Article A single bacterial genus maintains root growth in a complex microbiome

­­­­­­­­­­ https://doi.org/10.1038/s41586-020-2778-7 Omri M. Finkel1,2,9,12, Isai Salas-González1,2,3,12, Gabriel Castrillo1,2,10,12, Jonathan M. Conway1,2,12, Theresa F. Law1,2, Paulo José Pereira Lima Teixeira1,2,11, Ellie D. Wilson1,2, Connor R. Fitzpatrick1,2, Received: 17 January 2020 Corbin D. Jones1,3,4,5,6,7 & Jeffery L. Dangl1,2,3,6,7,8 ✉ Accepted: 3 July 2020

Published online: xx xx xxxx Plants grow within a complex web of species that interact with each other and with the Check for updates plant1–10. These interactions are governed by a wide repertoire of chemical signals, and the resulting chemical landscape of the rhizosphere can strongly afect root health and development7–9,11–18. Here, to understand how interactions between microorganisms infuence root growth in Arabidopsis, we established a model system for interactions between plants, microorganisms and the environment. We inoculated seedlings with a 185-member bacterial synthetic community, manipulated the abiotic environment and measured bacterial colonization of the plant. This enabled us to classify the synthetic community into four modules of co-occurring strains. We deconstructed the synthetic community on the basis of these modules, and identifed interactions between microorganisms that determine root phenotype. These interactions primarily involve a single bacterial genus (Variovorax), which completely reverses the severe inhibition of root growth that is induced by a wide diversity of bacterial strains as well as by the entire 185-member community. We demonstrate that Variovorax manipulates plant hormone levels to balance the efects of our ecologically realistic synthetic root community on root growth. We identify an auxin-degradation operon that is conserved in all available genomes of Variovorax and is necessary and sufcient for the reversion of root growth inhibition. Therefore, metabolic signal interference shapes bacteria–plant communication networks and is essential for maintaining the stereotypic developmental programme of the root. Optimizing the feedbacks that shape chemical interaction networks in the rhizosphere provides a promising ecological strategy for developing more resilient and productive crops.

Plant phenotypes, and ultimately fitness, are influenced by the micro- Plant hormones—in particular, auxins—are both produced and organisms that live in close association with them1–3. These micro- degraded by an abundance of plant-associated microorganisms15–18. organisms—collectively termed the plant microbiota—assemble on Microbially derived auxins can have effects on plants that range from the basis of plant and environmentally derived cues1,4–6, resulting in growth promotion to the induction of disease, depending on context myriad interactions between plants and microorganisms. Beneficial and concentration9. The intrinsic root developmental patterns of the and detrimental microbial effects on plants can be either direct3,7–9, or plant are dependent on finely calibrated auxin and ethylene concentra- an indirect consequence of microorganism–microorganism interac- tion gradients with fine differences across tissues and cell types19; it is tions2,10. Although antagonistic interactions between microorganisms not known how the plant integrates exogenous, microbially derived are known to have an important role in shaping plant microbiota and auxin fluxes into its developmental plan. protecting plants from pathogens2, another potentially important class Here we apply a synthetic community to axenic plants as a proxy of interactions is metabolic signal interference11,12: rather than direct for root-associated microbiomes in natural soils to investigate how antagonism, microorganisms interfere with the delivery of chemical interactions between microorganisms shape plant growth. We estab- signals produced by other microorganisms, which alters plant–micro- lish plant colonization patterns across 16 abiotic conditions to guide organism signalling12–14. stepwise deconstruction of the synthetic community, which led to the

1Department of Biology, University of North Carolina at Chapel Hill, Chapel Hill, NC, USA. 2Howard Hughes Medical Institute, University of North Carolina at Chapel Hill, Chapel Hill, NC, USA. 3Curriculum in Bioinformatics and Computational Biology, University of North Carolina at Chapel Hill, Chapel Hill, NC, USA. 4Department of Genetics, University of North Carolina at Chapel Hill, Chapel Hill, NC, USA. 5Lineberger Comprehensive Cancer Center, University of North Carolina at Chapel Hill, Chapel Hill, NC, USA. 6Carolina Center for Genome Sciences, University of North Carolina at Chapel Hill, Chapel Hill, NC, USA. 7Curriculum in Genetics and Molecular Biology, University of North Carolina at Chapel Hill, Chapel Hill, NC, USA. 8Department of Microbiology and Immunology, University of North Carolina at Chapel Hill, Chapel Hill, NC, USA. 9Present address: Department of Plant and Environmental Sciences, Institute of Life Science, The Hebrew University of Jerusalem, Jerusalem, Israel. 10Present address: Future Food Beacon of Excellence, School of Biosciences, University of Nottingham, Sutton Bonington, UK. 11Present address: Department of Biology, ‘Luiz de Queiroz’ College of Agriculture (ESALQ), University of São Paulo (USP), Piracicaba, Brazil. 12These authors contributed equally: Omri M. Finkel, Isai Salas-González, Gabriel Castrillo, Jonathan M. Conway. ✉e-mail: [email protected]

Nature | www.nature.com | 1 Article

IQR IQR Cut-off RGI abRoot versus Shoot versus No bacteria Full substrate substrate NB Full Single modules Module pairs

Pi NaClpH Pi NaClpH 8 Temp Temp T U Phosphate Salinity VW V UV UV 0 μM 50 mM W Module A 10 M 100 mM 6 μ X 30 μM 150 mM YZ 50 M 200 mM 4 YZ μ Y 100 μM Temperature Z 1,000 μM 10 °C 2 pH 21 °C 5.5 31 °C 7.0 0 8.2 Primary root elongation (cm) NB ABCD ABCD AB AC AD BC BD CD Module B cdSingle modules Module pairs Representative root growth inhibitors Module ABCDStrain Genus MF455 Mycobacterium MF220A Sphingomonas CL71 Acinetobacter MF45

Module C MF397 Pseudomonas ACDAC AD CL144 e Agar Soil Ralstonia CL21 8 YY MF384 Burkholderia CL11 6 YY MF4

Module D CL14 4 MF369 Z Z MF349 Variovorax 2

Module A isolates MF110 MF350 0 Primary root elongation (cm) MF160 Signi†cance q < 0.05 and MF375 CL14 CL28CL14CL28 CL14CL28CL14CL28 estimate > 2 5 MF278 Signi†cance –5.02.5 0 2. 5.0 True Actinobacteria Alphaproteobacteria Self q < 0.05 Bacteroidetes Firmicutes Fold change Gammaproteobacteria Primary root elongation (cm) CL28 MF48 MF107 MF224 Module A Module B Module C Module D Bacillus Burkholderiaceae Bacillaceae Rhizobiaceae Microbacteriaceae 0 2 4 6 8

Comamonadaceae Paenibacillaceae Micrococcaceae Arthrobacter Pseudomonas Moraxellaceae Streptomycetaceae Mycobacteriaceae Agrobacterium Nocardiaceae

Fig. 1 | Arabidopsis root length is governed by bacteria–bacteria correspond to a Tukey post hoc test. n = 75, 89, 68, 94, 87, 77, 76, 96, 82, 84, 89 interactions within a community. a, Fraction enrichment patterns of the and 77 (from left to right) biological replicates across 2 independent synthetic community across abiotic gradients. Each row represents a unique experiments. c, Binarized image of representative seedlings inoculated with sequence. The four modules of co-occurring strains (A, B, C and D) are modules A, C and D, and with module combinations A–C and A–D. d, Heat map represented. Dendrogram tips are coloured by taxonomy. Heat maps are coloured by average primary root elongation of seedlings inoculated with four coloured by log2-transformed fold changes derived from a fitted generalized representative RGI-inducing strains from each module (columns A–D) in linear model and represent enrichments in plant tissue (root or shoot) combination with isolates from module A (rows) or alone (self). Significance compared with substrate. Comparisons with false-discovery-rate was determined via ANOVA. e, Primary root elongation of seedlings inoculated (FDR)-corrected q-value < 0.05 are contoured in black. Enriched families within with Arthrobacter CL28 and Variovorax CL14 individually or jointly across two each module are listed below the heat map. n = 6 biological replicates across substrates. Significance was determined via ANOVA, letters are the results of a 2 independent experiments. Pi, inorganic phosphate; temp, temperature. Tukey post hoc test. n = 64, 64, 63, 17, 36 and 33 (from left to right) biological b, Primary root elongation of seedlings grown axenically (no bacteria, NB), with replicates across 2 independent experiments. In all box plots, the centre line the full synthetic community (ABCD) or its subsets: modules A, B, C and D alone represents the median, box edges show the 25th and 75th percentiles, and (single modules), and all pairwise combination of modules (module pairs). whiskers extend to 1.5× the interquartile range (IQR). Significance was determined via analysis of variance (ANOVA); letters identification of multiple levels of microorganism–microorganism composed of the major root-associated phyla1,4–6,20 (Extended Data Fig. 1a, interaction that interfere with the additivity of bacterial effects on root Supplementary Table 1). We exposed this microcosm to each of 16 abiotic growth. We demonstrate that a single bacterial genus (Variovorax) is contexts by manipulating one of four variables (salinity, temperature, a required for maintaining the intrinsically controlled developmental previously reported phosphate concentration gradient5 and pH). We meas- programme of the root, by tuning its chemical landscape. We establish ured the composition of the synthetic community in root, shoot and agar Variovorax as a core taxon in the root microbiota of diverse plants fractions 12 days after inoculation using 16S rRNA amplicon sequencing. grown in diverse soils. Finally, we identify a locus conserved across The composition of the resulting root and shoot microbiota reca- Variovorax strains that is responsible for this phenotype. pitulated phylum-level plant-enrichment patterns seen in soil-planted Arabidopsis1 (Extended Data Fig. 1b). We validated the patterns observed in the agar-based system using seedlings grown in sterilized Microbial interactions control root growth potting soil21 that were inoculated with the same synthetic community. To model plant–microbiota interactions in a fully controlled setting, we Both relative abundance and plant-enrichment patterns at the unique established a plant–microbiota microcosm that represents the native sequence level were significantly correlated between the agar- and bacterial root-derived microbiota on agar plates. We inoculated 7-day-old soil-based systems, which confirms the applicability of our relatively seedlings with a defined 185-member bacterial synthetic community5 high-throughput agar-based system as a model for the assembly of

2 | Nature | www.nature.com NB Full –Burk –Vario –Vario, –Burk

a c d JM agar MS agar Clay Soil –21 –11 –3 –3 C q = 4.1 × 10 q = 6.1 × 10 q = 6.6 × 10 q = 1.1 × 10 B B 6 6

4 4 b AA

2 2

Primary root elongation (cm) 0 Primary root elongation (cm) 0

, o o NB Full –Vario ri ri a a ario ario NB Full ario ario Full Full Full Full –V –V –V –V –Burk–V –V –Burk

e Module A Module C Module D 34-member f 10 μM Pi 150 mM NaCl pH 8.231 °C g Shoot P = 0.12 q = 6.9 × 10–4 q = 2.0 × 10–17 q = 1.4 × 10–23 q = 3.7 × 10–37 q = 2.4 × 10–6 q = 5.5 × 10–3 q = 1.4 × 10–21 q = 1.7 × 10–12 Substrate

6 6 Root CAP2(19.45%) 4 4 CAP1 (45.86%) h Variovorax RA (%) 6 2 2 4 2

Primary root elongation (cm ) 0 0 o l o i rio ri io Substrate RootShoot ario ar ario a ario ario ar ario ario ario ario Full Ful Va Full Full +V –V +V –V +V –V +V –V –V – –V –V

Fig. 2 | Variovorax maintains stereotypic root development. a, Phylogenetic Variovorax isolates. n = 40, 40, 15, 19, 22, 26, 25 and 29 (from left to right) tree of 185 bacterial isolates. Concentric rings represent isolate composition of biological replicates across 2 independent experiments. f, Primary root each synthetic community treatment. NB, uninoculated (no bacteria); full, full elongation of seedlings inoculated with the full synthetic community or with synthetic community; −Burk, Burkholderia drop-out; −Vario, Variovorax the Variovorax drop-out synthetic community across different abiotic drop-out; −Vario, −Burk, Burkholderia and Variovorax drop-out. b, Binarized stresses: low phosphate, high salt, high pH and high temperature. n = 43, 45, 37, image of representative seedlings inoculated or not with the full synthetic 37, 43, 44, 44 and 43 (from left to right) biological replicates across community or the Variovorax drop-out synthetic community. c, Primary root 2 independent experiments. Significance was determined within each elongation of seedlings grown axenically (NB) or inoculated with the different condition via ANOVA in d, f, and with a two-sided t-test in e. FDR-adjusted synthetic community treatments outlined in a. Significance was determined P values are displayed. g, Canonical analysis of principal coordinates (CAP) via ANOVA, letters correspond to a Tukey post hoc test. n = 26, 26, 24, 37 and 29 scatter plots comparing the compositions of the full synthetic community and (from left to right) biological replicates. d, Primary root elongation of Variovorax drop-out synthetic community (colour code as in a) across seedlings inoculated with the full synthetic community or with the Variovorax fractions (substrate (squares), root (circles) and shoot (triangles)). drop-out synthetic community across different substrates: Johnson medium Permutational multivariate ANOVA (PERMANOVA) P value is shown. n = 7 (JM) agar, Murashige and Skoog (MS) agar, sterilized clay or potting soil. n = 36, (substrate + full), 8 (substrate + −Vario), 6 (root + full), 6 (root + −Vario), 5 (shoot 47, 21, 24, 43, 48, 33 and 36 (from left to right) biological replicates across + full) or 5 (shoot + −Vario). h, Relative abundance (RA) of the Variovorax genus 2 independent experiments. e, Primary root elongation of seedlings inoculated within the full synthetic community across the agar, root and shoot fractions. with four subsets of the full synthetic community (module A, C, D or a n = 127, 119 and 127 biological replicates for agar, root and shoot, respectively, previously described 35-member synthetic community with its single across 16 independent experiments. Variovorax strain removed (34-member)1), with (+Vario) or without (−Vario)

plant microbiota (Extended Data Fig. 1c). Within the agar system, both and D singly or in all six possible pairwise combinations, and imaged fraction (substrate, root or shoot) and abiotic conditions significantly the seedlings 12 days after inoculation. We observed strong primary affected α- and β-diversity (Extended Data Fig. 1d–f). root growth inhibition (RGI) in seedlings inoculated with plant-enriched To guide the deconstruction of the synthetic community into mod- module C or D (Fig. 1b, c). RGI did not occur in seedlings inoculated with ules, we calculated pairwise correlations in relative abundance across module A or B, which do not contain plant-enriched strains (Fig. 1b). all samples, and identified four well-defined modules of co-occurring To test whether the root phenotype derived from each module is an strains that we termed modules A, B, C and D (Fig. 1a, Supplementary additive outcome of its individual constituents, we inoculated seedlings Table 2). These modules formed distinct phylogenetically structured in mono-association with each of the 185 members of the synthetic guilds in association with the plant. Module A contained mainly Gam- community. We observed that 34 taxonomically diverse strains, dis- maproteobacteria and was predominantly more abundant in the sub- tributed across all 4 modules, induced RGI (Extended Data Fig. 2a–c). strate than in the seedling; module B contained mainly low-abundance However, neither the full synthetic community nor derived synthetic Firmicutes, with no significant enrichment trend; and modules C and communities that consisted of module A or B exhibited RGI (Fig. 1b). D were composed mainly of Alphaproteobacteria and Actinobacte- Thus, binary plant–microorganism interactions were not predictive ria, respectively, and showed plant enrichment across all abiotic con- of interactions in this complex-community context. ditions. Both Alphaproteobacteria (module C) and Actinobacteria In seedlings inoculated with module pairs, we observed epistatic (module D) are consistently plant enriched across plant species4, which interactions: in the presence of module A, the RGI caused by modules C suggests that these clades contain plant-association traits that are and D was reverted (Fig. 1b, c). Thus, by deconstructing the synthetic deeply rooted in their evolutionary histories. community into four modules, we found that bacterial effects on root We next asked whether the different modules of co-occurring strains growth are governed by multiple levels of microorganism–microorgan- have different roles in determining plant phenotype. We inoculated ism interaction. This is exemplified by at least four instances: within seedlings with synthetic communities composed of modules A, B, C module A or B and between module A and module C or D. Because three

Nature | www.nature.com | 3 Article

Arthrobacter IAA 2,4-D ACC Arthrobacter –Vario a CL28 b CL28 8 1.25 NB mean CCC 1.00 IQR 6 C BC CC C B C C Arthrobacter B BC B A A 0.75 BB (CL28) BB + 4 Variovorax AA AA D CD 0.50 BC A A (CL14) 2 BC

ot elongation (cm) AB A root elongation 0.25 IQR

0 Standardized primary 0 Full P B4 B4 B4 B4 Sel f Sel f Sel f Sel f Col-0 Col-0 CL11 CL14 CL11 CL14 CL11 CL14 CL11 CL14 axr2-1 + MCP axr2-1 Primary ro + MC YR216 YR216 YR216 YR216 MF160 MF160 MF160 MF160

Col-0 + MCP Col-0 + MCP Self Burkholderia Variovorax axr2-1 axr2-1

Fig. 3 | Variovorax suppression of RGI is related to auxin and ethylene signalling. (Col-0 + MCP) or auxin- and ethylene-unresponsive (axr2-1 + MCP) seedlings a, Primary root elongation of seedlings grown with RGI-inducing Arthrobacter inoculated with RGI-inducing Arthrobacter CL28 or the Variovorax drop-out CL28 or three hormonal treatments (100 nM IAA, 2,4-dichlorophenoxyacetic synthetic community (−Vario). The blue dotted line marks the relative mean acid (2,4-D) or ACC), individually (self) or with Burkholderia CL11 or one of four length of uninoculated seedlings. Horizontal shading is the IQR of seedlings Variovorax isolates (CL14, MF160, B4 or YR216). Significance was determined grown with Arthrobacter CL28 +Variovorax CL14 (aqua) or the full synthetic within each treatment via ANOVA; letters correspond to a Tukey post hoc test. community (grey). Significance was determined via ANOVA; letters n = 74, 46, 61, 48, 49, 49, 45, 44, 46, 43, 49, 40, 22, 19, 22, 19, 20, 25, 28, 30, 29, 29, correspond to a Tukey post hoc test. n = 37, 25, 24, 23, 35, 21, 22 and 20 (from left 29 and 29 (from left to right) biological replicates across 2 independent to right) biological replicates across 2 independent experiments. In a, b, data experiments. b, Primary root elongation, standardized to axenic conditions, represent the mean ± 95% confidence interval. of wild-type (Col-0), auxin-unresponsive (axr2-1), ethylene-unresponsive of these interactions involve module A, we predicted that this module length of root network of the plant and its shoot size (Extended Data contains strains that strongly attenuate RGI and preserve stereotypic Fig. 4c, d). Importantly, the latter is considered a reliable proxy for root development. relative plant fitness22,23,which suggests that Variovorax-mediated suppression of RGI is adaptive. To ascertain the breadth of the ability of Variovorax to attenuate RGI, Variovorax maintain stereotypic root growth we tested additional Variovorax strains from across the phylogeny of To identify strains within module A that are responsible for intra- and this genus (Extended Data Fig. 5a, Supplementary Table 1). All 19 of the intermodule attenuation of RGI, we reduced our system to a tripar- Variovorax strains we tested reverted the RGI induced by Arthrobacter tite plant–microorganism–microorganism system. We individually CL28. A strain from the nearest plant-associated outgroup to this genus screened the 18 non-RGI strains from module A for their ability to (Acidovorax root21924) did not revert RGI (Extended Data Fig. 5a, b). attenuate RGI caused by representative strains from all 4 modules. Thus, all tested strains—representing the broad phylogeny of Variovo- We found that all the tested strains from the genus Variovorax (fam- rax—interact with a wide diversity of bacteria to enforce stereotypic ily Comamonadaceae) suppressed the RGI caused by representative root development within complex communities, independent of biotic RGI-inducing strains from module C (Agrobacterium MF224) and or abiotic contexts. Importantly, we found no evidence that this phe- module D (Arthrobacter CL28) (Fig. 1d). The strains from modules A notype is achieved by outcompeting or antagonizing RGI-inducing (Pseudomonas MF48) and B (Bacillus MF107) were not suppressed strains (Fig. 2g, h, Extended Data Fig. 6). by Variovorax, but rather by two closely related Burkholderia strains (CL11 and MF384) (Fig. 1d). A similar pattern was observed when we screened two selected RGI-suppressing Variovorax strains (CL14 and Variovorax manipulates auxin and ethylene MF160) and Burkholderia CL11 against a diverse set of RGI-inducing To study the mechanisms that underlie bacterial effects on root growth, strains. Variovorax attenuated 13 of the 18 RGI-inducing strains that we analysed the transcriptomes of seedlings colonized for 12 days with we tested (Extended Data Fig. 3a). the RGI-inducing strain Arthrobacter CL28 and the RGI-suppressing To test whether the RGI induction and suppression we observed strain Variovorax CL14, either in mono-association with the seedling or on agar occur in soil as well, we germinated Arabidopsis on sterile in a tripartite combination (Fig. 1e). We also performed RNA sequencing soil inoculated with an RGI-suppressing and -inducing pair of strains: (RNA-seq) on seedlings colonized with the full synthetic community (no the RGI-inducing Arthrobacter CL28 and the RGI-suppressing Vari- RGI) or the Variovorax drop-out synthetic community (RGI) (Extended ovorax CL14. As expected, Arthrobacter CL28 induced RGI, which Data Fig. 7a). Eighteen genes were significantly induced only under RGI was reverted by Variovorax CL14 in soil (Fig. 1e). We generalized this conditions (RGI-induced) across both experiments (Extended Data Fig. 7a, observation by showing that Variovorax-mediated attenuation of b). Seventeen of these genes are co-expressed with genes that have pro- RGI extended to tomato seedlings, in which Variovorax CL14 reverted posed functions related to the root apex25 (Extended Data Fig. 7b, c). The Arthrobacter CL28-mediated RGI (Extended Data Fig. 3b). Finally, we remaining gene, GH3.2, encodes indole-3-acetic acid-amido synthetase, tested whether the RGI-suppressing strains maintain their capacity which conjugates excess amounts of the plant hormone auxin and is a to attenuate RGI in the context of the full 185-member community. robust marker for late auxin responses26,27 (Extended Data Fig. 7b). The We compared the root phenotype of seedlings exposed to either the production of auxins is a well-documented mechanism by which bacteria full synthetic community or to the same community after dropping modulate plant root development13. Indeed, the top 12 auxin-responsive out all ten Variovorax strains and/or all six Burkholderia strains pre- genes from a previous RNA-seq study examining acute auxin response sent in the synthetic community (Fig. 2a). We found that Variovorax in Arabidopsis26 exhibited an average transcript increase in seedlings is necessary and sufficient to revert RGI within the full community exposed to our RGI-inducing conditions (Extended Data Fig. 7d). We (Fig. 2b, c, Extended Data Fig. 4a). This result was robust across a range hypothesized that the suppression of RGI by Variovorax is probably medi- of substrates (including soil), and under various biotic and abiotic ated by interference with bacterially produced auxin signalling. contexts (Fig. 2d–f, Extended Data Fig. 4b). Further, the presence We asked whether the suppression of RGI by Variovorax is directly of Variovorax in the synthetic community increases both the total and exclusively related to auxin signalling. Besides auxin, other small

4 | Nature | www.nature.com a Scaffold Ga0102008_10005 Variovorax CL14 Fold change (log2(CL14–CL28/CL14)) –3 –2 –1 012 3 45,000 50,000 55,000 60,000 65,000 70,000

53 5455 56 57 58 59 60 61 62 63 64 65 66 67 68 69 70 71 72 73 74 75 76 77 → Vector 1 (V1) Vector 2 (V2) ΔHS33 Treatment No bacteria Acidovorax Root219::V2 Variovorax CL14 ΔHS33 Acidovorax root219::EV Variovorax CL14 b c d )

–1 8 CC 8 C

0.10 6 6

B 4 B 4 0.05 A AA 2 2

0 0 0 IAA concentration (mg ml

0 20 40 60 Primary root elongation (cm) Primary root elongation (cm) Time (h) 2 CL14riovoraxHS33 CL14 HS33 Δ VariovoraxVariovoraxΔ AcidovoraxAcidovoraxVariovoraxVa AcidovoraxAcidovorax ot219::EV oot219::EVoot219::V2 ro root219::V r r CL14 CL14

Fig. 4 | An auxin-degrading operon in Variovorax is required for root b, In vitro degradation of IAA by Acidovorax root219::EV, Acidovorax development. a, A map of the auxin-degrading hotspot 33. Genes are Root219::V2, Variovorax CL14 and Variovorax CL14 ΔHS33. n = 3 biological annotated with the last two digits of their IMG gene identifier (26436136XX) replicates. c, d, Primary root elongation of seedlings treated with IAA (c) or

(Extended Data Fig. 9b), and are coloured by the log2-transformed fold change co-inoculated with Arthrobacter CL28 combined with Acidovorax root219::EV, in their transcript abundance in Variovorax CL14 cocultured with Arthrobacter Acidovorax root219::V2, Variovorax CL14 and Variovorax CL14 ΔHS33 (d). CL28 (CL14–CL28) relative to Variovorax CL14 (CL14) in monoculture Significance was determined via ANOVA; letters correspond to a Tukey post (measured by RNA-seq). Genes contained in vector 1 (V1) and vector 2 (V2), and hoc test. c, n = 49, 46, 46 and 49 (from left to right); d, n = 51, 41, 52 and 53 (from the region knocked-out in Variovorax CL14 ΔHS33, are shown below the map. left to right) biological replicates across 2 independent experiments.

molecules cause RGI. These include the plant hormones ethylene28 absence of Variovorax, a complex synthetic community can induce and cytokinin29, as well as microbial-associated molecular patterns severe morphological changes in root phenotypes via both auxin- and such as the flagellin-derived peptide flg2230. We tested the ability of ethylene-dependent pathways, but both are reverted when Variovorax diverse Variovorax strains and of the Burkholderia strain CL11 to revert is present. the RGI induced by auxins (indole-3-acetic acid (IAA) and the auxin To identify the bacterial mechanism or mechanisms that are involved analogue 2,4-dichlorophenoxyacetic acid), ethylene (the ethylene in the attenuation of RGI, we compared the genomes of the 10 Vari- precursor 1-aminocyclopropane-1-carboxylic acid (ACC)), cytokinins ovorax strains in the synthetic community to the genomes of the other (zeatin and 6-benzylaminopurine) and the flg22 peptide. All of the 175 members of the synthetic community. Using de novo orthologue Variovorax strains we tested suppressed the RGI induced by IAA or clustering across all 185 genomes, we identified 947 genes unique to ACC (Fig. 3a)—with the exception of Variovorax YR216, which did not Variovorax, with <5% prevalence across the 175 non-Variovorax members suppress ACC-induced RGI and does not contain an ACC deaminase of the synthetic community and 100% prevalence among all 10 Vari- gene, a plant-growth-promoting feature that is associated with this ovorax strains. We grouped these genes into regions of physically genus28 (Extended Data Fig. 5a). Burkholderia CL11 only partially contiguous genes (genomic hotspots) and focused on the 12 hotspots reverted ACC-induced RGI (Fig. 3a). None of the Variovorax strains that contained at least 10 genes unique to Variovorax (Extended Data attenuated the RGI induced by 2,4-dichlorophenoxyacetic acid, Fig. 9a, Supplementary Table 3). One of these hotspots (designated flg22 or cytokinins (Fig. 3a, Extended Data Fig. 8a). Importantly, this hotspot 33) contains weak homologues (average of about 30% identity) function is mediated by recognition of auxin by Variovorax and not by to the genes iacC, iacD, iacE, iacF and iacR of the IAA-degrading iac the plant auxin response per se, as the auxin response (RGI) induced operon of Paraburkholderia phytofirmans strain PsJN18, but lacks iacA, by 2,4-dichlorophenoxyacetic acid is not reverted. Indeed, we found iacB and iacI—which are known to be necessary for Paraburkholderia that Variovorax CL14 degrades IAA in culture (Extended Data Fig. 8b) growth on IAA17 (Fig. 4a, Extended Data Fig. 9b). To test whether the and quenches fluorescence of the Arabidopsis auxin reporter line hotspots we identified are responsive to RGI-inducing bacteria, we DR5::GFP caused by the RGI-inducing Arthrobacter CL28 (Extended analysed the transcriptome of Variovorax CL14 in monoculture and Data Fig. 8c, d). in coculture with the RGI-inducing Arthrobacter CL28. We observed Auxin and ethylene are known to act synergistically to inhibit root extensive transcriptional reprogramming of Variovorax CL14 when growth31. To ascertain the roles of both auxin and ethylene percep- cocultured with Arthrobacter CL28 (Supplementary Table 4). Among the tion by the plant in responding to RGI-inducing strains, we used the 12 hotspots we identified, the genes in hotspot 33 were the most highly auxin-insensitive axr2-1 mutant32 combined with a competitive inhibitor upregulated (Fig. 4a, Extended Data Fig. 9b, c). We thus hypothesized of ethylene receptors, 1-methylcyclopropene (1-MCP)33. We inoculated that hotspot 33 contains an uncharacterized auxin-degradation operon. wild-type seedlings and axr2-1 mutants, treated or not with 1-MCP, with In parallel, we constructed a Variovorax CL14 genomic library in the RGI-inducing Arthrobacter CL28 strain or the Variovorax drop-out Escherichia coli with >12.5-kb inserts in a broad host-range vector, and synthetic community. We observed in both cases that bacterial RGI is screened the resulting E. coli clones for auxin degradation. Two clones reduced in axr2-1 and 1-MCP-treated wild-type seedlings, and is fur- from the approximately 3,500 that we screened degraded IAA (denoted ther reduced in doubly insensitive 1-MCP-treated axr2-1 seedlings; this V1 and V2) (Supplementary Table 5). The Variovorax CL14 genomic demonstrates that both auxin and ethylene perception in the plant inserts in both of these clones contained portions of hotspot 33 (Fig. 4a, contribute additively to bacterially induced RGI (Fig. 3b). Thus, in the Extended Data Fig. 9b). The overlap common to both of these clones

Nature | www.nature.com | 5 Article contained nine genes, among them the weak homologues to Parabur- 1. Castrillo, G. et al. Root microbiota drive direct integration of phosphate stress and immunity. Nature 543, 513–518 (2017). kholedria iacC, iacD and iacE. To test whether this genomic region is 2. Durán, P. et al. Microbial interkingdom interactions in roots promote Arabidopsis survival. sufficient to revert RGI in plants, we transformed Acidovorax root219, a Cell 175, 973–983.e14 (2018). relative of Variovorax that does not cause or revert RGI (Extended Data 3. Herrera Paredes, S. et al. Design of synthetic bacterial communities for predictable plant phenotypes. PLoS Biol. 16, e2003962 (2018). Fig. 5a, b), with the shorter functional insert (V2) (Fig. 4a, Extended Data 4. Fitzpatrick, C. R. et al. Assembly and ecological function of the root microbiome across Fig. 9b) or with an empty vector (EV). The resulting gain-of-function angiosperm plant species. Proc. Natl Acad. Sci. USA 115, E1157–E1165 (2018). strain Acidovorax root219::V2 gained the ability to degrade IAA in cul- 5. Finkel, O. M. et al. The effects of soil phosphorus content on plant microbiota are driven by the plant phosphate starvation response. PLoS Biol. 17, e3000534 (2019). ture (Fig. 4b). We inoculated Acidovorax root219::V2 or the control 6. Thiergart, T. et al. Root microbiota assembly and adaptive differentiation among Acidovorax root219::EV onto plants treated with IAA or inoculated European Arabidopsis populations. Nat. Ecol. Evol. 4, 122–131 (2020). with the RGI-inducing Arthrobacter CL28. Acidovorax root219::V2 fully 7. Hogenhout, S. A., Van der Hoorn, R. A. L., Terauchi, R. & Kamoun, S. Emerging concepts in effector biology of plant-associated organisms. Mol. Plant Microbe Interact. 22, 115–122 reverted IAA-induced RGI (Fig. 4c) and partially reverted Arthrobacter (2009). CL28-induced RGI, despite colonizing roots at significantly lower levels 8. Mylona, P., Pawlowski, K. & Bisseling, T. Symbiotic nitrogen fixation. Plant Cell 7, 869–885 than Variovorax CL14 (Fig. 4d, Extended Data Fig. 9d). In addition, we (1995). 9. Ludwig-Müller, J. Bacteria and fungi controlling plant growth by manipulating auxin: deleted hotspot 33 from Variovorax CL14 (Fig. 4a) to test whether this balance between development and defense. J. Plant Physiol. 172, 4–12 (2015). putative operon is necessary for the reversion of RGI. The resulting 10. Carlström, C. I. et al. Synthetic microbiota reveal priority effects and keystone strains in strain Variovorax CL14 ΔHS33—which is not impaired in plant coloniza- the Arabidopsis phyllosphere. Nat. Ecol. Evol. 3, 1445–1454 (2019). 11. Faure, D., Vereecke, D. & Leveau, J. H. J. Molecular communication in the rhizosphere. tion (Extended Data Fig. 9e)—did not degrade IAA in culture (Fig. 4b), Plant Soil 321, 279–303 (2009). and did not revert IAA-induced (Fig. 4c) or Arthrobacter CL28-induced 12. Leadbetter, J. R. & Greenberg, E. P. Metabolism of acyl-homoserine lactone RGI (Fig. 4d). Thus, this Variovorax-specific gene cluster is necessary quorum-sensing signals by Variovorax paradoxus. J. Bacteriol. 182, 6921–6926 (2000). 13. Leveau, J. H. J. & Lindow, S. E. Utilization of the plant hormone indole-3-acetic acid for for the suppression of RGI and for auxin degradation. It is thus the criti- growth by Pseudomonas putida strain 1290. Appl. Environ. Microbiol. 71, 2365–2371 cal genetic locus required by Variovorax to maintain stereotypic root (2005). development in the context of a phylogenetically diverse microbiome. 14. Zúñiga, A. et al. Quorum sensing and indole-3-acetic acid degradation play a role in colonization and plant growth promotion of Arabidopsis thaliana by Burkholderia phytofirmans PsJN. Mol. Plant Microbe Interact. 26, 546–553 (2013). 15. Sun, S.-L. et al. The plant growth-promoting rhizobacterium Variovorax boronicumulans Conclusions CGMCC 4969 regulates the level of indole-3-acetic acid synthesized from indole- 3-acetonitrile. Appl. Environ. Microbiol. 84, e00298-18 (2018). Signalling molecules and other secondary metabolites are products 16. Gilbert, S. et al. Bacterial production of indole related compounds reveals their role in of adaptations that allow microorganisms to survive competition for association between duckweeds and endophytes. Front. Chem. 6, 265 (2018). primary metabolites. Our results illuminate the importance of a trophic 17. Donoso, R. et al. Biochemical and genetic bases of indole-3-acetic acid (auxin phytohormone) degradation by the plant-growth-promoting rhizobacterium layer of microorganisms that use these secondary metabolites for Paraburkholderia phytofirmans PsJN. Appl. Environ. Microbiol. 83, e01991-16 (2016). their own benefit, while potentially providing the unselected exapta- 18. Leveau, J. H. J. & Gerards, S. Discovery of a bacterial gene cluster for catabolism of the tion34 of interfering with signalling between the bacterial microbiota plant hormone indole 3-acetic acid. FEMS Microbiol. Ecol. 65, 238–250 (2008). 19. Brumos, J. et al. Local auxin biosynthesis is a key regulator of plant development. Dev. and the plant host. Such metabolic signal interference has previously Cell 47, 306–318.e5 (2018). been demonstrated in the case of quorum quenching12, the degrada- 20. Levy, A. et al. Genomic features of bacterial adaptation to plants. Nat. Genet. 50, 138–150 tion of microbial-associated molecular patterns35 and the degrada- (2018). 21. Kremer, J. M. et al. FlowPot axenic plant growth system for microbiota research. Preprint 13–15 tion of bacterially produced auxin, including among Variovorax . at https://www.biorxiv.org/content/10.1101/254953v1 (2018). Plant development relies on tightly regulated auxin concentration 22. Clauss, M. J. & Aarssen, L. W. Phenotypic plasticity of size–fecundity relationships in gradients19, which can be distorted by auxin fluxes emanating from Arabidopsis thaliana. J. Ecol. 82, 447 (1994). 23. Abreu, M. E. & Munné-Bosch, S. Salicylic acid deficiency in NahG transgenic lines and the microbiota. Some Variovorax strains have the capacity to both sid2 mutants increases seed yield in the annual plant Arabidopsis thaliana. J. Exp. Bot. 60, produce and degrade auxin, which suggests a capacity to fine-tune 1261–1271 (2009). auxin concentrations in the rhizosphere15,28. 24. Bai, Y. et al. Functional overlap of the Arabidopsis leaf and root microbiota. Nature 528, 364–369 (2015). We have shown here that the chemical homeostasis enforced by the 25. Klepikova, A. V., Kasianov, A. S., Gerasimov, E. S., Logacheva, M. D. & Penin, A. A. A high presence of Variovorax in a phylogenetically diverse, realistic synthetic resolution map of the Arabidopsis thaliana developmental transcriptome based on community allows the plant to maintain its developmental programme RNA-seq profiling. Plant J. 88, 1058–1070 (2016). 26. Uchida, N. et al. Chemical hijacking of auxin signaling with an engineered auxin–TIR1 pair. within a chemically complex matrix. Variovorax was recently found to Nat. Chem. Biol. 14, 299–305 (2018). have the rare property of improved plant colonization upon late arrival 27. Takase, T. et al. ydk1-D, an auxin-responsive GH3 mutant that is involved in hypocotyl and 10 root elongation. Plant J. 37, 471–483 (2004). to an established community , which suggests that they use bacte- 28. Chen, L., Dodd, I. C., Theobald, J. C., Belimov, A. A. & Davies, W. J. The rhizobacterium rially produced or induced—rather than plant-derived—substrates. Variovorax paradoxus 5C-2, containing ACC deaminase, promotes growth and Furthermore, after re-analysing a recent large-scale time and spatially development of Arabidopsis thaliana via an ethylene-dependent pathway. J. Exp. Bot. 64, 6 1565–1573 (2013). resolved survey of the Arabidopsis root microbiome and a common 29. Cary, A. J., Liu, W. & Howell, S. H. Cytokinin action is coupled to ethylene in its effects on 4 garden experiment including 30 plant species , we noted that Variovo- the inhibition of root and hypocotyl elongation in Arabidopsis thaliana seedlings. Plant rax is among a limited group of core bacterial genera found in 100% of Physiol. 107, 1075–1082 (1995). 30. Gómez-Gómez, L., Felix, G. & Boller, T. A single locus determines sensitivity to bacterial the sampled sites and plant species (Extended Data Fig. 10a, b). These flagellin in Arabidopsis thaliana. Plant J. 18, 277–284 (1999). ecological observations, together with our results using a reductionist 31. Robles, L., Stepanova, A. & Alonso, J. Molecular mechanisms of ethylene–auxin microcosm, reinforce the importance of Variovorax as a key player in interaction. Mol. Plant 6, 1734–1737 (2013). 32. Nagpal, P. et al. AXR2 encodes a member of the Aux/IAA protein family. Plant Physiol. 123, the bacteria–bacteria–plant communication networks that are required 563–574 (2000). to maintain root growth within a complex biochemical ecosystem. 33. Hall, A. E., Findell, J. L., Schaller, G. E., Sisler, E. C. & Bleecker, A. B. Ethylene perception by the ERS1 protein in Arabidopsis. Plant Physiol. 123, 1449–1458 (2000). 34. Gould, S. J. & Vrba, E. S. Exaptation—a missing term in the science of form. Paleobiology 8, 4–15 (1982). Online content 35. Bardoel, B. W. et al. Pseudomonas evades immune recognition of flagellin in both Any methods, additional references, Nature Research reporting sum- mammals and plants. PLoS Pathog. 7, e1002206 (2011). maries, source data, extended data, supplementary information, Publisher’s note Springer Nature remains neutral with regard to jurisdictional claims in acknowledgements, peer review information; details of author con- published maps and institutional affiliations. tributions and competing interests; and statements of data and code availability are available at https://doi.org/10.1038/s41586-020-2778-7. © The Author(s), under exclusive licence to Springer Nature Limited 2020

6 | Nature | www.nature.com Methods using a tissue homogenizer (MPBio). Agar from each plate was collected in 30-ml syringes with a square of sterilized Miracloth (Millipore) at No statistical methods were used to predetermine sample size. The the bottom and kept at −20 °C for 1 week. Syringes were then thawed experiments were randomized, and investigators were not blinded to at room temperature and samples were squeezed gently through the allocation during experiments and outcome assessment. Miracloth into 50-ml falcon tubes. Samples were centrifuged at maxi- mum speed for 20 min and most of the supernatant was discarded. The Arabidopsis with bacterial synthetic community microcosm remaining 1–2 ml of supernatant, containing the pellet, was transferred across four stress gradients into clean 1.5-ml Eppendorf tubes. Samples were centrifuged again, This section relates to Fig. 1a, Extended Data Fig. 1. supernatant was removed and pellets were stored at −80 °C until DNA extraction. DNA extractions were carried out on ground root and shoot Bacterial culture and plant inoculation. The 185-member bacterial tissue and agar pellets using 96-well-format MoBio PowerSoil Kit (MO- synthetic community used here contains genome-sequenced isolates BIO Laboratories; Qiagen) following the manufacturer’s instruction. obtained from surface-sterilized Brassicaceae roots, nearly all Arabi- Sample position in the DNA extraction plates was randomized, and this dopsis thaliana, planted in two soils from North Carolina (USA). A randomized distribution was maintained throughout library prepara- detailed description of this collection and isolation procedures can tion and sequencing. be found in ref. 20. One week before each experiment, bacteria were in- oculated from glycerol stocks into 600 μl KB medium in a 96-deep-well Bacterial 16S rRNA sequencing. We amplified the V3–V4 regions plate. Bacterial cultures were grown at 28 °C, shaking at 250 rpm. After of the bacterial 16S rRNA gene using the primers 338F (5′-ACTCCTACGGGA 5 days of growth, cultures were inoculated into fresh medium and re- GGCAGCA-3′) and 806R (5′-GGACTACHVGGGTWTCTAAT-3′). Two turned to the incubator for an additional 48 h, resulting in 2 copies of barcodes and six frameshifts were added to the 5′ end of 338F and six each culture (7 days old and 48 h old). We adopted this procedure to frameshifts were added to the 806R primers, on the basis of a pre- account for variable growth rates of different members of the synthetic viously published protocol36. Each PCR reaction was performed in community and to ensure that nonstationary cells from each strain triplicate, and included a unique mixture of three frameshifted prim- were included in the inoculum. After growth, 48-h and 7-day plates er combinations for each plate. PCR conditions were as follows: 5 μl were combined and optical density of cultures was measured at 600 nm Kapa Enhancer, 5 μl Kapa Buffer A, 1.25 μl 5 μM 338F, 1.25 μl 5 μM 806R,

(OD600) using an Infinite M200 Pro plate reader (TECAN). All cultures 0.375 μl mixed plant rRNA gene-blocking peptide nucleic acids (PNAs) were then pooled while normalizing the volume of each culture to (1:1 mix of 100 μM plastid PNA and 100 μM mitochondrial PNA36), 0.5 μl

OD600 = 1. The mixed culture was washed twice with 10 mM MgCl2 to Kapa dNTPs, 0.2 μl Kapa Robust Taq, 8 μl dH2O, 5 μl DNA; temperature remove spent medium and cell debris, and vortexed vigorously with cycling: 95 °C for 60 s; 24 cycles of 95 °C for 15 s; 78 °C (PNA) for 10 s; sterile glass beads to break up aggregates. OD600 of the mixed, washed 50 °C for 30 s; 72 °C for 30 s; 4 °C until use. Following PCR clean up, culture was then measured and normalized to OD600 = 0.2. The synthetic using AMPure beads (Beckman Coulter), the PCR product was indexed community inoculum (100 μl) was spread on 12 × 12-cm vertical square using 96 indexed 806R primers with the Kapa HiFi Hotstart readymix agar plates with amended Johnson medium (JM)1 without sucrose be- with the same primers as above; temperature cycling: 95 °C for 60 s; fore transferring seedlings. 9 cycles of 95 °C for 15 s; 78 °C (PNA) for 10 s; 60 °C for 30 s; 72 °C for 35 s; 4 °C until use. PCR products were purified using AMPure XP magnetic In vitro plant growth conditions. All seeds were surface-sterilized with beads (Beckman Coulter) and quantified with a Qubit 2.0 fluorometer 70% bleach, 0.2% Tween-20 for 8 min, and rinsed 3 times with sterile (Invitrogen). Amplicons were pooled in equal amounts and then diluted distilled water to eliminate any seed-borne microorganisms on the seed to 10 pM for sequencing. Sequencing was performed on an Illumina surface. Seeds were stratified at 4 °C in the dark for 2 days. Plants were MiSeq instrument using a 600-cycle V3 chemistry kit. DNA sequence germinated on vertical square 12 × 12-cm agar plates with JM containing data for this experiment are available at the NCBI Bioproject reposi- 0.5% sucrose, for 7 days. Then, 10 plants were transferred to each of the tory (accession PRJNA543313). The abundance matrix, metadata and agar plates inoculated with the synthetic community. The composition taxonomy are available at https://github.com/isaisg/variovoraxRGI. of JM in the agar plates was amended to produce environmental vari- ation. We added to the previously reported phosphate concentration 16S rRNA amplicon sequence data processing. Synthetic community gradient (0, 10, 30, 50, 100 and 1,000 μM Pi)5 three additional environ- sequencing data were processed with MT-Toolbox37. Usable read output mental gradients: salinity (50, 100, 150 and 200 mM NaCl), pH (5.5, 7.0 from MT-Toolbox (that is, reads with 100% correct primer and primer se- and 8.2) and incubation temperature (10, 21 and 31 °C). Each gradient quences that successfully merged with their pair) were quality-filtered was tested separately, in two independent replicas. Each condition using Sickle38 by not allowing any window with Q score under 20. The included three synthetic community + plant samples, two no-plant resulting sequences were globally aligned to a reference set of 16S controls and one no-bacteria control. Thus, the total sample size for rDNA sequences extracted from genome assemblies of members of the each condition was n = 6. Previous publications1,3,5 have shown that synthetic community. For strains that did not have an intact 16S rDNA an n ≥ 5 is provides sufficient power for synthetic community profil- sequence in their assembly, we sequenced the 16S rRNA gene using ing. Plates were placed in randomized order in growth chambers and Sanger sequencing. The reference database also included sequences grown under a 16-h dark/8-h light regime at 21-°C day/18-°C night for from known bacterial contaminants and Arabidopsis organellar se- 12 days. Upon collection, DNA was extracted from roots, shoots and quences. Sequence alignment was performed with USEARCH v.7.109039 the agar substrate. Here and hereafter, all measurements were taken with the option usearch_global at a 98% identity threshold. On average, from distinct samples. 85% of sequences matched an expected isolate. Our 185 isolates could not all be distinguished from each other based on the V3–V4 sequence DNA extraction. Roots, shoots and agar were collected separately, and were thus classified into 97 unique sequences. A unique sequence pooling 6–8 plants for each sample. Roots and shoots were placed in encompasses a set of identical (clustered at 100%) V3–V4 sequences 2-ml Eppendorf tubes with 3 sterile glass beads. These samples were coming from a single or multiple isolates. washed three times with sterile distilled water to remove agar particles Sequence mapping results were used to produce an isolate abun- and weakly associated microorganisms. Tubes containing the samples dance table. The remaining unmapped sequences were clustered were stored at −80 °C until processing. Root and shoot samples were into operational taxonomic units (OTUs) using UPARSE40 imple- lyophilized for 48 h using a Labconco freeze-dry system and pulverized mented with USEARCH v.7.1090, at 97% identity. Representative OTU Article sequences were taxonomically annotated with the RDP classifier41 trained on the Greengenes database42 (4 February 2011). Matches Deconstructing the synthetic community to four modules of to Arabidopsis organelles were discarded. The vast majority of the co-occurring strains remaining unassigned OTUs belonged to the same families as iso- This section relates to Fig. 1a–c. lates in the synthetic community. We combined the assigned unique sequence and unassigned OTU count tables into a single count table. Bacterial culture and plant-inoculation. Strains belonging to In addition to the raw count table, we created rarefied (1,000 reads each module A, B, C and D (‘Co-occurrence analysis’ in ‘Arabidopsis per sample) and relative abundance versions of the abundance matrix with bacterial synthetic community microcosm across four stress for further analyses. gradients’) were grown in separate deep 96-well plates and mixed The resulting abundance tables were processed and analysed with as described in ‘Bacterial culture and plant-inoculation’ in ‘Arabi- functions from the ohchibi package (https://github.com/isaisg/ dopsis with bacterial synthetic community microcosm across four ohchibi). An α-diversity metric (Shannon diversity) was calculated stress gradients’. The concentration of each module was adjusted 43 using the diversity function from the vegan package v.2.5-3 . We to OD600 = 0.05 (1/4 of the concentration of the full synthetic com- used ANOVA to test for differences in α-diversity between groups. munity). Each module was spread on the plates either separately, β-Diversity analyses (principal coordinate analysis and canonical or in combination with another module at a total volume of 100 μl. analysis of principal coordinates (CAP)) were based on Bray–Curtis In addition, we included a full synthetic community control and an dissimilarity calculated from the relative abundance matrices. We uninoculated control, bringing the number of synthetic community used the capscale function from the vegan R package v.2.5-343 to com- combinations to 12. We performed the experiment in two independ- pute the CAP. To analyse the full dataset (all fractions and all abiotic ent replicates and each replicate included five plates per synthetic treatments), we constrained by fraction and abiotic treatment while community combination. conditioning for the replica and experiment effect. We explored the abiotic conditions effect inside each of the four abiotic gradients In vitro plant growth conditions. Seed sterilization and germina- tested (phosphate, salinity, pH and temperature). We performed the tion conditions were the same as in ‘In vitro plant growth conditions’ fraction–abiotic interaction analysis within each fraction indepen- in ‘Arabidopsis with bacterial synthetic community microcosm dently, constraining for the abiotic conditions while conditioning across four stress gradients’. Plants were transferred to each of the for the replica effect. In addition to CAP, we performed PERMANOVA synthetic-community-inoculated agar plates containing JM without using the adonis function from the vegan package v.2.5-343. We used sucrose. Plates were placed in randomized order in growth chambers the package DESeq2 v.1.22.144 to compute the enrichment profiles for and grown under a 16-h dark/8-h light regime at 21-°C day/18-°C night unique sequences present in the count table. for 12 days. Upon collection, root morphology was measured. We estimated the fraction effect across all the abiotic conditions tested by creating a group variable that merged the fraction variable Root and shoot image analysis. Plates were imaged 12 days after trans- and the abiotic condition variable together (for example, Root_10Pi, fer, using a document scanner. Primary root length elongation was Substrate_10Pi). We fitted the following model specification using this measured using ImageJ45 and shoot area and total root network were group variable: abundance ~ rep + experiment + group. measured with WinRhizo software (Regent Instruments). From the fitted model, we extracted—for all levels within the group variables—the following comparisons: substrate versus root and sub- Primary root elongation analyses. Primary root elongation was com- strate versus shoot. A unique sequence was considered statistically pared across the no bacteria, full synthetic community, single mod- significant if it had a FDR-adjusted P value < 0.05. ules and pairs of modules treatments jointly using a two-sided ANOVA All scripts and dataset objects necessary to reproduce the synthetic model controlling for the replicate effect. We inspected the normality community analyses are deposited in the following GitHub repository: assumptions (here and elsewhere) using qqplots and Shapiro tests. https://github.com/isaisg/variovoraxRGI. Differences between treatments were indicated using the confidence letter display derived from the Tukey’s post hoc test implemented in Co-occurrence analysis. The relative abundance matrix (unique the package emmeans46. sequences × samples) was standardized across the unique sequences by dividing the abundance of each unique sequence in its sample Inoculating plants with all synthetic community isolates over the mean abundance of that unique sequence across all sam- separately ples. Subsequently, we created a dissimilarity matrix based on the This section relates to Extended Data Fig. 2a–c. Pearson correlation coefficient between all the pairs of strains in the transformed abundance matrix, using the cor function in the Bacterial culture and plant inoculation. Cultures from each strain stats base package in R. Finally, hierarchichal clustering (method in the synthetic community were grown in KB medium and washed ward.D2, function hclust) was applied over the dissimilarity matrix separately (‘Bacterial culture and plant inoculation’ in ‘Arabidopsis constructed above. with bacterial synthetic community microcosm across four stress

gradients’), and OD600 was adjusted to 0.01 before spreading 100 μl Heat map and family enrichment analysis. We visualized the results on plates. We performed the experiment in two independent repli- of the generalized linear model (GLM) testing the fraction effects across cates and each replicate included one plate per each of the 185 strains. each specific abiotic condition tested using a heat map. The rows in the In vitro growth conditions were the same as described in ‘In vitro plant heat map were ordered according to the dendrogram order obtained growth conditions’ in ‘Deconstructing the synthetic community to four from the unique sequences co-occurrence analysis. The heat map was modules of co-occurring strains’. Upon collection, root morphology coloured on the basis of the log2-transformed fold change output by was measured (‘Root and shoot image analysis’ in ‘Deconstructing the the GLM model. We highlighted in a black shade the comparisons that synthetic community to four modules of co-occurring strains’). Isolates were significant (q value < 0.05). Finally, for each of the four modules, we generating an average main root elongation of <3 cm were classified computed for each family present in that module a hypergeometric test as RGI-inducing strains. testing if that family was overrepresented (enriched) in that particular module. Families with an FDR-adjusted P value < 0.1 are visualized in Tripartite plant–microorganism–microorganism experiments the figure. This section relates to Fig. 1d, e, Extended Data Fig. 3a. quantification were performed using a 2100 Bioanalyzer instrument Experimental design. To identify strains that revert RGI (Fig. 1d), we (Agilent) and the Quant-iT PicoGreen dsDNA Reagent (Invitrogen), selected all 18 non-RGI-inducing strains in module A and co-inoculated respectively. Libraries were sequenced using Illumina HiSeq4000 them with each of four RGI-inducing strains, one from each module. sequencers to generate 50-bp single-end reads. The experiment also included uninoculated controls and controls consisting of each of the 22 strains inoculated alone, amounting to 95 RNA-seq read processing. Initial quality assessment of the Illumina separate bacterial combinations. RNA-seq reads was performed using FastQC v.0.11.7 (Babraham Bioin- To confirm the ability of Variovorax and Burkholderia to attenu- formatics). Trimmomatic v.0.3648 was used to identify and discard reads ate RGI induced by diverse bacteria (Extended Data Fig. 3a), three containing the Illumina adaptor sequence. The resulting high-quality RGI-suppressing strains were co-inoculated with a selection of 18 reads were then mapped against the TAIR10 Arabidopsis reference RGI-inducing strains. The experiment also included uninoculated genome using HISAT2 v.2.1.049 with default parameters. The feature- controls and controls consisting of each of the 21 strains inoculated Counts function from the Subread package50 was then used to count alone. Thus, the experiment consisted of 76 separate bacterial combina- reads that mapped to each one of the 27,206 nuclear protein-coding tions. We performed each of these two experiments in two independent genes. Evaluation of the results of each step of the analysis was per- replicates and each replicate included one plate per each of the strain formed using MultiQC v.1.151. Raw sequencing data and read counts combinations. are available at the NCBI Gene Expression Omnibus accession number GSE131158. Bacterial culture and plant-inoculation. All strains were streaked on agar plates, then transferred to 4 ml liquid KB medium for over-night Variovorax drop-out experiment growth. Cultures were then washed, and OD600 was adjusted to 0.02 This section relates to Fig. 2a–c, g, h, Extended Data Fig. 4a. before mixing and spreading 100 μl on each plate. Upon collection, root morphology was measured (‘Root and shoot image analysis’ in ‘Decon- Bacterial culture and plant-inoculation. The entire synthetic com- structing the synthetic community to four modules of co-occurring munity, excluding all 10 Variovorax isolates and all 5 Burkholderia strains’) and plant RNA was collected and processed from uninoculated isolates, was grown and prepared as described in ‘Bacterial culture samples, and from samples with Variovorax CL14, the RGI-inducing and plant inoculation’ in ‘Arabidopsis with bacterial synthetic com- strain Arthrobacter CL28 and the combination of both (‘Experimental munity microcosm across four stress gradients’. The Variovorax and design’ in ‘Tripartite plant–microorganism–microorganism experi- Burkholderia isolates were grown in separate tubes, washed and added ments’). to the rest of the synthetic community to a final OD600 of 0.001 (the cal-

culated OD600 of each individual strain in a 185-member synthetic com-

Primary root elongation analysis. We fitted ANOVA models for each munity at a total of OD600 of 0.2), to form the following five mixtures: RGI-inducing strain we tested. Each model compared the primary root (i) full community: all Variovorax and Burkholderia isolates added to elongation with the RGI-inducing strains alone against root elongation the synthetic community; (ii) Burkholderia drop-out: only Variovorax when the RGI-inducing strain was co-inoculated with other isolates. isolates added to the synthetic community; (iii) Variovorax drop-out: The P values for all the comparisons were corrected for multiple test- only Burkholderia isolates added to the synthetic community; (iv) Vari- ing using FDR. ovorax and Burkholderia drop-out: no isolates added to the synthetic community; (v) uninoculated plants: no synthetic community. The RNA extraction. RNA was extracted from A. thaliana seedlings fol- experiment consisted of six plates per synthetic community mixture, lowing previously published methods47. Four seedlings were pooled amounting to 30 plates. Upon collection, root morphology was meas- from each plate and 3–5 samples per treatment were flash frozen and ured and analysed (‘Root and shoot image analysis’ in ‘Deconstructing stored at −80 °C until processing. Frozen seedlings were ground using the synthetic community to four modules of co-occurring strains’, and a TissueLyzer II (Qiagen), then homogenized in a buffer containing in ‘Primary root elongation analysis’ in ‘Tripartite plant–microorgan- 400 μl of Z6-buffer; 8 M guanidine HCl, 20 mM MES, 20 mM EDTA at ism–microorganism experiments’); and bacterial DNA (‘DNA extrac- pH 7.0. Four hundred μl phenol:chloroform:isoamylalcohol, 25:24:1 tion’ and ‘Bacterial 16S rRNA sequencing’ in ‘Arabidopsis with bacterial was added, and samples were vortexed and centrifuged (20,000g, 10 synthetic community microcosm across four stress gradients’) and min) for phase separation. The aqueous phase was transferred to a plant RNA (‘RNA extraction’ and ‘Plant RNA sequencing’ in ‘Tripartite new 1.5-ml Eppendorf tube and 0.05 volumes of 1 N acetic acid and 0.7 plant–microorganism–microorganism experiments’) were collected volumes 96% ethanol were added. The RNA was precipitated at −20 °C and processed. overnight. Following centrifugation (20,000g, 10 min, 4 °C), the pellet was washed with 200 μl sodium acetate (pH 5.2) and 70% ethanol. The Variovorax drop-out under varying abiotic contexts RNA was dried and dissolved in 30 μl of ultrapure water and stored at This section relates to Fig. 2d, f, Extended Data Fig. 4c, d. −80 °C until use. Bacterial culture and plant-inoculation. The composition of JM in the Plant RNA sequencing. Illumina-based mRNA-seq libraries were pre- agar plates was amended to produce abiotic environmental variation. pared from 1 μg RNA following previously published methods3. mRNA These amendments included salt stress (150 mM NaCl), low phosphate was purified from total RNA using Sera-mag oligo(dT) magnetic beads (10 μM phosphate), high pH (pH 8.2) and high temperature (plates (GE Healthcare Life Sciences) and then fragmented in the presence of incubated at 31 °C), as well as an unamended JM control. Additionally, divalent cations (Mg2+) at 94 °C for 6 min. The resulting fragmented we tested a different medium (1/2-strength Murashige and Skoog (MS)) mRNA was used for first-strand cDNA synthesis using random hexamers and a soil-like substrate. As a soil-like substrate, we used calcined clay and reverse transcriptase, followed by second-strand cDNA synthe- (Diamond Pro), prepared as follows: 100 ml of clay was placed in Ma- sis using DNA polymerase I and RNaseH. Double-stranded cDNA was genta GA7 jars. The jars were then autoclaved twice. Forty ml of liquid end-repaired using T4 DNA polymerase, T4 polynucleotide kinase and JM was added to the Magenta jars, with the corresponding bacterial −4 Klenow polymerase. The DNA fragments were then adenylated using mixture spiked into the media at a final OD600 of 5 × 10 . Four 1-week Klenow exo-polymerase to allow the ligation of Illumina Truseq HT old seedlings were transferred to each vessel, and vessels were covered adapters (D501–D508 and D701–D712). All enzymes were purchased with Breath-Easy gas permeable sealing membrane (Research Products from Enzymatics. Following library preparation, quality control and International) to maintain sterility and gas exchange. Article The entire synthetic community, excluding all 10 Variovorax iso- tubes, washed and added to the rest of the synthetic community to a lates, was grown and prepared as described in ‘Bacterial culture and final OD600 of 0.001. plant inoculation’ in ‘Arabidopsis with bacterial synthetic community In a separate experiment, the 35-member synthetic community used microcosm across four stress gradients’. The Variovorax isolates were in ref. 1 was grown, excluding Variovorax CL14, to create a taxonomi- grown in separate tubes, washed and added to the rest of the synthetic cally diverse, Variovorax-free subset of the full 185-member commu- community to a final OD600 of 0.001 (the calculated OD600 of each indi- nity. The concentration of this synthetic community was adjusted to vidual strain in a 185-member synthetic community at an OD600 of 0.2), OD600 = 0.05. The Variovorax isolates were grown in separate tubes, to form the following five mixtures: (i) full community: all Variovorax washed and added to the rest of the synthetic community to a final isolates added to the synthetic community; (ii) Variovorax drop-out: no OD600 of 0.001. isolates added to the synthetic community; (iii) uninoculated plants: These two experiments included the following mixtures (Extended no synthetic community. Data Fig. 4b): (i) module A excluding Variovorax; (ii) module C; We inoculated all 3 synthetic community combinations in all 7 abiotic (iii) module D; (iv) module A including Variovorax; (v) module C + treatments, amounting to 21 experimental conditions. We performed all 10 Variovorax; (vi) module D + all 10 Variovorax; (vii) 35-member the experiment in 2 independent replicates and each replicate included synthetic community excluding the one Variovorax found therein; 3 plates per experimental conditions, amounting to 63 plates per rep- (viii) 34-member synthetic community + all 10 Variovorax; (ix) uninoculated licate. Upon collection, root morphology was measured (‘Root and control. The experiment with modules A, C and D was performed in two shoot image analysis’ in ‘Deconstructing the synthetic community independent experiments, with two plates per treatment in each. The to four modules of co-occurring strains’); and Bacterial DNA (‘DNA experiment with the 34-member synthetic community was performed extraction, ‘Bacterial 16S rRNA sequencing’ and ‘16S rRNA amplicon once, with 5 plates per treatment. Upon collection, root morphology sequence data processing’ in ‘Arabidopsis with bacterial synthetic com- was measured (‘Root and shoot image analysis’ in ‘Deconstructing munity microcosm across four stress gradients’) and plant RNA (‘RNA the synthetic community to four modules of co-occurring strains’). extraction’, ‘Plant RNA sequencing’ and ‘RNA-seq read processing’ in ‘Tripartite plant–microorganism–microorganism experiments’) were Primary root elongation analysis. We directly compared differences collected and processed. between the full synthetic community and Variovorax drop-out treat- ment using a t-test and adjusting the P values for multiple testing using Root image analysis. For agar plates, roots were imaged as described FDR. in ‘Root and shoot image analysis’ in ‘Deconstructing the synthetic community to four modules of co-occurring strains’. For calcined clay Phylogenetic inference of the synthetic community and pots, four weeks after transferring, pots were inverted, and whole root Variovorax isolates systems were gently separated from the clay by washing with water. This section relates to Figs. 1d, 2a, Extended Data Figs. 1a, 2b, 3a, 4b, Root systems were spread over an empty Petri dish and scanned using 5a, b, 9a. a document scanner. To build the phylogenetic tree of the synthetic community isolates, we used the previously described super matrix approach20. We scanned Primary root elongation and total root network analysis. Primary 120 previously defined marker genes across the 185 isolate genomes root elongation was compared between synthetic-community treat- from the synthetic community using the hmmsearch tool from the ments within each of the different abiotic contexts tested independent- hmmer v.3.1b252. Then, we selected 47 markers that were present as ly. Differences between treatments were indicated using the confidence single-copy genes in 100% of our isolates. Next, we aligned each indi- letter display derived from the Tukey’s post hoc test implemented in vidual marker using MAFFT53 and filtered low-quality columns in the the package emmeans. alignment using trimAl54. Then, we concatenated all filtered alignments into a super alignment. Finally, FastTree v.2.155 was used to infer the Bacterial 16S rRNA data analysis. To be able to compare shifts in phylogeny using the WAG model of evolution. For the tree of the rela- the community composition of samples treated with and without the tive of Variovorax, we chose 56 markers present as single copy across Variovorax genus, we in silico-removed the 10 Variovorax isolates from 124 Burkholderiales isolates and implemented the same methodology the count table of samples inoculated with the full community treat- described above. ment. We then merged this count table with the count table constructed from samples inoculated without the Variovorax genus (Variovorax Measuring how prevalent the RGI suppression trait is across the drop-out treatment). Then, we calculated a relative abundance of each Variovorax phylogeny unique sequence across all the samples using the merged count matrix. This section relates to Fig. 3a, Extended Data Fig. 5a, b. Finally, we applied CAP over the merged relative abundance matrix to control for the replica effect. In addition, we used the function adonis Bacterial culture and plant inoculation. Fifteen Variovorax strains from the vegan R package to compute a PERMANOVA test over the from across the phylogeny of the genus were each co-inoculated with merged relative abundance matrix and we fitted a model evaluating the RGI-inducing Arthrobacter CL28. All 16 strains were grown in sepa- the fraction and synthetic community (presence of Variovorax) effects rate tubes, then washed and OD600 was adjusted to 0.01 before mixing. over the assembly of the community. Pairs of strains were mixed in 1:1 ratios and spread at a total volume of 100 μl onto agar before seedling transfer. The experiment also included Variovorax drop-out under varying biotic contexts uninoculated controls and controls consisting of each of the 16 strains This section relates to Fig. 2e, Extended Data Fig. 4b. inoculated alone. Thus, the experiment consisted of 32 separate bac- terial combinations. We performed the experiment one time, which Bacterial culture and plant inoculation. Strains belonging to mod- included three plates per bacterial combination. Upon collection, pri- ules A (excluding Variovorax), C and D were grown in separate wells in mary root elongation was analysed as described in ‘‘Root and shoot deep 96-well plates and mixed as described in ‘Bacterial culture and image analysis’ in ‘Deconstructing the synthetic community to four plant inoculation’ in ‘Arabidopsis with bacterial synthetic community modules of co-occurring strains’. microcosm across four stress gradients’. The concentration of each module was adjusted to OD600 = 0.05 (1/4 of the concentration of the full Measuring RGI in tomato seedlings synthetic community). The Variovorax isolates were grown in separate This section relates to Extended Data Fig. 3b. performed the following contrasts, Variovorax drop-out versus NB, Experimental design. This experiment included the following treat- and Variovorax drop-out versus full synthetic community. The logic ments: (i) no bacteria, (ii) Arthrobacter CL28, (iii) Variovorax CL14 and behind these two contrasts was identical to the tripartite system: to (iv) Arthrobacter CL28 + Variovorax CL14. Each treatment was repeated identify genes that are associated with the RGI phenotype (Variovorax in three separate agar plates with five tomato seedlings per plate. The drop-out versus NB contrast) and repressed when Variovorax are pre- experiment was repeated in two independent replicates. sent (Variovorax drop-out versus full synthetic community contrast). For visualization purposes, we applied a variance stabilizing transfor- Bacterial culture and plant inoculation. All strains were grown in sepa- mation (DESeq2) to the raw count gene matrix. We then standardized rate tubes, then washed and OD600 was adjusted to 0.01 before mixing each gene expression (z-score) along the samples. We subset DEGs from and spreading (‘Bacterial culture and plant inoculation’ in ‘Tripartite this standardized matrix and calculated the mean z-score expression plant–microorganism–microorganism experiments’). Four hundred μl value for each synthetic community treatment. of each bacterial treatment was spread on 20 × 20 agar plates contain- To identify the tissue-specific expression profile of the 18 intersecting ing JM agar with no sucrose. genes between the tripartite and drop-out systems, we downloaded the spatial expression profile of each gene from the Klepikova atlas25 In vitro plant growth conditions. We used tomato cultivar Heinz using the bio-analytic resource of plant biology platform. Then, we 1706 seeds. All seeds were soaked in sterile distilled water for 15 min, constructed a spatial expression matrix of the 18 genes and computed then surface-sterilized with 70% bleach, 0.2% Tween-20 for 15 min, and pairwise Pearson correlation between all pairs of genes. Finally, we rinsed 5 times with sterile distilled water to eliminate any seed-borne applied hierarchical clustering to this correlation matrix. microorganisms on the seed surface. Seeds were stratified at 4 °C in the dark for 2 days. Plants were germinated on vertical square 10 × 10 cm Comparison with acute auxin response dataset. This section relates agar plates with JM containing 0.5% sucrose, for 7 days. Then, 5 plants to Extended Data Fig. 7d. We applied the variance stabilizing transfor- were transferred to each of the synthetic-community-inoculated agar mation (DESeq2) to the raw count gene matrix. We then standardized plates. Upon collection, root morphology was measured (‘Root and each gene expression (z-score) along the samples. From this matrix, shoot image analysis’ in ‘Deconstructing the synthetic community to we subset 12 genes that in a previous study26 exhibited the highest four modules of co-occurring strains’). fold change between auxin-treated and untreated samples. Finally, we calculated the mean z-score expression value of each of these 12 Primary root elongation analysis. Differences between treatments genes across the synthetic community treatments. We estimated the were indicated using the confidence letter display derived from the statistical significance of the trend of these 12 genes between a pair of Tukey’s post hoc test from an ANOVA model. synthetic community treatments (full synthetic community versus Variovorax drop-out, Arthrobacter CL28 versus Arthrobacter CL28 plus Determination of Arthrobacter CL28 colony forming units from Variovorax CL14) using a permutation approach: we estimated a P value roots by randomly selecting 12 genes 10,000 times from the expression ma- This section relates to Extended Data Fig. 6. trix and comparing the mean expression between the two synthetic Arabidopsis seedlings were inoculated with (i) Arthrobacter CL28 community treatments (for example, full synthetic community versus alone, (ii) Arthrobacter CL28 + Variovorax CL14 or (iii) Arthrobacter CL28 Variovorax drop-out) with the actual mean expression value from the + Variovorax B4, as described in ‘Bacterial culture and plantinoculation’ 12 genes reported as robust auxin markers. in ‘Tripartite plant–microorganism–microorganism experiments’. Each bacterial treatment included four separate plates, with nine seedlings Measuring the ability of Variovorax to attenuate RGI induced by in each plate. Upon collection, all seedlings were placed in pre-weighed small molecules IAA, 2,4-dichlorophenoxyacetic acid, ethylene 2-ml Eppendorf tubes containing 3 glass beads, 3 seedlings per tube (ACC), cytokinins (zeatin and 6-benzylaminopurine) and (producing 12 data points per treatment). Roots were weighed, and flagellin 22 peptide (flg22) then homogenized using a bead beater (MP Biomedicals). The resulting This section relates to Fig. 3a, Extended Data Fig. 8a. suspension was serially diluted, then plated on LB agar plates contain- ing 50 μg/ml of apramycin to select for Arthrobacter CL28 colonies and Bacterial culture and plant inoculation. We embedded each of the colonies were counting after incubation of 48 h at 28 °C. following compounds in JM plates: 100 nM IAA (Sigma), 100 nM 1- ACC) (Sigma), 100 nM 2,4-dichlorophenoxyacetic acid (Sigma), 100 nM flg22 Arabidopsis RNA-seq analysis (PhytoTech labs), 100 nM 6-benzylaminopurine (BAP) (Sigma) and 100 nM This section relates to Extended Data Fig. 7. zeatin (Sigma). As some of these compound stocks were initially solubilized in ethanol, we included comparable amounts of ethanol in Detection of RGI-induced genes. To measure the transcriptional re- the control treatments. Plates with each compound were inoculated sponse of the plant to the different synthetic community combinations, with one of the Variovorax strains CL14, MF160, B4 or YR216 or with we used the R package DESeq2 v.1.22.144. The raw count genes matrixes Burkholderia CL11. These strains were grown in separate tubes, then for the drop-out and tripartite experiments were used independently washed and OD600 was adjusted to 0.01 before spreading 100 μl on to define differentially expressed genes (DEGs). For the analysis of both plates. In addition, we included uninoculated controls for each com- experiments we fitted the following model specification: abundance pound. We also included unamended JM plates inoculated with the gene ~ synthetic community. RGI-inducing Arthrobacter CL28 co-inoculated with each of the Variovo- From the fitted models, we derived the following contrasts to obtain rax or Burkholderia strains, or alone. Thus, the experiment included 42 DEGs. A gene was considered differentially expressed if it had a q-value individual treatments. The experiment was repeated twice, with three < 0.1. For the tripartite system (‘Tripartite plant–microorganism– independent replicates per experiment. Upon collection, root morphol- microorganism experiments’), we performed the following contrasts: ogy was measured (‘Root and shoot image analysis’ in ‘Deconstructing Arthrobacter CL28 versus no bacteria (NB) and Arthrobacter CL28 ver- the synthetic community to four modules of co-occurring strains’). sus Arthrobacter CL28 co-inoculated with Variovorax CL14. The logic behind these two contrasts was to identify genes that were induced in Primary root elongation analysis. Primary root elongation was com- RGI plants (Arthrobacter CL28 versus NB) and repressed by Variovorax pared between bacterial treatments within each of RGI treatments CL14. For the drop-out system (‘Variovorax drop-out experiment’), we tested. Differences between treatments were estimated as described Article in ‘Primary root elongation analysis’ in ‘Tripartite plant–microorgan- ism–microorganism experiments’. We plotted the estimated means Primary root elongation analysis. Primary root elongation was stand- with 95% confidence interval of each bacterial treatment across the ardized to the no-bacteria control of each genotype, and compared different RGI treatments. between genotype and 1-MCP treatments within the Arthrobacter CL28 treatment and within the Variovorax drop-out synthetic community In vitro growth of Variovorax treatment, independently. Differences between treatments were esti- This section relates to Fig. 4b, Extended Data Figs. 8b, 9d, e. mated as described in ‘Primary root elongation analysis’ in ‘Tripartite Variovorax CL14 was grown in 5-ml cultures for 40 h at 28 °C in 1× plant–microorganism–microorganism experiments’. We plotted the M9 minimal salts medium (Sigma M6030) supplemented with 2 mM estimated means with 95% confidence interval of each bacterial treat-

MgSO4, 0.1 mM CaCl2, 10 μM FeSO4, and a carbon source: either 15 mM ment across the four genotypes. We calculated the IQR for the full and succinate alone, 0.4 mM IAA with 0.5% ethanol for IAA solubilization, Arthrobacter CL28 with Variovorax CL14 treatments, pooling the four or both. Optical density at 600 nm and IAA concentrations were meas- genotypes and MCP treatments. ured at six time points. IAA concentrations were measured using the Salkowski method modified from ref. 56. One hundred μl of Salkowski Preparation of binarized plant images reagent (10 mM FeCl3 in 35% perchloric acid) was mixed with 50 μl cul- This section relates to Figs. 1c, 2b, Extended Data Fig. 2c. ture supernatant or IAA standards and colour was allowed to develop To present representative root images, we increased contrast and for 30 min before measuring the absorbance at 530 nm. subtracted background in ImageJ, then cropped the image to select representative roots. Neighbouring roots were manually erased from Measuring plant auxin response in vivo using a bioreporter line the cropped image. This section relates to Extended Data Fig. 8c, d. Mining Variovorax genomes for auxin degradation operons and Bacterial culture and plant-inoculation. Seven-day-old trans- ACC-deaminase genes and comparative genomics of Variovorax genic Arabidopsis seedlings expressing the DR5::GFP reporter con- genomes against the other synthetic community members struct57 were transferred onto plates containing: (i) 100 nM IAA, This section relates to Extended Data Fig. 5a, Supplementary Table 4. (ii) Arthrobacter CL28, (iii) 100 nM IAA + Variovorax CL14, (iv) Arthrobacter We used local alignment (BLASTp)58 to search for the presence of CL28 + Variovorax CL14, (v) uninoculated plates. For treatments (ii) the 10 genes (iacA, iacB, iacC, iacD, iacE, iacF, iacG, iacH, iacI and iacY) and (iii), Bacterial strains were grown in separate tubes, then washed from a previously characterized auxin degradation operon in a different 17 and OD600 was adjusted to 0.01. For treatment (iv), OD-adjusted cul- genus across the 10 Variovorax isolates in our synthetic community. tures were mixed in 1:1 ratios and spread onto agar before seedling A minimal set of seven of these genes (iacA, iacB, iacC, iacD, iacE, iacF transfer. and iacI) was shown to be necessary and sufficient for auxin degrada- tion17. We identified homologues for these genes across the Variovorax Fluorescence microscopy. GFP fluorescence in the root elongation phylogeny (Extended Data Fig. 5a) at relatively low sequence identity zone of 8–10 plants per treatment were visualized using a Nikon (27–48%). Two genes of the minimal set of seven genes did not have any Eclipse 80i fluorescence microscope at days 1, 3, 6, 9 and 13 after homologues in most Variovorax genomes (iacB and iacI). In addition inoculation. The experiment was performed in two independent to the iac operon, we scanned the genomes for the auxin degrada- replicates. tion operon described in ref. 59 and could not identify it in any of the From each root imaged, 10 random 30 × 30 pixel squares were sam- Variovorax isolates. pled and average GFP intensity was measured using imageJ45. Treat- We also searched for the ACC deaminase gene by looking for the ments were compared within each time point using ANOVA tests with KEGG orthology identifier K01505 (1-aminocyclopropane-1-carboxylate Tukey’s post hoc in the R base package emmeans. For visualization deaminase) across the IMG annotations available for all our genomes. purposes, we plotted the estimated means of each bacterial across We used orthofinder v.2.2.160 to construct orthogroups (group the different time points. of homologue sequences) using all the coding sequences of the 10 Variovorax isolates included in the full 185-member synthetic commu- Measuring the dual role of auxin and ethylene perception in nity. We aligned each orthogroup using MAFFT v.70753 and proceeded synthetic-community-induced RGI to build hidden Markov model (HMM) profiles from the alignments This section relates to Fig. 3b. using hmmbuild v.3.1b252. We then used the HMM profiles that con- sist of core genes (present in the 10 isolates) in the Variovorax genus Bacterial culture and plant inoculation. We transferred four 7-day-old and scanned the 175 remaining genomes in the synthetic community wild-type seedling and four axr2-1seedlings to each plate in this for these HMM profiles using hmmsearch v.31.b252. We then selected experiment. The plates contained one of five bacterial treatments: orthogroups that were less than 5% prevalent in the 175 remaining iso- (i) Arthrobacter CL28, (ii) Arthrobacter CL28 + Variovorax CL14, lates scanned. Finally, taking the Variovorax CL14 genome as a refer- (iii) Variovorax drop-out synthetic community, (iv) full synthetic com- ence, we built hotspots of physically adjacent selected orthogroups. munity, (v) uninoculated, prepared as described in ‘Bacterial culture We used an iterative approach that extended adjacent orthogroups if and plant inoculation’ in ‘Variovorax drop-out under varying abiotic they were less than 10 kb from each other. As a final filtering step, we contexts’, and in ‘Bacterial culture and plant inoculation’ in ‘Measur- selected hotspots that contained more than 10 selected orthogroups. ing plant auxin response in vivo using a bioreporter line’. Plates were placed vertically inside sealed 86 × 68 cm Ziploc bags. In one of the Variovorax CL14 RNA-seq in monoculture and in coculture with bags, we placed an open water container with 80 2.5 g sachets contain- Arthrobacter CL28 ing 0.014% 1-MCP (Ethylene Buster, Chrystal International BV). In the This section relates to Fig. 4a, Extended Data Fig. 9b, c. second bag we added, as a control, an open water container. Both bags were placed in the growth chamber for 12 days. After 6 days of growth, Bacterial culture. Variovorax CL14 was grown either alone or in cocul- we added 32 additional sachets to the 1-MCP-treated bag to maintain ture with Arthrobacter CL28 in 5 ml of 1/10 2× YT medium (1.6 g/l tryp- 1-MCP concentrations in the air. Upon collection, root morphology tone, 1 g/l yeast extract, 0.5 g/l NaCl) in triplicate. The monoculture was was measured (‘Root and shoot image analysis’ in ‘Deconstructing inoculated at OD600 of 0.02 and the coculture was inoculated with OD600 the synthetic community to four modules of co-occurring s­tr­ai­ns’).­ of 0.01 of each strain. Cultures were grown at 28 °C to early stationary phase (approximately 22 h) and cells were collected by centrifugation from these clones using the ZymoPURE II Plasmid Midiprep Kit (Zymo at 4,100g for 15 min and frozen at −80 °C before RNA extraction. Research) and Sanger-sequencing the insert ends using primers JMC247 and JMC270 (Supplementary Table 6). Double digest of the purified RNA extraction and RNA-seq. Cells were lysed for RNA extraction plasmids with SacI and EcoRV confirmed the size of the inserts. Vector 1 using TRIzol Reagent (Invitrogen) according to the manufacturer in- contains a 35-kb insert and vector 2 contains a 15-kb insert (nucleotide structions. Following cell lysis and phase separation, RNA was purified coordinates 29,100–64,406 and 52,627–67,679, respectively, from using the RNeasy Mini kit (Qiagen) including the optional on column Variovorax CL14 scaffold Ga0102008_10005) (Fig. 4a, Extended Data DNase Digestion with the RNase-Free DNase Set (Qiagen). Total RNA Fig. 9b). The genes in both inserts are in the same direction as the IPT- quality was confirmed on the 2100 Bioanalyzer instrument (Agilent) and G0inducible Lac promoter used to drive LacZ-alpha expression for quantified using a Qubit 2.0 fluorometer (Invitrogen). RNA-seq libraries blue–white screening on pBBR1-MCS2. were prepared using the Universal Prokaryotic RNA-Seq, Prokaryotic AnyDeplete kit (Tecan, formerly NuGEN). Libraries were pooled and Acidovorax Root219::EV and Acidovorax root219::V2 sequenced on the Illumina HiSeq4000 platform to generate 50-bp construction and screening single-end reads. This section relates to Fig. 4b–d, Extended Data Fig. 9b. Triparental mating was used to mobilize vector 2 or control EV RNA-seq analysis. We mapped the generated raw reads to the Vari- pBBR1-MCS2 from E. coli to Acidovorax root219. Donor NEB 10-beta ovorax CL14 genome (fasta file available on https://github.com/isaisg/ E. coli containing the vector for conjugation and helper strain E. coli variovoraxRGI/blob/master/rawdata/2643221508.fna) using bowtie261 pRK201363 were grown in LB medium containing 50 μg/ml kanamycin with the ‘very sensitive’ flag. We then counted hits to each individual at 37 °C. Acidovorax root219 was grown in 2× YT medium containing coding sequence annotated for the Variovorax CL14 genome using 100 μg/ml ampicillin at 28 °C. Bacteria were washed 3 times with 2× YT the function featureCounts from the R package Rsubread, inputting medium without antibiotics, mixed in a ratio of approximately 1:1:10 the Variovorax CL14 gff file (available on https://github.com/isaisg/ donor:helper:recipient, centrifuged and resuspended in 1/10 the vol- variovoraxRGI/blob/master/rawdata/2643221508.gff) and using the ume and plated as a pool on LB agar plates without antibiotics and default parameters with the flag allowMultiOverlap = FALSE. Final- grown at 28 °C. Eighteen to thirty h later, exconjugantes were streaked ly, we used DESeq2 to estimate DEGs between treatments with the on LB agar plates containing 50 μg/ml kanamycin and 100 μg/ml ampi- corresponding fold change estimates and FDR-adjusted P values. cillin to select only Acidovorax root219 containing the conjugated vector. The resulting strains are designated Acidovorax root219::EV Variovorax CL14 genomic library construction and screening containing empty vector pBBR1-MCS2 and Acidovorax root219::V2 This section relates to Fig. 4a, Extended Data Fig. 9b, Supplementary containing vector 2. In vitro IAA degradation was performed as in ‘In Table 5. vitro growth of Variovorax’ using M9 medium with carbon sources: 15 mM succinate, 0.1 mg/ml IAA and 0.5% ethanol with the addition of Library construction. High-molecular -eight Variovorax CL14 genomic 50 μg/ml kanamycin and 1 mM IPTG. Primary root elongation meas- DNA was isolated by phenol–chloroform extraction. This genomic urement was performed as described in ‘Root and shoot image analy- DNA was partially digested with Sau3A1 (New England Biolabs), and sis’ in ‘Deconstructing the synthetic community to four modules of separated on the BluePippin (Sage Science) to isolate DNA fragments co-occurring strains’, on MS medium with 1 mM IPTG and RGI induced >12.5 kb. Vector backbone was prepared by amplifying pBBR-1MCS262 by either 100 nM IAA or Arthrobacter CL28. Acidovorax root219::V2 using Phusion polymerase (New England Biolabs) with primers JMC277– root colonization was compared to Variovorax CL14 colonization JMC278 (Supplementary Table 6), digesting the PCR product with by plating a subset of ground root samples from the root elongation BamHI-HF (New England Biolabs), dephosphorylating with Quick CIP experiment (see ‘Determination of Arthrobacter CL28 colony forming (New England Biolabs), and gel extracting using the QIAquick Gel Ex- units from roots’ for root collection and processing protocol) on LB traction Kit (Qiagen). The prepared Variovorax CL14 genomic DNA agar plates containing 100 μg/ml ampicillin, for which Arthrobacter fragments were ligated to the prepared pBBR1-MCS2 vector back- CL28 is susceptible and Variovorax CL14 and Acidovorax Root219 are bone using ElectroLigase (New England Biolabs) and transformed not. Number of colony-forming units (CFUs) was normalized to root by electroporation into NEB 10-beta Electrocompetent E. coli (New weight (Extended Data Fig. 9d). England Biolabs). Clones were selected by blue–white screening on LB plates containing 1.5% agar, 50 μg/ml kanamycin, 40 μg/ml X-gal Variovorax hotspot 33 knockout construction and screening (5-bromo-4-chloro-3-indolyl-β-d-galactopyranoside), and 1 mM isopro- This section relates to Fig. 4b–d, Extended Data Fig. 9b. pyl β-d-1-thiogalactopyranoside (IPTG) at 37 °C. White colonies were The unmarked deletion mutant Variovorax CL14 Δ2643613653– screened by colony PCR using Taq polymerase and JMC247–JMC270 2643613677 (Variovorax CL14 ΔHS33) was constructed based on a primers (Supplementary Table 6) to eliminate clones with small inserts. genetic system developed for Burkholderia spp. and its suicide vec- The screened library clones were picked into LB medium + 50 μg/ml tor pMo13064 kanamycin, grown at 37 °C, and stored at −80 °C in 20% glycerol. The Variovorax CL14 genomic library comprises approximately 3,500 Knockout suicide vector pJMC158 construction. The vector back- clones with inserts >12.5 kb in vector pBBR1-MCS2 in NEB 10-beta E. coli. bone was amplified from pMo130 using primers JMC203–JMC204 (Supplementary Table 6) with Q5 DNA polymerase (New England Bio- Library screening for IAA degradation. To screen the Variovorax CL14 labs), cleaned up and treated with DpnI (New England Biolabs). One-kb genomic library for IAA degradation, the E. coli clones were grown in LB regions for homologous recombination flanking Variovorax CL14 genes medium containing 50 μg/ml kanamycin, 1 mM IPTG, 0.05 mg/ml IAA, 2643613653–2643613677 were amplified using Q5 Polymerase (New and 0.25% ethanol from IAA solubilization for 3 days at 37 °C. Salkowski England Biolabs) and primers JMC533–JMC534 and JMC535–JMC536 reagent (10 mM FeCl3 in 35% perchloric acid) was mixed with culture (Supplementary Table 6). The vector was assembled with Gibson As- supernatant 2:1 and colour was allowed to develop for 30 min before sembly Mastermix (New England Biolabs) at 50 °C for 1 h, transformed measuring the absorbance at 530 nm. Two clones from the library into NEB 5-alpha chemically competent E. coli (New England Biolabs), (plate 8A well E8 and plate2A well F10, henceforth vector 1 and vector 2, and plated on LB agar with 50 μg/ml kanamycin. pJMC158 DNA was respectively) were identified as degrading IAA. The Variovorax CL14 isolated from a clone using the ZR Plasmid Miniprep Classic Kit (Zymo genes contained in vectors 1 and 2 were inferred by isolating plasmid Research), sequence confirmed, and transformed into biparental Article mating strain E. coli WM3064. E. coli strain WM3064 containing pJMC158 We used the same classification scheme to colour the ASVs across the was maintained on LB containing 50 μg/ml kanamycin and 0.3 mM scatter plots shown. ASVs present in >80% of the sites on average are diaminopimelic acid (DAP) at 37 °C. considered as widespread ASVs.

Conjugative transfer of pJMC158 into Variovorax CL14. For Different plant species in the same soil4. We used the ASV table pro- biparental mating, E. coli WM3064 containing pJMC158 was grown as vided within ref. 4 and followed the same pipeline described in ‘Natural above, and Variovorax CL14 was grown in 2× YT medium containing Arabidopsis populations across Europe’. For each sample in the dataset, 100 μg/ml ampicillin. Each strain was washed separately 3 times with an ASV was considered to be present if that ASV had an RA >0.01% in 2× YT medium, then mixed at ratios between 1:1–1:10 donor:recipient, that particular sample. To calculate the average relative abundance centrifuged and resuspended in approximately 1/10 the volume and (y-axis in Extended Data Figs. 10a, b) independently of prevalence, plated in a single pool on LB agar containing 0.3 mM DAP and grown at we calculated the mean RA using only the samples for which each ASV 28 °C overnight. Exconjugants were streaked onto LB plates containing was considered present. For each one of the thirty plant species in this 100 μg/ml ampicillin, 50 μg/ml kanamycin lacking DAP and grown at data set, an ASV was considered present in that plant species if it was 28 °C to select Variovorax CL14 strains that incorporated suicide vector present in >70% of samples coming from that plant species. An ASV pJMC158. First crossover strains were subsequently purified once by was considered core if it was present in all 30 plant species surveyed restreaking and then individual colonies grown in LB with 100 μg/ml in this experiment. ampicillin, 50 μg/ml kanamycin. Reporting summary Resolution of pJMC158 integration and knockout strain purification Further information on research design is available in the Nature and verification. To resolve the integration of pJMC158, first crossover Research Reporting Summary linked to this paper. strains were grown once in LB medium containing 100 μg/ml ampicillin and 1 mM IPTG then plated on medium containing 10 g/l tryptone, 5 g/l yeast extract, 100 g/l sucrose, 1.5% agar, 100 μg/ml ampicillin and 1 mM Data availability IPTG. Colonies were picked into the same liquid medium and grown The 16S rRNA amplicon sequencing data associated with this study once. The resulting strains were screened by PCR using Q5 polymerase have been deposited in the NCBI Sequence Read Archive under the for deletion of genes 2643613653–2643613677 using primers JMC568– project accession PRJNA543313. The raw transcriptomic data have JMC569 (Supplementary Table 6). These strains were subsequently been deposited in the Gene Expression Omnibus under the accession plate-purified at least 3 times on LB 100 μg/ml ampicillin plates. To GSE131158. We deposited all scripts and additional data structures ensure strain purity, PCR primers were designed to amplify from out- required to reproduce the results of this study in the following GitHub side into the genes that were deleted (primer pairs JMC571–JMC569 and repository: https://github.com/isaisg/variovoraxRGI. Source data are JMC568–JMC570) (Supplementary Table 6). These PCR reactions were provided with this paper. performed using Q5 polymerase with wild-type Variovorax CL14 as a control. All genomic DNA used for screening PCR was isolated using 36. Lundberg, D. S., Yourstone, S., Mieczkowski, P., Jones, C. D. & Dangl, J. L. Practical the Quick-DNA miniprep kit (Zymo Research). The resulting knockout innovations for high-throughput amplicon sequencing. Nat. Methods 10, 999–1002 (2013). 37. Yourstone, S. M., Lundberg, D. S., Dangl, J. L. & Jones, C. D. MT-Toolbox: improved strain was designated Variovorax CL14 ΔHS33. amplicon sequencing using molecule tags. BMC Bioinformatics 15, 284 (2014). 38. Joshi, N. & Fass, J. Sickle: a sliding-window, adaptive, quality-based trimming tool for Screening of Variovorax CL14ΔHS33. In vitro IAA degradation was FastQ files (version 1.33), https://github.com/najoshi/sickle (2011). 39. Edgar, R. C. Search and clustering orders of magnitude faster than BLAST. Bioinformatics performed as in ‘In vitro growth of Variovorax’ using M9 medium with 26, 2460–2461 (2010). carbon sources: 15 mM succinate, 0.1 mg/ml IAA and 0.5% ethanol. 40. Edgar, R. C. UPARSE: highly accurate OTU sequences from microbial amplicon reads. Primary root elongation measurement was performed as described Nat. Methods 10, 996–998 (2013). 41. Wang, Q., Garrity, G. M., Tiedje, J. M. & Cole, J. R. Naive Bayesian classifier for rapid in ‘Root and shoot image analysis’ in ‘Deconstructing the synthetic assignment of rRNA sequences into the new bacterial taxonomy. Appl. Environ. Microbiol. community to four modules of co-occurring strains’, on MS medium 73, 5261–5267 (2007). with 1mM IPTG and RGI induced by either 100 nM IAA or Arthrobacter 42. DeSantis, T. Z. et al. Greengenes, a chimera-checked 16S rRNA gene database and workbench compatible with ARB. Appl. Environ. Microbiol. 72, 5069–5072 (2006). CL28. To control for pleiotropic colonization effects, CFU counts were 43. Oksanen, J. et al. Vegan: Community Ecology Package, R package version 2.5-6, obtained for both CL14 and CL14 ΔHS33 in binary-association with the https://CRAN.R-project.org/package=vegan (2019). plant (see ‘Determination of Arthrobacter CL28 colony-forming units 44. Love, M. I., Huber, W. & Anders, S. Moderated estimation of fold change and dispersion for RNA-seq data with DESeq2. Genome Biol. 15, 550 (2014). from roots’ for root collection and processing protocol). CL14 ΔHS33 45. Schindelin, J. et al. Fiji: an open-source platform for biological-image analysis. Nat. colonization was not impaired (Extended Data Fig. 9e). Methods 9, 676–682 (2012). 46. Lenth, R. emmeans: Estimated Marginal Means, aka Least-Squares Means, R package version 1.4.7, https://CRAN.R-project.org/package=emmeans (2020). Screening existing 16S rRNA census data for Variovorax 47. Logemann, J., Schell, J. & Willmitzer, L. Improved method for the isolation of RNA from This section relates to Extended Data Fig. 10. plant tissues. Anal. Biochem. 163, 16–20 (1987). 48. Bolger, A. M., Lohse, M. & Usadel, B. Trimmomatic: a flexible trimmer for Illumina sequence data. Bioinformatics 30, 2114–2120 (2014). 6 Natural Arabidopsis populations across Europe . We used the 49. Kim, D., Langmead, B. & Salzberg, S. L. HISAT: a fast spliced aligner with low memory DADA265 pipeline to create amplicon sequence variants (ASVs) from requirements. Nat. Methods 12, 357–360 (2015). 6 50. Liao, Y., Smyth, G. K. & Shi, W. The Subread aligner: fast, accurate and scalable read the raw reads published in ref. . We then used the naive Bayes classifier mapping by seed-and-vote. Nucleic Acids Res. 41, e108 (2013). 66 implemented in mothur to taxonomically classify each ASV using the 51. Ewels, P., Magnusson, M., Lundin, S. & Käller, M. MultiQC: summarize analysis results for SILVA 132 database67. To determine prevalence and relative abundance multiple tools and samples in a single report. Bioinformatics 32, 3047–3048 (2016). 6 52. Wheeler, T. J. & Eddy, S. R. nhmmer: DNA homology search with profile HMMs. of ASVs, We followed the same approach as in ref. . The bacterial ASV Bioinformatics 29, 2487–2489 (2013). table was restricted to samples having >1,000 reads. The table was 53. Katoh, K. & Standley, D. M. MAFFT multiple sequence alignment software version 7: transformed to relative abundance (RA) by dividing each value in a improvements in performance and usability. Mol. Biol. Evol. 30, 772–780 (2013). 54. Capella-Gutiérrez, S., Silla-Martínez, J. M. & Gabaldón, T. trimAl: a tool for automated sample by the total reads in that sample. An ASV was considered to alignment trimming in large-scale phylogenetic analyses. Bioinformatics 25, 1972–1973 be present in a sample if it had an RA >0.01% in that particular sample. (2009). To calculate the average relative abundance (y-axis in Extended Data 55. Price, M. N., Dehal, P. S. & Arkin, A. P. FastTree 2—approximately maximum-likelihood trees for large alignments. PLoS ONE 5, e9490 (2010). Figs. 10a, b) independently of prevalence, we calculated the mean RA 56. Gordon, S. A. & Weber, R. P. Colorimetric estimation of indoleacetic acid. Plant Physiol. using only the sample for which each ASV was considered present. 26, 192–195 (1951). 57. Friml, J. et al. Efflux-dependent auxin gradients establish the apical–basal axis of P. Schulze-Lefert for strains; the Dangl laboratory microbiome group, A. Stepanova, J. Alonso, Arabidopsis. Nature 426, 147–153 (2003). J. Brumos, J. Kieber, J. Reed and I. Greenhut for useful discussions; and D. Lundberg and 58. Altschul, S. F., Gish, W., Miller, W., Myers, E. W. & Lipman, D. J. Basic local alignment A. Bishopp for critical comments on the manuscript. This work was supported by NSF grant search tool. J. Mol. Biol. 215, 403–410 (1990). IOS-1917270 and by Office of Science (BER), US Department of Energy, Grant DE-SC0014395 59. Ebenau-Jehle, C. et al. Anaerobic metabolism of indoleacetate. J. Bacteriol. 194, to J.L.D. J.L.D. is an Investigator of the Howard Hughes Medical Institute, supported by the 2894–2903 (2012). HHMI. O.M.F. was supported by NIH NRSA Fellowship F32-GM117758. C.R.F. was supported by 60. Emms, D. M. & Kelly, S. OrthoFinder: phylogenetic orthology inference for comparative an NSERC postdoctoral fellowship (532852-2019). genomics. Genome Biol. 20, 238 (2019). 61. Langmead, B. & Salzberg, S. L. Fast gapped-read alignment with Bowtie 2. Nat. Methods Author contributions J.L.D. supervised the project. O.M.F., I.S.-G., G.C. and J.L.D. 9, 357–359 (2012). conceptualized the project. O.M.F., I.S.-G., G.C., J.M.C. and J.L.D. designed experiments. O.M.F., 62. Kovach, M. E. et al. Four new derivatives of the broad-host-range cloning vector I.S.-G., G.C. and T.F.L. designed and performed most in planta agar-based experiments. C.R.F., pBBR1MCS, carrying different antibiotic-resistance cassettes. Gene 166, 175–176 (1995). O.M.F. and I.S.-G. designed and performed soil experiments. O.M.F., I.S.-G., G.C. and T.F.L. 63. Figurski, D. H. & Helinski, D. R. Replication of an origin-containing derivative of plasmid processed and sequenced bacterial DNA samples. J.M.C. and E.D.W. designed and performed RK2 dependent on a plasmid function provided in trans. Proc. Natl Acad. Sci. USA 76, bacterial gain- and loss-of function experiments and bacterial RNA-seq experiments. O.M.F., 1648–1652 (1979). T.F.L. and P.J.P.L.T. processed and sequenced plant RNA-seq samples. O.M.F. and G.C. 64. Hamad, M. A., Zajdowicz, S. L., Holmes, R. K. & Voskuil, M. I. An allelic exchange system performed plant morphometric measurements. I.S.-G., O.M.F., G.C. and J.M.C analysed data. for compliant genetic manipulation of the select agents Burkholderia pseudomallei and I.S.-G. performed bioinformatics analyses, including bacterial 16S rRNA, plant and bacterial Burkholderia mallei. Gene 430, 123–131 (2009). RNA-seq analysis and bacterial comparative genomics, with input from O.M.F., G.C., J.M.C. and 65. Callahan, B. J. et al. DADA2: high-resolution sample inference from Illumina amplicon J.L.D. I.S.-G. performed statistical analysis, with input from O.M.F., G.C., C.D.J. and J.L.D. I.S.-G. data. Nat. Methods 13, 581–583 (2016). curated the data used in the manuscript with help from O.M.F. O.M.F., I.S.-G. and G.C. wrote the 66. Schloss, P. D. et al. Introducing mothur: open-source, platform-independent, manuscript with input from all co-authors. community-supported software for describing and comparing microbial communities. Appl. Environ. Microbiol. 75, 7537–7541 (2009). Competing interests J.L.D. is a co-founder of, and shareholder in, AgBiome LLC, a corporation 67. Quast, C. et al. The SILVA ribosomal RNA gene database project: improved data with the goal of using plant-associated microorganisms to improve plant productivity. processing and web-based tools. Nucleic Acids Res. 41, D590–D596 (2013). 68. Winter, D. et al. An “Electronic Fluorescent Pictograph” browser for exploring and Additional information analyzing large-scale biological data sets. PLoS ONE 2, e718 (2007). Supplementary information is available for this paper at https://doi.org/10.1038/s41586-020- 2778-7. Acknowledgements We thank S. Barth, J. Shen, M. Priegel, D. Chudasma, D. Panda, I. Castillo, Correspondence and requests for materials should be addressed to J.L.D. N. Del Risco, C. Lindberg and R. Pérez-Torres for technical assistance; D. Pelletier and Reprints and permissions information is available at http://www.nature.com/reprints. Article a b Corynebacteriales Wild soil SynCom abundance pro les Propionibacteriales 1.00 1.00 Streptomycetales Bacterial Phyla (Panels a/b) Bacillales Acidobacteria 0.75 0.75 Actinobacteria Bacteroidetes

Micrococcales Chloroexi

Cyanobacteria 0.50 0.50 0.1 Firmicutes

oportion of ASVs Proteobacteria Relative abundanc e Pr Verrucomicrobia Pseudomonadales 0.25 0.25 Flavobacteriales Other Sphingobacteriales Xanthomonadales Caulobacterales

0.00 0.00 Burkholderiales Rhizobiales Root enriched e Experiment Fraction Root Shoot Phosphate Substrate Inoculum Substrat Salinity Root c d pH Shoot Relative abundance in substrate Relative abundance in root Enrichment (Root vs Substrate) Temperature R2=0.22 -1 1e-1 1e 5.0

-2 1e-2 1e m m m

2.5 1e-3 1e-3 Soil syste Soil syste Soil syste CAP2(20.94%) 1e-4 1e-4 0.0 r = 0.67 r = 0.80 r = 0.48

-13 -22 -6 1e-5 p = 1.4e 1e-5 p = 4.1e p = 1.6e -2.5

1e-5 1e-4 1e-3 1e-2 1e-1 1e-5 1e-4 1e-3 1e-2 1e-1 -4 04 8 Agar system Agar system Agar system CAP1(40.87%) e Phosphate Salinity f Substrate RootShoot Substrate Root Shoot Substrate Root Shoot 3.5 3.5 R2=0.30 R2=0.42 R2=0.40

BC DD BC C ABCD D BCDBCD 3.0 ABCD ABCD ABCD 3.0 ABCD CD ABCD ABC ABCD ABC ABC

AAAA Phosphate AA A AB ABC AB

A CAP2(17.97%) CAP2(17.16%) CAP2(15.33%) 2.5 2.5 CAP1(54.71%) CAP1(58.06%) CAP1(54.81%) R2=0.45 R2=0.49 R2=0.45

Shannon diversity inde x 2.0 Shannon diversity inde x 2.0 y

0 0 0 0 0 0 Salinit 10 30 50 10 30 50 10 30 50 50 50 50 100 100 100 100 150 20 10 150 200 100 150 20 1000 1000 1000 CAP2(24.79%) CAP2(27.82%) CAP2(21.81%) pH Temperature CAP1(66.79%) CAP1(60.44%) CAP1(70.24%) Substrate RootShoot Substrate Root Shoot R2=0.30 R2=0.24 R2=0.28 3.5 C C 3.5 C

BC BC BC BC BC BC

3.0 B 3.0 pH BB B B

AA A CAP2(46.51%) CAP2(34.42%) CAP2(30.46%)

2.5 2.5 CAP1(53.49%) CAP1(65.58%) CAP1(69.54%) R2=0.73 R2=0.43 R2=0.79 A

Shannon diversity inde x 2.0 Shannon diversity inde x 2.0 mperatu re

0 5 5 C Te

5.5 7. 8.2 5. 7.0 8.2 5. 7.0 8.2 CAP2(17.88%) CAP2(31.62%) CAP2(19.26%) 10°C 21°C 31°C 10° 21°C 31°C 10°C 21°C 31°C Phosphate (Panels e/f) 0μM 10μM 30μM 50μM 100μM 1000μM CAP1(82.12%) CAP1(68.38%) CAP1(80.74%) Salinity (Panels e/f) pH (Panels e/f) Temperature (Panels e/f) 50mM 100mM 150mM 200mM 5.5 7.0 8.2 10°C 21°C 31°C

Extended Data Fig. 1 | See next page for caption. Extended Data Fig. 1 | Synthetic community resembles the taxonomic P value are shown for each comparison. n = 24 (soil system) and 8 (agar system) make-up of natural communities. a, Phylogenetic tree of 185 genome- biological replicates. d, Canonical analysis of principal coordinates (CAP) sequenced isolates obtained from surface-sterilized Arabidopsis roots, showing the influence of the fraction (planted agar, root or shoot) on the included in the synthetic community. The composition of this synthetic assembly of the bacterial synthetic community across the four gradients used community captures the diversity of Actinobacteria, Proteobacteria and in this Article (phosphate, salinity, pH and temperature). Different colours Bacteroidetes; the three major root-enriched phyla1,4–6,20; and Firmicutes, differentiate between the fractions, and different shapes differentiate between which are abundant in plant-associated culture collections20. Tree tips are experiments. Ellipses denote the 95% confidence interval of each fraction. coloured according to phylum. The outer ring shows the distribution of the Fraction (substrate, root or shoot) explains most (22%) of the variance across 12 bacterial orders present in the synthetic community. b, Comparison of all abiotic variables. n = 94 (substrate), 90 (root) and 95 (shoot) biological proportions of Proteobacteria, Bacteroidetes and Actinobacteria in synthetic- replicates across 8 independent experiments. e, Abiotic conditions displayed community (SynCom)-inoculated roots to root microbiota derived from plants reproducible effects on α-diversity. Each panel represents bacterial α-diversity grown in natural soil1. Firmicutes, which are not plant-enriched, were reduced across the different abiotic gradients (phosphate, salinity, pH and to <0.1% of the relative abundance (Fig. 1a). The left panel (wild soil) shows the temperature) and fractions (substrate, root and shoot) used in this Article. proportion of ASVs enriched (q value < 0.1) in the plant root in comparison to Bacterial α-diversity was estimated using Shannon diversity index. Letters soil in a microbiota profiling study from the same soil from which the synthetic represent the results of the post hoc test of an ANOVA model testing the community strains were isolated. ASVs are coloured according to phylum and interaction between fraction and abiotic condition. f, Canonical analysis of proteobacterial ASVs are coloured by class. The right panel (synthetic principal coordinates scatter plots showing the effect on the composition of community) represents the relative abundance profiles of bacterial isolates the synthetic community of each of the four abiotic gradients (phosphate, across the initial inoculum, planted agar, and root and shoot in plants salinity, pH and temperature) within the substrate, root and shoot fractions. inoculated with the full synthetic community. c, Comparison of synthetic PERMANOVA R2 values are shown within each plot. e, f, Phosphate, n = 6, 5, 6, 6, community composition in agar versus the soil-based microcosms. Left, 6, 6, 6, 6, 6, 6, 6, 6, 6, 6, 6, 6, 5 and 6; salinity n = 6, 6, 6, 6, 6, 6, 6, 4, 5, 6, 4 and 5; pH, relative abundance in the substrate. Middle, relative abundance in root. Right, n = 6, 6, 6, 5, 5, 6, 6, 6 and 6; temperature n = 6, 6, 6, 6, 6, 6, 6, 6 and 6. All samples enrichment in root versus substrate. Each dot represents a single unique are biologically independent and represent two independent experiments. sequence. Pearson correlation line, 95% confidence intervals, r value and Article

a Module A Module B Module C Module D Panel a 8 IQR Module Panels a/b 6 IQR No bacteria

Cutoff RGI (Panels a/b) 4 Taxonomy (Panels a/b) Firmicutes 2 Actinobacteria Primary root elongation (cm) Bacteroidetes Alphaproteobacteria 0 Gammaproteobacteria b c

m r u as ri e on Paenibacillus a 9 m ri 5 9 e 38 23 o 4 6 F L6 domonas illus ud u M MF e MF C e MF36 ac 10 hodococcus cinetobacte bact F B Agrobact R Ps A Ps No M Bacillus

Nocardia Rhodococcus

Mycobacterium

Nocardioides 10 cm

Streptomyces s r um Terracoccus e nas na t ter Promicromonospora c 4 1 1 cteri mo 8 6 mo o 22 us 5 25 F 7 162 37 do ud F Arthrobacter F CL rkholderiaF throba F366 M cill 10 M MF13 a M u M seu M F Ar Luteibacter Agroba B Arthrobac Pse B P M Leucobacter Curtobacterium Leifsonia

Microbacterium Flavobacterium s r r as r r a a Chryseobacterium er e e te te t t Sphingomonas c c c 3 a 1 noma 4 a Caulobacter bacter 7 26 ob F6 F Brevundimonas L28 rob F r hrobact F231 Bosea C CL63 MF11 MF48 M M MF16 MF25 M M Methylobacterium Arthrobac Arthroba Arth Pseudomon Pseudo Art Arthro Arth Burkholderi Burkholderi

Ochrobactrum Phyllobacterium Rhizobium

Agrobacterium s Ralstonia illus llus illus c s Burkholderia ba 1 lus i ibaci ibac il 2 c enibacillu 49 L52 n L117 CL91 e L7 CL130 C C Bacillu Ba C Paen Pa Variovorax MF Paen Pa Stenotrophomonas Dyella Luteibacter Acinetobacter Pseudomonas

Phylogenetic tree Genus

Extended Data Fig. 2 | RGI trait is distributed across bacterial phylogeny. coloured on the basis of their genome-based taxonomy. The vertical blue a, b, Primary root elongation of seedlings inoculated with single bacterial isolates stripes across the panel correspond to the IQR of plants grown in sterile (one box plot per isolate). Isolates are coloured by taxonomy. a, Isolates are conditions. The vertical dotted line represents the 3-cm cut-off used to classify grouped by module membership. The strips across the panels correspond to strains as RGI strains. The bar on the right denotes the genus classification of the interquartile range (IQR), as noted at the far right. The dotted line represents each isolate. c, Binarized image of representative seedlings grown axenically the cut-off used to classify isolates as root-growth inhibiting (cutoff RGI). (no bacteria) or with 34 RGI strains individually. b, Isolates are ordered according to the phylogenetic tree on the left, and a Representative RGI reverters b Tomato Module AAA B B CL91

CL130 Paenibacillus CL52 Taxonomy dendrogram Firmicutes 12 MF107 A Actinobacteria Bacillus CL72 Alphaproteobacteria MF491 Gammaproteobacteria 8 MF131

MF254 Primary root elongation (cm) MF31 Arthrobacter MF231 0 2 4 6 8 Primary root elongation (cm) 4 owth inhibitors CL28

Root gr MF26 Signi cance MF10 No bacteria Arthrobacter Arthrobacter CL28 Agrobacterium q < 0.05 & CL28 Variovorax CL14 MF224 estimate > 2 TRUE MF48 Pseudomonas MF3 MF6 Burkholderia MF7 0 Self CL11 CL14 MF16 ariovorax ariovorax V V Burkholderia

Extended Data Fig. 3 | Variovorax-mediated reversion of RGI. a, Heat map second plant species. Primary root elongation of uninoculated tomato coloured by average primary root elongation of seedlings inoculated with seedlings (no bacteria) or seedlings inoculated with the RGI-inducer eighteen RGI-inducing strains (rows) alone (self) or in combination with Arthrobacter CL28 individually or along with Variovorax CL14 grown on vertical Burkholderia CL11, Variovorax CL14 or Variovorax MF160 (columns). agar plates. Significance was determined via ANOVA while controlling for Statistically significant RGI reversions were determined via ANOVA and are experiment; letters correspond to a Tukey post hoc test. n = 24, 25 and 23 outlined in black. b, Variovorax-mediated reversion of RGI is maintained in a biological replicates across 2 independent experiments. Article a b Module Module Module 34- Full SynCom-Vario SynCom Phylogenetic tree Genus A C D member

Paenibacillus

Bacillus

Representative plates Nocardia Rhodococcus

Mycobacterium Nocardioides Streptomyces Terracoccus Promicromonospora

Arthrobacter Taxonomy (Panel b) Firmicutes Leucobacter Curtobacterium Actinobacteria Leifsonia Bacteroidetes Microbacterium Alphaproteobacteria Flavobacterium Chryseobacterium Sphingomonas Gammaproteobacteria Caulobacter Brevundimonas Bosea Inoculation (Panel b) Methylobacterium Uninoculated Ochrobactrum Phyllobacterium Inoculated Rhizobium

Agrobacterium Ralstonia Burkholderia

Variovorax Stenotrophomonas Dyella Luteibacter Acinetobacter

Pseudomonas

SynCom c JM agar JM agar JM agar JM agar JM agar MS agar d Control 10μM Pi 150mM NaCl pH 8.2 31°C Control Soil 10 q = 3.1e-31 q = 2.8e-4 q = 8.6e-1 q = 1.5e-9 q = 3.9e-8 q = 3.6e-13 60 p = 0.00011 8 50 SynCom (Panel c/d) Full 6 40 -Vario

30

Total root network (cm) 4 Shoot circumference (cm)

20 2

Full-Vario Full-Vario Full -Vario Full -Vario Full -Vario Full -Vario Full -Vario

Extended Data Fig. 4 | Variovorax maintain stereotypic plant growth. (−Vario), across different abiotic conditions: JM agar control, low phosphate a, Representative plate images of plants grown with the full 185-member (JM agar 10 μM Pi), high salt (JM agar 150 mM NaCl), high pH (JM agar pH 8.2) synthetic community (full SynCom) or the Variovorax drop-out community and high temperature (JM agar 31 °C), as well as half Murashige and Skoog (−Vario SynCom) for 12 d. b, Bar graphs showing the isolate composition of medium (MS agar control). Significance was determined within each condition synthetic communities composed by module A, module C, module D and a via ANOVA while controlling for experiment. n = 110, 77, 42, 44, 41, 35, 48, 49, 44, previously described1 34-member synthetic community (34-member). Isolates 45, 33 and 26 biological replicates across 2 independent experiments. d, Shoot are ordered according to the phylogenetic tree on the left. The tips of the circumference of plants grown in pots with potting soil with the full synthetic phylogenetic tree are coloured on the basis of the genome-based taxonomy of community or with the full synthetic community excluding Variovorax. each isolate. Presence of an isolate across the different synthetic communities Significance was determined within each condition via ANOVA while is denoted by a black filled rectangle (labelled ‘inoculated’). c, Total root controlling for experiment. n = 30 and 36 biological replicates across 2 network quantification of Arabidopsis seedlings grown with the full synthetic independent experiments. community (full), or with the full synthetic community excluding Variovorax a Paraburkholderia phytofirmans PsJN iac Operon scanning Pseudomonas putida 1290

Burkholderia CL11 Acidovorax Root219 Genus (Phylogeny) Variovorax OV329 Variovorax BK119 Burkholderia Variovorax BT01 Variovorax NBRC 106424 Variovorax URHB0020 Acidovorax Variovorax YR216 Variovorax HW608 Variovorax CF079 Variovorax WDL1 Variovorax Variovorax KK3 Variovorax PAMC 28711 Variovorax PDC80 Variovorax NP4 Status (Bar graphs) Variovorax H084 Variovorax H061 Negative Variovorax H112 Variovorax H108 Variovorax H095 Variovorax H090 Positive Variovorax OV700 Variovorax Root318D1 Variovorax T529 Variovorax TBEA6 Untested Variovorax B4 Variovorax NBRC 15149 Variovorax MF4 Variovorax CL14 Percent Variovorax MF369 Variovorax MF349 identity to best Variovorax MF295 Variovorax MEDvA23 hit in each genome Variovorax 770b2 Variovorax 54 (Heatmap) Variovorax Root473 Variovorax J1 50 Variovorax NBRC 103145 Variovorax EL159 Variovorax Root411 Variovorax Root434 40 Variovorax CF313 Variovorax GV014 Variovorax GV008 Variovorax OK605 Variovorax OK212 30 Variovorax OK202 Variovorax GV004 Variovorax GV042 Variovorax GV035 20 Variovorax GV040 Variovorax GV051 Variovorax PMC12 Not found Variovorax KB5 Variovorax BK460 Variovorax OV084 Variovorax YR266 Variovorax YR752 Annotation (Bottom panel) Variovorax YR634 Variovorax YR750 Gene required Variovorax NFACC29 Variovorax NFACC28 for growth on IAA Variovorax NFACC27 ¤ Variovorax NFACC26 (Donoso et al. 2017) Variovorax MF110 Variovorax BK151 Variovorax BK370 Homolog present in Variovorax MF160 Variovorax Mf350 hotspot 33 (Figure 4a) Variovorax MF375 * Variovorax MF278

E iacA iacB iacC iacD iac iacF iacH iacI iacY ACC iacG iacR1 iacR2

CL28 RGI deaminase ¤¤¤¤¤¤ ¤ reversion b Co-inoculation** with* Arthrobacter** CL28 NS Burkholderia CL11 NS Genus Acidovorax Root219 *** Variovorax YR216 Burkholderia *** Variovorax PDC80 ** Acidovorax Variovorax Root318D1 ** ** Variovorax B4

** Variovorax Variovorax MF4 ** Variovorax CL14 ** ** Variovorax MF369 ** IQR Variovorax MF349 ** No bacteria ** Variovorax Root473 ** Cutoff RGI Variovorax CF313 ** Variovorax Root411 ** **

Variovorax OK605 ** Variovorax YR266 **

** Signi cance Variovorax YR750 ** Variovorax MF110 ** NS q > 0.05 ** ** Variovorax MF160 ** q < 0.01 Variovorax MF350 ** **

Variovorax MF375 ** *** q < 0.001 Variovorax MF278 **

02468 Primary root elongation (cm)

Extended Data Fig. 5 | See next page for caption. Article Extended Data Fig. 5 | Reversion of RGI is prevalent across the Variovorax genomes. b,All tested Variovorax isolates reverted RGI. Phylogenetic tree of 19 phylogeny. a, Phylogenetic tree of 69 publically available Variovorax genomes Variovorax genomes and 2 outgroup isolates (Acidovorax root 219 and and 2 outgroup isolates, Acidovorax root 219 and Burkholderia CL11. The CL28 Burkholderia CL11) that were tested for their ability to revert the RGI imposed RGI reversion bar categorizes (positive, negative or untested) the ability of by Arthrobacter CL28. The blue vertical stripe denotes the IQR of plants treated each isolate in the phylogeny to revert the RGI caused by Arthrobacter CL28. solely with Arthrobacter CL28. The dotted vertical line denotes the 3-cm cut-off The ACC deaminase bar denotes the presence of the KEGG Orthology (KO) term used to classify a treatment as an RGI. Each box plot is coloured according to KO1505 (1-aminocyclopropane-1-carboxylate deaminase) in each of the the genus classification of each isolate. Significance was determined via genomes. The heat map denotes the per cent identity of BLASTp hits in the ANOVA while controlling for experiment, letters correspond to a Tukey post genomes to the genes from the auxin-degrading iac operon in hoc test. n = 59, 9, 55, 71, 10, 10, 10, 9, 10, 10, 57, 10, 10, 48, 10, 10, 10, 9, 10, 10 and 9 Paraburkholderia phytophirmans, described by ref. 17. Synteny is not biological replicates across 2 independent experiments. necessarily conserved, as these BLAST hits may be spread throughout the = 12 biologically independent samples. n = 12 independent biologically test. Tukey hoc post a to correspond letters ANOVA, via determined was Significance inducers. RGI outcompeting or byantagonizing growth root stereotypic enforces nor nor root community (Fig. community root strains. 6|Variovorax Fig. Data Extended presence of Variovorax of presence Arthrobacter (Fig. Variovorax the to response in strains RGI-inducing of abundances the the Variovorax with colonized seedlings of that to community synthetic full the with colonized seedlings in profiles abundance relative bacterial the compared we strains, RGI-inducing representatives: representatives: agar plates containing 50 μg ml μg 50 containing plates agar selectively count Arthrobacter Arthrobacter of CFUs log-transformed Variovorax 2g ). In addition, we measured in planta absolute abundance of of abundance absolute planta in measured we addition, ). In To test whether Variovorax To whether test

CL28 when inoculated alone or with two Variovorax two with or alone inoculated when CL28 log CFUs/mg CL14 grow. Arthrobacter grow. CL14 10 0 1 2 3 4 5 6 Variovorax Arthrobacter Arthrobacter 2h . Notably, Variovorax . Notably, CL28 ). These results rule out the possibility that Variovorax that possibility the out rule results ). These A B4 and Variovorax and B4 drop-out community. We found no changes in CL28, CFUs were counted on Luria Bertani (LB) Bertani Luria on counted were CFUs CL28, −1 CL28 absoluteabundanceinro Inoculation tr of apramycin, on which neither Variovorax neither which on apramycin, of Arthrobacter does not compete with or antagonize RGI RGI antagonize or with compete not does attenuates RGI by inhibiting the growth of of growth the byinhibiting RGI attenuates Variovorax CL28 normalized to root weight. To weight. root to normalized CL28 CL28 CFUs are not reduced in the the in reduced not are CFUs CL28 B account for only about 1.5% of the the of 1.5% about only for account eatment CL28 B4 CL14. Here we show show we Here CL14.

Arthrobacter Variovorax B

CL14 CL28 ot

drop-out B4 B4

Article

a Tripartite system (45 genes) Drop-out system (29 genes) b Tripartite Drop out Pearson Correlation 0 0.5 1 1 27 18 11 GH3.2 ZIP3 EXT6 ATCAPE5 Root p = 1.5e-43 ATXTH26 0 AT4G26010 RHS7 apex AT4G08400

Without AT5G06630 RHS13 EXT13 EXT12 With apex Gene expression (z-score) -1 ATHRGP1 PRP3 EXT10 AGP3 XTH14 AT4G08410 6 NB NB ZIP3 EXT PRP3 AGP3 RHS7 GH3.2 EXT10 EXT12 EXT13 XTH14 XTH26 riovoraxCL14 CL28 RHS13 CAPE5

SynCom HRGP1 Va ario/-Burk 4G08410 5G06630 4G08400 4G26010

CL14/CL28 AT AT Arthrobacter ario SynCom-V Full SynCom AT

-Burk SynCom-V AT AT AT AT d c root morphogenesis Tripartite system Drop-out system root epidermal cell differentiation trichoblast differentiation -16 cation homeostasis p < 2e p = 0.017 cellular metal ion homeostasis GH3.2 inorganic ion homeostasis plant epidermal cell differentiation metal ion homeostasis GH3.2 cellular ion homeostasis GH3.3 iron ion transport transition metal ion homeostasis WES1 transition metal ion transport root hair cell differentiation 1 MSG2 LBD29

response to nitrate ion homeostasis -score) AT5G06865 ATACS4 cellular response to iron ion starvation cell maturation PGIP1 GH3.3 inorganic anion transport root hair cell development LBD29 MAKR4 cellular cation homeostasis cellular transition metal ion homeostasis trichoblast maturation ion (z IAA1 cellular response to extracellular stimulus cellular iron ion homeostasis IAA29 MSG2 iron ion homeostasis root hair elongation IAA1 cellular homeostasis ATACS4 WES1 cellular chemical homeostasisdevelopmental cell growth AT5G67430 AT5G67430 WES1 response to starvation IAA29 cellular response to nutrient levels 0 IAA1 chemical homeostasis response to nutrient levels MAKR4 ATACS4 nitrate transport IAA1 PGIP1 MAKR4 cellular response to starvation Gene express MSG2 PGIP1 cellular response to external stimulus MAKR4 ATACS4 GH3.3 AT5G06865 AT5G67430 IAA29 GH3.2 MSG2 homeostatic process response to extracellular stimulus cellular response to zinc ion transport LBD29 LBD29 auxin stimulus WES1 IAA29 AT5G67430 PGIP1 GH3.3 GH3.2 AT5G06865 AT5G06865 Arthrobacter Arthrobacter CL28 -Vario Full CL28 Variovorax CL14 SynCom SynCom response to auxin

auxin homeostasis

Extended Data Fig. 7 | Auxin-responsive genes are induced in response to shown, illustrating the spatial distribution of transcripts from the 17 highly RGI strains. a, Box plots showing the average standardized expression of correlated root apex-associated genes in the Klepikova Atlas25. Significance of genes significantly induced only under RGI conditions (RGI-induced), across the overlap of enriched genes was determined via hypergeometric test. the following treatments. Left (tripartite system), uninoculated seedlings (NB) c, RGI-related genes share gene ontologies. Network of statistically significant or seedlings inoculated with Variovorax CL14, Arthrobacter CL28 or both gene ontology terms contained within the 18 overlapping RGI-induced Arthrobacter CL28 and Variovorax CL14 (CL14/CL28). Right (drop-out system), genes (a, b). The network was computed using the emapplot function from the uninoculated seedlings (NB), Burkholderia drop-out synthetic community package clusterProfiler in R. A P value for terms across the gene ontology was (−Burk SynCom), Variovorax drop-out synthetic community (−Vario SynCom), computed using a hypergeometric test (only significant ontologies are shown). Variovorax and Burkholdria drop-out synthetic community (−Vario −Burk Point size (gene ontology term) denotes the number of genes mapped to that SynCom) or the full synthetic community (full SynCom). RGI-induced genes are particular term. d, Standardized expression of 12 late-responsive auxin genes26 defined as genes that are significantly overexpressed in RGI treatments. Left, across the tripartite and drop-out systems. Each dot represents a gene. genes overexpressed in Arthrobacter CL28-inoculated seedling versus NB and Identical genes are connected between bacterial treatments with a black line. in Arthrobacter CL28-inoculated seedlings versus CL14/CL28. Right, genes Mean expression (95% confidence intervals) of the aggregated 12 genes in each overexpressed in −Vario versus NB and in −Vario versus full. n = 3 (left); 5 (right). treatment is highlighted in red and connected between bacterial treatments b, Venn diagram showing the overlap of enriched genes between the tripartite with a red line. Significance was calculated using a resampling approach, in and drop-out systems. The heat map shows the pairwise correlation in which we compared the calculated difference between means across groups. expression of these 18 genes across tissues on the basis of the Klepikova Atlas25. After 10,000 resamplings, we calculated the P value by comparing the Seventeen of the 18 genes show high correlation across the atlas, with the distribution of means calculated against the real observed differences exception of the auxin-conjugating gene GH3.2. A root apex diagram from the between means. Arabidopsis eFP Browser (http://bar.utoronto.ca/efp/cgi-bin/efpWeb.cgi)68 is a Arthrobacter IAA 2,4-D ACC g22 BAP Zeatin CL28 100nM 100nM 100nM 100nM 100nM 100nM 8 CCC Genus 6 C BC CC C B Self B BC B A A BB 4 B AB AB B B AB AB AB Burkholderia A A AA D AB A A A ot elongation (cm) CD AB A A A BC Variovorax 2 BC AB A

0 Primary ro B4 B4 B4 B4 B4 B4 B4 Self Self Self Self Self Self Self CL11 CL11 CL11 CL11 CL11 CL11 CL11 CL14 CL14 CL14 CL14 CL14 CL14 CL14 YR21 6 YR21 6 YR21 6 YR21 6 YR21 6 YR21 6 YR21 6 MF160 MF160 MF160 MF160 MF160 MF160 MF160 Treatment

M9 IAA (+ 0.5% ethanol) b M9 IAA (+ 0.5% ethanol) M9 succinate c + succinate Tripartite system 0.6 60

0 *** 0.4

OD60 40 *** 0.2 ***

20

DR5::GFP intensity *** 0.06

Day 1Day 3Day 6Day 9Day 13 0.04 Treatment Bacterial treatment Signi cance No bacteria (CL28 vs CL28+CL14) 0.02 Variovorax CL14 Arthrobacter CL28 q < 0.001 IAA concentration Arthrobacter CL28 *** 0.00 No bacteria Variovorax CL14 010203040 0102030400 10 20 30 40 Hours d IAA + Arthrobacter CL28 + No bacteria IAA Variovorax CL14 Arthrobacter CL28 Variovorax CL14

Day 1

Day 3

Day 6

100 µm

Extended Data Fig. 8 | Variovorax degrades auxin and quenches auxin quenches induction of the auxin bioreporter DR5::GFP. Quantification of GFP perception by the plant. a, Primary root elongation of seedlings grown with intensity in DR5::GFP Arabidopsis seedlings grown with no bacteria, six hormonal or microbial-associated molecular pattern RGI treatments Arthrobacter CL28 and Arthrobacter CL28 + Variovorax CL14. GFP fluorescence (panels) individually (self) or with either Burkholderia CL11 or each of four was imaged 1, 3, 6, 9 and 13 d after inoculation, and quantified in the root Variovorax isolates (CL14, MF160, B4 and YR216). Significance was determined elongation zone. Significance was determined via ANOVA within each time via ANOVA within each panel; letters correspond to a Tukey post hoc test. n = 74, point, while controlling for experiment, and denoted with asterisks. n = 20 46, 61, 48, 49, 49, 45, 44, 46, 43, 49, 40, 22, 19, 22, 19, 20, 25, 28, 30, 29, 29, 29, 29, biological replicates across 2 independent experiments. d, Representative 12, 12, 21, 20, 19, 18, 26, 30, 30, 29, 30, 30, 29, 30, 30, 26, 29 and 28 biological primary root images of DR5::GFP plants quantified in c, showing roots from 1, 3 replicates across 2 independent experiments. b, Variovorax degrades auxin. and 6 d after inoculation. In addition to the bacterial treatments shown in c, an

Growth curves showing optical density at OD600 (top) and IAA concentrations exogenous IAA control is shown (IAA, second column), as well as IAA-treated (mg ml−1) (bottom) in Variovorax CL14 cultures grown in M9 medium with plants inoculated with Variovorax CL14, illustrating that IAA-induced different carbon sources. Left, IAA (+ 0.5% ethanol as solvent); middle, fluorescence is quenched in the presence of Variovorax CL14 within 3 d. succinate; right, succinate and IAA (+ 0.5% ethanol as solvent). c, Variovorax Article a Hotspot Genus 145 33 148 146106 4116 86 89 179 76 53

Paenibacillus

Bacillus

Nocardia Rhodococcus Taxonomy (Phylogeny) Firmicutes Mycobacterium Actinobacteria Nocardioides Streptomyces Terracoccus Bacteroidetes Promicromonospora Alphaproteobacteria Arthrobacter Gammaproteobacteria Leucobacter Curtobacterium

Leifsonia Orthogroup presence in genome (Heatmap) Microbacterium Flavobacterium Chryseobacterium Negative Sphingomonas Positive Caulobacter Brevundimonas Bosea Methylobacterium Ochrobactrum Phyllobacterium Rhizobium Agrobacterium Ralstonia Burkholderia

Variovorax Stenotrophomonas Dyella Luteibacter Acinetobacter Pseudomonas

Orthogroup Id b Scaffold Ga0102008_10005 Variovorax CL14 Variovorax CL14 co-cultured 45000 50000 55000 60000 65000 70000 with Arthrobacter CL28 26436136- 53 54 55 56 57 58 59 60 61 62 63 64 65 66 67 68 69 70 71 72 73 74 75 76 77 vs

Variovorax CL14 → monocultured Vector 1 (V1) Vector 2 (V2) -3 -2 -1 0123 Aromatic-ring-hydroxylating ∆HS33 Amp-dependent RNASeq 53,67 59,68 dioxygenase beta subunit 66 Thiolase 74 Alcohol dehydrogenase log fold change synthetase and ligase 2 (PsJN iacD 22/26%) Rieske 2fe-2S Oxidoreductase Alpha beta 54,56 (ABC) transporter 60,69 domain-containing protein 70 75 (PsJN iacF 35%) hydrolase (PsJN iacC 28/31%) Short-chain Indolepyruvate 3-hydroxyacyl-CoA 55 Hypothetical protein 63 dehydrogenase reductase 71 76 oxidoreductase subunit B dehydrogenase (PsJN iacE 28%) Inner-membrane Indolepyruvate ferredoxin Octopine nopaline 57 64 S cyclase 72 77 translocator oxidoreductase dehydrogenase MarR Extracellular ligand-binding Dehydrogenase 58,61, 65 73 Transcriptional regulator 62 receptor (PsJN iacE 25%) (P. putida 1290 iacR 30%) c de Variovorax CL14 co-cultured Hotspot 8 8 with Arthrobacter CL28 vs 145 33 148 146106 4116 86 89 179 76 53 7 7 Variovorax CL14 6 6 monocultured 5 5 -3 -2 -1 0123 4 4 CFUs/mg CFUs/mg

RNASeq Orthogroup Id 10 3 10 3

log 2 log log2 fold change 2 1 1 0 0

2 4 4

ariovoraxCL1 ariovorax CL1 AcidovoraxV V Variovorax

Root219::V CL14 ∆HS33

Extended Data Fig. 9 | See next page for caption. Extended Data Fig. 9 | Detection of CL28-responsive Variovorax-unique unique genomic hotspots are presented, aligned with the genes in a. Genes are operons. a, Presence–absence matrix denoting the distribution of 12 coloured by the log2-transformed fold change in their transcript abundance in Variovorax-unique hotspots containing at least 10 genes across the 185 Variovorax CL14 cocultured with Arthrobacter CL28 vs Variovorax CL14 members of the synthetic community. Hotspots are defined using the monoculture. Note uniform upregulation of genes in cluster 33. d, log- Variovorax CL14 genome as a reference. Phylogeny of the 185 members of the transformed CFUs of Variovorax CL14 or with the Acidovorax gain-of-function synthetic community is shown to the left of the matrix. We determined the strain Acidovorax root219::V2 normalized to root weight. Each of these two presence of an orthogroup based on a hidden Markov model profile scanning strains was co-inoculated with Arthrobacter CL28 onto 7-d-old seedlings and of each core Variovorax (genus) orthogroup across the 185 genomes in the collected after 12 d of growth. CFU counts were determined on LB plates synthetic community. b, A map of the auxin-degrading hotspot 33. Genes are containing 100 μg ml−1 ampicillin, for which Arthrobacter CL28 is susceptible annotated with the last two digits of their IMG gene identifier (26436136XX) and Variovorax CL14 and Acidovorax root219::V2 are naturally resistant. and their functional assignments are shown below the map, including per cent n = 3 biologically independent samples. Significance was determined using identity of any to genes from a known auxin degradation locus. Genes are a two-sided Student’s t-test. P = 2.57 × 10−5. Line represents median. coloured by the log2-transformed fold change in their transcript abundance in e, log-transformed CFUs of Variovorax CL14 or of the loss-of-function strain Variovorax CL14 cocultured with Arthrobacter CL28 versus Variovorax CL14 Variovorax CL14 ΔHS33 normalized to root weight. Each of these two strains monoculture, as measured by RNA-seq (shown in c). The overlap of this region was inoculated individually onto 7-d-old seedlings and collected after 12 d of with vectors 1 and 2 and the region knocked out in Variovorax CL14 ΔHS33 are growth. CFU counts were determined on LB plates. n = 5 biologically shown below the map. Vector 1 extends beyond this region. c, Results of independent samples. Significance was determined using a two-sided RNA-seq on Variovorax CL14 transcripts. Variovorax CL14 was cocultured with Student’s t-test. P = 0.049 (mutant is slightly higher). Red lines represent Arthrobacter CL28 versus Variovorax CL14 monoculture. Only Variovorax- median. Article

a Rhizoplane 0

Prevalence Genus ) Low Medium -5 Aquaspirillum BIrii41_ge Bradyrhizobium elative abundance Burkholderia Galbitalea Massilia verage r -10 A

( Polaromonas 2 Rhizobacter log Variovorax

-15 0.0 0.2 0.4 0.6 0.81.0 Prevalence across 17 sampled sites

b Rhizosphere ) -8 Prevalence Low Medium

elative abundance High Core -10 Variovorax core verage r A ( 2 g lo

-12

0.0 0.2 0.4 0.6 0.81.0 Prevalence across 30 plant species

Extended Data Fig. 10 | Variovorax are highly prevalent across naturally samples taken across 3 years in 17 sites in Europe. b, Correlation plot of data occurring Arabidopsis microbiomes and across 30 plant species. reanalysed from ref. 4, comparing bacterial ASV prevalence to log-transformed a, Correlation plot of data reanalysed from ref. 6, comparing bacterial ASV relative abundance across 30 phylogenetically diverse plant species grown in a prevalence to log-transformed relative abundance in A. thaliana rhizoplane common garden experiment.      

()00123)4564789@A)0B2CD n1VV10Gq847U    E82@9358@15FG89@A)0B2CD  n94opaqrqr  

H13)0@647I9PP80G    

Q8@901H12180RAS62A12@)6P30)T1@A10130)59R6F6U6@G)V@A1S)0W@A8@S139FU62AXYA62V)0P30)T65122@09R@901V)0R)4262@14RG845@084238014RG   

64013)0@647X`)0V90@A1064V)0P8@6)4)4Q8@901H12180RA3)U6R612a211b9@A)02cH1V10112845@A1d56@)068Ue)U6RG(A1RWU62@X     

I@8@62@6R2   `)08UU2@8@62@6R8U848UG212aR)4V60P@A8@@A1V)UU)S6476@1P2801301214@64@A1V67901U17145a@8FU1U17145aP864@1f@a)0g1@A)5221R@6)4X

4h8 ()4V60P15

s YA11f8R@28P3U126i1BpCV)018RA1f3106P14@8U70)93hR)456@6)4a76T14828562R01@149PF10845946@)VP182901P14@

s b2@8@1P14@)4SA1@A10P182901P14@2S101@8W14V0)P562@64R@28P3U12)0SA1@A10@A128P128P3U1S82P182901501318@15UG YA12@8@62@6R8U@12@B2C9215bQqSA1@A10@A1G801)41r)0@S)r26515 s sptuvwxyyxpv€‚€‚v‚ƒx„t v†v ‚w‡ˆ† v‚xttuv†uvp‰yv ‚w‡ˆ†vyx‡vwxy‘t’v€wƒpˆ“„‚vˆpv€ƒv”€ƒx ‚v‚w€ˆxp•

s b512R063@6)4)V8UUR)T8068@12@12@15

s b512R063@6)4)V84G8229P3@6)42)0R)001R@6)42a29RA82@12@2)V4)0P8U6@G84585–92@P14@V)0P9U@63U1R)P38062)42 bV9UU512R063@6)4)V@A12@8@62@6R8U3808P1@10264RU95647R14@08U@14514RGB1X7XP1842C)0)@A10F826R12@6P8@12B1X7X01701226)4R)1VV6R614@C s bQqT8068@6)4B1X7X2@84580551T68@6)4C)0822)R68@1512@6P8@12)V94R10@864@GB1X7XR)4V6514R164@10T8U2C

`)049UUAG3)@A1262@12@647a@A1@12@2@8@62@6RB1X7X—a€a‡CS6@AR)4V6514R164@10T8U2a1VV1R@26i12a5170112)VV0115)P845˜T8U914)@15 s ™ˆdv˜vd‰t„‚v‰‚v’‰w€vd‰t„‚veƒpd‡v‚„ˆ€‰†t•

s `)0f8G12684848UG262a64V)0P8@6)4)4@A1RA)6R1)V306)02845g80W)TRA864g)4@1(80U)21@@6472

s `)0A61080RA6R8U845R)P3U1f5126742a6514@6V6R8@6)4)V@A18330)3068@1U1T1UV)0@12@2845V9UU013)0@647)V)9@R)P12

s d2@6P8@12)V1VV1R@26i12B1X7X()A14g2 ae1802)4g2‡Ca6456R8@647A)S@A1GS101R8UR9U8@15

s„‡ve†vwxttw€ˆxpvxpv‚€‰€ˆ‚€ˆw‚vhx‡v†ˆxtxiˆ‚€‚vwxp€‰ˆp‚v‰‡€ˆwt‚vxpvy‰puvxhv€ƒv‘xˆp€‚v‰†xd• I)V@S801845R)51 e)U6RG64V)0P8@6)48F)9@8T86U8F6U6@G)VR)P39@10R)51 q8@8R)UU1R@6)4 j64HA6i)TXqroqF845tP871nTXoXuq3S1019215V)03U84@P)03A)P1@06RP182901P14@2X`)0F6)64V)0P8@6R2848UG262S19215DgYr Y))UF)fTvXoXqaI6RWU1ToXppawIdbH(lTxXoryraHqeRU8226V610TqXooa`82@z(TrXooXxaY06PP)P8@6RTrXp{altIbYqTqXoXrag9U@6z(ToXoa APP10TpXoFqagb``YTxXvrxa@06PbtToXq01Tuya`82@Y011TqXoXoraF)S@61qTqXvXoa)0@A)V64510TqXqXoX `)02@8@62@6R8U848UG262S19215HTpX{XqS6@A@A1V)UU)S64738RW8712D)ARA6F6TrXrXrXyaT1784TqXuruaqdI1mqToXqvXra1PP1842ToXvXxa I9F0185ToX{XpaH29F0185TqXqXqX

q8@8848UG262 bUU2)V@S801845R)519215V)058@8848UG262801U62@1564@A1g1@A)5221R@6)48458018T86U8FU18@76@A9FXR)Ph628627hT806)T)08fHkt

`)0P8492R063@29@6U6i647R92@)P8U7)06@AP2)02)V@S801@A8@801R14@08U@)@A1012180RAF9@4)@G1@512R06F156439FU62A15U6@108@901a2)V@S801P92@F1P8518T86U8FU1@)156@)02h01T61S102X j12@0)47UG14R)90871R)51513)26@6)4648R)PP946@G013)26@)0GB1X7Xk6@l9FCXI11@A1Q8@901H12180RA79651U6412V)029FP6@@647R)51c2)V@S801V)0V90@A1064V)0P8@6)4X q8@8 

e)U6RG64V)0P8@6)48F)9@8T86U8F6U6@G)V58@8  

!

bUUP8492R063@2P92@64RU951858@88T86U8F6U6@G2@8@1P14@XYA622@8@1P14@2A)9U530)T651@A1V)UU)S64764V)0P8@6)4aSA101833U6R8FU1D " # $

rbRR1226)4R)512a946m916514@6V6102a)0S1FU64W2V)039FU6RUG8T86U8FU158@821@2 % &

rbU62@)VV679012@A8@A8T1822)R68@1508S58@8 ' rb512R063@6)4)V84G012@06R@6)42)458@88T86U8F6U6@G

YA1o{I0HQb8P3U6R)421m914R64758@8822)R68@15S6@A@A622@95GS82513)26@1564@A1Q(ftIHb80RA6T194510@A130)–1R@8RR1226)4eHnQbuvppopXYA108S @0842R063@)P6R58@8S82513)26@1564@A1k141df301226)4|P46F92Bkd|C94510@A18RR1226)4kIdopoou}XI)90R158@8V)08UUV6790126230)T6515822933U1P14@80G 64V)0P8@6)4@)@A6280@6RU1Xt8556@6)4aS1513)26@158UU2R063@28458556@6)48U58@82@09R@901201m96015@)0130)59R1@A10129U@2)V@A622@95G64@A1V)UU)S647k6@l9F

013)26@)0GDA@@32Dhh76@A9FXR)Ph628627hT806)T)08fHkt      

`61U5r231R6V6R013)0@647    

eU182121U1R@@A1)41F1U)S@A8@62@A1F12@V6@V)0G)90012180RAXtVG)98014)@2901a0185@A18330)3068@121R@6)42F1V)01P8W647G)9021U1R@6)4X  

s E6V12R614R12 f1A8T6)908Uc2)R68U2R614R12 dR)U)76R8Ua1T)U9@6)480Gc14T60)4P14@8U2R614R12  `)0801V1014R1R)3G)V@A15)R9P14@S6@A8UU21R@6)42a21148@901XR)Ph5)R9P14@2h40r013)0@647r29PP80GrVU8@X35V         E6V12R614R122@95G512674   bUU2@95612P92@562RU)21)4@A1213)64@21T14SA14@A1562RU)2901624178@6T1X  I8P3U126i1 bUU28P3U126i12801013)0@1564V67901U171452ag1@A)528452)90R158@8XQ)28P3U1r26i1R8UR9U8@6)4S82310V)0P15Xt4301T6)92P8492R063@2  B(82@06UU)1@8Uqroxa`64W1U1@8Uaqroyal100108e801512aqro}CS1)F210T15@A8@844u30)T651229VV6R614@2@8@62@6R8U3)S10V)0IG4()P  848UG212845S1921828P3U126i1)V4{A10164X`)0P)03A)P1@06R848UG262S1S1018FU1@)9218UU3U84@2@A8@S101U8@103))U15V)0)@A10 3903)212a2))9028P3U149PF10S828@P646P9P4qraSA6RAS8251P)42@08FUG29VV6R614@V0)PV0)P@A158@864`67901oFX`)0HQbr21m 1f3106P14@2aS1)F210T1564301T6)92P8492R063@2B(82@06UU)1@8UaqroxC@A8@828P3U126i1)V4p6451314514@28P3U1218RAR)4@8646478 3))U)Vv213808@13U84@230)T651529VV6R614@3)S10@))F210T1P8–)0@0842R063@)P6R2A6V@2Xl101aS1)4UGR)42651015@A1)T10U8336477141A6@2 V0)P@S)213808@11f3106P14@2S6@A4p8454u013U6R8@12a18RAR)4@864647v3))U153U84@2X`)03U84@r822)R68@15(`wR)94@2a301T6)92 2@956122A)S@A8@4{30)T651229VV6R614@3)S10B(A141@8UaQ8@901aqrqrCXl101S19214€oq@)m984@6VGb0@A@)F8R@10(Eq}8F94584R1 Bdf@14515q8@8`67X{CX`679012df@14515q8@8`67Xy5a180164@14515@))4UG51P)42@08@1@A8@(`w2)V@A101U1T84@2@08642R84F101R)T1015 V0)P@A10))@a845@A101V)01S192154p8454uF6)U)76R8U28P3U12a18RA28P3U162848T10871)Vv@1RA46R8U013U6R8@1284518RA@1RA46R8U 013U6R8@16283))U)Vv6451314514@F6)U)76R8U28P3U12X YA126i1)V)@A101f3106P14@2a64RU95647@A12)6UT8U658@6)41f3106P14@284564rT6@0)F8R@1068U1f3106P14@2S8212@8FU62A15F8215)4@A1 P8f6P8U8P)94@)V013U6R8@6)4@A8@R)9U5F10182)48FUG)F@86415R)4265106473AG26R8U238R1845U8F)0R)42@0864@2X

q8@81fRU926)42 b2P8UU49PF10)V3U84@28P3U12S1011fRU9515V0)P848UG262XYA12128P3U12S1011f@01P1)9@U6102SA101@A12115U64770)S@AS822@94@15 845A85RU180UG4)@51T1U)3150)21@@1U18T12F1G)45@A1R)@GU15)42XYA121R06@1068S101301r12@8FU62A15845P864@86415@A0)97A)9@X`)0o{I 848UG26228P3U12S1011fRU9515F8215)4m98U6@GR)4@0)UR06@1068)9@U641564@A1g1@A)5221R@6)4X

H13U6R8@6)4 bUU013)0@1528P3U126i1201301214@F6)U)76R8UUG6451314514@28P3U12Xt43U84@81f3106P14@264@A138310S101013U6R8@15@S6R1G61U5647 26P6U800129U@2XYA1013U6R8@1629215828@10P648UUP)51U2Xt4@A1R821)V@A1HQbr21m58@8a)4UG829F21@)VR)UU1R@1528P3U12S82 30)R12215a8450130)59R6F6U6@GS821429015FGR)P380647@A126746V6R84@A6@2V0)P@S)4)4r6514@6R8U1f3106P14@2Xt4P9U@63U1R8212aS1 013U6R8@151f3106P14@2823)26@6T1R)4@0)U26429F21m914@1f3106P14@2X df3106P14@8U013U6R8@6)4FGV67901D`67o8D@S)6451314514@1f3106P14@2BS6@A6418RA8F6)@6R7085614@C`67oFD@S)6451314514@1f3106P14@2 `67Xo5D@S)6451314514@1f3106P14@2`67Xo1D@S)6451314514@1f3106P14@2`67XqRD|411f3106P14@2a01318@156451314514@UG82@A1 R)4@0)U64q5a1845VX`67q5a1aVD@S)6451314514@1f3106P14@2X`67q7Dq8@830)R12215V0)P`67XqRX|411f3106P14@X`67XqADq8@8 877U)P108@15V0)Po{6451314514@1f3106P14@2X`67Xp8D@S)6451314514@1f3106P14@2X`67XpFD@S)6451314514@1f3106P14@2X`67XvFD)41 1f3106P14@aT8U658@647dq`67X}8X`67XvRa5D@S)6451314514@1f3106P14@2Xdq`67oRDYA62628R)P38062)4)V@S)6451314514@1f3106P14@2a 310V)0P15)4R1Xdq`67o5D877U)P108@6)4)V}6451314514@1f3106P14@2Xdq`67Xo1aVD@S)6451314514@1f3106P14@2Xdq`67q8aFD@S) 6451314514@1f3106P14@2Xdq`67Xp8D@S)6451314514@1f3106P14@2Xdq`67XpFDYS)6451314514@1f3106P14@2Xdq`67Xp5DYS)6451314514@ 1f3106P14@2Xdq`67XuFD@S)6451314514@1f3106P14@2Xdq`67X{D|411f3106P14@Xdq`67XxD|T10U833647A6@2V0)P@S)56VV1014@1f3106P14@2a 18RA310V)0P15)4R1Xdq`67X}8DYS)6451314514@1f3106P14@2dq`67X}FDYS)6451314514@1f3106P14@2aT8U658@1564`67XvFdq`67X}Ra5D YS)6451314514@1f3106P14@2Xdq`67Xy5a1D|411f3106P14@18RABS6@A26P6U80R)4@0)U2V)0T8U658@6)4CX

H845)P6i8@6)4 eU84@153U8@128453)@2S1013U8R1564@A170)S@ARA8PF108@80845)P)0510B841f3106P14@10FU645UG36RW15933)@2h3U8@128@0845)P845 3U8R15@A1P64@A108RW2CX`)0qQbU6F080GR)42@09R@6)4a28P3U12S1010845)P6i159264783AG26R8UP1@A)5F1V)01822674P14@@)3U8@12X

fU645647 YA101S822)P1FU64564764@A128P3U1A80T12@647rP9U@63U11f3106P14@102380@6R638@1564@A1A80T12@a845@A15126748@6)4)V18RA3U8@1S82 W4)S4)4UG@))41)V@A1PXbV@10A80T12@647@A128P3U12S101FU64515r@A1GS10182267415849PF10845R0)2201V1014R647S6@AP1@858@8 S82310V)0P158@58@8848UG262X

H13)0@647V)0231R6V6RP8@1068U2a2G2@1P2845P1@A)52 j101m960164V)0P8@6)4V0)P89@A)028F)9@2)P1@G312)VP8@1068U2a1f3106P14@8U2G2@1P2845P1@A)52921564P84G2@95612Xl101a6456R8@1SA1@A1018RAP8@1068Ua 2G2@1P)0P1@A)5U62@156201U1T84@@)G)902@95GXtVG)98014)@29016V8U62@6@1P833U612@)G)90012180RAa0185@A18330)3068@121R@6)4F1V)0121U1R@64780123)421X   

! " # $ % & '

~ g8@1068U2c1f3106P14@8U2G2@1P2 g1@A)52   

4h8 t4T)UT1564@A12@95G 4h8 t4T)UT1564@A12@95G   s b4@6F)5612 s (Ater21m   

s d9W80G)@6RR1UUU6412 s `U)SRG@)P1@0G   

s e8U81)4@)U)7G s gHtrF82154190)6P87647 

s b46P8U2845)@A10)078462P2  

s l9P84012180RA380@6R6384@2  

s (U646R8U58@8              

! " # $ % & '

‚