The family Ceratostomataceae (Winter 1887) includes nearly 100 species, often mycoparasitic and characterized by ostiolate and rostrate, or less frequently non-ostiolate, translucent ascomata, unitunicate and evanescent asci, brown or exceptionally hyaline, unicellular ascospores with a germ pore at each end, or less frequently with only one germ pore or a germ slit, and phialidic asexual morphs or bulbils. Currently, that family is included in the Melanosporales (Chaudhary et al. 2006, Zhang et al. 2006, Hibbett et al. 2007, Li et al. 2016, Schultes et al. 2017), although historically it had been placed in Aspergillales (Gaümann 1964), Hypocreales (Alexopoulos 1962, Spatafora and Blackwell 1994a, Rehner and Samuels 1995, Jones and Blackwell 1998, Zhang and Blackwell 2002) and Sphaeriales (Bessey 1950, Dennis 1968). This family comprises 11 sexually reproducing genera, i.e. Arxiomyces, Melanospora, Persiciospora, Pteridiosperma, Pustulipora, Rhytidospora, Scopinella, Setiferotheca, Sphaerodes, Syspastospora and Vittatispora. Melanospora, the largest genus of this family (more than 50 species), was established by Corda (1837) to accommodate Ceratostoma chionea and two new species, Melanospora zamiae and Melanospora leucotricha, with the former chosen later as the type species (Kowalski 1965). Melanospora is characterized by usually ostiolate ascomata with a long neck and a translucent, pale yellow to reddish brown ascomatal wall, and mostly smooth-walled, brown, ellipsoidal to citriform, rarely discoid or fusoid ascospores, with a depressed germ pore at each end, occasionally surrounded by a raised rim (Guarro et al. 2012). Related genera are Microthecium and Sphaerodes. The former was erected by Corda (1842) to distinguish Mi. zobelii from Melanospora spp. by the presence of non-ostiolate, usually immersed ascomata; and Sphaerodes was introduced by Clements (1909) to separate Melanospora episphaeria from Melanospora spp. by its reticulate ascospores. However, the generic boundaries between Melanospora and its relatives remained obscure. Doguet (1955) carried out a revision of Melanospora, synonymizing several species and transferring additional species from other genera, mostly from Sphaeroderma, which had been proposed by Fuckel (1877) and distinguished from Melanospora by the absence of an ascomatal neck. Doguet (1955) considered the production of a neck as a non-stable taxonomic character influenced by the nature of the substrate where the fungus grows, and segregated the genus in several sections on the basis of the morphology of the ascomata (presence or absence of neck, and its size when present) and ascospores (shape and ornamentation). The most comprehensive revision of Melanospora and related genera was carried out by Cannon and Hawksworth (1982), based mainly on the structure of the ascospore wall under SEM, resulting in the transfer of species of Microthecium to Melanospora and to Sphaerodes. However, recent molecular studies demonstrated that these two latter genera are polyphyletic (Zhang and Blackwell 2002, Fan et al. 2012, Li et al. 2016, Schultes et al. 2017). Other genera included in the family are: Arxiomyces, which produces ovoid to ellipsoidal ascospores with a rounded apex and a truncate base with a large sunken germ pore (Cannon and Hawksworth 1982, 1983); Persiciospora, characterized by ascospores with a pitted wall and a faint reticulation (Cannon and Hawksworth 1982); Pteridiosperma, with ascospores ornamented with longitudinal wing-like appendages (Krug and Jeng 1979); Pustulipora, distinguished by its ascospores with a germ pore at each end surrounded by a blistered, rarely cushion-like structure showing an irregular pustulate appearance (Cannon 1982); Rhytidospora, characterized by non-ostiolate ascomata with a cephalothecoid ascomatal wall (Krug and Jeng 1979); Scopinella, producing brown, cuboid-ellipsoidal ascospores with two prominent longitudinal germ slits (Cannon and Hawksworth 1982); Setiferotheca, which produces ascospores similar to those of Arxiomyces and ascomata with a crown of dark brown setae surrounding the ostiole (Matsushima 1995); Syspastospora, possessing ascomata with a long neck composed of parallel arranged hyphae and cylindrical ascospores with a large terminal slightly sunken germ pore at each end (Cannon and Hawksworth 1982); and Vittatispora, which produces ascomata similar to those of Syspastospora and citriform ascospores with a longitudinal, thick, hyaline ridge (Chaudhary et al. 2006). Practically all taxonomic studies on these fungi have been based exclusively on the morphological characterization of the reproductive structures of preserved fungarium specimens, since unfortunately due to their mycoparasitism, many of these fungi do not grow in pure culture or do not produce ascomata in absence of their hosts. On the other hand, obtaining reliable nucleotide sequences from members of the Melanosporales is also difficult because of the usually large amount of DNA of their hosts. Based on the study of several freshly-isolated soil-borne fungi and of reference and type strains obtained from various culture collections, we have re-examined the phylogenetic relationships of the most relevant genera of the Ceratostomataceae. Consequently, the genus Melanospora has been redefined, Microthecium has been re-established, and three new genera have been proposed.

The SSU phylogenetic study was based on an alignment of 1023 bp and produced a single ML tree (Fig. 1) inferred from a RAxML analysis. The members of the Melanosporales including our isolates were placed in a highly supported main clade (100 % BS / 0.99 PP), and the isolate CBS 137837, whose morphological features did not match any previously described taxon, occurred as a basal branch clearly separated from the other Melanosporales, which grouped together with a high support (95 % BS / 0.98 PP) and separated into three subclades. The first one (96 % BS / - PP), contained most of the isolates morphologically identified as Melanospora, Persiciospora and Sphaerodes, including the type and reference strains of Melanospora brevirostris, M. fimbriata, M. fusispora, M. levita, M. zobelii, Papulaspora sepedonioides, Pteridiosperma ciliatum, Sphaerodes compressa, S. fimicola, S. retispora, S. quadrangularis and S. tenuissima, without significant genetic variation among them. The second subclade (80 % BS / - PP) comprised the type strains of Sphaerodes ellipsospora and Sphaerodes singaporensis and a reference strain of Melanospora kurssanoviana, which resulted clearly separated from the other two, which grouped with high support (100 % BS / 0.98 PP). In the third subclade (100 % BS / 1 PP) were nested the type species of Melanospora (M. zamiae), the type strains of Melanospora verrucispora and Sphaerodes mycoparasitica, and reference strains of Melanospora damnosa and Melanospora tiffanii.

The lengths of the individual alignments used in the combined data set were 802 bp (LSU), 535 bp (ITS), 727 bp (act) and 846 bp (tef1), respectively, and the final total alignment was 2910 bp. In the ML tree derived from the RAxML analysis of the combined data set (Fig. 2), the Melanosporales were highly supported (100 % BS / 1 PP) and subdivided into seven lineages. The first clade (89 % BS / 1 PP; Clade Microthecium) grouped all our isolates, with the exception of CBS 137837, and type or reference strains of Melanospora fimbriata, M. fusispora, M. levita, M. zobelii, Papulaspora sepedonioides, Pteridiosperma ciliatum, Sphaerodes compressa, S. fimicola, S. retispora, S. quadrangularis and S. tenuissima. All the fungi belonging to this clade have non-ostiolate ascomata, or when a neck is present, it is short and composed of angular cells similar to those of the ascomatal wall. Also, bulbils (microsclerotial-like asexual propagules) are present in most of these species. In spite of the high morphological variability shown by members of this clade, the loci used in the phylogenetic analysis were not able to separate the species from each other. The second clade (100% BS / 1 PP; Clade Melanospora) comprised the type species of Melanospora, M. zamiae, the type strain of M. verrucispora and a reference strain of M. damnosa. The members of this clade produce ostiolate ascomata with a long neck composed of hyphae irregularly arranged and ending in a crown of setae. In addition, an asexual morph is commonly present, which is characterized by solitary, sessile, flask-shaped phialides producing from rounded to ellipsoidal conidia. The third lineage comprised only a reference strain of Melanospora kurssanoviana, which failed to sporulate in pure culture. The fourth clade (90 % BS / 0.98 PP; Clade Dactylidispora) was composed of the type strains of Sphaerodes ellipsospora and S. singaporensis, both characterized by ascospores with a raised rim surrounding the germ pores. Finally, the isolate CBS 137837 and the type strains of Melanospora subterranea and Vittatispora coorgii formed three independent branches. The isolate CBS 137837 produces globose, non-ostiolate, densely setose, dark ascomata and smooth-walled ascospores with a depressed germ pore at each end, while the other two species of this clade also possess morphological features unique in the Melanosporales, e.g. ascospores with indistinct germ pores in M. subterranea and with a longitudinal, thick, hyaline ridge in V. coorgii.

We have revised the taxonomy of relevant members of the family Ceratostomataceae based on the analyses of the SSU, LSU, ITS, act and tef1 nucleotide sequences. This study strongly supported the order Melanosporales proposed by Zhang and Blackwell in 2007 (Hibbett et al. 2007). The phylogenetic inference showed seven lineages corresponding to the genera Dactylidispora, Echinusitheca, Melanospora, Microthecium, Pseudomicrothecium and Vittatispora, and to Melanospora kurssanoviana. Our results agree with previous studies (Zhang and Blackwell 2002, Fan et al. 2012) which already suggested and demonstrated that the ornamentation of the ascospores under SEM, a feature traditionally used to delimit most of the genera in the Melanosporales, is not useful for estimating phylogenetic relationships among these fungal taxa. Similarly, the morphology of the ascospores is of weak taxonomic value and a poor predictor for the generic delimitation of members of the family Sordariaceae, resulting in the synonymy of two relevant genera, i.e. Gelasinospora and Neurospora (Dettman et al. 2001, García et al. 2004, Nygren et al. 2011). In our study, two of the largest genera of the Melanosporales, Melanospora and Microthecium, grouped species with both smooth and ornamented ascospore walls. By contrast, a phylogenetic study of the Lasiosphaeriaceae (Miller and Huhndorf 2005) revealed that the morphology of the ascomatal wall was more phylogenetically informative than that of the ascospores, with several new genera proposed (i.e. Immersiella) or emended (i.e. Lasiosphaeria, Lasiosphaeris and Schizothecium) (Miller and Huhndorf 2004, Cai et al. 2005). Here, the erection of the new genus Echinusitheca is a clear example of the relevance of the ascomatal morphology in the taxonomy of these fungi, and in fact this taxon together with Arxiomyces and Scopinella are the only genera in the Melanosporales that show dark semi-transluscent ascomata. In this context, although Echinusitheca has ascospores similar to those of Melanospora and Microthecium, this genus constitutes one of the lineages phylogenetically most distant within this order.

Another lineage considerably distant from the other members of the Melanosporales is constituted by the clade represented only by the species Melanospora kurssanoviana, suggesting that this fungus could represent a new genus. However, this new taxon is at this moment not proposed because its colonies, in spite of attempts to induce sporulation, remain sterile and a detailed morphological study was not possible. The infertility of the cultures is probably due to the fact that an important part of the members of this fungal group show a peculiar habitat developing a certain degree of mycoparasitism and requiring the presence of the host to complete the biologic cycle and develope reproductive structures. The mycoparasitism of Melanospora, Syspastospora and the species previously placed in Persiciospora and Sphaerodes has already been demonstrated by numerous authors (Doguet 1955, Calviello 1973, Jordan and Barnett 1978, Harveson and Kimbrough 2000, 2001), and this ability has been exploited in the biocontrol of phytopathogenic fungi (Vujanovic and Goh 2009, Goh and Vujanovic 2010, Kim and Vujanovic 2016, 2017).

The genus Sphaeronaemella, which is characterized by pale and translucent ascomata, was thought to be related to Melanospora (Cannon and Hawksworth 1982). However, we do not agree with this relationship because it differs from the Melanosporales in the production of hyaline ascospores, as opposed to the pigmented ones in the members of that order. By contrast, our results correlate with those of other authors that demonstrated a closer phylogenetic relationship of this genus with the members of the order Microascales (Spatafora and Blackwell 1994b, Hausner and Reid 2004). In fact, our SSU tree seems to indicate that Sphaeronaemella could represent a new family of the Microascales; however, further studies including more taxa and additional genes are needed to more accurately confirm its taxonomic status.

The placement of our isolate of Persiciospora japonicum in the Microthecium clade once more demonstrated that the ornamentation of the ascospores, which is pitted in Persiciospora spp., is of poor taxonomic value, and consequently all the species of Persiciospora should be transferred to Microthecium. As it was above mentioned, the species of this latter genus show a typical cellular ascomatal neck which is also present in Persiciospora and constitutes a common feature in both genera. Surprisingly, in some previous phylogenetic studies, the species of Persiciospora were placed in the Hypocreales, closely related to Nectria (Zhang and Blackwell 2002, Maharachchikumbura et al. 2015, Schultes et al. 2017). However, this could be probably explained by a possible contamination of the cultures of Persiciospora spp. with an hypocrealean host (Fan et al. 2012). The same situation may have occurred with the cultures of Scopinella and Syspastospora, which led to a possible erroneous classification of both taxa in the Hypocreales (Zhang and Blackwell 2002, Chaudhary et al. 2006, Fan et al. 2012, Maharachchikumbura et al. 2015, Schultes et al. 2017).

Pteridiosperma ciliatum, a member of the Melanosporales with ascospores ornamented with longitudinal wing-like ridges that anastomose each other to form a well defined reticulum (a distinctive feature of Pteridiosperma), was also found in the Microthecium clade, proving once again that the ascospore ornamentation is not phylogenetically informative. Consequently, we have synonymyzed the genus Pteridiosperma with Microthecium since Pteridiosperma spp. show non-ostiolate ascomata, or if ostiolate, they show a short neck composed of angular cells, which are typical morphological characteristics of Microthecium.

Another genus that our results demonstrated should be synonymized and included in Microthecium is Sphaerodes because its type species, S. episphaerium, shows morphological features (non-ostiolate ascomata) that fit with the current circumscription of that emended genus. Most of the species of Sphaerodes, with the exception of S. ellipsospora and S. singaporensis, which are now located in the new genus Dactylidispora, and S. mycoparasitica, which is now placed in Melanospora, are also transferred to Microthecium since these produce non-ostiolate or ostiolate ascomata without a neck, or less frequently with a short neck composed of angular cells similar to the ascomatal ones. Another relevant feature of the genus Microthecium is the production of bulbils. These propagules are typical of Papulaspora, an anamorphic genus that encompasses more than 40 species. Although it was initially accepted as a genus without a sexual morph (Hotson 1912), its link with species of Melanospora and Chaetomium has been reported (Roll-Hansen 1948, Zhang et al. 2004). In our phylogenetic study Papulaspora sepedonioides, the type species of the genus, was nested in the Microthecium clade, and therefore transferred to this genus. The relationship of this species with the Melanosporales had already previously been demonstrated by Davey et al. (2008) and Li et al. (2016). However, it has been demonstrated that Papulaspora is a polyphyletic genus, and other species of the genus have been reported as belonging to the classes Leotiomycetes and Sordariomycetes (Ascomycota). Therefore, the other species of Papulaspora not linked to the species of Microthecium should be transferred to other taxonomic groups. The relationship of some species of Papulaspora with the Melanosporales is also suggested by the production of similar phialidic synanamorphs (Van Beyma 1931, Hotson 1942).

The most recent new combination performed in Sphaerodes, S. inferior, was done to accommodate S. retispora var. inferior since it was not clustering with S. retispora var. retispora (Schultes et al. 2017). However, we suspected that the sequences of S. retispora var. retispora deposited in GenBank were contaminated with the hypocrealean host. In order to corroborate it, we studied such sequenced strain demonstrating that it was effectively contaminated. Therefore, S. inferior is here considered a synonym of Mi. retisporum since the morphological differences are insufficient to recognize this variety as a different species.

There are important morphological differences among the strains of Microthecium that suggest the presence of several additional cryptic species in the genus; however, our phylogenetic study, in spite of having used five loci, was not able to resolve the boundaries among them.