The effect of the fungus Atkinsonella hypoxylon (Clavicipitaceae) on the reproductive system and demography of the grass Danthonia spicata

NewPhytol. (\9SA) 98,165-175 165

THE EFFECT OF THE FUNGUS ATKINSONELLAifyPOA^yLOiV(CLAVICIPITACEAE)ON THE

REPRODUCTIVE SYSTEM AND DEMOGRAPHYOF THE GRASS DANTHONIA SPICATA

BY K. C L A Y *

Department of Botany, Duke University, Durham, North Carolina 27706, USA

(Accepted 1 May 1984)

SUMMARY

The Balansiae (Clavicipitaceae, Ascomycetes) consists of five genera of fungi that parasitizegrasses, sedges and rushes. Most or all species produce alkaloids that are toxic to mammalianand insect herbivores. Previous reports suggest that infected grasses are more vigorous thanuninfected grasses and that tbe proportion of infected individuals is greater in older populations.In this study individual plants or genotypes of the grass Danthonia spicata (L.) Beauv. infectedby the fungus Atkinsonella hypoxylon (Peck) Diehl. were compared with uninfected genotypesfrom the same population. Danthonia spicata typically bears dimorphic chasmogamous andcleistogamous flowers but in the study population infected plants were partially sterilized,producing only cleistogamous flowers; uninfected plants produced chasmogamous and cleisto-gamous flowers. Two experimental comparisons were conducted by propagating infected anduninfected genotypes into clones, or ramets, which were then planted back into their native site.Infected ramets had higher survival and growth rates, but lower flower production, than unin-fected ramets both in random field transects and in field competition studies between D. spicataand the grass Anthoxanthum odoratum L. The results indicate infected individuals have highersurvival and growth rates but reduced fecundity compared to uninfected individuals.

Keywords: Symbiosis, grass, fungus, Danthonia, Atkinsonella.

I N T R O D U C T I O N

With the exception of mycorrhizas, there is relatively little information about theimpact of fungi on the ecology of natural plant populations. This paper presentsexperimental data on the effect of the endophytic fungus Atkinsonella hypoxylon(Peck) Diehl. on the reproductive system and demography of the grass Danthoniaspicata (L.) Beauv.

The Balansiae (Clavicipitaceae, Ascomycetes) is a group of fungi that parasitizegrasses and, to a lesser extent, sedges and rushes (Edgerton, 1919; Diehl, 1950;Kilpatrick et al., 1961). Four genera (Atkinsonella, Balansia, Balansiopsis andEpichloe) were originally described in this group by Diehl (1950) and the genusMyriogenospora was recently added (Luttrell & Bacon, 1977). The fungi growwithin or upon the aerial plant body but not in the roots (Neill, 1940; Diehl, 1950).

The fungi frequently sterilize their hosts by inhibiting floral development or byaborting developing flowers owing to the formation of a sclerotium aroundimmature inflorescences. The species that inhibit flowering produce stromata onthe leaves of their hosts. Both types of fruiting bodies produce conidia andascospores (Diehl, 1950). The common names 'choke' and 'black stripe' have

* Present address: Department of Botany, Louisiana State University, Baton Rouge, Louisiana, 70803,USA.

0028-646X/84/090165 + 11 $03.00/0 © 1984 The New Phytologist

• • f

i66 K. C L A Y

been applied to the diseases caused by the Balansiae (Diehl, 1950; Sampson &V^estern, 1959). The endophytes of tall fescue (Festuca arundinacea) and perennialryegrass {Lolium perenne) are exceptional; there are no external manifestations ofinfection, hosts are not sterilized, and the fungus is passed down from maternalparent to progeny through the seed (Neill, 1940; Siegel, 1983). These endophyteshave recently been assigned to the genus Acremonium by Morgan-Jones & Gam(1982). Acremonium corresponds to the asexual state of Epichloe typhina.

The Balansiae produce a variety of toxic alkaloids similar to those produced bythe closely related ergot, Clavicepspurpurea (Bacon, Porter & Robbins 1975; 1981;Porter et al., 1977, 1979, 1981). A number of cattle diseases result from grazinginfected grasses (Porter, Bacon & Robbins, 1974; Bacon et al., 1977; Hovelandet al., 1980). More recently, resistance to insect herbivory in perennial ryegrasshas been associated with infection by Epichloe typhina (Funk et al., 1983; Gaynor& Hunt, 1983; Mortimer et al., 1984).

Anecdotal reports in the literature suggest that infection increases host vigour.Bradshaw (1959) reported Agrostis tenuis infected by Epichloe typhina was sterilebut vegetatively vigorous and dominated where vegetative growth, as opposed toseed production, was at a premium. Diehl (1950) reported Danthonia compressainfected by Atkinsonella hypoxylon, and Cenchrus echinatus infected by Balansiaobtecta, exhibited increased vegetative vigour compared to uninfected plants.Similar observations on D. spicata were made by the author during the course ofunrelated research (Clay, 1983) and provided the stimulus for this study. The datareported here represent the first experimental attempt to measure demographicdifferences between grasses infected by Balansiae fungi and uninfected grasses. Inaddition, the effect of the fungus on the dimorphic reproductive system ofD. spicata is reported.

MATERIALS AND METHODS

The site of this study was a mown field located on the Duke University campus,Durham, NC, USA. The vegetation consisted of perennial herbs and grasses withD. spicata being one of the dominant species. The population of D. spicataconsisted of tens of thousands of individuals, many of which were infected by A.hypoxylon. Uninfected plants produce infiorescences 40 to 60 cm in height withterminal panicles of wind-pollinated chasmogamous fiowers and axillary spikeletsof cleistogamous fiowers in the leaf sheaths (Clay, 1983). In contrast, infectedplants produce aborted infiorescences, only 10 to 20 cm in height, topped by a hardgreyish sclerotium bearing conidia and ascospores (Diehl, 1950). All infiorescencesfrom a single plant are entirely infected or entirely uninfected; plants bearing bothhealthy and infected infiorescences were not observed. Infection can be determinedwhen infiorescences are not evident only by microscopic examination of leaf or stemtissue.

In the spring of 1982 the extent of infection was estimated by sampling 900 pointson a square grid (rows and columns 15 cm apart) located in the centre of theaforementioned population. At each point the closest fiowering individual wasexamined for infection; if no fiowering individual occurred within 8 cm of the pointit was recorded as being blank. Infected plants producing aborted infiorescenceswere considered to be fiowering despite the absence of open, visible fiowers. A totalof 817 plants were sampled in this manner.

In order to compare infected and uninfected individuals, 100 of each wererandomly selected from the grid described above and numbers of vegetative and

Fungal infection of Danthonia spicata 167

reproductive tillers were counted. One reproductive tiller, or inflorescence, wasrandomly selected from each individual and the number of chasmogamous andcleistogamous flowers was counted. This number was multiplied by the numberof inflorescences to estimate total potential reproductive output per individual.Differences between infected and uninfected plants were analyzed by one-wayanalysis of variance and the nonparametric Wilcoxon sign test.

Two experimental studies were undertaken to compare the survival, growth andreproduction of infected and uninfected D. spicata. Twelve large plants wererandomly sampled from the population during late summer, 1981. Danthoniaspicata is a caespitose grass, so it was assumed that each clump represented adistinct individual. Moreover, an individual was assumed to represent a distinctgenotype. Which individuals were infected was not known at this time.

In the flrst experiment, four of the genotypes were each divided into 72 ramets,or clones, each of one tiller. Seventy-two ramets per individual were planted backinto the same site in late fall in two separate 1 m transects (36 ramets per transect),a total of eight transects and 288 ramets. The vegetation was minimally disturbedduring transplantation. The following spring, when D. spicata flowered, thenumber of vegetative and reproductive tillers of each surviving ramet was counted.The ramets were counted again the next fall and the following spring. Growth wasdefined as current tiller number divided by initial tiller number and was thereforeequal to flnal tiller number because all ramets started out as a single tiller.Reproductive effort was defined as reproductive tiller number divided by initialtiller number. Differences in the percent survival and flowering were analyzed bychi-square tests, whereas the remaining variables were analyzed by nested analysisof variance with transects nested within genotypes.

In the second experiment, the remaining eight genotypes of D. spicata weregrown in competition with eight genotypes of the grass Anthoxanthum odoratum L.collected from the same site. Both grasses commonly occur together and bothflower synchronously. A complete diallel design was utilized; ramets of each ofthe eight D. spicata genotypes were grown with ramets of each of the eightA. odoratum genotypes and each of the 64 combinations was replicated flve times.To serve as controls, two ramets of the same genotype were grown together andeach of these 16 combinations was also replicated five times. The ramets wereplanted in sterilized soil collected from the field in 10 cm plastic pots. After 2 weeksin the greenhouse, in late fall, 1981, the pots were transported to the fleld site, thebottoms were cut off, and the pots were sunk flush with the ground. One replicateof each combination was planted in each of the flve separate blocks. The openbottom of the pot allowed root growth downward but the sides enforced proximityof the two ramets. The size of each ramet was determined at planting (most wereone to three tillers in size). The following spring the numbers of reproductive andvegetative tillers produced by each surviving ramet of D. spicata and A. odoratumwere counted. Statistical analysis of the results followed the same methodsdescribed for the previous experiment except that two-way analysis of variance wasutilized (blocks crossed with genotypes) rather than nested analysis of variance.

Continuous variables in both experiments were log transformed before analysisto meet normality assumptions. The nonparametric Wilcoxon sign test was usedto test for differences between infected and uninfected ramets in characters thatcould not be normalized by log transformation (Snedecor & Cochran, 1967). Onlyliving and flowering ramets were utilized in the analysis of growth and reproductiveeffort, respectively.

i68 K. CLAY

RESULTS

Population characteristics

Out of 817 plants randomly sampled, 125 (15 %) were infected. The infectedplants were larger than uninfected plants; most uninfected plants had more than50 tillers whereas most infected plants had less than 50 tillers (Fig. 1). Averagetiller number was 57 and 22 for infected and uninfected plants, respectively,a highly significant difference as determined by one-way analysis of variance(Table 1). However, the data were not normally distributed, even following logtransformation, so they were tested with the nonparametric Wilcoxon sign test

,11,11,11,11,11,1,11, I, I, ,11,-4.-4

8 -

4 -

(b)

0 20

JLT-T—I—r^-i—I—I—I—I—r

60 80 100 120 140 160 232 380Number of tillers

Fig. 1. Size distributions of (a) infected and (b) uninfected individuals of Danthonia spicatarandomly selected from the study population. Sample size equals 100 for each type.

Table 1. Characteristics of randomly selected plants from the study population ofDanthonia spicata, uninfected or infected by Atkinsonella hypoxylon

Character Infected Uninfected F-value P<

Tillers per plantInflorescences per plantChasmogamous spikelets perinflorescence

Chasmogamous flowers perplant

Cleistogamous spikelets perinflorescence

Cleistogamous flowers perplant

57-14±5-2717-94+1-670 00 + 000

0-00+ 0-00

2-36+ 0-07

64-09 ±971

(100)(100)(100)

(100)

(100)

(100)

22-07± 1-70 (100)5-25 + 0-88(100)4-49 + 0-,18(99)

129-33 + 14-62(99)

4-62 + 0-09(99)

67-77 ±9-58 (99)

61-39107-31

1105-04

931-42

394-98

2-86

0-00010-00010-0001

0-0001

0-0001

0-09

Means for each variable are followed by ± one standard error, sample sizes in parentheses. Differenceswere analysed via one-way analysis of variance.

Fungal infection of Danthonia spicata 169

{t = 3-94, P < 0-001). Infected plants produced more infiorescences (reproductivetillers) and more infiorescences per vegetative tiller than did uninfected plants(Table 1).

The nature of the infiorescence differed dramatically between infected anduninfected plants. Uninfected plants of D. spicata produced infiorescences withtwo types of fiowers; open, wind-pollinated, chasmogamous fiowers in terminalpanicles and closed, self-pollinated, cleistogamous fiowers in axillary spikeletssurrounded by the leaf sheath (also see Clay, 1983). In contrast, infected plantsproduced only cleistogamous fiowers (Table 1) and were therefore incapable ofoutcrossing. Sclerotia developed at the apex of the infiorescences mechanicallychoking, or mummifying, all of the chasmogamous fiowers and the cleistogamousflowers in the upper leaf sheaths. In total, uninfected plants produced an averageof 200 fiowers and infected plants produced an average of 65 fiowers.

Experimental demographyAll fiowering ramets from one genotype produced sclerotia, whereas no

fiowering ramets from the other three genotypes produced sclerotia, indicating oneof the four original genotypes was infected by A. hypoxylon. The fungus waspropagated along with its host. Thus 72 of the 288 ramets originally planted wereinfected.

There were more infected than uninfected ramets at each of the three censuses,although this difference was not significant (Table 2). Over 65 % of the infectedramets survived 18 months compared to 55% of the uninfected ramets. Thepercentage of infected ramets fiowering in 1982, and in both 1982 and 1983, wassignificantly greater than the percentage of uninfected ramets fiowering (Table 3).

Table 2. Survival of infected and uninfected ramets of Danthonia spicataover 18 months

Date Infected Uninfected Chi-square

12/81 72(100%) 216(100%) —5/82 60(83-3%) 141(65-3%) 3-33 NS12/82 48(66-7%) 118(54-6%) 1-36 NS5/83 47(65-3%) 118(54-6%) 1-07 NS

Number of surviving plants (percentages in parentheses) are presented with a chi-square test fordifferences in survival between infected and uninfected ramets.

Table 3. Percentage of ramets o/Danthonia spicdita. fiowering in each year, either year,and both years

Percent flowering

Year Infected Uninfected Chi-square

1982 80-70% (46/57) 16-41 % (21/128) 45-02 (P < 0-0001)1983 100-00% (47/47) 87-29% (103/118) 0-60 (NS)

1982 or 1983 95-08% (58/61) 74-48% (108/145) 2-26 (NS)1982 and 1983 59-02% (36/44) 8-28% (12/101) 45-28 (P < 00001)

Number of flowering ramets divided by number of living ramets in parentheses. Ramets alive either yearwere used for the or/and comparisons. Discrepancies in sample sizes were due to ramets scored dead in1982 but alive in 1983.

I7O K. C L A Y

The infected ramets were larger and produced more inflorescences and moreinflorescences per tiller than uninfected ramets in both years (Table 4). Thedifferences among genotypes were highly significant with respect to tiller andinflorescence number in 1982, whereas only inflorescence number was significantin 1983 (Table 5). The differences between transects within genotypes weresignificant for both variables both years. Because only one of four genotypes wasinfected and analysis of variance indicated there were significant differences amonggenotypes, additional contrasts were calculated in order to compare the oneinfected genotype with the three uninfected genotypes. The contrasts were testedagainst the transect within genotype mean square. Inflorescence number in 1982was significantly higher for ramets from the infected genotype (Table 5).

Table 4. Growth and reproduction of infected and uninfected ramets of Danthoniaspicata through two flowering seasons

TillersInflorescences

31

May

Infected

-37±0-25(57)-02 ±0-09 (57)

1982

Uninfected

2-23±O-12(128)"0-16 ±0-03 (128)

95

May

Infected

-66 ±1-10 (47)-38 ±0-75 (47)

1983

Uninfected

6-93 ±0-41 (118)3-69 ±0-27 (118)

Means ± one standard error are given with sample sizes in parentheses.

Table 5. Analysis of variance of tiller and inflorescence production (log transformed)through two flowering seasons

May 1982 May 1983

Tiller number Inflorescence number Tiller number Inflorescence number

Model DF MS F P < DF MS F P < DF MS F P < DF MS F P <

Genotype 3 0-43 810 0-0001 3 0-69 43-64 0-0001 3 0-20 211 NS 3 0-41 3-87 0-01Transect 4 0-53 10-02 0-0001 4 0-04 2-510-04 4 0-34 3-85 0-005 4 0-31 3-69 0-007

(Genotype)Error 177 0-05 — — 177 0-02 — — 157 009 — — 157 008 — —Contrast 1 109 206 NS 1 2-00 5032 0002 1 017 0-51 NS 1 034 1-08 NS

Infected vs.Uninfected

The infected vs. uninfected contrast was tested using the transect (genotype) MS as an error term.

Competition with Anthoxanthum odoratumAll flowering ramets from three of the genotypes produced sclerotia, whereas

none of the flowering ramets from the remaining five genotypes produced sclerotia,indicating three of the D. spicata genotypes were infected and five were uninfected.

Infected ramets were competitively superior to uninfected ramets when grownwith the grass A. odoratum. Survival and flowering were 100 % for infected ramets,but only 81 and 55 % for uninfected ramets, both significant increases (Table 6).Where two ramets of the sanrn- genotype were grown together, the infected rametshad higher survival and flowering than uninfected ramets, although neitherdifference was significant at the P < 0-05 probability level (Table 6).

Fungal infection of Danthonia spicata 171

Table 6. Survival and flowering of ramets 0/Danthonia spicata r̂ozyw in competitionwith Anthoxanthum odoratum and another ramet of the same genotype

Percent

Percent

survival

flowering

Interspecific competition(with Anthoxanthum odoratum)

Infected

100%(73/73)100%

(73/73)

Uninfected

81-03%(188/232)54-79%

(103/188)

Chi-square

7-42(P<0-01)

15-92( P < 0-001)

Intragenotypic competition

Infected

90-91 %(20/22)100%

(20/20)

Uninfected

83-93 %(47/56)61-70%(29/47)

Chi-square

0-09(NS)2-81

(P<0-10)

Percent flowering based on living plants only. Sample sizes in parentheses.

Growth and reproductive effort of ramets competing with A. odoratum wereconsistent with previous results; infected ramets had higher values than uninfectedramets (Table 7). Two-way analysis of variance of growth and reproductive effortindicated there were significant block, genotype and interaction effects on growth,and genotype effects on reproductive effort (Table 8). Contrasts between the threeinfected genotypes and the five uninfected genotypes were calculated in order todetermine whether ranmets from the three infected genotypes, considered together,were significantly different from ramets of the five uninfected genotypes. Contrastscould not be calculated from the two-way analysis of variance because someblock X genotype cells contained one or no observations, so one-way analysis ofvariance was performed with genotype as the main effect and the contrast wascalculated and tested over the error mean square. The infected versus uninfectedcontrast was significant for both growth and reproductive effort (Table 8).

Table 7. Growth {final tiller number/initial tiller number) and reproductive effort{inflorescence number/initial tiller number) of infected and uninfected ramets ofDanthonia spicata grown in competition with Anthoxanthum odoratum and another

ramet of the same genotype

Growth

Reproductive effort

Interspecific competition(with Anthoxanthum odoratum)

Infected

4-45 + 0-28(73)

1-65 + 0-13(73)

Uninfected

3-51+0-17(188)

1-07 + 0-07(103)

Intragenotypic

Infected

6-58 + 0-75(20)

2-24 + 0-33(20)

competition

Uninfected

4-67 + 0-52(56)

1-58 + 0-20(28)

Growth was calculated only for living plants and reproductive effort only for flowering plants. Means+ one standard error (sample sizes in parentheses) are presented.

The ramets grown under conditions of intragenotypic competition (two rametsof the same genotype per pot) were subjected to similar analyses. Only genotypehad a significant effect on reproductive effort at the P < 0-05 probability level,although block and the interaction were significant at the P < 010 probabilitylevel. Neither contrast, calculated and tested as above, was significant.

172 K. CLAY

Table 8. Analysis of variance of growth and reproductive effort (log transformed)

Model

BlockGenotypeInteractionError

ContrastInfected vs. Uninfected

BlockGenotypeInteractionError

ContrastInfected vs. Uninfected

DF

Growth

MS

Interspecific47

27222

1

47

1737

1

0-450-130-100-07

0-41

0-020-110-13—

0-001

F p <

]

DF

Reproductive effort

MS F P <

competition (with Anthoxanthum odoratum)6-802-041-54—

5-29

0-00010-050-05

—

0-02

47

25175

1

0-050-500-050-05

1-82

Intragenotypic competition0-181-041-18—

0-01

NSNSNS—

NS

47

1125

1

0-140-180-10—

0-006

1-039-560-99—

34-46

2-543-461-89—

0-08

NS0-0001

NS—

0-0001

0-070-010-09

—

NS

Contrasts were calculated from a different model. Details in text.

D I S C U S S I O N

Infection by Atkinsonella hypoxylon was associated with greater vegetative vigourand reduced fecundity of its host, D. spicata. Infected plants were significantlylarger in random field samples, and infected ramets had higher survival, floweringand growth rates in experimental plantings. However, infected plants bore partlyaborted inflorescences resulting in reduced flower production and loss of outcrossingpotential.

The differences between genotypes and infection were confounded in this studybecause the infected or uninfected status of plants was not manipulated. Allinfected ramets were derived from plants that were naturally infected prior toexperimentation, and all uninfected ramets were derived from uninfected plants.Previous workers have had little success in artificially infecting healthy plants(Diehl, 1950). In contrast, elimination of infection through the use of fungicideshas been successful, but the effects of the fungicide on the host plant are difficultto control (Latch & Christensen, 1982). The possibility that a pleiotropic genecontrols both susceptibility to infection and vigour, although unlikely, should notbe discounted. *

The experimental results, while based upon only four infected genotypes, agreewith field samples and are consistent and repeatable over a range of experimentalconditions and environmental microsites. Moreover, similar observations ofadditional populations of D. spicata and other grasses infected by Balansiae fungiclosely related to A. hypoxylon have been reported. Bradshaw (1959) suggested thatthe prostrate growth habit and vigorous vegetative spread of the grass Agrostistenuis infected by Epichloe typhina was advantageous where vegetative growth ismore important than seed production, as in pastures, where infected individualspredominated. Diehl (1950) reported that Florida pastures often contained highproportions of the grass Cenchrus echinatus infected by Balansia obtecta and that

Fungal infection of Danthonia spicata 173

infected plants were large in comparison with uninfected plants. He reportedadditional observations on many diverse grasses, including D. spicata, andconcluded infection was more prevalent in older populations. Personal observationsof other grass species infected by members of the Balansiae corroborate hisconclusion. The data presented here and the reports of other researchers belie thenotion that differences between infected and uninfected plants are due entirely todifferences between genotypes.

One possible mechanism by which infection confers vigour is by increasingresistance to herbivory (Funk et al., 1983; Mortimer et al., 1984). Perennialryegrass infected by E. typhina has been shown to be more resistant to a varietyof insect pests than comparable uninfected varieties. It is currently being investi-gated whether D. spicata and other grasses infected by Balansiae exhibit increasedresistance to insect herbivory.

The partial fertility of infected plants observed in the study population is at oddswith Diehl's account of the group and the general pathology of Balansiae/grassinteractions (Sampson, 1933; Diehl, 1950; Bacon et al., 1977). Most grassesinfected by Balansiae are sterile, either by abortion of developing inflorescencesor inhibition of flower development (Sampson, 1933; Diehl, 1950). Exceptionsoccur in perennial ryegrass and tall fescue (Neill, 1940; Harvey, 1983; Siegel,1983). Here the host grasses are fertile but the fungus is sterile, relying uponvegetative growth into developing seeds for its dissemination. In this studycleistogamous flowers were regularly produced by infected D. spicata. Seeds frominfected plants have been shown to give rise to infected plants, indicating infectionis passed down from parent to offspring (Clay, unpublished). No other host specieshas been previously reported to exhibit a similar maintenance of fertility wheninfected, although infected plants of the congeneric host D. compressa have alsobeen observed to produce cleistogamous flowers (unpublished data). The onlyother host oi Atkinsonella hypoxylon is Stipa leucotricha, an unrelated grass (Diehl,1950). It is perhaps fortuitous that all three host grasses oi Atkinsonella hypoxylonare characterized by the same dimorphic chasmogamous and cleistogamousreproductive system found only infrequently in the Gramineae (Dyksterhuis,1945; Connor, 1979; Campbell et al, 1983).

Infection by Atkinsonella hypoxylon has consequences that accrue from theenforced inbreeding and reproductive isolation of host plants and the verticaltransmission of the fungus. The fungus subdivides the host population byreproductively isolating infected plants. The population is further subdivided bycomplete and continual inbreeding of infected plants. A number of completelyinbreeding infected lines should eventually develop among the interbreedinguninfected portion of the population. Because infection is 'inherited' in a mannersimilar to a maternally inherited gene, selection may act upon traits influencingthe nature of the symbiotic association between host grasses and Atkinsonellahypoxylon. Adaptive gene complexes may develop without the disruptive effectsof gene flow from the uninfected portion of the population. The loss of infectionfrom plants, while not observed, may occasionally occur and would serve to retardthese processes.

The question of whether the relationship between grass and fungus is mutualisticor parasitic is important in this regard. The distinction between the two is notmerely semantic but depends on the relative fltness of infected plants comparedto uninfected plants. If the fltness of infected plants is greater than infected plants,mutations enhancing the association would be selected for; if lower, mutations

174 K. C L A Y

reducing the association would be selected for. The data presented here showinfected plants produce fewer flowers but have greater survival and growth rates.Infection decreases short term fitness, but long term survival may eventuallyoutweigh short term losses in fecundity. The ecological status of Atkinsonellahypoxylon may be parasite or mutualist but it certainly fits Bradshaw's description(of E. typhina infecting Agrostis tenuis) 'a curious form of symbiosis'.

ACKNOWLEDGEMENTS

I thank Terry Johnson for identifying Atkinsonella hypoxylon, Janis Antonovics,Steve Kelley and Johanna Schmitt for help in the field, and Meredith Blackwelland an anonymous reviewer for suggestions on improving the manuscript.

REFERENCES

BACON, C . W . , PORTER, J. K. & ROBBINS, J. D. (1975). Toxicity and occurrence oi Balansia on grasses fromtoxic fescue pastures. Applied Microbiology, 29, 553-556.

BACON, C . W . , PORTER, J. K., ROBBINS, J. D. & LUTTRELL, E . S. (1977). Epichloe typhina from toxic tallfescue grasses. Applied and Environmental Microbiology, 34, 576-581.

BACON, C . W . , PORTER, J. K., & ROBBINS, J. D. (1981). Ergot alkaloids biosynthesis by isolates oi Balansiaepichloe and B. henningsianna. Canadian Journal of Botany, 59, 2534—2538.

BRADSHAW, A. D. (1959). Population differentiation in Agrostis tenuis Sibth. II. The incidence andsignificance of infection by Epichloe typhina. The New Phytologist, 58, 310-315.

CAMPBELL, C . S., QUINN, J. A., CHEPLICK, G . P. & BELL, T . J. (1983). Cleistogamy in grasses. AnnualReview of Ecology and Systematics, 14, 411^41.

CLAY, K . (1983). Variation in the degree of cleistogamy within and among species of the grass Danthonia.American Journal of Botany, 70, 835-843.

CLAY, K . (1984). Transmission of Atkinsonella hypoxylon (Claricipitaceae) by cleistogamous seed ofDanthonia spicata (Gramineae) Canadian Journal of Botany (in press).

CONNOR, H . E . (1979). Breeding systems in the grasses: a survey. New Zealand Journal of Botany, 17,547-574.

DIEHL, W . W . (1950). Balansia and the Balansiae in America. Agriculture Monograph 4. United StatesDepartment of Agriculture, Washington, D.C.

DYKSTERHUIS, E . J. (1945). Axillary cleistogenes in Stipa and their role in nature. Ecology, 26, 195-199.EDGERTON, G . W . (1919). A new Balansia on Cyperus. Mycologia, 11, 259-261.FUNK, C . R. , HALISKY, P. M., JOHNSON, M . C , SIEGEL, M . R. , STEWART, A. V., AHMAD, S., HURLEY, R. H .

& HARVEY, I. C. (1983). An endophytic fungus and resistance to sod webworms: association in Loliumperenne L. Bio/ Technology, 1, 189-191.

GAYNOR, D . L . & HUNT, W . F . (1983). The relationship between nitrogen supply, endophytic fungus, andArgentine stem weevil resistance in ryegrasses. Proceedings of the New Zealand Grassland Association,44, 257-263.

HARVEY, I. C. (1983). The Lolium endophyte: return from anonymity. Proceedings of the New ZealandGrassland Association, 44, 234—236.

HovELAND, C. S., HAALAND, R. L. , KING JR., C . C , ANTHONY, W . B., CLARK, E. M . , MCGUIRE, J. A.,GRIMES, H . W . & HOLLIMAN, J. L. (1980). Association oi Epichloe typhina fungus and steer performanceon tall fescue pasture. Agronomy Journal, 72, 1064-1065.

LATCH, G . C. M . & CHRISTENSEN, M . J. (1982). Ryegrass endophyte, incidence, and control. New ZealandJournal of Agricultural Research, 25, 443^48.

LUTTRELL, E . S. & BACON, C. W . (1977). Classification oi Myriogenospora in the Clavicipitaceae. CanadianJournal of Botany, 55, 2090-2097.

MORGAN-JONES, G . & GAM, W . (1982). Notes on hyphomycetes. XLI. An endophyte oiEestuca arundinaceaand the anamorph oi Epichloe typhina, new taxa in one of two new sections oi Acremonium. Mycotaxon,15, 311-318.

MORTIMER, P. H., BARKER, G . M . , CAMPBELL, A. G., DI MENNA, M . E. & SMITH, G . S. (1984). Theassociation of Lolium endophyte and resistance to Argentine stem weevil. New Zealand VeterinaryJournal (In press).

NEILL, J. C . (1940). The endophyte of ryegrass {Lolium perenne L.). New Zealand Science and Technology,21, 280-291.

Fungal infection of Danthonia spicata 175

PORTER, J. K., BACON, C. W . & ROBBINS, J. D. (1974). Major alkaloids of a Claviceps isolated from toxicBermuda grass. Journal of Agricultural and Food Chemistry, 22, 838-841.

PORTER, J. K., BACON, C. W . , ROBBINS, J. D., HIMMELSBACH, D . S. & HIGAMN, H . C . (1977). Indolealkaloids from Balansia epichloe (Weese). Journal of Agricultural and Food Chemistry, 25, 88-93.

PORTER, J. K., BACON, C. W . & ROBBINS, J. D. (1979). Lysergic acid derivatives from Balansia epichloe andB. claviceps (Clavicipitaceae). Journal of Natural Products, 42, 309-314.

PORTER, J. K., BACON, C. W . , ROBBINS, J. D. & BETOWSKI, D . (1981). Ergot alkaloid identification inClavicipitaceae systemic fungi of pasture grasses. Journal of Agricultural and Food Chemistry, 29,653-657.

SAMPSON, K . (1933). The systematic infection of grasses by Epichloe typhina (Pers.) Tul. Transactions ofthe British Mycological Society, 18, 30-47.

SAMPSON, K . & WESTERN, J. H. (1959). Diseases of British Grasses and Forage Legumes. CambridgeUniversity Press, London.

SNEDECOR, G . W . & COCHRAN, W . G . (1967). Statistical Methods. Iowa State University Press, Ames, Iowa.SIEGEL, M . R. (1983). Mode of transmission of the fescue endophyte. Proceedings of the Tall Fescue Toxicosis

Workshop, Atlanta, Georgia, 48-51.