Effect of drought on the growth of Lolium perenne genotypes with and without fungal endophytes

Authors


Abstract

  • 1Grass leaves are often inhabited by fungal endophytes that can enhance host growth. In some forage species, endophytes improve host resistance to, and recovery from, drought.
  • 2Our objective was to determine if the growth of genotypes of Lolium perenne L. was improved by endophytes during recovery from drought.
  • 3Thirteen infected genotypes were cloned into ramets. Half were treated with a systemic fungicide to eliminate the endophyte (E−); half were untreated and retained high endophyte levels (E+). In a glasshouse, half of all E− and E+ ramets were watered regularly, whilst half were exposed to a 2 week drought on two occasions, each followed by a 3 week recovery period.
  • 4After the first drought and recovery period, endophytes significantly reduced tiller production in the drought-stressed group.
  • 5After the second drought and recovery period, effects of drought on live leaf area and dry mass were highly dependent on host genotype, but not endophytes. The mean tiller mass of E+ ramets after drought was significantly less than that of watered E+ ramets, but this was not true in E− ramets. For six genotypes there was greater mass allocation to storage in the tiller bases of E− ramets after drought.
  • 6This perennial ryegrass population showed marked genotypic variation in the ability to recover from drought stress, but endophytes played little or no role in this ability. For some host genotypes there may be a metabolic cost of harbouring endophytes during environmentally stressful conditions.

Introduction

The leaves of widespread forage grasses (such as Festuca spp., Lolium spp.) are frequently infected by endophytic fungi in the genus Neotyphodium (formerly Acremonium;Glenn et al. 1996 ; White 1997). In addition to their importance in pasture communities, some of these endophyte-infected grasses also provide a major part of the vegetation cover on public grounds and suburban lawns ( Funk, White & Breen 1993; Funk, Belanger & Murphy 1994).

Lolium perenne L. (perennial ryegrass), arguably the most important forage/turfgrass in the world ( Jung et al. 1996 ), often contains the endophytic fungus Neotyphodium lolii (Latch, Christensen & Samuels) Glenn, Bacon, Price & Hanlin 1996 ( Hume & Brock 1997; Latch, Potter & Tyler 1987; Lewis & Clements 1986; Lewis et al. 1997 ; Neill 1940; Oldenburg 1997). Given the global importance of L. perenne in pastures and turf communities and its widespread distribution, it is critical to evaluate the potential role of endophytic fungi in affecting host growth processes under a variety of environmental conditions. For some species, such as the well investigated tall fescue (Festuca arundinacea Schreb.) endophyte system, the symbiosis appears to be mutualistic and the fungus enhances host growth and reproduction ( Belesky et al. 1987 ; Cheplick, Clay & Marks 1989; Clay 1990; Hill, Belesky & Stringer 1991; Rice et al. 1990 ). There is also evidence that the endophyte of tall fescue can improve host resistance to, and recovery from, drought ( Arachevaleta et al. 1989 ; Bacon 1993; West 1994; West et al. 1993 ), but this is not always the case for this species ( Elbersen & West 1996;Hill, Pachon & Bacon 1996; White et al. 1992 ).

It is now evident that endophyte–host interactions can be quite variable, depending both on environmental conditions ( Bouton et al. 1993 ; Clay 1990; Saikkonen et al. 1998 ; Siegel 1993) and host genotypes ( Belesky et al. 1987 ; Cheplick 1997, 1998; Elbersen & West 1996; Hill et al. 1996 ; Marks & Clay 1996; Rice et al. 1990 ). In a review of the physiology and drought tolerance of endophyte-infected grasses, West (1994) noted that ‘an enhancing effect of endophyte on ryegrass yield appears not to be as consistent as with tall fescue’. Thus the putative mutualism found in some endophyte–host systems may, in L. perenne, be highly contingent on environmental conditions ( Barker, Hume & Quigley 1997; Cheplick 1997; Cheplick et al. 1989 ; Clay, Marks & Cheplick 1993; Lewis 1992; Lewis et al. 1997 ; Marks & Clay 1990; Ravel, Charmet & Balfourier 1995). The symbiosis between L. perenne and its Neotyphodium endophyte may sometimes appear to be conditional and highly variable because of the individualistic nature of phenotypic responses which vary greatly between host genotypes ( Cheplick 1998).

The objective of the present study was to examine the effects of endophytic fungi on the ability of L. perenne genotypes to recover from severe drought stress. The working hypothesis was that endophytes would improve host growth during recovery from drought, but the extent of improvement would depend on the host genotype.

Materials and methods

Experimental system

The endophytic hyphae of N. lolii grow intercellularly within the stems and leaf sheaths of L. perenne. This endophyte is completely asexual and is transmitted only maternally within host seeds ( White, Morgan-Jones & Morrow 1993).

Seeds of L. perenne cv. Yorktown III were obtained from Loft’s Seed Company (Somerset, NJ, USA). Because this cultivar is highly infected, all genotypes used in this study began as seedlings from individual, infected seeds ( Hurley et al. 1996 ). Seedlings originating from separate seeds should constitute distinct genotypes because turfgrass cultivars are heterogeneous, self-incompatible populations of highly heterozygous plants ( Funk et al. 1993 ; Funk & White 1997). In addition, Sweeney & Danneberger (1997) have used RAPD markers to reveal that cultivars of perennial ryegrass, including Yorktown III, are composites of genotypes. Furthermore, Cheplick (1997, 1998) has shown that morphological parameters such as tiller production, leaf area and regrowth rates after clipping all show significant genotypic variation in L. perenne cv. Yorktown III.

generation of endophyte-free (E−) ramets

Several hundred seeds were originally sown into a flat containing a 2 : 2 : 1 mixture of topsoil, vermiculite and perlite on 30 December 1996. Two weeks later, 40 randomly selected seedlings were transplanted into deep, tubular pots (6·5 cm diameter × 25 cm depth) with the same soil mix. Such ‘conetainers’ provide good rooting depth and are commonly used to propagate turfgrasses ( Beard 1992).

Five months later, 20 genotypes were assessed microscopically to determine endophyte infection (see next section for technique). All genotypes except two were highly infected; 15 of the infected genotypes were selected for the experiment. In late November 1997 these were cloned into replicated ramets that were individually planted into tubular pots (4 cm diameter × 20·5 cm depth) containing a 3 : 1 mixture of fine sand and topsoil to permit rapid root proliferation ( Beard 1992) for the uptake of aqueous fungicide. All pots were maintained for 2 months in an incubator set on a daily cycle of 20 °C/12 h light, 10 °C/12 h dark. Fluorescent lights provided an average photosynthetic photon flux of 84·2 ± 7·52 µmol m−2 s−1 (±SD; n = 10).

Beginning on 30 January 1998, after all genotypes had produced multiple ramets (tillers), half the ramets of each genotype were subjected to fungicide treatment to eradicate the endophyte (E− ramets) using the systemic fungicide methyl 1-[butylcarbamoyl]-2-benzimidazolecarbamate (Benomyl, Dupont, Kansas City, KS, USA) at 2 g l−1 ( Cheplick 1998). Ramets were trimmed to 3 cm and soaked overnight in the fungicide solution; then each was planted into a plastic cup (207 ml) containing sand and topsoil (3 : 1) saturated with 40 ml fungicide solution. To encourage tillering, each ramet received 5 ml of Peter’s 20-20-20 N-P-K fertilizer (Grace–Sierra Horticultural Products Co., Milpitas, CA, USA) at 2 g l−1. Earlier work with infected L. perenne has shown that, although this fungicide may inhibit growth slightly while ramets are being treated ( Cheplick 1997), there is no lasting effect on subsequent growth months after plants are removed from the fungicide solution ( Cheplick 1997; Cheplick et al. 1989 ). Also, in our study the drought experiment began 4 months after fungicide treatment, using only newly formed daughter ramets (made by fungicide-treated parental ramets) that were controlled for intitial size (see Experimental protocol).

Ramets that were to remain infected (E+) were treated similarly to E− ramets, except that they were soaked in distilled water instead of fungicide. Both untreated and treated ramets were placed in a glasshouse for 2 months. The maximum and minimum temperatures (±SD) recorded on 17 randomly selected dates during the 2 months of fungicide treatment were 28·7 ± 3·2 and 19·2 ± 2·6 °C, respectively. No additional light was supplied beyond natural levels (≈1500 µmol m−2 s−1 on sunny days). Only newly produced tillers were used to assess endophyte infection levels.

Microscopic assessment of infection

To determine endophyte infection in untreated ramets and to assess the effectiveness of the fungicide in treated ramets, microscopic examination of leaf sections was necessary. Since the greatest concentration of endophytes is found within the leaf sheaths ( Welty, Azevedo & Cook 1986; White et al. 1993 ), 5 mm pieces of sheath, excised from ramets, were cleared in 70% ethanol for 24 h ( Hignight, Muilenburg & van Wijk 1993), stained with aniline blue–lactic acid (8–10 min), and examined at 400× ( Bacon & White 1994). After the 2 month fungicide treatment, infection was quantified by direct counts of endophytic hyphae (range 0–10) in an 800 µm diameter field of view ( Cheplick 1997) for 10 untreated E+ ramets and for 10–20 fungicide-treated ramets per genotype, in order to obtain sufficient numbers of E− ramets for the experiment.

Experimental protocol

For each genotype, 10 E+ and 10 E− ramets were planted separately into large circular pots (15 cm diameter, 14·5 cm depth) filled with a 1 : 1 mixture of topsoil and fine vermiculite to allow an additional 2 months of growth in the glasshouse via new tiller production. Because fungicide treatment is rarely 100% effective ( Cheplick 1997), only fungicide-treated ramets that were verified microscopically as endophyte-free were used for the E− group. Only 13 of the original 15 genotypes could be used in the experiment because most ramets of two genotypes were still infected after fungicide treatment.

Sixteen E+ and 16 E− ramets per genotype were planted in tubular pots (4 cm diameter × 20·5 cm depth) containing 1 : 1 topsoil and vermiculite from 8–15 June 1998. On the day of planting, 50 ml distilled water was added to each pot to saturate the soil mixture. To control for possible variation in size, each ramet was a single tiller and was weighed after trimming to 3–5 cm height and 3 cm of roots. Only ramets between 0·1 and 0·2 g FM were planted to standardize initial mass.

The experiment began with 208 ramets (13 genotypes × 16 ramets), 48 of which died or failed to grow within 2 weeks and were replaced. To encourage initial growth, 10 ml of the fertilizer described earlier (2 g l−1) were added on 14 and 23 July.

Half of the ramets were allocated to a drought treatment, whilst the other half served as a watered control group. Two days before the drought began, shoot size was non-destructively estimated as the total summed length of all tillers (TTL). Data from 60 extra ramets (30 E+ and 30 E−), not used in the experiment, showed that shoot DM was highly correlated with TTL (r= 0·75, P < 0·01) where shoot DM = 0·1987 (TTL) + 5·9289.

The experiment began on 29 July 1998. Soil in all pots was fully saturated with 30 ml distilled water. For the next 2 weeks, water was withheld from all droughted ramets; the control ramets were supplied with 30 ml water twice weekly. Droughted plants were re-watered (12 August) and allowed to recover for 3 weeks. A second 2 week drought was then imposed by withholding water again, beginning on 3 September. In the few days before the second drought, tiller numbers and TTL were recorded for all plants (droughted and control). Droughted plants were again re-watered (17 September) and allowed to recover for 3 weeks. The maximum and minimum temperatures (±SD) recorded on 13 randomly selected dates during the experiment were 27·6 ± 0·9 and 19·8 ± 2·6 °C, respectively.

Plants were harvested at the end of the second recovery period by clipping just above the soil level, and the live tiller number was recorded. The area of live, green leaves was measured with a leaf-area meter (LICOR, LI-3000, Lincoln, NE, USA). The dry mass of roots, shoots (tiller leaves and sheaths), and tiller bases was determined after drying at 60 °C for 48 h. Tiller bases (the stubble, sensuThomas 1980), extending slightly above and below the soil surface, had a stub length ( Thomas 1980) of ≈1 cm.

During the experiment, soil moisture was quantified using 40 additional E+ (n = 20) and E− (n = 20) ramets planted individually in tubular pots with similar soil in the same glasshouse as the experimental plants, and treated in the same way as droughted plants. Beginning on the first day of drought treatment (29 July), a soil-moisture probe (Lincoln Irrigation, Inc., Lincoln, NE, USA) was inserted each week to 10 cm depth in the soil in each pot, to obtain a calibration curve by which relative soil moisture readings on a scale of 1–10 were correlated with percentage soil moisture determined gravimetrically (r= 0·71, n = 61, P < 0·01). From this curve, meter readings were converted to percentages of moisture per soil DM ( Fig. 1).

Figure 1.

Time course for mean (±SE) soil moisture (%) at 10 cm depth in pots containing E+ (●) and E− (▪) ramets over the duration of the drought experiment. Each drought stress was 2 weeks and was followed by a 3 week recovery period. n = 20 per infection group.

Data analyses

Analysis of covariance ( ANOVA) was employed to examine the effects of major factors and their interactions on the response variables ( Underwood 1997). The major factors were treatment (watered or drought); infection status (E+ or E−); and host genotype. For the analysis of tiller numbers and TTL after the first drought and recovery period, initial FM of the original ramet planted was the covariate. For the analysis of variables recorded after the second drought and recovery period, the covariate was the number of live tillers or TTL immediately prior to the second drought. Because the covariates were often highly significant, least-squares means (adjusted for the covariate) ±SE are presented throughout ( Dowdy & Wearden 1991). To comply with ANOVA assumptions ( Underwood 1997), some data transformations were necessary: DM variables were log10-tranformed, including mean tiller size (= shoot DM/number live tillers) and root/shoot ratio, and percentage allocation to tiller bases (= DM of tiller bases × 100/total DM) was arcsine, square-root transformed. The GLM procedure of the Statistical Analysis System, Version 6·08 (SAS Institute, Cary, NC, USA) was used for all analyses.

Results

Microscopic assessment of infection

All ramets used for the E− group were verified microscopically as being free of the endophyte. For the remaining ramets in the E+ group, infection (assessed as the number of fungal hyphae at 400× for 10 ramets per genotype) varied from 2·67 ± 0·50 (±SE) for genotype L to 8·30 ± 0·42 for genotype H. Mean (±SE) infection across all 13 genotypes was 6·14 ± 0·39.

Recovery from the first drought

During the first 2 week drought, soil moisture declined from 24 to 12–13% ( Fig. 1). After the 3 week recovery period the number of live tillers depended on treatment (watered vs drought), genotype, and endophyte presence ( Table 1). There was significant genotypic variation in tiller production ( Fig. 2), but no interaction of infection (E+ or E−) with treatment. However, there was a significant infection × genotype interaction ( Table 1), indicating that the effect of endophytes on tiller production was influenced by host genotype, regardless of treatment ( Fig. 2). Post hoc comparisons showed that E+ ramets produced 7·99 ± 0·21 tillers in the watered control, whilst E− ramets had 8·41 ± 0·20 tillers (P = 0·16); however, in the drought treatment E+ ramets had significantly fewer tillers (7·11 ± 0·21) than E− ramets (7·81 ± 0·21; P = 0·02).

Table 1. ANOVA results for the number of live tillers and total tiller length after the first drought and recovery period. The covariate was initial fresh mass of the original ramet planted
 No. live tillersTotal tiller length
Source of variationdfMSFPMS (×104) FP
  1. INF, infection status (E+ or E−); GENO, genotype; TREAT, treatment (watered or drought-stressed).

INF132·157·150·00793·130·770·3805
GENO12295·8365·76<0·0001105·5825·96<0·0001
INF × GENO1212·582·800·00115·581·370·1763
TREAT155·1612·260·0005106·3626·16<0·0001
INF × TREAT12·090·470·49542·160·530·4667
GENO × TREAT125·821·290·21999·372·300·0077
INF × GENO × TREAT124·761·060·39564·371·070·3805
Covariate122·715·050·02537·551·860·1738
Error3564·50    4·07  
Figure 2.

Number of live tillers per genotype after the first drought and recovery period. Means (±SE) have been adjusted for the covariate (initial ramet fresh mass). n = 6–8 per bar.

For TTL, a non-destructive measure of size, both genotype and treatment were again significant ( Table 1). There was also a significant genotype × treatment interaction, but endophyte infection had no effect on TTL.

Recovery from the second drought

The second 2 week drought was more severe than the first, reducing soil moisture to 4–5% ( Fig. 1). This may have occurred because there were more sunny days during the second drought, which generally causes glasshouse soil to dry more rapidly (G. P. Cheplick, personal observation). After the 3 week recovery period, the number of live tillers and live leaf area were significantly affected by genotype, treatment and their interaction ( Table 2). Infection had no effect on tiller production and a marginally significant effect (P = 0·06) on live leaf area ( Fig. 3). Mean leaf area in the control was 40·0 ± 0·89 cm2 for E+ and 37·9 ± 0·91 cm2 for E− ramets; in the drought treatment it was 22·5 ± 0·95 cm2 for E+ and 21·1 ± 0·91 cm2 for E− ramets.

Table 2. ANOVA results for the number of live tillers and live leaf area after the second drought and recovery period. The covariate was the number of live tillers immediately prior to the second drought
 No. live tillersLive leaf area
Source of variationdfMSFPMSFP
  1. INF, infection status (E+ or E−); GENO, genotype; TREAT, treatment (watered or drought-stressed).

INF  1  1·55  0·54   0·4645   286·41  3·54   0·0609
GENO 12 19·18  6·65<0·0001   380·12  4·69<0·0001
INF × GENO 12  3·55  1·23   0·2595    71·99  0·89   0·5589
TREAT  1210·99 73·15<0·000127 685·77341·73<0·0001
INF × TREAT  1  5·39  1·87   0·1723    11·13  0·14   0·7111
GENO × TREAT 12 11·08  3·84<0·0001   155·64  1·92 0·0310
INF × GENO × TREAT 12  1·68  0·58   0·8572    50·20  0·62   0·8256
Covariate  1777·42269·54<0·0001   288·92  3·57   0·0598
Error3422·88  81·02  
Figure 3.

Live leaf area (cm2) per genotype after the second drought and recovery period. Means (±SE) have been adjusted for the covariate (number of live tillers prior to the second drought). n = 6–8 per bar.

Total DM was computed as the sum of the DM of roots, shoots and tiller bases. The only significant factors in the ANOVA of total mass were genotype and treatment ( Table 3). Analysis of root:shoot ratio showed a similar pattern: only genotype (F = 2·68, P = 0·0018) and treatment (F = 263·33, P < 0·0001) were significant. Infection had no impact on total mass ( Table 3), root:shoot ratio (F = 0·03, P = 0·87), or the mass of tiller bases (F = 0·70, P = 0·40). Dry root mass was significantly affected by genotype (F = 2·15, P = 0·01), treatment (F = 333·41, P < 0·0001), and genotype × treatment (F = 1·93, P = 0·03). The percentage allocation of mass to roots also varied with genotype (F = 2·37, P = 0·006) and treatment (F = 314·73, P < 0·0001). Averaged across all genotypes, allocation to roots was 57·8 ± 1·16% in the control, but only 39·2 ± 1·08% in the drought treatment.

Table 3. ANOVA results for total accumulated dry mass and the percentage allocation of dry mass to tiller bases after the second drought and recovery period. The covariate was total tiller length immediately prior to the second drought
 Total massPercentage allocation to tiller bases
Source of variationdfMSFPMSFP
  1. INF, infection status (E+ or E−); GENO, genotype; TREAT, treatment (watered or drought-stressed).

INF10·0026  0·120·72690·00490·580·4472
GENO120·0735  3·42<0·00010·04084·85<0·0001
INF × GENO120·0189  0·880·57120·01241·480·1305
TREAT17·2444336·54<0·00010·00230·280·5979
INF × TREAT10·0180  0·840·36100·01261·500·2219
GENO × TREAT120·0275  1·280·23010·01041·230·2573
INF × GENO × TREAT120·0149  0·690·75990·01812·150·0138
Covariate10·8106 37·66<0·00010·03934·660·0315
Error3410·0215  0·0084  

Two additional variables were derived from the dry mass data. The percentage allocation of mass to tiller bases (mass of tiller bases × 100/total mass) was calculated as a crude measure of resource storage. Genotype significantly affected allocation to tiller bases, and there was a significant three-way interaction term in the ANOVA ( Table 3). For some genotypes (e.g. G, I, L and R), allocation to tiller bases was greatest for E+ ramets (relative to E− ramets) in the watered control; however, in the drought treatment, E− ramets sometimes showed increased allocation to tiller bases compared to E+ ramets for some, but not all, genotypes (e.g. E, K, M, and N) ( Fig. 4).

Figure 4.

Dry mass allocation (%) to tiller bases per genotype after the second drought and recovery period. Means (±SE) have been adjusted for the covariate (total tiller lengths prior to the second drought). n = 6–8 per bar.

The second derived variable was mean tiller mass (shoot mass/number of tillers), which indicates the average size of a tiller. Genotype influenced mean tiller mass (F = 12·20, P < 0·0001), but infection (F = 0·26, P = 0·61) and treatment (F = 1·61, P = 0·21) did not. There was one significant interaction – that of infection by treatment (F = 3·96, P = 0·0473). The mean tiller mass of E+ ramets in the drought treatment (52·2 mg) was significantly less than that of E− ramets in both the drought treatment (56·4 mg) and the watered control (55·2 mg) ( Fig. 5).

Figure 5.

Mean tiller mass (mg) per infection group after the second drought and recovery period. Means (±SE) have been adjusted for the covariate (number of live tillers prior to the second drought) and averaged across genotypes. From left to right, n = 103, 94, 100 and 98. Different letters indicate means are significantly different (P < 0·05).

Discussion

Endophytic fungi and recovery from drought

Because the availability of soil water can be considered the single most critical factor affecting the productivity and persistence of cool-season forage grasses ( Frank, Bittman & Johnson 1996), considerable efforts have been devoted to exploring the effects of drought on forage growth and productivity (e.g. Frank et al. 1996 ; Smoliak 1956; Turner & Begg 1978; Volaire, Thomas & Lelievre 1998; Wilman, Gao & Leitch 1998). With the advent of research into the growth effects of endophyte-infected Festuca and Lolium species, the possibility of endophyte-mediated drought tolerance and resistance has been explored in some detail ( Arachevaleta et al. 1989 ; Bacon 1993; Barker et al. 1997 ; Buck, West & Elbersen 1997; Elbersen & West 1996; Elmi & West 1995; Hill et al. 1996 ; West 1994; White et al. 1992 ).

In the present study, regardless of infection status, water deficit caused declines in tiller production, tiller length, leaf area and dry mass that were evident upon recovery from drought. Such reductions in growth are typical of the responses of cool-season perennial grasses to water deficit ( Frank et al. 1996 ; Milnes et al. 1998 ; Passioura, Condon & Richards 1993; Volaire et al. 1998 ; Wilman et al. 1998 ).

Although perennial ryegrass is not generally considered to be a drought-tolerant species ( Jung et al. 1996 ), the cultivar studied here showed marked genotypic variation in ability to recover from drought. If L. perenne populations in pasture or turf consist of a mixture of genotypes ( Hayward & Nsowah 1969; McNeilly & Roose 1984), then there is the potential in habitats where periodic or seasonal drought occurs for selection for specific genotypes that are best able to persist and recover from significant soil water deficits. Although the specific morphological parameters conducive to recovery from drought remain elusive, genotypes varied greatly in the relative number and size (leaf area and dry mass) of individual tillers, which together could influence the relative water requirements of whole plants ( Kneebone, Kopec & Mancino 1992).

Despite the evidence showing that endophytes in some grasses, especially tall fescue, can improve host resistance to, and recovery from, drought ( Arachevaleta et al. 1989 ; Bacon 1993; Elmi & West 1995; West 1994; West et al. 1993 ), results of our study agree with those of others who have been unable to find a consistent effect of endophytes in perennial ryegrass on both morphological and physiological parameters ( Barker et al. 1997 ; Cheplick et al. 1989 ; Eerens et al. 1997 ; Elbersen & West 1996; Lewis 1992; Marks & Clay 1990). Nevertheless, broad-scale geographical trends in relation to climate can occur; for example, Lewis et al. (1997) found that levels of endophyte infection across 57 L. perenne populations in France were correlated with water-supply deficit, and suggested that summer drought conditions could impart a selection pressure in favour of infection. Within a single population such as that examined here, host genotype × endophyte interactions may sometimes render it difficult to detect a definitive average effect of endophytes in a genetically heterogeneous host population ( Cheplick 1998). However, in this experiment there were only a few detectable genotype × infection interactions, suggesting that endophyte presence had little differential impact on host genotypes in this cultivar, despite genotypic variation in morphology.

The benefits and costs of endophyte infection

It has become apparent that endophyte–host symbioses span a broad ‘continuum of interactions’ ranging from strong antagonisms to obligate mutualisms, depending to a large degree on host species ( Saikkonen et al. 1998 ). Even within a species, the endophyte–host relationship can vary depending on host (and endophyte?) genotype and environmental conditions ( Cheplick 1997, 1998; Cheplick et al. 1989 ; Clay 1990; Marks & Clay 1990; Siegel 1993). When attempting to discern the nature of the endophyte–host relationship the potential costs and benefits of infection need to be addressed.

In a set of experiments that examined the effects of the endophytes of L. perenne and F. arundinacea under variable soil nutrient levels, Cheplick et al. (1989) reported that the stimulatory effects of endophytes on host growth were most evident when nutrients were not limiting. Under extreme nutrient deprivation, they suggested that there may be a ‘metabolic cost’ to endophyte-infected hosts, possibly due to competition between host and endophyte for a limited supply of nutrients or photosynthate. In the present study, under severe drought there was some evidence that endophytes could reduce recovery growth, although effects were weak and sometimes genotype-specific. After the first drought and recovery period, E+ ramets had produced significantly fewer tillers than E− ramets. In addition, after the second drought and recovery period, mean tiller mass was significantly less only for E+ ramets in the drought treatment compared to E+ and E− ramets in the watered control, as well as droughted E− ramets. These trends may indicate reduced recovery growth in E+ plants due to decreased availability of the stored food reserves needed to initiate new, larger tillers after drought.

In forage grasses such as L. perenne, regrowth after environmental stresses is often dependent on non-structural carbohydrate reserves stored in tiller bases ( Donaghy & Fulkerson 1997, 1998; Hull 1992; Volaire et al. 1998 ). During water stress, non-structural carbohydrates may accumulate and facilitate rapid regrowth after the stress is relieved ( Frank et al. 1996 ). In the analysis of the percentage of dry mass allocated to tiller bases in this study, a significant three-way interaction indicated that the endophyte effect depended on both treatment and host genotype. For a number of genotypes that showed no difference in allocation to tiller bases between E+ and E− ramets in the watered control, there was a significant reduction in this variable after drought stress.

Although necessarily speculative, the possibility exists that the potential ‘costs of storage’ ( Heilmeier & Monson 1994) are exacerbated by energy-requiring endophytes under stressful conditions, resulting in reduced storage reserves available before and after the regrowth (recovery) period. Given the general acceptance of endophyte-enhanced drought tolerance for some well studied grasses such as F. arundinacea ( West 1994), it should be recognized that it may be premature to extrapolate to other endophyte–host systems where endophyte–mediated effects on hosts are much more variable and highly contingent on environmental conditions and host genotypes.

Acknowledgements

Thanks are extended to M. Pompei, Loft’s Seeds Inc., for providing the endophyte-infected seeds. D. P. Belesky, K. Clay, D. J. Gibson and two anonymous reviewers provided useful comments on the manuscript. This research was supported by an Undergraduate Summer Research Fellowship (A. Perera) from the Division of Science and Technology and the Honors Research Internship Program (K. Koulouris) of the Discovery Center at the College of Staten Island, City University of New York.

Received 11 October 1999; revised 27 March 2000;accepted 27 March 2000

Ancillary