In this study, we examined 144 herbarium specimens of the genus Subulicystidium, which were collected in several regions of Paleotropics (Réunion Island, Madagascar, Africa, South-East Asia) and Neotropics (Caribbean region, various countries of South America). This material was collected during the last six decades, with the oldest collection (PDD13816) from 1954 and the most recent ones from 2015 (e.g. KAS:L 1860). Collections are preserved in the following herbaria: O (Natural History Museum, Oslo University, Norway), GB (Gothenburg University, Sweden), MG (Museu Paraense Emílio Goeldi, Belém, Brasil), SP (Instituto de Botânica, São Paulo, Brasil), KAS (University of Kassel, Germany), FR (Senckenberg Research Institute and Natural History Museum, Frankfurt am Main, Germany) and LY (University of Lyon, France). We examined also holotype specimens of Subulicystidium meridense Oberw. (TUB, Tübingen University, Germany), S. nikau (G. Cunn.) Jülich (PDD, New Zealand Fungal Herbarium, Landcare Research, Auckland) and the collection of S. allantosporum Boidin and Gilles ad interim (Boidin and Gilles 1988) from LY. Attempts to obtain the type specimen of S. brachysporum from PREM (Plant Protection Research Institute, Queenswood, South Africa) were not successful.

For a better biodiversity data availability and reusability, in Suppl. material 1, the table with detailed information on all 144 specimens examined is provided. If missing on the original specimen labels, data on higher-and lower rank administrative units were mined and added to the corresponding columns in Suppl. material 1. The table also includes geographic coordinates for each specimen in decimal degrees format (DD) with minus signs used to indicate southern and western hemispheres. Originally, geographic coordinates were available for 65 specimens. For the other 69 specimens, an attempt to estimate the coordinates from a map was made using resources Google Maps (via http://www.gpskoordinaten.de/), OpenStreetMap (https://www.openstreetmap.org) and the georeferencing calculator of Wieczorek and Wieczorek (2015). In ten specimens, the locality was not precisely indicated to estimate the coordinates. The manner, in which coordinates were obtained, was specified for each specimen in the Suppl. material 1.

Field data and photos of recent collections from Réunion Island (stored in FR and KAS) are accessible via PlutoF workbench (Abarenkov et al. 2010) under the project “Ordynets_Fungi of Reunion Island” and as a part of GBIF occurrence dataset of the Senckenberg herbarium FR (Senckenberg 2018).

Sections from dried herbarium specimens were examined in 3% aqueous solution of potassium hydroxide (KOH) mixed with 1% aqueous solution of Phloxine, using 100× immersion oil lens of a Leica DM500 light microscope. Images were captured with a built-in ICC 50 HD Camera using Leica Application Suite EZ V.3.2.1 software (Leica Microsystems Ltd., Switzerland). Measurements were done with the software “Makroaufmaßprogramm” from Jens Rüdigs (https://ruedig.de/tmp/messprogramm.htm) and analysed with the software “Smaff” version 3.2 (Wilk 2012). At least 30 basidiospores per specimen were measured where possible for the sequenced specimens or otherwise important collections. When referring to the basidiospore measurement results, abbreviation L was used for the spore length, W for the spore width, Q for the length to width ratio and N x/y for the x number of spores measured from y specimens.

The raw spore measurements were undertaken as follows. First, for each collection, automated search for size outliers was performed with the “Smaff” software (Wilk 2012). To account for the outliers in both spore length and width simultaneously, the parameters of (i) length multiplied by width and (ii) volume were calculated for each spore by the programme. These values, (i) followed by (ii), were checked to represent the outliers in the sample on a 95% probability level, using simultaneously the tests of Verma and Quiroz-Ruiz (2006), David et al. (1954) and Grubbs (1950). Upon their detection, outliers were excluded from the sample as recommended by Wilk (2012). This procedure usually resulted in a better fit of the spore measurements to the normal distribution. The spore measurements after excluding outliers are provided in the Suppl. material 2.

These filtered spore measurements were used to calculate the spore size range of the species. The main range was presented as the interval into which 90% of non-outlier measurements fall, while 5% of the smallest and 5% of the largest non-outlier measurements were included in parentheses. For the species with more than one specimen available, the filtered spore measurements were pooled together and the main range (90% of the data) with 5% of the smallest and 5% of the largest values were defined for this pooled sample. Calculations were done in R version 3.3.3 (R Core Team 2017) and script is available from Ordynets and Denecke (2018). Additionally, for species with at least three sequenced specimens, hypothetical intervals were calculated within which 90% of all existing individuals’ specimen mean values lie. This way of representing basidiospore size variability in species was highly recommended by Parmasto and Parmasto (1987) and Raitviir (1972). These 90% tolerance intervals were calculated for the 90% probability level, with the method of Howe (1969) implemented in the “normtol.int” function of the “tolerance” R package version 1.3.0 (Young 2010). R script for these calculations is available in Ordynets (2018b).

At least 10 basidia and cystidia were measured per specimen and their size variation was presented simply as the range between minimum and maximum values for the pooled measurements of all collections belonging to one species. When basally swollen cystidia was a regular feature, both the largest diameter at the place of swelling and the diameter next to the swelling were noted. The protruding bow-tie crystals were included in the measurements of cystidium diameter. The shape of the cystidium apex followed terminology for sterile hymenial elements of Yurchenko and Wu (2016) and included the following options: tapering, acute and acuminate. Cystidial ornamentation was described as seen under the light microscope.

Sequences of two nuclear ribosomal DNA regions were considered in our study: internal transcribed spacer (ITS) and ribosomal large subunit-coding DNA (28S). Sequences were obtained from dried herbarium specimens. Total DNA was isolated according to the protocol of Izumitsu et al. (2012). For that, 1-2 mg of fungal fruit body tissue were suspended in 100 µl TE buffer in a 1.5 ml tube. The tubes were microwaved (700 Watt) for 1 min two times, with a 30 seconds pause while keeping the tubes at room temperature. Tubes were cooled at -20°C for 20 min and centrifuged at 10000 rpm for 5 min. The supernatants were 10 or 100 times diluted and in this form used in PCR.

Primer pairs used to amplify the complete ITS region were ITS1F/ITS4, ITS1/ITS4 and ITS1/ALR0 (White et al. 1990, Gardes and Bruns 1993, Collopy et al. 2001). The D1–D2 domains at the 5' end of 28S were amplified with primer pairs NL1/NL4 (O’Donnell 1992) and less frequently with LR0R/LR5 (Hopple and Vilgalys 1999). PCRs of the collections from Réunion Island (FR and KAS herbaria) were performed as explained in Ordynets et al. (2015). The PCR for the remaining material was performed on 53 µl solution containing 5 µl of extracted DNA and 48 µl Master Mix (BIOLINE GmbH, Luckenwalde, Germany). One Master mix portion contained 30.2 µl H₂O, 10 µl reaction buffer (30 mM MgCl₂) coloured with red and orange dyes, 2 µl MgCl₂ (50 mM), 2 µl dNTPs (6 mM), 2 µl Bovine Serum Albumin (20 µg/µl), 0.8 µl of each forward and reverse primers (25 pM) and 0.2 µl Mango-Taq DNA Polymerase (5 units/µl).

Amplifications were performed in 96-well TGradient Thermocycler (Biometra, Göttingen, Germany). PCR with primer pairs ITS1F/ITS4, ITS1/ALRO and NL1/NL4 was set as initial denaturation at 94°C for 3 min followed by 29 cycles of denaturation 94°C for 30 s, annealing 55°C for 45 s and extension 72°C for 60 s; final elongation was done at 72°C for 7 min. PCR with primer pair LR0R/LR5 differed only in having the annealing temperature as 48°C.

PCR products were checked on 1% agarose gel stained with GelRed fluorescence dye (BIOTIUM, Hayward, CA, USA) in the Transilluminator Biometra Ti5 equipped with BioDocAnalyze software (Biometra GmbH, Göttingen, Germany). PCR products were cleaned with QIAquick PCR Purification Kit according to manufacturer’s instructions (QIAGEN GmbH, Hilden, Germany). Sanger sequencing of purified products was performed in the facilities of the Senckenberg Research Institute and Natural History Museum (Frankfurt am Main, Germany) and by the company GATC Biotech AG (Constance, Germany). The primers used for sequencing were identical to those used for amplification.

The oldest specimen we succeeded to sequence, with regard to both ITS and 28S regions, was from the year 1978 (LR 15483 in O:F 918488). Attempts of DNA amplification from the type specimens of S. meridense, S. nikau and S. allantosporum Boidin ad interim (Boidin and Gilles 1988) were not successful, as well as an attempt to sequence the type specimen of the new species Subulicystidium ryvardenii Ordynets, Langer & K.H. Larss. sp. nov. We did not succeed in amplifying two protein-coding genes from any of the specimens: a partial segment (511 bp) of the translation elongation factor 1α (TEF1α) with EF-595f and EF-1160r primer pair (Kauserud and Schumacher 2001), as well as the largest subunit of RNA polymerase II gene (RPB1) with RPB1-Af and RPB1-Cr primer pair (Matheny et al. 2002).

All sequences obtained in this study went through the standard quality assessment steps outlined by Nilsson et al. (2012). Raw sequence data were processed with Geneious version 5.6.7 (http://www.geneious.com, Kearse et al. 2012). For various sequence format conversions and alignment viewing, Mesquite version 3.40 (Maddison and Maddison 2018), AliView version 1.19 (Larsson 2014) and Seaview version 4 (Gouy et al. 2010) were used.

In this study, 49 sequences of ITS rDNA region and 51 sequences of 28S rDNA region of Subulicystidium were generated and submitted to GenBank (Benson et al. 2013). They are available as accessions MH041511-MH041559 for ITS and MH041560-MH041610 for 28S region. Additional ten ITS and six 28S sequences of Subulicystidium, earlier available in GenBank and UNITE database (Kõljalg et al. 2013), were downloaded and used in our analyses. Finally, we included sequences of Brevicellicium exile (H.S. Jacks.) K.H. Larss. & Hjortstam and B. olivascens (Bres.) K.H. Larss. & Hjortstam to serve as an outgroup in our sequence-based analyses. All sequences used in the study are listed with brief metadata in Table 1.

Sequences from each locus, ITS and 28S, were pre-aligned in Geneious version 5.6.7 (Kearse et al. 2012) with MUSCLE algorithm (eight iterations) (Edgar 2004). Final alignments of each locus were produced in the online mode of MAFFT version 7 (Katoh et al. 2017), with L-INS-i algorithm and other settings as default.

The small fragments of 18S rDNA and 28S rDNA were automatically trimmed from the target ITS region with the ITSx software (Bengtsson-Palme et al. 2013) implemented in the PlutoF workbench (Abarenkov et al. 2010). The same tool was used to partition ITS into ITS1, 5.8 and ITS2 regions prior to phylogenetic analyses of ITS alignment, to estimate the evolutionary model parameters for each partition separately. The 28S alignment was trimmed manually to produce the sequences of the same lengths and with fewer (if any) gaps at both ends and was not partitioned. Key properties of the final alignments were explored and described using RAxML terminology (Stamatakis 2016).

Morphologically outlined species were compared in terms of genetic distances estimated separately for the trimmed ITS and 28S alignments. For this, raw (also called uncorrected) pairwise dissimilarities of sequences in each alignment were calculated, defined as the percentage of sites that differ between each two full-length sequences including gap positions (Schoch et al. 2012, Kõljalg et al. 2013). This procedure was done with the “dist.dna” function of “ape” R package (Paradis et al. 2004) with option pairwise.deletion=FALSE (i.e. without deleting the sites with missing data in a pairwise way). Results were visualised with the ggplot2 R graphics (Wickham 2009) and R script can be viewed in Ordynets (2018a). Pairwise sequence dissimilarities were further analysed on the intraspecific versus interspecific level. The two levels of sequence variability were segregated with the “sppDist” function of the “spider” R package (Brown et al. 2012) and plotted simultaneously as histograms in a search for the barcoding gap (Meyer et al. 2005). As this classical approach provides only a general overview of sequence variability, i.e. for the pooled dataset, we applied also the recommended alternative which considers the species identity. For each sequence, the maximum intraspecific distance was contrasted with the minimum interspecific distance as recommended by Collins and Cruickshank (2012). Both types of distance for each sequence were estimated, respectively, with the functions “maxInDist” and “nonConDist” of the “spider” R package (Brown et al. 2012) and visualised as a scatterplot. R script for these procedures is available in Ordynets (2018c).

All phylogenetic analyses were performed using the GTR+G evolutionary model. We performed separate analyses of the ITS alignment (partitioned into ITS1, 5.8 and ITS2 regions), unpartitioned 28S alignment and concatenated ITS+28S alignment partitioned into four regions (ITS1, 5.8, ITS2 and 28S). For Bayesian inference of phylogeny, MrBayes 3.2.3 (Ronquist et al. 2012) was used. Two independent MCMC processes, each in 4 chains, were run. Five million trees were generated, sample frequency was set to 1000 and burnin fraction to 0.2. The acceptability of selected settings and mixing sampled trees, were confirmed by the standard deviation of split frequencies, by the potential scale reduction factor, by the sum of average effective sample size in two runs and by tracing likelihood scores of generated trees with Tracer 1.6 (Ronquist et al. 2011, Rambaut et al. 2014). For 8002 sampled trees per analysis (burn-in fraction excluded), a majority rule consensus tree was computed with branch supports representing the relative frequencies of bipartitions (posterior probabilities). Maximum likelihood analyses were performed in RAxML 8.2.10 (Stamatakis 2014). The search for the best-scoring maximum likelihood tree and bootstrap analysis (1000 replicates) were performed in a single run. Both RAxML and MrBayes were run on CIPRES Science Gateway V 3.3 (Miller et al. 2010; http://www.phylo.org). Resulting phylogenetic trees were first viewed in FigTree v. 1.4.2 (Rambaut 2014). Further visualisation and annotation of the phylogenetic trees were done in R version 3.3.3 (R Core Team 2017) and R script is available in Ordynets (2018d). The multiple sequence alignments, details of phylogenetic analyses and trees generated in the study were deposited in TreeBASE: http://purl.org/phylo/treebase/phylows/study/TB2:S22473.

The aligned ITS dataset included 59 Subulicystidium sequences and two outgroup sequences of Brevicellicium. The dataset consisted of 671 characters (gaps included) and contained 477 distinct alignment patterns, namely 238 in ITS1, 28 in 5.8S and 211 in ITS2 region. The proportion of gaps and completely undetermined characters in this alignment was 20.02%.

Aligned ITS sequences fell into several dissimilarity categories. All the Subulicystidium sequences were at least 10% different from two Brevicellicium sequences (outgroup), as well as from single sequence of Subulicystidium oberwinkleri (Fig. 1a). The sequences of S. robustius were at least 3% and maximum 10% different from the rest of the genus. The sequences of S. harpagum and S. parvisporum were most distant from S. robustius (7–10%) and 3-7% distant from the rest of the genus. Sequences mostly belonging to morphospecies S. meridense and S. brachysporum formed four groups within which they were all 0-3% (in many cases only up to 1%) dissimilar. One of these groups included also sequences of S. fusisporum and S. tedersooi, another group—S. longisporum and S. grandisporum and the third—single sequence of S. obtusisporum. Therefore, both easier and harder distinguishable species, in terms of ITS region identity, were found in the dataset.

The pattern seen through a visual inspection of the ITS sequence dissimilarity matrix was confirmed by the barcoding gap analysis. Throughout the dataset, intraspecific and interspecific distances strongly overlap and no universal for the genus Subulicystidium barcoding gap could be detected (Fig. 2b). Mean and maximal intraspecific distances were 2.87 and 7.73%, while mean and minimal interspecific distances were 5.06 and 0%, respectively. At the level of individual species, a barcode gap existed for S. fusisporum, S. parvisporum, S. robustius and S. tedersooi (Fig. 2a).

The aligned 28S dataset included 57 Subulicystidium sequences and two outgroup sequences of Brevicellicium. The dataset consisted of 617 characters (gaps included) and contained 246 alignment patterns, while the proportion of gaps and completely undetermined characters was 7.54%.

Pairwise 28S sequence dissimilarities were structured differently compared to the ITS dataset (Fig. 1b). The most distinct species in terms of 28S identity was S. oberwinkleri. The dissimilarity of its two sequences from the rest of Subulicystidium and two Brevicellicium sequences was 10–20%. The next most distinct group was formed by the sequences of S. harpagum and S. parvisporum which were 7–10% dissimilar from the rest of the genus except one group containing S. meridense and S. brachysporum sequences (2–3%). The majority of dissimilarities lay in the range 1–5% and were clearly grouped (Fig. 1b).

In a whole 28S dataset, intraspecific and interspecific distances strongly overlapped and thus showed no universal for the genus Subulicystidium barcode gap (Fig. 2d). Mean and maximal intraspecific distances were 2.52 and 12.5%, while mean and minimal interspecific distances were 5.58 and 0%, respectively. At the level of individual species, the barcode gap was evident for S. oberwinkleri, S. fusisporum, and S. robustius (Fig. 2c).

Bayesian analysis (BA) of the ITS alignment was finished with the standard deviation of split frequencies of 0.008 (equals average) and was characterised by the average potential scale reduction factor 1.00 (maximal 1.002) and pooled effective sample size from two MCMC runs 4151.3494. Maximum likelihood analysis (ML) resulted in a tree with a final optimisation log likelihood of -7019.372. BA produced a tree with a partly similar topology to ML tree but contained large polytomy at one of the basal nodes. Hereinafter we present and discuss the topology of the BA tree plotted with both posterior probabilities (pp) from BA and bootstrap supports (bs) from ML.

The phylogenetic tree, generated for the ITS dataset, contains monophyletic and polyphyletic taxa as well as several species represented by a single sequence (Fig. 12). Subulicystidium oberwinkleri is the most basal member of the ingroup. Other singletons descending from the basal nodes are S. nikau, the most deviating sequence of the morphospecies S. brachysporum (TU 110416) (see Discussion for the explanation) and S. rarocrystallinum. The clade dominated by Reunionese collections (pp=1, bs=87%) contains the new species S. parvisporum (pp=1, bs=99%) and S. harpagum (pp=1, bs=98%). The latter includes also L. Ryvarden’s collection from Colombia (LR 15736 in O:F: 918487). Two sequences of S. perlongisporum are also placed in the basal part of the tree but do not form a separate clade and, moreover, one sequence (from LE 302156) forms a clade with S. boidinii (KAS:L 1584a). S. robustius is recovered as a distinct clade (pp=1, bs=100%) of four sequences from Neotropics subtended by a long branch. S. fusisporum, represented by three sequences from the Caribbean region (pp=1, bs=98%), is a sister species to S. tedersooi represented by two sequences from Vietnam (branch support pp=1, bs=98%).

The remaining three clades each contain a mixture of sequences belonging to the morphospecies S. brachysporum and S. meridense with their likes. One clade contains also single sequences of S. obtusisporum from Germany (FR: W213-3-I) and S. harpagum from Jamaica (GB:KHL 10733). This clade is joined by three sequences of S. brachysporum: first by two sequences from Réunion (KAS:L 1498 and 1795) and at the next ancestor node with one sequence from Costa Rica (GB:KHL 11216). Another large clade is roughly equally rich in sequences of S. brachysporum and S. meridense (pp=1, bs=100%) and joined by a single sequence of S. inornatum (pp=0.88, bs=41%). One large clade (pp=1, bs=99%) included more collections of S. brachysporum, mostly sensu Boidin and Gilles (1988) and less of S. meridense, but also a single sequence of S. grandisporum (LR 29162 in O:F 506781). One more sequence of S. brachysporum (GB:KHL 10411) formed a weakly supported clade with the three sequences of S. longisporum from Europe (pp=0.84, bs<50%).

Bayesian analysis (BA) of the 28S alignment was finished with the standard deviation of split frequencies of 0.004 (equals average) and was characterised by the average potential scale reduction factor 1.00 (maximal 1.005) and pooled effective sample size from two MCMC runs 4673.55. Maximum likelihood analysis (ML) resulted in a tree with a final optimisation log likelihood of -2209.83. BA produced the tree with the topology highly similar to that of the ML tree. Hereinafter we present and discuss the topology of the BA tree plotted with both posterior probabilities (pp) from BA and bootstrap supports (bs) from ML, mostly focusing on differences from the results obtained for the ITS dataset.

The most basal ingroup members on the 28S tree were S. oberwinkleri (clade with two sequences, pp=1, bs=100%), S. rarocrystallinum and S. harpagum from Jamaica (GB:KHL 10733) (Fig. 13). The species S. harpagum and S. parvisporum remained in the single clade (pp=1, bs=92%) but were recovered as polyphyletic due to the placement of sequence from the KAS:L 1226 amongst the sequences of S. harpagum. They were joined by the clade containing sequences from S. obtusisporum (GB: KHL 10622) and S. brachysporum (TU 110416). The remaining sequences, mostly belonging to S. brachysporum and S. meridense, occupy upper nodes of the tree without clear grouping by morphospecies. 28S dataset was importantly enriched by the sequences from specimens, for which the ITS region could not be sequenced: GB:KHL 10566, O:F 506782 and LY 12772 (S. brachysporum sensu Boidin and Gilles), LY 11378 (S. brachysporum sensu Talbot), GB:KHL 11355 and LY 12816 (S. meridense), KHL 15325 (S. aff. meridense) and LY 12750 (S. parvisporum).

Bayesian and Maximum likelihood phylogenetic analyses of the concatenated alignment (ITS+28S) resulted in a tree topology which was contributed by both ITS and 28S regions (Fig. 14). In line with the ITS-based tree, S. boidinii and S. harpagum were recovered as polyphyletic but S. parvisporum as monophyletic (cf. Fig. 12). Furthermore, the German sequence of S. obtusisporum was found in the brachysporum-meridense clade and not between the sequences of S. longisporum (cf. Fig. 13). On the other hand, similarly to 28S-based tree, S. fusisporum and S. tedersooi intruded the brachysporum-meridense clades. Additionally, the 28S portion of concatenated alignment contributed to resolving a polytomy at one of the basal node containing sequences of S. nikau and S. rarocrystallinum.

We measured in total 2840 basidiospores from 67 specimens of Subulicystidium. We defined three groups of species according to the principal basidiospore shape: species with fusiform, cylindric and allantoid basidiospores. We found that some of the species could be delimited based on the basidiospore morphology solely, while, for other species, this was not possible and additional morphological characters had to be considered.

The species with fusiform basidiospores are barely distinguishable according to the basidiospore length. It varied generally from 8 to 11 µm, while the mean value did not exceed 10 µm (Fig. 11a). The only exception was S. fusisporum which had spores 10.7–12.3 µm long (main range, i.e. 5–95% quantiles of measurements data) and 11.5 µm long in average. Three species, viz. S. robustius, S. inornatum and S. fusisporum were indistinguishable in the spore width which varied for three of them between 2.5 and 3.5 µm. In contrast, S. ryvardenii had broader basidiospores with the main range 3.5–4.2 µm and mean value 3.8 µm. S. naviculatum was distinguished by the broadest fusiform basidiospores (main range 4.3–5.0 µm, mean value 4.6 µm), while S. tedersooi by the narrowest fusiform basidiospores (main range 2.1–2.6 µm, mean value 2.4 µm). The spore length to width ratio was as useful as spore width to discriminate the species and was remarkably the lowest in S. naviculatum (2.0–2.5, mean 2.2) and the highest in S. tedersooi (3.5–5.0, mean 4.2).

Under allantoid basidiospores, we considered those with adaxial side clearly concave and having length to width ratio around 2, thus looking rather as reniform or phaseoliform. Amongst the species with such spores, S. oberwinkleri had distinctly the longest and the broadest spores: mean length and width were 9.2 and 4.7 µm, respectively (Fig. 11b). S. nikau could be distinguished from S. boidinii by broader spores. The spore width in S. nikau was 3.9–4.6 µm (mean 4.2 µm) and length-width ratio is 1.6–2.0 (mean=1.8), while S. boidinii has spores 2.8–3.3 µm broad (mean 3.1 µm) and length-width ratio 1.9–2.4 (mean=2.1). Comparing our own data with the Cunningham’s specimen of S. nikau (holotype) and Boidin’s specimen of S. boidinii showed that we have the same understanding of the respective species as the mentioned authors. S. harpagum and S. parvisporum had rather overlapping spore width and length to width ratio but differed in the spore length: 5.6–8.3 µm (mean 6.7 µm) in the former versus 5.0–6.2 µm (mean 5.6 µm) in the latter.

Species with cylindric basidiospores (Fig. 11c) were characterised by the average length to width ratio of at least 2.8 and were well distinguished by the mean spore length: 12.7 µm in S. grandisporum, 10.7 µm in S. obtusisporum and 9.2 µm in S. cylindrosporum. The collections of the meridense-brachysporum morphogroup were also characterised by cylindrical basidiospores but of the smaller length and were less clearly distinguishable. Examples of single collections representing each morphogroup (Fig. 11d) showed that S. brachysporum sensu Boidin and Gilles (1988) had on average slightly longer basidiospores than S. brachysporum sensu Talbot (1958), viz. 7.9 µm versus 7.3 µm. Curved spores of the classical S. meridense were on average shorter than straight spores of S. aff. meridense, viz. 6.9 µm versus 8.2 µm. Spores in the type specimen of S. meridense were of the intermediate average length compared to the two former examples (7.4 µm). The spore width and length to width ratio were very much overlapping in the material of meridense-brachysporum morphogroup.

With reference to the newly obtained data, in the next section we present the key to the genus Subulicystidium. We used the successful key of Gorjon et al. (2011) as the basis and a source of the information on the long-spored taxa.

In this study, we describe 11 new species of Subulicystidium based on morphological evidence and rDNA ITS and 28S sequence analyses. Ten of these species are characterised by a unique combination of basidiospore and cystidium morphology and rDNA sequence identity. One species (S. ryvardenii) could not be sequenced but the morphological evidence itself was sufficient for describing it as a new species. With our contribution, the number of the known species in the genus Subulicystidium now totals 20. The provided morphological key to all known species should facilitate identification of specimens previously treated as highly variable S. longisporum or left without species name. Such literature is urgently needed to assist in tropical fungal inventories.

We revised also the morphological and genetic borders of the five previously known species. One of them, S. naviculatum, could not be sequenced, while for S. nikau, only one specimen with amplifiable DNA was available. For the two morphospecies, S. brachysporum and S. meridense, numerous sequences from different localities were obtained. Our data show that, despite differences in the protologues, the species are hard to separate morphologically and molecularly. They share, to a large extent, basidiospore size and shape as well as highly similar ITS and 28S sequences, leading to strongly intermixed clades in phylogenetic trees. Therefore, S. brachysporum and S. meridense, in current understanding, are highly polyphyletic.

Our study is based on examination of a large set of specimens from numerous localities in Paleo- and Neotropics. Upon this, we could show that the diversity of the short-spored Subulicystidium species is much higher than previously known. The newly described S. robustius is in fact a frequently occurring species in the Caribbean region and in South America. Furthermore, we could report a multicontinental distribution for several species, verified by DNA sequence data. S. brachysporum, S. boidinii, S. harpagum and S. oberwinkleri are typified by material from Paleotropics (South Africa in the first species and Réunion in three others), but were found by us also in South America. The morphospecies S. meridense, described from Venezuela (Oberwinkler 1977) and later found in Costa Rica (Kisimova-Horovitz et al. 1997), was also found on Réunion Island by Boidin and Gilles (1988). In addition to sequenced collections from a few more countries in South America, we confirmed the species presence on Réunion by sequencing collections of Boidin and Gilles (1988). In this study, we also report S. meridense for the first time from South-East Asia (Taiwan). The new species S. fusisporum was first considered by us as a Caribbean endemic. Re-identification as S. fusisporum of the specimen collected by G. Gilles in Côte D’ivoire (LY 7375, originally labelled as S. longisporum, DNA could not be amplified) may suggest that the species is present in West Africa as well.

It was surprising for us to find the species occurring on more than one continent or on islands separated by thousands of kilometres. For fungi with spores carried by wind, dispersal limitation was shown to act strongly even at small spatial scales (Peay et al. 2010, Norros et al. 2012). Given the architecture and location of Subulicystidium fruit-bodies (next to the ground, not rarely underside of the logs), one would expect prevailing spore dispersal distance smaller than 1 m (Galante et al. 2011). However, in macrofungi, there remains a probability of spore travel on a distance of kilometres (Nordén and Larsson 2000, Peay et al. 2012, Norros et al. 2014) and also overseas (Geml et al. 2012). Spore morphology traits have been recently discussed in a connection with dispersal and arrival success of a species (Norros et al. 2014, Calhim et al. 2018). In this regard, the genus Subulicystidium, with a high diversity of spore size and shape between species, is an interesting object to correlate spore traits and biogeography in future studies.

Subulicystidium brachysporum (P.H.B. Talbot & V.C. Green) Jülich

Photos of fresh collections in PlutoF: link 1, link 2

Figs 8; 10l, m

Notes. Unfortunately, we were not able to study the type specimen of Peniophora longispora var. brachyspora (Talbot’s No. 40683 in PREM, Mycology Division, ARC-Plant Protection Research Institute, Queensland, South Africa). However, reviewing taxonomic literature on Subulicystidium brought us to delineating two morphogroups in the species S. brachysporum, according to the views of the earlier authors. Talbot (1958), while describing Peniophora longispora var. brachyspora, characterised its basidiospores as “elliptic-fusoid, 6.4–8 × 2.2–3.2 µm … sometimes with a faint band about the middle”. Boidin and Gilles (1988) described the basidiospores of S. brachysporum from Réunion as elliptic in frontal face and bananiform (cylindric with slightly attenuated apex, slightly curved) in lateral face, 7.5–10 × 2–2.5(–3) µm. Therefore, we differentiated groups of (i) S. brachysporum sensu Talbot, i.e. sensu typi, with straight oblong-elliptic basidiospores having long attenuated base, with the mean length below 7.5 µm and mean length to width ratio hardly reaching 3 (see Fig. 10m); and (ii) S. brachysporum sensu Boidin and Gilles with cylindric and slightly curved basidiospores with the mean length over 7.5 µm and length to width ratio between 3 and 4 (Fig. 10l). In our dataset, S. brachysporum sensu Boidin and Gilles was represented by far more specimens than S. brachysporum sensu Talbot.

The description and illustration of S. brachysporum by Duhem and Michel (2001) (specimen Bourdot No. 7986) and by Oberwinkler (1977, fig. 25, “authentical material”) correspond to the protologue and drawing of S. brachysporum by Talbot (1958). On the other hand, our examination of Venezuelan collection (FO15970 in TUB, see Oberwinkler 1977, fig. 29) let us assign it to S. brachysporum sensu Boidin and Gilles (1988). Descriptions by Liberta (1980), Maekawa (1998), Martini (2016) and illustrations by Eriksson et al. (1984, fig. 764 f, g – for single specimens from Canada and USA), match S. brachysporum sensu Boidin and Gilles (1988) as well.

DNA sequence similarity analyses and phylogenetic reconstructions did not support the presence of two morphogroups of S. brachysporum. Sequences of both morphotypes occured in the same clade and, moreover, shared the clade with sequences from the morphospecies S. meridense. For the moment, we prefer to retain two morphogroups of S. brachysporum, thus promoting further exploration of species limits and examination of the type specimen.

The most comprehensive overview of global occurrence of S. brachysporum was provided by Martini (2016). We can add that the species is present also in South America, Caribbean region and Madagascar.

Specimens examined: Subulicystidium brachysporum sensu Talbot (1958). ARGENTINA. Misiones: Iguazu National Park, Cataratas de Iguazu, -25.6748, -54.4532, on dead wood of angiosperm tree, 1-5 Mar 1982, L.Ryvarden (LR 19687 in O:F 506779). BRAZIL. Rondonia: Porto Velho, Rua Rio Madeira 7014, Nova Esperanca, -8.7160, -63.8785, on dead wood of angiosperm tree, 11 Mar 2012, K.-H.Larsson (KHL 15318 and 15327 in O:F). Sao Paulo: Santos, Ubatuba, Ilha Anchieta, -23.5500, -45.0667, on dead wood, 17-18 Jan 1987, D.Pegler, K.Hjortstam & L.Ryvarden (LR 24170 and LR 24203 in O:F). COLOMBIA. Magdalena: Parque Nacional Tayrona, Estacion de Gairaca, 0-30 m, 11.3170, -74.1063, on dead wood, 12 Jun 1978, L.Ryvarden (LR 15755 in O:F 918492). RÉUNION. Saint-Paul: Saint-Paul, St-Gilles-II, Ravine de St. Gilles, abandoned orchard, 200 m, on dead wood, 26 Apr 1985, J.Boidin (LY 11378).

Specimens examined: Subulicystidium brachysporum sensu Boidin and Gilles (1988). ARGENTINA. Misiones: Iguazu National Park, Cataratas de Iguazu, -25.6748, -54.4532, on dead wood of angiosperm tree, 1-5 Mar 1982, L.Ryvarden (LR 19533 in O:F 506782). BRAZIL. Pará: Belem, Museo Goeldii Scientific Centre, -1.4525, -48.4764, on dead wood, 25 Nov 2013, K.-H.Larsson (KHL 16461 in O:F). Paraiba: Areia, Reserva Estadual Mata do Pau-Fero, -6.9642, -35.7496, on dead wood of angiosperm tree, 28 Apr 2013, K.-H.Larsson (KHL 16100 in O:F). Rondonia: Porto Velho, Rua Rio Madeira 7014, Nova Esperanca, -8.7160, -63.8785, on dead wood of angiosperm tree, 11 Mar 2012, K.-H.Larsson (KHL 15330 in O:F); Rubber plantation park, -8.7324, -63.9008, on dead wood of angiosperm tree, 11 Mar 2012, K.-H.Larsson (KHL 15352 in O:F). Sao Paulo: Sao Paulo, Instituto de Botanica, -23.6450, -46.6261, on dead wood, 24 Jan 1987, K.Hjortstam (Hjm 16573 in GB). Colombia. Magdalena: Parque Nacional Tayrona, Estacion de Gairaca, 0-30 m, 11.3170, -74.1063, on dead wood, 12 Jun 1978, L.Ryvarden (LR 15784 in O:F 918490; LR 15744B in O:F 918491; LR 15784 in O:F 918493). COSTA RICA. Alajuela: Bijagua, Albergue Heliconias, Sendero Heliconias, 770 m a.s.l, 10.7181, -85.0453, on branch of angiosperm tree, 12 Jul 2001, K.-H.Larsson (KHL 11216 in GB); La Fortuna de San Carlos, Parque Nacional Volcán Arenal, Sendero Pilón, 650 m, 10.4589, -84.7644, on dead wood of angiosperm tree, 15 Jul 2001, K.-H.Larsson (KHL 11419 in GB). Puntarenas: Coto Brus, Sabalito, Zona Protectora Las Tablas, La Neblina, 8.9149, -82.7719, on stem of angiosperm tree, 5 Nov 2004, K.-H.Larsson (KHL 12764 in GB). DOMINICAN REPUBLIC. Provincia La Altagracia: Parque Nacional del Este, Playa Guaraguao, 18.3269, -68.8092, on dead wood, 3 Jun 1997, K.-H.Larsson (KHL 9920 in GB). ETHIOPIA. Oromia: West Shewa Zone, Ginchu, Chilomo Forest, 9.0900, 38.1700, on dead wood, 10 Jul 1990, L.Ryvarden (LR 28047 in O:F 909592). JAMAICA. Cornwall County: Trelawny parish, Cockpit Country, Ramgoat Cave, 350 m a.s.l., 18.3378, -77.5568, on twigs of angiosperm tree, 11 Jun 1999, K.-H.Larsson (KHL 10686 in GB); N of Crowlands, trail/road into park area, 18.2611, -77.6511, on branch of angiosperm tree, 10 Jun 1999, K.-H.Larsson (KHL 10633 in GB); Windsor Cave, along trail to Troy, 18.3564, -77.6472, on branch of angiosperm tree, 13 Jun 1999, K.-H.Larsson (KHL 10763 in GB). Middlesex County: Manchester parish, Marshalls Pen, Sutton Farm, 18.0589, -77.5308, on branches of angiosperm tree, 8 Jun 1999, K.-H.Larsson (KHL 10505 in GB), on stem of angiosperm tree, 8 Jun 1999, K.-H.Larsson (KHL 10566 in GB). Surrey County: Portland parish, between reach and Ecclesdown hillside to the east, 500 m, 18.0433, -76.3108, on branch of angiosperm tree, 16 Jun 1999, K.-H.Larsson (KHL 10855 in GB). MADAGASCAR. Anosy: Tolagnaro, Mandena Conservation Zone, -24.9529, 47.0028, on dead wood of angiosperm trunk, 10 Mar 2010, K.-H.Larsson (KHL 14359 and 14365 in O:F); Nahampoana, -24.9667, 46.9667, on dead wood of angiosperm trunk, 12 Mar 2010, K.-H.Larsson (KHL 14390 in O:F); Petriky Conservation Zone, -25.0667, 46.8500, on dead wood of angiosperm trunk, 14 Mar 2010, K.-H.Larsson (KHL 14486, 14505 and 14516 in O:F); Sainte Luce Conservation Zone, -24.8028, 47.1636, on dead wood of angiosperm trunk, 15 Mar 2010, K.-H.Larsson (KHL 14537 in O:F). Ihorombe: Isalo National Park, Namaza, along track from entrance to waterfall, -22.5583, 45.4000, on dead wood of angiosperm trunk, 7 Mar 2010, K.-H.Larsson (KHL 14293 in O:F). PAPUA NEW GUINEA. Morobe: Bawituc, Lae 12-Mile, Uphill of Bawituc, -6.6390, 146.9089, on dead twig, 3 Jul 1905, L.Tedersoo (TU 110416). PUERTO RICO. Municipio Isabela, Moñtanas Aymamón, Bosque Estatal de Guajataca, Verada Nueva Trail, 230 m, 18.4242, -66.9678, on dead fruitbodies of polypore, 26 Jun 1996, K.-H.Larsson (KHL 9544 in GB). Municipio Luquillo, Luquillo Mts, Bisley Experimental Watersheds, along the logging road, 215 m, 18.3161, -65.7467, on log, 6 Jun 1998, K.-H.Larsson (KHL 10270 in GB); Sabana, above Chicken Farm & Rio Sabana, 70 m, 18.3500, -65.7344, 10 Jun 1998, on woody branch, K.-H.Larsson (KHL 10406 in GB) and on wet log, K.-H.Larsson (KHL 10411 in GB), on wood of angiosperm tree, 7 Jun 1997, K.-H.Larsson (KHL 10088, 10095 and 10097 in GB). RÉUNION. Saint-Benoît: La Plaine-des-Palmistes, Palmistes III-87, en descendant vers St Benoit, alt 800 m, on dead wood, 23 Mar 1987, G.Gilles (LY 12456); Salazie, Hell Bourg, ca 1000 m, -21.0642, 55.5269, on dead branch, 23 Mar 2015, M.Striegel (L 1584b in FR and KAS). Saint Pierre: Cilaos, Cirque de Cilaos, Roche Merveilleux, Sentiere botanique, 1300 m, -21.1232, 55.4920, on strongly decayed wood, 15 Mar 2013, E.Langer (L 0134 in FR and KAS); L’Étang-Salé, Étang-Salé-87: forêt domaniale, alt 60-80 m, on dead wood of Fabaceae tree, 14 Mar 1987, G.Gilles (LY 12293); Saint-Philippe, Baril-II-87, depart du sentier botanique, 100-150 m, on dead branch, 6 Apr 1987, G.Gilles (LY 12772), Forêt de Mare Longue, 495 m, -21.3438, 55.7410, on dead branch, 28 Mar 2015, M.Striegel (L 1795 in FR and KAS); Saint-Pierre, Piton de Mont Vert, ca 560 m, -21.3279, 55.5413, on dead wood, 18 Mar 2015, M.Striegel (L 1498 in FR and KAS), Piton de Mont Vert, hiking path, ca 560 m, -21.3279, 55.5413, on dead branch, 18 Mar 2015, J.Riebesehl & M.Schroth (L 1147 in FR and KAS). VENEZUELA. Estado Amazonas: Manapiare, Yutajé, 110 m, 5.6142, -66.1236, on dead wood of angiosperm tree, 12-19 Jun 1997, L.Ryvarden (LR 40650A in O:F 909582). Estado Merida: Merida, 8.6249, -71.1395, on dead wood, 1969, F.Oberwinkler (FO 15970 in TUB).

Subulicystidium longisporum (Pat.) Parmasto

Note. The following specimen was used to illustrate the species on Fig. 10g:

UKRAINE. Zakarpatska: Carpathian Biosphere Reserve, vicinities of Mala Uholka village, 670 m, 48.2632, 23.6175, on decayed deciduous wood, 11 Sep 2013, A.Ordynets (CWU 6737).

Subulicystidium meridense Oberw.

Figs 9; 10n–p

Notes. According to Oberwinkler (1977), the basidiospore size of the type collection (FO13761 in TUB) is 6–8 × 2.5–3 µm, suggesting the mean basidiospore length, width and their ratio are 6.5 µm, 2.75 µm and 2.36, respectively. Our re-measuring revealed longer basidiospores, namely 6.6–8.1 × 2.4–2.8 µm, with the length to width ratio 2.6–3.2 (2.9) (see Supplementary files 2-4).

When describing S. meridense, Oberwinkler (1977) stressed the importance of allantoid, i.e. clearly curved, basidiospores. We adhered to this concept assigning our collections to S. meridense (see Figs 10 n, o). We named those with similar spore size but with straight cylindric spores “Subulicystidium aff. meridense” (Fig. 10p). However, molecular level sequence similarity analyses and phylogenetic reconstructions did not support the presence of two distinct groups.

In addition to the South American and Reunionese specimens, we examined also specimens of S. meridense from India and Taiwan. Boidin and Gilles (1988) reported LY 12456 and LY 12772 from Réunion as S. meridense with the note that the spores are larger and more elongated than in the Venezuelan type material. Our morphological examination (and DNA sequence data for LY 12772) showed that the Reunionese collection represent S. brachysporum sensu Boidin and Gilles. However, we concur with Boidin and Gilles that LY 12816 (sequenced) from Réunion and LY 9144 from Gabon should be named S. meridense. The description of the specimen TMI 25520 from Vanuatu (Maekawa 2002) corresponds to our concept of Subulicystidium aff. meridense.

Specimens examined: Subulicystidium aff. meridense. ARGENTINA. Misiones: Iguazu National Park, Cataratas de Iguazu, -25.6748, -54.4532, on wood of angiosperm tree, 1-5 Mar 1982, L.Ryvarden (LR 19581 in O:F). BRAZIL. Rondonia: Porto Velho, Rubber plantation park, -8.7324, -63.9008, on dead wood of angiosperm tree, 11 Mar 2012, K.-H.Larsson (KHL 15325 in O:F). Sao Paulo: Santos, Ubatuba, Ilha Anchieta, -23.5500, -45.0667, on dead wood, 17-18 Jan 1987, D.Pegler, K.Hjortstam & L.Ryvarden (LR 24201 in O:F). COLOMBIA. Magdalena: Parque Nacional Tayrona, Estacion de Gairaca, 0-30 m, 11.3170, -74.1063, on dead wood, 12 Jun 1978, L.Ryvarden (LR 15812 in O:F 918846). PUERTO RICO: Municipio de Cerro Alto, Montanas Aymamon, limestone magote near Parador Guajataca, 60 m, 18.4828, -66.9583, on strongly decayed log of angiosperm tree, 27 Jun 1996, K.-H.Larsson (KHL 9561 in GB). Municipio Luquillo, Luquillo Mts, Sabana, above Chicken Farm & Rio Sabana, 70 m, 18.3500, -65.7344, on log, 10 Jun 1998, K.-H.Larsson (KHL 10397 in GB).

Specimens examined: Subulicystidium meridense. ARGENTINA. Misiones: Iguazu National Park, Cataratas de Iguazu, -25.6748, -54.4532, dead wood of angiosperm tree, 1-5 Mar 1982, L.Ryvarden (LR 19688 in O:F 506784). BRAZIL. Rondonia: Porto Velho, Rua Rio Madeira 7014, Nova Esperanca, -8.7160, -63.8785, on dead wood of angiosperm tree, 11 Mar 2012, K.-H.Larsson (KHL 15322 in O:F). Sao Paulo: Santos, Ubatuba, Ilha Anchieta, -23.5500, -45.0667, on decayed wood, 17-18 Jan 1987, D.Pegler, K.Hjortstam & L.Ryvarden (Hjm 16400 in GB). CENTRAL AFRICAN REPUBLIC. Lobaye: Nola, Boukoko, 3.8929, 17.9153, on dead tree trunk, 18 May 1965, J.Boidin (in LY 5476). COSTA RICA. Guanacaste: Area cons. Tempisque, Reserva Biologica Lomas Barbudal, near the entrance, 20-40 m, 10.5103, -85.3744, on dead wood of angiosperm tree, 14 Jul 2001, K.-H.Larsson (KHL 11355, 11365 and 11368 in GB). Puntarenas: Coto Brus, Sabalito, Zona Protectora Las Tablas, Progreso, Camino a Cotoncito, 1560 m, 8.9306, -82.8031, on wood of angiosperm tree, 3 Nov 2004, K.-H.Larsson (KHL 12557 in GB), La Neblina, 1350 m, 8.9149, -82.7719, on wood of angiosperm tree, 5 Nov 2004, K.-H.Larsson (KHL 12732 in GB). San José: Dota, San Gerardo, around Hotel Savegre, ca 2000 m, 9.5643, -83.8016, on branch of angiosperm tree, 9 Nov 2004, K.-H.Larsson (KHL 12969 in GB). GABON. Estuaire: Libreville, Bush littoral, km 13 N Libreville, 0.5338, 9.4673, on bark of dead wood, 2 Feb 1979, G.Gilles (LY 9144). INDIA. Darjeeling: Sukna, About 4 km from Sukna towards RongTong, 26.8246, 88.3625, on bark of dead/decaying branch of angiosperm, 9 Aug 1980, G.S.Dhingra (19201 in O:F 909586). RÉUNION. Saint-Benoit: Saint-Benoit, Route forestière 3 de Takamaka, -21.1038, 55.5724, on dead wood, 8 Apr 1987, G.Gilles (LY 12816). TAIWAN. Nantou: Huisun Recreation Area, path to the “stone frog”, 24.0912, 121.0337, on dead wood, 26 Apr 1996, G.Langer, E.Langer & C.-J.Chen (GEL 3520 and 3530 in KAS). VENEZUELA. Estado Aragua: Maracay, National Park Henri Pittier, Rancho Grande, 10.3800, -67.6190, on hardwood, 22 Jun 1995, L.Ryvarden (LR 35544/C in O:F). Estado Merida: Merida, Vicinities of Instituto Forestal Latino-Americano, 8.6249, -71.1395, on dead twigs, 27 Nov 1968, F.Oberwinkler (FO 13761 in TUB, holotype).

Subulicystidium naviculatum Oberw.

Figs 3a, b; 10a

Notes. We examined a single collection from Costa Rica (KHL 11566 in GB) which had broad fusiform basidiospores (8.6–)8.8–11.2(–11.6) × (4.0–)4.3–5.0(–5.3) µm, i.e. slightly shorter than in the holotype specimen FO 12778 (TUB) from Venezuela: 10–12 × 4.5–5 µm (Oberwinkler 1977). Furthermore, the cystidia in our specimen were covered with rows of rectangular crystals while the ornamentation pattern of cystidia in the holotype resembled that of S. harpagum, i.e. short rod-like protrusions that project backwards under an acute angle, giving cystidia the resemblance of a harpoon (see fig. 32 in Oberwinkler 1977). Unfortunately, the holotype in TUB could not be located.

Kisimova-Horovitz et al. (1997) reported a collection from Costa Rica (207a-I, =FO 42968 in TUB) as S. naviculatum. However, we re-identified the collection as S. robustius.

Specimens examined. COSTA RICA. San José: Reserva Los Santos, Cerro de la Muerte, 1.5 km from Interamerican Highway along road to San Gerardo de Dota, 2850 m, 9.5964, -83.7986, on stem of angiosperm tree, 18 Jul 2001, K.-H.Larsson (KHL 11566 in GB).

Subulicystidium nikau (G. Cunn.) Jülich

Figs 5c,d; 10u

Notes. The species was described by Cunningham (1955) as Peniophora sororia based on material from the midribs of the dead leaves of nikau palm (Rhopalostylis sapida) which is endemic to New Zealand. Later Cunningham (1963) noticed that the name was occupied (Bourdot and Galzin 1912; p. 386) and provided the legitimate name Peniophora nikau.

The holotype of S. nikau (PDD 13816) has basidiospores 7–9 × 4–5 µm as reported by (Cunningham 1955). We measured the holotype basidiospores as slightly narrower, viz. (6.8–)6.9–8.6(–9.0) × 3.3–4.1(–4.4) µm (see Supplementary files 2–4), confirming the measurements by Oberwinkler (1977) and Liberta (1980).

As the basidiospores are similar, S. nikau has been confused with S. oberwinkleri. The former has cystidia with regular ornamentation (rows of rectangular crystals) typical for the genus Subulicystidium and the generitype S. longisporum. In contrast, S. oberwinkleri has larger cystidia with large, irregularly shaped crystalline plates. Cunningham (1955) illustrated cystidia of S. nikau correctly, while the characterisation by Stalpers and Buchanan (1991) as “covered with plate-like crystals” is misleading. Oberwinkler (1977) and Maekawa (1998) realised the discrepancies in cystidial ornamentation but did not provide a solution. We re-identified the record of S. nikau from Reunion, LY12488 (Boidin and Gilles 1988) as S. oberwinkleri. We also collected and sequenced a Reunionese specimen with regularly ornamented cystidia (KAS: L1296). Though sampled on dead wood and far from locus classicus, we keep it under the name S. nikau for the time being. Réunion is thus the second known locality of the species after New Zealand. The record from Venezuela reported by Liberta (1980) has not been studied and is hard to interpret because no illustration of cystidia was provided.

Specimens examined. RÉUNION. Saint-Pierre: Saint-Philippe, Sentier de Takamaka, ca 840 m, -21.0913, 55.6199, on dead wood, 26 Mar 2015, J.Riebesehl & M.Schroth (L 1296 in FR and KAS). NEW ZEALAND. Auckland: Cascades, Waitakere Ranges, on dead leaf midribs of palm Rhopalostylis sapida, 3 Apr 1954, S.D.Baker (PDD 13816, holotype).

Subulicystidium obtusisporum Duhem & H. Michel

Figs 6d, e; 10h

Notes.Duhem and Michel (2001) identified a collection from Venezuela (Oberwinkler 1977, fig. 29, FO15970 in TUB) as S. obtusisporum, while we regard the same specimen as S. brachysporum sensu Boidin and Gilles (1988). Maekawa (2002) reported S. obtusisporum from Vanuatu and Ghobad-Nejhad et al. (2009) from Russian Caucasus. Here we report the species from East Asia and Caribbean region.

The first sequenced material of S. obtusisporum is our collection from Frankfurt am Main, central Germany (FR:W213-3-I). Fruit-body morphology, as well as microhabitat (exposed dead wood) agree with the data for the type specimen and related collections from southern France (Duhem and Michel 2001). Another sequenced specimen GB:KHL 10622 from Jamaica was very distant from the German specimen in terms of ITS and 28S sequence identity and position on the phylogenetic tree, which means S. obtusisporum is polyphyletic. Sequencing additional specimens, not least from Asia, is needed to clarify the taxonomy of this morphospecies.

Specimens examined. CHINA. Jilin: Chang Bai Shan Forest Reserve, Hangcong hou, 750 m, on dead wood of Acer sp., 11-17 Sept 1983, L.Ryvarden (LR 21774 in O:F 909590). COSTA RICA. Guanacaste: Area cons. Tempisque, Reserva Biologica Lomas Barbudal, near the entrance, 20-40 m, 10.5103, -85.3744, on deadwood of angiosperm tree, 14 Jul 2001, K.-H.Larsson (KHL 11373 in GB). GERMANY. Hesse: Frankfurt, Science Park at the Campus Riedberg of Frankfurt University, 50.1701, 8.6300, on decayed trunk, 31 Mar 2016, O.Koukol (W213-3-I in FR). ITALY. Italy. Latina: Circeo Natural park, Selva de Circeo, 41.3429, 13.0534, on wood of Quercus sp., 22-25 Oct 1984, K.Hjortstam, K.-H.Larsson & L.Ryvarden (LR 22458 in O:F 505520). JAMAICA. Cornwall County: Trelawny parish, Crown Lands, trail/road into park area, 18.2611, -77.6511, on branches of angiosperm tree, 10 Jun 1999, K.-H.Larsson (KHL 10622 in GB). PUERTO RICO. Municipio Luquillo, Luquillo Mts, Bisley Experimental Watersheds, along track from parking place, 215 m, 18.3161, -65.7467, on branch of angiosperm tree, 6 Jun 1997, K.-H.Larsson (KHL 9955 in GB). TAIWAN. Chiayi: Shi Ding, road No. 18 in direction to Alishan at km 60, ca 1500 m alt, 23.4801, 120.4491, on dead wood, 1 May 1996, G.Langer, E.Langer & C.-J.Chen (GEL 3677 in KAS). Miaoli: Sheipa National Park, Kuanwu, forest ca 500 m in direction of Le Shan, trail on the left side of the road, ca 2100 m, 24.3624, 121.1252, on dead wood, 19 Apr 1996, G.Langer, E.Langer & C.-J.Chen (GEL 3409 in KAS).

Subulicystidium perlongisporum Boidin & Gilles

Notes. The following specimen was used to illustrate the species on Fig. 10j: GERMANY. Hesse: Biedenkopf, little valley of Martinsbach creek, mixed forest 280 m, 50.5410, 8.3037, on dead wood, 23 Apr 2016, A.Ordynets & M.Theiss (Ordynets 00158 in KAS).

In general, spore size and shape are of crucial importance for the taxonomy of fungi (Parmasto et al. 1987). In our study, however, we found that basidiospore morphology itself may be insufficient for species-rank identifications in Subulicystidium. In this regard, the usefulness of other morphological characters is worth discussing.

Jülich (1975) studied cystidia of S. brachysporum and S. longisporum under a scanning electron microscope and concluded the identity of their ornamentation pattern. Jülich (1975) hypothesised, after observing cystidia of S. nikau under a light microscope, that this ornamentation pattern was universal at the genus level. Regarding the shape of cystidia, Jülich (1975) noticed the less prominent basal swelling in S. brachysporum compared to S. longisporum.Oberwinkler (1977) characterised cystidia of Subulicystidium as uniform, but his remarks and especially illustrations displayed several deviations from the common pattern regarding both size and ornamentation. We further developed the idea of the importance of cystidial morphology and showed the presence of interspecific size differences as well as species-specific types of cystidial ornamentation (S. oberwinkleri, S. harpagum, S. robustius, S. rarocrystallinum). The important finding of Jülich (1975) is that the shape of single crystals can vary within a collection. The sharpness of the crystals is reduced with age, resulting in rounded instead of rectangular crystals as observed in a light microscope.

Beside cystidia, also hyphae and hymenial elements can have encrustation. Oberwinkler (1977) illustrated crystalline collars on the bases of basidia in the holotype of S. meridense and in the specimen of S. brachysporum sensu Boidin and Gilles (figs 30 and 29. respectively in Oberwinkler 1977) but not in other species. Kisimova-Horovitz et al. (1997) noticed a nearly ubiquitous presence of hymenial encrustation in Subulicystidium. We share this opinion after examining our collections. Jülich (1968) and Liberta (1980) observed repetobasidia and considered them to be a criterion of the genus Subulicystidium. On the contrary, Eriksson et al. (1984) did not observe any repetobasidia in the North European collections of S. longisporum. In our large set of tropical specimens we did not find any repetobasidia. Thus we suppose that repetobasidia observed by Jülich and Liberta are simply basidia with a well-developed crystalline collar.

Oberwinkler (1977) illustrated slightly thick-walled subicular hyphae in S. meridense but thin-walled in the rest of the species. We confirm this pattern for S. meridense but also observed similar deviating subicular hyphae in several other species, viz. S. brachysporum, S. robustius, S. ryvardenii and S. oberwinkleri. However, in all these cases, we found that the hyphal surface is rough and the wall is highly light-refractive, which means they are covered by crystalline material. In S. harpagum, this crystalline sheath around hyphae can reach a thickness of one and in S. oberwinkleri several micrometres. Under a light microscope, it is not possible to decide to what extent thickness depends on the cell wall or is due to the deposition of crystalline material.

Bourdot and Galzin (1912) reported a series of collections deviating from typical S. longisporum in fruit-body thickness and colouration and cystidial uniformity. The several varieties proposed by the French authors were not accepted by Jülich (1975) who considered them to represent normal variation and different developmental stages of S. longisporum. In line with Jülich (1975), after examining our specimens, we conclude that fruit-body thickness and density is variable within the species and thus of little value for the species-level identifications. Nevertheless, an experienced eye can differentiate the hirsute hymenial surface of the species with more robust cystidia (S. robustius, S. ryvardenii) from the velutinous hymenial surface of the rest of the genus. In a single species, S. robustius, we consistently observed slightly yellowish fruiting bodies, which was due to a yellowish hue of the hyphal and cystidial walls.

Partitioning sequence dissimilarity of both ITS and 28S into interspecific and intraspecific components revealed a clear barcode gap for some of the species but problems to delimit others. Therefore, both cases when morphology and available molecular information are congruent and cases when they are in conflict were found. This points to the importance of careful morphological examination and the need to combine morphology, rDNA barcode data and other DNA markers when defining species in Subulicystidium.