Introduction

Recently, the negative impact of intensive agriculture on the environment and human health has led to a reevaluation of agricultural systems. Agroecology aims to improve agricultural production by employing natural processes, thereby reducing the reliance on synthetic inputs. The objective of agroecology is to utilize ecological processes and ecosystem services to develop and implement agricultural practices (Wezel et al., 2020). Nature-based management of the agroecosystem is at the heart of sustainable development for the next decades (Foley et al., 2011). The functioning of these agroecosystems is contingent upon soil microorganisms, which play a pivotal role in the health of soils, plants, animals, and humans (Banerjee and van der Heijden, 2023).

Vineyards combine a high economic and cultural value. France is the first grape producer globally, with an output of approximately 6 million tons, representing 20% of the production in the world (IOV, 2023). However, wine growers are facing numerous challenges, caught between stronger environmental constraints (e.g., higher levels of heat and water stress), economic viability and pressure to use fewer chemical inputs. Thanks to the great biodiversity they harbor (Paiola et al., 2020), organic vineyards support a wide array of ecosystem services (Winkler et al., 2017). Therefore, a sustainable viticulture will require a good comprehension of the grapevine microbiome, in order to adapt management practices, including possible microbial inoculation (Ochoa-Hueso et al., 2024).

The endomycorrhizal interaction between grapevine and arbuscular mycorrhizal fungi (AMF) is essential for both partners and soil health (Trouvelot et al., 2015). The benefits of AMF for grapevines are manifold and include enhanced nutrient uptake, improved access to water resources and protection from pathogens (see Torres et al., 2018; Aguilera et al., 2022 for a comprehensive list of benefits). AMF are also a key ecological compartment to maintain berry quality in grapevines under changing environments (Torres et al., 2018). However, AMF effectiveness may vary with changes in environmental conditions. Indeed, the beneficial effects of AMF can be exacerbated by conditions of limited nutrient and water availability in different ways. Resource availability can directly influence AMF growth and diversity. Studies have shown that increased precipitation reduces the density of AMF extra-radical hyphae (Wang et al., 2021), and, inversely, drought conditions increase sporulation and AMF diversity (Jafarian et al., 2024). Additionally, the alteration of host plants physiology and diversity in response to constraint conditions can affect AMF by selecting species as a function of stress (Biasi et al., 2023).

The diversity and composition of AMF communities are shaped by numerous variables, including plant species and genotypes (Guzman et al., 2021; Martin and van der Heijden, 2024). Environmental factors also shape AMF communities through both biotic (e.g., competition among AMF, Maherali and Klironomos, 2007) and abiotic (e.g., soil copper concentration, Betancur-Agudelo et al., 2023) characteristics. Biotic and abiotic factors can also interact in shaping AMF communities (Frew and Aguilar-Trigueros, 2023).

Likewise, AMF communities are affected by agricultural practices such as monoculture, soil tillage and elevated levels of fertilizer or biocide inputs (Trouvelot et al., 2015). Monocultures apply a strong negative selection pressure on biodiversity, including AMF, resulting in communities dominated by few taxa, better adapted to intensive agricultural practices (Verbruggen et al., 2010). Given the vulnerability of AMF mycelial networks, tillage has been demonstrated to affect AMF spore density, species richness and diversity (Säle et al., 2015; Thomopoulos et al., 2023). Indeed, deep plowing can result in the disappearance of certain species or the dispersal of their propagules to a deeper layer of soil, thereby reducing the level of root colonization. It has also been shown that synthetic herbicides harm root mycorrhization in herbal plant species (Zaller et al., 2018) and grapevines (Baumgartner et al., 2005) compared to mechanical weeding. Chemical components exert direct effects on AMF development and physiological metabolism (Yu et al., 2023).

Conversely, the establishment of cover crops promotes the proliferation of natural mycorrhizal communities (Rivera-Becerril et al., 2017). As an alternative to herbicides and to soil tillage for weed control, cover crops provide benefits to AMF communities by preventing soil disturbance and consequently mycelial destruction. Similarly, as AMF have a positive effect on plant diversity (van der Heijden, 2002), cultural practices based on high crop diversity enrich AMF communities (Guzman et al., 2021).

In the field of viticulture, the distinctions between agricultural systems reside in specific practices employed, generally related to the use of biocide (pesticides, fungicides, and herbicides). However, some practices can be shared by various agricultural systems, such as tillage, use of cover or organic inputs such as compost. As conventional agriculture uses large quantities of chemical fertilizers, biocides, and tillage to maximize crop yields, AMF spore density (Lin et al., 2020), species richness and diversity (Sheng et al., 2013; Oehl et al., 2004) tend to decline in comparison to organic plots. On the contrary, organic farming promotes AMF proliferation (Hart and Reader, 2002; Radić et al., 2014), colonization activity (Jiang et al., 2020) and AMF diversity (Oehl et al., 2004; Lin et al., 2012). The extensive use of organic inputs may counteract the detrimental effect of intensive tillage in organic farming, which is often used to replace herbicides (Rivera-Becerril et al., 2017; Van Geel et al., 2017).

We use the official definition of terroir as: “a concept which refers to an area in which collective knowledge of the interactions between the identifiable physical and biological environment and applied vitivinicultural practices develops, providing distinctive characteristics for the products originating from this area. Terroir includes specific soil, topography, climate, landscape characteristics and biodiversity features” (Resolution OIV/Viti 333/2010). The terroir concept can be considered a “black box” we need to explore (Brillante et al., 2020).

Although literature about terroir abounds, the incorporation of the microbial components of the terroir is poorly documented (Bokulich et al., 2016; Gilbert et al., 2014). A number of these studies concentrate upon the impact of microbes involved in the sulfur cycle or yeasts on the characteristics of wine (Mocali et al., 2020). The specific contributions of individual terroir components in explaining AMF community composition variation remain poorly understood. Recently, in a study of 200 vineyards, Gobbi et al. (2022) found that spatial distance was the primary explanatory variable for beta diversity of fungal and prokaryotic communities, at both global and local scales. The diversity of AMF is mainly influenced by the direct effects of climate (Jafarian et al., 2024). Inversely, according to Betancur-Agudelo et al. (2021), the most impactful component of terroir is soil properties, as soil directly influences nutrient availability and microbial habitats.

There is a paucity of data regarding the significance of AMF as a terroir component. Torres et al. (2019) postulated that infection with AMF might augment the amino acid content of the grapes, which may in turn affect the aromatic characteristics of the wine. This raises the question of the influence of terroir on AMF communities in vineyards across the country and how this terroir effect will interact with new ecoagricultural practices such as application of AMF-based biostimulants (Jindo et al., 2022). In order to utilize AMF-based biostimulants in a manner that does not adversely affect the natural AMF communities of a terroir, it is essential to evaluate the influence of soil properties, geographical distance, and other terroir characteristics on the communities that are naturally present in vineyard soils.

We aimed to quantify the effect of agricultural practices and terroirs on AMF communities. This was achieved by examining four key facets of AMF ecology:

i. Spore numbers.

ii. Mycorrhization rate.

iii. Alpha-diversity (local AMF diversity).

iv. Beta-diversity (change in AMF communities’ composition across samples).

Materials and methods Study sites

The objective of our study was to investigate the abundance and diversity of AMF within the French vineyard. To this end, we conducted a comprehensive analysis of 75 different vineyards, distributed across 14 distinct wine terroirs and encompassing six major wine-producing regions in France: Bordeaux, Bourgogne, Camargue, Champagne, Côtes-du-Rhône, and Languedoc. Our study considers a wide range of pedoclimatic and management conditions, as detailed in Supplementary Table S1. Each terroir refers to a specific protected designation of origin (AOP), controlled designation of origin (AOC) or protected geographical indication (PGI) according to European Union regulation N° 1308/2013 dated 17th December 2013. Vineyard plots were selected in each terroir based on their agricultural practices (conventional, conversion, organic) and weed management in row and inter-row (chemical control by application of herbicide, mechanical control by scratching or plowing, grassed). The conventional plots primarily utilized synthetic products, though sulfur and copper treatments were intermittently employed, along with mineral fertilization. The organic plots were managed in accordance with the European Union Regulation (EEC) No. 834/2007, which excludes synthetic pesticides and inorganic fertilizers. The plots undergoing conversion to organic were in the first, second, or third year of conversion.

This large-scale study was conducted out over several years, spanning from 2019 to 2023. All plots within the same terroir were sampled in the same year. All soils and roots in each of the 75 plots were sampled at the same time, from late June to early July, regardless of the year. Indeed, early summer in France corresponds to grapevine fruit set, when arbuscular colonization reaches its highest level (Schreiner, 2005). Some previous studies carried out on AMF diversity in vineyards have had to deal with plots that had received AMF-based biostimulants (e.g., Bouffaud et al., 2016a). To avoid any bias in AMF communities’ composition, we verified beforehand with winegrowers that no fields had been inoculated with AMF in the past. In each of the 75 vineyards, we counted spores, measured mycorrhization rates, characterized AMF communities using metabarcoding. Additionally, we analyzed physicochemical soil characteristics for 49 vineyards.

Soil, root sampling and processing

In each vineyard, a total of 12 plants per plot were sampled by selecting 4 consecutive plants in three homogenous rows across the vineyard, avoiding border rows. Soil (approximately 150 g) and grapevine roots (enough to fill a 2 mL tube) were sampled early after fruit set (between late June and early July) at a soil depth of 20 cm at the base of each selected plant. Only thin and young grapevine roots with approximately 1 mm in diameter were harvested. Grapevine black roots are easily visually distinguishable from the roots of other plant species. Soil and root samples were combined in 3 pools from 4 neighbors’ plants, each totalizing 3 soil and root composite samples per plot. For all Vergèze and Côte des Blancs plots, all 12 root samples were pooled into a single composite sample (Figure 1).

Figure 1. Schematic sampling strategy of the two AMF compartments. * For samples in Vergèze and Côte des Blancs terroirs, all three root samples were pooled before the unique extraction but 3 PCR were used to amplify DNA. eDNA, environmental DNA; AMF, Arbuscular Mycorrhizal Fungi.

All samples (soil and living roots) were stored in plastic bags at 4°C until processing, which was conducted within 1 week. Each root sample was washed free of soil and then divided into two portions. The first portion (a random subset of sixteen 1-cm-long root segments) was submerged in 70% ethanol and stored at 4°C before being colored for AM fungal colonization determination. The second portion was stored at −20°C until DNA extraction and subsequent molecular analyses of AM fungal communities. A 200 g soil aliquot was separated from each soil replicate sample, sieved at 2 mm and reserved for spore density and diversity analysis. To reach a good representation of AMF spore bank diversity, all spores found in soil samples were sorted out before DNA extraction. As we want to study the AMF communities in the Vitis vinifera rhizosphere as a whole, we combined environmental DNA (eDNA) samples from roots and from the spore bank (Figure 1) at the end of the bioinformatic pipeline.

Soil characterization

For each plot (with the exception of some Vergèze plots and all plots of Côte des Blancs), soil physicochemical properties were analyzed from a 500 g soil aliquot. The resulting 49 soil samples were analyzed in the Laboratoire d’Analyses de Terres, de Végétaux et Environnementales of the Chambre d’Agriculture de l’Aude (France). In brief, air-dried soil was sieved through 2 mm sieves. In 1:5 soil-to-water (w/v) suspension, the pH and electrical conductivity (EC) of the soil were determined using a pH-EC meter. Particle-size distribution was determined by the hydrometer method (Bouyoucos, 1962). Organic carbon was determined using the wet-oxidation method by Walkley and Black and expressed as organic matter using Vant Hoff’s factor (1.72). Total nitrogen was quantified using the Semi-Micro Kjeldahl method (Bremner and Mulvaney, 1982). The Joret-Hébert method was used for phosphorus extraction. Exchangeable bases, calcium (Ca), magnesium (Mg), potassium (K), and sodium (Na) were assessed by soil saturation with neutral 1 M ammonium acetate and measured via atomic absorption spectrophotometry (Sumner and Miller, 1996). Cation exchange capacity (CEC) was determined by the Kjeldahl distillation method. The determination of available micronutrients, iron (Fe), copper (Cu), manganese (Mn) and zinc (Zn) involved diethylene triamine pentacetic acid (DTPA) (Lindsay and Norvell, 1978). Measurement of these micronutrients was performed using an atomic absorption spectrophotometer. EC1:2.5 was potentiometrically measured in a 1:2.5 soil-to-water ratio according to Okalebo et al. (2002).

Determination of AMF spore density

AMF spores occurring in soil samples were extracted following the wet sieving method adapted from Gerdemann and Nicolson (1963) and Daniels and Skipper (1982). For each soil replicate sample, 50 g was sieved through three nested sieves with meshes of 1,000, 400, and 45 μm. Then, spores were purified by re-suspending the sieving in a 60% sucrose solution and centrifugation was carried out at 3,000 rpm for 2 × 3 min. The supernatant was removed and poured into the 45 μm sieve. Retrieved AMF spores were re-suspended in 15 mL of water. A 1 mL aliquot was placed in Petri dishes and spores were counted under a stereomicroscope (40 × magnification). Average numbers were calculated per 100 g of dry soil. The remaining 14 mL were stored at −20°C until DNA extraction and subsequent molecular analyses. As a result, each one of the 75 spore bank sample for eDNA diversity analysis correspond to spore bank of 140 g of soil (50 g × 14 mL / 15 mL × 3 replicates).

Determination of AM fungal colonization

Sampled root fragments (3 replicates of 16 fragments of 1 cm long for each plot) were cleared in 10% KOH at ambient temperature for 10 h. Highly pigmented grapevine roots were additionally cleared in 3% w/v H2O2 for 40 min at 70°C and rinsed with distilled water. Root fragments were then colored with Schaeffer black ink, as described in Vierheilig et al. (1998). Thereafter, the samples were immersed in a mixture of 50% glycerol in water. Roots were mounted onto microscope slides and examined under 200–800 × magnification. The number of sections where mycorrhizal arbuscules, vesicles or hyphae were observed was noted separately for each structure type. For each replicate, the frequency of mycorrhiza (F %), root mycorrhization rate (M %), and arbuscular (A %) abundance in the root system were evaluated according to Trouvelot et al. (1986) using the MYCOCALC program.

DNA extraction, amplification, and sequencing

AMF spores were extracted from each of the three soil replicate and ground in buffer solution (0.4 M NaCl, 10 mM Tris–HCl pH = 8, SDS 0.2%, 2 mM EDTA pH = 8). Spore DNA was extracted using the FastDNA Spin kit for Soil (MP Biomedicals, Europe) according to the manufacturer’s instructions. Elution of DNA was done using 200 μL of DNase-free water. The 3 DNA extracts replicates were mixed and diluted 1:50 in DNase-free water.

Root samples were ground in liquid nitrogen using a mortar and pestle. Genomic DNA was extracted from 250 mg of roots using the FastDNA Spin kit for Soil (MP Biomedicals, Europe) according to the manufacturer’s instructions. Elution of DNA was done using 100 μL of TE and stored at −20°C. The DNA extracts were diluted 1:10 in DNase-free water.

The DNA of arbuscular mycorrhizal fungi was amplified using 18S rRNA gene primers. The first PCR reactions were performed in triplicates. A fragment of 510–570 bp covering a variable region of the SSU was amplified using the universal eukaryotic primer NS31 (Simon et al., 1992; TTGGAGGGCAAGTCTGGTGCC) in combination with the AMF-specific primer AML2 (Lee et al., 2008; GAACCCAAACACTTTGGTTTCC) and included overhang adaptor sequences for the Nextera primer (Illumina Inc., CA, United States).

PCRs were performed in a total volume of 20 μL with 1 μL DNA, 0.2 μL of each specific primer (10 μM), 4 μL 5X Platinum II PCR buffer (Thermo Fisher, Massachusetts, Etats-Unis), 0.4 μL 10 mM dNTP mix, 0.32 μL Platinum II Taq Hot-Start DNA Polymerase, 13.88 μl of DNase-free water. The PCR cycle was as follows: 2 min at 94°C, (15 s at 94°C, 15 s at 55°C, 30 s at 72°C) for 35 cycles and a final elongation step at 72°C for 10 min. The PCR products were purified with magnetic beads (AMPure XP).

The second PCR was performed using a Nextera® XT Index Kit (Illumina, San Diego, United States) following the manufacturer’s instructions. After purification with magnetic beads (AMPure XP), these final PCR products were merged by triplicate (Figure 1), dosed with kit KAPA Library Quantification kit (Roche), multiplexed and sequenced on a MiSeq Illumina sequencer using MiSeq Reagent Kit v3 (600-cycle, Illumina).

Bioinformatics

DNA sequences were analyzed through the bioinformatics pipeline described in Supplementary Report S1. This pipeline draws 4.32 kWh (see Supplementary Note S1 for details) which results in a carbon footprint of 221 gCO2e (calculated using R package greenAlgoR, Taudière, 2024, algorithm based on Lannelongue et al., 2021). In short, primers were removed using cutadapt (v. 4.5, Martin, 2011). Sequences were quality filtered using filterAndTrim function from the dada2 package (v. 1.30.0; Callahan et al., 2016a) discarding sequences with default parameters. Then we followed dada2 classic pipeline (Callahan et al., 2016b) to obtain chimera-free amplicon sequence variants (ASV) using single forward (R1) sequences. Each ASV longer than 300 pb was then taxonomically assigned to two taxonomic databases with the assignTaxonomy function from dada2, which implements the RDP classifier of Wang et al. (2007). First, we used the PR2 database (v. 5.0.0; Guillou et al., 2012) to assign the taxonomy at the scale of Eukaryota. Second, we used the AMF specific database Maarjam (Öpik et al., 2010) to assign more specifically arbuscular mycorrhizal fungal OTUs.

Following recommendation by Tedersoo et al. (2022), we added a step of reclustering on ASV sequences to obtain a more classical version of OTU using the function asv2otu from the MiscMetabar package (v. 0.9.4; Taudière, 2023). The idea is to denoise using dada and then to cluster into taxonomic unit using vsearch software (v. 2.22.1; Rognes et al., 2016) at a 97% identity level. We also repeated the analysis on the ASV dataset in Supplementary Report S2 to identify potential differences in key results between the two approaches, as recommended by Joos et al. (2020).

After reclustering, we filtered out all non-AMF sequences using two filters. First, all sequences with less than 80% identity similarity with at least one sequence in Maarjam database were discarded (MiscMetabar::blast_pq function). Second, we also discarded OTUs assigned to other families than Mucoromycota by the PR2 database. Numbers of sequences across the major step are present in Table 1. Except for soil compartment analysis (spores vs. roots), we merged paired samples of spores and roots in one sample. We decided to pool root and spore samples because the two compartments bring different views on AMF communities, and we are interested in the whole rhizosphere of Vitis vinifera. Moreover, the different nature of the two compartments makes the result from metabarcoding difficult to compare and would bring a non-necessary level of complexity.

Table 1. Number of sequences, clusters (i.e., unique sequences, ASV or OTUs depending on the step) and samples across the main step of the bioinformatic pipeline.

Statistical analysis

All statistical analyses were carried out using R Studio software (Posit Team, 2024, version 2023.12.1) and R version 4.3.3 (R Core Team, 2023). Code for statistical analysis, tables, and figures is available in Supplementary Report S3. Most important packages are dada2 (v. 1.30; Callahan et al., 2016a), MiscMetabar (v. 0.9.4; Taudière, 2023), phyloseq (v. 1.46.0; McMurdie and Holmes, 2013), targets (v. 1.4.1 Landau, 2021), ggstatsplot (v. 0.12.3 Patil, 2021) and vegan (v. 2.6-4; Oksanen et al., 2022). Minimal graphical adjustments to improve the figures’ visibility were performed in Inkscape (Inkscape Team, 2023).

The local biodiversity of AMF (alpha-diversity) were assessed using the Hill number framework (Hill, 1973) recommended by Alberdi and Gilbert (2019) for DNA-based diversity analyses. The importance of the abundance distribution increases with increasing Hill order q. The Hill number for q = 0 (H0) is the richness, when q = 1 (H1), it is the exponential Shannon entropy and for q = 2 (H2), it is the inverse Simpson index.

To describe soil chemical properties, principal component analysis (PCA) was performed using ade4 R package (Dray and Dufour, 2007) and visualization with package FactoMiner (Le et al., 2008) and factoextra (Kassambara and Mundt, 2020). The PCAtest function from the PCAtest package (Camargo, 2024) was used to test for significance of PCA dimensions after correcting p-value for multiple-testing.

To study beta-diversity, we accounted for spatial autocorrelation in samples using distance-based Moran’s eigenvector maps (function dbmem from the adespatial package; Dray et al., 2023). The effect of terroir and practice on AMF beta-diversity was computed using a Permanova with the formula:

Distance~nb_seq+MEM_1+MEM_2+Dim.1+Dim.2+Dim.3+practice+rank+inter_rank+terroir

where nb_seq is the square roots of the number of reads per sample, MEM_1 and MEM_2 are the dbmem dimension; Dim.1, Dim.2, and Dim.3 correspond to the first three axes of soil PCA; practice corresponds to either organic, conventional, or conversion agricultural; rank and inter_rank correspond to weed management in the vineyard row and inter-row, terroir is the list of 14 terroirs.

Permanova were computed on bray-curtis distance. This allows us to consider differences in sample sequencing depth without discarding so many sequences. We show in supplementary tables the result of Permanova on bray-curtis distance after rarefaction, as well as the result of Permanova on robust-Aitchison distance. Test for multivariate homogeneity of groups dispersions (variances) were done using function vegan::betadisper (Supplementary Report S3).

Furthermore, we plotted beta-diversity results using Non-metric Multi-Dimensional Scaling (NMDS) ordination and upset plot (Lex et al., 2014). Indicator species were identified using the multipatt function from the indicspecies package (Cáceres and Legendre, 2009) with the IndVal.g metrics. Finally, soil, distance, terroir and practice influence on beta-diversity were measured using partition of variance of distance-based redundancy analysis function vegan::varpart (Oksanen et al., 2022). To circumvent the stochastic phenomenon of rarefaction, we ran 99 distinct rarefactions and report the mean value of adjust R2 only when at least 95% of rarefaction runs resulted in significant adjust R2 values (function MiscMetabar::var_par_rarperm_pq). As soil properties were only available for 49 samples, we computed Permanova and variance partioning on both the 75 samples without information about soil and the 49 samples with information about soil.

Results Soil properties vary between terroirs, but not between agricultural practices

A PCA on 17 soil variables (Table 2) was computed for the 49 samples for which both soil properties and AM diversity are known. The first three dimensions (72.8% of variance explained) were selected using manual inspection of scree plot. The first axis contrasts samples with sand versus samples with clay, high nitrogen (N) and high carbon (C). The second axis mostly represents differences in phosphorus pentoxide (P2O5), copper (Cu) and silt. Lime, pH and coarse silt drive the third axis (Figure 2). The three first axes significantly vary across terroirs, but not practices (Supplementary Figure S1, Kruskal-Wallis test p-values in Supplementary Figure S1). There is also no effect of practice on copper concentration alone (Kruskal-Wallis χ2 = 0.61, p-value = 0.73, Supplementary Figure S2).

Table 2. Mycorrhization, diversity and soil characteristics for each terroir and practice.

Figure 2. PCA of soil physicochemical variables (49 samples and 17 soil variables). Correlation graph of soil variables for the first two dimensions (A) and for dimension 3 and 4 (B).

Mycorrhization rates and spore counts vary with terroirs and practices

There is no correlation between the number of spores and the number of AM fungal species (Spearman rank correlation test, p-value = 0.58; Supplementary Figure S3). We found no correlation between the three Hill numbers and the three mycorrhization measures A %, F %, and M % (Pearson test; Supplementary Figure S4). Spore counts are correlated with the three mycorrhization measures A % (Spearman test, p-value = 6e-3), F % (p-value = 0.001) and M % (p-value = 6e-4).

The number of spores significantly differs among terroirs (Kruskal-Wallis χ2 = 48.54, df = 14, p-value = 5e-6; Supplementary Figure S5). Using pairwise tests (Dunn with p-value Holm-adjustement), the terroir of Côte des Blancs shows a significant higher number of spores compared to Vergèze, Langoiran, Montcalm, Aigues-mortes and Vallée de la Marne. Mycorrhizal frequency (F %, p-value = 6e-10), intensity of the mycorrhizal colonization (M %, p-value = 1e-03), and arbuscule abundance in the root system (A %, p-value = 0.001) vary across terroirs.

The number of spores slightly varies with practices with organic vineyards presenting a significant higher number of spores than non-organic vineyards (Kruskal-Wallis, p-value = 0.044, Supplementary Figure S6). Organic agricultural practices (Kruskal-Wallis test and pairwise Dunn test) improve mycorrhizal frequency F % (mean = 92.5 vs. 88.4 for conversion and 89.4 for conventional; p-value = 0.046, Table 2), colonization intensity M % (mean = 20.5 vs. 9.7 for conversion and 10.2 for conventional; p-value = 5.3e-05), and arbuscules abundance A % (mean = 13.9 vs. 9 for conversion and 6.7 for conventional; p-value = 1e-03).

Taxonomic filtering of eDNA (environmental DNA) sequencing

The bioinformatic pipeline identified 788 OTUs (5,937,106 sequences) from which 213 (3,617,248 sequences) were assigned to arbuscular mycorrhizal (AM) fungi using two criterions: 80% identity method to Maarjam database and PR2 assignation to Mucoromycota Family (Table 1). The first step filters out 37.1% of sequences and 70.4% of OTUs. Second, we also discarded OTUs assigned to other family than Mucoromycota in PR2, leading to an additional removal of 20 OTUs (113,338 sequences). The majority of non-Mucoromycota sequences are classified as Arthropoda (335 OTUs and 1,764,837 sequences), Nematoda (73 OTUs and 167,513 sequences), and Tardigrada (29 OTUs and 133,628 sequences, Supplementary Figures S7–S9). After merging paired spore and root samples, the sequencing depth varies from 6,768 to 101,183 (Table 1; mean = 48,230, sd = 19,620) AMF sequences per sample. Note that the 213 OTUs clustered a total of 3,695 ASV. All following results are robust to the decision to re-cluster ASV into OTUs (Supplementary Report S2).

Among the 9 families of the 213 OTUs, Glomeraceae is the most abundant (64.8% of OTUs, 81% of sequences) followed by Claroideoglomeraceae (8.5% of OTUs, 10.8% of sequences), Diversisporaceae (7% of OTUs, 3.9% of sequences), Paraglomeraceae (5.1% of OTUs, 2.5% of sequences) and Archaeosporaceae (5.1% of OTUs, 0.7% of sequences). The most abundant OTU is a Glomus species and accounts for 33% of the total number of sequences. Three Genus dominate the AMF communities using PR2 assignation: Glomus is the most abundant (42.3% of OTUs, 59.5% of sequences) followed by Rhizophagus (10.8% of OTUs, 18.8% of sequences) and Funneliformis (7.5% of OTUs, 12.7% of sequences).

Effect of soil compartment

Samples from spore and root compartments highly differ in terms of alpha and beta-diversity (Supplementary Figure S10). Root samples present a higher richness but similar hill numbers H1 and H2 (test de Mann–Whitney). For subsequent analysis, samples from roots and spores were merged by pairs leading to a total of 75 samples, 213 OTUs and 3,617,248 sequences (Table 1).

Terroir and, to a lesser extent, practices, have an impact on the diversity of AMF

Terroir drives the AMF diversity (Table 2; Figure 3A; Supplementary Figure S11). The Gigondas terroir shows the highest richness (median = 33; 35.5 without rarefaction) whereas Côte des Blancs is the poorest (median = 9 with or without rarefaction). If we focus on the total diversity present in a given terroir, Cévennes (79 OTUs, Figure 4A) and Faugères (78 OTUs) present the highest diversity. Note that accumulation plots (Figure 3) are not here to assess the absolute soundness of our sampling because as our pipeline discards singletons, we care unable to draw a correct accumulation curve. However, comparison of curve shapes indicates that Rian, Faugères and Cévennes are not fully sampled with our sampling effort.

Figure 3. Pseudo-accumulation curves of AMF across terroirs (A) and practices (B). We rarefied the number of samples per modality as well as the total number of sequences per modality using 999 permutations (function MiscMetabar::accu_plot_balanced_modality). Filled areas show the 90% quantile distribution.

Figure 4. Distribution of OTUs across terroirs (A) and practices (B,C). Panel (B) is a venn diagramm depicting the number of shared OTUs. Panel (A,C) are upset plot. The matrix with point at the bottom of the figure represents the intersection between modalities (terroir or practice). In panel (A) only the case with at least 3 OTUs are plot. Colors depict OTUs Family. For example, 7 OTUs (6 Glomeraceae and 1 Archaeosporaceae) are found in vineyards under conversion and conventional practices but not in organic vineyards. Concerning terroir, 6 OTUs (5 Glomeraceae and 1 Acaulosporaceae) are specific to Rian. Cévennes terroir harbor 79 OTUs while 85 OTUs were found in all conventional farming samples.

Practices drive Hill number 1 and 2 (Figure 5 and Supplementary Figure S12 for analysis without rarefaction) with organic samples tending to harbor a higher AMF diversity than conventional ones. Focusing on the total diversity per farming practice gives a complementary vision for the richness facet. Organic vineyards and vineyards under conversion harbor a higher richness if we regroup samples per modality (Figures 4B, 3B). Organic samples display a total of 151 OTUs (39 samples), whereas conversion samples are represented by 119 OTUs (17 samples), and conventional ones by only 85 OTUs (19 samples). Thus, samples in organic and conversion vineyards harbor on average the same number of OTUs as conventional vineyards, but when considering all samples together, conventional vineyards harbor far less diversity.

Figure 5. Diversity of AMF across practices. Hill number 0 is equivalent to richness (number of Species), Hill number 1 to the exponential of the Shannon index and Hill number 2 to the inverse of the Simpson index.

Terroirs and, to a lesser extent, practices, drive community composition of AM fungi

Terroir is the major driver of AMF community composition, even if we control for soil properties and spatial autocorrelation (Table 3; Figures 6, 7). Spatial autocorrelation, and soil properties beyond terroir characteristics, also shape AMF communities. Eight OTUs were common to all 14 terroirs. Cévennes and Faugères terroirs, respectively, harbor 26 and 25 specific OTUs despite these terroirs being represented by only 4 samples. On the contrary, Côtes des blancs harbor only 5 unique OTUs despite 15 samples (Figure 4A). Moreover, among OTUs specific to Faugères, 7 out of 9 AMF families are present. Specific OTUs are OTUs found in only one terroir, but can be represented by only one sample. Thus, these specific OTUs may not be indicators of terroir if the number of samples is not sufficient to achieve statistical significance. Ten OTUs are indicators of terroir (Figure 8). Eight terroirs out of 14 present at least one indicator species. Gigondas (OTU_120, OTU_21, OTU_28 and OTU_65) and Aigues-mortes (OTU_198, OTU_43 and OTU_99) are the only terroirs characterized by specific indicator species.

Table 3. Result of the Permanova (A with all 75 samples and B including soil physicochemical characteristics leading to only 49 samples) on Bray-curtis distance.

Figure 6. Variance partitioning of AMF community composition with (A) or without (B) soil component. This is the mean result of 99 rarefaction permutations (function MiscMetabar::var_par_rarperm_pq) on Bray Distance. The effect of soil, spatial and terroir components are statistically supported. When 5% of rarefaction permutations are below 0, values are not shown. See Supplementary Figure S13 for analysis with robust Aitchison distance.

Figure 7. Non-metric dimensional scaling (NMDS) of AMF community using Bray-Curtis distance. Conventional farming: triangles pointing downwards, conversion farming: diamond, organic farming: triangles pointing upwards. Different colors represent different terroirs.

Figure 8. Indicator OTUs for terroir. Significant association between one, two or three terroirs with indicator OTUs are displayed using circles for IndVal.g metric. Terroir without indicator OTUs are not shown. None of indicator OTUs were found in association to agricultural practice.

Agricultural practices (organic, conventional, conversion) as well as rank and inter-rank work, slightly impact the AMF communities (Figure 6; Supplementary Table S4). These effects are even non-significant when we first control for soil effect and/or spatial autocorrelation in the Permanova (Table 3; Supplementary Table S5) even if we use robust Aitchison distance (Supplementary Tables S6, S7). Fifty-one OTUs (24%) are shared between all practices, less than the number of OTUs only found in organic vineyards (67 OTUs; Figures 4B,C). Fifteen OTUs (7%) are only present in vineyards under conventional practices, whereas 40 (19%) are specific to vineyards under conversion. Using indicators species analysis, we found no OTUs to be significantly linked to practice for both presence/absence and abundance-based analysis.

Which vineyards are poor in AMF diversity?

Whatever the practice and terroir, there is a high variability in the diversity of AMF (Table 2). Thirteen samples present a low diversity of AMF (H1 or H2 < 2; Table 4). Most low-diversity samples are conventional (8 vs. 3 conversion and 2 organic samples) and are located in Côte des Blancs terroir (7 out of 13, the 6 other samples represent 6 different terroirs). Interestingly, 5 out of 19 conventional samples were not treated with herbicide and none of these samples belong to the 8 conventional samples with low-diversity (χ2 = 4.94; p-value = 0.026).

Table 4. Characteristics of low-diversity samples.

Discussion AMF communities in vineyards

One of the most promising solutions for a sustainable viticulture is the incorporation of mutualist interactions between plants and microorganisms, such as arbuscular mycorrhizal fungi, into agricultural practice (Aguilera et al., 2022). Here, we characterize AMF communities in the wine rhizosphere across France, examining the influence of different agricultural practices in 14 terroirs (Figure 9). The taxonomy of AMF observed in French vineyards is consistent with that reported in previous studies conducted in France (Drain et al., 2019), Brazil (Bezerra et al., 2021), Canada (Holland et al., 2014), and Portugal (Fors et al., 2023). These studies have identified the genus Glomus, Rhizophagus, and Funneliformis as the dominant AMF genera in these communities.

Figure 9. Schematic summary of the effect of practice and terroir on the four facets of AMF interactions. F %: mycorrhizal frequency in the root system, M %: intensity of the mycorrhizal colonization in the root system, A %: arbuscule abundance in the root system. All links are supported by correlations tests except for beta-diversity whose links are supported by Permanova and variance partitioning.

Spore-bank samples contain less AMF OTUs than root samples. This outcome may be attributed to variations in sampling, extraction, or amplification techniques between the two compartments. In particular, a notable discrepancy between spore and root samples is that we only sequenced a few thousand spores per sample, which is insufficient to encompass the total diversity present in the spore bank. Biological mechanisms can also