Variation in horizontal and vertical transmission of the endophyte Epichloë elymi infecting the grass Elymus hystrix


Author for correspondence: T. Tintjer Tel: +1 812 535 5262 Fax: +1 812 535 1131 Email:


  • • Systemic fungal endophytes (Clavicipitaceae) of grasses reproduce sexually when the fungus forms stromata and contagious ascospores, or asexually by vertical transmission of hyphae into seeds and seedlings. Vertical transmission is predicted to favor reduced virulence compared with horizontal transmission in systems with both types of transmission.
  • • Here, variation in vertical and horizontal transmission and its potential heritability in a host grass–endophyte interaction, Elymus hystrix infected with Epichloë elymi, were examined in natural populations and two common garden experiments using field-collected host tillers and seed progeny of maternal plants with known infection phenotypes.
  • • Transmission mode exhibited year-to-year variation in field and common garden environments. In the common garden there were consistent differences among maternal plant families in stroma production and significant correlations between stroma production in the common garden and in natural populations. Transmission mode differed among maternal families, spanning a continuum from pure vertical transmission to a high proportion of stroma production and horizontal transmission potential. Vertical transmission to seeds occurred at high rates in all maternal families regardless of their stroma production.
  • • Observed patterns of variation indicate that endophyte transmission mode and correlated changes in virulence can respond to selection by biotic and abiotic factors.


Transmission to new hosts is a major component of parasite fitness. ‘Parasite’ is used here in the sense of nutritional dependence. Along with the length of the infection, transmission determines the number of new infections initiated by an individual parasite (Frank, 1996) and the level of infection in host populations. It is generally assumed to be positively correlated with detrimental effects on host fitness; that is, its virulence (May & Anderson, 1983; Ebert & Herre, 1996; Day, 2001). By contrast, vertical transmission through host eggs may select for less harmful parasites, or even beneficial parasites, because parasite reproductive success is completely dependent on host reproductive success (Ewald, 1987; Dunn & Smith, 2001). Thus, selection on parasites may differ depending on transmission mode, with differing consequences for host fitness. In general, vertical transmission is predicted to favor reduced virulence compared with horizontal transmission (Lipsitch et al., 1996; Ferdy & Godelle, 2005; Stewart et al., 2005).

Both vertical and horizontal transmission success can depend on environmental conditions (Agnew & Koella, 1999; Restif & Kaltz, 2006). For example, humidity may be critical for successful infection by contagious spores (Beyer et al., 2004) whereas successful establishment of vertically infected seedlings may depend on soil moisture (Abbott & Roundy, 2003). Meijer & Leuchtmann (2000) found that horizontal transmission was more common in shady versus sunny conditions (see also Fowler & Clay, 1995). Parasite allocation to horizontal versus vertical transmission may itself be variable and dependent on environmental conditions (Kaltz & Koella, 2003). Plant populations may differ in light intensity, humidity, soil moisture and nutrients, and these could vary across years. Differences in transmission mode among host populations may also be attributable to genetic differences among parasites or host plants (Bucheli & Leuchtmann, 1996; Kover & Clay, 1998). If environmental conditions favorable to the establishment of vertically infected seeds versus contagious infection by spores vary among populations, genetic differences could result from selective responses of transmission mode to environmental factors.

Fungal endophytes in the Ascomycete family Clavicipitaceae systemically infect many grass (Poaceae) species (Clay, 1990). During sexual reproduction the fungus envelops and aborts the developing inflorescence to form a stroma, which produces perithecia and ascospores responsible for horizontal transmission (Chung & Schardl, 1997). Sterilized hosts may nevertheless spread vigorously by clonal growth, propagating the endophyte simultaneously (Clay, 1986; Pan & Clay, 2003). By contrast, some endophytes remain asymptomatic and are vertically transmitted by hyphal growth into host seeds and seedlings. Thus, grass–endophyte interactions span a continuum of pathogenic to mutualistic relationships (Selosse & Schardl, 2007). However, many grass–endophyte associations are characterized by both vertical and horizontal transmission (Clay & Schardl, 2002). In mixed associations some infected plants produce only fertile inflorescences, some produce a mixture of stroma-bearing and fertile inflorescences and others produce only stroma-bearing inflorescences. Stroma number per plant provides an estimate of potential horizontal transmission by ascospores, whereas the number of fertile inflorescences reflects potential vertical transmission through seeds. Infected plants could produce both more seeds and more stromata by producing more inflorescences.

The goal of this study was to determine the degree of variation in vertical and horizontal transmission and its potential heritability in Elymus hystrix infected with Epichloë elymi, which exhibits both types of transmission. Understanding variation in allocation to vertical and horizontal transmission is the first step towards predicting the potential for evolution of transmission mode and correlated responses in virulence. This paper addresses five questions.

  • • Do populations of E. hystrix infected with E. elymi differ in the proportion of stromata?
  • • Is there a relationship between infection frequency and proportion of stromata?
  • • Does stroma production vary across years?
  • • Do plants differ in stroma production (horizontal transmission) and production of infected seed (vertical transmission)?
  • • Finally, do genotypes of the host–endophyte combinations determine transmission mode?

We surveyed stroma production in natural populations and addressed the genetic and environmental contributions to transmission mode in two common garden experiments. Our results provide insight into the amount and basis of variation in endophyte transmission in the polymorphic E. hystrix–E. elymi system.

Materials and Methods

Population surveys

The number of plants with stromata was recorded in eight populations of Elymus hystrix L. infected with Epichloë elymi Schardl et Leuchtmann near Bloomington, Indiana, USA (Table 1). Elymus hystrix is a clump-forming, self-fertilizing perennial woodland grass, native to temperate North America (Gleason & Cronquist, 1991; Tintjer & Rudgers, 2006). It does not produce rhizomes or stolons for vegetative reproduction such as reported for Glyceria striata infected by Epichloë glyceriae (Pan & Clay, 2003). Surveys of stroma-bearing plants were conducted during flowering in June and July of 2001 and 2002. In both years, counts were made of the number of plants with stromata. The first year total population size was estimated whereas in the second year we counted the number of plants in each population. Population sizes were similar over the 2 yr of observations.

Table 1.  Locations and estimates of infection frequencies of populations near Bloomington (IN, USA) surveyed for Elymus hystrix plants bearing stromata (populations arranged alphabetically)
PopulationLatitudeLongitude2000 (leaf)2004 (seed)
Infection frequencynInfection frequencyn
  1. Proportions of samples infected were based on leaf material (in 2000) and on seeds (in 2004).NA, not available.

D39°11′49.6″N86°31′37.8″W0.92510.86 7
J39°11′57.3″N86°31′26.5″W1.00 51.00 5
L39°11′56.8″N86°31′38.9″W1.00 5NA −
T39°11′53.1″N86°31′25.6″WNA1.00 5

Infection frequency estimates

Because many asymptomatic (stroma-free) plants were potentially infected, infection frequencies could not be accurately estimated solely by visual inspection. Infection frequency was initially estimated by microscopic examination for five of the eight surveyed populations and nearby populations in 2000. Infection frequencies in seven of the eight surveyed populations and nearby populations were estimated again in 2004 after stroma surveys (Tables 1, 2). Populations were defined by discontinuous distributions and were separated by > 200 m.

Table 2.  Estimates of overall Epichloë elymi infection frequency in Bloomington (IN, USA) area Elymus hystrix populations
SiteLatitudeLongitude2000 (leaf)2004 (seed)
Populations containing infected plants (number of populations examined)Plants infected (number of plants examined)Populations containing infected plants (number of populations examined)Plants infected (number of seeds examined)
  1. Proportions of samples infected were based on leaf material (in 2000) and on seeds (in 2004).NA, not available.

Griffy Lake-west39°11′57″N86°31′39″W60% (n = 15)47% (n = 122)75% (n = 12)47% (n = 166)
Griffy Lake-east39°11′60″N86°30′47″W80% (n = 5)23% (n = 68)50% (n = 4)15% (n = 40)
TC Steele39°08′17″N86°20′44″W0% (n = 2)0% (n = 6)NA
Yellowwood SF39°13′28″N86°20′32″W0% (n = 7)0% (n = 21)NA
Cedar Bluffs NP39°02′13″N86°33′52″W25% (n = 4)20% (n = 12)NA
Leonard Springs Park39°07′07″N86°35′21″WNA33% (n = 3)19% (n = 41)
Overall  42% (n = 33)27% (n = 229)63% (n = 19)24% (n = 247)

Infection was determined in two ways: (1) by staining leaf sheaths of vegetative tillers (spring 2000); and (2) by staining seeds (autumn 2004) and then carrying out a microscopic examination for characteristic fungal hyphae (Clark et al., 1983). In 2000, a single vegetative tiller per plant was sampled from a haphazard sample of plants in each population (n = 229 tillers, including 144 tillers from populations surveyed for stromata). Five to twenty-five per cent of plants in each population dispersed across the entire population were sampled, and there was no correlation between sampling intensity and infection frequency (data not shown). Tillers were collected before stromata had formed and therefore sampling was blind with respect to infection. In 2004, seeds were collected from another sample of plants (n = 247 seeds, including 104 seeds from populations surveyed for stromata). One seed per plant was collected to avoid sampling a plant more than once. Whereas occasional uninfected tillers on infected plants are found in the field, results from common garden experiment 2 (see Results section ‘Common garden 2: maternal plant families’ below) indicated that all tillers from infected plants were infected and that nearly 100% of seeds from infected tillers were infected. Sampling uninfected tillers or seeds from infected plants would slightly underestimate the true infection frequency.

Common garden 1: plants from surveyed populations

To assess the genetic contribution to stroma production, stroma production of plants from four of the eight populations was measured in a common garden (populations D, J, L and T; locations are given in Table 1). A single tiller was collected from individual plants (n = 4 or 5 per population) in summer 2002. The tillers were grown in a glasshouse until the resulting plants were large enough to divide. Several plants died before division, reducing the number of plants from two of the populations to two and three, respectively. Each plant was divided into several clones (n = 3–6). Clones (n = 60 total; D: 13 clones; T: 7 clones; J: 19 clones; L: 21 clones) were planted in a 12 m × 15 m plot at the Indiana University Botany Experimental Field, Bloomington, IN, USA (39°10′26.6″N, 86°30′23.2″W) in September 2003. Clones were positioned in each row at 0.5-m intervals, which approximates natural densities of E. hystrix plants, and rows were 1.25 m apart. Resident vegetation was controlled by hand-weeding and mowing to minimize competitive interactions. Although old fields are not the typical habitat for this woodland grass, E. hystrix is commonly found in high-light areas along paths and roadsides, and in canopy gaps. During the flowering season in spring 2004, the number of stromata and the total number of tillers were counted for each clone.

Common garden 2: maternal plant families

To assess variation in stroma formation under controlled environmental conditions, a second common garden experiment was established from offspring of E. hystrix plants. Seeds were collected from three populations (39°12′02.2″N, 86°30′45.3″W; 39°11′52.9″N, 86°30′41.4″W; 39°11′55.3″N, 86°30′43.9″W) not included in the stroma surveys described in the previous section. Seeds were collected in July 1998 from 48 plants, and the presence or absence of stromata was noted (n = 25 symptomatic plants and n = 23 asymptomatic). Plants with stromata were endophyte-infected, but plants without stromata could have been either infected and asymptomatic or uninfected. Seeds were planted from inflorescences having at least 10 filled seeds (n = 64 inflorescences from 29 maternal plants: 48 inflorescences from 20 asymptomatic plants (1–9 inflorescences per plant) and 16 inflorescences from nine symptomatic plants (1–6 inflorescences per plant)). Seeds were first cold-stratified at 4°C in moist sand for 14 wk before planting in Metro Mix (Scotts Company, Marysville, OH, USA) in pots (2 cm × 2.5 cm × 5 cm) in a glasshouse in March 1999. Seedling infection was determined by microscopic examination of stained sheath peels (Clark et al., 1983) approx. 4 months after planting.

During September 1999, 161 infected seedlings (n = 80 from asymptomatic maternal plants and 81 from symptomatic maternal plants) and 159 uninfected seedlings (n = 129 from asymptomatic maternal plants and 40 from symptomatic maternal plants, where uninfected tillers were detected on infected plants) were grown in a 10 m × 25 m random array within a common garden at the Indiana University Botany Experimental Field as described in the previous section. During flowering in 2000 and 2001, the numbers of stromata and fertile inflorescences per plant were counted. Prior research demonstrated that fertile inflorescences produce an average of 36 seeds (T. Tintjer, unpublished data). Additionally, the total number of tillers per plant was counted in 2001 to estimate plant size. To estimate efficiency of vertical transmission, seeds were collected from the common garden plants during autumn 2000. A subset of seeds (n = 2–18 per inflorescence; mean = 5.36) were microscopically examined for characteristic fungal hyphae. Lack of infected seeds and stromata on uninfected common garden plants in both years confirmed (1) the accuracy of initial determination of seedling infection and (2) lack of horizontal transmission in the common garden environment. Vertical transmission, in contrast, could occur after the first year of flowering when new seedlings established around experimental plants. However, we carefully examined the size and proximity of tillers relative to the parent plant to avoid sampling newly established seedlings. Seedling tillers were smaller than tillers of established plants and often retained remains of the seed coat at the base.

Genetic analysis of Epichloë elymi isolates

To confirm that we were examining a single endophyte species (E. elymi), 21 isolates were obtained from different infected plants and grown in pure culture as previously described (Leuchtmann & Clay, 1990). Procedures for DNA extraction and amplification and the primers used were those described in Craven et al. (2001). Sequencing and analysis of sequence products were as described in Leuchtmann (2007). Phylogenetic analyses of aligned sequences were performed by heuristic search and maximum-parsimony (MP) implemented in paup* 4.0b10 (Swofford, 2003). Representative sequences of E. elymi isolates are deposited in GenBank (National Center for Biotechnology Information, Bethesda, MD, USA; under the accession numbers EU429345 (isolate #4) and EU429344 (isolate #11).

Data analysis

Variation in stroma production in natural populations  To determine if stroma production of populations differed within and between years, a log-linear model was used to test the effect of population, year, and their interaction on the presence or absence of stromata. Contrasts were used to compare the differences between individual populations within years and between years within populations. To determine if populations differed in infection frequency, a log-linear model was used to test the effect of population and year on number of infected samples, based on examination of tillers in 2000 and examination of seeds in 2004. An ANCOVA model was used to test the fixed effect of year and the effect of population infection frequency, as a covariate, on the number of stroma. Stroma number differences may result from differences in the proportion of infected plants if the proportion of plants in the population with stromata increases with the proportion of infected plants. Infection frequency was arcsine square-root transformed to fulfil the assumptions of the model, and we used a single estimate of infection frequency based on the weighted average from the 2000 and 2004 samples as the covariate. Finally, the direction of the relationship between infection frequency and proportion of plants with stromata each year was determined with a Spearman rank correlation test.

Common garden 1: plants from surveyed populations  To determine the effects of plant genotype and population of origin on stroma production, a nested ANOVA model was used to test the effects of plant nested within population and population on the proportion of stromata in a common garden. Because population variation was expressed at the plant level, we used plant as the corrected error term for the population effect. Spearman rank correlation was used to test the relationship between proportion of plants with stromata in the natural populations and proportion of tillers with stromata per plant in the common garden.

Common garden 2: maternal plant families  To determine whether there were genetic differences between maternal families in stroma production, we quantified offspring plant stroma production. Stroma production was measured in two ways: using the number of stromata and the proportion of stromata. The number of stromata is an estimate of the number of spores produced and thus potential horizontal transmission. Differences between maternal plant families in number of stromata could be caused merely by differences in plant size if plants with more tillers have more stromata. Alternately, some endophytes may induce production of proportionally more stromata, or some plants may be able to resist stroma production and so produce fewer stromata for a given plant size. Therefore, we first considered the effect of maternal family on the number of stromata per plant and then secondly on the proportion of stromata per plant. Repeated-measures ANOVA was used to test the effect of the maternal family on the number of stromata produced by offspring over 2 yr. Maternal plant family was nested within stroma condition of the maternal plants in natural populations (stroma present or absent). Number of stromata was rank-transformed to fulfil the assumptions of the model. We then tested the relationship between plant size (total number of tillers recorded in 2001 only) and number of stromata with a Spearman correlation test. A significant positive relationship would suggest that differences in stroma production could be caused by differences in plant size. We then controlled for the effect of plant size on stroma number by using the proportion of total tillers with stromata. The effect of maternal plant family, nested within maternal stroma condition, on the proportion of stromata (stroma-bearing inflorescences/total tillers (= fertile inflorescences + stroma-bearing inflorescences + vegetative tillers)) was tested with a nested ANOVA for 2001 when total number of tillers was determined. To fulfil the assumptions of ANOVA, proportion of stromata was rank-transformed.

Maternal families may also differ in the relative amounts of stroma production (horizontal transmission) versus inflorescence production (vertical transmission). To determine if there were differences between maternal families in stroma production relative to inflorescence production across years, the proportion of stroma-bearing inflorescences was calculated as the proportion of reproductive tillers (stroma-bearing inflorescences/(fertile inflorescences + stroma-bearing inflorescences)). We used repeated-measures ANOVA to test the effect of maternal plant family, nested within maternal plant stroma condition, on the rank-transformed proportion of stroma-bearing inflorescences repeated over 2 yr.

There could be a trade-off between producing stromata and vertical infection of seeds. We tested the relationship between stroma production and vertical seed infection with a linear regression model where mean proportion of stroma-bearing infloresences per maternal plant family in 2000 (the year seeds were collected) was the independent variable and the proportion of seeds colonized with fungal hyphae (number of seeds with hyphae/number of seeds examined) was the dependent variable.


Population surveys

Populations differed significantly in the proportion of plants with stromata, ranging from 0% to over 20% (inline image = 211.77, P < 0.0001). The log-linear model indicated that years also differed significantly (year: inline image = 13.85, P = 0.0002). The proportion of plants with stromata, averaged across the eight populations, was lower in 2001 (2.9%) than in 2002 (6.0%). In addition, there was a significant population × year interaction effect on the proportion of plants with stromata (inline image = 19.96, P = 0.0056). Specific contrasts indicated that six populations differed between 2001 and 2002, producing significantly fewer stromata in 2001 than in 2002, whereas two populations produced few stromata in both years (Fig. 1).

Figure 1.

Proportions of uninfected, infected and asymptomatic, and infected and symptomatic Elymus hystrix plants in each population in 2001 (left bar) and in 2002 (right bar): white, uninfected (based on a single overall estimate for both years); gray, infected without stromata (asymptomatic); dark gray, infected with stromata. Population locations are given in Table 1.

Populations also differed in infection frequency based on microscopic examination of plant tillers in 2000 (inline image = 95.634, P < 0.001) and seeds in 2004 (inline image = 32.819, P < 0.001) (Table 1). The two estimates of infection frequency were highly correlated (Pearson's correlation coefficient = 0.994, P = 0.006).

The ANCOVA model indicated that there was no effect of year, infection frequency or their interaction on the average proportion of plants with stromata in the populations (year: F1,15 = 0.12, P = 0.731; infection frequency: F1,15 = 2.69, P = 0.127; year × infection: F1,15 = 0.02, P = 0.881) (Fig. 1). In 2001, there was a marginally significant negative correlation between infection frequency and the proportion of stromata in the populations (Spearman correlation coefficient = −0.68, P = 0.066), but no relationship in 2002 (Spearman correlation coefficient = 0.02, P = 0.954).

Genetic analysis of Epichloë elymi isolates

Isolates obtained from plants in four populations used in common garden experiment 1 and eight isolates from plants of experiment 2 grouped in two closely related clades based on β tubulin gene (tubB) sequences (Fig. 2). Both clades contained isolates that formed either no stromata or low to high levels of stromata. Moreover, a stroma-producing reference strain of E. elymi (AF457468) was identical with our isolates of the first clade whereas another (AF250744) was intermediate between the two clades, clearly indicating that different stroma phenotypes found on E. hystrix are not different cryptic species. Further, no gene duplications of tubB or of eight isozyme loci (data not shown) were observed in any of the isolates, suggesting that they were not hybrids.

Figure 2.

Phylogenetic relationships of Epichloë elymi isolates with different levels of a choke expression on Elymus hystrix clones. The cladogram is based on maximum-parsimony (MP) analysis of sequences from intron-rich portions of β tubulin gene (tubB) resulting in a single tree that is rooted with the outgroup taxon Epichloë glyceriae. Numbers associated with branches are branch lengths. Isolates originated from plants of experiment 1 (populations D, L, J and T) or experiment 2 (II) and are numbered with numerals in parentheses. Two reference sequences (GenBank accession numbers AF457468 and AF250744) from stroma-forming E. elymi are included for comparison. Open circle, no stromata; one closed circle, < 1% stromata; two closed circles, 1–10% stromata; three closed circles, > 10% stromata.

Common garden 1: plants from surveyed populations

There was a significant plant genotype effect on proportion of stromata, but populations were not significantly different (plant: F10,59 = 7.41, P < 0.0001; population: F3,9 = 2.61, P = 0.1096). There was a significant positive correlation between the proportion of stroma-bearing plants in natural populations and the proportion of stromata per plant measured in the common garden (Spearman correlation10 = 0.548, P < 0.0001) (Fig. 3).

Figure 3.

Mean proportion of stromata (stroma number/tiller number) per Elymus hystrix plant in the common garden in relation to the proportion of stroma-bearing plants in the natural population. Plants were collected from one of four populations, D, T, J and L (locations given in Table 1), that differed in the proportion of stroma-bearing plants. The population proportion of stroma-bearing plants (number of plants with stromata/total number of plants) is shown on the x-axis. Error bars represent ± 1 SE.

Common garden 2: maternal plant families

Maternal plant families exhibited significant differences in the total number of stromata per plant in both years (2000: F6,109 = 7.05, P < 0.001; 2001: F6,109 = 4.94, P = 0.0002) (data not shown). Plants of maternal plant family 23 produced no stromata in either year. By contrast, every plant of maternal plant family 42 had between two and 15 stromata in 2000 and between one and 21 stromata in 2001. Stroma number per plant increased from an average of 2.93 per stroma-bearing plant in 2000 to 14.04 in 2001 (within subjects year effect: F1,109 = 4.03, P = 0.0405). However, the proportion of stroma-bearing infected plants was similar in the two years (39.9% in 2000 and 39.3% in 2001). In 2001, when total tiller number was recorded, stroma number was highly correlated with total tiller number per plant (Spearman's correlation coefficient117 = 0.711, P < 0.0001). However, maternal families still differed significantly in the proportion of stromata per plant (F6,109 = 5.31, P < 0.0001) when controlling for differences in plant size.

Maternal plant families differed significantly in the proportion of stroma-bearing inflorescences (2000: F6,109 = 6.68, P < 0.0001; 2001: F6,109 = 5.38, P < 0.0001) (Fig. 4). The proportion of stroma-bearing inflorescences was not significantly different across the 2 yr of the experiment (within subjects year effect: F1,153 = 0.56, P = 0.457). There was a significant interaction between stroma production across years and the presence or absence of stromata on the maternal plant (F1,109 = 4.67, P = 0.0329), but no interaction between maternal plant family and stroma production across years (year × maternal plant family: F6,109 = 1.70, P = 0.127).

Figure 4.

Mean proportion of stroma-bearing inflorescences (stroma-bearing inflorescences/(stroma-bearing inflorescences + fertile inflorescences)) per Elymus hystrix plant from different maternal plant families in 2000 (closed circles) and 2001 (open circles). Maternal plant families are arranged according to the presence or absence of stromata on the maternal plant. Numbers represent the names of the maternal plant families and sample sizes are in parentheses below. Error bars represent ± 1 SE.

There was no significant relationship between proportion of stroma-bearing inflorescences and the rate of vertical transmission to seeds (Spearman's correlation coefficient8 = −0.408, P = 0.316). Infected plants produced nearly 100% infected seeds (99.6% ± 0.04, mean ± SE) regardless of maternal plant stroma condition. Only two maternal plant families produced less than 100% infected seeds, one producing 91% infected seeds and another 99%.


There were significant differences among natural populations of E. hystrix in the proportion of plants bearing E. elymi stromata, an estimate of potential horizontal transmission by ascospores. There were also significant effects of year and a year by population interaction, suggesting that horizontal transmission can vary with environmental conditions. The proportion of stromata per plant was measured precisely in the common gardens and agreed with observations from natural populations, suggesting that plant or fungal genotypes determine transmission mode. Further, the proportion of stromata was similar within but differed among naturally infected maternal plant families, also supporting a genetic basis for transmission mode.

Variation in stroma production in natural populations

Year-to-year variation suggests that stroma production is affected by environmental conditions. Groppe et al. (2001) demonstrated that stroma formation increased in Epichloë bromicola infecting Bromus erectus following experimental fragmentation of populations (see also Meijer & Leuchtmann, 2000). Environmental differences between populations could lead to genetic differentiation of the grass–endophyte associations. For example, Hesse et al. (2003) found that endophytes of perennial ryegrass (Lolium perenne) were adapted to moisture conditions in their host populations. The data presented here are consistent with environmental differences among populations but argue also for a genetic contribution to transmission mode given the significant maternal family effects on stroma formation in the common garden. Other studies support a genetic basis for population differences in stroma production (White & Chambless, 1991; Bucheli & Leuchtmann, 1996; Kover & Clay, 1998).

Population infection frequencies

Differences in the proportion of stroma-bearing plants between natural populations could be explained merely by differences in infection frequencies if there were a positive relationship between the proportion of stroma-bearing plants and infection frequency. However, this was not the case. Although populations differed in infection frequency, those differences did not predict the proportion of stroma-bearing plants. In the first year there was marginally significant negative relationship such that populations with the highest levels of endophyte infection had the lowest proportion of stroma-bearing plants. In the second year there was no relationship between infection levels and proportion of stroma-bearing plants. Infection was quantified by leaf peels in 2000 and seed staining in 2004, and we have found that these measures are highly concordant. Population differences in proportion of stroma-bearing plants therefore result from differences in stroma production by infected plants and not the proportion of infected plants.

Populations characterized by predominant horizontal transmission and those characterized by vertical transmission might be expected to exhibit infection frequency differences. Although greater numbers of stromata could lead to higher infection levels in populations, the effect of transmission mode on population infection levels will depend on rates of successful transmission via both routes of infection. Furthermore, infection levels will depend on the costs and benefits to the host associated with horizontal and vertical routes of infection. Endophytes that are only transmitted horizontally generally have lower infection frequencies than vertically transmitted endophytes (Clay & Leuchtmann, 1989), perhaps because they reduce host competitiveness, because infected hosts produce no seed or because continuously changing host and endophyte genotypic combinations in each generation limit co-adaptation. However, stroma-producing endophytes might enhance host growth rate or survival, favoring their persistence (Pan & Clay, 2003). In related grass–endophyte associations with complete vertical transmission, such as found in Lolium arundinaceum and Lolium perenne, high infection frequencies are common, as predicted by Lipsitch et al. (1996). These associations are often mutualistic and population infection frequencies often increase with time (Clay, 1996; Clay et al., 2005). Unlike horizontal transmission, vertical transmission is independent of both host density and infection prevalence, so is unaffected when host populations are small and infection frequency is low (Jaenike, 2000). Vertical transmission is predicted to be the most important mode of parasite reproduction when infection frequencies are high, because there are few uninfected individuals to colonize by horizontal transmission (Lipsitch et al., 1996). The finding here of nearly 100% infection in several populations lacking stroma production suggests that vertical transmission may be an important means of endophyte reproduction, especially when infection levels are high.

Transmission mode differences in the common garden

Differences in transmission mode among plants were also apparent in two common garden experiments where all plants experienced similar environmental conditions. Plants from populations with proportionally few stroma-bearing plants had few stromata in the common garden, whereas those from populations with a greater proportion of stroma-bearing plants produced more stromata in the common garden. The lack of a significant population effect was probably a result of a lack of power to detect differences because a relatively small number of plants were collected initially and then plant deaths further reduced that number. Nevertheless, there was a significant correlation between the proportion of plants with stromata in the natural population and stroma production in the common garden, suggesting that population differences in stroma production had a genetic basis. In the mixed Epichloë–Brachypodium association, population differences in stroma formation were correlated with allozyme differences among endophytes (Bucheli & Leuchtmann, 1996). Another experiment suggested that both host plant and endophyte genotype determine stroma formation (Meijer & Leuchtmann, 2001).

Differences in stroma expression reflect differences in the plant or fungal genotypes of the infected maternal plant families. Most models of parasite evolution assume that the parasite controls its transmission mode, but the host genotype may also determine parasite transmission (Gandon & Michalakis, 2000). In addition to resistance to becoming infected, infected hosts may be able to alter the transmission mode of the parasite. For example, there was a genetic basis for host (mosquito) effects on parasite (microsporidian) transmission mode (Koella & Agnew, 1999). Host and endophyte genotypes were completely correlated here. The question of which partner (or their interaction) controls the differences in transmission mode is being investigated, but the results presented here indicate that stroma production is a genetically variable trait.

The relationship between horizontal and vertical transmission

The plants in the common garden experiment all acquired their infection by vertical transmission, as they were the seed offspring of naturally infected maternal plants. In spite of this and the fact that one half of the infected maternal plants from natural populations were asymptomatic, all but one maternal plant family in the common garden produced stromata. This indicates that virtually all infected plants maintain the capacity for horizontal transmission. The greatest difference among the maternal families was not in the presence or absence of stromata, but in the proportion of stromata produced by infected plants. The maternal plant families in the common garden exhibited not only continuous variation in the number of stromata but also in the proportion of stroma-bearing versus fertile inflorescences. Thus, maternal families differed significantly in their potential for vertical versus horizontal transmission.

In spite of abundant stroma production in the common garden, there was no evidence of horizontal transmission in this environment. During the second year of the experiment, none of the initially uninfected plants produced stromata or infected seeds, although they were growing in close proximity with spore-producing plants the previous year. However, horizontal transmission is an important part of endophyte reproduction in a related endophyte–woodland grass association with both spore and seed transmission (Epichloë sylvatica infecting Brachypodium sylvaticum). In that system there were high rates of horizontal transmission to uninfected plants when they were transplanted into natural populations with stroma-bearing plants, but no transmission to uninfected plants when they were transplanted into populations with infected, asymptomatic plants (Brem & Leuchtmann, 1999). Horizontal transmission to uninfected hosts occurred more commonly when plants were growing in the shade (Meijer & Leuchtmann, 2000). The lack of horizontal transmission here may indicate a need for specific environmental conditions for contagious spread which was not met in the common garden. Alternatively, some plants or tillers may have become infected by horizontal transmission and remained asymptomatic, despite our failure to detect the new infections in tiller and seed samples. Vertical transmission of the endophyte was less constrained in the common garden given that seeds dropped from experimental plants readily established. However, we did not investigate how common garden population characteristics would change over time with seedling recruitment.

All infected plants in the common garden, regardless of their proportion of stromata, produced between 91 and 100% infected seeds (average 99.6%). There was no relationship between stroma production and the ability to infect seeds. One hundred per cent infection occurred in maternal families with no stromata as well as in the families with the highest proportion of stromata. The two maternal plant families that produced < 100% infected seed were intermediate in their stroma production. If there was a trade-off between the two modes of parasite reproduction, we would expect that high stroma-forming combinations would produce a lower proportion of infected seeds. Alternately, it could be that only with a high level of stroma production, possibly associated with high within-host growth rates, are fungal hyphae able to vertically infect seeds, as was found with the fungus Atkinsonella hypoxylon infecting Danthonia compressa (Kover & Clay, 1998). However, in E. glyceriae infecting Glyceria striata there is no vertical transmission to seeds in the occasional fertile inflorescence (Pan & Clay, 2003). Overall, our results with E. hystrix suggest that vertical transmission to seeds is highly effective over a wide range of host and endophyte genotypes and phenotypes.


The results of this study, in both natural and common garden environments, demonstrate variation in the transmission mode of a plant parasite that is both vertically and horizontally transmitted. Transmission mode exhibited year-to-year variation in both environments, suggesting an environmental effect on transmission mode. However, consistency within and differences among maternal plant families in stroma production in the common garden, as well as similar stroma production in the common garden and in the natural populations, support the hypothesis that this variation has a genetic basis. The differences in transmission mode among maternal plant families spanned a continuum from pure vertical transmission, with only asymptomatic inflorescences and infected seeds, to a high proportion of stroma production and potential horizontal transmission. In spite of differences in stroma production, vertical transmission occurred in all maternal families. This variation indicates that endophyte transmission and, if correlated with transmission mode, virulence can respond to selection by biotic and abiotic factors.


The authors thank B. Fishman, S. Drake and W. Drake for assistance in the field. We also thank members of the Clay Lab (J. Alers-Garcia, L. Flory, J. Holah, J. Koslow, J. Matthews, A. Packer, J. Pan, J. Price, K. Reinhart, C. Richardson and J. Rudgers) for their advice and support. This research was supported by NSF grant DEB 0309240 to KC and TT.