l. species, the Hrp-2, Hrp-1, Ysc, and SPI-2 type encoding genetic regions are lacking several genes coding for effectors and chaperones compared to the pathogenic bacteria where these systems have been described (Figures 4A–D). The same trend can be observed for the SPI-1 type of P. ribeironis STM7296, a legume nodulating species (Bournaud et al., 2017), which also underwent substantial gene loss compared to other SPI-1 carriers such as B. pseudomallei (Figure 4E).

Using a database of 515 validated effectors from animal and plant pathogens but also rhizobia (Hu et al., 2017), we screened for homologies in the genomes of Burkholderia s.l. species. We used the effector predictive tool EffectiveT3 to cross-check the resulting list of putative effectors (Eichinger et al., 2016). This approach yielded hits for 62 Burkholderia s.l. species (Supplementary Table 6). The species for which we detected the most effectors are the well-studied plant pathogens T. caryophylli and R. andropogonis (17 and 16 hits, respectively). We also detect the animal pathogens of the Burkholderia mallei complex, the plant pathogens of the Burkholderia glumae complex and the fungal endosymbionts of the Mycetohabitans genus among the species with the most effectors (Supplementary Table 6). Our survey identified several putative effectors in commensal Paraburkholderia and Trinickia species. Mostly, these candidates share homology with previously validated effectors originating from Ralstonia. However, we observed no homology to nodulation enabling Nop effectors amongst the legume nodulating Paraburkholderia and Trinickia species.

As the Hrp-2 is the dominant T3SS of Burkholderia s.l. species, we screened the genomes used in this study for the presence of Plant Inducible Promoter regions (PIP-box; TTCG[CGT]N(15)TTCG[CGT]N(30,32)[CT]AN(3T), indicative of T3SS-linked genetic regulation (Koebnik et al., 2006). Two to eight of these regulatory sequences were found within the operon coding for the T3SS machinery depending on the species (Figure 2). As expected, the total number of PIP-box matching sequences is significantly higher in species possessing an Hrp-2 type injectisome compared to those lacking this type (Wilcoxon test, p < 1 × 10^–6). However, the three polyphyletic Hrp-2 bearing amoeba symbionts (B. bonniea, B. hayleyella and P. agricolaris) are not enriched in PIP-box sequences despite the presence of the PIP-box binding regulator HrpB, suggesting that their Hrp-2 has drifted from its initial function to another purpose.

Two model species of the Burkholderia group, P. kururiensis strain M130 (Pk) and B. vietnamiensis strain LMG10929 (Bv), are known to form beneficial and intricate relationships with rice. Both heavily colonize the root surface and have been observed as endophytes (Mattos et al., 2008; King et al., 2019). They are potent plant growth promoters which is linked to their ability to transfer fixed atmospheric nitrogen to the plant and to synthetize growth promoting phytohormones (Trân Van et al., 2000; Dias et al., 2019).

Pk and Bv possess a single injectisome of Hrp-2 and BCI type, respectively (Figure 3). We constructed insertion mutants targeting the sctC and sctV genes in Pk and Bv, respectively (Figure 1). The deletion of any of these structural components has repeatedly been used to efficiently inactivate T3 secretion in Burkholderia s.l. species (Muangsombut et al., 2008; Lackner et al., 2011a; Gavrilin et al., 2012; Swain et al., 2017).

We then compared their colonization efficiency to that of WT strains at different time-points (Figure 5). The mutation has no impact on the colonization capacity of rice roots by Pk in the first 2 weeks after inoculation. Bv however, is strongly affected by the loss of its T3SS. At 7 dpi, the mutant’s colonization capacity is only at 13% of the WT stain’s capacity. At 14 dpi the mutant’s population collapses further to 0.75% compared to the WT strain.

T3SS are one major driver of virulence in pathogenic strains and are increasingly described to play a central role in several beneficial associations between bacteria and their host. Besides the mammal pathogens associated to the Burkholderia mallei complex, almost no attention has been brought to the T3SS of Burkholderia s.l. species. These species were described to engage in a multitude of interactions ranging from symbiosis to pathogenicity with a variety of hosts. Burkholderia s.l. species present a strong diversity in T3SS types, with some strains possessing up to four T3SS-coding clusters, suggesting a central role for this macromolecular tool in the interaction capacity of Burkholderia s.l. strains. Such diversity is however not unique in the bacterial kingdom. Hu et al. (2017) had shown in a large survey that 38 genera carried multiple T3SSs, all having independent evolutionary origins (Hu et al., 2017). Among these 38 genera, Burkholderia s.l. has, however, the greatest diversity of T3SS, even though the original sampling did not cover the entire available species diversity. The T3SS plurality could represent a strategy to achieve an ecologic diversity as an adaptation to multiple hosts or to different levels of interaction with a same host. Indeed, several Burkholderia s.l. species have a broad host range that can include organisms from different kingdoms. Naturally, the use of a T3SS is by no means the only mechanism allowing an efficient host colonization. Additional secretion systems involved in the export of proteins are also considerably contributing to the success of a bacteria’s interaction with its host (Tseng et al., 2009). Interestingly, several Burkholderia s.l. species bearing one or more injectisome-coding regions have only been isolated from environmental settings to date. Thus, there are strong chances that an unknown or not yet discovered eukaryotic partner exists for those strains. For strains that have so far been isolated from a single host but bear several T3SS, alternative hosts could exist.

T3SS types display a complex distribution both between Burkholderia s.l. species but also between different strains of a same species, demonstrating how readily these systems can be acquired or lost. The selective pressures forcing these acquisition/loss dynamics could be set by the diversity of hosts these bacteria are set to colonize. Hence, intra-specific host diversity can be suggested by the heterogeneous distribution of T3SS among strains of a same species. Again, we are limited by the available knowledge on Burkholderia s.l. strain ecology to confidently interpret the consequences of intra-specific T3SS heterogeneity.

Among the most frequent injectisomes of Burkholderia s.l. are the BCI and PSI types. Both have been noted before in individual strains and some of their particularities have already been reported (Tomich et al., 2003; Mitter et al., 2013; Angus et al., 2014) but they had never been subjected to a wide scale characterization. Here, we report the distribution and conservation of the BCI and PSI among Burkholderia s.l. species. Both could be involved in non-virulent plant-colonization as suggested by the ecology of the strains these systems are found in. This supposition is corroborated for the BCI by the rice colonization deficiency of a T3SS insertional mutant of B. vietnamiensis LMG10929. Whether it is used for adhesion, as support for biofilm formation or to inject effector proteins, remains to be investigated.

If effectors are secreted through the BCI, they are unlike the effectors which have been identified hitherto for the other T3SS types as we could not detect any of them by homology. Their de novo prediction based on the detection of T3 secretion signals was also unsuccessful. This could be resulting from the limits of these computational approaches when screening for non-canonical effectors. Still, an alternative role to protein secretion cannot be excluded at this point.

The BCI is abundant in species of the BCC which are known to efficiently colonize the rhizosphere. However, these are not described to be phytopathogens. With B. vietnamiensis as leading figure, several BCC members have demonstrated plant growth promotion as they are potent phytohormone producers and nutrient providers (Trân Van et al., 2000; Rojas-Rojas et al., 2018). To date, the most common T3SS of plant interacting bacteria were those of the Hrp-1 and Hrp-2 type as well as the Rhc for rhizobia (Abby and Rocha, 2012).

Here we suggest that the BCI might be an additional major player in plant interaction but for which effectors remain to be investigated. On the other side, and quite unexpectedly, the rice colonization efficiency of the P. kururiensis M130 Hrp-2 mutant was not impacted in the first 14 days of interaction. This result was surprising regarding all the literature on this T3SS type reporting its major role in plant-bacteria interactions. Contrarily to Bv, Hrp-2 positive Paraburkholderia species such as Pk, are often bearing a few homologs of effectors from plant pathogens such as Ralstonia or Pseudomonas. While interpretations of negative results must always be done with care, we can hypothesize that in the specific case of the Pk/Nipponbare association, the interaction does not require the use of an injectisome for colonization. We can further suggest that the T3SS plays little or no role in this precocious stage of plant colonization, but that its absence could rather have a major impact in the later stages, particularly during the establishment of endophytic colonization. The absolute functionality of Pk’s Hrp-2 would first need to be validated to safely interpret the present results, although it seems very unlikely that a bacterium would conserve a 24 kb genomic region showing no sign of degeneration if it was not functioning.

There is still much to discover on the general implication of T3SS in beneficial interactions and specifically for Burkholderia s.l. species which are promising plant growth promoting strains. A focus on the secreted effector proteins, using a combination of in silico predictive tools, transcriptomics, and secretome-proteomics, could provide valuable knowledge on the frontier between pathogenicity and mutualism resulting from the type III secretion.