Spore associated bacterial communities display patterns related to the arbuscular mycorrhizal host genotype in Mediterranean sand dunes

Abstract

Background

Arbuscular mycorrhizal fungi (AMF) are beneficial microorganisms establishing symbioses with the roots of most plants, supporting their nutrition and health, mainly in nutrient-poor and harsh environments. AMF spores host diverse bacterial communities whose distribution in the different taxa is poorly understood. To this aim we determined the entire diversity of spore associated bacteria thriving in Ammophila arenaria rhizosphere from Mediterranean maritime sand dunes by Illumina sequencing.

Results

The two most abundant AMF species, Racocetra fulgida and Racocetra persica disclosed a highly heterogeneous composition of spore bacterial communities. The endosymbiont Ca. Moeniiplasma glomeromycotorum was abundant in both AMF species and detected in R. fulgida for the first time. Ca. Glomeribacter gigasporarum was present only in R. persica. Distinct highly diversified spore-associated bacterial communities were also identified, with 251 Amplicon Sequence Variants (ASVs) affiliated to 69 and 34 genera in case of R. persica and R. fulgida, respectively. The regular detection of Actinomycetota confirmed a relatively stable relationship with AMF spores. Most of the identified bacteria belonged to culturable genera and species, reported to possess key beneficial activities.

Conclusions

Our study showed that different AMF genotypes differentially recruited highly diversified and rich spore associated bacterial communities in a natural sand dune system. The isolation in pure culture of such bacteria could boost in-depth studies aimed at revealing their functional role in the complex interactions between AMF, associated bacteria and host plants and their possible exploitation in harsh agroecosystems.

Supplementary Information

The online version contains supplementary material available at 10.1186/s40793-026-00858-y.

Background

Maritime sand dunes represent a harsh environment where plants face several different stresses, from drought and salinity to seasonal extreme temperatures, high evaporation rates, dune instability and low levels of nutrients and organic matter [1, 2]. In such environment, only few plant species are able to survive, the most studied being represented by the xerophytes Ammophila arenaria and Ammophila breviligulata (European and American beachgrass, respectively) which are dominant sand-colonizing plant species in maritime sand dunes of Europe, Mediterranean basin and North America [3]. Moreover, given beachgrasses ability to trap and accumulate windblown sand—thus contributing to make dunes more resistant to erosion by wind and water—they are considered dune-building grasses and utilised for their ecosystem engineering traits [4, 5].

It has long been known that beachgrass can thrive in nutrient-poor and challenging environments, like maritime sand dunes, thanks to the support of arbuscular mycorrhizal fungi (AMF), a group of beneficial soil microorganisms that establish mutualistic symbioses with the roots of ca. 80% of land plants, including the major food and industrial crops, such as cereals, legumes, fruit trees, vegetables, cotton, sunflower and tobacco [6]. AMF obtain carbon compounds from host plants, while the extraradical networks of hyphae, growing from mycorrhizal roots into the nearby soil, uptake water together with mineral nutrients and transport them into the roots, representing a highly efficient auxiliary absorbing system [6]. In sand dune ecosystems, AMF are able to improve the efficiency of soil essential nutrients and water use by plants, and protect plants against oxidative damages caused by drought and other environmental stresses [7–11]. For this reason, field and greenhouse studies have been carried out in order to determine AMF ability to promote beachgrass growth and development and to withstand periods of limited water availability [3, 12–16].

The key functions of AMF are facilitated and complemented by bacterial populations living associated with their hyphae and spores and playing important roles as plant growth promoting (PGPB) and mycorrhiza helper (MHB) bacteria [6]. Overall, the identification and the isolation in pure culture of hundreds of AMF associated bacteria extended our understanding of their functional significance, providing new clues for their exploitation in sustainable agroecosystems [17–19]. The large and highly diverse and mostly culturable bacterial communities tightly associated with AMF spores, often embedded in spore wall layers [20], showed two main functions: as PGPB, being able to enhance plant nutrition and health by nitrogen fixation, phosphorus solubilisation and the production of plant hormones, siderophores and antibiotics, and as MHB, increasing spore germination, germling hyphal length and mycorrhizal establishment, thus indirectly promoting plant performance [20–25]. However, no such studies have been carried out on AMF spores living in harsh environments, such as maritime sand dunes, aimed at detecting and functionally characterizing the associated bacteria possibly beneficial to fungal symbionts and/or host plants.

Besides such spore-associated bacteria, AMF spores host intracellular unculturable endosymbionts, that were detected by ultrastructural studies as early as 1970 [26–29], later confirmed by molecular methods [30, 31], and assigned to the species Candidatus Glomeribacter gigasporarum (CaGg, Betaproteobacteria) [32] and Candidatus Moeniiplasma glomeromycotorum (CaMg, Mollicutes) [33]. CaGg was reported to positively affect germling growth and protein expression, increase fungal sporulation and ATP production, and elicit mechanisms to detoxify reactive oxygen species in the isolate BEG34 of Gigaspora margarita [34–36]. On the other hand, the role of CaMg is still to be revealed, being currently considered conditional mutualists of AMF [37].

Given the great functional significance of spore associated bacteria for AMF ecological fitness, and consequently for the fitness of their host plants, in this study we assessed the diversity of such complex microbial communities in A. arenaria rhizosphere growing in Mediterranean maritime sand dunes, an environment subject to seasonal drought, located in a Biosphere Reserve on the coast of Tuscany, near Pisa, Italy. To this aim we i) morphologically and molecularly characterized the community composition of AMF spores, the level of mycorrhizal colonization and the inoculum potential in A. arenaria rhizosphere, ii) determined the entire diversity and community composition of bacteria associated with the spores of the two most abundant AMF species, using Illumina high-throughput metabarcoding sequencing.

Materials and methods

Description of the sampling site

The selected site is located in a protected area in the Migliarino-San Rossore-Massaciuccoli Natural Park, inside the UNESCO Biosphere Reserve “Selve Costiere di Toscana”, in the north coast of Tuscany, near Pisa (Italy) (approximately 43°48′N, 10°15′E, Fig. S1). It is an old fixed dune system, with many dunes not exceeding 2–3 m in height, extending some 200 m inland, where a sclerophyllous forest together with plantations of pine trees grow. The dune system vegetation is well preserved and includes many herbaceous species, such as A. arenaria (L.) Link, Cakile maritima Scop., Eryngium maritimum L., Helichrysum stoechas (L.) Moench, Medicago littoralis Rohde ex Loisel and Medicago marina L. [38]. The foredunes are dominated by A. arenaria, an erect perennial grass growing in dense tussocks on the top of sand dunes, where extensive networks of underground rhizomes and adventitious roots represent the sand-binding main structures, contributing to building up the volume of the dunes (Fig. S1). The climate is typically Mediterranean, with a precipitation average of about 879 mm year⁻¹ and a mean annual temperature of 15.2 °C (11.5 °C in January; 30.2 °C in August). The main soil physicochemical properties were: sand 96.9%; silt 0.5%; clay 2.6%; pH (H₂O), 8.0; organic C, 0.10%; organic N, 0.09‰; available P (Olsen), 3.5 ppm.

Sampling

Samples of roots and rhizosphere sand were collected in May 2023 inside a 10 × 20 m collection area, 50 m far from the water’s edge, where there was a monospecific cover of A. arenaria in the foredune crests. Late spring was reported to be the most appropriate season for the direct isolation of the highest number of AMF spores in the investigated dune ecosystem, although in May spore density was found to vary from 13 to 47 per 100 g sand [39]. Thus, in order to obtain an adequate spore yield, approximately 15 kg of rhizosphere sand and roots were extracted using a cylindrical sampler to minimize environmental disturbance. As the sampling area was a monospecific cover of A. arenaria tussocks on the top of sand dunes (see Fig. S1), we obtained 30 samples by extracting horizontal cores from adjacent dunes, after removing the external sand layer before digging each hole. The samples containing rhizosphere sand and roots were pooled and transported to the laboratory, where three root samples (3 g each) were prepared for assessing mycorrhizal colonization. Three samples of rhizosphere sand (~ 5 kg each) were maintained in polyethylene bags at 4 °C prior to spore extraction.

Mycorrhizal colonization of Ammophila arenaria roots

Mycorrhizal colonization was assessed on three aliquots of A. arenaria roots (3 g each). The collected roots were carefully washed with tap water, cleared with 10% KOH in water bath at 80 °C for 15 min, neutralized in 2% aqueous HCl, and stained with 0.05% Trypan blue in lactic acid at 80 °C for 5 min. Percentage of colonized root length was assessed under a dissecting microscope (Wild, Leica Milano, Italy) at magnification 25 × or 40 × by the gridline intersect method [40]. Colonised roots were selected, mounted on microscope slides in lactic acid and observed under a Reichert-Jung (Wien, Austria) Polyvar light microscope to detect intraradical fungal structures. Images were captured using a Leica DFC 450C color video camera and the software package LAS X 5.2.0261303 (Life Science Microscope Software Platform, Leica Microsystems srl, Italy) connected to the microscope.

Mycorrhizal inoculum potential of rhizosphere sand

The activity of native AMF in the rhizosphere sand was assessed by the mycorrhizal inoculum potential (MIP) bioassay, using Cichorium intybus L. as biotest plant, as described in Turrini et al. (2018) [17], using three sand aliquots (200 g each). In brief, seeds of C. intybus L. “Zuccherina di Trieste” were sown into sterile 50 mL Falcon tubes, each filled with 40 mL of sand samples. Four replicate tubes were prepared for each sand aliquot. After germination, the seedlings were thinned to three per tube. The tubes were sealed in transparent Sun bags (Sigma-Aldrich, St. Louis, MO, USA) and placed in a growth chamber set to 24/21 °C (day/night) with a 16-h light and 8-h dark cycle. Plants were watered as needed weekly. After four weeks, plants were removed from the tubes, and their roots were rinsed with tap water, cleared, stained, and analysed for AMF colonization following the procedure previously outlined.

Spore extraction

AMF spores of rhizosphere sand were separated by the wet sieving and decanting procedure [41]. Soil was suspended in water, stirred and poured through stacked sieves with 300, 150 and 50 μm mesh size. Materials retained on the sieves were transferred to Petri dishes and observed under the Wild dissecting microscope at magnifications up to 40 × . The spores were manually collected using capillary pipettes for further morphological and molecular analyses.

Morphological spore identification and spore abundance

The spores, collected from three samples of rhizosphere sand (100 g each), were observed under the dissecting microscope and separated on the basis of their morphological traits. The spores were then mounted on slides in polyvinyl-lactic acid-glycerine (PLVG) [42], and their number and types recorded. Only intact, healthy spores were counted. The slides were systematically examined under the Polyvar compound microscope, equipped with Nomarski differential interference contact optics, at magnification up to 400 × , in order to count all AMF spores and separate the morphologically distinct types. The measure of spore diameter and capture of images were performed using the Leica DFC 450C color video camera and the software package LAS X 5.2.0261303 (Life Science Microscope Software Platform, Leica Microsystems srl, Italy) connected to the microscope. Qualitative and quantitative spore traits were determined on at least 50 spores for each morphotype, when possible. With the aim of detecting auxiliary cells production, Gigasporaceae spores were germinated and checked as described in Turrini et al. (2008) [43]. Images were acquired with a Reichert 100 × magnification oil immersion lens. The spores were identified using original species descriptions, updated taxonomic studies on the described taxa, available identification manuals and keys [44–46]. Other information was retrieved in the sites: https://invam.ku.edu/; http://www.amf-phylogeny.com/.

Molecular identification of AMF spores

Intact, healthy spores of the different morphotypes were manually collected with a capillary pipette under the dissecting microscope and surface decontaminated by sonication (two times, 1 min each) in a B-1210 cleaner (Branson Ultrasonics, Soest, The Netherlands), antibiotic treatment (streptomycin 400 µg/ml and chloramine T 2%) [47], followed by five washings in sterile deionized distilled water (SDW). Spores (1–15) were transferred into Eppendorf PCR tubes, crushed with a glass pestle, and their DNA was extracted using MasterPureTM Yeast DNA Purification kit (Epicentre, Madison, USA) according to manufacturer’s instructions, and the ribosomal region comprising the SSU-ITS-LSU fragment was amplified using a nested PCR protocol described by Krüger et al. (2009) [48]. The primer pairs SSUmAf1-LSUmAr2 and SSUmAf1-LSUmAr3 were used for the amplification of gigasporoid and glomoid morphotypes, respectively [48]. The first amplification was carried out in 25 μl reaction mix composed of: 1–5 ng genomic DNA, 2.5 μl Ex Taq buffer 10X, 0.2 mM each dNTP mixture, 0.5 μM of each primer (SSUmAf1-LSUmAr2/Ar3), and 1.25 U of TaKaRa Ex Taq DNA Polymerase (Takara Bio Inc., Kusatsu, Japan). Thermocycling conditions used were a hot start initial denaturation at 94 °C for 1 min, followed by 35 cycles of denaturation at 95 °C for 30 s, primer annealing at 60 °C for 30 s, extension at 72 ◦C for 90 s, and a final extension at 72 °C for 5 min. The nested PCR reactions were performed by diluting (1:100) the first PCR amplicons and using 1 μl of dilutions as template for the second reaction in a final volume of 25 μl. The primer pairs SSUmCf3-LSUmBr3 and SSUmCf3-LSUmBr1 [48] (0.5 μM) for the amplification of gigasporoid and glomoid morphotypes, respectively, were added to the PCR mix. Taq DNA polymerase, dNTPs, buffer concentrations were the same as those described above. Amplification conditions for both primer sets were as follows: 94 °C for 1 min, 35 cycles at 94 °C for 30 s, 63 °C for 45 s, 72 °C for 90 s, and a final extension step at 72 °C for 5 min.

PCR amplicons were visualized in a GelRed® (Biotium, USA) stained 1% agarose gel and samples were run at 80 V for 60 min. 1500 bp long amplicons were selected for downstream purification and concentration. Amplicons were purified by QIAquick® PCR Purification Kit (Qiagen GmbH, Hilden, Germany), according to the manufacturer’s instructions with a final elution volume of 50 μl, and purified products (2 μl) were quantified by a BioPhotometer (Eppendorf SE, Hamburg, Germany).

Purified products were cloned into pGem®-T Easy vector (Promega, Milan, Italy) according to the manufacturer’s instructions and XL10-Gold® Ultracompetent Escherichia coli cells (Stratagene, La Jolla, CA, USA) were used for transformation. Putative positive clones were screened by standard SP6/T7 colony PCR, followed by a nested PCR using the SSUmCf3-LSUmBr3 and SSUmCf3-LSUmBr1 primer pairs. Concentration of PCR mix components and PCR conditions were the same as those described above for the nested PCR reactions. Clones for the two gigasporoid morphotypes were purified by Wizard® Plus SV Minipreps (Promega) and the recombinant plasmids were sequenced at GATC Biotech (Köln, Germany) using SP6/T7 vector primers under standard conditions. As the glomoid morphotype was heterogeneous, possibly entailing different species, the libraries were RFLP screened by HinfI and MboI, and the relevant plasmids were sequenced as described above. Sequences were edited in MEGA11 software v11.0.13 [49] and their similarities determined using the Basic Local Alignment Search Tool (BLASTn) provided by NCBI. Then, they were aligned with those corresponding to the closest matches from GenBank as well as with some sequences from Glomeromycotina, covering the entire SSU-ITS-LSU region or a portion of it (ITS region for Gigasporaceae sequences), using MUSCLE [50, 51] as implemented in Mega11. The phylogenetic tree was inferred in MEGA11 using the Neighbor-Joining algorithm, based on the Maximum Composite Likelihood method [52] with Acaulospora denticulata (AJ239115) and Claroideoglomus claroideum (FR750058) inserted as an outgroup for Gigasporaceae and Glomeraceae sequences, respectively. The confidence of branching was assessed using 1000 bootstrap resamplings.

Sequences of the subphylum Glomeromycotina reported in this study were deposited in the GenBank database with accession numbers PQ843026-PQ843045 and PQ845095-PQ845123.

Metabarcoding of spore associated bacteria

Three replicates of 100 intact and healthy spores of the two most abundant AMF species, identified in the present work as Racocetra persica and Racocetra fulgida, were selected under the dissecting microscope. Each 100 spores replicate was suspended in 1 ml sterile physiological solution (9 g l⁻¹ NaCl) in a 1.5 ml Eppendorf tube, and vigorously washed 15 times in succession, using a vortex mixer at 1500 rpm for 1 min, following the method described in Agnolucci et al. (2015) [53]. Using such a method, only bacteria strictly associated to AMF spores—in the spore cytoplasm and/or within spore wall layers—were retrieved. After spore homogenization using a glass pestle, bacterial DNA was extracted utilizing the DNeasy® PowerSoil® Pro Kit (Qiagen), according to the manufacturer’s instructions and eluted in 50 μl 10 mM Tris buffer. The extracted DNA was checked for successful amplification and sent to BMR Genomics (Padua, Italy) for Illumina MiSeq sequencing. The hypervariable region V3-V4 of the 16S rRNA gene was amplified using the Pro341F and Pro805R primers [54]. The Platinum™ Taq DNA Polymerase High Fidelity (Thermo Fisher Scientific, Waltham, MA, USA) was used in the PCR mix. Thermocycling conditions were as follows: 94 °C for 1 min, 25 cycles at 94 °C for 30 s, 55 °C for 30 s, 68 °C for 45 s, and a final extension step at 68 °C for 7 min. Agencourt AMPure XP (LABPLAN; Naas, Ireland) was used for libraries purification according to the Illumina metagenomic sequencing library protocol. Dual indices and Illumina sequencing adapters from the Illumina Nextera XT index kits v2 B and C (Illumina, San Diego, USA) were added to the target amplicons in a second index PCR step to generate sequencing index libraries. The Illumina® MiSeq™ platform was used for producing 2 × 300 bp paired-end runs. The NGS procedure was performed by BMR Genomics. Libraries were processed using the microbial ecological tool QIIME2 [55] version 2021.4.0 pipeline for trimming, using Cutadapt v.10 [56], denoising, chimeras’ removal, dereplication and ASVs construction using DADA2 tool [57]. For analyses focusing on AMF spore endosymbionts, the resulting sequences were additionally clustered into operational taxonomic units (OTUs) at a 94% sequence similarity threshold. Taxonomic assignment of bacterial ASVs was done against the Silva [58] database version 138.

Illumina MiSeq libraries were submitted to the NCBI Sequence Read Archive under the BioProject accession number PRJNA1208279, containing the following Biosamples: SAMN46172275-80.

Phylogenetic analyses were carried out on sequences related to AMF spore-associated bacteria, using MEGA11, as described above for Glomeromycotina sequences.

Statistics of bacterial communities

Bacterial alpha-diversity was calculated using PAleontological STatistics (PAST) software v4.13 [59] taking ASV as a proxy of species Richness (S), Simpson dominance index (D), Shannon index (H_s) and Evenness (J_p). In addition, we used the Hill numbers of order 1 (the exponential of Shannon diversity) and 2 (the reciprocal of Simpson dominance), which are the number of equally abundant species necessary to produce the observed value of the relevant diversity index [60]. Data related to diversity indices were subjected to one-way ANOVA after checking the homogeneity of variances, with the two AMF spore morphotypes as the source of variation. Analyses were carried out using PAST software. Rarefaction curves based on the total number of bacterial communities (ASV) from R. fulgida and R. persica spores were calculated by 1000 simulated datasets, using re-sampling with replacement in PAST software. Such analysis was performed to estimate whether the number of screened sequences was sufficient to capture bacterial diversity occurring in the different AMF spore samples (Fig. S2).

Indicator species analysis [61] was carried out on ASV sequences from each spore type (R. persica and R. fulgida), whose occurrence, absence or abundance indicate possible environmental impact of the variables. The analysis was assessed by the PAST software v4.13. The indicator value (IndVal) pointed out the strength of association, while the statistical significance of association was tested using a permutation test (permutation number = 999) at P ≤ 0.05. Moreover, percent similarity (SIMPER) [62] analysis with Bray–Curtis dissimilarity metric [63], was used to examine which taxa contributed to the differences between non-endosymbiotic spore bacterial communities associated to each AMF species (R. persica and R. fulgida) and calculated based on ASV relative abundances > 1%. Spore associated bacterial communities shared between R. persica and R. fulgida were defined by Venn diagrams, drawn using BioVenn at https://www.biovenn.nl/index.php.

Results

Percentage of mycorrhizal root length in Ammophila arenaria roots and inoculum potential of rhizosphere sand

Percentage of mycorrhizal root length in A. arenaria was 28 ± 10 (mean ± SEM, Standard Error of the Mean). Typical fungal structures were detected, such as intraradical hyphae, arbuscules and vesicles and auxiliary cells (Fig. S3). The activity of AMF propagules occurring in the rhizosphere sand, evaluated by MIP bioassay, did not differ among the three root samples, as the percentage of colonized root length of the test plant C. intybus ranged from 15 to 21%. Beyond intraradical arbuscules, hyphae and vesicles, many auxiliary cells occurred on extraradical hyphae, confirming the presence of Gigasporaceae (Fig. 1a-e). Interestingly, the production of gigasporoid spores at different stages of development was detected, as shown in Fig. 1f-h.

Fig. 1.

Light micrographs of arbuscular mycorrhizal structures occurring in the roots of Cichorium intybus plants growing in sandy soil from the study site and of Racocetra persica and Racocetra fulgida spores. a Intraradical hyphae and vesicles. Scale bar: 360 µm; b arbuscules. Scale bar: 36 µm; c, d, e knobby auxiliary cells, produced on coiled hyphae, indicating the presence of AMF belonging to the Gigasporaceae family. Scale bar: 22.5 µm (c, d); 60 µm (e); f, g, h sequential production of gigasporoid spores at different stages of development. Scale bar: 45 µm. i R. persica whole spore showing the bulbous sporogenous cell. Scale bar: 40 µm; j, k Details of R. persica spore wall and bulbous sporogenous cells. Scale bars:16, and 40 µm, respectively. l R. fulgida whole spore showing the bulbous sporogenous cells. Scale bar: 37 µm, respectively; m, n Details of R. fulgida spore wall and bulbous sporogenous cell. Scale bar:12.5 and 44 µm, respectively

AMF spores in Ammophila arenaria rhizosphere

Morphological identification and abundance Mean number of AMF spores was 77 ± 17 (mean ± SEM) in 100 g sand. Different AMF morphotypes were retrieved in the rhizosphere sand of A. arenaria. The most abundant morphotype (43 ± 9 spores in 100 g sand) showed singly, dark orange-dark copper globose/subglobose spores, with a diameter of 311 ± 41 × 329 ± 44 μm, a bulbous sporogenous cell 58 ± 9 μm wide, producing pale yellow auxiliary cells in aggregates of 5–10, with tuberculate surfaces, identified as Racocetra persica (Fig. 1i–k). The second morphotype (15 ± 5 spores in 100 g sand, showed singly, pale cream subglobose spores, with a diameter of 217 ± 25 × 245 ± 33 μm and a bulbous sporogenous cell 38.0 ± 6.0 μm wide, producing pale yellow auxiliary cells in aggregates of 5–10, with tuberculate surfaces, identified as Racocetra fulgida (Fig. 1l–n). Other morphotypes were represented by glomoid spores (19 ± 11 spores in 100 g sand) whose traits were heterogeneous and not useful to allow a preliminary morphological description.

Molecular identification The DNA extracted from the spores was successfully amplified, obtaining a fragment of the expected size (ca. 1500 bp). The use of DNA cloning and sequencing allowed the identification of AMF spores to the species level. The sequences obtained from the gigasporoid morphotypes were identified as R. persica and R. fulgida after BLASTn and phylogenetic analyses (Fig. 2a). Interestingly, ITS sequences revealed two clusters (boostrap value 85 and 94) containing R. persica and R. fulgida sequences (accessions: AM503939/AM503940/AM503941 and accessions: AM503935/AM503936/AM503937/AM503938, respectively) previously isolated from the same Biosphere Reserve [43] with more than 99% of identity values.

Fig. 2.

Neighbour-Joining (NJ) phylogenetic trees of Glomeromycotina sequences. a NJ phylogenetic analysis focusing on Gigasporaceae obtained from the alignment of ITS sequences. Bootstrap values are shown when they exceed 50% (1000 replications). A sequence of Acaulospora denticulata was used as outgroup. b NJ phylogenetic analysis focusing on Glomeraceae, based on the alignment of partial nuclear SSU-ITS-LSU rRNA region (~1500 bp) sequences. Bootstrap values are shown when they exceed 50% (1000 replications). A sequence of Claroideoglomus claroideum was used as outgroup. Sequences obtained in this work are in bold and accession numbers are indicated

The sequences obtained from the glomoid spores allowed the identification of Septoglomus jasnowskae, Kamienskia bistrata (formerly Glomus bistratum) and Microkamienskia perpusilla (formerly Glomus perpusillum) (accessions: 99.06% identity with KF060323, 99.22% with KJ564137, 99.61% with KJ564142, respectively) and an unidentified species of Septoglomus sp., whose sequences formed a separate cluster supported by 97 boostrap value in the phylogenetic tree (Fig. 2b).

Diversity and community composition of spore associated bacteria as revealed by Illumina MiSeq sequencing

The two most abundant AMF species, R. persica and R. fulgida, were selected for metabarcoding analyses of spore associated bacteria. Illumina sequencing produced a total of 130,490 reads relative to the V3-V4 region of 16S rRNA gene, 71,956 from R. persica and 58,534 from R. fulgida, which grouped in 277 ASVs (Table S2). In particular, 107,616 reads were assigned to 26 ASVs matching with sequences of described AMF endosymbionts species, while 22,874 reads (251 ASVs) were taxonomically ascribed to the other spore-associated bacteria. The highest number of ASVs was retrieved from R. persica, 227, while in R. fulgida only 69 ASVs occurred, with 19 ASVs shared, as reported in Venn diagram in Fig. S4. Interestingly, bacterial endosymbionts did not share any ASV.

Overall, the numbers of bacterial phyla found in R. persica and R. fulgida were 15 and 10, respectively, and the most represented phylum was Mycoplasmatota, comprising Candidatus Moeniiplasma glomeromycotorum endosymbionts (CaMg), with mean relative abundance of 46.6% and 90.1% of total reads in R. persica and R. fulgida, respectively (Fig. 3). In R. persica two additional phyla were more abundant, Pseudomonadota (39.8%), mostly represented by the endosymbiont Candidatus Glomeribacter gigasporarum (CaGg), 29.7%, and Actinomycetota, 10.4%, while other phyla were represented with relative abundance, lower than 3%. In R. fulgida, beyond Mycoplasmatota, other phyla, including Actinomycetota (5.5%) and Pseudomonadota (3.4%), occurred in low abundance (Fig. 3).

Fig. 3.

Distribution of bacterial phyla retrieved in Racocetra persica and Racocetra fulgida spores

Alpha diversity analyses (Fig. 4, Table S1), based on ASVs, revealed that both Richness (S) and Shannon (H_s) indices were significantly higher in R. persica, compared with R. fulgida (P = 0.014 and P = 0.032, respectively). In particular, bacterial Richness was 99.33 ± 15.62 in R. persica and 30.00 ± 5.29 in R. fulgida, while H_s values were 2.36 ± 0.24 and 1.50 ± 0.11, in R. persica and R. fulgida, respectively. A similar trend was observed for Hill₁, which also differed significantly (P = 0.050), with values of 11.18 ± 2.34 and 4.54 ± 0.54 for R. persica and R. fulgida, respectively.

Fig. 4.

Alpha diversity of bacterial communities associated with Racocetra persica and Racocetra fulgida spores, showing Richness (S), Simpson dominance index (D), Shannon index (Hs), Evenness (Jp), Hill numbers of order1 (H₁) and 2 (H₂). In each boxplot, mean ( ×) and median (black line) are shown. The asterisks indicate significant differences (P < 0.05) between the two Racocetra species

Endosymbionts The ASVs assigned to both CaMg (25 ASVs) and CaGg (1 ASV) represented about 76.3% of all sequences retrieved from R. persica, and more than 90% from R. fulgida. R. persica hosted a large diversity of CaMg endosymbionts, 20 ASVs (33,499 reads), and only 1 ASV (21,376 reads) of CaGg. By contrast, R. fulgida hosted a lower number of highly abundant (52,741 reads) CaMg (5 ASVs). For the first time CaMg sequences were retrieved in R. fulgida spores, while no CaGg sequences were detected. Interestingly, the NJ phylogenetic analysis showed that the unique sequence of CaGg (ASV173) from R. persica clustered with CaGg sequences from R. persica isolated from the same dune system (accessions: AJ251634-35, 99.8% identity) [30] (Fig. S5).

The 20 CaMg ASVs detected in R. persica were grouped into highly heterogeneous phylotypes (13 OTUs, Fig. 5), two of which (OTU5 and OTU14) were shared with R. fulgida and may represent a core set of endosymbiotic sequences within Racocetra species in the sand dune ecosystem. Among CaMg OTUs, only five of them, including a variable number of highly similar ASVs (with a similarity percentage ≥ 94%), generally exhibited a high average number of reads per ASV, with values greater than 900. In particular, in R. persica ASV70 was the most abundant sequence type retrieved in the OTU1 with 11,417 total reads, followed by ASV103 (OTU14), ASV191 (OTU2) and ASV116 (OTU1), which showed a total number of reads of 7,379, 4,693 and 3,451, respectively.

Fig. 5.

Neighbour-joining phylogenetic analysis of Candidatus Moeniiplasma glomeromicotorum (CaMg) sequences based on the alignment of the V3–V4 region of the 16S rRNA gene. Bootstrap values are shown for branches with support above 50% (based on 1000 replicates). Sequences obtained in this study are highlighted in bold red and light blue for Racocetra persica and Racocetra fulgida spores, respectively. Red stars indicate sequences associated with indicator species. Different OTUs, defined by a similarity percentage ≥ 94% among ASVs, are indicated in brackets

By contrast, in R. fulgida fewer ASVs were detected, originating four OTUs, which were highly abundant and consistently present across the three replicates. In particular, OTU14 comprising the ASV250, showed a very high number of reads (38,544). Interestingly, OTU4 (ASV19) and OTU15 were exclusive to R. fulgida.

The analysis of indicator species based on ASVs (Fig. 6) showed that ASV173, identified as CaGg, was the only indicator species of R. persica (InVal = 100, P < 0.05), whereas ASV22, 137, 214 and 250, identified as CaMg, were the indicator species for R. fulgida (InVal = 100, P < 0.05). ASV137 and ASV250 were very similar and grouped into OTU15 (Fig. 5). Interestingly, no shared endosymbiont ASVs between the two fungal species were found. Specifically, 21 ASVs were exclusive to R. persica, while the remaining five ASVs (ASV19, 22, 137, 214, and 250) were exclusive to R. fulgida (Fig. 6a).

Fig. 6.

Significant indicator species of a endosymbionts and b non-endosymbiotic spore-associated bacterial communities associated with Racocetra persica and Racocetra fulgida spores. The scale indicates the IndVal values. The boxed IndVal values are significant at P < 0.05

Other spore-associated bacteria Overall, the number of ASVs were 206 (17,081 reads) and 64 (5,793 reads), in R. persica and R. fulgida, respectively, with 19 shared ASVs. Among the shared ASVs, four (ASV24, ASV141, ASV230, ASV273), were ascribed to genera of Actinomycetota (Lentzea and Streptomyces) and two ASVs (ASV119 and ASV277) to the genus Dongia (class of alphaproteobacteria), generally at high frequencies in the two AMF species. It is interestingly to note the occurrence of taxa related to Hyphomicrobiales (Pedomicrobium and Pseudorhodoplanes), Cytophagales (Microscillaceae) and Pseudomonadales, with the genus Pseudomonas (ASV260), in both AMF spore types. Such ASVs can represent a core set of bacteria, presumably functionally important for both AMF species. The majority of ASVs displayed an erratic distribution across the different replicate samples, with a low number of sequence reads, while some ASVs were uniformly distributed and differentially occurred in the two AMF species. For example, the ASVs corresponding to species belonging to the genera Rhizobium (ASV195), Bradyrhizobium (ASV183), Dongia (ASV119), Kibdelosporangium (ASV17), were significantly more represented in R. persica and were identified as R. persica indicator species (IndVal values = 100, 100, 93.77, 100 respectively, P < 0.05, Fig. 6b). On the other hand, the ASV228, identified as Amycolatopsis sp. (99.76% identity with the sequence LR584381), represented the unique indicator species of R. fulgida (IndVal = 100, P < 0.05, Fig. 6b). This sequence was the main contributor to the total dissimilarity between the two Racocetra species, as evidenced by the SIMPER test (Table 1), representing a cumulative contribution of 18.73%. Interestingly, other four Actinomycetota sequences closely matching Streptomyces spp. (ASV273, ASV129, ASV141, ASV46) mainly contributed to the dissimilarity between R. persica and R. fulgida, differentially occurring in the two species (Table 1).

Table 1.

Percent similarity (SIMPER) test showing the taxa contributing to the differences between non-endosymbiotic spore-associated bacterial communities detected in the two Racocetra species and calculated based on ASV relative abundances > 1%

Taxon	Species	Average dissimilarity	Individual contribution (%)	Cumulative contribution (%)	Mean R. persica	Mean R. fulgida
ASV228	Amycolatopsis sp.	17.970	18.730	18.730	0.000	0.245
ASV106	Rickettsiella sp.	11.880	12.380	31.110	0.156	0.000
ASV273	Streptomyces sp.	10.940	11.400	42.510	0.001	0.171
ASV129	Streptomyces sp.	6.801	7.086	49.590	0.102	0.000
ASV141	Streptomyces sp.	5.913	6.161	55.750	0.079	0.023
ASV46	Streptomyces sp.	5.004	5.214	60.970	0.070	0.000
ASV260	Pseudomonas sp.	4.647	4.842	65.810	0.006	0.069
ASV24	Lentzea sp.	4.308	4.488	70.300	0.018	0.067
ASV271	Steroidobacter sp.	2.628	2.738	73.040	0.036	0.000
ASV172	Ferruginibacter paludis	2.557	2.664	75.700	0.000	0.031
ASV265	Saccharimonadales	2.475	2.579	78.280	0.002	0.029
ASV119	Dongia sp.	2.192	2.284	80.560	0.033	0.002
ASV258	Amycolatopsis sp.	2.034	2.119	82.680	0.025	0.000
ASV107	Pseudomonas sp.	1.981	2.064	84.750	0.000	0.031
ASV227	Pseudomonas sp.	1.967	2.049	86.800	0.000	0.024
ASV78	Citrifermentans sp.	1.893	1.973	88.770	0.000	0.023
ASV17	Kibdelosporangium sp.	1.670	1.740	90.510	0.023	0.000
ASV89	Hydrogenophaga sp.	1.531	1.596	92.100	0.000	0.019
ASV192	Rhizobiaceae	1.353	1.410	93.510	0.019	0.000
ASV210	Haliangium sp.	1.310	1.365	94.880	0.000	0.016
ASV253	Pseudomonas sp.	1.114	1.161	96.040	0.000	0.017
ASV195	Rhizobium sp.	1.113	1.160	97.200	0.014	0.000
ASV183	Bradyrhizobium sp.	1.047	1.091	98.290	0.014	0.000
ASV136	Ohtaekwangia sp.	0.864	0.899	99.190	0.013	0.000
ASV108	Rhizobiaceae	0.778	0.810	100	0.011	0.000

The 251 ASVs were affiliated to 69 and 34 genera for R. persica and R. fulgida, respectively. In particular, the most represented genus was Streptomyces, with a relative abundance of about 37–38% both in R. persica and in R. fulgida (Fig. 7). In R. persica other abundant genera were also Steroidobacter (7.3%) and Dongia (5.2%), very poorly represented (0.4% and 0.2%, respectively) in R. fulgida. On the other hand, in R. fulgida other most abundant genera were Pseudomonas, Amycolatopsis and Lentzea, which accounted for 25.6%, 10.7% and 7.3%, respectively (Fig. 7).

Fig. 7.

Pie charts showing the relative abundance of the different predominant non-endosymbiotic spore-associated bacterial genera occurring in Racocetra persica and Racocetra fulgida spores

In order to better characterize the bacterial communities not known as endosymbionts, phylogenetic analyses were carried out on ASVs belonging to the most represented phyla, i.e. Actinomycetota, Pseudomonadota (classes: alpha- and gammaproteobacteria) and Bacteroidota (Figs. S6–S9). In the Actinomycetota phylogenetic tree (Fig. S6), four orders (Pseudonocardiales, Propionobacteriales, Streptomycetales and Solirubrobacterales) were detected, two of which (Pseudonocardiales and Streptomycetales), included the highest number of ASVs. Interestingly, in the Pseudonocardiales cluster, the ASV228 (Amycolatopsis sp.) and the ASV17 (Kibdelosporangium sp.), were ascribed to the indicator species found in this work (Fig. 6b). The phylogenetic analysis of the ASVs belonging to the Alphaproteobacteria class revealed the presence of eight orders, with Hyphomicrobiales containing the majority of ASVs, including those ascribed to the genera Mesorhizobium, Ensifer, Rhizobium and Bradyrhizobium, which are well known for establishing symbiotic association with plant roots (Fig. S7). In the phylogenetic analysis related to Gammaproteobacteria, the genera Steroidobacter, Pseudomonas and Lysobacter were well represented (Fig. S8), while in Bacteroidota tree Terrimonas spp., Niastella spp. and Chitinophaga spp. (Chitinophagales) and Ohtaekwangia spp. and Chryseolinea spp. (Cytophagales) occurred (Fig. S9).

Discussion

This work analysed the bacterial communities closely associated with AMF spores in a natural environment characterized by high levels of drought and salinity and low levels of nutrient availability, namely the coastal sand dunes system within a UNESCO Biosphere Reserve in Tuscany, Italy. The analysis of the AMF community in A. arenaria rhizosphere revealed the dominance of two species, R. persica and R. fulgida, that differentially recruited highly diversified and rich bacterial communities, as detected by high-throughput sequencing. The most abundant bacterial taxa were represented by the AMF endosymbionts CaMg and CaGg. Conversely, the other spore-associated bacterial taxa were less abundant but highly heterogeneous, with Actinomycetota as the most represented phylum, entailing the genera Streptomyces, Amycolatopsis and Lentzea, which could play a functional role in the fitness of fungal symbionts and host plants in the harsh sand dune ecosystem.

Occurrence of AMF in A. arenaria rhizosphere

Our data supported and complemented previous findings reporting that the two Racocetra species were the dominant components of A. arenaria rhizosphere in an in situ collection of native AMF established in the coastal sand dunes of the UNESCO Biosphere Reserve (AM503939/AM503940/AM503941) [43]. R. persica sequences grouped with those originated from the coastal dune system of Porto Caleri (Rovigo, Italy) (AJ410739/AJ410740, isolate E28) [30]. The sequences obtained from R. fulgida spores grouped with those derived from the quoted in situ collection [43].

The other species identified by cloning and sequencing were Septoglomus jasnowskae, Kamienskia bistrata and Microkamienskia perpusilla and another unidentified species of Septoglomus sp. Only Microkamienskia perpusilla (syn. Glomus perpusillum) was already found in the same Reserve, associated with A. arenaria roots [64].

Previous studies identified a number of AMF species in the rhizosphere of A. arenaria growing in European and Mediterranean maritime sand dunes, the dominant ones being represented, in addition to R. fulgida and R. persica, by Diversispora aurantia, Rhizoglomus fasciculatum, Rhizoglomus irregulare, Scutellospora dipurpurescens and Septoglomus constrictum [3, 64–69]. It is interesting to note that R. persica was the regular component of AMF communities in A. breviligulata rhizosphere, in North Atlantic sand dunes [3, 69–73]. Total spore numbers compare well with those reported from the Netherlands and Spain [66, 74], and those previously found in the same site [43].

The values of mycorrhizal colonization (20–56%) confirmed that A. arenaria is a mycotrophic species, as previously shown by studies performed in other European coastal sand dunes [39, 66, 68, 74, 75]. MIP of the test plant C. intybus (15–21%) showed that AMF propagules in A. arenaria rhizosphere sand were active, in particular those of the two Racocetra species, given the high number of auxiliary cells, characterizing Gigasporaceae mycelium [44].

Occurrence of highly rich and diversified bacterial communities within Racocetra species

High-throughput sequencing revealed that R. persica and R. fulgida host highly diverse and rich bacterial communities, with differences in community composition detected between the two species.

The dominant and most abundant bacteria were the Glomeromycotina CaMg endosymbionts, coccoid bacteria widely distributed in different AMF families, and CaGg, a rod-shaped bacterium previously reported to be restricted to members of the Gigasporaceae family [32], but recently detected in AMF species belonging to the Diversisporaceae, Paraglomeraceae and Ambisporaceae families [37].

CaMg sequences represented about 47% of all sequences retrieved from R. persica and more than 90% from R. fulgida. Interestingly, in this work CaMg sequences were detected for the first time in R. fulgida spores, while the unique CaGg sequence type occurred only in R. persica. The different abundance of the two endosymbionts sequences is consistent with recent data by Lastovetsky et al. [37, 76], who, using deep sequencing, found that CaMg sequences were highly abundant, ranging from 70 to 88% across diverse AMF spores from the field, whereas CaGg sequences were extremely rare (2%). Here, the co-occurrence of CaMg and CaGg sequences in R. persica spores had an average ratio of 2.8. Such results are consistent with data from qPCR and confocal microscope image analyses of Gigaspora margarita spore samples containing the two bacterial populations, which revealed that CaMg were 1.62–3.15 times more abundant than CaGg [31]. The co-occurrence of CaGg and CaMg sequences was consistently observed also in a survey of environmental AMF spores from agricultural grassland and dune ecosystems [37].

In this study, only one sequence type was detected belonging to CaGg, while CaMg exhibited a high level of intra-host sequence heterogeneity. Overall, no shared endosymbiont ASVs between the two fungal species were found. In particular, we found 25 ASVs related to CaMg sequences grouped into 15 OTUs at 94% sequence similarity, following the standard for this taxonomic group [77]. Only four OTUs were detected in R. fulgida (two shared with R. persica), while a higher number (13 OTUs) was found in R. persica. These data are consistent with other studies, where 15 haplotypes and three OTUs, clustered at 94% sequence similarity, were detected in individual AMF spores [78]. Recently, a higher number of CaMg OTUs (an average of 22 OTUs per single spore) was found in natural ecosystems [37]. Some of the ASVs detected in this study belong to sequences previously found in other AMF species. For instance, sequence corresponding to CaMg ASV250 from R. fulgida had already been described in Cetraspora pellucida (accession: KP77634, 100% similarity), and CaMg ASV117 was previously reported in R. verrucosa (accession: KM0658, 100% similarity). Other ASVs, however, were identified for the first time in this study. Specifically, CaMg ASVs 99, 238, and 240 (grouped in OTU3), were found in R. persica, while CaMg ASV19 was found for the first time in R. fulgida. It has been hypothesized that the extensive intra-host diversity of CaMg 16S rRNA gene sequences may be linked to a very rapid rate of mutation accumulation, resulting from the loss of DNA repair mechanisms [79, 80], as well as to potential horizontal gene transfer via recombination across highly divergent AMF hosts [78], together with the ability to harbour multiple genes horizontally acquired from their AMF host [79, 81]. Intriguingly, such high diversity may be an evolutionary adaptation aimed at outcompeting other bacteria, evolving from antagonists to conditional mutualists [37], especially in harsh environments, where different stressors can affect the fungal host, mainly when it exhibits large dimensions, like species of Gigasporaceae.

In contrast to CaMg, CaGg sequences did not exhibit intra-host diversity, confirming previous studies [37, 76, 82]. However, its occurrence only in R. persica spores is closely linked to the host genotype, as demonstrated by the very high similarity with the sequences found in R. persica spores (strains E09 and E28) collected from the same dune system more than 25 years ago (accessions: AJ251634-35, 99.8% identity, [30]).

In this work we found highly diversified spore-associated bacterial communities predominantly composed of taxa not previously reported as endosymbionts. The prevalence of low-abundance ASVs with variable occurrence among replicates suggests a complex and possibly dynamic bacterial assemblage or a transient or context-dependent association. Nevertheless, a subset of ASVs exhibited a more consistent distribution pattern and showed differential occurrence between the two AMF species, pointing to the possibility of more specific or functionally relevant associations.

Some studies suggested that the highly diverse bacterial communities associated with AMF spores were driven more by the different fungal host genotype—not only species but also isolates of the same species—than environmental parameters or host plants [53, 83, 84]. In particular, Agnolucci et al. (2015) [53] reported that AMF host genotype accounted for about 50% of total variation observed in the bacterial community profiles associated to the spores of six AMF isolates (three species) from different biomes, maintained for several generations in pot cultures with the same host plant, under the same environmental conditions. Consistent results were obtained by Pandit et al. [84], which analysed the bacterial communities associated with the spores of eight isolates from in vitro transformed AMF root organ co-cultures. The diverse community assembly observed in this work for the two Racocetra species might be related both to metabolic activities of the spores, which can release variable quantity and quality of chemical compounds, and to morphological traits of the spores, such as size, shape, spore wall structure, ornamentation and chemical composition.

What is the functional significance of spore-associated bacteria?

While CaMg may be probably hosted in the AMF spore cytoplasm as the result of a co-evolutionary strategy, allowing them to multiply in a protected niche and outcompete other bacteria, CaGg can influence the AMF-plant symbiosis at the molecular level, by regulating the fungal cell cycle, modulating fungal proteome and lipid profile, while enhancing fungal sporulation and ATP production and triggering mechanisms to detoxify reactive oxygen species [35, 36]. In particular, endosymbionts can stimulate pre-symbiotic growth of AMF by enhancing mycelial elongation and branching in response to root exudates and inducing stress-responsive proteins [34, 35]. Recent studies using metatranscriptomics and metabolomics techniques provided new insights on the role of endosymbionts influencing their fungal host during both pre- and post-stages of mycorrhizal symbiosis [36, 85, 86]. However, their possible role as MHB and/or PGPB remains to be further investigated in the years to come.

The highly diversified AMF spore-associated bacterial taxa found in this work encompass species known to possess key traits providing fitness benefits to both fungal and plant host. It is remarkable that the most relatively abundant bacteria belonged to Actinomycetota, containing members reported as resistant to the dominant physicochemical conditions in arid areas—desiccation, high temperature, salinity—such as the genera Amycolatopsis, Lentzea and Streptomyces [87]. The shared taxon consistently found in both Racocetra species, ASV24, was affiliated to actinobacterial species of the genus Lentzea, such as L. atacamensis, isolated from hyper-arid soils of the Atacama Desert in Chile, one of the most extreme environments on Earth, described as resistant to desiccation and high salinity [88, 89]. Among the taxa consistently detected in R. persica, ASV17, included among the indicator species, was ascribed to the actinobacterial genus Kibdelosporangium, encompassing species like K. aridum isolated from desert soils collected in Pima County, Arizona, USA [90], K. persicum, isolated from hot desert in Iran [91], K. banguiense isolated from the forest of Pama, in Central African Republic [92] and other root endophytic species [93, 94]. The indicator species for R. fulgida, ASV228, was a close relative of Amycolatopsis orientalis subsp. lurida and A. speibonae, two antibiotic-producing actinobacteria [95, 96], a trait which could play an important role in biocontrol activity against phytopathogens. It is interesting to note that the genus Amycolatopsis consistently occurred in G. margarita surface-decontaminated spores, suggesting a relatively stable relationship with Gigasporaceae [97, 98].

Other interesting taxa were consistently represented in R. persica, such as ASV119 (Rhodospirillaceae), affiliated to the genus Dongia, whose members were isolated from freshwater, soil and from a dry sandy soil sample collected in the Gurbantunggut Desert in Xinjiang, China, ASV42 Terrimonas sp. (Chitinophagaceae), whose members have been frequently retrieved in the root endosphere and rhizosphere [99], ASV195 and ASV183, affiliated to the N-fixers Rhizobium and Bradyrhizobium species. Such traits confirm the key multifunctional roles possibly played by spore-associated bacteria in mediating the acquisition of major plant nutrients and in plant resilience [21, 23, 53].

Our study showed that different AMF genotypes differentially recruited highly diversified and rich spore associated bacterial communities in a natural sand dune system, as revealed by high-throughput sequencing. The isolation in pure culture of such bacteria could boost in-depth studies on the complex interactions between AMF, associated bacteria and host plants, aimed at selecting effective individual bacterial strains and synthetic communities, to be exploited in the restoration of degraded sand dune systems and in sustainable food production in dry and harsh agroecosystems.

Author contributions

AG, MG, AT and MA conceived the study. AG, IP, GC, LA and AT performed experiments and analyses. AG, MG, AT drafted the manuscript. AG, MG, AT, LA, and MA contributed to the manuscript. This paper is part of AG doctoral thesis work.

Funding

The research has received funding from PhD Programme (Agriculture, Food and Environment) and Fondi di Ateneo of the University of Pisa.

Data availability

Illumina sequencing data have been submitted to the NCBI Sequence Read Archive under the BioProject accession number PRJNA1208279, containing the following Biosamples: SAMN46172275-80. Glomeromycotina sequences were deposited in the GenBank database with accession numbers PQ843026-PQ843045 and PQ845095-PQ845123.

Declarations

Conflict of interest

The authors declare no competing interests.