Environmental factors affect the distribution of two Epichloë fungal endophyte species inhabiting a common host grove bluegrass (Poa alsodes)

Abstract Aim The endophyte Epichloë alsodes, with known insecticidal properties, is found in a majority of Poa alsodes populations across a latitudinal gradient from North Carolina to New York. A second endophyte, E. schardlii var. pennsylvanica, with known insect‐deterring effects, is limited to a few populations in Pennsylvania. We explored whether such disparate differences in distributions could be explained by selection from biotic and abiotic environmental factors. Location Along the Appalachian Mountains from North Carolina to New York, USA. Taxon Fungi. Methods Studied correlations of infection frequencies with abiotic and biotic environmental factors. Checked endophyte vertical transmission rates and effects on overwintering survival. With artificial inoculations for two host populations with two isolates per endophyte species, tested endophyte–host compatibility. Studied effects of isolates on host performances in greenhouse experiment with four water‐nutrients treatments. Results Correlation analysis revealed positive associations of E. alsodes frequency with July Max temperatures, July precipitation, and soil nitrogen and phosphorous and negative associations with insect damage and soil magnesium and potassium. Plants infected with E. alsodes had increased overwintering survival compared to plants infected with E. schardlii or uninfected (E−) plants. Artificial inoculations indicated that E. alsodes had better compatibility with a variety of host genotypes than did E. schardlii. The experiment with reciprocally inoculated plants grown under different treatments revealed a complexity of interactions among hosts, endophyte species, isolate within species, host plant origin, and environmental factors. Neither of the endophyte species increased plant biomass, but some of the isolates within each species had other effects on plant growth such as increased root:shoot ratio, number of tillers, and changes in plant height that might affect host fitness. Main conclusion In the absence of clear and consistent effects of the endophytes on host growth, the differences in endophyte‐mediated protection against herbivores may be the key factor determining distribution differences of the two endophyte species.

phytic fungi of cool season grasses, has been shown to mitigate the effects of environmental stress such as drought and nutrient deficiencies as well as anthropomorphic stresses such as elevated CO 2 associated with climate change and resisting invasive species (Brosi et al., 2011;Compant, Heijden, & Sessitsch, 2010;Craig et al., 2011;. Moreover, these fungi may produce alkaloid compounds that have toxic or deterrent effects on various herbivores, thus reducing environmental stress from insect herbivory and vertebrate grazing (Brosi et al., 2011;Cheplick & Faeth, 2009;Compant et al., 2010;Craig et al., 2011;Hunt, Rasmussen, Newton, Parsons, & Newman, 2005;Schardl, Balestrini, Florea, Zhang, & Scott, 2009). The mode of transmission of Epichloë endophytes varies, with some species transmitted either vertically (via hyphae growing into seeds) or horizontally (by forming stromata and causing disease symptoms) or via both modes depending on the environment (Clay & Schardl, 2002). Epichloë endophytes that are thought to be strictly vertically (maternally) transmitted are considered more strongly mutualistic because host plant and endophyte reproduction, and hence fitness, are closely linked (Cheplick & Faeth, 2009;Clay & Schardl, 2002).
Generally, little is known about the effects of endophytes on their hosts across natural populations from different environments (e.g., Cheplick & Faeth, 2009;Hamilton, Faeth, & Dowling, 2009;Novas, Collantes, & Cabral, 2007;Wei et al., 2007). Basic knowledge of the variation in endophyte species and strains and their frequencies over a geographic range of environmental conditions may provide insights into the long-term nature of the interactions of endophytes and their hosts. Genetics of host plants also varies over the range of a grass species and may interact with variation in endophyte species or strain to affect persistence of the plant-endophyte symbiota. Indeed, host and endophyte genotypic combinations, especially in maternally transmitted endophytes, may have co-evolved with each other to increase fitness, and thus may be adapted to local environmental conditions (Cheplick & Faeth, 2009;Oberhofer, Gusewell, & Leuchtmann, 2014;Saikkonen et al., 2010).
Correlational studies may provide insight into what environmental factors are associated with different endophyte species or strains within a common host grass.
Thus, this P. alsodes host grass system is unique because it is the only grass host species known to date where an interspecific or intraspecific hybrid Epichloë species co-occur.
Selection by the biotic and abiotic environment largely controls whether the costs of harboring Epichloë endophytes outweigh the benefits or vice versa, and the outcomes of this selection over time may be reflected in endophyte distributions and frequencies across the populations. Correlation with environmental factors can point to possible factors that may determine the distribution and relative frequency of the endophyte species. However, the assumption that higher relative frequencies of an endophyte species reflect greater benefits may be misleading because other factors such as differences in rate of endophyte transmission (Afkhami & Rudgers, 2008;Sneck, Rudgers, Young, & Miller, 2017), timing of species origin or host-endophyte associations, meta-population or meta-community dynamics, or differences in dispersal may affect frequencies (Faeth & Sullivan, 2003;Saikkonen, Faeth, Helander, & Sullivan, 1998;Saikkonen, Lehtonen, Helander, Koricheva, & Faeth, 2006;Saikkonen, Wali, Helander, & Faeth, 2004 (Jia, Oberhofer, Shymanovich, & Faeth, 2016;Jia et al., 2015;Oberhofer et al., 2014;Vandegrift et al., 2015).
We hypothesized that key environmental factors affect the presence and frequency of Epichloë endophyte species in natural F I G U R E 1 (a) Poa alsodes plants inoculates with Epichloë alsodes, A1 and A2 isolates (b, c), and Epichloë schardlii var. pennsylvanica, S1 and S2 isolates (d, e) from the greenhouse experiment with different water-nutrient treatments

| Plant host
Poa alsodes A. Gray (common name, grove bluegrass), family Poaceae, is a perennial woodland grass species. P. alsodes is distributed in eastern North America from Canada to South Carolina, USA. In the southern part of its range, it is restricted to mountainous areas and becomes more widespread in northern regions. Flowering occurs in spring, and plants are mainly out-crossing via wind pollination, but self-fertilization is also possible. P. alsodes has not been used in agriculture (Hill, 2007).

| Endophyte species
The widespread and common endophyte inhabiting P. alsodes is E. alsodes, which is an interspecific hybrid of E. typhina subsp. poae and E. amarillans. This species has two mating type idiomorphs, MTA and MTB, and genes for production of N-acetylnorloline, a loline alkaloid. Genes for ergot alkaloids and peramine biosynthetic pathways are not functional (Shymanovich et al., 2017). The less common and more range restrictive endophyte inhabiting P. alsodes, E. schardlii var. pennsylvanica, is closely related to, and most likely is synonymous with, E. schardlii, which was described previously from Cinna arundinacea hosts (Ghimire et al., 2011;Shymanovich et al., 2017).
For simplicity and clarity, we use the E. schardlii name for this endophyte here. This endophyte is an intraspecific hybrid of two strains of E. typhina subsp. poae. This species has the MTB idiomorph and the peramine alkaloid gene. However, based on chemical analyses, peramine is not produced (Shymanovich et al., 2017). Both endophytes, like most hybrid Epichloë species, appear to be strictly vertically transmitted by hyphae growing into seeds and no stromata have been observed on P. alsodes in nature.

| Correlations of infection frequencies with abiotic and biotic environmental factors
We determined whether Epichloë species frequencies in the natural

| Vertical infection transmission rates
Transmission rates were estimated for each endophyte species in each population because observed population infection frequencies may depend on the effectiveness of vertical transmission, and transmission efficiency may be affected by environmental factors (Hill & Roach, 2009;Rolston, Hare, Moore, & Christensen, 1986;Siegel, Latch, & Johnson, 1985). For example, imperfect transmission (failure of hyphae to grow into seed ovaries or loss of endophyte viability in plants or seeds due to high temperatures), has been used to explain variation in endophyte frequencies in nature (Afkhami & Rudgers, 2008;Liu, Nagabhyru, & Schardl, 2017;Ravel, Michalakis, & Charmet, 1997). To determine transmission rates, infection status of about 24 (depending on availability) seeds from each of three infected mother plants per population was determined with immunoblot assay (Phytoscreen Immunoblot Kit #ENDO7971 Seed; Agrostics, Watkinsville, GA, USA). Mean transmission rate for each population was calculated from the three mother plants for each Epichloë species.

| Inoculations to test endophyte-host compatibility
To test for difference in endophyte-host compatibility for the two endophyte species, different isolates of each species, host plants from different populations, and reciprocal inoculations with endophytes were used. Inoculation success should be positively associated with endophyte species-host plant compatibility (Latchs & Christensen, 1985;Oberhofer et al., 2014). To control for the plant population effects, naturally uninfected seeds (collected in 2012-2013) from the two widely separated P. alsodes populations were used (Table S1) (modified from Shymanovich et al., 2017). One population is located at the southern limit of P. alsodes' distributional range in North Carolina (NC). This population is found at a high elevation with high precipitation and relatively low summer temperatures. In this NC population, only one endophyte, E. alsodes, was observed at relatively low infection frequency (26%). The second, northern P. alsodes population is in Pennsylvania (PA), where the two endophyte species co-occur. However, because of the lower elevation of this population, summer temperatures are higher and precipitation is lower compared to the NC population. To incorporate endophyte variation within species, two mycelial isolates for each species were obtained from different populations for the artificial inoculations (Table S2).
For the E. alsodes endophyte, one isolate (A1) was from the NC population, and the second (A2) was from the PA population. For E. schardlii var. pennsylvanica, one isolate (S1) was from a different population in Pennsylvania where only this endophyte species was present, and the second (S2) was from the PA population described above where the two endophyte species co-occur. In this experiment, due to time and budget limitations, we were unable to take into account possible genetic variation within a given population of plants between naturally uninfected and plants infected with a specific endophyte. The latter requires removing the endophyte and growing these plants at least for one year in a natural environment to produce seeds. Therefore, for the NC seedlings, we attempted to introduce A1, a residential isolate, and A2, S1, and S2, three alien isolates. For the PA seedlings, we attempted to introduce A2 and S2, residential isolates, and A1 and S1, alien isolates ( Figure 1).
Two endophyte inoculation techniques were employed: with and without seedling puncturing ( Figure S1). On 18 September 2014, for each isolate, 17 potato dextrose agar plates were inoculated by pouring on to their surface a suspension of fresh fungal mycelium stirred in sterile water by a pestle. Plates were kept in the dark at 24°C.
For each population, seeds from four naturally uninfected mother pants were used. Infection status of each mother plant was verified by PCR (Shymanovich et al., 2017). About 2,300-2400 surface sterilized seeds (1 min 70% ethanol, 4 min 4% sodium hypochlorite, 1 min 70% ethanol, 1 min sterile water), from each population, were split into four isolate groups, evenly placed on ten-day-old cultures, and kept in the dark, 24°C for the next 10 days (similarly to Tadych light/dark schedule ( Figure S2). When germination began during three weeks, each 3-6 mm seedling was punctured under laminar flow with sterile BD PrecisionGlide™ 0.4 × 13 mm needle into a hypocotyl near the seed coat, and a small portion of surrounding mycelium was introduced into a wound using a microscope at 400× and light source (puncturing treatment) as described in Latchs and Christensen (1985) and Oberhofer et al. (2014). Plates were checked for germination every 2-3 days, and newly processed seedlings were marked on the lid ( Figure S2). After 7-8 days, inoculated seedlings were individually removed from the agar and planted in 50 ml pots with potting soil ( Inoculation success was evaluated for each plant-isolate combination as number of positively infected seedlings/total number of survived seedlings for each inoculation procedure (puncturing and mycelia) separately × 100%. Total inoculation success was calculated as total number of positive seedlings/total number of seedlings survived × 100%.

| Effects of endophytes on plant performances
To test the effects of endophyte species and plant genotype on plant performance, we used infected seedlings from the inoculations and negative controls (seedlings that were inoculated but remained negative) from NC and PA populations (NC-E-and PA-E-). For E. alsodes infected plants, we had all the expected combinations: NC-A1, NC-A2, PA-A1, and PA-A2. For E. schardlii var. pennsylvanica infected plants, we only had sufficient numbers for NC-S1 and NC-S2. Due to poor inoculation success for PA-S1 and PA-S2 groups, they were excluded from this experiment. Therefore, we were unable to compare the effects of E. schardlii var. pennsylvanica infections on plants from the two populations.
Fifty clones were produced for each remaining seed-endophyte  (Jia et al., 2015;Saari & Faeth, 2012) to achieve significant differences in plant growth. Plant positions were rotated every 10 days to minimize any microclimatic differences within the greenhouse.
The experiment continued for 97 days after treatments began.
On 5 June 2015, plant height and number of tillers were recorded, and then plants were harvested. Aboveground and belowground biomass was separated, dried (three days at 65°C in a drying oven), and shoot and root dry biomass were determined, and root: shoot ratio, as a measure of plant resource allocation, was calculated. A few plants did not survive to the end of the experiment and were excluded from the statistical analyses. Infection status for each plant was confirmed with immunoblot assay (as described above).
The infection status of all plants except one (negative instead of positive) was as expected. This plant was excluded from the statistical analyses.

| Statistical analyses
Statistical analyses were performed with R i386 3.3.2 software with "R commander" package (R Development Core Team, 2008).

| Multiple regression analyses
To explore the relationship of endophyte frequencies with environ-

| Overwintering survival
For overwinter survival comparisons, Pearson's Chi-squared tests were applied for three groups and pairwise combinations.

Comparisons of inoculation success were performed with similar
Pearson's Chi-squared tests. To determine the effects of the specific infections within each treatment, one-way ANOVA comparisons for all variables were used for each plant population with endophyte as a fixed factor. The same transformations as above were used.

| Regression analyses of endophyte infection frequencies with environmental factors
Epichloë alsodes infection frequencies across the latitudinal populations of P. alsodes were associated positively with July Max temperature, July precipitation, soil organic matter, phosphorous, and pH, and negatively with soil magnesium, potassium, and mean insect damage (best-fit regression model, F = 10.93 on 9 and 9 df, pvalue 0.0007, R 2 = 0.83) (

| Residential versus alien isolate effects within treatments
Two E. alsodes isolates, when inoculated into plants from the NC population had different effects on total dry biomass (multi-way ANOVA, p = 0.009), leaf dry biomass (multi-way ANOVA, p = 0.006), root: shoot ratio (multi-way ANOVA, p = 0.01), and number of tillers (multi-way ANOVA, p = 0.01). Treatments were always significant as expected   Figure 3). Also root: shoot ratios were higher in two treatments for A2 plants than for A1 inoculated plants.
Height values for presumably residential endophyte (A2) infected plants were lower in the two low nutrients treatments than for plants with alien isolate (A1) (one-way ANOVAs, p = 0.0009, p = 0.016).

| Plant population
Genetic differences between plants from NC and PA populations affected only plant height and number of tillers (  Figure 2). Also, population and treatment interacted to affect plant heights (Table 4). The A1 and A2 isolates affected the height only of PA plants in the HWLN and LWHN treatments (Table 5).

| Treatments
As expected, the water-nutrient treatments strongly affected all growth variables (Table 5). Leaf and root biomass were lower in the LWHN than in HWLN treatment. Plants in the most stressful treatment, LWLN, had the smallest leaf and root biomass (Figure 4).
The effects of treatments on plants with a specific infection are discussed below (

| Effects of E. alsodes and E. schardlii var. pennsylvanica isolates on North Carolina plants
Endophyte infection affected all growth variables for NC plants infected with one of the two isolates for either endophyte species, E. alsodes or E. schardlii var. pennsylvanica (Table 6). As expected, treatments had strong effects on all growth variables. The interaction between endophyte status and treatment was not significant.

| Epichloë schardlii effects
Neither of the E. schardlii isolates had significant effects on total, leaf, and root biomass compared to uninfected NC plants (Figure 1).  (Figure 2).

| Effects of the four isolates
The effects of endophyte infection depended more on the specific isolate than on the Epichloë species. Isolates of each endophyte species had variable effects on host growth parameters, and this variation was often greater than variation between endophyte species ( Figure 2). Plants inoculated with the A1 and S2 isolates did of tillers was greater in plants inoculated with the S1 than with the A2 and S1 isolates.

| Effects of the isolates within treatments
When comparing plants from the same population with different infection types within treatments, several interesting effects were observed (Table 5; Figure 4). For NC population plants, all infection groups had similar total, leaf, and root dry biomasses in each treatment combination ( None of four isolates in PA plants had any effects on the tiller number at any treatment (Table 5).

| D ISCUSS I ON
The E. alsodes endophyte occurs commonly over a wide range of associations also can increase the frequency, persistence, and range of host plants (Klironomos, 2003;Smith & Read, 2010). However, the benefits of mycorrhizal associations depend on environmental conditions such as soil moisture, pH, temperature, and limiting nutrients, especially phosphorous (Bentivenga & Hetrick, 1992;Entry et al., 2002;Tuomi, Kytöviita, & Härdling, 2001). Mycorrhiza may also alleviate host stresses to various anthropogenic pollutants (Entry et al., 2002).
Asexual Epichloë are transmitted vertically and are not free-living, so their frequency and distribution might be determined indirectly via selection by environmental factors on host plant fitness.
If harboring the endophyte increases host fitness relative to uninfected plants across environments, then frequency and range of infected plants should increase with time (Clay, 1988(Clay, , 1990. For example, Clay (1988) showed that the frequency of E. coenophialum in agronomic tall fescue increased in heavily grazed pastures over time because livestock avoided infected plants. If, alternatively, the cost of infection outweighs the benefit in certain environments, then infection frequencies should decrease relative to uninfected plants.
For example, Novas et al. (2007) observed that in extremely harsh conditions in south Patagonia, Epichloë infection frequencies were reduced in several grass species. The same arguments apply to host grass species that harbor more than one Epichloë species. If infection by one endophyte species increases host plant fitness in certain TA B L E 5 Summary of significant effects of isolates from E. alsodes and E. schardlii var. pennsylvanica endophyte species on a host plant, Poa alsodes, growth parameters for North Carolina (NC) and Pennsylvania (PA) populations under specific treatments

Ln (number of tillers)
Treatments: HWHN-high water high nutrients, LWHN-low water high nutrients, HWLN-high water low nutrients, LWLN-low water low nutrients. b Infections: E− uninfected; S1, S2-infected with E. schardlii var. pennsylvanica isolates 1 and 2; A1, A2-infected with E. alsodes isolates 1 and 2. Our study showed that the frequencies of the two endophyte species, E. alsodes and E. schardlii var. pennsylvanica, were also correlated with key environmental factors. Frequency of the widespread E. alsodes in the southern populations was associated with increased July Max temperatures (Tables 2 and 3). However, positive correlation with July precipitation may indicate that this endophyte may not mediate drought stress. This finding contrasts with previous experimental studies that showed that infection with an undetermined Epichloë sp. (but based upon its wide distribution, probably E. alsodes) from Indiana may increase drought resistance in P. alsodes (Kannadan & Rudgers, 2008).
The frequency of E. alsodes was also positively associated with soil nitrogen or organic matter (both variables are highly collinear).
E. alsodes infected host plants may be associated with high nitrogen and phosphorous soils because of the increased nitrogen and phosphorous demand of producing high levels of NANL, a loline alkaloid. Alkaloids are nitrogen-rich compounds and phosphorous is required in their synthesis (Faeth & Fagan, 2002;Schardl, Grossman, Nagabhyru, Faulkner, & Mallik, 2007). Alternatively, Epichloë infection itself may also enhance uptake of phosphorous from nutrient poor soils (Malinowski, Alloush, & Belesky, 2000). Increased phosphorous content in Festuca rubra plant tissues was demonstrated F I G U R E 4 Mean (±SE) leaf dry and root dry biomasses for North Carolina (NC) population plants (a, c, respectively) and for Pennsylvania (PA) population plants (b, d, respectively). Naturally uninfected, some NC plants were successfully inoculated with either of two E. alsodes isolates (residential A1 and alien A2) or two E. schardlii var. pennsylvanica isolates (S1 and S2, both alien), or remained uninfected after procedures (E−). Some PA plants were successfully inoculated with either of two E. alsodes isolates (alien A1 and presumably residential A2) or remained uninfected after procedures (E−). After infection status check and cloning, plants were randomly assigned into four treatments-HWHN, HWLN, LWHN, LWLN for 97 days. For each symbiotum combination, there were 11-13 plants per treatment. Letters represent statistically significant differences among infection groups within each treatment (one-way ANOVAs, p < 0.05) for Epichloë festucae infection (Zabalgogeazcoa, Ciudad, Vázquez de Aldana, & Criado, 2006 (Jensen, Popay, & Tapper, 2009;Popay, Tapper, & Podmore, 2009;Shymanovich et al., 2019).  Figure 4a,c).
Alternatively, the opposite direction of the correlation may be a statistical artifact because as the relative frequency of one endophyte species such as E. alsodes increases, the second endophyte frequency may decrease by default.

| Transmission rates
Differences in transmission rates among Epichloë endophytes provide another explanation for differences in frequency and distribution that does not involve natural selection by the environment (Faeth & Sullivan, 2003;Ravel et al., 1997). Epichloë infection may be lost due to imperfect transmission (failure of hyphae to grow into seeds; Ravel et al., 1997) & Faeth, 2009;Hill & Roach, 2009;Rolston et al., 1986;Siegel et al., 1985). Imperfect transmission can result in decreasing infection frequencies over time, even if endophytes increase fitness, if the rate of transmission failure is high (Ravel et al., 1997)

. Various
Epichloë species in native grasses may have very different rates of transmission which could contribute to differences in frequency and range (Afkhami & Rudgers, 2008). However, the transmission rate hypothesis does not appear to explain differences in E. alsodes and E. schardlii var. pennsylvanica frequency and distribution, or the relative rarity of E. schardlii. Both species hosted by P. alsodes had high transmission rates (95%-100%) across all populations in our study. Chung et al. (2015) also detected high transmission rates in the populations in Indiana populations of P. alsodes infected with unspecified (but likely E. alsodes) Epichloë endophyte.

| Compatibility
Similarly to other studies (e.g., Friesen et al., 2011;Oberhofer et al., 2014;Saikkonen et al., 2010), our inoculation trials provided additional evidence that plant genetic characteristics may control the compatibility with specific endophytes (  (Ghimire et al., 2011) and this may partially explain the restrictive distribution of E. schardlii var.

pennsylvanica.
Increased compatibility of host-endophyte genetic combinations may have improved host growth parameters. Several plant growth parameters indicated that resident host-endophyte combinations, which may be co-adapted, were more beneficial to the host.  Figure 3). Greater root biomass may indicate better drought resistance and enhanced nutrient uptake (e.g., . Host plant co-adaptation with their residential endophytes may also depend on local environmental conditions. For example, the A1 isolate of E. alsodes that originated from the wettest habitat (Table 1) did not increase root biomass allocation in any plants. However, A2 isolate from the driest habitat (based on annual and July precipitation, Table 1) increased biomass allocation to roots in plants from both populations in intermediate stress level treatments ( Figure 3) and thus may potentially increase host resistance to drought stress.
Nevertheless, caution is necessary for two reasons. First, just a few isolates were tested in this study and, second, seeds from naturally uninfected genotypes were used for inoculations. Thus, additional inoculation experiments with other isolates and host plant populations and also with initially naturally infected genotypes may provide a stronger support for the hypothesis that plants and endophyte genotypes are co-adapted.

| Effects on host performance
Similar to other studies describing host-endophyte interactions as a mutualism-parasitism continuum (Junker, Draeger, & Schulz, 2012;Schulz & Boyle, 2005), our growth performance experiments with reciprocally inoculated plants from the NC and PA populations revealed the complexity of host and endophyte genotype and environment interactions on plant growth parameters. Different isolates from the same endophyte species may have different effects on plants from a given population. Moreover, effects of an endophyte on growth parameters were dependent on specific waternutrient conditions. In the resource-rich treatment environment (HWHN treatment), infected plants did not differ much in growth parameters than uninfected plants, except height and tiller number (Table 5). Some differences in growth parameters between infected and uninfected plants, and between plants infected with different isolates, were detected when plants were grown in the moderately stressful treatments (HWLN and LWHN) or in some cases when in highly stressful environments (the LWLN treatment (Table 5; Figures 3 and 4)).
However, the major result of the performance experiment is that neither of two endophyte species or their isolates increased total plant biomass compared to uninfected plants, and in some cases, infection even reduced biomass (Figures 2 and 3 Our experiment also revealed interactions of plant population origin and endophyte isolates ( Table 5). The effects of the E. alsodes isolates differed when introduced into plants from the North Carolina and Pennsylvania populations. For example, when infected with A2 isolate, NC plants showed only root biomass reductions, but the same isolate inoculated into PA plants showed reduced root and leaf biomass compared to uninfected plants from the same population ( Figure 2). Root: shoot ratios increased for PA plants infected with the A2 isolate compared to uninfected plants but root: shoot ratios of NC plants infected with the same isolate did not differ from uninfected plants (Figure 2). Likewise, PA plants infected with A1 isolate had fewer tillers than uninfected plants, but tiller number of NC plants infected with the same isolate tended to be greater than in uninfected plants ( Figure 2).
Overall, our growth performance experiment showed complex outcomes of infection depending on endophyte species, isolate within species, population origin of the host plant and environmental factors. We did not find consistent or clear benefits of the endophyte infection by either species.
Our approach with artificial inoculations and a performance experiment with controlled water-nutrient environments provided valuable results but had several limitations. First, because inoculations were made in naturally uninfected seedlings, we were not able to strictly control for plant genotypic variation within the population. These naturally uninfected plants may have once been infected with Epichloë, or may have been from plant lineages that had never been infected. Our inoculation and compatibility results suggest that plants infected by specific species and their isolates may be genetically distinct. Second, just a few plant and endophyte genotypes were tested for co-adaptation. Third, our greenhouse experiment with potted plants in uniform potting soil, and controlled temperature, water, and nutrient conditions may or may not simulate natural environments. Fourth, we were unable to document seed production by plants infected with isolates of the endophyte species. None of the plants produced florets during the course of the experiment. Therefore, the growth parameters we measured are only assumed to affect reproduction and fitness. Fourth, we were unable to compare plant population effects for E. schardlii var. pennsylvanica because this endophyte was not successfully inoculated into PA plants.

| CON CLUS IONS
Our study explored several explanations for the broader distribution range and higher frequency of the interspecific hybrid, E. alsodes, compared to the limited distribution of intraspecific hybrid species, E. schardlii var. pennsylvanica. Increased overwintering survival and better compatibility with a P. alsodes host from across the latitudinal gradient we sampled, may allow E. alsodes to persist over a broad latitudinal range. That the distribution and frequency of E. alsodes is correlated with maximum and minimum temperatures supports the overwintering success hypothesis. We did not find evidence that either endophyte species or their isolates provide consistent benefits in terms of growth parameters that would explain differences in distribution. However, our previous work (Shymanovich et al., 2017), showed that E. alsodes has another important benefit: production of loline alkaloids which may significantly reduce plant damage due to toxic effects on insect herbivores. E. schardlii var. pennsylvanica has insect deterrence properties, but does not have significant effects on insect survival and does not appear to produce alkaloids. Variation in insect defense mechanisms may be a key factor for variation in the distribution ranges. That E. alsodes, which produces high levels of NANL, a loline alkaloid that is nitrogen-rich and may compete with plant functions for nitrogen, is positively associated with high-nitrogen soils, suggests that the costs and benefits of alkaloid production may be important in dictating its distribution and frequency. Our overall results also support the more general hypothesis that interspecific hybridization provides greater genetic variation than intraspecific hybridization (e.g., Schardl & Craven, 2003) and thus greater potential for adaptation to wider range of, and more stressful, environments. Infection by the interspecific hybrid species, E. alsodes, appears to enable its host plant to persist across a wide variety of local environments across the 1,200 km latitudinal range that we sampled. In contrast, plants infected with the intraspecific hybrid species, E. schardlii var. pennsylvanica, appears restricted to a limited environments within this latitudinal range.
Our correlational and experimental tests suggest that the broader range of E. alsodes infected grove bluegrass may be related to greater variation in alkaloid production and enhanced overwintering survival, as well as changes in some growth parameters.
However, other hypotheses that do not involve natural selection by the environment, such as recent origination or host jump of E. schardlii var. pennsylvanica in Pennsylvania, or limited dispersal of E. schardlii var. pennsylvanica, cannot be excluded without further experimentation and observation.

AUTH O R CO NTR I B UTI O N S
All authors contributed to the project design, data analyses, and writing of the manuscript. TS conceived the project.

DATA ACCE SS I B I LIT Y
Data available in the Supplementary Material file.