Experimental design
We selected alfalfa cultivars from the three main commercial seed suppliers for western Canada: BrettYoung (Manitoba), DLF-Pickseed (Saskatchewan), and Northstar seeds (Manitoba), with the aim to maximize the trait diversity based on trait descriptions, fall dormancy, and winter hardiness ratings as reported in the seed catalogs from each respective seed supplier in 2020 (Table S1). One additional salt-tolerant cultivar (AC-Bridgeview) was sourced from Agriculture and Agri-food Canada. Most cultivars suitable for Canadian agriculture have a fall dormancy rating below 4. This means that the plants slow growth, metabolic activity, and increase carbohydrate storage near the end of the growing season. Lower fall dormancy ratings are linked to better winter hardiness, though there are additional mechanisms that support winter survival (Claessens et al. 2022). Some descriptions of cultivar characteristics overlap between cultivars and suppliers, and these characteristics were not independently verified, but we believe this list is a good representation of the alfalfa types available in Canada in 2020.
Growth conditions
This experiment consisted of a fully factorial combination of two AMF treatments (with or without AMF inoculation) on nine cultivars subjected to two stress treatments and an unstressed control replicated five times (n=270). Planting began on June 22nd, 2020, in the University of Saskatchewan Agriculture greenhouses (Saskatoon, SK). Planting was staggered over the next three weeks, with two full replicates planted per week. All plants were grown for four months. Harvest began on October 23rd, staggered so that each replicate had an equal growth period, with the final replicate harvested on November 22nd, 2020. Daylength and temperature fluctuated over the season; natural sunlight was supplemented with sodium halogen lamps keeping a minimum of 15h daylight. Daytime maximum temperature was 26 °C throughout most of the growth period, with a minimum nighttime temperature of 18°C.
We mixed soil from equal parts (by volume) screened commercial topsoil, sand, and sphagnum peat moss. The soil was moistened then sterilized at 150°C for 4 hours in a drying oven twice, ensuring that the internal temperature reached at least 120 °C each time. The commercially sourced seeds had a manufacturer-applied coating containing layers of rhizobia, fungicide, and fertilizers. We removed this coating to isolate the effects of our selected AMF inoculant from any enhancement from the coating, or damage from the fungicides. To dissolve this seed coating, we soaked the seeds in water, then 70% EtOH for one minute each, then sterilized the seeds in 5% hypochlorite for five minutes. We rinsed the seeds well then coated with Sinorhizobium meliloti inoculant (Exceed® alfalfa and true clover inoculant, Visjon Biologics) immediately prior to planting.
For each replicate, three alfalfa seeds were planted in half gallon pots filled with 2L of the soil mix. Half the pots were inoculated with 0.04g of AGTIV® forages powder (PremierTech ; 8000 spores/g of Rhizophagus irregularis), approximately 320 spores per pot, as per the recommended rate of application from the manufacturer. We planted inoculated alfalfa on alternate days to uninoculated alfalfa to minimize the risk of cross contamination in addition to sterilizing tools with 70% EtOH between planting sessions. Pots were thinned to two seedlings a week after they sprouted.
We applied two stress treatments – drought and salinity – beginning 40 days after seedlings sprouted to allow them to become well established before becoming stressed. All plants were fertilized weekly with 250mL of half strength Hoagland’s solution. After initial establishment, all plants except the drought treatment group were watered to saturation three times a week, the watering interval increased to every other day as the plants increased in size, and the daily greenhouse temperature increased. Drought was simulated by reducing watering to once a week initially, increasing to every five days as the plants grew. For salinity, we dissolved NaCl in the weekly fertilizer. The salt treatment began at 20mM NaCl and increased by 40mM each week (to prevent shock) to a final to a final concentration of 140mM.
Pest control:
Thrips, aphid, and spider mite infestations occurred periodically during the trial. We used Amblyseius cucumeris mites (Biobest® ABS-Mini sachet) throughout the growing period to control thrips. Near the end of the growing period (September 28th and October 2nd) all plants were sprayed with a Beauveria bassiana based biological insecticide (Botanitguard, BioWorks®), and a pymetrozine based insecticide (Endeavor, Syngenta®) to control an aphid and spidermite outbreak. No plants were lost to insect infestation, but there was some dieback. There was no clear pattern of insect damage between treatments, so the dead shoot biomass was clipped, but not included in the final biomass at harvest.
Data collection
Beginning 10 days after treatment initiation, we measured plant height and the number of non-tripped, unwilted flowering stems present on each plant weekly for seven weeks. In October, one standard bumblebee hive (Bombus impatiens, Biobest®, Ancaster ON.) was placed in the greenhouse chamber to facilitate pollination. The hive contained a ~150 worker brood box and a food source. Bees remained in the greenhouse chamber with free access to their brood box until the end of the experiment.
All plants were harvested after four months of growth. At harvest, we removed and dried seed pods for later seed extraction. Alfalfa shoot mass was clipped, then dried for 48h at ~70°C. After drying, leaves and stems were separated and weighed. All roots were washed, dried for 48h at ~70° C, then weighed, except for small subsamples (approximately 1g fresh weight) used for AMF colonization, which we stored in 70% EtOH prior to processing.
To measure root colonization we used a root staining method modified from Vierheilig et al. (1998): roots were heated to 90°C in beaker of 10% w/v KOH solution for 30 minutes, followed by 2% v/v HCl solution for a further 15 minutes. The heat was reduced to 80°C, then the roots were stained in a dye composed of 5% v/v black ink and 5% v/v acetic acid for 15 minutes. Roots were well rinsed in DI water between each step. Finally, the roots were stored in a mixture of equal parts glycerol, 5% acetic acid, and DI water for a minimum of two days to remove excess ink. Using 25 cm of root tissue, we estimated percent colonization using the line-intercept method, noting hyphae, arbuscules, and vesicles for 100 intercepts (modified from McGonigle et al., 1990).
We quantified the nitrogen, phosphorous, and sodium content in leaves from six cultivars that differed most in their growth responses in preliminary results across all treatment combinations (n=180). Dried leaf tissue was finely ground then 0.15 ± 0.01g was digested in sulphuric acid as in Lindner (1944). Nitrogen and phosphorous concentrations were measured colorimetrically with an AA3 Segmented flow analyzer (SEAL analytical). Sodium levels in the same samples were measured by atomic absorption spectrometry with a 200 Series AA systems analyzer (Agilent).
We measured the final electroconductivity (EC) of soil collected after plants had been harvested to see the extent of soil salinization and determine if AMF affected soil EC. Due to logistical constraints, soil was pooled across cultivars for each replicate within each AMF and treatment group combination (n=30). To measure EC, we mixed a 250mL soil sample with 500mL of DI water, stirring for three minutes. The aqueous solution was then filtered for electroconductivity measurement with an Orion Star A212 Conductivity Benchtop Meter (Thermo scientificTM, Waltham, Massachusetts, United States).
Statistical analysis
Prior to analysis, 8 plants were removed due to labeling errors leaving 262 individuals. All analysis was done in R version 3.6.3 with mixed models conducted with the package ‘lme4’ (Bates et al. 2015) unless otherwise noted. We checked model homoscedasticity via Levene’s test in the ‘car’ package (Fox and Weisberg 2019). Model residuals were checked visually using the package DHARMa (Hartig 2022), and using a Shapiro-Wilks test of normality in base R. To conduct F tests, we estimated degrees of freedom using Satterthwaite’s method in the lmerTest package (Kuznetsova et al. 2017) unless otherwise noted. When predictors were found to have significant or marginally significant (p<0.1) effects in the above models we used the package ‘emmeans’ to compare means or slopes in post-hoc tests using Tukey’s method for adjusting p values (Lenth 2022). Partial residual plots were created using the interactions package (Long 2019) and model-estimated means plots in the afex package (Singman et al. 2021).
To find the relationship between cultivar, stress treatment, and AMF colonization (where “colonization” includes the presence of hyphae, vesicles, and / or arbuscules) we used a square root transformed (to reduce skew) liner mixed effect model with stress treatment and cultivar as the predictor variables.
To determine the effect of AMF inoculation on plant responses independent of stress, we ran five models, focusing on shoot biomass, root biomass, seed mass, leaf nitrogen content, and leaf phosphorus content. Each model included only data from the unstressed plants, with AMF inoculation and cultivar as fixed effects and replicate as a random effect. For seed mass, we focused only on those plants that did set seed and excluded two cultivars (Assalt and TH2) as they did not set seed in either the inoculate or uninoculated treatments. We also log transformed both seed mass and leaf P to normalize the residuals.
We calculated the biomass and nutrient stress response for each plant in the two stress treatments as the biomass or nutrient concentration of a stressed plant divided by the corresponding value in an unstressed plant within the same replicate block. We use stress responses, rather than more typical mycorrhizal responses, to focus on how AMF influence plant responses to stress instead of how stress influences the effects of AMF.
We determined the relationships between biomass stress response and mycorrhizal inoculation in different treatment conditions by creating four separate mixed models, each testing the interaction of cultivar and AMF inoculation on either shoot or root stress responses to drought or salinity. Each model had replicate as a random effect to control for temporal differences in planting and harvest time. We used this same model set up for testing N and P stress response models, with the addition of a random effect to control for separate acid digestion batches. All stress responses were log transformed to satisfy to normality requirements.
Alfalfa growth rates (measured as height) were linear across the first 5 weeks before plateauing across weeks 6 and 7. Therefore, we modeled alfalfa height over time across the first 5 weeks to determine how AMF affected growth rates, and how growth rate differed among cultivars using separate models for each treatment condition, with plant identity as a random factor to account for repeated measures.
We modeled the number of flowers at three time points in relation to peak flowering time: early, mid peak and late peak. These time points are two weeks apart which minimized the number of flowers that may be counted twice. Flower number was modeled using three Poisson generalized linear mixed models (one per treatment) with AMF, cultivar, and time point as interacting fixed effects and block and plant ID (to account for repeated measures) as random effects. we tested for significant difference in means with type three Wald chisquare tests using ‘car’ package (Fox and Weisberg 2019) on each model.
Out of 262 plants, 148 produced seeds, with multiple cultivars failing to produce seeds in different treatment combinations. Due to these data limitations, we focus on the total mass of seed produced by plants that set seed. We calculated the mass of produced seeds per gram total plant mass to control for higher seed production in larger plants. Seed mass was square root transformed to normalize the residuals. We included cultivar and AMF as fixed effects and replicate as a random effect in three linear mixed models (one per treatment). Cultivars that failed to produce seed in any combination of AMF and stress treatment were excluded from the model for the corresponding stress treatment.