Cantharocybe specimens (Iqbal568 and 693) were collected from Madhupur upazila of Bangladesh on ground near to or associated with the roots of Cocos nucifera, a tree of the plant family Arecaceae during the monsoon (June to August) of 2012–2013. Specimens examined are deposited in the Cryptogamic Herbarium of Kunming Institute of Botany of the Chinese Academy of Sciences(KUN), China; and in the private herbarium of Iqbal (PHI). Cantharocybe brunneovelutina was previously deposited at BRH and CFMR (Ovrebo et al. 2011).

The morphological description of the basidiomata is based on field notes and documented by photographs. Color codes are according to Kornerup and Wanscher (1978). A small fragment of dried specimen was revived in H₂O, 5% KOH, and Congo red. The notation [n/m/p] is used in the descriptions of basidiospores measurements, which means n basidiospores from m basidiomata of p collections were measured; 20 basidiospores were measured from each voucher specimen. Dimension for basidiospores are given as (a–)b–c(–d), in which ‘b–c’ contains a minimum of 90% of the measured values and extreme values ‘a’ and ‘d’ are given in parentheses. Q_m = Q ± SD: Q indicates the length/width ratio of a measured basidiospore, Q_m indicates to the average of Q basidiospores and SD is the standard deviation. For the pileipellis and stipitipellis observations radial-vertical section were made halfway of the pileus and stipe, respectively. Line drawings were done free hand.

The protocol for DNA extraction followed that of Doyle and Doyle (1987). ITS1/ITS4 or ITS1/ITS5 (White et al. 1090) and LROR/LR5 (Vilgalys and Hester 1990) primer pairs were used for the amplification of the internal transcribed spacer region (ITS) and the large subunit nuclear ribosomal RNA (nrLSU), respectively. PCR amplification was carried out following the protocol of Hosen et al. (2013). PCR confirmation was confirmed on 1% agarose electrophoresis gels stained with ethidium bromide. The amplified PCR products were sent to a commercial sequencing provider company (BGI, China) for sequencing.

Three sequences (nrLSU: KF303143 and KX452406, ITS: KX452403) were generated from the Bangladeshi Cantharocybe. Additionally, two ITS sequences were also obtained from the type materials of C. brunneovelutina (BZ-1883: KX452404) and C. virosa (TENN-63483: KX452405). ITS sequences generated for this study were cloned in pMD18-T following manufacturer’s instructions. The newly generated sequences were deposited in GenBank. An initial BLASTn search of the nrLSU sequence obtained from the Bangladeshi material against the NCBI database (http://www.ncbi.nlm.nih.gov/) gave C. virosa (=Megacollybia virosa Manim. & K.B. Vrinda), C. brunneovelutina and C. gruberi as closest hits, with maximum similarities of 96%, 96% and 95%, respectively. The closest nrLSU sequences including Ampulloclitocybe and Cuphophyllus were retrieved from GenBank and additional taxa were chosen after consulting Lodge et al. (2014) and then combined with nrLSU sequence from Bangladesh materials. Two additional datasets were constructed: ITS and ITS+nrLSU to clarify relationships between Indian and Bangladesh collections of Cantharocybe. All datasets were aligned with Mafft v.6.8 (Katoh et al. 2005) and manually adjusted with BioEdit v.7.0.9 (Hall 1999) using default settings. Maximum Likelihood (ML) and Bayesian Inference (BI) methods followed those in Hosen et al. (2013). Phyllotopsis nidulans (Pers.) Singer was served as outgroup for all dataset analyses as inferred from other phylogenetic studies (Ovrebo et al. 2011, Kumar and Manimohan 2013, Lodge et al. 2014).

Both ML and BI analyses were conducted using RAxML v.7.2.6 (Stamatakis 2006) and MrBayes v.3.1.2 (Ronquist and Huelsenbeck 2003) with default settings. For ML analyses, the General Time Reversible Model of evolution with estimated gamma distribution was selected, and statistical support values were obtained using nonparametric bootstrapping (BS) with 1000 replicates. For BI analysis, the substitution model suitable for ITS and nrLSU datasets were determined using the Akaike Information Criterion (AIC) implemented in MrModeltest v.2.3 (Nylander 2004), and the models were SYM+G and GTR+I+G, respectively. Bayesian Inference analyses were conducted using the selected evolutionary model with four chains and generations set to 0.5 million for the ITS and nrLSU datasets, and four million for the combined dataset (ITS+nrLSU). Runs were terminated once the average standard deviation of split frequencies went below 0.01. Trees were sampled every 100 generations, with the first 25% of trees discarded as burn-in. Posterior probabilities (PP) were calculated using the “sump” and “sumt” commands implemented in MrBayes.

Three datasets (ITS, nrLUS and ITS+nrLSU) are constructed and analysed separately. The ITS dataset includes 17 sequences of fungal taxa (Table 1) and consisting of 1052 nucleotide sites (gaps included) of which 417 are parsimony informative. The nrLSUdataset contains 19 sequences of the same taxa (Table 1) as well as two additional sequences of Cantharocybe with 934 aligned nucleotide sites (gaps included) of which 131 are parsimony informative. The combined (ITS+nrLSU) dataset includes 17 sequences of fungal taxa (Table 1) and consists of 1986 nucleotide (first 1052 for ITS and the next 934 for LSU, gaps included) sites including 541 that are parsimony informative. The aligned datasets are deposited in TreeBASE (S19556).

The ITS sequences of C. brunneovelutina and C. virosa are longer than normal because a ca. 210 bp intron is inserted about 70 bp after the ITS1 primer. This intron was not previously known as Blast searches turned up no matches, but the intron sequences were similar between Cantharocybe species. The GenBank ITS sequence DQ200927 of C. gruberi was found to be truncated on the 3' end, where the intron was likely inserted. We infer that the partial ITS sequence of C. gruberi deposited in GenBank as part of the Assembling the Fungal Tree of Life (AFTOL) project was truncated because the first 70 bp of the ITS are missing beginning at a point which coincides with the position of the intron insertion in C. brunneovelutina and C. virosa. Lodge et al. (2014) found that introns which were inserted within 100–150 bp of a primer disrupted replication unless the DNA was cloned using a vector such as the method we used to obtain our ITS sequences. As there is no notation in GenBank that the AFTOLITS sequence of C. gruberi was cloned, we infer that forward reads were disrupted by the same intron that was found in C. brunneovelutina and C. virosa. Thus, it is likely that Matheny and Hibbett’s AFTOL program only obtained back-reads from the 5' end using the ITS4 primer. The authors of the C. gruberiDQ200927 sequence may or may not have obtained a partial read of an intron, but if so, it would have been of diminishing quality with distance from the ITS4 primer, it would not have matched any known ITS sequences, and it would have been impossible to correct or corroborate without a forward read from the 3' end. We therefore infer that if a partial read of an intron was obtained by Matheny and Hibbett for C. gruberi, that it was trimmed from the GenBank submission because it did not match ITS1 sequences and it could not be corrected or corroborated. Blast searches of GenBank using the ITS sequence of C. virosa, after removing the intron, retrieved C. gruberi sequences with highest similarity.

Tree topologies obtained from both ML and BI methods of phylogenetic analyses are congruent, the ML trees are shown in Fig. 1. The Bangladesh sample of Cantharocybe clusters in a strongly supported clade with the Indian sample of C. virosa in all three datasets, indicating that they are conspecific (Fig. 1).

Figure 1. 

Phylogenetic relationships of the genus Cantharocybe inferred from nrLSU (a), ITS (b) and ITS+nrLSU (c) sequences using maximum likelihood (ML). RAxML bootstrap supports values (>50%) and Bayesian posterior probabilities (>0.95) are indicated on the branches at nodes (ML/PP). Newly generated sequences are highlighted in bold. GenBank accession or voucher numbers are provided after the species name.

The Bangladeshi C. virosa is characterized by its gray to grayish brown basidiomata, moderately crowded lamellae, fine squamules on stipe surface formed from clusters of lecythiform caulocystidia, ellipsoid to broadly ellipsoid basidiospores, and a trichoderm pileipellis.

Based on molecular analyses, C. virosa, a species recently described from India is conspecific with the Bangladeshi collection (Fig. 1). However, the Indian C. virosa has a pale grayish brown pileus, long cheilocystidia which can be up to 63 µm with a long neck up to 35 µm, a cutis pileipellis or occasionally disrupted with trichodermal patches (Kumar and Manimohan 2013). In comparison, the Bangladeshi collection has a grayish brown to dark brown pileus, cheilocystida with short neck up to 15 µm long, and clearly defined trichoderm pileipellis. Geographically, C. virosa is distributed in the Kerala state (South-West region) of India, while the new collection was collected from Tangail district of the Dhaka division of Bangladesh. Moreover, the nrLSU sequence obtained from the Bangladeshi collection does not perfectly match (96%) with the holotype C. virosa retrieved from GenBank, and the genetic distance between them is 0.96% (8 bases differences and 23 deletions in the Indian collection) of 831 nucleotide sites. These morphological variations, geographic distance and nrLSU sequence inferred suggest that they may have diverged recently from each other due to its allopatric speciation.

Surprisingly, when we blasted the newly generated ITS sequences from the holotypes of C. virosa (TENN 63483) and C. brunneovelutina (BZ-1883) individually against the NCBI database, we did not find C. gruberi as the closest sister species among the first 100 matched species, even 80–88% matched only with some taxa of Tricholoma, Lepista, Macrolepiota, Lepiota, etc. The ITS sequence from the newly collected material from Bangladesh also gave a similar result. These results are caused by the presence of a large intron in C. brunneovelutina and C. virosa, and the truncated sequence of C. gruberi deposited in GenBank (missing the intron and 3' end). Subsequently, we retrieved the closest ITS sequences of those taxa including C. gruberi from GenBank and reconstructed an ITS phylogenetic tree where all Cantharocybe taxa were nested together within the same clade with strong BS value (99% MLBS, 1.0 PP) and apart from Tricholoma and Lepista clades (data not shown). In the ITS sequence analyses, the Indian entity C. virosa showed only 5 base pair difference with 4 deletions out of 742 nucleotides (genetic distance 0.68%) from the Bangladeshi material. Further extended combined dataset (ITS+nrLSU) also showed little divergence (genetic distance 0.83%) between them (Fig. 1c). This small variation in ITS sequences, with negligible differences in the color of the basidioma, size of basidia and cheilocystidia which were possibly due to environmental variations, do not warrant a new variety or species, suggesting that both south Asian entities are conspecific. Furthermore, both collections were the same ecology and associated with the roots of C. nucifera in a tropical region. Neither coconut trees nor palms in general have been shown to associate with ectomycorrhizal fungi. Halbwachs et al. (2013) however, found hyphae of a Cuphophyllus species (a genus near Cantharocybe) and several species of Hygrocybe as endophytes in plant roots including those of a monocot (Plantago major) so another type of root symbiosis with C. nucifera is possible. The considerable variation in the nrLSU sequence (96% match, 8 bases differences with 23 deletion) of the Indian C. virosa may be explained by the fact that the nrLSU sequence obtained from that collection was not clean, showing evidence of a contaminating sequence or minor indel (K.W. Hughes, pers. comm.).

Cantharocybe brunneovelutina and C. gruberi can also be separated from C. virosa morphologically. Cantharocybe brunneovelutina differs from C. virosa by its velutinous basidioma, unusual cheilocystidia with multiple prong-like appendages at the apex resembling a basidia-like structure, and a trichoderm pileipellis (Ovrebo et al. 2011). The type species of this genus, C. gruberi, has a pale yellow to lemon yellow pileus and narrowly elliptical to oblong basidiospores measuring 11–16(–17) × (4.5–) 6–7.5 µm (Bigelow and Smith 1973).

Molecular phylogenetic analyses indicated that the genus Cantharocybe is monophyletic, with strong bootstrap values (Fig. 1). Likewise, Ovrebo et al. (2011) and Lodge et al. (2014) showed that the monophyletic clade of Cantharocybe has strong BS value comprising C. gruberi and C. brunneovelutina using single locus or multi-locus sequence analyses. Although recent phylogenetic studies (Ovrebo et al. 2011, Kumar and Manimohan 2013, Lodge et al. 2014) showed the monophyly of Cantharocybe, the sister relationship with other genera remains unclear. Cantharocybe nests at the base of the hygrophoroid clade together with Ampulloclitocybe and Cuphophyllus (Binder et al. 2010, Matheny et al. 2006, Ovrebo et al. 2011, Lodge et al. 2014), but their relationships were not confidently resolved. In our combined dataset (ITS+nrLSU) analysis, Ampulloclitocybe is only weakly supported (65% MLBS, PP = 0.96) as sister to the Cuphophylloid clade (Fig. 1c). This is accordance with the recent phylogenetic study of Lodge et al. (2014).

Cantharocybe is an uncommon genus that only consists of three species. Cantharocybe gruberi has wide distribution from America (New Mexico, western North America and British Columbia) to Europe (Spain). Cantharocybe brunneovelutina is reported from tropical Central America (Belize) whereas C. virosa is from tropical South Asia (Bangladesh and India). Based on the branching order with strong bootstrap support at all nodes in our phylogeny in which C. gruberi is basal, we infer that the genus Cantharocybe may have originated in America or Europe and then migrated independently to Central America and South Asia.

The south Asian species usually occurs with the roots of Cocos nucifera (Manimohan et al. 2010, Kumar and Manimohan 2013, this study as well), the North American and European collections were found on needle beds or ground under conifers and Pinus nigra, respectively (Bigelow and Smith 1973, Esteves-Raventós et al. 2011), and the Central American species was found in humus around dead palm trees (Ovrebo et al. 2011). However, their symbiotic association with trees is still unknown. To facilitate identification of Cantharocybe taxa worldwide, a key to the species is given bellow.