Reclassification of Pterulaceae Corner (Basidiomycota: Agaricales) introducing the ant-associated genus Myrmecopterula gen. nov., Phaeopterula Henn. and the corticioid Radulomycetaceae fam. nov.

Pterulaceae was formally proposed to group six coralloid and dimitic genera: Actiniceps (=Dimorphocystis), Allantula, Deflexula, Parapterulicium, Pterula, and Pterulicium. Recent molecular studies have shown that some of the characters currently used in Pterulaceae do not distinguish the genera. Actiniceps and Parapterulicium have been removed, and a few other resupinate genera were added to the family. However, none of these studies intended to investigate the relationship between Pterulaceae genera. In this study, we generated 278 sequences from both newly collected and fungarium samples. Phylogenetic analyses supported with morphological data allowed a reclassification of Pterulaceae where we propose the introduction of Myrmecopterula gen. nov. and Radulomycetaceae fam. nov., the reintroduction of Phaeopterula, the synonymisation of Deflexula in Pterulicium, and 53 new combinations. Pterula is rendered polyphyletic requiring a reclassification; thus, it is split into Pterula, Myrmecopterula gen. nov., Pterulicium and Phaeopterula. Deflexula is recovered as paraphyletic alongside several Pterula species and Pterulicium, and is sunk into the latter genus. Phaeopterula is reintroduced to accommodate species with darker basidiomes. The neotropical Myrmecopterula gen. nov. forms a distinct clade adjacent to Pterula, and most members of this clade are associated with active or inactive attine ant nests. The resupinate genera Coronicium and Merulicium are recovered in a strongly supported clade close to Pterulicium. The other resupinate genera previously included in Pterulaceae, and which form basidiomes lacking cystidia and with monomitic hyphal structure (Radulomyces, Radulotubus and Aphanobasidium), are reclassified into Radulomycetaceae fam. nov. Allantula is still an enigmatic piece in this puzzle known only from the type specimen that requires molecular investigation. A key for the genera of Pterulaceae and Radulomycetaceae fam. nov. is also provided here.


INTRODUCTION
The history of Pterulaceae begins with the hesitant proposal of the genus Pterula (hereinafter abbreviated as Pt.) in the early 19th century by Fries (1821Fries ( , 1825Fries ( , 1830. The typification of this genus was addressed by Lloyd (1919) and this was followed by discussion between Doty (1948), Donk (1949), and Rogers (1949Rogers ( , 1950. Ultimately Corner (1952c) provided a thorough discussion of the timeline of Fries' decisions, which was later confirmed with further clarification by (Donk 1954;Donk 1963).
The ecological roles of Pterulaceae are not well understood, most being classified from superficial observations as saprotrophs, growing on wood or leaf litter, with wood decay potentially being the ancestral state. Whilst many species are found inhabiting soil or litter, two species are reported to associate with living plants, namely Pterula cf. tenuissima, endophytic in asymptomatic leaves of Magnolia grandiflora, and Pterulicium xylogenum, causal agent of culm rot disease of bamboo (Munkacsi et al. 2004;Villesen et al. 2004;Harsh et al. 2005) and possibly also a pathogen of sugarcane (Corner, 1952b).
Whilst these earlier phylogenetic studies did not focus on resolving evolutionary relationships of the genera, they did demonstrate that the coralloid genera of Pterulaceae are clearly polyphyletic. Amongst the morphological characters previously used to separate the genera, but now known to be phylogenetically unreliable, is the orientation of basidiome growth that differentiates Pterula from Deflexula and the presence of a corticioid patch at the base of the basidiome in Pterulicium (Corner 1950(Corner , 1952a(Corner , 1970. Therefore, the reclassification of Pterulaceae is required to restore the monophyly of the genera. We aimed to clarify the phylogenetic relationships of the various genera within Pterulaceae through collection of new samples during fieldwork campaigns in Brazil and additionally sampling of fungarium specimens. This has yielded sequence data from many specimens not included in previous phylogenetic analyses, permitting a comprehensive reappraisal of the phylogeny of Pterulaceae. Here we present a proposal for a new classification based on the phylogeny inferred from three nuclear loci (nrITS, nrLSU and RPB2), including representatives of all genera currently accepted in Pterulaceae except Allantula. Despite several attempts for recollecting Allantula in its type locality, the monotypic genus is still only known from the type specimen collected by Corner (1952a).

Collections and morphological observations
Several field campaigns between 2011 and 2017 have obtained new specimens from > 15 locations in nine states across Brazil (Amazonas, Espírito Santo, Minas Gerais, Pará, Paraíba, Paraná, Rio de Janeiro, Rio Grande do Sul and Santa Catarina). The samples were dried in a low-heat food dehydrator and deposited at Aberystwyth University (ABS), Instituto Nacional de Pesquisas da Amazônia (INPA), Jardim Botânico do Rio de Janeiro (RB), Royal Botanic Gardens -Kew (K), Universidade Federal do Oeste do Pará (HSTM) and Universidade Federal de Santa Catarina (FLOR). Morphological identification and taxonomy of Pterulaceae are treated sensu Corner. Microscopic observations followed the methods described in Leal-Dutra (2015) and Leal-Dutra et al. (2018).
DNA extraction, amplification, cloning and sequencing DNA was extracted from dried basidiomes or freeze-dried cultures by first grinding with liquid nitrogen and then lysis in CTAB buffer (100 mM Tris-HCl pH 8.0, 1.4 M NaCl, 20 mM EDTA, 2% CTAB), clean-up with chloroform:isoamyl alcohol (24:1), precipitation with isopropanol (0.6 vol.) and a final wash with 70% ethanol. Partial sequences of the nrITS, nrLSU and RPB2 were amplified by PCR using the primer pairs listed on Table 1 and following the cycling conditions in the original publications. PCR products were purified using 2 U of Exonuclease I (Thermo Fisher Scientific) and 0.2 U FastAP Thermosensitive Alkaline Phosphatase (Thermo Fisher Scientific) per 1 μl of PCR product, incubated at 37°C for 15 min, followed by heat inactivation at 85°C for 15 min. The samples were then sent for Sanger sequencing at the IBERS Translational Genomics Facility (Aberystwyth University) or Jodrell Laboratory (Royal Botanic Gardens, Kew). The same PCR primers were used for sequencing; additional primers were used to sequence the nrLSU and RPB2 (Table 1).
Chromatograms were manually checked and sequences assembled and edited using GENEIOUS 10.0.2 (Kearse et al. 2012). Samples presenting indels were cloned using pGEM-T Easy Vector Systems (Promega) into Subcloning Efficiency DH5α Competent Cells (Invitrogen). Up to five clones from each sample were amplified and sequenced as above. For each sample clone sequences were aligned to generate one or more consensus sequences and polymorphisms were replaced by respective IUPAC code for ambiguous nucleotide; in cases where indels were found, two different sequences were saved (Leal-Dutra et al. 2018).
Moreover, 27 sequences of nrITS (4), nrLSU (10) and RPB2 (13) were mined from 13 previously assembled and unpublished genomes using NCBI BLAST+ package v2.7.1 (Camacho et al. 2009). Two sequences of each Pterulaceae genus were used as query and the best hit based on the combination of e-value and bit score was selected; the same hit should usually appear for all query sequences. In one case (sample KM190547), more than one optimal hit was found; the subject sequences were compared for occurrence of indels and treated as virtual clones (VC). These sequences are included in the dataset ( Table 2). The sequences generated in this study have been submitted to GenBank (Table 2).

Phylogenetic analyses
A preliminary maximum-likelihood (ML) analysis was conducted with the sequences generated in this study alongside GenBank sequences to find the best outgroup for Pterulaceae based on previous studies (Dentinger et al. 2016;Zhao et al. 2016;Matheny et al. 2006;Larsson 2007) and to assess the similarities between the cloned sequences (Additional file 1; Additional file 2).
A reduced version of the previous dataset with only one sequence from each cloned sample was created. After removing near-identical sequences with no phylogenetic resolution, the final dataset comprised 119 sequences, including 32 sequences from GenBank and four sequences of Stephanospora as outgroups, and was divided into five partitions for further analyses: ITS1, 5.8S, ITS2, LSU and RPB2. Each partition was aligned separately with MAFFT v7.311 (Katoh and Standley 2013) using the E-INS-i algorithm for ITS1 and ITS2, and L-INS-i for 5.8S, LSU and RPB2. The alignments were examined and corrected manually in AliView v1.5 (Larsson 2014) and trimmed to remove uneven ends. Following the simple indel coding (Simmons and Ochoterena 2000), a morphological matrix were constructed using Seq-State (Müller 2005) where indels were coded as binary characters. The nucleotide alignments were then trimmed with trimAl v1.4.rev22 (Capella-Gutiérrez et al. 2009) with the option -gappyout to remove unaligned regions.
Maximum-likelihood tree reconstruction was performed with IQ-TREE v1.6.7.1 (Nguyen et al. 2015). The best-fit evolutionary models and partitioning scheme for this analysis were estimated by the built-in ModelFinder (option -m MF + MERGE) allowing the partitions to share the same set of branch lengths but with their own evolution rate (−spp option) (Chernomor et al. 2016;Kalyaanamoorthy et al. 2017). Branch support was assessed with 1000 replicates of ultrafast bootstrapping (UFBoot) (Hoang et al. 2018) and allowing resampling partitions and then sites within these partitions to reduce the likelihood of false positives on branch support (option -bspec GENESITE).
Bayesian Inference (BI) was implemented using MRBAYES v3.2 (Ronquist et al. 2012) with two independent runs, each one with four chains and starting from random trees. The best-fit evolutionary models and partitioning scheme for these analyses were estimated as for the ML analysis but restricting the search to models implemented on MRBAYES (options -m TESTMERGEONLY -mset mrbayes). Chains were run for 10 7 generations with tree sampling every 1000 generations. The burn-in was set to 25% and the remaining trees were used to calculate a 50% majority consensus tree and Bayesian Posterior Probability (BPP). The convergence of the runs was assessed on TRACER v1.7 (Rambaut et al. 2018) to ensure the potential scale reduction factors (PSRF) neared 1.0 and the effective sample size values (ESS) were sufficiently large (> 200). Nodes with BPP ≥0.95 and/or UFBoot ≥95 were considered strongly supported. Alignment and phylogenetic trees are deposited in Treebase (ID: 24428).

RESULTS
From this section, all taxa are referred to by the names proposed in this study.

Field data
Fieldwork resulted in the discovery of approximately 100 new specimens, now placed within Pterulaceae (Table 2). Axenic culture isolation was also possible from several of these specimens.

Phylogenetic analyses
A total of 278 sequences from 123 samples were generated in this study: 153 nrITS, 74 nrLSU and 51 RPB2; 61 from cloning and 40 from genome mining. The final alignment  Table 2 Details of new sequences generated in this study used in the tree of       consisted of 113 sequences with 2737 characters and 1050 parsimony-informative sites. The BI analysis converged both runs as indicated by the effective sample sizes (ESS) of all parameters above 2800 and the potential scale reduction factors (PSRF) equal 1.000 for all the parameters according to the 95% HPD Interval.
No members of the three genera within this superclade are pteruloid (i.e. coralloid basidiomes with dimitic hyphal system) in their morphology and consequently we introduce the family name Radulomycetaceae fam. nov. to accommodate them, as discussed further below. The current sister clade to Pterulaceae in our analyses is Stephanosporaceae, from which members of the Radulomycetaceae clade are clearly distinct phylogenetically and morphologically.

Phaeopterula (UFBoot = 100; BPP = 1)
Phaeopterula received maximum support in both analyses. It includes Pterula stipata, Pt. anomala, Pt. juruensis and other species which all have dark brown basidiomes. This clade is the first coralloid lineage to diverge within Pterulaceae. As these species render Pterula paraphyletic, a reclassification is needed. The generic name Phaeopterula was originally proposed as a subgenus of Pterula to accommodate Ph. hirsuta and Ph. juruensis (Hennings 1900;Hennings 1904). We propose its reintroduction below to distinguish these brownpigmented taxa from Pterula s. str.

Coronicium superclade (UFBoot = 98; BPP = 1)
This clade groups the remaining two resupinate genera of Pterulaceae, the monospecific Merulicium and Coronicium (UFBoot = 100; BPP = 1). Both genera form resupinate basidiomes but differ in the hyphal system present (dimitic in Merulicium, monomitic in Coronicium). Some Pterulicium species also show transitions in their morphology to a resupinate state. Corner (1950) showed that Pm. xylogenum Corner could form monomitic corticioid patches independent of the coralloid state and even in its absence, thus appearing to be truly corticioid. Furthermore, experimental studies on Pm. echo show a dimitic, resupinate, fertile corticioid phase both on agar and when cultured on cocoa twigs (McLaughlin and McLaughlin 1972;McLaughlin et al. 1978;McLaughlin and McLaughlin 1980). Despite the morphological distinctiveness from the rest of Pterulaceae, there is a trend in the morphology and strong phylogeny support for the placement of the Coronicium superclade among Pterula/Myrmecopterula and Pterulicium clades within Pterulaceae.

Pterulicium (UFBoot = 99; BPP = 1)
Two type species, Pterulicium xylogenum and Deflexula fascicularis, are nested within this clade alongside several species currently assigned to Pterula but which all have simple basidiomes (unbranched or limited branching). The Pterula species are interspersed with some Deflexula, rendering both genera polyphyletic. Pterulicium xylogenum forms a wellsupported subclade with Pterula secundiramea (= Pt. palmicola). Deflexula fascicularis forms a subclade with other Deflexula species that share globose spores, an unusual feature within Pterulaceae, most of which form ellipsoid to subamygdaliform spores.

Pterula (UFBoot = 100; BPP = 1)
This clade groups the true Pterula spp. that are represented by very bushy coralloid basidiomes, usually robust and taller than those of Pterulicium, stipe concolorous with hymenophore and lacking a cottony subiculum. Pterula has a mainly pantropical and pansubtropical distribution, with occurrence reported to all continents except Antarctica (Corner 1970).

Myrmecopterula (UFBoot = 97; BPP = 1)
This sister clade of Pterula represents the newly proposed genus (see below). It groups the two species cultivated by attine ants in the Apterostigma pilosum group with M. moniliformis and several unidentified free-living species. The species in this clade are only known from the Neotropics. Myrmecopterula is divided into seven subclades (Fig. 3) representing the two mutualists (MUTV and MUTN), three closely related to M. velohortorum (SAPV 1-3) and two closely related to M. nudihortorum (SAPN 1-2). Diagnosis: Differs from resupinate forms of Pterulaceae in the monomitic hyphal system and the absence of cystidia. Cystidia may be either present or absent in Pm. xylogenum, in the latter case the amygdaliform spores differentiate the species from Radulomyces that has ellipsoid to globose spores.
Etymology: From the type genus Radulomyces.
Notes: Radulomyces, Aphanobasidium and Radulotubus are placed in Radulomycetaceae. Larsson (2007) suggested that Lepidomyces had affinities to Aphanobasidium and could possibly be placed in Pterulaceae. However, no sequence data for the genus are available.
Lepidomyces is described as bearing pleurobasidia as in Aphanobasidium, but also leptocystidia as in Coronicium and Merulicium. Given its morphological similarities to Aphanobasidium and the Coronicium superclade, we retain Lepidomyces as incertae sedis until molecular data are available to confirm its phylogenetic position Description: Basidiome if present bushy, pteruloid, white-cream to light brown and greyish surface, normally concolorous or stipe with a darker tone than the hymenophore, arising from cottony subiculum with mycelial cords. Stipe surface sterile. Hyphal system, dimitic hyphal system. Basidiospores relatively small spores, usually less than 7 μm wide.
Ecology: Usually associated with the nests of ants, growing on top, or from a living or dead nest, or being cultivated by the ants.
Notes: Basidiomes of Myrmecopterula species are very similar to those of Pterula in habit, shape, and colour, but differ in the presence of mycelial cords and a cottony subiculum from which the basidiomes emerge. Some species of Myrmecopterula arise from soil, while others superficially appear to grow on wood. Closer observation of basidiomes formed on wood revealed that, rather than being lignicolous, they instead grow from a loose, granular substrate within a cavity inside the wood. This substrate in some cases resembles the substrate in the fungus gardens of the Apterostigma pilosum group of ants. In addition, M. moniliformis, which arises from soil, has been found emerging from active and inactive attine nests, (S. Sourell, pers. comm.; M. C. Aime, pers. comm.). Thus, all but one of the Myrmecopterula clades found to date had some association with attine ants, of which the two farmed mutualist species (M. nudihortorum and M. velohortorum) are best known. The five other species (of which only M. moniliformis is named) are less well studied and may play a role in decomposition of residual substrates in abandoned fungus garden, or potentially even as mycoparasites of the ant cultivar. In (See figure on previous page.) Fig. 3  Diagnosis: In the field, recognized by the absence of any veil on the fungus garden in Apterostigma nests, usually inside decomposing trunks or underground. In culture, it forms very little aerial mycelium and exhibits very slow growth (2-3 mm/week radial growth rate on PDA at 25C). Hyphal clamps abundant.
Notes: This species was formerly known as the ant cultivar G4. It is only known from the nest of fungusgrowing ants in the Apterostigma pilosum group in the A. manni subclade (Schultz 2007). ( Description: Basidiomes pteruloid, solitary or gregarious, scarcely branched to almost bushy, monopodial and slightly symmetric, branches from light brownish pink or greyish to pale brown and stipe dark reddish to rusty brown. Stipe surface glabrous with agglutinated hyphae (not sclerotioid) to villosetomentose. Dark brown mycelial cords usually present. Hyphal system dimitic with thick-walled skeletal hyphae, generative hyphae thin-walled and often clamped. Hymenial cystidia absent, caulocystidia sometimes present. Basidia terminal, clavate to suburniform. Basidiospores less than 9 μm varying between pip-shaped, subamygdaliform and ellipsoid.

Myrmecopterula velohortorum
Ecology: Growing on dead twigs or dead wood.
Notes: Hennings (1900) recognized the subgenus Phaeopterula to accommodate Pterula hirsuta that was distinguished from other Pterula species by the reportedly brown spores. Hennings (1904) later described a second species in the subgenus, Ph. juruensis, but noted that it was morphologically quite distinct from Ph. hirsuta. Phaeopterula was raised to generic level by Saccardo and Saccardo (1905) who cited only Ph. juruensis. Pterula hirsuta was recombined in Dendrocladium by Lloyd (1919) but later returned to Pterula by Corner (1950), even though Corner did not confirm the presence of brown spores in the samples he examined. Although we also have not observed pigmented spores in any of these taxa, dark brown pigments in the stipe hyphae are a consistent and diagnostic feature in this group, so we resurrect the name Phaeopterula. The term 'Phaeo-' relates to brown-pigmented basidiospores, but while members of this genus do not have brown basidiospores, they do contain brown hyphal pigments.

Introduction of Radulomycetaceae
We consider that it is better to erect a new family for these three genera (i.e. Radulomyces, Radulotubus and Aphanobasidium) than to leave them in Pterulaceae where they are clearly phylogenetically and morphologically distinct from nearly all the other members of Pterulaceae. In contrast, Merulicium (Fig. 2b-c) and Coronicium (Fig. 2a) form corticioid basidiomes but our phylogenetic analyses place them clearly within Pterulaceae. Two Pterulicium species, Pm. echo and Pm. xylogenum, also form both pteruloid and corticioid basidiomes, either independently or together (McLaughlin and McLaughlin 1980;Corner 1950).
Whilst the corticioid basidiomes of Merulicium and Pm. echo contain a dimitic hyphal system, typical of Pterulaceae, those of Coronicium spp. and Pterulicium xylogenum form a monomitic hyphal system, like all members of Radulomycetaceae. However, no members of Radulomycetaceae form cystidia, whereas these cells are found in most Pterulaceae (Corner 1950(Corner , 1952a(Corner , 1952b(Corner , 1967(Corner , 1970McLaughlin and McLaughlin 1980;Bernicchia and Gorjón 2010), including Coronicium spp. Thus, Radulomycetaceae is morphologically characterized by the combination of resupinate basidiomes, monomitic hyphal system and lack of cystidia. Moreover, our phylogenetic analyses strongly support the segregation of Radulomycetaceae from Pterulaceae.

Reintroduction of Phaeopterula
Phaeopterula spp. are distinct from other pterulaceous genera due to the distinctive brown colour of the main axis of the basidiome and monopodial/symmetric branching of these structures. This contrasts with other Pterulaceae which are either highly branched (bushy) and of uniform colour (Pterula and Myrmecopterula) or pigmented only at the stipe base, and (mostly) unbranched (Pterulicium). Hennings (1900) originally defined Phaeopterula by its brown spores. Corner (1950) cast doubt on the significance of this trait, but our results show that, despite an apparently misguided justification, Hennings was correct to group Ph. juruensis with Ph. hirsuta.
All Phaeopterula spp. are exclusively found on decaying wood, whereas members of other genera of Pterulaceae inhabit more diverse lignocellulosic substrates. Given the basal position of Phaeopterula in Pterulaceae, and the fact that all members of the sister family Radulomycetaceae are also lignicolous on wood, this habit is parsimoniously the ancestral condition. The reintroduction of Phaeopterula aims to pay tribute to Paul Hennings' work and his contribution to the taxonomy of Pterulaceae.

Synonymy of Deflexula with Pterulicium
Besides the paraphyly represented by Phaeopterula, the Pterulicium clade shows polyphyly of Pterula and Deflexula. Several species in the two latter genera are intermixed in a strongly supported subclade (Fig. 3). The presence of the type species of both Deflexula and Pterulicium within this clade requires that only one name be kept. Both genera were proposed by Corner (1950), to accommodate the dimitic and coralloid (but non-bushy) species, not fitting the description of Pterula. The name Pterulicium was based on a 'portmanteau' combination of Pterula and Corticium to reflect the presence of a corticioid patch at the stipe base (Corner 1950). However, this patch has only been reported in two species, Pterulicium xylogenum (Corner 1950) and Pm. echo (McLaughlin and McLaughlin 1980). Deflexula was named for the downward-oriented (positively geotropic) basidiomes (Corner 1950). Corner (1950)  Cottony subiculum absent, without association with attine ants Pterula close similarity between Deflexula and Pterulicium in the way the resupinate patch develops from the base of the basidiome. He also made a case for the formation of a fertile hymenium when facing downward in the two genera as supporting this similarity. Nonetheless, experimental studies on Pm. echo show that orientation of the hymenium does not affect the ability to produce spores, i.e., the hymenium is ageotropic (McLaughlin et al. 1978) and raised doubts about the validity of the genus Deflexula. This morphological distinction is not supported by phylogenetic analysis (Dentinger et al. 2009, Fig. 3) and its emphasis through taxonomic preservation would perpetuate misunderstanding. Accordingly, we propose to retain Pterulicium for this clade to avoid major misinterpretations of the species morphology.
Introduction of Myrmecopterula gen. nov.
Two species of Pterulaceae are cultivated by fungusfarming ants of the Apterostigma pilosum group in South and Central America (Dentinger et al. 2009;Munkacsi et al. 2004;Villesen et al. 2004;Mueller et al. 2018). Despite intensive investigation, neither has been observed to form basidiomes, but M. velohortorum is characterised by the formation of a veil of mycelium around the fungus garden, whilst M. nudihortorum lacks this veil. We recovered both species in a strongly supported clade, as a sister clade of Pterula, alongside five other subclades containing fertile, apparently free-living species. All the samples in this clade were collected from neotropical habitats (Fig. 1a-f), mostly as part of our recent fieldwork. During sampling campaigns by ourselves and others, it was observed that many of the 'free-living' specimens were associated in some way with living ant colonies or abandoned attine nests. Two Myrmecopterula samples belonging to subclade SAPV1 (CALD170307-02 and CALD170307-03; Fig. 1a) were found forming basidiomes atop two distinct but adjacent (1 m apart) living Apterostigma nests in Amazonian Rainforest. The cultivated mutualists from both nests were also analysed and found to belong to M. velohortorum confirming that the basidiomes were not linked to the cultivated mycelia in these nests. The third member of subclade SAPV1 was also reported forming a nascent basidiome on a living Apterostigma nest in Panama (Munkacsi et al. 2004). M. moniliformis (SAPN1; Fig. 1e) has been reported to be found outside both active and apparently inactive (see Myrmecopterula: Notes on Taxonomy section above) attine nests (S. Sourell, pers. comm.; M.C. Aime, pers.comm.) as was CALD170315-04 (SAPV2; Fig. 1b) and CALD170122-04 (SAPV3; Fig. 1c). Lastly, the mycelium of one sample (JSP 07-03 B 5.1; SAPV3) was isolated from a living Atta capiguara nest by Pereira et al. (2016).
The observations above and the phylogenetic analyses suggests that association with attine ants is a widespread trait amongst members of this clade, hence its naming as Myrmecopterula.
Most recent attention on Pterulaceae has been lavished on the ant-cultivated mutualists M. nudihortorum and M. velohortorum. These were once thought to be sister clades (Munkacsi et al. 2004;Villesen et al. 2004) but are now known to be only distantly related within the Myrmecopterula clade (Dentinger et al. 2009 , Fig. 3). This suggests two possibilities for the evolution of the Myrmecopterula-Apterostigma mutualism: (1) that it evolved independently on two occasions, or (2) that it is an ancestral condition of all Myrmecopterula. However, it is at present unclear whether the extant mutualistic association found for M. nudihortorum and M. velohortorum is ancestral, implying that the other taxa escaped the mutualism, or whether the looser association with ant nests widespread amongst members of Myrmecopterula was more recently elevated to a higher level of interdependence for these two species, as suggested by Dentinger et al. (2009). It is also possible that the free-living species within the Myrmecopterula may be specialised parasites specifically targeting their sister species that have formed a mutualism with the ants. An analogous situation is found in the leaf-cutting ants species Acromyrmex echinatior and its sister species Acromyrmex insinuator, the latter a highly specialised social parasite of the former (Sumner et al. 2004).
The basis of the association of 'free-living' species with attine ants and/or their abandoned nests is unclear. Given the apparent preference of some for abandoned nests, they may be specialised early stage colonisers of ant nest debris. A further possibility is that they are cheaters, deriving nutrition from the ant-collected biomass but not reciprocating by producing hyphae palatable to ants. This would represent a novel form of fungal mimicry, perhaps achieved by the ants' inability to differentiate hyphae of closely related species. Lastly, they may be mycoparasitic, including on ant cultivars, although there is currently no direct evidence supporting this hypothesis.

Re-delimitation of Pterulaceae
All the accepted genera in Pterulaceae were sampled in this study except for the monotypic Allantula. One specimen, with morphology consistent with Corner's description of Allantula diffusa, with pteruloid basidiomes borne on slender mycelial cords as curved intercalary swellings, was collected during our fieldwork (Fig. 1m). Phylogenetic reconstruction placed this specimen firmly within Phaeopterula. However, we have been unable to obtain the type specimen (no other collections authenticated exist) for more detailed analysis.