In TNS, specimens labeled as Erioscyphella were initially searched, and closely related specimens to Erioscyphella were searched based on the sequence similarities of ITS. Selected specimens were tentatively identified based on morphology following Dennis (1954), Haines (1980), Haines and Dumont (1984), Spooner (1987), and Perić and Baral (2014).

Micromorphology was examined using cotton blue (CB) dissolved in lactic acid (LA) (CB/LA; 0.5 g CB and 99.5 mL LA) as a mounting fluid. To check the ascal apex iodine reaction, Melzer’s reagent (MLZ; 0.5 g I₂, 1.5 g KI, 20 g chloral hydrate, and 20 g water) was initially used without KOH pretreatment, and Lugol’s iodine (IKI; 1 g I₂ and 1 g KI, and 100 mL H₂O) and MLZ with 3% KOH pretreatment were used when necessary. World Geodetic System 84 was used for the geographic coordinates. URLs herein shown were accessed on April 15, 2021, except for GBIF website accessed on Feb 10, 2020.

DNA was extracted from cultivated isolates in 2% malt extract broth (MEB) using the modified cetyltrimethylammonium bromide (CTAB) method (Hosaka and Castellano 2008; Tochihara and Hosoya 2019). When isolates are not available, DNA was extracted directly from a crushed apothecium using DNA extraction buffer following Tochihara and Hosoya (2019). The isolates from which DNA extracted were deposited in the NITE National Biological Resource Center (NBRC) (Kisarazu, Japan), except for isolates with restriction on transition by Japanese laws and those unavailable because of contracts with private companies.

Polymerase chain reaction (PCR) was used to amplify the following regions: ITS (= ITS1-5.8S-ITS2), the partial large subunit nuclear ribosomal RNA gene (LSU), the partial mitochondrial small subunit (mtSSU), and section ‘6–7’ of the second largest subunit of the nuclear RNA polymerase II gene (RPB2). Primer pairs for PCR reactions of ITS, LSU and mtSSU were ITS1F (5’–CTTGGTCATTTAGAGGAAGTAA–3’) (Gardes and Bruns 1993) or ITS1 (5’–TCCGTAGGTGAACCTCGGG–3’) (White et al. 1990) and ITS4 (5’–TCCTCCGCTTATTGATATGC–3’) (White et al. 1990), LR0R (5’–ACCCGCTGAACTTAAGC–3’) and LR5 (5’–TCCTGAGGGAAACTTCG–3’) (Vilgalys and Hester 1990), and mrSSU1 (5’–AGCAGTGAGGAATATTGGTC–3’) and mrSSU3R (5’–ATGTGGCACGTCTATAGCCC–3’) (Zoller et al. 1999) respectively. The PCR program consisted of an initial denaturation at 95 °C for 3 min, followed by 30 cycles of 94 °C for 35 s, 51 °C for 30 s, and 72 °C for 1 min, and a final extension at 72 °C for 10 min. When appropriate PCR products were not obtained, a modified PCR program was applied first, and then alternative primer pairs were tested. For RPB2, an alternative forward primer fRPB2-5F (5’–GAYGAYMGWGATCAYTTYGG–3’) (Liu et al. 1999) or RPB2-P6Fa (5’–TGGGGRYTK GTBTGYCCKGCHGA–3’) (Hansen et al. 2005) and a reverse primer bRPB2-7.1R2 (5’–CCCATNGCYTGYTTVCCCATDGC–3’) (modified from bRPB2-7.1R) (Matheny 2005; Matheny et al. 2007; Gelardi et al. 2015) were used.

Sequencing was conducted on an ABI PRISM 3500xL Genetic Analyzer (Applied Biosystems; Thermo Fisher Scientific, Waltham, MA, USA) with a BigDye Terminator 3.1 Cycle Sequencing Kit (Applied Biosystems). The obtained sequences were assembled using ATGC 7 (Genetyx, Tokyo, Japan). Assembled sequences were deposited in the International Nucleotide Sequence Database Collaboration (INSDC) via the DNA Data Bank of Japan (DDBJ), and acquired INSDC accession numbers. Assembled ITS sequences were also deposited in the UNITE database (https://unite.ut.ee/) via the PlutoF workbench (https://plutof.ut.ee/) (Abarenkov et al. 2010) and acquired UNITE accession numbers.

The specimens obtained from TNS were included in the phylogenetic analyses as candidate members of Erioscyphella (‘‡’ in Table 1). From other genera of the family Lachnaceae, four species of Lachnum, two species of Albotricha, Brunnipila, Capitotricha, Dasyscyphella, Incrucipulum, and Lachnellula, and one species of Neodasyscypha and Proliferodiscus were used (‘†’ in Table 1). Among the eight genera, seven of them (except Proliferodiscus) included type species. Three species of Helotiales were selected as outgroups following Tochihara and Hosoya (2019) (Table 1).

A concatenated dataset of ITS, LSU, mtSSU, and RPB2 was used in the phylogenetic analyses. Each region was aligned separately using MAFFT 7 (Katoh and Standley 2013). The Q-INS-i option was used for ITS, LSU, and mtSSU to accommodate the secondary structures of RNA, and the G-INS-1 option was used for RPB2 to assume global alignment using the entire region. The aligned sequences were edited manually using BioEdit 7.0.5.2 (Hall 1999).

Phylogenetic conflicts among gene partitions were checked before the phylogenetic analyses using the concatenated matrix. Maximum likelihood (ML) trees with 1,000 bootstrap replications (Felsenstein 1985) using the ITS, LSU, mtSSU, and RPB2 datasets separately were constructed using MEGA X (Kumar et al. 2018) with the GTR+G model; branches with bootstrap values > 70% were compared among trees. For mtSSU and RPB2, specimens containing missing data were excluded from the analyses.

The concatenated dataset was analyzed using ML, maximum parsimony (MP), and Bayesian inference (BI). For the ML and BI analyses, substitution models were estimated for each partition (ITS, LSU, mtSSU, and each codon position of RPB2) based on Akaike’s information criterion (AIC) (Akaike 1974) using Modeltest-NG 0.1.6 (Darriba et al. 2019).

ML tree search (Felsenstein 1984) and bootstrapping (Felsenstein 1985; Lemoine et al. 2018) was performed using RAxML-NG 0.9.0 (Kozlov et al. 2019) with 1,000 bootstrap replications under the substitution model SYM+I+G4 for ITS, TIM1+I+G4 for LSU, TPM1uf+I+G4 for mtSSU and RPB2 third codon position, GTR+I+G4 for RPB2 first codon position, and TPM3uf+I+G4 for RPB2 second codon position. Sequence matrix containing missing data typically yield multiple trees residing on a phylogenetic terrace (Sanderson et al. 2011; Biczok et al. 2018). Therefore, we checked if the best-scored-tree did not lie on a terrace using the Python tool called ‘terraphy’ implemented in RAxML-NG 0.9.0.

MP analysis was conducted using PAUP* 4.0a 167 (Swofford 2002). All substitutions were treated as unordered and of equal weights. All gaps were treated as missing data. A heuristic parsimony search was carried out with 1,000 replicates of random step addition, with a tree bisection reconnection (TBR) branch swapping algorithm, Multrees option on, Steepest descent modification option on, and branch collapse option set to MinBrlen. Bootstrap values (MPBP; Felsenstein 1985) were estimated from 1,000 replicates of heuristic searches, with random taxon addition, TBR branch swapping, and Multrees options off.

Figure 1. 

ML best-scored phylogenetic tree based on the concatenated dataset of ITS, LSU, mtSSU, and RPB2 constructed using RAxML-NG. MLBP/MPBP/BPP are represented on branches in this order. In MLBP/MPBP < 50% or BPP < 0.95, a hyphen appears. No evaluation values are shown on branches when MLBP and MPBP < 50% and BPP < 0.95. The branch of a clade TNS-F-17245 + 17249 to its most recent common ancestor is only one-third of the intended length due to space limitation.

BI analysis was based on MrBayes 3.2.7a (Ronquist et al. 2012) under the substitution model SYM+I+G4 for ITS, GTR+I+G4 for LSU and RPB2 first codon positions, HKY+I+G4 for mtSSU and RPB2 third codon positions, and F81+I for RPB2 second codon position. Two separate Metropolis-Coupled Markov Chains of Monte Carlo (MCMCMC) ran simultaneously starting from random trees for 20 million generations, and trees were sampled every 500 generations. The average standard deviation of split frequencies (ASDSF) and effective sample size (ESS) were checked using Tracer 1.7.1 (Rambaut 2018a) as an indication of convergence. Using post-burn-in trees, a 50% majority rule consensus tree was generated, and Bayesian posterior probabilities (BPP) were calculated to evaluate node supports. Trees were visualized using FigTree 1.4.4 (Rambaut 2018b) based on the ML, MP, and BI analyses respectively. Branches with MLBP and MPBP > 90% and BPP > 0.95 were regarded as strongly supported.

ITS-based species delimitation analyses (Fig. 2)

Figure 2. 

Diagrammatic representation showing the species delimitation analyses using ITS sequences.

To maximize the number of ITS sequences, we used the UNITE Species Hypotheses (SH) system provided by the UNITE database (Kõljalg et al. 2013; Nilsson et al. 2015; GBIF 2018; Kõljalg et al. 2020). In the UNITE SH system, all fungal ITS sequences are periodically divided into species-level clusters (species hypothesis; SH) at optional sequence-distance thresholds (0%–3% in 0.5% steps), each of which is assigned to a unique UNITE SH code represented by a digital objective identifier (DOI) accessible from internet (Kõljalg et al. 2016, 2020; Nilsson et al. 2015).

Based on the UNITE SH system, we collected ITS sequences of Erioscyphella in the following process: a) selectivity of closely related sequences: for every ITS sequence newly obtained from TNS specimens (= query sequences, 49 sequences), UNITE SH code at the 3% threshold value were searched in the UNITE database to gather sequences in wider scope, and all sequences within the UNITE SH code were downloaded. b) selectivity based on taxon names: using the UNITE search page, ITS sequences named Erioscyphella were searched, because only closely related sequences to query sequences are filtered under the a) criterion. Sequences with synonyms of Eriosyphella species were also searched, because the UNITE lookup function is not supported by any backbone taxonomies to integrate synonyms. Sequences satisfying criterion a) or b) were downloaded for ITS-based species recognition. The obtained ITS sequences were clustered into SHs based on an OTU clustering method, hierarchical clustering method, and two coalescent-based methods. For all ITS sequences, ITS1, 5.8S, and ITS2 regions were extracted using ITSx (Nilsson et al. 2010) to construct an accurate ITS dataset, because the inclusion of segments of adjacent regions (such as a small subunit of 18S rRNA or LSU) may decrease the accuracy of the calculation of ITS distances (Nilsson et al. 2010). OTU clustering was executed using VSEARCH v2.17.2 (Rognes et al. 2016) implemented in the Qiime 2 microbiome analysis platform (Bolyen et al. 2019).

The concatenated dataset of extracted ITS1, 5.8S, and ITS2 was incorporated into VSEARCH, and OTU clustering at 97% and 98.5% similarity thresholds were performed using the ‘-cluster_fast’ option. Hierarchical clustering based on pairwise sequence distances was executed using the Assemble Species by Automatic Partitioning (ASAP) method (Puillandre et al. 2021). The datasets of extracted ITS1, 5.8S, and ITS2 were separately aligned using MAFFT 7 under the Q-INS-i option and edited using trimAl v1.2 (Capella-Gutiérrez et al. 2009) under the ‘-gappyout’ option. The concatenated dataset of the three aligned partitions was analyzed using ASAP web (https://bioinfo.mnhn.fr/abi/public/asap/asapweb.html). Jukes-Cantor (JC69) was selected as a substitution model for computing pairwise distances of sequences. As phylogeny-based species delimitation methods, the generalized mixed Yule-coalescent (GMYC) model (Pons et al. 2006; Fujisawa and Barraclough 2013) and the Poisson Tree Processes (PTP) model (Zhang et al. 2013) were used. In both models, speciation (species-level differentiation) and coalescence (population-level differentiation) are identified based on the length of phylogenetic trees. GMYC requires the use of phylogenetic trees following the molecular clock model (= ultrametric tree) because it detects transition points from speciation to coalescence focusing on the time axis, while PTP does not require ultrametric tree as it focuses on the number of nucleotide substitutions. Ultrametric trees were estimated using BEAST v2.6.3. (Bouckaert et al. 2019). The ITS dataset was divided into ITS1, 5.8S, and ITS2, and suitable substitution models GTR+G for ITS1 and JC+G for 5.8S and ITS2 estimated using Modeltest-NG 0.1.6. were applied. To estimate branch length, a Yule model and a relaxed clock with a log-normal distribution were selected. MCMC chains were run for 1.5×10⁸ generations and sampled every 1,000 generations. After each run, convergence was checked using Tracer 1.7.1, and the first 10% were discarded as burn-in. A consensus tree was generated using TreeAnnotator v1.10.4 in BEAST package, from 150,000 generated trees except for the first 10% regarded as burn-in. A single-threshold species delimitation analysis based on GMYC was conducted using the R package ‘splits’ (Fujisawa and Barraclough 2013).

For the species delimitation analyses using PTP, an unrooted ML phylogenetic tree was constructed using RAxML-NG 0.9.0. The analysis used ITS1, 5.8S, and ITS2 partitions, aligned as previously described, under the substitution models TIM2+G4 for ITS1, TPM2+I+G4 for 5.8S, and GTR+I+G4 for ITS2, estimated using Modeltest-NG 0.1.6. based on the AIC. The species delimitation analysis was executed using the generated ML best-scored tree with the bPTP web server (https://species.h-its.org/). The MCMC run was set to 500,000 generations and burn-in rate was set to 0.1. The convergence of MCMC runs was visually checked. In ML and Bayesian results, a result generating fewer SHs was adopted to avoid excessive species division.

SHs generated in the species delimitation analyses and the UNITE SHs at 3% and 1.5% threshold values were compared with one another.

In the present study, we initially recognized species boundaries based on the two criteria:

1. Forming a monophyletic group in the phylogenetic analyses based on multigene data (Fig. 1).
2. Members can be distinguished based on morphological and/or common ecological features (such as host plants).

Species boundaries recognized by 1.and 2. were cross-checked based on the results of ITS-based species delimitation analyses. When the species boundaries are supported by the majority (= more than four methods) of the seven species delimitation methods (UNITE SH at 3% threshold, UNITE SH at 1.5% threshold, VSEARCH 97% similarity, VSEARCH 98.5% similarity, ASAP, GMYC, and PTP) (Fig. 3), we regard the species as reasonable and carry out taxonomic treatments if necessary.

Figure 3. 

Species delimitation analyses using ITS sequences of Erioscyphella and its potential members. Clusters based on UNITE SH at 3% and 1.5% threshold values at UNITE v8.2, VSEARCH at 97% and 98.5% threshold values, ASAP, GMYC, and PTP are displayed. Schematic phylogenetic relationships are shown using the ultrametric tree constructed for the GMYC analysis. The taxon names shown on the tree branches follow the results of the present study.

Forty-nine specimens in TNS were identified as candidates of Erioscyphella and morphologically identified as E. abnormis, E. brasiliensis, E. sclerotii, Lachnum hainanense W.Y. Zhuang & Zheng Wang, L. mapirianum (Pat. & Gaillard) M.P. Sharma, Lachnum mapirianum var. sinense Z.H. Yu, W.Y. Zhuang, Lachnum novoguineense var. yunnanicum W.Y. Zhuang, and L. palmae (Kanouse) Spooner (Table 1), together with six species of Erioscyphella described here as new ([E. boninensis, E. insulae, E. otanii, E. papillaris, E. paralushanensis, and E. sasibrevispora], Table 1).

The molecular phylogenetic analyses were based on 70 specimens selected from TNS (Table 1). The concatenated sequence matrix was composed of 2488 bp (sites 1–332 for ITS, 333–1108 for LSU, 1109–1828 for mtSSU, and 1829–2488 for RPB2). In the matrix, the following parts were treated as missing data: TNS-F-17245, 17249, and 81229 for mtSSU, and TNS-F-17567 for RPB2. The matrix was registered in TreeBase (http://purl.org/phylo/treebase/phylows/study/TB2:S28477).

Among the four ML trees based on each region, no conflicts were found in clades with support > 70% (Suppl. material 1: Fig. S1). Therefore, we considered these four regions to be combinable, and phylogenetic analyses were based on the concatenated sequence matrix. In the ML analysis, the best-scored tree generated did not reside on the phylogenetic terrace. In the MP analysis, 766 nucleotide substitution sites were detected, 601 of which were parsimony-informative. A total of 182,630 equally parsimonious trees were generated with tree length = 2,985 steps, consistency index (CI) = 0.38, retention index (RI) = 0.73, and rescaled consistency index (RC) = 0.28. In the BI analysis, when two runs reached 20 million generations and the first 10,000 trees (25%) of generated trees were excluded, ASDSF was observed to fall below 0.004 and ESS of all parameters was over 200. The first 10,000 trees were discarded as burn-in. A 50% majority rule consensus tree was constructed and BPP was calculated based on the remaining 30,000 trees.

As no topological contradictions occurred among the ML best-scored tree, MP 50% majority-rule consensus tree, and BI 50% majority-rule consensus tree, only ML tree was illustrated, and MLBS, MPBS, and BPP were plotted on its branches (Fig. 1).

Based on the phylogenetic analyses, 49 candidates of Erioscyphella formed a strongly supported clade (= Clade A, MLBP = 100%/MPBP = 100%/BPP = 1.00), apart from the clade of Lachnum sensu stricto (= L. asiaticum (Y. Otani) Raitv., L. pudibundum (Quél.) J. Schröt., L. rachidicola J.G. Han, Raitv. & H.D. Shin, and L. virgineum (Batsch) P. Karst.) [type of Lachnum]) (Fig. 1). Clade A and Proliferodiscus alboviridis formed a relatively strongly supported clade (Clade B, MLBP = 78%, MPBP = 82%, BPP = 1.00).

Within Clade A, each morphologically identified species and variety formed strongly supported monophyletic groups of their own (Fig. 1), and five strongly supported subclades were recognized (Clade I–V, Fig. 1). Lachnum mapirianum (TNS-F-17545, 17249) and E. insulae (TNS-F-26500, 39720) did not belong to any subclade. Clade I was composed of E. boninensis, E. paralushanensis, L. hainanense, and L. mapirianum var. sinense. Within Clade I, only E. paralushanensis occurred on bamboo sheaths, while others occurred on fallen leaves of broad-leaved trees. Clade II was composed only of L. palmae, which occurred on the palm petioles. Clade III was composed of E. otanii and E. papillaris occurring on bamboo leaves. Clade IV was composed of L. novoguineense var. yunnanicum, and E. sasibrevispora, occurring on bamboo sheaths. Clade V was composed of E. abnormis, E. brasiliensis, and E. sclerotii, occurring on wood.

Members of Clade A had totally and densely granulate, hyaline to brown, thin-walled hairs, fusiform to long filiform ascospores, ectal excipulum composed of textura prismatica to textura angularis, asci lacking croziers at the bases, and smooth walled ectal excipulum cells. Exceptionally, E. sasibrevispora, L. hainanense (Hosoya et al. 2013), and L. novoguineense var. yunnanicum W.Y. Zhuang had croziers and E. boninensis had granulated ectal excipulum.

Moreover, hairs of Clade A lacked crystals, but were equipped with apical amorphous materials and/or resinous materials. In the present study, “crystals” refers to amber colored materials that positioned near the hair apices and were regular-shaped (e.g. tetrahedral materials, masses of needle-like materials, or cross-shaped materials), described by Raitviir (2002), Suková (2005) or Tochihara and Hosoya (2019). “Resinous materials” refers to colored, refractive, irregular-shaped materials attached on any parts of hairs, described by Spooner (1987). Crystals and resinous materials are easily detatched from hairs and broken into fragments in the squash mount. “Apical amorphous materials” is termed uniquely in this study, and refers to hyaline to brown, refractive, irregular-shaped materials positioned outside the hair apices. They are usually small and inconspicuous cap-like shaped, and conspicuously globular in some species. Apical amorphous materials do not grow to big masses and are not easily detached from hairs in the squash mount.

In Clade A, members except for E. boninensis, E. sasibrevispora and L. novoguineense var. yunnanicum had apical amorphous materials, and E. boninensis, E. paralushanensis, and L. palmae complex also had resinous materials (see figures of described species and Suppl. material 1: Fig. S2).

In UNITE v8.3, 87 ITS sequences were clustered into 23 SHs at 3% and 26 SHs at 1.5% threshold values (Table 2, Fig. 3). The UNITE SH code for each SH is presented in Table 2. In OTU clustering using VSEARCH, 87 ITS sequences were clustered into 25 SHs at 97% similarity and 28 SHs at 98.5% similarity (Table 2, Fig. 3). VSEARCH SH codes (allocated in this study uniquely; VSH97_1 to VSH97_25, VSH985_1 to VSH985_28) are shown in Table 2.

The extracted and aligned ITS sequences were composed of three partitions, ITS1 (162 bp), 5.8S (157 bp), and ITS2 (142 bp). The concatenated ITS sequence matrix was registered in TreeBase (http://purl.org/phylo/treebase/phylows/study/TB2:S28473). In the ASAP analysis, the concatenated dataset of these partitions (461 bp) was input, and 87 ITS sequences were clustered into 18 SHs with the lowest asap-score, reflecting better partitioning (Suppl. material 1: Fig. S3). In the GMYC analysis, 29 SHs were delimited (Suppl. material 1: Fig. S4). The ultrametric tree constructed for the GMYC analysis is available in TreeBase (http://purl.org/phylo/treebase/phylows/study/TB2:S28473). For the PTP analyses, an ML best-scored tree was constructed (Suppl. material 1: Fig. S5). PTP analyses delimited 23 SHs in the Bayesian support and 26 SHs in the ML support (Suppl. material 1 Fig. S6), and the former was adopted.

Comparing the number of SHs generated by different clustering methods and applied thresholds, 18 SHs by ASAP, and 23 SHs by UNITE SH at 3% threshold represented the lowest SH numbers (Fig. 3; Table 2). The ASAP results were too rough to delimit the boundaries of E. abnormis, E. boninensis, E. brasiliensis, E. curvispora, and E. sclerotii. SH-classification recognized by UNITE SH at 3% threshold mostly corresponded to taxon names originally assigned to sequences.

Comparing the results of seven species delimitation methods (UNITE SH at 3% threshold, UNITE SH at 1.5% threshold, VSEARCH 97% similarity, VSEARCH 98.5% similarity, ASAP, GMYC, and PTP), sequences labeled as E. alba, E. brasiliensis, E. curvispora, E. euterpes, E. fusiformis, E. lunata, E. sclerotii, L. mapirianum, L. mapirianum var. sinense, L. novoguineense var. yunnanica, and six new species candidates were distinguished as separate clusters by more than four delimitation methods (Fig. 3). These species clusters did not contradict with morphological/ecological and phylogenetic relationships (Fig. 1). Seven sequences labeled as L. hainanense were clustered into one SH by four species delimitation analyses, and part of the SHs included a sequence labeled as Lachnum albidulum (Fig. 3).

Erioscyphella abnormis, E. aseptate, and L. palmae did not form separate clusters supported by majority of four species delimitation analyses (Fig. 3). Sequences labeled as E. abnormis were clustered into one to four SHs, and some SHs included sequences labeled as Chapsa patens (Nyl.) Frisch, E. aseptata, E. brasiliensis, and E. sclerotii (Fig. 3). Twelve sequences labeled as L. palmae were clustered into four to six SHs (Fig. 3).

We accepted Clade A as a monophyletic unit for Erioscyphella which is supported by morphology. Although Clade B comprised Clade A together with P. alboviridis, Clade B should not be regarded as a genus delimitation of Erioscyphella, because Proliferodiscus differs from members of Clade A in having apothecia proliferating from the margins continuously and thick-walled and coarsely warted hairs (Haines and Dumont 1983; Spooner 1987). All members of Clade A are distinguishable from the other lachnacenous genera. In contrast to Erioscyphella, Albotricha and Dasyscyphella are distinguished by hair apices with no granulation (Hosoya et al. 2010), Brunnipila, Capitotricha, and Incrucipulum by hair-crystals (Baral and Krieglsteiner 1985; Tochihara and Hosoya 2019), and Lachnellula by ectal excipulum composed of textura globose to textura oblita (Dharne 1965). Typical members of Clade A can be easily segregated from Neodasyscypha, because the characteristic features of Neodasyscypha, such as dark-brown hairs, ectal-excipulum structure, and ellipsoid to fusoid ascospores < 10 µm long (Spooner 1987), are rare in Clade A. Among members of Clade A and Lachnum sensu stricto, the shape and length of ascospores were continuous (Fig. 4), as indicated by Haines and Dumont (1984). However, ascospores longer than 15–20 µm were restricted to Clade A (Fig. 4). Moreover, most members of Clade A have hairs with apical amorphous materials, which are not seen in Lachnum sensu stricto. Members of Clade A usually also have hairs not swelling at the apices and distantly septate, as Perić and Baral (2014) pointed out for three tropical members, while members of Lachnum have swelling apices. The combination of such characters allows us to differentiate typical members of Erioscyphella from Lachnum.

Figure 4. 

Comparison of ascospores of Clade A (= Erioscyphella) and the clade of Lachnum sensu stricto in Fig. 1. Subclade numbers for members of Clade A in Fig. 1 are shown in parentheses. Bars show variation of ascospore length within each species.

In summary, Erioscyphella is still difficult to define solely based on morphology because of multiple exceptional characters continuous to other genera, but its typical members could be recognizable mainly by the hair structures and ascospore length. Based on members of Clade A, Erioscyphella is tentatively described as follows: apothecia occurring on dead hardwood leaves, rotten wood, bamboo sheaths, bamboo leaves or palm leaves; asci mostly arising from simple septa, but occasionally from croziers; ascospores fusiform to long needle-shaped, aseptate to multi-septate; paraphyses filiform to narrowly lanceolate, shortly exceeding the asci, but rarely lanceolate and long exceeding the asci; hairs straight or irregularly curved, usually not swollen at the apices, thin-walled, hyaline, but sometimes brown, totally and densely granulated, usually distantly septate, without needle-like or three-dimensional shaped crystals but mostly equipped with hyaline to brown apical amorphous materials, and/or resinous materials at any part of hairs; walls of ectal excipulum cells smooth but granulate in one species.

Perić and Baral (2014) pointed out that “yellow hymenium derived from carotenoid” is one of the common characters of Erioscyphella. This feature was not discussed in this study because some specimens were not observed when fresh; the hymenium color is variable (usually white hymenium becomes yellow) between fresh and dried states in lachnaceous species.

In Erioscyphella, the tendency of selectivity of species to host plants or parts occurs across the genus. Each subclade within Erioscyphella (Clade I–V) generally shared tendencies toward host selectivity as follows: Clade I on leaves of broad-leaved trees, except for E. paralushanensis occurring on bamboo sheaths, Clade II on palm leaves, Clade III on bamboo leaves, Clade IV on bamboo sheaths, and Clade V on rotten wood (Fig. 1). The results showed that selectivity to host plants, and parts of Erioscyphella, was acquired as apomorphic characters during speciation.

Erioscyphella (long-spored Lachnum) has long been known as the tropical genus in Lachnaceae (Dennis 1954; Spooner 1987; Guatimosim et al. 2016). Most long-spored species were described from tropical areas of Latin America (Dennis 1954) and tropical to temperate areas of Australasia (Spooner 1987). However, the new species or new combinations proposed in this study were reported from Japan in subtropical areas (E. boninensis and E. insulae), temperate area (E. hainanensis, E. palalushanensis, and E. sinensis) and cool-temperate to subarctic areas (E. otanii, E. papillaris, and E. sasibrevispora), showing that Erioscyphella is not limited to tropical zones, but is also distributed in temperate to subarctic zones in the northern hemisphere.

Iodine reactions of the ascus apical apparatus have been classified into several types (inamyloid, hemiamyloid [Type RB and RR, and euamyloid Type BB]) (Baral 2009), and the reaction ‘MLZ- without KOH pretreatment and MLZ+ with KOH pretreatment’, observed in E. papillaris (Fig. 11E1 and Fig. 11E2) has been restricted to the type of hemiamyloid. However, the apical apparatus of E. papillaris showed a dark blue reaction in IKI without KOH pretreatment (Fig. 11E3), while the hemiamyloid apparatus usually shows a red reaction under these conditions. The hemiamyloid ascal apparatus could show IKI-blue without KOH pretreatment due to long storage in the herbarium (Baral 2009), but this is not applicable for the material of E. papillaris, which has been maintained for only two years in herbarium until observed. Therefore, we assessed the iodine reaction of E. papillaris as a new type, and color reactions with various solutions of the species should be further examined using new materials, because there are few apothecia in the type specimen.

In this study, we carried out taxonomic treatment for species which were distinguished by morphology/ecology and phylogenetic analyses, and formed single clusters in species delimitation analyses. Based on this criteria, six undescribed species of Erioscyphella have been proposed as new species of Erioscyphella [E. boninensis, E. insulae, E. otanii, E. papillaris, E. paralushanensis, and E. sasibrevispora], and Lachnum hainanense and L. mapirianum var. sinense have been proposed as new members of Erioscyphella. Interpretation of species boundaries of L. hainanense was discussed in the taxonomy chapter. For new species and new combinations, Japanese names were also denominated for wider use of Japanese mycologists or amateurs.

In the phylogenetic analyses, Malaysian materials of L. mapirianum (TNS-F-17245, 17249) and Japanese materials of L. novoguineense var. yunnanicum (TNS-F-16442, 16642) were also found to be members of Erioscyphella (Fig. 1). However, we hesitate to transfer the two species into Erioscyphella, as we cannot guarantee the identification accuracy of the materials, because of inadequate type information of the two species.

Taxonomic assessments of E. abnormis, L. aseptate, and L. palmae, which were not accepted as independent species in species delimitation analyses, are discussed below.

In the species delimitation analyses, sequences labeled as E. abnormis formed a single SH at UNITE SH 3% threshold (DOI: SH1155612.08FU) and divided into two to four SHs at UNITE SH 1.5% threshold, VSEARCH, and GMYC (Fig. 3).

In ASAP, sequences labeled as E. abnormis belong to a single SH, but the SH also contained sequences labeled as Chapsa patens, E. aseptata, E. brasiliensis, E. curvispora, and E. sclerotii (Fig. 3). However, the phylogenetic analyses revealed that E. brasiliensis, and E. sclerotii are separate from the clade of E. abnormis (Fig. 1), suggesting that the two species are different from E. abnormis. Although E. curvispora was not included in the phylogenetic analyses (Fig. 1), the apparent morphological and ecological differentiation (Perić and Baral 2014) and low similarity of ITS (< 97%) with members of E. abnormis (Fig. 3) suggest that E. curvispora is different from E. abnormis.

Erioscyphella aseptata was originally described in Thailand and characterized by having aseptate ascospores, unlike E. abnormis or E. sclerotii with septate ascospores (Ekanayaka et al. 2019). However, the species delimitation analyses in this study suggested the difficulty of delimiting E. aseptata (MK584957) from E. abnormis (Fig. 3), suggesting that E. aseptata is a morphologically atypical (aseptate-ascospored) individual of E. abnormis.

Although two ITS sequences of C. patens (MT995055 = specimen no. FJ19131 and MW007918 = specimen no. FJ19049) were positioned in SHs dominated by E. abnormis, LSU and mtSSU sequences of FJ19131 and LSU sequence of FJ19049 were closely related to Chapsa spp. [Graphidaceae, Ostropales]. Since Lachnaceae and Graphidaceae are phylogenetically distant, the two ITS sequences MT995055 and MW007918 have been misidentified.

Considering that the monophyly of E. abnormis is strongly supported (Fig. 1) and members of the species share high ITS similarities (> 97%, compiled into SH1155612.08FU) (Fig. 3, Table 2), E. abnormis is accepted here as a species with some intraspecific morphological and phylogenetic variation.

Lachnum palmae formed a strongly supported clade in the phylogenetic analyses (Clade II in Fig. 1). They also shared strong selectivity to palm leaves and characteristic morphology such as thick-walled asci, hairs with resinous materials and apical amorphous materials (Suppl. material 1: Fig. S2) and ectal excipulum composed of thick-walled prismatic cells and interwoven hyphae. However, sequences labeled as L. palmae were divided into 4 to 7 SHs in all species delimitation analyses (Fig. 3), indicating that L. palmae is a species complex that includes multiple potential sister species. At present, we avoid creating new species from the complex, because the morphological and ecological differences detected among SHs are not enough to delimit species boundaries, although the size of asci and ascospores differ among some SHs, as shown in Fig. 4. Phylogenetic analyses revealed that members of the L. palmae complex belonged to Erioscyphella (Fig. 1). However, we could not judge which SH within the complex is equivalent to L. palmae as originally described from Honduras by Kanouse (1941) and redescribed by Spooner (1987) from the type plus another specimen from New Zealand. There are no L. palmae sequences from the tropical American type locality, so phylogenetic characterization and recombination of the species were avoided in the present study.