Testing Morphological Concepts of Orders of Pleurocarpous Mosses (Bryophyta) Using Phylogenetic Reconstructions Based on TRNL-TRNF and RPS4 Sequences William R

Home , Acroporium, Campylium, Cryphaeaceae, Cryphaea (plant), Ctenidium (plant), Fontinalis

Molecular Phylogenetics and Evolution Vol. 16, No. 2, August, pp. 180–198, 2000 doi:10.1006/mpev.2000.0805, available online at http://www.idealibrary.com on

Testing Morphological Concepts of Orders of Pleurocarpous Mosses (Bryophyta) Using Phylogenetic Reconstructions Based on TRNL-TRNF and RPS4 Sequences William R. Buck,* Bernard Gofﬁnet,†,‡ and A. Jonathan Shaw† *Institute of Systematic Botany, New York Botanical Garden, Bronx, New York 10458-5126; †Department of Botany, Duke University, Box 90339, Durham, North Carolina 27708-0338; and ‡Department of Ecology and Evolutionary Biology, Box U-43, University of Connecticut, Storrs, Connecticut 06269-3043

Received May 6, 1999; revised January 19, 2000

INTRODUCTION The ordinal classification of pleurocarpous mosses rests on characters such as branching mode and architecture of Familial definition and generic inclusions in mosses, the peristome teeth that line the mouth of the capsule. The the Bryophyta sensu Vitt et al. (1998), are undergoing Leucodontales comprise mainly epiphytic taxa, character- frequent reevaluation based on morphological (e.g., ized by sympodial branching and reduced peristomes, Buck, 1988, 1994; Hedena¨s, 1994, 1995, 1996a,b) and whereas the Hypnales are primarily terricolous and molecular (e.g., Goffinet and Vitt, 1998; Goffinet et al., monopodially branching. The third order, the Hookeriales, 1998; Cox and Hedderson, 1999) data. Until recently, is defined by a unique architecture of the endostome. We several generations of bryologists have utilized the sampled 78 exemplar taxa representing most families of classification of mosses proposed by Brotherus (1924– these orders and sequenced two chloroplast loci, the trnL- 1925), and for the most part showed little interest in trnF region and the rps4 gene, to test the monophyly and higher level classification (see Vitt et al., 1998). As relationships of these orders of pleurocarpous mosses. Es- classifications have come to be accepted as reflections timates of levels of saturation suggest that the trnL-trnF of phylogeny, older classifications subsequently have spacer and the third codon position of the rps4 gene have come to be viewed as phylogenetically informative reached saturation, in at least the transitions. Analyses of when there is little evidence that the originators even the combined data set were performed under three opti- thought about evolutionary history. These older classi- mality criteria with different sets of assumptions, such as fications have thus become infused with meaning never excluding hypervariable positions, downweighting the intended by their original proponents. Such is the case most likely transformations, and indirect weighting of with pleurocarpous mosses. rps4 codon positions by including amino acid translations. The pleurocarps, those mosses characterized by ex- Multiple parallelism in nonsynonymous mutations led to tensive branching and lateral sporophyte placement, little or no improvement in various indices upon inclusion compose one of the major groups of mosses. Typical of amino acid sequences. Trees obtained under likelihood pleurocarps are immediately recognizable on the basis were significantly better under likelihood than the trees of the elaborate branching pattern alone. The term derived from the same matrix under parsimony. Our phy- “pleurocarpy” refers to the lateral female gametangia logenetic analyses suggest that (1) the pleurocarpous (i.e., the perichaetia) versus the terminal perichaetia mosses, with the exception of the Cyrtopodaceae, form a in acrocarpous mosses (Bridel, 1826–1827). The rela- monophyletic group which is here given formal recogni- tionship of pleurocarpy to cladocarpy, a situation in tion as the Hypnidae; (2) the Leucodontales are at least which the perichaetia are produced terminally on lat- paraphyletic; and (3) the Hypnales form, with most mem- eral branches rather than being sessile, has recently bers of the Leucodontalean grade, a monophyletic group been examined by LaFarge-England (1996). Her study sister to a Hookerialean lineage. The Hypopterygiaceae, was prompted by the presence within typical acrocar- Hookeriales, and a clade composed of Neorutenbergia, pous lineages of isolated occurrences of taxa with fe- Pseudocryphaea, and Trachyloma likely represent a basal male gametangia that are produced laterally and are clade or grade within the Hypnidae. These results suggest sessile or nearly so (e.g., Molendoa, Pottiales). She that mode of branching and reduced peristomes are homo- distinguished pleurocarpy from analogous forms of re- plastic at the ordinal level in pleurocarpous mosses. © 2000 duced cladocarpy based on features of the juvenile Academic Press leaves at the base of the gametangia and of the short axis supporting the gametangia that lacks sub-

180 1055-7903/00 $35.00 Copyright © 2000 by Academic Press All rights of reproduction in any form reserved. MOLECULAR PHYLOGENY OF PLEUROCARPOUS MOSSES 181 perichaetial innovations. Other characters used for de- of the pleurocarpous lineage and the three orders com- fining a clade of pleurocarpous mosses include the posing it. Based on phylogenetic inferences from anal- mode of branching, the presence of pseudoparaphyllia yses of two chloroplast loci, the trnL-trnF region and that surround lateral primordia, and the anatomical the rps4 gene, the following issues are addressed: (1) homogeneity of the costa. Modes of branch develop- Do pleurocarpous mosses form a monophyletic group ment per se (i.e., sympodial or monopodial) are not as defined by Buck and Vitt (1986)? (2) Are the three informative for defining a clade of pleurocarpous orders (Hypnales, Hookeriales, and Leucodontales) as mosses, as both types occur in pleurocarpous (e.g., Leu- defined by Buck and Vitt (1986) and amended by Buck codontales) and acrocarpous (e.g., Orthotrichaceae) (1994), monophyletic? taxa (LaFarge-England, 1996). Buck and Vitt (1986) suggested the presence of pseudoparaphyllia around branch primordia as another synapomorphy for defin- MATERIAL AND METHODS ing the pleurocarpous lineage. Hedena¨s (1994) defined Taxon sampling. Exemplars of most pleurocarp- the monophyly of the pleurocarps based on the homo- ous families (sensu Vitt, 1984 and Buck, 1998) have geneous anatomy of the costa of the leaves. Neither been sampled. For speciose genera one or more rep- definition appears satisfactory since pseudoparaphyl- resentatives were included (Table 1). A special effort lia are lacking is some pleurocarps (ϭpseudopleuro- was made to include taxa of tropical latitudes in carps sensu Buck and Vitt, 1986), and costae are ab- order not to introduce a geographical bias into the sent in many other taxa. sampling. Potential outgroup taxa were selected Pleurocarpous mosses, which as a group lack formal based on the phylogenetic hypothesis proposed by nomenclatural recognition, have long been accommo- Cox and Hedderson (1999) for arthrodontous mosses dated in three orders, the Isobryales (ϭLeucodontales), and on ongoing phylogenetic analyses for all mosses Hookeriales, and Hypnobryales (ϭHypnales). Buck (B. Goffinet et al., unpublished). Vouchers for all and Vitt (1986) defined these orders by characters per- taxa are deposited in the herbarium of the New York taining to the mode of branching and associated leaf Botanical Garden (NY). differentiation between primary and secondary mod- ule, the differentiation of alar cells, the number of DNA extraction, PCR amplification, and sequencing. costae in the leaf, and the architecture of the inner row Apical portions of stems or branches or occasionally oper- of teeth that line the mouth of the sporophyte. The culate capsules were removed from dried herbarium col- Hookeriales are best defined by their endostome with lections or from samples placed in silica in the field. The the baffle-like crosswalls (Buck, 1998). This architec- dry weight of the material used for extraction likely never ture is unique among mosses and likely identifies a exceeded 5 mg. DNA was extracted from plant tissues monophyletic group. Whether this group is restricted following a modification of Doyle and Doyle’s (1987) pro- to the Hookeriales as defined by Buck (1998) remains, tocol. Plant material was ground using a small teflon however, uncertain. pestle in 700 ␮Lof2ϫ hexadecyltrimethylamonium bro- The Leucodontales typically lack well-developed mide (CTAB)–0.2% beta-mercaptoethanol, heated to peristome teeth, branch sympodially and in many 60°C, and incubated at this temperature for at least 30 cases also monopodially, and are epiphytic. Sympodial min. An equal volume of chloroform–isoamyl (24:1) was branching is likely a primitive state in mosses, being added. The emulsified solution was centrifuged for 1 min the dominant mode of branching in all other mosses, at 6500 rpm, and the aqueous phase was transferred to a i.e., the acrocarps. The apparent correlated occurrence new tube to which an equal volume of ice-cold isopropanol of epiphytism and sympodial branching for all Leuc- was added. DNA was precipitated at 0°C for 30–60 min. odontales led Buck (1991) to revise relationships pro- Tubes were centrifuged first for 10 min at 13,000 rpm. posed earlier (Buck and Vitt, 1986) and to suggest that The pellet was washed with 70% ethanol, and the tubes this order likely represents an evolutionary grade were then centrifuged for 3 min at 13,000 rpm. The pellet reached via convergent evolution from various terri- was dried in a vacuum centrifuge and suspended in 30 ␮L colous and monopodially branching taxa of the Hyp- Tris–EDTA (TE; pH 8.0). Amplifications of the trnL-trnF nales. This hypothesis is implicitly based on the as- and rps4 regions were performed in 50-␮L reaction vol- sumption that monopodial branching was acquired umes containing 1ϫ PCR buffer (Gibco BRL), 0.2 mM early in the evolution of the pleurocarpous lineage. The dNTPs in equimolar ratio, 2.5 mM MgCl2, 5% glycerol, hypothesis formulated by Buck (1991) was later sup- 1.0 unit Taq Polymerase (Gibco BRL), and 0.5 mM each ported by Hedena¨s (1995) based on a cladistic analysis primer. The primer sets trnC and trnF (Taberlet et al., of morphological characters. A polyphyletic origin of 1991) and rps5 (Nadot et al., 1994) and trnas (5ЈTAC- the Leucodontales implies that the Hypnales are a CGAGGGTTCGAATC3Ј) were used for the amplification paraphyletic assemblage. of the trnL-trnF region (trnL 5Јexon-trnF) and the rps4 The present study represents the first attempt using gene, respectively. To each reaction 0.5 ␮L of stock DNA molecular characters to elucidate the circumscription was added as template. The PCR was accomplished with 182 BUCK, GOFFINET, AND SHAW

TABLE 1 Taxa for Which the rps4 Gene and the trnL-trnF Region Has Been Included in This Analysis (All Vouchers for Which Sequences Were Generated during the Course of This Study Are Deposited in NY Unless Otherwise Noted; Classiﬁcation Follows Buck and Vitt [1986] Amended by Buck [1994])

GenBank Accession No. Taxon voucher Voucher or reference (rps4/trnL)

Bryales Aulacomniaceae Aulacomnium androgynum (Hedw.) Schwa¨gr. Cox and Hedderson, 1999 AF023811/AF023728 Leptostomum macrocarpum (Hedw.) R. Br. Cox and Hedderson, 1999 AF023790/AF023744 Bartramiaceae Breutelia scoparia (Schwa¨gr.) Jaeg. McDowell 4039 AF143075/AF161168 Bryaceae Leptobryum pyriforme (Hedw.) Wils. Cox and Hedderson, 1999 AF023802/AF023736 Pohlia cruda (Hedw.) Lindb. Cox and Hedderson, 1999 AF023795/AF023760 Cyrtopodaceae Bescherellia cryphaeoides (Mu¨ ll. Hal.) M. Fleisch. Streimann 44233 AF143081/AF161174 *Hypopterygiaceae Hypopterygium tamarisci (Sw.) Mu¨ ll. Hal. Churchill & Betancur 18102 AF143077/AF161170 Mniaceae Mnium hornum Hedw. Cox and Hedderson, 1999 AF023796/AF023767 Phyllodrepaniaceae Phyllodrepanium falcifolium (Schwa¨gr.) Crosby Buck 32969 AF143074/AF161167 Rhizogoniaceae Pyrrhobryum vallis-gratiae (Hampe) Manuel Cox and Hedderson, 1999 AF023825/AF023754 Rhizogonium lindigii (Hamp.) Mitt. Cox and Hedderson, 1999 AF023826/AF023755 Timmiaceae Timmia siberica Lind. & Arnell Cox and Hedderson, 1999 AF023775/AF02715 Hookeriales Adelotheciaceae Adelothecium bogotense (Hampe) Mitt. Buck 26301 AF143073/AF161166 Hookeriaceae Hookeria acutifolia Hook. & Grev. Allen 20123 (MO) AF143071/AF161164 Daltoniaceae Leskeodon cubensis (Mitt.) The´r. Buck 29474 AF143072/AF161165 Leucomiaceae Leucomium strumosum (Hornsch.) Mitt. Buck 33077 AF143068/AF161161 Pilotrichaceae Crossomitrium rotundifolium Herzog Buck 33043 AF143070/AF161163 Lepidopilum scabrisetum (Schwa¨gr.) Steere Buck 33081 AF143066/AF161159 Lepidopilum surinamense Mu¨ ll. Hal. Buck 33082 AF143067/AF161160 Pilotrichum fendleri Mu¨ ll. Hal. Buck 32966 AF143069/AF161162 Leucodontales Cryphaeaceae Cryphaea glomerata Sull. Buck 31329 AF143007/AF161100 Garovagliaceae Garovaglia elegans (Dozy & Molk.) Bosch & Lac. Streimann 40482 AF143017/AF161110 Leptodontaceae Forsstroemia trichomitria (Hedw.) Lindb. Buck 32619 AF143006/AF161099 *Pseudocryphaea domingensis (Spreng.) Buck Djan-Che´kar 94-96 AF143063/AF161156 Lepyrodontaceae Lepyrodon pseudolagurus B. H. Allen Streimann 51300 AF143014/AF161107 Leucodontaceae *Curviramea mexicana (The´r.) Crum Buck 28242 AF143062/AF161155 Leucodon andrewsianus (Crum & Anders.) Reese Buck 32502 AF143005/AF161098 Leucodon brachypus Brid. Buck AF143004/AF161097 Myuriaceae Myurium hochstetteri (Schimp.) Kindb. O’Shea 90B3 AF143018/AF161111 Neckeraceae Neckera pennata Hedw. Buck 32503 AF143008/AF161101 Neckeropsis disticha (Hedw.) Kindb. Buck 33041 AF143010/AF161103 Thamnobryum alleghaniense (Mu¨ ll. Hal.) Nieuwl. Buck 32721 AF143009/AF161102 Phyllogoniaceae Phyllogonium viride Brid. Buck 26430 AF143020/AF161113 MOLECULAR PHYLOGENY OF PLEUROCARPOUS MOSSES 183

TABLE 1—Continued

GenBank Accession No. Taxon voucher Voucher or reference (rps4/trnL)

Prionodontaceae Prionodon densus (Hedw.) Mu¨ ll. Hal. Churchill et al. 19068 AF143076/AF161169 Pterobryaceae Henicodium geniculatum (Mitt.) Buck Buck 33024 AF143011/AF161104 Orthostichopsis tetragona (Hedw.) Broth. Buck 33036 AF143012/AF161105 Pterobryon densum Hornsch. Linares & Churchill 3649 AF143013/AF161106 Ptychomniaceae Glyphothecium sciuroides (Hook.) Hampe Streimann 55558 AF143016/AF161109 Ptychomnion aciculare (Brid.) Mitt. Hiscox 3 AF143015/AF161108 Rutenbergiaceae Neorutenbergia usagarae (Dixon) Bizot & Poćs Poćs et al. 88110/A AF143019/AF161112 Trachylomataceae Trachyloma diversinerve Hampe Streimann 53949 AF143021/AF161114 Hypnales Amblystegiaceae Campylium chrysophyllum (Brid.) Lange Buck 32532 AF143048/AF161141 Hygroamblystegium tenax (Hedw.) C. Jensen Buck 33464 AF143047/AF161140 Sanionia georgico-uncinata (Mu¨ ller) Ochyra & Hedena¨s Anderson 27704 AF143049/AF161142 Anomodontaceae Anomodon rugelii (Mu¨ ll. Hal.) Keissl. Buck 32519 AF143023/AF161116 Haplohymenium triste (De Not.) Kindb. Buck 32601 AF143022/AF161115 Brachytheciaceae Brachythecium austro-glareosum (Mu¨ ller) Paris Anderson 27702 AF143026/AF161119 Brachythecium oxycladon (Brid.) Jaeg. Buck 32504 AF143025/AF161118 Brachythecium plumosum (Hedw.) Schimp. Buck 33463 AF143078/AF161171 Brachythecium salebrosum (Web. & Mohr) Schimp. Goffinet 4723 AF143027/AF161120 Bryhnia novae-angliae (Sull.) Grout Buck 32561 AF143029/AF161122 Homalotheciella subcapillata (Hedw.) Broth. Buck 32517 AF143061/AF161154 Pseudoscleropodium purum (Hedw.) M. Fleisch. Goffinet 4720 AF143030/AF161123 Climaciaceae Climacium americanum Brid. Buck 20951 AF143065/AF161158 Echinodiaceae Echinodium umbrosum (Mitt.) Jaeg. var. glaucoviride (Mitt.) S. P. Churchill Streimann 49668 AF143044/AF161137 Entodontaceae Entodon brevisetus (Wils.) Lindb. Buck 32483 AF143057/AF161150 Fabroniaceae Anacamptodon splachnoides (Brid.) Brid. Buck 32568 AF143031/AF161124 Clasmatodon parvulus (Hampe) Sull. Buck 33446 AF143032/AF161125 Fontinalaceae Fontinalis dalecarlica Bruch & Schimp. Allen 20153 (MO) AF143064/AF161157 Hypnaceae Ctenidium malacodes Mitt. Buck 33458 AF143036/AF161129 Hypnum imponens Hedw. Buck 32496 AF143034/AF161127 Hypnum lindbergii Mitt. Buck 33459 AF143035/AF161128 Isopterygium tenerum (Hedw.) Sw. Buck 33462 AF143037/AF161130 Platygyrium repens (Brid.) Schimp. Buck 33448 AF143038/AF161131 Hylocomiaceae Loeskeobryum brevirostre (Brid.) Broth. Buck 32522 AF143079/AF161172 Rhytidiadelphus squarrosus (Hedw.) Warnst. Goffinet 4719 AF143033/AF161126 Lembophyllaceae Lembophyllum divulsum (Hook.f. & Wils.) Lindb. Streimann 50840 AF143045/AF161138 Pilotrichella flexilis (Hedw.) Ångstr. Buck 26385 AF143046/AF161139 Leskeaceae Haplocladium virginianum (Brid.) Broth. Buck 32482 AF143040/AF161133 Leskea gracilescens Hedw. Buck 30102 AF143042/AF161135 Meteoriaceae Papillaria nigrescens (Hedw.) Jaeg. Mann 32 AF143051/AF161144 Trachypus bicolor Reinw. & Hornsch. Buck 23918 AF143052/AF161145 Zelometeorium patulum (Hedw.) Manuel Buck 33046 AF143050/AF161143 184 BUCK, GOFFINET, AND SHAW

TABLE 1—Continued

GenBank Accession No. Taxon voucher Voucher or reference (rps4/trnL)

Myriniaceae Helicodontium capillare (Hedw.) Jaeg. Buck 29550 AF143043/AF161136 Schwetschkeopsis fabronia (Schwa¨gr.) Broth. Buck 33461 AF143041/AF161134 Plagiotheciaceae Plagiothecium cavifolium (Brid.) Z. Iwats. Buck 32520 AF143080/AF161173 Plagiothecium laetum Bruch & Schimp. Gofﬁnet 4721 AF143058/AF161151 Sematophyllaceae Acroporium pungens (Hedw.) Broth. Buck 33028 AF143028/AF161121 Pylasiadelpha tenuirostris (Sull.) Buck Buck 32500 AF143053/AF161146 Sematophyllum demissum (Wils.) Mitt. Buck 32607 AF143055/AF161148 Taxithelium planum (Brid.) Mitt. Buck 33094 AF143054/AF161147 Trichosteleum papillosum (Hornsch.) Jaeg. Buck 33002 AF143056/AF161149 Stereophyllaceae Entodontopsis leucostega (Brid.) Buck & Irel. Djan-Che´kar 94-726 AF143060/AF161153 Pilosium chlorophyllum (Hornsch.) Mu¨ ll. Hal. Buck 32979 AF143059/AF161152 Theliaceae Thelia lescurii Sull. Buck 32864 AF143024/AF161117 Thuidiaceae Thuidium delicatulum (Hedw.) Bruch & Schimp. Buck 32594 AF143039/AF161132

* Uncertain systematic position. the same temperature profile for both regions: 95°C for 1 Phylogenetic analysis. TrnL-trnF and rps4 data min, 52°C for 1 min, and 72°C for 3 min. After 30 cycles, sets were combined in all analyses. Incongruence of a final 7-min extension at 72°C was performed, and re- data sets results either from distinct evolutionary his- actions were held at 4°C until further processing. PCR tories of two loci or from significant variation of molec- products were screened on 1% agarose gels and, when ular rates within each of the loci. Tests for incongru- successful, were cleaned and concentrated with filter car- ence, such as Farris et al.’s (1995) test, do not tridges (30,000 NMWL Low-Binding Regenerated Cellu- distinguish these sources of incongruence and, further- lose; Millipore). Ten to 50 ng of cleaned PCR product more, may be too conservative when large data sets are served as template in dRhodamine Dye Terminator Cycle compared (Soltis et al., 1998). If higher rates of evolu- sequencing reactions (Perkin–Elmer), performed accord- tion for certain regions result in high levels of ho- 1 moplasy, the topological ambiguities that they cause in ing to the manufacturer’s protocol modified for 4 -size reactions. TrnC and trnF (for the trnL-trnF amplification analyses of independent data sets are likely to apply to product) and rps5 and trnas (for the rps4 amplification distinct portions of the tree. Variation in molecular product) were used as primers in separate sequencing rates in all partitions of the nucleotide sequences was first examined empirically by calculating the propor- reactions, cycled as recommended by the manufacturer. tion of potentially parsimony-informative sites (Table To remove unincorporated reaction components, result- 2). Homogeneity of base composition across taxa was ing 10-mL products were cleaned using Centri-Sep spin examined using Puzzle 3.1 (Strimmer and von Hae- columns (Princeton Separations) containing G-50 Fine seler, 1997). Furthermore, the trnL-trnF intergenic Sephadex. Labeled fragments were separated on poly- spacer, trnL intron, and codon positions one and three acrylamide gels (Long Range Singel; FMC Bioproducts), were examined for saturation. Substitutional equilib- using an ABI Prism 373 or 373 automated DNA se- rium for each character partition, as defined by Ts quencer (Perkin–Elmer). Sequences obtained were edited sat and Tvsat, was estimated using the following formulas using Sequencher 3.0 (Gene Codes Corp.), entered and Tssat ϭ 2f(␲T␲C ϩ ␲A␲G) and Tvsat ϭ 2f␲CT␲AG (Hase- manually aligned, if necessary, in PAUP 4.0b2a (Swof- gawa et al., 1995; Friedrich and Tautz, 1997), where f ford, 1999). Delimitation of the intron, exon, and spacer is set equal to the proportion of variable sites (inferred composing the trnL-trnF product obtained here was done from Table 2). The saturation point was compared to by comparing the sequences with available GenBank ac- the distribution of the probability of transition and cessions. Aligned trnL-trnF sequences were trimmed of transversion per nucleotide derived for each pairwise the 5Јexon of the trnL and the trnF exon, and the first 27 comparison of partitioned sequences. sites (including the annealing site of the primer) and the Using the beta versions 1 and 2 of PAUP 4.0b2a intergenic spacer following the stop codon were excluded (Swofford, 1999), three optimality criteria were used from the rps4 sequences. All sequences obtained in this for reconstructing relationships of the taxa, namely study were submitted to GenBank (Table 1). maximum-parsimony (MP; Fitch, 1971; Table 3, anal- MOLECULAR PHYLOGENY OF PLEUROCARPOUS MOSSES 185

TABLE 2 Distribution of Constant and Phylogenetically Informative Sites, Base Composition, and Bias for the trnL-trnF Region, the rps4 Gene, and Nucleotide Partitions Thereof

TRNL (unambiguously aligned only) RPS4 (codon position) Total Intron Exon Spacer Total First Second Third Total

Size variation 348–372 241–265 49–50 42–57 — — — — Number of sites in matrix 375 268 50 57 570 190 190 190 945 Variation constant 186 129 42 15 305 110 137 58 491 variable-uninformative 69 54 4 11 101 38 20 43 170 parsimony informative 120 85 4 31 164 42 33 89 284 Base composition A 0.38 0.40 0.22 0.42 0.40 0.35 0.38 0.48 0.39 C 0.13 0.11 0.30 0.10 0.13 0.20 0.17 0.03 0.13 G 0.16 0.16 0.20 0.12 0.14 0.21 0.15 0.06 0.15 T 0.33 0.33 0.28 0.36 0.33 0.24 0.30 0.43 0.33 AT/GC 71/29 73/27 50/50 78/22 73/27 59/41 68/32 91/9 72/28 yses I–VIII), maximum-likelihood (ML; Felsenstein, among-site rate variation following a gamma distribu- 1981; Table 3, analyses IX and X), and neighbor-joining tion or breaking down rates of evolution according to (NJ; Saitou and Nei, 1987; Table 3, analyses XI and character partition are computationally intensive for XII). Parsimony analyses were performed with equal large matrices or not feasible at all. Therefore, among- (I–IV) and unequal (V–VIII) weighting. For each par- site variation was incorporated into the HKY-85 model simony analysis searches were replicated either 50 or only through an estimate of the proportion of constant 100 times, with random sequence addition, steepest sites. As a starting tree for heuristic search under the descent off, and branches with identical most-parsimo- likelihood criterion we used a parsimony tree obtained nious reconstruction sets for all characters of incident using the NNI swapping option. A distance parameter nodes collapsed (AMB ϭ option). Parsimony analyses was chosen following recommendations by Kumar et al. were not expected to swap to completion (see (1993). Observation of unequal average base frequen- Soltis et al., 1998). From each replicate, 1000 trees cies and a transition bias suggested inferring the (NCHUCK ϭ 1000) were saved with a maximum of HKY-85 model of sequence evolution for calculating 50,000 trees total (MAXTREES ϭ 50,000). Combining distances (analyses XI and XII). The weighted least the amino acid translations of the rps4 gene sequences squares with power ϭ 2 option was chosen for the with the nucleotide data was used as a method for objective function. The proportion of invariable sites indirectly weighting nonsynonymous mutations com- was estimated via likelihood from the single NNI tree pared to the synonymous or silent mutations (Agosti et used as starting tree in the likelihood analysis above. al., 1996). Weighting was also applied directly to char- This parameter was set in proportion to base frequen- acter transformations by execution of stepmatrices cies estimated for constant sites only, i.e., to 0.4932 for (analyses V–VIII). Based on a randomly chosen most- the full data set (analyses IX and XI) and 0.5726 for the parsimonious Wagner tree, weights were calculated reduced data set (analyses X and XII). Sites with miss- from estimates obtained from PAUP 4.0b2a (Swofford, ing data, gaps, and ambiguities were not distributed 1999) as the product of the mean instantaneous sub- proportionally, but rather ignored only in affected pair- stitution rate and the relative rate parameters corre- wise sequence comparisons. sponding to each possible substitution type (R-matrix; Jackkniﬁng analyses (Farris et al., 1996; with 50% see Swofford et al., 1996). Natural logs of these prod- deletion) were performed under the parsimony crite- ucts were used in the stepmatrices. A stepmatrix was rion, whereas bootstrap analyses were done for dis- generated for each of the six character partitions (i.e., tance trees. Relative support within most-parsimoni- the trnL intron, exon, and spacer and the three codon ous Fitch trees was further estimated for branches positions of the rps4 gene) based on rate parameters deﬁning relationships at the base of the pleurocarpous estimated from the same most-parsimonious tree de- lineage by Bremer support analysis (Bremer, 1994), rived from the combined data set. Internally inconsis- following the converse constraint method (Baum et al., tent stepmatrices were adjusted automatically upon 1994). Neighbor-joining and unweighted parsimony execution in PAUP, to satisfy the triangle of inequality. analyses were run on a Power MacIntosh 8600/200, The Hasegawa et al. (1985) model of sequence evo- and likelihood analyses were performed on an Ultra 60 lution (hereafter HKY-85) was inferred when using the with dual 360 ultrasparc processors. likelihood criterion (analyses IX and X). Inferences of In total, 12 analyses were run (Table 3). For each set 186 BUCK, GOFFINET, AND SHAW

TABLE 3 Summary of Tree Statistics for the Analyses Performed Using the Complete Nucleotide Sequences, the Reduced Sequences (i.e., with the trnL-trnF Intergenic Spacer and the Third Codon Position of the rps4 Gene Excluded), or Either Set Complemented by the Amino Acid Translation of the rps4 Gene

Parsimony Unweighted Weighted Likelihood Neighbor-Joining

Number of analysis I V IX XI Number of characters 945 945 945 945 Number of trees 19000 24 335 Tree score 1584 2330 9946.18983 16.078 Ϯ 6.633 Complete sequences Consistency Index* 0.3237 0.3267 Retention Index 0.5145 0.5225 Rescaled consistency 0.2098 0.2128 Number of analysis II VI Number of characters 1135 1135 Number of trees 1150 7000 Tree score 1882 2636 Complete sequences ϩ Consistency Index* 0.3260 0.3269 amino acids Retention Index 0.5254 0.5268 Rescaled consistency 0.2155 0.2148 Number of analysis III VII X XII Number of characters 698 698 698 698 Number of trees 33000 42000 2975 Tree score 877 1303 5968.35354 21.739 Ϯ 7.713 Reduced sequences Consistency Index* 0.3499 0.3414 Retention Index 0.5575 0.5674 Rescaled consistency 0.2492 0.2473 Number of analysis IV VIII Number of characters 888 888 Number of trees 26000 21000 Tree score 1174 1607 Reduced sequences ϩ Consistency Index* 0.3473 0.3394 amino acids Retention Index 0.5635 0.5665 Rescaled consistency 0.2486 0.2446

Note. Tree scores are number of steps for parsimony analyses, likelihood scores under the HYK85 model, and weighted least squares with power ϭ 2 for the distance analyses (see text for full description of assumptions). * Calculated with autapomorphies excluded. of assumptions (analyses I–XII), topologies congruent the Kishino and Hasegawa test (1989). The effect of with Buck and Vitt’s (1986) ordinal classification of enforcing the molecular clock onto the model of se- pleurocarpous mosses were generated (100 replicates, quence evolution was examined using the log likeli- TBR swapping, keeping five trees per replicate, each hood ratio test (Kishino and Hasegawa, 1989). compatible with a constraint of monophyly for the Hookeriales, Leucodontales, and Hypnales sensu Buck RESULTS and Vitt, 1986, and no specified relationships between the orders). Tree scores of constrained trees obtained Sequence variation. The polymerase chain reaction under each parsimony assumption were compared to yielded a single amplification product for the trnL-trnF that of unconstrained trees by nonparametric testing region and for the rps4 locus for all taxa included in (Templeton, 1983). Although the number of con- this study. Sequences for 86 taxa were used for phylo- strained trees was compared to an equal number of genetic analyses. The length of the trnL-trnF product unconstrained trees taken from the first trees in the varied between 349 and 432 nucleotides (nts), with unconstrained tree file, PAUP recognizes the best trees most of the variation found in the intron of the trnL (i.e., the first unconstrained most-parsimonious tree) gene (Table 2). Alignment of the sequences required and compares all constrained trees to it. Constraints the inclusion of numerous gaps, yielding a final size of were not enforced in actual likelihood searches, but 580 sites (final matrix is available at http://www. instead the most-parsimonious trees satisfying the herbaria.harvard.edu/treebase/index.html). Alignment constraint were described under the likelihood model. within one large indel area (92 sites) was not fully The likelihood scores obtained for trees generated us- resolved (Table 4) and this portion of the sequences ing the same set of characters were compared under was omitted from the analyzes. Several smaller indels MOLECULAR PHYLOGENY OF PLEUROCARPOUS MOSSES 187

TABLE 4 Portion of the trnL Operon between Positions 237 and 353 in the Aligned Matrix

222222222222222222222222222222222222222222222222222222222222222333333333333333333333333333333333333333333333333333333 333444444444455555555556666666666777777777788888888889999999999000000000011111111112222222222333333333344444444445555 Taxon 789012345678901234567890123456789012345678901234567890123456789012345678901234567890123456789012345678901234567890123

Timmia siberica AAGATTT-TATT-AATTAAGTAATTAATTAATAAAAA-TAAATTTAACT------AAAATTTTATA- Leptobryum pyriforme AAAACTT-ATTA-GTTGAATTTATTAAT-AATAATAAACAAATTTATT------AAAAA------Mnium hornum AAATTTT-ATTTCTATTAATAAAT------AAAATTTCATTTT Pohlia cruda AAATCTT-ATTTCTATTAATAAAT------AAAATTTTATTTT Leptostomum macrocarpum AAATTTT-ATTT-AATTAAAATTTATTTTAATTTTAT------TTC------TAATTTCATTTT Phyllodrepanium falcifolium CAATTTT-ACTT-AT------TTATTT------AATTTTTATTTT Breutelia scoparium AAAACTT-ATTT-AATT------GAATAATGGAAATAAAAAAA------TTACTT------AAATTTTTCTTTT Bescherellia cryphaeoides AAAACTT-ATTT-AATT-AAATTTAATTTATTAATTAATATATTA------ATTAATAGAAGTAAAA------ATTT------AAAA-TTCATTT Aulacomnium androgynum AAAAATT-ATTT-AATTT-AATTTTATTT------ATTAATATAAGTAAAAAAAAATT----TTATTT------AAAATTAATTTT Pyrrhobryum vallis- gratiae AAAATTT-ATTT-AATTTT-ATTTCATTTATTAATAGAAATAAAAAA------ATTAATAGAAATAAAAAAATT----TTATTT------AAAATTAATTTT Rhizogonium lindigii TGAATTT-AATT------GAA-TAAAAATAAATTTC------ATTAATAGAAGTAAAAAAAAT----TTATTT------AAAATTTATTTT Hypopterygium tamariscii AAATTTT-ATTT-AATCAAATGA------GTAAAAAAATGAAAATTATTAAA------ATTTTATTTTTT Hookeriales Leskeodon cubense AAAGTTT-ATTT-AATT-AAATTT------AATCTATTAAATAA------TAAAAAA------TTATTTT------AATTTAAATTTT Adelothecium bogotense AAAGTTT-AGTT-AATT-AAATTT------CATTT-AATTAA------CTAATAAAAA------TAAAAAAT------TTAAATAA------ATTTTTTATTTT Lepidopilum scabrifolium AAATATT-ATTG-AGATAAAATGT-----ATATCTTT-AATAAATACAATCTATTAAAAAA--AAATAATAAAAAGAA------AAAAATATTTTT Lepidopilum surinamense AAATATT-ATTG-AGTTAAAATTT-----ATATCTTT-ACTAAATACAATATATTAAAAAATAATATAATAAAAAGAA------AAAAATATTTTT Leucomium strumosum AAAGTGT-ATTT-AGTT-AAATTT------TATTT-AGTCAATAAATAAATAG------ATTAGAAAAAGT------TATTTTT Pilotrichum fendleri AAAACTT-ATTT-AATT-AAATTG------TCTTTTAGTCAATAAATAAAATT------TAATAAAAGTAAA------AAATTATTTTT Hookeria acutifolia ---GAAT-ATTT-AGCTAATATTGAT------TAATACAATAAAAGTAAAAAAAAAA---TTATTTAAATTTTATTTTTTATATTTT- Crossomitrium rotundifolium AAAGCTT-ATTT-AATT-AAATTTAATTTATTA------ACTAAGAAAAGTAAAAAAAAAAA--TTA------AATAATTTTT Hypnales and Leucodontales Pseudocryphaea domingensis AAAACTT-ATTT-ACTTTTATTA------ATTAATAAAAGTAAAAAAAAA----TTAT------AAAATTTATTTT Neorutenbergia usagarae AAAACTT-ATTT-AATTT-ATTA------ATTAATAAAAGTAAAAAAAA-----TTATTT------AAAATTTATTTT Trachyloma diversinerve AAAACTT-ATTT-AATTT-ATTA------ATTAATAAAAGTAAAAAAAA-----TTATTT------AAAATTTATTTT Ptychomnion aciculare ------T-ATTT-AATTT------AAATTAAATTTT Glyphothecium sciuriodes ------T-ATTT-AATTT------AAATTGAATTTT Garovaglia elegans ----TAT-ATTT-AATTT------AGATTGA-TTTT Acroporium pungens ----ATT-ATTTATTTTAAGTTTTATTT------AAAATTTATTTT Sematophyllum demissum ----ATT-ATTTATAATAAATTTTATTT------AAAATTTATTTT Trichosteleum papillosum ------ATTT----CAAATTTTATTT------AAAATTTATTTT Fontinal dalicarlica ACAAATT-ATTACATTTTTTAAAATTATTA------AAAATATAGTTT Other Leucodontales and Hypnales (56 taxa) TAAAATT-ATTT------AAAATTTATTTT

Note. Sites that are underlined were excluded from all phylogenetic analyses. Framed areas highlight portion shared by putative primitive pleurocarps with “closely related” outgroup taxa (Figs. 2–4), but absent in most Hypnales and Leucodontales. Clear box defines insertion found in putatively derived Leucodontales (Fontinalis) or Hypnales (Sematophyllaceae). were also omitted because of difficulties in assessing (Table 2). Complete rps4 sequences were obtained for positional correspondence (regions likely to result from the same set of taxa, with the exception of Curviramea duplication of contiguous sites) or simply for these mexicana, Entodontopsis leucostega, and Lepidopilum being autapomorphic and thus uninformative. Three scabrisetum. Sequences of the rps4 products were hundred and seventy-five sites were retained for the readily aligned within their coding region, requiring no phylogenetic analyses. Base frequencies were nearly at insertion of gaps. As expected based on previous stud- equilibrium in the 3Ј exon but strongly biased in favor ies of chloroplast protein-coding genes (e.g., Goffinet et of As and Ts in the intron and the spacer (Table 2). al., 1998), the observed variation is highest in the third Within the unambiguously alignable portions of the codon position and the least in the second position sequences slightly over half the sites were variable, (Table 2). Base composition bias also differed among and of these nearly two-thirds (Ϯ32% of total) ap- the three codon positions, with the strongest bias ob- peared phylogenetically informative. Most (71%) po- served in the third codon position (Table 2). Although tentially informative sites were found within the in- base frequencies varied between partitions of both the tron (22% of total characters), although a higher trnL-trnF region and the rps4, base composition within proportion of sites in the spacer (54% of spacer sites) each partition appeared homogenous among taxa (P Ͼ included potential synapomorphic transformations 0.95). The spacer of the trnL-trnF region and the third 188 BUCK, GOFFINET, AND SHAW

FIG. 1. Observed average transition (Ts) and transversion (Tv) divergence per nucleotide in the trnL-trnF spacer (A) and the third codon positions of the rps4 gene (B). White squares correspond to sequence comparisons involving at least one outgroup taxon; black diamonds represent comparisons between ingroup taxa. Crosses mark the expected level at which saturation in transitions and transversions is reached. codon position partition of the rps4 gene carried a high used. In terms of absolute patristic distances, the percentage of potential phylogenetically informative Hookeriales, Ptychomniaceae, Garovagliaceae, Phyllo- characters (Table 2). Saturation in transitions and drepanium (Bryales), and to a lesser degree the Sema- transversions was estimated to be at 0.127 and 0.366, tophyllaceae appeared most derived from both the out- and 0.052 and 0.345, respectively, for the trnL-trnF groups and most ingroup taxa. spacer and the third codon positions of the rps4 gene. Weighting alternative character transformations re- Transversions did not reach saturation in either data sulted in no or only minimal improvement (2% of the partition. In contrast, the number of transitions ex- original value, at most) of the indices when the full ceeds in many pairwise comparisons the estimated data set was analyzed (Table 3). When the indices were value of saturation in the spacer of the tnrL region and calculated from equally weighted transformations, the the third codon positions of the rps4 gene (Fig. 1). Of indices for the trees of analyses V and VI were lower the 178 comparisons exceeding the Tssat estimate for than those of the unweighted analyses (I and II). When the spacer region, 140 involved only ingroup taxa. Sim- stepmatrices were invoked for the reduced data set, the ilarly, 58% of the pairwise comparisons based on third indices were slightly lower, except for the retention codon positions were characterized by a level of tran- index whose value increased slightly. The addition of sitions exceeding the estimated saturation point, and amino acid sequences resulted in slightly higher indi- of these only 25% of the comparisons included at least ces when the complete sequences were analyzed with one outgroup taxon. The overall level of saturation is equal or differential weighting of alternative transfor- likely even underestimated, as multiple hits may yield mations and when the reduced data set was analyzed homoplasies or reversals and thus decrease the patris- with all transformations equally likely (Table 3). Dif- tic distance between two taxa, when in fact these taxa ferential weighting transformations and inclusion of may only be distantly related. amino acid sequences led to slightly smaller indices. Phylogenetic analyses. Tree scores and statistics The exclusion of hypervariable regions resulted in all for all 12 analyses are summarized in Table 3. Likeli- four analyzed indices improving by at least 7% of their hood analyses were aborted as swapping failed to com- original value (Table 3). Furthermore, attributing zero plete after running for over 60 days (analysis IX after weight to the characters of this partition had a more 1656 h while swapping on tree 62 of 335, and analysis pronounced effect on the retention index, which mea- X after 1542 h while swapping on tree 146 of 2975). sures congruence among the data (Maddison and Mad- Robustness of branches as measured by the jackknife dison, 1999), than weighting nonsynonymous muta- percentage and the Bremer support index was low or tions by including amino acid sequence. completely lacking for most branches under the parsi- Since likelihood analyses were not allowed to run to mony criterion: Bremer support varied from 0 to 1 (3 in completion, little can be said about the effect of exclud- one case; Fig. 2), whereas Jackkniﬁng and bootstrap ing the spacer from the trnL-trnF sequences and the values were in all cases below 60%. Branch lengths third codon positions from the rps4 sequences, except were in general very short for most branches deﬁning that under the reduced set of data the optimal recon- relationships around the base of the pleurocarpous lin- struction is twice as likely than under the full set of eage. Terminal branches leading to the OTUs were in characters. Enforcing the molecular clock under as- contrast generally longer, regardless of the data set sumptions IX and X yielded likelihood estimates that MOLECULAR PHYLOGENY OF PLEUROCARPOUS MOSSES 189

FIG. 2. Summary of 50% majority rule trees calculated from optimal topologies obtained under assumptions of analyses I, II, III, V, VII, IX (see Table 3). Bremer support is indicated for trees obtained under unweighted parsimony (I–III). Shaded boxes identify pleurocarpous lineages. were significantly worse in both cases (Ϫ2 log ⌳ϭ mony (analyses I and II) or the complete DNA se- 457.80 and 361.06, respectively; P Ӷ 0.001). Analyses quences alone in likelihood analyses (analysis IX), the under the distance criterion yield identical topologies Ptychomniaceae–Garovagliaceae clade occurred sister with respect to higher level relationships, regardless of to Aulacomnium, an acrocarpous bryalean taxon, with the data set used. These phenetic topologies are iden- which it formed a sister group to the pleurocarpous tical to those obtained when the same models of evolu- lineage (Fig. 2). Pseudopleurocarpous mosses (i.e., tion were inferred under likelihood. “pleurocarpous” taxa lacking pseudoparaphyllia; see Phylogenetic relationships. In all analyses (I–XII), Buck and Vitt, 1986), here represented by Bescherellia, the pleurocarpous taxa (i.e., the Hypnales, Leucodon- are always nested among acrocarps, except when com- tales, and Hookeriales) always form a monophyletic plete sequences are analyzed under the distance crite- group. One exception regards the position of the Pty- rion (XI), in which case Bescherellia is sister to the chomniaceae–Garovagliaceae clade (Figs. 2–4). Using pleurocarps and could potentially be considered part of the full nucleotide sequences alone or in combination the pleurocarpous clade. According to all analyses the with the amino acid sequences in unweighted parsi- affinities of the Hypopterygiacae lie within the pleuro- 190 BUCK, GOFFINET, AND SHAW MOLECULAR PHYLOGENY OF PLEUROCARPOUS MOSSES 191 carpous lineage. One analysis suggests that the Hy- maining pleurocarps. In contrast, in analyses II, V, VI, popterygiaceae are a sister taxon to the pleurocarps, in and VIII, the Hookeriales were nested within the Leuc- which case they can be interpreted as acrocarps, as odontalean–Hypnalean clade. A clade composed of suggested by Buck and Vitt (1986). In all other analy- Pseudocryphaea, the Rutenbergiaceae, and the Trachylo- ses the affinities of the Hypopterygiaceae lay primarily mataceae is shown in all analyses to have some relation- with the Hookerialean clade (Figs. 2–4). ship with the Hookeriales (Figs. 2–4). This clade forms The relationships at the base of the pleurocarpous either a sister group to the Hookeriales (e.g., analyses II lineage differ depending on the data partitions in- and IX) or a grade basal to the Hookeriales (analyses IV, cluded in the analyses and the optimality criterion V, VI, and VIII) or a transitional group between the used to infer relationships (Fig. 2). In analyses II and Hookeriales and the remaining pleurocarps (analyses X, VIII Pterobryon, Fontinalis, Leucodon, Cryphaea, and XI, and XII). Prionodon form a grade at the base of the pleurocarp Trees from analyses enforcing a constraint of mono- clade. Differential weighting of alternative transfor- phyly for the three orders Leucodontales, Hookeriales, mations (analysis V) resulted in two main lineages and Hypnales were generally significantly (P Ͻ 0.05%) with the Hookerialean clade rooted with Prionodon longer that the most-parsimonious unconstrained trees. and Leucodon (Leucodontales). In most analyses (I, III, Under assumption set III, the Templeton test yielded VII, IX, X, XI, and XII) the basal dichotomy distin- ambiguous results for rejecting the monophyly of the guishes the Hookeriales sensu lato (see below) from the orders; only 2 of 15 constrained trees were significantly Leucodontalean–Hypnalean clade. This dichotomy oc- longer than the unconstrained trees. Tree length of most curs in the strict consensus of analyses of the full and likely trees (IX and X) were not significantly different reduced data sets under the likelihood and unweighted from those of the most-parsimonious topologies (I and parsimony, as well as upon analyzing the reduced data III), whereas the likelihood scores of the latter were sig- set with differential weighting applied to alternative nificantly worse according to the Kishino and Hasegawa transformations. The Hookerialean lineage includes, test. in many optimal topologies, a clade composed of the Ptychomniaceae–Garovagliaceae and another clade or DISCUSSION grade comprising Pseudocryphaea–Neorutenbergia– Trachyloma, both assemblages traditionally consid- Phylogenetic analyses of the combined trnL-trnF and ered of Leucodontalean affinity. rps4 sequences yield optimal trees that consistently The Leucodontales included here never form a mono- reject a systematic concept of three major monophy- phyletic assemblage (Figs. 3 and 4). Even when the Pty- letic lineages of pleurocarpous mosses. The consensus chomniaceae–Garovagliaceae clade and the Pseudocry- of these trees, within or among analyses, reveals, how- phaea–Neorutenbergia–Trachyloma clade are excluded, ever, significant incongruence among characters and the remaining Leucodontalean taxa fail to assemble into leaves the circumscription of and thus the relation- a monophyletic group (Figs. 3 and 4). Instead, the Leuc- ships among the emerging lineages ambiguous. Under odontales are completely scattered among Hypnalean a certain set of assumptions incongruence among trees taxa. Both orders are of para- or polyphyletic nature. The may reflect inadequate taxon sampling or significant Hookerialean taxa (excluding the Hypopterygiaceae) homoplasy in character transformations. Pleurocarps form either a monophyletic (analyses I, II, IV, IX, XI, and compose a diverse group of mosses, distributed among XII) or a paraphyletic (analyses III, V, VI, VII, VIII, and approximately 40 families (Vitt, 1982; Buck and Vitt, X) assemblage (Figs. 2–4). Monophyly of the group ap- 1986). Taxa selected for this study represent at least 34 pears thus to be primarily supported by inclusion of vari- of these families and include all of the most diverse able partitions of the sequences (analyses I, II, IX, and genera. Furthermore, as familial concepts of pleuro- XI). Paraphyly of the Hookeriales is always due to the carps are often labile, two genera were chosen to rep- inclusion of the clade comprising the Ptychomniaceae resent diverse families. In total, 66 ingroup species and Garovagliaceae, which have traditionally been con- were sampled for this study. Finally, taxa were se- sidered of Leucodontalean alliance (e.g., Vitt, 1984; Buck lected to encompass the broad cosmopolitan distribu- and Vitt, 1986). The affinities of the Hookeriales are tion of the pleurocarpous mosses, to avoid a geograph- ambiguous when all trees are considered. In at least some ical bias in the analyses. Considering one of the issues trees retained in analyses I, III, VII, IX, X, XI, and XII, that we are addressing (i.e., monophyly of the major the Hookerialean clade, including the Ptychomniaceae– orders of pleurocarpous mosses), the sampling cer- Garovagliaceae clade, formed a sister group to the re- tainly appeared more than appropriate. As suggested

FIG. 3. Phylogram of 1 of 2975 most likely trees inferred from sequences of the trnL-trnF region and the rps4 gene from which the spacer and third codon position have been excluded, respectively (analysis X). Shaded boxes identify traditional Leucodontales. 192 BUCK, GOFFINET, AND SHAW MOLECULAR PHYLOGENY OF PLEUROCARPOUS MOSSES 193 below, taxa that are patristically distant may compose characterize just pleurocarpous taxa rather than the key elements for reconstructing basal dichotomies, and acrocarpous outgroup taxa. clearly future attempts to reconstruct the phylogeny The application of different optimality criteria, dif- should emphasize such morphologically odd taxa in the ferent weights to alternative transformations or differ- sampling. ent weights to particular data partitions, may account Incongruence among optimal trees may also result for incongruence of phylogenetic signals recovered un- from high levels of homoplasy, creating noise that over- der different sets of assumptions. Analysis of the full shadows the phylogenetic signal of the data. Both the data set using either the parsimony or the likelihood trnL-trnF region and the rps4 gene comprise partitions criterion yields trees or at least a majority thereof that that differ in their degree of variation. A priori exam- resolve two main lineages within the ingroup (analyses ination of the variability of the sequence partitions I and IX), as does the tree obtained under the distance (trnL intron, trnL 3Ј exon, trnL-trnF spacer, and all criterion (XI). These trees differ primarily in the affin- three codon positions of the rps4 gene) suggest that 70 ities of two small clades, one of which is composed of and 73% of the third codon positions and the intergenic three taxa defined by long branches. When hypervari- spacer, respectively, are variable (Table 2). The prob- able regions were excluded the two major lineages were ability of homoplastic mutations may be affected by recovered in the strict consensus trees (III and X), and differences in the probability of alternative transfor- one of the smaller clades with previously ambiguous mations, with the more likely transformations between affinities was consistently resolved as part of one of the the character states having a greater likelihood of oc- major lineages (Fig. 2). Weighting character transfor- curring in independent sites (see discussion below re- mations significantly decreased the incongruence garding “weighted parsimony”). The degree of ho- among most-parsimonious trees when the full data set moplasy is also a function of the overall variability of was analyzed, but also resulted in the paraphyly of one the sequence. As the number of variable sites in- major lineage (analyses I and V; see Fig. 2). In con- creases, the probability for a single site to mutate more trast, when analyzing the reduced data set, the overall than once, and thus for homoplasy to occur, increases. topology of two main lineages was not altered when Comparisons of an estimate of the level of transition weights were applied to specific transformations (anal- and transversion at saturation against the distribution yses III and VII; see Fig. 2). Among the four parsimony of transitions and transversions per site for each pair of analyses of nucleotide sequences, (I, III, V, and VII) the taxa suggest, indeed, that some sequences are satu- application of weights to transformations for all char- rated in transitions (Fig. 1). One may expect that the acters yields the most congruence among trees. Should probability of multiple hits increases in parallel to the this congruence be interpreted as reflecting improved time of divergence and thus that saturation is expected accuracy of the data or convergence toward the wrong particularly when comparing ingroup and outgroup topology? Since we do not know the phylogeny, the taxon sequences. From Fig. 1 it is, however, apparent accuracy of the various phylogenies cannot be rigor- that saturation is reached by ingroup taxa, as compar- ously addressed. Nevertheless, the following discus- isons within the ingroup indicate possible levels of sion is warranted. Topological congruence among trees transitions higher than the estimated saturation level. obtained under various models may reveal a certain This observation may be explained by differences in robustness of the data and thus be considered indica- evolutionary rates, that is, in this case, in higher rates tive of the actual signal carried by the data. Further- of molecular evolution in at least some ingroup lin- more, it should be noted that the weighting scheme eages. Enforcing the molecular clock (i.e., identical used here may be inadequate due to a heterogeneity in rates of evolution for all taxa) results in a significantly rates of variation among the characters. Indeed, worse likelihood estimate of the optimal tree, suggest- weights were derived to account for differences in like- ing that, indeed, rates of evolution differ among lin- lihoods of transformations within character partitions eages. Reconstruction of phylogenetic relationships rather than differences in evolutionary rates among among major lineages of mosses (i.e., primarily acro- the partitions themselves. In other words, the least carpous mosses) based on rps4 and trnL-trnF se- likely and thus uncommon transformations for the quences revealed little incongruence among characters three codon partitions were attributed identical (Cox and Hedderson, 1999), suggesting that the in- weights, regardless of rate differences between the creases in evolutionary rates invoked here may indeed codon positions. Consequently, certain transforma-

FIG. 4. 50% Majority rule consensus tree of 2975 most likely trees inferred from sequences of the trnL-trnF region and the rps4 gene from which the spacer and third codon position have been excluded, respectively (analysis X). Dashed lines represent branches not present in the strict consensus tree. Characters coded are peristome (present versus absent), primary branching mode (sympodial versus monopodial; sensu Buck and Vitt, 1986), and habitat (terricolous versus epiphytic), with the first state represented by an open circle and the second state by a filled circle. Shaded boxes identify traditional Leucodontales. 194 BUCK, GOFFINET, AND SHAW tions within hypervariable regions were weighted more The lack of saturation does, however, not preclude than transformations within highly conserved charac- multiple hits occurring at any given site. Indeed, the ters, thereby accentuating the negative effect of at interpretation of the saturation level is dependent on least some homoplasic changes. The higher congruence the amount of overall variable sites across all lineages. among optimal trees in weighted parsimony (V) should A given number of nucleotide differences between two therefore tentatively be regarded as an artifact of the taxa may be close or remote to the level of saturation, weighting scheme rather than reflecting an increase in depending on the number of variable sites observed the accuracy of the data. across all sequences. In other words, a given distance Agosti et al. (1996) proposed to compensate for the in transitions and transversions between two se- effect of multiple hits in the third codon positions un- quences approaches saturation level as the proportion der parsimony by complementing the nucleotide se- of variable sites across all taxa decreases. The proba- quences by their amino acid translations, thereby in- bility of any site mutating is statistically independent directly downweighting synonymous mutations in the of the past history of this site. As time of divergence third codon position. When the protein sequences were increases, the likelihood that a site has changed in- added to the full data set, the effect was minimal creases but the probability of change per unit of time is (Table 3), and when the reduced data set was comple- the same throughout the elapsed time period, unless mented by these sequences, the consistency index even the rates of evolution change. Consequently, estimat- decreased. The matrix composed solely of amino acid ing the level of saturation may allow for a preliminary sequences comprises 27% of phylogenetically informa- assessment of the degree of homoplasy to be expected tive sites, and of these 75% remained potentially infor- but cannot provide an accurate prediction on the extent mative within the ingroup. Parsimony analysis of this of parallelism. Estimates of saturation in transitions in data set resulted in an extensive polytomy involving the third codon positions lend support to complement- most of the terminal ingroup taxa (only the Hookeria- ing nucleotide sequences with protein sequences. Phy- les and a group of Brachytheciaceae were resolved as logenetic analyses reveal, however, extensive ho- monophyletic; result not shown), revealing significant moplasy in amino acids, suggesting that their inclusion conflict and thus homoplasy among characters (ho- in addition to nucleotide sequences (e.g., analysis II) or moplasy index ϭ 0.63 and 0.61 for ingroup taxa alone). in replacement of third codon position sites (e.g., anal- Homoplasy in amino acids arises in various ways. Most ysis IV) is not appropriate for our study. transitions or transversions in the first or second codon Maximum-likelihood methods are considered robust position translate into changes of the amino acid. If to violations of various assumptions, including changes parallel nonsynonymous changes occur at either codon in evolutionary rate among lineages. The model of evo- position in two lineages, such parallelism would be lution invoked here (HKY-85) integrates differences in weighted more by including the amino acid sequences. likelihood between transitions and transversions, un- Changes between a serine and a leucine or a proline, equal base frequencies, and variation among sites in for example, require two transversions, one in the first evolutionary rates. Likelihood reconstructions yielded and one in the second codon position. Some sites of the optimal trees that were not significantly longer than protein sequence are composed primarily of leucine alternative hypotheses obtained directly under parsi- (including in outgroup taxa) with serine characterizing mony but that are significantly better in terms of their several a priori unrelated taxa (e.g., at codon 57: likelihood scores compared to the most-parsimonious Glyphothecium, Pilosium, Climacium, Ctenidium, and trees. Inclusion of the hypervariable partitions had Neckeropsis). These taxa thus share at least two con- little effect on the overall topology reconstructed under vergent transversions; complementing the nucleotide the likelihood criterion (analyses IX and X). The consequences with the protein sequences will accentuate sensus trees in both cases resolve the pleurocarps as this degree of homoplasy by introducing another con- composed of two major lineages (see below). The only vergent character, the shared amino acid. For 9 of the difference between the analyses is the relationship of 20 amino acids, transversions in the third codon posi- the clade composed of the Ptychomniaceae and Garo- tion also result in the coding for another amino acid. vagliaceae. Analysis of the full data set yields this The introduction of the amino acid sequences will ac- clade as a sister group to Aulacomnium, an outgroup centuate the negative effect of any parallelism in these taxon. When the trnL-trnF spacer and the third codon sites on the phylogenetic signal. positions are excluded, the pleurocarps form a mono- Extensive parallelism in amino acids resulting in phyletic lineage. This inconsistency may represent a ambiguous relationships among ingroup lineages is a phylogenetic artifact due to long-branch attractions to priori surprising: estimates of saturation levels sug- which maximum-likelihood may be susceptible (Sid- gest that neither the first nor the second position is dall, 1998). The Ptychomniaceae and Garovagliaceae saturated in transitions or transversions (not shown) are, indeed, characterized by the most divergent (pa- and that the level of transversions in the third codon tristically distant) sequences. Maximum-likelihood is position also has not reached saturation level (Fig. 1). generally considered immune to long-branch attraction MOLECULAR PHYLOGENY OF PLEUROCARPOUS MOSSES 195 as it tends to favor parallelism in mutations as the lowed Robinson (1971), excluding the family from the length of a branch increases (Swofford et al., 1996). pleurocarps, suggesting instead affinities to the Raco- Considering that Aulacomium and the Ptychomni- pilaceae and even to the Rhizogoniaceae. All analyses aceae and Garovagliaceae tend, regardless of the as- except one analysis (IX) resolve Hypopterygium as a sumptions, to occur near the base of the pleurocarpous member of a pleurocarpous lineage (Fig. 2), as sug- lineages (albeit on either side of the transition) it is gested by Hedena¨s (1994) and LaFarge-England (1996) possible that under these circumstances two long based on morphological characters. branches may be resolved as sister branches. Since LaFarge-England (1996) recently suggested that trees generated under maximum-likelihood appear sig- Cryphaea (Cryphaeaceae) may be pivotal to the evolu- nificantly more likely than those obtained under par- tion of pleurocarpy. Several species of Cryphaeaceae simony, and considering that the inclusion of hyper- exhibit features of the perichaetial branches that are variable regions has possibly resulted in phylogenetic considered typical of cladocarpy, rather than pleuro- artifacts, our discussion will refer primarily to infer- carpy (LaFarge-England, 1996). The occurrence of ences made from the reduced data set under the like- cladocarpy in various bryalean lineages lends credence lihood criterion (Figs. 3 and 4). to the hypothesis of the Cryphaeaceae being basal in a Circumscription of a pleurocarpous lineage. Most pleurocarpous lineage. None of the reconstructions ob- analyses presented here suggest that the Hypnales, tained from our analyses, however, yields Cryphaea in Leucodontales, and Hookeriales, with the exception of such a basal position. The development of short Bescherellia (Cyrtopodaceae), together form a mono- branches supporting the perichaetia in these Crypha- phyletic group, identifiable as the pleurocarps. The eaceae is best interpreted as a convergence to clado- hypothesis that the Hookeriales are of independent carpy. Consequently, the development of sessile lateral origin relative to the remainder of the pleurocarps perichaetia is likely a synapomorphy for the Hypnidae, (Buck, 1998) is not supported by our data. A clade but further studies are needed to distinguish it either composed of pleurocarpous mosses currently lacks for- morphologically or ontogenetically from analogous lat- mal nomenclatural subclass recognition (Vitt et al., eral perichaetia in the Cyrtopodaceae and allied fami- 1998). Using Vitt et al.’s (1998) classification as a lies. framework, we propose the name Hypnidae Buck, Ordinal circumscription. All reconstructions of Goffinet, and Shaw subclass nova (plantae pleurocar- chloroplast DNA evolution used here do not resolve a pae, pseudoparaphyllia plerumque praesentia, anato- monophyletic concept of the Leucodontales and Hyp- mia costae ubi praesentis homogenea; typus: Hypnum nales. A monophyletic Leucodontales are rejected un- Hedw.) to represent this clade. Bescherellia is the only der all but one set of assumptions (i.e., analysis III). genus rather unambiguously excluded from the Hypni- Furthermore, unconstrained topologies resolving a dae (Fig. 2). A sister group relationship to the pleuro- polyphyletic Leucodontales are significantly shorter carps was suggested only when all data partitions were under the Templeton test. The Leucodontales have included under the distance criterion (analysis XI). been characterized by sympodial branching, reduced Under all other assumptions, Bescherellia was sepa- peristomes, and epiphytic habitats (Buck, 1998). In rated from the pleurocarps by at least one typical ac- contrast, the Hypnales have monopodial branching, rocarpous bryalean genus, i.e., Aulacomnium. These well-developed peristomes, and terrestrial habitats. latter results are congruent with those obtained by Our analyses show the Leucodontales composing a Withey (1996) based on rbcL sequences, whereby Be- grade rather than a clade and thus support the hypoth- scherellia was included in a clade of pseudopleurocar- esis raised by Hedena¨s (1995) based on morphological pous mosses (see Buck and Vitt, 1986) composed of the characters. Spiridentaceae, Hypnodendraceae, Racopilaceae, and The polyphyly of the Leucodontales is accentuated Rhizogoniaceae that was nested within a lineage of further by the close relationship between some of its acrocarpous taxa. Consequently, pleurocarpy as de- taxa (i.e., Neorutenbergia, Pseudocryphaea, Trachy- fined by LaFarge-England (1996) is not homologous loma, Ptychomnium, Glyphothecium, and Garovaglia) between the Hypnidae and the Cyrtopodaceae and pos- and the Hookeriales (Figs. 3 and 4). The Ptychomni- sible other so-called pseudopleurocarpous mosses. Con- aceae and Garovagliaceae are resolved either sister to sequently, pleurocarpy as expressed in these distinct Aulacomnium (Bryales) or nested within the Hookeria- clades results from convergent evolution and should be les. The latter association is reconstructed under like- treated as such when included in phylogenetic analy- lihood when all characters are included, whereas the ses. exclusion of hypervariable partitions resolves the al- The Hypopterygiaceae were considered by Fleischer ternative topology. Under parsimony, alignment (1904–1923), Brotherus (1924–1925), Walter (1983), within the Hookeriales also required the exclusion of and Vitt (1984) to be closely related to the Hookeri- these partitions or indirect weighting of codon posi- aceae sensu lato, whereas Buck and Vitt (1986) fol- tions of the rps4 via inclusion of amino acid sequences. 196 BUCK, GOFFINET, AND SHAW

In terms of patristic distance based on absolute dis- The Hookeriales are characterized by the unique tance, a sister group relationship of Aulacomnium with architecture of the inner peristome lining the capsule the Ptychomniaceae–Garovagliaceae may be an exam- mouth. The horizontal walls separating the inner cells ple of long-branch attraction under parsimony (Hendy of this endostome are left in place and form baffle-like and Penny, 1989) or long-branch repulsion under like- walls, much like a longitudinal section of a bamboo lihood (Siddall, 1998). Affinities to the Hookeriales are culm (Buck, 1998). Taxa that share this attribute form surprising, but have been suggested recently based on a cohesive clade in most analyses of nucleotide se- analysis of morphological characters (Hedena¨s, 1995). quences. Although the order never appears polyphyl- Phylogenetically more significant may be the posi- etic, satisfying the criterion of monophyly may under tion of the Rutenbergiaceae, Trachylomatacaeae, and some assumptions require extending the current cir- Pseudocryphaea. These three taxa are resolved either cumscription of the order. Three taxa or groups of taxa as a monophyletic group or as a tight paraphyletic are resolved in at least some analyses as allied to the assemblage at the base of either the Hookeriales or the Hookeriales, namely Hypopterygium, a clade composed Hypnales–Leucodontales. Affinities to the Hookeriales of Neorutenbergia–Pseudocryphaea–Trachyloma, and or a basal position of these taxa within the Hypnalean– one comprising Ptychomnium–Glyphothecium–Garo- Leucodontalean clade may be supported by these taxa vaglia (see above). None of these taxa or their relatives lacking an unambiguous deletion characteristic of the shares the endostomial architecture of the Hookeria- Leucodontales and Hypnales (Table 4). Although the les. Testing a close relationship of the Hypopterygi- alignment within this region among the outgroup taxa, aceae with the Hookeriales further, Hedena¨s (1996) the Hypopterygiaceae, the Hookeriales, and the above suggested that this family is nested within the order. three families is not unambiguous, at least portions of This hypothesis is not supported by our data. The the sequences are alignable, appearing homologous be- results of Hedena¨s (1996) may have been biased by the tween these taxa. Exclusion of the hypervariable par- hypnalean outgroups chosen for his analysis. In terms titions tends to resolve this group of families as sister of strong affinities to the Hookeriales, the Hypopteri- or basal to the Hypnalean–Leucodontalean clade. giaceae are resolved in our analyses, at best, as a sister A polyphyletic Leucodontales may reflect multiple group to the former and may thus represent a basal independent colonizations of an entirely new habitat taxon within the Hookeriales. Following Hedena¨s following the advent of angiospermous tropical forests (1996), this expanded hookeriaceous clade could be de- in the Cretaceous. Similar explosive radiation has been fined by as many as 18 morphological character states postulated for the Lejeuneaceae (Hepaticophyta), most considered to occur exclusively or almost exclusively of which are epiphytic or epiphyllous in humid tropical among the Hookeriales and Hypopterygiaceae. The forests (Schuster, 1966), as well as for obligatory epi- plain endostomial crosswalls may thus have originated phytic pteridophytes (Rothwell, 1996; Wikstro¨m and during the differentiation of the Hookeriales sensu Kenrick, 1997). Leucodontales are completely lacking stricto. The Neorutenbergia clade is another potential from the pre-Cretaceous fossil record (Oostendorp, ingroup taxon of the Hookeriales, although it too does 1987), and pleurocarps in general are only question- not share the peristomial architecture of the Hookeria- ably present prior to the Cretaceous (Busche, 1968; les. A basal position for the Hypopterygiaceae, as well Krasilov and Schuster, 1984; Oostendorp, 1987). Pleu- as for the Neorutenbergia clade, in the Hypnidae or rocarpous mosses are likely the most derived major within a hookerialean lineage may be supported by lineage of mosses, as inferred from both morphological these taxa lacking the unambiguous deletion (Table 4, (Vitt, 1984) and molecular (Capesius and Stech, 1997; characters 249–341) present in the trnL intron of the Cox and Hedderson, 1999) characters, and are of fairly Hypnales and Leucodontales (see above). recent origin (i.e., Cretaceous). Furthermore, the relationships among major lineages within the Hypnidae CONCLUSIONS are poorly resolved. Most of these branches are very short and are based on one or two nucleotide substitu- Sequences of the rps4 gene and of the trnL-trnF region tions. Low character support for basal branches and comprise a significant percentage of variable characters. high levels of homoplasy may result from ancient di- Exclusion of hypervariable character partitions that are vergence under a stochastic mode of evolution or from estimated to be saturated in at least their transitions rapid radiation accompanied by an increase in the sub- allows for a rather consistent phylogenetic signal to stitution rate (Lanyon, 1988). The second causation emerge under either maximum-parsimony or likelihood. appears more plausible considering the general lack of These analyses resolve one hookerialean and one hyp- pre-Cretaceous fossils. Investigations into the best pos- nalean clade. Although further work is needed to address sible sister group to the pleurocarps are currently un- the familial circumscription of these clades, we propose to derway in our laboratories. These studies should even- retain the Hookeriales as a distinct order sister to the tually permit formal tests of acceleration of relative clade comprising the Hypnales and Leucodontales. The rates of evolution. recognition of the Hypnales and the Leucodontales, if MOLECULAR PHYLOGENY OF PLEUROCARPOUS MOSSES 197 circumscribed anything like that in Brotherus (1925) or acrotonous. Nevertheless, if changes in branching types even Buck (1998), is untenable. In all analyses the genera have occurred as a consequence of becoming epiphytic, traditionally placed in the Leucodontales, based on sym- the scattered distribution of epiphytes in the Hypnidae podial branching, differentiated stem and branch leaves, implies that some of these character states are bound to reduced peristomes, and epiphytic habitat (e.g., Leuc- be homoplasic. Sequence data have allowed us to test the odon, Cryphaea, Neckera, Henicodium, Pterobryon, and circumscription of pleurocarpous lineages and to reject Papillaria), are scattered among the genera of Hypnales. the ordinal concepts based on simple concepts of complex Our sampling does not allow for familial circumscription morphological characters. Homology of characters or and relationships to be addressed extensively, as the their states, however, cannot be addressed solely on the primary objective was to test the monophyly of the Hyp- basis of gene trees. Based on the hypothesis that these nales and Leucodontales. Nomenclaturally, the Hypnales traits are not homologous among taxa that share them, and Leucodontales have equal priority. With the merging critical reexaminations of these morphological characters of the two orders, a single name must be chosen to des- are needed if the gene trees are to be used to understand ignate the group. We choose Hypnales. Putative basal character evolution in this diverse group of mosses. taxa within the Hypnidae, i.e., the Hypopterygiaceae and Rutenbergiaceae clade, may need to be accommodated in ACKNOWLEDGMENTS a third and new order. At present we retain them as taxa of uncertain affinities. Financial support from Duke University and NSF Grant DEB- Early suprageneric classifications of mosses are based 9806955 to Jon Shaw and Bernard Goffinet are acknowledged. Cliff on few characters (e.g., Bridel, 1826–1827; Brotherus, Cunningham kindly provided assistance with the construction of the 1924–1925). Mode of branch development has been a stepmatrices. Dave Swofford kindly allowed us to use test versions of the upcoming PAUP software. Finally, Sandra Boles and Randy central character in the ordinal classification of pleuro- Downer are gratefully acknowledged for providing help during all carpous mosses. Most if not all Leucodontales sensu Buck stages of this study. and Vitt (1986; amended Buck, 1994) can be characterized by the following growth and branching pattern: a REFERENCES stem often with monopodial branching assumes ortho- tropic growth; upon cessation of apical meristematic ac- Agosti, D., Jacobs, D., and DeSalle, R. (1996). On combining protein tivity basal innovations are developed that will assume sequences and nucleic acid sequences in phylogenetic analysis: the function of a stem. In the Hypnales sensu stricto The homeobox protein case. Cladistics 12: 65–82. growth relies almost exclusively on monopodial branch- Baum, D. A., Systma, K. J., and Hoch, P. C. (1994). A phylogenetic ing, and sympodial branching is mostly lacking. As dis- analysis of Epilobium (Onagraceae) based on nuclear ribosomal cussed above the distinction of two orders on the basis of DNA sequences. Syst. Bot. 19: 363–388. primary branching types is phylogenetically untenable. Bremer, K. (1994). Branch support and tree stability. Cladistics 10: 295–304. Furthermore, all our analyses suggest that neither state Bridel, S. E. (1826–1827). “Bryologia Universa seu Systematica ad defines a major lineage within the Hypnales (including Novam Methodum Dispositio, Historia et Descriptio Omnium the Leucodontales) and that even within the hookeri- Muscorum Frondosorum Hucusque Cognitorum cum Synonymia alean lineage, if the Ptychomniaceae or Garovagliaceae ex Auctoribus Probatissimis.” Vols. 1 and 2, Barth, Leipzig. (Fig. 4) were included, the type of branching is not con- Brotherus, V. F. (1924–1925). Musci (Laubmoose). In “Die Natu¨ rli- stant. Instead, either or both sympodial and monopodial chen Pflanzenfamilien” (A. Engler, Ed.), Vols. 10 and 11, 2nd ed. branching are homoplasic within the Hypnidae. The phy- Engelmann, Leipzig. logenetic significance of this character seems thus rather Buck, W. R. (1988). Another view of familial delimitation in the limited. Furthermore, since both states are not exclusive Hookeriales. J. Hattori Bot. Lab. 64: 29–36. Buck, W. R. (1991). The basis for familial classification of pleurocar- in pleurocarpous mosses, they may not be homologous pous mosses. Adv. Bryol. 4: 169–185. states of a single character; instead two characters each Buck, W. R. (1994). A new attempt at understanding the Meteori- with two or more states should be considered. Indeed aceae. J. Hattori Bot. Lab. 75: 51–72. more relevant to the evolution of pleurocarpous taxa may Buck, W. R. (1998). Pleurocarpous mosses of the West Indies. Mem. be the absence of a mode of branching, as evolution from N. Y. Bot. Gard. 82: 1–400. a polymorphic taxon could proceed via a loss of either Buck, W. R., and Vitt, D. H. (1986). Suggestions for a new familial state. Other informative characters may be retrieved by classification of pleurocarpous mosses. Taxon 35: 21–60. examining the position of monopodial or sympodial Busche, R. (1968). Als Laubmoosreste gedeutete Pflanzenfossilien branches (acrotonous versus basitonous and cauline ver- aus den Lebacher Schichten (Autunien) von St. Wendel, Saar. sus axillary; see LaFarge-England, 1996). Buck and Vitt Argumenta Palaeobot. 2: 1–14. (1986), for example, considered true pleurocarps to have Capesius, I., and Stech, M. (1997). Molecular relationships within mosses based in 18S rRNA gene sequences. Nova Hedwigia 64: cauline branches, whereas in acrocarps and pseudopleu- 525–533. rocarps (e.g., Bescherellia) the branches are in an axillary Cox, C. J., and Hedderson, T. J. (1999). Phylogenetic relationships position. Sympodial branches in the Hypnidae are always among the ciliate arthrodontous mosses: Evidence from chloro- basitonous, whereas in some acrocarps they can also be plast and nuclear DNA sequences. Plant Syst. Evol. 215: 119–139. 198 BUCK, GOFFINET, AND SHAW

Cunningham, C. W. (1997). Is congruence between data partitions a Phylogeny and Character, version 3.08. Evolution. Sinauer, Sun- reliable predictor of phylogenetic accuracy? Empirically testing an derland, MA. iterative procedure for choosing among phylogenetic methods. Mitten, W. (1859). Musci Indiae Orientalis: An enumeration of the Syst. Biol. 46: 464–478. mosses of the East Indies. Proc. Linn. Soc. Bot. Suppl. 1: 1–171. Doyle, J. J., and Doyle, J. L. (1987). A rapid DNA isolation procedure Nadot, S., Bajon, R., and Lejeune, B. (1994). The chloroplast gene for small quantities of fresh leaf tissue. Phytochem. Bull. 19: rps4 as a tool for the study of Poaceae phylogeny. Plant Syst. Evol. 11–15. 191: 27–38. Eyre-Walker, A. (1998). Problems with parsimony in sequences of Oostendorp, C. (1987). The bryophytes of the Paleozoic and the biased base composition. J. Mol. Evol. 47: 686–690. Mesozoic. Bryophyt. Biblioth. 34: 1–112. Farris, J. S., Kallersjo, M., Kluge, A. G., and Bult, C. (1995). Testing Robinson, H. (1971). A revised classification for the orders and fam- significance of incongruence. Cladistics 10: 315–319. ilies of mosses. Phytologia 21: 289–293. Farris, J. S., Albert, V. A., Ka¨llersjo¨, M., Lipscomb, D., and Kluge, Rothwell, G. W. (1996). Phylogenetic relationships of ferns: A palaeo- A. G. (1996). Parsimony jackknifing outperforms neighbor-joining. botanical perspective. In “Pteridology in Perspective: Proceedings Cladistics 12: 99–124. of the Holttum Memorial Pteridophyte Symposium, Kew 1995” Felsenstein, J. (1981). Evolutionary trees from DNA sequences: A (J. M. Camus, M. Gibby, and R. J. Johns, Eds.), pp. 395–404. Royal maximum likelihood approach. J. Mol. Evol. 17: 368–376. Botanic Gardens, Kew. Fitch, W. M. (1971). Toward defining the course of evolution: Mini- Rzhetsky, A., and Nei, M. (1995). Test of applicability of several mum change for a specific tree topology. Syst. Zool. 20: 406–416. substitution models for DNA sequence data. Mol. Biol. Evol. 12: Fleischer, M. (1904–1923). “Die Musci der Flora von Buitenzorg 131–151. (Zugleich Laubmoosflora von Java),” Vol. 1–4, Brill, Leiden. Saitou, N., and Nei, M. (1987). The neighbor-joining method: A new method for reconstructing phylogenetic trees. Mol. Biol. Evol. 4: Friedrich, M., and Tautz, D. (1997). Evolution and phylogeny of the 406–425. Diptera: A molecular phylogenetic analysis using 28S rDNA sequences. Syst. Biol. 46: 674–698. Schuster, R. M. (1966). “The Hepaticae and Anthocerotae of North America,” Vol. 1, Columbia Univ. Press, New York. Goffinet, B., and Vitt, D. H. (1998). A revised classification of the Orthotrichales (Bryopsida) based on comparative morphology and a Siddall, M. E. (1998). Success of parsimony in the four-taxon case: rbcL based phylogeny. In “Bryology for the Twenty-First Century: Long-branch repulsion by likelihood in the Farris zone. Cladistics Proceedings of the Centenary Symposium of the British Bryological 14: 209–220. Society” (J. W. Bates, N. W. Ashton, and J. G. Duckett, Eds.), pp. Soltis, D. E., Soltis, P. S., Mort, M. E., Chase, M. W., Savolainen, V., 143–159. Maney and The British Bryological Society, Leeds. Hoot, S. B., and Morton, C. M. (1998). Inferring complex phylog- Goffinet, B., Bayer, R. J., and Vitt, D. H. (1998). Circumscription and enies using parsimony: An empirical approach using three large molecular phylogeny of the Orthotrichales (Musci) based on se- DNA sets for angiopserms. Syst. Biol. 47: 32–42. quences of the chloroplast gene rbcL. Am. J. Bot. 85: 1324–1337. Strimmer, K., and von Haeseler, A. (1997). PUZZLE: Maximum Hasegawa, M., Kishino, H., and Yano, T.-A. (1985). Dating the hu- Likelihood Analysis for Nucleotide and Amino Acid Alignments, man–ape splitting by a molecular clock of mitochondrial DNA. J. version 3.1. Zoological Institute, University of Munich, Germany. Mol. Evol. 22: 160–174. Swofford, D. L., Olsen, G. J., Waddell, P. J., and Hillis, D. M. (1996). Hedena¨s, L. (1994). The basal pleurocarpous diplolepidous Phylogenetic inference. In “Molecular Systematics” (D. M. Hillis, C. mosses—A cladistic approach. Bryologist 97: 225–243. Moritz, and B. K. Mable, Eds.), pp. 407–514. Sinauer, Sunderland, MA. Hedena¨s, L. (1995). Higher taxonomic level relationships among diplolepidous pleurocarpous mosses—A cladistic overview. J. Swofford, D. L. (1999). PAUP4.0b2a: Phylogenetic Analysis Using Bryol. 18: 723–781. Parsimony (and Other Methods), version 4.0. Sinauer, Sunder- land, MA. Hedena¨s, L. (1996a). A cladistic evaluation of relationships between the Hookeriales, the Sematophyllaceae and some other taxa. Lind- Taberlet, P., Gielly, L., Patou, G., and Bouvet, J. (1991). Universal bergia 21: 49–82. primers for amplification of three non-coding regions of chloroplast DNA. Plant Mol. Biol. 17: 1105–1109. Hedena¨s, L. (1996b). A cladistic overview of the “Hookeriales.” Lind- bergia 21: 107–143. Templeton, A. R. (1983). Phylogenetic inference from restriction endonuclease cleavage site maps with particular reference to the Hendy, M. D., and Penny, D. (1989). A framework for the quantita- humans and apes. Evolution 37: 221–244. tive study of evolutionary trees. Syst. Zool. 38: 297–309. Vitt, D. H. (1984). Classification of the Bryopsida. In “New Manual of Kishino, H., and Hasegawa, M. (1989). Evaluation of the maximum Bryology” (R. M. Schuster, Ed.), Vol. 2, pp. 676–759. The Hattori likelihood estimate of the evolutionary tree topologies from DNA Botanical Laboratory, Nichinan. sequence data, and the branching order in Homonoidea. J. Mol. Vitt, D. H., Goffinet, B., and Hedderson, T. A. (1998). The ordinal Evol. 29: 170–179. classification of the mosses: Questions and answers for the 1990’s. Krassilov, V. A., and Schuster, R. M. (1984). Paleozoic and Mesozoic In “Bryology for the Twenty-First Century: Proceedings of the fossils. In “New Manual of Bryology” (R. M. Schuster, Ed.), Vol. 2, Centenary Symposium of the British Bryological Society” (J. W. pp. 1172–1193. The Hattori Botanical Laboratory, Nichinan. Bates, N. W. Ashton, and J. G. Duckett, Eds.), pp. 113–123. Maney Kumar, S., Tamura, K., and Nei, M. (1993). MEGA: Molecular Evo- and The British Bryological Society, Leeds. lutionary Genetic Analysis, version 1.0. Computer program dis- Walther, K. (1983). Bryophytina, Laubmoose. In “A. Englers Sylla- tributed by Pennsylvania State University, University Park, PA. bus der Pflanzenfamilien: Dreizehnte Vollsta¨ndig Neubearbeitete LaFarge-England, C. (1996). Growth form, branching pattern, and Auflage” (A. Gerloff and J. Poelt, Eds.), Berlin. perichaetial position in mosses: Cladocarpy and pleurocarpy rede- Withey, A. (1996). “Systematic Studies of the Spiridentaceae fined. Bryologist 99: 170–186. (Musci),” Ph.D. dissertation, Duke University, Durham, NC. Lanyon, S. M. (1988). The stochastic mode of molecular evolution: What Wikstro¨m, N., and Kenrick, P. (1997). Phylogeny of Lycopodiaceae consequences for systematic investigation? Auk 105: 565–573. (Lycopsida) and the relationships of Phylloglossum drummondii Maddison, W. P., and Maddison, D. R. (1999). MacClade: Analysis of Kunze based on rbcL sequences. Int. J. Plant Sci. 158: 862–871.