Plant Material
Seeds from naturally occurring, wild tall fescue populations from three areas that were separated approx. 500 km from each other by the Baltic Sea were collected in August 2003. We collected seeds from 8, 9, and 6 populations, respectively, from the island of Åland (A), the island of Gotland (G), and the West Coast of Sweden (S), 10–50 individual plants from each population. Three seeds from each individual plant were stained (Saha et al. 1988), and endophyte infections were determined by microscopic examination. All the tall fescue populations in the study had seed borne Epichloë sp., and infection frequencies varied between 85 and 100 %. We combined all E- (uninfected) seeds from populations within each area (A, G, and S) and all E+ (infected) seeds from populations within each area. In addition to plants from these natural tall fescue populations, both infected and uninfected ‘Kentuky-31’ (KY31) cultivar seeds were obtained from the University of Kentucky. Half of all the seeds were shipped from Finland to Kentucky or vice versa, so that both study sites had identical sets of seeds. At both locations, tall fescue seeds were germinated on moist tissue paper in Petri-dishes and planted in individual pots in a greenhouse. When plants had 2–3 tillers, they were transplanted to field plots.
Common Garden Experiments
Parallel field experiments were carried out in the fields of Turku Botanical Garden, University of Turku (60°26′0″N, 22°10′19″E), Finland and the University of Kentucky‘s experimental Eden Shale Farm (38°32′22″N, 84°44′24″W), Kentucky, USA. The field site in Finland is at the edge of the northern distribution range of natural tall fescue populations, while the Kentucky field site is in the zone of intensive tall fescue cultivation in the USA, also known as the ‘fescue belt’. Both experimental field sites had been in cultivation in the past and were tilled without nutrient application in the summer 2004, immediately preceding experimental establishment. The experimental areas were fenced to prevent large vertebrates (e.g., rabbits, deer) access; however, smaller vertebrates (e.g., voles) and invertebrates were able to move freely into the experimental area and among plants.
In the control plots, pH of the experimental soils were 6.7 and 5.9, total nitrogen 0.15 % and 0.14 %, phosphorus 7 and 14 mg/kg, potassium 132 and 81 mg, calcium 1800 and 2080 mg/kg, and magnesium 208 and 120 mg/kg for Finland and Kentucky sites, respectively.
Both experimental set-ups were randomized block designs consisting of 10 blocks, each divided to 4 plots. In August 2004, E+ plants from four origins (A, G, S, KY31) were randomly planted in each experimental plot about 0.5 m apart and 0.5 m from the edge of the plot. The space between the experimental plants was either mowed, hand weeded, or sprayed with herbicide two times during the growing season to prevent interspecific competition. E- plants of the same origin also were planted in the plots, but were not included in this study, because uninfected plants are not capable of producing alkaloids.
The four plots in each block were designated randomly to one of the four treatments: Control (C) - receiving only ambient rainfall and nutrients, Water treatment (W) - with 3 L of water applied to each plant separately three times a week from June to August in Finland, and in Kentucky twice per week at 3.8 L per plant at each watering from April to October using trickle emitters and an irrigation timer. In Finland, nutrient treatment (N) was 1 dl of N-P-K-fertilizer (Nurmen Y2, Kemira KnowHow) (N-P-K/20–6-6) applied two times during the growing season for each plant. In Kentucky fertilization consisted of 50 kg/ha of N/per application in form of urea. Water and nutrient treatments (WN) combined both water and nutrient applications. These experiments allowed us to compare the effects of water and nutrient treatments within the experiments in Finland and Kentucky. However, comparisons between the two experiments should be done with caution because the sites inevitably differ in terms of numerous abiotic and biotic factors.
In 2005, all the experimental plants were re-checked to verify their endophyte status. In July, we sampled a pseudostem from each plant for immunoblot assay detection using monoclonal antibodies specific to Epichloë (Phytoscreen Immunoplot Kit #ENDO7973, Agrinostics, Watkinsville, GA, USA). In addition, at the end of the summer, three seeds from each plant were stained (Saha et al. 1988), and endophytic infections were checked by microscopic examination. All plants included in this analysis were infected with Epichloë (N = 40, 38, 40, 34 and N = 35, 38, 40, 38 for A, G, S, and KY31 origins in Finland and Kentucky experiments, respectively).
The phenological timing of the two experiments is difficult to compare, and therefore we decided to sample regrowth leaf biomass of the experimental plants. The samples were collected from spring regrowth and fall regrowth in Finland and Kentucky, respectively. We obtained samples from all the 152 E+ plants in the Finnish experiment in mid-June 2006. A handful of green growth was cut at the height of 5 cm from the ground, placed into a plastic bag, and transported immediately to a deep-freezer (−80 C°). The samples were ground, and the ground plant material from Finland was sent to Kentucky for alkaloid analyses. In Kentucky, 151 E+ plants were sampled in early November 2006. The samples from both experiments were analyzed using the same protocol.
Alkaloid Analyses
Pyrrolizidine alkaloids [N-acetylloline (NAL), N-formylloline (NFL), and N-acetylnorloline (NANL)] were extracted from powdered plant material with ethanol:methylene chloride (4:1, v/v) containing the internal standard quinoline and sodium bicarbonate following protocol of Blankenship et al. (2001). Individual alkaloids were resolved and quantified by gas chromatography equipped with FID detector. Chromatographic conditions were 15 m × 0.53 mm DB5 column with initial oven temperature of 70 °C increased to 160 °C at 45 min-1, held for 5 min, and increased to 290 °C at 45 min-1 and held for 7 min.
Ergovaline, ergovalinine, and lysergic acid were measured by a modification of the procedure of Yates and Powell (1988). The HPLC was equipped with a fluorescence detector with excitation at 310 nm and measurement at 420 nm. Separation was on an Alltech Alltima C18 150 × 4.6 mm column with 3 μ particle size. Elution solutions were (A) 75 mM ammonium acetate in water:acetonitrile (3:1, v/v) and (B) acetonitrile. Elution gradient was 95:5 (A:B) for 1 min; linear change to 60:40 (A:B) during the next 15 min and maintained for 5 min; changed to 0:100 (A:B) during the next 1.5 min and maintained for 5 min; changed to 100:0 (A:B) during the next 1 min and maintained for 6 min before returning to 95:5.
Statistical Analyses
Statistical analyses were performed using R 2.15.2 statistical software (R Core Team 2012) by applying linear mixed models (Pinheiro et al. 2013). The experiments conducted in Finland and Kentucky were analyzed separately because the sites inevitably differ in terms of numerous abiotic and biotic factors.
Total ergot alkaloids and lysergic acids were used as the final response variables, as results based on individual compounds did not differ from models based on their sums. Lolines were analyzed as sums of N-acetylloline (NAL), N-formylloline (NFL), and N-acetylnorloline (NANL). Normality of the residuals was gained after transformation as suggested by the Box-Cox analysis. All ergot alkaloids were log transformed (λ = 0), and to take into account zero values, one was first added to each compound before transformation. Lysergic acids were square root transformed (λ = 0.5). There was no need for transformation in lolines.
The statistical model for alkaloids included the fixed factors fertilization (N) and water treatment (W) treatments (included in models as, “N”, nitrogen and nitrogen + water combined, “W”, control and water treatment combined, and interaction, “N*W” between N and W), plant origin (KY31, A, G, S) and interactions between treatments and origin. Block was used as a random factor in all models. Association between different compounds within treatments (N and W) and origins were analyzed in separate linear mixed models (i.e., comparison of ergot totals vs. loline totals in nitrogen and control treatments for each origin).
The full model with treatments, origins, and their interactions was fitted, and the non-significant factors were removed. Out of these models, we constructed the final model by including all factors that were significant at least in one of the experiments. This model is shown for experiments conducted in each site. Interaction plots were used in illustrating the significant interactions between the factors left in the models.