Arid Soil Bacterial Legacies Improve Drought Resilience of the Keystone Grass, Themeda triandra

Abstract

Plant-microbe interactions are critical to ecosystem functioning and impact soil legacies, where plants exert a lasting influence on the microbial and physicochemical conditions of the soils in which they grow. These soil legacies can affect subsequent plant growth and fitness. Specifically, biotic soil legacies can influence microbially associated plant fitness through the movement of soil microbiota in a two-step selection process: Microbes are recruited from bulk soil into the rhizosphere (the space around roots) and then into the endosphere (within plant roots). Furthermore, these endosphere root microbiota can also influence plant behaviour, shaping bulk soil communities over time. However, the potential of these soil legacies to provide host plant drought tolerance remains poorly understood. In a drought stress greenhouse trial, we show that arid soil legacies increased the biomass of the keystone grass Themeda triandra under both drought and control conditions. We report strong positive associations between T. triandra biomass and bacterial alpha diversity across soils, rhizospheres and endospheres. These findings show that bacterial soil legacies have an important but underappreciated role in grassland species resilience to drought and could be better harnessed to support resilient grassland restoration efforts.

Keywords: amplicon sequencing; drought resilience; endosphere; rhizosphere; soil microbiota; two‐step selection.

FIGURE 1

High‐ and low‐aridity sampling sites, and Themeda triandra plant trait responses to treatment effects. (a) High‐ and low‐aridity sampling sites for the collection of soil microbiota for experimental manipulation (yellow points). Mean annual aridity index data layer (ADM) is sourced from the Soil and Landscape Grid of Australia (Searle et al. 2022), where aridity index is calculated via annual precipitation/annual potential evaporation. T. triandra plant growth responses to soil aridity, sterilised and water stress treatments showing: T. triandra (b) total biomass, (c) aboveground biomass, (d) belowground biomass and (e) root–mass fraction differences. We chose not to overlay statistical groupings on raw‐data boxplots, as differences were tested via a mixed model that accounts for random structure (e.g., soil aridity). Statistical comparisons and groupings based on model‐estimated differences are provided in Table S6.

FIGURE 2

Mean relative abundance of major bacterial phyla across plant‐present pots within Themeda triandra compartments over time. (a) Compartment and time point included were the initial soil sampling period, soils at plant harvest, T. triandra rhizospheres at plant harvest and T. triandra endospheres at plant harvest. Treatments include sterilisation (live, sterile), soil aridity (high, low‐aridity soils) and watering regime (water stress as red text labels, control as blue text labels). We did not sequence viable DNA from sterilised low‐aridity soils. (b) Heatmap of ASV occurrence per 1000 ASVs within each phylum across each sample. Values represent the number of ASVs assigned to each phylum, scaled by the total number of ASVs per sample. Samples are annotated by compartment, water regime, sterilisation treatment and aridity. (c) Differential abundance analysis comparing changes in phyla within each time point and compartment across treatments. Each category compares differences to a reference group (the high aridity, live, control soil treatment). Log fold changes for the reference groups identify differences from the grand mean of each phyla.

FIGURE 3

Bacterial alpha diversity differences across Themeda triandra compartments and soil physicochemical influences (a) Alpha diversity (effective number of ASVs) across treatments, time and plant‐present versus soil‐only pots. Sample library sizes were rarified to 18,738 reads. Pairwise significance between treatments is indicated by unique letters within each subplot (see Table S5). (b) Canonical correspondence analysis (CCA) of bacterial communities at t1 (plant harvest) compared with significant soil physicochemical variables. Points are coloured by site.

FIGURE 4

Bacterial community differences across T. triandra compartments and time points. Nonmetric multidimensional scaling (NMDS) plot showing bacterial community composition differences for each sampling treatment. Each point represents a sample, and closer points have more similar communities. Sample library sizes were rarified to 18,738 reads.

FIGURE 5

Bacterial community differences using Bray–Curtis dissimilarity across each experimental treatment and comparisons to soil‐only pots. Nonmetric multidimensional scaling (NMDS) plot showing bacterial community composition differences across treatments in (a) sample types from plant‐present pots and (b) soil‐only containing low versus high‐aridity soils. All NMDS ordinations are based on Bray–Curtis distances that emphasise relative abundance of taxa (sample library sizes were rarified to 18,738 reads). Each point represents a sample, and closer points have more similar communities.

FIGURE 6

Bacterial diversity is correlated with Themeda triandra total biomass. (a) Alpha diversity (effective number of ASVs) is positively correlated with postharvest T. triandra biomass across all plant compartments and watering treatments. Soil aridity is denoted by colour (red = high‐aridity soils, blue = low‐aridity soils), and soils exposed to sterilisation at the beginning of the trial are shown with point shape (sterilisation = triangles, live = circles). (b) Relative Interaction Index (RII) for T. triandra plants grown under water stress and control conditions, and in soils from high‐ and low‐aridity soils. The RII ranges from −1 to 1 and reflects the net effect of soil microbial communities on plant growth. Values > 0 indicate a beneficial effect of live microbial communities compared to sterilised soils, while values < 0 indicate improved growth in the absence of live microbes. Squares represent treatment means, and error bars indicate 95% confidence intervals. Pairwise significance between treatments is indicated by unique letters within each subplot (see Table S3).