|Home | About | Journals | Submit | Contact Us | Français|
The Diervilla and Lonicera clades are members of the family Caprifoliaceae (Dipsacales sensu Donoghue et al., 2001, Harvard Papers in Botany 6: 459–479). So far, the intergeneric relationships of the Lonicera clade and the systematic position of Heptacodium remain equivocal. By studying fruit and seed morphology and anatomy, an attempt is made to clarify these issues. In addition, this study deals with the evolution of fruit and seed characters of the Diervilla and Lonicera clades with reference to allied taxa.
Light and scanning electron microscopy were used for the morphological and anatomical investigations. Phylogenetic analyses were carried out by applying the parsimony and Bayesian inference optimality criteria. Character evolution was studied by means of parsimony optimization and stochastic character mapping.
Diervilla and Weigela (Diervilla clade) are characterized by several unique traits in Dipsacales, including capsules with numerous seeds, seed coats without sclerified outer tangential exotestal cell walls, and dehiscent fruits. Seeds with completely sclerified exotestal cells and fleshy fruits characterize the Lonicera clade. Leycesteria and Lonicera have berries, ovaries without sterile carpels and several seeds per locule, whereas Symphoricarpos and Triosteum have drupes, ovaries with one or two sterile carpels and a single seed per locule. Heptacodium shares several characteristics with members of the Linnina clade, e.g. achenes, single-seeded fruits and a compressed, parenchymatous seed coat.
The results confirm the monophyly of the Diervilla and Lonicera clades and allow us to hypothesize a close relationship between Leycesteria and Lonicera and between Symphoricarpos and Triosteum. Fruit and seed morphology and anatomy point to a sister relationship of Heptacodium with the Linnina clade, rather than with the Lonicera clade.
Dipsacales fall within euasterids II and accommodate Adoxaceae and Caprifoliaceae (Donoghue et al., 1992, 2001, 2003; Pyck et al., 1999; Bell et al., 2001; Bremer et al., 2001; APG II, 2003; Zhang et al., 2003; Winkworth et al., 2008a, b; see Fig. 1). Recently, Winkworth et al. (2008a) postulated that Dipsacales are sister to Paracryphiaceae, and Dipsacales and Paracryphiaceae are together sister to Apiales.
Traditionally, Caprifoliaceae include the tribes Caprifolieae, Diervilleae and Linnaeeae, and the genera Sambucus and Viburnum (Hutchinson, 1967, 1973; Thorne, 1976; Takhtajan, 1980; Cronquist, 1981; Hara, 1983). Recent investigations based on molecular and morphological data (Donoghue et al., 1992, 2001, 2003; Gustafsson et al., 1996; Backlund and Pyck, 1998; Pyck et al., 1999; Bell et al., 2001; Bremer et al., 2001; Zhang et al., 2003; Winkworth et al., 2008b) have indicated that in this traditional circumscription Caprifoliaceae are polyphyletic. In 1998, Backlund and Pyck proposed a new classification in which two new families were erected, Diervillaceae and Linnaeaceae, corresponding to the former tribes Diervilleae and Linnaeeae, respectively. Caprifoliaceae corresponded to the former tribe Caprifolieae. In 2001, Donoghue et al. proposed a new classification for Caprifoliaceae accommodating the former tribes Caprifolieae, Diervilleae and Linnaeeae, and the former families Dipsacaceae, Morinaceae and Valerianaceae (Table 1). They proposed a rank-free classification in which traditional names of tribes and families were preserved to avoid confusion. In both the classifications of Backlund and Pyck (1998) and Donoghue et al. (2001), Sambucus and Viburnum were included in Adoxaceae with Adoxa, Sinadoxa and Tetradoxa. In this study, we adopt the upcoming classification of APG III (in preparation) in which two families are recognized, Adoxaceae and Caprifoliaceae (cf. Donoghue et al., 2001). We decided not to adopt the rank-free classification of Caprifoliaceae as proposed by Donoghue et al. (2001) as the use of informal names based on names of former tribes and families causes confusion. Instead, we have assigned informal names to the major clades of the family (Table 1, Fig. 1).
Several investigations (Backlund, 1996; Backlund and Pyck, 1998; Pyck et al., 1999; Donoghue et al., 2001, 2003; Bell et al., 2001; Zhang et al., 2003; Winkworth et al., 2008b) have dealt with the phylogenetic relationships of Caprifoliaceae (sensu Donoghue et al., 2001; Fig. 1). There is general consensus with regard to the sister relationship of Adoxaceae and Caprifoliaceae. Furthermore, the clade containing Diervilla and Weigela is sister to the remainder of Caprifoliaceae, and the Lonicera clade is sister to the Linnina clade sensu Donoghue et al. (2001; including the Dipsacus, Linnaea, Morina and Valeriana clades). The Lonicera clade comprises the genera Leycesteria, Lonicera, Symphoricarpos and Triosteum (Fig. 1). Although the position of Heptacodium is still ambiguous, the results of the analyses of Pyck and Smets (2000), Donoghue et al. (2003) and Winkworth et al. (2008b) suggest a position as sister to the Lonicera clade (Fig. 1); however, the possibility that Heptacodium is sister to the Linnina clade should not be discarded (Winkworth et al., 2008b).
The Diervilla clade contains deciduous shrubs and small trees with simple, opposite leaves (Hara, 1983). Flowers are organized in a cyme and have a cylindrical, bilocular ovary with numerous fertile ovules (Hara, 1983; Backlund, 1996; Donoghue et al., 2003). The clade comprises the genera Diervilla, including three species from eastern North America, and Weigela, including twelve species from East Asia (Hara, 1983; Backlund, 1996; Backlund and Pyck, 1998). Diervilla differs from Weigela by being stoloniferous and bearing capsules that split weakly or do not split at all (Hara, 1983). The seeds of Diervilla are lenticular and wingless, whereas those of Weigela are mostly cylindroid and morphologically characterized by a more-or-less prominent wing.
The intergeneric relationships of the Lonicera clade remain ambiguous, most likely due to the rapid diversification of its four genera (Winkworth et al., 2008b). Several studies have tackled this problem, starting with work by Backlund (1996) in which he analysed a data set of rbcL sequence data and 109 morphological characters. The results hypothesized a sister relationship between Lonicera plus Leycesteria and a clade of Heptacodium and Triosteum plus Symphoricarpos. In 2001, Donoghue et al. performed two analyses based on rbcL sequence data. The first analysis (rbcL only) placed Heptacodium as sister to a clade in which Leycesteria was sister to a polytomy of the remainder of the Lonicera clade. In the second analysis, in which the rbcL sequence data were supplemented with the morphological data of Backlund (1996), the authors suggested a clade consisting of Leycesteria and Lonicera being sister to a clade comprising Symphoricarpos and Triosteum. Once again, Heptacodium was sister to the Lonicera clade. In 2003, Donoghue et al. published the results of a maximum likelihood analysis based on nuclear (ITS) and plastid (matK, trnL) sequence data. The results showed strong support (95 %) for the sister relationship of Heptacodium and the Lonicera clade and 100 % support for the sister relationship of Leycesteria and the remaining three genera of the Lonicera clade. Recently, Winkworth et al. (2008b) carried out a series of analyses based on mitochondrial and plastid sequence data. Two datasets indicated a sister relationship of Heptacodium and the Lonicera clade with strong support. A third dataset supported a sister relationship of Heptacodium and the Linnina clade. The latter hypothesis was weakly supported by both parsimony and maximum likelihood, but strongly corroborated by Bayesian inference. The systematic equivocality surrounding the Lonicera clade was confirmed by incongruency between data sets, which largely disappeared when the Lonicera clade was excluded from the statistical tests. Theis et al. (2008) conducted a phylogenetic study focused on the phylogenetic relationships of the Lonicera clade and the genus Lonicera in particular. Based on ITS and five plastid markers, they concluded that Triosteum was sister to Symphoricarpos plus Leycesteria and Lonicera; however, sampling of the remainder of Dipsacales was rather limited.
Leycesteria, comprising five species distributed in the Himalayas and West China, resembles Lonicera, but has been considered more primitive (Horne, 1914) due to the development of a gynoecium with five (rarely four) fertile locules with numerous ovules in each locule. The gynoecium matures into a fleshy berry with numerous seeds. Chemotaxonomically, however, Leycesteria has been considered to be rather advanced in comparison with the other three genera of the clade (Bohm and Glennie, 1971). Lonicera comprises about 200 species of shrubs, trees and woody vines occurring in temperate and subtropical regions of Europe, North and Central America, North Africa and Asia. The genus is subdivided into two subgenera (Hara, 1983), Lonicera (180 species) and Caprifolium (22 species). The gynoecium is composed of two or three locules (rarely five), each locule containing three-to-eight ovules (Hara, 1983; Roels and Smets, 1996). The gynoecium matures into a fleshy berry. Triosteum comprises 6–7 species of perennial herbs with woody rhizomes found in East Asia and North America (Hara, 1983; Wilkinson, 1949). The zygomorphic flowers are sessile and located in the leaf axils or in a terminal spike (Hara, 1983). The gynoecium is composed of four carpels, of which one is sterile (Wilkinson, 1949). The fertile carpel holds a single pendent ovule and develops into a drupe containing three pyrenes (Wilkinson, 1949). Symphoricarpos contains about 15 species of shrubs mostly distributed in North America, but with one species, S. sinensis, in parts of China (Hara, 1983). The flowers are organized in a raceme or spike and have a tetralocular gynoecium, of which two carpels are infertile (Hara, 1983; Roels and Smets, 1996). The gynoecium develops into a drupe with two pyrenes (Hara, 1983; Roels and Smets, 1996).
Heptacodium holds two species, H. miconioides and H. jasminoides. Both are shrubs occurring in central China (Hara, 1983; Pyck and Smets, 2000). The flowers have a persistent calyx and a slightly curved, tubular corolla (Hara, 1983; Pyck and Smets, 2000; Zhang et al., 2002). The gynoecium is composed of three locules, of which two are abortive (Pyck and Smets, 2000; Zhang et al., 2002). Several ovules are present in the fertile locule, but only one matures into a long, spindle-shaped seed. The mature fruit is single-seeded and is often described as an achene (Pyck and Smets, 2000; Zhang et al., 2002; Donoghue et al., 2003). As mentioned earlier, the phylogenetic position of Heptacodium remains uncertain.
This current study documents the morphology and anatomy of the fruits and seeds of the Diervilla and Lonicera clades and Heptacodium. The impact of fruit and seed characters on the phylogenetic relationships of the Diervilla and Lonicera clades and the systematic position of Heptacodium is studied using sequence data (ITS, trnK and matK), and using a combined dataset composed of this sequence data and 17 fruit and seed characters plus 12 morphological characters from the study of Backlund (1996). The resulting topologies are used to study the evolution of fruit and seed characters. The decision to carry out original analyses instead of adopting a previously published phylogenetic analyses was motivated by our aim of matching more accurately the sampling of the morphological study. The primary focus of this paper is the study of the evolution of fruit and seed characters. Although our aim is not to present a better-supported or more-resolved phylogenetic analysis of Dipsacales, we try to contribute to a better understanding of the phylogenetic relationships.
Material for morphological and anatomical investigation was collected in the field and botanic gardens or acquired through collaboration with seed banks and herbaria (see Appendix 1). This study includes seven species of the Diervilla clade, 25 species of the Lonicera clade, one specimen of Heptacodium miconioides, 13 species of the Linnina clade and seven species of Adoxaceae. For a complete list of all specimens, please refer to Appendix 1. Leaf material for DNA extraction was collected in the field or in botanic gardens and preserved in silica gel.
DNA was extracted using a modified cetyltrimethylammonium bromide (CTAB) protocol (Tel-Zur et al., 1999). The tissue was ground and washed three times with extraction buffer (100 mm TrisHCL pH 8, 5 mm EDTA pH 8, 0·35 m sorbitol) to remove secondary metabolites. 700 µL CTAB lysis buffer (as described in Chase and Hills, 1991, with addition of 3% PVP-40) and 30 µL Sarkosyl were added to the samples, after which they were incubated for 1 h at 60°C. The aqueous phase was extracted twice with chloroform-isoamylalcohol (24/1, v/v) and subsequently subjected to an ethanol–salt precipitation (1/10 volume sodium acetate 3 m, 2/3 volume absolute ethanol). After centrifugation, the pellet was washed twice (70% ethanol), air-dried and dissolved in 100 µL TE buffer (10 mm TrisHCl pH 8, 1 mm EDTA pH 8).
Primers used for amplification and sequencing of ITS, trnK and matK are listed in Table 2, and statistics for the aligned sequence data are given in Table 3. Amplification of double-stranded copies of all three regions was done using standard PCR in 25-μL volume reactions. All reactions included an initial heating at 95 °C for 3 min. For ITS, the initial heating was followed by 30 cycles consisting of 95 °C for 60 s, 50 °C for 30 s and 72 °C for 30 s. For trnK and matK, the initial heating was followed by 30 cycles consisting of 95 °C for 60 s, 50–52 °C for 60 s and 72 °C for 60 s. All reactions were terminated with a final incubation of 72 °C for 3 min. To prevent the formation of secondary structures, we added 5% dimethylsulphoxide (DMSO) to the reaction mixture for ITS (Geuten et al., 2004). Samples were sequenced by the Macrogen sequencing facilities (Macrogen, Seoul, South Korea). Sequencing files were edited and assembled using Staden for Mac OS X (Staden et al., 1998).
Rehydration and fixation of the material was done by immersion in glutaraldehyde (2·5%) buffered with sodium cacodylate buffer (0·05 m, 24 h) and a subsequent wash in sodium cacodylate buffer (0·05 m, 24 h). The material was put through an ethanol series for dehydration purposes.
For light microscopy, the material was embedded in a hydroxyethylmethacrylate-based resin (Technovit 7100, Kulzer Histo-Technik, Wehreim, Germany), cut with a rotation microtome (HM360, Microm, Walldorf, Germany) and stained with toluidine blue. Longitudinal and cross-sections (5 µm in thickness) were observed and photographed with a Leitz Dialux 20 (Leitz, Wetzlar, Germany) equipped with a PL-B622CF PixeLINK digital camera and Microscopica v1·3 (Orbicule, Leuven, Belgium).
For scanning electron microscopy, the material was critical-point dried and sputtered with gold (Spi-Supplies, Walldorf, USA) prior to mounting on stubs. A JSM-6360 scanning electron microscope was used to observe and photograph the specimens. Pyrenes of Sambucus, Symphoricarpos, Triosteum and Viburnum were subjected to a hydrogen peroxide treatment (35%, 3–4 h, 60 °C) and subsequently cleaned with a toothbrush to remove the mesocarp.
A Leica MZ6 stereomicroscope (Leica Microsystems Ltd, Heerbrugg, Switzerland) was used to measure seeds and pyrenes. Seed-coat thickness and endocarp thickness were measured using a Leitz Dialux 20 equipped with a PL-B622CF PixeLINK digital camera and Microscopica v1·3 in combination with Macnification v0·2 (Orbicule, Leuven, Belgium).
Mean and s.d. values of seed and pyrene dimensions are based on five specimens per species; mean and s.d. values of seed coat and endocarp thickness are based on ten measurements on one specimen.
Table 4 provides a summary of the fruit and seed characters and their respective character states. Characters 18–29 were adopted from Backlund (1996) and we refer to this publication for a more extensive definition of these characters. The delimitation of four characters is arbitrary and deserves further explanation. We chose to define the second character (maximum number of carpels) as maximum number of carpels instead of number of carpels, because SIMMAP v1·0 beta 2·4 (Bollback, 2006; build 04082008–1·0-B2·4; Intel version) does not allow multiple character states as a single entry (e.g. ‘3–4’ is not a valid entry). The delimitation of character 4 (number of seeds) is based on the presence of clear gaps in the data. Although the bounds of character states 5 and 6 seem arbitrary, seed number for Lonicera fruits does not exceed 20 (state 5), whereas seed number for fruits of Leycesteria, Diervilla and Weigela varies from 50 to more than 100 (state 6). The delimitation of character 15 (embryo size) is based on observations and the presence of clear gaps. An unmistakable size and morphological difference is apparent between the large embryos of Sambucus, the Dipsacus clade and the Valeriana clade and the smaller embryos of the rest of the order. Additionally, we made a distinction between the embryos of Sambucus and the Dipsacus clade and the embryos of the Valeriana clade, which occupy the entire seed (no endosperm). Character 16 (seed-coat thickness) is continuous, and instead of using an algorithm we chose to delimit states based on the presence of gaps in the continuous data and the overall anatomy of the seed coat.
ITS, trnK and matK sequences of 32 species were aligned using MUSCLE v3·6 (Edgar, 2004), with default settings applied, after which small adjustments were made in MacClade v4·04 (Maddison and Maddison, 2002) to improve the alignment. The aligned matrices were submitted to TreeBASE (www.treebase.org). Five members of Adoxaceae were assigned to the outgroup (Appendix 2). Maximum parsimony (MP) and Bayesian inference (BI) were chosen to analyse three data sets: (1) molecular sequence data; (2) morphological data (Table 4, and see Table 8); and (3) molecular and morphological data combined. Parsimony analyses were carried out using PAUP* v4·0b10 (Swofford, 2002). Mr. Bayes v3·1·2 (Ronquist and Huelsenbeck, 2003) was used for BI analyses. Parsimony analyses were conducted on 1000 random addition replicates with tree-bisection-reconnection (TBR) branch swapping applied. Five trees were held at each step. Characters were unordered and equally weighted. Support for individual clades in the optimal tree was tested by a bootstrap analysis with 100 pseudoreplicates with settings identical to those of the original analysis except for the number of repetitions (100). Prior to the BI analyses, the molecular and morphological data were placed in separate partitions and a model was assigned to each partition. Mr. Modeltest v2·2 (Nylander, 2004) suggested a General Time Reversible model (GTR) with an invariable gamma-shaped distribution of rates across sites (GTR + I + G) for ITS and a GTR model with a variable gamma-shaped distribution of rates across sites (GTR + G) for trnK and matK. The standard discrete model was chosen for the analysis of the morphological data set (Ronquist and Huelsenbeck, 2003). The BI analyses were run for one million generations with partitions unlinked and sample frequency and burn-in set to 100 and 2500, respectively.
To search for incongruencies between matrices and topologies, we performed incongruence length difference (ILD; Farris et al., 1995) tests along with approximately unbiased (AU; Shimodaira, 2002) and Simodaira–Hasegawa (SH; Shimodaira and Hasegawa, 1999) tests. For the ILD tests, all data partitions were compared with each other (Table 5) using PAUP* v4·0b10 (Swofford, 2002). Additionally, we performed the same ILD tests with the exclusion of (1) the Lonicera clade, (2) the Linnaea clade and (3) Heptacodium in order to investigate the impact of these taxa on the incongruency of the data sets (Table 5), and we visually inspected the MP and BI topologies of the separate data sets to trace the cause of the incongruence. Furthermore, we used the AU and SH tests to compare three data sets (all sequence data, ITS sequence data, and plastid sequence data; Table 6) with eight alternative hypotheses: (1) the consensus topology of the BI analysis based on combined data; (2) the consensus topology of the BI analysis based on all molecular data; (3) all shortest MP trees based on combined data; (4) all shortest MP trees based on morphological data; and the maximum likelihood (ML) topology of (5) ITS, (6) trnK, (7) matK, and (8) all molecular data. PAUP* v4·0b10 (Swofford, 2002) was used to calculate the site-wise log-likelihoods, whereafter we used Consel v0·1j (Shimodaira and Hasegawa, 2001) to perform multiscale bootstrap resampling (ten sets of 10 000 replicates with scale parameters between 0·5 and 1·4). The ML topologies were calculated using GARLI v0·951 (Zwickl, 2006).
Parsimony optimization (PO) and stochastic character mapping (SCM) were used for the study of character evolution. The consensus tree of the combined BI analysis was chosen for the PO analysis in MacClade v4·04 (Maddison and Maddison, 2002) because of higher branch support and resolution. Additionally, a modified phylogenetic tree based on that of Winkworth et al. (2008b; see Fig. 1) was used for a second PO analysis in order to investigate the impact of an alternative placement of Heptacodium. The modified topology is identical to the consensus tree of our combined BI analysis except for the placement of Heptacodium, i.e. as sister to the Lonicera clade instead of sister to the Linnina clade. Stochastic character mapping was done through SIMMAP v1·0 beta 2·4 (Bollback, 2006; build 04082008–1·0-B2·4; Intel version) and based on the final 5000 trees of the combined BI analysis. The choice ofSCM was motivated by three factors: (1) multiple shifts along a single branch are possible; (2) potential under-estimation of variation inherent to the parsimony algorithm is avoided; and (3) BI deals with uncertainty with respect to phylogenetic reconstruction (Bollback, 2006). The results of the SCM analysis are available as a Supplementary Data, online.
In the Diervilla clade (Fig. 2A–D), Diervilla (Fig. 2A, B) and Weigela (Fig. 2C, D) differ in several aspects regarding seed shape and size. Generally, seeds of Diervilla are slightly smaller than those of Weigela (Table 7, Fig. 2A–D). Seeds of Diervilla are elliptic in lateral view (Fig. 2A) and ovate to elliptic in cross-section (Fig. 2B). Seeds of Weigela differ in shape from those of Diervilla as most seeds are angular and slightly more elongated (Fig. 2C, D). Some species of Weigela (e.g. W. hortensis) are characterized by having seeds with lateral outgrowths, often called wings (Fig. 2C). These outgrowths are absent in Diervilla (Fig. 2A, B). The hilum in both genera is terminal to subterminal (Fig. 2A, C).
In the Lonicera clade (Fig. 2E–L), seeds of Leycesteria (Fig. 2E, F) resemble those of Diervilla (Fig. 2A, B) in being bilaterally symmetrical and slightly dorsoventrally compressed. Generally, seed shape in lateral view varies from circular to elliptic to ovate and clavate (Fig. 2E). In cross-section, seed shape ranges from elliptic to ovate and clavate (Fig. 2F). Seed size differs significantly between Leycesteria formosa and L. crocothyrsos (Table 7). The seeds of Lonicera (Fig. 2G, H) are dorsiventrally compressed and irregular in shape (Fig. 2G, H). Seed shape in lateral view ranges from circular to elliptic to ovate (Fig. 2G), whereas seed shape in cross-section (Fig. 2H) is highly variable, even within species, although the seeds are dorsiventrally compressed in most cases. The hilum is terminal to subterminal. In Triosteum (Fig. 2I, J), seed shape in lateral view ranges from spindle-shaped (Fig. 2I) to elliptic or ovate, whereas seed shape in cross-section is predominantly elliptic (Fig. 2J). In Symphoricarpos (Fig. 2K, L), seed shape in cross-section is elliptic and uniform throughout the genus (Fig. 2L). Seed shape in lateral view is typically spindle-shaped and can be slightly curved (Fig. 2K).
The seed surface of Diervilla (Fig. 3A) and Weigela (Fig. 3B) is defined by the exotestal cells, which are characterized by a U-shaped sclerification pattern (Figs. 2B, D and and4A,4A, C). The outline of the cells is mostly pentagonal or hexagonal (Fig. 3A, B). Cell shape and size differ considerably within species, and no clear organizational pattern is discernable (Fig. 3A, B). The thin outer tangential cell wall is not sclerified, which often causes it to be concave and easily removable, uncovering the cell lumen (Fig. 3A, B). In Weigela, the seed-coat cells located at the lateral edges of the seed typically have uneven radial cell walls, which are taller than those of typical seed-coat cells (Fig. 2C). These cells give rise to the seed wing (Fig. 2C).
In the Lonicera clade (Fig. 3C–F), Leycesteria formosa has smooth seeds with a subtle sculpturing ascribed to the convex outer tangential walls of the exotestal cells (Fig. 3C). The exotestal cells of L. crocothyrsos have a thinner outer tangential cell wall than that in L. formosa. This causes the outer tangential cell walls to be concave when dehydrated. Characteristic for both genera is the subtle undulation of the cell outline of the exotestal cells (Fig. 3C). In Lonicera (Fig. 3D), the seed surface sculpture is defined by the anatomy of the outer tangential cell wall of the exotestal cells. The cell outline of the exotestal cells varies between hexagonal (Fig. 3D), circular and elliptic. In some species (e.g. L. canadensis), the cell outline is slightly undulate. Exotestal cell size varies strongly inter- and intraspecifically. The seed surface in Triosteum (Fig. 3E) is mostly characterized by exotestal cells with convex outer tangential cell walls (Fig. 4I) and a strongly undulate cell outline (Fig. 3E). Although anatomical cross-sections look highly similar when compared with other Triosteum spp., the undulate pattern is lacking in T. hirsutum. The exotestal cell outline of T. hirsutum is mostly square or rectangular, although slightly undulate exotestal cells were also observed. Exotestal cells of Symphoricarpos have moderately sclerified radial and inner tangential cell walls and a weakly sclerified outer tangential cell wall (Figs 3F and and4K,4K, L). The exotestal cell outline is modestly undulate (Fig. 3F) and varies between rectangular, square and slightly elongated. Observation of the seed surface of Symphoricarpos is difficult as exotesta and endocarp are closely associated (Fig. 4K, L) and endocarp removal generally results in a simultaneous removal of the exotesta. In Heptacodium (Fig. 3G), the seeds have a smooth surface with a subtle sculpturing due to the outlines of the compressed exotestal cells (Figs 3G and and44M–O).
In the Diervilla clade, seed-coat anatomy of both genera is highly comparable (Fig. 4A–D). The seed coat consists of three layers, i.e. an outer, well-developed exotesta, a mesotesta composed of compressed parenchyma cells, and an inner layer of endotesta cells (Fig. 4A, C). The exotesta functions as the mechanical layer and is composed of a single-cell layer (Fig. 4A–D). The radial and inner tangential cell walls are sclerified (Fig. 4A, C). The outermost part of the radial walls and the outer tangential walls are not sclerified, resulting in a U-shaped sclerification pattern (Fig. 4A, C). The adjacent mesotesta is reduced to a fine layer of compressed cells, which is not always visible (Fig. 4C). The inner endotesta is composed of a single layer of small, weakly sclerified cells (Fig. 4A, C). With the exception of W. japonica, none of the species studied in the Diervilla clade has seed-coat crystals.
In the Lonicera clade (Fig. 4E–L), the main difference between Leycesteria formosa and L. crocothyrsos is the degree of exotestal sclerification. The cell walls of L. formosa are heavily sclerified except for the moderately sclerified outer tangential wall (Fig. 4E, F). The exotestal seed-coat cells of L. crocothyrsos are moderately sclerified, whereas the outer tangential wall is not sclerified at all. The mesotestal cells in Leycesteria are reduced to a compressed layer of parenchyma cells (Fig. 4E, F). At the lateral edges, one or more cell layers of parenchymatous mesotestal cells are discernable, and the raphe is embedded in this mesotestal layer. The inner endotesta is slightly sclerified and consists of small dorsiventrally flattened cells (Fig. 4E, F). In L. crocothyrsos, one or two layers of endotestal cells are present, whereas in L. formosa only a single-cell layer is present at maturity (Fig. 4E). In Lonicera (Fig. 4G, H), the degree of exotestal cell sclerification varies greatly, from weakly sclerified cell walls in L. implexa to heavily sclerified cell walls in L. etrusca. In the majority of the species examined, the outer tangential wall is not sclerified (Fig. 4G, H). Although several species have cuboid exotestal cells (e.g. L. canadensis, Fig. 4G), exotestal cells are mostly more tall than wide. The mesotesta is reduced to a layer of compressed cells (Fig. 4G, H) except at the vascular bundles where multiple mesotestal cell layers can be present (cf. Leycesteria). In most species, the inner tangential cell walls are shaped by the mesotestal cells during seed maturation (Fig. 4G, H). The expanding endosperm subsequently compresses the mesotesta, resulting in a thin layer of cells. In most Lonicera spp., the mesotestal layer is characterized by the presence of druses. The endotesta is composed of small, dorsiventrally compressed cells, which are weakly sclerified (Fig. 4G). The seed-coat anatomy of Symphoricarpos and Triosteum is similar to that of Leycesteria and Lonicera, i.e. an outer layer of sclerified exotestal cells, a compressed layer of parenchymatous mesotestal cells, and an inner layer of small, sclerified, dorsiventrally compressed endotestal cells (Fig. 4I–L). In both Symphoricarpos and Triosteum, the mesotesta is characterized by druses. In some species (e.g. T. hirsutum), exotesta and endotesta can be multi-layered at the lateral edges of the seed. With the exception of Leycesteria, T. hirsutum and two species of Lonicera (L. canadensis and L. involucrata), all species studied in the Lonicera clade are characterized by seed-coat crystals (both druses and prismatic crystals were observed). The concave inner tangential exotestal cell wall of both Symphoricarpos and Triosteum is a clear indication that the mesotesta is composed of large, well-developed parenchymatous cells during seed development (Fig. 4I–K).
The seed coat of Heptacodium (Fig. 4M–P) is composed of compressed parenchymatous cells (Fig. 4M–O). During seed maturation, a single-layered exotesta and multilayered mesotesta is present (Fig. 4P). Although not observed, an endotesta is possibly present early in seed development. During seed development, the seed coat contains druses. The crystals, however, are absent at maturity.
In the Diervilla clade, endosperm cells of Diervilla (Fig. 5A) and Weigela (Fig. 5B) are similar in shape and size. The cells have thin cell walls and contain numerous starch grains. Although anatomically identical, cells located at the periphery of the endosperm are slightly smaller in size.
In the Lonicera clade, the endosperm of Leycesteria is composed of large, isodiametric cells with several, large starch grains (Fig. 5C). No significant differences were observed between the peripheral endosperm layers and the rest of the endosperm. Endosperm of Lonicera (Fig. 5D) is the most variable within the Lonicera clade. Cell shape and size differ among species, as do number and size of the starch grains. Lonicera is the only genus of the tribe in which some species have endosperm cells with slightly thickened cell walls. Endosperm cells of Triosteum (Fig. 5E) and Symphoricarpos (Fig. 5F) are similar to those of Leycesteria (Fig. 5C). Starch grains of the peripheral cells are markedly smaller than those of the rest of the endosperm. The number of starch grains per cell differs from species to species and does not seem to be genus-specific. Triosteum sinuatum, for example, has only a few starch grains per cell, whereas endosperm cells of T. hirsutum are filled with numerous, smaller starch grains.
The endosperm of Heptacodium (Fig. 5G) consists of large, isodiametric cells, which are slightly smaller at the periphery of the endosperm. Endosperm cells are filled with numerous starch grains. Endosperm anatomy is similar to that of Leycesteria (Fig. 5C), Triosteum (Fig. 5E) and Symphoricarpos (Fig. 5F).
In the Diervilla clade, seeds of both Diervilla and Weigela (Fig. 5H) have a straight, linear embryo. Embryo length varies between 1/2 to 1/3 of seed length and embryo width is about 1/4 of seed width.
In the Lonicera clade, Leycesteria is characterized by a straight, linear embryo. Embryo length is about 1/3 of seed length and embryo width is about 1/4 of seed width. The embryos of Lonicera (Fig. 5I) and Symphoricarpos are <1/3 of the length of the seed and shape and size is similar to that of Leycesteria. The embryo of Triosteum (Fig. 5J) is slender, straight and linear to ovate. The length of the embryo is <1/4 of seed length (or smaller), whereas embryo width is about 1/8 of seed width.
Embryo length in Heptacodium is about 1/10 of the length of the seed and embryo width is about 1/4 of seed width. The embryo is straight, linear and drop-shaped (Fig. 2O).
In the Diervilla clade, seeds of Diervilla (Figs 2B and and5K)5K) and Weigela (Fig. 2D) have a raphal bundle (no anti-raphal bundle) located in the mesotesta, containing an amphicribral vascular bundle. In the mature seed, the sclerified spiral tracheids are easily noticed and the thin-walled phloem cells are mostly compressed (Fig. 5K). At the location of the raphal bundle, often one or more mesotestal cell layers are present (Fig. 5K). Rexigenous cavities were observed, although not in all species.In the Lonicera clade (Figs 2F, H, J, L and and5L,5L, M), all members except for Leycesteria are marked by seeds with both a raphal and an anti-raphal vascular bundle located in the mesotesta. The vascular bundles are of the amphicribral type. The raphal bundle of Leycesteria (Fig. 2F) is composed of a few spiral tracheids surrounded by one or two layers of phloem cells. In Lonicera (Figs 2H and and5L),5L), the raphal bundle is generally larger than the anti-raphal bundle and both vascular bundles are often characterized by a rexigenous cavity (Fig. 5L). The vascular bundles in Symphoricarpos (Fig. 2L) are rather small in comparison with Triosteum (Figs 2J and and5M)5M) and size and shape differs interspecifically. Rexigenous cavities occur in Symphoricarpos, although not in all species. In some species of Triosteum (e.g. T. sinuatum), the anti-raphal bundle splits in two strands. Large rexigenous cavities are present in most species of Triosteum (Fig. 5M).
Seeds of Heptacodium lack an anti-raphe. In young stages, the raphal bundle appears to be amphicribral, although no clear phloem cells could be observed (Fig. 5N). The metaxylem surrounds a rexigenous cavity and is itself surrounded by multiple layers of collenchyma (Fig. 5N). At maturity, only the metaxylem of the vascular bundle is visible. The other tissues are compressed as a result of seed growth.
Triosteum (Fig. 6A–C, G, H) and Symphoricarpos (Fig. 6D–F, I, J) both produce seeds surrounded by a sclerified endocarp, i.e. pyrenes. Pyrene shape differs between the genera as most species of Triosteum (except for T. hirsutum) have pyrenes with five or six prominent ribs (Fig. 6A–C, G, H), whereas pyrenes of Symphoricarpos (Fig. 6D–F, 6I, J) lack such ribs. Another distinct difference is endocarp anatomy. The endocarp of Symphoricarpos is composed of three distinct layers (Fig. 6I, J): an inner layer of fibres orientated perpendicular to the pyrene's length axis; a layer of one (rarely more) cell layer of crystal-containing sclereids; and an outer layer of fibres orientated parallel to the length axis of the pyrene. The endocarp of Triosteum is composed of two layers (Fig. 6G, H) similar to the fibrous innermost and outermost layers of Symphoricarpos. The middle layer is lacking in Triosteum, although in T. angustifolium and T. perfoliatum a small number of crystal-containing sclereids are scattered throughout the endocarp. In T. angustifolium the latter sclereids contain druses, whereas the sclereids of T. perfoliatum hold prismatic crystals. Due to the presence of ribs in Triosteum, the organization of the two layers of fibres (Fig. 6G, H) differs from that of Symphoricarpos (Fig. 6I, J). Although the fibres are organized in large strands, a pattern, comparable to that of Symphoricarpos, is not discernable. The fibres of the outer layer seem to run dorsiventrally (Fig. 6C, G, H). The endocarp of Triosteum is much harder and more robust than the endocarp of Symphoricarpos, which is most likely due to the strands of fibres and the different organization of the fibres in Triosteum (Fig. 6C, G, H). The endocarp of Symphoricarpos tends to be slightly flexible.
The endocarp of Heptacodium is composed of a single layer of moderately sclerified fibres (Fig. 4P). Since the fruit of Heptacodium does not dehisce, the entire pericarp functions as mechanical layer (Fig. 2O).
The MP analysis resulted in two most-parsimonious trees of 1904 steps (CI = 0·70; RI = 0·77). The aligned matrix consisted of 2403 characters, of which 613 were potentially parsimony informative. The sister relationship of the Diervilla clade and the remainder of Caprifoliaceae gained strong supported (bootstrap, BS = 100; posterior probability, PP = 1·00). Both the MP and BI analysis validate the monophyly of the Diervilla (97 BS; 1·00 PP) and Lonicera clades (99 BS; 1·00 PP). Intergeneric relationships of the Lonicera clade, however, are different. Based on BI, the intergeneric relationships of the four genera remain unclear due to a polytomy that unites all four genera. The MP topology hypothesizes Triosteum to be sister to the remainder of the Lonicera clade (99 BS).The other intergeneric relationships in the Lonicera clade gain low support. The two shortest trees of the MP analysis suggest a sister relationship (no bootstrap support) of Lonicera and Symphoricarpos plus Leycesteria (no bootstrap support). The monophyly of all genera of the Lonicera clade, however, gained strong support. In both analyses, the Lonicera clade is sister to a clade consisting of Heptacodium plus the Linnina clade (94 BS; 1·00 PP). The sister relationship of Heptacodium and Linnina gained moderate to strong support (74 BS; 1·00 PP). In Linnina, the relationships between the Linnaea, Morina, Dipsacus and Valeriana clades are identical in both analyses, i.e. the Linnaea clade is sister (98 BS; 1·00 PP) to Valerina and the Morina clade is sister (62 BS; 0·66 PP) to the Dipsacus clade plus the Valeriana clade (98 BS; 0·99 PP).
The addition of 29 morphological characters to the molecular data matrix generally resulted in an increase in resolution and support in both analyses. The MP analysis resulted in four shortest trees of 2067 steps (CI = 0·71; RI = 0·77). The monophyly of Diervilla, Weigela and the Diervilla and Lonicera clades remained strongly supported (100 BS; 1·00 PP). The intergeneric relationships in the Lonicera clade differed from the resultant topologies based on molecular data. The BI analysis based on combined data suggested that Leycesteria plus Lonicera (0·89 PP) and Triosteum plus Symphoricarpos (0·87 PP) are sister groups (1·00 PP). Three out of four of the shortest MP trees are congruent with the BI hypothesis. The other shortest tree is congruent with the shortest MP trees based on molecular data (see above). The sister relationship of the Lonicera clade and the clade consisting of Heptacodium and Linnina gained lower support than in the analyses based on sequencing data only (83 BS; 1·00 PP). The sister relationship of Heptacodium and Linnina, however, gained support (89 BS; 1·00 PP). Support and resolution of the intergeneric relationships of the Linnaea clade increased slightly. Based on BI, Abelia and Linnaea borealis are hypothesized to be sisters (0·75 PP) with Dipelta being sister to the pair (0·99 PP). Kolkwitzia amabilis is sister to the remainder of the the Linnaea clade (1·00 PP). The four shortest MP trees are congruent with the BI topology. Bootstrap support, however, is low except for the sister relationship of K. amabilis and the remainder of the Linnaea clade (100 BS).
ILD testing (Table 5) shows significant incongruencies are present between ITS and trnK, trnK and matK, and between the individual molecular data sets and the morphological data set. The tests indicate that the exclusion of Heptacodium or the Linnaea clade has little to no impact on the incongruence of the data sets. The exclusion of the Lonicera clade, however, resulted in the disappearance of all significant incongruence between the data sets except for the incongruence between trnK and matK. The AU and SH tests (Table 6) largely confirm the ILD test results (Table 5). As indicated by the ILD tests, all morphological topologies (MP and BI) differ significantly from the ML topologies based on the three combinations of sequence data. The ML topologies of the individual molecular data sets (ITS, trnK and matK) differ significantly from the ML topology of the combined molecular data set. Comparison of topologies and ILD, AU and SH tests indicate significant incongruencies are present between the data sets. Visual inspection of the topologies, however, clearly indicates the incongruencies are primarily related to uncertainty regarding the intergeneric relationships of the Lonicera and Linnaea clades and the placement of Heptacodium. The incongruence of the molecular and morphological data is primarily due to the relatively weak phylogenetic signal of the latter data set.
The composition and monophyly of the Diervilla clade, Diervilla and Weigela are strongly supported in our analyses (Figs 7 and and8).8). Fruit and seed morphology and anatomy of both genera are quite different from what is encountered in the remainder of the order. Diervilla and Weigela have bicarpellate, dehiscent capsules with numerous seeds, whereas the remainder of the order is characterized by drupes, berries or achenes. Like the genera of the Lonicera clade, Diervilla and Weigela have seeds with a sclerified exo- and endotesta. The degree of sclerification of the outer tangential exotestal cell wall, however, differs from that of the Lonicera clade. The mesotesta of both genera is reduced to a layer of compressed cells, in which a vascular bundle or raphe is located. Abortive carpels are absent in the Diervilla clade (Backlund and Pyck, 1998), although a fruit-developmental study is required to address this question confidently. The fruits and seeds of Diervilla and Weigela are much alike in overall anatomy. Several morphological differences, however, are apparent. Seeds of Weigela are in close contact with each other, which causes the shape of the seeds to be partially determined by adjacent seeds, resulting in flattened lateral sides and seed wings (Fig. 2C, D). Seeds of Diervilla are less angular. It is still unclear whether this morphological difference is due to abortive ovules or an initial smaller number of ovules. Capsules of Weigela are long, slender and cylindrical, whereas those of Diervilla are shorter and broader at the base, resulting in bottle-shaped fruits.
Whereas the monophyly of all four genera is well supported, intergeneric relationships are less clear (Figs 7 and and8).8). Our results support a close affinity between Leycesteria and Lonicera and between Symphoricarpos and Triosteum. Although the latter hypothesis gained only moderate to poor support in our combined analyses, fruit and seed morphology and anatomy provide strong evidence. Leycesteria and Lonicera lack sterile carpels and a sclerified endocarp, whereas Symphoricarpos and Triosteum have ovaries with one (Triosteum) or two (Symphoricarpos) sterile carpels and fruits with a sclerified endocarp. In Leycesteria and Lonicera, each carpel contains several fertile ovules, whereas carpels of Symphoricarpos and Triosteum contain a single pendent ovule. Consequently, fruits of Leycesteria and Lonicera contain numerous seeds, whereas those of Symphoricarpos and Triosteum contain two and three seeds, respectively. Furthermore, Leycesteria and Lonicera have berries, whereas Symphoricarpos and Triosteum have drupes.
Several studies have dealt with the intergeneric relationships of the Lonicera clade (e.g. Backlund, 1996; Pyck et al., 1999; Donoghue et al., 2001; Zhang et al., 2003). Recently, Winkworth et al. (2008b) tackled the question by analysing mitochondrial and plastid sequence data using several strategies, i.e. separately analysing coding and non-coding data and a total evidence approach, but found the results to be conflicting. The underlying cause for the lack of a stable phylogenetic hypothesis for the Lonicera and Linnaea clades might be the rapid diversification of the taxa. This hypothesis is supported by a dating study of Dipsacales by Bell and Donoghue (2005). The authors found that the genera of the Lonicera clade diverged within a time frame of 3–4 million years, whereas other major clades of the order diverged within a period of 15–60 million years. The branch lengths of the BI topologies confirm this hypothesis (Fig. 8).
Within Dipsacales, the fruits and seeds of the Lonicera clade are unique in several ways. Members of the tribe have a sclerified exotesta and a weakly sclerified endotesta as in the Diervilla clade, but the outer tangential exotestal cell wall is weakly to moderately sclerified in the Lonicera clade, whereas it is not sclerified in the Diervilla clade. The Lonicera clade is the only clade in the family characterized by the occurrence of fruits with a fleshy mesocarp, i.e. berries and drupes. Seeds of Diervilla and Weigela show some resemblance to those of Leycesteria and Lonicera. All four genera have multiple fertile ovules per carpel, produce fruits with numerous seeds and have seeds with a similar seed-coat anatomy (see above).
Zhang et al. (2002) hypothesized that Heptacodium might have evolved as a hybrid from ancestors of the Lonicera and Linnaea clades. This hypothesis provided an explanation for the morphological similarities that Heptacodium shares with both tribes. Winkworth et al. (2008b), however, commented that it is less likely that differences between functional partitions in uniparentally inherited plastid or mitochondrial genomes could be explained by hybridization. Nevertheless, hybridization could explain why our analyses suggest a sister relationship of Heptacodium and the Linnina clade. Heptacodium shares specific floral characteristics with the Lonicera clade and ovary-related characteristics with the Linnaea clade (Donoghue et al., 2003; Winkworth et al., 2008b). Our results indicate a sister relationship between Heptacodium and the Linnina clade.
The broad fruit diversity in Caprifoliaceae is linked to several functional shifts of the mechanical layer, which is the tissue that protects the embryo and endosperm from the environment. Diervilla, Leycesteria, Lonicera and Weigela have seeds with a sclerified exotesta acting as a mechanical layer. In Symphoricarpos and Triosteum, a first transfer of function occurred with the sclerification of the endocarp and the resulting development of pyrenes. In spite of the presence of a sclerified endocarp acting as a mechanical layer, the exo- and endotesta are also sclerified in Symphoricarpos and Triosteum. We believe this might be a strong argument for hypothesizing that the ancestor of the Lonicera clade had fruits with a sclerified exo- and endotesta and an unsclerified endocarp. Sclerification of the exo- and endotesta should be regarded as a plesiomorphic condition in the Lonicera clade and possibly in the order as a whole (see below). In the clade including Heptacodium and Linnina, three independent evolutionary shifts are apparent. A first shift toward the development of a protective pericarp and the simultaneous development of a parenchymatous seed coat occurred in the Linnaea and Valeriana clades and Heptacodium. In the Morina and Dipsacus clades, independent second and third evolutionary shifts resulted in the functional transfer of the mechanical layer to the epicalyx. In the Dipsacus clade, for example, the pericarp is reduced to a thin, papery layer surrounding the single seed, whereas epicalyx morphology has diversified into a broad range of shapes and sizes. Finally, in Adoxaceae, the sclerified endocarp functions as a mechanical layer except in Sinadoxa. The fruits of Sinadoxa have been described as achene-like (Wu et al., 1981), which means the entire pericarp protects the seed. Thus, in Dipsacales a general trend is apparent in which the function of a mechanical layer is transferred to outer tissues in more derived clades.
Due to the morphological differences that separate Adoxaceae and Caprifoliaceae, the uncertain systematic position of several key taxa and the equivocal intergeneric relationships within the Linnaea and Lonicera clades, the morphology and anatomy of the ancestral Dipsacales is difficult to infer (Donoghue et al., 2003). We can be rather confident, however, that the ancestor of Dipsacales was woody and had simple, opposite leaves without stipules (Donoghue et al., 2003). Floral morphology of the first Dipsacales is more difficult to infer due to the contrasting flower morphology of Adoxaceae and Caprifoliaceae. The first Adoxaceae most likely had actinomorphic flowers with small calyx lobes, rotate corollas, short styles and lobed stigmas, but lacked distinct nectaries, whereas the first Caprifoliaceae had zygomorphic flowers with larger calyx lobes, tubular corollas, elongate styles, capitate stigmas and nectaries composed of unicellular hairs at the base of the corolla (Donoghue et al., 2003). We can also be fairly confident that the ancestor of Dipsacales had perfect, fertile flowers with five corolla lobes and five stamens (Donoghue et al., 2003). A biogeographical study indicated that the first Dipsacales most probably originated in East Asia, where they inhabited the understory of temperate forests (Bell and Donoghue, 2005).
In the following paragraphs, we discuss the evolution of several key fruit and seed characters (Tables 4 and and8)8) based on parsimony optimization and stochastic character mapping (Fig. 9). If appropriate, we indicate the posterior probability (PP) scores of the SCM analysis. The results of the SCM analysis are available as Supplementary Data, online.
The plesiomorphic fruit type for Dipsacales (PP 0·99) is a drupe. Based on this hypothesis, several shifts have occurred: (1) a shift to capsules along the branch leading to the Diervilla clade; (2) a shift to berries at the origin of the clade including Leycesteria and Lonicera; (3) a first shift to achenes toward the lineage comprising Heptacodium and Linnina; and (4) a second shift to achenes after the split of Sinadoxa and the clade containing the drupe-bearing genera Adoxa and Tetradoxa.
The PO and SCM analyses generate conflicting results with respect to carpel number of the frist Dipsacales. Ovaries with three carpels are hypothesized to be the plesiomorphic condition for Dipsacales based on PO. Based on this hypothesis, three reductions took place: (1) a reduction to two carpels along the branch leading to Dipelta; (2) a first reduction to a single carpel at the origin of the Dipsacus clade; and (3) a second reduction to a single carpel after the split of Sinadoxa plus the pentacarpellate genera Adoxa and Tetradoxa. Besides these reductions, three increases occurred: (1) an increase to four carpels along the branch leading to Symphoricarpos and Triosteum; (2) a first increase to five carpels at the origin of Leycesteria; and (3) a second increase to five carpels at the base of subfamily Adoxoideae. The number of carpels in Lonicera varies intraspecifically, ranging from two to three per ovary. However, SCM hypothesizes that the first Dipsacales had pentacarpellate ovaries as observed in Adoxa, Leycesteria and Sambucus (and possibly Tetradoxa). Six reductions took place: (1) a reduction to two carpels along the branch leading to the Diervilla clade; (2) a reduction at the origin of the clade comprising Symphoricarpos and Triosteum; (3) a reduction along the branch leading to Lonicera; (4) a reduction to three carpels at the origin of the Linnina clade; (5) a reduction along the branches leading to Dipelta (bicarpellate); and (6) a reduction in the Dipsacus clade (monocarpellate). In Adoxaceae, Viburnum is characterized by monocarpellate ovaries (Wilkinson, 1948) as is Sinadoxa (Wu et al., 1981).
Sterile carpels are absent in the Diervilla and Dipsacus clades, Adoxoideae (with the exception of Sinadoxa), Leycesteria and Lonicera. The plesiomorphic condition for both Dipsacales and the Lonicera clade is equivocal based on PO. Triosteum has ovaries with a single sterile carpel, whereas Symphoricarpos, Viburnum and the clade including Heptacodium and Linnina are characterized by ovaries with two sterile carpels. Based on the modified phylogenetic tree (based on that of Winkworth et al., 2008b; see Fig. 1), PO indicates that the ancestor giving rise to Caprifoliaceae minus the Diervilla clade had two sterile carpels. We believe that the first Dipsacales had flowers lacking sterile carpels and that the presence of sterile carpels is a derived feature in the order. This hypothesis is strongly confirmed by our SCM analysis (PP 0·99). The only Adoxaceae with sterile carpels are Sinadoxa and Viburnum, and the shift toward the occurrence of sterile carpels most likely took place along the branches leading to these taxa. Moreover, the Diervilla clade lacks sterile carpels, as do Leycesteria and Lonicera, which are often considered as the most ‘primitive’ genera of the Lonicera clade (e.g. Wilkinson, 1949). A pentacarpellate dipsacalean ancestor without sterile carpels as hypothesized by our SCM analysis seems likely.
SCM assigns a PP score of 0·99 to the hypothesis of fruits with five seeds. Our PO analysis, however, hypothesizes single-seeded fruits to have characterized the first Dipsacales. However, we believe the latter result is an artefact as Heptacodium, Viburnum and the entire Linnina clade (excluding Dipelta) produce single-seeded fruits. Ovary development in Viburnum is characteristic for the genus (e.g. Wilkinson, 1948, 1949; Jacobs et al., 2008; see above) and must therefore have occurred along the evolutionary path leading to Viburnum. An evolutionary link with the single-seeded fruits of Heptacodium and Linnina is therefore most unlikely as ovary development differs dramatically from ovary development in Viburnum (Wilkinson, 1948). The evolution of seed number within the Lonicera clade is equivocal. We believe, however, that the ancestor of Caprifoliaceae (and possibly the dipsacalean ancestor) had fruits with numerous seeds, as fruits with fewer (or a single) seed are predominantly found in derived clades (e.g. Linnina and Viburnum). Additional support for this hypothesis is provided by the many ovules present in all three carpels of Heptacodium, of which only a single ovule in a single carpel matures into a fertile seed. A first shift to carpels with a single fertile ovule occurred at the origin of the clade comprising Heptacodium and Linnina, and a second similar shift took place along the branch leading to Symphoricarpos and Triosteum. These shifts led to fruits with fewer seeds, a single seed in Heptacodium and Linnina, two seeds in Symphoricarpos and three seeds in Triosteum.
Based on SCM, the plesiomorphic condition for Dipsacales is the presence of three endocarp layers (PP 0·99), whereas PO hypothesizes two endocarp layers as the plesiomorphic condition for the order. With SCM in mind, a reduction to a single (unsclerified) layer occurred along the branch leading to Leycesteria and Lonicera. A similar shift took place at the origin of the clade containing Heptacodium and Linnina. Viburnum and the Diervilla clade are marked by fruits with two endocarp layers. When Heptacodium is considered sister to the the Lonicera clade, the shift to a single layer took place after the Diervilla clade separated from the rest of Caprifoliaceae.
Based on PO, it is unclear whether the ancestor of Dipsacales had fleshy fruits or not. However, SCM provides strong support for the ancestral condition for the order being a fleshy fruit (PP 1·00). Adoxaceae (except for Sinadoxa) and the Lonicera clade have fleshy fruits. Dry fruits are found in the Diervilla clade, Linnina and Heptacodium. When the shift toward dry fruits took place is unclear, and it is likely that multiple shifts took place.
Although both PO and SCM are unclear about the plesiomorphic condition for Dipsacales, an evolutionary pattern within the order is obvious. The ancestor of Caprifoliaceae is hypothesized to have had seeds with a sclerified exotesta and endotesta (PP 0·99), whereas the branch leading to Heptacodium and Linnina is marked by a shift to a compressed, parenchymatous seed coat (PP 0·99). In Adoxaceae, Adoxa and Viburnum have uncompressed, parenchymatous seed coats, whereas seeds of Sambucus have a compressed, parenchymatous seed coat (Jacobs et al., 2008). If we assume Heptacodium is sister to the Lonicera clade, the plesiomorphic condition for Caprifoliaceae and the clade holding the Lonicera clade and Heptacodium is equivocal.
In Dipsacales, a trend toward a larger embryo is apparent. Although not significantly supported by SCM (PP 0·84), based on PO the plesiomorphic condition for the order is an embryo less than 3/4 of seed length. Three shifts toward a larger embryo took place: (1) a first shift to an embryo larger than 3/4, but not occupying the entire seed (endosperm present) occurred at the origin of Sambucus; (2) a similar shift happened at the origin of the Dipsacus clade; and (3) a third shift toward an embryo occupying the entire seed (no endosperm present) characterizes the Valeriana clade. Although the character state of Sambucus and the Dipsacus clade is coded identically, embryo morphology is quite different. Sambucus has a long, slender, cylindrical embryo, whereas members of the Dipsacus clade have a comparatively larger embryo, which is slightly flattened dorsiventrally.
The Diervilla clade is formed of two genera, as confirmed by several features including the presence of capsules, a sclerified exo- and endotesta, and dehiscent fruits. Fruit and seed morphology and anatomy support a sister relationship between Leycesteria and Lonicera and between Symphoricarpos and Triosteum. The monophyly of the Lonicera clade is supported by several features including a sclerified exo- and endotesta and fleshy fruits. Our results also support the hypothesis of Heptacodium being sister to the Linnina clade, rather than to the Lonicera clade.
We believe the first Dipsacales had pentacarpellate, fleshy fruits with numerous (>20) seeds, characterized by sclerified seed coats (whether both exo- and endotesta were sclerified is unclear) and small embryos (less than 3/4 of seed length). A shift to bicarpellate capsules with numerous seeds took place along the branch leading to the Diervilla clade, which coincided with a shift to dry, dehiscent fruits. This hypothesis implies that the fruit of Leycesteria has all the characteristic features of a primitive dipsacalean fruit (Wilkinson, 1949). The main difference with Lonicera is the occurrence of sterile ovules in the latter. Within the Lonicera clade, a shift to one or two sterile carpels occurred along the branch leading to Symphoricarpos and Triosteum. This coincided with the sclerification of the endocarp, which means this sclerification occurred independently of the sclerification at the origin of Adoxaceae. Finally, a second shift to dry fruits together with a shift to three carpels (of which two are sterile) took place at the origin of the clade comprising Heptacodium and Linnina. This evolution occurred simultaneously with the maturing of a single ovule in the only fertile carpel.
A future study will deal with the evolution of fruits and seeds of the order Dipsacales in more detail. Such a study will need to include members of the sister group of Dipsacales in order to address more accurately questions related to the plesiomorphic character states of the order. Furthermore, an expanded sampling of the Linnina clade is required for the construction of a more highly resolved evolutionary map.
Supplememtary data are available online at www.aob.oxfordjournals.org and consists of a figure with the constraints used in the analysis, and the raw results of the stochastic character-mapping analysis.
Many thanks go the seed banks of the Botanic Garden of Jena, the Institute of Ecology and Botany of the Hungarian Academy of Sciences, the St. Andrews Botanic Garden, the Botanic Garden of Krefeld, the Botanical Garden of Ljubljana, the National Botanic Garden of Belgium, the Cruickshank Botanic Garden, the Botanical Garden of Nantes, the Paris National Museum for Natural History, the Botanic Garden of the University of Hamburg, the Botanic Garden of Smith College, the Nationaal Herbarium Nederland (Utrecht University Branch), the Sir Harold Hillier Gardens and Arboretum, the Botanical Garden of Copenhagen and the Botanical Garden Uppsala University for sending us dried fruits, seeds or leaf material. Additionally, we would like to thank the National Botanic Garden of Belgium and the Arboretum of Kalmthout for providing us with opportunity to collect fresh material, and the Nationaal Herbarium Nederland (Leiden University Branch) for allowing us to collect several key taxa in their herbarium. This work was supported by research grants of the Research Council of the Katholieke Universiteit Leuven (OT/05/35). Frederic Lens is a postdoctoral fellow of the Fund for Scientific Research – Flanders (Belgium) (F.W.O. – Vlaanderen).
Classification and collection details of sampled species
|Taxon||Classification||Herbarium, seed bank or locality1||Collector||Accession||Type2|
|Abelia chinensis R. Br.||Linnaea clade||National Botanic Garden of Belgium, Belgium||Jacobs B.||19881652||F|
|Abelia parvifolia Hemsl.||Linnaea clade||National Botanic Garden of Belgium, Belgium||Jacobs B.||19850252||F|
|Adoxa moschatellina L.||Adoxaceae||Heverlee, Belgium||Jacobs B.||290||F|
|Centranthus ruber (L.) DC.||Valeriana clade||Botanical Garden of Jena, Germany||NA||5471||S|
|Diervilla rivularis Gatt.||Diervilla clade||Botanical Garden of Jena, Germany||NA||2079||S|
|Diervilla sessilifolia Buckley||Diervilla clade||L||Boom B.K.||6117||H|
|Dipelta floribunda Maxim.||Linnaea clade||Shangai Botanical Garden, Shangai, China||NA||79||S|
|Dipelta yunnanensis Franch.||Linnaea clade||National Botanic Garden of Belgium, Belgium||Jacobs B.||19921864-04||F|
|Dipsacus fullonum L.||Dipsacus clade||Utrecht University Botanic Garden, Utrecht, The Netherlands||NA||1991ZE00259||S|
|Heptacodium miconioides Rehder||–||Arboretum Kalmthout, Belgium||Jacobs B.||19990134||F|
|Kolkwitzia amabilis Graebn.||Linnaea clade||Institute of Ecology and Botany of the Hungarian Academy of Sciences, Hungary||NA||562||S|
|Leycesteria crocothyrsos Airy/Shaw||Lonicera clade||L||–||C1123A||H|
|Leycesteria formosa Wall.||Lonicera clade||L||Boom B.K.||956062836||H|
|Linnaea borealis L.||Linnaea clade||Linnaeus Garden, Uppsala University, Uppsala, Sweden||Hansson L.||HL20080001||F|
|Lonicera alpigena L.||Lonicera clade||Botanical Garden of Jena, Germany||NA||2081||S|
|Lonicera canadensis Bartr. ex Marshall||Lonicera clade||L||Senn H.A. & Zinck M.N.||420||H|
|Lonicera caprifolium L.||Lonicera clade||L||Sotiaux P.||811162||H|
|Lonicera chrysantha Turcz.||Lonicera clade||Institute of Ecology and Botany of the Hungarian Academy of Sciences, Hungary||NA||505||S|
|Lonicera dioica L.||Lonicera clade||L||Moldenke H.N.||17763||H|
|Lonicera etrusca Santi||Lonicera clade||L||–||9991||H|
|Lonicera implexa Aiton||Lonicera clade||L||De Langhe J.E.||973727||H|
|Lonicera involucrata Banks ex Spreng.||Lonicera clade||L||Grant J.M.||502710||H|
|Lonicera japonica Thunb.||Lonicera clade||L||Koyama H.||2906||H|
|Lonicera javanica DC.||Lonicera clade||L||Afriastini J.J.||–||H|
|Lonicera maackii (Rupr.) Herder||Lonicera clade||Botanical Garden of Jena, Germany||NA||2083||S|
|Lonicera maximowiczii Maxim.||Lonicera clade||St. Andrews Botanic Garden, Great Britain||NA||80||S|
|Lonicera muscaviensis Rehder||Lonicera clade||L||Boom B.K.||902710||H|
|Lonicera vesicaria Kom.||Lonicera clade||Institute of Ecology and Botany of the Hungarian Academy of Sciences, Hungary||NA||571||S|
|Lonicera xylosteum L.||Lonicera clade||Botanischer Garten Krefeld, Germany||NA||80||S|
|Morina longifolia Wall.||Morina clade||Cruickshank Botanic Garden, Great Britain||NA||52||S|
|Morina persica L.||Morina clade||St. Andrews Botanic Garden, Great Britain||NA||265||S|
|Sambucus ebulus L.||Adoxaceae||Institute of Ecology and Botany of the Hungarian Academy of Sciences, Hungary||NA||2001-228||S|
|Sambucus racemosa L.||Adoxaceae||Botanische Garten der Universität Hamburg, Germany||NA||297||S|
|Scabiosa columbaria L.||Dipsacus clade||Institute of Ecology and Botany of the Hungarian Academy of Sciences, Hungary||NA||862||S|
|Succisa pratensis Moench.||Dipsacus clade||Botanical Garden of Nantes, France||NA||187||S|
|Symphoricarpos albus (L.) S.F. Blake var. laevigatus (Fernald) G.N. Jones||Lonicera clade||Botanical Garden of Ljubljana, Slovenia||NA||191||S|
|Symphoricarpos mollis Nutt.||Lonicera clade||St. Andrews Botanic Garden, Great Britain||NA||190/1964||S|
|Symphoricarpos occidentalis Hook.||Lonicera clade||St. Andrews Botanic Garden, Great Britain||NA||191/1964||S|
|Symphoricarpos oreophilus Gray||Lonicera clade||St. Andrews Botanic Garden, Great Britain||NA||192/1964||S|
|Triosteum angustifolium L.||Lonicera clade||L||Wilhelm N. Suksdorf||899||H|
|Triosteum hirsutum Roxb.||Lonicera clade||L||J. D. H.||202710||H|
|Triosteum perfoliatum L.||Lonicera clade||L||–||845875||H|
|Triosteum sinuatum Maxim.||Lonicera clade||L||Shimizu & N. Fukuoka||499||H|
|Valeriana officinalis L.||Valeriana clade||National Botanic Garden of Belgium, Belgium||NA||19721065||S|
|Valerianella locusta (L.) Betcke||Valeriana clade||National Botanic Garden of Belgium, Belgium||NA||19922077-23||S|
|Viburnum acerifolium L.||Adoxaceae||St. Andrews Botanic Garden, Great Britain||NA||84||S|
|Viburnum plicatum var. tomentosum (Thunb.) Miquel||Adoxaceae||L||Fukuoka N.||972060502||H|
|Weigela floribunda C. A. Mey.||Diervilla clade||Botanical Garden of Jena, Germany||NA||9262||S|
|Weigela florida DC.||Diervilla clade||National Botanic Garden of Belgium, Belgium||Jacobs B.||19392515||F|
|Weigela hortensis C. A. Mey.||Diervilla clade||Botanical Garden of Ljubljana, Slovenia||NA||192||S|
|Weigela japonica Thunb.||Diervilla clade||Botanical Garden of Jena, Germany||NA||6240||S|
|Weigela subsessilis L. H. Bailey||Diervilla clade||National Botanic Garden of Belgium, Belgium||Jacobs B.||19931547-84||F|
1 For herbarium material the particular herbarium (acronym) is indicated; for seed bank material the particular institute is indicated; for fresh material the particular locality or botanic garden is indicated.
2 H, herbarium; S, seed bank; F, fresh material.
na, not applicable; –, missing data.
Voucher and accession details of sampled species
|Taxon||Collection and voucher information||ITS||matK||trnK|
|Abelia chinensis||Sir Harold Hillier Gardens and Arboretum, N. Pyck 1989-2220||FJ745388||AY310461||–|
|Abelia parvifolia||Botanical Garden of Copenhagen, N. Pyck 1943-5025||FJ745387||FJ745398||–|
|Centranthus ruber||Institute of Botany and Microbiology, N. Pyck 001||FJ745391||AF446926||AY794313|
|Diervilla sessilifolia||National Botanical Garden of Belgium, N. Pyck 82-6494||AY236177||AF446907||FJ745402|
|Dipelta floribunda||Sir Harold Hillier Gardens and Arboretum, N. Pyck 1978-4099||FJ745389||FJ745399||–|
|Dipsacus fullonum||National Botanical Garden of Belgium, N. Pyck 80-1959||FJ745390||FJ745400||–|
|Heptacodium miconioides||National Botanical Garden of Belgium, N. Pyck 92-0130-16||AY236176||AF446906||FJ745412|
|Kolkwitzia amabilis||National Botanical Garden of Belgium, DDM/88/0215FB/R67||AY236182||AF446912||FJ745413|
|Leycesteria crocothyrsos||Sir Harold Hillier Gardens and Arboretum, N. Pyck 1992-1691||AF265277||FJ745393||FJ745406|
|Leycesteria formosa||National Botanical Garden of Belgium, N. Pyck 82-6395||AF265276||AF446902||FJ745405|
|Lonicera dioica||National Botanical Garden of Belgium, N. Pyck 51-3590||EU240713||FJ745395||–|
|Lonicera involucrata||National Botanical Garden of Belgium, N. Pyck 53-6481||FJ745386||FJ745397||FJ745408|
|Lonicera maackii||National Botanical Garden of Belgium, N. Pyck 88-1731||FJ217883||FJ745394||FJ745407|
|Lonicera maximowiczii||National Botanical Garden of Belgium, N. Pyck 81-1860||FJ745385||FJ745396||–|
|Succisa pratensis||National Botanical Garden of Belgium, N. Pyck 1975-2365||AY290018||FJ745401||AY290033|
|Symphoricarpos albus||Kasteelpark Arenberg, P. Roels 004||AF265282||AY310459||FJ745410|
|Symphoricarpos occidentalis||National Botanical Garden of Belgium, N. Pyck 90-1416||FJ217824||–||FJ745411|
|Triosteum perfoliatum||Botanical Garden Uppsala University, N. Pyck 1963-1028||AY236175||AF446905||FJ745409|
|Weigela florida||National Botanical Garden of Belgium, N. Pyck 51-0632||AF078711||FJ745392||FJ745404|
|Weigela subsessilis||National Botanical Garden of Belgium, N. Pyck 93-1547-84||AF078706||–||FJ745403|
Boldface accessions refer to sequences obtained for this study.
na, not applicable; –, missing data.