The soil for the greenhouse experiment was collected from an approximately 25-year-old site of a coal mining spoil bank wherein the transition of vegetation from the dominance of AM hosts to EcM hosts was ongoing (50°14′32″ N, 12°40′30′′ E, northwestern Bohemia, Czech Republic). Environmental characteristics of this site were previously described by Rydlová et al. (2016) and Knoblochova et al. (2017). Briefly, during early succession this site was colonized by AM-forming grasses and herbs, mainly Calamagrostis epigejos, H. caespitosum, Tussilago farfara, Daucus carota, and Centaurea stoebe. At the same time, EcM trees started to appear, in particular Salix caprea and B. pendula. At the time of sampling, there was a dense understory of AM hosts but EcM trees had begun visibly to dominate the site. Soil was collected from six sampling points (to depth of 0–20 cm), homogenized (while removing large root segments and non-weathered pieces of parent substrate), air-dried, and then stored in a cold and dark place until further use (approximately 3 months).

Betula pendula and H. caespitosum were selected as model species for EcM and AM hosts, respectively, because they are abundant and naturally coexist on the 25-year-old coal mine spoil bank site. For performing the experiment, plastic rhizoboxes (18 × 9 × 16 cm) separated into three equal compartments (6 × 9 × 16 cm) by nylon mesh with 42 μm pore size were used. Hyphae but not roots were able to spread between the compartments (Janoušková et al., 2011). The substrate used for filling the two lateral compartments was a mixture (1:1 v/v) of the non-sterile spoil bank soil and autoclaved (30 min at 121°C) zeolite. The middle compartment was filled with the same substrate, but it had been sterilized by γ-irradiation (25 kGy). Seeds of H. caespitosum were collected on different sites on the spoil banks, while seeds of B. pendula were purchased from a seed producer (Lesy České Republiky, Týniště nad Orlicí, Czech Republic). All seeds were surface-sterilized with a 10% solution of sodium hypochlorite for 5 min. Seeds were germinated and seedlings pre-grown under semi-sterile conditions in transparent plastic boxes containing autoclaved river sand.

The experiment consisted of two stages (Figure 1). In Stage I (conditioning phase), 7 weeks old seedlings of AM and/or EcM hosts were transplanted into the lateral compartments (hereinafter referred to as “large plants”) to establish the following treatments: (a) lateral compartments without plants (00, the lateral rhizoboxes contained bare soil throughout the experiment), (b) only one lateral compartment occupied by B. pendula (B0) or H. caespitosum (H0), or (c) both lateral compartments planted with the same plant species (BB and HH) or with a combination of both (BH, hereinafter referred to also as the “interaction treatment”). To avoid airborne contamination, the middle compartment was covered with an aluminum foil until seeds were sown into it. Sixteen replicates were established per treatment (00, B0, H0, HH, BB, and BH), and thus there were 96 rhizoboxes in total. The plants were cultivated for 5 months in a temperature-controlled greenhouse (18–28°C) with supplemental lighting (400 W metal halide bulbs) and watered with deionized water according to need. After 5 months, soil samples were collected from the middle compartments to describe the fungal communities. Two subsamples were taken from diagonal corners of the middle compartment using a sterile laboratory spoon to the depth of about 4–5 cm. The two subsamples were pooled (ca 10 g of soil fresh weight in total), homogenized, then separated into two parts: one of them for phospholipid fatty acid (PLFA) and neutral lipid fatty acid (NLFA) analyses and the second for DNA-based characterization of the soil fungal communities (Figure 1).

For Stage II, one-half of the rhizoboxes of each treatment was planted with 20 seeds of B. pendula (EcM host) and the second half with 25 seeds of H. caespitosum (AM host). Seeds were evenly placed on the soil surface of the middle compartments and watered. To prevent desiccation, the middle compartments were covered for a week with transparent foil lids that enable gas flow and prevent mold growth. In this second part of the experiment, there were eight replicates per plant species and treatment. Seed germination was recorded for 2 weeks and did not differ between the experimental treatments (data not shown). The two most distant and healthy-looking seedlings present in each of the middle compartments were retained and extra seedlings were removed by cutting. After 4 months under the same greenhouse conditions, the seedlings from the middle compartments were harvested (Figure 1).

Analyses of NLFA and PLFA were performed for each soil sample in order to estimate the fungal biomasses. The former enables quantification of AMF biomass (Olsson, 1999; Sharma and Buyer, 2015), while the latter is a better estimator of the biomass of the other fungal groups present in the soil, including EcMF, saprophytic (SaprF), and pathogenic fungi (PathF). Soil samples were extracted using a chloroform–methanol–phosphate buffer mixture (1:2:0.8) as detailed by Frouz et al. (2016). Briefly, the extracted lipids were separated using solid-phase extraction cartridges (LiChrolut Si 60, Merck). The samples were eluted in three fractions containing neutral lipids, glycolipids, and phospholipids with 2 ml of chloroform, 6 ml of acetone, and 2 ml of methanol, respectively (Oravecz et al., 2004). The first and third fractions were subjected to mild alkaline methanolysis (Šnajdr et al., 2008). The free methyl esters of NLFA and PLFA were analyzed by gas chromatography–mass spectrometry (450-GC, 240-MS ion trap detector, Varian, Walnut Creek, CA, United States). The GC instrument was equipped with a split/splitless injector, and a DB-5MS column was used for separation (60 m, 0.25 mm i.d., 0.25 mm film thickness) according to the programs described by Frouz et al. (2016). Mass spectra were recorded at 1 scan s⁻¹ under electron impact at 70 eV, mass range 50–350 amu. Methylated fatty acids were identified according to their mass spectra and by using a mixture of chemical standards from Sigma-Aldrich (St. Louis, MO, United States) and Matreya LLC (Pleasant Gap, PA, United States). Fungal biomass in the PLFA fraction was quantified based on 18:2ω6.9 concentration (Stella et al., 2015) while the biomass of AMF was estimated using 16:1ω5 concentration in the NLFA fraction (Bååth, 2003; Frouz et al., 2016).

The composition of fungal communities was assessed in the soil samples taken at the end of the Stage I (Figure 1). DNA was extracted from 2 g of soil per sample using the PowerSoil^® DNA Isolation Kit (Mobio) according to the manufacturer’s instructions. Polymerase chain reaction (PCR) was performed using the primer pair gITS7ngs (5′-GTGARTGTGARTCATCRARTYTTTG-3′; Ihrmark et al., 2012) and ITS4 (5′-TCCTCCGCTTATTGATATGC-3′; White et al., 1990) to amplify the ITS2 region. Both forward and reverse primers were tagged by molecular identifiers containing 10–11 bases. Three separate PCR reactions were performed for each sample to reduce PCR bias. The PCR mix consisted of 2.5 μl of 10× Taq buffer, 1 μl of dNTPs mix (10 mM), 2 μl of MgCl2, 1 μl of each primer (5 μM), 0.2 μl of Taq polymerase (Thermo Scientific), 1 μl of bovine serum albumin, 2 μl of 10 times diluted DNA, and 15.8 μl of ddH2O in a total volume of 26.5 μl. The cycling conditions were 4 min at 94°C, followed by 35 cycles of 30 s at 94°C, 30 s at 49°C, and 40 s at 72°C, then a final extension of 10 min at 72°C. Technical amplicon replicates of suitable PCR products were mixed together per each sample and then purified using the Qiaquick PCR purification kit (Qiagen). The DNA concentration was then measured using Qubit (Life Technologies) and all the PCR products were equimolarly pooled and then sequenced using the Illumina MiSeq system at GATC Biotech (Cologne, Germany).

SEED pipeline v 2.1.05 (Vetrovský et al., 2018) was used for filtering and quality check of ~2,200,000 reads obtained from Illumina MiSeq. The reads were merged into paired-end sequences with at least 20 bp overlap and maximum difference 15%. All sequences shorter than 40 bp and average base quality scores lower than 38 were removed from the data set. Sequences without primers and identifiers as well as those with mismatched identifiers also were removed. The remaining sequences were sorted into samples according to the molecular identifier sequences. The fungal ITS2 was extracted using ITSx (Bengtsson-Palme et al., 2013) and the ITS2 sequences were clustered by implementing UPARSE in USEARCH on the 97% similarity level. Chimeric sequences together with singletons and doubletons were removed from the data set. From each cluster (altogether 840 OTUs), the most abundant sequence was selected for BLAST search against the NCBI GenBank (altogether 840 OTUs, Supplementary Table S1). Operational taxonomic units (OTUs) were assigned to fungal ecological guilds using FUNGuild v 1.0 (Nguyen et al., 2016) with subsequent manual corrections (Supplementary Table S1). To avoid the effect of unequal read numbers per sample in linear statistical analyses, 1,190 reads per sample were randomly subsampled in R (Kolaříková et al., 2017).

Shoot (SDW) and root dry weight (RDW) were determined in all plants by drying at 60°C to constant weight. For analyzing mycorrhizal colonization, root samples were stained with 0.05% Trypan blue in lactoglycerol (Koske and Gemma, 1989). Thirty root segments ca 1.5 cm in length were observed under a compound microscope (Olympus IX 51) at 200× magnification. AM colonization was evaluated according to Trouvelot et al. (1986), and three colonization parameters were estimated using the program “Mycocalc”:¹ frequency of mycorrhizal colonization (AM.F), abundance of arbuscules (AM.A), and abundance of vesicles (AM.V; Janoušková et al., 2011; Rydlová et al., 2016). For B. pendula, the presence of EcM structures (hyphal mantle and Hartig net) was scored in 100 microscopic fields per sample to calculate colonization frequency (EcM.F).

To test the effect of the experimental treatments on the concentrations of fungal PLFA and NLFA in soil, AM and EcM colonization rates (AM.F, AM.M, AM.A, AM.V, EcM.F), and plant growth (SDW, RDW), we used the IBM SPSS software v. 23.0 (IBM Corp.). The data sets were first checked for normality (Shapiro–Wilk test) and homogeneity of variance (Levene’s test). Non-parametric Kruskal–Wallis ANOVA followed by Nemenyi post-hoc tests were performed for variables that showed non-normal distribution even after transformation (i.e., PLFA and NFLA, numbers of sequences and OTUs per fungal guild, mycorrhizal colonization, and seedling growth). Additionally, mycorrhizal colonization rates and growth of large plants were analyzed using two-way ANOVAs followed by Holm–Sidak pairwise multiple comparison tests (Factor 1 = Treatment: 00, B0, BB, BH, H0, and HH; Factor 2 = plant species sown in the middle compartment: B. pendula and H. caespitosum). Pearson correlations were calculated to investigate the association between seedling growth (SDW, RDW) and mycorrhizal colonization.

Fungal communities were standardized by Hellinger transformation, and Bray–Curtis dissimilarity was used to construct a fungal community dissimilarity matrix. PERMANOVA using the adonis function of the “vegan” package in R was applied to address the effect of the experimental treatments on the fungal community composition, followed by pairwise PERMANOVA with 99,999 permutations to determine specific differences between treatments. Bonferroni correction was used to calculate the corrected values of p from those determined by the pairwise PERMANOVA analyses. The heatmap function in R v. 3.5.1 (R Core Team, 2018) was used for evaluating similarities between the fungal communities present in the soils of the different treatments.

Fungal biomass (as PLFA 18:2ω6.9) in soil was significantly greater in treatments with only B. pendula (BB, B0) than in treatments where H. caespitosum was present in at least one of the compartments (BH, H0, HH). The lowest values were recorded in the treatment without any plants (Figure 2A), thus indicating that most of the PLFA 18:2ω6.9 content corresponded to EcMF rather than to SaprF or PathF. On the other hand, AMF biomass (as NLFA 16:1ω5) was significantly greater in all the treatments with H. caespitosum (BH, H0, HH) than in the others (Figure 2B). These results show that the interaction between B. pendula and H. caespitosum had a more negative effect on the overall fungal biomass (mostly comprised of EcMF) than on AMF biomass.

The relative abundance of EcMF (measured as number of sequences) was significantly higher in treatments having B. pendula in at least one of the compartments (BB, B0, BH) than in treatments with only H. caespitosum (H0, HH; Figure 3A). The highest EcMF taxa richness (measured as number of OTUs) was recorded in treatments with only B. pendula (BB, B0) and the lowest in those within which the EcM host was not present (H0, HH, 00). Intermediate values of EcMF taxa richness were determined in the interaction treatment (BH; Figure 3B). On the other hand, the relative abundance and richness of AMF were significantly higher in all treatments with H. caespitosum, including the interaction treatment (H0, HH, BH), than in treatments without the AM host (B0, BB, 00). The highest relative abundance of PathF and SaprF was recorded in the treatment without any plants in the lateral compartments (00; Figure 3).

Fungal community composition differed significantly among almost all treatments (Figure 4; Supplementary Table S1), the only exception being those treatments where B. pendula was the only cultivated plant species (BB and B0). Fungal communities from BB and B0 treatments were similar and characterized by high abundance of EcMF, especially of Geopora arenicola. The BH treatment was characterized by a combination of EcMF and AMF that led to a significantly different composition as compared to the other treatments. The large relative share of AMF in samples corresponding to treatments with only H. caespitosum (H0 and HH) distinguished these treatments from the others, and the presence of SaprF differentiated the H0 treatment from HH. Samples without any cultivated plants (00) showed high relative abundance of SaprF and PathF as well as low occurrence of the mycorrhizal guilds (Figure 4). When PERMANOVA was performed separately for analyzing AMF communities in the treatments with H. caespitosum (H0, HH, and BH) and EcMF in those with B. pendula (B0, BB, and BH), the effect of the presence of the other mycorrhizal host in the cultivation system differed between the two mycorrhizal fungal guilds. There were no differences in the composition of AMF communities among any of those treatments having H. caespitosum in at least one of the lateral compartments. By contrast, the EcMF community composition was similar in both treatments with only B. pendula, but it differed significantly when this species was cocultivated with the AM host (BH treatment; Supplementary Table S1).

Growth and mycorrhizal colonization of H. caespitosum and B. pendula seedlings varied widely among the treatments. For both plant species, the largest seedlings were those growing in the treatment without plants in the lateral compartments (Figure 5). Growth and EcM colonization of B. pendula seedlings were significantly reduced in treatments where H. caespitosum was the only plant species (H0, HH) as compared to treatments with B. pendula only (B0, BB). While biomass of B. pendula seedlings was not significantly decreased in the interaction treatment (BH) as compared to treatments where B. pendula was the only cultivated plant (B0, BB; Figure 5), EcM colonization was significantly less in this treatment than in B0 (Figure 6). In treatments that included H. caespitosum in at least one of the lateral compartments (BH, 0H, HH), B. pendula seedlings were also colonized by AMF. However, arbuscules were never detected in B. pendula roots in our experiment. The AM colonization of B. pendula was significantly less in the interaction treatment (BH) as compared to HH and H0 (Figure 6).

Biomass of the H. caespitosum seedlings in treatment H0 was significantly larger than that in B0, HH, and BB, while those growing in the interaction treatment (BH) had intermediate SDW and RDW values (Figure 5). Arbuscular mycorrhizal colonization was significantly greater in all those treatments having H. caespitosum in at least one of the lateral compartments (H0, HH, BH) than in those with only B. pendula (B0, BB). Arbuscular mycorrhizal colonization of H. caespitosum in the treatment without neighboring plants (00) showed intermediate values (Figure 6).

The tendencies observed for the seedlings were consistent with the results obtained for the large plants (Supplementary Table S1). Plant biomass and EcM colonization were significantly less in B. pendula specimens from the interaction treatment (BH) than in conspecific treatments (BB, B0), while H. caespitosum was not negatively affected by the presence of the other mycorrhizal host (BH treatment).

In treatments with plants in the lateral compartments (excluding the 00 treatment), it was observed for both plant species that when the seedlings were grown together with the other mycorrhizal host, root colonization values (AM.F for H. caespitosum or EcM.F for B. pendula) were positively and significantly correlated with the seedlings’ growth (Table 1; Supplementary Table S1). By contrast, in absence of the other mycorrhizal host, growth was unaffected by mycorrhizal colonization (Table 1). In B. pendula it was also noticed that seedling growth (Table 1) and EcM.F (ρ = −0.906, p < 0.001) were significantly and negatively correlated with AM colonization values (Supplementary Table S1).

The asymmetric response of the two mycorrhizal fungal guilds to the presence of the other host (and associated fungi) was evident at the level of the fungal biomass in soil, root colonization, and community composition. It suggests that, in our selected plant species and experimental conditions, AMF were less affected by competition with EcMF than vice versa. Asymmetric response of EcMF and AMF abundance to the presence of the other mycorrhizal type has been described for other plant species combinations. For example, willows indirectly reduced AMF abundance in roots of herbaceous plants via feedbacks with leaf litter and EcMF (Becklin et al., 2012; Mudrák et al., 2016). Similarly, Knoblochova et al. (2017) determined that presence of the ectomycorrhizal host S. caprea negatively affected the abundance of AM fungi not only in soil but also in C. epigejos roots, while C. epigejos caused no suppression of EcMF. In contrast, and similarly to what was observed in our study, McHugh and Gehring (2006) established that belowground competition with AM shrubs negatively impacted the abundance of EcM in roots of pinyon pines (Pinus edulis). When different mycorrhizal guilds occur in a nutrient-limited environment they have to acquire and provide the same resources to their host plants, thus resulting in this type of asymmetric effects and antagonistic plant–plant interactions (Montesinos-Navarro et al., 2018). The outcome of the interaction may then be related to the different nutrient acquisition strategies of these mycorrhizal guilds. To use the language of Lambers et al. (2008), the “scavenging strategy” of AMF is more efficient in early successional soils while the contrasting “mining” strategy of EcMF is superior in utilizing nutrients in later successional soils, where nutrients are usually bound in organic compounds (Read, 1991; Lindahl and Tunlid, 2015; Peay, 2016). Because we studied a 20-year-old coal mine spoil bank in which organic matter was absent or very scarce (Rydlová et al., 2016), it is likely that AMF were better competitors for the available resources than were EcMF, thus outcompeting them in the interaction treatment. Our results are also consistent with the “mutualistic niche concept” developed by Peay (2016), who stated that AM hosts and fungi tend to exclude EcM plants and fungi when organic N supply is low with respect to inorganic N, as it was in the spoil bank soil.

In contrast to our results, however, preceding studies in the same successional system determined that when an AM host (C. epigejos) coexisted with an EcM host (S. caprea), the latter affected negatively the abundance and richness of AMF as well as altered their community composition, while the impact of the AM host on EcMF communities was mostly not significant (Mudrák et al., 2016; Knoblochova et al., 2017). There are several plausible explanations for this discrepancy. It can be a matter of developmental stage and/or environmental conditions (Lodge and Wentworth, 1990), because adult plants from natural ecosystems were examined in those studies while young plants and seedlings cultivated under controlled conditions were analyzed in our study. Another probable reason is that different plant species were considered in each of these works. Detrimental effects of EcM plants and associated root fungi on AM hosts and AMF communities have been largely proven for Salix (Becklin and Galen, 2009; Becklin et al., 2012; Mudrák and Frouz, 2012; Mudrák et al., 2016; Knoblochova et al., 2017), but the opposite situation has been reported for P. edulis (Haskins and Gehring, 2004; McHugh and Gehring, 2006). This is the first time that the interaction between EcM and AM was studied using B. pendula as the EcM model plant. A factor that cannot be neglected is the allelopathy associated with some plant species but not with others. For instance, it is known that S. caprea produces allelopathic compounds (Mudrák and Frouz, 2012), and, despite that their effects on AMF are not well known, it has been hypothesized several times that they negatively influence the abundance and diversity of these fungi in soil (Becklin and Galen, 2009; Becklin et al., 2012; Mudrák and Frouz, 2012; Mudrák et al., 2016; Knoblochova et al., 2017). In contrast, allelopathic effects of Betula on soil fungi have not been described (Michelsen et al., 1995; Huang et al., 2016).

For seedlings of both plant species, mycorrhizal colonization was lowest when the other mycorrhizal host was the only plant in the lateral compartments. This is probably a consequence of missing MNs and the necessity to establish the colonization from spores remaining in the soil of the lateral compartments. Interestingly, the need to mention that mycorrhizal colonization from spores was also relevant for seedlings growing without any neighboring plants (treatment 00). In this case, however, mycorrhizal colonization of H. caespitosum and B. pendula seedlings was as great as in the treatments with conspecific neighboring plants. This suggests that the presence of MNs formed by fungi of a certain mycorrhizal type suppresses development of the other mycorrhizal type, and, interestingly, this effect was symmetric between the two fungal groups in our experiment. This is in agreement with other studies that also have documented how different plant species influence the abundance of mycorrhizal fungi in the soil and roots of neighboring plants through the production of inhibitory compounds or competitive interactions (Nilsson et al., 1993; McHugh and Gehring, 2006), which probably are mediated by associated mycorrhizal networks. Janos et al. (2013) even demonstrated that AM networks can be actively antagonistic to potential EcM hosts.

It is noteworthy that B. pendula plants were colonized by AM fungal hyphae and vesicles only when they coexisted with H. caespitosum and had contact with MNs radiating from the AM host, but not from resting propagules in the soil. This phenomenon has been described previously for other plant species that usually do not form AM (Püschel et al., 2007a; Janoušková et al., 2011; Cosme et al., 2018; Teste et al., 2020). Cosme et al. (2018) proposed to classify as “rudimentary AM phenotypes” those plant species that suppress or have lost their ability to form prominent AM phenotypes but under specific circumstances can harbor some symbiotic structures in their roots. These plant species might therefore have sufficient genetic tools to activate components of the symbiotic behavior of AMF (Teste et al., 2020). B. pendula could be included into this group of rudimentary AM plants. It is also interesting that, despite the comparable abundance and richness of AMF in all the treatments with H. caespitosum in at least one of the lateral compartments, AM colonization in B. pendula seedlings from the interaction treatment was significantly less than in treatments with only H. caespitosum. This showed that presence of the ectomycorrhizal host increased the abundance of EcMF in the soil as well as EcM formation, whose mantle could have restricted the entrance of AMF (Lodge and Wentworth, 1990). Exclusion of AM colonization in usually pure EcM plants may reflect fungal competition rather than plant control inasmuch as EcMF may outcompete AMF using mechanisms, such as mycelial overgrowth or by colonizing roots prior to other competitors (Teste et al., 2020). Another particularity observed in the soil of treatments including B. pendula was the dominance of the EcMF G. arenicola. Geopora species have been mainly found in Pinus spp. (Flores-Rentería et al., 2014 and references within; Shemesh et al., 2019), and occasionally in other tree species (Southworth and Frank, 2011; Long et al., 2016). As far as we know this is the first time it has been registered in association with B. pendula. These EcMF species are considered as stress tolerant, since their relative abundance in plant roots increases significantly with drought (Flores-Rentería et al., 2014 and references within), and they can dominate the EcMF spore banks above mountain treelines (Shemesh et al., 2019). The fact that G. arenicola was dominant in spoil banks soils agrees with the high tolerance of these species to harsh environmental conditions.

Most of the experiment’s treatments were characterized by a particular assembly of fungi and, in addition to the observed differences regarding mycorrhizal guilds, we also identified shifts in the relative abundance of saprotrophic fungi. In general, relative abundance of saprotrophs was particularly low in treatments with B. pendula. This is in agreement with Bödeker et al. (2016), who demonstrated that even though SaprF and EcMF may have overlapping fundamental niches in forests, the latter can restrict saprotrophs mainly through competitive interactions. As these fungal guilds occupy similar vertical positions in the soil profile and target the same litter and soil organic matter substrates, SaprF are usually displaced by EcMF (the phenomenon being known as the “Gadgil effect”). This competition for resources between SaprF and EcMF may be strongest when soils are poorly developed (Fernandez and Kennedy, 2015), as was the soil considered in this study.

Asymmetric responses between different mycorrhizal guilds naturally occur in nature and seem to be independent of plant habit, been possible that EcM tree species suppress herbaceous or shrubby AM species (Becklin et al., 2012; Mudrák et al., 2016; Knoblochova et al., 2017) and vice versa (Haskins and Gehring, 2004; McHugh and Gehring, 2006; this study). Based on the presented information, it becomes evident that the interaction of different mycorrhizal hosts on their fungal partners is context dependent and reliant on multiple biotic and abiotic factors, such as soil properties, the mycorrhizal hosts involved, and neighboring plant species (Moora and Zobel, 1996; Hubert and Gehring, 2008; Gorzelak et al., 2015; Knoblochova et al., 2017). This highlights the importance of studying the interaction between EcM and AM hosts and their fungal partners using different model plants and experimental approaches, including field analyses and cultivation experiments.

For both plant species, the largest seedlings were present in the treatment with no neighboring large plant (treatment 00), which corresponds to the early successional life strategy of the studied plant species and their ability to grow rapidly under weak competition. Furthermore, the seedlings may have had access to higher nutrient levels, as the MNs radiating from the large plants previously established in the lateral compartments may have caused nutrient pre-depletion of the soils in the other treatments (Janoušková et al., 2011). In treatments having at least one plant in the lateral compartments the response of the seedlings depended upon the plant species.

The growth of B. pendula was significantly reduced when H. caespitosum was the only neighboring plant as compared to specimens that were growing together with conspecifics (treatments BB, B0, and BH). Negative effects of AM on EcM hosts have been described previously. For example, Jones et al. (1998) reported that Eucalyptus coccifera (dual host) seedlings were smaller and had lower P content when they were colonized by AMF. A question that arises from these observations is whether the negative effects exerted by H. caespitosum and its AMF on B. pendula seedlings were indirect or direct. Regarding the former case, it is conceivable that H. caespitosum and its AMF produced higher nutrient depletion in the middle compartment and/or were better competitors for resources than were B. pendula seedlings. The fact that B. pendula seedlings were not significantly affected by H. caespitosum in the interaction treatment (BH) can be attributed to the greater abundance of EcMF in this treatment. This may counteract the adverse effects exerted by the AM host and associated AMF. It is also possible that AM root colonization reduced EcM colonization and consequently the benefits obtained from this symbiosis. On the other hand, AMF may also have acted directly as “parasitic fungi” (Johnson et al., 1997; Hoeksema et al., 2010). When different plant species connect to existing MNs, a strong asymmetry in the terms of trade may occur (Walder et al., 2012). In our case, it might be that when B. pendula seedlings connected to the existing AM networks, they deprived them of significant amounts of carbon and gave little (if any) profit in return (Cosme et al., 2018 and references within), thus directly and negatively affecting the EcM host.

Differences in seedlings growth were also observed for H. caespitosum, but the beneficial effects of the conspecific large plants were not as clear as for B. pendula. This was probably a consequence of the interaction between different factors conditioning the H. caespitosum seedlings’ development, such as AMF inocula and root colonization (significantly lower in treatments with only B. pendula in the lateral compartments), nutrient depletion, and availability of MNs. This is in agreement with Püschel et al. (2007b), who suggested that the general growth response of plants to mycorrhizae is probably determined by multiple factors, mainly by soil nutrient content, other plants connected to the MNs, and sharing the costs of network maintenance. For instance, it may have happened that large plants of H. caespitosum produced greater nutrient depletion than did B. pendula, so that seedlings growing in the treatment with the AM host in both lateral compartments were smaller than those which had this species in only one lateral compartment (H0 and BH). It is also possible that seedlings that become interconnected into AM MNs do not obtain significant amounts of nutrients from the fungus because most of the resources are directed to the dominant carbon source, that is, the larger plants (Moora and Zobel, 1996; Janoušková et al., 2011).

We also observed that when any of the studied species was growing together with the other mycorrhizal host, mycorrhizal colonization was lower but nevertheless positively correlated with the seedlings’ growth. Positive correlations between AM or EcM colonization and plant growth have been widely reported (e.g., Thomson et al., 1994; Treseder, 2013; Cheng et al., 2020), including under harsh conditions (e.g., Hachani et al., 2020; Klinsukon et al., 2021) or when competing with other plant species (e.g., Lin et al., 2015; Zhou et al., 2018). Our results suggest that under the pressure of interspecific competition, EcM and AM may improve the performance of B. pendula and H. caespitosum offspring in ecosystems where both species are present. In this context, the presence of a larger conspecific neighbor or, at least, of a plant with the same mycorrhizal type functioning as inoculum source seems to be essential for optimal seedling development during ecological succession of mixed plant communities on spoil banks, as has been described for other plant species and ecosystems (Nara, 2006a; Casanova-Katny et al., 2011; Zhenxing et al., 2019).