Phylogenetic Classification at Generic Level in the Absence of Distinct Phylogenetic Patterns of Phenotypical Variation: A Case Study in Graphidaceae (Ascomycota)

Molecular phylogenies often reveal that taxa circumscribed by phenotypical characters are not monophyletic. While re-examination of phenotypical characters often identifies the presence of characters characterizing clades, there is a growing number of studies that fail to identify diagnostic characters, especially in organismal groups lacking complex morphologies. Taxonomists then can either merge the groups or split taxa into smaller entities. Due to the nature of binomial nomenclature, this decision is of special importance at the generic level. Here we propose a new approach to choose among classification alternatives using a combination of morphology-based phylogenetic binning and a multiresponse permutation procedure to test for morphological differences among clades. We illustrate the use of this method in the tribe Thelotremateae focusing on the genus Chapsa, a group of lichenized fungi in which our phylogenetic estimate is in conflict with traditional classification and the morphological and chemical characters do not show a clear phylogenetic pattern. We generated 75 new DNA sequences of mitochondrial SSU rDNA, nuclear LSU rDNA and the protein-coding RPB2. This data set was used to infer phylogenetic estimates using maximum likelihood and Bayesian approaches. The genus Chapsa was found to be polyphyletic, forming four well-supported clades, three of which clustering into one unsupported clade, and the other, supported clade forming two supported subclades. While these clades cannot be readily separated morphologically, the combined binning/multiresponse permutation procedure showed that accepting the four clades as different genera each reflects the phenotypical pattern significantly better than accepting two genera (or five genera if splitting the first clade). Another species within the Thelotremateae, Thelotrema petractoides, a unique taxon with carbonized excipulum resembling Schizotrema, was shown to fall outside Thelotrema. Consequently, the new genera Astrochapsa, Crutarndina, Pseudochapsa, and Pseudotopeliopsis are described here and 39 new combinations are proposed.


Introduction
Molecular data have revolutionized our understanding of the evolution of organisms and have had profound impact on classifications, especially in organisms lacking complex morphologies, such as fungi [1][2][3]. Traditionally, the classification of living organisms has worked under the paradigm that taxa should be recognizable, i.e. having phenotypic features that delimit them from other taxa. However, a major challenge of the results of molecular phylogenetic studies is that lineages often do not correlate well with phenotypic features [4][5][6][7][8][9][10][11]. In these cases, reexamination of phenotypical characters often fails to identify diagnostic characters, especially in organismal groups lacking complex morphologies. Reasons for the absence of phenotypical differences among clades include convergent evolution, parallel but independent transformations of morphological characters in related lineages, as well as morphostasis and retention of ancestral features [12][13][14][15][16]. Hence, delimitation based on morphology alone can be difficult or even impossible. This problem has long been recognized and accepted at higher taxonomic levels, such as orders and families, which often cannot be circumscribed by phenotypic characters. Even so, their formal recognition does not appear to pose any conceptual problem, as apparent from widely accepted classifications [1,17].
Due to the nature of binomial nomenclature introduced by Linnaeus, in which a species name is composed of the generic name and the epitheton, changes in the classification of an organism at the generic level lead to a change in the name of the organism. Thus, systematists have been reluctant to translate phylogenetic studies into classification at the generic level when monophyletic clades lack correlating phenotypical features. When a genus-level taxon is found to be poly-or paraphyletic, it can be either split or merged with another taxon to obtain monophyly. Since genera and all higher taxonomic ranks are arbitrary, both lumping or splitting would be possible and there is no a priori scientific argument to favor either solution. Such reclassifications often lead to genera that are not distinguishable by phenotypical characters, and these have been called ''cryptic genera'' [18][19][20][21] analogous to ''cryptic species'', which are morphologically undistinguishable [22][23][24][25][26][27][28].
Here we propose a new, quantitative approach to choose among alternative classifications, combining the technique of morphology-based phylogenetic binning with a multi-response permutation procedure (MRPP) [29][30][31]. Phylogenetic binning [32] is a method that determines the level of congruence between phenotypical site patterns and molecular phylogenies and then applies character weights in order to improve the accuracy of the classification of taxa for which no molecular data are yet available. The advantage of this method is the individual placement of taxa in a reference tree based on closest relationship, rather than overall difference between clades. Thus, this method yields better results than simultaneous clustering or cladistic analysis of many taxa based on morphological data. MRPP compares average distances between groups based on characters (putative taxa), using data randomization to obtain evidence of statistical significance. Both methods can be combined to test alternative classification models in order to evaluate which classification best fits both the phylogenetic topology and the morphological data, under the criterion that resulting taxa should be monophyletic.
We used the tribe Thelotremateae in Graphidaceae, a family of lichenized fungi, to illustrate our approach [33,34]. This clade includes four currently accepted genera: Chapsa, Chroodiscus, Leucodecton, and Thelotrema. Previous studies showed that Chroodiscus and Leucodecton are monophyletic [33][34][35][36], while Chapsa was shown to be highly polyphyletic. The core of Thelotrema was found to be monophyletic with one species having unclear phylogenetic reltionships. The genus Chapsa in its current sense is characterized morphologically by so-called chroodiscoid apothecia with widely open disc bordered by a splitting, lobulate margin, as well as the presence of lateral paraphyses, which are hyphae growing into the hymenium from the lateral margin of the fruiting body [37,38]. However, Chapsa was found to consist of unrelated lineages [33] that fall both inside and outside the Thelotremateae, and even the Chapsa species within Thelotremateae form at least two clades, one of which having low support but including three subclades with strong support, whereas the other clade includes two supported subclades. At first glance, there are no apparent phenotypic characters that would separate these clades, since thallus morphology, ascospore type, and secondary chemistry vary widely in each clade. As a consequence, one could recognize a single genus Thelotrema for all Thelotremateae, which would, however, not do justice to the morphologically and phylogenetically welldefined clades representing Chroodiscus, Leucodecton, and Thelotrema sensu stricto. The preferred alternative would be splitting Chapsa into more than one genus, but without any obvious, supporting morphological characters the decision for either two, four, or even five genera would be arbitrary. Our approach provides statistical evidence that helps to choose among alternatives and we consider this a model case in how to tackle classifications of morphologically complex groups with clear underlying phylogenetic topologies.

Phylogenetic Analyses
Seventy-five new sequences were generated for this study and aligned with 237 sequences downloaded from Genbank, most of them generated in our lab and included in a previous study ( Table 1). The combined data matrix of 2482 unambiguously aligned characters with 804 characters in the nuLSU rDNA, 800 characters in mtSSU rDNA and 878 characters in RPB2 was used for phylogenetic analyses. The single gene analyses did not show any conflicts and hence the concatenated data set was analyzed. The ML tree had a likelihood value of -38,803.262 and in the B/ MCMC analysis of the combined data set, the likelihood parameters in the sample had the following mean (Variance): LnL = -42,231.616 (0.17). The maximum likelihood tree did not contradict the Bayesian tree topologies and hence only the majority-rule consensus tree of the Bayesian tree sampling is shown (Fig. 1).
In the phylogenetic tree, the genus Chapsa is polyphyletic, separating into two major clades, one of which is unsupported but consists of three well-supported subclades. Distantly related species of Chapsa sensu lato also appear in other clades, such as C. platycarpa in the outgroup close to the genus Wirthiotrema. Chapsa clade I is a well-supported clade; it contains the type species, C. indica, and the morphologically similar C. leprocarpa, C. niveocarpa, C. patens, and C. pulchra, but also C. alborosella and the morphologically quite disparate C. sublilacina and relatives. The clade forms two supported subclades, one containing C. alborosella and C. sublilacina, among other species, and the other C. indica, C. leprocarpa, and C. patens, among other species. Chapsa clade II can be divided into subclades IIa, IIb, and IIc. Subclade IIa contains C. dilatata and C. phlyctidioides, which morphologically resemble species of Clade I; subclade IIb comprises the single species C. laceratula, which resembles a Topeliopsis in apothecial morphology but with well-developed, corticate thallus; and subclade IIc includes the morphologically disparate C. astroidea, C. mastersonii, and C. zahlbruckneri, among other species. Hypothesis testing using both the SH and ELW strongly rejected the monophyly of Chapsa, even if only considering the species falling within the Thelotremateae (p#0.0001 in both tests). The Thelotrema clade is supported as a monophyletic group, but excluding Thelotrema petractoides, which falls outside the main clade as an early diverging taxon with uncertain phylogenetic relationships.

Phylogenetic Binning and Multi-response Permutation Procedure
Phylogenetic binning of the 65 described Chapsa species for which no molecular data are available suggests placement of 14 species within Clade I (Chapsa sensu stricto) and 49 species within Clade II under a 2-clade solution with ML weighting. Two species, C. chionostoma and C. microspora, are suggested to not form part of tribe Thelotremateae ( Six of these are identical whereas seven are conflictive between the two weighting techniques, for a total of 20 conflictive placements in the 5-clade solution ( Table 2).
The MRPP analysis resulted in non-significant or spuriously significant differences between groups for the 2-clade solution but in highly significant differences for the 4-clade solution, independent of group assignment based on ML or MP weighting and of the distance measure employed (Table 3). Group assignments based on MP weighting gave slightly better correlations than based on ML weighting, as did the Euclidean distance measure compared to a linear correlation coefficient (Table 3). This suggests that the 4-clade solution and group assignment based on MP weighting is the best fit to the data.
Kruskal-Wallis ANOVA indicates five characters as significantly discriminating between groups in a 2-clade solution using ML weighting and an additional three characters as marginally significant (Table 4). MP weighting results in a similar pattern but with overall fewer discriminating characters. Both the 4-clade and the 5-clade solutions suggests a much higher number of discriminating characters, again with a higher total for ML weighting. This supports the 4-clade or 5-clade solutions providing a better fit to the data than the 2-clade solution, with a slight advantage for ML over MP weighting.
The best discriminating characters in the 4-clade and 5-clade solutions (Table 5) are the presence of soralia (MP weighting only), the nature of the thallus cortex, ascoma exposure and the shape of the proper and thalline margin, excipulum carbonization, ascospore number and dimensions (ML weighting only), ascospore endospore development and iodine reaction (best discriminating character under both ML and MP weighting), ascospore septation, and secondary chemistry (stictic and protocetraric acid; ML weighting only). According to these results, species of Clades Ia and IIa tend to have a loose cortex or lack a cortex altogether, whereas species of Clades Ib, IIb and IIc mostly have a dense cortex. Soralia are entirely confined to Clade I and particularly Clade Ib. Brown excipula are significantly more frequent in Clades IIa and IIc. Ascospores with amyloid endospore are particularly frequent in Clades Ib and IIa, and the latter clade also tends to have ascospores with a lower number of transverse septa than the other clades. The partial differences found in the level of character discrimination between taxon placement based on ML or MP weighting correlate with the weights determined for each character in the phylogenetic binning analysis. Characters that received high weights under an ML approach but low weights under MP include the thallus cortex, ascospore number and dimensions, and secondary chemistry, whereas the opposite was found for characters such as the presence of soralia.

Discussion
The detection of phylogenetically defined clades lacking clearly discriminant morphological characters is not rare and particularly common in fungal groups, including lichenized species, since these organisms are composed of rather simple structures and are less differentiated than higher plants and animals [39,40]. At higher taxonomic levels, such as family and order, this phenomenon has already found broad acceptance in fungal classifications [1,17,41]. Also, increasing evidence points to the frequent occurrence of cryptic species [22,23,25,27,28]. At the generic level, however, systematists have been highly reluctant to accept so-called cryptic genera, mainly because in lichenized fungi, the genus level has been the main taxonomic entity for classification purposes and herbaria collections are mostly organized using this taxonomic category. It is also commonly expected that classifications should result in the recognition of taxa, particularly at the genus level, that are phenotypically recognizable. However, increasing evidence from phylogenetic studies indicates that, while many monophyletically circumscribed genera are indeed recognizable, in other cases clades are inconsistent with morphological data. In many cases, particular morphotypes form either paraphyletic grades, with other morphotypes nested within, or are polyphyletic. Since the objective of molecular phylogenetic studies is to recognize natural groups, such para-or polyphyletic taxa cannot be maintained, except in the case of recently evolved species that have experienced the founder effect [42][43][44][45].
The genus Chapsa [37] had already been suspected to be not monophyletic, but the split into up to five clades within tribe Thelotremateae, and the placement of additional species outside this tribe, poses a challenge to classification, since there are no straightforward characters or a combination thereof that can be used to distinguish these lineages phenotypically. This situation is not rare in Graphidaceae and has also been found in the genera Graphis versus Allographa, Myriotrema versus Ocellis, Leucodecton versus Wirthiotrema, and Leucodecton versus Leptotrema, versus Ocellularia [15,32,33,46]. There are numerous other examples of this situation among fungi and they are usually accepted if the lineages are unrelated or distantly related, but disputed in case of closer relationships, even if the underlying problem is the same. Thus, the basidiolichen genera Multiclavula and Lepidostroma include species that cannot be separated by any phenotypical character, but their very distant position among the Basidiomycota provides grounds for their taxonomic separation [47,48]. The opposite phenomenon is also not rare: closely related lineages that are widely disparate morphologically, such as the genera Cruentotrema and Dyplolabia in Graphidaceae [15]. It is surprising that morphologically variable lineages merged into a single taxon are more readily accepted than separate, morphologically cryptic lineages, even if the underlying problem of lack of phenotypic consistency is the same.
In the case of the genus Chapsa, the situation is especially complex since up to five clades can be distinguished based on molecular phylogeny in the tribe Thelotremateae alone. Our approach shows that a 4-clade or 5-clade solution fits the data much better that a 2-clade solution, since the between-group differences are highly significant and the number of discriminating characters is much higher than in a 2-clade solution. The data do not allow to determine whether the 4-clade or 5-clade solution is the best fit (except that the latter has a slightly higher number of conflictive placements), since both are highly significant in terms of morphological discrimination and have about the same number of discriminating characters. Because of the lack of difference between the two alternatives, and since the branch leading to Clade I has high support, we opt for the more conservative solution here and maintain Clade I as a single genus, Chapsa. Our results suggest that there is a strong tendency for each of the four Figure 1. Bayesian 50% majority-rule consensus tree depicting relationships among genera in the tribe Thelotrematae on the basis of a concatenated data set including mtSSU rDNA, nuLSU rDNA and protein-coding RPB2. Posterior probabilities equal or above 0.95 are indicated as bold branches. ML-bootstrap support equal or above 70% is shown as number at branches. doi:10.1371/journal.pone.0051392.g001 Table 1. Species and specimens used in the present study, with location, reference collection details, and GenBank accession numbers. Newly obtained sequenced in bold.  clades to differ in thallus cortex type, excipular carbonization, and ascospore type and septation (especially endospore development and iodine reaction), which is consistent with each clade having a long stem node and hence having evolved internal morphological variation independently. However, each clade includes a few species that morphologically would better fit in another clade. Such cases must be interpreted as either ancestral characters retained in a clade or as examples of parallel evolution. For this reason, these characters, even if statistically discriminant between genus-level clades, cannot be used to actually key out the genuslevel taxa themselves.
The phylogenetic binning method allows both ML and MP weighting of the morphological characters, but the underlying algorithms are different [32]. For ML weighting, the bestscoring molecular tree is compared to a set of randomized trees (e.g. 100 trees) and the weight is derived by the number of random trees in which a particular morphological character mapped on the tree receives a worse log likelihood score. If a character has a strongly consistent distribution on the bestscoring molecular tree, any randomized tree will have a lower log likelihood score for this character, and the weight will be 100%. In contrast, MP weighting is derived only from the best-  19 20 scoring molecular tree, and parsimony scores are computed by mapping the morphological characters on the tree and converting the scores into weights. As a result, ML weighting will emphasize characters that are confined to particular major clades (absolute synapomorphies), whereas MP weighting will also give higher weights to characters that characterize more than one clade but are absent from others (relative synapomorphies). Which method works better depends on the context, but the slightly better MRPP results for MP weighting in the 2clade solution in this study confirm the findings of the original paper [32] that MP weighting might give slightly more consistent results. Interestingly, the characters found here to receive higher weights under ML are ascospore number and dimensions as well as number of septa, whereas under MP their weight was zero. These characters are known to vary strongly even within clades but are usually consistent between more closely related species, which could cause the effect that randomized trees consistently give lower log likelihood scores even if the overall character distribution over the tree is nearrandom. It is therefore recommended to use both ML and MP weighting in combination and closely inspect taxa with conflicting placement under both approaches, but if both methods yield quantitatively similar results overall, the MP solution is preferable as done here. In contrast to making phenotypical characters obsolete in systematics, our study underlines the importance of these data even in times where molecular data become increasingly available to reconstruct phylogenies. While phenotypical data itself should not be used in such reconstructions, they are indispensable when transforming phylogenies into classifications and, with powerful analytical methods, provide statistical evidence that can be used to compare alternative classification models based on an underlying phylogeny.

Taxonomic Conclusions
The results of our phylogenetic analysis and the combined binning/multiresponse permutation procedure support a classification accepting each of the four Chapsa clades as different genera. Also, Thelotrema petractoides was shown to be distantly related to the core genus. Consequently, the new genera Astrochapsa, Crutarndina, Pseudochapsa, and Pseudotopeliopsis are described below and 39 new combinations are proposed. We only propose new combinations for species without conflict regarding clade placement under either ML or MP weighting, whereas the conflictive species are provisionally retained in Chapsa until sequence data become available. Therefore, the number or proposed combinations is lower than the numbers indicated in Table 2 for each clade. For example, the results suggest to place 24 taxa under ML and 21 taxa under MP weighting in Pseudochapsa, but only 16 of these are identical with both approaches, and only these are recombined here. In addition, we refrained from formally recombining five taxa that did not exhibit conflict but are suspected to possibly fall outside the Thelotremateae and hence require sequencing to clarify their position. This might also apply to some of the taxa with conflicting placement, such as C. asteliae and C. lordhowensis (see Appendix S1). Two species were recombined in Pseudotopelipsis favoring the MP weighting solution.
Astrochapsa Differing from Chapsa s.str. in the more frequently densely corticate thallus, the mostly recurved apothecial margin, and the almost exclusively subdistoseptate, non-amyloid ascospores.
Thallus usually with dense cortex, rarely with loose cortex or ecorticate. Apothecia erumpent, rounded to irregular in outline; disc exposed; margin lobulate to usually recurved. Excipulum usually brown. Ascospores septate to muriform, fusiform-ellipsoid to oblong-cylindrical, with slightly thickened septa and angular lumina (subdistoseptate), colorless or rarely brown, almost exclusively I-. Secondary chemistry: no substances or frequently stictic acid and relatives; apothecial disc sometimes pigmented.
Etymology: Derived from ''astro'' (Greek, starry) because of the star-like morphology of the ascomata and the genus name Chapsa.

Taxon Sampling and Molecular Methods
We assembled a three-locus data set consisting of mtSSU rDNA, nuLSU rDNA, and the protein-coding genes RPB2. The taxon sampling contained 60 species focusing on the tribe Thelotremateae ( Table 1). The outgroup taxa were chosen based on previous phylogenetic results [33]. New sequences were generated for this study using the Sigma REDExtract-N-Amp Plant PCR Kit (St. Louis, Missouri, SA) for DNA isolation following the manufacturer's instructions, except that 40 mL of extraction buffer and 40 mL dilution buffer were used. DNA dilutions (5x) were used in PCR reactions of the genes coding for the nuLSU, mtSSU and RPB2, respectively.Primers for amplification were: (a) for nuLSU: AL2R [35], and nu-LSU-1125-39 ( = LR6) [49], (b) for mtSSU: mr-SSU1 and Mr-SSU3R [50], and (c) for RPB2: fRPB2-7cF and fRPB2-11aR [51]. The cycle sequencing conditions were as follows: 96uC for 1 minute, followed by 25 cycles of 96uC for 10 seconds, 50uC for 5 seconds and 60uC for 4 minutes. Samples were precipitated and sequenced using Applied Biosystems 3730 DNA Analyzer (Foster City, California, U.S.A.). Sequence fragments obtained were assembled with SeqMan 4.03 (DNASTAR) and manually adjusted.

Sequences Alignments and Phylogenetic Analyses
Alignments were done using Geneious Pro 5.5.2 [52]. Ambiguously aligned portions were removed manually. The single-locus and concatenated alignments were analyzed by maximum likelihood (ML) and a Bayesian approach (B/MCMC). To test for potential conflict, ML bootstrap analyses were performed on the individual data sets, and 75% bootstrap consensus trees were examined for conflict [2].
The ML analysis of the concatenated alignment was performed with the program RAxML-HPC2 (version 7.3.1) on XSEDE [53] using the default rapid hill-climbing algorithm. The model of nucleotide substitution chosen was GTRGAMMA. The data set was partitioned into five parts (mtSSU, nuLSU and each codon position of RPB2), so each gene partition was treated as an independent data set. Rapid bootstrap estimates were carried out for 1000 pseudoreplicates [54].
The B/MCMC analysis was conducted on the concatenated data set using MrBAYES 3.1.2 [55], with the same substitution model as in the ML analysis. A run with 10,000,000 generations, starting with a random tree and employing four simultaneous chains, was executed. No molecular clock was assumed. Heating of chains was set to 0.2. Posterior probabilities were approximated by sampling trees using a variant of Markov Chain Monte Carlo (MCMC) method. To avoid autocorrelation, only every 1000th tree was sampled. The first 4,000 generations were discarded as burn in. We used AWTY [56] to compare splits frequencies in the different runs and to plot cumulative split frequencies to ensure that stationarity was reached. Of the remaining19992 trees (9996 from each of the parallel runs) a majority rule consensus tree with average branch lengths was calculated using the sumt option of MrBAYES. Posterior probabilities were obtained for each clade. Clades with bootstrap support above 70% under ML and posterior probabilities $0.95 were considered as strongly supported. Phylogenetic trees were visualized using the program Treeview [57].

Anatomical and Chemical Studies
Anatomical studies were conducted using standard light microscopy on hand-cut sections mounted in water. Secondary lichen substances were identified by thin-layer chromatography (TLC) and high performance thin-layer chromatography (HPTLC) according to standard methods [58,59].

Hypothesis Testing
Our phylogenetic analyses revealed that the genus Chapsa did not form a monophyletic group. Thus we tested whether our data are sufficient to reject monophyly of this genus. For the hypothesis testing, we used two different methods: (1) Shimodaira-Hasegawa (SH) test [60] and (2) expected likelihood weight (ELW) test [61]. The SH and ELW test were performed using Tree-PUZZLE v.5.2 [62] with the concatenated dataset, comparing the best tree agreeing with the null hypotheses, and the unconstrained ML tree. These trees were inferred in Tree-PUZZLE using the GTR+I+G nucleotide substitution model.

Morphology-based Phylogenetic Binning
Since the molecular data set corresponding to the genus Chapsa included 21 species, but the entire genus considered here comprises 86 accepted species, molecular data were unavailable for 65 species or about 75% of all currently accepted species. In this case, the phylogenetic binning method provides a statistical approach for a predictive classification of species, by weighting the morphological characters based on their distribution on the phylogenetic tree of the sequenced species and then placing each additional species known from morphological characters only separately in the tree and testing alternative placements by means of bootstrapping [32]. The weighting can be applied using both an MP and an ML approach. In this case, we applied the binning method for three alternative solutions: a 2-genus, a 4-genus, and a 5-genus solution. We and used both MP and ML weighting, to generate six possible alternative classifications of Chapsa based on both molecular and morphological data: ML-2, MP-2, ML-4, MP-4, ML-5, and MP5.

Multi-response Permutation Procedure
A multiresponse permutation procedure is a simple and effective tool to test for differences between groups of entities, in this case the groups obtained by the four alternative classifications obtained from the molecular phylogeny and subsequent binning method [63]. Since the morphological data matrix included only binary and ordered multistate characters, both Euclidean distances and linear correlation coefficients between each element within and between each group were computed. Within-and between group distances were then compared and statistical significance was tested by random data permutation using random shuffling of group partitions [63]. If within-group distances are smaller than expected by chance, it supports the recognition of a group as taxon, since such a result is evidence for partially independent phenotypic evolution. The analysis was performed in PC-Ord 5.03 [63]. For each individual character, we also employed Kruskal-Wallis ANOVA using each of the alternative clade solutions as grouping variable, to test whether the character discriminates between the resulting groups; this analysis was done in STATISTICA TM 6.0.

Nomenclature
The electronic version of this article in Portable Document Format (PDF) in a work with an ISSN or ISBN will represent a published work according to the International Code of Nomenclature for algae, fungi, and plants, and hence the new names contained in the electronic publication of a PLOS ONE article are effectively published under that Code from the electronic edition alone, so there is no longer any need to provide printed copies.
In addition, new names contained in this work have been submitted to MycoBank from where they will be made available to the Global Names Index. The unique MycoBank number can be resolved and the associated information viewed through any standard web browser by appending the MycoBank number contained in this publication to the prefix http://www.mycobank. org/MB/. The online version of this work is archived and available from the following digital repositories: PubMed Central, LOCKSS.

Supporting Information
Appendix S1 Clade placement of taxa according to molecular phylogenetic analysis and phylogenetic binning according to the different classification solutions using 2, 4, or 5 clades under either ML or MP weighting. (DOC)