Comparative genomics and transcriptomic response to root exudates of six rice root-associated Burkholderia sensu lato species

Beyond being a reliable nutrient provider, some bacteria will perceive the plant as a potential host and undertake root colonization leading to mutualistic or parasitic interactions. Bacteria of the Burkholderia and Paraburkholderia genera are frequently found in the rhizosphere of rice. While the latter are often described as plant growth promoting species, Burkholderia are often studied for their human opportunistic traits. Here, we used root exudate stimulation on three Burkholderia and three Paraburkholderia strains isolated from rice roots to characterize their preliminary adaptation to the rice host at the transcriptomic level. Instead of the awaited genus-dependent adaptation, we observed a strongly species-speciﬁc response for all tested strains. While all bacteria originate from the rice environment, there are great disparities in their levels of adaptation following the sensing of root exudates. We further report the shared major functions that were diﬀerentially regulated in this early step of bacterial adaptation to plant colonization, including amino acids and putrescine metabolism, the Entner-Doudoroﬀ (ED) pathway as well as cyclic diguanylate monophosphate (c-di-GMP) cycling.

Both Burkholderia and Paraburkholderia are described as efficient rhizosphere colonizers and can achieve plant growth promotion through diverse strategies including nutrient solubilization, phytohormone production and interference, siderophore production as well as enhancing the plants tolerance to abiotic stresses. Several Paraburkholderia species are free living nitrogen fixers and some induce nitrogen-fixing nodules on legume plants (Dall'Agnol et al., 2017;Estrada-de los Santos et al., 2018). Further Paraburkholderia species are root endophytes and enhance the growth of a number of major crops such as corn, rice and sugarcane (Perin et al., 2006;Govindarajan et al., 2008;Mattos et al., 2008;Mitter et al., 2013). While several Burkholderia species e.g., B. glumae, B. gladioli, B. plantari, are responsible for serious diseases in rice crops (Kim et al., 2021), others seem to have an intrinsic ability to colonize roots which is independent of plant genotype and do not necessarily cause disease (Vidal-Quist et al., 2014). B. cenocepacia or B. vietnamiensis for instance have been observed to cause plant tissue water-soaking in onion but have otherwise been isolated from healthy plants (Engledow et al., 2004;Jacobs et al., 2008). While Burkholderia strains are mostly reported for their biocontrol efficiency, some achieve plant growth promotion through direct hormonal stimulation and nutrition (Trân Van et al., 2000;Batista et al., 2016;Wang, Liu, et al., 2016;Araújo et al., 2016;Mullins et al., 2019;Chávez-Ramírez et al., 2020). Genome mining approaches underlined the importance of secondary metabolites in Burkholderia species, including an important diversity of antimicrobial metabolites (Navarro et al., 2019). However, in contrast to Paraburkholderia, several Burkholderia species were found to colonize animal hosts in addition to plants, causing opportunistic infections and severe pathologies in humans (Galyov et al., 2010;Sousa et al., 2011).
Here, we investigated the early strategies directing plant colonization at the onset of the interaction between rhizobacteria and their host which is mediated by the perception of root exudates (RE). Through root exudation, plants can use up to 20-40% of their photosynthetically fixed carbon to actively shape their microbiome using chemical attractants, selective carbon sources and diverse restrictive metabolites (Badri & Vivanco, 2009). Primarily genes involved in chemotaxis and motility, but also nutrient transport and metabolism are required for an efficient response to exudates. Still, as exemplified by several past studies, a bacterium's response to exudates is highly host dependent. For instance, Pseudomonas aeruginosa was shown to have a varying transcriptomic reaction to RE originating from two different sugar beet cultivars (Mark et al., 2005). The perception of RE can induce major metabolic adaptation in bacteria. Rice RE were found to be responsible for the differential expression of 4,4% genes in the genome of Azoarcus sp. BH72 (Shidore et al., 2012). More significant, a legume nodulating strain of Paraburkholderia phymatum differentially regulated 19.2% of its genome in response to RE of Mimosa pudica (Klonowska et al., 2018). Upon inoculation with maize RE, Bacillus amyloliquefaciens displays a two-stage response. At 24h posttreatment, the bacteria directs its metabolism towards cellular proliferation while at 48h post-treatment, most differentially expressed (DE) genes are down-regulated and production of extracellular matrix is favored (Fan et al., 2012;Zhang et al., 2015). Typically, the bacterial reaction to RE involves metabolic genes for sugar and amino-acid import and catabolism. However, functions involved at later steps of the interaction such as type 3 secretion are modulated as well upon RE perception (Hassan & Mathesius, 2012;Klonowska et al., 2018).
The present work focuses on six rice root-isolated strains evenly distributed between Burkholderia and Paraburkholderia genera. Beside the two model bacteria for plant growth promotion Burkholderia vietnamiensis LMG 10929 (BvLMG10929) and Paraburkholderia kururiensis M130 (PkM130), we analyzed four new rice-isolated bacterial strains, for which we provide whole-genome sequences, as well as an assessment of their rice root colonization capacity. We hypothesized that the genetic frontier between the two bacterial taxa would influence and differentiate their adaptation strategies towards plant colonization in a genus specific manner. With this in mind, we aimed at revealing each strain's adaptation signatures to the plant environment by exploring their gene expression profiles when sensing rice root exudates using a transcriptomics approach. Considering the results presented here, no distinction can be made between bacteria according to their genus as each strain displayed a specific reaction upon plant root exudate sensing, with no significant pattern clustering con-generic bacteria, while the regulation of several metabolic functions was shared among all bacteria.

Results
Genomic analyses of new Burkholderia and Paraburkholderia species Figure 1 -Genome clustering of studied strains (in bold) with their closest species neighbors. The tree was constructed using the Genome clustering tool of the Microscope plateform (https://mage.genoscope.cns.fr/microscope/). Distances between whole genomes was computed using Mash (Ondov et al., 2016) and plotted using the rapid Neighbor-Joining method (Simonsen et al., 2008). The displayed distances are correlated to average nucleotide identity (ANI) such as D ≈ 1-ANI. Genbank genome accession numbers of strains under study are indicated in Table 1. The newly isolated strains described in this study (Table 1) were compared to genomic databases of Burkholderia sensu lato (s.l.) and the genomic distance to their closest relative was computed ( Figure 1). Whole genome comparisons using average nucleotide identity (ANI) estimations showed a 98.52% score (80.14% sequence coverage) when comparing strain ABIP441 (hereafter BrABIP441) to the proposed species Burkholderia reimsis sp. nov. BE51 T (Esmaeel et al., 2018), lying above the species threshold (> 95% identity on more than 70% cover).
We had previously proposed that ABIP444 (hereafter BoABIP444) belongs to a novel species, B. servocepacia sp. nov., closely related to Burkholderia cenocepacia . A recent publication updated our proposal to validate this separate species as B. orbicola, with TatI-371 as the type strain (Morales-Ruíz et al., 2022). Strain ABIP659 (hereafter PspABIP659) displays no determining similarity to any described Burkholderia s.l. species. ANI analyses suggest that it is most closely related to Paraburkholderia tropica without exceeding the required threshold to infer conspecificity (91.86% identity and 76.0% sequence coverage to P. tropica LMG 22274 T ). At the genome level, ABIP630 (hereafter PsABIP630) shares 98.5% ANI (>70% aligned) with P. sabiae LMG24235 T and can thus be considered as belonging to this species.

All isolates demonstrate robust rice colonization capacity
As all strains were isolated from the rice rhizosphere (Table 1), we validated their ability to colonize this environment de novo ( Figure 2). Plants growing in a sterile hydroponic environment were inoculated with each strain individually and the root-associated strains were enumerated by plate counting after 3, 7 and attachment and colonization.  Baldani et al., 1997 Most bacteria reach their highest density at 7 days post inoculation (dpi) and the ratio of bacterial cells to root mass then decreases. The three Burkholderia spp. all have very similar colonization efficiencies and development profiles over time. Paraburkholderia spp. in turn, have more individual colonization approaches. PsABIP630 displays the strongest abundance on roots while PspABIP659 is the overall weakest colonizer but has the greatest dynamic in colonizing populations. Indeed, PspABIP659 shows a striking 6fold burst in population from 3 dpi to 7 dpi. For all strains, we evaluated the impact of inoculation on Oryza sativa cv. Nipponbare growth (see Material and methods). No impact was observed on root and leaf weight at 60 dpi compared to the mock treatment (Supplementary Figure S1). This is consistent with results from similar treatments using PkM130 and BvLMG10929 . Nipponbare rice seedlings were inoculated at 5 days post-germination with 1.10 8 bacterial cells and grown in a hydroponic medium. Total roots were sampled at different time points (5 replicates each) and the attached bacterial population was enumerated. The trendlines (red) were inferred using the non-parametric regression LOESS method (Cleveland et al., 2017).

Paraburkholderia and Burkholderia species share 2443 genes
An orthology search was performed to infer levels of similarity between the three Paraburkholderia strains and the three Burkholderia strains (Figure 3). While all genomes are of relatively large size compared to their Proteobacteria congeners (Table 1), there are some discrepancies between them. The two ricegrowth promoting model endophytes PkM130 and BvLMG10929 have genomes of comparable size which are however the smallest among the cohort analyzed here. The four remaining species have all been isolated from the rice rhizosphere, and their genomes are considerably larger (15-20%) than those of the endophytic PkM130 and BvLMG10929.
Altogether, 2443 genes are conserved among the six strains, representing 34% to 44% of their total gene sets ( Figure 3). Still, given the present strain selection, PspABIP659 and PsABIP630 are best defined by their specific genome which is larger than any other intersection. The remaining strains also have a considerable number of specific genes although smaller than the core-genome shared between the six strains. Taken together, the Burkholderia specific genome is 48% larger than its counterpart in Paraburkholderia species.

The global transcriptomic response to root exudates is strain-specific
In order to unravel bacterial genes involved in the preliminary steps of plant interaction, we stimulated the six Burkholderia s.l. strains with Nipponbare rice RE and compared their transcriptomic response to that of bacteria grown in VSG medium using three replicates by strain and condition (see material and methods). A preliminary validation of the sequencing procedure confirmed that ribodepletion efficiency was > 70% for every strain and condition except for BvLMG10929 which only attained ~50% efficiency (Supplementary Table S1). Nonetheless, high enough read numbers mapping against a database of CDSs were obtained for robust bioinformatic analyses (Supplementary Table S1). Furthermore, we verified that the exposition to RE significantly and consistently affected the bacteria's transcriptomic responses throughout the replicates (Supplementary Figure S2).
At a quantitative level, the general response of bacteria to RE was strain-specific with large variations between strains. Besides BoABIP444, all strains display a majority of negatively regulated genes upon root exudate stimulation ( Figure 4). For PspABIP659, the amount of down-and up-regulated genes is close to identical and represents the strongest response in terms of number of genes involved which is 8-fold stronger than for the weakest responder BrABIP441. We further undertook an analysis of the conserved transcriptomic response across strains. Despite the level of genetic conservation between the six strains ( Figure 3), the core-genome is not used in the primary steps of plant sensing as only two genes are DE in all six bacteria ( Figure 5). One codes for an outer membrane porin (Table 2 & Supplementary Table S2) without further characterization. It is down-regulated in PsABIP630 and up-regulated in all remaining strains. The second encodes a TonB protein (Table 2 &  Supplementary Table S2) which is associated with outer membrane receptors ExbB and ExbD. The differential expression of exbBD is conserved in all strains besides BrABIP441 and follows the trend of the respective tonB regulation in each strain. It is up-regulated in PspABIP659 and BvLMG10929 and downregulated in PsABIP630 and PkM130. If we exclude BrABIP441 from the analysis, a final gene is detected to be up-regulated in every strain, the Entner-Doudoroff pathway enzyme coding gene edd (Table 2 &  Supplementary Table S2).
Beside the intersection between PspABIP659 and PkM130, the levels of co-regulation are often higher in inter-generic than in intra-generic duos ( Figure 5). The levels of conserved regulation within Burkholderia and Paraburkholderia trios is noticeably low and dominantly differential in Paraburkholderia ( Figure 5).

The transcriptomic response is mainly nutrient directed
The classification according to COG categories of the DE genes comforts the hypothesis that each strain has a dominantly individual response towards RE ( Figure 6). Still, some trends can be inferred. In each strain, most DE genes involved in carbohydrate transport and metabolism are upregulated, indicating that the strains are taking advantage of the increased source of carbon. The reaction to other nutrients such as amino acids, lipids and ions is more disparate from strain to strain. Still, it appears that Paraburkholderia strains are globally more sensitive to the presence of lipids contained in RE. Our experimental approach used RE that were recovered from rice seedlings having grown for 5 days in a 6-fold diluted Hoagland nutritive medium (Supplementary Table S3). This experimental setting has been used before to assess the plant's transcriptomic response elicited upon bacterial inoculation . However, diluted nutrients contained in this medium were assimilated to the recovered RE fraction. Thus, the bacterial response concerning elements contained in the hydroponic medium must be considered with care. As it was designed for plant nutrition the hydroponic medium does not contain any source of carbon. It does however hold macro-nutrients such as nitrogen, phosphorus and potassium. Additional elements eliciting a bacterial response are iron and sulfur. Still, the transcriptomic response to these nutrients varies depending on the bacterial strain (Table 2 & Supplementary Table S2).
Overall, we detected elements of the sulfite reductase complex (cys), the alkanesulfonate transport system (ssuBCD) and the taurine transporter tauABC which can be linked to sulfure nutrition and are DE only in Paraburkholderia strains. Elements linked to both assimilatory and respiratory nitrate transport and reduction were DE in several strains (Table 2 & Supplementary Table S2). Concerning iron nutrition, whole NRPS clusters, siderophore receptors and bacterioferritins are dominantly negatively regulated in most strains.
While nitrogen is a main constituent of the hydroponic medium, it is present as ammonium-conjugates and not nitrates. An increased nitrate metabolism was reported upon stimulation of PkM130 with rice macerate (Coutinho et al., 2015). The same study further reported the upregulation of a transporter complex involved in sulfate acquisition (ANSKv1_70577-81) which is also upregulated in the present study. Furthermore, the iron uptake complex is comparably downregulated in response to rice macerate (Coutinho et al., 2015) and what we observe with RE. These observations attest that the supposedly medium-induced regulations described above could in fact result from the composition of RE.

B. vietnamiensis LMG10929
B. reimsis ABIP441 Figure 5 -Specific and conserved regulation between strains upon RE stimulation. We calculated the conserved genome between different groups of strains to determine which genes are commonly regulated (either positively (red), negatively (blue) or differentially (grey) between each pair of strains, the Burkholderia and Paraburkholderia trios, all 6 strains and all 6 except BrABIP441. Single dots represent the regulation of strain-specific genes. For the PspABIP659 specific response, the histogram was reduced to fit with the figure (total regulated genes = 561, with a similar number of up and down genes)

Amino acid metabolism is activated upon exudate sensing
Several strains respond to the presence of amino acids from the exudates by downregulating the synthesis of dedicated transporters consistent with the strong adaptation of amino acid metabolism in response to RE which was repeatedly reported (Mark et al., 2005;Fan et al., 2012;Yi et al., 2017). PkM130, BvLMG10929 and PspABIP659 all reduce the transcription of glutamate, aspartate and branched chain amino acids transporters. PspABIP659 further reacts in a similar fashion to the presence of arginine, lysine and histidine (Table 2 & Supplementary Table S2). The latter is also eliciting a response in PkM130. Upregulation of amino acid transporters was only observed for methionine in PkM130 and glycine in BoABIP444 (BCEN4_v1_340003).
Interestingly, an operon coding for a sarcosine oxidase (soxBDAG), a serine deaminase (sdaB) and an aldolase (folB) is up-regulated in all Paraburkholderia strains and in BoABIP444 but not in the remaining Burkholderia despite them bearing homologues of the concerned genes (Table 2 & Supplementary Table  S2). Sarcosine is an intermediate in glycine synthesis which is generated by the sarcosine oxidase with hydrogen peroxide as a byproduct. The serine deaminase converts serine but also threonine into pyruvate and thus participates in gluconeogenesis. The FolB aldolase is an enzyme involved in one step of tetrahydrofolate synthesis, a cofactor used by the sarcosine oxidase.

Exudates have a limited impact on chemotaxis and motility in our experimental design
Flagellar synthesis was not significantly impacted in our experiment. The chemotaxis pathway transducing elements cheA, cheY and cheB were equally devoid of regulation in all strains except PspABIP659 where cheA, cheB but also cheW, cheD and motB are negatively regulated (Table 2 &  Supplementary Table S2). In other strains, chemotaxis elements present a more scattered regulation. The steadiness of genes involved in cell motility might be explained by their already high base level expression. The base mean of the flagellin component fliC reaches up to hundreds of thousands of reads depending on the strain. While, motility is usually activated upon root exudate sensing (Klonowska et al., 2018), in the present setting bacteria may have had already reached their induction maxima.

Root exudate perception induces a shift in glycolysis strategies
The strains display numerous and specific reactions to the presence of carbohydrates from RE. Several strains positively regulate their dedicated metabolism to the presence of xylulose, ribose and glucarate (Table 2 & Supplementary Table S2).
A more conserved metabolic shift was however observed for glycolysis processes. In all strains except BrABIP441, genes involved in gluconate import and metabolism are up regulated. These enzymes are part of the Entner-Doudoroff (ED) pathway, an alternative to EMP glycolysis with similar costs and outcomes ( Figure 7A). We previously demonstrated that ED pathway mutants of PkM130 and BvLMG10929 were significantly affected in their capacity to colonize rice roots (Wallner et al., 2022). The upregulation of the ED pathway by PkM130 and BvLMG10929 in the present study underlines its involvement in plant colonization as one of the early metabolic adaptations. Genes, downstream of the Entner-Doudoroff pathway, coding for a sugar ABC transporter are up-regulated in every strain except PspABIP659 where they are repressed.

The perception of putrescine induces strain-specific adaptations
We paid particular attention to genetic elements that could indicate the onset of biofilm production as these structures are often involved in surface colonization. We noticed that RE were impacting the expression of genetic pathway involved in putrescine metabolism and cyclic diguanylate monophosphate (c-di-GMP) cycling. Upon stimulation with RE, we observe three different types of reactions to putrescine. PkM130, PspABIP659 and BvLMG10929 have a conserved response consisting in the downregulation of the previously mentioned genes. In PkM130 and PspABIP659, genes involved in putrescine formation from the precursor agmatine are also downregulated. Finally, BoABIP444 displays a fragmented regulation of the putrescine metabolism pathway but with systematic up-regulation of the respective genes (Table 2 & Figure 7B). The remaining strains, while they possess the according genes, do not adapt the pathways regulating the levels of intracellular putrescine. The Entner-Doudoroff pathway uses glucose as starting product and branches into the glycolysis pathway with glyceraldehyde-3P which is ultimately converted to pyruvate. No significant regulation was found in B. reimsis ABIP441. (B) Putrescine synthesis, catabolism and uptake pathways in Burkholderia s.l. strains. No significant regulation was found in B. reimsis ABIP441 and P. sabiae ABIP630.

Cyclic di-GMP signaling is only activated in Paraburkholderia species
Intracellular c-di-GMP levels are regulated by diguanylate cyclases (DGC) for synthesis and diguanylate phosphodiesterases (PDE) for degradation. Burkholderia strains do not differentially express any of their genes involved in c-di-GMP signaling. This is coherent with what was previously observed through a Tn-seq approach in BvLMG10929 where no gene involved in c-di-GMP cycling was required for rice colonization (Wallner et al., 2022). For Paraburkholderia strains, PkM130 shows the most moderate response with the downregulation of a single DGC. PsABIP630 downregulates one DGC and upregulates one PDE. The strongest response is observed for PspABIP659 which differentially expresses eleven genes involved in cdi-GMP cycling. The majority of DGC are upregulated and the only two regulated PDE are repressed. Hence, there are different strategies at play for Paraburkholderia strains upon sensing of RE concerning the adaptation of their c-di-GMP signaling. PkM130 and PspABIP659 elevate the intra-cellular levels of the secondary messenger while PsABIP630 decreases them. Again, these results correlate with previous observations for PkM130 where a mutant involved in c-di-GMP cycling had a severe defect in rice colonization (Wallner et al., 2022).

Discussion
In this study we analyzed the transcriptomic response to RE of six strains belonging to Paraburkholderia and Burkholderia. Given the ecologic background of both taxa we hypothesized that genera-dependent adaptation patterns could appear in the transcriptomic regulation upon sensing of rice RE. Burkholderia and Paraburkholderia strains share a close taxonomic proximity and thus a substantial level of orthologous genes. However, their strain-specific genomes are also of considerable size which is a premonitory sign of the disparate transcriptomic response displayed by the different strains upon chemical stimulation with RE. No categorizing trend could be inferred by the amount and intensity of regulated genes or by the function of regulated genes. Each strain displays a dominantly individual response and the genetic proximity between strains within a genus is not reflected in their transcriptomic response. Thus, in the present cases, the adaption to the habitat is independent from the genetic background.
Care was taken to expose the different bacteria to RE for a duration that was relative to their respective metabolic rate. The purpose was to obtain a synchronized response for reliable comparison. It is however conceivable that the six strains differently prioritize their responses which could result in an overall similar strategy that cannot be detected if sampled at a single time point. To avoid such biases induced by temporal shifts, a transcriptomic kinetic analysis would have been required. However, given the extent of differences observed here, we feel confident that there is indeed a strain-specific response to RE.
As the reported observations were made in a gnotobiotic setting, they are unlikely to reflect the full spectrum of bacterial adaptations in a complex natural environment. Still, species of Burkholderia s.l. were repeatedly described to be dominant in the rice rhizosphere which suggests that they can exist in communities where they exert the most influence beside their host (Ikeda et al., 2014;Yu et al., 2018).
Putrescine represents a long-overlooked metabolite in plant-bacteria interactions. While it has been identified as a virulence factor in Ralstonia solanacearum (Shi et al., 2019) it is also involved in the signalization between plant beneficial bacteria and their host (Liu et al., 2018). Putrescine is recognized as an important signaling molecule in plants where it is involved in various situations from organ development to biotic stress resilience (Tiburcio et al., 2014). Its presence in RE has been reported for tomato and rice (Kuiper et al., 2001;Suzuki et al., 2009) but has otherwise rarely been investigated. A study on the fitness determinants of Pseudomonas sp. WCS365 in the rhizosphere of wild type Arabidopsis thaliana and immunity impaired mutants demonstrated that the absence of putrescine catabolism genes (spuC, pauC and gabT) were highly increasing the bacteria's fitness in the rhizosphere of WT but not immunocompromised plants (Liu et al., 2018). BvLMG10929 transposon mutants affected in putrescine metabolism and uptake were also detected to be less competitive during rice root colonization (Wallner et al., 2022). It is suggested that putrescine might act as a signaling molecule, informing the bacteria of the presence of a eukaryotic host. Responsive bacteria in turn promote biofilm formation which favors attachment to the root. This regulatory system must be extremely fine-tuned as to not trigger plant defenses by an excess production of biofilms and other physiological changes subsequent to putrescine signalization. In the transposon sequencing assay on Pseudomonas sp. WCS365, genes involved in putrescine uptake and synthesis were found to be negative fitness regulators (Liu et al., 2018). Consistently, PkM130, PspABIP659 and BvLMG10929 downregulate the according gene homologues upon exudate sensing. Thus, we could hypothesize that these strains take advantage of putrescine to regulate their metabolism and physiology towards a controlled surface colonization of roots.
C-di-GMP is another signal molecule positively regulating biofilm formation when accumulated (Romling et al., 2013). The regulation of enzymes involved in its metabolism can therefore be key for a successful colonization of plant roots and evasion of host defenses. We previously demonstrated the involvement of a c-di-GMP cycling enzyme in the root colonization process of rice by PkM130 (Wallner et al., 2022). Inversely, in BvLMG10929 several genes involved in c-di-GMP cycling were identified as being obstructive to a successful root colonization (Wallner et al., 2022). Recently, the importance of c-di-GMP in biofilm production and subsequent virulence on rice was demonstrated for Burkholderia glumae BGR1 (Kwak et al., 2020). The human opportunist B. cenocepacia, which is closely related to B. orbicola, received much attention for the role played by c-di-GMP signalization in virulence by regulation of biofilm formation and motility (Richter et al., 2019). Similarly, the absence of the PDE-receptor coding gene rpfR in Burkholderia lata SK875, reduces the strain's virulence and increase its biofilm production (Jung et al., 2017). RpfR is a receptor for the Burkholderia specific quorum sensing signal BDSF which implies a connection between c-di-GMP and cell-to-cell signaling (Schmid et al., 2017). Further studies linked the production of biofilms with the sensing of nitrate and temperature which are relying on a subsequent c-di-GMP dependent signaling pathway in B. pseudomallei Mangalea et al., 2017). The maize endophyte Burkholderia gladioli 3A12 displays c-di-GMP dependent signaling for biocontrol purposes against fungal pathogens (Shehata & Raizada, 2017). Although root exudate stimulation had no direct impact on the regulation of flagellum-related genes, the regulation of genes involved in putrescine uptake and metabolism as well as those involved in c-di-GMP signaling indicate an indirect and intricate regulation of motility (Liu et al., 2018;Richter et al., 2019). This is coherent with the supposition that bacteria evade host defenses by reducing the production of the common flagellin MAMPs. This behavior was described upon exudate stimulation in Pseudomonas aeruginosa PA01, Azoarcus sp. BH72 and P. kururiensis M130 (Mark et al., 2005;Shidore et al., 2012;Coutinho et al., 2015).
Until recently, the ED pathway had never been reported for its activation at any stage of plant colonization. The difference with the more classic EMP-glycolysis pathway is the generation of NAD(P)H cofactor during ED-glycolysis. This cofactor can be essential for enzymes that play a decisive role in rhizospheric adaptation as it is the case for thioredoxins, involved in oxidative stress tolerance (Cumming et al., 2004;Chavarría et al., 2013). The edd gene, coding for the NAD(P)H-generating enzyme, is amongst the only genes that are differentially regulated in all six strains. In a separate study, we identified edd as being critical for a successful interaction of BvLMG10929 and PkM130 with rice roots as its absence led to a significant decrease in the bacteria's colonization capacities (Wallner et al., 2022).
Using six different strains from the Burkholderia and Paraburkholderia genera, we could find no discerning pattern in their early response to the perception of plants through RE, suggesting that there is a strong intra-generic diversity. Thus, instead of analyzing one Burkholderia and one Paraburkholderia group composed of three species each, we rather have six distinct strains that independently underline the importance of certain pathways in plant-bacteria interactions. While some, such as c-di-GMP signaling were already recognized, the implication of putrescine and the ED-pathway should receive an increased attention and especially in later steps of plant colonization.

Plant cultivation and exudate collection
Oryza sativa ssp. japonica cv. Nipponbare seeds were sterilized, germinated and cultivated in a sterile, hydroponic system as described before . Five seedlings were planted per system and transferred to a growth chamber (28°C; 16 h light; 8 h dark). After 3 days of growth, the hydroponic medium was collected using a sterile graduated pipette and filtered using sterile Millipore membranes to remove residual perlite sand. The resulting medium was frozen at -80°C and lyophilized. Dry RE were resuspended in sterile distilled water at an 80x concentration and stored at -20°C until use.
For plant growth measurements, rice seedlings were sterilized and germinated as before and transferred to pots containing 400 mL of a 7/2/1 mix of sterile sand (0.6-1.6 mm)/ perlite/vermiculite. Five pots containing four rice seedlings were used per condition.

Bacterial strains and cultivation
Burkholderia reimsis ABIP441, Paraburkholderia sp. ABIP659 and Paraburkholderia sabiae ABIP630 were isolated from healthy Oryza sativa subsp. japonica roots. ABIP441 and ABIP630 were isolated from rice roots collected from a field in Cameroon and ABIP659 from a field in Vietnam. Additionally, we used the well characterized rice growth promoting species Burkholderia vietnamiensis LMG10929 and Paraburkholderia kururiensis M130 as well as the recently proposed Burkholderia orbicola sp. nov. (previously named B. servocepacia) ABIP444 Morales-Ruíz et al., 2022).
All bacterial strains were cultured as follows: Glycerol stocks (20%) of bacterial cells conserved at -80°C were plated on Luria's low salt LB (LBm; Sigma-Aldrich) agar plates and incubated for 72 h at 28°C. Single colonies were used to inoculate 10 mL of LBm broth in 50 mL Falcon tubes and incubated for various amounts of time allowing the different strains to reach an OD600=1.0. For DNA extraction purposes, cells were used at this stage.

Plant inoculation assays
For colonization assays, bacterial strains were grown overnight in liquid LBm as described above. Bacterial density was then adjusted to 1.10 8 cfu.mL -1 with regard to the relative amount of cells at OD600=1.0 for each strain. In a sterile setting, 5-day old rice seedlings, prepared as described above, were inoculated with 1 mL of the bacterial solution at the base of each plant. Two systems (10 plants total) were prepared for each condition and grown for 14 days (28°C; 16 h light; 8 h dark). Five plants were randomly harvested per condition as biological replicates. 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 seconds. A serial dilution of the resulting solution was spotted out on square LBm plates and incubated at 28°C for 48 hours.
For plant growth measurements overnight broth bacterial cultures in LBm were pelleted by centrifugation, resuspended in distilled water and adjusted to OD600=1.0. CFU measurements at this OD are reported in the legend of supplementary Figure S1. Twenty plants grown in a sand/perlite/vermiculite mix were inoculated with 1 ml of bacterial suspension per plant, 5 days after seedling transfer to pots. Plants were grown in a growth chamber (80% humidity, 28°C; 16 h light; 8 h dark) for 50 days before sampling, and fertilized with a 50% Hoagland solution, twice a week. The length of the longest leaf as well as the root's and leave's dry weights were measured for each plant. Statistics were performed in R using packages rstatix, multcompView and ggplot2 (Wickham, 2016;Kassambara, 2020;Graves et al., 2022) and differences were estimated with a pairwise Wilcoxon non-parametric test. R Script used to do statistics and produce Supplementary Figure S1 is accessible on bioRxiv at https://doi.org/10.1101/2022.10.04.510755 (Wallner, Klonowska et al., 2022).

Exudate stimulation of bacterial cells
For RNA extraction, cells were prepared as detailed above but using Vincent Succinate Glutamate (VSG) growth medium (Vincent, 1970). The pre-culture was then used to inoculate 50 mL VSG in 250 mL Erlenmeyer flasks at a final OD600=0.075. After 1 h of agitation at 150 rpm and 28°C, 800 µL of concentrated root exudate solution (62-fold) was added to obtain a final exudate concentration that was equivalent to that found in the hydroponic growth system. Cells were then grown until they reached mid-exponential phase. The volume required for 10 9 cells was collected and immediately transferred to a 15mL Falcon tube containing 95% ethanol and 5% phenol. Collected cells were centrifuged at 4000 rpm for 15 minutes. The supernatant was discarded and the pellet was used to extract total RNA using the RiboPure RNA purification kit for bacteria (ThermoFisher). After extraction, the sample were subjected to a TURBO DNase (ThermoFisher) treatment to remove any residual DNA. Extraction efficiency was qualified using the 2100 Bioanalyzer system with an RNA Nano Chip (Agilent Technologies).

Ribodepletion and RNA sequencing
Ribosomic RNA (rRNA) was removed from total RNA extracts following the guidelines of Roux et al., 2018. Briefly, we generated probes targeting 5S, 16S and 23S rRNA incorporating a biotin complex in 5' and locked nucleic acids (LNAs) for improved specificity towards rRNA of the burkholderia group (Supplementary Table S4). An in silico approach validated that these probes are universal to all Burkholderia s.l. species and conserve specificity in the neighboring taxa Ralstonia and Cupriavidus. 2 µL of probes (50 µM) were used in combination with reagents from the RiboMinus Transcriptome Isolation Kit, bacteria (ThermoFisher). We used 2 µg of total RNA and 71 µL of magnetic beads. Prior to use, magnetic beads were washed twice in nuclease free water and once in 2x SSC buffer and resuspended in a final volume of 31 µL hybridization buffer. Ribodepletion efficiency was validated using the 2100 Bioanalyzer system (Agilent Technologies) with an RNA Nano Chip (Supplementary Table S1).
RNA sequencing was performed at the high throughout sequencing platform at I2BC (CNRS, Gif-sur-Yvette, France) using Nextseq500 (Illumina) technology with a Nextseq 500/550 High Output Kit v2 (Illumina) and 75 sequencing cycles (single read, 75 bp) on ribodepleted RNA samples originating from three biological replicates per strain and per growth condition. De-multiplexing, adapter trimming and quality controls were performed by the platform using bcl2fastq2-2.18.12, Cutadapt 1.15 and FastQC v0.11.5 respectively (Andrew, 2010;Martin, 2011