Evaluation of combined root exudate and rhizosphere microbiota sampling approaches to elucidate plant–soil–microbe interactions

Summary

• Deciphering the root exudate‐driven interplay between plants and the rhizosphere microbiota is essential for understanding plant adaptation to the environment and future‐proofing crop production. However, sampling root exudates and rhizosphere soil remains challenging due to the low throughput and destructive nature of the process.

• We used the staple crop barley (Hordeum vulgare L.) as a model to benchmark different sampling approaches for simultaneous exudation and microbiota profiling of soil‐grown plants.

• Exudate profiles and total dissolved organic carbon exudation rates were consistent across different sampling approaches, whereas root biomass, root morphology measurements, and organic nitrogen exudation varied. High‐throughput amplicon sequencing and quantitative PCR (qPCR) of phylogenetic markers and nitrogen cycle‐selected genes revealed a protocol‐specific footprint in the composition and abundance of rhizosphere bacterial and fungal microbiota. Yet, on average, 75% of microbes enriched in and differentiating between barley rhizosphere and unplanted soil controls were recovered across all the sampling approaches evaluated.

• Our results demonstrated that, under the tested conditions, different sampling approaches produced comparable microbiota and exudation patterns, enabling the integrated study of root exudation and microbial profiles from the same plant. The observed differences across sampling approaches must be considered according to the experimental scope.

Introduction

Plant–soil–microbe interactions in the rhizosphere, or the interface between roots and soil, play a pivotal role in plant health and growth performance. Therefore, the rhizosphere can be regarded as an ‘extended’ root phenotype (de la Fuente Cantó et al., ²⁰²⁰). Soluble root exudates, that is, plant metabolites released via the root into the surrounding soil, are key drivers of biogeochemical processes in the rhizosphere, as they shape the abundance, function, and composition of the microbial communities populating the rhizosphere, designated ‘rhizosphere microbiota’ (Anderson et al., ²⁰²⁴). Harnessing beneficial rhizosphere traits, including root exudation and the rhizosphere microbiota, is a critical step toward more sustainable agriculture (Schneider & Lynch, 2020; Brooker et al., ²⁰²²; Oburger et al., ²⁰²²; Colombi et al., ²⁰²⁴). Significant progress has been made in phenotyping (i.e. measuring the plant growth, development, and physiology arising from interactions between genotypes and environment) of root architectural traits, paving the way for including these traits in breeding programmes (Paez‐Garcia et al., ²⁰¹⁵). However, investigating complex rhizosphere traits such as plant–soil–microbe interactions remains challenging, especially at a throughput relevant to plant breeding.

Diverse toolboxes to study root exudation (Oburger & Jones, 2018), the rhizosphere microbiota (Alegria Terrazas et al., ²⁰¹⁶), and additional rhizosphere traits such as the rhizosheath formation, pH, and elemental gradients have already been established (George et al., ²⁰⁰⁸; Neumann et al., ²⁰⁰⁹; Oburger & Schmidt, ²⁰¹⁶). However, most techniques are often tailored for in‐depth studies and/or handling a limited number of genotypes and/or treatments, preventing a full appreciation of how plant genetics impacts rhizosphere traits. Likewise, most techniques and sampling approaches focus on individual aspects of the rhizosphere, for example, either exudates or the microbiota, limiting a holistic insight into feedback processes. For instance, root exudation and microbiota composition, and their subsequent impact on plant performance, are often inferred from independent samples rather than the same plant specimen due to handling complexity (Escudero‐Martinez et al., ²⁰²²).

Phenotyping of complex rhizosphere traits, such as root exudation, is methodologically challenging and labour‐intensive. Efforts to quantify root exudation from soil‐grown plants are confounded because the collection of rhizosphere soil does not enable accurate determination of root exudation rates, as the samples also contain dissolved native soil organic matter (Oburger & Jones, 2018). Furthermore, exuded metabolites can immediately interact with the soil matrix (sorption) and be instantly metabolised by the rhizosphere microbiota (Oburger et al., ^{2009, 2011, 2016}). To circumvent soil matrix interactions during the sampling process, root exudates are often collected from plants grown in (sterile) hydroponics or inert, artificial growth substrates (e.g. Kawasaki et al., ²⁰¹⁸; Lopez‐Guerrero et al., ²⁰²²; Aleksza et al., ²⁰²⁴; Otxandorena‐Ieregi et al., ²⁰²⁴). Artificial growth conditions also enable better control of microbial activity, especially during the exudate sampling process. Microbial decomposition in non‐sterile exudate sampling has repeatedly been shown to alter metabolite concentrations (Valentinuzzi et al., ²⁰¹⁵; Otxandorena‐Ieregi et al., ²⁰²⁴). Therefore, the control of microbial activity during the sampling process is critical if the aim is to obtain unbiased information about the quantity and diversity of plant‐exuded metabolites (Oburger & Jones, 2018). However, the ecological relevance of artificial growth environments, which are distant from field conditions, may fail to recapitulate rhizosphere processes (Oburger & Jones, 2018). If the aim is to investigate plant–soil‐microbe interactions, working with soil‐grown plants is essential. Among protocols using soil‐grown plants, the soil‐hydroponic‐hybrid technique combines plant–soil growth with a short hydroponic exudate collection period. This approach has been frequently used for individual, excavated root segments (Smith, 1970; Phillips et al., ²⁰⁰⁸), but also for entire root systems (Oburger et al., ²⁰¹⁴; Santangeli et al., ²⁰²⁴), and is probably the most applicable approach for investigating root exudation in the context of rhizosphere processes at an ecologically relevant scale. Typically, exudates from large plants, such as mature trees, are collected from excavated yet intact root segments, whereas for small to medium‐sized plants, like agricultural crops, the entire root system is excavated and sampled for exudates. The soil‐hydroponic‐hybrid technique requires (1) the excavation of an intact root system, followed by (2) careful soil removal by root washing (to avoid root damage), (3) a short root damage capture/osmotic adjustment period, and (4) 2–4 h (h) of hydroponic exudate sampling. The replicate throughput of this protocol is sufficient for medium‐scale experiments (c. 100 samples), including genotypes and/or treatments. Also, despite the sudden change in environment from soil to solution, the root cell environment will still reflect natural soil growth conditions, especially when the sampling period is short.

Like root exudates, experimental protocols have been developed to sample the rhizosphere soil for microbiota analysis, designated ‘rhizosphere fractionation’. Typically, the bulk soil assumed to be unaffected by root activity is removed by gently ‘shaking’ the roots until only the soil tightly adhering to the root remains. This root‐adhered soil is then defined as the rhizosphere soil (Escudero‐Martinez et al., ²⁰²²; Oyserman et al., ²⁰²²; Chang et al., ²⁰²⁵). In a subsequent step, the rhizosphere soil is harvested from the roots. This can be achieved by gently brushing roots (e.g. with a sterile toothbrush) (Lucas et al., ²⁰¹⁸; Fukuda et al., ²⁰²²). However, this technique is time‐consuming, may cause root damage, and carries brushed root cellular material into the sample. To address these limitations, a root washing step to harvest the rhizosphere is commonly applied. In this approach, the bulk soil is removed via handshaking, with the remaining rhizosphere‐coated roots being submerged in a buffer solution, followed by vortexing to facilitate the rhizosphere separation. Finally, the rhizosphere soil is centrifuged to create a rhizosphere soil pellet for microbial DNA extraction. This high‐throughput method, capable of handling c. 100 samples, was found to consistently retrieve a set of microbes significantly different from the unplanted soil (Alegria Terrazas et al., ²⁰²²; Reid et al., ²⁰²⁴).

Both rhizosphere microbiota profile and root exudate composition data have already been successfully combined with other root and shoot traits, including root morphology, shoot nutrient content, and root gene expression (Escudero‐Martinez et al., ²⁰²²; Malacrinò & Bennett, ²⁰²⁴; Santangeli et al., ²⁰²⁴). However, despite the clear mechanistic link between root exudates and rhizosphere microbiota composition, these traits have typically been measured from separate plant specimens within the same experiment, rather than assessed on the same root system (e.g. Escudero‐Martinez et al., ²⁰²²) due to trait‐specific sampling requirements. In practice, this means that root exudates are collected from living, intact plants, while rhizosphere soil is usually sampled after destructive excavation of the root system. Integrating the sampling of root exudates and rhizosphere microbiota from the same plant specimen will (1) avoid the uncertainty of linking different rhizosphere traits collected from two individual plants (even within the same experiment), (2) increase the number of treatments that can be investigated within one experiment as only half the number of plants are needed to sample both exudates and rhizosphere microbiota and (3) deliver a fundamental understanding on the root exudate metabolites‐microbiota feedback processes.

Hence, the objective of this study was to develop, assess, and critically evaluate different sampling approaches that enable the collection of root exudates and the rhizosphere microbiota from the same soil‐grown plant. As a proof of concept, we used the staple crop barley (Hordeum vulgare L.), one of the reference plants for microbiota investigations in crops (Alegria Terrazas et al., ²⁰²⁰; Jiang et al., ²⁰²⁵). By adapting two well‐established sampling approaches, the soil‐hydroponic hybrid exudate sampling approach (Oburger et al., ²⁰¹⁴; Otxandorena‐Ieregi et al., ²⁰²⁴) and the root shaking/vortexing rhizosphere microbiota sampling approach (Robertson‐Albertyn et al., ²⁰¹⁷; Alegria Terrazas et al., ²⁰²⁰; Escudero‐Martinez et al., ²⁰²²; Alegria Terrazas et al., ²⁰²²), we tested different procedures of (1) bulk soil removal (rinse vs shake), (2) the rhizosphere harvest from roots (dip vs vortex), and the effect of the final rhizosphere fraction collection (gravitational sedimentation vs centrifugation), followed by a 3 h of hydroponic exudate collection period (Fig. 1). We compared the impact of the different approaches on plant biomass, root morphology, root exudate quantity (total C exudation rates), and the metabolic profile of root exudates (non‐targeted metabolomic fingerprinting using Reversed‐Phase Liquid Chromatography–Mass Spectrometry (RPLC‐MS), as well as on rhizosphere microbiota composition by metabarcoding of bacteria/archaea (16S rRNA gene) and fungi (ITS), absolute quantification of these microbial communities, and the abundance of functional microbial nutrient cycling genes (quantitative PCR).

Fig. 1.

Schematic representation of the steps involved in the different rhizosphere sampling protocols Rinse‐dip, Shake‐dip, Shake‐vortex, and Shake‐cut. Created in BioRender. Schwalm, H. (2026) https://BioRender.com/ppng2qg.

Materials and Methods

Plant growth

Barley (Hordeum vulgare L.) plants from the genotype cv. Barke were used to evaluate the different sampling approaches (Fig. 1). Barley seeds were surface sterilised by washing them in 70% ethanol (30 s) followed by 5% sodium hypochlorite (15 min). Sterilised seeds were rinsed four times with autoclaved deionised water and pre‐germinated on Petri dishes containing a semi‐solid 0.5% agar solution. After 2 d, germinated seeds with similar rootlet size were sown in individual 400 ml pots filled with 300 g of a sieved (10 mm) reference agricultural soil previously used for barley‐microbiota investigations, designated ‘Quarryfield’ (Invergowrie, UK, 56°27′5′N 3°4′29′W; Sandy Silt Loam, pH 6.2; Organic Matter 5%; Robertson‐Albertyn et al., ²⁰¹⁷; Alegria Terrazas et al., ²⁰²⁰; Maver et al., ²⁰²⁰; Escudero‐Martinez et al., ²⁰²²; Alegria Terrazas et al., ²⁰²²). Planted pots were arranged in a randomised design n = 6 replicates per sampling approach, plus n = 6 unplanted (bulk) controls for Rinse‐dip and Shake‐dip, and an additional n = 6 unplanted controls for Shake‐vortex and Shake‐cut. Watering was performed daily to 80% field capacity using sterilised deionised water by weighing the pots. Two weeks after planting, plants were supplied weekly with a modified Hoagland's solution where mineral nitrogen was reduced to 25% (Alegria Terrazas et al., ²⁰²²). Plants were grown until stem elongation, Zadok's stage 30–35 (c. 5‐wk post‐transplant) in a glasshouse at the James Hutton Institute, UK, under the following controlled environmental conditions: 18°C : 14°C (day : night) temperature regime with 16 h daylight that was supplemented with artificial lighting to maintain a minimum light intensity of 200 μmol quanta m⁻² s⁻¹.

Testing different sampling approaches for combined rhizosphere soil collection and exudation sampling

Considering the scope of this methodological study, we report the tested experimental approaches in a protocol‐like manner to facilitate the applicability and reproducibility of the different sampling approaches. An overview of the four different sampling approaches is presented in Fig. 1 and in Notes S1 as a practice abstract.

Protocol rhizosphere Rinse‐dip

Bulk soil removal

Rooted soil blocks were gently removed from pots to preserve root integrity. Bulk soil was removed by exposing the intact soil block to a steady and moderate stream of tap water over a mesh tray. Rinsing was gently performed from multiple angles to avoid dislodging rhizosphere particles. Rinsing stopped once a uniform, thin layer of tightly adhering rhizosphere soil remained on the root surface.

Rhizosphere soil collection–part I

Once the rhizosphere soil layer was visible, the root system was repeatedly dipped into a deionised water container to dislodge the rhizosphere soil gently. The entire root system was submerged in this solution. The dipping process was repeated using a second container with fresh deionised water to maximise rhizosphere soil collection. This soil slurry represented the rhizosphere fraction. Both dipping solutions were pooled into one container, and the rhizosphere fractions were left for sedimentation for at least 2 h at 4°C. The dipping solution volume was adapted to the target plant's species root system size. In this barley experiment, we used a 2 × 250 ml dipping solution.

Exudate sampling

After the initial washing step used to collect rhizosphere material, the roots underwent a thorough manual cleaning procedure to remove residual organic matter adhering to the surface. Each root system was individually handled and gently agitated by hand while immersed in water, allowing persistent organic debris to detach with the application of minimal mechanical force. This hand‐washing was gentler than the previous washing, aiming to remove persistent organic debris while minimising mechanical stress. Then the plant was placed into a fresh container with deionised water, awaiting further processing. Before submerging the roots into the final exudate sampling solution, they were placed into a fresh container with the same composition as the exudate sampling solution (in‐between‐sampling step). This aimed to capture metabolites released from damaged cells and promote a root osmotic adjustment. After 5 min, the roots were removed and gently placed on tissue paper for a few seconds to capture water droplets. This exudate sampling solution was discarded. The roots were then transferred to the final container with 80 ml of exudate sampling solution. The exudate sampling solution consisted of autoclaved deionised water containing 5 mg l⁻¹ Micropur (Roth, Katadyn), as a broad‐spectrum bactericide to avoid microbial decomposition of exuded metabolites during the exudate collection (Otxandorena‐Ieregi et al., ²⁰²⁴). The sampling solution containers were wrapped with aluminium foil to protect the roots from light and placed back into the glasshouse for 3 h (Aulakh et al., ²⁰⁰¹). We aimed for the root (dry weight) to sampling solution volume (L) ratio (RSVR) to be in the range of 2–4 to avoid any set‐up driven biases (Otxandorena‐Ieregi et al., ²⁰²⁴).

Rhizosphere soil collection–part II

While the plants were in the glasshouse for exudate collection, containers with the rhizosphere soil slurry were taken from the 4°C cold room. After gravitational sedimentation for at least 2 h at 4°C, the supernatant was discarded by pouring until a 50 ml sample was obtained. Six supernatant samples were kept as ‘supernatant controls’ for the Rinse‐dip and Shake‐dip protocols. The concentrated 50 ml samples were centrifuged at 1500  g for 20 min at room temperature, and the remaining supernatant was removed. The resultant rhizosphere pellet was flash‐frozen in liquid nitrogen and stored at −70°C.

Plant tissue harvest and processing of exudate samples

After an exudate collection period of 3 h, plants were removed from the containers. The containers with the exudate sample solution were closed and stored at 4°C until further processing. Then the plant material was harvested. First, roots and shoots were separated. The shoot fraction was collected in paper envelopes for dry weight determination by drying at 70°C for 2 d. The root system was briefly blotted with paper before fresh weight determination. Then, the root system was divided into two aliquots. This was achieved by cutting the root system into four equal transversal sections (S1–S4) along the longitudinal axis. Each of the four sections was further evenly divided into two halves by manually separating the roots (SX‐right and SX‐left). These sections were then combined into two aliquots following the longitudinal root axis to obtain samples representative of the entire root system (S1–S4 right and S1–S4 left): one section was dedicated to dry weight determination (2 d at 70°C), and the second section for the analysis of root morphological parameters by winrhizo (see ‘Root morphology analyses’) and was stored in 70% ethanol until processed. This root sampling strategy allowed for the collection of individual root sections that can be used for representative subsampling/upscaling.

Immediately after plant tissue harvest, exudate samples were filter sterilised (0.2 μm acetate filters, Chromafil®) and aliquoted as follows: 30 ml for total dissolved organic carbon (DOC) analysis, 30 ml for non‐targeted metabolite analysis, and the remaining c. 20 ml as backup. Samples were frozen at −20°C, freeze‐dried in an alpha 1–2 LD plus lyophiliser (Sci‐Quip, Osterode am Harz, Germany), and then shipped to BOKU, Vienna, Austria, for root exudate analysis.

Protocol rhizosphere Shake‐dip

Bulk soil removal

The rooted soil blocks were carefully removed from the pots, ensuring the plant remained intact. Roots were gently shaken to remove loosely bound soil particles until only the rhizosphere soil tightly adhered to the roots remained; that is, dry bulk soil removal by shaking. This soil was considered the rhizosphere fraction.

Rhizosphere soil collection–part I

The intact root system with the adhering rhizosphere soil was repeatedly dipped in deionised water, following the same procedure from this point on as described in Protocol rhizosphere Rinse‐dip .

Exudate sampling

Same procedure as described in Protocol rhizosphere Rinse‐dip .

Rhizosphere soil collection–part II

Same procedure as described in Protocol rhizosphere Rinse‐dip .

Plant tissue harvest and processing of exudate samples

Same procedure as described in Protocol rhizosphere Rinse‐dip .

Protocol rhizosphere Shake‐vortex

Bulk soil removal

Dry bulk soil removal by shaking – same procedure as described in Protocol rhizosphere Shake‐dip .

Rhizosphere soil collection

The root system from the intact plant, along with its adhering rhizosphere, was placed into a sterile 50 ml Falcon tube containing 20 ml of phosphate‐buffered saline (PBS). Samples were then vortexed for 30 s, and the rhizosphere soil was sedimented for 2–3 min, and incubated in ice. The roots were transferred to a new 50 ml Falcon tube with 20 ml PBS, in which the samples were vortexed again for 30 s to separate the remaining rhizosphere soil from the roots. The two falcon tubes containing the rhizosphere fraction were combined into a single tube, with the combined sample representing the rhizosphere fraction. This sample was then centrifuged at 1500  g for 20 min at room temperature. After centrifugation, the supernatant was discarded, and the pellet was flash‐frozen in liquid nitrogen and stored at −70°C.

Exudate sampling

Same procedure as described in Protocol rhizosphere Rinse‐dip .

Plant tissue harvest & processing of exudate samples

Same procedure as described in Protocol rhizosphere Rinse‐dip .

Protocol rhizosphere Shake‐cut

This protocol represents a well‐established reference for 16S rRNA gene and ITS metabarcoding sequencing that has been frequently used in previous investigations (Maver et al., ²⁰²¹; Escudero‐Martinez et al., ²⁰²²; Alegria Terrazas et al., ²⁰²²). However, sampling a section of the root system (the root system uppermost 6 cm) represents a destructive approach that cannot be combined with root exudation sampling.

Bulk soil removal

Dry bulk soil removal by shaking – same procedure as described in Protocol rhizosphere Shake‐dip .

Rhizosphere soil collection–part I

The root system was separated from the shoot and sectioned to retain just the uppermost 6 cm of the seminal root system. Then, the samples were vortexed and centrifuged as described in Protocol rhizosphere Shake‐vortex .

Exudate sampling

No exudates were collected as the root system did not remain intact.

Plant the harvest & processing of exudate samples

In this protocol, only the shoots were harvested and processed as described in Protocol rhizosphere Rinse‐dip .

Unplanted soil collection

Bulk soil collection corresponding to protocols Rinse‐dip and Shake‐dip

In the same area explored by the roots in the planted plots, an equivalent soil volume was collected from the corresponding unplanted pots. The soil was sampled with a spatula and mixed with 250 ml of deionised water. The bulk soil suspension was processed as in Rinse‐dip and Shake‐dip: rhizosphere soil collection – part II sections.

Bulk soil collection corresponding to protocols Shake‐vortex and Shake‐cut

In the same area explored by the roots, the same soil volume obtained from the planted samples was collected from the unplanted pots in the planted plots. The soil was collected with a spatula and mixed with 30 ml of PBS. The bulk soil suspension was further processed as in the Shake‐vortex and Shake‐cut: rhizosphere soil collection sections.

Root morphology analysis

The root sections stored in 70% ethanol were spread out in a clean, scratch‐free plastic tray filled with water. Roots were arranged to avoid overlapping and any meniscus bubbles. Images were scanned in black and white at 600 dpi on an Epson 12 000 XL flatbed scanner (Watford Hets England WD17 1JA).

Scanned images were analysed using winrhizo pro™ software (winrhizo pro 19, Regent Instruments Inc., Québec, QC, Canada) to obtain root morphological traits. Root diameter categories were set at increments of 0.5 mm. The detection threshold was set at ‘automatic’.

Several root morphology measurements were recorded (Table S1). For calculations, we first normalised the winrhizo outputs by the total root dry weight. The specific root length (SRL) was computed using the equation root length/root dried weight. Normality was checked using the Shapiro–Wilk test, and one‐way analysis of variance (ANOVA) and post hoc Tukey's Honest Significant Difference were used to assess differences in root morphology.

Microbiota analysis

Total DNA was extracted from the rhizosphere and unplanted soil samples for the four different sampling approaches using the FastDNA™ SPIN kit for soil (MP Biomedicals, Solon, OH, USA) following the manufacturer's instructions, eluted at 35 μl with deionised and distilled water, and subject to targeted amplification of microbiota phylogenetic markers.

Amplicon sequencing library preparation: 16S rRNA gene

16S rRNA gene phylogenetic marker from rhizosphere and unplanted soil preparations was amplified using the 515F‐806R primer pair (Caporaso et al., ²⁰¹²) (Notes S2). These PCR primer sequences were fused with Illumina flow cell adaptor sequences at their 5′ termini. The 806R primer contained a 12‐mer unique ‘barcode’ sequence used for each sample, enabling multiplexed sequencing of several samples in a single pool.

For each bulk and rhizosphere sample, a total of 50 ng of DNA was subjected to PCR amplification using the KAPA HiFi HotStart PCR kit (KAPA Biosystems, Wilmington, DE, USA). The individual PCR reactions were performed in a 20 μl final volume and contained: 4 μl of 5× KAPA HiFi Buffer, 10 μg Bovine Serum Albumin (Roche, Mannheim, Germany), 0.6 μl of a 10 mM KAPA dNTPs solution, 0.6 μl of 10 μM solutions of the individual PCR primers, and 0.25 μl of KAPA HiFi polymerase. Reactions were performed using the following programme: 94°C (3 min), followed by 35 cycles of 98°C (30 s), 50°C (30 s), 72°C (1 min), and a final step of 72°C (10 min) (Notes S2). For each primer combination, a no‐template control (NTC) was included. Individual PCR reactions were performed in triplicate, and two independent sets of triplicate reactions per barcode were included to minimise PCR bias.

Before purification, 6 μl aliquots of individual replicates and the corresponding NTCs were inspected on 1.5% agarose gel to detect amplification and/or any contamination. Individual replicates of the PCR amplicons were then pooled per replicate and purified using the Agencourt AMPure XP Kit (Beckman Coulter, Brea, CA, USA) with 0.7 times the volume of the sample beads. Purified samples were quantified using a Qubit fluorometer with the dsDNA HS assay (Invitrogen, Thermo Fisher Scientific, Waltham, MA, USA). Once quantified, individual barcoded samples were pooled into a new single tube in an equimolar ratio to generate the library.

Paired‐end Illumina sequencing (2 × 250 bp reads) (Maver et al., ²⁰²¹) was performed using the Illumina MiSeq system. Library pool quality was assessed using a Bioanalyzer (High Sensitivity DNA Chip; Agilent Technologies, Santa Clara, CA, USA) and quantified using a Qubit and qPCR (KAPA Biosystems, Wilmington, DE, USA). Amplicon libraries were spiked with 15% of a 4 pM phiX control solution. The resulting high‐quality libraries were run at 10 pM final concentration.

Amplicon sequencing library preparation: ITS region

A 2‐step PCR protocol was used, in which PCR 1 involved the amplification of the ITS2 phylogenetic marker using the ITS86F and ITS4R primer pair (Notes S2). Each reaction was carried out in a 25 μl volume containing 12.5 μl of KAPA HiFi HotStart ReadyMix Taq (Roche Sequencing, San Jose, CA, USA), 0.2 μM forward and reverse primers, and 2.5 μl of the template at 5 ng μl⁻¹. Cycling conditions for PCR1 were 95°C (3 min), 25 × (95°C (45 s), 54°C (45 s), 72°C (1 min)), and 72°C (5 min) (Notes S2). PCR1 was performed with three technical replicates per sample, which were then pooled for purification with 0.6 uL Ampure XP beads (Beckmann Coulter, Amersham, UK) per μl of DNA, followed by elution in 15 μl Tris‐EDTA (TE) buffer (Thermo‐Fisher Scientific, Cork, Ireland). Next, barcodes were added to the amplicons using the unique dual index (UDI; Integrated DNA Technologies, Munich Germany) (PCR2). Amendment of UDI indexing primers was performed in a 50 μl reaction volume containing 25 μl of KAPA HiFi HotStart ReadyMix Taq, 5 μl of PCR 1 template, and 5 μl of UDI. PCR conditions were 95°C (3 min), 7 × (95°C (30 s), 55°C (30 s), 72°C (30 s), and 72°C for (5 min) (Notes S2). Amplicons were purified using the Ampure XP beads as for PCR1. Product concentrations were measured using the Qubit fluorometer dsDNA HS assay (Thermofisher, Dublin, Ireland). The products were then pooled in equimolar concentration, the library quality was checked using a Bioanalyzer (High Sensitivity DNA Chip; Agilent Technologies), and concentrations were quantified using the Qubit fluorometer. Paired‐end reads (2 × 300 bp) were sequenced on the Illumina NextSeq 2000 system using the P2 flow cell (Teagasc sequencing facility, Ireland).

Amplicon sequencing reads processing

dada2 v.1.22 (Callahan et al., ²⁰¹⁶) and R 4.1.0 (R Core Team, ²⁰²¹) were used to generate the ASVs following the ‘DADA2 Pipeline Tutorial (1.10)’ and to assign taxonomy the SILVA 138 database for 16S rRNA (Quast et al., ²⁰¹³) and the UNITE v2023 database for ITS (Abarenkov et al., ²⁰²³). The creation of the Phyloseq object was completed as in Maver et al., ²⁰²¹. Subsequently, sequences classified as ‘Chloroplast’ or ‘Mitochondria’ from the host plant were pruned in silico. Further filtering criteria were applied for the 16S rRNA gene library; samples with < 10 000 reads were discarded. Only ASVs occurring at least 20 reads in 2% of the samples were retained. The 16S rRNA and ITS datasets were rarefied at equal sequencing depth across samples (64 000 reads for the 16S rRNA gene and 200 000 reads for the ITS). The resulting data sets contained 2376 bacterial ASVs across 39 samples and 2139 fungal ASVs across 19 samples.

Alpha‐diversity and beta‐diversity were calculated as described in Maver et al. (2021). deseq2 (Love et al., ²⁰¹⁴) was used to perform microbial differential abundance analysis to identify taxa differentially enriched in pairwise comparisons of the tested approaches and calculated by the Wald test (false discovery rate, FDR < 0.05). Microbial taxa significantly enriched in rhizosphere specimens compared to their cognate unplanted soil were visualised with the package upsetr (Conway et al., ²⁰¹⁷).

qPCR

Potential PCR inhibitors were tested by comparing amplification of an added plasmid to DNA samples with that in a distilled water control as per Duff et al. (2022) (Notes S2). Briefly, if the plasmid showed reduced amplification compared to the control, the sample was considered to have inhibitors. None of the samples in this study showed evidence of inhibitors, and further cleaning of samples was not undertaken.

qPCR was used to quantify the absolute abundance of bacterial 16S rRNA gene and fungal ITS2 copies, as well as nitrification and denitrification genes for microbial functional assessment: AmoA and nosZ clade II, respectively. A tenfold serial dilution of the standard was prepared, covering seven concentration points ranging from 10⁸ to 10² gene copies per ng. qPCR was performed in a 10 μl reaction volume containing 5 μl Takyon™ Low ROX SYBR 2× Master Mix blue dTTP (Eurogentec, Liège, Belgium), 2 μl of DNA template (1 ng μl⁻¹), and 1.5/0.2 μM each of the primers. Primer concentration, sequences, and cycling conditions are found in Notes S2. Five samples per sampling approach were analysed with qPCR, and each sample was run with three technical replicates. An NTC was also run for each plate. A positive control of known quantity was also included for the 16S rRNA, ITS and nosZ clade II to confirm accuracy of the assay; a positive control was not used for the AmoA qPCR assay due to lack of control availability.

The resulting average gene copies per ng DNA of the three technical replicates for each sample were assessed for normality using a Shapiro–Wilk test. One‐way ANOVA was performed on each gene, followed by a post hoc Tukey test using the car and agricolae packages in R 4.3.2 (R Core Team, ²⁰²¹). The residuals of each ANOVA model were visually assessed for homoscedasticity and normality with a residual and a quantile‐quantile test. Statistical outliers, that is, those points that statistically affect the output of each model, were identified using a Cook's distance test.

Root exudate analysis and data evaluation

Total DOC and total nitrogen bound (TNb) concentrations were determined using a liquid total organic carbon (TOC) elemental analyser (Elemental VarioTOC, Langenselbold, Germany). For the analysis, the freeze‐dried aliquot (original sample volume 30 ml) was resuspended in 15 ml of type 1 laboratory‐grade water. To determine total dissolved organic N (DON), inorganic N concentrations (only NO₃ ⁻, no NH₄ ⁺ were detected) were analysed using a spectrophotometric assay (Hood‐Nowotny et al., ²⁰¹⁰), and NO₃ ⁻ concentrations were subtracted from TNb. Total DOC (hereafter C) and total dissolved organic nitrogen (hereafter N) exudation rates (nmol cm⁻² root surface area (RSA) h⁻¹) were calculated by dividing the absolute quantity of C and N in each sample by the total RSA of the respective barley plant and by the sampling period (3 h).

For non‐targeted metabolomic analysis, the second freeze‐dried exudate sample aliquot (originally 5 ml) was resuspended in 1.0, 1.5, or 2.0 ml of LC‐MS grade water, depending on the original sample DOC concentration, aiming for a final DOC concentration of c. 85 mg C l⁻¹. Re‐suspensions were filtered (4 mm; 0.2 μm cellulose acetate; Nalgene™,Thermo‐Fisher Scientific, Waltham, MA, USA) and non‐targeted metabolomic analysis was conducted using a 6230b Time‐of‐Flight mass spectrometer (TOF‐MS) with a dual Jetstream ESI interface (Agilent Technologies) coupled with an Infinity II UHPLC system (Agilent Technologies). The separation was performed by Reversed‐Phase Liquid Chromatography (RPLC), using an Atlantis T3 C18 column (2.1 × 150 mm, 3 μm particle size; Waters Corporation, Milford, MA, USA) as described in Lohse et al. (2023). Briefly, chromatographic separation was carried out using gradient elution at a flow rate of 200 μl min⁻¹. Mobile phase A consisted of LC‐MS grade water with 0.1% v/v formic acid, and mobile phase B was 100% methanol. The initial gradient condition of 100% A was maintained for 2 min, then decreased to 60% A over the next 8 min and held constant for an additional 2 min. This was followed by a cleaning step with 100% B for 2.1 min, a rapid decrease to 0% B, and a re‐equilibration period of 5.9 min at 100% A. The injection volume was 5 μl, and the column temperature was maintained at 40°C. Mass spectra in the range of 90 to 1700 m/z were recorded in negative (−) polarity, utilising the 2 GHz extended dynamic range mode with a spectral acquisition rate of 2 Hz. Data were acquired with the masshunter acquisition software (v.10.1; Agilent Technologies), while masshunter profinder B.10.00 software (Agilent Technologies) was used for peak picking and chromatographic deconvolution with a Batch Recursive Feature Extraction (small molecules/peptides) workflow (for details see also SI of Lohse et al. (2023)). Samples were injected in a randomised sequence and quality control samples (QCs, prepared by combining equal volumes of each sample) were injected every five samples to monitor the stability of the TOF‐MS system. The first four injected samples were excluded from further data evaluation due to pressure instability of the binary pump during the measurement. Features showing a relative SD higher than 25% in the QCs, and features with an intensity of < 10 times the blank average, were excluded from further analysis. Thereafter, raw intensities were scaled with a sample‐specific factor calculated as pre‐concentration fold × RSA, followed by unit variance scaling. Finally, Principal Component Analysis (PCA) and PERMANOVA were performed on Euclidean distances to investigate the impact of the sampling approach on the metabolomic data.

Results

Different sampling approaches have an impact on root biomass and morphology, but not root exudate quantity or metabolic profile

We evaluated the impact of the different sampling approaches, summarised in Fig. 1, on plant biomass (roots and shoots) and root morphology. We used these parameters as indicators of plant integrity. For the Shake‐cut sampling approach, total root weight, morphology, or root exudate parameters were not measured, as roots were partially destroyed during sampling.

As expected, the shoot biomass did not differ between the protocols evaluated (ANOVA; P‐value = 0.0684). Conversely, we observed a decrease in the recovered root biomass in the following order: Rinse‐dip ≥ Shake‐dip ≥ Shake‐vortex, but only Rinse‐dip and Shake‐vortex were significantly different (ANOVA; P‐value = 0.005, Fig. 2a,b). Root morphology parameters, including total RSA (cm²), average root diameter (mm), and total root length (cm), followed the same trend as the root biomass (Table S1). However, the SRL was comparable across the sampling approaches.

Fig. 2.

Biomass and exudation across sampling approaches: Boxplot depicting the dry weight biomass of shoot (a), roots (b), carbon exudation rate (c), and nitrogen exudation rate (d). In each panel, individual dots depict individual biological replicates. Upper and lower edges of the box plots represent the upper and lower quartiles, respectively. The bold line within the box denotes the median. Letters indicate significant differences following an ANOVA and post hoc Tukey test (P‐value < 0.05). (d) Principal component analysis (PCA) of exuded metabolite features detected via non‐targeted metabolomics analysis. In the PCA plot, large dots represent group centroids, while small dots indicate individual samples of the four sampling approaches depicted in distinct colours. PERMANOVA (Adonis function) was used to test for significance between the sampling approaches (P‐value > 0.05). R ² indicates the proportion of variance explained by the sampling approach. Not significant is denoted by ns. Rinse‐dip (orange), Shake‐dip (yellow), Shake‐vortex (light blue), and Shake‐cut (blue).

An important factor for harvesting root exudates is a constant root (dry) weight‐to‐sampling solution volume (L) ratio (RSVR) (Otxandorena‐Ieregi et al., ²⁰²⁴). Due to the lower root biomass in Shake‐vortex and a constant exudate sampling volume, we observed a significantly (P = 0.0023) lower RSVR in Shake‐vortex (2.6 ± 0.6 g root dwt l⁻¹, mean ± SE) in comparison to Rinse‐dip (3.7 ± 0.2) and Shake‐dip (3.3 ± 0.2) (Fig. S1).

C root exudation rates did not differ between the different sampling approaches, with an average of 0.9 ± 0.05 μmol C cm⁻² RSA h⁻¹ (mean ± SE) (ANOVA; P‐value = 0.0917, Fig. 2c). However, N root exudation rates were significantly affected (ANOVA; P‐value = 0.0157, Fig. 2d). Rinse‐dip N exudation rates (0.1 ± 0.01 μmol N cm⁻² RSA h⁻¹) were comparable to Shake‐dip and Shake‐vortex, yet Shake‐dip (0.1 ± 0.02) N exudation rates were significantly lower than Shake‐vortex (0.1 ± 0.01).

Metabolomic fingerprinting on the root exudates revealed no significant differences among sampling approaches (PERMANOVA (Adonis function) test, P‐value = 0.629), further supported by the lack of protocol‐wise sample organisation in the ordination (Fig. 2e). Nevertheless, we observed that Shake‐dip and Shake‐vortex exudate composition showed more consistent exudate profiles, as indicated by the tighter clustering within those groups (Fig. 2e).

All sampling approaches distinguish rhizosphere microbial communities from those of unplanted soil

Absolute quantification of bacterial (16S rRNA gene) rhizosphere communities did not identify significant differences between sampling approaches (ANOVA P‐value > 0.05, Fig. 3a). Conversely, a significant sampling approach effect was observed on the abundance of the ITS phylogenetic marker (ANOVA P‐value = 0.0219, Fig. 3b). Shake‐vortex displayed greater ITS copy numbers compared with Shake‐dip. The Shake‐vortex approach was also associated with greater variability in microbial gene abundances regardless of the taxonomic marker quantified (Fig. 3a,b; Table S2).

Fig. 3.

Microbial quantification and diversity: (a) 16S rRNA gene (bacteria) and (b) ITS (fungi) gene copies per ng of DNA in unplanted soil and rhizospheres obtained with different sampling approaches. Box plots depict the total copy numbers of each taxonomic marker per ng of DNA (16 rRNA gene or ITS). Individual dots depict individual biological replicates. Different letters denote significantly distinct groups (ANOVA and post hoc Tukey Honest Significant Difference test) Canonical Analysis of Principal Coordinates (CAP) computed on Bray–Curtis dissimilarity matrix for the different rhizosphere sampling approaches and the unplanted bulk soil controls for (c) bacteria 16S rRNA gene and (d) fungal ITS data. Sample type is depicted by colours and shapes. Individual dots in the plot denote individual biological replicates of the different sampling methods or the unplanted soil samples. The number in the plot depicts the proportion of variance (R ²) explained by the factor ‘Sampling approach’ on rhizosphere samples. The asterisks associated with the R ² value denote its significance in the PERMANOVA (Adonis function) (P‐value < 0.05). Note that the y‐axis measures Metric Multidimensional Scaling (MDS) and the x‐axis represents Canonical Analysis of Principal Coordinates (CAP) ordination, differing one order of magnitude in (c). Boxplots representing the alpha‐diversity index Shannon in (e) bacteria 16S rRNA gene and (f) fungal ITS data for the sampling approaches and the unplanted bulk soil controls. In each panel, individual dots depict individual biological replicates. Empty dots are outliers. Upper and lower edges of the box plots represent the upper and lower quartiles, respectively. The bold line within the box denotes the median. Letters indicate significant differences following ANOVA and post hoc Tukey test (P‐value < 0.05). In boxplots, orange bulk soil samples are the controls for Rinse‐dip and Shake‐dip, while blue bulk soil samples are the controls for Shake‐vortex and Shake‐cut.

Beta‐diversity, or between‐sample microbial diversity, confirmed a significant rhizosphere effect for bacteria/archaea and fungi, irrespective of the sampling approach. Rhizosphere communities across all the sampling approaches were significantly different from the unplanted soil, as shown by sample separation along the x‐axis of the Canonical Analysis of Principal Coordinates (CAP) PERMANOVA (Adonis function), 16S rRNA R ² = 78%, P‐value = 2 × 10⁻⁴, ITS R ² = 86%, P‐value = 4 × 10⁻⁴, 5000 permutations, Fig. 3c,d. Hereafter we used the bulk soil of the ‘dipping’ approaches as unplanted reference for fungi, as the ‘vortexing’ bulk soil did not pass the ITS sequencing quality filters. Beta‐diversity within the rhizosphere sampling approaches showed differences in bacterial composition across all the approaches, except for the comparison between Shake‐vortex and Shake‐cut, with no significant variation (PERMANOVA (Adonis function), Table S3a). Pairwise comparisons of the fungal communities revealed differences between Rinse‐dip and Shake‐cut (PERMANOVA (Adonis function), Table S3b).

Alpha‐diversity indices, namely Observed and Chao, computed on bacteria ASV counts, were not different between the unplanted soil and the sampling approaches Rinse‐dip and Shake‐dip (Fig. S2a,c). Conversely, lower numbers of taxa were observed for the Shake‐vortex and Shake‐cut, indicating a more marked rhizosphere effect (ANOVA and Tukey post hoc test, Observed P‐value = 1.59 × 10⁻¹⁵, Chao P‐value = 2.67 × 10⁻¹⁴) (Fig. S2a,c). Fungal ITS2 richness was significantly higher in bulk unplanted soil compared to all sampling approaches (ANOVA and Tukey post hoc test, Observed P‐value = 3.30 × 10⁻⁵, Chao P‐value = 4.03 × 10⁻⁵) (Fig. S2b,d). Shake‐vortex had the lowest Chao richness; however, this was significant just in comparison to Shake‐dip (P‐value = 0.04) (Fig. S2d). All sampling approaches had similar Observed fungal richness (P‐value > 0.05) (Fig. S2b).

Considering the bacterial and fungal community diversity Shannon index, which accounts for the number of taxa and their abundances, all the rhizosphere sampling approaches showed significantly reduced microbial diversity compared to the unplanted bulk soil (ANOVA and Tukey post hoc test, bacterial 16S rRNA P‐value = 1.28 × 10⁻¹⁵; Fungal ITS P‐value = 7.95 × 10⁻¹⁰) (Fig. 3d,e). For bacterial communities, excluding bulk unplanted soil, the Shannon index was greater in the Rinse‐dip and Shake‐dip approaches than in Shake‐vortex and Shake‐cut, with a significant increase observed for Rinse‐dip.

We included a ‘dip‐supernatant control’ to capture bacteria that may have remained in the supernatant due to insufficient sedimentation for the approaches Rinse‐dip and Shake‐dip. (the solution volume was decanted before centrifuging the remaining 50 ml). The ‘dip‐supernatant control’ richness was significantly lower than in rhizosphere samples; however, it still contained > 1000 ASVs representing a consistent proportion (between 63 and 75%) of the number of taxa retrieved in the rhizosphere samples (Fig. S3).

Finally, we were interested in understanding whether any of the sampling approaches inadvertently introduced plant‐derived 16S rRNA contamination into the rhizosphere samples. We compared the number of reads assigned as ‘Chloroplast’ as a proxy for plant contamination for each of the unplanted bulk soils and sampling approaches (Fig. S4, ANOVA and Tukey post hoc test, P‐value = 1.3 × 10⁻⁸). The bulk unplanted soil samples displayed lower content of chloroplast reads than any of the rhizosphere samples (Fig. S4). Among the sampling approaches, the greatest number of reads assigned as ‘Chloroplasts’ was retrieved from Rinse‐dip, whereas the lowest was assigned to Shake‐dip. The number of chloroplast reads was not significantly different for Shake‐vortex or Shake‐cut compared to Rinse‐dip or Shake‐dip (Fig. S4).

Different sampling approaches capture a conserved microbiota reflecting most of the microbial abundance

We performed pairwise comparisons to assess microbial abundances across different sampling approaches (Fig. 4; Table S4). A higher share of rhizosphere bacteria ASVs were enriched in ‘vortexing’ approaches (25–38%) compared with the ‘dipping’ approaches (1–16%) (Table S4a). These differences were reflected at the phylum level (Fig. 4a). Although differences in the number of bacterial ASVs were detected in pairwise comparisons, the majority of rhizosphere ASVs (349 ASVs accounting for 78% of the reads) were recovered across all four sampling approaches, recapitulating the phyla composition of the rhizosphere with Proteobacteria, followed by Actinobacteria, Bacteroidota, and Firmicutes (Fig. 4b, Wald test Individual P‐values < 0.05, FDR corrected).

Fig. 4.

Heatmap of differential microbial enrichment at phylum level across protocols in (a) the bacteria 16S rRNA gene and (c) the fungal ITS. Phyla differentially enriched at individual P‐values < 0.05, Wald Test, FDR corrected. UpSet plots of taxa (ASVs) simultaneously retrieved in pairwise comparisons from the different rhizosphere sampling approaches in (b) the bacteria 16S rRNA gene and (d) the fungal ITS. Vertical bars denote the number of ASVs shared or unique for each comparison coloured by phylum, while the horizontal bars refer to the number of ASVs enriched in the indicated rhizosphere sampling approach. ASVs differentially enriched at individual P‐values < 0.05, Wald Test, FDR corrected.

In fungi, a low percentage of rhizosphere reads (0–3%) enriched between sampling approaches indicating even less marked impact in the choice of sampling approach compared with bacteria (Table S4b). Hierarchical clustering of the fungal taxa enriched from the rhizosphere revealed the presence of three main clusters (Fig. 4c). Two of the clusters differentiated Rinse‐dip and Shake‐vortex, while the third cluster contained one Shake‐cut sample. Shake‐dip and Shake‐cut samples showed distinct profiles and were not clustering (Fig. 4c). As for bacteria, most of the rhizosphere fungal ASVs (29 ASVs accounting for 73% of rhizosphere reads) were shared between sampling approaches. Its phylum composition was dominated by Olpidiomycota, followed by Ascomycota, and Basidiomycota (Fig. 4d Wald test individual P‐values < 0.05, FDR corrected).

The absolute abundances of microbial nitrogen cycle genes were unaffected by the sampling approach

We investigated the impact of sampling approaches on microbiome functional groups. The amoA gene encodes the catalytic subunit of ammonia monooxygenase, the enzyme that performs the first, rate‐limiting step of nitrification, and is therefore a marker for quantifying nitrifier communities (Hodgskiss et al., ²⁰²³). By contrast, nosZII encodes a nitrous oxide reductase in the denitrification pathway in many soils, reducing N₂O to N₂. These genes show strong correlations with field fluxes of nitrate and N₂O and have complementary roles as proxies for N₂O sources and sinks (Li et al., ²⁰²⁵, Yang et al., ²⁰¹⁷). Abundances of the amoA gene (nitrification) and the nosZII clade II gene (denitrification) were not significantly affected by the sampling approach (Fig. S5). However, differences were identified between the unplanted soil and some of the sampling approaches (Fig. S5). Overall, we found a (mostly non‐significant) trend of reduced nitrification–denitrification genetic potential in the rhizosphere compared to bulk soil. The amoA gene was significantly less abundant in Shake‐dip compared with its cognate unplanted control Bulk‐dip, whereas nosZII abundance was reduced in the rhizospheres of the sampling approaches Shake‐vortex and Shake‐cut compared with its unplanted control Bulk‐vortex (Fig. S5).

Discussion

In this study, we tested and evaluated different sampling approaches for root exudation and microbiota profiling from the same soil‐grown plant. We observed differences in biomass and morphology of the recovered roots and, to a lesser extent, in the rhizosphere microbiota composition. While we found minor differences in organic N root exudation, total dissolved C root exudation rates and non‐targeted metabolite profiles were unaffected by the different sampling procedures. We establish a coherent methodology that integrates previously applied approaches for sampling root exudates and characterising the rhizosphere microbiota into one joint protocol, while streamlining these procedures to support higher sampling throughput. Rather than proposing an entirely new set of techniques, we combine and refine well‐validated approaches to produce a workflow that is robust and operationally efficient.

Root system excavation and bulk soil removal – experimental considerations

A limitation of our integrated approach is that the excavation of intact root systems in field settings, while maintaining the root integrity, is challenging. Nevertheless, soil‐hydroponic‐hybrid approaches to collect root exudates from the whole root system have been successfully applied to fully grown maize plants in the field (Santangeli et al., ²⁰²⁴), as well as on excavated field grown rice (Aulakh et al., ²⁰⁰¹), and arctic tundra plants (Wegner et al., ²⁰²⁵). Depending on root thickness, excavation of individual root segments could potentially be less destructive than sampling an entire root system. However, the root segment approach is limited to only a very small part of the root system and therefore requires numerous replicates of an individual plant to obtain representative results. While roots can be damaged during excavation (irrespective of soil texture and whether only a root segment or the entire root system is targeted), the majority of cell contents can be captured by submerging the soil‐free root system in an aliquot of the sampling solution for several minutes before moving the root system to the final sampling solution (in‐between‐sampling step, see Materials and Methods). This in‐between‐sampling step also captures any potential osmotic adjustment reactions by the plant. Furthermore, by treating all investigated plants the same, whether they are retrieved from pots, soil columns or segments excavated in the field, the damage effect, if not entirely captured by the in‐between sampling step, will be uniform across the experiment and have no impact between‐treatment comparisons. Working in the field or with larger plants requires methodological adaptations/compromises that need to be considered when interpreting results, but these challenges should not discourage from conducting field experiments.

Sampling approach selection influences root biomass and morphology, and to a lesser extent, root exudation

The retrieved belowground biomass varied among the different sampling approaches, with Rinse‐dip resulting in a greater root biomass than Shake‐vortex (Fig. 2b). As the mode of bulk soil removal (i.e. Rinsing vs Shaking) did not result in significant differences in root biomass, this may indicate that ‘vortexing’ is more destructive than ‘dipping’, inducing root losses. A gradual root biomass loss was observed, which could indicate that shaking has also negative impact on root system integrity (Fig. 2b). Interrelated root morphology traits, including root length, surface area, average diameter, and root area, followed the same trend, confirming that root biomass was lost during ‘vortexing’ (Table S1). ‘Vortexing’ has been widely used when sampling rhizosphere microbiota to detach the rhizosphere soil from the roots (Maver et al., ²⁰²¹; Escudero‐Martinez et al., ²⁰²²; Alegria Terrazas et al., ²⁰²²). If the goal is to assess root morphology or collect root exudates, where preserving root system integrity is critical, the use of vortexing should be carefully considered based on the experimental objective. For example, we found that SRL, when normalised by root weight, was comparable across different processing methods. This suggests that vortexing can be suitable if the parameter of interest is normalised, but it is not appropriate for measuring absolute, non‐normalised, root morphological traits.

‘Vortexing’ is also operationally limited to smaller and/or younger root systems. Yet ‘vortexing’ has important advantages such as faster sample throughput in comparison to the ‘dipping’ approach, and a smaller volume of solution containing the rhizosphere soil slurry needs to be processed. We did not find differences between bulk soil removal by ‘dipping’ or ‘shaking’, but it is important to note that we used one soil type with a sandy texture in our study (Sandy Silt Loam). Differences in soil texture might have an impact on sampling (Barillot et al., ²⁰¹³). Bulk soil removal via ‘shaking’ might cause more root loss and damage in heavier soils with larger clay contents due to break‐off of large root‐containing heavy soil aggregates in comparison to ‘rinsing’.

We used chloroplast reads in the rhizosphere as a proxy for plant content in the microbiota samples. The number of reads was similar across sampling protocols; only Shake‐dip rhizosphere samples contained fewer chloroplast reads (Fig. S4). Therefore, even if ‘vortexing’ was associated with root biomass loss, this was insufficient to cross‐contaminate the rhizosphere microbiota sample preparation.

Overall, the different sampling approaches showed a minor effect on barley exudation patterns. N root exudation rates were affected, while C exudation rates remained comparable. (Fig. 2c,d). A recent study demonstrated that experimental factors such as the root (dry weight) to sampling solution volume (L) ratio RSVR, can bias the exudation results (Otxandorena‐Ieregi et al., ²⁰²⁴), with lower RSVRs resulting in higher exudation rates, presumably driven by the higher diffusion gradient into the (more diluted) sampling solution. In our experiment, the RSVR of the Shake‐vortex treatment was significantly lower (Fig. S1) due to the vortex‐induced root biomass loss. This may explain the slightly higher C (though not significant) and N exudation observed in this treatment (Fig. 2d). However, we consider the observed differences in N exudation to be of minor importance as overall non‐targeted metabolomics results revealed no significant differences between the treatments, suggesting comparable metabolite identity and quantity (Fig. 2e). Nevertheless, our findings highlight the importance of maintaining a similar RSVR across treatments to avoid an experimental setup‐driven bias.

Microbiota diversity showed a distinctive ‘rhizosphere effect’ in comparison to the unplanted soil, but it varies within sampling approaches

Sampling approaches did not affect the absolute abundances in the bacterial community but did impact the fungal abundances (Shake‐dip had a lower abundance than Shake‐vortex) (Fig. 3b). This may be due to a better retrieval of fungal biomass by ‘vortexing’ the entire root system. However, these differences did not impact fungal beta‐ or alpha‐diversity.

All sampling approaches showed a clear rhizosphere pattern across kingdoms (fungi, bacteria, and archaea), although the primers employed may underrepresent archaeal diversity (Parada et al., ²⁰¹⁶), in either alpha or beta‐diversity compared to the unplanted soil controls (Fig. 3). These observations confirm that the plant's signature on microbiota assembly is sufficiently robust to withstand biases induced by different sampling approaches. However, there were differences among the rhizosphere soil sampling approaches (Fig. 3c,d). Microbiota composition was influenced by both the method used to remove bulk soil (rinse or shake) and the approach taken to harvest rhizosphere soil from roots (dipping or vortex), with more pronounced differences in bacterial than fungal communities. Bacterial rhizosphere communities typically exhibit greater diversity and finer spatial heterogeneity than fungi (Schultes et al., ²⁰²⁵), meaning that small variations in soil handling, rhizosphere harvesting, or DNA extraction can disproportionately affect observed community composition.

An additional aspect that distinguishes each of these sampling approaches is how, once we harvested the rhizosphere soil from roots, the rhizosphere was collected: in ‘Vortexing’ sampling approaches (Rinse‐dip and Shake‐dip) the entire solution volume was centrifuged to retrieve the rhizosphere fraction from the solution, while in the ‘Dipping’ approaches (Shake‐vortex and Shake‐cut) we used gravitational sedimentation with the majority of the dipping solution being decanted after sedimentation and only the final 50 ml being centrifuged. Therefore, observed differences in results could be confounded by downstream differences in sampling steps. Perhaps centrifuging the entire dipping solution volume may be more critical for bacteria than for the fungal community, as marked differences were found between ‘Vortexing’ and ‘Dipping’ in the composition of this microbial community (Fig. 3c). Indeed, we investigated the bacterial composition of the ‘dip‐supernatant control’ for the ‘dipping’ sampling approaches (gravitational sedimentation). This control revealed that a fraction of the bacterial community was still present in the supernatant (Fig. S3). We did not attempt to amplify the supernatant of the ‘Vortexing’ sampling approaches (centrifugation). Taken together, this suggests that sample centrifugation is clearly superior from a methodological standpoint when harvesting the rhizosphere microbial fraction, and it is a common step in rhizosphere sampling across the literature (Barillot et al., ²⁰¹³, Tkacz et al., ²⁰²⁵). Conversely, alternative methods, such as gravitational sedimentation, which may recover a representative fraction of the microbial community, represent a reasonable and pragmatic compromise when logistical constraints apply; for example, when large volumes of rhizosphere solution are generated, which require multiple centrifugation steps that are highly time‐consuming, as in the case of Rinse/Dip. This should be considered when designing subsequent experiments beyond the scope of this manuscript.

A prominent ‘conserved barley microbiota’ was observed across the rhizosphere fractions for all sampling protocols

Due to the observed differences in microbial composition across sampling approaches (Fig. 4a; Table S3), we measured the number of taxa intersecting from each of the sampling approaches in the rhizosphere fraction. Across the sampling protocols tested, both the bacteria and the fungal communities displayed a ‘core microbiota’ representing the mayor proportion of bacterial and fungal ASVs (16S: 349 ASVs and ITS: 29 ASVs) and rhizosphere reads (16S: 78% and ITS: 73%) across sampling approaches (Fig. 4b). The taxonomic composition of the bacterial ‘core microbiome’ aligns well with previously reported phyla dominating the rhizosphere (Ling et al., ²⁰²²) and, in particular, the barley rhizosphere, such as Proteobacteria, Actinobacteria, and Bacteroidota (Escudero‐Martinez et al., ²⁰²²). These bacterial phyla dominating the rhizosphere were even more abundant in the ‘vortexing’ approaches (Fig. 4a). Regardless the sampling approach used, the rhizosphere core fungal community, typically dominated by Ascomycota and Basidiomycota (Hennecke et al., ²⁰²³), was rather dominated by the phylum Olpidiomycota, which was previously reported dominant in the rhizosphere of Chinese cabbage (Wei et al., ²⁰²³).

Altogether, this suggests that, regardless of the sampling protocol applied, a characteristic rhizosphere microbial community profile was present in the enriched rhizosphere fractions. Therefore, despite variations in microbial composition across sampling approaches, rhizosphere microbiota sampling appeared more consistent and less variable than other plant‐associated traits, such as root biomass and root morphology.

In conclusion, we demonstrate that, under the tested conditions, different sampling approaches produced comparable microbiota and exudation patterns, enabling an integrated study of root exudation and microbial profiles from the same plant. We presented a detailed investigation of the implications associated with employing different approaches to sample root exudates, which require minimal plant disturbance, and the rhizosphere microbiota fraction, which must be distinct from the unplanted soil. While the overall results were comparable, the different sampling steps, that is, rinsing, shaking, dipping, vortexing, gravitational sedimentation or centrifugation, introduced minor (but still statistically significant) differences in the obtained results. Ultimately, the experimental scope and aim should guide sampling protocol selection. Our results also provide important insights into the robustness of different sampling approaches, thereby enabling more informed decisions regarding the selection of experimental procedures for the joint sampling of root exudates and the rhizosphere microbiome from the same plant. Learning from the sampling approaches comparison presented here, we compiled a sampling approach for subsequent experimentation that we consider applicable to most experimental conditions and can be visualised in the format of a practice abstract (see Notes S1) and the following video: https://www.youtube.com/watch?v=Rkif6BbJmGc.

Competing interests

None declared.

Author contributions

DB, TSG, FPB and EO funding acquisition, project administration, mentoring, resources, and supervision. CEM and EO conceptualisation, methodology, and supervision. CEM, EYB, HS, MS, MB, LB, DR, JM and EO investigation. CEM, EYB, HS, MS, DA, JM, PEH, PT, JA, AMD and EO data curation and formal analysis. CEM, EYB and EO writing – original draft. All authors writing, review, and editing.

Disclaimer

The New Phytologist Foundation remains neutral with regard to jurisdictional claims in maps and in any institutional affiliations.

Supporting information

Fig. S1 Box plots depicting the root (dry) weight‐to‐sampling solution volume (L) ratio (RSVR) obtained for the different sampling approaches.

Fig. S2 Boxplots representing the alpha‐diversity community richness (Observed and Chao) of the sampling approaches and the unplanted soil.

Fig. S3 Boxplots representing the observed community richness of the supernatant control (green), the unplanted soil controls, and sampling approaches.

Fig. S4 Box plot showing the number of reads assigned to chloroplast as a proxy for plant tissue contamination across the different sample types of rhizosphere fractionation methods and the bulk unplanted soil.

Fig. S5 Microbiota quantification of nitrogen cycling microbial genes.

Notes S1 Practice abstract: Combined exudation and rhizobiome sampling from the same soil‐grown plant.

Notes S2 Primers, qPCR raw abundances, and cycling conditions used in this investigation.

Table S1 Root morphology parameters recorded after handling the plant samples with different sampling approaches.

Table S2 Results of qPCR determination of gene abundances of selected taxonomic and nitrogen cycling genes.

Table S3 Table illustrating the percentage of variance explained in pairwise comparisons between sampling approaches in bacteria and fungi.

Table S4 Microbial differential abundance between protocols expressed in taxa numbers and percentage of reads for the bacteria 16S and the fungal ITS.

Please note: Wiley is not responsible for the content or functionality of any Supporting Information supplied by the authors. Any queries (other than missing material) should be directed to the New Phytologist Central Office.

Data availability

The sequences generated in this study (16S rRNA and ITS reads) are publicly available in the European Nucleotide Archive (ENA), accession no. PRJEB97221. Scripts to reproduce figure and statistical analyses are available at: https://github.com/carmenmariaescudero/Root2Res_WP2_Rhizosphere‐Exudates_protocol_comparison and https://github.com/eybrowne/Root2Resilience.