Molecular Phylogenetics and Evolution 84 (2015) 205–219
Contents lists available at ScienceDirect
Molecular Phylogenetics and Evolution
journal homepage: www.elsevier.com/locate/ympev
Intercontinental long-distance dispersal of Canellaceae from the New to the Old World revealed by a nuclear single copy gene and chloroplast loci
Sebastian Müller a,1, Karsten Salomo a,1, Jackeline Salazar b, Julia Naumann a, M. Alejandra Jaramillo c, ⇑ Christoph Neinhuis a, Taylor S. Feild d,2, Stefan Wanke a, ,2 a Technische Universität Dresden, Institut für Botanik, Zellescher Weg 20b, 01062 Dresden, Germany b Escuela de Biología, Universidad Autónoma de Santo Domingo (UASD), C/Bartolomé Mitre, Santo Domingo, Dominican Republic c Centro de Investigación para el Manejo Ambiental y el Desarrollo, Cali, Colombia d Centre for Tropical Biodiversity and Climate Change, College of Marine and Environmental Science, Townsville 4810, Campus Townsville, Australia article info abstract
Article history: Canellales, a clade consisting of Winteraceae and Canellaceae, represent the smallest order of magnoliid Received 10 July 2014 angiosperms. The clade shows a broad distribution throughout the Southern Hemisphere, across a diverse Revised 16 December 2014 range of dry to wet tropical forests. In contrast to their sister-group, Winteraceae, the phylogenetic rela- Accepted 17 December 2014 tions and biogeography within Canellaceae remain poorly studied. Here we present the phylogenetic Available online 9 January 2015 relationships of all currently recognized genera of Canellales with a special focus on the Old World Canellaceae using a combined dataset consisting of the chloroplast trnK-matK-trnK-psbA and the nuclear Keywords: single copy gene mag1 (Maigo 1). Within Canellaceae we found high statistical support for the mono- Canellales phyly of Warburgia and Cinnamosma. However, we also found relationships that differ from previous Madagascar Maigo 1 studies. Cinnamodendron splitted into two clades, a South American clade and a second clade confined Tropical Gondwana Pattern to the Antilles and adjacent areas. Cinnamodendron from the Antilles, as well as Capsicodendron, South Winteraceae American Cinnamodendron and Pleodendron were not monophyletic. Consequently, Capsicodendron should be included in the South American Cinnamodendron clade and the genus Pleodendron merged with the Cinnamodendron clade from the Antilles. We also found that Warburgia (restricted to mainland east- ern Africa) together with the South American Cinnamodendron and Capsicodendron are sister to the Mal- agasy genus Cinnamosma. In addition to the unexpected geographical relationships, both biogeographic and molecular clock analyses suggest vicariance, extinction, and at least one intercontinental long-dis- tance-dispersal event. Our dating result contrasts previous work on Winteraceae. Diversification of Win- teraceae took place in the Paleocene, predating the Canellaceae diversification by 13 MA in the Eocene. The phylogenetic relationships for Canellaceae supported here offer a solid framework for a future taxo- nomic revision of the Canellaceae. Ó 2014 The Authors. Published by Elsevier Inc. This is an open access article under the CC BY-NC-ND license (http://creativecommons.org/licenses/by-nc-nd/4.0/).
1. Introduction increased with both more published molecular phylogenetic stud- ies and molecular age estimates, a few poorly studied families such Many extant lineages of the earliest angiosperms, including the as the Canellaceae (Canellales) remain poorly known. magnoliids, frequently display a Gondwana distribution pattern. Ordinal-level molecular phylogenies have supported the mono- Moreover, many of these lineages are old enough to be of Gondw- phyly of both Canellales and its two families Canellaceae and Win- ana origin. Thus, their evolution is possibly a result of the isolation teraceae (e.g. Soltis et al., 2000, 2005; Karol et al., 2000; Zanis et al., driven by the successive breakup of Gondwana. Although our 2002; Cai et al., 2006; Massoni et al., 2014), corresponding with understanding of the evolution of the earliest angiosperms has morphology (e.g., Endress et al., 2000). However, Canellaceae lack a well-resolved and statistically supported genus- and species- ⇑ level molecular phylogeny. Also, the fossil record is limited to very Corresponding author. Fax: +49 351 463 37032. few well established lines. Consequently, little is known about E-mail address: [email protected] (S. Wanke). 1 Canellaceae evolution, biogeography, and timing of their Shared first author. 2 Shared last author. diversification. http://dx.doi.org/10.1016/j.ympev.2014.12.010 1055-7903/Ó 2014 The Authors. Published by Elsevier Inc. This is an open access article under the CC BY-NC-ND license (http://creativecommons.org/licenses/by-nc-nd/4.0/). 206 S. Müller et al. / Molecular Phylogenetics and Evolution 84 (2015) 205–219
Canellaceae were established in 1832 by Martius, based on Canellaceae distribution matches a Tropical Gondwana Pattern Browne’s description of Canella (Browne and Ehret, 1756; (TGP) as reviewed in Sanmartín and Ronquist, 2004, while Winter- Martius, 1832). Canellaceae consists of about 23 species, which aceae, including the known fossils, cover most Gondwana areas. are traditionally placed in six genera (Cronquist, 1981; Takhtajan, One might assume that the extant TGP-like distribution of Canell- 1997; Hammel and Zamora, 2005): Canella P. Browne (monotypic), aceae together with a stem group age (SGA) of e.g. 127 MYA Capsicodendron Hoehne (monotypic), Cinnamodendron Endlicher (Naumann et al., 2013; Thomas et al., 2014) resulted from vicari- ( 12 species), Pleodendron Tieghem ( three species), Cinnamosma ance, just as hypothesized for Winteraceae (Marquínez et al., Baillon ( three species) and Warburgia Engler ( three species). 2009). Canella, Capsicodendron, Cinnamodendron and Pleodendron occur To test this hypothesis, a well-resolved and well-supported across the American tropics and subtropics with Canella in sub- molecular phylogeny is needed, which would also allow to address tropical forests as well as Capsicodendron from southern Brazil. taxonomy, biogeography and questions of evolutionary biology. Warburgia occurs in eastern and southern Africa, and Cinnamosma Here we provide a phylogenetic hypothesis for Canellales including is an endemic, but widely distributed genus in Madagascar. The all genera of Canellaceae. The use of a nuclear low copy gene region first phylogenetic data were published by Karol et al. (2000) using in addition to a chloroplast-based dataset especially increased the Canellaceae as an outgroup for a Winteraceae phylogeny based on phylogenetic resolution below the generic level. Based on the ITS and trnL-F. Within Canellaceae, Karol et al. (2000) recon- resulting phylogenetic hypothesis and considering the fossil structed a clade formed by Capsicodendron, Warburgia and Cinna- record, biogeographical patterns of the Canellaceae were recon- mosma and a sister relationship between Pleodendron and structed. The natural relationships recovered here will also provide Cinnamodendron. the basis for a taxonomic revision of the group to be published The first Canellaceae phylogeny with a broad sampling, based elsewhere. on morphological and molecular markers, did not resolve all nodes of the phylogenetic backbone with statistical support (Salazar and 2. Materials and methods Nixon (2008). This study focused on New World species and found that Cinnamodendron was polyphyletic, consisting of one clade in 2.1. Sampling South America and another in the Greater Antilles. The South American Cinnamodendron turned out to be paraphyletic with All known genera of both Canellaceae (Canella, Capsicodendron, regard to Capsicodendron. Salazar (2006) suggested that Capsico- Cinnamodendron, Pleodendron, Cinnamosma and Warburgia) and dendron should be included in the genus Cinnamodendron. She pro- Winteraceae (Drimys, Takhtajania, Pseudowintera, Tasmannia and posed to rename the Antillean clade ‘‘Antillodendron’’. However, Zygogynum (incl. Bubbia) were sampled. The outgroup was chosen ‘‘Antillodendron’’ requires valid publication (nom. inval.), and thus from Piperales (Piper cenocladum), which are sister to Canellales, is used here as operational taxonomic unit only. In addition, differ- from which 52 accessions are included, focusing on the previously ing phylogenetic hypotheses exist on the monophyly of Pleoden- poorly sampled Old World taxa of Canellaceae. Due to the prob- dron. Salazar and Nixon (2008) recovered Pleodendron as lems with species delimitation in Cinnamosma, we included as monophyletic based on morphological and molecular data whereas many accessions as possible to compare morphology-based sys- Zimmer et al. (2012) found evidence for Pleodendron being para- tematics and molecular phylogenies as a base for a new systematic phyletic because Cinnamodendron ekmanii is recovered in a clade treatment. Cinnamosma, traditionally contains three species but with Pleodendron macranthum and Pleodendron costaricense. Bioge- collections show considerable morphological variability in vegeta- ographically, the Old World lineages Warburgia and Cinnamosma tive anatomy and morphology (T.S. Feild, personal observations were found to be nested between the New World clades (Salazar 2011). Some accepted species inhabit a wide range of habitats and Nixon, 2008). However, these phylogenies are not necessarily and are known from e.g., tropical montane cloud forest and spiny contradicting because of the limited sampling (Zimmer et al., desert (T.S. Feild, personal observations 2011). Pleodendron ekmanii 2012) and the low statistical support of nodes (Salazar and (DP 106 and DP 137) accessions were collected and described Nixon, 2008). In contrast to the previous studies, Massoni et al. without flowers from Hispaniola. However J. Salazar considers this (2014) obtained statistically supported relationships for Warburgia species as an Antillean Cinnamodendron = ‘‘Antillodendron’’. DNA as sister to Capsicodendron and this clade again sister to Cinna- samples were obtained from silica dried leaves collected in the mosma. However, this recent study did not include the South field and were completed with herbarium material and fresh mate- American Cinnamodendron, limiting the significance of the phylo- rial taken from the Botanical Garden Dresden. A detailed sampling genetic hypothesis. list including GenBank accessions and voucher specimens is pro- The sister family to Canellaceae, the Winteraceae, were vided in Table 1. described by Lindley (1836) and currently compromise five genera (Vink, 1985): Drimys J.R. Forst. & G. Forst., Pseudowintera Dandy, Takhtajania Baranova & J.-F. Leroy, Tasmannia R. Br. ex DC and 2.2. Marker selection Zygogynum Baill. s.I., which according to Vink (1985, 1988), include Bubbia Tieghem and Exospermum Tieghem and was confirmed by Molecular phylogenetic hypotheses that include Canellales Marquínez et al. (2009). Winteraceae comprise about 80 species show comparatively low resolution and support for Canellaceae and again show a broad distribution in the southern hemisphere, (Salazar and Nixon, 2008) and a relatively low substitution rate including Australasia, Madagascar, Central and South America in Canellales compared to its sister clade Piperales (Suh et al., (Vink, 1988, 2003). Based on divergence time estimates published 1993; Wanke et al., 2007; Naumann et al., 2013; Massoni et al., by Marquínez et al. (2009), the earliest cladogenesis of Wintera- 2014). The chloroplast encoded matK gene was used, at least in ceae predated the fragmentation of Gondwana and thus it was part, in the aforementioned studies. Here we added the full trnK hypothesized that Winteraceae’s extant distribution was explained intron as well as the trnK-psbA spacer, which are among the most by continental vicariance. Current molecular phylogenies of Win- variable regions of the plastid chromosome (e.g. Shaw et al., 2007). teraceae indicate that Takhtajania branched first, followed by Tas- Single and low copy nuclear genes have been proposed to be par- mannia, and a clade consisting of Drimys that is sister to ticularly useful in resolving difficult relationships among angio- Pseudowintera and Zygogynum (Marquínez et al., 2009; Massoni sperms in general (Zhang et al., 2012), among specific flowering et al., 2014; Thomas et al., 2014). plant lineages (e.g. Sang, 2002; Álvarez and Wendel, 2003; S. Müller et al. / Molecular Phylogenetics and Evolution 84 (2015) 205–219 207
Table 1 List of all accessions used in this study including lab number, collector, GenBank accession number and origin.
Sample Lab. number GenBank accession number Collector Origin trnK-matK-psbA mag1 Bubbia queenslandiana DP 10 KP407455 T. Feild Mt. Bubbia 11-16-03 Canella winterana DP 25 KP407458 KP407503 K. Salomo Bot. Garden Dresden Capsicodendron dinisii DP 127 KP407470 KP407515 Morales Brazil Capsicodendron dinisii DP 138 KP407471 KP407516 R. Goldenberg Universidade Federal de Parana, Curitiba, Parana, Brazil Capsicodendron dinisii DP 139 KP407472 KP407517 F. Barros São Paulo: São Paulo, Parque Estadual das Fontes do Ipiranga Cinnamodendron axillare DP 140 KP407473 KP407518 C.N. Fraga, J. Salazar Río de Janeiro, Jardim Botânico de Río de Janeiro (planted), Brazil Cinnamodendron axillare EU669479.1 Cinnamodendron Brazil 3 EU669480.1 Cinnamodendron Brazil 3 EU669478.1 Cinnamodendron corticosum DP 108 KP407461 KP407506 T. Feild Jamaica Cinnamodendron cubense EU669485.1 Cinnamodendron DR EU669487.1 Cinnamodendron ekmanii DP 107 KP407460 KP407505 T. Feild & J. Salazar Dominican Republic Cinnamodendron ekmanii DP 135 KP407462 KP407507 J. Salazar Prov. Samaná; Loma Atravesada, Dominican Republic Cinnamodendron Haiti EU669488.1 Cinnamodendron occhionianum DP 141 KP407474 KP407519 F. Barros Ilha do Cardoso, Sao Paulo, Brazil Cinnamodendron sp. DP 136 KP407463 KP407508 J. Salazar Prov. Azua. Martín García, Dominican Republic Cinnamodendron sp. DP 143 KP407476 KP407521 J. Salazar Cinnamodendron sp. B DP 142 KP407475 KP407520 29-Oct 2004, M.A. Município de Ubatuba, Parque Estadual da de Assis s.n. (SP) Serra do Mar, Núcleo Picinguaba, Brazil Cinnamosma fragrans DP 115 KP407487 KP407531 T. Feild & G. Schatz Madagascar, Orangea, GES-4380 Cinnamosma fragrans DP 129 KP407479 KP407544 J. Rabenantoandro Antsiranana: DIANA Region MO-326950 & H. Razanatsoa 608 Cinnamosma fragrans DP 03 KP407478 KP407523 P.J. Hudson Madagascar, Diego Saurez PJH034 Cinnamosma macrocarpa DP 117 KP407489 KP407533 T. Feild Madagascar, Tampolo, Feild004 Cinnamosma macrocarpa DP 118 KP407490 KP407534 T. Feild Madagascar, Analalava, Feild001 Cinnamosma macrocarpa DP 130 KP407500 KP407545 R. Razakamalala 2182 (MO) Fianarantsoa: Atsimo-Atsinanana Region, MO- 1914366 Cinnamosma madagascariensis DP 101 KP407480 KP407524 P. J. Hudson Madagascar, Marojejy, PJH019 Cinnamosma madagascariensis DP 109 KP407481 KP407525 T. Feild Madagascar, Ambotovy, Feild012 Cinnamosma madagascariensis DP 113 KP407485 KP407529 T. Feild & G. Schatz Madagascar, Angavobe, GES-4384 Cinnamosma madagascariensis DP 116 KP407488 KP407532 T. Feild Madagascar, Ambotovy, Feild007 Cinnamosma madagascariensis DP 120 KP407492 KP407536 T. Feild Madagascar, Tampolo, Feild003 Cinnamosma madagascariensis DP 122 KP407494 KP407538 T. Feild Madagascar, Ambohitantaley, Feild006 Cinnamosma madagascariensis DP 02 KP407477 KP407522 P. J. Hudson Morojejy, 400-700m PJH 19 Cinnamosma perrieri DP 111 KP407483 KP407527 T. Feild & G. Schatz Madagascar, GES-4381 Cinnamosma sp. DP 110 KP407482 KP407526 T. Feild & G. Schatz Madagascar, GES-4382 Cinnamosma sp. DP 112 KP407484 KP407528 T. Feild Madagascar, Ambotovy, Feild011 Cinnamosma sp. DP 119 KP407491 KP407535 T. Feild Madagascar, Analalava, Feild002 Cinnamosma sp. DP 121 KP407493 KP407537 T. Feild Madagascar, Ambohitantaley, Feild005 Cinnamosma sp. DP 123 KP407495 KP407539 T. Feild Ambotovy, Feild008 Cinnamosma sp. DP 124 KP407496 KP407540 T. Feild & G. Schatz Madagascar, Angavobe, GES-4386 Cinnamosma sp. DP 125 KP407497 KP407541 T. Feild & G. Schatz Madagascar, GES-4373 Cinnamosma sp. DP 126 KP407498 KP407542 T. Feild & G. Schatz Madagascar, Petrikey, GES-4362 Cinnamosma sp. DP 128 KP407499 KP407543 R. Razakamalala Fianarantsoa: Atsimo-Atsinanana Region Cinnamosma sp. DP 131 KP407501 KP407546 J. Rabenantoandro 1084 (MO) Antsiranana: SAVA Region; Rabevohitra MO- 326565 Cinnamosma sp. DP 132 KP407502 KP407547 R. Rabevohitra 4510 (MO) Antsiranana: SAVA Region; Rabenantoandro MO-1058203 Cinnamosma sp. DP 114 KP407486 KP407530 T. Feild & G. Schatz Madagascar, GES-4379 Drimys granadensis – NC_008456.1 Drimys winterii BA105 KP407457 J.F. Smith BA105 Piper cenocladum – DQ887677.1 Pleodendron costaricense EU669476.1 Pleodendron ekmanii DP 106 KP407459 KP407504 T. Feild & J. Salazar Dominican Republic Pleodendron ekmanii DP 137 KP407464 KP407509 J. Salazar La Jibara, El Pe;ñonde Mundo Nuevo, R.C. Salcedoa, Dominican Republic Pleodendron macranthum EU669486.1 Pseudowintera axillaris DP 18 KP407456 T. Feild Waipoa forest, New Zealand Takhtajania perrieri DP 05 KP407452 P. J. Hudson Madagascar, Anjanaharibe, PJH 007 Tasmannia lanceolata DP 22 KP407454 S. Wanke 16999 Wanke Tasmannia insipida DP 15 KP407453 T. Feild Barrington Tops, New South Wales, Australia Warburgia stuhlmannii DP 103 KP407466 KP407511 T. Feild Tanzania, Chalinze Warburgia stuhlmannii DP 104 KP407467 KP407512 T. Feild Tanzania, South Dar es Salaam Warburgia ugandensis DP 102 KP407465 KP407510 T. Feild Ethiopia, Harena Forest Warburgia ugandensis DP 105 KP407468 KP407513 T. Feild Tanzania, Lushoto, Shume Rd Warburgia ugandensis DP 133 KP407469 KP407514 MWC: 14998 Collector: Chase (Nr. 9647), ex cult. 208 S. Müller et al. / Molecular Phylogenetics and Evolution 84 (2015) 205–219
Duarte et al., 2010; Naumann et al., 2013, and reviewed in Zimmer with a NucleoSpinÒ Extract II kit (Macherey–Nagel) following the and Wen, 2013), as well as down to the subspecies level (Naumann standard protocol, except that elution was done for 10 min at et al., 2011). 50 °C. Purified products were sent to and sequenced by Macrogen To evaluate nuclear single copy genes in Canellales, we selected Europe using the primers listed in Table 2 or alternatively six gene regions and primers according to Naumann et al. (2013), sequenced on a Beckman Coulter lab sequencer. which were then tested for Canella winterana. Specific primers for Canellales were subsequently designed and tested, based on results of Duarte et al. (2010) and Naumann et al. (2013), in three 2.4. Phylogenetic analyses rounds for all genes (Fig. S1). In the first round, we tested the pri- mer sequences using Canella winterana cDNA. For the second Sequences were edited and aligned manually using PhyDEÒ ver- round, we used genomic DNA as template and increased the test sion 0.9971 (Müller et al., 2005 http://www.phyde.de/). IUPAC sampling by including all five genera of Canellaceae (one accession nucleotide ambiguity code was used for unspecified nucleotide each). Amplification was successful for two single copy genes. In peaks in pherograms if needed (no length mutations recovered the third round, all amplified accessions were purified and within mag1 but few double peaks). Characters with uncertain sequenced. The MAIGO 1 (mag1) region (intron III and exon IV) homology, a poly A mononucleotide repeat (position 619–627), a revealed high quality sequences and was chosen for reconstructing poly T mononucleotide repeat (position 2834–2842) and a highly a molecular phylogeny of Canellales (Fig. S1). This gene is the Ara- variable sequence within ingroup and outgroup (position 2460– bidopsis thaliana homolog At3g47810, TAIR, of the yeast and mam- 2495), were excluded prior to analyses. Insertion and deletions malian VPS29 (vacuolar protein sorting 29) (Shimada et al., 2006), (indels) were scored using the simple indel coding procedure of part of the retromer protein complex, and links cell polarization Simmons and Ochoterena (2000) implemented in SeqState version and organ initiation in plants (Jaillais et al., 2007). 1.4.1 (Müller, 2005). Maximum Parsimony (MP), Maximum Likeli- hood (ML) and Bayesian Inference (BI) analyses were used to reconstruct phylogenetic hypotheses. MP tree searches were per- 2.3. DNA extraction, amplification and sequencing formed using PAUP version 4.0b10 (Swofford, 2003) using a ratchet with 10 random additional cycles of 200 iterations with 25% of the Genomic DNA was extracted using a CTAB isolation method fol- characters upweighted by a factor of 2 in iterations. Statistical sup- lowing Wanke et al. (2007). The RNA of Canella was extracted from port was assessed applying a bootstrap resampling method (1000 snap frozen tissue using CTAB and amplified via Reverse Transcrip- replicates). ML analyses were performed using the rapid bootstrap tase PCR (Promega kit). PCR of cpDNA was performed following algorithm in RAxML version 8 (Stamatakis, 2014) applying the Wanke et al. (2007) using several primer combinations to obtain GTR + G substitution model recovered by jModeltest version 2.1.1 two amplicons for the trnK-matK-trnK-psbA chloroplast region (Posada, 2008) as best fitting model for both datasets. Bootstrap with an overlap of about 450 bp (Table 2). resampling of the data was used to assess statistical support for
A50ll mastermix was used, composed of 30.5 llH20, 5 ll nodes (1000 replicates). Partitioned BI analyses for the cpDNA 10 Taq buffer (PeqLab), 1.2 ll MgCl2,2ll forward and reverse and the nDNA matrix were performed using MrBayes version 3.2 primer each (10 pmol/ll), 8 ll dNTP (each 1.25 mM) and 0.3 ll (Ronquist and Huelsenbeck, 2003). Four Markov chains (one cold Taq DNA polymerase. 1 ll genomic DNA was added ( 30–50 ng/ chain 3 heated chains) were run and calculated simultaneously ll). with 2,000,000 generations. In total, six runs were calculated. Trees For the amplification of the mag1 region a 50 ll mastermix was were saved every 200 generations and the first 2000 generations used, composed of 2875 llH2O, 10 ll reaction buffer (Promega), were discarded as burn-in from each run as evaluated with Tracer 1 ll forward and reverse primer each (50 pmol/ll) (Table 2), 8 ll version 1.5 (Rambaut and Drummond, 2007). Obtained trees were dNTP (each 1.25 mM) and 0.25 ll Go Taq DNA polymerase illustrated using FigTree version 1.3-1 (http://tree.bio.ed.ac.uk/ (Promega). software/figtree/) and TreeGraph version 2.0 (Stöver and Müller, The following PCR conditions were used for cpDNA: pre-dena- 2010). turation at 96 °C for 1.5 min, first primer annealing at 50 °C for 1 min, first elongation at 68 °C for 2 min, followed by 45 amplifica- tion cycles at 95 °C for 30 s, 50 °C for 1 min and 68 °C for 2 min. 2.5. Fossil record of Canellales The following settings for amplifications were used for mag1: pre-denaturation at 95 °C for 2 min followed by 45 amplification Molecular clock analyses were performed following Parham cycles at 95 °C for 45 s, 55 °C for 1 min and 72 °C for 1.5 min. To et al. (2012) using multiple fossil calibration points at different obtain sufficient PCR products of the nuclear gene multiple prod- taxonomic levels to increase the reliability of our analyses. At ucts of the same sample where pooled. PCR products were purified the ordinal level, the Walkeripollis gabonensis fossil was used to
Table 2 List of primers used for amplification and sequencing of the trnK-matK-trnK-psbA and the mag1 region.
Primer Name Region Direction Design Sequence (50-30) trnk-FAngio-1 Chloroplast Forward This study GGGTTGCTAACTCAACGGTAGAG DP-matK-2740-R Chloroplast Reverse This study CATCTGGAAATCTTGSTTC DP-matK-1950-F Chloroplast Forward This study GACCGTATCGYACTATG psbA-R Chloroplast Reverse Steele and Vilgalys (1994) CGCGTCTCTCTAAAATTGCAGTCAT NymatK-480-F Chloroplast Forward Borsch et al. (2003) CATCTGGAAATCTTGSTTC trnk-R-bryo-2 Chloroplast Reverse Wicke and Quandt (2009) TCGAACCCGGAACTHGTCGG DP-trnk-4530-F Chloroplast Forward This study GCAAYGAAAAATGCAAGCACGG Po-2580-F Chloroplast Forward This study CATAGAGAAAGCCGTGTGC MLC-matK-3240-R Chloroplast Reverse This study TATGTTTACGAGMCAAAGTTCTA CaMa-1337 Chloroplast Reverse This study CTCCCAAGCACACAGATTTTCT Ca-2907-910F Nuclear Forward This study TRTGYCATGGTCATCAGGTT Pi-2907-1560R Nuclear Reverse Naumann et al. (2013) CATCAATTCAAGGCMTACAAGCA S. Müller et al. / Molecular Phylogenetics and Evolution 84 (2015) 205–219 209 calibrate the age of the split of Canellaceae and Winteraceae However, the fossil record of Canellaceae is limited to Tertiary (node 0, Figs. 1 and 2). Walkeripollis gabonensis was reconstructed deposits. Hollick and Berry (1924) reported Canella leaves from as the sister to Winteraceae (Doyle and Endress, 2010), dated to Pliocene beds in Bahia, Brazil, and Graham and Jarzen (1969) the Barremian/Aptian (from Gabon, Doyle et al., 1990a,b; Doyle described Oligocene pollen of Pleodendron from Puerto Rico. and Endress, 2010), and was used to calibrate the respective split Recently, the Eocene Wilsonoxylon edenense from Wyoming, with a minimum age of 125 MYA (Fig. 2). Besides Walkeripollis USA, was discovered (Boonchai and Manchester, 2012). The Plio- gabonensis, more fossils representing the Canellales fossil line cene Canella fossil (Hollick and Berry, 1924) could not be consid- are known. Qatanipollis was described by Schrank (2013) and ered as calibration point because only a single accession of includes three species: Qatanipollis valentini, Qatanipollis sp. A Canella is included in our dataset whereas the full genetic and (formerly known as Walkeripollis sp. A, Walker et al., 1983, geographic diversity of Canella would be needed to confidently Doyle et al., 1990a,b) and Qatanipollis sp. B. All these latter fossils place the fossil at the most ancestral node of the resulting Canella were found in Israel in formations dated to the Albian (Walker clade. Wilsonoxylon edenense is structurally similar to Capsicoden- et al., 1983; Schrank, 2013). Additionally Barreda and dron, Warburgia and Cinnamodendron, with the closest similarity Archangelsky (2006) reported Walkeripollis pollen from the late to Warburgia, but could not unequivocally be assigned to one of Albian – Cenomanian from Patagonia. Walkeripollis gabonensis is the three genera (Boonchai and Manchester, 2012). Our phyloge- the oldest fossil and is used here as calibration point for the most netic hypothesis recovers two independent Cinnamodendron recent common ancestor (MRCA) of Canellales (mean age 127 clades (a South American and an Antillean clade, Fig. 1). As it is MYA, min age 125 MYA). Since sampling focuses on Canellaceae, unclear which Cinnamodendron material was used by Boonchai additional calibration points had to be found within Canellaceae. and Manchester (2012) (Nareerat Boonchai, personal communica-
Fig. 1. Bayesian phylogeny of Canellales based on 53 accessions, including one outgroup accession (Piper), with focus on the Old World genera of Canellaceae. The topology is based on the concatenated mag1-trnK-matK-trnK-psbA plus indel matrix. Family, genus and subclade levels are indicated by brackets. ‘‘Antillodendron’’ (⁄) is used as operational taxonomic unit for Cinnamodendron (Cnd) from the Antilles and Central America. Maximum Parsimony and Maximum Llikelyhood analyses did not differ from Bayesian Inferences (for exceptions see Section 3.2). (a) Phylogeny: important nodes (N0–N9) and posterior probabilities (PP) are indicated above branches, maximum likelihood boostrap values (ML-BS) and maximum parsimony bootstrap values (MP-BS) are given below the branch. MP-BS values are only shown for the backbone of the phylogeny at the genus level. Support values of <70 are not shown. (b) Phylogram shown without the outgroup (Piper cenocladum). Cinnamosma and Warburgia are monophyletic, Cinnamodendron is polyphyletic and divided into two clades, a South American and a Central American/Antillean clade. Capsicodendron is nested within the South American Cinnamodendron clade, Pleodendron is nested within the Central American clade. Little variability within Cinnamosma is observed. Abbreviations: Cap: Capsicodendron;C:Cinnamosma Cnd: Cinnamodendron;P:Pleodendron;W:Warburgia. 210 S. Müller et al. / Molecular Phylogenetics and Evolution 84 (2015) 205–219
Fig. 2. Divergence time estimates for Canellales as reconstructed by BEAST (scenario one). The outgroup species Piper cenocladum is excluded from the Figure. Mean ages and 95% highest posterior density (HPD) of important nodes are indicated, pentagons indicate constrained nodes based on fossil record. Information about the respective fossil age and its area of distribution is given. Results derived from the second and third fossil calibration set are provided in Table 5. The geological timescale is given in million years ago (MYA) and associated relative geological units are provided. Three important paleotectonic events are indicated below the geological timescale showing that the Gondwana breakup predated the diversification of both Canellaceae and Winteraceae with possibly the exception of Takthajania. Abbreviations: Cap: Capsicodendron;C: Cinnamosma; Cnd: Cinnamodendron; Ple: Pleodendron;W:Warburgia.
tion 2013), the fossil calibration point had to include all three 2.6. Molecular dating analyses genera and both Cinnamodendron clades, resulting in the calibra- tion of the MRCA of Canellaceae (mean age of 49 MYA, minimum Molecular dating analyses were performed using BEAST version age 48 MYA, node 2, Figs. 1 and 2). Since both Antillean taxa, 1.7.5 (Drummond et al., 2012) using the high performance Beagle ‘‘Antillodendron’’ and Pleodendron, form a well-supported clade Library version 1.0 (http://code.google.com/p/beagle-lib/). XML (Salazar, 2006) and Pleodendron is not monophyletic (Zimmer command files were generated in BEAUTi version 1.7.5 et al., 2012), we applied the Pleodendron fossil as a crown group (Drummond et al., 2012). Calibrated nodes were constrained to calibration point of the ‘‘Antillodendron’’ and Pleodendron clade be monophyletic within the individual BEAST runs. The relaxed (node 5, Figs. 1 and 2) with a mean age of 24 MYA (minimum clock model was set to ‘‘uncorrelated lognormal’’ as recommended age 23 MYA). We did not include any fossil calibration point by Drummond et al. (2006). The Markov chain was run for within Winteraceae, as we did not focus our sampling on Winter- 250,000,000 generations, sampling trees every 2500 generations. aceae and the phylogenetic affinity of the Pseudowinterapollis fos- The first 5,000,000 generations were discarded as burn-in as eval- sil is still under debate. No phylogenetic survey of pollen uated with Tracer version 1.5 (Rambaud and Drummond, 2007). characters in Winteraceae is yet available to decide whether this Three independent runs were performed using the three sets of fossil is sister to or nested within the Winteraceae crown group calibration points. A strict consensus tree with mean ages as (Doyle, 2000). branch lengths was constructed with Tree Annotator version In general, the fossil record provides only a glimpse into the 1.7.5 (Drummond et al., 2012). Obtained trees were illustrated evolution of a clade, and as outlined above, fossil placement within using FigTree version 1.3-1 (http://tree.bio.ed.ac.uk/software/fig- Canellales remains difficult. Thus, we performed three indepen- tree/) and Tree Graph version 2.0 (Stöver and Müller, 2010). dent molecular age estimation analyses with different fossil sets to examine the range of possible molecular age estimates. The first 2.7. Ancestral area reconstruction test included all three fossils, for the second we used Walkeripollis gabonensis and Wilsonoxylon edenense, and for the third, we used Reconstruction of historical biogeography was performed using Walkeripollis gabonensis only (Table 3). two different approaches: dispersal, local extinction and cladogene- S. Müller et al. / Molecular Phylogenetics and Evolution 84 (2015) 205–219 211
Table 3 Callibration scenarios calculated using BEAST. Calibrated nodes with the minimum age of the node, in MYA, and their respectively fossils are given. Numbering of nodes corresponds to the labelling in Figs. 1–3. We reduced the number of fossils with every scenario by one. Scenario one includes all three fossils (Walkeripollis, Wilsonoxylon and Pleodendron), for scenario two, Pleodendron is excluded, and in scenario three both Wilsonoxylon and Pleodendron are excluded.
Scenario Used fossils Calibrated nodes Minimum age (MYA) 1 Walkeripollis gabonensis, Split Canellaceae – Winteraceae (node 0) 125 Wilsonoxylon edenense Split Greater Antillean – South 48 American/African/Malagasy species (node 2) 23 Pollen of Pleodendron Crown Group Greater Antillean Species (node 5) 2 Walkeripollis gabonensis Split Canellales (node 0) 125 Wilsonoxylon edenense Split Greater Antillean – South 23 American/African/Malagasy species (node 3) 3 Walkeripollis gabonensis Split Canellales (node 0) 125
sis model (DEC) and Bayesian Binary Method (BBM). Both 2.4) due to uncertain homology, leaving a dataset of 3413 approaches were performed using RASP version 3.0 (Reconstruct characters. Ancestral State in Phylogenies) (Yu et al. 2014 http://mnh.scu.edu. The trnK-matK-trnK-psbA region accounted for 2728 bp, on cn/soft/blog/RASP/). average ranging from 1987 to 2818 bp for individual accessions. Five areas of distribution were specified: Madagascar (Cinna- All regions of uncertain homology, listed in Section 2.4, are located mosma and Takhtajania), Africa (Warburgia), South America (South within the 50-trnK intron. MP analysis based on cpDNA resulted in American Cinnamodendron and Drimys), Central America including seven most parsimonious trees (1040 steps) based on 351 parsi- the Antilles (Pleodendron, Canella and Antillean Cinnamoden- mony informative characters (PICs) with low homoplasy dron = ‘‘Antillodendron’’) and Australasia (Pseudowintera, Bubbia, (CI 0.878, RI 0.939, RC 0.824, HI 0.124). The mag1 matrix comprised and Tasmannia). The outgroup was coded as distributed in all five of 398 characters only with an average of 362 bp, ranging from 268 areas as is the case for Piperales and is a conservative treatment that to 397 bp for individual accessions. The GC content of mag1is does not introduce bias. Given a possible wide distribution of the lower (23.08%) than for the chloroplast dataset (34.86%). MP anal- MRCA of Canellales, we allowed five areas of distribution to reduce ysis based on mag1 revealed 218 most parsimonious trees (113 bias. steps) based on 43 PICs with very low homoplasy (CI 0.912, The DEC model (Ree and Smith, 2008) was performed using RI 0.959, RC 0.874, HI 0.088). For the combined dataset, 46 most three different scenarios for dispersal constraints. Additional con- parsimonious trees (1161 steps) are recovered by MP analysis straints were incorporated to account for the fossil record distribu- (data not shown but see Section 3.2) based on 394 parsimony tion. In scenario one (DEC1), dispersal rates were treated as equally informative characters (PICs) with low homoplasy (CI 0.875, RI likely between all areas. In scenario two (DEC2), dispersal probabil- 0.938, RC 0.820, HI 0.125). ities were modified to account for possible paleogeographic changes (Gondwana breakup) that might have influenced Canell- 3.2. Phylogenetic relationships ales distribution and evolution. We followed Buerki et al. (2011), using four time windows: before 120 MYA, 120–60 MYA, 60– The individual cpDNA and mag1 dataset revealed the same 30 MYA and 30 MYA – present day, with three possible dispersal backbone topology as the combined dataset using MP, ML and BI rates. A dispersal rate of 1.0 was used for land connections analyses (Table 4). These results indicated no significant conflicting between areas of distribution, 0.5 for dispersal over sea and 0.01 phylogenetic signals. Differences between analyses and matrices where a complex scenario would be required to explain dispersal are only recovered for unsupported nodes (e.g. relationships of (Buerki et al., 2011). In scenario three (DEC3) we additionally con- individual accessions within Cinnamosma). Important nodes strained the ancestral distribution at nodes N0 and N2 based on (Fig. 1 and Table 4) within Canellaceae mostly received moderate fossil age and occurence. Walkeripollis and Qatanipollis were found to strong statistical support with the individual marker sets, as in Israel, Gabon and later in Patagonia (see Section 2.5) and a land well as in the combined dataset (0.92–1.00 PP) (Fig. 1 and Table 4). connection between Africa and South America remained until 110– Phylogenetic results are shown for the BI phylogenetic tree with 95 MYA (Sanmartín and Ronquist, 2004). Consequently, node N0 combined dataset and coded length mutations as well as statistical was constraint as Africa and South America. Node N2 was con- support for key nodes from ML and MP analyses (Fig. 1). strained as Central America based on the Wilsonoxylon fossil distri- The overall phylogenetic hypothesis, based on the combined bution from Wyoming because Central- and North America were dataset (Fig. 1), supports the monophyly of Winteraceae and covered with broad leaved evergreen forest resulting in a continu- Canellaceae. Within Winteraceae, Takhtajania branches first, fol- ous floristic region during the respective time (Wing, 1987; lowed by Tasmannia. A sister group relationship is recovered as Graham, 2010). ((Zygogynum s.l. + Pseudowintera) Drimys)(Fig. 1). All nodes are sta- BBM was performed using 500,000 generations with 10 chains, tistically supported (>0.95 PP). Within Canellaceae, Canella sampling every 100 generations. The first 500 trees were elimi- branches first, followed by a clade consisting of ‘‘Antillodendron’’ nated (burn-in). State frequencies were set to fixed, Jukes-Cantor and Pleodendron (1.00 PP, 100 ML BS). The Warburgia clade is sister (JC) and among-site rate variation was set to equal (BBM1). A sec- to a clade consisting of Capsicodendron and the South American ond BBM analysis with a reduced dataset containing only seven Cinnamodendron (0.95 PP, 89 ML BS). Cinnamosma is sister to a Cinnamosma accessions was run to test sampling bias (BBM2). clade, consisting of Warburgia, Capsicodendron and the South American Cinnamodendron (1.00 PP, 100 ML BS, Fig. 1). The mono- 3. Results phyly of the clades Warburgia, Cinnamosma and the South Ameri- can Cinnamodendron + Capsicodendron is statistically supported by 3.1. Phylogenetic dataset all analyses (Fig. 1 and Table 4), but Cinnamodendron from the Antilles, as well as Capsicodendron, South American Cinnamoden- The concatenated matrix comprised, 3467 characters. Three dron and Pleodendron are not monophyletic (Figs. 1 and S1, Table 4). regions were excluded prior to phylogenetic analyses (see Section Within Cinnamosma, six clades (A–F) and one branch with a single 212 S. Müller et al. / Molecular Phylogenetics and Evolution 84 (2015) 205–219
Table 4 Comparison of individual and combined datasets within Canellaceae including alignment length, GC content, number of parsimony informative characters (PICs) and support for important nodes (N3–N7, see Fig. 1). Individual mag1 and trnK-matK-trnK-psbA datasets reveal an identical well supported backbone phylogeny for Canellaceae. Statistical support is ordered as followed: MP-BS: maximum parsimony boostrap value; ML-BS: maximum likelihood bootstrap value; PP: posterior probabilities.
Dataset Alignment length (bp) GC content (%) PICs Statistical support (MP-BS; ML-BS; PP) Node 3 Node 4 Node 5 Node 6 Node 7 mag1 398 23.1 43 100 94 100 68 95 – 87 100 83 96 1.00 1.00 1.00 0.93 1 trnK-matK-trnK-psbA 3015 35.1 351 100 100 100 70 100 91 100 100 71 100 0.95 1.00 1.00 0.84 1.00 Combined 3413 33.7 394 5 100 100 92 100 92 100 100 89 100 0.92 1.00 1.00 0.95 1.00
accession are recovered. With the exception of clade B, each of the in the individual analyses, the age estimates of the majority of the individual clades is strongly supported (0.99 – 1 PP). However, the nodes increases (Table 5). One of the most extreme age differences relationships among the main Cinnamosma lineages remain unsup- is recovered for the split of ((Pleodendron + Antillean Cinnamoden- ported. Within Cinnamosma the phylogenetic hypotheses derived dron)(Warburgia + South American Cinnamodendron + Cinna- from ML and MP are congruent with the reconstruction derived mosma)), with a mean age of 43 MYA (scenario one), 39 MYA from BI, although ML and MP analyses receive low individual sup- (scenario two) and 34 MYA (scenario three) (Table 5). In all three port (68–85 ML BS). analyses, the Winteraceae crown group is older than the Canella- A further BI analysis was performed using matK sequences from ceae crown group (Table 5). A more recent diversification is recov- eight additional accessions available in GenBank (Table 1). ered for the crown group age of Warburgia (8–6 MYA, mean ages), Pleodendron macranthum, Pleodendron costaricense, Cinnamoden- Capsicodendron/South American Cinnamodendron (6–4 MYA mean dron cubense, Cinnamodendron sp. Haiti and Cinnamodendron sp. ages) and Cinnamosma (14–10 MYA mean ages), depending on DR are nested within the Antillean Cinnamodendron and Pleoden- the scenario. Although estimated ages differ slightly, differences dron clade. Cinnamodendron sp 1 Brazil, Cinnamodendron sp 3 Brazil in age estimates are recovered congruently using individual cali- and Cinnamodendron axillare are nested within the South American bration scenarios. Combining these lines of evidence, our Cinnamodendron clade. As both clades receive strong support (1.00 biogeographic analyses are based on the results of scenario one PP, Fig. S2), the non-monophyly of ‘‘Antillodendron’’, Pleodendron, (Fig. 2). Capsicodendron and South American Cinnamodendron is further substantiated. 3.4. Biogeographic analyses 3.3. Age estimates In general, BBM tends to suggest single distribution areas for Individual mean age estimates for main nodes including 95% ancestral nodes more often than DEC and DEC favors a more wide- highest posterior density (HPD) are provided in Table 5 for the dif- spread ancestral distribution consisting of all distribution areas of ferent fossil calibration sets. If more calibration points are included descendant branches (Figs. 3 and 4, Table 6).
Table 5 Mean ages (including 95% highest posterior density (HPD)) for important nodes in million years ago (MYA) as calculated by the BEAST analyses. In all scenarios, diversification started after the Gondwana breakup. Scenario one includes all three fossils (Walkeripollis, Wilsonoxylon and Pleodendron), for scenario two, Pleodendron is excluded, and in scenario three both Wilsonoxylon and Pleodendron are excluded. Numbering of nodes corresponds to the labelling in Figs. 1–3. Including more fossils in the analysis increases the age of the majority of nodes. The simplified tree shows the main topology with the respective node numbers.’’Antillodendron’’ (⁄) is used as operational taxonomic unit for Cinnamodendron (Cnd) from the Antilles.
Node Scenario 1 Scenario 2 Scenario 3 Mean (95% HPD) N0 128 (134; 125) 128 (133; 125) 127 (132; 125) N1 62 (91; 35) 53 (73; 33) 49 (71; 28) N2 49 (52; 48) 49 (51; 48) 41 (59; 23) N3 43 (50; 35) 39 (48; 30) 34 (50; 19) N4 25 (35; 15) 21 (28; 13) 18 (27; 10) N6 19 (29; 10) 17 (24; 10) 15 (22; 8) N7 14 (20; 8) 10 (14; 7) 10 (14; 6) N8 8 (15; 3) 6 (10; 3) 6 (10; 2) N9 6 (10; 2) 4 (7; 2) 4 (6; 2) S. Müller et al. / Molecular Phylogenetics and Evolution 84 (2015) 205–219 213
Fig. 3. Biogeographic reconstruction for Canellales using the Bayesian Binary Method (BBM) based on the phylogeny derived from the BEAST analysis (scenario one). The outgroup species Piper cenocladum is excluded. Pentagons indicate fossil placement in the BEAST and DEC analyses. The pentagons follow the color code of the analysis. Pie charts indicate the most likely distribution areas of the most recent common ancestor (MRCA). The current distribution area of accessions is provided at the tips of the branches. The areas are coded as follows: A, Madagascar; B, Africa; C, South America; D, Antilles and Central America (and includes North America for Wilsonoxylon); E, Australasia. The world map visualizes distribution areas, color code of BBM and DEC (Fig. 4) analyses are identical. Abbreviations: Bub: Bubbia; Cap: Capsicodendron; Cin: Cinnamosma; Cnd: Cinnamodendron; Drim: Drimys; Pip: Piper; Ple: Pleodendron; Pse: Pseudowintera; Tak: Takhtajania; Tas: Tasmannia;W:Warburgia.
BBM analyses with reduced and full sampling reveal identical and DEC2 analyses results are provided in Table 6 and differences results (Table 6). The origin of the Winteraceae MRCA remains to DEC3 are discussed later. The origin of the Winteraceae MRCA unclear in BBM, as Australasia, Madagascar and Central America was a broad distribution area with different combinations of areas are nearly equally likely (N1, Fig. 3), the MRCA of Tasmannia, (N1, Table 6). Highest likelihood is reconstructed for the combina- Pseudowintera, Bubbia and Drimys as well as the MRCA of Pseudo- tion of Madagascar, South America, Central America and Austral- wintera, Bubbia and Drimys were both distributed in Australasia asia. Madagascar is always part of the ancestral area followed by (Fig. 3). Canellaceae’s origin as well as the subsequent nodes (N2, Australasia and to a smaller fraction South America. The ancestors N3, N5) are reconstructed as Central America and adjacent areas. of the Tasmannia, Pseudowintera, Bubbia, Drimys clade as well as the Madagascar and Central America are the two most likely distribu- Pseudowintera, Bubbia, Drimys clade was most likely distributed in tion areas for the ancestor of the Warburgia, Cinnamosma and Caps- Central America, South America and Australasia (Fig. 4 and Table 6). icodendron/South American Cinnamodendron clade (N4, Fig. 3). The The ancestor of all extant Canellaceae excluding Canella (N3) was ancestor of the Capsicodendron/South American Cinnamodendron most likely distributed in Madagascar, Africa, and Central- and clade occurred in Africa or South America (N6, Fig. 3). South America (DEC3, N3, Fig. 4) with Central America recovered Modifying dispersal rates at different time slices and con- in most reconstructed combinations of ancestral areas followed straints of distribution areas at ancestral nodes in DEC returned by Madagascar. The ancestor of Warburgia, Capsicodendron and discrepant results for only two nodes (N0 and N2) (Table 6). Bio- South American Cinnamodendron (N6) is exclusively reconstructed geographic results are shown for the DEC3 analysis (Fig. 4), DEC1 as an African and South American origin (N6, Fig. 4, Table 6). 214 S. Müller et al. / Molecular Phylogenetics and Evolution 84 (2015) 205–219
Fig. 4. Biogeographic reconstruction for Canellales using Dispersal Extinction Cladogenesis (DEC) scenario 3 using modified dispersal rates and fossil constrains on ancestral nodes (see 2.7 Ancestral area reconstruction, Table 6), based on the phylogeny derived from the BEAST analysis (scenario one). The outgroup species Piper cenocladum is excluded. Pentagons indicate fossil placement in the BEAST and DEC analyses. The pentagons follow the color code of the analysis. The current distribution area of accessions is provided at the tips of the branches. The areas are coded as follows: A, Madagascar; B, Africa; C, South America; D, Antilles and Central America (and includes North America for Wilsonoxylon); E, Australasia. The world map visualizes distribution areas, color code of BBM and DEC analyses is identical. Abbreviations: Bub: Bubbia; Cap: Capsicodendron; Cin: Cinnamosma; Cnd: Cinnamodendron; Drim: Drimys; Pip: Piper; Ple: Pleodendron; Pse: Pseudowintera; Tak: Takhtajania; Tas: Tasmannia;W:Warburgia.
4. Discussion Wendel, 2003, reviewed in Zimmer and Wen, 2013). The tremen- dous numbers of deposited ITS sequences in public databases 4.1. Performance of nuclear and chloroplast loci and readily available primers for virtually any plant group seems to be most efficient for reconstructing phylogenetic hypotheses. Although the number of molecular phylogenetic studies using However, utilizing nuclear single copy genes as described here non-ribosomal nuclear DNA sequences for phylogenetics is stea- requires only a little more effort. Only a tenth of sequenced charac- dily growing, they have been used in molecular dating approaches ters of mag1 compared to the chloroplast trnK-matK-psbA region only in a limited number of studies (e.g. LFY and ACO in Guo et al., (398 bp and 3015 bp respectively) reveals the same phylogenetic 2012, GPI in Marcussen et al., 2012, at103 in Désamoré et al., 2012, backbone and provides strong statistical support for all ‘‘important 14 genes in Naumann et al., 2013). The need for orthologous low or nodes’’ within Canellaceae (Table 4). In some cases, the statistical single copy nuclear markers has long been recognized (e.g. Small support from the nuclear marker is even higher than the support et al., 1998; Sang, 2002; Hughes and Eastwood, 2006). Neverthe- received by the chloroplast marker. The single copy gene also pro- less, besides various chloroplast markers, ITS still dominates vides resolution on both high and low taxonomic levels, leaving molecular clock approaches, although the community has been only the relationships in Cinnamosma unresolved. A comparatively well aware of its drawbacks for a long time (e.g. Álvarez and recent and rapid diversification of Cinnamosma might be the reason S. Müller et al. / Molecular Phylogenetics and Evolution 84 (2015) 205–219 215
Table 6 Comparison of results derived from biogegraphic analyses. Bayesian Binary Method (BBM) and Dispersal Extinction Cladogenesis (DEC) results are provided with constrained dispersal rates (DEC1), with unconstrained dispersal rates (DEC2), with constrained dispersal rates and placed fossils (DEC3). Constrained dispersal rates are calculated based on
Buerki et al. (2011). BBM1 using the full data set and BBM2 with reduced dataset. Relative probabilities (RP) of the ancestral areas are given as a fraction of the global likelihood of a split and only ancestral areas with RP > 10% are shown. Most likely events for each node (dispersal/vicariance/extinction) are provided including the likelihood in brackets. However, it should be noted that in DEC, vicariance explanations are reconstructed for nodes that are younger than the split of the areas involved, which might potentially be a reconstruction artifact. (⁄) indicates nodes with fossil placement. The areas are coded as follows: A, Madagascar; B, Africa; C, South America; D, Antilles and Central America; E, Australasia.
DEC1 DEC2 DEC3 BBM1 BBM2 Area RP Event Area RP Event Area RP Event Area RP Event Area RP Event (%) (likelihood) (%) (likelihood) (%) (likelihood) (%) (likelihood) (%) (likelihood) N0 A 30.7 6/0/0 (0.076) ABCDE 32.0 3/0/0 (0.145) BC⁄ 100 5/0/0 D 66.2 2/1/0 (0.261) D 67.2 2/1/0 D 18.2 ACDE 17.7(0.3644) A 14.4 A 17.9 (0.2918) CD 15.7 ABCD 17.0 N1 ACDE 39.5 0/1/0 (0.179) ACDE 45.2 0/1/0 (0.227) CDE 37.4 0/1/0 A 40.5 2/1/0 (0.354) A 45.00 2/1/0 (0.1504) (0.3932) AE 24.0 AE 22.7 ACE 19.4 D 25.0 D 24.4 ADE 18.6 ADE 16.1 ABCDE 15.1 E 22.7 E 19.4 ACE 17.9 ACE 16.0 AE 14.4 ADE 13.8 N2 ABCD 63.0 1/0/0 (0.481) ABCD 100 1/0/0 (1.000) C⁄ 100 3/0/0 D 97.8 0/0/0 (0.934) D 96.6 0/0/0 ACD 14.6 (0.2752) (0.9126) ABD 12.9 N3 ABCD 76.3 0/1/0 (0.763) ABCD 100 0/1/0 (1.000) ABCD 27.5 0/1/0 D 95.6 2/1/0 (0.410) D 94.5 2/1/0 ACD 12.7 ABD 12.0 (0.2421) (0.4435) ABD 11.0 ACD 11.9 D 11.7 AD 11.5 N4 ABC 100 0/1/0 (1.000) ABC 100 0/1/0 (1.000) ABC 88.0 0/1/0 A 43.0 2/1/0 (0.194) A 47.1 2/1/0 AB 12.0(0.8798) D 35.4 D 31.4 (0.2124) N6 BC 100 0/1/0 (1.000) BC 100 0/1/0 (1.000) BC 100 0/1/0 (1.000) B 45.4 2/1/0 (0.439) B 45.3 2/1/0 C 38.8 C 38.8 (0.4384)
for a limited sequence divergence resulting in short branches, a species from the Antilles and adjacent areas should be included finding that is most likely independent from the studied genome in Pleodendron. The clade including South American species of Cin- (plastid, mitochondrial, nuclear). ‘‘Short branch clades’’ like Cinna- namodendron is well supported, but these species also form a para- mosma are found in many plant lineages and are the most chal- phyletic group, since the clade includes Capsicodendron as well. lenging to resolve (Richardson et al., 2004). In the future, more Thus, we suggest the inclusion of Capsicodendron (Hoehne, 1934, non-coding nuclear single copy data will be applied, as these Capsicodendron dinisii) into the South American Cinnamodendron regions are most promising at the Cinnamosma species level. There (Endlicher, 1840, Cinnamodendron axillare). Such a clade would also is a large pool of over 900 single copy nuclear loci (Duarte et al., be supported by two morphological synapomorphies: a stipe 2010) that are potentially orthologous in any angiosperm lineage below the fruit and a connective projection above the anther and can be extracted from rapidly growing databases (e.g. http:// (Salazar, 2006). The relationship ((Warburgia + Capsicodendron/ onekp.com/project.html). We suggest identification and applica- South American Cinnamodendron) Cinnamosma) is recovered here tion of nuclear single copy genes as the most efficient and valuable for the first time (Fig. 1) and is congruent with the composition strategy to complement plastid genome derived phylogenetic of secondary metabolites (Bastos et al., 1999). Thorough morpho- hypotheses for any phylogenetic study within flowering plants. logical re-investigation, possibly including pollen characteristics as well as the inclusion of fossils, in the light of our molecular phy- logenetic results, are advisable prior to a genus level revision. 4.2. Phylogeny, systematics and morphology
The phylogenetic results demonstrate the power of the combin- 4.3. BBM versus DEC ancestral area reconstructions ing chloroplast and nuclear DNA datasets to resolve and statisti- cally support previously ambiguous relationships among Our BBM analyses are free from sampling bias as both BBM Canellales. The monophyly of both Cinnamosma and Warburgia as analyses returned identical results. BBM analyses provide mostly previously shown (Karol et al., 2000; Salazar and Nixon, 2008)is one distribution with higher likelihood for areas resulting in sev- substantiated and statistically supported by our diverse taxon eral long distance dispersal (LDD) events. The reconstruction of diverse sampling. In addition, the phylogenetic hypothesis sup- the deepest node is beyond validity of the method because no fossil ports the proposal of Salazar (2006) to split Cinnamodendron into distribution data can be implemented in BBM and calculated distri- two well-supported groups. The first group contains the Central bution areas of node 0 do not match the fossil distribution of this American species and those from the Greater Antilles, whereas node. In contrast, DEC allows constraints to incorporate a priori fos- the second contains South American taxa (Figs. 1 and S2). Salazar sil distribution data and dispersal probabilities. In general, DEC (2006) proposed the name ‘‘Antillodendron’’ for the species from analyses suggest broader distribution areas and it might be argued the Antilles, replacing Cinnamodendron. However, as shown by that this is a clear evidence for vicariance. However, tectonic sep- Zimmer et al. (2012) and confirmed here, ‘‘Antillodendron’’ is not aration of the respective distribution areas occurred much earlier monophyletic but also includes Pleodendron (Figs. 1 and S2). Conse- than the respective splits at particular nodes, e.g. (1) N3 the Antil- quently, instead of establishing a new genus, Cinnamodendron lean species split from the South American, African and Malagasy 216 S. Müller et al. / Molecular Phylogenetics and Evolution 84 (2015) 205–219 lineages by 43 MYA (mean age) although the separation of the age between N2 and N3) and therefore Central America and Antil- Antilles from South America by the Venezuelan and Haitian Basin les are likely part of the ancestral distribution area, and other areas occurred around 72 MYA (Meschede and Frisch, 1998); and (2) are less likely but not impossible. At node 3 a Central American N4 splits Madagascar from Africa and South America at 25 MYA clade, resulting in Pleodendron and Antillean Cinnamodendron, (mean age) but South America separated from Africa 135 MYA, and a South American, Malagasy and African clade splitted. For with a connection that remained until 110–95 MYA, and Madagas- the MRCA of Cinnamosma, Warburgia, Capsicodendron and South car separated from Africa 121 MYA (Sanmartín and Ronquist, American Cinnamodendron (N4) multiple distribution areas and 2004). DEC thus favors broad distributions resulting in vicariance LDD routes are possible. DEC reconstructed a broad ancestral dis- explanation that do not match the fossil record or tectonic events. tribution area with South America, Africa, and Madagascar at node The limitations of both methods together with the unique phyloge- 4. However this distribution area is a result of the broad, but as netic relationships, the distribution of extant lineages, and the lim- mentioned above unlikely, ancestral distribution area of node 3. ited fossil record require the consideration of all lines of evidence BBM is in favor of LDD from the Central America and the Antilles in conjunction to obtain the most plausible biogeographic picture. to Madagascar without colonization of Africa. We also consider a colonization of Madagascar via Africa as likely, because it might 4.4. Spatio-temporal origin of Canellales well be that the ancestor got extinct in Africa and possible fossils have not yet been discovered. Extinction can imprint the biogeog- Vicariance scenarios are challenged by our age estimates and raphy but cannot be falsified with a likely incomplete fossil record biogeographic results, which instead suggest that the majority and therefore difficult to include in biogeographic reconstructions. diversified after the Gondwana breakup. However, with Walkeri- pollis and Qatanipollis fossils found in Israel, Gabon, and Patagonia (Walker et al., 1983; Doyle et al., 1990a,b; Barreda and 4.5. Extant lineages of Canellaceae Archangelsky, 2006; Doyle and Endress, 2010; Schrank, 2013) and a land connection between Africa and South America until Our dating and biogeography results are partly in accordance 110–95 MYA (Sanmartín and Ronquist, 2004) a distribution for with ideas published previously with respect to a late divergence the MRCA of Canellales in Africa and South America is proven. of South American, African and Malagasy Canellaceae as well as Fragmentation of current distribution areas predates our age dispersal to Madagascar (Karol et al., 2000). MacArthur and estimates for Winteraceae (Fig. 2 and Table 5), except possibly Wilson (1967) predicted that a high percentage of a species migrat- for the split of Takhtajania. The separation of India and Madagascar ing onto an island derive from the nearest mainland. In addition to took place about 84 MYA (Sanmartín and Ronquist, 2004) which is the theory of MacArthur and Wilson, several modes of dispersal, within the 95% HPD of node 1 (Fig. 2 and Table 5). Therefore, colonization and migration between Africa and Madagascar, such except for Takhtajania, vicariance is a rather unlikely explanation as land bridges (e.g. McCall, 1997) and swimming or rafting step for the distribution of extant Winteraceae, which is in contrast to by step (e.g. Paulian, 1984; Krause, 2003) have been suggested to Marquínez et al. (2009). A clear explanation for that discrepancy explain the exchange of the biotas of Africa and Madagascar. These is the treatment of all fossil calibrations as minimum crown age concepts support our current understanding of Malagasy plant constraints by Marquínez et al. (2009), resulting in significantly groups as not being old enough to be affected by the Gondwana older nodes in their study. In addition, the Walkeripollis fossil breakup. It is generally accepted that Madagascar’s flora is mostly was placed as a crown group age constraint instead of a stem age derived from descendants of Cenozoic dispersed biota, typically constraint for Winteraceae (see Doyle, 2000; Doyle and Endress, from Africa (Yoder and Nowak, 2006). 2010; Thomas et al., 2014), resulting in even older nodes. Malagasy Cinnamosma species are either a result of dispersal Thomas et al. (2014) used Walkeripollis to calibrate the stem age from a New World ancestor, or originated from an African ancestor of Winteraceae, resulting in much younger age estimates, suggest- that went extinct whereas the Malagasy descendant survived. ing a Gondwana vicariance pattern for Takhtajania, but LDD for Extinction events may have played an important role in the distri- other nodes. With a sampling focused on Canellaceae, our study bution of Canellaceae. Occurrence and origin of Warburgia in supports a possible vicariance pattern for Takhtajania and LDD Africa, and Capsicodendron and Cinnamodendron in South America for other nodes, but with a limited Winteraceae sampling our data- (N6) are explained by extinction (N2) and recolonization. Extinc- set does not provide dispersal details for Winteraceae. tion events are supported by ecology patterns of extant Canella- Our age estimates and biogeographic reconstructions for ceae lineages, which are rarely dominant in their habitats and Canellaceae also indicate a post Gondwana-breakup diversifica- many species are extremely rare e.g. Pleodendron costaricense tion. Given a MRCA of Canellales that was at least distributed in (Hammel and Zamora, 2005). These ecological patterns suggest South America, Africa (according to the fossil record), Central that Canellaceae in general may potentially be prone to extinction America and the Antilles (BBM) we have to assume extinction in (Field pers. observation). Given the sparse fossil record of Canella- South America and Africa before node 2 (the MRCA of Canellaceae). ceae, we might assume that their ancestors showed similar ecolog- Distribution in the northern hemisphere of the New World is not ical patterns. Especially climate changes during Miocene and the only supported by the Wilsonoxylon fossil and DEC, but also by resulting need for dry adaptation may have a strong impact on the BBM analysis (Central America and the Antilles, Figs. 3 and the biogeography. Drying in much of tropical Africa led to a more 4). Furthermore, the Antilles were suggested to be the center of savanna-like character and several rainforest species became diversity of Canellaceae because most species occur in the area extinct (e.g. Winteraceae in Africa) (Coetzee and Muller, 1984; (Salazar and Nixon, 2008). Node 3 is either reconstructed as being Cerling et al., 1997). In contrast to Winteraceae, Canellaceae clearly of Central American and Antillean distribution (BBM) or multiple have abilities to evolve much greater drought tolerance (Feild et al., different combinations of ancestral distributions but always 2011) and therefore radiate into drier areas. Interestingly, we including Central America and the Antilles for reconstructed areas found some evidence for this pattern, because dry adapted Cinna- with a likelihood of >10% (DEC). Given a Central American and mosma species are derived from wet adapted taxa within Madagas- Antillean distribution at node 2 only, massive LDD would be car (T.S. Feild, unpublished data). Therefore, Cinnamosma species needed to give rise to a broad distribution area possibly including may be descendants from a wet adapted and extinct, possibly Afri- all areas of the extant Canellaceae (node 3, DEC). However, massive can Canellaceae rainforest lineage. Finally, recent (15–10 MYA) LDD must have occurred in a relative short time frame (6 MY mean Cinnamosma diversification may be caused by local climate S. Müller et al. / Molecular Phylogenetics and Evolution 84 (2015) 205–219 217 changes due to increased tectonic activity in Madagascar (Wit, America, the American Society of Plant Taxonomists, the Interna- 2003). tional Association for Plant Taxonomy, the Linnean Society of Lon- don, and the Botanical Garden of Santo Domingo. We appreciate 4.6. Long-distance dispersal and potential vectors the support of the Governments of Kenya, Costa Rica, Dominican Republic, Ethiopia, Madagascar, Jamaica, Cuba, Tanzania and Since most fruits of Canellales are berries and eaten by birds Puerto Rico for research and collection permits. The study was sup- (Kubitzki et al., 1993 and our own observations), the dispersal of ported by the German Research Foundation (NE 681/11-1). Canellaceae from the New to the Old World is facilitated. Canella- ceae fruits are frequently red to dark purple (e.g. Canella winterana, Appendix A. Supplementary material Pleodendron macranthum, Cinnamodendron occhionii) and endozo- ochory by birds is reported (Hoehne, 1947; Occhioni, 1948; Supplementary data associated with this article can be found, in Snow, 1981; Roosmalen, 1985; Nunes et al., 2003) as well as by the online version, at http://dx.doi.org/10.1016/j.ympev.2014.12. fruit bats feeding on Cinnamosma (Bollen et al., 2004). Since 010. fruit-bats colonized Madagascar at least 20 times independently, mostly from Africa, they possess a great potential as vectors for dispersal (Ruedi et al., 2012). Warburgia ugandensis, however, is References dispersed by apes in Africa (Majid et al., 2011 and our own obser- vations), and in some Malagasy areas certain lemurs are believed Álvarez, I., Wendel, J.F., 2003. Ribosomal ITS sequences and plant phylogenetic inference. Mol. Phylogenet. Evol. 29, 417–434. to be the sole dispersers of Cinnamosma species. Sato (2012) found Barreda, V., Archangelsky, S., 2006. The southernmost record of tropical pollen that in Malagasy tropical dry forest Eulemur fulvus fulvus is the sole grains in the mid-Cretaceous of Patagonia, Argentina. Cretac. Res. 27, 778–787. disperser for plants with large seeds, e.g. Cinnamosma fragrans http://dx.doi.org/10.1016/j.cretres.2006.02.002. Bastos, J.K., Kaplan, M.A.C., Gottlieb, O.R., 1999. Drimane-type sesquiterpenoids as (seed diameter 10.54 ± 1.36 mm). Bollen (2007) stated that Cinna- chemosystematic markers of Canellaceae. J. Braz. Chem. Soc. 10, 136–139. mosma madagascariensis var. namoronensis is exclusively dispersed Bollen, A., 2007. Fruit characteristics: fruit selection, animal seed dispersal and by Eulemur fulvus collaris. Nevertheless, dispersal from Africa to conservation matters in the Sainte Luce forests. Biodiversity, Ecology, and Madagascar by ancient lemurs is unlikely, because the MRCA of Conservation of Littoral Ecosystems in the Region of Tolagnaro (Fort Dauphin), Southeastern Madagascar. Smithsonian Institution, Washington, DC, pp. 127– lemurs colonized Madagascar only once (Karanth et al., 2005) 145. and considerably earlier (60 MYA, 70–41 95% HPD, Poux et al., Bollen, A., Van Elsacker, L., Ganzhorn, J.U., 2004. Tree dispersal strategies in the 2005) than the respective Canellaceae split. Combining these lines littoral forest of Sainte Luce (SE-Madagascar). Oecologia 139, 604–616. http:// dx.doi.org/10.1007/s00442-004-1544-0. of evidence we suggest extant Canellaceae distribution being Boonchai, N., Manchester, S.R., 2012. Systematic affinities of early Eocene petrified shaped by extinction events, including at least Africa and South woods from Big Sandy Reservoir, southwestern Wyoming. Int. J. Plant Sci. 173, America; LDD through birds and/or bats, and dispersal within sin- 209–227. Borsch, T., Hilu, K.W., Quandt, D., Wilde, V., Neinhuis, C., Barthlott, W., 2003. gle landmasses by apes (Africa) and lemurs (Madagascar). Noncoding plastid trnT-trnF sequences reveal a well resolved phylogeny of basal angiosperms. J. Evol. Bio. 16, 558–576. http://dx.doi.org/10.1046/j.1420- 9101.2003.00577.x. 5. Conclusion Browne, P., Ehret, G.D., 1756. The Civil and Natural History of Jamaica: in three parts. Buerki, S., Forest, F., Alvarez, N., Nylander, J.A.A., Arrigo, N., Sanmartín, I., 2011. An Canellaceae, as many of the earliest angiosperm lineages, show evaluation of new parsimony-based versus parametric inference methods in a potential Gondwana distribution. They are not useless relicts, but biogeography: a case study using the globally distributed plant family play an important role for understanding the earliest flowering Sapindaceae. J. Biogeogr. 38, 531–550. plant evolution. This is due to their very dynamic evolution and Cai, Z., Penaflor, C., Kuehl, J.V., Leebens-Mack, J., Carlson, J.E., W dePamphilis, C., Boore, J.L., Jansen, R.K., 2006. Complete plastid genome sequences of Drimys, biogeography. Thus, the most parsimonious explanation may not Liriodendron, and Piper: implications for the phylogenetic relationships of necessarily be correct. We propose that the distribution of Canell- magnoliids. BMC Evol. Biol. 6, 77. aceae is a result of vicariance, extinction and at least one intercon- Cerling, T.E., Harris, J.M., MacFadden, B.J., Leakey, M.G., Quade, J., Eisenmann, V., Ehleringer, J.R., 1997. Global vegetation change through the Miocene/Pliocene tinental LDD event. While molecular dating and fossil evidence boundary. Nature 389, 153–158. http://dx.doi.org/10.1038/38229. together prove the distribution of the MRCA of Canellales in Africa Coetzee, J.A., Muller, J., 1984. The phytogeographic significance of some extinct and South America before the breakup of these continents, our data Gondwana pollen types from the Tertiary of the southwestern Cape (South Africa). Ann. Mo. Bot. Gard. 71, 1088. http://dx.doi.org/10.2307/2399246. suggests extinction in both continents before the MRCA of Canell- Cronquist, A., 1981. An Integrated System of Classification of Flowering Plants. aceae evolved. Central America and the Antilles are likely the Columbia University Press, New York. diversity center and ancestral area of Canellaceae, while patterns Wit, M.J. de, 2003. Madagascar: heads it’s a continent, tails it’s an island. Ann. Rev. Earth Planet. Sci. 31, 213–248. http://dx.doi.org/10.1146/ and timing of colonization of South America, Africa and Madagas- annurev.earth.31.100901.141337. car remain more speculative with potential parallel colonization of Désamoré, A., Laenen, B., González-Mancebo, J.m., Jaén Molina, R., Bystriakova, N., Africa and Madagascar. Finally, our molecular data suggest that Martinez-Klimova, E., Carine, M.A., Vanderpoorten, A., 2012. Inverted patterns of genetic diversity in continental and island populations of the heather Erica natural relationships, species delimitation, and systematics of Cin- scoparia s.l. J. Biogeogr. 39, 574–584. http://dx.doi.org/10.1111/j.1365-2699. namosma will require further studies applying more nuclear 2011.02622.x. encoded low copy genes, and in particular their introns. Doyle, J.A., 2000. Paleobotany, relationships, and geographic history of Winteraceae. Ann. Mo. Bot. Gard. 87, 303–316. Doyle, J.A., Endress, P.K., 2010. Integrating Early Cretaceous fossils into the Acknowledgments phylogeny of living angiosperms: magnoliidae and eudicots. J. Syst. Evol. 48, 1–35. Doyle, J.A., Hotton, C.L., Ward, J.V., 1990a. Early Cretaceous tetrads, zonasulculate Sincere thanks to J.A. Doyle and an unknown reviewer whose pollen, and Winteraceae. II. Cladistic analysis and implications. Am. J. Bot. 77, comments substantially improved the manuscript. Field assistance 1558–1568. http://dx.doi.org/10.2307/2444488. and general support were provided by A. Veloz, B. Peguero, I. de los Doyle, J.A., Hotton, C.L., Ward, J.V., 1990b. Early Cretaceous tetrads, zonasulculate pollen, and Winteraceae. I. Taxonomy, morphology, and ultrastructure. Am. J. Angeles, G. Feliz, F. Pichardo Marcano, G. Proctor, K. Campbell, H. Bot. 77, 1544–1557. http://dx.doi.org/10.2307/2444487. Vilmond, E. Santiago-Valentin, M. Caraballo, T. Carlos, R. Berazain, Drummond, A.J., Ho, S.Y.W., Phillips, M.J., Rambaut, A., 2006. Relaxed phylogenetics B. Hammel, I. Pérez, F. de Barros, R. Goldenberg, G. Schatz, Y. Ezige- and dating with confidence. PLoS Biol. 4, 0699–0710. http://dx.doi.org/10.1371/ journal.pbio.0040088. ber, H. Ndangalsi and C. N. Fraga. JS received funding for studying Drummond, A.J., Suchard, M.A., Xie, D., Rambaut, A., 2012. Bayesian phylogenetics Canellaceae from Cornell University, the Botanical Society of with BEAUTi and the BEAST 1.7. Mol. Biol. Evol., 1069–1973 218 S. Müller et al. / Molecular Phylogenetics and Evolution 84 (2015) 205–219
Duarte, J.M., Wall, P.K., Edger, P.P., Landherr, L.L., Ma, H., Pires, J.C., Leebens-Mack, J., Occhioni, P., 1948. Contribuição ao estudo da família Canellaceae Arquivos do 2010. Identification of shared single copy nuclear genes in Arabidopsis, Populus, Jardim Botânico de Rìo de Janeiro, vol. 8, pp. 8–165. Vitis and Oryza and their phylogenetic utility across various taxonomic levels. Parham, J.F., Donoghue, P.C.J., Bell, C.J., Calway, T.D., Head, J.J., Holroyd, P.A., Inoue, J.G., BMC Evol. Biol. 10, 61. Irmis, R.B., Joyce, W.G., Ksepka, D.T., Patané, J.S.L., Smith, N.D., Tarver, J.E., van Endlicher, S.F.L., 1840. Genera Plantarum Secundum Ordines Naturales Disposita. Tuinen, M., Yang, Z., Angielczyk, K.D., Greenwood, J.M., Hipsley, C.A., Jacobs, L., Endress, P.K., Igersheim, A., Sampson, F.B., Schatz, G.E., 2000. Floral structure of Makovicky, P.J., Müller, J., Smith, K.T., Theodor, J.M., Warnock, R.C.M., Benton, M.J., Takhtajania and its systematic position in Winteraceae. Ann. Mo. Bot. Gard. 87, 2012. Best Practices for Justifying Fossil Calibrations. Syst. Biol. 61, 346–359. 347–365. Paulian, R., 1984. Madagascar: A microcontinent between Africa and Asia. Key Feild, T.S., Hudson, P.J., Balun, L., Chatelet, D.S., Patino, A.A., Sharma, C.A., McLaren, Environments: Madagascar, pp. 1–26. K., 2011. The ecophysiology of xylem hydraulic constraints by ‘‘basal’’ vessels in Posada, D., 2008. JModelTest: phylogenetic model averaging. Mol. Biol. Evol. 25, Canella winterana (Canellaceae). Int. J. Plant Sci. 172, 879–888. http://dx.doi.org/ 1253–1256. 10.1086/660106. Poux, C., Madsen, O., Marquard, E., Vieites, D.R., de Jong, W.W., Vences, M., 2005. Graham, A., 2010. A Natural History of the New World: The Ecology and Evolution of Asynchronous colonization of madagascar by the four endemic clades of Plants in the Americas. University of Chicago Press, Chicago, IL. primates, tenrecs, carnivores, and rodents as inferred from nuclear genes. Syst. Graham, A., Jarzen, D., 1969. Studies in Neotropical paleobotany. I. The Oligocene Biol. 54, 719–730. http://dx.doi.org/10.1080/10635150500234534. communities of Puerto Rico. Ann. Mo. Bot. Gard. 56, 308–357. Rambaut, A., Drummond, A.J., 2007. Tracer v1.4.
Vink, W., 1985. The Winteraceae of the Old World. V. Exospermum links Bubbia to Yu, Y., Harris A.J., Xingjin, H., 2014. RASP (Reconstruct Ancestral State in Zygogynum. Blumea 31, 39–55. Phylogenies). 3.0.