Fungi were collected in February and March 2018 in South African Provinces KwaZulu-Natal and Mpumalanga. The description of Octospora conidiophora is based on 11 collections belonging to the most frequent genotype. Observations of apothecial features were made on vital (marked by *) or rehydrated (†) material mostly in tap water, cresyl blue (CRB), lactophenol cotton blue (LPCB) or lactic acid cotton blue (LACB). Absence of amyloidity of asci was confirmed in Lugol´s solution. Infection structures were observed on rehydrated material. Parts of the host plants (leaves and rhizoids) close to an apothecium were separated, pulled apart, treated with LPCB and studied by light microscopy. The preparations were screened at 100× to 200× magnification for the presence of conidia. Infection structures and conidia usually occurred in the same mounts. Illustrations and measurements of hyphae, appressoria and haustoria, as well as conidia, were done in LPCB. The mosses were identified as hosts, based on the presence of appressoria on leaves or rhizoids. The host species were determined using standard techniques for bryophytes (Magill 1981). Collections are deposited in the Mycological department of the National Museum in Prague (PRM) and the herbarium of the Botanische Staatssammlung München (M).

DNA was extracted from dried apothecia by the CTAB method as outlined by Doyle and Doyle (1987). Up to three apothecia were homogenised by a pestle and incubated in 300 μl extraction buffer at 65 °C for one hour; the extract was subsequently purified in chloroform-isoamyl alcohol mixture, precipitated by isopropanol and finally dissolved in water and incubated with RNase for 30 min at 37 °C. DNA quality was checked on agarose gel. Molecular sequence data were generated for three loci: the 28S subunit of ribosomal DNA (LSU) was amplified with primers LR0R and LR6 (Vilgalys and Hester 1990), the 18S subunit of rDNA (SSU) with primers NS1 and NS6 (White et al. 1990) and translation elongation factor-1alpha (EF1α) with primers EF1-983F and EF1–1567R (Rehner and Buckley 2005). PCR was performed with Kapa polymerase (Kapa Biosystems, Wilmington, USA) following a standard protocol with 37 cycles and annealing temperature of 54 °C. The PCR products were purified by precipitation with polyethylene glycol (10% PEG 6000 and 1.25 M NaCl in the precipitation mixture) and sequenced from both directions using the same primers by the Sanger method at Macrogen Europe, Amsterdam, The Netherlands.

Newly generated sequences were assembled, edited and aligned in Geneious 7.1.7. (Biomatters, New Zealand) using the MAFFT plugin, manually corrected and deposited in NCBI GenBank under accession numbers MK569288–MK569376. Datasets were compiled from these and previously published sequences (Table 1), aligned, trimmed in order not to contain too many missing data at the ends and concatenated in Geneious 7.1.7. Bayesian Inference for concatenated data was computed in Mrbayes (ver. 3.2.4; Ronquist et al. 2012) with 2×10⁷ generations, sampling every 1000^th tree, in two independent runs, each with 4 chains, the first 50% (10⁷) generations being excluded as burn-in. The most suitable substitution model for each locus was determined in Partitionfinder 2.1.1 (Lanfear et al. 2017) using the AIC corrected for small samples (AICc) and a greedy search. Single-locus phylogenies were computed with similar settings, but with 6×10⁶ MCMC generations and the parameter temp. = 0.01.

Divergence times were estimated with Beast 2.5.1 (Bouckaert et al. 2014) using the LSU and SSU data from our sample set (one sample per species or phylogenetic lineage) and six additional species: Caloscypha fulgens (Pers.) Boud., Scutellinia scutellata (L.) Lambotte, Cheilymenia stercorea (Pers.) Boud., Aleuria aurantia (Pers.) Fuckel, Pyronema domesticum (Sowerby) Sacc. and Sarcoscypha coccinea (Gray) Boud. (all sequences obtained from Beimforde et al. 2014; EF1α was not analysed by these authors and, therefore, not included in our molecular dating). Four calibration points were used for the analysis and the divergence times, together with their confidence intervals, were also taken from Beimforde et al. (2014), namely divergence Cheilymenia-Scutellinia, divergence Aleuria-(Cheilymenia+Scutellinia), split-off of Sarcoscypha and split-off of Caloscypha. Monophyly was forced for all of the points except the second one due to an unclear position of the Octospora clade. Analysis was run under GTR+I+G substitution model (as for Mrbayes), with relaxed clock log normal model and 10⁸ MCMC generations, but the first 50% were excluded as burn-in. Priors included the Yule model with uniform birth rate and exponential gamma shape. Convergence and stationarity were analysed using Tracer v1.7.1 (Rambaut et al. 2018) and results were considered when effective sample size (ESS) ≥ 1000. Statistical uncertainty of divergence time estimates was assessed through the calculation of highest probability density (HPD) values.

After trimming, the total length of the concatenated alignment was 2702 bp (539 bp from EF1α, 1102 bp from LSU and 1061 bp from SSU, including gaps). Every studied locus provided sufficient polymorphism both amongst and within previously phenotypically delimited groups (Suppl. material 1: Table S1). Four distinct phylogenetic lineages were detected in the concatenated data, as well as in single-locus data within the group of specimens that were hosted by Sematophyllaceae (Fig. 1, Suppl. material 2: Fig. S1). Divergence between them was between 4 and 59 nucleotide differences at every locus (Suppl. material 1: Table S1). The four South African lineages formed a highly supported and distinct clade together with O. kelabitiana (Fig. 1). Molecular dating analysis estimated the basal split of bryophilous Pezizales to be 87–172 Ma old (95% confidence interval; mean = 149 Ma), the basal split of the South African accessions was estimated at 23–73 Ma (mean = 47 Ma; Fig. 2).

Figure 1. 

Bayesian phylogeny inference, based on concatenated alignment of EF1α, LSU and SSU sequences. Bayesian posterior probabilities are shown above branches; Otidea leporina serves as outgroup; trees based on analysis of each locus are shown in Suppl. material 2: Fig. S1.

Figure 2. 

Maximum clade credibility tree with estimated divergence times, based on SSU and LSU data; calibration points are marked by red circles, posterior probabilities shown above branches, bars indicate the 95% highest posterior density (HPD) intervals.

No significant differences in phenotypic traits were detected amongst the South African lineages using standard characters and methods. They shared the structure of excipulum, stiff, thick-walled hyaline hairs, ellipsoid hyaline ascospores which can be either smooth or ornamented with fine warts and which contain 1 or 2 guttules, warted mycelial hyphae, appressoria, haustoria and presence of anamorph. Although differences amongst individual collections were observed, phenotypic characters did not correspond to the molecular markers and many characters exhibited variability both amongst and within the four phylogenetic lineages (Table 2).

The four lineages could not be distinguished phenotypically on the basis of characters that are normally studied in bryophilous Pezizales, although genetic differentiation was very high at all of the three studied loci (Suppl. material 1: Table S1). Such great genetic distances are usually observed amongst different species or even genera. The observed genetic distances, together with molecular dating, imply that the phenotypically more or less homogeneous morphotype actually represents a group of several cryptic species that have already become reproductively isolated in the Tertiary (Fig. 2). Similar cryptic diversity is probably quite common in fungi, including many genera of Pezizales, e.g. Genea Vittad. (Smith et al. 2006, Alvarado et al. 2016), Geopyxis (Pers.) Sacc. (Wang et al. 2016), Helvella L. (Nguyen et al. 2013, Skrede et al. 2017), Terfezia (Tul. & C.Tul.) Tul. & C.Tul. (Ferdman et al. 2009), Trichophaea Boud. (Van Vooren 2016) and Tuber P.Micheli ex F.H.Wigg. (Bonuso et al. 2010). In bryophilous Pezizales, intraspecific sequence variability was observed, e.g. in Octosporopsis nicolai (Maire) U.Lindem., M.Vega & T.Richt. (Lindemann et al. 2014) and Octospora kelabitiana (Egertová et al. 2018). Each of the species comprised two genetic lineages that, nevertheless, were relatively weakly diverged and were therefore not treated taxonomically. Besides the significant genetic distances amongst the South African populations, another fact speaks against the possibility that the four lineages could be treated as a single species; the whole clade includes Octospora kelabitiana (Fig. 1), a distinct species from Borneo infecting liverwort Riccardia. A widely defined species (i.e. including the four lineages but excluding O. kelabitiana) would therefore be paraphyletic.

The current approach of many authors to delimitation of species is based primarily or solely on DNA sequence data and sequence-based diagnoses have become almost a common practice in macromycetes (e.g. Buyck et al. 2016, Leacock et al. 2016, Taşkın et al. 2016, Wang et al. 2016, Korhonen et al. 2018). Some authors even aim to base descriptions of new species on environmental sequence data only (e.g. Hibbett et al. 2011). Although molecular phylogenetics is an excellent tool for evaluation of biodiversity, assignment of scientific binominal to molecularly defined species leads to several practical problems, mainly those related to limited accessibility of the methods for many field mycologists. Especially in developing countries, in which even standard optical microscopy can be barely affordable at the leading institutes, determination of species via DNA sequencing is still a matter for the distant future. This methodological obstacle may soon result (or has already resulted in some groups) in the split of traditional phenotype-based taxonomy and molecular taxonomy. Until recently, molecular taxonomy mostly worked with groups, such as molecular operational taxonomic unit (MOTU; Hibbett et al. 2011), phylogenetic species (O’Donnell et al. 2011), virtual taxon (Öpik et al. 2010) etc. and designated an alphanumeric code to them. Nevertheless, many of the molecular taxa are currently given traditional scientific names, often without studying related, validly described species that cannot be sequenced for various reasons. This process, although justified by the aim of cataloguing of global biodiversity, makes the resulting taxonomy impractical or even unusable for field mycologists (and sometimes also for molecular biologists). Another problem with descriptions of species, based on molecular data, is the fact that the borderline between intraspecific and interspecific molecular variation is often unclear (Thines et al. 2018), dependent on many evolutionary factors (e.g. Leliaert et al. 2014) and may become fuzzy after a more intensive and/or extensive sampling is performed, particularly if only one or few molecular markers are used. Nevertheless, this problem also exists with traditional taxonomy (e.g. Flynn and Miller 1990, Paal et al. 1998, Benkert 2001). One solution to the problems mentioned above is an integrative approach. This takes advantage of both multiple characters (morphology, DNA, ecology etc.) and results in robust, phylogeny-based taxonomy that is accessible to various users (e.g. Araújo et al. 2015, Skrede et al. 2017, Haelewaters et al. 2018).

After thorough consideration of the above-mentioned facts, we decided not to formally describe all of the four discovered cryptic species at the present moment. Instead, we prefer to establish two taxa: O. conidiophora (s.str.), which refers to the most common phylogenetic lineage A and the informal taxon O. conidiophora agg., which applies to all of the four South African cryptic species, but also to the morphologically distinct and host-specific Bornean O. kelabitiana. Although the name O. kelabitiana is older and should therefore be selected for the aggregate, we believe that the name O. conidiophora agg. better suits the pragmatic purposes of this informal taxon. Our approach enables field mycologists to determine their specimens at least on the aggregate level and, at the same time, preserves a monophyletic taxonomical system. Detailed studies may reveal phenotypic differences between the South African lineages of O. conidiophora agg., which can then be formally described as species. Until then, we prefer to leave lineages B, C and D without a Latin binominal.

Conidia have been reported in several genera of Pezizales. The most frequent type of conidia are amerospores which are produced, e.g. in Caloscypha Boud. (Paden et al. 1978), Desmazierella Lib. (Hughes 1951), Iodophanus Korf (Korf 1958, sub Ascophanus Boud.), Pachyphlodes Zobel (Healy et al. 2015), Peziza Fr. (Berthet 1964a, Paden 1967, 1972), Ruhlandiella Henn. (Warcup and Talbot 1989, sub Muciturbo P.H.B.Talbot), Thecotheus Boud. (Conway 1975), Urnula Fr. (Davidson 1950), Cookeina Kuntze, Phillipsia Berk. (Paden 1975), Pithya Fuckel (Paden 1972), Nanoscypha Denison (Pfister 1973), Sarcoscypha (Fr.) Boud. (Harrington 1990), Geopyxis (Paden 1972), Pyropyxis Egger (Egger 1984, Filippova et al. 2016) and Trichophaea (Hennebert 1973). Staurospores can be found in Miladina lecithina (Cooke) Svrček (Descals and Webster 1978). The conidia of O. conidiophora can be classified as scolecospores or phragmospores and are therefore unique amongst Pezizales with known teleomorph. In their shape, they resemble the conidia of the anamorphic genus Spermospora R. Sprague (Ascomycota, Pezizomycotina), a parasite of grasses (Sprague 1948, Seifert et al. 2011).

Detached conidia were regularly found between the rhizoids and leaves in almost all collections of Octospora conidiophora agg. (with the exception of specimens ZE38/18, ZE53/18 and ZE71/18, probably due to limited material). The distal part of the conidia is sometimes short and straight. It is not clear whether this is an artefact caused by breaking off during preparation, although tail fragments have not been found. Germinating conidia are not rare. Longer germination tubes look like normal hyphae with the characteristic warty surface structure (Figs 5F, 7F). Conidia germinating by a two-celled appressorium (Fig. 7F) or connected to a mycelial hypha by an anastomosis (Figs 7E, G, 9A) have been repeatedly observed. Conidiogenous cells (Fig. 7D) are much more difficult to detect than conidia and have only been found in a few collections. The scars formed by detachment of conidia must not be confused with the ends of torn-off hyphae, which inevitably result during preparation. Fully developed conidia still connected to the conidiogeneous cells have not been found. Apparently, mature conidia easily detach from their conidiogenous cells. A developing, still attached conidium was observed once (Fig. 7D).

Octospora conidiophora agg. is the first case amongst bryophilous Pezizales in which an anamorph has been detected. The absence of records of anamorphic states in other species can be caused either by their real rarity or only by their difficulty in detection. The latter can have many reasons. First, bryophilous ascomycetes, in general, stand rather on the periphery of researchers´ interest (see Döbbeler 1997). Second, anamorphs are usually inconspicuous and therefore not easy to encounter. Even if an anamorph is found, it can be difficult to link it with the corresponding teleomorph, because many fungal species commonly occur together. Moreover, anamorphs and teleomorphs are often formed in different environmental conditions (Kendrick 1979) and often at different times. And third, anamorphs are often studied in aseptic cultures and subsequently cultures are used for confirmation of their identity by molecular methods; unfortunately, cultivation of bryophilous Pezizales seems to be problematic (Berthet 1964b) and is not commonly attempted. Although an anamorph has not been confirmed by cultivation methods in O. conidiophora agg., the connection of anamorph and teleomorph is based on the evidence discussed above: conidia were repeatedly found amongst the moss plants near the teleomorph; germinating conidia have hyphae with the same ornamentation as observed in the mycelium bearing apothecia; conidiogenous cells occur on the mycelial hyphae; conidia anastomose with mycelial hyphae; the germlings form appressoria.

Sematophyllum brachycarpum (Hampe) Broth.

Syn: Hypnum brachycarpum Hampe

Sematophyllum brachycarpum can be distinguished from other species of Sematophyllum in southern Africa by the complanate, straight leaves with relatively large groups of alar cells (in 3–4 rows) that are not much inflated or coloured (Fig. 10, see also Câmara et al. 2019).

Figure 10. 

Sematophyllum brachycarpum. A Plants B Typical shoot with leaves C Leaf D Leaf base with alar cells.

The species is by far the most common and widespread species of Sematophyllum in South Africa; S. brachycarpum is found in forests and wooded areas of the Limpopo, Mpumalanga, North West, Gauteng, Free State, KwaZulu-Natal, Eastern Cape and Western Cape Provinces (Fig. 11, see also Câmara et al. 2019). It occurs as an epiphyte or occasionally on soil or rocks, from sea level up to 1900 m alt. The species is widely distributed throughout the Afromontane Region, as defined by Van Rooy and Van Wyk (2010) and was found to belong to the Widespread Afromontane Subelement, a subdivision of the Afromontane Forest Element (Van Rooy and Van Wyk 2011). The Widespread Afromontane Subelement is centred in the Midlands of KwaZulu-Natal and the Drakensberg escarpment of Mpumalanga as well as in forests in the south-western Cape. The species has also been recorded from Lesotho, Swaziland, Mozambique, Zimbabwe, Zambia, Uganda and Kenya (O’Shea 2006).

Figure 11. 

Geographical distribution of Sematophyllum brachycarpum in southern Africa based on records in BM, L, MO and PRE.

Trichosteleum perchlorosum Broth. & Bryhn

Trichosteleum perchlorosum is the only southern African species of Sematophyllaceae (sensu stricto) with papillose leaf cells. However, the papillae are sometimes difficult to see or may be absent on some leaves. The falcate leaves with enlarged, inflated and coloured alar cells will also help to identify the species (Fig. 12, see also Câmara et al. 2019).

Figure 12. 

Trichosteleum perchlorosum. A. Plants. B. Leaf. C. Leaf papillae. D. Alar cells.

The species is endemic to the southern part of Africa and occurs as an epiphyte and also on decaying logs or rocks from sea level up to 3090 m high (Drakensberg of KwaZulu-Natal). It is most frequently collected in the KwaZulu-Natal Province of South Africa, but it is also known from Limpopo, Mpumalanga, Eastern Cape and Western Cape Provinces, as well as Swaziland (Fig. 13, see also Câmara et al. 2019). Trichosteleum perchlorosum is widespread throughout the Afromontane Region sensu Van Rooy and Van Wyk (2010), but unknown from Afromontane outliers in the Magaliesberg of Gauteng and the North West, the eastern Free State and the Waterberg of Limpopo. It was therefore included in the Tropical Afromontane Subelement (Van Rooy and Van Wyk 2011), which is centred in the Drakensberg escarpment of Mpumalanga and the Midlands of KwaZulu-Natal. This species was also reported from Zimbabwe (O’Shea 2006).

Figure 13. 

Geographical distribution of Trichosteleum perchlorosum in southern Africa based on records in BM, L, MO and PRE.