Studied species and seed collection
For this study, we selected a pair of congeneric species, Cirsium vulgare (the invasive species) and Cirsium oleraceum (the non-invasive species), Asteraceae, Carduoideae. Both species are native to Europe, with one of their ecological optima being in ruderal vegetation. Specifically in the Czech Republic, C. oleraceum is more frequent and abundant than C. vulgare [occurrence frequency in vegetation plots 4.1% vs 1%, mean percentage cover 7.8% vs. 2.3%, and maximum percentage cover 88% vs. 38% (Wild et al. 2019)]. However, globally, and especially in North America, C. vulgare is reported to be a noxious weed and highly invasive species (Julien and Griffiths 1998; Sieg et al. 2003; Tenhumberg et al. 2008), while C. oleraceum has never been reported as an invader elsewhere.
Seeds of both species were collected in the field in the Czech Republic in 2017. For each species, we collected mature seeds from at least 10 individuals. Seeds from all individuals were mixed and mother plants were not further distinguished in the experiment. All collected seeds were surface sterilized with a diluted SAVO Originál (a 4.7% sodium chlorite-based disinfectant) prior to the experiment to reduce the chance of soil contamination via seed surface fungi.
Experimental design
Following a commonly used methodology (Bever et al. 1997; Kulmatiski et al. 2008), the plants were grown in a two-phase experiment. In the first (conditioning) phase, conditioned soil was prepared. Soil biotic and abiotic characteristics were assessed after the conditioning phase to compare the effect of the two species on the soil. In the second (feedback) phase, we studied intraspecific PSF by growing the plants in 12 different types of soil, including conditioned and unplanted control soil with full, partial or no soil biota.
Conditioning phase
The aim of the conditioning phase was to prepare the soil, conditioned by the species, for the upcoming feedback phase. The conditioning phase was carried out between April 2018 and July 2018 in the experimental garden of the Institute of Botany of the Czech Academy of Sciences (49°59′38.972′′N, 14°33′57.637′′E), 320 m above sea level, temperate climate zone, where the mean annual temperature is 8.6 °C and the mean annual precipitation is 610 mm.
To set up the conditioning phase, we used a mixture of topsoil (purchased from JENA company) and sand (AGRO Jesenice) in 1:1 ratio (for chemical characteristics of the soil mixture see Table S1). For each species, we used 150, 1-liter pots (10×10×10 cm) in the conditioning phase. Half of the pots were sown with 10 seeds of one of the species in April 2018, the other half of the pots remained unsown and served as controls. It is important to stress that even though the control pots remained unplanted during the conditioning phase, the soil was still a live soil in which a mixture of plant species was previously grown, and it thus contained species non-specific soil biota. Each pot with conditioned soil was randomly assigned its unplanted control pot. The pairs of pots were always kept in close proximity to each other throughout the experiment so that they were exposed to exactly the same conditions. Both pots with and without plants were kept under the same conditions, regularly watered with tap water, and weeded on a weekly basis to avoid any effects of other species on the soil.
After the seeds germinated and the seedlings emerged, we randomly removed some of the seedlings to keep just one seedling per pot. The soil was conditioned for 12 weeks, similar to a range of previous studies (e.g., Chiuffo et al. 2015; Florianova and Munzbergova 2018; Meijer et al. 2011; van de Voorde et al. 2011; van Grunsven et al. 2007; van Grunsven et al. 2010). After 12 weeks, in July 2018, all plants were harvested, divided into aboveground and belowground parts (all larger roots were carefully taken out from the soil by hand), dried to a constant weight, and weighed.
After the harvest, the pots with each species as well as their paired unplanted control pots were randomly divided into ten groups of 7-8 pots and their soil was mixed. For each species we thus had 10 heaps of conditioned soil and 10 paired heaps of control soil. Heap served as a replicate and was further treated as such. From each heap, one sixth of the soil was collected for analysis of soil chemistry and soil biota, one third was kept untreated to serve as source of specific biota for soil inoculation in the feedback phase, and the rest was sterilized by gamma irradiation (sterilization dose 25 kGy, performed by Bioster a.s. in Veverská Bítýška) and used as a background soil in the feedback phase.
Feedback phase
The feedback phase was carried out between September 2018 and March 2019 in a greenhouse of the Institute of Botany of the Czech Academy of Sciences. The greenhouse was heated to 18°C and daylight was extended by two hours every day.
In the feedback phase, we grew each species in six treatments of conditioned soil and six treatments of unplanted control soil. These treatments included sterilized soil, sterilized soil inoculated with microbial filtrate of conditioned or control soil, sterilized soil inoculated with whole inoculum of conditioned or control soil, and non-sterilized whole soil (Fig. 1). Inoculum and filtrate always originated from the same heap of soil as the background soil or from their paired heap with different soil conditioning. Using these treatments, we can assess the role of individual components in the PSF. By comparing growth in control and conditioned non-sterilized whole soil we can assess total net PSF. By comparing growth in control and conditioned sterilized soil, we can assess the effect of abiotic PSF [although there are complications with nutrient enrichment due to soil sterilization by gamma irradiation (McNamara et al. 2003; Troelstra et al. 2001), see Discussion for more details]. By comparing growth in sterilized soil with growth in soil with microbial filtrate we can quantify the effect of microbiota (bacteria and non-mycorrhizal fungi). By comparing growth in soil with microbial filtrate with growth in soil with whole inoculum the effect of other groups of soil biota can be assessed. By comparing growth in sterilized soil with growth in soil with whole inoculum total biotic PSF effects can be assessed. By comparing growth in soils with filtrates or inocula from conditioned and control soils, we can assess the effect of soil biota abundance and/or specificity, assuming conditioned soils have higher abundances of soil biota, as well as more specific soil biota compared to control soils.
To set up the feedback phase, we used 10, 1-liter pots (10×10×10 cm) per species, soil conditioning and treatment, resulting in 120 pots per species, 240 pots in total. The bottoms of the pots were covered with keramzit sterilized in autoclave up to the height of 2 cm to compensate for soil lost during the harvest of the conditioning phase, and the rest of the pots was filled with 500 ml of soil mixed depending on the treatment. For non-sterilized whole soil treatments, we used untreated soil from the conditioning phase of the experiment. For whole inoculum treatments, we mixed sterilized soil and untreated soil from the conditioning phase in a 9:1 ratio. For the treatments with microbial filtrate, we filled the pots with sterilized soil and we watered them with the microbial filtrate. The filtrate was created by mixing 50 ml of untreated soil in 500 ml of distilled water, homogenizing the mixture, and filtering it through two filter papers with pore size of 11 μm. Therefore, the microbial filtrate does not contain micro-arthropods, nematodes, or arbuscular mycorrhizal fungi, whereas it should contain soil bacteria and fungi (van de Voorde et al. 2012). For sterilized treatments, we filled the pots with sterilized soil and we watered them with microbial filtrate sterilized in autoclave.
Each pot was sown with 9 seeds of the same species as in the conditioning phase. The pots were kept in the greenhouse, regularly watered, and weeded when needed. All pots originating from one pair of heaps were kept in the same block within the greenhouse. Seedling emergence was followed. Three weeks after the first seedlings emerged in all pots, all seedlings but one were removed from each pot to avoid competition. All seedlings emerging afterwards were recorded and removed as well. Twelve weeks after germination, the plants were harvested, divided into above- and below-ground biomass and weighed. All pots of both species were harvested at the same time.
Soil characteristics
Soil characteristics were analyzed after the conditioning phase for three types of soil: soil conditioned by the invasive species, soil conditioned by the non-invasive species, and the control soil. For each of the soil conditioning types, samples from six out of the ten heaps were randomly selected for the analyses. In addition, the analyses were performed also on soil collected before the conditioning phase (Table S1).
From abiotic soil characteristics, we measured actual and exchangeable pH, total C, N, P and available P, Ca, Mg, and K. From biotic characteristics, we determined soil microbial community composition using phospholipid and neutral fatty acids analysis (PLFA and NLFA) and we assessed the infection potential of arbuscular mycorrhizal fungi (AMF) by commonly used procedures: the most probable number [MPN (Adelman and Morton 1986; Wilson and Trinick 1983)] and mean infection percentage [MIP (Giovannetti and Mosse 1980; Moorman and Reeves 1979)].
Abiotic soil characteristics
Actual and exchangeable pH was measured using deionized water and 0.1M solution of KCl as extracting solutions, respectively (ISO 10390: Soil quality – Determination of pH. International Organization for Standardization, ISO 2000). Total C and N contents were determined by methods of Ehrenberger and Gorbach (1973) using CHN catalyst (Carlo Erba NC 2500), total P was measured according to the method of Olsen and Sommers (1982). Available P was measured in filtrate of 5 g of soil with 50 ml of 0.5M K2SO4 solution by flow injection analysis with spectrophotometric detection using the instrument QuikChem FIA + 8000 Series (Ammerman 2001; Egan 2001). Concentrations of available Ca2+ and K+ were measured using atomic emission spectrometry method and available Mg2+ using atomic absorption spectrometry according to methods of Moore and Chapman (1986) and Dědina (1987), with solution of 1M ammonium acetate as the extractant. All analyses were performed by the Analytical Laboratory of Institute of Botany of the Czech Academy of Sciences in Průhonice.
Soil microbial community
Soil microbial community composition was assessed using PLFA analysis performed by the Laboratory of Environmental Biotechnology, Institute of Microbiology of the Czech Academy of Sciences, following the methodology described in Garcia-Sanchez et al. (2019). The PLFA were extracted from 1 g of freeze-dried soil samples with a mixture of chloroform-methanol-phosphate buffer (1:2:0.8, v/v/v), as previously described by Bligh and Dyer (1959). The lipids were fractionated into neutral lipids (NLFA), glycolipids and polar lipids (PLFAs) using an extraction cartridge (LiChrolut Si-60, Merck, White-house Station, USA), and NLFA and PLFA were subjected to mild alkaline methanolysis as described in Snajdr et al. (2008). The free methyl esters of NLFA and PLFAs were analyzed by gas chromatography-mass spectrometry (450-GC, 240-MS ion trap detector, Varian, Walnut Creek, CA) following the same procedure described by Sampedro et al. (2009).
The soil microbial community composition was characterized using the following PLFAs: fungal biomass was estimated on the basis of 18:2w6,9 content (Snajdr et al. 2008), bacterial biomass was quantified as the sum of i14:0, i15:0, a15:0, 16:1w5, 16:1w7; 16:1w9, 10Me-16:0, i16:0, i17:0, a17:0, cy17:0, 17:0, 10Me-17:0, 18:1w7, 10Me-18:0, and cy19:0. Actinobacterial biomass was determined as the sum of 10Me-16:0, 10Me-17:0, and 10-Me18:0, Gram-positive bacteria (G+) as sum of i14:0, i15:0, a15:0, i16:0, i17:0, and a17:0, and Gram-negative bacteria (G-) as the sum of 16:1w7, 16:1w9, 18:1w7, cy17:0, and cy19:0. The NLFA 16:1w5 was assigned as a marker for the quantification of AMF and total PLFA concentration was used to estimate the total viable microbial biomass (Olsson et al. 2003). Last, we calculated microbial ratios F:B (fungi : bacteria), G+:G- (Gram-positive bacteria : Gram-negative bacteria), and F:AMF (fungi : AMF).
Infection potential of AMF
Infection potential of AMF was assessed by MIP and MPN methods. In the MIP assay, the colonization intensity of AMF is measured after a certain period of bait plant cultivation and the index of root colonization is the percentage of the number of 1-cm root segments showing detectable AMF colonization (Moorman and Reeves 1979). In MPN method, the test plants are grown in serial dilutions of the inoculum and the propagule density of the original material is statistically calculated from MPN scores (Feldmann and Idczak 1992).
To assess MPN and MIP, we evaluated mycorrhizal colonization of maize roots [standardly used for assessing MPN and MIP as its roots are strongly colonized by AMF (Moorman and Reeves 1979)] that was grown in each of the studied types of soil in 1:0, 1:10, 1:100, 1:1000, 1:10000 dilutions with soil sterilized in autoclave, in five replicates per dilution. Maize seeds (Zea mays convar. saccharata, var. Ashworth) were purchased from a commercial supplier (ReinSaat KG company, St. Leonhard am Hornerwald, Austria), they were germinated in Petri dishes in sterile conditions, replanted into 100 ml plastic containers (4×14 cm), and left growing in a greenhouse. After six weeks, the plants were harvested, fine roots from the middle part of the root system were collected, placed in 10% KOH for three months to bleach, and stained (left for 12h in 2% lactic acid, 12h in 0.05% trypan blue in lactoglycerol, rinsed in water, and soaked into lactoglycerol prepared from glycerol, 80% lactic acid and distilled water in 3:2:5 ratio).
The stained roots from 1:10, 1:100, 1:1000, and 1:10000 dilutions were observed using a binocular magnifier and presence of AMF propagules was recorded. MPN/ml was calculated using a program MPN Calculator, Build 23 using information on types of dilutions, number of replicates per dilution and number of replicates per dilution in which AMF propagules were recorded. To assess MIP, only the 1:10 dilution was used. Stained roots were placed into a Petri dish with a 1x1 cm grid and presence of AMF propagules at 200 intersections of roots with the grid was recorded using a binocular magnifier. An average value from the five replicates was calculated both for MPN and MIP, resulting in one MPN and one MIP value per soil sample and six replicates per soil conditioning type.
Statistical analyses
Differences in soil biotic and abiotic characteristics between soil conditioned by C. arvense, and by C. oleraceum were studied using linear direct gradient analysis (Redundancy Analysis, RDA) and Monte-Carlo permutation tests (Ter Braak and Šmilauer 2012) with 499 permutations. Dependent variables used in this analysis were all the studied soil characteristics except for actual pH, K content, total microbial and bacterial biomass, which were excluded due to high correlations with other variables (Table S2). The variables were standardized prior to the analysis. The independent variable was conditioning species. We repeated the analysis with all three soil conditioning types including the control soil and we present the results in the appendix (Fig. S1). The analyses were performed in Canoco 5 (Ter Braak and Šmilauer 2012). As a supplementary analysis, we also performed ANOVA using R 3.6.1 (R Core Team 2019) always with one of the studied soil characteristics as dependent variable and tested the differences between multiple levels of soil conditioning type using Tukey post hoc tests (Fig. S2).
Differences in plant performance between individual treatments and soil conditioning types in the feedback phase were tested using a linear (square root transformed biomass and root-shoot ratio) or generalized linear (seedling emergence as number of emerged seedlings out of the number of seeds sown, with binomial error distribution) mixed effect models in the R package ‘lmerTest‘ (Kuznetsova et al. 2017) with identity of the soil heap as random effect, and species, soil conditioning, treatment (sterilized soil, filtrate from control soil, filtrate from conditioned soil, inoculum from control soil, inoculum from conditioned soil, non-sterilized whole soil), and their interactions as explanatory variables. To estimate p-values, we used F-tests comparing two models with and without a tested term, using a ‘drop1’ function in the ‘lmerTest’ package (Kuznetsova et al. 2017). To assess differences between pairs of group means, we used Tukey post-hoc tests adapted to mixed effect models using ‘glht’ function in ‘multcomp’ R-package (Hothorn et al. 2008).
Afterwards, we used subset of data excluding the sterilized treatment and the non-sterilized whole soil treatment and tested the effect of the type of soil biota (filtrate vs inoculum) and conditioning of soil biota (from control vs from conditioned soil) as explanatory variables instead of the treatment variable, otherwise there were no changes in the analyses. We obtained very similar results when including only the subset of treatments in the analyses and so we only present these results in the Results section. Results of the analyses including all the treatments and not differentiating between type and conditioning of soil biota can be found in the appendix (Table S3). The two treatments which are excluded from the main analyses are, however, visualized in some of the graphs and compared using multiple comparisons with the rest.
Last, we used structural equation modeling [performed in the ‘lavaan’ package (Rosseel 2012) in R] to assess how individual components of soil, i.e. amount of soil nutrients, bacterial, fungal and AMF biomass, affect biomass of the two species. For the analysis, we only used data on plant biomass from the non-sterilized whole soil treatment as detailed soil analyses are only available for this treatment. A separate model was created for each species. The assumed relationships were as follows: (i) plant performance is affected by the amount of soil nutrients and by bacterial, fungal and AMF biomass, (ii) bacterial, fungal and AMF biomass are affected by the amount of soil nutrients, and (iii) bacterial, fungal and AMF biomass are correlated.