YMPEV 5234 No. of Pages 9, Model 5G 3 July 2015
Molecular Phylogenetics and Evolution xxx (2015) xxx–xxx 1 Contents lists available at ScienceDirect
Molecular Phylogenetics and Evolution
journal homepage: www.elsevier.com/locate/ympev
5 6
3 Phylogeny and species delimitations in European Dicranum q 4 (Dicranaceae, Bryophyta) inferred from nuclear and plastid DNA
a,b,⇑ c a,b 7 Annick Lang , Gaëlle Bocksberger , Michael Stech
8 a Naturalis Biodiversity Center, P.O. Box 9517, 2300 RA Leiden, The Netherlands 9 b Leiden University, Leiden, The Netherlands 10 c Department of Primatology, Max Planck Institute for Evolutionary Anthropology, Deutscher Platz 6, 04103 Leipzig, Germany
1211 13 article info abstract 2915 16 Article history: DNA sequences have been widely used for taxonomy, inferring phylogenetic relationships and identifying 30 17 Received 28 July 2014 species boundaries. Several specific methods to define species delimitations based on molecular phyloge- 31 18 Revised 24 June 2015 nies have appeared recently, with the generalized mixed Yule coalescent (GMYC) method being most 32 19 Accepted 29 June 2015 popular. However, only few studies on land plants have been published so far and GMYC analyses of 33 20 Available online xxxx bryophytes are missing. Dicranum is a large genus of mosses whose (morpho-)species are partly 34 ill-defined and frequently confused. To infer molecular species delimitations, we reconstructed 35 21 Keywords: phylogenetic trees based on five chloroplast markers and nuclear ribosomal ITS sequences from 27 out 36 22 Bryophyte of 30 species occurring in Europe. We applied the species delimitation methods GMYC and Poisson tree 37 23 DNA-based taxonomy 24 Species boundaries processes (PTP) in order to compare their discriminatory power with species boundaries inferred from 38 25 Morphological plasticity the molecular phylogenetic reconstructions and with the morphological species concept. Phylogenetic 39 26 Chloroplast markers circumscriptions were congruent with the morphological concept for 19 species, while eight species were 40 27 nrITS molecularly not well delimited, mostly forming closely related species pairs. The automated species 41 28 delimitation methods achieved similar results but tended to overestimate the number of potential 42 species and exposed several incongruences between the morphological concept and inference from 43 molecular phylogenetic reconstructions. It is concluded that GMYC and PTP methods potentially provide 44 a useful and objective way of delimiting bryophyte species, but studies on further bryophyte data sets are 45 necessary to infer whether incongruences might ensue from evolutionary processes and to test the suit- 46 ability of these approaches. 47 Ó 2015 Published by Elsevier Inc. 48 49
50 51 52 1. Introduction on the differences in branching rates at species and population 64 levels, assuming that the number of substitutions within a species 65 53 DNA sequence data are widely used for inferring species delim- is significantly lower than between species. The main difference is 66 54 itations and phylogenetic relationships. Specific methods to ana- that GMYC requires an ultrametric tree that relies on a Bayesian 67 55 lyze species boundaries based on molecular phylogenetic tree sampling using MCMC methods to fit both Yule and coales- 68 56 reconstructions without prior species information, however, have cence models (Hudson, 1990; Yule, 1925) and finally delimit evo- 69 57 been developed only recently (cf. Carstens et al., 2013 for review). lutionary species units (ESU; Tang et al., 2014). PTP, in contrast, 70 58 Most popular is the generalized mixed Yule coalescent (GMYC) uses directly the number of substitutions (instead of time) to sim- 71 59 method (Fontaneto et al., 2007; Monaghan et al., 2009; Pons ulate speciation and coalescent events, and ESU delimitations are 72 60 et al., 2006), while the Poisson tree processes (PTP) method has based on heuristic search algorithms to estimate species bound- 73 61 recently been proposed by Zhang et al. (2013) as an alternative aries with the maximum likelihood scores (Tang et al., 2014; 74 62 to GMYC. Both methods estimate the point of transition between Zhang et al., 2013). Because PTP does not require the input tree 75 63 species and population, i.e. they infer species boundaries based to be ultrametric, the method is much less computing-intensive 76 than GMYC. 77 Generally, automated species delimitation methods are consid- 78 q This paper was edited by the Associate Editor Elizabeth Zimmer. ered especially useful in organisms with unclear species boundaries, 79 ⇑ Corresponding author at: Naturalis Biodiversity Center, P.O. Box 9517, 2300 RA due to poor taxonomic knowledge or signals in phylogenetic recon- 80 Leiden, The Netherlands. structions being obscured by lineage sorting or introgression 81 E-mail address: [email protected] (A. Lang).
http://dx.doi.org/10.1016/j.ympev.2015.06.019 1055-7903/Ó 2015 Published by Elsevier Inc.
Please cite this article in press as: Lang, A., et al. Phylogeny and species delimitations in European Dicranum (Dicranaceae, Bryophyta) inferred from nuclear and plastid DNA. Mol. Phylogenet. Evol. (2015), http://dx.doi.org/10.1016/j.ympev.2015.06.019 YMPEV 5234 No. of Pages 9, Model 5G 3 July 2015
2 A. Lang et al. / Molecular Phylogenetics and Evolution xxx (2015) xxx–xxx
82 (O’Meara, 2010 and references therein). Most GMYC studies so far three D. flagellare Hedw., 11 D. flexicaule Brid., four D. fragilifolium 144 83 focused on different animal groups (e.g. Poulakakis et al., 2012; Lindb., six D. fuscescens Turner, two D. groenlandicum Brid., 11 D. 145 84 Zaldívar-Riverón et al., 2010) and very few examples of analyses of laevidens R.S. Williams, three D. leioneuron Kindb., eight D. majus 146 85 other organisms such as algae (e.g. Leliaert et al., 2009), fungi (e.g. Turner, four D. montanum Hedw., four D. polysetum Sw., 65 D. sco- 147 86 Parnmen et al., 2012) and land plants (e.g. Hernández-León et al., parium Hedw., two D. scottianum Turner ex R. Scott, nine D. septen- 148 87 2013) have been published. GMYC analyses of bryophyte datasets trionale,15D. spadiceum J.E. Zetterst., three D. spurium Hedw., four 149 88 are still missing. Bryophytes are an important component of terres- D. tauricum Sapjegin, four D. undulatum Schrad. ex Brid. and four D. 150 89 trial ecosystems and count up to 18,000 known species (Goffinet and viride (Sull. and Lesq.) Lindb. specimens. Fourty specimens were 151 90 Shaw, 2009). The limited number of morphological characters avail- newly sequenced for all six markers employed here, except four 152 91 able, high morphological plasticity, and often broad geographical specimens which ITS sequences have been generated by 153 92 distributions pose serious problems on species delimitations and Tubanova et al. (2010), Ignatova and Fedosov (2008). The other 154 93 taxonomy in many bryophyte lineages. Therefore, species delimita- 162 specimens were sequenced for previous studies (Lang and 155 94 tion methods such as GMYC and PTP could potentially make an Stech, 2014; Lang et al., 2014a, 2014b; Stech, 1999; Stech et al., 156 95 important contribution to delimit bryophyte species and evaluate 2006). We chose as outgroup four specimens of Holomitrium, sister 157 96 the significance of morphological characters for species identifica- genus of the Dicranum s.l. clade (La Farge et al., 2002; Stech et al., 158 97 tion, but their performance on bryophyte datasets remains to be 2006). 159 98 tested. While molecular data can facilitate the circumscription of 99 (closely related) bryophyte species, (e.g. Dong et al., 2012; 2.2. DNA extraction, amplification and sequencing 160 100 Hedenäs and Eldenäs, 2007; Heinrichs et al., 2009; Stech et al., 101 2013), multiple DNA markers are often required in order to obtain The greenest parts of single gametophyte stems were selected 161 102 supported species delimitations due to low levels of genetic variabil- for DNA extraction and cleaned manually with demineralised 162 103 ity (Hollingsworth et al., 2009, 2011; Lang et al., 2014a), which may water under a binocular. Total DNA extraction was carried out 163 104 pose a problem on the accuracy of species delimitation methods. using the NucleoSpinÒ Plant II Kit (Macherey–Nagel, Düren, 164 105 Species circumscription and identification in the Holarctic moss Germany). Six markers employed to delimit closely related 165 106 genus Dicranum (Dicranaceae, Bryophyta) has been notoriously dif- Dicranum species in Lang and Stech (2014), Lang et al. (2014a, 166 107 ficult. The genus counts more than 90 species (www.tropicos.org; 2014b) were amplified and sequenced, i.e. five chloroplast regions 167
108 Frey and Stech, 2009), many of which are broadly distributed and (partial rpoB gene, trnHGUG-psbA, rps19-rpl2, rps4-trnTUGU and 168 109 display a great range of morphological plasticity, with only few trnLUAA–trnFGAA intergenic spacer) and the nuclear ribosomal 169 110 habitat-specific species (Hedenäs and Bisang, 2004). Moreover, nrITS1-5.8S-ITS2 region. PCR amplifications were performed as 170 111 Dicranum and related genera display little molecular variation described in Lang and Stech (2014). All PCR products were purified 171 112 (Cox et al., 2010; La Farge et al., 2002; Stech, 1999; Stech et al., and sequenced at Macrogen Inc. (www.macrogen.com). GenBank 172 113 2012). Thus, assessing species delimitations in Dicranum is chal- accession numbers of all sequences are listed in Appendix 1. 173 114 lenging both at the morphological and molecular level. Our recent 115 studies on the Dicranum scoparium and D. acutifolium species com- 2.3. Alignment and phylogenetic reconstruction 174 116 plexes (Lang and Stech, 2014; Lang et al., 2014b) as well as on 117 boreal-arctic Dicranum species (Lang et al., 2014a) showed that Sequences were aligned in Geneious v5.3.6 (Biomatters, 2010) 175 118 in several cases conclusive species delimitations could only be using 65% similarity matrix costs, and manually adjusted. Short 176 119 obtained from combined analyses of several chloroplast markers hairpin-associated inversions in the trnH-psbA spacer, which can 177 120 and nuclear ribosomal ITS sequences. flip at the population level and may significantly reduce phyloge- 178 121 The present study aims to elucidate species boundaries within netic structure if undetected (Borsch and Quandt, 2009; Quandt 179 122 Dicranum on a broad geographic scale, including 27 of the 29 and Stech, 2004; Whitlock et al., 2010), were positionally separated 180 123 Dicranum species occurring in Europe (Hedenäs and Bisang, in the alignment and the corresponding indels were excluded. 181 124 2004) plus D. septentrionale Tubanova and Ignatova, a newly Phylogenetic inferences were based on maximum likelihood 182 125 recorded species in Scandinavia (Lang et al., 2014b). Molecular (ML) and Bayesian inference (BI) analyses. Gaps were coded as 183 126 phylogenetic reconstructions based on five chloroplast markers informative by a simple indel coding strategy (SIC) (Simmons 184
127 (trnHGUG-psbA, rps4-trnTUGU and trnLUAA–trnFGAA intergenic spac- and Ochoterena, 2000) implemented in SeqState (Müller, 2004). 185 128 ers, rps19-rpl2, rpoB) plus the nrITS1-5.8S-ITS2 region will be used To check for incongruence, phylogenetic reconstructions based 186 129 to test, to our knowledge for the first time in bryophytes, the con- on chloroplast and nuclear sequences were visually compared. In 187 130 gruence of two automated species delineation approaches, the addition, an incongruence length difference test (ILD, Farris et al., 188 131 general mixed Yule-coalescent (GMYC) and Poisson tree processes 1994) as implemented in PAUP⁄ 4.0b10 (Swofford, 2002) was per- 189 132 (PTP) methods. formed with 100 replicates. As both visual inspections and the ILD 190 test indicated that the plastid and nuclear tree topologies were 191 congruent (p = 0.06), the two datasets were combined. 192 133 2. Material and methods Three nucleotide partitions were used in ML and BI, namely the 193 non-coding chloroplast markers (rps4-trnT, trnL–trnF, trnH-psbA, 194 134 2.1. Sampling rps19-rpl2), the coding chloroplast region rpoB and the nuclear 195 ribosomal ITS region. Partitions were unlinked in both ML and BI 196 135 A total of 202 Dicranum specimens were sampled (Appendix 1), inferences. ML analyses were carried out with RAxML v.7.2.6 197 136 representing 27 species of the 29 European species recognized by (Stamatakis, 2006) employing the graphical user interface 198 137 Hedenäs and Bisang (2004) and including the new European spe- raxmlGUI v.0.93 (Silvestro and Michalak, 2012) with the default 199 138 cies record of D. septentrionale (Lang et al., 2014b): six Dicranum GTR model of nucleotide substitution and C rate heterogeneity 200 139 acutifolium (Lindb. and Arnell) C.E.O. Jensen, nine D. angustum for all partitions. Bootstrap analyses under ML were done using 201 140 Lindb., six D. bonjeanii De Not., five D. brevifolium (Lindb.) Lindb., the thorough bootstrap heuristics algorithm with 20 runs and 202 141 three D. canariense Hampe ex Müll.Hal., five D. crassifolium 1000 replicates. BI analyses were run on the CIPRES science gate- 203 142 Sérgio, Ochyra and Séneca, one D. dispersum Engelmark, one way (Miller et al., 2010). Bayesian posterior probabilities were cal- 204 143 D. drummondii Müll.Hal., four D. elongatum Schleich. ex Schwägr., culated based on the Markov chain Monte Carlo (MCMC) method, 205
Please cite this article in press as: Lang, A., et al. Phylogeny and species delimitations in European Dicranum (Dicranaceae, Bryophyta) inferred from nuclear and plastid DNA. Mol. Phylogenet. Evol. (2015), http://dx.doi.org/10.1016/j.ympev.2015.06.019 YMPEV 5234 No. of Pages 9, Model 5G 3 July 2015
A. Lang et al. / Molecular Phylogenetics and Evolution xxx (2015) xxx–xxx 3
206 using MrBayes v3.2.1 Â 64 (Huelsenbeck and Ronquist, 2001; Datasets where the number of sampled individuals per species 269 207 Ronquist and Huelsenbeck, 2003). The a priori probabilities sup- is unbalanced can affect the calculations and might lead to overes- 270 208 plied were those specified in the default settings of the program. timation of the number of potential species (Bergsten et al., 2012; 271 209 Best-fit models of nucleotide sequence evolution were selected Zhang et al., 2013; Talavera et al., 2013). By increasing the sam- 272 210 according to the Akaike information criterion in MrModeltest pling, for example with a broader geographical coverage, the max- 273 211 (Posada and Crandall, 1998) executed through PAUP⁄, namely imum intraspecific divergence within the dataset might increase 274 212 HKY + C for the non-coding chloroplast partition and HKY + I for and can result in extra splitting (cf. Talavera et al., 2013). 275 213 coding and nuclear partitions. Sequence and indel data were trea- Besides, over-sampled species with little intraspecific variation 276 214 ted as separate and unlinked partitions, employing the restriction will cause phylogenetic inferences containing large sub-trees with 277 215 site model (‘F81’) for the indel matrix as recommended by numerous extremely short branches and small sub-trees with 278 216 Ronquist et al. (2005). Two runs with four chains were run simul- short branches. In such cases, the under-sampled species cannot 279 217 taneously (11 Â 106 generations), with the temperature of the sin- be identified properly and each individual may be identified as a 280 218 gle heated chain set to 0.5. Chains were sampled every 1000 separate species (Zhang et al., 2013). Therefore, in the present 281 219 generations and the respective trees written to a tree file. study GMYC and PTP analyses were additionally conducted on a 282 220 Fifty-percent majority-rule consensus trees and posterior probabil- reduced alignment containing only specimens with unique 283 221 ities of clades were calculated by combining the two runs and sequences. This reduced alignment, automatically obtained from 284 222 using the trees sampled after the chains converged. Trace plots the raxmlGUI interface, contained 145 specimens, with the stron- 285 223 generated in Tracer v1.5 (Rambaut and Drummond, 2007) were gest reduction in D. scoparium sequences retaining 18 out of the 286 224 used to check for convergence of the runs (plateaus of all runs at initial 65 specimens. The ultrametric and RaxML trees were recon- 287 225 comparable likelihoods) and to infer the ‘burnin’, which was set structed following the above-mentioned methods. 288 226 to 25%.
3. Results 289
227 2.4. Sequence-based species delimitation 3.1. Phylogenetic reconstruction 290
228 Species boundaries were estimated using the generalized mixed The total chloroplast alignment comprised 1914 positions, of 291 229 Yule coalescence (GMYC) and Poisson tree processes (PTP) meth- which 222 were variable, and 132 of the variable characters were 292 230 ods. For GMYC we reconstructed an ultrametric phylogenetic tree parsimony-informative. Of the 1142 positions in the ITS alignment, 293 231 with a strict molecular clock using parameters specified in 124 ambiguous positions were removed from the subsequent cal- 294 232 BEAUti v2 and implemented in BEAST v2.1.1 (Bouckaert et al., culations. The remaining 1019 positions comprised 217 variable 295 233 2014). Branch lengths were estimated under a Yule prior with characters, of which 139 were parsimony-informative. Simple 296 234 HKY nucleotide substitution model for each data partition. We indel coding of the combined dataset yielded 240 additional char- 297 235 included a C rate heterogeneity and no invariant sites for the acters (excluding three corresponding to an inversion in 298 236 non-coding chloroplast partition and no C rate heterogeneity but psbA-trnH), of which 148 were parsimony-informative. 299 237 estimated invariant sites for both the rpoB and ITS partitions. In The single optimal ML tree of the combined markers is shown in 300 238 the absence of appropriate fossil records for calibration, priors on Fig. 1, with bootstrap support (>80% BS) from likelihood analyses 301 239 the rate of molecular evolution were used for chloroplast and and posterior probabilities (PP > 0.95) from Bayesian inference 302 240 nuclear partitions following calibration rates found in the litera- indicated on the branches. The phylogenetic reconstruction 303 241 ture (Villarreal and Renner, 2014). Plastid and nuclear genes harbor resolved 19 clades that correspond to morphological species, 304 242 different substitution rates but few substitution rates are available including the two species with only one sample (Fig. 1). While 305 243 for bryophytes. We applied a plastid substitution rate of 5.0 Â 10À4 the clades of D. acutifolium, D. angustum, D. bonjeanii, D. brevifolium, 306 244 (SD of 2.0–8.0 Â 10À4 subst./site/My) following Villarreal and D. flagellare, D. laevidens, D. majus, D. montanum, D. polysetum, D. 307 245 Renner (2014) for the non-coding chloroplast and rpoB partitions septentrionale, D. spadiceum, D. spurium, D. tauricum, D. undulatum 308 246 and a substitution rate of 1.35 Â 10À3 (SD 0–3.0 Â 10À3 and D. viride were strongly supported (>80% BS, PP > 0.95), D. flex- 309 247 subst./site/My) for ITS as used in Hartmann et al. (2006). The icaule and D. fuscescens received high support only in the Bayesian 310 248 MCMC chains were run with 20 Â 106 generations, saving the reconstruction (59% BS, PP 0.99 and 67% BS, PP 1, respectively). Of 311 249 results every 2000th generation. The convergence of the runs the other Dicranum species, D. groenlandicum was not resolved as 312 250 was examined in Tracer v1.5. The maximum clade credibility tree monophyletic. Seven species were molecularly indistinguishable 313 251 was built from the combined runs after eliminating 25% of the from other closely related species: D. fragilifolium and D. elongatum 314 252 trees for burnin in TreeAnnotator v1.7.2. The GMYC approach samples were mixed within a strongly supported clade (93% BS, PP 315 253 was carried out in R 2.15 (R Development Core Team, 2013) using 1). While D. crassifolium was intermingled with D. scoparium s.s., D. 316 254 the splits (Ezard et al., 2009) and ape (Paradis et al., 2004) pack- leioneuron clustered with North American specimens of D. cf. sco- 317 255 ages. The number of clusters and singletons were estimated by parium in a highly supported clade (92% BS, PP 1; cf. Lang and 318 256 running both single and multiple threshold optimisations and Stech, 2014). The samples of two further species, D. scottianum 319 257 using a multimodel Akaike information criterion with a model cut- and D. canariense, clustered in subclades (67% BS, PP 0.8 and 94% 320 258 off of deltaAICc = 7 (Monaghan et al., 2009; Pons et al., 2006; BS, PP 1, respectively) within a clade with maximum support 321 259 Powell, 2012). For PTP we used the RaxML trees as input data. (100% BS, PP 1). 322 260 The calculations were conducted on the bPTP websever 261 (http://species.h-its.org/ptp/), with 500,000 MCMC generations, 262 thinning set to 100 and burnin at 25% and performing a bayesian 3.2. Sequence-based species delimitation 323 263 search. The probability of each node to represent a species node 264 was calculated in two ways: (1) the Bayesian solution (PTP_sh) The lineage-through-time plots (Fig. 2b and c) indicated an 324 265 considered the frequency of the nodes across the sampling; (2) exponential increase in branching rate near the tip of the tree, 325 266 the maximum likelihood solution (PTP_ML) considered the most which corresponds to the transition between species and popula- 326 267 likely solution among the sampling (P. Kapli, pers. tion processes. The single threshold GMYC (sGMYC) model using 327 268 communication). the ultrametric phylogenetic tree created in BEAST resulted in 328
Please cite this article in press as: Lang, A., et al. Phylogeny and species delimitations in European Dicranum (Dicranaceae, Bryophyta) inferred from nuclear and plastid DNA. Mol. Phylogenet. Evol. (2015), http://dx.doi.org/10.1016/j.ympev.2015.06.019 YMPEV 5234 No. of Pages 9, Model 5G 3 July 2015
4 A. Lang et al. / Molecular Phylogenetics and Evolution xxx (2015) xxx–xxx
Fig. 1. Single optimal maximum likelihood phylogenetic reconstruction of 27 European Dicranum species inferred from the partitioned (non-coding chloroplast, coding chloroplast, and ITS) nucleotide matrix including indels coded by simple indel coding (SIC). Branch thickness and gray shades indicate bootstrap support and posterior probabilities from ML and Bayesian analyses, respectively.
Please cite this article in press as: Lang, A., et al. Phylogeny and species delimitations in European Dicranum (Dicranaceae, Bryophyta) inferred from nuclear and plastid DNA. Mol. Phylogenet. Evol. (2015), http://dx.doi.org/10.1016/j.ympev.2015.06.019 YMPEV 5234 No. of Pages 9, Model 5G 3 July 2015
A. Lang et al. / Molecular Phylogenetics and Evolution xxx (2015) xxx–xxx 5
Fig. 2. (A) Ultrametric tree depicting the relationship of Dicranum based on Bayesian analysis using a Yule model in BEAST and with fit of the general mixed Yules coalescent single threshold (sGMYC) model from plastid and nuclear data. Branch length fitted a strict molecular clock. Estimated entities are indicated in gray. The gray columns indicate the estimated entities from the single and multiple-threshold model (sGMYC and mGMYC) as well as the maximum likelihood and simple heuristic solutions of the Poisson tree processes method (PTP_ML and PTP_sh, respectively). The black column indicates the phylogenetic boundaries of the Dicranum taxa (Phylo). (B and C) Lineages- through-time plot for single (b) and multiple (c) GMYC thresholds. The vertical dashed line represent the timing of the earliest coalescent event.
Please cite this article in press as: Lang, A., et al. Phylogeny and species delimitations in European Dicranum (Dicranaceae, Bryophyta) inferred from nuclear and plastid DNA. Mol. Phylogenet. Evol. (2015), http://dx.doi.org/10.1016/j.ympev.2015.06.019 YMPEV 5234 No. of Pages 9, Model 5G 3 July 2015
6 A. Lang et al. / Molecular Phylogenetics and Evolution xxx (2015) xxx–xxx
Table 1 Type of alignment, species delimitation method and number of estimated entities obtained for Dicranum. LR and LR test of the GMYC single-threshold (sGMYC) and multiple- threshold (mGMYC) analyses are also mentioned. Significant values are indicated with an asterisk. The species delimitation results are compared with the number of supported phylogenetic entities (phylo) obtained from maximum likelihood analyses.
Alignment Method Number of sequences Number of estimated clusters Number of estimated entities LR LR test Extended Phylo 206 21 23 sGMYC 24 34 9.723 0.021⁄ mGMYC 38 58 19.778 0.011⁄ PTP_ML 18 37 PTP_sh 19 45 Reduced Phylo 145 20 22 sGMYC 23 31 7.778 0.051 mGMYC 30 47 14.660 0.023⁄ PTP_ML 23 45 PTP_sh 22 78
329 the identification of 24 Dicranum clusters with high probabilities markers and nuclear ribosomal ITS sequences (Fig. 1), albeit not 376 330 (confidence interval [CI] = 23–26, lnL of null model = 741.079, ML all with significant statistical support. The results support our 377 331 of GMYC model = 748.162, p = 0.00269⁄⁄) and 10 additional lin- recent phylogenetic studies on Dicranum species complexes and 378 332 eages consisting of single sequences, resulting in a total of 34 enti- Arctic Dicranum species (Lang and Stech, 2014; Lang et al., 2014a, 379 333 ties, excluding the outgroup (Table 1; Fig. 2a; Supplementary data 2014b) in that a combination of molecular markers data can clarify 380 334 S1). The multiple threshold method (mGMYC) gave four threshold species circumscriptions in Dicranum, and that the low resolution 381 335 times, resulting in a total of 58 entities that consisted of 38 clusters and clade support within Dicranum in earlier analyses (e.g. La 382 336 (CI = 30–39, lnL of null model = 741.079, ML of GMYC Farge et al., 2002; Stech et al., 2006; Tubanova et al., 2010; 383 337 model = 752.849, p = 0.000634⁄⁄⁄) and 20 singletons, excluding Tubanova and Ignatova, 2011) was a result of too few molecular 384 338 the outgroup (Table 1; Fig. 2a; Supplementary data S2). Although markers analyzed (cf. also Stech and Quandt, 2010). Furthermore, 385 339 the multiple-threshold option was statistically preferred over the the present study shows that, at least for Europe, the molecular 386 340 single-threshold option (deltaAIC = 2.944), neither model was sig- data to a large extent support the morphological species concept, 387 341 nificantly different (Chi-square = 9.375, d.f. = 9, p = 0.40339). despite morphological confusions and subtle diagnostic characters 388 342 Since the mGMYC model considered a higher number of clusters in several species (e.g. Lang et al., 2014b; Tubanova et al., 2010). 389 343 from samples that belonged to single lineages (Table 1; Fig. 2a), In contrast to these results, eight species showed discrepancies 390 344 we took a more conservative approach and discussed only the between their morphological concepts and their molecular 391 345 results of the sGMYC method. circumscription, namely D. crassifolium, D. groenlandicum, 392 346 The trees resulting from PTP gave similar results to GMYC D. elongatum, D. fragilifolium, D. scottianum, D. canariense, 393 347 (Table 1; Fig. 2a). 18 clusters and 19 lineages consisting in single D. leioneuron and D. scoparium. Dicranum groenlandicum was 394 348 sequences were recovered by the best-fit ML search (PTP_ML), resolved as paraphyletic but without significant statistical support 395 349 excluding the outgroup, resulting in a total of 37 putative entities (Fig. 1), although both specimens formed a clade in the ultrametric 396 350 (Table 1; Fig. 2a, Supplementary data S3). However, a total of 45 tree (Fig. 2). This arctic species is morphologically very similar to 397 351 putative entities were recovered by simple heuristic search, con- D. laevidens and, in the absence of sporophytes, both species are 398 352 sisting of 19 clusters and 26 singletons (PTP_sh; Table 1; Fig. 2a; essentially differentiated based on growth form only. Recent 399 353 Supplementary data S4). molecular studies on arctic Dicranum species suggested that 400 354 GMYC results based on the reduced alignment were similar to D. groenlandicum and D. laevidens represent separate entities 401 355 the results based on the extended alignment. The sGMYC model (Lang et al., 2014a,b), which is supported in the present study, in 402 356 indicated the presence of 23 clusters and a total of 31 entities while particular by GMYC (Fig. 2). However, additional sequences of D. 403 357 the multiple models resulted in four threshold times and resulted groenlandicum are necessary to infer its delimitation. Dicranum 404 358 in 30 clusters for a total of 48 entities, excluding the outgroup elongatum and D. fragilifolium are morphologically different and 405 359 (CI = 9–40/38–68, lnL of null model = 259.1778/259.1778, ML of occupy different habitats (Ireland, 2007). Moreover, D. elongatum 406 360 GMYC model = 263.0662/266.5076, p = 0.051/0.023⁄, respectively; is frequently confused with D. groenlandicum, while D. fragilifolium 407 361 Table 1; Supplementary data S5). The tree resulting from PTP with shares morphological similarities with D. tauricum (Hedenäs and 408 362 best-fit ML search recovered 23 clusters and 22 additional single- Bisang, 2004; Ireland, 2007). Despite their clear morphological 409 363 tons, resulting in 45 entities (PTP_ML). 22 clusters and 56 single- distinctions, the present molecular phylogenetic reconstruction 410 364 tons, resulting in a total of 78 entities, were recovered by the indicates that D. elongatum and D. fragilifolium belong to the same 411 365 simple heuristic search (PTP_sh), excluding the outgroup taxon (Fig. 1). The two Macaronesian-Atlantic European species D. 412 366 (Table 1; Supplementary data S6). canariense and D. scottianum were resolved as sister clades within 413 one well-supported clade. Because of their morphological resem- 414 367 4. Discussion blance, D. canariense has been considered as a subspecies or variety 415 of D. scottianum (www.tropicos.org). In the current concept, D. 416 368 4.1. Phylogenetic reconstruction versus morphological species canariense differs from the latter by its strongly denticulate mar- 417 gins as well as a broad and denticulate costa (Hedenäs and 418 369 The present study comprises the largest molecular dataset of Bisang, 2004). The sampling included in this study confirms their 419 370 Dicranum available so far, including all but two Dicranum species close relationship and indicates that both taxa should be distin- 420 371 occurring in Europe following Hedenäs and Bisang (2004), plus D. guished at subspecies level, however, a larger sampling would be 421 372 septentrionale, recently described from Russia and newly identified necessary to confirm these results. Morphological and ecological 422 373 in Scandinavia (Tubanova et al., 2010; Lang et al., 2014b). The characters of D. leioneuron have been discussed several times, as 423 374 majority of the analyzed species (19 out of 27), were molecularly it is frequently confused with either D. bonjeanii or D. scoparium 424 375 recognisable based on the combined analysis of five chloroplast (Ahti and Isoviita, 1962; Corley, 1991). Consequently, D. leioneuron 425
Please cite this article in press as: Lang, A., et al. Phylogeny and species delimitations in European Dicranum (Dicranaceae, Bryophyta) inferred from nuclear and plastid DNA. Mol. Phylogenet. Evol. (2015), http://dx.doi.org/10.1016/j.ympev.2015.06.019 YMPEV 5234 No. of Pages 9, Model 5G 3 July 2015
A. Lang et al. / Molecular Phylogenetics and Evolution xxx (2015) xxx–xxx 7
426 has been sometimes considered as an ecotype of D. scoparium or a such as GMYC or PTP is the objective estimation of phylogenetic 490 427 variety of D. bonjeanii (Ahti and Isoviita, 1962), a hypothesis that is entities and the circumscription of taxa based on branch length 491 428 rejected by the present phylogenetic reconstructions (Fig. 1), dynamics rather than sequence similarities (Monaghan et al., 492 429 which in turn confirm the observations of Corley (1991). Despite 2009; Pons et al., 2006). Although GMYC and PTP performances 493 430 being molecularly separated from D. bonjeanii and D. scoparium have been proven to be stable under a wide range of conditions, 494 431 s.str., the D. leioneuron specimens included in this study clustered the accuracy of species delimitation methods will principally 495 432 in a well-supported lineage together with North American depend on the singularities of the data set and the initial species 496 433 samples, named as D. cf. scoparium in Lang and Stech (2014). concept used (Talavera et al., 2013; Zhang et al., 2013). In this 497 434 Morphology and habitat of these two groups are, however, clearly study, 34 Dicranum species were recovered by the single threshold 498 435 different: the North American specimens have falcate-secund GMYC method, which corresponds generally well with the phylo- 499 436 leaves with serrate margins and a lamellate costa. The D. leioneuron genetic reconstruction. However, disagreements were observed, 500 437 specimens, on the other hand, have all the characteristics of this such as in D. scoparium but also D. viride, D. fragilifolium/D. elonga- 501 438 species, i.e. small and erect-patent leaves; very thin costa and lack tum, D. flexicaule, D. fuscescens and D. polysetum, where overestima- 502 439 of dorsal lamellae. Additionally, flagellate shoots are common in tions in the number of entities occurred compared to the molecular 503 440 this species. Possibly the use of other molecular methods or more and morphological delimitations (Fig. 2a). Each of these species 504 441 variable markers could bring new insights in understanding the counted one additional entity when compared to the phylogenetic 505 442 relationship between D. leioneuron and D. scoparium. Finally, tree, except for D. fragilifolium/D. elongatum, which counted two 506 443 D. crassifolium is a species that has been described recently additional entities and D. scoparium, which was splitted into four 507 444 (Sérgio et al., 1995) and found only in few places in Europe. This entities. GMYC calculations considered both samples of D. groen- 508 445 species resembles D. scoparium but is most similar to landicum as one species and both D. brevifolium and D. acutifolium 509 446 D. transylvanicum (not included here) due to a bi- or even tris- were considered as belonging to the same lineage. 510 447 tratose leaf lamina and denticulate leaf margins. The present Applied to our dataset, GMYC and PTP methods tended to over- 511 448 molecular phylogenetic inferences, however, show that this spe- perform when compared to phylogenetic delineations and the mor- 512 449 cies actually corresponds to D. scoparium. D. scoparium is known phological species concept. The number of clusters was generally 513 450 to be very plastic morphologically and occurs in a broad range of consistent across methods, ranging between 18 and 38 and between 514 451 habitats (Hedenäs and Bisang, 2004; Ireland, 2007; Lang and 20 and 30 for the extended and reduced alignment, respectively 515 452 Stech, 2014; Smith, 2004), including soil or humus, rocks or tree (Table 1). However, the number of estimated entities varied consid- 516 453 bases, in open and shady places where D. crassifolium grows as well erably. The total number of entities obtained from the PTP_ML of the 517 454 (Sérgio et al., 1995). What triggers the deviating leaf lamina mor- extended dataset was similar to the results obtained from sGMYC 518 455 phology of D. crassifolium, and how D. transylvanicum relates to methods and relatively close to the phylogenetic entities, whereas 519 456 D. crassifolium and D. scoparium, remains to be tested. the number of PTP_sh entities was closer to the mGMYC results 520 457 Deviating morphologies are frequently observed in bryophytes, (Table 1). On the contrary, the number of estimated entities based 521 458 especially in species growing under stressful environmental condi- on the reduced dataset was much higher than the number of clades 522 459 tions (Buryová and Shaw, 2005; Hedenäs et al., 2006; Pereira et al., corresponding to morphological species in the phylogenetic recon- 523 460 2013; Såstad, 1998; Såstad et al., 1999; Spitale and Petraglia, structions, for both GMYC and PTP methods (Table 1). Simulations 524 461 2010). Most of the Dicranum species are widespread and found in have shown that an unbalanced sampling are likely to increase the 525 462 a great range of habitats. Hence, local adaptation could partly estimates of haplotypes of the oversampled species (Bergsten 526 463 explain the morphological differences of genetically similar taxa, et al., 2012; Zhang et al., 2013) and each specimen of an undersam- 527 464 such as observed in D. crassifolium, D. scottianum and D. canariense, pled species might be counted as separate entity (Zhang et al., 2013). 528 465 D. fragilifolium and D. elongatum,orD. leioneuron. Statistical analy- In our study, the reduced sampling of D. scoparium did not decrease 529 466 ses can be used to investigate the value of morphological charac- the number of potential species. On the contrary, most of the haplo- 530 467 ters by revealing discontinuities. For example, analyses of the D. types or unique sequences, in particular within D. scoparium, were 531 468 acutifolium complex pointed to a continuous range of morphologi- considered as single lineages (Supplementary data S5 and S6). 532 469 cal variation within the complex. However, the complex was Therefore, the effect of unbalanced sampling in our dataset has prob- 533 470 clearly differentiated from the other species included in the analy- ably less impact on the species delimitation than to the generally 534 471 sis (Lang et al., 2014b). An extended morphological analysis includ- low variability in Dicranum. Indeed, weak signals and high levels 535 472 ing all Dicranum species would be helpful for testing the of uncertainty can explain the low Bayesian support values of the 536 473 morphological species concept against molecular species nodes and the large range of estimated species in PTP estimations 537 474 circumscriptions. (J. Zhang, pers. communication). 538 475 Although the present data does not indicate any hybridisation Overestimations under the GMYC approach have been observed 539 476 events, this genetic process is known to influence the morphology in previous studies (e.g. Miralles and Vences, 2013; Puillandre 540 477 as well (Draper and Hedenäs, 2009; Hedenäs, 2008; Natcheva and et al., 2012; Talavera et al., 2013) and were often related with 541 478 Cronberg, 2004; Sotiaux et al., 2009). Moreover, the consequences errors in the GMYC method or the construction of the ultrametric 542 479 of the special sexual reproduction of several Dicranum species, i.e. tree, rather than to taxonomic knowledge gaps (Talavera et al., 543 480 dwarf males growing on the branch of a female plant (pseu- 2013; Zhang et al., 2013). As our PTP estimates obtained from a 544 481 domonoicy), are largely unknown and would deserve further RaxML tree were not substantially different from the GMYC results 545 482 investigations, in order to explain genetic relationships of closely obtained from an ultrametric tree, we considered that errors in the 546 483 related species. ultrametric tree construction had little effects on the species 547 delimitation. As for now, the GMYC and PTP analysis revealed mul- 548 484 4.2. Species delineation using GMYC and PTP tiple lineages within species in Dicranum that lack morphological 549 and ecological support. Simultaneously, these methods showed 550 485 The definition of boundaries between species clusters is essen- an absence of DNA divergences between D. acutifolium and 551 486 tial, as it will influence the interpretation of the phylogenetic D. brevifolium as well as between D. scottianum and D. canariense, 552 487 reconstructions (Powell, 2012). However, one major drawback of which indicates that these four morpho-species might belong to 553 488 molecular taxonomy is putting an arbitrary threshold for delineat- two single taxa. Regardless of the need of further analyses of the 554 489 ing species. The main advantage of species delimitation methods morphological species concept, the automated species delimitation 555
Please cite this article in press as: Lang, A., et al. Phylogeny and species delimitations in European Dicranum (Dicranaceae, Bryophyta) inferred from nuclear and plastid DNA. Mol. Phylogenet. Evol. (2015), http://dx.doi.org/10.1016/j.ympev.2015.06.019 YMPEV 5234 No. of Pages 9, Model 5G 3 July 2015
8 A. Lang et al. / Molecular Phylogenetics and Evolution xxx (2015) xxx–xxx
556 methods would benefit from refinement as well, in order to over- Ezard, T., Fujisawa, T., Barraclough, T., 2009. Splits: SPecies’ LImits by Threshold 618 Statistics R package. 619 557 come the problem of analyzing datasets with low sequence Farris, J.S., Källersjö, M., Kluge, A.G., Bult, C., 1994. Testing significance of 620 558 variability. congruence. Cladistics 10, 315–319. 621 Fontaneto, D., Herniou, E.A., Boschetti, C., Caprioli, M., Melone, G., Ricci, C., 622 Barraclough, T.G., 2007. Independently evolving species in asexual bdelloid 623 559 5. Conclusions rotifers. PLoS Biol. 5, e87. 624 Frey, W., Stech, M., 2009. Marchantiophyta, bryophyta, anthocerotophyta. In: Frey, 625 W. (Ed.), Syllabus of Plant Families. A. Engler´s Syllabus Der Pflanzenfamilien, 626 560 The identification of bryophyte species is largely based on mor- 13th ed., Part 3 Bryophytes and Seedless Vascular Plants. Gebr. Borntraeger 627 561 phological characters, which are often subtle and difficult to apply, Verlagsbuchhandlung, Stuttgart, p. 419. 628 562 or prone to plasticity induced by environmental conditions. Goffinet, B., Shaw, A.J., 2009. Bryophyte Biology, second ed. Cambridge University 629 630 563 Phylogenetic species delimitations, on the other hand, also rely Press, p. 565. Hartmann, F.A., Wilson, R., Gradstein, S.R., Schneider, H., Heinrichs, J., 2006. Testing 631 564 on a certain degree of subjectivity. Automated methods such as hypotheses on species delimitations and disjunctions in the liverwort Bryopteris 632 565 GMYC and PTP may provide a more objective approach to molecu- (Jungermanniopsida: Lejeuneaceae). Int. J. Plant Sci. 167, 1205–1214. 633 634 566 lar species delineation based on maximum likelihood inferences, Hedenäs, L., 2008. Molecular variation in Drepanocladus aduncus s.l. does not support recognition of more than one species in Europe. J. Bryol. 30, 108–120. 635 567 although inferred boundaries are only putative. Our results showed Hedenäs, L., Bisang, I., 2004. Key to European Dicranum species. Herzogia 17, 179– 636 568 that DNA-based circumscriptions were generally congruent with 197. 637 569 morphological species delimitations. Nevertheless, GMYC and Hedenäs, L., Eldenäs, P., 2007. Cryptic speciation, habitat differentiation, and 638 geography in Hamatocaulis vernicosus (Calliergonaceae, Bryophyta). Plant Syst. 639 570 especially PTP exposed several incongruences between morpho- Evol. 268, 131–145. 640 571 logical concepts and inference from molecular phylogenetic recon- Hedenäs, L., Bisang, I., Lüth, M., Schnyder, N., 2006. Variation in Dicranum majus in 641 572 structions. These incongruences might ensue from evolutionary central, western and northern Europe. J. Bryol. 28, 293–298. 642 Heinrichs, J., Klugmann, F., Hentschel, J., Schneider, H., 2009. DNA taxonomy, cryptic 643 573 processes, but also display the need for further testing on other speciation and diversification of the Neotropical-African liverwort, Marchesinia 644 574 bryophyte data sets to infer the suitability of GMYC and PTP meth- brachiata (Lejeuneaceae, Porellales). Mol. Phylogenet. Evol. 53, 113–121. 645 575 ods for species delimitation in bryophytes. Hernández-León, S., Gernandt, D.S., Pérez de la Rosa, J.A., Jardón-Barbolla, L., 2013. 646 Phylogenetic relationships and species delimitation in Pinus section Trifoliae 647 inferrred from plastid DNA. PLoS ONE 8, e70501. http://dx.doi.org/10.1371/ 648 576 Acknowledgments journal.pone.0070501. 649 Hollingsworth, M.L., Clark, A., Forrest, L.L., Richardson, J., Pennington, R.T., Long, 650 D.G., Cowan, R., Chase, M.W., Gaudeul, M., Hollingsworth, P.M., 2009. Selecting 651 577 We sincerely thank V. Merckx and J.C. Villarreal for their help barcoding loci for plants: evaluation of seven candidate loci with species-level 652 578 with BEAST. We also thank the curator of the herbaria FCO, ICEL, sampling in three divergent groups of land plants. Mol. Ecol. Resour. 9, 439– 653 654 579 MO, NY, O, S, TRH, LISU and UBC for loan of specimens and Dr. D. 457. Hollingsworth, P.M., Graham, S.W., Little, D.P., 2011. Choosing and using a plant 655 580 Tubanova, Dr. C. Sérgio, H. van Melick, and P. Sollman for providing DNA barcode. PLoS ONE 6, e19254. 656 581 additional material. Hudson, R.R., 1990. Gene genealogies and the coalescent process. Oxf. Surv. Evol. 657 Biol. 7, 1–44. 658 Huelsenbeck, J.P., Ronquist, F., 2001. MRBAYES: Bayesian inference of phylogenetic 659 582 Appendix A. Supplementary material trees. Bioinforma. Appl. Note 17, 754–755. 660 Ignatova, E.A., Fedosov, V.F., 2008. Species of Dicranum (Dicranaceae, Bryophyta) 661 with fragile leaves in Russia. Arctoa 17, 63–83. 662 583 DNA numbers, herbarium location (Herb), voucher information, Ireland, R.R., 2007. Dicranum (Family Dicranaceae). In: Zander, R.H. (Ed.), Flora of 663 584 collection date, geographic origin, coordinates, original identifica- North America. Oxford University Press, New York and Oxford, pp. 397–420. 664 585 tion and GenBank accession numbers (rps4-trnT, trnL–trnF, La Farge, C., Shaw, A.J., Vitt, D.H., 2002. The circumscription of the Dicranaceae 665 (Bryopsida) based on the chloroplast regions trnL—trnF and rps4. Syst. Bot. 27, 666 586 trnH-psbA, rps19-rpl2, rpoB, nrITS) of the specimens included in 435–452. 667 587 the study. Lang, A.S., Stech, M., 2014. What’s in a name? Disentangling the Dicranum scoparium 668 588 Supplementary data associated with this article can be found, in species complex (Dicranaceae, Bryophyta). Syst. Bot. 39, 369–379. 669 Lang, A.S., Kruijer, J.D., Stech, M., 2014a. DNA barcoding of arctic bryophytes – an 670 589 the online version, at http://dx.doi.org/10.1016/j.ympev.2015.06. example from the moss genus Dicranum (Dicranaceae, Bryophyta). Polar Biol. 671 590 019. Lang, A.S., Tubanova, D.Y., Stech, M., 2014b. Species delimitations in the Dicranum 672 acutifolium complex (Dicranaceae, Bryophyta). J. Bryol. 673 Leliaert, F., Verbruggen, H., Wysor, B., De Clerck, O., 2009. DNA taxonomy in 674 591 References morphologically plastic taxa: algorithmic species delimitation in the Boodlea 675 complex (Chlorophyta: Cladophorales). Mol. Phylogenet. Evol. 53, 122–133. 676 592 Ahti, T., Isoviita, P., 1962. Dicranum leioneuron Kindb. and the other Dicranum Miller, M.A., Pfeiffer, W., Schwartz, T., 2010. Creating the CIPRES Science Gateway 677 593 mosses inhabiting raised bogs in Finland. Arch. Soc. Zool. Bot. Fenn. ‘‘Vanamo’’ for inference of large phylogenetic trees. Proceedings of the Gateway 678 594 17, 68–79. Computing Environments Workshop (GCE), 14 Nov. 2010, New Orleans, LA, 679 595 Bergsten, J., Bilton, D.T., Fujisawa, T., Elliott, M., Monaghan, M.T., 2012. The effect of pp. 1–8.
Please cite this article in press as: Lang, A., et al. Phylogeny and species delimitations in European Dicranum (Dicranaceae, Bryophyta) inferred from nuclear and plastid DNA. Mol. Phylogenet. Evol. (2015), http://dx.doi.org/10.1016/j.ympev.2015.06.019 YMPEV 5234 No. of Pages 9, Model 5G 3 July 2015
A. Lang et al. / Molecular Phylogenetics and Evolution xxx (2015) xxx–xxx 9
703 Pons, J., Barraclough, T.G., Gomez-Zurita, J., Cardoso, A., Duran, D.P., Hazell, S., Stamatakis, A., 2006. RAxML-VI-HPC: maximum likelihood-based phylogenetic 748 704 Kamoun, S., Sumlin, W.D., Vogler, A.P., 2006. Sequence-based species analyses with thousands of taxa and mixed models. Bioinformatics 22, 2688– 749 705 delimitation for the DNA taxonomy of undescribed insects. Syst. Biol. 55, 2690. 750 706 595–609. Stech, M., 1999. A reclassification of Dicranaceae (Bryopsida) based on non-coding 751 707 Posada, D., Crandall, K.A., 1998. MODELTEST: testing the model of DNA substitution. cpDNA sequence data. J. Hattori Bot. Lab. 86, 137–159. 752 708 Bioinformatics 14, 817–818. Stech, M., Quandt, D., 2010. 20,000 species and five key markers: the status of 753 709 Poulakakis, N., Russello, M., Geist, D., Caccone, A., 2012. Unravelling the peculiarities molecular bryophyte phylogenetics. Phytotaxa 9, 196–228. 754 710 of island life: vicariance, dispersal and the diversification of the extinct and Stech, M., Pfeiffer, T., Frey, W., 2006. Molecular relationships and divergence of 755 711 extant giant Galápagos tortoises. Mol. Ecol. 21, 160–173. palaeoaustral Dicranoloma species (Dicranaceae, Bryopsida). Studies in austral 756 712 Powell, J.R., 2012. Accounting for uncertainty in species delineation during the temperate rain forest bryophytes 31. J. Hattori Bot. Lab. 100, 451–464. 757 713 analysis of environmental DNA sequence data. Methods Ecol. Evol. 3, 1–11. Stech, M., McDaniel, S.F., Hernández-Maqueda, R., Ros, R.M., Werner, O., Muñoz, J., 758 714 Puillandre, N., Modica, M.V., Zhang, Y., Sirovich, L., Boisselier, M.-C., Cruand, C., Quandt, D., 2012. Phylogeny of haplolepideous mosses—challenges and 759 715 Holford, M., Samadi, S., 2012. Large-scale species delimitation method for perspectives. J. Bryol. 34, 173–186. 760 716 hyperdiverse groups. Mol. Ecol. 21, 2671–2691. Stech, M., Veldman, S., Larraín, J., Muñoz, J., Quandt, D., Hassel, K., Kruijer, H., 2013. 761 717 Quandt, D., Stech, M., 2004. Molecular evolution of the trnTUGU–trnFGAA region in Molecular species delimitation in the Racomitrium canescens complex 762 718 bryophytes. Plant Biol. 6, 545–554. (Grimmiaceae) and implications for DNA barcoding of species complexes in 763 719 R Development Core Team, 2013. R: A language and environment for statistical mosses. PLoS ONE 8, e53134. 764 720 computing. Swofford, D.L., 2002. PAUP⁄. Phylogenetic analysis using parsimony (⁄and other 765 721 Rambaut, A., Drummond, A.J., 2007. Tracer. methods). Version 4. 766 722 Ronquist, F., Huelsenbeck, J.P., 2003. MRBAYES 3: Bayesian phylogenetic inference Talavera, G., Dinca˘, V., Vila, R., 2013. Factors affecting species delimitations with the 767 723 under mixed models. Bioinformatics 19, 1572–1574. GMYC model: insights from a butterfly survey. Methods Ecol. Evol. 4, 1101– 768 724 Ronquist, F., Huelsenbeck, J.P., van Dermark, P., 2005. MrBayes3.1 manual, Draft 5/ 1110. 769 725 17/2005.
Please cite this article in press as: Lang, A., et al. Phylogeny and species delimitations in European Dicranum (Dicranaceae, Bryophyta) inferred from nuclear and plastid DNA. Mol. Phylogenet. Evol. (2015), http://dx.doi.org/10.1016/j.ympev.2015.06.019