Molecular phylogeny of Acremonium and its taxonomic implications.

Acremonium is generally considered to be a highly polyphyletic form genus containing distantly related fungi. Sectional divisions within Acremonium distinguish the clavicipitaceous grass endophytes of sect. Albolanosa from the generally saprobic species of sections Acremonium, Chaetomioides, Gliomastix, and Nectrioidea. In an effort to assess the possible number of lineages currently placed within Acremonium and to determine which groups of sexual ascomycetes are phylogenetically affiliated with Acremonium species, maximum parsimony and neighborjoining analyses were performed using partial sequences of the nuclear small subunit ribosomal DNA (18S rDNA). Acremonium was shown to be a polyphyletic taon with affiliations to at least three ascomycetous orders: 1) most of the examined species from the sections Acremonium, Gliomastix, and Nectrioidea showed a relationship to the Hypocreaceae even though many of these species have never been associated with any teleomorph; 2) the grass endophytes of sect. Albolanosa and other taxa from the Clavicipitaceae formed a monophyletic group derived from within the Hypocreales; 3) the thermophilic A. alabamense of sect. Chaetomioides was derived from within the Sordariales. Acremonium alternatum, the tye species of the genus, was one of the species showing affiliation to the Hypocreaceae. In order to eliminate some of the heterogeneity within Acremonium while also emphasizing the unique biological, morphological, and ecological characteristics of the grass en doAccepted for publicauon February 8, 1996. 1 email: aglennC¡uga.cc.uga.edu. phytes, we are proposing that the anamorphs of Epichloë and closely related asexual grass endophytes be reclassified into the new form genus Neotyphodium. Phylogenetic and taxonomic considerations are also presented for other taxa.


INTRODUCTION
Members of the family Clavicipitaceae (Hypocreales; Ascomycotina) are well known pathogens of a diverse assemblage of hosts including grasses, sedges, other ascomycetes, and insects. These fungi are found throughout the tropical and temperate regions of the world (Diehl, 1950;White, 1994b). Rogerson (1970) listed 31 genera in the family, but eight of these are considered to be synonyms. The family is divided into three subfamilies with the insect and fungal pathogens variously placed in the Oomycetoideae and Cordycipitoideae (Diehl, 1950). The plant pathogens are by far the most widely studied and economically important members of the family and are classified as the subfamily Clavicipitoideae (Diehl, 1950). Diehl (1950) further divides the taxa of the Clavicipitoideae into the three tribes Clavicipiteae, Balansieae, and Ustilaginoideae. The fungus-host interactions of these taxa can be broadly categorized as being epibiotic or endophytic, systemic or localized, and heavily pathogenic, moderately pathogenic, or mutualistic.
Most of the grass pathogens are in the tribe Balansieae (Diehl, 1950) and are either localized to epibiotic reproductive stromata on leaves or inflorescences or form systemic endophytic infections intercellularly within their hosts in addition to forming external reproductive stromata. The Balansieae include the teleomorphic taxa Atkinsonella, Balansia, Balansiopsis, Epichloë, and Myriogenospora (Diehl, 1950;Luttrell and Bacon, 1977;Rykard et aI., 1984).
Myriogenospora is completely superficial without any penetration of the host epidermis (Luttrell and Bacon, 1977;White and Glenn, 1994), but Atkinsonella has some localized intercellular hyphae associated with the stroma (Leuchtmann and Clay, 1988;Mor-gan:Jones and White, 1989). Balansia contains a few epibiotic species (Leuchtmann and Clay, 1988; and Fren tz, 1993), but most species have a well de- veloped endophytic habit (White et aI., 1995). All species of the Epichloë typhina (Pers. : Fr.) TuI. group, including the apparent asexual derivatives of Epichloe", are intercellular grass endophytes (White, 1994b). Asexual grass endophytes were largely overlooked as being related to Epichloë until Bacon et aI. (1977) found intercellular hyphae growing within tall fescue. They referred to the fescue endophyte as a biotye of E. typhina based on concurrent in vitro examinations and the earlier report by Sampson (1933). The obscurity of these asexual endophytes was a result of the host grasses never showing any symptoms or signs of a fungal infection. For the anamorph of the fungus, Bacon et aI. (1977) applied the name Sphacelia typhina Sacco which was the binomial designated by Saccardo (1881) for the anamorph of E. typhina. However, S. typhina was later considered a nomen dubium by Morgan:Jones and Gams (1982) because of insufficient type materiaL Typhodium (Link, 1826) is another genus of uncertain application to the anamorph of E. typhina and is now considered a synonym of Epichloë (Hawksworth et aI., 1983). However, as a "convention of convenience," Diehl (1950) applied Typhodium as a form genus and used the term "tyhodial" when referring to the anamorph of Epichloë.
Based on in vitro similarities, Morgan:Jones and Gams (1982) placed the anamorphs of both the fescue endophyte and E. typhina in the form genus Acremonium. The authors erected the new sect. Albolanosa for these endophytic fungi with the in vitro characteristics of slow-growing white to yellow colonies, solitary phialides narrower at the base than the subtending hyphae, hyaline and smooth-walled conidia, and teleomorphs in the Clavicipitaceae. They stated that the exclusively solitary phialides of the endophytes preclude them from.'dassification in Verticillium sect. Prostrata which was restricted to those taxa, such as Cordyceps spp., that possess verticilate as well as solitary phialides (see also Gams, 1971). Morgan-Jones and Gams (1982) reported slight in vitro morphological differences between the anamorph of E. typhina and those of Holcus mollis L., Dactylis glomerata L., and Sphenopholis obtusata (Michx.) Scribn., which they treated as A. typhinum Morgan:Jones & W. Gams, and the tall fescue endophyte, which they Acremonium is a cosmopolitan, morphologically simple genus (see Gams, 1971). Acremonium alternatum Link: Fr. is the lectotye and produces conidia in chains or heads. The genus Cephaiosporium has often been distinguished from Acremonium because of the accumulation of conidia in slimy heads. However, Gams (1971)  untenable as a generic character and merged the two under the older Acremonium. The form genus Gliorn_ astix and monophialidic species of Paecilomyces were also merged in to Acremonium (Gams, 1971), thereby expanding its definition to comprise generally slowgrowing species with hyaline or pigmented conidia that are one-celled or exceptionally two-celled, in chains or slimy heads, and produced from onhophialides or basitonously branched conidiophores.
Acremonium is perceived to be a heterogeneous taxon because several morphologically distinct teleomorphic genera have Acremonium-like anamorphs.

Nectriopsis, npichloë, Emercellopsis, Mycoarachis, and
Nigrosabulum also have anamorphs classified in Acremonium (Malloch and Cain, 1970;Gams, 1971;Morgan:Jones and Gams, 1982;Samuels, 1988;Lowen, 1995). Gams (1971Gams ( , 1975 divided Acremonium into three sections: Acremonium, Nectrioidea, and Gliomastix. The majority of the species in these sections are saprobic in a wide variety of habitats, but some are plant parasites. Some species of sect. Nectrioidea have teleomorphs in the genus Nectria, but most of the species in the three sections do not have any known association to teleomorphs. Along with the establishment of sect. Albolanosa, Morgan:Jones and Gams (1982) also erected the new sect. Chaetomioides for anamorphs of ascomycetes within the family Chaetomiaceae that have, among other characteristics, short-aculeate to lageniform phialides. Acremonium alabamense Morgan:Jones, the type species of sect. Chaetomioides, is a thermophilic fungus originally isolated from fallen pine needles (Morgan- , 1974), and it is the anamorph of Thielavia terrestris (Apinis) Malloch & Cain (Morgan:Jones and Cams, 1982). Other species of this section are anamorphs of several Chaetomium species and cannot be distinguished without their teleomorphs (Morgan-Jones and Gams, 1982). Most recently, sect. Lichenoidea was erected for lichenicolous species (Lowen, 1995), and some of these species have hypocreaceous teleomorphs while others are not associated with any teleomorph.

Jones
Because of the possible heterogeneity of Acrernonium, the placement of the grass endophytes in sect. Albolanosa was done with no claim of phylogenetic relationship to the non endophytic species in the other sections (Morgan: Jones and Cams, 1982;'Nhite and Morgan-Jones, 1987). This classification scheme was based strictly on in vitro morphological similarities. Some disagreement with this classification of the endophytes was initially expressed, and it continues e See Samuels et al. (1991) and Cams and Van Zaayen (1982).

A. furcatum
Nectria vilior' (= A. berkeleyanum) to be debated (Latch et aI., 1984;Rykard et aI., 1984; Morgan:Jones et aI., 1992; Gams, 1995). The simple morphology of the anamorphs may have been derived multiple times in the evolution of fungi. However, the connection of so many Acremonium species to Nectria and other genera of the Hypocreaceae suggests that while there may be some heterogeneity created by sect. Albolanosa and sect. Chaetomioides, Acremonium may be more homogeneous than postulated.
The application of strict monophyly has been suggested for fungal systematics and taxonomy so as to reflect the phylogenetic history of a group of organisms (Vilgalys et aI., 1993). If a monophyletic classification scheme is to be the ultimate goal for any group of fungi, then the current delimitation of Acremonium is open to question. Molecular phylogenetic data are particularly helpful in resolving relationships of morphologically simple organisfns such as those in Acremonium. Therefore, using sequences of the nuclear encoded small subunit ribosomal DNA (18S rDNA), this project was undertaken to 1) assess the minimum number of separate lineages currently placed within the form genus Acremonium, 2) determine which groups of sexual ascomycetes are closely affiliated with Acremonium species in the current classification, 3) determine if Acremonium sect. Albolanosa is appropriate for classification of the grass endophytes, and if not, 4) propose a new taxon for this group.

MATERIS AND METHODS
Fungal isolates.-A total of fifteen taxa having anamorphs currently classified in Acremonium were selected from sections Acremonium, Albolanosa, Chae-tomioides, Gliomastix, and Nectrioidea for inclusion in the rDNA comparisons (TABLES I, II). Isolates from sect. Lichenoidea were not sampled because of the recent establishment of this newest section. All isolates of Acremonium were either tye or authenticated cultures, and additional taxa were either personally collected or obtained from other researchers (TABLE II).
Nucleic acid extraction.-For extraction of total genomic DNA from mycelium, all isolates were grown in M102 liquid medium (Rykard et aI., 1982) on a rotary shaker (200 rpm) at room temperature until adequate growth occurred (usually 1-2 wk). The mycelium was collected by centrifugation, and the pelleted tissue was washed once with sterile distiled water to remove excess medium. The tissue was then ground in liquid nitrogen and stored at -80 C until ready for nucleic acid extraction. Extractions were made from 0.2-0.5 g (wet weight) of ground mycelium.
The nucleic acid extraction procedure was a modification of Lee and Taylor (1990) in which the chloroform:phenol step was repeated at least once so as to remove as much cellular debris as possible. The extracted nucleic acid samples were diluted to provide solutions with a DNA concentration range of 0.1 to 1.0 ng f1L -I.
Polymerase chain reaction and sequencing.-Fifty f1L of each diluted sample was used as template for polymerase chain reactions (PCR) (Mullis and Fallona, 1987;Saiki et aI., 1988). For each amplification, a total reaction volume of 100 f1L was made containing diluted template, 10 mM Tris-HCI (pH 8.3), 50 mM  (Swofford, 1993) on a Macintosh Performa 6115CD. Alignment gaps were treated as missing data (GAP-MODE=MISSING). However, one gap was included as an additional character in the data matrix (absence and presence of gap were coded as 0 and 1, respectively). In total, 43 isolates were included in the analysis (TABLE II). As a result of the large data set, only heuristic searches were performed with the following options in effect: tree-bisection-reconnection (TBR) swapping algorithm, collapsing zero length branches, and saving all minimal length trees (MUL-PAR). Ten replications with random addition of taxa were performed for each heuristic search in order to find any additional islands of minimum length trees (Maddison, 1991). To measure the relative support and stability of the resulting clades, bootstrap values (Efron, 1982;Felsenstein, 1985) and decay indices (Bremer, 1988;Donoghue et aI., 1992)  Phenetic analyses were also performed on the aligned 18S rDNA sequences using MEGA v. 1.01 (Kumar et aI., 1993). Since MEGA isn't capable of evaluating binary code, the alignment gap was not included as an additional character in the data set.
For neighborjoining analyses (Saitou and Nei, 1987), the gamma (Kimura 2-parameter) distance method was used with complete deletion of all sites containing gaps or missing information. Bootstrapping of the neighborjoining tree was performed with 500 replications. Again, S. cerevisiae, T deformans, and C.
tropicalis were used as outgroup taxa.

RESULTS
Segments of the ssrRNA gene (18S rDNA) encompassing 937 bp were analyzed. Alignment of sequences was easily accomplished by direct examination due to the gene's conserved nature. An alignment gap of one base pair, common to the outgroup taxa and the Clavicipitaceae, was added to the data set as an additional character. Of the 43 taxa included in this study (TABLE II), sequences from 16 of these were obtained from either GenBank or other researchers. The remaining 27 were sequenced for this study. Our sequencing efforts focused primarily on the form ge- Based on these results, Acremonium is a polyphyletic taxon having species associated with at least three or four currently recognized ascomycete orders   1). Six conidial fungi Classified in Acremonium, two Emerceiiopsis species which have Acremonium anamorphs, and Nectria vilior Starbäck, which also has an Acremonium anamorph (see  . 2). MEGA's neighborjoining analysis gave a bootstrap value of 87% for the comparable clade (FIG. 3). ..

Among the Acremonium species having affiliation
to the Hypocreales is the type species of the genus, 3), and support from maximum parsimony analysis was only slightly weaker with a bootstrap of 85% and a decay index of + 3 (FIG. 2)   In general, there was a high degree of congruence between the resulting phylogenies of the cladistic (FIG. 2) and phenetic (FIG. 3)  -furcatum and the Microascales. These differences are probably inherent in the individual algorithms of the synapomorphy-based cladistical analysis and the similarity-based phenetic analysis and how they each affect the areas of the trees where the degree of support for the topology is limited.

New taxon and combinations.-Based on the 18S
rDNA sequence analyses presented here, the present classification of the anamorphs of Epichloë and related mutualists in Acremonium is untenable. Acremonium, as tyified by A. alternatum, appears to be restricted to the family Hypocreaceae. Epichloë and related genera of the Clavicipitaceae form a well circumscribed, monophyletic family that appears to be derived from within the Hypocreales. The monophyly of this family is supported by the unique morphology, ecology, and obligate parasitism of the Clavicipitaceae. The anamorphs of this family are unusual in that they are associated with stromatic tissue formed on a host (Diehl, 1950;Luttrell, 1980;Rykard et aI., 1984;White and Morgan:Jones, 1987;Leuchtmann and Clay, 1988;Morgan:Jones and White, 1989). Therefore, we are proposing the erection of a new genus to accommodate anamorphs of clavicipitaceous fungi that form a palisade of simple phialidic conidiogenous cells over the surface of a stroma and in culture produce simple, strictly aculeate phialides often lacking a basal septum and usually arising as lateral branches from aerial hYPJ:ae. As thus defined, this genus should include the conidial fungi currently classified in Acremonium sect. Albolanosa. While not included in our analyses, N. starrii and N. lolii have been shown to be closely related to N. coenophialum based upon maximum parsimony analysis of the two internal transcribed spacer (ITS) regions ofrDNA (Schardl et aI., 1991). Unfortunately, N. chilense, N. chisosum, and N. huerfanum were also unavailable to us for examination, and these three have yet to be characterized by DNA-based phylogenetic analyses. We do, however, propose the transfer of these species into Neotyphodium based upon the examinations and personal communications of James F. White, Jr. and Walter Gams.
The conidial stromata of species of Epichloë are easily and logically addressed by the teleomorphic nomen even though anamorphic binomials and varieties have been erected for the various species (Morgan:Jones and Gams, 1982;White, 1992). However, the abundance of asexual enaophytes that rarely or never produce teleomorphic stromata (clonal tye-III endophytes of White (1988) J means that a dual system of nomenclature must be maintained whereby the tye-III endophytes are classified in Neotyphodium and the sexually reproducing species are classified in Epichloë.
As defined, Neotyphodium is the most appropriate form genus for the tyhodial stage of Atkinsonella (Rykard et aI., 1984). However, the applicability of  . 'i pi iont is simi ar to Neotyphodium in that a palisade of phialidic conidiogenous cells are produced over the surface of a host-associated stroma, but the conidia are differe~t in being didymosporous (White, 1993b). White (1993b) suggested that Echinodothis and Epichloë were closely related and might deserv classification together in a separate tribe. While lackĩ ng statistical s~pport, cladistical analysis suggested that Echinodothis may be intermediate between Cor-dycef.s and Claviceps and is not closely related to Epichloe. (FI.G. 1). As with Atkinsonella, the holomorphic applicatlOn of E. tuberformis to its anamorph and teleomorph means there is no need to establish an ana-

morphic binomiaL
The Clavicipitales-Hypocreales relationship.-The Clavicipitales is a morphologically well circumscribed order of mostly parasitic ascomycetes possessing morphologically distinctive characters such as deliquescent lateral paraphyses, cylindrical asci with a thickened apical cap, and filiform ascospores that are often septate and may disarticulate into part-spores (Rogerson, 1970). However, the ordinal affinity of the clavicipitaceous genera has been evaluated and interpreted differently by several researchers (see Rogerson (1970) for a thorough historical review of the taxonomic literatureJ.
In recent years, the ordinal relationships among hypocreaceous and clavicipitaceous genera has again come into question as a result of phylogenetic evaluations using molecular data. Maximum parsimony analyses using the nuclear encoded 18S rDNA have indicated a monophyletic relationship exists between genera of the Clavicipitales and the Hypocreales, suggesting that all taxa may be classifiable under the single order Hypocreales with the monophyletic Clavicipitaceae being a separate family from the paraphy-Ie tic Hypocreaceae (Spatafora and Blackwell, 1993). Our results, which also utilized 18S rDNA, show the same general relationship (FIGS. 1-3). Additionally, gene phylogenies based on both the nuclear encoded large subunit ribosomal DNA (28S rDNA) and nuclear encoded orotidine-5' -monophosphate decarboxylase also support the derivation of the monophyletic Clavicipitaceae from within the Hypocreales (Rehner a~d Samuels, 1994b. All these gene phylogenies support the treatment by Kreisel (1969) with a single order incorporating the families Hypocreaceae (sensu Rogerson) and Clavicipitaceae. among hypocreaceous and clavicipitaceous genera, the same approaches are allowing systematists to propose phylogenetically informative holomorphic connections between anamorphs and teleomorphs (Schardl et aI., 1991;Rehner and Samuels, 1994a).
Such phylogenetic analyses are also generating proposals of novel intraordinal teleomorphic relation-, ships which are expanding the delimitation of the Hypocreales (Rehner andSamuels, 1994a, 1995; results reported herein).
Our approach to the form genus Acremonium was similar to that used to investigate other polyphyletic form genera (Rehner andSamuels, 1994a, 1995).
Our results (FIGS. 1-3) suggest Acremonium is indeed polyphyletic as currently circumscribed. Mfiliation of the tye species, A. alternatum, and other common species such as A. kiliense, A. strictum, A. chrysogenum, A. murorum (Corda) W. Gams, and A. rutilum W. Cams to the Hypocreaceae suggests that Acremonium should be restricted to anamorphs of only this family. Since many Acremonium species, including A.
alternatum, do not produce ascomata in vitro, applying the taxonomic criterion of relationship to the Hypocreaceae is currently not feasible based on morphological characteristics. In order to strictly reserve Acremonium as anamorphs of the Hypocreaceae, DNA sequence analyses would be needed for each of the species of Acremonium that do not have a known teleomorph. While perhaps justified, such an immense task is not yet practicable. However, it is possible that most of the orphaned Acremonium species may be phylogenetically affiliated with the Hypocreales, as suggested by FIG. 1 which shows six orphaned species of Acremonium from the sections Acremonium, Gliomastix, and Nectrioidea derived from within the Hypocreales. Of those sampled, only one orphaned species, A. furcatum of sect. Nectrioidea, is not derived from within the Hypocreales.
At present the best we can do to eliminate some of the heterogeneity within Acremonium is to remove those species which are known to be associated with teleomorphs that are not within the Hypocreaceae (Cams, 1995). Though not examined, some, if not all, species of sect. Lichenoidea would be expected to show affiliation to the Hypocreaceae since four of the nine species have hypocreaceous teleomorphs.
Among the Acremonium species that were examined by us, the grass endophytes (sect. Albolanosa) and A.
alabamense (sect. Chaetomioides) have teleomorphs that are in the Clavicipitaceae and Chaetomiaceae (Sordariales), respectively. Therefore, the generic placement of these anamorphs is in need of reevaluation. The morphological characteristics of A. alabamense are very similar to those of other 'true' Acremonium species in that it produces guttulate conidia in chains or heads from simple phialides, but A. alabamense is different in that it grows much more rapidly than other species and is thermophilic (Morgan- Jones, 1974;Morgan:Jones and Gams, 1982). Before any formal generic reassignment of this anamorph is proposed, a detailed study is needed to determine if other morphological characteristics exist which distinguish it from other species of Acremonium. However, because of the abundance of morphological and biological information that is available concerning the grass endophytes, the generic placement of the anamorphs of Epichloë and related grass symbionts is now ready to be addressed. A discussion of pertinent biological and, taxonomic information is presented below.
Numerous systems of higher categories for the Ascomycotina have been suggested and are summarized in Hawksworth et aI. (1983). Common to a few of these systems are the categorical divisions Plectomyc etes and Pyrenomycetes at varying hierarchical levels. These two divisions are based on characters of the ascomata. Plectomycetes includes those fungi having tyically globose, nonostiolate cleistothecia that produce asci at varying levels throughout the centrum (Fennell, 1973;Benny and Kimbrough, 1980), and Pyrenomycetes includes those fungi having flask-shaped, mostly ostiolate perithecia that produce asci from a single basal hymenium (Müller and von Arx, 1973). Suggestions have been made that the plectomycetes are a heterogeneous assemblage of morphologically similar fungi resulting from convergent evolution, and many plectomycetous genera are suggested to be derived from within the Pyre no mycetes (Malloch, 1981;Benny and Kimbrough, 1980). Recent molecular phylogenetic studies have confirmed that such a heterogeneity indeed exists. An initial maximum parsimony analysis indicated that there was a natural sister-group relationship between selected taxa corresponding well to the traditional ascomycete classes Plectomycetes and Pyrenomycetes (Berbee and Taylor, 1992). Such traditional plectomycetous taxa as Ascosphaera, Ajellomyces, Monascus, Talaromyces, and Thermoascus were found to form a monophyletic sister group to a monophyletic group of pyrenomycetes including Ophiostoma, Chaetomium, and Neurospora.
While a true higher level division appears to naturally exist between pyrenomycetes and some plectomycetes, molecular analyses are identifyng other "plectomycetes" that appear to be derived from within the pyrenomycetes. Roumeguerella and Heleococcum are cleistothecioid genera that have been previously placed in the Hypocreales (Rogerson, 1970;Malloch, 1981). This placement was supported by analysis of 28S rDNA sequences (Rehner andSa-380 MYCOLOGIA muels, 1994a, 1995). Mycoarachis (Rehner and Samuels, 1994a) and Emerceiiopsis (FIGS. 1-3), so far classified in the cleistothecioid family Pseudeurotiaceae, also show affinity to the Hypocreales. The Pseudeurotiaceae was suggested by Malloch (1981) to be related to the Diaporthales on the basis of centrum development, but some members of this family now appear to be phylogenetically related to the Hypocreales. This relationship is directly relevant to this study in that Mycoarachis, Emercellopsis, and other members of the Pseudeurotiaceae possess Acremonium anamorphs (Malloch and Cain, 1970). Heleococcum also produces an Acremonium anamorph (Udagawa et aI., 1995).
While ascomatal characteristics appear to vary dramatically between hypocreaceous genera, the possession of an Acremonium anamorph by many of these teleomorphs appears to be a phylogenetically informative character supporting their link to the Hypocreaceae. It is interesting to note that Wu and Kimbrough (1990) found much similarity in the ascogonial and ascogenous systems of Emerceiiopsis, Ascosphaera, and Monascus even though Emercellopsis is phylogenetically distinct from the more closely related Ascosphaera and Monascus (FIG. 1). Such similarity in ascomatal development despite the phylogenetic separation implies that these ascomatal characteristics may be subject to convergent evolution. Also, the anamorphs of Ascosphaera (= Chrysosporium) and Monascus (= Basipetospora) are distinctly different from the Acremonium anamorph of Emerceiiopsis.
Some of the other Acremonium-producing genera of the Pseudeurotiaceae that may prove to be hypocreaceous once molecular analyses are performed include Nigrosabulum, Hapsidospora, and Leucosphaerna (Malloch and Cain, 1970;Malloch, 1989). The inclusion of cleistothecioid forms within the Hypocreales is creating a need for reevaluation and refinement of the distinguishing characters defining this order.
Taxonomic considerations.-The fact that the teleomorph of A. alabamense is T terrestris of the family Chaetomiaceae is in opposition to the criterion of Acremonium being re~tricted to anamorphs of the Hypocreaceae. The thermophilic nature of A. alabamense, its rapid growth rate, and its truncated, dacryoid conidia were all considered by Morgan:Jones and Gams (1982) to be characters distinctive enough to warrant the creation of sect. Chaetomioides within Acremonium. While A. alabamense is tyically or thophialidic like most species of Acremonium, perhaps its unique morphological and biological characters are more distinctive at the generic versus sectional leveL Before any formal reclassification of A. alabamense is considered, more detailed morphological studies are needed to determine if other tangible characters exist that could be used for its generic distinction from Acremonium. A comparative analysis of conidium ontogeny of A. alternatum and A. alabamense may provide additional needed characters.
The potential taxonomic problems posed by A. furcatum are complex. Since its affiliation to an ascomycete order is currently vague, it must be retained in the genus Acremonium. If further molecular or morphological analyses should indicate a phylogenetic affiliation to the Hypocreaceae, then its current status would be maintained. If additional molecular analyses should confirm an affiliation to the Microascales as suggested by our results, or if its teleomorph is discovered and classified within the Microascales, a more phylogenetically appropriate classification for A. furcatum may then be justified.
The gene phylogenies presented here indicate that the Clavicipitaceae is a distinctive, monophyletic family derived from within the Hypocreales. Other phylogenetic analyses have also indicated this same relationship (Spatafora and Blackwell, 1993;Rehner and Samuels, 1995). The monophyletic nature of the Clavicipitaceae is substantiated by its unique ecology and morphology. Mating compatibility studies have shown that graminicolous species of Balansia, Atkinsonella, Echinodothis, and Epichloë are heterothallic, requiring the transfer of conidia (= spermatia) from a stroma of one mating type to a stroma of the opposite mating type (White and Bultman, 1987;Leuchtmann and Clay, 1989;White, 1993a, b;White et aI., 1995). Mating compatibility of Myriogenospora has not been examined. Conidia of Claviceps are infective and do not function as spermatia (Luttrell, 1980). Diehl (1950) emphasized the importance of conidial fructification for taxonomic and systematic evaluations of these fungi. Sphacelia, the form genus for the anamorph of Claviceps, is characterized by a palisade of doliiform to lageniform phialides produced along the convoluted surface of the developing sclerotium, and this palisade bears a wet mass of amerosporous conidia (Diehl, 1950;Luttrell, 1980). In contrast, the "typhodial" conidial fructifications of Epichloe, herein classified in the form genus Neotyphodium, produce dry masses of amerosporous conidia from a continuous, nonconvoluted palisade of narrowly aculeate phialides over the surface of the host-associated stroma (Diehl, 1950;White and Morgan:Jones, 1987). The anamorphs of Balansia and J\1yriogenospora are classified in the form genus Ephel- 1950; Rykard et aI., 1984;White et aI., 1995). Atkinsonella is intermediate in that it initially has a Neotyphodium-like anamorph which is followed by an Ephelis anamorph (Rykard et aI., 1984).
The close evolutionary connection between asexual grass endophytes and the sexually reproducing species of Epichloë has been examined from an ecological perspective (White, 1988), but a greater amount of attention has been given to molecular evaluations of relationships. The isozyme and molecular analyses previously performed on the asexual endophytes have mainly dealt with the variation and relationships that exist among some of the species. Leuchtmann and Clay (1990) found that isozyme profiles based on ten enzymes are virtually uniform throughout most isolates of N. coenophialum (= A. coenophialum). They also found that no distinction could be made between the Epichloë endophytes and the asexual endophytes, thereby suggesting that they jointly comprise a monophyletic group. Schardl et aI. (1991) performed maximum parsimony analyses on sequences of the internal transcribed spacer (ITS) regions of rDNA and showed that the asexual en dophytes formed a monophyletic group with Epichloë and that asexual species apparently arose from Epichloë on multiple occasions. Additionally, they found that sequence comparisons do not support some of the morphologically based species classifications of asexual endophytes. Based on maximum parsimony analyses of ITS sequences, An et aI. (1992) reported that at least two distinct evolutionary origins of Neotyphodium-endophytes from Epichloë had occurred in the single host species Festuca arizonica Vasey. The studies by Leuchtmann and Clay (1990) and Schardl et aI. (1991) are supportive of the existence of several different species of Epichloë (see also White, 1993a;White, 1994a;Leuchtmann et aI., 1994).
The monophyletic nature of the 6lavicipitaceae, the separation of Epichloëfrom A. alternatum, the obligate parasitism of Epichloë and its asexual derivatives, and the formation of a sporodochium-like palisade of phialides on an external stroma all indicate the distinct nature of Epichloë and support the reclassification of its anamorphs. The doubtful identity of Sphacelia as a form genus for these anamorphs (Diehl, 1950;Morgan:Jones and Gams, 1982) and the lack of any other valid or appropriate form genus means a new genus is needed to accommodate these fungi. Typhodium was erected by Link (1826), but his meager description is vague and confusing so there is uncertainty as to whether he was describing the anamorph of Epichloë or the teleomorph. Diehl (1950) felt that the conidial fructifications of Epichloë were significantly different from those of Claviceps and chose not to apply the form genus Sphacelia to the anamorph of Epichloë. As an informal "convention of convenience," Diehl (1950) applied Typhodium as the form genus for Epichloë conidial fructifications. In following Diehl's concept, we have proposed the form genus Neotyphodium for the an amorphs of Epichloë and its related asexual grass endophytes.
The establishment of Neotyphodium helps to alleviate some of the heterogeneity of Acremonium while emphasizing the unique, monophyletic nature of the grass endophytes. The employment of molecular data to substantiate this reclassification indicates the utility of DNA-based phylogenetic techniques in evaluating the taxonomic affiliations of morphologically simple taxa such as Acremonium. However, limitations do exist. For example, the basic morphology of A. alabamense is essentially the same as that of A. alternatum, but the two are affiliated with different ascomycete orders. More detailed morphological studies are now needed to determine if informative characters exist which could be used for their generic distinction. Results of phylogenetic analyses are useful as frameworks for further, more focused morphological and molecular comparisons. Such comparisons may help to alleviate some of the remaining heterogeneity of Acremonium. ACKNOWLEDGMENTS We are grateful to James F. White, Jr. for supplying us with several cultures and for his comments concerning the project and manuscript, to Joseph Spataora and Denise Silva for providing to us DNA sequences prior to their deposition into GenBank, and to Walter Gams and an unidentified reviewer for offering suggestions to improve the manuscript.