Streptomyces cocklensis DSM 42063 and Actinacidiphila bryophytorum DSM 42138 Colonize Arabidopsis thaliana and Modulate Its Proteome

Plants host a vast array of microorganisms that collectively form a microbiota. The microbiota and the host plant are considered an ecological unit called a holobiont, in which microorganisms interact with their host and between them in a regulated functional network (Hassani et al. 2018; Uroz et al. 2019). The microbiota fulfill functions that are important for the health and development of the host and the homeostasis of the holobiont. A well-described beneficial effect of certain microorganisms interacting with plants is the so-called plant growth-promoting (PGP) effect (Backer et al. 2018; Berg et al. 2020; Gaiero et al. 2013; Vacheron et al. 2013). The mode of action of PGP microorganisms (PGPMs) is a hot research topic, partly because implementing PGPMs in agriculture may help reduce the use of agrochemicals. PGPMs exert their effects through direct and indirect mechanisms. Directly, free-living or symbiotic nitrogen-fixing bacteria, phosphorus-solubilizing microorganisms, and hydrogen cyanide-, organic acids-, and siderophore-producing bacteria increase the availability of nutrients for the plant. Indirectly, PGPMs act on plant growth by producing metabolites or chemical signals that modulate the host hormone balance, inhibit pathogen attacks, or promote tolerance to abiotic stresses (e.g., drought, heat, or salinity). In particular, PGP bacteria (PGPB) contribute to phytohormone homeostasis with the synthesis of auxin (indol acetic acid IAA) and the degradation of the ethylene precursor 1-aminocyclopropane-1-carboxylate (ACC) by the enzyme ACC-deaminase to reduce the production of the stress hormone ethylene in the host (Afzal et al. 2019). Other metabolites or signals, such as polyamines, volatile organic compounds (VOCs), or acyl-homoserine lactones (AHL) are produced by PGPB to improve stress tolerance and stimulate growth of the host. In addition, these compounds limit the proliferation of pathogens and therefore produce a beneficial effect on plant fitness (Hassani et al. 2018). More generally, the beneficial effect of microbiota on plant resistance to biotic or abiotic stresses originates first from its ability to produce hydrogen cyanide, VOC, AHL, antibiotics, or cell-wall-degrading enzymes; second from the negative interaction of its members with pathogens by competing for nutrients or by regulating the bacterial quorum sensing-mediated responses; and finally from its ability to modulate phytohormone signaling pathways and trigger the induced systemic response or the systemic acquired resistance (Vlot et al. 2021).

Beneficial microorganisms interacting with host plants may be specific to one or a few plant genera, species, or genotypes, whereas others may be ubiquitously present as members of the core microbiota (Bai et al. 2015; Schlaeppi et al. 2014). Actinomycetes are found ubiquitously in plant microbiota. Among this phylum, Streptomycetaceae is the most abundant taxon (Bai et al. 2015; Bulgarelli et al. 2013; Lundberg et al. 2012; Schlaeppi et al. 2014). Streptomycetaceae form an important group of filamentous soil bacteria split into several genera, as shown by recent phylogenomic analysis (Labeda et al. 2017; Madhaiyan et al. 2022). These bacteria produce mycelium, allowing for their efficient colonization of the root surface or internal tissues as endophytes. Streptomycetaceae PGPs synthetize plant growth regulators and VOCs. They also produce compounds exhibiting antifungal or antibacterial activities and excrete enzyme-degrading polymers such as cellulose and hemicellulose, xylan, or chitin (Świecimska et al. 2022). Plant-Streptomycetaceae interactions have recently attracted significant research efforts to elucidate the yet unknown molecular mechanisms.

Plants regulate the composition of their microbiota by mechanisms controlling recruitment and management of beneficial microorganisms (Jones et al. 2019; Vannier et al. 2019). One of these mechanisms is the plant innate immune system, which affects host colonization by microorganisms (Hacquard et al. 2017; Vlot et al. 2021). Another well-described mechanism for the selection of microorganisms by a host plant is the production of primary and secondary (also called specialized) metabolites in root exudates. These metabolites that attract beneficial microorganisms and stimulate their growth may also act as defense compounds against pathogens or as specific signaling molecules to favor the colonization of host organs by specific beneficial microorganisms. Recently, the so-called “cry-for-help” mechanism has been proposed, according to which beneficial bacteria are specifically recruited when plants are faced with pathogen attacks or abiotic stresses (Rizaludin et al. 2021).

In a previous work, we reported the effect of an impaired isoprenoid metabolism on Arabidopsis thaliana-bacteria interactions and microbiota selection (Graindorge et al. 2022). The microbiota of the wild-type genotype was compared with that of a mutant called chs5 (chilling sensitive 5) that had a D627N missense mutation in the enzyme 1-deoxy-D-xylulose-5 phosphate synthase DXS1/CLA1/CHS5 (Araki et al. 2000; Schneider et al. 1995; Wright et al. 2014). This is a limiting enzyme in the plastidial isoprenoid biosynthesis pathway (Rodríguez-Concepción and Boronat 2015; Rohmer 1999). This chs5 mutant displays moderate phenotypic alterations at 22°C and a chlorotic phenotype when temperature is shifted to 15°C (Araki et al. 2000). In addition, untargeted metabolomic analyses revealed changes in the lipid and phenylpropanoid (flavonoids, cinnamates, coumarins) profiles in addition to deficiencies in carotenoids and chlorophylls (Araki et al. 2000; Graindorge et al. 2022; Schneider et al. 1995; Wright et al. 2014). Interestingly, the colonization of several Streptomycetaceae affiliated with Actinacidiphila bryophytorum (previously named Streptomyces bryophytorum) and Streptomyces cocklensis was reduced in the chs5 mutant regardless of growth conditions and plant developmental stages. Consequently, this study (Graindorge et al. 2022) suggested that the metabolic status of A. thaliana was crucial for the specific recruitment of Streptomycetaceae into the microbiota.

The aim of the work presented here was to study the A. thaliana-Streptomycetaceae interaction in vivo using A. bryophytorum DSM 42138 and S. cocklensis DSM42063, which are strains previously isolated from moss and soil, respectively (Kim et al. 2012; Li et al. 2016). To search for genes involved in plant-bacteria interaction, the genomes of both bacteria were sequenced and compared. To study the effects of these two bacteria on their host, we compared the proteomes of inoculated and non-inoculated wild-type or chs5 plants. These data, combined with growth phenotypes of inoculated seedlings and microscopic observations of the bacteria in planta, provide evidence that these bacteria specifically modulate the physiology of the A. thaliana chs5 mutant.

Materials and Methods

A. thaliana axenic plant growth and inoculation with Streptomycetaceae spp.

S. cocklensis DSM 42063 and A. bryophytorum DSM 42138 were cultivated at 28°C on a solid optimized medium for Streptomyces species to obtain colonies of the appropriate size (>2 mm) (“Streptomyces medium,” cat. 85883, Sigma-Aldrich, Saint-Louis, MO, U.S.A.; the composition is not given by the manufacturer). Colonies scraped with a sterile scraper were resuspended in 3 ml of 10 mM MgCl_2. The suspension was centrifuged at 2,500 × g for 10 min (repeated twice). For seed inoculation, the bacterial suspension in 10 mM MgCl₂ was adjusted to 1 × 10⁶ to 5 × 10⁶ cells/ml.

Seeds of A. thaliana ecotype Colombia-0 (Col-0) and the chs5 mutant of the same genetic background were surface sterilized in 70% ethanol for 2 min, then washed in sterile milliQ H₂O for 1 min. The supernatant was removed, and the seeds were treated for 5 min with a commercial sodium hypochlorite solution (4%) supplemented with 0.1% Tween 20 (Sigma-Aldrich). Seeds were washed eight times with sterile milliQ H₂O and then air-dried. Dried seeds were placed individually with a sterile toothpick on the surface of solid sterile MS medium plates (MS, 4.3 g/liter of Murashige & Skoog Medium M0221; Duchefa Biochemie, Haarlem, the Netherlands), 10 g/liter of sucrose (Euromedex, Strasbourg, France), 100 mg/liter of myo-inositol (Duchefa Biochemie), 1 mg/liter of thiamine HCl (Duchefa Biochemie), 0.5 mg/liter of pyridoxine HCl (Duchefa Biochemie), 0.5 mg/liter of nicotinic acid (Duchefa Biochemie), and 7 g/liter of agar (Sigma-Aldrich, pH 5.8). Plates with seeds were kept for 2 days at 4°C in the dark before inoculation with S. cocklensis DSM 42063 or A. bryophytorum DSM 42138. For inoculation, each seed was covered with 5 to 7 μl (5 × 10³ to 25 × 10³ cells) of bacterial suspension or sterile 10 mM MgCl₂ as a negative control. Plates were then incubated in a Sanyo MLR-351 incubator (Etten-Leur, the Netherlands) under a 16 h of light (160 μmol photon s⁻¹ m⁻²) at 18°C and 8 h of light (100 μmol photon s⁻¹ m⁻²) at 16°C regime. Seedlings of equivalent size (two to four leaves) were transferred onto fresh MS plates after 14 days of culture. Seedlings were photographed, and their phenotype was scored 24 or 31 days after inoculation (dai) using Image J (segmentation based on the “Excessive Green” method described previously; Woebbecke et al. 1995) to calculate the total leaf area and to measure the intensity of the green color of leaves.

Extraction and quantification of chlorophylls and carotenoids

Seedlings (three replicates of 15 seedlings each) were weighed and crushed in liquid nitrogen (with mortar and pestle). Extracts were prepared by incubating seedling materials at −20°C for 24 h in 80% acetone (Sigma-Aldrich) in water (vol/vol) as previously described (Graindorge et al. 2022). Supernatants (200 μl) were transferred to 96-well microplates (PS, U-bottom, MICROLON, Greiner Bio-one, Frickenhausen, Germany). Optical density was measured for each well (in triplicate) at 470, 646, and 663 nm on a FLUOstar Omega spectrometer (BMG Labtech, Ortenberg, Germany). The concentration of chlorophylls (c_a, c_b) and carotenoids (c_(x+c)) in samples was calculated with established formulas (Lichtenthaler and Buschmann 2001):

where c_a is the concentration of chlorophyll a; c_b is the concentration of chlorophyll b; and c_(x+c) is the concentration of xanthophylls and carotenes. The concentrations of chlorophylls and carotenoids were expressed in μg/mg dry weight of plant samples.

Microbiota profiling

Seedlings (three replicates of 15 seedlings each per condition) inoculated with Streptomycetaceae spp. or mock inoculated (24 days postinoculation) were gently removed from the agar plate and crushed in liquid nitrogen (with mortar and pestle) and stored at −80°C for further DNA extraction. DNA was extracted from 40 mg of frozen material using the PowerPlant DNA Isolation kit (MO BIO Laboratories, Carlsbad, CA, U.S.A.). The DNA concentration and quality were estimated by measuring the OD at 260 and 280 nm using a NanoDropTM 2000 spectrophotometer (Thermo Fisher Scientific, Waltham, MA, U.S.A.). The 16S rRNA-encoding gene was PCR-amplified from 12.5 ng of template DNA as described previously (Graindorge et al. 2022) using a mix of primers targeting the V5-V6 region (Supplementary Table S1). Libraries were constructed according to the 16S Metagenomic Sequencing Library Preparation protocol (Illumina Part # 15044223 Rev. B, https://support.illumina.com/documents/documentation/chemistry_documentation/16s/16s-metagenomic-library-prep-guide-15044223-b.pdf) except for some modifications as described previously (Graindorge et al. 2022). After sequencing, bioinformatic processing was performed using the FROGS 3.1 pipeline (Escudié et al. 2018). Briefly, it started with preprocessing of the sequencing read data with “VSEARCH” (to delete PCR duplicates, too long or too short reads). Then, quality sequences were clustered to operational taxonomic units (OTUs, >97% sequence similarity) with “Swarm.” Chimeric OTU sequences were removed using “VSEARCH.” Filtering was performed to keep OTUs present in at least three samples with a minimal coverage of five sequences and to suppress contaminants (phiX). Quantification was performed with the Phyloseq package version 1.30.0 (sourced from Migale Bioinformatics Facility, Jouy-en-Josas, France) (McMurdie and Holmes 2013).

Genome sequencing and mining

A. bryophytorum DSM 42138 and S. cocklensis DSM 42063 genomic DNA were prepared using the Wizard genomic kit (Promega, Madison, WI, U.S.A.). Genomic DNA sequencing was performed de novo on an Illumina instrument (Illumina, San Diego, California, U.S.A.) as follows. Libraries were prepared from 100 ng of genomic DNA with the Nextera Flex kit (Illumina). Library quality was assessed on an Agilent Bioanalyzer (Agilent, Agilent Technologies, Santa Clara, CA, U.S.A.) using a High Sensitivity DNA chip. These libraries were then loaded onto a MiSeq sequencing flowcell and sequenced in 2 × 150 paired ends mode, using a MiSeq V2 reagent kit. The 2.4 × 10⁶ trimmed reads produced 87.5 and 102.3× coverage for DSM 42138 and DSM 42063, respectively. The quality of the data was checked with FastQC version 0.11.8 (Andrews 2010). The PCR duplicates were removed with fastuniq version 1.1 (Xu et al. 2012). This step excluded the paired reads that were identical to keep only one. The low-quality and adapter sequences were trimmed with trimgalore version 0.5.0 (https://www.bioinformatics.babraham.ac.uk/projects/trim_galore/, based on cutadapt version 1.18; Martin 2011). The SPAdes software was used for short read assembly (v3.13.3) (Bankevich et al. 2012; Gurevich et al. 2013), with k-mers of 21, 33, 55, 77, and 99. Probable contaminant scaffolds were identified (and excluded) by comparing the obtained scaffolds (>500 nt) against the Microbial Genome Database (http://mbgd.genome.ad.jp/), which contains 6,318 genomes (2,547 species, 1,019 genera) including 5,861 bacteria, 254 archaea, and 203 eukaryota (last update 2018/09/20) using the megablast tool of the blast+ suite, version 2.9.0+.

A second sequencing was performed in the case of A. bryophytorum using the MinION technology (Oxford Nanopore, Oxford, U.K.) to obtain a better assembly quality. Libraries were prepared with 4 μg of gDNA using the Nanopore SQKLSK109 Ligation Sequencing kit and multiplexing with Nanopore EXP-NBD104 Native Barcoding Expansion kit (Oxford Nanopore). Sequencing was performed with this DNA library (21.88 fmol) on a Minion device with a R9.4.1 flowcell, for 48 h as run time and real-time Guppy fast basecalling. The different fastq files produced by the MinKNOW software (Oxford Nanopore) were fused with the Illumina reads to obtain one fastq file. A hybrid assembly was performed by combining short and long reads. The data were processed with the following steps to perform an assembly with high-quality reads: (i) PCR duplicates for Illumina paired reads were removed; (ii) remaining adapters from both sequences were trimmed; (iii) low-quality reads were trimmed (Q < 30 for Illumina; Q < 10 for Nanopore); and (iv) only paired-end reads with length greater than 100 nt for Illumina or reads with length greater than 1,000 nt for Nanopore were considered. For Nanopore reads, the quality of the data was checked with NanoPlot version 0.32.1 (https://pypi.org/project/NanoPlot/). First, the adapters were trimmed with Porechop version 0.2.4 (https://github.com/rrwick/Porechop), and then a read selection was performed with FiltLong version 0.2.0 (https://github.com/rrwick/Filtlong, length greater than 1,000 nt/best quality to have a coverage of 100×. (v) Contamination and taxonomic affiliation were checked by screening the library (fastq files) against the standard Kraken database (https://ccb.jhu.edu/software/kraken/) containing RefSeq genomes (archaea, bacteria, virus, human; last use 14/09/2020; https://benlangmead.github.io/aws-indexes/k2), using the default parameters, and UniVec (the NCBI vector sequence databank, https://www.ncbi.nlm.nih.gov/tools/vecscreen/univec/) (Langmead and Salzberg 2012). The completeness was evaluated using CheckM (Parks et al. 2015). The classification of the genome was performed with Kraken2 version 2.0.9beta (Wood et al. 2019). Two programs were used for de novo hybrid assembly: Unicycler (v0.4.8) (https://github.com/rrwick/Unicycler) (Wick et al. 2017) and SPAdes (v3.15.2) (Bankevich et al. 2012). Different k-mer sizes to assemble genomes with SPAdes (k-mers of 21, 33, 55, 77, and 99) were chosen. Unicycler is an assembly pipeline for bacterial genomes, based on SPAdes. Both tools compile best contigs and scaffolds for all the k-mers. The genomes analyzed for this study were integrated into the MicroScope platform (https://mage.genoscope.cns.fr/microscope/home/index.php) and analyzed with the tools available on this platform (Vallenet et al. 2006, 2009, 2013, 2017, 2020). The genomic similarity was estimated with 187 Streptomycetaceae genomes available on the MaGe platform (https://mage.genoscope.cns.fr/microscope/, accessed on 20/08/2021) using Mash and correlated to the average nucleotide index (Konstantinidis and Tiedje 2005; Ondov et al. 2016). Genes previously described by several authors (Afzal et al. 2019; Levy et al. 2017; Pinski et al. 2019) as genes allowing bacteria to associate with plants were searched in the genome of S. cocklensis and A. bryophytorum using the “search by keywords” option of the MaGe platform. We compared the metabolic capacities of these bacteria using the “Metabolic Profiles” tool of the MaGe platform (Vallenet et al. 2017). The comparative genomic analysis was performed to search for genes included in the core genome using the following MICFAM parameters: 80% amino acid identity and 80% alignment coverage. The search for genes involved in the synthesis of secondary metabolites was performed using the AntiSMACH prediction of the MaGe platform (Blin et al. 2019). Genomic sequences were submitted to EMBL (PRJEB44811 and PRJEB44812 for S. cocklensis and A. bryophytorum, respectively).

Proteomic analysis

Sample preparation.

The experimental design of the proteomic analysis consisted of protein extracts from wild-type and chs5 seedlings inoculated with S. cocklensis DSM 42063 or A. bryophytorum DSM 42138 or mock inoculated in triplicate (18 samples). Seedlings (24 days postinoculation) were crushed in liquid nitrogen with a pestle in a mortar and stored at −80°C for further protein extraction. Lysis buffer (1 ml: Tris 50 mM pH8, SDS 1%) was added to 100 mg of frozen material in 2-ml tubes containing 0.25- to 0.5-mm glass beads for grinding the mixture for 3 min with a Cell Disruptor Genie (catalog number SKU: SI-DD38, Scientific Industries, Bohemia, NY, U.S.A.) at 27,850 rpm. The ground mixture was centrifuged at 12,000 × g for 15 min at 4°C. The supernatants were transferred into new tubes, and protein precipitation was performed by adding five volumes of 0.1 M ammonium acetate in 100% methanol, mixing, and incubating overnight at −20°C. The pellet obtained after a 15-min centrifugation step at 12,000 × g was washed once with 0.1% ammonium acetate in 100% methanol and dried. Unless specified, all chemicals were from Merck Sigma (Merck, Darmstadt, Germany). Protein extracts were suspended in 100 to 130 µl of Tris 10 mM pH 6.8, EDTA 1 mM, SDS 5%, glycerol 10% buffer. Protein concentration was assessed to 2 to 6 µg/µl using a Detergent Compatible Assay (Catalog number 5000111, DC Protein Assay, Bio-Rad, Hercules, CA, U.S.A.). Reductant (10 mM DTT) and bromophenol blue (0.01%) were added to the samples, and 40 µg of proteins were loaded to in-house poured 5% acrylamide stacking gels. Gels were stained with Coomassie Blue, and the stacking bands were manually excised. Proteins were then reduced, alkylated, and digested overnight at 37°C with modified trypsin in a 1:100 enzyme/protein ratio (Sequencing-grade Trypsin, Promega). Peptides were extracted from the gel by gentle shaking in 80 μl of 8/2 acetonitrile/water for 1 h. Peptide mixtures were then dried and resuspended in 20 µl of water acidified with 0.1% formic acid.

Liquid chromatography-tandem mass spectrometry (LC-MS/MS) analyses.

Nano LC-MS/MS analyses of peptide extracts were performed on a NanoAcquity LC-system (Waters, Milford, MA, U.S.A.) coupled to a Q-Exactive plus Orbitrap (Thermo Fisher Scientific) mass spectrometer equipped with a nanoelectrospray ion source. Mobile phase A (99.9% water and 0.1% FA) and mobile phase B (99.9% acetonitrile and 0.1% FA) were delivered at 400 nl/min. Samples were loaded into a Symmetry C18 precolumn (0.18 × 20 mm, 5 μm particle size, Waters) over 3 min in 1% buffer B at a flow rate of 5 μl/min. This step was followed by reverse-phase separation at a flow rate of 400 nl/min using an ACQUITY UPLC BEH130 C18 separation column (250 mm × 75 μm id, 1.7 μm particle size, Waters). Nine hundred nanograms of peptide mixtures were eluted using a gradient from 1 to 35% B in 120 min and from 35 to 90% B in 1 min, maintained at 90% B for 5 min, and the column was reconditioned at 1% B for 20 min. The Q-Exactive plus Orbitrap instrument was operated in data-dependent acquisition mode by automatically switching between full MS and consecutive MS/MS acquisitions. Survey full-scan MS spectra (mass range 300 to 1,800) were acquired with a resolution of 70,000 at 200 m/z with an automatic gain control fixed at 3 × 10⁶ ions and a maximum injection time set at 50 ms. The 10 most intense peptide ions in each survey scan with a charge state of 2 or greater were selected for MS/MS fragmentation. MS/MS scans were performed at resolution 17,500 at 200 m/z with a fixed first mass at 100 m/z, automatic gain control was fixed at 1 × 10⁵, and the maximum injection time was set to 100 ms. Peptides were fragmented by higher-energy collisional dissociation with a normalized collision energy set to 27. Peaks selected for fragmentation were automatically put on a dynamic exclusion list for 60 s, and peptide match selection was turned on. MS data were saved in .raw file format using XCalibur software 3.1.66.10 (Thermo Fisher Scientific).

LC-MS/MS data interpretation and validation.

Raw files were converted to .mgf peaklists using MSconvert and were submitted to Mascot database searches (version 2.6.2, MatrixScience, London, U.K.) against an A. thaliana protein sequences database downloaded from The Arabidopsis Information Resource TAIR site (TAIR10 gene model; 27,416 entries) (Lamesch et al. 2012) concatenated with a selection of bacterial sequences downloaded on July 10, 2020, from UniprotKB (only Streptomyces-tagged proteomes with more than 1,000 proteins were kept, in addition to the phylogenetically related genus Achromobacter, for a total of 5,889,978 entries). In-house assembled S. cocklensis and A. bryophytorum predicted proteomes (using the Illumina-sequenced genomes, 17,237 entries, using the MaGe platform annotation tools and process; https://mage.genoscope.cns.fr/microscope/) derived from the in-house genome sequences (section 2.5) were also added to the database, as well as common contaminants (118 entries) and decoy sequences. The concatenated target/decoy database contains 2 × 5,934,749 protein entries. Spectra were searched with a mass tolerance of 5 ppm in MS mode and 0.07 Da in MS/MS mode. One trypsin missed cleavage was tolerated. Carbamidomethylation of cysteine residues was set as a fixed modification. Oxidation of methionine residues and acetylation of proteins’ N-termini were set as variable modifications. Identification results were imported into Proline software (http://proline.profiproteomics.fr/) (Bouyssié et al. 2020) for validation. Peptide spectrum matches with a pretty rank equal to one were retained. The false discovery rate (FDR) was then optimized to be below 1% at the peptide spectrum match level using the Mascot-adjusted E-value and below 1% at the protein level using the Mascot Mudpit score. To obtain unambiguous bacterial identifications, peptide sequences shared between A. thaliana (or contaminant proteins) and bacterial sequences were removed, as well as shared spectra (leading to close but different sequences). The quantification methods and statistical analysis are presented in the statistical methods section. The mass spectrometry proteomics data have been deposited to the ProteomeXchange Consortium via the PRIDE (Perez-Riverol et al. 2022) partner repository with the dataset identifiers PXD041221 and 10.6019/PXD041221.

Electron microscopy

For transmission electron microscopy (TEM), leaves were incubated overnight at 4°C in 3% glutaraldehyde-3% paraformaldehyde in 0.15 M phosphate buffer at pH 7.5 for fixation. Post-fixation was carried out for 2 h at room temperature in 0.5% OsO4. After staining with 0.5% uranyl acetate, samples were dehydrated through an ethanol series from 50 to 100% and embedded in EPON resin (Electron Microscopy Sciences, Hatfield, PA, U.S.A.). Embedded leaves were sectioned with a Reichert-Jung microtome (Leica Microsystems GmbH, Wetzlar, Germany), and sections were observed with a Hitachi H7500 TEM (Hitachi, Tokyo, Japan) at 80 kV. For scanning electron microscopy (SEM) of seeds, samples were handled at a minimum to prevent bacteria dissociating from a seed. Seeds underwent a graded dehydration with ethanol (70 to 100%). Final dehydration was obtained in hexamethyldisilazane. Samples were gold-palladium sputter coated and examined with a Hitachi TM1000 SEM at 15 kV.

Statistical methods

Analysis of chlorophyll accumulation.

The normality of the distributions and equality of variances (homoscedasticity) were verified using R (version 1.4.1103) (https://www.r-project.org) with a Shapiro-Wilk test and a Bartlett test, respectively, and a statistical t test or one-way ANOVA was used to analyze the chlorophyll concentrations.

Microbiota profiling.

For the differential abundance of OTUs, the DEseq2 package (Love et al. 2014), implemented in R, was used with a Wald test to compare groups.

Growth of the wild type (WT) or the chs5 mutant inoculated with A. bryophytorum DSM 42138 or S. cocklensis DSM 42063.

To compare the surface or green intensity of seedlings, the normality of distributions and equality of variances (homoscedasticity) were estimated with a Shapiro-Wilk test and a Bartlett test, respectively. Because at least one of the two conditions was not met for some samples, significant differences were evaluated using a non-parametric Kruskal-Wallis test, and the P values were evaluated with a Dunn test with Bonferroni correction.

Proteomic analysis statistics.

For label-free quantifications, peptide abundances were extracted with the Proline software version 2.1 (http://proline.profiproteomics.fr/; Bouyssié et al. 2020) using an m/z tolerance of 5 ppm. Alignment of the LC-MS runs was performed using Loess smoothing and peptide identity. Cross assignment was performed between all runs using only confident features with an m/z tolerance of 5 ppm and a retention time tolerance of 42 s. MS-derived protein set abundances were calculated by summing the extracted ion chromatogram peak intensities of all identified peptides uniquely matching to each protein set. For statistical analysis, protein abundances were loaded into Prostar software version 1.22.6 (http://www.prostar-proteomics.org/; Wieczorek et al. 2017) and associated with their conditions. Proteins with three values in at least one condition (three replicates) were kept for further statistical analysis. Proteins identified as contaminants were removed. Protein abundances were normalized using 15% quantile centering overall. Residual missing values were assigned in a conservative way (quantile 2.5%). Pairwise Limma tests were performed among the whole set of analyses. P value calibration was corrected using an adapted Benjamini-Hochberg method, and the FDR was optimized for each pairwise comparison. Functional annotation enrichment analysis of differential proteomics data was performed in an unbiased way using the desktop version of DAVID (Ease v2.0) and an updated version of the Gene Ontology (GO) and KEGG pathway databases (October 18, 2021) (Huang et al. 2009a, b). The list of differential proteins from each comparison of the label-free experiment was compared with the list of all protein matches identified. Enriched GO terms and KEGG maps were filtered by only considering those with an Ease score lower than 0.1 and a Benjamini P value lower than 0.05. Enriched GO terms and KEGG maps were thereafter grouped together into broad functional categories, which were considered enriched broad functions.

Results

S. cocklensis DSM 42063 and A. bryophytorum DSM 42138 genome analysis

Bacterial genomes were sequenced on an Illumina MiSeq sequencer. The assembly yielded 156 contigs and 139 scaffolds (N50 = 356,135, 8.06 Mb, with 43 scaffolds >1,000 nt) and 232 contigs and 220 scaffolds (N50 = 178,610, 9.15 Mb, with 117 scaffolds >1,000 nt) for A. bryophytorum DSM 42138 and S. cocklensis DSM 42063, respectively. The completeness for S. cocklensis was 99.4709%. The A. bryophytorum genome was also sequenced using the Nanopore (MinION) technology, leading to 23 scaffolds (N50 = 974, 8.07 Mb). The completeness for A. bryophytorum was 99.4709%. The general features of these genomes are described in Supplementary Table S2. The GC contents (63.7 to 63.9%) were similar, whereas the size differed by 1 Mb. The S. cocklensis genome had more duplications (129 tandem duplications versus 113) and a higher percentage of repetitive sequences compared with the A. bryophytorum genome (Supplementary Table S2). The genome similarity based on the average nucleotide identity was estimated with 187 Streptomycetaceae genomes. The resulting phylogram revealed that A. bryophytorum DSM 42138 was part of a clade also including Actinacidiphila paucisporea CGMCC 4.2025 and Actinacidiphila guanduensis CGMCC 4.2022 (Supplementary Fig. S1). These data support the reclassification of these species into the Actinacidiphila genus as previously suggested (Labeda et al. 2017; Madhaiyan et al. 2022). Our results revealed that S. cocklensis also belongs to this clade, suggesting that this bacterium may be included in the Actinacidiphila genus, but further biochemical and morphological experiments are required to demonstrate this. A comparative genomic analysis showed that the pan genome of the two Streptomycetaceae strains had 17,163 genes, whereas the core genome encompassed 4,918 families of genes, representing 54.7 and 61.7% of the S. cocklensis and A. bryophytorum genes, respectively. The S. cocklensis and A. bryophytorum variable genomes encompassed 45.3 and 38.3% of the predicted ORFs, respectively, which may confer specific capacities to each of these strains (Supplementary Fig. S2).

Genomic features of S. cocklensis and A. bryophytorum adaptation to plants

We searched for genetic determinants found in plant-associated bacteria that could contribute to plant colonization or PGP capacities (Afzal et al. 2019; Levy et al. 2017; Pinski et al. 2019) (Fig. 1; Supplementary Table S2). S. cocklensis and A. bryophytorum exhibited genes involved in the metabolism of fructose, D-mannose, L-arabinose, lactose, ribose, and xylose, which are carbohydrates found in root exudates. In Streptomycetaceae, γ-butyrolactone (GBL) has been described as part of a quorum sensing system; this process is known to be important for plant-bacteria interactions and for regulating antibiotic production (Du et al. 2011; Polkade et al. 2016). In the S. cocklensis genome, we identified a gene homolog to barS (SCOCK_v1_30036) encoding an A-factor type GBL 1'-reductase and two genes (SCOCK_v1_60217 and SCOCK_v1_70112) involved in A-factor biosynthesis. These genes were previously described in other bacteria as essential for the biosynthesis of GBL (Intra et al. 2016; Salehi-Najafabadi et al. 2014; Shikura et al. 2002). In A. bryophytorum, SBRY_v2_50115 is also predicted to be involved in A-factor biosynthesis. Moreover, we found 29 and 27 genes in S. cocklensis and A. bryophytorum, respectively, bearing more than 30% identity with GBL receptors or GBL-binding proteins previously characterized in other Actinomycetes, suggesting that such signaling molecules may play a role in S. cocklensis and A. bryophytorum. Biofilm formation plays an important role in the interaction between plants and bacteria (Naseem et al. 2018). Both bacteria have all genes required for colanic acid biosynthesis, a polysaccharide that contributes to biofilm formation in Enterobacteriaceae and also found in one Pseudomonas strain isolated from root endosphere (Barak et al. 2007; Prigent-Combaret et al. 2000; Zhang et al. 2020). Furthermore, we found one gene in each S. cocklensis and A. bryophytorum genome (SCOCK_v1_190032 and SBRY_v2_60465) displaying 49% identity with vbfA from Bacillus subtilis. vbfA encodes an acetate Na+-dependent symporter subunit involved in the formation of biofilms under the control of volatile molecular signals (Chen et al. 2015). We also identified a bdcA homolog (more than 43% identity, SBRY_v2_11100, SCOCK_v1_100057, and SCOCK_v1_250124) of E. coli that controls biofilm-related phenotypes such as cell motility, cell size, cell aggregation, and production of extracellular DNA and extracellular polysaccharides (Ma et al. 2011). Bacteria interacting with plants often produce cell-wall-degrading enzymes. We noticed that both S. cocklensis and A. bryophytorum have genes required for polymer degradation, such as 1,4-β-xylan, cellulose, glycogen, and chitin.

Download as PowerPoint

We searched for PGP-related genes required for bacterial mechanisms enhancing plant nutrient acquisition and growth. Both bacteria exhibited the ability to acquire phosphate through four acid phosphatases (Fig. 1; Supplementary Table S2). Both bacterial genomes also carried genes homologous to rhbACDF involved in the synthesis of the siderophore rhizobactin produced by the symbiotic nitrogen-fixing bacteria Ensifer meliloti (Lynch et al. 2001) and feuV/yusV, gobMP, involved in iron(III)-siderophore transport. S. cocklensis carried pvsABCDE homologs (SCOCK_v1_100083- SCOCK_v1_100086) required in Vibrio alginolyticus for iron acquisition (Wang et al. 2007). Bacterial genes involved in phytohormone homeostasis were found in S. cocklensis and A. bryophytorum genomes. The genes SCOCK_v1_360062 and SBRY_v2_10788 had more than 30% identity with the acdS genes from several Gram-negative symbiotic bacteria encoding the ACC-deaminase associated with plant ethylene homeostasis (Tittabutr et al. 2008). S. cocklensis had two genes, SCOCK_v1_140087 and SCOCK_v1_270097, encoding an indole acetamide hydrolase and a tryptophan 2-monooxygenase, respectively, required for IAA biosynthesis via the indole-3-acetamide synthesis. A. bryophytorum had only the gene encoding a tryptophan 2-monooxygenase (SBRY_v2_70106). The gene SBRY_v2_10668, SCOCK_v1_390035 shared 66.5 and 62% identity, respectively, with the “Lonely guy” (Log) gene from Corynebacterium glutamicum ATCC 13032, acting as a cytokinin-activating protein (Seo and Kim 2017). In addition, the genes SBRY_v2_11141 and SCOCK_v1_130169 shared about 36% identity with the LOG5 protein from A. thaliana involved in cytokinin biosynthesis (Kuroha et al. 2009). Other phytohormone-related genes were found in the A. bryophytorum genome, such as SBRY_v2_100168, which shared 39% identity with the fas5 gene of the Rhodococcus fascians D188 strain required for fasciation of its host plants (Crespi et al. 1994). SBRY_v2_100168 also shared 33% identity with the CKX4 gene of A thaliana encoding a cytokinine dehydrogenase (Werner et al. 2003). PGP effects may be indirectly due to the inhibition of pathogen growth by specialized metabolites and antibiotics produced by beneficial microorganisms (Kim et al. 2012; van der Meij et al. 2017). We identified 37 and 30 biosynthetic gene clusters (BGCs) for specialized metabolites in the genome of S. cocklensis and A. bryophytorum, respectively. BCGs with high completion were predicted for isorenieratene/carotenoid biosynthesis, for antibiotic or antifungal compounds such as griseusin, and curamycin, and for other bioactive molecules such as alkylresorcinol, pristinol, and 1-heptadecene, or for VOCs (geosmin, 2-methylisoborneol). BGCs predicted to produce the antibacterial lipopeptide rhizomide were found in the case of A. bryophytorum, and BGCs predicted to produce ketomemicin were found in the case of A. cocklensis.

Lastly, we searched for genes encoding proteins with Pfam domains enriched in plant-associated Streptomycetaceae genomes and previously predicted with five independent statistical approaches (Levy et al. 2017). We also searched for plant-resembling plant-associated and root-associated domains (PREPARADOs) enriched in plant-associated Streptomycetaceae as highlighted by (Levy et al. 2017). We found most of the genes encoding such proteins (18 and 17 genes in S. cocklensis and A. bryophytorum, respectively) (Supplementary Table S2). Overall, the analysis of novel genomic data generated here suggests that S. cocklensis and A. bryophytorum have the capacity to engage various processes necessary to interact with host plants. Nevertheless, 38.3 and 45.3% of the genes of A. bryophytorum and S. cocklensis, respectively, were found in only one genome (Supplementary Fig. S2). Of these strain-specific genes, 3 (among 88 genes, 3.4%) and 10 (among 98 genes, 10.2%) genes from A. bryophytorum and S. cocklensis, respectively, were proposed to be involved in plant-bacteria interactions, and at least 16 and 20 genes (considering only complete pathways) from A. bryophytorum and S. cocklensis, respectively, were proposed to be involved in secondary metabolite biosynthesis (Supplementary Table S2). These observations indicate that the metabolic, adaptive, or plant-interacting capabilities of S. cocklensis and A. bryophytorum may be different.

Exploring plant microbiota

To find out whether bacteria affiliated with S. cocklensis or A. bryophytorum occur in the microbiota of other plants, we compared the 16S rRNA sequences obtained from genomic data of these two strains with those found in plant metagenomic datasets using the Joint Genome Institute (JGI) BLAST tools (https://img.jgi.doe.gov/cgi-bin/mer/main.cgi; accessed August 17, 2020). We found several sequences sharing 100% identity with the 16S rRNA sequences of S. cocklensis DSM 42138 and A. bryophytorum DSM 42063 (with a coverage of more than 350 nt) in the Arabidopsis metagenomic datasets but also in the metagenomic data of Populus, Miscanthus, switchgrass, corn, or agave microbiota. Consequently, S. cocklensis and A. bryophytorum are most probably found ubiquitously.

S. cocklensis DSM 42063 and A. bryophytorum DSM 42138 colonize A. thaliana and have differential effects on host growth

To evaluate colonization and effects of either Streptomycetaceae strain on A. thaliana growth, sterilized seeds were inoculated with A. bryophytorum DSM 42138 and S. cocklensis DSM42063 and grown under controlled conditions (Supplementary Fig. S3). Bacterial colonization of A. thaliana seedlings inoculated with S. cocklensis or A. bryophytorum was visualized by TEM (Fig. 2A; Supplementary Fig. S4A). Bacteria were visible on the seed surface as soon as 4 dai, although in the case of S. cocklensis, only a few bacteria were observed. Streptomycetaceae cells also colonized the surface of leaves after 24 days of seedling growth, but no bacteria were detected within the leaf tissues (Fig. 2B; Supplementary Fig. S4B). At the same stage, bacterial colonization of A. thaliana seedlings was quantified by metabarcoding analysis (Fig. 3; Supplementary Fig. S5). Interestingly, the presence of bacteria affiliated with Actinomycetes, Proteobacteria, Firmicutes, Bacteroidota, and Acidobacteriota was demonstrated in non-inoculated seedlings, even though seeds underwent a surface-sterilization treatment (Supplementary Fig. S5A). Two major OTUs, named Cluster 2 and Cluster 3, detected in the inoculated seedlings were affiliated with Streptomyces spp. and A. bryophytorum, respectively. As expected, these OTUs were more abundant in inoculated than in non-inoculated seedlings (ANOVA, P = 0.003 and P = 0.009 for Cluster 2 and Cluster 3, respectively) (Fig. 3) and far more abundant than other bacteria present in non-inoculated control conditions (Supplementary Fig. S5B). Therefore, this quantitative analysis indicated that S. cocklensis as well as A. bryophytorum colonized the chs5 mutant less efficiently than the wild-type seedlings.

Download as PowerPoint

Download as PowerPoint

The effect of A. bryophytorum and S. cocklensis on seedling growth was evaluated 24 or 31 dai by measuring their leaf surfaces (Fig. 4). S. cocklensis had a small but significant growth-promoting effect on wild-type plants at 24 dai, which became non-significant at 31 dai. S. cocklensis had no significant effect on chs5 growth at 24 or 31 dai. A. bryophytorum had no significant effect on wild-type seedlings at 24 or 31 dai but had a significant negative effect on growth of the chs5 mutant at 24 and 31 dai (Fig. 4). Overall, these observations provide evidence that both bacterial species were capable of colonizing A. thaliana wild type and the chs5 mutant, although significantly less for the latter. Both bacteria had different effects on either Arabidopsis wild type or the chs5 mutant, and only A. bryophytorum had a permanent negative effect on growth of the chs5 mutant host plant.

Download as PowerPoint

S. cocklensis DSM 42063 and A. bryophytorum DSM 42138 express proteins involved in amino acid and carbohydrate metabolism when associated with A. thaliana in vitro

To investigate the interaction between S. cocklensis or A. bryophytorum and Arabidopsis seedlings, we carefully analyzed the proteomes of wild-type or mutant holobionts to identify the bacterial proteins expressed in inoculated wild-type or mutant plants. A proper differential analysis was not possible because the number of bacteria colonizing the two genotypes was dramatically different (Fig. 3). We identified 66 proteins expressed by bacteria affiliated with the Streptomycetaceae family. As the non-inoculated plants were still colonized by few bacteria after seed sterilization (Supplementary Fig. S5, non-inoculated plants), we focused only on 23 proteins that were associated with A. bryophytorum or S. cocklensis based on genome annotations and because they were detected in wild-type plants inoculated with A. bryophytorum or S. cocklensis but not in non-inoculated plants (Supplementary Table S3). For seven of these proteins, at least one specific peptide was unambiguously identified as originating from A. bryophytorum or S. cocklensis proteins in the corresponding samples inoculated with A. bryophytorum or S. cocklensis. For the remaining 16 proteins, no peptide specific to one of those two bacteria was found (i.e., we could not distinguish whether the identified protein originated from A. bryophytorum or S. cocklensis). Because we inoculated with one bacterial species separately, we could identify the origin of proteins ambiguously identified as part of either bacteria proteome.

The bacterial proteins expressed in planta were associated with several major biological processes (Supplementary Table S3). Interestingly, S. cocklensis expressed an adenosylhomocysteinase (SCOCK42063_v1_650034) sharing 76.1% identity with the adenosylhomocysteinase encoded by the flI gene from Streptantibioticus cattleyicolor (previously named Streptomyces cattleya; Madhaiyan et al. 2022). This enzyme catalyzes the reversible hydrolysis of S-adenosyl-L-homocysteine (AdoHcy) into homocysteine and adenosine, preventing the inhibition of the fluorinase FlA required for the synthesis of fluoroacetate in this bacterium (Huang et al. 2006). A. bryophytorum expressed a protein, SBRY_v2_40138, exhibiting 33.65% identity with a perakine reductase of Rauvolfia serpentina, which is an aldo-keto reductase involved in the biosynthesis of the monoterpenoid indole alkaloids such as ajmaline. S. cocklensis and A. bryophytorum expressed two proteins involved in xylose import, suggesting a bacterial capacity to use this carbohydrate when associated with plants.

The chs5 mutant is mainly impaired in its plastid/thylakoid function when grown in axenic conditions

In the next step, we compared the abundance of plant proteins in the following samples: wild-type or mutant plants inoculated or not with S. cocklensis or A. bryophytorum. A total of 3,039 Arabidopsis proteins expressed in wild-type or chs5 seedlings were found. About one-tenth (331 proteins) of those showed an abundance significantly different (FDR < 6%) between chs5 and the wild type or between inoculated and non-inoculated plants (Supplementary Table S4).

A differential analysis of the proteomes of non-inoculated wild-type versus chs5 seedlings (1.46% FDR, P value below 0.000447) indicated that the abundance of 94 proteins was significantly different between these two genotypes grown in axenic conditions (Supplementary Table S4, columns D and E). A functional annotation enrichment analysis of both proteomes (Fig. 5) revealed that 14 of the 94 proteins (14.9%) were plastid/thylakoid components or were involved in RNA binding (11 were more abundant and 3 were less abundant in chs5 than in wild-type seedlings), suggesting that the chs5 mutant is strongly impaired in its plastid/thylakoid function when grown in axenic conditions. Among 15 proteins that were more abundant in the wild type, we identified proteins involved in photosynthesis (two proteins), in general metabolism (two proteins), in specialized metabolism (two proteins), in stress response (two proteins), and in plant development (one protein) (Supplementary Table S4, columns D and E, positive Log2FC). Interestingly, among these proteins, we identified a β-1,3-glucanase 3, BGLU21, which is a β-D-glucosidase involved in scopolin, esculin, 4-MU-glucoside, or DIMBOA-glucoside metabolism, and the 12-oxophytodienoate reductase 1 involved in the biosynthesis of oxylipins, a class of compounds active in plant-bacteria interactions (Poloni and Schirawski 2014; Schenk and Schikora 2015; Stassen et al. 2021). The remaining 79 differential proteins were more abundant in chs5 than in wild-type seedlings (Supplementary Table S4, columns D and E, negative Log2FC). Three proteins were enzymes involved in the plastidial isoprenoid biosynthesis: 4-hydroxy-3-methylbut-2-enyl diphosphate synthase and 4-hydroxy-3-methylbut-2-enyl diphosphate reductase acting in the isopentenyldiphosphate pathway and homogentisate prenyltransferase implied in the synthesis of plastoquinone-9. The other proteins were involved in general metabolism, translation, protein folding and turnover, stress response, plant development, hormone-mediated signaling pathway, or photosynthesis. In addition, 12 ribosomal proteins were more abundant in chs5 when compared with wild-type plants. These results indicated that, under axenic and non-inoculated conditions, the chs5 mutant proteome differs from the wild-type proteome, especially regarding proteins involved in isoprenoid synthesis and photosynthesis, but also in plastidial translation, stress response, and in the synthesis of compounds that may play a role in bacteria-plant interactions.

Download as PowerPoint

S. cocklensis DSM 42063 and A. bryophytorum DSM 42138 affect the abundance of a common set of plant proteins

To assess the impact of bacteria on the wild-type or the chs5 proteome, we first searched for proteins whose abundance was different between inoculated and non-inoculated plants. We did not detect any proteins with significant differences in abundance between Streptomycetaceae-inoculated and non-inoculated wild-type plants. In contrast, A. bryophytorum- or S. cocklensis-inoculated chs5 plants displayed 24 and 23 differentially abundant proteins, respectively, compared with non-inoculated chs5 plants (Supplementary Table S4, columns F and G, and J and K, respectively). The abundance of four proteins (two involved in photosynthesis [AT1G31330.1 (PSAF) and AT5G64040.2 (PSAN)], one involved in translation, and one with an unknown function) was lower when inoculated with either A. bryophytorum or S. cocklensis as compared with non-inoculated chs5. Conversely, the abundance of the ELIP1 protein (AT3G22840.1) inhibiting the entire chlorophyll biosynthesis pathway was higher in inoculated as compared with non-inoculated mutant plants. These observations suggested that both bacteria could modulate these same five proteins of the chs5 mutant, and three are involved in photosynthesis. We observed a significant reduction of the green intensity (Supplementary Fig. S6A) and a reduction of chlorophylls a and b and xantophyll and carotenoid contents in chs5 (Supplementary Fig. S6B) inoculated with those bacteria as compared with the inoculated wild-type plants or in chs5 inoculated with A. bryophytorum as compared with non-inoculated chs5. However, this decrease was correlated with the decrease of the plant surface (Supplementary Fig. S7). When pigments were quantified per milligram of dry weight, a higher amount of those pigments was observed in chs5 inoculated with S. cocklensis.

For plants inoculated with each bacterial strain, we compared the proteomes of wild-type versus chs5 (Supplementary Table S4, columns H and I or L and M). In these comparisons, we checked whether plant proteins were common when inoculated with either bacterium. The abundance of a common set of 81 proteins varied between wild-type or chs5 seedlings inoculated with A. bryophytorum or S. cocklensis, and 21 of those proteins were already described as differentially abundant between non-inoculated wild-type and mutant plants. The 60 other proteins are known to exert functions in photosynthesis (13 proteins), general metabolism (7 proteins), translation (7 proteins), protein folding or turnover (8 proteins), and stress response (9 proteins), and 8 proteins were related to phytohormone metabolism: abscisic acid (ABA) metabolism, perception, or functions (AT3G08590.1, phosphoglycerate mutase, AT5G13630.1 magnesium-chelatase subunit chlH) or proteins implied in biosynthesis or signaling pathways of jasmonate (AT4G22240.1 plastid-lipid-associated protein PAP, AT4G04020.1 fibrillin) and auxin (AT1G06000.1 UDP-glycosyltransferase superfamily protein, AT5G64110.1 and AT1G05240.1 peroxidase superfamily protein, AT3G44300.1 nitrilase 2) (Supplementary Table S4). These results strongly suggest that both bacteria have some common effect on plant stress responses and phytohormone homeostasis.

S. cocklensis modulates the abundance of specific proteins involved in various cellular processes in chs5

Beyond the common core proteomes that vary upon colonization of both bacteria strains (see below), the abundance of 18 other proteins was different in the chs5 seedlings inoculated with S. cocklensis compared with the non-inoculated ones (Supplementary Table S4, columns J and K) (5.82% FDR, P value below 0.000437). Inoculation of chs5 with S. cocklensis led to an increase of two proteins as compared with non-inoculated seedlings: the photoreceptor PHOT1 (AT3G45780.1) that regulates stomata opening and leaf photomorphogenesis and a chloroplast thylakoid protein kinase AT5G01920.1. Inoculation of chs5 with S. cocklensis also led to a decrease of 16 proteins, four of them being related to general metabolism and, among those, a 1,2-alpha-L-fucosidase FUC95A (AT4G34260.1) possibly acting on cell wall remodeling. Other proteins are involved in oxidative stress response and translocation of proteins or lipids across chloroplasts or mitochondria membranes (AT2G28900.1, AT2G38540.1, and AT5G43970.1), and one protein, profilin 2 (AT4G29350.1), regulates the organization of the actin cytoskeleton.

Analysis of functional annotation enrichment in chs5 plants inoculated with S. cocklensis relative to wild-type inoculated plants, and in comparison with the same analysis in non-inoculated plants (Fig. 5), suggested that S. cocklensis disrupts several biological processes or molecular functions in wild-type or chs5 seedlings. A comparison of the proteomes of wild-type and mutant seedlings inoculated with S. cocklensis pointed out a variation in the abundance of 195 proteins (0.99% FDR, P value below 0.000661) (Supplementary Table S4, columns L and M): 101 had already been identified in the non-inoculated seedlings or were also differentially abundant between wild-type and mutant plant inoculated with A. bryophytorum (Supplementary Table S4). Among the remaining 94 S. cocklensis-specific proteins, the abundance of 58 proteins was higher in chs5 than in the wild-type plants inoculated with S. cocklensis. Some of these upregulated proteins were related to general metabolism (9 proteins), translation or ribosome biogenesis (27 proteins), stress response (3 proteins), defense of plants against pathogens (3 proteins), and ABA homeostasis (1 proteins). The abundance of 36 specific proteins was higher in the inoculated wild-type than in chs5 plants. These proteins had functions in photosynthesis (11 proteins), general metabolism (7 proteins), stress response (2 proteins), and defense of plants against pathogens (5 proteins) (Supplementary Table S4).

In the presence of A. bryophytorum, the abundance of specific proteins involved in the ABA-mediated stress response was altered in the chs5 mutant

Beyond the common core of proteins that vary upon colonization of both bacteria strains (see below), the abundance of 19 specific proteins (4.55% FDR, P value below 0.000363) was different in the chs5 seedlings inoculated with A. bryophytorum compared with the non-inoculated ones (Supplementary Table S4, columns F and G). Seedlings of the chs5 mutant exhibited an increase in the abundance of eight proteins when inoculated with A. bryophytorum. Among those were fibrillarin 2, a negative regulator of the expression of immune responsive genes; AT1G13930.1, a protein categorized as salt stress responsive; and AT1G06000.1, a UDP-glycosyltransferase acting in flavonol biosynthesis. Seedlings of the chs5 mutant inoculated with A. bryophytorum also exhibited a decrease in the abundance of 11 proteins, among which were the oxylipin biosynthetic enzyme allene oxide synthase CYP74A (AT5G42650.1), two proteins probably associated with the light/cold stress-related jasmonate biosynthesis (AT1G77490.1 and AT2G35490.1), and PCAP1 (plasma-membrane-associated cation-binding protein, AT4G20260.4) and PATL1 (AT3G14210.1) and PATL2 (AT1G22530.1), which are phosphoinositide-interacting proteins involved in membrane-trafficking events associated with cell plate formation during cytokinesis.

Analysis of functional annotation enrichment in chs5 plants inoculated with A. bryophytorum relative to wild-type inoculated plants, and in comparison with the same analysis in non-inoculated plants, suggests that this bacterium modulates the abundance of abiotic stress-related proteins (Fig. 5). A comparison of the proteomes of wild-type and mutant plants inoculated with A. bryophytorum revealed that the abundance of 157 proteins varied (0.98% FDR, P value below 0.000525) (Supplementary Table S4, columns H and I), among which 90 had already been discussed previously when comparing non-inoculated wild-type and mutant seedlings or were identified as differentially abundant in wild-type versus mutant plants inoculated with S. cocklensis (see previous paragraphs). Among the remaining 67 A. bryophytorum-specific proteins, the abundance of 20 proteins was higher in chs5 than in the wild type when inoculated with A. bryophytorum, especially proteins involved in general metabolism (four proteins), translation or ribosome biogenesis (three proteins), stress response (two proteins), and ABA homeostasis (one protein) (Supplementary Table S4). A set of 47 proteins was specifically more abundant in the wild-type than in mutant plants, some of which are involved in photosynthesis (18 proteins), general metabolism (3 proteins), stress response (6 proteins), and auxin, ABA, and jasmonate homeostasis (3 proteins). These results suggest that in the presence of A. bryophytorum, chs5 seedlings exhibit different metabolism, photosynthesis capacities, and responses to stress, in particular those related to ABA, jasmonate, and auxin, as compared with the wild type.

In conclusion, it appears that proteomes from chs5 seedlings colonized by S. cocklensis or A. bryophytorum are broadly modified when compared with the proteome of axenic chs5 seedlings. Overall, the differences between wild-type and mutant proteomes of inoculated seedlings are proteins of a wide range of functions, even more so than in non-inoculated seedlings. Both bacteria apparently affect the abundance of a common set of proteins known to play a role in photosynthesis. Nevertheless, in addition to this common set, these two bacteria modify the abundance of different Arabidopsis proteins associated with various cellular processes, such as phytohormone homeostasis or stress responses. Such a change in the proteome profiles of chs5 is most likely responsible, at least in part, for its reduced growth, particularly in the presence of A. bryophytorum (Fig. 4).

Discussion

Plant organs and tissues are the hosts of microbial communities (or microbiota) that consequently form holobiont entities. Our previous work revealed that the metabolic status of A. thaliana was crucial for the colonization of Streptomycetaceae into the microbiota, in particular those affiliated with A. bryophytorum and S. cocklensis (Graindorge et al. 2022). The aim of this work was to study the interaction between wild-type or mutant A. thaliana and A. bryophytorum DSM 42138 and S. cocklensis DSM 42063 isolated from moss and soil, respectively (Kim et al. 2012; Li et al. 2016). We indeed confirmed that these two bacteria colonize A. thaliana. The results demonstrate a decreased colonization of A. bryophytorum and S. cocklensis in the chs5 mutant that carries a weak allele of the gene coding for DXS1, a key enzyme of the plastidial MEP pathway, as compared with the wild type. This is in full accordance with our previous observations suggesting that colonization and/or growth of A. bryophytorum or S. cocklensis is reduced when these bacteria interact with this chs5 mutant (Graindorge et al. 2022). Because the chs5 mutant has defects in the synthesis of isoprenoids, phenylpropanoids, and lipids, these metabolites could participate in the recruitment of A. bryophytorum and S. cocklensis by the wild-type plant and/or colonization or bacterial growth when associated with the host plant. In this study, in addition to metabolites, we observed significant variations in the chs5 proteome compared with the wild type. The detailed proteome analysis of the chs5 mutant confirms a major deficit of the plastid/thylakoid function. Interestingly, we also observed variation in the abundance of proteins related to stress responses, photosynthesis, and hormone homeostasis or perception. We cannot exclude that those variations may directly or indirectly affect bacterial colonization.

This study enabled us to study the mechanisms of interactions between Streptomycetaceae and Arabidopsis. When associated with their host, A. bryophytorum and S. cocklensis expressed proteins involved in xylose import and a glyceraldehyde-3-phosphate dehydrogenase, suggesting that xylose can be used by the bacteria and further transformed via the glycolysis pathway. Interestingly, xylose is one of the most important carbohydrates produced in Arabidopsis roots (Peng et al. 2000). Other proteins expressed by the bacteria may be important for their interaction with their host. For example, S. cocklensis expressed an adenosylhomocysteinase, and A. bryophytorum expressed a protein, SBRY_v2_40138, which is an aldo-keto reductase involved in the biosynthesis of monoterpenoid indole alkaloids. Interestingly, proteins containing an aldo-keto reductase domain (Pfam00248) were enriched in the genome of plant-associated bacteria, suggesting that metabolites generated by such enzymatic activity may be important for plant-bacteria interactions (Levy et al. 2017). These proteomic data suggest that a metabolic interaction is possible between A. thaliana and A. bryophytorum or S. cocklensis.

Streptomycetaceae have the capacity to promote growth of plants (PGP) (Olanrewaju and Babalola 2019). We observed that S. cocklensis had a temporary minor but significant PGP effect on wild-type A. thaliana using the seedling growth bioassay described in this work. This bacterium exhibits genes for the synthesis of IAA, which may be responsible for this PGP effect, as previously observed with other Streptomycetaceae (Fu et al. 2022). This minor PGP effect was not observed when the chs5 mutant seedlings were inoculated with S. cocklensis, and a negative effect on growth was visible in the case of chs5 inoculated with A. bryophytorum. These observations suggest that the chs5 seedlings, but not the wild-type seedlings, have several defects and must withstand adverse conditions when inoculated with A. bryophytorum. Thus, the interaction between Arabidopsis and these two Streptomycetaceae could be defined as neutral in a wild-type context but, in the case of A. bryophytorum, harmful in the chs5 background where isoprenoid, phenylpropanoid, and lipid syntheses are partially impaired.

Both bacteria modulate the plant chs5 proteome, revealing that colonization by these bacteria influences plant physiology. A. bryophytorum and S. cocklensis share a common effect on plant stress responses and phytohormone homeostasis. However, each bacterium also has specific effects on the Arabidopsis proteome, as shown in Figure 5. The abundance of proteins involved in general metabolism was different between the wild-type and the chs5 seedlings inoculated with S. cocklensis DSM 42063, whereas the abundance of proteins implied in the response to abiotic stresses was different between the wild-type and chs5 seedlings inoculated with A. bryophytorum DSM 42138. These variations could be at least partly responsible for the reduced growth of chs5 in the presence of A. bryophytorum. Inoculation of chs5 with this strain led to an increase of the fibrillarin 2 that acts as a negative regulator of the expression of immune responsive genes, and of a protein involved in the regulation of ABA biosynthesis (AT1G13930.1). This observation indicates that A. bryophytorum may directly or indirectly manipulate the immune response of its mutated host or modulate ABA homeostasis. This effect on plant immune response or ABA homeostasis may indirectly play a role in the abiotic stress response (drought or salinity for example) and/or change the capacity of the plant to resist pathogens. A. bryophytorum may therefore modulate the induced systemic response (Pinski et al. 2019; Vlot et al. 2021). Remarkably, chs5 plants were less efficiently colonized by bacteria affiliated with Streptomycetaceae and were more sensitive to Pseudomonas syringae infection when grown in holoxenic conditions (Graindorge et al. 2022). The Streptomycetaceae-dependent tuning of the Arabidopsis immune response observed in this study under gnotoxenic conditions could be in part responsible for the increased pathogen susceptibility of the chs5 mutant observed previously under holoxenic conditions (Graindorge et al. 2022). However, further studies are needed to characterize a possible modulation of the immune response induced by A. bryophytorum. In addition, other indirect effects could be proposed to explain the above observations. For example, these bacteria have the capacity to produce antibiotics, which is crucial for the biocontrol effect previously described for such bacteria (Olanrewaju and Babalola 2019; van der Meij et al. 2017). S. cocklensis produces the antibiotic dioxamycin (Kim et al. 2012). S. cocklensis and A. bryophytorum probably produce other antibiotics because we found in their genomes more than 30 clusters of genes with a putative role in the synthesis of antibiotic and specialized metabolites.

Our data revealed that each bacterial strain had a distinct effect on its host, and the genomic content of A. bryophytorum and S. cocklensis was different in terms of gene content, for example, genes involved in plant-bacteria interactions. A. bryophytorum and S. cocklensis are two phylogenetically closely related strains, but the genomic data indicate that the variable genome otherwise known as the dispensable genome of these two Streptomycetaceae strains encompass 45.3 and 38.3% of the predicted ORFs. In fact, genomes of Streptomycetaceae strains are indeed quite divergent, with large dispensable regions localized in the sub-telomeric part of their linear genome (Bu et al. 2020; Kim et al. 2015).

Finally, the results presented here are consistent with a specific selection of Streptomycetaceae by Arabidopsis via signals or perception mechanisms involving isoprenoids, phenylpropanoids, and/or lipids. Database searches suggest that bacteria affiliated with S. cocklensis or A. bryophytorum are found in the microbiota of other plants. It is therefore most likely that the bacterial effect observed here in the case of Arabidopsis could also be observed with other plants. Overall, the work presented here provides a better understanding of the molecular mechanisms involved in the interaction between Streptomycetaceae and A. thaliana and, more broadly, the selection mechanisms of the plant microbiota.

Acknowledgments

H. Schaller thanks Toshiya Muranaka and Koh Iba for the gift of A. thaliana chs5 seeds.