Fungal endophyte infection and host genetic background jointly modulate host response to an aphid-transmitted viral pathogen


Correspondence author. E-mail:


  1. Despite their ubiquitous nature, interactions between multiple micro-organisms and their effects on host growth and each other's success have received limited scientific attention. In particular, grasses can be commonly infected by both endophytic fungi and viruses, which are typically transmitted by aphids. In this study, we investigated how an aphid-transmitted viral pathogen and a symbiotic endophytic fungus altered host growth and allocation. We hypothesized that, by reducing aphid feeding, endophyte infection would lower viral titre and consequently decrease the negative effects of virus infection on host biomass production.
  2. In a greenhouse experiment, we manipulated fungal endophyte status and virus infection (Barley Yellow Dwarf Virus – PAV) of two tall fescue cultivars with different genetic backgrounds [KY 31 and pasture demonstration farm (PDF)]. In one cultivar (PDF), we also manipulated endophyte strain, using two strains that had been selected for differences in alkaloid production. We assessed host, virus and vector responses.
  3. As hypothesized, endophyte infection decreased reproduction and abundance of aphid vectors; however, in contrast to our hypothesis, this response by aphids did not impact viral titre. For both tall fescue cultivars, endophyte infection alleviated the negative effect of virus infection on the proportion of total plant biomass allocated to roots. On the other hand, for the KY 31 cultivar only, virus infection decreased tillering in endophyte-infected individuals, but not endophyte-free individuals. Within the PDF cultivar, both endophyte strains produced similar effects on host, virus and vector responses.
  4. Synthesis These results indicate that some of the beneficial effects provided by endophyte infection, particularly alleviating the negative effect of virus infection on the proportion of total plant biomass allocated to roots, do not arise strictly from altering host interactions with herbivores (aphids), but also occur by changing host responses to viral infection. Furthermore, these results emphasize the importance of exploring multispecies microbial interactions and genetic controls on these interactions in order to more fully understand their role in community- and ecosystem-level dynamics.


Plant hosts are often confronted simultaneously with a diverse array of micro-organisms, including both pathogens and mutualists (Arnold 2007; Friesen et al. 2011; Pieterse & Dicke 2007). The close relationships between hosts and their microbes are characterized by a high degree of recognition and signalling between the plant and the associated microbe at molecular, morphological and physiological levels (Harrison 2005). Furthermore, association with microbes can alter plant phenotypes by supplying novel nutritional and defence pathways for the plant as well as influencing plant biochemical pathways (Friesen et al. 2011). Such alterations in plant phenotypes due to association with one microbe may in turn alter plant relationships with other microbes. These relationships may be altered either directly via the shared host or indirectly via a third player such as an arthropod vector. For example, mutualistic microbes may help protect plants against pathogens either by increasing plant defence against pathogens themselves, or by increasing plant defence against herbivores, including arthropods that transmit pathogens (Clay & Schardl 2002; Hartley & Gange 2009). Thus, a broad community context may be important for understanding at least some of these microbial interactions (Saunders, Glenn & Kohn 2010). Despite this recognition, relatively few studies examine the impact of interactions among multiple micro-organisms on host growth, biomass allocation or the impact of different micro-organisms on each other's success. Here, we investigate how the interaction of a foliar endophytic fungus and an arthropod-transmitted plant virus interact with plant genetic background to influence host growth, biomass allocation and the performance of both the virus and its arthropod vector.

A majority of plant-infecting viruses are dependent upon arthropod vectors for transmission between hosts (Hogenhout et al. 2008; Nault 1997). Therefore, virus ecology is often dependent on the population dynamics, host preference and movement of vectors (Power & Flecker 2008). Barley and cereal yellow dwarf viruses (B/CYDVs) are a widespread group of aphid-transmitted, generalist viral pathogens that have provided a model system for plant–virus–vector interactions (Gray & Gildow 2003). For example, consumption of B/CYDV-infected host tissue often increases aphid fecundity, with some variation among host, vector and virus species (Power & Gray 1995). Additionally, increased abundance of aphid vectors generally increases the rate at which B/CYDVs are transmitted to healthy plants (Burnett & Gill 1976; Jensen & D'Arcy 1995; Power & Gray 1995). Thus, plant characteristics that alter vector population dynamics are likely to alter their transmission of viruses.

Many grass species host endophytic fungi in the ascomycete family Clavicipitaceae. These endophytes receive nutrients, protection, reproduction and dissemination via seeds from the plant (Schardl, Leuchtmann & Spiering 2004). In return, the host may receive competitive advantages from the symbiont, including enhanced soil nutrient uptake (Malinowski, Alloush & Belesky 2000) and drought resistance (Arachevaleta et al. 1989; Malinowski & Belesky 2000). In addition, many of these endophytes are thought to provide herbivore deterrence via the production in planta of several distinct classes of biologically active alkaloids that can reduce arthropod feeding, population size and consequent damage for the host plant (Clay 1990; Schardl, Leuchtmann & Spiering 2004). However, benefits to the host provided by fungal alkaloid production can vary among herbivore species, host species and genotypes, endophyte strains and environmental conditions (Bultman & Bell 2003; Cheplick 1998; Faeth 2002; Swarthout et al. 2009). Endophyte-produced alkaloids may influence aphid-transmitted plant pathogens because, among insect herbivores, aphids are some of the most negatively affected by endophyte infection (Hartley & Gange 2009). Endophytic alkaloids commonly deter aphid consumption and reduce aphid fecundity (Hartley & Gange 2009; Schardl & Phillips 1997).

For viruses transmitted by aphids and other arthropods, the arthropod deterrence that results from endophyte infection may in turn decrease the severity of virus infection for the host plant. Transmission of B/CYDVs to the plant from the aphid typically requires several hours of aphid feeding, so decreased aphid feeding duration as a result of endophyte infection can decrease transmission of B/CYDVs to the plant (Power & Gray 1995). Furthermore, a decreased number of feeding aphids can decrease the titre of resulting virus infections (Power & Gray 1995), so impacts of endophytes on both aphid population size and feeding duration may reduce the titre of resulting virus infections in endophyte-infected hosts. In turn, reduced virus titre can both decrease the negative impacts of infection on the host plant and increase the amount of feeding time necessary for uninfected aphids to acquire the virus from the plants (Power & Gray 1995).

Within grass–fungal endophyte associations, such as that of tall fescue (Schedonorus arundinaceus (Schreb.) Dumort = Schedonorus phoenix (Scop.) Holub = Lolium arundinaceum Darbyshire = Festuca arundinacea Schreb.) and Neotyphodium coenophialum (Morgan-Jones & W. Gams) Glenn, C. W. Bacon & Hanlin, endophytes can produce a suite of alkaloid compounds that deters both mammalian and insect herbivory (Schardl, Leuchtmann & Spiering 2004). So-called common toxic strains of these endophytes typically reduce arthropod and mammalian feeding on tall fescue in agroecosystems (Breen 1994), although such deterrence may depend on abiotic conditions (Bultman & Bell 2003; Hunt & Newman 2005; Rasmussen et al. 2007). ‘Novel’ strains of these endophytes have been identified and isolated from naturally occurring host populations (Clement et al. 2001). These novel strains generally produce no or very low concentrations of the mammalian-active compounds but retain, to varying degrees, the production of compounds important in deterring arthropod herbivores (Malinowski & Belesky 2006). Several of these novel endophyte strains have been introduced into endophyte-free lines of agriculturally important hosts, such as tall fescue, and developed for commercial use [e.g. Jesup MaxQ and pasture demonstration farm (PDF)/Texoma MaxQ II; Bouton et al. 1997; Hopkins et al. 2011]. Such technological innovations can produce high-yielding host cultivar populations that benefit from endophyte-produced insect-deterring alkaloids but do not harm grazing mammals, particularly livestock. However, it is possible that novel endophytes may differ from common toxic strains in ways that are not evaluated in the breeding and commercial screening processes, including impacts on plant response to virus infection.

Previous research suggests that hosts infected with novel endophytes may be at a competitive disadvantage compared with hosts infected with the common toxic strain when exposed to biotic stresses, such as herbivory and abiotic stresses, such as drought (Malinowski & Belesky 2006). Additionally, there is evidence to suggest that novel endophytes do not provide the same degree of protection from aphids as the common toxic strain (Hunt & Newman 2005). Specifically, intrinsic rates of growth for enclosed populations of aphids were greatest on endophyte-free plants, slower on novel endophyte-infected plants and slowest (or no growth at all) on the plants infected with the common toxic strain of endophyte (Hunt & Newman 2005). Therefore, we predicted that novel endophyte infection will provide less aphid deterrence and consequently less protection for the host from virus infection, than the common toxic endophyte strain.

Much of the previous research on virus–endophyte–aphid interactions has centred on community-level studies performed on agriculturally important host species, with varying results (Guy 1992; Guy & Davis 2002; Mahmood et al. 1993). Individual host genotypes or cultivars (plant varieties produced via selective breeding) may interact with fungal endophyte presence and strain in ways that modify host response to viral infection. In one experiment, endophyte infection and virus infection masked each other's effects on the growth of perennial ryegrass for some combinations of host genotype and endophyte strain (Hesse & Latch 1999), but virus infection status was based on visual symptoms, which are highly unreliable (D'Arcy 1995). While this same methodological issue limits what can be inferred from some other previous studies (Lehtonen et al. 2006; Mahmood et al. 1993), the total available evidence strongly suggests the potential for strong interactions among host genetic background, endophytes and viruses of grasses (Guy & Davis 2002; Hesse & Latch 1999; Lehtonen et al. 2006; Mahmood et al. 1993). Yet, no study has combined experimental manipulations of both endophyte infection and virus infection with measurements of the performance of not only hosts but also the virus and its insect vectors. Such an integrated approach may provide an important insight into the dynamics and consequences of microbial interactions in grasses.

Here, we present an experiment evaluating the interaction of virus and endophyte infections as they relate to impacts on a widespread and ecologically important host species, tall fescue. Specifically, we explore how endophyte presence and host genetic background (as contained in two fescue cultivars), interact with virus infection to alter vector abundance, host biomass, allocation and tillering. Using a recently released cultivar, we also evaluate the effects of fungal endophyte strain on these parameters. The goal of this investigation was to document the magnitude of interactions among microbes and plants that have been little quantified, yet are widespread, and so may play a crucial role in grassland ecosystems around the world.

Materials and methods

Study System

Barley and cereal yellow dwarf viruses are a group of aphid-transmitted generalist viral pathogens that infect over 150 crop and noncrop grasses (D'Arcy 1995; Halbert & Voegtlin 1995). B/CYDV infection is systemic and localized to the phloem where it causes necrosis and disruption of carbohydrate translocation (D'Arcy 1995; Irwin & Thresh 1990). Impacts of infection include stunted plant growth, reduced root/shoot ratio and reduced longevity (Kolb et al. 1991; Malmstrom et al. 2005). B/CYDVs are obligately transmitted by aphids, including the globally common aphid species Rhopalosiphum padi (L.).

Tall fescue is a cool-season grass that has been introduced from Europe to the United States where it has been widely planted and bred for use as forage in pastures due to its ability to tolerate high temperatures, drought conditions and grazing (Stuedemann & Hoveland 1988). Many of the properties that make tall fescue attractive for use as a forage species can be attributed to the symbiotic fungal endophyte Neotyphodium coenophialum (Clay & Schardl 2002). It is estimated that between 75% and 85% of tall fescue in the US is infected with the common toxic form of N. coenophialum (Ball, Pedersen & Lacefield 1993; Clay & Schardl 2002). Tall fescue provides a valuable model system to investigate microbe–microbe interactions because pairwise host–fungus interactions and mechanisms for microbe–microbe competition have been relatively well described in this system (Saunders, Glenn & Kohn 2010).

Experimental Design, Treatments and Conditions

We used two tall fescue cultivars, KY 31 (Kentucky 31) and PDF, a.k.a, Texoma. Experimental seed for the KY 31 cultivar was either endophyte free (E−) or contained the common toxic strain of N. coenophialum (CTE+). Seed for the PDF cultivar was either endophyte free (E−), infected with the common toxic strain of endophyte (CTE+), or was infected with a novel strain of N. coenophialum (AR 584E+). Seed from the PDF cultivar was obtained from the Noble Foundation in Ardmore, Oklahoma, and seed from the KY 31 cultivar was obtained from the University of Kentucky. Plants were germinated in experimental pots. When multiple germinates were observed, plants were thinned down to one plant per pot. Plants were watered every 3 days.

The experiment was conducted in the greenhouse at the University of North Carolina at Chapel Hill. In each of the five host endophyte treatments above (KY 31 CTE−, KY 31 CTE+, PDF CTE−, PDF AR 584E+, PDF CTE+), we manipulated virus infection (infected and uninfected) at the individual pot level. This was replicated three times per block for five blocks, yielding a total of 150 experimental plants (5 host endophyte treatments × 2 virus treatments × 3 replicates × 5 blocks). Individual plants were grown in D60 Deepots (Steuwe and Sons Inc, Tangent, Oregon, USA). Each plant received 800 g of steam sterilized soil in a mixture of one part sandy loam soil with two parts of pure sand (by mass).

To infect plants with virus, we used the FA2K298 isolate of Barley yellow dwarf virus – PAV (hereafter referred to as BYDV for brevity). This isolate was collected on 21 June 1998 from Avena sativa in Central NY State and has previously been used in inoculation experiments (Hall et al. 2010; Power & Mitchell 2004). Since collection, it has been maintained (approximately three transmission cycles per year) in laboratory plants of A. sativa cultivar Coast Black oats. The virus isolate has been partially sequenced, see GenBank accession numbers DQ285674 and DQ286379 (Hall 2006). Virus inoculations occurred approximately 2 weeks after plant germination, when all plants had one tiller and two to three leaves. BYDV is solely transmitted by aphid feeding; therefore, to infect plants with virus, uninfected aphids of the species R. padi were fed in petri dishes for 72 h on infected plant tissue. Five infected aphids, primarily in the 5th (adult) instar and occasionally in the 4th instar, were then transferred to each experimental plant, at which time a cap, constructed of clear plastic and nylon mesh, was placed on plants to prevent the spread of aphids. Aphids were allowed to feed on each experimental plant for 48 h and then uncapped. In order to determine vector feeding responses, the number of apterous (unwinged) adult aphids, alate (winged) adult aphids and juvenile nymph aphids (whether apterous or alate) was counted for each plant. Plants were then sprayed with a horticultural oil solution (SAF-T-SIDE; ClawEl Specialty Products, Pleasant Plains, IL, USA) to kill remaining aphids. In order to ensure that all plants received herbivore pressure from aphid feeding, mock-inoculated plants received the same treatment, but uninfected aphids were fed on uninfected tissue prior to being transferred to experimental plants. Therefore, all plants received exposure to the same low level of aphid feeding (for 48 h, with five initial aphids) regardless of infection status. To test the plants for BYDV infection, a compound indirect double-antibody sandwich enzyme-linked immunosorbent assay (ELISA; Agdia Inc, Elkhart, IN, USA) was used. From each experimental plant, we tested 0.15–0.30 g wet above-ground tissue (a mixture of leaves, including pseudostem tissue, of varying ages), maintaining a constant tissue: extraction/buffer ratio of 1 mg: 10 μL (Cronin et al. 2010). Five plants that were inoculated with infected aphids but did not become infected with BYDV were removed from the analysis for a total of 145 experimental plants.

Plants were allowed to grow for 6 weeks after inoculation and then harvested. At harvest, plants were separated into above- and below-ground portions. and two tillers from each plant were visually scored for the presence of endophyte by staining leaf sheaths with lactophenol cotton blue (Clark, White & Patterson 1983). All endophyte-infected individuals tested positive, and all endophyte-free individuals tested negative. Soils were frozen and stored at −20 °C until they could be washed. The below-ground fraction was washed to separate roots from soil. Both above- and below-ground biomass samples were oven-dried at 60 °C for a minimum of 72 h to obtain dry biomass values.

Statistical Analysis

We used several response variables to assess experimentally induced changes in plant performance. To assess changes in plant allocation, we used root fraction, root biomass divided by total plant biomass. BYDV is known to suppress root allocation (Irwin & Thresh 1990; D'Arcy & Burnett 1995); therefore, root fraction is an important indicator of virus impact. Total plant biomass was the sum of all above- and below-ground biomass. To account for the portion of above-ground tissue removed for ELISA, a wet/dry conversion factor was calculated based on the ratio of wet/dry biomass and applied to the ELISA weight. This estimated dry mass was then added to complete the total biomass metric. We quantified tillering, a component of vegetative growth that can be sensitive to damage from natural enemies (Jewiss 1972), by counting the number of tillers per plant. In order to assess viral responses, we used relative viral titre. Viral titre is the measure of the concentration of virus present in plant tissue. ELISAs generate optical density (OD) values that can be used as a measure of the relative viral titre (Cronin et al. 2010). To assess the impacts of endophyte strain and host cultivar on viral titre, we considered only those plants infected with virus.

We performed two sets of statistical analyses to answer two different sets of questions. In order to assess cultivar × endophyte interactions, we excluded plants of the PDF cultivar infected with the novel endophyte AR 584 because there was no equivalent cultivar–endophyte combination for the KY 31 cultivar. For the same reason, to assess the role of endophyte strain in altering plant–virus–vector interactions, we excluded plants of the KY 31 cultivar and considered only the PDF cultivar, which contained both the common toxic and novel strains.

All data were analysed using r (v.2.13.1; R Foundation for Statistical Computing, Vienna Austria) with the ‘lme4’ package and the ‘glmer’ and ‘lmer’ functions (Bates & Maechler 2009). Data from the experiment was subjected to analysis of variance using general linear models with glasshouse block as a random effect. Response variables were log transformed to fit model assumptions of homogeneity of variances when necessary. The response variables of plant biomass, root fraction and viral titre were analysed using linear regressions while tiller number, and aphid number were analysed with Poisson regression employing the standard log link. Differences between means were determined using Tukey HSD with the ‘glht’ function of the ‘multcomp’ package for all measured parameters (Hothorn et al. 2010). Tukey's HSD test is a one-step procedure for examining pairwise comparisons and does not require a significant F statistic because conditioning it on the F statistic effectively decreases the α level required for significance to a level below 0.05, making the test overly conservative (Hancock & Klockars 1996; Hochberg & Tamhane 2009). All statistical tests were considered significant at P ≤ 0.05 and are included in Tables 1 or 2. Appendix A includes tables of the full statistical models for all response variables.

Table 1. Significant responses for full statistical models that were used to assess the effects of the three experimental factors: infection with Barley yellow dwarf virus PAV (−BYDV vs. +BYDV), endophyte infection [E− vs. CTE+; excluding the novel endophyte Pasture Demonstration Farm (PDF) 584], host cultivar (PDF vs. KY 31) and their interactions (indicated with colons)
Response variableTreatment or interaction
BYDVEndophyteCultivarBYDV : endophyteBYDV : cultivarEndophyte : cultivarBYDV : endophyte : cultivar
  1. P values < 0.05 are bold, while P values 0.1 < × < 0.05 are italicized. Blank cells indicate non-significant results.

Total biomass0.0001 0.0178
Root biomass0.0001 0.0727
Shoot biomass 0.0008 0.0167
Root fraction 0.0097 0.0952
Tiller number 0.0095 0.0198
Nymph abundance 0.0030 0.0598 0.0106
Adult abundance 0.0306
Total aphid abundance 0.0022 0.0018 0.0733
Table 2. Significant responses for full statistical models that were used to assess the effects of two experimental factors within the Pasture Demonstration Farm (PDF) cultivar: virus presence (−BYDV vs. +BYDV) and endophyte strain (seed type: PDF E− vs. PDF E+ vs. PDF 584) and their interaction (indicated with a colon)
Response variableTreatment
BYDVSeed typeBYDV : seed type
  1. a

    Seed type = PDF E− vs. PDF584.

  2. b

    Seed type = PDF E− vs. PDFE+.

  3. P values < 0.05 are bold, while P values 0.1 < × < 0.05 are italicized. Blank cells indicate non-significant results.

Total biomass 0.0021
Root biomass 0.0006 0.0717
Shoot biomass 0.0044
Root fraction 0.0104
Tiller number 0.0849 0.0003 a, 0.0575 b
Aphid abundance


Plant Biomass and Allocation

The effect of the common toxic endophyte on plant biomass depended on plant cultivar (endophyte × cultivar: F1,105 = 5.794, P = 0.018, Fig. 1A, Table 1). While common toxic endophyte infection significantly decreased total plant biomass of the KY 31 cultivar (Tukey HSD: P = 0.017), it did not significantly alter plant biomass of the PDF cultivar (Tukey HSD: P = 0.619). Within the PDF cultivar, the effect of endophyte presence and strain on total plant biomass was not significant in the anova (F2,79 = 1.709, P = 0.188, Table 2), but Tukey's HSD tests indicated that there was an effect of endophyte infection. Specifically, total plant biomass did not differ between plants infected with the novel vs. the common toxic endophyte strain (Tukey HSD: = 0.938), but endophyte-free plants produced less biomass than plants infected with the novel endophyte (Tukey HSD: = 0.043; Fig. 1B). Across both host cultivars, virus infection decreased total plant biomass by 70% (F1,105 = 15.65, P = 0.0001, Table 1), including both root (F1,105 = 21.1, < 0.0001, Table 1) and shoot biomass (F1,105 = 11.94, P = 0.0008, Table 1), none of which were significantly altered by common toxic endophyte infection (> 0.60).

Figure 1.

Common toxic endophyte infection decreased total plant biomass for the KY 31 cultivar but did not have a significant effect on the Pasture Demonstration Farm (PDF) cultivar (A). Within the PDF cultivar, infection with the novel endophyte increased total biomass in comparison with endophyte-free plants (B). Data shown are means ± SEM; letters indicate significant pairwise differences between means within each panel (Tukey HSD;< 0.05).

We examined impacts of microbial infections on plant allocation in terms of root fraction. While the endophyte-by-virus interaction was only marginally significant in the anova (F1,105 = 2.835, P = 0.0952, Table 1), Tukey's HSD tests indicated that the effect of virus infection depended on whether the host was also infected by the endophyte. Specifically, while virus infection decreased the root fraction of endophyte-free plants (Tukey HSD: P = 0.0097), common toxic endophyte infection greatly reduced the magnitude of this effect and rendered it statistically non-significant (Tukey HSD: P = 0.3648; Fig. 2A). This pattern was mirrored within the PDF cultivar, including plants infected with both endophyte strains. While the endophyte-by-virus interaction was also not significant in the anova (F2,79 = 2.098, P = 0.1295, Table 2), Tukey's HSD tests indicated that endophyte infection mitigated the impact of virus infection. Specifically, while virus infection decreased the root fraction of endophyte-free plants (Tukey HSD: P = 0.009), both the common toxic endophyte (Tukey HSD: P = 0.365) and the novel endophyte (Tukey HSD: P = 0.918) reduced the magnitude of this effect, so it was statistically non-significant (Fig. 2B).

Figure 2.

Virus infection decreased root fraction of endophyte-free plants but not of endophyte-infected plants. This was observed across both host cultivars, considering only the common toxic endophyte strain (A), and within the Pasture Demonstration Farm cultivar, for both the novel endophyte and the common toxic endophyte strain (B). Data shown are means ± SEM; letters indicate significant pairwise differences between means (Tukey HSD;< 0.05).

Tiller Production

Virus infection, plant cultivar and common toxic endophyte infection all significantly interacted to alter the number of tillers produced per plant (virus × cultivar × endophyte: z = 2.329, P = 0.0198; Fig. 3A, Table 1). Specifically, virus infection decreased the number of tillers produced per plant for endophyte-infected plants from the KY 31 cultivar (Tukey HSD: P = 0.022), but not endophyte-free KY 31 (Tukey HSD: P = 1.000), or the PDF cultivar regardless of endophyte status (Tukey HSD: > 0.89). Within the PDF cultivar, including plants infected with both endophyte strains, virus infection tended to reduce tiller production, but this effect was only marginally significant (z = −1.723, P = 0.0849, Table 2), and did not depend on endophyte presence or strain (> 0.4). Across virus treatments, infection of PDF by either the common toxic (z = 3.663, P = 0.0003, Table 2) or the novel endophyte (z = 1.900, P = 0.0575, Table 2) tended to increase the number of tillers produced compared to the endophyte-free PDF plants, but there was no difference in tiller production between plants infected by the novel endophyte or the common toxic endophyte (z = −0.880, P = 0.6530, Fig. 3B).

Figure 3.

Considering only the common toxic endophyte, virus infection significantly decreased tiller production for endophyte-infected plants in the KY 31 cultivar only (A). Across virus infection status, infection with the novel endophyte or the common toxic endophyte increased tiller production for the Pasture Demonstration Farm cultivar (B). Data shown are means ± SEM; letters indicate significant pairwise differences between means (Tukey HSD;< 0.05).

Aphid Abundance

Across host cultivars and virus infection status, common toxic endophyte infection tended to decrease the number of nymphs (z = −1.882, P = 0.059; Fig. 4A and Table 1) and significantly reduced the number of apterous adult aphids (z = −2.162, P = 0.031; Fig. 4B and Table 1) and the number of total aphids (nymphs + apterous adults + alate adults; z = −3.068, P = 0.0022; Fig. 4C and Table 1). The production of aphid nymphs (juveniles) was lower for those aphids that fed on the PDF cultivar (z = −2.557, P = 0.0106) compared with those that fed on KY 31. Furthermore, within the PDF cultivar alone, neither endophyte strain decreased aphid abundances relative to endophyte-free plants (> 0.66). Finally, the production of aphid nymphs was lower for those aphids that fed on virus-infected rather than virus-free tissue (z = −2.965, P = 0.003, Table 1). There were no treatment interactions that significantly influenced nymph abundance (> 0.2).

Figure 4.

Averaged across host cultivar, common toxic endophyte infection (‘y’) tended to decrease the abundance of aphid nymphs (A), and decreased both the abundance of apterous adult aphids (B), and the total abundance of aphids (C), relative to endophyte-free plants (‘n’). Data shown are means ± SEM.

Viral Titre

As an indicator of relative viral titre in leaf tissue, we analysed OD values from ELISAs for virus-infected hosts only. Averaged across endophyte statuses, OD values for the KY 31 cultivar were 91 per cent higher than for the PDF cultivar (F1,48 = 11.98, P = 0.0011; Fig 5, Table 1). OD values averaged 33% lower in common toxic endophyte-infected plants than in endophyte-free plants, but this was not statistically significant (F1,48 = 1.775, P = 0.1891), and there was no effect of endophyte infection on ODs within the PDF cultivar (> 0.3).

Figure 5.

The KY 31 cultivar had significantly higher relative viral titre as measured by optical density (OD) value then the Pasture Demonstration Farm cultivar. Data shown are means ± SEM.


While BYDV infection universally reduced plant biomass in our experiment, our results indicate that endophyte infection benefited the plant by reducing the severity of virus impacts on below-ground plant allocation. Furthermore, infection by the common toxic endophyte decreased the abundances of the aphids that act as vectors for the virus, although such impacts did not translate to significant impacts on viral titre. Finally, virus infection, endophyte infection and host cultivar interacted to control production of new tillers, a key component of growth in grasses such as tall fescue.

The primary mechanism by which we predicted virus–endophyte interactions to occur was via alterations to the arthropod vector. Consumption of B/CYDV-infected host tissue commonly increases aphid fecundity (Jensen & D'Arcy 1995; Mauck et al. 2012), but endophytes typically deter aphid consumption and reduce fecundity (Hartley & Gange 2009; Schardl & Phillips 1997). Despite the short duration of aphid feeding time allowed in this experiment, our results supported our initial predictions. As expected, when plants of both cultivars were analysed together, those infected with the common toxic endophyte supported less aphid production, abundance of adult aphids and total number of aphids. There were also additional differences due to host cultivar, in which the PDF cultivar supported lower production of aphid nymphs compared with KY 31. Virus-infected plants produced fewer nymphs than virus-free plants, as has been observed in some previous studies (Mauck et al. 2012).

Since arthropod vectors play a pivotal role in the transmission of most known plant viruses (Power & Flecker 2008), we predicted that the arthropod deterrence that typically results from endophyte infection (Schardl, Leuchtmann & Spiering 2004) would decrease the severity of virus infection for endophyte-infected hosts. Specifically, we predicted that endophyte presence would lower virus titre by decreasing aphid feeding time (Power & Gray 1995). Lower titres should then result in plants that are less severely impacted by virus infection than endophyte-free, viral-infected plants. However, contrary to our expectations, we did not observe lower viral titres in endophyte-infected vs. endophyte-free plants, despite observing that endophyte presence did reduce aphid abundances. One possible explanation for this result is that the virus may have had enough time to replicate after inoculation that it was able to compensate for reduced inoculum from decreased aphid feeding on endophyte-infected plants. Additionally, total plant biomass was not influenced by endophyte–virus interactions, but in terms of biomass allocation, our results indicate that endophyte infection enhanced plant tolerance to virus infection by reducing virus impacts on root allocation for both cultivars. While infection with B/CYDVs typically decreases root allocation for infected hosts (Irwin & Thresh 1990), endophyte infection ameliorated this effect. Furthermore, within the PDF cultivar, both endophyte strains mitigated virus impacts on root allocation. Although these results were only supported by Tukey's HSD tests, the fact that endophyte infection mitigated the impacts of the virus in both host cultivars, and regardless of endophyte strain, suggests that these results were robust. Such effects may allow endophyte-infected hosts to tolerate virus infection better and survive longer in a field setting.

Our results indicate that virus–endophyte interactions can also be shaped by host genetic background. Relative viral titre was much higher in the KY 31 cultivar than in the PDF cultivar. This difference between cultivars in virus titre corresponded to a difference in the effect of virus infection on tillering of endophyte-infected and endophyte-free plants. Specifically, while there was no interactive effect of endophyte infection and cultivar on virus titre, virus infection reduced tillering in common toxic endophyte-infected plants of the KY 31 cultivar, but not of the PDF cultivar. These results comparing host cultivars are consistent with previous studies comparing host genotypes, such as Hesse & Latch (1999), where endophyte infection appeared to reduce the effects of BYDV infection on plant biomass for some host genotypes but not others. Furthermore, while endophyte infection tended to increase overall plant biomass for plants from the PDF cultivar, endophyte infection significantly decreased overall plant biomass for the more common KY 31 cultivar. Variability in endophyte effects reflect host and endophyte genetic controls interacting with the surrounding environment, which together determine where the symbiosis falls on the continuum from mutualism to parasitism (Cheplick & Faeth 2009). For example, variation in alkaloid and metabolic profiles among different host genotype-endophyte strain combinations (Faeth, Bush & Sullivan 2002; Rasmussen et al. 2008) and different environmental conditions may explain why endophyte infection increases herbivory in some cases (Faeth & Shochat 2010; Jani, Faeth & Gardner 2010) but decreases it in others (Clay & Schardl 2002; Saikkonen, Saari & Helander 2010; Schardl, Leuchtmann & Spiering 2004). Thus, the alkaloid and metabolic profiles of different cultivar–endophyte strain combinations may have differential effects on the virus and/or the vector, perhaps explaining the observed virus × cultivar × common toxic endophyte effect on tiller production.

Novel endophyte strains can invoke different degrees of protection from herbivores and environmental stresses compared to the common toxic strain (Hunt & Newman 2005; Malinowski & Belesky 2006). In our study, the novel endophyte and common toxic endophyte did not invoke different host responses to viral infection or create differences in aphid abundances. Contrary to expectations, within the PDF cultivar, neither common toxic nor novel endophyte presence significantly affected aphid abundance, perhaps reflecting the lower overall aphid abundances supported by this cultivar. However, both the common and novel endophyte increased overall plant biomass, reduced the negative effect of virus infection on root fraction and stimulated tiller production similarly compared to endophyte-free plants. This indicates that in our study, the novel endophyte AR 584 had similar impacts on the host as the common toxic endophyte. Furthermore, there were no significant interactions between virus infection and endophyte strain for overall plant biomass, root fraction or tiller production. Thus, the differences between these two endophyte strains, whether in terms of alkaloid profiles or other characteristics, were not important in altering viral dynamics in this host cultivar during this study.

In conclusion, these results indicate that both host genetic background and endophyte infection can play important roles in determining host–herbivore–virus interactions. Endophyte infection, in addition to potentially providing protection against virus infection by decreasing vector abundance, may also mitigate viral effects on host below-ground allocation and thereby enhance host tolerance to viral infection. Thus, our work provides a largely unconsidered, but perhaps general, mechanism by which one microbe can alter plant phenotypic response to other microbes (Friesen et al. 2011). While further studies will be needed to assess the robustness of these results, they illustrate the complex interactions between host genetic background and microbial composition that challenge our understanding of microbial interactions in plants and their effects on community- and ecosystem-level processes (Saunders, Glenn & Kohn 2010).


We thank the Noble Foundation for providing the PDF material and Dr Tim Phillips at the University of Kentucky for the KY 31. We are grateful to Jack Weiss for statistical advice. We would also like to thank the Mitchell lab for assistance. Lauren Buckley, Alicia Frame and Josie Reinhart provided comments on earlier versions of this manuscript. This research was partially supported by the joint NSF-NIH Ecology of Infectious Disease programme through NSF Grants EF-05-25641 and DEB-10-15909 to C.E.M., Kentucky Agricultural Experiment Station funds (KY006045) to R.L.M., an NSF GRFP to M.A.R and the National Science Foundation Postdoctoral Research Fellowship in Biology under Grant No. DBI- 12-02676 to M.A.R.

Appendix A: Full statistical models

Tables A1, A3, A5, A7, A9, A11, A12, A13 and A17 present the full statistical models that were used to assess the effects of the three experimental factors: infection with Barley yellow dwarf virus – PAV (−BYDV vs. +BYDV), host cultivar (PDF vs. KY 31) and endophyte infection (E− vs. CTE+; excluding the novel endophyte PDF 584). Tables A2, A4, A6, A8, A10, A14, A15, A16, and A18b present the models that were used to assess the effects of two experimental factors within the PDF cultivar: virus presence (−BYDV vs. +BYDV) and endophyte strain (Seed type: PDF E− vs. PDF E+ vs. PDF 584). For analysis details, see the Methods section. (Significance codes for all tables: P < 0.001‘***’; 0.001 < P < 0.01‘**’; 0.01 < P < 0.05‘*’; 0.05 < P < 0.1‘.’).

A1 Model of total plant biomass

numDFdenDF F-value P-value
BYDV : Endophyte11050.154900.6947
BYDV : Cultivar11050.783170.3782
Endophyte : Cultivar11055.793550.0178*
BYDV : Endophyte : Cultivar11050.027260.8692

A2 Model of total plant biomass: within the PDF cultivar

numDFdenDF F-value P-value
Intercept17994.81344< 0.0001***
Seed type2791.709560.1876
BYDV : Seed type2790.574090.5655

A3 Model of root biomass

numDFdenDF F-value P-value
Intercept1105157.94545< 0.0001***
BYDV110521.10247< 0.0001***
BYDV : Endophyte11050.005160.9429
BYDV : Cultivar11050.004970.9439
Endophyte : Cultivar11053.287630.0727
BYDV : Endophyte : Cultivar11050.029550.8638

A4 Model of root biomass: within the PDF cultivar

numDFdenDF F-value P-value
Intercept179103.49826< 0.0001***
Seed type2792.725370.0717
BYDV : Seed type2790.489000.6151

A5 Model of shoot biomass

numDFdenDF F-value P-value
Intercept1105108.43510< 0.0001***
BYDV : Cultivar11051.253130.2655
BYDV : Endophyte11050.270550.6041
Cultivar : Endophyte11055.914320.0167*
BYDV : Cultivar : Endophyte11050.021650.8833

A6 Model of shoot biomass: within the PDF cultivar

numDFdenDF F-value P-value
Intercept179104.30478< 0.0001***
Seed type2791.365170.2613
BYDV : Seed type2790.585240.5594

A7 Model of root fraction

numDFdenDF F-value P-value
Intercept1105571.1717< 0.0001***
BYDV : Endophyte11052.83540.0952
BYDV : Cultivar11050.86260.3551
Endophyte : Cultivar11051.27490.2614
BYDV : Endophyte : Cultivar11050.52250.4714

A8 Model of root fraction: within the PDF cultivar

numDFdenDF F-value P-value
Intercept179608.4623< 0.0001***
Seed type2790.29020.7489
BYDV : Seed type2792.09790.1295

A9 Model of tiller number

Estimate SDError z ValuePr(>|z|)
Intercept2.402140.0833728.813< 0.0001***
Cultivar : Endophyte0.174900.149381.1710.2416
Cultivar : BYDV−0.185780.17017−1.0920.2749
Endophyte : BYDV−0.445730.17173−2.5950.0095**
Cultivar : Endophyte : BYDV0.556880.239062.3290.0198*

A10 Model of tiller number: within the PDF cultivar

Estimate SDError z ValuePr(>|z|)
Intercept2.284470.0930424.553< 2e−16***
Seed type PDF5840.375890.102633.6630.0003***
Seed type PDFE+0.202030.106361.9000.0575
BYDV : Seed type PDF584−0.040870.16249−0.2520.8014
BYDV : Seed type PDFE+0.126630.166420.7610.4468

A11 Model of nymph production by the aphid Rhopalosiphum padi

Estimate SDError z valuePr(>|z|)
Intercept1.05310.15256.906< 0.0001***
Cultivar : BYDV0.49830.42881.1620.2452
Cultivar : Endophyte0.44700.39841.1220.2618
BYDV : Endophyte−0.20490.4764−0.4300.6672
Cultivar : BYDV : Endophyte0.34900.66370.5260.5990

A12 Model of adult aphid abundance for the aphid Rhopalosiphum padi

Estimate SDError z ValuePr(>|z|)
Intercept0.875470.166675.253< 0.0001***
Cultivar : BYDV−0.419770.34483−1.2170.2235
Cultivar : Endophyte0.503910.399351.2620.2070
BYDV : Endophyte0.019420.376040.0520.9588
Cultivar : BYDV : Endophyte0.388510.531380.7310.4647

A.13 Model of total aphid abundance for the aphid Rhopalsiphum padi

Estimate SDError z ValuePr(>|z|)
Intercept1.856300.1020618.188< 0.0001***
Cultivar : BYDV0.115830.242540.4780.6330
Cultivar : Endophyte0.463030.258511.7910.0733
BYDV : Endophyte−0.062720.25581−0.2450.8063
Cultivar : BYDV : Endophyte0.246650.371340.6640.5066

A14 Model of nymph production by the aphid Rhopalosiphum padi: within the PDF cultivar

Estimate SDError z valuePr(>|z|)
Seed type PDF5840.127830.292330.4370.6619
Seed type PDFE+−0.026320.30896−0.0850.9321
BYDV : Seed type PDF5840.235070.437530.5370.5911
BYDV : Seed type PDFE+0.144100.462070.3120.7551

A15 Model of adult aphid abundance for the aphid Rhopalosiphum padi: within the PDF cultivar

Estimate SDError z valuePr(>|z|)
Seed type PDF5840.066690.258340.2580.796290
Seed type PDFE+−0.120250.27595−0.4360.663010
BYDV : Seed type PDF5840.444130.354961.2510.210859
BYDV : Seed type PDFE+0.407930.375451.0870.277255

A.16 Model of total aphid abundance for the aphid Rhopalsiphum padi: within the PDF cultivar

Estimate SDError z valuePr(>|z|)
Seed type PDF5840.261010.176231.4810.139
Seed type PDFE+−0.062040.19376−0.3200.749
BYDV : Seed type PDF5840.106710.249790.4270.669
BYDV : Seed type PDFE+0.183930.269180.6830.494

A17 Model of model of optical density values, estimating relative virus titer. Analyses for only virus-infected plants

Estimate SDError t valuePr(>|t|)
Endophyte : Cultivar0.25560.53250.4800.6334

A18 Model of model of optical density values, estimating relative virus titer. Analyses for only virus-infected plants

Estimate SDError t valuePr(>|t|)
(a) Within the KY cultivar
(b) Within the PDF cultivar
Seed type PDF584−0.31870.3131−1.0180.315
Seed type PDFE+−0.27020.3182−0.8490.401