Metatranscriptomic response of the wheat holobiont to decreasing soil water content

The study investigates how the components of the wheat holobiont (plant and its associated microbiota) respond at the transcriptomic level to decreasing soil water content. Drought is a major and intensifying threat to crop productivity. While both plants and their microbiota respond to water stress, integrated studies across plant and microorganisms are scarce, leaving unclear which partners are the best targets to improve drought resilience. Based on the hologenome/holobiont framework, which emphasizes the combined host-microbiome adaptive potential, the authors hypothesized that most of the holobiont’s transcriptomic response to reduced soil water availability would be microbial (reflecting both gene expression changes and shifts in metagenomic composition) rather than plant-driven (limited to gene expression changes). To test this, wheat was grown under field rainout shelters allowing 25–100% of natural precipitation, and root and rhizosphere metatranscriptomes were profiled to compare the most contrasting conditions (25% vs 100%).

Prior work shows numerous microbial taxa can enhance plant tolerance to drought and salinity, including many Actinobacteria and Proteobacteria, as well as fungal endophytes and mycorrhizal fungi, which improve water use efficiency and drought resilience in cereals. Mechanisms include modulation of plant stress-responsive genes, reduction of ethylene via bacterial ACC deaminase, induction of osmolyte-related genes by bacterial VOCs, and epigenetic modulation. The holobiont/hologenome concept posits hosts and their microbiomes as integrated evolutionary units; microbial components can rapidly change via partner recruitment, relative abundance shifts, and horizontal gene transfer, enabling adaptive responses. Microbial taxa from drought-prone environments often confer greater drought resistance to plants. Plant responses to water stress involve genetic, molecular, and physiological pathways; however, the microbial metagenome is more dynamic and plastic than the host genome within a generation, suggesting a potentially larger microbial contribution to holobiont responses.

Field experiment and sampling: Rainfall manipulation was established in 2016 at the INRS Centre Armand-Frappier Santé Biotechnologie (Laval, Québec, Canada) using 2 m × 2 m passive rain-out shelters transmitting 25%, 50%, 75%, or 100% of natural precipitation via transparent plastic sheeting (nine, six, three, or zero 2 m × 16.7 cm sheets, respectively) with gutters and 20 L collection buckets. Two wheat genotypes were seeded (drought-sensitive Triticum aestivum cv. AC Nass and drought-tolerant Triticum turgidum ssp. durum cv. Strongfield) in six randomized blocks (4 treatments × 2 genotypes × 6 blocks = 48 plots). Seeding density was 500 seeds m−2 on May 18, 2016 and May 23, 2017; harvested seeds were reseeded into the same plots the following year. For this study, only Strongfield plots were used. On July 26, 2017, rhizosphere soil and roots were collected: plants were uprooted, loosely adhering soil removed, and tightly adhering soil collected as rhizosphere. Roots were washed, excised, and both rhizosphere and roots were flash frozen in liquid nitrogen within 2 minutes and stored at −80 °C. Bulk soil from plot centers was collected to determine gravimetric soil water content (SWC) after drying overnight at 105 °C. RNA extraction and sequencing: Total RNA was extracted from 2 g rhizosphere soil (RNeasy PowerSoil Total RNA Kit) and 0.5 g roots (RNeasy Plant Mini Kit). DNA was removed with DNase (ThermoFisher), and absence of DNA was confirmed by lack of 16S rRNA gene PCR amplification. Libraries: rhizosphere libraries used microbial ribosome subtraction to capture microbial transcripts; root libraries used poly-dT reverse transcription to focus on plant and fungal transcripts. Sequencing was performed on Illumina HiSeq4000 (2 × 100 bp paired-end). Data availability: NCBI BioProject PRJNA880647. Bioinformatics: Reads were quality trimmed (Trimmomatic v0.32; bases with Q < 30 at read ends removed) and screened for adapter contaminants (DUK) to obtain QC-passed reads. Co-assembly across samples was performed with MEGAHIT v1.1.2 using iterative k-mer sizes 31–91. Transcripts were predicted with Prodigal v2.6.2. Annotations followed DOE-JGI MAP v4 guidelines, including KEGG ortholog assignments. QC reads were mapped to contigs with BWA mem v0.7.15; alignments were coordinate-sorted (samtools v1.2), and only properly paired alignments were retained. Coverage and transcript counts were derived with bedtools v2.17.0 using a transcript coordinate BED file, counting only paired reads both overlapping the target. Abundance matrices (contigs × samples) were generated and normalized to CPM using TMM (edgeR v3.10.2). Taxonomy: Each contig was BLASTn-searched (v2.6.0+) against NCBI nt (downloaded Jan 9, 2019), and best-hit taxIDs were used for lineage assignments; summaries with MicrobiomeUtils v0.9. Statistical analyses: R v4.1.0 was used. Differential transcript abundance between 25% and 100% precipitation was assessed with EBSeq (EBTest) at FDR 0.05. ANOVA for SWC used aov. Reproducibility: Analysis code on GitHub (https://github.com/le-labo-yergeau/MT_Holobiont_Wheat); abundance/annotation tables, metadata, and SWC files on Zenodo (10.5281/zenodo.7096909); additional workflow outputs on Zenodo (10.5281/zenodo.7121038).

• Soil water content manipulation: SWC differed significantly among treatments (p = 0.000367), decreasing from 23% (100% precipitation) to 11% (25%).
• Sequencing and assembly: 1,069,108,624 clean reads (per sample: mean 22,746,992; min 13,943,395; max 43,912,775) yielded 1,269,055 transcripts; detected transcripts per compartment: rhizosphere 1,193,501; roots 1,136,587.
• Kingdom-level composition: Roots—Archaea 1.1% (12,792), Bacteria 32.2% (365,435), Fungi 17.6% (200,233), Plants 17.7% (200,823), other Eukaryotes 11.6% (132,313), Viruses 0.3% (3,660), Unclassified 19.5% (221,331). Rhizosphere—Archaea 1.3% (14,943), Bacteria 36.1% (430,984), Fungi 15.6% (186,745), Plants 14.2% (169,788), other Eukaryotes 11.1% (132,042), Viruses 0.4% (4,255), Unclassified 21.3% (254,744).
• Differential abundance (25% vs 100% precipitation): Roots—42,001 DA transcripts (3.70%). Taxa among DA: Bacteria 2,309 (5.50%), Fungi 23,274 (55.41%), Plants 4,357 (10.37%), Unclassified 5,303 (12.63%), Other taxa 6,758 (16.09%). Direction: Fungi 23,042 more vs 232 less; Bacteria 2,231 more vs 78 less; Plants 1,295 more vs 3,061 less.
• Rhizosphere—21,765 DA transcripts (1.82%). Taxa among DA: Bacteria 14,178 (65.14%), Archaea 159 (0.73%), Fungi 402 (1.85%), Plants 219 (1.01%), Unclassified 5,224 (24.00%), Other taxa 1,583 (7.27%). Direction: Bacteria 7,938 more vs 6,240 less; Fungi 149 more vs 253 less; Plants 41 more vs 178 less.
• Taxonomic trends: Positive DA overrepresentation included Actinobacteria (both compartments), Ascomycota in rhizosphere, and Dothideomycetes in roots. Negative DA overrepresentation included Proteobacteria, Bacteroidetes, Eurotiomycetes (both compartments), Acidobacteria in rhizosphere, and Agaricomycetes in roots. Some taxa (e.g., Sordariomycetes, Chloroflexi, Gemmatimonadetes) showed limited DA response.
• Functional (COG) trends: Generally upregulated with reduced water—roots: Carbohydrate transport and metabolism, Lipid metabolism; rhizosphere: Cell envelope biogenesis/outer membrane, Signal transduction mechanisms, Transcription. Generally downregulated—both compartments: Translation, ribosomal structure and biogenesis; roots: Posttranslational modification, protein turnover, chaperones. Some categories showed limited response (e.g., rhizosphere: Amino acid transport and metabolism; roots: Cell envelope biogenesis, DNA replication/repair, Signal transduction, Transcription).
• Notable DA transcripts: Roots—many Agaricomycetes (notably Coprinopsis cinerea) transcripts more abundant at 25% (e.g., Neutral trehalase, Hexokinase, AA transporters, Glycerol uptake facilitators, Beta-glucanase), while plant transcripts (wheat/relatives) tended to be less abundant. Rhizosphere—many Proteobacteria DA transcripts with reduced abundance at 25% related to motility and secretion (pili, flagella, Type II/VI systems); fungal Agaricomycetes transcripts related to carbohydrate/amino acid metabolism increased at 25%.
• Shared DA between compartments: 513 positive DA transcripts shared (392 Actinobacteria; also Basidiomycota, Ascomycota, Proteobacteria), enriched in Translation, Transcription, Carbohydrate and Amino acid transport/metabolism, and Chaperones. Forty-seven negative DA transcripts shared (many unclassified; also Streptophyta and Basidiomycota), with affiliations to Cytoskeleton, Energy production, and Translation.

The wheat holobiont’s transcriptomic response to decreased precipitation was dominated by microbial partners: fungi responded most strongly in roots and bacteria in the rhizosphere, whereas plant transcripts showed comparatively smaller shifts and often decreased under low water. This aligns with the greater plasticity of the microbial metagenome (via community composition shifts and gene regulation) compared to the host genome (limited to gene regulation within a generation). Taxonomic patterns match known drought sensitivities: Actinobacteria increased (or upregulated functions) under drying, while Proteobacteria and Acidobacteria decreased. Functional responses highlighted microbial strategies for drought tolerance, including upregulation of osmolyte-associated carbohydrate and amino acid metabolism (e.g., trehalose, transporters, hexokinase) in root-associated fungi (notably Agaricomycetes such as Coprinopsis), and reduced expression of motility and secretion machinery in rhizosphere bacteria, consistent with shifts toward biofilm lifestyles under stress. Heat shock proteins were more abundant under low water in both compartments. Despite an overall downregulation of translation-related transcripts, Actinobacteria featured among positive DA translation-related functions, potentially contributing to their drought dominance. While plant gene expression changes were more limited—possibly influenced by the drought-tolerant Strongfield cultivar and moderate SWC levels—the observed microbial transcriptomic shifts likely influence holobiont function, potentially mediated by plant exudation effects on microbes.

Under field-imposed reductions in precipitation, the wheat holobiont exhibits a predominantly microbial metatranscriptomic response: root fungi and rhizosphere bacteria contribute most DA transcripts, with functional shifts consistent with drought-coping mechanisms (osmolyte metabolism, stress proteins, reduced motility). Plant transcripts generally show smaller or negative responses under low water. These findings support targeting the microbiome—especially drought-responsive taxa like Actinobacteria and beneficial root-associated fungi—in strategies to improve crop resilience to water stress. Future work should include holobionts of drought-sensitive genotypes and experiments imposing more extreme water deficits to assess whether plant responses become more pronounced and to disentangle community composition changes from gene expression changes.

Metatranscriptomics cannot disentangle whether DA transcripts arise from community composition shifts or gene expression changes. The differential abundance analysis produced a single DA list per compartment, limiting statistical testing of taxon over/under-representation beyond descriptive trends. The driest treatment achieved moderate SWC (~11–12%) without visible plant stress, and the cultivar studied (Strongfield) is drought-tolerant; both factors may have attenuated plant transcriptomic responses and constrained generalizability to more severe drought or sensitive cultivars. Some transcripts remained unclassified at higher taxonomic or functional levels, and potential database biases may affect taxonomic/functional assignments.