Herb-Drug Interactions

The seedlings of susceptible rye cultivar Słowiańskie were inoculated with Prs single spore isolate No. 1.1.6. Leaf samples were collected at 4, 8, 12, 16, 20, 24, 36, 48, and 72 hpi and stained with calcofluor white as described by Orczyk et al. [46 (link)]. Briefly, samples were fixed for 24 h with an ethanol:dichloromethane (3:1) solution supplemented with 0.15% trichloroacetic acid, after which they were rinsed twice with 50% ethanol, twice with 0.05 M sodium hydroxide, three times with water, and once with 0.1 M Tris. They were then stained with calcofluor white (3.5 mg/ml).
The stained leaf fragments were examined with the Diaphot fluorescence microscope (Nikon) for the presence of germinating spores, HMCs, and micronecrosis symptoms. Additionally, the number of infection sites was calculated. Observations were made in 80 on average (but not less than 30) infection sites per leaf sample. For selecting time-points for additional analyses of plant–pathogen interactions, the germinating spores and appressoria at infection sites were counted. The following four infection site profiles were used to reflect plant–pathogen interactions: (i) appressoria; (ii) appressoria and HMCs; (iii) appressoria, HMCs, and micronecrosis; and (iv) appressoria and micronecrosis. The analysis of pathogenesis and the percentage of profiles in the preliminary experiment with cv. Słowiańskie (Fig 1) were calculated according to: Eqs 1, 2, 3 and 4. The analysis of pathogenesis and the percentage of profiles in the main experiment with lines L318, D33 and D39 (Fig 2) were calculated according to: Eqs 5, 6, 7 and 8.
$\text{Percentage}\text{of}\text{profile}\text{i}=\frac{n_i}{(n_{gs}+n_i+n_{ii}+n_{iii}+n_{iv})}\times 100\%$
$\text{Percentage}\text{of}\text{profile}\text{ii}=\frac{n_{ii}}{(n_{gs}+n_i+n_{ii}+n_{iii}+n_{iv})}\times 100\%$
$\text{Percentage}\text{of}\text{profile}\text{iii}=\frac{n_{iii}}{(n_{gs}+n_i+n_{ii}+n_{iii}+n_{iv})}\times 100\%$
$\text{Percentage}\text{of}\text{profile}\text{iv}=\frac{n_{iv}}{(n_{gs}+n_i+n_{ii}+n_{iii}+n_{iv})}\times 100\%$
$\text{Percentage}\text{of}\text{profile}\text{i}=\frac{n_i}{(n_i+n_{ii}+n_{iii}+n_{iv})}\times 100\%$
$\text{Percentage}\text{of}\text{profile}\text{ii}=\frac{n_{ii}}{(n_i+n_{ii}+n_{iii}+n_{iv})}\times 100\%$
$\text{Percentage}\text{of}\text{profile}\text{iii}=\frac{n_{iii}}{(n_i+n_{ii}+n_{iii}+n_{iv})}\times 100\%$
$\text{Percentage}\text{of}\text{profile}\text{iv}=\frac{n_{iv}}{(n_i+n_{ii}+n_{iii}+n_{iv})}\times 100\%$
where: n_gs−number of infection sites with germinating spores, n_i−number of infection sites with appressoria, n_ii−number of infection sites with appresoria and HMC, n_iii−number of infection sites with appresoria, HMC and micronecrosis, n_iv−number of infection sites with appresoria and micronecrosis.
The number of infection sites with germinating spores and the rates of the four profiles were used to select the following time-points for further analyses of plant–pathogen interactions: 8, 17, 24, and 48 hpi. The first time-point (8 hpi) was selected because of the considerable abundance of appressoria at established infection sites. The 17 and 24 hpi time-points were associated with HMC formation and intense pathogen growth, respectively. The final time-point (48 hpi) corresponded to the beginning of the resistance reaction. Leaf samples were collected at the selected time-points, stained, and analysed regarding plant–pathogen interaction profiles.

Generalized linear models (GLMs) were used to determine the best predictors of community plant height, total cover, plant–plant interactions, and PLR. For community plant height and total cover, the predictors were aridity and soil variables (SOC, pH and sand content); for plant–plant interactions, the predictors increased community plant height and total cover; for PLR, all variables were involved in the prediction. Because plant–plant interactions are binary variables, we used GLMs with a family of binomial distributions. In order to obtain the best models, a model selection procedure was applied based on minimizing the Akaike's Information Criterion corrected for small sample size (AICc, Burnham & Anderson, 2002 ). The averaged coefficients of predictors were computed by averaging models with ΔAICc (difference from the minimum AICc) less than two units. As none of the two‐way interactions between environmental conditions (climate and soil) and biological factors (community plant height, total cover and plant–plant interactions) significantly affected PLR, we removed all the interactions and only considered the major effects. This procedure was performed by the package MuMIn in R 4.0.2 (Bartoń, 2022 ).
A structural equation model (SEM) was used to explore the direct and indirect effects of environmental conditions and biological factors on PSDs. We established a priori model (Supporting information Appendix E) based on the previous literatures (Supporting information Appendix F). This model included: (i) a direct effect of environmental conditions, community plant height, and plant–plant interactions on PLR; (ii) a direct effect of environmental conditions on community plant height, total cover, and plant–plant interactions; and (iii) the influence of community plant height and total cover on plant–plant interactions. Based on the best model selection process described above, the potential pathways for the environmental and biological factors to drive PSDs were explored, thereby supporting our a priori model and benefiting the simplification of SEM (Carvajal et al., 2022 (link); García‐Palacios et al., 2018 (link); Wang et al., 2022 (link)). We included only the best predictors and removed the hypothesized pathways that were not effective. The SEM was bult by the piecewiseSEM package in R 4.0.2 (Lefcheck, 2016 (link)). To assess the goodness‐of‐fit of SEM, Fisher's C statistic was used, with significant values (p < .05) indicating that the model cannot fit the data. The standardized total effect of each explanatory variable was also calculated in order to demonstrate the total impact of each variable.

We employed the mark correlation function calculated as the normalized mark pair density to infer plant–plant interactions at short distances (<5 m, Getzin et al., 2008 (link); Illian et al., 2007 (link)). The patch sizes were attached to the patch locations as quantitative marks. The test function of mark pair density is pi×pj, where pi and pj are the patch sizes of two points i and j that are distance r apart. The normalization is squared mean patch size per plot and pair density at distance r. Thus, the quantitative marked correlation function focuses on the correlation between two patch sizes at a certain distance, while separating it from the effect of patch locations (Law et al., 2009 (link)). The null model is independent marking, which assumes that patch sizes are randomly distributed between the locations. We performed 999 simulations of the null model and took the 25 highest and lowest values of the function as envelopes with an error rate of 0.05 (Wiegand et al., 2013 (link)). We also used Goodness‐of‐fit to test the probability of type I error (Grabarnik et al., 2011 (link)), a departure of empirical observation from the null model was accepted once it reached significant level (p < .05).
With flat terrain and small observed plots, we assumed a homogeneous environment within a plot. Therefore, the significant departure was only considered as evidence of plant–plant interactions (Stoyan & Penttinen, 2000 (link); Velázquez et al., 2016 (link)). At short distances, if the empirical observations are significantly greater than the simulated envelopes, this is interpreted as plants developing well in size due to facilitation (e.g., microclimate amelioration by above‐ground biomass, Trautz et al., 2017 (link)). If the empirical observations are significantly smaller, indicating plants may be limited in sizes due to competition for water or light by neighboring patches (Getzin et al., 2008 (link); Schenk & Jackson, 2002 (link)). We used two binary indicators to represent the presence of facilitation and competition (or absence) per plot, respectively. The analysis was carried out in Programita software (Wiegand & Moloney, 2014 (link)).