Experimental setup
We conducted a greenhouse experiment to study Lolium multiflorum Lam. (var. Lucero) plants grown in monoculture or in mixture with Trifolium repens L. (var. Junín). L. multiflorum plants were either endophyte free or associated with the endophyte Epichloë occultans (C.D. Moon, B. Scott & M.J. Chr.) Schardl, (E- or E + plants, respectively). The plants were grown in pots filled with sterilized soil, which was either inoculated or not with a combination of three species of arbuscular mycorrhizal fungi (AM + or AM-, respectively). Thus, the experiment had a factorial design with three factors with two levels each: neighbor identity, endophytic status of L. multiflorum plants, and pot AMF inoculation. Each treatment was repeated six times (n = 48).
On July 31st, eight seeds were sown in each 1.5 l pots. In monocultures, eight seeds of L. multiflorum of the corresponding endophytic status were sown, while in mixture treatments, four seeds of T. repens and four seeds of L. multiflorum of the corresponding endophytic status were sown. After two weeks, some plants were removed so as only four plants were retained in each pot (two of each species in mixture pots). The pots were kept in greenhouse and watered daily until September 20th, when they were harvested. Shoots were cut at the soil surface and the roots were washed carefully. The number of active nodules in T. repens roots was visually assessed by checking for pink coloration. Small root samples from T. repens and L. multiflorum were cleared and stained with Tryphan Blue. Then, they were observed under an optical microscope at ×200 magnification to identify AMF structures such as hyphae, arbuscules and vesicles (Phillips and Hayman 1970), which were detected in plants grown in AM + soils, but not in those grown in AM- soils. All shoots and the remaining roots were oven dried at 70°C for 48 h, and their dry weight was recorded. Ground dried tissues were then analyzed for C, N and P concentration.
To assess plant and symbiont fitness, we installed additional AM- and AM + pots, with E- or E + grass plants in mixture with rhizobia inoculated legume plants including the same plant density. These pots, resembling mixture treatments, were repeated six times (n = 24). Pots were kept until December 18th, when grass plants were senesced. We measured L. multiflorum seed production, the endophytic status of those seeds to estimating plant-to-seed endophyte transmission efficiency and the number of endophyte-associated seeds produced by E + plants (i.e. endophyte fitness), and T. repens flower production. For this, we collected spikes or inflorescences from L. multiflorum and T. repens plants, respectively. Each grass spike was kept in an individual bag, and then, each spike was threshed, and the number of full seeds was counted and weighted. From a subsample of ten seeds per spike, we estimated plant-to-seed endophyte transmission efficiency (see Symbionts measurements below). Inflorescences of T. repens were also counted and weighted. To estimate AMF and rhizobia fitness we measured soil abundance of AMF spores and rhizobia in a subsample of 150 g (wet weight) of the sieved soil (see Symbionts measurements below).
Substrate and biological material
We obtained endophyte-free and endophyte-associated L. multiflorum seeds following the methods previously employed in other studies (e.g. Omacini et al. 2006, García-Parisi et al. 2015, 2017). Briefly, one year prior to the experiment, we collected seeds from an old-field Pampean grassland (34°060S, 60°250W) dominated by a L. multiflorum population with ≈ 95% endophytic association. Half of the collected seeds were treated with triadimenol fungicide (0.5 g per 100 g seeds) to eliminate the endophyte. Fungicide-treated and untreated seeds were cultivated in 1 m2 plots, and the seeds produced by those plants were harvested and used in the experiment as E- and E + seeds, respectively. Endophyte presence in the seeds was evaluated through microscopic observation of 30 seeds collected from each plot, and stained with bengal rose (Bacon and White 1994). Results showed that about 95% of the E + seeds had the endophyte present, while no endophyte was observed in any of the E- seeds.
The AMF inoculum was composed of a mixture of internal and external hyphae and spores (32 ± 3.4 spores/g) of three fungal species known to colonize grasses and clovers: Funneliformis mosseae, Simiglomus hoi and Rhizophagus irregularis (as in García-Parisi et al. 2017). The inoculum was obtained by multiplying pure cultures of each fungus in plants of Plantago lanceolata L., Lotus tenuis L., and Bromus unioloides HBK, which were grown together in pots with sterile perlite and vermiculite. Plants received distilled water during the first week, followed by a modified Hoagland’s solution (0.02 mM P, as in García-Parisi et al. 2017) afterwards. Watering was interrupted once the plants showed > 60% root length colonization by AMF. Therefore, the inoculum consisted of the substrate, plant roots, and spores present in the pots. Additional pots were sown with the same plant species that grew under the same conditions but without AMF to obtain an inoculum control for AM- treatments in the experimental pots. Rhizobia were inoculated into T. repens plants in the mixture pots using a liquid commercial product (Ribol, Rizobacter Argentina S.A., Pergamino, Argentina) that exclusively contained Rhizobium leguminosarum bv. trifolii (> 109 bateria.ml− 1).
The experimental pots were filled with a mixture of sterile soil and sand (1:1, total C: 11.8 mg.g − 1, total N: 0.93 mg.g − 1, extractable P : 22.8 mg.g− 1). The soil was taken from the top 10 cm of a Mollisol, whose plant community was a dominated by exotic dicots, without presence of L. multiflorum or T. repens. Moreover, to reduce the amount AMF propagules the soil was sterilized by autoclaving at a pressure of 1013 hPa and a temperature of 100°C for 1 hour, thrice with a 24-hour interval between cycles (i.e. tindalization).
Carbon, nitrogen and phosphorous acquisition
Carbon (C), nitrogen (N), and phosphorus (P) acquisition (g.pot− 1) were estimated based on each element content in plant biomass, calculated as the product of the element concentration (%) and biomass (g.pot− 1) divided by 100. For the elements analysis, dried material of the shoot and root samples was ground to a fine powder in a ball mill. Tissue C and N concentration (%C and %N) and N isotopic composition (δ15N) were determined on aliquots of 0.7 ± 0.05 mg dry mass weighed into tin cups (IVA Analysetechnik, Meerbusch, Germany), combusted in an elemental analyser (NA1110, Carlo Erba Instruments, Milan, Italy) interfaced to a continuous-flow isotope ratio mass spectrometer (IRMS, Delta Plus, Finnigan MAT, Bremen, Germany). Samples were measured against a working gas standard previously calibrated against a secondary isotope standard. A laboratory standard (wheat flour) was run after every 10 samples to estimate the precision of the isotope analyses (± 0.14‰ s.d.).
P concentration (% of d wt) was determined on 10–20 mg of shoot and root samples by a modified phosphovanado-molybdate colorimetric method following acid digestion (Hanson 1950). P content (g.plant− 1) was calculated as P concentration (%) * aboveground biomass (g.plant− 1) divided by 100. Finally, N:P ratio was calculated to evaluate nutrient limitation considering that N:P ratio lower than 14 indicates N limitation and N:P ratio higher than 16 indicates P limitation (Koerselman and Meuleman 1996). Also, the N and P content (mg.pot− 1) were divided into C content (mg.pot− 1) to obtain nutrient concentration on a C basis (mg.g− 1 C)
Isotopic composition of N in tissues (δ15N) was used to evaluate the contribution of different N sources. In monocultures pots, all the N content (g N. pot− 1) in the grass tissues was acquired solely from the soil. However, in mixtures with the legume, N can be acquired either from the soil or from the legume via the transfer of biologically fixed N (N derived from biological fixation: Nbf). To estimate the contribution of soil N uptake and biologically fixed N transfer from the legume to the grass N acquisition/content, we employed the 15N natural abundance technique. This technique is based on the difference in the N isotopic composition [δ15N (‰) = (15N/14Nsample)/(15N/14Nstandard) – 1) x 1000] of atmospheric N and N derived from soil organic matter (Högberg 1997). The percentage of the total N content on the grass that was acquired through transferences from the legume (%Nbf) was estimated as:
%Nbf= (δ15Nplant monoculture – δ15Nplant mixture) / (δ15Nplant monoculture – B) x 100 (4)
where δ15Nplant mixture is the δ15N of the sample of L. multiflorum growing in mixture with T. repens, B is the δ15N of a plant whose N supply depends completely on atmospheric fixation, and δ15Nplant monoculture is the δ15N of a L. multiflorum plant that whose supply depends completely on uptake of soil N (i.e. growing in monoculture with other grasses).
The value of B was measured in additional T. repens plants inoculated with rhizobia, and grown in a perlite/vermiculite substrate watered with a modified Hoagland's solution containing no N. The values of δ15Nplant monoculture were measured in a set of E- or E + L. multiflorum plants cultivated in the monocultures, either inoculated or not with AMF. B values (mean ± SEM) were 2.1 ± 0.49‰ in mycorrhizal plants and 2.71 ± 0.68‰ in non-mycorrhizal plants. Values of δ15Nplant monoculture were 11.07 ± 0.23‰ in mycorrhizal plants and 10.71 ± 0.42‰ in non-mycorrhizal plants.
The contribution of soil N uptake and N derived from biologically fixation and afterwards transferred (Nbf) were calculated as:
Nbf (g.pot− 1) = N content (g.pot− 1) * %Nbf / 100 (5)
Soil N uptake (g.pot− 1) = N content (g.pot− 1) * (100 – %Nbf) / 100 (6)
Symbionts measurements
Endophyte fitness was estimated as the number of endophyte-associated seeds produced by E + plants and plant-to-seed transmission. Ten seeds from each spike were stained with bengal rose and examined under a microscope to estimate the transmission efficiency of each spike (Bacon and White 1994; García-Parisi et al. 2012). The number of endophyte-associated seeds produced by E + plants of each extra pot was calculated by adding the transmission efficiency of each spike multiplied by the number of seeds of the corresponding spike. Endophyte plant-to-seed transmission efficiency at pot level was obtaining by dividing the result by the total number of seeds in the pot. AMF fitness was estimated as the spore density remaining in the soil at the end of the experiment in the extra pots. For spore density estimation, a 50 g sub-sample of air-dried soil was wet-sieved and decanted according to the Gerdemann and Nicolson (1963) method, and the supernatant was centrifuged in a sucrose gradient following the Walker et al. (1982) protocol. Only non-empty spores were counted under a stereomicroscope. The total spore count in each sample was corrected for moisture content to express the value per gram of dry soil.
Rhizobia fitness was estimated as the rhizobia density remaining in the soil at the end of the experiment in the extra pots. To estimate this, we determined the most probable number (MPN) of soil rhizobia capable of nodulating T. repens in soil samples. For this, five serial ten-fold soil dilutions with four repetitions each were prepared according to the method described by Somasegaran and Hoben (1985), using 10 g of soil in 90 ml of physiological solution for the first dilution step. Test plants were grown under sterile conditions and inoculated with a 0.2 ml aliquot of the corresponding dilution. These plants were kept in sterile growing chambers supplied with an adequate volume of a sterile N-free nutrient solution. After 21 days of inoculation, plants were scored for nodulation. A plant with at least one functional nodule (determined by a pink color) was considered "positive," while a plant without nodules was considered "negative." The MPN of rhizobia was determined based on the number of positive and negative plants in the serial dilutions (Somasegaran and Hoben 1985). To ensure accuracy, four control plants were inoculated with the physiological solution used to prepare the dilutions, and all resulted in negative nodulation scores. The MPN of rhizobia was adjusted for soil moisture content to express it per gram of dry soil.
Statistical Analyses
Analyses were performed with linear mixed effect models with the package nlme (Pinheiro et al. 2023) using statistical software R (R Core Team 2023). For L. multiflorum plants, total C acquisition per pot (gC.pot− 1), shoot:root ratio, N concentration in shoot and root (mg N.g− 1 C), P concentration in shoot and root (mg P.g− 1 C), N:P ratio, total N content (mg.pot− 1), and total P content (mg.pot− 1) were analyzed with gls models (nlme) including the neighboring plant identity, endophytic status of L multiflorum plants and AMF inoculation as fixed factors. From mixture pots, L. multiflorum Nbf and Nsoil in shoot, root and total (% and mg.pot− 1), T. repens total C acquisition (g C.pot− 1), Nbf, Nsoil (mg N.pot− 1), and seed production (number.pot− 1) of L. multiflorum plants, T. repens flower biomass (g.pot− 1) and rhizobia density (log(bacteria.g− 1soil) were analyzed with gls models (nlme) including endophytic status of L. multiflorum plants and AMF inoculation as fixed factors. The spore density (number.g− 1 dry soil) in soils of AMF-inoculated mixture pots included only endophytic status of L. multiflorum plants, while plant-to-seed endophyte transmission efficiency on endophyte-associated plants growing in mixtures included only AM-inoculation as fixed factors.
Normality and homogeneity of variances were evaluated with shapiro.test and leveneTest. When necessary, the varFunc = varIdent function was used to stratify the variances within the levels of a factor (Pinheiro and Bates 2009). The estimates produced by gls allowed modelling the variances and the correlation of errors within each group (factors levels) for the fixed effects models (Pinheiro and Bates 2009). Number of seeds (discrete counting data) and transmission efficiency were analyzed with gls models including the specification of data distribution (Poisson distribution and binomial distribution). Overdispersion in each model was analytically evaluated. The significance of the fixed factors of each model were assessed through likelihood ratio test (LRT) with the function Anova(). When the interaction between factors was significant, pairwise test was performed using estimated marginal means (EMMs, also known as least-squares means) with the function emmeans() (package emmeans; Lenth 2024).