Melampyrum arvense L. and Rhinanthus alectorolophus (Scop.) Pollich are annual xylem-feeding root-hemiparasitic species native to Europe. Melampyrum typically grows in dry grasslands and steppes, while Rhinanthus alectorolophus (Scop.) Pollich favours dry to mesic grasslands. Solidago gigantea Aiton and Symphyotrichum lanceolatum (Willd.) G. L. Nesom are perennial rhizomatous species from the Asteraceae family, originating from North America (Pyšek et al. 2012). They began spreading across Europe in the 19^th century and have become serious invaders (Weber and Jakobs 2005; Jedlička and Prach 2006; Axmanová et al. 2021). Solidago and Symphyotrichum are considered typical wetland species, but they also occur in disturbed anthropogenic habitats, poorly-managed fields, pastures and meadows within their native range (Chmielewski and Semple 2001; Weber and Jakobs 2005). Solidago has a broader ecological niche in the invaded areas, also occupying drier and nutrient-poorer soils (Weber and Jakobs 2005). Both species have a perennial rhizome, which, in the spring, produces a cohort of shoots that start to flower in late summer (Solidago) or early autumn (Symphyotrichum) and yield numerous tiny wind-dispersed seeds. Jedlička and Prach (2006) noted the high viability of Symphyotrichum lanceolatum seeds, which, combined with the effective ability to penetrate established vegetation, triggers the high invasive potential of this species.

We initiated a pilot cultivation trial to examine the anatomy of haustorial connections between the hemiparasites and the two invasive hosts. The cultivation was set up in the experimental garden of the Department of Botany and Zoology at Masaryk University in Brno, Czech Republic. The hemiparasites’ seeds were collected from species-rich vegetation in the summer of 2019 (see Suppl. material 1: appendix S1 for localisations). In the autumn of 2019, we transplanted rhizomes of host species into 15 × 15 × 20 cm pots (narrower at the bottom, corresponding to 3.6 litres), filled with a mixture of peat and garden soil (ratio 1:3). In October, we sowed 20 hemiparasite seeds on each pot. We established five replicates for each hemiparasite-host combination.

In June 2020, we rinsed the hosts’ roots, harvested the haustoria and preserved them in 70% ethanol. Following the method of Soukup and Tylová (2014), we dehydrated the samples, transferred them to anhydrous butanol, infiltrated and then embedded them in paraffin. We prepared 12 µm sections using a sliding microtome (Reichert, Wien, Austria) and de-waxed and stained them with phloroglucinol-HCl (Wiesner solution) (Liljegren 2010) to colour the lignified cell walls.

We established the main pot experiment in autumn 2021 to investigate and quantify the outcome of the novel interactions for the hemiparasites and the extent of host suppression. For each hemiparasitic species, we used three seed sources: (i) seeds from a wild population growing in a species-rich grassland and (ii) seeds from plants originally obtained from the same populations as in (i), but which had been growing since 2019 in monostands of the two invasive host species. The aim was to investigate the potential selection of genotypes more adapted to the specific hosts. More specifically, the monostands were mown in the autumn of 2019, after which we sowed the hemiparasites’ seeds. In 2020, the monostands with hemiparasites were mown in July and October. We collected ripe hemiparasite seeds from all populations from June to July 2021. The seeds were stored at room temperature before use. As both host species produce a dense rhizome network in the topsoil layer, we collected soil blocks with rhizomes from monostands of each host species to establish host cultivation in September 2021. First, we removed the above-ground biomass and then cut approx. 12 × 12 cm rhizome blocks with a spade. The rhizomes were then inserted into the same pots and soil substrate described in the chapter ‘Haustorial connection’. See Suppl. material 1: appendix S1 for GPS coordinates of the sites of hemiparasites’ seed and host plants’ origin.

The experimental design comprised: (i) an uninfected control treatment (host species without hemiparasite seed addition) and three types of ‘infected’ treatments (with hemiparasite seeds addition), i.e. treatments (ii) ‘naïve’ (seeds of hemiparasites originating from a wild population), (iii) ‘home’ (seeds from hemiparasites growing for two years in a monostand of a host species and then sown with the same host species in the pot) and (iv) ‘cross’ (seeds from hemiparasites growing for two years in a monostand of one host species and then sown into the pot with the other host species) (see the scheme of the origin of hemiparasites’ seeds in Fig. 1). Both hemiparasites were sown with both invasive species, resulting in 14 treatments. Each treatment consisted of 10 replicates of the pots, totalling 140 pots. Each pot in the ‘infected’ treatments (treatments ii–iv) received 40 seeds of one of the hemiparasitic species. Seeds were spread on the surface and gently mixed with the topsoil layer. The pots were then placed in the experimental garden in Brno, following a completely randomised design and irrigated. During spring 2022, the pots were irrigated as necessary. In April 2022, seedlings of non-target species were removed from the pots. The pots were spaced 30 cm apart and their position within the experimental matrix was changed three times before harvest at the beginning of June 2022.

Figure 1.

Scheme of origin of the hemiparasites’ seeds used in the cultivation experiment. In October 2019, seeds of Melampyrum arvense and Rhinanthus alectorolophus from a single population per species, originating from a species-rich grassland, were sown in monostands of the host species Solidago gigantea and Symphyotrichum lanceolatum. By 2021, hemiparasite seeds collected from the host species’ monostands and the original hemiparasite population were used in the cultivation experiment resulting in three types of hemiparasite seed sources: ‘naïve’, ‘home’ and ‘cross’.

The experiment was harvested during hemiparasite flowering. We cut the above-ground biomass and counted the number of host shoots and hemiparasitic plants that survived in each pot. The hemiparasite and host biomass from each pot were dried separately at 80 °C and weighed. Schmid et al. (1995) revealed a strong dependency of sexual reproduction and clonal growth on plant size as well as a threshold size for sexual reproduction in Symphyotrichum lanceolatum and Solidago canadensis, a species closely related to Solidago gigantea. We thus expected the vegetative biomass to reflect host fitness and reproductive potential sufficiently.

Initially, we conducted an exploratory analysis of patterns in counts of hemiparasite individuals, host ramets and above-ground biomass production to identify pots that were not representative due to insufficient host or hemiparasite recruitment. Only pots with at least six host shoots and three hemiparasite individuals (in infected treatments) were subsequently included in the analysis (n = 132 out of 140 pots). Scatterplots of biomass vs. individual or shoot counts (Suppl. material 1: appendices S2, S3) demonstrated low correlations, indicating compensatory growth in pots retained for the analysis.

We used linear models to analyse the following parameters: hemiparasite above-ground biomass, the number of individuals, mean biomass per individual and host above-ground biomass, the number of shoots and mean biomass per shoot. All variables were log-transformed before analysis to improve the normality of residuals and homogeneity of variances. The analysis of each parameter, used as response variables, was conducted at two levels: (i) the species-level model included hemiparasite, host species and their interaction as predictors. Seed-source treatments were disregarded in this analysis; (ii) seed-source analysis consisted of a series of linear models, one for each host–hemiparasite combination, with seed-source treatment as a single predictor. In this analysis, we set treatment contrasts with the ‘naïve’ treatment as the baseline level, to which the two other treatments were compared. Only biomass data were tested in the seed-source level analysis.

We first built a saturated model for each analysis with all candidate predictors and interactions. Individual terms of the saturated models were tested by an F-test, the results of which are reported in ANOVA tables as in a classical two-way ANOVA with interactions. Non-significant (P > 0.05) terms were subsequently removed from the models in the backward predictor selection procedure. Non-significant main effects were retained if a predictor was involved in a significant interaction. The resulting minimal adequate models were then used to extract regression coefficients and their associated tests of significance. This approach was allowed by the nature of our data coming from a manipulative experiment with a balanced design, which implies the orthogonality of the predictors. We acknowledge that the orthogonality was not perfect because we removed a few pots with low establishment of hosts or parasites. Still, the collinearity between the tested effects (host and parasite predictors) was minimal as measured by the phi-coefficient (φ = 0.026; χ₁ = 0.0084; P = 0.927), which justifies the validity of the interaction-term testing and supports backward selection as a suitable model-selection approach. All analyses were performed in R, version 4.2.2 (R Core Team 2022).

Both hemiparasitic species formed fully developed haustoria on the roots and rhizomes of both host species. In all cases, the xylem bridge from hemiparasite haustoria reached the xylem vessels of the hosts. No signs of a defensive reaction by the hosts were observed (Fig. 2).

Figure 2.

Cross sections of haustorial connections between two root-hemiparasitic species and their hosts. In the hemiparasite haustoria (ha), there is a hyaline body (hb), the vascular core of the haustorium (vc) and a xylem bridge (xb) leading to host xylem vessels (xv) in the host root (hr); xx – xylem–xylem contact.

Hosts successfully resprouted from rhizomes in the transferred soil blocks; only four pots had to be omitted because of insufficient sprouting (Fig. 3, Suppl. material 1: appendix S4). The number of hemiparasite plants varied in the pots, but their establishment was generally successful, with only four pots omitted from the experiment due to poor hemiparasite establishment. On average, 10.9 Melampyrum plants were harvested in pots with both host species (max. 20 individuals). In contrast, significantly higher average numbers of Rhinanthus plants, 16.2 and 13.2, were harvested in pots with Solidago and Symphyotrichum (max. 23 individuals), respectively (Table 1, Fig. 4, Suppl. material 2 for the primary data). Hemiparasite biomass production differed between the two species and was also significantly affected by the host identity (Table 1). Specifically, Melampyrum grew larger than Rhinanthus (t₁₁₀ = 11.25, P < 10^-6) and Solidago supported a more vigorous hemiparasite growth than Symphyotrichum (t₁₁₀ = 10.12, P < 10^-6). These effects were additive, i.e. the difference in the host quality had a similar impact on both hemiparasitic species (Fig. 4). Similar trends and significant interactions were also found concerning the average biomass of hemiparasite individuals (Table 1). Melampyrum individuals were consistently larger than Rhinanthus and both hemiparasitic species produced larger specimens on Solidago than on Symphyotrichum. However, this trend was less pronounced in Rhinanthus, i.e. Rhinanthus individuals growing with Symphyotrichum were larger than expected by additive effects (t₁₀₉ = 2.57, P = 0.012; Fig. 4).

Figure 3.

Representative pots for each hemiparasite seed-source treatment (‘cross’, ‘home’, ‘naïve’) and the uninfected control. Solidago gigantea (left) and Symphyotrichum lanceolatum (right) are infected by Melampyrum arvense (top) or Rhinanthus alectorolophus (bottom). The bottom photo is flipped vertically for clarity of the experiment presentation. Photographic documentation of all experimental pots is provided in Suppl. material 1: appendix S4. Scale bars: 50 cm.

Figure 4.

Effects of host species on the total biomass, number of individuals per pot and average biomass of the individuals of the two hemiparasitic species. Boxplots represent median, quartiles and ranges. See Table 1 for the ANOVA tables summarising significance tests. Note the logarithmic scale of the y-axes.

Regarding host suppression, we identified strong interactive effects of host and hemiparasite species identities on the host biomass (Table 1). The hemiparasitic species significantly reduced host biomass relative to uninfected controls (Melampyrum: t₁₂₆ = -10.1, P < 10^-6, Rhinanthus: t₁₂₆ = -4.53, P < 10^-4), but the suppression was significantly more pronounced in Solidago infected by Melampyrum (t₁₂₆ = 4.50, P < 10^-5; Fig. 5). Overall, Solidago biomass was reduced by 77.6% and 49.1% on average when infected by Melampyrum and Rhinanthus, respectively. Symphyotrichum biomass was reduced by 31.6% and 35.2% on average by Melampyrum and Rhinanthus, respectively. Host biomass was reduced by decreasing the number of host shoots or reducing the average biomass of host shoots. While Melampyrum acted in both ways, Rhinanthus mainly decreased the average host shoot biomass (Fig. 5). In detail, Melampyrum reduced the number of host shoots per pot (t₁₂₈ = -4.05, P < 10^-4) by 33% in Solidago and 21% in Symphyotrichum. The effect of Rhinanthus on the host shoot number was not significant (t₁₂₈ = -0.76, P = 0.45). Both Melampyrum (t₁₂₆ = -7.07, P < 10^-6) and Rhinanthus (t₁₂₆ = -4.00, P < 0.001) reduced the average biomass of host shoots. While Rhinanthus reduced the average shoot biomass of both hosts to a similar extent, Melampyrum was significantly less deleterious to Symphyotrichum than to Solidago (t₁₂₆ = 3.98, P < 0.001).

Figure 5.

Effect of hemiparasite infection on total biomass, number of shoots per pot and average shoot biomass of the two host species. Boxplots represent median, quartiles and non-outlier ranges, with outliers displayed as points outside the non-outlier ranges. Note the logarithmic scale of the y-axes. See Table 1 for the ANOVA tables summarising significance tests.

We identified the significant effects of the hemiparasite seed-source treatments on some interactions. Total hemiparasite biomass was affected in the case of Melampyrum growing on Solidago (R² = 0.29, F_2,24 = 4.98, P = 0.016) and Rhinanthus growing on Symphyotrichum (R² = 0.32, F_2,25 = 5.81, P = 0.008). Specifically, Melampyrum plants in the ‘cross’ treatment (seeds from plants previously grown with the alternative invasive host) produced significantly less biomass (t₂₄ = -2.80, P = 0.010) compared to the ‘naïve’ treatment (seeds from species-rich vegetation), while the biomass of Melampyrum on Solidago from the ‘home’ (seeds from plants previously grown with the same host species) and ‘naïve’ treatment did not significantly differ (Fig. 6). Conversely, the biomass of Rhinanthus on Symphyotrichum was significantly higher in the ‘home’ treatment compared to the ‘naïve’ treatment (t₂₅ = 3.09, P = 0.005) and the hemiparasite biomass in the ‘cross’ and ‘naïve’ treatment did not differ (Fig. 6).

Figure 6.

Effect of seed-source treatments on hemiparasite biomass production categorised by the individual host–hemiparasite combinations. Boxplots represent median, quartiles and non-outlier ranges, with outliers displayed as points outside the non-outlier ranges. P-values indicate significant effects of seed-source treatments. Note the logarithmic scale of the y-axes.

Host biomass was significantly affected only in the case of Solidago infected by Rhinanthus (R² = 0.27, F_2,27 = 5.09, P = 0.013) (Fig. 7). Here, Rhinanthus of ‘home’ and ‘cross’ treatments suppressed Solidago biomass more than ‘naïve’ Rhinanthus plants (t₂₇ = -2.73, P = 0.011 and t₂₇ = -2.80, P = 0.009 for ‘home’ and ‘cross’ treatments, respectively).

Figure 7.

Effect of seed-source treatments on host biomass production in infected pots categorised by the individual host–hemiparasite combinations. Boxplots represent median, quartiles and non-outlier ranges, with outliers displayed as points outside the non-outlier ranges. P-values indicate significant effects of seed-source treatments. Note the logarithmic scale of the y-axes.