ER body-resident myrosinases and tryptophan specialized metabolism modulate root microbiota assembly

root – microbe interactions, we hypothesized that ER bodies in roots are important for interaction with soil-borne microbes at the root – soil interface. (cid:1) We used mutants impaired in ER bodies ( nai1 ), ER body-resident myrosinases ( pyk10b-glu21 ), IG biosynthesis ( myb34/51/122 ), and Trp specialized metabolism ( cyp79b2b3 ) to proﬁle their root microbiota community in natural soil, evaluate the impact of axenically collected root exudates on soil or synthetic microbial communities, and test their response to fungal endophytes in a mono-association setup. (cid:1) Tested mutants exhibited altered bacterial and fungal communities in rhizoplane and endo-sphere, respectively. Natural soils and bacterial synthetic communities treated with mutant root exudates exhibited distinctive microbial proﬁles from those treated with wild-type (WT) exudates. Most tested endophytes severely restricted the growth of cyp79b2b3 , a part of which also impaired the growth of pyk10bglu21 . (cid:1) Our results suggest that root ER bodies and their resident myrosinases modulate the proﬁle of root-secreted metabolites and thereby inﬂuence root – microbiota interactions.


Introduction
Endoplasmic reticulum (ER) bodies (Fig. 1a; Matsushima et al., 2003a) are ER-derived spindle-shaped organelles observed in plants from a monophyletic group of families within the order Brassicales, that is, Brassicaceae, Cleomaceae, and Capparaceae (Hayashi et al., 2001;Hara-Nishimura & Matsushima, 2003).These structures were originally discovered in radish roots using an electron microscope and named 'dilated cisternae' after their characteristics as a part of the ER cisternae (Bonnett & Newcomb, 1965).ER bodies are constitutively developed in the ground tissue of roots (epidermis, cortex, and endodermis), implying their role in root physiology.In aboveground tissues, ER bodies are found mainly in seedlings (cotyledon and hypocotyl), especially in the epidermal cell layers, but not in mature tissues, such as mesophyll cells and rosette leaves.Interestingly, ER body formation is strongly induced by wounding or jasmonic acid treatment in mature rosette leaves (Matsushima et al., 2002;Ogasawara et al., 2009), suggesting a role in wounding stress response.Consistently, recent studies showed that leaf ER bodies are important for defense against herbivory (Nakazaki et al., 2019;Yamada et al., 2020).Another study suggested that leaf ER bodies triggered by a pathogen-derived jasmonate mimicry, coronatine, negatively regulate immune responses against a bacterial pathogen Pseudomonas syringae pv tomato DC3000 (Rufian et al., 2021).These findings suggest that leaf ER bodies are involved in environmental interactions, while the role of root ER bodies remains ununderstood.
ER body formation in leaves and roots is regulated by a transcription factor NAI1 in Arabidopsis thaliana, and b-glucosidases, namely PYK10 and BGLU21 in roots/seedlings and BGLU18 in rosette leaves, are accumulated inside the ER bodies (Matsushima et al., 2003b;Ogasawara et al., 2009).These enzymes were recently shown to be myrosinases (Nakano et al., 2017;Nakazaki et al., 2019), which is a class of b-glucosidases that are capable of hydrolyzing sulfur-containing specialized thio-glucosides called glucosinolates.Glucosinolates accumulate in plants of the order Brassicales and are crucial for defense against herbivores and microbial pathogens (Wittstock & Halkier, 2002).While intact glucosinolates are biologically inert, their hydrolysis by myrosinases removes the glucose moiety and produces unstable aglycons that decompose into a wide range of terminal products, such as isothiocyanates and nitriles, which are thought to be actively influencing the behavior of insects and microbes.It is, therefore, conceivable that the terminal products of glucosinolate catabolism initiated by these ER body-resident myrosinases mediate the defensive role of ER bodies against the woodlice.
Glucosinolates are categorized into several groups based on the side chain structures.Glucosinolates with an aliphatic side chain (aliphatic glucosinolates, AGs) and with a benzyl side chain (benzyl glucosinolates; BGs; also known as aromatic glucosinolates) are mainly produced from methionine (Met) and phenylalanine (Phe), respectively, while glucosinolates with an indolic side chain (indole glucosinolates, IGs) are primarily produced from tryptophan (Trp).These glucosinolates and their catabolic products are well known for their role in antiherbivorous defense in leaves and seeds, as well as in defense against microbial pathogens.For instance, genetic disruption of leaf IG metabolism results in severe disease susceptibility to a wide range of filamentous pathogens (Bednarek et al., 2009;Clay et al., 2009;Hiruma et al., 2010;Frerigmann et al., 2016).Furthermore, genetic disruption of CYTOCHROME P450 (CYP) 79B2 and 79B3 enzymes, which are responsible for the production of all Trp-derived specialized metabolites including IGs, triggers aberrant interaction with soil-borne nonpathogenic microbes at the root-soil interface (Lahrmann et al., 2015;Hiruma et al., 2016;Koprivova et al., 2019;Frerigmann et al., 2021;Wolinska et al., 2021).For example, this mutant cannot properly manage the accommodation of beneficial endophytic fungi, such as Serendipita vermifera, Serendipita indica, and Colletotrichum tofieldiae, and exhibits severe growth defects in the presence of these fungi (Lahrmann et al., 2015;Hiruma et al., 2016).Likewise, the cyp79b2b3 mutant in the presence of a synthetic microbial community (SynCom) composed of commensal bacteria and endophytic fungi fails to control the overall abundance of fungi in roots ('fungal load'), and its growth is strongly restricted (Wolinska et al., 2021).
Notably, Trp-derived specialized metabolites, including IG catabolites, are reported to be secreted into the rhizosphere (Schreiner et al., 2011;Strehmel et al., 2014;Xu et al., 2017).A. thaliana roots express several myrosinases, including but not limited to PYK10 and BGLU21.Namely, THIOGLUCO-SIDE GLUCOHYDROLASE 4 (TGG4) and TGG5 and PENETRATION 2 (PEN2) myrosinases are also strongly expressed in roots, among which PEN2 is known to be a crucial component of the plant innate immune system against microbial pathogens (Lipka et al., 2005;Bednarek et al., 2009), while the functional roles of TGG4 and TGG5 are yet to be addressed.Given the well-characterized roles of IG catabolites in interaction with microbes, it is plausible that these root-secreted compounds derived from glucosinolates and/or Trp play a role in rootmicrobe interactions.This idea is supported by a previous study showing that metabolic engineering of a foreign glucosinolate in A. thaliana impacted the assembly of the root-associated microbial community (root microbiota; Bressan et al., 2009).
Here, we hypothesized that PYK10 mediates the hydrolysis of IGs and other Trp-derived specialized metabolites and the secretion of their terminal products into the rhizosphere, ultimately affecting the composition of root microbiota.By profiling the root microbiota shaped in the roots of mutants impaired in ER bodies and Trp-derived metabolic pathways grown in natural soil, we provide genetic evidence that these pathways are involved in the root microbiota assembly.We also show that this effect on microbial communities is partially mediated by root-exuded compounds and that lack of these pathways has a direct impact on (1) bacterial community composition in the absence of fungi and (2) fungal behavior in the absence of bacteria.Overall, our results demonstrate a physiological role for root ER bodies in modulating root microbiota assembly at the root-soil interface.

Root harvesting and fractionation
Plants were cultivated in pots containing CAS under long-day conditions in a glasshouse, where pots with different genotypes of plants were placed in a randomized design.Roots were then harvested and fractionated into the rhizosphere, rhizoplane, and endosphere fractions, as described previously (Duran et al., 2018).Briefly, the soil particles loosely attached to the root ('rhizosphere') were collected by quickly shaking by hand in sterile water.Microbes on the root surface ('rhizoplane') were collected by vigorously shaking roots in a buffer containing a detergent by hand for 2 min, followed by filtration through a 0.22-lm membrane.Roots were then surface-sterilized with 70% (v/v) ethanol and bleach solution (sodium hypochlorite; c. 2% active chlorine) for 45-60 s each to obtain the 'endosphere' fraction.Soil from unplanted pots that underwent the same watering procedures next to plant-containing pots were collected as 'bulk soil' samples.All samples were immediately frozen in liquid nitrogen and stored at À80°C until processing.

Collection of root exudates
Root exudates were collected from plants hydroponically grown in axenic glass jars containing glass beads, as described previously (Wippel et al., 2021).Seeds of Col-0 WT, as well as pyk10bglu21 and cyp79b2b3 mutant plants, were surface-sterilized with 70% (v/v) ethanol and sodium hypochlorite with c. 2% (w/v) active chlorine and germinated on metal meshes placed on half-strength Murashige and Skoog (MS) media (Sigma-Aldrich) supplemented with 1% (w/v) sucrose and 1% (w/v) agar.Seeds were stratified for 48 h at 4°C in the dark and cultured for 4 d under short-day conditions (10 h under light at 21°C and 14 h under dark at 19°C).The 4-d-old seedlings on the mesh were transferred aseptically into glass jars containing sterile glass beads (1 mm) and 26 ml of half-strength liquid MS media.Glass jars were placed in a breathable Microbox (Sac O2, Belgium) and cultivated for 5 wk under short-day conditions.The hydroponic medium was collected using an aseptic stainless needle into 50ml tubes and concentrated to c. 1/10 volume by a lyophilizer.

Obtaining roots for metabolite profiling
To collect roots for metabolite analysis, Col-0 WT and pyk10b-glu21 and cyp79b2b3 mutant plants were cultured on halfstrength MS media supplemented with 1% (w/v) sucrose and 1% (w/v) agar for 5 wk under short-day conditions as described previously.Plates were placed into the D-Root system (Silva-Navas et al., 2015).Roots were weighed, harvested, frozen in liquid nitrogen, and stored at À80°C until further processing.Frozen roots were homogenized in DMSO (Sigma-Aldrich) containing 0.5 mM camphorsulfonic acid and 0.5 mM lidocaine (Sigma-Aldrich) as internal standards, using 1-mm zirconia beads as described earlier (Wolinska et al., 2021) and subjected to LC-MS analysis.

Metabolite profiling of root extracts and root exudates
The LC-MS system consisted of the UltiMate 3000 RS (Thermo Fisher Scientific, Waltham, MA, USA) hyphenated to a TIMS-TOF mass spectrometer (Bruker Daltonics, Hamburg, Germany).Chromatographic separation was carried out on a ACQUITY UPLC HSS T3 C18 column (Waters, Milford, MA, USA; 2.1 9 150 mm, 1.8 lm particle size) at 22°C with a mobile phase flow rate of 0.25 ml min À1 .The elution was conducted using water containing 0.1% of formic acid (Sigma-Aldrich; Solvent A) and acetonitrile (VWR Chemicals, Fontenay-sous-Bois, France) containing 0.01% of formic acid (Solvent B) in the following gradient: 0-5 min from 10% to 30% B, 5-12 min to 100% B, and 12-15 min maintained at 100% B. The mass spectrometer was operated using the following settings: ion source voltage of À4.5 or 4.5 kV, nebulization of nitrogen at a pressure of 2.2 bar, and a gas flow rate of 10 l min À1 .Ion source temperature was 220°C.The spectra were scanned in positive and negative ionization in fragmentation mode (ddMSMS) at a range of 95-1000 m/z at a resolution >30 000 FWHM (full width at half maximum).The spectrometer was calibrated with sodium formate salt clusters as internal calibrant before each analysis.The samples (5 ll) were injected into the inlet port after purging and rinsing the system.LC-MS system operation and data acquisition was supervised by Compass HyStar software (v.6.0;Bruker Daltonics).Obtained LC-MS data were processed for peak detection, deisotoping, alignment, and gap-filling by MS-DIAL v.4.74 (Tsugawa et al., 2015) separately for positive and negative ionization modes, and then, data from both modes were combined.The prepared data table was postprocessed for missing values imputation, log transformation, and data filtering by MetaboAnalyst (Pang et al., 2021).Then, data were visualized by sparse partial least squares discriminant analysis and Venn diagrams showing signals selected by one-way ANOVA with Benjamini-Hochberg correction with the following criteria: FDR ≤ 0.05, and |log fold change| ≥ 1.0, where fold change is calculated for specific comparison of cyp79b2b3 vs Col-0 and pyk10bglu21 vs Col-0.Glucosinolates and indole-3-acetamide were identified based on the MS/MS fragmentation spectra.Standard compounds were used for the identification of indole-3acetonitrile, camalexin, indole-3-carboxylic acid, raphanusamic acid, esculetin, fraxetin, and scopoletin (Sigma-Aldrich).Heights of respective molecular ion peaks were used for quantification of identified metabolites.In case of root extracts, peak heights of signals of interest were additionally normalized to the peak height of the internal standard.

Collection of root extracts
To collect root extracts, Col-0 WT and pyk10bglu21 and cyp79b2b3 mutant plants were cultured on half-strength MS media supplemented with 1% (w/v) sucrose and 1% (w/v) agar for 21 d under short-day conditions as described previously.Roots were weighed, harvested, frozen in liquid nitrogen, and stored at À80°C until further processing.Roots were homogenized in phosphobuffered saline using 1-mm zirconia beads to obtain 10 mg ml À1 of root extracts immediately before treatment.

Soil treatment with root exudates
Approximately 500 mg of CAS soil was transferred into 2-ml screw-cap tubes and treated with 50 ll of root exudates or root extracts.Tubes were covered by breathable tape and incubated at 28°C under dark conditions.Root exudates or freshly prepared root extracts were added every 3 d to replenish the weight loss due to the evaporation of the moisture content.Soils were freezedried and stored at À80°C until further processing.

Bacterial microbiota reconstitution experiment
Bacterial cells were first grown on plates containing 50% trypsin soy broth (TSB; 15 g l À1 ) supplemented with 1.5% (w/v) agar.Single colonies were then used to inoculate 50% TSB liquid media in 96-deep-well plates and cultured for 4 d at 25°C.Sixty microliters of cultures were transferred to another 96-well plate containing 600 ll of the sterile TSB media cultured for another 3 d in parallel to the original culture plates.Cultures from 7-d-old and 3-d-old plates were pooled, and optical absorbance at 595 nm was roughly normalized to each other using a plate reader (TECAN, M€ annedorf, Switzerland).Equal volumes of strains were then pooled into a new tube and centrifuged at 1800 g for 10 min to remove the TSB, followed by two washes in 10 mM MgCl 2. The obtained pellet was dissolved in MgCl 2 at the OD 600 of 0.15-0.25,followed by overnight incubation at 25°C, such that bacteria would have consumed nutrients carried over from previous cultures and utilize the root exudate as their sole nutrient source during the treatment.On the day of treatment, the bacterial cells were collected by centrifuge and resuspended in 1/10 volume of 10 mM MgCl 2 and mixed with root exudates from respective genotypes to obtain OD 600 of 0.05, a part of which was further 1 : 10 diluted with the same root exudates to obtain OD 600 of 0.005.These bacterial SynCom samples were distributed to sterile 96-well microbial culture plates (50 ll per well) and incubated at 25°C without agitation.At the time of harvest (24 or 72 h postinoculation; 24 or 72 hpi), a fixed amount of Escherichia coli DH5a cells (at final OD 600 of 0.01 or 0.001 depending on the initial inoculum titer), whose 16S rRNA is fully distinguishable from all strains included in our SynCom, was added to each sample to enable quantitative abundance (QA) estimation (Lundberg et al., 2021).The cultures were processed for DNA isolation immediately after adding the E. coli cells.

DNA extraction and amplicon sequencing
Total DNA from the rhizosphere, rhizoplane, and root samples was extracted from collected soil, filter paper, and surfacesterilized root samples, respectively, using the FastDNA SPIN Kit for Soil (MP Biomedicals, Solon, OH, USA), as described previously (Duran et al., 2018).For DNA extraction from bacterial SynCom, cells were lysed in alkaline sodium hydroxide and directly used as a PCR template, as described previously (Bai et al., 2015).The V5-V7 region of the bacterial 16S rRNA gene and the ITS1 region of fungal DNA were PCR-amplified by specific primer sets containing adapter sequences for sequencing and barcode sequences for multiplexing (Tables S3-S6).Approximately the same amounts of PCR product were pooled, purified twice using AMPure XP beads (Beckman Coulter Agencourt, Brea, CA, USA), and sequenced by an Illumina MiSeq platform (MiSeq Reagent Kit V3, 600-cycle).

Statistical analysis for microbiome profiling
All statistical analyses were performed in R (https://www.rproject.org/).Unconstrained and constrained principal coordinates analyses (PCoA and CPCoA) were performed based on Bray-Curtis dissimilarities using the cmdscale and capscale functions in the STATS and VEGAN packages.Differential abundance of ASVs and aggregated ASVs was conducted using the EDGER package by fitting relative abundance to a generalized linear model with a negative binomial distribution, controlling for sequencing batch, technical replicates, and the experimental batch as random factors.The quantitative abundance of bacteria in SynCom was calculated by dividing the read counts per strain or for the overall community by the read counts of E. coli.

Plant-fungi binary interaction assay
Fungal inoculation was performed as described previously (Frerigmann et al., 2021;Mesny et al., 2021).Approximately 50 mg of fungal mycelium was collected and homogenized in 10 mM MgCl 2 , which was then used to inoculate surface-sterilized seeds.Plants and fungi were co-cultured on half-strength MS media supplemented with 1% (w/v) agar for 21 d under short-day conditions, and their shoot fresh weights were measured after cultivation.

ER bodies and Trp-derived specialized metabolites together contribute to root microbiota assembly
To investigate the impact of ER bodies and Trp-derived specialized metabolites on root microbiota assembly, we prepared natural soils (Bulgarelli et al., 2012) and grew A. thaliana mutants impaired in ER body-resident myrosinases (pyk10bglu21), the formation of ER bodies (nai1-1), the biosynthesis of IGs as well as other Trp-derived specialized metabolites to a milder extent (myb34/51/122), and the entire specialized Trp metabolism (cyp79b2b3), along with the Col-0 WT plants (Fig. 1b), and performed amplicon sequencing analysis of bacterial (16S rRNA) and fungal (ITS1) microbiota community compositions of roots (rhizoplane and endosphere fractions) as well as rhizosphere and bulk soil fractions.A constrained principal coordinates analysis (cPCoA) based on Bray-Curtis dissimilarities between samples at the amplicon sequence variant (ASV) level (Fig. 1c) revealed significant differences between the plant genotypes in the bacterial communities in the rhizoplane (P = 0.009; 5.40% of variation explained by genotypes) and in the fungal communities in the endosphere (P = 0.021; 11.41%; Table S7).On the contrary, bacterial and fungal communities in the endosphere and rhizoplane, respectively, did not significantly differ across genotypes.Bacterial communities in the rhizoplanes of cyp79b2b3 and myb34/51/122 and of pyk10bglu21 and nai1 were similar to each other, respectively (Fig. 1c; top left), pointing to an active role of IGs and ER bodies in bacterial community assembly rather than a stochastic variation between different host genotypes.In fungal communities, only cyp79b2b3 and pyk10bglu21 mutants, in which enzyme-encoding genes were disrupted, showed substantial differences from the WT, while myb34/51/122 and nai1, the transcription factor mutants, exhibited a milder impact (Fig. 1c; bottom right).This may be explained by Trp-derived metabolites other than IGs accumulating to levels similar to Col-0 and/or residual amounts of IGs and PYK10 and BGLU21 in these mutant roots (Matsushima et al., 2004;Frerigmann et al., 2016).A similar trend of microbial community shifts in mutants compared with the WT was observed when we aggregated the relative abundance (RA) of ASVs at the family level (Fig. S1).We also observed that the root-associated microbial communities were not dominated by a small subset of ASVs (Fig. S2).These analyses showed that the observed community shift occurred at the family or higher taxonomic level rather than at the strain-or species-specific level.
We noted that the bacterial community shift observed in the rhizoplane of mutants compared with the WT was in the same direction along the first axis, while the ER body mutants and Trp metabolism mutants were separated along the second axis (Fig. 1c; top left).This implied that the loss of ER bodies and Trp-derived metabolites exerted a partly similar impact, if not entirely identical, on the root-associated microbiota structure.This was also supported by the significant positive correlation between the log 2 -scale fold changes (logFC) in RA values of ASVs or families between each mutant and the WT (Figs S3-S5).Overall, these data demonstrate that the ER body-resident myrosinases and Trp metabolism contribute to root microbial community assembly, affecting the behavior of distinct but overlapping sets of microbes.

ER body-resident myrosinases contribute to the secretion of Trp specialized metabolites, potentially produced from IGs
The significant bacterial community shift at the rhizoplane (root exterior) suggested a role for root-exuded compounds in bacterial community assembly.However, it remained unknown whether and how PYK10 and BGLU21 intracellularly stored in ER bodies could contribute to the extracellular secretion of metabolites into the rhizosphere.We aimed to test this directly by collecting root exudates from Col-0 WT and pyk10bglu21 and cyp79b2b3 mutant plants, axenically grown in a hydroponic culture system (Wippel et al., 2021) and analyzing it by untargeted HPLC-MS/ MS.This revealed a significant difference in the compositions of metabolites between exudates collected from mutants and the WT (Fig. 2a), indicating that ER body-resident PYK10 and BGLU21 are involved in the cellular metabolism, intracellularly and/or extracellularly, and influence the profile of metabolites ultimately secreted to the rhizosphere.We observed up to 696 Ó 2023 The Authors New Phytologist Ó 2023 New Phytologist Foundation New Phytologist (2023) www.newphytologist.comand 1307 signals with significant differences in their RA values in pyk10bglu21 and/or cyp79b2b3 root exudates compared with Col-0 exudates, among which 34 and 273 signals were commonly depleted/enriched in both mutant exudates (Fig. 2b).The overall difference in the metabolomic profiles was larger in pyk10bglu21 than in cyp79b2b3 (Fig. 2a), consistent with the broad range of substrates for ER-resident myrosinases, including but not limited to Trp-derived specialized metabolites (Matsushima et al., 2003b;Ahn et al., 2010;Nakano et al., 2017;Nakazaki et al., 2019;Yamada et al., 2020).
We then identified a wide range of glucosinolates and Trp specialized metabolites, along with coumarins, in our dataset based on standards or MS/MS fragmentation spectra (Figs 2c, S6a-d; Table S8).Among the tested compounds, we revealed that indole-3-carboxylic acid (I3CA) and camalexin were significantly different between genotypes.We also noted that the abundance of raphanusamic acid tends to be lower in pyk10bglu21 exudates compared with Col-0 exudates, albeit not statistically significant after correction for multiple comparisons (P = 0.068).On the contrary, we did not observe a significant difference in the abundance of IGs, indole-3-acetamide (IAM), and indole-3-acetonitrile (IAN).We noted that the abundance normalized to the total ion chromatogram of these metabolites is low and that even cyp79b2b3 exudates did not show any difference compared with Col-0 exudates, while these compounds were highly abundant in root extracts harvested from plants grown without light exposure (Fig. S6e).It was therefore inferred that these compounds are not substantially secreted to the rhizosphere in our experimental setup, and the peaks we observed were within the range of detection background.Notably, I3CA and camalexin are both exclusively produced from IAN, while IAN can be produced as a catabolic product of I3G hydrolysis by myrosinases as well as directly from indole-3-acetaldoxime (IAOx).Nitrile formation from IG aglycones is facilitated by a group of proteins called NITRILE SPECIFIER PROTEINS (NSPs), many of which are strongly co-expressed with PYK10 and other ER bodyrelated genes (Nakano et al., 2017).This, together with our results, suggests that PYK10-mediated I3G catabolism produces IAN, which is then fed into the biosynthetic pathways of I3CA and camalexin, although in planta camalexin production from I3G-derived IAN has not been characterized to date.Overall, this result demonstrates that Trp-derived specialized metabolites are Trp-derived metabolites secreted into the rhizosphere in a manner dependent on PYK10 and BGLU21 are capable of modulating microbiota community structure We then tested whether the observed difference in root exudation profiles would affect microbial community assembly processes by treating natural soils with these different root exudates and analyzing the resulting bacterial and fungal community structures (Fig. 3a; Table S9).Given the major variation between experimental batches (Fig. S6f,g), while the overall trend was similar, we decided to focus on one batch of exudate samples for further experiments (note that each batch contained three technical replicates, where plants were grown independently in different pots).We found that, after 15 d of consecutive treatments with 2-d intervals, the soil microbial communities exhibited a shift compared with those from mock-treated soils, which is consistent with a previous study (Badri et al., 2013), in a manner dependent on the genotypes of origin of the root exudates (Fig. 3a; middle panels).Both bacterial and fungal communities in the soils treated with pyk10bglu21 or cyp79b2b3 exudates were significantly different from each other and from the communities in the soils treated with Col-0 exudates.The logFC in their RA values relative to the WT controls was positively correlated between pyk10bglu21 and Col-0 and cyp79b2b3 and Col-0, at both the ASV and the family levels (Fig. 3a; right panel), suggesting the presence of a group of microbes that are targeted by the metabolites whose secretion to the rhizosphere is dependent of Trp specialized metabolism and ER body-resident myrosinases.
Next, to test whether PYK10 and BLGLU21 also contribute to the biosynthesis or accumulation in roots or only to the secretion of the bioactive metabolites responsible for the observed community shifts or also their accumulation in roots, we prepared axenic root crude extracts (homogenates) from these genotypes and performed the same soil treatment experiment (Fig. 3b; Table S9).The root extract treatments also triggered significant community shifts in the soil microbial community compared with mock-treated soils (Fig. 3b; middle panels).Interestingly, the cPCoA plots showed that Col-0 and pyk10bglu21 root extracts triggered similar shifts in the microbial community compositions, while treatments with cyp79b2b3 root extracts resulted in significantly different community structures.Comparison of logFC in RA values of ASVs and families also pointed to the presence of a group of microbes that were specifically affected by the CYP79B2 and CYP79B3 pathway but not by the PYK10 and BGLU21 pathway (Fig. 3b, right panel).These findings suggest that the Trp specialized metabolites in Col-0 root extracts that were responsible for the observed community assembly are not secreted from pyk10bglu21 roots but they (or their glycosylated precursors) accumulate in roots.A. thaliana roots accumulate many myrosinases that are not stored in ER bodies, such as TGG4, TGG5, and PEN2 (Millet et al., 2010;Fu , 2016), which are capable of catalyzing the same reaction as PYK10 and BGLU21 when subcellular membrane partitions are artificially disrupted in a homogenate.Therefore, the active compounds may be accumulated in pyk10bglu21 roots either in an already active form or in an inactive glycosylated form.Overall, these results suggest a role for Trp-derived metabolites whose secretion but not accumulation depends on ER body myrosinases in modulating microbial community structures.It needs to be noted that, besides the microbes that are commonly affected in pyk10b-glu21 and cyp79b2b3 roots, root exudates, and root extracts, these mutants also triggered a community shift unique to each genotype (note the separation of these genotypes in CPCoA plots).This, along with the metabolomic profiling data of root exudates, points to the presence of PYK10 and BGLU21 substrates that are derived from other precursors than Trp, which are also capable of manipulating microbial community compositions.

Root-exuded compounds directly affect bacterial community structure independently of fungi
Trp-derived metabolites, including IGs, have been reported to be crucial for interaction with endophytic fungi (Hiruma et al., 2016;Wolinska et al., 2021), while their impact on bacteria is less understood.To assess the direct impact of root exudates on bacterial community assembly, we performed a synthetic community (SynCom) profiling experiment in these root exudates (Fig. 4a), using 171 bacterial strains isolated from healthy A. thaliana roots grown in natural soil and 29 strains isolated from the same, unplanted soil (Bai et al., 2015).The analysis of b-diversity based on RA values revealed that the SynCom compositions in the Col-0 root exudates were significantly different from the composition in the mutant root exudates (Fig. 4b; Table S10).The effect on bacterial community composition was more different between Col-0 and cyp79b2b3 exudates than between Col-0 and pyk10bglu21 exudates, which is similar to what was observed in the soil treatment experiment (Fig. 3).The community shift in mutants compared with Col-0 was significant with both initial titers (OD 600 of 0.05 and 0.005) and at both time points (24 and 72 hpi).The overall variance explained by genotypes was larger when we used a lower titer as a starting inoculum, and we observed a higher level of overall community growth (increase in the total number of bacterial cells within the community measured by the quantitative abundance normalized to E. coli cells spiked in before DNA extraction) under the lowtiter than high-titer conditions, while between genotypes, the overall community growth was largely similar (Fig. S7a).When

Research
New Phytologist we compared the community structure dynamics over time within genotypes, we observed a more dynamic community assembly at 24 hpi when we used low-titer inoculum, with eventual convergence on a similar community structure at 72 hpi (Fig. S7b).These data suggest that the contribution of ER bodies and Trp-derived specialized metabolites to community assembly is larger when bacterial cells are metabolically more active and undergoing dynamic community re-assembly.We compared the growth of each bacterial strain in Col-0 and mutant exudates (Fig. S7c) and found a small group of strains that showed a similar response in mutant exudates in comparison with Col-0 exudates (Fig. S7d), pointing to a set of bacteria commonly targeted by these pathways via root-exuded compounds.Overall, these results suggest that Trp-derived metabolites secreted to the rhizosphere in a PYK10 and BGLU21-dependent manner can directly impact the bacterial community assembly in the absence of eukaryotic organisms, such as plants and fungi.

ER bodies and Trp-derived metabolites directly influence plant-fungus interactions
Lastly, we tested whether the ER body pathway and the Trp metabolic pathway have a direct impact also on the plant-fungus interaction in the absence of bacteria.Toward this end, we inoculated the same set of plants (Col-0, pyk10bglu21, and cyp79b2b3) in a mono-association setup (Frerigmann et al., 2021;Fig. 5) with fungal strains isolated from roots or leaves of healthy A. thaliana and related species grown in natural soils (Hiruma et al., 2016;Mesny et al., 2021).We found that the growth of cyp79b2b3 mutant plants was severely impaired by more than half of the strains compared with the WT plants (24 isolates), especially by those belonging to the classes Leotiomycetes and Dothideomycetes.Furthermore, seven strains were found to significantly impair the growth of both cyp79b2b3 and pyk10bglu21 mutants compared with Col-0.On the contrary, 16 isolates, including all those belonging to the class Leotiomycetes, specifically restricted the growth of cyp79b2b3 but not pyk10bglu21 plants, and three strains showed a negative impact only on pyk10bglu21 compared with the Col-0 WT.These results demonstrate that the myrosinases stored in ER bodies and Trp-derived metabolites directly regulate plantfungus interactions and target distinct but overlapping sets of fungal strains.Combined with the results that the ER body and Trpderived metabolic pathways together contribute to the bacterial and fungal community assembly at the rhizoplane and in the endosphere, respectively, our overall findings illustrate a role for ER body-resident myrosinase-mediated Trp metabolism in rootmicrobiota interactions.

Discussion
The physiological role of root ER bodies in root-microbiota interactions via root-secreted metabolites ER bodies in A. thaliana are abundant both in leaves (inducible ER bodies) and roots (constitutive ER bodies).While the role of leaf ER bodies in interaction with herbivorous animals and pathogenic bacteria has been characterized, the role of root ER bodies remained unclear.Here, we provide clear evidence that root ER bodies and the myrosinases therein are involved in rootmicrobiota interactions, whose processes are at least partially mediated by root-secreted Trp specialized metabolites.In the early era of ER body/dilated cisternae research, a functional link between ER bodies and glucosinolates was already proposed, given their co-occurrence in a group of families within the order Brassicales (Iversen, 1970), which was further supported by the recent discovery of myrosinase activity in PYK10 and BGLU18 (Nakano et al., 2014(Nakano et al., , 2017;;Nakazaki et al., 2019).This also explains the role of leaf ER bodies in antiherbivorous defense (Nakazaki et al., 2019;Yamada et al., 2020).Overall, our findings are in line with previous predictions that the physiological functions of ER bodies are associated with glucosinolates.Furthermore, we clearly demonstrated that ER body-resident myrosinases are involved in the secretion of metabolites into the rhizosphere, whose underlying mechanisms remain to be addressed.A plausible scenario is that the deglycosylation of intracellularly accumulated metabolites by PYK10 and BGLU21 stimulates the secretion of their aglycons to the apoplast.This possibility has been reported for scopolin, whose deglycosylation by a cytosolic enzyme BGLU42 is necessary for the rhizospheric secretion of its aglycon, scopoletin (Stringlis et al., 2018).Alternatively, it also remains possible that a part of PYK10 and BGLU21 proteins are secreted to the apoplast or deposited to the rhizoplane through the secretory pathway from the ER, thereby metabolizing the glucosides secreted from root tissues.
Potential in planta substrates of PYK10 and BGLU21 responsible for the microbiota assembly Trp-derived specialized metabolites are widely crucial for interactions with microbes (Pedras et al., 2011;Pastorczyk & Bednarek, 2016), including root microbiota (Wolinska et al., 2021), and our findings suggest that a part of the physiological substrates of ER body-resident myrosinases are derived from Trp. Partially similar community shift observed in myb34/51/122 and pyk10b-glu21 suggests IGs as potential in planta substrates of PYK10 and BGLU21, which is consistent with the transcriptional coregulation of these pathways (Nakano et al., 2017).On the contrary, we did not detect a striking difference in the abundance of intact glucosinolates or its well-characterized terminal product IAN in the root-secreted metabolome.In contrast, we observed a reduction in the relative abundance of other Trp-derived metabolites, including camalexin and I3CA.Both metabolites can be produced from IAN, which is one of the terminal products of I3G catabolism, implying the possibility that PYK10-mediated I3G hydrolysis feeds the precursor for Trp metabolism apart from IG biosynthesis (Fig. 2).On the contrary, PYK10 and other ER body-related genes are strongly co-expressed with CYP81F4 (Nakano et al., 2017), encoding an enzyme to produce 1methoxy I3G (1MI3G) from I3G, and 1MI3G is the most abundant form of IGs in the roots of Col-0 plants (Brown et al., 2003).As methoxylated IAN cannot be a precursor for I3CA, and camalexin, it is inferred that there are other catabolic Ó 2023 The Authors New Phytologist Ó 2023 New Phytologist Foundation New Phytologist (2023) www.newphytologist.comproducts than those potentially fed into I3CA/camalexin pathways, which may also be responsible for the root-associated microbiota assembly.Importantly, however, it needs to be noted that the compounds we identified by standards or MS/MS spectra are a small minor proportion of root exudate metabolome, and there are yet many other unidentified metabolites whose abundance appeared to be dependent on PYK10 and BGLU21, and/or CYP79B2 and CYP79B3, which might play a crucial role in root microbiota assembly.Moreover, albeit a partial similarity, mutants impaired in Trp metabolism and mutants impaired in ER body system accommodate substantially distinctive microbial communities in their roots, demonstrating a role of pathwayspecific impact on microbiota assembly, which can be explained by the fact that PYK10 is also capable of hydrolyzing nonglucosinolate glucosides (Ahn et al., 2010).A future study is needed to specify the identity of metabolites responsible for the observed community shift and to reveal its metabolic pathway(s).
Camalexin operates as an antimicrobial compound and affects microbial behavior directly (Bednarek et al., 2005), while I3CA has been shown to mediate b-aminobutyric acid-triggered priming of immune responses and to covalently bind to cell wall oligosaccharides upon infection in A. thaliana (Hagemeier et al., 2001;Tan et al., 2004;Gamir et al., 2012Gamir et al., , 2018)).The role of root-produced camalexin in regulating root-commensal interaction has also been reported (Koprivova et al., 2019(Koprivova et al., , 2023)), which is in line with the role of specialized metabolites in interaction with microbes in a wide range of plant species (Singh et al., 2022).Interestingly, camalexin exudation appears to play a positive role in accommodating a plant growth-promoting bacterial commensal CH267 (Koprivova et al., 2019(Koprivova et al., , 2023)), illustrating its function not only as an antimicrobial but also for recruiting beneficial microbes.Therefore, it is possible that camalexin and I3CA produced from IAN via IG metabolism directly influence root microbiota assembly.On the contrary, however, it needs to be noted that camalexin and I3CA are among the variety of metabolites whose relative abundance in root exudates is altered in pyk10bglu21 and cyp79b2b3 mutants (Fig. 2a), and the characteristics of other Trp metabolites whose secretion depends on ER body-resident myrosinases are yet to be identified.Further studies are needed to address which of these catabolic products are crucial for the interaction with root microbiota and to disentangle whether PYK10 and BGLU21 only contribute to the secretion of these metabolites or also to their biosynthesis.
It is also important to consider the fact that the exudates used in our experiments were collected from axenic, na€ ıve plants, that is, in the absence of microbes or any immune elicitors.It has been well described that plant immune status has an impact on root metabolomic profiles (Millet et al., 2010;Koprivova et al., 2019) as well as on root exudation profiles (Bednarek et al., 2005;Badri & Vivanco, 2009;Koprivova et al., 2020Koprivova et al., , 2023)).Therefore, it is likely that the metabolomic profile of root exudates may be substantially different between axenic and soil conditions.It also needs to be noted that the root exudation profiles were significantly distinct between the two biological replicates performed in this study.In order to avoid overly stochastic microbial behavior triggered by the metabolic fluctuation, we decided to focus on one biological replicate and rather repeated the soil and SynCom treatments multiple independent times to support reproducibility.Our approach has provided meaningful insight into how microbial community assembly is affected by metabolites that were differentially abundant in these exudate samples, while an independent experiment using independent exudate samples might trigger quantitatively different microbial responses.Moreover, root exudation is one of many aspects of root physiology, and our soil treatment experiments may not necessarily recapitulate the entire interaction occurring at the root-soil interface.Nonetheless, given that the three independent approaches employed in this study all suggested an overlapping role of Trpderived metabolites and ER body-resident myrosinases, we propose a model according to which Trp-derived metabolites that are secreted in a manner dependent on PYK10 and BGLU21 under axenic na€ ıve conditions are capable of manipulating a microbial community.Further studies will address to which extent our insights into the activity of axenic root exudates can be generalized to explain the actual microbial community shift observed in the glasshouse experiment.

Fig. 1
Fig. 1 Endoplasmic reticulum (ER) bodies and Trp metabolism play a role in root-associated microbiota assembly.(a) Roots of Arabidopsis thaliana developing a large amount of ER bodies, visualized by ER-localized GFP (SP-GFP-HDEL).Cell wall was stained by propidium iodide.The bar corresponds to 50 lm.(b) Representation of IG biosynthetic and catabolic pathways.Arrows and dotted arrows indicate metabolic reactions and transcriptional regulation, respectively.(c) Constrained principal coordinates analysis (cPCoA) of the bacterial and fungal community structures in the roots of Col-0 as well as mutants impaired in ER body formation (nai1-1), ER body-resident myrosinases (pyk10bglu21), IG biosynthesis (myb34/51/122) and Trp metabolism (cyp79b2b3) based on Bray-Curtis dissimilarities.Ordination was constrained by genotypes and conditions by soil batches, biological replicates, and sequencing runs.Colors and shapes represent the genotypes and biological replicates, respectively.Variation explained by genotypes based on permutational analyses of variance (PERMANOVA; n = 1000) is indicated at the top right, and asterisks indicate its statistical significance (a = 0.05).GST, glutathione-S-transferase; IAOx, indole-3-acetaldoxime; IG, indole glucosinolates; SOT, sulfotransferase; Trp, tryptophan; UGT, uridine diphosphateglycosyltransferase.

Fig. 2
Fig. 2 Root secretion of Trp-derived metabolites is partly dependent on PYK10 and BGLU21.(a) Sparse partial least squares discriminant analysis plot of profiles of signals corresponding to root-exuded metabolites.Colors and shapes represent the genotypes and biological replicates, respectively.(b) Venn diagrams representing the differentially abundant MS signals detected in cyp79b2b3 and pyk10bglu21 mutant media compared with Col-0 for each biological replicate (moderated t-test; false discovery rate ≤ 0.05 and |log 2 fold change| ≥ 1.0).(c) The abundance of indole-3-carboxylic acid (I3CA), camalexin, and raphanusamic acid, normalized to respective Col-0 samples, are shown as box plots.A schematic overview of the Trp metabolic pathway is also shown.Blue letters indicate the metabolites whose abundances are shown in the boxplots, and blue arrows indicate possible paths through which these metabolites can be affected by PYK10 and BGLU21.Letters indicate statistical significance corresponding to the analysis of variance (ANOVA) and post hoc Tukey's honestly significant difference (HSD) tests (a = 0.05).Marginal P values are shown with numbers.Median, interquartile range (IQR), and data points within AE1.5*IQR are indicated by horizontal lines, boxes, and whiskers, respectively, along with the raw data points whose shapes correspond to individual experiments.

Fig. 3
Fig. 3 Compounds secreted into the rhizosphere in a PYK10-and BGLU21-dependent manner have an impact on microbial community assembly in the soil environment.Soils were treated with root exudates (a) or root extracts (b) consecutively for 15 d with 2-d intervals.Constrained principal coordinates analysis (cPCoA) plots are based on Bray-Curtis dissimilarities of bacterial (left) and fungal (right) communities, and the ordination was constrained by genotypes and conditioned by biological and technical replicates.The scatter plots show a comparison of changes in the relative abundance of amplicon sequence variants (ASVs; magenta) or families (green) in soils treated with mutant root exudates/extracts compared with soils treated with Col-0 root exudates/extracts.Size corresponds to their mean relative abundance in soils treated with Col-0 root exudates/extracts.Variation explained by genotypes based on permutational analysis of variance (PERMANOVA; n = 1000) is indicated at top right, and asterisks indicate its statistical significance (a = 0.05).

Fig. 4
Fig. 4 Compounds secreted into the rhizosphere in a PYK10-and BGLU21-dependent manner have a direct impact on bacterial community assembly.(a) Schematic representation of experiments using a bacterial SynCom composed of 171 root-derived and 29 soil-derived isolates.(b) Constrained principal coordinates analysis (cPCoA) analysis of bacterial communities in root exudates from Col-0 (black), pyk10bglu21 (blue) and cyp79b2b3 (red) at 24 and 72 h postinoculation (hpi) using Bray-Curtis dissimilarities based on the relative abundance.Shapes correspond to the individual replicates of root exudates.Ordination was constrained by genotypes and conditioned by biological and technical replicates.Variance explained by genotypes based on the permutational analysis of variance (PERMANOVA; n = 1000) is indicated at top right, and asterisks indicate its statistical significance (a = 0.05).

Fig. 5
Fig. 5 ER bodies and Trp metabolism are needed for an appropriate accommodation of endophytic fungi.(a) Schematic representation of experimental setup using a set of endophytic fungi isolated from healthy plant roots.(b) Representative images of plants inoculated with fungi along with axenic control plants.(c) A phylogenetic tree of fungi used in this study along with their impact on Col-0 growth compared with the axenic plants (left panel) and the impact of pyk10bglu21 and cyp79b2b3 mutations on shoot growth with respect to Col-0 plants inoculated with the same fungal strain.Marked with bold lines are false discovery rate ≤ 0.05 and |log 2 fold changes| ≥ 0.5.