In our study we were able to demonstrate indirect positive effects of microbial plant symbionts on a vertebrate herbivore via reduced predation. Similar effects have been previously demonstrated with invertebrate predators and parasites as natural enemies in food webs where herbivores feed on endophyte infected plants [14], [15], [20]. In cases where species can tolerate mycotoxins produced by the endophyte and are less at risk of predation due to endophyte consumption, the net effect of the endophyte on the host grass will be negative. Therefore the traditional view of endophytes as defensive plant mutualists may not hold if a third trophic level is included. Thus, our results provide evidence that the nature of the relationship between grass endophytes and their hosts may range from mutualism to parasitism depending on the complexity of the food web.

All procedures involving voles were carried out in accordance with the Act on the Use of Animals for Experimental Purposes established by the Ministry of Agriculture and Forestry, Finland. The study was approved and supervised by the Animal Experiment Committee of Finland (License number: STH393A).

Meadow ryegrass, the experimental plant species, is one of the most important forage grasses in Finland. It is a native grass species in Europe which occurs commonly outside of agronomic use in meadows, roadsides and wastelands in Finland [21]. Several widely used meadow ryegrass cultivars in Finland are commonly infected Neotyphodium uncinatum endophyte [22], which grows systemically in all parts of the host plant. N. uncinatum produces lolines which may cause variable responses in invertebrates and small vertebrates [7], [11] but the loline appears to be non-toxic to large mammal herbivores [23].

The sibling vole is a common and widely distributed species in southern and western Finland. The individuals used in the experiment were laboratory-born individuals, whose parents were trapped from natural populations in nearby fields of MTT Agrifood Research Finland, Jokioinen (60° 48′ 15″ N, 23° 29′ 10″ E), in autumn 2006. Prior to the experiment, voles were housed in ca. 60×40×40 cm³ cages (3–5 same sex individuals in a cage) and provided with ad libitum potatoes, water and twice a week with endophyte free fresh grass cut from the wild. Bedding was provided in the form of wood shavings and hay. Temperature in the laboratory was ca. 20°C and photoperiod 16 h light ∶ 8 h dark.

The least weasel is a common specialist predator of voles and their single most important source of mortality in natural populations [24], [25]. The least weasels used in the experiment were either first generation laboratory-born individuals from the Konnevesi Research Station of the University of Jyväskylä, Finland, or trapped from the wild but kept under similar conditions like lab-born ones for weeks before the experiment. Prior to the predation and olfactory experiments, the least weasels were housed individually in 60×80×60 cm³ cages in an outdoor shelter and provided with a rooster chick per day with one fasting day a week, and occasionally voles of the genera Microtus and Myodes. Bedding was provided in the form of wood shavings and hay.

The experimental field was established to study the importance of endophytes on the population development of voles. Seeds (cultivar ‘Kasper’) were obtained in various seed lots from seed production farms via the Plant Production Inspection Centre, Loimaa, Finland. As samples, we stained 50 seeds per seed lot and examined them microscopically for endophyte status [26]. We chose two seed lots for the experiment: one uninfected (E−, 0% endophyte frequency) and the other infected (E+, 79% endophyte frequency). We sowed E+ and E− seeds in a field in five plot pairs (each plot 39×25 m²) so that we randomized E+ and E− treatments separately within each pair. The field was established in Jokioinen in May 2006. Each plot was surrounded with a sheet metal fence in order to keep the experimental voles inside and voles of natural populations and small mammal predators out of the experimental areas. The sheet metal was embedded 60 cm below ground while 60 cm remained above ground. Before sowing, the field was fertilized with cow manure (30 000 kg/ha) and again in June 2007 with a commercial fertilizer [16∶9∶22 (N ∶ P ∶ K) with micronutrients, Kemira, product number: 0647334]. All other plants except for meadow ryegrass were regularly rooted up from the field during the experiment.

Endophyte infection status of nine plant individuals per field plot was verified before the experiment using immunoplot assay to detect monoclonal antibodies specific to Neotyphodium (Phytoscreen Immunoplot Kit #ENDO7973, Agrinostics, Watkinsville, Georgia, USA). Alkaloid extracts of the plants were analysed [27], [28] and E+ plants were detected to harbor active endophyte infections producing loline alkaloids.

Vole individuals (50 males and 50 females) were randomly selected from a larger pool of males and females, all of which were sexually mature and had a body mass of >20 g. Individuals were randomly assigned either to an E+ or an E− treatment. Thereafter, in August, five male and five female voles were released into each of the ten enclosures. Four and a half months later, which approximates the annual length of the reproductive period of voles in Finland, vole population sizes were estimated using Ugglan multiple live capture traps (Grahnab, Sweden). Five traps were placed in each enclosure under plywood shelter boxes to reduce exposure to the elements. The traps were baited with carrots and checked twice a day for three days. Population size was estimated for each enclosure with the minimum number alive –method [29].

Voles (72 individuals) were selected from a larger pool of males and females, all of which had a body mass of >16 g, and housed singly in cages. The selected voles were assigned into 36 pairs (15 female and 21 male pairs) based on similarity in body mass. At this stage, voles were provided ad libitum potato and water. On the following day, we assigned the vole individuals randomly either to an E+ or an E− diet treatment within each pair and recorded their body mass to the nearest 0.1 g with electronic scales. The mean differences in the body mass (body mass_E− − body mass_E+) of females and males (and standard deviations of the differences) were 0.01 g (0.73 g) and 0.03 g (0.44 g), respectively. After this, we removed potatoes from the cages and provided the corresponding experimental diets, namely ad libitum fresh meadow ryegrass three times a day cut from respective E+ or E− plots in the field (see “Vole population size”). The body mass of the voles was again recorded when the voles had been maintained on the experimental diet for seven days.

Prior to the predation experiment six female and 18 male vole pairs (one fed on E+ grass and one fed on E− grass within each pair) were fed with the experimental diets as in the body mass bioassay for 7–30 days. The length of the feeding period varied between vole pairs due to logistical reasons but was recorded for use a covariate in the analyses.

The experiment was conducted at the Konnevesi Research Station of the University of Jyväskylä (62° 37′ 40″ N, 26° 17′ 15″ E), Finland, in an experimental enclosure (10×10 m²) on a field naturally vegetated by meadow plant species (the field did not include meadow ryegrass). The enclosure was divided into six sectors separated by ca. 30 cm wide short-grass belts (cut 2 cm above ground surface) to enable monitoring of vole mobility. One experimental trial consisted of exposing both a vole fed on E+ and a vole fed on E− grass from the same feeding pair to least weasel predation in the enclosure. Thus, a replicate consisted of a pair of voles that were both the same sex, similar in weight in the beginning of the experiment and had been on the experimental diet for the same period of time. The voles were marked with fiber strips of different colors (15 cm long, 2 cm wide) bound around the pelvis, where it does not hinder movements, and released into the cut middle belt between the sectors of the enclosure. We randomized the strip color between voles fed on E+ and E− grass in every replicate. Simultaneously with releasing voles, we placed a least weasel in an Ugglan live trap in the same middle belt where the voles were released. The least weasel was released from the trap five minutes after releasing the voles. We observed and recorded the number of belt crossings of the voles in the enclosure until one of the voles was captured by a least weasel. A single trial lasted from 15 min to 9 hours. Seven out of the 24 trials had to be terminated for external reasons (darkness at night, heavy rain etc.) before the weasel had captured a vole. Different voles and least weasels were used in every replicate. After each trial we trapped the surviving vole (or both voles if a replicate had to be terminated) and removed them from the enclosure back to the laboratory.

We studied whether least weasels, that are known to be olfactory hunters [19], distinguish and prefer olfactory cues of voles fed with E+ grass from voles fed with E− grass. As a source of odour we used urine- and faeces-soaked vole bedding material from cages of voles that had been feeding on either E− or E+ grass for the weasel predation experiments (see above). The collected bedding material was stored in air-tight plastic bags at −22°C for ca. three months. The bedding was thawed in the bags at room temperature before the experiment.

The vole odour experiment was carried out in a Y-maze arena [19], which consisted of three transparent Perspex (Perspex, Rotterdam, The Neatherlands) plastic tubes of 80 mm inside diameter and 80, 60 and 60 cm length, forming a Y. The weasel entered the 80 cm long tube and came to a bifurcation of ∼60°. There it had to choose to continue into either of the 60 cm tubes until it reached a target “nest box” containing vole odour at the end of the tube. The nest boxes were small laboratory cages (25×10×10 cm³) covered with a Perspex roof. The ends of the Perspex tubes were separated from the nest box by a Perspex door with 12 holes of 4-mm diameter to allow airflow.

In each nest box we placed a 2×10×10 cm³ wire mesh basket filled with vole bedding from either the E+ or the E− treatment. We randomized E+ and E− bedding from the same vole feeding pair to either end of the Y-maze prior each trial. Thus, we collected the E+ and E− bedding in each replicate from voles that had been on the experimental diet for the same amount of time. The entrance to the tube was a wooden box, in which we acclimatized the weasel for five minutes prior to the trial. The box was separated from the tube by a Perspex door that could be opened from outside the experimentation room with a monofilament line. The door had holes to allow airflow from the arena to the weasel box during the acclimatization period. Above the arena we mounted infrared light sources and a camera that was connected to a monitor in an adjacent room where the behavior of the weasel was monitored on screen. Altogether, we tested nine female and 12 male weasels. All weasels slowly approached the tube and selected one of the tubes at the bifurcation. The test ended when the weasel reached the end of either branch of the Y-tube and sniffed the holes at the door separating the tube from the nest box. After each trial we cleaned all parts of the arena with water and ca. 50% ethanol.

The response variables used in the statistical analyses and other details of the experiments are summarized in Table 2.

We analyzed the body mass of voles separately for females and males because variation in male body mass was higher than in females. The E+ and E− diet treatments were compared using a paired t-test and 95% confidence interval (CI) for the mean difference [30]. The analyses were performed by the MIXED procedure in version 9.1.3 of the SAS/STAT software. The rest of the data were analyzed through exact statistical methods because the data were small and non-normally distributed. We based the statistical analyses for least weasel predation on logistic regression models for binary data. Let if E+ vole was captured and if E− vole was captured in pair . Each binary outcome variable is assumed to have a Bernoulli distribution with parameter , where is the probability of E+ vole being captured in a pair . The value indicates an equal probability of capture for E+ and E− voles. When examining whether least weasels prefer E+ voles over E− voles as prey, the logistic regression model had the following form:(1)where is a constant and the ratio is the odds of E+ vole being captured in a pair . On fitting the model to the data, an estimate of , , is obtained, and the estimated capture probability (proportion) . The 95% confidence limits for were computed similarly from the estimated limits of . When examining the dependence of on the values of each potential explanatory variable (one at a time), the model was of the following form:(2)where and were the unknown parameters estimated by the data. We fitted the models by using the approach of conditional exact inference [31], and performed the analyses with version 8 of the LogXact software. The analysis of the binary choice data for vole odour was based on the corresponding model than the model (1) above. We tested the difference in the amount of captured voles per enclosure between E+ and E− grasses and also the difference in the belt crossings of voles/hour by using the exact Wilcoxon signed ranks test, and estimated the median difference and its 95% confidence interval (CI) by the Hodges-Lehmann procedure [32]. We performed the analysis with version 8 of the StatXact software.