Conserved immunomodulation and variation in host association by Xanthomonadales commensals in Arabidopsis root microbiota

Main

Soil is one of the most species-rich environments¹, and plants growing in this natural substrate are constantly in contact with a high density and diversity of microorganisms. Soil-dwelling microorganisms constitute the principal inoculum of the plant microbiota—that is, the entire community of plant-inhabiting microorganisms, which consists of only a fraction of the soil biome. Despite substantial variation between soil biomes, a similar taxonomic structure is found in the bacterial plant microbiota on roots and leaves, which is dominated by Proteobacteria, followed by Actinobacteria, Bacteroidetes and Firmicutes, collectively termed the core microbiota^2,3,4,5,6,7. Photoassimilates produced by plant hosts are consumed during microbiota establishment and support heterotrophic microbial growth throughout the host’s lifetime^8,9. In return, members of the bacterial plant microbiota improve plant performance by mobilizing mineral nutrients from soil for root uptake^10,11, provide pathogen protection in roots and leaves^12,13,14,15 and increase abiotic stress tolerance^16,17. Numerous host-derived specialized metabolites or defence phytohormones such as salicylic acid, as well as bacterial substrate preferences for root-secreted molecules, influence root microbiota profiles at lower taxonomic ranks^{18,19,20,21,22,23,24}.

The proliferation of invading, pathogenic microorganisms in plants is controlled by a two-tiered innate immune system, with immune responses triggered by families of cell-surface and intracellular immune receptors²⁵. The presence of microbe-associated molecular patterns (MAMPs), which are often conserved among widely related taxa, and host-released damage-associated molecular patterns (DAMPs) are perceived by plasma-membrane-resident pattern recognition receptors (PRRs) in the extracellular space²⁶. The majority of characterized PRRs belong to the families of leucine-rich repeat (LRR)-containing receptor-like kinases (RLKs) or LRR-containing receptor-like proteins (RLPs); the latter are characterized by the absence of a cytoplasmic kinase domain and constitutive interaction with the kinase SOBIR1 (refs. ^27,28). Upon extracellular binding of the respective MAMP or DAMP, most characterized LRR-RLKs form heterodimers with the co-receptor BAK1 (ref. ²⁹). Pattern-triggered immunity (PTI) is associated with a stereotypic array of defence-associated responses that ultimately restrict pathogen proliferation, including Ca²⁺ influx transients³⁰, extensive transcriptional reprogramming³¹, the biosynthesis of the phytohormone ethylene³², NADPH-oxidase (for example, RBOHD)-dependent spikes in reactive oxygen species^33,34, callose deposition³⁵ and metabolic changes³⁶. In axenic Arabidopsis, chronic exposure to immunogenic peptides such as flg22 (a bacterial peptide derived from flagellin that is perceived by the LRR-RLK FLS2) results in a dose-dependent reduction in plant growth³⁵. This is interpreted as a trade-off, in which the host allocates its limited resources towards defence or growth³⁷. Host-adapted phytopathogenic bacteria suppress PRR-triggered immune signalling by delivering typically strain-specific effectors into plant cells via the bacterial type III secretion system (T3SS)³⁸. Intracellular immune receptors either bind to or detect the activity of these pathogen effectors, inducing effector-triggered immunity³⁹.

Higher-order PRR mutants and defence phytohormone mutants in Arabidopsis exhibit altered bacterial community composition^19,40. An intact PTI contributes to preventing Arabidopsis phyllosphere dysbiosis characterized by a higher bacterial load and a shift of the endophytic community⁴¹. This dysbiosis requires not only genetic depletion of multiple PRR complexes but also the MIN7 vesicle trafficking pathway or a specific mutation in a membrane-attack-complex/perforin-domain protein. Experiments with root commensals, including strains of a culture collection derived from Arabidopsis thaliana grown in Cologne agricultural soil (At-SPHERE)⁵, have demonstrated that interference with PTI outputs induced by exogenous application of synthetic flg22 is taxonomically widespread but typically strain-specific within a phylogenetic lineage^42,43,44. In roots, the induction of PTI by treatment with flg22 alters the profile of synthetic communities (SynComs) containing only commensal strains that do not interfere with PTI, whereas the presence of immunomodulatory strains attenuates this effect⁴⁴. SynComs consisting of immunomodulatory strains specifically downregulate a subset of immune-related genes^43,44. However, our understanding of the mechanisms by which members of the microbiota interfere with PTI remains fragmentary⁴⁵. While most commensal bacteria encode flagellin isoforms whose flg22 peptides evade detection by FLS2, others have evolved peptide variants that inhibit FLS2 activation through partial or complete antagonism^46,47. The rhizobacteria Pseudomonas capeferrum WCS358 and Pseudomonas simiae WCS417 lower environmental pH through the production of gluconic acid and its derivative keto-gluconic acid and through amino acid biosynthesis pathways, respectively, resulting in reduced flg22-induced defence marker gene expression^42,48,49. Type II secretion system (T2SS) mutants of the Dyella japonica MF79 commensal are impaired in suppressing PTI-associated defence marker gene expression in roots following acute flg22 treatment, while its T3SS mutants retain wild-type-like immunomodulatory activity⁴³. However, the causative effectors exported through the T2SS of D. japonica and their host targets remain undefined.

In terrestrial environments, Xanthomonadales constitute a core bacterial order in the microbiota of photoautotrophic unicellular and multicellular eukaryotes^50,51. The order is classified into Xanthomonadaceae and Rhodanobacteraceae and comprises both commensal and pathogenic strains⁵². The genus Xanthomonas consists mostly of pathogens infecting various aboveground plant organs, and these cause systemic vascular and local non-vascular diseases in nearly 400 plant species, with vascular infection probably being the ancestral mechanism^53,54.

Using phylogenetic analysis of more than 1,700 Xanthomonadales genomes and strains of microbiota culture collections isolated from diverse photoautotrophic hosts, we show here that immunomodulation by Xanthomonadales is a conserved, ancestral trait that preceded their specialization as host-adapted plant pathogens. We demonstrate that Arabidopsis perceives the presence of Rhodanobacter R179, a root commensal belonging to the deepest branch of the Xanthomonadales, by detecting at least three immunogenic elicitors through the redundant activity of the immune modules EFR and SOBIR1. However, analyses of root transcriptomes and PTI-associated responses show that R179 hides itself and other commensals from host recognition upon prolonged root colonization. We generated and screened a mini-Tn5 mutant library of Rhodanobacter R179 and identified commensal defence suppression system (dss) genes needed for immunomodulation and microbiota establishment in planta. A further mechanism of R179 camouflage involves clearance of extracellular immunogenic peptides by secreted peptidase(s), which probably serves a function in hiding the producer and other microbiota members from host recognition. We discuss the idea that the conservation of immunomodulation by xanthomonads might be related to their adaptation to terrestrial habitats and, together with strain-specific diversification to colonize roots, explains why they represent a core lineage of the plant microbiota.

Results

Immunomodulation is conserved in Xanthomonadales

All 12 Xanthomonadales commensals of the At-R-SPHERE culture collection alleviate root growth inhibition (RGI) induced by flg22 (representative strains are shown in Fig. 1a)^5,44. To determine whether the suppression of this immune response extends to Xanthomonadales strains derived from Lotus japonicus roots, the soil-borne unicellular alga Chlamydomonas reinhardtii or Arabidopsis grown in a US soil with a different inoculum source, we tested 17 additional isolates for their ability to suppress flg22-induced RGI in the Arabidopsis line fls2/pWER::FLS2–GFP, overexpressing FLS2 in the root epidermis (pWER::FLS2; Fig. 1a)^50,55,56,57. These 17 isolates alleviated flg22-induced RGI to similar degrees within 3 weeks of co-cultivation upon direct germination of Arabidopsis seeds on agar matrix containing flg22 (Fig. 1a).

Conserved immunomodulation and variation in host association by Xanthomonadales commensals in Arabidopsis root microbiota — Fig. 1: Phylogenetically diverse commensal and pathogenic Xanthomonadales strains suppress MAMP/DAMP-induced RGI in *A. thaliana* and *L. japonicus.*

Three phytopathogenic strains, Xanthomonas campestris pv. vesicatoria (Xcv 85-10), Xanthomonas campestris pv. campestris (Xcc 8004) and Xanthomonas hortorum pv. pelargonii (Xhp), likewise alleviated flg22-induced RGI (Fig. 1a). Xcc and Xhp moderately inhibited root growth even in the absence of flg22 in our experimental setup (Fig. 1a). This growth inhibition was more pronounced with two leaf-derived Xanthomonas species, L131 and L148 (Fig. 1a and Extended Data Fig. 1a,b), recently described as opportunistic pathogens on immunocompromised Arabidopsis rbohD mutant plants^5,40,58,59. Furthermore, L131 and L148 impaired seed germination (Extended Data Fig. 1c). This suggests that their pathogenic potential is not limited to leaves. Reduced seed germination and root growth were retained upon flg22 co-application, indicating that flg22-induced PTI does not protect the host against the detrimental impact of these opportunistic pathogens. However, L131-mediated detrimental activities were abolished by co-inoculation with a non-immunosuppressive SynCom consisting of five phylogenetically diverse Arabidopsis-root-derived bacteria that do not alleviate flg22-induced RGI (NS3 SynCom; Extended Data Fig. 1d and Supplementary Table 1)⁴⁴, similar to a previously tested SynCom comprising 136 leaf-derived commensal bacteria⁴⁰. Notably, co-inoculation of L131 with the NS3 SynCom but not the NS3 SynCom alone alleviated flg22-induced RGI (Extended Data Fig. 1e). Collectively, this indicates that the detrimental activities of L131 and the suppression of MAMP-induced RGI are independent traits.

We constructed a rooted phylogenetic tree comprising 1,765 publicly available Xanthomonadales draft genomes, including the genomes of our microbiota culture collections from A. thaliana, L. japonicus and C. reinhardtii (Fig. 1b and Supplementary Table 2). This revealed a global overrepresentation of phytopathogenic Xanthomonas species, particularly Xcv, Xcc and Xanthomonas oryzae pv. oryzae strains, as well as the opportunistic human pathogen Stenotrophomonas maltophilia. Strains isolated from A. thaliana, L. japonicus and C. reinhardtii, each cultured in the same Cologne agricultural soil, predominantly belong to Rhodanobacter and Lysobacter, Pseudoxanthomonas and Stenotrophomonas, respectively, indicating a potential preference of different Xanthomonadales genera for colonizing these three photoautotrophic hosts in Cologne agricultural soil (Fig. 1b). The strains tested for suppression of flg22-induced RGI belong to 7 of the 11 Xanthomonadales genera, each represented by at least three genomes (Fig. 1b). The suppression of immunity-associated RGI is thus highly conserved among Xanthomonadales regardless of the host species, organs or soil types from where they originate. Considering that both commensal and phytopathogenic Xanthomonadales can alleviate MAMP-induced RGI, this activity could be an ancestral trait that preceded specialization within Xanthomonadales to become plant-adapted pathogens. An exception to the conserved suppression of RGI was found in one of the two abundant soil-derived Xanthomonadales tested⁵, Rhodanobacter S772, which we failed to recover from 14-day-old inoculated roots (Fig. 1a,c). This suggests that the colonization of roots by Xanthomonadales is needed to relieve MAMP-induced RGI and that there is natural variation among soil-borne members of the Xanthomonadales in their ability to colonize Arabidopsis roots.

We next sought to explore the distribution of Xanthomonadales species across different environments (soil, aquatic, plant-associated and animal-associated). To that end, we leveraged the Microbe Atlas Project (MAP), which compiles over two million globally distributed 16S rRNA profiles. Xanthomonadales reference species (n = 57) were largely enriched in soil and plant-associated environments (78.9% of the samples, median across species; Methods). We also explored the distribution of potentially uncultivated Xanthomonadales across large-scale ocean, lake, animal-associated and soil metagenome-assembled genome (MAG) datasets, as well as a smaller plant-associated metagenomic survey. We found a total of 620 Xanthomonadales MAGs spanning 374 species in a global soil metagenomic dataset (40,039 MAGs spanning 21,076 species; 1.5% and 1.8%, respectively)⁶⁰ and 110 Xanthomonadales MAGs out of 910 in total (12.1%) in a metagenomic study of leafy greens⁶¹. By contrast, we did not identify any genome belonging to the Xanthomonadales order in a continent-scale lake metagenomic dataset (1,184 MAGs spanning 1,008 species)⁶² or a rumen-associated metagenomic dataset (4,941 genomes)⁶³, while we found only 69 Xanthomonadales MAGs from 10 species in an integrated human gut dataset (287,002 MAGs spanning 4,645 species; 0.02% and 0.2%, respectively)⁶⁴ and 7 Xanthomonadales genomes from 6 species in a global open-ocean dataset (26,309 MAGs spanning 5,098 species; 0.03% and 0.1%, respectively)⁶⁵. Taken together, these results provide evidence that terrestrial environments are the main reservoir for Xanthomonadales and suggest that soil, soil-dwelling algae and flowering plants may be the primary natural habitats for this entire order of Gammaproteobacteria.

We tested a subset of pathogenic or commensal Xanthomonadales strains derived from plant or algal hosts for suppression of RGI induced by the DAMP AtPep1 (Figs. 1a and 2b). All seven tested strains suppressed AtPep1-induced RGI, showing that the suppressive activity is not restricted to the flg22 elicitor. Treatment of L. japonicus with flg22 is known to elicit early immune responses such as reactive oxygen species spiking and transcriptional reprogramming^66,67. We observed reduced root growth in L. japonicus accession Gifu upon chronic exposure to flg22. All seven strains tested for suppression of AtPep1-induced RGI also alleviated flg22-induced L. japonicus RGI (Fig. 1d and Extended Data Fig. 2). This indicates that immunosuppressive strategies employed by Xanthomonadales are effective on at least two plant species.

**Fig. 2: *Rhodanobacter* R179 suppresses plant growth inhibition induced by various MAMPs/DAMPs and modulates SynCom-induced root transcriptional responses.**

Immunosuppressive Rhodanobacter modulates root transcription

We selected one suppressive Xanthomonadales strain, the Arabidopsis root commensal Rhodanobacter R179, which belongs to the deepest branch of the Xanthomonadales, for further in-depth analysis. As PTI-associated RGI was suppressed by leaf-derived Xanthomonas species in roots (Figs. 1a and 2a), we tested whether, vice versa, the alleviation of growth inhibition by root-derived R179 manifests in shoots (Fig. 2c). Shoot growth inhibition caused by the bacterial MAMP elf18, known to be predominantly recognized in leaves of Arabidopsis⁶⁸, was suppressed by R179 (Fig. 2c). Thus, interference with PTI-associated growth restriction by R179 is conserved in roots and leaves for the MAMPs flg22 and elf18, as well as the DAMP AtPep1.

Co-inoculation of Arabidopsis with the five-member NS3 SynCom together with R179 resulted in the suppression of flg22-induced RGI (Fig. 2d). This prompted us to investigate whether R179 influences host defence-associated transcriptional responses during plant–microbiota interactions in the absence of exogenously applied immunogenic peptides. We assessed the root transcriptome of Arabidopsis Col-0 colonized for 2 weeks by R179 alone, or together with either a five-member immunosuppressive SynCom (S3 SynCom) or the NS3 SynCom, designed on the basis of the members’ ability to alleviate flg22-induced RGI in mono-associations (Supplementary Table 1). Variation in root transcriptomes was predominantly explained by SynCom colonization regardless of immunomodulatory activities—that is, root transcriptomes of plants colonized by the NS3 or S3 SynCom are separated from axenic plants or those colonized by R179 alone (first principal component, 25% of the variation explained; Fig. 2e). The second component distinguished transcriptional responses to the S3 SynCom and NS3 SynCom (14%; Fig. 2e). Interestingly, co-inoculations of the S3 or NS3 SynCom with R179 resulted in an overall convergence of the corresponding root transcriptomes—that is, R179 reduced the dissimilarity of transcriptional changes induced by either SynCom alone (arrows in Fig. 2e).

k-means hierarchical clustering of differentially expressed genes (DEGs) obtained by comparing treated roots with axenic roots identified clusters c2 and c7 as specifically induced by the NS3 SynCom. Consistent with an immunomodulatory role of R179, all DEGs in both clusters were absent upon R179 co-inoculation (Fig. 2f). Gene Ontology (GO) enrichment analyses revealed that clusters c2 and c7 are enriched in genes with functions related to regulation of defence response, defence response to fungus, response to salicylic acid and cell-surface receptor signalling (Supplementary Table 3). Upon colonization with the S3 SynCom, we observed downregulation of another partially defence-related cluster, c4, and upregulation of cluster c14, which is enriched in genes with functions related to regulation of phenylpropanoid metabolism and detection of bacteria (Supplementary Table 3). Differential transcriptional regulation of genes in these two clusters was also attenuated by R179 co-colonization. Taken together, the impact of R179 on the expression of SynCom-regulated clusters c2, c7, c4 and c14 explains how the commensal reduces the dissimilarity of root transcriptomes. Comparative transcriptomic analysis indicated that these clusters overlapped significantly with those identified in a previous study using the identical NS3 and S3 SynComs⁴⁴. Genes shared between both studies show similar expression patterns and are enriched with functions related to defence, response to hypoxia, response to inorganic substance and cell wall modification (Supplementary Table 4). These results suggest that R179 root colonization suppresses defence responses triggered by either exogenously applied immunogenic peptides or live bacterial communities.

In contrast to prolonged co-cultivation of roots with SynComs for 2 weeks (381–495 DEGs), prolonged co-cultivation with R179 alone resulted in only minor transcriptional changes compared with axenic roots (95 DEGs; Fig. 2e,g). This was corroborated by proteomic analysis in which the relative abundance of none of the approximately 7,000 root proteins detected was significantly enriched or depleted upon R179 colonization (Extended Data Fig. 3). To investigate whether the apparent lack of root responses to live R179 colonization results from the evasion of host immune perception, we applied either live bacteria or an excess of heat-killed R179 (3,000-fold; optical density at 600 nm (OD₆₀₀) = 1.5) to the plant growth medium (Fig. 2g). On the basis of a linear regression correlating OD₆₀₀ and colony-forming units (CFU) (Extended Data Fig. 4) and the assumption that the activity of heat-killed bacteria will be reduced after 14 days of incubation, we expected the excess of heat-killed bacteria to be considerably less than the final titre of live R179 bacteria. Treatment with heat-killed R179 resulted in marked transcriptional responses exceeding responses to live R179 alone (256 DEGs) that are primarily attributable to four gene clusters, s1, s2, s3 and s8 (Fig. 2g). Cluster s8 was enriched in genes involved in cell-surface receptor signalling, pointing to the presence of heat-stable MAMPs in heat-killed R179 whose immunogenic activity is suppressed by live R179 during plant colonization (Supplementary Table 3).

Immunomodulatory determinants enhance R179 competitiveness

We used long-read sequencing to assemble and annotate a finished R179 genome. This assembly lacks operons for known virulence factors of bacterial phytopathogens, such as the T3SS machinery⁶⁹, biosynthesis genes of the phytohormone mimic coronatine⁷⁰, the flg22-degrading protease aprA⁷¹ and the PQQ biosynthesis operon (required for gluconic acid production)⁴². We conducted a mutant screen using Rhodanobacter R179 in planta to identify bacterial determinants contributing to the immunomodulatory activity. We generated 7,100 mini-Tn5 (ref. ⁷²) insertion mutants in the R179 background and tested them individually for their ability to suppress flg22-induced RGI in the Arabidopsis pWER::FLS2 line using a high-throughput phytostrips-based assay⁷³. A total of 67 candidates that showed partial loss of suppressive activity in two independent biological replicates were re-evaluated for impaired suppressive activity in Arabidopsis using conventional square agar plates (Fig. 3a). We identified ten dss mutant candidates with varying levels of reduced suppression of flg22-induced RGI (Fig. 3a and Extended Data Fig. 5a) and further characterized two of them.

**Fig. 3: Immunosuppressive determinants *dssA* and *dssB* provide *Rhodanobacter* R179 with a competitive advantage in a community context.**

Pre-colonization with wild-type R179, but neither the dssA nor the dssB mutant, inhibited induced expression of the defence marker gene FRK1 in Arabidopsis Col-0 roots 5 h after co-application of flg22 and AtPep1 peptides (Fig. 3b). The R179 dssA and dssB mutants have mini-Tn5 transposon insertions in the ABC transporter permease and TonB-dependent transporter genes, respectively (Extended Data Fig. 5a and Supplementary Table 5). The ABC transporter permease dssA is encoded within an operon encompassing 17 genes, whereas the TonB-dependent transporter dssB is predicted to be transcribed as monocistronic mRNA. Independent in-frame deletion mutants of dssA and dssB generated by homologous recombination exhibited impaired suppression of RGI induced by flg22, similar to the corresponding mini-Tn5 transposon insertion mutants, suggesting that the partial loss of RGI suppression is due to the disruption of the dssA and dssB genes rather than potential off-target effects (Fig. 3a). We generated a double mutant (dssAB) via in-frame deletion of dssA through homologous recombination in the dssB mini-Tn5 insertion mutant background. The resulting R179 dssAB double mutant also failed to suppress flg22-induced RGI (Fig. 3a). The in-frame deletions and the mini-Tn5 insertions were verified by long-read PacBio sequencing, and no additional chromosomal rearrangements were detected. The R179 dssA and dssB single mutants and the dssAB double mutant showed wild-type-like growth in liquid XVM2 minimal medium as well as wild-type-like colonization of Arabidopsis roots when inoculated in mono-associations (Extended Data Fig. 5b,c).

We next tested whether mutations of R179 dssA or dssB or combined mutations of both genes alter the establishment of synthetic microbiota on Arabidopsis Col-0 roots. Wild-type R179 and mutants were tagged with a recently developed MoBacTag barcode, facilitating strain-specific tracking during community profiling in planta⁷⁴. Upon co-inoculation of Arabidopsis with four phylogenetically matching five-member suppressive (S2 and S3) or non-suppressive SynComs (NS3 and NS4) with either wild-type R179 or the dssAB mutant, we conducted a constrained principal component analysis of Bray–Curtis dissimilarities on the SynCom-member-specific 16S rRNA reads normalized to spike reads (Fig. 3c, Extended Data Fig. 6a and Supplementary Table 1). This revealed that wild-type R179 and its dssAB double mutant differentially influenced the composition of both non-suppressive SynComs, whereas the compositions of the suppressive SynComs were indistinguishable regardless of co-colonization with wild-type R179 or the dssAB mutant. Co-colonization of R179 dssA or dssB single mutants resulted in an intermediate NS3 composition compared with the composition in the presence of wild-type R179 or the dssAB double mutant (Extended Data Fig. 6c). The inactivation of dssA and dssB thus appears to perturb the establishment of the community more strongly than that of the individual mutants.

Spike-normalized abundance of wild-type R179 was significantly elevated compared with the dssAB mutant on roots when inoculated together with both non-suppressive SynComs (NS3 and NS4) and the S2 suppressive SynCom (Fig. 3d and Extended Data Fig. 6b). By contrast, in liquid medium without plants, the abundances of wild-type R179 and the dssAB double mutant were comparable for both SynComs (Extended Data Fig. 6e). Furthermore, in liquid cultures, a similar composition of the non-suppressive SynCom (NS4) was observed regardless of co-cultivation with wild-type R179 or the dssAB double mutant, and the composition of the suppressive SynCom (S2) shifted only slightly when wild-type R179 or the dssAB mutant were co-cultivated (Extended Data Fig. 6d). This suggests that wild-type alleles of dssA and dssB provide R179 with a competitive advantage specifically in root-associated bacterial communities.

Orthogroup prediction using genomes of the At-R-SPHERE commensal culture collection and Xanthomonadales isolated from diverse environments showed that dssA and dssB are highly conserved within the Xanthomonadales and typically lacking in other phylogenetic lineages of the Arabidopsis bacterial root microbiota (Extended Data Fig. 5d). Within the Xanthomonadales, only a representative genome of the specialized pathogen Xylella fastidiosa, which has a significantly smaller genome than the average Xanthomonas species⁷⁵, lacked orthologues of dssA and dssB. Outside the Xanthomonadales clade within the At-R-SPHERE culture collection⁵, only some Oxalobacteriaceae harbour genes belonging to both dssA and dssB orthogroups. A BLAST-based search for dssA and dssB orthologous genes in publicly available Xanthomonadales draft genomes revealed that 74.7% and 88.9% of Xanthomonadales genomes harbour dssA and dssB orthologues, respectively (Supplementary Tables 10 and 11). Evolutionary conservation of dssA and dssB orthologues in Xanthomonadales thus correlates with the conserved capability of Xanthomonadales to suppress PTI-associated RGI (Fig. 1a and Extended Data Fig. 5d).

Rhodanobacter R179 eliminates immunogenic peptides

Since the dssAB double mutant exhibits residual suppression of flg22-induced RGI (Fig. 3a), we hypothesized that R179 has additional means to interfere with PTI. We pre-cultured R179 in XVM2, a minimal medium, which is known to induce virulence gene expression in phytopathogenic Xcv⁷⁶, and tested the bacteria-free R179 supernatant for the ability to suppress flg22-induced RGI in Arabidopsis pWER::FLS2. The supernatant of wild-type R179 contained immunosuppressive activities that were retained in the supernatant of the dssAB mutant (Extended Data Fig. 7a). Heat treatment of the supernatant abolished the immunosuppressive activity, suggesting the involvement of a secreted, soluble and heat-labile factor (Extended Data Fig. 7a). As suppression of flg22-induced RGI was also observed for the supernatant of the R179 dssAB mutant, R179 probably interferes with PTI via at least two distinct mechanisms.

We hypothesized that the observed immunosuppressive activity of the R179 supernatant may be caused by secreted peptidase(s) that are often sensitive to heat treatment. We quantified flg22 via mass spectrometry after 1 h of incubation in supernatant collected from wild-type R179 or the dssAB mutant or heat-treated supernatant of wild-type R179. We detected flg22 peptide after incubation in the heat-treated supernatant but not the supernatant of wild-type R179 or the dssAB mutant, indicating efficient and rapid (1 h) proteolytic degradation of the flg22 peptide in vitro (Fig. 4a and Extended Data Fig. 7b). Of note, two other immunogenic peptides, N-acetylated elf18 and AtPep1 were depleted in the supernatants of both wild-type R179 and the R179 dssAB mutant, whereas the root-growth-promoting sulfated PSY1 peptide and the de-sulfated dRGF1 peptide were still detected (Fig. 4a)^77,78. Unlike flg22 and elf18, minute traces of AtPep1 (<0.01%) were detected, and its reduced abundance correlated with the detection of a new peak corresponding to an AtPep1 cleavage product (ATKVKAKQRGK), indicating that AtPep1 reduction is probably caused by proteolytic cleavage (Extended Data Fig. 7c). To exclude possible starvation-induced expression of peptidases upon cultivation in minimal medium, we repeated the experiment using supernatant of R179 grown in XVM2 medium supplemented with 10% carbon-rich medium (tryptic soy broth) and obtained similar results (Extended Data Fig. 7d). We concluded that the R179 secretome suppresses PTI through selective elimination of MAMPs/DAMPs via mechanisms independent of dssA and dssB.

**Fig. 4: Selective peptide elimination by the cell-free supernatant of Rhodanobacteraceae.**

The culture supernatant from root commensal D. japonica MF79, a relative of Rhodanobacter R179, was similarly effective in degrading flg22, elf18 and AtPep1 after 1 h of incubation (Fig. 4b). This indicates that the secretion of peptidase(s) eliminating immunogenic peptides is a common mechanism employed by these two Xanthomonadales strains. The supernatant of the D. japonica MF79 gspD mutant also depleted the immunogenic peptides flg22, AtPep1 and elf18. Hence, an intact D. japonica MF79 T2SS is dispensable for the elimination of the immunogenic peptides tested (Fig. 4b).

Strain-specific factors determine Xanthomonadales root load

The transcriptomic analysis of Arabidopsis roots subjected to prolonged exposure to live and heat-killed R179 provided evidence that the commensal can both induce and suppress immune sectors in roots (Fig. 2g). The finding that R179 harbours immunogenic patterns was supported by the concentration-dependent induction of the expression of the FRK1 defence marker gene in roots 5 h after treatment with live or heat-killed R179 (Fig. 5a and Extended Data Fig. 8a). The supernatant of the Rhodanobacter R179 dssAB mutant induced FRK1 expression to a similar extent as the supernatant of wild-type R179, suggesting that the immunogenic activity of R179 functions independently of dssA and dssB (Extended Data Fig. 8d). We sought to identify Arabidopsis determinants required for R179 recognition using immune receptor kinase mutants. We found wild-type-like inducible FRK1 root expression in plant mutants lacking the receptors FLS2 or EFR or the adaptor kinase SOBIR1 (Fig. 5a). Wild-type-like responsiveness of the fls2 mutant is not surprising as the R179 genome lacks fliC, the source of the flg22 epitope. FRK1 expression induced by wild-type R179 or the dssAB double mutant was essentially abolished in the roots of the efr fls2 sobir1 triple mutant (Fig. 5a and Extended Data Fig. 8e), suggesting redundant immune perception of the commensal by the EFR and SOBIR1 cell-surface receptor complexes, which is independent of dssA and dssB.

**Fig. 5: PRR-dependent recognition of *Rhodanobacter* R179 does not affect the colonization of *Arabidopsis* in mono-association.**

The colonization of wild-type Arabidopsis Col-0 with live R179 suppressed PTI-associated RGI and did not negatively affect root growth (Fig. 2a–d). By contrast, heat-killed R179 induced RGI, which was abolished in efr fls2 sobir1 plants (Fig. 5b), confirming that root growth restriction is a bona fide PRR-triggered and PTI-associated response to the commensal. Although EFR is generally believed to be expressed in shoots but not in roots⁶⁸, we verified an R179-inducible root signal in an EFR transcriptional reporter line (pEFR::NLS–3*mVenus)^79,80, which was further supported by the detection of endogenous EFR root transcripts upon R179 inoculation, but not elf18 application, in wild-type Arabidopsis (Extended Data Fig. 8b,c). Thus, R179-derived molecules other than elf18 seem to induce the expression of EFR, which subsequently contributes to R179 detection by the plant immune system in roots. Taken together, these data show redundant recognition of the R179 commensal by EFR and SOBIR1 cell-surface receptor complexes in roots.

As R179 also alleviated shoot growth inhibition upon prolonged exposure to elf18 (Fig. 2c), we examined ethylene biosynthesis as another early plant immune response in leaves 5 h after treatment with native or heat-inactivated lysates of R179. Dose-dependent ethylene production was indistinguishable between wild-type Arabidopsis and the fls2 mutant, but it was almost abolished in efr fls2 sobir1 (Fig. 5c and Extended Data Fig. 9a). Unexpectedly, ethylene production in the efr mutant was significantly reduced upon treatment with heat-inactivated but not native R179 (Fig. 5c). This suggests that a dominant heat-stable elicitor (or elicitors) of R179 is detected through EFR to induce ethylene production in leaves, which is masked by heat-labile MAMPs recognized in an EFR-independent but SOBIR1-dependent manner. This differs from the observed induction of defence marker gene expression in roots, where no heat-labile immunogenic activity was found with live R179, and where the response to heat-killed R179 was not affected in the efr single mutant. The genome of R179 encodes the cognate ligand of EFR, EF-Tu. We synthesized the corresponding peptide epitope elf18^R179, differing from the commonly used elf18 elicitor by two amino acids, and verified EFR-dependent shoot growth inhibition induced by elf18^R179 (Extended Data Fig. 9b,c). Moreover, R179 colonization suppressed shoot growth inhibition induced by exogenously administered elf18^R179 or elf18^{E. coli} (Fig. 2c and Extended Data Fig. 9c). This supports the notion that camouflage by R179 in shoots involves the elimination of elf18^R179 (Fig. 4a).

We observed ethylene production in efr fls2 sobir1 leaves following AtPep1 treatment that is indistinguishable from that in wild-type Arabidopsis (Extended Data Fig. 9d), indicating that the triple mutant can still mount PTI-associated defence upon PRR activation. We tested whether R179 produces the Xanthomonas-specific, heat-labile MAMP eMAX, whose perception is SOBIR-dependent, but we did not detect eMAX receptor (RLP1)-dependent ethylene production (Extended Data Fig. 9e)^81,82. Instead, bacterial initiation factor 1, recently identified as a SOBIR1-dependent heat-labile bacterial MAMP, is encoded by R179 and might contribute to commensal perception (Extended Data Fig. 9f)⁸³. Collectively, our data reveal the detection of at least three MAMPs of R179 by Arabidopsis (elf18^R179, heat-labile and heat-stable MAMPs perceived by SOBIR1-dependent PRRs).

We next tested whether PRR-dependent recognition of wild-type R179 or the dssAB mutant, which is impaired in interference with PTI-associated root responses, limits bacterial proliferation on roots in mono-associations. Neither the fls2, efr or sobir1 single mutant nor the efr fls2 sobir1 triple mutant showed differential root and shoot colonization by wild-type R179 or the dssAB mutant (Fig. 5d and Extended Data Fig. 10a). Moreover, simultaneous application of synthetic flg22 or elf18 peptides with bacteria did not restrict wild-type R179 or dssAB mutant growth on wild-type Arabidopsis roots (Fig. 5e). Together, this shows that in mono-associations MAMP treatment and PRR-dependent commensal perception do not effectively limit wild-type R179 or dssAB mutant root load. To investigate whether the results obtained with R179 have broader relevance within Xanthomonadales, we determined the root load on Arabidopsis pWER::FLS2 plants in mono-associations with 19 root microbiota members derived from A. thaliana or L. japonicus and two C. reinhardtii phycosphere members. The root load of individual Xanthomonadales strains varied markedly by almost 1,000-fold (ranging between 10⁶ and 10⁹ CFU per g root tissue; Extended Data Fig. 10b) and was unchanged upon flg22 treatment for most tested strains. We determined for each strain the difference in root length between flg22- and mock-treated plants in relation to the difference in root length between axenic flg22- and mock-treated plants in relation to its root titre in mono-association (suppressiveness score; Extended Data Fig. 10c). This revealed that Xanthomonadales root load does not correlate with the suppressiveness score and that Arabidopsis-derived Xanthomonadales do not colonize their cognate host to higher levels compared with Xanthomonadales strains isolated from other photoautotrophic hosts. Thus, in mono-associations, Xanthomonadales root colonization is determined by strain-specific factors that act dominantly over their conserved immunomodulating activity.

Discussion

We have shown here that the root commensal Rhodanobacter R179, belonging to the deepest branch of the Xanthomonadales, suppresses pattern-triggered immune responses in Arabidopsis roots and leaves (Fig. 2a–c). Other Xanthomonadales commensals, recovered from diverse soil-dwelling photoautotrophic eukaryotic hosts, alleviated PTI-associated RGI in A. thaliana and L. japonicus roots. This shows that immunosuppression is ubiquitous in this core order of the plant microbiota. The near absence of Xanthomonadales in free-living ocean and lake prokaryotic communities, as well as in metazoan guts, supports the notion that the primary reservoirs for this entire order of Gammaproteobacteria are soil and soil-dwelling photoautotrophic hosts^{50,51,62,65,84,85}. Hence, maintaining immunomodulatory activities is associated with the prevalence of Xanthomonadales in their primary habitat and community competitiveness in the plant microbiota. The latter conclusion is supported by Rhodanobacter R179 dssAB-dependent immunomodulation, which gives the commensal a competitive advantage in a bacterial community on Arabidopsis roots even in the absence of exogenous stimulation with immunogenic elicitors.

Comparative analysis of immunity-associated responses in wild-type Col-0 and efr fls2 sobir1 triple mutant plants to live and heat-killed Rhodanobacter R179 demonstrated that Arabidopsis detects the presence of R179 through the perception of at least three MAMPs. This must be reconciled with our finding that prolonged root colonization in mono-association by live R179 but not heat-killed commensal induced only minor root transcriptional reprogramming (Fig. 2g). Our results suggest that immunoevasion by R179 via the secretion of heat-labile extracellular peptidase(s) for the selective elimination of immunogenic peptides and dssAB immunomodulation leads to R179 camouflage. This could partly account for subtle host transcriptional changes induced by R179 colonization despite presenting multiple MAMPs.

The capacity to eliminate or modify immunogenic peptides is unlikely to be R179-specific. Phytopathogenic Pseudomonas syringae DC3000 releases AprA protease via the type I secretion system to deplete flg22 for full virulence on Arabidopsis leaves^71,86. In addition, Actinobacteria root commensals deplete flg22 via an unknown mechanism⁴⁴. We have shown here that the immunomodulatory root commensal D. japonica MF79 eliminates multiple immunogenic peptides (Fig. 4b), similar to R179. This is in line with a recently identified immunosuppressive subtilase A (IssA) in MF79 that cleaves flg22 and AtPep1 (ref. ⁸⁷). This suggests that clearance of immunogenic peptides is conserved in at least two Xanthomonadales genera. We found that peptide clearance is retained in an MF79 T2SS mutant, which is impaired in the suppression of flg22-induced CYP71A12 defence marker gene expression in roots⁴³. This implies that MAMP-eliminating peptidase(s) can be exported by a T2SS-independent mechanism and that D. japonica MF79 deploys at least two immunomodulatory mechanisms.

The proposed elimination of flg22 by secreted peptidase(s) for R179 camouflage is counterintuitive as the R179 genome lacks fliC, whose product contains the flg22 epitope. Functional diversification of the flg22 epitope has been shown in commensals of the Arabidopsis root microbiota, with most peptide variants evading recognition by FLS2 or acting as antagonists of immunogenic flg22^Pao (refs. ^46,47). Thus, clearance of immunogenic flg22 epitopes by R179-secreted peptidase(s) does not directly benefit R179 but in principle impacts resident microbiota members expressing fliC. This could explain why R179 acts dominantly over the five-member NS3 SynCom containing at least one flg22-presenting Pseudomonas commensal to reduce root transcriptomic responses⁴⁶. Hence, R179-secreted peptidase(s) function in hiding the producer and other microbiota members from host recognition. R179-mediated immunosuppression through combined elimination of the DAMP AtPep1 and MAMPs is also in accordance with a recently proposed model for PTI in roots, in which the coincidence detection of cell damage and microbial non-self patterns controls local immune outputs⁷⁹.

To our knowledge, two immunomodulatory mechanisms operating in individual plant commensals, D. japonica MF79 and Rhodanobacter R179, have not been reported before. MAMP/DAMP elimination by extracellular bacterial peptidase(s) is probably an immunoevasive mechanism, whereas dssAB-dependent secretion could be part of an immunoevasive or immunosuppressive strategy. Given that we identified dssA and dssB in a genetic screen as bacterial determinants for the suppression of PTI-associated RGI in the absence of other commensals and found elevated competitiveness of R179 in SynComs in the root compartment, these bacterial genes probably contribute to host immunosuppression rather than acting directly on other community members. Orthologues of dssA and dssB are largely conserved in Xanthomonadales (Extended Data Fig. 5d) and may contribute to their success in colonizing plant hosts as a core lineage of the microbiota⁵¹. The wild-type-like colonization of the R179 dssAB mutant on roots and the presence of dssA and dssB orthologues in the genome of Rhodanobacter S772, which fails to colonize roots, argues against a further alternative explanation that these genes act solely as root competence factors. The conservation of dssA and dssB across the entire order of Xanthomonadales contrasts with the presence of the T3SS, which is restricted to the genus Xanthomonas within the Xanthomonadales^58,88. The T3SS of phytopathogenic Xanthomonas species is essential for virulence and was potentially acquired in three independent ancestral acquisition events, followed by multiple sequential losses in commensals^88,89. As all known T3SS-containing phytopathogenic Xanthomonadales belong to the genus Xanthomonas, it is perhaps no coincidence that the opportunistic pathogens L131 and L148 also belong to Xanthomonas, despite the absence of the T3SS in these two strains^40,58,59. L131 also suppressed PTI-associated RGI (Extended Data Fig. 1d,e). Together, these results indicate that T3SS-independent but dssAB-dependent immunosuppression is an ancient trait whose evolution preceded the emergence of T3SS-associated pathogen specialization and that is maintained throughout the evolutionary history of Xanthomonadales.

TonB-dependent and ABC transporters frequently function in concert for substrate translocation across the outer and inner membranes of Gram-negative bacteria⁹⁰. TonB-dependent transporters, which are overrepresented in pathogenic Xanthomonas genomes, are induced in the genus Sphingomonas within the leaf microbiota and are involved in the import of various substrates, including plant-derived carbohydrates, vitamins, and iron and nickel chelates^9,91,92,93. Furthermore, TonB-dependent transporters were reported to function in the export of a protease⁹⁴. Future work is needed to identify the dssB substrate(s) involved in the immunomodulatory activity and test the conservation of substrate specificity for dssB orthologues in Xanthomonadales.

The application of immunogenic peptides to Arabidopsis or the use of the efr fls2 sobir1 triple mutant, which barely perceives R179, does not restrict or enhance root colonization by wild-type R179 or the dssAB mutant (Fig. 5d,e). The root load of 17 Xanthomonadales commensals in mono-associations with Arabidopsis varied 1,000-fold up to 10⁹ CFU g⁻¹, a bacterial titre that is comparable to that of phytopathogens, and was not limited by flg22 treatment. By contrast, the proliferation of endophytic P. syringae DC3000 pathogen in Arabidopsis leaves is restricted by flg22 co-treatment despite T3SS-mediated immunosuppressive activities and AprA protease-mediated flg22 degradation^71,95,96. The unaltered commensal colonization by Xanthomonadales upon MAMP treatment observed here might result from the plant developmental stage at inoculation—that is, starting from germination—and epiphytic root colonization. In addition, tissue-specific PTI responses might contribute to the apparent uncoupling of Xanthomonadales root load from PTI-mediated bacterial growth restriction demonstrated for pathogenic bacteria.

We have shown that in mono-associations, root colonization by Xanthomonadales commensals is determined by strain-specific factors that act dominantly over their conserved immunomodulating activity. As heterotrophic bacterial proliferation on plants is primarily driven by plant-derived structural and non-structural carbohydrates^8,9, the strain-specific root load of Xanthomonadales commensals in these mono-associations might reflect diversified host resource use. However, immunomodulation by R179 through dssAB was found to exert a dominant influence and elevate R179 root load when the commensal co-colonized with distantly related microbiota members. In this community context, Rhodanobacter R179 can be expected to experience enhanced metabolic constraints for root colonization compared with a mono-association. It is conceivable that conserved immunomodulation by Xanthomonadales may function in parallel with or facilitate the diversification of substrate preferences for host resources, thereby contributing to the co-existence of multiple Xanthomonadales strains in the plant microbiota.

Methods

Plant cultivation conditions

A. thaliana Col-0, rbohD³⁴, fls2 SAIL_691-C4 (ref. ⁹⁵), efr-1 SALK_044334 (ref. ⁹⁷) and sobir1-12 SALK_050715 (ref. ⁹⁸) and L. japonicus Gifu B-129 seeds were obtained from the institute stocks. A. thaliana pWER::FLS2–GFP⁵⁶ and pEFR::NLS–3*mVenus⁷⁹ were provided by N. Geldner, and the efr-1 fls2 sobir1-12 mutant, provided by L. P. Maier and G. Felix (University Tübingen, Germany), was generated by crossing the efr-1 fls2 mutant⁹⁹ with the sobir1-12 mutant. The newly generated mutant line will be made available by the authors on request. The Arabidopsis efr fls2 rlp1 mutant was present at the University Tübingen, Germany⁸¹. Plants were cultivated on vertical agar plates under short-day conditions (10 h light, 21 °C; 14 h darkness, 19 °C) with 65% relative air humidity and a light intensity of 120 mE m⁻² s⁻¹. For ethylene measurements, 7-week-old plants grown in potting soil under short-day conditions (8 h light, 24 °C; 16 h darkness, 22 °C) with 40–60% relative air humidity were used.

Bacterial cultivation conditions

A list of all the bacterial strains used can be found in Supplementary Table 1. Bacteria were obtained from the At-SPHERE⁵, Lj-SPHERE⁵⁵ and Cr-SPHERE⁵⁰ culture collections. MF92 and MF178 were retrieved from Levy et al.⁵⁷. Xanthomonas campestris pv. vesicatoria 85-10, Xanthomonas campestris pv. campestris 8004 and Xanthomonas hortorum pv. pelargonii were provided by U. Bonas (Martin-Luther-University, Halle-Wittenberg, Germany)¹⁰⁰, L. Noel (lipme Toulouse, France)¹⁰¹ and K. Richter (Julius Kühn Institute, Quedlinburg, Germany), respectively. Xanthomonas arboricola pv. pruni and Xanthomonas campestris pv. raphani were present at the University Tübingen (Germany). Bacteria were cultivated in 0.5× TSB (15 g l⁻¹ tryptic soy broth, Sigma Aldrich; supplemented, if necessary, with 10 g l⁻¹ Bacto Agar, Duchefa Biochemie) at 25 °C with shaking at 180 rpm for liquid cultures only. E. coli DH5α λ-pir and pRK600 were cultivated in LB medium. For ethylene measurements, agar-grown Rhodanobacter R179 was resuspended (OD₆₀₀ ≈ 1), plated on NYG medium (0.5% (w/v) peptone, 0.3% (w/v) yeast extract, 2% (w/v) glycerol, 1% (w/v) agar) and incubated at 28 °C for 2 days.

Agar-based cultivation of axenic plants and co-cultivation with bacteria, bacterial supernatants and elicitors

The inoculation of seeds with bacteria was performed as previously described⁷³. In brief, A. thaliana seeds were surface-sterilized using two 5-min incubations with 70% ethanol and 15 s of incubation in 96% ethanol followed by five to seven washes in H₂O. Sterilized A. thaliana seeds were stratified for 2 days at 4 °C. The surface of L. japonicus seeds was abraded using sandpaper, followed by incubation for 20 min in 2% (v/v) bleach; the seeds were then washed five times with H₂O. Sterilized L. japonicus seeds were placed on H₂O-wetted Whatman paper in square petri dishes and germinated for 4 days under short-day conditions. Saturated overnight bacterial cultures were harvested by centrifugation (11,600 g) for 2 min and washed twice in 10 mM MgSO₄. Bacteria were inoculated into 0.5× Murashige and Skoog (MS) medium (2.22 g l⁻¹ MS basal salts medium, Sigma; for testing in-frame mutants, MS medium mod. Mikro-Komplex Phygenera (MD0830); 0.1 g l⁻¹ MES, BioChemica; pH 5.7; 10 g l⁻¹ Bacto Agar, Duchefa Biochemie) prior to medium solidification at a final concentration of OD₆₀₀ = 0.0005 (per strain for SynComs). The suppression of flg22-induced RGI by wild-type R179 and the mutants varied depending on the agar source and 0.5× MS medium used¹⁰². We recommend the calibration of plant phenotypes on a given batch of agar and MS medium combination with at least four replicate plates per condition.

Elicitors were also supplemented into the medium prior to solidification at a final concentration of 1 µM. The flg22 (QRLSTGSRINSAKDDAAGLQIA) and AtPep1 (ATKVKAKQRGKEKVSSGRPGQHN) peptides were synthesized by EZbiolab. elf18^R179 (Ac-AKGKFERTKPHVNVGTIG) and elf18 (Ac-SKEKFERTKPHVNVGTIG) were synthesized by GenScript and EZbiolab, respectively. For inoculation with the bacterial supernatant, bacteria were cultivated overnight in XVM2 minimal medium (20 mM NaCl, 10 mM (NH₄)₂SO₄, 5 mM MgSO₄, 1 mM CaCl₂, 0.16 mM KH₂PO₄, 0.32 mM K₂HPO₄, 0.01 mM FeSO₄, 10 mM fructose, 10 mM sucrose, 0.03% Casamino acids, pH 6.7)⁷⁶ to a final concentration of OD₆₀₀ = 0.4. Bacteria were pelleted by centrifugation (11,600 g, 2 min), and the supernatant was sterile-filtered using filter units with MF-Millipore membrane (0.22 µm; Millex). The supernatant was incubated in boiling water for 30 min for heat treatment. Then, 25 ml of 1× MS with 20 g l⁻¹ Bacto Agar was mixed with 25 ml of supernatant prior to medium solidification and poured into square petri dishes. Sterilized seeds were placed on solidified agar in square petri dishes and vertically incubated for 14 days.

Quantification of bacterial abundance via live colony counts

Roots of 14-day-old seedlings were briefly washed in 10 mM MgSO4, dried with autoclaved Whatman paper and stored in weighed, sterilized screw-cap tubes containing metal beads. The tubes were weighed again to calculate root fresh weights prior to sample homogenization using a tissue homogenizer (Bertin Precellys; 6,200 rpm, 2 × 30 s, 15 s pause in between). Then, 500 µl of 10 mM MgSO₄ was added, and the samples were homogenized again (same conditions). Dilution series were prepared in 96-well plates, and 10 µl portions were dropped onto 0.5× TSB in square petri dishes and spread across half of the plate by tilting the plate. Bacteria were incubated at 25 °C until colonies could be counted⁷³.

Acute treatment with bacteria or elicitors

Axenic 14-day-old seedlings (see ‘Agar-based cultivation of axenic plants and co-cultivation with bacteria, bacterial supernatants and elicitors’) were submerged in 12 ml of 10 mM MgSO₄ supplemented with 1 µM elicitor or bacteria, which were directly harvested from 0.5× TSB plates and resuspended in 10 mM MgSO₄. Bacterial concentration was adjusted to OD₆₀₀ = 1.5, if not indicated differently. For a heat-killed bacterial inoculum, resuspended bacteria were incubated in boiling water for 30 min. After incubation for 5 h on a shaker, roots were directly harvested into screw-cap tubes containing metal beads, frozen in liquid nitrogen and stored at −80 °C until further processing. For co-exposure to flg22 and AtPep1, seedlings were submerged as described for 1 h, residual liquid was removed and the plants were incubated in a cultivation chamber for 4 h prior to harvesting.

Quantification of defence marker gene expression by RT–qPCR

For reverse transcription quantitative PCR (RT–qPCR), roots of eight to ten plants grown on one square plate were pooled for one sample. Frozen roots were homogenized using a TissueLyser II (QIAGEN, 2 × 29 pps for 40 s) with pre-cooled adaptors. RNA was extracted using the Plant RNA Kit (my-Budget) according to the user manual and eluted in 35 µl of elution buffer. RNA concentrations were measured using a spectrometer (Nanodrop OneC Microvolume UV-Vis Spectrophotometer). Complementary DNA was synthesized from 300–500 ng of RNA using the RevertAid H Minus First Strand cDNA Synthesis Kit (ThermoScientific) in a 20-µl reaction according to the user manual. In brief, 1 µl of oligo(dT)₁₈ primer was mixed with RNA in a final volume of 12 µl and incubated at 65 °C for 5 min. Afterwards, 4 µl of 5× Reaction buffer, 2 µl of 10 mM dNTPs, 1 µl of RiboLock RNase Inhibitor (20 U µl⁻¹) and 1 µl of RevertAid H Minus M-MuLV Reverse Transcriptase (200 U µl⁻¹) were added, and the reaction was incubated at 42 °C for 60 min followed by heat inactivation at 70 °C for 5 min. cDNA was diluted 1:10 prior to qPCR using the iQ SYBR Green Supermix (Bio-Rad). Reactions (20 µl) were prepared with 5 μl of cDNA, 10 μl of SYBR Green Supermix and 0.4 μl of each primer (10 µM; Supplementary Table 6). qPCR was performed using the CFX ConnectTM Real-Time System (Bio-Rad) with the following conditions: 95 °C for 3 min; 95 °C for 15 s, 65 °C for 15 s and 72 °C for 15 s for four cycles; and 95 °C for 15 s, 57 °C for 15 s and 72 °C for 15 s for 39 cycles, followed by a melting curve.

Confocal microscopy

Six-day-old axenic seedlings cultivated on 0.5× MS agar plates were flush-inoculated with live R179 (OD₆₀₀ = 1.5) and incubated for 1 h with gentle shaking, followed by removal of extensive liquid. The plates were incubated overnight (see ‘Plant cultivation conditions’), and the roots were visualized using confocal microscopy. Confocal laser scanning microscopy was performed on a Zeiss LSM880 inverted confocal scanning microscope. Pictures were taken with an LD C-Apochromat ×40/1.1 water immersion objective. To image root colonization, Z-stacks were generated, and maximum-intensity projections were compiled. The following excitation and detection window was used: GFP 488 nm, 493–598 nm.

16S rRNA amplicon sequencing and community profiling

After 14 days of co-cultivation of bacteria with A. thaliana plants, roots from approximately 20 seedlings grown on three different square plates inoculated with the same bacterial inoculum were pooled, briefly washed in 10 mM MgSO₄, dried using Whatman paper, transferred into Lysing Matrix E tubes (MP Biomedicals) and frozen in liquid nitrogen. For community profiling upon cultivation in liquid cultures, the indicated SynComs were inoculated in parallel in XVM2 medium (see ‘Bacterial cultivation conditions’) supplemented with 10% TSB at an OD₆₀₀ of 0.05. The next day, bacteria were pelleted by centrifugation (11,600 g, 2 min) and frozen in liquid nitrogen. Sample preparation and 16S rRNA community profiling were performed according to Bulgarelli et al.⁴. In brief, roots were homogenized as before (see ‘Quantification of bacterial abundance via live colony counts’), and total DNA was extracted using the FastDNA SPIN Kit for Soil (MP Biomedicals) according to the user manual. DNA was eluted in 80 µl of H₂O, and concentrations were measured using the Quant-iT PicoGreen dsDNA Assay (Thermo Fisher Scientific) and adjusted to a concentration of 3.5 ng µl⁻¹.

PCR I and II were both performed in 25-µl reaction volumes containing 10.5 ng of sample DNA, 2 U of DFS-Taq DNA polymerase (Bioron), 1× incomplete buffer, 2 mM MgCl₂, 0.3% bovine serum albumin, 0.2 mM dNTPs (Life Technologies) and 0.3 μM forward and reverse primers. PCR I was performed in triplicates with the primers 799F and 1192R and an additional 0.004 ng of pBCC23 plasmid DNA per reaction (Supplementary Table 6) using the following conditions: 94 °C for 2 min; 94 °C for 30 s, 55 °C for 30 s and 72 °C for 30 s for 25 cycles; and a final extension at 72 °C for 10 min. Afterwards, 20 µl of the pooled triplicates of PCR I were purified by primer and protein digestion with 5 U of Antarctic phosphatase, 20 U of exonuclease I and 1× Antarctic phosphatase buffer (New England Biolabs) at 37 °C for 30 min, followed by enzyme deactivation at 85 °C for 15 min. The reactions were centrifuged for 15 min (4,200 g), and 3 μl of supernatant was used for PCR II. In PCR II, the samples were double-indexed with one indexed forward primer per full factorial replicate and an indexed reverse primer for each sample per full factorial replicate. PCR reactions and conditions were similar to those for PCR I, but the cycles were reduced to ten. Afterwards, the PCR quality and quantity were checked on a 1.5% agarose gel, indexed amplicons of biological replicates were pooled and separated from plant 16S rRNA amplicons on a 1.5% agarose gel and bacterial 16S rRNA was cut out together with the pBCC23 amplicon and purified using the QIAquick Gel Extraction Kit (QIAGEN) according to the user manual. DNA concentrations were again measured using the Quant-iT PicoGreen dsDNA Assay (Thermo Fisher Scientific), and equal amounts of DNA were pooled. The library was then again purified and concentrated using Agencourt AMPure XP beads, and the DNA concentration was determined using the Quantus Fluorometer (Promega). Finally, 17.5 ng µl⁻¹ of the final library were subjected to paired-end Illumina sequencing in-house, using the MiSeq sequencer and custom sequencing primers.

Sequencing reads were demultiplexed according to the respective index sequence and quality-filtered using the QIIME pipeline¹⁰³. Then, forward and reverse sequencing reads were merged using flash2 software¹⁰⁴ and aligned to reference V5V7 16S rRNA sequences of all SynCom members and the pBCC23-specific amplicon sequence using Rbec¹⁰⁵. For experiments with S2 and NS4, reads were manually curated for the inoculated strains for each sample. Finally, an amplicon sequence variant table was generated. First, counts were normalized by the pBCC23 spike in counts for each sample. Bray–Curtis dissimilarities were calculated after in silico depletion of R179-specific counts using the vegdist function of the vegan package (v2.5.7)¹⁰⁶, and the constrained principal component analysis was performed using the vegan capscale function constraining by the interaction between the R179 derivative (wt/dssAB) and SynCom (suppressive/non-suppressive) and conditioning by technical and biological replicates. Finally, all amplicon data were visualized using ggplot2 (ref. ¹⁰⁷) implemented in the tidyverse package (v.1.3.1)¹⁰⁸.

Quantification of peptides by mass spectrometry

Bacterial supernatants were prepared as described in ‘Agar-based cultivation of axenic plants and co-cultivation with bacteria, bacterial supernatants and elicitors’. Here, additionally, the peptides dRGF1 (DYSNPGHHP-Hyp-RHN) and PSY1 (D-Y(SO₃H)-GDPSANPKHDPGV-Hyp-Hyp-S) synthesized by Scilight Peptide were incubated for 1 h with 1 µM of peptides in a plant cultivation light chamber (day-light condition; see ‘Plant cultivation conditions’), and the samples were frozen at −20 °C until further processing. For the investigation of AtPep1 degradation, the AtPep1 peptide was first modified by acetylation (six to seven times). To this end, 100-µl aliquots of peptides were treated with 10 µl of NHS acetate solution (1 M NHS acetate in 50% ACN/PBS) and incubated for 30 min with shaking. The reaction was quenched by the addition of 100 µl of 100 mM Tris-HCl (pH 8.5) and incubated for 10 min. For comparison of the non-acetylated peptides with the above-described modified AtPep1 peptides, aliquots of the non-acetylated peptides were processed as described above, but without the modifying NHS reaction. To this end, 100-µl aliquots of the sample were mixed with 100 µl of PBS buffer (137 mM NaCl, 2.7 mM KCl, 10 mM Na₂HPO₄, 1.8 mM KH₂PO₄; pH 7.4). The samples were then further diluted with 100 µl of 100 mM Tris-HCl (pH 8.5) and incubated for 10 min. All samples were acidified with 5 µl of TFA and desalted using StageTips with C18 Empore disk membranes (3M)¹⁰⁹. The samples were dried in a vacuum evaporator and dissolved in 10 µl of 2% ACN (0.1% TFA for analysis).

The samples were analysed using an EASY-nLC 1200 (Thermo Fisher) coupled to a QExactive Plus mass spectrometer (Thermo Fisher). Peptides were separated on 16-cm frit-less silica emitters (New Objective, 75 µm inner diameter), packed in-house with reversed-phase ReproSil-Pur C18 AQ 1.9 µm resin (Dr. Maisch). Peptides were loaded on the column and eluted for 50 min using a segmented linear gradient of 5% to 95% solvent B (0 min, 5% B; 0–5 min, 15% B; 5–35 min, 50% B; 35–45 min, 95% B; 45–50 min, 95% B) (solvent A: 0% ACN, 0.1% FA; solvent B: 80% ACN, 0.1% FA) at a flow rate of 300 nl min⁻¹. Mass spectra were acquired in data-dependent acquisition mode with a TOP10 method. Mass spectrometry spectra were acquired in the Orbitrap analyser with a mass range of 300–1,500 m/z at a resolution of 70,000 full width at half maximum (FWHM) and a target value of 3 × 10⁶ ions. Precursors were selected with an isolation window of 1.3 m/z. Higher-energy collision dissociation (HCD) fragmentation was performed at a normalized collision energy of 25. Tandem mass spectrometry (MS/MS) spectra were acquired with a target value of 5 × 10⁵ ions at a resolution of 17,500 FWHM, a maximum injection time of 85 ms and a fixed first mass of 100 m/z. Peptides with a charge of 1 or greater than 6 or with an unassigned charge state were excluded from fragmentation for MS/MS; dynamic exclusion for 20 s prevented repeated selection of precursors. The raw data were analysed on the MS1 level using Skyline (https://skyline.ms)¹¹⁰, and the results were filtered for the respective intact peptides, or the acetylated version in the case of AtPep1. Peaks were checked and, if needed, integrated manually. Finally, peak areas were exported for further processing.

Root proteomics by mass spectrometry

Sample preparation

Approximately 20 mg of root tissue from 14-day-old seedlings co-cultivated with R179 on three square plates was harvested into Lysing Matrix E tubes (MP Biomedicals) and immediately frozen in liquid nitrogen. Frozen root samples were disrupted using a Retsch mill at 50 Hz for 5 min. Then, 350 µl of hot SDT extraction buffer (4% SDS in 100 mM Tris-HCl, pH 8.5, 0.1 M DDT) was added, and the samples were incubated at 95 °C for 10 min. Next, the samples were centrifuged for 5 min at 5,000 g, and the supernatants were again centrifuged at full speed for 10 min. Protein concentrations were determined using a Pierce 660 mn protein assay, and aliquots corresponding to 50 µg of total protein were digested using a filter-aided sample preparation method¹¹¹. Briefly, 450-µl extracts were loaded onto a Vivacon 500 (Sartorius Stedim Biotech), centrifuged at 14,000 g and washed once with 450 µl of UA (8 M in 100 mM Tris-HCl, pH 8.5). Next, the samples were alkylated with CAA (100 µl, 55 mM CAA) for 20 min at room temperature in the dark. After centrifugation at 14,000 g for 10 min, the samples were washed three times with UA (450 µl), and filters were exchanged onto fresh tubes to collect flow-through for the following steps. Next, 50 µl of LysC solution (1:100 enzyme/protein in 100 mM Tris-HCl, pH 8.5) was added, and the samples were incubated for 3 h at room temperature. Then, 250 µl of trypsin solution (1:100 enzyme/protein in 100 mM Tris-HCl, pH 8.5) was added onto the filter, the filter was filled with 100 mM Tris-HCl pH 8.5 to the mark, the solutions were mixed well and the samples were incubated overnight at 37 °C. After digestion, the filters were centrifuged to collect the flow-through and were washed once with 50 µl of 100 mM Tris-HCl, pH 8.5. The flow-throughs were combined and acidified/desalted as described in ‘Quantification of peptides by mass spectrometry’. Peptide concentrations were adjusted to 0.2 µg µl⁻¹.

Data acquisition

The samples were analysed using an Ultimate 3000 RSLC nano (Thermo Fisher) coupled to an Orbitrap Exploris 480 mass spectrometer equipped with an FAIMS Pro interface for field asymmetric ion mobility separation (Thermo Fisher). Peptides were pre-concentrated on an Acclaim PepMap 100 pre-column (75 µm × 2 cm, C18, 3 µM, 100 Å, Thermo Fisher) using the loading pump and buffer A** (water, 0.1% TFA) with a flow of 7 µl min⁻¹ for 5 min. Peptides were separated as described in ‘Quantification of peptides by mass spectrometry’. Peptides were loaded on the column and eluted for 130 min using a segmented linear gradient of 5% to 95% solvent B (0–5 min, 5% B; 5–65 min, 20% B; 65–90 min, 35% B; 90–100 min, 55% B; 100–105 min, 95% B; 105–115 min, 95% B; 115–115.1 min, 5% B; 115.1–130 min, 5% B) (solvent A: 0% ACN, 0.1% FA; solvent B: 80% ACN, 0.1% FA) at a flow rate of 300 nl min⁻¹. Mass spectra were acquired in data-dependent acquisition mode with a TOP_S method using a cycle time of 2 s. For field asymmetric ion mobility separation, two compensation voltages (−45 and −60) were applied with a cycle time of 1 s for each experiment. Mass spectrometry spectra were acquired in the Orbitrap analyser with a mass range of 320–1,200 m/z at a resolution of 60,000 FWHM and a normalized AGC target of 300%. Precursors were filtered using the MIPS option (MIPS mode = peptide), the intensity threshold was set to 5,000 and precursors were selected with an isolation window of 1.6 m/z. Higher-energy collision dissociation fragmentation was performed at a normalized collision energy of 30%. MS/MS spectra were acquired with a target value of 75% ions at a resolution of 15,000 FWHM, inject time set to auto and a fixed first mass of 120 m/z. Peptides with a charge of +1 or greater than 6 or with an unassigned charge state were excluded from fragmentation for MS².

Data analysis

The raw data were processed using MaxQuant software (v.1.6.3.4, http://www.maxquant.org/)¹¹² with label-free quantification (LFQ) and iBAQ enabled¹¹³. MS/MS spectra were searched using the Andromeda search engine against an A. thaliana database (TAIR10_pep_20101214; ftp://ftp.arabidopsis.org/home/tair/Proteins/TAIR10_protein_lists/) and sequences of 248 common contaminant proteins and decoy sequences. Trypsin specificity was required, and a maximum of two missed cleavages were allowed. Minimal peptide length was set to seven amino acids. Carbamidomethylation of cysteine residues was set as fixed, and oxidation of methionine and protein amino-terminal acetylation were set as variable modifications. The match-between-runs option was enabled. Peptide-spectrum-matches and proteins were retained if they were below a false discovery rate of 1% in both cases. Statistical analysis of the MaxLFQ values was carried out using Perseus (v.1.5.8.5, http://www.maxquant.org/). Quantified proteins were filtered for reverse hits, and hits ‘identified by site’ and MaxLFQ values were log₂-transformed. Quantified proteins were grouped by condition, and only those hits were retained that had three valid values in one of the conditions. Missing values were imputed from a normal distribution (1.8 downshift, separately for each column).

PacBio genome sequencing, assembly and analysis

Genomic DNA was isolated using the Macherey-Nagel Bacteria DNA Kit, and the quality of the extracted DNA was assessed with FEMTOpulse and Agilent technologies. A long-insert Tn5 library was then prepared following the LongPlex Long Fragment Multiplexing Kit (seqWell) and pooled before sequencing on the PacBio Revio platform at the Max Planck Genome Centre Cologne. PacBio reads were filtered and subjected to quality control using Filtlong v.0.2.1 (available at https://github.com/rrwick/Filtlong), applying the following parameters: min_length, 1,000; keep_percent, 90; target_bases, 500,000,000. The filtered reads were subsequently assembled with Flye v.2.9.5-b1801 (ref. ¹¹⁴) using parameters that included the following: genome-size, 4 m; asm-coverage, 70; iterations, 4. To ensure accurate start positions for the assembled genomes, the assembly was anchored at the dnaA gene with dnaapler v.0.8.1 (ref. ¹¹⁵). Finally, annotations for the assembled genomes were generated using Bakta v.1.9.4 (ref. ¹¹⁶).

In-frame deletions and mini-Tn5 insertions were confirmed through BLASTP v.2.16.0 (ref. ¹¹⁷) using the default parameters, comparing the wild-type and mutant R179 assembled genomes against the amino acid sequences of dssA and dssB. Further verification of the mutations was conducted via multiple genome alignment using the progressiveMauve algorithm, implemented in Mauve v.2015-02-25 (ref. ¹¹⁸), also with the default parameters. The results from Mauve were manually inspected to exclude any unintended structural rearrangements within the chromosome.

Phylogenetic trees

We compiled a set of 1,765 genomes including all genomes classified as Xanthomonadales found in the NCBI database¹¹⁹ as of October 2019, as well as the available genomes from isolated Arabidopsis strains from the United States and the Cr-, At– and Lj-SPHERE culture collections^5,50,55,57. We also selected 20 P. syringae genomes from the NCBI database as an outgroup. We searched within the final set of 1,785 genome assemblies for the presence of 31 conserved, single-copy marker genes, known as AMPHORA genes¹²⁰, using HMMER (v.3.3.1)¹²¹. Next, the sequences of each of the 31 genes were aligned separately using Clustal Omega (v.1.2.0)¹²². The alignments for each AMPHORA gene were concatenated into continuous alignments, and missing sequences were filled with alignment gaps. On the basis of this concatenated alignment, a maximum likelihood phylogeny was calculated using FastTree (v.2.1.3)¹²³ and sequentially rooted with the Pseudomonas outgroup. The rooted tree was visualized using the Interactive Tree of Life web tool¹²⁴. We predicted orthologous groups of R179 dssA and dssB from protein-coding regions of genomes from the At-R-SPHERE and a selection of Xanthomonadales strains (Supplementary Table 7) using Orthofinder (v.2.3.7, default parameters)¹²⁵. The presence of orthologues of dssA and dssB in all publicly available Xanthomonadales genomes was investigated by querying the coding sequences of dssA and dssB against the coding sequences of 1,772 Xanthomonadales genomes using BLASTX v2.12.0 (ref. ¹¹⁷). Percent identity value > 30 and e < 1 × 10⁻¹⁰ were used as thresholds for true orthologue prediction. The phylogenetic tree was generated as before using AMPHORA marker genes and visualized using the ggtree package (v3.1.4) in R (v4.0.5)¹²⁶. Species, which were members of the orthogroups, were marked on the phylogenetic tree. All software used was executed with the default parameters.

Distribution of reference Xanthomonadales in the MAP

We queried the MAP¹²⁷ for Operational Taxonomic Units (OTUs) clustered at 99% identity (based on the full-length 16S rRNA gene) that belonged to the order Xanthomonadales. We then selected reference OTUs, which we defined as containing a reference genome and having a defined species as well as a type strain. This led to a set of 57 OTUs (40 belonging to the Xanthomonadaceae family and 17 to the Rhodanobacteraceae family). For each of these OTUs, we downloaded the associated abundance table from the MAP website (https://microbeatlas.org/). We grouped the samples by environment type (animal, aquatic, plant, soil and unknown) and, for each OTU, normalized the number of samples from each environment type by the total number of samples from that environment included in the MAP. We then computed the normalized proportion of samples belonging to soil and plant-associated environments for each OTU (Supplementary Table 8).

Distribution of Xanthomonadales MAGs in aquatic, terrestrial and host-associated environments

To estimate the ecological distribution of potentially uncultivated Xanthomonadales lineages, we searched large-scale metagenomic datasets for MAGs belonging to that order, including soil (40,039 MAGs spanning 21,076 species from 3,304 globally distributed soil metagenomes)⁶⁰, leafy greens (910 MAGs from 107 metagenomes across three host species)⁶¹, lake (1,184 MAGs spanning 1,008 species from 308 Canadian lakes)⁶², marine (26,309 MAGs spanning 5,098 species from 1,028 globally distributed, mostly free-living, ocean microbial communities)⁶⁵, rumen (4,941 MAGs across 283 cattle)⁶³ and human (287,002 MAGs spanning 4,645 species from over 10,000 metagenomes)⁶⁴ datasets. For each of these datasets, we selected MAGs belonging to the Xanthomonadales order on the basis of the Genome Taxonomy Database¹²⁸ annotations provided with the release of the dataset. Whenever possible, the number of Xanthomonadales MAGs was also translated into species-level units as defined by the original dataset (Supplementary Table 9).

Establishment of an R179 mini-Tn5 mutant library

R179, E. coli carrying pUTmTn5Km2 (ref. ⁷²) and the conjugation-competent helper E. coli PRk600 were mixed in a 2:1:1 ratio and co-cultivated overnight on 0.5× TSB plates at 25 °C. The conjugation mix was resuspended in 0.5× TSB, supplemented with 20% glycerol and frozen at −80 °C. Subsequently, aliquots of the conjugation mix were plated on 0.5× TSB plates supplemented with 50 µg ml⁻¹ kanamycin and 100 µg ml⁻¹ nitrofurantoin and incubated at 25 °C for 48 h. Afterwards, single colonies were inoculated into 100 µl of 0.5× TSB (same antibiotics) in 96-well culture plates, incubated at 25 °C and 180 rpm for 48 h and immediately frozen at −80 °C after adding 20% glycerol.

High-throughput screening for the loss of flg22-induced RGI

Phytostrip-based high-throughput screening of R179 mini-Tn5 mutants was performed as previously described⁷³. In brief, R179 mini-Tn5-transposon mutants were inoculated in 0.5× TSB in 96-deep-well plates and incubated at 25 °C and 180 rpm for 48 h. Bacteria were pelleted by centrifugation (10 min, 4,200 g) and washed twice with 10 mM MgSO₄. After the last centrifugation, bacterial growth was qualitatively documented by scanning 96-deep-well plates. Washed bacteria were then resuspended in 0.5× MS medium (see ‘Agar-based cultivation of axenic plants and co-cultivation with bacteria, bacterial supernatants and elicitors’) to reach a bacterial concentration of approximately OD₆₀₀ = 0.015. Afterwards, sealed phytostrips were filled with 100 µl of resuspended bacterial culture. Then, 0.5× MS (2.22 g l⁻¹ MS basal salts, Sigma; 0.1 g l⁻¹ MES, BioChemica; pH 5.7; 1.2% (w/v) Phytagel, Sigma-Aldrich) was supplemented with 1 µM flg22 (see ‘Agar-based cultivation of axenic plants and co-cultivation with bacteria, bacterial supernatants and elicitors’), and 230 µl was added to the resuspended bacteria in the phytostrips. Phytostrip-compatible 96-deep-well plates were filled with liquid 0.5× MS medium, and the phytostrips were placed on top after the medium solidified. Finally, five sterilized A. thaliana pWER::FLS2–GFP seeds were added to each well, and the assembled system was placed into an autoclaved rectangular box and incubated for 2 to 3 weeks (see ‘Plant cultivation conditions’). Root length was documented by scanning and qualitatively scored.

Identification of mini-Tn5 insertion sites

Chromosomal mini-Tn5 transposon integration sites were determined similarly as described before using TAIL PCR^72,129. The first PCR was conducted using the primers JO4/JO28, 10 ng of gDNA (see ‘Targeted mutagenesis’) and the DFS-Taq DNA polymerase (Bioron) using the following conditions: 94 °C for 2 min; 94 °C for 30 s, 30 °C for 30 s and 72 °C for 1 min for 6 cycles; 94 °C for 30 s, 45 °C for 30 s and 72 °C for 1 min for 30 cycles; and a final extension at 72 °C for 10 min. For the second PCR, a 1:10 dilution of the PCR product was used with the JO1/JO5 primers. The following conditions were used: 94 °C for 2 min; 94 °C for 30 s, 57 °C for 30 s and 72 °C for 1 min for 30 cycles; and a final extension at 72 °C for 10 min. Finally, the PCR products were separated using gel electrophoresis, and the most prominent band was extracted and Sanger-sequenced with the JO1 primer (Eurofins Scientific). The sequencing results were analysed in CLC Main Workbench (QIAGEN). A list of the oligonucleotides used can be found in Supplementary Table 6.

Targeted mutagenesis

Targeted mutants were generated using the Golden Gate-compatible pOGG2 vector¹³⁰. In brief, R179 DNA was extracted by resuspending a bacterial colony in 25 µl of buffer I (25 mM NaOH, 0.2 mM EDTA, pH 12) followed by incubation at 95 °C for 30 min and the addition of 25 µl of buffer II (40 mM Tris-HCl). Approximately 750-nucleotide sequences flanking dssA and dssB upstream and downstream were amplified using the primers JO105/JO106 and JO107/JO108 and JO184/JO185 and JO186/JO195, respectively (Phusion Hot Start High-Fidelity DNA polymerase; Thermo Scientific, according to the user manual). PCR products were assembled into pOGG2 by a restriction/ligation reaction with BsaI (Thermo Scientific) and T4 DNA ligase (New England Biolabs) as previously described¹³¹. Colony PCR-positive plasmids (DFS-Taq DNA polymerase; Bioron, primer JO127/128)) were extracted using a NucleoSpin Plasmid kit (Macherey-Nagel), and the final vectors were validated by Sanger sequencing (Eurofins Scientific). The sequencing results were analysed in CLC Main Workbench (QIAGEN). Vectors were integrated by R179 conjugation with the respective pOGG2 derivate harbouring E. coli DH5α λ-pir and the E. coli helper PRk600. The conjugation mix was incubated at 25 °C overnight, and transformants/recombinants were identified by sequential selection on 0.5× TSB supplemented with spectinomycin (25 µg ml⁻¹), nitrofurantoin (100 µg ml⁻¹) and 10% sucrose. The R179 mutants were validated using the primers JO147/148 for the deletion of dssA and the primers JO182/JO183 for the deletion of dssB followed by Sanger sequencing.

RNA sequencing for root transcriptomics

Sample preparation and sequencing

The roots of approximately 30 seedlings cultivated on three square plates were pooled for one sample, briefly washed in 10 mM MgSO₄, dried using Whatman paper and immediately frozen in liquid nitrogen in Lysing Matrix E tubes (MP Biomedicals). The roots were homogenized using a TissueLyser II (QIAGEN, 2 × 29 pps for 40 s) with pre-cooled adapters. RNA was extracted using the RNeasy Plant Mini Kit (QIAGEN) according to the user manual with optional on-column DNase digestion, and RNA concentrations were measured using a spectrometer (Nanodrop OneC Microvolume UV-Vis Spectrophotometer). Library preparation and sequencing were conducted at Novogene (Cambridge, UK) using the Illumina HiSeq2500 platform. Approximately 10 million paired-end reads were obtained per sample.

Analysis of the raw data

The quality of the raw FASTQ files was assessed using fastQC (http://www.bioinformatics.babraham.ac.uk/projects/fastqc). Quality-controlled files were mapped to the A. thaliana reference genome version TAIR10 (ref. ¹³²) using STAR (v.2.7.10b)¹³³. The aligned read counts were obtained with featureCounts in the R package Rsubread (v.2.12.0)¹³⁴. The R package sva (v.3.46.0)¹³⁵ was used to account for hidden batch effects. Using the function sva, the number of surrogate variables was determined to be 6. Differential analysis was performed using the R package DESeq2 (v.1.38.0)¹³⁶. To reduce the number of false positive DEGs, only the genes that had DESeq2 normalized counts ≥10 in at least three samples were analysed. The thresholds for DEGs were set to |log₂(fold change)| ≥ 1 and false-discovery-rate-corrected P ≤ 0.05. The functional annotations of genes based on Araport11 annotations¹³⁷ were downloaded from the TAIR website (https://www.arabidopsis.org/).

Clustering and GO term enrichment analysis

The clustering of DEGs was performed by taking the raw log₂(fold change) values with hierarchical clustering, using the euclidean distance and ward.D method as parameters¹³⁸. The number of optimal clusters was identified by using the Silhouette score¹³⁹ calculated by the R package NbClust v.3.0.1 (ref. ¹⁴⁰). The R package ComplexHeatmap (v.2.14.0)¹⁴¹ was used for visualization. The GO term enrichment analyses of the gene clusters were performed using the R package topGO (v.2.50.0)¹⁴² with the enrichment threshold of P = 0.01. The significance of the overlaps between clusters from this study and from the previous study⁴⁴ were assessed by a one-sided Fisher’s exact test, with the threshold of false-discovery-rate-adjusted P = 0.05.

Preparation of microbial extracts and ethylene measurement

Bacteria incubated for 2 days were resuspended at a concentration of 10% (w/v) in ice-cold water. Bacterial cells were pelleted by centrifugation at 4,500 g and 4 °C for >30 min, washed with ice-cold water and lysed by sonication (1 kJ ml⁻¹ over 20 min) on ice. Cell debris was removed by centrifugation at >14,000 g and 4 °C for 20 min. The supernatant was used either without treatment or after heat treatment at 95 °C for 30 min. Serial dilutions of native and heat-treated lysates were prepared in 1% (w/v) BSA and 100 mM NaCl. The same solution without lysate was used for mock treatments.

Ethylene biosynthesis was measured as previously described¹⁴³. In brief, A. thaliana leaves were cut into square pieces (4 × 4 mm²) and floated overnight on water. Three pieces each were then transferred to 6-ml glass tubes containing 500 µl of water. After bacterial lysates (or AtPep1) were added, the reaction tubes were sealed with rubber stoppers and agitated on a shaker at 100 g. At 5 h post treatment, 1 ml of head-space volume was sampled and subjected to gas chromatographic analysis (GC 14 A, by Shimadzu). The gas sample was separated on an alumina column (1.2 m × 3 mm, packed with Al₂O₃ grains) via isocratic elution at 140 °C and with a constant N₂ carrier gas stream (250 kPa). Signals were detected with an FID (50 kPa of H₂, 25 kPa of air). Ethylene peaks were identified according to their retention time, integrated and quantified by comparison to an analytical standard.

Reporting summary

Further information on research design is available in the Nature Portfolio Reporting Summary linked to this article.