Molecular Phylogenetics and Evolution 87 (2015) 91–104
Contents lists available at ScienceDirect
Molecular Phylogenetics and Evolution journal homepage: www.elsevier.com/locate/ympev
Systematics, biogeography, and character evolution of Deutzia (Hydrangeaceae) inferred from nuclear and chloroplast DNA sequences Changkyun Kim a, Tao Deng a,b,c, Jun Wen d, Ze-Long Nie a, Hang Sun a,⇑ a
Key Laboratory for Plant Biodiversity and Biogeography of East Asia, Kunming Institute of Botany, Chinese Academy of Sciences, Kunming 650204, Yunnan, PR China School of Life Science, Yunnan University, Kunming 650091, PR China c University of Chinese Academy of Sciences, Beijing 100039, PR China d Department of Botany, MRC-166, Smithsonian Institution, PO Box 37012, Washington, DC 20013-7012, USA b
a r t i c l e
i n f o
Article history: Received 23 June 2014 Revised 23 February 2015 Accepted 2 March 2015 Available online 14 March 2015 Keywords: Biogeography Character evolution Dating analysis Deutzia Molecular phylogeny Infrageneric classification
a b s t r a c t The genus Deutzia (Hydrangeaceae), containing ca. 60 species circumscribed in three sections, is disjunctly distributed in eastern Asia and Central America (Mexico). Although the genus is well delimited, its subdivisions into sections and series have not been the subject of an explicit test of monophyly based on molecular data. A comprehensive examination of the evolutionary relationships within the genus is thus still lacking. We present a fossil-calibrated, molecular phylogeny of Deutzia based on two nuclear ribosomal DNA (ITS and 26S) and three chloroplast DNA regions (matK, rbcL, and trnL-F intergenic spacer). Within this framework, we examine character evolution in petal arrangement, filament shape, and the number of stamens, and infer the ancestral area and biogeographic history of the genus. Our molecular phylogeny suggests that Deutzia is monophyletic. Two major clades are recovered: one composed of the species of sect. Neodeutzia from Mexico, and the other containing all remaining Deutzia species of sections Mesodeutzia and Deutzia from SW China and Northeast Asia. The latter two Asian sections were each revealed to be polyphyletic. The induplicate petals, 2-dentate filaments, and polystemonous androecia are inferred to be ancestral character states. Biogeographic reconstructions suggest a Northeast Asian origin for the genus and subsequent spread to Mexico during the Oligocene and to SW China during the Miocene. Based on our results, a new infrageneric classification of Deutzia inferred from molecular phylogeny is required. We propose to merge sections Mesodeutzia and Deutzia to ensure the monophyly at the sectional level. Cooling trends during the Oligocene resulted in isolation, separating eastern Asian and Mexican taxa, while the warm period during the middle Miocene stimulated the diversification from Northeast Asia to SW China. The uplift in the Qinghai-Tibetan Plateau and monsoon regimes are important in promoting high species diversification of Deutzia in SW China. Ó 2015 Elsevier Inc. All rights reserved.
1. Introduction The temperate flora of the regions of the Northern Hemisphere exhibits striking intercontinental floristic similarities (Qian, 1999; Wen et al., 2010). Numerous temperate plant genera exhibit a disjunct distribution pattern in two or more of the five now isolated areas: eastern Asia, eastern North America, western North America, Central America, and Europe (Li, 1952; Tiffney, 1985; Qian, 1999; Wen, 1999; Wen et al., 2010). Of these, the similarity is most remarkable between eastern Asia and North America (Qian, 1999). Because of its particular significance to biogeography, ⇑ Corresponding author at: 132 Lanhei Road, Key Laboratory for Plant Biodiversity and Biogeography of East Asia, Kunming Institute of Botany, Chinese Academy of Sciences, Kunming 650204, Yunnan, PR China. Fax: +86 871 65215002. E-mail address:
[email protected] (H. Sun). http://dx.doi.org/10.1016/j.ympev.2015.03.002 1055-7903/Ó 2015 Elsevier Inc. All rights reserved.
the group of disjunct eastern Asian–North American genera has been much studied since the time of Linnaeus (reviewed by Boufford and Spongberg, 1983). The floristic disjunction between eastern Asia and North America has been considered to be the result of the fragmentation of a once continuous mixed mesophytic forest, a process that occurred throughout the Northern Hemisphere due to the climatic and geological changes during the late Tertiary and Quaternary, although long distance dispersal has also been proposed to explain the observed distributional pattern (e.g., Sulman et al., 2013). Despite long-standing interest in the disjunction between eastern Asia and eastern/western North America, only a few taxa have been examined using molecular phylogenetic and biogeographic methods to determine the evolution between eastern Asia and Central America (Baird et al., 2010; Guo et al., 2013; Wang et al., 2015). Moreover, evolutionary history among taxa displaying the disjunction between eastern
92
C. Kim et al. / Molecular Phylogenetics and Evolution 87 (2015) 91–104
Asia and Central America is controversial. Philadelphus L. was inferred to have originated in western North America, and later dispersed into Central America and Asia (Guo et al., 2013), whereas Leibnitzia Cass. originated in eastern Asia and later reached to North and Central America (Baird et al., 2010). To better understand the disjunction between eastern Asia and Central America, we need to evaluate phylogenetic relationships and estimate divergence times in more lineages that exhibit this distribution pattern. Deutzia Thunb. is the second largest genus (ca. 60 species) of the tribe Philadelpheae (Hydrangeaceae) and its economically important as ornamentals (Huang et al., 2001; Hufford, 2004). Deutzia species are distributed across eastern Asia and Mexico (Central America), but most species (>50) occur in eastern Asia (Fig. 1). By far the highest species diversity is in SW China including on the Qinghai-Tibetan Plateau (QTP) and adjacent areas, where 35 species occur (28 endemic) (He, 1990; Huang et al., 2001). Four species endemic to Mexico (D. mexicana, D. pringlei, D. occidentalis, D. oaxacana) have been reported (Hwang, 1993). With the richest species diversity, it has been suggested that the center of diversity of the genus is in SW China (Hwang, 1993). Some of the genera exhibiting the highest species diversification in SW China originated on the QTP and in adjacent regions and then migrated to other regions of the Northern Hemisphere (Xu et al., 2010; Li et al., 2014), others showed the reverse pattern, originating in other regions of the world and diversifying greatly after their ancestors arrived on the QTP (Tu et al., 2010; Wen et al., 2014). Thus, the diversity center of a taxon does not necessarily correspond to the center of origin. It is also unclear whether the Mexican species, which are uniquely polystemonous in the genus, are a relict group originated from an initial radiation of the genus in Mexico (Hufford, 2004). Thus, biogeographic analysis is required to elucidate the phylogeographic history of the genus using DNA markers. Previous phylogenetic studies of Hydrangeaceae have supported the placement of Deutzia in the tribe Philadelpheae (Soltis et al., 1995; Fan and Xiang, 2003; Xiang et al., 2011). Within Philadelpheae, Deutzia is most closely related to Kirengeshoma or Philadelphus according to the morphological and molecular evidence (Styer and Stern, 1979a; Soltis et al., 1995; Hufford, 1997, 2004; Fan and Xiang, 2003). However, the members of Deutiza can be easily distinguished from Philadelphus by having stellate hairs on their leaves, five petals, and 10–15 stamens in their
flowers (Hufford, 1997, 2004). Deutzia is also distinguished from Kirengeshoma by habit (shrub vs. herb) and leaf shape (not palmate [ovate, elliptic, and/or lanceolate] vs. palmate; Hufford, 2004). Despite the morphological distinctiveness of Deutzia as a genus, its monophyly and sister group are unclear because previous phylogenetic studies focused on family- or genus-level relationships or were based on sampling that was not representative of the entire genus, including few and/or relatively distantly related outgroups (Soltis et al., 1995; Fan and Xiang, 2003). Thus, the monophyly and the sister of this genus need to be evaluated by a broader taxon sampling scheme of Deutzia. Until now, the infrageneric classification of Deutzia has been based on morphology (Zaikonnikova, 1966; Hwang, 1993). Three sections are recognized within the genus – Deutzia, Mesodeutzia, and Neodeutzia – based on petal arrangements and the number of stamens (Zaikonnikova, 1966; Hwang, 1993): sect. Deutzia (with induplicate petals, 2-dentate filaments, and 10 stamens per flower), sect. Mesodeutzia (with imbricate petals, edentate or 2dentate filaments, and 10 stamens per flower), and sect. Neodeutzia (with induplicate petals, edentate filaments, and 12– 15 stamens per flower). Section Deutzia is subdivided into four subsections: subsect. Deutzia, Grandiflorae, Stenosepalae, and Cymosae based on the filament shape and length. Mesodeutzia is also subdivided into three series: Parviflorae, Rubentes, Corymbosae based on the shape of filaments and the number of flowers per inflorescence. Attempts to subdivide the genus using morphological data have been hampered by the extensive quantitative and overlapping characters (Huang et al., 2001; Hufford, 2004). Thus, many questions remain unanswered, in particular the relationships between major lineages and infrasectional groups within Deutzia. In addition to being one of the largest genera in Hydrangeaceae, Deutzia is also remarkably diverse in terms of petal arrangements, shape of filaments, and number of stamens (Hwang, 1993; Huang et al., 2001). There are two types of petal arrangement: induplicate and imbricate which have been relied upon heavily in the infrageneric classification of Deutzia (Hwang, 1993). The filaments exhibit two types: edentate and 2-dentate, with the majority 2-dentate, a state not typically found in the genera of Hydrangeaceae. A transition in stamen number between 10 and 12–15 occurs within Deutzia, although the number of such independent shifts is unknown. These variations in morphology have undoubtedly contributed to controversies surrounding circumscription of the
Fig. 1. Distribution of Deutzia species indicated by green shading, with species numbers shown in each geographical area. The dark blue shaded area represents the current center of species diversity in SW China. (For interpretation of the references to colour in this figure legend, the reader is referred to the web version of this article.)
C. Kim et al. / Molecular Phylogenetics and Evolution 87 (2015) 91–104
93
species- and higher-level relationships (Hwang, 1993). However, the evolutionary transition of characters has not been examined in a phylogenetic context within Deutzia. Here, we employ a combined set of nuclear (nrDNA) and chloroplast (cpDNA) gene regions, temporally fossil-calibrated, to investigate a series of questions related to systematics and biogeography in Deutzia. The data set includes two regions of nrDNA (ITS and 26S) and three cpDNA regions (matK, rbcL, trnL-F intergenic spacer [IGS]). We examined the five gene regions across the broadest taxon sampling of Deutzia and outgroups in order to shed light on the following questions about Deutzia evolution. (1) Is the genus Deutzia monophyletic? (2) Which genera are the closest relatives of the genus? (3) Are the sections and subsections composed of natural clades? (4) When and where did divergence events among Deutiza species occur? (5) How many transitions have occurred in key characters of Deutzia species, and what are the ancestral states?
TGTGTGGACCGATGGACTTA) and DerbcLR (50 -TACGGAATCATCCC CAAAGA). All PCR amplifications were carried out in a total volume of 25 lL containing 2.5 lL of 10 PCR reaction buffer, 0.25 mM of each dNTP, 10 pmol of each primer, 1U Taq DNA polymerase (Takara, Dalian, China), and 200 ng of DNA template. Amplifications of a specific region were performed on a PTC-200 thermal cycler (MJ Research, Waltham, USA) under the following reaction conditions: initial denaturation for 4 min at 94 °C; 35 cycles of 1 min at 94 °C, 1 min at 53 °C (for 26S, matK, and rbcL) or 55 °C (for ITS and trnL-F), and 1 min at 72 °C; and terminal extension for 10 min at 72 °C. The amplified DNA samples were purified for sequencing using a PCR Purification Kit (Qiagen, Valencia, CA, USA) in accordance with the supplier’s specifications. All PCR products were directly sequenced in both directions using the amplification primers on an ABI3730 automated sequencer (Applied Biosystems, Foster City, CA, USA).
2. Materials and methods
2.3. Phylogenetic analysis
2.1. Taxon sampling
DNA Baser v.3 (http://www.DnaBaser.com) was used to evaluate chromatograms for base confirmation and to edit contiguous sequences. Multiple-sequence alignment was performed using MAFFT (http://www.genome.jp/tools/mafft/; Katoh and Toh, 2008) with the default alignment parameters and then edited manually. We excluded the poly A or poly T regions from the trnL-F data set because homology assessment can be very difficult for these repeated nucleotides (Kelchner, 2000) and they might be technical artifacts of the PCR amplification (Clarke et al., 2001). Preliminary phylogenetic analyses were conducted to explore the distribution of phylogenetic signals in the nrDNA and cpDNA data matrix with and without coded gaps. Neither resolution nor support was strikingly different between analyses including and excluding gaps (results not shown), and therefore gaps in the alignments were treated as missing data and indels were not coded. The alignments used in this study are available from the online database TreeBASE (http://www.treebase.org/; study accession number, 15843). The incongruence length difference (ILD) test was used to assess data congruency between the nrDNA and cpDNA datasets (Farris et al., 1995) by using PAUP⁄ v.4.0b10 and 1000 heuristic search replications. The ILD test indicated significant incongruence (P = 0.012) between the two data sets (nrDNA and cpDNA). Because of the stochastic manner in which lineages sort during speciation and hybridization (Maddison, 1997; Nakhleh, 2010), it is common that gene trees differ in topology from each other. Whether conflicting data sets should be analyzed separately or combined in a simultaneous analysis is a complicated and contentious decision. In the current study, the combined data set was significant for the phylogenetic resolution of Deutzia although some nodes were still weakly supported (Fig. 2). Furthermore, few strongly supported (BP > 90) incongruent clades were found upon comparison of parsimony bootstrap consensus trees generated from the two data sets. Thus, we chose to combine nrDNA and cpDNA data sets with their weakly-incongruent trees, as suggested by Sheahan and Chase (2000) and Nie et al. (2008). Phylogenetic reconstructions of nrDNA, cpDNA, and combined datasets were performed by maximum parsimony (MP) methods with PAUP⁄ v.4.0b10 (Swofford, 2002). All characters and character states were weighted equally and unordered. Searches were conducted using 100 random-taxon-addition replicates with tree bisection-reconnection (TBR) branch swapping, and MulTrees in effect, using maxtrees = 10 000. Bootstrap analyses (1000 pseudoreplicates) were conducted to examine the relative level of support for individual clades on the cladograms of each search
A total of 153 accessions were included in the phylogenetic analyses, altogether representing 56 species (106 accessions) of Deutzia and 37 outgroup taxa (47). Within Deutzia, 10 species (21 accessions) were included in sect. Mesodeutzia, 44 species (82) in sect. Deutzia, and two species (3) in sect. Neodeutzia. Based on results from previous phylogenetic analyses of Hydrangeaceae (Soltis et al., 1995; Fan and Xiang, 2003; Xiang et al., 2011), our outgroups were selected in order to sample taxa that are closely related to Deutzia as well as some that are more distantly related within the Hydrangeaceae. They include five genera (Carpenteria, Fendlerella, Kirengeshoma, Philadelphus, and Whipplea from Hydrangeoideae tribe Philadelpheae) that are closely related to Deutzia and five more distantly related genera (Dichroa, Decumaria, and Hydrangea [Hydrangeoideae tribe Hydrangeae], Fendlera and Jamesia [Jamesioideae]). We also included five species (Cajophora cirsiifolia, Eucnide bartonioides, E. urens, Mentzelia lindleyi, Petalonyx nitidus) from Losaceae and one species (Cornus controversa) from Cornaceae as outgroups. Taxa sampled, voucher information, and GenBank accession numbers for the five data sets are listed in Appendix A.
2.2. DNA extraction, PCR amplification, and sequencing Genomic DNA was extracted from silica-dried plant material and herbarium specimens using a Universal Genomic DNA Extraction Kit (Takara, Dalian, China) according to the manufacturer’s specifications. The five gene regions, including two nuclear ribosomal DNA regions (nrDNA; ITS and 26S) and three chloroplast DNA regions (cpDNA; matK, rbcL, and trnL-F IGS), were employed. The ITS region was amplified and sequenced using the primers ITS1 and ITS4 (White et al., 1990). The cpDNA trnL-F region was amplified and sequenced using primers ‘‘e’’ and ‘‘f’’ (Taberlet et al., 1991). A new primer combination for the amplification of nrDNA 26S and cpDNA matK and rbcL was designed based on the sequences of D. scabra available from GeneBank (accession numbers JF321107 [26S], JF308685 [matK], and JF308656 [rbcL]). A small portion of the 26S region (892 bp long) was amplified and sequenced using primers De26SF (50 -TCCGTCCAAGGCTAAATACG) and De26SR (50 GATTCGGCAGGTGAGTTGTT). The matK region was amplified using primers DematKF (50 -TCCCATCCATCTGGAA ATCT) and DematKR (50 -CCCCTGCGAAGTAGAAGAAG). PCR amplification of the rbcL region was performed with primers DerbcLF (50 -
94
C. Kim et al. / Molecular Phylogenetics and Evolution 87 (2015) 91–104
Fig. 2. A 50% majority-rule consensus tree of the 10 000 most parsimonious trees from parsimony analysis of the combined sequence data from two nrDNA (nrITS and 26S) and three cpDNA (matK, rbcL, and trnL-F) regions. (A) Outgroup families and Hydrangeaceae outside Deutzia. (B) The genus Deutzia. Sections and infrasections of Deutzia are shown as delimited in this study. The numbers near nodes indicate corresponding support values (maximum parsimony bootstrap [BP]/Bayesian posterior probability [PP]); an en dash (–) indicates that a node did not receive P75% BP in the maximum parsimony analysis.
C. Kim et al. / Molecular Phylogenetics and Evolution 87 (2015) 91–104
(Felsenstein, 1985). Bootstrap percentage (BP) was categorized as strong (>90%), moderate (80–90%), or weak (70–79%). Phylogenetic analyses of the individual and combined sequences were also conducted under Bayesian Markov Chain Monte Carlo (MCMC) inference (BI; Rannala and Yang, 1996) using MrBayes v.3.12 (Ronquist and Huelsenbeck, 2003). Based on the Akaike information criterion (AIC; Akaike, 1974), MODELTEST v.3.1 (Posada and Crandall, 1998) assigned the GTR + I + G model of molecular evolution to the nrDNA, cpDNA, and the combined data set. Four MCMC simulations were run simultaneously and sampled every 100 generations for a total of 20 million generations. The first 50 000 (25%) of the sample trees from each run were discarded (they represented the burn-in) as determined by Tracer v.1.5 (Rambaut and Drummond, 2007). The remaining trees were used to construct a 50% majority-rule consensus tree, with the proportion bifurcations found in this consensus tree given as posterior probability (PP) to estimate robustness of the BI trees.
2.4. Estimates of divergence times To estimate the divergence time of the clades and subclades, we used a reduced data set of 75 species, including 56 species of Deutzia and 19 outgroups with one sample per species (Appendix A) because all accessions were grouped according to their taxonomic species. We used BEAST v.1.5.4 (http://beast.bio.ed.ac.uk; Drummond and Rambaut, 2007) based on the combined nrDNA and cpDNA sequence data. In order to generate input files for BEAST, the BEAUti interface was used, in which a selected model (GTR + I + G) for the combined dataset was applied with a Yule speciation tree prior and an uncorrelated lognormal molecular clock model. Two runs of 10 million generations of the MCMC chains were produced, sampling every 1000 generations. Convergence of the stationary distribution was checked by visual inspection of plotted posterior estimates using Tracer v.1.5 (Rambaut and Drummond, 2007). After discarding the first 1000 trees as burn-in, the samples were summarized in the maximum clade credibility tree using TreeAnnotator v1.6.1 (Drummond and Rambaut, 2007) with a PP limit of 0.50 and summarizing mean node heights. Means and 95% higher posterior densities (HPDs) of age estimates are obtained from the combined outputs using Tracer. The results were visualized using FigTree v.1.3.1 (Rambaut, 2009). We constrained the ages of four nodes in the phylogeny of Deutzia and its close relatives. First, the crown age of Jamesioideae (Fedlera and Jamesia) was constrained with a uniform distribution from 23 to 28.4 million years ago [mya] based on the fossil leaves of Jamesia caplanii Axelrod. These fossil leaves are very similar to those of the extant J. americana, which is a common deciduous shrub in the mixed conifer forest of southwestern America (Axelrod, 1987). Second, the crown age of the macrophylla clade of Hydrangeoideae was constrained to be 28.4 (±1) mya with a normal distribution based on the fossil seeds of Dichroa bornensis Mai, described from the Middle Oligocene of Europe (Mai, 1998) and related to the extant D. febrifuga species. Due to the insufficient character states of the fossil species, however, it can be reasonably assigned to the macrophylla clade of Hydrangea (Hufford et al., 2001). Third, the node of Schizophragma was calibrated with a uniform distribution from 2.6 to 5.3 mya based on Pliocene fossils fruits of Schizophragma from Europe (Mai, 1985; Martinetto, 2001). Also, fossil leaves very similar to those of extant S. hydrangeoides were identified from the Pliocene Kabutoiwa Formation of Japan (Ozaki, 1991). Lastly, the divergence between Decumaria and Pileostegia was constrained to 2.38 (±1.69) mya with a normal distribution following Xiang et al. (2000), who estimated the divergence time of Decumaria using the rbcL sequence data.
95
2.5. Reconstruction of ancestral areas Biogeographic data for species within Deutzia were compiled from distributions described in the literature and for herbarium specimens. The distribution range of Deutzia and outgroup species was divided into four areas: A, SW China; B, other parts of eastern Asia (mainly Northeast Asia; E China, Korea, Japan, and far eastern Russia); C, Mexico (Central America); D, North America. We coded each species with the entire range of the species regardless of their biogeographic sources (Appendix A). Ancestral areas reconstruction (AAR) and estimating spatial patterns of geographic diversification within Deutzia were inferred using the Bayesian Binary Method (BBM) and Statistical dispersalvariance analysis (S-DIVA) as implemented in RASP v. 2.1b (Reconstruct Ancestral State in Phylogenies, formerly S-DIVA; Yu et al., 2010, 2013). The BBM was run with the fixed state frequencies model (Jukes-Cantor) with equal among-site rate variation for two million generations, ten chains each, and two parallel runs. In S-DIVA, the frequencies of an ancestral range at a node in ancestral reconstructions are averaged over all trees (Yu et al., 2010). For these analyses, we used all post burn-in trees that resulted from the BEAST analysis. The consensus tree used to map the ancestral distribution on each node was obtained with the Compute Condense option in RASP from the stored trees. The maximum number of ancestral areas was set to three. 2.6. Morphological character evolution Three morphological characters (petal arrangement, filament shape, and stamen number) were included in the morphological character evolution analysis. All three characters were selected because of their potential for inferring relationships among species in Deutzia (Hwang, 1993). Petal arrangement was coded as: (0) induplicate, (1); imbricate; filament shape as: (0) edentate, (1) 2dentate; and the number of stamens as: (0) 6 10, (1) 12–15, (2) > 20 (Appendix B). To infer patterns of character evolution, we used the phylogenetic framework provided by BEAST analysis of the combined nrDNA and cpDNA data set. Character reconstructions were performed under the assumption of unordered and unweighted character states using parsimony with the Ancestral State Reconstruction Package in Mesquite v.2.75 (Maddison and Maddison, 2011). 3. Results 3.1. Analysis of nrDNA sequence data analysis The aligned nrDNA data matrix comprised 1569 characters (nrITS, 677; 26S, 892). There were 494 (31.5%) variable sites, of which 349 (22.2%) were informative for parsimony (Table 1). Parsimony analysis of the nrDNA data resulted in >10 000 equally parsimonious trees (tree length = 1257; consistency index, CI = 0.518; retention index, RI = 0.899). The BI phylogram was identical in topology to the 50% majority-rule consensus tree sampled by MP analysis (trees not shown). The monophyly of Deutzia was supported (BP = 100, PP = 1.00). Kirengeshoma was found to be sister to Deutzia with the highest support. The genus was divided into two clades. Clade I comprised two species (D. mexicana and D. pringlei) of sect. Neodeutzia from Central America but with low support (BP = 52, PP = 0.78). Clade II, with a 90% bootstrap value, included 54 species of sections Mesodeutzia and Deutzia from SW China and Northeast Asia. Within this clade, four subclades were recognized (a–d in Fig. 2), but two sections Mesodeutzia and Deutzia were each polyphyletic.
96
C. Kim et al. / Molecular Phylogenetics and Evolution 87 (2015) 91–104
Table 1 Statistics from Modeltest and maximum parsimony (MP) analyses obtained from separate and combined datasets. Parameters
No. of DNA sequences (ingroup/outgroup) Aligned length (bp) No. of variable sites (%) No. of parsimony-informative sites (%) No. of trees (MP) No. of steps Consistency indexa Retention index Model selected by Akaike information criterion a
nrDNA
cpDNA
Combined
nrITS
26S
matK
rbcL
trnL-F
nrDNA
cpDNA
nrDNA + cpDNA
133 (105/28)
153 (106/47)
153 (106/47)
153 (106/47)
153 (106/47)
153 (106/47)
153 (106/47)
153 (106/47)
677 330 (48.7) 250 (36.9)
892 164 (18.4) 99 (11.1)
909 278 (30.6) 149 (16.4)
892 153 (17.2) 88 (9.9)
447 159 (35.6) 109 (24.4)
1569 494 (31.5) 349 (22.2)
2248 590 (26.2) 346 (15.4)
3817 1084 (28.4) 695 (18.2)
>10 000 873 0.541 0.912 GTR + I + G
>10 000 372 0.479 0.861 GTR + I + G
>10 000 416 0.676 0.948 GTR + G
50 273 0.525 0.923 GTR + I + G
>10 000 261 0.660 0.937 GTR + G
>10 000 1257 0.518 0.899 GTR + I + G
>10 000 989 0.593 0.928 GTR + I + G
>10 000 2344 0.520 0.901 GTR + I + G
The consistency index is calculated excluding uninformative characters.
3.2. Analysis of cpDNA sequence data analysis The aligned cpDNA data matrix included 2248 characters (matK, 909; rbcL, 892; trnL-F, 447). The cpDNA matK, rbcL, and trnL-F regions contained 278 (30.6%), 153 (17.2%), and 159 (35.6%) variable sites, and 149 (16.4%), 88 (9.9%), and 109 (24.4%) parsimony-informative sites, respectively (Table 1). MP analysis of the combined cpDNA data matrix resulted in >10 000 equally parsimonious trees (tree length = 989; CI = 0.593; RI = 0.928). A 50% majority-rule consensus tree was identical in topology to the BI phylogram (trees not shown). Duetzia was supported as a clade and sister to Kirengeshoma. The cpDNA tree showed two strongly-supported clades: D. mexicana and D. pringlei (sect. Neodeutzia [BP = 100, PP = 1.00]) and all other species (sections Mesodeutzia and Deutzia [BP = 100, PP = 1.00]). Neither of the two traditional sections Mesodeutzia and Deutzia is monophyletic.
3.3. Combined DNA analysis Among the 3817 characters of the combined data set, 695 (18.2%) were parsimony-informative (Table 1). Phylogenetic analysis of the combined data set resulted in >10 000 equally parsimonious trees, each of 2344 steps (CI = 0.520, RI = 0.901). The BI phylogram was identical in topology to the 50% majority-rule consensus tree sampled by the MP analysis (BI phylogram not shown). Strong support for the monophyly of Deutzia and the two main clades (I and II) common to both nrDNA and cpDNA trees remained in the combined analysis (Fig. 2). The combined tree shows the monophyly of D. mexicana and D. pringlei from Mexico but with weak support (BP = 76, PP = 1.00; clade I in Fig. 2). In clade II, the four subclades (a–d in Fig. 2) diverged from the remainder in a sequential fashion. Subclade d included five species (D. mollis, D. grandiflora, D. baroniana, D. hypoglauca, and D. albida) from Northeast Asia (BP = 84, PP = 0.93). Within this subclade, D. grandiflora was closely related to D. baroniana (BP = 96, PP = 1.00). Subclade c comprised three species of sect. Mesodeutzia (D. glabrata, D. amurensis, D. parviflora) and two species of sect. Deutzia (D. glauca, and D. schneideriana) from Northeast Asia (BP = 76, PP = 1.00). Subclade b included one species of Mesodeutzia (D. rubens) and 12 species of sect. Deutzia from Northeast Asia (BP = 95, PP = 1.00). Within this subclade, three groups were recognized. The Deutzia scabra complex (D. scabra and D. crenata) formed a group (BP = 83, PP = 1.00). Three Taiwan species – D. pulchra, D. ningpoensis, and D. cordatula – were sister to the D. scabra complex but with weak support. The East Asian species including D. maximowicziana, D. hypoleuca, D. taiwanensis, D. sieboldiana, and D. gracilis formed a group (BP = 86, PP = 1.00).
Subclade a comprised four species of Mesodeutzia and 29 species of sect. Deutzia from Northeast Asia and SW China, but with low support (BP < 50, PP = 0.97). The combined data did not resolve the monophyly of the two sections Deutzia and Mesodeutzia nor any of the infrasectional groups except for subsect. Grandiflorae of sect. Deutzia (Fig. 2). 3.4. Divergence time analysis Based on the combined nrDNA and cpDNA sequence data, the divergence time estimates for the main clades (I, II) and subclades (a–d) of Deutzia are shown in Fig. 3. The BEAST dating analysis results for the stem and crown node of Deutzia were estimated at 32 mya (95% HPD = 21.1–43.1 mya) in the early Oligocene and at 25.2 mya (95% HPD = 15.7–34.8 mya) in the late Oligocene, respectively. Within the genus, the age estimate for the crown node of clade II, including sections Mesodeutzia and Deutzia species from eastern Asia and SW China, was estimated at 19.8 mya (95% HDP = 11.6–28.1 mya) in the early Miocene while that of clade I comprising sect. Neodeutzia species from Mexico was dated to be 12.3 mya (95% HDP = 3.0–23.3 mya) in the middle Miocene. The ages of the nodes of the subclades were estimated to be the middle Miocene (subclade a, 13.1 mya, 95% HPD = 7.9–19.3 mya; subclade d, 12.5 mya, 95% HPD = 3.4–22.4 mya) or late Miocene (subclade b, 10.9 mya, 95% HPD = 6.5–15.9 mya; subclade c, 9.4 mya, 95% HPD = 3.0–17.1 mya). 3.5. Ancestral area reconstruction The BBM reconstruction suggests Northeast Asia (B) followed by Mexico (C), as the most probable ancestral area for Deutzia, whereas the S-DIVA reconstruction for this node is Northeast Asia + Mexico (BC) (Fig. 4). Both analyses indicated Northeast Asia as the ancestral area for clade II, including sections Mesodeutzia and Deutzia species from Northeast Asia and SW China. Similar results were obtained for the nodes of the subclades b, c, and d. BBM suggests Northeast Asia (B) as the most probable ancestral area at the node of the subclade a, which included the taxa of SW China and Northeast Asia, whereas S-DIVA indicates SW China (A) or SW China + Northeast Asia (AB) as the ancestral range at the node of this subclade (Fig. 4). 3.6. Morphological character evolution A schematic representation of the results from the ancestral state reconstructions of the three morphological characters (petal arrangement, filament shape, and stamen number) is shown in
C. Kim et al. / Molecular Phylogenetics and Evolution 87 (2015) 91–104
97
Fig. 3. Chronogram showing divergence times estimated in BEAST based on combined nrDNA and cpDNA sequence data. Divergence times of the clades and subclades are shown near each node. Yellow bars represent 95% high posterior density for the estimated mean dates. Nodes labeled C1–C4 are calibration points used in the analysis (for more details, see Section 2). Geological epoch is shown below the tree. The clades (I and II) and subclades (a–d) correspond to those in Fig. 2. An asterisk indicates the branches with maximum parsimony bootstrap value >75% and posterior probability >0.95.
98
C. Kim et al. / Molecular Phylogenetics and Evolution 87 (2015) 91–104
Fig. 4. Bayesian Binary Method (BBM) and S-DIVA models of ancestral area reconstruction in Deutzia based on a reduced BEAST combined-gene chronogram. BBM ancestral area reconstructions with the highest likelihood are shown as large pies for each clade and subclade. S-DIVA ancestral area reconstructions are shown by boxes at nodes; two boxes separated by a branch indicate the ancestral ranges inherited by each of the daughter lineages arising from the node. The clades (I and II) and subclades (a–d) correspond to those in Fig. 2. Biogeographic regions used in BBM and S-DIVA are shown on the map: A, SW China; B, other parts of eastern Asia; C, Mexico (Central America); D, North America. A color key for ancestral ranges at different nodes is provided in the figure. The numbers near nodes indicate corresponding support values (maximum parsimony bootstrap [BP]/Bayesian posterior probability [PP]); an en dash (–) indicates that a node did not receive P75% BP in the maximum parsimony analysis. (For interpretation of the references to colour in this figure legend, the reader is referred to the web version of this article.)
C. Kim et al. / Molecular Phylogenetics and Evolution 87 (2015) 91–104
Fig. 5. The induplicate type of petal arrangement (character A) was reconstructed as the ancestral form in Deutzia, with imbricate petals derived six times independently in clade II. The edentate filaments (character B) were plesiomorphic but they have changed to being 2-dentate in clade II, with two reversals. The ancestral state of the number of stamens for the genus is polystemonous. The 12–15 stamens were reconstructed unequivocally for the ancestor of clade I and ten stamens for the common ancestor of clade II, including the sections Mesodeutzia and Deutzia; there were no further transitions in each clade. 4. Discussion 4.1. Monophyly of Deutzia According to the current sampling, the monophyly of Deutzia is strongly supported by phylogenetic analyses of individual and combined data sets (Fig. 2). This result is in line with a previous study of the molecular phylogenetic relationships of Hydrangeaceae based on limited sampling (4 cm), inflorescences with more than five flowers, and shorter filament teeth than anther stalks. Subsection Grandiflorae is described as appearing similar to subsect. Deutzia based on its monomorphic filaments (2-dentate type), but differs in having shorter flowering branchlets to 4 cm and inflorescences with 1–3 flowers. Subsections Cymosae and Stenosepalae share the character of dimorphic filaments (only outer filaments are 2-dentate). They differ in that Cymosae species have globose or subglobose capsules and incurved calyx lobes. Subsection Stenosepalae, on the other hand, has hemispheric capsules and erect calyx lobes. Within subclade a, we could not find subsectional groupings of sect. Deutzia, i.e., all three subsections (Deutzia, Cymosae, Stenosepalae) are polymorphic (Fig. 2). For example, D. silvestrii and D. discolor of subsect. Deutzia are sisters to D. nanchuanensis of Stenosepalae (BP = 92, PP = 1.00). Deutzia esquirolii of Cymosae is sister to D. rehderiana of Stenosepalae (BP = 88, PP = 1.00). In contrast, some relationships seem to be defined by geographic distribution. Three species (D. triadiata, D. coreana, D. uniflora) from Korea (Northeast Asia) form a strongly supported group (BP = 100, PP = 1.00). Five species (D. fargesii, D. pilosa, D. setchuenensis, D. coriacea, D. multiradiata) from SW China form a clade, although only with weak support (BP = 0.78, PP = 0.94). The fact that subclade a is morphologically heterogeneous and has a disjunct range lends credence to the suggestion that it should be reevaluated with more extensive sampling of species from a wider geographical range. The strongly supported subclade b (BP = 95, PP = 1.00), including subsect. Deutzia species and D. rubens of sect. Mesodeutzia, had no clear unambiguously optimized synapomorphies. The
100
C. Kim et al. / Molecular Phylogenetics and Evolution 87 (2015) 91–104
Fig. 5. Reconstruction of the evolution of selected morphological characters in Deutzia based on a reduced BEAST combined-gene chronogram. The character state at the node of Deutzia indicates the ancestral state of the Deutzia clade. Transitions appear as filled boxes on the branches, yellow boxes indicate reversals; characters are shown above boxes and state transitions below. Descriptions of characters and character states are provided in the figure. The two clades (I, II) and four subclades (a–d) correspond to those in Fig. 2. An asterisk indicates the branches with maximum parsimony bootstrap value >75% and posterior probability >0.95. (For interpretation of the references to colour in this figure legend, the reader is referred to the web version of this article.)
widely distributed D. scabra complex, including D. scabra and D. crenata, is represented by several accessions in our study from
throughout its distributional range and form a clade (BP = 83, PP = 1.00) suggesting that it can be considered as a single variable
101
C. Kim et al. / Molecular Phylogenetics and Evolution 87 (2015) 91–104
species. In addition to its broad distribution, this species is morphologically variable in several characters including inflorescence type (compound racemes and panicles), the number of flowers per inflorescence (5–50; Huang et al., 2001; McGeogr, per. comm.), and the number of horizontal rays per stellate hair on the leaf epidermis (2–20; Styer and Stern, 1979a). Four accessions of D. scabra sequenced here are paraphyletic, suggesting that there may be cryptic species in this complex. Additional molecular studies should be pursued to determine the taxonomic status of D. scabra using more genetic markers. Subclade c comprises five species from Northeast Asia and two well-established segregations, but their relationship is not strongly supported (BP = 76, PP = 1.00). Three species (D. grabrata, D. amurensis, D. parviflora; BP = 94, PP = 1.00), corresponding to ser. Parviflorae of Mesodeutzia, are recognized by their edentate filaments and white petals (Hwang, 1993). However, the monophyly of ser. Parviflorae is not supported by our current sampling. Three species of ser. Parviflorae do not group with D. mollis which is the remaining species of this series and which, in fact, was included in subclade d (Fig. 2). According to previous morphological analysis of ser. Parviflorae species, three species (D. grabrata, D. amurensis, and D. parviflora) are clearly distinguished from D. mollis by petal color (white vs. pink), calyx length (0.6–0.8 mm vs. 2.0–2.3 mm), and pubescence of leaves (sparsely covered vs. densely covered; Kim and Chang, 2003). Deutzia glauca from Northeast Asia and D. schneideriana also from Northeast Asia, are strongly supported as sister taxa (BP = 97, PP = 1.00). Although they differ in that D. schneideriana has stellate hairs on its flowering branchlets and abaxial surface of its leaves (vs. without stellate hairs in D. glauca), these species are very similar in appearance and might be included in subsect. Deutzia of sect. Deutzia. Species of subclade d (BP = 84, PP = 0.93) are mainly restricted to Northeast Asia, with the exception of D. hypoglauca, which extends to SW China. However, it is difficult to identify morphological synapomorphies for this subclade. Within this subclade, a sister relationship between D. grandifolia and D. baroniana was recovered (BP = 96, PP = 1.00) and is also supported by their similar morphology, including short flowering branchlets (
