Complex Interactions among Sheep, Insects, Grass, and Fungi in a Simple New Zealand Grazing System

Epichloë fungi (Ascomycota) live within aboveground tissues of grasses and can have important implications for natural and managed ecosystems through production of alkaloids. Nonetheless, vertebrate herbivores may possess traits, like oral secretions, that mitigate effects of alkaloids. We tested if sheep saliva mitigates effects of Epichloë alkaloids on a beetle pest of perennial ryegrass (Lolium perenne L.) in a New Zealand pasture setting. Plants with one of several fungal isolates were clipped with scissors, grazed by sheep, or clipped with sheep saliva applied to cut ends of stems. We then assessed feeding damage by Argentine stem weevils on blade segments collected from experimental plants. We found that clipping plants induced synthesis of an alkaloid that reduces feeding by beetles and that sheep saliva mitigates this effect. Unexpectedly, the alkaloid (perloline) that explains variation in beetle feeding is one produced not by the endophyte, but rather by the plant. Yet, these effects depended upon fungal isolate. Such indirect, complex interactions may be much more common in both managed and natural grassland systems than typically thought and could have implications for managing grazing systems.


Introduction
Epichloë endophytes are clavicipitacean fungi that infect most cool season grasses (Rodriguez et al. 2009) and can have important implications for grazing agricultural systems, which is the focus of our study. The fungi live intercellularly within host shoot tissue and often provide fitness benefits to hosts, particularly defense against herbivores (Clay 1988). The mechanism of herbivore defense is production of various noxious alkaloids; ergot alkaloids, indole diterpenes (including lolitrems and janthitrems), lolines and peramine being the most common (Schardl et al. 2013). Ergot alkaloids and lolitrems primarily affect vertebrates, while lolines and peramine affect invertebrates (Schardl et al. 2013). Thus, the interaction between Epichloë endophytes and grasses has been termed a defensive mutualism (Clay 1988). Many Epichloë endophytes produce some or all of the alkaloids constitutively, that is, continuously whether herbivory is occurring or not (Clay 1988). In addition, at least some Epichloë-produced alkaloids are wound-induced, as was shown for loline in tall (Schedonorous arundinaceus) and meadow (Lolium pratense) fescue (Bultman et al. 2004;Sullivan et al. 2006;Zhang et al. 2009), and for peramine and lolitrem B in perennial ryegrass (Fuchs et al. 2017).
Alkaloid production by Epichloë endophytes reduces the need for synthetic pesticides in grazing and turf systems and thus, the Epichloë-grass interaction will play an increasingly important role globally in integrated pest management strategies (Kauppinen et al. 2016). Sustainable practices are especially important in grasslands, which make up 70% of the world's agricultural area (Ramankuttry et al. 2008). Development of forage grasses with Epichloë strains (i.e., so called Bselected^endophytes) that provide pest resistance but lack negative effects on livestock has progressed in the last two decades, particularly in the USA (Bouton et al. 2002), Australia and New Zealand (Johnson et al. 2013). These isolates, which produce no or low levels of alkaloids affecting vertebrates, were collected from grasses living in their native range and then inoculated into specific cultivars of the same host species, to form novel endophyte-grass associations.
Even though endophytes can provide defense against herbivores (Clay 1988), these animals may possess traits that circumvent the defensive mutualism. For example, evolutionary pressures would favor herbivores that mitigate effects of alkaloids through oral secretions that have anti-fungal activity (Bellamy et al. 1994). Indeed, some evidence suggests that saliva of moose (Alces alces) can reduce ergot alkaloid production by Epichloë-infected red fescue (Tanentzap et al. 2014). Here we test if sheep (Ovis aries) saliva mitigates effects of Epichloë alkaloids and how this might affect the mutualism between fungus and plant. Our study is set within a pasture agroecosystem; an agricultural system that occupies 30% of the planet's arable land surface (Steinfeld et al. 2006). It typically has a fairly simplistic food web (Goldson et al. 2014), characterized by far fewer species than found in natural grasslands and thus, should have fewer direct and indirect interactions than its natural counterparts (Strauss 2013).
Paddocks for sheep grazing in New Zealand are often dominated by just two plant species: perennial ryegrass and white clover (Trifolium repens) (Daley 1990). A relatively small set of pest insects occurs; one of the most damaging and common of which is the Argentine stem weevil (hereafter weevil) (Listronotus bonariensis) (Popay et al. 2011). We tested the hypotheses that: 1) the fungal isolate affects weevil preference, 2) simulated sheep grazing induces alkaloid production, 3) alkaloid (specifically lolines and peramine) induction reduces weevil preference, and 4) sheep saliva mitigates alkaloid production and therefore modulates weevil preference.

Materials and Methods
Plant Care Perennial ryegrass seeds (cultivar Grasslands Samson obtained from Margot Forde Germplasm Centre, Palmerston North, New Zealand) were sown into 14 cm dia pots in December 2014. Seeds were infected with one of three isolates of Epichloë festucae var. lolii: common strain (CS), AR1 or NEA2 (AR104), or were uninfected (Nil). Isolates AR1 and NEA2 are endophytes that produce no (AR1) or reduced (NEA2) ergot and lolitrem alkaloids (Rupport et al. 2017). Protocols were employed to ensure plants in all isolate groups were nearly identical genetically (see below). We established two seedlings within pots that were filled with a 50:50 mixture of potting peat and Wakanui silt loam soil. The peat was mixed with washed crusher dust and fertilisers (calcium nitrate dolomite, lime, Osmocote, sulphate of potash, superphosphate along with zeolite) and added to the soil. Plants were housed in a shadehouse and watered as needed. We used one plant per pot in our study; the second plant in pots was used for a related but separate experiment.
Inoculations to produce Bselected^endophyte infected plants carries with it the possibility of narrowing the genetic diversity of a cultivar, or selecting a genetic subset of the cultivar. For this reason, we used 35-50 inoculated plants for AR1 and NEA2 isolates. The CS strain is the one that typically infects Samson perennial ryegrass. Genetic testing across a range of CS endophytes in New Zealand showed they were genetically the same at SSR sites (Simpson et al. 2012). In addition, for all inoculations we used the same base uninfected line of Samson plants. Further, when we multiplied the seed we interpollinated it with other plants from the same cultivar to dilute any narrowing/change in genetics. Thus, genetic differences in plants among isolate lines were minimized. Assessment of these procedures in several different cultivars using vegetative and reproductive plant growth characteristics, like seed production found no evidence of genetic change between lines (Easton 2007).

Sheep Saliva Collection
We collected saliva from twenty-five 2-tooth ewes (Coopworth cross) that had routinely grazed paddocks of endophyte-infected perennial ryegrass and white clover. Cotton disks and swabs were held inside their cheeks for ca. 1 min and then removed and the absorbed saliva squeezed into a vial. This was done over a 2 day period. Samples were centrifuged at 2000 rpm for 2 min, resulting in 21 mL of saliva. To obtain adequate volume for application to plants, we diluted saliva (1:5.3) with 0.98% saline.
Damage Treatments Seven weeks after seeds were sown, plants were treated in one of four ways. A quarter of the plants were clipped at 5 cm above the soil surface and had 0.98% saline applied to cut ends with a paint brush (ca. 300 ul applied to each plant). A second group was clipped and had the sheep saliva solution (above) applied to cut ends with a paint brush (300 ul applied to each plant). A third group of plants was grazed by sheep. This was accomplished by sinking potted plants into holes dug in the soil in a paddock of perennial ryegrass at the AgResearch research farm at Lincoln, New Zealand (S43.6318, E172.4703). Plants were placed in a block design with each block containing four replicates of each isolate type. Irrigation was supplied as needed. Plants were moved to the paddock 5 days prior to sheep grazing. Metal grating was placed over pots (with shoots of plants emerging through grates) to prohibit sheep from pulling whole plants from the ground. Twenty ewes (same as used for saliva collection above) grazed plants down to ca. 5 cm from the soil surface in about 45 min. A fourth group of 20 pots served as controls and were not damaged.
Insect Bioassay One 4 cm section of the newest fully expanded leaf blade from the largest tiller was collected from one plant within each pot for all treatment groups 2 weeks after damage to plants. Blade sections were placed individually into 90 mm dia petri dishes with filter paper and 1 mL distilled water. Five weevil adults (starved for 8 h prior to the experiment) were placed into each petri dish and dishes were wrapped in parafilm to retain moisture. This was a no-choice design. Petri dishes were placed in an environmental chamber at 20 o C and a 14:10 L:D cycle. Percent damage to individual leaf blade sections was visually scored in 10% increments from 0 to 10% to 90-100% after 24 and 48 h. At 24 h many blades had less than 10% feeding. To distinguish between blades with no feeding from those with small amounts of feeding we counted the number of feeding scars by weevils and added this to the percent damage score in the following manner: 0 scars = 0 added to damage score; 1-3 scars = 0.1 added; 4-7 scars = 0.2 added; 8-10 scars = 0.3 added. Adult weevils can fly and are able to select host plants for feeding (Goldson 1982); thus, we feel our first measurement of feeding damage (at 24 h) is most valuable (nonetheless, results at 48 h were similar to those at 24 h, see Fig. S1) as plants that receive little or no initial feeding would likely be ones weevils would move from in the field while searching for suitable hosts. Beetles did not have this option in our assays and could only feed on the plant tissue in their petri dish. Due to the time required to establish assays, collecting leaf blade sections and set up with adult weevils was spread over 2 days.
Assessment of Endophyte Infection Fifty seeds from lots for each isolate line used in planting were assessed for endophyte infection using aniline blue stain as in Saha et al. (1988). Infection rates ranged from 92 to 94% in all isolate groups, while Nil plants had 2% infection. Further, all plants were assessed for endophyte infection following their use in bioassay trials. Tillers of plants were cut 3-5 cm from the soil surface and the end blotted on immunoblot paper to detect presence of fungi within tillers as in Simpson et al. (2012). Plants (n = 15 out of 320 total) from isolate treatments found to lack infection and those from the nil treatment that were infected were omitted from the analysis.
Chemical Analyses Simultaneous with collecting leaf blade sections from plants for weevil bioassays, we also harvested ten 4 cm basal sections of leaf blades from tillers of the same plants. Twenty plants per treatment combination were sampled and blades from 5 plants were combined into one sample to provide tissue mass (>100 mg) adequate for chemical analyses. This provided 4 replicates for most treatment combinations; some had small plants and thus fewer bulked replicates could be made. Samples were placed on liquid nitrogen in a portable cooler and then transferred to a − 80°C freezer. They were lyophilized and homogenized by adding ball bearings to sample vials on a shaker. Samples were analysed for a wide range of endophyte alkaloids using methods described by Moore et al. (2015). The plant alkaloid perloline was detected simultaneously with peramine using the chromatographic conditions described by Moore et al. (2015). The alkaloids were monitored as extracted MS 1 ion chromatograms at 248.3 m/z [M + H] + peramine, 333.3 m/z [M -OH + H] + perloline, and 262.3 m/z [M + H] + the internal standard homoperamine using a single-quadrupole detector (MSQ Plus, Thermo Fisher Scientific, Waltham, MA, USA) with the following parameters; ESI in positive ion mode, probe temperature 350°C, cone voltage 75 V, 0.10 s dwell time, and a 1 amu span. The limit of detection (LOD) and limit of quantitation (LOQ) were 0.1 ppm and 0.3 ppm for peramine, and 0.3 ppm and 1 ppm for perloline, respectively. Lolitrem B was extracted for 1 h with 1 mL of 2:1dichloromethane/methanol. The sample was then centrifuged (5000 g, 5 min) and a 500 μL aliquot of the supernatant transferred via a 0.45 μm syringe filter (PVDF) to an HPLC vial for analysis. Separation was achieved with an isocratic flow (1 mL/min, 80% dichloromethane/20% acetonitrile) using a Luna Silica 250 × 2.0 mm (5 μm) column (Phenomenex, Torrance, CA, USA), with the lolitrem B peak detected with a Shimadzu RF-10Axl fluorescence detector (excitation at 260 nm, emission detection at 410 nm). The limit of quantitation of this technique was 0.1 μg g − 1 DM (0.1 ppm).
Statistical Analyses To test the hypothesis (#1 above) that fungal isolate affects weevil preference we analyzed feeding damage by weevils to leaf blade segments after 24 and 48 h by fixed factor 2-way ANOVA (with main effects of fungal isolate and damage treatment and their interaction) using SPSS (IBM Corp 2011). Data approximated a normal distribution, so no transformation was performed. Multiple contrasts between treatments within an isolate group were performed to test a priori hypotheses (#2-4 above); we corrected for possible inflation of making a type I error using a Bonferroni adjustment of calculated confidence intervals (Dunn 1961). The hypothesis that alkaloid induction reduces weevil preference (#3 above) was further assessed by regressing weevil feeding on quantities of each alkaloid using multiple step-wise regression in SPSS (IBM Corp 2011) in which predictors were added to the model (forward selection) based on their contribution to the explaining variability in feeding score. Further, a linear regression of weevil feeding and concentrations of perloline was conducted within each of the four isolates.

Results
We found weevils fed less on grasses infected with CS or AR1 isolates (Fig. 1). This was expected because weevils tend to avoid peramine (Rowan and Gaynor 1986) and both CS and AR1 isolates produced this alkaloid, while NEA2 did so at reduced levels and Nil produced none (Fig. 2). For this reason AR1, which produces no ergot alkaloids or lolitrems (Fig. 2c,d) that are detrimental to livestock, is a commonly used Bselected^endophyte in grazing systems in New Zealand (Easton and Tapper 2005). No other alkaloids had overall patterns of expression that explained differences in weevil feeding damage (Fig. S2).
In contrast to the effect of fungal isolate, we did not find an overall effect of damage type on weevil preference ( Fig. 1; Table S1). Rather, weevil response to damage depended upon fungal isolate; that is, there was an interaction between the main effects of fungal isolate and damage (F 9,285 = 1.92; P = 0.049; Table S1). Prior damage in both AR1 and CS isolates had little impact on weevil feeding damage (Fig. 1), while plants infected with the NEA2 isolate displayed induced resistance to weevil feeding; clipped plants were more resistant to weevils than were control plants (t = 2.80, df = 19, P < .05 (contrast controlling for type I error); Fig. 1). We found that clipping Nil plants led to increased feeding by weevils, but that the difference in feeding from control plants was not significant (t = 1.54, df = 19, P > .05 (contrast controlling for type I error); Fig. 1).
Interestingly, weevil feeding damage increased on NEA2 blades that came into contact with sheep salivaeither by natural grazing or through manual application (t = 3.38, df = 19, P < .05 (contrast controlling for type I error); Fig. 1). In fact, weevil feeding was as high on plants exposed to sheep saliva (15.5% ± 1.10, mean ± SE) as it was on plants with no prior damage (15.2% ± 1.19, mean ± SE). So, sheep saliva completely mitigated the wound induced response we observed due to clipping.
The chemical mechanism for both induced resistance and mitigation of resistance by saliva is likely the levels of perloline in NEA2-infected plants (Fig. 2a). Perloline increased from 945.4 ± 168.6 ppm to 1438.8 ± 229.2 ppm (mean ± SE) when plants were clipped, but if sheep saliva was applied (either naturally or manually) quantities of perloline were comparable to those found in undamaged, control plants (Fig. 2a). No other alkaloid shared this pattern of expression across the treatments (Fig. 2). Furthermore, variation in perloline concentration in NEA2-infected plants explained variation in feeding by weevil, while it did not for grasses infected with other isolates (Fig. 3). Yet, patterns of perloline in Nil plants did resemble those of NEA2-infected plants, however concentrations were 3× higher in NEA2infected plants (Fig. 2a). Fig. 1 Percent feeding damage after 24 h by adult Argentine stem weevils (weevil) to blade sections of perennial ryegrass infected or uninfected (Nil) with various isolates of Epichloë festucae var. lolii following damage treatments to grass. Fungal isolate had an effect on weevil preference (F 3,285 = 7.76, P < .001), while damage did not (F 3,285 = 0.77, P = .514). However, weevil response to isolate depended upon the damage treatment (F 9,285 = 1.92, P = .049). Histograms with common letters are not significantly different as evidenced by a priori contrasts.

Discussion
While the defensive mutualism between endophytes and their grass hosts was originally postulated to be provided via constitutive resistance to herbivores (Clay 1988), it is becoming clear that wound-induced resistance can also be operating, as shown for loline alkaloids in endophyte-infected tall fescue which confer resistance to bird cherry-oat aphids (Bultman et al. 2004;Sullivan et al. 2006). And, more recently, Fuchs et al. (2017) showed endophyte-infected perennial ryegrass can even exhibit herbivore-specific induction of some alkaloids. Our results here are the first to our knowledge to report plant-mediated interactions (Denno et al. 2008) between sheep and beetles that are modified by an isolate of a microbial symbiont of the plants.
The impact of herbivore oral secretions on plants and their response to herbivory has been reported for some plant-animal interactions. For example, vertebrate saliva can promote growth in some plants (Detling et al. 1980;Liu et al. 2012). In contrast, there is some evidence that vertebrate saliva can also contain antifungal compounds, like lactoferrin peptides and lysozymes (Bellamy et al. 1994) that can protect herbivores against fungal pathogens and their metabolites. Indeed, moose saliva reduced growth of Epichloë endophyte infecting red fescue and reduced quantities of an ergot alkaloid produced by the fungus as well (Tanentzap et al. 2014). Our results show that sheep saliva has similar effects on the response by perennial ryegrass. This is all the more impressive given that we diluted the saliva solution by more than 5× prior to application to plants (see Materials and Methods). Components of sheep saliva that affect the L. perenne-Epichloë symbiotum are unknown, but likely involve salivary enzymes or perhaps even metabolites from living microbes within the saliva. An implication of this effect is that sheep grazing can reduce the level of defense against a major pest of perennial ryegrass. This indirect interaction could have implications for use of particular grazing regimes and pest control practices in managed grazing systems.
Perloline is an unexpected mediator of weevil preference, as it is produced by the grass host, not the fungal endophyte (Hovin and Buckner 1983). It is a diazaphenanthrone alkaloid that can occur at high levels (more than 10,000 mg/kg) in grasses. Its presence in perennial ryegrass (Grimmett and Waters 1943) has been known for over 60 years, but its role in herbivore preference and performance has received little attention (Bush 2001). Early studies showed it inhibited in vitro activity of rumen bacteria (Bush et al. 1970). More recent work showed it, along with two endophyte-produced alkaloids, explained 70% of the variation in dry weight of fall armyworm caterpillars feeding on perennial ryegrass (Salminen et al. 2005). Our data clearly point to perloline as the chemical mechanism responsible for the response we observed in weevil damage. Perloline concentration in NEA2 was negatively associated with weevil feeding (Fig. 3a); this relationship did not hold for any of the other isolates (Fig. 3). We caution that these analyses depend upon few data points, but they nonetheless plainly show a significant pattern between perloline concentration and weevil feeding for only NEA2-infected plants (Fig. 3).
Our results clearly show perloline was produced in plants lacking endophyte (Nil) and that its concentration varied across fungal isolates (Fig. 2a). Why might perloline concentrations differ among grasses harboring different fungal isolates? One explanation could be that genetic differences among grasses harboring different isolates contributed to different expression of the alkaloid. We feel this explanation is unlikely due to precautions we took to ensure plants from all isolate treatments were nearly identical genetically (see Materials and Methods). A second, and more likely explanation is that close biochemical coupling between plant and fungus leads to differences in perloline biosynthesis as isolate genotype varies. As obligate endosymbionts, Epichloë endophytes are finely adapted to their hosts. They receive all of their nutrients from their hosts and can produce compounds that their hosts undoubtedly respond to, such as plant hormones, glycosidases, proteases (DeBattista et al. 1990;Liu et al. 2012;Reddy et al. 1996;Fig. 3 Relationship between perloline in perennial ryegrass and feeding by Argentine stem weevils (weevil) at 24 h. Data points are means of replicates within each damage treatment.; a Feeding on grass infected with NEA2 isolate, b Feeding on grass infected with AR1 isolate, c Feeding on grass infected with CS isolate, d Feeding on grass lacking infection by endophyte. Only isolate NEA2 showed a significant relationship between weevil feeding and perloline concentration (R 2 = 0.982, test of slope different from 0: t = −10.4, df = 2, P = 0.009) Yue et al. 2000), the alkaloids we assessed, and likely many other compounds that allow a fungus to live and grow within the confines of another organism. An intricacy of the intimate fungus-plant relationship was recently uncovered by Pan et al. (2014) who showed that while loline alkaloids are synthesized by many Epichloë endophytes, one step in the diverse production of lolines in meadow fescue requires a host plant-produced enzyme. Thus, both fungal and plant genes can be required for expression of the full array of loline alkaloids produced by the endophyte in the Epichloë-meadow fescue symbiotum. A similar type of coupling may occur between perennial ryegrass and Epichloë, such that certain genotypes (i.e., NEA2) of the fungus promote greater perloline production by the host than do other fungal genotypes (i.e., AR1). This might occur for example, through varied contribution or modulation by different fungal isolates of enzymes needed for perloline biosynthesis. Another, and perhaps more plausible possibility, is that Epichloë elicits a defensive response by perennial ryegrass that includes perloline production and that the level of response differs with fungal genotype. It is well known that many grasses respond to Epichloë endophytes with enhanced levels of antioxidant compounds, like phenolics (White and Torres 2010). These may provide plants with greater protection from oxidative stress associated with plant diseases and other challenges. In a similar way, perennial ryegrass may respond to Epichloë infection with production of perloline and as endophyte isolate varies, plant perloline production could also vary.
One might question why weevils did not respond to damage treatments in Nil plants as they did in plants infected with the NEA2 isolate. Indeed, Nil plants produced perloline and in a pattern across damage treatments that resembled that in plants infected with NEA2. Yet, levels were 2-3× lower in Nil compared to NEA2 isolate-infected plants (Fig. 2a). These low levels are likely below a threshold for response by weevils (Fig. 1). Levels of perloline in AR1-infected plants were also low ( Fig. 2a), yet weevils tended to avoid these plants (Fig. 1); a response likely due to the high levels of peramine (Fig. 2b), as mentioned above.
Thus, even in a relatively simple ecosystem vertebrate and invertebrate herbivores interact indirectly through their common host plant and this interaction is mediated by a symbiont living within the plant. Further, the interaction between herbivores is influenced by genetic variation in the symbiont. Such indirect, complex interactions may be much more common, even in relatively simple managed grassland systems than typically thought and could have important implications for managing grazing systems.