The relationship between vertically transmitted asexual fungal grass endophytes and their hosts is considered to be mutualistic. Results from agronomic field support this line of reasoning but recent studies have shown more variable results in natural systems. We investigated how high and low nutrient and water treatments affected biomass allocation patterns of endophyte‐infected and uninfected Festuca pratensis and F. rubra in greenhouse experiments over two growing seasons. Irrespective of infection status, both grass species showed improved performance on highly fertilized and watered soils. However, infected F. pratensis plants produced larger tillers than endophyte‐free plants on soil low in nutrients and water in the first growing season, although they (E+) otherwise showed decreased performance on nutrient‐poor soil. In low nutrient and water conditions, endophyte‐infected plants produced less tillers and had lower total biomass compared to uninfected plants, and displayed a negative phenotypic correlation between seed production and vegetative growth. The latter indicates costs of reproduction when the plant shares common resources with the fungal endophyte. However, endophyte infection status (E+, E−) interacted significantly with the soil fertilisation in terms of plant growth, having a stronger positive effect on growth in infected F. pratensis plants. In F. rubra, endophyte‐infected plants showed higher vegetative growth in fertilized and watered soils compared to uninfected plants. However, infected plants tended to produce fewer inflorescences. This had no effect on seed production, perhaps because seed production was partly replaced by asexual pseudovivipary. Contrary to the general assumption in the literature that fungal endophytes are plant mutualists, these findings suggest that the costs of endophytes may outweigh their benefits in resource limited conditions. However, the costs of endophyte infections appear to differ among the grass species studied; costs of endophytes were mainly detected in F. pratensis under low nutrient conditions. We propose that differences in response to endophyte infection in these species may depend on the differences in life‐history strategies and environmental requirements of these two fescue and fungal species and may change during the life span of the plant.

We examined the occurrence of vertically via host seeds transmitted endophyte infections of 14 grass species in natural populations in Finland and totally 97 agricultural cultivars of 13 grass species. Although endophyte infections were widespread in native grass species, overall endophyte occurrence and frequencies were lower than published reports have suggested. In natural populations, 10 out of 14 grass species examined harbor fungal endophytes in their seeds. The highest species‐specific mean incidences of endophyte infected plants in infected populations were found in Agrostis capillaris. Festuca arundinacea. F. ovina. F. Pratensis. F. rubra and Phleum pratense (67%, 98%, 29%., 42%. 32% and 33%, respectively). Mean incidences were < 20% in Dactylis glomerata. Deschampsia flexuosa. D. cespitosa and Elymus repens. and no infections were detected in Calamagrostis lapponica. C. epigejos. Alopecurus pratensis and Phalaris arundinacea. However, we detected a very high variation in infection incidences among natural populations and a large proportion of populations was, indeed, endophyte‐free. This supports the ideas that 1) endophytic fungi provide selective advantage of infected grasses to their uninfected eonspecifics in some habitats, and/or 2) fungi are occasionally transmitted horizontally by spores. In grass cultivars, endophyte infected seeds were detected only in F. Pratensis and Lolium perenne. and endophyte frequencies were either very high or very low. Cultivars of 11 other grass species were endophyte‐free.