Drought response of the maize plant–soil–microbiome system is influenced by plant size and presence of root hairs

Abstract

Background and Aims

We have abundant knowledge on drought responses of plants or soil microorganisms individually. However, there is a severe lack of knowledge regarding interactions in the plant–soil–microbiome continuum, and specifically root–soil interface traits including the role of root hairs. Here we investigated how water limitation propagates in a plant–soil–microbiome system upon stopping irrigation. We used two Zea mays genotypes [rth3 and its isogenic wild type (WT), B73], differing in root hair formation, to elucidate the effect of rhizosphere extension under water limitation.

Methods

For 22 d, WT and rth3 plants were grown in a climate chamber, with irrigation stopped for drought treatment during the last 7 d. Daily measurements included soil water status, plant evapotranspiration and gas exchange. At harvest, root exudates, shoot relative water content, osmolality and nutrients, root morphological traits and transcriptomics, and soil microbial β-diversity and enzyme activity were determined.

Key Results

In line with a larger plant size, drought stress developed more rapidly and the number of differentially expressed genes was higher in the WT compared with rth3. Under water limitation, root exudation rates increased and soil enzyme activities decreased more strongly in the WT rhizosphere. In both genotypes, water level significantly altered microbial β-diversity in the bulk soil, particularly affecting fungi more than bacteria/archaea. The genotype affected only bacteria/archaea and was more pronounced in rhizosphere than in bulk soil.

Conclusions

This interdisciplinary study assessed how a short drought stress manifested in a plant–soil–microbiome system. Water limitation altered microbial (fungal) diversity in distance from the root surface. Genotype-specific stress-induced increases in exudation rates modified microbial activity in root proximity, possibly pointing to root hair functions under water limitation. Less intense drought responses of rth3 were confirmed at all levels of investigation and may be due at least in part to its smaller plant size.

INTRODUCTION

Agricultural production is adversely affected by climate change, with drought being singled out as one of the main causes of food insecurity (Caretta et al., 2022). Crops grow in diverse soil types, influenced by various abiotic environments that propagate drought stress in different ways (Draye et al., 2010), while also interacting with a wide range of microbiota. Despite a large body of knowledge on the responses to water limitation of plants and microbial communities separately (Naylor and Coleman-Derr, 2018), we still lack an integral understanding of the plant–soil–microbe system under water limitation (Vetterlein et al., 2020). The main challenge is that water limitation not only affects plants or microbes directly, but that the alteration of one ‘player’ (plant or microbes) can indirectly change the stress response of the other (Naylor and Coleman-Derr, 2018). For both phytobiome and microbiome, adaptive responses to stress rely heavily on temporal developments (Kuromori et al., 2022; Metze et al., 2023) and the chosen experimental conditions.

Well-described early effects of water limitation on water relations include reduced plant water content, increased osmotic potential or stomatal closure (Wedeking et al., 2017; Kuromori et al., 2022; Sato et al., 2024) and differential gene expression related to the production of reactive oxygen species (ROS) (Opitz et al., 2014) or associated with signalling, photosynthesis, carbohydrate metabolism and oxidoreductase activities (Zhang et al., 2018). Other effects occur more slowly, such as reduction of photosynthesis and changes in overall metabolism (Tripathi et al., 2016), root morphology and root exudates (reviewed in Fang and Xiong, 2015; Tiziani et al., 2022). The soil microbiome can be directly affected by water supply, as some taxa are simply more tolerant to desiccation than others and have a relative advantage during water limitation (Xu et al., 2018). Changes in soil moisture also alter nutrient and exudate diffusion rates, soil pH and redox state, which determine the nutritional status and physiology of plants, as well as the soil microbial community composition (Hartman and Tringe, 2019; Jaeger et al., 2023). To complicate matters further, the quality and quantity of root exudates, which are decisive for microbiota recruitment, are highly dependent on environmental conditions. Under water limitation, both the amount and the composition of root exudates can change (reviewed in Naylor and Coleman-Derr, 2018). These changes can be expected to alter the composition of the rhizosphere microbiome, as rhizosphere-associated bacterial communities depend on the profile of exudates released (Shi et al., 2011; Zhao et al., 2021). Rhizosphere microbial extracellular enzyme activity is also affected by water limitation (Rajper et al., 2024).

Root hairs develop from specific epidermal cells called trichoblasts and protrude into the rhizosphere, where they extend the contact area between roots and soil, absorb nutrients and exude organic compounds and enzymes (Bienert et al., 2021). Under water-limited conditions, root hairs increase the depletion zone for immobile nutrients, facilitate the diffusion of exudates and alter the microbial composition of the rhizosphere (Vetterlein et al., 2022). Their ability to improve root penetration into harder soil layers (Kong et al., 2024) may represent another benefit in drying soils. There is conflicting evidence for a direct contribution of root hairs to root water uptake. While barley root hairs contributed to the avoidance of steep soil water potential gradients around roots (Carminati et al., 2017), a similar function was not shown for maize root hairs (Cai et al., 2021). On the other hand, root hairs were shrinking in drying soils at relatively high water potentials, likely limiting their contact with the soil and a possible function in water uptake (Duddek et al., 2023). The presence of root hairs in maize had a limited but measurable effect on the composition of the rhizosphere microbiome (Gebauer et al., 2021), but it is still largely unknown how and to what extent they contribute to the formation of the maize rhizosphere microbiome during water deficit. Better knowledge of root hair functions might be key for selecting root surface traits in breeding strategies to optimize the resilience of crop plants to abiotic stresses.

Noticeably, observations are often contradictory between different drought stress studies. For example, some studies report an increase in root exudates (Canarini et al., 2016; Karlowsky et al., 2018), while others report a decrease (Gargallo-Garriga et al., 2018; Staszel et al., 2022). Similarly, inconsistent results are also reported with respect to microbiome alterations under water limitation (reviewed in Naylor and Coleman-Derr, 2018). Such contrasting results might in part be due to differences in drought intensities and durations, but also to the use of different species or genotypes at different developmental stages and with different adaptation strategies to water limitation (Canarini et al., 2016). Leaf area in particular is often not reported, even though it largely defines the plant’s water requirement and hence the temporal development of stress, and may be significantly different even between closely related genotypes (Jorda et al., 2022). Likewise, microbes differ in their potential interaction with plants, ranging from improving the availability of nutrients to direct effects on root growth by releasing or altering plant hormone levels (Passari et al., 2016), thus influencing plant response to water limitation. The largest part of the inconsistencies is most likely related to the poor definition of drought stress across the literature. Drought stress treatments for gene expression studies range from purely osmotic stress induced by polyethylene glycol (Opitz et al., 2014) to sudden stress induced by removing plants from hydroponics (Tripathi et al., 2016). Other studies investigate drought in soil or substrate-grown plants without providing any soil physical information, which would be key to understanding temporal development of soil water availability (Moshelion et al., 2024).

The present study aims to understand how water limitation propagates in a plant–soil–microbiome system by examining physiological parameters, gene expression, exudation, microbiome composition and microbiome functions for two maize genotypes differing in their ability to form root hairs. Specific attention is given to the fact that the presence of root hairs had induced slight differences in plant size already at the start of the drought stress treatment. We address the following hypotheses: (1) genotypic differences in plant size at the onset of water limitation lead to different stress intensities and thus stress responses, which extend to the rhizosphere microbiome; and (2) the effect of water limitation on soil microbial communities is mediated by root-based processes and the presence of root hairs, resulting in genotype-specific alterations of microbiome activity and composition in different soil compartments.

MATERIALS AND METHODS

Experimental setup

Zea mays plants of the genotypes roothairless3 mutant (rth3) (Hochholdinger et al., 2008) and its corresponding wild type (WT), B73, were grown in clear acrylic glass columns with a height of 25 cm and an inner diameter of 7 cm. To protect roots from light, columns were wrapped in aluminium foil. The bottom of the columns was covered by nylon mesh with 30-µm mesh size to prevent roots from growing out and to facilitate watering. The columns were filled with air-dry loam, sieved to 1-mm particle size, resulting in a soil volume of 885 cm³ with a bulk density of 1.26 g cm⁻³. The loam substrate was classified as a Haplic Phaeozem and was under agricultural use. It was obtained by excavating the top 50 cm of soil from an area in Schladenbach, Germany (Vetterlein et al., 2021). The water retention curve for the substrate and details of texture are likewise available in Vetterlein et al. (2021). Briefly, the permanent wilting point at pF 4.2 (pF representing the decadic logarithm of the absolute value of a soil matric potential of −15 800 hPa) for this substrate corresponds to a volumetric water content (VWC) of 6.2 % (v/v). The soil surface was covered with ~2 cm of gravel to reduce evaporation. Before filling, the soil was fertilized with 55.7 mg N as NH₄NO₃, 55.7 mg K as K₂SO₄, 27.9 mg Mg as MgCl₂ × 6H₂O and 44.6 mg P as CaHPO₄ per column. Seeds were sown after sterilization with 10 % hydrogen peroxide for 10 min, followed by rinsing three times with deionized water (dH₂O) and soaking in saturated CaSO₄ solution (2.65 g L⁻¹) for 3 h. Plants were grown for a total of 22 days after sowing (DAS) in a climate chamber with a 12/12-h day/night cycle. Average day/night temperature, relative humidity (RH) and photosynthetically active radiation (PAR) at column height were 29/20 °C (± 1.0 °C), 45/58 % (± 2.6 %) and 350/0 µmol m⁻² s⁻¹, respectively.

Initially, all columns were adjusted with dH₂O to a VWC of 22 % (v/v), corresponding to a soil matric potential of 282 hPa and a pF value of 2.45. An even water distribution within the columns was achieved by adding part of the water to the top and part to the bottom of the columns. This VWC was maintained by weighing every second day (DAS 1–9) and every day (DAS 10–15). On DAS 16, two treatments representing two different water levels (WLs) were established. One half of the plants were assigned to the well-watered treatment (abbreviated W) and watered twice per day from DAS 16 to DAS 22. For the other half of the plants, watering was completely stopped on DAS 16, which resulted in an increasing water limitation over time. To keep legends in figures readable, this treatment is termed the ‘drought’ treatment (abbreviated D). Both treatments lasted from DAS 16 to 22 (corresponding to day of treatment [DOT] 1–7), and were established with six replicates for each genotype (WT and rth3).

The experiment was carried out in two batches of maize plants grown in parallel in the same climate chamber to ensure comparable growth conditions, and the second batch was used for root exudate sampling. The use of two batches with six replicates each was necessary because the collection of root exudates required destructive sampling, involving rinsing of the roots. Consequently, it was not possible to collect soil and root samples for microbial and transcriptomic analyses, respectively, from the same batch of columns. To facilitate the root washing procedure, batch 2 columns were cut lengthwise into two halves held together by two metal ties. The plant growth of both batches was nearly identical, as indicated by similar final shoot dry weight of the plants (Supplementary Data Fig. S2).

Harvest of plant and soil material

On DAS 22, shoots were cut off close to the soil surface and total shoot fresh weight was determined. For every leaf of each plant, length and width were measured, and leaf area was calculated according to Lippold et al. (2022). Individual leaves were then cut off, weighed and used for different measurements as follows: the second youngest leaf was stored at −20 °C until measurement of osmolality. Two pieces (4 cm length each) of the third (middle part) and fourth (basal part) youngest leaves were immediately used for determination of the relative water content (RWC). The remaining fourth youngest leaf was dried at 65 °C until constant weight to determine nutrient contents. The remaining shoot was weighed again fresh and after drying to determine shoot water content and to quantify the total shoot dry weight.

Simultaneously, roots and soil fractions were sampled. The soil column was pushed out of the acrylic cylinder and was cut horizontally into 4-cm thick slices except the bottom slice, which was 7 cm in height. The 8- to 12-cm slice was halved vertically and used immediately after cutting to collect root samples and three different soil fractions. Roots of one half were quickly washed out with tap water using a sieve with 1-mm mesh size, shock-frozen and later used for RNA sequencing and determination of abscisic acid (ABA) and aquaporins. Roots of the other half with adhering soil were carefully taken out, and three soil fractions were subsequently collected. Soil that was not adhering to the roots was collected after removing the roots, and is termed ‘bulk soil’. Roots were then shaken gently to collect soil that was only loosely adhered to the roots, termed hereafter ‘loosely root-attached soil’ (LRA soil). Thereafter, roots were washed by vortexing, using a ratio of fresh root material (g) to sterile 0.3 % NaCl (mL) of 1:10. The resulting suspension represented soil that was tightly adhering to the roots, and is termed ‘rhizosphere’. The three soil fractions thus represent a gradient of increasing impact of roots on the soil. Loosely root-attached soil and rhizosphere were used for the analysis of soil extracellular enzyme activity, while microbiome composition was determined in bulk soil and rhizosphere as described below. All other soil slices were washed out as described above and roots were blotted dry to determine the root fresh weight. These roots were then stored in 50 % (v/v) ethanol to later determine further root morphological traits (length, surface, volume, diameter).

Root exudates were collected on DAS 22 starting 4 h after light was turned on using the soil–hydroponic hybrid approach as described previously (Oburger and Jones, 2018). The soil columns were opened, and the soil was removed by gentle rinsing with tap water for ~10 min. Once all soil particles were completely removed, the plants underwent three cycles of soaking in fresh dH₂O, each lasting 5 min. Root exudates were then collected for 2 h in 0.5 L of Milli-Q water (MQ water) containing 0.01 g L⁻¹ Micropur classic (Katadyn^®, Switzerland) as a microbial inhibitor agent. Root exudates were sampled under growth chamber conditions with sampling buckets being covered with aluminium foil to shield roots from direct light. At the end of the 2-h sampling period, plants were removed from the sampling solution and shoots were separated from the roots. A portion of the fresh root system was weighed and stored in 50 % (v/v) ethanol at 4 °C for determination of root morphological traits (see above). Root and shoot dry mass were determined after drying the remaining biomass at 65 °C for 72 h. The exudate samples were vacuum-filtered (0.2 µm, Cellulose Acetate OE 66, Whatman, UK) to remove all root debris. The filtered exudates were divided into multiple aliquots, frozen at −20 °C for 72 h and then stored at −80 °C until analysis.

Non-destructive monitoring of gas exchange and soil water properties

Evapotranspiration was determined daily by weighing every column from DAS 1 to 22 and was used to calculate plant cumulative water uptake. Average soil matric potential was calculated based on the measured volumetric water content (from weighing the columns) and the water retention curve that has been published previously in Vetterlein et al. (2021) and was obtained from HYPROP apparatus (UMS GmbH, Munich, Germany) by fitting the measured data to the bimodal Mualem–van Genuchten model (Durner, 1994).

Starting on DAS 15 (last day of watering for the drought treatment, corresponding to DOT 0), carbon assimilation and transpiration rates were measured daily starting 2 h after light was turned on and completed within the following 2 h, using the portable gas exchange fluorescence system GFS-3000 (Heinz Walz GmbH, Germany). Instrument settings of the measuring head were: PAR 1500 µmol m⁻² s⁻¹, 10 000 ppm H₂O, 29 °C, 400 ppm CO₂, and area 4 cm². The measurements were taken on the youngest fully expanded leaf at a distance of 10 cm from the leaf tip and avoiding the midrib.

Destructive physiological measurements at harvest

For determination of osmolality, the frozen second youngest leaf was subjected to two thawing/freezing cycles, and cell sap was collected by centrifuging the leaf for 5 min at 3600 g. Osmolality was measured in two technical replicates using a vapour pressure osmometer (Vapro, Model 5600, Elitech).

The RWC of the shoot was assessed using two 4-cm long pieces of the third and fourth youngest leaves, of which the midrib was removed. The leaf pieces were then weighed (FW), placed overnight into Petri dishes filled with MQ water, blotted dry, weighed to determine turgid weight (TW), dried for 3 d at 60 °C and weighed again to determine the dry weight (DW). RWC was calculated according to Wedeking et al. (2017) using the following formula:

RWC (%) = (FW – DW)/(TW – DW) × 100

The dried fourth youngest leaf was ground (mixer mill, type MM301, Retsch GmbH, Germany), and 50 mg of ground material was digested in a microwave (Ethos 550, MWT AG) together with 1.5 mL HNO₃ (65 %) and 0.5 mL hydrogen peroxide (30 %) for 3 min at 1400 W and 70 °C followed by 62 min at 1400 W and 210 °C. Nutrient concentrations were determined using inductively coupled plasma optical emission spectrometry (ICP–OES) (Optima 8300, Perkin Elmer, MA, USA).

Abscisic acid was quantified using a liquid chromatography–mass spectrometry (LC–MS/MS) method according to Balcke et al. (2012). Thirty milligrams of root powder was extracted with 500 µL methanol supplied with [²H₆]ABA (OlChemIm, Olomouc, Czech Republic) as the internal standard. After extraction and centrifugation (9610 g for 10 min), the samples were diluted with nine volumes of MQ water and subjected to solid-phase extraction on Chromabond HR-XC (Macherey-Nagel, Düren, Germany). The samples were eluted in 900 µL acetonitrile and concentrated at 35 °C to a final volume of ~100 µL (Concentrator Plus, Eppendorf, Hamburg, Germany). Ten microlitres of the eluate was subjected to LC–MS/MS and the ABA level was calculated using the ratio of analyte and internal standard peak heights.

Statistics of physiological measurements were determined using R statistical software (v.4.3.1; R Core Team, 2023) using the packages car, emmeans and multcomp. Normality and homogeneity of variance was tested with the Shapiro–Wilk test and Levene’s test, followed by a two-way ANOVA and Tukey’s HSD test (P ≤ 0.05). Statistical differences between ABA levels were calculated with the Kruskal–Wallis test and pairwise Wilcoxon test (P ≤ 0.05). Figures except Fig. 1 were created using the package ggplot2 and ggpubr.

Fig. 1.

Time course of cumulative water uptake (A, B), soil water content and soil matric potential (C, D) during the growth period of 21 d for WT (A, C) and rth3 mutant (B, D) plants. Plants were grown under well-watered conditions (W) or without further irrigation from DAS 15 onwards (D). Data points represent means (n = 6) and error bars represent the respective standard errors.

Root exudate analysis and root morphological traits

Total organic carbon (TOC) and total bound nitrogen (TNb) concentrations were determined with a Vario Elementar TOC analyser (Elementar Analysensysteme GmbH, Germany). Before analysis, 10 mL of all exudate samples was acidified with 80 µL of 10 % HCl to remove carbonates from the solution. The instrument was calibrated according to the manufacturer’s instructions using potassium phthalate (KHP, Elementar 35.00-0151) for TOC and a mix of ammonium chloride (Elementar, 35.00-0157) with sodium nitrate (Elementar, 38.00-0099) for TNb. To quantify the organic nitrogen content in the exudate samples, the concentrations of NO₃⁻ and NH₄⁺ were determined separately as described in Hood-Nowotny (2010). The NH₄⁺ concentration was below the limit of quantification in all samples and was therefore not included in the calculations. Organic nitrogen was then calculated by subtracting the concentrations of NO₃⁻ from TNb. Different target compound classes (sugars, phenolics, amino acids and proteins) were determined photometrically using a multiwell plate reader (Infinite^® 200 PRO Nano, Tecan, Switzerland). Prior to analysis, a 40 mL frozen exudate aliquot was lyophilized (−45 °C, 0.070 mbar; Alpha 1-4 LSCplus, Christ, Osterode am Harz, Germany) and resuspended in 2.5 mL of MQ water. Sugar concentrations were determined using the sulfuric acid–UV method as described by Albalasmeh et al. (2013). Briefly, 250 µL of exudate sample was mixed with 750 µL of 98 % sulfuric acid and the solution was vortexed for 30 s and quickly cooled down on ice. Absorbance was measured at 315 nm in a UV-suitable microwell plate (UV-Star^® flat-bottom, Greiner Bio-One, Austria). A calibration curve was made using glucose as a standard, and results are expressed as glucose equivalents. Total phenolics were quantified using the Folin–Ciocalteau method as described by Ainsworth and Gillespie (2007). Results were expressed as chlorogenic acid equivalents. Total free amino acids were determined fluorometrically with the o-phthalaldehyde and β-mercaptoethanol method (Jones et al., 2002) as described by Santangeli et al. (2024). Protein concentrations were assessed according to the Bradford dye-binding method (Bradford, 1976) using a Coomassie Protein Assay Kit (Thermo Scientific, Waltham, MA, USA). Bovine serum albumin (BSA) was used as calibration standard. Exudation rates of carbon, nitrogen and all compound classes were calculated taking into account the concentration factor, normalization per root surface area and sampling duration (cm⁻² h⁻¹). The relative contribution of each exudate compound class to the total exuded carbon was estimated as described in Santangeli et al. (2024).

Differences between experimental factors (i.e. genotype and water level) were evaluated by two-way ANOVA, followed by Tukey’s post hoc test (P ≤ 0.05). Data not following a normal distribution were log-transformed to meet the assumptions of parametric statistical tests. R² values were calculated by dividing the between-group sum of squares by the total sum of squares. Statistical analyses and graphs of root exudation were conducted using GraphPad Prisms 10.3.1 for Windows (GraphPad Software, San Diego, CA, USA).

To determine root morphological traits, roots collected from each soil slice were rinsed with dH₂O and evenly distributed in a flat tray on a flatbed scanner (Epson Expression 12000XL, Epson America Inc., Long Beach, CA, USA) to avoid overlapping of the roots. Scans were made with a resolution of 720 dpi, calibrated for image analysis with LA2400 Regent Instruments scanner software (Regent Instruments Inc., Quebec, Canada) and analysed using WinRHIZO software (WinRHIZO Pro 2022a, Regent Instruments Inc., Quebec, Canada). Root morphological traits of the 8- to 12-cm slice of the column used for root and soil sampling were estimated by calculating the average of the adjacent two slices. For batch 2 plants, morphological traits were calculated by extrapolating the values of the scanned root aliquot to the entire root system as described in Santangeli et al. (2024).

Analysis of root gene expression

Total root gene expression was analysed by RNA sequencing. Root material of four biological replicates per treatment was powdered under liquid nitrogen. Fifty milligrams of root powder was extracted with a NucleoSpin RNA Plant kit (Macherey-Nagel, Germany), following the manufacturer’s instructions. Elution was carried out with 40 µL RNase-free H₂O and performed twice to obtain a higher total amount of RNA. RNA quantity and quality were verified using a NanoDrop 8000 spectrophotometer (Thermo Fisher Scientific, USA) and a Bioanalyzer 2100 with a Plant RNA Nano chip (Agilent, USA). An RNA integrity number (RIN) >8 was obtained for all samples. Sequencing was carried out by the Genewiz sequencing facility (Leipzig, Germany) with paired-end 150-bp library design on the Illumina NovaSeq 6000 (Illumina, USA). Reads were trimmed using Trimmomatic (v.0.36) with default parameters. Mapping of sequences onto the Zea mays B73 RefGen_v5 reference genome was performed using HISAT2 (v.2.1.0). The featureCounts software of Subread (v.1.6.3) was used to obtain read counts. Statistical analyses were performed and figures generated using R (v.4.3.0; R Core Team, 2023). Differential gene expression was determined using the DESeq2 package (v.1.38.2), and the Benjamini–Hochberg method was applied to adjust P-values (Benjamini and Hochberg, 1995). Genes were considered as differentially expressed with P_adj < 0.05 and absolute log₂ fold change >1 for upregulated and log₂ fold change <−1 for downregulated genes. Differentially expressed genes (DEGs) were analysed for their biological functions using Gene Ontology (GO) annotation. Enriched GO terms were determined with the goseq package, utilizing the maize-GAMER GO annotation. Permutational analysis of variance (PERMANOVA) was performed using the vegan package (v.2.6-4) applying the Adonis test with Euclidian distances. RNA sequencing raw data were deposited in the NCBI Short Read Archive (SRA) under the BioProject PRJNA1190671.

The expression of selected genes was additionally determined by RT–qPCR. cDNA synthesis was performed with M-MuLV-Reverse Transcriptase (New England Biolabs, USA) with a total RNA concentration of 1 µg in a total reaction volume of 20 µL and a subsequent 1:10 dilution using nuclease-free water (Promega, USA). RT–qPCR was performed using GoTaq qPCR Master Mix (Promega, USA) and 2 µL cDNA dilution in 10 µL total reaction volume. The following protocol was applied in a 384-well thermocycler (CFX384 Touch TM Real-Time PCR Detection System, Bio-Rad, UK): 95 °C for 3 min, 45 cycles at 95 °C for 10 s followed by 58 °C for 40 s. The melting curve was created by augmenting the temperature from 60 to 95 °C in 0.5 °C steps, each step lasting for 5 s. To determine primer efficiency, a standard curve was generated including a pool of all analysed cDNAs diluted stepwise from a 1:1 dilution up to a 1:64 dilution. Primer efficiency as well as primer sequences are listed in Supplementary Data Table S1. Six biological replicates per treatment and per primer pair were applied in two technical replicates, from which the mean was used for further calculations. The geometric mean of two internal references (actin1 and glyceraldehyde 3-phosphate dehydrogenase) was used for data normalization of the relative expression of the analysed genes. To identify significant differences (P ≤ 0.05), a two-way ANOVA and subsequent post hoc Tukey test in OriginPro 2021 (v.9.8.0.200) were performed. Plotting was done with RStudio (v.4.3.0), including the packages ggplot2, ggpubr, rstatix, dplyr, grid, gridExtra and cowplot.

Analysis of soil microbiome composition and enzymatic activity

Amplicon sequencing of 16S rRNA gene fragments and ITS regions were used to determine the diversity and community composition of bacteria/archaea and fungi, respectively. Novogene (Cambridge, UK) both generated and sequenced amplicon libraries using NovaSeq PE250, as described by (Yim et al., 2022). The primers Uni341F (5′-CCTAYGGGRBGCASCAG-3′) and Uni806R (5′-GGACTACNNGGGTATCTAAT-3′) were used for the 16S rRNA gene fragments (Sundberg et al., 2013), while the primers gITS7 (5′-GTGARTCATCGARTCTTTG-3′) and ITS4 (5′-TCCTCCGCTTATTGATATGC-3′) were used for the ITS2 regions (Ihrmark et al., 2012). The DADA2 pipeline (Callahan et al., 2016) was followed to generate the amplicon sequence variants (ASVs). The representative sequence of each ASV was annotated using the SILVA SSU release 138.1 (Quast et al., 2013) and UNITE (Abarenkov et al., 2023) databases to obtain the corresponding taxonomic information and the abundance distributions for the 16S rRNA gene fragments and ITS regions, respectively. The ASV tables (abundance and taxonomy) of the 16S rRNA gene fragments and ITS regions were imported using the phyloseq package (McMurdie and Holmes, 2013). For subsequent analyses, all ASVs affiliated to chloroplasts, mitochondria and singletons were removed before rarefied reads at 36 219 and 60 993 per sample for bacteria/archaea and fungi, respectively, using the vegan package (Oksanen et al., 2020) . The bacterial/archaeal and fungal β-diversity was visualized using principal coordinates analysis based on the Bray–Curtis dissimilarity. The PERMANOVA test (vegan package, 10 000 permutations) was used to determine the impact of the water level on the analysed microbial community composition. The Wald test from Deseq2 was used to reveal significantly different relative abundances of bacterial, archaeal and fungal ASVs influenced by the water level. All analyses were performed using the statistical software R v.4.3.2. All raw sequences of the 16S rRNA gene fragments and ITS regions are available at NCBI within the sequence read archive accession PRJNA1065820.

Loosely root-attached soil and rhizosphere were further analysed for potential enzymatic activities related to organic carbon, phosphorus and nitrogen. To this end, fluorogenic substrates for β-glucosidase (4-methylumbelliferone-β-d-glucoside), acid phosphatase (4-methylumbelliferone-phosphate), leucine aminopeptidase (l-leucine 7-amino-4- methylcoumarin-hydrochloride) and N-acetylglucosaminidase (4-methylumbelliferone-N-acetyl-β-d-glucosaminide) were used. All substrates and chemicals were purchased from Sigma–Aldrich (Germany). To obtain a suspension of LRA soil with a soil/solution ratio comparable to that of the rhizosphere, 0.2 g of LRA soil was suspended in 20 mL of MQ water using low-energy sonication (40 J s⁻¹ output energy) for 1 min (Tian et al., 2020). Enzyme activity was determined in 96-well microplates using 50 µL of soil suspension, 50 µL of 0.1 m MES [2-(N-morpholino)ethanesulfonic acid] or 0.05 m Trizma buffer for 4-methylumbelliferone (MUF) or 7-amino-4-methylcoumarin-hydrochloride (AMC)-based substrates, and 100 µL of substrate at the saturating concentration of 200 µm (which was tested in preliminary experiments). Fluorescence was measured at 360/465 nm excitation/emission wavelengths and at a bandwidth of 35 nm with a plate reader (Tecan Infinite F200 Pro) after 150 min of incubation in darkness at room temperature and under continuous orbital shaking. Calibration curves based on either MUF or AMC were obtained for all treatments to express enzymatic activity as MUF or AMC release over time (nmol MUF/AMC g⁻¹ dry soil h⁻¹). To identify significant differences (P ≤ 0.05), a two-way ANOVA and subsequent post hoc Fisher LSD test were performed using R statistical software (v.4.3.1; R Core Team, 2023) with the packages rstatix and emmeans.

RESULTS

Plant physiological stress response and soil water status

Cumulative water uptake increased non-linearly over the course of the experiment, and was larger for WT compared with rth3 plants (Fig. 1). A separation of the well-watered and drought-stressed plants was observed from DOT 4 onwards for WT and from DOT 5 onwards for rth3. The difference between well-watered and drought-stressed plants on DOT 6 was significantly larger (P = 0.02) for WT than for rth3 (94 and 28 mL per plant per day, respectively). Exponentially declining soil water content and soil matric potential were observed after DOT 1 for both genotypes. This decrease was stronger for WT than for rth3, resulting in significantly lower soil water content (10 % v/v in WT versus 12 % v/v in rth3; P < 0.001) and soil matric potential at the end of the drought treatment in WT.

Transpiration and assimilation rates were similar in both genotypes until DOT 3 (Fig. 2A, B). Between DOT 3 and 7, rates decreased on average from 4.6 to 0.3 mmol (H₂O) m⁻² s⁻¹ (transpiration) and from 31.4 to 1.7 µmol (CO₂) m⁻² s⁻¹ (assimilation), reaching similar final values for both drought-treated genotypes. Responses in transpiration and assimilation rates under drought were delayed by 1 d for rth3 compared with WT. Transpiration rates and assimilation rates of well-watered plants were on a stable level during the treatment period without differences between genotypes.

Fig. 2.

Transpiration rates (A) and assimilation rates (B) of well-watered (W) and drought-stressed (D) WT and rth3 mutant plants during the final 7 d of the experiment. Data points represent means (n = 6), error bars represent standard errors, and different letters of the same format indicate significant differences between treatments on individual DOT (two-way ANOVA followed by Tukey’s HSD; P ≤ 0.05).

Under drought, leaf area (LA) and RWC were significantly lower compared with well-watered plants (Fig. 3A, D) for both genotypes. The decrease was stronger for WT compared with rth3, with a reduction of 36 % (RWC) and 44 % (LA) for WT, and 24 % (RWC) and 23 % (LA) for rth3. Under drought conditions, root length decreased for both genotypes, even though this was statistically significant only for WT (Fig. 3B). Again, the relative decrease was stronger for the WT (−35 %) compared with rth3 (−17 %). In contrast, root/shoot ratio was significantly reduced under water limitation for rth3, but not for WT (Fig. 3C), indicating a larger decline in root compared with shoot biomass for rth3. Shoot osmolality and root ABA levels increased significantly under drought for both genotypes (Fig. 3E, F). Again, the increase was more pronounced for WT compared with rth3, with a 2.3-fold increase in osmolality and 39-fold increase in ABA level for WT and a 1.6-fold increase in osmolality and 38-fold increase in ABA level for rth3.

Fig. 3.

Leaf area (A), total root length (B), root/shoot ratio based on dry weight (DW) (C), leaf relative water content (D), leaf osmolality (E) and root ABA concentration based on fresh weight (FW) (F) of well-watered (W) and drought-stressed (D) WT and rth3 plants at the last day of drought treatment. Different small letters indicate significant differences between treatments (two-way ANOVA followed by Tukey’s HSD; n = 6; P ≤ 0.05).

Root gene expression

Principal component analysis (PCA) of gene expression levels resulted in a separation between plant roots under drought and well-watered conditions along PC1, as well as a higher variance between samples under drought (Fig. 4A). According to PERMANOVA, 61.5 % of the variance between all samples can be explained by the factor water level, uncovering this factor as the main driver of gene expression levels (Fig. 4B). Between the water levels, 6042 genes were DEGs, of which 2012 were up- and 4030 downregulated under drought. Identified DEGs were screened for gene functions by performing a GO enrichment analysis. Among over-represented GO terms of genes that were upregulated under drought (Fig. 4C), we found the nutrition-related terms sucrose synthase activity, sugar transmembrane transporter activity, polysaccharide catabolic process, and carbohydrate transport, the cell wall-related term cell wall macromolecule catabolic process, and the ROS-related terms response to hydrogen peroxide and removal of superoxide radicals, as well as the terms chitin binding, response to heat, and citrate metabolic pathway. Over-represented GO terms of downregulated genes under drought (Fig. 4D) included the secondary metabolite-related terms aromatic compound biosynthetic process and terpene synthase activity, the oxidative stress-related terms peroxidase activity and response to oxidative stress terms, but also the terms cell wall, haem binding, calcium ion binding, response to auxin, oxidoreductase activity, and defence response. We found 550 DEGs between WT and rth3 roots, even though the genotype effect on root gene expression was not statistically significant (Fig. 4B; P = 0.051 explaining 6 % of the variance). A screening for genes related to osmotic adjustment and stress tolerance revealed 55 to be differentially expressed, the majority (39) of which appeared to be downregulated in rth3 (Supplementary Data Table S3). By analysing both genotypes separately, we observed a higher number of DEGs under drought in WT (5765) compared with rth3 (3427) (data not shown). However, GO term enrichment analysis revealed a highly similar pattern of DEGs between WT and rth3 (Supplementary Data Fig. S7), with the exceptions of the terms removal of superoxide radicals, xylose isomerase activity, citrate metabolic process, and defence response to fungus (up in WT only; Supplementary Data Fig. S7A), raffinose α-galactosidase activity, response to hydrogen peroxide, sucrose metabolic process, and carbohydrate transport (up in rth3 only; Supplementary Data Fig. S7B), acid phosphatase activity and response to auxin (down in WT only) (Supplementary Data Fig. S7C), and defence response to bacterium and terpene synthase activity (down in rth3 only; Supplementary Data Fig. S7D).

Fig. 4.

PCA (A) and PERMANOVA (B) of all expressed genes of 22-d-old WT and rth3 maize plants under well-watered (W) and drought (D) conditions. WL, water level; G, genotype; WL:G, interaction of water level and genotype. ***P < 0.001. (C, D) GO enrichment analysis of over-represented terms of upregulated (C) and downregulated (D) genes of both genotypes analysed together under drought compared with well-watered conditions. Number.DE, number of differentially expressed genes in this category; P.value, P-value of enrichment; Ratio.DE, ratio between differentially expressed and total genes in the GO category (n = 4) with adjusted P-value <0.05, absolute log₂ fold change >1.

Root exudation rates

The drought treatment significantly increased both carbon and nitrogen exudation rates of the WT, but not of the rth3 plants (Fig. 5A, B), resulting in significantly higher carbon exudation rates in WT compared with rth3 when irrigation was withheld. A similar pattern was obtained under drought for different compound classes (i.e. sugars, amino acids, phenols and proteins) in the root exudate samples, with significant increases in all classes for the WT plants, while only a trend towards increased exudation rates of these classes was observed for rth3 (Fig. 5C–F). Overall, water level explained most of the variance in the compound classes dataset (25–47 %). For exudation rates of phenols and proteins, the factor genotype was not significant and accounted for <3 % of the observed variability. In contrast, a significant genotype effect was found for sugars and amino acids, accounting for 29 and 10 % of the observed variance, respectively. For all compound classes, as well as for carbon and nitrogen exudation rates, a significant interaction was observed between water level and genotype. Under drought, the relative contribution of each compound group to the total carbon exudation was increased at the expense of unknown substances (Supplementary Data Fig. S4). While the relative contribution of sugars increased to a similar extent in both genotypes (+32 % in WT, +38 % in rth3), this relative increase was higher in the WT for phenolics (+72 % in WT, +13 % in rth3) and amino acids (+84 % in WT, +34 % in rth3).

Fig. 5.

Root exudation rates of 22-d-old WT and mutant (rth3) maize under two water levels (W, well-watered; D, drought), representing total organic carbon (µmol C cm⁻² h⁻¹) (A), organic nitrogen (nmol N cm⁻² h⁻¹) (B), sugars (nmol glucose equivalent cm⁻² h⁻¹) (C), proteins (µg protein cm⁻² h⁻¹) (D), total phenolic compounds (nmol chlorogenic acid equivalent cm⁻² h⁻¹) (E) and total free amino acids (nmol alanine equivalent cm⁻² h⁻¹) (F). All units are based on root surface area (RSA) in cm⁻². Differences between experimental factors [water level (WL), genotype (G) and the interaction of water level and genotype (WL:G)] were evaluated by two-way ANOVA followed by Tukey’s post hoc test (P ≤ 0.05). Different small letters indicate significant differences. ***P < 0.001, **P < 0.01, *P < 0.05 (n = 5 for well-watered treatments and n = 6 for drought treatments).

Soil enzyme activity and microbial community structure

Across all treatments, the activity of enzymes was 5.5, 5.1, 1.7 and 1.2 times higher in rhizosphere compared with LRA soil for acid phosphatase, β-glucosidase, N-acetylglucosaminidase and leucine aminopeptidase, respectively (Fig. 6A–D). No significant differences in enzyme activity were observed in any of the soil compartments between the well-watered WT and rth3 plants (Fig. 6A–D). In LRA soil, drought decreased the activity of β-glucosidase (significant for both genotypes) and of acid phosphatase (significant only for rth3) (Fig. 6A, B, right sides). In the rhizosphere, the drought-induced decreases in β-glucosidase and acid phosphatase activities were significant for WT, but not for rth3, resulting in a significantly higher β-glucosidase activity for rth3 compared with WT under drought (Fig. 6A, B, left side). For the WT, the relative reduction in enzyme activity under drought in the rhizosphere was much stronger (acid phosphatase, −39 %; β-glucosidase, −6 % of well-watered) than in the LRA soil (acid phosphatase, −7 %; β-glucosidase, −50 %), while for rth3 it was similar (acid phosphatase, −21 %) or even less (β-glucosidase, −19 %) than in LRA soil (acid phosphatase, −18 %; β-glucosidase, −59 %) (Fig. 6). No significant drought or genotype effect was observed for N-acetylglucosaminidase and leucine aminopeptidase activities in any of the soil compartments (Fig. 6C, D).

Fig. 6.

Enzymatic activity for acid phosphatase (A), β-glucosidase (B), N-acetyl-glucosaminidase (C) and leucine aminopeptidase (D) in LRA soil and rhizosphere (RS) of the 22-d-old maize genotypes WT and rth3 under two water levels: well-watered (W) and drought (D). Differences between experimental factors [water level (WL), genotype (G) and the interaction of water level and genotype (WL:G)] were evaluated by two-way ANOVA followed by Fisher’s LSD test (P ≤ 0.05, n = 6) individually for each soil fraction. Different letters of the same format between treatments in rhizosphere or LRA soil indicate significant differences within one soil fraction. ***P < 0.001, **P < 0.01, *P < 0.05.

Microbial β-diversity was analysed for bacteria/archaea and fungi either in the two soil compartments individually or for both compartments together (Fig. 7). When all samples were analysed together, the soil compartment had the strongest effect on the microbiome composition, explaining 21 % (archaea/bacteria) and 13 % (fungi) of the variance (Fig. 7E). The effect of water level on microbial composition was also significant and stronger for fungi (6 % of the variance) than for bacteria/archaea (3 % of the variance). The genotype effect was significant for bacteria/archaea, but not for fungi. When rhizosphere and bulk soil were analysed individually, the genotype effect on bacterial/archaeal composition was significant in both compartments (Supplementary Data Fig. S5A, C), even though the effect was stronger in the rhizosphere (9 % of the variance) compared with the bulk soil (6 % of the variance). At the level of ASVs, more fungal ASVs had a substantial change in relative abundance due to water level than did bacterial and archaeal ASVs (Supplementary Data Table S2).

Fig. 7.

Principal coordinate analyses (PCoA) based on Bray–Curtis dissimilarity for bacterial and archaeal (left, A, C) and fungal (right, B, D) community composition in the two soil compartments (Co) bulk soil (bottom) and rhizosphere (top) analysed individually (A–D) of 22-d-old maize genotypes (G) WT and rth3 under the two water levels (WL): well-watered (W) and drought (D). PERMANOVA results are shown in the upper left corners for individually analysed soil compartments and in the table for both soil compartments analysed together (E). *P < 0.05, **P < 0.01, ***P < 0.001; n = 6. No significant interactions were detected between WL, G (WL:G) and Co for any treatments (bacteria and archaea or fungi).

Shoot nutrient concentrations

Multiple nutritional elements were analysed in the leaf material, but only nitrogen, phosphorus and potassium concentrations are shown, as these were the most interesting in terms of growth, root hairs and water relations (Fig. 8). Under well-watered conditions, leaf nitrogen concentration was significantly lower for WT compared with rth3, whereas no significant difference was observed between the two genotypes for phosphorus and potassium (Fig. 8A–C). In both genotypes, drought reduced the concentrations of phosphorus (significant for both genotypes) and potassium (significant only for rth3) (Fig. 8B, C). The largest reduction was observed for phosphorus concentrations, with a 30 % decrease for WT and a 41 % decrease for rth3 compared with the respective well-watered treatment, resulting in a significantly lower phosphorus concentration in rth3 compared with WT (Fig. 8B). Drought also reduced the nitrogen concentration in rth3, but increased that in WT (Fig. 8A).

Fig. 8.

Leaf nitrogen (A), phosphorus (B) and potassium (C) concentrations of the well-watered (W) and drought stressed (D) maize genotypes WT and rth3. Different letters indicate significant differences between treatments (two-way ANOVA followed by Tukey’s HSD; n = 6; P ≤ 0.05).

DISCUSSION

Genotypic differences in plant size may contribute to different stress intensities

After 7 d of water limitation, both genotypes exhibited clear signs of drought stress, including reduced leaf area, root length and biomass, as well as physiological changes like reduced RWC, increased osmolality or ABA levels (Fig. 3), reduced transpiration and assimilation rates (Fig. 2), and differential expression of >6000 root genes. Both genotypes responded to the water limitation in an overall similar way, as indicated by similar GO term enrichment patterns (Supplementary Data Fig. S7), similar wilting phenotypes at harvest (Supplementary Data Fig. S1) and the lack of a significant genotype effect on root gene expression (Fig. 4).

Despite this similarity in response, the drought-induced fold change was typically larger in WT compared with rth3 for most parameters (Fig. 3, Supplementary Data Fig. S2), suggesting a greater intensity of drought stress in the larger WT plants. This is also reflected in an earlier onset of declines in assimilation and transpiration rates in the WT (starting on DOT 4) compared with rth3 (starting on DOT 5) (Fig. 2), corresponding to slightly (though not significantly) higher ABA levels in WT roots at harvest. Soil water content, and thus soil matric potential, seemed to hit critical thresholds on DOT 4/5 (WT/rth3), when plants began to reduce the transpiration rate (Fig. 1). These thresholds of ~14 % (v/v) soil water content or −135 kPa soil matric potential were also observed in other drought stress experiments conducted with the same genotypes and soil (e.g. Cai et al., 2021). This soil matric potential aligns well with the threshold potential of −0.1 MPa, below which the soil may control root water uptake in most soils (Draye et al., 2010). Still, at harvest, transpiration and assimilation rates of both drought-stressed genotypes had reached comparable values near zero, and the strongly enriched GO terms related to ROS detoxification in roots of both WT and rth3 support the occurrence of cellular damage due to oxidative stress in both genotypes.

Carbohydrate-related GO terms were over-represented among upregulated genes (Fig. 4C, D; Supplementary Data Fig. S7) in both genotypes, which is consistent with previous reports (Opitz et al., 2014; Kang et al., 2022) suggesting a relation to osmoregulatory processes. Indeed, at the gene expression level, 55 DEGs related to osmoregulation and stress were found between the two genotypes, of which 39 changed to a larger extent in WT compared with rth3 (Supplementary Data Table S3), again confirming the stronger stress response perceived by the WT plants. In addition, we also observed an over-representation of GO terms related to modification of cell wall structure, which might enable roots to maintain elongation under water limitation by cell wall loosening despite reduced turgor (Opitz et al., 2015) .

It would be short-sighted to interpret the observed differences between genotypes in drought stress intensity only as differences in physiology. Rather, the higher stress intensity in WT compared with rth3 can at least in part be explained by the larger plant size and shoot biomass production of WT under well-watered conditions (Supplementary Data Fig. S2), which goes along with a greater leaf area and root length (Fig. 3A, B). Especially the greater leaf area is responsible for the significantly higher cumulative water loss of WT, despite similar transpiration rates per unit leaf area (Fig. 2). The smaller biomass of rth3 compared with its corresponding WT was previously also observed in pot experiments (Ma et al., 2021) as well as in the field (Vetterlein et al., 2022). Interestingly, rth3 displays a higher root/shoot ratio compared with the WT irrespective of the water level (Fig. 3C), possibly related to a higher relative investment of rth3 plants in root growth, as previously also observed under field conditions (Vetterlein et al., 2022).

Larger plants consume the available water in a limited soil volume more rapidly than smaller plants, especially during the exponential vegetative growth phase. In order to disentangle the effects of plant size from those of physiology (e.g. presence of root hairs), we correlated stress responses with cumulative transpiration (Supplementary Data Fig. S6). We chose the cumulative transpiration on DOT 6 as a proxy for plant size at harvest, based on its highly significant correlation with the shoot dry weight (Supplementary Data Fig. S5). The significant correlations between cumulative transpiration and root ABA levels (P ≤ 0.001) or leaf osmolality (P ≤ 0.05) (Supplementary Data Fig. S6G, H) indicate that plant size could at least to some extent explain the observed genotypic differences in these stress responses. Notably, plant size explained to a lesser extent the genotypic differences in exudation rates, although for carbon, sugars and amino acids comparable trends were observed (correlation coefficient between 0.43 and 0.53; not significant at P ≤ 0.05) (Supplementary Data Fig. S6A–F). This suggests that additional factors, e.g. root hairs, might be involved in these responses. The problem of plant size is, however, often overlooked in drought stress pot experiments using different genotypes. Ways to overcome this issue include a different duration of the drought treatment for different genotypes, e.g. based on targeted reductions in leaf gas exchange, but this results in plants of different age and development at the time of harvest. Alternatively, small amounts of water could be added to the larger genotypes under drought, to maintain similar soil water contents. However, the daily addition of water (even in small amounts) would likely interrupt the progression of the drought response, which is also complicating the interpretation of the results. In the present study and for the questions addressed here, the benefit of harvesting both genotypes at the same age and developmental stage was considered more important than obtaining identical stress levels. However, applying similar stress levels to different genotypes in drought experiments remains an important challenge which is often underestimated, especially when screening larger selections of genotypes.

Interestingly, overall gene expression analysis (comparing all DEGs of both genotypes) did not show a significant difference between the genotypes (P = 0.051), whereas physiological measurements (RWC and osmolality) already revealed highly significant differences (P < 0.010). These results are fully consistent with an overall similar direction of stress response, i.e. similar metabolic pathways are up- or downregulated in both genotypes, though to a different extent, as would be expected for the same plant species. This could mask more subtle differences in the amplitude of the response. Combining gene expression and physiological analyses may thus be beneficial in unravelling smaller differences between genotypes, as both provide unique and complementary pieces of information.

Effect of water limitation on microbial composition and enzyme activities differs with distance from the roots

Roots can shape their rhizosphere through the release of exudates and rhizodeposition, enhancing nutrient availability, water retention, root growth, interactions with beneficial microorganisms and soil stability during soil drying (Carminati et al., 2017; Vives-Peris et al., 2020). Our results support this, showing significant differences in bacterial/archaeal and fungal community composition between rhizosphere and bulk soil, regardless of water levels (Fig. 7E). Additionally, rhizosphere activities of β-glucosidase and acid phosphatase were 5.1–5.5 times higher than in LRA soil, irrespective of the water level (Fig. 6). Even though these enzymes can be produced by both microorganisms and roots, the fraction released by microbes tends to be larger than that released from plant roots (Zhang et al., 2023). Our results thus indicate that root-derived exudates/rhizodeposits may have boosted rhizosphere microbial activity, similar to what was reported previously (Ahmed et al., 2018).

Water level significantly impacted microbial β-diversity in bulk soil and β-glucosidase and acid phosphatase activities in LRA soil (Figs 6 and 7E), emphasizing the universal requirement of water for enzyme functionality and substrate availability (Sardans and Peñuelas, 2005). Interestingly, microbial diversity was not significantly affected by water level in the rhizosphere (Fig. 7A, B), possibly related to generally higher moisture levels in the rhizosphere compared with the bulk soil. Roots are able to release water into the rhizosphere to facilitate nutrient uptake, especially in drying soil (Caldwell et al., 1998). In addition, the observed increase in root exudates may indirectly also contribute to a higher water content of the rhizosphere, since e.g. mucilage can enhance the water-holding capacity of the soil under water limitation (Carminati et al., 2017). Soil microbes may benefit from this, as they rely on adequate moisture for growth (Naylor and Coleman-Derr, 2018). Thus, the buffering capacity of the rhizosphere under water limitation may stabilize microbial diversity compared with the bulk soil, resulting in a stronger water level effect more distant from the roots (Fig. 7A–D). However, strong daily fluctuations in rhizosphere soil water content, required gradients of water availability for hydraulic lift and the temporal offset between cutting the shoot and collecting the rhizosphere samples at harvest all add unknowns. Hence, these points would need to be investigated in more detail.

Short-time water limitation leads to shifts in microbial diversity and a stronger effect on fungal than bacterial/archaeal composition

In the bulk soil, the stronger water level effect on the β-diversity of fungi compared with bacteria/archaea (Fig. 7C, D) and the higher number of significantly altered fungal ASVs from the Ascomycota phylum in both genotypes (Supplementary Data Table S2) were unexpected, as fungi are usually more resilient to low water availability than bacteria/archaea (de Vries et al., 2018; Ochoa-Hueso et al., 2018), due to their chitin cell wall and their capability to mobilize nutrients via the hyphal network (Homa et al., 2022). Possibly, water limitation altered the community composition in a manner that increased its resilience to water limitation (Preece and Peñuelas, 2016). Overall, it is interesting that a short drought period of just a few days was sufficient to induce significant shifts in microbial diversity, which is typically associated with longer drought periods in the existing literature (Carbone et al., 2021; Jaeger et al., 2023; Swift et al., 2024).

Water level causes reallocation of assimilates modifying microbial activity and plant nutrition

In the rhizosphere, water limitation reduced β-glucosidase and acid phosphatase activities, with a significant response only for the WT (Fig. 6A, B). This water level × genotype interaction contrasts with the LRA soil, where relative reductions in enzymatic activities were similar between the two genotypes (at least for β-glucosidase). The main differences between the two genotypes were the root hair presence and the more intense stress of the larger WT plants (as discussed above). In the present setting, we cannot differentiate between the impact of these two components of the stress response, as they may be interrelated, and thus we refrain from speculations regarding the contribution of root hairs in the following sections, and focus on the genotypic differences in root exudates, irrespective of their cause. However, root exudation rates were less tightly correlated with plant size compared with other parameters (e.g. RWC or osmolality), suggesting a possible contribution of root hairs to exudation. We would like to also point out that the postulated effect of root exudates does not replace the general effect of reduced water content on enzymatic activities, but is an overlaying process further explaining the complex situation in the rhizosphere.

In the WT, the initial drought response was a decrease in transpiration and assimilation rates from DOT 4 to 7 (Fig. 2), followed by increased exudation rates of all compound classes under water limitation (Fig. 5) and an increased contribution of sugars, amino acids and phenolics to the overall carbon exudation (Supplementary Data Fig. S4). At the same time, genes related to sugar transporter activity and carbohydrate transport in roots were upregulated under water limitation (Fig. 4). Especially carbohydrates are easily accessible carbon sources for microorganisms. Their increased exudation has been previously observed under water limitation (Karlowsky et al., 2018; Santangeli et al., 2024) and can be one reason for shifts in microbial communities (Lopes et al., 2022; Seitz et al., 2022). Typically, the redistribution of assimilates from shoots to roots occurs during water limitation (Prescott et al., 2020), serving three different functions. Firstly, it may contribute to an increased passive exudation of osmotically active compounds in roots (Oburger and Jones, 2018). The increased exudation rates of sugars and amino acids in WT roots, as well as their increased contribution to total exuded carbon in both genotypes are in line with this and indicate that both genotypes adjusted to the water limitation by lowering root water potential via accumulation of osmotically active compounds and increasing soil water-holding capacity via mucilage accumulation. Secondly, plant growth is more rapidly inhibited at the onset of water limitation compared with photosynthesis, and thus release of carbon compounds from roots might also help to get rid of excess assimilates (Muller et al., 2011). Thirdly, it provides carbon and energy to enhance root growth and the ability to forage for water in deeper soil layers under water limitation (Kang et al., 2022). As the root/shoot ratio and root length of our plants was rather decreased under water limitation (Fig. 3B, C), the increased carbon allocation to the roots in the present experiment is unlikely to have been driven by root growth and thus root sink strength. Typically root/shoot ratios increase under water limitation (Dietz et al., 2021), but this is only feasible in soils drying out from top to bottom, where roots have the possibility to reach deeper and moister soil layers. The unusual reduction of root/shoot ratios observed in the present study might be related to the experimental setup, which restricted deeper root growth by pot geometry, and where soil drying occurred almost simultaneously throughout the soil columns.

Functions of root hairs are most likely strongest in the close proximity of roots, where we observed significant differences in β-glucosidase activity between genotypes (Fig. 6B, rhizosphere soil). β-Glucosidases are enzymes produced by microorganisms that degrade various glycosides and oligosaccharides, and their activity may be reduced under water limitation due to a reduction in microbial growth (Metze et al., 2023) and substrate diffusion. Increased sugar exudation rates in the WT might have enhanced the availability of easily degradable organic carbon, reducing the need for β-glucosidase as a tool to use more complex organic compounds for microbial growth, and compensating in part the drought effect on β-glucosidase activity. The lack of effect in rth3 is in line with no change in root exudation rates, and possibly related to the absence of root hairs. However, β-glucosidase activity declined for both genotypes in LRA soil, indicating overall water deficit effects on enzyme activity, irrespective of the genotype.

Similar to β-glucosidase, the activity of acid phosphatase was reduced more strongly in the rhizosphere compared with LRA soil for WT, but not for rth3 plants (Fig. 6A). Acid phosphatase is an important enzyme in the soil, hydrolysing organic phosphorus and thus helping to mobilize phosphorus (Ma et al., 2021), which is often limited under water limitation (He and Dijkstra, 2014). The reduction in acid phosphatase activity aligns with an over-representation of acid phosphatase activity in the differentially downregulated genes of the WT, but not of rth3 (Supplementary Data Fig. S7). Phenolics are also associated with phosphorus mobilization by decreasing rhizosphere pH and increasing phosphorus solubilization (Hu et al., 2005). It is tempting to speculate that increased exudation rates of phenolics in WT, but not rth3, improved the availability of phosphorus as substrate for microbes and reduced the need for microbial acid phosphatase activity in the rhizosphere of WT plants. Root hairs facilitate phosphorus uptake for plants (Klamer et al., 2019; Ma et al., 2021), and they harbour several phosphorus uptake transporters (Bienert et al., 2021). The higher expression of the phosphorus uptake transporter PHT12 in rth3 compared with WT under well-watered conditions, and its stronger upregulation under water limitation in rth3, might indicate compensation for the lack of phosphorus uptake related to missing root hairs (Supplementary Data Fig. S3). The lower leaf phosphorus concentrations in rth3 compared with WT under water limitation, but not when well-watered (Fig. 8), suggests that the lack of root hairs was disadvantageous primarily under water-limited conditions, where upregulation of phosphorus uptake transporters was not sufficient to overcome this deficit. Still, since stress intensity was lower in rth3 compared with WT, it remains unclear whether rth3 could enhance phosphorus uptake under more severe drought conditions via an increased root exudation rate.

Different stress intensities are reflected in regulation of oxidative stress response and aquaporin expression

Oxidative stress is the most damaging aspect of drought stress in plants and associated microorganisms (Hartman and Tringe, 2019), marking the transition from medium to severe stress level and potential cellular damage (Wedeking et al., 2018). In the present study, genes involved in superoxide radical removal and response to hydrogen peroxide were among the most enriched GO terms in drought-stressed roots of both genotypes (Fig. 4), indicating sufficient stress to activate antioxidative mechanisms. The enrichment of GO terms related to ROS detoxification in both up- and downregulated genes suggests complex regulation of ROS processes in drought-affected maize roots (Opitz et al., 2016; Li et al., 2017; Yang et al., 2022). Prior research has shown that plant roots can produce and exude ROS under drought stress (Huang et al., 2017), which can inhibit specific microorganisms, particularly Gram-negative bacteria (Xu et al., 2018), leading to characteristic changes in bacterial community compositions (Hartman and Tringe, 2019). In our study, a significant upregulation was observed for the tonoplast intrinsic protein TIP3;4 in drought-stressed WT, but not in rth3, suggesting once more a higher stress intensity in this genotype (Supplementary Data Fig. S3). Increased expression of TIP proteins under water limitation was found previously by Romero-Munar et al. (2024). Belonging to aquaporins, TIPs can transport other molecules except water, like hydrogen peroxide (Mukherjee et al., 2024), indicating potential hydrogen peroxide detoxification under oxidative stress. Even though ROS were not directly determined in this study, it is tempting to speculate that exudation of ROS or ROS scavengers contributed to the observed shifts in bacterial/archaeal composition (Fig. 7A, C). The larger increases in root exudation rates of WT plants may have further contributed to the genotypic differences in β-diversity of bacteria/archaea (Fig. 7A, C). Unlike TIP3;4, we found significant drought-induced downregulation of the aquaporin PIP2;3 in both genotypes and PIP2;1 in WT roots (Supplementary Data Fig. S3). PIPs are involved in water uptake and release (Hachez et al., 2006), and the observed downregulation aligns with previous findings on drought-stressed maize (Mahdieh et al., 2008; Protto et al., 2024). This may be an adaptive response aimed at minimizing water loss and backflow into the drying soil (Šurbanowski et al., 2013). PIP2;1 is expressed in maize root hairs (Fetter et al., 2004), though its function remains debated (Bienert et al., 2021). In our study, the expression of PIP2;1 was significantly lower in rth3 roots under well-watered conditions compared with the WT and was not further reduced under drought conditions, potentially reflecting the absence of root hairs in the rth3 genotype.

Conclusions

By withholding water for just a few days, we induced changes in maize biology that extended to the functionality and community composition of the rhizosphere microbiome. In root proximity (rhizosphere soil), the impact of water limitation on the microbiome was mediated by the presence of root hairs.

Genotype-specific reallocation of assimilates modified extracellular enzyme activities by supplying easily degradable C-rich compounds and facilitating phosphorus mobilization in the WT plants. Conversely, at larger distance from roots (bulk and LRA soil), the overall reduction in soil water content was the driving force for shifts in microbial diversity. Here, water limitation more strongly altered the composition of the fungal compared with the bacterial/archaeal microbiome, possibly resulting in a more drought-resilient community structure.

While both maize genotypes responded overall similarly to water limitation, the stress intensity differed, with the root hairless genotype exhibiting a less drastic response at all levels of investigation. This milder reaction could be attributed, at least in part, to its smaller size at the onset of water limitation, underscoring the need to consider genotype-specific size differences in drought stress screening trials.

Our interdisciplinary approach provides a comprehensive snapshot of how plants interact with the surrounding soil and microorganisms during short-term water limitation. The complex stress response network highlighted in this study emphasizes the importance of integrative research for enhancing plant resilience and supporting their adaptation to the increasing climate change challenges.

SUPPLEMENTARY DATA

Supplementary data are available at Annals of Botany online and consist of the following.

Figure S1: well-watered and drought-stressed 22-d-old maize of the genotypes WT and rth3 grown in soil columns on the last day of drought. Figure S2: shoot dry weight (DW) at the last day of drought treatment of batch 1 (A) and batch 2 (B) of 22-d-old well-watered (W) and drought-stressed (D) maize genotypes WT and rth3. Figure S3: RT–qPCR (a) and RNAseq (b) based expression analysis of aquaporin (PIP, TIP), potassium channel (TPPC-b) and phosphate transporter (PHT) genes in roots of 22-d-old maize WT and mutant rth3 genotypes grown under well-watered (W) and drought (D) conditions. Figure S4: estimated proportional contributions of various exudate compound groups to the overall carbon exudation of the two maize genotypes, WT and rth3, grown under the two water levels, well-watered (W) and drought (D). Figure S5: cumulative transpiration on DOT 6 as a function of shoot DW for batch 1 (A) and batch 2 (B) maize plants of the genotypes WT and rth3 under the water levels well-watered (W) and drought (D), respectively. Figure S6: root exudation rates of total organic carbon (µmol C cm⁻² h⁻¹) (A), organic nitrogen (nmol N cm⁻² h⁻¹) (B), proteins (µg protein cm⁻² h⁻¹) (C), sugars (nmol glucose equivalent cm⁻² h⁻¹) (D), total phenolic compounds (nmol chlorogenic acid equivalent cm⁻² h⁻¹) (E) and total free amino acids (nmol alanine equivalent cm⁻² h⁻¹) (F) based on root surface area (RSA) in cm⁻² and root ABA concentration (G) and leaf osmolality (H) of 22-d-old maize genotypes WT and rth3 under the water level drought (D) as a function of cumulative transpiration on DOT 6. Figure S7: GO enrichment analysis of over-represented terms of upregulated (A, B) and downregulated (C, D) genes under drought for WT (A, C) and rth3 (B, D) compared with well-watered conditions. Table S1: RT–qPCR primer sequences used in this study and their efficiencies. Table S2: relative abundance (enrichment or depletion) of ASVs (mean ± s.e.m.) with significant differences between the pairwise comparisons, log₂ fold change at P_adj < 0.05 and n = 6 (Wald test) of bacterial and archaeal and fungal ASV levels in the two soil compartments (bulk soil and rhizosphere) of the two maize genotypes WT (a) and rth3 (b) under the two different water levels (WLs): well-watered (W) and drought (D). Table S3: DEGs in maize roots related to osmotic adjustment and stress tolerance.

FUNDING

This work was supported by the Deutsche Forschungsgemeinschaft (DFG; German Research Foundation) within the framework of the Priority Programme 2089 Rhizosphere Spatiotemporal Organisation – A Key to Rhizosphere Functions (403626025 to M.A.W., 403637238 to K.S., 403664478 to E.B., 403625794 to P.B., 403803214 to E.O., 403641192 to M.T. and 403640293 to D.V.).