Jacobaea vulgaris Gaertn. (Asteraceae, formerly Senecio jacobaea) is a predominantly biennial herbaceous plant species regularly observed with annual or perennial life-cycles (Wardle 1987). Flowering in the second year is positively influenced by size of the first-year rosette (van der Meijden and van der Waals-Kooi 1979) and nutrient availability (Prins et al. 1990). Each flowering individual may produce up to 30,000 achenes (Harper and Wood 1957). Jacobaea vulgaris is native to Eurasia (Harper and Wood 1957). After several introductions to various places, J. vulgaris today, is considered invasive in North America (Isaacson 1973) and Australia (Schmidl 1972).

Seed collection was carried out in the summer of 2018 in the Pacific Northwest (invasive range) and Central Europe (native range) at the same time. For species identification, we referred to “Rothmaler - Exkursionsflora von Deutschland” (Jäger 2017) and local plant identification literature. We sampled 22 invasive and 24 native populations, spanning a maximum distance of 524 km and 742 km, respectively. For each population, achenes (hereinafter seeds) of 20 maternal plants were collected, if available. The seeds were stored separately by seed family (i.e., seeds from the same maternal plant) in paper bags at room temperature in the lab of the Institute for Ecosystem Research (Kiel University) until use.

Six populations each by range of origin (invasive – native) were chosen according to seed quality and availability to be included in this experiment. In addition, we intentionally included populations varying in size and density in order to cover a broad range of variation within ranges. The sites in the native range served both as donor populations for seed sourcing and target sites for (re-)transplantation. For the selection of these six native populations, we thus additionally had to acquire permission from local authorities, landowners, and the tenant farmers for conducting a transplant experiment on their sites. All field sites for this experiment are owned by the Stiftung Naturschutz Schleswig-Holstein (for population information, s. Suppl. material 1: fig. S1).

From each population, seeds from seven randomly selected seed families were sown in potting soil (TKS 2 pot Medium Coarse, Floragard Vertriebs-GmbH, Oldenburg, Germany) on germination trays in April of 2019. The seeds were covered by 1 cm of soil layer to prevent them from drying out. The germination trays were placed in a greenhouse cabinet with ambient temperature and a photoperiod of 12:12 (night/day) hours and watered daily in the following days. After four weeks, five seed families with the highest germination success within each population were chosen to be included in the experiment and seedlings were thinned to allow optimal growth. Once established, the germination trays were placed outside to allow acclimatization of the separated individuals to outdoor conditions.

The field experiment was designed to estimate performance of invasive individuals compared to native individuals in the species’ original native range. The location of the six native populations used for seed material sampling also served as transplantation sites. In each of these six sites, we established five experimental plots. Plot locations were assigned randomly within site with coordinates marking the southwest corner of each plot. Starting from this corner, an area of 0.9 m × 1.2 m was established, where transplants were arranged in 4 × 5 rows (all plants were 0.3 m apart) leading to a total of 20 planting positions. One individual from each of the six invasive populations and two individuals from each of the six native populations (and therefore also originating from the experimental sites (= at their population home)) were randomly assigned to the planting positions leaving out the southwest and northwest spots. Thus, a total of 18 individuals were planted per plot. Each two individuals from the six native populations were replicates from the same seed family. For each of those replicates, one individual was a priori randomly chosen for the present experiment while the second one was assigned to remain into the summer of 2020 as part of an additional experiment (s. Suppl. material 1: fig. S2 for schematic overview). Therefore, 12 of the 18 individuals per plot (6 native and 6 invasive) originally belonged to the experiment described here. In case of mortality, however, the native replicate individual served as a substitute in the present experiment (except for the assessment of survival). All remaining individuals were removed before flowering in summer 2020 to ensure prevention of seed set and dispersal.

Planting was carried out starting June 15^th 2019 (approx. 2 months after sowing). The vegetation in the plots was cut to approx. 0.3 m to reduce heterogeneity during early establishment of the experimental plants. Subsequently, experimental plants were brought out with the adhering potting soil and labelled for recognition. After planting, each experimental plot was watered with 10 l water right away and two additional times after one week to assure establishment and survival of the planted individuals.

Next to each of the five experimental plots, a 2 m × 2 m monitoring plot was established following the diagonal extension 5 m apart in a northeastern direction. The monitoring plots served for recording vegetation composition and structure, including information on overall vegetation height (as a measure for productivity), coverage using a modified Londo scale (Londo 1976) and percentage of bare soil. Soil samples were taken next to the monitoring plot for soil C:N determination (with a EURO Elemental Analyzer) and determination of pH in the laboratory at the Institute for Ecosystem Research, Kiel University. In the monitoring plots, Jacobaea abundance (i.e., naturally occurring individuals) was additionally determined to infer local population density. Soil moisture was measured at the four corners of the experimental plot using Time Domain Reflectometry with ML3 Theta-soil moisture sensor (Delta-T devices Ltd, Cambridge, United Kingdom). Mean relative light intensity was determined from four evenly distributed measurements taken within each experimental plot (LI-1500 Light Sensor Logger, Li-COR Biosciences, Lincoln, United States).

The monitoring of the field experiment ran from June 28^th until September 28^th. After 6 weeks and 14 weeks of experimental runtime, we determined specific leaf area (SLA) and leaf dry matter content (LDMC) of transplants. For this, the third fully developed leaf from the top was taken from each plant and stored in a moisturized plastic bag in a cooler box for transportation to the lab. Fresh leaves were scanned (Expression 11000XL, EPSON Deutschland GmbH, Meerbusch, Germany) and leaf area was determined using WinFolia (WinFolia Pro 2015, Regent Instruments Inc., Quebec, Canada). Leaf fresh weight was determined using a precision scale (Sartorius 1702MP8, Sartorius AG, Göttingen, Germany). All leaves were dried at 65 °C for 48 hours afterwards for subsequent dry weight determination. At the end of the experimental runtime, we determined the transplants’ expansion in two directions (to calculate rosette size), the number of healthy leaves, and length of the longest leaf. Herbivory was assessed as a binary trait and considered present when parts of the leaf were missing or by the appearance of characteristic “bullet-holes” caused by Longitarsus jacobaeae. For biomass determination, all invasive individuals and half of the native individuals were dug up (i.e. 12 individuals per plot). Dry weight was separately determined for aboveground and belowground biomass after drying in a drying oven at 65 °C for 48 hours. We additionally calculated the root:shoot ratio as a measure for resource distribution strategy.

The greenhouse experiment was set up analogous to the field experiment and ran from August 21^st (approx. 4 months after sowing of seeds) to November 13^th. To decouple the influence of soil biota effects from other environmental factors varying with the field sites, soil samples were taken from all six native field sites used in the field experiment. These soil samples served as an inoculum for soil-biota treatments to all native and invasive individuals. For this, soil material was sampled about 0.5 m south of the southwestern corner of each experimental plot. After careful sod removal, a volume of 1 l soil was taken per plot, sieved through a 2 mm mesh and collected in a sterilized bucket. Separate soil samples from all plots were pooled and merged by site and served as the site-specific donor substrate. Soil sampling equipment was sterilized between sites to avoid cross-contamination.

All 12 population origins incorporated in the field experiment were also used in the greenhouse with three seed families randomly chosen out of the five used in the field. For each seed family, each one individual was grown with soil addition from one of the six field sites or only using standard substrate (control). Standard substrate consisted of 60% fine sand provided by the Botanical Garden of Kiel University and 40% unfertilized potting soil (F.E. Typ Nullerde, HAWITA Gruppe GmbH, Vechta, Germany) constituting an environment especially low in nutrients. This led to a total of seven different treatments for each seed family, thus resulting in a total of 252 individuals in the greenhouse experiment. All individuals were transferred to 1.5 l planting pots filled with 1.26 l standard substrate supplemented either by 0.14 l of soil collected from one of the six field sites (9:1 standard substrate:field soil) or an additional 0.14 l standard substrate for the control group. The standard substrate was processed by an autoclave (Webeco Dampf-Sterilisator, Matachana Germany GmbH, Selmsdorf, Germany) to reduce already present soil biota to a minimum. Each pot additionally received 3 g slow-release fertilizer (2.14 g/l) (Basacote Plus 6M 16 + 8 + 12 (+ 2 + 5), Compo Expert GmbH, Münster, Germany) corresponding to low levels of nutrient availability as per the manufacturer’s specifications. Planting pots were put on saucers and distributed in the greenhouse. Their position on benches was randomized every week. Predatory mites and sticky traps were installed at the beginning of the experiment to reduce infestation risk with insects. After one month, an insecticide was used on all plants (Spruzit Schädlingsfrei, W. Neudorff GmbH KG, Emmerthal, Germany), and milk and neem oil were applied to all individuals to prevent the spread of mildew. Throughout the experiment all plants were watered with 75 ml of tap water every 1–3 days as needed. Excess water from the saucers was emptied after every watering.

After the experimental runtime, monitoring and biomass harvest were carried out analogous to the field experiment. SLA and LDMC were assessed after 4 weeks and 12 weeks, respectively.

Statistical analyses were performed with R (Version 4.1.1) (R Core Team 2019). For the field experiment, a linear mixed effects model was fitted using lme4 (Version 1.1.27) (Bates et al. 2015) and lmerTest (Version 3.1.3) (Kuznetsova et al. 2017). Response variables were transformed if necessary (s. Suppl. material 1: table S1 for transformations used). We included origin (native/invasive) as a fixed effect and the random intercept for plot nested in site and seed family nested in population. We performed a correlation test (function rcorr from the package Hmisc (Harrel 2021), type = “pearson”) with the environmental variables. Mean moisture was significantly correlated with maximum plant height, C:N ratio as well as cover of bare soil, whereas relative light was significantly correlated with maximum plant height, α-diversity as well as cover of bare soil, respectively. Soil pH was significantly correlated with α-diversity. Subsequently we excluded mean moisture, pH and relative light intensity as covariates in model fitting. We included number of leaves one week after planting, population density of J. vulgaris individuals as well as α-diversity (i.e. number of species), C:N ratio, maximum vegetation height and percentage of bare soil in the monitoring plot as covariates in the full model. The best model was subsequently selected using the step function (lmerTest) and retained covariates were additionally tested for their interaction with the fixed effect origin in the final model. For survival and herbivory, we fitted generalized linear mixed effects models using the binomial-family and the same structure as for the other variables; however, model selection was done manually.

Similarly, for the greenhouse experiment, we fitted a lmer with origin and provenance of the soil (site) as fixed effects. As a covariate we included either the respective response variables’ value at the beginning of the experiment or the initial number of leaves if no starting value was available (for biomass variables, SLA, LDMC). We added the random intercept for seed family nested in the population. For the number of leaves, we fitted a glmer with the poisson family. Differences between the treatments were examined using the Tukey post-hoc test in emmeans (Version 1.7.0) (Lenth 2019). All plots were created using ggplot2 (Version 3.3.5) (Wickham 2016). Predicted values for response variables used in plots were obtained using effects (Version 4.1.4) (Fox and Weisberg 2018, 2019).

The datasets generated during and/or analyzed during the current study as well as the code used for analysis are available from the corresponding author upon reasonable request.

After 14 weeks of experimental runtime, individuals originating from the invasive range had developed larger rosettes (Table 1, Suppl. material 1: table S2) and showed a significantly higher aboveground biomass than those from the native range (Fig. 1a). There was no difference in the belowground biomass (Fig. 1b), which led to a smaller root:shoot ratio in invasive origins (Fig. 1c).

Figure 1. 

Origin effects (field experiment). Response of performance (a–c) traits in relation to origin of the seeds for each individual. Data shown are predicted values from the model ± SE. Native individuals (left, blue) originated from the field sites where the experimental plots were located. Invasive individuals (right, orange) originate from the Pacific Northwest. Herbivory (d) was assessed as a binary trait (presence/absence) only. N = 319 (a–c) and n = 322 (d). For depiction of raw data s. Suppl. material 1: fig. S3.

Irrespective of origin, C:N ratio and maximum height of the vegetation in the monitoring plot displayed a significantly negative relationship with J. vulgaris belowground and total biomass as well as with length of the longest leaf (not shown, Table 1). The higher the maximum vegetation, the higher was also the number of individuals with signs of herbivory (Table 1) and this pattern was more pronounced in native than in invasive individuals (Fig. 1d, Table 1). There was no difference in survival depending on origin (Table 1).

For functional leaf traits, no significant origin effect could be detected (Table 2, Suppl. material 1: table S3). Specific leaf area (SLA) after 6 weeks of experimental runtime, however, was negatively correlated with the number of species in the monitoring plot (α-diversity) for individuals originating from the species’ invasive range (Fig. 2a). After 14 weeks, SLA was higher with increasing height of the surrounding vegetation, irrespective of plant origin (Fig. 2b). In addition, after 14 weeks, there was an interaction effect of LDMC with increasing α-diversity displaying an increase in LDMC in invasive individuals but no such pattern in native origins (Fig. 2c).

Figure 2. 

Origin effects × covariate (field experiment). Effects of origin in interaction with α-diversity (a, c) and maximum vegetation height of surrounding vegetation within the experimental plot (b). Data shown are predicted values from the model with upper and lower range. O = origin, α-Div = α-diversity, mvh = maximum vegetation height (of surrounding vegetation), SLA = specific leaf area, LDMC = leaf dry matter content. Week 6 and Week 14 indicates that the leaves for analysis were harvested after 6 weeks of experimental runtime or at the final harvest of the plants after the entire experimental runtime, respectively. N = 193 (a), 311 (b), 319 (c). SLA after 6 weeks was only taken for plants that had a sufficient number of healthy leaves and therefore constitutes a reduced subset. For depiction of raw data s. Suppl. material 1: fig. S4.

In the greenhouse experiment, there was no difference in biomass depending on the origin of the individuals (Table 3, Suppl. material 1: table S4). Provenance of the soil (treatment) did not affect the aboveground biomass (Fig. 3a) and root:shoot ratio (Fig. 3d) but belowground biomass (Fig. 3b) and total biomass differed significantly between the treatments (Fig. 3c). Number of leaves was the only trait exhibiting significant soil provenance × population origin interaction effects, but post-hoc test revealed no significant difference for origins within any given site.

Figure 3. 

Treatment effects on performance traits (greenhouse experiment). Effects of soil provenance in the greenhouse experiment. Data shown are predicted values from the model ± SE. Different colors represent different soil origins (only soil from native sites were included in this experiment). Different letter combinations in the panels indicate significant differences according to the Tukey post-hoc test. ALB = Albersdorf, ARB = Arpsdorf, BUN = Bünsdorf/Wittensee, Con = Control (no soil added from any field site), PRE = Preetz, ROT = Rotenhahn/Eidertal, VOL = Vollstedter See. For location information see Suppl. material 1: fig. S1. N(a) = 250, n(b) = 238, n(c) = 251, n(d) = 230. For depiction of raw data s. Suppl. material 1: fig. S5.

For the functional traits SLA and LDMC, no difference depending on the origin of the individuals could be detected (Table 4, Suppl. material 1: table S5). However, soil provenance incurred differences in SLA after four (Fig. 4a) and 12 weeks (Fig. 4b), and LDMC differed significantly depending on soil provenance after four weeks (Fig. 4c). No differences in LDMC were observed after 12 weeks of experimental runtime (Fig. 4d).

Figure 4. 

Treatment effects on functional traits (greenhouse experiment). Effects of soil provenance in the greenhouse experiment. Data shown are predicted values from the model ± SE. Different colors represent different soil origins (only soil from native sites were included in this experiment). Different letter combinations in the panels indicate significant differences according to the Tukey post-hoc test. ALB = Albersdorf, ARB = Arpsdorf, BUN = Bünsdorf/Wittensee, Con = Control (no soil added from any field site), PRE = Preetz, ROT = Rotenhahn/Eidertal, VOL = Vollstedter See. For location information see Suppl. material 1: fig. S1. N(a) = 215, n(b) = 248, n(c) = 251, n(d) = 249. For depiction of raw data s. Suppl. material 1: fig. S6.

There is some evidence that specialist herbivores prefer invasive individuals of J. vulgaris over native ones (Lin et al. 2015). In the experimental area of Northern Germany, these specialist herbivores are naturally present. However, the specialist Tyria jacobaeae usually prefers larger individuals (van der Meijden 1976) while additionally, theory and evidence suggest that generalist herbivores prefer native origins (Keane 2002; Joshi and Vrieling 2005). The fact that our transplants were relatively unattractive to specialist herbivores while previous studies predict higher attractiveness of native origins to generalists, might have contributed to the observed increased performance in invasive origins and increased occurrence of herbivory in native origins. For moths and butterflies, specialist species were found to prefer lower vegetation compared to generalists (Pöyry et al. 2006), which in combination with the significant herbivory × maximum vegetation height interaction in our experiment, further corroborates this assumption.

In contrast, the absence of any origin-dependent differences in the greenhouse experiment was unexpected given previous studies with J. vulgaris showing higher performance of invasive origins (Joshi and Vrieling 2005). Different factors could explain this outcome. First of all, greenhouse experiments, by definition, involve herbivory to be kept at a minimum. This design therefore neither favors native (more specialists) nor invasive (more generalists) individuals of J. vulgaris in a greenhouse. Secondly, the greenhouse experiment was set up later in the year while relying on the same growing material. Accordingly, the experimental individuals for the greenhouse experiment were older than the transplants. Seasonal timing might thus have additionally affected plant development differently.

In the field experiment, the invasive individuals might also have benefitted from atypically high temperature and decreased precipitation during the experimental runtime (Broennimann et al. 2007; Early and Sax 2014). During the summer of 2019, the average temperature in Germany was 2.9 °C higher and precipitation on average was 27.7% lower than predicted according to the international climatological reference period (Deutscher Wetterdienst 2020). For the populations included in this experiment, the invasive range is both warmer (native: 15.6 ± 1.38 C; invasive: 16.21 ± 1.18 °C) and drier (native: 232.5 ± 6.75 mm; invasive: 125.5 ± 34.49 mm) during the warmest quarter on average (Data from BioClim, accessed Nov. 10^th 2020; Fick and Hijmans 2017), so the present finding might be indicative of a possible climatic niche shift in invasive origins (Broennimann et al. 2007). However, previous studies with J. vulgaris showed that, inter alia, competitive ability and regrowth (Lin et al. 2015), growth and regrowth (Lin et al. 2018) as well as growth, photosynthetic rate, and LMR (Lin et al. 2019) differ between native and invasive origins of J. vulgaris, but these differences were explicitly not driven by climatic conditions.

In the present study, the provenance of soil (treatment) differently impacted J. vulgaris performance (e.g., aboveground biomass), thereby confirming that soil-borne biotic effects contribute to differentiation among populations, as expected (hypothesis III). Strong negative feedbacks on J. vulgaris populations themselves have previously been shown in native populations (van de Voorde et al. 2011, 2012a). They furthermore increased with population density (van de Voorde et al. 2012a) and over time in an interspecific competition setting (Bezemer et al. 2018). Additionally, other species were found to exert PSFs on J. vulgaris (Bezemer et al. 2006; van de Voorde et al. 2011; Wubs and Bezemer 2018), showing that PSFs are a factor not to be neglected when studying this species’ performance and population dynamics.

Knowledge about soil provenance × plant origin interactions in general is lacking for this model species to date and we found no signs for enemy release on the belowground level as shown for Centaurea maculosa performance (Callaway et al. 2004). Our findings are more in line with a study on Verbascum thapsus that found signs of coevolution between plants and soil microbes on the between-population level but not between ranges of origin (Dieskau et al. 2020). To ensure that the patterns observed in our experiment are not co-affected by differences in soil properties, soil washes as applied by Dieskau et al. (Dieskau et al. 2020) should preferably be used in future experiments to address pure biotic effects most precisely.

In the present experiment, maximum vegetation height (strongly linked to light availability with r = - 0.62, p<0.001 with Pearson’s rank) was a relevant environmental factor for both origins, with the typical responses of increasing specific leaf area with decreasing light availability (Cornelissen 1992). LDMC varied more in the further course of the experiment, whereas variation in SLA was higher earlier in the experiment with different responses to increasing α-diversity depending on the range origin of the transplants. Given that the invasive individuals allocated more biomass by the end of the experiment, this pattern points towards a reaction to different environmental conditions rather than a sign of stress. Higher phenotypic variation was not only shown to pay off for invasive species compared to native congeners (Funk 2008), but may also apply within a species when comparing native and invasive origins (Caño et al. 2008; Hock et al. 2019). Under these considerations, our results suggest that high phenotypic plasticity might be one important factor contributing to the invasion success of J. vulgaris as well.

In summary, we cannot conclude explicitly which factors are the main drivers of increased performance of invasive transplants in the species’ native range (hypothesis II). It is, therefore, also difficult to accurately predict the long-term consequences of back-introduction of propagules or individuals of invasive origin into the species’ native range. However, genetic admixture might accelerate geographic expansion and invasion (Qiao et al. 2019), thus re-introduction of these invasive genotypes could therefore be problematic. Re-introduction might return originally common but then modified (“adapted”) genotypes that have undergone selection in the invasive range. At the same time, it could also add genotypes from other parts of the species’ native range, especially in J. vulgaris. In our experiment, we cannot determine whether the invasive transplants are re-introduced in the strict sense or originally came from other parts of the species’ native range, especially as invasive populations of J. vulgaris were found to probably be admixed before spreading within the invasive range (Doorduin 2012). In Northern Germany, J. vulgaris is assumed to form large panmictic metapopulations (Jung et al. 2020). However, for other species, intraspecific hybridization was previously shown to promote invasion processes (Kolbe et al. 2004; Geiger et al. 2011) and these findings could also apply to cryptic invasions. Therefore, introducing genetic material from the invasive range irrespective of its history, may evoke concerns among nature conservationists for the species’ native range.