Phylogenetic Relationships of Agaric Fungi Based on Nuclear ...

Syst. Biol. 49(2):278-305,2000

Phylogenetic Relationships of Agaric FungiBased on Nuclear Large Subunit Ribosomal DNA Sequences

JEAN-MARC MONCALVO, FRANCOIS M. LUTZONI,1 STEPHEN A. REHNER,2

JACQUI JOHNSON,3 AND RYTAS VILGALYS

Department of Botany, Duke University, Durham, North Carolina 27708, USA,E-mail: [email protected] (J.-M.M.),[email protected] (R.V.)

Abstract.—Phylogenetic relationships of mushrooms and their relatives within the order Agaricaleswere addressed by using nuclear large subunit ribosomal DNA sequences. Approximately 900bases of the 5' end of the nucleus-encoded large subunit RNA gene were sequenced for 154 selectedtaxa representing most families within the Agaricales. Several phylogenetic methods were used, in-cluding weighted and equally weighted parsimony (MP), maximum likelihood (ML), and distancemethods (NJ). The starting tree for branch swapping in the ML analyses was the tree with the high-est ML score among previously produced MP and NJ trees. A high degree of consensus was ob-served between phylogenetic estimates obtained through MP and ML. NJ trees differed accordingto the distance model that was used; however, all NJ trees still supported most of the same terminalgroupings as the MP and ML trees did. NJ trees were always significantly suboptimal when evalu-ated against the best MP and ML trees, by both parsimony and likelihood tests. Our analyses sug-gest that weighted MP and ML provide the best estimates of Agaricales phylogeny. Similar supportwas observed between bootstrapping and jackknifing methods for evaluation of tree robustness.Phylogenetic analyses revealed many groups of agaricoid fungi that are supported by moderate tohigh bootstrap or jackknife values or are consistent with morphology-based classification schemes.Analyses also support separate placement of the boletes and russules, which are basal to the maincore group of gilled mushrooms (the Agaricineae of Singer). Examples of monophyletic groups in-clude the families Amanitaceae, Coprinaceae (excluding Coprinus comatus and subfamily Panae-olideae), Agaricaceae (excluding the Cystodermateae), and Strophariaceae pro parte (Stropharia,Pholiota, and Hypholoma); the mycorrhizal species of Tricholoma (including Leucopaxillus, also mycor-rhizal); Mycena and Resinomycena; Termitomyces, Podabrella, and Lyophyllum; and Pleurotus with Ho-henbuehelia. Several groups revealed by these data to be nonmonophyletic include the families Tri-cholomataceae, Cortinariaceae, and Hygrophoraceae and the genera Clitocybe, Omphalina, andMarasmius. This study provides a framework for future systematics studies in the Agaricales andsuggestions for analyzing large molecular data sets. [Fungal evolution; higher phylogeny; homoba-sidiomycete; large-scale molecular phylogeny; tree support.]

It is particularly hoped . . . that DNA Gilled mushrooms and their allies (Basid-analysis can be methodically extended iomycota: Agaricales) are among the mostto generic taxonomy and that ways familiar of fungi. Often cryptic, they are ac-will be discovered to add new ap- tually a prominent component of most ter-proaches to the solution of problems. restrial ecosystems and perform a wide va-Singer, (1986:viii) riety of ecological roles as saprophytes,

mutualists, and parasites. Taxonomy ofWhence cometh the Agarics? Miller mushrooms has traditionally relied on mor-and Watling (1987) phological characters that are known to be

subject to parallel evolution and pheno-typic plasticity; as a result, many moderngenera and families are artificial, and my-

iPresent address: Department of Botany, The Field cologists Still disagree about taxonomicMuseum of Natural History, Chicago, Illinois 60605, limits of the Agaricales and the identity ofUSA; E-mail: [email protected] natural groups within the order.

Present address: Department of Biology, Univer- j ^ e a r l i e s t classification system for agar-sity of Puerto Rico, Box 23360, San Juan, Puerto Rico • j ,i i - j - i u r • / i o m \nno^i c -i I.* u ^ u 1CS an(^ o ther bas idiomycetes by Fries (1821)00931; E-mail: [email protected] , , . .J , . J . . \ . '

3Present address: Louis Calder Biological Field Sta- w a s m o s t n o t a b l e f o r l t s c l a n t v ' l o g i c a l s i m "tion, Fordham University, Armonk, New York 10504, pllClty, and complete artificiality. In his lastUSA. general work , Fries (1874) recognized one

278

Dow

nloaded from https://academ

ic.oup.com/sysbio/article/49/2/278/1687075 by guest on 24 July 2022

2000 MONCALVO ET AL.—RIBOSOMAL DNA PHYLOGENY OF MUSHROOMS 279

family and 20 genera of gilled (agaricoid)fungi. Beginning with Patouillard (1887,1900), mycologists began to revise Fries'staxonomy and to recognize increasinglylarger numbers of segregate genera basedon ever-more-restricted sets of characters. Acentury after Fries, modern taxonomic sys-tems recognize up to 230 genera (Singer,1986) that have been classified into as manyas 80 families and 25 orders (Singer, 1962,1975, 1986; Moser, 1967; Kreisel, 1969; Kuh-ner, 1980; Jiilich, 1981; Locquin, 1984).

The most comprehensive modern taxo-nomic treatment for the Agaricales is that ofSinger (1986; see also Appendix), based onthat author's exhaustive first-hand knowl-edge of mushrooms from virtually all re-gions of the world. In Singer's (1986)system the Agaricales encompass three sub-orders. Most modern mycological floristictreatments largely follow Singer's system(Moser, 1967; Pegler, 1977, 1983; Bas et al.,1988). However, at least several alternativesto Singer's classification system have beenproposed, most notably by Kiihner and Ro-magnesi (1978), Kiihner (1980), Jiilich(1981), and Redhead (1986). Until recently,it has been difficult to evaluate the weak-nesses and strengths of each classificationsystem in the absence of knowledge aboutphylogenetic relationships.

A growing number of phylogenetic stud-ies have started to address the evolutionaryrelationships of mushrooms and their rela-tives, using evidence from ribosomal RNAgene sequences (Bruns et al., 1992, 1998;Moncalvo et al., 1993, 1995; Chapela et al.,1994; Hopple and Vilgalys, 1999; Hibbettand Donoghue, 1995; Binder et al., 1997;Hibbett et al., 1997; Johnson and Vilgalys,1998; Lutzoni, 1997). Monophyly of Singer'sAgaricales has been rejected by several ofthese studies (Hibbett and Vilgalys, 1993; Hi-bbett et al., 1997; Thorn et al., 2000; Hibbettand Thorn, in press), which instead supportthe recognition of a clade of "euagarics"(gilled mushrooms) corresponding largelyto Singer's concept for the Agaricineae butalso including a number of nonagaric fungisuch as puffballs and coral mushrooms.

In this paper we analyze phylogenetic re-lationships among the major evolutionarylines of agaric fungi, using sequence data

from nuclear-encoded large subunit riboso-mal RNA genes (nLSU-rDNA) from 154 di-verse taxa. Taxonomic sampling for thisstudy has been focused on the Agaricineae(as circumscribed by Singer, 1986), whichalso constitutes the core group of euagaricsrecognized by Hibbett et al. (1997). Phylo-genetic analyses were conducted with sev-eral methods, including maximum parsi-mony (MP), maximum likelihood (ML),and neighbor-joining (NJ). Our results pro-vide molecular support for many well-defined groups of agarics and also help toresolve phylogenetic relationships for somecontroversial taxa.

We also explored several general aspectsof large-scale phylogenetic analysis involv-ing molecular data by using a variety ofsearch strategies aimed at finding optimaltrees. The reconstruction of complex phylo-genies is one of the most rapidly changingand controversial areas of systematics, andmany uncertainties surround the choice ofappropriate algorithms, models, and searchstrategies for dealing with ever-larger datasets, for which optimal solutions mightnever be known (Hillis, 1996; Rice et al.,1997; Graybeal, 1998; Kim, 1998; Soltis et al.,1998). Earlier studies that have reported ondetails and problems associated with largeDNA sequence data sets were mainly fromplant groups (Chase et al., 1993; Soltis et al.,1997, 1998; Soltis and Soltis, 1997). Thisstudy is the first to explore this issue by us-ing an intensive sampling of sequencesfrom a single group of fungi.

MATERIALS AND METHODS

Taxonomic Sampling

Taxa for this study were classified ac-cording to Singer's (1986) Agaricales in Mod-ern Taxonomy. A large collection of nLSU se-quences obtained from previous studies(Chapela et al., 1994; Vilgalys and Sun,1994; Lutzoni, 1997; Johnson and Vilgalys,1998; Hopple and Vilgalys, 1999) was com-bined into a single data matrix. Sequencesof additional taxa were sampled to includerepresentatives from each of Singer's (1986)families; only two families (Crepidotaceaeand Gomphidiaceae) were not sampled.The choice of taxa for phylogenetic analysis

Dow



280 SYSTEMATIC BIOLOGY VOL. 49

was eventually made from a database of>300 sequences and was based on three cri-teria: (1) taxonomic diversity (emphasis oncurrent classifications), (2) sequence quality(absence of regions with missing data, hy-pervariability, or insertions/deletions [in-dels]), and (3) relative sequence diversityand nonredundancy (emphasis on sequencedivergence and phylogenetic breadth). Thespecies selected for phylogenetic analysisare listed in Appendix, along with taxo-nomic classification according to Singer(1986) and GenBank accession numbers ofthe nLSU-rDNA sequences. For rootingpurposes, two species of the polyporegenus Ganoderma were included in theanalysis as an outgroup to represent thePolyporaceae, which has been suggested asa possible outgroup to the Agaricales ac-cording to the molecular phylogeny of ho-mobasidiomycetes (Hibbett et al., 1997).

Molecular Techniques

The region targeted for phylogeneticanalysis of sequence data was the 5'end ofthe nLSU-rDNA gene, which encompassesdivergent domains Dl to D3 as defined inMichot et al. (1984,1990). The D1-D3 regionhas been shown to contain most of the phy-logenetically informative sites in this por-tion of the nLSU gene (Hillis and Dixon,1991; Kuzoff et al., 1998; Hopple and Vil-galys, 1999). Sequences were produced by avariety of different enzymatic sequencingstrategies over a period of nearly 10 years,including both manual and automatedmethods (see Appendix for details of themolecular techniques used to produce pre-viously published sequences). StandardDNA isolation procedures were used witheither CTAB (hexadecyltrimethyl-ammo-nium bromide; Zolan and Pukkila, 1986) orsodium dodecyl sulfate (SDS) buffers (Leeand Taylor, 1990). Polymerase chain reac-tion (PCR) amplifications were as describedby Vilgalys and Hester (1990). PCR andsequencing reactions used the primers5.8SR(5'-TCGATGAAGAACGCAGCG-3'),LROR (5'-ACCCGCTGAACTTAAGC-3'),LR3R (5'-GTCTTGAAACACGGACC-3'),LR5 (5'-TCCTGAGGGAAACTTCG-3'), LR7(5'-TACTACCACCAAGATCT-3'), and LR16(5'-TTCCACCCAAACACTCG-3'); additional

information about these primers is givenat http://www.botany.duke.edu/fungi/mycolab/primers.htm. Most of the se-quences were produced with the use of au-tomated sequencers (model ABI373 orABI377; Perkin-Elmer) and dye terminatorsequencing chemistries (Perkin-Elmer/ABI), according to the manufacturer's in-structions. Sequence chromatograms gener-ated by the automated sequencers werecompiled by using Sequencher softwareversion 2.0 (Gene Codes Corp.).

Phylogenetic Analyses

Sequences were aligned manually by us-ing the editor window of PAUP* (Swofford,1998). DNA regions with ambiguous align-ment were initially removed from theanalyses, and regions with distinctive pat-terns of length variation were recoded toprovide additional information for parsi-mony analysis through the use of severalstrategies (Bruns et al., 1992; Hibbett et al.,1995; Moncalvo et al., 1995): single gapsthat aligned unambiguously were scored asa "fifth character state", gaps forminglarger indels were scored as "missing" ex-cept for one position scored as "fifth state"(to score an indel represented by severalcontiguous gaps as a single evolutionaryevent), and all ambiguously aligned gaps innonexcluded regions were treated as "miss-ing". Phylogenetic analyses were conductedby using MP with changes among characterstates having equal weights ("equallyweighted parsimony"; UP) or differentweights ("weighted parsimony"; WP), max-imum likelihood (ML), and distance meth-ods (NJ), as performed in PAUP* (Swofford,1998) with a Power Macintosh 8600/300MHz or a UNIX Sun Sparc 20 Station withdual 150 MHz hypersparc processors.

Equally weighted parsimony.—The follow-ing heuristic search settings in PAUP* wereused: steepest descent option not in effect,branches allowed to collapse (creatingpolytomies) if maximum branch lengthequals 0, multistate taxa interpreted as un-certainty using polymorphism coding inthe formal IUPAC codes (which PAUP* rec-ognizes in "datatype = DNA"), and allcharacters of type unordered. Heuristicsearches are subject to local optima (tree is-

Dow




lands) and may not always find the shortesttrees (Maddison, 1991; Stewart, 1993; Swof-ford et al., 1996); in addition, many searchesfrom our data matrix could not be com-pleted with MAXTREES unlimited andtree-bisection-reconnection (TBR) branchswapping. We therefore used several strate-gies to find most-parsimonious trees. Weconducted parsimony searches with TBRbranch swapping, using starting trees ob-tained by way of (1) simple addition se-quences with MULPARS "on" and MAX-TREES unlimited; (2) random additionsequences along with different MAXTREESsettings, including searches with MUL-PARS off; or (3) multiple random trees(Maddison et al., 1992) with low MAX-TREES (typically 1-100). Several sets oftrees obtained with low MAXTREES set-tings were used as starting trees in recur-sive searches (Olmstead et al., 1993; Olm-stead and Palmer, 1994) in which wealternated swapping algorithms betweensearches (TBR, subtree pruning-regrafting[SPG], and nearest neighbor interchanges[NNI]) along with increased MAXTREES(typically 1,000-5,000). If there was im-provement in tree scores, the resulting besttrees were saved and then used as startingtrees for TBR branch swapping with MAX-TREES set to higher values (from >10,000 tounlimited). Though many searches couldnot be completed, more than a year of CPUtime was logged during the course of theseanalyses. Finally, the strict consensus tree ofthe most-parsimonious trees obtained inUP searches was used as a constraint for asearch of 500 replicates of random additionsequences with TBR branch swapping andMAXTREES set to 5, and PAUP was in-structed to save only trees that did notmatch the constraint; the resulting besttrees were used as starting trees for TBRbranch swapping with unlimited MAX-TREES and the same constraint and also ina search where the constraint was removed.This search strategy was developed to sam-ple a large neighborhood of equal-lengthtrees without exhaustive searching (Catalanet al., 1997; Rice et al., 1997).

Support for phylogenetic groups was es-timated by bootstrapping (Felsenstein,1985) and parsimony jackknifing (Farriset al., 1996) with 50% character deletion,

with inclusion of parsimony-uninformativecharacters during resampling. Bootstrapand jackknife analyses were conductedwith 100 replications of random additionsequences, with the use of "fast" proce-dures (no branch swapping and MULPARSoff) and more-optimized procedures withTBR branch swapping and MAXTREES setto 100. Groups compatible with 50% major-ity-rule consensus were retained in the re-sulting consensus trees.

Weighted parsimony.—Weighted parsi-mony is often used to compensate for un-equal base frequencies, transition/trans ver-sion biases, or rate heterogeneity insequence data sets (Albert and Mishler,1992; Allard and Carpenter, 1996; Yoder etal., 1996; Cunningham, 1997). To avoid arbi-trary weighting of nucleotide substitutionratios and circularity in estimating these ra-tios from a given tree, we developedweightings based on nucleotide frequenciesand substitution biases through pairwisecomparison of sequences in the data matrix,using the option "pairwise base difference"in PAUP*. Base differences were estimatedfor all possible pairwise combinations ofthe 154 taxa in the data matrix, and weightsfor each substitution type were averagedacross all comparisons and scored as a step-matrix, as suggested by Maddison andMaddison (1992:60; when a substitutiontype was not observed between two taxa,that substitution type was arbitrarily scoredas 0.5 in the calculation of the In value; gapsscored as fifth character states were givenhigher weights to account for their rarity).The WP analysis used simple addition se-quence with TBR branch swapping andMAXTREES unlimited. Additional searcheswere conducted by using trees found in UPand ML searches as starting trees.

Maximum likelihood.—A variety of in-creasingly complex models of molecularevolution were evaluated by using likeli-hood ratio tests (LRT; Goldman, 1993;Huelsenbeck and Rannala, 1997) to identifya simple and robust substitution model forthe nLSU-rDNA data set. LRT were per-formed by using trees obtained from MPand NJ methods.

The large size of our data matrix made itimpractical for PAUP* to use starting treesgenerated from sequence additions in ML

Dow




analyses. Therefore, starting trees for branchswapping were selected from among earlierMP and NJ analyses by choosing as startingtrees those with the best ML score. ML pa-rameter settings were reoptimized for eachstarting tree and later held constant duringbranch swapping to find more optimal trees(invariable sites were included in all MLanalyses). Most branch swapping with MLused NNI because it is the least computa-tionally intensive swapping algorithm avail-able in PAUP*; subsequent searches usingTBR swapping had to be aborted beforecompletion (on a Power Macintosh Com-puter with 300 MHz processor, TBR searcheson a single tree with 154 taxa were estimatedto take as long as 3 or 4 months to complete).

Distance methods.—NJ trees (Saitou andNei, 1987) were estimated under a varietyof distance measures, based on a priori as-sumptions about rates and modes of nu-cleotide substitutions (distances describedin Swofford et al., 1996, and estimated inPAUP*), including uncorrected distance"p", LogDet/paralinear distance (Lockhartet al., 1994), Jukes-Cantor (JC) distances(Jukes and Cantor, 1969), Hasegawa-Kishino-Yano (HKY85) distance (Hase-gawa et al., 1985), Kimura two-parameter(K2) distance (Kimura, 1980), and Tajima-Nei (TN) distance (Tajima and Nei, 1984).Several distance measures incorporatingrate heterogeneity were also included, us-ing a gamma distribution (shape parameterarbitrarily assigned as 0.5) based on JC,HKY85, K2, and TN distances, as well asthe general-time-reversible model (GTR)with nucleotide frequencies estimated fromthe data, as many as six categories of nu-cleotide substitution types with substitu-tion rate parameters estimated by ML, andas many as eight rate categories with ratesassumed to follow a gamma distributionwith shape parameter 0.5. Ties encounteredduring tree reconstruction were set to breaksystematically (taxon-order dependent).

Estimation of topological differences betweentrees.—Statistical significance of topologicaldifferences among trees was estimated byusing MP as the optimality criterion, withthe Templeton (1983) and winning-sitestests implemented by PAUP*. Significanttopological differences under ML crite-ria were estimated from the Kishino-

Hasegawa test (Kishino and Hasegawa,1989).

To compare tree lengths and topologiesunder alternative phylogenetic hypotheses,we also searched for trees by using phylo-genetic constraints (Swofford, 1991). Con-straint searches used UP with simple addi-tion sequence, TBR branch swapping, andMAXTREES set to 10. The resulting treeswere evaluated against the best MP trees.

RESULTS

Sequence Alignment andNucleotide Sequence Variation

An alignment of 154 nLSU-rDNA se-quences was 1,190 positions long. Overall,there was excellent agreement in sequencealignment between sequences producedmanually and by computer, with all dis-crepancies in alignment occurring withinhypervariable regions. After removal of theextreme 5'and 3'positions, which were in-complete for several taxa, 934 positions re-mained in the analyses. Alignment over abroad taxonomic sampling was not attain-able within three hypervariable, indel-richregions, and these regions were also re-moved from the analyses: positions 126-138in variable domain Dl, and 475-492 and635-686 in variable domain D2 (however,subsets of related taxa aligned nicely withinthese divergent domains, suggesting thatthese regions have phylogenetic informa-tion at lower taxonomic levels). A few re-maining single-gap regions were removedbefore sequence analysis, because of thepossibility that they represented sequenc-ing errors, as suggested from their occur-rence in only one or a very few sequences.In contrast, several gap regions with shortindels could be recoded as phylogeneticallyinformative characters (mostly by scoringsingle gaps as a "fifth character state"). Inall, 140 positions corresponding to regionswith problematic alignments were re-moved, and 35 indel positions were re-coded. Of the 826 characters included in theanalyses, 422 characters were constant, 93variable characters were parsimony-unin-formative, and 311 characters were parsi-mony-informative. The final alignment for154 taxa has been deposited on the internetas a NEXUS file, together with more de-

Dow




tailed explanations about the recodingstrategies used, at http://www.botany.duke.edu/fungi/mycolab/agarical.htm.

Data Analysis

Equally weighted parsimony.—Results ofUP searches using simple addition se-quences, TBR branch swapping, and MAX-TREES unlimited yielded 60 equally parsi-monious trees of length 3,100 (consistencyindex [CI] = 0.207, retention index [RI] =0.555). Searches using random addition se-quences with TBR and MAXTREES unlim-ited were aborted during the sixth replicate(and >300 hr of CPU time), and yieldedtrees of length 3,102-3,112. In subsequentsearches MAXTREES settings were reducedto allow searches to continue to completionwithin a reasonable amount of time. Twosearches using 100 replicates of random-addition sequences with MAXTREES set to100 and 10 were completed within 71 and20 hr, respectively; both searches producedtrees 3,101 steps long, which were poten-tially located in three different tree islands(Maddison, 1991).

Another search strategy was attemptedwith MAXTREES set to 1 (= MULPARS off)and TBR branch swapping. Searches begin-ning with either 100 or 500 replicates of ran-dom addition sequences yielded slightlylonger trees than previous UP analyses(lengths 3,104-3,121 and 3,102-3,130, re-spectively); searches starting from randomtrees (100 replicates) yielded trees of length3,103-3,120, which were also very similar inlength to trees produced by searches usingrandom addition sequences. In contrast,when no branch swapping was performed,trees produced from random addition se-quences were at least 43 steps longer thanthe shortest trees even though the numberof replicates used was high (10,000).

Recursive swapping with unlimitedMAXTREES and using starting trees fromone of the three islands of score 3,101yielded 72 trees of length 3,100. These 72trees were topologically different from thefirst 60 MP trees (of equal length), accord-ing to comparisons of symmetric tree dis-tances (data not shown), and therefore be-long to a different tree island (Maddison,1991). We also eventually found a third tree

island consisting of 180 trees of length 3,100when trees produced by WP searches (seebelow) were used as starting trees for TBRbranch swapping in a UP search. No othersets of trees of length 3,100 or shorter werefound, although numerous sets of treeswere used as starting trees for recursiveswapping. We subsequently refer to thethree equally parsimonious tree islandsfound by using UP as UP60, UP72, andUP180 (referring to the number of equallyparsimonious trees present in each island).

Symmetric differences among trees fromthese islands were calculated in PAUP*.Most differences occurred between treesfrom island UP60 and those from UP72 orUP180. Little average symmetric differ-ences were found between trees from the is-lands UP72 and UP180. Within a single treeisland, strict consensus trees were relativelywell resolved. The strict consensus treefrom the 252 trees of the combined islandsUP72 and UP180 was also well resolved(Fig. la), which reflected lower averagesymmetric differences between trees fromthese two islands. In contrast, a strict con-sensus tree containing trees from all threeislands contains many collapsed (especiallydeep) branches (Fig. lb), largely because ofdifferences between UP60 and the othertwo islands. When the strict consensus treefrom all three islands was used as a nega-tive constraint in a search with 500 randomaddition sequences, TBR branch swapping,and MAXTREES set to 5, the shortest treesproduced were of length 3,103; when thesetrees were used as starting trees with TBRbranch swapping and MAXTREES unlim-ited, 2,700 trees of length 3,101 (i.e., onestep longer than the best trees in UP60,UP72, and UP180) were produced insearches that either kept, or removed, theconstraint. The latter results provide a highdegree of confidence (Catalan et al., 1997)that islands UP60, UP72, and UP180 repre-sent sets of shortest trees.

Both bootstrapping and jackknifing pro-vided good support for many terminalclades as well as several internal nodes(Figs. 2 and 3a). As expected for statisticallyrelated procedures, both bootstrap andjackknife values are strongly correlated(Fig. 3a; Pearson R2 = 0.98). Both bootstrapand jackknife trees showed nearly identical

Dow



284 SYSTEMATIC BIOLOGY VOL. 4 9

(a)

FIGURE 1. Tree topologies for (a) the strict consensus of 252 trees for two tree islands UP72 and UP180 and (b)the strict consensus of 312 trees for all three tree islands UP72, UP180, and UP60. Identities of groups A to BB aregiven in Figures 2 and 5. Groups that are paraphyletic in Figure 2 or 5 are labeled in quotation marks.

Dow




Cotlybia polyphyllaCollybia dryoplula

Collybia maculata• Marasmiellis ramealis

I Lentimla eclodesI Micranpliale perforans

1 Omplialotus nidiformisNotliopanus eugrammus

60 ^ ^ ^ ^ ^ ™ Marasmius capillaris

A

Marasmius deleclaisCrinipellismaxima

Gernnema strcmbodes

"attim fungus" G2.U 11905 — '

Cyptotrama aspraia

I Marasmius pyrroceplialus1 Slrobilurus mdlisitus

I Flammulina veluipes1 Rliodolus paimatusI Xendafiufuracea

I Armilkria tabescemI Baeospora myriadophylla , _

1 Hydropus scabripes 1 E' Pleurotopsis laiginqua

D

Amanita rubesceinAmanitajlavocoiiia

Amanita citrimAmanita peckitna

I Pleurotus djamo• Pleiirotus purpurenolivaceus

I Pltyllaopsis nidulansI Pleurocybellaporrigens

I Resipinatus sp.I Resupinatus alboniger

Omplialina py.xidataClitccybe lateritia

Caulorliiza hygroplwroidesCancliomycesbursaefcrmis

99 Hygroplionis sorditlusHygroplwrus biJiereisis

Chrysamphalina clwysnphylla

Chrysanplialim grossulaA rrhenia auriscatpitm

Phaeaellis griseipallidisOmptialim veluipes

Ompluiina luteovitelliiuOmplialinu velnina

K

I Clitocybe dealbataI Hypsizygus ulmarius

i Clitocvbe nuda

M

Triclolnma atravidaceim — i\Triclolcma pardinum

Leucopaxillm albissimmI Triclnloma imbrication

I Triclnloma caligalwnI Triclnloma porteitasum

I Triclnloma subaumm —I Lyopliyllun semitate

N

Lyopliyllun atratunLyopliyllun decastes

Tennitanyces cyliidricieTermitnnyceslieimii

PodabrellamicrccarpaMycem nailauliiformis I

Mycem galericilalu pResiwmycemacadiensis

M\cem clavicUaris 1

o

FIGURE 2. Agaricales phylogeny based on MP analysis of large nuclear subunit rDNA sequences. The figureshows one most-parsimonious tree found in both equally weighted parsimony (UP) and weighted parsimony(WP) searches. Bold lines indicate branches present in the strict consensus of all most-parsimonious UP and WPtrees. UP searches yielded 312 equally parsimonious trees of length 3,100 (CI = 0.207, RI = 0.555) located in three

Dow




Clitipilie pnmulusCystoderma granulosum

Callistosporium luteoolivaceitnCortinarius iodes

Cortinarius marylandensisLaccariabicolor

Cortinariiis sp.Bdbitiie vitelliius

ConocyberickeriiPanaedimfoenisecii

Panaeolus acuminatusAnellaria semiovata

Inocvbe geophylla var. lilaceaInocybesp.

Coprinus bisporusPsatltyrellagracilis

PsathyrellacandolleanaCoprinus atramentarius

Coprinis cinereusCoprinus kimwae

PsatmrelladelineataCoprinus nudiceps

Lacrymaria veltfina

Cystolepidata cystidosaCoprinis sterquilinis

Coprinus comatusattiti: fungus" GIACU11893

attinefimgus" GI.SBU 11902Lepiotacristaa

Macrdepiota racliodesAgaricus bisporus

Agaricus pocilktororophyllun molybdites

LeucoagariciB rubrolinctusicrolepiotaproceraLeucocoprinus cepaesiipss

LeucocopritnisfragilissimisLeucoagariciB naucinus

"attire fungus" G3.CFU 11895Upiota acutesifiamosa

Hebelcma crustuliniforme 1 _ .™ Agrocybe praecax ' VPsilccybestuntzii

KuelmeromycesmutabilisI Psiltcybe sylvctica

Stropltaria rugosoaiuiulataPMiota squarrosoides

Hypholoma sublateritiunHvplioloma subviride —

100 Omplialina rosellaRickerellamellea

RickerellapseudogrisellaBondarzewia meserterica

unknown hasidiomyceteRussula earlei

RussulamaireiRussula viresceis

Russula romagnesiiLactarius corrugis

Lactarius piperatusLactarius volemus

Phyllcporus rlwdoxamhis

Boletus retipes

Suilhe luteusCanoderma lucidum

Canoderma australe

IAA

BB (OUTGROUP)

FIGURE 2. (continued) different tree islands. WP searches yielded four equally parsimonious trees of score12,655.9 (CI = 0.220, RI = 0.563). Branch lengths were estimated by using ACCTRAN with character-state changeshaving equal weight. Bootstrap values S;50% are shown above each branch (bootstrap values based on 100 repli-cates with MAXTREE set to 100 in each replicate, random addition sequence, TBR branch swapping, and charac-ter-state changes having equal weight).

topologies at nodes supported at >50% con-fidence level, though below this value somebranches that are present in one tree may beabsent in the other (data not shown). Sev-eral computationally faster alternative pro-cedures for calculating bootstrap and jack-knife values were also implemented inPAUP*, including "fast" bootstrapping and"fastjac" jackknifing, which yielded verysimilar topologies to those of the more opti-

mized (and more computationally inten-sive) search strategies, albeit with slightlylower support values for most nodes (Fig.3b; results are shown only for fast bootstrapvs. normal bootstrap).

The Templeton and winning-sites testswere used to evaluate trees obtained withdifferent search strategies. All searches thatused TBR branch swapping were able tofind trees within 30 steps of the most-parsi-

Dow




100

0 10 20 30 40 50 60 70 80 90 100bootstrap

uu -90 -80 :

70 -60 -50 -40 -30 -20 -10 -

0 -

a

.. v^v •oo.Q

0 10 20 30 40 50 60 70 80 90 100

bootstrap

FIGURE 3. Comparison between bootstrap and jack-knife measures of support (based on 100 replicates), (a)Pairwise comparison between bootstrap and jackknifewith 50% character deletion, TBR branch swapping,random addition sequence, and MAXTREE set to 100in each replicate (Pearson R2 = 0.94). (b) Pairwise com-parison between "fast" bootstrap (no branch swap-ping) versus bootstrap with TBR branch swapping,random addition sequence, and MAXTREE set to 100in each replicate (Pearson R2 = 0.93).

monious trees, and nearly all of these trees(>99.9 %) were not markedly different fromthe most-parsimonious trees. Heuristicsearches performed without swapping,however, always yielded trees that were atleast 43 steps longer and always signifi-cantly worse than the best tree (P < 0.02 inthe Templeton test). The efficiency of eachsearch strategy can also be gauged by ex-amining the computer time required (usingeither UNIX or PowerMac computer). WithMAXTREES set at 10 or higher, both com-puters were able to find trees within 2 stepsof the most-parsimonious tree after TBRswapping—but only after considerabletime (at least 19 hr of CPU time). In con-trast, searches storing a single tree (MAX-TREES = 1 or MULPARS off) at each step ofTBR swapping ran very quickly on a Power

Macintosh computer (as little as 7 hr for 100replicates) and quickly converged toward aset of near most-parsimonious trees thatwere not statistically different from thethree islands of most-parsimonious trees.

Weighted parsimony.—A stepmatrix basedon dinucleotide frequencies (Fig. 4a) wasused to perform WP analyses. Nine equallyparsimonious trees of length 12,670.4 wereproduced from a search using simple addi-tion sequences. When these trees were eval-uated under the UP criterion, the length ofall nine trees was 3,106 steps (six stepslonger than the best UP trees). Both Temple-ton and winning-sites tests found no signif-icant differences between UP and WP treesunder either UP (P = 0.565-0.834, Temple-ton test, and P = 0.460-1.000, winning-sites

(a)

[A][C][G][T]

[A]1.45.7

3.85.16.7

[C]5.7

1.75.93.5

6.9

[G]3.85.91.35.26.9

[T]5.13.55.21.56.2

6.76.96.96.22.7

(b)[A][C][G][T]

[A] [C] [G] [T]

- 6 . 7 6 5 0 .919 4 .613 1.2290 .919 - 1 0 . 3 4 2 0 . 6 5 1 8 .7724 . 6 1 3 0 . 6 5 1 - 6 . 2 6 4 1.0001.229 8 .772 1.000 - 1 1 . 0 0 0

(c)

StepmatrixFIGURE 4. Nucleotide substitution matrices used

for WP and ML analyses, (a) Stepmatrix based ondinucleotide frequencies used for WP analysis (seetext for treatment of gaps: [-]). (b) Substitution ratematrix estimated from ML with a GTR-6 model ofevolution, (c) Matrix correlation plot for (a) and (b)above (Pearson R2 = 0.98).

Dow




test) or WP models (P = 0.273- 1.000, Tem-pleton test, and P = 0.526-1.000, winning-sites test).

Under the WP model, 17 of the UP60 andUP72 trees were shorter than the best WPtree obtained with simple addition se-quences (12,670.4 steps in length), and thefour equally shortest trees from this set(12,669.3 steps) were used as starting treesfor TBR branch swapping in a new WPsearch. That search found four slightlyshorter trees (12,655.9 steps), which scored3,100 under the UP criterion (same as thebest UP trees). These trees eventuallyyielded the tree island UP180 (see above).

One of the four best WP trees (length12,655.9 steps) revealed by these analyses isshown in Figure 2. This tree is also one ofthe 312 best UP trees (length 3,100, locatedin island UP180). Because it is supported byboth UP and WP analyses, this tree repre-sents our best hypothesis of Agaricalesphylogeny based on parsimony (MP tree).

Distance methods.—Eleven NJ trees wereestimated from distance matrices calculatedby using various evolutionary models.Though many topological differences wereobserved between these trees, many termi-nal groups were generally similar and cor-respond well with clades also found in par-simony analyses (data not shown). Whenscored under the UP criterion, however, allNJ trees (length = 3,166-3,179 steps) weresignificantly worse than the most-parsimo-nious trees under both the Templeton (P <0.001) and winning-sites tests (P < 0.028).

Maximum likelihood.—Table 1 outlines theprocedure followed in choosing the mostadequate ML model for this data set. Forevery evolutionary model investigated, thetoken UP tree always had higher likelihoodscores than the NJ-JC tree (Table 1). LRTbased on either tree suggest the use of asimilar model of evolution with five cate-gories for base-substitution types andseven rate categories for the variable sitesdistributed according to a gamma-parame-ter (Table 1). In evaluating model complex-ity, the greatest improvements in ML scoresin the LRT were obtained by increasing thenumber of substitution types from 1 to 2 toaccount for transition/trans version bias(6.8% improvement), and in taking into ac-count among-site rate variation (17% im-provement: 10% improvement after distin-

guishing variable vs. invariable sites, andan additional 7% improvement when vari-able sites were allowed to vary according toa gamma distribution). ML estimates ofsubstitution rates for a model with differentrate categories for each substitution type(GTR-6 model; Yang, 1994) are shown inFigure 4b. The smallest difference in substi-tution rates was between A-C and G-T sub-stitutions, and combining these two kindsof substitution into a single rate categorydid not significantly affect ML scores (Table1); therefore, a GTR-5 model of evolutionwas chosen for subsequent ML analyses.

ML estimates of substitution rates foreach substitution type (Fig. 4b) were com-pared with the weights assigned to thesame substitution types for WP (deter-mined from dinucleotide frequencies) (Fig.4a). Both methods gave estimates that wereinversely correlated (Pearson R2 = 0.98),with the lowest ML estimate of substitutionrate (G-C) having the most weight in WPanalyses (Fig. 4c).

Likelihood scores were calculated fortrees from all of the previous UP, WP, andNJ searches by using likelihood parametersoptimized for either a UP or NJ model tree.The tree with the highest likelihood scorewas also one of the nine best trees obtainedfrom WP with simple addition sequence(ML-WP tree). The likelihood parameterswere reoptimized to fit the ML-WP tree un-der the GTR-5 model of evolution, resultingin the following parameters: assumed nu-cleotide frequencies A = 0.270, C = 0.194,G = 0.298, and T = 0.238; substitution-ratematrix with G-C substitutions = 0.677, A-Cand G-T = 1, A-T = 1.275, A-G = 4.792, andC-T = 9.110; proportion of sites assumed tobe invariable = 0.455; rates for variable sitesassumed to follow a gamma distributionwith shape parameter = 0.749, and numberof rate categories = 7. With these settings,the ML-WP tree was used as a starting treefor NNI branch swapping. This search wascompleted within 2 days of computationand resulted in a single tree of score15,689.295, which was only 0.15% betterthan the input tree. This tree was used as astarting tree for TBR branch swapping. Af-ter 1,009 hr of CPU time on a UNIX SunSparc 20 Station (150 MHz), 132,748 re-arrangements were performed, and the MLtree still had the same score. This best ML

Dow



TABL

E 1.

A

lter

nati

ve e

volu

tion

ary

mod

els

used

to e

stim

ate

likel

ihoo

ds s

core

s (-

In L

) and

like

liho

od ra

tio

test

s fo

r on

e of

the

mos

t-pa

rsim

onio

us tr

ee o

f th

eun

wei

ghte

d pa

rsim

ony

anal

ysis

wit

h si

mpl

e ad

diti

on s

eque

nce

(tre

e U

Pl)

and

for

a n

eigh

bor-

join

ing

tree

bas

ed o

n Ju

kes-

Can

tor

dist

ance

s (t

ree

NJ-

JC).

Dat

afo

r th

e be

st-f

it m

odel

are

giv

en in

bol

d.

Evo

luti

onm

odel

a

JC F81

K2

P

HK

Y85

GT

R-3

GT

R-4

GT

R-5

GT

R-6

GT

R-5

GT

R-5

GT

R-5

GT

R-5

GT

R-5

GT

R-5

GT

R-5

GT

R-5

No.

of

subs

titu

tion

type

s

1 I-H 2 2 3 4 5 6 5 5 5 5 5 5 5 5

Ass

umed

base

freq

uenc

ies

Equ

alE

stim

ated

d

Equ

alE

stim

ated

d

Equ

alE

qual

Equ

alE

qual

Equ

alE

qual

Equ

alE

stim

ated

"1

Est

imat

edd

Est

imat

ed"1

Est

imat

ed"1

Est

imat

ed"1

Am

ong-

site

s ra

te v

aria

tion

Inva

riab

lesi

tes

Est

imat

ede

0 Est

imat

ede

Setf

Setf

Setf

Set

f

Setf

Rat

es o

f va

riab

lesi

tesb

Equ

alE

stim

ated

e

Est

imat

ed6

Setf

Setf

Setf

Set

f

Setf

No.

of

rate

cate

gori

es

4 4 4 5 6 7 8

Tre

e U

Pl

-ln

L

19,7

30.5

2219

,797

.629

18,5

41.0

5218

,649

.802

18,4

25.1

3418

,416

.157

18,4

09.8

7318

,409

.584

16,7

15.0

6515

,865

.837

15,7

72.3

5315

,766

.148

15,7

56.6

0715

,751

.878

15,7

49.3

3615

,747

.874

Pc - <0.0

01- <0

.001

<0.0

01<0

.001

n.s.

<0.0

01<0

.001

<0.0

1<0

.01

<0.0

01<0

.01

<0.0

5n.

s.

Tree

NJ-J

C

-lnL

19,9

86.0

75

20,0

59.0

49

18,7

60.6

84

18,8

75.1

48

18,6

33.7

99

18,6

25.1

74

18,6

16.8

92

18,6

15.6

95

16,8

66.6

60

15,9

61.6

86

15,8

65.4

62

15,8

57.7

15

15,8

47.7

04

15,8

43.0

74

15,8

40.5

46

15,8

38.9

87

Pc -

<0.

001

-

<0.

001

<0.

001

<0.

001

n.s.

<0.

001

<0.

001

<0.

05

<0.

01

<0.

001

<0.

01

<0.

05

n.s.

a Lis

ted

in o

rder

of

incr

easi

ng c

ompl

exity

. GT

R-3

= tw

o ca

tego

ries

for

tra

nsiti

ons

and

one

for

tran

sver

sion

s; G

TR

-4 =

two

cate

gori

es f

or t

rans

itio

ns a

nd t

wo

for

tran

sver

sion

s: (A

-T),

(A-C

, G

-T, G

-C);

GT

R-5

= tw

o ca

tego

ries

for

tra

nsit

ions

and

thr

ee f

or t

rans

vers

ions

: (A

-T),

(G-C

), (A

-C,

G-T

); G

TR

-6 =

two

cate

gori

es f

or t

rans

itio

ns a

nd f

our

for

tran

sver

sion

s.b V

aria

ble

site

s as

sum

ed t

o fo

llow

a g

amm

a di

stri

buti

on w

ith

a ga

mm

a sh

ape

para

met

er.

c Com

pare

d w

ith

the

best

val

ue p

revi

ousl

y ca

lcul

ated

, lik

elih

ood

was

eith

er s

igni

fica

ntly

im

prov

ed, n

ot s

igni

fica

ntly

im

prov

ed (

n.s.

), o

r w

orse

(-)

.d E

stim

ated

em

piri

call

y fr

om t

he d

ata.

e Lik

elih

ood

esti

mat

e fr

om t

he d

ata.

f Set a

s es

tim

ated

abo

ve.

Dow




tree is shown in Figure 5. When scored un-der the UP criterion, this ML tree is 3,128steps in length, which is not significantlydifferent from the UP trees (3,100 steps) inboth the Templeton (P = 0.167) and winning-sites (P = 0.262) tests. Reciprocally, treesfound in the most-parsimonious islandsUP60, UP72, UP180 and in the WP analysiswere also not significantly different fromthe ML tree under the Kishino-Hasegawatest (scores 15,717.375-15,755.618, P =0.358-0.065). In contrast, NJ trees were al-ways significantly worse than the ML tree(P < 0.010, Kishino-Hasegawa test).

Comparison between Topologies ofMP,ML, and Distance Trees

Results of the Templeton, winning-sites,and Kishino-Hasegawa tests (see above) in-dicated that trees produced by UP, WP, andML analyses were all highly congruent,whereas unoptimized distance trees wereall significantly worse than the best UP, WP,and ML trees. We therefore consider the op-timal MP and ML trees (Figs. 2 and 5) torepresent our best current estimates ofAgaricales phylogeny. To facilitate discus-sion, the groups in Figures 1, 2, and 5 havebeen labeled A to BB. We use quotationmarks to refer to groups that were para-phyletic in Figure 2 or 5 (e.g., group "B").

DISCUSSION

Phylogenetic Relationships in the Agaricales

The phylogeny of the Agaricales is stilla controversial field. Singer (1986:124)

Progress toward a phylogenetic systemof classification for the agaricoid fungi hasbeen slower than for other organisms andis still hampered by many factors, mostnotably our still-too-poor knowledge ofundiscovered taxa, especially in the tropics.This study represents a first comprehensiveattempt to analyze phylogenetic relation-ships within the Agaricales, including exem-plars from 16 of the 18 families recognizedby Singer (1986). Our results demonstratethat nLSU-rDNA sequences provide suit-able resolution for identifying major lin-eages of agaric fungi, with good support formany terminal clades and many internalbranches (Fig. 2). The evolutionary lineages

identified from this study (see Figs. 2 and 5)may also be considered as a starting pointfor defining phylogenetic taxa. In most in-stances, phylogenetic groupings recognizedby our analysis already correspond inwhole or part to existing taxonomic groups(e.g., Amanitaceae and Russulaceae). Ofcourse, results of this study also suggestnew relationships among certain groups, orotherwise suggest that some traditional tax-onomic views are in need of revision.

Our two best estimates of phylogeneticrelationships (Figs. 2 and 5) are largely con-sistent with Singer's system (1986). Bothparsimony and likelihood trees supportSinger's division of the Agaricales intothree major lineages, corresponding to thesuborders Agaricineae, Boletineae, andRussulineae. Relationships among thesethree groups are not completely resolved,however, because MP trees (Fig. 2) suggestsister-group relationships between the Bo-letineae and the Russulineae lineages(group X-AA), whereas the ML analysisplaces the Boletineae lineage basal to theAgaricineae (Fig. 5). The ML tree (Fig. 5)agrees with a recently published MP tree ofhomobasidiomycetes that is based on 18Snuclear and 12S mitochondrial rDNA se-quences (Hibbett et al, 1997) and supportsa sister-group relationship between bole-toid and agaricoid lineages. That study wasbased on analysis that would be expected toresolve phylogenetic groupings at higherlevels (i.e., at deeper levels of branching)than the faster-evolving nLSU-rDNA gene.Combined analysis using data from a moreslowly evolving gene (18S RNA) along withour nLSU-rDNA should offer better resolu-tion of the basal relationships in this agaric-bolete-russuloid triangle.

The family Russulaceae (group Z), con-taining Russula and Lactarius, forms thecore group of suborder Russulineae (Singer,1986) and represents a distinct clade with100% bootstrap support (Figs. 2 and 5).Phylogenetic analysis also supports group-ing of the Russulaceae into a larger cladethat includes a polypore (Bondarzewia), agroup of small omphalinoid agarics {Rick-enella spp., placed by Singer in the Agari-cineae), and an unidentified basidiomycete(group X-Z). This "russuloid" clade (Hib-bett and Thorn, in press) is supported byboth MP (61% bootstrap support) and ML

Dow




FIGURE 5. Agaricales phylogeny estimated by using ML. Tree score = 15,689.295. Groups labeled A to BB arethe same as in Figures 1 and 2.

Dow




analyses (MP and ML trees differ in thebranching order within the Russulaceae-Bondarzewia-Rickenella clade). Bondarzewiahas been proposed to be closely related tothe Russulaceae on the basis of basidio-spore morphology and the presence of lati-ciferous hyphae (Redhead and Norvell,1993; Singer, 1986). The grouping of severalRickenella species (group X) with other rus-suloid fungi is surprising from a morpho-logical standpoint, because Rickenella doesnot possess any distinctive characters (amy-loid ornamented spores, heteromerous trama,spherocysts, laticifers) that might show evi-dence of recent coancestry. Because of theirsimple agaric morphology, Rickenella spp.instead have traditionally been classified inthe Agaricineae (family Tricholomataceae),"doubtfully distinct from Mycena" accord-ing to Kiihner (1980:730), and as a synonymof Gerronema in Singer (1986).

The boletes and their relatives (groupAA) include mostly poroid {Boletus andSuillus) as well as gilled (Phylloporus) mush-rooms, which Singer placed in the Bo-letineae (Figs. 2 and 5). Although our sam-pling of boletoid fungi for this study waslimited, the Boletineae (represented by thethree genera named) are supported as amonophyletic group with 100% bootstrapsupport in all analyses. Monophyly of theBoletaceae and their close agaricoid rela-tives has been demonstrated in several pre-vious molecular phylogenetic studies (Brunsand Szaro, 1992; Bruns et al, 1992, 1998;Hibbett et al., 1997). Our results do not sup-port a close evolutionary relationship be-tween Omphalotus (an agaric, group "B")with other boletes as once suggested (Bre-sinsky and Besl, 1979; see also discussion ofOmphalotus below).

The majority of taxa sampled for thisstudy belong to the Agaricineae sensuSinger (1986), which includes most speciesof agaric fungi. Except for the Rickenellaclade (group X), monophyly of the Agari-cineae (groups A-W) is supported by allanalyses (including bootstrap and jackknifeconsensus trees). The node for monophylyof the Agaricineae, however, is not sup-ported beyond 30% in either bootstrap orjackknife analyses (data not shown). Eventhough the Agaricineae are the largest andbest known group of mushroom taxa, weare not aware of any obvious morphologi-

cal synapomorphies for this group, becausemost of the traditional characters used todefine these fungi (lamellae, fleshy fruitbodies, etc.) are known to have evolvedmultiple times in other groups of fungi(Hibbett et al., 1997). Agreement support-ing the monophyly of the Agaricineae be-tween this study and earlier ones suggeststhat additional characters may be foundthat unambiguously define this lineage.

The polypores and their relatives havebeen suggested at different times to beclosely related with agarics. Singer (1986)even included several genera of gilled agar-ics together with the genus Polyporus in thefamily Polyporaceae on the basis of similar-ities in dimitic hyphal structure. Althoughthe outgroup chosen for this study (Gano-derma) appears to be a good exemplar fromthe Polyporaceae, based on molecular evi-dence (Hibbett and Vilgalys, 1993; Hibbettand Donoghue, 1995), our results supportthe view that several taxa placed by Singerin the Polyporaceae belong instead to theAgaricineae, including Lentinula (group A),Nothopanus (group "B"), Phyllotopsis (groupI), and Pleurotus (group H).

Phylogeny within the Agaricineae

Within the Agaricineae, many terminaland higher-level phylogenetic groupingsare supported by the nLSU-rDNA data. Thephylogenetic hypothesis based on MP inFigure 2 divides the Agaricineae into twomajor groups that correspond with sporeprint color: Groups A to P include onlypale-spored agarics, which are derivedfrom within a paraphyletic group (R to"W") that includes mostly dark-sporedspecies. This pattern is similar to the earli-est Friesian classifications (Fries, 1821,1874), which emphasized spore print coloras an important character at higher taxo-nomic levels. A different set of relationships,however, is suggested by ML analysis (Fig.5), which places a group of dark-sporedagarics (clade R) within the pale-sporedagarics, and the Pleurotus clade (pale-spored; clade H) within the dark-spored.Statistical tests (Templeton and winning-sites test under MP criterion andKishino-Hasegawa test under ML model)of MP versus ML trees did not show anydifference between these alternative hy-

Dow




potheses. We agree with Singer's (1986:3)observation that "Fries' discovery of sporeprint colors as a taxonomic character of firstgrade importance was certainly a fortunateand valuable contribution to the systemat-ics of the Agaricales, however it should beused with reason".

Tricholomataceae (groups A-F, H-P, X).—The Tricholomataceae is the largest familyof the Agaricales. In Figures 2 and 5, mem-bers of the Tricholomataceae are not mono-phyletic: They composed the core of groupsA through F, H through P, and X (Rick-enella), with Laccaria and Callistosporium be-ing in group "Q". Group "Q" forms a het-erogeneous, poorly supported group thatalso includes members of the Cortinari-aceae, Entolomataceae, and Agaricaceae (itis monophyletic in the MP tree in Fig. 2, butdoes not include Clitopilus prunulus; it is pa-raphyletic in the ML analysis in Fig. 5).Trees constraining monophyly of the Tri-cholomataceae (including Phyllotopsis, Len-tinula, and Omphalotus but not Hohenbuehe-lia, which strongly groups with Pleurotus inclade H) are 3,135 steps in length and sig-nificantly different from MP trees (Tem-pleton test, P = 0.038-0.047). If only theRickenella clade (group X) is excluded, con-straint trees (3,123 steps) are no longer sig-nificantly different from MP trees (Tem-pleton test, P = 0.184-0.209). In summary,though our analysis cannot unequivocallysupport or reject monophyly of the Tri-cholomataceae, phylogenetic analysisstrongly suggest placement of certain taxain other families. Also, although attemptshave been made to redivide the Tricholo-mataceae into smaller families (Kuhner,1980; Pegler, 1983; Redhead, 1986), none ofthe clades in Figure 2 or 5 fully correspondswith any of these. However, further studyof cases where rDNA phylogeny and sup-porting characters from anatomy are inconcordance are likely to provide addi-tional justification for redividing the Tri-cholomataceae based on monophyletic lin-eages. Several examples are discussedbelow.

All MP trees (Figs. 1 and 2) and the MLtree (Fig. 5) show exemplar sequences fromthe Tricholomataceae divided into twosmaller clades (groups A-E with inclusion ofPleurotopsis longiqua, and M-P), with similarsets of relationships among groups within

each clade. The first clade (A-E) includes ex-emplars from Singer's tribes Collybieae,Marasmieae, Biannularieae, Pseudohiatu-leae, and Rhodoteae, whereas the secondclade (M-P) mostly contains members ofthe Lyophylleae, Termitomyceteae, Leuco-paxilleae, Tricholomatinae, and Clitocybi-nae (see Appendix). In Figures 2 and 5, ex-emplar sequences for tribe Myceneae fallinto two monophyletic groups, one (Baeo-spora with Hydropus, clade E) being sister tothe A-D clade, the other (Mycena andResinomycena, clade P) sister to the M-Oclade. This overall topology does not fullycorrespond with Singer's (1986) proposedtribes and subtribes (see Appendix) or anyother proposed classifications aimed atdividing the Tricholomataceae. However,Kuhner's (1980) families Marasmiaceae(with the inclusion of the Rhodotaceae) andTricholomataceae roughly correspond togroups A-D and M-O, respectively. Not allTricholomataceae are unambiguously re-solved, particularly groups F (Macrocybeand Ossicaulis), I (Resupinatus, Pleurocybella,and Phyllotopsis), "]" (Clitocybe lateritia, Cau-lorhiza, and Conchomyces), and L (Xerom-phalina and Clitocybe clavipes), for which weare unaware of any significant morphologi-cal synapomorphies. In the absence ofstronger evidence, we consider these lastgroupings based on nLSU-rDNA phylog-eny to be provisional.

Phylogenetic analyses also show thatseveral genera placed in the Tricholomat-aceae by Singer (1986) are not mono-phyletic. For example, exemplar taxa repre-senting different sections of Collybia areplaced into two phylogenetically separategroups (sections Levipedes and Striipedeswithin clade A, and section Collybia withinclade M). The genera Omphalina, Maras-mius, Tricholoma, and Clitocybe are also poly-phyletic according to our analyses. Maras-mius species possessing broom cells form amonophyletic group (in clade C), whereasthose lacking broom cells (M. pyrrocephalus,in section Chordales) belong to clade D, closeto Gloiocephala. Similarly, the mycorrhizalspecies of genus Tricholoma form a mono-phyletic group with Leucopaxillus (also amycorrhizal genus—clade N, 91% support),whereas Tricholoma giganteum (a nonmycor-rhizal species) is sister to Ossicaulislignatilis—group F, 100% support. Our re-

Dow




suits support a recent phylogenetic analysisof partial nLSU-rDNA sequence data by Pe-gler et al. (1998), who transferred T. gigan-teum to a new genus, Macrocybe.

The taxonomic position of the termite-associated agaric mushroom Termitomyces(Heim, 1941, 1977) has seen a certainamount of controversy. The original de-scription of Termitomyces (Heim, 1942) indi-cated morphological similarities with bothAmanita and Lepiota. Termitomyces speciesare characterized by a bilateral hy-menophore trama, as in Amanita and Plu-teus; pink spores, as in Pluteus and En-toloma; a spore hilum of the open-type, as inPluteus (Pegler and Young, 1971); andcyanophilic spores and siderophilous gran-ulation in the mature basidia, as in Lyophyl-lum (Clemengon, 1978,1985). Consequently,the genus has been tentatively classifiedin the Amanitaceae (Singer, 1962; Pegler,1977), Pluteaceae (Pegler and Young, 1971),and Tricholomataceae (Singer, 1975). In ourstudy both MP and ML analyses suggestthat Termitomyces is best classified close toLyophyllum (in clade O) in the Tricholomat-aceae.

The genus Omphalotus includes severalspecies of pleurotoid, white-spored, biolu-minescent, toxin-containing mushrooms,including the well-known Jack-o-lanternmushroom (O. olearius). Before 1970, mycol-ogists often placed Omphalotus in the Tri-cholomataceae, often as a synonym of othertaxa, including Clitocybe, Panus, Armillaria,or Pleurotus (see Corner, 1981; Singer, 1986).Similarities in pigment chemistry not foundin other agarics led Bresinsky (1974) andBresinsky and Besl (1979) to suggest a linkbetween Omphalotus and the boletes, a pro-posal accepted by later taxonomists, whofollowed in placing Omphalotus in the Paxil-laceae within the Boletineae (Singer, 1986;Bas et al., 1988). Our analyses place Om-phalotus with another bioluminescent fun-gus, Nothopanus eugrammus, in the para-phyletic group "B", which is deeply nestedwithin the A-E clade (Figs. 2 and 5). Biolu-minescence is not well documented infungi but has often been reported for otheragaric fungi, including Armillaria, Flam-mulina, Panellus, and Mycena. Another re-cent analysis of nLSU-rDNA phylogenysupports conclusions similar to ours, sug-gesting that Omphalotus is better placed

with white-spored agarics than with the bo-letes (Binder et al., 1997).

Amanitaceae and Pluteaceae (group G).—The Amanitaceae (Amanita and Limacella)are monophyletic in both MP and ML trees(Figs. 2 and 5). Monophyly of the genusAmanita is strongly supported (81% boot-strap support). Mycologists have disagreedabout placement for the genus Limacella,which was included in the Pluteaceae byLocquin (1984) and later transferred to theAmanitaceae by Singer (1986). Both MP andML analyses support placement of Limacellaas a sister-group to other Amanitaceae. Thethree members of the genus Pluteus weremonophyletic (86% bootstrap support) andnested with the Amanitaceae (in clade G) inthe MP and ML trees, although with lowersupport. The Amanitaceae and Pluteaceaediffer in their spore anatomy (Kiihner, 1984;Singer, 1986; Clemenc.on, 1997) but sharesimilarly structured lamellar trama, freeand crowded lamellae, and inamyloid hy-phae in the context tissue (Bas, 1969; Cle-mengon, 1997). A close relationship be-tween these two families was suspected byearlier taxonomic treatments, with Pluteusclassified in the Amanitaceae by Maire(1933) and Pegler (1977), and by Singer'searlier taxonomic system for the Agaricales(Singer, 1962).

The placement of Humidicutis marginataas a sister-group of the Amanitaceae in boththe MP and ML trees is surprising, giventhat most taxonomic treatments place itwith other members of the Hygrophora-ceae. Given that none of the exemplar taxafrom the Hygrophoraceae were clearlyplaced by our analyses, this odd placementof H. marginata may be the result of poorresolution for nLSU-rDNA data or evenpossible misidentification. Additional taxonand data sampling should help address thisissue.

Entolomataceae.—The two members ofEntolomataceae sampled (Entoloma strictiusand Clitopilus prunulus) were placed intogroups F and "Q" by the MP analysis (Fig.2) but do not cluster strongly with anytaxon. In the ML tree (Fig. 5), Entoloma stric-tius is also in group F but Clitopilus prunulusstands alone between groups G and T. TheEntolomataceae differ from other agarics inhaving spores that are angular and have acharacteristically complex wall structure as

Dow




seen in electron microscopy (Clemenc,on,1997); the monophyly of the family hasnever been challenged. We attribute to asampling artifact the nonmonophyly of thetwo Entolomataceae included here; the En-tolomataceae appear to be monophyleticfrom ongoing analyses from a larger dataset that includes many more taxa from thisfamily (Baroni, Moncalvo, and Vilgalys,data not shown).

Hygrophoraceae (groups G, K).—Membersof the Hygrophoraceae are nested amonggroups G (Humidicutis) and K (Hygrophorusand Hygrocybe) but do not cluster stronglywith any taxa (Fig. 2). A MP analysis con-straining the Hygrophoraceae to be mono-phyletic produced a tree of length 3,108,which is not significantly different from theshortest unconstrained trees of length 3,100(Templeton test, P = 0.893-0.943). Since thedescription of Hygrophorus by Fries (1835,1838) and the creation of the family Hy-grophoraceae by Roze (1876), the mono-phyly of this assemblage of white-spored,waxy agarics with long and narrow basidiahas never been questioned. Trees in Figs. 2and 5 (group K) suggest that the Hygro-phoraceae may be related to Omphalina andallied genera; these taxa share bright col-ored pigments and decurrent lamellae thatare sometimes thick and distant.

Pleurotaceae (group H).—Although severalpleurotoid fungi were included in thisstudy (Pleurotus, Pleurotopsis, Conchomyces,Resupinatus, Hohenbeuhelia, Phyllotopsis), onlyPleurotus and Hohenbeuhelia species are sup-ported as monophyletic groups, with 81%bootstrap support. Both Pleurotus and Ho-henbuehelia share the ability to trap and di-gest nematodes (Thorn, 1986; Thorn andTsuneda, 1993), which could be a synapo-morphy for recognizing the family Pleuro-taceae (Thorn et al, 2000).

Cortinariaceae (groups "Q", S, V).—Mem-bers of the Cortinariaceae (Cortinarius ingroup "Q", Inocybe in group S, and Hebe-loma in group V) were polyphyletic in bothMP and ML analyses. However, nLSU dataprovide only weak support for most rela-tionships. Constraining the Cortinariaceaeto be monophyletic resulted in trees 3,117steps in length, not significantly differentfrom the shortest unconstrained trees oflength 3,100 (Templeton test, P = 0.187-0.234). However, if relationships of Corti-

narius and Inocybe to other agarics remainlargely unresolved, members of Hebelomacluster with Agrocybe in both ML and MPtrees (clade V) with moderate (58%) boot-strap support in MP analyses.

Bolbitiaceae (groups R, V).—The smallfamily Bolbitiaceae includes Bolbitius, Cono-cybe, and Agrocybe. In both MP (Fig. 2) andML (Fig. 5) analyses, Bolbitius and Conocybeare monophyletic and form the sister group(group R) of the Panaeoloideae (Coprina-ceae), whereas Agrocybe groups with Hebe-loma (Cortinariaceae) in group V close to aparaphyletic Strophariaceae (group "W").

Coprinaceae (group T).—The Coprinaceae(group T), with the exclusion of subfamilyPanaeoloideae (group R) and two species(C. comatus and C. sterquilinus, both placedin the Agaricaceae and discussed below),are monophyletic in both ML and MP (71%bootstrap support) trees. This result is alsoin agreement with earlier studies by Hop-ple and Vilgalys (1994; 1999). SubfamilyPanaeoloideae (Panaeolina, Panaeolus, andAnelleria) form a monophyletic group with100% bootstrap support and clustered withBolbitius and Conocybe (Bolbitiaceae proparte) in both ML and MP trees but withlow bootstrap support in the MP analysis.Constraining the monophyly of the Panae-oloideae with the Coprinaceae (excludingCoprinus comatus and C. sterquilinus) re-sulted in a tree of length 3,112 that is notsignificantly different from the shortest un-constrained trees of length 3,100 (P = 0.593-0.683, Templeton test). Therefore, althoughour results suggest closer relationships be-tween the Panaeoloideae and Bolbitiusrather than between the Panaeoloideae andCoprinus, our results do not fully resolvethe controversy about whether the Panae-oloideae should be classified in the Copri-naceae (Singer, 1986; Bas et al., 1988; Kemp,1995) or in the Bolbitiaceae (Kiihner andRomagnesi, 1978).

Agaricaceae (group U).—The light- anddark-spored family Agaricaceae (excludingtribes Cystodermateae and including Copri-nus comatus and C. sterquilinus) form amonophyletic group (group U) in both MLand MP (58% bootstrap support) trees. Phy-logenetic evidence linking C. comatus andC. sterquilinus with the Agaricaceae hasbeen presented previously (Johnson andVilgalys, 1998; Hopple and Vilgalys, 1999).

Dow




In our analysis, two groups of ant sym-bionts (Gl and G3 from Chapela et al., 1994)are also included in the Agaricaceae. Theposition of both Cystoderma (in group "Q")and Ripartitella (in group S) remains unre-solved. Constraining the monophyly ofCystoderma and Ripartitella with the Agari-caceae resulted in a tree of length 3,118,which is not significantly different from theshortest MP trees (Templeton test, P =0.148-0.173).

Strophariaceae (group "W").—Members ofthe Strophariaceae form a paraphyleticgroup (group " W") at the base of the Agari-cineae in both MP and ML analyses (Figs. 2and 5). Monophyly of the genera Stropharia,Pholiota, and Hypholoma is fairly well sup-ported by bootstrapping (71% of bootstrapconfidence level); thus, neither subfamiliesStropharioideae nor Pholiotideae sensuSinger (1986) (see Appendix) appear to bemonophyletic.

One application for the nLSU database isfor identification and taxonomic placementof unknown sequences originating fromfungi, including mycorrhizas, plant tissues,and other environmental samples. Duringthe course of this study we identified oneinstance of misidentification involving aculture originally identified as Asterophoraparasita, a parasitic mushroom that often oc-curs on decaying fruit bodies of Russu-laceae, which was later determined to be anunidentified environmental isolate. Phylo-genetic analysis places "unknown basid-iomycete" together with Bondarzewia (group"Y") (Figs. 2 and 5), not with other Tricholo-mataceae (in the Agaricineae) as suggestedby morphological evidence (Kiihner, 1980;Singer, 1986). Because evidence from othermolecular studies (Bruns et al., 1998; Cull-ings et al., 1996) also supports placement ofAsterophora in the Tricholomataceae, weconsider this sequence to represent an envi-ronmental contaminant. Subsequent analy-ses of independently derived nLSU se-quences from bona fide material ofAsterophora have confirmed this conclusion(V. Hofstetter and H. Clemen^on, pers.comm.).

Phylogenetic Inference of Large Data Sets

With the recent birth of automated se-quencing, analysis of large molecular data

sets has come within the reach of many lab-oratories. At the same time, systematists arecollectively realizing the importance ofboth adequate taxon and data samplingfor accurate phylogenetic inference (Hillis,1996; Lecointre et al., 1993; Graybeal, 1998;Poe, 1998). In this study we performed ex-tensive analyses on a moderately large mol-ecular data set (154 taxa and 826 characters)to determine empirically which heuristicmethods might also be useful for futurephylogenetic studies involving larger num-bers of taxa and sequence data. Below wediscuss several aspects of complex phyloge-netic data sets that relate to our data.

In this study, taxa were selected from ob-servation of branch lengths in preliminaryanalyses, to avoid pitfalls resulting fromunequal branch lengths. Unequal branchlengths may have two origins: taxon sam-pling and among-taxa variation in molecu-lar rates of evolution.

Taxon sampling.—Taxon sampling canhave a strong effect on phylogenetic infer-ence (Lecointre et al., 1993; Poe, 1998). Weobserved this frequently during prelimi-nary analyses in which the bolete (groupAA) and Amanita (group G) clades were al-ways placed together within the Agari-cineae. A sister-group relationship betweenboletes and Amanita conflicted with nearlyevery other classification system proposedto date and made little sense from a mor-phological standpoint. Examination ofbranch lengths showed that both the bo-letes and Amanita spp. occurred on longbranches (Figs. 2 and 5), which suggestedthat "long branch attraction" might causespurious grouping of these taxa (Felsen-stein, 1978). One solution proposed for thelong branch attraction problem was sam-pling of additional taxa, which was ex-pected to break up long branches and resultin more stable phylogenetic estimates(Swofford et al., 1996; Graybeal, 1998). Wetherefore sampled Limacella (which hadbeen suggested to be the sister taxon ofAmanita in Singer, 1986) and Suillus (repre-senting another bolete lineage) for inclu-sion in this study. Phylogenetic analysis ofthe resulting data matrix (Figs. 2 and 5) sep-arated the bolete and Amanita clades, inagreement with morphologically basedclassifications. Although this was a particu-larly noteworthy example of how taxon

Dow




sampling can affect tree topologies, otherexamples appear to be present in our data.Our observations agree with those of Hillis(1996), who has demonstrated that in-creased taxon sampling density appears toalso increase phylogenetic accuracy.

Evaluating support: bootstrapping or jack-knifing?—One commonly encounteredproblem with large data sets concerns theapplicability or accuracy of standard de-scriptors used to assess branch robustness.The simplest measure of support used forparsimony trees, the decay or support in-dex (Bremer, 1994), is not practical for largedata sets because of the large number oftrees that cannot be sampled. Other mea-sures of "goodness of fit", such as CI andhomoplasy, are also sensitive to sample size(Sanderson and Donoghue, 1989). Evenbootstrapping has come under increasedscrutiny because of questions regardingwhat constitutes a "significant" value(Hillis and Bull, 1993; Sanderson, 1989) andalso because bootstrap replication often re-quires longer amounts of computing timeas the number of taxa increases. As an alter-native to bootstrapping, Farris et al. (1996)proposed the use of jackknife resampling toinfer branch support.

In this study little difference was ob-served between support values obtained bybootstrap or jackknife analyses (Fig. 3a).Also, bootstrapping with fast search strate-gies (no branch swapping) gave values thatwere slightly lower (and thus more conser-vative) than more-optimized methods (Fig.3b), which is probably a consequence of bothquick convergence of parsimony searches to-ward the shortest trees and inclusion ofsuboptimal trees in the bootstrap consen-sus. Overall topologies of bootstrap andjackknife trees were similar to that of thestrict consensus tree of combined islandsUP60, UP72, and UP180 (Fig. lb). There-fore, we regard bootstrapping and jack-knifing (including "fast" methods) as twosimilar tools for assessment of branchrobustness, both of which are useful (al-though not fully satisfactory) in analyses oflarge molecular data sets.

Choice of evolutionary models and searchstrategies.—Controversies abound concern-ing the choice of appropriate algorithms,models, or search strategies in reconstruct-ing complex phylogenies. The applicability

of parsimony to large data sets has beenquestioned because of the large number oftrees that must be examined in searchingfor shortest trees (other problems of parsi-mony in reconstructing phylogenies havebeen discussed by Felsenstein, 1978, andStewart, 1993). ML searches appear to beeven more limited by computational con-straints than parsimony. Though distancemethods such as NJ have the advantage ofrapidity for phylogenetic reconstruction, NJperformed poorly when compared againstother methods that used MP or ML as opti-mality criteria.

To reduce the time necessary for heuristicsearches to find islands of shortest trees andalso to explore the universe of parsimo-nious solutions, we conducted UP searchesfrom many different starting trees, usingvarious settings for MAXTREES and branchswapping. Two major observations regard-ing analysis of large data sets are also evi-dent from this study: (1) Multiple islands ofmost-parsimonious trees (Maddison, 1991)were found; the trees from two of them(UP72 and UP180) were very similar toeach other, whereas trees from the third is-land (UP60) were quite different from thosefrom UP72 and UP180. Discovery of addi-tional islands would also be expected to re-duce the overall resolution of basal rela-tionships in strict consensus trees (see Fig.1). (2) Several search strategies are neces-sary when analyzing large data sets. In par-ticular, we found it useful to vary severalparameters for heuristic searches, includingdifferent settings for MAXTREES (<100versus unlimited), as well as differentswapping algorithms (TBR and NNI) whileperforming successive searches, saving thebest trees at each. In a recursive fashion, ini-tial searches use less-greedy algorithmsfirst, which quickly converge toward localoptima; these are followed by more compu-tationally extensive algorithms, which aremore likely to find optimal trees. In thisstudy, all quick searches with 100 replicatesof random addition sequences (as well as asearch starting from random trees) thatused TBR branch swapping and MAX-TREES in the range 1-100 yielded at leastone tree within 0.13% of the length of MPtrees. In addition, all trees were <1% longerthan the length of the MP trees; though notoptimal, 99.9% of these trees were statisti-

Dow




cally similar to the MP trees, according tothe Templeton and winning-sites tests, andscores generally quickly improved whenthese trees were recursively submitted toNNI swapping with higher MAXTREESand then again to TBR swapping (data notshown). We do not know whether the quickconvergence toward MP trees as observedin our analyses resulted from our taxon-intensive sampling strategy or is a generalproperty of large data sets, or both; but ifthis is a general property of large data sets,then increasing the number of taxa wouldbe expected to increase accuracy of phylo-genies without necessarily increasing com-putation time, as has already been sug-gested by Hillis (1996,1998).

This study also demonstrates that appli-cation of ML approaches to moderatelylarge data sets is feasible with personalcomputers and small Unix workstations.Using a restricted set of preselected startingtrees (based on their likelihood scores), wewere able to swap ML trees to completionby using the NNI algorithm. SubsequentTBR swapping for > 1,000 hours did not al-ter the topology of the tree yielded fromNNI swapping, suggesting that the ML tree(Fig. 5) is a fairly good estimate of phy-logeny under the ML model of evolution. Insimulation studies with limited numbers oftaxa, ML performs extremely well in recon-structing phylogenies (Huelsenbeck andHillis, 1993); however, little is known aboutthe behavior of ML with complex, largedata sets in comparison with parsimonymodels. In this study, we observed goodcongruence between the ML tree and treesproduced from either UP or WP.

WP analysis yielded trees with slightlyhigher ML scores than UP did (-In>15,712.679 versus >15,717.376), suggest-ing that WP may perform better than UPwhen appropriate weighting schemes areapplied: This would agree with simulationstudies, which have shown that WP can re-cover the correct phylogeny with fewercharacters than UP requires (Hillis et al.,1994). If MP and ML have the same founda-tion, as suggested by Goldman (1990), thenthe use of WP with appropriate weightingschemes might also be expected to estimatetrees with higher likelihood scores. Thehigh correlation observed between nucleo-

tide weighting matrices used in WP versusML (Fig. 4) also suggests that the stepma-trix values based on dinucleotide frequen-cies provide a reasonably accurate estimateof nucleotide substitution biases in the data.

In conclusion, our experience with thesedata suggests several approaches in analyz-ing large data sets. We recommend the de-sign of a careful taxon sampling that maxi-mizes homogeneity of branch length acrossphylogenetic trees. For MP, our study sug-gests the following: (1) WP using a stepma-trix for nucleotide substitution types withparameters derived from the data set or es-timated by way of ML; and (2) extensive ex-ploration of the universe of all possible bi-nary trees with TBR branch swapping andmultiple replicates of random addition se-quences (at least 100) with low MAXTREES(e.g., 1 to 100). Evaluation of topologicaldifferences between trees obtained afterfastidious searches (e.g., with MAXTREESunlimited) versus those obtained in quicksearches (e.g., MULPARS off) provides in-formation about the extent (length range) ofthe universe of solutions (which are likelyto be found in different tree islands) andabout topological differences within thisuniverse. For ML searches, we also recom-mend using as starting trees those MP (orNJ) trees that exhibit higher ML scores, fol-lowed by extensive NNI branch swappingand exploration with TBR swapping. Stan-dard descriptors for branch robustness(bootstrapping and jackknifing) may not befully satisfactory when applied to largedata sets; however, even "fast" proceduresof these descriptors can quickly revealstrongly supported clades for developinguseful phylogenetic hypotheses.

ACKNOWLEDGMENTS

We thank Scott Redhead, Dennis Desjardin, OrsonK. Miller, Ronald Peterson, Jean Lodge, Albert Eicker,Heinz Clemencon, Thomas Kuyper, and Denise Lam-oure for helping with taxonomic questions and pro-viding specimens. Tom Bruns, David Hibbett, JimJohnson, Paul Manos, and Cliff Cunningham mademany valuable suggestions to the manuscript and re-search program. Randall Downer was of great assis-tance with computing facilities. David Swofford pro-vided tester versions of PAUP* and shared insightsinto ML models and search strategies. This work wassupported in part by several grants from the NationalScience Foundation and training grants from the A. W.Mellon Foundation.

Dow




REFERENCES

ALBERT, V. A., AND B. D. MISHLER. 1992. On the ratio-nale and utility of weighting nucleotide sequencedata. Cladistics 8:73-83.

ALLARD, M. W., AND J. M. CARPENTER. 1996. Onweighting and congruence. Cladistics 12:183-198.

BAS, C. 1969. Morphology and subdivision of Ama-nita and a monograph of its section Lepidella. Persoo-nia 5:285-579.

BAS, C, T. W. KUYPER, M. E. NOORDELOOS, AND E. C.VELLINGA. 1988. Flora Agaricina Neerlandica, Vol-ume 1. A. A. Balkema, Rotterdam.

BINDER, M., H. BESL, AND A. BRESINSKY. 1997 Agari-cales or Boletales? Molecular evidence towards theclassification of some controversial taxa. Z. Mykol.63:189-196.

BREMER, K. 1994. Branch support and tree stability.Cladistics 10:295-304.

BRESINSKY, A. 1974. Zur Frage der taxonomische Rel-evanz chemischer Merkmale bei hoheren Pilzen.Bull. Soc. Linn. Lyon R. Kiihner 1974:61-84.

BRESINSKY, A., AND H. BESL. 1979. Zum verwandt-schlaftlichen Anschluss von Omphalotus. Sydowia8:98-109.

BRUNS, T. D., AND T. M. SZARO. 1992. Rate and modedifferences between nuclear and mitochondrialsmall-subunit rRNA genes in mushrooms. Mol. Biol.Evol. 9:836-855.

BRUNS, T. D., T. M. SZARO, M. GARDES, K. W. CULLINGS,J. PAN, D. L. TAYLOR, T. R. HORTON, A. KRETZER, M.GARBELOTTO, AND Y. LI. 1998. A sequence databasefor identification of ectomycorrhizal basidiomycetesby phylogenetic analysis. Mol. Ecol. 7:257-272.

BRUNS, T. D., R. VILGALYS, S. M. BARNS, D. GONZALEZ,D. S. HIBBETT, D. J. LANE, L. SIMON, S. STICKEL, T. M.SZARO, W. G. WEISBURG, AND M. L. SOGIN. 1992.Evolutionary relationships within the fungi: Analy-ses of nuclear small subunit rRNA sequences. Mol.Phylogenet. Evol. 1:231-241.

CATALAN, P., E. A. KELLOG, AND R. G. OLMSTEAD. 1997.Phylogeny of Poaceae subfamily Pooideae based onchloroplast ndhF gene sequences. Mol. Phylogenet.Evol. 8:150-166.

CHAPELA, I. H., S. A. REHNER, T. R. SCHULTZ, AND U. G.MUELLER. 1994. Evolutionary history of the sym-biosis between fungus-growing ants and their fungi.Science 266:1691-1694.

CHASE, M. W., D. E. SOLTIS, R. G. OLMSTEAD, D. MOR-GAN, D. H. LES, B. D. MISHLER, M. R. DUVALL, R. A.PRICE, H. G. HILLS, Y. L. QIU, K. A. KRON, J. H. RETTIG,E. CONTI, J. D. PALMER, J. R. MANHART, K. J. SYTSMA,H. J. MICHAELS, W. J. KRESS, K. G. KAROL, W. D.CLARK, M. HEDREN, B. S. GAUT, R. K. JANSEN, K.-J.KIM, C. F. WIMPEE, J. F. SMITH, G. R. FURNIER, S. H.STRAUSS, Q.-Y. XIANG, G. M. PLUNKETT, P. S. SOLTIS,S. M. SWENSEN, S. E. WILLIAMS, P. A. GADEK, C. J.QUINN, L. E. EGUIARTE, E. GOLENBERG, G. H. LEARN,JR., S. W. GRAHAM, S. C. H. BARRETT, S. DAYANANDAN,AND V. A. ALBERT. 1993. Phylogenetics of seedplants: An analysis of nucleotide sequences from theplastid gene rbcL. Ann. Mo. Bot. Gard. 80:528-580.

CLEMENCON, H. 1978. Siderophilous granules in thebasidia of Hymenomycetes. Persoonia 10:83-86.

CLEMENCON, H. 1985. Siderophiles granules in thebasidia of Termitomyces. Mycol. Helv. 1:267-270.

CLEMENCON, H. 1997. Anatomy of the Hymeno-mycetes. Lausanne Univ. Press, Lausanne, Switzer-land.

CORNER, E. J. H. 1981. The agaric genera Lentinus,Panus, and Pleurotus with particular reference toMalaysian species. Beih. Nova Hedwigia 69:1-169.

CULLINGS, K. W., T. M. SZARO, AND T. D. BRUNS. 1996.Evolution of extreme specialization within a lineageof ectomycorrhizal epiparasites. Nature 379:63-66.

CUNNINGHAM, C. W. 1997. Is congruence betweendata partitions a reliable predictor of phylogeneticaccuracy? Empirically testing an iterative procedurefor choosing among phylogenetic methods. Syst.Biol. 46:464-478.

FARRIS, J. S., V. A. ALBERT, M. KALLERSJO, D. LIPSCOMB,AND A. G. KLUGE. 1996. Parsimony jackknifing out-performs neighbor-joining. Cladistics 12:99-124.

FELSENSTEIN, J. 1978. Cases in which parsimony andcompatibility will be positively misleading. Syst.Zool. 27:401^10.

FELSENSTEIN, J. 1985. Confidence limits on phyloge-nies: An approach using the bootstrap. Evolution39:783-791.

FRIES, E. M. 1821. Systema mycologicum. Lund, Swe-den. Johnson Reprint Corporation, New York, 1952.

FRIES, E. M. 1835. Corpus floranum provincialumSueciae. Uppsala, Sweden. Johnson Reprint Corpo-ration, New York, 1952.

FRIES, E. M. 1838. Epicrisis systematis mycologici.Uppsala, Sweden. Johnson Reprint Corporation,New York, 1952.

FRIES, E. M. 1874. Hymenomycetes europaei. Upp-sala, Sweden. Johnson Reprint Corporation, NewYork, 1952.

GOLDMAN, N. 1990. Maximum likelihood inferenceof phylogenetic trees, with special reference to aPoisson process model of DNA substitution and toparsimony analyses. Syst. Zool. 39:345-361.

GOLDMAN, N. 1993. Simple diagnostic statistical testsof models for DNA substitution. J. Mol. Evol.37:650-661.

GRAYBEAL, A. 1998. Is it better to add taxa or charac-ters to a difficult phylogenetic problem? Syst. Biol.47:9-17.

HASEGAWA, M., H. KISHINO, AND T.-A. YANO. 1985.Dating of the human-ape splitting by a molecularclock of mitochondrial DNA. J. Mol. Evol. 21:160-174.

HEIM, R. 1941. Etudes descriptives et exp6rimentalessur les agarics termitophiles d'Afrique tropicale.M£m. Acad. Sci. Inst. Fr. 64:1-72.

HEIM, R. 1942. Nouvelles Etudes descriptives sur lesagarics termitophiles d'Afrique tropicale. Arch.Mus. Natl. Hist. Nat. Paris 6:1-133.

HEIM, R. 1977. Termites et champignons. Boube'e, Paris.HIBBETT, D. S., AND M. J. DONOGHUE. 1995. Progress

toward a phylogenetic classification of the Polypo-raceae through parsimony analyses of mitochon-drial ribosomal DNA sequences. Can. J.. Bot.73:s853-s861.

HIBBETT, D. S., Y. F. NAKAI, A. TSUNEDA, AND M. J.DONOGHUE. 1995. Phylogenetic diversity in shi-itake inferred from nuclear ribosomal DNA se-quences. Mycologia 87:618-638.

HIBBETT, D. S., E. M. PINE, E. LANGER, G. LANGER, ANDM. J. DONOGHUE. 1997. Evolution of gilled mush-

Dow




rooms and puffballs inferred from ribosomal DNAsequences. Proc. Natl. Acad. Sci. USA 94:12002-12006.

HIBBETT, D. S., AND R. G. THORN. In press. Basidiomy-cota: Homobasidiomycetes. In The Mycota VII: System-atics and evolution (D. S. McLaughlin, ed.).Springer-Verlag, Berlin.

HIBBETT, D. S., AND R. VILGALYS. 1993. Phylogeneticrelationships of Lentinus (Basidiomycotina) inferredfrom molecular and morphological characters. Syst.Bot. 18:409-433.

HILLIS, D. M. 1996. Inferring complex phylogenies.Nature 383:130-131.

HILLIS, D. M. 1998. Taxonomic sampling, phyloge-netic accuracy, and investigator bias. Syst. Biol. 47:3-8.

HILLIS, D. M., AND J. J. BULL. 1993. An empirical testof bootstrapping as a method for assessing confi-dence in phylogenetic analysis. Syst. Biol. 42:182-192.

HILLIS, D. M., AND M. T. DIXON. 1991. RibosomalDNA: Molecular evolution and phylogenetic infer-ence. Q. Rev. Biol. 66:411^453.

HILLIS, D. M., J. P. HUELSENBECK, AND C. W. CUNNING-HAM. 1994. Application and accuracy of molecularphylogenies. Science 264:671-677.

HOPPLE, J. S., JR., AND R. VILGALYS. 1994. Phylogeneticrelationship among coprinoid taxa and allies basedon data from restriction site mapping of nuclearrDNA. Mycologia 86:96-107.

HOPPLE, J. S., JR., AND R. VILGALYS. 1999. Phylogeneticrelationships in the mushroom genus Coprinus anddark-spored allies based on sequence data from thenuclear gene coding for the large ribosomal subunitRNA: divergent domains, outgroups and mono-phyly. Mol. Phylogenet. Evol. 13:1-19.

HUELSENBECK, J. P., AND D. M. HILLIS. 1993. Success ofphylogenetic methods in the four-taxon case. Syst.Biol. 42:247-264.

HUELSENBECK, J. P., AND B. RANNALA. 1997. Phyloge-netic methods come of age: Testing hypotheses in anevolutionary context. Science 276:227-232.

JOHNSON, J., AND R. VILGALYS. 1998. Phylogenetic sys-tematics of Lepiota sensu lato based on nuclear largesubunit rDNA evidence. Mycologia 90: 971-979.

JUKES, T. H., AND C. R. CANTOR. 1969. Evolution ofprotein molecules. Pages 21-132 in Mammalian pro-tein metabolism (H. N. Munro, ed.). Academic Press,New York.

JULICH, W. 1981. Higher taxa of Basidiomycetes. Bib-lioteka mycologica, volume 85. J. Kramer, Vaduz,Liechtenstein.

KEMP, R. F. 1995. Incompatibility in basidiomycetes:The heterogenic Pentax. Edinburgh J. Bot. 52:71-89.

KIM, J. 1998. Large-scale phylogenies and measuringthe performance of phylogenetic estimators. Syst.Biol. 47:43-60.

KIMURA, M. 1980. A simple method for estimatingevolutionary rate of base substitutions trough com-parative studies of nucleotide sequences. J. Mol.Evol. 16:111-120.

KISHINO, H., AND M. HASEGAWA. 1989. Evaluation ofthe maximum likelihood estimates of the evolution-ary tree topologies from sequence data, and thebranching order in Hominoidea. J. Mol. Evol. 29:170-179.

KREISEL, H. 1969. Grundziige eines naturliches Sys-tems der Pilze. Cramer, Lehre, Germany.

KUHNER, R. 1980. Les Hymenomycetes agaricoides:etude generale et classification. Societe Linneennede Lyon, Lyon, France.

KUHNER, R. 1984. Some mainlines of classification inthe gill fungi. Mycologia 76:1059-1074.

KUHNER, R., AND H. ROMAGNESI. 1978. Flore Analy-tique des Champignons Superieurs, 3rd edition.Masson, Paris.

KUZOFF, R. K., J. A. SWEERE, D. E. SOLTIS, P. S. SOLTIS, ANDE. A. ZIMMER. 1998. The phylogenetic potential ofentire 26S rDNA sequences in plants. Mol. Biol.Evol. 15:251-263.

LECOINTRE, G., H. PHILIPPE, H. L. VAN LE, AND H. LEGUYADER. 1993. Species sampling has a major im-pact on phylogenetic inference. Mol. Phylogenet.Evol. 2:205-224.

LEE, S. B., AND J. W. TAYLOR. 1990. Isolation of DNAfrom fungal mycelia and single spores. Pages282-287 in PCR protocols: A guide to methods andapplications (M. A. Innis, D. H. Gelfand, J. J. Sninsky,and T. J. White, eds.). Academic Press, New York.

LOCKHART, P. J., M. A. STEELE, M. D. HENDY, AND D.PENNY. 1994. Recovering evolutionary trees undera more realistic model of sequence evolution. Mol.Biol. Evol. 11:605-612.

LOCQUIN, M. 1984. Mycologie generate et structurale.Masson, Paris.

LUTZONI, F. M. 1997. Phylogeny of lichen- and non-lichen-forming omphalinoid mushrooms and theutility of testing for combinability among multipledata sets. Syst. Biol. 46:373-406.

MADDISON, D. R. 1991. The discovery and impor-tance of multiple islands of most-parsimonioustrees. Syst. Zool. 40:315-328.

MADDISON, W. P., AND D. R. MADDISON. 1992. Mac-Clade, version 3. Analysis of phylogeny and charac-ter evolution. Sinauer Associates, Sunderland,Massachusetts.

MADDISON, D. R., M. RUVOLO, AND D. L. SWOFFORD.1992. Geographic origins of human mitochondrialDNA: Phylogenetic evidence from control region se-quences. Syst. Biol. 41:111-124.

MAIRE, R. 1933. Fungi Catalaunici. Treballs JuntaMus. Cienc. Nat. Barcelona 3:1-120.

MICHOT, B., N. HASSOUNA, AND J. BACHELLERIE. 1984.Secondary structure of mouse 28S rRNA and a gen-eral model for the folding of the large rRNA in eu-caryotes. Nucleic Acids Res. 12:4259-4279.

MICHOT, B., L. HU QU, AND J. P. BACHELLERIE. 1990.Evolution of large-subunit ribosomal RNA struc-ture: The diversification of divergent D3 domainamong major phylogenetic groups. Eur. J. Biochem.188:219-230.

MILLER, O. K., JR., AND R. WATLING. 1987. Whencecometh the agarics? A reappraisal. Pages 435-448 inEvolutionary biology of the fungi (A. D. M. Rayner,C. M. Brasier, and D. Moore, eds.). Cambridge Univ.Press, Cambridge, UK.

MONCALVO, J. M., S. A. REHNER, AND R. VILGALYS. 1993.Systematics of Lyophyllum section Difformia based onevidence from culture studies and ribosomal DNAsequences. Mycologia 85:788-794.

MONCALVO, J. M., H. H. WANG, AND R. S. HSEU. 1995.Phylogenetic relationships in Ganoderma inferred

Dow




from the internal transcribed spacers and 25S ribo-somal DNA sequences. Mycologia 87:223-238.

MOSER, M. 1967. Basidiomycetes. II. Kleine Kryp-togamenflora, volume IIb/2. Gustav Fisher Verlag,Stuttgart.

OLMSTEAD, R. G., B. BREMER, K. M. SCOTT, AND J. D.PALMER. 1993. A molecular systematic analysis ofthe Asteridae sensu lato based on rbcL sequences.Ann. Mo. Bot. Gard. 80:700-722.

OLMSTEAD, R. G., AND J. D. PALMER. 1994. ChloroplastDNA systematics: A review of methods and dataanalysis. Am. J. Bot. 81:1205-1224.

PATOUILLARD, N. 1887. Les Hymenomycetes d'Eu-rope: Anatomie generate et classification des cham-pignons sup£rieurs. Librairie Paul Klincksieck,Paris.

PATOUILLARD, N. 1900. Essai taxonomique sur lesfamilies et les genres d'Hymenomycetes. Lons-le-Saunier, Paris.

PEGLER, D. N. 1977. A preliminary agaric flora of EastAfrica. Kew Bull. Addit. Ser. 6:1-615.

PEGLER, D. N. 1983. Agaric flora of the Lesser An-tilles. Kew Bull. Addit. Ser. 9:1-668.

PEGLER, D. N., D. J. LODGE, AND K. K. NAKASONE. 1998.The pantropical genus Macrocybe gen. nov. Mycolo-gia 90:494-504.

PEGLER, D. N., AND T. W. K. YOUNG. 1971. Basidio-spore morphology in the Agaricales. Beih. NovaHedwigia 35:1-210.

POE, S. 1998. The effect of taxonomic sampling on ac-curacy of phylogeny estimation: Test case of aknown phylogeny. Mol. Biol. Evol. 15:1086-1090.

REDHEAD, S. A. 1986. The Xerulaceae (Basidio-mycetes), a family with sarcodimitic tissues. Can. J.Bot. 65:1551-1562.

REDHEAD, S. A., AND L. L. NORVELL. 1993. Notes onBondarzewia, Heterobasidion and Pleurogala, newgenus. Mycotaxon 48:371-380.

RICE, K. A., M. J. DONOGHUE, AND R. G. OLMSTEAD.

1997. Analyzing large data sets: rbcL 500 revisited.Syst. Biol. 46:554-563.

ROZE, E. 1876. Essai d'une nouvelle classification desAgaricinees. Bull. Soc. Bot. Fr. 23:45-54.

SAITOU, N., AND M. NEI. 1987. The neighbor-joiningmethod: A new method for reconstructing phyloge-netic trees. Mol. Biol. Evol. 4:406-425.

SANDERSON, M. J. 1989. Confidence limits on phylo-genies: The bootstrap revisited. Cladistics 5:113-129.

SANDERSON, M. J., AND M. J. DONOGHUE. 1989. Pat-terns of variation in levels of homoplasy. Evolution43:1781-1795.

SINGER, R. 1962. The Agaricales in modern taxon-omy. Weinheim, Germany.

SINGER, R. 1975. The Agaricales in modern taxon-omy, 3rd edition. J. Cramer, Vaduz, Liechtenstein.

SINGER, R. 1986. The Agaricales in modern taxon-omy, 4th edition. Koeltz, Koenigstein, Germany.

SOLTIS, D. E., P. S. SOLTIS, M. E. MORT, M. W. CHASE, V.SAVOLAINEN, S. B. HOOT, AND C. M. MORTON. 1998.Inferring complex phylogenies using parsimony: Anempirical approach using three large DNA data setsfor angiosperms. Syst. Biol. 47:32-42.

SOLTIS, D. E., P. S. SOLTIS, D. L. NICKRENT, L. A. JOHNSON,W. J. HAHN, S. B. HOOT, J. A. SWEERE, R. K. KUZOFF,

K. A. KRON, M. W. CHASE, S. M. SWENSEN, E. A. ZIM-MER, S. M. CHAW, L. J. GILLESPIE, W. J. KRESS, AND K. J.SYTSMA. 1997. Angiosperm phylogeny inferredfrom 18S ribosomal DNA sequences. Ann. Mo. Bot.Gard. 84:1^9.

SOLTIS, P. S., AND D. E. SOLTIS. 1997. Phylogeneticanalysis of large molecular data sets. Bol. Soc. Bot.Mex. 59:99-113.

STEWART, C.-B. 1993. The powers and pitfalls of parsi-mony. Nature 361:603-607.

SWOFFORD, D. L. 1991. PAUP: Phylogenetic analysisusing parsimony, version 3.1.1. Illinois Natural His-tory Survey, Champaign.

SWOFFORD, D. L. 1998. PAUP*: Phylogenetic analysisusing parsimony (*and other methods), beta version4.0d64. Sinauer, Sunderland, Massachusetts.

SWOFFORD, D. L., G. J. OLSEN, P. J. WADDELL, AND D. M.HILLIS. 1996. Phylogenetic inference. Pages 407-514 in Molecular systematics, 2nd edition (D. M.Hillis, C. Moritz, and B. K. Mable, eds.). Sinauer,Sunderland, Massachusetts.

TAJIMA, E, ANDM. NEI. 1984. Estimation of evolution-ary distance between nucleotide sequences. Mol.Biol. Evol. 1:269-285.

TEMPLETON, A. R. 1983. Phylogenetic inference fromrestriction endonuclease cleavage site maps withparticular reference to the evolution of humans andthe apes. Evolution 37:221-244.

THORN, R. G. 1986. The Pleurotus silvanus complex.Mycotaxon 25:27-66.

THORN, R.G., J. M. MONCALVO, C. A. REDDY, AND R. VIL-GALYS. 2000. Phylogenetic analyses and the distri-bution of nematophagy support a monophyleticPleurotaceae within the polyphyletic pleurotoid-lentinoid fungi. Mycologia, 92:241-252.

THORN, R. G., AND A. TSUNEDA. 1993. Interactions be-tween Pleurotus species, nematodes, and bacteria onagar and in wood. Trans. My'col. Soc. Jpn. 34:449-464.

VILGALYS, R., AND M. HESTER. 1990. Rapid geneticidentification and mapping of enzymatically ampli-fied ribosomal DNA from several Cryptococcusspecies. J. Bacteriol. 172:4238-4246.

VILGALYS, R., AND B. L. SUN. 1994. Ancient and recentpatterns of geographic speciation in the oystermushroom Pleurotus revealed by phylogeneticanalysis of ribosomal DNA sequences. Proc. Natl.Acad. Sci. USA 91:4599-4603.

YANG, Z. 1994. Estimating the pattern of nucleotidesubstitution. J. Mol. Evol. 39:105-111.

YODER, A. D., R. VILGALYS, AND M. RUVOLO. 1996.Molecular evolutionary dynamics of cytochrome bin strepsirrhine primates: The phylogenetic signifi-cance of third-position transversions. Mol. Biol.Evol. 13:1339-1350.

ZOLAN, M. E., AND P. J. PUKKILA. 1986. Inheritance ofDNA methylation in Coprinus cinereus. Mol. Cell.Biol. 6:195-200.

Received 21 October 1998; accepted 2 February 1999

Associate Editor: D. Hibbett

Dow



302

Taxa

SYSTEMATIC BIOLOGY

APPENDIX

Organisms included in the study are listed according to Singer's3 (1986) taxonomic system

DNACollection no.a sourceb

VOL. 49

GenBank no.

Suborder Agaricineae (11 families)Polyporaceae (2 tribes)

Tribe Lentineae (5 genera)Pleurotus ostreatus (Jacqu.: Fr.) KummerPleurotus djamor (Fr.) BoedijnPleurotus purpureoolivaceus (Stev.) Segedin et al.Nothopanus eugrammus (Mont.) SingerPhyllotopsis nidulans (Pers.: Fr.) SingerLentinula edodes (Berk.) Pegler4

Hygrophoraceae (3 tribes)Tribe Hygrophoreae (1 genus)

Hygrophorus sordidus PeckHygrophorus bakerensis Smith & Hesler

Tribe Hygrocybeae (5 genera)Humidicutis marginata (Peck.) SingerHygrocybe citrinopallida (Smith & Hesler) Kobay

Tricholomataceae (12 tribes)Tribe Lyophylleae (4 genera)

Lyophyllum decastes (Fr.: Fr.) SingerLyophyllum atratum (Fr.) SingerLyophyllum semitale (Fr.) KuhnerHypsizygus ultnarius (Bull.: Fr.) Redhead

Tribe Termitomyceteae (2 genera)Termitomyces cylindricus S.E. HeTermitomyces heimii NatarajanPodabrella microcarpa (Berk. & Br.) Singer

Tribe Tricholomateae (4 subtribes)Subtribe Laccariinae (1 genus)

Laccaria bicolor (Maire) OrtonSubtribe Clitocybinae (3 genera)

Clitocybe nuda (Bull.: Fr.) CookeClitocybe connata (Schum.: Fr.) GilletClitocybe dealbata (Sow.: Fr.) KummerClitocybe clavipes (Pers.: Fr.) KummerClitocybe lateritia FavreOssicaulis lignatilis (Pers.: Fr.) Redhead & Ginnsf

Subtribe Tricholomatinae (1 genus)Tricholoma atroviolaceum A.H. SmithTricholotna subaureum OvreboTricholoma imbricatum (Fr.: Fr.) KummerTricholoma pardinum QueletTricholoma portentosum (Fr.) QueletTricholoma caligatum (Viv.) RickenMacrocybe gigantea (Berk.) Pegler & Lodgeh

Subtribe Omphalinae (13 genera)Arrhenia auriscalpium (Fr.) Fr.Omphalina luteovitellina (Pilat & Nannf.) M. LangeOmphalina rosella (M. Lange) MoserOmphalina velutina (Quel.) QueletOmphalina velutipes OrtonOmphalina ericetorum (Fr.) M. LangeOmphalina pyxidata (Pers.: Fr.) QueletGerronema strombodes (Berk. & Mont.) SingerRickenella mellea (Sing. & Clem.) Lamoure'Rickenella pseudogrisella (A.H. Smith) Gulden'Chrysomphalina chrysophylla (Fr.) Cle'mencon'Chrysomphalina grossula (Pers.) Novell et al.)Phaeotellus griseopallidus (Desm.) Kuhner & Lamourek

Callistosporium luteoolivaceum (Berk.& Curt.) Singer

RV 83/233t = D261RV 95/920RHP 3588.8 = D2342JLPR1308RV96/1ATCC 42962

RV 94/178SAR s.n.

JM 96/33LUTZ 930731-1

JM87/16HC 79/133HC 85/13JM-HW

JM leg. R.S.Hseu s.n.JM leg. S.Muid s.n.PRU3900

JM96/19

RV84/1JM90cJMs.n.JM 96/22LUTZ 930803-1DAOM 191173

KMS400KMS590KMS356KMS278KMS591KMS452IFO 31860

LUTZ 930731-3LUTZ 930816-8REDHEAD 7501LUTZ 930812-1LAM L77-166hllXh4LUTZ 930817-2LAM L66-118hl4KUYPER2984LAM 74-20h 1/9.91LUTZ 930728-3MICH A.H.Smith 76299MICH A.H.Smith 82899LUTZ & LAM 910828-4RV10-1

CCCCBC

BB

BB

CCCC

BBB

B

BBBBBC

BBBBBBC

BBBBCBCBCBBBBC

U04140c

AF042575AF042576AF042577AF042578AF042579

AF042562AF042623

AF042580U66435e

AF042583AF042582AF042581AF042584

AF042585AF042586AF042587

AF042588

AF042624AF042590AF042589AF042564U66431e

AF042625

U76457SU76466SU76458SU764628U764648U76467SAF042591

U66428*U66447*U66452*U66454*U66455*U66445*U66450*U66433e

U66438e

U66437*U66430e

U66457*U66436e

AF042627

Dow




APPENDIX (CONTINUED)

Taxa Collection no.aDNA

sourceb GenBank no.

Tribe Leucopaxilleae (8 genera)Leucopaxillus albissimus (Peck) Singer

Tribe Biannularieae (2 genera)Armillaria tabescens (Scop.: Fr.) Emel

Tribe Collybieae (15 genera)Pleurocybella porrigens (Pers.: Fr.) SingerCollybia dryophila (Bull.: Fr.) KummerCollybia polyphylla (Peck) Sing.Collybia maculata (Fr.) Kumm.Collybia racemosa (Pers.: Fr.) Que"l.Marasmiellus ramealis (Bull.: Fr.) SingerMicromphale perforans (Hofm.: Fr.) SingerCampanella subdendrophora Redhead

Tribe Resupinateae (6 genera)Hohenbuehelia tristis StevensonHohenbuehelia sp.Resupinatus alboniger (Pat.) SingerResupinatus sp.Conchomyces bursaeformis (Berk.) Horak1

Tribe Panelleae (3 genera)Pleurotopsis longinqua (Berk.) Horakm

Tribe Marasmieae (3 subtribes, 18 genera)Marasmius delectans MorganMarasmius capillaris MorganMarasmius pyrrocephalus BerkeleyStrobilurus trullisatus (Murr.) LennoxGloiocephala menieri (Boud.) SingerXerula furfuracea (Peck) Redhead et al. °Crinipellis maxima A.H. Smith & Walter

Tribe Myceneae (16 genera)Baeospora myriaodophylla (Peck) SingerHydropus scabripes (Murr.) SimgerCaulorhiza hygrophoroides (Peck) HallingpMycena galericulata (Scop.: Fr.) S.F. GrayMycena clavicularis (Fr.) SaccardoMycena rutilanthiformis (Murr.) MurrillResinomycena acadiensis Redhead & SingerXeromphalina cauticinalis (With.: Fr.) Kuhner & Maire

Tribe Pseudohiatuleae (4 genera)Cyptotrama asprata (Berk.) Redhead & GinnsFlammulina velutipes (Curt.: Fr.) Singer

Tribe Rhodoteae (1 genus)Rhodotus palmatus (Bull.: Fr.) Maire

Amanitaceae (2 genera)Amanita muscaria (L.: Fr.) HookerAmanita citrina (Schaeff.) S.F. GrayAmanita rubescens (Pers.: Fr.) S.F. GrayAmanita flavoconia AtkinsonAmanita peckiana KauffmannLimacella glishra (Morg.) Murrill

Pluteaceae (3 genera)Pluteus primus BonnardPluteus petasatus (Fr.) GilletPluteus sp.

Agaricaceae (4 tribes)Tribe Leucocoprineae (7 genera)

Leucocoprinus cepaestipes (Sow.: Fr.) PatouillardLeucocoprinus fragilissimus (Berk. & Rav.) PatouillardLeucoagaricus rubrotinctus (Peck) SingerLeucoagaricus naucinus (Fr.) Singer

SAR1-2-90

D290

OKM 19644RV 83/180RV 182.01RV 94/175DED 5575DED 3973RV 83/67DAOM 175393

RV 95/214 •RV 95/573RV/JM s.n.VT1520 = D 596RV 95/302

RV 95/473

DED 4518DED 4345DED 4503DAOM 188775DAOM 170087JM 96/42DAOM 196019

DAOM 188774DAOM 192847DAOM 172075RV 87/14.01RV 87/6JM96/26DAOM 169949RV 86/11

DAOM 157066SAR s.n.

VT356

SAR s.n.J188RV "5Aug96"RV "5Aug96"RV 94/143VT-GB505

JM leg. JB 94-24JM leg. JB 91-21JM 96/28

EFM 548JJ84JJ100OKM 15134

B

C

BCCBCCCC

CCBCC

B

CCCCCBC

CCCCCBCB

CB

C

BBBBBB

BBB

CBBB

AF042592

AF042593

AF042594AF042595AF042596AF042597AF042598AF042626AF042628AF042629

AF042601AF042602AF042600AF042599AF042603

AF042604

U11922"AF042631AF042605AF042633AF042632AF042566AF042630

AF042634AF042635AF042640AF042636AF042637AF042606AF042638AF042639

AF042642AF042641

AF042565

AF042643AF041547qAF042607AF042609AF042608U85301'

AF042610AF042611AF042612

U85286r

U85289'U85281'U85280r

Dow



304

Taxa

SYSTEMATIC BIOLOGY


Collection no.aDNA

sourceb

VOL. 49

GenBank no.

Macrolepiota procera (Scop.: Fr.) SingerMacrolepiota rachodes (Vitt.) SingerChlorophyllum molybdites (Meyer: Fr.) Massee

Tribe Agariceae (5 genera)Agaricus bisporns (Lange) SingerAgaricus pocillator Murrill

Tribe Lepioteae (6 genera)Cystolepiota cystidiosa (A.H. Smith) BonLepiota acutesquamosa (Weinm.) SingerLepiota cristata (Bolt.: Fr.) KummerLepiota clypeolaria (Bull.: Fr.) Kummer

Tribe Cystodermateae (7 genera)Cystoderma granulosnm (Batsch) FayodRipartiteHa brasiliensis (Speg.) Singer

Coprinaceae (4 subfamilies)Subfamily Coprinoideae (1 genus)

Coprinus atramentarius (Bull.: Fr.) FriesCoprinus cinereus (Schaeff.: Fr.) FriesCoprinus kimurae Hongo & AokiCoprinus nudiceps P.D. OrtonCoprinus bisporus LangeCoprinus comatus (Mull.: Fr.) S.F. GrayCoprinus sterquilinus (Fr.) Fries

Subfamily Psathyrelloideae (2 genera)Psathyrella candolleana (Fr.) MairePsathyrella delineata (Peck) A.H. SmithPsathyrella gracilis (Fr.) QueletLacrymaria velutina (Pers.: Fr.) Singers

Subfamily Panaeoloideae (4 genera)Panaeolina foensecii (Pers.: Fr.) MairePanaeolus acuminatus (Schaeff.: Seer.) QueletAnellaria semiovata (Sow.: Fr.) Pearson & Dennis

Bolbitiaceae (6 genera)Bolbitius vitellinus (Pers.) FriesConocybe rickenii (Schaeff.) KiihnerAgrocybe praecox (Pers.: Fr.) Fayod

Strophariaceae (2 subfamilies)Subfamily Stropharioideae (4 genera)

Stropharia rugosoannulata Farlow ex. MurrillPsilocybe stuntzii Guzman & OttPsilocybe silvatica (Peck) Singer & SmithHypholoma sublateritium (Fr.) Quelet 'Hypholoma subviride (Berk. & Curt.) Dennis t

Subfamily Pholiotoideae (5 genera)Pholiota squarrosoides PeckKuehneromyces mutabilis (Schaeff.: Fr.) Singer & Smith

Cortinariaceae (3 tribes)Tribe Inocybeae (1 genus)

Inocybe sp.Inocybe geophylla (Sow.: Fr.) Kummer

Tribe Hebelomateae (3 genera)Hebeloma crustiliniforme (Bull.: Fr.) Quelet

Tribe Cortinarieae (10 genera)Cortinarius iodes Berkeley & CurtisCortinarius sp.Cortinarius marylandensis Ammirati & Smith

Entolomataceae (3 genera)Clitopilus prunulus (Scop.: Fr.) KummerEntoloma strictius (Peck) Sacc.

JJ168OKM 19588JJ162

SAR 88/411J173

MICH 18884JJ177DUKE HN 1582OKM 22029

BPI752511EFM 744

C 114 = VT 1131C 13 (from R.KEMP)C 78 = KEMP 1553C 159 = KEMP 737/1C 148 = KEMP 1659/1C 116 = D 252C 123 = TF916 (VPI)

J181J156J130JlOO

SAR 87/378J129SAR s.n.

SAR 84/100J183SAR 84/159

D258VT 1263 = D 216RV 5-7-1989JM 96/20JJ69

JJ7DSM 1684

RV7/4JM96/25

SAR 87/408

JM 96/23JM 96/40JM 96/24

RV 88/109JM 96/10

CQ

C

Q

CO

03

O3

BBBB

BC

UU

UU

UU

U

cBCC

CO

U

CQ

CO

00

03

U

UU

CQ CQ

BC

03

03

B

CO

C

Q

CQ

CB

U85304r

U85277r

U85274'

U11911"AF041542q

U85298'U85293r

U85292'U85291'

U85299'U85300r

AF041484qAF041494qAF041500qAF041517qAF041523qAF041529qAF0415301

AF041531qAF041532qAF041533qAF041534q

U11924"AF041535qAF041536q

U11913"AF041546qAF042644

AF0415441AF042567AF042618AF042569AF042570

AF042568AF042619

AF042617AF042616

U11918"

AF042613AF042614AF042615

AF042645AF042620

Dow





Taxa

Suborder Boletineae (3 families)Paxillaceae (7 genera)

Omphalotus nidiformis BerkeleyBoletaceae (24 genera)

Suillus luteus (L.: Fr.) S.F. GrayPhylloporus rhodoxanthus (Schwein.) BresadolaBoletus retipes Berkeley & Curtis"

Suborder Russulineae (2 families)Bondarzewiaceae (1 genus)

Bondarzewia mesenterica (Schaeff.) KreiselRussulaceae (2 genera)

Russula earleiPeckRussula mairei SingerRussula virescens (Schaeff.: Zanted.) FriesRussula romagnesii SingerLactarius corrugis PeckLactarius piperatus (L.: Fr.) S.F. GrayLactarius volemus (Fr.) Fries

Unclassified:Attine fungus G1.U11893Attine fungus G1.U11902Attine fungus G2.U11905Attine fungus G3.U11895Unknown basidiomycetev

Outgroups:Ganoderma lucidum gr.Ganoderma australe gr.

Collection no.aDNA

source6

VT 1946.8 = OKM 23886 C

JM 96/41SAR 89/457SAR91/1

SARs.n.

RV Isp92RV 89/62

JH s.n.JJ60RV 88/61RV"6jul. 1994"RV 94/150

RV 94/140

JM RSH-RZJM RSH-0705

BBB

B

BBBBBBB

C

CC

GenBank no.

AF042621

AF042622U11925"U11914"

AF042646

AF042571U11926"AF041548ciAF042572U11919"AF042573AF042574

U11893"U11902"U11905" .U11895"AF042563

X78776W

X78780w

R̂V, D, JM, SAR, LUTZ, JJ, JH, J, C, KMS = authors' collections or Duke University Herbarium and Culture Collection; VT, VP1,OKM = Orson K. Miller, Virginia Tech.; DED = Dennis Desjardin, San Francisco; RHP = Ron Peterson, University of Tennessee;DAOM, REDHEAD = Scott Redhead, Ottawa, Canada; DSM, HC = Heinz Clemengon, University of Lausanne, Switzerland; JL =Jean Lodge, Puerto Rico; KUYPER = Thomas Kuyper, The Netherlands; LAM = Denise Lamoure, France; MICH = Herbarium ofthe University of Michigan; EFM = New York Botanical Garden; PRU = Herbarium of the University of Pretoria, South Africa; BPI= U.S. National Fungus Collection; ATCC = American Type Culture Collection; IFO = Institute for Fermentation, Osaka, Japan.

bDNA was isolated from culture (C) or basidiome (B) tissues.cVilgalys and Sun (1994).ALentinus in Singer (1986).cLutzoni (1997).<Clitoajbe in Singer (1986).RShanks and Vilgalys, unpublished.hTriclwloma in Singer (1986).'Gerronema in Singer (1986).iOmphalina zvynniae (Berk. & Br.) Ito in Lutzoni, 1997.kLeptoglossum in Singer (1986).]Hohenbuehelia in Singer (1986).mPanellus in Singer (1986)."Chapela et al. (1994)."Oudemansiella in Singer (1986).PHydropus in Singer (1986).qHopple and Vilgalys, 1999.••Johnson and Vilgalys (1998).sPsatlnjrella in Singer (1986).iNaematoloma in Singer (1986)."Pulveroboletus in Singer (1986).vUnidentified culture isolated from Asteroplwra parasitica."Moncalvo et al. (1995).

Dow



Phylogenetic Relationships of Agaric Fungi Based on Nuclear ...

Documents

Transcript of Phylogenetic Relationships of Agaric Fungi Based on Nuclear ...