Origin of seeds and soil
Seeds of both C. stoebe and C. jacea were collected from wild populations in the Netherlands. Seeds were surface-sterilized in a 0.5% hypochlorite solution and germinated in sterilised medium. Emerged seedlings were planted in pots directly after the inoculation of soil organisms and dead seedlings were replaced only during the first two weeks.
We collected soil from Slovenia, where the range expanding C. stoebe is native (representing the original range), and from the Netherlands where the species recently has established (representing the new range). In both countries, we collected soil from 3–15 cm below the soil surface from a riverine grassland, avoiding the microsites where either of the plant species occurred at that moment. Soil was collected from Slovenia and the Netherlands at the end of May 2014, stored cool and transported to the laboratory of NIOO-KNAW in Wageningen, the Netherlands. Soil collected from each range (either original or new range) was mixed before the experiment to create uniform soil communities per range. Thereafter specific groups of soil organisms were isolated, or the abundances of soil organisms were reduced, using the approaches described below. Additionally, soil from a similar riverine grassland in the Netherlands was collected, sieved using 4 mm mesh, homogenized and gamma-sterilized (> 25 KGray) at Steris AST (Ede, The Netherlands). This sterilized soil was used as a background substrate for adding all specific organism groups or soil fractions as described below.
Overview of experimental approach
We used two approaches to estimate separate and combined effects of nematode and AMF communities from the original and new range on the growth of range-expanding plant species. In Experiment I, we first isolated nematodes and AMF from the field-collected soil and added them, separately or combined, to the sterilized background soil. After 16 weeks, we determined the biomass production of C. stoebe and C. jacea grown in these inoculated soils. In Experiment II, we first reduced the abundance of nematodes and AMF from field-collected soil by a wet-sieving method (Wagg et al. 2014). Then we inoculated the sterilized soil with these wet-sieved soil communities and determined biomass production of the test plants grown in these soils after 13 weeks. Because isolation of nematode and AMF communities and manipulation of soil communities with wet sieving may also influence the communities of small-size non-target microbes, the effect of non-target microbes was also tested in both experiments. During plant growth, soil biota had the opportunity to increase in abundance, similarly to a conditioning phase in a plant-soil feedback experiment (Bever et al. 1997). After harvesting the plants in Experiment II, the soil in the pots was split into two halves: one half was subjected to drought and the other was kept under regular water regime. As in the feedback phase of a plant-soil experiment, a second generation of plants was grown in the conditioned soils and biomass production was determined after 8 weeks.
Experiment I: Addition of soil organisms
The design of the first experiment consisted of a factorial combination of five treatments: plant species (range expanding C. stoebe or native C. jacea), origin of soil organisms (original or new range of the range expander), inoculation of nematodes (yes or no), inoculation of AMF (yes or no) and inoculation of non-target microbes (yes or no). Each treatment combination was replicated 5 times. In addition, there were 5 replicate pots with each plant species grown in sterilized soils, serving as a control. This resulted in 170 one-litre pots in total.
As we were interested in biotic effects, all soil communities were inoculated into 630 g of background soil, which had been sterilized as described above. To test the effect of AMF, we extracted AMF spores from soils of both origins following the protocol of International Culture Collection of (Vesicular) Arbuscular Mycorrhizal Fungi (INVAM). We aimed to apply spores from 70 g of soil per pot, but as the numbers of spores in inocula were variable, we applied a fixed amount (100 spores) to each pot. Collected spores were stored at 4 ˚C in 1 ml distilled water until they were pipetted next to seedling roots during planting. To test the effect of the non-target microbial community, we suspended 70 g of soil in 200 ml of demineralized water per pot, and left it settle for half an hour so that large particles sank. Then, the supernatant was sieved using a 20 μm mesh. The treatment combination with the addition of only non-target microbes was prepared as the small fraction treatment in Experiment II (below). All pots that did not receive this non-target microbial community received 200 ml of distilled water when seedlings were planted in order to ensure that all pots had received the same amount of water. Plants, non-target microbial communities and AMF communities were allowed to establish for two weeks before nematodes were added to avoid AMF spores being damaged by nematodes before germination. Nematode communities used in the experiment were collected from soil samples (using 70 g of soil per pot) using an Oostenbrink elutriator (Oostenbrink 1960). Nematodes were added to the pots assigned to nematode treatment using a volume of 200 ml of tap water while other pots received the same amount of tap water.
Experiment II: Reduction of the abundance of soil organisms
The design of the second experiment consisted of factorial combination of three treatments: plant species identity (range expanding C. stoebe or native C. jacea), origin of soil organisms (original or new range of range expander) and fraction size of soil organisms (full, medium and small). Each treatment combination was replicated five times. In addition, 5 replicates of each species were grown in sterilized soil, resulting into 70 pots in total. As in the first experiment, we were interested only in biotic effects, and we established pots with 560 g (dry weight) of sterilized background soil and 70 g of alive soil inocula from either original or new range. Additionally, 70 g of sterilized soil inocula from the opposite range of alive soil inocula was added to test the effect of soil communities in uniform abiotic conditions. To obtain sterilized soil inocula, range-specific soil inocula were gamma-sterilized as indicated above. Plants in control treatments were grown in 100% gamma-sterilized soil. We decreased the abundance of soil organisms in an experimental unit by sieving 70 g of live inoculum soil (mixed in 200 ml water) with either 1000 µm, 50 µm or 20 µm size mesh. The soil that did not pass selected mesh size was sterilized by two autoclaving cycles (30 min at 121ºC with 24 h between the cycles) and mixed with the sterilized soil in pots before seedlings were planted. Although such sieving method influences soil communities in general, selected mesh sizes have shown to have greatest effect on specific groups of soil organisms (Wagg et al. 2014). Specifically, we expect 1000 µm (full) fraction to consist of the total soil community, 50 µm (medium) fraction to contain decreased abundance of AMF, and 20 µm (small) fraction to have non-target microbes and decreased abundance of AMF and nematodes (Wagg et al. 2014). Seedlings were planted on 26th of May 2014. After above- and belowground plant biomass was harvested, soil from each pot was kept separately in a plastic bag at 4 ºC and used in the feedback phase described below.
The experimental design of the feedback phase consisted of factorial combinations of four treatments: drought (yes or no), plant species identity that was constant in the conditioning and feedback phase (range expanding C. stoebe or native C. jacea), origin of soil organisms (original or native range of the range expander) and fraction size (full, medium, small and sterilized). In addition, 20 pots (5 replicates for each plant species for original and new range) were included as a control. To maintain similar abiotic conditions for control plants, these plants were grown in soil collected from pots where plants were growing with non-target microbes in the Experiment I and this was sterilised by two cycles of autoclaving as indicated above before the start of the feedback phase. Each treatment combination had five replicates, resulting into 160 pots in total.
Soil from each pot from the conditioning phase of the Experiment II was divided into two halves of 350 g each. One-litre pots were thereafter filled with the equal parts (350 g) of inocula from the first phase and sterilized soil (collected and sterilized with gamma irradiation as indicated above). To evaluate whether the interactions between plant species and soil organisms change in extreme weather conditions, we subjected half of the pots to drought. By weighing the pots and providing water up to the pre-specified weight, the water level in drought treatments was kept at 30% of dry weight while water level at regular conditions was kept at 60%.
The root colonisation by AMF was estimated after harvesting Experiment I and after harvesting the conditioning phase of the Experiment II using magnified grid-line intersection method (McGonigle et al. 1990). Stained roots were mounted on microscope slides and the percentage of root length colonised by AMF was estimated by scoring the presence or absence of AMF structures at each intersection of root and the vertical crosshair for 120 intersections per plant individual. An intersection was considered mycorrhizal if the vertical crosshair intersected a hypha, an intercellular coil, an arbuscule or a vesicle. Counting was performed using an Olympus CH20 microscope at 400 × magnification.
All plant biomass was dried at 70 ˚C to constant weight and weighed. To analyse the effect of soil organisms on plant growth, we compared the growth response of each plant in a treatment with growth in sterilized conditions, using random pairing of treatment and control plants; more specifically calculating ln(treatment/control) following Brinkman et al. (2010). This paired approach was used in all experiments. Using this paired approach, negative values indicate growth reduction by soil organisms compared to sterilized conditions while positive values indicate growth promotion. Aboveground and belowground biomass response to treatments was analysed using ANOVA with all experimental treatments as fixed factors. More specifically, fixed factors in Experiment I were plant species identity (C. jacea, C. stoebe), the origin of soil organisms (original and new range), inoculation with nematodes (yes or no), AMF (yes or no), non-target microbes (yes or no). In the conditioning phase of the Experiment II the fixed factors were plant species identity, origin of soil organisms and fraction size (full, medium, small). In the feedback phase of the Experiment II, linear mixed-effects models, using package “lmerTest” (Kuznetsova et al. 2017), were used to analyse plant biomass response. In these models, plant species identity, origin of soil organisms, the fraction size of soil communities and drought (yes or no) served as fixed factors and pot number from the conditioning phase as random factor (Brinkmann et al. 2010).
AMF colonisation in control pots was always 0 (data not shown), therefore we analysed the change of AMF colonization to the addition of soil organisms (Experiment I) and to the wet-sieving treatments (Experiment II). As the abundances of different AMF structures were highly correlated, the analyses of total AMF colonisation (presence of hyphae, arbuscules or vesicules) is presented. To meet the assumptions of general linear models, AMF colonisation was arcsine-transformed prior to the analyses, while measured colonisation % is presented as mean values. The colonisation of AMF in plant roots was analysed using ANOVA with all experimental factors as fixed factors as for the analyses of plant biomass response. In case of a significant interaction between the fixed factors, the difference between all treatment combinations was estimated using Tukey’s HSD test. The composition of AMF and general fungal communities were also screened with molecular methods (see Supplementary information). All statistical analyses were performed in R (version 3.6.3, R Core Development Team 2020).