A microbiota–root–shoot circuit favours Arabidopsis growth over defence under suboptimal light

Bidirectional root–shoot signalling is probably key in orchestrating stress responses and ensuring plant survival. Here, we show that Arabidopsis thaliana responses to microbial root commensals and light are interconnected along a microbiota–root–shoot axis. Microbiota and light manipulation experiments in a gnotobiotic plant system reveal that low photosynthetically active radiation perceived by leaves induces long-distance modulation of root bacterial communities but not fungal or oomycete communities. Reciprocally, microbial commensals alleviate plant growth deficiency under low photosynthetically active radiation. This growth rescue was associated with reduced microbiota-induced aboveground defence responses and altered resistance to foliar pathogens compared with the control light condition. Inspection of a set of A. thaliana mutants reveals that this microbiota- and light-dependent growth–defence trade-off is directly explained by belowground bacterial community composition and requires the host transcriptional regulator MYC2. Our work indicates that aboveground stress responses in plants can be modulated by signals from microbial root commensals.

U nlike animals, plants are sessile organisms that must simultaneously integrate various responses to biotic and abiotic stresses to prioritize either growth or defence depending on rapidly changing surrounding conditions 1 . Since the colonization of land by ancestral plant lineages 450 million years ago, complex multikingdom microbial consortia have interacted with and colonized the roots of healthy plants 2 . Direct integration of microbial and environmental cues by plants is therefore probably key in orchestrating their ability to overcome environmental stresses [3][4][5] . The beneficial activities conferred by microbial root commensals have been primarily described in the context of nutritional stress 4,6-9 or pathogen attack [10][11][12][13][14] , revealing that root colonization by complex microbial communities can shape host phenotypic plasticity and protect plants from diseases. By analogy with the microbiota-gut-brain axis described in animals, the concept of the microbiota-root-shoot axis was recently introduced in plants 15 ,16 , suggesting that bidirectional root-shoot signalling might represent a key mechanism linking root microbiota assembly to aboveground stress responses.
Carbohydrate biosynthesis, energy production and therefore plant growth depend on the amount and quality of photosynthetically active radiation (PAR) absorbed by chloroplasts in leaves 17 . Spectral PAR irradiance ranges from 400 to 700 nm under natural conditions, and a reduction of both the amount of PAR (light quantity) and the ratio of red to far-red light (light quality) is perceived as a warning signal in plants, thereby triggering adaptive changes in shoot morphology [18][19][20][21] . Light-dependent resource allocation has been shown to be tightly linked with defence signalling in Arabidopsis thaliana and tomato leaves, suggesting that aboveground responses to light and pathogens are interconnected [22][23][24][25][26] . However, the extent to which belowground commensals can modulate or even dictate aboveground stress responses remains unknown. Given the fact that a substantial proportion of photosynthetically fixed carbohydrates (that is, >15%) is invested in the rhizosphere 27,28 , we hypothesized that the aboveground response to light and the belowground response to microorganisms are interconnected along the microbiota-root-shoot axis.
Here, we show that microbial root commensals can alleviate A. thaliana growth deficiency under suboptimal light, and we observe a direct link between bacterial community composition and the prioritization of microbiota-induced growth over defence responses in shoots. We report that this microbiota-and light-dependent growth-defence trade-off requires the host transcriptional regulator MYC2. Our results imply that belowground responses to microorganisms and aboveground responses to light are integrated along a microbiota-root-shoot circuit to boost plant growth when the light condition is suboptimal.

Results
A synthetic root microbiota rescued Arabidopsis growth under low PAR. We hypothesized that a microbiota-root-shoot axis exists in plants, allowing them to take advantage of belowground microbial commensals to orchestrate aboveground stress responses. First, we assembled a synthetic microbial consortium (SynCom; Extended Data Fig. 1a and Supplementary Table 1) with phylogenetically diverse bacteria (B, 183 strains), fungi (F, 24 strains) and oomycetes (O, 7 strains) that naturally colonize the roots of healthy A. thaliana 12,29 . Eight versions of this BFO SynCom containing B, F and O communities mixed at different relative abundances (RAs) in the starting inoculum were used to recolonize the roots of germ-free A. thaliana in the gnotobiotic FlowPot system 30 (2) by exposing shoots to far-red light (740 nm) for 15 minutes at the end of the day (EODFR (ref. 31 ), reduction in light quality). Hypocotyl length (Extended Data Fig. 2d,e) and relative petiole length (petiole length divided by corresponding leaf length; LP: 1.9-fold, P < 2.2 × 10 −16 ; EODFR: 2.8-fold, P < 2.2 × 10 −16 ; Mann-Whitney U-test; Fig. 1a) were increased under suboptimal light conditions, consistent with stereotypical shade phenotypes previously described [32][33][34] . Compared with five-week-old plants grown under NC, those exposed to LP and EODFR showed a significant reduction in canopy size (LP: 2.4-fold, P = 5.45 × 10 −13 ; EODFR: 2.1-fold, P = 2.88 × 10 −4 ; Mann-Whitney U-test; Fig. 1b,c and Extended Data Fig. 2f) and shoot fresh weight (LP: 2.2-fold, P = 2.32 × 10 −8 ; EODFR: 2.2-fold, P = 4.11 × 10 −5 ; Mann-Whitney U-test; Fig. 1d) in the absence of the BFO SynCom. Remarkably, the presence of BFO rescued LP-and EODFR-mediated decreases in canopy size and shoot fresh weight to control levels ( Fig. 1b-d). Furthermore, recolonization experiments with B in the presence or absence of F and O communities indicated that B commensals alone partially rescued plant growth under LP, whereas the synergistic contribution of B and filamentous eukaryotes was needed for the full rescue (Kruskal-Wallis with Dunn's post hoc test, P < 0.05; Fig. 1e). Experiments with a natural soil (Cologne agricultural soil (CAS)) and with peat repopulated with a microbial wash from CAS validated the robustness of the plant growth rescue under LP, irrespective of microbial diversity and soil conditions (Extended Data Fig. 3a and Supplementary Table 1). The presence of BFO also led to a significant increase in leaf number (Fig. 1f) and the leaf length/width ratio (a proxy of leaf shape; Fig. 1g) (Kruskal-Wallis with Dunn's post hoc test, P < 0.05). Inspection of the magnitude of BFO-mediated modification of leaf traits indicated quantitative differences among the tested parameters and revealed that the amplitude of the BFO effect was always greater under LP and   Fig. 3d,e), and the aforementioned taxonomic changes between LP and NC were quantitatively similar in the context of the CAS wash but not when the natural soil was used (Extended Data Fig. 3c,f,g). Given the possible ectopic leaf colonization by root microbiota members in our gnotobiotic system 35 , we also tested whether BFO root commensals can be detected in leaves and the extent to which their RAs can be modulated by light conditions. Diversity analysis (Extended Data Fig. 4a), PCoA (Extended Data Fig. 4b) and enrichment tests (Extended Data Fig. 4c,d) revealed that 44%, 15% and 0% of B, F and O strain variants detected belowground were colonizing aboveground leaf tissues, and that the composition of ectopic B assemblages was also modulated by light conditions, with 13 strain variants showing light-dependent differential growth (Extended Data Fig. 4d). Sequencing of blank controls from DNA extractions and control samples from germ-free plants confirmed the sterility of our system and validated our sequencing-based detection method (Extended Data Fig. 4e). Our results indicate that shoot exposure to suboptimal light triggers host-dependent modulation of the growth of root-associated and ectopic leaf-associated bacterial taxa in a complex multikingdom synthetic microbiome.
Host prioritizes microbiota-induced growth over defence under LP. We hypothesized that plant responses to BFO commensals and light are interconnected, thereby orchestrating resource investment into shoots under LP. We profiled the root and shoot transcriptomes of BFO-colonized and germ-free A. thaliana exposed to LP and NC in the gnotobiotic FlowPot system five weeks after BFO inoculation (Supplementary Table 3). PERMANOVA (Supplementary Table 3) and pairwise correlations among samples (Extended Data Fig. 5a) indicated that the presence/absence of the BFO SynCom explained transcriptome differentiation in root samples more than the light condition (microorganisms: R 2 = 0.361, P < 0.001; light:  Fig. 3a) and eight in shoot samples (S1-S8, n = 2,790 genes; Fig. 3b). The presence of the BFO SynCom triggered light-independent upregulation of genes involved in ion homeostasis (root, R2) and cell differentiation (root, R2), as well as downregulation of genes involved in the response to salicylic acid (SA) (root, R9) and anthocyanin biosynthesis (shoot, S6) ( Fig. 3a Table 3). In contrast, LP triggered BFO-independent upregulation of genes involved in the response to jasmonic acid (JA) and gibberellin (GA) (shoot, S5) as well as downregulation of genes involved in the flavonoid metabolic process (shoot, S3) and response to high light intensity (root, R4) ( Fig. 3a,b, Extended Data Fig. 5c,d and Supplementary Table 3). Furthermore, we identified clusters for which gene expression was modulated by both light and BFO conditions (R1, R3, R7, R8, S1 and S4), probably explaining BFO-mediated host rescue in LP (Fig. 3a,b). Particularly, genes belonging to clusters R1, R3 and S1 were upregulated in the presence of the BFO SynCom under NC but not under LP (Fig. 3a,b). A substantial fraction of the genes (227) did overlap between root clusters R1 and R3 and the shoot cluster S1 (Fig. 3c), and Gene Ontology (GO) term analysis revealed consistent enrichments of immune-related processes, including the response to chitin; the response to SA, JA and ethylene; the response to organonitrogen compounds; and the regulation of immune responses, among others (hypergeometric test with Bonferroni correction, P < 0.05; Fig. 3d, Extended Data Fig. 5c,d and Supplementary Table 3). Transcripts with conserved expression patterns in roots and shoots primarily encode transcription factors (WRKY33, WRKY40, MYB15, MYB51, ANAC042 and ANAC055), receptor-like protein/kinase (WAKL10, RLK3 and CRK9), ethylene-responsive elements (ERF6 and ERF11) or calmodulin binding proteins (CBP60G and CML38) (Fig. 3e). Immune response activation in response to the BFO SynCom under NC was more extensive aboveground than belowground and involved several well-known immune-related genes that act through multiple pathways (Fig. 3e) Fig. 3f) or colony counting (Pst, Fig. 3g) revealed major influences of both light and SynCom conditions on disease resistance five days after pathogen inoculation (ANOVA followed by post hoc Tukey's HSD, P < 0.05; Fig. 3f,g). Plants colonized by the BFO SynCom under NC were the most resistant to both leaf pathogens, which is consistent with extensive BFO-triggered immune responses observed in the RNA-seq data and the presence of putatively protective commensals that ectopically colonize aboveground shoot organs 36 (Fig. 3b,  Only samples with more than 1,000 reads were considered. the box plot boundaries reflect the interquartile range, the centre line is the median and the whiskers represent 1.5× the interquartile range from the lower and upper quartiles. b, Variation in B, F and O community composition explained by 'light' (NC versus LP) in root and matrix samples five weeks after BFO inoculation in the FlowPot system. CAP was used to quantify the percentage of variance explained by 'light' for each condition. ***P < 0.001; *P < 0.05; NS, not significant (CAP, ANOVA-like permutation test). c, Between-group variation in microbial community composition assessed using PLS-DA. Only samples with more than 1,000 reads were considered. the percentage of variance explained by the grouping factor is indicated along the x and y axes (B, n = 47 samples; F, n = 48; O, n = 26). Comp., component. d, Normalized RAs of prevalent B strain variants detected in roots across NC and LP. Sample-wise normalized RA was calculated for each strain variant and is depicted in the heat map next to the taxonomic tree. Only strain variants with an average RA ≥ 0.1% across samples were considered (n = 24 samples). the taxonomic tree was constructed on the basis of a maximum likelihood method using full-length 16S rRNA sequences. e, Sample-wise log fold change in RA measured for each prevalent strain variant between LP and NC in root and matrix samples. Differential RA with statistically significant P values is shown (edgeR, generalized linear model, P < 0.05).
to both pathogens and failed to mount effective immune responses.
Although the amplitude of BFO-induced leaf pathogen protection (versus germ-free control) was quantitively similar in NC and LP, plants were more susceptible when the light condition was suboptimal (ANOVA followed by post hoc Tukey's HSD, P < 0.05; Fig. 3f    the gene set was split into nine gene expression clusters in roots (R1 to R9) and eight gene expression clusters in shoots (S1 to S8). there were three independent biological replicates (n = 24 samples). c, Number of genes shared between root clusters R1 and R3 and shoot cluster S1 or specific to each of the two groups. d, GO term enrichment network depicting the top 12 most significantly enriched GO terms (hypergeometric test with Bonferroni correction, P < 0.05) detected in clusters R1, R3 and S1. Each GO term is represented as a circle, and the contribution of each cluster to the overall GO term enrichment is shown. the size of the GO term reflects the number of genes enriched in the GO term. R, response; reg., regulation; Et, ethylene. e, Gene-concept network depicting linkages between genes and the associated top 12 most significantly enriched GO terms detected in clusters R1, R3 and S1. Each node represents a gene and is colour-coded according to the different clusters. f, qPCR-based quantification of Bc growth in A. thaliana leaves five days after pathogen droplet inoculation in the FlowPot system. there were four independent biological replicates (n = 48 plants). g, Colony-count-based quantification of Pst growth in A. thaliana leaves five days after pathogen spray inoculation in the FlowPot system. there were three independent biological replicates (n = 36 plants). In f and g, the box plot boundaries reflect the interquartile range, the centre line is the median and the whiskers represent 1.5× the interquartile range from the lower and upper quartiles. the letters indicate statistical significance (one-way ANOVA followed by post hoc tukey's HSD, α = 0.05). cpm, counts per million; c.f.u., colony-forming units.
we observed mutant versus Col-0 wild-type (WT) plant variation in vegetative growth under control conditions (NC − BFO), with enhanced growth for dellaP and cry1 cry2 mutants and reduced growth for jazQ and bri1-301 mutants (Extended Data Fig. 6a-j,o; Mann-Whitney U-test, P < 0.05; statistical analysis not depicted). Irrespective of these differences in growth rate, LP-mediated reduction in canopy size was quantitatively similar across all mutants tested in the absence of the BFO SynCom (0.48-fold to 0.71-fold decrease in canopy size), except for the sav3-3 mutant, which retained high growth under LP (Kruskal-Wallis with Dunn's post hoc test; Fig. 4a). The similarity in Col-0 canopy size observed between LP and NC in the presence of BFO (+BFO; ratio of LP to NC close to 1; Fig. 4a) was largely retained in the sid2-2, del-laP and sav3-3 mutants, indicating that the corresponding genes are dispensable for BFO-mediated canopy size enlargement under LP ( Fig. 4a and Extended Data Fig. 6a,b). In contrast, canopy size was dramatically reduced under LP compared with NC for dde2-2, myc2-3, jazQ, bri1-301 and cry1 cry2 mutants (+BFO; ratio of LP to NC close to 0.5; Mann-Whitney U-test, P < 0.01; Fig. 4a and Extended Data Fig. 6a,b), indicating that JA biosynthesis/signalling, brassinosteroid signal transduction and cryptochromes are needed for BFO-induced growth under LP. Integration of the belowground response to microbiota and the aboveground response to light therefore involved multiple points of control along the root-shoot axis. In  Fig. 4c), validating that LP-mediated reduction in canopy size in the absence of root commensals was also associated with impaired resistance to unrelated leaf pathogens (Fig. 4b,c). In contrast, most of the mutants impaired in BFO-mediated growth rescue in LP remained as resistant to Bc and Pst (myc2-3, jazQ and bri1-301) or Pst only (cry1 cry2) as under NC (Mann-Whitney U-test, P < 0.05). Our results underline the potential roles of MYC2 and BRI1 as regulatory nodes balancing investment into microbiota-induced growth at the expense of defence under LP.
A link between differential canopy size and root microbiota composition under LP. We asked whether differentiation in B community composition in the roots of WT plants was altered in the different mutants and could correlate with aboveground change in canopy size. We took advantage of the previous experiment in which the different mutants were grown in the gnotobiotic system under NC and LP to simultaneously monitor BFO assemblages in roots and the surrounding peat matrix five weeks after BFO inoculation. PERMANOVA confirmed the effect of the light condition on B community composition in the roots but not in the matrix samples (root, 'light': R 2 = 0.014, P = 0.024) and revealed that B community differentiation was more extensively shaped by 'genotype' than by 'light' (root, 'genotype': R 2 = 0.117, P < 0.001), which was validated by CAP (Supplementary Table 5 and Extended Data Fig.  7a). CAP analysis constrained by 'light' for each genotype indicated that light-mediated differentiation in B community composition was greater for jazQ (12.8%, P = 0.004), dde2-2 (8.07%, P = 0.0265), myc2-3 (7.87%, P = 0.005) and bri1-301 (6.26%, P = 0.049) than for the WT (3.35%, P = 0.044) and was not significant for the other mutants (Extended Data Fig. 7b). Pairwise-enrichment tests conducted between LP and NC conditions for each genotype (edgeR, generalized linear model, P < 0.05; Fig. 5a) validated the increase in RA of root-associated Pseudomonadales and Actinomycetales observed in the WT, and revealed mutant-specific differences in abundance profiles at the strain (Fig. 5a) and class (Extended Data Fig. 7c) levels. To further test for potential associations between aboveground canopy phenotypes and belowground B community composition, we calculated canopy size variation for each mutant between LP − BFO and LP + BFO conditions and asked whether these quantitative differences were significantly linked to corresponding root microbiota composition under LP. Using a linear regression model, we observed a significant link between BFO-induced canopy growth across mutants and B community composition in roots (coordinate on PCoA axis 1, 46.9% of the variance in Bray-Curtis dissimilarities), which explained 47.7% of the variation in B community composition along PCoA1 (R 2 = 0.4776,  . the ratio of canopy sizes (five-week-old plants) was computed between LP and NC across three independent biological replicates (n = 1,188, −BFO and +BFO samples). Statistical significance across genotypes is indicated with letters (Kruskal-Wallis with Dunn's post hoc test, α = 0.05). Statistical significance between LP and NC is indicated with asterisks (Mann-Whitney U-test, ***P < 0.001, **P < 0.01, *P < 0.05). b, Differential Bc growth between LP and NC in leaves of A. thaliana Col-0 and eight mutants grown in the FlowPot system (left, −BFO; right, +BFO). Bc growth was quantified by qPCR in A. thaliana leaves five days after pathogen droplet inoculation. the ratio of Bc loads was computed between LP and NC across three independent biological replicates for each genotype (n = 336, −BFO and +BFO samples). c, Differential Pst growth between LP and NC in leaves of A. thaliana Col-0 and eight mutants grown in the FlowPot system (left, −BFO; right, +BFO). Pst growth was quantified by colony counting in A. thaliana leaves five days after pathogen spray inoculation. the ratio of Pst loads was computed between LP and NC across three independent biological replicates for each genotype (n = 324, −BFO and +BFO samples). In b and c, statistical significance across genotypes is indicated with letters (one-way ANOVA followed by post hoc tukey's HSD, α = 0.05). Statistical significance between LP and NC is indicated with asterisks (Mann-Whitney U-test, ***P < 0.001, **P < 0.01, *P < 0.05). In a-c, the box plot boundaries reflect the interquartile range, the centre line is the median and the whiskers represent 1.5× the interquartile range from the lower and upper quartiles.  2.0 P < 2.2 × 10 -16 P = 5.9 × 10 -14 P = 5.5 × 10 -10 P = 5.9 × 10 -9 P = 1.7 × 10 -6 P = 1.3 × 10 -5 P = 3.1 × 10 -8 P = 3.8 × 10 -13 P = 0.007 P = 2.7 × 10 -6 P = 1.8 × 10 -10 P = 1.4 × 10 -5 P = 1.3 × 10 -9 P = 4.2 × 10 -10 B community composition and that host modulation of belowground community assembly influences investment into growth over defence under LP.
Priority to microbiota-induced growth over defence under LP requires MYC2. Given the fact that MYC2 is a central node controlling the cross-talk between JA and other phytohormone signalling pathways (that is, GA, SA, abscisic acid and indole-3-acetic acid) and regulating responses to light and the circadian clock 49 , we hypothesized that this transcription factor might coordinate the prioritization of microbiota-induced growth over defence under suboptimal light conditions. Importantly, impaired growth and increased resistance to Bc observed in the myc2-3 mutant were quantitatively similar in the independent myc234 triple mutant (Extended Data Fig. 6o) and reverted in the jin1-8 pMYC2::MYC2-FLAG line, in which a MYC2-FLAG fusion protein is expressed under control of the native MYC2 promoter (MYC2-FLAG, myc2 background) 50 ( Fig. 6a and Extended Data Fig. 8a). We also demonstrated that the alleviation of the EODFR-mediated decrease in canopy size by BFO also required MYC2, illustrating that MYC2-dependent rescue of plant growth by BFO is robust across multiple light-limiting conditions, including when light quantity and quality are altered (Fig. 6a). Western blot assays revealed that MYC2 protein abundance in root and shoot samples was modulated by BFO under LP but not under NC (MYC2-FLAG line), with expression levels in the shoot matching canopy size phenotypes ( Fig. 6b and Extended Data Fig. 8b). MYC2 protein abundance in roots showed an opposite trend under LP compared with that observed in shoots ( Fig. 6b and Extended Data Fig. 8b). We profiled the root and shoot transcriptomes of BFO-colonized and germ-free myc2-3 and MYC2-FLAG lines exposed to LP and NC in the gnotobiotic FlowPot system five weeks after BFO inoculation. Inspection of 571 shared MYC2-regulated genes in shoots and roots under LP validated that MYC2 drives opposite responses between shoots and roots (Extended Data Fig.  8c), which explained the opposite trend of MYC2 protein abundance in shoots and roots ( Fig. 6b and Extended Data Fig. 8b). Remarkably, presence/absence of the BFO SynCom explained transcriptome differentiation in shoot samples more than light in the myc2-3 mutant (Extended Data Fig. 8d), which was validated by PERMANOVA (myc2-3, microorganisms: R 2 = 0.394, P < 0.001; light: R 2 = 0.179, P = 0.006; MYC2-FLAG, microorganisms: R 2 = 0.085, P = 0.078; light: R 2 = 0.439, P < 0.001; Supplementary Table 6). These results suggest that mutation of MYC2 resulted either in the attenuation of shoot response to light or in the exacerbation of shoot response to microorganisms. Mutation of MYC2 led to dramatic shifts in the root and shoot transcriptome, with 5,231 and 5,038 genes differentially regulated in myc2-3 compared with MYC2-FLAG in roots and shoots, respectively (referred to as MYC2-differentially expressed or MDEs, |log 2 FC| ≥ 1, empirical Bayes statistics, FDR < 0.05; Fig. 6c and Extended Data Fig. 8e-g). A large fraction of these MDEs were previously reported as direct targets of MYC2 (refs. [51][52][53] ) (MDEs root, 52%; MDEs shoot, 58%; Fig. 6d and Extended Data Fig. 9a). Hierarchical clustering of the gene expression profiles of MDEs revealed eight gene expression clusters for both root and shoot samples. In roots, GO term enrichment analyses of MDEs for each cluster revealed that processes related to ion homeostasis (MDER7 and MDER8), circadian rhythm (MDER5), response to sugar (MDER5), defence (MDER1), photosynthesis and response to light (MDER3) or cytokinesis processes (MDER2) were modulated by MYC2 across the different conditions (Extended Data Fig. 8g). In shoots, terms related to photosynthesis (MDES1), defence (MDES2), cytokinesis (MDES3), secondary metabolite biosynthesis (MDES4), response to JA (MDES5) or sugar biosynthesis (MDES7) were also altered in the myc2-3 mutant (Extended Data Fig. 8g). Notably, clusters MDES2 and MDES7 (shoot) contained GO terms for which the most significant enrichments were observed, and genes in these clusters were either induced (MDES2) or repressed (MDES7) in the myc2-3 mutant compared with MYC2-FLAG under LP in the presence of BFO ( Fig. 6c and Extended Data Fig. 8g). Therefore, MYC2-dependent transcriptional reprogramming between these two clusters probably explained the lack of BFO-mediated growth rescue under LP in the mutant. Inspection of the GO terms and associated genes revealed that BFO-triggered shoot defence responses were retained in the myc2-3 mutant under LP (MDES2), with 28% of the genes shared between this cluster and the previously identified cluster S1 (Fig. 6e,f; see also Fig. 3b). Consistent with the high levels of free SA (that is, active) measured in leaves of the myc2-3 mutant under LP (Extended Data Fig. 10), these responses involved primarily SA-related and/or SAR-related genes (PR1, BGL2, FRK1, EDS5, SARD1, AZI1 and FMO1), indicating that BFO-triggered immunity in leaves remained activated under LP in this mutant (Fig. 6g). In contrast, genes involved in starch and sugar metabolic processes were downregulated under the same condition, illustrating that priority to defence in shoots of the myc2-3 mutant was associated with altered sugar metabolic processes. These genes encode enzymes primarily involved in starch biosynthesis (SS1 and SS3), starch accumulation (ADG1 and PGI) or starch breakdown (LSF1, LSF2, PTPKIS1 and BAM3) in shoots (Fig. 6g). Our results indicate that the trade-off between starch/carbon metabolism and defence in shoots is modulated by microbiota and light in a MYC2-dependent manner.

Discussion
Reminiscent of the bidirectional communication mechanisms described along the microbiota-gut-brain axis in animals 15,54,55 , we report here that aboveground stress responses in plants can be orchestrated through long-distance communication with belowground commensals. We demonstrate that changes in aboveground light can cascade along the shoot-root axis to alter the composition of root commensal communities. Reciprocally, the presence of BFO commensals triggered leaf activation of defence or growth responses in a light-dependent manner. The integration of these responses to microorganisms and light along the root-shoot axis dictates the trade-off between growth and defence, thereby boosting plant growth under suboptimal light conditions (Fig. 7). Over the course of evolution, the roots of land plants have continuously interacted with multikingdom microbial commensals 2,56 . Consequently, microbial molecules in the rhizosphere (such as hormones, microorganism-associated molecular patterns, volatile organic compounds and lipo-chitooligomers) have been shown to serve as developmental signals for plant growth and immune system maturation 36,57,58 . However, whether microbial signals belowground can modulate aboveground stress responses and vice versa remains poorly understood. Bidirectional communica-tion mechanisms between shoot and root organs have been previously described in the context of aboveground biotic stresses 16 . Leaf colonization by microbial pathogens or herbivores results in shifts in the rhizosphere microbiota through host-induced modulation of root exudation profiles [59][60][61] . Using manipulation experiments, these pathogen/herbivore-induced shifts in rhizosphere commensal communities were shown to be the direct cause protecting the next plant generation through the promotion of systemic defence responses 11,60,61 . Modulation of the rhizosphere microbiota via leaf-pathogen-induced change in root exudation profiles therefore seems to dictate the survival and performance of the offspring (that is, the cry-for-help hypothesis). Here, we identified a link between differential canopy size across mutants and bacterial community composition under LP and validated that the composition of bacterial root commensals can drive differential investment in microbiota-induced growth or microbiota-induced defence responses. Our results suggest that host-driven recruitment of specific root commensals drives or at least contributes to plant prioritization of growth over defence responses under LP. Although investment in growth under LP is microbiota-dependent, it remains difficult to determine the exact contribution of microbial metabolites, microbiota-induced plant prioritization of growth over defence or microbiota-induced reallocation of resources from root to shoot to the growth-rescue phenotype.
The induction of systemic defence responses, including pathogen-triggered SAR (SA-dependent) and commensal-triggered induced systemic resistance (JA-and ethylene-dependent), has been extensively reported in response to specific commensal or pathogenic microorganisms [62][63][64][65][66] . Here we showed that root colonization by a multikingdom consortium of microorganisms originally isolated from the roots of healthy plants triggered aboveground induction of defence responses effective against Bc and Pst. Compared with germ-free plants, the presence of the BFO SynCom provided extensive protection from leaf pathogens, which was largely independent of the light condition. It is therefore important to note that although a significant part of BFO-induced pathogen resistance is modulated by light and MYC2, most of this protective activity occurred through a MYC2-and light-independent mechanism. Since ectopic leaf colonization by a few root microbiota members was also noted, it is possible that direct antagonism towards pathogens and local defence responses activated in response to these ectopic colonizers also contribute to aboveground pathogen protection. The observed defence responses induced by BFO commensals in the leaves of WT (cluster S1, NC + BFO) and myc2-3 (cluster MDES2, LP + BFO) plants resemble more stereotypical SAR than induced systemic resistance responses 67 , which is consistent with the fact that induced systemic resistance responses were previously found to be abolished in the myc2-3 mutant 68 . The induction of immune responses is costly for plants 69,70 , and, consistent with our data (Fig. 6), these responses are known to be associated with the downregulation of genes involved in chloroplast functions, including photosynthetic light reactions, the Calvin cycle, photorespiration and starch metabolism [70][71][72] . As chloroplasts are central integrators of multiple functions linked to photosynthesis, defence and development 62,63,73 , complex metabolic trade-offs in these organelles probably dictate plant investment into microbiota-induced growth or microbiota-induced defence responses.
A direct link between induced defences and light conditions is supported by previous work showing an absolute requirement of phytochromes, but not cryptochromes, for biological induction of SAR 22,74 . Since phytochrome-mediated light signalling is interconnected with JA signalling, probably via MYC2 (refs. 24,[75][76][77], it has been suggested that MYC2 might orchestrate the regulation of plant growth and development by light quality 49 . Our results are consistent with this hypothesis and suggest that LP-dependent downregulation of microbiota-induced defence in leaves is orchestrated     Fig. 3b (S1 to S8). NF, not found in Fig. 3b. f, GO term enrichment network depicting the top 12 most significantly enriched GO terms (hypergeometric test with Bonferroni correction, P < 0.05) in MDES2 and MDES7. the size of the GO term reflects the number of genes enriched. R, response; BP, biosynthetic process; MP, metabolic process. g, Gene-concept network (cnetplot function in R) depicting linkages between genes and the associated top 12 most significantly enriched GO terms detected in clusters MDES2 and MDES7.
through MYC2 via a cross-talk between photoreceptor signalling and defence signalling. Our observation that MYC2 is needed for LP-triggered reduction of immune activity against Pst and Bc in the presence of BFO commensals but not under axenic conditions indicates that this MYC2-dependent growth-defence trade-off requires signals from microbial commensals. This result further supports the hypothesis that, over evolutionary time, the direct integration of microbial and environmental cues by plants has been key for plant adaptation to environmental constraints. Our results suggest that the interference of MYC2 (the JA signalling pathway) with light, SA/SAR and GA signalling pathways is key to prioritizing investment in shoot growth over shoot defence under LP. In our gene expression network (Fig. 6), we identified a few genes connecting the clusters MDES2 (defence cluster) and MDES7 (growth cluster). These MDEs include EDS5 (ref. 78 ), a transporter required for SA export from chloroplasts and needed for SA signalling (induced in myc2-3 in LP); AOS 79 , a key chloroplastic enzyme needed for JA biosynthesis (repressed in myc2-3 in LP); and RGL3 (ref. 80 ), a repressor of GA responses (induced in myc2-3 in LP). As these genes are well-known MYC2 target genes 51,53,81,82 , it is conceivable that differential regulation of the expression of these genes by MYC2 orchestrates a complex cross-talk between JA, SA and GA to prioritize either microbiota-induced growth or defence according to the surrounding light conditions. Importantly, this survival trade-off might not necessarily occur in plant species that naturally grow at high density or under suboptimal light conditions, as exemplified by the concomitant activation of defence and shade avoidance responses observed in Geranium robertianum 83 .
Taken together, our data suggest that plant growth and survival in nature probably depend on the ability of these sessile organisms to utilize belowground microbial signals to prioritize either growth or defence depending on light quantity and quality perceived by leaves. Phenotypic plasticity and aboveground stress responses in plants can therefore be governed by microbial root commensals.

Methods
All experiments were performed at least three independent times, except RNA-seq experiments, which were carried out once with three biological replicates. All findings were consistent across replicates.
Microbial strains and culture conditions. The bacterial, fungal and oomycete strains used in this study have been previously reported 12,35 and are summarized in Supplementary Table 1. Pst and Bc (a benomyl derivative of the strain SAS56) were used as model pathogens for pathogen assays in A. thaliana 48,84 . Bacterial strains were routinely cultured at 25 °C in liquid 50% Tryptic Soy Broth (Sigma-Aldrich) and stored at −80 °C in 25% glycerol. Fungal strains were cultured at 20 °C in solid PGA media (Sigma-Aldrich), and agar plugs containing mycelia were stored at −80 °C in 30% glycerol. Oomycete strains were continuously propagated in solid PGA media. Pst was cultured at 28 °C in NYGA medium and stored at −80 °C in 7% dimethyl sulfoxide. Bc spores were obtained from −80 °C glycerol stocks in which the concentration was adjusted to 10 7 spores per ml in Vogelbuffer (in 1 l: 15 g sucrose, 3 g Na-citrate, 5 g K 2 HPO 4 , 0.2 g MgSO 4 7H 2 O, 0.1 g CaCl 2 2H 2 O, 2 g NH 4 NO 3 ).

Plant material and growth conditions. A. thaliana
Col-0 and the mutants used in this study are provided in this section and Supplementary Table 4. A. thaliana Col-0 WT (N60000) was obtained from the Nottingham Arabidopsis Stock Centre. The dellaP mutant (della pentuple) was previously generated by crossing ga-28, rgl1-SK62, rgl2-SK54, rgl3-3 and introgressed gai-t6 (ref. 40 ). The jazQ mutant (jaz quintuple) was previously obtained from T-DNA insertion mutants of jaz1-2, jaz3-4, jaz4-1, jaz9-4 and jaz10-1 (ref. 24 ). The mutants dde2-2 (CS65993) 38 , sid2-2 (CS16438) 37 , myc2-3 (salk_061267) 39 , sav3-3 (ref. 43 ), bri1-301 (ref. 41 ), myc234 (ref. 85 ), cry1-304 cry2-1 (ref. 86 ) and jin1-8 pMYC2::MYC2-FLAG 50 were previously reported. Seeds of A. thaliana Col-0 and mutant plants were surface-sterilized in 70% ethanol for 18 min followed by a brief wash with 100% ethanol (1 min). The seeds were dried out under sterile bench conditions and were incubated for two days at 4 °C in the dark. Individual seeds were sown onto the surface of FlowPots 30 by pipetting one seed at a time. Before seed sowing, the FlowPots were inoculated with half-strength Murashige and Skoog medium (without sucrose, pH 5.5; Sigma-Aldrich) with or without microbial commensals. Microbiota reconstitution experiments in the FlowPot system. The bacterial strains were cultivated in 96-deep-well plates containing 400 μl of 50% Tryptic Soy Broth (Sigma-Aldrich) with 180 r.p.m. shaking speed for seven days at 25 °C and subsequently pooled at equal volume ratios. This bacterial pool was centrifuged at 4,000 g for 10 min and resuspended in 10 mM MgCl 2 to remove residual media and bacteria-derived metabolites. Prior to inoculation, OD 600 was adjusted to 0.5. Individual fungal and oomycete strains were cultivated on solid PGA medium for 14 days. Then, 100 mg of fungal and 40 mg of oomycete mycelium was harvested for each strain and aliquoted into 2 ml Eppendorf tubes containing 1 ml of MgCl 2 and one sterile stainless-steel bead (3.2 mm). The mycelium was subsequently crushed with a paint shaker (SK450, Fast & Fluid Management) for 10 min. Fragmented fungal or oomycete mycelia were then pooled at equal volume ratios at a concentration of 100 mg ml −1 for fungi and 40 mg ml −1 for oomycetes. For the microbial ratio experiment, either 2 ml (high concentration) or 0.2 ml (low concentration) of the bacterial suspension (OD 600 = 0.5; see above), 1 ml (high) or 0.1 ml (low) of the fungal suspension (100 mg ml −1 ; see above), and 1 ml (high) or 0.1 ml (low) of the oomycete suspension (40 mg ml −1 ; see above) was transferred into a falcon containing 50 ml of Murashige and Skoog medium without sucrose, which was used to repopulate sterile peat in the FlowPot. For all the other experiments in the FlowPot system, 0.2 ml (low) of the bacterial suspension (OD 600 = 0.5), 0.1 ml (low) of the fungal suspension (100 mg ml −1 ) and 0.1 ml (low) of the oomycete suspension (40 mg ml −1 ) were used. The procedures for setting up the gnotobiotic FlowPot system were carried out as previously described 12,30 . One week after incubation, the seedlings were randomly thinned out under a sterile bench to keep three plants per FlowPot. The plants were harvested at the vegetative stage five weeks after seed sowing for all experiments.

CAS soil and CAS wash repopulation experiments in the FlowPot system.
The natural CAS mixed with sterile vermiculites (volume of soil:vermiculite, 2:1) was used as the matrix in the FlowPot. The procedures for setting up the FlowPot system were carried out as described above. The properties of CAS soil and peat were measured by Labor für Boden-und Umweltanalytik and are provided in Supplementary Table 1. For the CAS wash, 10 ml of soil was washed with a sterile detergent (sterile 1× TE + 0.1% Triton X-100) in a 1:10 ratio (soil:detergent) in a 50 ml falcon tube and shaken vigorously until the soil pellet was well mixed. The tube was then placed in a rotator for 30 min at 40 r.p.m. at room temperature. To remove big soil particles, the tube was centrifuged for 1 min at 1,500 r.p.m. The supernatant was transferred to a new falcon tube and centrifuged for 20 min at 4,000 r.p.m. The supernatant was then discarded, and the pellet was resuspended in the same initial volume of 10 mM MgCl 2 as the CAS wash. Then the CAS wash was repopulated in the sterile peat matrix. Meanwhile, a heat-killed soil wash (CAS wash HK) was prepared by incubating it at 95 °C for 45 min. The sterility of the heat-killed soil wash was validated by plating it on TSA and PGA plates. The other procedures for setting up the FlowPot system were carried out as described above.
Shoot trait measurements. Shoots of individual plants were cut, and their fresh weight was measured first. The shoots were placed on white paper, sealed with polyester non-sterile transparent film (VWR) and scanned (Perfection V600 Photo, Epson). Canopy size, petiole length, leaf length, leaf width, leaf numbers and hypocotyl length were measured using ImageJ (Fiji) 87 . After the first PCR, the enzymes and ssDNA were digested using a mixture of 1 μl of Antarctic phosphatase, 1 μl of Exonuclease I and 2.44 μl of Antarctic phosphatase buffer that was added to 20 μl of the first-step PCR product (37 °C for 30 min and 85 °C for 15 min). After centrifugation, 3 μl of this solution was used as a template for a second PCR (ten cycles) 12 . This step involved the same aforementioned PCR mix except that the universal primers were barcoded (reverse primers only) and included P5 and P7 Illumina adaptors. After PCR, 5 µl of PCR product for each barcoded sample was mixed with 5 µl of 6× Orange G DNA Loading Dye (Orange G, 6×, Sigma) and run on a 1% agarose gel in TAE 1× buffer. After control gel checking, the remaining PCR product (around 70 μl) was mixed with 20 μl of Gel Loading Dye (Orange G) and run on a 1.5% agarose gel for 2 h at 80 V to excise the bacterial 16S rRNA band (500 base pairs (bp)). The bands were cut and purified using the QIAquick gel extraction kit (QIAGEN), and DNA concentration was measured using the PicoGreen method as described above. For fungi and oomycetes, DNA purification was performed using the Agencourt AMPure XP-PCR Purification method (Beckman Coulter). DNA sample concentration was then quantified by the Quant-iT PicoGreen dsDNA method, as described above.

Microbial
For each microbial group (B, F and O), 30 ng of DNA of each of the barcoded samples were mixed, resulting in three pooled samples that were purified twice (AMPure XP-PCR Purification, Beckman Coulter). The final DNA concentration of each pool was measured using the Quantus Fluorometer (Promega), and an equal quantity of DNA (that is, 100 ng) was used to assemble a single library from the three bacterial, fungal and oomycete DNA pools. Finally, paired-end Illumina DNA sequencing was performed using the Illumina MiSeq system available on-site at the Max Planck Institute for Plant Breeding Research.
16S rRNA and ITS read processing. Paired amplicon sequencing reads were joined using Qiime (v.1.9.1) (join_paired_reads) 88 . In the case of ITS reads, for unjoined read pairs, corresponding forward reads were retained for demultiplexing. Demultiplexing and quality filtering were done using Qiime (split_libraries_fastq, phred = 30). All quality filtered and demultiplexed reads were trimmed to an equal length. Reference sequences were obtained from all used strains (183 bacteria, 24 fungi and 7 oomycetes; see Supplementary Table 1) using the respective resources of sequenced genomes, if available. The reference sequences were then trimmed to contain only the amplified region. Trimmed reference sequences that were 100% identical were grouped in so-called strain variants. This led to 115 bacterial, 24 fungal and 7 oomycete strain variant sequences. The reference sequences of strain variants were mapped against all trimmed amplicon reads using usearch (v.8.0.1517) 89 , allowing for one mismatch (usearch -usearch_global, with max diff = 1). Unmapped reads were discarded. Count tables were made from these mapping results.
Microbial community profile statistical analysis. To calculate alpha diversity indices, the count tables were rarefied to 1,000 reads. For the microbial profiling data, only samples with more than 1,000 reads were considered. Significant differences between alpha diversity indices were determined by an ANOVA followed by post hoc Tukey's HSD (P < 0.05). Distance matrices were calculated by using a normalized count table (CSS) 90 , which was used for Bray-Curtis distance calculation. Distance tables were used as an input for PCoA and CAP (capscale function in R). The condition function was used to correct for batch effects and when applicable for treatments (LP or NC) when performing CAP. The significance of the CAP results was tested using ANOVA (anova.cca in R, P < 0.05). PLS-DA was used to discriminate groups of samples on the basis of the composition of their microbial communities. PLS-DA consists of a partial least squares regression analysis where the response variable is categorical (y-block; describing the grouping factor), expressing the class membership of the statistical units 91  The enrichment of strain variants across LP and NC conditions was calculated using the following steps (all functions are from the R package EdgeR). Raw read counts were normalized using TMM normalization (calcNormFactors). A generalized linear model was fitted to integrate the batch effect (glmFit). Enrichment was then determined with a likelihood ratio test (glmLRT, P < 0.05). A predictive SVM model with a linear kernel was trained (SVC function of Scikit-learn) 92 to link standardized RAs of bacteria to plant rescue. Recursive feature elimination with cross-validation (RFECV function) was performed to identify the smallest set of bacteria discriminating the two plant phenotypes and to estimate the model accuracy using a leave-one-out approach (K-fold cross-validator, with K equalling the number of samples). The species identified in the RFECV were then used to compute a PLS-DA.
Transcriptome sequencing experiments. Shoots and roots of five-week-old A. thaliana Col-0, myc2-3 and jin1-8 pMYC2::MYC2-FLAG plants growing in the FlowPot system were harvested separately at 10 a.m. The roots were washed quickly (<1 min) to detach the surrounding peat matrix using 10% RNAlater (QIAGEN) in 1× PBS as the capture buffer to mitigate RNA degradation. Total RNA from all samples was extracted using the RNeasy Plant Mini Kit (QIAGEN). DNA removal was performed using RNase-Free DNase Set (QIAGEN). Eleven samples from myc2-3 plants detected more than 50% reads mapped to latent genome. On the basis of the hierarchical relationship between samples from myc2-3 mutants (Extended Data Fig. 8b), the contamination with latent virus did not significantly affect the plant transcriptome. RNA-seq libraries were prepared and sequenced at the Max Planck Genome Centre with an Illumina HiSeq2500. For samples from Col-0 plants, the run conditions were 1 × 150 bp (single read), and the total reads per sample were 20 million. For samples from myc2-3 plants, the run conditions were 1 × 150 bp (single read), and the total reads per sample were 55 million. For samples from jin1-8 pMYC2::MYC2-FLAG plants, the run conditions were 1 × 150 bp (single read), and the total reads per sample were 10 million.
Transcriptome sequencing data analysis. The FastQC suite (http://www. bioinformatics.babraham.ac.uk/projects/fastqc/) was performed to check the quality of the sequenced reads. The RNA-seq reads were then mapped to the annotated genome of A. thaliana (TAIR10) using Tophat2 (v.2.1.1) (tophat2 parameters: "p, 20; a, 10; g, 10') 93 with Bowtie2 (ref. 94 ) (v.2.2.3) building the genome index. The mapped RNA-seq reads were subsequently transformed into a fragment count per gene per sample using the htseq-count script (s, reverse; t, exon) in the package HTSeq 95 (v.0.11.2). Count tables of Col-0 samples were concatenated to one count matrix. Count tables of myc2-3 and jin1-8 pMYC2::MYC2-FLAG samples were concatenated to one count matrix. Raw counts were normalized via TMM normalization. Then, the genes with more than 100 counts were extracted. The extracted TMM-normalized count data were transformed to log 2 cpm via the voom function in the limma package 96 in R (R v.3.6.3; https://www.r-project.org/). Subsequently, the log 2 cpm data were used to calculate log 2 fold changes and P values (F-test) for individual comparisons. The resulting P values were adjusted for false discoveries due to multiple hypothesis testing via the Benjamini-Hochberg procedure. To identify significantly differently expressed genes, a threshold of |log 2 FC| ≥ 1 and FDR < 0.05 was applied. Heat maps of the expression profiles of significantly regulated genes were generated using the pheatmap package in R. The Euclidean distance was used to show the distance among clustering rows. The values in the heat maps were scaled by row. GO enrichment was conducted using the enrichGO function in the clusterProfiler package 97 in R. Biological process and 0.05 P value cut-offs were chosen. The P values were adjusted via the Benjamini-Hochberg method.
Variance partitioning of microbial community composition and transcriptomic profiles. Variance partitioning between experimental factors was tested with a permutational ANOVA approach with the Adonis function (R package vegan) in all experiments. For the stress experiment, the effect of compartment and light condition factors on microbial community composition was tested in a global model. The effect of light conditions was also tested in separate models for root and shoot samples. These models were constructed separately on each of the bacterial, fungal and oomycete datasets using Bray-Curtis dissimilarity matrices between pairs of samples produced with the vegdist function (R package vegan). For the mutant experiment, the effects of genotype, compartment and light condition factors on bacterial community composition were tested in a global model. The effects of genotype and light conditions were also tested in separate models of root and matrix samples. These models were constructed using Bray-Curtis dissimilarity matrices between pairs of samples produced with the vegdist function (R package vegan). For the transcriptomic experiments, the effects of the experimental factors on the transcriptomic expression profiles of A. thaliana plants were tested using Euclidean distance matrices between pairs of samples produced with the vegdist function (R package vegan). Models were constructed separately for Col-0 plants and for myc2-3 and jin1-8 pMYC2::MYC2-FLAG plants, as these were harvested in separate experiments. For Col-0 plants, the effects of compartment, microorganism inoculation and light conditions on the plant transcriptomic profile were tested in a global model. The effects of microorganism inoculation and light conditions were also tested in separate models for root and shoot samples. For myc2-3 and jin1-8 pMYC2::MYC2-FLAG plants, the effects of compartment, genotype, microorganism inoculation and light conditions on the plant transcriptomic profile were tested in a global model. The effects of genotype, microorganisms and light were also tested in separate models for root and shoot samples. These models were also constructed for the myc2-3 and jin1-8 pMYC2::MYC2-FLAG datasets separately to test the effects of compartment, microorganism inoculation and light independently of the genotype effect.

Pathogen inoculation and symptom quantification.
For Bc inoculation of A. thaliana, the spores were diluted in Vogelbuffer (in 1 l: 15 g of Suc, 3 g of Na-citrate, 5 g of K 2 HPO 4 , 0.2 g of MgSO 4 ·7H 2 O, 0.1 g of CaCl 2 ·2H 2 O and 2 g of NH 4 NO 3 ) to 5 × 10 5 spores per ml. For droplet inoculations, 2 μl droplets containing 1 × 10 3 spores were applied to each leaf of four-week-old A. thaliana grown in the presence or absence of the BFO SynCom under either NC or LP. The entire infection process was conducted under a sterile clean bench. Five days after pathogen inoculation in the FlowPot system, the shoots were cut, washed using Milli-Q water twice and dried out with sterilized Whatman glass microfibre filters (GE Healthcare Life Sciences). The shoots were weighed, placed into 2 ml sterile Eppendorf tubes and snap-frozen in liquid nitrogen. For quantification of fungal growth, DNA from the plant shoots was extracted using the DNeasy Plant Mini Kit (QIAGEN). The relative amounts of Bc and A. thaliana DNA were determined by qPCR, employing specific primers for cutinase A and SKII, respectively, as described previously 48 .
Pst inoculation was carried out by spraying Pst at 0.2 OD in 10 mM MgCl 2 on the leaves of four-week-old A. thaliana grown in the presence or absence of the BFO SynCom under either NC or LP. The entire infection process was conducted under a sterile clean bench. Five days after pathogen inoculation in the FlowPot system, the plant shoots were cut, washed using 70% ethanol once and rinsed in Milli-Q water twice. The shoots were dried with sterilized Whatman glass microfibre filters and weighed under a sterile clean bench. Then, the shoots were put into 2 ml sterile Eppendorf tubes containing 1.5 ml of 10 mM MgCl 2 /0.01% Silwet. The tubes were then directly shaken at 650 r.p.m. for 1 h at 28 °C. For colony counting, a series of dilutions were conducted with 10 mM MgCl 2 to 10 −1 ,10 −2 ,10 −3 ,10 −4 and 10 −5 . Then, 20 μl of diluted liquid was spotted on NYGA plates, and colonies were counted after two days of incubation at 28 °C.
Immunoblot analysis. To analyse MYC2 protein levels of plants under NC or LP in the presence or absence of BFO, the shoots and roots of A. thaliana Col-0 and jin1-8 pMYC2::MYC2-FLAG plants (grown in the FlowPot system under NC or LP conditions as described above) were harvested separately into 2 ml Eppendorf tubes with one 3.2 mm stainless-steel bead five weeks after −BFO or +BFO inoculation and frozen in liquid nitrogen. Samples with the same fresh weight were ground to powder in liquid nitrogen and boiled in 150 μl of 2× Laemmli buffer for 10 min at 95 °C. The proteins were resolved on 10% SDS-PAGE (1610156, Bio-Rad) and transferred using the wet transfer method onto a nitrocellulose membrane (10600001, GE Healthcare Life Sciences). Monoclonal anti-β-actin-peroxidase antibody produced in mouse (A3854, Sigma-Aldrich; clone: AC-15, monoclonal) was used to detect actin and to adjust the total protein concentration. Monoclonal anti-FLAG M2-peroxidase antibody produced in mouse (A8592, Sigma-Aldrich; clone: M2, monoclonal) was used to detect MYC2 protein. Both antibodies were used at a dilution of 1:5,000 (1× TBST, 5% milk powder). First, a full membrane was used to check MYC2 and actin separately. Then, the membranes performed later were cut at 50 kDa to separate MYC2 (roughly 75 kDa) and actin (roughly 40 kDa) to detect both proteins together. Detection of the signal was performed with SuperSignal West Pico and Femto (34080 and 34095, ThermoFisher Scientific) using ChemiDoc Imaging Systems (Bio-Rad). The protein concentration was calculated on the basis of the band thickness using Fiji.
Phytohormone measurement. To measure phytohormones of plants under NC or LP in the presence or absence of BFO, the shoots and roots of A. thaliana Col-0 and myc2-3 plants (grown in the FlowPot system under NC or LP conditions as described above) were harvested separately into 2 ml Eppendorf tubes with one 3.2 mm stainless-steel bead. Phytohormone concentrations were determined from 50 mg of tissue (fresh weight). Sample processing, data acquisition, instrumental set-up and quantifications (using 5 ng of [ 2 H] 6  Reporting Summary. Further information on research design is available in the Nature Research Reporting Summary linked to this article.

Data availability
The sequencing reads from the microbiota reconstitution experiments (MiSeq 16S rRNA and ITS reads) have been deposited in the European Nucleotide Archive (bacteria at PRJEB40980, fungi at PRJEB40981 and oomycetes at PRJEB40982). The sequencing reads from the transcriptome sequencing experiments have been deposited in the Gene Expression Omnibus (Col-0 data at GSE160106 and myc2-3 and jin1-8 pMYC2::MYC2-FLAG data at GSE160115). All raw data in this study are available at https://github.com/ShijiHou/Light-limitation-Paper.

Code availability
All scripts for computational analysis are available at https://github.com/ShijiHou/ Light-limitation-Paper. side stands for sample distances (using "dist" function in "DESeq" R package to the transpose of the transformed count matrix to get sample-to-sample distances) corresponding to the colour range in the heatmap. b, Bar plot showing the numbers of differentially regulated genes among all possible pairwise comparisons. Red colour and blue colour filled bars represent significantly up-and down-regulated genes between the corresponding comparisons, respectively. the numbers highlighted inside the bars indicate the total number of up-or down-regulated genes between the corresponding comparisons. the numbers above the bars represent the numbers of genes shared between the corresponding conditions. c, and d, Gene Ontology (GO) term enrichment depicting the top 12 most significantly enriched GO terms (Hypergeometric test with Bonferroni correction, P < 0.05) detected in all clusters for root samples and shoot samples. the size of point reflects the amount of gene numbers enriched in this GO term. the colour of point means the p value ((Hypergeometric test with Bonferroni correction). R: response, Reg. Regulation, Et: ethylene, SA: salicylic acid, JA: jasmonic acid, SAR: systemic acquired resistance, CW: cell wall, CG: cell growth, P: process, Com: compound, Hex: hexakisphosphate, inv.: involved in. e, Gene-concept network (cnetplot function in R) depicting linkages between genes and associated top 12 most significantly enriched GO terms detected in clusters R7 (root) and S4 (shoot). Each node represents a gene and is colour-coded according the different cluster names. MP: metabolic process. BP: biosynthetic process. Fig. 6 | BFO-induced growth over defence in lP depends on JA and BR pathways. Canopy size (a), shoot fresh weight (c), mean petiole length (e), total leaf number (g), leaf length/width ratio (i) of five-week-old A. thaliana plants grown under NC and LP, in the absence (-BFO) or presence of the BFO SynCom (+BFO). three independent biological replicates are shown (a: n = 1188 plants; c: n = 1188 plants; e: n = 1095 plants; g: n = 925 plants; i: n = 9,708 leaves). Letters indicate statistical significance (Kruskal-Wallis with Dunn's post hoc test, α = 0.05). Summary heatmap depicting the mean of canopy size (b), shoot fresh weight (d), mean petiole length per plant (f), total leaf number per plant (h), leaf length/width ratio (j) measured across plant genotypes and conditions. B. cinerea B05.10 (Bc) growth (k) and Pseudomonas syringae pv. tomato DC3000 (Pst) growth (m) in leaves of A. thaliana Col-0 and eight mutants grown under NC and LP, in the absence (-BFO) or presence of the BFO SynCom (+BFO), in the FlowPot system. three independent biological replicates are shown (k: n = 336; m: n = 324). k and m, Statistical significance across conditions for each genotype is indicated with letters (oneway ANOVA followed by post hoc tukey´s HSD, α = 0.05). Summary heatmap depicting relative Bc growth (l) and Pst growth (n) measured across plant genotypes and conditions. o, Canopy size (top) and B. cinerea B05.10 (Bc) growth (bottom) in leaves of five-week-old myc2-3 and myc234 mutants grown under NC and LP, in the absence (-BFO) or presence of the BFO SynCom (+BFO), in the FlowPot system. three independent biological replicates are shown (Canopy size: n = 216; Bc growth: n = 72). Letters indicate statistical significance (Canopy size: Kruskal-Wallis with Dunn's post hoc test, α = 0.05; Bc growth: one -way ANOVA followed by post hoc tukey´s HSD, α = 0.05). a, c, e, g, i, k, m, Boxplot boundaries reflect the inter-quartile range, the centre line is the median, and whiskers represent 1.5× interquartile range of lower and upper quartiles.

nature research | reporting summary
April 2020 Corresponding author(s): Dr. Stéphane Hacquard Last updated by author(s): May 27, 2021 Reporting Summary Nature Research wishes to improve the reproducibility of the work that we publish. This form provides structure for consistency and transparency in reporting. For further information on Nature Research policies, see our Editorial Policies and the Editorial Policy Checklist.

Statistics
For all statistical analyses, confirm that the following items are present in the figure legend, table legend, main text, or Methods section.

n/a Confirmed
The exact sample size (n) for each experimental group/condition, given as a discrete number and unit of measurement A statement on whether measurements were taken from distinct samples or whether the same sample was measured repeatedly The statistical test(s) used AND whether they are one-or two-sided Only common tests should be described solely by name; describe more complex techniques in the Methods section.
A description of all covariates tested A description of any assumptions or corrections, such as tests of normality and adjustment for multiple comparisons A full description of the statistical parameters including central tendency (e.g. means) or other basic estimates (e.g. regression coefficient) AND variation (e.g. standard deviation) or associated estimates of uncertainty (e.g. confidence intervals) For null hypothesis testing, the test statistic (e.g. F, t, r) with confidence intervals, effect sizes, degrees of freedom and P value noted

Software and code
Policy information about availability of computer code Data collection ImageJ (Fiji) was used to measure canopy size, Petiole length, leaf length, leaf width and quantify the band thickness of western blot. Data analysis R (version 3.6.3) was used for data transformation, analysis, and visualization. Qiime (v1.9.1, http://qiime.org/) and Usearch (v8.0.1517, https://www.drive5.com/usearch/) were used to process 16S rRNA and ITS read alignment. DADA2 (version 1.12.1) and Vegan (version 2.5.6) were used for denoising and analyzing 16s rRNA amplicon reads, and community profiling analysis diversity analyses. edgeR (version 3.24.3) was used to identify differentially enriched Sequence Variants. tidyverse (version 1.2.1) was used for data formating. RNASeq reads were checked with fastqc suite (http://www.bioinformatics.babraham.ac.uk/projects/fastqc/), aligned using Tophat2 (version 2.1.1) and Bowtie2 (version 2.2.3), and normalized by limma package (version 3.42.2). pheatmap package (version 1.0.12) was used to generate heatmaps. Gene ontology analysis was performed with the clusterProfiler package (version 3.14.3). ggplot2 (version 3.3.0) and ggpubr (version 0.4.0) were used for data visualization. Final figures were prepared in Adobe Illustrator 2020 for combining, positioning, and resizing plot and text. Code & raw data: https://github.com/ShijiHou/Light-limitation-Paper For manuscripts utilizing custom algorithms or software that are central to the research but not yet described in published literature, software must be made available to editors and reviewers. We strongly encourage code deposition in a community repository (e.g. GitHub). See the Nature Research guidelines for submitting code & software for further information.

Field-specific reporting
Please select the one below that is the best fit for your research. If you are not sure, read the appropriate sections before making your selection.

Life sciences Behavioural & social sciences Ecological, evolutionary & environmental sciences
For a reference copy of the document with all sections, see nature.com/documents/nr-reporting-summary-flat.pdf

Life sciences study design
All studies must disclose on these points even when the disclosure is negative.

Sample size
Statistical methods were not used to predetermine necessary sample size. Rather, efforts were made to ensure large replicate numbers per experiment (usually n>24 per experiment), and replicate experiments were performed. For microbial profiling analyses, root samples from six plants grown in two single pots were pooled to create a single sample. Soil samples from two single pot were pooled to create a single sample. We aimed to have at least three root samples and three soil samples per treatment, and three independent biological replicate experiments. Overall, for each genotype and each condition, 18 samples (9 samples from roots and 9 samples from matrixes) were chosen from 3 independent biological replicates (3 technical replicates per independent biological replicate).
For RNASeq experiments, samples (shoot and root) from 12 plants grown in four single pots were pooled to create a single sample. In total, we had three root samples and three shoot samples per treatment.
For microbiota reconstitution and plant growth assays experiments, experiments were performed with n=54 plants per condition, split across 3 replicate plates each containing 18 individual plants. In this way, each experiment contained replicate microbiota reconstitution conditions (n=3 plates) and n=18 plants per plate. Experiments were repeated at least three times and the results pooled.
Data exclusions Very few samples where 16S rRNA or ITS sequences did not meet predetermined quality checks (i.e. low quality or low read counts) were excluded. Very few samples where the leaf was broken during harvesting were excluded for measuring leaf traits.

Replication
Experiments were performed at least three independent times to confirm reproducibility. All findings from replicate experiments were consistent and confirmed.
Randomization Seeds of each genotype were allocated into different conditions at random without note of the different conditions. Each replicate pot or plate of each condition of the experiment was distributed at random into the growth chamber, and randomly shuffled every two to three days to minimize location-based effects. Samples were collected randomly for each condition in all experiments.

Blinding
During sample harvesting and measuring shoot fresh weight, no procedure was taken to blind experimenters to the different conditions since different light treatments are obviously distinguished. Harvested tissues were stored in numbered tubes without sample condition or identity labels. The sample metadata was stored in a separate file. Thereby, during the processing of samples for microbial profiling, RNASeq analysis, leaf traits, leaf pathogen quantification, and phytohormone content, experimenters did not know the identity of the samples.

Reporting for specific materials, systems and methods
We require information from authors about some types of materials, experimental systems and methods used in many studies. Here, indicate whether each material, system or method listed is relevant to your study. If you are not sure if a list item applies to your research, read the appropriate section before selecting a response.