Fungal endophytes of grasses: hybrids rescued by vertical transmission? An evolutionary perspective

Neotyphodium species are asexual ascomycetes growing as endophytes in cool-season grasses (Poaceae). They do not induce symptoms, and sporulate sparsely or not at all on their hosts; they are vertically transmitted through plant seeds (Fig. 1; cycle A). Neotyphodium spp. constitute fascinating symbiotic models offering various benefits to the host plants. In addition to accumulating toxic alkaloids (Bush et al., 1997) protecting against grazers and herbivorous insects (Omacini et al., 2001; Clay et al., 2005), Neotyphodium spp. also improve plant mineral nutrition (Lyons et al., 1990), tolerance to water-deficit stress (Elmi & West, 1995) and toxic soil compounds (Monnet et al., 2001; reviewed by Clay & Schardl, 2002). This symbiosis profoundly affects ecological performances of host plants and trophic webs, for example phytophagous insects (Omacini et al., 2001) or litter-degradation microbes (Lemons et al., 2005). Costs entailed in American and Australian pastures by endophytes of tall fescue (Lolium arundinaceum = Festuca arundinacea), that strongly reduce livestock performance, have been largely popularized (Ball et al., 1993). Several recent studies suggest that a continuum between purely mutualistic and pathogenic (i.e. detrimental to the host plant) associations might exist (Faeth & Sullivan, 2003; Müller & Kraus, 2005). An intriguing feature is the genomic origin of Neotyphodium endophytes. They are most often hybrids between pathogenic fungal species, and some have even undergone at least two successive hybridization events (Moon et al., 2004). Here, we focus on this unexpected evolutionary origin of Neotyphodium spp. We (1) briefly summarize evidence that they are often hybrids between pathogenic fungal species; (2) propose new explanations accounting for the unexpectedly high frequency of hybrids among the asexual Neotyphodium endophytes; and (3) tentatively describe a biological scenario where such hybridizations could occur. Our model emphasizes the role and consequences of hybridization in fungal evolution, and may apply to a much wider range of fungi. Epichloë spp. can reproduce asexually by invading their host's seeds (A), through a sexual cycle where ascospores are formed (B) on a stroma or ‘choke’ (inset picture) that destroys the inflorescence, and more rarely through asexual spores (C). Transmission is thus vertical (A) or horizontal (B, C). Neotyphodium species, endophytes derived from Epichloë spp., mostly carry out asexual reproduction through seeds (A), resulting in a dominant or exclusive vertical transmission. Conceivably, rare horizontal transmission could occur via asexual spores (C), but has never been demonstrated. Neotyphodium is a form genus representing the mitotic state of ascomycetes related to the Clavicipitaceae, an order including the ergot fungus Claviceps purpurea. Strictly asexual Neotyphodium spp. arose repeatedly from the sexual genus Epichloë, which causes ‘choke’ disease of grasses (Moon et al., 2004; Gentile et al., 2005). Epichloë spp. colonize grass tissues biotrophically (Fig. 1). Their sexual stage arrests development of the plant inflorescence (White et al., 1991) by producing a ‘choke’, namely a stroma surrounding the arrested inflorescence. Haploid conidia acting as male gametes are produced by the young stroma, but heterothallism does not allow any selfing. Following reception of compatible conidia from another thallus, plasmogamy takes place in the stroma, followed by formation of mature fruiting structures (perithecia) containing sacs (asci) of meiotic spores. These ascospores are forcibly ejected to infect new host plants (horizontal transmission, Fig. 1, cycle B). Conceivably, conidia formed on stromata or the epiphytic hyphae sometimes observed (White et al., 1996) might very rarely allow an asexual horizontal transmission (Fig. 1, cycle C). Depending on the host–fungus combination, stromata may occur on all inflorescences (thus annihilating seed production), or on a limited number of tillers, allowing some seed production (Pan & Clay, 2002). In the latter case, vertical transmission usually occurs as well, since fungal hyphae invade the seeds (Fig. 1, cycle A). At the opposite extreme, asexual Neotyphodium spp. cause little or (typically) no choke, and mostly transmit vertically through seeds and stolons (Fig. 1, cycle A). They thus do not cause symptoms and are nonpathogenic. Phylogenetic analyses of the genes tub2 (for beta-tubulin), act1 (for actin) and tef1 (for transcription elongation factor 1-alpha) have revealed an unexpected feature: most asexual Neotyphodium spp. display several gene copies with different phylogenetic origins among Epichloë spp. (Craven et al., 2001; Moon et al., 2004; Gentile et al., 2005; Fig. 2). They thus are interspecific polyploids (= allopolyploids) arising from two parental Epichloë spp., or more. From the 59 endophytes isolated by Moon et al. (2004) and Gentile et al. (2005), 44 were hybrids. In some Neotyphodium spp., different numbers of gene copies of divergent origin are found from one locus to another, indicating that the hybrids subsequently lost some redundant gene copies. Hybrid origin and heteroploidy (polyploidy and aneuploidy) have been confirmed in many endophytes by allozyme studies (Leuchtmann & Clay, 1990), amplified fragment length polymorphism (AFLP) studies (Tredway et al., 1999) and microsatellite analysis (Moon et al., 2004). Evolutionary dynamic of the Neotyphodium–Epichloë complex. Asexual Neotyphodium can arise directly from haploid Epichloë strains (on the left). Other asexual Neotyphodium spp. are hybrids that can be converted into heteroploids (= aneuploids); the diagram depicts a possible scenario for their origin (on the right). Up to now, hybridization has been suggested to occur in strains that would have first lost the ability to form stromata and thus transmit vertically and asexually. Hybrids would be secondarily selected for several reasons (Moon et al., 2002, 2004). First, hybridization would help mask or reverse effects of Muller's ratchet, that is, the irreversible accumulation of deleterious mutations that cannot be purged by recombination in clonal species. However, many biotrophic or pathogenic asexual ascomycetes are not hybrids, demonstrating that parasitic fungi can survive asexually, without hybridization (Taylor et al., 1999). Second, hybridization allows accumulation of genes for alkaloid production, a defence that improves the host's fitness and, owing to vertical transmission, the fungal fitness itself. Actually, more toxins or higher toxin levels improve the deterrent or toxic effect on herbivores (Wilkinson et al., 2000; Tanaka et al., 2005). However, the occurrence of true haploid strains among Neotyphodium spp. (Brem & Leuchtmann, 2003; Moon et al., 2002) suggests that hybridization is not necessary for emergence and survival of vertically transmitted, mutualistic endophytes. We do not ignore the fact that hybridization contributes to evolution of fungi and symbiosis. However, we suggest that answering the reverse question (next) provides a more convincing explanation of the apparent correlation between hybridization and mutualism + vertical transmission. Because interspecific hybrids with such features obviously have better fitness! Hybrids are both male- and female-infertile (Moon et al., 2004) and cannot produce ascospores (Schardl & Leuchtmann, 1999), perhaps because correct chromosome pairing for meiosis is not possible as a result of heteroploidy (Fig. 1, cycle B). If a hybrid still produces stromata, it can neither produce ascospores nor survive through seeds because of arrested development of the inflorescence. Since there is no evidence that conidia from stromata can infect new hosts, it does not reproduce. Any chromosome doubling to restore meiotic fertility would be useless, since diploid vegetative nuclei are unstable in ascomycetes (leading to the parasexual cycle, see later). Investment in stroma formation is thus counter-selected, as it decreases the fungal fitness. In other words, inability to form stromata is an evolutionarily stable strategy for hybrids. Once strict vertical transmission is achieved, the two partners’ fitnesses are linked, that is (1) any increase of the plant fitness (stolon and/or seed production) in turn increases the fungal fitness, and (2) once the fungus is beneficial to its host, any increase of the fungal fitness also profits the plant. Selection for mutualism by vertical transmission has been already discussed (Ewald, 1987). The very diverse positive effects of Neotyphodium spp. on hosts reported so far, ranging from growth traits to nutrition and protection (already discussed), can be viewed as evolutionary ‘bricolages’, recruiting any mechanism to increase plant fitness. Recent works, mostly in Festuca arizonica, suggest that some seed-transmitted Neotyphodium spp. might not be mutualistic. Several differences between partners might account for this. (1) It is conceivable that horizontal transmission can occur owing to conidia formation on vegetative hyphae (White et al., 2001; Fig. 1, cycle C), although such a transmission route is far from demonstrated in Neotyphodium spp. (Latch & Christensen, 1985). (2) Whereas the fungal fitness directly depends on seeds (and thus, ovule) production, the fitness of plant genes depends on both ovule and pollen production (Faeth, 2002); pollen production diverts resources from ovule production and therefore from fungal reproduction. These two factors (1) and (2) can subtly dissociate host and fungal fitness. (3) Plant sexual reproduction leads to new allele combinations at each generation, whereas the asexual fungal genome remains identical; changing genome of each partner at each generation might limit the evolution of mutualism (Faeth et al., 2004). Therefore, although many Neotyphodium endophytes are mutualistic (Fig. 2), antagonism can be selected. Our scenario (hybridization followed by selection for vertical transmission and mutualism) is not the only one leading to asexual Neotyphodium spp., but explains why all hybrids are asymptomatic and often mutualistic. Some haploid, apparently nonhybrid strains also occur among mutualistic, vertically transmitted Neotyphodium spp. (Moon et al., 2002), so that direct selection from haploid Epichloë spp. can occur (Fig. 2). For example, Neotyphodium lolii is clearly a derivative of Epichloë festucae, as indicated by multigene phylogenetic analyses (Moon et al., 2004). Might cryptic hybrids exist among these ‘haploid’Neotyphodium spp.? Indeed, heteroploid hybrids might undergo a progressive loss of chromosomes (or chromosome fragments) by drift, to eventually produce haploid nuclei – at this step, chromosome loss is counter-selected since there is no more genetic redundancy. This is the parasexual cycle described within ascomycete species by Pontecorvo (1956) and recently rediscovered in the human pathogenic ascomycete Candida albicans (Bennett & Johnson, 2003): haploid nuclei are produced after a vegetative karyogamy, via transient aneuploid states. If interspecific hybrids of Epichloë spp. subsequently haploidize, how can such secondary haploids be identified? Indirect controls of ploidy might exist: in hybrids, hyphae still produce conidia (Fig. 1, cycle C) whose size is larger than that of conidia from true haploid strains (Kuldau et al., 1999; Craven et al., 2001). Molecular markers provide easier testing methods; however, the probability that such haploids would go undetected can be calculated as follows. During gene loss, the probability of retaining the gene copy from a given parent for a given locus is 0.5, assuming random loss. For n genetically independent loci, the probability of looking exactly as one parent is thus 0.5n. As there are two parents, the probability, P, of looking genetically identical to one of them at n loci is thus P = 2 × 0.5n = 0.5n−1. If we assume no interchromosomic recombination, then each chromosome behaves as a single locus and, since Epichloë species tend to have five or six chromosomes (Kuldau et al., 1999), P is at best 6.25%. If interchromosomic recombination occurs, as in true parasexual cycle, then all markers behave independently and P is 25% for three loci and only 0.1024% for 10 loci. This risk is underestimated if there is little polymorphism between the two ancestral species, but so far, Epichloë species have been readily distinguishable by molecular polymorphism. Several loci are thus required to assess that an apparently haploid strain is not a secondarily haploidized hybrid. Typical analyses of Epichloë and Neotyphodium genotypes involve two or three sequenced loci, plus at least three and as many as eight multisatellite or isozyme loci (Schardl et al., 1994; Moon et al., 2004). It is remarkable that, despite the large number of interspecific heteroploid hybrids detected to date, there has never been any evidence of a haploidized hybrid. We can conclude either that haploidization rarely occurs, or that it involves selective loss of genetic material from one parent only. So, why are there so many hybrids among seed-transmitted Neotyphodium endophytes? Our scenario also predicts why hybrids predominate, although hybridization is not a parsimonious mechanism. Haploid endophytes directly evolved from Epichloe spp. might revert to form stromata and transmit horizontally (Fig. 2) because their euploidy allows meiosis and ascospore formation. Conversely, hybrids cannot, because they are heteroploid, and are thus more stable as vertically transmitted endophytes. Finally, as previously mentioned, hybrids produce more diverse and abundant alkaloids than sexual species, most of which produce low or undetectable amounts of alkaloids (Leuchtmann et al., 2000). Confined to a vertical transmission, hybrids can thus profit their hosts, and improve their host's fitness more easily than haploid endophytes. Does hybridization occur at the fungal vegetative state or during mating? Vegetative hybridization would involve hyphal fusion followed by nuclear fusion during endophytic growth. By contrast, sexual hybridization would first require exchange of conidia between stromata. For sexual Epichloë spp., flies of the genus Botanophila generally mediate this exchange. These flies are not strictly specific (Bultman & Leuchtmann, 2003), and may sometimes facilitate illegitimate interspecific matings, but this rarely if ever leads to ascospore production (Schardl & Leuchtmann, 1999). It is unknown whether such matings abort before or after formation of a diploid nucleus. Even if a diploid hybrid nucleus forms, it is unclear how to convert the diploid cell (that usually undergoes meiosis; Fig. 1) into a vegetative hypha. Out of a very large number of experimental mating attempts between all described Epichloë species, only one generated hybrids (Schardl & Leuchtmann, 1999). Furthermore, these sexual hybrids did not produce heteroploid Neotyphodium spp. because the progeny were haploid recombinants, and they had to be rescued from the stroma since ascospores did not eject (Schardl & Leuchtmann, 1999). Sexual hybridization between different Epichloë spp. thus appears problematic. A more likely scenario for hybrid origin is a vegetative nuclear fusion between two Epichloë spp. growing within a single host. Hyphal fusion is common among ascomycetes (Glass et al., 2004) including Epichloë spp. (Christensen et al., 2000). Hyphal fusions can lead to a heterokaryotic condition, whereby the two parental nuclei coexist in the same hypha. Although vegetative incompatibility barriers make heterokaryons between distantly related strains or different species unlikely in most ascomycetes, experimental evidence supports the ability of different Epichloë species to form heterokaryons, at least in culture (Chung et al., 1997). A vegetative event is further supported by the fact that some hybrids have three, not just two, parents (Moon et al., 2004), that is, one of the ancestors was already a hybrid. Considering that no hybrid has been observed to be male fertile (in test matings to their ancestral species) or to produce stromata (a prerequisite for female fertility), it is very unlikely that a hybrid could generate sexual hybrids. It follows that such an event must have occurred vegetatively, on a plant harbouring a hybrid that was super-infected by a second Epichloë species, likely via ascospore (Fig. 1). Whether such super-infections favour hybridization needs to be experimentally tested. Coinfection of individual plants with two endophyte species has been achieved experimentally, but competitive exclusion in tillers tends to eliminate one or the other (Wille et al., 1999; Christensen et al., 2000). Therefore, hybridization events must occur shortly after coinfection or in rare instances of long-term coinfection. The former scenario seems more likely, since no naturally occurring grass plants simultaneously bearing two species of the Neotyphodium/Epichloë complex were observed. An intriguing observation is that many hybrids live in grass species that are quite distantly related to all of their parents’ known hosts (Chung et al., 1997; Moon et al., 2004; Gentile et al., 2005). Furthermore, endophytes can cause a defence response if introduced into nonhost species, sometimes leading to spontaneous endophyte extinction after inoculation (Koga et al., 1993; Christensen, 1995; Christensen et al., 2000). To account for these observations, we propose that most hybridizations occur soon after a rare event of dual colonization of an ‘unusual’ host (i.e. a novel host) species (Fig. 2) by the two parental strains. Maladaptation of parental strains to that host could reduce competition and thus allow coexistence, hyphal fusion and survival of the resulting hybrid. Hybrids may be more fit and finally competitively exclude parental strains. Indeed, a ‘genomic shock’, including a transposon burst and new genetic regulations, arises when genomes of different species coexist in allopolyploids (Comai et al., 2003), producing new phenotypes (Osborn et al., 2003). Such changes make hybrids a good substrate for selection of genetic combinations more adapted to the unusual host. Indeed, a simple meiosis was shown to modify epistasis between genes and reveal new host compatibilities in offspring compared with their parents (Chung et al., 1997). Moreover, on an unusual host might also the parents from stromata, allow seed formation and for vertical transmission of the hybrid. An intriguing on this scenario is that an endophyte might colonize a new host by vertical transmission. hybridization is a common and evolutionary in plants that occurs very the form of in grasses & 2002). Such a might the survival of endophytes from the species by vertical transmission. by horizontally transmitted ascospores from another Epichloë species may then allow hybridization and formation of a new genomic more adapted to the the and role of this further experimental testing but may account for several For example, the grass arundinaceum is a hybrid of (= Festuca and a but the endophytes of and arundinaceum are unstable in each host it is fascinating (1) ancestors of Neotyphodium a hybrid common on has genomes related to of hybrid and endophytes in related and Festuca and the of grasses related to arundinaceum hosts many different hybrid species of endophyte et al., 1994; Schardl, It also be that host hybridization is the only for hybrid Neotyphodium to colonize a host it strictly and thus, the exclusive to undergo secondary To hybrids might be selected as a of with an unusual host plant. the experimental could test for hybrid formation and competition with parental strains on host and nonhost grass species, to better the origins of hybrid Neotyphodium spp. We suggest evolutionary (Fig. 2) that account for the emergence of hybrid Neotyphodium spp. and experimental A genomic of hybridization on plants that are unusual hosts for the parental species provides material from which to the new host can meiotic for vertical transmission and therefore this evolution only for endophytes of cool-season We that it might be only the of an First, many endophytic fungi for their hosts et al., 2003), and their is et al., 2000). Several fungi previously as were recently reported as asymptomatic such as et al., or the and spp. et al., Second, in many and pathogenic species, asymptomatic may result from genetic or & 2005). For example, a single can convert a pathogenic strain of into an & 1993). Conversely, a single can a mutualistic Neotyphodium strain into a et al., may thus often arise from many of these endophytes are vertically Does hybridization to the of vertically transmitted fungal by the of Does vertical transmission of these endophytes favour selection of mutualistic studies may to Neotyphodium endophytes from cool-season grasses provide a model for evolution among fungi. To and two for on of this